meillo@58: .H0 "Discussion
meillo@0: .P
meillo@169: This main chapter discusses the practical work accomplished in the
meillo@169: mmh project.
meillo@217: It is structured along the goals chosen for the project.
meillo@217: A selection of the work undertaken
meillo@217: is described.
meillo@217: .P
meillo@217: This discussion compares the present version of mmh with the state of
meillo@217: nmh at the time when the mmh project had started, i.e. fall 2011.
meillo@224: Recent changes in nmh are rarely part of the discussion.
meillo@187: .P
meillo@187: For the reader's convenience, the structure of modern email systems
meillo@217: is depicted in the following figure.
meillo@187: It illustrates the path a message takes from sender to recipient.
meillo@217: 
meillo@217: .sp 1.5
meillo@187: .KS
meillo@187: .in 2c
meillo@187: .so input/mail-agents.pic
meillo@187: .KE
meillo@217: .sp 1.5
meillo@217: 
meillo@187: .LP
meillo@217: The ellipses denote mail agents, i.e. different jobs in email processing.
meillo@217: These are:
meillo@187: .IP "Mail User Agent (MUA)
meillo@217: The only program users directly interact with.
meillo@187: It includes functions to compose new mail, display received mail,
meillo@187: and to manage the mail storage.
meillo@217: It is called a \fImail client\fP as well.
meillo@187: .IP "Mail Submission Agent (MSA)
meillo@187: A special kind of Mail Transfer Agent, used to submit mail into the
meillo@187: mail transport system.
meillo@217: Often it is also called an MTA.
meillo@187: .IP "Mail Transfer Agent (MTA)
meillo@187: A node in the mail transport system.
meillo@217: It transfers incoming mail to a transport node nearer to the
meillo@217: final destination.
meillo@217: An MTA may be the final destination itself.
meillo@187: .IP "Mail Delivery Agent (MDA)
meillo@217: Delivers mail according to a set of rules.
meillo@217: Usually, the messages are stored to disk.
meillo@187: .IP "Mail Retrieval Agent (MRA)
meillo@217: Initiates the transfer of mail from a remote location to the local machine.
meillo@217: (The dashed arrow in the figure represents the pull request.)
meillo@217: .LP
meillo@217: The dashed boxes represent entities that usually reside on single machines.
meillo@217: The box on the lower left represents the sender's system.
meillo@187: The box on the upper left represents the first mail transfer node.
meillo@224: The box on the upper right represents the transfer node responsible
meillo@224: for the destination address.
meillo@217: The box on the lower right represents the recipient's system.
meillo@187: Often, the boxes above the dotted line are servers on the Internet.
meillo@217: Many mail clients, including nmh, include all of the components below
meillo@217: the dotted line.
meillo@217: This is not the case for mmh; it implements the MUA only.
meillo@187: 
meillo@187: 
meillo@187: 
meillo@58: 
meillo@58: 
meillo@58: 
meillo@133: .\" --------------------------------------------------------------
meillo@125: .H1 "Streamlining
meillo@58: 
meillo@0: .P
meillo@217: MH once provided a complete email system.
meillo@217: The community around nmh tries to keep nmh in similar shape.
meillo@178: In fundamental contrast, mmh shall be an MUA only.
meillo@87: I believe that the development of all-in-one mail systems is obsolete.
meillo@173: Today, email is too complex to be fully covered by a single project.
meillo@173: Such a project will not be able to excel in all aspects.
meillo@159: Instead, the aspects of email should be covered by multiple projects,
meillo@87: which then can be combined to form a complete system.
meillo@169: Excellent implementations for the various aspects of email already exist.
meillo@223: Just to name three examples: Postfix is a specialized MTA, Procmail
meillo@223: is a specialized MDA, and Fetchmail is a specialized MRA.
meillo@89: I believe that it is best to use such specialized tools instead of
meillo@217: providing the same function once more as a side component.
meillo@58: .P
meillo@169: Doing something well requires focusing on a small set of specific aspects.
meillo@217: Under the assumption that development which is focussed on a particular
meillo@217: area produces better results there, specialized projects will be superior
meillo@87: in their field of focus.
meillo@87: Hence, all-in-one mail system projects \(en no matter if monolithic
meillo@87: or modular \(en will never be the best choice in any of the fields.
meillo@217: Even in providing the most consistent all-in-one system, they are likely
meillo@217: to be beaten by projects that focus exclusively on the creation
meillo@217: of a homogeneous system by integrating existing mail components.
meillo@87: .P
meillo@217: Usually, the limiting resource in the community development of
meillo@217: free software is man power.
meillo@217: If the development effort is spread over a large development area,
meillo@217: it becomes more difficult to compete with the specialists in the
meillo@87: various fields.
meillo@87: The concrete situation for MH-based mail systems is even tougher,
meillo@169: given their small and aged community, concerning both developers and users.
meillo@87: .P
meillo@87: In consequence, I believe that the available development resources
meillo@100: should focus on the point where MH is most unique.
meillo@87: This is clearly the user interface \(en the MUA.
meillo@125: Peripheral parts should be removed to streamline mmh for the MUA task.
meillo@60: 
meillo@60: 
meillo@100: .H2 "Mail Transfer Facilities
meillo@154: .Id mail-transfer-facilities
meillo@60: .P
meillo@217: The removal of the mail transfer facilities, effectively dropping the
meillo@217: MSA and MRA, had been the first work task in the mmh project.
meillo@217: The desire for this change initiated the creation of the mmh project.
meillo@60: .P
meillo@169: Focusing on one mail agent role only, is motivated by Eric Allman's
meillo@105: experience with Sendmail.
meillo@217: He identified the limitation of Sendmail
meillo@217: to the MTA task as one reason for its success:
meillo@105: .[ [
meillo@105: costales sendmail
meillo@105: .], p. xviii]
meillo@105: .QS
meillo@105: Second, I limited myself to the routing function \(en
meillo@110: I wouldn't write user agents or delivery back-ends.
meillo@217: This was a departure of the dominant thought of the time,
meillo@105: in which routing logic, local delivery, and often the network code
meillo@105: were incorporated directly into the user agents.
meillo@105: .QE
meillo@105: .P
meillo@187: In nmh, the MSA is called \fIMessage Transfer Service\fP (MTS).
meillo@105: This facility, implemented by the
meillo@105: .Pn post
meillo@217: command, establishes network connections and spoke SMTP to submit
meillo@159: messages to be relayed to the outside world.
meillo@217: When email transfer changed, this part needed to be changed as well.
meillo@89: Encryption and authentication for network connections
meillo@87: needed to be supported, hence TLS and SASL were introduced into nmh.
meillo@217: This added complexity without improving the core functions.
meillo@217: Furthermore, keeping up with recent developments in the field of
meillo@87: mail transfer requires development power and specialists.
meillo@217: In mmh, this whole facility was simply cut off
meillo@76: .Ci f6aa95b724fd8c791164abe7ee5468bf5c34f226
meillo@76: .Ci fecd5d34f65597a4dfa16aeabea7d74b191532c3
meillo@217: .Ci 156d35f6425bea4c1ed3c4c79783dc613379c65b .
meillo@87: Instead, mmh depends on an external MSA.
meillo@217: All outgoing mail in mmh goes through the
meillo@60: .Pn sendmail
meillo@87: command, which almost any MSA provides.
meillo@87: If not, a wrapper program can be written.
meillo@87: It must read the message from the standard input, extract the
meillo@87: recipient addresses from the message header, and hand the message
meillo@87: over to the MSA.
meillo@223: For example, a wrapper script for qmail would be:
meillo@87: .VS
meillo@87: #!/bin/sh
meillo@138: exec qmail-inject  # ignore command line arguments
meillo@87: VE
meillo@87: The requirement to parse the recipient addresses out of the message header 
meillo@217: may be removed in the future.
meillo@217: Mmh could pass the recipient addresses as command line arguments.
meillo@100: This appears to be the better interface.
meillo@60: .P
meillo@60: To retrieve mail, the
meillo@60: .Pn inc
meillo@217: command in nmh acts as MRA.
meillo@217: It establishes network connections
meillo@217: and speaks POP3 to retrieve mail from remote servers.
meillo@76: As with mail submission, the network connections required encryption and
meillo@217: authentication, thus TLS and SASL were added to nmh.
meillo@169: Support for message retrieval through IMAP will soon become necessary
meillo@169: additions, too, and likewise for any other changes in mail transfer.
meillo@217: But not in mmh because it has dropped the support for retrieving mail
meillo@217: from remote locations
meillo@217: .Ci ab7b48411962d26439f92f35ed084d3d6275459c .
meillo@76: Instead, it depends on an external tool to cover this task.
meillo@169: Mmh has two paths for messages to enter mmh's mail storage:
meillo@100: (1) Mail can be incorporated with
meillo@60: .Pn inc
meillo@87: from the system maildrop, or (2) with
meillo@60: .Pn rcvstore
meillo@87: by reading them, one at a time, from the standard input.
meillo@60: .P
meillo@217: With the removal of the MSA and MRA, mmh converted from a complete
meillo@217: mail system to only an MUA.
meillo@60: Now, of course, mmh depends on third-party software.
meillo@87: An external MSA is required to transfer mail to the outside world;
meillo@60: an external MRA is required to retrieve mail from remote machines.
meillo@217: Excellent implementations of such software exist.
meillo@217: They likely are superior to the internal versions that were removed.
meillo@217: Additionally, the best suiting programs can be chosen freely.
meillo@60: .P
meillo@217: As it had already been possible to use an external MSA and MRA,
meillo@217: why should the internal version not be kept for convenience?
meillo@224: Transfered to a different area, the question,
meillo@224: whether there is sense in having a fall-back pager in all
meillo@224: the command line tools for the cases when
meillo@60: .Pn more
meillo@60: or
meillo@60: .Pn less
meillo@173: are not available, appears to be ridiculous.
meillo@100: Of course, MSAs and MRAs are more complex than text pagers
meillo@87: and not necessarily available but still the concept of orthogonal
meillo@217: design holds: ``Write programs that do one thing and do it well''.
meillo@87: .[
meillo@87: mcilroy unix phil
meillo@87: p. 53
meillo@87: .]
meillo@87: .[
meillo@87: mcilroy bstj foreword
meillo@87: .]
meillo@87: Here, this part of the Unix philosophy was applied not only
meillo@87: to the programs but to the project itself.
meillo@87: In other words:
meillo@164: Develop projects that focus on one thing and do it well.
meillo@169: Projects which have grown complex should be split, for the same
meillo@169: reasons that programs which have grown complex should be split.
meillo@100: If it is conceptionally more elegant to have the MSA and MRA as
meillo@87: separate projects then they should be separated.
meillo@217: In my opinion, this is the case.
meillo@217: The RFCs suggest this separation by clearly distinguishing the
meillo@217: different mail handling tasks [RFC\|821].
meillo@217: The small interfaces between the mail agents support the
meillo@217: separation as well.
meillo@76: .P
meillo@217: Once, email had been small and simple.
meillo@100: At that time,
meillo@60: .Pn /bin/mail
meillo@169: had covered everything there was to email and still was small and simple.
meillo@100: Later, the essential complexity of email increased.
meillo@224: (Essential complexity is the complexity defined by the problem itself
meillo@217: .[ [
meillo@87: brooks no silver bullet
meillo@224: .]].)
meillo@217: Consequently, email systems grew.
meillo@100: RFCs started to introduce the concept of mail agents to separate the
meillo@217: various roles because they became more extensive and because
meillo@217: new roles appeared.
meillo@217: As mail system implementations grew, parts of them were split off.
meillo@169: For instance, a POP server was included in the original MH;
meillo@169: it was removed in nmh.
meillo@217: Now is the time to go one step further and split off the MSA and MRA,
meillo@217: as well.
meillo@87: Not only does this decrease the code size of the project,
meillo@169: more importantly, it unburdens mmh of the whole field of
meillo@217: message transfer, with all its implications for the project.
meillo@169: There is no more need for concern with changes in network transfer.
meillo@217: This independence is gained by depending on external components
meillo@217: that cover the field.
meillo@60: .P
meillo@217: In general, functionality can be added in three different ways:
meillo@171: .LI 1
meillo@217: By implementing the function in the project itself.
meillo@171: .LI 2
meillo@217: By depending on a library that provides the function.
meillo@171: .LI 3
meillo@217: By depending on a program that provides the function.
meillo@171: .LP
meillo@169: While implementing the function in the project itself leads to the
meillo@169: largest increase in code size and requires the most maintenance
meillo@169: and development work,
meillo@217: it keeps the project's dependence on other software lowest.
meillo@169: Using libraries or external programs requires less maintenance work
meillo@217: but introduces dependencies on external projects.
meillo@169: Programs have the smallest interfaces and provide the best separation,
meillo@87: but possibly limit the information exchange.
meillo@169: External libraries are more strongly connected than external programs,
meillo@169: thus information can be exchanged in a more flexible manner.
meillo@87: Adding code to a project increases maintenance work.
meillo@217: As implementing complex functions in the project itself adds
meillo@217: a lot of code, this should be avoided if possible.
meillo@217: Thus, the dependencies only change in their character,
meillo@169: not in their existence.
meillo@66: In mmh, library dependencies on
meillo@66: .Pn libsasl2
meillo@66: and
meillo@66: .Pn libcrypto /\c
meillo@66: .Pn libssl
meillo@159: were traded against program dependencies on an MSA and an MRA.
meillo@159: This also meant trading build-time dependencies against run-time
meillo@87: dependencies.
meillo@169: Besides providing stronger separation and greater flexibility,
meillo@169: program dependencies also allowed
meillo@66: over 6\|000 lines of code to be removed from mmh.
meillo@66: This made mmh's code base about 12\|% smaller.
meillo@87: Reducing the project's code size by such an amount without actually
meillo@87: losing functionality is a convincing argument.
meillo@87: Actually, as external MSAs and MRAs are likely superior to the
meillo@87: project's internal versions, the common user even gains functionality.
meillo@66: .P
meillo@169: Users of MH should not have problems setting up an external MSA and MRA.
meillo@60: Also, the popular MSAs and MRAs have large communities and a lot
meillo@169: of available documentation.
meillo@217: 
meillo@223: Choices for MSAs range from small forwarders such as \fIssmtp\fP and
meillo@223: \fInullmailer\fP, over mid-size MTAs including \fImasqmail\fP and
meillo@223: \fIdma\fP, up to full-featured MTAs as for instance \fIPostfix\fP.
meillo@223: MRAs are provided for example by \fIfetchmail\fP, \fIgetmail\fP,
meillo@223: \fImpop\fP, and \fIfdm\fP.
meillo@60: 
meillo@60: 
meillo@100: .H2 "Non-MUA Tools
meillo@60: .P
meillo@217: One goal of mmh is to remove the tools that do not significantly
meillo@217: contribute to the MUA's job.
meillo@217: Loosely related and rarely used tools distract from a lean appearance,
meillo@217: and require maintenance work without adding much to the core task.
meillo@217: By removing these tools, mmh became more streamlined and focused.
meillo@62: .BU
meillo@58: .Pn conflict
meillo@87: was removed
meillo@76: .Ci 8b235097cbd11d728c07b966cf131aa7133ce5a9
meillo@217: because it is a mail system maintenance tool and not MUA-related.
meillo@87: It even checked
meillo@58: .Fn /etc/passwd
meillo@58: and
meillo@58: .Fn /etc/group
meillo@87: for consistency, which is completely unrelated to email.
meillo@87: A tool like
meillo@87: .Pn conflict
meillo@87: is surely useful, but it should not be shipped with mmh.
meillo@62: .BU
meillo@58: .Pn rcvtty
meillo@87: was removed
meillo@87: .Ci 14767c94b3827be7c867196467ed7aea5f6f49b0
meillo@89: because its use case of writing to the user's terminal
meillo@200: on reception of mail is obsolete.
meillo@87: If users like to be informed of new mail, the shell's
meillo@58: .Ev MAILPATH
meillo@87: variable or graphical notifications are technically more appealing.
meillo@217: Writing to terminals directly is hardly ever desired today.
meillo@169: If, though, one prefers this approach, the standard tool
meillo@58: .Pn write
meillo@58: can be used in a way similar to:
meillo@82: .VS
meillo@58: scan -file - | write `id -un`
meillo@82: VE
meillo@62: .BU
meillo@58: .Pn viamail
meillo@87: was removed
meillo@87: .Ci eda72d6a7a7c20ff123043fb7f19c509ea01f932
meillo@87: when the new attachment system was activated, because
meillo@58: .Pn forw
meillo@76: could then cover the task itself.
meillo@62: The program
meillo@58: .Pn sendfiles
meillo@62: was rewritten as a shell script wrapper around
meillo@58: .Pn forw .
meillo@76: .Ci 0e82199cf3c991a173e0ac8aa776efdb3ded61e6
meillo@62: .BU
meillo@58: .Pn msgchk
meillo@87: was removed
meillo@87: .Ci bb9360ead7eb7a3fedcce2eeedfc660014e41dbe ,
meillo@87: because it lost its use case when POP support was removed.
meillo@76: A call to
meillo@58: .Pn msgchk
meillo@87: provided hardly more information than:
meillo@82: .VS
meillo@58: ls -l /var/mail/meillo
meillo@82: VE
meillo@217: Yet, it distinguished between old and new mail, but
meillo@169: these details can be retrieved with
meillo@76: .Pn stat (1),
meillo@62: too.
meillo@100: A small shell script could be written to print the information
meillo@76: in a similar way, if truly necessary.
meillo@76: As mmh's
meillo@76: .Pn inc
meillo@87: only incorporates mail from the user's local maildrop,
meillo@62: and thus no data transfers over slow networks are involved,
meillo@169: there is hardly any need to check for new mail before incorporating it.
meillo@62: .BU
meillo@58: .Pn msh
meillo@87: was removed
meillo@76: .Ci 916690191222433a6923a4be54b0d8f6ac01bd02
meillo@87: because the tool was in conflict with the philosophy of MH.
meillo@217: It provided an interactive shell to access the features of MH.
meillo@217: However, it was not just a shell tailored to the needs of mail handling,
meillo@217: but one large program that had several MH tools built in.
meillo@217: This conflicted with the major feature of MH of being a tool chest.
meillo@76: .Pn msh 's
meillo@159: main use case had been accessing Bulletin Boards, which have ceased to
meillo@62: be popular.
meillo@62: .P
meillo@62: Removing
meillo@169: .Pn msh
meillo@212: together with the truly archaic code relics
meillo@58: .Pn vmh
meillo@58: and
meillo@169: .Pn wmh
meillo@62: saved more than 7\|000 lines of C code \(en
meillo@66: about 15\|% of the project's original source code amount.
meillo@100: Having less code \(en with equal readability, of course \(en
meillo@76: for the same functionality is an advantage.
meillo@63: Less code means less bugs and less maintenance work.
meillo@76: As
meillo@63: .Pn rcvtty
meillo@63: and
meillo@63: .Pn msgchk
meillo@87: are assumed to be rarely used and can be implemented in different ways,
meillo@87: why should one keep them?
meillo@217: Removing them streamlined mmh.
meillo@63: .Pn viamail 's
meillo@63: use case is now partly obsolete and partly covered by
meillo@63: .Pn forw ,
meillo@217: hence there is no reason to still maintain it.
meillo@63: .Pn conflict
meillo@76: is not related to the mail client, and
meillo@63: .Pn msh
meillo@63: conflicts with the basic concept of MH.
meillo@169: These two tools might still be useful, but they should not be part of mmh.
meillo@63: .P
meillo@220: .Id slocal
meillo@169: Finally, there is
meillo@217: .Pn slocal ,
meillo@217: which is an MDA and thus not directly MUA-related.
meillo@217: It should be removed from mmh because including it conflicts with
meillo@178: the idea that mmh is an MUA only.
meillo@87: However,
meillo@76: .Pn slocal
meillo@76: provides rule-based processing of messages, like filing them into
meillo@76: different folders, which is otherwise not available in mmh.
meillo@87: Although
meillo@76: .Pn slocal
meillo@169: neither pulls in dependencies, nor does it include a separate
meillo@154: technical area (cf. Sec.
meillo@154: .Cf mail-transfer-facilities ),
meillo@169: it still accounts for about 1\|000 lines of code that need to be maintained.
meillo@76: As
meillo@76: .Pn slocal
meillo@76: is almost self-standing, it should be split off into a separate project.
meillo@76: This would cut the strong connection between the MUA mmh and the MDA
meillo@76: .Pn slocal .
meillo@87: For anyone not using MH,
meillo@87: .Pn slocal
meillo@87: would become yet another independent MDA, like
meillo@87: .I procmail .
meillo@100: Then
meillo@87: .Pn slocal
meillo@217: could be installed without a complete MH system.
meillo@76: Likewise, mmh users could decide to use
meillo@76: .I procmail
meillo@217: without having a second, unused MDA, i.e.
meillo@87: .Pn slocal ,
meillo@76: installed.
meillo@100: That appears to be conceptionally the best solution.
meillo@76: Yet,
meillo@76: .Pn slocal
meillo@87: is not split off.
meillo@100: I defer the decision over
meillo@78: .Pn slocal
meillo@169: out of a need for deeper investigation.
meillo@217: In the meanwhile, it remains part of mmh
meillo@217: as its continued existence is not significant;
meillo@100: .Pn slocal
meillo@100: is unrelated to the rest of the project.
meillo@0: 
meillo@58: 
meillo@133: 
meillo@134: .H2 "Displaying Messages
meillo@155: .Id mhshow
meillo@131: .P
meillo@133: Since the very beginning, already in the first concept paper,
meillo@58: .Pn show
meillo@62: had been MH's message display program.
meillo@58: .Pn show
meillo@76: mapped message numbers and sequences to files and invoked
meillo@58: .Pn mhl
meillo@89: to have the files formatted.
meillo@173: With MIME, this approach was not sufficient anymore.
meillo@217: MIME messages can consist of multiple parts.
meillo@217: Some parts, like binary attachments or text content in foreign charsets,
meillo@217: are not directly displayable.
meillo@58: .Pn show 's
meillo@76: understanding of messages and
meillo@58: .Pn mhl 's
meillo@173: display capabilities could not cope with the task any longer.
meillo@62: .P
meillo@88: Instead of extending these tools, additional tools were written from
meillo@217: scratch and were added to the MH tool chest.
meillo@88: Doing so is encouraged by the tool chest approach.
meillo@88: Modular design is a great advantage for extending a system,
meillo@88: as new tools can be added without interfering with existing ones.
meillo@62: First, the new MIME features were added in form of the single program
meillo@58: .Pn mhn .
meillo@58: The command
meillo@82: .Cl "mhn -show 42
meillo@217: had then shown the message number
meillo@217: .Fn 42 ,
meillo@217: interpreting MIME.
meillo@58: With the 1.0 release of nmh in February 1999, Richard Coleman finished
meillo@58: the split of
meillo@58: .Pn mhn
meillo@88: into a set of specialized tools, which together covered the
meillo@88: multiple aspects of MIME.
meillo@88: One of them was
meillo@69: .Pn mhshow ,
meillo@88: which replaced
meillo@88: .Cl "mhn -show" .
meillo@88: It was capable of displaying MIME messages appropriately.
meillo@62: .P
meillo@88: From then on, two message display tools were part of nmh,
meillo@76: .Pn show
meillo@76: and
meillo@76: .Pn mhshow .
meillo@88: To ease the life of users,
meillo@69: .Pn show
meillo@69: was extended to automatically hand the job over to
meillo@69: .Pn mhshow
meillo@69: if displaying the message would be beyond
meillo@69: .Pn show 's
meillo@69: abilities.
meillo@88: In consequence, the user would simply invoke
meillo@69: .Pn show
meillo@69: (possibly through
meillo@69: .Pn next
meillo@69: or
meillo@69: .Pn prev )
meillo@69: and get the message printed with either
meillo@69: .Pn show
meillo@69: or
meillo@69: .Pn mhshow ,
meillo@69: whatever was more appropriate.
meillo@69: .P
meillo@217: Having two similar tools for basically the same task is redundancy.
meillo@217: Usually, users do not distinguish between
meillo@88: .Pn show
meillo@88: and
meillo@88: .Pn mhshow
meillo@88: in their daily mail reading.
meillo@217: Having two separate display programs was therefore unnecessary
meillo@88: from a user's point of view.
meillo@88: Besides, the development of both programs needed to be in sync,
meillo@76: to ensure that the programs behaved in a similar way,
meillo@76: because they were used like a single tool.
meillo@76: Different behavior would have surprised the user.
meillo@69: .P
meillo@69: Today, non-MIME messages are rather seen to be a special case of
meillo@100: MIME messages, although it is the other way round.
meillo@69: As
meillo@69: .Pn mhshow
meillo@217: already had been able to display non-MIME messages, it appeared natural
meillo@69: to drop
meillo@69: .Pn show
meillo@69: in favor of using
meillo@69: .Pn mhshow
meillo@217: exclusively
meillo@217: .Ci 4c1efddfd499300c7e74263e57d8aa137e84c853 .
meillo@88: Removing
meillo@88: .Pn show
meillo@217: is no loss in function, because
meillo@88: .Pn mhshow
meillo@88: covers it completely.
meillo@217: Yet, the old behavior of
meillo@88: .Pn show
meillo@88: can still be emulated with the simple command line:
meillo@88: .VS
meillo@88: mhl `mhpath c`
meillo@88: VE
meillo@88: .P
meillo@76: For convenience,
meillo@76: .Pn mhshow
meillo@88: was renamed to
meillo@88: .Pn show
meillo@88: after
meillo@88: .Pn show
meillo@88: was gone.
meillo@88: It is clear that such a rename may confuse future developers when
meillo@88: trying to understand the history.
meillo@88: Nevertheless, I consider the convenience on the user's side,
meillo@217: to outweigh the inconvenience for understanding the evolution
meillo@217: of the tools.
meillo@69: .P
meillo@88: To prepare for the transition,
meillo@69: .Pn mhshow
meillo@69: was reworked to behave more like
meillo@69: .Pn show
meillo@217: first (cf. Sec.
meillo@217: .Cf mhshow ).
meillo@88: Once the tools behaved more alike, the replacing appeared to be
meillo@88: even more natural.
meillo@88: Today, mmh's new
meillo@69: .Pn show
meillo@217: has become the one single message display program once again,
meillo@159: with the difference
meillo@88: that today it handles MIME messages as well as non-MIME messages.
meillo@217: The outcomes of the transition are one program less to maintain,
meillo@88: no second display program for users to deal with,
meillo@88: and less system complexity.
meillo@69: .P
meillo@88: Still, removing the old
meillo@69: .Pn show
meillo@88: hurts in one regard: It had been such a simple program.
meillo@159: Its lean elegance is missing from the new
meillo@159: .Pn show ,
meillo@225: but there is no alternative;
meillo@159: supporting MIME demands higher essential complexity.
meillo@58: 
meillo@131: 
meillo@133: 
meillo@100: .H2 "Configure Options
meillo@58: .P
meillo@76: Customization is a double-edged sword.
meillo@76: It allows better suiting setups, but not for free.
meillo@76: There is the cost of code complexity to be able to customize.
meillo@76: There is the cost of less tested setups, because there are
meillo@171: more possible setups and especially corner cases.
meillo@159: Additionally, there is the cost of choice itself.
meillo@76: The code complexity directly affects the developers.
meillo@173: Less tested code affects both users and developers.
meillo@217: The problem of choice affects the users, for once by having to choose
meillo@159: but also by more complex interfaces that require more documentation.
meillo@159: Whenever options add few advantages but increase the complexity of the
meillo@159: system, they should be considered for removal.
meillo@72: I have reduced the number of project-specific configure options from 
meillo@217: 15 to 3.
meillo@74: 
meillo@76: .U3 "Mail Transfer Facilities
meillo@74: .P
meillo@217: With the removal of the mail transfer facilities 5 configure
meillo@85: options vanished:
meillo@85: .P
meillo@85: The switches
meillo@85: .Sw --with-tls
meillo@85: and
meillo@85: .Sw --with-cyrus-sasl
meillo@89: had activated the support for transfer encryption and authentication.
meillo@159: .\" XXX cf
meillo@159: .\" XXX gruende kurz wiederholen
meillo@217: They are not needed anymore.
meillo@85: .Ci fecd5d34f65597a4dfa16aeabea7d74b191532c3
meillo@85: .Ci 156d35f6425bea4c1ed3c4c79783dc613379c65b
meillo@85: .P
meillo@159: .\" XXX cf
meillo@159: .\" XXX ``For the same reason ...''
meillo@85: The configure switch
meillo@85: .Sw --enable-pop
meillo@85: activated the message retrieval facility.
meillo@217: Whereas the code area that had been conditionally compiled in
meillo@217: for TLS and SASL support was small,
meillo@217: the conditionally compiled code area for POP support was much larger.
meillo@217: The code base had only changed slightly on toggling TLS or SASL
meillo@217: support but it had changed much on toggling POP support.
meillo@85: The changes in the code base could hardly be overviewed.
meillo@159: By having POP support togglable, a second code base had been created,
meillo@85: one that needed to be tested.
meillo@85: This situation is basically similar for the conditional TLS and SASL  
meillo@85: code, but there the changes are minor and can yet be overviewed.
meillo@85: Still, conditional compilation of a code base creates variations
meillo@85: of the original program.
meillo@85: More variations require more testing and maintenance work.
meillo@85: .P
meillo@85: Two other options only specified default configuration values:
meillo@100: .Sw --with-mts
meillo@217: defined the default transport service
meillo@217: .Ci f6aa95b724fd8c791164abe7ee5468bf5c34f226 .
meillo@85: With
meillo@100: .Sw --with-smtpservers
meillo@217: default SMTP servers could be specified
meillo@217: .Ci 128545e06224233b7e91fc4c83f8830252fe16c9 .
meillo@164: Both of them became irrelevant when the SMTP transport service was removed.
meillo@164: In mmh, all messages are handed over to
meillo@164: .Pn sendmail
meillo@164: for transportation.
meillo@164: 
meillo@72: 
meillo@74: .U3 "Backup Prefix
meillo@74: .P
meillo@76: The backup prefix is the string that was prepended to message
meillo@76: filenames to tag them as deleted.
meillo@173: By default it had been the comma character (`\fL,\fP').
meillo@78: In July 2000, Kimmo Suominen introduced
meillo@78: the configure option
meillo@78: .Sw --with-hash-backup
meillo@173: to change the default to the hash character `\f(CW#\fP'.
meillo@217: This choice was probably personal preference, but,
meillo@217: being related or not, words that start with the hash character
meillo@78: introduce a comment in the Unix shell.
meillo@72: Thus, the command line
meillo@72: .Cl "rm #13 #15
meillo@72: calls
meillo@72: .Pn rm
meillo@217: without arguments because the first hash character starts a comment
meillo@72: that reaches until the end of the line.
meillo@72: To delete the backup files,
meillo@72: .Cl "rm ./#13 ./#15"
meillo@72: needs to be used.
meillo@217: Thus, using the hash as backup prefix may be seen as a precaution
meillo@217: against backup loss.
meillo@78: .P
meillo@159: First, I removed the configure option but added the profile entry
meillo@217: .Pe Backup-Prefix ,
meillo@217: which allowed to specify an arbitrary string as backup prefix
meillo@217: .Ci 6c40d481d661d532dd527eaf34cebb6d3f8ed086 .
meillo@76: This change did not remove the choice but moved it to a location where
meillo@217: it suited better, in my eyes.
meillo@76: .P
meillo@217: Eventually however, the new trash folder concept
meillo@154: (cf. Sec.
meillo@154: .Cf trash-folder )
meillo@164: removed the need for the backup prefix completely.
meillo@78: .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
meillo@133: .Ci ca0b3e830b86700d9e5e31b1784de2bdcaf58fc5
meillo@133: 
meillo@76: 
meillo@76: .U3 "Editor and Pager
meillo@74: .P
meillo@74: The two configure options
meillo@74: .CW --with-editor=EDITOR
meillo@74: .CW --with-pager=PAGER
meillo@74: were used to specify the default editor and pager at configure time.
meillo@109: Doing so at configure time made sense in the eighties,
meillo@76: when the set of available editors and pagers varied much across
meillo@76: different systems.
meillo@89: Today, the situation is more homogeneous.
meillo@74: The programs
meillo@74: .Pn vi
meillo@74: and
meillo@74: .Pn more
meillo@76: can be expected to be available on every Unix system,
meillo@74: as they are specified by POSIX since two decades.
meillo@74: (The specifications for
meillo@74: .Pn vi
meillo@74: and
meillo@74: .Pn more
meillo@74: appeared in
meillo@74: .[
meillo@74: posix 1987
meillo@74: .]
meillo@74: and,
meillo@74: .[
meillo@74: posix 1992
meillo@74: .]
meillo@74: respectively.)
meillo@217: As a first step, these two tools were hard-coded as defaults
meillo@217: .Ci 5d43a99db70c12a673028c7758c20cbe3e13ef5f .
meillo@74: Not changed were the
meillo@74: .Pe editor
meillo@74: and
meillo@74: .Pe moreproc
meillo@76: profile entries, which allowed the user to override the system defaults.
meillo@217: Later, the concept was reworked again to respect the standard
meillo@217: environment variables
meillo@74: .Ev VISUAL
meillo@74: and
meillo@74: .Ev PAGER
meillo@76: if they are set.
meillo@74: Today, mmh determines the editor to use in the following order,
meillo@217: taking the first available and non-empty item
meillo@217: .Ci f85f4b7ae62e3d05a945dcd46ead51f0a2a89a9b :
meillo@171: .LI 1
meillo@74: Environment variable
meillo@74: .Ev MMHEDITOR
meillo@171: .LI 2
meillo@74: Profile entry
meillo@74: .Pe Editor
meillo@171: .LI 3
meillo@74: Environment variable
meillo@74: .Ev VISUAL
meillo@171: .LI 4
meillo@74: Environment variable
meillo@74: .Ev EDITOR
meillo@171: .LI 5
meillo@74: Command
meillo@74: .Pn vi .
meillo@171: .LP
meillo@217: The pager to use is determined in a similar order
meillo@217: .Ci 0c4214ea2aec6497d0d67b436bbee9bc1d225f1e :
meillo@171: .LI 1
meillo@74: Environment variable
meillo@74: .Ev MMHPAGER
meillo@171: .LI 2
meillo@74: Profile entry
meillo@74: .Pe Pager
meillo@74: (replaces
meillo@74: .Pe moreproc )
meillo@171: .LI 3
meillo@74: Environment variable
meillo@74: .Ev PAGER
meillo@171: .LI 4
meillo@74: Command
meillo@74: .Pn more .
meillo@171: .LP
meillo@76: By respecting the
meillo@74: .Ev VISUAL /\c
meillo@74: .Ev EDITOR
meillo@74: and
meillo@74: .Ev PAGER
meillo@76: environment variables,
meillo@217: the new behavior complies with the common style on Unix systems.
meillo@217: It is more uniform and clearer for users.
meillo@72: 
meillo@72: 
meillo@76: .U3 "ndbm
meillo@72: .P
meillo@74: .Pn slocal
meillo@217: used to depend on the database library
meillo@217: .I ndbm .
meillo@217: The database is used to store the
meillo@217: .Hd Message-ID
meillo@217: header field values of all messages delivered.
meillo@217: This enabled
meillo@74: .Pn slocal
meillo@74: to suppress delivering the same message to the same user twice.
meillo@217: This features was enabled by the
meillo@74: .Sw -suppressdup
meillo@217: switch.
meillo@74: .P
meillo@217: As a variety of versions of the database library exist,
meillo@78: .[
meillo@78: wolter unix incompat notes dbm
meillo@78: .]
meillo@217: complicated autoconf code was needed to detect them correctly.
meillo@181: Furthermore, the configure switches
meillo@74: .Sw --with-ndbm=ARG
meillo@74: and
meillo@74: .Sw --with-ndbmheader=ARG
meillo@74: were added to help with difficult setups that would
meillo@217: not be detected automatically or not correctly.
meillo@74: .P
meillo@74: By removing the suppress duplicates feature of
meillo@74: .Pn slocal ,
meillo@74: the dependency on
meillo@74: .I ndbm
meillo@217: vanished and 120 lines of complex autoconf code could be saved
meillo@217: .Ci ecd6d6a20cb7a1507e3a20d6c4cb3a1cf14c6bbf .
meillo@217: The change removed functionality but that is considered minor to the
meillo@217: improvement of dropping the dependency and the complex autoconf code.
meillo@225: 
meillo@72: 
meillo@217: .U3 "MH-E Support
meillo@72: .P
meillo@74: The configure option
meillo@74: .Sw --disable-mhe
meillo@217: was removed when the MH-E support was reworked. 
meillo@217: MH-E is the Emacs front-end to MH.
meillo@76: It requires MH to provide minor additional functions.
meillo@76: The
meillo@76: .Sw --disable-mhe
meillo@217: configure option had switched off these extensions.
meillo@217: After removing the support for old versions of MH-E,
meillo@74: only the
meillo@74: .Sw -build
meillo@76: switches of
meillo@74: .Pn forw
meillo@74: and
meillo@74: .Pn repl
meillo@217: are left to be MH-E extensions.
meillo@76: They are now always built in because they add little code and complexity.
meillo@76: In consequence, the
meillo@74: .Sw --disable-mhe
meillo@76: configure option was removed
meillo@217: .Ci a7ce7b4a580d77b6c2c4d980812beb589aa4c643 .
meillo@217: Dropping the option also removed a variant of the code base
meillo@217: that would have needed to be tested.
meillo@217: This change was undertaken in January 2012 in nmh and
meillo@217: thereafter merged into mmh.
meillo@217: 
meillo@72: 
meillo@74: .U3 "Masquerading
meillo@72: .P
meillo@74: The configure option
meillo@74: .Sw --enable-masquerade
meillo@76: could take up to three arguments:
meillo@217: .Ar draft_from ,
meillo@217: .Ar mmailid ,
meillo@217: and
meillo@217: .Ar username_extension .
meillo@74: They activated different types of address masquerading.
meillo@74: All of them were implemented in the SMTP-speaking
meillo@74: .Pn post
meillo@217: command.
meillo@76: Address masquerading is an MTA's task and mmh does not cover
meillo@76: this field anymore.
meillo@76: Hence, true masquerading needs to be implemented in the external MTA.
meillo@74: .P
meillo@74: The
meillo@74: .I mmailid
meillo@74: masquerading type is the oldest one of the three and the only one
meillo@74: available in the original MH.
meillo@74: It provided a
meillo@74: .I username
meillo@74: to
meillo@74: .I fakeusername
meillo@217: mapping, based on the
meillo@217: .Fn passwd 's
meillo@217: GECOS field.
meillo@217: Nmh's man page
meillo@181: .Mp mh-tailor (5)
meillo@74: described the use case as being the following:
meillo@98: .QS
meillo@74: This is useful if you want the messages you send to always
meillo@74: appear to come from the name of an MTA alias rather than your
meillo@74: actual account name.  For instance, many organizations set up
meillo@74: `First.Last' sendmail aliases for all users.  If this is
meillo@74: the case, the GECOS field for each user should look like:
meillo@74: ``First [Middle] Last <First.Last>''
meillo@98: .QE
meillo@74: .P
meillo@74: As mmh sends outgoing mail via the local MTA only,
meillo@76: the best location to do such global rewrites is there.
meillo@74: Besides, the MTA is conceptionally the right location because it
meillo@74: does the reverse mapping for incoming mail (aliasing), too.
meillo@181: Furthermore, masquerading set up there is readily available for all
meillo@74: mail software on the system.
meillo@76: Hence, mmailid masquerading was removed.
meillo@74: .Ci 0836c8000ccb34b59410ef1c15b1b7feac70ce5f
meillo@74: .P
meillo@74: The
meillo@74: .I username_extension
meillo@76: masquerading type did not replace the username but would append a suffix,
meillo@76: specified by the
meillo@74: .Ev USERNAME_EXTENSION
meillo@76: environment variable, to it.
meillo@76: This provided support for the
meillo@74: .I user-extension
meillo@217: feature of qmail
meillo@217: .[ [
meillo@217: sill qmail handbook
meillo@217: .], p. 141]
meillo@217: and the similar
meillo@74: .I "plussed user
meillo@217: processing of Sendmail.
meillo@217: .[ [
meillo@217: sendmail costales
meillo@217: .], p. 476]
meillo@217: The decision to remove this username_extension masquerading
meillo@217: was motivated by the fact that
meillo@74: .Pn spost
meillo@217: had not supported it yet.
meillo@217: Username extensions can be used in mmh, but less convenient.
meillo@225: .\" XXX In the format file: %(getenv USERNAME_EXTENSION)
meillo@217: .Ci 2abae0bfd0ad5bf898461e50aa4b466d641f23d9
meillo@74: .P
meillo@74: The
meillo@74: .I draft_from
meillo@74: masquerading type instructed
meillo@74: .Pn post
meillo@84: to use the value of the
meillo@84: .Hd From
meillo@84: header field as SMTP envelope sender.
meillo@76: Sender addresses could be replaced completely.
meillo@76: Mmh offers a kind of masquerading similar in effect, but
meillo@74: with technical differences.
meillo@76: As mmh does not transfer messages itself, the local MTA has final control
meillo@217: over the sender's address.
meillo@217: Any masquerading mmh introduces may be reverted by the MTA.
meillo@76: In times of pedantic spam checking, an MTA will take care to use
meillo@76: sensible envelope sender addresses to keep its own reputation up.
meillo@84: Nonetheless, the MUA can set the
meillo@84: .Hd From
meillo@217: header field and thereby propose a sender address to the MTA.
meillo@74: The MTA may then decide to take that one or generate the canonical sender
meillo@74: address for use as envelope sender address.
meillo@217: .Ci b14ea6073f77b4359aaf3fddd0e105989db9
meillo@74: .P
meillo@74: In mmh, the MTA will always extract the recipient and sender from the
meillo@84: message header (\c
meillo@74: .Pn sendmail 's
meillo@74: .Sw -t
meillo@74: switch).
meillo@84: The
meillo@84: .Hd From
meillo@84: header field of the draft may be set arbitrary by the user.
meillo@74: If it is missing, the canonical sender address will be generated by the MTA.
meillo@74: 
meillo@74: .U3 "Remaining Options
meillo@74: .P
meillo@74: Two configure options remain in mmh.
meillo@74: One is the locking method to use:
meillo@74: .Sw --with-locking=[dot|fcntl|flock|lockf] .
meillo@217: The idea of removing all methods except the portable
meillo@217: .I "dot locking
meillo@76: and having that one as the default is appealing, but this change
meillo@76: requires deeper technical investigation into the topic.
meillo@76: The other option,
meillo@74: .Sw --enable-debug ,
meillo@217: compiles the programs with debugging symbols.
meillo@74: This option is likely to stay.
meillo@72: 
meillo@72: 
meillo@58: 
meillo@63: 
meillo@100: .H2 "Command Line Switches
meillo@58: .P
meillo@217: The command line switches of MH tools follow a style similar to
meillo@217: the X Window System style.
meillo@217: The switches consist of a single dash (`\fL-\fP') followed by a word.
meillo@217: For example
meillo@217: .Cl -truncate .
meillo@212: To ease typing, the word can be abbreviated, given the remaining
meillo@217: prefix is unambiguous.
meillo@212: If no other switch starts with the letter `t', then any of
meillo@212: .Cl "-truncate" ,
meillo@212: .Cl "-trunc" ,
meillo@212: .Cl "-tr" ,
meillo@212: and
meillo@212: .Cl "-t
meillo@212: is equal.
meillo@212: As a result, switches can neither be grouped (as in
meillo@212: .Cl "ls -ltr" )
meillo@212: nor can switch arguments be appended directly to the switch (as in
meillo@212: .Cl "sendmail -q30m" ).
meillo@212: Many switches have negating counter-parts, which start with `no'.
meillo@212: For example
meillo@212: .Cl "-notruncate
meillo@212: inverts the
meillo@212: .Cl "-truncate
meillo@212: switch.
meillo@212: They exist to override the effect of default switches in the profile.
meillo@93: Every program in mmh has two generic switches:
meillo@93: .Sw -help ,
meillo@93: to print a short message on how to use the program, and 
meillo@159: .Sw -Version
meillo@164: (with capital `V'), to tell what version of mmh the program belongs to.
meillo@93: .P
meillo@93: Switches change the behavior of programs.
meillo@93: Programs that do one thing in one way require no switches.
meillo@93: In most cases, doing something in exactly one way is too limiting.
meillo@217: If one task should be accomplished in various ways,
meillo@217: switches are a good approach to alter the behavior of a program.
meillo@93: Changing the behavior of programs provides flexibility and customization
meillo@217: to users, but at the same time it complicates the code,
meillo@217: the documentation, and the usage of the program.
meillo@93: Therefore, the number of switches should be kept small.
meillo@217: A small set of well-chosen switches is best.
meillo@217: Usually, the number of switches increases over time.
meillo@93: Already in 1985, Rose and Romine have identified this as a major
meillo@93: problem of MH:
meillo@93: .[ [
meillo@93: rose romine real work
meillo@93: .], p. 12]
meillo@98: .QS
meillo@93: A complaint often heard about systems which undergo substantial development
meillo@93: by many people over a number of years, is that more and more options are
meillo@93: introduced which add little to the functionality but greatly increase the
meillo@93: amount of information a user needs to know in order to get useful work done.
meillo@93: This is usually referred to as creeping featurism.
meillo@93: .QP
meillo@93: Unfortunately MH, having undergone six years of off-and-on development by
meillo@93: ten or so well-meaning programmers (the present authors included),
meillo@93: suffers mightily from this.
meillo@98: .QE
meillo@93: .P
meillo@217: Being reluctant to adding new switches (or \fIoptions\fP,
meillo@217: as Rose and Romine call them) is one part of a counter-action,
meillo@97: the other part is removing hardly used switches.
meillo@217: Nmh's tools have lots of switches already implemented.
meillo@217: Hence, cleaning up by removing some of them was the more important part
meillo@97: of the counter-action.
meillo@93: Removing existing functionality is always difficult because it
meillo@93: breaks programs that use these functions.
meillo@93: Also, for every obsolete feature, there'll always be someone who still
meillo@93: uses it and thus opposes its removal.
meillo@93: This puts the developer into the position,
meillo@93: where sensible improvements to style are regarded as destructive acts.
meillo@97: Yet, living with the featurism is far worse, in my eyes, because
meillo@97: future needs will demand adding further features,
meillo@93: worsening the situation more and more.
meillo@93: Rose and Romine added in a footnote,
meillo@93: ``[...]
meillo@93: .Pn send
meillo@217: will no doubt acquire an endless number of switches in the years to come''
meillo@217: .[ [
meillo@217: rose romine real work
meillo@217: .], p. 12].
meillo@97: Although clearly humorous, the comment points to the nature of the problem.
meillo@97: Refusing to add any new switches would encounter the problem at its root,
meillo@97: but this is not practical.
meillo@97: New needs will require new switches and it would be unwise to block
meillo@97: them strictly.
meillo@97: Nevertheless, removing obsolete switches still is an effective approach
meillo@97: to deal with the problem.
meillo@97: Working on an experimental branch without an established user base,
meillo@97: eased my work because I did not offend users when I removed existing
meillo@110: functions.
meillo@93: .P
meillo@93: Rose and Romine counted 24 visible and 9 more hidden switches for
meillo@93: .Pn send .
meillo@97: In nmh, they increased up to 32 visible and 12 hidden ones.
meillo@182: At the time of writing, no more than 4 visible switches and 1 hidden switch
meillo@97: have remained in mmh's
meillo@97: .Pn send .
meillo@217: These numbers include the two generic switches,
meillo@182: .Sw -help
meillo@182: and
meillo@183: .Sw -Version .
meillo@217: .P
meillo@183: Hidden switches are ones not documented.
meillo@183: In mmh, 12 tools have hidden switches.
meillo@183: 9 of them are
meillo@183: .Sw -debug
meillo@183: switches, the other 6 provide special interfaces for internal use.
meillo@93: .P
meillo@217: The following figure displays the number of switches for each of the tools
meillo@159: that is available in both nmh and mmh.
meillo@100: The tools are sorted by the number of switches they had in nmh.
meillo@217: Both visible and hidden switches were counted,
meillo@97: but not the generic help and version switches.
meillo@93: Whereas in the beginning of the project, the average tool had 11 switches,
meillo@93: now it has no more than 5 \(en only half as many.
meillo@93: If the `no' switches and similar inverse variant are folded onto
meillo@100: their counter-parts, the average tool had 8 switches in pre-mmh times and
meillo@100: has 4 now.
meillo@93: The total number of functional switches in mmh dropped from 465
meillo@182: to 233.
meillo@58: 
meillo@93: .KS
meillo@93: .in 1c
meillo@93: .so input/switches.grap
meillo@93: .KE
meillo@58: 
meillo@93: .P
meillo@93: A part of the switches vanished after functions were removed.
meillo@93: This was the case for network mail transfer, for instance.
meillo@97: Sometimes, however, the work flow was the other way:
meillo@97: I looked through the
meillo@97: .Mp mh-chart (7)
meillo@97: man page to identify the tools with apparently too many switches.
meillo@217: Then I considered the benefit of each switch by examining
meillo@217: the tool's man page and source code, aided by literature research
meillo@217: and testing.
meillo@97: 
meillo@58: 
meillo@93: .U3 "Draft Folder Facility
meillo@93: .P
meillo@100: A change early in the project was the complete transition from
meillo@217: the single draft message to the draft folder facility
meillo@217: .Ci 337338b404931f06f0db2119c9e145e8ca5a9860 .
meillo@164: .\" XXX ref to section ...
meillo@109: The draft folder facility was introduced in the mid-eighties, when
meillo@100: Rose and Romine called it a ``relatively new feature''.
meillo@93: .[
meillo@93: rose romine real work
meillo@93: .]
meillo@217: Since then, the facility was included, inactive by default.
meillo@217: By making it permanently active and by related rework of the tools, the
meillo@93: .Sw -[no]draftfolder ,
meillo@93: and
meillo@93: .Sw -draftmessage
meillo@217: switches could be removed from
meillo@93: .Pn comp ,
meillo@93: .Pn repl ,
meillo@93: .Pn forw ,
meillo@93: .Pn dist ,
meillo@93: .Pn whatnow ,
meillo@93: and
meillo@217: .Pn send
meillo@217: .Ci 337338b404931f06f0db2119c9e145e8ca5a9860 .
meillo@217: The only flexibility lost with this change is having multiple
meillo@97: draft folders within one profile.
meillo@97: I consider this a theoretical problem only.
meillo@159: At the same time, the
meillo@93: .Sw -draft
meillo@93: switch of
meillo@93: .Pn anno ,
meillo@93: .Pn refile ,
meillo@93: and
meillo@93: .Pn send
meillo@93: was removed.
meillo@159: The special treatment of \fIthe\fP draft message became irrelevant after
meillo@217: the rework of the draft system
meillo@159: (cf. Sec.
meillo@217: .Cf draft-folder ).
meillo@164: Furthermore,
meillo@95: .Pn comp
meillo@164: no longer needs a
meillo@95: .Sw -file
meillo@164: switch as the draft folder facility together with the
meillo@95: .Sw -form
meillo@164: switch are sufficient.
meillo@93: 
meillo@95: 
meillo@102: .U3 "In Place Editing
meillo@93: .P
meillo@93: .Pn anno
meillo@93: had the switches
meillo@93: .Sw -[no]inplace
meillo@100: to either annotate the message in place and thus preserve hard links,
meillo@217: or annotate a copy to replace the original message.
meillo@217: The latter approach broke hard links.
meillo@97: Following the assumption that linked messages should truly be the
meillo@217: same message and annotating it should not break the link, the
meillo@93: .Sw -[no]inplace
meillo@93: switches were removed and the previous default
meillo@93: .Sw -inplace
meillo@217: was made the definitive behavior
meillo@217: .Ci c8195849d2e366c569271abb0f5f60f4ebf0b4d0 .
meillo@93: The
meillo@93: .Sw -[no]inplace
meillo@93: switches of
meillo@93: .Pn repl ,
meillo@93: .Pn forw ,
meillo@93: and
meillo@93: .Pn dist
meillo@217: could be removed, as well, as they were simply passed through to
meillo@93: .Pn anno .
meillo@93: .P
meillo@93: .Pn burst
meillo@93: also had
meillo@93: .Sw -[no]inplace
meillo@217: switches, but with a different meaning.
meillo@95: With
meillo@95: .Sw -inplace ,
meillo@95: the digest had been replaced by the table of contents (i.e. the
meillo@110: introduction text) and the burst messages were placed right
meillo@95: after this message, renumbering all following messages.
meillo@95: Also, any trailing text of the digest was lost, though,
meillo@95: in practice, it usually consists of an end-of-digest marker only.
meillo@217: Nonetheless, this behavior appeared less elegant than the
meillo@95: .Sw -noinplace
meillo@95: behavior, which already had been the default.
meillo@95: Nmh's
meillo@95: .Mp burst (1)
meillo@95: man page reads:
meillo@98: .QS
meillo@164: If
meillo@164: .Sw -noinplace
meillo@164: is given, each digest is preserved, no table
meillo@93: of contents is produced, and the messages contained within
meillo@93: the digest are placed at the end of the folder. Other messages
meillo@93: are not tampered with in any way.
meillo@98: .QE
meillo@95: .LP
meillo@93: The decision to drop the
meillo@93: .Sw -inplace
meillo@95: behavior was supported by the code complexity and the possible data loss
meillo@95: it caused.
meillo@93: .Sw -noinplace
meillo@95: was chosen to be the definitive behavior.
meillo@97: .Ci 68a686adeb39223a5e1ad35e4a24890ec053679d
meillo@93: 
meillo@95: 
meillo@95: .U3 "Forms and Format Strings
meillo@93: .P
meillo@95: Historically, the tools that had
meillo@95: .Sw -form
meillo@95: switches to supply a form file had
meillo@95: .Sw -format
meillo@95: switches as well to supply the contents of a form file as a string
meillo@95: on the command line directly.
meillo@95: In consequence, the following two lines equaled:
meillo@95: .VS
meillo@95: scan -form scan.mailx
meillo@217: scan -format "`cat /path/to/scan.mailx`"
meillo@95: VE
meillo@95: The
meillo@95: .Sw -format
meillo@95: switches were dropped in favor for extending the
meillo@95: .Sw -form
meillo@217: switches
meillo@217: .Ci f51956be123db66b00138f80464d06f030dbb88d .
meillo@217: If their argument starts with an equal sign (`\fL=\fP'),
meillo@95: then the rest of the argument is taken as a format string,
meillo@95: otherwise the arguments is treated as the name of a format file.
meillo@95: Thus, now the following two lines equal:
meillo@95: .VS
meillo@95: scan -form scan.mailx
meillo@217: scan -form "=`cat /path/to/scan.mailx`"
meillo@95: VE
meillo@95: This rework removed the prefix collision between
meillo@95: .Sw -form
meillo@95: and
meillo@95: .Sw -format .
meillo@217: Typing `\fL-fo\fP' is sufficient to specify form file or format string.
meillo@95: .P
meillo@95: The different meaning of
meillo@95: .Sw -format
meillo@95: for
meillo@217: .Pn forw
meillo@217: and
meillo@95: .Pn repl
meillo@95: was removed in mmh.
meillo@95: .Pn forw
meillo@95: was completely switched to MIME-type forwarding, thus removing the
meillo@217: .Sw -[no]format
meillo@217: .Ci 6e271608b7b9c23771523f88d23a4d3593010cf1 .
meillo@95: For
meillo@95: .Pn repl ,
meillo@95: the
meillo@95: .Sw -[no]format
meillo@95: switches were reworked to
meillo@95: .Sw -[no]filter
meillo@217: switches
meillo@217: .Ci 67411b1f95d6ec987b4c732459e1ba8a8ac192c6 .
meillo@95: The
meillo@95: .Sw -format
meillo@95: switches of
meillo@95: .Pn send
meillo@95: and
meillo@95: .Pn post ,
meillo@217: which had a third meaning, were removed likewise
meillo@217: .Ci f3cb7cde0e6f10451b6848678d95860d512224b9 .
meillo@95: Eventually, the ambiguity of the
meillo@95: .Sw -format
meillo@217: switches is resolved by not having such switches anymore in mmh.
meillo@95: 
meillo@95: 
meillo@95: .U3 "MIME Tools
meillo@95: .P
meillo@217: The MIME tools, which once were part of
meillo@100: .Pn mhn
meillo@164: (whatever that stood for),
meillo@95: had several switches that added little practical value to the programs.
meillo@95: The
meillo@95: .Sw -[no]realsize
meillo@95: switches of
meillo@95: .Pn mhbuild
meillo@95: and
meillo@95: .Pn mhlist
meillo@217: were removed
meillo@217: .Ci 8d8f1c3abc586c005c904e52c4adbfe694d2201c .
meillo@217: Real size calculations are done always now because nmh's
meillo@159: .Mp mhbuild (1)
meillo@217: man page states that
meillo@217: ``This provides an accurate count at the expense of a small delay''
meillo@217: with the small delay not being noticable on modern systems.
meillo@95: .P
meillo@95: The
meillo@95: .Sw -[no]check
meillo@95: switches were removed together with the support for
meillo@95: .Hd Content-MD5
meillo@217: header fields [RFC\|1864]
meillo@154: (cf. Sec.
meillo@154: .Cf content-md5 )
meillo@217: .Ci 31dc797eb5178970d68962ca8939da3fd9a8efda .
meillo@95: .P
meillo@95: The
meillo@95: .Sw -[no]ebcdicsafe
meillo@95: and
meillo@95: .Sw -[no]rfc934mode
meillo@95: switches of
meillo@95: .Pn mhbuild
meillo@217: were removed because they are considered obsolete
meillo@97: .Ci 01a3480928da485b4d6109d36d751dfa71799d58
meillo@217: .Ci 3363e2624dce0eb8164cf8b3f1ab385c8ff72e88 .
meillo@95: .P
meillo@95: Content caching of external MIME parts, activated with the
meillo@95: .Sw -rcache
meillo@95: and
meillo@95: .Sw -wcache
meillo@217: switches was completely removed
meillo@217: .Ci d1fefd9f614e4dc3cda16da6c69133c1b2005269 .
meillo@97: External MIME parts are rare today, having a caching facility
meillo@159: for them appears to be unnecessary.
meillo@95: .P
meillo@95: In pre-MIME times,
meillo@95: .Pn mhl
meillo@95: had covered many tasks that are part of MIME handling today.
meillo@95: Therefore,
meillo@95: .Pn mhl
meillo@95: could be simplified to a large extend, reducing the number of its
meillo@217: switches from 21 to 6
meillo@97: .Ci 350ad6d3542a07639213cf2a4fe524e829c1e7b6
meillo@217: .Ci 0e46503be3c855bddaeae3843e1b659279c35d70 .
meillo@95: 
meillo@95: 
meillo@95: 
meillo@95: 
meillo@95: .U3 "Header Printing
meillo@95: .P
meillo@95: .Pn folder 's
meillo@95: data output is self-explaining enough that
meillo@159: displaying the header line makes little sense.
meillo@95: Hence, the
meillo@95: .Sw -[no]header
meillo@217: switch was removed and headers are never printed
meillo@217: .Ci 601cc73d1fa05ce96faa728f036d6c51b91701c7 .
meillo@95: .P
meillo@95: In
meillo@95: .Pn mhlist ,
meillo@95: the
meillo@95: .Sw -[no]header
meillo@217: switches were removed, as well
meillo@217: .Ci b24f96523aaf60e44e04a3ffb1d22e69a13a602f .
meillo@217: In this case, the headers are printed always because the output
meillo@217: is not self-explaining.
meillo@95: .P
meillo@95: .Pn scan
meillo@95: also had
meillo@95: .Sw -[no]header
meillo@95: switches.
meillo@217: Printing this header had been sensible until the introduction of
meillo@217: format strings made it impossible to display column headings.
meillo@95: Only the folder name and the current date remained to be printed.
meillo@217: As this information can be perfectly generated with
meillo@95: .Pn folder
meillo@95: and
meillo@95: .Pn date ,
meillo@217: the switches were removed
meillo@217: .Ci c477dc5d1d03fa6d9a8ab3dd3508c63cbddc044e .
meillo@95: .P
meillo@95: By removing all
meillo@95: .Sw -header
meillo@95: switches, the collision with
meillo@95: .Sw -help
meillo@95: on the first two letters was resolved.
meillo@95: Currently,
meillo@95: .Sw -h
meillo@95: evaluates to
meillo@95: .Sw -help
meillo@95: for all tools of mmh.
meillo@95: 
meillo@95: 
meillo@139: .U3 "Suppressing Edits or the Invocation of the WhatNow Shell
meillo@95: .P
meillo@95: The
meillo@95: .Sw -noedit
meillo@100: switch of
meillo@95: .Pn comp ,
meillo@95: .Pn repl ,
meillo@95: .Pn forw ,
meillo@95: .Pn dist ,
meillo@95: and
meillo@95: .Pn whatnow
meillo@217: was removed and replaced by specifying
meillo@95: .Sw -editor
meillo@217: with an empty argument
meillo@217: .Ci 75fca31a5b9d5c1a99c74ab14c94438d8852fba9 .
meillo@95: (Specifying
meillo@159: .Cl "-editor /bin/true
meillo@217: is nearly the same. It differs only in setting the previous editor.)
meillo@95: .P
meillo@95: The more important change is the removal of the
meillo@95: .Sw -nowhatnowproc
meillo@217: switch
meillo@217: .Ci ee4f43cf2ef0084ec698e4e87159a94c01940622 .
meillo@217: This switch had once introduced an awkward behavior,
meillo@217: as explained in nmh's man page for
meillo@95: .Mp comp (1):
meillo@98: .QS
meillo@164: The
meillo@164: .Sw -editor
meillo@164: .Ar editor
meillo@164: switch indicates the editor to use for
meillo@164: the initial edit. Upon exiting from the editor,
meillo@164: .Pn comp
meillo@164: will invoke the
meillo@164: .Pn whatnow
meillo@164: program. See
meillo@164: .Mp whatnow (1)
meillo@164: for a discussion of available options.
meillo@164: The invocation of this program can be
meillo@164: inhibited by using the
meillo@164: .Sw -nowhatnowproc
meillo@164: switch. (In truth of fact, it is the
meillo@164: .Pn whatnow
meillo@164: program which starts the initial edit.
meillo@164: Hence,
meillo@164: .Sw -nowhatnowproc
meillo@164: will prevent any edit from occurring.)
meillo@98: .QE
meillo@95: .P
meillo@95: Effectively, the
meillo@95: .Sw -nowhatnowproc
meillo@217: switch caused only only a draft message to be created.
meillo@95: As
meillo@159: .Cl "-whatnowproc /bin/true
meillo@217: does the same, the
meillo@95: .Sw -nowhatnowproc
meillo@95: switch was removed for being redundant.
meillo@95: 
meillo@95: 
meillo@95: 
meillo@95: .U3 "Various
meillo@95: .BU
meillo@139: With the removal of MMDF maildrop format support,
meillo@139: .Pn packf
meillo@139: and
meillo@139: .Pn rcvpack
meillo@139: no longer needed their
meillo@139: .Sw -mbox
meillo@139: and
meillo@139: .Sw -mmdf
meillo@139: switches.
meillo@217: The behavior of
meillo@139: .Sw -mbox
meillo@217: is the sole behavior now
meillo@217: .Ci 3916ab66ad5d183705ac12357621ea8661afd3c0 .
meillo@171: Further rework in both tools made the
meillo@139: .Sw -file
meillo@217: switch unnecessary
meillo@217: .Ci ca1023716d4c2ab890696f3e41fa0d94267a940e .
meillo@139: 
meillo@139: .BU
meillo@217: Mmh's tools do no longer clear the screen (\c
meillo@139: .Pn scan 's
meillo@139: and
meillo@139: .Pn mhl 's
meillo@139: .Sw -[no]clear
meillo@139: switches
meillo@139: .Ci e57b17343dcb3ff373ef4dd089fbe778f0c7c270
meillo@139: .Ci 943765e7ac5693ae177fd8d2b5a2440e53ce816e ).
meillo@217: Neither does
meillo@139: .Pn mhl
meillo@139: ring the bell (\c
meillo@139: .Sw -[no]bell
meillo@139: .Ci e11983f44e59d8de236affa5b0d0d3067c192e24 )
meillo@217: nor does it page the output itself (\c
meillo@139: .Sw -length
meillo@139: .Ci 5b9d883db0318ed2b84bb82dee880d7381f99188 ).
meillo@159: .\" XXX Ref
meillo@139: Generally, the pager to use is no longer specified with the
meillo@139: .Sw -[no]moreproc
meillo@139: command line switches for
meillo@139: .Pn mhl
meillo@139: and
meillo@139: .Pn show /\c
meillo@217: .Pn mhshow
meillo@217: .Ci 39e87a75b5c2d3572ec72e717720b44af291e88a .
meillo@139: 
meillo@139: .BU
meillo@96: In order to avoid prefix collisions among switch names, the
meillo@95: .Sw -version
meillo@95: switch was renamed to
meillo@95: .Sw -Version
meillo@217: (with capital `V')
meillo@217: .Ci 32b2354dbaf4bf934936eb5b102a4a3d2fdd209a .
meillo@95: Every program has the
meillo@95: .Sw -version
meillo@95: switch but its first three letters collided with the
meillo@95: .Sw -verbose
meillo@95: switch, present in many programs.
meillo@95: The rename solved this problem once for all.
meillo@95: Although this rename breaks a basic interface, having the
meillo@95: .Sw -V
meillo@95: abbreviation to display the version information, isn't all too bad.
meillo@139: 
meillo@95: .BU
meillo@95: .Sw -[no]preserve
meillo@95: of
meillo@95: .Pn refile
meillo@168: was removed
meillo@168: .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
meillo@168: because what use was it anyway?
meillo@168: Quoting nmh's man page
meillo@165: .Mp refile (1):
meillo@98: .QS
meillo@95: Normally when a message is refiled, for each destination
meillo@95: folder it is assigned the number which is one above the current
meillo@95: highest message number in that folder. Use of the
meillo@164: .Sw -preserv
meillo@164: [sic!] switch will override this message renaming, and try
meillo@95: to preserve the number of the message. If a conflict for a
meillo@164: particular folder occurs when using the
meillo@164: .Sw -preserve
meillo@164: switch, then
meillo@164: .Pn refile
meillo@164: will use the next available message number which
meillo@95: is above the message number you wish to preserve.
meillo@98: .QE
meillo@139: 
meillo@95: .BU
meillo@95: The removal of the
meillo@95: .Sw -[no]reverse
meillo@95: switches of
meillo@95: .Pn scan
meillo@97: .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
meillo@217: is a bug fix.
meillo@217: This is supported by the comments
meillo@95: ``\-[no]reverse under #ifdef BERK (I really HATE this)''
meillo@95: by Rose and
meillo@95: ``Lists messages in reverse order with the `\-reverse' switch.
meillo@217: This should be considered a bug'' by Romine in the changelogs.
meillo@217: The question remains why neither Rose nor Romine have fixed this
meillo@217: bug in the eighties when they wrote these comments.
meillo@93: 
meillo@93: 
meillo@93: 
meillo@102: 
meillo@102: 
meillo@95: 
meillo@95: 
meillo@133: .\" --------------------------------------------------------------
meillo@74: .H1 "Modernizing
meillo@102: .P
meillo@164: In the more than thirty years of MH's existence, its code base was
meillo@159: increasingly extended.
meillo@118: New features entered the project and became alternatives to the
meillo@118: existing behavior.
meillo@217: Relics from several decades have gathered in the code base
meillo@118: but seldom obsolete features were dropped.
meillo@118: This section describes the removing of old code
meillo@118: and the modernizing of the default setup.
meillo@118: It focuses on the functional aspect only;
meillo@154: the non-functional aspects of code style are discussed in Sec.
meillo@154: .Cf code-style .
meillo@58: 
meillo@58: 
meillo@212: .H2 "Code Relics
meillo@0: .P
meillo@217: My position regarding the removal of obsolete code
meillo@159: is much more revolutional than the nmh community appreciates.
meillo@217: Working on an experimental version, I was able to quickly drop
meillo@217: functionality that I considered ancient.
meillo@104: The need for consensus with peers would have slowed this process down.
meillo@104: Without the need to justify my decisions, I was able to rush forward.
meillo@217: .P
meillo@110: In December 2011, Paul Vixie motivated the nmh developers to just
meillo@104: do the work:
meillo@104: .[
meillo@104: paul vixie edginess nmh-workers
meillo@104: .]
meillo@104: .QS
meillo@104: let's stop walking on egg shells with this code base. there's no need to
meillo@104: discuss whether to keep using vfork, just note in [sic!] passing, [...]
meillo@104: we don't need a separate branch for removing vmh
meillo@104: or ridding ourselves of #ifdef's or removing posix replacement functions
meillo@164: or depending on pure ansi/posix ``libc''.
meillo@104: .QP
meillo@164: these things should each be a day or two of work and the ``main branch''
meillo@104: should just be modern. [...]
meillo@104: let's push forward, aggressively.
meillo@104: .QE
meillo@104: .LP
meillo@104: I did so already in the months before.
meillo@104: I pushed forward.
meillo@104: I simply dropped the cruft.
meillo@104: .P
meillo@104: The decision to drop a feature was based on literature research and
meillo@159: careful thinking, but whether having had contact with this particular
meillo@104: feature within my own computer life served as a rule of thumb.
meillo@159: I explained my reasons in the commit messages
meillo@109: in the version control system.
meillo@104: Hence, others can comprehend my view and argue for undoing the change
meillo@104: if I have missed an important aspect.
meillo@109: I was quick in dropping parts.
meillo@179: I rather include falsely dropped parts again, than going at a slower pace.
meillo@179: Mmh is experimental work; it requires tough decisions.
meillo@12: 
meillo@102: 
meillo@217: .U3 "Process Forking
meillo@12: .P
meillo@109: Being a tool chest, MH creates many processes.
meillo@104: In earlier times
meillo@104: .Fu fork()
meillo@104: had been an expensive system call, because the process's image needed
meillo@159: to be completely duplicated at once.
meillo@200: This expensive work was especially unnecessary in the commonly occurring
meillo@159: case wherein the image is replaced by a call to
meillo@104: .Fu exec()
meillo@104: right after having forked the child process.
meillo@104: The
meillo@104: .Fu vfork()
meillo@104: system call was invented to speed up this particular case.
meillo@104: It completely omits the duplication of the image.
meillo@104: On old systems this resulted in significant speed ups.
meillo@104: Therefore MH used
meillo@104: .Fu vfork()
meillo@104: whenever possible.
meillo@12: .P
meillo@104: Modern memory management units support copy-on-write semantics, which make
meillo@104: .Fu fork()
meillo@104: almost as fast as
meillo@104: .Fu vfork() .
meillo@104: The man page of
meillo@104: .Mp vfork (2)
meillo@104: in FreeBSD 8.0 states:
meillo@104: .QS
meillo@104: This system call will be eliminated when proper system sharing mechanisms
meillo@104: are implemented. Users should not depend on the memory sharing semantics
meillo@104: of vfork() as it will, in that case, be made synonymous to fork(2).
meillo@104: .QE
meillo@104: .LP
meillo@104: Vixie supports the removal with the note that ``the last
meillo@104: system on which fork was so slow that an mh user would notice it, was
meillo@104: Eunice. that was 1987''.
meillo@104: .[
meillo@104: nmh-workers vixie edginess
meillo@104: .]
meillo@104: I replaced all calls to
meillo@104: .Fu vfork()
meillo@104: with calls to
meillo@217: .Fu fork()
meillo@217: .Ci 40821f5c1316e9205a08375e7075909cc9968e7d .
meillo@104: .P
meillo@104: Related to the costs of
meillo@104: .Fu fork()
meillo@104: is the probability of its success.
meillo@109: In the eighties, on heavy loaded systems, calls to
meillo@104: .Fu fork()
meillo@104: were prone to failure.
meillo@104: Hence, many of the
meillo@104: .Fu fork()
meillo@104: calls in the code were wrapped into loops to retry the
meillo@104: .Fu fork()
meillo@217: several times, to increase the chances to succeed eventually.
meillo@109: On modern systems, a failing
meillo@104: .Fu fork()
meillo@109: call is unusual.
meillo@104: Hence, in the rare case when
meillo@104: .Fu fork()
meillo@217: fails, mmh programs simply abort
meillo@217: .Ci 5fbf37ee68e018998ada61eeab73e035b26834b6 .
meillo@12: 
meillo@12: 
meillo@109: .U3 "Header Fields
meillo@104: .BU
meillo@84: The
meillo@84: .Hd Encrypted
meillo@104: header field was introduced by RFC\|822,
meillo@109: but already marked as legacy in RFC\|2822.
meillo@109: Today, OpenPGP provides the basis for standardized exchange of encrypted
meillo@104: messages [RFC\|4880, RFC\|3156].
meillo@109: Hence, the support for
meillo@104: .Hd Encrypted
meillo@217: header fields is removed in mmh
meillo@217: .Ci 064527f7b57ab050e5af13e15ad99aeeab125857 .
meillo@104: .BU
meillo@159: The native support for
meillo@84: .Hd Face
meillo@217: header fields has been removed, as well
meillo@217: .Ci 8e5be81f784682822f5e868c1bf3c8624682bd23 .
meillo@104: This feature is similar to the
meillo@84: .Hd X-Face
meillo@84: header field in its intent,
meillo@21: but takes a different approach to store the image.
meillo@84: Instead of encoding the image data directly into the header field,
meillo@109: it contains the hostname and UDP port where the image
meillo@109: date can be retrieved.
meillo@159: There is even a third Face system,
meillo@109: which is the successor of
meillo@109: .Hd X-Face ,
meillo@109: although it re-uses the
meillo@104: .Hd Face
meillo@217: header field name.
meillo@109: It was invented in 2005 and supports colored PNG images.
meillo@104: None of the Face systems described here is popular today.
meillo@104: Hence, mmh has no direct support for them.
meillo@104: .BU
meillo@154: .Id content-md5
meillo@104: The
meillo@104: .Hd Content-MD5
meillo@104: header field was introduced by RFC\|1864.
meillo@104: It provides detection of data corruption during the transfer.
meillo@104: But it can not ensure verbatim end-to-end delivery of the contents
meillo@104: [RFC\|1864].
meillo@104: The proper approach to verify content integrity in an
meillo@166: end-to-end relationship is the use of digital signatures.
meillo@104: .\" XXX (RFCs FIXME).
meillo@104: On the other hand, transfer protocols should detect corruption during
meillo@109: the transmission.
meillo@109: The TCP includes a checksum field therefore.
meillo@104: These two approaches in combinations render the
meillo@104: .Hd Content-MD5
meillo@104: header field superfluous.
meillo@109: Not a single one out of 4\|200 messages from two decades
meillo@217: in the nmh-workers mailing list archive
meillo@217: .[
meillo@217: nmh-workers mailing list archive website
meillo@217: .]
meillo@217: contains a
meillo@104: .Hd Content-MD5
meillo@104: header field.
meillo@104: Neither did any of the 60\|000 messages in my personal mail storage.
meillo@217: Removing the support for this header field
meillo@217: .Ci 31dc797eb5178970d68962ca8939da3fd9a8efda ,
meillo@104: removed the last place where MD5 computation was needed.
meillo@104: Hence, the MD5 code could be removed as well.
meillo@104: Over 500 lines of code vanished by this one change.
meillo@104: 
meillo@104: 
meillo@104: .U3 "MMDF maildrop support
meillo@21: .P
meillo@217: This type of maildrop format is conceptionally similar to the mbox format,
meillo@139: but uses a different message delimiter (`\fL\\1\\1\\1\\1\fP',
meillo@139: commonly written as `\fL^A^A^A^A\fP', instead of `\fLFrom\0\fP').
meillo@104: Mbox is the de-facto standard maildrop format on Unix,
meillo@159: whereas the MMDF maildrop format is now forgotten.
meillo@217: Mbox remains as the only packed mailbox format, supported in mmh.
meillo@104: .P
meillo@109: The simplifications within the code were moderate.
meillo@109: Mainly, the reading and writing of MMDF mailbox files was removed.
meillo@109: But also, switches of
meillo@109: .Pn packf
meillo@104: and
meillo@109: .Pn rcvpack
meillo@217: could be removed
meillo@217: .Ci 3916ab66ad5d183705ac12357621ea8661afd3c0 .
meillo@109: In the message parsing function
meillo@109: .Fn sbr/m_getfld.c ,
meillo@217: knowledge of MMDF packed mail boxes was removed
meillo@217: .Ci 684ec30d81e1223a282764452f4902ed4ad1c754 .
meillo@109: Further code structure simplifications may be possible there,
meillo@109: because only one single packed mailbox format is left to be supported.
meillo@104: I have not worked on them yet because
meillo@104: .Fu m_getfld()
meillo@104: is heavily optimized and thus dangerous to touch.
meillo@104: The risk of damaging the intricate workings of the optimized code is
meillo@104: too high.
meillo@104: 
meillo@12: 
meillo@101: .U3 "Prompter's Control Keys
meillo@20: .P
meillo@20: The program
meillo@20: .Pn prompter
meillo@104: queries the user to fill in a message form.
meillo@217: When used as
meillo@104: .Cl "comp -editor prompter" ,
meillo@20: the resulting behavior is similar to
meillo@20: .Pn mailx .
meillo@51: Apparently,
meillo@20: .Pn prompter
meillo@173: had not been touched lately.
meillo@104: Otherwise it's hardly explainable why it
meillo@20: still offered the switches
meillo@84: .Sw -erase
meillo@84: .Ar chr
meillo@20: and
meillo@84: .Sw -kill
meillo@84: .Ar chr
meillo@20: to name the characters for command line editing.
meillo@21: The times when this had been necessary are long time gone.
meillo@20: Today these things work out-of-the-box, and if not, are configured
meillo@20: with the standard tool
meillo@20: .Pn stty .
meillo@104: The switches are removed now
meillo@104: .Ci 0bd9750710cdbab80cfb4036dd87af20afe1552f .
meillo@20: 
meillo@104: 
meillo@109: .U3 "Hardcopy Terminal Support
meillo@21: .P
meillo@109: More of a funny anecdote is a check for being connected to a
meillo@109: hardcopy terminal.
meillo@159: It remained in the code until spring 2012, when I finally removed it
meillo@104: .Ci b7764c4a6b71d37918a97594d866258f154017ca .
meillo@21: .P
meillo@109: The check only prevented a pager to be placed between the printing
meillo@104: program (\c
meillo@104: .Pn mhl )
meillo@104: and the terminal.
meillo@109: In nmh, this could have been ensured statically with the
meillo@104: .Sw -nomoreproc
meillo@109: at the command line, too.
meillo@121: In mmh, setting the profile entry
meillo@104: .Pe Pager
meillo@104: or the environment variable
meillo@104: .Ev PAGER
meillo@104: to
meillo@109: .Pn cat
meillo@159: is sufficient.
meillo@104: 
meillo@104: 
meillo@21: 
meillo@12: 
meillo@58: .H2 "Attachments
meillo@22: .P
meillo@101: The mind model of email attachments is unrelated to MIME.
meillo@217: Although the MIME RFCs [RFC\|2045\(en2049] define the technical
meillo@217: requirements for having attachments, they do not mention the term.
meillo@101: Instead of attachments, MIME talks about ``multi-part message bodies''
meillo@101: [RFC\|2045], a more general concept.
meillo@101: Multi-part messages are messages
meillo@101: ``in which one or more different
meillo@101: sets of data are combined in a single body''
meillo@101: [RFC\|2046].
meillo@101: MIME keeps its descriptions generic;
meillo@101: it does not imply specific usage models.
meillo@217: Today, one usage model is prevalent: attachments.
meillo@101: The idea is having a main text document with files of arbitrary kind
meillo@101: attached to it.
meillo@101: In MIME terms, this is a multi-part message having a text part first
meillo@110: and parts of arbitrary type following.
meillo@101: .P
meillo@101: MH's MIME support is a direct implementation of the RFCs.
meillo@101: The perception of the topic described in the RFCs is clearly visible
meillo@101: in MH's implementation.
meillo@159: As a result,
meillo@159: MH had all the MIME features but no idea of attachments.
meillo@173: But users do not need all the MIME features,
meillo@109: they want convenient attachment handling.
meillo@109: 
meillo@102: 
meillo@102: .U3 "Composing MIME Messages
meillo@102: .P
meillo@102: In order to improve the situation on the message composing side,
meillo@217: Jon Steinhart had added an attachment system to nmh in 2002
meillo@217: .Ci 7480dbc14bc90f2d872d434205c0784704213252 .
meillo@102: In the file
meillo@102: .Fn docs/README-ATTACHMENTS ,
meillo@217: he described his motivation to do so:
meillo@101: .QS
meillo@159: Although nmh contains the necessary functionality for MIME message
meillo@159: handing [sic!], the interface to this functionality is pretty obtuse.
meillo@101: There's no way that I'm ever going to convince my partner to write
meillo@101: .Pn mhbuild
meillo@101: composition files!
meillo@101: .QE
meillo@102: .LP
meillo@102: With this change, the mind model of attachments entered nmh.
meillo@102: In the same document:
meillo@101: .QS
meillo@101: These changes simplify the task of managing attachments on draft files.
meillo@101: They allow attachments to be added, listed, and deleted.
meillo@101: MIME messages are automatically created when drafts with attachments
meillo@101: are sent.
meillo@101: .QE
meillo@102: .LP
meillo@217: Unfortunately, the attachment system, like every new facilities in nmh,
meillo@110: was inactive by default.
meillo@101: .P
meillo@217: During my time in Argentina, I tried to improve the attachment system.
meillo@217: But, after long discussions my patch died as a proposal on the
meillo@217: mailing list because of great opposition in the nmh community.
meillo@101: .[
meillo@101: nmh-workers attachment proposal
meillo@101: .]
meillo@217: In January 2012, I extended the patch and applied it to mmh
meillo@217: .Ci 8ff284ff9167eff8f5349481529332d59ed913b1 .
meillo@102: In mmh, the attachment system is active by default.
meillo@102: Instead of command line switches, the
meillo@102: .Pe Attachment-Header
meillo@102: profile entry is used to specify
meillo@102: the name of the attachment header field.
meillo@102: It is pre-defined to
meillo@102: .Hd Attach .
meillo@101: .P
meillo@159: To add an attachment to a draft, a header line needs to be added:
meillo@101: .VS
meillo@101: To: bob
meillo@101: Subject: The file you wanted
meillo@101: Attach: /path/to/the/file-bob-wanted
meillo@101: --------
meillo@101: Here it is.
meillo@101: VE
meillo@101: The header field can be added to the draft manually in the editor,
meillo@102: or by using the `attach' command at the WhatNow prompt, or
meillo@102: non-interactively with
meillo@101: .Pn anno :
meillo@101: .VS
meillo@102: anno -append -nodate -component Attach -text /path/to/attachment
meillo@101: VE
meillo@102: Drafts with attachment headers are converted to MIME automatically by
meillo@102: .Pn send .
meillo@102: The conversion to MIME is invisible to the user.
meillo@159: The draft stored in the draft folder is always in source form with
meillo@101: attachment headers.
meillo@179: If the MIMEification fails (e.g. because the file to attach
meillo@179: is not accessible) the original draft is not changed.
meillo@101: .P
meillo@102: The attachment system handles the forwarding of messages, too.
meillo@173: If the attachment header value starts with a plus character (`\fL+\fP'),
meillo@101: like in
meillo@101: .Cl "Attach: +bob 30 42" ,
meillo@159: the given messages in the specified folder will be attached.
meillo@101: This allowed to simplify
meillo@217: .Pn forw
meillo@217: .Ci f41f04cf4ceca7355232cf7413e59afafccc9550 .
meillo@101: .P
meillo@101: Closely related to attachments is non-ASCII text content,
meillo@217: because it requires MIME as well.
meillo@102: In nmh, the user needed to call `mime' at the WhatNow prompt
meillo@101: to have the draft converted to MIME.
meillo@102: This was necessary whenever the draft contained non-ASCII characters.
meillo@101: If the user did not call `mime', a broken message would be sent.
meillo@101: Therefore, the
meillo@101: .Pe automimeproc
meillo@101: profile entry could be specified to have the `mime' command invoked
meillo@102: automatically each time.
meillo@179: Unfortunately, this approach conflicted with the attachment system
meillo@101: because the draft would already be in MIME format at the time
meillo@101: when the attachment system wanted to MIMEify it.
meillo@102: To use nmh's attachment system, `mime' must not be called at the
meillo@102: WhatNow prompt and
meillo@101: .Pe automimeproc
meillo@102: must not be set in the profile.
meillo@101: But then the case of non-ASCII text without attachment headers was
meillo@101: not caught.
meillo@102: All in all, the solution was complex and irritating.
meillo@168: My patch from December 2010
meillo@168: .[
meillo@168: nmh-workers attachment proposal
meillo@168: .]
meillo@168: would have simplified the situation.
meillo@102: .P
meillo@101: Mmh's current solution is even more elaborate.
meillo@101: Any necessary MIMEification is done automatically.
meillo@101: There is no `mime' command at the WhatNow prompt anymore.
meillo@102: The draft will be converted automatically to MIME when either an
meillo@102: attachment header or non-ASCII text is present.
meillo@173: Furthermore, the hash character (`\fL#\fP') is not special any more
meillo@159: at line beginnings in the draft message.
meillo@159: .\" XXX REF ?
meillo@159: Users need not concern themselves with the whole topic at all.
meillo@101: .P
meillo@102: Although the new approach does not anymore support arbitrary MIME
meillo@102: compositions directly, the full power of
meillo@101: .Pn mhbuild
meillo@101: can still be accessed.
meillo@217: Given no attachment headers are included, users can create
meillo@101: .Pn mhbuild
meillo@102: composition drafts like in nmh.
meillo@217: Then, at the WhatNow prompt, they can invoke
meillo@101: .Cl "edit mhbuild
meillo@217: to convert the draft to MIME.
meillo@217: Because the resulting draft neither contains non-ASCII characters
meillo@102: nor has it attachment headers, the attachment system will not touch it.
meillo@101: .P
meillo@159: The approach taken in mmh is tailored towards today's most common case:
meillo@159: a text part, possibly with attachments.
meillo@159: This case was simplified.
meillo@102: 
meillo@112: 
meillo@102: .U3 "MIME Type Guessing
meillo@102: .P
meillo@159: From the programmer's point of view, the use of
meillo@101: .Pn mhbuild
meillo@159: composition drafts had one notable advantage over attachment headers:
meillo@159: The user provides the appropriate MIME types for files to include.
meillo@217: The new attachment system needs to find out the correct MIME type itself.
meillo@217: This is a difficult task.
meillo@102: Determining the correct MIME type of content is partly mechanical,
meillo@102: partly intelligent work.
meillo@102: Forcing the user to find out the correct MIME type,
meillo@102: forces him to do partly mechanical work.
meillo@179: Letting the computer do the work can lead to bad choices for difficult
meillo@102: content.
meillo@217: For mmh, the latter option was chosen to spare the user the work
meillo@217: .Ci 3baec236a39c5c89a9bda8dbd988d643a21decc6 .
meillo@102: .P
meillo@102: Determining the MIME type by the suffix of the file name is a dumb
meillo@102: approach, yet it is simple to implement and provides good results
meillo@102: for the common cases.
meillo@217: If no MIME type can be determined, text content is sent as `text/plain',
meillo@217: anything else under the generic fall-back type `application/octet-stream'.
meillo@102: Mmh implements this approach in the
meillo@102: .Pn print-mimetype
meillo@217: script
meillo@217: .Ci 4b5944268ea0da7bb30598a27857304758ea9b44 .
meillo@102: .P
meillo@112: A far better, though less portable, approach is the use of
meillo@102: .Pn file .
meillo@102: This standard tool tries to determine the type of files.
meillo@102: Unfortunately, its capabilities and accuracy varies from system to system.
meillo@102: Additionally, its output was only intended for human beings,
meillo@102: but not to be used by programs.
meillo@102: Nevertheless, modern versions of GNU
meillo@102: .Pn file ,
meillo@217: which are prevalent on the popular GNU/Linux systems,
meillo@159: provide MIME type output in machine-readable form.
meillo@217: Although this solution is system-dependent,
meillo@102: it solves the difficult problem well.
meillo@102: On systems where GNU
meillo@102: .Pn file ,
meillo@102: version 5.04 or higher, is available it should be used.
meillo@102: One needs to specify the following profile entry to do so:
meillo@102: .VS
meillo@102: Mime-Type-Query: file -b --mime
meillo@102: VE
meillo@102: .LP
meillo@102: Other versions of
meillo@102: .Pn file
meillo@217: might possibly be usable with wrapper scripts that reformat the output.
meillo@102: The diversity among
meillo@102: .Pn file
meillo@102: implementations is great; one needs to check the local variant.
meillo@102: .P
meillo@102: It is not possible in mmh to override the automatic MIME type guessing
meillo@102: for a specific file.
meillo@159: To do so, either the user would need to know in advance for which file
meillo@217: the automatic guessing fails or the system would require interaction.
meillo@102: I consider both cases impractical.
meillo@102: The existing solution should be sufficient.
meillo@102: If not, the user may always fall back to
meillo@102: .Pn mhbuild
meillo@217: composition drafts and bypass the attachment system.
meillo@101: 
meillo@102: 
meillo@102: .U3 "Storing Attachments
meillo@102: .P
meillo@169: Extracting MIME parts of a message and storing them to disk is performed by
meillo@108: .Pn mhstore .
meillo@108: The program has two operation modes,
meillo@108: .Sw -auto
meillo@108: and
meillo@108: .Sw -noauto .
meillo@108: With the former one, each part is stored under the filename given in the
meillo@108: MIME part's meta information, if available.
meillo@108: This naming information is usually available for modern attachments.
meillo@108: If no filename is available, this MIME part is stored as if
meillo@108: .Sw -noauto
meillo@108: would have been specified.
meillo@108: In the
meillo@108: .Sw -noauto
meillo@108: mode, the parts are processed according to rules, defined by
meillo@108: .Pe mhstore-store-*
meillo@108: profile entries.
meillo@108: These rules define generic filename templates for storing
meillo@108: or commands to post-process the contents in arbitrary ways.
meillo@108: If no matching rule is available the part is stored under a generic
meillo@108: filename, built from message number, MIME part number, and MIME type.
meillo@108: .P
meillo@108: The
meillo@108: .Sw -noauto
meillo@108: mode had been the default in nmh because it was considered safe,
meillo@108: in contrast to the
meillo@108: .Sw -auto
meillo@108: mode.
meillo@108: In mmh,
meillo@108: .Sw -auto
meillo@108: is not dangerous anymore.
meillo@108: Two changes were necessary:
meillo@171: .LI 1
meillo@108: Any directory path is removed from the proposed filename.
meillo@108: Thus, the files are always stored in the expected directory.
meillo@108: .Ci 41b6eadbcecf63c9a66aa5e582011987494abefb
meillo@171: .LI 2
meillo@108: Tar files are not extracted automatically any more.
meillo@108: Thus, the rest of the file system will not be touched.
meillo@108: .Ci 94c80042eae3383c812d9552089953f9846b1bb6
meillo@217: .P
meillo@217: In mmh, the result of
meillo@108: .Cl "mhstore -auto
meillo@110: can be foreseen from the output of
meillo@108: .Cl "mhlist -verbose" .
meillo@217: Although the
meillo@108: .Sw -noauto
meillo@217: mode is considered to be more powerful, it is less convenient and
meillo@108: .Sw -auto
meillo@217: is safe now.
meillo@217: Additionally, storing attachments under their original name
meillo@217: is intuitive.
meillo@108: Hence,
meillo@108: .Sw -auto
meillo@217: serves better as the default option
meillo@217: .Ci 3410b680416c49a7617491af38bc1929855a331d .
meillo@108: .P
meillo@108: Files are stored into the directory given by the
meillo@108: .Pe Nmh-Storage
meillo@108: profile entry, if set, or
meillo@108: into the current working directory, otherwise.
meillo@108: Storing to different directories is only possible with
meillo@108: .Pe mhstore-store-*
meillo@108: profile entries.
meillo@108: .P
meillo@217: Still existing files get overwritten silently in both modes.
meillo@108: This can be considered a bug.
meillo@108: Yet, each other behavior has its draw-backs, too.
meillo@108: Refusing to replace files requires adding a
meillo@108: .Sw -force
meillo@217: switch.
meillo@108: Users will likely need to invoke
meillo@108: .Pn mhstore
meillo@108: a second time with
meillo@159: .Sw -force .
meillo@159: Eventually, only the user can decide in the specific case.
meillo@108: This requires interaction, which I like to avoid if possible.
meillo@108: Appending a unique suffix to the filename is another bad option.
meillo@108: For now, the behavior remains as it is.
meillo@108: .P
meillo@108: In mmh, only MIME parts of type message are special in
meillo@108: .Pn mhstore 's
meillo@108: .Sw -auto
meillo@108: mode.
meillo@108: Instead of storing message/rfc822 parts as files to disk,
meillo@108: they are stored as messages into the current mail folder.
meillo@159: The same applies to message/partial, although the parts are
meillo@159: automatically reassembled beforehand.
meillo@159: MIME parts of type message/external-body are not automatically retrieved
meillo@159: anymore.
meillo@159: Instead, information on how to retrieve them is output.
meillo@217: Not supporting this rare case saved nearly one thousand lines of code
meillo@217: .Ci 55e1d8c654ee0f7c45b9361ce34617983b454c32 .
meillo@217: The MIME type `application/octet-stream; type=tar' is not special anymore.
meillo@217: The automatically extracting of such MIME parts had been the
meillo@217: dangerous part of the
meillo@108: .Sw -auto
meillo@217: mode
meillo@217: .Ci 94c80042eae3383c812d9552089953f9846b1bb6 .
meillo@108: 
meillo@102: 
meillo@102: 
meillo@102: .U3 "Showing MIME Messages
meillo@102: .P
meillo@114: The program
meillo@114: .Pn mhshow
meillo@217: was written to display MIME messages.
meillo@114: It implemented the conceptional view of the MIME RFCs.
meillo@114: Nmh's
meillo@114: .Pn mhshow
meillo@217: handles each MIME part independently, presenting them separately
meillo@114: to the user.
meillo@114: This does not match today's understanding of email attachments,
meillo@114: where displaying a message is seen to be a single, integrated operation.
meillo@114: Today, email messages are expected to consist of a main text part
meillo@114: plus possibly attachments.
meillo@217: They are no more seen to be arbitrary MIME hierarchies with
meillo@114: information on how to display the individual parts.
meillo@114: I adjusted
meillo@114: .Pn mhshow 's
meillo@114: behavior to the modern view on the topic.
meillo@114: .P
meillo@224: One should note that this section completely ignores the original
meillo@114: .Pn show
meillo@114: program, because it was not capable to display MIME messages
meillo@114: and is no longer part of mmh.
meillo@179: .\" XXX ref to other section
meillo@114: Although
meillo@114: .Pn mhshow
meillo@114: was renamed to
meillo@114: .Pn show
meillo@114: in mmh, this section uses the name
meillo@114: .Pn mhshow ,
meillo@224: in order to avoid confusion.
meillo@114: .P
meillo@114: In mmh, the basic idea is that
meillo@114: .Pn mhshow
meillo@114: should display a message in one single pager session.
meillo@114: Therefore,
meillo@114: .Pn mhshow
meillo@114: invokes a pager session for all its output,
meillo@217: whenever it prints to a terminal
meillo@217: .Ci a4197ea6ffc5c1550e8b52d5a654bcaaaee04a4e .
meillo@114: In consequence,
meillo@114: .Pn mhl
meillo@217: does no more invoke a pager
meillo@217: .Ci 0e46503be3c855bddaeae3843e1b659279c35d70 .
meillo@114: With
meillo@114: .Pn mhshow
meillo@114: replacing the original
meillo@114: .Pn show ,
meillo@217: the output of
meillo@114: .Pn mhl
meillo@217: no longer goes to the terminal directly, but through
meillo@114: .Pn mhshow .
meillo@114: Hence,
meillo@114: .Pn mhl
meillo@114: does not need to invoke a pager.
meillo@114: The one and only job of
meillo@114: .Pn mhl
meillo@114: is to format messages or parts of them.
meillo@114: The only place in mmh, where a pager is invoked is
meillo@114: .Pn mhshow .
meillo@114: .P
meillo@217: In the intended setup, only text content is be displayed,
meillo@217: in a single pager session.
meillo@217: Non-text content needs to be converted to text by appropriate
meillo@217: .Pe mhshow-show-*
meillo@217: profile entries before, if this is possible and wanted.
meillo@217: A common example for this are PDF files.
meillo@217: In mmh, MIME parts are always displayed serially.
meillo@114: The request to display the MIME type `multipart/parallel' in parallel
meillo@114: is ignored.
meillo@217: It is simply treated as `multipart/mixed'
meillo@217: .Ci d0581ba306a7299113a346f9b4c46ce97bc4cef6 .
meillo@217: This was already possible to requested with the, now removed,
meillo@114: .Sw -serialonly
meillo@114: switch of
meillo@114: .Pn mhshow .
meillo@179: As MIME parts are always processed exclusively, i.e. serially,
meillo@217: the `\fL%e\fP' escape in
meillo@114: .Pe mhshow-show-*
meillo@217: profile entries became useless and was thus removed
meillo@217: .Ci a20d405db09b7ccca74d3e8c57550883da49e1ae .
meillo@114: .P
meillo@114: Other kinds of attachments are ignored.
meillo@114: With
meillo@114: .Pe mhshow-show-*
meillo@114: profile entries for them, they can be displayed serially along
meillo@114: the message.
meillo@114: For parallel display, the attachments need to be stored to disk first.
meillo@114: .P
meillo@114: To display text content in foreign charsets, they need to be converted
meillo@114: to the native charset.
meillo@114: Therefore,
meillo@114: .Pe mhshow-charset-*
meillo@217: profile entries were needed.
meillo@169: In mmh, the conversion is performed automatically by piping the
meillo@169: text through the
meillo@114: .Pn iconv
meillo@217: command, if necessary
meillo@217: .Ci 2433122c20baccb10b70b49c04c6b0497b5b3b60 .
meillo@114: Custom
meillo@114: .Pe mhshow-show-*
meillo@114: rules for textual content might need a
meillo@114: .Cl "iconv -f %c %f |
meillo@114: prefix to have the text converted to the native charset.
meillo@114: .P
meillo@121: Although the conversion of foreign charsets to the native one
meillo@114: has improved, it is not consistent enough.
meillo@114: Further work needs to be done and
meillo@114: the basic concepts in this field need to be re-thought.
meillo@114: Though, the default setup of mmh displays message in foreign charsets
meillo@114: correctly without the need to configure anything.
meillo@114: 
meillo@114: 
meillo@114: .ig
meillo@114: 
meillo@114: .P
meillo@114: mhshow/mhstore: Removed support for retrieving message/external-body parts.
meillo@173: These tools will not download the contents automatically anymore. Instead,
meillo@114: they print the information needed to get the contents. If someone should
meillo@114: really receive one of those rare message/external-body messages, he can
meillo@114: do the job manually. We save nearly a thousand lines of code. That's worth
meillo@114: it!
meillo@114: (The profile entry `nmh-access-ftp' and sbr/ruserpass.c for reading
meillo@114: ~/.netrc are gone now.)
meillo@114: .Ci 55e1d8c654ee0f7c45b9361ce34617983b454c32
meillo@114: 
meillo@114: ..
meillo@102: 
meillo@58: 
meillo@58: 
meillo@166: .H2 "Signing and Encrypting
meillo@22: .P
meillo@166: Nmh offers no direct support for digital signatures and message encryption.
meillo@157: This functionality needed to be added through third-party software.
meillo@217: In mmh, the functionality is included because it
meillo@217: is a part of modern email and is likely wanted by users of mmh.
meillo@217: A fresh mmh installation supports signing and encrypting
meillo@157: out-of-the-box.
meillo@157: Therefore, Neil Rickert's
meillo@157: .Pn mhsign
meillo@157: and
meillo@157: .Pn mhpgp
meillo@157: scripts
meillo@157: .[
meillo@157: neil rickert mhsign mhpgp
meillo@157: .]
meillo@217: were included
meillo@177: .Ci f45cdc98117a84f071759462c7ae212f4bc5ab2e
meillo@177: .Ci 58cf09aa36e9f7f352a127158bbf1c5678bc6ed8 .
meillo@177: The scripts fit well because they are lightweight and
meillo@177: similar of style to the existing tools.
meillo@217: Additionally, no licensing difficulties appeared
meillo@157: as they are part of the public domain.
meillo@112: .P
meillo@157: .Pn mhsign
meillo@157: handles the signing and encrypting part.
meillo@157: It comprises about 250 lines of shell code and interfaces between
meillo@223: .Pn gnupg
meillo@217: and the MH system.
meillo@177: It was meant to be invoked manually at the WhatNow prompt, but in mmh,
meillo@157: .Pn send
meillo@177: invokes
meillo@217: .Pn mhsign
meillo@177: automatically
meillo@177: .Ci c7b5e1df086bcc37ff40163ee67571f076cf6683 .
meillo@177: Special header fields were introduced to request this action.
meillo@157: If a draft contains the
meillo@157: .Hd Sign
meillo@157: header field,
meillo@157: .Pn send
meillo@177: will initiate the signing.
meillo@217: The signing key is either chosen automatically or it is specified by the
meillo@157: .Pe Pgpkey
meillo@157: profile entry.
meillo@157: .Pn send
meillo@217: always create signatures using the PGP/MIME standard [RFC\|4880],
meillo@217: but by invoking
meillo@217: .Pn mhsign
meillo@217: manually, old-style non-MIME signatures can be created as well.
meillo@177: To encrypt an outgoing message, the draft needs to contain an
meillo@157: .Hd Enc
meillo@157: header field.
meillo@177: Public keys of all recipients are searched for in the gnupg keyring and
meillo@177: in a file called
meillo@177: .Fn pgpkeys ,
meillo@177: which contains exceptions and overrides.
meillo@157: Unless public keys are found for all recipients,
meillo@177: .Pn mhsign
meillo@177: will refuse to encrypt it.
meillo@157: Currently, messages with hidden (BCC) recipients can not be encrypted.
meillo@171: This work is pending because it requires a structurally more complex
meillo@171: approach.
meillo@157: .P
meillo@177: .Pn mhpgp
meillo@177: is the companion to
meillo@177: .Pn mhsign .
meillo@177: It verifies signatures and decrypts messages.
meillo@217: Encrypted messages can be either temporarily decrypted and displayed
meillo@177: or permanently decrypted and stored into the current folder.
meillo@177: Currently,
meillo@177: .Pn mhpgp
meillo@177: needs to be invoked manually.
meillo@177: The integration into
meillo@177: .Pn show
meillo@177: and
meillo@177: .Pn mhstore
meillo@217: to verify signatures and decrypt messages as needed
meillo@217: is planned but not yet realized.
meillo@177: .P
meillo@217: Both scripts were written for nmh.
meillo@217: Hence they needed to be adjust
meillo@177: according to the differences between nmh and mmh.
meillo@177: For instance, they use the backup prefix no longer.
meillo@181: Furthermore, compatibility support for old PGP features was dropped.
meillo@177: .P
meillo@157: The integrated message signing and encrypting support is one of the
meillo@157: most recent features in mmh.
meillo@217: It has not had the time to mature.
meillo@177: User feedback and personal experience need to be accumulated to
meillo@177: direct the further development of the facility.
meillo@217: Already it seems to be worthwhile to consider adding
meillo@157: .Sw -[no]sign
meillo@157: and
meillo@157: .Sw -[no]enc
meillo@157: switches to
meillo@157: .Pn send ,
meillo@177: to be able to override the corresponding header fields.
meillo@177: A profile entry:
meillo@157: .VS
meillo@157: send: -sign
meillo@157: VE
meillo@177: would then activate signing for all outgoing messages.
meillo@177: With the present approach, a
meillo@177: .Hd Send
meillo@177: header component needs to be added to each draft template
meillo@177: to achieve the same result.
meillo@177: Adding the switches would ease the work greatly and keep the
meillo@177: template files clean.
meillo@157: 
meillo@58: 
meillo@58: 
meillo@102: 
meillo@133: .H2 "Draft and Trash Folder
meillo@131: .P
meillo@58: 
meillo@131: .U3 "Draft Folder
meillo@154: .Id draft-folder
meillo@131: .P
meillo@131: In the beginning, MH had the concept of a draft message.
meillo@217: This was a file named
meillo@131: .Fn draft
meillo@217: in the MH directory, which was treated special.
meillo@131: On composing a message, this draft file was used.
meillo@131: When starting to compose another message before the former one was sent,
meillo@131: the user had to decide among:
meillo@171: .LI 1
meillo@168: Using the old draft to finish and send it before starting with a new one.
meillo@171: .LI 2
meillo@168: Discarding the old draft and replacing it with a new one.
meillo@171: .LI 3
meillo@168: Preserving the old draft by refiling it to a folder.
meillo@171: .LP
meillo@217: Working on multiple drafts was only possible in alternation.
meillo@217: For that, the current draft needed to be refiled to a folder and
meillo@168: another one re-used for editing.
meillo@131: Working on multiple drafts at the same time was impossible.
meillo@131: The usual approach of switching to a different MH context did not
meillo@168: help anything.
meillo@131: .P
meillo@131: The draft folder facility exists to
meillo@131: allow true parallel editing of drafts, in a straight forward way.
meillo@131: It was introduced by Marshall T. Rose, already in 1984.
meillo@131: Similar to other new features, the draft folder was inactive by default.
meillo@131: Even in nmh, the highly useful draft folder was not available
meillo@131: out-of-the-box.
meillo@131: At least, Richard Coleman added the man page
meillo@131: .Mp mh-draft (5)
meillo@131: to better document the feature.
meillo@131: .P
meillo@131: Not using the draft folder facility has the single advantage of having
meillo@131: the draft file at a static location.
meillo@131: This is simple in simple cases but the concept does not scale for more
meillo@131: complex cases.
meillo@217: The concept of the draft message is too limited for the problem
meillo@217: it tries to solve.
meillo@131: Therefore the draft folder was introduced.
meillo@131: It is the more powerful and more natural concept.
meillo@131: The draft folder is a folder like any other folder in MH.
meillo@131: Its messages can be listed like any other messages.
meillo@131: A draft message is no longer a special case.
meillo@131: Tools do not need special switches to work on the draft message.
meillo@171: Hence corner cases were removed.
meillo@131: .P
meillo@131: The trivial part of the work was activating the draft folder with a
meillo@131: default name.
meillo@131: I chose the name
meillo@217: .Fn +drafts ,
meillo@131: for obvious reasons.
meillo@131: In consequence, the command line switches
meillo@131: .Sw -draftfolder
meillo@131: and
meillo@131: .Sw -draftmessage
meillo@131: could be removed.
meillo@131: More difficult but also more improving was updating the tools to the
meillo@131: new concept.
meillo@131: For nearly three decades, the tools needed to support two draft handling
meillo@131: approaches.
meillo@217: By fully switching to the draft folder, the tools could be
meillo@217: simplified by dropping the awkward draft message handling code.
meillo@131: .Sw -draft
meillo@131: switches were removed because operating on a draft message is no longer
meillo@131: special.
meillo@131: It became indistinguishable to operating on any other message.
meillo@168: .Ci 337338b404931f06f0db2119c9e145e8ca5a9860
meillo@168: .P
meillo@168: There is no more need to query the user for draft handling
meillo@168: .Ci 2d48b455c303a807041c35e4248955f8bec59eeb .
meillo@131: It is always possible to add another new draft.
meillo@131: Refiling drafts is without difference to refiling other messages.
meillo@168: All of these special cases are gone.
meillo@131: Yet, one draft-related switch remained.
meillo@131: .Pn comp
meillo@131: still has
meillo@131: .Sw -[no]use
meillo@131: for switching between two modes:
meillo@171: .LI 1
meillo@217: Modifying an existing draft, with
meillo@217: .Sw -use .
meillo@171: .LI 2
meillo@217: Composing a new draft, possibly taking some existing message as template,
meillo@217: with
meillo@217: .Sw -nouse ,
meillo@217: the default.
meillo@171: .LP
meillo@131: In either case, the behavior of
meillo@131: .Pn comp
meillo@131: is deterministic.
meillo@131: .P
meillo@131: .Pn send
meillo@131: now operates on the current message in the draft folder by default.
meillo@131: As message and folder can both be overridden by specifying them on
meillo@131: the command line, it is possible to send any message in the mail storage
meillo@131: by simply specifying its number and folder.
meillo@131: In contrast to the other tools,
meillo@131: .Pn send
meillo@131: takes the draft folder as its default folder.
meillo@131: .P
meillo@131: Dropping the draft message concept in favor for the draft folder concept,
meillo@217: replaced special cases with regular cases.
meillo@131: This simplified the source code of the tools, as well as the concepts.
meillo@131: In mmh, draft management does not break with the MH concepts
meillo@131: but applies them.
meillo@133: .Cl "scan +drafts" ,
meillo@133: for instance, is a truly natural request.
meillo@217: .P
meillo@169: Most of the work was already performed by Rose in the eighties.
meillo@133: The original improvement of mmh is dropping the old draft message approach
meillo@217: and thus simplifying the tools, the documentation,
meillo@217: and the system as a whole.
meillo@131: Although my part in the draft handling improvement was small,
meillo@217: it was important.
meillo@131: 
meillo@131: 
meillo@131: .U3 "Trash Folder
meillo@154: .Id trash-folder
meillo@131: .P
meillo@131: Similar to the situation for drafts is the situation for removed messages.
meillo@131: Historically, a message was ``deleted'' by prepending a specific
meillo@173: \fIbackup prefix\fP, usually the comma character,
meillo@173: to the file name.
meillo@164: The specific file would then be ignored by MH because only files with
meillo@164: names consisting of digits only are treated as messages.
meillo@131: Although files remained in the file system,
meillo@168: the messages were no longer visible in MH.
meillo@168: To truly delete them, a maintenance job was needed.
meillo@168: Usually a cron job was installed to delete them after a grace time.
meillo@131: For instance:
meillo@131: .VS
meillo@131: find $HOME/Mail -type f -name ',*' -ctime +7 -delete
meillo@131: VE
meillo@168: In such a setup, the original message could be restored
meillo@131: within the grace time interval by stripping the
meillo@217: backup prefix from the file name \(en usually but not always.
meillo@168: If the last message of a folder with six messages (\fL1-6\fP) was removed,
meillo@131: message
meillo@131: .Fn 6 ,
meillo@168: became file
meillo@131: .Fn ,6 .
meillo@168: If then a new message entered the same folder, it would be named with
meillo@168: the number one above the highest existing message number.
meillo@168: In this case the message would be named
meillo@217: .Fn 6 ,
meillo@217: reusing the number.
meillo@168: If this new message would be removed as well,
meillo@217: then the backup of the former message becomes overwritten.
meillo@168: Hence, the ability to restore removed messages did not only depend on
meillo@181: the sweeping cron job but also on the removing of further messages.
meillo@131: It is undesirable to have such obscure and complex mechanisms.
meillo@168: The user should be given a small set of clear assertions, such as
meillo@131: ``Removed files are restorable within a seven-day grace time.''
meillo@131: With the addition ``... unless a message with the same name in the
meillo@131: same folder is removed before.'' the statement becomes complex.
meillo@217: A user will hardly be able to keep track of all removals to know
meillo@131: if the assertion still holds true for a specific file.
meillo@164: In practice, the real mechanism is unclear to the user.
meillo@131: .P
meillo@217: Furthermore, the backup files were scattered within the whole mail storage.
meillo@217: This complicated managing them.
meillo@217: It was possible with the help of
meillo@131: .Pn find ,
meillo@217: but everything is more convenient
meillo@217: if the deleted messages are collected in one place.
meillo@131: .P
meillo@131: The profile entry
meillo@131: .Pe rmmproc
meillo@131: (previously named
meillo@131: .Pe Delete-Prog )
meillo@131: was introduced very early to improve the situation.
meillo@164: It could be set to any command, which would be executed to remove
meillo@131: the specified messages.
meillo@217: This had overridden the default action, described above.
meillo@217: Refiling the to-be-removed files to a trash folder was the usual example.
meillo@131: Nmh's man page
meillo@131: .Mp rmm (1)
meillo@131: proposes to set the
meillo@131: .Pe rmmproc
meillo@131: to
meillo@131: .Cl "refile +d
meillo@217: to move messages to the trash folder
meillo@217: .Fn +d
meillo@131: instead of renaming them with the backup prefix.
meillo@217: The man page additionally proposes the expunge command
meillo@131: .Cl "rm `mhpath +d all`
meillo@168: to empty the trash folder.
meillo@131: .P
meillo@217: Removing messages in such a way has advantages:
meillo@217: .LI 1
meillo@131: The mail storage is prevented from being cluttered with removed messages
meillo@131: because they are all collected in one place.
meillo@131: Existing and removed messages are thus separated more strictly.
meillo@217: .LI 2
meillo@131: No backup files are silently overwritten.
meillo@217: .LI 3
meillo@217: Most important, however, removed messages are kept in the MH domain.
meillo@131: Messages in the trash folder can be listed like those in any other folder.
meillo@131: Deleted messages can be displayed like any other messages.
meillo@169: .Pn refile
meillo@169: can restore deleted messages.
meillo@131: All operations on deleted files are still covered by the MH tools.
meillo@131: The trash folder is just like any other folder in the mail storage.
meillo@131: .P
meillo@131: Similar to the draft folder case, I dropped the old backup prefix approach
meillo@131: in favor for replacing it by the better suiting trash folder system.
meillo@131: Hence,
meillo@131: .Pn rmm
meillo@131: calls
meillo@131: .Pn refile
meillo@131: to move the to-be-removed message to the trash folder,
meillo@131: .Fn +trash
meillo@131: by default.
meillo@164: To sweep it clean, the user can use
meillo@131: .Cl "rmm -unlink +trash a" ,
meillo@131: where the
meillo@131: .Sw -unlink
meillo@131: switch causes the files to be unlinked.
meillo@168: .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
meillo@168: .Ci ca0b3e830b86700d9e5e31b1784de2bdcaf58fc5
meillo@131: .P
meillo@217: Dropping the legacy approach and converting to the new approach
meillo@217: completely, simplified the code base.
meillo@131: The relationship between
meillo@131: .Pn rmm
meillo@131: and
meillo@131: .Pn refile
meillo@131: was inverted.
meillo@131: In mmh,
meillo@131: .Pn rmm
meillo@131: invokes
meillo@217: .Pn refile .
meillo@217: That used to be the other way round.
meillo@131: Yet, the relationship is simpler now.
meillo@164: Loops, like described in nmh's man page for
meillo@131: .Mp refile (1),
meillo@164: can no longer occur:
meillo@131: .QS
meillo@131: Since
meillo@131: .Pn refile
meillo@131: uses your
meillo@131: .Pe rmmproc
meillo@131: to delete the message, the
meillo@131: .Pe rmmproc
meillo@131: must NOT call
meillo@131: .Pn refile
meillo@131: without specifying
meillo@131: .Sw -normmproc
meillo@131: or you will create an infinite loop.
meillo@131: .QE
meillo@131: .LP
meillo@131: .Pn rmm
meillo@131: either unlinks a message with
meillo@131: .Fu unlink()
meillo@131: or invokes
meillo@131: .Pn refile
meillo@131: to move it to the trash folder.
meillo@131: .Pn refile
meillo@131: does not invoke any tools.
meillo@131: .P
meillo@136: By generalizing the message removal in the way that it became covered
meillo@136: by the MH concepts made the whole system more powerful.
meillo@131: 
meillo@131: 
meillo@131: 
meillo@131: 
meillo@131: 
meillo@133: .H2 "Modern Defaults
meillo@133: .P
meillo@133: Nmh has a bunch of convenience-improving features inactive by default,
meillo@217: although one can expect every new user to want them active.
meillo@133: The reason they are inactive by default is the wish to stay compatible
meillo@133: with old versions.
meillo@217: But what are old versions?
meillo@136: Still, the highly useful draft folder facility has not been activated
meillo@136: by default although it was introduced over twenty-five years ago.
meillo@133: .[
meillo@133: rose romine real work
meillo@133: .]
meillo@136: The community seems not to care.
meillo@217: .P
meillo@217: In nmh, new users are required to first build up
meillo@217: a profile before they can access the modern features.
meillo@136: Without an extensive profile, the setup is hardly usable
meillo@133: for modern emailing.
meillo@133: The point is not the customization of the setup,
meillo@136: but the need to activate generally useful facilities.
meillo@217: Yet, the real problem lies less in enabling the features,
meillo@217: as this is straight forward as soon as one knows what he wants.
meillo@168: The real problem is that new users need deep insight into the project
meillo@217: to discover the available but inactive features.
meillo@133: To give an example, I needed one year of using nmh
meillo@133: before I became aware of the existence of the attachment system.
meillo@133: One could argue that this fact disqualifies my reading of the
meillo@133: documentation.
meillo@133: If I would have installed nmh from source back then, I could agree.
meillo@217: Yet, I had used a pre-packaged version and had expected that it would
meillo@133: just work.
meillo@133: Nevertheless, I had been convinced by the concepts of MH already
meillo@133: and I am a software developer,
meillo@133: still I required a lot of time to discover the cool features.
meillo@133: How can we expect users to be even more advanced than me,
meillo@217: just to enable them to use MH in a convenient and modern way?
meillo@133: Unless they are strongly convinced of the concepts, they will fail.
meillo@133: I have seen friends of me giving up disappointed
meillo@133: before they truly used the system,
meillo@133: although they had been motivated in the beginning.
meillo@217: New users suffer hard enough to get used to the tool chest approach,
meillo@179: we developers should spare them further inconveniences.
meillo@133: .P
meillo@136: Maintaining compatibility for its own sake is bad,
meillo@217: because the code base will collect more and more compatibility code.
meillo@200: Sticking to the compatibility code means remaining limited;
meillo@168: whereas adjusting to the changes renders the compatibility unnecessary.
meillo@217: Keeping unused alternatives in the code for longer than a short
meillo@217: grace time is a bad choice as they likely
meillo@217: gather bugs by not being constantly tested.
meillo@136: Also, the increased code size and the greater number of conditions
meillo@136: increase the maintenance costs.
meillo@133: If any MH implementation would be the back-end of widespread
meillo@133: email clients with large user bases, compatibility would be more
meillo@133: important.
meillo@133: Yet, it appears as if this is not the case.
meillo@133: Hence, compatibility is hardly important for technical reasons.
meillo@217: Its importance originates from personal reasons rather.
meillo@133: Nmh's user base is small and old.
meillo@217: Changing the interfaces causes inconvenience to long-term users of MH.
meillo@217: It forces them to change their many years old MH configurations.
meillo@168: I do understand this aspect, but by sticking to the old users,
meillo@217: new users are kept from entering the world of MH.
meillo@217: But the future lies in new users.
meillo@168: In consequence, mmh invites new users by providing a convenient
meillo@168: and modern setup, readily usable out-of-the-box.
meillo@133: .P
meillo@136: In mmh, all modern features are active by default and many previous
meillo@217: approaches are removed or only accessible in a manual way.
meillo@136: New default features include:
meillo@133: .BU
meillo@133: The attachment system (\c
meillo@217: .Hd Attach )
meillo@217: .Ci 8ff284ff9167eff8f5349481529332d59ed913b1 .
meillo@133: .BU
meillo@133: The draft folder facility (\c
meillo@217: .Fn +drafts )
meillo@217: .Ci 337338b404931f06f0db2119c9e145e8ca5a9860 .
meillo@133: .BU
meillo@133: The unseen sequence (`u')
meillo@133: .Ci c2360569e1d8d3678e294eb7c1354cb8bf7501c1
meillo@217: and the sequence negation prefix (`!')
meillo@217: .Ci db74c2bd004b2dc9bf8086a6d8bf773ac051f3cc .
meillo@133: .BU
meillo@217: Quoting the original message in the reply
meillo@217: .Ci 67411b1f95d6ec987b4c732459e1ba8a8ac192c6 .
meillo@133: .BU
meillo@217: Forwarding messages using MIME
meillo@217: .Ci 6e271608b7b9c23771523f88d23a4d3593010cf1 .
meillo@171: .LP
meillo@217: An mmh setup with a profile that defines only the path to the
meillo@136: mail storage, is already convenient to use.
meillo@217: Again, Paul Vixie's supports the direction I took:
meillo@136: ``the `main branch' should just be modern''.
meillo@136: .[
meillo@136: paul vixie edginess nmh-workers
meillo@136: .]
meillo@131: 
meillo@133: 
meillo@133: 
meillo@133: 
meillo@133: 
meillo@133: .\" --------------------------------------------------------------
meillo@131: .H1 "Styling
meillo@22: .P
meillo@118: Kernighan and Pike have emphasized the importance of style in the
meillo@219: preface of \fPThe Practice of Programming\fP:
meillo@118: .[ [
meillo@118: kernighan pike practice of programming
meillo@118: .], p. x]
meillo@118: .QS
meillo@118: Chapter 1 discusses programming style.
meillo@219: Good style is so important to good programming that we have chosen
meillo@118: to cover it first.
meillo@118: .QE
meillo@168: This section covers changes in mmh that were guided by the desire
meillo@118: to improve on style.
meillo@219: Many of them follow the advice given in the quoted book.
meillo@118: 
meillo@118: 
meillo@127: 
meillo@127: 
meillo@127: .H2 "Code Style
meillo@154: .Id code-style
meillo@118: .P
meillo@118: .U3 "Indentation Style
meillo@118: .P
meillo@219: Indentation styles are the holy cow of programming.
meillo@219: Kernighan and Pike write:
meillo@118: .[ [
meillo@118: kernighan pike practice of programming
meillo@118: .], p. 10]
meillo@118: .QS
meillo@118: Programmers have always argued about the layout of programs,
meillo@118: but the specific style is much less important than its consistent
meillo@118: application.
meillo@121: Pick one style, preferably ours, use it consistently, and don't waste
meillo@118: time arguing.
meillo@118: .QE
meillo@118: .P
meillo@118: I agree that the constant application is most important,
meillo@118: but I believe that some styles have advantages over others.
meillo@118: For instance the indentation with tab characters only.
meillo@219: The number of tabs corresponds to the nesting level \(en
meillo@118: one tab, one level.
meillo@219: Tab characters provide flexible visual appearance because developers
meillo@219: can adjust their width as prefered.
meillo@219: There is no more need to check for the correct mixture of
meillo@219: tabs and spaces.
meillo@219: Two simple rules ensure the integrity and flexibility of the visual
meillo@219: appearance:
meillo@219: .LI 1
meillo@219: Leading whitespace must consist of tabs only.
meillo@219: .LI 2
meillo@219: All other whitespace should be spaces.
meillo@219: .LP
meillo@121: Although reformatting existing code should be avoided, I did it.
meillo@200: I did not waste time arguing; I just reformatted the code.
meillo@118: .Ci a485ed478abbd599d8c9aab48934e7a26733ecb1
meillo@118: 
meillo@118: .U3 "Comments
meillo@118: .P
meillo@219: Kernighan and Pike demand: ``Don't belabor the obvious''.
meillo@118: .[ [
meillo@118: kernighan pike practice of programming
meillo@118: .], p. 23]
meillo@219: Following the advice, I removed unnecessary comments.
meillo@219: For instance, I removed all comments in the following code excerpt
meillo@219: .Ci 426543622b377fc5d091455cba685e114b6df674 :
meillo@118: .VS
meillo@120: context_replace(curfolder, folder);  /* update current folder  */
meillo@120: seq_setcur(mp, mp->lowsel);  /* update current message */
meillo@120: seq_save(mp);  /* synchronize message sequences */
meillo@120: folder_free(mp);  /* free folder/message structure */
meillo@120: context_save();  /* save the context file */
meillo@120: 
meillo@120: [...]
meillo@120: 
meillo@120: int c;  /* current character */
meillo@120: char *cp;  /* miscellaneous character pointer */
meillo@120: 
meillo@120: [...]
meillo@120: 
meillo@120: /* NUL-terminate the field */
meillo@120: *cp = '\0';
meillo@118: VE
meillo@118: .P
meillo@219: The information in each of the comments was present in the code
meillo@219: statements already, except for the NUL-termination, which became
meillo@219: obvious from the context.
meillo@136: 
meillo@118: 
meillo@118: .U3 "Names
meillo@118: .P
meillo@219: Regarding this topic, Kernighan and Pike suggest:
meillo@118: ``Use active names for functions''.
meillo@118: .[ [
meillo@118: kernighan pike practice of programming
meillo@118: .], p. 4]
meillo@118: One application of this rule was the rename of
meillo@118: .Fu check_charset()
meillo@118: to
meillo@219: .Fu is_native_charset()
meillo@219: .Ci 8d77b48284c58c135a6b2787e721597346ab056d .
meillo@219: The same change additionally fixed a violation of ``Be accurate'',
meillo@181: .[ [
meillo@181: kernighan pike practice of programming
meillo@181: .], p. 4]
meillo@219: as the code did not match the expectation the function suggested.
meillo@219: It did not compare charset names but prefixes of them only.
meillo@219: In case the native charset was `ISO-8859-1', then
meillo@219: .VS
meillo@219: check_charset("ISO-8859-11", strlen("ISO-8859-11"))
meillo@219: VE
meillo@219: had returned true although the upper halves of the code pages
meillo@219: are different.
meillo@118: .P
meillo@118: More important than using active names is using descriptive names.
meillo@145: .VS
meillo@145: m_unknown(in);  /* the MAGIC invocation... */
meillo@145: VE
meillo@145: Renaming the obscure
meillo@118: .Fu m_unknown()
meillo@219: function was a delightful event, although it made the code less funny
meillo@219: .Ci 611d68d19204d7cbf5bd585391249cb5bafca846 .
meillo@118: .P
meillo@118: Magic numbers are generally considered bad style.
meillo@118: Obviously, Kernighan and Pike agree:
meillo@118: ``Give names to magic numbers''.
meillo@118: .[ [
meillo@118: kernighan pike practice of programming
meillo@118: .], p. 19]
meillo@219: .P
meillo@219: The argument
meillo@219: .CW outnum
meillo@219: of the function
meillo@219: .Fu scan()
meillo@219: in
meillo@219: .Fn uip/scansbr.c
meillo@219: holds the number of the message to be created.
meillo@219: As well it encodes program logic with negative numbers and zero.
meillo@219: This led to obscure code.
meillo@219: I clarified the code by introducing two variables that extracted
meillo@219: the hidden information:
meillo@219: .VS
meillo@219: int incing = (outnum > 0);
meillo@219: int ismbox = (outnum != 0);
meillo@219: VE
meillo@219: The readable names are thus used in conditions;
meillo@219: the variable
meillo@219: .CW outnum
meillo@219: is used only to extract ordinary message numbers
meillo@219: .Ci b8b075c77be7794f3ae9ff0e8cedb12b48fd139f .
meillo@219: .P
meillo@219: Through the clarity improvement of the change detours in the program
meillo@219: logic of related code parts became apparent.
meillo@219: The implementation was simplified.
meillo@219: This possibility to improve had been invisible before
meillo@219: .Ci aa60b0ab5e804f8befa890c0a6df0e3143ce0723 .
meillo@219: .P
meillo@219: The names just described were a first step, yet the situation
meillo@219: was further improved by giving names to the magic values of
meillo@219: .CW outnum :
meillo@118: .VS
meillo@118: #define SCN_MBOX (-1)
meillo@118: #define SCN_FOLD 0
meillo@118: VE
meillo@219: The two variables were updated thereafter as well:
meillo@219: .VS
meillo@219: int incing = (outnum != SCN_MBOX && outnum != SCN_FOLD);
meillo@219: int scanfolder = (outnum == SCN_FOLD);
meillo@219: VE
meillo@219: Furthermore,
meillo@219: .CW ismbox
meillo@219: was replaced by
meillo@219: .CW scanfolder
meillo@219: because that matched better to the program logic.
meillo@118: .Ci 7ffb36d28e517a6f3a10272056fc127592ab1c19
meillo@219: 
meillo@118: 
meillo@133: 
meillo@133: 
meillo@133: .H2 "Structural Rework
meillo@133: .P
meillo@219: Although the stylistic changes described already improve the
meillo@219: readability of the source code, all of them were changes ``in the small''.
meillo@219: Structural changes, in contrast, affect much larger code areas.
meillo@219: They are more difficult to accomplish but lead to larger improvements,
meillo@219: especially as they often influence the outer shape of the tools as well.
meillo@118: .P
meillo@118: At the end of their chapter on style,
meillo@118: Kernighan and Pike ask: ``But why worry about style?''
meillo@181: .[ [
meillo@181: kernighan pike practice of programming
meillo@219: .], p. 28].
meillo@219: Following are two examples of structural rework that demonstrate
meillo@136: why style is important in the first place.
meillo@136: 
meillo@136: 
meillo@136: .U3 "Rework of \f(CWanno\fP
meillo@118: .P
meillo@120: Until 2002,
meillo@120: .Pn anno
meillo@219: had six functional command line switches:
meillo@120: .Sw -component
meillo@120: and
meillo@120: .Sw -text ,
meillo@219: each with an argument,
meillo@120: and the two pairs of flags,
meillo@120: .Sw -[no]date
meillo@120: and
meillo@120: .Sw -[no]inplace .
meillo@120: Then Jon Steinhart introduced his attachment system.
meillo@120: In need for more advanced annotation handling, he extended
meillo@120: .Pn anno .
meillo@120: He added five more switches:
meillo@120: .Sw -draft ,
meillo@120: .Sw -list ,
meillo@120: .Sw -delete ,
meillo@120: .Sw -append ,
meillo@120: and
meillo@120: .Sw -number ,
meillo@219: the last one taking an argument
meillo@219: .Ci 7480dbc14bc90f2d872d434205c0784704213252 .
meillo@120: Later,
meillo@120: .Sw -[no]preserve
meillo@219: was added as well
meillo@219: .Ci d9b1d57351d104d7ec1a5621f090657dcce8cb7f .
meillo@120: Then, the Synopsis section of the man page
meillo@120: .Mp anno (1)
meillo@120: read:
meillo@120: .VS
meillo@219: anno [+folder] [msgs] [-component f(CIfieldfP] [-inplace | -noinplace]
meillo@120: 	[-date | -nodate] [-draft] [-append] [-list] [-delete]
meillo@219: 	[-number [f(CInumfP|fPallfP]] [-preserve | -nopreserve] [-version]
meillo@219: 	[-help] [-text f(CIbodyfP]
meillo@120: VE
meillo@120: .LP
meillo@120: The implementation followed the same structure.
meillo@120: Problems became visible when
meillo@120: .Cl "anno -list -number 42
meillo@219: worked on the current message instead of on message number 42,
meillo@120: and
meillo@120: .Cl "anno -list -number l:5
meillo@124: did not work on the last five messages but failed with the mysterious
meillo@120: error message: ``anno: missing argument to -list''.
meillo@121: Yet, the invocation matched the specification in the man page.
meillo@120: There, the correct use of
meillo@120: .Sw -number
meillo@120: was defined as being
meillo@120: .Cl "[-number [num|all]]
meillo@120: and the textual description for the combination with
meillo@120: .Sw -list
meillo@120: read:
meillo@120: .QS
meillo@164: The
meillo@164: .Sw -list
meillo@164: option produces a listing of the field bodies for
meillo@120: header fields with names matching the specified component,
meillo@164: one per line. The listing is numbered, starting at 1, if the
meillo@164: .Sw -number
meillo@164: option is also used.
meillo@120: .QE
meillo@120: .LP
meillo@120: The problem was manifold.
meillo@120: Semantically, the argument to the
meillo@120: .Sw -number
meillo@120: switch is only necessary in combination with
meillo@120: .Sw -delete ,
meillo@120: but not with
meillo@120: .Sw -list .
meillo@219: The code, however, required a numeric argument in any case.
meillo@219: If the argument was missing or non-numeric,
meillo@219: .Pn anno
meillo@219: aborted with an error message that additionally had an off-by-one error.
meillo@219: It printed the name of the switch one before the concerned one.
meillo@120: .P
meillo@219: Trying to fix these problems on the surface would not have solved them.
meillo@219: They originate from a discrepance between the
meillo@120: structure of the problem and the structure implemented in the program.
meillo@219: Such structural differences can only be solved by adjusting the
meillo@219: structure of the implementation to the structure of the problem.
meillo@120: .P
meillo@219: Steinhart had added the new
meillo@120: .Sw -list
meillo@120: and
meillo@120: .Sw -delete
meillo@219: switches in a style similar to the other switches though
meillo@219: they are of structural different type.
meillo@120: Semantically,
meillo@120: .Sw -list
meillo@120: and
meillo@120: .Sw -delete
meillo@219: introduce operation modes.
meillo@120: Historically,
meillo@120: .Pn anno
meillo@120: had only one operation mode: adding header fields.
meillo@219: With the extension, two more modes were added:
meillo@120: listing and deleting header fields.
meillo@120: The structure of the code changes did not pay respect to this
meillo@219: fundamental change.
meillo@120: Neither the implementation nor the documentation did clearly
meillo@219: declare the exclusive operation modes as such.
meillo@120: Having identified the problem, I solved it by putting structure into
meillo@120: .Pn anno
meillo@219: and its documentation
meillo@219: .Ci d54c8db8bdf01e8381890f7729bc0ef4a055ea11 .
meillo@120: .P
meillo@173: The difference is visible in both the code and the documentation.
meillo@219: For instance in the following code excerpt:
meillo@120: .VS
meillo@120: int delete = -2;  /* delete header element if set */
meillo@120: int list = 0;  /* list header elements if set */
meillo@120: [...]
meillo@121: 	case DELETESW:  /* delete annotations */
meillo@121: 		delete = 0;
meillo@121: 		continue;
meillo@121: 	case LISTSW:  /* produce a listing */
meillo@121: 		list = 1;
meillo@121: 		continue;
meillo@120: VE
meillo@121: .LP
meillo@219: which was replaced by:
meillo@120: .VS
meillo@120: static enum { MODE_ADD, MODE_DEL, MODE_LIST } mode = MODE_ADD;
meillo@120: [...]
meillo@121: 	case DELETESW:  /* delete annotations */
meillo@121: 		mode = MODE_DEL;
meillo@121: 		continue;
meillo@121: 	case LISTSW:  /* produce a listing */
meillo@121: 		mode = MODE_LIST;
meillo@121: 		continue;
meillo@120: VE
meillo@120: .LP
meillo@121: The replacement code does not only reflect the problem's structure better,
meillo@121: it is easier to understand as well.
meillo@121: The same applies to the documentation.
meillo@120: The man page was completely reorganized to propagate the same structure.
meillo@219: This is already visible in the Synopsis section:
meillo@120: .VS
meillo@219: anno [+folder] [msgs] [-component f(CIfieldfP] [-text fPbodyfP]
meillo@120: 	[-append] [-date | -nodate] [-preserve | -nopreserve]
meillo@120: 	[-Version] [-help]
meillo@120: 
meillo@219: anno -delete [+folder] [msgs] [-component fPfieldfP] [-text
meillo@219: 	fPbodyfP] [-number fPnum fP| fPall fP] [-preserve | -nopreserve]
meillo@120: 	[-Version] [-help]
meillo@120: 
meillo@219: anno -list [+folder] [msgs] [-component fPfieldfP] [-number]
meillo@120: 	[-Version] [-help]
meillo@120: VE
meillo@118: 
meillo@58: 
meillo@58: 
meillo@133: .U3 "Path Conversion
meillo@133: .P
meillo@134: Four kinds of path names can appear in MH:
meillo@171: .LI 1
meillo@134: Absolute Unix directory paths, like
meillo@134: .Fn /etc/passwd .
meillo@171: .LI 2
meillo@134: Relative Unix directory paths, like
meillo@134: .Fn ./foo/bar .
meillo@171: .LI 3
meillo@134: Absolute MH folder paths, like
meillo@219: .Fn +projects/mmh .
meillo@171: .LI 4
meillo@134: Relative MH folder paths, like
meillo@134: .Fn @subfolder .
meillo@171: .LP
meillo@219: Relative MH folder paths, are hardly documented
meillo@219: although they are useful for large mail storages.
meillo@134: The current mail folder is specified as `\c
meillo@134: .Fn @ ',
meillo@134: just like the current directory is specified as `\c
meillo@134: .Fn . '.
meillo@134: .P
meillo@134: To allow MH tools to understand all four notations,
meillo@219: they need to be able to convert between them.
meillo@134: In nmh, these path name conversion functions were located in the files
meillo@134: .Fn sbr/path.c
meillo@134: (``return a pathname'') and
meillo@134: .Fn sbr/m_maildir.c
meillo@134: (``get the path for the mail directory'').
meillo@134: The seven functions in the two files were documented with no more
meillo@134: than two comments, which described obvious information.
meillo@219: The signatures of the four exported functions did not explain their
meillo@219: semantics:
meillo@219: .LI 1
meillo@219: .CW "char *path(char *, int);
meillo@219: .LI 2
meillo@219: .CW "char *pluspath(char *);
meillo@219: .LI 3
meillo@219: .CW "char *m_mailpath(char *);
meillo@219: .LI 4
meillo@219: .CW "char *m_maildir(char *);
meillo@134: .P
meillo@219: My investigations provided the following descriptions:
meillo@171: .LI 1
meillo@134: The second parameter of
meillo@134: .Fu path()
meillo@219: defines the type as which the path given in the first parameter should
meillo@219: be treated.
meillo@134: Directory paths are converted to absolute directory paths.
meillo@134: Folder paths are converted to absolute folder paths.
meillo@173: Folder paths must not include a leading `\fL@\fP' character.
meillo@134: Leading plus characters are preserved.
meillo@134: The result is a pointer to newly allocated memory.
meillo@171: .LI 2
meillo@134: .Fu pluspath()
meillo@134: is a convenience-wrapper to
meillo@134: .Fu path() ,
meillo@134: to convert folder paths only.
meillo@134: This function can not be used for directory paths.
meillo@134: An empty string parameter causes a buffer overflow.
meillo@171: .LI 3
meillo@134: .Fu m_mailpath()
meillo@134: converts directory paths to absolute directory paths.
meillo@173: The characters `\fL+\fP' or `\fL@\fP' at the beginning of the path name are
meillo@134: treated literal, i.e. as the first character of a relative directory path.
meillo@134: Hence, this function can not be used for folder paths.
meillo@219: In any case, the result is an absolute directory path,
meillo@219: returned as a pointer to newly allocated memory.
meillo@171: .LI 4
meillo@134: .Fu m_maildir()
meillo@134: returns the parameter unchanged if it is an absolute directory path
meillo@173: or begins with the entry `\fL.\fP' or `\fL..\fP'.
meillo@134: All other strings are prepended with the current working directory.
meillo@219: Hence, this function can not be used for folder paths.
meillo@134: The result is either an absolute directory path or a relative
meillo@219: directory path, starting with dot or dot-dot.
meillo@134: In contrast to the other functions, the result is a pointer to
meillo@134: static memory.
meillo@134: .P
meillo@134: The situation was obscure, irritating, error-prone, and non-orthogonal.
meillo@219: Additionally, no clear terminology was used to name the different
meillo@219: kinds of path names.
meillo@219: Sometimes, the names were even misleading, much as the first argument of
meillo@134: .Fu m_mailpath() ,
meillo@219: which was named
meillo@219: .CW folder ,
meillo@219: although
meillo@134: .Fu m_mailpath()
meillo@219: could not be used with MH folder arguments.
meillo@134: .P
meillo@219: I clarified the path name conversion by complete rework.
meillo@134: First of all, the terminology needed to be defined.
meillo@134: A path name is either in the Unix domain, then it is called
meillo@219: \fIdirectory path\fP (\fIdirpath\fP for short) or it is in the MH domain,
meillo@219: then it is called \fIfolder path\fP (\fIfolpath\fP for short).
meillo@134: The two terms need to be used with strict distinction.
meillo@219: Often a clear terminology indicates that the problem is understood.
meillo@134: Second, I exploited the concept of path type indicators.
meillo@219: By requiring every path name to start with a distinct type identifier,
meillo@219: the conversion between the types could be fully automated.
meillo@219: This allows the tools to accept paths of any type from the user.
meillo@134: Therefore, it was necessary to require relative directory paths to be
meillo@134: prefixed with a dot character.
meillo@134: In consequence, the dot character could no longer be an alias for the
meillo@134: current message.
meillo@134: .Ci cff0e16925e7edbd25b8b9d6d4fbdf03e0e60c01
meillo@134: Third, I created three new functions to replace the previous mess:
meillo@171: .LI 1
meillo@134: .Fu expandfol()
meillo@219: converts folder paths to absolute folder paths.
meillo@134: Directory paths are simply passed through.
meillo@134: This function is to be used for folder paths only, thus the name.
meillo@134: The result is a pointer to static memory.
meillo@171: .LI 2
meillo@134: .Fu expanddir()
meillo@134: converts directory paths to absolute directory paths.
meillo@134: Folder paths are treated as relative directory paths.
meillo@134: This function is to be used for directory paths only, thus the name.
meillo@134: The result is a pointer to static memory.
meillo@171: .LI 3
meillo@134: .Fu toabsdir()
meillo@134: converts any type of path to an absolute directory path.
meillo@134: This is the function of choice for path conversion.
meillo@134: Absolute directory paths are the most general representation of a
meillo@134: path name.
meillo@134: The result is a pointer to static memory.
meillo@134: .P
meillo@134: The new functions have names that indicate their use.
meillo@134: Two of the functions convert relative to absolute path names of the
meillo@134: same type.
meillo@134: The third function converts any path name type to the most general one,
meillo@134: the absolute directory path.
meillo@134: All of the functions return pointers to static memory.
meillo@219: The file
meillo@219: .Fn sbr/path.c
meillo@219: contains the implementation of the functions;
meillo@134: .Fn sbr/m_maildir.c
meillo@219: was removed.
meillo@168: .Ci d39e2c447b0d163a5a63f480b23d06edb7a73aa0
meillo@134: .P
meillo@134: Along with the path conversion rework, I also replaced
meillo@134: .Fu getfolder(FDEF)
meillo@134: with
meillo@134: .Fu getdeffol()
meillo@134: and
meillo@134: .Fu getfolder(FCUR)
meillo@134: with
meillo@134: .Fu getcurfol() ,
meillo@219: which only wraps
meillo@219: .Fu expandfol(""@"")
meillo@219: for convenience.
meillo@134: This code was moved from
meillo@134: .Fn sbr/getfolder.c
meillo@219: into
meillo@219: .Fn sbr/path.c
meillo@219: as well.
meillo@168: .Ci d39e2c447b0d163a5a63f480b23d06edb7a73aa0
meillo@134: .P
meillo@134: The related function
meillo@134: .Fu etcpath()
meillo@219: is now included in
meillo@134: .Fn sbr/path.c ,
meillo@168: too
meillo@168: .Ci b4c29794c12099556151d93a860ee51badae2e35 .
meillo@134: Previously, it had been located in
meillo@219: .Fn config/config.c .
meillo@134: .P
meillo@219: Now,
meillo@134: .Fn sbr/path.c
meillo@219: contains all path handling code.
meillo@219: Besides being less code, its readability is highly improved.
meillo@219: The functions follow a common style and are well documented.
meillo@133: 
meillo@133: 
meillo@133: 
meillo@133: 
meillo@133: .H2 "Profile Reading
meillo@133: .P
meillo@220: The MH profile contains the configuration of a user-specific MH setup.
meillo@220: MH tools read the profile right after starting up
meillo@220: because it contains the location of the user's mail storage
meillo@138: and similar settings that influence the whole setup.
meillo@220: Furthermore, the profile contains the default switches for the tools
meillo@220: as well.
meillo@220: The context file is read along with the profile.
meillo@138: .P
meillo@138: For historic reasons, some MH tools did not read the profile and context.
meillo@138: Among them were
meillo@138: .Pn post /\c
meillo@138: .Pn spost ,
meillo@138: .Pn mhmail ,
meillo@138: and
meillo@138: .Pn slocal .
meillo@138: The reason why these tools ignored the profile were not clearly stated.
meillo@220: During a discussion on the nmh-workers mailing list,
meillo@181: David Levine posted an explanation, quoting John Romine:
meillo@138: .[
meillo@138: nmh-workers levine post profile
meillo@138: .]
meillo@220: 
meillo@138: .QS
meillo@138: I asked John Romine and here's what he had to say, which
meillo@138: agrees and provides an example that convinces me:
meillo@138: .QS
meillo@164: My take on this is that
meillo@164: .Pn post
meillo@164: should not be called by users directly, and it doesn't read the
meillo@164: .Fn .mh_profile
meillo@138: (only front-end UI programs read the profile).
meillo@138: .QP
meillo@164: For example, there can be contexts where
meillo@164: .Pn post
meillo@164: is called by a helper program (like `\c
meillo@164: .Pn mhmail ')
meillo@164: which may be run by a non-MH user.
meillo@164: We don't want this to prompt the user to create an MH profile, etc.
meillo@138: .QP
meillo@164: My suggestion would be to have
meillo@164: .Pn send
meillo@164: pass a (hidden) `\c
meillo@164: .Sw -fileproc
meillo@164: .Ar proc '
meillo@164: option to
meillo@164: .Pn post
meillo@164: if needed.
meillo@164: You could also
meillo@164: use an environment variable (I think
meillo@164: .Pn send /\c
meillo@164: .Pn whatnow
meillo@164: do this).
meillo@138: .QE
meillo@220: .sp \n(PDu
meillo@164: I think that's the way to go.
meillo@164: My personal preference is to use a command line option,
meillo@164: not an environment variable.
meillo@138: .QE
meillo@220: 
meillo@138: .P
meillo@220: To solve the problem that
meillo@138: .Pn post
meillo@220: does not honor the
meillo@138: .Pe fileproc
meillo@138: profile entry,
meillo@138: the community roughly agreed that a switch
meillo@138: .Sw -fileproc
meillo@138: should be added to
meillo@138: .Pn post
meillo@138: to be able to pass a different fileproc.
meillo@138: I strongly disagree with this approach because it does not solve
meillo@138: the problem; it only removes a single symptom.
meillo@220: The actual problem is that
meillo@138: .Pn post
meillo@220: does not behave as expected,
meillo@220: though all programs should behave as expected.
meillo@220: Clear and general concepts are a precondition for this.
meillo@220: Thus, there should be no separation into ``front-end UI programs''
meillo@220: and ones that ``should not be called by users directly''.
meillo@220: The real solution is having all MH tools read the profile.
meillo@138: .P
meillo@220: But the problem has a further aspect,
meillo@220: which originates from
meillo@220: .Pn mhmail
meillo@220: mainly.
meillo@138: .Pn mhmail
meillo@138: was intended to be a replacement for
meillo@138: .Pn mailx
meillo@138: on systems with MH installations.
meillo@220: In difference to
meillo@220: .Pn mailx ,
meillo@138: .Pn mhmail
meillo@220: used MH's
meillo@138: .Pn post
meillo@220: to send the message.
meillo@220: The idea was that using
meillo@138: .Pn mhmail
meillo@220: should not be influenced whether the user had
meillo@138: MH set up for himself or not.
meillo@220: Therefore
meillo@138: .Pn mhmail
meillo@220: had not read the profile.
meillo@138: As
meillo@138: .Pn mhmail
meillo@138: used
meillo@138: .Pn post ,
meillo@138: .Pn post
meillo@220: was not allowed to read the profile neither.
meillo@138: This is the reason for the actual problem.
meillo@220: Yet, this was not considered much of a problem because
meillo@138: .Pn post
meillo@138: was not intended to be used by users directly.
meillo@220: To invoke
meillo@220: .Pn post ,
meillo@138: .Pn send
meillo@220: was used an a front-end.
meillo@138: .Pn send
meillo@138: read the profile and passed all relevant values on the command line to
meillo@138: .Pn post
meillo@138: \(en an awkward solution.
meillo@138: .P
meillo@138: The important insight is that
meillo@138: .Pn mhmail
meillo@220: is a wolf in sheep's clothing.
meillo@220: This alien tool broke the concepts because it was treated like
meillo@220: a normal MH tool.
meillo@138: Instead it should have been treated accordingly to its foreign style.
meillo@220: .P
meillo@220: The solution is not to prevent the tools from reading the profile but
meillo@220: to instruct them to read a different profile.
meillo@138: .Pn mhmail
meillo@220: could have set up a well-defined profile and caused the following
meillo@138: .Pn post
meillo@220: to use this profile by exporting an environment variable.
meillo@220: With this approach, no special cases would have been introduced
meillo@220: and no surprises would have been caused.
meillo@220: By writing a wrapper program to provide a clean temporary profile,
meillo@220: the concept could have been generalized orthogonally to the whole
meillo@220: MH tool chest.
meillo@220: .P
meillo@220: In mmh, the wish to have
meillo@220: .Pn mhmail
meillo@220: as a replacement for
meillo@220: .Pn mailx
meillo@220: is considered obsolete.
meillo@220: Mmh's
meillo@220: .Pn mhmail
meillo@220: does no longer cover this use-case
meillo@220: .Ci d36e56e695fe1c482c7920644bfbb6386ac9edb0 .
meillo@220: Currently,
meillo@220: .Pn mhmail
meillo@220: is in a transition state
meillo@220: .Ci 32d4f9daaa70519be3072479232ff7be0500d009 .
meillo@220: It may become a front-end to
meillo@220: .Pn comp ,
meillo@220: which provides an alternative interface which can be more convenient
meillo@220: in some cases.
meillo@220: This would convert
meillo@220: .Pn mhmail
meillo@220: into an ordinary MH tool.
meillo@220: If, however, this idea does not convince, then
meillo@220: .Pn mhmail
meillo@220: will be removed.
meillo@220: .P
meillo@220: In the mmh tool chest, every program reads the profile.
meillo@220: (\c
meillo@220: .Pn slocal
meillo@220: is not considered part of the mmh tool chest (cf. Sec.
meillo@220: .Cf slocal ).)
meillo@220: Mmh has no
meillo@220: .Pn post
meillo@220: program, but it has
meillo@220: .Pn spost ,
meillo@220: which now does read the profile
meillo@220: .Ci 3e017a7abbdf69bf0dff7a4073275961eda1ded8 .
meillo@220: Following this change,
meillo@220: .Pn send
meillo@220: and
meillo@220: .Pn spost
meillo@220: can be considered for merging.
meillo@220: Besides
meillo@220: .Pn send ,
meillo@220: .Pn spost
meillo@220: is only invoked directly by the to-be-changed
meillo@220: .Pn mhmail
meillo@220: implementation and by
meillo@220: .Pn rcvdist ,
meillo@220: which requires rework anyway.
meillo@220: 
meillo@220: .P
meillo@220: Jeffrey Honig quoted Marshall T. Rose explaining the decision that
meillo@220: .Pn post
meillo@220: ignores the profile:
meillo@138: .[
meillo@197: nmh-workers honig post profile
meillo@138: .]
meillo@138: .QS
meillo@138: when you run mh commands in a script, you want all the defaults to be
meillo@138: what the man page says.
meillo@138: when you run a command by hand, then you want your own defaults...
meillo@138: .QE
meillo@138: .LP
meillo@220: The explanation neither matches the problem concered exactly
meillo@220: nor is the interpretation clear.
meillo@220: If the described desire addresses the technical level,
meillo@220: then it conflicts fundametally with the Unix philosophy,
meillo@220: precisely because the indistinquishability of human and script
meillo@220: input is the main reason for the huge software leverage in Unix.
meillo@220: If, however, the described desire addresses the user's view,
meillo@220: then different technical solutions are more appropriate.
meillo@220: The two cases can be regarded simply as two different MH setups.
meillo@220: Hence, mapping the problem of different behavior between interactive and
meillo@220: automated use on the concept of switching between different profiles,
meillo@220: marks it already solved.
meillo@133: 
meillo@133: 
meillo@127: 
meillo@121: .H2 "Standard Libraries
meillo@22: .P
meillo@121: MH is one decade older than the POSIX and ANSI C standards.
meillo@121: Hence, MH included own implementations of functions
meillo@220: that were neither standardized nor widely available, back then.
meillo@220: Today, twenty years after POSIX and ANSI C were published,
meillo@220: developers can expect that systems comply with these standards.
meillo@121: In consequence, MH-specific replacements for standard functions
meillo@121: can and should be dropped.
meillo@220: Kernighan and Pike advise: ``Use standard libraries''.
meillo@121: .[ [
meillo@121: kernighan pike practice of programming
meillo@121: .], p. 196]
meillo@121: Actually, MH had followed this advice in history,
meillo@220: but it had not adjusted to more recent changes in this field.
meillo@121: The
meillo@121: .Fu snprintf()
meillo@121: function, for instance, was standardized with C99 and is available
meillo@121: almost everywhere because of its high usefulness.
meillo@220: Thus, the project's own implementation of
meillo@121: .Fu snprintf()
meillo@123: was dropped in March 2012 in favor for using the one of the
meillo@220: standard library
meillo@220: .Ci 0052f1024deb0a0a2fc2e5bacf93d45a5a9c9b32 .
meillo@123: Such decisions limit the portability of mmh
meillo@173: if systems do not support these standardized and widespread functions.
meillo@123: This compromise is made because mmh focuses on the future.
meillo@121: .P
meillo@220: As I am still in my twenties, have no programming experience from
meillo@220: past decades.
meillo@220: I have not followed the evolution of C through time.
meillo@220: I have not suffered from the the Unix wars.
meillo@121: I have not longed for standardization.
meillo@121: All my programming experience is from a time when ANSI C and POSIX
meillo@121: were well established already.
meillo@220: Thus, I needed to learn about the history in retrospective.
meillo@121: I have only read a lot of books about the (good) old times.
meillo@220: This put me in a difficult position when working with old code.
meillo@123: I need to freshly acquire knowledge about old code constructs and ancient
meillo@123: programming styles, whereas older programmers know these things by
meillo@123: heart from their own experience.
meillo@123: Being aware of the situation, I rather let people with more historic
meillo@220: experience do the transition from ancient code constructs to
meillo@220: standardized ones.
meillo@121: Lyndon Nerenberg covered large parts of this task for the nmh project.
meillo@121: He converted project-specific functions to POSIX replacements,
meillo@121: also removing the conditionals compilation of now standardized features.
meillo@220: Ken Hornstein and David Levine had their part in this work, as well.
meillo@220: Often, I only pulled the changes over from nmh into mmh.
meillo@220: These changes include many commits, among them:
meillo@121: .Ci 768b5edd9623b7238e12ec8dfc409b82a1ed9e2d
meillo@121: .Ci 0052f1024deb0a0a2fc2e5bacf93d45a5a9c9b32 .
meillo@102: .P
meillo@220: Nevertheless, I worked on the task as well, tidying up the
meillo@220: \fIMH standard library\fP,
meillo@220: .Fn libmh.a .
meillo@220: It is located in the
meillo@123: .Fn sbr
meillo@220: (``subroutines'') directory in the source tree and
meillo@220: includes functions that mmh tools usually need.
meillo@123: Among them are MH-specific functions for profile, context, sequence,
meillo@123: and folder handling, but as well
meillo@123: MH-independent functions, such as auxiliary string functions,
meillo@123: portability interfaces and error-checking wrappers for critical
meillo@123: functions of the standard library.
meillo@220: .BU
meillo@123: I have replaced the
meillo@121: .Fu atooi()
meillo@121: function with calls to
meillo@220: .Fu strtoul() ,
meillo@220: setting the third parameter, the base, to eight.
meillo@123: .Fu strtoul()
meillo@220: is part of C89 and thus considered safe to use
meillo@220: .Ci c490c51b3c0f8871b6953bd0c74551404f840a74 .
meillo@220: .BU
meillo@121: I did remove project-included fallback implementations of
meillo@121: .Fu memmove()
meillo@121: and
meillo@220: .Fu strerror()
meillo@220: .Ci b067ff5c465a5d243ce5a19e562085a9a1a97215 ,
meillo@121: although Peter Maydell had re-included them into nmh in 2008
meillo@121: to support SunOS 4.
meillo@121: Nevertheless, these functions are part of ANSI C.
meillo@121: Systems that do not even provide full ANSI C support should not
meillo@121: put a load on mmh.
meillo@220: .BU
meillo@121: The
meillo@121: .Fu copy()
meillo@180: function copies the string in parameter one to the location in
meillo@180: parameter two.
meillo@121: In contrast to
meillo@121: .Fu strcpy() ,
meillo@121: it returns a pointer to the terminating null-byte in the destination area.
meillo@123: The code was adjusted to replace
meillo@121: .Fu copy()
meillo@123: with
meillo@121: .Fu strcpy() ,
meillo@121: except within
meillo@121: .Fu concat() ,
meillo@121: where
meillo@121: .Fu copy()
meillo@123: was more convenient.
meillo@123: Therefore, the definition of
meillo@121: .Fu copy()
meillo@123: was moved into the source file of
meillo@121: .Fu concat()
meillo@220: and its visibility it limited to that
meillo@220: .Ci 552fd7253e5ee9e554c5c7a8248a6322aa4363bb .
meillo@220: .BU
meillo@121: The function
meillo@121: .Fu r1bindex()
meillo@121: had been a generalized version of
meillo@121: .Fu basename()
meillo@121: with minor differences.
meillo@121: As all calls to
meillo@121: .Fu r1bindex()
meillo@220: had the slash (`\fL/\fP') as delimiter anyway,
meillo@121: replacing
meillo@121: .Fu r1bindex()
meillo@121: with the more specific and better-named function
meillo@121: .Fu basename()
meillo@121: became desirable.
meillo@121: Unfortunately, many of the 54 calls to
meillo@121: .Fu r1bindex()
meillo@123: depended on a special behavior,
meillo@121: which differed from the POSIX specification for
meillo@121: .Fu basename() .
meillo@121: Hence,
meillo@121: .Fu r1bindex()
meillo@121: was kept but renamed to
meillo@123: .Fu mhbasename() ,
meillo@220: setting the delimiter to the slash
meillo@220: .Ci 240013872c392fe644bd4f79382d9f5314b4ea60 .
meillo@121: For possible uses of
meillo@121: .Fu r1bindex()
meillo@121: with a different delimiter,
meillo@121: the ANSI C function
meillo@121: .Fu strrchr()
meillo@121: provides the core functionality.
meillo@220: .BU
meillo@121: The
meillo@121: .Fu ssequal()
meillo@121: function \(en apparently for ``substring equal'' \(en
meillo@121: was renamed to
meillo@121: .Fu isprefix() ,
meillo@220: because this is what it actually checked
meillo@220: .Ci c20b4fa14515c7ab388ce35411d89a7a92300711.
meillo@220: Its source file had included both of the following comments, no joke.
meillo@220: .in -\n(PIu
meillo@121: .VS
meillo@121: /*
meillo@121:  * THIS CODE DOES NOT WORK AS ADVERTISED.
meillo@121:  * It is actually checking if s1 is a PREFIX of s2.
meillo@121:  * All calls to this function need to be checked to see
meillo@121:  * if that needs to be changed. Prefix checking is cheaper, so
meillo@121:  * should be kept if it's sufficient.
meillo@121:  */
meillo@121: 
meillo@121: /*
meillo@121:  * Check if s1 is a substring of s2.
meillo@121:  * If yes, then return 1, else return 0.
meillo@121:  */
meillo@121: VE
meillo@220: .in +\n(PIu
meillo@220: Eventually, the function was completely replaced with calls to
meillo@220: .Fu strncmp()
meillo@220: .Ci b0b1dd37ff515578cf7cba51625189eb34a196cb .
meillo@121: 
meillo@102: 
meillo@102: 
meillo@102: 
meillo@133: 
meillo@133: .H2 "User Data Locations
meillo@133: .P
meillo@133: In nmh, a personal setup consists of the MH profile and the MH directory.
meillo@133: The profile is a file named
meillo@133: .Fn \&.mh_profile
meillo@133: in the user's home directory.
meillo@133: It contains the static configuration.
meillo@133: It also contains the location of the MH directory in the profile entry
meillo@133: .Pe Path .
meillo@133: The MH directory contains the mail storage and is the first
meillo@220: place to search for form files, scan formats, and similar
meillo@133: configuration files.
meillo@133: The location of the MH directory can be chosen freely by the user.
meillo@220: The usual name is a directory named
meillo@133: .Fn Mail
meillo@220: in the user's home directory.
meillo@133: .P
meillo@200: The way MH data is split between profile and MH directory is a legacy.
meillo@133: It is only sensible in a situation where the profile is the only
meillo@133: configuration file.
meillo@133: Why else should the mail storage and the configuration files be intermixed?
meillo@220: They are of different kind:
meillo@220: One kind is the data to be operated on and the other kind is
meillo@220: the configuration to change how tools operate.
meillo@133: Splitting the configuration between the profile and the MH directory
meillo@220: is inappropriate, as well.
meillo@220: I improved the situation by breaking compatibility.
meillo@133: .P
meillo@220: In mmh, personal data is grouped by type.
meillo@220: This results in two distinct parts:
meillo@171: the mail storage and the configuration.
meillo@220: The mail storage directory still contains all the messages,
meillo@133: but, in exception of public sequences files, nothing else.
meillo@133: In difference to nmh, the auxiliary configuration files are no longer
meillo@133: located there.
meillo@133: Therefore, the directory is no longer called the user's \fIMH directory\fP
meillo@220: but the user's \fImail storage\fP.
meillo@133: Its location is still user-chosen, with the default name
meillo@220: .Fn Mail
meillo@133: in the user's home directory.
meillo@220: The configuration is grouped together in the hidden directory
meillo@133: .Fn \&.mmh
meillo@133: in the user's home directory.
meillo@133: This \fImmh directory\fP contains the context file, personal forms,
meillo@133: scan formats, and the like, but also the user's profile, now named
meillo@133: .Fn profile .
meillo@220: The path to the profile is no longer
meillo@133: .Fn $HOME/.mh_profile
meillo@220: but
meillo@133: .Fn $HOME/.mmh/profile .
meillo@220: (The alternative of having file
meillo@133: .Fn $HOME/.mh_profile
meillo@220: and a configuration directory
meillo@133: .Fn $HOME/.mmh
meillo@220: appeared to be inconsistent.)
meillo@220: .P
meillo@133: The approach chosen for mmh is consistent, simple, and familiar to
meillo@133: Unix users.
meillo@220: The main achievement of the change is the clear and sensible separation
meillo@220: of the mail storage and the configuration.
meillo@168: .Ci 7030d7edb099bff36ded7548bb5380f7acab4f9b
meillo@133: .P
meillo@220: As MH allows users to have multiple MH setups,
meillo@220: it is necessary to switch the profile.
meillo@133: The profile is the single entry point to access the rest of a
meillo@133: personal MH setup.
meillo@133: In nmh, the environment variable
meillo@133: .Ev MH
meillo@220: is used to specify a different profile.
meillo@220: To operate in the same MH setup with a separate context, the
meillo@133: .Ev MHCONTEXT
meillo@220: environment variable is used.
meillo@220: This allows having a separate current folder in each terminal at
meillo@220: the same time, for instance.
meillo@220: In mmh, three environment variables replace the two of nmh.
meillo@133: .Ev MMH
meillo@133: overrides the default location of the mmh directory (\c
meillo@133: .Fn .mmh ).
meillo@133: .Ev MMHP
meillo@133: and
meillo@133: .Ev MMHC
meillo@220: override the paths to the profile and context file, respectively.
meillo@133: This approach allows the set of personal configuration files to be chosen
meillo@220: independently of the profile, context, and mail storage.
meillo@133: The new approach has no functional disadvantages,
meillo@133: as every setup I can imagine can be implemented with both approaches,
meillo@220: possibly even easier with the new one.
meillo@220: .Ci 7030d7edb099bff36ded7548bb5380f7acab4f9b
meillo@133: 
meillo@133: 
meillo@133: 
meillo@133: 
meillo@133: 
meillo@118: .H2 "Modularization
meillo@220: .Id modularization
meillo@118: .P
meillo@123: The source code of the mmh tools is located in the
meillo@122: .Fn uip
meillo@123: (``user interface programs'') directory.
meillo@180: Each tool has a source file with the name of the command.
meillo@122: For example,
meillo@122: .Pn rmm
meillo@122: is built from
meillo@122: .Fn uip/rmm.c .
meillo@123: Some source files are used for multiple programs.
meillo@122: For example
meillo@122: .Fn uip/scansbr.c
meillo@173: is used for both
meillo@122: .Pn scan
meillo@122: and
meillo@122: .Pn inc .
meillo@122: In nmh, 49 tools were built from 76 source files.
meillo@123: This is a ratio of 1.6 source files per program.
meillo@123: 32 programs depended on multiple source files;
meillo@123: 17 programs depended on one source file only.
meillo@122: In mmh, 39 tools are built from 51 source files.
meillo@123: This is a ratio of 1.3 source files per program.
meillo@123: 18 programs depend on multiple source files;
meillo@123: 21 programs depend on one source file only.
meillo@123: (These numbers and the ones in the following text ignore the MH library
meillo@123: as well as shell scripts and multiple names for the same program.)
meillo@122: .P
meillo@123: Splitting the source code of a large program into multiple files can
meillo@220: increase the readability of its source code,
meillo@220: but most of the mmh tools are small and straight-forward programs.
meillo@220: In exception of the MIME handling tools (i.e.
meillo@122: .Pn mhbuild ,
meillo@122: .Pn mhstore ,
meillo@122: .Pn show ,
meillo@220: etc.),
meillo@220: .Pn pick
meillo@220: is the only tool with more than one thousand lines of source code.
meillo@220: Splitting programs with less than one thousand lines of code into
meillo@220: multiple source files leads seldom to better readability.
meillo@220: For such tools, splitting still makes sense
meillo@220: when parts of the code are reused in other programs
meillo@179: and the reused code fragment is (1) not general enough
meillo@179: for including it in the MH library
meillo@179: or (2) has dependencies on a library that only few programs need.
meillo@122: .Fn uip/packsbr.c ,
meillo@122: for instance, provides the core program logic for the
meillo@122: .Pn packf
meillo@122: and
meillo@122: .Pn rcvpack
meillo@122: programs.
meillo@122: .Fn uip/packf.c
meillo@122: and
meillo@122: .Fn uip/rcvpack.c
meillo@122: mainly wrap the core function appropriately.
meillo@122: No other tools use the folder packing functions.
meillo@123: As another example,
meillo@123: .Fn uip/termsbr.c
meillo@220: accesses terminal properties, which requires linking with the
meillo@220: \fItermcap\fP or a \fIcurses\fP library.
meillo@220: If
meillo@123: .Fn uip/termsbr.c
meillo@220: is included in the MH library, then every program needs to be linked
meillo@220: with termcap or curses, although only few of the programs use
meillo@220: the library.
meillo@122: .P
meillo@122: The task of MIME handling is complex enough that splitting its code
meillo@122: into multiple source files improves the readability.
meillo@122: The program
meillo@122: .Pn mhstore ,
meillo@122: for instance, is compiled out of seven source files with 2\|500
meillo@122: lines of code in summary.
meillo@122: The main code file
meillo@122: .Fn uip/mhstore.c
meillo@220: consists of 800 lines; the other 1\|700 lines are code reused in
meillo@123: other MIME handling tools.
meillo@123: It seems to be worthwhile to bundle the generic MIME handling code into
meillo@123: a MH-MIME library, as a companion to the MH standard library.
meillo@220: This is left to be done.
meillo@122: .P
meillo@169: The work already accomplished focussed on the non-MIME tools.
meillo@122: The amount of code compiled into each program was reduced.
meillo@123: This eases the understanding of the code base.
meillo@122: In nmh,
meillo@122: .Pn comp
meillo@122: was built from six source files:
meillo@122: .Fn comp.c ,
meillo@122: .Fn whatnowproc.c ,
meillo@122: .Fn whatnowsbr.c ,
meillo@122: .Fn sendsbr.c ,
meillo@122: .Fn annosbr.c ,
meillo@122: and
meillo@122: .Fn distsbr.c .
meillo@122: In mmh, it builds from only two:
meillo@122: .Fn comp.c
meillo@122: and
meillo@122: .Fn whatnowproc.c .
meillo@123: In nmh's
meillo@123: .Pn comp ,
meillo@123: the core function of
meillo@122: .Pn whatnow ,
meillo@122: .Pn send ,
meillo@122: and
meillo@122: .Pn anno
meillo@220: were all compiled into
meillo@122: .Pn comp .
meillo@123: This saved the need to execute these programs with
meillo@220: the expensive system calls
meillo@122: .Fu fork()
meillo@122: and
meillo@220: .Fu exec() .
meillo@171: Whereas this approach improved the time performance,
meillo@171: it interwove the source code.
meillo@122: Core functionalities were not encapsulated into programs but into
meillo@122: function, which were then wrapped by programs.
meillo@122: For example,
meillo@122: .Fn uip/annosbr.c
meillo@122: included the function
meillo@122: .Fu annotate() .
meillo@122: Each program that wanted to annotate messages, included the source file
meillo@123: .Fn uip/annosbr.c
meillo@123: and called
meillo@123: .Fu annotate() .
meillo@123: Because the function
meillo@123: .Fu annotate()
meillo@123: was used like the tool
meillo@123: .Pn anno ,
meillo@123: it had seven parameters, reflecting the command line switches of the tool.
meillo@122: When another pair of command line switches was added to
meillo@122: .Pn anno ,
meillo@122: a rather ugly hack was implemented to avoid adding another parameter
meillo@220: to the function
meillo@220: .Ci d9b1d57351d104d7ec1a5621f090657dcce8cb7f .
meillo@122: .P
meillo@122: In mmh, the relevant code of
meillo@122: .Pn comp
meillo@122: comprises the two files
meillo@122: .Fn uip/comp.c
meillo@122: and
meillo@122: .Fn uip/whatnowproc.c ,
meillo@220: together 210 lines of code,
meillo@220: whereas in nmh,
meillo@122: .Pn comp
meillo@122: comprises six files with 2\|450 lines.
meillo@220: Not all of the code in these six files is actually used by
meillo@122: .Pn comp ,
meillo@220: but the reader needed to read it all to know which parts are relevant.
meillo@220: Understanding nmh's
meillo@122: .Pn comp ,
meillo@123: required understanding the inner workings of
meillo@122: .Fn uip/annosbr.c
meillo@122: first.
meillo@123: To be sure to fully understand a program, its whole source code needs
meillo@122: to be examined.
meillo@123: Not doing so is a leap of faith, assuming that the developers
meillo@122: have avoided obscure programming techniques.
meillo@220: Here, it should be recalled that information passed in obscure ways
meillo@220: through the program's source base, due to the aforementioned hack
meillo@220: to save an additional parameter in nmh's
meillo@220: .Pn anno .
meillo@220: .P
meillo@220: In mmh, understanding
meillo@220: .Pn comp
meillo@220: requires to read only 210 lines of code to read, whereas the amount
meillo@220: is ten times more for nmh's
meillo@220: .Pn comp .
meillo@220: .P
meillo@220: By separating the tools on the program-level,
meillo@220: the boundaries are clearly visible, as the interfaces are calls to
meillo@122: .Fu exec()
meillo@122: rather than arbitrary function calls.
meillo@220: Additionally, this kind of separation is more strict because 
meillo@220: it is technically enforced by the operating system;
meillo@220: it can not be simply bypassed with global variables.
meillo@220: Good separation simplifies the understanding of program code
meillo@220: because the area influenced by any particular statement is small.
meillo@220: As I have read a lot in nmh's code base during the last two years,
meillo@220: I have learned about the easy and the difficult parts.
meillo@220: In my observation, the understanding of code is enormously eased
meillo@220: if the influenced area is small and clearly bounded.
meillo@123: .P
meillo@220: Yet, the real problem is another:
meillo@123: Nmh violates the golden ``one tool, one job'' rule of the Unix philosophy.
meillo@123: Understanding
meillo@122: .Pn comp
meillo@123: requires understanding
meillo@123: .Fn uip/annosbr.c
meillo@123: and
meillo@123: .Fn uip/sendsbr.c
meillo@123: because
meillo@123: .Pn comp
meillo@220: annotates and sends messages.
meillo@220: In nmh, there surely exist the tools
meillo@220: .Pn anno
meillo@220: and
meillo@122: .Pn send ,
meillo@220: which cover these jobs,
meillo@220: but
meillo@122: .Pn comp
meillo@123: and
meillo@122: .Pn repl
meillo@122: and
meillo@122: .Pn forw
meillo@122: and
meillo@122: .Pn dist
meillo@122: and
meillo@122: .Pn whatnow
meillo@122: and
meillo@220: .Pn viamail
meillo@220: \(en they all (!) \(en
meillo@220: have the same annotating and sending functions included, once more.
meillo@220: As a result,
meillo@123: .Pn comp
meillo@123: sends messages without using
meillo@123: .Pn send .
meillo@123: The situation is the same as if
meillo@123: .Pn grep
meillo@220: would page its output without using
meillo@123: .Pn more
meillo@123: just because both programs are part of the same code base.
meillo@123: .P
meillo@220: The clear separation on the surface of nmh
meillo@220: \(en the tool chest approach \(en
meillo@123: is violated on the level below.
meillo@122: This violation is for the sake of time performance.
meillo@220: Decades ago, sacrificing readability and conceptional beauty
meillo@220: for speed might have been necessary to prevent MH from being
meillo@220: unusably slow, but today this is not the case anymore.
meillo@220: No longer should speed improvements that became unnecessary be kept.
meillo@220: No longer should readability or conceptional beauty be sacrificed.
meillo@220: No longer should the Unix philosophy's ``one tool, one job''
meillo@220: guideline be violated.
meillo@123: Therefore, mmh's
meillo@123: .Pn comp
meillo@220: no longer sends messages.
meillo@220: .P
meillo@123: In mmh, different jobs are divided among separate programs that
meillo@122: invoke each other as needed.
meillo@123: In consequence,
meillo@123: .Pn comp
meillo@123: invokes
meillo@123: .Pn whatnow
meillo@123: which thereafter invokes
meillo@220: .Pn send
meillo@168: .Ci 3df5ab3c116e6d4a2fb4bb5cc9dfc5f781825815
meillo@220: .Ci c73c00bfccd22ec77e9593f47462aeca4a8cd9c0 .
meillo@123: The clear separation on the surface is maintained on the level below.
meillo@220: Human users and other tools use the same interface \(en
meillo@123: annotations, for example, are made by invoking
meillo@123: .Pn anno ,
meillo@220: no matter if requested by programs or by human beings
meillo@168: .Ci 469a4163c2a1a43731d412eaa5d9cae7d670c48b
meillo@168: .Ci aed384169af5204b8002d06e7a22f89197963d2d
meillo@220: .Ci 3caf9e298a8861729ca8b8a84f57022b6f3ea742 .
meillo@123: The decrease of tools built from multiple source files and thus
meillo@123: the decrease of
meillo@123: .Fn uip/*sbr.c
meillo@220: files confirm the improvement
meillo@168: .Ci 9e6d91313f01c96b4058d6bf419a8ca9a207bc33
meillo@168: .ci 81744a46ac9f845d6c2b9908074d269275178d2e
meillo@168: .Ci f0f858069d21111f0dbea510044593f89c9b0829
meillo@168: .Ci 0503a6e9be34f24858b55b555a5c948182b9f24b
meillo@168: .Ci 27826f9353e0f0b04590b7d0f8f83e60462b90f0
meillo@168: .Ci d1da1f94ce62160aebb30df4063ccbc53768656b
meillo@220: .Ci c42222869e318fff5dec395eca3e776db3075455 .
meillo@220: This is also visible in the complexity of the build dependency graphs:
meillo@145: 
meillo@220: .sp
meillo@220: Nmh:
meillo@220: .BP input/deps-nmh.eps .5i
meillo@220: .EP
meillo@220: .sp
meillo@220: Mmh:
meillo@220: .BP input/deps-mmh.eps .8i
meillo@220: .EP
meillo@145: 
meillo@220: The figures display all program to source file relationships
meillo@220: that are not one-to-one,
meillo@220: i.e. all programs that are built from multiple source files.
meillo@220: The primary source file of each program is omited from the graph.