docs/master

view discussion.roff @ 235:e58400695ae2

Added tag final version for changeset 348b92755bef
author markus schnalke <meillo@marmaro.de>
date Mon, 16 Jul 2012 11:25:04 +0200
parents 96778c1afc3e
children
line source
1 .H0 "Discussion
2 .P
3 This main chapter discusses the practical work accomplished in the
4 mmh project.
5 It is structured along the goals chosen for the project.
6 A selection of the work undertaken
7 is described.
8 .P
9 This discussion compares the present version of mmh with the state of
10 nmh at the time when the mmh project had started, i.e. fall 2011.
11 Recent changes in nmh are rarely part of the discussion.
12 .P
13 Whenever lines of code are counted, David A. Wheeler's \fIsloccount\fP
14 was used to measure the amount in a comparable way.
15 .P
16 For the reader's convenience, the structure of modern email systems
17 is depicted in the following figure.
18 It illustrates the path a message takes from sender to recipient.
20 .sp 1.5
21 .KS
22 .in 2c
23 .so input/mail-agents.pic
24 .KE
25 .sp 1.5
27 .LP
28 The ellipses denote mail agents, i.e. different jobs in email processing.
29 These are:
30 .IP "Mail User Agent (MUA)
31 The only program users directly interact with.
32 It includes functions to compose new mail, display received mail,
33 and to manage the mail storage.
34 It is called a \fImail client\fP as well.
35 .IP "Mail Submission Agent (MSA)
36 A special kind of Mail Transfer Agent, used to submit mail into the
37 mail transport system.
38 Often it is also called an MTA.
39 .IP "Mail Transfer Agent (MTA)
40 A node in the mail transport system.
41 It transfers incoming mail to a transport node nearer to the
42 final destination.
43 An MTA may be the final destination itself.
44 .IP "Mail Delivery Agent (MDA)
45 Delivers mail according to a set of rules.
46 Usually, the messages are stored to disk.
47 .IP "Mail Retrieval Agent (MRA)
48 Initiates the transfer of mail from a remote location to the local machine.
49 (The dashed arrow in the figure represents the pull request.)
50 .LP
51 The dashed boxes represent entities that usually reside on single machines.
52 The box on the lower left represents the sender's system.
53 The box on the upper left represents the first mail transfer node.
54 The box on the upper right represents the transfer node responsible
55 for the destination address.
56 The box on the lower right represents the recipient's system.
57 Often, the boxes above the dotted line are servers on the Internet.
58 Many mail clients, including nmh, include all of the components below
59 the dotted line.
60 This is not the case for mmh; it implements the MUA only.
67 .\" --------------------------------------------------------------
68 .H1 "Streamlining
70 .P
71 MH once provided a complete email system.
72 The community around nmh tries to keep nmh in similar shape.
73 In fundamental contrast, mmh shall be an MUA only.
74 I believe that the development of all-in-one mail systems is obsolete.
75 Today, email is too complex to be fully covered by a single project.
76 Such a project will not be able to excel in all aspects.
77 Instead, the aspects of email should be covered by multiple projects,
78 which then can be combined to form a complete system.
79 Excellent implementations for the various aspects of email already exist.
80 Just to name three examples: Postfix is a specialized MTA, Procmail
81 is a specialized MDA, and Fetchmail is a specialized MRA.
82 I believe that it is best to use such specialized tools instead of
83 providing the same function once more as a side component.
84 .P
85 Doing something well requires focusing on a small set of specific aspects.
86 Under the assumption that development which is focussed on a particular
87 area produces better results there, specialized projects will be superior
88 in their field of focus.
89 Hence, all-in-one mail system projects \(en no matter if monolithic
90 or modular \(en will never be the best choice in any of the fields.
91 Even in providing the most consistent all-in-one system, they are likely
92 to be beaten by projects that focus exclusively on the creation
93 of a homogeneous system by integrating existing mail components.
94 .P
95 Usually, the limiting resource in the community development of
96 free software is man power.
97 If the development effort is spread over a large development area,
98 it becomes more difficult to compete with the specialists in the
99 various fields.
100 The concrete situation for MH-based mail systems is even tougher,
101 given their small and aged community, concerning both developers and users.
102 .P
103 In consequence, I believe that the available development resources
104 should focus on the point where MH is most unique.
105 This is clearly the user interface \(en the MUA.
106 Peripheral parts should be removed to streamline mmh for the MUA task.
109 .H2 "Mail Transfer Facilities
110 .Id mail-transfer-facilities
111 .P
112 The removal of the mail transfer facilities, effectively dropping the
113 MSA and MRA, had been the first work task in the mmh project.
114 The desire for this change initiated the creation of the mmh project.
115 .P
116 Focusing on one mail agent role only, is motivated by Eric Allman's
117 experience with Sendmail.
118 He identified the limitation of Sendmail
119 to the MTA task as one reason for its success:
120 .[ [
121 costales sendmail
122 .], p. xviii]
123 .QS
124 Second, I limited myself to the routing function \(en
125 I wouldn't write user agents or delivery back-ends.
126 This was a departure of the dominant thought of the time,
127 in which routing logic, local delivery, and often the network code
128 were incorporated directly into the user agents.
129 .QE
130 .P
131 In nmh, the MSA is called \fIMessage Transfer Service\fP (MTS).
132 This facility, implemented by the
133 .Pn post
134 command, establishes network connections and spoke SMTP to submit
135 messages to be relayed to the outside world.
136 When email transfer changed, this part needed to be changed as well.
137 Encryption and authentication for network connections
138 needed to be supported, hence TLS and SASL were introduced into nmh.
139 This added complexity without improving the core functions.
140 Furthermore, keeping up with recent developments in the field of
141 mail transfer requires development power and specialists.
142 In mmh, this whole facility was simply cut off
143 .Ci f6aa95b724fd8c791164abe7ee5468bf5c34f226
144 .Ci fecd5d34f65597a4dfa16aeabea7d74b191532c3
145 .Ci 156d35f6425bea4c1ed3c4c79783dc613379c65b .
146 Instead, mmh depends on an external MSA.
147 All outgoing mail in mmh goes through the
148 .Pn sendmail
149 command, which almost any MSA provides.
150 If not, a wrapper program can be written.
151 It must read the message from the standard input, extract the
152 recipient addresses from the message header, and hand the message
153 over to the MSA.
154 For example, a wrapper script for qmail would be:
155 .VS
156 #!/bin/sh
157 exec qmail-inject # ignore command line arguments
158 VE
159 The requirement to parse the recipient addresses out of the message header
160 may be removed in the future.
161 Mmh could pass the recipient addresses as command line arguments.
162 This appears to be the better interface.
163 .P
164 To retrieve mail, the
165 .Pn inc
166 command in nmh acts as MRA.
167 It establishes network connections
168 and speaks POP3 to retrieve mail from remote servers.
169 As with mail submission, the network connections required encryption and
170 authentication, thus TLS and SASL were added to nmh.
171 Support for message retrieval through IMAP will soon become necessary
172 additions, too, and likewise for any other changes in mail transfer.
173 But not in mmh because it has dropped the support for retrieving mail
174 from remote locations
175 .Ci ab7b48411962d26439f92f35ed084d3d6275459c .
176 Instead, it depends on an external tool to cover this task.
177 Mmh has two paths for messages to enter mmh's mail storage:
178 (1) Mail can be incorporated with
179 .Pn inc
180 from the system maildrop, or (2) with
181 .Pn rcvstore
182 by reading them, one at a time, from the standard input.
183 .P
184 With the removal of the MSA and MRA, mmh converted from a complete
185 mail system to only an MUA.
186 Now, of course, mmh depends on third-party software.
187 An external MSA is required to transfer mail to the outside world;
188 an external MRA is required to retrieve mail from remote machines.
189 Excellent implementations of such software exist.
190 They likely are superior to the internal versions that were removed.
191 Additionally, the best suiting programs can be chosen freely.
192 .P
193 As it had already been possible to use an external MSA and MRA,
194 why should the internal version not be kept for convenience?
195 Transferred to a different area, the question,
196 whether there is sense in having a fall-back pager in all
197 the command line tools for the cases when
198 .Pn more
199 or
200 .Pn less
201 are not available, appears to be ridiculous.
202 Of course, MSAs and MRAs are more complex than text pagers
203 and not necessarily available but still the concept of orthogonal
204 design holds: ``Write programs that do one thing and do it well''.
205 .[
206 mcilroy unix phil
207 p. 53
208 .]
209 .[
210 mcilroy bstj foreword
211 .]
212 Here, this part of the Unix philosophy was applied not only
213 to the programs but to the project itself.
214 In other words:
215 Develop projects that focus on one thing and do it well.
216 Projects which have grown complex should be split, for the same
217 reasons that programs which have grown complex should be split.
218 If it is conceptionally more elegant to have the MSA and MRA as
219 separate projects then they should be separated.
220 In my opinion, this is the case.
221 The RFCs suggest this separation by clearly distinguishing the
222 different mail handling tasks [RFC\|821].
223 The small interfaces between the mail agents support the
224 separation as well.
225 .P
226 Once, email had been small and simple.
227 At that time,
228 .Pn /bin/mail
229 had covered everything there was to email and still was small and simple.
230 Later, the essential complexity of email increased.
231 (Essential complexity is the complexity defined by the problem itself
232 .[ [
233 brooks no silver bullet
234 .]].)
235 Consequently, email systems grew.
236 RFCs started to introduce the concept of mail agents to separate the
237 various roles because they became more extensive and because
238 new roles appeared.
239 As mail system implementations grew, parts of them were split off.
240 For instance, a POP server was included in the original MH;
241 it was removed in nmh.
242 Now is the time to go one step further and split off the MSA and MRA,
243 as well.
244 Not only does this decrease the code size of the project,
245 more importantly, it unburdens mmh of the whole field of
246 message transfer, with all its implications for the project.
247 There is no more need for concern with changes in network transfer.
248 This independence is gained by depending on external components
249 that cover the field.
250 .P
251 In general, functionality can be added in three different ways:
252 .LI 1
253 By implementing the function in the project itself.
254 .LI 2
255 By depending on a library that provides the function.
256 .LI 3
257 By depending on a program that provides the function.
258 .LP
259 While implementing the function in the project itself leads to the
260 largest increase in code size and requires the most maintenance
261 and development work,
262 it keeps the project's dependence on other software lowest.
263 Using libraries or external programs requires less maintenance work
264 but introduces dependencies on external projects.
265 Programs have the smallest interfaces and provide the best separation,
266 but possibly limit the information exchange.
267 External libraries are more strongly connected than external programs,
268 thus information can be exchanged in a more flexible manner.
269 Obviously, adding code to a project increases the maintenance work.
270 As implementing complex functions in the project itself adds
271 a lot of code, this should be avoided if possible.
272 Thus, the dependencies only change in their character,
273 not in their existence.
274 In mmh, library dependencies on
275 .Pn libsasl2
276 and
277 .Pn libcrypto /\c
278 .Pn libssl
279 were traded against program dependencies on an MSA and an MRA.
280 This also meant trading build-time dependencies against run-time
281 dependencies.
282 Besides providing stronger separation and greater flexibility,
283 program dependencies also allowed
284 over 6\|000 lines of code to be removed from mmh.
285 This made mmh's code base about 12\|% smaller.
286 Reducing the project's code size by such an amount without actually
287 losing functionality is a convincing argument.
288 Actually, as external MSAs and MRAs are likely superior to the
289 project's internal versions, the common user even gains functionality.
290 .P
291 Users of MH should not have problems setting up an external MSA and MRA.
292 Also, the popular MSAs and MRAs have large communities and a lot
293 of available documentation.
295 Choices for MSAs range from small forwarders such as \fIssmtp\fP and
296 \fInullmailer\fP, over mid-size MTAs including \fImasqmail\fP and
297 \fIdma\fP, up to full-featured MTAs as for instance \fIPostfix\fP.
298 MRAs are provided for example by \fIfetchmail\fP, \fIgetmail\fP,
299 \fImpop\fP, and \fIfdm\fP.
302 .H2 "Non-MUA Tools
303 .P
304 One goal of mmh is to remove the tools that do not significantly
305 contribute to the MUA's job.
306 Loosely related and rarely used tools distract from a lean appearance,
307 and require maintenance work without adding much to the core task.
308 By removing these tools, mmh became more streamlined and focused.
309 .BU
310 .Pn conflict
311 was removed
312 .Ci 8b235097cbd11d728c07b966cf131aa7133ce5a9
313 because it is a mail system maintenance tool and not MUA-related.
314 It even checked
315 .Fn /etc/passwd
316 and
317 .Fn /etc/group
318 for consistency, which is completely unrelated to email.
319 A tool like
320 .Pn conflict
321 is surely useful, but it should not be shipped with mmh.
322 .BU
323 .Pn rcvtty
324 was removed
325 .Ci 14767c94b3827be7c867196467ed7aea5f6f49b0
326 because its use case of writing to the user's terminal
327 on reception of mail is obsolete.
328 If users like to be informed of new mail, the shell's
329 .Ev MAILPATH
330 variable or graphical notifications are technically more appealing.
331 Writing to terminals directly is hardly ever desired today.
332 If, though, one prefers this approach, the standard tool
333 .Pn write
334 can be used in a way similar to:
335 .VS
336 scan -file - | write `id -un`
337 VE
338 .BU
339 .Pn viamail
340 was removed
341 .Ci eda72d6a7a7c20ff123043fb7f19c509ea01f932
342 when the new attachment system was activated, because
343 .Pn forw
344 could then cover the task itself.
345 The
346 .Pn sendfiles
347 shell script was rewritten as a wrapper around
348 .Pn forw
349 .Ci 0e82199cf3c991a173e0ac8aa776efdb3ded61e6 .
350 .BU
351 .Pn msgchk
352 was removed
353 .Ci bb9360ead7eb7a3fedcce2eeedfc660014e41dbe ,
354 because it lost its use case when POP support was removed.
355 A call to
356 .Pn msgchk
357 provided hardly more information than:
358 .VS
359 ls -l /var/mail/meillo
360 VE
361 Yet, it distinguished between old and new mail, but
362 these details can be retrieved with
363 .Pn stat (1),
364 too.
365 A small shell script could be written to print the information
366 in a similar way, if truly necessary.
367 As mmh's
368 .Pn inc
369 only incorporates mail from the user's local maildrop,
370 and thus no data transfers over slow networks are involved,
371 there is hardly any need to check for new mail before incorporating it.
372 .BU
373 .Pn msh
374 was removed
375 .Ci 916690191222433a6923a4be54b0d8f6ac01bd02
376 because the tool was in conflict with the philosophy of MH.
377 It provided an interactive shell to access the features of MH.
378 However, it was not just a shell tailored to the needs of mail handling,
379 but one large program that had several MH tools built in.
380 This conflicted with the major feature of MH of being a tool chest.
381 .Pn msh 's
382 main use case had been accessing Bulletin Boards, which have ceased to
383 be popular.
384 .P
385 Removing
386 .Pn msh
387 together with the truly archaic code relics
388 .Pn vmh
389 and
390 .Pn wmh
391 saved more than 7\|000 lines of C code \(en
392 about 15\|% of the project's original source code amount.
393 Having less code \(en with equal readability, of course \(en
394 for the same functionality is an advantage.
395 Less code means less bugs and less maintenance work.
396 As
397 .Pn rcvtty
398 and
399 .Pn msgchk
400 are assumed to be rarely used and can be implemented in different ways,
401 why should one keep them?
402 Removing them streamlined mmh.
403 .Pn viamail 's
404 use case is now partly obsolete and partly covered by
405 .Pn forw ,
406 hence there is no reason to still maintain it.
407 .Pn conflict
408 is not related to the mail client, and
409 .Pn msh
410 conflicts with the basic concept of MH.
411 These two tools might still be useful, but they should not be part of mmh.
412 .P
413 .Id slocal
414 Finally, there is
415 .Pn slocal ,
416 which is an MDA and thus not directly MUA-related.
417 It should be removed from mmh because including it conflicts with
418 the idea that mmh is an MUA only.
419 However,
420 .Pn slocal
421 provides rule-based processing of messages, like filing them into
422 different folders, which is otherwise not available in mmh.
423 Although
424 .Pn slocal
425 neither pulls in dependencies, nor does it include a separate
426 technical area (cf. Sec.
427 .Cf mail-transfer-facilities ),
428 it still accounts for about 1\|000 lines of code that need to be maintained.
429 As
430 .Pn slocal
431 is almost self-standing, it should be split off into a separate project.
432 This would cut the strong connection between the MUA mmh and the MDA
433 .Pn slocal .
434 For anyone not using MH,
435 .Pn slocal
436 would become yet another independent MDA, like
437 .I procmail .
438 Then
439 .Pn slocal
440 could be installed without a complete MH system.
441 Likewise, mmh users could decide to use
442 .I procmail
443 without having a second, unused MDA, i.e.
444 .Pn slocal ,
445 installed.
446 That appears to be conceptionally the best solution.
447 Yet,
448 .Pn slocal
449 is not split off.
450 I defer the decision over
451 .Pn slocal
452 out of a need for deeper investigation.
453 In the meanwhile, it remains part of mmh
454 as its continued existence is not significant;
455 .Pn slocal
456 is unrelated to the rest of the project.
460 .H2 "Displaying Messages
461 .Id mhshow
462 .P
463 Since the very beginning, already in the first concept paper,
464 .[
465 original memo rand mh shapiro gaines
466 .]
467 .Pn show
468 had been MH's message display program.
469 .Pn show
470 mapped message numbers and sequences to files and invoked
471 .Pn mhl
472 to have the files formatted.
473 With MIME, this approach was not sufficient anymore.
474 MIME messages can consist of multiple parts.
475 Some parts, like binary attachments or text content in foreign charsets,
476 are not directly displayable.
477 .Pn show 's
478 understanding of messages and
479 .Pn mhl 's
480 display capabilities could not cope with the task any longer.
481 .P
482 Instead of extending these tools, additional tools were written from
483 scratch and were added to the MH tool chest.
484 Doing so is encouraged by the tool chest approach.
485 Modular design is a great advantage for extending a system,
486 as new tools can be added without interfering with existing ones.
487 First, the new MIME features were added in form of the single program
488 .Pn mhn .
489 The command
490 .Cl "mhn -show 42
491 had then shown the message number
492 .Fn 42 ,
493 interpreting MIME.
494 With the 1.0 release of nmh in February 1999, Richard Coleman finished
495 the split of
496 .Pn mhn
497 into a set of specialized tools, which together covered the
498 multiple aspects of MIME.
499 One of them was
500 .Pn mhshow ,
501 which replaced
502 .Cl "mhn -show" .
503 It was capable of displaying MIME messages appropriately.
504 .P
505 .ZZ
506 From then on, two message display tools were part of nmh,
507 .Pn show
508 and
509 .Pn mhshow .
510 To ease the life of users,
511 .Pn show
512 was extended to automatically hand the job over to
513 .Pn mhshow
514 if displaying the message would be beyond
515 .Pn show 's
516 abilities.
517 In consequence, the user would simply invoke
518 .Pn show
519 (possibly through
520 .Pn next
521 or
522 .Pn prev )
523 and get the message printed with either
524 .Pn show
525 or
526 .Pn mhshow ,
527 whatever was more appropriate.
528 .P
529 Having two similar tools for basically the same task is redundancy.
530 Usually, users do not distinguish between
531 .Pn show
532 and
533 .Pn mhshow
534 in their daily mail reading.
535 Having two separate display programs was therefore unnecessary
536 from a user's point of view.
537 Besides, the development of both programs needed to be in sync,
538 to ensure that the programs behaved in a similar way,
539 because they were used like a single tool.
540 Different behavior would have surprised the user.
541 .P
542 Today, non-MIME messages are rather seen to be a special case of
543 MIME messages, although it is the other way round.
544 As
545 .Pn mhshow
546 already had been able to display non-MIME messages, it appeared natural
547 to drop
548 .Pn show
549 in favor of using
550 .Pn mhshow
551 exclusively
552 .Ci 4c1efddfd499300c7e74263e57d8aa137e84c853 .
553 Removing
554 .Pn show
555 is no loss in function, because
556 .Pn mhshow
557 covers it completely.
558 Yet, the old behavior of
559 .Pn show
560 can still be emulated with the simple command line:
561 .VS
562 mhl `mhpath c`
563 VE
564 .P
565 For convenience,
566 .Pn mhshow
567 was renamed to
568 .Pn show
569 after
570 .Pn show
571 was gone.
572 It is clear that such a rename may confuse future developers when
573 trying to understand the history.
574 Nevertheless, I consider the convenience on the user's side,
575 to outweigh the inconvenience for understanding the evolution
576 of the tools.
577 .P
578 To prepare for the transition,
579 .Pn mhshow
580 was reworked to behave more like
581 .Pn show
582 first (cf. Sec.
583 .Cf mhshow ).
584 Once the tools behaved more alike, the replacing appeared to be
585 even more natural.
586 Today, mmh's new
587 .Pn show
588 has become the one single message display program once again,
589 with the difference
590 that today it handles MIME messages as well as non-MIME messages.
591 The outcomes of the transition are one program less to maintain,
592 no second display program for users to deal with,
593 and less system complexity.
594 .P
595 Still, removing the old
596 .Pn show
597 hurts in one regard: It had been such a simple program.
598 Its lean elegance is missing from the new
599 .Pn show ,
600 but there is no alternative;
601 supporting MIME demands higher essential complexity.
605 .H2 "Configure Options
606 .P
607 Customization is a double-edged sword.
608 It allows better suiting setups, but not for free.
609 There is the cost of code complexity to be able to customize.
610 There is the cost of less tested setups, because there are
611 more possible setups and especially corner cases.
612 Steve Johnson confirms:
613 .[ [
614 eric raymond the art of unix programming
615 .], p. 233]
616 .QS
617 Unless it is done very carefully, the addition of an on/off
618 configuration option can lead to a need to double the amount of
619 testing. Since in practice one never does double the amount of testing,
620 the practical effect is to reduce the amount of testing that any
621 given configuration receives. Ten options leads to 1024 times as much
622 testing, and pretty soon you are talking real reliability problems.
623 .QE
624 .LP
625 Additionally, there is the cost of choice itself.
626 The code complexity directly affects the developers.
627 Less tested code affects both users and developers.
628 The problem of choice affects the users, for once by having to choose
629 but also by more complex interfaces that require more documentation.
630 Whenever options add few advantages but increase the complexity of the
631 system, they should be considered for removal.
632 I have reduced the number of project-specific configure options from
633 15 to 3.
635 .U3 "Mail Transfer Facilities
636 .P
637 With the removal of the mail transfer facilities (cf. Sec.
638 .Cf mail-transfer-facilities )
639 five configure options vanished:
640 .P
641 The switches
642 .Sw --with-tls
643 and
644 .Sw --with-cyrus-sasl
645 had activated the support for transfer encryption and authentication.
646 They are not needed anymore.
647 .Ci fecd5d34f65597a4dfa16aeabea7d74b191532c3
648 .Ci 156d35f6425bea4c1ed3c4c79783dc613379c65b
649 .P
650 The configure switch
651 .Sw --enable-pop
652 had activated the message retrieval facility.
653 Whereas the code area that had been conditionally compiled in
654 for TLS and SASL support was small,
655 the conditionally compiled code area for POP support was much larger.
656 The code base had only changed slightly on toggling TLS or SASL
657 support but it had changed much on toggling POP support.
658 The changes in the code base could hardly be overviewed.
659 By having POP support togglable, a second code base had been created,
660 one that needed to be tested.
661 This situation is basically similar for the conditional TLS and SASL
662 code, but there the changes are minor and can yet be overviewed.
663 Still, conditional compilation of a code base creates variations
664 of the original program.
665 More variations require more testing and maintenance work.
666 .P
667 Two other options had only specified default configuration values:
668 .Sw --with-mts
669 defined the default transport service
670 .Ci f6aa95b724fd8c791164abe7ee5468bf5c34f226 .
671 With
672 .Sw --with-smtpservers
673 default SMTP servers could be set
674 .Ci 128545e06224233b7e91fc4c83f8830252fe16c9 .
675 Both of them became irrelevant when the SMTP transport service was removed.
676 In mmh, all messages are handed over to
677 .Pn sendmail
678 for transportation.
681 .U3 "Backup Prefix
682 .P
683 The backup prefix is the string that was prepended to message
684 filenames to tag them as deleted.
685 By default it had been the comma character (`\fL,\fP').
686 In July 2000, Kimmo Suominen introduced
687 the configure option
688 .Sw --with-hash-backup
689 to change the default to the hash character `\f(CW#\fP'.
690 This choice was probably personal preference, but,
691 being related or not, words that start with the hash character
692 introduce a comment in the Unix shell.
693 Thus, the command line
694 .Cl "rm #13 #15
695 calls
696 .Pn rm
697 without arguments because the first hash character starts a comment
698 that reaches until the end of the line.
699 To delete the backup files,
700 .Cl "rm ./#13 ./#15"
701 needs to be used.
702 Thus, using the hash as backup prefix may be seen as a precaution
703 against backup loss.
704 .P
705 First, I removed the configure option but added the profile entry
706 .Pe Backup-Prefix ,
707 which allowed to specify an arbitrary string as backup prefix
708 .Ci 6c40d481d661d532dd527eaf34cebb6d3f8ed086 .
709 This change did not remove the choice but moved it to a location where
710 it suited better, in my eyes.
711 .P
712 Eventually however, the new trash folder concept
713 (cf. Sec.
714 .Cf trash-folder )
715 removed the need for the backup prefix completely.
716 .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
717 .Ci ca0b3e830b86700d9e5e31b1784de2bdcaf58fc5
720 .U3 "Editor and Pager
721 .Id editor-pager
722 .P
723 The two configure options
724 .CW --with-editor=EDITOR
725 .CW --with-pager=PAGER
726 were used to specify the default editor and pager at configure time.
727 Doing so at configure time made sense in the eighties,
728 when the set of available editors and pagers varied much across
729 different systems.
730 Today, the situation is more homogeneous.
731 The programs
732 .Pn vi
733 and
734 .Pn more
735 can be expected to be available on every Unix system,
736 as they are specified by POSIX since two decades.
737 (The specifications for
738 .Pn vi
739 and
740 .Pn more
741 appeared in
742 .[
743 posix 1987
744 .]
745 and,
746 .[
747 posix 1992
748 .]
749 respectively.)
750 As a first step, these two tools were hard-coded as defaults
751 .Ci 5d43a99db70c12a673028c7758c20cbe3e13ef5f .
752 Not changed were the
753 .Pe editor
754 and
755 .Pe moreproc
756 profile entries, which allowed the user to override the system defaults.
757 Later, the concept was reworked again to respect the standard
758 environment variables
759 .Ev VISUAL
760 and
761 .Ev PAGER
762 if they are set.
763 Today, mmh determines the editor to use in the following order,
764 taking the first available and non-empty item
765 .Ci f85f4b7ae62e3d05a945dcd46ead51f0a2a89a9b :
766 .LI 1
767 Environment variable
768 .Ev MMHEDITOR
769 .LI 2
770 Profile entry
771 .Pe Editor
772 .LI 3
773 Environment variable
774 .Ev VISUAL
775 .LI 4
776 Environment variable
777 .Ev EDITOR
778 .LI 5
779 Command
780 .Pn vi .
781 .LP
782 The pager to use is determined in a similar order
783 .Ci 0c4214ea2aec6497d0d67b436bbee9bc1d225f1e :
784 .LI 1
785 Environment variable
786 .Ev MMHPAGER
787 .LI 2
788 Profile entry
789 .Pe Pager
790 (replaces
791 .Pe moreproc )
792 .LI 3
793 Environment variable
794 .Ev PAGER
795 .LI 4
796 Command
797 .Pn more .
798 .LP
799 By respecting the
800 .Ev VISUAL /\c
801 .Ev EDITOR
802 and
803 .Ev PAGER
804 environment variables,
805 the new behavior complies with the common style on Unix systems.
806 It is more uniform and clearer for users.
809 .U3 "ndbm
810 .P
811 .Pn slocal
812 used to depend on the database library
813 .I ndbm .
814 The database is used to store the
815 .Hd Message-ID
816 header field values of all messages delivered.
817 This enabled
818 .Pn slocal
819 to suppress delivering the same message to the same user twice.
820 This features was enabled by the
821 .Sw -suppressdup
822 switch.
823 .P
824 Because a variety of versions of the database library exist,
825 .[
826 wolter unix incompat notes dbm
827 .]
828 complicated autoconf code was needed to detect them correctly.
829 Furthermore, the configure switches
830 .Sw --with-ndbm=ARG
831 and
832 .Sw --with-ndbmheader=ARG
833 were added to help with difficult setups that would
834 not be detected automatically or not correctly.
835 .P
836 By removing the suppress duplicates feature of
837 .Pn slocal ,
838 the dependency on
839 .I ndbm
840 vanished and 120 lines of complex autoconf code could be saved
841 .Ci ecd6d6a20cb7a1507e3a20d6c4cb3a1cf14c6bbf .
842 The change removed functionality but that is considered minor to the
843 improvement of dropping the dependency and the complex autoconf code.
846 .U3 "MH-E Support
847 .P
848 The configure option
849 .Sw --disable-mhe
850 was removed when the MH-E support was reworked.
851 MH-E is the Emacs front-end to MH.
852 It requires MH to provide minor additional functions.
853 The
854 .Sw --disable-mhe
855 configure option had switched off these extensions.
856 After removing the support for old versions of MH-E,
857 only the
858 .Sw -build
859 switches of
860 .Pn forw
861 and
862 .Pn repl
863 are left to be MH-E extensions.
864 They are now always built in because they add little code and complexity.
865 In consequence, the
866 .Sw --disable-mhe
867 configure option was removed
868 .Ci a7ce7b4a580d77b6c2c4d980812beb589aa4c643 .
869 Dropping the option also removed a variant of the code base
870 that would have needed to be tested.
871 This change was undertaken in January 2012 in nmh and
872 thereafter merged into mmh.
875 .U3 "Masquerading
876 .P
877 The configure option
878 .Sw --enable-masquerade
879 could take up to three arguments:
880 .CW draft_from ,
881 .CW mmailid ,
882 and
883 .CW username_extension .
884 They activated different types of address masquerading.
885 All of them were implemented in the SMTP-speaking
886 .Pn post
887 command.
888 Address masquerading is an MTA's task and mmh does not cover
889 this field anymore.
890 Hence, true masquerading needs to be implemented in the external MTA.
891 .P
892 The
893 .I mmailid
894 masquerading type is the oldest one of the three and the only one
895 available in the original MH.
896 It provided a
897 .I username
898 to
899 .I fakeusername
900 mapping, based on the
901 .Fn passwd 's
902 GECOS field.
903 Nmh's man page
904 .Mp mh-tailor (5)
905 described the use case as being the following:
906 .QS
907 This is useful if you want the messages you send to always
908 appear to come from the name of an MTA alias rather than your
909 actual account name. For instance, many organizations set up
910 `First.Last' sendmail aliases for all users. If this is
911 the case, the GECOS field for each user should look like:
912 ``First [Middle] Last <First.Last>''
913 .QE
914 .P
915 As mmh sends outgoing mail via the local MTA only,
916 the best location to do such global rewrites is there.
917 Besides, the MTA is conceptionally the right location because it
918 does the reverse mapping for incoming mail (aliasing), too.
919 Furthermore, masquerading set up there is readily available for all
920 mail software on the system.
921 Hence, mmailid masquerading was removed.
922 .Ci 0836c8000ccb34b59410ef1c15b1b7feac70ce5f
923 .P
924 The
925 .I username_extension
926 masquerading type did not replace the username but would append a suffix,
927 specified by the
928 .Ev USERNAME_EXTENSION
929 environment variable, to it.
930 This provided support for the
931 .I user-extension
932 feature of qmail
933 .[ [
934 sill qmail handbook
935 .], p. 141]
936 and the similar
937 .I "plussed user
938 processing of Sendmail.
939 .[ [
940 sendmail costales
941 .], p. 476]
942 The decision to remove this username_extension masquerading
943 was motivated by the fact that
944 .Pn spost
945 had not supported it yet.
946 Username extensions can be used in mmh, but less convenient.
947 .\" XXX In the format file: %(getenv USERNAME_EXTENSION)
948 .Ci 2abae0bfd0ad5bf898461e50aa4b466d641f23d9
949 .P
950 The
951 .I draft_from
952 masquerading type instructed
953 .Pn post
954 to use the value of the
955 .Hd From
956 header field as SMTP envelope sender.
957 Sender addresses could be replaced completely.
958 Mmh offers a kind of masquerading similar in effect, but
959 with technical differences.
960 As mmh does not transfer messages itself, the local MTA has final control
961 over the sender's address.
962 Any masquerading mmh introduces may be reverted by the MTA.
963 In times of pedantic spam checking, an MTA will take care to use
964 sensible envelope sender addresses to keep its own reputation up.
965 Nonetheless, the MUA can set the
966 .Hd From
967 header field and thereby propose a sender address to the MTA.
968 The MTA may then decide to take that one or generate the canonical sender
969 address for use as envelope sender address.
970 .Ci b14ea6073f77b4359aaf3fddd0e105989db9
971 .P
972 In mmh, the MTA will always extract the recipient and sender from the
973 message header (\c
974 .Pn sendmail 's
975 .Sw -t
976 switch).
977 The
978 .Hd From
979 header field of the draft may be set arbitrary by the user.
980 If it is missing, the canonical sender address will be generated by the MTA.
982 .U3 "Remaining Options
983 .P
984 Two configure options remain in mmh.
985 One of them is the file locking method to use:
986 .Sw --with-locking=[dot|fcntl|flock|lockf] .
987 The idea of removing all methods except the portable
988 .I "dot locking
989 and having that one as the default is appealing, but this change
990 requires deeper technical investigation into the topic.
991 The other remaining option,
992 .Sw --enable-debug ,
993 compiles the programs with debugging symbols.
994 This option is likely to stay.
999 .H2 "Command Line Switches
1000 .P
1001 The command line switches of MH tools are similar in style to the
1002 switches in the X Window System.
1003 They consist of a single dash (`\fL-\fP') followed by a word.
1004 For example
1005 .Cl -truncate .
1006 To ease typing, the switch can be abbreviated, given the remaining
1007 prefix is unambiguous.
1008 If no other switch starts with the letter `t', then any of
1009 .Cl "-truncate" ,
1010 .Cl "-trunc" ,
1011 .Cl "-tr" ,
1012 and
1013 .Cl "-t
1014 is equal.
1015 As a result, switches can neither be grouped (as in
1016 .Cl "ls -ltr" )
1017 nor can switch arguments be appended directly to the switch itself
1018 (as in
1019 .Cl "sendmail -q30m" ).
1020 Many switches have negating counter-parts, which start with `no'.
1021 For example
1022 .Cl "-notruncate
1023 inverts the
1024 .Cl "-truncate
1025 switch.
1026 They exist to override the effect of default switches in the profile.
1027 Every program in mmh has two generic switches:
1028 .Sw -help ,
1029 to print a short message on how to use the program, and
1030 .Sw -Version
1031 (with capital `V'), to tell what version of mmh the program belongs to.
1032 .P
1033 Switches change the behavior of programs.
1034 Programs that do one thing in one way require no switches.
1035 In most cases, doing something in exactly one way is too limiting.
1036 If one task should be accomplished in various ways,
1037 switches are a good approach to alter the behavior of a program.
1038 Changing the behavior of programs provides flexibility and customization
1039 to users, but at the same time it complicates the code,
1040 the documentation, and the usage of the program.
1041 Therefore, the number of switches should be kept small.
1042 A small set of well-chosen switches is best.
1043 Usually, the number of switches increases over time.
1044 Already in 1985, Rose and Romine have identified this as a major
1045 problem of MH:
1046 .[ [
1047 rose romine real work
1048 .], p. 12]
1049 .QS
1050 A complaint often heard about systems which undergo substantial development
1051 by many people over a number of years, is that more and more options are
1052 introduced which add little to the functionality but greatly increase the
1053 amount of information a user needs to know in order to get useful work done.
1054 This is usually referred to as creeping featurism.
1055 .QP
1056 Unfortunately MH, having undergone six years of off-and-on development by
1057 ten or so well-meaning programmers (the present authors included),
1058 suffers mightily from this.
1059 .QE
1060 .P
1061 Being reluctant to adding new switches (or \fIoptions\fP,
1062 as Rose and Romine call them) is one part of a counter-action,
1063 the other part is removing hardly used switches.
1064 Nmh's tools have lots of switches already implemented.
1065 Hence, cleaning up by removing some of them was the more important part
1066 of the counter-action.
1067 Removing existing functionality is always difficult because it
1068 breaks programs that use these functions.
1069 Also, for every obsolete feature, there'll always be someone who still
1070 uses it and thus opposes its removal.
1071 This puts the developer into the position,
1072 where sensible improvements to style are regarded as destructive acts.
1073 Yet, living with the featurism is far worse, in my eyes, because
1074 future needs will demand adding further features,
1075 worsening the situation more and more.
1076 Rose and Romine added in a footnote,
1077 ``[...]
1078 .Pn send
1079 will no doubt acquire an endless number of switches in the years to come''
1080 .[ [
1081 rose romine real work
1082 .], p. 12].
1083 Although clearly humorous, the comment points to the nature of the problem.
1084 Refusing to add any new switches would encounter the problem at its root,
1085 but this is not practical.
1086 New needs will require new switches and it would be unwise to block
1087 them strictly.
1088 Nevertheless, removing obsolete switches still is an effective approach
1089 to deal with the problem.
1090 Working on an experimental branch without an established user base,
1091 eased my work because I did not offend users when I removed existing
1092 functions.
1093 .P
1094 Rose and Romine counted 24 visible and 9 more hidden switches for
1095 .Pn send .
1096 In nmh, they increased up to 32 visible and 12 hidden ones.
1097 At the time of writing, no more than 4 visible switches and 1 hidden switch
1098 have remained in mmh's
1099 .Pn send .
1100 These numbers include the two generic switches,
1101 .Sw -help
1102 and
1103 .Sw -Version .
1104 .P
1105 Hidden switches are ones not documented.
1106 In mmh, 12 tools have hidden switches.
1107 9 of them are
1108 .Sw -debug
1109 switches, the other 6 provide special interfaces for internal use.
1110 .P
1111 The following figure displays the number of switches for each of the tools
1112 that is available in both nmh and mmh.
1113 The tools are sorted by the number of switches they had in nmh.
1114 Both visible and hidden switches were counted,
1115 but not the generic help and version switches.
1116 Whereas in the beginning of the project, the average tool had 11 switches,
1117 now it has no more than 5 \(en only half as many.
1118 If the `no' switches and similar inverse variant are folded onto
1119 their counter-parts, the average tool had 8 switches in pre-mmh times and
1120 has 4 now.
1121 The total number of functional switches in mmh dropped from 465
1122 to 233.
1124 .sp
1125 .KS
1126 .in 1c
1127 .so input/switches.grap
1128 .KE
1129 .sp
1131 .P
1132 .ZZ
1133 A part of the switches vanished after functions were removed.
1134 This was the case for network mail transfer, for instance.
1135 Sometimes, however, the work flow was the other way:
1136 I looked through the
1137 .Mp mh-chart (7)
1138 man page to identify the tools with apparently too many switches.
1139 Then I valued the benefit of each switch by examining
1140 the tool's man page and source code, aided by literature research
1141 and testing.
1144 .U3 "Draft Folder Facility
1145 .P
1146 A change early in the project was the complete transition from
1147 the single draft message to the draft folder facility
1148 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860
1149 (cf. Sec.
1150 .Cf draft-folder ).
1151 The draft folder facility was introduced in the mid-eighties, when
1152 Rose and Romine called it a ``relatively new feature''.
1153 .[
1154 rose romine real work
1155 .]
1156 Since then, the facility was included, inactive by default.
1157 By making it permanently active and by related rework of the tools, the
1158 .Sw -[no]draftfolder ,
1159 and
1160 .Sw -draftmessage
1161 switches could be removed from
1162 .Pn comp ,
1163 .Pn repl ,
1164 .Pn forw ,
1165 .Pn dist ,
1166 .Pn whatnow ,
1167 and
1168 .Pn send
1169 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860 .
1170 The only flexibility lost with this change is having multiple
1171 draft folders within one profile.
1172 I consider this a theoretical problem only.
1173 At the same time, the
1174 .Sw -draft
1175 switch of
1176 .Pn anno ,
1177 .Pn refile ,
1178 and
1179 .Pn send
1180 was removed.
1181 The special treatment of \fIthe\fP draft message became irrelevant after
1182 the rework of the draft system
1183 (cf. Sec.
1184 .Cf draft-folder ).
1187 .U3 "In Place Editing
1188 .P
1189 .Pn anno
1190 had the switches
1191 .Sw -[no]inplace
1192 to either annotate the message in place and thus preserve hard links,
1193 or annotate a copy to replace the original message.
1194 The latter approach broke hard links.
1195 Following the assumption that linked messages should truly be the
1196 same message and annotating it should not break the link, the
1197 .Sw -[no]inplace
1198 switches were removed and the previous default
1199 .Sw -inplace
1200 was made the definitive behavior
1201 .Ci c8195849d2e366c569271abb0f5f60f4ebf0b4d0 .
1202 The
1203 .Sw -[no]inplace
1204 switches of
1205 .Pn repl ,
1206 .Pn forw ,
1207 and
1208 .Pn dist
1209 could be removed, as well, as they were simply passed through to
1210 .Pn anno .
1211 .P
1212 .Pn burst
1213 also had
1214 .Sw -[no]inplace
1215 switches, but with a different meaning.
1216 With
1217 .Sw -inplace ,
1218 the digest had been replaced by the table of contents (i.e. the
1219 introduction text) and the burst messages were placed right
1220 after this message, renumbering all following messages.
1221 Also, any trailing text of the digest was lost, though,
1222 in practice, it usually consists of an end-of-digest marker only.
1223 Nonetheless, this behavior appeared less elegant than the
1224 .Sw -noinplace
1225 behavior, which already had been the default.
1226 Nmh's
1227 .Mp burst (1)
1228 man page reads:
1229 .QS
1230 If
1231 .Sw -noinplace
1232 is given, each digest is preserved, no table
1233 of contents is produced, and the messages contained within
1234 the digest are placed at the end of the folder. Other messages
1235 are not tampered with in any way.
1236 .QE
1237 .LP
1238 The decision to drop the
1239 .Sw -inplace
1240 behavior was supported by the code complexity and the possible data loss
1241 it caused.
1242 .Sw -noinplace
1243 was chosen to be the definitive behavior.
1244 .Ci 68a686adeb39223a5e1ad35e4a24890ec053679d
1247 .U3 "Forms and Format Strings
1248 .P
1249 Historically, the tools that had
1250 .Sw -form
1251 switches to supply a form file had
1252 .Sw -format
1253 switches as well to supply the contents of a form file as a string
1254 on the command line directly.
1255 In consequence, the following two lines equaled:
1256 .VS
1257 scan -form scan.mailx
1258 scan -format "`cat /path/to/scan.mailx`"
1259 VE
1260 The
1261 .Sw -format
1262 switches were dropped in favor for extending the
1263 .Sw -form
1264 switches
1265 .Ci f51956be123db66b00138f80464d06f030dbb88d .
1266 If their argument starts with an equal sign (`\fL=\fP'),
1267 then the rest of the argument is taken as a format string,
1268 otherwise the arguments is treated as the name of a format file.
1269 Thus, now the following two lines equal:
1270 .VS
1271 scan -form scan.mailx
1272 scan -form "=`cat /path/to/scan.mailx`"
1273 VE
1274 This rework removed the prefix collision between
1275 .Sw -form
1276 and
1277 .Sw -format .
1278 Typing `\fL-fo\fP' is sufficient to specify form file or format string.
1279 .P
1280 The different meaning of
1281 .Sw -format
1282 for
1283 .Pn forw
1284 and
1285 .Pn repl
1286 was removed in mmh.
1287 .Pn forw
1288 was completely switched to MIME-type forwarding, thus removing the
1289 .Sw -[no]format
1290 .Ci 6e271608b7b9c23771523f88d23a4d3593010cf1 .
1291 For
1292 .Pn repl ,
1293 the
1294 .Sw -[no]format
1295 switches were reworked to
1296 .Sw -[no]filter
1297 switches
1298 .Ci 67411b1f95d6ec987b4c732459e1ba8a8ac192c6 .
1299 The
1300 .Sw -format
1301 switches of
1302 .Pn send
1303 and
1304 .Pn post ,
1305 which had a third meaning, were removed likewise
1306 .Ci f3cb7cde0e6f10451b6848678d95860d512224b9 .
1307 Eventually, the ambiguity of the
1308 .Sw -format
1309 switches is resolved by not having such switches anymore in mmh.
1312 .U3 "MIME Tools
1313 .P
1314 The MIME tools, which once were part of
1315 .Pn mhn
1316 (whatever that stood for),
1317 had several switches that added little practical value to the programs.
1318 The
1319 .Sw -[no]realsize
1320 switches of
1321 .Pn mhbuild
1322 and
1323 .Pn mhlist
1324 were removed
1325 .Ci 8d8f1c3abc586c005c904e52c4adbfe694d2201c .
1326 Real size calculations are done always now because nmh's
1327 .Mp mhbuild (1)
1328 man page states that
1329 ``This provides an accurate count at the expense of a small delay''
1330 with the small delay not being noticeable on modern systems.
1331 .P
1332 The
1333 .Sw -[no]check
1334 switches were removed together with the support for
1335 .Hd Content-MD5
1336 header fields [RFC\|1864]
1337 (cf. Sec.
1338 .Cf content-md5 )
1339 .Ci 31dc797eb5178970d68962ca8939da3fd9a8efda .
1340 .P
1341 The
1342 .Sw -[no]ebcdicsafe
1343 and
1344 .Sw -[no]rfc934mode
1345 switches of
1346 .Pn mhbuild
1347 were removed because they are considered obsolete
1348 .Ci 01a3480928da485b4d6109d36d751dfa71799d58
1349 .Ci 3363e2624dce0eb8164cf8b3f1ab385c8ff72e88 .
1350 .P
1351 Content caching of external MIME parts, activated with the
1352 .Sw -rcache
1353 and
1354 .Sw -wcache
1355 switches was completely removed
1356 .Ci d1fefd9f614e4dc3cda16da6c69133c1b2005269 .
1357 External MIME parts are rare today, having a caching facility
1358 for them appears to be unnecessary.
1359 .P
1360 In pre-MIME times,
1361 .Pn mhl
1362 had covered many tasks that are part of MIME handling today.
1363 Therefore,
1364 .Pn mhl
1365 could be simplified to a large extend, reducing the number of its
1366 switches from 21 to 6
1367 .Ci 350ad6d3542a07639213cf2a4fe524e829c1e7b6
1368 .Ci 0e46503be3c855bddaeae3843e1b659279c35d70 .
1373 .U3 "Header Printing
1374 .P
1375 .Pn folder 's
1376 data output is self-explaining enough that
1377 displaying the header line makes little sense.
1378 Hence, the
1379 .Sw -[no]header
1380 switch was removed and headers are never printed
1381 .Ci 601cc73d1fa05ce96faa728f036d6c51b91701c7 .
1382 .P
1383 In
1384 .Pn mhlist ,
1385 the
1386 .Sw -[no]header
1387 switches were removed, as well
1388 .Ci b24f96523aaf60e44e04a3ffb1d22e69a13a602f .
1389 In this case, the headers are printed always because the output
1390 is not self-explaining.
1391 .P
1392 .Pn scan
1393 also had
1394 .Sw -[no]header
1395 switches.
1396 Printing this header had been sensible until the introduction of
1397 format strings made it impossible to display column headings.
1398 Only the folder name and the current date remained to be printed.
1399 As this information can be perfectly generated with
1400 .Pn folder
1401 and
1402 .Pn date ,
1403 the switches were removed
1404 .Ci c477dc5d1d03fa6d9a8ab3dd3508c63cbddc044e .
1405 .P
1406 By removing all
1407 .Sw -header
1408 switches, the collision with
1409 .Sw -help
1410 on the first two letters was resolved.
1411 Currently,
1412 .Sw -h
1413 evaluates to
1414 .Sw -help
1415 for all tools of mmh.
1418 .U3 "Suppressing Edits or the Invocation of the WhatNow Shell
1419 .P
1420 The
1421 .Sw -noedit
1422 switch of
1423 .Pn comp ,
1424 .Pn repl ,
1425 .Pn forw ,
1426 .Pn dist ,
1427 and
1428 .Pn whatnow
1429 was removed and replaced by the ability to specify
1430 .Sw -editor
1431 with an empty argument
1432 .Ci 75fca31a5b9d5c1a99c74ab14c94438d8852fba9 .
1433 (Using
1434 .Cl "-editor /bin/true
1435 is nearly the same. It differs only in setting the previous editor.)
1436 .P
1437 The more important change is the removal of the
1438 .Sw -nowhatnowproc
1439 switch
1440 .Ci ee4f43cf2ef0084ec698e4e87159a94c01940622 .
1441 This switch had once introduced an awkward behavior,
1442 as explained in nmh's man page for
1443 .Mp comp (1):
1444 .QS
1445 The
1446 .Sw -editor
1447 .Ar editor
1448 switch indicates the editor to use for
1449 the initial edit. Upon exiting from the editor,
1450 .Pn comp
1451 will invoke the
1452 .Pn whatnow
1453 program. See
1454 .Mp whatnow (1)
1455 for a discussion of available options.
1456 The invocation of this program can be
1457 inhibited by using the
1458 .Sw -nowhatnowproc
1459 switch. (In truth of fact, it is the
1460 .Pn whatnow
1461 program which starts the initial edit.
1462 Hence,
1463 .Sw \%-nowhatnowproc
1464 will prevent any edit from occurring.)
1465 .QE
1466 .P
1467 Effectively, the
1468 .Sw -nowhatnowproc
1469 switch caused only only a draft message to be created.
1470 As
1471 .Cl "-whatnowproc /bin/true
1472 does the same, the
1473 .Sw -nowhatnowproc
1474 switch was removed for being redundant.
1478 .U3 "Various
1479 .BU
1480 With the removal of MMDF maildrop format support,
1481 .Pn packf
1482 and
1483 .Pn rcvpack
1484 no longer needed the
1485 .Sw -mbox
1486 and
1487 .Sw -mmdf
1488 switches.
1489 The behavior of
1490 .Sw -mbox
1491 became the definitive behavior
1492 .Ci 3916ab66ad5d183705ac12357621ea8661afd3c0 .
1493 Further rework in both tools made the
1494 .Sw -file
1495 switch unnecessary
1496 .Ci ca1023716d4c2ab890696f3e41fa0d94267a940e .
1498 .BU
1499 Mmh's tools do no longer clear the screen (\c
1500 .Pn scan 's
1501 and
1502 .Pn mhl 's
1503 .Sw -[no]clear
1504 switches
1505 .Ci e57b17343dcb3ff373ef4dd089fbe778f0c7c270
1506 .Ci 943765e7ac5693ae177fd8d2b5a2440e53ce816e ).
1507 The message formating tool
1508 .Pn mhl
1509 does neither ring the bell (\c
1510 .Sw -[no]bell
1511 .Ci e11983f44e59d8de236affa5b0d0d3067c192e24 )
1512 nor does it page the output itself (\c
1513 .Sw -length
1514 .Ci 5b9d883db0318ed2b84bb82dee880d7381f99188 )
1515 anymore.
1516 Generally, the pager to use is no longer specified with the
1517 .Sw -[no]moreproc
1518 command line switches for
1519 .Pn mhl
1520 and
1521 .Pn show /\c
1522 .Pn mhshow
1523 .Ci 39e87a75b5c2d3572ec72e717720b44af291e88a .
1525 .BU
1526 In order to avoid prefix collisions among switch names, the
1527 .Sw -version
1528 switch was renamed to
1529 .Sw -Version
1530 (with capital `V')
1531 .Ci 32b2354dbaf4bf934936eb5b102a4a3d2fdd209a .
1532 Every program has the
1533 .Sw -version
1534 switch but its first three letters collided with the
1535 .Sw -verbose
1536 switch, present in many programs.
1537 The rename solved this problem once for all.
1538 Although this rename breaks a basic interface, having the
1539 .Sw -V
1540 abbreviation to display the version information, isn't all too bad.
1542 .BU
1543 .Sw -[no]preserve
1544 of
1545 .Pn refile
1546 was removed
1547 .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
1548 because what use was it anyway?
1549 Quoting nmh's man page
1550 .Mp refile (1):
1551 .QS
1552 Normally when a message is refiled, for each destination
1553 folder it is assigned the number which is one above the current
1554 highest message number in that folder. Use of the
1555 .Sw -preserv
1556 [sic!] switch will override this message renaming, and try
1557 to preserve the number of the message. If a conflict for a
1558 particular folder occurs when using the
1559 .Sw -preserve
1560 switch, then
1561 .Pn refile
1562 will use the next available message number which
1563 is above the message number you wish to preserve.
1564 .QE
1566 .BU
1567 The removal of the
1568 .Sw -[no]reverse
1569 switches of
1570 .Pn scan
1571 .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
1572 is a bug fix.
1573 This is supported by the comments
1574 ``\-[no]reverse under #ifdef BERK (I really HATE this)''
1575 by Rose and
1576 ``Lists messages in reverse order with the `\-reverse' switch.
1577 This should be considered a bug'' by Romine in the changelogs.
1578 The question remains why neither Rose nor Romine have fixed this
1579 bug in the eighties when they wrote these comments.
1587 .\" --------------------------------------------------------------
1588 .H1 "Modernizing
1589 .P
1590 In the more than thirty years of MH's existence, its code base was
1591 increasingly extended.
1592 New features entered the project and became alternatives to the
1593 existing behavior.
1594 Relics from several decades have gathered in the code base
1595 but seldom obsolete features were dropped.
1596 This section describes the removing of old code
1597 and the modernizing of the default setup.
1598 It focuses on the functional aspect only;
1599 the non-functional aspects of code style are discussed in Sec.
1600 .Cf code-style .
1603 .H2 "Code Relics
1604 .P
1605 My position regarding the removal of obsolete code
1606 is much more revolutional than the nmh community appreciates.
1607 Working on an experimental version, I was able to quickly drop
1608 functionality that I considered ancient.
1609 The need for consensus with peers would have slowed this process down.
1610 Without the need to justify my decisions, I was able to rush forward.
1611 .P
1612 In December 2011, Paul Vixie motivated the nmh developers to just
1613 do the work:
1614 .[
1615 paul vixie edginess nmh-workers
1616 .]
1617 .QS
1618 let's stop walking on egg shells with this code base. there's no need to
1619 discuss whether to keep using vfork, just note in [sic!] passing, [...]
1620 we don't need a separate branch for removing vmh
1621 or ridding ourselves of #ifdef's or removing posix replacement functions
1622 or depending on pure ansi/posix ``libc''.
1623 .QP
1624 these things should each be a day or two of work and the ``main branch''
1625 should just be modern. [...]
1626 let's push forward, aggressively.
1627 .QE
1628 .LP
1629 I did so already in the months before.
1630 I pushed forward.
1631 I simply dropped the cruft.
1632 .P
1633 The decision to drop a feature was based on literature research and
1634 careful thinking, but whether having had contact with this particular
1635 feature within my own computer life served as a rule of thumb.
1636 I explained my reasons in the commit messages
1637 in the version control system.
1638 Hence, others can comprehend my view and argue for undoing the change
1639 if I have missed an important aspect.
1640 I was quick in dropping parts.
1641 I rather include falsely dropped parts again, than going at a slower pace.
1642 Mmh is experimental work; it requires tough decisions.
1645 .U3 "Process Forking
1646 .P
1647 Being a tool chest, MH creates many processes.
1648 In earlier times
1649 .Fu fork()
1650 had been an expensive system call, because the process's image needed
1651 to be completely duplicated at once.
1652 This expensive work was especially unnecessary in the commonly occurring
1653 case wherein the image is replaced by a call to
1654 .Fu exec()
1655 right after having forked the child process.
1656 The
1657 .Fu vfork()
1658 system call was invented to speed up this particular case.
1659 It completely omits the duplication of the image.
1660 On old systems this resulted in significant speed ups.
1661 Therefore MH used
1662 .Fu vfork()
1663 whenever possible.
1664 .P
1665 Modern memory management units support copy-on-write semantics, which make
1666 .Fu fork()
1667 almost as fast as
1668 .Fu vfork() .
1669 The man page of
1670 .Mp vfork (2)
1671 in FreeBSD 8.0 states:
1672 .QS
1673 This system call will be eliminated when proper system sharing mechanisms
1674 are implemented. Users should not depend on the memory sharing semantics
1675 of vfork() as it will, in that case, be made synonymous to fork(2).
1676 .QE
1677 .LP
1678 Vixie supports the removal with the note that ``the last
1679 system on which fork was so slow that an mh user would notice it, was
1680 Eunice. that was 1987''.
1681 .[
1682 nmh-workers vixie edginess
1683 .]
1684 I replaced all calls to
1685 .Fu vfork()
1686 with calls to
1687 .Fu fork()
1688 .Ci 40821f5c1316e9205a08375e7075909cc9968e7d .
1689 .P
1690 Related to the costs of
1691 .Fu fork()
1692 is the probability of its success.
1693 In the eighties, on heavy loaded systems, calls to
1694 .Fu fork()
1695 were prone to failure.
1696 Hence, many of the
1697 .Fu fork()
1698 calls in the code were wrapped into loops to retry the
1699 .Fu fork()
1700 several times, to increase the chances to succeed eventually.
1701 On modern systems, a failing
1702 .Fu fork()
1703 call is unusual.
1704 Hence, in the rare case when
1705 .Fu fork()
1706 fails, mmh programs simply abort
1707 .Ci 5fbf37ee68e018998ada61eeab73e035b26834b6 .
1710 .U3 "Header Fields
1711 .BU
1712 The
1713 .Hd Encrypted
1714 header field was introduced by RFC\|822,
1715 but already marked as legacy in RFC\|2822.
1716 Today, OpenPGP provides the basis for standardized exchange of encrypted
1717 messages [RFC\|4880, RFC\|3156].
1718 Hence, the support for
1719 .Hd Encrypted
1720 header fields is removed in mmh
1721 .Ci 064527f7b57ab050e5af13e15ad99aeeab125857 .
1722 .BU
1723 The native support for
1724 .Hd Face
1725 header fields has been removed, as well
1726 .Ci 8e5be81f784682822f5e868c1bf3c8624682bd23 .
1727 This feature is similar to the
1728 .Hd X-Face
1729 header field in its intent,
1730 but takes a different approach to store the image.
1731 Instead of encoding the image data directly into the header field,
1732 it contains the hostname and UDP port where the image
1733 date can be retrieved.
1734 There is even a third Face system,
1735 which is the successor of
1736 .Hd X-Face ,
1737 although it re-uses the
1738 .Hd Face
1739 header field name.
1740 It was invented in 2005 and supports colored PNG images.
1741 None of the Face systems described here is popular today.
1742 Hence, mmh has no direct support for them.
1743 .BU
1744 .Id content-md5
1745 The
1746 .Hd Content-MD5
1747 header field was introduced by RFC\|1864.
1748 It provides detection of data corruption during the transfer.
1749 But it can not ensure verbatim end-to-end delivery of the contents
1750 [RFC\|1864].
1751 The proper approach to verify content integrity in an
1752 end-to-end relationship is the use of digital signatures [RFC\|4880].
1753 On the other hand, transfer protocols should detect corruption during
1754 the transmission.
1755 The TCP includes a checksum field therefore.
1756 These two approaches in combinations render the
1757 .Hd Content-MD5
1758 header field superfluous.
1759 Not a single one out of 4\|200 messages from two decades
1760 in the nmh-workers mailing list archive
1761 .[
1762 nmh-workers mailing list archive website
1763 .]
1764 contained a
1765 .Hd Content-MD5
1766 header field.
1767 Neither did any of the 60\|000 messages in my personal mail storage.
1768 Removing the support for this header field
1769 .Ci 31dc797eb5178970d68962ca8939da3fd9a8efda ,
1770 removed the last place where MD5 computation was needed.
1771 Hence, the MD5 code could be removed as well.
1772 Over 500 lines of code vanished by this one change.
1775 .U3 "MMDF maildrop support
1776 .P
1777 This type of maildrop format is conceptionally similar to the mbox format,
1778 but uses a different message delimiter (`\fL\\1\\1\\1\\1\fP',
1779 commonly written as `\fL^A^A^A^A\fP', instead of `\fLFrom\0\fP').
1780 Mbox is the de-facto standard maildrop format on Unix,
1781 whereas the MMDF maildrop format is now forgotten.
1782 Mbox remains as the only packed mailbox format, supported in mmh.
1783 .P
1784 The simplifications within the code were moderate.
1785 Mainly, the reading and writing of MMDF mailbox files was removed.
1786 But also, switches of
1787 .Pn packf
1788 and
1789 .Pn rcvpack
1790 could be removed
1791 .Ci 3916ab66ad5d183705ac12357621ea8661afd3c0 .
1792 In the message parsing function
1793 .Fn sbr/m_getfld.c ,
1794 knowledge of MMDF packed mail boxes was removed
1795 .Ci 684ec30d81e1223a282764452f4902ed4ad1c754 .
1796 Further code structure simplifications may be possible there,
1797 because only one single packed mailbox format is left to be supported.
1798 I have not worked on them yet because
1799 .Fu m_getfld()
1800 is heavily optimized and thus dangerous to touch.
1801 The risk of damaging the intricate workings of the optimized code is
1802 too high.
1805 .U3 "Prompter's Control Keys
1806 .P
1807 The program
1808 .Pn prompter
1809 queries the user to fill in a message form.
1810 When used as
1811 .Cl "comp -editor prompter" ,
1812 the resulting behavior is similar to
1813 .Pn mailx .
1814 Apparently,
1815 .Pn prompter
1816 had not been touched lately.
1817 Otherwise it's hardly explainable why it
1818 still offered the switches
1819 .Sw -erase
1820 .Ar chr
1821 and
1822 .Sw -kill
1823 .Ar chr
1824 to name the characters for command line editing.
1825 The times when this had been necessary are long time gone.
1826 Today these things work out-of-the-box, and if not, are configured
1827 with the standard tool
1828 .Pn stty .
1829 The switches are removed now
1830 .Ci 0bd9750710cdbab80cfb4036dd87af20afe1552f .
1833 .U3 "Hardcopy Terminal Support
1834 .P
1835 More of a funny anecdote is a check for being connected to a
1836 hardcopy terminal.
1837 It remained in the code until spring 2012, when I finally removed it
1838 .Ci b7764c4a6b71d37918a97594d866258f154017ca .
1839 The check only prevented a pager to be placed between the printing
1840 program (\c
1841 .Pn mhl )
1842 and the terminal.
1843 In nmh, this could have been ensured statically with the
1844 .Sw -nomoreproc
1845 at the command line, too.
1846 In mmh, setting the profile entry
1847 .Pe Pager
1848 or the environment variable
1849 .Ev PAGER
1850 to
1851 .Pn cat
1852 is sufficient.
1857 .H2 "Attachments
1858 .P
1859 The mind model of email attachments is unrelated to MIME.
1860 Although the MIME RFCs [RFC\|2045\(en2049] define the technical
1861 requirements for having attachments, they do not mention the term.
1862 Instead of attachments, MIME talks about ``multi-part message bodies''
1863 [RFC\|2045], a more general concept.
1864 Multi-part messages are messages
1865 ``in which one or more different
1866 sets of data are combined in a single body''
1867 [RFC\|2046].
1868 MIME keeps its descriptions generic;
1869 it does not imply specific usage models.
1870 Today, one usage model is prevalent: attachments.
1871 The idea is having a main text document with files of arbitrary kind
1872 attached to it.
1873 In MIME terms, this is a multi-part message having a text part first
1874 and parts of arbitrary type following.
1875 .P
1876 .ZZ
1877 MH's MIME support is a direct implementation of the RFCs.
1878 The perception of the topic described in the RFCs is clearly visible
1879 in MH's implementation.
1880 As a result,
1881 MH had all the MIME features but no idea of attachments.
1882 But users do not need all the MIME features,
1883 they want convenient attachment handling.
1886 .U3 "Composing MIME Messages
1887 .P
1888 In order to improve the situation on the message composing side,
1889 Jon Steinhart had added an attachment system to nmh in 2002
1890 .Ci 7480dbc14bc90f2d872d434205c0784704213252 .
1891 In the file
1892 .Fn docs/README-ATTACHMENTS ,
1893 he described his motivation to do so:
1894 .QS
1895 Although nmh contains the necessary functionality for MIME message
1896 handing [sic!], the interface to this functionality is pretty obtuse.
1897 There's no way that I'm ever going to convince my partner to write
1898 .Pn mhbuild
1899 composition files!
1900 .QE
1901 .LP
1902 With this change, the mind model of attachments entered nmh.
1903 In the same document:
1904 .QS
1905 These changes simplify the task of managing attachments on draft files.
1906 They allow attachments to be added, listed, and deleted.
1907 MIME messages are automatically created when drafts with attachments
1908 are sent.
1909 .QE
1910 .LP
1911 Unfortunately, the attachment system, like every new facilities in nmh,
1912 was inactive by default.
1913 .P
1914 During my time in Argentina, I tried to improve the attachment system.
1915 But, after long discussions my patch died as a proposal on the
1916 mailing list because of great opposition in the nmh community.
1917 .[
1918 nmh-workers attachment proposal
1919 .]
1920 In January 2012, I extended the patch and applied it to mmh
1921 .Ci 8ff284ff9167eff8f5349481529332d59ed913b1 .
1922 In mmh, the attachment system is active by default.
1923 Instead of command line switches, the
1924 .Pe Attachment-Header
1925 profile entry is used to specify
1926 the name of the attachment header field.
1927 It is pre-defined to
1928 .Hd Attach .
1929 .P
1930 To add an attachment to a draft, a header line needs to be added:
1931 .VS
1932 To: bob
1933 Subject: The file you wanted
1934 Attach: /path/to/the/file-bob-wanted
1935 --------
1936 Here it is.
1937 VE
1938 The header field can be added to the draft manually in the editor,
1939 or by using the `attach' command at the WhatNow prompt, or
1940 non-interactively with
1941 .Pn anno :
1942 .VS
1943 anno -append -nodate -component Attach -text /path/to/attachment
1944 VE
1945 Drafts with attachment headers are converted to MIME automatically by
1946 .Pn send .
1947 The conversion to MIME is invisible to the user.
1948 The draft stored in the draft folder is always in source form with
1949 attachment headers.
1950 If the MIMEification fails (e.g. because the file to attach
1951 is not accessible) the original draft is not changed.
1952 .P
1953 The attachment system handles the forwarding of messages, too.
1954 If the attachment header value starts with a plus character (`\fL+\fP'),
1955 like in
1956 .Cl "Attach: +bob 30 42" ,
1957 the given messages in the specified folder will be attached.
1958 This allowed to simplify
1959 .Pn forw
1960 .Ci f41f04cf4ceca7355232cf7413e59afafccc9550 .
1961 .P
1962 Closely related to attachments is non-ASCII text content,
1963 because it requires MIME as well.
1964 In nmh, the user needed to call `mime' at the WhatNow prompt
1965 to have the draft converted to MIME.
1966 This was necessary whenever the draft contained non-ASCII characters.
1967 If the user did not call `mime', a broken message would be sent.
1968 Therefore, the
1969 .Pe automimeproc
1970 profile entry could be specified to have the `mime' command invoked
1971 automatically each time.
1972 Unfortunately, this approach conflicted with the attachment system
1973 because the draft would already be in MIME format at the time
1974 when the attachment system wanted to MIMEify it.
1975 To use nmh's attachment system, `mime' must not be called at the
1976 WhatNow prompt and
1977 .Pe automimeproc
1978 must not be set in the profile.
1979 But then the case of non-ASCII text without attachment headers was
1980 not caught.
1981 All in all, the solution was complex and irritating.
1982 My patch from December 2010
1983 .[
1984 nmh-workers attachment proposal
1985 .]
1986 would have simplified the situation.
1987 .P
1988 Mmh's current solution is even more elaborate.
1989 Any necessary MIMEification is done automatically.
1990 There is no `mime' command at the WhatNow prompt anymore.
1991 The draft will be converted automatically to MIME when either an
1992 attachment header or non-ASCII text is present.
1993 Furthermore, the hash character (`\fL#\fP') is not special any more
1994 at line beginnings in the draft message.
1995 Users need not concern themselves with the whole topic at all.
1996 The approach taken in mmh is tailored towards today's most common case:
1997 a text part, possibly with attachments.
1998 This case was simplified.
1999 .P
2000 Although the new approach does not anymore support arbitrary MIME
2001 compositions directly, the full power of
2002 .Pn mhbuild
2003 can still be accessed.
2004 Given no attachment headers are included, users can create
2005 .Pn mhbuild
2006 composition drafts like in nmh.
2007 Then, at the WhatNow prompt, they can invoke
2008 .Cl "edit mhbuild
2009 to convert the draft to MIME.
2010 Because the resulting draft neither contains non-ASCII characters
2011 nor has it attachment headers, the attachment system will not touch it.
2014 .U3 "MIME Type Guessing
2015 .P
2016 From the programmer's point of view, the use of
2017 .Pn mhbuild
2018 composition drafts had one notable advantage over attachment headers:
2019 The user provides the appropriate MIME types for files to include.
2020 The new attachment system needs to find out the correct MIME type itself.
2021 This is a difficult task.
2022 Determining the correct MIME type of content is partly mechanical,
2023 partly intelligent work.
2024 Forcing the user to find out the correct MIME type,
2025 forces him to do partly mechanical work.
2026 Letting the computer do the work can lead to bad choices for difficult
2027 content.
2028 For mmh, the latter option was chosen to spare the user the work
2029 .Ci 3baec236a39c5c89a9bda8dbd988d643a21decc6 .
2030 .P
2031 Determining the MIME type by the suffix of the file name is a dumb
2032 approach, yet it is simple to implement and provides good results
2033 for the common cases.
2034 If no MIME type can be determined, text content is sent as `text/plain',
2035 anything else under the generic fall-back type `application/octet-stream'.
2036 Mmh implements this approach in the
2037 .Pn print-mimetype
2038 script
2039 .Ci 4b5944268ea0da7bb30598a27857304758ea9b44 .
2040 .P
2041 A far better, though less portable, approach is the use of
2042 .Pn file .
2043 This standard tool tries to determine the type of files.
2044 Unfortunately, its capabilities and accuracy varies from system to system.
2045 Additionally, its output was only intended for human beings,
2046 but not to be used by programs.
2047 Nevertheless, modern versions of GNU
2048 .Pn file ,
2049 which are prevalent on the popular GNU/Linux systems,
2050 provide MIME type output in machine-readable form.
2051 Although this solution is system-dependent,
2052 it solves the difficult problem well.
2053 On systems where GNU
2054 .Pn file ,
2055 version 5.04 or higher, is available it should be used.
2056 One needs to specify the following profile entry to do so:
2057 .VS
2058 Mime-Type-Query: file -b --mime
2059 VE
2060 .LP
2061 Other versions of
2062 .Pn file
2063 might possibly be usable with wrapper scripts that reformat the output.
2064 The diversity among
2065 .Pn file
2066 implementations is great; one needs to check the local variant.
2067 .P
2068 It is not possible in mmh to override the automatic MIME type guessing
2069 for a specific file.
2070 To do so, either the user would need to know in advance for which file
2071 the automatic guessing fails or the system would require interaction.
2072 I consider both cases impractical.
2073 The existing solution should be sufficient.
2074 If not, the user may always fall back to
2075 .Pn mhbuild
2076 composition drafts and bypass the attachment system.
2079 .U3 "Storing Attachments
2080 .P
2081 Extracting MIME parts of a message and storing them to disk is performed by
2082 .Pn mhstore .
2083 The program has two operation modes,
2084 .Sw -auto
2085 and
2086 .Sw -noauto .
2087 With the former one, each part is stored under the filename given in the
2088 MIME part's meta information, if available.
2089 This naming information is usually available for modern attachments.
2090 If no filename is available, this MIME part is stored as if
2091 .Sw -noauto
2092 would have been specified.
2093 In the
2094 .Sw -noauto
2095 mode, the parts are processed according to the rules that are defined by
2096 .Pe mhstore-store-*
2097 profile entries.
2098 These rules define generic filename templates for storing
2099 or commands to post-process the contents in arbitrary ways.
2100 If no matching rule is available the part is stored under a generic
2101 filename, built from message number, MIME part number, and MIME type.
2102 .P
2103 The
2104 .Sw -noauto
2105 mode had been the default in nmh because it was considered safe,
2106 in contrast to the
2107 .Sw -auto
2108 mode.
2109 In mmh,
2110 .Sw -auto
2111 is not dangerous anymore.
2112 Two changes were necessary:
2113 .LI 1
2114 Any directory path is removed from the proposed filename.
2115 Thus, the files are always stored in the expected directory.
2116 .Ci 41b6eadbcecf63c9a66aa5e582011987494abefb
2117 .LI 2
2118 Tar files are not extracted automatically any more.
2119 Thus, the rest of the file system will not be touched.
2120 .Ci 94c80042eae3383c812d9552089953f9846b1bb6
2121 .P
2122 In mmh, the result of
2123 .Cl "mhstore -auto
2124 can be foreseen from the output of
2125 .Cl "mhlist -verbose" .
2126 Although the
2127 .Sw -noauto
2128 mode is considered to be more powerful, it is less convenient and
2129 .Sw -auto
2130 is safe now.
2131 Additionally, storing attachments under their original name
2132 is intuitive.
2133 Hence,
2134 .Sw -auto
2135 serves better as the default option
2136 .Ci 3410b680416c49a7617491af38bc1929855a331d .
2137 .P
2138 Files are stored into the directory given by the
2139 .Pe Nmh-Storage
2140 profile entry, if set, or
2141 into the current working directory, otherwise.
2142 Storing to different directories is only possible with
2143 .Pe mhstore-store-*
2144 profile entries.
2145 .P
2146 Still existing files get overwritten silently in both modes.
2147 This can be considered a bug.
2148 Yet, each other behavior has its draw-backs, too.
2149 Refusing to replace files requires adding a
2150 .Sw -force
2151 switch.
2152 Users will likely need to invoke
2153 .Pn mhstore
2154 a second time with
2155 .Sw -force .
2156 Eventually, only the user can decide in the specific case.
2157 This requires interaction, which I like to avoid if possible.
2158 Appending a unique suffix to the filename is another bad option.
2159 For now, the behavior remains as it is.
2160 .P
2161 In mmh, only MIME parts of type message are special in
2162 .Pn mhstore 's
2163 .Sw -auto
2164 mode.
2165 Instead of storing message/rfc822 parts as files to disk,
2166 they are stored as messages into the current mail folder.
2167 The same applies to message/partial, although the parts are
2168 automatically reassembled beforehand.
2169 MIME parts of type message/external-body are not automatically retrieved
2170 anymore.
2171 Instead, information on how to retrieve them is output.
2172 Not supporting this rare case saved nearly one thousand lines of code
2173 .Ci 55e1d8c654ee0f7c45b9361ce34617983b454c32 .
2174 The MIME type `application/octet-stream; type=tar' is not special anymore.
2175 The automatically extracting of such MIME parts had been the
2176 dangerous part of the
2177 .Sw -auto
2178 mode
2179 .Ci 94c80042eae3383c812d9552089953f9846b1bb6 .
2183 .U3 "Showing MIME Messages
2184 .Id showing-mime-msgs
2185 .P
2186 The program
2187 .Pn mhshow
2188 was written to display MIME messages.
2189 It implemented the conceptional view of the MIME RFCs.
2190 Nmh's
2191 .Pn mhshow
2192 handles each MIME part independently, presenting them separately
2193 to the user.
2194 This does not match today's understanding of email attachments,
2195 where displaying a message is seen to be a single, integrated operation.
2196 Today, email messages are expected to consist of a main text part
2197 plus possibly attachments.
2198 They are no more seen to be arbitrary MIME hierarchies with
2199 information on how to display the individual parts.
2200 I adjusted
2201 .Pn mhshow 's
2202 behavior to the modern view on the topic.
2203 .P
2204 One should note that this section completely ignores the original
2205 .Pn show
2206 program, because it was not capable to display MIME messages
2207 and is no longer part of mmh
2208 (cf. Sec.
2209 .Cf mhshow ).
2210 Although
2211 .Pn mhshow
2212 was renamed to
2213 .Pn show
2214 in mmh, this section uses the name
2215 .Pn mhshow ,
2216 in order to avoid confusion.
2217 .P
2218 In mmh, the basic idea is that
2219 .Pn mhshow
2220 should display a message in one single pager session.
2221 Therefore,
2222 .Pn mhshow
2223 invokes a pager session for all its output,
2224 whenever it prints to a terminal
2225 .Ci a4197ea6ffc5c1550e8b52d5a654bcaaaee04a4e .
2226 In consequence,
2227 .Pn mhl
2228 does no more invoke a pager
2229 .Ci 0e46503be3c855bddaeae3843e1b659279c35d70 .
2230 With
2231 .Pn mhshow
2232 replacing the original
2233 .Pn show ,
2234 the output of
2235 .Pn mhl
2236 no longer goes to the terminal directly, but through
2237 .Pn mhshow .
2238 Hence,
2239 .Pn mhl
2240 does not need to invoke a pager.
2241 The one and only job of
2242 .Pn mhl
2243 is to format messages or parts of them.
2244 The only place in mmh, where a pager is invoked is
2245 .Pn mhshow .
2246 .P
2247 Only text content is displayed.
2248 Other kinds of attachments are ignored.
2249 Non-text content needs to be converted to text by appropriate
2250 .Pe mhshow-show-*
2251 profile entries before, if this is possible and wanted.
2252 A common example for this are PDF files.
2253 .P
2254 MIME parts are always displayed serially.
2255 The request to display the MIME type `multipart/parallel' in parallel
2256 is ignored.
2257 It is simply treated as `multipart/mixed'
2258 .Ci d0581ba306a7299113a346f9b4c46ce97bc4cef6 .
2259 This was already possible to request with the
2260 .Sw -serialonly
2261 switch of
2262 .Pn mhshow ,
2263 which is now removed.
2264 As MIME parts are always processed exclusively, i.e. serially,
2265 the `\fL%e\fP' escape in
2266 .Pe mhshow-show-*
2267 profile entries became useless and was thus removed
2268 .Ci a20d405db09b7ccca74d3e8c57550883da49e1ae .
2269 For parallel display, the attachments need to be stored to disk first.
2270 .P
2271 To display text content in foreign charsets, they need to be converted
2272 to the native charset.
2273 Therefore,
2274 .Pe mhshow-charset-*
2275 profile entries were needed.
2276 In mmh, the conversion is performed automatically by piping the
2277 text through the
2278 .Pn iconv
2279 command, if necessary
2280 .Ci 2433122c20baccb10b70b49c04c6b0497b5b3b60 .
2281 Custom
2282 .Pe mhshow-show-*
2283 rules for textual content might need a
2284 .Cl "iconv -f %c %f |
2285 prefix to have the text converted to the native charset.
2286 .P
2287 Although the conversion of foreign charsets to the native one
2288 has improved, it is not consistent enough.
2289 Further work needs to be done and
2290 the basic concepts in this field need to be re-thought.
2291 Though, the default setup of mmh displays message in foreign charsets
2292 correctly without the need to configure anything.
2295 .ig
2297 .P
2298 mhshow/mhstore: Removed support for retrieving message/external-body parts.
2299 These tools will not download the contents automatically anymore. Instead,
2300 they print the information needed to get the contents. If someone should
2301 really receive one of those rare message/external-body messages, he can
2302 do the job manually. We save nearly a thousand lines of code. That's worth
2303 it!
2304 (The profile entry `nmh-access-ftp' and sbr/ruserpass.c for reading
2305 ~/.netrc are gone now.)
2306 .Ci 55e1d8c654ee0f7c45b9361ce34617983b454c32
2308 ..
2312 .H2 "Signing and Encrypting
2313 .P
2314 Nmh offers no direct support for digital signatures and message encryption.
2315 This functionality needed to be added through third-party software.
2316 In mmh, the functionality is included because it
2317 is a part of modern email and is likely wanted by users of mmh.
2318 A fresh mmh installation supports signing and encrypting
2319 out-of-the-box.
2320 Therefore, Neil Rickert's
2321 .Pn mhsign
2322 and
2323 .Pn mhpgp
2324 scripts
2325 .[
2326 neil rickert mhsign mhpgp
2327 .]
2328 were included
2329 .Ci f45cdc98117a84f071759462c7ae212f4bc5ab2e
2330 .Ci 58cf09aa36e9f7f352a127158bbf1c5678bc6ed8 .
2331 The scripts fit well because they are lightweight and
2332 similar of style to the existing tools.
2333 Additionally, no licensing difficulties appeared
2334 as they are part of the public domain.
2335 .P
2336 .Pn mhsign
2337 handles the signing and encrypting part.
2338 It comprises about 250 lines of shell code and interfaces between
2339 .Pn gnupg
2340 and the MH system.
2341 It was meant to be invoked manually at the WhatNow prompt, but in mmh,
2342 .Pn send
2343 invokes
2344 .Pn mhsign
2345 automatically
2346 .Ci c7b5e1df086bcc37ff40163ee67571f076cf6683 .
2347 Special header fields were introduced to request this action.
2348 If a draft contains the
2349 .Hd Sign
2350 header field,
2351 .Pn send
2352 will initiate the signing.
2353 The signing key is either chosen automatically or it is specified by the
2354 .Pe Pgpkey
2355 profile entry.
2356 .Pn send
2357 always create signatures using the PGP/MIME standard [RFC\|4880],
2358 but by invoking
2359 .Pn mhsign
2360 manually, old-style non-MIME signatures can be created as well.
2361 To encrypt an outgoing message, the draft needs to contain an
2362 .Hd Enc
2363 header field.
2364 Public keys of all recipients are searched for in the gnupg keyring and
2365 in a file called
2366 .Fn pgpkeys ,
2367 which contains exceptions and overrides.
2368 Unless public keys are found for all recipients,
2369 .Pn mhsign
2370 will refuse to encrypt it.
2371 Currently, messages with hidden (BCC) recipients can not be encrypted.
2372 This work is pending because it requires a structurally more complex
2373 approach.
2374 .P
2375 .Pn mhpgp
2376 is the companion to
2377 .Pn mhsign .
2378 It verifies signatures and decrypts messages.
2379 Encrypted messages can be either temporarily decrypted and displayed
2380 or permanently decrypted and stored into the current folder.
2381 Currently,
2382 .Pn mhpgp
2383 needs to be invoked manually.
2384 The integration into
2385 .Pn show
2386 and
2387 .Pn mhstore
2388 to verify signatures and decrypt messages as needed
2389 is planned but not yet realized.
2390 .P
2391 Both scripts were written for nmh.
2392 Hence they needed to be adjust
2393 according to the differences between nmh and mmh.
2394 For instance, they use the backup prefix no longer.
2395 Furthermore, compatibility support for old PGP features was dropped.
2396 .P
2397 The integrated message signing and encrypting support is one of the
2398 most recent features in mmh.
2399 It has not had the time to mature.
2400 User feedback and personal experience need to be accumulated to
2401 direct the further development of the facility.
2402 Already it seems to be worthwhile to consider adding
2403 .Sw -[no]sign
2404 and
2405 .Sw -[no]enc
2406 switches to
2407 .Pn send ,
2408 to be able to override the corresponding header fields.
2409 A profile entry:
2410 .VS
2411 send: -sign
2412 VE
2413 would then activate signing for all outgoing messages.
2414 With the present approach, a
2415 .Hd Send
2416 header component needs to be added to each draft template
2417 to achieve the same result.
2418 Adding the switches would ease the work greatly and keep the
2419 template files clean.
2424 .H2 "Draft and Trash Folder
2425 .P
2427 .U3 "Draft Folder
2428 .Id draft-folder
2429 .P
2430 In the beginning, MH had the concept of a draft message.
2431 This was a file named
2432 .Fn draft
2433 in the MH directory, which was treated special.
2434 On composing a message, this draft file was used.
2435 When starting to compose another message before the former one was sent,
2436 the user had to decide among:
2437 .LI 1
2438 Using the old draft to finish and send it before starting with a new one.
2439 .LI 2
2440 Discarding the old draft and replacing it with a new one.
2441 .LI 3
2442 Preserving the old draft by refiling it to a folder.
2443 .LP
2444 Working on multiple drafts was only possible in alternation.
2445 For that, the current draft needed to be refiled to a folder and
2446 another one re-used for editing.
2447 Working on multiple drafts at the same time was impossible.
2448 The usual approach of switching to a different MH context did not
2449 help anything.
2450 .P
2451 The draft folder facility exists to
2452 allow true parallel editing of drafts, in a straight forward way.
2453 It was introduced by Marshall T. Rose, already in 1984.
2454 Similar to other new features, the draft folder was inactive by default.
2455 Even in nmh, the highly useful draft folder was not available
2456 out-of-the-box.
2457 At least, Richard Coleman added the man page
2458 .Mp mh-draft (5)
2459 to better document the feature.
2460 .P
2461 Not using the draft folder facility has the single advantage of having
2462 the draft file at a static location.
2463 This is simple in simple cases but the concept does not scale for more
2464 complex cases.
2465 The concept of the draft message is too limited for the problem
2466 it tries to solve.
2467 Therefore the draft folder was introduced.
2468 It is the more powerful and more natural concept.
2469 The draft folder is a folder like any other folder in MH.
2470 Its messages can be listed like any other messages.
2471 A draft message is no longer a special case.
2472 Tools do not need special switches to work on the draft message.
2473 Hence corner cases were removed.
2474 .P
2475 The trivial part of the work was activating the draft folder with a
2476 default name.
2477 I chose the name
2478 .Fn +drafts ,
2479 for obvious reasons.
2480 In consequence, the command line switches
2481 .Sw -draftfolder
2482 and
2483 .Sw -draftmessage
2484 could be removed.
2485 More difficult, but also more improving, was the updating of the
2486 tools to the new concept.
2487 By fully switching to the draft folder, the tools could be
2488 simplified by dropping the awkward draft message handling code.
2489 .Sw -draft
2490 switches were removed because operating on a draft message is no longer
2491 special.
2492 It became indistinguishable to operating on any other message.
2493 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860
2494 .P
2495 There is no more need to query the user for draft handling
2496 .Ci 2d48b455c303a807041c35e4248955f8bec59eeb .
2497 It is always possible to add another new draft.
2498 Refiling drafts is without difference to refiling other messages.
2499 All of these special cases are gone.
2500 Yet, one draft-related switch remained.
2501 .Pn comp
2502 still has
2503 .Sw -[no]use
2504 for switching between two modes:
2505 .LI 1
2506 Modifying an existing draft, with
2507 .Sw -use .
2508 .LI 2
2509 Composing a new draft, possibly taking some existing message as template,
2510 with
2511 .Sw -nouse ,
2512 the default.
2513 .ZZ
2514 .RT
2515 .sp \n(PDu
2516 In either case, the behavior of
2517 .Pn comp
2518 is deterministic.
2519 .P
2520 .Pn send
2521 now operates on the current message in the draft folder by default.
2522 As message and folder can both be overridden by specifying them on
2523 the command line, it is possible to send any message in the mail storage
2524 by simply specifying its number and folder.
2525 In contrast to the other tools,
2526 .Pn send
2527 takes the draft folder as its default folder.
2528 .P
2529 Dropping the draft message concept in favor for the draft folder concept,
2530 replaced special cases with regular cases.
2531 This simplified the source code of the tools, as well as the concepts.
2532 In mmh, draft management does not break with the MH concepts
2533 but applies them.
2534 .Cl "scan +drafts" ,
2535 for instance, is a truly natural request.
2536 .P
2537 Most of the work was already performed by Rose in the eighties.
2538 The original improvement of mmh is dropping the old draft message approach
2539 and thus simplifying the tools, the documentation,
2540 and the system as a whole.
2541 Although my part in the draft handling improvement was small,
2542 it was important.
2545 .U3 "Trash Folder
2546 .Id trash-folder
2547 .P
2548 Similar to the situation for drafts is the situation for removed messages.
2549 Historically, a message was ``deleted'' by prepending a specific
2550 \fIbackup prefix\fP, usually the comma character,
2551 to the file name.
2552 The specific file would then be ignored by MH because only files with
2553 names consisting of digits only are treated as messages.
2554 Although files remained in the file system,
2555 the messages were no longer visible in MH.
2556 To truly delete them, a maintenance job was needed.
2557 Usually a cron job was installed to delete them after a grace time.
2558 For instance:
2559 .VS
2560 find $HOME/Mail -type f -name ',*' -ctime +7 -delete
2561 VE
2562 In such a setup, the original message could be restored
2563 within the grace time interval by stripping the
2564 backup prefix from the file name \(en usually but not always.
2565 If the last message of a folder with six messages (\fL1-6\fP) was removed,
2566 message
2567 .Fn 6 ,
2568 became file
2569 .Fn ,6 .
2570 If then a new message entered the same folder, it would be named with
2571 the number one above the highest existing message number.
2572 In this case the message would be named
2573 .Fn 6 ,
2574 reusing the number.
2575 If this new message would be removed as well,
2576 then the backup of the former message becomes overwritten.
2577 Hence, the ability to restore removed messages did not only depend on
2578 the sweeping cron job but also on the removing of further messages.
2579 It is undesirable to have such obscure and complex mechanisms.
2580 The user should be given a small set of clear assertions, such as
2581 ``Removed files are restorable within a seven-day grace time.''
2582 With the addition ``... unless a message with the same name in the
2583 same folder is removed before.'' the statement becomes complex.
2584 A user will hardly be able to keep track of all removals to know
2585 if the assertion still holds true for a specific file.
2586 In practice, the real mechanism is unclear to the user.
2587 .P
2588 Furthermore, the backup files were scattered within the whole mail storage.
2589 This complicated managing them.
2590 It was possible with the help of
2591 .Pn find ,
2592 but everything is more convenient
2593 if the deleted messages are collected in one place.
2594 .P
2595 The profile entry
2596 .Pe rmmproc
2597 (previously named
2598 .Pe Delete-Prog )
2599 was introduced very early to improve the situation.
2600 It could be set to any command, which would be executed to remove
2601 the specified messages.
2602 This had overridden the default action, described above.
2603 Refiling the to-be-removed files to a trash folder was the usual example.
2604 Nmh's man page
2605 .Mp rmm (1)
2606 proposes to set the
2607 .Pe rmmproc
2608 to
2609 .Cl "refile +d
2610 to move messages to the trash folder
2611 .Fn +d
2612 instead of renaming them with the backup prefix.
2613 The man page additionally proposes the expunge command
2614 .Cl "rm `mhpath +d all`
2615 to empty the trash folder.
2616 .P
2617 Removing messages in such a way has advantages:
2618 .LI 1
2619 The mail storage is prevented from being cluttered with removed messages
2620 because they are all collected in one place.
2621 Existing and removed messages are thus separated more strictly.
2622 .LI 2
2623 No backup files are silently overwritten.
2624 .LI 3
2625 Most important, however, removed messages are kept in the MH domain.
2626 Messages in the trash folder can be listed like those in any other folder.
2627 Deleted messages can be displayed like any other messages.
2628 .Pn refile
2629 can restore deleted messages.
2630 All operations on deleted files are still covered by the MH tools.
2631 The trash folder is just like any other folder in the mail storage.
2632 .P
2633 Similar to the draft folder case, I dropped the old backup prefix approach
2634 in favor for replacing it by the better suiting trash folder system.
2635 Hence,
2636 .Pn rmm
2637 calls
2638 .Pn refile
2639 to move the to-be-removed message to the trash folder,
2640 .Fn +trash
2641 by default.
2642 To sweep it clean, the user can use
2643 .Cl "rmm -unlink +trash a" ,
2644 where the
2645 .Sw -unlink
2646 switch causes the files to be unlinked.
2647 .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
2648 .Ci ca0b3e830b86700d9e5e31b1784de2bdcaf58fc5
2649 .P
2650 Dropping the legacy approach and converting to the new approach
2651 completely, simplified the code base.
2652 The relationship between
2653 .Pn rmm
2654 and
2655 .Pn refile
2656 was inverted.
2657 In mmh,
2658 .Pn rmm
2659 invokes
2660 .Pn refile .
2661 That used to be the other way round.
2662 Yet, the relationship is simpler now.
2663 Loops, like described in nmh's man page for
2664 .Mp refile (1),
2665 can no longer occur:
2666 .QS
2667 Since
2668 .Pn refile
2669 uses your
2670 .Pe rmmproc
2671 to delete the message, the
2672 .Pe rmmproc
2673 must NOT call
2674 .Pn refile
2675 without specifying
2676 .Sw -normmproc
2677 or you will create an infinite loop.
2678 .QE
2679 .LP
2680 .Pn rmm
2681 either unlinks a message with
2682 .Fu unlink()
2683 or invokes
2684 .Pn refile
2685 to move it to the trash folder.
2686 .Pn refile
2687 does not invoke any tools.
2688 .P
2689 By generalizing the message removal in the way that it became covered
2690 by the MH concepts made the whole system more powerful.
2696 .H2 "Modern Defaults
2697 .P
2698 Nmh has a bunch of convenience-improving features inactive by default,
2699 although one can expect every new user to want them active.
2700 The reason they are inactive by default is the wish to stay compatible
2701 with old versions.
2702 But what are old versions?
2703 Still, the highly useful draft folder facility has not been activated
2704 by default although it was introduced over twenty-five years ago.
2705 .[
2706 rose romine real work
2707 .]
2708 The community seems not to care.
2709 .P
2710 In nmh, new users are required to first build up
2711 a profile before they can access the modern features.
2712 Without an extensive profile, the setup is hardly usable
2713 for modern emailing.
2714 The point is not the customization of the setup,
2715 but the need to activate generally useful facilities.
2716 Yet, the real problem lies less in enabling the features,
2717 as this is straight forward as soon as one knows what he wants.
2718 The real problem is that new users need deep insight into the project
2719 to discover the available but inactive features.
2720 To give an example, I needed one year of using nmh
2721 before I became aware of the existence of the attachment system.
2722 One could argue that this fact disqualifies my reading of the
2723 documentation.
2724 If I would have installed nmh from source back then, I could agree.
2725 Yet, I had used a pre-packaged version and had expected that it would
2726 just work.
2727 Nevertheless, I had been convinced by the concepts of MH already
2728 and I am a software developer,
2729 still I required a lot of time to discover the cool features.
2730 How can we expect users to be even more advanced than me,
2731 just to enable them to use MH in a convenient and modern way?
2732 Unless they are strongly convinced of the concepts, they will fail.
2733 I have seen friends of me giving up disappointed
2734 before they truly used the system,
2735 although they had been motivated in the beginning.
2736 New users suffer hard enough to get used to the tool chest approach,
2737 we developers should spare them further inconveniences.
2738 .P
2739 Maintaining compatibility for its own sake is bad,
2740 because the code base will collect more and more compatibility code.
2741 Sticking to the compatibility code means remaining limited;
2742 whereas adjusting to the changes renders the compatibility unnecessary.
2743 Keeping unused alternatives in the code for longer than a short
2744 grace time is a bad choice as they likely
2745 gather bugs by not being constantly tested.
2746 Also, the increased code size and the greater number of conditions
2747 increase the maintenance costs.
2748 If any MH implementation would be the back-end of widespread
2749 email clients with large user bases, compatibility would be more
2750 important.
2751 Yet, it appears as if this is not the case.
2752 Hence, compatibility is hardly important for technical reasons.
2753 Its importance originates from personal reasons rather.
2754 Nmh's user base is small and old.
2755 Changing the interfaces causes inconvenience to long-term users of MH.
2756 It forces them to change their many years old MH configurations.
2757 I do understand this aspect, but by sticking to the old users,
2758 new users are kept from entering the world of MH.
2759 But the future lies in new users.
2760 In consequence, mmh invites new users by providing a convenient
2761 and modern setup, readily usable out-of-the-box.
2762 .P
2763 In mmh, all modern features are active by default and many previous
2764 approaches are removed or only accessible in a manual way.
2765 New default features include:
2766 .BU
2767 The attachment system (\c
2768 .Hd Attach )
2769 .Ci 8ff284ff9167eff8f5349481529332d59ed913b1 .
2770 .BU
2771 The draft folder facility (\c
2772 .Fn +drafts )
2773 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860 .
2774 .BU
2775 The unseen sequence (`u')
2776 .Ci c2360569e1d8d3678e294eb7c1354cb8bf7501c1
2777 and the sequence negation prefix (`!')
2778 .Ci db74c2bd004b2dc9bf8086a6d8bf773ac051f3cc .
2779 .BU
2780 Quoting the original message in the reply
2781 .Ci 67411b1f95d6ec987b4c732459e1ba8a8ac192c6 .
2782 .BU
2783 Forwarding messages using MIME
2784 .Ci 6e271608b7b9c23771523f88d23a4d3593010cf1 .
2785 .LP
2786 An mmh setup with a profile that defines only the path to the
2787 mail storage, is already convenient to use.
2788 Again, Paul Vixie's supports the direction I took:
2789 ``the `main branch' should just be modern''.
2790 .[
2791 paul vixie edginess nmh-workers
2792 .]
2798 .\" --------------------------------------------------------------
2799 .H1 "Styling
2800 .P
2801 Kernighan and Pike have emphasized the importance of style in the
2802 preface of \fIThe Practice of Programming\fP:
2803 .[ [
2804 kernighan pike practice of programming
2805 .], p. x]
2806 .QS
2807 Chapter 1 discusses programming style.
2808 Good style is so important to good programming that we have chosen
2809 to cover it first.
2810 .QE
2811 This section covers changes in mmh that were guided by the desire
2812 to improve on style.
2813 Many of them follow the advice given in the quoted book.
2818 .H2 "Code Style
2819 .Id code-style
2820 .P
2821 .U3 "Indentation Style
2822 .P
2823 Indentation styles are the holy cow of programming.
2824 Kernighan and Pike write:
2825 .[ [
2826 kernighan pike practice of programming
2827 .], p. 10]
2828 .QS
2829 Programmers have always argued about the layout of programs,
2830 but the specific style is much less important than its consistent
2831 application.
2832 Pick one style, preferably ours, use it consistently, and don't waste
2833 time arguing.
2834 .QE
2835 .P
2836 I agree that the constant application is most important,
2837 but I believe that some styles have advantages over others.
2838 For instance the indentation with tab characters only.
2839 The number of tabs corresponds to the nesting level \(en
2840 one tab, one level.
2841 Tab characters provide flexible visual appearance because developers
2842 can adjust their width as preferred.
2843 There is no more need to check for the correct mixture of
2844 tabs and spaces.
2845 Two simple rules ensure the integrity and flexibility of the visual
2846 appearance:
2847 .LI 1
2848 Leading whitespace must consist of tabs only.
2849 .LI 2
2850 All other whitespace should be spaces.
2851 .LP
2852 Although reformatting existing code should be avoided, I did it.
2853 I did not waste time arguing; I just reformatted the code.
2854 .Ci a485ed478abbd599d8c9aab48934e7a26733ecb1
2856 .U3 "Comments
2857 .P
2858 Kernighan and Pike demand: ``Don't belabor the obvious''.
2859 .[ [
2860 kernighan pike practice of programming
2861 .], p. 23]
2862 Following the advice, I removed unnecessary comments.
2863 For instance, I removed all comments in the following code excerpt
2864 .Ci 426543622b377fc5d091455cba685e114b6df674 :
2865 .VS
2866 context_replace(curfolder, folder); /* update current folder */
2867 seq_setcur(mp, mp->lowsel); /* update current message */
2868 seq_save(mp); /* synchronize message sequences */
2869 folder_free(mp); /* free folder/message structure */
2870 context_save(); /* save the context file */
2872 [...]
2874 int c; /* current character */
2875 char *cp; /* miscellaneous character pointer */
2877 [...]
2879 /* NUL-terminate the field */
2880 *cp = '\0';
2881 VE
2882 .P
2883 The information in each of the comments was present in the code
2884 statements already, except for the NUL-termination, which became
2885 obvious from the context.
2888 .U3 "Names
2889 .P
2890 Regarding this topic, Kernighan and Pike suggest:
2891 ``Use active names for functions''.
2892 .[ [
2893 kernighan pike practice of programming
2894 .], p. 4]
2895 One application of this rule was the rename of
2896 .Fu check_charset()
2897 to
2898 .Fu is_native_charset()
2899 .Ci 8d77b48284c58c135a6b2787e721597346ab056d .
2900 The same change additionally fixed a violation of ``Be accurate'',
2901 .[ [
2902 kernighan pike practice of programming
2903 .], p. 4]
2904 as the code did not match the expectation the function suggested.
2905 It did not compare charset names but prefixes of them only.
2906 In case the native charset was `ISO-8859-1', then
2907 .VS
2908 check_charset("ISO-8859-11", strlen("ISO-8859-11"))
2909 VE
2910 had returned true although the upper halves of the code pages
2911 are different.
2912 .P
2913 More important than using active names is using descriptive names.
2914 .VS
2915 m_unknown(in); /* the MAGIC invocation... */
2916 VE
2917 Renaming the obscure
2918 .Fu m_unknown()
2919 function was a delightful event, although it made the code less funny
2920 .Ci 611d68d19204d7cbf5bd585391249cb5bafca846 .
2921 .P
2922 Magic numbers are generally considered bad style.
2923 Obviously, Kernighan and Pike agree:
2924 ``Give names to magic numbers''.
2925 .[ [
2926 kernighan pike practice of programming
2927 .], p. 19]
2928 .P
2929 The argument
2930 .CW outnum
2931 of the function
2932 .Fu scan()
2933 in
2934 .Fn uip/scansbr.c
2935 holds the number of the message to be created.
2936 As well it encodes program logic with negative numbers and zero.
2937 This led to obscure code.
2938 I clarified the code by introducing two variables that extracted
2939 the hidden information:
2940 .VS
2941 int incing = (outnum > 0);
2942 int ismbox = (outnum != 0);
2943 VE
2944 The readable names are thus used in conditions;
2945 the variable
2946 .CW outnum
2947 is used only to extract ordinary message numbers
2948 .Ci b8b075c77be7794f3ae9ff0e8cedb12b48fd139f .
2949 .P
2950 Through the clarity improvement of the change detours in the program
2951 logic of related code parts became apparent.
2952 The implementation was simplified.
2953 This possibility to improve had been invisible before
2954 .Ci aa60b0ab5e804f8befa890c0a6df0e3143ce0723 .
2955 .P
2956 The names just described were a first step, yet the situation
2957 was further improved by giving names to the magic values of
2958 .CW outnum :
2959 .VS
2960 #define SCN_MBOX (-1)
2961 #define SCN_FOLD 0
2962 VE
2963 The two variables were updated thereafter as well:
2964 .VS
2965 int incing = (outnum != SCN_MBOX && outnum != SCN_FOLD);
2966 int scanfolder = (outnum == SCN_FOLD);
2967 VE
2968 Furthermore,
2969 .CW ismbox
2970 was replaced by
2971 .CW scanfolder
2972 because that matched better to the program logic.
2973 .Ci 7ffb36d28e517a6f3a10272056fc127592ab1c19
2978 .H2 "Structural Rework
2979 .P
2980 .ZZ
2981 Although the stylistic changes described already improve the
2982 readability of the source code, all of them were changes ``in the small''.
2983 Structural changes, in contrast, affect much larger code areas.
2984 They are more difficult to accomplish but lead to larger improvements,
2985 especially as they often influence the outer shape of the tools as well.
2986 .P
2987 At the end of their chapter on style,
2988 Kernighan and Pike ask: ``But why worry about style?''
2989 .[ [
2990 kernighan pike practice of programming
2991 .], p. 28].
2992 Following are two examples of structural rework that demonstrate
2993 why style is important in the first place.
2996 .U3 "Rework of \f(CWanno\fP
2997 .P
2998 Until 2002,
2999 .Pn anno
3000 had six functional command line switches:
3001 .Sw -component
3002 and
3003 .Sw -text ,
3004 each with an argument,
3005 and the two pairs of flags,
3006 .Sw -[no]date
3007 and
3008 .Sw -[no]inplace .
3009 Then Jon Steinhart introduced his attachment system.
3010 In need for more advanced annotation handling, he extended
3011 .Pn anno
3012 and added five more switches:
3013 .Sw -draft ,
3014 .Sw -list ,
3015 .Sw \%-delete ,
3016 .Sw -append ,
3017 and
3018 .Sw -number ,
3019 the last one taking an argument
3020 .Ci 7480dbc14bc90f2d872d434205c0784704213252 .
3021 Later,
3022 .Sw -[no]preserve
3023 was added as well
3024 .Ci d9b1d57351d104d7ec1a5621f090657dcce8cb7f .
3025 Then, the Synopsis section of the man page
3026 .Mp anno (1)
3027 read:
3028 .VS
3029 anno [+folder] [msgs] [-component f(CIfieldfP] [-inplace | -noinplace]
3030 [-date | -nodate] [-draft] [-append] [-list] [-delete]
3031 [-number [f(CInumfP|fPallfP]] [-preserve | -nopreserve] [-version]
3032 [-help] [-text f(CIbodyfP]
3033 VE
3034 .LP
3035 The implementation followed the same structure.
3036 Problems became visible when
3037 .Cl "anno -list -number 42
3038 worked on the current message instead of on message number 42,
3039 and
3040 .Cl "anno -list -number l:5
3041 did not work on the last five messages but failed with the mysterious
3042 error message: ``anno: missing argument to -list''.
3043 Yet, the invocation matched the specification in the man page.
3044 There, the correct use of
3045 .Sw -number
3046 was defined as being
3047 .Cl "[-number [num|all]]
3048 and the textual description for the combination with
3049 .Sw -list
3050 read:
3051 .QS
3052 The
3053 .Sw -list
3054 option produces a listing of the field bodies for
3055 header fields with names matching the specified component,
3056 one per line. The listing is numbered, starting at 1, if the
3057 .Sw -number
3058 option is also used.
3059 .QE
3060 .LP
3061 The problem was manifold.
3062 Semantically, the argument to the
3063 .Sw -number
3064 switch is only necessary in combination with
3065 .Sw -delete ,
3066 but not with
3067 .Sw -list .
3068 The code, however, required a numeric argument in any case.
3069 If the argument was missing or non-numeric,
3070 .Pn anno
3071 aborted with an error message that additionally had an off-by-one error.
3072 It printed the name of the switch one before the concerned one.
3073 .P
3074 Trying to fix these problems on the surface would not have solved them.
3075 They originate from a discrepance between the
3076 structure of the problem and the structure implemented in the program.
3077 Such structural differences can only be solved by adjusting the
3078 structure of the implementation to the structure of the problem.
3079 .P
3080 Steinhart had added the
3081 .Sw -list
3082 and
3083 .Sw -delete
3084 switches in the same way as the other switches though
3085 they are of structural different type.
3086 Semantically,
3087 .Sw -list
3088 and
3089 .Sw \%-delete
3090 introduce operation modes.
3091 Historically,
3092 .Pn anno
3093 had only one operation mode: adding header fields.
3094 With the extension, two more modes were added:
3095 listing and deleting header fields.
3096 The structure of the code changes did not pay respect to this
3097 fundamental change.
3098 Neither the implementation nor the documentation did clearly
3099 declare the exclusive operation modes as such.
3100 Having identified the problem, I solved it by putting structure into
3101 .Pn anno
3102 and its documentation
3103 .Ci d54c8db8bdf01e8381890f7729bc0ef4a055ea11 .
3104 .P
3105 The difference is visible in both the code and the documentation.
3106 For instance in the following code excerpt:
3107 .VS
3108 int delete = -2; /* delete header element if set */
3109 int list = 0; /* list header elements if set */
3110 [...]
3111 case DELETESW: /* delete annotations */
3112 delete = 0;
3113 continue;
3114 case LISTSW: /* produce a listing */
3115 list = 1;
3116 continue;
3117 VE
3118 .LP
3119 which was replaced by:
3120 .VS
3121 static enum { MODE_ADD, MODE_DEL, MODE_LIST } mode = MODE_ADD;
3122 [...]
3123 case DELETESW: /* delete annotations */
3124 mode = MODE_DEL;
3125 continue;
3126 case LISTSW: /* produce a listing */
3127 mode = MODE_LIST;
3128 continue;
3129 VE
3130 .LP
3131 The replacement code does not only reflect the problem's structure better,
3132 it is easier to understand as well.
3133 The same applies to the documentation.
3134 The man page was completely reorganized to propagate the same structure.
3135 This is already visible in the Synopsis section:
3136 .VS
3137 anno [+folder] [msgs] [-component f(CIfieldfP] [-text fPbodyfP]
3138 [-append] [-date | -nodate] [-preserve | -nopreserve]
3139 [-Version] [-help]
3141 anno -delete [+folder] [msgs] [-component fPfieldfP] [-text
3142 fPbodyfP] [-number fPnum fP| fPall fP] [-preserve | -nopreserve]
3143 [-Version] [-help]
3145 anno -list [+folder] [msgs] [-component fPfieldfP] [-number]
3146 [-Version] [-help]
3147 VE
3151 .U3 "Path Conversion
3152 .P
3153 Four kinds of path names can appear in MH:
3154 .LI 1
3155 Absolute Unix directory paths, like
3156 .Fn /etc/passwd .
3157 .LI 2
3158 Relative Unix directory paths, like
3159 .Fn ./foo/bar .
3160 .LI 3
3161 Absolute MH folder paths, like
3162 .Fn +projects/mmh .
3163 .LI 4
3164 Relative MH folder paths, like
3165 .Fn @subfolder .
3166 .LP
3167 Relative MH folder paths, are hardly documented
3168 although they are useful for large mail storages.
3169 The current mail folder is specified as `\c
3170 .Fn @ ',
3171 just like the current directory is specified as `\c
3172 .Fn . '.
3173 .P
3174 To allow MH tools to understand all four notations,
3175 they need to be able to convert between them.
3176 In nmh, these path name conversion functions were located in the files
3177 .Fn sbr/path.c
3178 (``return a pathname'') and
3179 .Fn sbr/m_maildir.c
3180 (``get the path for the mail directory'').
3181 The seven functions in the two files were documented with no more
3182 than two comments, which described obvious information.
3183 The signatures of the four exported functions did not explain their
3184 semantics:
3185 .LI 1
3186 .CW "char *path(char *, int);
3187 .LI 2
3188 .CW "char *pluspath(char *);
3189 .LI 3
3190 .CW "char *m_mailpath(char *);
3191 .LI 4
3192 .CW "char *m_maildir(char *);
3193 .P
3194 My investigations provided the following descriptions:
3195 .LI 1
3196 The second parameter of
3197 .Fu path()
3198 defines the type as which the path given in the first parameter should
3199 be treated.
3200 Directory paths are converted to absolute directory paths.
3201 Folder paths are converted to absolute folder paths.
3202 Folder paths must not include a leading `\fL@\fP' character.
3203 Leading plus characters are preserved.
3204 The result is a pointer to newly allocated memory.
3205 .LI 2
3206 .Fu pluspath()
3207 is a convenience-wrapper to
3208 .Fu path() ,
3209 to convert folder paths only.
3210 This function can not be used for directory paths.
3211 An empty string parameter causes a buffer overflow.
3212 .LI 3
3213 .Fu m_mailpath()
3214 converts directory paths to absolute directory paths.
3215 The characters `\fL+\fP' or `\fL@\fP' at the beginning of the path name are
3216 treated literal, i.e. as the first character of a relative directory path.
3217 Hence, this function can not be used for folder paths.
3218 In any case, the result is an absolute directory path,
3219 returned as a pointer to newly allocated memory.
3220 .LI 4
3221 .Fu m_maildir()
3222 returns the parameter unchanged if it is an absolute directory path
3223 or begins with the entry `\fL.\fP' or `\fL..\fP'.
3224 All other strings are prepended with the current working directory.
3225 Hence, this function can not be used for folder paths.
3226 The result is either an absolute directory path or a relative
3227 directory path, starting with dot or dot-dot.
3228 In contrast to the other functions, the result is a pointer to
3229 static memory.
3230 .P
3231 The situation was obscure, irritating, error-prone, and non-orthogonal.
3232 Additionally, no clear terminology was used to name the different
3233 kinds of path names.
3234 Sometimes, the names were even misleading, much as the first argument of
3235 .Fu m_mailpath() ,
3236 which was named
3237 .CW folder ,
3238 although
3239 .Fu m_mailpath()
3240 could not be used with MH folder arguments.
3241 .P
3242 I clarified the path name conversion by complete rework.
3243 First of all, the terminology needed to be defined.
3244 A path name is either in the Unix domain, then it is called
3245 \fIdirectory path\fP or it is in the MH domain,
3246 then it is called \fIfolder path\fP.
3247 The two terms need to be used with strict distinction.
3248 Second, I exploited the concept of path type indicators.
3249 By requiring every path name to start with a distinct type identifier,
3250 the conversion between the types could be fully automated.
3251 This allows the tools to accept path names of any type from the user.
3252 Therefore, it was necessary to require relative directory paths to be
3253 prefixed with a dot character.
3254 In consequence, the dot character could no longer be an alias for the
3255 current message
3256 .Ci cff0e16925e7edbd25b8b9d6d4fbdf03e0e60c01 .
3257 Third, I created three new functions to replace the previous mess:
3258 .LI 1
3259 .Fu expandfol()
3260 converts folder paths to absolute folder paths.
3261 Directory paths are simply passed through.
3262 This function is to be used for folder paths only, thus the name.
3263 The result is a pointer to static memory.
3264 .LI 2
3265 .Fu expanddir()
3266 converts directory paths to absolute directory paths.
3267 Folder paths are treated as relative directory paths.
3268 This function is to be used for directory paths only, thus the name.
3269 The result is a pointer to static memory.
3270 .LI 3
3271 .Fu toabsdir()
3272 converts any type of path to an absolute directory path.
3273 This is the function of choice for path conversion.
3274 Absolute directory paths are the most general representation of a
3275 path name.
3276 The result is a pointer to static memory.
3277 .P
3278 The new functions have names that indicate their use.
3279 Two of the functions convert relative to absolute path names of the
3280 same type.
3281 The third function converts any path name type to the most general one,
3282 the absolute directory path.
3283 All of the functions return pointers to static memory.
3284 The file
3285 .Fn sbr/path.c
3286 contains the implementation of the functions;
3287 .Fn sbr/m_maildir.c
3288 was removed.
3289 .Ci d39e2c447b0d163a5a63f480b23d06edb7a73aa0
3290 .P
3291 Along with the path conversion rework, I also replaced
3292 .Fu getfolder(FDEF)
3293 with
3294 .Fu getdeffol()
3295 and
3296 .Fu getfolder(FCUR)
3297 with
3298 .Fu getcurfol() ,
3299 which only wraps
3300 .Fu expandfol(""@"")
3301 for convenience.
3302 This code was moved from
3303 .Fn sbr/getfolder.c
3304 into
3305 .Fn sbr/path.c
3306 as well.
3307 .Ci d39e2c447b0d163a5a63f480b23d06edb7a73aa0
3308 .P
3309 The related function
3310 .Fu etcpath()
3311 is now included in
3312 .Fn sbr/path.c ,
3313 too
3314 .Ci b4c29794c12099556151d93a860ee51badae2e35 .
3315 Previously, it had been located in
3316 .Fn config/config.c .
3317 .P
3318 Now,
3319 .Fn sbr/path.c
3320 contains all path handling code.
3321 Besides being less code, its readability is highly improved.
3322 The functions follow a common style and are well documented.
3327 .H2 "Profile Reading
3328 .P
3329 The MH profile contains the configuration of a user-specific MH setup.
3330 MH tools read the profile right after starting up
3331 because it contains the location of the user's mail storage
3332 and similar settings that influence the whole setup.
3333 Furthermore, the profile contains the default switches for the tools
3334 as well.
3335 The context file is read along with the profile.
3336 .P
3337 For historic reasons, some MH tools did not read the profile and context.
3338 Among them were
3339 .Pn post /\c
3340 .Pn spost ,
3341 .Pn mhmail ,
3342 and
3343 .Pn slocal .
3344 The reason why these tools ignored the profile were not clearly stated.
3345 During a discussion on the nmh-workers mailing list,
3346 David Levine posted an explanation, quoting John Romine:
3347 .[
3348 nmh-workers levine post profile
3349 .]
3351 .QS
3352 I asked John Romine and here's what he had to say, which
3353 agrees and provides an example that convinces me:
3354 .QS
3355 My take on this is that
3356 .Pn post
3357 should not be called by users directly, and it doesn't read the
3358 .Fn .mh_profile
3359 (only front-end UI programs read the profile).
3360 .QP
3361 For example, there can be contexts where
3362 .Pn post
3363 is called by a helper program (like `\c
3364 .Pn mhmail ')
3365 which may be run by a non-MH user.
3366 We don't want this to prompt the user to create an MH profile, etc.
3367 .QP
3368 My suggestion would be to have
3369 .Pn send
3370 pass a (hidden) `\c
3371 .Sw -fileproc
3372 .Ar proc '
3373 option to
3374 .Pn post
3375 if needed.
3376 You could also
3377 use an environment variable (I think
3378 .Pn send /\c
3379 .Pn whatnow
3380 do this).
3381 .QE
3382 .sp \n(PDu
3383 I think that's the way to go.
3384 My personal preference is to use a command line option,
3385 not an environment variable.
3386 .QE
3388 .P
3389 To solve the problem that
3390 .Pn post
3391 does not honor the
3392 .Pe fileproc
3393 profile entry,
3394 the community roughly agreed that a switch
3395 .Sw -fileproc
3396 should be added to
3397 .Pn post
3398 to be able to pass a different fileproc.
3399 I strongly disagree with this approach because it does not solve
3400 the problem; it only removes a single symptom.
3401 The actual problem is that
3402 .Pn post
3403 does not behave as expected,
3404 though all programs should behave as expected.
3405 Clear and general concepts are a precondition for this.
3406 Thus, there should be no separation into ``front-end UI programs''
3407 and ones that ``should not be called by users directly''.
3408 The real solution is having all MH tools read the profile.
3409 .P
3410 But the problem has a further aspect,
3411 which originates from
3412 .Pn mhmail
3413 mainly.
3414 .Pn mhmail
3415 was intended to be a replacement for
3416 .Pn mailx
3417 on systems with MH installations.
3418 In difference to
3419 .Pn mailx ,
3420 .Pn mhmail
3421 used MH's
3422 .Pn post
3423 to send the message.
3424 The idea was that using
3425 .Pn mhmail
3426 should not be influenced whether the user had
3427 MH set up for himself or not.
3428 Therefore
3429 .Pn mhmail
3430 had not read the profile.
3431 As
3432 .Pn mhmail
3433 used
3434 .Pn post ,
3435 .Pn post
3436 was not allowed to read the profile neither.
3437 This is the reason for the actual problem.
3438 Yet, this was not considered much of a problem because
3439 .Pn post
3440 was not intended to be used by users directly.
3441 To invoke
3442 .Pn post ,
3443 .Pn send
3444 was used an a front-end.
3445 .Pn send
3446 read the profile and passed all relevant values on the command line to
3447 .Pn post
3448 \(en an awkward solution.
3449 .P
3450 The important insight is that
3451 .Pn mhmail
3452 is a wolf in sheep's clothing.
3453 This alien tool broke the concepts because it was treated like
3454 a normal MH tool.
3455 Instead it should have been treated accordingly to its foreign style.
3456 .P
3457 The solution is not to prevent the tools from reading the profile but
3458 to instruct them to read a different profile.
3459 .Pn mhmail
3460 could have set up a well-defined profile and caused the following
3461 .Pn post
3462 to use this profile by exporting an environment variable.
3463 With this approach, no special cases would have been introduced
3464 and no surprises would have been caused.
3465 By writing a wrapper program to provide a clean temporary profile,
3466 the concept could have been generalized orthogonally to the whole
3467 MH tool chest.
3468 .P
3469 In mmh, the wish to have
3470 .Pn mhmail
3471 as a replacement for
3472 .Pn mailx
3473 is considered obsolete.
3474 Mmh's
3475 .Pn mhmail
3476 does no longer cover this use-case
3477 .Ci d36e56e695fe1c482c7920644bfbb6386ac9edb0 .
3478 Currently,
3479 .Pn mhmail
3480 is in a transition state
3481 .Ci 32d4f9daaa70519be3072479232ff7be0500d009 .
3482 It may become a front-end to
3483 .Pn comp ,
3484 which provides an alternative interface which can be more convenient
3485 in some cases.
3486 This would convert
3487 .Pn mhmail
3488 into an ordinary MH tool.
3489 If, however, this idea does not convince, then
3490 .Pn mhmail
3491 will be removed.
3492 .P
3493 .ZZ -1
3494 In the mmh tool chest, every program reads the profile.
3495 (\c
3496 .Pn slocal
3497 is not considered part of the mmh tool chest (cf. Sec.
3498 .Cf slocal ).)
3499 Mmh has no
3500 .Pn post
3501 program, but it has
3502 .Pn spost ,
3503 which now does read the profile
3504 .Ci 3e017a7abbdf69bf0dff7a4073275961eda1ded8 .
3505 Following this change,
3506 .Pn send
3507 and
3508 .Pn spost
3509 can be considered for merging.
3510 Besides
3511 .Pn send ,
3512 .Pn spost
3513 is only invoked directly by the to-be-changed
3514 .Pn mhmail
3515 implementation and by
3516 .Pn rcvdist ,
3517 which requires rework anyway.
3519 .P
3520 Jeffrey Honig quoted Marshall T. Rose explaining the decision that
3521 .Pn post
3522 ignores the profile:
3523 .[
3524 nmh-workers honig post profile
3525 .]
3526 .QS
3527 when you run mh commands in a script, you want all the defaults to be
3528 what the man page says.
3529 when you run a command by hand, then you want your own defaults...
3530 .QE
3531 .LP
3532 The explanation neither matches the problem concerned exactly
3533 nor is the interpretation clear.
3534 If the described desire addresses the technical level,
3535 then it conflicts with the Unix philosophy,
3536 precisely because the indistinquishability of human and script
3537 input is the main reason for the huge software leverage in Unix.
3538 If, however, the described desire addresses the user's view,
3539 then different technical solutions are more appropriate.
3540 The two cases can be regarded simply as two different MH setups.
3541 Hence, mapping the problem of different behavior between interactive and
3542 automated use on the concept of switching between different profiles,
3543 marks it already solved.
3547 .H2 "Standard Libraries
3548 .P
3549 MH is one decade older than the POSIX and ANSI C standards.
3550 Hence, MH included own implementations of functions
3551 that were neither standardized nor widely available, back then.
3552 Today, twenty years after POSIX and ANSI C were published,
3553 developers can expect that systems comply with these standards.
3554 In consequence, MH-specific replacements for standard functions
3555 can and should be dropped.
3556 Kernighan and Pike advise: ``Use standard libraries''.
3557 .[ [
3558 kernighan pike practice of programming
3559 .], p. 196]
3560 Actually, MH had followed this advice in history,
3561 but it had not adjusted to more recent changes in this field.
3562 The
3563 .Fu snprintf()
3564 function, for instance, was standardized with C99 and is available
3565 almost everywhere because of its high usefulness.
3566 Thus, the project's own implementation of
3567 .Fu snprintf()
3568 was dropped in March 2012 in favor for using the one of the
3569 standard library
3570 .Ci 0052f1024deb0a0a2fc2e5bacf93d45a5a9c9b32 .
3571 Such decisions limit the portability of mmh
3572 if systems do not support these standardized and widespread functions.
3573 This compromise is made because mmh focuses on the future.
3574 .P
3575 .ZZ
3576 As I am still in my twenties, have no programming experience from
3577 past decades.
3578 I have not followed the evolution of C through time.
3579 I have not suffered from the the Unix wars.
3580 I have not longed for standardization.
3581 All my programming experience is from a time when ANSI C and POSIX
3582 were well established already.
3583 Thus, I needed to learn about the history in retrospective.
3584 I have only read a lot of books about the (good) old times.
3585 This put me in a difficult position when working with old code.
3586 I need to freshly acquire knowledge about old code constructs and ancient
3587 programming styles, whereas older programmers know these things by
3588 heart from their own experience.
3589 Being aware of the situation, I rather let people with more historic
3590 experience do the transition from ancient code constructs to
3591 standardized ones.
3592 Lyndon Nerenberg covered large parts of this task for the nmh project.
3593 He converted project-specific functions to POSIX replacements,
3594 also removing the conditionals compilation of now standardized features.
3595 Ken Hornstein and David Levine had their part in this work, as well.
3596 Often, I only pulled the changes over from nmh into mmh.
3597 These changes include many commits, among them:
3598 .Ci 768b5edd9623b7238e12ec8dfc409b82a1ed9e2d
3599 .Ci 0052f1024deb0a0a2fc2e5bacf93d45a5a9c9b32 .
3600 .P
3601 Nevertheless, I worked on the task as well, tidying up the
3602 \fIMH standard library\fP,
3603 .Fn libmh.a .
3604 It is located in the
3605 .Fn sbr
3606 (``subroutines'') directory in the source tree and
3607 includes functions that mmh tools usually need.
3608 Among them are MH-specific functions for profile, context, sequence,
3609 and folder handling, but as well
3610 MH-independent functions, such as auxiliary string functions,
3611 portability interfaces and error-checking wrappers for critical
3612 functions of the standard library.
3613 .BU
3614 I have replaced the
3615 .Fu atooi()
3616 function with calls to
3617 .Fu strtoul() ,
3618 setting the third parameter, the base, to eight.
3619 .Fu strtoul()
3620 is part of C89 and thus considered safe to use
3621 .Ci c490c51b3c0f8871b6953bd0c74551404f840a74 .
3622 .BU
3623 I did remove project-included fallback implementations of
3624 .Fu memmove()
3625 and
3626 .Fu strerror()
3627 .Ci b067ff5c465a5d243ce5a19e562085a9a1a97215 ,
3628 although Peter Maydell had re-included them into nmh in 2008
3629 to support SunOS 4.
3630 Nevertheless, these functions are part of ANSI C.
3631 Systems that do not even provide full ANSI C support should not
3632 put a load on mmh.
3633 .BU
3634 The
3635 .Fu copy()
3636 function copies the string in parameter one to the location in
3637 parameter two.
3638 In contrast to
3639 .Fu strcpy() ,
3640 it returns a pointer to the terminating null-byte in the destination area.
3641 The code was adjusted to replace
3642 .Fu copy()
3643 with
3644 .Fu strcpy() ,
3645 except within
3646 .Fu concat() ,
3647 where
3648 .Fu copy()
3649 was more convenient.
3650 Therefore, the definition of
3651 .Fu copy()
3652 was moved into the source file of
3653 .Fu concat()
3654 and its visibility it limited to that
3655 .Ci 552fd7253e5ee9e554c5c7a8248a6322aa4363bb .
3656 .BU
3657 The function
3658 .Fu r1bindex()
3659 had been a generalized version of
3660 .Fu basename()
3661 with minor differences.
3662 As all calls to
3663 .Fu r1bindex()
3664 had the slash (`\fL/\fP') as delimiter anyway,
3665 replacing
3666 .Fu r1bindex()
3667 with the more specific and better-named function
3668 .Fu basename()
3669 became desirable.
3670 Unfortunately, many of the 54 calls to
3671 .Fu r1bindex()
3672 depended on a special behavior,
3673 which differed from the POSIX specification for
3674 .Fu basename() .
3675 Hence,
3676 .Fu r1bindex()
3677 was kept but renamed to
3678 .Fu mhbasename() ,
3679 setting the delimiter to the slash
3680 .Ci 240013872c392fe644bd4f79382d9f5314b4ea60 .
3681 For possible uses of
3682 .Fu r1bindex()
3683 with a different delimiter,
3684 the ANSI C function
3685 .Fu strrchr()
3686 provides the core functionality.
3687 .BU
3688 .ZZ
3689 The
3690 .Fu ssequal()
3691 function \(en apparently for ``substring equal'' \(en
3692 was renamed to
3693 .Fu isprefix() ,
3694 because this is what it actually checked
3695 .Ci c20b4fa14515c7ab388ce35411d89a7a92300711.
3696 Its source file had included both of the following comments, no joke.
3697 .in -\n(PIu
3698 .VS
3699 /*
3700 * THIS CODE DOES NOT WORK AS ADVERTISED.
3701 * It is actually checking if s1 is a PREFIX of s2.
3702 * All calls to this function need to be checked to see
3703 * if that needs to be changed. Prefix checking is cheaper, so
3704 * should be kept if it's sufficient.
3705 */
3706 sp .5
3707 /*
3708 * Check if s1 is a substring of s2.
3709 * If yes, then return 1, else return 0.
3710 */
3711 VE
3712 .in +\n(PIu
3713 Eventually, the function was completely replaced with calls to
3714 .Fu strncmp()
3715 .Ci b0b1dd37ff515578cf7cba51625189eb34a196cb .
3721 .H2 "User Data Locations
3722 .P
3723 In nmh, a personal setup consists of the MH profile and the MH directory.
3724 The profile is a file named
3725 .Fn \&.mh_profile
3726 in the user's home directory.
3727 It contains the static configuration.
3728 It also contains the location of the MH directory in the profile entry
3729 .Pe Path .
3730 The MH directory contains the mail storage and is the first
3731 place to search for form files, scan formats, and similar
3732 configuration files.
3733 The location of the MH directory can be chosen freely by the user.
3734 The usual name is a directory named
3735 .Fn Mail
3736 in the user's home directory.
3737 .P
3738 The way MH data is split between profile and MH directory is a legacy.
3739 It is only sensible in a situation where the profile is the only
3740 configuration file.
3741 Why else should the mail storage and the configuration files be intermixed?
3742 They are of different kind:
3743 One kind is the data to be operated on and the other kind is
3744 the configuration to change how tools operate.
3745 Splitting the configuration between the profile and the MH directory
3746 is inappropriate, as well.
3747 I improved the situation by breaking compatibility.
3748 .P
3749 In mmh, personal data is grouped by type.
3750 This results in two distinct parts:
3751 the mail storage and the configuration.
3752 The mail storage directory still contains all the messages,
3753 but, in exception of public sequences files, nothing else.
3754 In difference to nmh, the auxiliary configuration files are no longer
3755 located there.
3756 Therefore, the directory is no longer called the user's \fIMH directory\fP
3757 but the user's \fImail storage\fP.
3758 Its location is still user-chosen, with the default name
3759 .Fn Mail
3760 in the user's home directory.
3761 The configuration is grouped together in the hidden directory
3762 .Fn \&.mmh
3763 in the user's home directory.
3764 This \fImmh directory\fP contains the context file, personal forms,
3765 scan formats, and the like, but also the user's profile, now named
3766 .Fn profile .
3767 The path to the profile is no longer
3768 .Fn $HOME/.mh_profile
3769 but
3770 .Fn $HOME/.mmh/profile .
3771 (The alternative of having file
3772 .Fn $HOME/.mh_profile
3773 and a configuration directory
3774 .Fn $HOME/.mmh
3775 appeared to be inconsistent.)
3776 .P
3777 The approach chosen for mmh is consistent, simple, and familiar to
3778 Unix users.
3779 The main achievement of the change is the clear and sensible separation
3780 of the mail storage and the configuration.
3781 .Ci 7030d7edb099bff36ded7548bb5380f7acab4f9b
3782 .P
3783 As MH allows users to have multiple MH setups,
3784 it is necessary to switch the profile.
3785 The profile is the single entry point to access the rest of a
3786 personal MH setup.
3787 In nmh, the environment variable
3788 .Ev MH
3789 is used to specify a different profile.
3790 To operate in the same MH setup with a separate context, the
3791 .Ev MHCONTEXT
3792 environment variable is used.
3793 This allows having a separate current folder in each terminal at
3794 the same time, for instance.
3795 In mmh, three environment variables replace the two of nmh.
3796 .Ev MMH
3797 overrides the default location of the mmh directory (\c
3798 .Fn .mmh ).
3799 .Ev MMHP
3800 and
3801 .Ev MMHC
3802 override the paths to the profile and context file, respectively.
3803 This approach allows the set of personal configuration files to be chosen
3804 independently of the profile, context, and mail storage.
3805 The new approach has no functional disadvantages,
3806 as every setup I can imagine can be implemented with both approaches,
3807 possibly even easier with the new one.
3808 .Ci 7030d7edb099bff36ded7548bb5380f7acab4f9b
3814 .H2 "Modularization
3815 .Id modularization
3816 .P
3817 The source code of the mmh tools is located in the
3818 .Fn uip
3819 (``user interface programs'') directory.
3820 Each tool has a source file with the name of the command.
3821 For example,
3822 .Pn rmm
3823 is built from
3824 .Fn uip/rmm.c .
3825 Some source files are used for multiple programs.
3826 For example
3827 .Fn uip/scansbr.c
3828 is used for both
3829 .Pn scan
3830 and
3831 .Pn inc .
3832 In nmh, 49 tools were built from 76 source files.
3833 This is a ratio of 1.6 source files per program.
3834 32 programs depended on multiple source files;
3835 17 programs depended on one source file only.
3836 In mmh, 39 tools are built from 51 source files.
3837 This is a ratio of 1.3 source files per program.
3838 18 programs depend on multiple source files;
3839 21 programs depend on one source file only.
3840 (These numbers and the ones in the following text ignore the MH library
3841 as well as shell scripts and multiple names for the same program.)
3842 .P
3843 Splitting the source code of a large program into multiple files can
3844 increase the readability of its source code,
3845 but most of the mmh tools are small and straight-forward programs.
3846 In exception of the MIME handling tools (i.e.
3847 .Pn mhbuild ,
3848 .Pn mhstore ,
3849 .Pn show ,
3850 etc.),
3851 .Pn pick
3852 is the only tool with more than one thousand lines of source code.
3853 Splitting programs with less than one thousand lines of code into
3854 multiple source files leads seldom to better readability.
3855 For such tools, splitting still makes sense
3856 when parts of the code are reused in other programs
3857 and the reused code fragment is (1) not general enough
3858 for including it in the MH library
3859 or (2) has dependencies on a library that only few programs need.
3860 .Fn uip/packsbr.c ,
3861 for instance, provides the core program logic for the
3862 .Pn packf
3863 and
3864 .Pn rcvpack
3865 programs.
3866 .Fn uip/packf.c
3867 and
3868 .Fn uip/rcvpack.c
3869 mainly wrap the core function appropriately.
3870 No other tools use the folder packing functions.
3871 As another example,
3872 .Fn uip/termsbr.c
3873 accesses terminal properties, which requires linking with the
3874 \fItermcap\fP or a \fIcurses\fP library.
3875 If
3876 .Fn uip/termsbr.c
3877 is included in the MH library, then every program needs to be linked
3878 with termcap or curses, although only few of the programs use
3879 the library.
3880 .P
3881 The task of MIME handling is complex enough that splitting its code
3882 into multiple source files improves the readability.
3883 The program
3884 .Pn mhstore ,
3885 for instance, is compiled out of seven source files with 2\|500
3886 lines of code in summary.
3887 The main code file
3888 .Fn uip/mhstore.c
3889 consists of 800 lines; the other 1\|700 lines are code reused in
3890 other MIME handling tools.
3891 It seems to be worthwhile to bundle the generic MIME handling code into
3892 a MH-MIME library, as a companion to the MH standard library.
3893 This is left to be done.
3894 .P
3895 The work already accomplished focussed on the non-MIME tools.
3896 The amount of code compiled into each program was reduced.
3897 This eases the understanding of the code base.
3898 In nmh,
3899 .Pn comp
3900 was built from six source files:
3901 .Fn comp.c ,
3902 .Fn whatnowproc.c ,
3903 .Fn whatnowsbr.c ,
3904 .Fn sendsbr.c ,
3905 .Fn annosbr.c ,
3906 and
3907 .Fn distsbr.c .
3908 In mmh, it builds from only two:
3909 .Fn comp.c
3910 and
3911 .Fn whatnowproc.c .
3912 In nmh's
3913 .Pn comp ,
3914 the core function of
3915 .Pn whatnow ,
3916 .Pn send ,
3917 and
3918 .Pn anno
3919 were all compiled into
3920 .Pn comp .
3921 This saved the need to execute these programs with
3922 the expensive system calls
3923 .Fu fork()
3924 and
3925 .Fu exec() .
3926 Whereas this approach improved the time performance,
3927 it interwove the source code.
3928 Core functionalities were not encapsulated into programs but into
3929 function, which were then wrapped by programs.
3930 For example,
3931 .Fn uip/annosbr.c
3932 included the function
3933 .Fu annotate() .
3934 Each program that wanted to annotate messages, included the source file
3935 .Fn uip/annosbr.c
3936 and called
3937 .Fu annotate() .
3938 Because the function
3939 .Fu annotate()
3940 was used like the tool
3941 .Pn anno ,
3942 it had seven parameters, reflecting the command line switches of the tool.
3943 When another pair of command line switches was added to
3944 .Pn anno ,
3945 a rather ugly hack was implemented to avoid adding another parameter
3946 to the function
3947 .Ci d9b1d57351d104d7ec1a5621f090657dcce8cb7f .
3948 .P
3949 In mmh, the relevant code of
3950 .Pn comp
3951 comprises the two files
3952 .Fn uip/comp.c
3953 and
3954 .Fn uip/whatnowproc.c ,
3955 together 210 lines of code,
3956 whereas in nmh,
3957 .Pn comp
3958 comprises six files with 2\|450 lines.
3959 Not all of the code in these six files is actually used by
3960 .Pn comp ,
3961 but the reader needed to read it all to know which parts are relevant.
3962 Understanding nmh's
3963 .Pn comp ,
3964 required understanding the inner workings of
3965 .Fn uip/annosbr.c
3966 first.
3967 To be sure to fully understand a program, its whole source code needs
3968 to be examined.
3969 Not doing so is a leap of faith, assuming that the developers
3970 have avoided obscure programming techniques.
3971 Here, it should be recalled that information passed in obscure ways
3972 through the program's source base, due to the aforementioned hack
3973 to save an additional parameter in nmh's
3974 .Pn anno .
3975 .P
3976 In mmh, understanding
3977 .Pn comp
3978 requires to read only 210 lines of code to read, whereas the amount
3979 is ten times more for nmh's
3980 .Pn comp .
3981 .P
3982 By separating the tools on the program-level,
3983 the boundaries are clearly visible, as the interfaces are calls to
3984 .Fu exec()
3985 rather than arbitrary function calls.
3986 Additionally, this kind of separation is more strict because
3987 it is technically enforced by the operating system;
3988 it can not be simply bypassed with global variables.
3989 Good separation simplifies the understanding of program code
3990 because the area influenced by any particular statement is small.
3991 As I have read a lot in nmh's code base during the last two years,
3992 I have learned about the easy and the difficult parts.
3993 In my observation, the understanding of code is enormously eased
3994 if the influenced area is small and clearly bounded.
3995 .P
3996 Yet, the real problem is another:
3997 Nmh violates the golden ``one tool, one job'' rule of the Unix philosophy.
3998 Understanding
3999 .Pn comp
4000 requires understanding
4001 .Fn uip/annosbr.c
4002 and
4003 .Fn uip/sendsbr.c
4004 because
4005 .Pn comp
4006 annotates and sends messages.
4007 In nmh, there surely exist the tools
4008 .Pn anno
4009 and
4010 .Pn send ,
4011 which cover these jobs,
4012 but
4013 .Pn comp
4014 and
4015 .Pn repl
4016 and
4017 .Pn forw
4018 and
4019 .Pn dist
4020 and
4021 .Pn whatnow
4022 and
4023 .Pn viamail
4024 \(en they all (!) \(en
4025 have the same annotating and sending functions included, once more.
4026 As a result,
4027 .Pn comp
4028 sends messages without using
4029 .Pn send .
4030 The situation is the same as if
4031 .Pn grep
4032 would page its output without using
4033 .Pn more
4034 just because both programs are part of the same code base.
4035 .P
4036 The clear separation on the surface of nmh
4037 \(en the tool chest approach \(en
4038 is violated on the level below.
4039 This violation is for the sake of time performance.
4040 Decades ago, sacrificing readability and conceptional beauty
4041 for speed might have been necessary to prevent MH from being
4042 unusably slow, but today this is not the case anymore.
4043 No longer should speed improvements that became unnecessary be kept.
4044 No longer should readability or conceptional beauty be sacrificed.
4045 No longer should the Unix philosophy's ``one tool, one job''
4046 guideline be violated.
4047 Therefore, mmh's
4048 .Pn comp
4049 no longer sends messages.
4050 .P
4051 In mmh, different jobs are divided among separate programs that
4052 invoke each other as needed.
4053 In consequence,
4054 .Pn comp
4055 invokes
4056 .Pn whatnow
4057 which thereafter invokes
4058 .Pn send
4059 .Ci 3df5ab3c116e6d4a2fb4bb5cc9dfc5f781825815
4060 .Ci c73c00bfccd22ec77e9593f47462aeca4a8cd9c0 .
4061 The clear separation on the surface is maintained on the level below.
4062 Human users and other tools use the same interface \(en
4063 annotations, for example, are made by invoking
4064 .Pn anno ,
4065 no matter if requested by programs or by human beings
4066 .Ci 469a4163c2a1a43731d412eaa5d9cae7d670c48b
4067 .Ci aed384169af5204b8002d06e7a22f89197963d2d
4068 .Ci 3caf9e298a8861729ca8b8a84f57022b6f3ea742 .
4069 .P
4070 .ZZ -1
4071 The decrease of tools built from multiple source files and thus
4072 the decrease of
4073 .Fn uip/*sbr.c
4074 files confirm the improvement
4075 .Ci 9e6d91313f01c96b4058d6bf419a8ca9a207bc33
4076 .ci 81744a46ac9f845d6c2b9908074d269275178d2e
4077 .Ci f0f858069d21111f0dbea510044593f89c9b0829
4078 .Ci 0503a6e9be34f24858b55b555a5c948182b9f24b
4079 .Ci 27826f9353e0f0b04590b7d0f8f83e60462b90f0
4080 .Ci d1da1f94ce62160aebb30df4063ccbc53768656b
4081 .Ci c42222869e318fff5dec395eca3e776db3075455 .
4082 This is also visible in the complexity of the build dependency graphs:
4084 .sp
4085 Nmh:
4086 .BP input/deps-nmh.eps .5i
4087 .EP
4088 .sp
4089 Mmh:
4090 .BP input/deps-mmh.eps .8i
4091 .EP
4093 The figures display all program to source file relationships where
4094 programs (ellipses) are built from multiple source files (rectangles).
4095 The primary source file of each program is omited from the graph.