docs/master

view discussion.roff @ 152:1b9bdc178610

style: Switch to french spacing. *yay*
author markus schnalke <meillo@marmaro.de>
date Fri, 06 Jul 2012 15:00:54 +0200
parents b7b81ae9c9d8
children 63e885fb48ba
line source
1 .H0 "Discussion
2 .P
3 This main chapter discusses the practical work done in the mmh project.
4 It is structured along the goals to achieve.
5 The concrete work done
6 is described in the examples of how the general goals were achieved.
7 The discussion compares the current version of mmh with the state of
8 nmh just before the mmh project started, i.e. Fall 2011.
9 Current changes of nmh will be mentioned only as side notes.
10 .\" XXX where do I discuss the parallel development of nmh?
14 .\" --------------------------------------------------------------
15 .H1 "Streamlining
17 .P
18 MH had been considered an all-in-one system for mail handling.
19 The community around nmh has a similar understanding.
20 In fundamental difference, mmh shall be a MUA only.
21 I believe that the development of all-in-one mail systems is obsolete.
22 Today, email is too complex to be fully covered by single projects.
23 Such a project won't be able to excel in all aspects.
24 Instead, the aspects of email should be covered my multiple projects,
25 which then can be combined to form a complete system.
26 Excellent implementations for the various aspects of email exist already.
27 Just to name three examples: Postfix is a specialized MTA,
28 Procmail is a specialized MDA, and Fetchmail is a specialized MRA.
29 I believe that it is best to use such specialized tools instead of
30 providing the same function again as a side-component in the project.
31 .P
32 Doing something well, requires to focus on a small set of specific aspects.
33 Under the assumption that focused development produces better results
34 in the particular area, specialized projects will be superior
35 in their field of focus.
36 Hence, all-in-one mail system projects \(en no matter if monolithic
37 or modular \(en will never be the best choice in any of the fields.
38 Even in providing the best consistent all-in-one system they are likely
39 to be beaten by projects that focus only on integrating existing mail
40 components to a homogeneous system.
41 .P
42 The limiting resource in Free Software community development
43 is usually man power.
44 If the development power is spread over a large development area,
45 it becomes even more difficult to compete with the specialists in the
46 various fields.
47 The concrete situation for MH-based mail systems is even tougher,
48 given the small and aged community, including both developers and users,
49 it has.
50 .P
51 In consequence, I believe that the available development resources
52 should focus on the point where MH is most unique.
53 This is clearly the user interface \(en the MUA.
54 Peripheral parts should be removed to streamline mmh for the MUA task.
57 .H2 "Mail Transfer Facilities
58 .P
59 In contrast to nmh, which also provides mail submission and mail retrieval
60 agents, mmh is a MUA only.
61 This general difference initiated the development of mmh.
62 Removing the mail transfer facilities had been the first work task
63 in the mmh project.
64 .P
65 Focusing on one mail agent role only is motivated by Eric Allman's
66 experience with Sendmail.
67 He identified limiting Sendmail the MTA task had be one reason for
68 its success:
69 .[ [
70 costales sendmail
71 .], p. xviii]
72 .QS
73 Second, I limited myself to the routing function \(en
74 I wouldn't write user agents or delivery back-ends.
75 This was a departure of the dominant through of the time,
76 in which routing logic, local delivery, and often the network code
77 were incorporated directly into the user agents.
78 .QE
79 .P
80 In mmh, the Mail Submission Agent (MSA) is called
81 \fIMessage Transfer Service\fP (MTS).
82 This facility, implemented by the
83 .Pn post
84 command, established network connections and spoke SMTP to submit
85 messages for relay to the outside world.
86 The changes in email demanded changes in this part of nmh too.
87 Encryption and authentication for network connections
88 needed to be supported, hence TLS and SASL were introduced into nmh.
89 This added complexity to nmh without improving it in its core functions.
90 Also, keeping up with recent developments in the field of
91 mail transfer requires development power and specialists.
92 In mmh this whole facility was simply cut off.
93 .Ci f6aa95b724fd8c791164abe7ee5468bf5c34f226
94 .Ci fecd5d34f65597a4dfa16aeabea7d74b191532c3
95 .Ci 156d35f6425bea4c1ed3c4c79783dc613379c65b
96 Instead, mmh depends on an external MSA.
97 The only outgoing interface available to mmh is the
98 .Pn sendmail
99 command, which almost any MSA provides.
100 If not, a wrapper program can be written.
101 It must read the message from the standard input, extract the
102 recipient addresses from the message header, and hand the message
103 over to the MSA.
104 For example, a wrapper script for qmail would be:
105 .VS
106 #!/bin/sh
107 exec qmail-inject # ignore command line arguments
108 VE
109 The requirement to parse the recipient addresses out of the message header
110 is likely to be removed in the future.
111 Then mmh would give the recipient addresses as command line arguments.
112 This appears to be the better interface.
113 .\" XXX implement it
114 .P
115 To retrieve mail, the
116 .Pn inc
117 command acted as Mail Retrieval Agent (MRA).
118 It established network connections
119 and spoke POP3 to retrieve mail from remote servers.
120 As with mail submission, the network connections required encryption and
121 authentication, thus TLS and SASL were added.
122 Support for message retrieval through IMAP will become necessary
123 to be added soon, too, and likewise for any other changes in mail transfer.
124 Not so for mmh because it has dropped the support for retrieving mail
125 from remote locations.
126 .Ci ab7b48411962d26439f92f35ed084d3d6275459c
127 Instead, it depends on an external tool to cover this task.
128 In mmh exist two paths for messages to enter mmh's mail storage:
129 (1) Mail can be incorporated with
130 .Pn inc
131 from the system maildrop, or (2) with
132 .Pn rcvstore
133 by reading them, one at a time, from the standard input.
134 .P
135 With the removal of the MSA and MRA, mmh converted from an all-in-one
136 mail system to being a MUA only.
137 Now, of course, mmh depends on third-party software.
138 An external MSA is required to transfer mail to the outside world;
139 an external MRA is required to retrieve mail from remote machines.
140 There exist excellent implementations of such software,
141 which do this specific task likely better than the internal
142 versions had done it.
143 Also, the best suiting programs can be freely chosen.
144 .P
145 As it had already been possible to use an external MSA or MRA,
146 why not keep the internal version for convenience?
147 The question whether there is sense in having a fall-back pager in all
148 the command line tools, for the cases when
149 .Pn more
150 or
151 .Pn less
152 aren't available, appears to be ridiculous.
153 Of course, MSAs and MRAs are more complex than text pagers
154 and not necessarily available but still the concept of orthogonal
155 design holds: ``Write programs that do one thing and do it well.''
156 .[
157 mcilroy unix phil
158 p. 53
159 .]
160 .[
161 mcilroy bstj foreword
162 .]
163 Here, this part of the Unix philosophy was applied not only
164 to the programs but to the project itself.
165 In other words:
166 ``Develop projects that focus on one thing and do it well.''
167 Projects grown complex should be split for the same reasons programs grown
168 complex should be split.
169 If it is conceptionally more elegant to have the MSA and MRA as
170 separate projects then they should be separated.
171 This is the case here, in my opinion.
172 The RFCs propose this separation by clearly distinguishing the different
173 mail handling tasks.
174 .[
175 rfc 821
176 .]
177 The small interfaces between the mail agents support the separation.
178 .P
179 In the beginning, email had been small and simple.
180 At that time,
181 .Pn /bin/mail
182 had covered anything there was to email and still had been small
183 and simple.
184 Later, the essential complexity of email increased.
185 (Essential complexity is the complexity defined by the problem itself.\0
186 .[[
187 brooks no silver bullet
188 .]])
189 Email systems reacted to this change: They grew.
190 RFCs started to introduce the concept of mail agents to separate the
191 various tasks because they became more extensive and new tasks appeared.
192 As the mail systems grew even more, parts were split off.
193 In nmh, for instance, the POP server, which was included in the original
194 MH, was removed.
195 Now is the time to go one step further and split the MSA and MRA off, too.
196 Not only does this decrease the code size of the project,
197 but, more important, it unburdens mmh of the whole field of
198 message transfer with all its implications for the project.
199 There is no more need to concern with changes in network transfer.
200 This independence is received by depending on an external program
201 that covers the field.
202 Today, this is a reasonable exchange.
203 .P
204 Functionality can be added in three different ways:
205 .BU
206 Implementing the function originally in the project.
207 .BU
208 Depending on a library that provides the function.
209 .BU
210 Depending on a program that provides the function.
211 .P
212 Whereas adding the function originally to the project increases the
213 code size most and requires most maintenance and development work,
214 it makes the project most independent of other software.
215 Using libraries or external programs require less maintenance work
216 but introduces dependencies on external software.
217 Programs have the smallest interfaces and provide the best separation
218 but possibly limit the information exchange.
219 External libraries are stronger connected than external programs,
220 thus information can be exchanged more flexible.
221 Adding code to a project increases maintenance work.
222 .\" XXX ref
223 Implementing complex functions originally in the project adds
224 a lot of code.
225 This should be avoided if possible.
226 Hence, the dependencies only change in kind, not in their existence.
227 In mmh, library dependencies on
228 .Pn libsasl2
229 and
230 .Pn libcrypto /\c
231 .Pn libssl
232 were treated against program dependencies on an MSA and an MRA.
233 This also meant treating build-time dependencies against run-time
234 dependencies.
235 Besides program dependencies providing the stronger separation
236 and being more flexible, they also allowed
237 over 6\|000 lines of code to be removed from mmh.
238 This made mmh's code base about 12\|% smaller.
239 Reducing the project's code size by such an amount without actually
240 losing functionality is a convincing argument.
241 Actually, as external MSAs and MRAs are likely superior to the
242 project's internal versions, the common user even gains functionality.
243 .P
244 Users of MH should not have problems to set up an external MSA and MRA.
245 Also, the popular MSAs and MRAs have large communities and a lot
246 of documentation available.
247 Choices for MSAs range from full-featured MTAs like
248 .I Postfix
249 over mid-size MTAs like
250 .I masqmail
251 and
252 .I dma
253 to small forwarders like
254 .I ssmtp
255 and
256 .I nullmailer .
257 Choices for MRAs include
258 .I fetchmail ,
259 .I getmail ,
260 .I mpop
261 and
262 .I fdm .
265 .H2 "Non-MUA Tools
266 .P
267 One goal of mmh is to remove the tools that are not part of the MUA's task.
268 Further more, any tools that don't improve the MUA's job significantly
269 should be removed.
270 Loosely related and rarely used tools distract from the lean appearance.
271 They require maintenance work without adding much to the core task.
272 By removing these tools, the project shall become more streamlined
273 and focused.
274 In mmh the following tools are not available anymore:
275 .BU
276 .Pn conflict
277 was removed
278 .Ci 8b235097cbd11d728c07b966cf131aa7133ce5a9
279 because it is a mail system maintenance tool that is not MUA-related.
280 It even checked
281 .Fn /etc/passwd
282 and
283 .Fn /etc/group
284 for consistency, which is completely unrelated to email.
285 A tool like
286 .Pn conflict
287 is surely useful, but it should not be shipped with mmh.
288 .\" XXX historic reasons?
289 .BU
290 .Pn rcvtty
291 was removed
292 .Ci 14767c94b3827be7c867196467ed7aea5f6f49b0
293 because its use case of writing to the user's terminal
294 on receiving of mail is obsolete.
295 If users like to be informed of new mail, the shell's
296 .Ev MAILPATH
297 variable or graphical notifications are technically more appealing.
298 Writing directly to terminals is hardly ever wanted today.
299 If though one wants to have it this way, the standard tool
300 .Pn write
301 can be used in a way similar to:
302 .VS
303 scan -file - | write `id -un`
304 VE
305 .BU
306 .Pn viamail
307 was removed
308 .Ci eda72d6a7a7c20ff123043fb7f19c509ea01f932
309 when the new attachment system was activated, because
310 .Pn forw
311 could then cover the task itself.
312 The program
313 .Pn sendfiles
314 was rewritten as a shell script wrapper around
315 .Pn forw .
316 .Ci 0e82199cf3c991a173e0ac8aa776efdb3ded61e6
317 .BU
318 .Pn msgchk
319 was removed
320 .Ci bb9360ead7eb7a3fedcce2eeedfc660014e41dbe ,
321 because it lost its use case when POP support was removed.
322 A call to
323 .Pn msgchk
324 provided hardly more information than:
325 .VS
326 ls -l /var/mail/meillo
327 VE
328 It did distinguish between old and new mail, but
329 this detail information can be retrieved with
330 .Pn stat (1),
331 too.
332 A small shell script could be written to print the information
333 in a similar way, if truly necessary.
334 As mmh's
335 .Pn inc
336 only incorporates mail from the user's local maildrop,
337 and thus no data transfers over slow networks are involved,
338 there's hardly any need to check for new mail before incorporating it.
339 .BU
340 .Pn msh
341 was removed
342 .Ci 916690191222433a6923a4be54b0d8f6ac01bd02
343 because the tool was in conflict with the philosophy of MH.
344 It provided an interactive shell to access the features of MH,
345 but it wasn't just a shell, tailored to the needs of mail handling.
346 Instead it was one large program that had several MH tools built in.
347 This conflicts with the major feature of MH of being a tool chest.
348 .Pn msh 's
349 main use case had been accessing Bulletin Boards, which have seized to
350 be popular.
351 .P
352 Removing
353 .Pn msh ,
354 together with the truly archaic code relicts
355 .Pn vmh
356 and
357 .Pn wmh ,
358 saved more than 7\|000 lines of C code \(en
359 about 15\|% of the project's original source code amount.
360 Having less code \(en with equal readability, of course \(en
361 for the same functionality is an advantage.
362 Less code means less bugs and less maintenance work.
363 As
364 .Pn rcvtty
365 and
366 .Pn msgchk
367 are assumed to be rarely used and can be implemented in different ways,
368 why should one keep them?
369 Removing them streamlines mmh.
370 .Pn viamail 's
371 use case is now partly obsolete and partly covered by
372 .Pn forw ,
373 hence there's no reason to still maintain it.
374 .Pn conflict
375 is not related to the mail client, and
376 .Pn msh
377 conflicts with the basic concept of MH.
378 Theses two tools might still be useful, but they should not be part of mmh.
379 .P
380 Finally, there's
381 .Pn slocal .
382 .Pn slocal
383 is an MDA and thus not directly MUA-related.
384 It should be removed from mmh, because including it conflicts with
385 the idea that mmh is a MUA only.
386 .Pn slocal
387 should rather become a separate project.
388 However,
389 .Pn slocal
390 provides rule-based processing of messages, like filing them into
391 different folders, which is otherwise not available in mmh.
392 Although
393 .Pn slocal
394 does neither pull in dependencies nor does it include a separate
395 technical area (cf. Sec. XXX), still,
396 it accounts for about 1\|000 lines of code that need to be maintained.
397 As
398 .Pn slocal
399 is almost self-standing, it should be split off into a separate project.
400 This would cut the strong connection between the MUA mmh and the MDA
401 .Pn slocal .
402 For anyone not using MH,
403 .Pn slocal
404 would become yet another independent MDA, like
405 .I procmail .
406 Then
407 .Pn slocal
408 could be installed without the complete MH system.
409 Likewise, mmh users could decide to use
410 .I procmail
411 without having a second, unused MDA,
412 .Pn slocal ,
413 installed.
414 That appears to be conceptionally the best solution.
415 Yet,
416 .Pn slocal
417 is not split off.
418 I defer the decision over
419 .Pn slocal
420 in need for deeper investigation.
421 In the meanwhile, it remains part of mmh.
422 That does not hurt because
423 .Pn slocal
424 is unrelated to the rest of the project.
428 .H2 "Displaying Messages
429 .P
430 Since the very beginning, already in the first concept paper,
431 .Pn show
432 had been MH's message display program.
433 .Pn show
434 mapped message numbers and sequences to files and invoked
435 .Pn mhl
436 to have the files formatted.
437 With MIME, this approach wasn't sufficient anymore.
438 MIME messages can consist of multiple parts. Some parts are not
439 directly displayable and text content might be encoded in
440 foreign charsets.
441 .Pn show 's
442 understanding of messages and
443 .Pn mhl 's
444 display capabilities couldn't cope with the task any longer.
445 .P
446 Instead of extending these tools, additional tools were written from
447 scratch and added to the MH tool chest.
448 Doing so is encouraged by the tool chest approach.
449 Modular design is a great advantage for extending a system,
450 as new tools can be added without interfering with existing ones.
451 First, the new MIME features were added in form of the single program
452 .Pn mhn .
453 The command
454 .Cl "mhn -show 42
455 would show the MIME message numbered 42.
456 With the 1.0 release of nmh in February 1999, Richard Coleman finished
457 the split of
458 .Pn mhn
459 into a set of specialized tools, which together covered the
460 multiple aspects of MIME.
461 One of them was
462 .Pn mhshow ,
463 which replaced
464 .Cl "mhn -show" .
465 It was capable of displaying MIME messages appropriately.
466 .P
467 From then on, two message display tools were part of nmh,
468 .Pn show
469 and
470 .Pn mhshow .
471 To ease the life of users,
472 .Pn show
473 was extended to automatically hand the job over to
474 .Pn mhshow
475 if displaying the message would be beyond
476 .Pn show 's
477 abilities.
478 In consequence, the user would simply invoke
479 .Pn show
480 (possibly through
481 .Pn next
482 or
483 .Pn prev )
484 and get the message printed with either
485 .Pn show
486 or
487 .Pn mhshow ,
488 whatever was more appropriate.
489 .P
490 Having two similar tools for essentially the same task is redundant.
491 Usually,
492 users wouldn't distinguish between
493 .Pn show
494 and
495 .Pn mhshow
496 in their daily mail reading.
497 Having two separate display programs was therefore mainly unnecessary
498 from a user's point of view.
499 Besides, the development of both programs needed to be in sync,
500 to ensure that the programs behaved in a similar way,
501 because they were used like a single tool.
502 Different behavior would have surprised the user.
503 .P
504 Today, non-MIME messages are rather seen to be a special case of
505 MIME messages, although it is the other way round.
506 As
507 .Pn mhshow
508 had already be able to display non-MIME messages, it appeared natural
509 to drop
510 .Pn show
511 in favor of using
512 .Pn mhshow
513 exclusively.
514 .Ci 4c1efddfd499300c7e74263e57d8aa137e84c853
515 Removing
516 .Pn show
517 is no loss in function, because functionally
518 .Pn mhshow
519 covers it completely.
520 The old behavior of
521 .Pn show
522 can still be emulated with the simple command line:
523 .VS
524 mhl `mhpath c`
525 VE
526 .P
527 For convenience,
528 .Pn mhshow
529 was renamed to
530 .Pn show
531 after
532 .Pn show
533 was gone.
534 It is clear that such a rename may confuse future developers when
535 trying to understand the history.
536 Nevertheless, I consider the convenience on the user's side,
537 to call
538 .Pn show
539 when they want a message to be displayed, to outweigh the inconvenience
540 on the developer's side when understanding the project history.
541 .P
542 To prepare for the transition,
543 .Pn mhshow
544 was reworked to behave more like
545 .Pn show
546 first.
547 (cf. Sec. XXX)
548 Once the tools behaved more alike, the replacing appeared to be
549 even more natural.
550 Today, mmh's new
551 .Pn show
552 became the one single message display program again, with the difference
553 that today it handles MIME messages as well as non-MIME messages.
554 The outcome of the transition is one program less to maintain,
555 no second display program for users to deal with,
556 and less system complexity.
557 .P
558 Still, removing the old
559 .Pn show
560 hurts in one regard: It had been such a simple program.
561 Its lean elegance is missing to the new
562 .Pn show .
563 But there is no chance;
564 supporting MIME demands for higher essential complexity.
566 .ig
567 XXX
568 Consider including text on scan listings here
570 Scan listings shall not contain body content. Hence, removed this feature.
571 Scan listings shall operator on message headers and non-message information
572 only. Displaying the beginning of the body complicates everything too much.
573 That's no surprise, because it's something completely different. If you
574 want to examine the body, then use show(1)/mhshow(1).
575 Changed the default scan formats accordingly.
576 .Ci 70b2643e0da8485174480c644ad9785c84f5bff4
577 ..
582 .H2 "Configure Options
583 .P
584 Customization is a double-edged sword.
585 It allows better suiting setups, but not for free.
586 There is the cost of code complexity to be able to customize.
587 There is the cost of less tested setups, because there are
588 more possible setups and especially corner-cases.
589 And, there is the cost of choice itself.
590 The code complexity directly affects the developers.
591 Less tested code affects both, users and developers.
592 The problem of choice affects the users, for once by having to
593 choose, but also by more complex interfaces that require more documentation.
594 Whenever options add little advantages, they should be considered for
595 removal.
596 I have reduced the number of project-specific configure options from
597 fifteen to three.
599 .U3 "Mail Transfer Facilities
600 .P
601 With the removal of the mail transfer facilities five configure
602 options vanished:
603 .P
604 The switches
605 .Sw --with-tls
606 and
607 .Sw --with-cyrus-sasl
608 had activated the support for transfer encryption and authentication.
609 This is not needed anymore.
610 .Ci fecd5d34f65597a4dfa16aeabea7d74b191532c3
611 .Ci 156d35f6425bea4c1ed3c4c79783dc613379c65b
612 .P
613 The configure switch
614 .Sw --enable-pop
615 activated the message retrieval facility.
616 The code area that would be conditionally compiled in for TLS and SASL
617 support had been small.
618 The conditionally compiled code area for POP support had been much larger.
619 Whereas the code base changes would only slightly change on toggling
620 TLS or SASL support, it changed much on toggling POP support.
621 The changes in the code base could hardly be overviewed.
622 By having POP support togglable a second code base had been created,
623 one that needed to be tested.
624 This situation is basically similar for the conditional TLS and SASL
625 code, but there the changes are minor and can yet be overviewed.
626 Still, conditional compilation of a code base creates variations
627 of the original program.
628 More variations require more testing and maintenance work.
629 .P
630 Two other options only specified default configuration values:
631 .Sw --with-mts
632 defined the default transport service, either
633 .Ar smtp
634 or
635 .Ar sendmail .
636 In mmh this fixed to
637 .Ar sendmail .
638 .Ci f6aa95b724fd8c791164abe7ee5468bf5c34f226
639 With
640 .Sw --with-smtpservers
641 default SMTP servers for the
642 .Ar smtp
643 transport service could be specified.
644 .Ci 128545e06224233b7e91fc4c83f8830252fe16c9
645 Both of them became irrelevant.
647 .U3 "Backup Prefix
648 .P
649 The backup prefix is the string that was prepended to message
650 filenames to tag them as deleted.
651 By default it had been the comma character `\f(CW,\fP'.
652 In July 2000, Kimmo Suominen introduced
653 the configure option
654 .Sw --with-hash-backup
655 to change the default to the hash symbol `\f(CW#\fP'.
656 The choice was probably personal preference, because first, the
657 option was named
658 .Sw --with-backup-prefix.
659 and had the prefix symbol as argument.
660 But giving the hash symbol as argument caused too many problems
661 for Autoconf,
662 thus the option was limited to use the hash symbol as the default prefix.
663 This supports the assumption, that the choice for the hash was
664 personal preference only.
665 Being related or not, words that start with the hash symbol
666 introduce a comment in the Unix shell.
667 Thus, the command line
668 .Cl "rm #13 #15
669 calls
670 .Pn rm
671 without arguments because the first hash symbol starts the comment
672 that reaches until the end of the line.
673 To delete the backup files,
674 .Cl "rm ./#13 ./#15"
675 needs to be used.
676 Using the hash as backup prefix can be seen as a precaution against
677 data loss.
678 .P
679 I removed the configure option but added the profile entry
680 .Pe backup-prefix ,
681 which allows to specify an arbitrary string as backup prefix.
682 .Ci 6c40d481d661d532dd527eaf34cebb6d3f8ed086
683 Profile entries are the common method to change mmh's behavior.
684 This change did not remove the choice but moved it to a location where
685 it suited better.
686 .P
687 Eventually, however, the new trash folder concept
688 .Cf "Sec. XXX
689 obsoleted the concept of the backup prefix completely.
690 .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
691 .Ci ca0b3e830b86700d9e5e31b1784de2bdcaf58fc5
694 .U3 "Editor and Pager
695 .P
696 The two configure options
697 .CW --with-editor=EDITOR
698 .CW --with-pager=PAGER
699 were used to specify the default editor and pager at configure time.
700 Doing so at configure time made sense in the eighties,
701 when the set of available editors and pagers varied much across
702 different systems.
703 Today, the situation is more homogeneous.
704 The programs
705 .Pn vi
706 and
707 .Pn more
708 can be expected to be available on every Unix system,
709 as they are specified by POSIX since two decades.
710 (The specifications for
711 .Pn vi
712 and
713 .Pn more
714 appeared in
715 .[
716 posix 1987
717 .]
718 and,
719 .[
720 posix 1992
721 .]
722 respectively.)
723 As a first step, these two tools were hard-coded as defaults.
724 .Ci 5d43a99db70c12a673028c7758c20cbe3e13ef5f
725 Not changed were the
726 .Pe editor
727 and
728 .Pe moreproc
729 profile entries, which allowed the user to override the system defaults.
730 Later, the concept was reworked to respect the standard environment
731 variables
732 .Ev VISUAL
733 and
734 .Ev PAGER
735 if they are set.
736 Today, mmh determines the editor to use in the following order,
737 taking the first available and non-empty item:
738 .IP (1)
739 Environment variable
740 .Ev MMHEDITOR
741 .IP (2)
742 Profile entry
743 .Pe Editor
744 .IP (3)
745 Environment variable
746 .Ev VISUAL
747 .IP (4)
748 Environment variable
749 .Ev EDITOR
750 .IP (5)
751 Command
752 .Pn vi .
753 .P
754 .Ci f85f4b7ae62e3d05a945dcd46ead51f0a2a89a9b
755 .P
756 The pager to use is determined in a similar order,
757 also taking the first available and non-empty item:
758 .IP (1)
759 Environment variable
760 .Ev MMHPAGER
761 .IP (2)
762 Profile entry
763 .Pe Pager
764 (replaces
765 .Pe moreproc )
766 .IP (3)
767 Environment variable
768 .Ev PAGER
769 .IP (4)
770 Command
771 .Pn more .
772 .P
773 .Ci 0c4214ea2aec6497d0d67b436bbee9bc1d225f1e
774 .P
775 By respecting the
776 .Ev VISUAL /\c
777 .Ev EDITOR
778 and
779 .Ev PAGER
780 environment variables,
781 the new behavior confirms better to the common style on Unix systems.
782 Additionally, the new approach is more uniform and clearer to users.
785 .U3 "ndbm
786 .P
787 .Pn slocal
788 used to depend on
789 .I ndbm ,
790 a database library.
791 The database is used to store the `\fLMessage-ID\fP's of all
792 messages delivered.
793 This enables
794 .Pn slocal
795 to suppress delivering the same message to the same user twice.
796 (This features was enabled by the
797 .Sw -suppressdup
798 switch.)
799 .P
800 A variety of versions of the database library exist.
801 .[
802 wolter unix incompat notes dbm
803 .]
804 Complicated autoconf code was needed to detect them correctly.
805 Further more, the configure switches
806 .Sw --with-ndbm=ARG
807 and
808 .Sw --with-ndbmheader=ARG
809 were added to help with difficult setups that would
810 not be detected automatically or correctly.
811 .P
812 By removing the suppress duplicates feature of
813 .Pn slocal ,
814 the dependency on
815 .I ndbm
816 vanished and 120 lines of complex autoconf code could be saved.
817 .Ci ecd6d6a20cb7a1507e3a20d6c4cb3a1cf14c6bbf
818 The change removed functionality too, but that is minor to the
819 improvement by dropping the dependency and the complex autoconf code.
821 .U3 "mh-e Support
822 .P
823 The configure option
824 .Sw --disable-mhe
825 was removed when the mh-e support was reworked.
826 Mh-e is the Emacs front-end to MH.
827 It requires MH to provide minor additional functions.
828 The
829 .Sw --disable-mhe
830 configure option could switch these extensions off.
831 After removing the support for old versions of mh-e,
832 only the
833 .Sw -build
834 switches of
835 .Pn forw
836 and
837 .Pn repl
838 are left to be mh-e extensions.
839 They are now always built in because they add little code and complexity.
840 In consequence, the
841 .Sw --disable-mhe
842 configure option was removed
843 .Ci a7ce7b4a580d77b6c2c4d980812beb589aa4c643
844 Removing the option removed a second code setup that would have
845 needed to be tested.
846 This change was first done in nmh and thereafter merged into mmh.
847 .P
848 The interface changes in mmh require mh-e to be adjusted in order
849 to be able to use mmh as back-end.
850 This will require minor changes to mh-e, but removing the
851 .Sw -build
852 switches would require more rework.
854 .U3 "Masquerading
855 .P
856 The configure option
857 .Sw --enable-masquerade
858 could take up to three arguments:
859 `draft_from', `mmailid', and `username_extension'.
860 They activated different types of address masquerading.
861 All of them were implemented in the SMTP-speaking
862 .Pn post
863 command, which provided an MSA.
864 Address masquerading is an MTA's task and mmh does not cover
865 this field anymore.
866 Hence, true masquerading needs to be implemented in the external MTA.
867 .P
868 The
869 .I mmailid
870 masquerading type is the oldest one of the three and the only one
871 available in the original MH.
872 It provided a
873 .I username
874 to
875 .I fakeusername
876 mapping, based on the password file's GECOS field.
877 The man page
878 .Mp mh-tailor(5)
879 described the use case as being the following:
880 .QS
881 This is useful if you want the messages you send to always
882 appear to come from the name of an MTA alias rather than your
883 actual account name. For instance, many organizations set up
884 `First.Last' sendmail aliases for all users. If this is
885 the case, the GECOS field for each user should look like:
886 ``First [Middle] Last <First.Last>''
887 .QE
888 .P
889 As mmh sends outgoing mail via the local MTA only,
890 the best location to do such global rewrites is there.
891 Besides, the MTA is conceptionally the right location because it
892 does the reverse mapping for incoming mail (aliasing), too.
893 Further more, masquerading set up there is readily available for all
894 mail software on the system.
895 Hence, mmailid masquerading was removed.
896 .Ci 0836c8000ccb34b59410ef1c15b1b7feac70ce5f
897 .P
898 The
899 .I username_extension
900 masquerading type did not replace the username but would append a suffix,
901 specified by the
902 .Ev USERNAME_EXTENSION
903 environment variable, to it.
904 This provided support for the
905 .I user-extension
906 feature of qmail and the similar
907 .I "plussed user
908 processing of sendmail.
909 The decision to remove this username_extension masquerading was
910 motivated by the fact that
911 .Pn spost
912 hadn't supported it already.
913 .Ci 2abae0bfd0ad5bf898461e50aa4b466d641f23d9
914 Username extensions are possible in mmh, but less convenient to use.
915 .\" XXX format file %(getenv USERNAME_EXTENSION)
916 .P
917 The
918 .I draft_from
919 masquerading type instructed
920 .Pn post
921 to use the value of the
922 .Hd From
923 header field as SMTP envelope sender.
924 Sender addresses could be replaced completely.
925 .Ci b14ea6073f77b4359aaf3fddd0e105989db9
926 Mmh offers a kind of masquerading similar in effect, but
927 with technical differences.
928 As mmh does not transfer messages itself, the local MTA has final control
929 over the sender's address. Any masquerading mmh introduces may be reverted
930 by the MTA.
931 In times of pedantic spam checking, an MTA will take care to use
932 sensible envelope sender addresses to keep its own reputation up.
933 Nonetheless, the MUA can set the
934 .Hd From
935 header field and thereby propose
936 a sender address to the MTA.
937 The MTA may then decide to take that one or generate the canonical sender
938 address for use as envelope sender address.
939 .P
940 In mmh, the MTA will always extract the recipient and sender from the
941 message header (\c
942 .Pn sendmail 's
943 .Sw -t
944 switch).
945 The
946 .Hd From
947 header field of the draft may be set arbitrary by the user.
948 If it is missing, the canonical sender address will be generated by the MTA.
950 .U3 "Remaining Options
951 .P
952 Two configure options remain in mmh.
953 One is the locking method to use:
954 .Sw --with-locking=[dot|fcntl|flock|lockf] .
955 The idea of removing all methods except the portable dot locking
956 and having that one as the default is appealing, but this change
957 requires deeper technical investigation into the topic.
958 The other option,
959 .Sw --enable-debug ,
960 compiles the programs with debugging symbols and does not strip them.
961 This option is likely to stay.
966 .H2 "Command Line Switches
967 .P
968 The command line switches of MH tools follow the X Window style.
969 They are words, introduced by a single dash.
970 For example:
971 .Cl "-truncate" .
972 Every program in mmh has two generic switches:
973 .Sw -help ,
974 to print a short message on how to use the program, and
975 .Sw -Version ,
976 to tell what version of mmh the program belongs to.
977 .P
978 Switches change the behavior of programs.
979 Programs that do one thing in one way require no switches.
980 In most cases, doing something in exactly one way is too limiting.
981 If there is basically one task to accomplish, but it should be done
982 in various ways, switches are a good approach to alter the behavior
983 of a program.
984 Changing the behavior of programs provides flexibility and customization
985 to users, but at the same time it complicates the code, documentation and
986 usage of the program.
987 .\" XXX: Ref
988 Therefore, the number of switches should be kept small.
989 A small set of well-chosen switches does no harm.
990 But usually, the number of switches increases over time.
991 Already in 1985, Rose and Romine have identified this as a major
992 problem of MH:
993 .[ [
994 rose romine real work
995 .], p. 12]
996 .QS
997 A complaint often heard about systems which undergo substantial development
998 by many people over a number of years, is that more and more options are
999 introduced which add little to the functionality but greatly increase the
1000 amount of information a user needs to know in order to get useful work done.
1001 This is usually referred to as creeping featurism.
1002 .QP
1003 Unfortunately MH, having undergone six years of off-and-on development by
1004 ten or so well-meaning programmers (the present authors included),
1005 suffers mightily from this.
1006 .QE
1007 .P
1008 Being reluctant to adding new switches \(en or `options',
1009 as Rose and Romine call them \(en is one part of a counter-action,
1010 the other part is removing hardly used switches.
1011 Nmh's tools had lots of switches already implemented,
1012 hence, cleaning up by removing some of them was the more important part
1013 of the counter-action.
1014 Removing existing functionality is always difficult because it
1015 breaks programs that use these functions.
1016 Also, for every obsolete feature, there'll always be someone who still
1017 uses it and thus opposes its removal.
1018 This puts the developer into the position,
1019 where sensible improvements to style are regarded as destructive acts.
1020 Yet, living with the featurism is far worse, in my eyes, because
1021 future needs will demand adding further features,
1022 worsening the situation more and more.
1023 Rose and Romine added in a footnote,
1024 ``[...]
1025 .Pn send
1026 will no doubt acquire an endless number of switches in the years to come.''
1027 Although clearly humorous, the comment points to the nature of the problem.
1028 Refusing to add any new switches would encounter the problem at its root,
1029 but this is not practical.
1030 New needs will require new switches and it would be unwise to block
1031 them strictly.
1032 Nevertheless, removing obsolete switches still is an effective approach
1033 to deal with the problem.
1034 Working on an experimental branch without an established user base,
1035 eased my work because I did not offend users when I removed existing
1036 functions.
1037 .P
1038 Rose and Romine counted 24 visible and 9 more hidden switches for
1039 .Pn send .
1040 In nmh, they increased up to 32 visible and 12 hidden ones.
1041 At the time of writing, no more than 7 visible switches and 1 hidden switch
1042 have remained in mmh's
1043 .Pn send .
1044 (These numbers include two generic switches, help and version.)
1045 .P
1046 Fig. XXX
1047 .\" XXX Ref
1048 displays the number of switches for each of the tools that is available
1049 in both, nmh and mmh.
1050 The tools are sorted by the number of switches they had in nmh.
1051 Visible and hidden switches were counted,
1052 but not the generic help and version switches.
1053 Whereas in the beginning of the project, the average tool had 11 switches,
1054 now it has no more than 5 \(en only half as many.
1055 If the `no' switches and similar inverse variant are folded onto
1056 their counter-parts, the average tool had 8 switches in pre-mmh times and
1057 has 4 now.
1058 The total number of functional switches in mmh dropped from 465
1059 to 234.
1061 .KS
1062 .in 1c
1063 .so input/switches.grap
1064 .KE
1066 .P
1067 A part of the switches vanished after functions were removed.
1068 This was the case for network mail transfer, for instance.
1069 Sometimes, however, the work flow was the other way:
1070 I looked through the
1071 .Mp mh-chart (7)
1072 man page to identify the tools with apparently too many switches.
1073 Then considering the value of each of the switches by examining
1074 the tool's man page and source code, aided by recherche and testing.
1075 This way, the removal of functions was suggested by the aim to reduce
1076 the number of switches per command.
1079 .U3 "Draft Folder Facility
1080 .P
1081 A change early in the project was the complete transition from
1082 the single draft message to the draft folder facility.
1083 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860
1084 The draft folder facility was introduced in the mid-eighties, when
1085 Rose and Romine called it a ``relatively new feature''.
1086 .[
1087 rose romine real work
1088 .]
1089 Since then, the facility had existed but was inactive by default.
1090 The default activation and the related rework of the tools made it
1091 possible to remove the
1092 .Sw -[no]draftfolder ,
1093 and
1094 .Sw -draftmessage
1095 switches from
1096 .Pn comp ,
1097 .Pn repl ,
1098 .Pn forw ,
1099 .Pn dist ,
1100 .Pn whatnow ,
1101 and
1102 .Pn send .
1103 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860
1104 The only flexibility removed with this change is having multiple
1105 draft folders within one profile.
1106 I consider this a theoretical problem only.
1107 In the same go, the
1108 .Sw -draft
1109 switch of
1110 .Pn anno ,
1111 .Pn refile ,
1112 and
1113 .Pn send
1114 was removed.
1115 The special-casing of `the' draft message became irrelevant after
1116 the rework of the draft system.
1117 (See Sec. XXX.)
1118 Equally,
1119 .Pn comp
1120 lost its
1121 .Sw -file
1122 switch.
1123 The draft folder facility, together with the
1124 .Sw -form
1125 switch, are sufficient.
1128 .U3 "In Place Editing
1129 .P
1130 .Pn anno
1131 had the switches
1132 .Sw -[no]inplace
1133 to either annotate the message in place and thus preserve hard links,
1134 or annotate a copy to replace the original message, breaking hard links.
1135 Following the assumption that linked messages should truly be the
1136 same message, and annotating it should not break the link, the
1137 .Sw -[no]inplace
1138 switches were removed and the previous default
1139 .Sw -inplace
1140 was made the only behavior.
1141 .Ci c8195849d2e366c569271abb0f5f60f4ebf0b4d0
1142 The
1143 .Sw -[no]inplace
1144 switches of
1145 .Pn repl ,
1146 .Pn forw ,
1147 and
1148 .Pn dist
1149 could be removed, too, as they were simply passed through to
1150 .Pn anno .
1151 .P
1152 .Pn burst
1153 also had
1154 .Sw -[no]inplace
1155 switches, but with different meaning.
1156 With
1157 .Sw -inplace ,
1158 the digest had been replaced by the table of contents (i.e. the
1159 introduction text) and the burst messages were placed right
1160 after this message, renumbering all following messages.
1161 Also, any trailing text of the digest was lost, though,
1162 in practice, it usually consists of an end-of-digest marker only.
1163 Nontheless, this behavior appeared less elegant than the
1164 .Sw -noinplace
1165 behavior, which already had been the default.
1166 Nmh's
1167 .Mp burst (1)
1168 man page reads:
1169 .sp \n(PDu
1170 .QS
1171 If -noinplace is given, each digest is preserved, no table
1172 of contents is produced, and the messages contained within
1173 the digest are placed at the end of the folder. Other messages
1174 are not tampered with in any way.
1175 .QE
1176 .LP
1177 The decision to drop the
1178 .Sw -inplace
1179 behavior was supported by the code complexity and the possible data loss
1180 it caused.
1181 .Sw -noinplace
1182 was chosen to be the definitive behavior.
1183 .Ci 68a686adeb39223a5e1ad35e4a24890ec053679d
1186 .U3 "Forms and Format Strings
1187 .P
1188 Historically, the tools that had
1189 .Sw -form
1190 switches to supply a form file had
1191 .Sw -format
1192 switches as well to supply the contents of a form file as a string
1193 on the command line directly.
1194 In consequence, the following two lines equaled:
1195 .VS
1196 scan -form scan.mailx
1197 scan -format "`cat .../scan.mailx`"
1198 VE
1199 The
1200 .Sw -format
1201 switches were dropped in favor for extending the
1202 .Sw -form
1203 switches.
1204 .Ci f51956be123db66b00138f80464d06f030dbb88d
1205 If their argument starts with an equal sign (`='),
1206 then the rest of the argument is taken as a format string,
1207 otherwise the arguments is treated as the name of a format file.
1208 Thus, now the following two lines equal:
1209 .VS
1210 scan -form scan.mailx
1211 scan -form "=`cat .../scan.mailx`"
1212 VE
1213 This rework removed the prefix collision between
1214 .Sw -form
1215 and
1216 .Sw -format .
1217 Now, typing
1218 .Sw -fo
1219 suffices to specify form or format string.
1220 .P
1221 The different meaning of
1222 .Sw -format
1223 for
1224 .Pn repl
1225 and
1226 .Pn forw
1227 was removed in mmh.
1228 .Pn forw
1229 was completely switched to MIME-type forwarding, thus removing the
1230 .Sw -[no]format .
1231 .Ci 6e271608b7b9c23771523f88d23a4d3593010cf1
1232 For
1233 .Pn repl ,
1234 the
1235 .Sw -[no]format
1236 switches were reworked to
1237 .Sw -[no]filter
1238 switches.
1239 .Ci 67411b1f95d6ec987b4c732459e1ba8a8ac192c6
1240 The
1241 .Sw -format
1242 switches of
1243 .Pn send
1244 and
1245 .Pn post ,
1246 which had a third meaning,
1247 were removed likewise.
1248 .Ci f3cb7cde0e6f10451b6848678d95860d512224b9
1249 Eventually, the ambiguity of the
1250 .Sw -format
1251 switches was resolved by not anymore having any such switch in mmh.
1254 .U3 "MIME Tools
1255 .P
1256 The MIME tools, which were once part of
1257 .Pn mhn
1258 [sic!],
1259 had several switches that added little practical value to the programs.
1260 The
1261 .Sw -[no]realsize
1262 switches of
1263 .Pn mhbuild
1264 and
1265 .Pn mhlist
1266 were removed, doing real size calculations always now
1267 .Ci 8d8f1c3abc586c005c904e52c4adbfe694d2201c ,
1268 as
1269 ``This provides an accurate count at the expense of a small delay.''
1270 This small delay is not noticable on modern systems.
1271 .P
1272 The
1273 .Sw -[no]check
1274 switches were removed together with the support for
1275 .Hd Content-MD5
1276 header fields.
1277 .[
1278 rfc 1864
1279 .]
1280 .Ci 31dc797eb5178970d68962ca8939da3fd9a8efda
1281 (See Sec. XXX)
1282 .P
1283 The
1284 .Sw -[no]ebcdicsafe
1285 and
1286 .Sw -[no]rfc934mode
1287 switches of
1288 .Pn mhbuild
1289 were removed because they are considered obsolete.
1290 .Ci 01a3480928da485b4d6109d36d751dfa71799d58
1291 .Ci 3363e2624dce0eb8164cf8b3f1ab385c8ff72e88
1292 .P
1293 Content caching of external MIME parts, activated with the
1294 .Sw -rcache
1295 and
1296 .Sw -wcache
1297 switches was completely removed.
1298 .Ci d1fefd9f614e4dc3cda16da6c69133c1b2005269
1299 External MIME parts are rare today, having a caching facility
1300 for them is appears to be unnecessary.
1301 .P
1302 In pre-MIME times,
1303 .Pn mhl
1304 had covered many tasks that are part of MIME handling today.
1305 Therefore,
1306 .Pn mhl
1307 could be simplified to a large extend, reducing the number of its
1308 switches from 21 to 6.
1309 .Ci 350ad6d3542a07639213cf2a4fe524e829c1e7b6
1310 .Ci 0e46503be3c855bddaeae3843e1b659279c35d70
1315 .U3 "Header Printing
1316 .P
1317 .Pn folder 's
1318 data output is self-explaining enough that
1319 displaying the header line makes few sense.
1320 Hence, the
1321 .Sw -[no]header
1322 switch was removed and headers are never printed.
1323 .Ci 601cc73d1fa05ce96faa728f036d6c51b91701c7
1324 .P
1325 In
1326 .Pn mhlist ,
1327 the
1328 .Sw -[no]header
1329 switches were removed, too.
1330 .Ci b24f96523aaf60e44e04a3ffb1d22e69a13a602f
1331 But in this case headers are always printed,
1332 because the output is not self-explaining.
1333 .P
1334 .Pn scan
1335 also had
1336 .Sw -[no]header
1337 switches.
1338 Printing the header had been sensible until the introduction of
1339 format strings made it impossible to display the column headings.
1340 Only the folder name and the current date remained to be printed.
1341 As this information can be perfectly retrieved by
1342 .Pn folder
1343 and
1344 .Pn date ,
1345 consequently, the switches were removed.
1346 .Ci c477dc5d1d03fa6d9a8ab3dd3508c63cbddc044e
1347 .P
1348 By removing all
1349 .Sw -header
1350 switches, the collision with
1351 .Sw -help
1352 on the first two letters was resolved.
1353 Currently,
1354 .Sw -h
1355 evaluates to
1356 .Sw -help
1357 for all tools of mmh.
1360 .U3 "Suppressing Edits or the Invocation of the WhatNow Shell
1361 .P
1362 The
1363 .Sw -noedit
1364 switch of
1365 .Pn comp ,
1366 .Pn repl ,
1367 .Pn forw ,
1368 .Pn dist ,
1369 and
1370 .Pn whatnow
1371 was removed, but it can now be replaced by specifying
1372 .Sw -editor
1373 with an empty argument.
1374 .Ci 75fca31a5b9d5c1a99c74ab14c94438d8852fba9
1375 (Specifying
1376 .Cl "-editor true
1377 is nearly the same, only differing by the previous editor being set.)
1378 .P
1379 The more important change is the removal of the
1380 .Sw -nowhatnowproc
1381 switch.
1382 .Ci ee4f43cf2ef0084ec698e4e87159a94c01940622
1383 This switch had introduced an awkward behavior, as explained in nmh's
1384 man page for
1385 .Mp comp (1):
1386 .QS
1387 The \-editor editor switch indicates the editor to use for
1388 the initial edit. Upon exiting from the editor, comp will
1389 invoke the whatnow program. See whatnow(1) for a discussion
1390 of available options. The invocation of this program can be
1391 inhibited by using the \-nowhatnowproc switch. (In truth of
1392 fact, it is the whatnow program which starts the initial
1393 edit. Hence, \-nowhatnowproc will prevent any edit from
1394 occurring.)
1395 .QE
1396 .P
1397 Effectively, the
1398 .Sw -nowhatnowproc
1399 switch creates only a draft message.
1400 As
1401 .Cl "-whatnowproc true
1402 causes the same behavior, the
1403 .Sw -nowhatnowproc
1404 switch was removed for being redundant.
1405 Likely, the
1406 .Sw -nowhatnowproc
1407 switch was intended to be used by front-ends.
1411 .U3 "Various
1412 .BU
1413 With the removal of MMDF maildrop format support,
1414 .Pn packf
1415 and
1416 .Pn rcvpack
1417 no longer needed their
1418 .Sw -mbox
1419 and
1420 .Sw -mmdf
1421 switches.
1422 .Sw -mbox
1423 is the sole behavior now.
1424 .Ci 3916ab66ad5d183705ac12357621ea8661afd3c0
1425 In the same go,
1426 .Pn packf
1427 and
1428 .Pn rcvpack
1429 were reworked (see Sec. XXX) and their
1430 .Sw -file
1431 switch became unnecessary.
1432 .Ci ca1023716d4c2ab890696f3e41fa0d94267a940e
1434 .BU
1435 Mmh's tools will no longer clear the screen (\c
1436 .Pn scan 's
1437 and
1438 .Pn mhl 's
1439 .Sw -[no]clear
1440 switches
1441 .Ci e57b17343dcb3ff373ef4dd089fbe778f0c7c270
1442 .Ci 943765e7ac5693ae177fd8d2b5a2440e53ce816e ).
1443 Neither will
1444 .Pn mhl
1445 ring the bell (\c
1446 .Sw -[no]bell
1447 .Ci e11983f44e59d8de236affa5b0d0d3067c192e24 )
1448 nor page the output itself (\c
1449 .Sw -length
1450 .Ci 5b9d883db0318ed2b84bb82dee880d7381f99188 ).
1451 Generally, the pager to use is no longer specified with the
1452 .Sw -[no]moreproc
1453 command line switches for
1454 .Pn mhl
1455 and
1456 .Pn show /\c
1457 .Pn mhshow .
1458 .Ci 39e87a75b5c2d3572ec72e717720b44af291e88a
1460 .BU
1461 In order to avoid prefix collisions among switch names, the
1462 .Sw -version
1463 switch was renamed to
1464 .Sw -Version
1465 (with capital `V').
1466 .Ci 32b2354dbaf4bf934936eb5b102a4a3d2fdd209a
1467 Every program has the
1468 .Sw -version
1469 switch but its first three letters collided with the
1470 .Sw -verbose
1471 switch, present in many programs.
1472 The rename solved this problem once for all.
1473 Although this rename breaks a basic interface, having the
1474 .Sw -V
1475 abbreviation to display the version information, isn't all too bad.
1477 .BU
1478 .Sw -[no]preserve
1479 of
1480 .Pn refile
1481 was removed because what use was it anyway?
1482 .QS
1483 Normally when a message is refiled, for each destination
1484 folder it is assigned the number which is one above the current
1485 highest message number in that folder. Use of the
1486 \-preserv [sic!] switch will override this message renaming, and try
1487 to preserve the number of the message. If a conflict for a
1488 particular folder occurs when using the \-preserve switch,
1489 then refile will use the next available message number which
1490 is above the message number you wish to preserve.
1491 .QE
1493 .BU
1494 The removal of the
1495 .Sw -[no]reverse
1496 switches of
1497 .Pn scan
1498 .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
1499 is a bug fix, supported by the comments
1500 ``\-[no]reverse under #ifdef BERK (I really HATE this)''
1501 by Rose and
1502 ``Lists messages in reverse order with the `\-reverse' switch.
1503 This should be considered a bug.'' by Romine in the documentation.
1504 The question remains why neither Rose and Romine had fixed this
1505 bug in the eighties when they wrote these comments nor has anyone
1506 thereafter.
1509 .ig
1511 forw: [no]dashstuffing(mhl)
1513 mhshow: [no]pause [no]serialonly
1515 mhmail: resent queued
1516 inc: snoop, (pop)
1518 mhl: [no]faceproc folder sleep
1519 [no]dashstuffing(forw) digest list volume number issue number
1521 prompter: [no]doteof
1523 refile: [no]preserve [no]unlink [no]rmmproc
1525 send: [no]forward [no]mime [no]msgid
1526 [no]push split [no]unique (sasl) width snoop [no]dashstuffing
1527 attach attachformat
1528 whatnow: (noedit) attach
1530 slocal: [no]suppressdups
1532 spost: [no]filter [no]backup width [no]push idanno
1533 [no]check(whom) whom(whom)
1535 whom: ???
1537 ..
1540 .ig
1542 .P
1543 In the best case, all switches are unambiguous on the first character,
1544 or on the three-letter prefix for the `no' variants.
1545 Reducing switch prefix collisions, shortens the necessary prefix length
1546 the user must type.
1547 Having less switches helps best.
1549 ..
1552 .\" XXX: whatnow prompt commands
1557 .\" --------------------------------------------------------------
1558 .H1 "Modernizing
1559 .P
1560 In the over thirty years of MH's existence, its code base was
1561 extended more and more.
1562 New features entered the project and became alternatives to the
1563 existing behavior.
1564 Relicts from several decades have gathered in the code base,
1565 but seldom obsolete features were dropped.
1566 This section describes the removing of old code
1567 and the modernizing of the default setup.
1568 It focuses on the functional aspect only;
1569 the non-functional aspects of code style are discussed in
1570 .\" FIXME REF
1571 Sec. XXX.
1574 .H2 "Code Relicts
1575 .P
1576 My position to drop obsolete functions of mmh, in order to remove old code,
1577 is much more revolutional than the nmh community likes to have it.
1578 Working on an experimental version, I was able to quickly drop
1579 functionality I considered ancient.
1580 The need for consensus with peers would have slowed this process down.
1581 Without the need to justify my decisions, I was able to rush forward.
1582 In December 2011, Paul Vixie motivated the nmh developers to just
1583 do the work:
1584 .[
1585 paul vixie edginess nmh-workers
1586 .]
1587 .QS
1588 let's stop walking on egg shells with this code base. there's no need to
1589 discuss whether to keep using vfork, just note in [sic!] passing, [...]
1590 we don't need a separate branch for removing vmh
1591 or ridding ourselves of #ifdef's or removing posix replacement functions
1592 or depending on pure ansi/posix "libc".
1593 .QP
1594 these things should each be a day or two of work and the "main branch"
1595 should just be modern. [...]
1596 let's push forward, aggressively.
1597 .QE
1598 .LP
1599 I did so already in the months before.
1600 I pushed forward.
1601 I simply dropped the cruft.
1602 .P
1603 The decision to drop a feature was based on literature research and
1604 careful thinking, but whether having had contact to this particular
1605 feature within my own computer life served as a rule of thumb.
1606 Always, I explained my reasons in the commit messages
1607 in the version control system.
1608 Hence, others can comprehend my view and argue for undoing the change
1609 if I have missed an important aspect.
1610 I was quick in dropping parts.
1611 I rather re-included falsely dropped parts than going a slower pace.
1612 Mmh is experimental work; it required tough decisions.
1615 .U3 "Forking
1616 .P
1617 Being a tool chest, MH creates many processes.
1618 In earlier times
1619 .Fu fork()
1620 had been an expensive system call, because the process's image needed
1621 to be duplicated completely at once.
1622 This was especially painful in the common case when the image gets
1623 replaced by a call to
1624 .Fu exec()
1625 right after having forked the child process.
1626 The
1627 .Fu vfork()
1628 system call was invented to speed up this particular case.
1629 It completely omits the duplication of the image.
1630 On old systems this resulted in significant speed ups.
1631 Therefore MH used
1632 .Fu vfork()
1633 whenever possible.
1634 .P
1635 Modern memory management units support copy-on-write semantics, which make
1636 .Fu fork()
1637 almost as fast as
1638 .Fu vfork() .
1639 The man page of
1640 .Mp vfork (2)
1641 in FreeBSD 8.0 states:
1642 .QS
1643 This system call will be eliminated when proper system sharing mechanisms
1644 are implemented. Users should not depend on the memory sharing semantics
1645 of vfork() as it will, in that case, be made synonymous to fork(2).
1646 .QE
1647 .LP
1648 Vixie supports the removal with the note that ``the last
1649 system on which fork was so slow that an mh user would notice it, was
1650 Eunice. that was 1987''.
1651 .[
1652 nmh-workers vixie edginess
1653 .]
1654 I replaced all calls to
1655 .Fu vfork()
1656 with calls to
1657 .Fu fork() .
1658 .Ci 40821f5c1316e9205a08375e7075909cc9968e7d
1659 .P
1660 Related to the costs of
1661 .Fu fork()
1662 is the probability of its success.
1663 In the eighties, on heavy loaded systems, calls to
1664 .Fu fork()
1665 were prone to failure.
1666 Hence, many of the
1667 .Fu fork()
1668 calls in the code were wrapped into loops to retry the
1669 .Fu fork()
1670 several times, to increase the changes to succeed, eventually.
1671 On modern systems, a failing
1672 .Fu fork()
1673 call is unusual.
1674 Hence, in the rare case when
1675 .Fu fork()
1676 fails, mmh programs simply abort.
1677 .Ci 5fbf37ee68e018998ada61eeab73e035b26834b6
1680 .U3 "Header Fields
1681 .BU
1682 The
1683 .Hd Encrypted
1684 header field was introduced by RFC\|822,
1685 but already marked as legacy in RFC\|2822.
1686 Today, OpenPGP provides the basis for standardized exchange of encrypted
1687 messages [RFC\|4880, RFC\|3156].
1688 Hence, the support for
1689 .Hd Encrypted
1690 header fields is removed in mmh.
1691 .Ci 064527f7b57ab050e5af13e15ad99aeeab125857
1692 .BU
1693 Native support for
1694 .Hd Face
1695 header fields has been removed, as well.
1696 .Ci 8e5be81f784682822f5e868c1bf3c8624682bd23
1697 This feature is similar to the
1698 .Hd X-Face
1699 header field in its intent,
1700 but takes a different approach to store the image.
1701 Instead of encoding the image data directly into the header field,
1702 it contains the hostname and UDP port where the image
1703 date can be retrieved.
1704 There exists even a third Face system,
1705 which is the successor of
1706 .Hd X-Face ,
1707 although it re-uses the
1708 .Hd Face
1709 header field.
1710 It was invented in 2005 and supports colored PNG images.
1711 None of the Face systems described here is popular today.
1712 Hence, mmh has no direct support for them.
1713 .BU
1714 The
1715 .Hd Content-MD5
1716 header field was introduced by RFC\|1864.
1717 It provides detection of data corruption during the transfer.
1718 But it can not ensure verbatim end-to-end delivery of the contents
1719 [RFC\|1864].
1720 The proper approach to verify content integrity in an
1721 end-to-end relationship is the use of digital cryptography.
1722 .\" XXX (RFCs FIXME).
1723 On the other hand, transfer protocols should detect corruption during
1724 the transmission.
1725 The TCP includes a checksum field therefore.
1726 These two approaches in combinations render the
1727 .Hd Content-MD5
1728 header field superfluous.
1729 Not a single one out of 4\|200 messages from two decades
1730 in an nmh-workers mailing list archive contains a
1731 .Hd Content-MD5
1732 header field.
1733 Neither did any of the 60\|000 messages in my personal mail storage.
1734 Removing the support for this header field,
1735 removed the last place where MD5 computation was needed.
1736 .Ci 31dc797eb5178970d68962ca8939da3fd9a8efda
1737 Hence, the MD5 code could be removed as well.
1738 Over 500 lines of code vanished by this one change.
1741 .U3 "MMDF maildrop support
1742 .P
1743 This type of format is conceptionally similar to the mbox format,
1744 but uses a different message delimiter (`\fL\\1\\1\\1\\1\fP',
1745 commonly written as `\fL^A^A^A^A\fP', instead of `\fLFrom\0\fP').
1746 Mbox is the de-facto standard maildrop format on Unix,
1747 whereas the MMDF maildrop format became forgotten.
1748 I did drop MMDF maildrop format support.
1749 Mbox is the only packed mailbox format supported in mmh.
1750 .P
1751 The simplifications within the code were moderate.
1752 Mainly, the reading and writing of MMDF mailbox files was removed.
1753 But also, switches of
1754 .Pn packf
1755 and
1756 .Pn rcvpack
1757 could be removed.
1758 .Ci 3916ab66ad5d183705ac12357621ea8661afd3c0
1759 In the message parsing function
1760 .Fn sbr/m_getfld.c ,
1761 knowledge of MMDF packed mail boxes was removed.
1762 .Ci 684ec30d81e1223a282764452f4902ed4ad1c754
1763 Further code structure simplifications may be possible there,
1764 because only one single packed mailbox format is left to be supported.
1765 I have not worked on them yet because
1766 .Fu m_getfld()
1767 is heavily optimized and thus dangerous to touch.
1768 The risk of damaging the intricate workings of the optimized code is
1769 too high.
1770 .\" XXX: move somewhere else
1771 This problem is know to the developers of nmh, too.
1772 They also avoid touching this minefield.
1775 .U3 "Prompter's Control Keys
1776 .P
1777 The program
1778 .Pn prompter
1779 queries the user to fill in a message form.
1780 When used by
1781 .Pn comp
1782 as
1783 .Cl "comp -editor prompter" ,
1784 the resulting behavior is similar to
1785 .Pn mailx .
1786 Apparently,
1787 .Pn prompter
1788 hadn't been touched lately.
1789 Otherwise it's hardly explainable why it
1790 still offered the switches
1791 .Sw -erase
1792 .Ar chr
1793 and
1794 .Sw -kill
1795 .Ar chr
1796 to name the characters for command line editing.
1797 The times when this had been necessary are long time gone.
1798 Today these things work out-of-the-box, and if not, are configured
1799 with the standard tool
1800 .Pn stty .
1801 The switches are removed now
1802 .Ci 0bd9750710cdbab80cfb4036dd87af20afe1552f .
1805 .U3 "Hardcopy Terminal Support
1806 .P
1807 More of a funny anecdote is a check for being connected to a
1808 hardcopy terminal.
1809 It remained in the code until Spring 2012, when I finally removed it
1810 .Ci b7764c4a6b71d37918a97594d866258f154017ca .
1811 I would be truly happy to see such a terminal in action today,
1812 maybe even being able to work on it.
1813 But I fear my chances are null.
1814 .P
1815 The check only prevented a pager to be placed between the printing
1816 program (\c
1817 .Pn mhl )
1818 and the terminal.
1819 In nmh, this could have been ensured statically with the
1820 .Sw -nomoreproc
1821 at the command line, too.
1822 In mmh, setting the profile entry
1823 .Pe Pager
1824 or the environment variable
1825 .Ev PAGER
1826 to
1827 .Pn cat
1828 does the job.
1833 .H2 "Attachments
1834 .P
1835 The mind model of email attachments is unrelated to MIME.
1836 Although the MIME RFCs (2045 through 2049) define the technical
1837 requirements for having attachments, they do not mention the word
1838 ``attachment''.
1839 Instead of attachments, MIME talks about ``multi-part message bodies''
1840 [RFC\|2045], a more general concept.
1841 Multi-part messages are messages
1842 ``in which one or more different
1843 sets of data are combined in a single body''
1844 [RFC\|2046].
1845 MIME keeps its descriptions generic;
1846 it does not imply specific usage models.
1847 One usage model became prevalent: attachments.
1848 The idea is having a main text document with files of arbitrary kind
1849 attached to it.
1850 In MIME terms, this is a multi-part message having a text part first
1851 and parts of arbitrary type following.
1852 .P
1853 MH's MIME support is a direct implementation of the RFCs.
1854 The perception of the topic described in the RFCs is clearly visible
1855 in MH's implementation.
1856 In result, MH had all the MIME features but no idea of attachments.
1857 But users don't need all the MIME features,
1858 they want convenient attachment handling.
1861 .U3 "Composing MIME Messages
1862 .P
1863 In order to improve the situation on the message composing side,
1864 Jon Steinhart had added an attachment system to nmh in 2002.
1865 .Ci 7480dbc14bc90f2d872d434205c0784704213252
1866 In the file
1867 .Fn docs/README-ATTACHMENTS ,
1868 he described his motivation to do so as such:
1869 .QS
1870 Although nmh contains the necessary functionality for MIME message handing,
1871 the interface to this functionality is pretty obtuse.
1872 There's no way that I'm ever going to convince my partner to write
1873 .Pn mhbuild
1874 composition files!
1875 .QE
1876 .LP
1877 With this change, the mind model of attachments entered nmh.
1878 In the same document:
1879 .QS
1880 These changes simplify the task of managing attachments on draft files.
1881 They allow attachments to be added, listed, and deleted.
1882 MIME messages are automatically created when drafts with attachments
1883 are sent.
1884 .QE
1885 .LP
1886 Unfortunately, the attachment system,
1887 like any new facilities in nmh,
1888 was inactive by default.
1889 .P
1890 During my work in Argentina, I tried to improve the attachment system.
1891 But, because of great opposition in the nmh community,
1892 my patch died as a proposal on the mailing list, after long discussions.
1893 .[
1894 nmh-workers attachment proposal
1895 .]
1896 In January 2012, I extended the patch and applied it to mmh.
1897 .Ci 8ff284ff9167eff8f5349481529332d59ed913b1
1898 In mmh, the attachment system is active by default.
1899 Instead of command line switches, the
1900 .Pe Attachment-Header
1901 profile entry is used to specify
1902 the name of the attachment header field.
1903 It is pre-defined to
1904 .Hd Attach .
1905 .P
1906 To add an attachment to a draft, simply add an attachment header:
1907 .VS
1908 To: bob
1909 Subject: The file you wanted
1910 Attach: /path/to/the/file-bob-wanted
1911 --------
1912 Here it is.
1913 VE
1914 The header field can be added to the draft manually in the editor,
1915 or by using the `attach' command at the WhatNow prompt, or
1916 non-interactively with
1917 .Pn anno :
1918 .VS
1919 anno -append -nodate -component Attach -text /path/to/attachment
1920 VE
1921 Drafts with attachment headers are converted to MIME automatically by
1922 .Pn send .
1923 The conversion to MIME is invisible to the user.
1924 The draft stored in the draft folder is always in source form, with
1925 attachment headers.
1926 If the MIMEification fails, for instance because the file to attach
1927 is not accessible, the original draft is not changed.
1928 .P
1929 The attachment system handles the forwarding of messages, too.
1930 If the attachment header value starts with a plus character (`+'),
1931 like in
1932 .Cl "Attach: +bob 30 42" ,
1933 The given messages in the specified folder will be attached.
1934 This allowed to simplify
1935 .Pn forw .
1936 .Ci f41f04cf4ceca7355232cf7413e59afafccc9550
1937 .P
1938 Closely related to attachments is non-ASCII text content,
1939 because it requires MIME too.
1940 In nmh, the user needed to call `mime' at the WhatNow prompt
1941 to have the draft converted to MIME.
1942 This was necessary whenever the draft contained non-ASCII characters.
1943 If the user did not call `mime', a broken message would be sent.
1944 Therefore, the
1945 .Pe automimeproc
1946 profile entry could be specified to have the `mime' command invoked
1947 automatically each time.
1948 Unfortunately, this approach conflicted with with attachment system
1949 because the draft would already be in MIME format at the time
1950 when the attachment system wanted to MIMEify it.
1951 To use nmh's attachment system, `mime' must not be called at the
1952 WhatNow prompt and
1953 .Pe automimeproc
1954 must not be set in the profile.
1955 But then the case of non-ASCII text without attachment headers was
1956 not caught.
1957 All in all, the solution was complex and irritating.
1958 My patch from December 2010 would have simplified the situation.
1959 .P
1960 Mmh's current solution is even more elaborate.
1961 Any necessary MIMEification is done automatically.
1962 There is no `mime' command at the WhatNow prompt anymore.
1963 The draft will be converted automatically to MIME when either an
1964 attachment header or non-ASCII text is present.
1965 Further more, the special meaning of the hash character (`#')
1966 at line beginnings in the draft message is removed.
1967 Users need not at all deal with the whole topic.
1968 .P
1969 Although the new approach does not anymore support arbitrary MIME
1970 compositions directly, the full power of
1971 .Pn mhbuild
1972 can still be accessed.
1973 Given no attachment headers are included, the user can create
1974 .Pn mhbuild
1975 composition drafts like in nmh.
1976 Then, at the WhatNow prompt, he needs to invoke
1977 .Cl "edit mhbuild
1978 to convert it to MIME.
1979 Because the resulting draft does neither contain non-ASCII characters
1980 nor has it attachment headers, the attachment system will not touch it.
1981 .P
1982 The approach taken in mmh is tailored towards todays most common case:
1983 a text part with possibly attachments.
1984 This case is simplified a lot for users.
1987 .U3 "MIME Type Guessing
1988 .P
1989 The use of
1990 .Pn mhbuild
1991 composition drafts had one notable advantage over attachment headers
1992 from the programmer's point of view: The user provides the appropriate
1993 MIME types for files to include.
1994 The attachment system needs to find out the correct MIME type itself.
1995 This is a difficult task, yet it spares the user irritating work.
1996 Determining the correct MIME type of content is partly mechanical,
1997 partly intelligent work.
1998 Forcing the user to find out the correct MIME type,
1999 forces him to do partly mechanical work.
2000 Letting the computer do the work, can lead to bad choices for difficult
2001 content.
2002 For mmh, the latter option was chosen.
2003 .P
2004 Determining the MIME type by the suffix of the file name is a dumb
2005 approach, yet it is simple to implement and provides good results
2006 for the common cases.
2007 Mmh implements this approach in the
2008 .Pn print-mimetype
2009 script.
2010 .Ci 4b5944268ea0da7bb30598a27857304758ea9b44
2011 Using it is the default choice.
2012 .P
2013 A far better, though less portable, approach is the use of
2014 .Pn file .
2015 This standard tool tries to determine the type of files.
2016 Unfortunately, its capabilities and accuracy varies from system to system.
2017 Additionally, its output was only intended for human beings,
2018 but not to be used by programs.
2019 It varies much.
2020 Nevertheless, modern versions of GNU
2021 .Pn file ,
2022 which is prevalent on the popular GNU/Linux systems,
2023 provides MIME type output in machine-readable form.
2024 Although this solution is highly system-dependent,
2025 it solves the difficult problem well.
2026 On systems where GNU
2027 .Pn file ,
2028 version 5.04 or higher, is available it should be used.
2029 One needs to specify the following profile entry to do so:
2030 .Ci 3baec236a39c5c89a9bda8dbd988d643a21decc6
2031 .VS
2032 Mime-Type-Query: file -b --mime
2033 VE
2034 .LP
2035 Other versions of
2036 .Pn file
2037 might possibly be usable with wrapper scripts to reformat the output.
2038 The diversity among
2039 .Pn file
2040 implementations is great; one needs to check the local variant.
2041 .P
2042 If no MIME type can be determined, text content gets sent as
2043 `text/plain' and anything else under the generic fall-back type
2044 `application/octet-stream'.
2045 It is not possible in mmh to override the automatic MIME type guessing
2046 for a specific file.
2047 To do so, the user would need to know in advance for which file
2048 the automatic guessing does fail, or the system would require interaction.
2049 I consider both cases impractical.
2050 The existing solution should be sufficient.
2051 If not, the user may always fall back to
2052 .Pn mhbuild
2053 composition drafts and ignore the attachment system.
2056 .U3 "Storing Attachments
2057 .P
2058 Extracting MIME parts of a message and storing them to disk is done by
2059 .Pn mhstore .
2060 The program has two operation modes,
2061 .Sw -auto
2062 and
2063 .Sw -noauto .
2064 With the former one, each part is stored under the filename given in the
2065 MIME part's meta information, if available.
2066 This naming information is usually available for modern attachments.
2067 If no filename is available, this MIME part is stored as if
2068 .Sw -noauto
2069 would have been specified.
2070 In the
2071 .Sw -noauto
2072 mode, the parts are processed according to rules, defined by
2073 .Pe mhstore-store-*
2074 profile entries.
2075 These rules define generic filename templates for storing
2076 or commands to post-process the contents in arbitrary ways.
2077 If no matching rule is available the part is stored under a generic
2078 filename, built from message number, MIME part number, and MIME type.
2079 .P
2080 The
2081 .Sw -noauto
2082 mode had been the default in nmh because it was considered safe,
2083 in contrast to the
2084 .Sw -auto
2085 mode.
2086 In mmh,
2087 .Sw -auto
2088 is not dangerous anymore.
2089 Two changes were necessary:
2090 .BU
2091 Any directory path is removed from the proposed filename.
2092 Thus, the files are always stored in the expected directory.
2093 .Ci 41b6eadbcecf63c9a66aa5e582011987494abefb
2094 .BU
2095 Tar files are not extracted automatically any more.
2096 Thus, the rest of the file system will not be touched.
2097 .Ci 94c80042eae3383c812d9552089953f9846b1bb6
2098 .LP
2099 Now, the outcome of mmh's
2100 .Cl "mhstore -auto
2101 can be foreseen from the output of
2102 .Cl "mhlist -verbose" .
2103 .P
2104 The
2105 .Sw -noauto
2106 mode is seen to be more powerful but less convenient.
2107 On the other hand,
2108 .Sw -auto
2109 is safe now and
2110 storing attachments under their original name is intuitive.
2111 Hence,
2112 .Sw -auto
2113 serves better as the default option.
2114 .Ci 3410b680416c49a7617491af38bc1929855a331d
2115 .P
2116 Files are stored into the directory given by the
2117 .Pe Nmh-Storage
2118 profile entry, if set, or
2119 into the current working directory, otherwise.
2120 Storing to different directories is only possible with
2121 .Pe mhstore-store-*
2122 profile entries.
2123 .P
2124 Still, in both modes, existing files get overwritten silently.
2125 This can be considered a bug.
2126 Yet, each other behavior has its draw-backs, too.
2127 Refusing to replace files requires adding a
2128 .Sw -force
2129 option.
2130 Users will likely need to invoke
2131 .Pn mhstore
2132 a second time with
2133 .Sw -force
2134 then.
2135 Eventually, only the user can decide in the concrete case.
2136 This requires interaction, which I like to avoid if possible.
2137 Appending a unique suffix to the filename is another bad option.
2138 For now, the behavior remains as it is.
2139 .P
2140 In mmh, only MIME parts of type message are special in
2141 .Pn mhstore 's
2142 .Sw -auto
2143 mode.
2144 Instead of storing message/rfc822 parts as files to disk,
2145 they are stored as messages into the current mail folder.
2146 The same applies to message/partial, only, the parts are reassembled
2147 automatically before.
2148 Parts of type message/external-body are not automatically retrieved
2149 anymore. Instead, Information on how to retrieve them is output.
2150 Not supporting this rare case saved nearly one thousand lines of code.
2151 .Ci 55e1d8c654ee0f7c45b9361ce34617983b454c32
2152 .\" XXX mention somewhere else too: (The profile entry `nmh-access-ftp'
2153 .\" and sbr/ruserpass.c for reading ~/.netrc are gone now.)
2154 Not special anymore is `application/octet-stream; type=tar'.
2155 Automatically extracting such MIME parts had been the dangerous part
2156 of the
2157 .Sw -auto
2158 mode.
2159 .Ci 94c80042eae3383c812d9552089953f9846b1bb6
2163 .U3 "Showing MIME Messages
2164 .P
2165 The program
2166 .Pn mhshow
2167 had been written to display MIME messages.
2168 It implemented the conceptional view of the MIME RFCs.
2169 Nmh's
2170 .Pn mhshow
2171 handled each MIME part independently, presenting them separately
2172 to the user.
2173 This does not match today's understanding of email attachments,
2174 where displaying a message is seen to be a single, integrated operation.
2175 Today, email messages are expected to consist of a main text part
2176 plus possibly attachments.
2177 They are not any more seen to be arbitrary MIME hierarchies with
2178 information on how to display the individual parts.
2179 I adjusted
2180 .Pn mhshow 's
2181 behavior to the modern view on the topic.
2182 .P
2183 Note that this section completely ignores the original
2184 .Pn show
2185 program, because it was not capable to display MIME messages
2186 and is no longer part of mmh.
2187 Although
2188 .Pn mhshow
2189 was renamed to
2190 .Pn show
2191 in mmh, this section uses the name
2192 .Pn mhshow ,
2193 in order to avoid confusion.
2194 .P
2195 In mmh, the basic idea is that
2196 .Pn mhshow
2197 should display a message in one single pager session.
2198 Therefore,
2199 .Pn mhshow
2200 invokes a pager session for all its output,
2201 whenever it prints to a terminal.
2202 .Ci a4197ea6ffc5c1550e8b52d5a654bcaaaee04a4e
2203 In consequence,
2204 .Pn mhl
2205 does no more invoke a pager.
2206 .Ci 0e46503be3c855bddaeae3843e1b659279c35d70
2207 With
2208 .Pn mhshow
2209 replacing the original
2210 .Pn show ,
2211 output from
2212 .Pn mhl
2213 does not go to the terminal directly, but through
2214 .Pn mhshow .
2215 Hence,
2216 .Pn mhl
2217 does not need to invoke a pager.
2218 The one and only job of
2219 .Pn mhl
2220 is to format messages or parts of them.
2221 The only place in mmh, where a pager is invoked is
2222 .Pn mhshow .
2223 .P
2224 .Pe mhshow-show-*
2225 profile entries can be used to display MIME parts in a specific way.
2226 For instance, PDF and Postscript files could be converted to plain text
2227 to display them in the terminal.
2228 In mmh, the displaying of MIME parts will always be done serially.
2229 The request to display the MIME type `multipart/parallel' in parallel
2230 is ignored.
2231 It is simply treated as `multipart/mixed'.
2232 .Ci d0581ba306a7299113a346f9b4c46ce97bc4cef6
2233 This could already be requested with the, now removed,
2234 .Sw -serialonly
2235 switch of
2236 .Pn mhshow .
2237 As MIME parts are always processed exclusively , i.e. serially,
2238 the `%e' escape in
2239 .Pe mhshow-show-*
2240 profile entries became useless and was thus removed.
2241 .Ci a20d405db09b7ccca74d3e8c57550883da49e1ae
2242 .P
2243 In the intended setup, only text content would be displayed.
2244 Non-text content would be converted to text by appropriate
2245 .Pe mhshow-show-*
2246 profile entries before, if possible and wanted.
2247 All output would be displayed in a single pager session.
2248 Other kinds of attachments are ignored.
2249 With
2250 .Pe mhshow-show-*
2251 profile entries for them, they can be displayed serially along
2252 the message.
2253 For parallel display, the attachments need to be stored to disk first.
2254 .P
2255 To display text content in foreign charsets, they need to be converted
2256 to the native charset.
2257 Therefore,
2258 .Pe mhshow-charset-*
2259 profile entries used to be needed.
2260 In mmh, the conversion is done automatically by piping the text through
2261 the
2262 .Pn iconv
2263 command, if necessary.
2264 .Ci 2433122c20baccb10b70b49c04c6b0497b5b3b60
2265 Custom
2266 .Pe mhshow-show-*
2267 rules for textual content might need a
2268 .Cl "iconv -f %c %f |
2269 prefix to have the text converted to the native charset.
2270 .P
2271 Although the conversion of foreign charsets to the native one
2272 has improved, it is not consistent enough.
2273 Further work needs to be done and
2274 the basic concepts in this field need to be re-thought.
2275 Though, the default setup of mmh displays message in foreign charsets
2276 correctly without the need to configure anything.
2279 .ig
2281 .P
2282 mhshow/mhstore: Removed support for retrieving message/external-body parts.
2283 These tools won't download the contents automatically anymore. Instead,
2284 they print the information needed to get the contents. If someone should
2285 really receive one of those rare message/external-body messages, he can
2286 do the job manually. We save nearly a thousand lines of code. That's worth
2287 it!
2288 (The profile entry `nmh-access-ftp' and sbr/ruserpass.c for reading
2289 ~/.netrc are gone now.)
2290 .Ci 55e1d8c654ee0f7c45b9361ce34617983b454c32
2292 ..
2296 .H2 "Digital Cryptography
2297 .P
2298 Signing and encryption.
2299 .P
2300 FIXME
2304 .H2 "Draft and Trash Folder
2305 .P
2307 .U3 "Draft Folder
2308 .P
2309 In the beginning, MH had the concept of a draft message.
2310 This is the file
2311 .Fn draft
2312 in the MH directory, which is treated special.
2313 On composing a message, this draft file was used.
2314 As the draft file was one particular file, only one draft could be
2315 managed at any time.
2316 When starting to compose another message before the former one was sent,
2317 the user had to decide among:
2318 .BU
2319 Use the old draft to finish and send it before starting with a new one.
2320 .BU
2321 Discard the old draft, replacing it with the new one.
2322 .BU
2323 Preserve the old draft by refiling it to a folder.
2324 .P
2325 This was, it was only possible to work in alternation on multiple drafts.
2326 Therefore, the current draft needed to be refiled to a folder and
2327 another one re-using for editing.
2328 Working on multiple drafts at the same time was impossible.
2329 The usual approach of switching to a different MH context did not
2330 change anything.
2331 .P
2332 The draft folder facility exists to
2333 allow true parallel editing of drafts, in a straight forward way.
2334 It was introduced by Marshall T. Rose, already in 1984.
2335 Similar to other new features, the draft folder was inactive by default.
2336 Even in nmh, the highly useful draft folder was not available
2337 out-of-the-box.
2338 At least, Richard Coleman added the man page
2339 .Mp mh-draft (5)
2340 to better document the feature.
2341 .P
2342 Not using the draft folder facility has the single advantage of having
2343 the draft file at a static location.
2344 This is simple in simple cases but the concept does not scale for more
2345 complex cases.
2346 The concept of the draft message is too limited for the problem.
2347 Therefore the draft folder was introduced.
2348 It is the more powerful and more natural concept.
2349 The draft folder is a folder like any other folder in MH.
2350 Its messages can be listed like any other messages.
2351 A draft message is no longer a special case.
2352 Tools do not need special switches to work on the draft message.
2353 Hence corner-cases were removed.
2354 .P
2355 The trivial part of the work was activating the draft folder with a
2356 default name.
2357 I chose the name
2358 .Fn +drafts
2359 for obvious reasons.
2360 In consequence, the command line switches
2361 .Sw -draftfolder
2362 and
2363 .Sw -draftmessage
2364 could be removed.
2365 More difficult but also more improving was updating the tools to the
2366 new concept.
2367 For nearly three decades, the tools needed to support two draft handling
2368 approaches.
2369 By fully switching to the draft folder, the tools could be simplified
2370 by dropping the awkward draft message handling code.
2371 .Sw -draft
2372 switches were removed because operating on a draft message is no longer
2373 special.
2374 It became indistinguishable to operating on any other message.
2375 There is no more need to query the user for draft handling.
2376 It is always possible to add another new draft.
2377 Refiling drafts is without difference to refiling other messages.
2378 All these special cases are gone.
2379 Yet, one draft-related switch remained.
2380 .Pn comp
2381 still has
2382 .Sw -[no]use
2383 for switching between two modes:
2384 .BU
2385 .Sw -use :
2386 Modify an existing draft.
2387 .BU
2388 .Sw -nouse :
2389 Compose a new draft, possibly taking some existing message as a form.
2390 .P
2391 In either case, the behavior of
2392 .Pn comp
2393 is deterministic.
2394 .P
2395 .Pn send
2396 now operates on the current message in the draft folder by default.
2397 As message and folder can both be overridden by specifying them on
2398 the command line, it is possible to send any message in the mail storage
2399 by simply specifying its number and folder.
2400 In contrast to the other tools,
2401 .Pn send
2402 takes the draft folder as its default folder.
2403 .P
2404 Dropping the draft message concept in favor for the draft folder concept,
2405 removed special cases with regular cases.
2406 This simplified the source code of the tools, as well as the concepts.
2407 In mmh, draft management does not break with the MH concepts
2408 but applies them.
2409 .Cl "scan +drafts" ,
2410 for instance, is a truly natural request.
2411 Most of the work was already done by Rose in the eighties.
2412 The original improvement of mmh is dropping the old draft message approach
2413 and thus simplifying the tools, the documentation and the system as a whole.
2414 Although my part in the draft handling improvement was small,
2415 it was an important one.
2418 .U3 "Trash Folder
2419 .P
2420 Similar to the situation for drafts is the situation for removed messages.
2421 Historically, a message was ``deleted'' by prepending a specific
2422 \fIbackup prefix\fP, usually the comma character,
2423 to the file name.
2424 The specific message would vanish from MH because only files with
2425 non-digit characters in their name are not treated as messages.
2426 Although files remained in the file system,
2427 the messages were no more visible in MH.
2428 To truly delete them, a maintenance job is needed.
2429 Usually a cron job is installed to delete them after a grace time.
2430 For instance:
2431 .VS
2432 find $HOME/Mail -type f -name ',*' -ctime +7 -delete
2433 VE
2434 In such a setup, the original message can be restored
2435 within the grace time interval by stripping the
2436 the backup prefix from the file name.
2437 But one can not rely on this statement.
2438 If the last message of a folder with six messages (1-6) is removed,
2439 message
2440 .Fn 6 ,
2441 becomes file
2442 .Fn ,6 .
2443 If then a new message enters the same folder, it will be given
2444 the number one higher than the highest existing message.
2445 In this case the message is named
2446 .Fn 6
2447 then.
2448 If this message is removed as well,
2449 then the backup of the former message gets overwritten.
2450 Hence, the ability to restore removed messages does not only depend on
2451 the ``sweeping cron job'' but also on the removing of further messages.
2452 It is undesirable to have such obscure and complex mechanisms.
2453 The user should be given a small set of clear assertions.
2454 ``Removed files are restorable within a seven-day grace time.''
2455 is such a clear assertion.
2456 With the addition ``... unless a message with the same name in the
2457 same folder is removed before.'' the statement becomes complex.
2458 A user will hardly be able to keep track of any removal to know
2459 if the assertion still holds true for a specific file.
2460 The the real mechanism is practically obscure to the user.
2461 The consequences of further removals are not obvious.
2462 .P
2463 Further more, the backup files are scattered within the whole mail storage.
2464 This complicates managing them.
2465 It is possible, with help of
2466 .Pn find ,
2467 but everything would be more convenient
2468 if the deleted messages would be collected in one place.
2469 .P
2470 The profile entry
2471 .Pe rmmproc
2472 (previously named
2473 .Pe Delete-Prog )
2474 was introduced very early to improve the situation.
2475 It could be set to any command, which would be executed to removed
2476 the specified messages.
2477 This would override the default action, described above.
2478 Refiling the to-be-removed files to a garbage folder is the usual example.
2479 Nmh's man page
2480 .Mp rmm (1)
2481 proposes to set the
2482 .Pe rmmproc
2483 to
2484 .Cl "refile +d
2485 to move messages to the garbage folder,
2486 .Fn +d ,
2487 instead of renaming them with the backup prefix.
2488 The man page proposes additionally the expunge command
2489 .Cl "rm `mhpath +d all`
2490 to empty the garbage folder.
2491 .P
2492 Removing messages in such a way has advantages.
2493 The mail storage is prevented from being cluttered with removed messages
2494 because they are all collected in one place.
2495 Existing and removed messages are thus separated more strictly.
2496 No backup files are silently overwritten.
2497 Most important is the ability to keep removed messages in the MH domain.
2498 Messages in the trash folder can be listed like those in any other folder.
2499 Deleted messages can be displayed like any other messages.
2500 Restoring a deleted messages can be done with
2501 .Pn refile .
2502 All operations on deleted files are still covered by the MH tools.
2503 The trash folder is just like any other folder in the mail storage.
2504 .P
2505 Similar to the draft folder case, I dropped the old backup prefix approach
2506 in favor for replacing it by the better suiting trash folder system.
2507 Hence,
2508 .Pn rmm
2509 calls
2510 .Pn refile
2511 to move the to-be-removed message to the trash folder,
2512 .Fn +trash
2513 by default.
2514 To sweep it clean, one can use
2515 .Cl "rmm -unlink +trash a" ,
2516 where the
2517 .Sw -unlink
2518 switch causes the files to be unlinked.
2519 .P
2520 Dropping the legacy approach and completely converting to the new approach
2521 simplified the code base.
2522 The relationship between
2523 .Pn rmm
2524 and
2525 .Pn refile
2526 was inverted.
2527 In mmh,
2528 .Pn rmm
2529 invokes
2530 .Pn refile ,
2531 which used to be the other way round.
2532 Yet, the relationship is simpler now.
2533 No more can loops, like described in nmh's man page for
2534 .Mp refile (1),
2535 occur:
2536 .QS
2537 Since
2538 .Pn refile
2539 uses your
2540 .Pe rmmproc
2541 to delete the message, the
2542 .Pe rmmproc
2543 must NOT call
2544 .Pn refile
2545 without specifying
2546 .Sw -normmproc
2547 or you will create an infinite loop.
2548 .QE
2549 .LP
2550 .Pn rmm
2551 either unlinks a message with
2552 .Fu unlink()
2553 or invokes
2554 .Pn refile
2555 to move it to the trash folder.
2556 .Pn refile
2557 does not invoke any tools.
2558 .P
2559 By generalizing the message removal in the way that it became covered
2560 by the MH concepts made the whole system more powerful.
2566 .H2 "Modern Defaults
2567 .P
2568 Nmh has a bunch of convenience-improving features inactive by default,
2569 although one can expect every new user wanting to have them active.
2570 The reason they are inactive by default is the wish to stay compatible
2571 with old versions.
2572 But what is the definition for old versions?
2573 Still, the highly useful draft folder facility has not been activated
2574 by default although it was introduced over twenty-five years ago.
2575 .[
2576 rose romine real work
2577 .]
2578 The community seems not to care.
2579 This is one of several examples that require new users to first build up
2580 a profile before they can access the modern features of nmh.
2581 Without an extensive profile, the setup is hardly usable
2582 for modern emailing.
2583 The point is not the customization of the setup,
2584 but the need to activate generally useful facilities.
2585 .P
2586 Yet, the real problem lies less in enabling the features, as this is
2587 straight forward as soon as one knows what he wants.
2588 The real problem is that new users need deep insights into the project
2589 before they find out what they are missing and that nmh actually
2590 provides it already, it just was not activated.
2591 To give an example, I needed one year of using nmh
2592 before I became aware of the existence of the attachment system.
2593 One could argue that this fact disqualifies my reading of the
2594 documentation.
2595 If I would have installed nmh from source back then, I could agree.
2596 Yet, I had used a prepackaged version and had expected that it would
2597 just work.
2598 Nevertheless, I had been convinced by the concepts of MH already
2599 and I am a software developer,
2600 still I required a lot of time to discover the cool features.
2601 How can we expect users to be even more advanced than me,
2602 just to allow them use MH in a convenient and modern way?
2603 Unless they are strongly convinced of the concepts, they will fail.
2604 I have seen friends of me giving up disappointed
2605 before they truly used the system,
2606 although they had been motivated in the beginning.
2607 They suffer hard enough to get used to the toolchest approach,
2608 we should spare them further inconveniences.
2609 .P
2610 Maintaining compatibility for its own sake is bad,
2611 because the code base collects more and more compatibility code.
2612 Sticking to the compatiblity code means remaining limited;
2613 not using it renders it unnecessary.
2614 Keeping unused alternative in the code is a bad choice as they likely
2615 gather bugs, by not being well tested.
2616 Also, the increased code size and the greater number of conditions
2617 increase the maintenance costs.
2618 If any MH implementation would be the back-end of widespread
2619 email clients with large user bases, compatibility would be more
2620 important.
2621 Yet, it appears as if this is not the case.
2622 Hence, compatibility is hardly important for technical reasons.
2623 Its importance originates rather from personal reasons.
2624 Nmh's user base is small and old.
2625 Changing the interfaces would cause inconvenience to long-term users of MH.
2626 It would force them to change their many years old MH configurations.
2627 I do understand this aspect, but it keeps new users from using MH.
2628 By sticking to the old users, new users are kept away.
2629 Yet, the future lies in new users.
2630 Hence, mmh invites new users by providing a convenient and modern setup,
2631 readily usable out-of-the-box.
2632 .P
2633 In mmh, all modern features are active by default and many previous
2634 approaches are removed or only accessible in manual ways.
2635 New default features include:
2636 .BU
2637 The attachment system (\c
2638 .Hd Attach ).
2639 .Ci 8ff284ff9167eff8f5349481529332d59ed913b1
2640 .BU
2641 The draft folder facility (\c
2642 .Fn +drafts ).
2643 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860
2644 .BU
2645 The unseen sequence (`u')
2646 .Ci c2360569e1d8d3678e294eb7c1354cb8bf7501c1
2647 and the sequence negation prefix (`!').
2648 .Ci db74c2bd004b2dc9bf8086a6d8bf773ac051f3cc
2649 .BU
2650 Quoting the original message in the reply.
2651 .Ci 67411b1f95d6ec987b4c732459e1ba8a8ac192c6
2652 .BU
2653 Forwarding messages using MIME.
2654 .Ci 6e271608b7b9c23771523f88d23a4d3593010cf1
2655 .P
2656 In consequence, a setup with a profile that defines only the path to the
2657 mail storage, is already convenient to use.
2658 Again, Paul Vixie's ``edginess'' appeal supports the direction I took:
2659 ``the `main branch' should just be modern''.
2660 .[
2661 paul vixie edginess nmh-workers
2662 .]
2668 .\" --------------------------------------------------------------
2669 .H1 "Styling
2670 .P
2671 Kernighan and Pike have emphasized the importance of style in the
2672 preface of their book:
2673 .[ [
2674 kernighan pike practice of programming
2675 .], p. x]
2676 .QS
2677 Chapter 1 discusses programming style.
2678 Good style is so important to good programming that we have chose
2679 to cover it first.
2680 .QE
2681 This section covers changes in mmh that were motivated by the desire
2682 to improve on style.
2683 Many of them follow the rules given in the quoted book.
2684 .[
2685 kernighan pike practice of programming
2686 .]
2691 .H2 "Code Style
2692 .P
2693 .U3 "Indentation Style
2694 .P
2695 Indentation styles are the holy cow of programmers.
2696 Again Kernighan and Pike:
2697 .[ [
2698 kernighan pike practice of programming
2699 .], p. 10]
2700 .QS
2701 Programmers have always argued about the layout of programs,
2702 but the specific style is much less important than its consistent
2703 application.
2704 Pick one style, preferably ours, use it consistently, and don't waste
2705 time arguing.
2706 .QE
2707 .P
2708 I agree that the constant application is most important,
2709 but I believe that some styles have advantages over others.
2710 For instance the indentation with tab characters only.
2711 Tab characters directly map to the nesting level \(en
2712 one tab, one level.
2713 Tab characters are flexible because developers can adjust them to
2714 whatever width they like to have.
2715 There is no more need to run
2716 .Pn unexpand
2717 or
2718 .Pn entab
2719 programs to ensure the correct mixture of leading tabs and spaces.
2720 The simple rules are: (1) Leading whitespace must consist of tabs only.
2721 (2) Any other whitespace should consist of spaces.
2722 These two rules ensure the integrity of the visual appearance.
2723 Although reformatting existing code should be avoided, I did it.
2724 I did not waste time arguing; I just reformated the code.
2725 .Ci a485ed478abbd599d8c9aab48934e7a26733ecb1
2727 .U3 "Comments
2728 .P
2729 Section 1.6 of
2730 .[ [
2731 kernighan pike practice of programming
2732 .], p. 23]
2733 demands: ``Don't belabor the obvious.''
2734 Hence, I simply removed all the comments in the following code excerpt:
2735 .VS
2736 context_replace(curfolder, folder); /* update current folder */
2737 seq_setcur(mp, mp->lowsel); /* update current message */
2738 seq_save(mp); /* synchronize message sequences */
2739 folder_free(mp); /* free folder/message structure */
2740 context_save(); /* save the context file */
2742 [...]
2744 int c; /* current character */
2745 char *cp; /* miscellaneous character pointer */
2747 [...]
2749 /* NUL-terminate the field */
2750 *cp = '\0';
2751 VE
2752 .Ci 426543622b377fc5d091455cba685e114b6df674
2753 .P
2754 The program code explains enough itself, already.
2757 .U3 "Names
2758 .P
2759 Kernighan and Pike suggest:
2760 ``Use active names for functions''.
2761 .[ [
2762 kernighan pike practice of programming
2763 .], p. 4]
2764 One application of this rule was the rename of
2765 .Fu check_charset()
2766 to
2767 .Fu is_native_charset() .
2768 .Ci 8d77b48284c58c135a6b2787e721597346ab056d
2769 The same change fixed a violation of ``Be accurate'' as well.
2770 The code did not match the expectation the function suggested,
2771 as it, for whatever reason, only compared the first ten characters
2772 of the charset name.
2773 .P
2774 More important than using active names is using descriptive names.
2775 .VS
2776 m_unknown(in); /* the MAGIC invocation... */
2777 VE
2778 Renaming the obscure
2779 .Fu m_unknown()
2780 function was a delightful event, although it made the code less funny.
2781 .Ci 611d68d19204d7cbf5bd585391249cb5bafca846
2782 .P
2783 Magic numbers are generally considered bad style.
2784 Obviously, Kernighan and Pike agree:
2785 ``Give names to magic numbers''.
2786 .[ [
2787 kernighan pike practice of programming
2788 .], p. 19]
2789 One such change was naming the type of input \(en mbox or mail folder \(en
2790 to be scanned:
2791 .VS
2792 #define SCN_MBOX (-1)
2793 #define SCN_FOLD 0
2794 VE
2795 .Ci 7ffb36d28e517a6f3a10272056fc127592ab1c19
2796 .P
2797 The argument
2798 .Ar outnum
2799 of the function
2800 .Fu scan()
2801 in
2802 .Fn uip/scansbr.c
2803 defines the number of the message to be created.
2804 If no message is to be created, the argument is misused to transport
2805 program logic.
2806 This lead to obscure code.
2807 I improved the clarity of the code by introducing two variables:
2808 .VS
2809 int incing = (outnum > 0);
2810 int ismbox = (outnum != 0);
2811 VE
2812 They cover the magic values and are used for conditions.
2813 The variable
2814 .Ar outnum
2815 is only used when it holds an ordinary message number.
2816 .Ci b8b075c77be7794f3ae9ff0e8cedb12b48fd139f
2817 The clarity improvement of the change showed detours in the program logic
2818 of related code parts.
2819 Having the new variables with descriptive names, a more
2820 straight forward implementation became apparent.
2821 Before the clarification was done,
2822 the possibility to improve had not be seen.
2823 .Ci aa60b0ab5e804f8befa890c0a6df0e3143ce0723
2827 .H2 "Structural Rework
2828 .P
2829 Although the stylistic changes described up to here improve the
2830 readability of the source code, all of them are changes ``in the small''.
2831 Structural changes affect a much larger area.
2832 They are more difficult to do but lead to larger improvements,
2833 especially as they influence the outer shape of the tools as well.
2834 .P
2835 At the end of their chapter on style,
2836 Kernighan and Pike ask: ``But why worry about style?''
2837 Following are two examples of structural rework that show
2838 why style is important in the first place.
2841 .U3 "Rework of \f(CWanno\fP
2842 .P
2843 Until 2002,
2844 .Pn anno
2845 had six functional command line switches,
2846 .Sw -component
2847 and
2848 .Sw -text ,
2849 which took an argument each,
2850 and the two pairs of flags,
2851 .Sw -[no]date
2852 and
2853 .Sw -[no]inplace.,
2854 .Sw -component
2855 and
2856 .Sw -text ,
2857 which took an argument each,
2858 and the two pairs of flags,
2859 .Sw -[no]date
2860 and
2861 .Sw -[no]inplace .
2862 Then Jon Steinhart introduced his attachment system.
2863 In need for more advanced annotation handling, he extended
2864 .Pn anno .
2865 He added five more switches:
2866 .Sw -draft ,
2867 .Sw -list ,
2868 .Sw -delete ,
2869 .Sw -append ,
2870 and
2871 .Sw -number ,
2872 the last one taking an argument.
2873 .Ci 7480dbc14bc90f2d872d434205c0784704213252
2874 Later,
2875 .Sw -[no]preserve
2876 was added.
2877 .Ci d9b1d57351d104d7ec1a5621f090657dcce8cb7f
2878 Then, the Synopsis section of the man page
2879 .Mp anno (1)
2880 read:
2881 .VS
2882 anno [+folder] [msgs] [-component field] [-inplace | -noinplace]
2883 [-date | -nodate] [-draft] [-append] [-list] [-delete]
2884 [-number [num|all]] [-preserve | -nopreserve] [-version]
2885 [-help] [-text body]
2886 VE
2887 .LP
2888 The implementation followed the same structure.
2889 Problems became visible when
2890 .Cl "anno -list -number 42
2891 worked on the current message instead on message number 42,
2892 and
2893 .Cl "anno -list -number l:5
2894 did not work on the last five messages but failed with the mysterious
2895 error message: ``anno: missing argument to -list''.
2896 Yet, the invocation matched the specification in the man page.
2897 There, the correct use of
2898 .Sw -number
2899 was defined as being
2900 .Cl "[-number [num|all]]
2901 and the textual description for the combination with
2902 .Sw -list
2903 read:
2904 .QS
2905 The -list option produces a listing of the field bodies for
2906 header fields with names matching the specified component,
2907 one per line. The listing is numbered, starting at 1, if
2908 the -number option is also used.
2909 .QE
2910 .LP
2911 The problem was manifold.
2912 The code required a numeric argument to the
2913 .Sw -number
2914 switch.
2915 If it was missing or non-numeric,
2916 .Pn anno
2917 aborted with an error message that had an off-by-one error,
2918 printing the switch one before the failing one.
2919 Semantically, the argument to the
2920 .Sw -number
2921 switch is only necessary in combination with
2922 .Sw -delete ,
2923 but not with
2924 .Sw -list .
2925 In the former case it is even necessary.
2926 .P
2927 Trying to fix these problems on the surface would not have solved it truly.
2928 The problems discovered originate from a discrepance between the semantic
2929 structure of the problem and the structure implemented in the program.
2930 Such structural differences can not be cured on the surface.
2931 They need to be solved by adjusting the structure of the implementation
2932 to the structure of the problem.
2933 .P
2934 In 2002, the new switches
2935 .Sw -list
2936 and
2937 .Sw -delete
2938 were added in the same way, the
2939 .Sw -number
2940 switch for instance had been added.
2941 Yet, they are of structural different type.
2942 Semantically,
2943 .Sw -list
2944 and
2945 .Sw -delete
2946 introduce modes of operation.
2947 Historically,
2948 .Pn anno
2949 had only one operation mode: adding header fields.
2950 With the extension, it got two moder modes:
2951 listing and deleting header fields.
2952 The structure of the code changes did not pay respect to this
2953 fundamental change to
2954 .Pn anno 's
2955 behavior.
2956 Neither the implementation nor the documentation did clearly
2957 define them as being exclusive modes of operation.
2958 Having identified the problem, I solved it by putting structure into
2959 .Pn anno
2960 and its documentation.
2961 .Ci d54c8db8bdf01e8381890f7729bc0ef4a055ea11
2962 .P
2963 The difference is visible in both, the code and the documentation.
2964 The following code excerpt:
2965 .VS
2966 int delete = -2; /* delete header element if set */
2967 int list = 0; /* list header elements if set */
2968 [...]
2969 case DELETESW: /* delete annotations */
2970 delete = 0;
2971 continue;
2972 case LISTSW: /* produce a listing */
2973 list = 1;
2974 continue;
2975 VE
2976 .LP
2977 was replaced by:
2978 .VS
2979 static enum { MODE_ADD, MODE_DEL, MODE_LIST } mode = MODE_ADD;
2980 [...]
2981 case DELETESW: /* delete annotations */
2982 mode = MODE_DEL;
2983 continue;
2984 case LISTSW: /* produce a listing */
2985 mode = MODE_LIST;
2986 continue;
2987 VE
2988 .LP
2989 The replacement code does not only reflect the problem's structure better,
2990 it is easier to understand as well.
2991 The same applies to the documentation.
2992 The man page was completely reorganized to propagate the same structure.
2993 This is visible in the Synopsis section:
2994 .VS
2995 anno [+folder] [msgs] [-component field] [-text body]
2996 [-append] [-date | -nodate] [-preserve | -nopreserve]
2997 [-Version] [-help]
2999 anno -delete [+folder] [msgs] [-component field] [-text
3000 body] [-number num | all ] [-preserve | -nopreserve]
3001 [-Version] [-help]
3003 anno -list [+folder] [msgs] [-component field] [-number]
3004 [-Version] [-help]
3005 VE
3006 .\" XXX think about explaining the -preserve rework?
3010 .U3 "Path Conversion
3011 .P
3012 Four kinds of path names can appear in MH:
3013 .IP (1)
3014 Absolute Unix directory paths, like
3015 .Fn /etc/passwd .
3016 .IP (2)
3017 Relative Unix directory paths, like
3018 .Fn ./foo/bar .
3019 .IP (3)
3020 Absolute MH folder paths, like
3021 .Fn +friends/phil .
3022 .IP (4)
3023 Relative MH folder paths, like
3024 .Fn @subfolder .
3025 .P
3026 The last type, relative MH folder paths, are hardly documented.
3027 Nonetheless, they are useful for large mail storages.
3028 The current mail folder is specified as `\c
3029 .Fn @ ',
3030 just like the current directory is specified as `\c
3031 .Fn . '.
3032 .P
3033 To allow MH tools to understand all four notations,
3034 they need to convert between them.
3035 In nmh, these path name conversion functions were located in the files
3036 .Fn sbr/path.c
3037 (``return a pathname'') and
3038 .Fn sbr/m_maildir.c
3039 (``get the path for the mail directory'').
3040 The seven functions in the two files were documented with no more
3041 than two comments, which described obvious information.
3042 The function signatures were neither explaining:
3043 .VS
3044 char *path(char *, int);
3045 char *pluspath(char *);
3046 char *m_mailpath(char *);
3047 char *m_maildir(char *);
3048 VE
3049 .P
3050 My investigation provides the following description:
3051 .BU
3052 The second parameter of
3053 .Fu path()
3054 defines the type of path given as first parameter.
3055 Directory paths are converted to absolute directory paths.
3056 Folder paths are converted to absolute folder paths.
3057 Folder paths must not include a leading `@' character.
3058 Leading plus characters are preserved.
3059 The result is a pointer to newly allocated memory.
3060 .BU
3061 .Fu pluspath()
3062 is a convenience-wrapper to
3063 .Fu path() ,
3064 to convert folder paths only.
3065 This function can not be used for directory paths.
3066 An empty string parameter causes a buffer overflow.
3067 .BU
3068 .Fu m_mailpath()
3069 converts directory paths to absolute directory paths.
3070 The characters `+' or `@' at the beginning of the path name are
3071 treated literal, i.e. as the first character of a relative directory path.
3072 Hence, this function can not be used for folder paths.
3073 In any case, the result is an absolute directory path.
3074 The result is a pointer to newly allocated memory.
3075 .BU
3076 .Fu m_maildir()
3077 returns the parameter unchanged if it is an absolute directory path
3078 or begins with the entry `.' or `..'.
3079 All other strings are prepended with the current working directory.
3080 Hence, this functions can not be used for folder paths.
3081 The result is either an absolute directory path or a relative
3082 directory path, starting with a dot.
3083 In contrast to the other functions, the result is a pointer to
3084 static memory.
3085 .P
3086 The situation was obscure, irritating, error-prone, and non-orthogonal.
3087 No clear terminology was used to name the different kinds of path names.
3088 The first argument of
3089 .Fu m_mailpath() ,
3090 for instance, was named
3091 .Ar folder ,
3092 though
3093 .Fu m_mailpath()
3094 can not be used for MH folders.
3095 .P
3096 I reworked the path name conversion completely, introducing clarity.
3097 First of all, the terminology needed to be defined.
3098 A path name is either in the Unix domain, then it is called
3099 \fIdirectory path\fP, `dirpath' for short, or it is in the MH domain,
3100 then it is called \fIfolder path\fP, `folpath' for short.
3101 The two terms need to be used with strict distinction.
3102 Having a clear terminology is often an indicator of having understood
3103 the problem itself.
3104 Second, I exploited the concept of path type indicators.
3105 By requesting every path name to start with a clear type identifier,
3106 conversion between the types can be fully automated.
3107 Thus the tools can accept paths of any type from the user.
3108 Therefore, it was necessary to require relative directory paths to be
3109 prefixed with a dot character.
3110 In consequence, the dot character could no longer be an alias for the
3111 current message.
3112 .Ci cff0e16925e7edbd25b8b9d6d4fbdf03e0e60c01
3113 Third, I created three new functions to replace the previous mess:
3114 .BU
3115 .Fu expandfol()
3116 converts folder paths to absolute folder paths,
3117 without the leading plus character.
3118 Directory paths are simply passed through.
3119 This function is to be used for folder paths only, thus the name.
3120 The result is a pointer to static memory.
3121 .BU
3122 .Fu expanddir()
3123 converts directory paths to absolute directory paths.
3124 Folder paths are treated as relative directory paths.
3125 This function is to be used for directory paths only, thus the name.
3126 The result is a pointer to static memory.
3127 .BU
3128 .Fu toabsdir()
3129 converts any type of path to an absolute directory path.
3130 This is the function of choice for path conversion.
3131 Absolute directory paths are the most general representation of a
3132 path name.
3133 The result is a pointer to static memory.
3134 .P
3135 The new functions have names that indicate their use.
3136 Two of the functions convert relative to absolute path names of the
3137 same type.
3138 The third function converts any path name type to the most general one,
3139 the absolute directory path.
3140 All of the functions return pointers to static memory.
3141 All three functions are implemented in
3142 .Fn sbr/path.c .
3143 .Fn sbr/m_maildir.c
3144 is removed.
3145 .P
3146 Along with the path conversion rework, I also replaced
3147 .Fu getfolder(FDEF)
3148 with
3149 .Fu getdeffol()
3150 and
3151 .Fu getfolder(FCUR)
3152 with
3153 .Fu getcurfol() ,
3154 which is only a convenience wrapper for
3155 .Fu expandfol("@") .
3156 This code was moved from
3157 .Fn sbr/getfolder.c
3158 to
3159 .Fn sbr/path.c .
3160 .P
3161 The related function
3162 .Fu etcpath()
3163 was moved to
3164 .Fn sbr/path.c ,
3165 too.
3166 Previously, it had been located in
3167 .Fn config/config.c ,
3168 for whatever reasons.
3169 .P
3170 .Fn sbr/path.c
3171 now contains all path handling code.
3172 Only 173 lines of code were needed to replace the previous 252 lines.
3173 The readability of the code is highly improved.
3174 Additionally, each of the six exported and one static functions
3175 is introduced by an explaining comment.
3176 .Ci d39e2c447b0d163a5a63f480b23d06edb7a73aa0
3181 .H2 "Profile Reading
3182 .P
3183 The MH profile contains the configuration for the user-specific MH setup.
3184 MH tools read the profile right after starting up,
3185 as it contains the location of the user's mail storage
3186 and similar settings that influence the whole setup.
3187 Further more, the profile contains the default switches for the tools,
3188 hence, it must be read before the command line switches are processed.
3189 .P
3190 For historic reasons, some MH tools did not read the profile and context.
3191 Among them were
3192 .Pn post /\c
3193 .Pn spost ,
3194 .Pn mhmail ,
3195 and
3196 .Pn slocal .
3197 The reason why these tools ignored the profile were not clearly stated.
3198 During the discussion on the nmh-workers mailing list,
3199 .[
3200 nmh-workers levine post profile
3201 .]
3202 David Levine posted an explanation, quoting John Romine:
3203 .QS
3204 I asked John Romine and here's what he had to say, which
3205 agrees and provides an example that convinces me:
3206 .QS
3207 My take on this is that post should not be called by
3208 users directly, and it doesn't read the .mh_profile
3209 (only front-end UI programs read the profile).
3210 .QP
3211 For example, there can be contexts where post is called
3212 by a helper program (like 'mhmail') which may be run by
3213 a non-MH user. We don't want this to prompt the user
3214 to create an MH profile, etc.
3215 .QP
3216 My suggestion would be to have send pass a (hidden)
3217 `\-fileproc proc' option to post if needed. You could also
3218 use an environment variable (I think send/whatnow do
3219 this).
3220 .QE
3221 I think that's the way to go. My personal preference is to use a command line option, not an environment variable.
3222 .QE
3223 .P
3224 To solve the problem of
3225 .Pn post
3226 not honoring the
3227 .Pe fileproc
3228 profile entry,
3229 the community roughly agreed that a switch
3230 .Sw -fileproc
3231 should be added to
3232 .Pn post
3233 to be able to pass a different fileproc.
3234 I strongly disagree with this approach because it does not solve
3235 the problem; it only removes a single symptom.
3236 The problem is that
3237 .Pn post
3238 does not behave as expected.
3239 But all programs should behave as expected.
3240 Clear and simple concepts are a precondition for this.
3241 Hence, the real solution is having all MH tools read the profile.
3242 .P
3243 Yet, the problem has a further aspect.
3244 It mainly originates in
3245 .Pn mhmail .
3246 .Pn mhmail
3247 was intended to be a replacement for
3248 .Pn mailx
3249 on systems with MH installations.
3250 .Pn mhmail
3251 should have been able to use just like
3252 .Pn mailx ,
3253 but sending the message via MH's
3254 .Pn post
3255 instead of
3256 .Pn sendmail .
3257 Using
3258 .Pn mhmail
3259 should not be influenced by the question whether the user had
3260 MH set up for himself or not.
3261 .Pn mhmail
3262 did not read the profile as this requests the user to set up MH
3263 if not done yet.
3264 As
3265 .Pn mhmail
3266 used
3267 .Pn post ,
3268 .Pn post
3269 could not read the profile neither.
3270 This is the reason why
3271 .Pn post
3272 does not read the profile.
3273 This is the reason for the actual problem.
3274 It was not much of a problem because
3275 .Pn post
3276 was not intended to be used by users directly.
3277 .Pn send
3278 is the interactive front-end to
3279 .Pn post .
3280 .Pn send
3281 read the profile and passed all relevant values on the command line to
3282 .Pn post
3283 \(en an awkward solution.
3284 .P
3285 The important insight is that
3286 .Pn mhmail
3287 is no true MH tool.
3288 The concepts broke because this outlandish tool was treated as any other
3289 MH tool.
3290 Instead it should have been treated accordingly to its foreign style.
3291 The solution is not to prevent the tools reading the profile but
3292 to instruct them reading a different profile.
3293 .Pn mhmail
3294 could have set up a well-defined profile and caused all MH tools
3295 in the session use it by exporting an environment variable.
3296 With this approach, no special cases would have been introduced,
3297 no surprises would have been caused.
3298 By writing a clean-profile-wrapper, the concept could have been
3299 generalized orthogonally to the whole MH toolchest.
3300 Then Rose's motivation behind the decision that
3301 .Pn post
3302 ignores the profile, as quoted by Jeffrey Honig,
3303 .[
3304 nmh-workers post profile
3305 .]
3306 would have become possible:
3307 .QS
3308 when you run mh commands in a script, you want all the defaults to be
3309 what the man page says.
3310 when you run a command by hand, then you want your own defaults...
3311 .QE
3312 .LP
3313 Yet, I consider this explanation short-sighted.
3314 We should rather regard theses two cases as just two different MH setups,
3315 based on two different profiles.
3316 Mapping such problems on the concepts of switching between different
3317 profiles, solves them once for all.
3318 .P
3319 In mmh, the wish to have
3320 .Pn mhmail
3321 as as replacement for
3322 .Pn mailx
3323 is considered obsolete.
3324 Mmh's
3325 .Pn mhmail
3326 does no longer cover this use-case.
3327 Currently,
3328 .Pn mhmail
3329 is in a transition state.
3330 .Ci 32d4f9daaa70519be3072479232ff7be0500d009
3331 It may become a front-end to
3332 .Pn comp ,
3333 which provides an interface more convenient in some cases.
3334 In this case,
3335 .Pn mhmail
3336 will become an ordinary MH tool, reading the profile.
3337 If, however, this idea will not convince, then
3338 .Pn mhmail
3339 will be removed.
3340 .P
3341 Every program in the mmh toolchest reads the profile.
3342 The only exception is
3343 .Pn slocal ,
3344 which is not considered part of the mmh toolchest.
3345 This MDA is only distributed with mmh, currently.
3346 Mmh has no
3347 .Pn post
3348 program, but
3349 .Pn spost ,
3350 which now reads the profile.
3351 .Ci 3e017a7abbdf69bf0dff7a4073275961eda1ded8
3352 With this change,
3353 .Pn send
3354 and
3355 .Pn spost
3356 can be considered to be merged.
3357 Direct invocations of
3358 .Pn spost
3359 are only done by the to-be-changed
3360 .Pn mhmail
3361 implementation and by
3362 .Pn rcvdist ,
3363 which will require rework.
3364 .P
3365 The
3366 .Fu context_foil()
3367 function to pretend to have read an empty profile was removed.
3368 .Ci 68af8da96bea87a5541988870130b6209ce396f6
3369 All mmh tools read the profile.
3373 .H2 "Standard Libraries
3374 .P
3375 MH is one decade older than the POSIX and ANSI C standards.
3376 Hence, MH included own implementations of functions
3377 that are standardized and thus widely available today,
3378 but were not back then.
3379 Today, twenty years after the POSIX and ANSI C were published,
3380 developers can expect system to comply with these standards.
3381 In consequence, MH-specific replacements for standard functions
3382 can and should be dropped.
3383 Kernighan and Pike advise: ``Use standard libraries.''
3384 .[ [
3385 kernighan pike practice of programming
3386 .], p. 196]
3387 Actually, MH had followed this advice in history,
3388 but it had not adjusted to the changes in this field.
3389 The
3390 .Fu snprintf()
3391 function, for instance, was standardized with C99 and is available
3392 almost everywhere because of its high usefulness.
3393 In project's own implementation of
3394 .Fu snprintf()
3395 was dropped in March 2012 in favor for using the one of the
3396 standard library.
3397 .Ci 0052f1024deb0a0a2fc2e5bacf93d45a5a9c9b32
3398 Such decisions limit the portability of mmh
3399 if systems don't support these standardized and widespread functions.
3400 This compromise is made because mmh focuses on the future.
3401 .P
3402 I am not yet thirty years old and my C and Unix experience comprises
3403 only half a dozen years.
3404 Hence, I need to learn about the history in retrospective.
3405 I have not used those ancient constructs myself.
3406 I have not suffered from their incompatibilities.
3407 I have not longed for standardization.
3408 All my programming experience is from a time when ANSI C and POSIX
3409 were well established already.
3410 I have only read a lot of books about the (good) old times.
3411 This puts me in a difficult positions when working with old code.
3412 I need to freshly acquire knowledge about old code constructs and ancient
3413 programming styles, whereas older programmers know these things by
3414 heart from their own experience.
3415 .P
3416 Being aware of the situation, I rather let people with more historic
3417 experience replace ancient code constructs with standardized ones.
3418 Lyndon Nerenberg covered large parts of this task for the nmh project.
3419 He converted project-specific functions to POSIX replacements,
3420 also removing the conditionals compilation of now standardized features.
3421 Ken Hornstein and David Levine had their part in the work, too.
3422 Often, I only needed to pull over changes from nmh into mmh.
3423 These changes include many commits; these are among them:
3424 .Ci 768b5edd9623b7238e12ec8dfc409b82a1ed9e2d
3425 .Ci 0052f1024deb0a0a2fc2e5bacf93d45a5a9c9b32 .
3426 .P
3427 During my own work, I tidied up the \fIMH standard library\fP,
3428 .Fn libmh.a ,
3429 which is located in the
3430 .Fn sbr
3431 (``subroutines'') directory in the source tree.
3432 The MH library includes functions that mmh tools usually need.
3433 Among them are MH-specific functions for profile, context, sequence,
3434 and folder handling, but as well
3435 MH-independent functions, such as auxiliary string functions,
3436 portability interfaces and error-checking wrappers for critical
3437 functions of the standard library.
3438 .P
3439 I have replaced the
3440 .Fu atooi()
3441 function with calls to
3442 .Fu strtoul()
3443 with the third parameter, the base, set to eight.
3444 .Fu strtoul()
3445 is part of C89 and thus considered safe to use.
3446 .Ci c490c51b3c0f8871b6953bd0c74551404f840a74
3447 .P
3448 I did remove project-included fallback implementations of
3449 .Fu memmove()
3450 and
3451 .Fu strerror() ,
3452 although Peter Maydell had re-included them into nmh in 2008
3453 to support SunOS 4.
3454 Nevertheless, these functions are part of ANSI C.
3455 Systems that do not even provide full ANSI C support should not
3456 put a load on mmh.
3457 .Ci b067ff5c465a5d243ce5a19e562085a9a1a97215
3458 .P
3459 The
3460 .Fu copy()
3461 function copies the string in argument one to the location in two.
3462 In contrast to
3463 .Fu strcpy() ,
3464 it returns a pointer to the terminating null-byte in the destination area.
3465 The code was adjusted to replace
3466 .Fu copy()
3467 with
3468 .Fu strcpy() ,
3469 except within
3470 .Fu concat() ,
3471 where
3472 .Fu copy()
3473 was more convenient.
3474 Therefore, the definition of
3475 .Fu copy()
3476 was moved into the source file of
3477 .Fu concat()
3478 and its visibility is now limited to it.
3479 .Ci 552fd7253e5ee9e554c5c7a8248a6322aa4363bb
3480 .P
3481 The function
3482 .Fu r1bindex()
3483 had been a generalized version of
3484 .Fu basename()
3485 with minor differences.
3486 As all calls to
3487 .Fu r1bindex()
3488 had the slash (`/') as delimiter anyway,
3489 replacing
3490 .Fu r1bindex()
3491 with the more specific and better-named function
3492 .Fu basename()
3493 became desirable.
3494 Unfortunately, many of the 54 calls to
3495 .Fu r1bindex()
3496 depended on a special behavior,
3497 which differed from the POSIX specification for
3498 .Fu basename() .
3499 Hence,
3500 .Fu r1bindex()
3501 was kept but renamed to
3502 .Fu mhbasename() ,
3503 fixing the delimiter to the slash.
3504 .Ci 240013872c392fe644bd4f79382d9f5314b4ea60
3505 For possible uses of
3506 .Fu r1bindex()
3507 with a different delimiter,
3508 the ANSI C function
3509 .Fu strrchr()
3510 provides the core functionality.
3511 .P
3512 The
3513 .Fu ssequal()
3514 function \(en apparently for ``substring equal'' \(en
3515 was renamed to
3516 .Fu isprefix() ,
3517 because this is what it actually checks.
3518 .Ci c20b4fa14515c7ab388ce35411d89a7a92300711
3519 Its source file had included the following comments, no joke.
3520 .VS
3521 /*
3522 * THIS CODE DOES NOT WORK AS ADVERTISED.
3523 * It is actually checking if s1 is a PREFIX of s2.
3524 * All calls to this function need to be checked to see
3525 * if that needs to be changed. Prefix checking is cheaper, so
3526 * should be kept if it's sufficient.
3527 */
3529 /*
3530 * Check if s1 is a substring of s2.
3531 * If yes, then return 1, else return 0.
3532 */
3533 VE
3534 Two months later, it was completely removed by replacing it with
3535 .Fu strncmp() .
3536 .Ci b0b1dd37ff515578cf7cba51625189eb34a196cb
3542 .H2 "User Data Locations
3543 .P
3544 In nmh, a personal setup consists of the MH profile and the MH directory.
3545 The profile is a file named
3546 .Fn \&.mh_profile
3547 in the user's home directory.
3548 It contains the static configuration.
3549 It also contains the location of the MH directory in the profile entry
3550 .Pe Path .
3551 The MH directory contains the mail storage and is the first
3552 place to search for personal forms, scan formats, and similar
3553 configuration files.
3554 The location of the MH directory can be chosen freely by the user.
3555 The default and usual name is a directory named
3556 .Fn Mail
3557 in the home directory.
3558 .P
3559 The way MH data is splitted between profile and MH directory is a legacy.
3560 It is only sensible in a situation where the profile is the only
3561 configuration file.
3562 Why else should the mail storage and the configuration files be intermixed?
3563 They are different kinds of data:
3564 The data to be operated on and the configuration to change how
3565 tools operate.
3566 Splitting the configuration between the profile and the MH directory
3567 is bad.
3568 Merging the mail storage and the configuration in one directory is bad
3569 as well.
3570 As the mail storage and the configuration were not separated sensibly
3571 in the first place, I did it now.
3572 .P
3573 Personal mmh data is grouped by type, resulting in two distinct parts:
3574 The mail storage and the configuration.
3575 In mmh, the mail storage directory still contains all the messages,
3576 but, in exception of public sequences files, nothing else.
3577 In difference to nmh, the auxiliary configuration files are no longer
3578 located there.
3579 Therefore, the directory is no longer called the user's \fIMH directory\fP
3580 but his \fImail storage\fP.
3581 Its location is still user-chosen, with the default name
3582 .Fn Mail ,
3583 in the user's home directory.
3584 In mmh, the configuration is grouped together in
3585 the hidden directory
3586 .Fn \&.mmh
3587 in the user's home directory.
3588 This \fImmh directory\fP contains the context file, personal forms,
3589 scan formats, and the like, but also the user's profile, now named
3590 .Fn profile .
3591 The location of the profile is no longer fixed to
3592 .Fn $HOME/.mh_profile
3593 but to
3594 .Fn $HOME/.mmh/profile .
3595 Having both, the file
3596 .Fn $HOME/.mh_profile
3597 and the configuration directory
3598 .Fn $HOME/.mmh
3599 appeared to be inconsistent.
3600 The approach chosen for mmh is consistent, simple, and familiar to
3601 Unix users.
3602 .P
3603 MH allows users to have multiiple MH setups.
3604 Therefore, it is necessary to select a different profile.
3605 The profile is the single entry point to access the rest of a
3606 personal MH setup.
3607 In nmh, the environment variable
3608 .Ev MH
3609 could be used to specifiy a different profile.
3610 To operate in the same MH setup with a separate context,
3611 the
3612 .Ev MHCONTEXT
3613 environment variable could be used.
3614 This allows having own current folders and current messages in
3615 each terminal, for instance.
3616 In mmh, three environment variables are used.
3617 .Ev MMH
3618 overrides the default location of the mmh directory (\c
3619 .Fn .mmh ).
3620 .Ev MMHP
3621 and
3622 .Ev MMHC
3623 override the paths to the profile and context files, respectively.
3624 This approach allows the set of personal configuration files to be chosen
3625 independently from the profile, context, and mail storage.
3626 .P
3627 The separation of the files by type is sensible and convenient.
3628 The new approach has no functional disadvantages,
3629 as every setup I can imagine can be implemented with both approaches,
3630 possibly even easier with the new approach.
3631 The main achievement of the change is the clear and sensible split
3632 between mail storage and configuration.
3638 .H2 "Modularization
3639 .P
3640 The source code of the mmh tools is located in the
3641 .Fn uip
3642 (``user interface programs'') directory.
3643 Each tools has a source file with the same name.
3644 For example,
3645 .Pn rmm
3646 is built from
3647 .Fn uip/rmm.c .
3648 Some source files are used for multiple programs.
3649 For example
3650 .Fn uip/scansbr.c
3651 is used for both,
3652 .Pn scan
3653 and
3654 .Pn inc .
3655 In nmh, 49 tools were built from 76 source files.
3656 This is a ratio of 1.6 source files per program.
3657 32 programs depended on multiple source files;
3658 17 programs depended on one source file only.
3659 In mmh, 39 tools are built from 51 source files.
3660 This is a ratio of 1.3 source files per program.
3661 18 programs depend on multiple source files;
3662 21 programs depend on one source file only.
3663 (These numbers and the ones in the following text ignore the MH library
3664 as well as shell scripts and multiple names for the same program.)
3665 .P
3666 Splitting the source code of a large program into multiple files can
3667 increase the readability of its source code.
3668 Most of the mmh tools, however, are simple and straight-forward programs.
3669 With the exception of the MIME handling tools,
3670 .Pn pick
3671 is the largest tools.
3672 It contains 1\|037 lines of source code (measured with
3673 .Pn sloccount ), excluding the MH library.
3674 Only the MIME handling tools (\c
3675 .Pn mhbuild ,
3676 .Pn mhstore ,
3677 .Pn show ,
3678 etc.)
3679 are larger.
3680 Splitting programs with less than 1\|000 lines of code into multiple
3681 source files seldom leads to better readability.
3682 For such tools, splitting makes sense
3683 when parts of the code are reused in other programs,
3684 and the reused code fragment is not general enough
3685 for including it in the MH library,
3686 or, if the code has dependencies on a library that only few programs need.
3687 .Fn uip/packsbr.c ,
3688 for instance, provides the core program logic for the
3689 .Pn packf
3690 and
3691 .Pn rcvpack
3692 programs.
3693 .Fn uip/packf.c
3694 and
3695 .Fn uip/rcvpack.c
3696 mainly wrap the core function appropriately.
3697 No other tools use the folder packing functions.
3698 As another example,
3699 .Fn uip/termsbr.c
3700 provides termcap support, which requires linking with a termcap or
3701 curses library.
3702 Including
3703 .Fn uip/termsbr.c
3704 into the MH library would require every program to be linked with
3705 termcap or curses, although only few of the programs require it.
3706 .P
3707 The task of MIME handling is complex enough that splitting its code
3708 into multiple source files improves the readability.
3709 The program
3710 .Pn mhstore ,
3711 for instance, is compiled out of seven source files with 2\|500
3712 lines of code in summary.
3713 The main code file
3714 .Fn uip/mhstore.c
3715 consists of 800 lines; the other 1\|700 lines of code are reused in
3716 other MIME handling tools.
3717 It seems to be worthwhile to bundle the generic MIME handling code into
3718 a MH-MIME library, as a companion to the MH standard library.
3719 This is left open for the future.
3720 .P
3721 The work already done, focussed on the non-MIME tools.
3722 The amount of code compiled into each program was reduced.
3723 This eases the understanding of the code base.
3724 In nmh,
3725 .Pn comp
3726 was built from six source files:
3727 .Fn comp.c ,
3728 .Fn whatnowproc.c ,
3729 .Fn whatnowsbr.c ,
3730 .Fn sendsbr.c ,
3731 .Fn annosbr.c ,
3732 and
3733 .Fn distsbr.c .
3734 In mmh, it builds from only two:
3735 .Fn comp.c
3736 and
3737 .Fn whatnowproc.c .
3738 In nmh's
3739 .Pn comp ,
3740 the core function of
3741 .Pn whatnow ,
3742 .Pn send ,
3743 and
3744 .Pn anno
3745 were compiled into
3746 .Pn comp .
3747 This saved the need to execute these programs with
3748 .Fu fork()
3749 and
3750 .Fu exec() ,
3751 two expensive system calls.
3752 Whereis this approach improved the time performance,
3753 it interweaved the source code.
3754 Core functionalities were not encapsulated into programs but into
3755 function, which were then wrapped by programs.
3756 For example,
3757 .Fn uip/annosbr.c
3758 included the function
3759 .Fu annotate() .
3760 Each program that wanted to annotate messages, included the source file
3761 .Fn uip/annosbr.c
3762 and called
3763 .Fu annotate() .
3764 Because the function
3765 .Fu annotate()
3766 was used like the tool
3767 .Pn anno ,
3768 it had seven parameters, reflecting the command line switches of the tool.
3769 When another pair of command line switches was added to
3770 .Pn anno ,
3771 a rather ugly hack was implemented to avoid adding another parameter
3772 to the function.
3773 .Ci d9b1d57351d104d7ec1a5621f090657dcce8cb7f
3774 .P
3775 Separation simplifies the understanding of program code
3776 because the area influenced by any particular statement is smaller.
3777 The separating on the program-level is more strict than the separation
3778 on the function level.
3779 In mmh, the relevant code of
3780 .Pn comp
3781 comprises the two files
3782 .Fn uip/comp.c
3783 and
3784 .Fn uip/whatnowproc.c ,
3785 together 210 lines of code.
3786 In nmh,
3787 .Pn comp
3788 comprises six files with 2\|450 lines.
3789 Not all of the code in these six files was actually used by
3790 .Pn comp ,
3791 but the code reader needed to read all of the code first to know which
3792 parts were used.
3793 .P
3794 As I have read a lot in the code base during the last two years,
3795 I learned about the easy and the difficult parts.
3796 Code is easy to understand if:
3797 .BU
3798 The influenced code area is small.
3799 .BU
3800 The boundaries are strictly defined.
3801 .BU
3802 The code is written straight-forward.
3803 .P
3804 .\" XXX move this paragraph somewhere else?
3805 Reading
3806 .Pn rmm 's
3807 source code in
3808 .Fn uip/rmm.c
3809 is my recommendation for a beginner's entry point into the code base of nmh.
3810 The reasons are that the task of
3811 .Pn rmm
3812 is straight forward and it consists of one small source code file only,
3813 yet its source includes code constructs typical for MH tools.
3814 With the introduction of the trash folder in mmh,
3815 .Pn rmm
3816 became a bit more complex, because it invokes
3817 .Pn refile .
3818 Still, it is a good example for a simple tool with clear sources.
3819 .P
3820 Understanding
3821 .Pn comp
3822 requires to read 210 lines of code in mmh, but ten times as much in nmh.
3823 Due to the aforementioned hack in
3824 .Pn anno
3825 to save the additional parameter, information passed through the program's
3826 source base in obscure ways.
3827 Thus, understanding
3828 .Pn comp ,
3829 required understanding the inner workings of
3830 .Fn uip/annosbr.c
3831 first.
3832 To be sure to fully understand a program, its whole source code needs
3833 to be examined.
3834 Not doing so is a leap of faith, assuming that the developers
3835 have avoided obscure programming techniques.
3836 By separating the tools on the program-level, the boundaries are
3837 clearly visible and technically enforced.
3838 The interfaces are calls to
3839 .Fu exec()
3840 rather than arbitrary function calls.
3841 .P
3842 But the real problem is another:
3843 Nmh violates the golden ``one tool, one job'' rule of the Unix philosophy.
3844 Understanding
3845 .Pn comp
3846 requires understanding
3847 .Fn uip/annosbr.c
3848 and
3849 .Fn uip/sendsbr.c
3850 because
3851 .Pn comp
3852 does annotate and send messages.
3853 In nmh, there surely exists the tool
3854 .Pn send ,
3855 which does (almost) only send messages.
3856 But
3857 .Pn comp
3858 and
3859 .Pn repl
3860 and
3861 .Pn forw
3862 and
3863 .Pn dist
3864 and
3865 .Pn whatnow
3866 and
3867 .Pn viamail ,
3868 they all (!) have the same message sending function included, too.
3869 In result,
3870 .Pn comp
3871 sends messages without using
3872 .Pn send .
3873 The situation is the same as if
3874 .Pn grep
3875 would page without
3876 .Pn more
3877 just because both programs are part of the same code base.
3878 .P
3879 The clear separation on the surface \(en the toolchest approach \(en
3880 is violated on the level below.
3881 This violation is for the sake of time performance.
3882 On systems where
3883 .Fu fork()
3884 and
3885 .Fu exec()
3886 are expensive, the quicker response might be noticable.
3887 In the old times, sacrificing readability and conceptional beauty for
3888 speed might even have been a must to prevent MH from being unusably slow.
3889 Whatever the reasons had been, today they are gone.
3890 No longer should we sacrifice readability or conceptional beauty.
3891 No longer should we violate the Unix philosophy's ``one tool, one job''
3892 guideline.
3893 No longer should we keep speed improvements that became unnecessary.
3894 .P
3895 Therefore, mmh's
3896 .Pn comp
3897 does no longer send messages.
3898 In mmh, different jobs are divided among separate programs that
3899 invoke each other as needed.
3900 In consequence,
3901 .Pn comp
3902 invokes
3903 .Pn whatnow
3904 which thereafter invokes
3905 .Pn send .
3906 The clear separation on the surface is maintained on the level below.
3907 Human users and the tools use the same interface \(en
3908 annotations, for example, are made by invoking
3909 .Pn anno ,
3910 no matter if requested by programs or by human beings.
3911 The decrease of tools built from multiple source files and thus
3912 the decrease of
3913 .Fn uip/*sbr.c
3914 files confirm the improvement.
3915 .P
3916 .\" XXX move this paragraph up somewhere
3917 One disadvantage needs to be taken with this change:
3918 The compiler can no longer check the integrity of the interfaces.
3919 By changing the command line interfaces of tools, it is
3920 the developer's job to adjust the invocations of these tools as well.
3921 As this is a manual task and regression tests, which could detect such
3922 problems, are not available yet, it is prone to errors.
3923 These errors will not be detected at compile time but at run time.
3924 Installing regression tests is a task left to do.
3925 In the best case, a uniform way of invoking tools from other tools
3926 can be developed to allow automated testing at compile time.
3929 .ig
3930 XXX consider writing about mhl vs. mhlproc
3932 sbr/showfile.c
3934 23 /*
3935 24 ** If you have your lproc listed as "mhl",
3936 25 ** then really invoked the mhlproc instead
3937 26 ** (which is usually mhl anyway).
3938 27 */
3940 Sat Nov 24 19:09:14 1984 /mtr (agent: Marshall Rose) <uci@udel-dewey>
3942 sbr/showfile.c: if lproc is "mhl", use mhlproc for consistency
3943 (Actually, user should use "lproc: show", "showproc: mhl".)
3944 ..