docs/master
view discussion.roff @ 169:f4ffe121a0a2
Applied excellent suggestions and corrections by Kate.
| author | markus schnalke <meillo@marmaro.de> |
|---|---|
| date | Tue, 10 Jul 2012 13:08:51 +0200 |
| parents | 277eeb5ba223 |
| children | 346ff7e201f5 |
line source
1 .H0 "Discussion
2 .P
3 This main chapter discusses the practical work accomplished in the
4 mmh project.
5 It is structured along the goals set for the project.
6 The concrete work undertaken
7 is described in the examples of how the general goals were achieved.
8 The discussion compares the current version of mmh with the state of
9 nmh just before the mmh project started, i.e. Fall 2011.
10 Current changes of nmh will be mentioned only as side notes.
11 .\" XXX where do I discuss the parallel development of nmh?
15 .\" --------------------------------------------------------------
16 .H1 "Streamlining
18 .P
19 MH once provided anything necessary for email handling.
20 The community around nmh has the similar understanding that nmh should
21 provide a complete email system.
22 In fundamental contrast, mmh shall be a MUA only.
23 I believe that the development of all-in-one mail systems is obsolete.
24 Today, email is too complex to be fully covered by single projects.
25 Such a project won't be able to excel in all aspects.
26 Instead, the aspects of email should be covered by multiple projects,
27 which then can be combined to form a complete system.
28 Excellent implementations for the various aspects of email already exist.
29 Just to name three examples: Postfix is a specialized MTA,
30 .\" XXX homepages verlinken
31 Procmail is a specialized MDA, and Fetchmail is a specialized MRA.
32 I believe that it is best to use such specialized tools instead of
33 providing the same function again as a side-component in the project.
34 .\" XXX mail agent picture here
35 .P
36 Doing something well requires focusing on a small set of specific aspects.
37 Under the assumption that development focussed on a particular area
38 produces better results there, specialized projects will be superior
39 in their field of focus.
40 Hence, all-in-one mail system projects \(en no matter if monolithic
41 or modular \(en will never be the best choice in any of the fields.
42 Even in providing the best consistent all-in-one system, they are likely
43 to be beaten by projects that focus only on integrating existing mail
44 components to create a homogeneous system.
45 .P
46 The limiting resource in the community development of Free Software
47 is usually man power.
48 If the development power is spread over a large development area,
49 it becomes even more difficult to compete with the specialists in the
50 various fields.
51 The concrete situation for MH-based mail systems is even tougher,
52 given their small and aged community, concerning both developers and users.
53 .P
54 In consequence, I believe that the available development resources
55 should focus on the point where MH is most unique.
56 This is clearly the user interface \(en the MUA.
57 Peripheral parts should be removed to streamline mmh for the MUA task.
60 .H2 "Mail Transfer Facilities
61 .Id mail-transfer-facilities
62 .P
63 In contrast to nmh, which also provides mail submission and mail retrieval
64 agents, mmh is a MUA only.
65 This general difference initiated the development of mmh.
66 The removal of the mail transfer facilities was the first work task
67 in the mmh project.
68 .P
69 Focusing on one mail agent role only, is motivated by Eric Allman's
70 experience with Sendmail.
71 He identified the limitation of Sendmail to the MTA task as one reason for
72 its success:
73 .[ [
74 costales sendmail
75 .], p. xviii]
76 .QS
77 Second, I limited myself to the routing function \(en
78 I wouldn't write user agents or delivery back-ends.
79 This was a departure of the dominant through of the time,
80 in which routing logic, local delivery, and often the network code
81 were incorporated directly into the user agents.
82 .QE
83 .P
84 In nmh, the Mail Submission Agent (MSA) is called
85 \fIMessage Transfer Service\fP (MTS).
86 This facility, implemented by the
87 .Pn post
88 command, established network connections and spoke SMTP to submit
89 messages to be relayed to the outside world.
90 The changes in email demanded changes in this part of nmh as well.
91 Encryption and authentication for network connections
92 needed to be supported, hence TLS and SASL were introduced into nmh.
93 This added complexity to nmh without improving it in its core functions.
94 Also, keeping up with recent developments in the field of
95 mail transfer requires development power and specialists.
96 In mmh, this whole facility was simply cut off.
97 .Ci f6aa95b724fd8c791164abe7ee5468bf5c34f226
98 .Ci fecd5d34f65597a4dfa16aeabea7d74b191532c3
99 .Ci 156d35f6425bea4c1ed3c4c79783dc613379c65b
100 Instead, mmh depends on an external MSA.
101 The only outgoing interface available to mmh is the
102 .Pn sendmail
103 command, which almost any MSA provides.
104 If not, a wrapper program can be written.
105 It must read the message from the standard input, extract the
106 recipient addresses from the message header, and hand the message
107 over to the MSA.
108 For example, a wrapper script for qmail would be:
109 .VS
110 #!/bin/sh
111 exec qmail-inject # ignore command line arguments
112 VE
113 The requirement to parse the recipient addresses out of the message header
114 is likely to be removed in the future.
115 Then mmh would pass the recipient addresses as command line arguments.
116 This appears to be the better interface.
117 .\" XXX implement it
118 .P
119 To retrieve mail, the
120 .Pn inc
121 command acted as an Mail Retrieval Agent (MRA).
122 It established network connections
123 and spoke POP3 to retrieve mail from remote servers.
124 As with mail submission, the network connections required encryption and
125 authentication, thus TLS and SASL were added.
126 Support for message retrieval through IMAP will soon become necessary
127 additions, too, and likewise for any other changes in mail transfer.
128 Not so for mmh because it has dropped the support for retrieving mail
129 from remote locations.
130 .Ci ab7b48411962d26439f92f35ed084d3d6275459c
131 Instead, it depends on an external tool to cover this task.
132 Mmh has two paths for messages to enter mmh's mail storage:
133 (1) Mail can be incorporated with
134 .Pn inc
135 from the system maildrop, or (2) with
136 .Pn rcvstore
137 by reading them, one at a time, from the standard input.
138 .P
139 With the removal of the MSA and MRA, mmh converted from an all-in-one
140 mail system to being a MUA only.
141 Now, of course, mmh depends on third-party software.
142 An external MSA is required to transfer mail to the outside world;
143 an external MRA is required to retrieve mail from remote machines.
144 Excellent implementations of such software exist,
145 which likely are superior than the internal version.
146 Additionally, the best suiting programs can be freely chosen.
147 .P
148 As it had already been possible to use an external MSA or MRA,
149 why not keep the internal version for convenience?
150 .\" XXX ueberleitung
151 The question whether there is sense in having a fall-back pager in all
152 the command line tools, for the cases when
153 .Pn more
154 or
155 .Pn less
156 aren't available, appears to be ridiculous.
157 Of course, MSAs and MRAs are more complex than text pagers
158 and not necessarily available but still the concept of orthogonal
159 design holds: ``Write programs that do one thing and do it well.''
160 .[
161 mcilroy unix phil
162 p. 53
163 .]
164 .[
165 mcilroy bstj foreword
166 .]
167 Here, this part of the Unix philosophy was applied not only
168 to the programs but to the project itself.
169 In other words:
170 Develop projects that focus on one thing and do it well.
171 Projects which have grown complex should be split, for the same
172 reasons that programs which have grown complex should be split.
173 If it is conceptionally more elegant to have the MSA and MRA as
174 separate projects then they should be separated.
175 In my opinion, this is the case here.
176 The RFCs propose this separation by clearly distinguishing the different
177 mail handling tasks.
178 .[
179 rfc 821
180 .]
181 The small interfaces between the mail agents support the separation.
182 .P
183 Email once had been small and simple.
184 At that time,
185 .Pn /bin/mail
186 had covered everything there was to email and still was small and simple.
187 Later, the essential complexity of email increased.
188 (Essential complexity is the complexity defined by the problem itself.\0
189 .[[
190 brooks no silver bullet
191 .]])
192 Email systems reacted to this change: They grew.
193 RFCs started to introduce the concept of mail agents to separate the
194 various tasks because they became more extensive and new tasks appeared.
195 As the mail systems grew even more, parts were split off.
196 For instance, a POP server was included in the original MH;
197 it was removed in nmh.
198 Now is the time to go one step further and split off the MSA and MRA, too.
199 Not only does this decrease the code size of the project,
200 more importantly, it unburdens mmh of the whole field of
201 message transfer with all its implications for the project.
202 There is no more need for concern with changes in network transfer.
203 This independence is gained by depending on an external program
204 that covers the field.
205 Today, this is a reasonable exchange.
206 .P
207 .\" XXX ueberleitung ???
208 Functionality can be added in three different ways:
209 .BU
210 Implementing the function in the project itself.
211 .BU
212 Depending on a library that provides the function.
213 .BU
214 Depending on a program that provides the function.
215 .P
216 .\" XXX Rework sentence
217 While implementing the function in the project itself leads to the
218 largest increase in code size and requires the most maintenance
219 and development work,
220 it increases the project's independence of other software the most.
221 Using libraries or external programs requires less maintenance work
222 but introduces dependencies on external software.
223 Programs have the smallest interfaces and provide the best separation,
224 but possibly limit the information exchange.
225 External libraries are more strongly connected than external programs,
226 thus information can be exchanged in a more flexible manner.
227 Adding code to a project increases maintenance work.
228 .\" XXX ref
229 Implementing complex functions in the project itself adds
230 a lot of code.
231 This should be avoided if possible.
232 Hence, the dependencies only change in their character,
233 not in their existence.
234 In mmh, library dependencies on
235 .Pn libsasl2
236 and
237 .Pn libcrypto /\c
238 .Pn libssl
239 were traded against program dependencies on an MSA and an MRA.
240 This also meant trading build-time dependencies against run-time
241 dependencies.
242 Besides providing stronger separation and greater flexibility,
243 program dependencies also allowed
244 over 6\|000 lines of code to be removed from mmh.
245 This made mmh's code base about 12\|% smaller.
246 Reducing the project's code size by such an amount without actually
247 losing functionality is a convincing argument.
248 Actually, as external MSAs and MRAs are likely superior to the
249 project's internal versions, the common user even gains functionality.
250 .P
251 Users of MH should not have problems setting up an external MSA and MRA.
252 Also, the popular MSAs and MRAs have large communities and a lot
253 of available documentation.
254 Choices for MSAs range from full-featured MTAs such as
255 .\" XXX refs
256 .I Postfix ,
257 over mid-size MTAs such as
258 .I masqmail
259 and
260 .I dma ,
261 to small forwarders such as
262 .I ssmtp
263 and
264 .I nullmailer .
265 Choices for MRAs include
266 .I fetchmail ,
267 .I getmail ,
268 .I mpop
269 and
270 .I fdm .
273 .H2 "Non-MUA Tools
274 .P
275 One goal of mmh is to remove the tools that are not part of the MUA's task.
276 Further more, any tools that don't significantly improve the MUA's job
277 should be removed.
278 Loosely related and rarely used tools distract from the lean appearance.
279 They require maintenance work without adding much to the core task.
280 By removing these tools, the project shall become more streamlined
281 and focused.
282 In mmh, the following tools are not available anymore:
283 .BU
284 .Pn conflict
285 was removed
286 .Ci 8b235097cbd11d728c07b966cf131aa7133ce5a9
287 because it is a mail system maintenance tool that is not MUA-related.
288 It even checked
289 .Fn /etc/passwd
290 and
291 .Fn /etc/group
292 for consistency, which is completely unrelated to email.
293 A tool like
294 .Pn conflict
295 is surely useful, but it should not be shipped with mmh.
296 .\" XXX historic reasons?
297 .BU
298 .Pn rcvtty
299 was removed
300 .Ci 14767c94b3827be7c867196467ed7aea5f6f49b0
301 because its use case of writing to the user's terminal
302 on receival of mail is obsolete.
303 If users like to be informed of new mail, the shell's
304 .Ev MAILPATH
305 variable or graphical notifications are technically more appealing.
306 Writing directly to terminals is hardly ever desired today.
307 If, though, one prefers this approach, the standard tool
308 .Pn write
309 can be used in a way similar to:
310 .VS
311 scan -file - | write `id -un`
312 VE
313 .BU
314 .Pn viamail
315 .\" XXX was macht viamail
316 was removed
317 .Ci eda72d6a7a7c20ff123043fb7f19c509ea01f932
318 when the new attachment system was activated, because
319 .Pn forw
320 could then cover the task itself.
321 The program
322 .Pn sendfiles
323 was rewritten as a shell script wrapper around
324 .Pn forw .
325 .Ci 0e82199cf3c991a173e0ac8aa776efdb3ded61e6
326 .BU
327 .Pn msgchk
328 .\" XXX was macht msgchk
329 was removed
330 .Ci bb9360ead7eb7a3fedcce2eeedfc660014e41dbe ,
331 because it lost its use case when POP support was removed.
332 A call to
333 .Pn msgchk
334 provided hardly more information than:
335 .VS
336 ls -l /var/mail/meillo
337 VE
338 It did distinguish between old and new mail, but
339 these details can be retrieved with
340 .Pn stat (1),
341 too.
342 A small shell script could be written to print the information
343 in a similar way, if truly necessary.
344 As mmh's
345 .Pn inc
346 only incorporates mail from the user's local maildrop,
347 and thus no data transfers over slow networks are involved,
348 there is hardly any need to check for new mail before incorporating it.
349 .BU
350 .Pn msh
351 was removed
352 .Ci 916690191222433a6923a4be54b0d8f6ac01bd02
353 because the tool was in conflict with the philosophy of MH.
354 It provided an interactive shell to access the features of MH,
355 but it wasn't just a shell tailored to the needs of mail handling.
356 Instead, it was one large program that had several MH tools built in.
357 This conflicts with the major feature of MH of being a tool chest.
358 .Pn msh 's
359 main use case had been accessing Bulletin Boards, which have ceased to
360 be popular.
361 .P
362 Removing
363 .Pn msh
364 together with the truly archaic code relicts
365 .Pn vmh
366 and
367 .Pn wmh
368 saved more than 7\|000 lines of C code \(en
369 about 15\|% of the project's original source code amount.
370 Having less code \(en with equal readability, of course \(en
371 for the same functionality is an advantage.
372 Less code means less bugs and less maintenance work.
373 As
374 .Pn rcvtty
375 and
376 .Pn msgchk
377 are assumed to be rarely used and can be implemented in different ways,
378 why should one keep them?
379 Removing them streamlines mmh.
380 .Pn viamail 's
381 use case is now partly obsolete and partly covered by
382 .Pn forw ,
383 hence there's no reason to still maintain it.
384 .Pn conflict
385 is not related to the mail client, and
386 .Pn msh
387 conflicts with the basic concept of MH.
388 These two tools might still be useful, but they should not be part of mmh.
389 .P
390 Finally, there is
391 .Pn slocal .
392 .Pn slocal
393 is an MDA and thus not directly MUA-related.
394 It should be removed from mmh, because including it conflicts with
395 the idea that mmh is a MUA only.
396 .Pn slocal
397 should rather become a separate project.
398 However,
399 .Pn slocal
400 provides rule-based processing of messages, like filing them into
401 different folders, which is otherwise not available in mmh.
402 Although
403 .Pn slocal
404 neither pulls in dependencies, nor does it include a separate
405 technical area (cf. Sec.
406 .Cf mail-transfer-facilities ),
407 it still accounts for about 1\|000 lines of code that need to be maintained.
408 As
409 .Pn slocal
410 is almost self-standing, it should be split off into a separate project.
411 This would cut the strong connection between the MUA mmh and the MDA
412 .Pn slocal .
413 For anyone not using MH,
414 .Pn slocal
415 would become yet another independent MDA, like
416 .I procmail .
417 Then
418 .Pn slocal
419 could be installed without the complete MH system.
420 Likewise, mmh users could decide to use
421 .I procmail
422 without having a second, unused MDA,
423 .Pn slocal ,
424 installed.
425 That appears to be conceptionally the best solution.
426 Yet,
427 .Pn slocal
428 is not split off.
429 I defer the decision over
430 .Pn slocal
431 out of a need for deeper investigation.
432 In the meanwhile, it remains part of mmh.
433 However, its continued existence is not significant because
434 .Pn slocal
435 is unrelated to the rest of the project.
439 .H2 "Displaying Messages
440 .Id mhshow
441 .P
442 Since the very beginning, already in the first concept paper,
443 .\" XXX ref!!!
444 .Pn show
445 had been MH's message display program.
446 .Pn show
447 mapped message numbers and sequences to files and invoked
448 .Pn mhl
449 to have the files formatted.
450 With MIME, this approach wasn't sufficient anymore.
451 MIME messages can consist of multiple parts. Some parts are not
452 directly displayable and text content might be encoded in
453 foreign charsets.
454 .Pn show 's
455 understanding of messages and
456 .Pn mhl 's
457 display capabilities couldn't cope with the task any longer.
458 .P
459 Instead of extending these tools, additional tools were written from
460 scratch and added to the MH tool chest.
461 Doing so is encouraged by the tool chest approach.
462 Modular design is a great advantage for extending a system,
463 as new tools can be added without interfering with existing ones.
464 First, the new MIME features were added in form of the single program
465 .Pn mhn .
466 The command
467 .Cl "mhn -show 42
468 would show the MIME message numbered 42.
469 With the 1.0 release of nmh in February 1999, Richard Coleman finished
470 the split of
471 .Pn mhn
472 into a set of specialized tools, which together covered the
473 multiple aspects of MIME.
474 One of them was
475 .Pn mhshow ,
476 which replaced
477 .Cl "mhn -show" .
478 It was capable of displaying MIME messages appropriately.
479 .P
480 From then on, two message display tools were part of nmh,
481 .Pn show
482 and
483 .Pn mhshow .
484 To ease the life of users,
485 .Pn show
486 was extended to automatically hand the job over to
487 .Pn mhshow
488 if displaying the message would be beyond
489 .Pn show 's
490 abilities.
491 In consequence, the user would simply invoke
492 .Pn show
493 (possibly through
494 .Pn next
495 or
496 .Pn prev )
497 and get the message printed with either
498 .Pn show
499 or
500 .Pn mhshow ,
501 whatever was more appropriate.
502 .P
503 Having two similar tools for essentially the same task is redundant.
504 Usually,
505 users wouldn't distinguish between
506 .Pn show
507 and
508 .Pn mhshow
509 in their daily mail reading.
510 Having two separate display programs was therefore mainly unnecessary
511 from a user's point of view.
512 Besides, the development of both programs needed to be in sync,
513 to ensure that the programs behaved in a similar way,
514 because they were used like a single tool.
515 Different behavior would have surprised the user.
516 .P
517 Today, non-MIME messages are rather seen to be a special case of
518 MIME messages, although it is the other way round.
519 As
520 .Pn mhshow
521 had already been able to display non-MIME messages, it appeared natural
522 to drop
523 .Pn show
524 in favor of using
525 .Pn mhshow
526 exclusively.
527 .Ci 4c1efddfd499300c7e74263e57d8aa137e84c853
528 Removing
529 .Pn show
530 is no loss in function, because functionally
531 .Pn mhshow
532 covers it completely.
533 The old behavior of
534 .Pn show
535 can still be emulated with the simple command line:
536 .VS
537 mhl `mhpath c`
538 VE
539 .P
540 For convenience,
541 .Pn mhshow
542 was renamed to
543 .Pn show
544 after
545 .Pn show
546 was gone.
547 It is clear that such a rename may confuse future developers when
548 trying to understand the history.
549 Nevertheless, I consider the convenience on the user's side,
550 to call
551 .Pn show
552 when they want a message to be displayed, to outweigh the inconvenience
553 on the developer's side when understanding the project history.
554 .P
555 To prepare for the transition,
556 .Pn mhshow
557 was reworked to behave more like
558 .Pn show
559 first.
560 (cf. Sec.
561 .Cf mhshow )
562 .\" XXX code commits?
563 Once the tools behaved more alike, the replacing appeared to be
564 even more natural.
565 Today, mmh's new
566 .Pn show
567 has become the one single message display program once more,
568 with the difference
569 that today it handles MIME messages as well as non-MIME messages.
570 The outcome of the transition is one program less to maintain,
571 no second display program for users to deal with,
572 and less system complexity.
573 .P
574 Still, removing the old
575 .Pn show
576 hurts in one regard: It had been such a simple program.
577 Its lean elegance is missing from the new
578 .Pn show ,
579 .\" XXX
580 however there is no alternative;
581 supporting MIME demands higher essential complexity.
583 .ig
584 XXX
585 Consider including text on scan listings here
587 Scan listings shall not contain body content. Hence, removed this feature.
588 Scan listings shall operator on message headers and non-message information
589 only. Displaying the beginning of the body complicates everything too much.
590 That's no surprise, because it's something completely different. If you
591 want to examine the body, then use show(1)/mhshow(1).
592 Changed the default scan formats accordingly.
593 .Ci 70b2643e0da8485174480c644ad9785c84f5bff4
594 ..
599 .H2 "Configure Options
600 .P
601 Customization is a double-edged sword.
602 It allows better suiting setups, but not for free.
603 There is the cost of code complexity to be able to customize.
604 There is the cost of less tested setups, because there are
605 more possible setups and especially corner-cases.
606 Additionally, there is the cost of choice itself.
607 The code complexity directly affects the developers.
608 Less tested code affects both, users and developers.
609 The problem of choice affects the users, for once by having to choose,
610 but also by more complex interfaces that require more documentation.
611 Whenever options add few advantages but increase the complexity of the
612 system, they should be considered for removal.
613 I have reduced the number of project-specific configure options from
614 fifteen to three.
616 .U3 "Mail Transfer Facilities
617 .P
618 With the removal of the mail transfer facilities five configure
619 options vanished:
620 .P
621 The switches
622 .Sw --with-tls
623 and
624 .Sw --with-cyrus-sasl
625 had activated the support for transfer encryption and authentication.
626 .\" XXX cf
627 .\" XXX gruende kurz wiederholen
628 This is not needed anymore.
629 .Ci fecd5d34f65597a4dfa16aeabea7d74b191532c3
630 .Ci 156d35f6425bea4c1ed3c4c79783dc613379c65b
631 .P
632 .\" XXX cf
633 .\" XXX ``For the same reason ...''
634 The configure switch
635 .Sw --enable-pop
636 activated the message retrieval facility.
637 The code area that would be conditionally compiled in for TLS and SASL
638 support had been small.
639 The conditionally compiled code area for POP support had been much larger.
640 Whereas the code base changes would only slightly change on toggling
641 TLS or SASL support, it changed much on toggling POP support.
642 The changes in the code base could hardly be overviewed.
643 By having POP support togglable, a second code base had been created,
644 one that needed to be tested.
645 This situation is basically similar for the conditional TLS and SASL
646 code, but there the changes are minor and can yet be overviewed.
647 Still, conditional compilation of a code base creates variations
648 of the original program.
649 More variations require more testing and maintenance work.
650 .P
651 Two other options only specified default configuration values:
652 .Sw --with-mts
653 defined the default transport service.
654 .Ci f6aa95b724fd8c791164abe7ee5468bf5c34f226
655 With
656 .Sw --with-smtpservers
657 default SMTP servers could be specified.
658 .Ci 128545e06224233b7e91fc4c83f8830252fe16c9
659 Both of them became irrelevant when the SMTP transport service was removed.
660 .\" XXX code ref
661 In mmh, all messages are handed over to
662 .Pn sendmail
663 for transportation.
666 .U3 "Backup Prefix
667 .P
668 The backup prefix is the string that was prepended to message
669 filenames to tag them as deleted.
670 By default it had been the comma character `\f(CW,\fP'.
671 .\" XXX Zeitlich ordnen
672 In July 2000, Kimmo Suominen introduced
673 the configure option
674 .Sw --with-hash-backup
675 to change the default to the hash symbol `\f(CW#\fP'.
676 The choice was probably personal preference, because first, the
677 option was named
678 .Sw --with-backup-prefix.
679 and had the prefix symbol as argument.
680 But giving the hash symbol as argument caused too many problems
681 for Autoconf,
682 thus the option was limited to use the hash symbol as the default prefix.
683 This supports the assumption, that the choice for the hash was
684 personal preference only.
685 Being related or not, words that start with the hash symbol
686 introduce a comment in the Unix shell.
687 Thus, the command line
688 .Cl "rm #13 #15
689 calls
690 .Pn rm
691 without arguments because the first hash symbol starts the comment
692 that reaches until the end of the line.
693 To delete the backup files,
694 .Cl "rm ./#13 ./#15"
695 needs to be used.
696 Using the hash as backup prefix can be seen as a precaution against
697 data loss.
698 .P
699 First, I removed the configure option but added the profile entry
700 .Pe backup-prefix ,
701 which allows to specify an arbitrary string as backup prefix.
702 .Ci 6c40d481d661d532dd527eaf34cebb6d3f8ed086
703 Profile entries are the common method to change mmh's behavior.
704 This change did not remove the choice but moved it to a location where
705 it suited better.
706 .P
707 Eventually, however, the new trash folder concept
708 (cf. Sec.
709 .Cf trash-folder )
710 removed the need for the backup prefix completely.
711 .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
712 .Ci ca0b3e830b86700d9e5e31b1784de2bdcaf58fc5
715 .U3 "Editor and Pager
716 .P
717 The two configure options
718 .CW --with-editor=EDITOR
719 .CW --with-pager=PAGER
720 were used to specify the default editor and pager at configure time.
721 Doing so at configure time made sense in the eighties,
722 when the set of available editors and pagers varied much across
723 different systems.
724 Today, the situation is more homogeneous.
725 The programs
726 .Pn vi
727 and
728 .Pn more
729 can be expected to be available on every Unix system,
730 as they are specified by POSIX since two decades.
731 (The specifications for
732 .Pn vi
733 and
734 .Pn more
735 appeared in
736 .[
737 posix 1987
738 .]
739 and,
740 .[
741 posix 1992
742 .]
743 respectively.)
744 As a first step, these two tools were hard-coded as defaults.
745 .Ci 5d43a99db70c12a673028c7758c20cbe3e13ef5f
746 Not changed were the
747 .Pe editor
748 and
749 .Pe moreproc
750 profile entries, which allowed the user to override the system defaults.
751 Later, the concept was reworked to respect the standard environment
752 variables
753 .Ev VISUAL
754 and
755 .Ev PAGER
756 if they are set.
757 Today, mmh determines the editor to use in the following order,
758 taking the first available and non-empty item:
759 .IP (1)
760 Environment variable
761 .Ev MMHEDITOR
762 .IP (2)
763 Profile entry
764 .Pe Editor
765 .IP (3)
766 Environment variable
767 .Ev VISUAL
768 .IP (4)
769 Environment variable
770 .Ev EDITOR
771 .IP (5)
772 Command
773 .Pn vi .
774 .P
775 .Ci f85f4b7ae62e3d05a945dcd46ead51f0a2a89a9b
776 .P
777 The pager to use is determined in a similar order,
778 also taking the first available and non-empty item:
779 .IP (1)
780 Environment variable
781 .Ev MMHPAGER
782 .IP (2)
783 Profile entry
784 .Pe Pager
785 (replaces
786 .Pe moreproc )
787 .IP (3)
788 Environment variable
789 .Ev PAGER
790 .IP (4)
791 Command
792 .Pn more .
793 .P
794 .Ci 0c4214ea2aec6497d0d67b436bbee9bc1d225f1e
795 .P
796 By respecting the
797 .Ev VISUAL /\c
798 .Ev EDITOR
799 and
800 .Ev PAGER
801 environment variables,
802 the new behavior confirms better to the common style on Unix systems.
803 Additionally, the new approach is more uniform and clearer to users.
806 .U3 "ndbm
807 .P
808 .Pn slocal
809 used to depend on
810 .I ndbm ,
811 a database library.
812 The database is used to store the `\fLMessage-ID\fP's of all
813 messages delivered.
814 This enables
815 .Pn slocal
816 to suppress delivering the same message to the same user twice.
817 (This features was enabled by the
818 .Sw -suppressdup
819 switch.)
820 .P
821 A variety of versions of the database library exist.
822 .[
823 wolter unix incompat notes dbm
824 .]
825 Complicated autoconf code was needed to detect them correctly.
826 Further more, the configure switches
827 .Sw --with-ndbm=ARG
828 and
829 .Sw --with-ndbmheader=ARG
830 were added to help with difficult setups that would
831 not be detected automatically or correctly.
832 .P
833 By removing the suppress duplicates feature of
834 .Pn slocal ,
835 the dependency on
836 .I ndbm
837 vanished and 120 lines of complex autoconf code could be saved.
838 .Ci ecd6d6a20cb7a1507e3a20d6c4cb3a1cf14c6bbf
839 The change removed functionality too, but that is minor to the
840 improvement by dropping the dependency and the complex autoconf code.
841 .\" XXX argument: slocal ist sowieso nicht teil vom mmh kern
843 .U3 "mh-e Support
844 .P
845 The configure option
846 .Sw --disable-mhe
847 was removed when the mh-e support was reworked.
848 Mh-e is the Emacs front-end to MH.
849 It requires MH to provide minor additional functions.
850 The
851 .Sw --disable-mhe
852 configure option could switch these extensions off.
853 After removing the support for old versions of mh-e,
854 only the
855 .Sw -build
856 switches of
857 .Pn forw
858 and
859 .Pn repl
860 are left to be mh-e extensions.
861 They are now always built in because they add little code and complexity.
862 In consequence, the
863 .Sw --disable-mhe
864 configure option was removed
865 .Ci a7ce7b4a580d77b6c2c4d980812beb589aa4c643
866 Removing the option removed a second code setup that would have
867 needed to be tested.
868 .\" XXX datum?
869 This change was first accomplished in nmh and thereafter merged into mmh.
870 .P
871 The interface changes in mmh require mh-e to be adjusted in order
872 to be able to use mmh as back-end.
873 This will require minor changes to mh-e, but removing the
874 .Sw -build
875 switches would require more rework.
877 .U3 "Masquerading
878 .P
879 The configure option
880 .Sw --enable-masquerade
881 could take up to three arguments:
882 `draft_from', `mmailid', and `username_extension'.
883 They activated different types of address masquerading.
884 All of them were implemented in the SMTP-speaking
885 .Pn post
886 command, which provided an MSA.
887 Address masquerading is an MTA's task and mmh does not cover
888 this field anymore.
889 Hence, true masquerading needs to be implemented in the external MTA.
890 .P
891 The
892 .I mmailid
893 masquerading type is the oldest one of the three and the only one
894 available in the original MH.
895 It provided a
896 .I username
897 to
898 .I fakeusername
899 mapping, based on the password file's GECOS field.
900 The man page
901 .Mp mh-tailor(5)
902 described the use case as being the following:
903 .QS
904 This is useful if you want the messages you send to always
905 appear to come from the name of an MTA alias rather than your
906 actual account name. For instance, many organizations set up
907 `First.Last' sendmail aliases for all users. If this is
908 the case, the GECOS field for each user should look like:
909 ``First [Middle] Last <First.Last>''
910 .QE
911 .P
912 As mmh sends outgoing mail via the local MTA only,
913 the best location to do such global rewrites is there.
914 Besides, the MTA is conceptionally the right location because it
915 does the reverse mapping for incoming mail (aliasing), too.
916 Further more, masquerading set up there is readily available for all
917 mail software on the system.
918 Hence, mmailid masquerading was removed.
919 .Ci 0836c8000ccb34b59410ef1c15b1b7feac70ce5f
920 .P
921 The
922 .I username_extension
923 masquerading type did not replace the username but would append a suffix,
924 specified by the
925 .Ev USERNAME_EXTENSION
926 environment variable, to it.
927 This provided support for the
928 .I user-extension
929 feature of qmail and the similar
930 .I "plussed user
931 processing of sendmail.
932 The decision to remove this username_extension masquerading was
933 motivated by the fact that
934 .Pn spost
935 hadn't supported it already.
936 .Ci 2abae0bfd0ad5bf898461e50aa4b466d641f23d9
937 Username extensions are possible in mmh, but less convenient to use.
938 .\" XXX covered by next paragraph
939 .\" XXX format file %(getenv USERNAME_EXTENSION)
940 .P
941 The
942 .I draft_from
943 masquerading type instructed
944 .Pn post
945 to use the value of the
946 .Hd From
947 header field as SMTP envelope sender.
948 Sender addresses could be replaced completely.
949 .Ci b14ea6073f77b4359aaf3fddd0e105989db9
950 Mmh offers a kind of masquerading similar in effect, but
951 with technical differences.
952 As mmh does not transfer messages itself, the local MTA has final control
953 over the sender's address. Any masquerading mmh introduces may be reverted
954 by the MTA.
955 In times of pedantic spam checking, an MTA will take care to use
956 sensible envelope sender addresses to keep its own reputation up.
957 Nonetheless, the MUA can set the
958 .Hd From
959 header field and thereby propose
960 a sender address to the MTA.
961 The MTA may then decide to take that one or generate the canonical sender
962 address for use as envelope sender address.
963 .P
964 In mmh, the MTA will always extract the recipient and sender from the
965 message header (\c
966 .Pn sendmail 's
967 .Sw -t
968 switch).
969 The
970 .Hd From
971 header field of the draft may be set arbitrary by the user.
972 If it is missing, the canonical sender address will be generated by the MTA.
974 .U3 "Remaining Options
975 .P
976 Two configure options remain in mmh.
977 One is the locking method to use:
978 .Sw --with-locking=[dot|fcntl|flock|lockf] .
979 The idea of removing all methods except the portable dot locking
980 and having that one as the default is appealing, but this change
981 requires deeper technical investigation into the topic.
982 The other option,
983 .Sw --enable-debug ,
984 compiles the programs with debugging symbols and does not strip them.
985 This option is likely to stay.
990 .H2 "Command Line Switches
991 .P
992 The command line switches of MH tools follow the X Window style.
993 They are words, introduced by a single dash.
994 For example:
995 .Cl "-truncate" .
996 Every program in mmh has two generic switches:
997 .Sw -help ,
998 to print a short message on how to use the program, and
999 .Sw -Version
1000 (with capital `V'), to tell what version of mmh the program belongs to.
1001 .P
1002 Switches change the behavior of programs.
1003 Programs that do one thing in one way require no switches.
1004 In most cases, doing something in exactly one way is too limiting.
1005 If there is basically one task to accomplish, but it should be done
1006 in various ways, switches are a good approach to alter the behavior
1007 of a program.
1008 Changing the behavior of programs provides flexibility and customization
1009 to users, but at the same time it complicates the code, documentation and
1010 usage of the program.
1011 .\" XXX: Ref
1012 Therefore, the number of switches should be kept small.
1013 A small set of well-chosen switches does no harm.
1014 But usually, the number of switches increases over time.
1015 Already in 1985, Rose and Romine have identified this as a major
1016 problem of MH:
1017 .[ [
1018 rose romine real work
1019 .], p. 12]
1020 .QS
1021 A complaint often heard about systems which undergo substantial development
1022 by many people over a number of years, is that more and more options are
1023 introduced which add little to the functionality but greatly increase the
1024 amount of information a user needs to know in order to get useful work done.
1025 This is usually referred to as creeping featurism.
1026 .QP
1027 Unfortunately MH, having undergone six years of off-and-on development by
1028 ten or so well-meaning programmers (the present authors included),
1029 suffers mightily from this.
1030 .QE
1031 .P
1032 Being reluctant to adding new switches \(en or `options',
1033 as Rose and Romine call them \(en is one part of a counter-action,
1034 the other part is removing hardly used switches.
1035 Nmh's tools had lots of switches already implemented,
1036 hence, cleaning up by removing some of them was the more important part
1037 of the counter-action.
1038 Removing existing functionality is always difficult because it
1039 breaks programs that use these functions.
1040 Also, for every obsolete feature, there'll always be someone who still
1041 uses it and thus opposes its removal.
1042 This puts the developer into the position,
1043 where sensible improvements to style are regarded as destructive acts.
1044 Yet, living with the featurism is far worse, in my eyes, because
1045 future needs will demand adding further features,
1046 worsening the situation more and more.
1047 Rose and Romine added in a footnote,
1048 ``[...]
1049 .Pn send
1050 will no doubt acquire an endless number of switches in the years to come.''
1051 Although clearly humorous, the comment points to the nature of the problem.
1052 Refusing to add any new switches would encounter the problem at its root,
1053 but this is not practical.
1054 New needs will require new switches and it would be unwise to block
1055 them strictly.
1056 Nevertheless, removing obsolete switches still is an effective approach
1057 to deal with the problem.
1058 Working on an experimental branch without an established user base,
1059 eased my work because I did not offend users when I removed existing
1060 functions.
1061 .P
1062 Rose and Romine counted 24 visible and 9 more hidden switches for
1063 .Pn send .
1064 In nmh, they increased up to 32 visible and 12 hidden ones.
1065 At the time of writing, no more than 7 visible switches and 1 hidden switch
1066 have remained in mmh's
1067 .Pn send .
1068 (These numbers include two generic switches, help and version.)
1069 .P
1070 The figure displays the number of switches for each of the tools
1071 that is available in both nmh and mmh.
1072 The tools are sorted by the number of switches they had in nmh.
1073 Visible and hidden switches were counted,
1074 but not the generic help and version switches.
1075 Whereas in the beginning of the project, the average tool had 11 switches,
1076 now it has no more than 5 \(en only half as many.
1077 If the `no' switches and similar inverse variant are folded onto
1078 their counter-parts, the average tool had 8 switches in pre-mmh times and
1079 has 4 now.
1080 The total number of functional switches in mmh dropped from 465
1081 to 234.
1083 .KS
1084 .in 1c
1085 .so input/switches.grap
1086 .KE
1088 .P
1089 A part of the switches vanished after functions were removed.
1090 This was the case for network mail transfer, for instance.
1091 Sometimes, however, the work flow was the other way:
1092 I looked through the
1093 .Mp mh-chart (7)
1094 man page to identify the tools with apparently too many switches.
1095 Then considering the value of each of the switches by examining
1096 the tool's man page and source code, aided by recherche and testing.
1097 This way, the removal of functions was suggested by the aim to reduce
1098 the number of switches per command.
1101 .U3 "Draft Folder Facility
1102 .P
1103 A change early in the project was the complete transition from
1104 the single draft message to the draft folder facility.
1105 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860
1106 .\" XXX ref to section ...
1107 The draft folder facility was introduced in the mid-eighties, when
1108 Rose and Romine called it a ``relatively new feature''.
1109 .[
1110 rose romine real work
1111 .]
1112 Since then, the facility had existed but was inactive by default.
1113 The default activation and the related rework of the tools made it
1114 possible to remove the
1115 .Sw -[no]draftfolder ,
1116 and
1117 .Sw -draftmessage
1118 switches from
1119 .Pn comp ,
1120 .Pn repl ,
1121 .Pn forw ,
1122 .Pn dist ,
1123 .Pn whatnow ,
1124 and
1125 .Pn send .
1126 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860
1127 The only flexibility removed with this change is having multiple
1128 draft folders within one profile.
1129 I consider this a theoretical problem only.
1130 At the same time, the
1131 .Sw -draft
1132 switch of
1133 .Pn anno ,
1134 .Pn refile ,
1135 and
1136 .Pn send
1137 was removed.
1138 The special treatment of \fIthe\fP draft message became irrelevant after
1139 the rework of the draft system.
1140 (cf. Sec.
1141 .Cf draft-folder )
1142 Furthermore,
1143 .Pn comp
1144 no longer needs a
1145 .Sw -file
1146 switch as the draft folder facility together with the
1147 .Sw -form
1148 switch are sufficient.
1151 .U3 "In Place Editing
1152 .P
1153 .Pn anno
1154 had the switches
1155 .Sw -[no]inplace
1156 to either annotate the message in place and thus preserve hard links,
1157 or annotate a copy to replace the original message, breaking hard links.
1158 Following the assumption that linked messages should truly be the
1159 same message, and annotating it should not break the link, the
1160 .Sw -[no]inplace
1161 switches were removed and the previous default
1162 .Sw -inplace
1163 was made the only behavior.
1164 .Ci c8195849d2e366c569271abb0f5f60f4ebf0b4d0
1165 The
1166 .Sw -[no]inplace
1167 switches of
1168 .Pn repl ,
1169 .Pn forw ,
1170 and
1171 .Pn dist
1172 could be removed, too, as they were simply passed through to
1173 .Pn anno .
1174 .P
1175 .Pn burst
1176 also had
1177 .Sw -[no]inplace
1178 switches, but with different meaning.
1179 With
1180 .Sw -inplace ,
1181 the digest had been replaced by the table of contents (i.e. the
1182 introduction text) and the burst messages were placed right
1183 after this message, renumbering all following messages.
1184 Also, any trailing text of the digest was lost, though,
1185 in practice, it usually consists of an end-of-digest marker only.
1186 Nontheless, this behavior appeared less elegant than the
1187 .Sw -noinplace
1188 behavior, which already had been the default.
1189 Nmh's
1190 .Mp burst (1)
1191 man page reads:
1192 .QS
1193 If
1194 .Sw -noinplace
1195 is given, each digest is preserved, no table
1196 of contents is produced, and the messages contained within
1197 the digest are placed at the end of the folder. Other messages
1198 are not tampered with in any way.
1199 .QE
1200 .LP
1201 The decision to drop the
1202 .Sw -inplace
1203 behavior was supported by the code complexity and the possible data loss
1204 it caused.
1205 .Sw -noinplace
1206 was chosen to be the definitive behavior.
1207 .Ci 68a686adeb39223a5e1ad35e4a24890ec053679d
1210 .U3 "Forms and Format Strings
1211 .P
1212 Historically, the tools that had
1213 .Sw -form
1214 switches to supply a form file had
1215 .Sw -format
1216 switches as well to supply the contents of a form file as a string
1217 on the command line directly.
1218 In consequence, the following two lines equaled:
1219 .VS
1220 scan -form scan.mailx
1221 scan -format "`cat .../scan.mailx`"
1222 VE
1223 The
1224 .Sw -format
1225 switches were dropped in favor for extending the
1226 .Sw -form
1227 switches.
1228 .Ci f51956be123db66b00138f80464d06f030dbb88d
1229 If their argument starts with an equal sign (`='),
1230 then the rest of the argument is taken as a format string,
1231 otherwise the arguments is treated as the name of a format file.
1232 Thus, now the following two lines equal:
1233 .VS
1234 scan -form scan.mailx
1235 scan -form "=`cat .../scan.mailx`"
1236 VE
1237 This rework removed the prefix collision between
1238 .Sw -form
1239 and
1240 .Sw -format .
1241 Now, typing
1242 .Sw -fo
1243 suffices to specify form or format string.
1244 .P
1245 The different meaning of
1246 .Sw -format
1247 for
1248 .Pn repl
1249 and
1250 .Pn forw
1251 was removed in mmh.
1252 .Pn forw
1253 was completely switched to MIME-type forwarding, thus removing the
1254 .Sw -[no]format .
1255 .Ci 6e271608b7b9c23771523f88d23a4d3593010cf1
1256 For
1257 .Pn repl ,
1258 the
1259 .Sw -[no]format
1260 switches were reworked to
1261 .Sw -[no]filter
1262 switches.
1263 .Ci 67411b1f95d6ec987b4c732459e1ba8a8ac192c6
1264 The
1265 .Sw -format
1266 switches of
1267 .Pn send
1268 and
1269 .Pn post ,
1270 which had a third meaning,
1271 were removed likewise.
1272 .Ci f3cb7cde0e6f10451b6848678d95860d512224b9
1273 Eventually, the ambiguity of the
1274 .Sw -format
1275 switches was resolved by not anymore having any such switch in mmh.
1278 .U3 "MIME Tools
1279 .P
1280 The MIME tools, which were once part of
1281 .Pn mhn
1282 .\" XXX
1283 (whatever that stood for),
1284 had several switches that added little practical value to the programs.
1285 The
1286 .Sw -[no]realsize
1287 switches of
1288 .Pn mhbuild
1289 and
1290 .Pn mhlist
1291 were removed, doing real size calculations always now
1292 .Ci 8d8f1c3abc586c005c904e52c4adbfe694d2201c ,
1293 as nmh's
1294 .Mp mhbuild (1)
1295 man page states
1296 ``This provides an accurate count at the expense of a small delay.''
1297 This small delay is not noticable on modern systems.
1298 .P
1299 The
1300 .Sw -[no]check
1301 switches were removed together with the support for
1302 .Hd Content-MD5
1303 header fields.
1304 .[
1305 rfc 1864
1306 .]
1307 .Ci 31dc797eb5178970d68962ca8939da3fd9a8efda
1308 (cf. Sec.
1309 .Cf content-md5 )
1310 .P
1311 The
1312 .Sw -[no]ebcdicsafe
1313 and
1314 .Sw -[no]rfc934mode
1315 switches of
1316 .Pn mhbuild
1317 were removed because they are considered obsolete.
1318 .Ci 01a3480928da485b4d6109d36d751dfa71799d58
1319 .Ci 3363e2624dce0eb8164cf8b3f1ab385c8ff72e88
1320 .P
1321 Content caching of external MIME parts, activated with the
1322 .Sw -rcache
1323 and
1324 .Sw -wcache
1325 switches was completely removed.
1326 .Ci d1fefd9f614e4dc3cda16da6c69133c1b2005269
1327 External MIME parts are rare today, having a caching facility
1328 for them appears to be unnecessary.
1329 .P
1330 In pre-MIME times,
1331 .Pn mhl
1332 had covered many tasks that are part of MIME handling today.
1333 Therefore,
1334 .Pn mhl
1335 could be simplified to a large extend, reducing the number of its
1336 switches from 21 to 6.
1337 .Ci 350ad6d3542a07639213cf2a4fe524e829c1e7b6
1338 .Ci 0e46503be3c855bddaeae3843e1b659279c35d70
1343 .U3 "Header Printing
1344 .P
1345 .Pn folder 's
1346 data output is self-explaining enough that
1347 displaying the header line makes little sense.
1348 Hence, the
1349 .Sw -[no]header
1350 switch was removed and headers are never printed.
1351 .Ci 601cc73d1fa05ce96faa728f036d6c51b91701c7
1352 .P
1353 In
1354 .Pn mhlist ,
1355 the
1356 .Sw -[no]header
1357 switches were removed, too.
1358 .Ci b24f96523aaf60e44e04a3ffb1d22e69a13a602f
1359 But in this case headers are always printed,
1360 because the output is not self-explaining.
1361 .P
1362 .Pn scan
1363 also had
1364 .Sw -[no]header
1365 switches.
1366 Printing the header had been sensible until the introduction of
1367 format strings made it impossible to display the column headings.
1368 Only the folder name and the current date remained to be printed.
1369 As this information can be perfectly retrieved by
1370 .Pn folder
1371 and
1372 .Pn date ,
1373 consequently, the switches were removed.
1374 .Ci c477dc5d1d03fa6d9a8ab3dd3508c63cbddc044e
1375 .P
1376 By removing all
1377 .Sw -header
1378 switches, the collision with
1379 .Sw -help
1380 on the first two letters was resolved.
1381 Currently,
1382 .Sw -h
1383 evaluates to
1384 .Sw -help
1385 for all tools of mmh.
1388 .U3 "Suppressing Edits or the Invocation of the WhatNow Shell
1389 .P
1390 The
1391 .Sw -noedit
1392 switch of
1393 .Pn comp ,
1394 .Pn repl ,
1395 .Pn forw ,
1396 .Pn dist ,
1397 and
1398 .Pn whatnow
1399 was removed, but it can now be replaced by specifying
1400 .Sw -editor
1401 with an empty argument.
1402 .Ci 75fca31a5b9d5c1a99c74ab14c94438d8852fba9
1403 (Specifying
1404 .Cl "-editor /bin/true
1405 is nearly the same, only differing by the previous editor being set.)
1406 .P
1407 The more important change is the removal of the
1408 .Sw -nowhatnowproc
1409 switch.
1410 .Ci ee4f43cf2ef0084ec698e4e87159a94c01940622
1411 This switch had introduced an awkward behavior, as explained in nmh's
1412 man page for
1413 .Mp comp (1):
1414 .QS
1415 The
1416 .Sw -editor
1417 .Ar editor
1418 switch indicates the editor to use for
1419 the initial edit. Upon exiting from the editor,
1420 .Pn comp
1421 will invoke the
1422 .Pn whatnow
1423 program. See
1424 .Mp whatnow (1)
1425 for a discussion of available options.
1426 The invocation of this program can be
1427 inhibited by using the
1428 .Sw -nowhatnowproc
1429 switch. (In truth of fact, it is the
1430 .Pn whatnow
1431 program which starts the initial edit.
1432 Hence,
1433 .Sw -nowhatnowproc
1434 will prevent any edit from occurring.)
1435 .QE
1436 .P
1437 Effectively, the
1438 .Sw -nowhatnowproc
1439 switch creates only a draft message.
1440 As
1441 .Cl "-whatnowproc /bin/true
1442 causes the same behavior, the
1443 .Sw -nowhatnowproc
1444 switch was removed for being redundant.
1445 Likely, the
1446 .Sw -nowhatnowproc
1447 switch was intended to be used by front-ends.
1451 .U3 "Various
1452 .BU
1453 With the removal of MMDF maildrop format support,
1454 .Pn packf
1455 and
1456 .Pn rcvpack
1457 no longer needed their
1458 .Sw -mbox
1459 and
1460 .Sw -mmdf
1461 switches.
1462 .Sw -mbox
1463 is the sole behavior now.
1464 .Ci 3916ab66ad5d183705ac12357621ea8661afd3c0
1465 In the same go,
1466 .Pn packf
1467 and
1468 .Pn rcvpack
1469 were reworked and their
1470 .Sw -file
1471 switch became unnecessary.
1472 .Ci ca1023716d4c2ab890696f3e41fa0d94267a940e
1474 .BU
1475 Mmh's tools will no longer clear the screen (\c
1476 .Pn scan 's
1477 and
1478 .Pn mhl 's
1479 .Sw -[no]clear
1480 switches
1481 .Ci e57b17343dcb3ff373ef4dd089fbe778f0c7c270
1482 .Ci 943765e7ac5693ae177fd8d2b5a2440e53ce816e ).
1483 Neither will
1484 .Pn mhl
1485 ring the bell (\c
1486 .Sw -[no]bell
1487 .Ci e11983f44e59d8de236affa5b0d0d3067c192e24 )
1488 nor page the output itself (\c
1489 .Sw -length
1490 .Ci 5b9d883db0318ed2b84bb82dee880d7381f99188 ).
1491 .\" XXX Ref
1492 Generally, the pager to use is no longer specified with the
1493 .Sw -[no]moreproc
1494 command line switches for
1495 .Pn mhl
1496 and
1497 .Pn show /\c
1498 .Pn mhshow .
1499 .Ci 39e87a75b5c2d3572ec72e717720b44af291e88a
1501 .BU
1502 In order to avoid prefix collisions among switch names, the
1503 .Sw -version
1504 switch was renamed to
1505 .Sw -Version
1506 (with capital `V').
1507 .Ci 32b2354dbaf4bf934936eb5b102a4a3d2fdd209a
1508 Every program has the
1509 .Sw -version
1510 switch but its first three letters collided with the
1511 .Sw -verbose
1512 switch, present in many programs.
1513 The rename solved this problem once for all.
1514 Although this rename breaks a basic interface, having the
1515 .Sw -V
1516 abbreviation to display the version information, isn't all too bad.
1518 .BU
1519 .Sw -[no]preserve
1520 of
1521 .Pn refile
1522 was removed
1523 .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
1524 because what use was it anyway?
1525 Quoting nmh's man page
1526 .Mp refile (1):
1527 .QS
1528 Normally when a message is refiled, for each destination
1529 folder it is assigned the number which is one above the current
1530 highest message number in that folder. Use of the
1531 .Sw -preserv
1532 [sic!] switch will override this message renaming, and try
1533 to preserve the number of the message. If a conflict for a
1534 particular folder occurs when using the
1535 .Sw -preserve
1536 switch, then
1537 .Pn refile
1538 will use the next available message number which
1539 is above the message number you wish to preserve.
1540 .QE
1542 .BU
1543 The removal of the
1544 .Sw -[no]reverse
1545 switches of
1546 .Pn scan
1547 .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
1548 is a bug fix, supported by the comments
1549 ``\-[no]reverse under #ifdef BERK (I really HATE this)''
1550 by Rose and
1551 ``Lists messages in reverse order with the `\-reverse' switch.
1552 This should be considered a bug.'' by Romine in the documentation.
1553 .\" XXX Ref: welche datei genau.
1554 The question remains why neither Rose and Romine had fixed this
1555 bug in the eighties when they wrote these comments nor has anyone
1556 thereafter.
1559 .ig
1561 forw: [no]dashstuffing(mhl)
1563 mhshow: [no]pause [no]serialonly
1565 mhmail: resent queued
1566 inc: snoop, (pop)
1568 mhl: [no]faceproc folder sleep
1569 [no]dashstuffing(forw) digest list volume number issue number
1571 prompter: [no]doteof
1573 refile: [no]preserve [no]unlink [no]rmmproc
1575 send: [no]forward [no]mime [no]msgid
1576 [no]push split [no]unique (sasl) width snoop [no]dashstuffing
1577 attach attachformat
1578 whatnow: (noedit) attach
1580 slocal: [no]suppressdups
1582 spost: [no]filter [no]backup width [no]push idanno
1583 [no]check(whom) whom(whom)
1585 whom: ???
1587 ..
1590 .ig
1592 .P
1593 In the best case, all switches are unambiguous on the first character,
1594 or on the three-letter prefix for the `no' variants.
1595 Reducing switch prefix collisions, shortens the necessary prefix length
1596 the user must type.
1597 Having less switches helps best.
1599 ..
1602 .\" XXX: whatnow prompt commands
1607 .\" --------------------------------------------------------------
1608 .H1 "Modernizing
1609 .P
1610 In the more than thirty years of MH's existence, its code base was
1611 increasingly extended.
1612 New features entered the project and became alternatives to the
1613 existing behavior.
1614 Relicts from several decades have gathered in the code base,
1615 but seldom obsolete features were dropped.
1616 This section describes the removing of old code
1617 and the modernizing of the default setup.
1618 It focuses on the functional aspect only;
1619 the non-functional aspects of code style are discussed in Sec.
1620 .Cf code-style .
1623 .H2 "Code Relicts
1624 .P
1625 My position regarding the removal of obsolete functions of mmh,
1626 .\" XXX ``in order to remove old code,''
1627 is much more revolutional than the nmh community appreciates.
1628 Working on an experimental version, I was quickly able to drop
1629 functionality I considered ancient.
1630 The need for consensus with peers would have slowed this process down.
1631 Without the need to justify my decisions, I was able to rush forward.
1632 In December 2011, Paul Vixie motivated the nmh developers to just
1633 .\" XXX ugs
1634 do the work:
1635 .[
1636 paul vixie edginess nmh-workers
1637 .]
1638 .QS
1639 let's stop walking on egg shells with this code base. there's no need to
1640 discuss whether to keep using vfork, just note in [sic!] passing, [...]
1641 we don't need a separate branch for removing vmh
1642 or ridding ourselves of #ifdef's or removing posix replacement functions
1643 or depending on pure ansi/posix ``libc''.
1644 .QP
1645 these things should each be a day or two of work and the ``main branch''
1646 should just be modern. [...]
1647 let's push forward, aggressively.
1648 .QE
1649 .LP
1650 I did so already in the months before.
1651 I pushed forward.
1652 .\" XXX semicolon ?
1653 I simply dropped the cruft.
1654 .P
1655 The decision to drop a feature was based on literature research and
1656 careful thinking, but whether having had contact with this particular
1657 feature within my own computer life served as a rule of thumb.
1658 I explained my reasons in the commit messages
1659 in the version control system.
1660 Hence, others can comprehend my view and argue for undoing the change
1661 if I have missed an important aspect.
1662 I was quick in dropping parts.
1663 I rather re-included falsely dropped parts than going at a slower pace.
1664 Mmh is experimental work; it required tough decisions.
1665 .\" XXX ``exp. work'' schon oft gesagt
1668 .U3 "Forking
1669 .P
1670 Being a tool chest, MH creates many processes.
1671 In earlier times
1672 .Fu fork()
1673 had been an expensive system call, because the process's image needed
1674 to be completely duplicated at once.
1675 This expensive work was especially unnecessary in the commonly occuring
1676 case wherein the image is replaced by a call to
1677 .Fu exec()
1678 right after having forked the child process.
1679 The
1680 .Fu vfork()
1681 system call was invented to speed up this particular case.
1682 It completely omits the duplication of the image.
1683 On old systems this resulted in significant speed ups.
1684 Therefore MH used
1685 .Fu vfork()
1686 whenever possible.
1687 .P
1688 Modern memory management units support copy-on-write semantics, which make
1689 .Fu fork()
1690 almost as fast as
1691 .Fu vfork() .
1692 The man page of
1693 .Mp vfork (2)
1694 in FreeBSD 8.0 states:
1695 .QS
1696 This system call will be eliminated when proper system sharing mechanisms
1697 are implemented. Users should not depend on the memory sharing semantics
1698 of vfork() as it will, in that case, be made synonymous to fork(2).
1699 .QE
1700 .LP
1701 Vixie supports the removal with the note that ``the last
1702 system on which fork was so slow that an mh user would notice it, was
1703 Eunice. that was 1987''.
1704 .[
1705 nmh-workers vixie edginess
1706 .]
1707 I replaced all calls to
1708 .Fu vfork()
1709 with calls to
1710 .Fu fork() .
1711 .Ci 40821f5c1316e9205a08375e7075909cc9968e7d
1712 .P
1713 Related to the costs of
1714 .Fu fork()
1715 is the probability of its success.
1716 In the eighties, on heavy loaded systems, calls to
1717 .Fu fork()
1718 were prone to failure.
1719 Hence, many of the
1720 .Fu fork()
1721 calls in the code were wrapped into loops to retry the
1722 .Fu fork()
1723 several times, to increase the chances to succeed, eventually.
1724 On modern systems, a failing
1725 .Fu fork()
1726 call is unusual.
1727 Hence, in the rare case when
1728 .Fu fork()
1729 fails, mmh programs simply abort.
1730 .Ci 5fbf37ee68e018998ada61eeab73e035b26834b6
1733 .U3 "Header Fields
1734 .BU
1735 The
1736 .Hd Encrypted
1737 header field was introduced by RFC\|822,
1738 but already marked as legacy in RFC\|2822.
1739 Today, OpenPGP provides the basis for standardized exchange of encrypted
1740 messages [RFC\|4880, RFC\|3156].
1741 Hence, the support for
1742 .Hd Encrypted
1743 header fields is removed in mmh.
1744 .Ci 064527f7b57ab050e5af13e15ad99aeeab125857
1745 .BU
1746 The native support for
1747 .Hd Face
1748 header fields has been removed, as well.
1749 .Ci 8e5be81f784682822f5e868c1bf3c8624682bd23
1750 This feature is similar to the
1751 .Hd X-Face
1752 header field in its intent,
1753 but takes a different approach to store the image.
1754 Instead of encoding the image data directly into the header field,
1755 it contains the hostname and UDP port where the image
1756 date can be retrieved.
1757 There is even a third Face system,
1758 which is the successor of
1759 .Hd X-Face ,
1760 although it re-uses the
1761 .Hd Face
1762 header field.
1763 It was invented in 2005 and supports colored PNG images.
1764 None of the Face systems described here is popular today.
1765 Hence, mmh has no direct support for them.
1766 .BU
1767 .Id content-md5
1768 The
1769 .Hd Content-MD5
1770 header field was introduced by RFC\|1864.
1771 It provides detection of data corruption during the transfer.
1772 But it can not ensure verbatim end-to-end delivery of the contents
1773 [RFC\|1864].
1774 The proper approach to verify content integrity in an
1775 end-to-end relationship is the use of digital signatures.
1776 .\" XXX (RFCs FIXME).
1777 On the other hand, transfer protocols should detect corruption during
1778 the transmission.
1779 The TCP includes a checksum field therefore.
1780 These two approaches in combinations render the
1781 .Hd Content-MD5
1782 header field superfluous.
1783 Not a single one out of 4\|200 messages from two decades
1784 in an nmh-workers mailing list archive contains a
1785 .Hd Content-MD5
1786 header field.
1787 Neither did any of the 60\|000 messages in my personal mail storage.
1788 Removing the support for this header field,
1789 removed the last place where MD5 computation was needed.
1790 .Ci 31dc797eb5178970d68962ca8939da3fd9a8efda
1791 Hence, the MD5 code could be removed as well.
1792 Over 500 lines of code vanished by this one change.
1795 .U3 "MMDF maildrop support
1796 .P
1797 This type of format is conceptionally similar to the mbox format,
1798 but uses a different message delimiter (`\fL\\1\\1\\1\\1\fP',
1799 commonly written as `\fL^A^A^A^A\fP', instead of `\fLFrom\0\fP').
1800 Mbox is the de-facto standard maildrop format on Unix,
1801 whereas the MMDF maildrop format is now forgotten.
1802 By dropping the MMDF maildrop format support,
1803 mbox became the only packed mailbox format supported in mmh.
1804 .P
1805 The simplifications within the code were moderate.
1806 Mainly, the reading and writing of MMDF mailbox files was removed.
1807 But also, switches of
1808 .Pn packf
1809 and
1810 .Pn rcvpack
1811 could be removed.
1812 .Ci 3916ab66ad5d183705ac12357621ea8661afd3c0
1813 In the message parsing function
1814 .Fn sbr/m_getfld.c ,
1815 knowledge of MMDF packed mail boxes was removed.
1816 .Ci 684ec30d81e1223a282764452f4902ed4ad1c754
1817 Further code structure simplifications may be possible there,
1818 because only one single packed mailbox format is left to be supported.
1819 I have not worked on them yet because
1820 .Fu m_getfld()
1821 is heavily optimized and thus dangerous to touch.
1822 The risk of damaging the intricate workings of the optimized code is
1823 too high.
1826 .U3 "Prompter's Control Keys
1827 .P
1828 The program
1829 .Pn prompter
1830 queries the user to fill in a message form.
1831 When used by
1832 .Pn comp
1833 as
1834 .Cl "comp -editor prompter" ,
1835 the resulting behavior is similar to
1836 .Pn mailx .
1837 Apparently,
1838 .Pn prompter
1839 hadn't been touched lately.
1840 Otherwise it's hardly explainable why it
1841 still offered the switches
1842 .Sw -erase
1843 .Ar chr
1844 and
1845 .Sw -kill
1846 .Ar chr
1847 to name the characters for command line editing.
1848 The times when this had been necessary are long time gone.
1849 Today these things work out-of-the-box, and if not, are configured
1850 with the standard tool
1851 .Pn stty .
1852 The switches are removed now
1853 .Ci 0bd9750710cdbab80cfb4036dd87af20afe1552f .
1856 .U3 "Hardcopy Terminal Support
1857 .P
1858 More of a funny anecdote is a check for being connected to a
1859 hardcopy terminal.
1860 It remained in the code until spring 2012, when I finally removed it
1861 .Ci b7764c4a6b71d37918a97594d866258f154017ca .
1862 .P
1863 The check only prevented a pager to be placed between the printing
1864 program (\c
1865 .Pn mhl )
1866 and the terminal.
1867 In nmh, this could have been ensured statically with the
1868 .Sw -nomoreproc
1869 at the command line, too.
1870 In mmh, setting the profile entry
1871 .Pe Pager
1872 or the environment variable
1873 .Ev PAGER
1874 to
1875 .Pn cat
1876 is sufficient.
1881 .H2 "Attachments
1882 .P
1883 The mind model of email attachments is unrelated to MIME.
1884 Although the MIME RFCs (2045 through 2049) define the technical
1885 requirements for having attachments, they do not mention the word
1886 ``attachment''.
1887 Instead of attachments, MIME talks about ``multi-part message bodies''
1888 [RFC\|2045], a more general concept.
1889 Multi-part messages are messages
1890 ``in which one or more different
1891 sets of data are combined in a single body''
1892 [RFC\|2046].
1893 MIME keeps its descriptions generic;
1894 it does not imply specific usage models.
1895 One usage model became prevalent: attachments.
1896 The idea is having a main text document with files of arbitrary kind
1897 attached to it.
1898 In MIME terms, this is a multi-part message having a text part first
1899 and parts of arbitrary type following.
1900 .P
1901 MH's MIME support is a direct implementation of the RFCs.
1902 The perception of the topic described in the RFCs is clearly visible
1903 in MH's implementation.
1904 .\" XXX rewrite ``no idea''.
1905 As a result,
1906 MH had all the MIME features but no idea of attachments.
1907 But users don't need all the MIME features,
1908 they want convenient attachment handling.
1911 .U3 "Composing MIME Messages
1912 .P
1913 In order to improve the situation on the message composing side,
1914 Jon Steinhart had added an attachment system to nmh in 2002.
1915 .Ci 7480dbc14bc90f2d872d434205c0784704213252
1916 In the file
1917 .Fn docs/README-ATTACHMENTS ,
1918 he described his motivation to do so as such:
1919 .QS
1920 Although nmh contains the necessary functionality for MIME message
1921 handing [sic!], the interface to this functionality is pretty obtuse.
1922 There's no way that I'm ever going to convince my partner to write
1923 .Pn mhbuild
1924 composition files!
1925 .QE
1926 .LP
1927 With this change, the mind model of attachments entered nmh.
1928 In the same document:
1929 .QS
1930 These changes simplify the task of managing attachments on draft files.
1931 They allow attachments to be added, listed, and deleted.
1932 MIME messages are automatically created when drafts with attachments
1933 are sent.
1934 .QE
1935 .LP
1936 Unfortunately, the attachment system,
1937 like any new facilities in nmh,
1938 was inactive by default.
1939 .P
1940 During my work in Argentina, I tried to improve the attachment system.
1941 But, because of great opposition in the nmh community,
1942 my patch died as a proposal on the mailing list, after long discussions.
1943 .[
1944 nmh-workers attachment proposal
1945 .]
1946 In January 2012, I extended the patch and applied it to mmh.
1947 .Ci 8ff284ff9167eff8f5349481529332d59ed913b1
1948 In mmh, the attachment system is active by default.
1949 Instead of command line switches, the
1950 .Pe Attachment-Header
1951 profile entry is used to specify
1952 the name of the attachment header field.
1953 It is pre-defined to
1954 .Hd Attach .
1955 .P
1956 To add an attachment to a draft, a header line needs to be added:
1957 .VS
1958 To: bob
1959 Subject: The file you wanted
1960 Attach: /path/to/the/file-bob-wanted
1961 --------
1962 Here it is.
1963 VE
1964 The header field can be added to the draft manually in the editor,
1965 or by using the `attach' command at the WhatNow prompt, or
1966 non-interactively with
1967 .Pn anno :
1968 .VS
1969 anno -append -nodate -component Attach -text /path/to/attachment
1970 VE
1971 Drafts with attachment headers are converted to MIME automatically by
1972 .Pn send .
1973 The conversion to MIME is invisible to the user.
1974 The draft stored in the draft folder is always in source form with
1975 attachment headers.
1976 If the MIMEification fails, for instance because the file to attach
1977 is not accessible, the original draft is not changed.
1978 .P
1979 The attachment system handles the forwarding of messages, too.
1980 If the attachment header value starts with a plus character (`+'),
1981 like in
1982 .Cl "Attach: +bob 30 42" ,
1983 the given messages in the specified folder will be attached.
1984 This allowed to simplify
1985 .Pn forw .
1986 .Ci f41f04cf4ceca7355232cf7413e59afafccc9550
1987 .P
1988 Closely related to attachments is non-ASCII text content,
1989 because it requires MIME too.
1990 In nmh, the user needed to call `mime' at the WhatNow prompt
1991 to have the draft converted to MIME.
1992 This was necessary whenever the draft contained non-ASCII characters.
1993 If the user did not call `mime', a broken message would be sent.
1994 Therefore, the
1995 .Pe automimeproc
1996 profile entry could be specified to have the `mime' command invoked
1997 automatically each time.
1998 Unfortunately, this approach conflicted with attachment system
1999 because the draft would already be in MIME format at the time
2000 when the attachment system wanted to MIMEify it.
2001 To use nmh's attachment system, `mime' must not be called at the
2002 WhatNow prompt and
2003 .Pe automimeproc
2004 must not be set in the profile.
2005 But then the case of non-ASCII text without attachment headers was
2006 not caught.
2007 All in all, the solution was complex and irritating.
2008 My patch from December 2010
2009 .[
2010 nmh-workers attachment proposal
2011 .]
2012 would have simplified the situation.
2013 .P
2014 Mmh's current solution is even more elaborate.
2015 Any necessary MIMEification is done automatically.
2016 There is no `mime' command at the WhatNow prompt anymore.
2017 The draft will be converted automatically to MIME when either an
2018 attachment header or non-ASCII text is present.
2019 Furthermore, the hash character (`#') is not special any more
2020 at line beginnings in the draft message.
2021 .\" XXX REF ?
2022 Users need not concern themselves with the whole topic at all.
2023 .P
2024 Although the new approach does not anymore support arbitrary MIME
2025 compositions directly, the full power of
2026 .Pn mhbuild
2027 can still be accessed.
2028 Given no attachment headers are included, the user can create
2029 .Pn mhbuild
2030 composition drafts like in nmh.
2031 Then, at the WhatNow prompt, he needs to invoke
2032 .Cl "edit mhbuild
2033 to convert it to MIME.
2034 Because the resulting draft does neither contain non-ASCII characters
2035 nor has it attachment headers, the attachment system will not touch it.
2036 .P
2037 The approach taken in mmh is tailored towards today's most common case:
2038 a text part, possibly with attachments.
2039 This case was simplified.
2042 .U3 "MIME Type Guessing
2043 .P
2044 From the programmer's point of view, the use of
2045 .Pn mhbuild
2046 composition drafts had one notable advantage over attachment headers:
2047 The user provides the appropriate MIME types for files to include.
2048 The attachment system needs to find out the correct MIME type itself.
2049 This is a difficult task, yet it spares the user irritating work.
2050 Determining the correct MIME type of content is partly mechanical,
2051 partly intelligent work.
2052 Forcing the user to find out the correct MIME type,
2053 forces him to do partly mechanical work.
2054 Letting the computer do the work, can lead to bad choices for difficult
2055 content.
2056 For mmh, the latter option was chosen.
2057 .P
2058 Determining the MIME type by the suffix of the file name is a dumb
2059 approach, yet it is simple to implement and provides good results
2060 for the common cases.
2061 Mmh implements this approach in the
2062 .Pn print-mimetype
2063 script.
2064 .Ci 4b5944268ea0da7bb30598a27857304758ea9b44
2065 Using it is the default choice.
2066 .P
2067 A far better, though less portable, approach is the use of
2068 .Pn file .
2069 This standard tool tries to determine the type of files.
2070 Unfortunately, its capabilities and accuracy varies from system to system.
2071 Additionally, its output was only intended for human beings,
2072 but not to be used by programs.
2073 It varies much.
2074 Nevertheless, modern versions of GNU
2075 .Pn file ,
2076 which is prevalent on the popular GNU/Linux systems,
2077 provide MIME type output in machine-readable form.
2078 Although this solution is highly system-dependent,
2079 it solves the difficult problem well.
2080 On systems where GNU
2081 .Pn file ,
2082 version 5.04 or higher, is available it should be used.
2083 One needs to specify the following profile entry to do so:
2084 .Ci 3baec236a39c5c89a9bda8dbd988d643a21decc6
2085 .VS
2086 Mime-Type-Query: file -b --mime
2087 VE
2088 .LP
2089 Other versions of
2090 .Pn file
2091 might possibly be usable with wrapper scripts to reformat the output.
2092 The diversity among
2093 .Pn file
2094 implementations is great; one needs to check the local variant.
2095 .P
2096 If no MIME type can be determined, text content gets sent as
2097 `text/plain' and anything else under the generic fall-back type
2098 `application/octet-stream'.
2099 It is not possible in mmh to override the automatic MIME type guessing
2100 for a specific file.
2101 To do so, either the user would need to know in advance for which file
2102 the automatic guessing fails, or the system would require interaction.
2103 I consider both cases impractical.
2104 The existing solution should be sufficient.
2105 If not, the user may always fall back to
2106 .Pn mhbuild
2107 composition drafts and ignore the attachment system.
2110 .U3 "Storing Attachments
2111 .P
2112 Extracting MIME parts of a message and storing them to disk is performed by
2113 .Pn mhstore .
2114 The program has two operation modes,
2115 .Sw -auto
2116 and
2117 .Sw -noauto .
2118 With the former one, each part is stored under the filename given in the
2119 MIME part's meta information, if available.
2120 This naming information is usually available for modern attachments.
2121 If no filename is available, this MIME part is stored as if
2122 .Sw -noauto
2123 would have been specified.
2124 In the
2125 .Sw -noauto
2126 mode, the parts are processed according to rules, defined by
2127 .Pe mhstore-store-*
2128 profile entries.
2129 These rules define generic filename templates for storing
2130 or commands to post-process the contents in arbitrary ways.
2131 If no matching rule is available the part is stored under a generic
2132 filename, built from message number, MIME part number, and MIME type.
2133 .P
2134 The
2135 .Sw -noauto
2136 mode had been the default in nmh because it was considered safe,
2137 in contrast to the
2138 .Sw -auto
2139 mode.
2140 In mmh,
2141 .Sw -auto
2142 is not dangerous anymore.
2143 Two changes were necessary:
2144 .BU
2145 Any directory path is removed from the proposed filename.
2146 Thus, the files are always stored in the expected directory.
2147 .Ci 41b6eadbcecf63c9a66aa5e582011987494abefb
2148 .BU
2149 Tar files are not extracted automatically any more.
2150 Thus, the rest of the file system will not be touched.
2151 .Ci 94c80042eae3383c812d9552089953f9846b1bb6
2152 .LP
2153 Now, the outcome of mmh's
2154 .Cl "mhstore -auto
2155 can be foreseen from the output of
2156 .Cl "mhlist -verbose" .
2157 .P
2158 The
2159 .Sw -noauto
2160 mode is seen to be more powerful but less convenient.
2161 On the other hand,
2162 .Sw -auto
2163 is safe now and
2164 storing attachments under their original name is intuitive.
2165 Hence,
2166 .Sw -auto
2167 serves better as the default option.
2168 .Ci 3410b680416c49a7617491af38bc1929855a331d
2169 .P
2170 Files are stored into the directory given by the
2171 .Pe Nmh-Storage
2172 profile entry, if set, or
2173 into the current working directory, otherwise.
2174 Storing to different directories is only possible with
2175 .Pe mhstore-store-*
2176 profile entries.
2177 .P
2178 Still, in both modes, existing files get overwritten silently.
2179 This can be considered a bug.
2180 Yet, each other behavior has its draw-backs, too.
2181 Refusing to replace files requires adding a
2182 .Sw -force
2183 option.
2184 Users will likely need to invoke
2185 .Pn mhstore
2186 a second time with
2187 .Sw -force .
2188 Eventually, only the user can decide in the specific case.
2189 This requires interaction, which I like to avoid if possible.
2190 Appending a unique suffix to the filename is another bad option.
2191 For now, the behavior remains as it is.
2192 .P
2193 In mmh, only MIME parts of type message are special in
2194 .Pn mhstore 's
2195 .Sw -auto
2196 mode.
2197 Instead of storing message/rfc822 parts as files to disk,
2198 they are stored as messages into the current mail folder.
2199 The same applies to message/partial, although the parts are
2200 automatically reassembled beforehand.
2201 MIME parts of type message/external-body are not automatically retrieved
2202 anymore.
2203 Instead, information on how to retrieve them is output.
2204 Not supporting this rare case saved nearly one thousand lines of code.
2205 .Ci 55e1d8c654ee0f7c45b9361ce34617983b454c32
2206 .\" XXX mention somewhere else too: (The profile entry `nmh-access-ftp'
2207 .\" and sbr/ruserpass.c for reading ~/.netrc are gone now.)
2208 `application/octet-stream; type=tar' is not special anymore.
2209 Automatically extracting such MIME parts had been the dangerous part
2210 of the
2211 .Sw -auto
2212 mode.
2213 .Ci 94c80042eae3383c812d9552089953f9846b1bb6
2217 .U3 "Showing MIME Messages
2218 .P
2219 The program
2220 .Pn mhshow
2221 had been written to display MIME messages.
2222 It implemented the conceptional view of the MIME RFCs.
2223 Nmh's
2224 .Pn mhshow
2225 handled each MIME part independently, presenting them separately
2226 to the user.
2227 This does not match today's understanding of email attachments,
2228 where displaying a message is seen to be a single, integrated operation.
2229 Today, email messages are expected to consist of a main text part
2230 plus possibly attachments.
2231 They are not any more seen to be arbitrary MIME hierarchies with
2232 information on how to display the individual parts.
2233 I adjusted
2234 .Pn mhshow 's
2235 behavior to the modern view on the topic.
2236 .P
2237 One should note that this section completely ignores the original
2238 .Pn show
2239 program, because it was not capable to display MIME messages
2240 and is no longer part of mmh.
2241 Although
2242 .Pn mhshow
2243 was renamed to
2244 .Pn show
2245 in mmh, this section uses the name
2246 .Pn mhshow ,
2247 in order to avoid confusion.
2248 .\" XXX ref to other section
2249 .P
2250 In mmh, the basic idea is that
2251 .Pn mhshow
2252 should display a message in one single pager session.
2253 Therefore,
2254 .Pn mhshow
2255 invokes a pager session for all its output,
2256 whenever it prints to a terminal.
2257 .Ci a4197ea6ffc5c1550e8b52d5a654bcaaaee04a4e
2258 In consequence,
2259 .Pn mhl
2260 does no more invoke a pager.
2261 .Ci 0e46503be3c855bddaeae3843e1b659279c35d70
2262 With
2263 .Pn mhshow
2264 replacing the original
2265 .Pn show ,
2266 output from
2267 .Pn mhl
2268 does not go to the terminal directly, but through
2269 .Pn mhshow .
2270 Hence,
2271 .Pn mhl
2272 does not need to invoke a pager.
2273 The one and only job of
2274 .Pn mhl
2275 is to format messages or parts of them.
2276 The only place in mmh, where a pager is invoked is
2277 .Pn mhshow .
2278 .P
2279 .Pe mhshow-show-*
2280 profile entries can be used to display MIME parts in a specific way.
2281 For instance, PDF and Postscript files could be converted to plain text
2282 to display them in the terminal.
2283 In mmh, MIME parts will always be displayed serially.
2284 The request to display the MIME type `multipart/parallel' in parallel
2285 is ignored.
2286 It is simply treated as `multipart/mixed'.
2287 .Ci d0581ba306a7299113a346f9b4c46ce97bc4cef6
2288 This could already be requested with the, now removed,
2289 .Sw -serialonly
2290 switch of
2291 .Pn mhshow .
2292 As MIME parts are always processed exclusively , i.e. serially,
2293 the `%e' escape in
2294 .Pe mhshow-show-*
2295 profile entries became useless and was thus removed.
2296 .Ci a20d405db09b7ccca74d3e8c57550883da49e1ae
2297 .P
2298 In the intended setup, only text content would be displayed.
2299 Non-text content would be converted to text by appropriate
2300 .Pe mhshow-show-*
2301 profile entries before, if possible and wanted.
2302 All output would be displayed in a single pager session.
2303 Other kinds of attachments are ignored.
2304 With
2305 .Pe mhshow-show-*
2306 profile entries for them, they can be displayed serially along
2307 the message.
2308 For parallel display, the attachments need to be stored to disk first.
2309 .P
2310 To display text content in foreign charsets, they need to be converted
2311 to the native charset.
2312 Therefore,
2313 .Pe mhshow-charset-*
2314 profile entries used to be needed.
2315 In mmh, the conversion is performed automatically by piping the
2316 text through the
2317 .Pn iconv
2318 command, if necessary.
2319 .Ci 2433122c20baccb10b70b49c04c6b0497b5b3b60
2320 Custom
2321 .Pe mhshow-show-*
2322 rules for textual content might need a
2323 .Cl "iconv -f %c %f |
2324 prefix to have the text converted to the native charset.
2325 .P
2326 Although the conversion of foreign charsets to the native one
2327 has improved, it is not consistent enough.
2328 Further work needs to be done and
2329 the basic concepts in this field need to be re-thought.
2330 Though, the default setup of mmh displays message in foreign charsets
2331 correctly without the need to configure anything.
2334 .ig
2336 .P
2337 mhshow/mhstore: Removed support for retrieving message/external-body parts.
2338 These tools won't download the contents automatically anymore. Instead,
2339 they print the information needed to get the contents. If someone should
2340 really receive one of those rare message/external-body messages, he can
2341 do the job manually. We save nearly a thousand lines of code. That's worth
2342 it!
2343 (The profile entry `nmh-access-ftp' and sbr/ruserpass.c for reading
2344 ~/.netrc are gone now.)
2345 .Ci 55e1d8c654ee0f7c45b9361ce34617983b454c32
2347 ..
2351 .H2 "Signing and Encrypting
2352 .P
2353 Nmh offers no direct support for digital signatures and message encryption.
2354 This functionality needed to be added through third-party software.
2355 In mmh, the functionality should be included because it
2356 is a part of modern email and likely wanted by users of mmh.
2357 A fresh mmh installation should support signing and encrypting
2358 out-of-the-box.
2359 Therefore, Neil Rickert's
2360 .Pn mhsign
2361 and
2362 .Pn mhpgp
2363 scripts
2364 .[
2365 neil rickert mhsign mhpgp
2366 .]
2367 were included into mmh.
2368 The scripts fit well into the mmh, because they are lightweight and
2369 of style similar to the existing tools.
2370 Additionally, no licensing difficulties appeared,
2371 as they are part of the public domain.
2372 .P
2373 The scripts were written for nmh, hence I needed to adjust them according
2374 to the differences of mmh.
2375 For instance, I removed the use of the backup prefix and dropped support
2376 for old PGP features.
2377 .P
2378 .Pn mhsign
2379 handles the signing and encrypting part.
2380 It comprises about 250 lines of shell code and interfaces between
2381 .Pn gnupg
2382 and
2383 the MH system.
2384 It was meant to be invoked at the WhatNow prompt, but in mmh,
2385 .Pn send
2386 does the job automatically.
2387 Special header fields were introduced to request the action.
2388 If a draft contains the
2389 .Hd Sign
2390 header field,
2391 .Pn send
2392 will sign it.
2393 The key to be used is either chosen automatically or specified by the
2394 .Pe Pgpkey
2395 profile entry.
2396 .Pn send
2397 always signes messages using the PGP/MIME standard, \" REF XXX
2398 but by manually invoking
2399 .Pn mhsign ,
2400 old-style non-MIME signatures can be created as well.
2401 To sign an outgoing message, the draft needs to contain a
2402 .Hd Enc
2403 header field.
2404 Public keys of all recipients are taken from the gnupg keyring or
2405 from an overrides files, called
2406 .Fn pgpkeys .
2407 Unless public keys are found for all recipients,
2408 .Pn send
2409 will refuse to encrypt and send it.
2410 Currently, messages with hidden (BCC) recipients can not be encrypted.
2411 This corner-case requires a more complex solution.
2412 Covering it is left to do.
2413 .P
2414 The integrated message signing and encrypting support is one of the
2415 most recent features in mmh.
2416 Feedback from users and the experience I will gather myself
2417 will direct the further development of the facility.
2418 It is worthwhile to consider adding
2419 .Sw -[no]sign
2420 and
2421 .Sw -[no]enc
2422 switches to
2423 .Pn send ,
2424 to override the corresponding header fields.
2425 The profile entry:
2426 .VS
2427 send: -sign
2428 VE
2429 .LP
2430 would then activate signing of all outgoing messages.
2431 With the present approach, the line
2432 .VS
2433 Send:
2434 VE
2435 .LP
2436 needs to be added to all message forms to achieve the same result.
2437 Yet, the integration of
2438 .Pn mhsign
2439 into mmh is too recent to have enough experience to decide this
2440 question now.
2441 .P
2442 .Pn mhpgp
2443 is the contrary part to
2444 .Pn mhsign .
2445 It verifies signatures and decrypts messages.
2446 .P
2447 FIXME: Add it to mmh first, then write about it here.
2448 .P
2449 The integration of
2450 .Pn mhpgp
2451 into
2452 .Pn show ,
2453 to automatically verify signatures and decrypt messages as needed,
2454 is a task left open.
2455 .Pn show 's
2456 current structure does not allow such an integration on basis of
2457 the existing code.
2458 Extensive programming work is required. ... FIXME
2464 .H2 "Draft and Trash Folder
2465 .P
2467 .U3 "Draft Folder
2468 .Id draft-folder
2469 .P
2470 In the beginning, MH had the concept of a draft message.
2471 This is the file
2472 .Fn draft
2473 in the MH directory, which is treated special.
2474 On composing a message, this draft file was used.
2475 When starting to compose another message before the former one was sent,
2476 the user had to decide among:
2477 .BU
2478 Using the old draft to finish and send it before starting with a new one.
2479 .BU
2480 Discarding the old draft and replacing it with a new one.
2481 .BU
2482 Preserving the old draft by refiling it to a folder.
2483 .P
2484 It was only possible to work in alternation on multiple drafts.
2485 Therefore, the current draft needed to be refiled to a folder and
2486 another one re-used for editing.
2487 Working on multiple drafts at the same time was impossible.
2488 The usual approach of switching to a different MH context did not
2489 help anything.
2490 .P
2491 The draft folder facility exists to
2492 allow true parallel editing of drafts, in a straight forward way.
2493 It was introduced by Marshall T. Rose, already in 1984.
2494 Similar to other new features, the draft folder was inactive by default.
2495 Even in nmh, the highly useful draft folder was not available
2496 out-of-the-box.
2497 At least, Richard Coleman added the man page
2498 .Mp mh-draft (5)
2499 to better document the feature.
2500 .P
2501 Not using the draft folder facility has the single advantage of having
2502 the draft file at a static location.
2503 This is simple in simple cases but the concept does not scale for more
2504 complex cases.
2505 The concept of the draft message is too limited for the problem.
2506 Therefore the draft folder was introduced.
2507 It is the more powerful and more natural concept.
2508 The draft folder is a folder like any other folder in MH.
2509 Its messages can be listed like any other messages.
2510 A draft message is no longer a special case.
2511 Tools do not need special switches to work on the draft message.
2512 Hence corner-cases were removed.
2513 .P
2514 The trivial part of the work was activating the draft folder with a
2515 default name.
2516 I chose the name
2517 .Fn +drafts
2518 for obvious reasons.
2519 In consequence, the command line switches
2520 .Sw -draftfolder
2521 and
2522 .Sw -draftmessage
2523 could be removed.
2524 More difficult but also more improving was updating the tools to the
2525 new concept.
2526 For nearly three decades, the tools needed to support two draft handling
2527 approaches.
2528 By fully switching to the draft folder, the tools could be simplified
2529 by dropping the awkward draft message handling code.
2530 .Sw -draft
2531 switches were removed because operating on a draft message is no longer
2532 special.
2533 It became indistinguishable to operating on any other message.
2534 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860
2535 .P
2536 There is no more need to query the user for draft handling
2537 .Ci 2d48b455c303a807041c35e4248955f8bec59eeb .
2538 It is always possible to add another new draft.
2539 Refiling drafts is without difference to refiling other messages.
2540 All of these special cases are gone.
2541 Yet, one draft-related switch remained.
2542 .Pn comp
2543 still has
2544 .Sw -[no]use
2545 for switching between two modes:
2546 .BU
2547 .Sw -use
2548 to modify an existing draft.
2549 .BU
2550 .Sw -nouse
2551 to compose a new draft, possibly taking some existing message as template.
2552 .P
2553 In either case, the behavior of
2554 .Pn comp
2555 is deterministic.
2556 .P
2557 .Pn send
2558 now operates on the current message in the draft folder by default.
2559 As message and folder can both be overridden by specifying them on
2560 the command line, it is possible to send any message in the mail storage
2561 by simply specifying its number and folder.
2562 In contrast to the other tools,
2563 .Pn send
2564 takes the draft folder as its default folder.
2565 .P
2566 Dropping the draft message concept in favor for the draft folder concept,
2567 removed special cases with regular cases.
2568 This simplified the source code of the tools, as well as the concepts.
2569 In mmh, draft management does not break with the MH concepts
2570 but applies them.
2571 .Cl "scan +drafts" ,
2572 for instance, is a truly natural request.
2573 Most of the work was already performed by Rose in the eighties.
2574 The original improvement of mmh is dropping the old draft message approach
2575 and thus simplifying the tools, the documentation and the system as a whole.
2576 Although my part in the draft handling improvement was small,
2577 it was an important one.
2580 .U3 "Trash Folder
2581 .Id trash-folder
2582 .P
2583 Similar to the situation for drafts is the situation for removed messages.
2584 Historically, a message was ``deleted'' by prepending a specific
2585 \fIbackup prefix\fP, usually the comma character, to the file name.
2586 The specific file would then be ignored by MH because only files with
2587 names consisting of digits only are treated as messages.
2588 Although files remained in the file system,
2589 the messages were no longer visible in MH.
2590 To truly delete them, a maintenance job was needed.
2591 Usually a cron job was installed to delete them after a grace time.
2592 For instance:
2593 .VS
2594 find $HOME/Mail -type f -name ',*' -ctime +7 -delete
2595 VE
2596 In such a setup, the original message could be restored
2597 within the grace time interval by stripping the
2598 backup prefix from the file name.
2599 But the user could not rely on this statement.
2600 If the last message of a folder with six messages (\fL1-6\fP) was removed,
2601 message
2602 .Fn 6 ,
2603 became file
2604 .Fn ,6 .
2605 If then a new message entered the same folder, it would be named with
2606 the number one above the highest existing message number.
2607 In this case the message would be named
2608 .Fn 6
2609 then.
2610 If this new message would be removed as well,
2611 then the backup of the former message is overwritten.
2612 Hence, the ability to restore removed messages did not only depend on
2613 the ``sweeping cron job'' but also on the removing of further messages.
2614 It is undesirable to have such obscure and complex mechanisms.
2615 The user should be given a small set of clear assertions, such as
2616 ``Removed files are restorable within a seven-day grace time.''
2617 With the addition ``... unless a message with the same name in the
2618 same folder is removed before.'' the statement becomes complex.
2619 A user will hardly be able to keep track of any removal to know
2620 if the assertion still holds true for a specific file.
2621 In practice, the real mechanism is unclear to the user.
2622 The consequences of further removals are not obvious.
2623 .P
2624 Further more, the backup files are scattered within the whole mail storage.
2625 This complicates managing them.
2626 It is possible with the help of
2627 .Pn find ,
2628 but everything would be more convenient
2629 if the deleted messages would be collected in one place.
2630 .P
2631 The profile entry
2632 .Pe rmmproc
2633 (previously named
2634 .Pe Delete-Prog )
2635 was introduced very early to improve the situation.
2636 It could be set to any command, which would be executed to remove
2637 the specified messages.
2638 This would override the default action described above.
2639 Refiling the to-be-removed files to a trash folder is the usual example.
2640 Nmh's man page
2641 .Mp rmm (1)
2642 proposes to set the
2643 .Pe rmmproc
2644 to
2645 .Cl "refile +d
2646 to move messages to the trash folder,
2647 .Fn +d ,
2648 instead of renaming them with the backup prefix.
2649 The man page proposes additionally the expunge command
2650 .Cl "rm `mhpath +d all`
2651 to empty the trash folder.
2652 .P
2653 Removing messages in such a way has advantages.
2654 The mail storage is prevented from being cluttered with removed messages
2655 because they are all collected in one place.
2656 Existing and removed messages are thus separated more strictly.
2657 No backup files are silently overwritten.
2658 But most important is the ability to keep removed messages in the MH domain.
2659 Messages in the trash folder can be listed like those in any other folder.
2660 Deleted messages can be displayed like any other messages.
2661 .Pn refile
2662 can restore deleted messages.
2663 All operations on deleted files are still covered by the MH tools.
2664 The trash folder is just like any other folder in the mail storage.
2665 .P
2666 Similar to the draft folder case, I dropped the old backup prefix approach
2667 in favor for replacing it by the better suiting trash folder system.
2668 Hence,
2669 .Pn rmm
2670 calls
2671 .Pn refile
2672 to move the to-be-removed message to the trash folder,
2673 .Fn +trash
2674 by default.
2675 To sweep it clean, the user can use
2676 .Cl "rmm -unlink +trash a" ,
2677 where the
2678 .Sw -unlink
2679 switch causes the files to be unlinked.
2680 .Ci 8edc5aaf86f9f77124664f6801bc6c6cdf258173
2681 .Ci ca0b3e830b86700d9e5e31b1784de2bdcaf58fc5
2682 .P
2683 Dropping the legacy approach and converting to the new approach completely
2684 simplified the code base.
2685 The relationship between
2686 .Pn rmm
2687 and
2688 .Pn refile
2689 was inverted.
2690 In mmh,
2691 .Pn rmm
2692 invokes
2693 .Pn refile ,
2694 which used to be the other way round.
2695 Yet, the relationship is simpler now.
2696 Loops, like described in nmh's man page for
2697 .Mp refile (1),
2698 can no longer occur:
2699 .QS
2700 Since
2701 .Pn refile
2702 uses your
2703 .Pe rmmproc
2704 to delete the message, the
2705 .Pe rmmproc
2706 must NOT call
2707 .Pn refile
2708 without specifying
2709 .Sw -normmproc
2710 or you will create an infinite loop.
2711 .QE
2712 .LP
2713 .Pn rmm
2714 either unlinks a message with
2715 .Fu unlink()
2716 or invokes
2717 .Pn refile
2718 to move it to the trash folder.
2719 .Pn refile
2720 does not invoke any tools.
2721 .P
2722 By generalizing the message removal in the way that it became covered
2723 by the MH concepts made the whole system more powerful.
2729 .H2 "Modern Defaults
2730 .P
2731 Nmh has a bunch of convenience-improving features inactive by default,
2732 although one can expect every new user wanting to have them active.
2733 The reason they are inactive by default is the wish to stay compatible
2734 with old versions.
2735 But what is the definition for old versions?
2736 Still, the highly useful draft folder facility has not been activated
2737 by default although it was introduced over twenty-five years ago.
2738 .[
2739 rose romine real work
2740 .]
2741 The community seems not to care.
2742 This is one of several examples that require new users to first build up
2743 a profile before they can access the modern features of nmh.
2744 Without an extensive profile, the setup is hardly usable
2745 for modern emailing.
2746 The point is not the customization of the setup,
2747 but the need to activate generally useful facilities.
2748 .P
2749 Yet, the real problem lies less in enabling the features, as this is
2750 straight forward as soon as one knows what he wants.
2751 The real problem is that new users need deep insight into the project
2752 to find out about inactive features nmh already provides.
2753 To give an example, I needed one year of using nmh
2754 before I became aware of the existence of the attachment system.
2755 One could argue that this fact disqualifies my reading of the
2756 documentation.
2757 If I would have installed nmh from source back then, I could agree.
2758 Yet, I had used a prepackaged version and had expected that it would
2759 just work.
2760 Nevertheless, I had been convinced by the concepts of MH already
2761 and I am a software developer,
2762 still I required a lot of time to discover the cool features.
2763 How can we expect users to be even more advanced than me,
2764 just to allow them use MH in a convenient and modern way?
2765 Unless they are strongly convinced of the concepts, they will fail.
2766 I have seen friends of me giving up disappointed
2767 before they truly used the system,
2768 although they had been motivated in the beginning.
2769 They suffer hard enough to get used to the toolchest approach,
2770 we should spare them further inconveniences.
2771 .P
2772 Maintaining compatibility for its own sake is bad,
2773 because the code base collects more and more compatibility code.
2774 Sticking to the compatiblity code means remaining limited;
2775 whereas adjusting to the changes renders the compatibility unnecessary.
2776 Keeping unused alternatives in the code is a bad choice as they likely
2777 gather bugs, by not being well tested.
2778 Also, the increased code size and the greater number of conditions
2779 increase the maintenance costs.
2780 If any MH implementation would be the back-end of widespread
2781 email clients with large user bases, compatibility would be more
2782 important.
2783 Yet, it appears as if this is not the case.
2784 Hence, compatibility is hardly important for technical reasons.
2785 Its importance originates rather from personal reasons.
2786 Nmh's user base is small and old.
2787 Changing the interfaces would cause inconvenience to long-term users of MH.
2788 It would force them to change their many years old MH configurations.
2789 I do understand this aspect, but by sticking to the old users,
2790 new users are kept away.
2791 Yet, the future lies in new users.
2792 In consequence, mmh invites new users by providing a convenient
2793 and modern setup, readily usable out-of-the-box.
2794 .P
2795 In mmh, all modern features are active by default and many previous
2796 approaches are removed or only accessible in manual ways.
2797 New default features include:
2798 .BU
2799 The attachment system (\c
2800 .Hd Attach ).
2801 .Ci 8ff284ff9167eff8f5349481529332d59ed913b1
2802 .BU
2803 The draft folder facility (\c
2804 .Fn +drafts ).
2805 .Ci 337338b404931f06f0db2119c9e145e8ca5a9860
2806 .BU
2807 The unseen sequence (`u')
2808 .Ci c2360569e1d8d3678e294eb7c1354cb8bf7501c1
2809 and the sequence negation prefix (`!').
2810 .Ci db74c2bd004b2dc9bf8086a6d8bf773ac051f3cc
2811 .BU
2812 Quoting the original message in the reply.
2813 .Ci 67411b1f95d6ec987b4c732459e1ba8a8ac192c6
2814 .BU
2815 Forwarding messages using MIME.
2816 .Ci 6e271608b7b9c23771523f88d23a4d3593010cf1
2817 .P
2818 In consequence, a setup with a profile that defines only the path to the
2819 mail storage, is already convenient to use.
2820 Again, Paul Vixie's ``edginess'' call supports the direction I took:
2821 ``the `main branch' should just be modern''.
2822 .[
2823 paul vixie edginess nmh-workers
2824 .]
2830 .\" --------------------------------------------------------------
2831 .H1 "Styling
2832 .P
2833 Kernighan and Pike have emphasized the importance of style in the
2834 preface of their book:
2835 .[ [
2836 kernighan pike practice of programming
2837 .], p. x]
2838 .QS
2839 Chapter 1 discusses programming style.
2840 Good style is so important to good programming that we have chose
2841 to cover it first.
2842 .QE
2843 This section covers changes in mmh that were guided by the desire
2844 to improve on style.
2845 Many of them follow the rules given in the quoted book.
2846 .[
2847 kernighan pike practice of programming
2848 .]
2853 .H2 "Code Style
2854 .Id code-style
2855 .P
2856 .U3 "Indentation Style
2857 .P
2858 Indentation styles are the holy cow of programmers.
2859 Kernighan and Pike
2860 .[ [
2861 kernighan pike practice of programming
2862 .], p. 10]
2863 wrote:
2864 .QS
2865 Programmers have always argued about the layout of programs,
2866 but the specific style is much less important than its consistent
2867 application.
2868 Pick one style, preferably ours, use it consistently, and don't waste
2869 time arguing.
2870 .QE
2871 .P
2872 I agree that the constant application is most important,
2873 but I believe that some styles have advantages over others.
2874 For instance the indentation with tab characters only.
2875 Tab characters directly map to the nesting level \(en
2876 one tab, one level.
2877 Tab characters are flexible because developers can adjust them to
2878 whatever width they like to have.
2879 There is no more need to run
2880 .Pn unexpand
2881 or
2882 .Pn entab
2883 programs to ensure the correct mixture of leading tabs and spaces.
2884 The simple rules are: (1) Leading whitespace must consist of tabs only.
2885 (2) Any other whitespace should consist of spaces.
2886 These two rules ensure the integrity of the visual appearance.
2887 Although reformatting existing code should be avoided, I did it.
2888 I did not waste time arguing; I just reformated the code.
2889 .Ci a485ed478abbd599d8c9aab48934e7a26733ecb1
2891 .U3 "Comments
2892 .P
2893 Section 1.6 of
2894 .[ [
2895 kernighan pike practice of programming
2896 .], p. 23]
2897 demands: ``Don't belabor the obvious.''
2898 Hence, I simply removed all the comments in the following code excerpt:
2899 .VS
2900 context_replace(curfolder, folder); /* update current folder */
2901 seq_setcur(mp, mp->lowsel); /* update current message */
2902 seq_save(mp); /* synchronize message sequences */
2903 folder_free(mp); /* free folder/message structure */
2904 context_save(); /* save the context file */
2906 [...]
2908 int c; /* current character */
2909 char *cp; /* miscellaneous character pointer */
2911 [...]
2913 /* NUL-terminate the field */
2914 *cp = '\0';
2915 VE
2916 .Ci 426543622b377fc5d091455cba685e114b6df674
2917 .P
2918 The program code explains enough itself, already.
2921 .U3 "Names
2922 .P
2923 Kernighan and Pike suggest:
2924 ``Use active names for functions''.
2925 .[ [
2926 kernighan pike practice of programming
2927 .], p. 4]
2928 One application of this rule was the rename of
2929 .Fu check_charset()
2930 to
2931 .Fu is_native_charset() .
2932 .Ci 8d77b48284c58c135a6b2787e721597346ab056d
2933 The same change fixed a violation of ``Be accurate'' as well.
2934 The code did not match the expectation the function suggested,
2935 as it, for whatever reason, only compared the first ten characters
2936 of the charset name.
2937 .P
2938 More important than using active names is using descriptive names.
2939 .VS
2940 m_unknown(in); /* the MAGIC invocation... */
2941 VE
2942 Renaming the obscure
2943 .Fu m_unknown()
2944 function was a delightful event, although it made the code less funny.
2945 .Ci 611d68d19204d7cbf5bd585391249cb5bafca846
2946 .P
2947 Magic numbers are generally considered bad style.
2948 Obviously, Kernighan and Pike agree:
2949 ``Give names to magic numbers''.
2950 .[ [
2951 kernighan pike practice of programming
2952 .], p. 19]
2953 One such change was naming the type of input \(en mbox or mail folder \(en
2954 to be scanned:
2955 .VS
2956 #define SCN_MBOX (-1)
2957 #define SCN_FOLD 0
2958 VE
2959 .Ci 7ffb36d28e517a6f3a10272056fc127592ab1c19
2960 .P
2961 The argument
2962 .Ar outnum
2963 of the function
2964 .Fu scan()
2965 in
2966 .Fn uip/scansbr.c
2967 defines the number of the message to be created.
2968 If no message is to be created, the argument is misused to transport
2969 program logic.
2970 This lead to obscure code.
2971 I improved the clarity of the code by introducing two variables:
2972 .VS
2973 int incing = (outnum > 0);
2974 int ismbox = (outnum != 0);
2975 VE
2976 They cover the magic values and are used for conditions.
2977 The variable
2978 .Ar outnum
2979 is only used when it holds an ordinary message number.
2980 .Ci b8b075c77be7794f3ae9ff0e8cedb12b48fd139f
2981 The clarity improvement of the change showed detours in the program logic
2982 of related code parts.
2983 Having the new variables with descriptive names, a more
2984 straight forward implementation became apparent.
2985 Before the code was clarified, the possibility to improve had not be seen.
2986 .Ci aa60b0ab5e804f8befa890c0a6df0e3143ce0723
2990 .H2 "Structural Rework
2991 .P
2992 Although the stylistic changes described up to here improve the
2993 readability of the source code, all of them are changes ``in the small''.
2994 Structural changes affect a much larger area.
2995 They are more difficult to do but lead to larger improvements,
2996 especially as they influence the outer shape of the tools as well.
2997 .P
2998 At the end of their chapter on style,
2999 Kernighan and Pike ask: ``But why worry about style?''
3000 Following are two examples of structural rework that show
3001 why style is important in the first place.
3004 .U3 "Rework of \f(CWanno\fP
3005 .P
3006 Until 2002,
3007 .Pn anno
3008 had six functional command line switches,
3009 .Sw -component
3010 and
3011 .Sw -text ,
3012 which have an argument each,
3013 and the two pairs of flags,
3014 .Sw -[no]date
3015 and
3016 .Sw -[no]inplace .
3017 Then Jon Steinhart introduced his attachment system.
3018 In need for more advanced annotation handling, he extended
3019 .Pn anno .
3020 He added five more switches:
3021 .Sw -draft ,
3022 .Sw -list ,
3023 .Sw -delete ,
3024 .Sw -append ,
3025 and
3026 .Sw -number ,
3027 the last one taking an argument.
3028 .Ci 7480dbc14bc90f2d872d434205c0784704213252
3029 Later,
3030 .Sw -[no]preserve
3031 was added.
3032 .Ci d9b1d57351d104d7ec1a5621f090657dcce8cb7f
3033 Then, the Synopsis section of the man page
3034 .Mp anno (1)
3035 read:
3036 .VS
3037 anno [+folder] [msgs] [-component field] [-inplace | -noinplace]
3038 [-date | -nodate] [-draft] [-append] [-list] [-delete]
3039 [-number [num|all]] [-preserve | -nopreserve] [-version]
3040 [-help] [-text body]
3041 VE
3042 .LP
3043 The implementation followed the same structure.
3044 Problems became visible when
3045 .Cl "anno -list -number 42
3046 worked on the current message instead on message number 42,
3047 and
3048 .Cl "anno -list -number l:5
3049 did not work on the last five messages but failed with the mysterious
3050 error message: ``anno: missing argument to -list''.
3051 Yet, the invocation matched the specification in the man page.
3052 There, the correct use of
3053 .Sw -number
3054 was defined as being
3055 .Cl "[-number [num|all]]
3056 and the textual description for the combination with
3057 .Sw -list
3058 read:
3059 .QS
3060 The
3061 .Sw -list
3062 option produces a listing of the field bodies for
3063 header fields with names matching the specified component,
3064 one per line. The listing is numbered, starting at 1, if the
3065 .Sw -number
3066 option is also used.
3067 .QE
3068 .LP
3069 The problem was manifold.
3070 The code required a numeric argument to the
3071 .Sw -number
3072 switch.
3073 If it was missing or non-numeric,
3074 .Pn anno
3075 aborted with an error message that had an off-by-one error,
3076 printing the switch one before the failing one.
3077 Semantically, the argument to the
3078 .Sw -number
3079 switch is only necessary in combination with
3080 .Sw -delete ,
3081 but not with
3082 .Sw -list .
3083 In the former case it is even necessary.
3084 .P
3085 Trying to fix these problems on the surface would not have solved it truly.
3086 The problems discovered originate from a discrepance between the semantic
3087 structure of the problem and the structure implemented in the program.
3088 Such structural differences can not be cured on the surface.
3089 They need to be solved by adjusting the structure of the implementation
3090 to the structure of the problem.
3091 .P
3092 In 2002, the new switches
3093 .Sw -list
3094 and
3095 .Sw -delete
3096 were added in the same way, the
3097 .Sw -number
3098 switch for instance had been added.
3099 Yet, they are of structural different type.
3100 Semantically,
3101 .Sw -list
3102 and
3103 .Sw -delete
3104 introduce modes of operation.
3105 Historically,
3106 .Pn anno
3107 had only one operation mode: adding header fields.
3108 With the extension, it got two moder modes:
3109 listing and deleting header fields.
3110 The structure of the code changes did not pay respect to this
3111 fundamental change to
3112 .Pn anno 's
3113 behavior.
3114 Neither the implementation nor the documentation did clearly
3115 define them as being exclusive modes of operation.
3116 Having identified the problem, I solved it by putting structure into
3117 .Pn anno
3118 and its documentation.
3119 .Ci d54c8db8bdf01e8381890f7729bc0ef4a055ea11
3120 .P
3121 The difference is visible in both, the code and the documentation.
3122 The following code excerpt:
3123 .VS
3124 int delete = -2; /* delete header element if set */
3125 int list = 0; /* list header elements if set */
3126 [...]
3127 case DELETESW: /* delete annotations */
3128 delete = 0;
3129 continue;
3130 case LISTSW: /* produce a listing */
3131 list = 1;
3132 continue;
3133 VE
3134 .LP
3135 was replaced by:
3136 .VS
3137 static enum { MODE_ADD, MODE_DEL, MODE_LIST } mode = MODE_ADD;
3138 [...]
3139 case DELETESW: /* delete annotations */
3140 mode = MODE_DEL;
3141 continue;
3142 case LISTSW: /* produce a listing */
3143 mode = MODE_LIST;
3144 continue;
3145 VE
3146 .LP
3147 The replacement code does not only reflect the problem's structure better,
3148 it is easier to understand as well.
3149 The same applies to the documentation.
3150 The man page was completely reorganized to propagate the same structure.
3151 This is visible in the Synopsis section:
3152 .VS
3153 anno [+folder] [msgs] [-component field] [-text body]
3154 [-append] [-date | -nodate] [-preserve | -nopreserve]
3155 [-Version] [-help]
3157 anno -delete [+folder] [msgs] [-component field] [-text
3158 body] [-number num | all ] [-preserve | -nopreserve]
3159 [-Version] [-help]
3161 anno -list [+folder] [msgs] [-component field] [-number]
3162 [-Version] [-help]
3163 VE
3164 .\" XXX think about explaining the -preserve rework?
3168 .U3 "Path Conversion
3169 .P
3170 Four kinds of path names can appear in MH:
3171 .IP (1)
3172 Absolute Unix directory paths, like
3173 .Fn /etc/passwd .
3174 .IP (2)
3175 Relative Unix directory paths, like
3176 .Fn ./foo/bar .
3177 .IP (3)
3178 Absolute MH folder paths, like
3179 .Fn +friends/phil .
3180 .IP (4)
3181 Relative MH folder paths, like
3182 .Fn @subfolder .
3183 .P
3184 The last type, relative MH folder paths, are hardly documented.
3185 Nonetheless, they are useful for large mail storages.
3186 The current mail folder is specified as `\c
3187 .Fn @ ',
3188 just like the current directory is specified as `\c
3189 .Fn . '.
3190 .P
3191 To allow MH tools to understand all four notations,
3192 they need to convert between them.
3193 In nmh, these path name conversion functions were located in the files
3194 .Fn sbr/path.c
3195 (``return a pathname'') and
3196 .Fn sbr/m_maildir.c
3197 (``get the path for the mail directory'').
3198 The seven functions in the two files were documented with no more
3199 than two comments, which described obvious information.
3200 The function signatures were neither explaining:
3201 .VS
3202 char *path(char *, int);
3203 char *pluspath(char *);
3204 char *m_mailpath(char *);
3205 char *m_maildir(char *);
3206 VE
3207 .P
3208 My investigation provides the following description:
3209 .BU
3210 The second parameter of
3211 .Fu path()
3212 defines the type of path given as first parameter.
3213 Directory paths are converted to absolute directory paths.
3214 Folder paths are converted to absolute folder paths.
3215 Folder paths must not include a leading `@' character.
3216 Leading plus characters are preserved.
3217 The result is a pointer to newly allocated memory.
3218 .BU
3219 .Fu pluspath()
3220 is a convenience-wrapper to
3221 .Fu path() ,
3222 to convert folder paths only.
3223 This function can not be used for directory paths.
3224 An empty string parameter causes a buffer overflow.
3225 .BU
3226 .Fu m_mailpath()
3227 converts directory paths to absolute directory paths.
3228 The characters `+' or `@' at the beginning of the path name are
3229 treated literal, i.e. as the first character of a relative directory path.
3230 Hence, this function can not be used for folder paths.
3231 In any case, the result is an absolute directory path.
3232 The result is a pointer to newly allocated memory.
3233 .BU
3234 .Fu m_maildir()
3235 returns the parameter unchanged if it is an absolute directory path
3236 or begins with the entry `.' or `..'.
3237 All other strings are prepended with the current working directory.
3238 Hence, this functions can not be used for folder paths.
3239 The result is either an absolute directory path or a relative
3240 directory path, starting with a dot.
3241 In contrast to the other functions, the result is a pointer to
3242 static memory.
3243 .P
3244 The situation was obscure, irritating, error-prone, and non-orthogonal.
3245 No clear terminology was used to name the different kinds of path names.
3246 The first argument of
3247 .Fu m_mailpath() ,
3248 for instance, was named
3249 .Ar folder ,
3250 though
3251 .Fu m_mailpath()
3252 can not be used for MH folders.
3253 .P
3254 I reworked the path name conversion completely, introducing clarity.
3255 First of all, the terminology needed to be defined.
3256 A path name is either in the Unix domain, then it is called
3257 \fIdirectory path\fP, `dirpath' for short, or it is in the MH domain,
3258 then it is called \fIfolder path\fP, `folpath' for short.
3259 The two terms need to be used with strict distinction.
3260 Having a clear terminology is often an indicator of having understood
3261 the problem itself.
3262 Second, I exploited the concept of path type indicators.
3263 By requesting every path name to start with a clear type identifier,
3264 conversion between the types can be fully automated.
3265 Thus the tools can accept paths of any type from the user.
3266 Therefore, it was necessary to require relative directory paths to be
3267 prefixed with a dot character.
3268 In consequence, the dot character could no longer be an alias for the
3269 current message.
3270 .Ci cff0e16925e7edbd25b8b9d6d4fbdf03e0e60c01
3271 Third, I created three new functions to replace the previous mess:
3272 .BU
3273 .Fu expandfol()
3274 converts folder paths to absolute folder paths,
3275 without the leading plus character.
3276 Directory paths are simply passed through.
3277 This function is to be used for folder paths only, thus the name.
3278 The result is a pointer to static memory.
3279 .BU
3280 .Fu expanddir()
3281 converts directory paths to absolute directory paths.
3282 Folder paths are treated as relative directory paths.
3283 This function is to be used for directory paths only, thus the name.
3284 The result is a pointer to static memory.
3285 .BU
3286 .Fu toabsdir()
3287 converts any type of path to an absolute directory path.
3288 This is the function of choice for path conversion.
3289 Absolute directory paths are the most general representation of a
3290 path name.
3291 The result is a pointer to static memory.
3292 .P
3293 The new functions have names that indicate their use.
3294 Two of the functions convert relative to absolute path names of the
3295 same type.
3296 The third function converts any path name type to the most general one,
3297 the absolute directory path.
3298 All of the functions return pointers to static memory.
3299 All three functions are implemented in
3300 .Fn sbr/path.c .
3301 .Fn sbr/m_maildir.c
3302 is removed.
3303 .Ci d39e2c447b0d163a5a63f480b23d06edb7a73aa0
3304 .P
3305 Along with the path conversion rework, I also replaced
3306 .Fu getfolder(FDEF)
3307 with
3308 .Fu getdeffol()
3309 and
3310 .Fu getfolder(FCUR)
3311 with
3312 .Fu getcurfol() ,
3313 which is only a convenience wrapper for
3314 .Fu expandfol("@") .
3315 This code was moved from
3316 .Fn sbr/getfolder.c
3317 to
3318 .Fn sbr/path.c .
3319 .Ci d39e2c447b0d163a5a63f480b23d06edb7a73aa0
3320 .P
3321 The related function
3322 .Fu etcpath()
3323 was moved to
3324 .Fn sbr/path.c ,
3325 too
3326 .Ci b4c29794c12099556151d93a860ee51badae2e35 .
3327 Previously, it had been located in
3328 .Fn config/config.c ,
3329 for whatever reasons.
3330 .P
3331 .Fn sbr/path.c
3332 now contains all path handling code.
3333 Only 173 lines of code were needed to replace the previous 252 lines.
3334 The readability of the code is highly improved.
3335 Additionally, each of the six exported and one static functions
3336 is introduced by an explaining comment.
3341 .H2 "Profile Reading
3342 .P
3343 The MH profile contains the configuration for the user-specific MH setup.
3344 MH tools read the profile right after starting up,
3345 as it contains the location of the user's mail storage
3346 and similar settings that influence the whole setup.
3347 Further more, the profile contains the default switches for the tools,
3348 hence, it must be read before the command line switches are processed.
3349 .P
3350 For historic reasons, some MH tools did not read the profile and context.
3351 Among them were
3352 .Pn post /\c
3353 .Pn spost ,
3354 .Pn mhmail ,
3355 and
3356 .Pn slocal .
3357 The reason why these tools ignored the profile were not clearly stated.
3358 During the discussion on the nmh-workers mailing list,
3359 .[
3360 nmh-workers levine post profile
3361 .]
3362 David Levine posted an explanation, quoting John Romine:
3363 .QS
3364 I asked John Romine and here's what he had to say, which
3365 agrees and provides an example that convinces me:
3366 .QS
3367 My take on this is that
3368 .Pn post
3369 should not be called by users directly, and it doesn't read the
3370 .Fn .mh_profile
3371 (only front-end UI programs read the profile).
3372 .QP
3373 For example, there can be contexts where
3374 .Pn post
3375 is called by a helper program (like `\c
3376 .Pn mhmail ')
3377 which may be run by a non-MH user.
3378 We don't want this to prompt the user to create an MH profile, etc.
3379 .QP
3380 My suggestion would be to have
3381 .Pn send
3382 pass a (hidden) `\c
3383 .Sw -fileproc
3384 .Ar proc '
3385 option to
3386 .Pn post
3387 if needed.
3388 You could also
3389 use an environment variable (I think
3390 .Pn send /\c
3391 .Pn whatnow
3392 do this).
3393 .QE
3394 I think that's the way to go.
3395 My personal preference is to use a command line option,
3396 not an environment variable.
3397 .QE
3398 .P
3399 To solve the problem of
3400 .Pn post
3401 not honoring the
3402 .Pe fileproc
3403 profile entry,
3404 the community roughly agreed that a switch
3405 .Sw -fileproc
3406 should be added to
3407 .Pn post
3408 to be able to pass a different fileproc.
3409 I strongly disagree with this approach because it does not solve
3410 the problem; it only removes a single symptom.
3411 The problem is that
3412 .Pn post
3413 does not behave as expected.
3414 But all programs should behave as expected.
3415 Clear and simple concepts are a precondition for this.
3416 Hence, the real solution is having all MH tools read the profile.
3417 .P
3418 Yet, the problem has a further aspect.
3419 It mainly originates in
3420 .Pn mhmail .
3421 .Pn mhmail
3422 was intended to be a replacement for
3423 .Pn mailx
3424 on systems with MH installations.
3425 .Pn mhmail
3426 should have been able to use just like
3427 .Pn mailx ,
3428 but sending the message via MH's
3429 .Pn post
3430 instead of
3431 .Pn sendmail .
3432 Using
3433 .Pn mhmail
3434 should not be influenced by the question whether the user had
3435 MH set up for himself or not.
3436 .Pn mhmail
3437 did not read the profile as this requests the user to set up MH
3438 if not done yet.
3439 As
3440 .Pn mhmail
3441 used
3442 .Pn post ,
3443 .Pn post
3444 could not read the profile neither.
3445 This is the reason why
3446 .Pn post
3447 does not read the profile.
3448 This is the reason for the actual problem.
3449 It was not much of a problem because
3450 .Pn post
3451 was not intended to be used by users directly.
3452 .Pn send
3453 is the interactive front-end to
3454 .Pn post .
3455 .Pn send
3456 read the profile and passed all relevant values on the command line to
3457 .Pn post
3458 \(en an awkward solution.
3459 .P
3460 The important insight is that
3461 .Pn mhmail
3462 is no true MH tool.
3463 The concepts broke because this outlandish tool was treated as any other
3464 MH tool.
3465 Instead it should have been treated accordingly to its foreign style.
3466 The solution is not to prevent the tools reading the profile but
3467 to instruct them reading a different profile.
3468 .Pn mhmail
3469 could have set up a well-defined profile and caused all MH tools
3470 in the session use it by exporting an environment variable.
3471 With this approach, no special cases would have been introduced,
3472 no surprises would have been caused.
3473 By writing a clean-profile-wrapper, the concept could have been
3474 generalized orthogonally to the whole MH toolchest.
3475 Then Rose's motivation behind the decision that
3476 .Pn post
3477 ignores the profile, as quoted by Jeffrey Honig,
3478 .[
3479 nmh-workers post profile
3480 .]
3481 would have become possible:
3482 .QS
3483 when you run mh commands in a script, you want all the defaults to be
3484 what the man page says.
3485 when you run a command by hand, then you want your own defaults...
3486 .QE
3487 .LP
3488 Yet, I consider this explanation short-sighted.
3489 We should rather regard theses two cases as just two different MH setups,
3490 based on two different profiles.
3491 Mapping such problems on the concepts of switching between different
3492 profiles, solves them once for all.
3493 .P
3494 In mmh, the wish to have
3495 .Pn mhmail
3496 as as replacement for
3497 .Pn mailx
3498 is considered obsolete.
3499 Mmh's
3500 .Pn mhmail
3501 does no longer cover this use-case.
3502 Currently,
3503 .Pn mhmail
3504 is in a transition state.
3505 .Ci 32d4f9daaa70519be3072479232ff7be0500d009
3506 It may become a front-end to
3507 .Pn comp ,
3508 which provides an interface more convenient in some cases.
3509 In this case,
3510 .Pn mhmail
3511 will become an ordinary MH tool, reading the profile.
3512 If, however, this idea will not convince, then
3513 .Pn mhmail
3514 will be removed.
3515 .P
3516 Every program in the mmh toolchest reads the profile.
3517 The only exception is
3518 .Pn slocal ,
3519 which is not considered part of the mmh toolchest.
3520 This MDA is only distributed with mmh, currently.
3521 Mmh has no
3522 .Pn post
3523 program, but
3524 .Pn spost ,
3525 which now reads the profile.
3526 .Ci 3e017a7abbdf69bf0dff7a4073275961eda1ded8
3527 With this change,
3528 .Pn send
3529 and
3530 .Pn spost
3531 can be considered to be merged.
3532 .Pn spost
3533 is only invoked directly by the to-be-changed
3534 .Pn mhmail
3535 implementation and by
3536 .Pn rcvdist ,
3537 which will require rework.
3538 .P
3539 The
3540 .Fu context_foil()
3541 function to pretend to have read an empty profile was removed.
3542 .Ci 68af8da96bea87a5541988870130b6209ce396f6
3543 All mmh tools read the profile.
3547 .H2 "Standard Libraries
3548 .P
3549 MH is one decade older than the POSIX and ANSI C standards.
3550 Hence, MH included own implementations of functions
3551 that are standardized and thus widely available today,
3552 but were not back then.
3553 Today, twenty years after the POSIX and ANSI C were published,
3554 developers can expect system to comply with these standards.
3555 In consequence, MH-specific replacements for standard functions
3556 can and should be dropped.
3557 Kernighan and Pike advise: ``Use standard libraries.''
3558 .[ [
3559 kernighan pike practice of programming
3560 .], p. 196]
3561 Actually, MH had followed this advice in history,
3562 but it had not adjusted to the changes in this field.
3563 The
3564 .Fu snprintf()
3565 function, for instance, was standardized with C99 and is available
3566 almost everywhere because of its high usefulness.
3567 In project's own implementation of
3568 .Fu snprintf()
3569 was dropped in March 2012 in favor for using the one of the
3570 standard library.
3571 .Ci 0052f1024deb0a0a2fc2e5bacf93d45a5a9c9b32
3572 Such decisions limit the portability of mmh
3573 if systems don't support these standardized and widespread functions.
3574 This compromise is made because mmh focuses on the future.
3575 .P
3576 I am not yet thirty years old and my C and Unix experience comprises
3577 only half a dozen years.
3578 Hence, I need to learn about the history in retrospective.
3579 I have not used those ancient constructs myself.
3580 I have not suffered from their incompatibilities.
3581 I have not longed for standardization.
3582 All my programming experience is from a time when ANSI C and POSIX
3583 were well established already.
3584 I have only read a lot of books about the (good) old times.
3585 This puts me in a difficult positions when working with old code.
3586 I need to freshly acquire knowledge about old code constructs and ancient
3587 programming styles, whereas older programmers know these things by
3588 heart from their own experience.
3589 .P
3590 Being aware of the situation, I rather let people with more historic
3591 experience replace ancient code constructs with standardized ones.
3592 Lyndon Nerenberg covered large parts of this task for the nmh project.
3593 He converted project-specific functions to POSIX replacements,
3594 also removing the conditionals compilation of now standardized features.
3595 Ken Hornstein and David Levine had their part in the work, too.
3596 Often, I only needed to pull over changes from nmh into mmh.
3597 These changes include many commits; these are among them:
3598 .Ci 768b5edd9623b7238e12ec8dfc409b82a1ed9e2d
3599 .Ci 0052f1024deb0a0a2fc2e5bacf93d45a5a9c9b32 .
3600 .P
3601 During my own work, I tidied up the \fIMH standard library\fP,
3602 .Fn libmh.a ,
3603 which is located in the
3604 .Fn sbr
3605 (``subroutines'') directory in the source tree.
3606 The MH library includes functions that mmh tools usually need.
3607 Among them are MH-specific functions for profile, context, sequence,
3608 and folder handling, but as well
3609 MH-independent functions, such as auxiliary string functions,
3610 portability interfaces and error-checking wrappers for critical
3611 functions of the standard library.
3612 .P
3613 I have replaced the
3614 .Fu atooi()
3615 function with calls to
3616 .Fu strtoul()
3617 with the third parameter, the base, set to eight.
3618 .Fu strtoul()
3619 is part of C89 and thus considered safe to use.
3620 .Ci c490c51b3c0f8871b6953bd0c74551404f840a74
3621 .P
3622 I did remove project-included fallback implementations of
3623 .Fu memmove()
3624 and
3625 .Fu strerror() ,
3626 although Peter Maydell had re-included them into nmh in 2008
3627 to support SunOS 4.
3628 Nevertheless, these functions are part of ANSI C.
3629 Systems that do not even provide full ANSI C support should not
3630 put a load on mmh.
3631 .Ci b067ff5c465a5d243ce5a19e562085a9a1a97215
3632 .P
3633 The
3634 .Fu copy()
3635 function copies the string in argument one to the location in two.
3636 In contrast to
3637 .Fu strcpy() ,
3638 it returns a pointer to the terminating null-byte in the destination area.
3639 The code was adjusted to replace
3640 .Fu copy()
3641 with
3642 .Fu strcpy() ,
3643 except within
3644 .Fu concat() ,
3645 where
3646 .Fu copy()
3647 was more convenient.
3648 Therefore, the definition of
3649 .Fu copy()
3650 was moved into the source file of
3651 .Fu concat()
3652 and its visibility is now limited to it.
3653 .Ci 552fd7253e5ee9e554c5c7a8248a6322aa4363bb
3654 .P
3655 The function
3656 .Fu r1bindex()
3657 had been a generalized version of
3658 .Fu basename()
3659 with minor differences.
3660 As all calls to
3661 .Fu r1bindex()
3662 had the slash (`/') as delimiter anyway,
3663 replacing
3664 .Fu r1bindex()
3665 with the more specific and better-named function
3666 .Fu basename()
3667 became desirable.
3668 Unfortunately, many of the 54 calls to
3669 .Fu r1bindex()
3670 depended on a special behavior,
3671 which differed from the POSIX specification for
3672 .Fu basename() .
3673 Hence,
3674 .Fu r1bindex()
3675 was kept but renamed to
3676 .Fu mhbasename() ,
3677 fixing the delimiter to the slash.
3678 .Ci 240013872c392fe644bd4f79382d9f5314b4ea60
3679 For possible uses of
3680 .Fu r1bindex()
3681 with a different delimiter,
3682 the ANSI C function
3683 .Fu strrchr()
3684 provides the core functionality.
3685 .P
3686 The
3687 .Fu ssequal()
3688 function \(en apparently for ``substring equal'' \(en
3689 was renamed to
3690 .Fu isprefix() ,
3691 because this is what it actually checks.
3692 .Ci c20b4fa14515c7ab388ce35411d89a7a92300711
3693 Its source file had included the following comments, no joke.
3694 .VS
3695 /*
3696 * THIS CODE DOES NOT WORK AS ADVERTISED.
3697 * It is actually checking if s1 is a PREFIX of s2.
3698 * All calls to this function need to be checked to see
3699 * if that needs to be changed. Prefix checking is cheaper, so
3700 * should be kept if it's sufficient.
3701 */
3703 /*
3704 * Check if s1 is a substring of s2.
3705 * If yes, then return 1, else return 0.
3706 */
3707 VE
3708 Two months later, it was completely removed by replacing it with
3709 .Fu strncmp() .
3710 .Ci b0b1dd37ff515578cf7cba51625189eb34a196cb
3716 .H2 "User Data Locations
3717 .P
3718 In nmh, a personal setup consists of the MH profile and the MH directory.
3719 The profile is a file named
3720 .Fn \&.mh_profile
3721 in the user's home directory.
3722 It contains the static configuration.
3723 It also contains the location of the MH directory in the profile entry
3724 .Pe Path .
3725 The MH directory contains the mail storage and is the first
3726 place to search for personal forms, scan formats, and similar
3727 configuration files.
3728 The location of the MH directory can be chosen freely by the user.
3729 The default and usual name is a directory named
3730 .Fn Mail
3731 in the home directory.
3732 .P
3733 The way MH data is splitted between profile and MH directory is a legacy.
3734 It is only sensible in a situation where the profile is the only
3735 configuration file.
3736 Why else should the mail storage and the configuration files be intermixed?
3737 They are different kinds of data:
3738 The data to be operated on and the configuration to change how
3739 tools operate.
3740 Splitting the configuration between the profile and the MH directory
3741 is bad.
3742 Merging the mail storage and the configuration in one directory is bad
3743 as well.
3744 As the mail storage and the configuration were not separated sensibly
3745 in the first place, I did it now.
3746 .P
3747 Personal mmh data is grouped by type, resulting in two distinct parts:
3748 The mail storage and the configuration.
3749 In mmh, the mail storage directory still contains all the messages,
3750 but, in exception of public sequences files, nothing else.
3751 In difference to nmh, the auxiliary configuration files are no longer
3752 located there.
3753 Therefore, the directory is no longer called the user's \fIMH directory\fP
3754 but his \fImail storage\fP.
3755 Its location is still user-chosen, with the default name
3756 .Fn Mail ,
3757 in the user's home directory.
3758 In mmh, the configuration is grouped together in
3759 the hidden directory
3760 .Fn \&.mmh
3761 in the user's home directory.
3762 This \fImmh directory\fP contains the context file, personal forms,
3763 scan formats, and the like, but also the user's profile, now named
3764 .Fn profile .
3765 The location of the profile is no longer fixed to
3766 .Fn $HOME/.mh_profile
3767 but to
3768 .Fn $HOME/.mmh/profile .
3769 Having both, the file
3770 .Fn $HOME/.mh_profile
3771 and the configuration directory
3772 .Fn $HOME/.mmh
3773 appeared to be inconsistent.
3774 The approach chosen for mmh is consistent, simple, and familiar to
3775 Unix users.
3776 .Ci 7030d7edb099bff36ded7548bb5380f7acab4f9b
3777 .P
3778 MH allows users to have multiple MH setups.
3779 Therefore, it is necessary to select a different profile.
3780 The profile is the single entry point to access the rest of a
3781 personal MH setup.
3782 In nmh, the environment variable
3783 .Ev MH
3784 could be used to specifiy a different profile.
3785 To operate in the same MH setup with a separate context,
3786 the
3787 .Ev MHCONTEXT
3788 environment variable could be used.
3789 This allows having own current folders and current messages in
3790 each terminal, for instance.
3791 In mmh, three environment variables are used.
3792 .Ev MMH
3793 overrides the default location of the mmh directory (\c
3794 .Fn .mmh ).
3795 .Ev MMHP
3796 and
3797 .Ev MMHC
3798 override the paths to the profile and context files, respectively.
3799 This approach allows the set of personal configuration files to be chosen
3800 independently from the profile, context, and mail storage.
3801 .Ci 7030d7edb099bff36ded7548bb5380f7acab4f9b
3802 .P
3803 The separation of the files by type is sensible and convenient.
3804 The new approach has no functional disadvantages,
3805 as every setup I can imagine can be implemented with both approaches,
3806 possibly even easier with the new approach.
3807 The main achievement of the change is the clear and sensible split
3808 between mail storage and configuration.
3814 .H2 "Modularization
3815 .P
3816 The source code of the mmh tools is located in the
3817 .Fn uip
3818 (``user interface programs'') directory.
3819 Each tools has a source file with the same name.
3820 For example,
3821 .Pn rmm
3822 is built from
3823 .Fn uip/rmm.c .
3824 Some source files are used for multiple programs.
3825 For example
3826 .Fn uip/scansbr.c
3827 is used for both,
3828 .Pn scan
3829 and
3830 .Pn inc .
3831 In nmh, 49 tools were built from 76 source files.
3832 This is a ratio of 1.6 source files per program.
3833 32 programs depended on multiple source files;
3834 17 programs depended on one source file only.
3835 In mmh, 39 tools are built from 51 source files.
3836 This is a ratio of 1.3 source files per program.
3837 18 programs depend on multiple source files;
3838 21 programs depend on one source file only.
3839 (These numbers and the ones in the following text ignore the MH library
3840 as well as shell scripts and multiple names for the same program.)
3841 .P
3842 Splitting the source code of a large program into multiple files can
3843 increase the readability of its source code.
3844 Most of the mmh tools, however, are simple and straight-forward programs.
3845 With the exception of the MIME handling tools,
3846 .Pn pick
3847 is the largest tools.
3848 It contains 1\|037 lines of source code (measured with
3849 .Pn sloccount ), excluding the MH library.
3850 Only the MIME handling tools (\c
3851 .Pn mhbuild ,
3852 .Pn mhstore ,
3853 .Pn show ,
3854 etc.)
3855 are larger.
3856 Splitting programs with less than 1\|000 lines of code into multiple
3857 source files seldom leads to better readability.
3858 For such tools, splitting makes sense
3859 when parts of the code are reused in other programs,
3860 and the reused code fragment is not general enough
3861 for including it in the MH library,
3862 or, if the code has dependencies on a library that only few programs need.
3863 .Fn uip/packsbr.c ,
3864 for instance, provides the core program logic for the
3865 .Pn packf
3866 and
3867 .Pn rcvpack
3868 programs.
3869 .Fn uip/packf.c
3870 and
3871 .Fn uip/rcvpack.c
3872 mainly wrap the core function appropriately.
3873 No other tools use the folder packing functions.
3874 As another example,
3875 .Fn uip/termsbr.c
3876 provides termcap support, which requires linking with a termcap or
3877 curses library.
3878 Including
3879 .Fn uip/termsbr.c
3880 into the MH library would require every program to be linked with
3881 termcap or curses, although only few of the programs require it.
3882 .P
3883 The task of MIME handling is complex enough that splitting its code
3884 into multiple source files improves the readability.
3885 The program
3886 .Pn mhstore ,
3887 for instance, is compiled out of seven source files with 2\|500
3888 lines of code in summary.
3889 The main code file
3890 .Fn uip/mhstore.c
3891 consists of 800 lines; the other 1\|700 lines of code are reused in
3892 other MIME handling tools.
3893 It seems to be worthwhile to bundle the generic MIME handling code into
3894 a MH-MIME library, as a companion to the MH standard library.
3895 This is left open for the future.
3896 .P
3897 The work already accomplished focussed on the non-MIME tools.
3898 The amount of code compiled into each program was reduced.
3899 This eases the understanding of the code base.
3900 In nmh,
3901 .Pn comp
3902 was built from six source files:
3903 .Fn comp.c ,
3904 .Fn whatnowproc.c ,
3905 .Fn whatnowsbr.c ,
3906 .Fn sendsbr.c ,
3907 .Fn annosbr.c ,
3908 and
3909 .Fn distsbr.c .
3910 In mmh, it builds from only two:
3911 .Fn comp.c
3912 and
3913 .Fn whatnowproc.c .
3914 In nmh's
3915 .Pn comp ,
3916 the core function of
3917 .Pn whatnow ,
3918 .Pn send ,
3919 and
3920 .Pn anno
3921 were compiled into
3922 .Pn comp .
3923 This saved the need to execute these programs with
3924 .Fu fork()
3925 and
3926 .Fu exec() ,
3927 two expensive system calls.
3928 Whereis this approach improved the time performance,
3929 it interweaved the source code.
3930 Core functionalities were not encapsulated into programs but into
3931 function, which were then wrapped by programs.
3932 For example,
3933 .Fn uip/annosbr.c
3934 included the function
3935 .Fu annotate() .
3936 Each program that wanted to annotate messages, included the source file
3937 .Fn uip/annosbr.c
3938 and called
3939 .Fu annotate() .
3940 Because the function
3941 .Fu annotate()
3942 was used like the tool
3943 .Pn anno ,
3944 it had seven parameters, reflecting the command line switches of the tool.
3945 When another pair of command line switches was added to
3946 .Pn anno ,
3947 a rather ugly hack was implemented to avoid adding another parameter
3948 to the function.
3949 .Ci d9b1d57351d104d7ec1a5621f090657dcce8cb7f
3950 .P
3951 Separation simplifies the understanding of program code
3952 because the area influenced by any particular statement is smaller.
3953 The separating on the program-level is more strict than the separation
3954 on the function level.
3955 In mmh, the relevant code of
3956 .Pn comp
3957 comprises the two files
3958 .Fn uip/comp.c
3959 and
3960 .Fn uip/whatnowproc.c ,
3961 together 210 lines of code.
3962 In nmh,
3963 .Pn comp
3964 comprises six files with 2\|450 lines.
3965 Not all of the code in these six files was actually used by
3966 .Pn comp ,
3967 but the code reader needed to read all of the code first to know which
3968 parts were used.
3969 .P
3970 As I have read a lot in the code base during the last two years,
3971 I learned about the easy and the difficult parts.
3972 Code is easy to understand if:
3973 .BU
3974 The influenced code area is small.
3975 .BU
3976 The boundaries are strictly defined.
3977 .BU
3978 The code is written straight-forward.
3979 .P
3980 .\" XXX move this paragraph somewhere else?
3981 Reading
3982 .Pn rmm 's
3983 source code in
3984 .Fn uip/rmm.c
3985 is my recommendation for a beginner's entry point into the code base of nmh.
3986 The reasons are that the task of
3987 .Pn rmm
3988 is straight forward and it consists of one small source code file only,
3989 yet its source includes code constructs typical for MH tools.
3990 With the introduction of the trash folder in mmh,
3991 .Pn rmm
3992 became a bit more complex, because it invokes
3993 .Pn refile .
3994 Still, it is a good example for a simple tool with clear sources.
3995 .P
3996 Understanding
3997 .Pn comp
3998 requires to read 210 lines of code in mmh, but ten times as much in nmh.
3999 Due to the aforementioned hack in
4000 .Pn anno
4001 to save the additional parameter, information passed through the program's
4002 source base in obscure ways.
4003 Thus, understanding
4004 .Pn comp ,
4005 required understanding the inner workings of
4006 .Fn uip/annosbr.c
4007 first.
4008 To be sure to fully understand a program, its whole source code needs
4009 to be examined.
4010 Not doing so is a leap of faith, assuming that the developers
4011 have avoided obscure programming techniques.
4012 By separating the tools on the program-level, the boundaries are
4013 clearly visible and technically enforced.
4014 The interfaces are calls to
4015 .Fu exec()
4016 rather than arbitrary function calls.
4017 .P
4018 But the real problem is another:
4019 Nmh violates the golden ``one tool, one job'' rule of the Unix philosophy.
4020 Understanding
4021 .Pn comp
4022 requires understanding
4023 .Fn uip/annosbr.c
4024 and
4025 .Fn uip/sendsbr.c
4026 because
4027 .Pn comp
4028 does annotate and send messages.
4029 In nmh, there surely exists the tool
4030 .Pn send ,
4031 which does (almost) only send messages.
4032 But
4033 .Pn comp
4034 and
4035 .Pn repl
4036 and
4037 .Pn forw
4038 and
4039 .Pn dist
4040 and
4041 .Pn whatnow
4042 and
4043 .Pn viamail ,
4044 they all (!) have the same message sending function included, too.
4045 In result,
4046 .Pn comp
4047 sends messages without using
4048 .Pn send .
4049 The situation is the same as if
4050 .Pn grep
4051 would page without
4052 .Pn more
4053 just because both programs are part of the same code base.
4054 .P
4055 The clear separation on the surface \(en the toolchest approach \(en
4056 is violated on the level below.
4057 This violation is for the sake of time performance.
4058 On systems where
4059 .Fu fork()
4060 and
4061 .Fu exec()
4062 are expensive, the quicker response might be noticable.
4063 In the old times, sacrificing readability and conceptional beauty for
4064 speed might even have been a must to prevent MH from being unusably slow.
4065 Whatever the reasons had been, today they are gone.
4066 No longer should we sacrifice readability or conceptional beauty.
4067 No longer should we violate the Unix philosophy's ``one tool, one job''
4068 guideline.
4069 No longer should we keep speed improvements that became unnecessary.
4070 .P
4071 Therefore, mmh's
4072 .Pn comp
4073 does no longer send messages.
4074 In mmh, different jobs are divided among separate programs that
4075 invoke each other as needed.
4076 In consequence,
4077 .Pn comp
4078 invokes
4079 .Pn whatnow
4080 which thereafter invokes
4081 .Pn send .
4082 .Ci 3df5ab3c116e6d4a2fb4bb5cc9dfc5f781825815
4083 .Ci c73c00bfccd22ec77e9593f47462aeca4a8cd9c0
4084 The clear separation on the surface is maintained on the level below.
4085 Human users and the tools use the same interface \(en
4086 annotations, for example, are made by invoking
4087 .Pn anno ,
4088 no matter if requested by programs or by human beings.
4089 .Ci 469a4163c2a1a43731d412eaa5d9cae7d670c48b
4090 .Ci aed384169af5204b8002d06e7a22f89197963d2d
4091 .Ci 3caf9e298a8861729ca8b8a84f57022b6f3ea742
4092 The decrease of tools built from multiple source files and thus
4093 the decrease of
4094 .Fn uip/*sbr.c
4095 files confirm the improvement.
4096 .Ci 9e6d91313f01c96b4058d6bf419a8ca9a207bc33
4097 .ci 81744a46ac9f845d6c2b9908074d269275178d2e
4098 .Ci f0f858069d21111f0dbea510044593f89c9b0829
4099 .Ci 0503a6e9be34f24858b55b555a5c948182b9f24b
4100 .Ci 27826f9353e0f0b04590b7d0f8f83e60462b90f0
4101 .Ci d1da1f94ce62160aebb30df4063ccbc53768656b
4102 .Ci c42222869e318fff5dec395eca3e776db3075455
4103 .P
4104 .\" XXX move this paragraph up somewhere
4105 One disadvantage needs to be taken with this change:
4106 The compiler can no longer check the integrity of the interfaces.
4107 By changing the command line interfaces of tools, it is
4108 the developer's job to adjust the invocations of these tools as well.
4109 As this is a manual task and regression tests, which could detect such
4110 problems, are not available yet, it is prone to errors.
4111 These errors will not be detected at compile time but at run time.
4112 Installing regression tests is a task left to do.
4113 In the best case, a uniform way of invoking tools from other tools
4114 can be developed to allow automated testing at compile time.
4117 .ig
4118 XXX consider writing about mhl vs. mhlproc
4120 sbr/showfile.c
4122 23 /*
4123 24 ** If you have your lproc listed as "mhl",
4124 25 ** then really invoked the mhlproc instead
4125 26 ** (which is usually mhl anyway).
4126 27 */
4128 Sat Nov 24 19:09:14 1984 /mtr (agent: Marshall Rose) <uci@udel-dewey>
4130 sbr/showfile.c: if lproc is "mhl", use mhlproc for consistency
4131 (Actually, user should use "lproc: show", "showproc: mhl".)
4132 ..