Differences between mail-mode and message-mode (was: message-mode is now default?)

Differences between mail-mode and message-mode (was: message-mode is now default?)

[Reiner Steib]

On Sat, Jul 18 2009, Eli Zaretskii wrote:

> I had a few spare moments, so I compared mail-mode and message-mode.
> The differences I found are below.
>
> Disclaimers:
>
>     Please note that I don't necessarily think all of them should be
>     changed to match mail-mode, in order for message-mode to be
>     compatible.  But I do think they all need to be _considered_.

Thanks for your analysis.  No matter in which direction a change will
go, this list will be a good basis for NEWS entries (either for
mail-mode or message-mode users).

>     I also don't pretend to have found all the important differences,
>     nor that all my conclusions below are correct.

>     Finally, what's below is based on inspection of the *Help* buffer
>     and defcustoms; I didn't actually try using message-mode, and
>     don't know anything about it, except what I saw just now.
>
> Here's what I found:
[ key bindings ]

Obviously the key bindings cannot be resolved without a change for
either mail-mode or message-mode users.  Maybe set the bindings
depending on some compatibility variable?

>  . In mail-mode C-c C-f C-f moves to FCC; in message-mode C-c C-f C-f
>    moves to Followup-To, and C-c C-f C-w moves to FCC.
>
>  . In mail-mode C-c C-f C-a moves to Mail-Reply-To; in message-mode
>    C-c C-f C-a runs the command
>    `message-generate-unsubscribed-mail-followup-to' and there doesn't
>    seem to be a command to move to Mail-Reply-To.
>
>  . In mail-mode C-c C-f C-l moves to Mail-Followup-To; in message-mode
>    C-c C-f C-m moves to Mail-Followup-To and C-c C-f C-l is undefined
>
>  . In mail-mode C-c C-t moves to message text; in message-mode this is
>    on C-c C-b, while C-c C-t runs `message-insert-to'.
>
>  . In mail-mode C-c C-v runs `mail-sent-via'; in message-mode it runs
>    `message-delete-not-region'.

What is "Sent-Via" about?  Neither Google nor grepping through the
RFCs gave me any useful result. [1]

>  . mail-mode runs `text-mode-hook' and `mail-mode-hook'; message-mode
>    says it runs `message-mode-hook' "in addition to any hooks its
>    parent mode `text-mode' might have run" -- it isn't clear to me if
>    these two are compatible wrt what text-mode does.

This text is from `derived-mode-make-docstring'.  message-mode is
derived from text-mode.

>  . message-mode overrides the following global key bindings:
>
>    - C-a runs `message-beginning-of-line'

This is somewhat similar to C-a in the minibuffer.  If you want to
move beyond the mail header (e.g. "Subject: "), just press C-a again.

>    - TAB runs `message-tab'

`message-tab' expand partially typed headers,
cf. `message-completion-alist'.

>    - M-; runs `comment-region'

AFAICS, in mail-mode, `M-;' runs `comment-dwim', which first asks for
a comment syntax since none is defined.

>  . message-mode's default value of `message-generate-new-buffers'
>    seems to generate buffer names in a way that is different from what
>    mail-mode does.

AFAICS, if I do `M-x message-mail RET', I get also get a "*mail*"
buffer.

>  . message-mode's default value of `message-fill-column' overrides the
>    global value of `fill-column'.

I used the recommendation from the relevant RFCs or GNKSA docs, IIRC.
Unlike `fill-column' which affects the user (and his co-workers, ...),
the fill column in mails (and news) should try to make sense for the
recipients.

>  . mail-mode uses `mail-setup-hook'; message-mode uses
>    `message-setup-hook'.

On Sat, Jul 18 2009, Miles Bader wrote:
| One issue, of course is hook names; would it work to simply have
| message-mode use the mail-mode hooks directly, and use variable-aliasing
| to accommodate people that use the message-foo-hook names?

Miles, is `defvaralias' what you have in mind?

  @@ -905,12 +923,13 @@
   (defcustom message-setup-hook nil
     "Normal hook, run each time a new outgoing message is initialized.
   The function `message-setup' runs this hook."
  [...] 
  +(defvaralias 'mail-setup-hook 'message-setup-hook)
  [...] 

>  . mail-mode uses `mail-yank-hooks'; message-mode doesn't appear to
>    have such a facility.

,----
| ;; FIXME make it really obsolete.
| (defvar mail-yank-hooks nil
|   "Obsolete hook for modifying a citation just inserted in the mail buffer.
| Each hook function can find the citation between (point) and (mark t).
| And each hook function should leave point and mark around the citation
| text as modified.
| 
| This is a normal hook, misnamed for historical reasons.
| It is semi-obsolete and mail agents should no longer use it.")
`----

Should we care about it if is obsolete?

>  . message-mode's default value of `message-citation-line-format'
>    seems to produce different results compared to what `r' in Rmail
>    produces.

AFAICS, `r' doesn't cite at all and produce not citation line (neither
does `C-c C-y' (`mail-yank-original').

>  . message-mode's default values of `message-yank-cited-prefix' and
>    `message-yank-empty-prefix' are different from the results produced
>    by the default value of `mail-yank-prefix', and the names of the
>    options are also different.

You mean the indentation?  I think message mode's default are much
better.

>  . message-mode's default value of `message-signature' is different
>    from mail-mode's `mail-signature', 

I think using ~/.signature by default is reasonable.

>    and the names of the options are different.

See above -> Alias?

>  . mail-mode uses `mail-signature-file'; message-mode uses
>    `message-signature-file'.

See above -> Alias?

>  . mail-mode uses `mail-default-directory' for auto-saving mail
>    buffers; message-mode uses `message-auto-save-directory', 

See above -> Alias?

>    and the default values are also different.

message-mode tries to write all it's files under ~/Mail .  Many users
like to sync their mail stuff between different computers.
message-mode's default makes this easier.

>  . mail-mode uses `mail-mode-syntax-table'; message-mode uses
>    `message-mode-syntax-table.  The default values are also
>    different.

I don't know why message mode additionally contains `>' and `<'.
message-mode also copies `text-mode-syntax-table'.

>  . mail-mode uses `mail-send-hook'; message-mode uses
>    `message-send-mail-hook' and `message-send-hook'.

See above -> Alias?

>  . mail-mode's `mail-bury' tries to show the Rmail buffer if
>    `mail-bury-selects-summary' is non-nil.  message-mode's
>    `message-bury' doesn't seem to have a comparable feature.  (This is
>    important when reading mail on a text terminal.)

Probably we should call code like `mail-bury' to `message-bury' if
`rmail-summary-buffer' is non-nil.

>  . message-mode doesn't seem to have an equivalent of `mail-use-dsn'.

We could add the `mail-use-dsn' code to `message-send-mail-with-sendmail'.

Bye, Reiner.

[1] http://www.google.com/search?q=%2B"Sent-Via"+mail+header+rfc

    $ dpkg -l 'doc-rfc-*'
    [...]
    ii  doc-rfc-0001-0999        20030621-1               Other RFCs
    ii  doc-rfc-1000-1999        20030621-1               Other RFCs
    ii  doc-rfc-2000-2999        20030621-1               Other RFCs
    ii  doc-rfc-3000-3999        20030621-1               Other RFCs
    ii  doc-rfc-experimental     20030621-1               Experimental RFCs
    ii  doc-rfc-fyi-bcp          20030621-1               FYI and BCP RFCs
    ii  doc-rfc-misc             20030621-1               Miscellaneous RFCs
    ii  doc-rfc-old-std          20030621-1               Old Standard RFCs
    ii  doc-rfc-std              20030621-1               Standard RFCs
    ii  doc-rfc-std-proposed     20030621-1               Proposed Standard RFCs
    
    /usr/share/doc/RFC$ zgrep -ie Sent-Via */*.txt.gz
    /usr/share/doc/RFC$ 
-- 
       ,,,
      (o o)
---ooO-(_)-Ooo---  |  PGP key available  |  http://rsteib.home.pages.de/

Differences between mail-mode and message-mode - sending mail (was: message-mode is now default?)

[Reiner Steib]

On Sun, Jul 19 2009, Alfred M. Szmidt wrote:

> message-mode should atleast follow the standard conventions for
> hooking in different methods for sending mail, it doesn't.  

Please describe what these "standard conventions" are and what doesn't
work (for you).

> That way, it will work for those who prefer mail-mode, and those who
> prefer message-mode.  Otherwise, it will break things as I reported
> before.

Bye, Reiner.
-- 
       ,,,
      (o o)
---ooO-(_)-Ooo---  |  PGP key available  |  http://rsteib.home.pages.de/

Re: Differences between mail-mode and message-mode

[Sven Joachim]

On 2009-07-19 13:41 +0200, Reiner Steib wrote:

> On Sat, Jul 18 2009, Eli Zaretskii wrote:

>>  . In mail-mode C-c C-v runs `mail-sent-via'; in message-mode it runs
>>    `message-delete-not-region'.
>
> What is "Sent-Via" about?  Neither Google nor grepping through the
> RFCs gave me any useful result. [1]

See also bug #1776¹ about that header.  Probably only the long-term users
of mail-mode know the answer, if anyone.

Sven


¹ http://emacsbugs.donarmstrong.com/cgi-bin/bugreport.cgi?bug=1776

Re: Differences between mail-mode and message-mode

[Eli Zaretskii]

> From: Reiner Steib <reinersteib+gmane@imap.cc>
> Cc: emacs-devel@gnu.org, ding@gnus.org
> Date: Sun, 19 Jul 2009 13:41:32 +0200
> 
> >  . mail-mode runs `text-mode-hook' and `mail-mode-hook'; message-mode
> >    says it runs `message-mode-hook' "in addition to any hooks its
> >    parent mode `text-mode' might have run" -- it isn't clear to me if
> >    these two are compatible wrt what text-mode does.
> 
> This text is from `derived-mode-make-docstring'.  message-mode is
> derived from text-mode.

So is mail-mode, but the blurb is somehow still different, I don't
know why.

> >  . mail-mode uses `mail-yank-hooks'; message-mode doesn't appear to
> >    have such a facility.
> 
> ,----
> | ;; FIXME make it really obsolete.
> | (defvar mail-yank-hooks nil
> |   "Obsolete hook for modifying a citation just inserted in the mail buffer.
> | Each hook function can find the citation between (point) and (mark t).
> | And each hook function should leave point and mark around the citation
> | text as modified.
> | 
> | This is a normal hook, misnamed for historical reasons.
> | It is semi-obsolete and mail agents should no longer use it.")
> `----
> 
> Should we care about it if is obsolete?

It depends whether people are using it or not, IMO.  And we usually
declare a feature obsolete formally, and keep it that way for some
time, before removing it.

> >  . message-mode's default value of `message-citation-line-format'
> >    seems to produce different results compared to what `r' in Rmail
> >    produces.
> 
> AFAICS, `r' doesn't cite at all and produce not citation line (neither
> does `C-c C-y' (`mail-yank-original').

I mean the "John Doe writes" stuff vs the "From: ..." etc. produced by
Rmail's `r'.

Re: Differences between mail-mode and message-mode

[Eli Zaretskii]

> Date: Sun, 19 Jul 2009 20:51:44 +0300
> From: Eli Zaretskii <eliz@gnu.org>
> Cc: ding@gnus.org, emacs-devel@gnu.org
> 
> > >  . message-mode's default value of `message-citation-line-format'
> > >    seems to produce different results compared to what `r' in Rmail
> > >    produces.
> > 
> > AFAICS, `r' doesn't cite at all and produce not citation line (neither
> > does `C-c C-y' (`mail-yank-original').
> 
> I mean the "John Doe writes" stuff vs the "From: ..." etc. produced by
> Rmail's `r'.

Sorry, you are right, of course: It's C-c C-y that produces "From:...".

Re: message-mode is now default?

[Reiner Steib]

On Mon, Jul 20 2009, Chong Yidong wrote:

> Miles Bader <miles@gnu.org> writes:
>
>> In the most important senses -- the general UI and method of operation
>> -- message-mode is very much compatible... There are clearly things
>> that need to be ironed out, so we should iron them out.
>>
>> The hook situation seems one of the more important issues; is anybody
>> going to comment on my suggestion for merging the hooks?
>
> How troublesome would it be to make message-mode derived from mail-mode?
> This would solve the hooks problem in a clean way.  

Wouldn't this imply that message-mode depends on mail-mode so we would
keep the (partial) code duplication?

> It might also bring other minor conveniences, such as letting us
> merge their syntax tables and/or font-lock keywords.

Bye, Reiner.
-- 
       ,,,
      (o o)
---ooO-(_)-Ooo---  |  PGP key available  |  http://rsteib.home.pages.de/

Differences between mail-mode and message-mode (was: message-mode is now default?)

[Stephen J. Turnbull]

Reiner Steib writes:

 > Obviously the key bindings cannot be resolved without a change for
 > either mail-mode or message-mode users.  Maybe set the bindings
 > depending on some compatibility variable?

How about the presence of Newsgroups?

 > >  . In mail-mode C-c C-f C-f moves to FCC; in message-mode C-c C-f C-f
 > >    moves to Followup-To, and C-c C-f C-w moves to FCC.

Followup-To is rarely useful in a mail message.

 > >  . In mail-mode C-c C-f C-a moves to Mail-Reply-To; in message-mode
 > >    C-c C-f C-a runs the command
 > >    `message-generate-unsubscribed-mail-followup-to' and there doesn't
 > >    seem to be a command to move to Mail-Reply-To.

There is no Mail-Reply-To header in the mail standards AFAIK.

 > >  . In mail-mode C-c C-f C-l moves to Mail-Followup-To; in message-mode
 > >    C-c C-f C-m moves to Mail-Followup-To and C-c C-f C-l is undefined

There is no Mail-Followup-To header in the mail standards AFAIK,
although there is a plausible argument that it is useful for mailing
lists.

 > >  . In mail-mode C-c C-t moves to message text; in message-mode this is
 > >    on C-c C-b, while C-c C-t runs `message-insert-to'.
 > >
 > >  . In mail-mode C-c C-v runs `mail-sent-via'; in message-mode it runs
 > >    `message-delete-not-region'.
 > 
 > What is "Sent-Via" about?  Neither Google nor grepping through the
 > RFCs gave me any useful result. [1]
 > 
 > >  . mail-mode runs `text-mode-hook' and `mail-mode-hook'; message-mode
 > >    says it runs `message-mode-hook' "in addition to any hooks its
 > >    parent mode `text-mode' might have run" -- it isn't clear to me if
 > >    these two are compatible wrt what text-mode does.
 > 
 > This text is from `derived-mode-make-docstring'.  message-mode is
 > derived from text-mode.
 > 
 > >  . message-mode overrides the following global key bindings:
 > >
 > >    - C-a runs `message-beginning-of-line'
 > 
 > This is somewhat similar to C-a in the minibuffer.  If you want to
 > move beyond the mail header (e.g. "Subject: "), just press C-a again.
 > 
 > >    - TAB runs `message-tab'
 > 
 > `message-tab' expand partially typed headers,
 > cf. `message-completion-alist'.
 > 
 > >    - M-; runs `comment-region'
 > 
 > AFAICS, in mail-mode, `M-;' runs `comment-dwim', which first asks for
 > a comment syntax since none is defined.
 > 
 > >  . message-mode's default value of `message-generate-new-buffers'
 > >    seems to generate buffer names in a way that is different from what
 > >    mail-mode does.
 > 
 > AFAICS, if I do `M-x message-mail RET', I get also get a "*mail*"
 > buffer.
 > 
 > >  . message-mode's default value of `message-fill-column' overrides the
 > >    global value of `fill-column'.
 > 
 > I used the recommendation from the relevant RFCs or GNKSA docs, IIRC.
 > Unlike `fill-column' which affects the user (and his co-workers, ...),
 > the fill column in mails (and news) should try to make sense for the
 > recipients.
 > 
 > >  . mail-mode uses `mail-setup-hook'; message-mode uses
 > >    `message-setup-hook'.
 > 
 > On Sat, Jul 18 2009, Miles Bader wrote:
 > | One issue, of course is hook names; would it work to simply have
 > | message-mode use the mail-mode hooks directly, and use variable-aliasing
 > | to accommodate people that use the message-foo-hook names?
 > 
 > Miles, is `defvaralias' what you have in mind?
 > 
 >   @@ -905,12 +923,13 @@
 >    (defcustom message-setup-hook nil
 >      "Normal hook, run each time a new outgoing message is initialized.
 >    The function `message-setup' runs this hook."
 >   [...] 
 >   +(defvaralias 'mail-setup-hook 'message-setup-hook)
 >   [...] 
 > 
 > >  . mail-mode uses `mail-yank-hooks'; message-mode doesn't appear to
 > >    have such a facility.
 > 
 > ,----
 > | ;; FIXME make it really obsolete.
 > | (defvar mail-yank-hooks nil
 > |   "Obsolete hook for modifying a citation just inserted in the mail buffer.
 > | Each hook function can find the citation between (point) and (mark t).
 > | And each hook function should leave point and mark around the citation
 > | text as modified.
 > | 
 > | This is a normal hook, misnamed for historical reasons.
 > | It is semi-obsolete and mail agents should no longer use it.")
 > `----
 > 
 > Should we care about it if is obsolete?
 > 
 > >  . message-mode's default value of `message-citation-line-format'
 > >    seems to produce different results compared to what `r' in Rmail
 > >    produces.
 > 
 > AFAICS, `r' doesn't cite at all and produce not citation line (neither
 > does `C-c C-y' (`mail-yank-original').
 > 
 > >  . message-mode's default values of `message-yank-cited-prefix' and
 > >    `message-yank-empty-prefix' are different from the results produced
 > >    by the default value of `mail-yank-prefix', and the names of the
 > >    options are also different.
 > 
 > You mean the indentation?  I think message mode's default are much
 > better.
 > 
 > >  . message-mode's default value of `message-signature' is different
 > >    from mail-mode's `mail-signature', 
 > 
 > I think using ~/.signature by default is reasonable.
 > 
 > >    and the names of the options are different.
 > 
 > See above -> Alias?
 > 
 > >  . mail-mode uses `mail-signature-file'; message-mode uses
 > >    `message-signature-file'.
 > 
 > See above -> Alias?
 > 
 > >  . mail-mode uses `mail-default-directory' for auto-saving mail
 > >    buffers; message-mode uses `message-auto-save-directory', 
 > 
 > See above -> Alias?
 > 
 > >    and the default values are also different.
 > 
 > message-mode tries to write all it's files under ~/Mail .  Many users
 > like to sync their mail stuff between different computers.
 > message-mode's default makes this easier.
 > 
 > >  . mail-mode uses `mail-mode-syntax-table'; message-mode uses
 > >    `message-mode-syntax-table.  The default values are also
 > >    different.
 > 
 > I don't know why message mode additionally contains `>' and `<'.
 > message-mode also copies `text-mode-syntax-table'.
 > 
 > >  . mail-mode uses `mail-send-hook'; message-mode uses
 > >    `message-send-mail-hook' and `message-send-hook'.
 > 
 > See above -> Alias?
 > 
 > >  . mail-mode's `mail-bury' tries to show the Rmail buffer if
 > >    `mail-bury-selects-summary' is non-nil.  message-mode's
 > >    `message-bury' doesn't seem to have a comparable feature.  (This is
 > >    important when reading mail on a text terminal.)
 > 
 > Probably we should call code like `mail-bury' to `message-bury' if
 > `rmail-summary-buffer' is non-nil.
 > 
 > >  . message-mode doesn't seem to have an equivalent of `mail-use-dsn'.
 > 
 > We could add the `mail-use-dsn' code to `message-send-mail-with-sendmail'.
 > 
 > Bye, Reiner.
 > 
 > [1] http://www.google.com/search?q=%2B"Sent-Via"+mail+header+rfc
 > 
 >     $ dpkg -l 'doc-rfc-*'
 >     [...]
 >     ii  doc-rfc-0001-0999        20030621-1               Other RFCs
 >     ii  doc-rfc-1000-1999        20030621-1               Other RFCs
 >     ii  doc-rfc-2000-2999        20030621-1               Other RFCs
 >     ii  doc-rfc-3000-3999        20030621-1               Other RFCs
 >     ii  doc-rfc-experimental     20030621-1               Experimental RFCs
 >     ii  doc-rfc-fyi-bcp          20030621-1               FYI and BCP RFCs
 >     ii  doc-rfc-misc             20030621-1               Miscellaneous RFCs
 >     ii  doc-rfc-old-std          20030621-1               Old Standard RFCs
 >     ii  doc-rfc-std              20030621-1               Standard RFCs
 >     ii  doc-rfc-std-proposed     20030621-1               Proposed Standard RFCs
 >     
 >     /usr/share/doc/RFC$ zgrep -ie Sent-Via */*.txt.gz
 >     /usr/share/doc/RFC$ 
 > -- 
 >        ,,,
 >       (o o)
 > ---ooO-(_)-Ooo---  |  PGP key available  |  http://rsteib.home.pages.de/
 > 

Re: Differences between mail-mode and message-mode

[Reiner Steib]

On Wed, Jul 22 2009, Stephen J. Turnbull wrote:

> Reiner Steib writes:
>
>  > Obviously the key bindings cannot be resolved without a change for
>  > either mail-mode or message-mode users.  Maybe set the bindings
>  > depending on some compatibility variable?
>
> How about the presence of Newsgroups?

>  > >  . In mail-mode C-c C-f C-f moves to FCC; in message-mode C-c C-f C-f
>  > >    moves to Followup-To, and C-c C-f C-w moves to FCC.
>
> Followup-To is rarely useful in a mail message.

A message-mode buffer can be news, mail or both.  E.g. when adding Cc
to a news posting, the buffer also becomes a mail buffer (see
`message-news-p' and `message-mail-p').  I think modifying key
bindings depending on `message-news-p'/`message-mail-p' would be
confusing.

>  > >  . In mail-mode C-c C-f C-a moves to Mail-Reply-To; in message-mode
> There is no Mail-Reply-To header in the mail standards AFAIK.
>  > >  . In mail-mode C-c C-f C-l moves to Mail-Followup-To; in message-mode
>  > >    C-c C-f C-m moves to Mail-Followup-To and C-c C-f C-l is undefined
> There is no Mail-Followup-To header in the mail standards AFAIK,
> although there is a plausible argument that it is useful for mailing
> lists.

AFAIK, both aren't in any RFC.  But at least Mail-Followup-To is quite
useful (and also supported by other MUA like mutt).

Bye, Reiner.
-- 
       ,,,
      (o o)
---ooO-(_)-Ooo---  |  PGP key available  |  http://rsteib.home.pages.de/

Re: Differences between mail-mode and message-mode

[Stephen J. Turnbull]

Reiner Steib writes:
 > On Wed, Jul 22 2009, Stephen J. Turnbull wrote:
 > 
 > > Reiner Steib writes:
 > >
 > >  > Obviously the key bindings cannot be resolved without a change for
 > >  > either mail-mode or message-mode users.  Maybe set the bindings
 > >  > depending on some compatibility variable?
 > >
 > > How about the presence of Newsgroups?
 > 
 > >  > >  . In mail-mode C-c C-f C-f moves to FCC; in message-mode C-c C-f C-f
 > >  > >    moves to Followup-To, and C-c C-f C-w moves to FCC.
 > >
 > > Followup-To is rarely useful in a mail message.
 > 
 > A message-mode buffer can be news, mail or both.

Indeed.  That's why I mention "Newsgroups" above.  I'm conditioning
all my suggestions on "compatibility with the mail-only MUA that has
been default until now".  Ie, they apply only when it's *not* a news
post.  (As far as Emacs can tell so far: of course that can change if
the user desires.)

Of course this is a hack.  I think it's a shame that Gnus chose to
cater to 1-keystroke savings rather than keep a regular, mnemonic
keymap for default, but any attempt to change the Gnus keymap now will
of course meet a response greatly resembling in all ways the response
of mail-mode advocates to suggestions of change.

Re: Differences between mail-mode and message-mode

[Miles Bader]

"Stephen J. Turnbull" <stephen@xemacs.org> writes:
> any attempt to change the Gnus keymap now will of course meet a
> response greatly resembling in all ways the response of mail-mode
> advocates to suggestions of change.

It would differ in one respect, at least, though:  there would be an
obvious and understandable reason for it...

I'm not sure there would be all that much of response though -- as a
long-time user of message-mode, I only use the barest fraction of its
bindings, and really wouldn't care much if they got re-arranged for
compatibility, or even just for sanity.

-miles

-- 
"... The revolution will be no re-run brothers; The revolution will be live."