Re: /srv/bzr/emacs/trunk r105648: Make sendmail-query-once update send-mail-function directly.

Re: /srv/bzr/emacs/trunk r105648: Make sendmail-query-once update send-mail-function directly.

[Lars Magne Ingebrigtsen]

Chong Yidong <cyd@stupidchicken.com> writes:

>   Make sendmail-query-once update send-mail-function directly.

The reason for using the second variable was, as previously discussed
and agreed-to on the mailing list, that after using the new sendmail.el,
all Emacs users will have `sendmail-send-function' set.  This makes it
impossible to change the default at a later date, if we find out that we
should default to something else sometime in the future.  (Which is not
unlikely -- mail delivery methods change.)

Please revert.

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

Re: /srv/bzr/emacs/trunk r105648: Make sendmail-query-once update send-mail-function directly.

[Chong Yidong]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

>>   Make sendmail-query-once update send-mail-function directly.
>
> The reason for using the second variable was, as previously discussed
> and agreed-to on the mailing list, that after using the new
> sendmail.el, all Emacs users will have `sendmail-send-function' set.
> This makes it impossible to change the default at a later date, if we
> find out that we should default to something else sometime in the
> future.  (Which is not unlikely -- mail delivery methods change.)

One problem with sendmail-query-once-function is that it's not a
defcustom, so it's inappropriate to save it via customize-save-variable.

Making it a defcustom is not a good solution either.  The problem of
forward compatibility doesn't seem serious enough to justify yet another
layer on top of send-mail-function.

If, in the future, we come up with another method to deliver mail, the
fact that some users have set send-mail-function via sendmail-query-once
should not bother us any more than those who've customized it directly.
So it is better to just do the straightforward thing.

Re: /srv/bzr/emacs/trunk r105648: Make sendmail-query-once update send-mail-function directly.

[Lars Magne Ingebrigtsen]

Chong Yidong <cyd@stupidchicken.com> writes:

> If, in the future, we come up with another method to deliver mail, the
> fact that some users have set send-mail-function via sendmail-query-once
> should not bother us any more than those who've customized it directly.
> So it is better to just do the straightforward thing.

The difference is that we're forcing the users to a single value now
that we can't ever change again.  So if, for instance, the Linux people
pull themselves together and create a proper API for querying how mail
is supposed to be sent out, this is something that few Emacs users will
be using, since if we change the default in Emacs 26.1 to
`send-mail-using-the-clever-new-linux-api', they'll never see it.

-- 
(domestic pets only, the antidote for overdose, milk.)
  bloggy blog http://lars.ingebrigtsen.no/

Re: /srv/bzr/emacs/trunk r105648: Make sendmail-query-once update send-mail-function directly.

[Richard Stallman]

    The difference is that we're forcing the users to a single value now
    that we can't ever change again.  So if, for instance, the Linux people
    pull themselves together and create a proper API for querying how mail
    is supposed to be sent out,

Linux has nothing to do with this issue.  Such a convention would
not be implemented in a kernel.

Are you talking about the GNU system and calling it "Linux"?


-- 
Dr Richard Stallman
President, Free Software Foundation
51 Franklin St
Boston MA 02110
USA
www.fsf.org  www.gnu.org
Skype: No way! That's nonfree (freedom-denying) software.
  Use free telephony http://directory.fsf.org/category/tel/

Re: /srv/bzr/emacs/trunk r105648: Make sendmail-query-once update send-mail-function directly.

[Chong Yidong]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> The difference is that we're forcing the users to a single value now
> that we can't ever change again.  So if, for instance, the Linux
> people pull themselves together and create a proper API for querying
> how mail is supposed to be sent out, this is something that few Emacs
> users will be using, since if we change the default in Emacs 26.1 to
> `send-mail-using-the-clever-new-linux-api', they'll never see it.

Users who have customized send-mail-function independently, prior to
Emacs 24, would encounter the same problem.  In the event, it woud be
more reliable to hard code a check into the mail-sending functions,
similar to what we did in compose-mail when switching from mail-mode to
message-mode as the default mail user agent.