Eliminating mail-transport-agent from standard
In addition to determining the MTA pulled in by default for packages
which require mail-transport-agent in order to function (the provider of
default-mta), I'd like to propose as a release goal that we not have any
MTA in standard anymore. I've actually worked towards this goal for a
while now, and made a fair bit of progress; this mail documents the
remaining work required, most of which is simple dependency/priority
changes and patch application.
Only one package in standard or above currently Depends on a
mail-transport-agent:
- bsd-mailx: should have the same priority as an MTA (optional).
Two packages in standard or above Recommends bsd-mailx (indirectly via
the mailx virtual package):
- exim4-base: moved to optional as part of this goal.
- logrotate; could just Suggests mailx (or use sendmail directly and
Suggests an MTA).
Four packages in standard or above Recommends mail-transport-agent:
- cron: I've already filed bugs on cron (and anacron) with patches to
support sending cronjob output to syslog, so that it will not
disappear into /dev/null without an MTA installed.
- at: I'd argue for this becoming priority optional, though it wouldn't
be particularly hard to write a patch like the one for cron instead.
- procmail: should have the same priority as an MTA (optional)
- mutt: can easily Suggests a mail-transport-agent, since it supports
IMAP and SMTP, leaving aside more exotic configurations like
getmail/fetchmail. (That leaves aside the question of whether mutt
should be standard or optional, but I think either way it should only
Suggests an MTA.)
With the above changes made, all providers of mail-transport-agent could
become priority optional or lower, including the provider of
default-mta.
(To the extent this affects the selection of default-mta, I'd suggest
that it might argue in favor of making default-mta one that only
supports smarthosts and does not queue locally, leaving the choice of
what MTA to run on a mail server up to the end user, but that question
matters a lot less to me than removing MTAs from standard.)
- Josh Triplett
Reply to: