I think what people are looking for is a positive reason to junk
sendmail.  Yes: "The sendmail developers decided to sell sendmail to
Sun".  No: "Sendmail sucks and has always sucked and I hate it I hate it
die".

Someone offered the fact that sendmail has moved to the GPL license as
justification for it's eventual removal from the base distribution.  Well,
there is no prospect of the entire /usr/src/gnusrc tree being removed, so
I believe that's a red herring.

It seems someone has decided NetBSD is going to change SMTP MTAs from
something to accomplish the goal of a better engineered system.  It seems
also that NetBSD has accepted the fact that it will thusly burn the people
who require the standard Unix MTA to be in the base distribution.  If
anyone really believes that sendmail isn't the standard unix MTA
historically and across the majority of unix-type operating systems, go
ahead.

When I use the mailer that comes in the base operating system, I am
basically setting up a dumb node.  If I am building an SMTP server that
requires the functionality of a Postfix or Exim, then I am going to be
compiling the software anyhow, *even if it is sendmail*.  A mail server of
any kind will require tweaking, just like apache or mysql, so let's
acknowledge the fact.

Therefore, the logical choice for me is to have a no-brainer SMTP daemon
that fulfills the requirements of, say, a dumb desktop user, and to have
any mailer that fulfills the extensive needs of a server to be in ports.  
I am trying to imagine a scenario where someone would be badly served by
that approach, and I can't think of one.

"If it ain't broken, don't fix it" is a great rule.  If NetBSD is going to
change that, then I think it should opt for the least-common denominator
of MTAs, whether that be mail.local or nbsmtp or whatever.  It certaintly
should not go from one cover-all-possible-situations MTA that's crufty to
another cover-all-situations MTA that is less crufty, because yhat
achieves no real advancement of modularity in the OS.

--N.

P.S.  What mailing list should I have been subscribed to to watch the
discussions of this befor they occurred?  It obviously isn't
current-users, and I have no desire to sibscribe to tech-*, so if someone
could point out the proper list for major changes of direction to be
mooted before they occur I'd be happy to lurk on it.

P.P.S. when did the statements in this email become deprecated, if at all?

--- begin included message ---
Subject: Re: sendmail and netbsd
To: Andrew Brown <atatat@atatdot.net>
From: Thor Lancelot Simon <tls@rek.tjls.com>
List: current-users
Date: 05/01/2000 00:37:09 

On Mon, May 01, 2000 at 12:21:45AM -0400, Andrew Brown wrote:
> >It *is* allowed in the tree.
> >
> >It's just in gnusrc because it has a different license now which is,
> >from the perspective of NetBSD, roughly equivalent to the GPL.
> 
> i looked a little more gpl-ish to me, as opposed to the bsd licence it
> used to have.
> 
> so how come we jumped from 8.8.8 to 8.9.3 with no intervening
> integrations?  and what objections might others have to sendmail being
> in a release tree?

I think sendmail's a piece of crap.  It's ancient, it's been fixed with
bailing wire and prayer about a thousand times too many, its performance
is awful, its configuration syntax is obtuse, and the only reason I know
of that most people even considered using it as long as they did was that
there was no reasonable alternative.

Postfix looks like a pretty darned reasonable alternative to me, but
unfortunately while we failed to include it in our OS, and thus lost
whatever leverage we might have had, its license became unacceptable,
too.

What it boils down to is that there's *no* acceptably-licensed, modern
mailer available.  Given that sendmail's license is among the most
objectionable, and its quality is so low, I wish we would ship just
about anything *else*...
--- end included message ---