[Mailman-Developers] chunkify suggestion, with patch.

[Darrell Fuhriman]

> I don't think there's time to get this into 2.0 final.  I'd like to
> get more feedback from people on whether this works and is a good
> idea.  

well, i'm using it, and it seems to work.  ;)

> My own intuition says that Mailman should do as little work as
> possible here.  I think a good MTA can always do a better job than
> Mailman at deciding the best delivery strategy.  Ideally, you'd pass
> the whole recip list off to the MTA in one huge chunk.

Yah, but the world isn't ideal.  :)  sendmail certainly doesn't cope very
well.

As an example, assume we have list with 5 recipients, in this order:

a@foo.com
b@bar.com
c@foo.com
d@baz.com
e@xyzzy.com

now, let us further assume that foo.com is not reachable.

when sendmail tries to connect, it will take 5 mins. (default) to decide
that foo.com is down.  that means that bar.com, baz.com and xyzzy.com all
have to wait that five minutes before sendmail attempts to deliver to
them.

Now, if the list is sorted on the domain name, bar.com and baz.com get
right through and only xyzzy.com has to wait the five minutes.

Now imagine there's a list with 500 recipients (default SMTP_MAX_RCPTS),
which in my experience means there will be at least 2 or 3 domains
temporarily unreachable.  That means poor little xyzzy.com has to wait
some 15 minutes for propogation.

By turning down SMTP_MAX_RCPTS we're decreasing the number of people who
will be punished by a host being down.

> I'm less inclined to coddle inefficient MTAs.  Still, this may be a
> good idea if it's properly configurable.

Well, I don't know how other MTAs will handle it, so I can't say how many
we'd be coddling, and I certainly have no interest in an MTA religious
war.  But the fact remains that there are an awful lot of people using
sendmail still.  :)  

I can work on adding it as an option, certainly.  And I would like to make
the modifications I mentioned in my original post.

> My suggestion at this point would be to upload the patch to
> SourceForge for future consideration, and so that others can play with
> it.

done!

d.