Mailman 3 December 2001 - Mailman-Developers

Re: [Mailman-Developers] Re: way to minimize IO load with MTA supported VERP
by Dan Mick 06 Dec '01

06 Dec '01

> Okay, I agree, your way is better. Let's make it as simple as possible: > > HELO shrek.org > 250 smtp.aol.com Hello shrek.org [123.45.6.7], glad to see you. > MAIL FROM:news@shrek.org > 250 news(a)shrek.org... Sender ok > RCPT TO:ogre@aol.com > 250 ogre(a)aol.com... Recipient ok > VERP FROM:ogre-aol.com-1007679391-1f04c30dabd-news@shrek.org > 250 ogre-aol.com-1007679391-1f04c30dabd-news(a)shrek.org... Sender ok > RCPT TO:witch@aol.com > 250 witch(a)aol.com... Recipient ok > VERP FROM:witch-aol.com-1007679391-297c0dd13a3-news@shrek.org > 250 witch-aol.com-1007679391-297c0dd13a3-news(a)shrek.org... Sender ok > DATA The way it's already implemented in Postfix is: MAIL FROM: sender(a)sender.dom XVERP everything else is as normal, but the envelopes get VERPed.

1 0

Re: way to minimize IO load with MTA supported VERP
by Tom Neff 06 Dec '01

06 Dec '01

Peter W <peterw(a)usa.net> wrote: >> I guess I was a little afraid that MTA's would get lost matching up >> separately-issued RCPT TO: and OVRD commands that were supposed to >> function as logical pairs. > > Yes, that makes sense. But couldn't that be clarified in the new > RFC/draft? yes, it might even be obeyed :) but yes. > A couple problems I see with that: > > 1) MTAs would have to be 100% uniform in the way they constructed > VERP return paths, or apps like mailman would not be able to use them > > 2) this takes aways some flexibility from the sending MTA, which > currently has the flexibility to deal with VERP in arbitrarily > complex ways... Okay, I agree, your way is better. Let's make it as simple as possible: HELO shrek.org 250 smtp.aol.com Hello shrek.org [123.45.6.7], glad to see you. MAIL FROM:news@shrek.org 250 news(a)shrek.org... Sender ok RCPT TO:ogre@aol.com 250 ogre(a)aol.com... Recipient ok VERP FROM:ogre-aol.com-1007679391-1f04c30dabd-news@shrek.org 250 ogre-aol.com-1007679391-1f04c30dabd-news(a)shrek.org... Sender ok RCPT TO:witch@aol.com 250 witch(a)aol.com... Recipient ok VERP FROM:witch-aol.com-1007679391-297c0dd13a3-news@shrek.org 250 witch-aol.com-1007679391-297c0dd13a3-news(a)shrek.org... Sender ok DATA

1 0

Re: [Mailman-Developers] MM Bouncer
by J C Lawrence 06 Dec '01

06 Dec '01

On Thu, 6 Dec 2001 10:37:36 -0500 Jay R Ashworth <jra(a)baylink.com> wrote: > I don't see that there *is* any theoretical way to *keep* loads > down with VERP, by it's very nature. External IO, yes, disk IO, no. QMail allows you to initiate a single spool file with a list of addresses to VERP against, and from there every message is delivered individually. This cuts local disk IO, the ultimate constraining performance factor, drastically. -- J C Lawrence ---------(*) Satan, oscillate my metallic sonatas. claw(a)kanga.nu He lived as a devil, eh? http://www.kanga.nu/~claw/ Evil is a name of a foeman, as I live.

2 1

Re: [Mailman-Developers] FWIW: Postfix does VERP
by Dan Mick 06 Dec '01

06 Dec '01

> > At least as of the snapshot I'm running, 20010714, postfix does VERP just > > as you'd expect, merely by adding -V to the "sendmail" invocation. > > (It works with multiple addresses, too.) Hmmm; maybe it's time to > > play around with Sendmail.py again. > > > > There seems to be no way to invoke such behavior through an > > SMTP connection (not surprisingly). > > I lied: EHLO reports that XVERP is an optional command, and indeed > MAIL FROM: <from-addr> XVERP is legal and generates a VERPed address. > > Hmm. Maybe this SMTP extension is already "defined". Sure looks > like one could hack SMTPDirect.py to use it. > > Hmm, hmm, hmm. At least Courier and Postfix seem to implement this extension. http://courier.sourceforge.net/intro.html http://www.postfix.org continuing 'hmm'-ing.

1 0

Re: [Mailman-Developers] FWIW: Postfix does VERP
by Dan Mick 06 Dec '01

06 Dec '01

> At least as of the snapshot I'm running, 20010714, postfix does VERP just > as you'd expect, merely by adding -V to the "sendmail" invocation. > (It works with multiple addresses, too.) Hmmm; maybe it's time to > play around with Sendmail.py again. > > There seems to be no way to invoke such behavior through an > SMTP connection (not surprisingly). I lied: EHLO reports that XVERP is an optional command, and indeed MAIL FROM: <from-addr> XVERP is legal and generates a VERPed address. Hmm. Maybe this SMTP extension is already "defined". Sure looks like one could hack SMTPDirect.py to use it. Hmm, hmm, hmm.

1 0

FWIW: Postfix does VERP
by Dan Mick 06 Dec '01

06 Dec '01

At least as of the snapshot I'm running, 20010714, postfix does VERP just as you'd expect, merely by adding -V to the "sendmail" invocation. (It works with multiple addresses, too.) Hmmm; maybe it's time to play around with Sendmail.py again. There seems to be no way to invoke such behavior through an SMTP connection (not surprisingly).

1 0

Re: [Mailman-Developers] MM Bouncer
by J C Lawrence 06 Dec '01

06 Dec '01

On Thu, 6 Dec 2001 09:44:07 -0800 Peter C Norton <spacey-mailman(a)lenin.nu> wrote: > It also unloads the disk. The wire is rarely where you need to > save time. I've seen a single mailhost with over a million > messages queued empty out overnight over a t1. The biggest delays > had nothing to do with mail traffic over the wire. ObStats: Running Postfix on a mostly unloaded and minimally tuned box with a T3 tier II connection I've repeatedly seen sustained mail delivery rates of 1,500 messages per minute (fairly flat MX distribution; AOL/MSN/Hotman together are less than 3% of the base). I've also seen occasional peaks over 3,200 messages per minute, but I've never sustained that sort of transfer rate (not enough mail traffic). Not great, not too shabby. Now, would you mind NOT setting a personal Mail-FollowUp-To? -- J C Lawrence ---------(*) Satan, oscillate my metallic sonatas. claw(a)kanga.nu He lived as a devil, eh? http://www.kanga.nu/~claw/ Evil is a name of a foeman, as I live.

1 0

Re: [Mailman-Developers] MM Bouncer
by J C Lawrence 06 Dec '01

06 Dec '01

On Wed, 5 Dec 2001 23:02:01 -0800 Peter C Norton <spacey-mailman(a)lenin.nu> wrote: > I have a feeling that VERPs can be implemented in a way that they > can work on any MTA. The problem isn't making them work with any MTA -- that's actually fairly trivial. The problem is keeping IO loads on the host reasonable with VERP with any MTA which is quite difficult. -- J C Lawrence ---------(*) Satan, oscillate my metallic sonatas. claw(a)kanga.nu He lived as a devil, eh? http://www.kanga.nu/~claw/ Evil is a name of a foeman, as I live.

6 12

New FAQ entry: broken autoresponders
by Greg Ward 06 Dec '01

06 Dec '01

3.6. What can I do about users with broken autoresponders? http://www.python.org/cgi-bin/faqw-mm.py?req=show&file=faq03.006.htp

3 2

Virtual Hosting Implementation
by Hubbard, Matt 06 Dec '01

06 Dec '01

Hi, I'm looking at implementing the Virtual Hosting support as described in the relevant design notes on the Wiki. I would like to query the usefulness / necessity of having global lists. Surely from an implementation point of view, things would be far simpler if all hosted domains (including the "default") are managed in the same way as the virtual domains? In regard to directory layout, would it not be a good idea to place the lists, archives and "virtual domain site configuration" under one directory, rather than "./lists/vdomain/" etc. For instance: vhosts/python.org vhosts/python.org/site-config vhosts/python.org/lists/ vhosts/python.org/public_archive/ vhosts/python.org/private_archive/ vhosts/virtual.org vhosts/virtual.org/site-config .. etc... Cheers, Matt.

1 0