--On September 14, 2005 10:30:35 PM +0200 Patrick von der Hagen <[log in to unmask]> wrote: > Am Mittwoch, den 14.09.2005, 12:15 -0600 schrieb Michael Loftis: > [...] >> Mailman's bounce handling is a joke. Mailmans list "database" handling >> is also a joke. It can NOT handle large lists. It reads the entire >> Python pickle of the list in and rewrites that every single time a >> change is made or a bounce is processed. > I'd like to thank you for your well-founded thoughts on mailman and its > deficiencies. When evaluating alternatives to listserv, I'll use your > thoughts to guide my analysis and concentrate on those issues. Well it is. When moving several VERY large lists to mailman (100k+ subscribers) they each shed in excess of 10k users on their first sends, all of which were bouncing with 550 mailbox doesn't exist errors. These lists would also more or less take down the list machine every single time they sent due to the unprocessed bounces and the way in which mailman handles bounces (to rewrite the entire subscriber database for each and every mail) -- this takes 1-2, or more seconds per mail on large lists. Locking contention builds up because any send of a large list prompts people to request unsubscribes and subscribes as well, and the webUI basically stops completely. Even on relatively fast hardware using a RAM drive it becomes CPU bound. For small, simple lists, I do like it. But anything larger, requires something a lot better. The UI on mailman is atleast as confusing as ListServ's too for the list owners. I don't think either is all that great, honestly. Atleast ListServ list owners can change the subscriber UI without impinging upon our staff here to do it. Neither solution is perfect honestly. My big beef with mailman is the way it handles it's subscriber database. The second one is the very inneffective bounce handling. I honestly didn't think it was too bad until I saw how many subscribers were pulled off of these lists when moved to ListServ. And yes I realise that it could be a settings issue, and in one of the several cases I think it was, but in the others, the settings should have caused most, or all, of those subscribers to have been removed by Mailman due to bounces. Anyway I don't intend to start a flame/troll war. I've simply had (and continue to have) very rotten performance from mailman for large lists, and very poor bounce handling from mailman as well. ListServ is ... better it processes and handles more of the bounces, and doesn't' completely get blocke dup on large lists that end up with a large bounce rate...the bounces dont' tend to block out web clients forever...however large mailing list sends and imports do block out web clients, unless you have HPO (obviously, and not really ListServ's fault, just pushing the limits of the algorithms in non HPO) ... and it's also worse WRT bounce processing... there's no way to say (or rather none that I've ascertained), leave auto bounce processing on, but don't inform me of any of it, or don't inform me of auto processed bounces but do inform me when you fail to, or any combination thereof. Instead if auto processing is on you get the successfully processed bounces. Atleast that's what the documentation seems to say, and seems to be my experience, I've not taken the time to do direct match ups though.