(Aside: Is there some reason why you To: me and CC: the list rather than having the list address in the To: field? I ask because I'm wondering if it's a gmail thing, or something about your MUA, and because I suppress the list copy if I'm CC'd directly, I don't get a List-Post: header, so my MUA's reply-to-list won't work on your messages. I'm mostly curious because I can work around it as you see here, but just wanted to point it out in case you weren't aware of it and/or could easily fix it -- if you even agree it's broken. ;)

On Mar 29, 2012, at 11:46 PM, Stephen J. Turnbull wrote:

On Thu, Mar 29, 2012 at 8:07 AM, Barry Warsaw <barry@list.org> wrote:

...

Mailman 3 itself requires unique Message-IDs.

So? FWIW, I don't think I agree with that requirement (even RFC 5322 doesn't make it a "MUST"), but I'm not going to argue with you about Mailman 3 design, that's your pidgin.

We had a fairly lengthy discussion about this a year or so ago, and I'm pretty sure we came to the consensus (even if not unanimous) that it was reasonable for Mailman to impose this restriction. It's 2012 so if your MUA/MTA can't generate unique message id's there's no reason for us to think you're not a spammer <wink>.

But there's nothing particularly Mailman-3-dependent about archiver web UIs, though. I don't see any reason why the front end shouldn't be used on my several gigs of personal archives going back to about 1980, eg, or as a poor man's webmail.

True, the archiver as an independent component can be more lenient. I'm just pointing out that as fed from Mailman, it will only get unique message ids. And if it doesn't, that's a bug in Mailman.

...

 IIRC, the Mail Archive guys found a very very low collision rate over millions of messages, and I think all such cases were basically spam.

Sure, but XEmacs archives go back to at least 1994. mailarchive.com is a more recent phenomenon. In the early days of Linux/*BSD diffusion, there were lots of buggy MUAs/very simple MTAs out there.

Sure. And I suspect that there will be plenty of mailing lists that get fed messages from programs, e.g. think vcs -commit diff lists. Those programs can also be buggy, but again I'd prefer that Mailman not compromise on this issue for their sake.

It doesn't do that for subobject content IDs, and more important, users don't necessarily have the X-Message-ID-Hash (they may have not-metoo set, they may have gotten the message as a direct Cc). True, it's easy enough to compute -- if you're a Mailman 3 developer and know it's present.<wink/> And, of course, why have a Mailman 3 dependency that is absolutely unnecessary?

Right. The point being that messages that flow through Mailman will have that hash in the message URLs in the Archived-At header and possibly in the decorated footers. An archiver should certainly provide an interface to look up a message by pure Message-ID or the hash. The hash is just a scheme to regularize the message id and is a tiny fraction more user-friendly (because of its limited alphabet and manageable, known-in-advance length).

-Barry