Re: Duplicate message ids

From: Paul Haldane <Paul.Haldane_at_newcastle.ac.uk_at_hypermail-project.org>
Date: Thu, 6 May 1999 10:35:01 +0100 (GMT)
Message-ID: <Pine.GSO.3.95-960729.990506101630.25188A-100000_at_carr6.ncl.ac.uk>

On Wed, 5 May 1999, John Finlay wrote:

> Paul Haldane wrote:

....
> But how can you tell which one has the broken duplicate and which is the OK
> duplicate? Seems like you'd have to keep both with the duplicate msgid and try to
> separate by subject for replies.

I don't think that (at this stage) we should go overboard on complicated solutions to this problem.

It should only affect a small number of messages (if you're getting many duplicate msgids then you probably want to work on fixing that :->. I realise that this often isn't under your control, and we have to be prepared to deal with reality as opposed to what the RFC says, but the bottom line is that there should be a 1 to 1 correspondence between mail messages and msgids).

At the moment hypermail's data structures/logic assume a 1:1 msgid:message correspondence (at least that's my understanding) so we _don't_ want to store two messages with the same msgid.

I think the bext thing for the moment is to (at the user's choice) either dump duplicates (with an error message - allowing the user to fix things by hand) or do a simplistic replacement of any duplicate msgids with unique hypermail-generated ones (with a warning so that the user knows what's happening). This _will_ mean that replies to the second message with the dup. msgid will be attached to the thread of the first message with that msgid.

John's suggestion of using the subject lines to distinguish between the two threads would work, but this is probably too much effort _at this stage_ if you're only seeing a small number of dupes.

Paul Received on Thu 06 May 1999 12:00:44 PM GMT