[Mailman-Developers] In-Reply-To vs. References

Harald Meland Harald.Meland@usit.uio.no
15 May 1999 00:18:23 +0200


[Harald Meland]

> (Not having read c.l.p,) I don't understand how such a proposition
> could generate much of a discussion :)

Having read the discussion in c.l.p, I now understand :)

> If an email contains an In-Reply-To: header, I'd guess it is due to it
> being a reply.  If this email is gated to news, the fact that it is a
> reply, and preferably what message it is a reply to, should be
> indicated somehow.
> 
> If such an indication can only be given by having the news-gate muck
> around with the headers of the message, _and_ such mucking won't
> destroy any of the _original_ headers of the message, I don't really
> see the harm.

... and before anyone gets religious on me here, I'll rush to explain
that I _do_ see the harm of inserting "faulty" References: headers --
any In-Reply-To: candidate should at the very least be in the form of
a valid message ID.

I think there was a slightly similar discussion on the (ding) Gnus
mailing list a while back -- it was spurred by the fact that Gnus'
threading sometimes broke down, as some mailers put several
message-id-look-a-like-tags (e.g. mail addresses in angle brackets) in
their In-Reply-To: header.

Thus, Mailman should be very conservative when moving
MessageID-candidates from In-Reply-To: to References: -- it might even
be a good idea to query the newsserver whether it has the candidate
replied-to article (even though that probably won't work with all
newsserver configurations).

If the appropriate level of conservatism can be reached without making
the involved code *too* complex, I still believe this particular case
of header mucking could be a Good Thing.
-- 
Harald