Re: threading is a risky business

From: Daniel Stenberg <Daniel.Stenberg_at_frontec.se_at_hypermail-project.org>
Date: Fri, 15 Oct 1999 10:24:26 +0200 (MET DST)
Message-ID: <Pine.GSO.4.20.9910150903370.18437-100000_at_stanley>

On Thu, 14 Oct 1999, John Finlay wrote:

> My guess is that the problem is related to matching to the same subject
> "Re: Beschwerde" when there is no replyto field.

That's my conclusion too.

> I think that oneunre() should be done before the first subject match is
> attempted. My thinking is that without a msgid in the replyto field you
> have to assume that the original subject must be at least oneunre() down
> ("Beschwerde" in this case). In this case, 0 would be the reply to 2
> and 1 and 2 would not be replies.

Hm, yes. That might not be such a bad idea. Pretty straight-forward to implement too...

> When I rewrote the threading code for my variant of hypermail, if a
> message should be a reply to another but no match for the other is found
> then a dummy message is created to act as the anchor for the thread and
> possible future matches. The dummy messages are not printed out but the
> likely related mesages are printed together in the thread list. Your
> example would end up threaded like:

I like that concept. It will effectly (attempt to) gather all replies to a mail that is not present into one thread, yes.

In this particular mailbox I think it would work.

> dummy message
> |
> +-1
> |
> +-2
> |
> +-0

Yep, this is how I figure it should look like

> It's not perfect since it's possible to have an unresolved msgid in the
> replyto field of 1 that would create threading like:

I don't follow you here.

> dummy message
> |
> +-1
> |
> dummy message
> |
> +-2
> |
> +-0
>
> This result has the potential to have possibly related messages
> separated while printing.

If message 1 and 2 had different and non-existing replyto-ids, then how can we possibly know that they are replies to the same mail? Of course, based on the subjects, we could probably identify this too, but then we're getting quite advanced and I'm not so sure that is the right thing to do.

-- 
 Daniel Stenberg -- http://www.fts.frontec.se/~dast/
 You'll find my hypermail pages at http://www.fts.frontec.se/~dast/hypermail

Received on Fri 15 Oct 1999 10:26:35 AM GMT