Re: [Evolution] how are threads sorted?

[Not Zed <notzed ximian com>]

On 25 Jul 2001 11:10:46 +0200, Thomas Emmel wrote:
> Am 25 Jul 2001 09:42:21 +0100 schrieb Ross Burton:
> > On 25 Jul 2001 10:06:06 +0200, Thomas Emmel wrote:
> > > Hi all,
> > >
> > > all the time I am just wondering how the algorithm to build a thread in
> > > evo works...
> > > It works mostly realy fine but sometimes there are curious threads like
> > > (I have to type it, cause I will not add a picture of evo itself ;-)
> > >
> > > janne...    - [Evolution] _Very_ strange sizing problem
> > > jeffrey...    - Re: [Evolution] _Very_ strange sizing problem
> > > steven...     | - [Evolution] Launching
> > > peter...      |   |_ Re: [Evolution] Launching
> > > ritchie...    |   |_ Re: [Evolution] Launching
> > > den...        |   |_ Re: [Evolution] Launching
> > > luis...       |_ Re: [Evolution] _Very_ strange sizing problem
> >
> > I don't know the algorithm as I'm not an Evo developer, but I do know
> > that sometimes that happens because people reply to an existing message
> > and then change the subject, so in this example steven might have
> > replied to jeffrey's mail and changed the subject to "Launching".  Look
> > at the headers of the messages for the References: field, this is used
> > to build the thread tree.
>
> Yes, but there is no reference-field for the first "Launching" mail
> the others are obviously referenced to that mail...

It checks for a References header, and if not, it checks In-Reply-To.

It is possible that someone just hit reply to the message, it is also
possible that the message-id isn't unique (whichever mailer is sending
them isn't being careful perhaps?), and finally the threading aglorithm
doesn't actually compare the whole messageid, it only compares a 64 bit
hash of the messageids, so there is a (rather slim) possibility of a
collision.

Looking at the messages in question (i still have them in my evolution
folder, shown the same way), In-reply-to matches the messageid of the
message from jeff, so it is actually threading as expected.  Since
in-reply-to is set at all, it means the first 'launching' message was
indeed created by replying to an existing message.

I know of one bug with the threading code, that can misorder a reply
above its parent, but that's definetly unrelated.