Re: Threading algorithm

[Earl Hood <ehood@hydra.acs.uci.edu>]

On August 22, 2001 at 15:57, Jeff Squyres wrote:

> Is there any way to disable or tune the subject-matching aspects of the
> threading algorithm?  I realize that disabling subject-matching may
> mis-thread some messages that *should* be threaded (e.g., from icky mail
> clients that don't include a In-Reply-To or References lines), but I think
> that I would prefer such messages show up as a top-level post rather than
> have unrelated messages show up buried in a very old thread index.

The SUBJECTTHREADS resource can be turned off to disable subject-based
thread detection.  Only in-reply-to and references data will be used.

> Or perhaps it would be possible to do threading solely based on subject
> matching (in the absence of In-Reply-To or References) only on some
> maximum amount of time -- e.g, in the absence of IRT of Ref lines, only
> thread a new message to an old message if:
> 
> - the subject matches
> - the date difference between the new and old message is < N days (perhaps
>   N can be user-definable)

Good idea, but may be somewhat complicated to implement.  I'll need to
do some evaluation to see if it is possible without hurting performance.
Changes to the temporary data structures used to determine thread
ordering will have to be changed to mark multiple "first" occurance
of subject and them to make sure a subject-based follow-up falls into
the right thread bucket according to its date.

Another suggestion that has been proposed in the past is to treat
a message without a "Re: ", or more specifically, that does not
match SUBJECTREPLYRXP, as the start of a new thread, but it also
requires the 'multiple "first" occurance' that your idea suggests.
Actually, this idea and your idea is very similiar in how the
implementation would be done.  Again, performance considerations
crop up on supporting this kind of feature.

--ewh