Re: A T does not work in nnimap

[Andrew Cohen <cohen@andy.bu.edu>]

>>>>> "Lars" == Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

    Lars> Andrew Cohen <cohen@andy.bu.edu> writes:
    >> Firstly, this may be beyond my minimal skill set.  But in any
    >> case, I wonder what the backend functions should do. They could
    >> return the headers as you suggest. Or they might just return the
    >> list of articles. But I noticed that there is an imap extension
    >> for returning a threaded list of messages. It might be useful to
    >> choose something that would allow for backend specific speedups
    >> in threading.

    Lars> Gnus does its own threading according to its own algo.  It's
    Lars> better to just return a buffer full of article headers and let
    Lars> Gnus do its magic.

OK. Like I said, probably beyond my skill set, but I'm still trying.

Once I have retrieved the headers (gnus-retrieve-headers), how do I get
them incorporated into the thread? I think I see how to do this with
NOV, but not with the full headers in the nnimap-buffer.

The existing gnus-summary-refer-thread doesn't work very well even for
nntp groups. For example I have one of Julien's messages in this thread
ticked (the only ticked message in gnus.general). If I enter the group
this is the only message showing in the summary buffer. If I then 'A T'
on this message, I get a subset of thread only. From what I can tell the
existing code tries to download a whole bunch of headers and simply
hopes that the full thread is present. Sometimes it is, and sometimes it
isn't. I can guarantee the full thread in nnimap, but only if I do
something special for the nnimap backend, which I think we would like to
avoid. 

Regards,
Andy