Re: UW imap c-client

[Simon Josefsson <jas@pdc.kth.se>]

Mark Moll <mmoll@cs.cmu.edu> writes:

> I have used both nnimap 0.95 and Pine 4.10. Although nnimap is quite usable, 
> the speed difference between nnimap and Pine is *huge*.

I agree with John, this wouldn't win us that much speed. The speed
difference between pine/gnus is because of the very different design
decisions.

> I was wondering if anybody has looked at using Pine's imap back-end,
> c-client. I'm not a real xemacs wizard, but I can imagine that this
> c-client library can be plugged into xemacs 21.2 as a module. Would
> this be a good idea?

Probably not, the imap parser isn't the bottleneck so replacing that
wouldn't improve much.

The main bottleneck in nnimap/gnus right now is imho that it takes a
long time to check for new mail. No other imap clients I've tested
does this at all, so perhaps we can make that part optional. (It can
also be made faster by using a different strategy).

The major bottleneck would then be that it takes a long time to enter
a folder, this is impossible to fix without heavy modifications of
Gnus, and for me this isn't a major problem anyway (entering a folder
with ~1200 articles takes ~25 seconds with 9s spent in nnimap for
me. I can live with that).

Do you agree with my observations of where the bottlenecks are? Which
things in nnimap are slow for you?