A few questions

[Tom Hughes]

In message <889c158749.blaughto at blaughton.cam.pace.co.uk>
          Ben Laughton <ben.laughton at pace.co.uk> wrote:

> In message <yekd7qnhglw.fsf at elva.cyberscience.com>
>           Tom Hughes <tom at compton.nu> wrote:
>
> > In article <6748ec8649.blaughto at blaughton.cam.pace.co.uk>, Ben
> > Laughton <ben.laughton at pace.co.uk> writes:
>
> > > Marking articles as read seems very slow.
> [snip]
>
> > Marking as read shouldn't involve RemoteNB at all at the moment as
> > that is something which Messenger tracks itself.
>
> Ah right.  <fx: digs out IMAP spec which I looked at ages ago> I thought
> that maybe RemoteNB was using the IMAP \Seen flag to mark messages as
> read on the server, which might have accounted for the speed.

Well the \Seen flag does usually get set simply because fetching the
message body sets it as a side effect.

There is ongoing work to get Messenger to keep it's flags in sync with
the server. The message protocol is defined and I have implemented the
necessary support in RemoteNB which is currently disabled until Mark
does the necessary to Messenger.

> > It really shouldn't freeze the machine as it uses non-blocking socket
> > calls. You may find that the next version will help somewhat though
> > as it allows client requests from Messenger to interupt scans for new
> > mail.
>
> I remember you saying this when I complained about the polling speed
> before.  I'd always meant to compile a debug build of the Internet
> module to see if that could reveal more about what was going on, but
> it's one of the many things I never got around to...

That issue has gone away now as I've stopped using blocking sockets
and relying on Upcall 6 to do a context switch. It now uses non-blocking
sockets and does a context switch itself if it gets EWOULDBLOCK back.

Certainly I'm using Internet 5 at the moment and havn't noticed any
problems that I'd put down to the stack.

Tom

-- 
Tom Hughes (tom at compton.nu)
http://www.compton.nu/
...Kix are for trids.