Re: [Evolution] Evolution takes very long to read IMAP folders

[Not Zed <notzed ximian com>]

On Fri, 2002-06-14 at 06:42, Alexander Skwar wrote:
> Hello.
>
> I just tried out evolution 1.05 again.  I'm using the Mandrake
> RPM package evolution-1.05-1mdk on a current MandrakeLinux cooker
> system.
>
> When I load some semi large mailboxes over an IMAP connection to
> a courier IMAP 1.4.5 server running on localhost, Evolution
> nearly always takes at least twice as long to read a given
> mailbox as Mozilla 1.0 does - and mutt is almost always ~4 times
> as fast as Evolution, but since mutt doesn't cache anything, it
> always needs to reload the complete folder, so overall Evolution
> will be fast that mutt.
>
> Anyhow - why is it, that Evolution is so very slow in reading
> IMAP folders?

Becuase the imap code sux?  Uh.  Not entirely.

1. evolution gets all of the headers (not sure if the others do, mutt
probably doesn't), which for most imap implementations isn't real cheap,
and also means more network traffic.
2. evolution stores each header on disk, which in my testing slows
things down quite a bit esp for bigger mailboxes (where bigger is >5K or
so).
3. the imap code does suck, at least a bit.

We get all of the headers mainly so 'vfolder on mailing list' is more
reliable, and the threaded message display works more reliably too.  The
information we use to calculate both are not part of the information
obtained in the imap request (FETCH BODY) which was designed for
'getting enough info to display a list of messages'.  I'm not sure
caching the headers on disk was really useful, we should probably throw
them away and rely on the summary content, or store them some other,
more efficient way (like a null mbox).

There are also some overheads in evolution not related to imap at all,
but the displaying of the list of messages, which start to impact more
once you get above a few thousand messages.  We're looking at addressing
these issues too, but they probably dont rate very highly in your
timings anyway.

I've been rewriting imap in my spare time to address some issues like
this, but its not going terribly fast (and my spare time lately is worth
more to me than better imap code!).  But if it sticks to still getting
all (or even an 'interesting subset' of) headers, so that 'vfolder on
mailing list' continues to work usefully, its still not going to be as
fast as mutt, although it should be faster than currently.

Might have to make it an option so users can decide what is more
important to them - speed or features.


> Here are the timings I've done:
>
> Mandrake Changelog List
> - 1902 messages
> - 9 MB
> -> mutt: 5 seconds
> -> Mozilla: 10 seconds
> -> Evolution: 20 seconds
>
> Mandrake Cooker List
> - 11574 messages
> - 64 MB
> -> mutt: 30 seconds
> -> Mozila: 70 seconds
> -> Evolution: 190 seconds
>
> MySQL List
> - 13136 messages
> - 73 MB
> -> mutt: 30 seconds
> -> Mozilla: 70 seconds
> -> Evolution: 190 seconds
>
> The timings have been taken with a "stop watch" from the time I
> clicked/opened the folder until the time the MUA was done with
> reading.
>
> Alexander Skwar
> -- 
> How to quote: http://learn.to/quote (german) http://quote.6x.to (english)
> Homepage:     http://www.iso-top.de      |    Jabber: askwar a-message de
>    iso-top.de - Die günstige Art an Linux Distributionen zu kommen
>                        Uptime: 16 hours 41 minutes
>
> _______________________________________________
> evolution maillist  -  evolution ximian com
> http://lists.ximian.com/mailman/listinfo/evolution