[please keep the conversation on the users@ list]
On Jan 25, 2005, at 11:55 AM, Earl A. Killian wrote:
> Ben Collins-Sussman wrote:
>
>> Ah! Excellent! Then the mystery is probably solved. If you upgrade
>> the server to 1.1, the problem will probably be gone.
> I see how the change you described makes the timeout go
> away, but the broader problem is the total time it
> takes (it looks like I could increase the timeout to 90m
> and make the timeout go away, but I didn't bother because even
> 40m is too long). If before it took 60m to
> produce the report, and then an unknown amount of time
> to transmit and act upon that report (let's just be kind and
> say 10 minutes), then streaming the report will eliminate
> the timeout, but it will reduce the hypothetical time from
> 70m to 60m, unless the report generation has also been
> sped up. Do you agree with this analysis?
>
Are you saying that it takes 60m just for the client to stream a
description of its working copy to the server? That sounds ridiculous
to me, I don't believe it. 'svn update' merely needs to open each
directory in the working copy, read .svn/entries, and send that
information to the server. It shouldn't take that long. It should
definitely be much faster than 'svn status'.
The bug I'm describing was a huge delay on the *server* side, whereby
it used to take a long time to process the client's report before
sending a response. (The server should no longer be doing all that
work.) But merely sending the report the server shouldn't take very
long.
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Tue Jan 25 19:36:11 2005