[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: [TSVN] Thoughts on performance

From: Nicklas Norling <exinor_at_exinor.net>
Date: 2005-04-14 17:36:50 CEST

Will Dean wrote:

> At 15:28 14/04/2005 +0200, you wrote:
>> Regarding performance issues in TSVN I've been thinking a lot of the
>> filesystem events on windows. I know that there is a new cache in the
>> upcomming 1.2, but as far as I understand it can still age out etc.
> My plan for the 'new' cache was really that there wouldn't be *any*
> aging out on it, it would be sufficiently diligent about checking
> things that you wouldn't need to discard stuff just because it was old.
> I did put an aging mechanism in there, but I'd really like to take it
> out again. If they're working properly, the other checks in there
> should keep the data fresh. I only wish I had taken it out before I
> ever released any of the cache code.
>> Have there been any thoughts into using windows filesystem events
>> to trap changes to files and alter the content of the cache this way
>> to keep it fresh.
> Yes. I think that the user-mode file/directory watching mechanism is
> unsuitable for watching arbitrarily large numbers of directories.
> (It's apparently what Explorer uses, but Explorer only needs to watch
> a very small number of directories at once.) A kernel-mode
> file-system filter would be a great way to do it, and there's a new
> framework for file-system filters coming along from MS, but to be
> honest I doubt that anyone will ever have the time to write such a
> thing for TSVN, at least under the existing framework.
Not being knowledgeable in user-mode/kernel-mode events I can't argue
with you here.
It's a pity you feel it would be next to impossible. How sweet for the
cache to know that
a user just touched a file :) I'm guessing kernel-mode is what virus
scanners use then?

>> I'm guessing svn might encapsulate some of the functionallity needed
>> but if we can make a good case maybe svn devs would be inclined to
>> accept implementations using filesystem events in svn libraries as well?
> I very much doubt it, as they don't really have the continuous need to
> supply up-to-date status info which TSVN has. Maybe I'm
> misunderstanding you, though.
> Cheers,
> Will

But they do get such a need if we can demonstrate such a case :) They
provide our library and
hence, in a way, we dictate what they need to supply. The svn command
line client is only one
implementation using the libraries.

Btw. thanks for a great answer!


To unsubscribe, e-mail: dev-unsubscribe@tortoisesvn.tigris.org
For additional commands, e-mail: dev-help@tortoisesvn.tigris.org
Received on Thu Apr 14 17:37:16 2005

This is an archived mail posted to the TortoiseSVN Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.