On 5/20/05, Branko Èibej <brane@xbc.nu> wrote:> John Peacock wrote:> > > Mark Phippard wrote:> >> >> I think there have been plenty of complaints about the performance of> >> svn status on Windows. This is nothing new.> >> > Thus laying the blame firmly on NTFS, where it belongs... ;-)
As much as I would love to blame NTFS/Windows, it is not purely the case.Just strace svn commands some time on Unix. Look at all of the redundantcalls. The only reason it is reasonable in performance is due to the amountand quality of caching that many Unix systems perform.
> "Stercus tauri", as the ancients used to say.> > In fact, Subversion WC performance sucks on Unix, too. We make far too> many system calls to get the information we need. As to Windows> performance, some of the WC design decisions we made (that have no> noticeable effect on Unix) cause slow-down Windows. We can't easily back> down on those decisiions now without a WC redesign, but saying it's all> Windows' fault is missing the point.
I would like to see some internal (within SVN) cleanup of redundantsystem calls before we go and blame the WC design. (Not that thereare not some issues, but lets fix the outliers first)
(PS - Depending on my work load, I may be able to start looking atthat problem again - it is not clear, however, that I will have muchtime given family and work issues)
-- Michael Sinz Technology and Engineering Director/Consultant"Starting Startups" mailto:Michael.Sinz_at_sinz.orgMy place on the web http://www.sinz.org/Michael.Sinz
Received on Sat May 21 17:55:29 2005