>>>>> "Kjell" == Kjell Harald Andersen <k.h.andersen@chipcon.com> writes:
Kjell> I've seen that, and it would work. However, wouldn't that use
Kjell> a lot of server resources. I'm not an expert on these things
Kjell> (kind of thrown into it), but it looks to me like that command
Kjell> will stream the entire file through a byte counter. For
Kjell> smaller files that wouldn't be a problem, but I have users who
Kjell> regularly commits 10-30Mb and even more.
Yes, it does mean that. But that's not a big deal. My Mac does
"cat foo | wc -c" on a 20 MB file in about 0.6 seconds; on a 500 MB
file it takes about 15 seconds.
Kjell> My point is that this information is already available on the
Kjell> local file system at commit time. Why can't it be transferred
Kjell> to the repository. It shouldn't add more than a couple of
Kjell> bytes to each file transmitted, and if would save a lot of
Kjell> server time compared to the approach mentioned below.
You're quite right that it would save some server time to use the
information already available in the file system. My earlier point
was that the capabitility *does* exist today. It isn't quite as fast
as it could be, but if a normal commit is under 50 MB or so, it's
still pretty fast. And yes, if someone tries a 600 MB commit they
will wait 15 seconds before getting the error message. I can't see
that as a problem.
paul
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Dec 1 16:34:35 2005