On Feb 13, 2009, at 12:33 PM, Steve Bakke wrote:
> On Feb 13, 2009, at 9:37 AM, Bert Huijben wrote:
>>> -----Original Message-----
>>> From: Bolstridge, Andrew [mailto:andy.bolstridge_at_intergraph.com]
>>> Sent: vrijdag 13 februari 2009 14:44
>>> To: users_at_subversion.tigris.org
>>> Subject: RE: Speeding up workspace
>>>> -----Original Message-----
>>>> From: Bert Huijben [mailto:rhuijben_at_sharpsvn.net]
>>>> Sent: Friday, February 13, 2009 8:03 AM
>>>> To: Listman
>>>> Cc: users_at_subversion.tigris.org
>>>> Subject: RE: Speeding up workspace
>>> Is it possible to get such a fix into 1.6, maybe pushing back the
>>> release a little? This sounds like it's a worthwhile, and relatively
>>> safe, update. I'm sure many people would prefer a delay to the
>>> if it contained a significant speedup.
>> I suggested implementing this fix for 1.6 as the actual
>> implementation of a
>> stable fix is only a few hours of work (I could have completed this
>> a few
>> days ago).
>> The sentiment on the development list (and irc) was however that the
>> fix should wait for 1.7. (Reasoning: Not needed for 1.7 and the
>> is not worse than that of 1.0-1.5.. and that this should be checked
>> thoroughly of course).
>> (See the dev_at_s.t.o archive:
>> 4340 and follow ups)
> Not to mention that doing this would kill any cases where multiple
> clients want to access non-overlapping parts of the same working
> copy. Because 1.7 is going to use sqlite for the metadata, the
> locking should be "built-in" if I understand correctly. That means it
> should no longer create one lock file for each file under revision
[ it's one lock file for each directory under version control, not
file. But the point still holds. ]
There will still be some internal locking within subversion, but
SQLite will handle lots of the concurrency issues which we currently
have to do ourselves.
To unsubscribe from this discussion, e-mail: [users-unsubscribe_at_subversion.tigris.org].
Received on 2009-02-13 19:44:53 CET