Philip Martin writes:
> I have just read your attachment, and I don't know what to think.
> It's appears to be simply the same rough ideas as before. This time
> with a claim that it will improve some unspecified quantity by 30%.
> Amazing! I wish I could predict the behaviour of code that has not
> been written with the same accuracy. I suppose if you don't specify
> what you are measuring it gets a bit easier...
As a case in point, I got a bee in my bonnet last week about the fact
that 'svn status' sometimes stats the same file multiple times in a
row, and implemented a stat-caching method. When I measured it, it
turned out that the overhead of a) caching a stat entry and b)
figuring out that you could reuse it was the same as just re-statting.
> You also want to have a BDB implementation of the working copy admin
> area. While this is possible I see no evidence that it will solve
> your problem with the speed of working copies on network disks. Does
> BDB even support your network filesystem? Have you any evidence that
> it will be faster? Have you even considered the problems like backup,
> log files, recovery?
The single biggest thing I can see slowing down the client is multiple
reading of the entries file, and we seem to be doing a lot less of
that than we used to. I doubt that putting that stuff in a BDB would
speed it up much (and I *like* being able to fix my screwups with a
text editor ;-)
Robert
--
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Mar 2 23:36:25 2003