On 9/27/07, Daniel L. Rall <dlr@collab.net> wrote:
> On Wed, 26 Sep 2007, Garrett Rooney wrote:
>
> > On 9/26/07, Greg Hudson <ghudson@mit.edu> wrote:
> >
> > > I think their "distributed development" row (which is actually just
> > > about slow network links) kind of inverts the story. My muddled
> > > understanding is that Perforce is much more dependent upon a fast
> > > network link because you typically NFS-mount your working copy.
> >
> > Actually it's more the fact that you have to contact the server for
> > almost all actions, you don't actually NFS mount your working copy in
> > p4 (perhaps you're thinking of clearcase?).
>
> ...because all p4 WC meta data is stored server-side; only the equivalent
> of Subversion's text-base is stored client-side.
Yes, exactly. Also note that this is one of p4's big scaling issues,
as the amount of data the server needs easy access to (i.e. needs to
be cached in RAM to be usefully fast) is proportional to the combined
size of all active working copies (they call them "client views") at
any given time. Once you have more than can comfortably fit in RAM on
the server you're screwed.
-garrett
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Thu Sep 27 15:22:56 2007