[svn.haxx.se] · SVN Dev · SVN Users · SVN Org · TSVN Dev · TSVN Users · Subclipse Dev · Subclipse Users · this month's index

Re: A forward-thinking thought about deltas in FS design

From: Stefan Monnier <monnier+gmane.comp.version-control.subversion.devel/news/_at_cs-www.cs.yale.edu>
Date: 2003-07-07 17:13:06 CEST

> You can really only make the strong guarantees if you always store a
> skip-delta immediately upon commit and never throw them away.

I'd expect that the currently committed revision (i.e. the latest) would
simply always be immediately added to the cache. I don't understand your
comment about skip-deltas since AFAIU there wouldn't be any such thing
on-disk any more (you'd only have one-step-deltas and a cache composed of
actual files).

> That's a very degenerate case of "caching."

It's very common to do such opportunistic caching.

> For a large project, this means for that period of time (which may well
> be days, not hours) the system doesn't work. If it takes two days
> instead of half an hour to check out the latest version of Mozilla, the
> system is not carrying on gracefully.

I must admit that I have no idea what impact the skip-deltas have
in reality. But my experience with CVS is that retrieving revision
1.1 from the repository (which is O(n)) is not 96 times slower
as your example assumes. Maybe Mozilla's repository has an `n'
much larger than Emacs's.


To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Jul 7 17:15:12 2003

This is an archived mail posted to the Subversion Dev mailing list.

This site is subject to the Apache Privacy Policy and the Apache Public Forum Archive Policy.