On Tue, Feb 05, 2002 at 06:25:23PM -0800, Zack Weinberg wrote:
> On Tue, Feb 05, 2002 at 08:10:34PM -0600, Karl Fogel wrote:
> > The Berkeley atomicity helps, but it didn't give us commit atomicity
> > completely for free. We still had to do some of the same things in
> > the db that we would have had to do in the regular filesystem, to
> > achieve commit atomicity.
> >
> > I have no idea if it would have been harder or easier to do just using
> > filesystem primitives. We only had time to do it once. :-)
>
> Well, of course you had to do some work to implement it. I was
> primarily thinking of the work saved by being able to code to one API,
> with reliable cross-platform semantics. Also, the way to use the
> txn_* routines to implement atomic transactions is a lot more obvious
> than the way to use link(2)...
Well... we got quite a few other things from using Berkeley :-) Hot
backup and replication for starters. All kinds of existing tools that know
about BDB databases (e.g. Python or Perl bindings). A body of "community"
knowledge. etc
But I think best of all, we didn't have to write and debug the darned thing.
As a result, we haven't had one incident of data loss/corruption in the past
five months that we've been self-hosting. We haven't had to do any surgery
on the repository to extract bits that got lost. Nada.
(we have fixed some bugs in the *app* that has had some interesting
repercussions on the data *in* the repository, but that isn't berkeley's
fault at all; at some point, we'll probably just export/import the repo to
clean out some of those tidbits)
Cheers,
-g
--
Greg Stein, http://www.lyra.org/
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sat Oct 21 14:37:04 2006