On Wed, 2005-12-21 at 09:50 -0500, Greg Hudson wrote:
> On Wed, 2005-12-21 at 09:05 -0500, Daniel Berlin wrote:
> > I'm a bit confused then.
> > You said "we haven't bumped the repos format number since *before* we
> > created the concept of separate FS format numbers".
>
> > That implies that it wasn't bumped when separate FS format numbers were
> > created.
>
> Right, we didn't want to prevent old code from accessing new
> repositories with FS format numbers.
>
> > Which means the internal format numbers can't ever be the only thing
> > bumped, because older versions don't look at them.
>
> > Which brings me to the question.
> > If we can't actually just bump the internal FS format numbers when
> > completely internal changes are made, why are they there and checked?
>
> Did we jump the gun adding FS formats? Possibly. It seemed reasonable
> to get an early start doing checking at the right layer, but now that it
> comes time to actually take advantage of that checking, it's not clear
> that getting an early start actually helped us any.
>
> Luckily, we are bumping both FS format versions at once, so it makes
> sense to just bump the repos format,
So uh, should i just bump the repo format and undo the bump of the
internal versions, or bump all three (repo, bdb, fsfs)?
> while allowing new code to work on
> the old repos version and making the new svndiff0-only svnadmin create
> option create a repos with the old repos version.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Sun Dec 25 02:54:27 2005