On Tue, 20 Sep 2005, kfogel@collab.net wrote:
> Daniel Rall <dlr@finemaltcoding.com> writes:
> > How would this work with a third-party FS backend which is not yet part of
> > the SVN core (not that they're falling from the sky or anything)? It
> > doesn't seem particularly pluggable -- that might be okay given the
> > complexity of implementing a FS backend, however.
>
> Hmm. That's a good point. The API could return a constant string
> instead, so that each back end would be free to return whatever value
> it wants.
>
> There wouldn't be any central registry guaranteeing that the namespace
> remains free of conflict, if we did it that way. But maybe that's
> okay. People implementing new repository back ends should be at least
> enough in touch with each other to avoid choosing the same names!
>
> Would you be +1 on getting rid of the enum type and just returning a
> constant string (i.e., allocated in static data segment, not pool)
> from svn_fs_type()? I'm happy to make the change...
As you say, there are trade-offs, but I think your suggestion of strings
instead of an enum would be preferable, as it doesn't lend the (false)
appearance that all possible FS backend types are known.
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Wed Sep 21 01:10:43 2005