I wrote:
> Leon Zandman <lzandman@lode.nl> writes:
> > I don't understand why you don't checksum the properties. I do
> > understand why the properties/log messages aren't versioned, but I'd
> > still like to know when my repository has been corrupted somehow.
>
> I think it's because we just didn't think of it.
>
> We could checksum them, I suppose. But so far there hasn't been a
> real-world instance of them being corrupted (yet no other part of the
> repository being corrupted), so we haven't bothered. Unless there's
> been a recent report of such and I just missed it? Let us know if so.
>
> How's that for a direct answer? :-)
I forgot, there's another reason we don't checksum revision
properties.
Most data in Subversion is reflected in the working copy. File
contents, directory contents, and regular versioned properties are all
"cached" in the working copy. (I'm not talking about .svn/text-base/
here, I mean that the working copy *itself* is a cache of data that's
already in the repository).
One purpose of our checksums is to make sure that the two sides agree.
For example, we take a checksum on the client side, then send the data
to the server, then make sure the server gets the same checksum for
the same data, or vice versa.
But with revision properties, there's no client-side cache anyway,
except very briefly in your editor or in a tmp file or something.
There's very little opportunity for the values to become corrupted in
situ, so checksums are correspondingly less important.
I'm not saying checksums on revision properties would be useless, of
course, just that they're less important for revision properties than
for other kinds of data.
Best,
-Karl
--
www.collab.net <> CollabNet | Distributed Development On Demand
---------------------------------------------------------------------
To unsubscribe, e-mail: users-unsubscribe@subversion.tigris.org
For additional commands, e-mail: users-help@subversion.tigris.org
Received on Thu Oct 6 21:24:15 2005