On Fri, Mar 14, 2003 at 07:27:53AM -0800, Ben Collins-Sussman wrote:
> In a nutshell: there's no point in sending UUIDs this way. The other
> update-report 'entry' props are dynamic things; if the update report
> says a file has changed, then by gum, the created-rev, author, and
> date have probably changed as well. But the UUID is invariant.
>
The reason I added this functionality was that I noticed that
an "svn up" over ra_dav in an existing (non-UUID-containing) WC
didn't add UUIDs. It's certainly ironic that this nod to
compatibility triggered a massive incompatibility!
People with old WCs can simply re-checkout. That will be much
nicer. The update functionality was, in retrospect, a bad idea.
> Furthermore, the UUID live-property still exists on every resource.
> That means you'll get them when you do a fresh checkout of a tree, and
> any time you receive a new item from the server. That's plenty of
> coverage. There's just no need to marshall these invariant things to
> the client during an update.
>
> Any objections? If not, I'm just going to remove this UUID update
> stuff from /trunk.
Sounds great to me.
--ben
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Fri Mar 14 16:54:09 2003