On Sun, 2005-04-10 at 14:11 -0400, Scott Palmer wrote:
> On Apr 10, 2005, at 1:22 PM, Greg Hudson wrote:
>
> > On Sun, 2005-04-10 at 13:16, Scott Palmer wrote:
> >
> >> "The repository doesn't care" - you can commit changes without having
> >> the file locked.
> >
> > Correct, in the sense that svn:needs-lock will not cause the server to
> > enforce that you have the file locked in order to commit to it. But
> > someone else having the file locked will cause the server to enforce
> > that you have to break or steal the lock in order to commit to it.
>
> Ahhhhhhh.
>
> So needs-lock is a hint to the client that you should get a lock if you
> want to avoid someone else making a change in parallel. If you do get
> the lock, then the server will not allow commits by others, but if
> others don't pay attention to the needs-lock they still are able to
> make a change locally (with some effort to by-pass read-only
> attributes, assuming the client was aware of svn:needs-lock) that they
> might have to throw away.
>
> I think I get it now. Thanks.
The salient point here (a point that we need to make abundantly clear
when we write the documentation for the locking functionality) is that
there are two parts to locking: The actual locking part (which can be
enforced to varying degrees by the server), and the
advisory/communication part (accomplished by setting:needs-lock on a
file).
Until we get this documented, I think that we're going to see more
misunderstandings like yours--thanks for your patience. :)
-Fitz
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@subversion.tigris.org
For additional commands, e-mail: dev-help@subversion.tigris.org
Received on Mon Apr 11 19:07:16 2005