RE: Can you BIND lock-null resources?

   From: Jim Whitehead [mailto:ejw@cse.ucsc.edu]

   > Is it expected that you should be able to BIND a lock-null resource?

   Interesting question. Since it is a resource, it makes sense that
   you can bind to it.

I disagree (strongly :-).  There are a large variety of things that
you cannot do to a lock null resource.  BIND should be one of them.
(In fact, just about everything should be one of them :-).  A
lock null resource is just a mechanism (hack :-) to expose a URL that
is locked but that currently is not bound to a resource.
If you want to see just how bad the lock null mechanism really is,
just compare any two implementations of lock null resources, and
notice how different/non-interoperable they are.

   So I'd say, yes. But, I'd understand if servers
   decided to fail such a request.

I'd say "no", and all servers should be required to fail such a request.

   > (1) Does the spec prohibit BIND-ing to a lock-null resource?
   >     (OK, I am being lazy and have not reread the spec again to
   >     verify this question - but the next question is the real question)

   I just gave it a quick look, and it seems like the BIND protocol
   specification does not mention lock null resources.

That is an omission (somewhat conscious, because I try to ignore
lock null resources in hopes that they will be replaced with a
more sensible form of lock discovery).  If it looks like we can't
make lock null resources go away, we need to update the BIND
protocol to explicitly disallow binding to a lock null resource.

   > (2) Regardless of the spec, do people think its acceptable to fail
   >     BIND requests to a lock-null resource with an error? (Is there
   >     any reason why someone would want to do it before doing a PUT
   >     or MKCOL etc?)

More than acceptable, commendable (:-).

Cheers,
Geoff

Received on Saturday, 23 June 2001 23:18:53 UTC