Should DASL have structure queries?

Some of the things you'd like to do with DASL require the ability to form
queries over structured objects.  The best example is the lockdiscovery
property.  How could a client find all resources locked by a given owner?
The information is available in the lockdiscovery property, but none of the
currently defined DASL query operators can work with structured values. 

What should we do about this?  I see four possibilites

 1 - Omit structured query in the version.  Some people think it's too hard
to define for now (a "rathole"), and that most of DASL's value will come
from queries on 'ordinary' values.

 2 - Define ad-hoc properties that flatten those DAV properties that have
structured values.  This is what the current draft does, with the
'iscollection' property.

 3 - Define our own structure query.

 4 - Wait for the XML community to produce one.  They may or may not
 produce one in time for the DASL spec.

We need to choose a course of action.  I would like some debate from the
list on this point.

I prefer option 3.  Option 2, while deceptively easy for the simple cases
(resourcetype) would be a pain for properties with multiple values
(lockdiscovery can potentially list multiple locks), and it does not
generalize.  Option 4 is tempting, but we have no reason to think that the
XML community will produce a design soon, and in any case our task is to
search data from the DAV data model, which is NOT XML.  DAV uses XML as an
on-the-wire syntax, but the model is not XML.  Our task is more constrained
than theirs.

I think it's possible to define a few simple operators that can access DAV
structure. I've seen at least one such proposal (I sent one  Sun, 30 Aug
1998 13:43:39 PDT) and  I believe Alan is working on another.  There is
little downside to designing our own, especially if it's small and cheap.
If the XML community later produces a better one, we can switch to it, and
will have lost little even if our work is replaced.  Moreoever,
constructing a prototype that actually implements this simple structural
query may produce useful insight into the larger and more general problem.

I'd like the list to come to a concensus about what to do.

Your opinions solicited.

Jim

Received on Wednesday, 16 September 1998 18:54:07 UTC