[INFOD-WG] XML thoughts

[Fisher, SM (Steve)]

Hi,

Since reading my new Xquery book, I think that some of the choices we
have made are not good.

I would like to use the id/idref mechanism to relate XML elements and to
use EPRs only for external references.

I would like to treat the whole registry as a single XML document (as we
currently have it in the spec) rather than moving to collection
functions to do the get meta data. The reason for treating the registry
as a single document are:

 it makes the constraints much easier to explain and to use

 it makes queries simpler

 it allows us to use the id/idref mechanism which is more natural and
should be implemented more efficiently as it is the way to do cross
relationships outside the XML hierarchy.

To some extent the problem is the hierarchical XML model which is not
very well suited to describing the INFOD registry - the relational model
would be much better - however we clearly have to go the XML route to go
along with the current trends :-(

The original spec did not (to my mind) have quite the right structure
for the document. I think it should have top nodes of consumer,
subscriber etc without the infod$$ bit and then the individual
resources.

This can all be well described by an XML schema and/or DTD  which will
help a lot to explain the system to people.

Dieter has concerns about the efficiency of using one big document
rather than one document per resource. I hope this is not a problem
othewise the XML technology seems to be badly broken!

One other problem I found is that we don't define exactly what a
property constraint should look like. We must do this and explain the
processing rules for combining those constraints. Clearly each
constraint is a partial xquery body. However each may be preceded by a
number of xquery declarations which are presumably local to that
constraint. We currently have some text explaining how the constraints
are applied simultaneously - but I don't think that what we have is
sufficiently precise.  Can someone come up with a way of defining what
an acceptable constraint looks like - and what it means.

I hope to phone in today - but it will be on my mobile as I am at a
conference in Budapest

Steve