Re: updated schema

Shivaram,

I submitted your question but today being a US holiday, it will take some
time to get the answer.

I didn't find a link in the archive to a previous comment I had done
on schema modification, so I'll write it here again for the record.

After consulting with my colleagues some time ago about the schema
changes, the conclusion was that if the change was due to a better
understanding, or evolution of a related spec., we could do it
without changing the namespace. See it more like a rewriting of some
text to remove ambiguity, but not a change that adds new features. This
doesn't require a namespace change.

Also, its up to the WG's discretion and responsability to decide if we 
should go back to Last Call or republish the spec. It all depends on
the nature of the change, how many existing implementations will break,
how easy it will be to integrate the change, and what's the benefit 
of the change. The WG should not just justify the change by saying "we
thought it would be easy" but give proof that the change can be done.

If there has been substantive changes in the schema and spec, then
there's no choice, we need to go back to LC and have a new namespace. I
don't believe this is the situation today.

Thus, when you ask:

<quote>
Due to advances in schema writing for better error handling and parsing, 
use of QNames are being discouraged.  Hence, we have fixed this 
problem to do the right thing.  But, this has happened after CR and before 
PR.
</quote>

My answer is that if its a rewriting to add enumerations and remove
the use of QNAME, we can do the change. If it's adding new features, 
we need to go back to Last Call. I will wait for my other colleagues
to verify.

The second question you asked was how can we update the schema in
datespace. As the schema is tightly associated with a spec, we can't
publish one without the other in those final URLs. Look at the spec
and see how much it quotes the schema. It just becomes confusing if
the spec says one thing and the schema another. 

In this case, IMO, we would need to publish a new edition of the spec
that integrates the change so that the schema is aligned to it.

If you follow the 2002/03/xkms link, you'll see it actually points to
the XKMS CR. In fact, it was pointing to the WD before I realized this
and updated the spec. I have also asked for clarification to know
where that URL should be pointing to, to the schema or to the spec.

I think that nothing keeps us from continuing adding the schema changes,
even if the schema is not officially published until we get to PR.

Maybe I'm missing something and something is breaking down because of
this (validation, ... ?). What is the reason why the schema
needs to be updated right now outside of the draft?

Hope this helps! I forwarded your complete question to my colleagues and
gave some background too.

-jose

Received on Monday, 11 October 2004 14:33:07 UTC