Re: Feedback on draft spec

[Terry Allen <tallen@sonic.net>]

Ravi wrote:
| Having reviewed the draft spec here is my feedback.
| 
| Based on our experience of having a ZIP file format for enabling schema
| submission although was a start, I do not see how having a registry
| specification which focuses on yet another and just a ZIP file packaging
| format for schema submission is going to help in providing a mechanism for
| solution design/developers to register, search, and use these schemas.

Thanks for your feedback, Ravi; you're the first to comment on the draft.

There are a couple issues here.

First, I erred in not labelling the zip format nonnormative.  It's
intended only to get the XML.org registry and repository up and running 
 - and if the folks who will be doing that work have a better idea, 
that's fine, too.  (The DTDs, except for the manifest, are
needed for normative purposes elsewhere in the spec and aren't
included only to support submission.)

Second, we hadn't identified this piece of workflow as something that
doesn't have to be the same across registries in order to enable
interoperability.  The interoperability we want is related to dealing
with entities already registered.  So I think we should add language
to that effect, and provide this (or an improved) submission format
as a suggestion, and perhaps move it to an appendix or another 
document.  Does anyone have preferences on this point?

You mention registration, search, and use.

Registration involves setting up business relations, submitting
the entities to be registered, and providing necessary metadata.
For Biztalk.org, just to give an example, I think this setting up
business relations maps to signing up as a member and reading the 
terms of business.  

Ravi, could you tell us about your experiences with zip as a schema
submission packaging mechanism, and how you are capturing metadata
about submitted schemas today for Biztalk.org?

Searching is to be done on an interface generated from the metadata
that is supplied by both submittor and registrar.  We've generated
lots of ideas about what would be good to have in an interface,
and some functional requirements, but the normative spec will have
only the core functional requirements.  Building a good interface
will be an interesting design challenge, but we'll leave it to
XML.org.

Ravi, if you have any advice about interface (I haven't seen the
Biztalk.org library), we'd be most happy to hear it.

When we come to use of the contents of the repository, I believe
XML.org and Biztalk.org may have slightly different requirements.
Both share the requirement that the repository has to be able
to serve a registered entity when requested (I have agreed to 
revise RFC 2469 to synch with later URN specs, and the query
formats in that document will be method of requesting an entity
that go into our normative spec).  But their requirements for
integration of the repository into a development environment differ,
insofar as I understand them.

My impression is that the contents of the Biztalk.org repository
are intended to be interoperable with each other (thus the restriction
to XDR, for example), and in a specific method of using XML.

XML.org, I believe, wants to enable reuse of entities in the
repository, not unlike Biztalk.org, but is much more latitudinarian
about the contents of the repository.  We'll have SGML in there,
for goodness sakes; we'll have competing schemas in the same domain.

So while Biztalk.org is naturally focussed on use (in XML development)
tightly coupled with tools, XML.org, which might be interested in
similar tools, must also provide a much less disciplined method of 
discovering what's in the repository, reusing it, and registering 
derivatives.  

Anyway, that's how I've thought of it; I'd really like to hear
from others, too.  Ravi, if you'd like to propose an alternate
method for submissions, you're welcome to do so; I hope you would
not oppose a tech spec that includes a nonnormative suggestion 
to use zip.

regards, Terry


regards, Terry