RQ 011-014 Taxonomy items - Minutes 20031014

[Mary Nishikawa <nisikawa@fuchinobe.oilfield.slb.com>]

Feedback on RQ 011-014 Taxonomy items

 From minutes:
 >>..we discussed whether relationship information would impact the 
validation of the various value sets (OWL may require new constructs).

*Mary Nishikawa
Yes, I would say so. It would depend on the kind of validation you are 
referring to. I was under the impression that when you speak about 
validation, you are referring to validation against a W3C Schema -- so this 
is a structural validation, not a semantic validation requirement. You 
would need to trust the publisher on the validation of the contents before 
publishing semantic content, or you need a full-fledged constraint 
language, query language, and support for them in your system. You won't be 
able to do this with XQuery either .

 >>Would the same hold true for XTM? At issue is whether UDDI has a 
requirement to keep the  relationship information.

*Mary Nishikawa
I need to know what you want kind of information as relationships you want 
to validate. I will give you an example:

http://psi.oasis-open.org/iso/639/

The OASIS GeoLang Published Subjects TC has this draft on the OASIS site. 
ISO will take over the publication of this set, once we finalize the standard.

There is no relationship information here -- only core assertions -- 
essentially typing of information.  I can see a great benefit in UDDI using 
these kinds of XTM files, and they should be published by the authorities. 
They will be soon, and by the time your V4 comes out, this should be ready 
for publication by ISO. This seems very reasonable to me.

Some terminology for you:
The published subject information set is here
http://psi.oasis-open.org/iso/639/  (this is probably comparable to your 
overviewURL)

The published subject indicator is what a person can read to determine what 
the subject is. (in the published subject information set)
English Name: Catalan
French Name: catalan
Published Subject Identifier: http://psi.oasis-open.org/iso/639/#cat
A2: ca
A3B: cat
A3C: cat

If you take a look at the resource, you will get the general idea, I think. 
There are links to the XTM representations.

You can do a structural validation against the constraints specified by the 
DTD, XML Schema, or Relax NG, but that does not mean you can check if the 
publisher has valid, semantic instances (here, languages that belong to 
this ISO set)  in his or her set. That is why we are working on a Topic Map 
constraint language, to be able do "semantic" validation. It doesn't seem 
to me that you can get these kinds of complexities in UDDI yet on the 
server; however, this is reasonable to assume that clients can validate 
their taxonomies this way.

Even for the UN-SPSC, there can be various approaches to represent it in 
XTM. I gave one in my Montreal paper last year.
http://www.idealliance.org/papers/extreme03/html/2002/Nishikawa01/EML2002Nishikawa01.html
I think that for UDDI, you would need to take a very conservative approach. 
My paper has relationship information, but it is very basic, and does not 
go beyond what the UN-SPSC already has. However, if you want to make use of 
this information on your server,  and want to do merging, querying, 
semantic validation, which was the whole point of representing the UN-SPSC 
in XTM in the first place, you would need a topic map engine on  your UDDI 
server.

The same related scenario would be true for RDF/OWL, I would guess.

I hope this is useful for your meeting and I have understood the nature of 
your requirements. Sorry that I can't be there.

Cheers,
Mary