[ubl-ndrsc] Feedback on code list doc and template: UN/ECE,ebXML Reg/Rep

["Eve L. Maler" <eve.maler@sun.com>]

I've been communicating further with the UN/ECE folks on whether the 
code list rules work for them, and have also explored code lists further 
in the ebXML Reg/Rep context.  I wanted to give a brief summary here, 
and catalogue some comments that have come in.

- UN/ECE have brought their experimental code list schema modules into 
compliance with our draft rules.  You can see the results here:

http://www.unece.org/etrades/unedocs/repository/codelistintegration.htm
http://www.unece.org/etrades/unedocs/repository/codelist.htm

They have a canonical format for the "real" definition of a code list, 
and a doctored version that meets the needs of UBL and other code list 
consumers.

- They pointed out the following errors in the rules document and the 
template, carried over from the position paper days:

   . The simpleType should contain a restriction, not an extension, for 
supplying the enumerations.

   . The names of the types should not be QNames.

- They asked whether we have rules for the naming of their types, and I 
said no, as the names make no difference to our usage of them.  (Does 
anyone feel differently?)

- Finally, I had a conversation this week with my colleague Farrukh 
Najmi, who is involved in the ebXML Registry/Repository effort.  I 
discovered that the Registry Information Model (RIM) spec defines a 
vocabulary for "classification schemes" (of which country codes and 
currency codes are examples), and that anyone can use this vocabulary to 
mark up a canonical version of their code lists.  You can see V2.1 of 
that spec here:

http://www.oasis-open.org/committees/regrep/documents/2.1/specs/ebrim_v2.1.pdf

They haven't been very active in advertising this model, but even though 
it's fairly simple I think it's rich enough to accommodate the needs 
that code list producers have, e.g. to match up code values with names 
(even multiple versions of names in various languages, so that DE can be 
associated with both "Germany" and "Deutschland").

The ebXML Reg/Rep project in SourceForge even has some experimental 
versions of code lists done in this ClassificationScheme vocabulary.  A 
version of ISO 3166 is here:

http://cvs.sourceforge.net/cgi-bin/viewcvs.cgi/ebxmlrr/ebxmlrr/misc/samples/SubmitObjectsRequest_ISO3166.xml

(Note that the author was playing around a bit with the capabilities of 
the vocabulary, and gave the country codes a little hierarchy -- they 
know that's not standard.)

I can see where it would be useful to encode the "base/normative" 
version of a code list as an instance of the ClassificationScheme 
vocabulary, and then use XSLT to transform automatically -- and possibly 
even dynamically -- to the schema format required by UBL and other "code 
list consumers".  What do you all think of this idea?  We would need to 
make sure that the ClassificationScheme vocabulary has all the right 
ingredients for such a transform, but it nearly does already...

	Eve

-- 
Eve Maler                                        +1 781 442 3190
Sun Microsystems                            cell +1 781 883 5917
XML Web Services / Industry Initiatives      eve.maler @ sun.com