IRIbis and OWL

I wanted to make sure that you know that a revision to RFC 3987 is in
the works. In the event 3987bis changes the set of syntactically
correct IRIs relative to 3987, this could be a problem for
interoperability between OWL (which cites 3987 normatively, see [1])
and other standards. I don't consider this to be a major risk, as the
problem cases, if they exist at all, will in my assessment be highly
unlikely. So my message is just a heads-up, not a call for any kind of

The IRI WG charter [2] suggests (to me) that compatibility is to be
maintained by vigilance, not by construction:

"* The IRI specification(s) must (continue to) be suitable
for normative reference with Web and XML standards from W3C
specifications. The group should coordinate with the W3C working
groups on HTML5, XML Core, and Internationalization, as well
as with IETF HTTPBIS WG to ensure acceptability."

You might want to try to satisfy yourselves that this gives adequate
protection for OWL.

If I understand correctly, 3987bis may change the way IRIs are
converted to URIs, even if it doesn't change the set of allowable
IRIs. This has no impact on the OWL or RDF specs as such a translation
never comes into play, but it could have an effect on deployed OWL

If the reference in RDF Concepts to "[IRI draft] or its successors"
[3] would allow an RDF processor to follow 3987bis, one could
conceivably end up with a situation of an RDF+3987bis processor unable
to interoperate with an OWL+3987 processor.

(On a related note I haven't verified that OWL/3987 IRIs are the same
set of strings as "RDF URI References"; if not then OWL and RDF
already have an interoperability problem.)

IRI issues have received a lot of attention in W3C. In the XML world
we have 'anyURI' as defined for XSD 1.1 [4] and many other documents
normatively referring to the LEIRI specification [5]. I don't know
whether LEIRI would be appropriate for OWL or RDF. Another hedge
against changing IRIs would be to adopt a generic policy for normative
reference to changing specifications, but no consensus has emerged on
how to do this or even whether it's a good idea.



Received on Thursday, 18 November 2010 22:02:16 UTC