Re: Time to reconsider namespace URIs?

I certainly don't seek to reopen a complex, lengthy debate.  The choice of 
a new namespace seems to me like a simple binary decision, and if nobody 
else thinks it's worth considering then I'm happy to let it drop.

On the other hand, I don't think we'd be doing proper service to the 
community if we didn't at least consider the consequences of what we have 
done so far.

The "changes" here are not primarily syntactic in nature, and have already 
been made, at least with respect to a considerable body of RDF published on 
the Internet.  I refer mainly to Mike Dean's DAML work:  if I understand 
correctly, large amounts of this data assume schema-derived datatyping of 
literals, and the decisions on datatyping mean that these are no longer 
conformant with the clarified RDF specification.

So I thought that updating the namespace URI would lay the basis of a 
migration path for this body of data:  the interpretation of literals in 
the current (old) namespace remains ambiguous, and interpretation in the 
new namespace is subject to the new stricter rules.  This means that Mike 
Dean's data, and others, isn't overnight declared wrong, just 
ambiguous.  If and when old data is brought into conformance with the 
clarified specification, updating the namespace URI will be an indication 
that can be exploited by inference engines, rather than trying to work out 
in some unspecified way whether the data they're reading is legacy data for 
which expected entailments may not hold, or new data for which they can be 
assumed.

#g
--

At 02:10 PM 10/21/02 +0100, Dave Beckett wrote:

> >>>Graham Klyne said:
> >
> > I'm beginning to think we may have crossed that line now with the latest
> > round of decisions.
>
>No.  We have made minmal changes to the M&S syntax-related terms.
>
> > I don't have my thoughts together right now, but I think it was some of
> > Mike Dean's concerns that made me think that maybe we should reconsider
> > adopting a new namespace URI, so that the volumes of existing RDF data can
> > be migrated rather than invalidated.
>
>What concerns?  At this point, I'd really ask you not to open new issues.
>This will delay all our work by months.
>
>Dave

-------------------
Graham Klyne
<GK@NineByNine.org>

Received on Tuesday, 22 October 2002 00:27:30 UTC