Re: incorporating datatypes into the MT

On Wed, 2002-10-30 at 22:19, pat hayes wrote:
> 
> Er..... guys, I need guidance. I was under the impression that our 
> editoral task included incorporating the various aspects of 
> datatyping into our various documents, and that rather than being a 
> separate appendix, as it were, to RDF, that datatyping was now to be 
> fully integrated into the main thread. In the context of the MT, this 
> means that datatyping is pretty much the first thing that gets 
> mentioned, since one needs it to define what a typed literal means, 
> and one needs that in order to state the basic triple semantics for 
> RDF in section 1.5. In other words, in the document I am now working 
> on, there will be no such thing as a non-datatyped interpretation: 
> datatyping will be built into the very foundation of the language. 
> RDF will *include* datatyping.
> 
> Recent messages from Dan C and Jeremy and Jos, however, have made me 
> realize that some of us apparently expect the MT to be structured 
> rather like it has been in the past, in that there would be a simple 
> basic RDF notion of interpretation which had no such built-in stuff, 
> and datatyping would be one of the later additions.
> 
> So my question is, will incorporating datatyping into the basic RDF 
> MT cause anyone grief?

Well, it would cause me disappointment with respect
to the expectations I came away with from our
18 Oct meeting:

====
  Proposal: to distinguish between RDF-entailment (without DT knowledge,
  where DTyped literals in the abstract syntax don't have an equality
  based on value mappings) and a parameterised RDF+DT-entailment
  (that is, parameterised by the set of datatypes involved in the
particular
  entailment)

  There seemed to be general agreement that this distinction was
  important and useful.

http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Oct/0220.html
====

Or... hmm... would I be pleasantly surprised, rather
than disappointed?

i.e. what you're talking about now is consistent with
what I expected as of a vew days previous...

# details of rdf:datatype? Dan Connolly (Mon, Oct 14 2002)
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Oct/0133.html

but the responses to that message showed
that other WG members were expecting something else... more
like what we discussed on 18Oct.


> In particular, will it break the proposed API 
> designs apparently being developed?

I suspect it will, but I would need time to investigate
the details...

Ah! yes, now I recall: it's fine
as long as your API knows about all the datatypes
it's ever going to see in an RDF/XML document.
But as soon as the code reads a document with
a datatype URI that it doesn't recognize, it
sort has to throw an "I can't make sense of
this literal" exception; I don't see any
way to continue gracefully, i.e. without
acting non-monotonic.

That really doesn't bother me that much;
it means that datatyping is less
flexible/extensible than it appeared
in many of our discussions... but I can
live with that.

(if that looks like a jumbled mess rather
than a coherent answer to your question, you have
accurately understood the level of clarity
in my head at this point.)

> Because if so, we have some hard 
> thinking to do. I really don't see how I can make sense of typed 
> literals without talking about datatypes and datatype mappings.

Yes, quite; hence my hesitation to get into all this stuff,
and my rdfs:format proposal.

# a low-impact datatypes proposal: rdfs:format
Dan Connolly (Thu, Oct 03 2002)
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Oct/0031.html

I'm starting to wonder if working out the details
of our datatypes decision is giving me enough
new information to ask to re-open it. Hmm...
certainly not before the next WD, but after that,
I hope we'll all take a deep breath and re-evaluate
what we came up with.

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Thursday, 31 October 2002 00:06:43 UTC