Telecon 2001-11-02 Draft minutes

RDFCore WG minutes for the Telecon 2001-11-02

Transcript:
	http://ilrt.org/discovery/chatlogs/rdfcore/2001-11-02
Agenda:
	http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Nov/0007.html
  "Datatypes" was the only item for discussion.

Roll call -

Participants:
   - Art Barstow
   - Dave Beckett
   - Jeremy Carroll
   - Ron Daniel
   - Bill dehOra
   - Jan Grant (scribe)
   - Martin Horner
   - Graham Klyne
   - Frank Manola
   - Brian McBride (chair)
   - Eric Miller
   - Mike Dean
   - Pat Hayes (thankfully recovered from 'flu)
   - Sergei Melnik

Regrets:
   Dan Brickley, Dan Connolly, Jos De Roo, Steve Petschulat,
   Aaron Schwarz

Absent:
   Frank Boumphrey, Rael Dornfest, Yoshiyuki Kitahara, Michel Kopchenov,
   Kwon Hyung-Jin, Ora Lassila, Satoshi Nakamura, Pierre Richard,
   Guha.

Also absent was Pat Stickler (who will be a new member of the WG
representing Nokia and is likely to be present in the new year).

Review minutes of previous meeting:
	http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Oct/0581.html
	http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Oct/0582.html
	http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Oct/0595.html

   In our eagerness to dive into the datatyping discussion, this was
   overlooked (postponed until next meeting).

ITEM Data typing:

   (Scribe attempted to make a reasonably complete transcript; an
   attempt at summary is given here. I use the abbreviation DT for
   "data type". Technical corrections may appear as followups to
   this and are invited if I put things badly)

Pending Sergei's arrival, Pat gave a brief summary of his proposal,
outlining its advantages (there's no need for a special treatment of
literals), requirements (URIs denoting DT classes are distinguishable
from URIs denoting other arbitrary classes), and problems:
  1. in the absence of any datatyping information, the graph is
     ambiguous
  2. permitting literals as subjects means that DT information
     can be indicated (eg, with rdf:type arcs).

Pat also reiterated that his new MT used graphs tidy on resources
(ie, nodes labelled with URIs), but not tidy on nodes with literal
labels.

Pat then began to outline his perceived problems with Sergei's proposal,
at which point Sergei made a timely arrival and demonstrated the idea:
that instead of
	foo eg:someproperty _:bn
	_:bn rdf:value "12"
we use
	foo eg:someproperty _:bn
	_:bn xsd:integer "12"
thus indicating the interpretation to be used for "12".

Sergei indicated that a high-level problem with the discussion was that
we lacked a clear idea of use cases, scope, and intended deliverables.
General consensus was that we might look for a mechanism for embedding
datatyping schemes in RDF, rather than spend the effort doing such an
embedding for (say) XMLSchema.

Jeremy suggested a case to consider: inconsistent or ill-formed
documents that gave incompatible typing for a literal, and that that
seemed to be a schema question: can an object be a member of both these
classes at once?

Other cases were discussed: in particular, xsd:integer (or "the
integers" as a data type) and "month" seemed examples; one being
finitely enumerable.

At this point the chair stepped in with a suggestion that Pat and Sergei
collaborate on an informal document (ie, outlining the various
options) that might be circulated to implementers to ask for feedback.

Pat was reluctant to scope the document beyond the WG in the first
instance, citing an order of magnitude more work being involved to make
the document ready for general consumption.

Jan said that while Pat's proposal seemed good, he had some doubts as to
API and storage issues, asked that the document might sketch API
illustrations for proposals where appropriate.

Graham said the CCPP might be a useful source of use cases.

The following actions came out of this discussion:

   ACTION: 2001-11-02#01 : Sergei, Pat, Jeremy: to collaborate in the
     preparation of a document (initially for circulation to WG,
     possibly beyond) outlining in a "primer" each of the proposals,
     together with examples of how the approach would work with
     a few use cases.

[Eric asked the time-scale for this. It is a short-term item to feed
back into this process]

   ACTION: 2001-11-02#02 : Graham: to see if he can extract DT use
     cases from CCPP.



-- 
jan grant, ILRT, University of Bristol. http://www.ilrt.bris.ac.uk/
Tel +44(0)117 9287088 Fax +44 (0)117 9287112 RFC822 jan.grant@bris.ac.uk
Whose kung-fu is the best?

Received on Friday, 2 November 2001 12:44:31 UTC