Closed Containers (was Re: ADMIN: Agenda/Logistics April 18 Telecon)

DanC:
> ACTION (Mar 28) DanC, Lynn Stein (prov.), Jos De Roo, to participate
> in RDF core discussions on construct for closed lists

> hmm... perhaps overtaken by

> action (Apr 8) : PeterPS + FrankVH + JeremyC write down a paragraph for
> RDFCore w.r.t. these issues [1) LIST 2) DARK TRIPLES]

No, not at all (perhaps socially).

The dark triples and closed containers are orthogonal issues.

The closed containers issue is "why do we have both daml:container and
rdf:Bag?"

RDF Core seems to want the minimum change needed to rdf:Bag for it to
satisfy WebOnt's needs.

So far their have been two proposals:

Pat [1] suggested an end marker so that rdf:_4=<end> would indicate that the
Bag was closed with only the members given by rdf:_1, rdf:_2 and rdf:_3.

Jeremy [2] suggested a size property so that rdfx:size="3" would indicate
that the Bag had precisely three elements. ([2] also suggests a model theory
for RDF containers that attempts to derive appropriate entailments).

We have not yet started on a comparison of the pros and cons of these two
approaches.
(I think in a nutshell that the first fits in well with current RDF
containers when used as envisaged in M&S; the second is significantly more
flexible and less commited to the current approach to containers, but seems
to have integers within the graph that isn't particularly pleasing).

Problematic test cases occur like:

A
<rdf:Bag rdf:about="#ex" rdf:_1="a" rdf:_2="b"/>

B
<rdf:Bag rdf:about="#ex" rdf:_2="a" rdf:_1="b"/>

My reading of M&S says that these are different ways of writing the same
bag, how do we do the closed container stuff in such a way as to continue to
permit that reading. Or perhaps we want to prohibit that reading, (despite a
reluctance to clairfy containers in general).

Worse is

C
<rdf:Bag rdf:about="#ex" rdf:_10="a" rdf:_99="b"/>

which, on my reading of M&S, is still the same bag.

Adding a size property to A and C can be plausibly thought of as equivalent
([2] shows how). Adding an end marker to C at rdf:_3 makes it an
inconsistent Bag, and hence different from A.


Jeremy

[1] (near end of message)
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Apr/0063.html

[2]
http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2002Apr/0095.html

Received on Thursday, 18 April 2002 06:16:52 UTC