Issue of support of Bag/Alt in RDF

Gentlepeople, this is to raise the issue of support of Containers in the formal semantics of RDF, a.k.a. "Bag/Alt issue".
BTW, apologies for not having had time at all to reply in time to the corresponding prior thread on this (cf.
http://lists.w3.org/Archives/Public/www-rdf-comments/2003JanMar/0180.html ) and neighbors, but I didn't have time at all (in the
deepest sense of the world "didn't"...).

I don't have time even now, but the comments deadline is approaching, so I'll sketch the issue and ask for its consideration.


*****


Issue: support of Bag/Alt now is not present in the semantics, at the level of entailment.
RDF graphs that would seem to be equivalent (like differing in the order of the rdf_n's in a Bag container) are not semantically
equivalent for the RDF Semantics (as defined by the entailment relationship).

This would lead to think support for Bag/Alt's has been dropped, with consequent discussion on the reasons (why) and impact on
backward-compatibility.
But in fact the situation is less clear, as the RDF Semantics doc talks about an "intended mode of use" for containers, and then
goes on citing with natural language the semantics of Bag/Alt (which is not reflected in the entailment).

This leads to two three sorts of problems:
+ first (inadequacy of the semantics); if the "intended semantics" of RDF is something (including Bag/Alt in their intended mode of
use), and the RDF formal semantics (entailment) doesn't/cannot support it, then there is a problem with the current semantics.
(Otherwise, what is the goal of having a semantics...?)

+ second (interoperability); are reasoners app's allowed to do the kind of graph transformations allowed by the "intended mode of
use" of the Bag/Alt (like, say, shuffling a bag's elements)? Can they be considered to be behaving perfectly fine (i.e., to
correctly process RDF), or such use is deprecated, as it produces things that are in no relationship, according to the entailment?

+ third (precision); while the kind of intended-mode-of-use graph transformations are rather clear in the Bag case (permutations),
they're not that clear in the Alt case. Because, this kind of "intended use" transformations should in any case be specified at
formal level, and not just left in natural language as now
(quoted: "things of type rdf:Alt are considered to represent a collection of alternatives, possibly with a preference ordering").

There are deeper implication/technicalities esp. behind the first point, but this email is just to raise the issue, not to discuss
it.

Thanks,
-M

Received on Friday, 21 February 2003 14:01:13 UTC