Re: AS&S - next steps

From: Jeremy Carroll <jjc@hpl.hp.com>
Subject: AS&S - next steps
Date: Fri, 24 Jan 2003 06:30:14 +0100

> 
> Hi Peter
> 
> this is the beginning of the offlist thread on how to move forward.
> I find it at lot easier if we keep all the discussion on www-archive, mainly 
> because I don't like getting thru the HP firewall. Procedurally it's also 
> better given its a public WG.
> 
> I suggest that:
> - we try to characterise where we differ
> - we try to generate a plan of action
> - I write most and you yes/no given our current availability for OWL
> 
> Proposed outputs:
> - two or three new ISSUEs for the issue list (possibly all about annotations).

Has this been done already?

> - a workpackage to proceed on OWL DL syntax in graph form.
>   (i.e. an ACTION for me to do and you to review, your review scheduled to fit 
> in with your other commitments)
> - a mutually acceptable decision on rdf:XMLLiteral
> 
> =====
> 
> My understanding is that we agree on the main thrusts of AS&S, and also tend 
> to agree on detailed bug fixes once it is clear what we are trying to do.
> 
> However, I think we have differences at the "what we are trying to do" level - 
> my current take is we disagree:
> 
> - about the importance of simplifying the description of OWL DL RDF Graphs
> - about how to resolve the tension between a direct description and the 
> implicit description

I don't know what a direct description or an implicit description are.

> - about rdf:XMLLiteral   - we probably should have the technical discussion as 
> to why I want it, and why you don't - that may be one we could have on 
> webont, rather than off-list. This feels like an issuette rather than an 
> issue, i.e. not really worth the full ISSUE process. (except I really can't 
> live without it!).

Well, my view is on this is precisely that OWL Lite and OWL DL processors
should not have to do XML canonicalization, or understand anything about
XML documents.  If there is a way of keeping rdf:XMLLiteral under these
conditions I would like to hear about it.

> - about annotations in general
> 
>   my position roughly: try and do as much as RDFS - i.e. free and easy, but 
> there will have to be some reduction somewhere, and the annotations *do* mean 
> something (not a lot - but not nothing).
>    my understanding of your position: do as little as possible, because there 
> is a danger that users might think the annotations actually mean something

Agreed.

Also, there has to be an easy way of distinguishing annotations from other
stuff.

>    So I see you as comfortable with simple text valued annotations as long as 
> they are not understood via rdfms-assertion.

More or less.

> Summary of my current understanding of preferred positions:
> 
> Syntax:
> pfps: minimal, as in AS&S editor's draft (or probably a bit less)
> jjc: annotation properties can participate in rdfs:subPropertyOf with each 
> other, can be themselves annotated; and can be used on individuals.

Annotations are already possible on individuals.  

> Formal Semantics
> pfps: none? 
> jjc: RDFS compatible 
>     The thread "Re: Annotations and entailments" died a death - on 
> rdfms-assertion.

My view on the semantics of *annotations* (except imports) is that they can
be no more than isolated triples.  I would, of course, prefer no meaning.
RDF(S) compatible means a *very* heavy meaning for annotations.

> Social Semantics:
> pfps: none
> jjc: rdfms-assertion ('with a vengence')

Agree to disagree.  Good luck implementing social semantics.  :-)


> Jeremy

peter

Received on Wednesday, 29 January 2003 15:29:16 UTC