Re: Fwd: Re: LANG: owl:import - Two Proposals

Jim Hendler wrote:
> 
[snip]
> 
> thanks Jeff - all - here's a message I should have sent to you all -
> esp the call at the end for other people to weigh in on this issue!
> 

Here's one thing Jim and I do agree on! ;-)  We need more input on this
issue, because I doubt Jim or I are going to convince each other.

> >
> >Jim Hendler wrote:
> >>
> >>  >Please see my responses inline...
> >>  >
> >>  >Jim Hendler wrote:
> >>  >>
> >>  >>  At 10:46 AM -0400 9/30/02, Jeff Heflin wrote:

[snip]

> >>
> >>  Jeff -it my seem like I'm quibbling - but it is important.  If we put
> >>  the RDF in an XML document, then it is no longer a "RDF/XML" document
> >>  (by definition) - it is an XML document.  This violates the decision
> >>  made by the WG earlier that our transfer protocol will be RDF/XML,
> >>  and that XML will bw non-normative.  I think reopening that issue is
> >>  problematic at this late date, and the above would require the WG to
> >>  rethink earlier decisions, which is a problem to me given we have an
> >>  alternative.
> >>

Maybe I'm quibbling now, but from the minutes of the F2F2 [1], the
decision was:

RESOLVED:

    1.that there is a presentation syntax and an underlying syntax and a
transform 
    2.some form of presentation syntax is requirement 
    3.RDF is underlying syntax 

This decision does not say that RDF will be the root node of all OWL
documents. It doesn't even say that RDF will be the presentation syntax.
I think that proposal #1 stays as close as possible to the intent of
this resolution without running into some thorny problems (the various
drawbacks of proposal #2).

[1] http://www.w3.org/2001/sw/WebOnt/ftf2.html#What:1

[snip]

> >>
> >>  okay, analogy wasn't great - but I think you underestimate RDF use -
> >>  however, that's an external issue.  I think the important thing to me
> >>  is that the ontologies and instances we create in OWL get pulled into
> >>  a graph, that can be manipulated separately from using a reasoner on
> >>  the ontology per se. We have LOTS of examples of this sort of OWL
> >>  tool growing - and thus whether we say "RDF" per se, or RDF-like
> >  > graphs, is indeed quibbling - the idea of saying you have to
> >>  reprocess graphs to use them in OWL makes me nervous indeed.
> >>

Well let's pretend for a moment that all that was important was RDF
graphs and that that documents don't exist. An OWL application has some
number of graphs representing various ontologies and instances. In order
to do reasoning, it merges some selected subset of these graphs that are
relevant to the task. In this environment, imports can be interpreted as
saying if you merge in this one graph, then you should merge in this
other one as well. However, in such an environment you will still need
at a minimum to have a URI associated with each graph (so that you know
which graph to include when you see a particular imports statement).
This information will not be in your graph so you are already relying on
something that is somehow outside of RDF. The point is you cannot
completely represent the necessary information in RDF, and until RDF
allows you to make statements about graphs, you won't be able to.

[snip]

> >>  >>
> >>  >>  I suspect we could discuss with RDF Core the idea of their being
> >>  >>  something in the <rdf:RDF> to help - perhaps an "RDF-Profile" or a
> >>  >>  pointer to an (optional) "RDF Header" graph - if it appears we need
> >>  >>  such mechanisms I'll be happy to bring that up in the SWCG - I'm not
> >>  >  > yet convinced we couldn't do this with RDF as is.
> >>  >
> >>  >Such an approach may alleviate many of my concerns with proposal #2. If
> >>  >the group is leaning strongly that way, we may wish to investigate this
> >>  >further.
> >>
> >>  We can discuss this at the f2f - I'm hoping Dave Beckett will join us
> >>  for at least the social events, and we (either you and I, or the WG)
> >>  might wish to discuss this with him.

Unfortunately, I can't be at the f2f in person. I will be calling in for
the imports discussion and the telecons though.

[snip]

> >>  >
> >>  >No, I don't think I missed the argument. I just have a different idea of
> >>  >what it means to parse an OWL document. You think the only result is a
> >>  >set of RDF triples. I think the results is a data structure which
> >>  >consists mostly of RDF triples, but there might be a few extra
> >>  >components as well (such as imports information or your pointer back to
> >>  >the original document). If you are storing this in a database (or Parka
> >>  >for that matter), then you might have one or more tables for storing the
> >>  >RDF triples and then another "meta"-table for the imports information
> >>  >(this is basically what I did with SHOE in Parka). If you then want to
> >>  >exchange this information with another application then you should
> >>  >either use the OWL presentation syntax or define custom data structures
> >>  >and/or file formats that preserve all relevant aspects of the language,
> >>  >whatever they may be.
> >>
> >>  i.e. we should lose interoperability, defeating the whole purpose of
> >>  moving towards standards

Well sure, if you decide not to use the exchange syntax (whatever it may
be) then you risk losing interoperability. If you exchange triples in
Parka's syntax (or N3 for that matter) then you are not conforming to
the standards and have lost interoperability. My point was if you want
to exchange OWL information with other people then use OWL's syntax. If
you just want to pass it between your own applications, then use
whatever you like, but make sure you pass all relevant information, or
at least that you don't care about the stuff you don't pass.

Jeff

Received on Thursday, 3 October 2002 10:32:57 UTC