Re: Requirements Document

From: Jim Hendler <hendler@cs.umd.edu>
Subject: Re: Requirements Document
Date: Thu, 14 Feb 2002 18:00:48 -0500

> (Chair neutrality remains on for this one - these are attempts to 
> clarify what I think sense of the WG was at the f2f, and requests for 
> clarification with respect to the comments - in many cases I agree 
> with Peter's point, but want to make sure the WG understands what was 
> voted on at the f2f (since I took the notes) and make sure the points 
> match up)
> 
> At 2:03 PM -0500 2/14/02, Peter F. Patel-Schneider wrote:
> >From: Jeff Heflin <heflin@cse.lehigh.edu>
> >Subject: Re: Requirements Document
> >Date: Thu, 14 Feb 2002 13:23:11 -0500
> >
> >>  Peter,
> >>
> >[...]


> >>  3) You had a problem with the "Referencing with URIs" requirement. In
> >>  subsequent discussion with Dan, it appears that you would be happy if we
> >>  change the wording to "URIs + frag[]ment ids."
> >>
> >>  ACTION: Wording for the requirement will mention "frag[]ment ids."
> >
> >No, this would not make me happy.  In my discussion with Dan, I stated that
> >I wanted to be able to refer to anything on the web, and things outside of
> >the web, as well.  Unfortunately, URIs plus fragments does not get us
> >there, because XML Schema document definitions cannot be so addressed.
> >
> >I don't know why this unfortunate state of affairs exists, but the
> >*requirement* should therefore be, in my mind at least, the ability to
> >reference what we need to reference, including XML Schema document
> >definitions.  URIs plus fragments gets us 99% of the way there, and may be
> >what we end up using, but changing from what we want to a partial solution
> >for what we want is, in my view, not something that should be done in a
> >requirements document.
> 
>    The requirement says we must be able to reference with URIs, it 
> doesn't say anything about that being the only thing to use. I 
> believe this is consistent with what I heard at the f2f.  Also,  I 
> don't see what in your discussions makes you want to remove this 
> requirement.

The requirement says that 

	It must be possible to uniquely identify a term in an ontology
	using a URI.

This says to me that URIs have universal coverage for everything we
want to reference, which is just not the case, as it does not hold for XML
Schema document definitions. 

>   If you're suggesting we could add to the objectives the ability to 
> refer to things on and off the web, etc - that is fine with me and I 
> believe it would be consistent with discussion at the f2f, however, I 
> think what is written is a requirement that got an A vote at the f2f, 
> which means you did not vote against it previously -  Please clarify 
> what change should be made to the document and why you are changing 
> your vote if you are.
> (Remember we are talking necessary conditions, not sufficient).

I'm changing my vote for the simple reason that URIs (plus fragment IDs)
are not adequate for the purpose.  (Well, maybe you could concoct a URI
scheme along the lines of ``the XML Schema definition at character position
n in document d''.)

> >[...]



> >>  5) You believe the "Ability to state closed worlds" requirement is too
> >>  strong. Once again, this was a requirement that a significant number of
> >>  people think is essential. Obviously, we can't make a closed-world
> >>  assumption, but the ability to infer negative information from the
> >>  absence of positive information is useful. However, I understand your
> >>  concern in that this is not a feature found in existing langauges and
> >>  may be difficult to implement.
> >>
> >>  ACTION: We will demote "Ability to state closed worlds" to an objective.
> >>  If anyone wishes to argue for it remaining a requirement, please do so.
> >
> >It is not just that it is difficult to implement, but that the closure of
> >a general ontology is not well defined.
> >
> >For example, what is the closure of
> >
> >	John is a Person whose sisters include either Jill or Susan or
> >	whose brother is Bill.
> 
> my understanding of the requirement expressed at the f2f was that by 
> closing the world we wanted someway to designate a document (or other 
> designation of a set of Owl triples) and let a reasoner know that it 
> could consider this a closed world if it wished.  If you feel the 
> document doesn't explain this right, you should suggest wording 
> changes.  Also, I recorded that the discussion at the f2f suggested 
> this could be "extra-logical" (although to be honest, I'm not quite 
> sure what that would mean).

Well, yes, I am stating that the requirement is ill-stated.  I would be
much happier with wording something along the lines of:

	Some documents contain definitive information about a particular
	set of facts, such as the membership of a class.  When this is
	the case the language must be able to so specify.

However, I don't view even the above as a good way of stating the
requirement.

What is needed is some idea of what sorts of things can be closed and under
what conditions they can (and cannot) be closed.

[...]
> -- 
> Professor James Hendler				  hendler@cs.umd.edu
[...]

Peter F. Patel-Schneider
Bell Labs Research

Received on Friday, 15 February 2002 12:40:26 UTC