RE: AS&S and WG consensus (was Re: abstract syntax and RDFS)

> This sort of late game posturing is what can make good specs go bad. You
> need to give technical reasons why you "can't live with" what has been
> done -- coming from a person who at present doesn't give a hoot
> either way.
> Shrug -- I've never developed any sort of strong affinity for
> rdfs:seeAlso -- seems like a wishy washy sort of thing to me. What is your
> technical thinking?

I am sorry that I am posturing.
I do have genuine concerns about how well the document matches WG
consensus - which is achieved through following, not to the letter
admittedly, some process.

I had deliberately decided not to offer any reasons, since, as far as I can
see, such discussion is out of order - it is about topics that have never
been raised as issues.

However, since you ask -

I hope that semantic web users will be able to migrate from RDF thru RDFS
thru OWL Lite thru OWL DL thru OWL Full - possibly stopping off where their
needs are met; possibly skipping a stage if they have specific needs that
are only met at higher levels.

Every feature of RDF and RDFS that we decide is inappropriate in OWL Lite
hence has a cost. Another user will find migrating more difficult. We need
to make a case that there are significant benefits in doing so.

For many of the features of RDF this case has been made, primarily under the
semantic layering issue - I have often resisted dropping RDF features, but
have been open to reasoned argument.

Simply dropping things from the RDFS and RDF vocabulary, apparantly on an
editorial whim, does not involve any sort of "case".

So I can't live with OWL Lite prohibiting rdfs:seeAlso and rdfs:isDefinedBy
because of the additional migration cost for users moving from RDFS to OWL
Lite. I offer no other defense of these properties - I am not sure that
there is much of one (I agree with your "wishy washy").

This concern about increased migration costs also applies to rdf:XMLLiteral.
However, I am also prepared to defend that on internationalization grounds.

I can illustrate this with respect to one of our requirements:
 R18. User-displayable labels
The language should support the specification of multiple alternative
user-displayable labels for the objects within an ontology. This can be
used, for example, to view the ontology in different natural languages.

Motivation: Internationalization, Ease of use

The W3C has significant expertise in the area of user-displayable text in
different natural languages. The state of the art is to use xhtml markup for
certain advanced features such as bidirectional text (using xhtml span
elements with dir attributes) and ruby annotation [1]. Ruby is particularly
relevant to labels. Integrating this with OWL is done using
rdf:parseType="Literal", which uses the rdf:XMLLiteral datatype.

I use this route in constructing the OWL Test Cases document from the
manifest files, since I wish to use mark up in the test descriptions.

In summary, it is important to me that the migration path works, which means
that RDF and RDFS features should not be arbitrarily dropped from OWL Lite.

I am certainly prepared to make a more substantial case if anyone really
wants to have a discussion on the merits or not of having rdf:XMLLiteral in
OWL Lite; but it should be up to them to propose it, rather than to me.

> I am a little peplexed why folks who were all hot on OWL full in the first
> place cannot live with a limited OWL Lite or OWL DL.

The WG suffers from folks who only care about OWL Full - this way we will
end up with three different languages rather than three sublanguages of one

> It is "Lite" for a
> reason, no?

Yes, essentially semantic ones.

> Jonathan



Received on Thursday, 23 January 2003 08:19:40 UTC