W3C home > Mailing lists > Public > w3c-wai-er-ig@w3.org > July 2001

Re: EARL In RDFS(FA)

From: Sean B. Palmer <sean@mysterylights.com>
Date: Sat, 14 Jul 2001 16:43:44 +0100
Message-ID: <011801c10c7b$c46d6960$d9da93c3@z5n9x1>
To: "Dan Brickley" <danbri@w3.org>
Cc: <w3c-wai-er-ig@w3.org>
> [...] Lots of researchers have looked at RDFS and found it
> OK, some others looked at it and didn't like it.

Um... in case you haven't noticed, I am one of these people who have
found RDF Schema to be "O.K." - i.e. "stable enough to be
implemented" - as the EARL 0.95 schema proves. But I am open-minded to
suggestions about how RDF Schema can be improved, and there is some
evidence to suggest that the most basic properties in RDFS should be
primitive. I'm not saying that I necessarily agree with that
suggestion.

Languages such as UML (which have proved very easy to implement) have
these properties as primitives. I'm sure you're also aware of the RDF
issue about this:-

   http://www.w3.org/2000/03/rdf-tracking/#rdfs-primitive-properties

and the works of Wolfgang Nejdl:-

http://lists.w3.org/Archives/Public/www-rdf-comments/2000AprJun/0041
http://www.kbs.uni-hannover.de/Arbeiten/Publikationen/2000/modeling200
0/wolpers.pdf

As this issue is sufficient enough to merit RDF Core to take a look at
it, I don't think you can hold a steadfast view that RDF Schema is at
its optimum without at least considering these viewpoints.

Once again, my personal opinion is that RDFS is well-written, and easy
to understand. But a closed mind outlook on that is not healthy for
RDF.

[...]
> [...] not all of the (now documented) faults of  RDFS are carried
> through to DAML+OIL (eg. they cleaned up range/domain stuff
> nicely), and it doesn't follow from the fact that properties have
> multiple uses that they're ambiguous.

Well, think about it. If they have multiple uses, then the argument is
that for any one instance of their use, you can't tell what particular
role they are being used for.

[...]
> domain and range is a known problem; the RDF Issue List includes
> this and other issues [...] I'm glad to say we've (at last) got a
> working group in place to fix this stuff.

The domain and range thing is a no brainer really - all
implementations that I know of treat multiple calsses and/or ranges as
intersections. How could it be any other way?

[...]
> > > Do you  know if RDFS(FA) is being considered by the
> > > RDF groups?
> >
> > I very much doubt it.
>
> Ahem. It hasn't been submitted to the WG, but I'll link it from
> the issue summary when we get to those topics. [...]

And what is the submission process to the RDF Core WG? Please point me
to some publically available reference material on the process.

[Mild digression]

Also, I know that RDF Core is only chartered to provide a "second
edition" type thing for RDF, but I don't think the logistics of the
future of RDF have been made clear enough to the public. Are the
layers of RDF to be arranged in different specifications, or are they
going to remain relatively the same, or what? This is something that
we need to be informed about.

> I can't read PDF files comfortably so it hasn't worked its
> way to the top of my reading list yet.

Who was it that said "if you're going to publish a press report on the
Web in PDF, you may as well bury it in your back yard"? :-)

[...]
> > > Also, I just realized that RDFS is a Candidate Recommendation
> > > - is that true?
> >
> > Yep, that's true. It's been like that for some time now.
>
> Yes, the Metadata Activity and old RDF Schema WG expired before
> RDFS was entirely finalised; we've now got the Semantic Web Activity
> and RDF Core WG (see charters at http://www.w3.org/2001/sw/) in
> place to complete this work.

And so five years down the line we might have a stable W3C
recommendation to implement? C'mon now, there's plenty of low hanging
fruit that should take RDF Core all of five minutes to resolve.
Multiple domains and ranges are to be interpreted as intersections,
and cyclic preoperties and classes are allowed, expressing
equivalence.

The only other stuff of any real concern is the "do sub properties
inherit range and domain semantics?". I'd suggest they they inherit a
sub class (i.e. which may be equivalent) to the original ranges and
domains. I have no mathematical proof for that, but a few simple
examples will show that this is common sense.

--
Kindest Regards,
Sean B. Palmer
@prefix : <http://webns.net/roughterms/> .
:Sean :hasHomepage <http://purl.org/net/sbp/> .
Received on Saturday, 14 July 2001 11:42:39 GMT

This archive was generated by hypermail 2.2.0 + w3c-0.30 : Thursday, 9 June 2005 12:10:39 GMT