W3C home > Mailing lists > Public > public-awwsw@w3.org > March 2010

Re: Harry Halpin's PhD dissertation

From: Jonathan Rees <jar@creativecommons.org>
Date: Wed, 17 Mar 2010 17:49:10 -0400
Message-ID: <760bcb2a1003171449m4ae3e33dn26be259fbe454045@mail.gmail.com>
To: Harry Halpin <hhalpin@w3.org>
Cc: AWWSW TF <public-awwsw@w3.org>
I wish we had more time to talk, as I too consider myself a
Wittgensteinian and I think figuring out why we appeared to disagree
would be informative.

Maybe this has to do with attitudes toward the specifications. I try
to be careful to say: "If you follows specs X, Y, and Z, then such and
such follows." Sometimes the antecedent is the background of a
conversation. I give weight to rule of law (hard to argue with city
hall), and to attempt to change laws that are wrong. I think where we
agree is that URIs and RDF should form a language with socially
determined meaning. The question is exactly what "authority" our web
browsers (which are slavish to ICANN/DNS/HTTP) will have in the social
system. It ends up being political; browsers are strong w.r.t. RDF
because people will use them as dictionaries. But then language is
always political...

I can't read through the whole dissertation, so will have to pick up
various pieces of it from time to time. One nit to pick is the
"NonInformationResource" class in IRW. I agree with TimBL that it
makes no more sense to have this class than to have the class
NonFlower or NonGazebo, since the IR/NIR split would not be at the top
of *any* upper ontology that I'm aware of, and no property unites all
of its members.

Except that your diagram shows NIR as the range of the two properties
IsAbout and identifies. I think the ranges of *any* property that
thinks it wants to have domain or range NIR needs to be Resource
instead. You should be able to name IRs using # URIs, or to use the
303 hack with them. Several reasons: (a) given the vagueness of the
definition of IR, you may have your hands on something and not even
know whether it's an IR or not - so you should be able to go the safe
route and use # or 303. (b) the IR may not be on the web at all, in
which case you have to describe it. (c) the description of the IR may
be more interesting or important than the IR itself, so # or 303 may
be the preferred result of an attempt to access. (d) <link> and Link:
create the opportunity to deploy descriptions of IRs, which means they
have no particular deficit with respect to the semantic web.

This is supported by the careful wording of the httpRange-14
resolution: "... with a 303 (See Other) response, then the resource
identified by that URI could be any resource".  In particular it could
be an information resource.

So please don't conclude from a # or 303 that it's not an information resource.

I realized something recently that should have been quite obvious
before, which is that the httpRange-14 decision is still meaningful
and (in the view of those who like it) necessary even in the absence
of the 303 hack. Without 303, when you name something, you would still
have to decide whether to name it with or without a #. It ought to
always be safe to use a #.


On Wed, Mar 10, 2010 at 7:37 PM, Harry Halpin <hhalpin@w3.org> wrote:
>> Sense and Reference on the Web
>> http://www.ibiblio.org/hhalpin/homepage/thesis/
>> on my reading list.
> Thanks and do send me any comments! Small changes can still be made :)
> Note that I'm happy to discuss this thesis (in particular, the distinction
> between "sense" and "reference" is crucial and very hard to make, as is
> the neo-Fregean/Wittgensteinian case for public meaning, which derailed
> the last conversation I had on this topic).
> Also, myself and Valentina will try to take another shot at seeing how the
> latest work from JAR can mix in with the IRW ontology. We're both rather
> busy right now, but would be happy to do a telecon on this in April.
>      thanks,
>            harry
Received on Wednesday, 17 March 2010 21:49:44 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:21:08 UTC