W3C home > Mailing lists > Public > public-awwsw@w3.org > March 2011

Re: a different way to interpret URIs that confuse IRs with their subjects

From: Jonathan Rees <jar@creativecommons.org>
Date: Wed, 30 Mar 2011 10:57:40 -0400
Message-ID: <AANLkTim9Ofc4Ynb9W4O_80MovRrVUsCJjKFsCY-ELaqp@mail.gmail.com>
To: AWWSW TF <public-awwsw@w3.org>
and for those of you who like this kind of this... the approach is
reminiscent of two things

(1) a monad construction, with rdfs:isDefinedBy (or maybe its inverse)
as the functor

(2) coercions in Algol 68... which are only barely consistent.  In the
1968 version of the language there was a 'proceduring' coercion which
had an apparatus for preventing inconsistencies and loops that was so
complicated probably only van Wijngaarden understood it; it was
explained by a fabulously contorted diagram.  Proceduring was removed
for the 1972 version of the language, but other confusing coercions
remained such as dereferencing. Even the updated spec had to
distinguish five different classes of evaluation context for the
purpose of sorting out which coercions were allowed where.

Jonathan

On Wed, Mar 30, 2011 at 10:16 AM, Jonathan Rees <jar@creativecommons.org> wrote:
> I've been considering a different way to describe the approach
> (advanced by Ed Summers and I think by Harry) where you say that there
> is no conflict between saying that a single URI refers to both a canoe
> and a document. Instead of interpreting "node" URIs as referring to
> chimera entities, as in the current draft, you redefine all of your
> properties so that they automagically coerce documents to the things
> that they're about.
>
> Attached is a picture of how this works. We start with graphs created
> by two parties with incompatible models: one of them thinks the URI
> refers to the document, the other thinks it refers to a canoe. A third
> party tries to interpret the merged graph. All of the relevant
> properties get extended so that they coerce an IR to the thing it
> defines.
>
> You'd have trouble dealing with statements of domain and range of
> these properties; this would require reinterpreting domain and range
> classes to include IR (or some subset of IR) as a subclass.  And if
> there's any hint of functional properties, disjointness, etc. the
> whole thing falls apart, but that's acknowledged - the people who like
> this kind of thing also say they don't care about inference.
>
> I have no idea what a proof of soundness of this approach would look
> like, but maybe the burden of proof would be on someone who likes this
> approach, not on us.
>
> But still this approach might be easier to explain than a "chimera"
> interpretation, which of course suffers most of the same drawbacks.
>
> I know most of the people on this list are not keen on this approach.
> I actually sort of like it, in a perverse way. In any case it has to
> be represented and explained somehow. I'm just trying to find the
> simplest possible way.  I could just cop out and say "maybe Carol can
> interpret this mess without making mistakes" and leave it at that.
>
> Jonathan
>
Received on Wednesday, 30 March 2011 14:58:13 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 30 March 2011 14:58:14 GMT