Re: FragIds in semantic web (ACTION-543)

On Thu, Apr 28, 2011 at 5:06 AM, Graham Klyne <GK-lists@ninebynine.org> wrote:
> Tim Berners-Lee wrote:
>> We need to stop people letting connect go to seed, and building
>> systems where hypertext and data are muddled and not both accessible
>> unless
>> on the client th e user can control the accept headers.
>> Conneg should only be used where the data is identical in all cases (or a
>> subset due
>>  to the limitations of one of the languages).
>
> I'm probably missing something here, but in the context of the preceding
> text, esp. the distinction between cases B and C, this seems to discourage
> the common(?) practice of publishing both HTML and RDF at the same URI,
> selectable by conneg?
>
> (My take has generally been that in such cases hypertext and RDF should
> generally describe the same things, for human- and machine- consumption
> respectively.)
>
> #g

What Tim says does *not* discourage HTML/RDF conneg - except in the
case where the RDF and the HTML carry different information
(translated into HTML for human consumption or RDF for machine
consumption in the same way that any document might be translated into
different natural languages or document formats). Unfortunately, in
practice, the two variants almost never carry the same information
(e.g. compare the HTML and RDF versions of FOAF). This is very
difficult to arrange even if you want to do it. So in effect you are
right. Personally, I have been advising against HTML/RDF conneg for
this reason.

Putting different information in different conneg variants is asking
for trouble, because we use URIs to name the generic resource, not
individual representations. You don't know which variant your
interlocutor will be consulting, so whatever you state or assume about
the generic resource has to hold for all its representations. E.g. if
I say that <x> says the domain of <x#y> is z (in any language or
notation), then *all* current representations of <x> had better say
that the domain of <x#y> is z.

Ideally we would have a more appropriate protocol for providing a
human-friendly view of an ontology or namespace document. XSLT is one
candidate, a statement like <> :presentation <...> would be another,
web linking (Link: or .well-known) might do it, RDFa would do it.

Jonathan

Received on Thursday, 28 April 2011 11:58:46 UTC