Re: New issue - Meaning of URIs in RDF documents (Tim Berners-Lee) writes:
>On the contrary, his aspect

Is that supposed to be "his" or "this"?  I find Patrick Hayes' approach
far more plausible on "let it break" as he seems to start out _without_
the Platonic pretensions.  If it's "his" I agree here, but it doesn't
seem coherent with the following sentences.

>of the semantic web has exactly that
>"let it break: flavor, in that in fact if there is no document or
>in fact there are local misunderstandings about what things mean,
>the web in general goes on.
>Those old hypertext systems required complete consistency across the
>But still both web and semantic web are based on a design where when
>you point to a document you can retrieve a document.
>The web design didn't allow a link to be deemed to point to whatever
>the reader wanted to philosophically argue it might be considered
>to point to.

I don't think most people expected URLs to morph into the philosophical
demilitarized zone of URIs.  From my perspective, the philosophy isn't
the fault of your opponenents - it's the necessary consequence of the
entire URI approach, especially as it enters new arenas for which it's
not clear it's appropriate.

It was tough for readers to argue with URLs.  It's very very easy for
readers to argue with URIs, especially when those URIs move between

>> Perhaps that approach qualifies as "perturbation theory", but I'd
>> always thought that the genius of the Web was that it simply didn't
>> bother with perfection, even especially at the URL/URI level.
>Your idea of why the web worked is just right, I think, and am
>in fact proposing the same thing for the semantic web.

I think you've wandered miles from that, to say the least.  You may
think you're proposing the same thing, but in fact you're building an
enormous system of pieces which don't accept breakage easily - and then
saying "well, it may not always conform to reality, so hopefully people
will fix it."

>(That was what I meant by perturbation theory - a theory which allows
>one to start off by considering what happens when it doesn't break,
>and then work out what happens when it does break by considering the
>local effect of deviations.  It works eg in phsics when considering
>what happens to the energy levels of an electron when the  atom is
>placed in a magnetic field. It does *not* work on logic, when someone
>writes just one small assertion that 1=2 - the whole thing falls 
>apart and you can prove anything.  On the semantic web  you deal with
>a subset of documents which don't say 1=2, and if you find they do you
>fix it.  You require a certain amount of consistency locally, and it
>is the combined efforts to produce local consistency which tend to
>help but not completely generate global consistency. )

I think you need to start from broken/breakable rather than pretend it's
just not broken.  Adding exception handling after the project is
complete doesn't make any sense to me.

Simon St.Laurent
Ring around the content, a pocket full of brackets
Errors, errors, all fall down! --

Received on Thursday, 17 July 2003 08:12:01 UTC