Re: "On the Web" vs "On the Semantic Web" (was Re: resources and URIs)

On Thu, 2003-07-17 at 14:08, Sandro Hawke wrote:
> Tim Bray wrote:
> > Michael Mealling wrote:
> > 
> > > In my 'layered' view, the SW is a layer above the web, and as such a SW
> > > 'resource' contains at its heart, a Web resource. You _could_ think of
> > > it this way: it's the same object with multiple interfaces, the
> > > Uri-Resource view found in 2396 being the equivalent of an IUnknown
> > > interface (just without the ability to query for  the other
> > > interfaces)'. As you go up the layers you end up with more available
> > > interfaces to pick from....
> > 
> > I have a lot of sympathy with this world-view.  Is there anyone who 
> > really doesn't like it?
> 
> Alas, I don't find it compelling at all.  (And yes, I've been thinking
> in terms of network layering, where appropriate, for ~20 years now.)

I know you have, I'm just not certain others have been...

> As I imagine the Semantic Web, we're likely to have two kinds of URIs:
> those which identify servers (via ResponsePoints, mostly something
> like an information-providing server or query-answering service, of
> which a normal web server offering RDF/XML files is a simple instance)
> and those which identify other things.  The "other things" URIs will
> often be strongly associated with the URI of an information-providing
> server, so that you can easily find out more information about those
> things.  

Sounds a lot like the concepts of resolution found in the DDDS stuff,
I'm with you.....

> The forms of "strong association" I'm comfortable with
> involve choping off the part of the URI after the hash mark, or
> following an HTTP redirect (ie [1] [2]).  I just don't see any kind of
> interfaces or even serious protocols around that second class of URI,
> and the first class is a refinement/extension of web protocols more
> than a layering.

In that light I could see how you might not find it compelling. My point
of view is based on cleanly delineating where the SW starts and stops so
that things that don't understand it don't have to deal with it. My
concern is that if we determine that the Web architecture includes all
URIs and not just 'http' protocol and scheme that we will start
attempting to normalize non-SW applications with that point of view. 
I know I'm bringing some of my concerns from the uri@w3.org list into
this but its important to be clear about what the web (and the TAG's)
scope is here. And the above layered approach does that.

For me I see both existing just fine. CORBA has an entirely consistent
object model that is very much not layer based, but IOP works over TCP
which works over IP and CORBA just doesn't need to care. I would hope
that the SW could be just 'clean'....

-MM

Received on Thursday, 17 July 2003 14:49:00 UTC