Re: namespace usage as assertions

At 03:56 PM 6/21/00 -0400, John Cowan wrote:
>*sigh*
>
>A URI *identifies* a resource.  A resource may be instantiated by zero, 
>one, or
>many entity bodies.  Thus the URI http://www.ccil.org/~cowan/scritchifschisted
>identifies a resource with zero entity bodies: attempts to fetch the entity
>body will fail.  (You cannot fetch a resource, because resources are 
>abstractions;
>you can only fetch the corresponding entity bodies).

I agree that you fetch an entity body, not a resource; that's why I said 
you "pour out the *contents* of the bucket" when you fetch.

What I was pouncing on (sorry, Dan...) was Dan's use of "points to a 
resource".  I'm very familiar with the impulse to put it in these terms, 
but we have got to starting admitting that pointing at something implies 
the ability to get it.  URIs were *built* for fetching, despite the 
philosophical stuff in RFC 2396 about resources not always being network 
retrievable.  In fact, they're defined as being "conceptual mappings to one 
or more entities" (note, not none), with the obvious implication that they 
*can* resolve to something.  It's widely considered an error when they don't.

So I contend, after years of holding the other position, that our choice of 
URIs for namespace names was a mistake because it relied on a hopelessly 
theoretical view of the Web universe, and that some other kind of naming 
scheme would have served us all much better because it wouldn't have come 
with the fetching/accessing/dereferencing baggage.  A separate mapping 
mechanism could then have been defined at any time, once we decided whether 
we wanted this.

I'm sorry if this rehashes other discussions held here; a person with a day 
job can't possibly keep up with this list. :-)

         Eve
--
Eve Maler                                    +1 781 442 3190
Sun Microsystems XML Technology Center    elm @ east.sun.com

Received on Wednesday, 21 June 2000 16:36:19 UTC