Re: Terminology Question concerning Web Architecture and Linked Data

>Pat Hayes <phayes@ihmc.us> writes:

...

>  > Try this for size.
>>

...

>That's not bad.  I don't think it gets the "old web" right, though.

Possibly not. I realize that the old Web is pretty darn complicated, 
and admit that others (such as  Tim and Roy) have a much better grasp 
of its intricacies than I ever want or will have. So maybe my 'http 
endpoint' criterion is too architecturally simplistic. Nevertheless, 
I think the spirit is clear: an information resource is some 
computational network entity that can deliver responses to transfer 
protocols, even if this 'entity' is distributed, virtual (like a 
hypothetical web server that knows the abstract text of some 
international agreement and can deliver it in any European language, 
which is in fact a bunch of servers with a content negotiator 
standing in front of them) and maybe other things I have never heard 
of: still, it has to be able to somehow be suitably active in the 
matter of moving information around the internet. As I say, for full 
details ask someone who knows the details, probably Tim or Roy. But 
in any case, things that aren't active in this way, aren't 
information resources.

>  It
>doesn't really explain the many web pages which look completely
>different depending on your cookies or IP address.

What have web pages got to do with it, still less what they look 
like? The representation that REST talks about in cases like this is 
still the (single) representation of the (single) resource. If it 
gets hacked around by your cookies on your machine, that's not the 
Web's fault.

>  And it suggests that
>the end-point of an HTTP request corresponds one-to-one to the URI, but
>in fact the mapping between URIs and web server processes is
>many-to-many.

OK, I admit I get lost in the weeds at this point. But see above.

>
>And it's still pretty darn complicated.    :-)
>
>I'd love to see a New Architecture Of the WWW, that covers old and
>Semantic web in a few simple pages, but the old one isn't quite broken
>enough yet to motivate its acceptance, even if we could figure it out.

Im not saying its broken at all. What is broken is getting 
architecture muddled up with semiotics. I don't think the 
*architecture* of the SWeb is any different from that of the Web. 
Just admit that the description of the architecture is exactly what 
it says it is, and leave the semantics to a different document. Then 
we wouldn't have honest folk trying to understand the architecture 
document using intuitions from semantics, and getting utterly 
confused. And, we can write a Semantics of the Semantic Web which 
will refer to the Architecture of the Web but won't get confused with 
it.

Here's how to do it:

(1) distinguish at the get-go between reference and access.
(2) have the architecture document talk about access and not mention 
reference at all. Admit that its all about 'information resources' 
and give up on this crap about resources being anything in the 
universe.

(3) In the semantics document, point out how amazingly convenient and 
natural it is to have a URI refer to whatever it accesses, (so we can 
just talk about 'identifying'), but ...
(4) ... since we are now talking about reference, and we might want 
to use URIs to refer to inaccessible things, we need to handle this 
somehow, which leads to the 303-redirect recommendation.

Pat


-- 
---------------------------------------------------------------------
IHMC		(850)434 8903 or (650)494 3973   home
40 South Alcaniz St.	(850)202 4416   office
Pensacola			(850)202 4440   fax
FL 32502			(850)291 0667    cell
phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes

Received on Friday, 27 July 2007 20:25:28 UTC