RE: The self-describing web...

Big can of worms this, but the answer I like best is 
the reader is obligated to understand that which they obligate  
themselves to understand:  aka, ontological commitment.  (See Gruber). 

My short response is as it was for http range: a system is 
defined in terms of itself (see last question) so any 
'understanding' per se it in terms of system symbols and 
functions (returns null, returns 404, returns.... given 
named function).

In effect, given a set of specifications or standards that have 
a provable means to check assertions of compliance or conformance, 
a party can determine if understanding exists for 
some time interval (might be a third party or might be a 
communication party).   So ultimately, you need a test, a 
time and an occurrence and you may repeat this as needed. 

The tough problem is to know when it is needed.  For that 
reason, DTDs were mandatory in SGML (defaulted to sender 
contract always validated by receiver) and the standard 
limits what can be formally known.  Then the 'text' is interpreted 
exactly as Elliotte says.   XML reversed that by making 
syntax obligatory through Draconian mandate, then left everything 
else to the negotiating/communicating parties.  Namespaces throw 
a wrench into that understanding by introducing a third symbol/sign 
the meaning of which is in some minds, not explicit beyond 
being syntax sugar for preventing name collisions.  The Arch 
then says, 'and you might want to facilitate negotiation/understanding' 
by putting a representation of a resource at the identified 
location should you so choose to use this as a locator'. 

It's harder to say than to understand.  In short, "when in 
doubt peek and poke or layout".  The first problem is the 
self-realization of 'doubt'.  One might say that the web 
provides the means of initiating a negotiation/communication 
to resolve doubt if not the means of realizing it.

I don't think that works very well here, but then like Elliotte, 
I'm not quite sure what is being asked.  I don't think you want 
to go down the path of attempting to formalize 'self-aware systems' 
or 'sentient computing' although I have a CTO that would enjoy 
that discussion.  See Peter Batty.   A web that is 'self-describing' 
would be 'sentient' in the sense it has to have a 'self'.  Does it?

Started the New Year with a question that this list will still be 
discussing next New Year, eh?

len

-----Original Message-----
From: www-tag-request@w3.org [mailto:www-tag-request@w3.org]On Behalf Of
Dan Connolly
Sent: Tuesday, January 03, 2006 3:28 PM
To: Elliotte Harold
Cc: Norman Walsh; www-tag@w3.org
Subject: Re: The self-describing web...



On Tue, 2006-01-03 at 15:52 -0500, Elliotte Harold wrote:
[...]
> Bottom line: the reader of a document is ultimately responsible for 
> understanding the document. Different readers will understand different 
> things.

In extreme cases, yes; but mostly, they'll understand the same
thing; that's where the web gets its value. It facilitates
shared understanding by providing mechanisms to bind (relatively) small
symbols like http://en.wikipedia.org/wiki/BAR_Camp to interesting and
useful meanings.

I can't be certain that the bytes I see when I visit wikipedia's
BAR_Camp page will be exactly the same as the bytes you get; anybody
could edit them in the the mean time. I can't be 100% sure your
browser will render them the same way. I can't be sure your
understanding of English is just like mine. But it's a good bet
that you will understand my meaning if I use that symbol as
a reference, because, by and large, we do share quite a bit
of context: URI syntax, HTTP, TCP, DNS, IP, HTML, and English
(and I think ... yes... in other cases, such as
http://en.wikipedia.org/wiki/United_States , I can use a URI symbol
to cross natural languages too).

If it were really the case that given two readers of a document,
there was no correlation in the information they'd get from it,
the web would be of little value.


>  The document author cannot force the reader to understand any 
> particular thing.

Indeed, but there are some understandings that readers can
hold the author accountable for, and some that they cannot;
those understandings are the ones that the author invokes
by implicit reference to ubiquitous standards or explicit
references to linguistic constructs described elsewhere in the Web.


>  Author's intent does not outweigh the reader's 
> presumption.

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

Received on Tuesday, 3 January 2006 22:25:25 UTC