RE: The self-describing web...

Sorry Norm, but that reads like skepticism about the streets 
of a given city at night being unsafe.  It makes for a polite 
and politically correct position in polite and politically 
correct company, but honest residents will tell you differently 
and then it comes down to risk management over some n of 
tests or 'what source do you trust for this iteration'.

So far Panglossian assumptions have not worked well for 
the web.  Witness click fraud, splogs, spam and the whole 
rot of malicious opportunism.   

The problem of authority and assertions about such 
on the web is that it does not reckon with two conjoined 
problems:

1.  Criminal and malicious behavior are permanent aspects 
of the web landscape.

2.  The web is an amplifier enabling such behaviors to become 
instantly and hugely profitable.

Therefore, a permanent business will be offsetting the effects 
of these.  Whatever the arch group does that helps with that 
will be more useful than assumptions that it won't occur or 
that the significance of it occurring is minimal.  It is in 
fact, a very significant problem.

Again, what is this 'self' in the subject?  If one has to 
manage a reputation to obtain and maintain trust, I suspect 
this will come back to location, time, identity and named 
verifiable types because the concept 'self' has to be 
combined with 'aware' and awareness has the distinctions 
'what is in my head? what is my head in?' so 'boundary'.

What does this statement mean:

"nesting things inside each other in arbitrary ways is core to the power of
XML."
http://www.w3.org/DesignIssues/XML

and pay attention to 'arbitrary' because I dispute that this is true even 
if polite and politically correct. Some arbitrary combinations will be
nonsense.  
For that reason, SGML provides notations and DTDs.  Now a savvy street user 
will look for ways to communicate 'correct' interpretations to 'committed' 
partners.  That is all a MIMETYPE does.

"...to use with any sort of document, without it having to be foreseen in
the schema for the original document"

Sounds good.  Doesn't work unless the combinatorics are specified apriori.
In 
other words, what TimBL says is not of necessity so, but can be made so by 
agreement.  What one says about a city's safety at night cannot be made so 
without agreement at scales that cannot be obtained or maintained.  So we 
come to:

"This way of specifying n independent schemas, or rather schemas which have
back-references to earlier schemas in some cases, allows a product to simply
quote the set of XML technologies which it supports. This has to be
negotiated between the sender and receiver of XML. It is not the same in the
general case to the set of namespaces used in the document, because function
elaboration may change that. All the same, the namespaces may be a useful
way of indirectly referring to the features."

Again:  a system is defined in terms of itself.  Apparently, DTDs and
Notations 
did work and self-description is a term for saying 'has a schema of some
form 
with everything you need to know to start interpreting this in the opinion 
of the author'.

Does it have to be harder than that?  Why?

len


From: www-tag-request@w3.org [mailto:www-tag-request@w3.org]On Behalf Of
Norman Walsh

Indeed. But if we assume that most authors are not malicious and most
readers don't carry around pernicious assumptions, I see some appeal
in the notion of being able to follow your nose and usually find what
the author intended.

Received on Wednesday, 4 January 2006 15:47:40 UTC