W3C home > Mailing lists > Public > xml-uri@w3.org > May 2000

XML semantics was: Web Architecture, 'XML Autonomy'

From: Tim Berners-Lee <timbl@w3.org>
Date: Wed, 24 May 2000 15:04:13 -0400
Message-ID: <001201bfc5c8$f72e21f0$07a45c8b@ridge.w3.org>
To: "Rick JELLIFFE" <ricko@geotempo.com>, <xml-uri@w3.org>

-----Original Message-----
From: Rick JELLIFFE <ricko@geotempo.com>
To: xml-uri@w3.org <xml-uri@w3.org>
Date: Wednesday, May 24, 2000 12:08 PM
Subject: Re: Web Architecture, 'XML Autonomy'

>> Based on this discussion and outliers to it, I'd strongly suggest that
>> W3C step back for a while to clarify both their basic architecture and
>> approach to achieving it.  Alternatively, I'd suggest that clarification
>> the relatively simple layering issue - what is the role of XML in the
>> semantic Web? - might be enough to get this wagon out of the mud.
>Tim BL makes interesting statements in
>When I first read them, I thought he was just being loose (since the
>documents are
>just sketched out) but in fact it seems he is being exact.
>"In XML, a language is a "namespace", and the document about the
>language is called a "schema".
>In XML, one document can contain a mixture of languages, and so the
>schema if written in XML may contain information about syntactic
>constraints (in XML-schema language)
>and/or RDF properties (in rdf-schema language), or any combination of
>the above. (note)
>"XML puts no contraints on a language apart from syntactic structure.
>There is not (without RDF
>and logic or some other higher level) any overall framework into which
>new languages can be introduced. So, the question of what an XML
>document means depends
>first upon the fully qualified name of the document element. No
>can be attached to any of its descendents in the document tree except in
>as much as is
>defined by the specification of that element type in that namespace. One
>cannot talk
>about the "meaning" of a subtree of a document without understanding the
>semantics of the language. In fact, because languages only necessarily
>define meaning for
>documents, the only way one can talk about the meaning of a subset of a
>document is
>to define a how those parts of the document can be reassembled into a
>second whole
>document. This is what must be done when a digital signature is applied
>to a document. "
>According to this vision:
> * a language = namespace  (therefore derefencing a namespace URI should
>yield a language definition, I suppose)

Yes, if it is one which can be dereferenced.

> * the document element ultimately determines semantics of subelements

Not ultimately, but primarily. Let me tell you where Iwas coming from here.
You have to be able to write languages which qualify somethig quoted.
 I must be able to write "<not><b>The sky is green</b></not>" and send it to
you without
you holding me to "<b>the sky is green</b>"

Also, it is no good for XML signature to cover just parts of the document.
(This was suggested at one point). What would it mean? You can't
give a meaning to disconnected bits of document in place - you can
give a transformation which forms a new valid document and sign that.

> * semantics are only determined by element name and namespace name, in

You said "only", I didn't.  The attributes and body and stuff aren't there
for no reason!

I said that if you understand foo: but not bar: then you don't understand


This was related to the discussions in digital signature.

>I cannot imagine that this vision is shared by many people: the more
>common position would perhaps be
> * a vocabulary = namespace

I take this as a definition of "vocaulary".

> * any information item can contribute to the semantics of any other
>information, whether in the same branch or not

I can't make this into something I can reason with. "Contribute to"?
If this means "add new information you didn't know before
but without contradicting anything you knew" then I agree.

If it allows contradiction, then I would propose that the rule in
which <not href="foo.xml#sentence2"/> allows you to
invert the sense of a phrase in someone else's document is
an unsound basis for semantics of the language. The semantics
of a document depend on all other documents.
Not a sound basis for ecommerce

The next step is to allow it within a document.  This means you have to
have the whole document before you can deal wiht any of it.
This might work for some applications, but it is also nice to deal
sometimes with forumale hich have properties which allow
conclusions to be drawn more generally. Wehen you raed a library
catalog card, you only have to read the title. You don't expect something
on the back to say "oh, the title wasn't really that, it's this...".
If one level of a logical document is a logical conjuction of statements
you can treat them individually without this possibility.

Most languages have some properties like these and they are very useful.
XML doesn't insist on any at the XML level:

> * semantics can be determined by many other factors including position,
>presence or absense of attributes, etc.


>XML Schemas seems to be neutral to both views, I think.

XML schemas don't talk about semantics, unless you hang some
semantics onto the subtyping of elements I suppose. I don't think that
is the intent.

>Rick Jelliffe

This is getting a bit off topic for the list.  I think so long as we keep
subject lines in synch with the subject of the message we should be able to

Tim BL
Received on Wednesday, 24 May 2000 17:40:47 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 23:13:58 UTC