Re: Layering XPath/XSLT namespaces is unacceptable

At 10:07 AM 2000-06-15 -0400, keshlam@us.ibm.com wrote:
>>Consider the XML 1.0 Recommendation.  It defined the xml:lang attribute.
>>There is syntax here, but there is also semantics that goes beyond the
>>syntax to constrain what sort of content is OK to be within the tags.
>
>This doesn't constrain; per the XML REC, it _specifies_. In other words, it
>gives guidance about the expected content of this element at this time; and
>individual tools can attempt to do something useful with that "language
>tagging". But the definition of "something useful" is not stated.
>

I would be very interested in your distinction between constrain and
specify.  Since what I need is your word for the union of the two.

>I agree that this is a semantic hint. But it's a pretty darned weak
>semantic. At best it's equivalent to using XML Schema to specify that the
>field is a number -- OK, that's fine, but what does the number _mean_? (And
>how does one type-check langauges -- put them through an appropriate
>spellchecker?)
>

Manually.  Compare <http://www.w3.org/TR/AERT>.  Markup language is in the
main semi-formal language, not formal language.  Processing the full
language is an automated (i.e. man/machine) not automatic (machine alone)
process.  The automatic processing of syntax in the parser is a [severable,
encapsulated] supporting subprocess in the overarching automated process.

>I  consider number, and xml:lang, to be extended syntax rather than
>semantics. But I grant that the line is fuzzy. De gustibus non disputandum
>est. (OK, now what language is this paragraph in?)

The paragraph is macaronic.  The _paragraph_ per se is not in one language
identifiable within the type definition of xml:lang.  Fortunately in HTML
one can put a lang attribute on a SPAN and you can say what language you
are in where you are in each language.  This is not idle chatter.  Check
the W3C Rec. on web content accessibility where it reminds you that you not
only can, you should.

But the bottom line that you got to, that the dividing line between syntax
and semantics is placed differently by different people, is the point I
would like folks to take away from this exchange.  We need more objective
ways of bundling computing and communication concerns into articulable
domains, and work within a web of these ontologies, not a binary layer
stack of semantics over syntax, nor even in the end a uniform layer stack.
The layer stacks in different processes need to respect the precedence
[dependency] relationships among domains, but not linearize this graph
always in the same linear flow nor segment it always in the same chunks.

Al

>
>______________________________________
>Joe Kesselman  / IBM Research
>

Received on Thursday, 15 June 2000 10:43:28 UTC