Re: Infoset as subset of useful

["Steven R. Newcomb" <srn@coolheads.com>]

[Simon St. Laurent:]
> To some extent, I find the Infoset to be an odd
> imposition of expectations from SGML which were never
> fully expressed in the XML 1.0 specification, which
> itself conflates lexical expression with processing
> expectations.

> Given that the Infoset already allows applications to
> be compliant by expressing how they do or don't
> implement portions of the Infoset - a refrigerator
> can be compliant, with a Post-It note expressing that
> it needs none of the Infoset for proper
> implementation[1] - it seems like a wasted
> opportunity for the Infoset to be written only as a
> subset of XML's lexical potential.

> Starting with the complete set - and then allowing
> developers to specify which parts they do or don't
> care about, perhaps with guidance from the Infoset
> spec - seems like a much wiser idea.

> I can't say I've found any use for the Infoset as
> currently written, except as a contributor of such
> fine terms as "element information item" which litter
> the landscape of XML verbiage.  And yes, I'd strongly
> suggest that the W3C discard the Infoset as presently
> written and start again, perhaps from XSet[2], or
> drop the project entirely.  The XPath model seems
> quite popular if not without its own flaws.

I still believe in the value of the Infoset work.  I
would very much like to see a machine-processable
formal expression of the Infoset adopted as part of the
Infoset Recommendation.  I believe that it would be
very appropriate and very useful to use the ISO
"Property Set" formalism for this purpose.  All of
Simon's concerns and complaints can be successfully and
favorably resolved in a property set for XML, based on
the Infoset, and expressed (like all property sets for
all notations) in XML.

As I recall, the reason we needed the Infoset was to
provide a firm basis for XML addressing.  For example,
it is necessary to eliminate the ambiguities in the DOM
spec that have made it impossible to guarantee that,
given the same XML document, the same calls to
different DOM implementations will always return the
same information.  In other words, we need a firm
answer to the question, "What counts as a node?" in all
contexts and circumstances.  The need for reliable
addressing has not diminished; quite the contrary, in
fact.

The grove paradigm meets Simon's above-implied
requirements of:

* distinguishing lexical expression from processing
  expectation,

* modularity, with users able to choose to use and/or
  ignore modules,

* minimality, with foundational concepts and
  terminology clearly distinguished from concepts and
  terms that are coined for specific semantics (the
  grove paradigm is, among other things, a toolkit for
  defining such concepts and terms), and

* many other real-world requirements, too, including
  all the rigor, formality, and machine-readability
  that is needed to support direct translation of
  grove schemas (called "property sets") into the
  relevant aspects of real implementations.

We need information to be maintainably addressable.
The grove paradigm is the internationally standard
enabler of universal and maintainable addressability
for all kinds of information, including but not limited
to information that happens to be expressed in XML.
Only the grove paradigm covers the territory.  Someday,
the Web community will have to embrace the grove
paradigm, although perhaps under another name.  The
name is not important.  Reliable addressing, however,
is vital now, and it will always be vital.

For those who are interested in learning more, I'm
still recommending Paul Prescod's Web-oriented
introduction, which (I just checked) is still at
http://www.prescod.net/groves/shorttut/

-Steve

--
Steven R. Newcomb, Consultant
srn@coolheads.com

voice: +1 972 359 8160
fax:   +1 972 359 0270

1527 Northaven Drive
Allen, Texas 75002-1648 USA