xml-dev - It's the syntax. . .

["W. E. Perry" <wperry@fiduciary.com>]

Jon Bosak's standing in the XML world is (rightly) such that
proponents of every markup ideology or mechanism rush to graft
their arguments onto his aphorisms. Most recently, a spate of
references to his keynote address at last week's XTech
demonstrates again the hope that a Bosak epigram may offer the XML
imprimatur. I shall act no differently in asserting that I heard
him say in that keynote that XML is syntax, not semantics, and I
contend that the most important work which we should be doing is
grounded in that distinction.

The ancillary standards promulgated or considered since XML 1.0
are all (except SAX) attempts to canonize various
semantically-intentioned structures. This is not a bad thing: well
designed semantic shorthand can usefully simplify communication
and drastically reduce the preliminaries required between
well-acquainted correspondents. The original intent of XML itself,
though, is different. SGML-on-the-Web was specifically intended
for the autonomous nodes of an Internet topology who did not share
the semantic infrastructure of a tight-knit markup community.
Well-formedness was an explicit extension of the benefits of
markup to those without the shared semantics of the DTD. Since
those earliest days of XML, however, each new proposal has
increased the requirement for semantics shared a priori, and
thereby effectively subsetted the range of application for
'conformant' XML. The monstrous complexity of XML Schemas and, at
the apparent other end of the spectrum, the putative simplicity of
SML are both pre-agreed semantic structures which brutally curtail
what might otherwise be expressed, albeit in novel or unexpected
ways, in the simple syntax of XML 1.0.

The immediate motivation for these observations is the 'XML
Messaging and Protocols' BOF held at XTech. Whether or not that
meeting was a first step toward canonizing SOAP as a W3C
recommendation, it unquestionably revealed a determination to
constrain the syntactical possibilities of 'messaging' to the
narrowly semantic ones of an RPC vocabulary. In other words, in
the name of interoperability and with genuflection to the vast
potential of ecommerce, we are now subsetting both the
'interchange' form of XML (as RPC) and the 'compositional' form
(with XML Schemas, etc.). At this moment we should take a look at
our roots, from which I (alone?) conclude that:
    --XML was originally motivated by an Internet topology of
autonomous, largely anonymous nodes. Given that in many cases the
humans behind many of those nodes do not share the history nor the
assumptions of the established players (whether 'established'
means as standards-setters or, from the ecommerce perspective, as
dominant forces in their vertical markets, or both) there is no
easy--maybe no workable--way to cast the needs and assumptions of
these newly-arrived participants into semantic structures which
they had no hand in defining. One effect will surely be to
exclude, at least from the more standard fora and markets, those
for whom, presumably, XML went to the Internet in the first place.
The irony is that XML 1.0, as syntax, continues to offer them the
tools with which they might create a truly worldwide forum from
which established players would be (self-)excluded by their own a
priori semantic expectations.

    --The fundamental functions of an XML processor are syntactic.
I am talking here about processors more ambitious than parsers,
which are simply plumbing, The processor which I believe is
described by that term in XML 1.0 takes XML syntax as its input,
and only expectation, and performs some fundamentally local work
upon it. If that were not so, why would the document/data/message
need to be handled by that processor, or that node, at all? If the
source of the XML were obliged to share, and therefore to know,
sufficient semantics to be able perfectly to describe its intent
to a separate processing node, why not simply do the processing
itself and avoid the specifically-remote possibilities for failure
which the RPC people have been drawing attention to these past few
days? In other words, why shouldn't the
assumption/expectation/prerequisite of XML processing be simply
XML syntax, precisely as described in XML 1.0?

I believe that this is the processor which we should be building
as the general case, and for which standards grounded in semantic
infrastructure are merely a sideshow. The past two years have
allowed the XML community to plumb fairly thoroughly the
possibilities of pre-agreed semantics and perhaps usefully defer
the really hard work of building on the broader syntactic base.
That time of doodling has now passed. We need to build an XML
processor that works like yacc, or, to take a more nearby example,
which allows, as RDF intended, saying anything about anything in
an XML syntax. Notice that in this case the XML is a great deal
more than mere serialization format between established semantic
structures:  it is effectively the source code for semantic
structures purpose-built to a particular processing instance.
Specifically instance semantics can be assembled from simple
predications and then communicated to an autonomous, anonymous
node. Careful construction can provide reasonable communication of
the intended semantics, but cannot dictate their specific
interpretation in the unique, and quite possibly opaque, context
of the receiving node. Nor can the transmitter dictate--perhaps
not even know--the specific processing done at a receiving node,
nor should it be entitled to. If the two nodes are engaged in a
useful transactional dialogue, there will be a return of data
which the original node is then free to interpret and process in
its own way.

Finally a plug (eleemosynary, and therefore I hope not offensive
to the list):  this question of what standards should be built
upon seems so pressing that I will be devoting significant
attention to it in the New York XML SIG over the next several
months (subject to the continuing acquiescence and, I hope,
enthusiasm, of the members of the SIG). Those with an interest in
this family of topics (and a non-virtual presence in New York) are
welcome to join us as speakers, audience or both. The first such
session will be Monday 20 March 7-9 p.m. in downtown Manhattan.
Please email me directly [DO NOT RESPOND TO THE XML-DEV LIST] to
reserve a place, propose a topic, offer to speak or to debate,
etc.

Respectfully,

Walter Perry

XML SIG Leader
NY Object Developers Group


***************************************************************************
This is xml-dev, the mailing list for XML developers.
To unsubscribe, mailto:majordomo@xml.org&BODY=unsubscribe%20xml-dev
List archives are available at http://xml.org/archives/xml-dev/
***************************************************************************