[
Lists Home |
Date Index |
Thread Index
]
"Bullard, Claude L (Len)" <clbullar@ingr.com> writes:
> If I understand you correctly, Steve, you are
> turning the markup world on it's head a bit.
> Instead of the element type declaration dominating
> the instance (here is the contract, it is enforced
> on any one of these you encounter), you are saying
> the instance declares the classes to which it belongs
> (here is my contract, pick the right declaration)
> and the instance dominates the contract?
That's right, Len. This is about who controls what's
allowed to be said in an XML document. It's about
allowing the document to say, "I contain data that can
be understood within the X, Y, and Z processing
contexts. (Substitute for X, Y, and Z the favorite
software vendors of the likely customers for the
information in the document. The information's vendor
doesn't force all of its customers to buy the same
software, just so they can use the information.)
> Is that it? I understand that you advocate arch
> forms, and don't dispute they are a solution to
> that. I want to first discuss the requirement
> itself and ensure I understand that.
> Then the archform solution such as
> <?xml version="1.0"?>
> <?IS10744 arch
> name="somearch"
> public-id="+//IDN me.com//NOTATION Some Architecture//EN"
> dtd-public-id="+//IDN me.com//DTD Some Architecture//EN"
> doc-elem-form="somedoc"
> renamer-att="somedocnames"
> options="option1 option2"
> ?>
> <mydoc somearch="somedoc"/>
> which will immediately invoke the wrath of the PIhaters,
> but regardless of that, how does that example illustrate
> a solution to the requirement (multiple classes for
> one instance)?
Len, please forgive me for minimizing your example of a
"base architecture declaration" as follows:
> <?IS10744 arch
> name="somearch"
> dtd-system-id="http://www.somewhere.com/somedtd.dtd";
> ?>
The above Processing Instruction declares a "base
architecture" -- that is, it establishes a connection
between the document that contains the PI and the DTD
that it refers to. (It's not so very
different from an XML Namespace declaration.) In order
to support *multiple* inheritance by a single element,
you'd first have to have two or more PIs, referring to two or
more DTDs. For example, let's say that we have two
PIs, one with name="somearch" and the other with
name="somearch2". If an element in the same document
says:
<myelement somearch="foo" somearch2="bar"...
then that particular element will be subject to
(1) all of the syntactic and semantic constraints
imposed on <foo> elements by the information
architecture whose DTD is referenced by the
name="somearch" DTD, and
(2) all of the syntactic and semantic constraints
imposed on <bar> elements by the information
architecture whose DTD is referenced by the
name="somearch2" DTD, .... *and*, of course,
(3) all of the *additional* syntactic and semantic
constraints imposed by the primary architecture --
the actual DTD in effect for this document -- on
<myelement> elements.
In other words:
* one element, with one explicit syntax,
* plus two additional implicit syntaxes,
* so, in effect, three syntactically- and
semantically-interrelated interpretations. The
choice of interpretation is made when the user of
the document chooses a processing (application)
context within which the information will be used.
Why is this capability important? For all the reasons
that XML Namespaces are said to be important. But
there are telling differences between the
"architectural forms" paradigm and the "XML Namespaces"
paradigm. Here are two of them:
(1) The "architectural forms" paradigm offers a single
standard way, implementable in an
XML-1.0-conforming XML parser, to validate the
syntactic conformance of an element instance with
all of the constraints involved in all of its
various interpretations (one interpretation per
element type (in the DTD of each base architecture)
of which the element claims to be an instance). As
with any other DTD, the syntactic constraints that
a base architecture imposes on an element type
include constraints on its context(s), as well as
its content and attributes.
Why is this important? Remember, a DTD (or XML
Schema or whatever) represents a kind of contract
between information providers and software
providers. For the sake of all information
providers and consumers, it's good and necessary to
have a way to tell whether the element instance --
as it exists in the document instance -- will be
understandable by software that is designed to
process data that conforms to the base
architecture. The "architectural forms" paradigm
requires the XML parser to be able to report not
only the parsed document as it was written, but
also the parsed document as it would be reported if
it had been explicitly marked up according to each
of its "base architectures". If it's a validating
parser, then it can also report whether the
reported "architectural instance" conforms to its
DTD.
When information interchange fails, the public
interest demands that there be a way to point the
finger of blame at either the information provider
*or* the software provider, or both. Without such
a publicly-available functionality, the information
industry will never be independent of the
information processing industry, the nature of what
can be said will be limited to what the major
software houses deign to allow us to say, and the
evolution of civilization's nervous system will be
distorted. (I'm reminded of the fact that there is
legislation now before Congress that would allow
the owners of the communications networks to give
precedence and speedier delivery to packets in
which they have a special business interest.
Should the network owners have the power to censor
the public bandwidth? I hope not. And neither
should the major software vendors have the power to
censor the interchange of information based on
their de facto hegemony over the world's desktops.)
There is no room here for doubters who say, "Show
me the code." Standard, super-quality, open-source
software for this purpose already exists, and has
existed for years. James Clark's "SP" parser
already has everything necessary to do the job, and
it has been upgraded by Luis Martinez and Peter
Newcomb to support inheritable XML architectures,
using the ISO standard Processing Instruction-based
syntax for declaring "base architectures". There's
a link to this software, compiled for Linux,
Windows, and Solaris, at
http://www.hytime.org/htnews.html (first paragraph)
(2) Similarly, it means that software that understands
a given information architecture can be re-used, as
an engine, in the context of more specialized
applications. The same idea appears in XLink.
People who wonder why XLink isn't more popular
should consider how unfriendly the current XML
syntactic environment is to the use of such
"engine" software. There's no way for developers
to prove that their engines function correctly, and
there's no way for information providers to have
assurance that they are using a particular "element
template" (as the XLink Recommendation calls
architectural forms) correctly. Indeed, it's hard
to see how XML Namespaces could be more perfectly
contrived to preclude such general accountability.
Why am I still beating this dead horse?
Just because XML people should pay attention to this.
Either:
(1) XML will provide a real platform for collaborative
work, in which real control over syntax can be
really distributed, without losing the syntactic
rigor necessary to guarantee effective information
interchange, or
(2) significant portions of the potential for XML
syntax to serve as a basis for collaborative work
in the sunshine will be lost.
Maybe #2 would be the best outcome, all things
considered. Personally, however, I think that with
"XML Namespaces" the XML community has been shooting
itself in the foot, from the beginning to the present.
Regardless whether I'm right or wrong, the XML
community would be well advised to consider: Whose
interests are served by the technology of XML
Namespaces, and how? What vision of the future is
guiding the technical development, here? And is that
the future we really want?
-- 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
|
