RE: [xml-dev] Defining an XML vocabulary: specify syntax, semanti cs, an

[Len Bullard <len.bullard@uai.com>]

>From: Fraser Goffin [mailto:goffinf@googlemail.com] 
 
>Perhaps other complementary vocabularies that are closer to process
>modelling are required, including maybe BPEL and other workflow
>languages that provide a target for the data instance ?

Pick one or several.  My description was networks of processors, but one or
several, my point is simply that it is disingenuous to consider XML
application languages or even data serializations without understanding the
need for the processor.   It is a convenience of specification, not
practical systems design where the goal is to actually use the data.

If the question is 'should we specify the process in XML', my answer is,
fine.  XML doesn't care.  IMO that is a clunky way to do it but there are
examples of this.   

If the question is, should the behavior be specified in the same XML as the
data, I say, XML doesn't care but common practice as Noah points out
indicates that is not a good design if the intent is to reuse the same data
for multiple applications.  

If the intent is to provide the name of the processor type and even to
indicate where an instance of that type can be found, then NOTATION
declarations can be applied, processing instructions, out-of-band comments
and so forth.

There are plenty of XML applications who's specifications and standards
describe the expected behaviors in detail.  Prominent among these are the
graphics languages that use XML as an encoding.   They typically do that in
the specification/standard itself.  In the case of X3D, an object model is
specified.  When we discuss extensions to it such as the new NetworkSensor,
we don't define it in say the XSD abstract type initially.  An object
interface spec is defined first.  Once there is consensus on that, the
editor or their designate creates the XSD and provides it for the XML
encoding.  For the Classic VRML coding, the interface spec is sufficient.

So the question might be, is a behavior specification necessary?  The answer
is, your mileage may vary.  For rendering applications, it typically is
because fidelity of rendering and fidelity of behavior are both desirable.
Otherwise we get cul de sacs such as Second Life or Forterra Olive:
proprietary non-interoperable or we get open source initiatives which wind
up being the same thing: a cul de sac that traps the content.   To avoid
that, the Web3DC uses the object model approach and it works about as well
as any other approach.  X3D files for the most part do interoperate and
efforts are underway every year to improve that such as the open conformance
test evaluation going on at the upcoming Siggraph conference.

It would be useful to hear from other application language design
communities about how they do this and what requirements create the need for
this?  I agree that if one is sharing names, phone numbers etc, it isn't all
that useful.  But then that is information best stored in a relational
database where there are mountains of standards already describing it.

Cheers,

len




This email and any files transmitted with it are confidential and intended solely for the use of the individual or entity to whom they are addressed. If you have received this email in error please notify the sender. This message contains confidential information and is intended only for the individual named. If you are not the named addressee you should not disseminate, distribute or copy this e-mail.