RE: Proposed restatement of syntax-based interoperability princip le ( was RE: Action item on syntax-based interoperability)

Let's just be careful with the 'why' of truth in practice. 

1. HTML came with implicit and easy to recognize semantics. 
   The existence case isn't based on the syntax but on the 
   ubiquity of the application.  It is cheaper than sending 
   objects across the wire.  On the other hand, in the beginning, 
   it was a presentation app and that isn't very complex.  The 
   response (eg, ASP) page is where the action is and it is full 
   of syntaxes.  Ugly but so.

2.  This I can agree with the following wholeheartedly

  "..acting as though syntax is an axiomatic prerequisite 
   gets me to a place where interoperation happens between systems at 
   reasonable cost."

Yet I don't consider that 'architectural' in the sense of 'necessity'. 
I consider it a smart effective means to lower the costs of systems 
interoperation by getting an agreement at a low level such that 
implementing that agreement is cheap, and once implemented uniformly, 
many conveniences fall out, such as the infoset, namespaces, XSLT 
and so on.  Now the footprint shrinks and flattens (complexity 
goes down and costs follow).

The problem of declaring it fundamental or that one syntax is 
basic is that it flies in the face of legacies or the problems 
of impedance matching.  We will and do have cases of systems 
that support multiple syntaxes and we shouldn't be so sure 
of ourselves that we think XML or any other syntax is always 
the right answer.  I do know that where it can be used, things 
are better overall, but in the case of VRML, both the syntax 
and the structure of XML were barriers to overcome, and in the 
case of RNG, the XML syntax obscures the meaning, so the alternative 
was provided.  I think that practice is outing the monoculture of 
the XML syntax.

len


From: Bill de hOra [mailto:dehora@eircom.net]
Sent: Monday, October 27, 2003 3:08 PM

Bullard, Claude L (Len) wrote:

> Start with VRML97 and X3D.  As long as the abstract model is 
> solid and the encodings support that, you can have multiple 
> syntaxes and interoperate successfully.  

I'm well aware of this argument. I've seen it espoused by bodies 
such as FIPA in their abstract architecture or by the OMG in UML or 
the W3C in the DOM IDL. I've seen analogous arguments about 
abstracting applications away from transports.

> so how fundamental is the syntax compared to the functional 
> model?  Not much.

Yet, as fond as I am of stuff like the FIPA AA, and other models 
such as RDF, what I haven't seen the widepsread adoption and 
proliferation of such things. Even SOAP was syntactic before it 
jumped to the Infoset.


> Because I never have been able to get this coop to do more than cluck 
> when asked for a definition of 'interoperation', I have to agree with 
> the last statement.  It isn't sufficient.  But I have to add, nor 
> is it fundamental.  Very useful, cheap, convenient, yes, but nothing 
> about the web except scaling costs make it necessary, and necessity is 
> my definition of fundamental.

There aren't many neccessary truths to be had. But I'd need a 
compelling explanation as to why something that is "true in 
practice" for the web cannot be captured as architectural truth of 
the web. Certainly I've no proof to put forward that a syntax driven 
approach is an axiomatic prerequisite for interoperation. All I can 
say is that acting as though syntax is an axiomatic prerequisite 
gets me to a place where interoperation happens between systems at 
reasonable cost. In other words, it's an empirical observation, not 
a theory.

If we said "syntax is primary" instead of "syntax is fundamental", 
would that work?

Bill de hOra

Received on Monday, 27 October 2003 16:36:42 UTC