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

No force.  The data is in a syntactical form to make it easy 
to push or pull across a wire and unpack it an use it on 
the other end.  But yes, when it comes down to standardization 
for a particular and shared purpose, it is about rendering 
and behavioral fidelity.  Otherwise, there is no incentive 
to invest effort in the format.  There are degrees of 
this.  For HTML 1.x, the rendering behavior could be initially 
very loose.  The behavior was crude, so it was of not too 
much concern.  Since then, the requirements have gotten 
more complex and the specification has also accordingly.  For 
VRML97, a real time rendering and simulation in 3D, the 
requirements started complex and have remained at about 
the same level since with the changes being made to enhance 
the behavioral fidelity, tighten up the rendering, and 
provide multiple syntaxes.  If you want to use the data 
to do things with it other than what the specification calls 
out, you can.  Just don't expect interoperation based on 
that reliably.  No problem.  Local rules still prevail 
and you can use it as you wish.

In other words, syntax is enabling not interoperability, 
but portability.  Data can be shipped to you or from 
you, packed and unpacked, and used at will to desire. 
What it is not doing is enabling interoperability because 
you are not interoperating; you are operating.  No problem.

len


From: Elliotte Rusty Harold [mailto:elharo@metalab.unc.edu]

At 1:17 PM -0600 10/27/03, Bullard, Claude L (Len) wrote:

>A standard that provides an abstract model for the
>interoperating **system** (eg, X3D) is sufficient to enable
>two different implementors to create two different
>implementations for which a third author can write
>a document in any of the authorized encodings and expect
>it to operate with either of those implementations to some
>degree of rendering or behavioral fidelity.

This perhaps is the crux of the matter. Efforts to standardize data 
models and semantics are effectively efforts to standardize behavior. 
And I don't want to behave. (Bad Rusty! No biscuit!)

I may have very different needs than you have. I may want to do 
something very different with the data you send me than you expect me 
to do. For instance, I may want to render in black and white instead 
of color. I may want to compute the minimal enclosing volume of all 
your VRML objects, and render that. I may want to spell check the 
text data and throw the rest away. I may want to search the data for 
structures that meet only some criteria, and render only those. I may 
want to spider the links in the data. Or maybe I do want to do 
something you'd recognize as acceptable given a fixed set of 
semantics. But it's my choice. There's no reason I have to accept 
your meaning for the document. I'll have my own.

You're seeking interoperation by making everyone do the same thing. 
I'm an anarchist. Let everyone do whatever they want with the data. 
It's not my business to tell someone what they can or cannot do with 
the data I send them.

Received on Tuesday, 28 October 2003 09:10:00 UTC