RE: language modularity, with semantics

From: Sandro Hawke [mailto:sandro@w3.org]

>IMHO the W3C should provide a system by which application languages
>can be usefully combined in ways which were not specifically planned
>for by the designers of the languages.  

Desirable, perhaps, but also perhaps a bridge too far.  The syntactic 
combination of XML languages using namespaces (sans some issues 
of ownership) is straightforward.  Actually processing these to 
do useful work is difficult without settling the problems of 
assigning handlers to tags.  Microsoft has a way to do that 
with stylesheets that works pretty well.  I don't know how 
Moz does that and that is an interesting topic but not for 
this list.

>I think this is a feature many
>XML enthusiasts erroneously think it has.

I agree.  Namespaces in and of themselves don't do this.

>This is of course different from providing application languages.  I
>think the W3C has this problem already just within the infrastructure
>with HTML+(stuff).  of course, that user community isn't about to
>switch to RDF/XML, but I think if we get the semantics figured out
>well enough, some of the techniques may work for them too.   (I have
>some experimental code to parse HTML as into RDF triples with a
>hypertext ontology, but it's rough/early work.)

Fine, but it comes down to yet another way to tell a framework 
which DLL to load.  It comes down to runtime libraries eventually. 
So RDF is yet another way to define an object model.

> The application semantics modularity model is a 
> problem of the object models, not XML or namespaces 
> any more than they are a problem of URIs.  We 
> can't treat interoperation by the same means as 
> we join relational tables.  It's nonsense.  Try 
> putting an SVG handler inside an X3D texture. 
> Doable?  Yes.  Meaningful? Depends.  Good engineering? 
> Undetermined: the performance of the objects is 
> doubtful but that is something for experimentation.

>Whether such a combination is meaningful should be a formal question,
>and it should not depend on either language committee knowing about
>the other, although in practice there will need to be some kind of
>sufficient ontological path between them.   

Verily, but in the last part of that sentence, I sense "a miracle". 
I'm not saying it isn't doable, just that it will come down the 
same old rats nest of incompatible implementations of handlers, 
aka, DLL Hell. 

>Whether it will turn out
>to have adequite performance, I agree, is a subject for research (and
>will of course depend on many non-architectural features).

Yes, and practically, IME, something that one ends up being 
forced to evaluate logistically and operationally, not formally.

len

Received on Monday, 28 July 2003 10:58:00 UTC