RE: Why Model Concepts? (Was RE: Object Role Modelling (ORM) or U ML or ??for designing Schemas)

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

Something more generic than syntax?  Well, I guess we could 
offer up the SGML Declaration once again to the masses.  
Nah, then the XML inventors would have to be hanged too. 
RDF made the semantic web AI unless we want to use AI like 
Republicans in the States use the term "liberal", a demonization 
for conning the unwashed into electing them.  Let's not.

Tell me how to get rid of the communicative a priori, the bootstrap,   
then, I can agree with the rest of what you are saying.

We can route information without knowing it's meaning.  That 
is easy.  We just can't use it safely.  My guess is, practicality is 
in the direction of common means for choosing meanings.  Generic 
tools are common means.  Choose a notation and you  
chose a model too (IDEF(relational) vs UML(objects)).   Should 
we include groves here too? *Hiding* information under a 
stack of abstract names is still just a means.  We can agree 
on UML just like a good sized chunk of the CSHeads once 
agreed on IDEF, others on Express.  But until we agree, 
all bets are off above syntax parsing (and for that, we 
had to agree on an SGML Declaration).

So once again, why do we need conceptual models (we 
don't always need them but when we do....)?  To me, 
this is an abstract layer, so I agree with you, Rick, 
but I really don't think the semantic web does much 
until some model for this gets buy-in.  AI is irrelevant 
except insofar as we can identify and learn from the 
experience of the systems builders who used that name 
to label their products.

Len 
http://www.mp3.com/LenBullard

Ekam sat.h, Vipraah bahudhaa vadanti.
Daamyata. Datta. Dayadhvam.h


-----Original Message-----
From: Rick Jelliffe [mailto:ricko@allette.com.au]

From: Bullard, Claude L (Len) <clbullar@ingr.com>

>IMO, we need it to make notions like the semantic
>web work reliably.  Long before the machines process machine
>processable descriptions, humans have to agree
>on what those descriptions mean.

I wonder.  Perhaps (as well?) we need to go the other way: concentrating on
information hiding--figuring out how to notate data so that it can be
manipulated by as generic tools as possible.  Perhaps we need to be reducing
the need for human agreement on meanings as much as possible.   To be able
to route information without knowing its meaning; to be able to build
structural dependencies into schemas so that users can take document types
for granted (and not need know why a stuctural constraint is--unless they
want to)

Without the widespread availability of generic tools, it is too difficult to
have specific data; and generic tools=layered standards for most intents and
purposes.

If the semantic web was AI, then human-machine descriptions would be king; I
suspect the semantic web will emerge out of pragmatic and incremental
development--raising the bar for generic manipulation of generic structures
each time.