RE: Are we losing out because of grammars? (Re: Schema ambiguitydetection algorithm for RELAX (1/4))

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

I don't know what a Linda-system is.  As for the rest, 
that is the domain of ontological hermeneutics, the interpretation 
of texts, originally.  It doesn't matter if the systems share 
rules globally.  It matters that each systems acts and reacts 
in accordance with its own rules one of which may be to 
issue requests for verification to the other agency.  The discovery 
based systems rely on the interpretation of discoverable 
descriptions.  They can set up "ground rules" as they go.  
This becomes a protocol of behaviors by communicating types 
of information.   The schemas are adequate for that.  The 
issue is associating these with locally derived and 
maintained rules.   I think ultimately this is why namespaces 
get resolved to something like RDDL in many cases and 
directly to schema in some.   One can't sensibly rely 
on schemas without coupling a process of use for the 
instances.   

Simple exchange of documents or messages is not hard 
to do.  The problem is one of shared command over 
resource allocation.  Consider the problem of multiple 
stores that share the same delivery system, eg, trucks. 
A truck is en route to deliver a package for its owner 
agency when another agency that also has packages on 
the truck receives an order critical to a customer.  
Should the second agency negotiate with the first 
agency over rerouting the truck?  No.  It 
should negotiate with the truck driver.  Local rules 
for command and control.  Global rules for intelligence.

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

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


-----Original Message-----
From: Thomas B. Passin [mailto:tpassin@home.com]

This thought experiment might highlight what's needed in these cases Len
keeps
talking about.  Imagine that there is a Linda-like system, and its tuple
space
is relatively public, or at least shared by Len's agencies.  One agency puts
out a tuple with slots that are query templates asking to be filled.

Later, the agency retrieves the tuple and pulls out the field values.  How
can
it have some confidence that those field values are what it asked for?  Or,
to
bring the question closer to this thread, what has to be included with the
query so that another system can undertand what is being asked for and
evaluate whether it should respond?

Remember, in this Linda-like system, a responding agency only knows the
contents of the field in the tuple.  Nothing more, unless the whole system
is
set up with some ground rules.