Re: Tentative contribution to the "URI issue"

I (again!) complete agree with Michael on all counts, and disagree as
completely with Jos. Local name scoping is a fundamental concept of
*abstraction* - namely establishing clearly what is, and isn't,
independent of the reste of the galaxy, and have only local meaning.
(Thank you Alonzo Church!)  If "merging" of modules is needed (for
whatever reason, and whatever "merging" means"), then as Michael
appropriately explained, consistent renaming - a.k.a. (beta) substitutions -
is the cleanest and most efficient way of doing it. Are we reinventing
the squalre wheel here? :-)

Anyway, to cool down the debate and make a self-evident statement:
this issue is indeed controversial and warrant sfurther (educated and
educating) discussion.

-hak

Michael Kifer wrote:
>>>>[...]
>>>>RIF is an exchange language for rule sets that are not necessarily
>>>>web-related.
>>>>There are tons of such sets.
>>>
>>>I agree with Michael here.
>>
>>I agree that there are lots of rule sets which are not Web-related. I do
>>not agree with Michael's conclusion that we not use URIs in such a case.
>>I support Dave's argument earlier in this threat.
>>
>>In fact, if you want to use local symbols for your predicates are
>>constants, you have to be really careful about how you want to treat them.
>>RDF treats such local symbols as existentially quantified variables.  I
>>guess we do not want to do that here.
>>However, the question is then what to do when merging rule sets which
>>have local symbols with the same name.
> 
> 
> Merging rule sets is a very bad way of thinking about integration, IMO.
> 
> 
>>If you use URIs for the interchange, you circumvent these problems.
> 
> 
> You don't circumvent anything. You are just imposing an undue burden
> on the developer and bloat your standards.
> 
> If you really want to merge rules (which is a bad way of thinking, as I
> said) then it is really easy. Have the merging tool consistently rename the
> local predicates. Voila!
> 
> 
>>I guess the point here is that the use of URIs really facilitates the
>>interchange, no matter whether a rule base to be interchanged is
>>"Web-based" (whatever that may mean).
> 
> 
> Disagree completely. This is just a big burden.
> 
> 
> 
>>>>Second, in a complex KB, you always have internal predicates that
>>>>shouldn't
>>>>be visible outside. Why should they be given a URI?
>>>
>>>I agree even more! One detail of momentous importance that seems to
>>>escape
>>>anyone envioning a system where *all* identifiers and constants are URIs
>>>is that these are precisely that - *universal*. In other words, they
>>>defeat the concepts of local scoping, hiding, and modularity - something
>>>desirable for any respectable programming idiom. Namespaces are a
>>>poor-man
>>>way of somehow working around this flatness.
>>
>>I agree with Dave's argument that the concepts of scoping and modularity
>>are completely orthogonal to the use of URIs. At the moment I am not
>>sure whether we need hiding or not.
> 
> 
> Strange (that you think that it is possible to do things without hiding).
> 
> 
> 	--michael  
> 
> 


-- 
Hassan Aït-Kaci
ILOG, Inc. - Product Division R&D
tel/fax: +1 (604) 930-5603 - email: hak @ ilog . com
http://koala.ilog.fr/wiki/bin/view/Main/HassanAitKaci

Received on Wednesday, 21 March 2007 16:06:26 UTC