RE: Priorities for F2F (completion of ACTION: 2001-06-22#4)

>:Brian McBride:
>:
>:On the question of literals as resources, I've looked through
>:the recent
>:email traffic and the logs and minutes of last Friday's
>:discussion.  I'd
>:like to test support for this position:
>:
>:  o the WG agrees that URI's can be assigned to denote literals.
>
>Is this to say: the wg indicates that people may assign URIs
>to Literals in their RDF data? I think I would be fine with
>this, but I still don't understand what is being described.
>Is a Literal simply a resource that can be present in the
>web (ie on a computer)?

Seems to me that there are at least three possible positions one 
could adopt here.

1. URI's may be assigned to literals, in the sense that the language 
(in some future version) allows one to assert that something denoted 
by a literal is also the referent of a URI, so that an equality of 
the form (= <URI> <literal> ) could be asserted (maybe not with this 
syntax, but you get the idea.)

2. At the other extreme, the language could be understood relative to 
a semantics in which the things denoted by URI's and the things 
denoted by literals were required to come from disjoint (subsets of 
the) universe(s), so that it would be logically impossible for such 
an identity to every be true. ON this view, providing literals for 
some range of values effectively removes those things from the normal 
domain of quantification, so that it is impossible to have the same 
relation between, say, thingies and integers as it is between 
thingies and thingies.

3. An agnostic version where there was nothing which *required* that 
a URI could not denote a literal value, but on the other hand the 
langauge does not *provide* any way to express a URI-to-literal link, 
so that there are no URI-to-literal assignments and no ways to 
express identities of the above kind between URI-denoted things and 
literal-denoted things (such as strings or integers).

I would urge the advantages, in a language that purports to be very 
general-pupose, of agnosticism of the third kind. That leaves the 
door open for a future extension or modification of the langauge of 
the first kind.

DAML+OIL took the second route, drawing a sharp semantic contrast 
between objects and literal values, for essentially computational 
reasons: to let the two domains 'mix' would have produced language 
constructs that would have had unknown inferential complexity (or 
known intractability), and there was seen to be an overarching 
requirement to keep DAML reasoning tractable, or at least decideable. 
But other languages might take other directions.

Pat Hayes

---------------------------------------------------------------------
(650)859 6569 w
(650)494 3973 h (until September)
phayes@ai.uwf.edu 
http://www.coginst.uwf.edu/~phayes

Received on Thursday, 12 July 2001 16:07:17 UTC