RE: Possible simplification of RDF formal model

Desmond,
 
You've clearly been eating too much Smalltalk for breakfast! :-). Seriously,
other than the fact that I don't believe "7" or any other literal is a valid
URI (unless you could construe it to be a relative URL), I like the idea. I
personally strongly dislike the dichotomy between values and references and
values and resources. I'm not sure precisely what the solution should be but
I think you're pointing in the right direction.
 
Jeff

-----Original Message-----
From: Desmond D'Souza [mailto:desmond-dsouza@austin.rr.com]
Sent: Monday, January 10, 2000 12:07 AM
To: ora.lassila@research.nokia.com
Cc: www-rdf-comments@w3.org
Subject: Possible simplification of RDF formal model


Ora,
 
In the RDF formal model, have you considered treating Literals as a subset
of Resources? The string representation of that resource (some people
confuse this for "the value") actually is the URI of the resource.
 
So everything is a resource (Object), referred to by its URI. And all
properties are directed edges on this sea of objects. For example: 

*	The number 7 always existed as a unique individual, before any
particular document referred to it. The string representation "7" is its
URI. Its 2's compliment representation is an alternate way of referring to
7. (This does not fit the current syntactic definition of URI, but I'll
ignore that here). 

*	So, what useful properties does 7 have? 

*	7 + 1 is no more than the "next" property of 7 i.e. 8.  (or see 7 +
N below). 

*	7 +  5 = 12 is done as follows: 


*	The + property of 7 is another resource +_of_7 

*	+_of_7 is a function that maps any number N to 7 + N. Using the
triple notation: 

*	(0   +_of_7  7) 

*	(1   +_of_7  8) 

*	(2   +_of_7  9) etc. 


*	(A common functional programming technique)

*	All these are optimized away in practice e.g. 2's complement
representation of 7 ("0111") is an efficiently encoded URI of 7 that
exploits preknowledge of the encoded URI of 8 ("1000") ... so that to get
the (URI of the) "next" property of 7, you do not have to lookup the "next"
tag, but directly manipulate the URI of 7 itself to get the URI of 8.  


*	This is just a variation of what relative pathnames do: take prefix
of current URL, append relative offset, viola! URL of result.

*	Based on the above, I think rdf:_1, rdf:_2, rdf:_3 etc. are just 1,
2, 3, ... 

*	This might simplify dealing with other typical "value" types: dates,
money, etc. They have special URI syntax, a special way of comparing
normalized URIs, and of looking up propertis to determine other related
URIs. Thus, the date (pointed to by) 12/12/2999 always existed; as did the
amount (pointed to by) $7654321; and so on. 

*	This seems to fit in with RDF schema as well.

We use this in a cleaner approach to object modeling with UML (which we call
Catalysis). Everything is an object, including numbers, strings, etc. All
properties are links between objects. And all operations simply create or
delete links and objects, though some objects and links cannot be deleted.
 
Thoughts?
 
Regards,
 
Desmond D'Souza

Received on Monday, 10 January 2000 12:42:55 UTC