- From: Sergey Melnik <melnik@DB.Stanford.EDU>
- Date: Wed, 17 Nov 1999 12:18:27 -0800
- To: Janne Saarela <js@pro-solutions.com>
- CC: www-rdf-interest@w3.org
Janne,
in my understanding RDF nodes cannot exist independently of the arcs
they are hanging on. In this sense, the semantics of set operations is
straight-forward:
> For example: remove a model of one triple (A,B,C) from model (A,D,E).
> Is the result an empty model or one single node E?
>
> Or: what is the intersection of (A,B,C) and (A,D,E)?
{(A,B,C)} minus {(A,D,E)} = {(A,B,C)}
{(A,B,C)} intersect {(A,D,E)} = {}
{(A,B,C)} union {(A,D,E)} = {(A,B,C), (A,D,E)}
etc.
Thus, removing and adding nodes, i.e. specific resource URI, to the
model is not necessary.
However, thinking more about the core API, I'm tending to say that the
set-oriented view on RDF models may be not appropriate in some
circumstances. For example, if your models represent datasources a la
Mozilla, it arguably may not make sense to intersect an email folder
with a file system directory.
The question is whether is it reasonable to reduce the core API even
further, to remove union, intersection and difference and to place them
into a separate interface that extends the core Model.
Sergey
Received on Wednesday, 17 November 1999 15:13:58 UTC