- From: Michael Kifer <kifer@cs.sunysb.edu>
- Date: Tue, 25 Mar 2008 19:39:44 -0400
- To: Jos de Bruijn <debruijn@inf.unibz.it>
- Cc: RIF WG <public-rif-wg@w3.org>
> > I think the current proposal for meta data is inadequate for the current > > and future needs because it is not part of the syntax of the language, its > > proposed is too limited, and in the current form it cannot be given a > > semantics. > > Right. That is why I also proposed the "directives" [1] (even though > the term is probably not appropriate; so if you have anything better). > The examples you mentioned are actually good examples of the kinds of > things directives can be used for. So, we have the current list of > examples for the use of directives: > > These are the ones we I think we really want to have in BLD: > > - import of RIF rule sets > - reference to external data sources (e.g. RDF) > - reference to external data models (e.g. RDFS/owl) > > These are things to keep in mind for extensions of BLD: > > - sorting of query answers > - priorities > - preferences > > If we would need variables in directives, extensions to BLD can extend > the notion of "directive". > > Note that still the same discussion (we've had already several times) > applies: whether we want to have a generic syntax element "directives" > or whether we want to extend the syntax for each of the possible directives. OK. But the more general syntax should then be added to FLD -- to keep us honest. Also, this still does not address the issue that the meta syntax is not accessible to processing by rules. --michael > Best, Jos > > [1] http://www.w3.org/2005/rules/wiki/Annotations > > > > > Here are some concrete problems. > > > > 1. Meta data can be attached to various parts of the rules, not just rules > > themselves, and this metadata can affect the semantics. > > A simple example is specifying that the output to a query must be sorted. > > This changes the semantics, since the answer is not a set any more > > but a list. > > > > 2. Some important types of rule systems are based on prioritized defaults > > (defeasible, courteous, preference LP). > > > > - Here metadata is part of the syntax and of semantics. > > - Typically metadata consists of rule labels and priority or > > preference information. > > - Metadata items are often *terms with variables* and not just strings. > > > > 3. The current proposal does not offer any obvious or natural way for the > > metadata to be processable by a (possibly different) rule set. > > > > > > > > > > --michael > > > > -- > debruijn@inf.unibz.it > > Jos de Bruijn, http://www.debruijn.net/
Received on Tuesday, 25 March 2008 23:40:40 UTC