- From: Dave Reynolds <der@hplb.hpl.hp.com>
- Date: Wed, 17 May 2006 14:28:19 +0100
- To: Bijan Parsia <bparsia@isr.umd.edu>
- CC: public-rif-wg@w3.org
Bijan Parsia wrote: > > On May 16, 2006, at 10:46 PM, Vincent, Paul D wrote: > >> +1 >> >> [Actually, I find it worrisome that this should be questioned at all!] > > Well, I have some questioning. The W3C already has a plethora of > languages and apis for dealing with XML...are we going to reuse them? > Exactly how? What does "working with/against XML data" actually mean? > (And XSLT is already a rules language...do we need to include it?) XPath > integration? I assume the requirement is to do "other stuff" with the result of the XML data matching than is normal with XSLT et al, e.g. deduce new logical statements about business processes, which fires further business rules or actions. > It's not like that we can *prevent* rules from working on something > which reasonably counts as XML data. On the other hand, there's reason > to be cautious in our positioning since there are other "products" at > the W3C that we definitely don't want to be seen to compete with. Seems to me there are at least three approaches to how we might support matching over XML data (directly analogous to the options with RDF): 1. We define a built-in which takes an XQuery (or whatever) specification and a set of variables to bind. Similar to the proposed SPARQL built-in. [Modular, little work for us, minimal conflict with existing tools.] 2. We build XPath (or whatever) primitives directly into the condition language and include XML node sets (or whatever) in the universe of data values to which variables can be bound. [More powerful (?), more work, more overlap and potential for confusion.] 3. We say that the mapping from XML data to Atoms in RIF language is not part of RIF itself, some other layer does that mapping. Then someone can use their own favourite XML tool (there are so many) to extract values from the XML but just pass those values onto their RIF processor. [Maximum freedom, no work for us, doesn't lead to much interoperability since the rule users will also need some way to agree on the mapping before their rule exchange becomes effective.] > I suspect that some sort of XPath integration would be relatively > straightforward and uncontroversial (mostly), but specifying it would be > seriously nontrivial (in terms of amount of work). Is this important > enough to standardize as part of RIF per se? (As opposed to leaving it > to idiosyncratic application/library support that is opaque to the RIF > level?) How many rule languages currently have specific language level > support for XML? If none, there, assuming we're about "interchange", > there's nothing to do. That's a really interesting question. As you say XSLT is a kind of rules language. Schematron is described as a rule language (for integrity rules at least). Would people regard schematron as a legitimate rule language to transfer via RIF? Dave
Received on Wednesday, 17 May 2006 13:28:35 UTC