- From: Jonathan Robie <jonathan.robie@softwareag.com>
- Date: Fri, 11 Jan 2002 16:05:02 -0500
- To: Vadim Draluk <vadim@draluk.net>, www-ql@w3.org
At 10:23 AM 12/27/2001 -0800, Vadim Draluk wrote: >Hi, > >here are my initial 2c on the new XQuery/XPath specs. > >1. External functions, mentioned in issues 124 and >223, is a very important feature indeed. However it >would be insufficient simply to define them as >outlined in 223. The invocation paradigm has also to >be addressed somehow: system library function, Java >class, EJB or COM interface method, SOAP invocation >etc, just to name some possibilities. The rest of XQuery is not defined in terms of the software used to realize it. Why should this be different for external functions? I think there is a real advantage to saying that an implementation may implement external functions any way it chooses, while still providing function prototypes that give the type information to the query processor. Naturally, implementations will have to provide a way of making external functions available. But does that need to be specified in the standard? >2. Recursion in dereferencing. As I understand it such >recursion is not allowed, unlike one over conventional >navigation axes. So the manager-employee example in >2.3.3 could not be generalized, say, to look for a >person some place up the reporting chain. This could >become more conspicuous an omission if dereferencing >syntax is merged with one of axes, as considered in >227. > >The recursion in question can become very useful, for >example, for advanced UDDI registry querying, allowing >to traverse tModel hierarchies. I certainly understand why you would think that. There are some important concerns for recursion in references, though. One is that references on the World Wide Web can create very large networks, with high latency times, and dreadful consequences for performance. Recursive traversal of references is unlikely to be well optimized in the general case. Given that, using recursive functions for traversing references seems to work pretty well. This is a strategy I have used successfully for both RDF and Topic Maps - see the following paper: http://www.idealliance.org/papers/xml2001/papers/html/03-01-04.html These are my opinions right now. They may be quite different from the opinions of Software AG, the W3C XML Query Working Group, or the opinions that I will have after reading and considering your response.
Received on Friday, 11 January 2002 16:05:23 UTC