Archives for: May 2008

Monday, May 19th 2008

Permalink 03:23:26 pm, Categories: Meeting summaries

Meeting Summary 12 May 2008

The detail in several of the group's documents were discussed today, in particular the new one concerning the formal semantics of POWDER and the Test Suite, both of which should be published as first working drafts very shortly. More than ever, each document depends on detail in another. Meanwhile work is progressing well with the XML Schema and datatypes that will aid the validation of POWDER docs. The first half of the meeting picked up on the recent e-mail exchanges [member only link] concerning the formal semantics model. The outstanding questions come down to things like: how important is it to be able to process non-Web URIs? Does a POWDER Processor necessarily have to support an extension mechanism such that it can process any POWDER-like document if the extension is declared properly or is it OK to require additional code? And is grouping by IP address and port number essential? If the answers to all these is no then it all gets much easier so, with time running out for the group, decisions need to be made about what really is essential and what is a nice to have. Many of these issues are really only coming to light as we tie down the details and start to work out actual tests we'll need to pass ourselves (never mind anyone else!). It was the Test Suite that filled most of the remainder of the meeting. What exactly are we testing? It seems that there are two 'products' to test: the POWDER to POWDER-S transform and the POWDER Processor. We need to be clear about the difference between the Test Suite and Candidate Recommendation phase. A document that discusses the sub divisions of specifications looks highly pertinent to both the formal semantics discussion and the Test Suite development. The suggestion was made that as well as providing test data we should also provide examples of predictable mistakes and how processors should handle them - the problem is that it's an open ended question and this seems unlikely to be included in the Test Suite.
Phil ARCHER

Friday, May 2nd 2008

Permalink 03:13:22 pm, Categories: Meeting summaries

Meeting Summary 28th April 2008

The group discussed the concept of 'a POWDER Processor' and agreed that such a thing needs to be defined. As a minimum a POWDER processor will handle two functions: For a given URI u, the function describe(u) will return RDF/XML that has u as the subject. For a given description, d, the function resourcesWith(d) will return one or more regular expressions, all of which must match for a resource to be described by d. There are many options that can be added to this - describe using a given dataset, only use DRs from particular creators etc, returning DRs relevant to the query and so on. Many of these options will be documented but only the two core functions will be mandatory. The group then turned to discussion of the work done recently by Stasinos Konstantopoulos. Subject to consultation outside the group, this work will become a separate new document called POWDER: Formal Semantics. This will underpin the semantics of POWDER-S which in turn will underpin the 'native' POWDER. This creates a 3 layered approach, however, the 3 layers all build on each other. Only POWDER and POWDER-S need be implemented, but the formal semantics, although largely theoretical, provide a benchmark against which the other two can be tested. Since a lot of work has been done in this area already, it will not add significantly to the group's workload. Finally, subject to final confirmation, the next face to face meeting will be on Monday 14th - Tuesday 15th July in London.
Phil ARCHER