Archives for: February 2008

Thursday, February 21st 2008

Permalink 09:37:17 am, Categories: Meeting summaries

Meeting Summary 18 February

The meeting this week focussed on one topic - how to apply POWDER where there is no structure in a Web site's URIs that can be used to create meaningful URISets. POWDER is about creating sets of URIs and then describing them - those sets being defined by the host, path components etc. But there are cases where this isn't possible and all one can say is something like 'example.org contains some resources that are blue and some resources that are red.' The danger is in creating two statements that conflict, i.e. 'all resources on example.org are blue' AND 'all resources are red.' After much discussion, it was provisionally agreed that the way around this would be to allow POWDER documents - that is, XML files with a root element in the POWDER namespace that contain an attribution element - not to contain any DRs at all. Rather they can contain simply descriptors elements that are not part of any DR and therefore have no scope. The attribution element MAY however include an aboutHosts element, the value of which is a white space separated list of hosts. If such an element is included, then the data in the document, whether in DRs or just descriptions, can only be applied to resources available from those hosts. Such a POWDER document provides some discovery information - example.org contains both red and blue resources - but does not provide any further information about where on that host those resources are. One would need to look at the resource (or perhaps the HTTP headers) to note a link pointing to a particular description. It's not ideal but a) it does not break the POWDER model; b) it does not lead to the creation of conflicting DRs; c) the expressivity is limited by the (lack of) URI structure across the described resources, not by the design of POWDER.
Phil ARCHER
Permalink 09:34:01 am, Categories: Meeting summaries

Meeting Summary 11th February

The meeting began with a brief check on progress towards our next outreach meeting. Invitations have been sent out, the agenda is evolving and all but one speaker has confirmed attendance. At the time of writing, there is likely to be space for a few more guests. The group then turned its attention to the new POWDER model that was discussed at the recent Athens face to face (see Jeremy Carroll's summary and previous entries on this blog). Some questions and answers: Is attribution element mandatory? Group resolved yes. Is maker element within attribution mandatory. Group resolved yes (which, as an aside, retains POWDER's dependence on the FOAF vocabulary) Why is the attribution element outside the DRs now? Because it's the only way to make statements about triples without using either reification or named graphs, both of which are unsafe for various reasons. So how can you have DRs with different attribution? By having multiple POWDER documents (and it's worth noting that one POWDER package may link to DRs in other POWDER documents). What about the human-readable text and/or icon - they need to be associated with a DR, not the document. True! Are descriptors RDF predicates and objects or XML attributes and values? Both. Literal values can be encoded either way, but where the object is a URI it must be encoded using the rdf:resource="..." attribute. It's probably better and less ambiguous to use RDF/XML notation for the descriptors. Actually, only the URISet is constrained - this must not include any element from any namespace other than POWDER. Development of the XML Schema for POWDER will be influential here. The conversation then moved on to look at the problem of 'switchable DRs' or cascading DRs. One solution was proposed and found wanting. We're still looking at how to enable a resource to include a link to a DR that describes, perhaps overriding a description already in cache. There was a brief discussion of the group's timeline: the expectation is of new draft documents being published very soon and the hope of reaching Last Call at Easter. Note - the WG includes several Greek members - so Easter may be according to the Orthodox calendar! We will need to seek a charter extension. Finally, e-mails received on the public list were acknowledged.
Phil ARCHER

Monday, February 4th 2008

Permalink 10:18:41 pm, Categories: Meeting summaries

Athens Meeting Decides Final POWDER Format

Several members of the Working Group gathered at the Institute of Informatics and Technology, part of NCSR "Demokritos," for this 2 day face to face meeting and were very pleased to welcome guests Irini Fundulaki of FORTH, Giorgos Stoilos of NTUA and HP's Jeremy Carroll who attended in his capacity as the OWL 1.1 WG's liaison to POWDER. Substantially as a result of Jeremy's input at the meeting and beforehand on both the GRDDL and POWDER mailing lists (with significant input from David Booth) the meeting has agreed a new representation of Description Resources. This will be written up in new drafts of the POWDER technical documents as soon as possible, with publication expected within two weeks or so. In brief, a POWDER document will be written in XML with a root element in the POWDER namespace. The format of such a document will be similar, but not identical, to that already proposed. These documents encode the operational semantics of POWDER – attribution, scope and description – and may be processed in a non-semantic environment. Critically, however, a GRDDL transform associated with the POWDER namespace, when executed, will create an RDF/XML instance with formal semantics that underpin and effectively define the operational semantics of POWDER. Two semantic extensions will be defined. The first will relate a string of the form "http://example.org" to the superficially similar URI <http://example.org>. The second defines the semantics of the white space-separated lists of strings envisaged in the Grouping of Resources document so that they can be interpreted as components of a URI. Thus, the following example can take on the intended meaning of the set of URIs that have a host component with a first and second level domain of example.org or example.com: <wdr:includeHosts>example.org example.com</wdr:includeHosts> More details will be presented in the technical documents, however, the general approach is now agreed. It meets the difficult challenge of creating a system that can work in the day to day, non-semantic web scenarios envisaged in the use cases whilst allowing the data to be processed within a semantic web environment. Several other issues were discussed, such as potential implementations, the test suite, exit criteria for Candidate Recommendation and so on, but the headline from the meeting is clear. After a period in which the group's fundamental approach has been questioned (and was very nearly jettisoned), there is a feeling that the work is back on track. The group records its thanks to NCSR for hosting the meeting, welcomes the interest from its guests, and records its gratitude to Jeremy Carroll for his insight and clear explanations.
Phil ARCHER