W3C RDFCore Working Group
Previous: 2001-07-06 (archive)
See also: Meeting Transcript
Minutes, as written by Jan Grant, are approved.
Frank has an item about the DAML Joint Committee's submission to the working group which we agreed to address early in the meeting.
The Working Group acknowledged receipt of the DAML coordination document (@@link) and agreed to send a response to DAML. Frank Manola was assigned an action to draft an assesment of the document and describe how it fits in with the issues we're discussing. Brian McBride was assigned an action to inform the DAML Committee of our receipt of their message and let them know that we hope to have a more detailed response by the end of the month.
The Working Group agreed that <http://purl.org/net/dajobe/2001/06/ntriples/> would become the official definition of N-Triples. Brian thought it would be likely that it would be a key part of the spec. The intention is to keep the format and document updated as we continue to define the abstract syntax. Dave Beckett and Art Barstow will meet to decide who will maintain the document. The document will be moved into W3C webspace with the proper copyright notice, etc.
It was agreed that we would link <http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jul/0039.html> from the issues list. Next week we will decide whether to accept it as an issue resolution.
Sergey Melnik sent <http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jul/0033.html> with a proposed ordering of topics for the Working Group to focus on, which people mostly agreed upon. However, there is also the issue of scope which we weren't all clear where to place in our priorities. All of the items are expensive topics but this first-cut summary should help us.
Brian sent <http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jul/0072.html> proposing the summarization and dividing the open issues into two groups. The top half are ones we should focus on, since they affect the abstract syntax. Everyone should speak up if they have other issues which they feel are important. Sergey raised the issue of namespaces which he volunteered to own.
It was also agreed that a model theory would help us answer questions. Pat Hayes volunteered to provide an incomplete strawman to give us a focus for discussion. Sergey pointed out that issues of terminology were also important, and might hamper progress on a model theory. Pat Hayes felt that both could proceed in parallel and help progress with the other. There was general agreement on this point.
Graham Klyne reminded the Working Group of <http://public.research.mimesweeper.com/RDF/RDFTerminologyConcepts.html> which he began to put together back on RDF Interest.
Aaron pointed out that the model theory should not try and take the place of answering the questions raised by the issues. Pat hayes agreed, saying that the model theory is a mathematical sandbox to focus the questions and try out the answers.
Brian explained that he wanted to begin reviewing issues with high priority to see if we could come to some sort of resolution. Graham Klyne summarized that he though there was some progress based on his recent exchange with Frank Manola, but he was especially curious if Dan Connolly had any comments. He felt that he had narrowed the issue down by dismissing query as out of scope that left only the question of what were the semantics of an anonymous resource. Pat thought that dismissing query may not be well accepted by many. Graham said he would wait until someone raised the issue. Frank Manola suggested query might need to go into the M&S while Pat Hayes suggested the document might have explicit warnings against common misunderstandings, something the DAML group plans to do. Frank suggested two major approaches: 1) recognize that people might use RDF in a query context but leave others to specify the semantics for that situation or 2) do the extension ourselves. Graham said he felt that the current spec uses anonymous resources to say 'there exists something with these properties'. In response to a question from Jos de Roo, he clarified that there could be more than one thing and referred to <http://lists.w3.org/Archives/Public/w3c-rdfcore-wg/2001Jul/0124.html> (?) for more precise wording. He felt that it really came down to two choices: 1) a form of skolemization -- generating IDs for anonymous resources or 2) introducing a more explicit scoped variable. He felt there were problems each way: with (1) the question was how to generate the ID, with (2) it was how to scope something in the abstract syntax. Brian pointed out that the rdfms-graph issue might provide a URI for an RDF graph, and thus might help in this area. Brian asked whether the IDs generated by skolemization would be distinguishable from normal URIs. Graham felt they would not necessarily be distinguishable, but that he wasn't sure about it. He said he didn't have a strong view on the issue but was leaning towards scoped variables. Jos de Roo commented that the scope for these variables could be the document itself, which he felt N-Triples made quire clear with its _:anonnode syntax. Graham felt that problems could arise when documents were combined and though you really need a way of capturing boundaries within the abstract syntax. Frank Manola wondered if this was a syntactic exercise or a more substantial function of merging graphs, bringing up the example of logical statements. Pat Hayes said that in logic you had to rename variables when you merged documents but said he thought that's what URIs were supposed to do.
Brian sent a summary of use cases late last night. He felt that M&S was quite clear on how a literal should be represented. He presented his summary saying that Jan Grant implemented it the way M&S described and preferred that. Martyn Horner found that they modeled language using anonymous resources and didn't use the M&S way. Eric Miller provided a similar use case, but we could not determine whether the M&S met his needs. He also raised some questions: Do we understand what M&S says about xml:lang? The answer seemed to be yes. Is what M&S describes broken - is there some fatal flaw that must be fixed? Pat Hayes felt that clearly not, since some have implemented it. Aaron Swartz raised that Tim Berners-Lee and Dan Connolly felt it was broken and thought we shouldn't go on without their feedback. Graham suggested three possible ways to do things: 1) Define literals as complex entities with a language property; 2) model it in the graph; 3) adopt both - let the market decide or let a thousand flowers bloom. Pat Hayes thought that this would lead to contradictions and inconsistencies. Brian volunteered to do a writeup of the pros and cons of the alternatives.