Re: DQL update, etc.

From: Ian Horrocks (horrocks@cs.man.ac.uk)
Date: 03/11/03

  • Next message: Richard Fikes: "KSL DQL Web Site"
    On March 10, Richard Fikes writes:
    > > During the query discussions at the W3C meetings last week
    > > in Boston, I realized that we never officially published an
    > > updated DQL [1] specification addressing the feedback
    > > (answer templates, etc.) from the October DAML PI Meeting.
    > > Should we try to publish an update before the April PI
    > > Meeting?
    > Yes, I think so.  I did some updating of the spec when we were writing 
    > the DQL paper.  I can dust that off and send it out as a proposed revision.
    > 
    > > A use case also came up that I don't believe DQL can clearly
    > > cover.  Suppose we have a knowledge base containing contact
    > > information (name, phone, emailAddress, etc.) for
    > > individuals.  A given individual may or may not have an
    > > instantMessageAddress.  Can we write a single DQL query that
    > > retrieves name, email, and the instantMessageAddress only if
    > > present?  Does this require some mechanism for indicating
    > > that a statement is optional, or can it just be handled
    > > using a may-bind variable for the object of the
    > > instantMessageAddress statement?
    > 
    > I don't see any way of doing that with a single DQL query.  May-bind 
    > variables are only operative when the server can prove that the property 
    > value (in this case a value for instantMessageAddress) exists.  I 
    > suppose the client would need to do a query asking for all of the 
    > person's contact information that the class definitions in the knowledge 
    > base say that the person is required to have (e.g., with cardinality 
    > restrictions).  For that query, may-bind variables could be used for 
    > values that may not be in the knowledge base but that the theory says 
    > must exist, and then do a follow-up query asking for the person's 
    > instantMesageAddress.
    
    I agree with Richard. Of course it would be easy for an application to
    ask two queries, one asking for the instantMesageAddress and the other
    not, and "merge" the results in outer-join style. In fact, extending
    conjunctive query languages (like DQL) to deal with disjunctions of
    conjunctive queries is in general relatively straightforward, and
    would allow cases like this to be handled (some dummy value would have
    to be returned for bindings with no instantMesageAddress).
    
    Ian
    
    > 
    > > There appears to be significant interest in forming some
    > > sort of W3C Semantic Web Query Working Group.  Should we try
    > > to publish DQL as a W3C Note that can be more easily used as
    > > an input to such a group?
    > 
    > Yes.
    > 
    > > It would probably be best if such
    > > a note included a concrete syntax.
    > 
    > Agreed.
    > 
    > Richard
    


    This archive was generated by hypermail 2.1.4 : 03/11/03 EST