courteous details Re: Semantic Web Services Framwork v. 1.0

From: Benjamin Grosof (bgrosof@MIT.EDU)
Date: 05/13/05

  • Next message: Sheila McIlraith: "AAAI-05 and ISWC"
    Hi Peter,
    At 11:03 AM 5/13/2005, Peter F. Patel-Schneider wrote:
    >I have a number of comments and questions concerning the document "Semantic
    >Web Services Language (SWSL) - Version 1.0".  Most of these questions have
    >to do with the SWSL-Rules language.
    >Note that this is by no means a complete catalog of my concerns concerning
    >the document.  Answers to these questions, particularly my fundamental
    >concern about semantics, may give rise to many other concerns.
    >Section 2.14
    >I am disappointed that the semantics for SWSL-Rules is not included in this
    >document.  Pointing to external documents is not a substitute.  I believe
    >that the external documents do not provide a semantics for the various
    >SWSL-Rules subsets.  I cannot find any transformation to define the
    >Courteous layer in Grosof2004a.
    It's given in three papers referenced (from Grosof2004a) rather than 
    directly in that paper.
    (See -> 
    Courteous Logic Programs)
    That's because there was an original version that was more expressively 
    then two steps of expressive generalization.
    Briefly, courteous LP semantics is defined there via a transformation to 
    ordinary/normal LP.
    This transformation is called courteous compilation.
    I'm currently working on:
    1. a new self-contained paper about courteous LP semantics, including some 
    additional theoretical results about it; and
    2. a new detailed description of the courteous compiler algorithmic design 
    used in SweetRules V2.1, with my student
            Shashi Ganjugunte who was lead developer for that.
    These will be available by sometime in June I expect.
    >VanGelder91 does not provide a semantics
    >for the = and != operators.  Without semantic definitions, it is hard to
    >determine just what is going on in the language.
    >Section 2.2
    >Is there any difference between f and f() as first-order atomic formulae?
    >The unification (=) and disunification (!=) operators are not appropriately
    >handled.   First, what does "identical" mean here?  Is it before or after
    >prefix expansion, for example?  Second, "substitution" has not been
    >defined.  Third, what is the scope of the subsitution?  For example,
    >         p(?x) = q(f(?y)) and p(f(?x)) = q(?y).
    >And/or formulae are ambiguous.  Consider
    >        p1 and p2 or p3
    >Is this the conjunction of an atom and a disjunction or the disjunction of
    >a conjunction and an atom?
    >Section 2.3
    >Any appeal to semantically-related notions (like "equivalently") is suspect
    >here, as no semantics has yet been defined.  This is particularly true for
    >illegal syntax.
    >I don't understand the stated allure of Horn rules.  Couldn't I make a
    >similar statement about arbitrary first-order formulae being independently
    >characterized by entailment, models, and deductive consequences?  How is
    >this in any way less desirable than being characterized by entailment,
    >a minimal model, and deductive consequences?
    >Section 2.4
    >The appeal to semantic reductions is not appropriate here.  First, no
    >semantics has been given yet.  Second, which semantics is to be used?
    >Section 2.5
    >No syntactic characterization of the extended syntax for this layer is
    >Transformations into a syntax which is not defined do not provide useful
    >Section 2.6
    >There is no definition of "equivalent" to use here, nor is the syntax
    >Without any proofs that the transformations preserve any useful
    >characteristics of the formulae, they should be treated very suspiciously.
    >Peter F. Patel-Schneider
    >Bell Labs Research
    Prof. Benjamin Grosof
    Web Technologies for E-Commerce, Business Policies, E-Contracting, Rules, 
    XML, Agents, Semantic Web Services
    MIT Sloan School of Management, Information Technology group or

    This archive was generated by hypermail 2.1.4 : 05/13/05 EST