Warning:
This wiki has been archived and is now read-only.

LC2 Responses/BG1

From OWL
Jump to: navigation, search

To: BenjaminG@vulcan.com
CC: public-owl-comments@w3.org
Subject: [LC response] To Benjamin Grosof

Dear Benjamin,

Thank you for your comment
     <http://lists.w3.org/Archives/Public/public-owl-comments/2009May/0024.html>
on the OWL 2 Web Ontology Language last call drafts.

Regarding the semantics of rule systems, a detailed discussion of these matters would be far beyond the scope of this document (which is, in the end, primarily specification). We would expect a reader interested in the design to consult the references provided. A few words and an additional reference have, however, been added [1].

Regarding establishing a mechanism for recognizing and naming useful subsets of expressiveness, it is outside the remit of the working group as we cannot set up a *process* of indefinite length at the W3C. We will pass the suggestion on to Semantic Web Coordination Group. Alternatively, the OWLED workshop series is a good place for this kind of work [2], and suitable subsets could be the subject of future W3C member submissions in this area.

Regarding a primer on implementation design considerations and techniques, it is both far beyond the scope of this Working Group and not possible given our current resource constraints. Furthermore, such a document should be kept up to date, which is difficult to do with a W3C technical report. This seems to be an ideal document for various third parties to produce.

Regarding theorem PR1, producing a worked out proof would be a tedious and, we believe, not especially enlightening task. The proof sketch should be sufficient to be convincing and to allow the interested reader to fill in any missing details.

Regarding the expressive restrictions in OWL 2 RL, some features have been omitted in order to facilitate "easy and efficient implementation using existing forward-chaining rule systems". Reflexive object properties, for example, would require rules that operate over all individuals, which is likely to compromise efficiency, and may not even be possible in some rule systems. It may be that some of these features could, in theory, be added, but we have been mindful of specific major systems (such as Oracle and Jena) and their implementation concerns.

[1] http://www.w3.org/2007/OWL/wiki/index.php?title=Profiles&diff=23681&oldid=23680
[2] http://www.webont.org/owled/

Please acknowledge receipt of this email to <mailto:public-owl-comments@w3.org> (replying to this email should suffice). In your acknowledgment please let us know whether or not you are satisfied with the working group's response to your comment.

Regards,
Ian Horrocks
on behalf of the W3C OWL Working Group



CUT AND PASTE THE BODY OF THE MESSAGE (I.E. FROM "Dear" TO "Group") INTO THE BODY OF AN EMAIL MESSAGE. SET THE To:, CC:, AND Subject: LINES ACCORDINGLY.

PLEASE TRY TO REPLY IN A WAY THAT WILL ALLOW THREADING TO WORK APPROPRIATELY, I.E., SO THAT YOUR REPLY CONTINUES THE THREAD STARTED BY THE ORIGINAL COMMENT EMAIL



% note with comments on OWL 2 RL Profile, public draft of 2009-04-21 % by Benjamin Grosof, May 12, 2009

Hi Boris, Ian, Zhe, and other OWL 2 Working Group folks,

It's with pleasure (but in a bit of a rush) that I have just read the OWL 2 RL Profile document, public draft of 2009-04-21 [1], in response to Ivan Herman's recent personal request that I review it as part of the 2nd Last Call process.

[1] http://www.w3.org/TR/owl2-profiles/ -> #OWL_2_RL (i.e., section 4)

Overall, the document looks like a quite good job. Below are my suggestion comments on it. No show stoppers. But imho they will help make the document stronger and the adoption a bit easier on implementers and users. They are given in descending order of importance.

Note these comments are by myself as an individual, and do not represent the views nor position of Vulcan Inc.

I've attached this email as a plaintext file, too, for your convenience.

Best, Benjamin


1. In section 4.3 (Reasoning in OWL RL ...), towards the end, it would be good to have a brief discussion about how the direct semantics of most existing implemented rule systems is based on (either rigorously, or in reasonably close hailing distance to) declarative logic programs (LP) rather than FOL, and then point to reference work describing the relationship of LP semantics to FOL semantics.

2. A lesson learned from the experience of OWL 1 is that it would be useful to have some kind of communal mechanism, ideally under the auspices of W3C, to recognize and name useful subsets of expressiveness other than the ones specified as part of the Recommendation itself (-DL, the Profiles, etc.), that emerge as well understood and useful only after the Recommendation. E.g., expressive relaxations that preserve desired characteristics. E.g., expressive restrictions that improve computational performance or improve simplicity of implementation. While not absolutely necessary for the Working Group to do it at this point, it would smooth the way. Let's learn from history, folks!

3. Ideally, there would be another document in the OWL 2 document suite, aimed at implementers as an audience, that acts as a primer on the implementation *design* considerations and techniques, including specifically for the RL Profile. I didn't find anything like this. (The Conformance document focuses on *test*, as opposed to design.)

4. In section 4.3, theorem PR1: The proof of this central result should be more than a short sketch, i.e., should be elaborated. This could be via pointer to a separate reference document.

5. In section 4.2 (Profile Specification): Why are the following three expressive restrictions imposed? Please explain. (Or perhaps generalize if the restrictions can indeed be relaxed.)

 a. No reflexive object property axioms (ReflexiveObjectProperty). Why?
 b. No self property restriction (ObjectHasSelf) in class expressions.  Why?
 c. Data range expressions (DataRange) may only be formed by intersection
    (DataIntersectionOf).  Why?  In particular, why not permit a data
    range formed by union (DataUnionOf or DataOneOf) in a
    DataSomeValuesFrom within a subClassExpression?

Benjamin Grosof, PhD -- Semantic Technologies. Sr. Research Program Manager, Vulcan Inc. Head of Project Halo Advanced Research (HalAR) program.