comments on OWL 1.1 charter

[I take Ivan's message as an invitation to resend this publicly]

This is a personal comment. I have discussed it with HP colleagues, and
I am not aware of dissenting opinion, but this should not be read as an
HP position.

I have looked at
http://www.w3.org/2007/05/WebOntCharter.html
Revision 1.33
and have a few comments.

OWL Full
========

I am pleased to see the constraint
[[
All extension features should have a clear semantics both in terms of
OWL DL and OWL Full.
]]
but would be happier for the 'should' to be a 'must'.

I believe that this constraint, and other conservative aspects of the
draft charter, mean that several more contentious features of the OWL
1.1 member submission are, as currently formulated, out-of-scope (e.g.
annotations on axioms, the OWL 1.1 member submission version of the
mapping rules to RDF/XML, and punning). However, experience suggests
that the WG will need to operate by constructing an issues list, with
each such feature on the list, and then spend time discussing each
feature and considering whether that feature is incompatible with the
charter. This takes time, and I think the time scale is not realistic.
At least another six months is needed before LC. Partly this will depend
on how many of the difficult features, the WG desires to redesign in
order to have adequate OWL Full compatibility.

Issues
======

In addition, some features of the member submission are contentious, for
example, n-ary datatypes, see [1]. Such contentious features will need
to be dealt with as issues by the WG.

The issue list should also be populated by comments from
public-webont-comments on the OWL 1.0 documents.

The charter should require the group to formally address such comments.

There are also a number of postponed issues from the original WebOnt
issues list [2]. The charter should encourage the new WG to address such
issues where possible.

Requirements Document
=====================

Given that the change from OWL 1.0 to OWL 1.1 is large enough that there
are a significant number of issues, it will probably be effective to
have the first publication of the working group be a discussion
document, for example "Requirements for OWL 1.1", and delay FPWD of the
main specification by a few months.


OWL Full and OWL DL: divergence or convergence
==============================================

Towards the end of the OWL 1.0 work it was realised that we were not
going to achieve our initial objective [3] of aligning the semantics of
OWL Full and the direct semantics on the OWL DL syntactic subset. The
result is the weaker one-way implication of theorems 1 and 2 of section
5.4 of the OWL Semantics. In my view, the charter for the new group
should discourage further divergence between the two semantics, and
encourage fixing the current divergence if possible.

For example,
[[
All extension features should have a clear semantics both in terms of
OWL DL and OWL Full.
]]
to
[[
All extension features must have a clear semantics both in terms of OWL
DL and OWL Full. Such extensions must not introduce new OWL DL
entailments that are not OWL Full entailments, and should not (in the
RFC 2117 sense of should) introduce new OWL Full entailments that are
not OWL DL entailments. Where possible the semantic correspondence
between OWL DL and OWL Full should be strengthened.
]]

Tractable Subsets
=================

I think that while the people who work on tractable fragments and/or owl
tiny may be a different set of people, that it is likely to be
beneficial to have all such people in a single WG. Experience from
working on OWL 1.0 was that while the consensus between different
parties was sometimes difficult and fragile, it was possible to bridge
the differences of opinion and outlook, and the end result was
significantly better because of this effort. Having two WGs may result
in the OWL 1.1 WG being heavily weighted towards a DL perspective, and
the fairly delicate compromise achieved in OWL 1.0 may be lost.

I would be inclined to run the tractable fragments work in parallel,
with perhaps a six month lead time, for brainstorming, and a FPWD due at
T+8 (note in my timeline the LC on OWL 1.1 has been pushed back to T+16)

I think that tractable fragments could be scoped as possibly allowing
subsetting in many different ways:
e.g.
- vocabulary subsetting (only some terms)
- syntactic subsetting  (only some graphs)
- OWL Full semantic subsetting (only some of the conditions on OWL Full
interpretations)
- OWL DL semantic subsetting (only some of the conditions on direct
interpretations)

Test Cases
==========

The charter should require the WG to develop test cases for new or
modified features in OWL 1.1. My belief that the test cases should be a
delta document, but there are obviously many different ways to manage
the construction of a test suite.

Timeline
========

     * Month T: First teleconference
     * Month T+2: First face-to-face meeting
     * Month T+3: First Public Working Draft for Requirements
     * Month T+6: Second face-to-face meeting
     * Month T+6: First Public Working Draft for OWL1.1
     * Month T+10: Third face-to-face meeting
     * Month T+14: Fourth face-to-face meeting
     * Month T+16: Working Group Note for the XML Exchange syntax for
OWL1.1 (in case the WG decides to publish it separately as a note)
     * Month T+18: Last Call for OWL1.1
     * Month T+18: Fifth face-to-face meeting
     * Month T+22: Candidate Recommendation for OWL1.1
     * Month T+22: Sixth face-to-face meeting
     * Month T+24: Proposed Recommendation for OWL1.1
     * Month T+26: OWL1.1 Recommendation

Jeremy

[1]
Turner, David; Carroll, Jeremy
Practical Conforming Datatype Groups
http://www.hpl.hp.com/techreports/2007/HPL-2007-37.html
[2]
http://www.w3.org/2001/sw/WebOnt/webont-issues.html
[3]
http://lists.w3.org/Archives/Public/www-archive/2002Oct/0022



-- 
Hewlett-Packard Limited
registered Office: Cain Road, Bracknell, Berks RG12 1HN
Registered No: 690597 England

Received on Friday, 8 June 2007 12:55:19 UTC