RE: DAWG Action SWBP teleconference - original mails from Gary NG





Further to my submission of a recent email thread to the Working Group
regarding a potential response to the DAWG’s request for help, I have
reformatted the associated text for further discussion. I trust that the
following format hence conforms to W3C WG protocols, makes for easier
reading and accurately reflects the communication that took place? If not,
I am more than happy to reformat…

I have, of course, copied both Gary and Jeremy personally on this mail and
would especially appreciate their comments.

*********************************************************************************************************

My initial comments were:

I have now had time to look at Gary NG's response
(http://lists.w3.org/Archives/Public/public-swbp-wg/2004Sep/0049.html) to
the DAWG's request for feedback on RDF Data Access Use Cases and
Requirements and found it a measured and thoughtful consideration of the
issues currently being faced. Nevertheless my reading of Gary's reply
suggests that a set of tabled responses still needs to be debated, even
though Gary more that adequately provides substantial, valuable material
towards this.

For the purposes of reiteration, the DAWG have asked for specific BP
comment on:-

1.    XQuery, syntax and integration: We're chartered to "... maximize W3C
technology re-use, while also taking account of differences between the RDF
graph data model and the XQuery data model" and to allow "... for RDF data
to be accessable within an XML Query context".
2.    Rules, Additional Semantic Information

      We have an objective

      "It should be possible for knowledge encoded in other semantic
      languages-for example: RDFS, OWL, and SWRL-to affect the results of
      queries executed against RDF graphs."

      http://www.w3.org/TR/rdf-dawg-uc/#d4.6


      and in discussion of rules and query
      http://www.w3.org/2001/sw/DataAccess/ftf2#qrdesigns


      we noted a connection between rules and a CONSTRUCT
      mechanism found in various contemporary designs, including
      our current draft

      http://www.w3.org/2001/sw/DataAccess/rq23/#construct

      $Revision: 1.25 $ of $Date: 2004/08/16 12:23:00 $

      Any rules/query integration experience to share? Thoughts
      on best practices for accessing RDF data, while rules
      work is still in the early stages of standardization?

Gary further suggests comment of the following:

3. I also noticed that they would like to address "data source
identification" within the query language.

In commenting on whether we - the SWBPWG - should, or be in a position to,
provide comment on the above issues I have made specific reference to our
charter and consider that there may well be some slight conflict involved.
Specifically we are tasked to 'to provide hands-on support for developers'
which implies advice on implementation issues - a level of agreement not
yet reached by the DAWG. Nevertheless the charter also states that we may
well remark on 'engineering guidelines' by applying 'combining experience’.
For this reason, and given the level of confusion currently apparent around
the choice of implementation route for Semantic Web Data Access, one might
suggest that our remit to comment using 'combined experience' should
prevail. For this reason there may well be some generalized, yet well
established, Best Practice concepts that apply here and on which we might
all agree. Hence, given the relative immaturity of Semantic Web Data Access
mechanisms (languages) perhaps we should stick to safe ground and table the
following 'open' responses to the DAWG?

Tabled Response 1 - On methods for exploiting metadata on the Semantic Web
Although the Semantic Web has been designed to address a specific set of
requirements around the storage and use of metadata, it must still be
remembered that metadata is still only a specialization on the generic data
theme. As such, Semantic Web metadata may well be used for a wide spectrum
of uses in the future, some of which may not have yet been envisioned yet
alone realized. To narrow this potential range by recommending specific
closed implementation standards around data exploitation (querying, rules
etc.) must, hence, be viewed as contradictory to the objectives of the
Semantic Web initiative. As such, the development of an abstract canonical
syntax, as currently advocated by [1], on top of which several concrete
syntaxes for Semantic Web metadata exploitation could be implemented is of
obvious merit. In establishing such a model a primary aim should be the
extensible accommodation of, translation between and possible combining of
valid concrete syntaxes (both present and future) around core data
embodiment and constraint concepts. This will then move the onus of Working
Group responsibility away from implementation specifics towards
guardianship of data embodiment, rules application and mediation between
implementation mechanisms etc.

Tabled Response 2 - On maximization of technology reuse
Reuse is a recognised and fundamental Best Practice concept that should be
promoted wherever possible. Nevertheless reuse should always be tempered
with a view towards relevance to targeted core concepts, the specific use
advantages offered and potential for future extensibility and further
reuse.

If significant overlap exists with target canonical representations,
significant advantage is to be gained or non-overlapping features can be
implemented without significant effort or investment, reuse should always
be the chosen route forward. By recommending reuse, however, this should
not imply that implementations that mature first are any better, or should
dominate over, those the take longer to reach mass take up. Nor should it
imply that standards or commercial unification around a specific
implementation is correct or desired.

Tabled Response 3 - On making trade-offs in accepting query requirements
that are practical and binding abstract syntax to a concrete syntax– as
outlined in [1]

Although it is recognised that a pragmatic approach to implementation is a
commendable goal, it is apparent that a number of potentially orthogonal
approaches to concrete syntax currently exist and the likelihood is that
this number will grow in the future. Attempting to compensate for such
misalignments directly must surely be an overwhelming and torturous
endeavour best suited to either commercial competition of organic
acceptance over time through routes like open source. Whether the DAWG
should be seen to side with particular concrete syntax at this moment is
hence debatable from a Best Practices perspective.

[1] http://www.w3.org/2003/12/swa/dawg-charter#concreteSyntax


*********************************************************************************************************

Gary NG’s comments on the above were:

a. On whether we should respond to the DAWG - Yes. I agree on this
approach. From the first paragraph of our charter: "consensus-based
guidance ... to facilitate Semantic Web deployment" would seem to be the
choice of capacity in which the SWBPDWG shall respond. My view is that
giving guidance on query design/scoping is in our scope towards indirectly
facilitating Semantic Web deployment in the long run.

b. On the use of closed implementation standards - From other conversations
I had, I thought this is already DAWG's philosophy. However, I cannot find
similar wording in their charter. It could be useful for them if BP WG also
agrees this is the way forward. In any case it is a kind of endorsement and
vote of confidence from us.

However, the above seem a little abstract and open to interpretation. It
may or may not imply some of their "out-of-scope" items (e.g. OWL
semantics, Rules) should be brought back in for consideration, wherever
such item is deemed relevant to the openness/extensible-ness of the
eventual query language.

c. On maximising technology reuse - Interesting choice of words. I think it
is interesting and important to set the 'mood' towards openness and
encourage 'let the best "implementation" wins' kind of thinking. By the
above thus far you are suggesting that the DAWG group shall concentrate
their effort on devising a sound, well justified and extensible abstract
model, be mindful of other overlapping possibilities, provide guidelines on
implementation and translation, and remove themselves from concrete model
implementation details. Leaving the public to implement their own concrete
syntaxes based on the abstract. Eventually, the best will evolve and be the
de facto standard. Am I correct?

Gary further commented on my thoughts as follows….All these seem to address
only point 1.  XQuery, syntax and integration. And perhaps part of 3. data
source identification. You got any views on 2.?

Regarding 2), Jeremy Caroll had a few comments slightly overlapping my
own in that message [2].

[2] http://lists.w3.org/Archives/Public/public-swbp-wg/2004Sep/0097.html


**********************************************************************************************************

My reply to Gary’s comments was:

I appreciate your input; you have filled in a number of gaps nicely. I
think there is a consensus forming - I'm pleased.

As for comment on point [2], I'm not sure I'm really qualified to speak.
Jeremy's mail appears to cover the required ground more than adequately.
Hence I have copied Jeremy on this mail and, if there are no further
comments, I propose to submit the history of our conversation to the BP
Working Group as the basis for our response to the DAWG.


Regards


Phil Tetlow
Senior Consultant
IBM Business Consulting Services
Mobile. (+44) 7740 923328

Received on Friday, 29 October 2004 14:35:38 UTC