HP reactions to AS&S and OWL

I have been talking with immediate colleagues about where OWL is at, 
particularly the OWL Lite, OWL DL and OWL Full layering.

The current reactions are not positive.

I'll first discuss process, then sketch the substantive worries.

===

Process:

I have two main concerns:

- I do not believe that I can give an HP position on AS&S before the 
planned last call period. Two obstacles are (a) the document keeps changing 
  (more change is still needed, I do not believe we can freeze the 12th 
January copy) and (b) I believe I need two to four weeks to allow my 
colleagues to read it, and for an HP internal discussion.

- Given that I have early indications of problems, it is unclear how I 
should vote on moving AS&S to last call. Since my preference is that we 
would have another round of WDs before last call, I am clear that I will 
not vote FOR last call. My choice is between abstaining and voting against.
In the f2f IRC Jim is recorded as saying:
http://ilrt.org/discovery/chatlogs/webont/2003-01-10.html#T09-54-30#
[[
the WG should realize that before LC is the last time for WG members to 
make objections
]]
Do I need to vote against to indicate that there may be substantive 
objection from HP during the last call period?

===

Substantive:

The key problem being raised by my colleagues is that OWL Lite and OWL DL 
are both too difficult to implement. OWL Lite because the iff semantics 
more or less commits implementors to use tableau reasoners, which seems to 
prioritise one part of the relevant implementation community over and above 
others; OWL DL because no one is signed up to implement it.
Indicative of this is that only Network Inference signed up to support OWL 
Lite.

This is tricky because there are two very different possible fixes:
- change the defn of OWL Lite and OWL DL to make them easier to implement. 
Goal would be to get many more of the implementors saying they intend to 
support OWL Lite.
- set the CR exit gate high enough. e.g. more than one OWL Lite 
implementation, and at least one OWL Lite implementation integrated with a 
system essentially targeting OWL Full rather than OWL DL (e.g. Protege, SWI 
Prolog or Jena). This seems to be necessary to ensure that OWL Lite will 
serve to provide interoperability between radically different 
implementations of OWL.

I have been unable to defend the chairs' enthusiasm for skipping CR; when 
we have so many unimplemented features in OWL. I can easily imagine a 
scenario in which OWL rushes to PR, some non-WG member objects that we have 
insufficient implementation experience; my AC rep asks for my opinion and I 
have to say that the objection is well-founded.


Note - I am trying to give early indication of problems ahead; this message 
does not yet reflect an HP consensus opinion, more my understanding of the 
reactions of some of my closer colleagues.

Jeremy

Received on Tuesday, 14 January 2003 12:07:26 UTC