on OWL test cases: good stuff, one CRITICAL change request

regarding...

"30 Jan (Thu)
        test reviews due (ACTION Connolly, Stanton)"
 -- http://www.w3.org/2001/sw/WebOnt/ftf5.html#Schedule

I spent some time looking over...

http://www.w3.org/2002/03owlt/editors-draft/snapshot
Tue, 28 Jan 2003 16:41:58 GMT

Only one CRITICAL issue, re conflict resolution stuff;
see below.

Otherwise, I didn't find any problems; I have
some editorial suggestions, but nothing major.

I like the idea of publishing it ASAP, without
"last call" status. It represents significant
progress since our Oct publication.


Detailed comments...


| [[EDITORS' NOTE: Do we need index files for the web site that clarify
which parts are part of the deliverable and which are not?]]

No, but we need a link to the zip file.
It goes with the "this version" link, I think...
yes...
http://www.w3.org/2001/06/manual/#formats

http://www.w3.org/2002/03owlt/editors-draft/snapshot#conformance

| 2.2. Conflict Resolution

The stuff about "If the OWL recommendation has passed Proposed
Recommendation then..." is out of scope for this WG.
This document can't sign W3C up to do work after we're gone,
and can't override errata processes otherwise established
by W3C. Please strike. CRITICAL.


| 3.2. Entailment Tests

Clarify that there's more than one entialment relationship
to test for. Something like...

  The conclusions are related to the premsises by one
  of three entailment relationships:
  -- OWL Full entailment (cf section 5.4. OWL Full of [AS&S])
   for those tests tagged "Full"
  ... similarly for Lite/DL

The same comment applies to the following sections,
up to 3.8. Import Entailment Tests,
though it may suffice to clarify just the one section.

The conformance stuff looks correct, though I'd like
to see a bit more examples/verbiage.

In particular...

| A complete OWL Lite consistency checker is an OWL consistency checker
that takes an OWL Lite document as input, and is logically complete with
respect to [OWL Abstract Syntax and Semantics] over the set of all OWL
Lite documents.


'logically complete' isn't clear enough in that context.
We're talking to hackers here, not logicians.
Say something about 'always returns yes or no'.

Also, consider separating 'resource exhausted' from "don't
know" in the case of complete consistency checkers.
And perhaps

	Complete consistency checkers SHOULD
	have resource limits clearly documented,
	and SHOULD offer parameters about how
	much resource to consume before giving up.

I'd really like to see the test descriptions refer
to specific parts of AS&S, but I didn't get around to
specific suggestions, so I'll understand if you don't
either.

I'd also like to run thru some of the tests
in the "C proposed" section, but I didn't get
to it yet. Perhaps tomorrow.

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/

Received on Wednesday, 29 January 2003 19:10:29 UTC