RE: RL/Full testsuite uploaded

Hi Mike!

>-----Original Message-----
>From: Mike Smith [mailto:msmith@clarkparsia.com]
>Sent: Tuesday, July 14, 2009 7:12 PM
>To: Michael Schneider
>Cc: Ian Horrocks; W3C OWL Working Group; mak@aifb.uni-karlsruhe.de
>Subject: Re: RL/Full testsuite uploaded
>
>On Tue, Jul 14, 2009 at 02:45, Michael Schneider<schneid@fzi.de> wrote:
>
>>>I tried to find these tests using [1], but it wasn't easy. Probably
>>>my fault. I tried browsing by species "OWL Full" and by profile "OWL
>>>RL", but that didn't work.
>>
>> Browsing for the "RL" flag won't give you the desired result, since my
>> tests generally do not satisfy the syntactic restrictions of the
>> OWL 2 RL profile, and therefore haven't been marked by that flag
>> (as settle in the past).
>>
>>>I then tried looking at proposed tests,
>>
>> My tests aren't marked as "proposed" yet. They are all "New Test"s.
>
>I've updated that status, all of these tests are now Proposed.

Thank you!

>I also updated the syntactic information.  All of these tests were not
>DL (and hence not EL,QL, or RL), at least because they lacked an
>ontology header.  

Yes, that's intended. 

Ontology headers would be redundant for testing RL and Full. They could even
break a positive entailment test, in particular for RL: the RL rules would
fail on different ontology URIs or different bNodes on both sides of the
test, or when there is an ontology URI on the LHS but a bNode on the RHS. 

In the latter two cases, a complete OWL 2 Full reasoner would be required to
succeed, but it would be non-trivial for it. And if the implementation fails
(due to an incomplete implementation of bNode semantics, for example), one
would get a very misleading testing result ("the reasoner fails on every
test"). 

In any case, testing with redundant information is not a good idea in
general, since an implementation might succeed only when having this
information, while it would fail without it. Since it would in fact be
required to succeed without the redundant information, one would then have
missed an actual failure (false positive). That's why I decided, in general,
to create pretty "stripped" tests that try to be as close as possible to the
RL rules (and the semantic conditions for the RDF-Based Semantics).
 
>Many of these test would be DL if ontology headers
>were added, but because it appears that you designed them unit tests
>for the RL rules, I did not add any ontology headers.

Yes, thanks! 

>--
>Mike Smith
>
>Clark & Parsia

Cheers,
Michael

--
Dipl.-Inform. Michael Schneider
Research Scientist, Dept. Information Process Engineering (IPE)
Tel  : +49-721-9654-726
Fax  : +49-721-9654-727
Email: michael.schneider@fzi.de
WWW  : http://www.fzi.de/michael.schneider
=======================================================================
FZI Forschungszentrum Informatik an der Universität Karlsruhe
Haid-und-Neu-Str. 10-14, D-76131 Karlsruhe
Tel.: +49-721-9654-0, Fax: +49-721-9654-959
Stiftung des bürgerlichen Rechts, Az 14-0563.1, RP Karlsruhe
Vorstand: Prof. Dr.-Ing. Rüdiger Dillmann, Dipl. Wi.-Ing. Michael Flor,
Prof. Dr. Dr. h.c. Wolffried Stucky, Prof. Dr. Rudi Studer
Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus
=======================================================================