LC Comment on Entailment Regimes: Illegal Handling, specifically OWL 2 Direct Regime

Dear all!

Document: SPARQL 1.1 Entailment Regimes
State: LCWD

The definition tables for the entailment regimes have an entry "Illegal 
Handling". It is generally defined that a system must signal an error, 
if either the queried graph or the query is not a syntactically valid 
RDF graph or BGP. In the case of the "OWL 2 Direct Semantics Entailment 
Regime" (Chap. 6), there is additional treatment for the special case 
that the graph or the query is not an OWL DL ontology:

     If the queried ontology is not an OWL 2 DL
     ontology or the query is not legal for the
     ontology, the system MAY refuse the query
     and raise a QueryRequestRefused error or the
     system MAY use only a subset of the triples
     in the ontology or query.

In general, I would be very cautious about specifying required ("MUST") 
behavior of a system in the case of syntactically invalid input. This 
may be too much of a requirement for an implementer to provide a 
compliant system, and might significantly affect the performance of a 
system due to additional checks. So, I suggest to reconsider your 
decision of how to handle syntactically invalid input: maybe a "SHOULD" 
is sufficient, or maybe nothing at all should be said. The latter would 
have the advantage that future extensions, such as support for 
generalized RDF (e.g. bnodes in predicate position) or named graphs 
beyond what SPARQL allows today would be possible without breaking 
compliance to SPARQL 1.1.

In any case, I strongly propose to /not/ make any concrete suggestions 
of the form: "on invalid OWL DL input, a system may only use a subset of 
the triples in the query". Not even with a "MAY". This makes a wrong 
behavior with unexpected results on wrong input into a kind of "best 
practice recommendation" to implementors and leads to certain 
expectations of users. If a system works outside its specification, 
there should be either a strong requirement to not process the input (as 
currently for illegal RDF data), or nothing at all should be said, so 
the system behavior is strictly implementation dependent. It is then up 
to the implementor to do what he considers best (and potentially to 
describe the behavior in the system's manual). Again, system 
implementers may decide to support more than plain OWL 2 DL (e.g. by 
relaxing on the global syntactic constraints to retain decidability), 
and there should not be a general expectation that the system will use 
only parts of the input query in such cases.

Best regards,
Michael

-- 
Dipl.-Inform. Michael Schneider
Research Scientist, 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
Stiftung Az: 14-0563.1 Regierungspräsidium Karlsruhe
Vorstand: Dipl. Wi.-Ing. Michael Flor, Prof. Dr. rer. nat. Ralf Reussner,
Prof. Dr. rer. nat. Dr. h.c. Wolffried Stucky, Prof. Dr. rer. nat. Rudi 
Studer
Vorsitzender des Kuratoriums: Ministerialdirigent Günther Leßnerkraus
==============================================================================

Received on Tuesday, 26 July 2011 12:35:06 UTC