[Fwd: Comments on SPARQL] (entailment, soundness, completeness)

Copying this to public-rdf-dawg comments for tracking purposes.
Follow-up discussion should happen there...

-- 
Dan Connolly, W3C http://www.w3.org/People/Connolly/
D3C2 887B 0F92 6005 C541  0875 0F91 96DE 6E52 C29E

attached mail follows:

Two comments/questions:

Firstly, I strongly support the suggestion to define query answers in 
terms of entailment rather than subgraph. Using an entailment based 
definition has numerous advantages, and no disadvantages that I can 
see.

Entailment based definition:

- is widely used and very well understood
- is concise, clear and unambiguous
- builds on existing semantic definitions
- can deal with a wide variety of languages, including RDF, RDFS, OWL, 
SWRL and FOL, simply by referring to the existing entailment semantics 
for the relevant language

Subgraph based definition:

- is not widely used, not well understood, and not *known* to work at 
all (at least not for more expressive languages)
- is verbose, obfuscated (it seems capable of confusing even expert 
logicians), and may even be ambiguous
- ignores existing semantic definitions
- cannot easily deal with more expressive languages, and may even be 
incapable of doing so

Can someone please explain to me why the subgraph based definition has 
been preferred?


Secondly, IMHO, the minimum requirement for a query language standard 
is that, independently of the features and capabilities of any given 
implementation, it should define what would constitute a sound and 
complete answer for a given language, dataset and query. It is not at 
all clear that, in its current form, SPARQL satisfies this requirement.

Am I missing something?

Ian