major technical comment: identifier length

2006-01-12T02:19:30Z Fred Zemke

Summary: Since arbitrary sized elements (such as, URIs, Literals, and blank node labels) need to be handled anyway, some members in DAWG wondered whether we need to do a special case for identifier-length just for variables.

major technical: semantics are poorly specified

2006-01-12T21:34:41Z Fred Zemke

Latest version includes many re-writings and clean-ups. The DAWG members are reluctant to take up additional changes in this regard at this time.

major technical: runaway queries

2006-01-12T21:37:34Z Fred Zemke

Summary: The solution outlined in the new version of the document is along the line of the provisional solution you suggested (that is, limiting the result to be made up of known RDF terms). Also, the point about SELECT clause allowing expression seems important. I need to find out as to why we are not allowing that.

major technical: underspecified errors

2006-01-12T21:41:10Z Fred Zemke

(I saw an ACTION ITEM for Dan Connolly to respond to this comment. In that case the following is unnecessary.)

I have been told by other members of DAWG that it is illegal to name the same dataset twice in FROM or FROM NAMED clauses of a query. I’d think that protocol service behaves similarly in case the same dataset appears multiple times in the RDF dataset specified in the invocation.

 

If an IRI does not identify a resource represented by a graph then I’d expect query request would be refused using QueryRequestRefused. The protocol document does not individually identify these errors, however.

minor technical on 5.4 Optional matching - formal definition

2006-01-12T17:31:59Z Fred Zemke

Summary:

The definition you proposed is very clean. There is probably a minor problem in the boundary case. I think that it needs to exclude the pathological case of VB = empty. That change will correctly handle a case like { ?x :hasBoss :Mary . OPTIONAL { ?x :plays :Soccer } } where VA = {?x} and VB = empty. This is a pathological case because the OPTIONAL part in this case example does not contribute anything to the result and therefore such a graph pattern, although valid, would yield same result as just {?x :hasBoss :Mary}

 

The definition in the latest version of the document works too if it says something like: “… S is a solution … if S is a pattern solution for the graph pattern same as Opt(A, B) but with B considered compulsory (i.e., not OPTIONAL), otherwise if S is a pattern solution for the graph pattern consisting of A only.”

 

I am assuming that the following assumption is true: A pattern solution for a graph pattern is a set of bindings for only those variables that are in the graph pattern and for no other variables. This assumption is needed to ensure that a pattern solution for A does not include a binding for a variable contained only in B (i.e., not in A, so in VB set using your notation). If this assumption does not hold then a pattern solution for A only may contain misleading bindings for variables in VB. Consider the following example: { ?x :hasBoss :Mary . OPTIONAL { ?x :hasFriend ?y } } Here, if we allow { ?x=:John, ?y=:David} to be a solution for A only, then that result may indicate that :John :hasFriend :David . is an assertion in the RDF dataset, which may or may not actually be the case.

minor technical on 9 Specifying RDF datasets

2006-01-12T17:33:12Z Fred Zemke

This may be okay the way it is in the document. This is more like use of default values for RDF datasets in the query that allows invocation-time overriding. This is somewhat similar to optional parameters with default values in PL/SQL procedures and functions. However, I think (and as Newman suggested in his reply e-mail) augmentation of RDF datasets should be considered too.

minor technical on 10.3 Constructing an output graph

2006-01-12T17:35:36Z Fred Zemke

The text “ and a warning may be generated” has since been removed in the new document version.

minor technical on 11.2.2 Effective boolean value

2006-01-12T17:38:09Z Fred Zemke

This behavior is aligned with how this is done in XQuery and XPath. So, they are reluctant to change.

minor technical: A.7 grammar, rule [25] Constraint

2006-01-12T17:44:37Z Fred Zemke

According to some members of the DAWG group this is a minor issue and they do not think a change is needed.

minor technical: omissions from Appendix B Conformance

2006-01-12T17:45:22Z Fred Zemke

The DAWG group does not want to put additional requirements for conformance regarding error behaviors. Regarding extension, the current thought is that anything more than SPARQL is not SPARQL.