RE: Requirement 3.1 rewording

>     The query language must include the capability to restrict matches
>     on the queried graph by providing two or more RDF triple patterns,
>     with at least one shared variable, to be satisfied in one query.

Not so good for me. 

1/ The variables may be implicit due to syntactic sugar and there are
different kinds of variables (not returned, may bind, ...)  I would not talk
about variables.

example (made up syntax): for a graph pattern:

(?x vCard:N-->vCard:familyName "Smith")

is the triple patterns 

(?x vCard:N  ?y)
(?y vCard:familyName "Smith")

and ?y is not a projection (SELECT) variable.

2/ not all the triple patterns may be connected by variables (this may be
bad design for other reasons nothing to do with this requriment).

How about:

  The query language must include the capability to restrict matches
  by graph pattern on the queried graph by providing a graph pattern.
  The graph pattern may consist of one, two or more RDF triple patterns,
  to be satisfied in one query.

That is, introduce graph pattern explicitly and say graph patterns are made
of triple patterns.

Note: graphs need not be connected so graph patterns need not be connected
and one pattern per query is enough for conjunctive queries.


I preferred the original but will put up with the reworded, "variable"
version.  These are requirements and we do not understand their full
implications yet.

	Andy


-------- Original Message --------
> From: Thompson, Bryan B. <mailto:BRYAN.B.THOMPSON@saic.com>
> Date: 5 May 2004 15:13
> 
> +1 on "queried graph".
> 
> And, yes, I think that this language is an improvement -- as long as
> it is reflecting the intention.  It at least removes one ambiguity
> concerning the meaning of the requirement.
> 
> -bryan
> 
> -----Original Message-----
> From: public-rdf-dawg-request@w3.org
> To: Thompson, Bryan B.
> Cc: 'Seaborne, Andy '; 'public-rdf-dawg@w3.org '
> Sent: 5/5/2004 9:52 AM
> Subject: Re: Requirement 3.1 rewording
> 
> 
> On Wed, May 05, 2004 at 09:38:20AM -0400, Thompson, Bryan B. wrote:
> > What seems to be missing is the notion that there are shared bindings
> > among the triples in the pattern.  Without that, there is no notion of
> > a series of connected edges in the graph.
> 
> Bryan,
> 
> Does this get at what you're suggesting?
> 
>     The query language must include the capability to restrict matches
>     on the queried graph by providing two or more RDF triple patterns,
>     with at least one shared variable, to be satisfied in one query.
> 
> Andy: does this seem an improvement to you?
> 
> (FWIW, I'm trying to find a short term to mean the "dataset that the
> query is executed against". I tried "query graph" in earlier drafts,
> but that's ambiguous. Someone said "queried graph" during the telcon
> yesterday, which I think is an improvement. We will evolve some
> specific language in this WG and in our documents, so comments on
> "queried graph" are welcome, but probably best directed to me (?).)
> 
> Kendall

Received on Wednesday, 5 May 2004 10:47:14 UTC