Minutes of RDF Data Access WG telcon 2005-04-12 for review

Minutes of RDF Data Access WG telcon 2005-04-12

RDF Data Access

Dave Beckett (scribe),
Jeen Broekstra,
Dan Connolly (chair)
Kendall Clark,
Steve Harris,
Pat Hayes,
Eric Prud'hommeaux,
Alberto Reggiori,
Hiroyuki Sato,
Andy Seaborne,
Kevin Wilkinson

Regrets: Bryan Thompson, Howard Katz


1. Convene, take roll, review records

Minutes of 5 April 2005 telcon

next meeting: 19 Apr. JeenB to scribe

ACTION: KendallC to work out remaining details in sparqlx.{rnc,xsd,rng}:
expression syntax, order-by stuff
ACTION: KendallC to incorporate EricP's privacy text, salting to taste
ACTION: KendallC to own wsdlAbstractProtocol (based on WSDL description
added to protocol to editor's draft).
ACTION: DaveB to consider use of xsi:dataType ala comment from Steer

all actions continued

2. issue: fromUnionQuery

Discussed Re: fromUnionQuery Dave Beckett (Monday, 11 April):
and background graph, it's use in protocol, keywords for them in the
query language (last WD) such as LOAD, FROM, WITH.

Discussed a proposal from SteveH of a design without a distinguished
background graph in:
which was to instead of having a single background graph, with no
name, to specify a set of graphs (a set of URIs with "use") which
together answer matches with no GRAPH keyword.  When a GRAPH keyword
is used, the "constrain" set of URIs (the named graphs) provide
matches, and bindings for the GRAPH ?var.  This makes the background
graph explicit.

The current protocol WD doesn't mention the background graph because
as it was understood from the query WD, it is implicit and unnamed.
The dataset in the query language may cover things the protocol
cannot specify.

SteveH noted that he was having problems with the un-named aspect of
the background graph which was one reason for his proposal.  The use
case is data management of the background graph.

ACTION SteveH: write up the use case that can't be handled by the
current design - data management of background graph

Discussion of the terminology of background graph, target graph,
default graph.  The current design of the dataset is:
  { G, (u1, G1), (u2, G2), . . . (un, Gn) }
where the first G has no URI.

Considering the two types of datasets, considered data/background,
trusted/untrusted (but objections to thas a overloaded),
named/unnamed (objections to it being emphasising naming),
named/default, foreground/background, others.

ACTION KendallC: draft protocol section on specifying datasets
(fromUnionQuery) including WSDL types and HTTP examples

ACTION EricP: try out the "2 list of URIs" datasets design from

DaveB asked about forming the dataset in the query language using
keywords, which was removed after a decision at F2F5 in Boston,
refering to use cases in

DaveB said passing along (data to query, query to operate) together in a
document is very handy - should be available in rdf query as it is in
xquery and xslt with document(). 

Discussion of the reason why it was removed at F2F5, and issues with
having the dataset specification in query and protocol.  Some support
for having it in the query language, could be a forward compatibility
point for any later work on subqueries, passing on query results to
other queries in the same query document.

If the dataset specification was in both query and protocol, then the
case of what happens if both are present turns up - conflicts, which
takes precendence, or an error given.

Took straw polls saw support for allowing dataset in both query and
protocol.  On how to resolve conflicts in dataset specifications, the
choices seen were:
  * error if both are present
  * QL overrides protocol
  * protocol overrides QL
  * error if they disagree
Some people had preferences for answers, several didn't care.

ACTION AndyS: add syntax to specify datasets in the query language

3. issue: valueTesting

ACTION: DaveB to work with EricP to clarify valueTesting proposal ETA 19

ACTION: HowardK to seek clarification on valueTesting w.r.t. differenced
from XQuery

Discussed Klyne' Comments on SPARQL 17-Feb-2005 draft

4. issue: punctuationSyntax

ACTION: DanC to propose to close punctuationSyntax (again) sans <<reifiction shorthand>>

ACTION: [PENDING] DaveB to consider dots in qnames, report on impact
on turtle

PROPOSED: to not have reification syntax
objections: none
abstain: KendallC, AndyS/KevinW (HP), PatH, JeenB
in favor: DanC/EricP (W3C), HiroyukiS, SteveH, DaveB, AlbertoR

ACTION AndyD: take out the reification syntax

5. issue sort

ACTION: PatH to review sort design

PatH: reviewed it and it's complicated, but it would have to be.
Needs a total ordering.  More concerned about what the motiviation
for sorting was.  AndyS said it had a total ordering in response to
earlier PatH feedback.  Discussion if a total ordering was needed,
and it may not be but it could be a consistent ordering.

ACTION JeenB: Write a few sort test cases.
Of the order of about 10 - 6 normal cases and 4 weird things users
want to do.

JeenB asked about ordering in the XML results format.  DaveB says it
matches the query language's result set, so at present is unordered.

The test result format would also need changing to add an extra
property to record the result index in the ordering, if sorting is

6. SPARQL QL publication candidate

ACTION: DaveB to review rq23 editor's draft this week
Design Review of SPARQL query editor's draft $Revision: 1.293 $ Dave
Beckett (Thursday, 7 April)

ACTION: PatH to internal review of rq23, starting monday

ACTION: AndyS to clarify 5.4 w/r/t closed world assumption

other comments, e.g.
V. odd example Bijan Parsia (Monday, 4 April)

AndyS and EricP are still catching up on the comments after
travelling and vacations.

PROPOSED: to publish editor's best effort this week, publication date
early next week (week of Apr18)
objections: none
abstentions: SH

Meeting ended

Received on Thursday, 14 April 2005 10:29:55 UTC