W3C home > Mailing lists > Public > public-rdf-dawg@w3.org > April to June 2005

Minutes of RDF Data Access WG telcon 2005-04-19 for review

From: Jeen Broekstra <jeen@aduna.biz>
Date: Wed, 20 Apr 2005 12:13:47 +0200
Message-ID: <42662B5B.4080307@aduna.biz>
To: public-rdf-dawg@w3.org

Minutes of RDF Data Access WG telcon 2005-04-19

RDF Data Access
http://www.w3.org/2001/sw/DataAccess/

Present:
Dave Beckett
Jeen Broekstra (scribe),
Dan Connolly (chair)
Steve Harris,
Pat Hayes,
Howard Katz,
Eric Prud'hommeaux,
Alberto Reggiori,
Andy Seaborne,
Janne Saarela,
Kevin Wilkinson

Regrets: Hiroyuki Sato, Kendall Clark, Jos De Roo

Agenda
http://lists.w3.org/Archives/Public/public-rdf-dawg/2005AprJun/0122.html

1. Convene, take roll, review records

Minutes of 12 April 2005 telcon
http://lists.w3.org/Archives/Public/public-rdf-dawg/2005AprJun/0095.html
http://lists.w3.org/Archives/Public/public-rdf-dawg/2005AprJun/0101.html
approved

next meeting: 26 Apr. ericP to scribe

regrets for 26 Apr meeting: Jos De Roo

open actions:

ACTION: KendallC to draft protocol section on specifying datasets
(fromUnionQuery) including WSDL types and HTTP examples
ACTION: EricP to try out the "2 list of URIs" datasets design from
http://lists.w3.org/Archives/Public/public-rdf-dawg/2005JanMar/0440.html
ACTION: KendallC to own wsdlAbstractProtocol (based on WSDL 
description added to protocol to editor's draft).
ACTION: KendallC to work out remaining details in sparqlx.{rnc,xsd,rng}:
expression syntax, order-by stuff (xmlAbstractSyntax)
ACTION: AndyS to define a term for abstract form of SPARQL query
ACTION: KendallC to incorporate EricP's privacy text, salting to taste
ACTION: DaveB to work with EricP to clarify valueTesting proposal
ACTION: JeenB to Write a few sort test cases.
ACTION: PatH to internal review of rq23
ACTION: EricP to pair with SteveH on making the HTML test results page
ACTION: DaveB to to propose source test to approve
ACTION: SteveH to prepare test cases for publication as WG Note (no 
deadline/urgency)
ACTION: AndyS to add the above graph test cases (analagous to valueTesting
         test cases) (don't expect quick delivery)
ACTION: DanC to follow up re optional test based on op:dateTime triple
ACTION: EricP to propose a test re "x"^^mytype != "y"^^mytpe

1. "news from SparqlingDays"

http://rdf4food.asemantics.org/SparqlingDays

DanC does quick round to see who's been to the SparqlingDays event and 
what has been achieved there, also invites other general remarks.

AndyS reports having a student looking at a mapping between SPARQL and
SQL.

DaveB was at Tuscany meeting, reports working on implementation of 
SPARQL query and protocol demo to perform queries over RDF and allow 
result transformation with XSLT. See http://librdf.org/2005/sparqling. 
DaveB notes that he does not think LIMIT and OFFSET belong in the 
protocol. SteveH agrees.

EricP was at Tuscany meeting, created an Algae CGI tool for RDF query. 
See http://www.w3.org/2004/10/18-RDF-WWW2005/
(http://www.w3.org/2004/10/18-RDF-WWW2005/algae for the actual script).

AlbertoR explains that the idea of the Sparqling Days meeting was a
get-together to try and set up examples and demos. The Gargonza
experiment (http://www.gargonza.org/) is an initiative to bring 
continuity to that idea. Interested people are welcome to join.

AndyS asks clarification from HowardK regarding XQuery comparisons, in
particular why there is a need for a separate operator for xs:integer 
when it can be mapped (through subtype substitution) to xs:decimal.

HowardK briefly clarifies his mail, noting that he is not sure why
xs:integer is singled out, unless for efficiency issues.

Several people note that a result of the SparqlingDays was a logo for
SPARQL, to be found at http://www.gargonza.org/images/sparql48.png. 
Logocreator is Zavisa Bjelogrlic (foaf profile:
 
http://xml.mfd-consult.dk/foaf/explorer/?foaf=http://foaf.asemantics.com/zac).

2. issue: fromUnionQuery

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

The FROM/FROM NAMED stuff only has one URI for the
background graph, no merging allowed. Requests feedback from the WG.

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

DaveB remarks that it is scary to readers to put WSDL 2.0 semantics 
right at the start of the protocol document. Alberto says HTTP 
bindings should go first eventually.

ACTION EricP: try out the "2 list of URIs" datasets design from
http://lists.w3.org/Archives/Public/public-rdf-dawg/2005JanMar/0440.html
CONTINUED

The question is raised what the reason is for not allowing a list of 
URIs for the background graph. AndyS explains that allowing this will 
raise user expectations of functionality like merging, etc. AndyS 
notes that he has implemented 2 lists: the foreground is a merge, the 
background is one name per graph. "Could have FROM NAMED <u1> <u2> AS 
<u3>". EricP notes that if we do named graph, he prefers this proposal 
by AndyS.

3. "issue: wsdlAbstractProtocol"

ACTION: KendallC to own wsdlAbstractProtocol (based on WSDL 
description added to protocol to editor's draft).
CONTINUED

4. issue: xmlAbstractSyntax

ACTION: KendallC to work out remaining details in 
sparqlx.{rnc,xsd,rng}: expression syntax, order-by stuff 
(xmlAbstractSyntax)
CONTINUED

DanC asks what the defined term for the abstract form of a SPARQL 
query. AndyS notes that there is no single term for this. DaveB points 
out that in the RNC (http://www.w3.org/2001/sw/DataAccess/sparqlx.rnc) 
the recursive term is 'query-pattern', but that in the query WD, he 
would have expected it to be 'graph-pattern'.

ACTION AndyS: define a term for abstract form of SPARQL query

5. protocol publication candidate

ACTION: KendallC to incorporate EricP's privacy text, salting to taste
CONTINUED

DanC solicits reviewers for upcoming publication. AndyS will read, but 
can not commit to review.

6. issue: valueTesting

ACTION: DaveB to work with EricP to clarify valueTesting proposal

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

HowardK made progress, but is not sure when to claim victory. Requests
feedback from AndyS and ericP.

HowardK's report at
http://lists.w3.org/Archives/Public/public-rdf-dawg/2005AprJun/0124.html
is discussed.

DanC wants to encourage checking of links from SPARQL spec to Xquery 
spec regarding value testing, a la
http://lists.w3.org/Archives/Public/public-rdf-dawg/2005AprJun/0106.html

DanC wants to know how close to done this issue is.

ericP remarks that some fuzzy issues are left, for example whether 
plain rdf literals match (or not) because are the same term or because 
the have the same (string) value. Current idea is to use a fallback 
setup where first RDF term matching is performed and if that fails for 
some reason, string comparison is performed. DaveB remarks that he 
does not quite follow the significance of this distinction and 
requests a test. DanC wants to know what would trigger the fallback: 
when the first test returns false or returns an error? EricP clarifies 
that it is the latter.

AndyS requests that a table with example value comparisons and their
evaluation outcome is put into the document, as clarification.

Danc asks what happens if "1.0"^^xsd:integer is compared to
"1"^^xsd:integer, or more in general, whether the current proposal
addresses users expectations. DaveB remarks that although the terms in
DanC's example are different, the comparison would result in equal.

EricP supplies a table:

bindings                    same    type-specific
?x          ?y              node    equivalence
-------------------------------------------------------------
<foo>       <foo>           1       error
_:a         _:a             1       error
"123"       "123"           1       error
"123"^^int  "123"^^int      1       1 =(int, int)
"123"^^int  "123"^^float    0       1 =(float, float)
"11:01"^^dt "11:01"^^dt     1       1 =(dateTime, dateTime)

DanC asks JanneS if any of this is a problem in his implementation. 
JanneS responds that they have not considered type value testing in 
detail yet.

AndyS asks clarification on the last column (specifically the errors) 
of ericP's table; SteveH explains that it is reasonable that a typed
comparison on untyped values would result in an error.

EricP Algae implementation to be found at : 
http://dev.w3.org/cvsweb/perl/modules/W3C/Rdf/Atoms.pm?rev=1.88&content-type=text/x-cvsweb-markup&only_with_tag=HEAD

DanC notes that in CWM the background graph values are by default
normalized and the 'orginal' value is forgotten.

DaveB notes that he is not (yet) coding this part of the design.

SteveH notes he has implemented part of it but not all.

7. issue: sort

JeenB has checked in a few simple test cases for ORDER BY. The first 
two are simple alphabetic sortings on untyped literal values. The 
third is more tricky: it sorts on a variable that is optional, and 
therefor not always has a value. This results in a partial ordering, 
which is a problem to write down in the current result set formats (see
 
http://lists.w3.org/Archives/Public/public-rdf-dawg/2005AprJun/0126.html).

DanC remarks that the spec does not specify what to do in this case, 
so to not have this as a testcase. JeenB will add additional test 
cases for sorting.

ACTION JeenB: Write a few sort test cases.
CONTINUED

DanC proposes that
http://www.w3.org/2001/sw/DataAccess/rq23/#specifyingDataset v 1.307
addresses the sort issue

DaveB remarks that he is unhappy with the wording and that many issues
have come with LIMIT and OFFSET etc. Specifically the phrase "solution
sequence" is unclear. Also request that the bulleted list in 10.1 
should be an ordered list (in processing order).

SteveH remarks that sorting is a feature he really wants to be in 
SPARQL, but that he doesn't have enough implementation experience on 
sorting to be confident on the current text.

DaveB remarks that from his users he mostly got requests for DISTINCT 
and LIMIT.

JeenB remarks that in Sesame's SeRQL language, a lot of users have 
asked for sorting.

AlbertoR remarks that some form of ordering/sorting is important and
needed today, and would rather spend more time implementing a 
(potentially difficult) part of the spec than not have sorting in 
there at all.

Strawpoll: who thinks postponing sort is a bad idea?
bad idea: AndyS, AlbertoR, JeenB
neutral: DaveB, SteveH

8. SPARQL QL publication candidate

DanC remarks that the next publication candidate will be ready this
afternoon.

AndyS notes that he still has some comments from Bijan that he has not
processed yet.

ACTION: PatH to internal review of rq23
CONTINUED

9. test suite maintenance (time permitting)

ACTION: EricP to pair with SteveH on making the HTML test results page
ACTION: DaveB to to propose source test to approve
ACTION: SteveH to prepare test cases for publication as WG Note (no 
deadline/urgency)

ACTION: SteveH to to revise test manifest w.r.t. "background" and 
named graphs
DONE

ACTION: AndyS to add the above graph test cases (analagous to 
valueTesting test cases) (don't expect quick delivery)

ACTION: DanC to follow up re optional test based on op:dateTime triple
CONTINUED

ACTION: EricP to propose a test re "x"^^mytype != "y"^^mytpe

Meeting ended


-- 
Jeen Broekstra          Aduna BV
Knowledge Engineer      Julianaplein 14b, 3817 CS Amersfoort
http://aduna.biz        The Netherlands
tel. +31 33 46599877
Received on Wednesday, 20 April 2005 10:11:50 GMT

This archive was generated by hypermail 2.3.1 : Tuesday, 26 March 2013 16:15:23 GMT