See also: IRC log
<scribe> ACTION: LeeF to note mf:requires of extension points when emailing out update to test suite [DONE] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action02]
<scribe> ACTION: LeeF to update approval information on all tests approved and re-approved today [DONE] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action03]
<AndyS> EricP : More on F&O (it gets worse!)
<AndyS> dt-equal is false for ANY non-timezoned DTs. because it say use teh XMl schema on DT timzones only. "Returns false otherwise."
<AndyS> And with the P/Q example all these are false at the same time: P<Q, P=Q and P>Q.
beautiful
<AndyS> It's a bit broken by exact spec text.
<AndyS> Could do with the history - XMLschema preadates F&O
Hi everyone - we're trying to dispatch of the timezone issue right now, will get on with the rest of the agenda momentarily
<chimezie> okay
<chimezie> I have a conflict at 11am and will have to leave (early) at that point
np
Minutes from last week: http://lists.w3.org/Archives/Public/public-rdf-dawg/2007JulSep/att-0118/04-dawg-minutes.html
PROPOSED: Approve minutes from last week http://lists.w3.org/Archives/Public/public-rdf-dawg/2007JulSep/att-0118/04-dawg-minutes.html
resolved
Next meeting 18 Sep
Regrets next week: chimezie
<scribe> ACTION: ericP to try to produce a diff from CR to current editor's draft [DONE] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action04]
SPARQL query language text changes ready for review by Orri and SteveH
<scribe> ACTION: ericP to have first cut of shiny implementation report by 11 Sep [CONTINUES] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action05]
What's the status of the test service?
<iv_an_ru> http://demo.openlinksw.com/sparql_endpoint_test/sws_main.vsp
LeeF: plan is to advertise service now, get folks (starting with Richard N) to try it out
<scribe> ACTION: Orri and Ivan to publish URL to test service along with a list of tests that aren't handled correctly yet [DONE] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action06]
<scribe> ACTION: Eric+Andy to identify and collect mf:requires URIs and put in documentation [CONTINUES] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action07]
<ericP> [[
<ericP> The EBV of any literal whose type is xsd:boolean, xsd:string or numeric is false if the lexical form is not valid for that datatype (e.g. "abc"^^xsd:integer).
<ericP> ]]
<scribe> ACTION: ericP to incorporate text to fix EBV of illformed numeric literals bug in spec [DONE] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action08]
<scribe> ACTION: ericP to answer "EBV of invalid numeric literals" email from Arjohn [CONTINUES] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action09]
<scribe> ACTION: ericP to poke IETF folks about registering SPARQL media types (esp. application/sparql-query) [CONTINUES] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action10]
<scribe> ACTION: EricP to publish xml format CR pending OK of the results schema [CONTINUES] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action11]
<ericP> [[
<scribe> ACTION: ericP to write explanatory text saying that, like xsd:inteter and xsd:dateTime, the relative order of simple literals and xsd:strings is not defined here [DONE] [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action12]
<ericP> SPARQL does not define a total ordering of all possible RDF terms. Here are a few examples of pairs of terms for which the relative order is undefined:
<ericP> * "a" and "a"@en_gb (a simple literal and a literal with a language tag)
<ericP> * "a"@en_gb and "b"@en_gb
<ericP> * "a" and "1"^^xsd:integer (a simple literal and a literal with a supported data type)
<ericP> * "1"^^my:integer and "2"^^my:integer (two unsupported data types)
<ericP> * "1"^^xsd:integer and "2"^^my:integer (a supported data type and an unsupported data type)
<ericP> ]]
<AndyS> case 2 needs a reason.
<AndyS> Earlier ref to simple literal is (1) wrong format and (2) first mention in the doc
<AndyS> Case 4 is not what is in rq25
<AndyS> EricP : you seem to have a different doc to rq25/cvs. Case 4 reason text is different. (Or caches are messed up)
<scribe> ACTION: LeeF to put together CR transition request for SPARQL XML format and EricP to publish [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action13]
<AndyS> Case 4 says -- "1"^^my:integer and "2"^^my:integer (a supported data type and an unsupported data type)"
<ericP> [[
<ericP> * "a" and "a"@en_gb (a simple literal and a literal with a language tag)
<ericP> * "a"@en_gb and "b"@en_gb (two literals with language tags)
<ericP> * "a" and "a"^^xsd:string (a simple literal and an xsd:string)
<ericP> * "a" and "1"^^xsd:integer (a simple literal and a literal with a supported data type)
<ericP> * "1"^^my:integer and "2"^^my:integer (two unsupported data types)
<ericP> * "1"^^xsd:integer and "2"^^my:integer (a supported data type and an unsupported data type)
<ericP> ]]
LeeF: With schema import issues resolved, decision is to prepare transition request and publish SPARQL Query Results CR ASAP
-> http://lists.w3.org/Archives/Public/public-rdf-dawg-comments/2007Sep/0004.html
Comment that shortcuts in Turtle for numeric datatypes are not defined to have a lexical form
:s :p 3.
:s :p 3.0 .
:z3 :p1 1.3e0 .
:z3 :p1 "1.3e0"^^xsd:decimal .
<iv_an_ru> But who cares what's the lexical representation at all?
<AndyS> It appears in the XML results format where value rules do not apply.
PROPOSED: Change all numeric shortcuts in the test suite Turtle data to use explicit lexical forms + datatype
ericP seconds
<iv_an_ru> OK, but after that we may provide a separate test that will demonstrate equalities between different forms.
<AndyS> Seems OK - Turtle does not permit canonucalisation of ""^^ form.
resolved
<scribe> ACTION: LeeF to update Turtle data to have explicit lexical forms and answer Gregory W's mail [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action14]
-> http://lists.w3.org/Archives/Public/public-rdf-dawg/2007JulSep/0128.html Andy's response re open-eq-12
Extension required for this test is to know about language tags, so that you know that plain literal with language tag is != to a simple literal
ericP: old mail with operator mapping table added rows to accomodate open world tests contains mf:requires
<ericP> http://www.w3.org/mid/20061028085031.GA4200@w3.org
PROPOSED: open-world/open-eq-12 requires an extension to the SPARQL operator mapping table for support for language tags and should be noted as such in the test manifest
ericP seconds
resolved
<scribe> ACTION: LeeF to update open-world manifest with mf:requires for open-eq-12 [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action15]
Nothing to look at yet - probably will be meat of next week's meeting
<AndyS> Beep - I have SOAP bindings
<iv_an_ru> Beep - we too
<AndyS> If you have a public access endpoint, I can use it.
<AndyS> I can run a test service for a few days.
<iv_an_ru> Same demo.openlinksw.com/sparql
<AndyS> I only do WSDL 1.1.
<iv_an_ru> I even don't know what 'WSDL' mean :)
<iv_an_ru> I'll cc to the developer
<AndyS> That's a SOAP endpoint? Cool - will try.
<iv_an_ru> It should recognize any sort of request.
<scribe> ACTION: AndyS to try SOAP protocol tests against Joseki and Virtuoso endpoints [recorded in http://www.w3.org/2007/09/11-dawg-minutes.html#action16]
<AndyS> I tried GET - it's not up for GET - is it POST only?
<iv_an_ru> (But I haven't tried it myself
<AndyS> Leave that to me :-)
LeeF: timezone issue needs a proposal for something to fix or something not to fix
ericP: need to know semantics of
CR, come up with proposal to clarify cases, and figure out if
its a clarification or a change
... LC had links to XSD datetime comparison functions in XPath
F&O
<AndyS> See http://www.w3.org/TR/timezone/ sec 1.5, point 2. (no anchor)
<ericP> http://www.w3.org/TR/xpath-functions/#func-dateTime-equal
from Andy's link: "#
Do not apply operations based on date or time types (such as indexing) to collections of data in which some data items may have zone offset information and other data items may not have zone offset information.
"
ericP: i believe we have to clarify whether dynamic context that provides timezone is consistent as you execute a query
<iv_an_ru> I agree, timezone is consistent.
<iv_an_ru> Finally, we intend to reuse XQuery core functions library.
AndyS: how do you define what the implicit timezone is? mapping from lexical to value space needs to be complete
ericP: Not our problem?
<iv_an_ru> Yeah
AndyS: Could send a query to different places and get different answers - if we explicitly permit that, it's bad
<iv_an_ru> Bad queries -- bad answers.
iv_an_ru: could specify implicit timezone in protocol?
(scenario: asking query over graphs C and D to endpoints A and B)
<AndyS> Scenario: A asks B to execute a query which reads C and D
<iv_an_ru> These issues are considered accurately in design of HTTP proxy servers.
<AndyS> End of -> http://lists.w3.org/Archives/Public/public-rdf-dawg/2007JulSep/0130.html has a use case
AndyS: Would like errors in cases that you don't positively know to be true/correct
ericP: doable but a substantive change
<iv_an_ru> Hm. What if neither client nor server know their accurate default timezone but they konw that they use same timezone?
<AndyS> See http://www.w3.org/TR/timezone/ 1.5 bullet 3
Souri: if we have a query with no
timezone, there are two places that it could be fixed
... (1) at storage time, DB could add a default TZ
... then no problem in SPARQL because we never see it w/o a
TZ
... (2) at query time, because a query gets processed in a
distributed fashion
... and we want a consistent answer
... protocol is one way of giving default TZ context for a
query
... that way default TZ is already controlled by the query
<AndyS> Re: (1) the dateTime may be a literal in the query.
(Chair asks that people stick around for 15 minutes or so to flesh this out)
<AndyS> I could read "http://www.w3.org/TR/timezone/ 1.5 bullet 3" as "use Z"
<AndyS> That would stable for SPARQL.
<iv_an_ru> What I'm afraid is that we need some extensible syntax to specify details of the context in the query. Timezone could be first but not the last issue as soon as core function library will grow :(
<AndyS> +1 to Ivan - let's avoid a context as it will grow.
<iv_an_ru> We do not avoid, we've specified a custom syntax extension. We have no other choice.
<iv_an_ru> Others may wish to avoid a context at all.
AndyS cites """
If you have data that includes implicit and fixed explicit zone offsets, before applying any date- or time-sensitive operations adjust the zone offset of the implicit data to UTC with the functions for zone offset adjustment, cf. sec. 10.7 in [XPathFO].
"""
<AndyS> This is not clear text here because it is so knotty.
(discussion of whether this means brand implciit data as Z or adjust from some implicit TZ _to_ UTC)
ericP: I believe what we've published already has a total ordering
AndyS: it's explicitly a different ordering in different places on the Web
LeeF is partial to iv_an_ru's observation - a bad query can get a bad answer
<iv_an_ru> Who the hell had invented the revolving Earth ?!
ericP: Can we change the spec to mandate errors where currently it defines a return value that is not an error?
<iv_an_ru> We may wish to introduce 'SUCCESS with info' sort of reply in the protocol.
ericP: suggesting text to clarify that there's an implicit timezone that is consistent over the life of a query
<iv_an_ru> I agree with ericP.
<iv_an_ru> (So reconfiguration of the timezone of the server should force re-compilation of cached queries.)
<iv_an_ru> I'd prefer to leave default timezone implementation specific plus an optional parameter in the protocol.
<Souri> we are trying to compare if any of the N possible dateTime values is =,<,or > to a specific dateTime value.
<iv_an_ru> Moreover we can explicitly state that SPARQL should provide consistent context for all calls of all XPath functions during the query call. No matter what's in that context.
<Souri> protocol may specify user's intent: be strict and return unknown, or use a specific time zone, or use a system-assigned (say Z) timezone
http://lists.w3.org/Archives/Public/public-rdf-dawg/2007JulSep/0130.html
<iv_an_ru> Souri, you're right, protocol may say 'be strict' instead of providing a timezone.
<Souri> A side question: What is the EBV of "abc"^^xsd:dateTime?
<iv_an_ru> false?
adjourned