RDF Data Access Weekly -- 9 Oct 2007

<LeeF> ericP, are you around?

<LeeF> SteveH, iv_an_ru - joining us today?

<SteveH> yes, just dialing

<LeeF> Scribe: ericP

<LeeF> minutes from 2 Oct

PROPOSED: accept above minutes as a true record of the last meeting

SECOND: chimezie

APPROVED

next meeting: 16 October

convene

actions

<LeeF> ACTION: AndyS to put replacement text for filter attachment into rq25 [DONE] [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action01]

<LeeF> ACTION: ericP to implement the conservative algorithm (fail a test -> [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action02]

<LeeF> fail the feature) in the implementation report

<LeeF> action -2

<LeeF> ACTION: ericP to implement the conservative algorithm (fail a test -> fail the feature) in the implementation report [DONE] [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action03]

<LeeF> ACTION: LeeF to mail out a complete test case to determine implementation behavior in the OPTIONAL+FILTER+{{...}} case [DONE] [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action04]

<LeeF> ACTION: LeeF to mark test in 0003 as approved, update test suite, notify implementors [DONE] [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action05]

<LeeF> ACTION: LeeF to solicit implementors' desired behavior for {{ ... }} case [DONE] [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action06]

<LeeF> ACTION: LeeF to solicit mechanically generated test results from pyrhho db [DONE] [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action07]

<LeeF> ACTION: LeeF to weed out basic facets from complex tests [DROPPED] [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action08]

<LeeF> ACTION: LeeF to investigate PR mechanics, put together draft transition request [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action09]

<LeeF> ACTION: LeeF to examine 7 tests affected by change in http://lists.w3.org/Archives/Public/public-rdf-dawg/2007JulSep/0177.html in view of implementors' experiences [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action10]

LeeF: AndyS found 7 tests that one would expect to fail if one implemented the letter of the [semantics part of] spec

<LeeF> ACTION: ericP to adapt impl report to include syntax tests [CONTINUES] [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action11]

<LeeF> ACTION: ericP to poke IETF folks about registering SPARQL media types (esp. application/sparql-query) [CONTINUES] [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action12]

query language @@ text

LeeF: @@ text in SPARQL Query has been visited?

AndyS: yes

<LeeF> editor's draft

<AndyS> http://www.w3.org/2001/sw/DataAccess/rq23/rq25.html

http://www.w3.org/2001/sw/DataAccess

optional, filter, {{ ... }}

LeeF: [re optional filter]
... AndyS noted that section 12 was ambiguous about when you can simplify {{}}s to {}

<AndyS> uses query that triggered the investigation

<LeeF> 3a: ... ?x ... OPTIONAL { ... FILTER (... ?x ...) }

<LeeF> 3b: ... ?x ... OPTIONAL { { ... FILTER ( ... ?x ... ) } }

LeeF: we composed queries that show the possible interpretations (note above options)
... we surveyed implementors:
... ... richardNs and chimezie's use the simplified semantics
... ... richard + 3 prefer non-simplied case
... ... chimezie and steve prefer the simplified
... ... glitter does the simplified

<LeeF> #dawg-optional-filter-005-simplified (same semantics)

<LeeF> #dawg-optional-filter-005-not-simplified

LeeF: Ralph's advice: don't hurry a decision and don't break any impls that don't agree to be impl'd

ericP: i pass simplified
... we have a split between simplified and not simplified
... it is a corner case [enumerates conditions]
... how do folks feel about leaving it ambiguous

chimezie: i find the non-simplified counter-intuitive

<LeeF> I find simplified counter-intuitive :-)

afs: "simplified" means that 3a and 3b (above) have the same semantics

chimezie: in this proposal, does text call out the ambiguity?

LeeF: yes, alternative behaviors explained

ericP: no obvious route to concensus on which behavior to choose?

LeeF: if we were in the design phase, yes

chimezie: in GRDDL, we labeled opposing test cases in the SOTD
... i would be content with that

steve: i find that the addition of {}s changes semantics unfortunate
... a bit of a hobson's choice

<SteveH> ...except it's not :) it's more like a rock and a hard place

<LeeF> PROPOSED: Kick EricP out of the DAWG

second

<LeeF> PROPOSED: Note the simplification of {{ ... }} to { ... } (with example: ... ?x ... OPTIONAL { { ... FILTER ( ... ?x ... ) } } as an explicitly ambiguous part of the SPARQL algebra, point to two possible (unapproved) test cases

<LeeF> PROPOSED: Note the simplification of {{ ... }} to { ... } (with example: ... ?x ... OPTIONAL { { ... FILTER ( ... ?x ... ) } } as a recognized ambiguous part of the transformation from SPARQL query to SPARQL algebra expression, point to two possible (unapproved) test cases

second

APPROVED, AndyS abstaining

<LeeF> ACTION: AndyS to draft text recognizing the ambiguity [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action13]

same graph in default and named graph parts of RDF dataset

<LeeF> LeeF on the samed graph twice on the dataset

LeeF: 4 tests assume that bnodes comeing from the same graph loaded into the default graph and into a named graph are distinct
... AndyS and i have seen nothing in the spec that says these graphs must be disjoint
... eliminating this restriction would allow me to pass these tests
... am not proposing that we specify either behavior

AndyS: [missed]

chimezie: instead of eliminating old ones, what if we leave one but unapprove it
... how does that affect PR transition?

LeeF: we weren't including unapproved tests before
... but we just decided to include them so we just won't require a green box

<LeeF> PROPOSED: Unapprove dataset/manifest#dawg-dataset-09,dataset/manifest#dawg-dataset-10, dataset/manifest#dawg-dataset-12, and graph/manifest#dawg-graph-10; approve dataset/manifest#dawg-dataset-09b, dataset/manifest#dawg-dataset-10b, dataset/manifest#dawg-dataset-12b, and graph/manifest#dawg-graph-10b, add informative text noting this possible interop snag to SPARQL QL spec

ericP: do these tests still have value?

LeeF: yes

AndyS: these tests are simply making wrong assumptions (that there are two calls to the parser)

chimezie: it calls out a behavior of web architecture

g1: { _:who foaf:name "Bob" }

query: SELECT ?x ?name FROM <g1> FROM <g1> WHERE { ?x foaf:name ?name }

<chimezie> I would think since the spec is explicit about the indirection of representation -> parse -> graph that how g1 is 'merged' depends completely on the protocol layer

<chimezie> i.e.: new representations over time, or caching headers

GET /g1 HTTP/1.1

if-newer-than: 2007-10-09T15:34:09Z

<LeeF> FROM <urn:lsid:...> :-)

<chimezie> DanC on web arch and dataset construction corelation

<chimezie> doh

<AndyS> { G, (<u1>, G1), (<u2>, G2), . . . (<un>, Gn) }

<AndyS> Each <ui> is distinct.

<chimezie> second

LeeF: would like to use a WBS to gather feedback on PR decision

<LeeF> ACTION: Chimezie to propose text to clarify that not all graphs in the dataset need to be disjoint/distinct [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action14]

<LeeF> APPROVED, chimezie seconds, SteveH and ericP abstain

<LeeF> ACTION: LeeF to fix approval on dataset and graph tests [recorded in http://www.w3.org/2007/10/09-dawg-minutes.html#action15]

- DRAFT -

RDF Data Access Weekly

9 Oct 2007

Attendees

Contents

convene

actions

query language @@ text

optional, filter, {{ ... }}

same graph in default and named graph parts of RDF dataset

Summary of Action Items

Scribe.perl diagnostic output