RDF DAWG Weekly

<LeeF> Proposed Minutes

Seconded by AndyS

Meet next time: in one week's time

Scribe next time: Souri

Ivan: Intro
... XSLT and XQuery impl experience
... implementing SPARQL on SQL for Virtuoso

LeeF: Please send intro email to DAWG list for linking

Convene

LeeF: Any Comments on agenda?

all: (silence)

Review ACTION Items

<scribe> ACTION: LeeF to update the issues list marking punction syntax [DONE] [recorded in http://www.w3.org/2007/01/09-dawg-minutes.html#action01]

<scribe> ACTION: Souri to review rq24 [DONE] [recorded in http://www.w3.org/2007/01/09-dawg-minutes.html#action02]

<scribe> ACTION: ericP to seek clarification from Bob McG [DONE] [recorded in http://www.w3.org/2007/01/09-dawg-minutes.html#action03]

<scribe> ACTION: LeeF to close formsOfDistinct issue [DONE] [recorded in http://www.w3.org/2007/01/09-dawg-minutes.html#action04]

<scribe> ACTION: EricP to sort out some string literal thing for the operator table [DONE] [recorded in http://www.w3.org/2007/01/09-dawg-minutes.html#action05]

<scribe> ACTION: Jeen to get started setting up new test directory [DONE] [recorded in http://www.w3.org/2007/01/09-dawg-minutes.html#action06]

<scribe> ACTION: LeeF to remember that the wee, lost filter tests should be put [CONTINUES] [recorded in http://www.w3.org/2007/01/09-dawg-minutes.html#action07]

<scribe> ACTION: PatH to change the entailment section around to talk about SPARQL first, then more general conditions in a normative appendix [CONTINUES] [recorded in http://www.w3.org/2007/01/09-dawg-minutes.html#action08]

<scribe> ACTION: EricP to run the yacker tool over and annotate the existing tests [CONTINUES] [recorded in http://www.w3.org/2007/01/09-dawg-minutes.html#action09]

<LeeF> Jeen test suite: http://lists.w3.org/Archives/Public/public-rdf-dawg/2007JanMar/0007.html

Test suite

Jeen has committed some tests and super manifests

... syntax tests and triple pattern tests

AndyS (email) There is a more recent set of syntax test.

LeeF: request to review Jeen's first test batch.

(no volunteers on the phone)

LeeF: this agenda item will become a standing "test approval" slot.

<iv_an_ru> I can try to review during the weekend.

<LeeF> iv_an_ru, that would be great

EricP: suggestion to make syntax only cover stuff not covered elsewhere
... because there is overlap.

LeeF: can see value in having broad coverage in syntax tests (e.g. parser writers)

<LeeF> ACTION: LeeF to check if SteveH can eyeball Jeen's first group of tests pre-WG approval (LeeF and iv_an_ru will also try to eyeball) [recorded in http://www.w3.org/2007/01/09-dawg-minutes.html#action10]

<iv_an_ru> can see value in having two outputs per every syntax test: list of parsed lexems plus the rest as usual.

<Zakim> SimonR, you wanted to comment about syntax testing and abstract syntax

AndyS: advocates having a separate set of tests for syntax and not ask implementers to extract tests themselves.

<ericP> yacker kinda gives us an "abstract" syntax, but it's ugly

<ericP> yacker "abstract syntax"

http://sparql.org/validator?query=PREFIX+books%3A+++%3Chttp%3A%2F%2Fexample.org%2Fbook%2F%3E%0D%0APREFIX+dc%3A++++++%3Chttp%3A%2F%2Fpurl.org%2Fdc%2Felements%2F1.1%2F%3E%0D%0A%0D%0ASELECT+%3Fbook+%3Ftitle%0D%0AWHERE%0D%0A+++%7B+%3Fbook+dc%3Atitle+%3Ftitle+%7D&languageSyntax=SPARQL&outputFormat=prefix&linenumbers=true

SimonR: (discusses a formalized abstract representation of a syntax tree)

<SimonR> I'm suggesting that we really do have to standardize a single abstract syntax if the language is going to be unambiguous.

AndyS: The purpose is to get implementation reports
... if a test fails, we don't know why

EricP: does it matter why they fail?

AndyS: Good to know why a test fails for feedback

<iv_an_ru> How can I use an SPARQL abstract syntax if I compile a mix of SPARQL and SQL? I have more nodes in my tree than it is expected for plain SPARQL, ditto more properties.

EricP: burden having more tests because need tracking

<SimonR> Both RDF/XML and NTriples are concrete syntaxes, but the tests that convert between the two do demonstrate that you're parsing the right abstract syntax. We don't quite have an equivalent to NTriples.

<LeeF> AndyS: I have a large test suite that I use against the syntax tests for developing the grammar -- it would be a burden on me to switch to using the evaluation tests to test the grammar

<Zakim> ericP, you wanted to say if the semantics pass, we're happy with whatever abstract syntax the impl used

<SimonR> So it's slightly different from the RDF case, which is standardized at the abstract syntax level with multiple concrete syntaxes. SPARQL is standardized at the concrete syntax level.

LeeF: Will bring this to Jeen's attention.

Path SPARQL and accessingCollections

<LeeF>

<LeeF>

<LeeF>

LeeF: feedback received from Steve and Kendall: both not in favour
... asked for opinions from the WG

<sdas2> This is an important requirement, but requires careful thinking and I'd be leaning towards postponing it.

DanC: prompted by OWL1.1 which is only just being setup - SPARQL 1.1 is 5year away.

Souri: important requirement but not trivial Oracle concerned about the delay.

<iv_an_ru> We (= OpenLink) can't imlement path expressions both easily and effeciently :(

EricP: conservative - nervous of the time taken

<SimonR> I'm in favor of postponing.

SimonR: don't want to precude future solutions

Ivan: can't implement easily and efficiently

<SimonR> the only really important thing is that we don't *preclude* any future solutions, and I think the new information actually supports the idea that we're not ruling out future extension to solve the problem.

Ivan: datalog input on recursion may be informative.
... no really good news here though
... recursion will be coming along sometime but not just yet.
... will have to ignore feature for 6 months

<SimonR> AndyS: (1) Are we really talking about just collections, or the whole general issue of regexps?

<SimonR> AndyS: 2) What are the alternatives?

<DanC> (indeed, Andy's ordering point is a user requirement that no design I know of meets.)

<SimonR> AndyS: 3) It's not just accessing collections, it's also ordering. Sometimes collections are used as closed sets, sometimes as lists -- in which case, order is important.

AndyS: HP is not opposed to reopening this issue but concerned with the size of the WG and energy leaking out.

<ericP> DanC, one could infer more than a magic predicate, but actually build a structure (<foo>) ==> [ :hasOrderedMember [ :member <Foo> ; :ordinal 0 ] ]

<SimonR> AndyS: There are a number of features on the postponed list. It would seem a good idea to prioritize this list so we can reopen them in the best order.

<ericP> DanC, also, our result set is order-able so i guess we can write special semantics that access order directly

<DanC> "write special semantics" <- there be dragons.

AndyS +1

<ericP> yeah, that's what i did in algae so i could walk through manifests

<ericP> hasn't been used elsewhere, though

rq24/rq25 status

email status report

Proposed to adjourn

<ericP> ADJOURN

Seconded: DanC

<SimonR> Adjourned at 15:27 Z.