<trackbot> Date: 09 January 2013

<SteveH_> I can hear andy...

<markus> zaki, ??P6 is me

<scribe> scribenick: cgreer

admin

resolved: minutes accepted from last meeting

open actions

Turtle

<sandro> (Ivan and I were in another meeting, sorry)

Guus: We have three open issues still for Turtle. First is ISSUE-95.

@ericp: I need to keep this in mind and follow up. I spoke with the commenter, who reluctantly agreed... to accept the resolution without overt happiness.

<Guus> RESOLVED: minutes of 19 Dec telecon accepted http://www.w3.org/2011/rdf-wg/meeting/2012-12-19

Guus: Second open issue for Turtle, ISSUE-100.

Sandro: I think the spec is just fine, and there's not enough information in the issue.

<AZ> is it related to ACTION-191: Add escaping to Turtle in HTML?

<sandro> action-191?

<trackbot> ACTION-191 -- Gavin Carothers to add escaping to Turtle in HTML -- due 2012-10-17 -- OPEN

<trackbot> http://www.w3.org/2011/rdf-wg/track/actions/191

@ericP: Trying to understand a document with several graphs embedded in it implicitly.

gkellogg: I put everything I find into the default graph
... I even look for microdata when searching for RDFa and include that.

Guus: So we need to add some links to this issue description and close it.
... It's still on Gavin's plate for ACTION-191

Subtopic: Well-formed lists

Sandro: I know this one, and it has nothing to do with turtle.
... It's a general RDF issue, with implications for turtle.

@ericp: Are there type constraints for well-formed lists?

<davidwood> Sorry - in a noisy airport and can't get a good connection. Will try to monitor, though. Please IRC me directly if you would like my attention.

Sandro: My proposal is to defined well-formed lists as what turtle does here.

<sandro> sandro: this is based on turtle, no affecting turtle

Guus: This in other words doesn't affect turtle.

<ericP> i think http://www.w3.org/mid/20130109161231.GD30410@w3.org closes ISSUE-95

@ericp: The feature at risk, second bullet point in http://lists.w3.org/Archives/Public/public-rdf-wg/2013Jan/0025.html

@ericp: Assuming I know what the feature at risk is, I think it's not an issue.

Guus: So we need a new resolution about what 'at risk features' are.

<sandro> cgreer, there should not be a "@" before "ericp" in scribing....

Guus: We'll look for a grammar resolution that resolves conflicts.

ericp: And we did that before publication.

<sandro> GRRRRRRRRRRRRR. Stupid css "Feature at Risk" means I can't search for them.

Sandro: Can I propose we not use CSS for "feature at risk?"

<sandro> ACTION: ericP change the AtRisk styling to allow searching documents for the words "feature" and "risk" to find Features At Risk [recorded in http://www.w3.org/2013/01/09-rdf-wg-minutes.html#action01]

<trackbot> Created ACTION-220 - Change the AtRisk styling to allow searching documents for the words "feature" and "risk" to find Features At Risk [on Eric Prud'hommeaux - due 2013-01-16].

<ericP> https://dvcs.w3.org/hg/rdf/raw-file/default/rdf-turtle/index.html#sec-grammar-grammar

Guus: Either today or after telecom I'd like proposal for "Features at Risk"

ericp: If we accept grammar issues I can remove diff markings.

gavinc: They are editorial and only changes are in line with features at risk.

<ericP> PROPOSAL: the Turtle CR will have one feature at risk: the grammar productions for SPARQL prefix and base

Guus: What determines whether this feature is in or out? Implementation?

ericp: Feedback is invited.

Guus: Usually features at risk are decided by implementation experience, is this OK?

Sandro: Yes

<ericP> +1

<ivan> +1

<gkellogg> +1

<yvesr> +1

+1

<markus> +1

<Guus> +1

<AndyS> +1

<Arnaud> +1

<AZ> +1

<ericP> RESOLVED: the Turtle CR will have one feature at risk: the grammar productions for SPARQL prefix and base

<gavinc> +1

Guus: A few more issues in my message... OK HTML5 link.
... in the transition request we must indicate whether there are normative references to

Just lost audio.

<ericP> Guus: the "normative ref" to HTML5 is actually in an informative section

<ericP> ... the link is regarding script tags

Guus: I'm unclear as to the status of this document.

<gavinc> gavinc: The HTML5 refrence being nominative is a bug.

Sandro: as long as it's not a normative reference it's OK.

gavinc: The link only leads to another link, to editor's draft.

Guus: The purpose of this whole scheme is that it should work in five years.

ivan: Having a reference to stable CR document is fine and that's what we should do.

gavinc: So we no longer refer to latest editor's draft?

ericp: I'm OK with that regardless of what the link end says.

<ivan> http://localhost:8001/TR/2012/CR-html5-20121217/

<gkellogg> http://www.w3.org/TR/html5/scripting-1.html#the-script-element

<ivan> http://www.w3.org/TR/2012/CR-html5-20121217/

gavinc: I'll get the link fixed.

Guus: What about the date of the documents?

sandro: When you check in, update the date.

ericp: I'll update CSS and Gavin will update link reference.

Guus: It would be good to send request Friday.
... Can you do this in next 24 hours?

gavinc: Sure.

subtopic: Coverage of Turtle Tests

ericp: The goal is to look through the grammar and concepts of language features and make sure we have tests that cover them all
... In the process I created a set of minimal tests. One n-triples statement, then features are minimal deviations from that.

<ericP> https://dvcs.w3.org/hg/rdf/raw-file/default/rdf-turtle/coverage/report.html

ericp: There are some redundant tests (with exactly same features)
... So another question - do we want to use existing test suite or minimal tests.
... Minimal tests have advantage of what feature is failing.

<gavinc> "minimal tests?"

<AndyS> where are the minimal tests as a suite online?

ericp: If we use Andy's tests, we know they're weathered, people have been using them for a while.

sandro: What's the advantage of not having both?

ericp: If we're pedantic about it, if somebody passes minimal tests, except the two at the end, then fail a couple of Andy's tests,
... Andy's tests may not pinpoint feature failures precisely.

<gavinc> That's a failure to implement Turtle :P

Sandro: I'd rather have both sets of tests.

ericp: We ran into this with SPARQL and didn't know how to resolve it.

<gavinc> Turtle is a great deal smaller then a query language.

ericp: another option is to use minimal tests when reporting on failure, and others as a second set.

Sandro: Submitter should be passing all tests anyhow.

ericp: From SPARQL, a lot of folks failed entailment tests because they didn't implement them.
... with the minimal tests then I'd rather see a failure than a disabled test.

<gavinc> Turtle is VERY VERY tiny

Sandro: I don't think this will matter for turtle. We expect full implementations.

<gavinc> I expect ALL implementations to implement ALL of Turtle :P

ericp: But it's still useful to know about what the test coverage is for a given language version.

gkellogg: In RDFa we tend to break features out, and have a feature-based test suite.
... Failing features are generally removed from the test reports.

Sandro: It seems that Eric's tests are unit tests for debugging, and Andy's are for regression, confidence in whole system

gkellogg: Interaction tests are useful. Beyond minimal feature tests.

ericp: But we've not been diligent to test feature interaction.

<AndyS> No test suite is complete. Lots of weird, obscure interactions possible e.g. all bnodes are all the same.

gkellogg: Also we'll find that people add tests when they hit failures. If we find processors that all pass, but differ in a particular use case, we add a test that covers that combination.

ericp: We can do that same thing here.

<PatH> Sorry, have to leave early fter arriving late.

<sandro> gavinc, I wonder if we can find someone to donate a prize to the writer of the smallest turtle parser. :-)

Guus: I think we list the minimal test suite in the transition request.

<gavinc> it exists inside a perl module which is now checked in with Turtle

<AndyS> So I can't execute it. Hmm.

<sandro> checked in with Turtle?

gavinc: Didn't Greg already do reporting for the larger test suite?

ericp: What I don't understand is how to map features to test reports.

gavinc: Why do we have to do that?

ericp: It's a method for finding which implementations have what shortcomings.

sandro: this situation is for when a big test fails, and we don't know what that means.

ericp: So if they've failed a small test... that justifies the division between minimal test suite and the larger one.

sandro: I agree that it's good to have a distinction.

gkellogg: Two manifests would help. In JSON-LD we decided that the tests that start at '1000' are intended to be large, hard tests.
... for something like SPARQL, you could break it into several manifests.
... but turtle is small, and bigger tests that fail should probably be rewritten into smaller ones.
... a single test manifest is probably best here, and more practical. We could partition the tests at some point that distinguishes between minimal tests and larger ones.

<ericP> minimal tests (3rd column)

<gavinc> ... again, does ANYONE think that Turtle implementations will NOT be complete?

sandro: all the approved test should be in the transition request.

<Zakim> AndyS, you wanted to ask where the minimal suite is

ericp: Do we want syntax tests that are the same as evaluation tests?

davidwood: My implemenations that had evaluation tests.. those were also treated as syntax tests.

<AndyS> the eval test, + and - should be syntax correct.

ericp: It sounds like all of these tests had the reference graph with them.

sandro: Back then, no, there were a lot of positive tests, negative tests, and entailment tests.
... And I treated the latter as syntax tests too.

oops

ericp: The difference between evaluation tests and syntax is reporting.

I'll get it back from Gavin later.

ericp: Do we take all of the evaluation tests, and duplicate them as syntax tests?
... or the opposite, eliminating redudandancies

<gavinc> No. We don't do that.

gkellogg: I don't see a reason to remove any tests. eval tests are a superset of syntax tests. If you were to fail evaluation tests that passed for syntax, what does that mean?

<AndyS> There should be no syntax tested in eval that is not elsewhere. May well have got coverage wrong and would like to see people submit new tests.

gkellogg: The more ways you test something the better your confidence level is.

<gavinc> Yes.

ivan: We have a bunch of other things to discuss. Testing details should be out of scope for telecon.

<gkellogg> agree with Ivan

ivan: The charter is expiring (an example of something we need to discuss as a WG)

<tlr> hi guys, any chance to get off the bridge?

Guus: If you check the open issues list, we seem to be almost done. There are some issues in cleanup tasks.
... We should mainly be completing the documents.

<AndyS> me AOB -- I have a Q about TriG and NQuads ... we are doing these aren't we?

ivan: But what about the semantic document discussions? We need to bring that back up.

I just scribed last time :)

<markus> bye

<Guus> trackbot, end meeting

This is scribe.perl Revision: 1.137  of Date: 2012/09/20 20:19:01  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/

Guessing input format: RRSAgent_Text_Format (score 1.00)

Succeeded: s/@erikp/@ericP/
Succeeded: s/propse/propose/
Succeeded: s/having/not having/
Succeeded: s/davidwood/sandro/
Found ScribeNick: cgreer
Inferring Scribes: cgreer
Default Present: +31.20.598.aaaa, Arnaud, AndyS, markus, gkellogg, Guus, SteveH, AZ, +1.707.874.aacc, cgreer, yvesr, ericP, Sandro, Ivan, +1.408.992.aadd, +1.603.897.aaee, Souri, PatH, GavinC, pchampin
Present: +31.20.598.aaaa Arnaud AndyS markus gkellogg Guus SteveH AZ +1.707.874.aacc cgreer yvesr ericP Sandro Ivan +1.408.992.aadd +1.603.897.aaee Souri PatH GavinC pchampin

WARNING: No meeting chair found!
You should specify the meeting chair like this:
<dbooth> Chair: dbooth

Found Date: 09 Jan 2013
Guessing minutes URL: http://www.w3.org/2013/01/09-rdf-wg-minutes.html
People with action items: ericp