See also: IRC log
<Ashok> rackbot, start the meeting
<Ashok> trackbot, start the meeting
<trackbot> Meeting: RDB2RDF Working Group Teleconference
<trackbot> Date: 06 March 2012
<Ashok> scribenick: nunolopes
PROPOSAL: Accept minutes of last meeting: http://www.w3.org/2012/02/14-RDB2RDF-minutes.html
<Ashok> PROPOSAL Approve minutes from Feb 28: http://www.w3.org/2012/02/28-RDB2RDF-minutes.html
RESOLUTION: Accepted minutes of last meeting: http://www.w3.org/2012/02/28-RDB2RDF-minutes.html
Ashok: Adobe said by email that they have no essential claims
http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2012Feb/0085.html
Ashok: The director was anxious regarding implementations
… lots of them demonstrate public uptake
… we agreed that we would have 4 implementations for each spec
… from which 2 would send formal implementation reports
<dmcneil> i think ashok said _4_ implementations for each spec
… can we have an idea on who would be doing these implementations?
… is D2R doing one?
<dmcneil> +q
<Ashok> Yes, 4 for each spec, 2 of them should submit formal implementation reports
<dmcneil> -q
cygri: we will be doing this but the timing might not fit with the WG timing
<juansequeda> +q
<dmcneil> +q
Ashok: ok, anybody else with a formal report for R2RML
juansequeda: UltraWrap are almost ready to have the report out
… for both specs
dmcneil: we are in a similar situation to what richard described
… we assume we'll do an implementation report but can't commit to timing
… probably second quarter 2012
<dmcneil> -q
boris: we can probably make the deadline
… after I finish the test cases I will look into this
… only for the R2RML spec
… before the end of next month
MacTed: We can try before the end of next month but cannot be sure
… for both specs
+q
<Zakim> ericP, you wanted to say i should have a reportable implementation in a week or so
ericP: I should have a DM implementation
… passing all the tests in about a week
… I'm patching locally all the tests and was planning to then submit the diffs
Ashok: are you going to do a formal report?
ericP: I intent to
Ashok: good news, as you make progress can you email the WG
<ericP> nunolopes: i have an implementation of both specs in XSPARQL
<ericP> ... i need to look at the current tests and send something in a week
juansequeda: can you explain the process of implementation report vs test-case report?
Ashok: what we require is one report, from 2 of the implementations at least
… a formal document stating that we passed the tests and which failed
<ericP> as i understand the WG concensus, that document should be the EARL format used by SPARQL
… and why failed
… 2 others would just require an email stating that the implementation passed the tests
juansequeda: we'd need the test harness code
<boris> http://www.w3.org/2001/sw/rdb2rdf/wiki/Submitting_Test_Results
Ashok: the report would be in EARL format
ericP: we follow the SPARQL WG on this
… everybody ran the tests and sends back turtle
… which implementation passes which tests
… looking up an example of EARL
juansequeda: the EARL report is part of the complete implementation report from the WG?
ericP: yes
... for the implementation report we can go through the tests
and determine which features are implemented
… for SPARQL we had a more formal requirements
… for us it's sufficient to have the feature names?
… assign URLs to each feature boris identified
cygri: I have some comments
… a lot of the test cases use R2RML in a way that is not the shortest
… like using the rr:constant construct, instead of rr:predicte
… it's not wrong and works but would be better to show the more compact way
… to guide people to some best practices
… I would propose the TC use the shorter version
juansequeda: we should have a combination of both
… some should test the long form
<ericP> example EARL
cygri: yes, at least one use the long version
… another point is that when there is an integer column has associated the type specifically
… which is something we don't want
… R2RML should use the datatype mapping
… to get the most appropriate datatype
… again have TC should show that this can be changed but the others should not use the datatyype specification
… would be good to keep the TC as simple as possible
<ericP> +1 to minimizing complexity of test cases
<Ashok> +1 to minimizing complexity of test cases
… since it's easier to check what goes wrong in an implementation
<ericP> minimized complexity will help the implementation report in the same way it helps fmplementors debug
juansequeda: I ran all the test cases and so far they worked nicely
… they are incremental in the way the schema is generated and features are introduced
… on the contrary I would expect more complicated test cases
… with combinations of advanced features
<Zakim> juansequeda, you wanted to ask about the formal report
boris: so far no one else has ran the test cases
… in the end we should include more complicated
… but I agree with richard
ericP: from an implementation report point of view it's easier to have simple test cases
… opposed to more complicated ones
… the problem with the complex ones is that we can't tell which features are not implemented
scribe: but if we can do it based on earlier tests that's ok
Ashok: the test cases on other WG have been very minimal
juansequeda: in TC 8C, with multiple predicate maps,
<boris> http://www.w3.org/2001/sw/rdb2rdf/test-cases/#R2RMLTC0008c
… if we combine with multiple object maps the result will be a combination of all?
cygri: yes
juansequeda: that's ok from the implementation side but it's strange
cygri: we had a discussion on this
Souri: this TC has 2 predicatempas and one object map but does not have the other case
… you say that N predicate maps and M object maps are weird
juansequeda: yes
Souri: that generates the cartesean product
… it is not specifically stated but is allowed
juansequeda: in TC 19a, it has a subject map
… if the data has IRIs then its fine
<boris> http://www.w3.org/2001/sw/rdb2rdf/test-cases/#R2RMLTC0019a
… but if it's a mix we have a data error
cygri: referring to section 7.4 of the R2RML spec
<boris> http://www.w3.org/2001/sw/rdb2rdf/r2rml/#termtype
… but you're right that this should not produce anything due to the space in the 3rd row
… because if a query has any data error you will not get a partial result
<boris> http://www.w3.org/2001/sw/rdb2rdf/test-cases/#R2RMLTC0019a is ok
<boris> http://www.w3.org/2001/sw/rdb2rdf/test-cases/#R2RMLTC0019b has data-error
… as david pointed out, data should not be silently dropped
juansequeda: ok, without the 3rd row it would be fine?
cygri: I don't think so
… there is a base declaration used as base IRI
… and is also being used in this TC to get the absolute IRI when you have relative IRIs in the data
… to be precise every test case should have 2 inputs
… database and base IRI
<Souri> Should we change "Jhon" to "John" ?
juansequeda: in the DM graph n.6 the input would need to be changed (?)
discussion on relative vs absolute IRIs
ivan: ntriples doesn't allow relative IRIs
… since we can't have @ prefixes
cygri: yes, this is not allowed syntactically
… we can just call them turtle and that would work
… I would prefer that R2RML would have absolute IRI as output
… I don't have an opinion for the DM
Souri: do we do percent encoding in … ?
cygri: for templates yes
… for a column we don't , giving users the choice of doing so using a view
… the user is responsible for producing valid iris
Souri: if we have an rr:template what is the role of the base?
<Souri> ../{FIRSTNAME}
cygri: even with the base you can have a relative IRI
… it's simply a concatenation with the base IRI
<cygri> rr:template "{\"Name\"}";
<Souri> rr:template "{FIRSTNAME}"
Souri: in a simple template can we use base and percent encoding?
cygri: checking the spec
juansequeda: if that is the case TC 19a would not be wrong
… since the space would be %20
cygri: no, it uses rr:column
Ashok: can we continue by email?
<Souri> first row: http://example.com/ns#Jhon -> would that be right?
<boris> very good question souri
<juansequeda> Souri, but then you have to check every single value if it's a valid IRI
boris: for R2RML I have to input the base IRI
cygri: most of the cases it does not matter
<boris> +1 to say in the begining
… we can just add a line in the beginning of the document
… for the ones where it si needed it should not be taken from the mapping file
Ashok: should we have a telco next week?
<ericP> http://people.apache.org/~andy/ARQ-earl-2012-02-02.ttl
<Souri> 3rd row: http://mappingpedia.org/rdb2rdf/r2rml/tc/Juan%20Daniel -> right?
ericP: an example of an earl report
Ashok: ok, let's have a telco next week and we can cancel if needed
RRSAgent: draft minutes
This is scribe.perl Revision: 1.136 of Date: 2011/05/12 12:01:43 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Succeeded: s/have 2/have 4/ Succeeded: s/IRIR/IRI/ Found ScribeNick: nunolopes Inferring Scribes: nunolopes Present: Ted Ashok Richard Nuno David Boris Seema Eric Juan Souri Ivan Regrets: Michael Found Date: 06 Mar 2012 Guessing minutes URL: http://www.w3.org/2012/03/06-RDB2RDF-minutes.html People with action items: WARNING: Input appears to use implicit continuation lines. You may need the "-implicitContinuations" option.[End of scribe.perl diagnostic output]