RDB2RDF Working Group Teleconference -- 19 Oct 2010

<trackbot> Date: 19 October 2010

<mhausenblas> scribenick: hhalpin

<boris> I'm having problems with the passcode

<ericP> dialing from skype?

Make sure you are typing it in correctly, but it does work via Skype.

(at least on linux skype debian)

Admin

<mhausenblas> PROPOSAL: Accept the minutes of last meeting, see http://www.w3.org/2010/10/12-rdb2rdf-minutes.html

<juansequeda> +1

RESOLUTION: Accepted the minutes of last meeting, see http://www.w3.org/2010/10/12-rdb2rdf-minutes.html

<mhausenblas> http://www.w3.org/2001/sw/rdb2rdf/track/actions/open

<Seema> dialing in right now

<mhausenblas> ACTION-63?

<trackbot> ACTION-63 -- Michael Hausenblas to incorporate Orri's input re datatypes into UCR doc -- due 2010-10-22 -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/actions/63

<mhausenblas> ACTION-69?

<trackbot> ACTION-69 -- Daniel Daniel Miranker to work on UC&R on OLAP application -- due 2010-10-22 -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/actions/69

Dan should have is Invited Expert Status, I forwarded his e-mail to Team IE process.

mhausenblas: will go after those two use-cases

R2ML mapping

<mhausenblas> http://www.w3.org/2001/sw/rdb2rdf/r2rml/

<Seema> latest version is at http://www.w3.org/2001/sw/rdb2rdf/r2rml/Overview.html

mhausenblas: update?

cygri: I did quite a bit of work on section 1
... the announcement was sent out over e-mail.
... tried to explain how the entire thing works, then goes into detail about how Turtle is being used.

<Seema> latest version od the document should be 1.17

cygri: and looking at the overall execution environment

Souri: I worked more on the examples that we have
... tried to make it clearer in terms of properties
... using in-line examples, from full example in appendix
... so they make sense as people are reading

Seema: we reworded things

mhausenblas: it seems to address all the issues

cygri: what is critical obstacle for putting this out to the public?

mhausenblas: I'm happy to propose to put this out as first public working draft
... you would need concrete suggestion to text change if you disagree

PatH: the version I'm looking for it

<Marcelo> Sorry I am late again (I am teaching until 16:00 UTC)

hhalpin: that is why can put the issues up

soeren: PatH's points in previous e-mail

hhalpin: the two points was the confusion over 2 uses of turtle

PatH: this document is the mapping, doesn't tell you.

<Souri> we have put some comments regarding the mapping in: http://www.w3.org/2001/sw/rdb2rdf/r2rml/#R2rmlOverview

PatH: the Turtle defines a class of mapping, needs to tell you more about the kinds of mapping

ericP: would it possible for an examples in the Turtle language and generated RDF help?
... would that put the reader on the mapping

<Souri> @@ Proposed text from the call: “A row in a relational table has <a row identifier> and a number of <column-name, value> pairs. Such a row may be translated to RDF as follows: subject := MAP(<row identifier>) and <predicate, object> := <MAP(column-name), MAP(value)>, for each <column-name, value> pair. An RDF triple is generated by combining the above as follows: <subject, predicate, object>.”

MacTed: The R2ML doc is in context of some other documentation

<soeren> +q

MacTed: in primary, the use-case document

soeren: how does this relate to the semantics of the mapping
... ideally we could define the mapping between SPARQL and SQL relational algebra
... I drafted this small test document
... where I have a very simple definition of this mapping
... but it may be difficult to align with the current mapping

PatH: Explain to the reader what is being talked about in this document

<soeren> +q

<juansequeda> +1 to hhalpin

<mhausenblas> +1 as well

<Zakim> betehess, you wanted to ask if someone can define what is "semantics" here? where are not defining an algebra, only a mapping

<juansequeda> hhalpin: the R2RML document is for database admin who are going to write the mapping language.

hhalpin: let's not define too much in a single document

BetHess: there is no semantics as function from input to output

Marcelo: we are trying to define the syntax and semantics of the mapping

<Marcelo> http://www.w3.org/2001/sw/rdb2rdf/wiki/Semantics_of_R2RML

<Marcelo> http://www.w3.org/2001/sw/rdb2rdf/wiki/Database-Instance-Only_and_Database-Instances-and-Schema_Mapping

BetHess: I don't see the goal, another document about semantics, another document about syntax

<Marcelo> http://www.w3.org/2001/sw/rdb2rdf/directGraph/

BetHess: I see no relation with current document

Marcelo: This is what we are trying to do
... so we have several proposals
... the right way to define semantics of mapping

BetHess: I see a semantics from Datalog

Marcelo: This is a proposal, but we haven't made a formal decision
... but it's just one alternative
... Datalog gives us a precise semantics, and EricP...

Juan: Datalog is equivlaent to SPARQL, SQL, and RIF, so it's fine

BetHess: Only if you accept your output is URI or a blank node

<Marcelo> +q

<Souri> just for quick reference, let me put the preliminary proposed text (from Sec 2) for informally explaining the input and output for the mapping

Souri: we proposed how to do this informally, and then we are describing with a row how to get a bunch of triples out of it.

<mhausenblas> Michael: I totally agree with hhalpin summary. let's define input/output on the syntax and then focus on finding the 'fitting' model/semantics

<betehess> please, can you show me how people define mappings from one model to another one? no mention of semantics is required in the answer

<betehess> hhalpin, give me an example

<betehess> [ hhalpin: you create a model that can be proven to express everything in the two models you want to map, and then you use that first model to express the mapping language ]

soeren: I think we should have the R2ML be purely for implementers of the mapping languge

<Zakim> MacTed, you wanted to talk about setting context within the doc... each must refer to the others that are key/important to its understanding

soeren: and that database admins will not read the R2ML language

MacTed: We can describe what you can say with this kind of langugae without actually saying it

<juansequeda> betehess: Consider the schema: student(s_id, name)

MacTed: this R2ML definition language, absolutely and unequivocally relies of comprehension and context of other documents
... we have to either restate those documents

<juansequeda> and consider instance student(1, Bob)

MacTed: we have to point to the other documents.

<juansequeda> This is the mapping: Triple(s, p, y) ← student(x, y), generateTupleURI("student", x, s), generateAttributeURI("student", "name", p)

<PatH> +1 to whoever is speaking right now.

<juansequeda> This is the output: Triple(http://www.example.com/DB/student/1#, http://www.example.com/DB/student/name#, "Bob")

MacTed: r2rl follows after the default mapping in a conceptual way

<PatH> tnx

<betehess> juansequeda, I want an example in real life, where people defined a mapping between 2 different models (HINT: it's call "compilation")

<betehess> +1 PatH

<juansequeda> betehess, I just gave you an example of mapping from relations to triples

hhalpin: so to clarify to BetHess, if you can show datamodel A can be expressed in datamodel C, and so can datamodel B, then you use datamodel C. Maybe Google "Abstraction Theorem"

MacTed: it's a related specification
... we need to make sure it says it relies on the mapping document

ericP: people should read the previous document first.

<betehess> juansequeda, a triple is not a relation, here it is http://www.w3.org/TR/2004/REC-rdf-concepts-20040210/#section-Graph-syntax

<juansequeda> betehess, I'm not saying that a triple is a relation. I just show you a way to map a relation to a triple

<betehess> hhalpin, I don't follow you re: datamodels A B C. Please give me an example in real life. No semantics, just "how?"

hhalpin: We should test for consensus next week

<Zakim> ericP, you wanted to propose a way to use examples to move this document forwards

hhalpin: but eveyrone should read the document next week.
... and send arguments to the list-serv.

<ericP> [[

<ericP> The RDB2RDF WG is producing a mapping from relational databases to RDF graphs.

<ericP> The mapping parameteters are controlled by another RDF graph.

<ericP> This document describes that graph and begins to elucidate the mapping with inline examples; it does not yet define the mapping.

hhalpin: and then we can see what parts of direct mapping can be put into R2ML document re normative dependencies

<ericP> ]]

<juansequeda> +1 to ericP. Add more examples to R2RML document to get this out

Marcelo: What we are hoping is that Datalog can express it as rules

<Zakim> mhausenblas, you wanted to propose next steps

ericP: If we kept it example driven and added in the above text

+1 ericP

<juansequeda> I would recommend to put the appendix example in the beginning as a running example

<PatH> +1 Eric

mhausenblas: people can propose complete textual changes
... but we need to make sure these very direct and valid questions can be put forward by Friday
... so this would allow the editors to know if they hvae addressed the documents
... then we review to see if they have all been addressed by next week.

juan: I really back up by what Eric just said
... and then we can have concrete input-output examples
... moving appendix to right after Section 1.

<MacTed> +1 running example

+1 running example, but maybe "build" it up over time.

<Zakim> ericP, you wanted to say that *all* of those examples might be a bunch to front-load; propose sprinkling them in

scribe: example up front makes much more sense

ericP: too much to read up front

<juansequeda> Proposal: move the example in Appendix A to the beginning of the document so it can be treated as a running example

Souri: I have a question

<MacTed> as I understand it ... *basic* statement of example -- "I have this RDB table, which I want to map to this RDF ontology" -- at the beginning

Souri: we have inline in the latest example

<MacTed> then as each bit comes along in the spec, the relevant bite of what is now in App A goes there

<MacTed> App A *could* (and perhaps *should*) be kept intact where it is as a summation

<PatH> +1 MacTed

+1 MacTed

mhausenblas: Can you revise Juan's proposal?
... Ted?

Seema: a first question for Juan or just the specification of the table

<MacTed> Proposal: copy *basic* statement of example from Appendix A to beginning of document; then inline relevant bite from App A with the document segment it applies to; then present complete example as Appendix A

Seema: i.e. put the tables and triples together at the beginning then each section building the mapping little by little

mhausenblas: are these instructions sufficient?

Souri: Your saying that the input to the mapping should be explained in the beginning and put right up front
... then we build up each of them
... and at the end we give the whole specification for the example,

<mhausenblas> Michael: copy *basic* statement of example from Appendix A to beginning of document; then inline relevant bite from App A with the document segment it applies to; then present complete example as Appendix A and publish this as FPWD

Souri: so we put input first, and then we show how it maps from the example
... and then at the end, we show the complete mapping specification.

<mhausenblas> PROPOSAL: copy *basic* statement of example from Appendix A to beginning of document; then inline relevant bite from App A with the document segment it applies to; then present complete example as Appendix A and publish this as FPWD

Souri: is that the proposal?

juan: yes, input should be first

Souri: sounds fine to me!

<MacTed> +1 proposal

mhausenblas: any objections to proposal?

<Souri> +1

<juansequeda> +1

<Seema> +1

RESOLUTION: copy *basic* statement of example from Appendix A to beginning of document; then inline relevant bite from Appendix A with the document segment it applies to; then present complete example as Appendix A and publish this as FPWD

<mhausenblas> ACTION: Hausenblas to take care of pubrule check and init publication with W3C team contacts [recorded in http://www.w3.org/2010/10/19-rdb2rdf-minutes.html#action01]

<trackbot> Created ACTION-75 - Take care of pubrule check and init publication with W3C team contacts [on Michael Hausenblas - due 2010-10-26].

<scribe> ACTION: hhalpin to help with publication process [recorded in http://www.w3.org/2010/10/19-rdb2rdf-minutes.html#action02]

<trackbot> Created ACTION-76 - Help with publication process [on Harry Halpin - due 2010-10-26].

hhalpin: we are OK with publishing, right?

<mhausenblas> Michael:

mhausenblas: yes, assuming that proposal is implemented

juan: any updates on semantics telecons?

ericP: Thursday 2 PM?
... 2PM EST
... shortname?

R2ML?

<mhausenblas> http://www.w3.org/TR/rdb2rdf-ucr/

<mhausenblas> http://www.w3.org/TR/rdb2rdf-r2rml/

hhalpin: or just www.w3.org/TR/r2rml?

short is better IMHO

<cygri> +1 hhalpin

<MacTed> short is better ... except that it removes the context of rdb2rdf docs

mhausenblas: 2 of Nov no telecon.

<betehess> +1 to rdb2rdf-r2rml

<mhausenblas> [adjourned]

trackbot, end meeting

- DRAFT -

RDB2RDF Working Group Teleconference

19 Oct 2010

Attendees

Contents

Admin

R2ML mapping

Summary of Action Items

Scribe.perl diagnostic output