RDB2RDF Working Group Teleconference

<trackbot> Date: 09 August 2011

<scribe> scribenick: mhausenblas

Admin

PROPOSAL: Accept the minutes of last meeting http://www.w3.org/2011/08/02-rdb2rdf-minutes.html

<soeren> +1

<juansequeda> +1

RESOLUTION: Accept the minutes of last meeting http://www.w3.org/2011/08/02-rdb2rdf-minutes.html

Current actions

ACTION-149?

<trackbot> ACTION-149 -- Nuno Lopes to look into how SQL 2008 handles CAST of user-defined types and constructed types to string -- due 2011-08-09 -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/actions/149

<nunolopes> http://www.w3.org/2001/sw/rdb2rdf/wiki/Mapping_SQL_datatypes_to_XML_Schema_datatypes

Nuno: I've updated the Wiki page

Michael: Richard, do you need anything more or is this fine now what Nuno put together?

(Nuno explains details what he did ...)

close ACTION-149

<trackbot> ACTION-149 Look into how SQL 2008 handles CAST of user-defined types and constructed types to string closed

Richard: Haven't seen the update yet

<scribe> ACTION: Richard to check http://www.w3.org/2001/sw/rdb2rdf/wiki/Mapping_SQL_datatypes_to_XML_Schema_datatypes and let WG know if something more is need [recorded in http://www.w3.org/2011/08/09-rdb2rdf-minutes.html#action01]

<trackbot> Created ACTION-151 - Check http://www.w3.org/2001/sw/rdb2rdf/wiki/Mapping_SQL_datatypes_to_XML_Schema_datatypes and let WG know if something more is need [on Richard Cyganiak - due 2011-08-16].

ACTION-150?

<trackbot> ACTION-150 -- Richard Cyganiak to come up with a proposal for ISSUE-61 (entity ID look-up table) -- due 2011-08-09 -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/actions/150

Richard: Working on a draft for a proposal
... will post it on the Wiki and share with WG

Michael: Once you've sent this out, Richard, feel free to close the action

Direct Mapping (DM)

http://www.w3.org/2001/sw/rdb2rdf/directMapping/EGP

Eric: I'm currently working on addressing the comments WG members have brought up so far

<dmcneil> the EGP link above is not working for me

<soeren> I get an XSLT stylesheet problem when trying to access the EGP

Eric: examples, formal notations

Souri: I've sent two DM-related proposals

<ericP> http://www.w3.org/2001/sw/rdb2rdf/directMapping/EGP.xml#notation

<ericP> fixed

thanks!

<Souri> My first proposal this morning for DM: http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2011Aug/0053.html

Eric: main thing was to turn the Normative text into plain English

<Souri> My second proposal this morning for DM: http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2011Aug/0055.html

Michael: thanks a million, Souri!
... Issues should be 'real' issues, that is raised in the tracker

ISSUE-63?

<trackbot> ISSUE-63 -- Normative expression of the Direct Mapping in Plain English -- open

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

<ivan> +1

<ivan> -1

PROPOSAL: To close ISSUE-63 with accepting http://www.w3.org/2001/sw/rdb2rdf/directMapping/EGP

<ivan> +1

<cygri> +1

<nunolopes> +1

<Marcelo> +1

<juansequeda> +1

<boris> +1

<ericP> +.5

<ericP> +½

RESOLUTION: To close ISSUE-63 with accepting http://www.w3.org/2001/sw/rdb2rdf/directMapping/EGP

<Souri> +1

<ericP> +ε

<dmcneil> it may not be important, but what does "EGP" mean?

<ericP> my initials

Eric: are we standardising the direct graph or the direct mapping

Richard: two questions - how is the doc called and what is the artefact we're standardising
... direct mapping is what I got used to
... for the document
... but I'd like to see direct graph somehow defined cleanly

<juansequeda> the output of a direct mapping is an RDF graph that represents the relational schema, and this RDF graph is also known has the Direct (RDF) graph

<ericP> the direct mapping is how you produce the direct graph by any means necessary

Richard: maybe something like: DM is an algo taking an RDB as input and produces a direct graph

<dmcneil> if the direct mapping were defined in terms of R2RML, then R2RML would define the graph

qck ericP

Eric: seems awkward to say both (scribe unsure ??)

<dmcneil> the way we use our equivalent of the direct mapping, is that we have an algorithm that produces a mapping

<dmcneil> so it is the production of a "direct mapping", not a "direct graph"

q>

Ivan: I agree with Michael that DM Editors can sort the DM vs. DG out themselves
... other comment - if I look at the two formal specs in Appendix they are more like graph not mapping

Richard: I agree with Ivan

Ivan: we should focus on the other remaining issues

Michael: is there an issue for the URI stuff Ivan was referring to?

<Souri> I'll raise two non-editorial issues: 1) current predicate IRI design cannot handle common cases AND 2) literal triple non-generation in the single-col foreign key case causes non-uniformity and need for unnecessary join with the parent table

<ivan> issue-40?

<trackbot> ISSUE-40 -- Express Direct Mapping in terms of R2RML -- open

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/40

<ivan> issue-42?

<trackbot> ISSUE-42 -- How is the direct mapping suppose to handle NULL values wrt Information Preserving -- open

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/42

ISSUE-40

Ivan: not sure if I totally understand the issue per se - might be underspecified

Michael: My understanding is that this is about externalising a DM

Juan: I agree with Michael

Richard: To Ivan - in good SQL impl you have the information schema
... a separate schema that tells you what exists in the DB
... R2RML + info schema is supported then you can impl. the DM in R2RML
... I think it *should* be possible, though might be complex
... but I'm unsure if this has anything to do with the spec

<ericP> cygri: you may be able to write an r2rml mapping over the information schema to produce an r2rml mapping for the [working] schema

<dmcneil> +q

Ivan: I agree with Richard and Michael - this has nothing to do with the spec

<Zakim> juansequeda, you wanted to say that a r2rml representation of the direct mapping will allow the user to not have to start from a blank sheet

Juan: I'm fine with what Ivan suggested

<cygri> i'd prefer to POSTPONE rather than CLOSE

<Souri> Created ISSUE-64: http://www.w3.org/2001/sw/rdb2rdf/track/issues/64

<ivan> PROPOSED: expressing DM in R2RML (ie, issue 40) is an implementation issue that should not be part of a specification (may be the topic of a separate Working Group Note)

David: When I raised the issue I didn't mean what you guys just discussed - only reusing the semantics

<ericP> +1 close

PROPOSAL: Close ISSUE-40 - expressing DM in R2RML (ie, issue 40) is an implementation issue that should not be part of a specification (may be the topic of a separate Working Group Note)
... Postpone ISSUE-40 - expressing DM in R2RML is an implementation issue that should not be part of a specification (but may be the topic of a separate Working Group Note)

<ericP> +1 to postpone, then

<ericP> +1 to close

<ericP> +1 to postpone

<ivan> +1

<ericP> +1

<cygri> +1

<Marcelo> +1

<boris> +1

<juansequeda> +1

<nunolopes> +1

RESOLUTION: Postpone ISSUE-40 - expressing DM in R2RML is an implementation issue that should not be part of a specification (but may be the topic of a separate Working Group Note)

<Souri> Created ISSUE-65: http://www.w3.org/2001/sw/rdb2rdf/track/issues/65

<ivan> issue-42?

<trackbot> ISSUE-42 -- How is the direct mapping suppose to handle NULL values wrt Information Preserving -- open

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/42

ISSUE-42

Michael: I think one of the co-chairs whom I see every day in the mirror is a slacker

<cygri> PROPSOAL: one of the co-chairs whom I see every day in the mirror is a slacker

-1

ISSUE-64

ISSUE-64?

<trackbot> ISSUE-64 -- Predicate IRI design for foreign key does not handle common cases where same column sequence may be used for multiple foreign key constraints -- raised

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/64

(Souri explains the issue)

Eric: the schema has a correlation already for this
... knows the target

Souri: consider just one table with an ID
... firstname and lastname
... can define FK constrains

<ericP> gotcha

(scribe lost)

<Souri> proposed design: <People,Department#deptName,name;deptCity,city>

(Souri explains how Oracle handles it)

Juan: I agree with Souri

<ericP> note, it also requires the referenced table name

<scribe> ACTION: Juan to figure out and solution with Souri to address ISSUE-64 [recorded in http://www.w3.org/2011/08/09-rdb2rdf-minutes.html#action02]

<trackbot> Created ACTION-152 - Figure out and solution with Souri to address ISSUE-64 [on Juan Sequeda - due 2011-08-16].

ISSUE-65

ISSUE-65?

<trackbot> ISSUE-65 -- For uniformity and performance, "literal" triples must be always generated for each child table column in a foreign key -- open

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/65

Souri: very simple issue

<ericP> s{<People/ID=7> <People#deptName,deptCity> <Department/ID=23> .}{<People/ID=7> <People#2,deptName,deptCity,Departments,name,city> <Department/ID=23> .}

Souri: in sec 2.1 ...

Marcelo: will be a redundant mapping then

<scribe> ACTION: Marcelo to draft a solution for ISSUE-65 and send out to WG [recorded in http://www.w3.org/2011/08/09-rdb2rdf-minutes.html#action03]

<trackbot> Created ACTION-153 - Draft a solution for ISSUE-65 and send out to WG [on Marcelo Arenas - due 2011-08-16].

[meeting adjourned]

trackbot, end telecon