RDB2RDF Working Group Teleconference

<trackbot> Date: 13 July 2010

<mhausenblas> aabb is mhausenblas

<juansequeda> ok, so I was a

<lima> I guessed I was IPCaller

<mhausenblas> scribenick: cygri

<mhausenblas> http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2010Jul/0002.html

Admin

mhausenblas: We have two new members

<mhausenblas> Nan Ma UTC, China Electronics Standardization Institute

<mhausenblas> Nophadol Jekjantuk from University of Aberdeen

PROPOSAL: Accept the minutes of last meeting, see

http://www.w3.org/2010/06/29-rdb2rdf-minutes.html

<soeren> +1

RESOLUTION: Accept the minutes of last meeting

mhausenblas: Reminder for the scribe, please send out minutes ASAP after the call

Action items review

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

<mhausenblas> ACTION-63?

<trackbot> ACTION-63 -- Michael Hausenblas to incorporate Orri's input re datatypes into UCR doc -- due 2010-07-06 -- OPEN

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

mhausenblas: ACTION-63 continues

ACTION-64?

<trackbot> ACTION-64 -- Harry Halpin to re-do scribelist to accurately reflect working group composition -- due 2010-07-06 -- OPEN

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

<mhausenblas> ACTION-64?

<trackbot> ACTION-64 -- Harry Halpin to re-do scribelist to accurately reflect working group composition -- due 2010-07-06 -- OPEN

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

mhausenblas: we assume it continues

ACTION-66?

<trackbot> ACTION-66 -- Richard Cyganiak to document issues on the SQL-based approach due July 8 -- due 2010-07-06 -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/actions/66

cygri: continues
... i'll do it shortly

ACTION-67?

<trackbot> ACTION-67 -- Souripriya Das to write first draft of SQL-based approach due July 22 -- due 2010-07-06 -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/actions/67

Souri: will likely take to the 29th ... continues

ACTION-68

ACTION-68?

<trackbot> ACTION-68 -- Juan Sequeda to write note on Ways to Leverage Ontologies due July 8 -- due 2010-07-06 -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/actions/68

<hhalpin> also, will update scribelist after this meeting based on who has attended and given explicit regrets.

juansequeda: continues ... should be able to do it by next week

ACTION-69?

<trackbot> ACTION-69 -- Daniel Daniel Miranker to work on UC&R on OLAP application -- due 2010-07-06 -- OPEN

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

juansequeda: i'll nag him ... continues

Telecons in July and August

<mhausenblas> http://doodle.com/u6qg5z7gmfhc2zak

mhausenblas: please all put availability in the doodle poll

<hhalpin> +1 mhausenblas

<hhalpin> However, given the amount of work, we need to keep going if possible :)

mhausenblas: we should have the calls, unless there is no one at all available
... Looks like there are 4+ people for all dates, so I suggest to have all the calls

Two alternative mapping languages

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

Marcelo: For the last weeks, juansequeda and me were thinking about mapping languages
... we have concluded that there are two alternatives
... (explaining the example database)
... alternative 1: Database-Instance-Only Mapping
... we just create triples, but no ontology or RDFS/OWL predicates
... the p in the triple is of course always a property but we don't say explicitly
... we use datalog in the document

<hhalpin> wonders about conversion of datalog to RIF...does anyone know about that? I would assume it's trivially covered, but we might want to ask Michael Kifer from the RIF WG.

Marcelo: (walks us through the examples, Case 1-3)

<hhalpin> Could just concat IRI with "name",..but with "foaf:name" maybe coming from a string look-up of "name" from say some web service that matches strings to popular URIs, i.e. Sindice.

Marcelo: (there is an assumption that a user writes the rules)
... so for each case there are three things: first a template; second an example rule that a mapping author might write (for the example db); the third is example output after processing
... second alternative: Database-Instances-and-Schema Mapping
... we also want to capture foreign keys etc in RDFS/OWL
... we have some predicates that represent the relational schema: Rel, Attr, PK, FK
... "Ontology Predicates" section has the desired deductions

Souri: I have a comment on BinRel
... not sure wether it's needed
... you could have a table with five FKs in one tuple
... what's special about BinRel?

Marcelo: Nothing special, really. If you want to compute BinRel, fine; but don't have to
... The language is expressive enough to map from the relational predicates to whatever you want

Souri: We may not want want to treat this BinRel case any special. Could just handle it as a normal rel with two FKs

Marcelo: Yes we could change the ontology definitions if that's not wantd
... (back to presentation)
... Value(...) predicate decomposes the relations into "atomics", see example
... (reading aloud all the rules)
... what's nice about this language: you can directly generate RDF triples, also for the RDFS/OWL predicates

mhausenblas: We need time to digest this, more discussion next week

Souri: nice to see this language in this form
... what about composite keys?
... what we did, we used the *name* of the multi-column key as the basis of the property

Marcelo: yes right, we just skipped that here

Ashok: How would I execute this? Would I translate the right hand to SQL?

Marcelo: Yes

Ashok: Why not just have the SQL on the right hand? Why a different syntax?

Marcelo: That's a tough question!
... There's less overhead in our syntax, and it's close to RIF
... Easy to translate between datalog, RIF, SQL etc
... This is just for the sake of semantics -- datalog has well-defined semantics

(last comment was juansequeda)

Marcelo: This is what people who do data integration use

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

mhausenblas: Eric has put together a proposal based on RIF, see this link

<juansequeda> Are you taking about this? http://www.w3.org/2001/sw/rdb2rdf/wiki/Direct_Mapping_in_RIF

<hhalpin> I am pretty sure Datalog can be transformed to SQL is a pretty straightforward manner.

mhausenblas: I'd like to understand the difference or overlap
... this is very new

ericP: I put this together to demonstrate how you go from direct mapping to a nice graph that you can expose to the world
... and second, shows how a small set of RIF constructs can be used to achieve this
... use case: you have a triple store, and have your stuff with different schema, but still want to be able to honor queries according to the old schema

juansequeda: datalog, RIF, you get the same thing out

<mhausenblas> scribenick: mhausenblas

Richard: How does this fit in with our deliverables?
... I'm a bit confused
... first is providing templates, second quite generic (no user options ?)

<scribe> scribenick: cygri

Marcelo: second is also intended as executable

<juansequeda> Marcelo is talking about this doc http://web.ing.puc.cl/~marenas/W3C/mapping_language.txt

juansequeda: imagine db with really complicated relations, keys etc
... the first language is too complicated in that case

mhausenblas: if this was prolog syntax, would i just run it in prolog?

Marcelo: yes exactly

Souri: we have this little software that actually implements these rules
... we look at the tables etc
... we wrote that system once, and it is generic and can be used for all schemas

Marcelo: the language allows users to write other transformations as well

Souri: assuming you have an implementation that understands this rule language and predicates, right?

Marcelo: yes

mhausenblas: Wrapping up
... here's a very interesting proposal
... let's consider discussion on the list
... adjourned

<mhausenblas> thanks a lot to Marcelo and juansequeda