RDB2RDF Working Group Teleconference -- 30 Mar 2010

<trackbot> Date: 30 March 2010

scribe?

http://www.w3.org/2010/03/23-rdb2rdf-minutes.html

<scribe> ACTION: eric to review the UC and report back to the WG with requests for more details [recorded in http://www.w3.org/2010/03/30-rdb2rdf-minutes.html#action01]

<trackbot> Created ACTION-38 - Review the UC and report back to the WG with requests for more details [on Eric Prud'hommeaux - due 2010-04-06].

<scribe> ACTION: [CONTINUES] Orri to sum up today's discussion [recorded in http://www.w3.org/2010/03/30-rdb2rdf-minutes.html#action02]

Any action in particular, Ahmed?

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

<scribe> ACTION: hhalpin to begin a draft of Working Group Note on use-cases, ordering them from most complete to least complete [recorded in http://www.w3.org/2010/03/30-rdb2rdf-minutes.html#action03]

<trackbot> Created ACTION-39 - Begin a draft of Working Group Note on use-cases, ordering them from most complete to least complete [on Harry Halpin - due 2010-04-06].

<scribe> ACTION: [DONE] Sum up today's discussion [recorded in http://www.w3.org/2010/03/30-rdb2rdf-minutes.html#action04]

<ericP> http://www.w3.org/2001/sw/rdb2rdf/wiki/Use_Cases_and_Requirements#Expressivity

<scribe> ACTION: [DONE] Create Wiki page for RDF team and link from homepage [recorded in http://www.w3.org/2010/03/30-rdb2rdf-minutes.html#action05]

<ericP> http://www.w3.org/2001/sw/rdb2rdf/wiki/Use_Cases_and_Requirements#Expressivity

<Marcelo> Sorry, I am late (I was in a meeting)

Marcelo - we are discussing use-cases right now, perhaps you should read http://www.w3.org/2001/sw/rdb2rdf/wiki/Use_Cases_and_Requirements#Expressivity

<scribe> scribe: hhalpin

<scribe> scribenick: hhalpin

<Marcelo> OK, thanks

ericP: I don't want to present people with too much info
... as they might get lost
... the DDL should be a link out, but we should keep the table right in people's faces

<juansequeda> Here is the url : http://www.w3.org/2001/sw/rdb2rdf/wiki/Use_Cases_and_Requirements#COI_Patient_Recruitment_Demo

Ashok: what about URLs?

<Souri> Table definitions is good enough

juan: its too verbose to keep SQL DDL in document
... but lets have a link out to the DDL and ER diagrams in document

ericP: Strike tables and have ER diagrams?

Ahmed: Lets have it so people can click on ER diagrams and see it.

ericP: haven't built ER diagrams yet

<Souri> "DDL" is often vendor specific -- just putting the "table definitions" should be good enough

Ashok: inline, hyperlink, both fine with me.

ericP: Does anyone want to make ER diagrams for all the databases?

<scribe> ACTION: juan will make ER diagrams for use-case document [recorded in http://www.w3.org/2010/03/30-rdb2rdf-minutes.html#action06]

<trackbot> Created ACTION-40 - Will make ER diagrams for use-case document [on Juan Sequeda - due 2010-04-06].

ericP: not sure how to specify the identity requirements of the "same" person

<Zakim> Souri, you wanted to say we just need definitions of tables and constraints -- *not* the DDL statements as such

Souri: I wanted to mention that we should say "Table Definitions and Constraint Definitions" rather than DDL
... as that can be vendor-specific
... DDL is a misnomer

ericP: I thought it was in ericP
... it was in SQL 1992

Souri: We just need table, vendor specific aspects
... can be kept out

juan: yes, we can keep it simple - just types, etc.

souri: what are we talking about identity resolution

angela: we create an okkam id for each patient
... in my use-case
... we could have a table for okkam ids

ericP: if we have a look-up table, then we think graph transformations and identity isomorphisms meet your requirements

angela: as long as I can do union of two graphs if these talk about same entity

ericP: or some table that specifies the table with unique identifier

souri: we may call the same person john or johnny with some equivalentces
... that is very voluminous
... so we have to create those as a separate look-up table
... and then maintain the lookup table.

,.. thats another part

scribe: of the entire thing
... so given one database, so we can map to SPARQL
... so we can do individual equivalence

Ahmed: the issue is metadata management
... this is an integration above the mapping language

ericP: we have tables that map from one identifier to another
... tuple identifiers

Ahmed: trying to distinguish between schema to another, one particular database to another

<MacTed> unmute me

Ahmed: then we want to match multiple databases from one mapping to another.

"The mapping language MUST allow for a mechanism to create identifiers for database entities. The generation of identifiers should be designed to support the implementation of the linked data principles. Where possible, the language will encourage the reuse of public identifiers for long-lived entities such as persons, corporations and geo-locations."

from charter

http://www.w3.org/2009/08/rdb2rdf-charter.html

<MacTed> mute me

angela: one way to solve this is to use OKKAM ID?

semantics of SPARQL 1.0?

so lets think of it as "best guess at this point in the experience of the WG" rather than "we have fulfilled this requirement of the charter by using particular technique X [where X could be OKKAM IDs, etc. etc.]

<angela_UNITN> i agree...

so lets have OKKAM IDs say, we fulfill this charter requirement, etc. and that graph expressivity may match it.

rather than MUST match it (personally worried about vendor-specific requirements)

ericP: I just want us to require as much as possible so the community can see if we are specifying the problem correctly

http://www.w3.org/TR/rdf-dawg-uc/

ericP: if we use languages that have semantics for sort of stuff like ID transformations
... like RIF, but maybe we don't want a language that leans on a second step
... then we need to couch the shared ontology in terms of relational schema
... here's the effect we want and lets be agnostic about what we want.

Ahmed: I like that, so far, but we need to set-up some engine

ericP: I just do transformations re language

<cygri> hhalpin, no my point has been made

ericP: not sure if group as a whole as bought into it, but want to push the ultimate semantic effect we want

MacTed: It seems to me that the whole concept of mapping IDs across databases goes way beyond of exposing RDB data to RDF
... figuring out whether or not is outside the scope.
... figuring out those rules and being able to change them as we go is part of what is wanted
... actually saying these ID in that schema is equivalent is outside of the schema

ericP: I think thats what we arrived at in discussing the tax case
... i.e. we have an extra mapping table with a simple join

notes that we need to give people an option to re-use identifiers, but this can be an option (i.e. use OKKAM, etc.)

MacTed: Doesn't this change things radically, we don't want to write to data?

ericP: Do you want a new requirement that is beyond graph expressivitiy, lookup, and simple joins?

MacTed: I don't think I'm asking for more than that
... was worried by Souri's argument about voluminous

juan: What EricP was going on, about the local ontology
... one thing is our mapping language

angela: we should have an option to have this mapping language

juan: I'm worried that would require lot of expressivity

angela: what I was saying that this maps a domain ontology to a local ontology
... we want option to map column of table to domain ontology

so you can imagine some people would want to control the URIs being generated on a more fine-grained level of detail rather than just a "genID" approach

and this could be provided by some direct control over the URI mapping of particular schemas in addition to the "voluminous" table identifiers

but the "linked automatically" part is

a "where possible" option, not a MUST requirement from the charter

daniel: you can imagine the linked data
... being either an identifier or a

<Souri> let us first solve the problem for one database, later if time permits we can try to tackle the mult-database aspects

<scribe> ... new one for the hub

+1 souri, but let's leave that option open in the use-case requirement, clearly specified as "where possible" and optional.

<Souri> I agree harry

it seems ericP own writing it

will help with pubrules and commonscribe mess.

Ahmed: lets try to finish on time so we can have more reviews

trackbot, end meeting

- DRAFT -

RDB2RDF Working Group Teleconference

30 Mar 2010

Attendees

Contents

Summary of Action Items

Scribe.perl diagnostic output