RDB2RDF Working Group Teleconference

26 Apr 2011

See also: IRC log


Ivan, Percy, Alex, Juan, David, Ashok, Eric, Ted, Alexandre
Michael, Soeren, Souri, Marcelo


<trackbot> Date: 26 April 2011

<Ashok> meeting: RDB2RDF

<Ashok> chair: Ashok

<ivan> ok

<privera> +??P10 is privera

<privera> privera is ??P10

<privera> ??P10 is privera

<privera> Zakim mute me

<Ashok> scribe: David

<ivan> scribenick: dmcneil

<Ashok> scribenick: dmcneil

<ivan> last week's minutes

<Ashok> RESOLUTION: Minutes for April 19 are approved


ashok: need to discuss a face-to-face meeting

who can summarize where we are and what to do?

ivan: first question: do we need a f2f?
... if current timing is realistic then last call will be in summer time

candidate recommendation will need to be in early autumn (?)

ashok: seems reasonable

ivan: a f2f to handle last call comments is an efficient way of dealing with them

the answers/decisions need to be well documented

planning a f2f in autumn makes sense if there will be a lot of last call comments... but nobody knows that

TPAC is an obvious place to do a f2f: end of Oct in Santa Clara, CA

would need to register for TPAC by end of this week

problem is not many on the call today, nor last week, so hard to decide

<ivan> TPAC: Oct 31 - 01 or 03-04 November

<alexdeleon> I cannot confirm now

<privera> kamin, unmute me

who could come to TPAC for a f2f?

ivan: yes

juan: think so

david: think so

alex: timing is good, but cannot confirm right now

eric: at risk, has a conflict

<privera> I cannot confirm now, first I need to talk with the department

ivan: need to have at least one editor of both documents present

ashok: maybe book the room and cancel if needed?

ted: could maybe attend a TPAC f2f

ivan: maybe send out an email and give people to Thursday to object

ashok: will send out an email after the telecon

Action Items

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

ashok: will close ACTION-83

ivan: re ACTION-101, looked at SemtTech London, but profile of the conference seemed different than SemTech San Francisco, much more business people oriented

ashok: what about a database conference?

<ericP> enterprise data world was a couple weeks ago in chicago

<ericP> VDLB is 19 Aug

<ericP> http://kcap11.stanford.edu/

<ericP> KCAP 25-29 June

<Ashok> Eric will present paper on DM

juan: marcello will be presenting at SIGMOD/PODS in Athens in June

<juansequeda> Marcelo will be giving an invited tutorial at SIGMOD/PODS on RDF/SPARQL. He will give a brief intro to RDB2RDF

ashok: keep ACTION-101 open as an ongoing effort

ACTION-102 will be kept open

ashok: status of Action 102

ivan: from RDF working group columnization (?) issue is still open

working group will add to the semantics a sentence that says if a triplestore wants to columnize BNodes when exporting the data then we may have a proposed way of doing it

two schemes being discussed

1) on HTTP

<ivan> http://example.org/.known-uri/{SOMETHING}

<ericP> well-known

<ericP> iirc

<ivan> http://example.org/.well-known/genid/{SOMETHING}

ivan: could register the ".well-known/genid" part

<ivan> tag:example.org,2011:genid:SOMETHING

<ivan> skolemize

above the word "columnize" should be "skolemize"

<ivan> urn:genid:XXXXX

<Ashok> ... refers to Skolem functions as a means of generating names

ivan: might be good for this working group to express a need for something like the above

ashok: one of us could write something and send it to the RDF working group

eric: we need to document our use case

ashok: since you are on the RDF working group, could you draft a note for us?

eric: to summarize the issue: we want to express rows in a database that we know to be volatile

want to give them labels that we won't honor (i.e. resolve) but we want to talk about them outside of a particular graph

but why?

ashok: a requirement was to take parts of a graph and cut and paste it (e.g. combining graphs)

eric: but to go to the RDF working group we need a more precise statement of what we need to do

ivan: also need to make it clear why blank nodes cannot be used for the purpose
... once you skolemize they are not blank nodes

ashok: looking for one or two people who can draft something for the RDF working group

eric: can help refine it if someone else drafts it

needs clarification on use case

ted: a blank node is still a thing, if it does not have a portable identifier, then description data about it is unportable. The whole point of RDB2RDF is to make portable descriptions.

<betehess> not sure what "portable" means

ivan: administrative point: tomorrow the RDF working group may decide this and we don't need to do anything

perhaps the real issue is the need to generate URIs on the fly that are universally unique?

blank nodes were used until now because that was an easy solution

need a way to generate identifiers for HTTP, that allows them to be recognized as generated

generating unique identifiers is different than skolemizing blank nodes

so maybe we need to define a scheme for generating identifiers for rows (inspired by what the RDF working group is doing)

ashok: ted will get help from eric on ACTION-102

<Zakim> betehess, you wanted to ask how you compare these rows with such a mechanism

<ericP> { <s1> <p> 5 } == { <s2> <p> 5 } ?

alex: if we use a URI instead of a blank node then if we have two graphs, there is not way to identify them as representing the same data

since the URIs are just strings

<ericP> 'cuse { _:s1 <p> 5 } = { _:s2 <p> 5 }

ivan: why not use blank nodes in this case? then the graphs will be isomorphic

alex: that is why I want blank nodes

ivan: then we need to discuss why not to use blank nodes

ashok: move to open issues

Open Issues

<Ashok> http://www.w3.org/2001/sw/rdb2rdf/track/issues/open

ashok: email from david about closing ISSUE-25

ivan: the R2RML spec is silent on the topic of what to do if there is no mapping specified for an RDB entity
... the decision is what to do if there is no mapping specified

<ericP> if the DM is either the default behavior or some user-selected behavior, R2RML needs a reference to DM

ashok: we don't have a quorum to discuss this now

should ask the editors of the R2RML spec to clarify this

ivan: but what would the paragraph say?

two options:

1) if column is not mapped, then produce no triple

2) is column is not mapped, then use the Direct Mapping

<ericP> q

ivan: needs to be addressed by the working group, not simply by the editors

<juansequeda> It should be option 1

<Zakim> ericP, you wanted to ask if we want to allow user-selected DM

eric: R2RML needs a reference to DM if it is either the default behavior or if the user can invoke the DM behavior

ivan: this is all open, undecided

<betehess> so it's a tool feature, not a language feature

ivan: another possibility is to realize the DM as an R2RML mapping, and then modify it

eric: should only specify an R2RML mapping in DM if it is a tested part of the spec

oh seems i misunderstood erics point

eric: if users are to expect DM to be an option then R2RML spec needs to reference the DM spec (?)

ashok: what if I have a huge database and only want to map part of it?
... our time is up

<ivan> trackbot, stop telcon

<trackbot> Sorry, ivan, I don't understand 'trackbot, stop telcon'. Please refer to http://www.w3.org/2005/06/tracker/irc for help

<ivan> trackbot, end telcon

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2011/04/26 17:12:39 $