RDB2RDF Working Group Teleconference

01 Feb 2011


See also: IRC log


boris, juansequeda, mhausenblas, +1.636.544.aaaa, Ashok_Malhotra, souri, +1.284.612.aabb, Alexandre, +1.603.891.aacc, Seema, EricP, dmcneil, soeren, +886201aadd, Ivan, nunolopes
Marcelo, Percy


<trackbot> Date: 01 February 2011

<boris> who is here

<boris> so

<mhausenblas> scribenick: cygri

<mhausenblas> http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2011Jan/0080.html

1. Admin

<mhausenblas> PROPOSAL: Accept the minutes of last meeting, see

<mhausenblas> http://www.w3.org/2011/01/25-rdb2rdf-minutes.html

<soeren> +1

<boris> I+1

<juansequeda> +1

<betehess> +1

RESOLUTION: Accept the minutes of last meeting, see http://www.w3.org/2011/01/25-rdb2rdf-minutes.html

2. Direct Mapping status


<trackbot> ACTION-85 -- Ted Thibodeau to review Direct Mapping ED -- due 2011-02-01 -- OPEN

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/actions/85

<mhausenblas> MacTed?

<mhausenblas> drop ACTION-85

<mhausenblas> close ACTION-85

<trackbot> ACTION-85 Review Direct Mapping ED closed

<juansequeda> http://www.w3.org/2001/sw/rdb2rdf/directMapping/

<mhausenblas> http://www.w3.org/2001/sw/rdb2rdf/track/issues/raised

mhausenblas: let's go through raised issues for the direct mapping
... for each issue, let's get them from "raised" to "open", "postponed", "closed" etc

<juansequeda> http://www.w3.org/2001/sw/rdb2rdf/track/issues/9

<mhausenblas> ISSUE-9?

<trackbot> ISSUE-9 -- Generate Blank Nodes for duplicate tuples -- raised

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/9

juansequeda: Marcelo and I currently use a predicate that states the cardinality
... in section 4.1, 4.2
... we ask people to review this

betehess: let's be careful not to call this solved, it interacts with other issues i have in the queue

<mhausenblas> Michael: looks like we should change the issue 9 from raised to open

betehess: i don't understand how to play with the datalog rules in order to execute them

mhausenblas: let's not do very generic issues

<betehess> uri?

<juansequeda> http://www.w3.org/2001/sw/rdb2rdf/directMapping/

juansequeda: we added examples for 4.1 and 4.2
... that should help clarify. if there are still problems with this, please raise

ericP: test cases please

<mhausenblas> Michael: I encourage everyone to raise issues no matter if there exists something on not, as long as it is concrete enough, please go ahead

betehess: i don't recognize RDB or RDF in the rules

juansequeda: RDB model is just some predicates

betehess: i don't think it's well-defined. two solutions in the document that work with sort of different data models
... rdb model is more than just predicates
... in datalog you need to state facts. i don't see this in the current version
... it's not exactly datalog
... don't see how to go from rdb to rdf using this

ivan: the document says that we present rules in datalog *syntax*, so it may not use the formal mathematical model
... it's just a set of rules. that's fine
... we should show that both formalism lead to same triples from same db

<juansequeda> +1 to a lot of testcases

ivan: how to do that? a boatload of testcases

mhausenblas: yes we need test cases to proceed. how to get them?

ivan: we started with specific issues. put them in the issue list.
... juan, are there ways for getting those rules rolling in a specific system? the other formalism has an implementation

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

ivan: if we had a processor, we could run them side by side

mhausenblas: could this writeup be used as input for test cases? http://www.w3.org/2001/sw/rdb2rdf/wiki/All_Cases_for_Default_Mapping

juansequeda: just a list of cases that had to be managed
... rules don't need to be implemented. it's an if-then case

ivan: we have to prove they do the same thing. how else would we do that?

juansequeda: this just specifies some cases. if this, then that. doesn't have to be implemented using a rules engine. we don't expect people to do that

boris: we started with very simple test cases, want to get on to more complex ones

mhausenblas: probably we want to prioritize direct mapping test cases

<ericP> TaskAssignments

<ericP> 7 "pencil survey" "accounting" "Cambridge"

<ericP> NULL "pencil survey" "accounting" "Cambridge"

Souri: as a practitioner, my experience is that blank nodes cause lots of problems
... let's minimize generation of blank nodes

<soeren> +1

Souri: we need to generate a unique label anyways. then we could just make a URI from it. makes life easier for implementers

<ericP> note that the above table preserves UNIQUE(worker, project), but not PRIMARY KEY(worker, project)

Souri: bnodes are local symbols in a graph, so if you make two triples in different graphs from same row, they don't merge

<Zakim> ericP, you wanted to say that unresolvable identifiers create issues

ericP: we can't calculate the same blank node twice

<Ashok> He just wants to distinguish between blank nodes

ericP: blank nodes are unpleasant in rdf
... lack of commitment towards identifiers

<ivan> +1 to eric

ericP: bnodes should be used for identifiers that you're not going to remember for future responses

<mhausenblas> http://www.w3.org/2001/sw/rdb2rdf/track/issues/raised

mhausenblas: how should we proceed with the issues?

juansequeda: let's go through them one by one, open them


<Souri> I'd like the issue of use or non-use of bNodes to be raised as an issue so that we can discuss it later

soeren: rename it to "avoid blank nodes"?

<ericP> +1

<soeren> +1

<ivan> +1

<boris> +1


<mhausenblas> ISSUE-10?

<trackbot> ISSUE-10 -- Hash vs Slash -- raised

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/10

<ericP> open

ericP: there's this linked data fury going on
... long and contentious issue about uris
... have an rdf file that has a node in it that identifies what you're talking about
... and label that node with a relative hash URI
... or alternative: don't use a hash, but just a slash URI, and a GET does http redirect to the file URI
... one is clearer, one is simpler

ivan: i see direct graph as intermediate graph, where i use rules or something else to transform it into something i really like
... that's my impression. might be rat hole

mhausenblas: let's not resolve issues now, but just open them

<ericP> i think the distinguishing situation comes when there's a predicate which applies both to the record and the document describing the record



<mhausenblas> +1

<boris> +1

<ericP> +1


<betehess> +1

<mhausenblas> ISSUE-11?

<trackbot> ISSUE-11 -- Primary Key is a Candidate Key -- raised

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/11

<betehess> IMO "Common data modeling" should not be addressed by the Direct Mapping

<boris> FK is a PK?

juansequeda: this should read: "Primary Key is Foreign Key"


<mhausenblas> +1

<boris> +1

<ivan> +1


<mhausenblas> ACTION: Juan to rename Issue 11 and fix it in the DM [recorded in http://www.w3.org/2011/02/01-rdb2rdf-minutes.html#action01]

<trackbot> Created ACTION-98 - Rename Issue 11 and fix it in the DM [on Juan Sequeda - due 2011-02-08].

<mhausenblas> ISSUE-12?

<trackbot> ISSUE-12 -- Hierarchical Tables -- raised

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/12

ericP: related to previous issue. first one is: what happens if pk is fk? second is: assuming we consider pk=fk a hierarchical table, then how do we model it?


<betehess> /me agrees to keep it somewhere, not opened

cygri: what does POSTPONED mean?

mhausenblas: we may discuss it again if someone comes with new information

<juansequeda> +1

<betehess> +1

<boris> +1

<mhausenblas> +1


<ericP> cygri, would you propose that issue 11 be resolved as this?: primary keys which are also foreign keys are treated as if that foreign key doesn't exist

<mhausenblas> ISSUE-13?

<trackbot> ISSUE-13 -- Foreign Key and Primary Key rearrangement -- raised

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/13

<ericP> (i have no problem with that)

ericP: this is again dependent on ISSUE-11

juansequeda: can we merge ISSUE-11, ISSUE-12, ISSUE-13

ivan: edit issues manually to add references between them

PROPOSAL: change ISSUE-13 to POSTPONED; add refs from ISSUE-11 to ISSUE-13

<mhausenblas> +1

<juansequeda> +1

<betehess> +1

<boris> +1

RESOLUTION: change ISSUE-13 to POSTPONED; add refs from ISSUE-11 to ISSUE-13

<scribe> ACTION: Juan to change ISSUE-13 to postponed and add ref from ISSUE-11 to ISSUE-13 [recorded in http://www.w3.org/2011/02/01-rdb2rdf-minutes.html#action02]

<trackbot> Created ACTION-99 - Change ISSUE-13 to postponed and add ref from ISSUE-11 to ISSUE-13 [on Juan Sequeda - due 2011-02-08].

<mhausenblas> ISSUE-14?

<trackbot> ISSUE-14 -- Many-to-Many tables -- raised

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/14

<betehess> +1



<boris> -1

<soeren> cygri: many to many tables are important

betehess: my objection is that the direct mapping should be kept really simple, just expose the relational model, simple expression of that in the RDF domain
... if we do one case, then we have to do a lot of cases

ashok: i agree with keeping the direct mapping simple. q to richard: how do you figure out that something is many to many?

<betehess> Ashok, you can identify some patterns, not all

juansequeda: i agree with everybody
... let's keep it simple
... we could have a direct mapping plus optional features
... many to many is usually two-column table
... that's how tools translate uml to sql ddl

ashok: what if extra columns?

<ericP> another prob with a special case for binary relations is that not all such relations are MxN (e.g ConventiallyMarriedCouple(Wife, Husband))

<ericP> i think this should be an R2RML feature

juansequeda: then it's not a many to many

<betehess> ericP, just like the hierarchical pattern? ;-)


<mhausenblas> +1

<soeren> +1

<betehess> +1

<boris> +1


<Souri> +1


<mhausenblas> ISSUE-15?

<trackbot> ISSUE-15 -- Formalism -- raised

<trackbot> http://www.w3.org/2001/sw/rdb2rdf/track/issues/15

<ericP> betehess, yeah, i'm totally game to ignore foreign keys which are also primary keys

juansequeda: this is an issue already written in the document

<betehess> mhausenblas, did you say "either"?

mhausenblas: three options: datalog, set-based notation, plain english

ivan: the three options address different communities
... having both in the document is viable

PROPOSAL: change ISSUE-15 to OPEN, and rephrase to highlight the three options (datalog, set-based notation, plain english)

<betehess> I would say "consistent"


<mhausenblas> +1

<juansequeda> +1

<boris> +1

<ivan> +1

<betehess> +1

RESOLUTION: change ISSUE-15 to OPEN, and rephrase to highlight the three options (datalog, set-based notation, plain english)


ericP, do you have a second?

re the hierarchical table thing?

<ericP> cygri, sure

<juansequeda> gotta go guys, have fun geeking off

<mhausenblas> cheers, and thanks a lot juansequeda

<mhausenblas> trackbot, end telecon

Summary of Action Items

[NEW] ACTION: Juan to change ISSUE-13 to postponed and add ref from ISSUE-11 to ISSUE-13 [recorded in http://www.w3.org/2011/02/01-rdb2rdf-minutes.html#action02]
[NEW] ACTION: Juan to rename Issue 11 and fix it in the DM [recorded in http://www.w3.org/2011/02/01-rdb2rdf-minutes.html#action01]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.135 (CVS log)
$Date: 2011/02/02 09:35:59 $