See also: IRC log
<trackbot> Date: 19 January 2010
<MacTed> mhausenblas - Orri will not be able to scribe
<mhausenblas> scribenick: hhalpin
<juansequeda> i'll be dialing in a bit late
<scribe> scribe: hhalpin
<mhausenblas> aaaa is whalb
<scribe> ScribeNick: hhalpin
<mhausenblas> cygri is with mhausenblas
<mhausenblas> nunolopes is mhausenblas
<MacTed> Ted Thibodeau @ OpenLink
http://www.w3.org/2010/01/12-RDB2RDF-minutes.html
PROPOSAL: Accept http://www.w3.org/2010/01/12-RDB2RDF-minutes.html as the true and correct minutes?
<angela_UNITN> +1
<soeren> +1
ACCEPTED: http://www.w3.org/2010/01/12-RDB2RDF-minutes.html accepted as correct minutes
<mhausenblas> http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects
<Marcelo> I am trying to connect to the conference call, but it is not working ... let me try again ...
mhausenblas: have people tried to
look at this?
... this conforms us to our charter
... my interpretation of what the charter means that we need to
fulfill these principles
... we need to make sure we deal with Linked Data
principles
... and solve it via best practices
Ahmed: We need to define the scope of provenance so we don't go to far
mhausenblas: I phrased it a bit
stronger
... so that we MUST conform to Linked Data principles
<ericP> ack
<Zakim> ericP, you wanted to propose "Use URLs for tuples with a primary/unique key'
ericP: We can't use URIs if there's no unique key
Soeren: Sometimes we don't define unique keys but there are unique keys in principle
<Marcelo> I finally manage to connect :-)
ericP: SQL-aware or not, not just SQL
cygri: Regarding your first principle, what do you mean with schema-level entities having URIs or not?
mhausenblas: I mean the classes
cygri: but if the schema is somewhere else, then we will just re-use URIs of existing vocabularies
ericP: going up a
meta-level
... is linked data principles critical for use-cases
... linked data principles means there should be a HTTP URI,
ideally served.
... nice practice but not critical
... but this won't effect the use-cases
... so shouldn't we just use *names* that are consistent with
linked data principles
... but not indicate that people MUST make thoese
dereferencable
cygri: As long as you can do the
mapping that if you want to implement the URIs, this is
possible.
... we must be able to configure it so that we CAN put a http
URI there and a domain we can control.
... but we can't enforce that particular configuration.
ericP: that makes sense to
me
... system must be able to support names compatible with linked
data
Ahmed: There are lots of deployments that are enterprises
<mhausenblas> ??P5 is Orri
Ahmed: so we should not make them pay that overhead
cygri: We need to take care of 3
scenarios
... RDF dump, SPARQL queries, and Linked Data
... we should do all 3
ericP: the language should allow conformance to linked data.
<mhausenblas> ACTION: ericP to update http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Linked_Data_Principles to say 'MUST allow to..." [recorded in http://www.w3.org/2010/01/19-RDB2RDF-minutes.html#action01]
<trackbot> Sorry, couldn't find user - ericP
<mhausenblas> ACTION: eric to update http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Linked_Data_Principles to say 'MUST allow to..." [recorded in http://www.w3.org/2010/01/19-RDB2RDF-minutes.html#action02]
<trackbot> Created ACTION-20 - Update http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Linked_Data_Principles to say 'MUST allow to..." [on Eric Prud'hommeaux - due 2010-01-26].
<mhausenblas> http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Well-known_Entities
rcygri: Another case that works
quite easily
... is that I could put a URI into that column by using some
arbitrary algorithm
... but do we want to specify another more algorithm
soeren: we could identify a URI using a web service or RPC call
cygri: I could see that as a 3rd case, so that the function is out of scope of language, with arbitrary invocations of this functions.
+1 3rd choice
cygri: I'd like to see these 3 cases covered.
<ericP> c!
soeren: in one case we could just
specify namespace prefix
... and the other we could do else.
... I can volunteer to draft this up
<mhausenblas> ACTION: soeren to update http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Well-known_Entities to reflect discussion (3 cases) [recorded in http://www.w3.org/2010/01/19-RDB2RDF-minutes.html#action03]
<trackbot> Sorry, couldn't find user - soeren
<mhausenblas> ACTION: sauer to update http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Well-known_Entities to reflect discussion (3 cases) [recorded in http://www.w3.org/2010/01/19-RDB2RDF-minutes.html#action04]
<trackbot> Created ACTION-21 - Update http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Well-known_Entities to reflect discussion (3 cases) [on Sören Auer - due 2010-01-26].
angela: any kind of URIs should be allowed
<mhausenblas> http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Provenance
mhausenblas: we have some notes
on provenance
... we can liason with the provenance XG
Ahmed: There are many situations
where you want to track who did one
... who did what
... such as who changed the data from 93 to 100
... very interesting area
... very good work
Ashok: Why is it relevant to our work?
Ahmed: I said it is very
important, and we should at least chat about it.
... to see if we can look into how this.
but another group is
chartered to look into it, so we just need to set up a group meeting
cygri: when I think about
provenance in RDB2RDF
... I can imagine two levels
... 1) THis RDF came from this database applying these mapping
rules
2) Sometimes you have provenance info in database
cygri: in some application
specific format
... which we might want to translate to a specific
vocaublary
mhausenblas: Once we go rec
track
... this will be one the question
... no
... and we punt to XG
... or we directly see how it would be useful
<mhausenblas> http://sig.ma/
mhausenblas: thinking that we
could show how this mash-up from diff sources
... but we only trust sources from some kind of govt/etc.
... in that case
... we should allow R2ML allow us to provide this kind of
info
soeren: would like to use linked
data
... for relational data already solves some provenance
... so we can at least dereference URIs and this helps
alot
... and just enabling people to use some provenance
vocabulary
... and attached institutional data to this problem.
Ahmed: Henderson just came up
with some work
... but you want to know where we got this data
... especially in caching situations
?: there is also work done at Stanford
scribe: but no standard way to specify provenance
Note that there are multiple kinds of provenance
in Edinburgh, we are mostly interested in looking at the changes in databases
the "why" provenance
and also the "where" provenance when data comes from different graphs.
Irini Fundalaki has ported some of this work, using the idea of colouring, to RDF.
cygri: we could attach metadata
to the named graph as well
... talking about the case where we map into the log
... or a set of named graphs
<Souri> multiple graphs
cygri: map to a single graph or a set of named graphs
Ahmed: I think both
Irini's work
cygri: data model is a set of
graphs, not a single graph
... which do we use?
souri: we would need to
... relational data needs to get to multiple graphs
... so we would want to say this view goes to one particular
graph, i.e. assoicate a graph name with that
orri: we have to use own
graph
... so we can all of this graph
<Souri> there is a mapping between <a list of views> to an RDF graph
orri: so in the view paradigm, we can then get different graphs, even from an ETL
<Souri> so v1 and v2 may go to graph G1, then v3 and v4 may go to graph G2, and so on ...
orri: in some cases, we can
control how this mapping is applied
... so we want only some for subgraph.
... this is also what we do
... when we map on demand wihtout ETL
... we project graph out of some arbitrary column from the
table
... graph is some column, say owner of blog post
<ericP> <update18> a :DbUpdate, mysql:logUpdate ; :updater [ :username "ericP" ] ; :rowsAffected ( <row7>, <row9> ) ; :properties ( :salary, :office ) .
<Zakim> ericP, you wanted to show off a logged update doodle
ericP: I want to show off an alternative to named graph
soeren: is this related to
triplify's
... update approach
<mhausenblas> see also http://triplify.org/vocabulary/update
soeren: we have discussed
this
... but we were hoping to discuss this after we get the
mappingg
... the data is sufficient for lots of prov. use cases
... after a few people implement it we can look back into
this
mhausenblas: anyone want to update the wiki here?
rcygri: map to one graph or set
of named graph based on views
... or do we map statemwnts to graph dynamically
... ala orri's approach
Ahmed: this should be part of the two teams task
orri: in the view approach, it's straightforward
mhausenblas: what about this decision regardless of imlementation
cygri: mapping a database into
diff. named graphs
... then we solve some of the provenance problem
... as we can attach information to each named graph
Ahmed: don't see mapping
language
... would like to see it on higher level
cygri: see my previous 2
cases
... what i want
... is to represent both
<Souri> +q
cygri: as the latter can happen in mapping language
Ahmed: there is this term data
lineage
... we handle this with handling schema a certain way
... can not too much in existing schema to do go for rdf
mapping
... big projects already inside databases, dont want to get
lost
<mhausenblas> ACTION: ericP to add Ahmed's plan to the Wiki [recorded in http://www.w3.org/2010/01/19-RDB2RDF-minutes.html#action05]
<trackbot> Sorry, couldn't find user - ericP
<mhausenblas> ACTION: eric to add Ahmed's plan to the Wiki [recorded in http://www.w3.org/2010/01/19-RDB2RDF-minutes.html#action06]
<trackbot> Created ACTION-22 - Add Ahmed's plan to the Wiki [on Eric Prud'hommeaux - due 2010-01-26].
Semtech face-to-face?
<angela_UNITN> +1
mhausenblas: we'll keep looking
<ericP> +1 to semtech (recognizing that the real decision is in doodle)
Souri: My only curiousity is that
we are already
... talking about this
<Ahmed> Bye
Yolanda Gil is the chair
<Souri> not me
s/Juan/Souri
<Juan> harry that was not me
<mhausenblas> ACTION: juansequeda to contact XG provenance [recorded in http://www.w3.org/2010/01/19-RDB2RDF-minutes.html#action07]
<mhausenblas> ACTION: cygri to update http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Provenance re today's discussion [recorded in http://www.w3.org/2010/01/19-RDB2RDF-minutes.html#action08]
<trackbot> Sorry, couldn't find user - cygri
Souri: database capturing
provenance
... that should be covered by mapping
... but am not sure
... if we want to do something special
cygri: we might want to map
certain columsn to graph names
... that is what I think we should o.
Souri: If we define a view
... to be a single poriton
... then we create a smaller view
... with only some slice
cygri: but what about 10,000 users, 10,000 views
<mhausenblas> ACTION: rcygania2 to update http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Provenance re today's discussion [recorded in http://www.w3.org/2010/01/19-RDB2RDF-minutes.html#action09]
<trackbot> Created ACTION-23 - Update http://www.w3.org/2001/sw/rdb2rdf/wiki/LinkedDataAspects#Provenance re today's discussion [on Richard Cyganiak - due 2010-01-26].
Meeting adjourned.
trackbot, end meeting
This is scribe.perl Revision: 1.135 of Date: 2009/03/02 03:52:20 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Succeeded: s/mhausenblas: As long/cygri: As long/ Succeeded: s/rcygri/cygri/ FAILED: s/Juan/Souri/ Succeeded: s/Souri/Juan/ Found ScribeNick: hhalpin Found Scribe: hhalpin Inferring ScribeNick: hhalpin Found ScribeNick: hhalpin WARNING: No "Present: ... " found! Possibly Present: ACCEPTED Ahmed Ashok Ashok_Malhotra IPcaller Juan MacTed Marcelo OpenLink_Software Orri P13 P16 P26 P5 PROPOSAL RDB2RDF Seema Soeren Souri aabb aacc aaee aaff aagg angela angela_UNITN cgi-irc cygri ericP hhalpin iv_an_ru joined juansequeda li_ma mdean mhausenblas nunolopes rcygri scribenick trackbot whalb You can indicate people for the Present list like this: <dbooth> Present: dbooth jonathan mary <dbooth> Present+ amy Agenda: http://lists.w3.org/Archives/Public/public-rdb2rdf-wg/2010Jan/0030.html Found Date: 19 Jan 2010 Guessing minutes URL: http://www.w3.org/2010/01/19-RDB2RDF-minutes.html People with action items: cygri eric ericp juansequeda rcygania2 sauer soeren[End of scribe.perl diagnostic output]