See also: IRC log
Previous: 2006-10-17 http://www.w3.org/2007/07/17-swd-minutes.html
<RalphS> Scribenick: edsu
<RalphS> Scribe: Ed
RESOLVED to accept minutes of 17 July telecon http://www.w3.org/2007/07/17-swd-minutes.html
<RalphS> Daniel: regrets for 7 Aug
RESOLVED next telecom 7 Aug, followed by 21 Aug
<scribe> ACTION: Guus to propose dates in Oct for Amsterdam meeting on SKOS [recorded in http://www.w3.org/2007/07/03-swd-minutes.html#action01] [DONE]
<scribe> http://www.w3.org/2002/09/wbs/39408/skos-ftf-amsterdam/
Guus: looks like 8th and 9th is
strongest preference
... suggests we keep poll open for a bit longer
-- Issue-26: RelationshipsBetweenLabels (http://www.w3.org/2006/07/SWD/track/issues/26)
Guus: had discussion with Antoine, seems to be the issue is to do with containment
aliman: agreement about containment
Guus: if i were propose we rely on general rdf mechanisms what would you think Alistair?
aliman: would have to think about
it a bit
... can i have concrete examples of trust in the context of
skos?
... lets say we don't address containment, and we dump various
taxonomies into a single store, then how could you display just
aat, or mesh, for example?
Guus: many tools support the idea
of quadruples
... maybe we are trying to solve a problem that creates more
problems than we can handle
... we can't solve the notion of named graphs
aliman: given the support there
is for named graphs in sparql, perhaps there are solutions
there
... it's definitely worth exploring
... understanding the relationship between traditional thesauri
and skos is important ... need a story of whether concepts are
contained within a scheme or not
... there are use cases where people will want to attach admin
information
... will send email
... if we look at British Standard Working Group who have a UML
model for thesaurus data, not dealing with open-world issues,
not a straight forward mapping
RalphS: could namespaces help
aliman: have we published any guidelines on what that means?
RalphS: formal semantics people
punted on it
... would be interested in what containment means in relation
to SKOS
aliman: difference between containment and aggregation in UML
<RalphS> Ralph: Alistair's description of the new UML model for thesaurii sounded like "containment" within a thesaurus is like rdfs:isDefinedBy for a namespace
<RalphS> rdfs:isDefinedBy
dlrubin: not sure these ideas fit with web architecture
Antoine: if concept scheme is defined with a namespace, there could be a problem with versioning
Guus: can't solve some of this stuff at the language level, from tbl
<RalphS> Daniel: in the SemWeb, deleting an ontology doesn't delete the instances
Guus: if you look at version contructs in owl there's not semantics
aliman: antoine's point about namespaces not being enough is valid
<RalphS> Daniel: the ontology owner has no control over the instances; if the ontology is deleted, the semantics for the instances just become harder to identify
<RalphS> Daniel: so SKOS has no control over deletion [of either ontology or instances]
<RalphS> Daniel: therefore, not clear that 'containment' idea fits in SemWeb architecture
Guus: I want a last call draft in 7-8 months, and if we take all this on we might not be able to finish it
<RalphS> [Daniel expressed that nicely, so I wanted it recorded :) ]
<scribe> ACTION: Guus to email a proposal to the list about the issue of containment [recorded in http://www.w3.org/2007/07/24-swd-minutes.html#action01]
-- ISSUE-38 CompatibilityWithOWL-DL
aliman: after discussion last week about skos semantics wiki draft made some minor edits to the skos semantics wiki draft
<aliman> http://www.w3.org/2006/07/SWD/wiki/SKOS/Semantics/Introduction
aliman: the basic idea is to use
the RDF compatible OWL semantics as the basis for skos
... if we do that then there is an issue about OWL DL, we have
users who would like to use SKOS within DL
... some of the semantics for SKOS are outside DL
<aliman> [OWL-Full]
aliman: currently next to any semantic condition that takes you out of DL we use the tag name 'OWL-Full'
dlrubin: everything required in
skos will be consistent with owl dl?
... what would the owl dl user be missing?
<RalphS> [?? I still see the text "DL-optional" in http://www.w3.org/2006/07/SWD/wiki/SKOS/Semantics/Introduction ]
<aliman> http://www.w3.org/2006/07/SWD/wiki/SKOS/Semantics/Labelling
aliman: if we look at the
labelling module as it is right now
... you get various things that aren't owl dl
... you'd lose some generalization inferences
dlrubin: subproperties aren't part of owl dl?
aliman: because they've been typed as annotation properties
<RalphS> Daniel: I'd rather keep subPropertyOf and drop Domain and Range
<RalphS> Alistair: another alternative is to type them as datatype properties, not as annotation properties
Guus: if we would go for my simple extension proposal one of the rationales there is that it would solve the compatibility issue
aliman: what we're talking about is how to mark some triples as optional
Guus: if you want to be in DL you
have to give up some of your freedom
... and people can live with that
RalphS: i'm more worried about people in the skos community being frustrated
aliman: Guus: are you saying that we should try to construct the semantics so that nothing is optional?
Guus: i missed some of the
discussion last week around that
... there seem to be other triples that need to be added to
make it OWL DL
aliman: we're not going to leave the type of prefLabel to be decided by an application are we?
Guus: the only way we can commit
would be to use something like my simple extension
proposal
... may need to think about this a bit more
RalphS: understandable given the
confusion
... in the wiki
aliman: the only way i could
think of to build the semantics docs so you could choose
dl/full
... as i understand it leaving that choice open is
valuable
... based on my understanding of the semantics
http://www.w3.org/2006/07/SWD/wiki/SKOS/Semantics/Labelling
http://www.w3.org/2006/07/SWD/wiki/SKOS/Semantics/Grouping
<RalphS> OWL-DL restrictions on the use of annotations
scribe: these semantics docs
could be used to go to rdfs, owl-dl or owl-full
... heavily influenced by the way rdfs is built up
Guus: why are data type properties used in the labeling module?
aliman: just a proposal
Guus: want to conclude discussion of SKOS
--skipping
<scribe> ACTION: Diego to write test to confirm what's broken about recipes as stated re: ISSUE 58 [recorded in http://www.w3.org/2007/07/24-swd-minutes.html#action03] [DONE]
<berrueta> http://lists.w3.org/Archives/Public/public-swd-wg/2007Jul/0130.html
diego: problems exist with regular expression
diego: not sure how to solve the problem
JonP: it looks like the best way to solve this is with a map
<RalphS> Diego: compare cases c and d -- the only different is the value of q, yet one works and the other doesn't
JonP: an alogorithm that matches the qs value on the server and the q value of the request
aliman: from last week i understood there 2 issues: 1. current implementations ignore q values ; 2. if you include more than one accept header you don't get a match
<RalphS> Ralph: perhaps case c works for the wrong reason; I'd like to see the tests repeated with no q values
<scribe> ACTION: Diego to repeat test without q values [recorded in http://www.w3.org/2007/07/24-swd-minutes.html#action02]
<RalphS> ISSUE 58
<RalphS> [I'm less concerned if the recipes don't account for q values but do work for multiple content-types]
--skipping
Guus: adjourn meeting, hope to speak in 2 weeks