See also: IRC log
- Carsten Lutz, Peter Patel-Schneider, Michael Schneider, Michael Smith, Ivan Herman, Rinke Hoekstra, Markus Krötzsch, Boris Motik, Uli Sattler, Jeremy Carroll, Achille Fokoue, Bernardo Cuenca Grau, Sandro Hawke, Alan Ruttenberg, Bijan Parsia, Ian Horrocks, Zhe Wu
- Martin Dzbor, Doug Lenat, Evan Wallace, Jim Hendler, Deborah McGuinness, Ratnesh Sahay
- Ian Horrocks, Peter Patel-Schneider
- Zhe Wu
PROPOSED: accept minutes http://www.w3.org/2007/OWL/wiki/Teleconference.2008.02.13/Minutes
RESOLVED: accept minutes of http://www.w3.org/2007/OWL/wiki/Teleconference.2008.02.13/Minutes
Pending review actions
action 87 completed.
action 88 completed (punning and owl lite)
action 89 completed--Investigate QA group advised recently to have "as few parts as possible"
Due and overdue actions
action 43 unlikely to happen soon
action 86 Send proposal for issue-91 ontology property
Ian Horrocks: jjc to update due date
Discussion: Fragments and Conformance (Continued)
Ian Horrocks: can we get a working agreement
Ian Horrocks: Jim is not here.
Ian Horrocks: everyone see the email from Alan?
Ian Horrocks: common understanding of fragments
Ian Horrocks: too many fragments is not a good idea
Ian Horrocks: Alan's original email: OWLPrime, EL++, DL Lite for rec track
Achille Fokoue: just sent an email. IBM's position is have limited # of fragments. not proposing a new one. need something that suits customer's needs. EL like seems to be nice. ... like OWLPrime rule based approach, not sure if it will scale though
Achille Fokoue: like OWLPrime because it is simple, also care about Alan's 3 point for real scalability, DL Lite + transitivity, translate to pure sql queries, close to Alan's position
Jeremy Carroll: consider tradeoffs
Jeremy Carroll: A Box scalability is a good measure
Jeremy Carroll: rationale is limit # of fragments. should have clear commercial interests
Peter Patel-Schneider: to HP, whether REC-ness is vital part for way forward?
Ian Horrocks: does it really matter for a fragment to be in rec or note
Jeremy Carroll: not sure.
Achille Fokoue: does matter. encourage tool and people to adopt
Jeremy Carroll: a rec should focus on interoperability
Ian Horrocks: commercial support on DL fragments is tricky
Ian Horrocks: do we count implementation of DL as implementation for fragments?
Zhe Wu: OWLPrime is indeed scalable. Oracle tested inference on LUBM8000 (with over 1 billion triples) on a simple desktop machine
Alan Ruttenberg: owlprime complements DL lite
Alan Ruttenberg: important to have one ties closely to completeness
Boris Motik: requirement on at least 2 interoperable implementations
Boris Motik: not sure if owlprime has 2 (or if they interoperate)
Boris Motik: even entire OWL FULL does not have such implementations
Uli Sattler: can we postpone rec track decision. ... rather, make clear what exactly is scalability... and how incomplete... be happy to study owlprime for example
Ian Horrocks: is the decision to postpone all fragments?
Ian Horrocks: or just the one (or ones) that is not totally clear
Achille Fokoue: second Uli's suggestion
Achille Fokoue: no rush here. Different levels of conformance, completeness, not totally clear
Alan Ruttenberg: comment on Uli's suggestion. concern about time line. OWLPrime fits into two spots: 1) rule based owl. need to specify clearly; 2) scalable. hard to imagine to have a second fragment like this. ... timeline is a factor to be considered.
Carsten Lutz: if we rec track something, we should do a good job
... try to have maximal DL fragment and FULL fragment
... OWLPrime, the benefit seems to be scalability
... hard to maximize it
Ian Horrocks: I don't think that WG can do a lot of research... fragments have to be based on what we know, instead of more research
Boris Motik: 1) a strong case for rule bases OWL reasoning.... seems to be DLP anyway... Achille's requirement should be accormodated by DLP... believe we should go to maximal subset... if vendors think it is hard to implement, we should not worry much.
Ivan Herman: we should worry
Bijan Parsia: production quality implementation.... more importantly, inter-operable implementations.
Jeremy Carroll: agree with bijan totally
Jeremy Carroll: maybe not research, but some quick check
Jeremy Carroll: on DL Lite fragment
Ian Horrocks: I did not mean not to do any investigation
Ian Horrocks: do not want lengthy research
Jeremy Carroll: sanity check DL Lite in the context of OWL 1.1
Ian Horrocks: anybody?
Uli Sattler: don't think a simple sanity on DL Lite is sufficient
Ian Horrocks: it could be that we cannot add anything.
Ian Horrocks: do we think it makes sense now to make decisions on fragments?
Ian Horrocks: make some provisional decision on which fragments
Ian Horrocks: first one OWLPrime. maybe we want to rename?
Ian Horrocks: don't worry about names for now
Ian Horrocks: do we think it is a good candiate for rec track
Ian Horrocks: OWLPrime folks need to do more convince work
Bijan Parsia: need a proposal before make a decision
Ian Horrocks: want to see a proposal for rule based fragment like OWLPrime before go forward
Ian Horrocks: EL++
Ian Horrocks: DL Lite kind of fragment? Scalable Abox space?
Ian Horrocks: what about DLP?
Ivan Herman: how much additional to merge DLP and OWLPrime
Ian Horrocks: yes it involves quite some work.
Ian Horrocks: it was an important link to make
Ian Horrocks: who think it is important to have a correspondance between rule based fragment and DL fragment
Boris Motik: to answer ivan, with minor work, we can come up with a clear definition when these two fragments coincide... this is something we really should look into... don't believe it is a lot of work.... pD* is not that far away from DLP anyway
Bijan Parsia: is the correspondence a requirement?
Ian Horrocks: it is an objective
Ian Horrocks: do we cover all fragments (for rec track candidates)?
Ian Horrocks: time is up. follow up with emails... in particular, we need specification for rule based fragment
Ian Horrocks: hand over chair to pfps
Peter Patel-Schneider: go over issues list
two proposals. most current one from bmotik
Boris Motik: allow bnode in funtional spec... could use freely just like in rdf... regarding semantics. could interpret them as unique, or through existential
Alan Ruttenberg: isn't this what we had before?... what is the difference
jjc, can you please put a summary in IRC?
ACTION: jjc send out email clarify
ACTION: jeremy to summarise problem with bnodes in ISSUE-3 vs bnodes in OWL list
want to defer because of the action item?
issue 95 No compatibility restrictions between the datatype being restricted and the facets in the DatatypeRestriction construct
Boris Motik: DatatypeRestriction takes data types
Boris Motik: change syntax a little bit
Peter Patel-Schneider: looks like this can be resolved next week.
ACTION: bmotik2 to Update the spec to resolve ISSUE-95
Boris Motik: can I update it now and send out an email for the resolution
issue 68: mapping rules are non-monotonic
Peter Patel-Schneider: I don't quite understand.
Jeremy Carroll: is a problem for Jena
Bijan Parsia: a proposal may solve the problem
Peter Patel-Schneider: this is not about QCR,
Peter Patel-Schneider: to m_schnei, this is a different issue
Peter Patel-Schneider: QCR is a separate issue
Jeremy Carroll: why it is a problem for Jena
Jeremy Carroll: knowledge is stored in triples
Jeremy Carroll: as you add more knowledge to Jena model, you don't retract triples... it is a fundamental design
Boris Motik: if Jena is a triple based, do we really need to retract triples?... I don't see why these mapping rules will require you to retract triples?
Jeremy Carroll: not clear
Peter Patel-Schneider: bijan proposed to do something?
Bijan Parsia: has no clue about the problem. not sure which email is relevant.
Peter Patel-Schneider: take an action to write up an email summarizing... my understanding of this problem
ACTION: pfps to Write up his understanding of ISSUE-68 (nonmonotonic mapping rules)
ACTION: patelschneider to Write up his understanding of ISSUE-68 (nonmonotonic mapping rules)
ACTION: peter to Write up his understanding of ISSUE-68 (nonmonotonic mapping rules)
ACTION: patel-schneider to Write up his understanding of ISSUE-68 (nonmonotonic mapping rules)
Peter Patel-Schneider: defer it
issue 69 punning is incompatible with OWL Full
Boris Motik: it is incompatible with owl full semantics.... in owl 1.0, DL and FULL are not completely aligned.
Boris Motik: don't see it as a show stopper
Bijan Parsia: have all sorts of punning in pellet... encountered no problems so far
Bijan Parsia: able to handle more graphs is really important... don't think punning is a really issue
Alan Ruttenberg: change the way owl dl works.
Peter Patel-Schneider: close to our time