Warning:
This wiki has been archived and is now read-only.
Teleconference.2007.11.21/Minutes
See also: IRC log
Contents
- Present
- Bijan Parsia, Rinke Hoekstra, Michael Smith, Markus Krötzsch, Martin Dzbor, Ian Horrocks, Uli Sattler, Alan Ruttenberg, Boris Motik, Sandro Hawke, Zhe Wu, Vit Novacek, Bernardo Cuenca Grau, Peter Patel-Schneider, Alan Ruttenberg, Carsten Lutz, Evan Wallace
- Regrets
- James Hendler, Ivan Herman, Ratnesh Sahay, Joanne Luciano
- Chair
- Ian Horrocks
- Scribe
- Rinke Hoekstra
(Scribe changed to Rinke Hoekstra)
Admin
Roll Call
Ian Horrocks: covered Roll call
Agenda Amendments
Ian Horrocks: any amendments?
Ian Horrocks: no amendments
(Previous) Previous Minutes
PROPOSED: Accept Previous Previous Minutes
Evan Wallace: previous previous minutes list me as evan wallbace
Sandro Hawke: accept them pending the change
PROPOSED: Accept previous previous minutes modula fixing Evan's name
Sandro Hawke: fixed them
RESOLVED: Accept previous previous minutes module fixing Evan's name
PROPOSED: Accept Previous Minutes
Sandro Hawke: haven't been there long enough to be reviewed
Ian Horrocks: who has reviewed them?
Ian Horrocks: leave them to next week
Action Items
Ian Horrocks: action item status?
Sandro Hawke: still mulling it over
Ian Horrocks: move to next week
Alan Ruttenberg: did the syntax document (Action 15)
Ian Horrocks: Action 17 qname approval?
Michael Smith: email has been sent, asking about schema components status
Ian Horrocks: Action 18, brain dump?
Alan Ruttenberg: haven't dumped, postponed to next week
Agenda for 1st F2F Meeting
Ian Horrocks: f2f agenda
Ian Horrocks: have collected a list of items to be discussed at the f2f, please look at the agenda
Ian Horrocks: currently on day 1 an overview of language features, imports
Ian Horrocks: two other topics dl and full alignment
Ian Horrocks: if you have comments, topics, then you might want to suggest them
Ian Horrocks: the schedule is going to be quite tight!
Ian Horrocks: cannot guarantee that every topic will be added
Alan Ruttenberg: where did you put the agenda?
Ian Horrocks: link from meetings page, in the irc as well
Alan Ruttenberg: if you feel that any knowledge needs to be made clear, we can put it in the tutorial
Ian Horrocks: over the course of the next week we will fix the agenda?
Evan Wallace: would like to discuss the Declaration
Evan Wallace: just the discussion of it
Bijan Parsia: what's the tutorial?
Alan Ruttenberg: orientation, looking who's coming, having a walkthrough of the specs, and doing a Q&A
Alan Ruttenberg: if you have ideas of what should be covered, let us know
Bijan parsia): is that the same as Overview of language features?
Alan Ruttenberg: yes
Ian Horrocks: are we done?
Ian Horrocks: alan and I will prepare a more concrete agenda
2nd F2F meeting
Ian Horrocks: 2nd F2F meeting
Ian Horrocks: quite a few have commented already
Ian Horrocks: any people on the phone who would like to add their response before we come to a decision?
Ian Horrocks: uli, boris, bernardo... you didn't respond yet?
Ian Horrocks: any preference Washington, Bejing, Sydney
Jeff Pan: beijing
Alan Ruttenberg: will put the raw responses on the voting list
Ian Horrocks: it is overwhelming for Washington... two people Beijing, one Sydney, everybody else Washington
Ian Horrocks: it's a no brainer... almost unanimous for Washington
Ian Horrocks: peter can go ahead with the OWLED colocation thing? That's my proposal
PROPOSED: F2F2 will be in Washington DC
Ian Horrocks: lots of agreement, no disagreement
RESOLVED: F2F2 will be in Washington DC
Ian Horrocks: finished with the admin
Alan Ruttenberg: if Conrad could put his work on the cleanups on the wiki
Task Forces reports and/or Discussions
Ian Horrocks: task forces stuff
Ian Horrocks: it would be useful to get some feedback from them
Ian Horrocks: jeremy to say something about the UFDTF
Jeremy Carroll: let me think
Jeremy Carroll: can I go second?
Rich Annotations
Ian Horrocks: Rich annotations? Bijan?
Bijan Parsia: I wrote up a proposal, some use cases, not finished
Bijan Parsia: I ran them by several people, e.g. Alan Rector
Bijan Parsia: somebody wanted to use them for patterns (value partitions)
Bijan Parsia: that's fine in my proposal, haven't written it up
Bijan Parsia: (Jeremy?) using reification for axiom annotations
Bijan Parsia: I guess in that exchange with Boris one option would be to ... the annotations
Bijan Parsia: would remove a lot of the usefulness of annotations
Bijan Parsia: proposal is 90% done...
Bijan Parsia: would be nice to get some sense of what we're going to do with it
Ian Horrocks: there is a fairly wellformed proposal on the table
Ian Horrocks: please give some comments
Alan Ruttenberg: concerned about the multiple domains of interpretation/ multiple worlds/ multiple models aspect of it
Alan Ruttenberg: part of the concern was that it's fairly new, and unexplored... want a warm fuzzy feeling that it's well understood
Alan Ruttenberg: second issue is that I don't know whether and how it would affect the RDF semantics
Alan Ruttenberg: could you comment on how it affects the OWL Full aspect of it, in the RDF sense
Bijan Parsia: I have not realised an RDF mapping
Bijan Parsia: one could use different extensions: multiple files with pointers analogous to owl:imports
Bijan Parsia: minimal mutilation of everything, and easy to understand
Bijan Parsia: other than that: is that enough to get you going?
Alan Ruttenberg: yes, just to say where I am on it: how much of the proposal is dependent on that
Alan Ruttenberg: is that the central part of the proposal? If it's left out, what will be left?
Bijan Parsia: it's a major part... another part is the idea to have blobs of annotations instead of nested or chained
Bijan Parsia: in the current annotations have to booted in the axiom
Bijan Parsia: a lot of people I talked to want to be able to associate fairly elaborate structures to an entity or axiom
Ian Horrocks: until more people have had a chance to have a look at that... we might as well leave that
Bijan Parsia: suggest that it becomes an official agenda item
Ian Horrocks: action item to go and read it?
Bijan Parsia: if I know its' going to be on the agenda I could (re)start an email discussion about this
Sandro Hawke: no need to do an action item
Bijan Parsia: if its on the agenda, I will start a discussion
(it was on the agenda for today)
Ian Horrocks: why don't you start a discussion, and if there's a significant response to that we can put it on the agenda
Alan Ruttenberg: who are the stakeholders on this?
Alan Ruttenberg: and have an action item to review this?
Alan Ruttenberg: for them (that holds for me)
Ian Horrocks: anyone else?
Alan Ruttenberg: more strongly, if you think it's important and you might object, you're a stakeholder
Ian Horrocks: why not let jeremy speak, and let people mull over whether they are stakeholders
Jeremy Carroll: by having the OWL Full semantics in multiple documents is very complicated
Jeremy Carroll: far away from owl 1.0
Jeremy Carroll: out of court. More about 'lets' put some wacky stuff in' instead of a conservative improvement over 1.0
Ian Horrocks: you're talking yourself into being a stakeholder
Jeremy Carroll: just articulating how a large part of the OWL community would feel about this move away from 1.0
Jeremy Carroll: this seems a big change from 1.0 in terms of the OWL Full semantics
Jeremy Carroll: sounds like a non starter (?)
Alan Ruttenberg: by putting it into different files, we are just making things independently and unconnected
Alan Ruttenberg: each separately is an OWL Full document
Alan Ruttenberg: that correct bijan?
Jeremy Carroll: strikes me as a 'big' change (in the eye of the beholder)
Jeremy Carroll: mouses and elephants
Ian Horrocks: clearly needs to be discussed in some more detail
Ian Horrocks: perhaps jeremy and bijan could exchange some emails on this topic?
Ian Horrocks: then the rest of us can eavesdrop
Ian Horrocks: how would that be?
Jeremy Carroll: not very enthusiastic, at least I get payed for it
Ian Horrocks: I'll take that as a yes
Datatypes
Ian Horrocks: move on to datatypes..
Ian Horrocks: a lot of email traffic, no taskforce...
Alan Ruttenberg: summarise where we came to in the first meeting
Michael Smith: two types of external datatypes... those that include ID's which can be externally referenced, and those that can't
Michael Smith: on the second we are waiting on the XML Schema WG
Michael Smith: moving forward on the approach outlined in the best practices document for the case where we do have IDs
Ian Horrocks: what kinds of things wouldn't we be able to do without ID's
Michael Smith: perfectly valid xml schema exists without IDs... if you want to reuse such schema types you have a problem
Michael Smith: on the semantic web
Michael Smith: if the author does have the semweb in mind, we have a solution.
Michael Smith: from the best practices
Ian Horrocks: we won't lose expressivity, just some rework
Michael Smith: yes, about reusing existing XML schema definitions
Ian Horrocks: unary datatypes?
Michael Smith: yes, the external datatypes only discussed about unary datatypes
Michael Smith: one other thing, that we discussed
Michael Smith: don't know if we want to put that in to a proposal first or
Ian Horrocks: try us with the other item first
Michael Smith: inline xml as opposed to what is in the member submission
Michael Smith: I think it hit some kind of completion point
Michael Smith: we understand what could be done,
Michael Smith: if we want to use xml schema types inline, ...
Michael Smith: we need feedback from people about this
Ian Horrocks: another case where people need to have a look and get feedback
Ian Horrocks: negative with respect to which option?
Alan Ruttenberg: I was advocating taking this as our first point strictly from a simplicity point of view
Alan Ruttenberg: doesn't require any vocabulary, might reduce the load at the expense of annoying but sufferable software adaptations
Ian Horrocks: first bijan, then jeffP
Bijan Parsia: I originally was thinking that using the XML schema syntax would be useful even with its limitations
Bijan Parsia: main argument against it: if we use XSD it restricts us in how we...
Bijan Parsia: would be apply new datatypes such as rational
Bijan Parsia: we could change it, but XSD guys might not like that
Bijan Parsia: unary datatypes with definitions, and totally different for n-ary: two different syntaxes
Bijan Parsia: if we do it ourselves
Bijan Parsia: some advantages if we do it ourselves, otherwise we have to take the XSD WG into account
Bijan Parsia: if we make our home-made one it would probably be more uniform and more under control
Ian Horrocks: sounds like a convincing argument
Jeff Pan: question for mike... can we see the unions as well?
Michael Smith: there hasn't been a proposal to include unions no
Jeff Pan: if we don't have unions for datatypes. Won't we have too complicated datatypes
Jeff Pan: users might really want to reuse existing datatypes that they have
Ian Horrocks: if people don't use unions, then ...
(sorry missed that)
Alan Ruttenberg: might at least support the XML syntax. At least vs. might
Ian Horrocks: if people don't use unions, then the kinds of datatypes they would be defining would be very simple, then not referencing
Ian Horrocks: external datatypes using wouldn't be so much of an issue
Ian Horrocks: Mike was saying something about a proposal
Michael Smith: proposal for modifying structural specification
Ian Horrocks: that must mean everyone else is on board as well
Ian Horrocks: this would need to be formed into a proposal that we might resolve
Alan Ruttenberg: is there a section that we could read, a section that says everything about it?
Michael Smith: the alternative would be to insert text into the structural document
Ian Horrocks: add it to the document, and then come back with a proposal to accept the edit in the document
Alan Ruttenberg: please comment in sandro's format
Ian Horrocks: peter objects? but I guess that when you come back with the proposal to accept your editing...
Peter Patel-Schneider: the message stands for itself
Peter Patel-Schneider: owl 1.1 already has its own syntax solution which eliminates the need for hacked-up xml schema documents
Alan Ruttenberg: there is our own syntax for defining datatypes?
Alan Ruttenberg: a requirement is to be able to reuse other people's documents
Peter Patel-Schneider: there's a vanishingly small number of xsd documents in that form
Alan Ruttenberg: bijan?
Bijan Parsia: not so much about reusing... I would be perfectly reasonable to define my own datatypes in xml schema.
Bijan Parsia: I have reasons for doing them inline sometimes, and for doing them outline in some cases
Jeremy Carroll: the data may well be useful for an owl or webservice application
Jeremy Carroll: since the rest of the world reads xml datatypes already, we might as well do it ourselves as well
Jeremy Carroll: yes, it's awful, but that's life
Ian Horrocks: bijan and jeremy are saying that there isn't actually a huge cost to it?
Ian Horrocks: we just need to add id's
Alan Ruttenberg: the documents need to be understood by the reasoners that support them
Ian Horrocks: the added overhead for implementers is that they need to add support for xml datatypes
Peter Patel-Schneider: you are going to require them to do that
Michael Smith: if you do reference them, you do it in the required way
Alan Ruttenberg: add a delimited specific set of datatypes... but implementers could hook into the syntax to support more complex datatypes
Michael Smith: owl 1.0 docs are a little bit inconsistent (integer + string, another doc has a longer list)
Michael Smith: we are not really clear on what's required for implementers and what's not
Michael Smith: maybe clarifying that is important
Ian Horrocks: a side chat between boris and jeff, trying to clarify whether we are only talking SimpleTypes
Ian Horrocks: I was presuming we're talking about more complex types, not unions, but facets
Michael Smith: when I was responding to Jeff, I was talking about inline... now we're talking about external datatypes
Michael Smith: I don't think there was a proposal to restrict what kinds of external xsd datatypes are allowed
Ian Horrocks: reuse the external xsd in your owl file
Jeremy Carroll: complextype and simpletype are technical terms
Jeremy Carroll: a simpletype can in fact be very complicated
Ian Horrocks: complicated rather than complex them
Ian Horrocks: does this answer boris' questions?
Ian Horrocks: perhaps Boris could say whether he understands everything now?
Bijan Parsia: we need to be a little bit careful wrt the types you are about to reference are types that are 'allowed'
Bijan Parsia: presumably no current reasoner can do anything with it
Bijan Parsia: the type that you reference must be definable in the inline syntax as well
Bijan Parsia: qnames wouldn't be types you can use and dereference from
Ian Horrocks: clarified a lot for me
Ian Horrocks: action on Mike
Ian Horrocks: to make a change that encapsulates his proposal
UFDTF
Ian Horrocks: skipped over the user facing documents
Jeremy Carroll: three telecons so far
Jeremy Carroll: quite a lot of disagreement, the main point of the telecons was sharing and exchanging views
Jeremy Carroll: agreement on different users prefer docs in terms of their domain and use cases
Jeremy Carroll: risk that we like to produce more docs than we could
Jeremy Carroll: some members are keen on docs that represent user communities
Jeremy Carroll: another opinion is that we should produce very little
Jeremy Carroll: another point of disagreement is to do with what the overview should look like
Jeremy Carroll: input is owl 1.0: brief summary + list of constructs
Jeremy Carroll: owl 1.1 summary which is much briefer
Jeremy Carroll: not anywhere near resolving anytime soon
Jeremy Carroll: a further issue is that non-wg members have interesting work that they would like to contribute to the group
Jeremy Carroll: how much can we interact with people not on the wg?
Ian Horrocks: not really a constraint on interaction, but if the documents reflect a lot of their input, then their names could not be on the docs
Vipul Kashyap: sent an email to michel and ... as to what their expectations are
Ian Horrocks: fair to say, ongoing, progress is made, but significant issues?
Jeremy Carroll: for the overall taskforce, yes, I think we're making disappointingly small progress so far
Vipul Kashyap: one thing we did agree on is to identify the set of users that would
Vipul Kashyap: be targeted by these docs
Vipul Kashyap: CIO's Enterprise Architects, developers: list on an external website
Ian Horrocks: we'll be discussing this more on the F2F... hopefully more progress on these issues then
Issues
Issue 14
Ian Horrocks: issues ... first is Issue 14: CURIES
Sandro Hawke: CURIES as in Marie Curie
Alan Ruttenberg: peter and I had a brief discussion about our original decision to use SPARQL
Alan Ruttenberg: peter thought it was limited, suggested to go for full CURIES instead
Alan Ruttenberg: no problem: a dependency as they're not a standard yet
Alan Ruttenberg: if others think that CURIES are good, then I'm happy too
Ian Horrocks: are we in a position that we could resolve
Sandro Hawke: the state of CURIES, we don't know they're ever going to be a rec
Peter Patel-Schneider: we could pull the stuff out of the CURIE spec
Alan Ruttenberg: someone from the group suggested we might monitor this
Alan Ruttenberg: and do a fallback when it turns out to not become a rec
Sandro Hawke: in RIF there was some discussion, in the end decided not to... maybe because it turned out to be not relevant... (don't remember exactly)?
Bijan Parsia: if CURIES are not legal element names, then how would this effect the RDF serialisation
Bijan Parsia: probably would require additional processing
Alan Ruttenberg: In my application(s), if the local part is not a valid part of a qname then it's expanded to an escaped uri
Alan Ruttenberg: I don't create extra namespaces, I fully quote them (in angle brackets)
Jeremy Carroll: certain uris that might occur as property uris can not be serialised as RDF uris
Jeremy Carroll: in alan's cases we're talking about subject/object uris, in that case no problem
Jeremy Carroll: in predicate position some curies cannot be serialised in RDF/XML
Jeremy Carroll: this a known limitation in RDF/XML, and we should ignore it
Bijan Parsia: that's fine jeremy, that's what I wanted to know
Bijan Parsia: if that's ok with you, then I'm fine with it
Jeremy Carroll: RDF/XML went to REC with this as a known bug
PROPOSED: to base abbreviated URIs on CURIES not QNAMES
Alan Ruttenberg: you can already use an arbitrary URI already in OWL 1.1 which cannot be serialised as RDf
Jeremy Carroll: not an OWL problem, an RDF problem
Sandro Hawke: is this just about the abstract syntax?
Alan Ruttenberg: only effect the functional syntax
Ian Horrocks: structural syntax
Alan Ruttenberg: functional style syntax
PROPOSED: to base functional style syntax abbreviated URIs on CURIES not QNAMES
Jeremy Carroll: have you seen my notes? I might change my vote tomorrow by email
Ian Horrocks: what's the best way forward? Resolve this now, backtrack? Push it to next week?
Ian Horrocks: put it on the agenda for next week. You'll be in a position to give a definitive yay or nay
Ian Horrocks: we resolved that we will resolve something
Issue 12
Ian Horrocks: Issue 2 is the final thing on the agenda
Ian Horrocks: we already resolved that we fix up this problem for alldisjoint wrt the RDF mapping
Ian Horrocks: the question arose as to whether we make the language symmetrical and add constructs for things in the structural syntax that use lists
Ian Horrocks: boris what are the other ones
Peter Patel-Schneider: union, intersection
Ian Horrocks: really correct peter?
Ian Horrocks: in RDF they are mapped to multiple pairwise..
Alan Ruttenberg: alldifferent was already in the first version?
Peter Patel-Schneider: if you want to go for a same-sized translation, there are only two that are lists
Ian Horrocks: not a direct syntax as we have for alldisjoint
Peter Patel-Schneider: the only rationale for having a direct syntax for some of them is
Peter Patel-Schneider: that the obvious translation is n^2 and the non-obvious translation needs a trick
Ian Horrocks: boris says that it makes the problem for roundtripping
Peter Patel-Schneider: we always have a roundtrip from functional to RDF and back, but if we already have a document in RDF then we cannot do roundtripping unless we have a direct mapping
Peter Patel-Schneider: quantitative vs. qualitative improvement
Peter Patel-Schneider: reason is bloat
Boris Motik: I understand the point regarding the bloat
Boris Motik: if we now extend the language with this new construct... new ontology, serialise as RDF then we can do roundtrip
Boris Motik: if you use owl 1.1 RDF and owl 1.1 AS then things should be round-trippable
Ian Horrocks: this wasn't the simple discussion I was anticipating
Ian Horrocks: are not going to get through this in the remaining 3 minutes of this telecon
Ian Horrocks: aim for email exchange to clarify the matter
Any other Business
Ian Horrocks: last item.... aob?
Ian Horrocks: any of it?
Ian Horrocks: declare us finished for this week