See also: IRC log
<bijan> Why not? You think I *like* to hear myself talk?
<calvanese> so, bijna you have to say sth
<bijan> What an aribtrary restriction!
<bijan> Like forbidding class/property punning!
<bijan> Is that Italian for STFU? :)
<calvanese> come on, usually you don't have problems in talking. the contrary would be harder :)
<bijan> Hence the need to mute myself!
<bijan> Oo, that' looks like I snubbed rinke
<bijan> You can talk and I'll listen :)
<Rinke> sure, as if I feel like talking after you've been so rude
<msmith> the scribee thinks they look good
<IanH> PROPOSED: accept previous minutes http://www.w3.org/2007/OWL/meeting/2008-06-25
<msmith> +1 to accept 2008-06-25 minutes
RESOLUTION: accept previous minutes http://www.w3.org/2007/OWL/meeting/2008-06-25
Act 163 completed
<bijan> Done and closed
Action 156 done
<trackbot> Sorry, couldn't find user - 42
<scribe> done and closed
action 159 continued
<trackbot> Sorry, couldn't find user - 159
<bijan> He is
action 161 continued
<trackbot> Sorry, couldn't find user - 161
<bmotik> OK, I'll just add it right away.
Diego will write up what his investigation revealed
<bijan> Subject lin contains ACTION-Number
<m_schnei> i step back from 159
jeffP: cmt on inconsistancy
<bcuencagrau> you need the data to have the inconsistency
Diego: Easy keys are compatible with key notion in DL-Lite
Diego: we need to restict these
keys in the same way
... the keys cannot be subtyped
diego to write up how easy-keys could be used in DL-lite
msmith: asked if we described unique names assumption in the profile document
<msmith> I see, I didn't realize this had changed
Boris: it is described
<alanr_> could we get a review of what the issue was?
<bijan> Peter's not here, and he's the issue raiser?
IanH: the issue was - could you annotate annotations?
boris: problem - you can annotate
entities and axioms, but not annotations
... peter proposed that annotations could contain a set of other annotations
... having an axiom that contains another axiom is hard in RDF
... my proposal is to can the issue because both proposed solutions are quite hard
alanr: I wonder if the question might go away with rich annotations
alanr: there are motivating use cases for this
alanr: how is the reification in one of the proposed solutions harder than where we have done this elsewhere
<alanr_> Mcdermott was convincing to me....
<alanr_> one + level of indirection
bijan: may need to recommend how
to construct annotations for meta-annotations
... set up your structure of annotations so that you always reify in a nice way
... we could work out the pattern for this
alan: the idea of structuring annotations works for new annotation but not for old rdf annotations
<bijan> data: uris could solve this
<bijan> But they are ugly
<bijan> Literals as well
ianH: are annotations inside annotations asserted in the KB?
boris: the problem is that there
is no way in rdf to say this axiom contains an axiom
... as soon as its in a bag of triples in rdf it is asserted
<bijan> there's a queue!
boris: can't tell after whether the triple occured at the top level or inside another triple
alan: I'd be happy to work through the example with Boris over email
<alanr_> Note: I will have to leave at 2pm.
bijan: this problem of not having
syntactic context is something I considered
... people who are tracking this should look at the reification table
alan: it seemed like there was
clear consensus on an underlying Real datatype
... and floating point is promoted to this for reasoning
<bijan> +1 to disagree with type promotion
alan: there was a question on whether or not non-numeric values of float like +inf were also promoted
<Zakim> msmith, you wanted to disagree on type promotion
msmith: I agree we want an underlying real datatype, but disagree promoting xsd: float
<alanr_> is it clear what "promotion" means? Perhaps Boris should explain.
<bijan> I also thing Reals shouldn't have NaN. Those aren't reals! Why make a clean datatype and then crude it up!
<alanr_> the argument is that floats are there to represent machine computations.
msmith: don't understand the point of having both xsd:float and xsd:decimal
<alanr_> This is an important use case for Science Commons
<alanr_> not 1+
boris: promotion means to
restrict float values
... I'm pretty happy with ditching float and double, but this will look bad
<bijan> "Ditching"? Isn't it that we "aren't adding"
boris: you might want to store these in an efficient way
<bijan> Floats aren't continuous
<bijan> I'm confused
boris: I would bet if we keep the continuous aspects of float, then now implementation will be correct
<bijan> It's arbitrary sized decimals
<MarkusK> yes, I also think that xsd:decimal supports no exponent notation
<bijan> It can't !
<bijan> There's too much here
<bijan> Can we chunk the discussion a littl
boris: a possible way out is to define an owl:float and owl:real
alan: wrt continuous aspect:
<MarkusK> (for the minutes) xsd:float also has non-numerical numbers, NaN and infinite
<MarkusK> s /numbers/values/
alan: effect on floats, the only
consequence to considering them real would be
... asked some folk, they would prefer real
<m_schnei> (for the minutes, too) IEEE floats also have +/- 0, do xsd:float have too?
alan: also asked about +-inf and
... they considered these to be essential
... the objective is to be able to transmit and contain numeric data in an OWL file
bijan: the first step I have is
if we are going to talk about something with a binary
... we can't avoid rounding, we can't separate the value space from the representation
<alanr_> OWL does not produce new floats in the course of reasoning.
ianH: it seems to me that we are proposing in owl to have a virtual float that is continuous
<alanr_> So precision issues are external to OWL - OWL would not disturb any precision or do any rounding.
bijan: so you are just treating the float rep as an idiosyncratic rep of reals
boris: the value space is the set of real numbers between the min and max of float
<alanr_> Ian asked my question
<alanr_> 2.0 float is not considered different than int float
carsten: I like this
... either dropping float completely, or treating them as reals for reasoning
<alanr_> no float predicate, I think.
boris: floats are a subset of reals
carsten: treat float as a property of a real number?
boris: the reason for doing this is so that you can ship data around as reals
<alanr_> consider: oracle than answers between a and b, how many values. For float we decide to answer: Infinity , always
<bijan> One question at a time!
<bijan> I wanted to respond to the carsten questiona nd now we're off track
boris: if you have something like 1 / 0 then the ontology is unsatisfiable
bijan: there are 3 options for the predicate thing
<alanr_> comment: Lexical float doesn't work - because of defined rounding.
bijan: no predicate
alan: we want to capture the result of an experiment and that may include NaN values
<Zakim> alanr_, you wanted to mention nan as data bottom
alan: they don't care about how many discrete values between here and there
<bijan> the �value space�s of all �primitive� datatypes are disjoint (they do not share any values)
msmith: people using XSD already make the choice between xsd:float and xsd:decimal
<bijan> (double and decimal are primitive datatypes in xsd)
msmith: given the fact that they chose xsd:float, we ought to respect the choice
<Zakim> msmith, you wanted to ask about the benefit of this proposal
boris: the reason that float is not put under decimal in xsd may be because of the 3 special values
<Zakim> bijan, you wanted to point to 4.2
<JeffP> the spec is somehow inconsistent
bijan: the spec says the value spaces of float and decimal are disjoint
<alanr> Does it actually say they are disjoint, or does it not say they have a shared value space (negation or naf)
<msmith> it says disjoint
<Zhe> let boris go first
boris: problem with OWL R profile is OWL R full version is not a syntactic fragment
the idea is to have basically 1 OWL R profile
boris: if the ontology is written in triples it is in OWL R if it is parseable as OWL R
Zhe: Yes. I do agree with Boris
on this. It seems a bit odd to have these two versions.
... I see the value of combining the syntactic restriction into the profile
m_schei: regarding confusion - it is not confusing from an rdf point of view because
m_schei: any sublanguage is a
... I don't share the argument.
boris: I think what is confusing is from an ontology point of view.
boris: you don't know what it means. You can't interpret it in an unambiguous way.
<Zakim> bijan, you wanted to talk about user perspective
<bijan> I'm still on the queue!
<bijan> No no!
boris: what is the point also from a user's perspective, when the meaning is ill-defined.
bijan: In my experience users
find the semantic subsetting confusing.
... In our spec.s, all the other profiles are syntactic subsets
bernardo: I totally agree with
Bijan on this. The purpose of OWL R is to define a
... that is easily implementable using production rules.
... What people real care about is this ability to implement the reasoning using a rule engine.
IanH: to M_schnei - what if they
use some rdfs syntax and no rdfs interpretation is made
... wouldn't the user be surprised?
<Rinke> +q to ask about relation with DLP
m_schei: you would restrict the
reasoning to rdf in a tool like Jena, you would be explicitly
aware of this
... run the reasoning and see what inference graph is produced
... you of course have to know which reasoner you are using
Bijan: you will still be able to do the RDF style reasoning
<Zakim> Rinke, you wanted to ask about relation with DLP
Bijan: In OWLland people are used to having certain syntax indicate the reasoning features in the interpretation
<bijan> (DLP and hornSHIQ are also syntactic fragments)
rinke: when we started we had
dlp, and horn-shiq and others
... How will this impact people who use DLP like stuff
<m_schnei> Motivation for OWL R was RDFS 3.0 / OWL-Prime
boris: Horn-SHIQ was dropped because there were too many fragments
<bijan> And a champion in the working group :)
boris: we just kept those that had larger user bases
<m_schnei> All the fragments in the beginning were *DL* fragments --> HENCE syntactic fragments
<bijan> All the fragments in the beginning were *OWL* fragments --> HENCE syntactic fragments
Boris: what remains is OWL R. You can still use production rules or other similar tools for it.
<bijan> OWL lite is a syntactic fragment of OWL DL which is a syntactic fragment of OWL Full
<m_schnei> Full fragements are always semantic fragments, they are always applyable on every RDF graph
Boris: We are just saying there is a syntactic check that can indicate which profile is being used.
<bijan> OWL DL and OWL Lite *are* fragments of full. What you say is false. And I'll stop the back chat ;)
Zhe: Oracle is planning to
support this profile in the future and it is probable that
... will include the capability to bypass the syntax check.
ivan: what I would like to
understand is if we go with Boris' proposal and I'm in
... what exactly do I lose?
IanH: I guess you lose the
ability to consider some graphs as OWL R.
... Like if you include SomeValuesFrom constructs.
Boris: you don't lose anything.
The rules will work exactly as they are. You don't lose
... expressive power. The syntax forbidden doesn't have rules for the corresponding reasoning.
bijan: in a way the fragment is saying these are the things we know how to do something interesting with.
<m_schnei> what is with the RDFS axiomatic triples?
Zhe: to Ivan's point, expressivity is not lost. Just some ontologies will be rejected, if syntactic checking is on.
bernardo: we have an additional
benefit from specifying this as a syntactic fragment
... you can know if you are in the fragment, and that is a desirable feature.
<bijan> Yay to evan!
This is scribe.perl Revision: 1.133 of Date: 2008/01/18 18:48:51 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Succeeded: s/156/42/ No ScribeNick specified. Guessing ScribeNick: ewallace Inferring Scribes: ewallace WARNING: No "Present: ... " found! Possibly Present: Boris Carsten Diego Evan_Wallace IPcaller IanH Ivan JeffP MarkusK MartinD Note P12 P14 P17 P9 PROPOSED Zhe aaaa aabb aadd alan alanr alanr_ baojie bcuencagrau bernardo bijan bmotik calvanese comment consider data ewallace m_schei m_schnei msmith ratnesh rinke sandro trackbot You can indicate people for the Present list like this: <dbooth> Present: dbooth jonathan mary <dbooth> Present+ amy WARNING: No meeting title found! You should specify the meeting title like this: <dbooth> Meeting: Weekly Baking Club Meeting WARNING: No meeting chair found! You should specify the meeting chair like this: <dbooth> Chair: dbooth Got date from IRC log name: 02 Jul 2008 Guessing minutes URL: http://www.w3.org/2008/07/02-owl-minutes.html People with action items:[End of scribe.perl diagnostic output]