F2F3 9 Jun 2006 Session 6

9 Jun 2006


RIF, Leora_Morgenstern, AlexKozlenkov, pfps, Evan_Wallace


Technical Specification


harold: I pasted unchanged what was in the e-mail then made one clarification change, so you can see it in the wiki history

mkifer: we had a poll on this, right?

csma: we agreed to start with a condition language, and examine this as a candidate

(in the telecon about 5 weeks ago, as I recall)

Axel: my review
... Slotted syntax, as in F-logic and WRL.
... My languages are WRL and DLV
... Has extensions for aggregates, etc
... so missing things:
... 1. slotted
... 2. only-conjuntions vs. complex formulas? does this language only allow negation in front of atoms?

Harold: that's an extension, in A2

Axel: Builtins

go for it, Paula!

<sandro> scribeNick: PaulaP

Axel: do you allow two free variables in the condition

Michael Kifer: syntactically, yes

Axel: some languages allow binding patterns and the question is whether we should allow such patterns

Michael Kifer: yes, I think this is an issue

Axel: Phase II issues are also considered
... aggregates are also an open issue

csma: do you have something useful?
... is there an essential part that you could map and this part is useful?

Axel: yes, there is a useful part

<AxelPolleres> For the records: http://lists.w3.org/Archives/Public/public-rif-wg/2006Jun/0035

Hassan: most of Axel's comments apply also for me
... missing slots as well
... types, because objects are types
... I took the object model of IRL and put it into a logical form
... this is a way to get around with slots and other features
... aggregates are missing

Michael Kifer: types for slots or for variables?

Hassan: for both
... types a la Java

ChrisW: Is Peter on the phone?

Sandro: he is on the phone
... I guess not

ChrisW: any more discussion on this?

Sandro: I'm tenpted to do such a mapping too
... I imagine we can have a web form for mappings to RIF

csma: the same question for Hassan as for Axel...
... are there easy extensions to the condition language that are needed to cover what is missing?

Hassan: yes, we need more elaboration on XML constructs that we need

csma: we can look at some concrete examples

Hassan: I think we should all do this together

Axel: how about mapping N3 to the condition language?
... or perhaps SPARQL?

<sandro> Sandro: binary or ternary mapping to RDF ?

Harold: the text on the condition language mentions some of the issues that are needed for such a mapping

Hassan: Prolog is trivialy mapped to the condition language

csma: so we are done for phase 1

ChrisW: how about reverse mappings?

no one tried this

ChrisW: is the condition language what people expect from RIF?
... does this looks like a RIF?

csma: you also need an abstract syntax

Hassan also agrees on this

<sandro> by Abstract Syntax them mean XML Schema

Hassan: I can work on this on IRL
... we should complete this and then define what we need
... I took the grammar, annotated it and got the XML serialization

csma: Hassan should show us what he means

ChrisW: the question is what happens if one uses the condition language to interchange between WRL and IRL

Michael Kifer: we need to define the taxonomies

csma: and then to go back to this

ChrisW: an issue is the default behaviour in this particular case

Hassan: Hyperdocumentation for grammar RCL.grm
... you can specify an XML serialization by providing annotations

Sandro: did you developed this only for this application?

Hassan: we can of course generalize this to other applications
... I tried the example given in the text of the condition language and it works
... more details on the grammar definitions
... the bindings are done by using object handlers
... test expression can be very complex
... you can have a referent expression, a partial expression
... my simple annotation breaks down here

ChrisW: all these activities are useful for various reasons
... this can impact RIFRAF

csma: can you interchange a useful part of your language?

Hassan: it is useful now

Michael Kifer: the binding patterns represent the only discriminator for RIFRAF

Michael Kifer: not sure how to RIFRAF built-ins

<scribe> ACTION: Michael Kifer to describe the issue on built-ins as RIFRAF discriminator

Hassan: I don't parse XML yet
... there is no way to provide the original syntax
... this is because I use the mapping
... from IRL to AST, then get XML

the problem consists in the mapping from AST to IRL

Hassan: this is a problem if one needs to do the reverse translation
... AST doesn't know what to do with the universal

ChrisW: do you have the same problem?
... what can't you translate into WRL but you have it in the condition language?

Hassan: I didn't do the IRL

Axel: I wrote an email on descriptions of rule systems
... this goes in the same direction
... what subset of RIF do I understand?
... I need an annotation of what I understand

ChrisW: descriptions of language capabilities are needed
... for this we need such mappings

Hassan: from XML to AST, we need an XML to AST parser
... as long as I see tags that I understand, it's no problem
... I see universal now
... I don't know what to do here

csma: this is a compliance issue

Hassan: there are a couple of possibilities

ChrisW: it is important for determining appropriate behaviours

Hassan: it is before that
... AST is a bunch of Java classes
... I am RIF compliant, so I should be able to have a default behaviour represented internally

csma: we are talking about the behaviour of the translator

ChrisW: the idea is to determine the behaviour

Hassan: we need representations for such kind of behaviours

csma: a possibility is to reject rule sets

ChrisW: we should be careful not requiring changes in the rule systems

<AxelPolleres> For the records, link to my mail mentioned above: http://lists.w3.org/Archives/Public/public-rif-wg/2006Jun/0102

csma: this could be a separate program
... so no need to change rule systems

Gary: we will have a large number of discriminators
... but we want a small number of dialects
... we need to structure these discriminators

csma: some features may be optional, so we need a way to handle default behaviour

Gary: default behaviour may be in most cases 'not proces these rules'

Sandro: there are also other possible behaviours
... perhaps I want fewer results

csma: this works for publishing rule sets and using them as they are, without communication

Sandro: difference between profiles and dialects
... I can see the reason for the difference now

csma: for some use cases we might need also a kind of protocol

Gary: a negotiation is not always possible

ChrisW: it is useful for a language to describe themself
... the point of the dialects is for interoperability

Sandro: dialects need to be in the intersection of the languages to be interchanged

ChrisW: that's fine
... but the languages may have a superset that is described in their profile

Hassan: RIF compliance means that all construct I understand I can represent in RIF
... but what do you do with things that you don't understand directly?

<pfps> I don't think that it is *necessary* to be in the intersection to have useful interchange. Perhaps it may be necessary to be in the intersection of the dialects to have totally faithful interchange, but I think that useful interchange can be done even if it is not totally faithful.

Hassan: I'm RIF compliant even if I don't understand everything from RIF

Sandro: there are two modes of deployment
... inside an enterprise or between different business partners
... there are different styles

Hassan: I see what you mean but I still don't understand what your definition is

csma: your definition is ok, but we should not have many dialects
... this is the point of this discussion
... I have a problem I would like to raise

<sandro> (mode 1 allows lots of extnesions; mode 2 is strict-to-standard-dialects)

csma: on Hassan's problem with getting the original syntax
... is this depends on your implementation?

Hassan: it depends on how concrete your abstract syntax is
... if the RIF XML is abstract, then is impossible

Gary: no, it is possible but is harder to read

Hassan: I agree with Gary

csma: so you could generate some sort of IRL from the AST

Hassan: yes, it is possible but I don't know if it's readable

Gary: an example is Java

Hassan: variable names will be a problem

Sandro: very good chairing if this meeting

csma: we are even before schedule
... thanks for scribing

<Darko> -Darko

ChrisW: thanks everybody


<EvanWallace> scribes did a great job today!

<PaulV> quit

Summary of Action Items

[NEW] ACTION: Michael Kifer to describe the issue on built-ins as RIFRAF discriminator
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.127 (CVS log)
$Date: 2006/06/17 20:13:49 $