See also: IRC log
<csma> Meeting RIF telecon 16 September 2008
<csma> agendum+ Admin
<csma> agendum+ Liaisons
<csma> agendum+ Publicity and public comments
<csma> agendum+ Actions review
<csma> agendum+ F2F11
<csma> agendum+ Core
<csma> Agendum+ Test Cases
<csma> Agendum+ AOB (pick scribe!)
<csma> Michael, do you remember that you scribe, today?
<csma> Scribe: Michael Kifer
<csma> scribenick: Michael_Kifer
<csma> PROPOSED: accept minutes of telecon Sept 9
<csma> RESOLVED: accept minutes of telecon Sept 9
<Harold> The OWL 2 RL profile was just a section in another doc.
<ChrisW> ACTION: chris to finish responses to public comments by sept 22 [recorded in http://www.w3.org/2008/09/16-rif-minutes.html#action01]
<trackbot> Created ACTION-575 - Finish responses to public comments by sept 22 [on Christopher Welty - due 2008-09-23].
<csma> PROPOSED: Core will not have equality in the head
<MichaelKifer> Discussion of the proposed resolution that Core will not have equality in the head
<MichaelKifer> Addl info: Dave R removed his objections to restricting equality in the body
<MichaelKifer> RESOLVED: No equality in the head
<csma> RESOLVED: Core will not have equality in the head
<MichaelKifer> equality in the body will be discussed at the Core telecon next Monday, 11am.
<Harold> PROPOSED: Core should keep unrestricted equality in rule bodies (cf. ISSUE-71).
<Harold> PROPOSED: Core should keep unrestricted equality and external function
<Harold> calls in rule bodies and keep external functions calls in rule heads.
<MichaelKifer> Note: external functions are NOT allowed in the head in BLD
<MichaelKifer> The above "PROPOSED" resolutions are examples of what is to come from the "Core" subgroup.
<Stella> let's start with EntailEverything
<MichaelKifer> Leora: need test cases that illustrate hard issues in the language of BLD
<Stella> 1 through 6
<MichaelKifer> Gary: the purpose the EntainEverything test cases is to test the basic features of BLD.
<Stella> that is not in proper format
<MichaelKifer> Regarding the Frames test case, the ? in the vars is in a wrong position
<Stella> Adrian intended to update this once the PS syntax is finalized
<MichaelKifer> Gary: the Frames example seems wrong conceptually. Customer should be an object or a variable (Sandro).
<Hassan> The choice of syntax must be justified and explained...
<Stella> yes, I think we do
<MichaelKifer> Csma: how complete should be the test suite?
<MichaelKifer> Sandro: somebody should go through the spec and suggest the places that are suitable for the testsuite.
<MichaelKifer> But this is hard to do, since we don't have an automatic way to check these tests.
<Hassan> I agree with Sandro regarding syntax
<Hassan> Otherwise semantics of the syntax is a guess...
<Stella> but then reviewers of the tests will have to learn multiple languages
<MichaelKifer> Issue: how to express the test cases? The PS is a moving target
<trackbot> Created ISSUE-77 - How to express the test cases? The PS is a moving target ; please complete additional details at http://www.w3.org/2005/rules/wg/track/issues/77/edit .
<Stella> there is currently a PS documented in BLD, we can use that
<Stella> and if it changes, update the test cases
<MichaelKifer> Sandro proposed that test cases would be labeled with timestamps that indicate the version of the PS. A number of people did not like that.
<ChrisW> who needs to imagine, i've DONE it
<Stella> why don't we use the PS for the human readable version, and XML for the runnable version, and a tool to translate between
<ChrisW> that is, i hope, what hassan is proposing (and doing)
<Leora_Morgenstern> Stella, the problem is the multiple versions of P
<Leora_Morgenstern> of PS
<ChrisW> but the PS is not in a form yet that can support that
<Stella> there are not multiple versions yet
<Hassan> XML is not "normal form" but "normative"
<Hassan> I agree with ChrisW on XML generated by people!
<Stella> people can even submit them in email in informal form for now
<Stella> we will translate them into ps for them
<Leora_Morgenstern> good point, Stella. We've said it before, but I think people have forgotten.
<Stella> submission does not have to be in xml. tool will create the xml, and the xml is the normative form that will be run
<Hassan> BTW - in that example it should be ?Name
<Hassan> not Name?
<Stella> +1 chris about making ps serializable
<MichaelKifer> Hassan: proposing to complete the PS to make it into a parsable syntax. Chris expressed support.
<MichaelKifer> Csma: wants to write everything in XML, against a parsable PS.
<Leora_Morgenstern> +1 to Sandro's point above. This is the whole point of RIF, after all.
<MichaelKifer> Sandro: Prolog is not good for writing test cases, as there is no standard mapping to RIF-XML.
<Stella> it's kind of ok, except the testers will have to get use to multiple presentation syntaxes to understand the cases
<MichaelKifer> CSMA: correction: don't want to write everything in XML. I think that the point of the test cases is to *document* the cases. XML is a good way for doing this.
<MichaelKifer> Leora: but we also need to test the cases.
<sandro> Sandro: let's say that each test case in the wiki exists in any language which has a well-defined mapping to RIF and for which code is installed by the test-case-managers, so that the downloadable version has the XML and that's what is run.
<sandro> Sandro: This degenerates to the current situation if we assume Hassan's PS->RIF-XML is our only defined+implemented language.
<Zakim> sandro, you wanted to say the problem with using other rule languages (eg Prolog) is that there is not a standard mapping to RIF XML. I suggest mappings must be defined and
<Hassan> Sandro: this is not the case - I think Aberdeen folks have an implementation as well. Am I right?
<MichaelKifer> Leora: disagree with Sandro's proposal that people should be able to write test cases in their own languages.
<Leora_Morgenstern> no, Michael, I disagree with Christian's proposal
<Leora_Morgenstern> Michael, I don't think Sandro and Christian are saying the same thing at all!
<Stella> I think we can get fairly broad (not deep) coverage
<Stella> I can work on it
<MichaelKifer> They are saying different things, but I thought that your argument was against Sandro's proposal (for the most part)
<GaryHallmark> stella, I hope you will use RIF-PS and not prolog or ilog or mylog, or ...
<Stella> and then we need the deeper coverage in the corner and difficult cases
<Leora_Morgenstern> Let me just write this rather than say this, because I think it will be clearer:
<Leora_Morgenstern> The presentation syntax is built on the underlying BLD semantics.
<Leora_Morgenstern> This is not the case for the syntaxes of other languages, right?
<Leora_Morgenstern> Christian, I don't think you can entirely separate the issue of a comprehensive set of features from the issue of corner cases.
<GaryHallmark> corner case = things that are hard for PRD :-)
<Leora_Morgenstern> I agree, Chris. It's when you look at things, feature by feature, that you may come up with corner cases.
<Harold> If someone has their own syntax, OS, but give a mapping to RIF's XML, OS->RIF/XML, then OS can be seen as just their 'shorthand': only the XML produced by their mapping will be relevant for RIF syntactically and semantically.
<Stella> they all involve features, but it's a matter of focus
<Stella> whether we only want hard or non-obvious cases, or the general coverage also
<GaryHallmark> Harold, assuming I'm not familiar with OS, I would not find this useful unless you can completely specify OS->RIF/XML in a way that is easy to understand (i.e. not just a black box service)
<MichaelKifer> Leora: "corner cases" correspond to features. To distill them one needs to go feature-by-feature.
<Harold> Sandro, right, I don't take any standpoint here.
<csma> nobody does, zakim
<MichaelKifer> Leora: test cases should "cover enough" situations. It does not need to be exhaustive.
<MichaelKifer> Gary: Stella's cases are real good.