RIF F2F 27 Feb 2007 -- 27 Feb 2007

UML Diagrams of CORE

<ChrisW> Scribe: Allen

<sandro> discussion over what the leading "-" means on the UML diagrams? it seems to mean something about public/private -- something we don't care about here.

<sandro> (prefixing the names of relations/properties)

discussion about UML diagram for structure of RIF Core Rules

question why can't we rename implies to rule?

<sandro> Christian shows outline of UML from PRR.

csma: why forall is a class?

<Hassan> Please everyone make sure to turn on your mikes! Thanks.

<Hassan> Thanks!

csma: what about rule-set?

harold: could be a level above

paul also questions forall class

<Hassan> mike?

<sandro> Sandro: "forall" as a class comes from the standard FOL syntactic nesting

sandro: this maps to scoping

<sandro> Allen: "forall" represents the class of universally quanitified formulas

chris: this mirrors fol syntax

csma: why is the rule associated with forall instead of implies?

<ChrisW> scribenick: Allen

harold: keep it general for extensibility

<sandro> error in diagram --- forall can take either an implies or a positive --- diagram says it has to have both.

harold: positive is a disjunction?

csma: should we extend this arbitrary formulas?

sandro: at some point yes

harold: you need disjunctions for integrity constraints

csma: link from forall to postive?

hassan: likes this diagram
... this covers prolog class of languagaes nicely

chris: but there is still a problem with the diagram

csma: that link allows for rules with empty body

harold: it is only a matter of brevity of expression

csma: straw poll on this

5 prefer as is

<sandro> straw poll preference 5-to-3 for having facts as themselves, instead of as degenerate rules

3 prefer remove that link

2 don't care

<Harold> Hereditary Harrop Formula: http://arxiv.org/abs/cs.PL/0404053

mike: why it a 1 on the formula side?

harold: to be consistent with Horn

csma: have the same problem as yesterday: syntax vs. metamodel points of view

hassan: to harold why do you want this "At all costs"
... don't understand arg for having facts without implies?

harold: it make is simpler to write facts

csma: so we keep it as it for now

dave: but you do need to get the disjunction in there

jos: there is a way to do that in uml

chris: let's use an intermediate class like in first diagram

harold: ok, we can use "clause"

csma: clause is either a rule or a fact

csma draws diagram on whiteboard

sandro recommends some changes

mike: what about duplication of positive?

harold: it is just a readability thing

<Hassan> mikes???

sandro: forall vs. rule,
... rules should be same as formula

csma: suppose we need existential rule variables (shared by body and head)

harold: that can definitely happen
... it would be side-by-side with forall

harold goes to whiteboard

harold: ruleset cotains 0 or more of univerally or existen clauses

csma redraws diagram

<Hassan> It's hard to follow : no mikes for most (except Christian) and no diagrams!

csma: postpone decision vis-a-avis core

csma describes simpler diagram

csma: does any object to having this in core wd1?

<Hassan> The description was too fast for me to catch all details ... :-(

sandro: where is rule?

jos: we need to be consistent. rename ruleset or use something called rule

mike: it is odd not to have "rule"

<Hassan> I second Mike's point...

<sandro> Christian objects to my proposal that Rule==Formula on the grounds that recursion is too much for WD1.

csma redraws diagram with rule inserted btwn ruleset and forall

<sandro> Christian proposes a replacement version where Rule is a superclass of Forall, but under Forall is the same as before, for now.

<Hassan> A pic of updated diagrams would be nice (anyone a camera)?

harold: rule is very general includes facts
... and integrity constraints

paul is taking a shot of the diagram

harold: allows non-ground facts

csma: any objections to new diagram?

sandro: can a ruleset directly contain a clause?

csma: yes

sandro: consider the xml

csma: concrete syntax not supplied by this diagram

<Hassan> (Thanks Paul!)

sandro: a fact would still need an empty forall list

<sandro> sandro: can a clause be a rule?

sandro: you don't want to recurse on rule

csma: i don't understand the implications of that
... keep it like this for wd1

harold: I will add some "blue" explanatory text about this

csma: no new material
... add new comments to draft in progress, not to released wd1

<Hassan> I do not have the info yet to vote

paul: can we identify these diagrams somehow

sandro: is fact a superclass of postive?

csma: fact is a kind of clause

<Hassan> Is there a better name than "positive"?

harold: may be a bit redundant

csma: we can't resolve all the issues, but can we agree to publish "that one" for wd1
... prefers to keep "fact" and positive separate to avoid recursion

harold: wants to merge them
... but don't call it either of those, might use "litform"

csma: straw poll on replacing fact with positive, currently call the merger "positive"

<sandro> straw poll: merge Fact and Positive 5 in favor, 3 against,

5 in favor, 3 against

csma: put this merger in 1st wd

mike does not object

no objects to changing

csma objects

csma withdraws objection

harold and michael: we like "Atom"

hassan: objects to other names too, like "uniterm" etc

csma: not discussing that now

csma describes change of fact to atom

hassan is ok with publishing in wd1 except for certain names

sandro: as its written here the role names are not in diagram

hassan: what about uniterm?

harold: a "universal term" atom or expression

csma: what do we need to add to diagram., roles?

harold: we need it to bridge communities

hassan: i differ
... if-then in production rules is not implies

paul: this is an "Abstract model" for an abstraction...

dave: but for nowthe xml syntax would contain these names

csma: don't add names for roles for now, avoid contention

harold: what about the minus signs

sandro: are you proposing giving up mapping to xml

csma: no, but don't include names for roles in wd1

<Hassan> I agree with Sandro

sandro: first wd should be implementable
... i thought we had an xml syntax from these diagrams, fully striped

harold: we need the roles

paul: will the syntax use class or role names

sandro: both

paul: shouldn't class be generic roles specific to domain

csma & paul: use body and head for roles

paul: the vocabulary can change for other dialects
... atomic formula ok, implies and forall no

hassan: antecedent, consequent, var to variable

sandro: sympathetic to it, but torn because everyone thinks in terms of if-then

john: if-part then-part

<sandro> scribe: Sandro

<scribe> scribenick: sandro

<Harold> Because of their identical content models, Uniterm is unification/merger of an Atom (in the sense of a predicate applied to arguments) and an Expression. As a minor new point, instead of POSITIVE we could say ATOMICFORMULA (in the sense of Uniterm or Equal).

csma: (reviewing diagram)
... "decalre" renamed to "variable"
... "implies" to "Conditional" (to match "Atomic")
... "ifpart", "thenpart".

Harold: we're had many versions of these names. hard to see all the consequences.....
... I'd object to "ifpart"

csma: so we stick with the old names for WD1
... so back to "if" and "then" and "implies",
... and still "atomic"
... and 'declare' instead of "variable".
... I want it on the record that this was discussed and may be discussed again. we are in no way committed to this version.

Hassan, are you calling in?

csma is working on getting diagram out in e-mail.

vpn troubles.

<Hassan> (Received - thanks Chris)

PROPOSED: in WD1 we'll publish this diagram, labeled as "still under discussion".

RESOLUTION: Use diagram in http://lists.w3.org/Archives/Public/public-rif-wg/2007Feb/0134, in Core WD1, labeled "still under discussion"

<Hassan> I can't hear

Reader Feedback on Core

<Hassan> Can someone post a pointer to the topic at hand if there is any (slides maybe?)

chris is working on it, Hassan, I think.

MikeDean: If there's a language designed for human consumption, than some people will implement it.

ChrisW: But it's not important in this WD.

<Harold> The RIF Human Readable BNF Syntax was modeled on the OWL Abstract Syntax (http://www.w3.org/TR/owl-semantics/syntax.html) regarding its Lisp-like prefix notation and its use of whitespace as separator.

csma: the question is how to address comments about BNF.

want to call in , MoZ? we're talking about your comments.

<ChrisW> slides are up on the wiki

csma: What is the proper way to deal with all these comments on the BNF?

<Hassan> (thanks - again! - Chris...)

Harold: DateTime may be more controvercial?
... We just wanted to just have something like OWL's S&AS abstract syntax.

<MoZ> sandro, Zakim France seems full for the moment...

MoZ, press 0 for an operator and ask them to add you -- they can over-ride the limit.

Oh, Zakim France. Huh..... I dunno about that.

<MoZ> sandro, tel:+33.4.89.06.34.99

csma: concrete syntax for types...... we could remove this, or more clearly label it as an examples.

michaelKifer: the reason for those is to show people how they play out, in a concrete way.

<LeoraMorgenstern> Sorry --- the irc had died on me for a while --- can you tell me which slides we're looking at now?

csma: the dangers is that it looks so thorough that it looks like the real syntax.

DaveR: Didn't we just agree we were using XML Schema datatypes?

<Hassan> Michael: please speek into your microphone - thanks.

<Hassan> (I did mean "speak" not "peek" :-)

MK: note it as "just for illustrative purposes"

csma: that not everything has been fixed, or decided by WG
... Maybe we need to label in the draft which things are decided and which are not......

PROPOSED: The concrete human-readable syntax, described is BNF, is: work in progress and under discussion. (It was already resolved as being For Illustrative Purposes Only).

Sandro: (sarcastically) maybe we should label the whole things as a "Working Draft"

<ChrisW> hearing noise on phone

PROPOSED: The concrete human-readable syntax, described in BNF, is: work in progress and under discussion. (It was already resolved as being For Illustrative Purposes Only).

csma: this resolution will let us skip many of the feedback comments.

RESOLUTION: The concrete human-readable syntax, described in BNF, is: work in progress and under discussion. (It was already resolved as being For Illustrative Purposes Only).

csma: so we can skip some bullets.
... reserved words?

mk: not a problem in the XML -- problem in HR syntax.

<scribe> ACTION: Harold to fix ForAll, FORALL inconsistencies [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action01]

<rifbot> Created ACTION-244 - Fix ForAll, FORALL inconsistencies [on Harold Boley - due 2007-03-06].

<scribe> ACTION: mkifer to delete DateTime text and use reference XSD instead [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action02]

<rifbot> Created ACTION-245 - Delete DateTime text and use reference XSD instead [on Michael Kifer - due 2007-03-06].

mk: You don't need to type uniterms because you already know type from signature

moz: But can you narrow the type?

mk: what's the point?

moz/mdean: something like integer 1..17

mk: that's already in the languages, as sorts are defined

points about xml syntax

dtd dropped.

default namespace rif = "http://www.w3.org/2007/01/rif#"

<Hassan> (sandro - pls speak UP! :-)

sandro: we can get rid of the 01 with permission.
... 01 is the month

PROPOSED: that xmlns in WD1 is "http://www.w3.org/2007/01/rif#"
... that xmlns in WD1 is "http://www.w3.org/2007/01/rif#" (consdiered preliminary)

mk: How do namespaces change when standards change, eg for XML Schema Datatypes?

DaveR: There haven't been any new versions...
... in RDF, they decided not to change the namespace, even though they changed the spec --- or you could change the namespace.
... There's no painless answer -- there are tradeoffs.

Hassan: We'll need to face that someday -- some kind of versioning control.
... if there are examples in the draft, they should use the NS

Harold: No, they'll make it look too official.

DaveR: We could just state it wherever we mention the NS -- say that it's implied everywhere else.

RESOLUTION: the xmlns to use for WD1 is "http://www.w3.org/2007/01/rif#"

<scribe> ACTION: Harold to change Core to include the xmlns namespace "http://www.w3.org/2007/01/rif#" [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action03]

<scribe> ACTION: Harold to change Core to include the xmlns namespace "http://www.w3.org/2007/01/rif#" [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action04]

<rifbot> Created ACTION-246 - Change Core to include the xmlns namespace \"http://www.w3.org/2007/01/rif#\" [on Harold Boley - due 2007-03-06].

"The sort name should be a URI"

DaveR: so use "xsd:integer" instead of "integer" in draft.

<ChrisW> noise on the phone (Hassan are you muted?)

Jos: *can* be URIs or *must* be URIs?

mk: Why?

Sandro: It's simpler to *always* use URIs

Harold: "import" will need to turn things into URIs.

Jos: That's normal & natural

PROPOSED: all sorts will be named with URIs

Chris: Are there user-defined sorts?

mk: I have some language, X, and I have my own sort -- how do I exchange it with someone else.

csma: If I defined shopping cards and customers, etc, am I defining sorts???

mk: I don't think so..... (hesitantly)

ChrisW: i thought sorts were there for how symbols are categorized in dialects -- in which case requiring URIs is fine. I don't want to force URIs for user-defined types.
... If you want to load in some data model for your application, are you including as sorts ........
... you do treat user defined types as sorts?

mk: The document is silent about that.

csma: We said earlier that identifiers would be URIs if they were not local.

sandro: sounds like that should extend to sorts.
... if they are local -- you don't interchange them....?

<Harold> Sorted logic example -- Schubert's steamroller: http://www.inferenzsysteme.informatik.tu-darmstadt.de/~walther/Paper/Schuberts_Steamroller_by_Many-Sorted_Resolution-AIJ-25-2-1985.pdf

csma: depends what you means by "local", cf, local variables.

mk: How about we say the sorts RIF-WG defines will be given URIs.

DaveR: Sorts as a mechanisms for extending syntaxes .... is different from application-specific types.

PROPOSED: Any sort defined in CORE MUST BE identified by a URI.
... Any sort defined in Core MUST BE identified by a URI.

RESOLUTION: Any sort defined in Core MUST BE identified by a URI.

mdean: Will we use URIs or cURIs, so you can tell whether http is a prefix or a URI scheme?
... so examples should say xsd:integer now.

<scribe> ACTION: kifer to make sure sorts are named with curis [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action05]

<rifbot> Created ACTION-247 - Make sure sorts are named with curis [on Michael Kifer - due 2007-03-06].

[i3] done.

[i4] already done

[i5] what is the sort URI --- is it essenially the string (ie xsd:anyURI), or something else....

mk: I meant it in the sense of xs:anyURI -- an URI is a kind of string.

jos: then we don't have a way to use URI to refer to abstract objects.

DaveR: there's a big difference between "Jos" and Jos himself -- the signature of a predicate might say it pertains to strings or people....

Allen: (workshop)

Dave: There's some muddyness about things vs pages -- that's not what we're talking about here.

Jos: this is well understood in RDF (example of different URIs)

csma: (incomprehensible)
... a predicate will be in a boolean sort and if it's identified by a URI, then.....

predicate-name can be a URI

mk: constants that identify cars, constants that identify people, constants that idenfity pencils, .....
... a database is a bunch of a symbols --- it's in the mind of the creator of the DB that those symbols are associated with people, etc.

<Harold> Following up on the discussion yesterday, and what Jos just indicated, the URIs http://example.org, http://example.org/ and http://example.org/index.html are all different as xsd:anyURIs but equivalent as RDF URIrefs.

DaveR: Suppose I'm writing a library of builtins. I'd write signatures for those functions. I want to create a strlen builtin, and some that apply to real-world things.

csma: first case sort is URI, second case sort is a Resource.

Dave: I think we need "Resource" as another sort.

mk: anyURI --- elements of the sort have internal structure (eg schema, path, host), and may have a method toString, and it can have a method "fetch". URI and String are different, but can be converted to each other.

Dave: Fine -- but that's all different from Resource.

mk: If you are using a URI to denote a person, that's your business, as in a db.

jos: Not true. In XSD an anyURI denotes itself, it cannot denote a person.

mk: but in a database it can.

Jos: We are not talking about databases here.

<AxelPolleres> if I might hook in here, I think that making this difference between resource and URI-typed literals in RDF doesn't seem to be such a good idea and makes quite some troubles, IMO. but this just as a side note.

<Harold> Besides proceeding from string-like anyURIs to equivalence URIrefs classes, we need also need 'dereference' URIrefs. The semantics for this dereferencing depends on the URI sort: for URIs denoting individuals, dereferencing just moves towards the semantic domain element; for URIs denoting another RIF Ruleset, dereferencing could be regarded as a importing it.

<AxelPolleres> ... well, but I see the point (of jos, dave)

Dave: example of RDF: "someURI"^^xs:anyURI vs someuri

Jos: I'm not sure we need a sort for this. These are just constants.

sandro: is there a universal sort?

<Harold> Sandro, we considered to introduce a universal rif:Any sort.

<AxelPolleres> owl:Thing?

<AxelPolleres> maybe not....

mk: I we're making statements about Chris, and he has a URI, why can't I say he's an anyURI ?

Jos: This is the usual way. Abstract domain and concrete domain.

<Harold> Axel, there was a discussion about 2 months ago with Dave about owl:Thing perhaps being rif:Any, but then he brought in rdf:Resource...

Jos: people are in abstract domain, concrete domain might have a URI in it.

csma: two separate discussions. 1 -- "URI" sort in core is xs:anyURI -- agreement **YES**
... 2 -- do we need a Resource sort some day -- unknown.

<AxelPolleres> thanks harold, can you paste the uri to the thread maybe?

Dave: the sort here might be rdfs:Resource, but I'm not sure that's exactly what we need here.
... but I think we're tabling this for now.

Jos: Why have anyURI in there? It's pretty obscure. Just have strings.

<Hassan> For what it is worth, I agree with Jos...

<Harold> Axel and Dave, I guess it was off-line, so if Dave is fine, I will search my mailbox and forward to you and everyone interested.

Sandro: it's just a subclass of string. Why bother?

MikeDean: Actually it's not a subclass of string.

Sandro: Ah, okay. Still, it' kind of obscure.

Jos: I think all the text about URIs in the Core is based on this misunderstanding.

<Harold> In http://www.w3.org/TR/xmlschema-2/#built-in-datatypes anyURI is *a sibling of* string (it's not *a* string).

Chris: We just recently agreed that sorts in Core would be named with URIs..... is that related?

mk: No.
... it's a name which looks like a URI

csma: we need it if we have predicates that apply to URIs.

mk: what sorts do predicate names come from? eg, maybe we want to restrict it to strings that look like URIs.

<Hassan> http://www.w3.org/TR/xmlschema-2/#anyURI

Jos: anyURI has a value space
... for naming predicated, we want strings, not anyURIs
... just quote the RDF specs about what URIs are -- don't use anyURIs.

mk: we might want to allow, eg, integers as names of predicates, but not floating point numbers. so for this kind of thing, we want URIs here.

Jos: use URIReference as in RDF

<Hassan> Very good analysis Dave! I agree ...

Dave: we don't have "this is a predicate, and here is its identifier _____" ---- we're talking about the mechanism.

<Hassan> To rephrase Dave's in French: "Nous mettons la charrue avant les boeufs!" ("we worry about the plow before the we have oxen!")

mk: we just need a lexical space, without any associated bagage of equality in the value space, etc.
... if you don't have sorts, then anything can be used in any contexts. Sorts allow us to say URIs can be used to name predicates, but for instance that floating point numbers cannot.

<allen> check out section 6.4 of http://www.w3.org/TR/rdf-concepts/#dfn-URI-reference

<Hassan> don't URIs have a canonical form?

I don't think so, Hassan.

<Harold> An equality theory for URI should look into rfc3986 "Uniform Resource Identifier (URI): Generic Syntax" (http://tools.ietf.org/html/rfc3986#section-6).

<DaveReynolds> The XSD section is at: http://www.w3.org/TR/xmlschema-2/#anyURI

Jos: typical use of sorts is just syntactic disambiguation -- however, we've also been using it for XML schema datatypes which suggests the value space semantics
... Two URIs for the same person cannot be stated to be equal because of course the strings are not equal.

mk: ah ha!

<Hassan> Sandro: isn't 6.2.2. Syntax-Based Normalization in the link Harold just posted defining such a canonical form?

sorry Hassan, I'm scribing.

or trying to scribe

csma: rif:URI as sub-sort of xs:string

Jos: but we need to be explicit about them being interpreted in some abstract domain.

mk: If we're are talking about the sort of integers, than all the equalities in xsd should be there.

csma: but not for uris.

Jos: just have to be careful not to use any unsorted names.

mk: all constants are sorted.
... So.....
... we'll have to define our own URI sort, with the lexical space coming from RFC 3986.

Dave: When push comes to shove, we'll have two different things here, with different value space.

Chris: The difference between a URIRef and a Resource.

Dave: Yes.

<Harold> Dave, isn't this like What is in the middle of "Paris"?

csma: let's raise an issue on this.

<Harold> (The distinction between names and their denotations has been discussed in philosophy for a while.)

<scribe> ACTION: Deborah to raise issue on rif:URI sort [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action06]

<rifbot> Created ACTION-248 - Raise issue on rif:URI sort [on Deborah Nichols - due 2007-03-06].

<Harold> s/"Paris?"/"Paris"?/

PROPOSED: replace uri with rif:URI in WD1 and link to issue.

RESOLUTION: replace uri with rif:URI in WD1 and link to issue.

<Hassan> when do we reconvene?

<scribe> ACTION: mkifer to update Core with rif:URI and link to ussue. [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action07]

<rifbot> Created ACTION-249 - Update Core with rif:URI and link to ussue. [on Michael Kifer - due 2007-03-06].

<Hassan> thanks - bon appetit

Reconvene at 1:30 (eastern(

Session continues after lunch break

<johnhall> scribe: johnhall

ChrisW: start with DAve Reynolds i6
... integer and decimal make more sense?

josb: just use integer and decimal

sandro: can' just change the charter

josb: charter says integer

csma: at least int

chrisW; charter required inte, is this proposal to support at least 'long'?

chrisw: go back to charter and discuss adding others for next WD

mk: implement long, have inplemented integer?

daveR: integer/decimal pair is sensible

mk: double or float exist and can be taken as decimal
... ... in fact decimal requires lot of work

chrisW: go back to charter

josb: charter includes ' decimal'

chrisW: anyon object to adding decimal?

no objections

daveR: also deal with float and double

chrisW: resolved - leave draft as is?

RESOLUTION: keep text as in draft, which changes datatype list from charter by replacing int with integer.

csma: charter "other primitive sorts ..."

DaveR i7

DAveR; had not defined RuleSet

scribe: now we have

DaveR: Issue in WD after second picture

ChrisW: add placeholder "WG has still to discuss ordering"?

josb: discussed in last F2F

,,, decided on not ordering

harold: 'ordered' could be XML attribute

chrisW: action on MK and Harold to replace diagram and remove issue

DAveR i8

DAveR: for WD2

<ChrisW> ACTION: harold to delete the issue below the rule diagram [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action08]

<rifbot> Sorry... I don't know anything about this channel

chrisW: postpone, also i9

<sandro> rifbot, help?

<rifbot> See http://www.w3.org/2005/06/tracker/ for help (use the IRC bot link)

<sandro> ACTION: Sandro to rest rifbot [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action09]

<rifbot> Created ACTION-250 - Rest rifbot [on Sandro Hawke - due 2007-03-06].

<sandro> ACTION: harold to delete the issue below the rule diagram [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action10]

<rifbot> Created ACTION-251 - Delete the issue below the rule diagram [on Harold Boley - due 2007-03-06].

Detailed CORE WD walkthrough

chrisW: someone edit wiki page as we go?

Harold volunteers

chrisW: focus mainly on green highlighted issues and respond
... address the first on for WD1?

csma: could it be resolved by just adding a sentence
... ?

mk: could say that dialiect is a logic-based language

csma: prefer 'rule-based'

chrisW: remove 'rule-based'?
... doesn't bother me

harold: rule language?

chrisW: 'rule-based' aand remove green

second green issue - fix agreed

first issue in section 2

scribe: The following paragraph should be elsewhere.

chrisW: remove following paragraph

correction - just remove para in green

daveR: some ed corrections - e.g. wrong URIs and suggestions for rephrasing
... para below links, strike para re. examples

mk: in core - have we decided?

chrisW: just strike examples?
... talking about eaxmples as well as core
... Delete blue text and presceding sentence
... fix "to support the web ..."

mk: will do off-line

harold: the parenthetical remarks
... remove "striped" and related issue

first green issue in "SYNTAX"

chrisw: remove reference to stripe skipping?

csma: BNF is instantiated into concrete syntax
... but we need to explian that it is not a transformation
... does not belong in the WD anymore

MK: ahreed that metamodel cannot be used to generate syntax

chrisw: do not have to explian the algorithm

csma: but may have to add some comments

second green issue in SYNTAX

<Harold> The concrete human-readable syntax, described in BNF, is: work in progress and under discussion. (It was already resolved as being For Illustrative Purposes Only).

chrisW: new para before the BNF box
... and delete the green

mk: it needs to be there

csma: we know we need to fix it

chrisW: if we have a BNF syntaxt it needs to be a good one

next green issue "Currently CONSTNAME is undefined..."

chrisW: move to next

next green issue "Should we allow certain special characters ..."

chrisW: can remove criticisms of BNF - we know it has to be fixed

harold: anonymous veriables were rejected

csma: we can deal with the action later
... we can deal with issues and remove some of the colored text, but not all actions

Semantic Structure

csma: blues boxes to end notes

chrisw: we should merge conditions with 'rule' section
... found section names confusing

harold: remove parentheses

csma: cannot see different levels in headings

chrisW: need to raise the levels
... need to see what are subsections of what

csma: can it be done offline?

daveR: "Other primitive sorts that are likely to be incorporated include long, double, date, and duration."
... delete 'duration'

mk: is needed

daveR: we will fixit but xsd:duration is not the answer

Issue "Need to provide BNF and XML syntax for arrow/Boolean sorts here"

MK: remove issue

issue: "Need to decide if sort symbols are also coming from Const."

harold: action 247

mk: did not decide where to define sort URIs

<sandro> CURIE reference seems to be http://www.w3.org/2001/sw/BestPractices/HTML/2005-10-27-CURIE

chrisW: delete all three green issues
... we have sections on sorted and unsorted core

csma: we decided some weeks ago to do this

chrisW: unsorted core semantics are irrelevant
... there only for explanation
... requires a big fix to move from 'how to add sorted to unsorted'

harold: add a subheading?
... main heading 'Semantic Structures' applies only to first para

<scribe> ACTION: fix heading structure on MK [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action11]

<rifbot> Sorry, couldn't find user - fix

<sandro> ACTION: mkifer to fix heading structure [recorded in http://www.w3.org/2007/02/27-rif-minutes.html#action12]

<rifbot> Created ACTION-252 - Fix heading structure [on Michael Kifer - due 2007-03-06].

mk: what are W3C onventions for headings?

csma: have to check

chrisW: Now in 'rules'

"RIF RULE LANGUAGE"

josb: resolved with RIF core to cover Horn logic, not higher order

ACTION on MK to add words on predicates, functions constant symbols, disjoint sorts

MOF/UML metamodel

chrisW: .. extending the metamodel of positivre conditions is show below

"SYNTAX"

chrisW: delete text, update symbols in examples

csma: and add words on "workin in progress ..."

"The following extends the mapping in 'Positive Conditions' ..."

chrisW: "The following extends the example syntax in Positive Conditions ..." and delete the DTD sentence

"SEMANTICS"

chrisW: blue text becomes end note

"RIF Compatibility"

chrisW: remove "here" in RIF-OWL and RIF-RDF compatibility

WIKI-TR diagnostics

<ChrisW> hassan, are you there now?

<LeoraM> Hassan got out around 45 minutes ago, I think ...

<LeoraM> I got off the phone around 20 minutes ago or so ...

<LeoraM> It was getting hard to follow ...

chrisW: actions to be completed by ...?

josb: done

harold: at least one week

chrisw: can work tomorrow on this
... we also have architecture and RIFRAF
... new UML diagrams?

harold: not yet

MK: not much time next week

csma: telecon 2 weeks from now?

MK: March 16

csma: for new version

chrisw: what kind of review to accept WD?
... for example - vote now to accept subject to harold and michael completing actions?

DaveR: see frozen doc and vote at telecon

chrisw: telecon on 27 March?
... review is go/no go
... prefer not another round
... can accept subject to typos

csma: what would cause "no"?

chrisW: actions unfulfilled
... no new issues
... working draft to let the world know what we are doing

<sandro> PROPOSED: to publish Core WD1, pending actions performed as discussed so far this meeting.

josb: new material - 2 paras

harold: fix in f2f

csma: have modified metamodel
... whole doc did change

chrisW: but changes agreed

<sandro> PROPOSED: to publish Core WD1, if ACTIONS assigned in this meeting so far are done to our satisfaction. (That is, no new issues should arise to block publication of Core WD1)

csma: clarification - if actions are done, accept document?

chrisW: yes

csma: actions done to WG's satisfaction

RESOLUTION: to publish Core WD1, if ACTIONS assigned in this meeting so far are done to our satisfaction. (That is, no new issues should arise to block publication of Core WD1)

chrisw: any objections to resolution?

RESOLVED

chrisw: new draft for March 16, one week for review

<sandro> expected vote to publish on the 27th.

chrisw: vote to publish March 27

Breakout on external data models

<PaulVincent> scribe: PaulVincent

<scribe> scribe: PaulVincent

External Data Model breakout

Mike: does "external" include OWL etc? yes

Jos: what vocabs are required and how much is required in RIF?
... need for vocab translation as part of RIF role?

Christian: example: shopping cart domain + rules to be interchanged reference domain object model - do they use the XML schema directly or translate to a form for interchange?
... one option is just to adopt a single data model used in interchange -- so burden is on implementer / translation which implies a new translator for each application

Jos: different (use) cases require different treatments for vocabularies

Mike: XML schema can be much harder than OWL/RDF for translators

Christian: an XML schema representing a data model [eg ACORD insurance model supported by rule tools from ILOG and Fair Isaac]

Paul: XML schema for domain specific languages represents a data model + vocabulary for the domain

Reference: http://www.acord.org/home/home.aspx for ACORD / insurance industry

Jos: lightweight approach: rules use vocab with particular URIs relevant to a schema

Christian: problem with this approach: does not fit model ie predicates

Paul: existing BREs use an object mapping mechanism to map disparate object/data/other data models to an OO model referenced by rules

Christian: qu how to map a relational (data ) model to the RIF Condition Metamodel

Andreas: Can use graph-directed model to represent other models

Jos: OWL-DL maps to relational model
... RDF is not just a graph...

Christian: what is OWL compatibility for RIF? OWL and RDF data is a part of the overall problem
... most industry-specific models are relational and therefore can map to the RIF Condition Language metamodel

Mike: ... but the metamodel displayed does not go into the detail for data model issues

Christian: how does RIF hook into externally defined data models?
... mapping an object model into a standardized model may be too expensive from a translator perspective

Christian: ... or can users plug in own data models

Jos: They can already plug in their own models via URIs

Christian: plug-in issue is that the plug-in interpreter takes on the cost of interpretation and needs to be the same on both provider and consumer of RIF

Correction: Christian: enforcement of a relational versus OO versus other model will be a translation issue

Jos: these concerns re Core may be pointless as Core is of limited practicability

Christian: ... but principles apply to all dialects
... assumption that there will be 2 customers who often share data model types

Andreas: RDF - data and meaning layers - may be way to go here

John: issue is that domain specific languages need to be usable directly in order to allow adoption

<johnhall> That wasn't quite my concern. I said it would be unfortunate if RIF actually precluded organizations from using solutions already in place.

Jos: which dialects require this issue

Christian: statement "if they have the same object model they don't need RIF" is wrong as they still need to interchange rules

Jos: ... but you also need things like variables

Allen: is this RIF Core? Phase 2?
... a new requirement not in RIF at present

Christian: need to enumerate mappings for external data

Mike: note even several mappings for RDF and tools like JESS

Christian: Example: XBRL for financial reporting: have a complex structure, interchange rules as text

Jos: propose: 2 dimensions; type of vocab language + degree of integration in RIF

Christian: how do we define compliance if there is a plug-in environment

<LeoraM> +1 with Mike Dean's suggestion to ground this in a concrete example

Jos: RDFs requirements are needed

<LeoraM> +1 also to instantiating the use cases

Mike: need to ground requirements in expanding use cases

Summary of breakout for external data models

Summary by Jos of breakout for external data models

1. Definition of external data models: data structure / vocab eg XML schema or OWL

2. How would data structure be represented in RIF rules

3. Proposed: plug-in for external data models

4. Should not focus on RIF Core limitations ie other dialects may require OO data structures

5. May need special treatments for RDFS and OWL

6. Working group needs some requirements for external data models use in RIF

3. correction: proposal was to indicate range of options from plug-in for arbitrary models to mapping everything to a single Core data model

<allen> dru

Dave: does this include option of eg using a single a URI to reference to what you mean eg complex types

<allen> McCandless, Dru

<sandro> thanks.

Dave: coverage of RDF and XML should cover most options

Christian: need examples to better understand mapping needs

<sandro> RIF Syntax Breakout

Summary of the breakout for syntax

Summary by Chris of the breakout for syntax

1. Different paradigms between metamodels and ASN abstract syntax - metamodel includes items not in syntax

2. Sandro can now generate near-UML diagrams from ASN06 so publication should specify these as "not metamodel"

3. From ASN06 will generate XML schema as XML syntax specification

4. Need for human-readable presentation syntax

<sandro> PROPOSED: We'll use UML to help people visualize our abstract syntax -- but we'll be clear that it's not a metamodel.

<sandro> PROPOSED: We'll use UML to help people visualize our abstract syntax -- but we'll be clear that these UML diagrams are not metamodels

5. Discussion on presentation syntaxes - Sandro will provide some examples to be generated from ASN06 (as "RIF Presentation Syntax")

Hassan: is there a BNF/grammar for ASN06 - yes - so Hassan can implement an XML output too

<sandro> PROPOSED: We'll use UML to help people visualize our abstract syntax -- but we'll be clear that these UML diagrams are not metamodels

Hassan: need semantics for ASN to be able to discuss

Chris: abstract syntax is not normative

<sandro> Chris: I want these not to confuse people used to metamodels.

<sandro> Chris: I want them not to find them lacking.

<sandro> csma: These are graphical views of the abstract syntax using UML notation.

<sandro> Sandro: it's not all of UML, but we what UML we use should be correct.

<sandro> PROPOSED: We'll use UML to help people visualize our abstract syntax. We'll say "these are graphical views of the abstract syntax using UML notation".

RESOLUTION: We'll use UML to help people visualize our abstract syntax. We'll say "these are graphical views of the abstract syntax using UML notation".

<sandro> PROPOSED: we need a presentation syntax

Christian: viewing a RIF Presentation Syntax example: would keep roles not classes

<sandro> PROPOSED: we need a presentation syntax -- to be used for examples and in the specification of the semantics.

Harold: Presentation Syntax is WD2 and later

Chris: this is not normative at this point in time (although examples etc in future will need a presentation syntax)

- DRAFT -

RIF F2F 27 Feb 2007

27 Feb 2007

Attendees

Contents