Slide: Semantic Community
... 'Definition'
ISO 1087-1 and 24707 (Common Logic)
Slide: Business Vocabulary
separate meaning from notation
non-normative SBVR structured English
enforcement requires rules about rules
Slide: Semantic Formulation
... Seamntic Formulation of a Simple Rule
strongly linguistics oriented
integrated with formal logic
Grosof: RIF must be able to support communication of non-automatable rules? comment field? support?
e.g. must wear hard had on construction site
enforcement can't be automated
Grosof: perhaps different term that non-automatable
can analyze for ambiguity, inconsistency, etc. even if machine can't enforce
?: rules should be in human readable form - XML is basically binary
<hak> who spoke?
<MarkusK> Igor Mozetic
Speaker: Gary , Oracle
DB glossary
Slide: query
optimize using algebraic rewrite rules
Gary Hallmark, Oracle
Slide: view
... integrity constraints
conditions that must hold at transaction boundaries
perhaps add glossary term for transaction
<sandro> glossary+ transaction
constraints resemble queries - same syntax, different interpretation
Slide: triggers
... how to evolve the KB
... role based access control
security
in scope for RIF?
Grosof: use of rules for security
policy?
... kitchen sink, without distinguishing layers
... is this satisfactory?
Gary: arbitrary security requirements, minimum result set size, etc.
Barkmeyer: hoping for more about
transactions
... can't evaluate rule safely until everything available -
unitary change
Gary: rules in database vs. rules
in middle tier
... ECA rules can't impact other rules
<hak> There is no such verb as "to inference"! :-)
Speaker: Chris Welty leading open discussion
Welty: What are we doing? How are we doing it?
Boley: scalability to scope of
the web?
... hard to control or debug, particularly for active rules
(production and reactive)
Welty: scalability is more of a requirement
RIF focus on interchange, not defining a new language
understanding will emerge over time
should be less fuzzy as a result of this meeting
tomorrow reserved almost exclusively for discussion of use cases and requirements
Boley: are we concerned about efficient execution of interchanged rules?
<sandro> Jon Pellant
big difference in scope for interchange of rules vs. rule instances running in live environment
Mala: like to see support for evolution of large rule bases
Welty: sounds like requirement
Cory Casanave: what isn't a rule? process modeling as rules about behavior? every UML element?
Dave Springgay: usability, human readability, authoring - outside RIF scope
Hassan: different conceptions of
formal semantics logicians as model theory semantics - doesn't
have to be model-theoretic; other formalisms possible
... rules discussed today mostly involve inference, vs. just
flexible way to express computation (production rules)
... just changes of state
... business rules don't require inference, just agile
computing - useful to distinguish evaluation/execution from
forward/backward inference
?: not a response but elaboration - what are we going to do in this working group
<MarkusK> ?^Pascal Hitzler
thanx
Grosof: think we should identify
KR expressiveness and type of generic computational tasks
associated with clusters of systems and use cases, then
identify superset of subsets that enables significant
translation and interoperability between pairs with
well-understood semantics
... model theory is good at specifying these functions
... could support multiple KRs - e.g. SWRL sits on cusp between
Horn FOL and Horn LP
... behaviors/ground conclusions often identical
... a lot known theoretically - should have 1 or 2 umbrella
KRs, at least for this phase
... perhaps annotated comments
Welty: clusters of existing rule
languages and systems (with possible overlap)
... classifying existing systems
... may not achieve 100% interoperability - standard approach
is to cluster things with common assumptions
Sandro: keeps the market for extensions simple (package per cluster)
<sandro> Is the speaker Joshua Engel?
yes
Engel: also accommodates non-rule systems - queries are a special case of rules
Barkmeyer: different tasks:
inference for the purpose of creating knowledge vs. immediate
information vs. validation of information set
... tools tend to cross between tasks (DBMS as kitchen
sink)
... 3 related but different axes
Welty: not hearing any dissent on classifying systems
Bijan: not sure - still
assimilating
... worried about scope - job of working group to survey
existing systems?
Welty: necessary for interchange?
Sandro: sides with Bijan - those communities are here, don't have to worry about others until Last Call
Bijan: would rather start with rule systems represented here
Christian: agree - but if some community is missing, may need to look outside for extensibility
Sandro: how does this relate to
use cases? top down vs. bottom up?
... meet at features (systems) or requirements (use cases)
Ginsberg: clustering good -
ontology about rules or reasoning may be helpful
... capturing features (e.g. types of nonmon) may help avoid
confusion
Kifer: series of languages?
interchange implies multiplicity
... define language(s) underlying format
Barkmeyer: define abstract concepts
Kifer: and concrete syntax
Welty: classification could drive requirements for syntax and semantics of interchange format
Kifer: OK
Bijan: use cases for rules vs. use cases for exchange
MDean: +1
Welty: most use cases for rule
language not interchange between rule languages
... use cases fairly standard in standards efforts - to assess
coverage, etc.
... no use cases focused on interchange
<bijan> http://lists.w3.org/Archives/Public/public-rif-wg/2005Dec/0055.html
Dieter: use cases provide guidance on relevant features
<bijan> Possible use cases from a interchange perspective
<bijan> I don't agree with that :)
Grosof: agree, but how
valuable/likely is exchange of certain rules?
... e.g. context required to run aprocs
... counter-argument is combination of local context with
exchanged information - need to semantically integrate
Dieter: too limited notion of exchange - could be within processes/vendors locally
Grosof: agree
Paul: need to define "exchange"
Bijan: need to prioritize features?
Ginsberg: use case showing
exchange discussed tomorrow
... also compare consistency of rules
Hassan: 1+
Welty: enumeration problem - use
case for each rule engine
... e.g. what kind of interopability can be achieved between
systems with different semantics
Boley: agree - should combine use cases - compare translated results - multiple groups working on same domain (e.g. bioinformatics)
Hassan: translate between
formalisms preserving structure (homomorphism) - PRR started
with this (meta-model)
... same issue here - PRR approach makes continuing sense
homomorphism can be forgetful - exchange only patterns
Hassan: first describe objects in common
Cory: relative stupidity of
meta-models can be an advantage - simplify synthesize common
meta-model
... use cases for rules have probably been accommodated in
existing languages
David Springgay: OMG goal to identify core then extend
David Springgay: solves goal of interoperability between vendors
Dave Reynolds: what do rule vendors expect from RIF that PRR with XML serialization doesn't provide?
Bijan: or RuleML?
Paul: excellent question - PRR
aimed at OMG and modelling - clear need for more real-time
non-modelling interchange closely related to modelling
interchange
... RIF should be run-time interchange format based on PRR -
production rule support required, otherwise irrelevant
... RIF in XMI possible, but other formats probably better
Reynolds: concrete syntax for PRR would meet requirement? semantics requires?
Paul: no concrete syntax for PRR core (main meta-model) - what other concrete syntax will be required for RIF
Welty: need concrete syntax for rule interchange - phase 1 for core subset
Paul: meta-model describes only
semantics - e.g. doesn't define expression language (e.g.
RDF)
... concrete syntax provides hub
Grosof: what semantics will PRR
have before RIF arrives?
... semantics expressed at level of meta-model rather than
logical conclusions and actions from premises
... should RIF provide that kind of stuff?
finer/deeper/different level
Welty: agree - PRR has different type of semantics
Barkmeyer: model-theoretic semantics?
Welty: not necessary - but want to ensure same conclusions
Paul: rule vendors won't change engines
Welty: semantics should be specified, allowing rule engine to know how to translate into their system
Paul: requirement is purely compatibility - OMG uses annotated meta-model
Hassan: converging - Structured Operational Semantics (SOS) are rules that transform syntactic constructs - formal semantics
<sandro> http://www.cs.vu.nl/~x/sos.html Structured Operational Semantics
<sandro> by J.C.M. Baeten and C. Verhoef ?
<hak> SOS is a formalism due to Gordon Plotkin
<sandro> http://portal.acm.org/citation.cfm?id=199472 ?
<hak> yes !
<sandro> G. Plotkin. A structured approach to operational semantics. Technical Report DAIMI FN-19, Computer Science Department, Aarhus University, 1981. 15
Uli: how many semantics of how many vendors of how many engines?
<hak> http://citeseer.ist.psu.edu/plotkin81structural.html
Welty: scope of this WG is primarily its membership
Sandro: charter discussed standard and non-standard (vendor specific) extensions
Welty: non-member rule engine features could be addressed when designing extension mechanisms
Grosof: any Prolog vendors here?
<phitzler> Concerning SOS I have some doubts whether this would cover e.g. non-monotonic logical semantics and other things important for logic-programming based rule languages. This is probably rather a case for research?
Paul: every PRR vendor has features not covered by PRR - subset/least common denominator
Engel: OntologyWorks is Prolog-like
Grosof: nice to interoperate with pure Prolog
Christian: need phase 1 use cases to identify extensibility mechanism requirements
Welty: how to proceed? classification of existing systems?
<hak> About phitzler doubts re: SOS and non-motonicity - no problems at all. Read Plotkin's paper (cited above)
Sandro: worked some on "feature matrix" over summer, inspired by workshop
Bijan: could proponents provide (preferably semantic) descriptions of their languages, particularly vendor-unique features
<hak> I second Chris Welty's "let's classify systems" initiative.
Bijan: free licenses helpful too :-)
Welty: need some documentation/understanding from vendors
pfps: user documents generally don't describe semantics - evocative not descriptive
<MarkusK> Offtopic: where are the logs of this chat to be found?
Ginsberg: benchmark results may apply here
Bijan: what parts of languages aren't usefully covered by PRR, RuleML, etc.?
Welty: want clear agenda for
tomorrow - sense relative satisfaction with
classification
... what about use cases? not interchange-centric? higher-level
use cases?
Christian: should include test cases from specific applications as use cases
Dieter: 80/20 rule - RIF needs to support common constructs
<sandro> hand raised for you, hak
Grosof: use use cases to weight different features
Dieter: strange RDF features to cover PICS use case - over-design
Welty: use cases not carved in stone - don't commit to cover everything
<GaryH> download a beta of Oracle Biz rules at http://www.oracle.com/technology/products/ias/business_rules/index.html
<GaryH> in the .zip file, rl.pdf is a "spec" of the java-like technical rule language
?: half of use cases can be satisified by web services
?: when do rules need to be explicitly interchanged
?: constraint programming engines also important
<MarkusK> ^?^Igor Mozetic^
<MarkusK> Last ?="Philippe Bernard"
<csma> Philippe Bonnard
<MarkusK> Oh, soory
Gary: 2 types of scenario: interchange (vendor/company/tool), rule use cases
Engel: another scenario -
vendor-neutral authoring tool (e.g. Eclipse or
SourceForge)
... need to motivate exchange of rules - some engines may not
be able to handle even if properly translated
Boley: collaborating rule engines on large distributed problem
Welty: challenge authors of existing use cases, etc. to think more generally and about how they relate to classification
Pellant: won't be here tomorrow - use case: very wide set of behaviors
Welty: please write up
<hak> who spoke?
Bijan: don't want interchange language, want Semantic Web Rules Language - others may also - how to accommodate this
Welty: have this conversation at
end of tomorrow, after more clear expectations
... view question as largely irrelevant
... interchange already complex, without introducing
divide
... can phrase desire as use case or requirement
Kifer: RIF not possible without defining (web) language
<hak> I disagree!
Welty: gather requirements based on need to interchange - resulting spec can be viewed as language
Bijan: different between rules language and (semantic) web rules language
^different^difference
Welty: RIF will be web-savvy interchange format (URIs, interoperability with OWL/RDF)
Barkmeyer: distinguish desire for
interchange with what to interchange
... focus on capabilities
Pascal Hitzler: ambiguous terms between communities: model, etc.
Welty: on-going and evolving process - can adapt as necessary
close of meeting
start promptly at 9am for "quick" 3 hour roundtable
This is scribe.perl Revision: 1.127 of Date: 2005/08/16 15:12:03 Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) No ScribeNick specified. Guessing ScribeNick: mdean Inferring Scribes: mdean WARNING: No "Topic:" lines found. Default Present: JeremyC, +1.650.696.aaaa, +1.604.930.aabb, meeting_room, jeremy, Giorgos_Stamou, hassan, Sandro, PedramAbrari Present: JeremyC +1.650.696.aaaa +1.604.930.aabb meeting_room jeremy Giorgos_Stamou hassan Sandro PedramAbrari 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 WARNING: No date found! Assuming today. (Hint: Specify the W3C IRC log URL, and the date will be determined from that.) Or specify the date like this: <dbooth> Date: 12 Sep 2002 Guessing minutes URL: http://www.w3.org/2005/12/09-rif-minutes.html People with action items: WARNING: Input appears to use implicit continuation lines. You may need the "-implicitContinuations" option. WARNING: No "Topic: ..." lines found! Resulting HTML may have an empty (invalid) <ol>...</ol>. Explanation: "Topic: ..." lines are used to indicate the start of new discussion topics or agenda items, such as: <dbooth> Topic: Review of Amy's report WARNING: IRC log location not specified! (You can ignore this warning if you do not want the generated minutes to contain a link to the original IRC log.)[End of scribe.perl diagnostic output]