See also: IRC log
Sandro: Start with a quick round of introductions. People introduce themselves.
People introduce themselves.
<DanC_DCA> list of registrants
Sandro: Focus on interoperability. All kinds of rules, all kinds of use cases. Cast a broad net.
... hosted by W3C, approach through web architecture ...
... approach from consensus standpoint ...
... objective of WS is to come to common understanding ...
... use cases - what's the common base? ...
... look at available technologies, two sessions on candidate technologies ...
... hope to emerge with common understanding how to come to standard ...
... community-building ...
... out of scope ...
... do not design the ultimate rule language ...
... may end up with a rule language, will probably be what you want ...
... focus on interoperability ...
... don't digress into "what's a rule", "what's a rule system" ...
... don't reopen cans of worms on W3C recommendations ...
... acknowledge existing consensus ...
... can get feedback for next versions, but respect existing consensus ...
http://www.w3.org/2004/12/rules-ws/submissions/dynamic/Program
Sandro: Rules of the game ...
... short presentations, then simple, clarifying questions ...
... each session has 1/2 hour for discussions ...
... handle questions of less general interest in the breaks ...
... there'll be scribing ...
... could use volunteers ...
... things said here are part of the public record ...
<DanC_DCA> scribe volunteers should find Sandro at a break
Christian de Ste Marie: Please give your name when speaking.
Sandro: Think time is right.
Anthony Finkelstein: Would like to see demos.
<kendall> Anthony Finkelstein asks for a demo time
Terry Moriarty: Large non-DoD govt agency has decided to go for business rules ...
scribe: got e-mail from project manager - "standard rule language next week?" ...
... lot of interest ...
... highlight that standards in one area don't live in isolation ...
... events, object state, business rules ...
... business rules are controls org puts in place to ensure consistent behavior ...
... about controlling behavior, ensuring transactions occur consistently ...
... maybe different from what semantic web is looking for ...
... looked like they were talking about data mining ...
... business rules about controlling behavior ...
... business perspective ...
... and technology perspective; focus on biz perspective ...
... use vocabulary of businesses -- whatever terms they use ...
... make sure information associated with tech environment ...
... associate with process ...
... do we start with process, informationrules? ...
... start with the process, drive to the rules ...
... policy and procedure manuals, product specifications, sometimes go to the code ...
... that's the kind of thing we look at ...
... rules ought to be expressed so they can be shared ...
... regulators! ...
... technical rule is something that supports biz rule ...
... clean differentiation ....
... technical rules expressed in whatever the system uses ...
<kendall> new synonym for "legacy system" -- "the current processing environment"
scribe: legacy -- ups, verboten -- that's the current processing environment ...
... reality: majority of applications in assembler in some corporations ...
... important to have standards integrated ...
... using BPMN ...
... for events ...
... OWL for knowledge base, happy with that ...
... rule piece? ...
... business rules: terms, facts, rules ...
... rules capture from business perspective ...
... introduces classification scheme ...
http://www.w3.org/2004/12/rules-ws/paper/35/ includes the table.
<DanC_DCA> hmm... business:elligibility = logic:inference ? hard to see it that way
Terry Moriarty: Eligibility -- WHERE clause in an SQL statement ...
scribe: validation rule -- is knowledge base correct? ...
... stated as a positive, results in a state ...
... exactly one car must be assigned to a rental agreement ...
... cardinality type constraints, domains, minimum and maximum values ...
... have seen people dismiss those types of rules ...
... "just data rules" ...
... important for business ...
... result: state "valid" or state "invalid" ...
... process rules - basic ECA ...
... if in state then this action, else other action ...
... process rule tells what to do ...
... go to process modeling environment; others go to rule engine ...
... authority statement ...
... missing piece: treatment of time ...
... if loyalty club member does 30% less this quarter than last quarter, then do something ...
... comparison of objects at same time ...
sandro: q?
<DanC_DCA> (well done fitting that into just 20 min or so)
http://www.w3.org/2005/Talks/0427-rules-em/
EricM: workshops like this one are an important step forward ...
... going to give semantic web perspective ...
... DanC is going to talk about this later today ...
... put in context who we are ...
... holistic view ...
... intl consortium directed by TimBL, who's here not as director, but as person working in the space ...
... goal many at consortium are focused on is making web work, leading to full potential ...
... folks all around the world at three hosts ...
... working on this from different cultural, country-specific perspectives ...
... intl perspective, not the "English first, i18n in v2" approach ...
... lots of different standards -- HTML, XML, ... -- ...
... create specs that work well together ...
... track record ...
... bring lots of different communities together ...
... http://www.w3.org/2005/Talks/0427-rules-em/?n=2
... http://www.w3.org/2005/Talks/0427-rules-em/?n=3
... Focus not on re-inventing the web ...
... integrate work with existing network infrastructure ...
... lots of different applications ...
http://www.w3.org/2005/Talks/0427-rules-em/?n=4
ericM: about a year old in terms of specs that are basis for sem web ...
... a lot is happening in the past year ...
... pick-up ...
... different communities -- life sci, high ed, govt ...
... breaks down social barriers for exchanging and sharing information ...
... people who have vocabularies that make sense to them ...
... not change, but expose in an interoperable way ...
... web of documents ...
http://www.w3.org/2005/Talks/0427-rules-em/?n=5
ericM: rules as key element of the semantic web ...
... good uptake on query (SPARQL) ...
... make it easy to join data from different sites ..
... seing a lot of good uptake ...
... rules to facilitate integration ...
... strong value in treating rules as data ...
... same sort of framework ...
http://www.w3.org/2005/Talks/0427-rules-em/?n=6
ericM: notion of grounding URIs ...
... like Ben's curteous logic stuff ...
... curteous rules, when in a presentation, turn off cell phone ...
<DanC_DCA> (hmm... "if you're in a theater, turn off your phone" isn't declarative)
ericM: don't have mobile devs agree on vocabulary, but have framework ...
... stitch rules in context that makes sense to them ...
... by basing in terms of web, leverage existing infrastructure ...
Sandro: What's a URI?
ericM: http:// ... -- click, and you get stuff back.
... use it as a network key ...
... Ground in context of web. Take advantage of web as delivery mechanism ...
... begin to take advantage of tools ...
... turn owl editors into rule editors ...
... build databases of rules ...
... when built rdf core: "is anyone using this"? ...
... use Swoogle to find out ...
... "what rules apply to my data" are questions you can start to ask ...
http://www.w3.org/2005/Talks/0427-rules-em/?n=7
ericM: Lessons learned -- simple things simple, but complex things possible ...
... simplify technologies, but make it possible they scale ...
... rule requirements from cell phone and router industries ...
... few here ...
... but their requirements can be addressed by rules ...
... W3C context ...
... P3P/APPEL, CC/PP, Semantic Web Svcs, ...
http://www.w3.org/2005/Talks/0427-rules-em/?n=8
ericM: Architecture, RDF, OWL, W3C recommendations ...
... build on existing infrastructure ...
... recombinant rules ...
... social reasons ...
... if we think we have a scope today, when it goes to a WG, we're going to get more requirements ...
... use architecture as grounding ...
... reference points that may help technical and social interaction ...
... minimize misunderstandings ...
http://www.w3.org/2005/Talks/0427-rules-em/?n=9
ericM: Talked to early RDF people ...
... free data from application ...
... it used to be the application, now it's the data ...
... apply rules and use it as web data will lead to the same benefits ...
... that's it.
Adrian...: Distinction between business rule and technical rule notations ...
... in Terry's talk ...
... ericM makes point to make data explicit ...
... didn't touch on Terry's distinction ...
... getting in to same thing where app knowledge is locked up at technical rule level ...
ericM: Valid observation...
... deployment issue ...
... transmitting across the wire, communicate with business folks ...
... have to permit transition from current systems.
Fensel: Integration process?
ericM: ... business process aspects become more sharable ...
... implications? don't know ...
??, Fannie Mae: [[ ununderstandable ]]
EricM: NYT article on IBM, patent strategy, standards work ...
... embrace open standards ...
... important shift ...
... in how they embrace open standards ...
... value of open community ...
... worth reading ...
Dave Goldstean, Freddy Mac: q -- What is being done? suggestion -- having done some of this work for telecom ...
scribe: clients' reaction, "way overkill" ...
... this is not Java or C++ ...
... advocacy and training needed ...
... preaching to the choir right now ...
... from W3C process, how buy-in to get people to volunteer ...
... to encode legacy apps in this? ...
... what's done by W3C to facilitate buy-in ...
ericM: First thing, no good suggestion goes unpunished. Good and important issue ...
... issue the community tries to grapple with this ...
... come up with best practices, educational material, ...
<DanC_DCA> "no good suggestion goes unpunished" is colloquial for "good idea; I hope to recruit you to work on it"
ericM: there are advocacy and education aspects of this ...
... bring early adopters together ...
... information of what works and what hurts ...
... starting to be a shift from where we were several years ago ...
... time's up ...
<timbl> _________________________________________
Michael Kifer, State Univ of New York at Stony Brook
Presents semantic web stack with rules on top of RDF(S), besides OWL.
MichaelK: Some overlap between OWL and Rules ...
... other exensions on top of this? (FOL++) ...
... mistakes of the past -- prolog ...
... semantics, not really declarative ...
... not easy to use ...
... few features ....
... swsl-rules, flora-2 ...
... feature laundry list ...
... adding frames -- f-logic ...
... meta-programming - 2nd order syntax, but not semantics ...
... HiLOG, rediscovery in form of SKIF ...
... logical updates ...
... incorporate actions ...
... ECA rules / triggers ...
... transaction logics ...
... powerful logic, can express planning strategies ...
... useful for sem web svcs ...
... annotated logic ...
... easy to use, paraconsistency, naturally combines with rules ...
<DanC_DCA> -> http:reliant.teknowledge.com/IJCAI01/HayesMenzel-SKIF-IJCAI2001.pdf A semantics for the knowledge interchange format P. Hayes and C. Menzel 2001
MichaelK: capable of dealing with time ...
<josdebruijn> Just to make this clear: SKIF is a rediscovery of HILOG (was not really clear from the notes)
MichaelK: constraint logic programming well-known ...
... web-izing: URIs a matter of syntax; modular system ...
... independent theories that interact in controlled ways ...
... underappreciated issue in web rules community ...
... only 2 or 3 systems doing it or doing it right ...
... scalable web rule language => issue has to be dealt with. ...
NIST: re annotated logic -- fuzzy logic?
<DanC_DCA> A semantics for the knowledge interchange format P. Hayes and C. Menzel 2001
MichaelK: Well-defined term, started 1985/84 ...
<DanC_DCA> re SKIF
MichaelK: basically a logic which assigns confidence factors to facts and rules ...
??: Marrying XSB and OWL -- different semantics?
MichaelK: That's why 1st slide had rules besides OWL.
... no known or acceptable semantic framework which can unify rules - important aspect is nonmonotonicity - not clear how to combine that with OWL which is based on monotonic FOL ...
... some things better done monotonically, some things better non-monotonically ...
Herry Halpin, Univ Edinburgh: Still teach PROLOG to students. Prolog 2 for Web would be fantastic for students ...
scribe: but what about functional paradigm? ...
... Curry? ...
MichaelK: Good feature to add. Not on radar too much ...
<DanC_DCA> A Truly Integrated
<DanC_DCA> Functional Logic Language
MichaelK: ability to specify functions is useful ...
<DanC_DCA> A Truly Integrated Functional Logic Language
MichaelK: WSML -- limited way to add functions ...
??: XSB -- tabeling, when to use, when not to use? Automatic?
MichaelK: Research.
Ian?: F-Logic as basis -- F-Logic and SKIF same thing
MichaelK: no, HiLog & SKIF same thing
Ian?: So HiLog and F-Logic not same thing ...
scribe: SKIF v. FOL?
MichaelK: Talking about rules language. Some things don't depend on negation as failure, so can extend FOL.
Sandro cuts of this discussion.
[[ some discussion of logics and decidability ]]
MichaelK: ... if you want to do interesting stuff, have to go beyond decidability ...
... that's the point with programming languages ...
DieterDERI: Nonmonotonicity as reason not to layer rules on top of OWL ...
... existing efficient rule languages ...
... but when you put them on top of OWL, you get powerful bastard language ...
... also, it would mean everyone who wants to implement rules has to purchase DL reasoner from Network Inference ...
Phil Archer, ICRA: Metadata important for the web. Advocacy. Rules have to be business related ...
scribe: million-a-hit web sites have no metadata? ...
... maybe not enough advocacy, or maybe not enough business value ...
DanC: Well, there's just the normal amount of metadata -- titles, standardized way to write links ...
PhilA: not a lot of Dublin Core...
DanC: A little bit goes a far way.
EricM: Simple tings simple.
DanC: Question to Terry -- how's eligibility like inference?
Terry: The word Metadata is horrible and meaningless. Metadata is just data.
... a lot of info we think of as metadata is business data -- product description ...
... expect to see it behind commercial web sites ...
... how does it matter that we don't see it? ...
PhilA: Would like to build things on top of that -- where's all the web sites written by X?
Terry: Tags behind the web site, rather than visible?
PhilA: Link to it on the source code.
Terry: Not sure what you meant. Biz rules -- inferencing as process of going through data based on rules, coming up with new information ...
<DanC_DCA> (hmm... what does google use for the title of an ms word or pdf document?)
Terry: preferred customer is customer who has 20,000$ purchases ...
... need inferencing to go over database and find these customers ...
[[ ununderstandable question ]]
Terry: From business perspective, use eligibility rule. Don't care what's used to find that out.
NIST: Class vs. class membership rule.
Terry: Find a lot of people use terms -- define customer, preferred customer.
... separating person from state that person is in is a lot of my work ...
<josdebruijn> about PDF and MS Word: both formats encode meta-data in a proprietary format
Dave Goldstein: Would be useful if people could describe the problem they'd like to address?
<DanC_DCA> but do enough ms word files do it the same way that google can extract it? or does HTML have a distinctive advantage?
scribe: first 50 hits on Google weren't what he'd looked for ...
... simple things simple ...
... considered antichrist for teaching Prolog ...
... to Kifer, what's the business rules applications beyond what Terry is solving?
... lots of smart people, but they often design things that less smart people shy away from ...
... haven't ever seen 2nd order logic in actual application ...
<josdebruijn> All MS Word documents do it the same way (file->properties), but people do not always include the correct title, because you normally don't see the meta-data when editing a word document
scribe: what do you see in foreseeable future? Where do you see logics come in in terms of requirements?
Michael Kifer: Problem is -- prolog is good language, but doesn't speak language programmers speak ...
scribe: people know how to do OO programs ...
... tell them to use prolog, and they are lost ...
... will feel more at home when writing a Java problem with some rules ...
Goldstein: Terry's users are business people, Kifer's users are programmers, business people don't write programs?
... so, what features of your logics would an MBA-background person look at?
<DanC_DCA> (spreadsheets were a breakthru in allowing business folks to represent their knowledge in such a way that machines could help them with it; i.e. programming)
<josdebruijn> For example, google for "semantic information integration" and check the 7th result, which is a ppt document with the incorrect meta-data
MichaelK: Biz people write programs? Vision of managers typing SQL statement 30 years ago ...
Terry: ... but then business intelligence tools came out that put front end in front of SQL.
... so free the data from the warehouse, so MBAs can write queries without going to IT ...
... majority of logic done is very simple ...
<DanC_DCA> [... discussion gets pretty interactive at this point... scribe is doing well to get bits and pieces ...]
Terry: build rule language for biz people, save complexity for programmers ...
MichaelK: If logic is not complex, no point to this. Go to SQL when it's as easy as now? (??)?
Goldstein: >= 2 types of users -- tiers of interoperable languages?
... big interest in having people who are domain experts in their industry to use these languages ...
<DanC_DCA> (heh... Goldstein's paper is listed as (no title) )
Goldstein: more complex problems that represent big value for businesses ...
Fannie Mae: Business people don't write programs? We have biz analysts who write rules.
[[ hard to understand ]]
scribe: compare requirements laundry list to business requirements ...
<DanC_DCA> ?: there are analysts writing rules... [missed a lot]
MichaelK: Not thinking of business people at this point.
??: Concern that legacy business environment isn't webified. Get at that even with webified rule language.
EricM: Exciting apps where people use URIs inside enterprise.
... fact that you can put web server or infrastructure in place inside firewall ...
... web tools reduce cost inside the firewall ...
... using URIs doesn't imply global access ...
... don't think just because there's URIs that means everything is free...
??: That's not the question. Rules language that speaks web services only? Nice if it had ability to speak IP natively?
scribe: something that's not so web specific ...
Terry: My applications not web based, either. Needs to be independent channel.
TimBL: all kinds of information in unified format. SPARQL ...
... SPARQL query from whatever API ...
EricM: Don't confuse URLs and URIs.
... using URI specification for minimizing name collisions ...
<DanC_DCA> er... hey... do confuse URLs with URIs! make your URIs defererenceable. use http and DNS
??: Take all of GM's data, URI it.
DanC: That's how the web started.
ericM: Don't boil the ocean, do it as you go.
... getting back to advocacy side ...
... make implicit names explicit when you expose data ...
Danny: Does response answer question?
Adrian?: Pick up on what tim said about wrapping legacy data ...
<DanC_DCA> DanC: do a little bit at a time. that's how the web started. a phone book here, a few physics abstracts there, and with each bit you get a little more value. [in particular, my "that" did _not_ refer to "take all of GM's data, URI it"]
scribe: seems to happen that in process of taking things out of SQL which has > 2 columns ...
... is shredding data ...
... get to point where you can ask the same questions, but how you ask the question gets more complicated ...
... have to unshred the data in the inference process ...
... less unshredding needed with columns ...
... conceptual tradeoff between wrapping legacy data and shredding it (??) ...
... if you want someone to look at an answer from shredded data, following inference through to RDF is more difficult than following through to legacy SQL ...
... trade-off that may happen ...
... RDF has tremendous advantages ...
... tradeoff, having business-level explanation of proof tree...
Bijan Parsia: Doesn't business rule stuff fall more naturally to W3C in the web services space?
scribe: do we have the right parties at the table?
<DanC_DCA> BP: IBM's line on BPEL is "it's supposed to do for apps what SQL did for data"
Said Tabeh: Optimistic. Don't think there will be KR v. biz rules divide.
scribe: re BPEL meeting on biz processes ...
... in that community, people also have requirements for rules ...
... many people not looking to create yet another rule language for yet another area ...
... interchange language ...
Gary (Network Inference): Came in late, missed Terry's talk. Would, from discussion, agree that biz user don't care about rules. Simple syntax.
<bijan> It is my prediction
scribe: real use case ...
... client with 100,000 products ...
<bijan> If it is correct I want us to reconize that fact as soon as possible
scribe: biz analyst has to write rules to convert this into financial data.
... rules in spreadsheet, then converted into SQL ...
<bijan> And I have no investment in the prediction. I'll be interested if it comes out false
scribe: rely on sequence of execution ...
... working with them to see if can use rules and DL to solve their problem ...
... don't want to know about SQL procedures etc ...
<DanC_DCA> (note to self, Q for Gary at the break or something: what did those spreadsheets look like? the ones that got converted to SQL stored procedures)
?? Abdullah: Lot of work XML schemas in vertical industry std orgs for business transactions ...
scribe: convert all these schemas to RDF? ...
... to make use of new technologies? ...
... approach or direction or any kind of suggestion? ...
EricM: If you can get people to agree on XML schema for one kind of transaction, that's great.
... if you have a closed world problem, and can get ALL THE PEOPLE in the room, then XML schema is great ...
... suggested process is to annotate, wrap schema ...
... so if you want to expose it in ways that others can take advantage of it ...
... annotate in a way that makes RDF available ...
... write things down in an explicit way that others can rely on ...
<DanC_DCA> (what ericm's talking about is the subject of some papers in progress... one to be presented at XTech in Amsterdam in May)
EricM: nice tools coming online -- not changing things you do in relational DBs ...
... but expose them in RDF ...
?? Abdullah: Expose rules in XML, not RDF?
<DanC_DCA> I think he was saying: a rules language should work on XML data that isn't RDF
EricM: Applying rules to closed word data is important. But apply rules *across* rules
Finkelstein: Kifer slid over one rather important point -- whether or not expressiveness is be all and end all in this situation.
... Important business areas where need to exchage simple rules ...
... how to deal with heterogeneous data? ...
... need to discuss this, not just the rule language ...
... also discuss mechanisms that interplay with rule language ...
... how do you manage records, how to you relate them to the rules ...
... when building practical rule language, that may be more important than some of the features of the concrete rule language ...
... third observation about XML and RDF: build on top of what's there ...
... biz users have big schemas, problems managing actions on top of these things ...
... address these pain points urgently ...
Bejamin Grosof: Business rules, knowledge representation.
scribe: terminology ...
... real issue is, to what extent is there overlap with KR? ...
... intersection should be our focus here ...
... a group of things around policies for contracting, ...
... benefits for life cycle mgmt etc. ...
... can be a productive focus ...
MichaelK: Expressiveness.
... 95% of rules people write are simple ...
... But sometimes need that little thing that's not in this sublanguage ...
... that's where the problem arises, and that's one problem of Prolog ...
... there's this one thing you can't write well, and that's where people give up ...
Finkelstein: ... 95% ...
MichaelK: But the 5% are what breaks things.
Christian de Ste. Marie: Could appear that we have different communities, biz rules cmty that wants it simple, v. KR community that cares about logics
scribe: no ...
... it's a continuum ...
... unsolved theoretical issues are of no use to business rules people ...
... on the other hand, have KR people who work on extending reach of expressiveness, while keeping good properties, so biz rule people can use it ...
... another remark on continuum ...
... more overlap between communities, overlap of interest in terms of s13nizing ...
... 20-25 years ago, production rules were KR issues ...
... what biz rules people use for implementing apps are things that came from KR community years ago ...
... What they will use in 5 years is what KR is working on now ...
?? IBM/Rational: Bridging gap between communities will be difficult ...
scribe: understood everything that Terry said, some things that eric said, nothing from Michael ...
... business processes ...
... data is not the only king ...
... in world I come from, process is key issue with customers ...
... important issue: workflow paradigm ...
... one way of expressing process ...
... services as abstraction on how to organize processes ...
<DanC_DCA> (good point about data not being the only king... process is important. Data that isn't consumed rots.)
scribe: rules fit in as way for expressing what services do that are choreographed ...
??: Relationship of rules and OWL. Don't just need ontology, also need rules. Complementary.
Terry: Can't do my job with out *three* pieces. +process
... looking at current process, modeling, ...
... all we're doing is moving the rules to an earlier stage ...
[[ scribe unsure he properly captured Terry's meaning ]]
Ushold: ... used OWL-DL to do semantic filtering ...
... had problems ...
... moved to XSB ...
... compatibility concern ...
... concerned about lock-in ...
... different roads, or coming together ...
... rules on top of owl, and get same performance as with ontobroker?
Stoutenberg, MITRE: Sit between KR and practical biz rules group ...
scribe: support fed govt, provide consulting services ...
... help govt build systems that are extensible ...
... need to share infomation across the govt ...
... hoping we can find overlap ...
Bijan: The gorilla's aren't in the room, aren't interested in expressivity.
... CC/PP not interested in expressivity, want to do piecemeal stuff ...
Goldstein: Comment on what Terry mentioned -- what are the kinds of reasons we have to support ...
... prjs that use biz rules are at reengineering stage ...
... see common themes ...
<DanC_DCA> (bijan, my experience is that the 800lb gorillas are often not the early adopter/drivers. Besides, there are some 400 or even 750lb gorillas here)
Goldstein: how do things differently going forward ...
----
Break.
<scribe> scribe: Thomas Roessler