<tlrDC> Scribe: sandro
HarryChen: Rules for Geospatial
(scribe missing this stuff ---- discussion about whether Fuzzy is necessary, can be done on top, etc. BenG, Vassilis, ...)
Chen: Our prototype reasoners, we tried doing geometry in lower-level builtins. Bridging to Fuzzy, & to OWL is interesting
JonPellant: If you use Fuzzy, your rules are much simpler.
Vassilis: We have a paper on Fuzzy OWL and FUzzy SWRL in this workshop
<tlrDC> Jeffrey Ritter: ...
<DanC_lap> "I don't write code, but I do write rules; I make laws"
JeffreyRitter: I'm intrigued by being invitied here. :-)
... Compliance: PROVE the behavior of a company, data-based
... Requirements from Policy
... "extended enterprise" is crashing -- relationships are not expressive enough
... Contracts often end up hundreds of pages long
... "That which is not recorded did not occur"
... "That which is not documented does not exist"
... "That which is not approved and audited is vulnerable"
... ==> "The company must be able to document, to an objective reviewer, tht compliance has occured in fact"
<DanC_lap> (I have encouraged W3C to operate that way pretty much all along: we agree to what the record says we agree to, no more, no less. Expectations are written down, and renegotiated in writing)
(Indeed. This talk is making me think of your approach, DanC)
scribe: Sprectrum: Constitution, Statute, Regulations, Trade Addns, Copr, Contract, Policies, Bus ules, System Rules
... Matching ISO stack from Layer 1 - Layer ?8? (data)
... Legel requirements for standards is increasing.
... We policy makers have lost the capacity to express ourselves to IT folks
... If you don't appreciate this problem, without expressing the relationship to a hierachichal rule bases society
Wow! We live in a rule-based system. Huh.
scribe: Huge failuer -- the DOZENS of pages attached to any outsourcing-to-India contract, about Security Controls
... You Folks can solve this
PatHayes: How can we do this, if no one will show us their data?
JR: Much data is considered confidential, yes.
DannyW: There is a huge sphere of public data, alarmingly public [ cf rules in IRS forms! ],
... Huge middle sphere, very confused though
... For Jeff --- Re: stack, from broad/vague to completely precise. What's your sense, given that range, how do you think about top-bottom connections.
Jeff: A fuzzy legal requirement ("adequate procedures") requires companies to make justified choices. We can get the fuzzy-rules writers to ANTICIPATE that the rules will be written in an IT environment.
... EG Banking rules require "adequate procedures" to verify customer identity. THEN a 3-page list came out of procedures which would qualify.
Kurt: Rules expressed in contracts, spreadsheets, software
... I want them in RDF, so I can attach them to domain knowledge
<anthonyf> did I hear the previous speaker say he wanted to do program property proofs in a rule language?
Kurt: "Warren Core"
... Vehicle representations, with Warantee information.
... Rules about warantees, volume, ... affect assignment of "red x engineers" to look at problem
... Documented in Lotus Notes app today
... Rules might in theory drive supply-chain re-negotiation
<DanC_lap> (this line of design looks familar. I think I presented something pretty similar at the DAML kick-off meeting. It was not well received ;-) I was... "educated" rather... enthusiastically. ;-)
Kurt: I thought about going to SAE, but that's a difficult process. It's hard enough within GM
BenG: SweetDeal is very much like this
Earlier questions: Bijan: I feel your pain
EricMiller: Are these scenarios (these simple rules), enough to help?
Mark: This is work I did with Donald Ferguson, our chief software architect
... Simple Business Rules, very simple, invoked in many places. if-then rules.
... also as decision tables and decision trees.
... This is 60%, 80%, 95%, .... of the users
<timbl> > 60% of cases benefit from Simple business riles.
Mark: Addition: Event Correlation Rules
<timbl> These can b e visualized as decision tables -- just display all input combinations.
Mark: Inference Rules (Prolog, etc -- KR stuff). They have a place, but these probably are all specialty problems. Not a broad need for one type of rule, so we're not so attracted to it (as a commercial vendor)
... In this space, it takes experts to work with these rules.
... Tivoli Event Console, came out with Prolog inside it, with starter rules. "if db and app server both go down; tell op to fix db first".
... Customers could not not handle the complexity
... So we re-did it as Event Correlation Rules. Customers CAN handle this.
... Inferencing Space's complexity is a real problem in commercial world
... Only needed in ~~5% of problem space!
DaveReynolds: OMG seems like a good match. Does it do what you want here???
Mark: OMG addresses the Simple Business Rules stuff. They dont cover ECR. Business Semantics stuff looks interesting too.
PatHayes: Is cant-hack-prolog solveable if customers had better interfaces?
Mark: I think it's intrinsic.
?: Who is going to write the rules?
icra: child advocacy groups, 10 rules, 10m qps
fannieMae: business analysist, tech folks, 100-1000 rules
cjm: Engineer with Subj matter experts, <100 rules, maybe 500; very fast but no numbers. maybe 10-20 ips
chen: geo rules -- 20ish, domainspecific -- domain experts; performance low level fast
Mark: very simple if-then, so why rules tech? Volatility! Who changes them? generally business people, to improve agility, and BP-IT communication gap. #rules tables may allow a low. perf req: trading systems need to be very fast.
Kurt: dozens-hundreds of rules, PER DOMAIN. It would be nice to have domain engineers write the rules; may also need a knowledge engineer working with them.
... Ontologies will be enormous
Jeff: Simple authoring tools are what's missing. Large.
Doug: hundreds of rules, skilled developer type for some types; you'd like project managers to be able to write the rules. Agents need speed
DaveR: 100-1000s rules, some web-based apps need more speed than we can deliver
Ed: (1) Need a standard for tool-independent-archiving
... (2) Exchange
Mostly A 4
SOme B -- all?
<DanC_lap> (er... when there's a poll, pls have the chair check how it's recorded, near-real-time)
Tim: Context? Is it reasonable to think of publishing rules, with pre-conditions in ruleset?
Mark: XACML has targets, like that
Ed: first decision in a decision tree
Chen: Good feature -- but what is a context??
... eg we have Location/Time context, but hard to do.
Ritter: Think about Open Source rule bases
... Payroll Service Provider, offers their rules, makes deal much faster.
<DanC_lap> +1 Ritter
Ritter: Phil: in germany, no nazis, no porn before 10pm. AOL conforms, etc.
Mark: These rules attach to the following part of an ontology
TimBL: Isn't that the same as "and the time now is > 10?"
<DanC_lap> there was initially a "the internet doesn't work that way' in response to the "before 10pm" idea, but AOL eventually made it work
BenG: Interesting, Jeff. We proposed that at eGov'99
... How much energy is there for govt agencies to drive this? Patient Records Protection.
... If the political process wants to get back in control, then it has to do what you;'re talking about -- but how much will is there? I dont see much energy to drive IT.
Jeff: NationalUS - "Late Adopter" is a favorable comment
... SoftLaw in Australia; using KR on laws. So some folks have momentum -- Competitive Advantage between nations/regions is imporant here.
... Could be Industry or Industry+Govt.
... Smart Developing economies are the place to look; USA not likely to be a leader here, as it's looking right now.
Doug: A11 is a 300 question business case
... $69 e+12 IT investment; OMB work in this space
BenG: historically, a signal comes out of an agency that a mandate will be coming. Where will that be happening next?
... the US seems strangely lagging here.
CSMA: folks with a need to share rules (Ed's case B), what is your tpyivcal case?
PhilArcher: German law has age bands -- which would be published by a suitable org, for all germany.
Srini: Parterns might not be outside org, but diff. groups inside org. We see rules for underwriting need to be met AGAIN during aftermarket sale.
... So that's implemented in two different systems. Right now that involves duplication of effort
... Could be done with internal proprietary language for now, but not if seconary market is eg FreddieMac
CJM: assumption is all users will be using same engine
Chen: I want users to use my engine, but I want to spread the cost of developing the laguage, and get tools users on board
Mark: Avoid vendor lock in!
... In web services policy -- I need to tell annotation about policy to tell down-stream users, eg "kerberos needed"
... If you use my services at this time, it'll cost that. Or pay me one of these three ways. Publishing those rules
Kurt: Need to send rules to suppliers, and would like to monitor processes
Jeff: Don't just tell me you use Kerberos -- show me evidence.
Doug: Web Services Based Rules is how standards should be published, obviously.
Jeff: Not enought to say "we are secure" -- need drill down into details.
csma: that's a different purpose -- analyze rules, not just execute
DaveR: that second column in my talk.
AnthonyF: FPML rules are published on their website, along with test cases. (financial)
... In natural language, with XML text cases
Ed: Open Source Rule Sets. We need a language for that.... Who knows who all might use it?
DevasisBassu: Ordering of telephone services; cost per order; if you fill out the form badly you get charged extra -- so there's a market for people to check over your forms. If the rules could just be published!
... It makes business much more easy.
scribe: Industry slump; they want us to open up more.
<anthonyf> FPML rules reference implementation http://www.systemwire.com/FpML/2003/FpML-4-0/validation/2004-04-02/rules-english-shared.html
scribe: XML adoption recent
... Huge demand from ILEC
... They want to lower their costs
... Rules are in every aspect of Telecom
<anthonyf> FPML business rules from ISDA http://www.fpml.org/2003/FpML-4-0/validation/2004-04-02/
JeffR: competitive advantage
... business effcieincy
... but reduces customer loyalty (lock in)
Ravi: What would it take to create manageable rule sets...
<DanC_lap> (er... "which all of your are welcome to participate in"... I wonder if that sets expectations a little high... how?)