W3C

Part of the Workshop Session Notes

- DRAFT -

Rules Workshop -- Session 9 -- Closing Panel, What's Next?

13 May 2005

Attendees

Present
Regrets
Chair
SV_MEETING_CHAIR
Scribe
tlrDC, josdebruijn

Contents


<tlrDC> Scribe: tlrDC

(EricM helping)

Panel -- what's next?

sandro: Next steps -- will finish the record ...
... workshop report ...
... then, activity report ...
... to get approval from membership ...
... panel: DannyJW to moderate ...
... Chairs have selected participants ...

Danny: Lightning tour through cross-section of perspectives ...
... ask panel to think from each person's perspective about top three functional requirements ...

<em>

Danny: from perspective of individual communities, not from guessing what consensus may look like ...
... thanks to chairs ...

[[ applause ]]

scribe: ILOG did the logistical work ...
... appreciate DARPA's support ...
... have asked each panel member to present three brief thoughts ...
... order of appearance on the page ....
... will rigorously enforce 3min rule ...

Jon Pellant, Pega: Came here without doing homework ...

scribe: going away with a lot of homework ...
... understand semantic web ...
... taking back that perspective ...
... customer demand: business rules long-term assets, don't get into vendor lock-in ...
... enabling factors ...
... rulesets are a valuable asset ...
... repurposed to diff. realities or fact sets ...
... bindings to fact space? ...
... or rules themselves? ...
... ability for author to be able to publish their own dictionary ...
... rule templates ...

<DanC_lap> closing panel

scribe: IP, assets? ...
... delegation model of rules ...
... generic rule set to be specialized as you come across exceptions ...
... define concept that can be overridden by more speclized behaviour ...
... important to how customers use technologies ...
... value to semweb from author's perspective ...
... rule authoring through two mechanisms -- top-down, define ontology as you go ...
... would be nice to provide mechanism within tools to use "standard" or existing terms ...
... obviate need to publish data dictionary ...
... process space, modeling environments ...
... see that in the busniess rules space as well ...
... modeling environments that create business rules ...

JFA: ILOG took load of organizing workshop, previous life in W3C ...
... starting a new initiative -- need to gather the community -- ...
... was obvious that in this particular case, this would be a very diverse participation ...
... think not been disappointed by the diversity during these two days ...
... challenge is here ...
... will try mostly with ILOG hat to give three points where we should go next ....
... what should we try to achieve? ...
... develop rule language that allows the use of rules in multi-application, multi-platform situations ...
... need that on the marketplace ...
... so far, offering more on single app, single pltfm ...
... need to deploy change of gears ...
... that's why we need rule std ...
... 2nd question challenging -- scope of language? ...
... from market standpoint start small ...
... in order to guarantee applicability so it fits 95% ...
... performance and deployment criteria ...
... 3rd questions, what are the external constraints to the development? ...
... compliance to existing efforts ...
... efforts in W3C -- Bijan asked why at W3C? ...
... because Web is universe of network information ...
... multi-app, multi-pltfm has no sense if it doesn't go to the web ...
... other initiatives -- pay attention to e.g., OMG; ...
... pay attention to not partitioning the world ...
... by having others do something almost the same ...

TimBL, director: Web architecture points ...

scribe: initially, simple enough to be widely deployed ...
... unbelievable amount of time to get group together and get anything done at all ...
... 80-20 rule ...
... simple doesn't mean small, simplicity more important ...
... not put in little features which seem important for early adopters ...
... RDF about-each prefix ...
... there was a particular early application ...
... avoid doing things like that ...
... good understanding of relationship between different logic languages ...
... get positive Horn out there, gather experience adopting to it ...
... gives people something to work with ...
... develop bearing in mind that it's part of a much larger thing ...
... plenty of people will make sure it's subset of FOL ...
... take URIs seriously ...
... what does it mean to webize something? ...
... symbols become global symbols ..
... that's not just syntax ...
... it's not just rules ...
... when attributes become URIs, other people can put attributes on your data ...
... make language for set of rules that are designed to be web of rules ...
... two people write rules, both die, others can still use their rules without changing them ...
... context -- know what context was meant ...
... maybe some work to be done ...
... how to move from system that have concept of context to those that don't ...
... partial understanding -- web-like ...
... if rule file ends up among lots of rule files ...
... will be part of system which are useful ...
... but which use terms you don't understand ...
... RDF designed so graphs entail subgraphs ....
... you're deducing things you didn't know about? don't panic ...
... system should be designed so it can process information it doesn't understand, be a carrier for such information ...

Don Chapin (Business Semantics Ltd): Business rule team ...

scribe: to a wider extent, biz rules cmty ...
... two key features to see in work that's done ...
... third one abt framework ...
... first, pick up context ...
... notion of community is core notion for setting context and scope ...
... people typically know whether they're in the SCUBA diving community ...
... important notion ...
... second, ability to explicitly state in ontology or rule definition ...
... if you have set of given expressions that mean the same thing, must be able to state that ...
... equally valuable at level of facts and rules ...
... pick up on last slide from yesterday's presentation ...
... really important that framework in this space -- in two dimensions ...
... one dimension, rules in some form of vocabulary ...
... second dimension ??

Kurt Godden, GM: Hope can remember what he wrote...

scribe: need interlingua ...
... vendor- and tool independent ...
... Pat Hayes advocating sth like that? ...
... if you have own favored environment, like to transfer into or out of that interlingua ...
... second, how do you handle rule exceptions? ...
... lambda operators as way of doing this ...
... without evaluating every time ...
... third, like about OWL the three diff. layers ...
... 2 layers at least for rules ...
... simplified form for domain engineers ...
... without requiring deep understanding ...
... for more expressive, complicated rules, rule-full, that would require knowledge engineer ...

Harold Boley, RuleML: Also about WSMO, SWSL ...

scribe: kernel, Datalog + NAF ...

<DanC_lap> (this interlingua notion... I think the web accomodates that to an extent, but it doesn't really support/encourage it. it's more of a competitive marketplace of syntaxes)

[[ goes through points on panel web page ]]

scribe: beautiful, colorful layer cake ...
... integrate rules with ontology layer ...
... DLP ...

<DanC_lap> (DLP is a great case for use-case/story-telling... I wonder if it's so small that it doesn't actually address an interesting set of use cases)

scribe: integration up to OWL-DL, OWL-full probably out of scope ...
... Semantic Web Services ...
... like to look into policy annotation, web service discovery, time and event process modeling ...
... extension layer for ECA rules ...

Paul Vincent (Fair Isaac): Pragmatism...

scribe: live in world where there are many different types of data ...
... any meaningful type of rule std -- don't want propagation of rule stds for interchange ...
... acknowledge multiple data rules ...
... company goes as far as generating COBOL code ...
... as far from 21st century as you can get ...
... some people want to write rules and run them in COBOL ...
... horrifyingly large community which does want to do that ...
... don't keep these people in their little black box, but embrace them ...
... issue for semantic web community ...
... rule standard that is compatible with the old world ...
... and can be used / transformed to work with the new world ...
... someone using rules in COBOL world -- give them upward path towards future.
... 2. role ...
... rules for defining structure vs behavior ...
... semantic web virgin, but there were lots of rules described that had to do with structure ...
... defining how things look rather than with what things do ...
... behavior ...
... enforcement ...
... business community ...
... careful on which to chose / respond to ...
... what's the use case for rule interop? ...
... number of different worlds ...
... pushing info to subsystems ...
... pulling it as part of larger semantic community ...
... various combinations ...
... that concept to bear in mind as well, in view of use cases ...

Ed Barkmeyer (NIST): there's a number of things he wants to agree ...

scribe: intent was to primarily identify program of work for would-be WG ...
... 1st, define abstract syntax ...
... that follows idea that there may be multiple surfc langs ...
... abstr synt as basis for exchange ...
... don't care whether exchange syntax f-based ...
... whatever define as abstract syntax, need >= 1 semantic models ...
... there may be sets that don't overlap, or they overlap and a symbol has different interpretation ...
... possible that PatH and Benjamin produce unifying model of all rules ...
... important to have well-defined semantics ...
... if you want to render it through RDF semantics, difficult for anyone to understand what the hell it's supposed to mean ...
... standardize what rule engines do ...
... many rule engines based on Horn Datalog + NAF or some form of NAF ...
... n-ary surface language issues? ...
... those things are not compatible directly with FOL ...
... so need to explain how they work ...
... have to be able in that environment to refer to rule sets which are in OWL ...
... or SWRL/anologies ...
... may need to be able to reference another engine ...
... wg also ought to standardize A surface syntax ...
... good many of users wnat a way in which everybody can write or read it ...
... if you force them to read lots of XML, they'll never going to follow it ...
... need a language to write it down and know what the mapping is ...
... like frames, but that's a minor concern ...

<DanC_lap> (what is "frames"? is it an ascii character syntax? what does it look like?)

Danny: Comments from audience?
... please be brief ...

<timbl_> _______________________________________

BenjaminG: not exactly that someone got sth totally wrong ...
... glaring omission -- everyone tip-toed around Bijan's provocative issue ...

Danny: That's why he's not on the panel.

[[ laughter ]

BenjaminG: Two communities. Some intersection in prod rules area ...
... people coming from biz rule space ...
... simple biz rule stuff ...
... OTOH, KR, semweb community, PROLOG community ...
... big challenge, what's going to be the big thing in common ...
... get people to comment on that ...
... two things -- one, negation ...
... two actions, procedural attachments, ?? ...
... nothing that's been previously been on the RDF/OWL-centric community's agenda ...
... that's at heart of business rule applications ...

Danny: Panel?

Jon Pellant: Nothing that users will see.

<DanC_lap> Jon Pellant

Jon Pellant: Exemplars, intermediate things in memory space ...

scribe: feel strongly that have to have actionable statements ...
... vendors ...
... dealing with this as part of processes that have to have actionable results...

P3: Actionable behavior superset of regular logic.

Ed: Have to have negation scheme -- NAF?
... address carefully ...
... timbl's approach ...
... beaver folks' approach ...

<scribe> ... closed relations, specific sets over which to do NAF ...

UNKNOWN_SPEAKER: maybe more constrainable scheme ...

Danny: f'up comments?

PatH: Agree with Ben's 2nd point -- computational machinery to be integrated with rule machinery ...
... interaction needs to be clear ...
... scared of fact that NAF is being used here as a term ...
... heared it mean several different things ...
... careful! ...
... say "sth is not in DB", "sth is not in set" ...
... that's not itself NAF ...
... perfectly classic ...
... not obvious that changing what's in shopping cart is changing assertion ...

Danny: room's sense on Pat's characterization of NAF?
... people decidedly uncomfortable with that approach? ...

Ian?: Not sufficiently well-characterized ...

scribe: perpetrator of DLP ...
... embarrassed by the layer cake with OWL/Rules intersecting ..
... "we're all happy" -- disingeneous ...
... DLP shows that the bit in the middle is very small and not very useful ..
... fudge wrt datalog ...
... thing in the middle is subset of FOL, horn rules ...
... in papers, weasel words about that ...
... this new picture is proposing two completely different and mainly incompatible languages for representing logics on the web ...
... saying to OWL/RDF people -- hold on, we're going sth completely different ...
... don't invest, incompatibility ...

Make: Emphasis on use cases. Whatever you do, drive by use case.

Jos de Bruijn: Comment on Ijan...

Danny: Get at requirements and scope for potential work. Obviously lot to debate, keep to restraint on that front ...

Jos de Bruijn: As have seen in some use cases which are using OWL ...

scribe: lots of these use cases use rules engines ...
... they are just working with that DLP fragment ...
... for them, not a big difference ...

Ian: Don't see any ...

Anthony: Stuck between Scylla and Charybdis -- large research efforts, large vendors ...
... as small vendor, squeezed ...
... looking for interoperable rule systems ...
... use rule engines alongside each other ...
... mistake to just standardize rule language ...
... it's not what's necessary to make rule language interoperable ...

Danny: proposing rules interchange + rule engine interoperability?

Anthony: absolutely

[ missing Ed's repsonse ]

scribe:

Anthony: stuff re getting rule engines to interoperate is ripe for standardization ...

Abdullah: key point -- rule language must be simple
... comment directed to RuleML guys ...
... tried honestly to understand language ...
... not expert in the theory behind it ...
... tried several times to understand syntax ...
... and idea ...
... that was hard ...
... so many acronyms, so many sublanguages ...
... got lost ...
... feel that anyone who wants to implement language won't be very excited if it's so hard to understand ...

Danny: Take that as general comment about desirable production from W3C.

Green Shirt guy in the back of the room: other important stuff besides NAF...

Gary: Is there a basic subset that would satisfy biz rule work ...
... as vendor, very interested in interop of really basic stuff ...

<scribe> Scribe: josdebruijn

Christian: discrepancy between use cases and candidate technologies
... standard should be use case driven
... seems from use cases that using RDF is more of a burden

TimBL: nobody proposes that rules are written in RDF

Christian: understood that rule language needs to be built on RDF

TimBL: for the SemWeb community, rules language needs to operate on RDF; language does not need to be written in RDF

Bijan; Amen

PatH: every language should have n-ary predicates

<timbl_> Unique Names Assumption.

PatH: NAF ok, but UNA is really important for many players
... has not been mentioned yet

Bijan: about use cases: don't talk about them; they don't resolve theoretical issues. It's hard to come up with decent use cases, but it doesn't mean progress

Daniel: we will have use cases

Gary: there is still a problem in understanding between different communities: two KR communities and the business community

<DanC_lap> (this table sounds an awful lot like good use cases to me0

<DanC_lap> )

Gary: proposal to make a table with features so that people get an overview

<sandro_> (Gary Ng, Network Inference)

<DanC_lap> (this is a good question; I think it merits discussion for approx 4 months)

Harry: does the business community understand what they want the rule language to be? A lot of features are mentioned in the requirements, e.g. uncertainty, are they required right now or can that be in later versions?

Suzette: we can agree to a small thing now

??: we need to consider rules other than business rules; we need to have a look at the different problem sets

MichaelU: an omission: support ontology-ontology mapping to enable information integration

??: start with small core and allow extensions

<timbl_> ??: There is a lot of stuff out there .. not going that route is like stnding with your finger in tye dyke.

AdrianW: omission: we need to extract explanations
... second omission: surface syntax which is googled

Donald: Roogle!

Mark: s/Donald/Daniel

MarkLineman: break problem down to smaller scopes

Daniel: parting thought for each member of the panel

Donald: have a look at SBVR

PaulVincent: has not seen any showstoppers

JonRobert: there is determination to define a commonality

TimBL: a lot of connections have been established and more mutual understanding has been reached
... gap seems to be shrinking; relation with other work is very important

EdB: we need to be sure beforehand what the scope is of a working group. We need to avoid creating competing standards.

KurtG: it would be great to have a W3C rule language
... if rules can make me a better chessplayer, I would be very greatful!

<tlrDC> I think he also asked for a syntax he can memorize...

HaroldB: feature table is very interesting. One particular issue are the built-ins.

JeanFrancois: thanks to BenG for pointing out the difficulty we are facing. We need keep the momentum going.

Daniel: we started with the ambitious goal to bring together different communities. What we do has to be well-defined and has to be simple and compatible with what is out there. We need to take existing standardization efforts into account.

<DanC_lap> ("simple" is... a tension all by itself. simple-to-use often competes with simple-to-specify)

Daniel: we need to evolve to a language which allows us to integrate as an ultimate goal.
... next step is report about the workshop; presentations are available online
... report done early June the latest
... looking for volunteers to develop an activity proposal
... contact Sandro if you're interested
... there is the fellows program; people can spend time at W3C

PatH: is designing rules like engineering carpentry or brick layering

Daniel: thank you all

<tlrDC> Thanks Jos for helping to scribe!

<DanC_lap> W3C fellows program

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.122 (CVS log)
$Date: 2005/05/13 19:35:55 $