Workshop on Frameworks for Semantics in Web Services (Day 2)

10 Jun 2005

Steve Battle & David Martin


Session 4: Looking Ahead (David Martin)

[Thanks and Acknowledgements]

List of functionalities displayed (as defined in session before lunch)

Presentation from Ken Laskey, MITRE

- Semantic Interoperability for We Services: The needs and challenges for the Demanding Customer
... MITRE is a defence contractor for the US Gov
... Presentation is from the customer perspective

Slide 1: perspective from typical customer

slide 2: typical problems, typical solutions

... US DoD moved to a single vocabulary approach... but didn't quite work
... as of 2004 data should assume a net-centric approach
... promoted by communities of interest

Slide 3: New solutions, new problems

... SOA introduces different vocabs/semantics etc
... single standard vocab across diverse domains not feasible, so how to share info

Slide 4: Underlying Challenges (1)

... avoid n-squared communication problem
... try to use small set of vocabularities, trade offs

Slide 5: Underlying Challenges (2)

... different evolving vocabularies, need to be timely
... expansion of vocab may be required before full scope of use understood

Slide 6: Underlying Challenges (3)

... imprecise description necessary (e.g. when selling house on 2 neighbourhood boundary - want to advertise in desirable one)
... degree of precision is variable, depending on domain/requirements

Slide 7: Specific Needs (1) - The Many Vocaularies of XML Tags and Metadata

... need both XML metadata tags and values between tags
... should disambiguate use of terms, at least

Slide 7: Specific Needs (2) - Versioning as a Variation of Multiple Vocabularies

... differences between two vocabularies, and two versions of the same vocabulary

Slide 9: Specific Needs (3) - The Many Uses of Constraints

... authoritative data sources (how defined and evaulated)

Slide 10: Overriding Need for Reuse Across Domains

... need standards to accomodate change

Slide 11: Possible Approaches of Interest

... annotations to connect documents to semantics that define the meaning
... upper level ontologies - don't want to redefine every time

Slide 12: Thoughts on moving forward

... for mapping, what information/structures/tools do we need?
... what is the difference between compromise integration ontology and upper ont
... can we come up with a soln, which can be reused by others

David - limit to short # of questions...

Florian: relate semantics to semantics to each other - crucial point
... would be interested in discussion of similarity

A Semantic Web Services Challenge - Stuart Williams, HP

Presentation by Stuart and Steve Battle - view on what the Workshop is all about, and giving advice to W3C

Slide 1: index slide

... many of the points in my slide have emerged in the last couple days

slide 2: Observations on Web Services

... WSDL doesn't say what a provider does!
... multiple different providers provide same service but with different interfaces
... thus multiple re-development of same service with different mechanical intetrfaces, may segregate communities

slide 3: Web Analogy

... e.g. B2C ecommerce example
... various genres of interfaces may lead to different standards themselves - but creative genie is out of the bottle

slide 4: Service Delivery Lifecycle

... work in Bristol (UK) at HP
... service is "exchange of value" (Chris Priest)
... window cleaner analogy (he cleans window, I give money)
... issues include:
... advertising and discovery
... negotiation
... post contractural (delivering service)
... contingencies, failures etc
... web service interactions throughout lifecycle

slide 5: vision!

.. want to see open economy of service providers/consumers
... remove technology friction, reduce cost/time to integrate
... SW brings adaptability
... current WS is static, rigid and at design time
... pair-wise
... scale from pragmatic to idealistic but unrealistic
... explicit operational semantics facilitates planning etc

slide 6: Why Bother (Mike Brodie - at ISWC2003)

... folks spending $500 BN per year on integration costs!!!

slide 7: success factors (for W3C in SWS space)

... need enthusiasm from WS community (pull, not push)
... should identify their requirements, and solve some of them
... may "borrow" some, but not all solutions from SemWeb community
... Maturity - are we too early for standardisation
... need growing consensus on where we are going
... Accessible - available to Joe on the street
... build only when necessary - REUSE

slide 8: A challenge

... to find and "buy" book
... also services "bake-off" - buy a book from "Amazon", then sell on "ebay"
... should try to do this, but then generalise to other vendors/search engines

slide 9: What should the W3C do?

... 1, 2, several working groups in this space???
... what is the focus of them (or even just one if we are lucky)
... how they relate to WSCL, BPEL, CDL etc
... like the OWL-S high level partitioning (profile/process-model/grounding)
... good to see WSDL-S
... some ultility would be to put in place vocabularies to ground descriptions, and encourage people to use them

slide 10: Concerns/Questions

... need academic/industry consensus on directions


Mark Nottingham: what is the right timing?

Dieter: you only mention OWL - what about the other languages - important?

Stuart: yes, don't read too much in it - others are important, OWL as generic term for SemWeb languages

James Clark: what about backward compatibility? how much of what is out there today is on the bargaining table? or do we tear down and rebuild?

Stuart: take a web service - define semantic rich interaction. Don't expect provider to alter it for you!!!
... my ambition at least

Rick Hull: on slide 5 you define a service... is that exchange a short or long duration

Stuart: later

Ben Grosof: SWSI etc extending existing work on contracts/policy etc

Suggestions from the Other Side of the Aisle - Mark Nottingham, BEA

Mark: works in office of CTO of BEA Systems - works on architecture and standards
... wearing hat of technologist :)

slide 1: Web

... me - I know only a few of you. I started with the web itself
... company was noticed when we did Web Apps

slide 2: Semantic Web

... we're interested in this. Have done early work e.g. URI space note

slide 3: Web Services

... have been involed in various WS groups. Worked in XMLP WG, Chaired the Constraints and Capabilities in fall 2004, Chair of WS-Addressing WG, etc

slide 4: Semantic Web Services

slide 5: what are web services

... architecture being defined in products (bottom up)
... example - task force (async task force):
... past efforts like Corba didn't involve everyone, so this is a unique opportunity

slide 6

... various architectural styles, inc SOA, REST, RPC
... web services itself doesn't have an "archtecture" but can be modelled across many

slide 7: Frameworks

... can you build a framework without an architecture?
... well, web didn't when timbl started!

slide 8: Risks and Rewards

... WS are a fast moving target
... semnweb has a lot to offer WS, but neither are fully defined
... complex and resource hungry from communities, so finding alignment may be risky (is it necessary)

slide 9: RSS (a cautionary tale)

... have been requirements to get RDF model into efforts such as WSDL and P3P, but difficult as little by-in from others
... RSS is more illustrative :)
... emerged in 1998, various folks developed it. Others came up with RDF variant, leading to two efforts
... confusion, infighting, blood-sweat etc
... different technologies with same name (yuck)
... and ATOM started at IETF

slide 10: Use Another Name

... "Web Services" is still ill defined
... candidates inc. Semantic SOA, Semantic Services, etc

slide 11: Why Standardise?

... some think that standardisation guarantee success!
... many examples where that doesn't happen
... Standardisation should be BORING - no intellectual challenge
... Standardisation is really hard, and really expensive, so think about why (and whether) you really need it!!!

slide 12: Why am I here?

... you should hear the vendor feedback.

slide 13 : diag from his paper, contrasting WS stack vs SW stack

<carine> http://www.w3.org/2005/04/FSWS/Submissions/32/stacks.png

... interesting to look at parallels between stacks. Doesn't mean they have to be mixed!

slide 14: Suggestions

... Good if Bijan carries on work in WSDL on RDF
... Use RDF model in WSDL
... issues on stateful services (look at REST)
... fix RDF/XML serialisation
... get the stuff deployed and out of the lab!


Rick Hull: your point on Describe the Web - do you mean services on the web - Documents that provide information, or things that affect the world?

MarkN: Rest blurs this distinction

Rick: is there the need for describing semantics, and maybe the SWS should go away for a while?

MarkN: maybe. More on REST, such as manipulating resources...

Jamie Clark (OASIS): are you saying - we have the stack - don't mess with it?

Mark: not exactly those words. But there are diminishing returns...

Ben Grosof: clarify "fix Datatypes" (about RDF types)

Mark: need to specify every time used
... not sure of history, but may have been a compromise
... as to XML schema, are there other ways of defining data models. RDF schema may want to look at this

Mark: XML ID coming along - add ref and graphs become possible
... ordering is an issue

Rick: you are saying don't call "them" Semantic Web Services". Looking at WSDL, no sequencing, etc, so that would be good to add...
... do you think semantics of sequencing stuff should be added to WSDL (e.g. WSDL-S)

Mark: I liked that stuff...

Rick: is sequencing info useful for web objects?

Mark: sure.
... reasonable way of adding semantics to Web Services
... implicit model for web services that appear in implementations today (not pretty)
... defining high level arch is not going to help of clarify...


Round Table Discussion and Summary

Session 4 Questions slide - to be addressed by panel

Peter Brown: Eat your own Dogfood. Concern - if we want to answer the questions, and have to find the standards (just using html), then that doesn't say much about SemWeb

John Domingue: comment on time frame - there was lots of money in US, but winding down. Lots of EU money now, but could be winding down

Mark: W3C concern is more from vendor and industry perspective - sorry

Rama Akkiraju (IBM): if WS community find out the key "pain" points in developing WS and what customers want, and bring them to the research community - would be valuable next step

Amit Sheth: I have several hats - from a vendor perspective, single operations are limited, and workflows powerful. Pain point is composition. Such pain points should be identified, hence our focus on use-cases
Other thing is that if this approach is ideal, current approaches may not get traction... need clear use cases which might be limited in nature
... recent W3C task force discussed challenges of coming up with something in the next 18 months [not sure what for?]

Ken: need to describe pain points in customer's view, not your view of where the pain points could be

Rick Hull: Straw Proposal for roadmap
... 1 linkage between services we invoke and stuff in the world they impact
... WSDL-S, OWL-S atomic process, and WSMO notion of capability as starting points
... 2 address issue of sequential nature of services and that we may want discovery for sequential (not transactional)
... longer runing transactions over time
... i) procedural style specification of the process (e.g. abstract state machine from WSMX)
... ii) relationship between messages
... iii) discovery over the previous two points (starting point could be PSL/FLOWS)

Joel Farrell from IBM: when talk to customers, they don't talk "Web Semantics" but we have problems in Enterprise Integration
... reconcile schemas etc, which make one think about semantics

Julie DeMeester (MITRE): people may not like WS-I but mandated across our customers and Dod
... hence WSDL-S can be adapted - start small, and gets customers on board
... can come back with lessons learned. Need W3C as facilitator
... currently having to teach the value of various specs so far...

Ben Grosof: Longer term - picture emerged from various efforts is that there are more things than there may actually be
... first OWL-S "it was good"
... 2nd SWSI
... 3rd day WSMO
... 4th day was WSDL-S
... but OWL-S had non-functional properties and IOPEs, grounding in WSML, but process model, with DL is possibly crippling
... SWSL more realistic in process modelling
... can take that work to serve the work on translation mediation or even describing a service
... WSDL-S is nice (bottom up)
... WSMO nice as it focusses on mediation
... few basic building blocks with depth, but not necessarily controvertial
... hence, question is - what makes sense for standardisation? Major Building Block? Are we in position now to do this
... but longer term - maybe good
... policy is good

Jacek Kopecky (DERI): Is WS-I so important? What about specs without ambiguity, such as SOAP1.2 etc

[out of scope discussion about WS-I]

Jeff Mischkinsky: before one standardises, one should be beyond the research point. Is there a platform out there?

Mark: that harks back to Bijan's point - where are the start-ups?
... small pieces that work together to make things

Axel P: have heard good proposals, so short term goals...
... focus on topics such as discovery, there is the incubator thing, etc

Ken: as member of W3C advisory board - that would be a good thing

Bijan: if you have problems, we would gladly mutilate ourselves to work on real problems
... if there were problems with (sanitised) datasets - then please provide them. Bijan would happily set up a repository for this

Rick Hull: cart before the horse? There is considerable research on profiles (OWL-S) etc so phase 1 has a body of research
... second point - schema wkshp may focus on data integration problems, but this community should not be diverted down that path (we're process focussed)

Hugo: not sure if people know W3C process - but not get into this.
... but would people be interested in getting into the standards space? Can ask questions and figure out how to proceed

Ken: you might want to outline the process...

Hugo: incubator process (lightweight) for small # of people to create seed for community and define area
... if cool stuff happens, then set up working group
... i.e. bootstrapping

Mark: need more than cool stuff happening. Show actual savings - e.g. reduction in time or cost to get solutions with Semantic Web Services

Rama: I concur - should perhaps start with large datasets in DB world

Mark: think of orders of magnitute that would be. To move to the SW would be significant investment for a vendor

Rama: just start the experimentation to make business case for ourselves first

Mark: yes

Ken: Vendors not actually showing good case for WS as it is

Jeff Mischkinsky: there is space for stuff for the Schema workshop

David Martin: creating standards from the bottom up may not be the only approach. OWL was valuable, took a differnt approach - whaddaya think?

Ben G: what is the value of an Ont if there are no techniques for using them?

Ken: techniques are useless unless there is a need!!! (being provocative)

Rick Hull (speaking for Lucent): I have seen at least one example (Fujitsu Labs) which is using OWL-S profiles - TASK COMPUTING and devices

Rick H: shut me up :) but... relational db query lang - what happened was that formal foundation emerged, and after that SQL etc happened

Alois R: you have raised use cases...
... but customers want to be asured that the technology will be adopted widely - where are the refernce docs
... I don't care about what the standard is called, but please provide an agreed model so we can proceed

Antonio?: we need a small standard to know we are solving a big problem; but for a small project I need something to make the EU happy!

Dieter: and what do you say to the commission?

Antonio?: I say WSMO :)

David: are things resonating, or junk ideas???

Mark: it sounds like there are pressures for standards, but 2 concerns: W3C is heavily involved in WS, we need to be distinguished from it; 2) W3C is web aligned, we need to mind possible questions about such an alignment from the TAG

Joel: ... something like taskforce would not loose momentum, yet avoid claim of "cart-before-the-horse".
... an outcome of that may lead to decision on standardisation

Carine: there is an IG!

Katia: I've spoken to industry about this stuff and case studies. If we are serious, I propose we should put some deadlines down on the roadmap
... don't forget that working groups are a time commitment and take time to results

Stuart: HP has presented a challenge case - ideally industry based, maybe at a forthcoming conference
... also there is some risk of defering use cases - they should still appear

Ken: we could create a short term group to define a challenge problem, solve it, and present to public in the next year

Katia: you want us to vote? Is this a formal proposal?

Ken: I propose this - that a group be formed to come up with at least 1 problem, and within 12 months come up with at least 1 solution

Katia: friendly ammendment - should demonstrate advantages of Semantic Web Services

Ken: the problem should also include points the WS community can't already solve (ideally)

Ben: I don't like the challenge problem mode. Takes rich field and compresses it. Good for narrow focus, but we are wide.

Mark: problem is that community is perceived as too broad. Would be good to demonstrate solid example to convince industry!

???: we should go deep - problem described formally, to show how services described, composed etc

Dieter: we're preparing a challenge problem and will publish one in Feb/March 2006

Jeff: there is an IG, right?

Hugo: yes

Ben G: Should we really have a community for the challenge problem? Fabien, Dieter etc are doing this

Katia: that's a different thing from a Global challenge problem
... Dieter is defining such a problem from DERI, but surely the industry people should define a problem from a vendor focus

Paul: you said "one or more problems" - I say scenarios
... 2nd we want scenario that not just demonstrates value of SWS, but that we have an implementation *that works*
... maybe we should pick a date and get people to propose demos

Hugo: we have 5 mins left, so would like to close the session.

Carine: we need real use cases, ideally from someone who is *not* involved in current efforts, should be from industry

Terry and Bijan agree.

Massimo P: I want to see out of this meeting, a set of people and deadlines. Am concerned that 6 months from now we are still fuzzy on the thing

<bijan> Use the sws-ig list!

David: we need a mailing list for this discussion

<sbattle> use the IG!!

Dieter (last word): the use-case was not just a DERI case. Should be one that all approaches use
... don't necessarily need the W3C involved in this

Carine: [describing how the XML Binary Characterization WG has worked]
... the XML Binary Characterization has put together lots of use cases to start the study around its 'challenge'

Hugo: it should be discussed in the IG
... have some questions for you:
... Who thinks the area is ready for standardisation? [13 hands counted]
... Who would be participating in a W3C activity for this (of those that voted yes)? [9 hands counted]
... for the people who didn't want to do standardization, who thinks W3C should do some pre-standardization work - incubator? [about half the room voted]
... Who would be interested in investigation work (Ken's proposal about challenge) ? [about 75% hands raised]
... (W3C members)? [10]
... Who thinks W3C should *not* do anything for the next year? [3 hands]
... We've heard 3 directions - OWL-S, WSMO, etc... so for people who want to do something - who wants to see stuff on the whole framework in the short term? [no hand raised, no interest in the complete space]
... Who thinks that W3C should start work on expressing semantics in WSDL as a first step? [20 hands voted]

Stuart: Remember this is a serious investment of time
... what about a phased approach (e.g. Rick's approach), i.e. W3C should start work on expressing semantics in WSDL as a first step? [20 hands voted]

Meeting Adjourned.

Minutes formatted by David Booth's scribe.perl version 1.126 (CVS log)
$Date: 2005/06/14 16:21:02 $