Workshop on Frameworks for Semantics in Web Services (Day 1)

9 Jun 2005


Steve Battle & David Martin


Welcome and Logistics

Axel Polleres: we will give out coupons for lunch

scribe: if you have not registered online, please come and see us at the break

Introduction of the workshop (Carine Bournez, W3C)

Carine's slides: http://www.w3.org/2005/Talks/0609-fsws-cb/

Slide 2: Scope of the Workshop

Digital Enterprise Research Institute (DERI), Innsbruck, Austria

Slide 3: Workshop Goal: http://www.w3.org/2005/Talks/0609-fsws-cb/#[3]

Slide 4: Background Material from W3C: http://www.w3.org/2005/Talks/0609-fsws-cb/#[4]

Slide 5: Criteria for Success: http://www.w3.org/2005/Talks/0609-fsws-cb/#[5]


Carine goes over the program

Carine: the last session is for discussions and gathering opinions

Introduction session: domain boundaries, definitions, challenges...

Steve Battle: goal is to discuss major issues in this area

Steve: structure of each session: papers presented for 20 minutes, with then 10 minutes of discussion

"The Semantic Web as Types, Web Services as Functions: Ten Points on a Semantic Framework for Web Services" - H. Halpin, University of Edinburgh

Harry Halpin: slide: Point 1: Build Upon a Common Formal Foundation

Harry: use pi-calculus, concentrate on inputs and outputs
... slide: Point 1: The Wild World of the Pi Calculus
... slide: Point 2: Hide Everything From the User Except Functions
... slide: Point 2: A bit of lambda calculus
... slide: Point 3: Ontologies as Types
... schema for data types for things like integer, the Semantic Web to describe things like age
... slide: Point 4: Binding Ontologies to XML
... WSDL only takes xsd types
... though you can use GRDDL, or XML Schema annotations
... slide: Point 5: Make XML and SOAP really self describing
... slide: Point 6: Embed Web Services in the Infoset
... example given of LambdaXML
... slide: Point 7: Types We Can Trust
... you can put a proof of your security in your type
... slide: Point 9: Give Web Services a Programming Language
... plea for a logical-functional programming language like Curry
... slide: Point 10: Simplification via Abstraction
... let's use some simple and well-studied formalisms as basis for Web services
... Web services will transform the Web from a universal information space to a universal computation space

Sheila McIlraith - University of Toronto: I'm afraid that your presentation doesn't take into account some of the work that happened in the SWS area
...OWL-S does address some of your issues with discovery and composition
... I believe that we need something richer than pi-calculus for typing
... you need to know more than just the type, but also consequences of actions
... this is where this desire to build richer ontologies comes from
... I think that OWL-S has very positive features going in the direction of your documentation

Harry: OWL-S provides a very rich ontology
... however, as a user, I prefer not to know about those details
... I believe that most of the details should be hidden somewhere
... I recognize that you need to take into account side effects (both client and server)

Sheila: we should talk about input, output and side effects more in detail

Benjamin Grosof - MIT, SWSI, RuleML: one of your points was about trust
... I disagree that trust should be put in types
... I believe that types should be used for trust management
... if you look at the way trust management is dealt with in the Web services area, no solution uses types
... even the rich type system that OWL-S provides doesn't fit the bill for that
... I think that using a rich type system is a very good starting point, but we have to realize the limitations to that approach

Harry: I agree that it doesn't solve world hunger, but it's a very good starting point

Benjamin: I don't think that it's a proof in the type, but it should be a proof about the type

<TerryPayne> There is the argument of provenance here - both provenance of data/assertions, as well as provenance of services (and the corresponding providers)

"Two main challenges in service description: Web service tunnel vision and Semantic myopia" - J. O'Sullivan - University of Queensland

Justin O'Sullivan: slide: Outline

Justin: slide: Who am I?
... running http://www.service-description.com/
... slide: Existing Catalogue
... my concerns are about ontologies for Web services
... description of non-functional aspects of Web services
... slide: What I'd like to ask?
... slide: Concern #1
... Conventional services are being ignored for a purely Web services view of services description
... slide: Concern #2
... Semantic richness of non-functional properties of services is not being exploined
... OWL-S is based on Dublin Core and lacking some real depth
... slide: Other Challenges
... we want to allow people to express things like repeating time intervals, but we don't want to make it too hard
... slide: Sources and References
... I have 2 concerns: that Web services are seen as informational services, and that non-functional properties of services are being largely ignored at this point

Katia Sycara - CMU, OWL-S, SWSI: non-functional properties in OWL-S and WSMO is an initial set that can be expanded
... also, we have developed an IDE supporting OWL-S
... this IDE can deal with extensions, and again, you can adapt it so that it fits your need

Justin O'Sullivan: I wasn't aware of this extensibility of OWL-S

Katia: different industries will have different needs, and they can come up with their own extension

Justin: I personally think that we need to get a broader set of basic non-functional properties before we get into extensions

Benjamin Grosof: I share your concerns. you're essentially talking about electronic contracts
... I believe that this touches the policy area
... but it's a very important topic

Rick Hull: we should try to work on standardizing this area

Rick Hull - Bell Labs: it would be nice if the SWS crowd was taking into account traditional services
... an important aspect is "what are the side effects to the world"?
... one of the components to take into account is that those side effects may take time (e.g. a reservation may take 2 weeks to be done)

Alois Reitbauer - Profactor Research: you are talking more about business services here versus computation-oriented services

"Discovery and Contracting of Semantic Web Services" - R. Lara (TIF)

Rubén Lara: in this presentation, I'll focus on the functional aspects of Web services

Rubén: slide 2: Problem statement
... given a requester goal G, (semi)automatically find a service S fulfilling G
... slide 3: Services ws. Web services
... service: provision of value in some domain [Preist, 2004]
... Web service: computational entity accessible over the Internet (using Web services standards and protocols) [Preist, 2004]
... slide 4: Web services descriptions
... slide 5: Conceptual model -- goal discovery
... goals must be semantically described
... slide 6: conceptual model -- WS discovery
... services as set of state transitions
... slide 7: conceptual model -- WS discovery (2)
... determine whether, given an initial state, the final state reached fulfills the goal
... slide 8: conceptual model -- WS discovery (3)
... re-introduce the state transition view
... slide 9: Conceptual model -- service contracting
... interact with the provider to have a complete guarantee
... slide 10: Conclusions
... slide 10: Acknowledgements

Katia Sycara: OWL-S takes this view that the profile describes the service at a high level
... I'm afraid that exposing all the flights doesn't scale
... in terms of input/output, in OWL-S we do partial matching
... there actually exists a methodology from the more general to the more detailed description

Rubén: this is true that both OWL-S and WSMO can have both a general and detailed view

Katia: the service provider provides an abstract description, until you start interacting and you want more details

Rubén: however, this doesn't work, as I may consider the details as very important
... we claim that it's not clear where the line between general and detailed description is

Carine Bournez - W3C: what you call discovery may only be a part of the discovery; there's not only matching, but indexing, etc.; what we have to concentrate is the lower part and how to make indexes of Web services

Amit Sheth - U. of Georgia: usually, it's impossible to describe real world objects completely
... where would you go to find the information to do the matching?
... the challenge is how to deal with incomplete information with a minimalistic framework

Benjamin Grosof: take booking flights as an example: I want reliable and cheap
... are those functional or non-functional? they're both
... OWL-S is a framework that has been designed to deal with that kind of aspects

BREAK until 11am

Discussion / Break-out: Problem Domain. Connections to other domains and limits.

Steve Battle: what scope should we be interested in?
... can we agree about a common terminology?
... challenges: what can a WG can realistically achieve?
... I'll be editing a list live which is projected on the screen
... there are many different outcomes to this workshop: 1 long-term WG, several short-term WGs
... is it the right time to do any kind of activity?

Peter Brown: importance of terminology; the WS Architecture WG tried for WS, and realized that the problem was much greater than they originally thought

<bijan> Two members of the WS-Arch working group say: It failed

<bijan> (jeffm and myself :))

Peter: same issue at OASIS
... the solution must be deployable to things beyond the W3C stack

Yasmine Charif - LIP6: there was discussion about naming functionalities
... we can suppose that there is a domain ontology, as it seems unconceivable that one applies to all
... and we should integrate all of those domain ontologies, which seems to be very hard

Amit Sheth: I believe that this integration is not practical, and I don't think that this work should not be standardized on at this point

Rick Hull: we need a simple presentation for developers on the street; and another being a formal foundation

Harry: we can agree on natural language and XML foundation to express properties

Peter Brown: I believe that your end user is a programmer, which is different from a real end user
... so I believe that we should strive for simple presentation

Rick Hull: I believe that it's a different class of users

Sheila: I'd like to add semantic integration, such as input / output matching

Stuart Williams - HP Labs: foaf and Dublin Core have become very pervasive
... they could be used as a starting point for domain ontologies

Alois: I see the formal foundation as a Semantic Web metamodel
... mental ontology translation is very expensive

<skw> My point was about approach.... putting simple potentially unambitious things in place provide things that people can be encouraged to use.

Paul Giangarra - IBM: the whole concept of composing services is going to need semantic capabilities, where the service consumer is a process

<bijan> Yeah, underspecified doesn't *quite* mean informal

Paul: how do you reserve a service? a lot of this is going to go back to business; we're going to need to meter and monitor

<JacekK> I think the problem is that Web services need to be formal, and the stuff WS seems to need cannot easily start small and informal

<bijan> Well, the real issues is what do you need to know about FOAF and DC in order to use them effectively?

<bijan> For example, take the Dublin Core Date Modified qualified element

<bijan> With an XML schema datetype type, that seems sufficient

<danbri> yes, i think that's the right question to ask, bijan

<bijan> For many, if not most, tasks

<bijan> Because all I might want to know is when the last update was

<carine> bijan: that was Harry's point I guess

<bijan> And I'm perfectly wiling to do a xsd datetime comparison

<bijan> to determine this.

Rama Akkiraju - IBM Research: from industry experience, there are two level: interface (using XML and XML Schema, that we need to support), and domain (mostly represented in UML)
... there's a lot of tooling aspects here too

<bijan> However, if the semantics are small and simple with major use being done by simple comparisons...what exactly are we doing? Isn't xml in WSDL sufficent for this?

Jorge Cardoso - Univ. Madeira: I don't see a lot of Web services on the summary we're working on
... instead of talking about Semantic Web Services, we should be talking about Semantic Resources
... maybe we should talk about adding semantics to pages too
... also, I don't see domain ontologies converging

Rick Hull: I don't think it's appropriate to include resources in the scope
... also, I think it's important to integrate the service ontologies with the domain ontologies
... another point: at times, you may want to express incomplete descriptions, for publishing and discovery purposes

<TerryP> Complete vs Incomplete - does that raise the need for negotiaion/arguamentation either directly with the agent provider, or as part of the provisioning stage with middlewear?

John Domingue: ontologies are not the problem, they're part of the solution
... for all the problems of integration, I see ontologies as a tool

???: the more expressive power you've got, the more complexity you have

<Alois_R> to terryP when do you know your descrption is complete in sense of all use cases anyway ;-) it is not a question but rather a fact

Paul Giangarra - IBM: two different topics: service discovery vs. service location discovery
... UDDI doesn't fit the bill as one is at design time and the other one is at run time, and this is why it's not widely used today
... also, SLA and QoS must be taken into account

Avi Bernstein - Univ. Zurich: want to reemphasize the importance of simplicity for the end user
... the end user doesn't understand most of the complexity that we've been talking about here

Benjamin Grosof: monitoring is very important
... it needs to be highly automated, because there's typically a huge flow of events
... also, we need to be able to manage legacy resources

Harry: we have 4 types of audiences: programmers, programs, users (who aren't programmers), organizations
... how should we do simple convergence?

Amit Sheth: wearing my Semagix hat, I'm hearing a lot of research-driven agenda
...I think that I'm representing the W3C perspective here, AFAIUI
... for conceptual model, we have to recognize things like OWL
... regarding ontologies: there's many ways we can cut the pie
... policies and agreements are kind of things that we could be doing

[ trying to summarize a very long point ]

<carine> Amit: domain ontologies seems out of scope

<carine> ... process ontologies possibly in

Pierluigi Plebani: UDDI requires few steps to be used: publish the WSDL description
... careful about designing a complex framework - functional description, non-functional description, ...,

Jacek Kopecky - DERI: SOAP 1.2 attempts to be Web friendly, WSDL 2.0 is working in this direction too
... however, people still using simple services without those
... this workshop is about semantics and Web services
... this can be seen different ways, and one of them is just bringing semantics to Web services

Mikhail Roshchin: [ out of scope not scribed technical question ]

<TerryP> More thoughts - do we describe and wrap conventional services that communicate in XML, or do we have services that share URI/literal/rdf fragments

<TerryP> very different beasts - requiring different infrastructure support

<JacekK> this workshop is about semantics for web services, but apparently also about connecting Web Services and Semantic Web

<JacekK> That can go two ways - Web Services with Semantics, or Services on the Semantic Web

<JacekK> I think the latter, something like lightweight sw services, should not be forgotten

<carine> it is "semantics in web services"

<JacekK> even though some may view it as description of resources

Giuseppe DeGiacomo: I believe that this group should be more concern about semantics integration than @@@

<JacekK> and if we can reuse the web-friendly parts of WSDL, that would just super

<carine> I don't think that adding a brick on top of WSDL should make usage of SOAP mandatory anyway....

<TerryP> Right - but a web service can exchange any XML data (defined in XML Schema)... and RDF is represented in XML. Hence a WS could talk RDF. Different from a non-SW constrained service using a full range of XML-Schema

Peter Brown: in one word: governance; we need not only to have an ontology, but to be able to show their validity, etc.

<TerryP> SOAP shouldn't be mandatory... I'm still waiting for someone to produce a KQML / FIPA ACL binding for WSDL :)

Yasmine: we're waiting in the next generation UDDI which has semantics integration

Benjamin Grosof: mediation is also important

<skw> Sheila uttered the word "Verification" to add to Steve's list.

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