WSWS Wednesday 11 April morning sessions
Go to the complete list of minutes.
Introduction - Tim Berners-Lee (W3C)
- Q (??): ontologies - been there - but people still trying to model cars
- sounds very ambitious - CYC tried this
- A: 10 years ago hypertext was in the same situation - existing systems
didn't scale - existing KR systems don't scale/central architectures -
there is only one ontology in these systems. Semantic web tech is not like
that - different concepts of car can coexist
- Q (Sankar, Crystaliz): what i'm seeing is repetition of OMG approach -
benchmark I want to use: any activity that gets started should be
implementable - that's level where http started - could do this - can't do
this with XML protocol, now adding even more stuff - cascading effect of
complexity - vision of semantic web is ok - but looking for 5 to 10 years
of development - would like to see small, independent effort that ... all
papers seem to be about replicating OMG effort
- A: concern about complexity has already been expressed - how do you go
back to existing specs to make them simpler
- Q (Lotus): complexity is sometimes necessary ...
- Q (Mitre): what is complexity of XML protocol ?
- A (Sankar, Crystaliz): if I want to implement whatever shape XML
protocol will be will be more complex than http - intermediaries etc. -
have to grok all complexity until i get a simple protocol - same for XML
schema - have to know infoset before I can do a simple date schema - need
to keep things simple to bootstrap Semantic Web - UDDI is complex
- Q (Andrew Layman, Microsoft): people don't want to have pointless
complexity, but need to ask what are the goals, then let's build minimal
amount of machinery to achieve goal - mere simplicity is not the goal,
simplicity in achieving a certain goal - this workshop should respond to
question what are the goals
The W3C: Organization and Technologies - Hugo Haas (W3C) - [slides]
- Q (Chevron): looks like you're operating in a vacuum - other groups are
working in same space (ebXML, IETF, Oasis) - way things end up in
different organizations looks capricious - as a member organization, we
have a strong stake in seeing standards bodies cooperate rather than
compete - two organizations in Oil industry talk to each other, W3C
doesn't seem to talk to other people
- A: we know other people are working on other things - one of the topics
of last session in this workshop is coordination with other groups
- A (Tim): liaison - first, we try to have well-defined boundary (e.g.
with IETF - won't do TCP, DNS in W3C) - also, try to reduce workload by
doing things elsewhere - HTTP was e.g. done in the IETF, but with W3C
resources - Digit. Sig. is done in both W3C and IETF as cooperation - with
Oasis, original understanding was that Oasis does vertical pieces and
coordination between vertical pieces - that's not how things have
developed - still spending lot of effort talking to Oasis - in the end,
we're guided by our members - W3C and Oasis have large overlap - one of
the questions that need to be answered in Activity proposal is why
activity has to be done at W3C - other organizations have to be listed -
other example is WAP - we're not interested in doing work that other
people are doing
- Q (Bob Sutor, IBM): we're in the middle of who does what were - I'm on
the board of Oasis - many other people of Oasis and ebXML are here -
coordination has not been great in the past - what i'd like to see out of
WS is to see better coordination - member companies are driving the agenda
in all organizations involved
- Q (Steve Zilles, Adobe): thanks for survey - providing context is
important - surprised that Xforms is there - uses other technologies that
you didn't mention - was surprised that work in other organizations is
last thing on the program
- A: Xforms not really related - other technologies that are used by
Xforms are what I called "core technologies" - doing other organizations
last thing was an organizational choice, but you do have a point
- Q (Schwab): problems we're talking about are going to cut across
organizations - seems we'll end up with different interface description
languages for different transports - want to move away from that - general
framework where some services are deployed in "web-way", others are
not
- David: if you know technologies that are already done in other
organizations, please bring them up in the discussion
- Glenn Davis: have done something very similar to WSFL - are you
publishing this ?
- A: just finishing documentation - may come out next month or so ... this
is an area that is right for standardization
- Q (Ann, Sun): there is a lot of work in ebXML that overlaps with WSFL
etc. - some of the work in ebXML is based on IBM submissions - why don't
pursue it in ebXML
- A (Bob Sutor): have to take a long time view - ebXML has a lot of input
from EDI community - we need convergence - need to have everyone have
their say at appropriate time - ebXML will strongly influence what will
happen here
- Q (Cisco): what happens if service provider can come and go, e.g. P2P
environment
- A: have thought about this somewhat - still work in progress
- Q (webMethods): web services will be provided by different vendors - how
do you charge people for using services ?
- A: this is where intermediaries come in - have thought about this a lot
- need metering, logging - won't hard-code one metering system ...
intermediary will log request - but is not intrusive
- Q (PDA systems (?)): question about scoping in W3C - talked about a lot
of broad aspects - seems you're suggesting a lot of standards - what are
priorities ? which areas overlap/are already addressed ? what is the
timeline ? could take five years
- A: spent a lot of time on roadmap/timeline on when people will adopt
different technologies - agree with you that there is a lot of work -
don't have an easy answer, but surprised on how fast people are adopting
technologies - think workflow is important - customers want to know what
our timeline is - want to know if we follow standards - i don't think
adoption roadmap is five years - example WSFL: will be adopted very
quickly - next thing is endpoint language - we don't want to outpace
customers though - but a lot of stuff sophisticated customers already ask
for
- Q (Andrew Layman): on W3C scope - look at what is big picture, how can
we modularize it so that we end up with something that is coherent -
splitting up over organizations won't reduce workload - too early in
conference to make conclusions/decisions on scoping in W3C
- Q (Crystaliz): WSFL - DPMI sounds related - how is this different ?
comment: you seem to be talking to advanced customers - thinking of doctor
practices - ideal case for web services - go to Chief Doctor and tell them
what they're supposed to do to use Web Services - that's the scenario I'm
looking for
- A: just talked about this at another conference yesterday -
interpro/protensios - looking at web services - we'll help them - I talk
to large companies, but also talk to startups
- Q (Andrew Layman): what does the service broker do ?
- A: will show in later slide
- Q (Intel, Joel Munter): what about scalability of service broker ?
- A: that is a platform feature - can be done by deploying caching -
replication - broker contacts catalogue, catalogue info can be cached -
that's how it scales
- Q (Sun): how many brokers do you envision ? is there communication
mechanism between brokers ?
- A: yes, brokers can become clients of services
- Q: are brokers and intermediaries the same ?
- A: yes, from the conceptual perspective
- Q (Ken Laskey, SAIC): we looked at brokers in a particular environment -
tracking packages as they went from one shipping unit to another - brokers
themselves can be domain-specific
- A: yes, you can have federation of brokers
- Q (Crystaliz): like specificity of your presentation - but if you
generalize brokers to intermediaries, lot of your assumptions don't seem
to be true ...
- A: certainly, openness and extensibility would allow you to do this -
extensibility in the framework - haven't shown specifics how it can be
done
- Q (Data Channel): proposal for W3C work - how does this work from
process - BPMI, ebXML, ... efforts - BTP for transaction etc.
- A (Tim): tool for coordination in our process is coordination group - no
reason that coordination group only includes W3C groups - can be across
organizations
- Q (???): if you're interested in XMLP, there is also XML Query, Semantic
Web - what is thought on integrating with those things in W3C ?
- A: total integration is goal - want to encompass more of W3C technology
- we would like to increase W3C involvement of Novell
- Q (Crystaliz): how does UDDI relate to DSML ?
- A: UDDI is not necessarily *a* directory, but mechanism to discover
services - UDDI just a registration and a look-up - not a lot of security
involved - what I'm proposing is more robust way to look at what UDDI is -
see convergence/merging - takes into account security and authentication
(as composed to UDDI)
- Q (Crystaliz): LDAP seems to have more semantics than UDDI
- Q (Andrew Layman): examples that you bring up - are they baked ?
- A: this is a product - this works for us now - but would like to see it
move forward to use namespaces, Xpath etc. - usage is what is
important
- Q (??): you said that DirXML has good security support - how do you
authorize proxy users ?
- A: all access to all objects are controlled by authorization elements -
each "driver" has "identity" - rights transferred e.g. via
inheritance
- Q: proxy users who stands in for some other user - how is this solved
?
- A: has another identity - we do that by roles - proxy user can take on
role of the entity it acts for
- Q (Ken Laskey): directory would hold all password info ?
- A: yes
- Q (Crystaliz): still on UDDI - use of directory services as presented is
good - in previous presentations, directory service was pretty vague -
recommendation service could be considered as directory service - can't do
everything in directory service - what is your view on what directory
service should do ?
- A: authentication, authorization, configuration information - you don't
go to directory to do what you want to do, but to find out where to do it
- brain of the web service - see UDDI on top of existing directory -
roles, relationships should be kept in directory
- Q (Tim): can you have one conceptual vocabulary, but competing services
?
- A: vocabulary should have an identity, i.e. somebody that guarantees
vocabulary is meaningful - it's the person that makes the vocabulary that
makes the difference ... if there was organization that does ISBN numbers,
they could endorse their vocabulary - if signed by organization that does
ISBN numbers, it is indeed an ISBN number - otherwise, no way to tell
whether it is actually a true number
- Q (Sun): how do I actually validate that a vocabulary is used in the
right way ?
- A: we use SPKI certificates (simple public key infrastructure - another
way of representing signed certificate, difference to PKI is that things
that describe what service can do are signed in the certificate - can be
shipped along with the identity - when you send somebody your public key,
you send along all certificates) ... lots of other ways to do it, e.g.
round-trip messages
- Q (??): what mechanism to extend query types ? are they fixed ?
- A: query language is trader query like language, drawn from Corba -
haven't changed that - trader defines function call method that calls with
matching constraint - haven't really tested that part of the system - what
I want to describe here are business motivations, what you need in order
to do business with results of web service lookup - idea of trust, idea of
looking up using vocabularies that can be extended dynamically is
important for any directory system that handles web services - extending
queries wasn't the first thing we tested
- A: WCDL does not overlap WSDL, meant to complement
- Q (Crystaliz): in previous pres., there was a lot of talk about service
metadata, but not about how it will be used - what kind of queries would
people do, e.g. - trucking user wants to find where trucks are
(location-based service) as opposed to computer manufacturer that has four
products - how can you put a box around these different services - by not
looking at types of queries, we're leaving things too open
- A: ...
- Q (Ken Laskey): brokers will tailor their services and queries to
particular domain .... you will need use cases to design rest of the
system - can we develop component set that brokers will be built of
- A: probably agree - there's any number of ways of judging, and therefore
querying, best business fits - will be more elaborate than name/value
pairs or RDF
- Q: you made reference to ad-hoc
- A: service selection is ad-hoc - likely to be selecting services at
design time of service - there may be standards for web service interfaces
(e.g. for buying pencils)
- Q: talking about manually crafted services
- Q: really agree with you about what sweet spot is - have real doubts
whether directory situation is going to work - looking at the technical
details, if web process can call web process over internet, it has to
expose more than its interface - one web service could secretly call
another web service, which is maybe not what you want
- A: like to suggest breakout on workflow
Philipp Hoschka