RE: Web service and Semantic Web

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

Yes such negotiation is needed.  How complex that is 
should be a function of the mission of the   
service, ontological or otherwise.

Global trust systems such as described for the 
semantic web envisioned may be expensive to create 
and maintain given the political and cultural 
difficulties already encountered by global trading 
partners.  There are complex interactions among the 
companies, the governments and the cultures of the 
globe.  I foresee simpler systems based on local 
network definitions initially until we have established 
the reliability of the local systems.  It is difficult 
to believe that companies that have difficulty establishing 
trust within and among their own divisions can open 
the kimono, so to speak, globally easily.  Still, 
we seem to be setting this as a task for ourselves, 
so putting my trepidation aside, my intuition is:

1.  Web global trust requires operational solutions.

2.  Operational solutions require standard public 
means for establishing credibility.  Ad hoc procedures 
are insufficient for some services.

3.  Ontological services have common features that 
can be used to establish operational credibility.  The 
article I cited in an earlier post on this subject  
described some of these features which experienced 
investigators critiqued as potential definitional 
deficiencies (e.g., hidden cycles).  Creating a global 
ontological service will require some form of vetting. 
For this, a testing authority may be required.

4.  The use of or mission of the service determines 
the level of testing required to ensure quality.  We 
have to look at complexity problems of the service 
in terms of types of services and duration of service.  
Consider the use of the service descriptions and 
schema at the end of the namespace URL/URN.  One might 
say a simple human readable description such as those 
that Tim Bray et al are describing are useable at 
the same level of business where one enters a restaurant 
and asks for a menu.  For a complex transaction, more 
stringent quality requirements will be needed.  In 
either case, the observable behaviors are described 
and tested.   

Not being familiar with the work of ebXML, I cannot 
comment.  I think the "simple bridges to allies" 
concept scales and therefore, I look for tools built 
over simple descriptions of services that can then 
engage scripted service transactions.  In other words, 
we have to look at levels of engagement as a protocol.

The work at Microsoft on .Net, XLang, etc. fits the kind of framework 
for such services as were described in our work at 
General Electric Aircraft Engines in the early late 
eighties and early 90s.
 
Our conclusions then were that markup based systems 
with public type definitions (XML Schemas were not 
available then, we were using SGML), and product/process 
models that nested and enabled the companies to use 
public interfaces were required.  This seems to me 
to be exactly where MS.net is leading.  We need more 
experience with the tools but a critical look at how 
the tools enable one to establish the trusted service 
will be the next logical step.

As I said earlier, go slow and deliberately into 
this.  We need practice.  A head long rush such as 
the HTML experience is ill-advised.

Len Bullard
clbullar@ingr.com
http://www.mp3.com/LenBullard

Ekam sat.h, Vipraah bahudhaa vadanti.
Daamyata. Datta. Dayadhvam.h


-----Original Message-----
From: URAMOTO Naohiko [mailto:uramoto@trl.ibm.co.jp]

Len, thank you very much for your comment.

> That is the credibility issue.  It cannot be decided 
> apriori.  You do as you would in a face to face negotiation 
> where no trusted provider (eg, a keiretsu member) already 
> exists (in western terms, a sole source providcer) or for 
> which no prior record of authority exists that attests to t
> the credibility of the provider (eg, a business reference):

I see. My question is does Semantic Web require a mechanism
of trust establishment as well as digital signature framework.

I think digital signature (and certificates from CAs) is not
enough for trust establishment for assertions, and we need 
extra mechanism to qualify assertions scattered on the
Web. If not, we might make trust network in local communities,
but it is very hard to establish Web-scale trust network (which
is a goal of SW), since it is (still) expensive to introduce a trust 
establishment system with PKI that can cover whole the Web. 

> 1.  Discover an entity that claims to provide a service 
> in a claim language you recognize.  This may be in response 
> to a query that serves the same function as a Request for 
..

The steps describe how to establish business trading
between business entities that don't know each other. 
For example, tpaML (Trading Partner Agreement ML), which 
is discussing in the ebXML community, aims to describe terms 
and condition for negotiation (I heard tpaML has been renamed, 
but I forgot the new  name). Do you think such negotiation is 
needed for SW?