Re: architecture question

Hi Steve

I suppose that it's possible to describe these services as just different
services, but that still won't capture the service level characteristics.
Won't using differing ontologies  mean that every representation of a real
world entity would be different as it would have different qualities of
currency (how out of date the representation is), accuracy (difference
between representation value and real world value), precision (how much
information about the real world entity can be captured), consistency (if I
navigate to the same entity through different routes how similar are the
representations), performance (how long do I wait for a response before I
can assume that the service won't deliver an answer), availability (is
information about this entity always there), etc?

It doesn't feel right that different representations of an entity should be
entirely distinct. Is that the intention, and mapping between services and
consumers is an exercise for the implementer? doesn't this make automated
discovery pointless.

If these (and other) dimensions are not described, what should the user of a
service assume about  them? And who picks up the bill when the assumptions
are wrong (maybe out of scope, but at the front of most business managers
I'm sure).

These are not application architecture issues (where the environment is
pretty homogeneous), but they do take a lot of energy when defining what I
would call technical architectures (eg POSIX.23 and subsequent work at large
systems integrators), where the focus is binding together (often very old)
islands of IT into a coherent whole for an enterprise.

tc
----- Original Message -----
From: "Vinoski, Stephen" <VINOSKI@iona.com>
To: "Tim Coote" <tim@coote.org>
Cc: <www-ws-arch@w3.org>
Sent: Wednesday, March 20, 2002 8:37 PM
Subject: RE: architecture question


Hi Tim,

The problems in your example are problems of semantics and ontology, not
strictly problems of architecture. You might want to look at the
recently-published Web Ontology Requirements working draft [1] to see if
it's addressing your concerns.

[1] http://www.w3.org/TR/2002/WD-webont-req-20020307/

--steve

> -----Original Message-----
> From: Tim Coote [mailto:tim@coote.org]
> Sent: Wednesday, March 20, 2002 1:30 PM
> To: David Booth
> Cc: www-ws-arch@w3.org
> Subject: Re: architecture question
>
>
> Hi David
> A typical example here would be the current account systems of most UK
> clearing banks whose 'current balance' is actually the amount
> in the account
> after yesterday's transactions have been applied and don't know about
> anything that has already happened today, or what
> transactions were queued
> up to be applied today.  These systems may also assume that
> they will only
> be interrogated between 8am and 8pm, the currency is assumed,
> not explicit,
> the applications have no real service levels for response
> times, and they
> can show confidential information that is expected to be filtered and
> interpreted by an employee.
>
> I think that my fear is that there'll be a different set of
> assumptions for
> each company's web services and that it won't be possible to
> capture these
> within the discovery protocols so that discovery will not be
> automatic and
> the interoperability challenges will be of a similar size to those of
> implementing EDI, but with an even larger legal bickering
> process over who's
> responsible for errors.
>
> tc
> ----- Original Message -----
> From: "David Booth" <dbooth@w3.org>
> To: "Tim Coote" <tim@coote.org>; <www-ws-arch@w3.org>
> Sent: Wednesday, March 20, 2002 4:21 PM
> Subject: Re: architecture question
>
>
> > Tim,
> >
> > At 12:50 PM 3/20/2002 +0000, Tim Coote wrote:
> > >The biggest challenges tend to come from very old
> assumptions (~20 years
> > >old) in line of business applications.
> >
> > Can you elucidate?   What are some of these assumptions?
> >
> >
> > David Booth
> > W3C Fellow / Hewlett-Packard
> > Telephone: +1.617.253.1273
> >
> >
>
>

Received on Wednesday, 20 March 2002 17:16:54 UTC