RE: Top cloud in triangle/rectangle diagram

At 01:47 PM 10/7/2002 -0400, jones@research.att.com wrote:
>There is no pure SOA without some service information being acquired
>by the requestor.

I totally agree.  And the important part is the service information itself 
(e.g., WSDL) -- not the means by which it is acquired, nor the number of 
hypothetical parties through which it is passed on its way from the Service 
Provider to the Service Requester.

>To identify a case where it involves human agency,
>prior business arrangement and SMTP is only to locate a particular
>point in the continuum.  There are other real-world variants, even
>more likely in my experience, in which the service provider explicitly
>communicates URL's to the business partner at which WSDL and English
>documents that further describe the service are to be found.  The
>requestor then gets these documents by using HTTP rather than SMTP as
>in David's example.

Actually, I would take this as an excellent illustration of my point.  The 
concept of "discovering" or "searching for" the service is totally 
unnecessary in these scenarios.

The only reason I can see for including a component in our architecture is 
if we need to describe what it DOES.  For scenarios like you just 
described, or the FredCo/Widgets-R-Us scenario[1] that I described, I see 
no benefit (and some harm!) in us trying to describe what a hypothetical 
third party might be or do.  In fact, we shouldn't even care whether a 
third party exists!  The only thing that matters is the service description 
itself.  Let's not make the BASIC architecture more complex than necessary.

[1] http://lists.w3.org/Archives/Public/www-ws-arch/2002Oct/0072.html

-- 
David Booth
W3C Fellow / Hewlett-Packard
Telephone: +1.617.253.1273

Received on Monday, 7 October 2002 21:35:30 UTC