RE: Top cloud in triangle/rectangle diagram

At 01:20 PM 10/7/2002 -0400, Heather Kreger wrote:
>. . . David's scenario is valid and works fine. I don't see it 'breaking' the
>roles in the triangle.

I agree.  The FredCo/Widgets-R-Us scenario[1] doesn't "break" the triangle 
roles at all.  The point is that the triangle defines an extra role that is 
not NECESSARY to this scenario.

>I have always supported that the 'top cloud' (I
>missed our meeting and haven't caught up with my mail to find out if we
>have a new official name yet) role is not specifically required in every
>scenario and that roles may be collapsed.

The spec draft[2] specifically states: "Valid implementations include 
subsets or parts of the stack, but must at least provide the components 
within the basic architecture."  I take this to mean that the BASIC 
architecture defines a minimum set of REQUIRED components.  Optional 
components should be defined in EXTENDED architectures.

I think we should be conservative about what we put into the BASIC 
architecture -- only including components that are clearly needed for the 
vast majority of our expected applications.  Other components should go in 
EXTENDED architectures.  But the whole concept of a "discovery role" is 
ONLY relevant if the Service Requester does not already know who it wishes 
to talk to.  For this reason I think it is a much better candidate for an 
EXTENDED architecture.

[1] http://lists.w3.org/Archives/Public/www-ws-arch/2002Oct/0072.html
[2] http://lists.w3.org/Archives/Public/www-ws-arch/2002Oct/0038.html

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