On Wed, 14 Jul 2004, Geetha Manjunath wrote:
Alex Rousskov wrote:
<stripped>... unless we want to expand P scope to affect things like preview
size dynamically, while the data is being retrieved from the network.
We will be able to control service _execution_ then.
I am sure this is right. If the service is say, "virus scanning", by
changing the preview size of the stream being sent, you cannot
probably tell the scanner to change its execution path . On the
other hand, the OPES service may be able to dictate the preview size
required for it to function.. Basically, it will boil down to
negotiation over the callout protocol .
I had an even more dynamic case in mind, where some P code will tell
the invoked service or OPES processor (after the service invocation),
that the service can have 10 more bytes, then 30 more bytes, etc.
(depending on some conditions expressed in P). That would be execution
control. I do not think we want to go that far for now.
Configuration before invocation is clearly in scope.
s.a = 1;
invoke s;
Negotiation over OCP might be in scope, but only if P hides the
details:
if (s.supportsFoo()) then {
s.a = 1;
} else {
s.b = "bar";
}
invoke s;
P interpreter should be free to implement "supportsFoo()" call via OCP
negotiation with the service if static information about Foo support
is not available, right?
Thanks,
Alex.