Also, 1.0 portlets may be "surprised" if they got multiple
performBlockingInteraction, e.g., if such an interaction results in an order
being submitted to an ERP system. IMHO, the spec language should be clear on
what v1.0 is - the "choreography" may indeed change in future versions, but I
think it would be clearer to change the spec then rather than leaving things
open before a consensus has been reached on how to address them moving
forward.
My two cents,
Eilon
Post 1.0 will have a
different binding and/or versioning information (actually that is missing in
the protocol, change request to follow). So it could be handled when consumer
and producer implement different WSRP
versions.
Alejandro
Michael Freedman wrote:
I think it does
-- as this is a requirement of how consumers interact with producers.
By making this statement we will require that all post 1.0 consumer
still adhere to this requirement when communicating with a 1.0 producer as
the 1.0 producer my have coded to this restriction.
-Mike-
Alejandro Abdelnur wrote:
The current
wording does not restrict ourselves for the future, in the future we could
say it could be more than one and it would not break backwards
compatibility.
Alejandro
Rich Thompson wrote:
Document: Spec Section:
6.3.2 Page/Line: 44/28 Requested by: Michael Freedman Old text:
The Consumer MUST initiate at most one performBlockingInteraction() for
any one client request. New text: remove this newly added
line
Reasoning: Do we really want to restrict ourselves by
this for the future? As we haven't addressed events yet, should we
be cautious to not handtie us if we find a use case/rationale that
through handling of events would cause another blocking action to be
called? What is the harm of NOT limiting ourselves. As long
as each adheres to blocking rule it would seem
safe.
|