I wanted to be sure of this since we are scheduled to work on the
glossary tomorrow, and I would like to capture at least as much as we
managed to reach consensus on.
At 11:11 AM -0700 6/11/02, Michael Freedman wrote:
I think you mostly got it. The key mistake (if I understand your e-mail)
is that the client/consumer session is equivalent to the sessionID in the
API. The client/consumer session is of no concern to the API as its
something between the client and the consumer vs. the consumer and the
producer. Yes, its likely that a consumer will tie its session
conversation with a particular producer to live within this
client/consumer scope but it need not.
Rex Brooks wrote:
Unless I had cotton in my ears, I think we neared consensus on what a
Session is for the purposes of having a sessionID for the interface.
Correct me please if I am wrong, but I believe we arrived at a point
where we can say that what we mean by Session is the conversation
started by one end-user requesting service from one consumer, such
that a sessionID is created by the consumer whose responsibility it
is to manage this session/conversation with the end-user, providing
for delivery of service from one or more producers.
There was also a related concept put forward for consideration called
a sharedSession. This concept had many possible configurations, but
the basic idea was that something like this was needed, in addition
to transientEntities for economical management of multiple portlets
within containers, multiple producers within a session, etc.
TransientEntities and sharedSessions were not quite narrowed down
enough for consensus to emerge, and I may be mistaken about
sessionID, but we certainly worked tenaciously at the interface