AI 130 - Attempt to clarify my interop questions on Pull Points

[Peter Niblett <peter_niblett@uk.ibm.com>]

A brief summary of my interoperation question. I would like to use this as
a test for the solutions to 2.58/2.59.

In our discussions we have talked about 3 possible Pull Subscriber types,
and 3 possible NP types.

S1. Simple subscriber. Just wants to create a single subscription and pull
from it. After a while it wishes to stop pulling messages and has no
further use for the subscription or pull point.
S2. Serial reuser. Subscriber wishes to create one subscription and a pull
point. It then wishes to change the subscription to something else and
continue to pull, using the same pull point (either for convenience or
because it wants to switch the subscription before it has finished pulling
the messages from the previous one.
S3. Complex subscriber. The most generic case where a subscriber wants to
create a long-lived pull point and register it with multiple subscriptions
simultaneously.. and to vary the set of subscriptions over time.

NP1. Simple NotificationProducer. Doesn't want to support pull points.
NP2. Minimal Pull supporter. Wants to support pull points, but without any
special optimisation
NP3. Optimised pull supporter. Wants to implement some kind of "virtual
consumer" to avoid holding duplicate messages in the case where there are
many pull subscriptions all for the same message.

Some of these may be a bit theoretical, but at least some of these 9
combinations will occur, including situations where the Subscriber has
limited knowledge of the NP. I am interested in investigating some use
cases, in particular S1/NP2 and S1/NP3. I think it is important that the S1
Subscriber should not have to know whether the NP is type 2 or type 3. So
my questions are

1. What are the steps that the S1 user has to take to get what they want,
regardless of whether it is NP2 or NP3.
2. Is this making the S1 subscriber too complex?
3. Are there any unfortunate things that can happen if the S1 subscriber
does not follow the steps precisely?

Peter Niblett