- From: binod pg <
- Subject: Meeting Notes
- Date: Thu, 22 Aug 2013 09:37:01 +0530
- Organization: Oracle Corporation
Attendees : Brian, Keith, Eric, Binod.
- We discussed application composition primarily.
- We felt that the containers can optimize the number of internal
objects that gets created as
a result of application composition. The 289 container already allows
for such optimization.
If there are any further relaxation further needed, we will make that
in the specification.
Eric mentioned that there may be one such case, for which, he will
dig up an old email thread.
- There was a short discussion on whether we should allow a mini
composition within a single
application itself. There are ways to do java level modularization
already. However this does
not solve the original composition performance.
- There was a discussion on whether distributed containers would be
benefited by a application
chain level locking mechanism. Some kind of a key identified for the
chain will be useful or not.
At the moment, containers are able to do such optimizations for a
large majority of cases already.
So, it does not seem necessary to make changes in the specification
for the same.
- We then had a discussion on state sharing between applications in the
chain. It does appear like
an issue, though it is against the pipe and filter architecture of
the composition. For the same
application that appear again in the chain, it does not break
application independence. However,
the most of us in the meeting seem to favor minimal state sharing
across all applications rather
than just the same application. At the moment, applications optimize
this by sending data as
custom headers, which can be avoided by a standardized mechanism for