On Feb 19, 2007, at 9:45 AM, Jim Marino wrote:
There has been quite a bit of activity over the last month-and-a-
half enhancing the Kernel. Based on this work, I'd like to cut a
release of Kernel, the Standalone Runtime, the Webap Runtime, and
the Maven iTest Plugin as a stepping stone to having a 1. release.
I was thinking we would call this release "1.0 alpha".
Works for me.
I'd suggest three release bundles (source):
* kernel
* runtime (which includes the three runtimes you mention above)
* core samples
with binary distributions of the standalone assembly plus maven
artifacts (including the war and itest plugins).
I see this "alpha" as evolving into a series of iterative releases
over the next month where we introduce some of the more
"compelling" features we have been discussing related to service
networks, federation, and deployment. The primary goals of the
first alpha release would be centered on enhancements to the
programming model that have been introduced with the recent Kernel
changes. Specifically, the alpha would provide an enhanced version
of Kernel that our users can experiment with, extend and provide us
feedback on. This will assist us in validating he programming model
supported by Kernel.
The key features of the alpha release would be:
1. SCA 1.0 APIs
-Support for many of the new SCA 1.0 Java APIs (ComponentContext,
Conversational annotations)
2. An enhanced standalone runtime with JMX support
3. An enhanced and SCA 1.0-based model for integration testing
(elimination of SCATestCase, which is not spec-compliant
I propose we remove the "test" module.
4. Simplified wiring
5. Simplified extension model
6. Architecture for support of federated deployment
7. Support for web applications using SCA 1.0 concepts
I'd like to follow the alpha with additional releases that
introduce additional support for federation, deployment, and the
SCA 1.0 APIs. To stage this, perhpaps we the following in the next
release after the alpha:
- Contribution service
- Refactor of Databinding (mentioned in a separate thread)
- Introduction of master/slave nodes and federated wiring
- More complete support for conversational APIs, including
ServiceReference
In terms of work items, I think we need the following (besides a
stable kernel :-) ):
1. Standalone runtime operational and able to deploy application
and extension SCDLS
2. At least two samples. I propose the Calculator Sample
(Standalone and Web app) and the Loan Application Sample
Feel free to suggest additional features. As a general principle,
I'd like to get a release out sooner rather than later with "big"
features introduced in the consecutive releases mentioned
previously. One thing I'd like to see if we can fit in but may have
to cut is the new PhysicalComponent builders. That may be something
we stage later.
Hopefully, we can cut the release this week.
Thoughts?
The extension model is a bit hokey at the moment requiring users to
create new or modify existing profiles which basically means
duplicating the installation every time. We've had this view for a
while that extensions should be dynamically and automatically loaded
based on intent and for the 1.0 timeframe I think we should provide
that. However, this really requires the Contribution service be fully
working to we can match intent to extension and I don't think that
will be ready soon.
We do support a very primitive extension mechanism where users can
add them by modifying the system scdl (which is now externalized as a
text file). I'm OK with releasing an alpha version like that with the
intent-based support coming later (i.e. 1.0 beta or final).
I think we need to do some clean-up on the poms. As Sebastien pointed
out, there is a lot of dependency cruft in the SCA pom which should
be stripped out - we can probably reduce that to just the
configuration of checkstyle etc. I'm happy to don my build-monkey hat
again and clean that up.
--
Jeremy
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
