My comments inline.
Rick Rineholt wrote:
I think I've asked this indirectly, but I'll be more blunt:
Would one of the proposed objectives of the branch be to eliminate the
current mix of kernel being published as snapshots and the rest of
Tuscany to use those snapshots ?
I am finding the current mix confusing and difficult to work with in my
Eclipse environment, even with source attached to the JARs, so I'm not
particularly interested in repeating this in a branch.
To be able at the head of this branch check out all of Tuscany SCA
(not SDO/DAS) and do a build from the top of that and expect all of
Tuscany to be running the code just compiled?
To try and to stabilize this to run most of the samples and iTest and
keep them running?
Yes that's one of the main goals, build and test the integration of all
the pieces in a stable environment.
To go forth from there to add the features we said we were interested
to work on this branch?
Then later at some point when those features that require more
involved refactoring of the kernel on the trunk have stabilized and
can run on a relatively regular basis the iTest and samples to merge
them?
I think people should be able to work on code in the environment that
works best for them and the community (trunk, branch or a sandbox), then
merge when possible. The main goal of this branch is integration and
stabilization, so I'd prefer to not use it for big new features that
would destabilize it, but I'm sure that we'll need incremental
improvements and fixes to the existing features in that branch to get
the end to end integration working.
If the answers to these are yes, I think that would work out better
for me and would prefer that approach.
Thanks.
Rick Rineholt wrote:
I'm fine with the content that both the list that Sebastien produced
to bring up Tuscany to an SCA 1.0 spec and the work that was
previously discussed for the improving the kernel referenced by Jim.
I think both sets add value for our users. However, for me the
branch is not about what content is right, or even how it's
implemented, but more how it would be developed.
I find I need to be able to work with running user samples and a
complete systems to help figure out how pieces and part fit together.
The current split between prespec published and snapshot core, and
using mostly an old kernel for the rest of Tuscany truly hampers that
and makes anything outside of the kernel confusing as what is
belonging where.
If it must be, I prefer a branch that can give the alternative back
to where Tuscany SCA as a whole can run without published snapshot
kernels and kernel functionality can be added and run immediately
with remaining Tuscany without having to republish it. I see this
even with some of the instability we had as a reason for moving to
the current approach still a preferable environment to work in.
Simon Nash wrote:
A March release with basic functional improvements in a consumable
package
(kernel, selected extensions, and tools) makes sense to me.
As well as the items suggested by Sebastien, I'm interested in adding
flexible ordering of elements in SCDL files as required by the SCA
spec.
Having work on these items proceed in a branch so that it does not
conflict
with the restructuring and distributed deployment work going on in
trunk
would allow people to be more productive, with less interference
between
the different activities in progress.
Simon
Jeremy Boynes wrote:
Guys,
I'm a little confused here - so far we seem to have 3 different
people volunteering to manage 3 different releases. We now have a
very very long list of "requirements" many of which have not been
discussed on the list and most of which do not have names against
them or really relate to the coding that is actually going on;
they also don't seem to apply to two out of the three releases.
Version numbers are being assigned to milestones, we have
stabilization branches and end-to-end scenarios, all without
meaningful discussion on this list.
I think we need to stop and figure out what we are doing as a
community. Here, on this list, with everyone involved.
--
Jeremy
On Feb 6, 2007, at 3:58 PM, Jean-Sebastien Delfino wrote:
Jim Marino wrote:
Sebastien,
I'm a little surprised that you have not referenced the previous
release discussion thread or any of the work that has been
ongoing in core over the past month and a half:
http://www.mail-archive.com/[email protected]/msg12291.html
http://www.mail-archive.com/[email protected]/msg13445.html
http://www.mail-archive.com/[email protected]/msg12238.html
Most of the work in core during this period has been aimed at
getting a release of kernel out that supports features outlined
in the first referenced thread. How does your proposal relate to
that release? I'm happy to have two simultaneous release
processes going at once and think it could even be beneficial.
However, it would be helpful if you put your proposal in context
so others such as myself can understand it a bit better.
Jim
On Feb 5, 2007, at 5:19 PM, Jean-Sebastien Delfino wrote:
Now that we have a list of requirements on our Wiki at http://
cwiki.apache.org/confluence/display/TUSCANY/Feature+areas+and+what
+folks+are+working+on, and a number of people are signing up
for some of the corresponding work, I'd like to start a
discussion on the content of our next milestone. Given that our
last milestone was in December, I'd like to have another
milestone soon, by March.
Here's the function that people have already signed up for on
the Wiki page + what I'm interested in for this milestone:
- Support for complex properties and multi valued properties
- Support for SCA deployment-contributions, and in particular
support for JAR based deployment contributions
- Ability to reference and resolve composites in an SCA domain
(would be nice to support recursive composition but I'm not
particularly interested in it)
- Ability to configure and override the configuration of
References, Services and Properties (again here I'd be happy if
this works with just one or two levels of composition)
- Support for wiring inside an SCA domain references to
services with bindings and have the wiring decide the endpoints
to use
- Support for business exceptions in end to end interactions
- Support for promoting services and references out of a
composite (without having to wire a reference to a reference or
a service to a service)
- Support for defining and configuring services and references
directly on components
- Interchangeability / mapping between Java and WSDL interfaces
- Ability to use, alter and write an SCDL model at deployment
- WSDL2Java and Java2WSDL support using the SDO databinding
- Core support for non-blocking invocations playing nicely with
bindings, and without having to send complete routing paths to
the services/references
- Databinding framework with support for conversions between
JAXB and SDO
- Working and modular build allowing to build subsets of the
project
- Services to add(/remove/query) compositions to an SCA domain
- Services to add(/remove/query) SCA deployment contributions
to an SCA domain
- Core support for addressing, resolving, loading artifacts
from SCA deployment contributions
Thoughts?
--Jean-Sebastien
Jim,
The idea is to bring together a number of pieces from the core
runtime, extensions like databinding and WSDL support, tools like
WSDL2Java and Java2WSDL etc. and stabilize them to get some of
the basic function that I listed in my earlier email working in
end to end scenarios. As a first step, we probably need a very
small subset of the new deployment story that is being built in
the trunk, starting with the ability to work with one SCA
composite and one JAR contribution.
To have a stable integration by March, I think we need to start
this effort now. In order to not disrupt the wider and more
innovative work going on in the trunk I'd like to do the
integration/stabilization work in a branch, starting with the
kernel from the pre-spec-changes branch or a stable level from
last week. This will allow the trunk to continue to evolve in
parallel and at a faster pace to support things like federated
deployment, new management services, JMX support, multiparent
classloading, and the latest changes to the Java C&I APIs.
--
Jean-Sebastien
--
Jean-Sebastien
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
