Hi Marcel,

We are agree, but maybe we can provide:
- a default runtime distribution which packages Karaf, ACE bundles/features, ACE shell commands. The newbie could start with this package, simply uncompress the archive and run ace. The starter is ready. As Karaf support WAR deployment, we can provide a ACE feature to also deploy the web console. - a "expert" distribution which packages ACE bundles. The "power" users could deploy this distribution in any OSGi container: a pure Felix or Equinox one, a custom Karaf, etc.

In both case, I would like to provide script to deploy agent to an OSGi instance.

So, the distributions could contain:
1/ Runtime
        1.1/ Karaf
        1.2/ ACE core resources (bundles/features)
        1.3/ ACE extra resources (bundles/features for webui/vaadim)
1.4/ ACE shell commands, including a command to deploy an agent into a target Karaf instance 1.5/ out of the box scripts to start ACE runtime, and deploy the agent on a target OSGi instance 1.6/ ace-maven-plugin containing ace:deploy-agent, ace:run goals which to quite the same as 1.5
2/ Expert
Provide all ACE bundles in jar format, ready to be installed into any OSGi container.
The web ui/vaadin war is also provided ready to be deployed.
The script to deploy ACE agent is also provided.
3/ Source
Just a packaging of all ACE sources.

WDYT ?

Regards
JB

On 02/23/2011 02:38 PM, Marcel Offermans wrote:
I would prefer a distribution that can also run on just any OSGi container. I 
don't mind also providing a feature so one can easily run it in Karaf, but I'm 
hesitant about making ACE somehow dependent on more than just the OSGi 
framework itself.

I know we currently use Pax Runner to bootstrap, which (especially when 
developing) is a very flexible way of launching frameworks, but with the new 
launcher API we should be able to write a very small piece of bootstrap code 
that launches ACE.

By the way, I completely agree with your end goals JB, run the server with a 
simple script and run the target by deploying an agent in an OSGi instance 
(and/or provide an OSGi instance with an agent).

Looking ahead a bit (not for now) we should be able to eat our own dogfood and 
be able to provision an ACE server with ACE. Of course the first ACE server 
would have to be bootstrapped with a script anyway. ;)

Greetings, Marcel


On Feb 23, 2011, at 12:50 ,<[email protected]>  <[email protected]>  wrote:

I would prefer to create a custom karaf distribution. I can do it very quickly.

Regards
JB
-----Original Message-----
From: "Geert Schuring"<[email protected]>
Date: Wed, 23 Feb 2011 12:49:21
To:<[email protected]>
Reply-To: [email protected]
Cc:<[email protected]>
Subject: Re: [PROPOSAL] ACE 0.8.0 ReleaseCandidate

I was thinking exactly the same. We can re-use the servicemix
configuration for creating a Karaf based Ace distribution.

Geert.

About the same topic, what do you think about provide a runtime
distribution based on Karaf ?
We can create a distribution kit based on Karaf + ACE feature.

I heavily think that it should be the target.
The end-user only need to launch bin/ace to have the console and all
runtime bundles up and running. He just has to deploy the ACE agent into
the target instances.

Regards
JB

On 02/23/2011 10:37 AM, Jean-Baptiste Onofré wrote:
Hi again,

FYI, I propose to add an assembly in 0.8.0 release. The assembly should
provide the following artifacts:
- ACE sources archive (zip/tar.gz)
- ACE agent runtime (zip/tar.gz)
- ACE consoles (zip/tar.gz)

The purpose to give the user the possibility to quickly start with ACE.

WDYT ?

Regards
JB

On 02/23/2011 09:08 AM, Jean-Baptiste Onofré wrote:
Hi all,

Geert and I are working on fully support of Maven 3 in ACE (properties,
cleanups, etc).

I hope to complete it end of this morning.
In the mean time, I deployed Felix Metatype, Dependency Manager and
Deployment Admin into the ACE m2 repo:
http://svn.apache.org/repos/asf/incubator/ace/repo/org/apache/felix/

and defined this repo in the ACE pom.

It allows users to be able to build ACE directly without building
previous dependencies.

I think that we are not so far from a release now.

I propose:
1/ to release ACE 0.8.0 asap (I can do it tomorrow). It's a
ReleaseCandidate version. The purpose is to check the release process
(identify if there are some particular points) and document this
release
process.
2/ The 0.8.0 will allow us to prepare demo/tests and identify what's
the
pending points. My purpose is to be able to prepare a ACE 1.0 release
for end of march, and allow Marcel and Karl to announce ACE 1.0 at
OSGiCon/EclipseCon during the ACE talk.

WDYT ?

Regards
JB







Reply via email to