Hi, On Jan 30, 2011, at 12:39 PM, Marcel Offermans wrote:
> Hello JB, > > The absolute minimum requirement is that we release the sources. I would > definitely be in favor of having a single archive that contains all of them > and that can be unpacked and built with a single command. The result of > "building" should at the minimum be a set of OSGi bundles, but maybe we > should also automatically generate some assemblies that can directly be "run". I agree we should aim for as few artifacts as possible. On that same note, we might want to look into reducing the number of bundles too; for instance, we have a number of dedicated API-bundles which can very well be merged into other ones. > > As binaries, I would like to release two "assemblies", one containing the > whole server, the other containing a target that consists of an OSGi > framework and the ACE management agent. Both must be "unzip and run". > > On 30 Jan 2011, at 7:58 , Jean-Baptiste Onofré wrote: > >> regarding the release preparation, I think that we need to clean >> distribution. I think that the users are a little lost with the current >> artifact delivery. > > I would love to hear what confuses users the most right now, and if the idea > mentioned above would help. > >> I'm working of an assembly, embedding Felix/Karaf to provide a runtime >> deployment platform. > > You did see the ones we currently have, (ace-target-devserver and > ace-target-devgateway)? > > One thing that Angelo is still working on is this one: > https://issues.apache.org/jira/browse/ACE-98 > > That could replace the current target with a small launcher that, using the > OSGi 4.2 launching API can launch any framework you supply, and run it with a > management agent. I think that's the most elegant and simple way of > bootstrapping OSGi with ACE. Work on ACE-98 is slowly moving along. When the launcher target is far enough along (see subtasks), we should be able to make this the de-facto way for starting a target, and drop the target-devgateway. We can then choose to release binaries for (a) the 'management agent' bundle, or (b) the roll-you-own management agent. > > The server now basically runs on a bare bones OSGi framework. Of course you > can deploy a lot of things alongside, and for the developer we do (things > like a shell). Ultimately, the server could also start out as a target and > should be provisionable by a different ACE instance. Of course you then get > into the question "who provisions the first provisioning server" and for that > we need this assembly. :) > >> My question is: >> - do we provide several artifacts (Web UI tarball/zip, file server >> tarball/zip, etc) or one providing all modules (one tarball/zip with web UI, >> etc) ? > > For one, I would not want to release the "file server" at all. It was more or > less an intermediate step to get ACE up and running and not intended for real > use. Agreed; the step from filebased server to 'repository based' server seems to be rather confusing anyway. > > The server with web UI and the target I would want to release separately as > stated above. > >> WDYT ? >> >> Thanks >> Regards >> JB >> >> On 01/29/2011 01:11 PM, Marcel Offermans wrote: >>> Hello JB, >>> >>> On 27 Jan 2011, at 6:45 , Jean-Baptiste Onofré wrote: >>> >>>> FYI, I created a Jira brainstorming umbrella to define the ACE roadmap: >>>> https://issues.apache.org/jira/browse/ACE-115 >>> >>>> The purpose is to kind in mind our discussions and create child tasks >>>> that we will pick up. >>> >>> My suggestion for implementing a roadmap would be to define versions in >>> Jira and start assigning issues to them. That way we can use the >>> "roadmap" feature of Jira to track progress: >>> https://issues.apache.org/jira/browse/ACE?report=com.atlassian.jira.plugin.system.project:roadmap-panel >>> >>> So our first release would be 0.8.0 (for example), and we can start >>> assigning issues to it. >>> >>>> Feel free to complete it. >>> >>> In general I feel we should try to cut our first release relatively >>> soon, and try to get into a rhythm. Therefore in general I think we >>> should look at what's fairly stable now and start with that. >>> >>> Another question I have is what we should release. ACE is very modular, >>> and can be assembled and configured in different ways. On the other >>> hand, a lot of bundles are related. In short I don't think it makes a >>> lot of sense releasing bundles one by one. I would be in favor of a >>> single release of all (stable) bundles. >>> >>> Also, I would like to take the first release as an opportunity to >>> properly baseline all bundle and package versions, use a version policy >>> that conforms to the one the OSGi Alliance recommends and only bump >>> versions when things actually change. That means that our next release >>> will probably contain bundles and packages that will not all be the >>> same. This is very similar to the OSGi specification itself, that is >>> released as for example 4.2 but contains packages with versions like >>> 1.3, 1.2, 2.1, etc. >>> >>>> @Angelo, I know that you're working hard on unit tests, etc. Feel free >>>> to append comments concerning next steps around this topic. >>> >>> It's mainly integration tests that still need porting, as described in: >>> https://issues.apache.org/jira/browse/ACE-79 >>> >>> From the looks of it, Angelo has already made great progress here! Some >>> help from a Maven expert would be nice so we can properly hook up the tests. >>> >>> Greetings, Marcel >>> >> >> -- >> Jean-Baptiste Onofré >> --------------------------------- >> HomePage >> http://www.nanthrax.net >> --------------------------------- >> Contacts >> [email protected] >> [email protected] >> --------------------------------- >> OpenSource >> BuildProcess/AutoDeploy >> http://buildprocess.sourceforge.net >> Apache ServiceMix >> http://servicemix.apache.org >> ----------------------------------- >> PGP : 17D4F086 >
