Am Freitag, den 23.06.2006, 01:30 +0200 schrieb Stefano Bagnara:
> I won't be there, but I would appreciate if this topic could be 
> discussed there:
> 
> 1) We have only 1 committer able to sign releases: imo this is *really 
> bad* to James. Is in our power (as PMC) to decide that only final 
> releases need to be signed with a trusted signature while at least 
> milestones, alphas (and I think even betas) could be released signed by 
> "untrusted" keys? I think that *MANY* apache projects currently have 
> untrusted official distributions so I don't think why James need to be 
> so "strict" in this regard.
> 
> 2) Decision system analysis. IMO James PMC have BIG problems in managing 
> the project. Most of the few developer/architect time we have at hand is 
> lost because we have not been able to be *agile* in our decisions. We 
> loose weeks in design discussions that *rarely* (or better never) ends 
> up in operative plans. Most times the discussion is stalled and 1, 2 
> years ago, when someone find the time to work on the topic it is hard to 
> understand what has been decided and most time things are no more up to 
> date because the world has moved forward in the mean time.
> I think we need to find a procedural solution to this problem, and here 
> are a few proposals:
>   a. Define macro goals and related priorities (IMAP support, Container 
> change, repository refactoring, and so on). Priorities are not a way to 
> sort operations, but way to how much the feature is important. If I work 
> on a primary goal I'm more allowed to create incompatibilities, if I 
> work on minor things now, and so on.
>   b. Leave at the developer or group of developers design and 
> implementation of this features working this way: if the change is 
> simple, short in time, and isolated just apply it and the we review it 
> (CTR), if the change is bigger the developer that understand this 
> automatically create a shortliving branch to commit, work, and show us 
> the work in progress. I would avoid at all costs long living branches.
>   c. I would leave to the developer the option to discuss implementation 
> (or even design) details before coding. Either way each PMC member can 
> cast a veto and the developer have the responsibility to revert the code 
> that has been vetoed. (Of course vetoes needs explanations as per Apache 
> rules)
>   d. We should definitely avoid to start design discussions if we don't 
> have the time to finish them, or at least make clear in the message that 
> it is only a stone thrown in the lake to count how many jumps we can make.
>      As an example I would like to bring the last talk about James 
> configurations. I created a JIRA issue (JAMES-495), everyone have been 
> excited by an "UML" word in a message.. we had a good hi level approach 
> proposed Steve (2006/05/27). I spent my time to understand the proposal 
> and make an analysis on our code and I saw possible flaws in the 
> hi-level design so I wrote asking for more details (2006/05/28). 4 weeks 
> later no one replied to my message and we have *Yet Another Unfinished 
> Proposal*. This is just an example. We all loose time if we don't take 
> the responsibility to finish what we started.
> 
> 3) 2.3.0, 2.3.1, 2.4, 3.0, TNG and what else: I already said this, but I 
> would like to bring again this to your attention.
>   a. Imho we should keep working only on the trunk and eventually 
> backporting code to the branch.
>   b. I would keep only 2 channels (trunk/branch).
>   c. Code is always committed first to trunk (at least if we don't have 
> branch specific issues) and eventually backported.
>   d. After 2.3.0 will be released we keep working on trunk. If needed 
> 2.3.x releases we'll be created on the branch by backporting code from 
> trunk or just for mantenaince code (we'll decide this at that time)
>   e. At the end of september we'll evaluate what we have in trunk and 
> how good has been the 2.3.0 release and we'll decide wether we should 
> try to branch for 2.4 or keep working on trunk for a while, until 2.3.X 
> is stable enough to close its branch.
> 
> 4) Code issues: I want a "vote like" to apply this issues, then I would 
> like to start back with the CTR. Some of them already contains the code, 
> for the others I'll try to attach the code soon, otherwise please vote 
> if I can start working on the trunk for a Commit-Then-Review session of 
> the issue.
>   a. Cleanup/Refactor FetchMail code
>      http://issues.apache.org/jira/browse/JAMES-509
>      Code is there (not tested and not completed, but enough to be voted)
>   b. Create a RemoteDelivery service
>      http://issues.apache.org/jira/browse/JAMES-520
>      I already have code also related to SpoolManager refactorings that 
> I would like to commit in trunk and proceed with step-by-step 
> refactorings in order to experiment this and the JAMES-491 (see at the 
> end) in relation to the thoughts Noel reported to the list the past week 
> about the processor/spool manager/queue.
>   c. Mail/Spool/Message repositories refactoring
>      http://issues.apache.org/jira/browse/JAMES-521
>      Main goal is IMAP here: I would like to commit the code from 
> Joachim and start working on this refactoring.
>   d. Refactor James services to extract common code and isolate 
> cornerstone/excalibur dependencies
>      http://issues.apache.org/jira/browse/JAMES-516
>      Some code is already there, much more can be done, but it is a step 
> (the code I attached is not my last code, but I'm completely out of 
> synch due to this procedural problems).
>   e. Refactor the service methods to inject services via setters
>      http://issues.apache.org/jira/browse/JAMES-494
>   f. SpoolManager refactorings
>      http://issues.apache.org/jira/browse/JAMES-491
> I also think that we should vote to decide wether I can work again on 
> the trunk with a CTR approach or not. I stopped my work on trunk in the 
> mean time. I want to understand why I cannot proceed with CTR, or my 
> codebase will be so different from james-trunk that I will be never able 
> to synchronize again with James. Incidentally I'm a lot more busy in 
> July and on holiday in August, so this is not an urgent issue, but I 
> really want this to be solved in a clear way as soon as possible.
> 
> 5) JSPF: we need to release it. Norman wrote a message titled "jspf 
> release" (2006/06/06). I replied with a test build, a test website and 
> more informations. No one else replied.
> We voted to include Jspf in the James project: I would expect help 
> making the first release, but if you can't help, please at least reply 
> that you agree to make a release or anything else.
> 
> 6) Postage: it needs a JIRA project so we can separate Postage issues 
> from James issues. We should also make clear (at least to Bernd) what we 
> expect from him and what can we do for Postage. I can help with the 
> maven2 build/website once we decide how to publish the JSPF website and 
> the jpsf release.
> 
> 7) OSGi: I crossposted a message from the felix list about this. I would 
>   really like to be there to talk with felix authors and hear their 
> suggestion on how James should ideally be "built" (refactored) in the 
> felix world. The main issue I would like to submit to them for Ideas is 
> how to manage configurations and reloadable configurations for james 
> services (take also complex examples like processors, mailets and 
> mailets configurations). I think we are far from moving to another 
> container, but I would also love to be there and being able to talk to 
> some OSGi/Felix guru about this. Maybe in future I'll try to open a 
> mailing list thread on felix about this, but I wanted to bring this 
> point to your attention because they offered to help at the ApacheCon 
> and I think that it doesn't happen so often that 3-5 james committers 
> are able to directly talk about similar issues.
> 
> 
> I hope that at least some of this issues will be evaluated at ApacheCon 
> and that you can at least find an agreement on this issue (to my 
> proposals or alternative proposals) so we can vote and finally solve 
> them soon.
> 
> 
> Stefano

8)SMTPServer: Discuss what will be the best "method" for support a
plugable SMTPHandler api like we did for mailets. Also discuss howto
support fastfail etc.

bye
Norman

Attachment: signature.asc
Description: Dies ist ein digital signierter Nachrichtenteil

Reply via email to