Hi,
Finally we move from philosophy to specifics.
I think there are 3 parties involved in here: iptel, voice-systems and users. Iptel and voice-systems have their own agenda's, and fights between them end up harming users.
The project model started by iptel took us very far and made ser a great application, but it's own success should make them realize that changes are needed. Greger's and Iqbal's email are I think mostly right and layout a good view of what is going on here. And Andrei's email is a good way to start "peace talks".
Clearly, SER has grown far so that different interests apply and this has caused friction. Friction means there is a problem. I do think that forking is not the best solution, but things MUST change. Contributions must have an easy way into SER: starting maybe as experimental, then moving into unstable, then as testing and finally into stable (kind of debian approach). And this process must be well-defined and NOT CONTROLLED exclusively by a company (be iptel or voice-systems, same-o, same-o). If the contribution is voted as desirable by a community of users, it is well tested and so on, it should be accepted. TLS is only the first, but what will happen when someone develops a free-LDAP module (actually, i think there is one ... just never became mainstream) or any other module which now is kept private by iptel/voice-systems?? If SER is opensource, let it be the whole way.
Summarizing ... I think the shake up by the voice-system's guys is good and desirable, but now it is time to stick together and work out a compromise which works for everybody (and specially for the user). I think Andrei's email is a good start. Let's work on it.
Cesc
On 6/15/05, Andrei Pelinescu-Onciul andrei@iptel.org wrote:
I think that a fork is bad for everybody. Whatever way you put it, it will still involve extra overhead at least for each code merge.
So this mail is an attempt to come to a compromise that would be acceptable for everybody.
My first proposal would be to have another branch on ser cvs (berlios). Something like openser_0.x.y (where the version number should reflect somehow that it's newer than 0.9 but not unstable). Bogdan and Daniel would mantain this branch and they would be able to release from it as often as they want. It would be some kind of stable on steroids :-)
This would have the main advantage that code merges/diffs/fixes propagation would be much easier and much faster. Administrating all the stuff will be also easier (like common bug tracking system a.s.o). It would be also easier for users, it will be much more easier to keep track of the changes and to choose an appropiate version.
There must be however some rules. Something like: backporting/forwadporting from unstable/stable should be ok, the same for minor patches (e.g. xlog colors), but not major code changes, not present in unstable and without the code maintainer approval (to maintain future compatibility). Adding new stuff (new modules) should also be ok, but it would be preferable that these module come from either unstable or experimental.
The mailing list for development should remain serdev. For users it could be a different one (if the changes or the mail traffic gets too big).
As far as integrating new contributions faster in unstable, I agree that this is a problem, but I think we can sort it out. The experimental repository is a beginning and if there are better ideeas on how to speed things up while maintaing sanity, please speak. Maybe some kind of reviewed patches (e.g: someone sends a patch that adds a new feature in core, Bogdan reviews it decides it's ok and blesses it => I will integrate much faster something blessed by other ser developer).
If you don't agree with my proposal, please say what you don't like and what you would like changed so that it would be acceptable for you. I'm sure that we can reach a compromise and hopefully improve ser development pace in the process.
Andrei
Serdev mailing list serdev@lists.iptel.org http://lists.iptel.org/mailman/listinfo/serdev