I hope you're all not too tired of my infrastructure emails, but it has to be done so we can get to the good stuff. This one is about how to organize the CVS repo. Here is my proposal, based on my own thoughts and some of the mailing list discussion.
Each module will have its own directory. The organization below this level is TBD (probably however avr-libc does it). The module directories will be organized into groups (we'll need a better name for this) based on functionality (perhaps a communications group, a time group, a protocol group, etc).
In order to distinguish between development and stable code, there will be a stable branch that modules will be merged into once their APIs are stable (some of this might have to be changed depending on the best way to use CVS). There was some worry about code forever remaining "unstable," but as long as the focus is on the API being stable, I don't think it will be a problem. We could even have a set amount of time for a "release candidate" to be in testing before it automatically moves into stable (unless serious problems were found). I dont' know if development code should just be in the trunk, or a separate branch, or even a separate branch for each new module (someone with more development experience should chime in here). There would also be another branch for contributed code, which would basically just be a code dump.
So that's what I'm envisioning. Send out any comments to the list. Ruddick