SCM proposal

[risc at volumehost.com]

This mail is in response to a conversation on IRC between myself, and lazyb0y in #fai.

If i were to describe an ideal SCM system for the FAI crew to use, it would be structured as the following:
i'm familiar with CVS, so i'm going to use it as an example. i realize we'll end up using something different. ;)


1 cvs server, for releases. thomas is the only one with write access to this repository.

1 cvs server, with multiple cvsroots structured as such:
1 tree per "fai-developer", which is to say, a /juri for me, to place our /usr/local/share/fai directories in (for reference). developer write/world read.
possibly limit the disk space these per-user trees occupy, i have all my FAI stuff stored on a 32 meg memory stick, but others may not be so...
1 tree with a shared copy of the FAI package, for the developers to "hack on". access to this tree would be default yes, unless someone continuously broke things for others. ;)
cvs commits to this tree would go to a fai-developer mailing list, for thomas to pick-and-chose among, for the stable cvs repository.
1 tree for "fai-utils", tools fai developers feel like sharing/modifying (for instance, i'm already started on a lint checker). again, developer write/world read.

of course, our /usr/local/share/fai repositories will have to be sanatized for passwords/sensitive_data.

anyone have any better ideas? ;)

Julia Longtin <risc at volumehost.com>