Hi,
over the last few months, release managers (d-i + “regular” release
team) have been quite open to various improvements in d-i components,
but it looks like we really should concentrate on finalizing things
for wheezy.
Things like improved consistency, or buildflags-related changes are
examples of things which can easily bring regressions for next to no
short-term benefits. Of course, if a real bug is fixed in the process,
those can be considered on a case by case basis.
[A] I think we could just keep on pushing things to regular 'master'
branches, and accumulate bugfixes and changes targeted at wheezy in
'wheezy' branches. Ideally, we would push releases from 'wheezy'
branches to unstable so that they get user-exposed for a while, and
only push releases (if needed) from 'master' branches to experimental.
[B] Another way would be to stage changes in 'jessie' branches and
keep everything we want for wheezy in 'master' (I don't think the l10n
update process can work with something which isn't master; Christian?).