Broken upgrade paths in F7

[Ralf Corsepius]

On Tue, 2007-05-15 at 10:50 -0400, Jeremy Katz wrote:
> On Mon, 2007-05-14 at 19:21 -0400, Adam Jackson wrote:
> > I've suggested that releases should branch from devel before they
> > actually release.  It hasn't been implemented.  I'm not sure why.
> 
> Because (as the discussion showed :-), that has other downsides.
Really, I don't buy this.

If you had taken a snapshot of FC7 and used this as basis for the DVDs
(aka freeze), simultaneously to forking rawhide, all this discussion
would have been mood. 

FC7 updates (after "freeze", before "official launch") could have gone
to FC7/updates without interrupting rel-engs work on the DVDs nor would
rel-eng's work have interfered with "maintainer's work" (as it now
does). 

The only thing which would have changed was "the release's repos being
exposed" to the public before an official "launch" (i.e. a marketing
event).

> Really, we need an easy way for maintainers to do branching themselves
> so that the branching can be done at the most suitable point on a
> package by package basis.
I also don't see this. At the very moment you fork rawhide and
<release>, you decouple rawhide and release, i.e. maintainers can chose
which version to push into which release version. ATM, you block
"upstream" due to the "freeze".

This is the working principle how FE had worked ever since it existed.

Ralf