Adam Williamson wrote:
As I said, the particular code isn't the issue. We ship a kernel
API. At
present, we consider it fine to break that API in stable releases. This
is not something that would be considered 'stable' in a traditional
definition. The kernel's just an example, we do the same kind of
non-stable updates all over the place. That's the issue I'm trying to
talk about, not just the specific example I happened to mention. Please
don't bog down in specifics.
Well, the specifics are that packages both within Fedora and in third-party
repositories which depend on the bumped API usually get rebuilt (and patched
if needed) fairly quickly, normally before the update even goes stable. Of
course that's only possible for software which can be patched, which is just
another example of how binary-only software is broken by design.
Kevin Kofler