On Thu, Jan 22, 2009 at 11:31 AM, Mirko Jahn <[email protected]> wrote:
> >From BJs point of view, it is quiet rational. As soon as any "no > matter what kind" of API break appears, the major version has to be > increased. This is a safe approach from the point of a bundle > provider. > Peters and my suggestion aims more at the re-usability of bundles but > puts more burden on the "user/consumer". The advantage in my point of > view is that clients of the interface (only working with the interface > and not implementing it) are more flexible and can stay compatible for > a longer time without the need to change the bundle. I agree that the difference between a "user" and an "implementor" is very significant and should not be under-estimated. A general "incompatible change" attitude doesn't cut it. Even a bug fix in a supporting class could be considered "incompatible" under extreme circumstances. Personally, I would call for implementors to do a "[1.2,1.3)" constraint, whilst the users of the interface can specify "[1.2.0,2.0.0)" with confidence.. Cheers Niclas -- http://www.qi4j.org - New Energy for Java _______________________________________________ OSGi Developer Mail List [email protected] https://mail.osgi.org/mailman/listinfo/osgi-dev
