On Wed, 27 Nov 2024, 22:00 Matthias Bläsing, <[email protected]> wrote:
> Am Mittwoch, dem 27.11.2024 um 21:39 +0000 schrieb Neil C Smith: > > > > As I've said before, I disagree with this. For seven releases 12.0 to > 12.6 > > we did not require this, nor could we. Switching to a single incrementing > > version every release was not intended to cause this extra overhead for > > plugin authors, and all it achieves is less testing of the APIs during > > releases. A plugin not functioning is just as, if not more, likely to be > an > > issue on our side! > > Author decide for which NB version they make which Plugin version > available. Consider the situation: > > Plugin Version 1.0 uses API available from NB 19 onwards > Plugin Version 2.0 uses API available from NB 24 onwards > > How, if not the author, should make the call which NB version is the > baseline? > Automatic validation of plugins copied from one IDE version to the next does not break that scenario. To achieve adequate testing at the moment we serve the previous plugin centre to the first two release candidates. That is not ideal for testing purposes. > And before you say: "A trivial!, just extend from last version!" What > happens when we to a point release in the order: > > NB 23 > NB 24 > NB 23.1 > No one has shown any interest in doing point releases, despite the conversations around JDK 8, etc. So it's neither a consideration, nor would actually be a problem (as 23.1 would retain 23's plugin link in any case) Best wishes, Neil
