On Mon, Oct 14, 2013 at 1:19 AM, Emmanuel Bourg <ebo...@apache.org> wrote:
> Le 14/10/2013 08:06, Henri Yandell a écrit : > > > What do others think? > > I quite like triaging the feature requests with a 'n.x' version and a > '(n+1).0' version. The former indicates that the request can be > implemented in a compatible manner, and the latter indicates this can't > be done without a major release. > > The pain I'm finding with that is that quickly it turns into a big bucket of n.x. I liked it at first, but it's become a backlog bucket (I'd argue 2.x, 3.x and 4.x are all one backlog bucket as the true test of their version will only be when the code is complete). The issues come in, end up in a bucket, and stay there. A smart user opens their issue against the next version and has a 50/50 chance of it getting in. Better I think to treat it like a pipeline. New issue, accepted, code needed, review needed, committed (against version 3.2 etc). I could imagine others - test needed is a common one for us and might be worth its own section. Perhaps 'research needed' for the nearly complete code that has some weird item or the bug that needs figuring out. Hen