Le jeu. 10 août 2017 à 15:52, Erick Erickson <erickerick...@gmail.com> a
écrit :

> I've never understood this. Why? I realize we don't want to keep
> release 6.x forever, but what is it about a week in this case that's
> so horrible? We're not talking about introducing back-compat issues
> here.
>

We only have backward-compatibility tests. If the new version to be
released is greater than all other versions that have ever be released,
this is just fine. However if we release an older version, we need to make
sure not only that it can read older indices, which tests check
automatically (they don't test all cases, but enough to give us good
confidence), but also that all greater versions can read indices generated
by this version (forward compatibility), which falls under the
responsibility of the release manager since it is not automatically tested.
In my opinion we should not take the risk: let's just produce sequential
releases and recommend users to use the latest version.

I see 7.0 needs to be delayed a bit more to fix some JMX issue, so this
gives us a window in case someone wants to be a RM for 6.7 without delaying
7.0, but in that case let's build a RC and start a vote as soon as possible.

Reply via email to