+1 - This is great for users who want to take longer to upgrade from hadoop-1 and care mainly for bug fixes and incremental features, rather than radical new features. - The ability to release initial 2.x releases marked as alpha/beta also helps to get users to try it out, and also lets them choose what is right for them. - This also lets developers focus on major new features without the burden of maintaining hadoop-1 compatibility.
On Tue, May 26, 2015 at 11:41 AM, Alan Gates <alanfga...@gmail.com> wrote: > We have discussed this for several weeks now. Some concerns have been > raised which I have tried to address. I think it is time to vote on it as > our release plan. To be specific, I propose: > > Hive makes a branch-1 from the current master. This would be used for 1.3 > and future 1.x releases. This branch would not deprecate existing > functionality. Any new features in this branch would also need to be put on > master. An upgrade path for users will be maintained from one 1.x release > to the next, as well as from the latest 1.x release to the latest 2.x > release. > > Going forward releases numbered 2.x will be made from master. The purpose > of these releases will be to enable users to get access to new features > being developed in Hive and allow developers to get feedback. It is > expected that for a while these releases will not be production ready and > will be clearly so labeled. Some legacy features, such as Hadoop 1 and > MapReduce, will no longer be supported in the master. Any critical bug > fixes (security, incorrect results, crashes) fixed in master will also be > ported to branch-1 for at least a year. This time period may be extended in > the future based on the stability and adoption of 2.x releases. > > Based on Hive's bylaws this release plan vote will be open for 3 days and > all active committers have binding votes. > > Here's my +1. > > Alan.
