Bump - does anyone have any opinions on this based on the information provided by Sergio?
On Fri, Sep 1, 2017 at 9:07 AM, Sergio Pena <sergio.p...@cloudera.com> wrote: > The Hive community started the discussion to support JDK8 in apr/2016 for > Hive 2.x versions. There was a mix of decisions to whether use JDK8 100% or > just keep compatibility for JDK8. The final decision was to build and test > with JDK8 on Hive 2.0 but keep JDK7 compatibility for one release. The Hive > community had some active maintenance releases as well, so I think one of > the decisions was to keep backports to maintenance releases easy. > https://lists.apache.org/thread.html/83d8235bc9547cc94a0d689580f20d > b4b946876b6d0369e31ea12b51@1460158490@%3Cdev.hive.apache.org%3E > > Later on feb/2017, the community started a vote to start using JDK8 full > features and drop support for JDK7. The vote passed, and we started dropped > JDK7 support on Hive 2.1. However, the community didn't introduce any JDK8 > features yet even they were allowed. > https://lists.apache.org/thread.html/dcd57844ceac7faf8975a00d5b8b18 > 25ab5544d94734734aedc3840e@%3Cdev.hive.apache.org%3E > > Recently in newer Hive 2.x releases and master, the community has started > to use these features. My experience hasn't been bad on backports as not > all developers use these new features (including me). There are other > backports issues regarding new code base for certain features that makes us > hard to backport, but we haven't had any problem with JDK8 at all. > > On Thu, Aug 31, 2017 at 6:19 PM, Alexander Kolbasov <ak...@cloudera.com> > wrote: > > > Sergio, can you share Apache Hive experience with this issue? > > > > Thanks, > > > > - Alex > > > > On Thu, Aug 31, 2017 at 1:10 PM, Sergio Pena <sergio.p...@cloudera.com> > > wrote: > > > > > Hi All, > > > > > > This thread discussion is a follow-up to the old thread related to > > > supporting JDK8 and Datanucleus 4 as the minimum version for Sentry > 2.0. > > > This is dedicated to discuss whether we should allow using JDK8 > specific > > > features (such as lambda functions and other useful API) in Sentry. > > > > > > Advantages are that we could start using these cool features that come > in > > > JDK8 and forget about JDK7 for all. > > > > > > Disadvantages are that doing backports on older releases and/or > allowing > > > other companies backporting fixes from Sentry 2.x will make these > > backports > > > harder because JDK7 is still in use. > > > > > > Maintenance releases are not active but companies are still pretty > active > > > on Sentry. > > > > > > Questions to answer: > > > - What should we do? > > > - If we decide to keep JDK7 compatibility, how long should we keep this > > > until we move completely to JDK8? > > > > > > In my opinion, companies will always be outdated with what we do as an > > > Apache community. Taking a look at what we are doing with SentryHA > > > redesign, this is a breaking change for companies too because fixes on > > this > > > new design may not work for other companies and/or the backports could > be > > > harder. > > > > > > However, the current Sentry 2.0 has been active with JDK7 as the > support, > > > and users active on 2.0 may be using JDK7 environments only. > > > > > > So, > > > Should we wait until Sentry 2.1 or newer releases to allow JDK8 > features? > > > Should we start in Sentry 2.0? > > > > > > Any thoughts? > > > > > > - Sergio > > > > > >
