Hey Micah, Looks like this hasn't gotten much attention, unfortunately. If Spark wants to support JDK8 for a while longer, since Spark is an important user of Arrow, we might also want/need to support JDK8, unless they're OK with using an older version of Arrow, or targeting different Arrow versions for different JDK versions? (Not sure if that is customary or reasonable, but as I understand it, Spark doesn't use too much of the Arrow APIs - so it might be workable?) Either way that would need support from maintainers and CI.
-David On Wed, Mar 16, 2022, at 13:02, Micah Kornfield wrote: > I don't think we've ever discussed a formal policy on which JDK versions we > intend to support in Java. > > JDK8 is ending active support this month (but still has premium/security > support available). Spark seems like it will continue to support JDK8 > through its 3.x versions which are still under active development. > > As a data point I think Python generally tries to be compatible with > versions that aren't end of life (and at least some other big projects that > depend on Arrow also follow this policy.). > > To a large extent this will boil down to contributors willing to set up and > maintain the necessary CI infrastructure to ensure that Arrow is working on > all the existing JDKs. > > My opinion is we should at least support the oldest LTS version that has > active support but would like to hear others' thoughts. > > Cheers, > Micah
