If it is just API compat, why not re-add the 0.19 API to 0.21. It hasn't even been a year since the new 0.20 API was released... Maintaining the old API costs nothing in terms of core code (ie: its not preventing us from doing features).
-ryan On Mon, Jan 25, 2010 at 6:40 PM, Andrew Purtell <apurt...@apache.org> wrote: > I have opened a branch for 0.20 (currently 0.20.3 RC2) updated to run on > Hadoop 0.21 > (currently 0.21-dev) It's passing all unit tests now. > > http://svn.apache.org/repos/asf/hadoop/hbase/branches/0.20_on_hadoop-0.21 > > The aim is for client API compatibility. This is an intermediate step for > upgrading > to 0.21 so HDFS improvements, e.g. hflush and we hope also HDFS-630, can be > incorporated immediately through an upgrade of the deployment but without > requiring > any change to users of the HBase client API or the Thrift or Stargate > connectors. > > Unlike the 0.20 on Hadoop 0.18.3 branch, the 0.20 on 0.21 branch has a more > aggressive strategy for back end changes -- things that touch HDFS and/or a > major > performance win may go in, evaluated on a case by case basis, as long as they > do > not change client side API or semantics. For example, the HLog and related > THBase > improvements from HBase trunk have been backported. > > I do not expect to maintain this branch beyond some transition period, maybe > until > 0.21.1. > > - Andy > > > > >
