I think it would be good to have a release -> maven jar mapping for hadoop 1 and hadoop2. That is why releasing two artifacts build against Hadoop 1 and 2 respectively makes sense.
See my poor old issue for background, and rightful arguments from downstream projects: https://issues.apache.org/jira/browse/HBASE-6929 Having hadoop version appended to the release version would be ugly though. Enis On Fri, Mar 22, 2013 at 3:59 PM, Andrew Purtell <[email protected]> wrote: > On 3/22/13, ramkrishna vasudevan <[email protected]> wrote: > > On Fri, Mar 22, 2013 at 4:49 AM, Jimmy Xiang <[email protected]> > wrote: > [...] > >> As to hadoop, how about we ship two: one with hadoop1, the other with > >> hadoop2? > >> > > I agree to this. Actually this will help a lot in case of automatic > > scripts that tries to use the tarball. If not every time the source has > to > > be recompiled with hadoop 2 and then need to create a tar ball and use > it. > > +1 on this. > > +1 here too > > I also like the idea of producing a single -bin tarball that unpacks > into something sane. Spent some time recently hacking on the Bigtop > RPM build for HBase 0.94 to try and deal with the 0.95 layout as is > right now, but had to move on before getting something that works. It > will be much easier for Bigtop, and any user, really, if 0.95 (and > trunk...) looks as much like 0.94 as possible when unpacked from a > dist tarball. >
