Allen, are all of these choices equally acceptable to you? If not, what is your preferred choice?
Some questions about each of these choices:
a. Are you proposing that things that are currently in various contrib directories move to the hadoop-tools module created by this proposal? If so, those sound like good follow-on issues to file.
b. Where specifically in the current tree would you prefer we place the new distcp code (
MAPREDUCE-2765)? In hadoop-mapreduce-project/src/tools, beside the old distcp?
c. Will you propose and drive a new subproject? Would the release cycle be separate from Hadoop?
d. Distcp is not currently in contrib. Would you prefer that the new version be committed in hadoop-mapreduce-project/src/contrib?
e. I think that's what's proposed here. The 'rules and process' are not currently formalized. Rather it was agreed that we should be more cautious than before about accepting things and that we be more aggressive about booting things when they're abandoned and/or stop passing tests. If you'd like to formalize this and build consensus around some rules and process that'd be great. Should we block the commit of
MAPREDUCE-2765 on that?