Folks - I feel as though we should uptake the policy that is used for Hadoop common commits where every commit includes a record of the change in CHANGES.txt.
This will make the release process easier since we won't have to backfill all of the changes at release candidate creation time. Also, as we take on more of the policies and procedures of the rest of Hadoop it will make us better committers on other projects when that is appropriate. thoughts? thanks, --larry