(2) is a good addition to Hive, given that we have views now and need to add the support of session-level temp tables.
Question, how are we going to support session level temp tables Files/tables in HDFS are going to be tied into SessionState or a session ID in someway?
(1) From my experience map/reduce does not lend itself well to recursive work. Currently if each recursion was a map/reduce job that would be a major strain on hadoop. (JobHistory would have explosive growth from a few queries) Maybe there is a simple way around this.
However it is possible/practical recursive processing does open up interesting queries on self joined tables.
id name parent
1 adam null
2 caan 1
3 able 1
Select "ALL DESCENDANTS OF ADAM"