Remove some completed TODOs

TODO.md

@@ -1,36 +1,20 @@
 TODO
 ====
 
-- [PX] Verify each step of process_incoming_mp.py, make sure that if we have
-       any failures, we do not end up publishing partial results
 - [P2] Add "number of records" to exported filenames
 - [P2] Figure out idle-daily de-duplication
 - [P2] Supply the correct Histograms.json spec for each record to the Mapper
-- [P2] MapReduce: make sure to create all the mapper_x_y files (since we might
-       end up with a file whose keys don't hash to all possible reducer
-       buckets).
 - [P2] MapReduce: delete downloaded data files after they have been processed.
 - [P2] Improve speed of the conversion process
-- [P2] Define data access policy
-  -  read access?
-  -  retention period
-  -  investigate [S3's Lifecycle policy][7] (send to glacier, eventually
-     expire)
 - [P3] Have the "process_incoming" job write bad input records back to S3
-- [P3] Create AMI images for bootstrapped server and mapreduce nodes
 - [P3] Stream data from S3 for MapReduce instead of downloading first
-- [P3] Investigate using Amazon ElasticMapReduce for MR jobs (instead of
-       fetching and running locally)
 - [P3] Add timeout/retry around fetching Histograms.json from hg.mozilla.org
 - [P3] Add many tests
 - [P3] Add runtime performance metrics
-- [P3] Add proper [logging][2]
 - [P3] Ensure things are in order to accept Addon Histograms, such as
        from [pdf.js][5]
 - [P4] Change the RevisionCache to fetch the entire history of Histograms.json
        and then convert incoming revisions to times to find the right version
-- [P4] Preopen the mapper input files in the parent process, pass fd's to child
-       process to avoid race condition with the compressor.
 
 [1]: https://github.com/Cue/scales "Scales"
 [2]: http://docs.python.org/2/library/logging.html "Python Logging"