I agree on the need, and on a duration at least that long.  These data also need to be organized to follow a consistent management life cycle that is applied by all ARRA participants.  If they are not all following a consistent management life cycle process, then their activities and products will be difficult, if not impossible to analyze and correlate.

These management life cycle process (i.e., methdology) activities also need to read and write their data from a consistent underlying metamodel, or the record of the activities will be fragmented, probably inaccessible, and thus again not be able to be analyzed as a whole and correlated.