We are going to be extremely conservative here, using the criteria that we
use for non-blocker approvals. To get in, a feature will need to be:

 - complete: no follow up fixes or initial implimentations that commit us to
extending the schedule further
 - proven: patches with tests, security reviews and performance impact
analysis, baked on trunk for a good amount of time
 - valuable: we need to understand the benefit of taking the change, and why
it's needed for 3.1 instead of later
 - removable: backout-ready, if anything goes wrong

There are certain things which are tempting targets of opportunity, and I'm
happy to talk about them when people are done their work on the 120 blockers
that remain, but we should not consider this as a re-opening for new
features.

At least, as I understand things at this time :)

cheers,
mike