Just wondering: 
How do the harvesters plan to deal with the modularization efforts?
There are a few problems that could rise pretty quickly if we don't
coordinate -

* As I discovered, it's not trivial to package modularizing code in a
manner that's understandable/appliable by clean images. If there are no
guideline, we'll get every variation possible.

* Modularization requires refactorings. These are notoriously brittle if
not inserted into the image fast. What do we do? prioritize the
integration of modularizations over fixes/functionality? 

* Real refactorings are hard to do. My modularization of Celeste, for
example, did change/remove some functionality. How are we going to deal
with this? If what's important is to keep every toy working all the
time, we might never get started...

If there are enough people willing to actually be real test pilots, I
think we should declare 3.3a an "almost-anything-flies" zone and make
the release cycle much quicker, even at a cost to stability. Send up any
changeset that allows to unload something more without breaking
everything, and wire things back together later. I think that's what's
usually meant by alpha cycles...

Daniel