Which means we should take extra effort and purge the git tree ... and waste a couple of months doing so. I can already see people rioting the moment i announce something like that.

Let me give you an example: when i implemented dvfs for the n810 i went through 3 different implementation. I think if you ask around you'll still find people pissed off by the first 2 incarnations.

All of that gets lost in the tarball.

The implementation was not deemed to be good enough to be merged in the main linux omap tree (which incidentally as of today still lacks dvfs support) and publishing the git tree would have made available the history of trial and errors to whoever wanted to pick up the activity.

But i am not fanatic enough to go through every commit (including others') to make sure they are acceptable.

Either we put out the real thing and behave consistently with what we preach or it's probably better to shut up and stop pretending to be open.