I rather like this idea, I like that we're taking block scaling back to a technical method rather than political.  BIP100 is frightening to me as it gives a disproportionate amount of power to the miners, who can already control their own blocksize with a soft cap.  It also seems silly to worry about a selfish mining attack if you're going to institute a miner vote that an entity with that much hashrate can noticeably influence anyway.

101 is better but is still attempting to make a guess as to technological progression quite far into the future.  And then when we do finally hit 8GB we will need yet another hard fork if we need to go bigger(or we may need to do it earlier if the increase schedule isn't aggressive enough).  And who knows how large the ecosystem may be at that time, a hard fork may be an undertaking of truly epic proportions due to the sheer number of devices and embedded firmware that operates on the block chain.

I've done no math on this(posting from mobile) but something similar to this would be reasonable, I think.  Unbounded growth, as Adam points out, is also undesirable.

Every 4032 blocks (~4 weeks), the maximum block size will be decreased by 10% *IF* a minimum of 2500 blocks has a size <= 40% of the maximum block size at that time.

This requires a larger threshold to be crossed to move downwards, that way we hopefully aren't oscillating back and forth constantly.  I'll try to do some blockchain research sometime this week and either back my plucked from the air numbers or change them.

Andrew Johnson