I've received some confused messages that whatever I was replying to didn't come through, I've reproduced Bob's e-mail below that I was responding to for context:
This, quite simply, is not a "pool". A pool is by definition a tool to reduce
profit variance by miners by collecting "weak blocks" that do not meet the
difficulty target, so as to get a better statistical measure of each miner's
hashrate, which is used to subdivide profits. These are called "shares" and are
entirely absent here.
The only available information here to decide payouts is the blocks themselves,
I do not have any higher statistics measurement to subdivide payments. If I
expect to earn 3 blocks within the window, sometimes I will earn 2 and sometimes
I will earn 4. Whether I keep the entire coinbase in those 2-4 blocks, or I have
100 other miners paying me 1/100 as much 100 times, my payment is the same and
must be proportional to the number of blocks I mine in the window. My variance
is not reduced.
Further, by making miners pay other miners within the window N, this results in
N^2 payments to miners which otherwise would have had N coinbase payments. So,
this is extremely block-space inefficient for no good reason. P2Pool had the
same problem and generated giant coinbases which competed with fee revenue.
"Congestion control" makes this somewhat worse since is it is an absolute
increase in the block space consumed for these N^2 payments.
The only thing this proposal does do is smooth out fee revenue. While hedging on
fee revenue is valuable, this is an extremely complicated and expensive way to
go about it, that simultaneously *reduces* fee revenue due to all the extra
block space used for miner payouts.