that said I was wondering if another thing we could do is limiting max tx size. Did you rember that no more than a one or two months ago Peter Todd said that chain someone put an entire book on the blockchain?
maybe not an absolute limit but a progressive fee per kb... just a thougth.
can you explain the technicalities of that attack?
was it only possible b/c of p2sh? also, did it only involve a non standard tx? if so, isn't that only something that a miner could include into a block he self generates? and why haven't we seen widespread usage of that attack if it's so effective that we have to worry about it?
nothing special actually, borrowing from here
http://bitcoin.stackexchange.com/a/25292There is a maximum standard transaction size since Bitcoin 0.8.2 of 100k per transaction.
There are a number of other limits that influence the validation and propagation of a transaction though. Specifically:
A block is limited to 20000 signature verifications.
The block itself can't be larger than 1Mb.
The standard Bitcoin client (Bitcoin Core / bitcoind) will refuse to relay transactions flagged as dust.
Enough fee should be included (0.0001 BTC/kb)
emph on standard is mine. not standard != invalid.
so there's not a direct limit to tx max size. I'm not expert enough to get all the details, though.
if memory serves in ptodd's git hub repo there's a python script that let you leverage those weak limits to fill the block chain with data (a lot of), maybe it even try to minimize the fee amount.
that said what I'm proposing is to increase the fee per KB progressively. e.g. the first 100 KB will cost you X satoshi each, from 100 to 200 X*2 etc etc
you could use whatever's monotonic increasing function you see fit.
but maybe I'm too naive and this is wishful thinking, maybe as soon as the block space will become more scarce the market will take care of the problem automatically
so b/c a book size tx is non-std, the only way to get it into a block is for a miner to self mine it into his own block. and, in today's situation, if a pool miner operator is caught doing that on a regular basis, his constituent hashing miners would leave him en masse.
therefore, this isn't a problem long run.
