@go1111111
If we had a diversity of implementations with none dominant, of course Greg's BIP would be of no concern. The context right now is that the Core devs hold undue sway, which is interfering in the market to a degree we can only hope will be remedied in time.
Like all other natural orders, markets take time to mature. They take time to develop the mechanisms to route around central control. Market communication mechanisms like fork futures are a key missing piece, along with decentralized communication platforms (decentralized reddit, etc.). Although to a large degree already active, the "magic" of markets to achieve the kinds of optimizations I often describe won't be *fully* reliable until Bitcoin and its ecosystem are more mature.
In the meantime, Core has been able to get away with things that from our perspective are obviously suboptimal (holding down the blocksize limit). This further convinces us that the market mechanisms are not fully matured, which of course implies there is still a chance for Core to do major damage.
It is yet possible for Bitcoin to die in the fires of its current antifragility trials. Miners still do optimize for pleasing the Core devs far more than is warranted (for you to say this is only a consequence of their "capabilities and reputation" surprised me, as I thought you agreed that a key factor - if not the main factor - is their riding on the historical goodwill, inertia, and code-complexity-driven network effects of the original repo). Greg is doing his level best to leverage this residual weakness.
As to the claims of "reverse engineering" firmware, that's exactly the kind of semantic dither I expected would be tried. The hardware claims are completely unproven, we have industry people saying it is highly unlikely to be done, and the whole thing is way too conveniently timed to target the next person to oppose Greg. That said, the lie I was referring to was the implication that Bitmain was actually mining using the covert method, and I admit I can't find that implication in Greg's mailing list post - at least not overtly. People have taken it that way, though, and he surely knew that would be the effect, especially given the timing. The result has been that even though Jihan had a plausibly good reason for having AB in the chips but not activated (awaiting patent matters), he has been skewered and accused of something unprovable in perfect timing for Greg. I will retract that it was an out and out lie until further evidence is shown, and rather say it was a very effective underhanded political maneuver in a long train of such tactics.
As for whether it is actually a good idea to prevent covert optimizations, I think the answer is probably no, as it disincentivizes research by businesses and hands governments free rein to monopolize research on an entire class of optimizations that some consider bad.
The idea of PoW is that there is no objective measure other than how fast you can solve the puzzle, so that any would-be "attacker" needs to play by the same rules as everyone else. If you open a hole up where research into certain types of optimizations becomes risky for a profit-driven business, you give governments a leg up. It gets messy with patent issues and such entailing government interference, but in general I doubt the market is going to want to make ad hoc changes of this nature in minor cases like this one. If there are any additional considerations about covertness I missed, I'm open to hearing them.
Finally, as for a hashing optimization incentivizing miners to block protocol changes (such as Segwit), it falls equally on both sides. Sure, if we assume the optimization factor is bigger than the perceived gain to the miner thanks to BTC price growth from the protocol change (if any), that miner has a perverse motive to refuse to support the change even if they judge it to be somewhat beneficial for the network as a whole...but every other miner has the opposite perverse motive: to *support* the change even if they believe it will be somewhat detrimental for the network as a whole. Since it hamstrings their more efficient competitors.
It is therefore equally specious to accuse Bitmain to be "blocking Segwit despite knowing it would be good for Bitcoin" while not also accusing Segwit miners of "pushing Segwit despite knowing it would be bad for Bitcoin."
Additionally, the above reasoning leads to a good litmus test for when the market would likely support special measures to invalidate a hashing optimization. I suggest it will be when all three of the following hold:
1) the hashing advantage of the optimization AND the anticipated price growth from the upgrade are both large [since otherwise it is trivial as either there is little miner incentive to block or little market incentive for the upgrade]
and
2) the miners that have access to the optimization are in the majority or would be in the majority if they used the optimization [since otherwise the majority of hashpower will have all the more incentive to push through the upgrade]
and
3) the percentage hashing speed advantage exceeds the percentage of anticipated price growth from available upgrades that would nullify the advantage (but only if those upgrades are the only way to achieve that - i.e., if LN can be enabled though a different means like extblk, the price difference between a Segwit and extblk future scenario may no longer be large, in which case 2 above no longer holds) [since otherwise even the optimizing miners will be incentivized to support the upgrade]
For example, if we expect Segwit to bring 2x price growth but AB confers only 20-30% advantage the market seems likely to reject the change, as AB-using miners (whether covert or overt) have a great incentive to support Segwit.
If instead Segwit is only expected to offer 10% price growth, or even a price fall, the market will likely again reject the change as trivial or detrimental.
The type of situation where I think the market would likely accept the change is for instance if an upgrade was thought to bring 5x price growth but the hashing advantage of the would-be-nullified optimization was 10x. I kind of doubt such optimizations exist, but if they do, sure the market may want to intervene there.
This now puts in even greater context Greg's underhandedness in bringing this BIP out with this all too convenient timing, for only a minor hashpower difference vs. his self-proclaimed "rocketship to the moon" Segwit proposal. The $100 million per year figure was also highly disingenuous and deliberately misleading in an industry moving as fast as ASIC development does.
