BIP-300/301. I read through them, and there's some good stuff and some bad stuff. Here's what I got so far, and please do clarify if I got something wrong.

Summary: Seems like only minor changes to Bitcoin code are necessary, and theoretically it enables a lot of innovation, by placing said innovation in a drivechain rather than the mainchain. This also helps to scale, theoretically, because 256 sidechains can carry a lot more transactions than the mainchain.

However:

1) There is a lot of complex "game theory" going on. And because that touches on human behavior, it's hard to predict. Also, because the "game theoretical" aspects cross borders between mainchain and drivechains, and between users, sidechain miners, and mainchain miners, the resulting complexity isn't appealing to me. I'd say it's nearly impossible to predict what will happen once real money is in play.

Example:

The sidechains don't need to do the hashing, but instead buy mainchain block space with the fees they generate on their sidechain ("Blind Merged Mining, BMM"). That fee is paid to the sidechain miner, and the sidechain miner then uses almost all of that fee to buy mainchain block space. That presupposes a reliable exchange value between sidechain coins and sats, and that reliability is only alledged, not proven. How the exchange rate is supposed to be stable across 6 months escapes me, especially given how much can go wrong on an "innovative" sidechain. @bitcoinerrorlog

2) How the sidechain miners are supposed to reach consensus, given that they don't hash themselves but borrow hashpower from the mainchain, isn't described. Why wouldn't a competing miner create a high-fee withdrawal transaction (WT) that favors himself, then have it validated (albeit after 6 months)? How would the mainchain miner even notice that something bad is going on, if he isn't running a node? And why would he care, as long as that WT pays the highest fee?

3) It also isn't described, how a miner would acknowledge a new DC or new WT. Is that an automated process? In that case, my tiny RasPi might want to spam the network with DC creation requests (M1s), and occupy all 256 slots right at the beginning, precluding all DC development. So it can't be automatic, right? But what is the decision basis for a miner acknowledging M1s manually? Phone calls? Twitter? Reddit? Sorry if I'm being dense, but I just don't see it, and would be glad if someone could explain. Same for WTs (see above).

4) The whole concept being rather new, and a *lot* of innovation (experimentation) being unlocked, means that there is potential for a great number of unforeseen things happening. The validity of a WT (and its associated mainchain block) hinges on several non-obvious conditions, and different implementations could easily lead to a fork, if implementations differ across versions or repos. Such a fork could even be willingly caused by an evil sidechain miner inserting a 0-day exploit of the block validity logic.

IMHO, the "innovation on sidechains" argument is technically correct, but the "shitcoins will move to sidechains" narrative is wrong. The reason shitcoins exist is not innovation; that's just how their marketing departments push their scam onto the naïve. The reason for shitcoinery is scamming peopple out of their money, and truth (especially auditable, reliable, consensus truth) is anathema to them.

So, that's what I think of drivechains as of now. Happy to discuss, and very happy to be proven wrong. As long as the proof passes my reality check. Because IFF all those points are invalid or wrong, drivechains might be a Big Thing™.