The problem with BABE and how Lichen solves it: In BABE, you set up a puzzle to be solved later, like "if you can at any time find a valid proof for the statement that this 1BTC belongs to you, you can claim it". The problem is that the amount, 1BTC is fixed right there, at setup, before anything happened. So the only system you can use that for are things like claiming collateral or whatnot. Very limiting. If you made the statement say "I have a valid claim on an amount x" with x being a variable, then they can claim the prize with x=0.01 while the prize is 1BTC; useless.
In Lichen we can exploit the fact that the hub is *required* to answer requests (hub's have to stay live; users don't, until timeout), as follows: first, the statement that BABE's WE works over becomes: "I have an amount x **which the hub signed** and a valid proof that my balance on L2 is > x" where the proof ensures both 'x's are the same. It's bizarre at first until you realize, the hub doesn't sign 'x' to say 'x is a valid withdrawal'; he signs 'x' to commit the user to that choice. So the process goes: user claims 'i want to withdraw x', hub (forced by liveness; loses after timeout) must winternitz-sign x. Then user can use the setup-at-contract-start BABE setup to extract the spending secret which then fires off into the bit decomposition, that actually pays out that same x.
Currently I'm about 80% sure this works in practice, not just theory. And if it does, dispute goes from worst-case 23 transactions onchain to worst-case 4 or 5 and *maybe* 20kb. In either case, it's zero normally. ( Lichen: if you don't know what it is, see the paper in
https://github.com/AdamISZ/lichen, also the explainers under docs/ ).
Right now setting up a shielded transfers L2 using a basic design, just to create a prototype. Already done a fair few mutinynet tests of the existing BitVMX dispute (with its 20+ transactions, a lot more bearable on that chain of course!).
