You probably figured out what the IIDs are: Instead of having a transaction specified as 343207abc972491:0 (first output of some random tx), you would specify in terms of the transaction that created the input used in 343207abc972491. Or the previous transaction, or its previous transaction, and so on.
Four different L2 protocols are proposed in that paper, but they all require the concept of IIDs to be implemented in Bitcoin protocol. Three of them create multi-party channels (2Stage is more like vanilla LN), and the concept is strikingly similar to the Settlements idea I formulated in the other thread.
Except one would assume there is a good reason for wanting to change the Bitcoin protocol.
Does your idea require similar changes, or do you claim to be able to do the same thing but without the protocol changes?
Do you think IIDs are necessary for scaling - why or why not?
I can see how IIDs work, but I am unsure about the implications for scaling.