Copy on write virtualization for L2

[euonymos]

Instead of cloning L2 ledger for block creation/validation purposes, "copy on write" optimization looks appealing. The idea is described by @GeorgeFlerovsky here.

Blocks L2 effect, which is now represented as a new set of utxos to be put in the ledger (utxosActive), can be optimized too. Though this can be done separately, it seems reasonable to piggyback it onto the copy-on-write implementation.

[GeorgeFlerovsky]

Blocks L2 effect, which is now represented as a new set of utxos to be put in the ledger (utxosActive), can be optimized too. Though this can be done separately, it seems reasonable to piggyback it onto the copy-on-write implementation.

Yup. Here's the update rule for a block's L2 effect on utxosActiveConfirmed:

utxosActiveConfirmed <-
   (utxosActiveConfirmed ∪ utxosAddedTxs ∪ utxosDeposited) \
   (utxosSpentTxs ∪ utxosWithdrawn)

[GeorgeFlerovsky]

Invariant (since every L2 genesis event comes after its block's eventsValid):

∅ = utxosDeposited ∩ (utxosSpentTxs ∪ utxosWithdrawn)

So we can also write:

utxosActiveConfirmed <- utxosDeposited ∪
   ( (utxosActiveConfirmed ∪ utxosAddedTxs) \
     (utxosSpentTxs ∪ utxosWithdrawn)
   )