Half-baked thoughts exploring a different way of implementing a fully-validating BCH node.
The idea is to shift the storage of full transaction data to wallets, and explore how little data a fully validating node could store. This isn't a problem today (the UTXO set easily fits in the RAM of an inexpensive server-class machine), but might eventually be at very large transaction volumes.
Initial block download is a problem today (it is annoying to have to wait several hours or days to sync up a new node), and this scheme could make it orders of magnitude faster by shifting the time when full transaction data is broadcast from initial block download to new transaction announcement.
