Only storing the merkle tree is useful for SPV nodes. The actual full nodes need to store the full chain to verify blocks (or else you could double spend). In m

Only storing the merkle tree is useful for SPV nodes. The actual full nodes need to store the full chain to verify blocks (or else you could double spend). In mimblewibmle, the full nodes has visibility over interactive transactions only while joined of the network. The final artifact is a combined tx that spans the whole block -- ie: you throw away lots of tx detail and use the kernel for validation. The grin wiki explains the basis of MW protocol: https://github.com/mimblewimble/grin/blob/master/doc/intro.md https://github.com/mimblewimble/grin/blob/master/doc/intro.m... The original whitepaper is : https://github.com/mimblewimble/docs/wiki/MimbleWimble-Origin https://github.com/mimblewimble/docs/wiki/MimbleWimble-Origi... Two implementations of this (slightly different) today are: https://grin-tech.org/ https://grin-tech.org/ and https://www.beam.mw/ https://www.beam.mw/