grin/doc/pruning.md

# Pruning Blockchain Data

*Read this in other languages: [Korean](pruning_KR.md).*

One of the principal attractions of MimbleWimble is its theoretical space
efficiency. Indeed, a trusted or pre-validated full blockchain state only
requires unspent transaction outputs, which could be tiny.

The grin blockchain includes the following types of data (we assume prior
understanding of the MimbleWimble protocol):

1. Transaction outputs, which include for each output:
    1. A Pedersen commitment (33 bytes).
    2. A range proof (over 5KB at this time).
2. Transaction inputs, which are just output references (32 bytes).
3. Transaction "proofs", which include for each transaction:
    1. The excess commitment sum for the transaction (33 bytes).
    2. A signature generated with the excess (71 bytes average).
4. A block header includes Merkle trees and proof of work (about 250 bytes).

Assuming a blockchain of a million blocks, 10 million transactions (2 inputs, 2.5
outputs average) and 100,000 unspent outputs, we get the following approximate
sizes with a full chain (no pruning, no cut-through):

* 128GB of transaction data (inputs and outputs).
* 1 GB of transaction proof data.
* 250MB of block headers.
* Total chain size around 130GB.
* Total chain size, after cut-through (but incl. headers) of 1.8GB.
* UTXO size of 520MB.
* Total chain size, without range proofs of 4GB.
* UTXO size, without range proofs of 3.3MB.

We note that out of all that data, once the chain has been fully validated, only
the set of UTXO commitments is strictly required for a node to function.

There may be several contexts in which data can be pruned:

* A fully validating node may get rid of some data it has already validated to
  free space.
* A partially validating node (similar to SPV) may not be interested in either
  receiving or keeping all the data.
* When a new node joins the network, it may temporarily behave as a partially
  validating node to make it available for use faster, even if it ultimately becomes
  a fully validating node.

## Validation of Fully Pruned State

Pruning needs to remove as much data as possible while keeping all the
guarantees of a full MimbleWimble-style validation. This is necessary to keep
a pruning node state's sane, but also on first fast sync, where only the
minimum amount of data is sent to a new node.

The full validation of the chain state requires that:

* All kernel signatures verify against their public keys.
* The sum of all UTXO commitments, minus the supply is a valid public key (can
  be used to sign the empty string).
* The sum of all kernel pubkeys equals the sum of all UTXO commitments, minus
  the supply.
* The root hashes of the UTXO PMMR, the range proofs PMMR and the kernels MMR
  match a block header with a valid Proof of Work chain.
* All range proofs are valid.

In addition, while not necessary to validate the full chain state, to be able
to accept and validate new blocks additional data is required:

* The output features, making the full output data necessary for all UTXOs.

At minimum, this requires the following data:

* The block headers chain.
* All kernels, in order of inclusion in the chain. This also allows the
  reconstruction of the kernel MMR.
* All unspent outputs.
* The UTXO MMR and the range proof MMR (to learn the hashes of pruned data).

Note that further pruning could be obtained by requiring the validation of
only a subset of the range proofs, chosen randomly by the validating node.