Module Mavryk_context_disk

Context comes with two variants: Context and Context_binary with different tradeoffs.

Both have different Merkle tree representations (i.e. when presented the same data, they don't produce the same hashes).

lib_context represents directories as a structured tree of inodes, instead of a flat list of files, to get efficient copy-on-write and optimised read patterns.

The context variants differ by the branching factors used for these inode trees:

• Context uses a branching factor of 32;
• Context_binary uses a branching factor of 2.

To represent a large directory, Context uses less but larger inodes than Context_binary.

As persisting inodes on disk have an overhead (i.e. the serialisation of an inode is prefixed by its 32 byte hash), Context is thus optimised for storing a large quantity of data on disk.

On the opposite, as the inodes in Merkle proofs contain the hashes of the shallow siblings, Context_binary is thus optimised for producing smaller Merkle proofs.