A blockchain is a distributed ledger that is not partitioned: one globally shared record, one authority calculus. Accounting was always the other kind: a partitioned distributed ledger, held pairwise, where partitioning partitions authority.
The usual opposition, blockchain vs "centralized" ledgers, misses the distinction that matters. Double-entry accounting has been a distributed system for centuries: the books are held pairwise between counterparties (nostro/vostro), consistency is created bilaterally (every entry carries its counter-entry), and the whole coheres without any global record existing anywhere. What distinguishes blockchain is not distribution; it is that its distribution is global: every node holds the same record, validity is decided by one consensus, meaning is defined by one grammar. Sharded and modular systems do not change this: shards synthesize back into a singular entity, one state, one authority model.
Partitioning changes what can exist. When the ledger is held in partitions, authority is partitioned too: different partitions can run different value logics, evolve at different paces, in different places; they tolerate partition by construction; and they cohere not through global state but through transitive querying and communication across ledger boundaries. A partitioned system can host plural semantics, many definitions of value, interoperable, where a globally-held system can host exactly one.
Blockchain's choice was not a mistake; it was a price. Distributed semantic consensus was computationally unaffordable, so integrity was bought the brute-force way: make everyone hold the same record. The proposal that follows from lifting that constraint is the distributed partitioned ledger: mirror accounting plus multilateral, stake-incentivized verification: authority boundaries and transitive communication, instead of global state preservation. It builds on what blockchains made possible, and opens what they could not: the semantics.
Related: · · · ·