Core to the Topos ecosystem, the Transmission Control Engine (TCE) serves as the foundation for consistent cross-subnet communication. Supported by an unlimitedly scalable network of dedicated nodes that run a fully asynchronous broadcast protocol, the TCE innovates by reaching agreement on certificates without relying on consensus: the TCE is consensusless.
- The TCE is not a blockchain!
- Certificates are not totally ordered!
To ensure a consistent global state with no conflict between certificates, Topos does not involve consensus at the ecosystem level. Consensus enforces total ordering of messages, but causal ordering is sufficient for Topos's goals, i.e., certificates from the same subnet do not commute, while two independent certificates from two different subnets do. Causal ordering makes sure that the protocol delivers certificates such that all cause-effect relations are respected.
Relying on causal—hence partial—ordering rather than total ordering as enforced by consensus algorithms allows the TCE to be built on top of a much weaker primitive: a reliable broadcast protocol.
The Topos Reliable Broadcast (TRB), the broadcast primitive powering the TCE, enforces consistent delivery of certificates without the need for a classic BFT protocol where known participants synchronously exchange messages to decide as a group on some data. This makes the TCE network much more scalable than common consensus-based blockchains that are inherently limited by a quadratic message complexity and as such cannot expand by more than a few hundreds or thousands of consensus participants. Indeed, the TRB's per-node message complexity is logarithmic in the size of the network (the total message complexity is quasilinear), making the TCE network extremely scalable.
Topos's TRB is fully asynchronous, and as such, is not vulnerable to adversaries arbitrarily delaying messages in order to slow down correct participants and thus break the correctness of the protocol. This broadcast primitive allows agreement to be reached without any time bound on message transmission, and doesn't halt to wait for synchronization between participants.
Causal ordering, unlimited scalability, and asynchrony are key differentiators which, all combined, result in a simpler, more efficient, and more robust protocol than total order broadcast- and consensus-based solutions.