Blockchains and legacy data sources
As an interoperability protocol, Topos bridges data layers, called subnets. The protocol makes near-zero assumption as to the internal structure of subnets: natively built for blockchains made with Subnet DevKits, the ecosystem is designed to seamlessly integrate any type of new and existing blockchains and, to a greater extent, all data systems.
Unified Certification Interface (UCI)
The abstraction over the data layer is made possible by means of a common interface, a shared language spoken by all subnets to interact with each other. The Unified Certification Interface lays the foundation for subnet authentication and state transitions validity, and is at the core of Topos.
Transmission Control Engine (TCE)
The Transmission Control Engine sets the groundwork for Topos cross-subnet communication. Moving away from common consensus-based solutions, the engine employs an asynchronous reliable broadcast primitive to reach agreement on data exchanged between subnets, and is supported by a network of dedicated nodes that can unlimitedly scale.
How does the TCE work?
With quadratic message complexity, traditional BFT algorithms are limited by significantly higher settlement latency as the number of system participants increases, hence validating groups cannot exceed a few hundred participants. Topos's leaderless reliable broadcast primitive has a logarithmic per-node message complexity, and thus allows the TCE network to scale unlimitedly.
Certificates—data structures defined in UCI to support cross-subnet communication—do not commute for a given sending subnet, i.e., they are totally ordered. On the other hand, those of different subnets do commute providing that dependencies are correctly observed. Causal ordering of certificates is what allows Topos to rely on a causal reliable broadcast primitive in place of stronger primitives such as consensus.
Topos's reliable broadcast primitive is fully asynchronous as there is no upper bound on the delivery time of certificates and no synchronized clocks. Its correctness does not depend on any timing assumptions, unlike synchronous systems which are vulnerable to adversaries delaying messages to slow down honest participants. This leads to a considerably more robust protocol.