FrogNet in Short
This is the short description of the complete FrogNet/Ribbit inventory. It is meant to tell an engineer what the system contains before they choose which part to read in detail.
Living Network
Self-forming FrogNet nodes discover reachable peers, derive routes and services from observed state, survive partition, continue as independent ponds, and restore connectivity when links return. Discovery is the part of FrogNet that converges.
Discovery and routing
Discovery proves what is reachable; routing is derived from those observations, measured path cost and stability rules rather than from a fixed topology plan.
Transport independence
The same application can operate over Ethernet, Wi-Fi, WireGuard, Internet paths, radio and constrained links without importing bearer-specific logic.
Services and elections
Services publish capability and use service-specific deterministic selectors. Each node applies the service rule to the state it can see; there is no central election coordinator.
Broker and WAN
The broker publishes inter-site channel state and provides WireGuard transit. Nodes pull that state; the broker is a data path, not network authority.
Memory
FrogNet Memory exposes independently addressable current state using service, variable and instance coordinates with write, read, held read, freshness and remove.
Ribbit
Applications are built from truths, references, containers and small participants that derive new truths. Real dependencies use held reads rather than application message choreography.
Regions and Internet RAM
A service can expose an independently governed Region over the Internet with its own namespace, schema, API, authorization, lifetime, residency and backing policy.
Residency and backing
RAM may be transient, persistent or materialized on demand. Re-homing does not imply copying, reconciliation or continuity unless Region/application policy supplies it.
UnREST and BLDC
Existing request/response applications can keep their local contract while FrogNet learns stable representation and moves semantic differences across the wire.
FNWP and semantic transport
The peer-facing wire carries learned semantic operations and compact SAME/DIFF/FULL representations over persistent sessions rather than repeating full application payloads.
Plane
Plane carries large, frequent or perishable current values such as media and tensors. Newer values supersede stale unsent values; the contract is current useful state, not history.
Application RAM
An application RAM server need not be a FrogNet node. It can expose the Memory contract directly; how a client reaches that RAM is the application vendor’s business.
Redis-on-Ribbit
Redis compatibility is used as an adversarial behavioral oracle for strings, containers, blocking, expiry, streams, consumer groups, transactions and related semantics.
Psychedelic FrogTorch
A PyTorch distributed backend re-expresses distributed computation through shared state while keeping PyTorch itself unchanged and measuring against Gloo controls.
Communicator
Presence and control use shared state while bidirectional audio/video uses the high-speed path with priority, segmentation, drop-before-stale behavior and adaptive quality.
Sourri
An ordinary application uses independently addressed users and turns, references, containers, held reads and participants rather than a growing serialized conversation object.
Monitor and Dashboard
Observability is published truth. Monitoring and topology views derive their display from the same state the system is actually using.
Physical control and automation
Sensors publish observations; software and AI derive state; actuators read the truths they depend on. Remote physical action has been demonstrated.
Language and machine independence
Python and C++ implementations interoperate on the Memory and Plane contracts; the architecture is not tied to one language or one machine class.
Qualification and characterization
Qualification asks whether the contract holds. Characterization measures the operating envelope of a particular host, path or backing store. The two are kept separate.
FrogSim and frogbench
The simulator exercises real network behavior against oracles. frogbench qualifies Memory behavior and characterizes the machine before installation or workload planning.
Explicit boundaries
FrogNet does not claim a globally coherent heap, global transactions, automatic cross-Region reconciliation, free contention, free fanout or persistence across re-homing.
This page is intentionally compressed. The detailed pages and evidence ledger expand these descriptions with implementation behavior, measurements and explicit limits.