====== 1. Overview ======

===== 1.1 Purpose =====

This document defines the high-level operating model of MTRP.

===== 1.2 System Model =====

An MTRP node MAY register one or more transports. Each registered transport is treated as a path usable by the routing layer.

MTRP separates:

  * direct-neighbor knowledge
  * learned forwarding knowledge
  * send-time forwarding preference

===== 1.3 Core Terms =====

==== 1.3.1 Path ====

A ''path'' is a registered transport instance exposed to the routing layer.

==== 1.3.2 Neighbor ====

A ''neighbor'' is a node reachable directly on a local path.

==== 1.3.3 Learned Route ====

A ''learned route'' is a route candidate derived from an advertisement received from another node.

==== 1.3.4 Preferred Route ====

A ''preferred route'' is a forwarding decision selected by policy during send-time route resolution.

===== 1.4 Architectural Separation =====

The forwarding table represents what a node knows.

The preferred route table represents what a node has selected to use.

This distinction is intentional and MUST be preserved when interpreting diagnostics.

===== 1.5 Current Boundaries =====

The current protocol does not define:

  * end-to-end reliable delivery semantics
  * reliable group delivery semantics
  * dynamic topology-map distribution
  * complete fragmentation and reassembly behavior