An IGP like OSPF or EIGRP answers one question inside a single administrative domain: what is the shortest path to this prefix? BGP answers a different question between domains: whose routes am I willing to accept, and whose am I willing to pass on? It carries reachability plus the policy that decides what to do with it, which is why it has no metric in the IGP sense and why it converges deliberately rather than quickly.

A BGP speaker belongs to an Autonomous System — a set of routers under one administrative policy, identified by an AS number. The AS is the unit BGP reasons about. Everything BGP does at the edge is ultimately "which AS did this come from, and what is my policy toward that AS?"

BGP is a path-vector protocol. Every route carries the list of Autonomous Systems it has traversed — the AS_PATH. That list is both the loop-prevention mechanism (an eBGP speaker rejects a route that already contains its own AS) and the raw material for policy (shorter paths are preferred, and the path can be lengthened on purpose to steer traffic). Distance vector trusts a neighbour's metric; link state shares a full map; path vector shares the whole journey and lets policy decide.