BGP was the protocol that joined the internet’s independently operated networks, called autonomous systems, by letting each announce which address blocks it could reach and letting each choose routes by its own policy rather than by the shortest path. An isolated settlement network does not need it, because its value lies in policy between operators who do not trust each other, and a settlement federation answerable to one council can use static routes or a link-state protocol throughout. Global routing of that kind does not return with a village, because the networks it connected are gone and there is no one to peer with.
Part of Networking
The border protocol of the old internet, what it did, the institutions it depended on, and the narrow cases in which a group of settlements might want something like it.
BGP Chose Routes By Agreement Between Operators, Not By Distance
Routing Basics states the contrast: inside one organisation OSPF picks the shortest path, but BGP is the exterior protocol between organisations, and it decides by policy, so an operator can prefer one provider, avoid a country, or choose a cheaper link over a faster one. Network Types shows why that was needed. The internet was organised in tiers: large backbone networks connected to each other without payment, regional networks paid them for transit, and local providers paid regional ones. Every boundary between two such networks was a commercial and political decision about whose traffic to carry, and BGP was the tool for writing those decisions down.
The unit was the autonomous system: a network under one administration with one routing policy. Each had a number, and the numbers were issued by registries (general knowledge, not from the other articles). Each system announced the address blocks it could reach to its neighbours, who passed them on, and each system accepted or rejected announcements according to its own rules. Everything depended on trust in what neighbours announced and on institutions that handed out blocks and numbers.
A Settlement Network Is One Administration, So Policy Is Not Its Problem
The carrier Internet Infrastructure suggests assigning autonomous system numbers to each settlement and running BGP or OSPF between settlement routers. That is workable. It is also more than a settlement federation needs, and the honest reading of the carrier’s line is “OSPF, or something like it”. The reason is that BGP solves a problem of mutual distrust and conflicting interests among operators, and a council-run federation of settlements does not have a routing policy problem; it has a link problem. What it needs is Static Routing on a tree, or Dynamic Routing with Link State where there is a loop.
There is no one to peer with, either. Peering means exchanging traffic and announcements with another network that exists, runs a router, and agrees to terms. The long-haul cable and carrier equipment that BGP stitched together was industrial plant needing continuous maintenance and the organisations that ran it, and a village that builds a radio link to the next valley has built a link, not joined an internet.
What You Need
- To know your own situation: how many separately governed networks you will join, and whether they trust each other
Method: Decide Whether You Need Anything Like BGP
Step 1: Count the administrations One council: no. Several settlements each with their own technical staff and rules about whose traffic they carry: possibly.
Step 2: Ask whether anyone must refuse to carry traffic If every link may carry every settlement’s traffic, shortest-path routing is enough.
Step 3: If policy is really needed, keep it small Use private autonomous system numbers, which are set aside for use inside a closed network (the range 64512 to 65534 is general knowledge, not from the other articles; check it before use), and write each rule on paper as an agreement between named people.
Step 4: Otherwise stop A written agreement and a link-state protocol cover almost everything.
Worked Example: Policy Against Shortest Path
Three settlements A, B and C. A has two ways to reach B: a slow radio link direct to B (1 hop), or fast cables through C (2 hops). Hop count picks the radio link. If A’s council would rather use the cables and C’s council agrees to carry the traffic, a policy protocol lets A simply prefer the route through C. If C’s council refuses to carry A’s traffic, C just does not announce that route to A, and A keeps the radio link. A shortest-path protocol could only do either by changing link costs, which changes the answer for every router. That is the point of policy: each operator states what it will use and what it will carry without altering what the others see. The cost is that it is easy to misconfigure, and a wrong announcement can pull traffic to the wrong network, since neighbours trust what they hear.
Common Mistakes
- Copying the internet’s structure into a village — autonomous system numbers and border routers add parts that can fail without adding a capability a single administration needs
- Assuming peering means anything now — it presumed other operators that no longer exist
- Announcing blocks you do not own — if a link to another network is ever made, an announcement of someone else’s address block pulls their traffic to you
What’s Next
- Learn next: Routing Tables — how the results of any routing method are stored
- Combine with: Private Ranges — the address space an isolated federation numbers itself from