Routing — Trivia & Interesting Facts¶
Surprising, historical, and little-known facts about IP routing and routing protocols.
BGP is held together by trust, not technology¶
BGP (Border Gateway Protocol) is the protocol that glues the entire Internet together, and it has essentially no built-in security. Any autonomous system can announce any IP prefix, and its neighbors will believe it by default. Route hijacks — where an AS announces prefixes it doesn't own — have redirected traffic for major organizations including YouTube (2008, Pakistan Telecom), Amazon (2018), and Google (2018). RPKI (Resource Public Key Infrastructure) was designed to fix this but adoption remains partial.
The Pakistan-YouTube incident took YouTube offline for two hours worldwide¶
On February 24, 2008, Pakistan Telecom announced a more-specific BGP route for YouTube's IP space, attempting to block YouTube domestically. Due to a configuration error, the announcement leaked to the global Internet via PCCW (a Hong Kong ISP). Because BGP prefers more-specific routes, the hijacked route propagated worldwide, black-holing YouTube's traffic into Pakistan's network for approximately two hours until the route was withdrawn.
OSPF was designed to replace RIP, which counted to infinity¶
RIP (Routing Information Protocol, RFC 1058, 1988) used hop count as its only metric, maxing out at 15 hops. Networks more than 15 hops away were unreachable. RIP also suffered from "count to infinity" — when a route disappeared, RIP routers would slowly increment the metric toward 15, one update at a time, taking minutes to converge. OSPF (RFC 2328, 1998) replaced RIP with a link-state algorithm that converges in seconds and uses bandwidth-based metrics.
The longest prefix match rule is the most important concept in IP routing¶
When a router has multiple routes matching a destination, it always uses the most specific (longest prefix) match. A /32 beats a /24, a /24 beats a /16, and a /16 beats a /0 (default route). This single rule — longest prefix match — is what makes CIDR work, enables route summarization, allows traffic engineering, and is also what makes BGP hijacking possible (announce a more-specific prefix and traffic follows).
ECMP changed routing from "one best path" to "multiple equal paths"¶
Traditional routing selected a single best path for each destination. Equal-Cost Multi-Path (ECMP) routing allows a router to install multiple routes with the same metric and distribute traffic across them using per-flow hashing. ECMP is the foundation of modern leaf-spine data center networks, where every spine provides an equal-cost path. Without ECMP, half of a leaf-spine fabric's bandwidth would be wasted.
IS-IS survived because it runs directly on Layer 2, not IP¶
IS-IS (Intermediate System to Intermediate System) is an OSI protocol that runs directly on the data link layer — it does not use IP. This was considered a disadvantage in the 1990s (when IP won the protocol wars), but it became an advantage for ISPs: IS-IS can route IPv4, IPv6, and MPLS labels without modification because it doesn't depend on any Layer 3 protocol. Most large ISP backbones run IS-IS rather than OSPF for this reason.
Policy-based routing can override the entire routing table¶
Normal routing decisions are based on the destination IP address. Policy-based routing (PBR) allows matching on source IP, protocol, port, DSCP marking, or even packet size, and then overriding the routing table entirely. PBR is powerful but dangerous — it creates routing behavior that is invisible in the routing table, making troubleshooting a nightmare. The common advice is to use PBR only as a last resort.
VRFs let a single router be multiple routers simultaneously¶
Virtual Routing and Forwarding (VRF) creates multiple independent routing tables on a single physical router. Each VRF is completely isolated — the same IP address can exist in different VRFs without conflict. VRFs are the foundation of MPLS VPN services (RFC 4364) and are increasingly used in data centers for multi-tenancy. The concept is essentially "network namespaces" implemented in router hardware, years before Linux network namespaces existed.
Route redistribution between protocols has caused some of the worst outages in networking¶
Redistributing routes between protocols (e.g., from OSPF into BGP) is necessary in complex networks but extremely dangerous. Without proper filtering, redistribution can create routing loops, inject thousands of unexpected routes, or cause metric-based suboptimal routing. The classic failure mode is mutual redistribution between two protocols, creating a "redistribution loop" that is notoriously difficult to debug and can bring down an entire network.
The Internet default-free zone has over 1 million BGP prefixes¶
The Default-Free Zone (DFZ) is the set of routers that carry the full global BGP routing table with no default route. As of the mid-2020s, the global BGP table contains over 1 million IPv4 prefixes and over 200,000 IPv6 prefixes. Each DFZ router must store all of these routes in memory and evaluate them for every forwarded packet. The steady growth of the routing table drives hardware upgrade cycles at ISPs and IXPs worldwide.