March 16, 2026
Peering wars and acronym beefs
How far can you go with IX Route Servers only?
Running the internet on “group chats” alone? Commenters: only 14% reach and regret
TLDR: Relying only on exchange “group chats” (route servers) lets you see lots of networks but actually reach only about 14%. Commenters say it’s fine for pushing content out, bad for pulling traffic in—sparking debates over strategy and even a FAANG vs “MAGNA” name fight.
Imagine running the internet through one big group chat. That’s the vibe of route servers at internet exchanges: plug in once, see lots of networks. The article flexes that these hubs can be even safer than hand‑built connections, with real‑deal checks (think ID validation for routes) and less human oops. Cue the crowd’s collective side‑eye as veterans joked their “artisan, hand‑crafted” configs just got roasted.
Then came the mic drop: building a network on route servers alone looks shiny… until you try to live on it. As pjf put it, even if you join 100 exchanges, you only see a lot of addresses on paper, but can actually reach about 14% of the internet. The commentariat pounced: great for blasting content out, awful for attracting traffic back in. And after the top five exchanges, the returns fizzle—cue memes about “collecting badges won’t save you.”
Drama bonus: a naming brawl. J‑Kuhn floated “MAGNA” (Meta, Amazon, Google, Netflix, Akamai), while others clung to the old finance‑world “FAANG.” Acronym stan wars ensued. Meanwhile, practical voices noted not everyone uses route servers, and policies differ, so you still need those messy one‑on‑one deals. Verdict from the peanut gallery: fun experiment, spicy charts, but RS‑only is a stunt, not a strategy. Check the maps before you burn your bridges—and maybe keep bgp.tools handy.
Key Points
- •IXs connect routers on a shared Layer 2 switch and often host shared services like AS112, NTP, route collectors, and route servers.
- •Route servers redistribute routes as BGP route reflectors, enabling one session to reach many peers while not forwarding traffic themselves.
- •RS implementations commonly exceed the routing security of many bilateral peerings, using IRR filters, RPKI ROV, Peerlock, next-hop checks, and AS_PATH validation.
- •Filtering RS feeds can be extremely large and complex; at DE-CIX Frankfurt, AS-DECIX generates ~2.2M IPv4 and ~885k IPv6 prefixes.
- •RS participation is optional and policies may differ from bilateral peers; sensitive networks (e.g., anycast) may restrict exports. bgp.tools provides visibility into RS peering status.