Usenet Peering: How Servers Exchange Articles

When you post an article to a Usenet server, that article needs to reach every other server on the network. The mechanism for this is called peering. Two Usenet servers establish a peering relationship and exchange articles with each other. Each server feeds new articles to its peers, and receives new articles from them. This is how a single post propagates to thousands of servers worldwide within minutes.

Peering is the backbone of how Usenet works as a decentralized system. There is no central server that holds all articles. Instead, every server has its own copy, kept in sync through peering relationships. The history of Usenet goes back to 1979 when this model was first designed.

IHAVE / CHECK / TAKETHIS

The NNTP protocol defines three commands for inter-server article exchange:

IHAVE is the classic method. Server A tells Server B: "I have article with Message-ID X." Server B either responds "send it" (if it does not already have it) or "I already have it" (if it does). This is a two-step conversation for every article.

CHECK / TAKETHIS is the modern streaming method (RFC 4644). Server A sends CHECK with a Message-ID, and Server B immediately responds with whether it wants the article. Then Server A uses TAKETHIS to send the article data. The streaming approach allows multiple CHECK/TAKETHIS pairs to be in-flight simultaneously, which is much faster on high-latency connections.

Large backbone operators exchange millions of articles per day with each peer. The efficiency of the peering protocol directly affects how quickly new posts propagate across the network.

Tier-1 peers

The largest Usenet operators peer directly with each other. These are the backbone providers: the servers that ingest the most content and have the most peering relationships. When an article is posted to a Tier-1 server, it reaches other Tier-1 servers within seconds through direct peering.

Downstream feeds

Smaller providers that do not have their own backbone receive articles through a feed from a larger provider. They are not peers in the traditional sense; they are customers of a backbone operator. Their article inventory is whatever the upstream provider gives them, minus whatever the upstream provider has already removed.

This is the key distinction between independent providers and resellers. An independent provider peers with other backbones directly. A reseller gets a downstream feed from one backbone and has no control over what it contains.

Article availability: A provider with more peering relationships ingests more articles from more sources. Articles that fail to propagate through one peering path may arrive through another. More peers means a more complete article inventory.

Propagation speed: A well-peered provider receives new articles faster. If you are using automation tools, this means new content is available on your provider sooner after it is posted.

Completion rate: Completion rate is partly a function of peering quality. A provider that ingests articles reliably from multiple peers will have fewer missing segments. NewsDemon maintains a 99%+ completion rate in part because of our direct peering relationships with other Tier-1 operators.

NewsDemon operates as a Tier-1 backbone with direct peering relationships to other major Usenet backbones. We ingest approximately 500TB of new articles per day across all peering sources. Our three server regions (US East, US West, EU Netherlands) each maintain full copies of our spool, so peering latency does not affect article availability by region.