Download websites that use MochiWeb

About MochiWeb

Nobody chooses an Erlang HTTP library while building a website. MochiWeb is infrastructure that arrives underneath something else, a low-level layer that other software imports in order to answer requests, and the operators of the sites where we record it could often not name it if asked. A detection here says more about a platform a company bought or a system it inherited than about anybody's framework preference.

The record supports that reading. MochiWeb is written in Erlang, released under the MIT license, and dates to 2007, the period when the Erlang runtime was moving out of telecoms and into general server work. A language choice of that kind is never incidental. Erlang gets picked for concurrency and long uptimes, usually by a team building a service rather than a page, so the deployments that surface tend to belong to messaging systems, storage layers and internal APIs that happen to expose an HTTP door to the outside.

Version data is unusually informative for this one. Where a version is published at all we record 1.0, out of 2 distinct strings in total. Almost no spread across an already small population points to a component that is stable and effectively finished rather than one under active churn, which is exactly what you would expect from a library that other projects pull in once and then leave alone for years. A narrow version spread can be read in the other direction as well, as a rough test of whether a host is running something maintained or an image that was built once and never rebuilt.

As of October 2026 we detect MochiWeb on 4 sites, and the caveat matters more here than the number does. Server software of this class is identified from response headers, and response headers are precisely what a reverse proxy or a content delivery network rewrites. Any deployment sitting behind nginx, a load balancer or a caching edge presents that layer's identity instead, so the visible population is a fraction of the real one, and the size of the fraction is not something we can estimate honestly.

The question a list answers, then, is not how much Erlang runs on the web. Nobody can answer that from the outside, and anyone claiming otherwise is guessing. What a dataset gives you is a verified set of hosts that openly admit to running the library, which is useful if you sell tooling, hosting or consulting into that ecosystem, if you are researching where a language genuinely reached production instead of where it was merely discussed, or if you are checking whether a component in your own supply chain turns up on domains you did not expect. Counting across successive exports is the more defensible use of a file like this, because a header that leaks in one export leaks in the next one too, and the direction of change survives an undercount even when the absolute figure does not.