Download websites that use CouchDB

About CouchDB

A database engine has no obvious business appearing in a survey of websites, and the reason this one does is the same reason the entry is worth reading. CouchDB answers HTTP directly, so its interface can be reached on a hostname the way a web server can, and a crawler that asks gets a reply naming the software and its release.

Detection is therefore a statement about exposure rather than about adoption. The great majority of installations sit on private networks, behind firewalls, or bound to a loopback interface where nothing outside the host can reach them, and none of those will ever enter our index. As of October 2026 we record 16 hosts. A figure of that kind measures how often the interface is reachable from the public internet, which is a different and far smaller quantity than how often the software is used, and no reader should treat it as an install base.

Reachability is exactly what makes the list interesting. An exposed database endpoint is a finding in its own right for a security researcher, and the version attached to it turns that finding into a specific question about patch level and configuration. We record 3.5.1 as the version string and 8 distinct versions across the population, a wide spread for a group of hosts this small, which points at long lived deployments that were configured once and then left alone. Older releases inside a spread like that are the rows worth checking against published advisories first.

The rest of the record explains why nobody has a better list. CouchDB is published by Apache Software Foundation under Apache 2.0 and has existed since 2005, written in Erlang, a language chosen for the concurrency and fault tolerance properties a replicating database needs. Nothing in that model produces a customer register. No licences are sold, no seats are counted, and no vendor sales pipeline exists to leak numbers into a press release, so an observed population drawn from a crawl is the only market map anybody can build, ours included, and it is admittedly a partial one.

Three groups get real use out of the file. Security teams and researchers want the reachable hosts with versions attached so they can prioritise disclosure work by risk rather than alphabetically. Anyone sizing the document database segment, or the ecosystem around the implementation language, wants a floor figure they can defend in front of a sceptical audience, and a crawl derived count is at least measurable in a way that download totals are not. Competing vendors and consultancies want the hosts as prospects, on the reasonable theory that an endpoint left open to the internet belongs to a team that would benefit from help. A per host check confirms one machine somebody already suspected. The file gives the population, and the population is what every one of those questions actually requires.