Download websites that use Debian

About Debian

The country breakdown for Debian does not look like the country breakdown for most web technologies. Germany carries 400,886 of the sites where we identify it, and the explanation has little to do with national preference among developers. A dense population of independent hosting companies and colocation providers, operating on an operating system with long support windows and no license cost, puts that system underneath every customer at once. Geography in this record is a map of where fleets are operated.

Which is the first thing to understand about server software in a technographic dataset. The site owner almost never makes this choice. It is made once, by whoever runs the machines, and then it appears beneath thousands of unrelated domains that have nothing else in common. For a company selling infrastructure tooling, endpoint security or managed migration, the target is the fleet operator rather than the individual site, and the value of the list is that it groups the domains that share an operator.

Version data behaves the same way. We separate 14 named releases, and the names are codenames rather than numbers, which is a reminder that identification comes from a string a machine chose to publish. A server answering with squeeze is naming a release that left support a long time ago, and a fair share of those answers are stale banners rather than accurate descriptions of what is actually running underneath.

All of which leads to the caveat that matters more than any number on this page. Identification depends on a server volunteering the information, through a version suffix in a response header, a default page left in place, or an artifact of the package layout. Operators strip and rewrite those signals routinely, reverse proxies replace them wholesale, and any site behind a content delivery network usually presents the network identity instead of the origin. A large number of installations are therefore invisible to us. The count is a floor, and the bias is not random, because the more carefully a fleet is configured the less likely it is to be counted here at all.

The project itself explains why observation is the only method available. Maintained since 1993 by a volunteer community under Open Source terms, it has no vendor, no license registry and no sales figures. Nobody can sell you an installed base number because nobody collects one. What exists is what machines say about themselves in public, gathered at scale and reported with its limits attached.

A per-domain lookup answers whether one server admits to running it, which is a thin answer for a technology this widely deployed. The downloadable list answers the questions that need a population. Which hosting networks standardize on it and how many domains each of them accounts for. How the release mix ages between snapshots, which is the closest available proxy for patch discipline across an entire provider. Which domains still sit on releases past end of life, in which country, under which suffix. Cross the operating system column with .com and the country field and you have a sized market rather than an anecdote, which is the difference between knowing a technology is common and knowing where it is.