Download websites that use Perl

About Perl

Very little of the web's working infrastructure predates Perl, and almost none of what does is still in production. First released in 1987 and distributed under GPL v2, the language turned static pages into applications, running the form handlers, mailers, counters and catalogues of the first commercial web. Most of that generation has been rewritten. A surprising amount of it has not.

As of September 2026 we detect Perl on 210,341 sites, and the number deserves a caveat before it deserves an interpretation. A server side language is only visible when it leaks, through a script suffix in a URL, a header naming the interpreter, a generator string or an error page. Sites that route requests through a modern front end or strip identifying headers vanish from the count entirely, so the figure is a lower bound on a population that is genuinely hard to measure from outside. What does survive detection skews towards the older and less curated end of the web, which happens to be the part worth finding if you are looking for systems due for replacement.

Version data sharpens the picture. We record 72 distinct releases in circulation, and the release we see most often is 5.26.3. A spread that wide across one interpreter is not the signature of active development. The pattern is what you get when a language is embedded in operating system images and left at whatever version the distribution shipped, upgraded only when the underlying host is rebuilt from scratch. A modal release several years behind current maintenance is exactly what that lifecycle produces.

Geography tells a similar story. The country column puts Germany at 77,377 sites, a concentration heavy enough to be worth explaining rather than ignoring, and it points at a long tail of shared hosting and small business sites that were built once and then simply kept running. The suffix column is concentrated in .com, which suggests the population is not confined to one national hosting culture even where one country supplies a lot of it.

The commercial reading is straightforward. A domain still running the interpreter today is rarely a greenfield project. More often the site is maintained legacy work, an internal system exposed to the public, a booking or membership application written for a business that still depends on it, a university department page, a hosting control panel. Everyone selling migration services, hosting consolidation, security auditing or replacement software wants that population, and so does anyone researching how long software actually stays in service after the industry has moved on.

A per domain lookup cannot produce any of that. Asking about one address tells you what one company runs, which is useful when you already have a target and useless when the target is the thing you are trying to find. The downloadable extract inverts the question. It hands over the whole detected population with version, country, suffix and co installed software on every row, so the segment can be filtered down to the exact configuration you can sell to or study, rather than sampled one guess at a time.