Download websites that use Shelf

About Shelf

Shelf is a record to read carefully before you use it. Our catalogue files it under Web server extensions, credits Google, marks the license BSD-3-Clause and dates the technology from 2015, while the reference address stored alongside those fields points at a package registry listing for a library in an entirely different ecosystem. More than one product answers to the name, and the fields disagree with each other. Rather than pick whichever description sounds more plausible, the useful thing to do is say so and work from what the match itself can support.

So the count is what it is and nothing more. As of October 2026 the pattern matches 181 sites. The figure counts fingerprint matches, not confirmed installations of any named company's product, and anyone who needs a specific vendor's customer base should treat the list as candidates to verify rather than as an answer to hand to a sales team.

Verification is less expensive than it sounds. Pull a sample from the file, fetch the pages and look at what the matched artefact actually is, because a hosted content product and a server side library leave very different traces on a page. One arrives as a hosted script or an admin path, the other as a build artefact or a dependency reference. Separating the two populations by what else the same rows carry works well too, since a site running a knowledge or content platform looks nothing like a site whose stack points at a developer toolchain. An afternoon of that work turns an ambiguous row into two clean segments, and it only has to be done once per refresh.

There is a broader point here about how to use a technology dataset at all. Detection is pattern matching, and a pattern is only ever as specific as the name and the artefacts behind it, so the honest way to handle an ambiguous row is to flag it and plan a check rather than inherit somebody else's confident description of what the software does. The file is worth having because it is the cheap way to get a population of candidate domains in the first place, and no other method puts that many starting points in front of you for the effort. What the file is not is a substitute for opening a handful of them and looking.