Download websites that use MochiKit

About MochiKit

Before promises were part of the language, JavaScript that had to wait for a network response was written as a nest of callbacks. MochiKit is a general purpose library from that era, a heavily documented suite of modules covering iteration, DOM work, formatting, logging and asynchronous calls, built by developers who took what they liked from Python and Objective-C and adapted it to the browser. Its asynchronous module was modelled on Twisted, the Python networking framework, and the deferred objects in it gave callback heavy code a consistent shape.

Nobody adds a library of that vintage to a new project. A page still serving it either has not been rebuilt since it was first built, or carries one subsystem nobody dares touch: an admin screen, a report generator, a form that works and whose author left. As of October 2026 we detect MochiKit on 37 sites, and the interesting property of that set is age rather than size.

Two of the recorded facts explain why the code lingers. The licence is MIT and the distribution was a plain JavaScript file to copy into a project, so nothing ever forced a review. No subscription lapses, no vendor emails a deprecation notice, and nothing breaks on its own schedule. Software survives in rough proportion to how cheap it was to acquire, and a free file acquired by download is as cheap as acquisition gets. A first release in 2005 therefore sets a lower bound on the age of the page around it rather than on the age of the code.

For a researcher, sites still serving the library are a proxy for something harder to measure directly, which is web pages that have fallen out of any maintenance cycle. Security teams building a picture of unpatched frontend code want that population. So do agencies looking for rebuild work, and anyone studying how long web technology actually survives in production rather than in release notes.

The count is a floor, and the shortfall is structural rather than accidental. Detection depends on the library arriving as a recognisable file. Once the same code is minified into a bundle with everything else on the page, the fingerprint disappears, so an old project that was later put through a build step becomes invisible even though the code is still running. Counting what can be proved from the delivered page keeps the number honest, and it also means the real tail of survivors runs longer than the one we can publish.

A per-domain lookup settles a suspicion about one site. A downloadable list shows where the survivors cluster: which sectors and which hosting environments keep them alive, what else those pages load alongside, and how a generation of frontend code actually ends, which is not with a shutdown announcement but with a long thin tail that never quite reaches zero.