Download websites that use MobX

About MobX

The geography attached to MobX is the part of this page worth reading twice. A concentration that extreme is not really a statement about where developers use the library. The concentration is a statement about how a narrow population gets sampled.

State management code is invisible to a crawler unless something leaks. A source map left in production, an unminified vendor chunk, a bundle built without name mangling, a framework devtools hook exposed in a release build: each of those makes a library legible from outside, and none of them is evenly distributed. One hosting platform, one agency boilerplate or one popular starter template can put an identifiable build on thousands of domains in a single market. The country figure should be read as a property of our detection method rather than as a property of the ecosystem.

The library itself is a JavaScript libraries written in TypeScript and released under the MIT license, in circulation since 2015. The approach is reactive rather than functional. State lives in observable objects, and the parts of an interface that read a value re-render when the value changes, with no explicit dispatch through a reducer and very little ceremony. Teams that choose it are usually building an application with a lot of interconnected state, often one with a long expected life, since the object oriented style reads naturally to developers arriving from other ecosystems and to teams working in a typed codebase.

Version data supports the same cautious reading as the geography. We record 3 distinct releases in production, with 5.15.7 appearing more often than the rest, and a heavy concentration on one older release is exactly what bundled code looks like. A front-end dependency does not update itself. The version a crawler observes is the version that was current on the day somebody last ran a build, which makes it a fair estimate of when an application was last deployed and a poor estimate of what the team develops against today.

For a researcher, the useful question is never whether a single domain uses the library, since one lookup settles that in a moment. The question a full export answers is structural. How many live deployments exist beyond the download statistics, how they cluster by market and by domain suffix, which toolchains leave the library visible, and how the population drifts between releases over time are all questions that need the whole set at once. For a vendor selling error tracking, performance monitoring, testing tools or contract engineering, a list of sites that leak a recognisable front-end bundle is also a list of teams shipping real applications rather than brochures, which is a stronger qualifier than industry or company size.

The honest limit deserves restating at the end rather than buried. Detection is inferential and biased toward builds that were never fully minified, so the real installed base is larger than the figure shown here, and its geography is almost certainly wider than the country breakdown suggests.