Positron

» see github.com/mozilla/positron as it lived, on the wayback machine

~ the goodbye ~

By 2016 the question in front of Mozilla was awkward. Developers wanting to ship a desktop application had settled on Electron, which meant every one of those applications carried its own copy of Chromium and its own copy of Node. Mozilla had a browser engine of its own, had been letting people build desktop apps on it since XULRunner, and was watching all of that traffic go to a competitor's engine.

Positron was the answer, announced on 22 April 2016. The idea was not to compete with Electron but to be compatible with it: import the Electron and Node modules, shim the native bindings, base the webview element on an iframe with mozbrowser, and have an existing Electron app start up on Gecko without its author rewriting anything. Early progress was genuinely good. The post announcing the project could already report the Electron Quick Start app running on the Positron trunk, with BrowserWindow support and the developer tools working.

It came in two halves that were built in parallel. SpiderNode handled the bottom, shimming enough of the V8 API that Node's C++ would compile against SpiderMonkey. Positron handled the top, the Electron surface itself. The plan was to merge them into one runtime once both stood up on their own.

Neither half got there. Myk Melez posted on 8 March 2017 that development was stopping, and his reasoning was the honest kind. Tofino, the browser-design experiment Positron had been built to support, was over, so the thing had no consumer. And full Electron compatibility was a bad target regardless: the API is large, it changes constantly, it assumes Node is wired into the same process, and Firefox's own architecture had moved somewhere else entirely.

~ the chronicle ~

~ the customer that disappeared ~

Tofino started in 2016 as a three-month look at what a browser designed that year would be. The team pulled at everything from the way you navigate to the way a browser front end gets built in the first place, and it was never going to ship as a product. By late summer it had been renamed the Browser Futures Group and redirected at underlying browser technology.

Positron had been scoped against Tofino's needs, so when Tofino's original form ended, the runtime was left holding a compatibility target that nobody inside the building needed anymore. Melez did not pretend otherwise in the post. He wrote that Tofino was dead and moved directly to what he wanted to do instead.

~ what he did next ~

The same post sketches the replacement idea: an experimental command-line tool that would take an existing Firefox distribution and reuse it as a runtime, so installing and packaging a desktop app got simpler without anyone having to reimplement Electron. That work later surfaced as Mozilla's web-app runtime experiments, and the underlying thought, that Firefox on disk is already a perfectly good application platform if you can get at it, is one Mozilla has restated in one form or another ever since.

I write software for a living and I have shipped things on Electron, so the trade Positron was trying to avoid is one I have paid. A small utility arrives as a 150-megabyte download because it brought a browser with it. Mozilla's version of the fix was to make the second engine unnecessary, and the fix died because the one team that wanted it stopped existing. The repositories are still on GitHub. SpiderNode stayed open another nineteen months before it was archived; Positron just stopped.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »