The Origin Story of Butterfloat

The origins of Butterfloat start, among other places, with major performance trouble in an Angular app I was working on. I had come from a background with things like ReactiveUI and Cycle.js. Angular was an "everybody knows it" compromise at the time, where I had been campaigning for React which I was using in other projects at the time. Part of the compromise was that my rich RxJS background gave me an opportunity to try to build "clever things" and solve some hard problems in Angular, such as how you get a complex real-time dashboard with a few CPU-heavy real-time aggregates to perform well in a user's browser.

One of my first big disappointments, from my background was how much Angular failed to live up to RxJS principles despite seeming to encourage RxJS use. It had too many escape hatches and compromises that broke reliability. It tried to be a framework for everyone and ended up a framework for no one. Use too much RxJS and "fine-grained" reactivity and suffer the performance consequences; escape hatch from RxJS too much and too often and feel the reliability consequences of too-heavy Promises and hard to debug timing scenarios.

But the biggest disappointment was finding out Angular was haunted as the production application I was working on started having direct performance fights where Release builds (and not development build) were falling into Zone.js holes and couldn't get out. That lead directly to building the Angular Component framework angular-pharkas to collect some best practices for fine-grained reactivity under Angular and build much more performant Zoneless components. (I'm glad that Angular 20 finally does a lot to go Zoneless, answering the warning I provided many years prior.)

As with many open source projects, the next kick in the pants was getting laid off from my employer at the time (the one with the performance embattled Angular-based live dashboard), I delivered one last angular-pharkas feature, and was happy to get off the Angular maintenance treadmill for at least a few months. I was proud to deliver a feature I considered to be easy that also reflected/worked like an advanced React feature delivered to the small subset of Angular devs that might stumble upon pharkas.

While my day job was searching for my next day job, I got to taking a long look at what the state of my GitHub portfolio was and wasn't happy with how much had succumbed to tech debt or looked as ancient as it felt. There were old tools like Bower and SystemJS in use that made sense at the time but hadn't aged well, if hadn't also just entirely broke in the meantime. Most of it was in Knockout, which I still liked, but had flaws like poor compile-time type checking. I got it into my head that I wanted to rewrite at least a couple of these old projects in something "modern", and a lot of things I had learned from pharkas and thought about if I were doing them from scratch (the running thread of "Project Gawky" thoughts and asides in the blog posts above, for instance) rather quickly spilled out into Butterfloat. My sketch pad for the API of Butterfloat 1.0 (titled "Project Dr. Mario" at that time) was from early-October and my Butterfloat 1.0 announcement was late November. (The name Butterfloat, as that post also details, came from thinking about Knockout while staring too much at the logo of the Louisville Muhammad Ali International Airport. Butterfloat seems a very Louisville name because it seems a lot more obvious as a relative to "Knockout" to technical people that grew up in Louisville, which seems fascinating to me, but that's part of what keeps the name fun.)

I still think I delivered a lot of the advanced features of React, both the runtime and the developer experience, in a cool, slim package with Butterfloat.

Butterfloat 1.5 with Stamps was first officially delivered almost exactly a year later. I had a laptop with me on a long-ish European trip and one of the days a lot of the ideas for Stamps hit me at once and "needed" to be built, and I had working prototype relatively quickly. Then it was a couple months of hobby time to polish it and get it released.