10 ms·

Thanks for the article, it provides very good insights. In the article, there are a couple of passes that mention “benchmarks” where Firefox was faster than Ch

Thanks for the article, it provides very good insights. In the article, there are a couple of passes that mention “benchmarks” where Firefox was faster than Chrome. It sounds like those benchmarks, albeit I assume they were correct, were actually providing a distorted reality of what “fast” actually means, because I distinctly remember people (including myself) switching to Chrome in early days because it was simply objectively faster. It sounds like the existence of those benchmarks have had a net negative on Firefox: it provided a false confidence that the browser was “fast”, and possibly delayed the refactorings that eventually managed to really speed up Firefox again. You mention that Chrome used “UX tricks” but I assume what this means is that Chrome developers were actually driven by more correct benchmarks: the ones that actually translate to a better experience for the end user. Since I’m always intrigued by the topic of false benchmarking and how complex it is to do correct benchmarks that don’t fall into traps, I’d love to hear what those benchmarks you mention were measuring, and if you agree that they gave a false assurance to the Firefox team.

Yoric 6y ago
To summarize, we were benchmarking for throughput: how long between the time you click and the time the operation terminates. For a long time, we absolutely crushed Chrome on most of these benchmarks. Firefox started faster, loaded pages faster, switched between tabs faster, JavaScript executed faster, downloads were faster, ... Chrome developers were benchmarking for responsiveness: how long between the time you click and the user interface shows that something is going on. And that's all that end users actually see. Chrome showed its initial window faster (although it took more time until you could actually do anything with it), was slower at loading pages (but stopped its throbber earlier), slower at switching tabs (but the screen changed faster to show you that it was switching tabs), ... Which ones were the "correct benchmarks" remains an open question, but it's clear which ones the users perceive/remember as being faster! Of course, the better architecture that Chrome used meant that they eventually got faster at throughput, too, on computers that had sufficient RAM.

Yoric 6y ago

To summarize, we were benchmarking for throughput: how long between the time you click and the time the operation terminates. For a long time, we absolutely crushed Chrome on most of these benchmarks. Firefox started faster, loaded pages faster, switched between tabs faster, JavaScript executed faster, downloads were faster, ... Chrome developers were benchmarking for responsiveness: how long between the time you click and the user interface shows that something is going on. And that's all that end users actually see. Chrome showed its initial window faster (although it took more time until you could actually do anything with it), was slower at loading pages (but stopped its throbber earlier), slower at switching tabs (but the screen changed faster to show you that it was switching tabs), ... Which ones were the "correct benchmarks" remains an open question, but it's clear which ones the users perceive/remember as being faster! Of course, the better architecture that Chrome used meant that they eventually got faster at throughput, too, on computers that had sufficient RAM.