3 ms·

I haven't ever seen any test where Chrome was more energy efficient than WebKit. Usually difference is so stark it is 2x. Performance wise, Chrome/Blink had a

I haven't ever seen any test where Chrome was more energy efficient than WebKit. Usually difference is so stark it is 2x. Performance wise, Chrome/Blink had a brief lead in March but since WebKit has again surpassed it. Currently WebKit is about 10% faster on macOS. Security wise - when is the last time you had to update your iPhone/iPad because of a security issue in WebKit? Even on desktop, check the rate of Chromium security vulnerabilities vs WebKit. > I think we're trying a little too actively to promote a sense of significant diversity here, where there is in fact a harsh & brutal monoculture. As someone building a macOS/iOS browser I can say that WebView != Browser. There is so much to be built around a rendering engine for something to be considered a browser today. > Safari just doesn't have modern web features. Web bundles? Nope. Early-hints? Modern web optimizations just aren't possible. This is not always true. https://wpt.fyi/interop-2022 https://wpt.fyi/interop-2022 Yes I am cherry picking, but so were you ;) Also consider that most users these days do not care about web bundles or early-hints. They care about speed, security, battery life and privacy. I feel as WebKit delivers on these essentials and is a good base for a browser. Having Blink/Gecko on iOS may bring a web feature or two but is also sure to compromise on other things.

rektide 4y ago
> Performance wise, Chrome/Blink had a brief lead in March but since WebKit has again surpassed it. Currently WebKit is about 10% faster on macOS. I've also heard no such claims, seen no evidence Safari has found new vast wins to "beat" chrome on, whatever that means. I think it's also absurd that we accept such primitive rankings in the first place, although I myself have shared some of these reductionist claims. Fast at what? There's so many different conditions, so many things we could measure. Is Safari the fastest WebGPU engine? Are they best time-to-first-paint? For what kind of site complexities? Is there JS or just pure HTML? Apple themselves invented Speedometer- an interesting & creative test of loading a lot of different TodoMVC apps, one after another- and for a while dominated it, but Chrome continues to excel[1] (day old article). What data were you looking at, if any, or just heresay? One category I'd point out: I haven't heard a single web-game developer ever thank Safari for anything. These people deserve good treatment, are doing great work. Safari isn't helping them, isn't fast there. (Also props to Chrome for optimizing on a brand new platform that they have had far far less time to optimize for.) > As someone building a macOS/iOS browser I can say that WebView != Browser. There is so much to be built around a rendering engine for something to be considered a browser today. As a web site developer, it all looks like the same shitty, low feature, piece of crap turtle that doesn't implement shit. I can offer my users no more features on your browser than any other turd bucket WebView based browser. It's not your fault, the platform is just deficient, grossly. As for wpt.fyi interop testing, I dig it. There's some very good stuff Safari is up to, thank you for sharing. Better colors is excellent. I would note however: > These scores represent how browser engines are doing in 15 focus areas and 3 joint investigation efforts. This seems to be 100% only agreed upon areas, where each browser already has some stake. It's all rendering, not platform features. So we basically ignore the fact that Safari doesn't have web bundles, doesn't have early hints, doesn't have http3, doesn't have gamepads, doesn't have midi. I'm barely getting started. Safari isn't always- but quite quite frequently is- an obstacle to modernity. > Also consider that most users these days do not care about web bundles or early-hints. They care about speed,... These two statements contradict. Webbundles and early-hints are techniques for performance optimization. I'm willing to bet by 2030 Safari is dragged into doing the right thing here, but right now they keep the web slow. [1] https://9to5google.com/2022/06/06/chrome-mac-speedometer/ https://9to5google.com/2022/06/06/chrome-mac-speedometer/
- freediver 4y ago
  > I think it's also absurd that we accept such primitive rankings in the first place, although I myself have shared some of these reductionist claims. Fast at what? There's so many different conditions, so many things we could measure. On macOS there is only one benchmark used by all sides to determine the 'fastest browser' title - Speedometer 2.0 benchmark. It is used in Apple's marketing [1] and in Google marketing [2] Blink had a brief lead in March when it reached 300, WebKit overtook, the just a few days ago (and a day before WWDC) Google claimed new best score (360 - I was not able to confirm [3]). [1] https://www.apple.com/safari/#footnote-1 https://www.apple.com/safari/#footnote-1 [2] https://twitter.com/googlechrome/status/1533524079717892098 https://twitter.com/googlechrome/status/1533524079717892098 [3] https://twitter.com/vladquant/status/1534301507566202880 https://twitter.com/vladquant/status/1534301507566202880
  - rektide 4y ago
    Thanks. FWIW, Speedometer started with & always seemed like an Apple specific test. It greatly favors low optimization paths, because it's just loading app after app after app. It's basically not worth having optimizing compilers, because Speedometer is just testing initial load speed, more or less. That's a valid number, and I love that others (like Chrome) accept the thrown gauntlet as is, withiut protest, & compete. But this is not really indicative of what life will be like after the first 2s. Maps (or any other app) may run way way faster on an engine that focuses more on optimizing compilers. Thanks though, thanks for the links. Im not trying to invalidate this. But I do think it is limited, I do think it's origin & what it tests is worth noting in here. It's only one way of looking at performance.

rektide 4y ago

> Performance wise, Chrome/Blink had a brief lead in March but since WebKit has again surpassed it. Currently WebKit is about 10% faster on macOS. I've also heard no such claims, seen no evidence Safari has found new vast wins to "beat" chrome on, whatever that means. I think it's also absurd that we accept such primitive rankings in the first place, although I myself have shared some of these reductionist claims. Fast at what? There's so many different conditions, so many things we could measure. Is Safari the fastest WebGPU engine? Are they best time-to-first-paint? For what kind of site complexities? Is there JS or just pure HTML? Apple themselves invented Speedometer- an interesting & creative test of loading a lot of different TodoMVC apps, one after another- and for a while dominated it, but Chrome continues to excel[1] (day old article). What data were you looking at, if any, or just heresay? One category I'd point out: I haven't heard a single web-game developer ever thank Safari for anything. These people deserve good treatment, are doing great work. Safari isn't helping them, isn't fast there. (Also props to Chrome for optimizing on a brand new platform that they have had far far less time to optimize for.) > As someone building a macOS/iOS browser I can say that WebView != Browser. There is so much to be built around a rendering engine for something to be considered a browser today. As a web site developer, it all looks like the same shitty, low feature, piece of crap turtle that doesn't implement shit. I can offer my users no more features on your browser than any other turd bucket WebView based browser. It's not your fault, the platform is just deficient, grossly. As for wpt.fyi interop testing, I dig it. There's some very good stuff Safari is up to, thank you for sharing. Better colors is excellent. I would note however: > These scores represent how browser engines are doing in 15 focus areas and 3 joint investigation efforts. This seems to be 100% only agreed upon areas, where each browser already has some stake. It's all rendering, not platform features. So we basically ignore the fact that Safari doesn't have web bundles, doesn't have early hints, doesn't have http3, doesn't have gamepads, doesn't have midi. I'm barely getting started. Safari isn't always- but quite quite frequently is- an obstacle to modernity. > Also consider that most users these days do not care about web bundles or early-hints. They care about speed,... These two statements contradict. Webbundles and early-hints are techniques for performance optimization. I'm willing to bet by 2030 Safari is dragged into doing the right thing here, but right now they keep the web slow. [1] https://9to5google.com/2022/06/06/chrome-mac-speedometer/ https://9to5google.com/2022/06/06/chrome-mac-speedometer/

freediver 4y ago
> I think it's also absurd that we accept such primitive rankings in the first place, although I myself have shared some of these reductionist claims. Fast at what? There's so many different conditions, so many things we could measure. On macOS there is only one benchmark used by all sides to determine the 'fastest browser' title - Speedometer 2.0 benchmark. It is used in Apple's marketing [1] and in Google marketing [2] Blink had a brief lead in March when it reached 300, WebKit overtook, the just a few days ago (and a day before WWDC) Google claimed new best score (360 - I was not able to confirm [3]). [1] https://www.apple.com/safari/#footnote-1 https://www.apple.com/safari/#footnote-1 [2] https://twitter.com/googlechrome/status/1533524079717892098 https://twitter.com/googlechrome/status/1533524079717892098 [3] https://twitter.com/vladquant/status/1534301507566202880 https://twitter.com/vladquant/status/1534301507566202880
- rektide 4y ago
  Thanks. FWIW, Speedometer started with & always seemed like an Apple specific test. It greatly favors low optimization paths, because it's just loading app after app after app. It's basically not worth having optimizing compilers, because Speedometer is just testing initial load speed, more or less. That's a valid number, and I love that others (like Chrome) accept the thrown gauntlet as is, withiut protest, & compete. But this is not really indicative of what life will be like after the first 2s. Maps (or any other app) may run way way faster on an engine that focuses more on optimizing compilers. Thanks though, thanks for the links. Im not trying to invalidate this. But I do think it is limited, I do think it's origin & what it tests is worth noting in here. It's only one way of looking at performance.