Hacker News Reader

Comment by Animats | Hacker News Reader

> Thanks to HTML5, WebGL, and WebAssembly, browser games are catching up to native ones in ways we couldn’t have imagined just a few years ago. Meta’s Oculus browser already delivers web games that rival native apps in performance. And once WebGPU becomes standard, the differences will be practically invisible.

If only. Have you tried Google Earth on Firefox lately? You can't even zoom in and out any more without waiting many seconds. It's CPU bound, not I/O or graphics bound. Anyone know what went wrong in there? Is it a deliberate attempt by Google to force people to Chrome?

As for games, the browser environment is still rather limited. WebGPU is (mostly) single thread. (You can share memory between processes, which is something else.) It's a subset of Vulkan - one command queue, and no bindless mode. Google proposes to fix that around December 2026.

On the other hand, there is absolutely no reason that Uber or Waymo needs to have an "app".

View on HN

> Is it a deliberate attempt by Google to force people to Chrome?

For the last several months, the YouTube player takes me several seconds to load on Firefox and freezes on me for a couple of seconds every other time I press play/pause...

And changing the user agent fixed it! So I say yes. But I found the extensions to do so quite clunky so I gave up.

> And changing the user agent fixed it!

Send a note about that to the legal team at the Department of Justice managing the Google antitrust lawsuit.[1]

[1] https://apnews.com/article/google-search-antitrust-case-5911...

This is really annoying when it happens. Next time it starts, I'll definitely try the user agent thing. When loading youtube, when this is happening, (and sometimes it doesn't) the whole page locks up for a few seconds half-way through loading/rendering/hydrating. After that, it seems to do a CSS recalc/reflow. While waiting, it's impossible to do anything.

> On the other hand, there is absolutely no reason that Uber or Waymo needs to have an "app".

Here’s one: I don’t give websites my location or the ability to interrupt me with notifications. Ever. I’ve blocked the browser from doing this entirely.

I have Uber and Lyft apps installed and granted those permissions though.

You could give only them location access then they wouldn't be able to see what apps are installed, contacts and family photos or listen on the mic.

You are afraid of a browser so you give system access?

What kind of phone are you running and how old is it? Apps don’t get any of that other information either unless you specifically approve it. If you want to give Uber just notifications and just location, you can do that.

And even then, it only gets location access while it's running (if that's the access that's set). And doesn't get to run in the background either (if you say it's not allowed to).

"When it's running" also means when it runs in the background, and apps can wake themselves up any time they want when you allow them to send you notifications.

Running location triggers an icon in the UI which puts it in a list of apps that have run location recently. An app abusing this in the background will eventually cause the operating system to ask you “hey, are you okay with this?” alongside a map of your current location as your phone sees it.

What you describe hasn't been the case for almost a decade. On iOS and Android apps need to explicitly request and be granted one-by-one permission for all those items.

Which is exactly the problem I'm encountering for web apps--typically the browser is granted all those privileges and then there is a per-site restriction in place. (1) I don't trust this sub-level restriction as much as I do the native per-app restrictions, and (2) it's often more difficult to configure and stay on top of (e.g. iOS will nag me if an app has been using my location in the background, the browser will not).

For these reasons, I restrict my browser to NOT be allowed access to bluetooth, location information, etc. Nobody gets to track me through the browser. If a site I use needs those capabilities, I install their app.

Here’s an example of how large companies abuse their power to push back against these privacy guards.

iOS has the ability to share your photos with an app in a special “picker” where you select the pics you want to share and then they’re sent to the app. The app can’t see the picker so it only has access to the photos you explicitly share.

WhatsApp deliberately ignores this integration path and gives you a choice of either manually choosing more photos to share each time, then selecting them a second time. Or sharing all photos forever, which most users will eventually choose because the friction of the first one is so massive. It’s done cynically and deliberately so they can have access to all your photographs. I shudder to think for what purpose.

Uber is almost the canonical example of something that should be an app! I need to do a frequent task, I punch up the app, I'm always logged in with my location pinpointed, I get a little live-updating widget on my phone with my ride location, plus notifications and Apple Pay and so on. When my ride's over I'm done with the app. Much more useful than say the "Hilton Honors" or "Ann Arbor News" apps which are literally website wrappers that I use once a year.

literally each of those things is also totally possible and easy with a PWA, except maybe the widget.

which approach would use less battery though? rendering a web view is always more expensive than native rendering.

What about older smartphones that may not work well with bloated web frontend frameworks?

I agree that most of it is possible but in this case I agree with parent, uber makes more sense as a native app.

Is there a reason why you think location access is OK for the Uber app but not ok for the uber.com website?

That's not what I was saying. Granting the permission to uber.com requires granting the permission first to the browser, then configuring and trusting the browser's own internal per-site restrictions. I don't trust this as much as the more fine grained and configurable OS-level restrictions.

Uber are the worst for abusing this kind of stuff and for dark UX patterns to get you to subscribe to Uber One. I’m repeatedly asked SEVERAL TIMES EVERY TIME I USE THE APP. and there is no way to pause or opt out from this nagging. They even abuse the push notifications to try and sell me stuff as well.

> I’m repeatedly asked SEVERAL TIMES EVERY TIME I USE THE APP.

You know what doesn't do that? The local cab company when I call them to ask for a pick up.

I feel like I have much more control over "when" I share my location, if permission is granted to "uber app" not "uber domain".

All the reasons I come up with, can be easily argued "you can do the same thing in a browser". But still, it is not the same.

Funny, I blocked notifications from the Uber and Lyft apps because they kept spamming me. And the notifications aren’t all that useful, since I’m actively checking the app whenever something time sensitive is happening anyway.

Location is super useful here, of course.

I was referring to lock-screen notifications (time-sensitive notifications on iOS?), which are quite useful in this context. It was a feature basically designed for apps like Uber, and I do t think it’s available for web apps.

> Anyone know what went wrong in there?

Knowing, I don’t know, could be related though…

Google Earth uses Flutter on web, too, which is in my opinion, the platform where Flutter is giving its users the worst experience.

Though the root cause of the bug could also be non-Flutter related, because in Google Earth’s case, Flutter mainly just wraps the interesting bits (maps stuff in c++???) with a unified UI skin.

And does Google break Firefox all the time on purpose? Good question…

https://medium.com/flutter/extreme-ui-adaptability-in-flutte...

Also, after a decade, browser vendors still don't provide any usable 3D developer tooling better than SpectorJS.

Yeah, your best bet is targeting both desktop and web with something like the rust wgpu library and using renderdoc for debugging/profiling. Which works well until the code stops working in the browser for whatever reason.

>It's a subset of Vulkan - one command queue, and no bindless mode. Google proposes to fix that around December 2026.

Yeah, WebGPU would be a good, modern API... If it was 2015 and not 2025. But now the differences between it and the desktop are widening and at a certain point, people will just break compatibility for the sake of convenience. I mean, theres not even an equivalent of vkCmdFillBuffer ffs! If you want to reset e.g. a indirect dispatch buffer in a command list, tour best bet is to have a buffer that contains uint32_t 1s and do copies.

Yes, Google has adopted lots of dark patterns.

I use YouTube via browser (Safari or Brave) on mobile and it's clear they don't care about the mobile web experience at all - to the point it seems like they are deliberately trying to force you to use the app.

I assume it's because they know the browsers support adblocking and limitations on tracking and notifications.

If you’re in the Apple ecosystem, I highly recommend the Vinegar extension which replaces the awful custom player with a standard HTML5 <video> element

Reply on news.ycombinator.com