Google makes YouTube much slower for Firefox and Edge

[LAwLz]

9 hours ago, Dan Castellaneta said:

I'll just say it: fuck VP9. It compresses slightly better, but at what cost? Not being able to play without dropping frames on something that isn't a good laptop or a decent desktop?

They really should keep the Opus audio, though.

Maybe it's because I am an anime fan, and the fansubbing community have always had the mentality of "if you can't play something, too bad. We will use the best video technologies available", but I am OK with that.

It's roughtly a 40% file reduction. So that "slightly better compression" as you put it will reduce a file from about 1GB to about 600MB, for the same quality.

 

In any case, VP9 is only used for some 720p, 1080p and 4K videos.

If you're going to watch Youtube on your ZX Spectrum then maybe you can't expect to watch 4K 60 FPS videos?

 

1 hour ago, Ryan_Vickers said:

In my testing, Edge can effortlessly play back 4K60 VP9 youtube while on the same machine, Firefox and chrome can't even manage 1080p60 of AVC or VP9, and yes, hardware acceleration is enabled in the settings and all that. 

How did you test it and in which version of Edge?

[Mihle]

11 hours ago, GoodBytes said:

So far Google  doesn't care about Opera or other web browsers as they are not a competitor to them, their market share is too low.

 

Opera has the same backend as Chrome I think. With some differences.

[Dabombinable]

2 hours ago, Zodiark1593 said:

My old laptop (i5-460M) played 1080P VP9 flawlessly with about 50-60% cpu usage. Needless to say though, my battery life suffered grievously. I was not happy with Google then.

 

My Atom tablet is also confined to software decode outside Edge, though the 4 Silvermont cores seem to do a pretty decent job of decoding as well.

 

Didn't know dual PIIIs could decode H.264.

They could-at 1.2GHz they actually come close to the old P4 in performance (with lower power consumption ). As it happens browsers love multiple CPU, as surprisingly do some games from back then (even with custom maps with AI in Halo CE, my Geforce 6200A is actually the performance limiter not the CPU)

[Emberstone]

16 hours ago, GoodBytes said:

Firefox user have an extension called YouTube Classic which they can install: https://addons.mozilla.org/en-GB/firefox/addon/youtube-classic/

This is one of the best extensions I've ever seen, next to my "View Image" extension (which was also downloaded because Google sucks).

 

Google is amazing at breaking things these days.

[LAwLz]

42 minutes ago, Sniperfox47 said:

Where's the outrage about this? Google's dogfooding polymer to show it off on their site. It's slower on the other browsers because it's using a polyfill... That's what happens when you use a polyfill... You get compatibility at the cost of performance...

 

If Google had, from the ground up redesigned YouTube to take advantage of these features sure, but they've been constantly improving Polymer for years, and by the time browsers have proper v1 API support for Shadow DOM Polymer should too, based on what the team has said at past events.

 

The API isn't "old" any more than WebGL 1 is "old", and it's depreciated because the final release version of the API has been finalized...

 

How do you expect Polymer to use a non-polyfill version of an API that nobody has a release of?

I am willing to bet that 9/10 people in this thread doesn't even know the basics of what is going on.

They read that something is faster in Chrome than Firefox/Edge, and then cry that Google is evil.

 

As soon as something gets more technical than "this processor gets a higher score than this one, so it's better" this forum kind of falls apart.

 

 

 

I'm going to try and give some background, as well as explain what is going on for those interested.

 

Background

Many years ago (to be more precise, 7 years ago), Google decided that they wanted web development to be more component based. The primary goal was to make web code more reusable, as well as some other benefits. In order to achieve this, they created what's called "Web Components". Several "web components" are being converted into official W3C standards right now.

 

However, Chrome is basically the only browser which has some level of native web component support today.

 

If it's been 7 years, why isn't it implemented in all browsers?

The short answer is that web components is mostly a push from Google, and other browser developers have disagreed with some aspects of it. If I understand and remember things correctly, Microsoft were mainly busy developing Edge and weren't really that interested. Apple wanted to develop their own features for Safari, Mozilla had some technical interjections, and kind of felt left out and got unenthusiastic...

Disagreements like if the API should be declarative or imperative, and stuff like that.

 

There has been so much disagreement regarding Shadow DOM that there are currently three different versions being developed/used, just because they have basically gone "OK, this is how we want it but let's settle with this for now".

As far as I know, all browser vendors has agreed to aim to make v1 work right now, and then we will see what happens with v2 in the future (standard is being worked on, even though v1 might not be fully done yet).

 

Polyfills and Polymer

What do you do when you want to use a feature not supported in all browsers? You polyfill!

Polyfill is code which enables features in browsers that don't natively support them, usually by using JavaScript. An example of this would be this website, which has a BPG image on it. No browser supports that image format, so it uses a JavaScript called bpgdec-0.9.4.js to do it instead. Your browser most likely supports JavaScript, so it is a reliable, albeit slow, way of making sure everyone can view your website regardless of which standards their browser supports.

 

Polymer is the name of a javascript library which implements (polyfill) web components.

 

YouTube and Polymer

Google has updated the design of Youtube to use shadow DOM. Contrary to what is said in the OP, Youtube does NOT use Shadow DOM v0. Doing shadow DOM in polyfill is possible, but it's complicated (lots of code), it's slow, it can't handle some stuff and so on. That is why it is NOT being used on for example YouTube. What Youtube uses instead is a feature of Polymer called Shady DOM. More info can be found here but the takeaway is that Shady DOM is really fast, even in Firefox.

 

Also worth noting that Google have said that they will switch to Shadow DOM v1 when it is more widely supported.

 

If Shadow DOM isn't the issue, what is?

Since Youtube uses Shady DOM (for Chrome, Firefox and Edge) it can't be the issue. It performs well in all browsers.

However, YouTube also uses a web component called HTML Import, which is currently the suspect for slowdowns in Firefox and Edge.

 

HTML Import is kind of up in the air if it's going to be a standard or not. Mozilla have said that they will not implement it in Firefox (a very hotly debated discussions as you can see in the comments on this blog), and there is currently a thread on Chromiums bug reporter discussing if Chromium should remove the support for it. What is slowing down the decision is the fact that about 2.5% of all web traffic uses HTML Import in some form. 

 

It is not confirmed that HTML Import polyfill is behind the slowdown though.

When things like this happens it is important to remember that Youtube is a massive and complex website, which uses several components which interact with each other in various ways. Troubleshooting something like a performance slowdown is not something you do some afternoon and then push out an update before you leave work for the day.

[Ashiella]

Sucks I use Opera which doesnt seem to have this issue

[Ashleyyyy]

14 hours ago, Mihle said:

One bonus about using Opera is that in situations like this Opera is most likely not affected, but it still isn't chrome.

same can be said for chromium, and any browser based on it including opera. 

[Ashleyyyy]

4 minutes ago, Ashiella said:

Sucks I use Opera which doesnt seem to have this issue

that's because it's based on the chromium browser, an open source browser that google chrome is based on too. 

[Doobeedoo]

Not on Opera it's actually faster

[Sniperfox47]

32 minutes ago, LAwLz said:

IYouTube and Polymer

Google has updated the design of Youtube to use shadow DOM. Contrary to what is said in the OP, Youtube does NOT use Shadow DOM v0. Doing shadow DOM in polyfill is possible, but it's complicated (lots of code), it's slow, it can't handle some stuff and so on. That is why it is NOT being used on for example YouTube. What Youtube uses instead is a feature of Polymer called Shady DOM. More info can be found here but the takeaway is that Shady DOM is really fast, even in Firefox.

 

Also worth noting that Google have said that they will switch to Shadow DOM v1 when it is more widely supported.

Thanks for the corrections you offered to my post, seems I misunderstood some of it but I do want to offer one of my own.

 

The Shady DOM API is compatible with Shadow DOM (just with additional restrictions) and will fall back to using full Shadow DOM where available, so YouTube in fact *is* using Shadow Dom v0 on chrome, because Shady DOM leverages it.

 

From Mozilla's comments it doesn't seem like they haven't got the Polymer update to let Shady DOM leverage Shadow DOM v1 finished yet, but Chrome also has a working implimentation of v1 for when it's finished.

[LAwLz]

51 minutes ago, Sniperfox47 said:

Thanks for the corrections you offered to my post, seems I misunderstood some of it but I do want to offer one of my own.

 

The Shady DOM API is compatible with Shadow DOM (just with additional restrictions) and will fall back to using full Shadow DOM where available, so YouTube in fact *is* using Shadow Dom v0 on chrome, because Shady DOM leverages it.

 

From Mozilla's comments it doesn't seem like they haven't got the Polymer update to let Shady DOM leverage Shadow DOM v1 finished yet, but Chrome also has a working implimentation of v1 for when it's finished.

According to Alex Russell, Shadow DOM is not used anywhere on YouTube, not even in Chrome.

 

 

 

 

[Sniperfox47]

2 hours ago, LAwLz said:

According to Alex Russell, Shadow DOM is not used anywhere on YouTube, not even in Chrome.

 

 

 

 

Not directly, but if he means indirectly that contradicts his earlier tweet about it using Shady DOM:

Since Shady DOM *does* shim to Shadow DOM.

 

Staight from the page he linked:

Quote

It’s important to note that we've made shady DOM and shadow DOM compatible. This means that the shady DOM API can optionally employ real shadow DOM where it's available. Therefore, you can write one code base that works on all platforms, but you enjoy improved performance and robustness on platforms that implement Shadow DOM.

 

[LAwLz]

Just now, Sniperfox47 said:

Not directly, but if he means indirectly that directly contradicts his earlier tweet about it using Shady DOM:

 

Since Shady DOM *does* shim to Shadow DOM.

Staight from the page he linked:

Quote

It’s important to note that we've made shady DOM and shadow DOM compatible. This means that the shady DOM API can optionally employ real shadow DOM where it's available. Therefore, you can write one code base that works on all platforms, but you enjoy improved performance and robustness on platforms that implement Shadow DOM.

 

The keyword there is OPTIONALLY though. I haven't looked at how Polymer works, but to me that sounds like you have to explicitly enable it to check for Shadow DOM compatibility, or else it will just run Shady DOM even in browsers that support Shadow DOM. If that is the case, and that optional feature isn't enabled on Youtube, then it would explain why Chrome runs Shady DOM, just like Firefox and Edge.

I don't know why Google wouldn't enable Shadow DOM in Chrome though, assuming I am correct.

[vanished]

9 hours ago, LAwLz said:

How did you test it and in which version of Edge?

Testing was very simple.  I'd watch a video in Edge, then watch the same video in Firefox and observe the difference.  "Maths Lab" on YouTube is an excellent source of detailed, 4K60 videos with incredibly smooth motion that make it perfect for spotting even the tiniest flaw in playback, but I hardly needed help noticing the difference Latest version of everything.  I expect to get a lot of questions and to meet a lot of resistance with these findings since, I mean, of course that can't be how it is, that's insane.  But yet, as far as I can tell, that's really how it is.  Edge has hardware acceleration (and strangely refuses to use anything else, even if it would be better), and conversely Firefox and Chrome don't have hardware acceleration, even though it would greatly improve the experience on low end hardware.  If you go looking around online, you can find forum posts and videos of many other people having noticed this but sadly, we seem to remain in the vast minority for now.  I guess so many people either just stick to one browser or the other and thus never end up comparing them and just assume they'd all be the same, and/or they have powerful computers that can play anything in any browser so they don't go looking for answers.  My hope is that one day this reaches "general awareness" to the point where there's a story about it like this, and maybe then some progress will be made.

[Sniperfox47]

2 hours ago, LAwLz said:

The keyword there is OPTIONALLY though. I haven't looked at how Polymer works, but to me that sounds like you have to explicitly enable it to check for Shadow DOM compatibility, or else it will just run Shady DOM even in browsers that support Shadow DOM. If that is the case, and that optional feature isn't enabled on Youtube, then it would explain why Chrome runs Shady DOM, just like Firefox and Edge.

I don't know why Google wouldn't enable Shadow DOM in Chrome though, assuming I am correct.

I just went through the webcomponent stuff on github to check.

 

https://github.com/webcomponents/shadydom if you want to check yourself.

 

According to the github page for the recent Polymer 3 version of the component (which does now support ShadowDOM V1 if youtube updates to it) you must set "ShadyDOM = {force: true}" prior to loading ShadyDOM in order to manually disable it.

 

This doesn't necessarily hold true for the Polymer 1 version and that's not available on Github it seems which is disappointing, but if I remember correctly from my playing around when Polymer was first starting to be a thing, I believe the same was true then. You had to manually force it off to avoid it using ShadowDOM if it was available.

 

The Polymer script for the desktop version of youtube is available here, and doesn't as far as I can see have anything to disable Shimming for ShadyDOM: https://s.ytimg.com/yts/jsbin/desktop_polymer-vfl9zSdzL/desktop_polymer.js

 

I may be incorrect here, or missing something, and he has been working on the web components team so he probably knows better. I just think it seems kinda sketch that he went "Look! We don't directly make use of ShadowDOM at all!" when A) the tool he pointed out doesn't search all the subscripts, and B) ShadyDOM leverages ShadowDOM for improved performance and security reasons unless you manually go to the trouble of disabling it.

[Sauron]

Wow. I hadn't paid it any attention but now I can tell it's slower on firefox.

23 hours ago, asus killer said:

If EU was a person it would be laughing at this topic. Vindicated.

This crap is totally unacceptable and Chrome is a piece of crap as IE was back in the day, i hope to never be forced to use it.

Chromium seems just fine to me, is there a specific reason you're comparing it to IE?

 

Other than that I agree, the EU will probably fine them with a vengeance as soon as they catch wind of this.

[asus killer]

20 minutes ago, Sauron said:

 

Chromium seems just fine to me, is there a specific reason you're comparing it to IE?

 

Other than that I agree, the EU will probably fine them with a vengeance as soon as they catch wind of this.

comparing in the sense it's being shoved down our throats. Google has become evil

[Sauron]

18 minutes ago, asus killer said:

comparing in the sense it's being shoved down our throats. Google has become evil

Oh then yes, I see your point. Chromium isn't as bad though, it's FOSS and pretty community driven.

[Sniperfox47]

1 hour ago, Ryan_Vickers said:

Testing was very simple.  I'd watch a video in Edge, then watch the same video in Firefox and observe the difference.  "Maths Lab" on YouTube is an excellent source of detailed, 4K60 videos with incredibly smooth motion that make it perfect for spotting even the tiniest flaw in playback, but I hardly needed help noticing the difference Latest version of everything.  I expect to get a lot of questions and to meet a lot of resistance with these findings since, I mean, of course that can't be how it is, that's insane.  But yet, as far as I can tell, that's really how it is.  Edge has hardware acceleration (and strangely refuses to use anything else, even if it would be better), and conversely Firefox and Chrome don't have hardware acceleration, even though it would greatly improve the experience on low end hardware.  If you go looking around online, you can find forum posts and videos of many other people having noticed this but sadly, we seem to remain in the vast minority for now.  I guess so many people either just stick to one browser or the other and thus never end up comparing them and just assume they'd all be the same, and/or they have powerful computers that can play anything in any browser so they don't go looking for answers.  My hope is that one day this reaches "general awareness" to the point where there's a story about it like this, and maybe then some progress will be made.

That has nothing to do with Edge supporting hardware acceleration and the others not...

 

That has to do with Edge disabling VP9 by default so YouTube uses HEVC for Edge. VP9 has no hardware acceleration on most laptop/desktop machines unless you're using Intel integrated graphics on Kaby Lake or newer. HEVC does with modern Intel, Nvidia and AMD. It's nothing to do with the browser and everything to do with the format.

 

Firefox and Chrome don't support h.265 because it's *not* an open standard and is patent hell. It's not a good technology for wide adoption on the web. VP9 doesn't have as good hardware adoption in the desktop world but switch to Intel integrated or check out the performance of 4k YouTube on a modern mobile device with Chrome or Firefox. You'll notice, boom, hardware acceleration.

[vanished]

15 minutes ago, Sniperfox47 said:

That has to do with Edge disabling VP9 by default so YouTube

Then why is VP9 enabled by default in Edge for me?

 

If anything, it's Firefox that avoids it in favour of AVC, but I can force it to use VP9.  If I do, it's even worse.

[Sniperfox47]

3 minutes ago, Ryan_Vickers said:

Then why is VP9 enabled by default in Edge for me?

Do you have VP9 hardware acceleration, or does your machine think you have VP9 hardware acceleration?

 

I know with my Titan X (Pascal), in Windows my system enables VP9 support despite the card not *actually* supporting VP9 decode, because the drivers are jerks and lie. 

 

Like Chrome's Chrome://GPU says it supports VP9 acceleration up to 4k despite not having any acceleration in reality and Edge enables VP9 which is only supposed to happen if you have hardware acceleration.

[vanished]

1 minute ago, Sniperfox47 said:

Do you have VP9 hardware acceleration, or does your machine think you have VP9 hardware acceleration?

 

I know with my Titan X (Pascal), in Windows my system enables VP9 support despite the card not *actually* supporting VP9 decode, because the drivers are jerks and lie. 

 

Like Chrome's Chrome://GPU says it supports VP9 acceleration up to 4k despite not having any acceleration in reality and Edge enables VP9 which is only supposed to happen if you have hardware acceleration.

I assume so or there's no way this chip could handle that playback.  It's an N3350.  I know for a fact there's differences between these browsers though because they're behaving differently on the same hardware, and by going from powerful but older stuff to newer yet low performance stuff, who does what totally switches.  It's very strange.  If you want to investigate this with me further perhaps we should go to my thread about it rather than taking over this one tho

[Tech_Dreamer]

been using a plugin to force load old YouTube for a while , have no issues comparatively.

[mark_cameron]

On 25/07/2018 at 4:41 PM, GoodBytes said:

Google is at again, YouTube run slower for those using Firefox or Edge. If you are using either web browser, you probably noticed that since the new YouTube redesign, YouTube is exceptionally slow. Well sharing light on the issue, Mozilla's Technical Program Manager, Chris Peterson, said on Twitter that YouTube new website uses Shadow DOM polyfill v0, which is a old, deprecated API, which only Chrome natively supports.

 

In addition, in the case of Firefox, Google implemented a polyfill that makes Firefox confused with the newer v1, making it having trouble to support v0. Chris added in another post on Twitter:

So it looks like it will be very tricky for the Mozilla team to fix this issue.

 

The good news, is that YouTube still serves a version of the site without the implementation of Polyfill for IE11 users, as that old web browser doesn't support any of it. So you can use that, or install an extension to your web browser to by pass this.

 

On Twitter Chris Peterson shares workarounds:

Firefox user have an extension called YouTube Classic which they can install: https://addons.mozilla.org/en-GB/firefox/addon/youtube-classic/

Edge users have the extension called Tampermonkey that they can install and use this script: https://openuserjs.org/scripts/Cpt_mathix/Youtube_-_Restore_Classic

 

Source: https://www.neowin.net/news/mozilla-executive-claims-that-google-has-made-youtube-slower-on-edge-and-firefox

 

 

Google have the been on the attack with its monopoly on services that people want from any competitor to its services. They did everything in their power to not provide Windows Phone its services back in 2013 with Windows Phone 7, which caused the device to fail to capture users. Now they are blocking Amazon Alexa fro using its services, they cut funding to Mozilla via Firefox default search engine which they previously supported. Recently Google was fined by the EU, $5 billion for Android anti-trust suit, and they continue their efforts at crushing the competition by ensuring YouTube runs best on Chrome.

 

 

As I previously said.

 

Google uses its software to damage competitors in other markets.

 

This is another illustration.

 

Anti competitive behaviour.

[Cornontheecob]

sure are a lot of people that come off like Google owes them something. I dont understand all the hate here towards them? Youtube is a free service last time I checked, a free service that actually in a way pays for this forums operation. Likewise Chrome is a free browser free to use by whomever. Last time i checked if you dont like either you are not forced to use them. I understand Google is the 800lb gorilla in the room and ppl tend to disagree/hate whomever is on top regardless but damn lol this is next level shit. Heres an idea, when the creators of firefox rollout their own youtube competitor they are free to follow the same practice just like if anyone here creates their own website/business they are free to do with it as they please within the confines of the law.