Firefox 115 Now Available with Intel GPU Video Decoding on Linux | Hacker News Reader

Firefox 115 Now Available with Intel GPU Video Decoding on Linux

phoronix.com

179 pointsBender72 comments

"Certain Firefox users may come across a message in the extensions panel indicating that their add-ons are not allowed on the site currently open. We have introduced a new back-end feature to only allow some extensions monitored by Mozilla to run on specific websites for various reasons, including security concerns."

https://support.mozilla.org/en-US/kb/quarantined-domains

Tin foil hat on

This will be used by Google to force Mozilla into disabling ad blockers on YouTube (dare I say: SponsorBlock as well).

Tin foil hat off

The problem with this feature is the user does not make the choice. Firefox should be alerting the user why they disabled the extensions and give them a choice to enable them.

From the linked page explaining the feature:

How can I re-enable the add-ons that are not allowed on some websites?

We understand that installing add-ons is a user choice and, as with your security, we also take this matter very seriously. If you are aware of the associated risk and still wish to allow the add-ons that have been disallowed on a website by Mozilla, you can do it from the configuration editor (about:config):

"“But the plans were on display…”

“On display? I eventually had to go down to the cellar to find them.”

“That’s the display department.”

“With a flashlight.”

“Ah, well, the lights had probably gone.”

“So had the stairs.”

“But look, you found the notice, didn’t you?”

“Yes,” said Arthur, “yes I did. It was on display in the bottom of a locked filing cabinet stuck in a disused lavatory with a sign on the door saying ‘Beware of the Leopard.” "

Requiring users to have non-trivial general computer skills before disabling security features seems completely fine to me.

If you see that an extension is blocked on a domain and don't think to Google "unblock extension domain Firefox", select an appropriate result, and follow straightforward instructions then you probably aren't capable of understanding the nuances of extension security risks.

These stupid users, can’t control their computers! Let it be us, corporation, who decide what you can run. It is safe and effective.

Apple, is that you?

The blocking UI shown to n the documentation of the feature has a “learn more” link which presumably leads to to exact the same documentation that I quoted. I consider that absolutely discoverable and fine. The one thing that I would wish for as an improvement is that I can unblock specific extensions individually instead of a single switch for all.

This about:config toggle seems to be a "nuclear option" that disables the entire "quarantined domains" functionality. It would be better if Mozilla offered users a choice to toggle only certain blocked addons on certain domains, though the normal UI.

I agree, that would be better. But the current situation is strictly better than before, even if it’s not perfect - it protects you until you flip that switch and when you do, you’re exactly where you were before they added the feature. So nobody is worse off and many are better off.

I don't think this is exactly new. For example the Add-ons website for Firefox already didn't allow any extensions to be run on it. It does seems this feature extends this list for new websites, but it doesn't mean Firefox will abuse (and even if they start abusing it, they have a explicit flag to opt-out).

They will abuse it. Then "apologize". And then abuse again.

https://github.com/gorhill/uBlock/wiki/Privileged-Pages

The screenshot they use for the example specifically has ublock origin as the vetted extension that is permitted to run.

On the page you linked it explains how to disable this completely. No tin foil hat needed.

I searched a bit through the documentation and code and these were my findings. I thought I'd share them for others that are interested and for future reference.

Currently, there are no domains blocked, they would appear on this API endpoint: https://firefox.settings.services.mozilla.com/v1/buckets/mai...

This is the json schema for this API endpoint: https://firefox.settings.services.mozilla.com/v1/buckets/mai...

More information on the remote settings in general: AMRemoteSettings Overview - quarantinedDomains: https://firefox-source-docs.mozilla.org/toolkit/mozapps/exte...

Remote Settings documentation: https://remote-settings.readthedocs.io/en/latest/index.html

Remote Settings Devtools - where you can see all the remote settings, that get set: https://github.com/mozilla-extensions/remote-settings-devtoo...

I truly miss the days when companies weren't able to remotely change any settings at all.

Makes sense though, the amount of malware extensions is... unbelievable.

This appears to have been implemented by the security team, with the bugzilla issue originally made private / Mozilla staff only a month ago.

Mozilla has earned some trust here, I think. I'll be interested to see why this was done, can't imagine that'll be quiet forever.

> Mozilla has earned some trust here, I think.

How? Was it the opt-out telemetry, putting ads in the browser, or the "experiments" that were also ads?

The 25 years of seeking better outcomes for netizens, using their small influence to help people to have a better experience with the internet.

If you always compare things to your version of utopia you'll be very disappointed. I appreciate their efforts in saving me from the alternatives like Chrome and IE, which are produced and pushed by groups with a million times their resources.

I thought Chrome saved us from slow and tarded Mozilla Suite and IE 7 past gen browsers. God bless Apple WebKit I guess. It made web 2.0 possible.

The old Mozilla suite was obsoleted by Firefox much earlier: Firefox 1.0 was in 2004, IE 7 in 2006, and Chrome didn't show up until 2008 when Firefox was on 3.0.

This is the Bugzilla issue:

https://bugzilla.mozilla.org/show_bug.cgi?id=1832791

> Mozilla has earned some trust here, I think.

Mozilla has already burned through a fair amount of the trust they earned, though. The trust that is left is why adverse reactions to this are mostly in the form of suspicion rather than knee-jerk condemnation.

In the past, Mozilla has quarantined several things outside of the reach of standard Firefox customizing tools. Every time it has been because compromising one of those things would result on the user not being able to manage the configurations or even seeing that something different is happening.

IMO, the most likely usage is that they will use it to disallow extensions on the extension store.

I can't imagine uBlock Origin not being on the list of "monitored extensions" that are allowed.

That said, I am concerned that there doesn't seem to be any easily findable information about what domains (or even what kind of domains) will be quarantined, or what extensions will be "monitored".

This thing is like there on the first day firefox have extension. If you look closely, you can't run extension on extension store page since day one. So basically nothing changed. The only difference is they now allow some extension to run on these forbidden domains. And show you clearly where the extension won't run.

I think Google would rather preinstall SponsorBlock in Chrome by default if it could, because this move would encourage use of YouTube Ad tools instead of in-video ads (from which Google gets nothing).

This is awesome, maybe we can get more user feedback for devices on hw video decoding.

HW video decode on Linux (just for browsers) has dragged at a snail's pace for the longest time. Firefox seems to take the lead in this regard but Chromium is just terrible.

Chromium devs create flags that break hw video every few months or so (and when the flags break you are forced to go on a random easter egg hunt to find out how to make hw video work again) and currently has a bug that leads to a memory leak. Sigh.

> Chromium devs create flags that break hw video every few months or so (and when the flags break you are forced to go on a random easter egg hunt to find out how to make hw video work again) and currently has a bug that leads to a memory leak. Sigh.

The Xorg leak was awful, it took me three days to find its cause.

I get this problem on Xwayland which is the only way [0] to enable HW video decode on official Chromium binaries on Wayland ATM. Am on Arch and if the Chrome team doesn't fix the issue in future releases I just might give this workaround [1] a chance.

[0] https://bugs.chromium.org/p/chromium/issues/detail?id=132675...

[1] https://aur.archlinux.org/packages/chromium-wayland-vaapi

This caused me to switch to FF. But, now I have native, well-functioning wayland & vaapi - wish I had switched sooner!

Once upon a time my Asus 1215B sold with Linux had drivers where everything worked, then came WiFi driver rewrite forcing months of LAN cable (but it was FOSS), then came AMD driver rewrite forcing software rendering and lower GL support (but it was FOSS), then came ... but it was FOSS.

Nowadays it is the remaining device that still has GNU/Linux on it, mostly because it is an aging netbook and I can't be bothered to install anything else on it before it dies.

Lets see how well Firefox 115 works on it.

The more things change the more they stay the same. Thanks for this short anecdote.

Wondering what kernel version you run that old device on :)

Nowadays, Ubuntu 20.04.6 LTS.

All those issues were coming up during various updates since 2009.

Official release notes: https://www.mozilla.org/en-US/firefox/115.0/releasenotes/

Also of note is that this is the last major version of Firefox for Windows 7 and Windows 8, as well as macOS 10.12, 10.13, and 10.14.

Thanks, I did not realize that. I have one hackintosh still on 10.13.

I've actually been waiting for this release to switch to ESR. The regular release has been getting updates three or more times a month, each of which prompts daily nags to update, which gets tiresome across three or four computers with a lot of tabs open. I'm hoping for more like one per month on ESR.

For people who haven’t read the linked release notes: Mozilla is not dropping support for Windows 7-8.1 or macOS 10.12-10.14 at this time. Those users will be migrated to Firefox ESR (Extended Support Release, an LTS branch) that will receive monthly security fixes and some feature fixes for at least another 14 months.

> Also of note is that this is the last major version of Firefox for Windows 7 and Windows 8

Note that this also means Windows 2012 and 2012R2.

This sort of improvement is extremely welcome. Good job Mozilla.

[deleted because I don't to be accused for ragebaiting]

Honestly, not that interesting, and I hope this doesn't spin up a ragebait thread that's a distraction from what is objectively some very good news.

It really isn't that interesting. Red Hat engineers do great work and have done so for years. That controversy has little to do with the engineers and everything to do with Red Hat corporate.

One has basically nothing to do with the other, but I'll humor your attempt to draw a connection and say that if I had to choose between Firefox having GPU video decoding and IBM/RH not trying to subvert the GPL, I'd choose the latter and forego video acceleration. This is to say, the former does not make up for the latter even if I humor the existence of a connection.

Besides, I'm still going to play videos in mpv anyway. In-browser players suck for various other reasons besides the acceleration/performance issue. The bare-bones vanilla in-browser player has virtually no features and controls cannot easily be remapped. Firefox's floating video player feature is a nice step in the right direction, but only a tiny step. Furthermore the video controls implemented by websites like youtube are even worse and fixing that on a per-site basis would be a huge chore. I'll give you a specific example: my laptop's universal volume keys are my F1, F2, and F3 keys and they're right next to my 1, 2, and 3 keys. In the control scheme youtube has implemented, if I ever miss my volume keys I jump the video back to 1:00, 2:00 or 3:00 because youtube for whatever reason thinks number keys should be shortcuts to seek to that minute mark. I guess some people probably like this but I certainly do not. I could try to fix this by injecting scripts into the youtube page, but that's a hassle I can simply avoid in the first place by using a video player that has mostly sane defaults and makes configuration straight forward besides. I'd rather configure mpv once then play wack-a-mole fixing numerous websites and keeping those fixes up-to-date and working.

Hi - I said well done Mozilla and they let me know who did the work. I don't see why you're saying anything you said here.

Since the other poster deleted their post, I will hazard a guess as to what they said.

This work was done by Martin Stránský from Red Hat, as with most of the other linux hardware acceleration and Wayland enablement work that has happened over the past several years.

But Red Hat is not very popular right now, so even mentioning that they might be due credit for this work is therefore considered potential "rage bait".

I very much don't want to see the thread devolve into another argument either, but I also hope that can be avoided without completely ignoring such details as the engineer that did this work (and lots of other work) in Firefox.

Disclosure: I work at Red Hat

Very cool.

I am a hard core open source guy and make a living with it. It gives me huge grief to see people complaining about red hat. Red hat employs many many people and contributes a big lot. I personally do not like RHEL but worked in a company that more than could justify the RHEL license and even so went for centos. It disgusted me even further because when there were bugs to the red hat bug trackers they went and found their solutions.

I support open source, I believe it needs to be sustainable financially otherwise it will just be replaced by closed source competitors and the world will be better off.

I don't understand this. I've had h264 decoding with VAAPI since it's been supported in Intel GPU drivers. This is telling me about a new feature I've had for years?

Maybe they flipped the switch om the default setting?

It's been disabled by "policy" so far. You could force enable it but it was a lot of hassle.

Great achievement! How many additional years did it take compared to FF on Windows?

The effort seems to have started 4 years ago but I'm probably wrong: https://bugzilla.mozilla.org/show_bug.cgi?id=1610199

Vulkan video decoding is just around the corner (already working with latest mpv/ffmpeg/mesa from git). I'd imagine it'd be easier to implement/maintain (single API across platforms). I wonder if Mozilla is looking into it yet?

Does that mean all intel GPUs such as integrated GPUs only Intel's discrete GPUs ?

Am I likely to find any kind of improvement on a intel 12th gen laptop ? I guess yes since firefox has a hard time handling 1080p videos on my machine, is this because it all happens on the CPUs ?

A modern CPU should be able to decode 1080p in software without breaking a sweat. I had hardware decoding disabled in Firefox until recently and the difference is unnoticeable to me.

hardware decoding generally improves battery life by a lot. why disable it?

Software decoding is generally more customizable and better quality, also better stability depending on the circumstances.

Adjustability of codec parameters and quality are reasons to prefer software over hardware for encoding, but I've never seen any evidence or explanation for how software decoding can likewise be significantly better—unless you're referring to post-processing that can just as easily be done after hardware decoding.

Encoding yes. Decoding the image is equal, at least with h.264

You probably read this about software encoding and got them mixed up in your head.

I've messed around in the anime fansubbing scene many years ago and still tinker around with codecs and filters since I care about that stuff.

I don't usually use hardware decoders because there are restrictions, such as what renderers can be used or what and where filters can be placed in the decoding and rendering pipeline.

Hardware decoding is great when power (be it electrical or processing) is limited, but if you really want to have all the knobs available software decoding is what you want.

I was hitting a bug that would randomly crash the GPU when hardware decoding VP9. I re-enabled hardware decoding after adding a Firefox extension that let me chose the codec Youtube uses.

Regarding battery life, I really haven't noticed a difference on a 6800U. Both hover around a disappointing 7W.

> A modern CPU should be able to decode 1080p in software without breaking a sweat.

Not true for modern codecs such as AV1

Strange, my fan (12th gen intel) spins up instantly on software, and my battery life takes a hit.

VA-API is available on integrated CPUs as well. You can run "vainfo" to check support on your system.

That said, even my slow Celeron can do 1080p fine, so not sure why your system can't.

I was pretty certain that Firefox did already support VA-API with Intel and I did test that with integrated Intel graphics. Maybe this release is more about enabling it by default, or to also support the discrete GPUs?

It's a pity the changelog does not link a more detailed writeup.

Now looking forward to Firefox adding Vulkan video as a better alternative to VAAPI.

Ok, tried Firefox again: 4k YouTube videos aren't a slide-show on my system anymore (using an old Nvidia card, though)! Yay! But GeoGuessr/Google Maps Street View is still unplayable. Well, still will continue to use a Chromium based browser until that is fixed (probably never will be at this point).

It used to be that Google Earth worked passably well for me in Chromium, and was a slideshow in Firefox. As of a few weeks ago, it's the other way round. No idea what's changed.

On a sidenote, is Google Earth still a maintained product? Seems like it's on life support. Is there any decent alternative?

I don't know, I don't use Google Earth (the Qt desktop application). I only use Google Maps and mostly through GeoTastic.

Reply on news.ycombinator.com