Google's $124.6m open codec hits Chrome dev build

I'd like to share my experiences with those who haven't experienced WebM on Linux yet:

On Chromium(I'm using build 48000, pre-compiled Chromium-bin on Gentoo) it works flawless. If you join the HTML5 beta at http://www.youtube.com/html5 and you click on a video which is available WebM (normally all those less then a week old, and HD at 720p), you are automatically served the WebM version. It doesn't crash like flash, and has about 20-50% CPU usage on my dual core AMD 4850e, pretty much if you ask me. But unlike Flash, it doesn't peak at 2x100% after ten minutes or so and cause the browser to hang.

WebM in the developer build of Firefox (3.7a4) is a bit worse however, it sometimes 'skips' a few tenths of seconds. Far better than flash however, because Flash always crashes Firefox. Also, scrolling up and down while the vid is playing works much better than with Flash. It seems load times are also a bit quicker.

One thing I haven't succeeded in yet, is to download the h.264 and WebM version of the same vid in the same resolution to compare file sizes. It seems Chromium video-cache is protected somehow, and I can't find it using strace. Does somebody have experience with this?

Quoting:because Flash always crashes Firefox

Rubbish. I watch flash for hours at a time in Firefox, with no crashes or slowdowns

The fact it works for you, doesn't make it work better for me, does it? Hence, my experiences above.

I already started with a fresh FF-profile (previous one was about five years old), problem persisted. Disabled all add-ons, problem persisted. Tried the Beta of the flash player plugin, problem persisted. Tried Chrome and Opera (both withoud add-ons) with the same Flash, problem persisted. Tried both symlinked and 'file in plugin map', problem persisted. Tried different resolutions, problem persisted. Tried the plugin from my distro and direct from Adobe website, problem persisted. Disabled Flash, all problems gone.

Hence I don't give a **** about the quality of Flash with h.264 vs. WebM: Anything is better than Flash currently. Browse the internet, and you'll see I'm not the only one having problems.

@Hans: Try throwing away your ~/.abobe folder. Flash has it's own cache and it's own cookies. When you delete your cache and cookies from Firefox, it doesn't delete the flash cache and flash cookies. Perhaps there's something in there that makes your flash choke.

Google "flash cookies" or "locally shared objects" for more information. For example: http://en.wikipedia.org/wiki/Local_Shared_Object

@hk

Quoting:The fact it works for you, doesn't make it work better for me, does it? Hence, my experiences above.

Equally the it doesn't work for me is not necessarily a valid argument against something.

TA: I know, and I'm even wel aware of the fact that it works for other people. I even know of people using the same distribution, packages and browser who have no problem with it.

But those experiencing software glitches / problems are usually a minority. Most of the times if you file a bug report, things are fixed, however, not for Flash. The fact that Adobe didn't fix the issue after over two years is an argument against Flash. If your browser became unresponsive and started taking 100% of your CPU, you killed it and before being able to restart, you had to use 'killall -9' - and all that about twenty times a day, you'd become frustrated too.

Anyway, I'll try to delete the .adobe folder and see what happens. Maybe that's the holy grail I'm looking for. I just compiled Firefox myself yesterday instead of using the precompiled binary to see if it makes any difference.

On the positive side of things: Yesterday I was watching some 'fullscreen' WebM, when after one minute it had some tiny problems. The image froze for three seconds, sound still worked, and then it worked fine again. It turned out (I forgot) I was compiling some programs, and therefore I had 100% CPU usage on both cores. But I could still watch WebM in fullscreen!

Apart from it, arguments can be made h.264 is an open standard while WebM isn't, probably I'll write an article about that issue next week.

Quoting:Yesterday I was watching some 'fullscreen' WebM, when after one minute it had some tiny problems. The image froze for three seconds, sound still worked, and then it worked fine again. It turned out (I forgot) I was compiling some programs, and therefore I had 100% CPU usage on both cores.

Try a real-time kernel. Just for giggles, I did a kernel build with "-j5" while watching a DVD, on an Athlon X2, running 2.6.33.3 with the real-time patches. The movie never once stuttered or stumbled.

If a real-time kernel solves multimedia problems (and seems to be a must for audio recording), why aren't all kernels "real-time" kernels? Presumably there must be a drawback to using one. Anybody?

Quoting:If a real-time kernel solves multimedia problems (and seems to be a must for audio recording), why aren't all kernels "real-time" kernels? Presumably there must be a drawback to using one. Anybody?

Performance. In the "throughput" meaning.

AFAIK the -rt tree gets merged bit by bit into the mainline, but it looks like there are some nasty hacks in there that don't meet Linus' quality standards.

Disclaimer: I'm not an expert, just an LWN reader ;)

BTW: What if you run the player with a FIFO or RR scheduling policy (both are "real time") on mainline kernel?

But if Flash crashes, I suppose giving it realtime-permissions is a stupid thing to do?

Quoting:Performance. In the "throughput" meaning.

Right. Everything gets slowed down an milligram or two, in order to guarantee I/O rates and response times more broadly.

So slower but predictible.

But not so much slower that it interferes with media playback.