State of Sound Development On Linux Not So Sorry After All 427
An anonymous reader writes "There have been past claims by Adobe and others that development on Linux is a jungle, particularly with regards to audio. However today, the author of the popular 'The Sorry State of Sound in Linux' has posted a follow up showing Adobe's claims to be FUD, as well as being a good update on where OSS and ALSA are holding today, and why PulseAudio isn't a good idea."
By saying that he proves his former point (Score:4, Insightful)
it's all relative (Score:5, Insightful)
PulseAudio... (Score:5, Insightful)
Pulse Audio: the best gift the Linux world gave M$ (Score:5, Insightful)
Pulse Audio is a bloody disaster. It breaks just about every audio application I have, and even when its not running, it creates over runs and under runs in other ALSA and SDL audio applications (like ZSNES). ALSA, and SDL audio was the perfect sound abstraction system. Pulse Audio screws EVERYTHING up. I have to makle my own patched RPMs to get rid of Pulse Audio hooks in applications. Its bad. Its really bad.
Audio applications should use ALSA but not lock the card. Games should use SDL. Everyone else should follow suit.
If an application is locking a card its the drivers fault. Fix the driver, fix the over runs, and ditch Pulse Audio!
ALSA was a mistake (Score:4, Insightful)
The fundamental problem (Score:5, Insightful)
Instead of attempting to write a one size fits all sound library that would interface directly with the sound hardware and provide the direct interface for applications who wish to play sound, what they should have been done was drafting a specification for an API that contains only the most basic audio features (creation of primary / secondary audio buffers, enumerating supported device buffer formats, etc.). The driver provides the implementation for the specification. If the device driver indicates the device is capable of hardware mixing, it should use hardware mixing internally, if it doesn't, it uses software mixing internally, if supports the use of hardware buffers for secondary buffers it can do so, but this all will take place within within the driver specific implementation of the standard specification. This should have been paired with a robust generic open source driver that (hopefully) supported as many generic audio devices as possible. Using the interface exposed by the spec directly might seem a little low level, but additional software libraries could be built on top of that interface for use by applications. The important advantage if they had gone down THIS road is that the single conduit, the arbiter of all things audio in the system would've been the device driver for the sound hardware, which would reside neatly in the kernel.
10 Years From Now You'll Be Writing The Same Thing (Score:1, Insightful)
Except the sound project names will be different.
The way they fail to match commercial OS sound systems will be somewhat different in detail but equally abysmal.
There will be the same "Works for me!(tm)" posts.
The only hope for Linux to get a commercial quality sound system is if one or more major commercial companies with competent grownups take on the task and have people sitting at desks 40 hours a week doing real work.
The open source "competing teams of bearded GNU freaks taking their sweet time working on competing projects because "choice is good(tm)" in their spare time between bong hits and World of Warcraft raids" isn't something anyone should be waiting around to start finally delivering commercial quality software engineering solutions.
KISS (Score:4, Insightful)
I hate PA. It's a complex mess and half the time it just doesn't want to work right. There is no way your average user could deal with it. Most of the time I have trouble with it not allowing multiple users to have audio at the same time seemingly due to some twisted sense of how security should work. ALSA is better than PA but still doesn't work a lot of the time.
It sounds like OSS is getting it's act together and just needs someone to hire the lead developer(s) and port all cards missing OSS support over. That sounds like a worthy goal for those selling distros or soundcards. If it works well and is easy for developers then it'll work well for end users. That is what matters. Sound has been my #1 embarrassment when pushing Linux. It has never worked well and it's time we get it fixed.
Re:What are we trying to achieve? (Score:5, Insightful)
So again, what was Linux hoping to achieve by dropping old "obsolete" OSS in favor of increasingly complex solutions?
As far as Ubuntu is concerned, its the same inane neophyte behavior that "obsoleted" Xmms and BMPx in Jaunty in favor of the iTunes wannabe Amarok, which I find much less stable and cumbersome to use. There was absolutely nothing wrong with Xmms as a Winamp-style media player that was quick, efficient and could handle Internet radio and almost all the popular DRM-free formats, yet it was automatically removed with other "obsolete" software. Yes, I can compile it again from source, but it just seems a bit obnoxious. BMPx was another simple media player that was quite nice, albeit with the occassional bug, and it too has been "obsoleted".
For all the evangelism of the Ubuntu community, why are we being driven towards solutions that mimic the proprietary soup-du-jour (iTunes in this case)?
Re:Main blocker (Score:4, Insightful)
Troll. I lolled.
About a gazillion years of linux use lie behind that comment. I love linux to death, its the greatest thing to have happened to 'PC' since .... ever.
The irony lies in the fact that modprobe - in fact the whole, loading unloading kernal modules on the fly - is nothing short of amazing.
I can't think of anything more impressive (in context) than being able to dynamically modify the core OS behaviour through a simple set of command line tools.
The only problem that remains is that 'everyone else' doesn't ever want to even know that stuff is possible.
Re:What are we trying to achieve? (Score:5, Insightful)
"So again, what was Linux hoping to achieve by dropping old "obsolete" OSS in favor of increasingly complex solutions?"
Linux deprecated OSS2, which everyone agrees sucks hard. It was a no-brainer.
OSS3 is significantly better but it was only recently open sourced. Frankly, if the OSS devs hadn't spent most of the last decade with their heads firmly wedged, audio on Linux would probably be in a much better state. Ah well.
Re:Main blocker (Score:4, Insightful)
kernel is fine, distros have problems (Score:3, Insightful)
TFA says that the way sound is implemented in the kernel is basically okay, but there are problems with how the kernel's facilities are used at higher levels by applications, and with the way the whole thing is integrated by distros. I think he's basically correct.
As an example of what's not broke about the kernel, and doesn't need to be fixed, it's a good thing that we still have support for OSS. OSS allows you to do sound I/O in exactly the way you would expect to do sound I/O based on the fundamental design principles of unix. You just do open(), ioctl(), read() or write() on devices like /dev/dsp. If you couldn't do that, it would be a failure to do the obvious, straightforward stuff to handle sound in the Unix Way.
As an example of what is broken at higher levels: I run Ubuntu Jaunty. Sound works fine every time I boot the computer, and I get the bongo sound as the login screen comes up. Then when I log in, master playback is muted, and the volume is down at 1/31. Also, the way the Gnome icon shows me that sound is muted (a tiny red box with a white x in it) is the same as the way the network icon would show me that I'd disconnected my ethernet cable or something; in other words, it makes it look like it's not just muted, but actually broken. Here's [ubuntuforums.org] my best attempt to characterize the bug: Here's [launchpad.net] a bug on launchpad that may or may not be the same thing:
A sure road to success ..... (Score:5, Insightful)
... when application developers or users express concern about a problem in your OS is to attack them, call them liars and FUD rakers, accuse them of being stooges for Microsoft or whatever.
I'm pretty sure the engineer who develops the Flash Linux player is probably on your side, and he was expressing a legitimate concern about a problem with Linux. As best I remember Adobe hired him out of the open source, Linux world. It would probably be more productive to listen to his concerns, and see if maybe, just maybe, there is a problem with audio on Linux. Having tried to write simple audio apps myself using OSS and ALSA I can assure you they have issues, OSS having no mixer at all was a nightmare to make play with more than one audio stream or more than one app at a time, that's why ESD, arts and pulse were created to hide these mixer deficiencies.
ALSA is a ridiculously overdone, convoluted audio API which makes it very painful for audio driver writers and application developers alike. It simply has too many knobs that can be tweaked and turned most of which never get implemented properly by driver writers and can't be trusted.
The simple fact that there must be a dozen different audio API's on Linux many of which exist solely to hide applications and users from the deficiencies in OSS and ALSA tells you something right there.
Rather than attacking this guy maybe you should have the empathy for the guy, he has to deploy an application that is used by probably millions of Linux users, most of whom are ticked off its not open source in the first place and then when it doesn't work perfectly they scream bloody murder. He has to try to make audio work in the face of the fact there are countless barely working or at least buggy ALSA drivers in the world, and there must be about a HUNDRED different ways to configure audio when you count OSS, ALSA, gstreamer, pulse, esd, arts, jack, OpenAL, and a MILLION different configurations when you count all the obscure options you can or in some cases HAVE to set on audio drivers.
As an end user I've suffered through painful, hard to fix audio bugs, in just about every PC I've owned over the last ten years due to audio driver bugs. Sure I could sift through "supported" hardware lists and try to find that rare new PC or laptop where everything is guaranteed to work on Linux, but I would actually prefer to just buy the hardware I want at the price I want. Of course in all fairness to the Linux developer community it is a total bitch to get working drivers on all the PC hardware being put out especially when the vast majority of hardware developers either just don't support Linux, support Linux badly, or actively obstruct Linux support.
You all seriously need to realize that if you want broader acceptance of your wonderful operating system:
A. You need applications and application developers to develop for your system, and not attack them if they point out problems deploying apps on your system. In a perfect world every app would be open source, but there may be some apps which aren't Linux would be better off having as closed source than not having at all.
B. it will have to actually work for ordinary people who aren't going to spend days/weeks/years fiddling with things to try to make it work right.
One of the beauties of the Mac is the hardware is tightly controlled. You may view that as confining and depriving you of your freedom, but it also helps insure the damn thing works out of the box, and most of the applications on it work pretty damn well. After years of fighting nagging bugs on Linux I decided it was in my own best interest to just switch to a Mac for my desktop system and I use my Linux box solely to develop code on. Linux on the desktop is a lot better than it was but unfortunately its just not a very good desktop experience by comparison.
Unless there is a major attitude adjustment in the Linux community that is unlikely to change. Either:
A. Be content that Linux is a niche OS for hardcore fans a
Re:it's all relative (Score:5, Insightful)
You know, when i read moronic bullshit like that I sometimes wonder for a moment why don't I just stick with OS X. Why should I spend endless hours every other month or so trying to get a satisfactory result out of various Linux distros, anyway? Then I remember, I love Linux in spite of assholes like you. And I hope one day to be able to say with pride, "I recorded, mixed, and mastered this project all in Linux, using nothing but FOSS!". And when that day arrives, it will be in spite of, not because of, idiots like you. Burying your head in the sand may make you feel better in the short term, but it doesn't get the work done.
Developer FAIL (Score:5, Insightful)
Wait. Claiming audio sucks on Linux is FUD because there's not one, not two, but three mutually incompatible and redundant APIs? How the hell is this not a clusterfuck?
Oh I'm sure there's some reason why someone prefers one to the other, but seriously. You're sending bits to a soundcard. That's it. Just make one API and be done with it. Got a beef with the API? Enhance it, don't just throw it away?
My god, audio was one of the reasons why I ditched Linux for a mac four years ago after running it as my primary OS for ten years prior. Frankly I got tired of having sound work in some applications, but not others. I got tired of guessing which mixer would adjust the sound, which mixer wouldn't. I got tired of seeing "No ALSA cards detected" in my startup, but someone how having `alsamixer` be the one mixer that worked most consistently.
This is a mess made by the developer community and developer community has so far failed to show that it is capable of solving it. If only there were a Benevolent Dictator or something...
Re:Main blocker (Score:3, Insightful)
Whoever modded you a troll is a moron.
The problem is most likely the video drivers. Download updates from NVidia's website. The free drivers for NVidia cards will get your display working, but it won't be fast.
Re:The fundamental problem (Score:1, Insightful)
What you describe is more-or-less what the Open Sound System is. Mostly less from what I seen, but like a lot of *nix based things, only a few minor tweaks should modernize it to what you describe.
Sadly, Linux switched over to ALSA, a complex beast that while improving audio on Linux initianlly (from the user's view, well mine as a user), it ended up doing little to improve the user-space situation: audio servers are still common and implement most of ALSA's most importent features (such as mixing for cheap hardware). In the end, ALSA is simply a complex mess thats a pain to configure to make it work better, and a pain to program for, and your programs will still only use ALSA thru a sound server, or at the very least a sound API wrapper library.
It would have been better to simply improve OSS, add the damn software mixing in it (that will fix most problems user's had with it), and leave the rest to userspace tools. Nice, clean, simple.
Re:By saying that he proves his former point (Score:5, Insightful)
If Pulse Audio really sucks, then Linux Audio really is in a sorry state
No, because you do not have to use Pulseaudio.
He says pulse sucks for games . Although he is exaggerating the latencies, I can believe it.
It is so, so for video (you can get occasional lack of sync)
It does audio very nicely - mixing works fine, you can play different streams to different cards (yes, I do that), you can play streams on remote servers, you can combine all local sound cards into a single virtual device etc.
So the problem is not that we do not have good solutions. It is that we have different solutions with different strengths and it is not clear which should be the default. He thinks pulse should not be the default. I like pulse although I would like the latency and reliabliity issues dealt with.
PulseAudio (Score:5, Insightful)
It is just the best possible counterexample of "Just Works(tm)". In other terms: each time I try it, it just "Doesn't Work(tm)". Without it, sound works more often than not; I don't care why or how as long as it does work. Simple observation: "apt-get install pulseaudio" breaks audio, "dpkg --purge pulseaudio" repairs audio.
Hm. Maybe that's how Linux audio is supposed to be brought to a (relatively) sane state: by breaking it so terribly that rolling everything back to the previous state would almost look like a step forward.
Re:it's all relative (Score:1, Insightful)
Not trying to troll here, but what's with all the "we use it and it works great" comments and not one link to any such projects? Can you say, "[citation needed]"?
Re:ALSA was a mistake (Score:5, Insightful)
You hear this a lot in the open source circle. Many projects have close competition (Gecko/KHTML, KDE/Gnome, etc.) where this comment might apply. The problem is that "fixing the few issues that existed" is frequently not only very hard, but also very boring. Put another way, if it was fun and/or easy, the original developers would've already done it. IOW, this is probably crappy work that you have to pay people to do, and unfortunately free software doesn't usually pay, so volunteers gravitate towards the fun and easy (at least, perceived to be easy), which is often to start a new exciting project.
Re:Main blocker (Score:5, Insightful)
All modern operating systems offer this functionality, most from the command line (ie on OS X it's kextload/kextunload). It's not some amazing Linux thing.
Re:A sure road to success ..... (Score:5, Insightful)
"This is not a police state"
The Linux and Slashdot communities are certainly not police states but sometimes they do degenerate in to "mob rule". In "mob rule" the people who excel at shouting others down and pandering to what the mob wants to hear often win even if they are wrong. I've excelled at it myself here sometimes :)
I'm not sure if the problem is even in any of the the actual fracking articles no one reads, but more the trollish way things were worded in the submission to Slashdot, which seems designed to provoke a fight. The submission seems crafted to make Adobe look evil and bad, Linux look good, for no actual reason other than its certain to be red meat to the slashdot mob and sure to win acceptance with the /. editors. I hate Adobe and Flash as much as the next guy, maybe more, but from what I read in the blog of the Adobe Linux guy he seems to be a pretty decent guy trying to make good on a tough situation.
IMHO, and it appears in the opinion of many others posting tonight, audio on Linux is deeply messed up, and its a leading factor in killing Linux acceptance on the desktop. Linus or other Linux kernel leaders seriously need to step up, lead a rational discussion of the problem, throw out all the old biases and misconceptions and come up with a rational fix. Audio on Linux has been bad for 10 years, its not getting any better and ALSA is more the problem than the solution, as are all the hacks like pulse, esd, arts, gstreamer, etc.
Audio more than any other issue turned me and several others I've read tonight from Linux to Mac for our desktop, and I've been a Linux fan and desktop user for a long time as you can tell from my 5 digit /. user ID so it wasn't a switch I made lightly.
The original submission makes it sound like audio on Linux is great and its only a bunch of whiners and evil corporations spreading FUD about Linux saying otherwise, which is pure mob rule pandering.
Re:What are we trying to achieve? (Score:2, Insightful)
I don't see what you see in Amarok that makes you think it's an iTunes wannabe. I see all the contrary. But I am with you with the fact that Amarok 2 is unstable and cumbersome. For some reason, going from 1.4 to 2 was a big regression in term of usability. But try Amarok 1.4, it's great, and far from being an iTunes clone. I hate iTunes, but love my Amarok 1.4.
On the other hand, Rythmbox IS an iTunes wannabe, for anyone who wants an iTunes clone on Linux, I think Rythmbox is the way to go.
Re:By saying that he proves his former point (Score:5, Insightful)
Someone please fork OSS4 (Score:4, Insightful)
I'm sure from all the rave reviews it's technically superior and all, but right now it's controlled by a paranoid schizo who hasn't got a clue how open source works: after GPLing it and whining that he hasn't suddenly started making money, he now thinks he can dictate what license apps using his API have to be released under.
So, when do we go ALSA -> OSSv4? (Score:3, Insightful)
Looking at the charts, and looking in a few other places, it is clear to me that OSSv4 is the way to go.
So, when does this start to happen? I tried this a few months ago, and I had to patch my kernel and do all sorts of other things that ended up hosing sound completely (since I'm not a developer, asking me to do developer-ly things is trouble).
When will it be a simple switch in the kernel config, or a simple matter of installing a package in the major distros?
PulseAudio can be fast (Score:2, Insightful)
But, more to the point, your latency relies on the program itself much more. I've been doing quite a bit of low-latency audio programming using both ALSA and PulseAudio, and they both get extremely small if you know what you're doing. With ALSA, I've gotten to around 1ms, while I can easily get PulseAudio to approximately 5ms. Using a PC that has two cores, I've gotten the numbers cut down to around 0.5ms and 2ms, respectively.
If you're having a real touchy time getting response that low, you likely have to bump up priority of the process. Using ALSA, simply setting the application to realtime scheduling will do it. With PulseAudio, you're going to need to set the audio server itself to realtime as well as the application (I haven't done too much testing with that, but that seems to be the consensus online). As a parent suggested, you probably want to look at using JACK. It does most of the dirty work for you.
I learned most of this working on a low-latency audio application. Just yesterday I got the thing to route a guitar at low latency using ALSA, and I'm looking to finish up the Pulse portion this next week or so.
Re:Main blocker (Score:3, Insightful)
A single bad moderation doesn't make a bad community. Hell, they're modded 4 at the moment. Let the system do its job.
Re:Main blocker (Score:3, Insightful)
incorrect. The proprietry nvidia drivers are not even installed by default, and never have been. There was talk about providing them out of the box but it never eventuated. They are, however, easily installable via the "hardware drivers" app.
Re:Main blocker (Score:3, Insightful)
OSS. It's standard on all UNIX-like systems except OS X. The API is painfully simple. There are no libraries, the only dependency is a single header (soundcard.h). To play sound, you open() /dev/dsp and write() the data there. If you want to change the format (e.g. sample size, rate, or number of channels), then you issue an ioctl().
If you use FreeBSD, then you've had in-kernel sound mixing with OSS since around 2001. It was a bit difficult to use back then; you got a different device for each channel and had to configure each app to use a different one (I had one for each of KDE and GNOME's sound daemons, one for xmms, and one for games, for example). With FreeBSD 5 (January 2003), each app to open /dev/dsp got a new vchannel, so all sound-outputting apps got a trivial-to-use interface that supported mixing.
Now, both OpenSolaris and FreeBSD support the newer OSS API, which is backwards compatible but also provides per-channel volume controls and a few other nice things. Linux does if you install the 4Front OSS implementation (which the author of TFA is recommending), but by default ships with ALSA, which is a Linux-only API that is not backwards-compatible with OSS (which was the standard Linux sound API until ALSA shipped).
It amuses me that I've had multiple applications able to play sound on a cheap soundcard, using a simple, well-supported, API on FreeBSD for almost a decade, and yet people still claim Linux is ready for the desktop but FreeBSD isn't.