Hacker News Reader

Comment by sneak | Hacker News Reader

If Linux idiots weren't senselessly married to terrible in-group traditions and the status quo, it wouldn't have taken 20 years (or Canonical) to replace the clusterfuck that is X.

View on HN

I use Linux and I don't even use a graphical environment most of the time because all it does is slow me down. (When I do use one, it's Ratpoison.) The most effective Linux developers I know work similarly. I believe this is a large cause of us being "senselessly married to terrible in-group traditions and the status quo". I don't know if you're calling all Linux users idiots or what, but I consider Linux idiots as the people who use Ubuntu and Unity and don't have any idea whats going on underneath. They obviously aren't going to be doing any replacing of X.

I am not calling all linux users idiots.

I made a picture for you to clarify:

http://i.imgur.com/ulGCde2.png (not to scale.)

Brutal.

Why?

Hang on, what have we got here?

TD;LR Linux is not one thing. Various projects are not mutually exclusive

We have GNU/Linux with a huge array of applications all licensed in such a way as to allow forking and customisation.

We have different groups of users of desktop/laptops; sysadmins; developers; scientific users; people who use administered desktops at work (e.g. French police &c)

We have a smallish and very pushy company in the UK (Canonical) who want to build on GNU/Linux to produce an operating system (Ubuntu) that will work on TVs, phones, tablets, fridges &c. Their target audience won't be developing alternatives to X, and probably won't be writing bash scripts or anything. (That isn't to say that some their users won't be doing those things).

We have a much larger company in the US (Redhat) who provide a stable conservative desktop as a compliment to their commercially supported server OS (RHEL). They part fund a desktop system that has changed radically recently (Gnome).

There is a German based company that make an enterprise desktop OS and server OS and who contribute to the development of a different desktop environment (KDE)

We have a not-for-profit foundation type entity who push out a version of GNU/Linux that can run on a huge range of architectures (Debian)

There is this thing called 'upstream'. Small projects push out all kinds of applications including alternative desktop environments.

So Canonical does its thing, some of the stuff they do may get back into the mainstream GNU/Linux, some may not. Some hardware manufacturer may or may not take up the system. I may be able to dock my phone to a keyboard/mouse/monitor and write documents in LibreOffice and edit podcasts/photos in a couple of years or 5. (bring that on).

The others carry on as they do now. You may have to choose the graphics card more carefully in future but I hope not. I suspect workstation class computers with separate monitor/bases will get more expensive and specialised anyway, and I imagine someone will do a 'high end workstation-OS' just for them.

The problem isn't Linux idiots (whatever that means) or their cautionary tendencies, it's Canonical.

Care to elaborate? Your comment reads as a sentence-long "NO, YOU!" right now. What about Canonical is a problem, and who are they causing the problem for?

As a "Linux idiot" myself, I hear your disapproval of the parent post, but am curious to hear the details that support your opinion (I might be able to learn something from it).

This is how I parsed the conversation:

    pilgrim689: Why is everybody hating on Canonical for
                ditching X11?
    sneak: Because they're "Linux idiots" who love X11.	
    ihsw: They don't love X11, they just hate Canonical.

ihsw is saying that the change isn't getting negative reactions because of the change itself, but because it is being made by Canonical.

edit: I accidentally using the wrong verb form.

I thought everyone was hating on Canonical for ditching X because rather than support the fledgling standard replacement for X, Wayland, they once again (just like with systemd vs upstart, compiz vs mutter/kdm) go off and do their own thing.

And if it turns out like either of those, they end up with an inferior barely maintained product that gets pushed to the sidelines once they show it off on a showroom floor.

Skimming the article, Canonical has some legitimate criticisms of Wayland, and are working on something that doesn't have those problems. Yay, this would be good! Except that it's by Canonical, and as you said, "Canonical never delivers". (That wasn't a quote from you, it was an iteration of "OP (never) delivers")

What I said was that when Canonical does deliver, they let the delivery wither and die without support because they are spread like a teaspoon of mayo thin on their footlong distro sandwich. See: Upstart, Compiz, Software Center.

We agree, I just called that not delivering.

Yes

The linux idiots have one glaring example: PulseAudio

Throw away user experience, system stability, judicious use of CPU for what?

It certainly improves certain details of the user experience, when it works correctly, also usually consuming more CPU than flash player

Am I the only person in the world who has never had anything but a pleasant experience with PulseAudio? The first time I noticed that I was using it, I had already been using it for a year...

Let me ask you some questions then:

- How often do you reboot your computer

- What audio softwares you use the most

For me, PA crashes around once a week (and this is with the latest version I tried), or just plain stops working (like muting the sound, garbled sound, etc)

Not to mention the delay that happens while switching songs in mplayer for example. Use ALSA and song switch is instantaneous.

> - How often do you reboot your computer

Once every two or three months, when I remember to apply updates.

> - What audio softwares you use the most

Only flash (youtube, primarily. Also youtube's html5 player probably, I haven't paid much attention to that) and mplayer. A few years ago, XMMS2. I used to use an XMMS2 client I wrote for music, now I just use some bash scripts and mplayer; I can't say I've noticed any delay while switching songs with mplayer or XMMS2.

When I play music from my laptop through my raspberry pi connected to my TV with pusleaudio, then I notice an audio delay. That is the only sore spot pulseaudio has ever presented me.

Very interesting. I thought some instability may be due to PA staying up for a long time, but apparently not.

About the delay, try this: mplayer *mp3 (in a directory with mp3 files, of course), then press >. There should be a significant delay in switching songs (like, 0.5s). Make sure mplayer shows it's using the [Pulse] driver

With ALSA this is very fast

"Not to mention the delay that happens while switching songs in mplayer for example. Use ALSA and song switch is instantaneous. "

There's a good possibility that this is mplayer's fault more than PA's.

No issue on a couple of machines. Running with a call to start-pulseaudio-x11 from my .xsession, works fine. I believe running it as a system daemon is not recommended, though.

My most recent attempt to use PulseAudio (as it came via Ubuntu) had the quirk of not activating the headphone jack when something was inserted. It dutifully turned off my speakers, but no sound was available elsewhere.

I previously had stock Debian on the machine, and everything was working fine. A half day's effort was put into toying with ALSA configs that Pulse was deferring to for whatever driver choices it made. Dozens of help articles or threads where everyone else in the world could proclaim "Thanks, that fixed it for me!" left me with less functionality than I started with.

The only thing Pulse has ever done right for me was was help to record some loopback audio on an otherwise-crippled sound card.

No, there are a decent chunk of users for whom it works fine. Maybe as many as 95%. But there's also a substantial number of users for whom it introduces major problems - which would be more forgiveable (at least to yours truly) if it provided any visible advantages over existing systems.

(After a few months of trying to make pulse work I left for FreeBSD, which works beautifully and avoids a lot of linux's change (seemingly) for the sake of change)

I haven't had any recent trouble with it. Back in the early days, it didn't always work well with SDL, but that was fixed years ago. I've been able to crash the daemon with stress tests, but it seems to restart automatically.

I can't understand all the dislike of PulseAudio either. It certainly has a nicer API than /dev/dsp.

No, you're not the only one. I'd been forewarned of the awfulness of pulse, but didn't even notice its installation until later, when I discovered all the useful things I could do with it: play music over the office Apple airplay thingy; use two sets of headphones simultaneously; send audio to another computer in the house. Without any typing, searching, or rebooting,

I remember OSS and Alsa, and I don't miss either. As far as I can see, it makes Linux audio work like any other computer in this millennium.

I agree with. The only problems I've ever had with it have turned out to either be problems with the actual hardware or problems with the clients, not problems with Pulse Itself.

I will admit having a lot of a hate for figuring out which audio 'profile' is which...

If you've got the plugs for 7.1 and a normal line in as well as possibly a digital one, they you've probably got 23 choices of profile for your built in audio device. Add in a GPU with HDMI and the NVIDIA driver and you probably get at least another 7+ choices all labeled Nvidia... I find I frequently have to guess multiple times whenever I hook up a TV via HDMI.

But it works quite well if you define works as works without adding on any usability riders.

PulseAudio is good now. It was a different story for me back in 2007/2008 though.

> The linux idiots have one glaring example: PulseAudio

PulseAudio works wonderfully for me. It allows software mixing, something that's never been possible for me prior to PulseAudio, and it never crashes.

It also gives me a single place to control volume on a per-app basis (pavucontrol). That's not something I had prior to PulseAudio, either.

I can also keep a single desktop session going for months.

I agree with this, I have similar experiences with pulse. I also use the pulse equalizer, which is super easy to set up.

I like it. I think it does things right, and they can optimize it more over time. It is better than having a fast audio subsystem that just can't do what I want. Pulse is a kitchen sink that at least covers its bases.

Not this again... You could have complained about this years ago, but it makes no sense today. PulseAudio got fixed.

> user experience

you can set up PulseAudio for LAN audio streaming with padevchooser (GUI). Could you do this with ESD? Arts? Tell me about this audio streaming solution with super UX for Linux that you know...

"you can set up PulseAudio for LAN audio streaming with padevchooser (GUI)."

Yes, and? I understand this is a useful feature for some people, but I'm just interested in listening to music and PA does not me allow to do that without frequent crashes and bugs

I don't care about the features of a car if it stops working at random. And yes, this has happened on every distro I tried since PA started being bundled. Every single distro, every single machine I've had only a stable system with PA OFF

Bugs can be fixed. Pulse is newer than Alsa and more complex, you should expect bugs. If you don't have a problem with the implementation details of PA (software mixing, optional software equalizing, networked audio, stateful audio adaption to devices) then just report the bugs and make it better. Don't whine and try to fragment the Linux space any worse than it already is, especially with Mir's announcement.

Yes, I should expect bugs soon after release

But PA is 8 years now.

I don't remember any software being so problematic as PA in the Linux world. Some have more problems, yes, but they usually have a limited deployment.

Flash Player? You didn't have problems with that?

Of course I did. Still less frequently than PA

The real question is why did you need something newer in the first place ? Are years of bugs, misconfigurations, random crashes worth the minimal improvement end-users will get in the end (when everything finally works for everyone) ?

Yes, pulseaudio has all sorts of awesome esoteric features. You can stream sound transparently from a linux box to a windows one. That's kind of useless if it brings down your computer.

Parent seems to have been experimenting this issue recently. Saying "PulseAudio got fixed" does not mean "it works perfectly for everybody" (see https://bugzilla.redhat.com/show_bug.cgi?id=813407 ). Most people will care more about getting a sound system which works without crashing and not sucking up CPU before getting sound system with streaming. The fact that you and me have had a good experience with it does not mean other people are not having issues with it (just as their bad experience may not reflect the overall quality of the software).

That's a Fedora bug. It wouldn't be bleeding-edge (enough!) if it didn't crash. Running Fedora and complaining in a public forum about crashes (without saying the distribution name in the first place!) makes no sense to me.

"without saying the distribution name in the first place"

RHEL 6, Fedora 16, Mint 11 (while it was not obsolete), Ubuntu Maverick Meercat (amongst others)

Oh ok, so Fedora users just have to suck it up and having a crashing system? Congratulations, you don't know anything about distros.

> you can set up PulseAudio for LAN audio streaming with padevchooser (GUI). Could you do this with ESD?

Considering that was pretty much the point of it, of course.

The point of ESD? It wasn't nearly as user friendly if I remember correctly. But I was a KDE user.

Oh, didn't realize you were asking about the UI specifically. I believe you would just set an environment variable of where you wanted the sound to go.

The part I liked was that you would start ESD and everything just worked.

> It certainly improves certain details of the user experience, when it works correctly, also usually consuming more CPU than flash player

You make a sweeping statement which is not true for every system. Looking at what happens playing a Youtube music video, according to htop:

- pulseaudio is using between 0 and 1% CPU

- plugin-container (wrapping Flash) is at 3%

- firefox oscillates between 20 and 40%

Now, my understanding is that, at the time pulseaudio came out, it was using untested functionalities in audio drivers which, for many chipsets, did not actually work and progressively got fixed. There was the same problem with graphics drivers when KDE 4.0 was introduced. Maybe you're out of luck and actually have a defective driver?

Thanks for this insightful answer

I'm not sure I have a bad driver, the last two chipsets I used were Intel (and no weird audio devices), works fine with other sw. Latest distro tested: RHEL 6 and Fedora 16

For me, PA would stay around 5%, 10% sometimes, and even more depending on the task. Not to mention crashing every week (at a random moment)

That's curious. You might be a victim of this bug:

https://bugzilla.redhat.com/show_bug.cgi?id=813407

I'm not sure if this could affect the CPU usage. I hear that you can play with the resample setting to lower it, but I haven't had to do that on a 5+ yo desktop.

You should avoid RH based distros for Desktop related activities. Perhaps you should try Ubuntu, Mint or Suse.

Nonsense. Fedora is a fine desktop Linux... I run it on a woefully underpowered laptop which I use for development, and it is every bit as usable as, say, Ubuntu (which I use on my corporate laptop). In fact, I'd take my Fedora/KDE setup over this Ubuntu/Unity setup any day.

Hum do you mean HDA by Intel ? If so, look closer because HDA is an umbrella for a lot of codecs.

Edit: to be clearer, HDA does not automatically mean Intel chip. I made that mistake before.

I'm going to guess you haven't used pulseaudio in the last 3-4 years?

I conducted an experiment on my last laptop, having noticed PulseAudio taking up more CPU than my media player. I found, quite consistently, that disabling PA led to my media player using more CPU on its own than the sum of the media player and PA when both were running.

Reply on news.ycombinator.com