A Journal Comes To systemd

[d4ddi0]

I'm sorry

I'm SORRY about my comment attributing network-manager to Lennart...

I'm not sure where I got the idea. It may be that the hatred I recently felt for systemd (before removing fedora from the machine) reminded me very much of the hatred I felt for network-manager.

Its a very cool idea whose implementation leaves the user (me) bleeding, because I got to experience all the sharp edges.

I never did figure out how to restart gdm without rebooting the machine... somehting that has been simple with every other init system I have ever used.

[liam]

Not just inefficient, extremely limited.

I'm all for mixer in kernel space. I'm not a fan of the entire way Pulseaudio + ALSA work, but it beats the crap out of anything OSS has come up with.

Pulseaudio is a good software mixer in the wrong place, and OSS has a bad mixer in the right place.

Correct me if I'm wrong, but doesn't oss4 require floating point in kernel space?

[DaemonFC]

I was actually not against Pulse. ALSA has very inefficient software mixer. OSS4 has complete mixer. Hence lots of trolls, using OSS4 themself yet claiming "ALSA+dmix" is enough.

But why is ALSA+dmix more powerhungry than pulse?? Pulse "wastes" 20-30Mb of RAM, so it has to be more powerhungry than simple mixer, no?
Upd: Ah, I think I understand what you mean. Pulse sends audio data in bigger packages then sleeps more, hence consuming less power than "brainless" dmix...?

Not just inefficient, extremely limited.

I'm all for mixer in kernel space. I'm not a fan of the entire way Pulseaudio + ALSA work, but it beats the crap out of anything OSS has come up with.

Pulseaudio is a good software mixer in the wrong place, and OSS has a bad mixer in the right place.

[crazycheese]

power-hungry ALSA + DMix

I was actually not against Pulse. ALSA has very inefficient software mixer. OSS4 has complete mixer. Hence lots of trolls, using OSS4 themself yet claiming "ALSA+dmix" is enough.

But why is ALSA+dmix more powerhungry than pulse?? Pulse "wastes" 20-30Mb of RAM, so it has to be more powerhungry than simple mixer, no?
Upd: Ah, I think I understand what you mean. Pulse sends audio data in bigger packages then sleeps more, hence consuming less power than "brainless" dmix...?

[DaemonFC]

You listened to only one side? Didn't you know about broken ALSA drivers? Pulseaudio have been revealed many bugs in ALSA drivers, some of them are buggy as hell even now. The overall drivers quality become better. Win-win situation, even if you use power-hungry ALSA + DMix instead of good-engineered Pulse with much lower power consumption and per-application volume control and better resampling and other cool features Pulse has.

Typical "It's not bugging someone in this particular way so it doesn't get fixed until it does". People that FUD Pulseaudio probably last used it in 2008 or something when Ubuntu started shipping their badly broken version of an already alpha quality (at the time) Pulseaudio server.

If they try it again on a modern Fedora or *buntu, they will see it's not bad and when it does fail it's usually because of some broken shit like VLC. Boo hoo. VLC is bad bad bad. Good thing there are options...

Edit, if you need VLC to work, you can set Pulseaudio to use old fashioned interrupt timing, which will cause the entire system to use more power from all the extra wakeup events, but it hides VLC's bugs pretty well. VLC is designed for Windows XP or FreeBSD style sound systems, not ones that work in modern operating systems like Linux + ALSA + Pulseaudio, or even god forbid Mac OS X with CoreAudio. (Yes, I just said OS X does something right. Shock!)

[reflexing]

And then remember how many many times he was trash-talked regarding pulseaudio? With critics either using OSS themself, or using text-configured single ALSA card, with non-audio-blocking applications cherry-picked :X
Constantly and continuously. Same here; you just wait.

You listened to only one side? Didn't you know about broken ALSA drivers? Pulseaudio have been revealed many bugs in ALSA drivers, some of them are buggy as hell even now. The overall drivers quality become better. Win-win situation, even if you use power-hungry ALSA + DMix instead of good-engineered Pulse with much lower power consumption and per-application volume control and better resampling and other cool features Pulse has.

[DaemonFC]

And then remember how many many times he was trash-talked regarding pulseaudio? With critics either using OSS themself, or using text-configured single ALSA card, with non-audio-blocking applications cherry-picked :X
Constantly and continuously. Same here; you just wait.

FWIW, OSS and ALSA's horrible to nonexistant software mixing was cited as a reason Jamie Zawinski went to OS X on his desktop. (In 2005, before Pulseaudio)

With OSS (More or less a BSD thing these days) or ALSA and no Pulseaudio, you go back to the 1990s where being able to access the soundcard becomes a race condition if the thing doesn't have a hardware mixer, which is only typically available on high end dedicated cards (the cheap Sound Blasters they sell at Walmart don't cut it, they're nto even an upgrade from what comes on your motherboard.). In some cases, integrated is all you have, due to lack of ports or being on some kind of portable or embedded system, or not wanting to throw $200 at a sound card, and mixing just has to work. Enter Pulseaudio. I don't know why it is so maligned, it took a situation that was unbearable and made it into one that mostly works, with a few minor glitches in some corner cases. Go use OSS if it suits you, it's complete and utter dog crap, even the version 4 one, but go ahead. Or rip out Pulseaudio and see how long until you're tearing your hair out with ALSA's mixer. Lennart Poettering has expressed frustration that OSS (again, BSD thing) isn't even good enough to use with Pulseaudio properly. If you want to avoid Pulseaudio by using a BSD that never improves as a desktop system as some kind of "strategy" to avoid new and improved stuff on Linux, be my guest, your system.

Back to systemd journal, there's already a lot of FUD flying, but let me discredit the two that make no sense at all.

1. It doesn't have to replace syslogd. It can be installed side by side if you need the old primitive syslog too for some reason.

2. It DOES save the journal to the disk, it simply stores better information than syslog does, and only the relevant bits, so less stuff ends up hitting the disk and crapflooding your logs with useless junk. Also worth mentioning is that if the system is under attack, they can't easily forge the journal like they can with syslog (Used by all Linux distributions and all BSDs) now.

When the journal comes along it will just be another excellent tool that BSD is too primitive/crude to make use of.

[crazycheese]

Why do you think Lennart have anything to do with network-manager? Anyway, IMHO Lennart's pulseaudio and systemd are one of the best services developed for Linux in recent years, very well-engineered.

And then remember how many many times he was trash-talked regarding pulseaudio? With critics either using OSS themself, or using text-configured single ALSA card, with non-audio-blocking applications cherry-picked :X
Constantly and continuously. Same here; you just wait.

[curaga]

Why yes, one binary to rule them all. Next it will include X, sshd, libreoffice, and doom3.

[peppepz]

Get off my lawn

They'll have to pry the syslog from my cold hands. My personal experience with these next-generation subsystems is not positive at all. I do hear that many people are happy with them, but I'm not. I'll resist installing them as long as they don't become de-facto mandatory because they get included as a dependency in some software I can't do without.