Mar 2009
ext4: latest linux braindamage
So what the fuck are the communists of the FOSS herd up to now?
Yea right, another braindamaged filesystem for their broken-beyond-repair "Linux". Now that Reiser is in jail for slaughtering his whiny bitch wife, someone really had to step in and fill the gap in the long line of psychopathic linux developers.
Here's some background insight, the thing you never get from your local linux zealot. So there we are using write caches ever since some guys from IBM apparently invented the principle. Yea, using fast medium for temporary storage of data is good. Why? Go on, disable your CPU's and disk's caches and you'll immediately feel the pain. Write cache is among us for the latest thirty-some years, and everybody adores it. Let's fast forward those old cold war years to the year the next cold war started: 2009.
After the data annihilator¹ murderfs thankfully disappeared into the morning mist, some linux zeals created 2 (yes, two) new wheel-reinventing filesystems, completely oblivious to the fact that there's already a gazillion of non-working craplementations for better filesystems in their crappy kernel (see the HFS+ and Files-11 drivers). The other reason for coding new shit is that their shitty, limitation-stacked viral GPL license does not allow the inclusion of Sun's (otherwise fucking great) ZFS. Linux and "free" software is just way to cool.
So two new
Apparently some Ubuntu users (yes, some insane people actually do install shit with a name like this) experienced that their usual data losses with the horseshit linux kernel that has just crashed actually got a lot worse than with last week's hypefs. All their open files were zeroed out and such. You think that's bad? Wait until the vietcong author Ted Tso steps in.
The usual stuff happens: there's a bug in linux, so they start a religious crusade around it. This chink actually gets the audacity to argue that everything in the previously super-hyped-überalles FOSS industry is just badly written code. Oh wait, he's actually mostly right on that, but this is not the reason.
His crusade bullshit has more attack vectors, as usual.
First, accusations of heresy: "This is what is required by POSIX" - he shouts in half-conscious religious zeal. The mentioning of "POSIX" is a callword, it makes the believers fall to their knees and murmur "thank thee Lord for POSIX" and all that crap, despite the well established fact that they have absolutely no idea of what it actually is, and how much braindamage does it contain (see Posix ACLs here earlier). Never mind the fact that POSIX doesn't say "don't use write caching or you will die a terrible death." It would be the stupidest thing.
Second, fucking over with people actually using his fs: "Application writers had gotten lazy." Yeah, right. Those lazy bastards have actually been happily using write-caching for the last few decades, completely ignoring the future possibility that some idiot might code such a crappy filesystem for such a crappy kernel, that using DRAM cache for making your miserable IT using life even somewhat bearable will result in catastrophic data loss.
Third, assuming similarities where there are none: "xfs has had delayed allocation for years" Yes. And you know where it came from: Silicon Graphics' IRIX. You know what are the major differences between an Octane and your PC? The Octane works and your PC is a stinking shithole. Not to mention IRIX, see, the guys using this (then-top-notch) architecture were actually busy creating Toy Story and other shit in the whole nineties, while you are doing what, ten years later? Busy waking up from nightmares where application writers don't explicitly fsync their data onto your latest work of "code" before their hardware burns away or their latest "enterprise" Bubuntu Kernel panics away in dying screams. All thanks to the swarm of politican coders like you, that were mass-injected during the late days of kernel 2.4, commiting
Fourth, jumping on the usual bandwagon by saying "apparently some Ubuntu users are happy using proprietary Nvidia drivers". So yeah, Shitty Teddy argues that it's mostly "Ubuntu gamers", whose linuz box freezes and dies unexpectedly. Look, Ted, in case you haven't been paying attention to the gamer scene's latest years, linux is just not a gaming platform, you realize that?.. If you wanna game, you don't choose Linux. So the guys who you've just belittled are the sweet little linux switchers who would actually play something just a little more 2009-ish than Soko-Ban or Nethack while they are not possibly busy with deploying linux servers... using your filesystem? Well, from now on, probably not. Ideally, you should be hoping for these guys, but since you actually code linux for big money from big companies, you don't actually care about this "community" bullshit. That's actually understandable and logical. Proprietary display driver and whatnot: since the super-high-quality FOSS-bazaar replacements envisioned by Richard "Rot beard" Stallman are obviously just not gonna happen, the situation is at hand. Do something about it instead of whining like a pussy.
So what does he propose? Oh yeah, simple:
#1: don't use the "delayed allocation" ext4 feature which is causing you problems. But of course this is just a theoretical possibility, we're the stable-api-nonsense guys so we just don't consider sane things viable! On to number two!
#2: "fix" everyone else! Yes! This is the really recommended Tso method. It's simply being: "application writers who are worried about file getting lost... really should use fsync." See, Ted, you've just sucked cock again. People who worry about their files don't use ext4 or even Linux anymore! You've just lost them, and you'll keep on losing. Who the fuck do you think you are? You've coded a filesystem that will never make it to any other mainstream UNIX due to its restrictive GPL license (see the fatal decision of AdvFS' GPL release), and you expect the whole world to adjust and recompile to your new little shitty feature? The only reason you've written your shitload of "code" in the first place is that Linux people did not use XFS in masses, precisely because the LinuxPC world is just not stable for this! Power outage? Get a grip on reality instead of FUD'ing, the world has tons more unexpected software halts than power loss!
People who are "actually worried about files" have been using VMS for the latest twenty-some years, and they are fully intent on keeping doing so. You'd love that, every I/O is synchronous and guaranteed. Painfully slow, but guaranteed. Plus it doesn't panic. Ever. (unnecessarily compared to the horse manure that is called Linux)
Idiot.
¹: if I didn't have 10 ocassions where a linux kernel panic+reboot zeroed out the
Yea right, another braindamaged filesystem for their broken-beyond-repair "Linux". Now that Reiser is in jail for slaughtering his whiny bitch wife, someone really had to step in and fill the gap in the long line of psychopathic linux developers.
Here's some background insight, the thing you never get from your local linux zealot. So there we are using write caches ever since some guys from IBM apparently invented the principle. Yea, using fast medium for temporary storage of data is good. Why? Go on, disable your CPU's and disk's caches and you'll immediately feel the pain. Write cache is among us for the latest thirty-some years, and everybody adores it. Let's fast forward those old cold war years to the year the next cold war started: 2009.
After the data annihilator¹ murderfs thankfully disappeared into the morning mist, some linux zeals created 2 (yes, two) new wheel-reinventing filesystems, completely oblivious to the fact that there's already a gazillion of non-working craplementations for better filesystems in their crappy kernel (see the HFS+ and Files-11 drivers). The other reason for coding new shit is that their shitty, limitation-stacked viral GPL license does not allow the inclusion of Sun's (otherwise fucking great) ZFS. Linux and "free" software is just way to cool.
So two new
EXPERIMENTAL filesystems in a kernel that similarly just gets even more experimental by every passing fucking day. Who would care huh?Apparently some Ubuntu users (yes, some insane people actually do install shit with a name like this) experienced that their usual data losses with the horseshit linux kernel that has just crashed actually got a lot worse than with last week's hypefs. All their open files were zeroed out and such. You think that's bad? Wait until the vietcong author Ted Tso steps in.
The usual stuff happens: there's a bug in linux, so they start a religious crusade around it. This chink actually gets the audacity to argue that everything in the previously super-hyped-überalles FOSS industry is just badly written code. Oh wait, he's actually mostly right on that, but this is not the reason.
His crusade bullshit has more attack vectors, as usual.
First, accusations of heresy: "This is what is required by POSIX" - he shouts in half-conscious religious zeal. The mentioning of "POSIX" is a callword, it makes the believers fall to their knees and murmur "thank thee Lord for POSIX" and all that crap, despite the well established fact that they have absolutely no idea of what it actually is, and how much braindamage does it contain (see Posix ACLs here earlier). Never mind the fact that POSIX doesn't say "don't use write caching or you will die a terrible death." It would be the stupidest thing.
Second, fucking over with people actually using his fs: "Application writers had gotten lazy." Yeah, right. Those lazy bastards have actually been happily using write-caching for the last few decades, completely ignoring the future possibility that some idiot might code such a crappy filesystem for such a crappy kernel, that using DRAM cache for making your miserable IT using life even somewhat bearable will result in catastrophic data loss.
Third, assuming similarities where there are none: "xfs has had delayed allocation for years" Yes. And you know where it came from: Silicon Graphics' IRIX. You know what are the major differences between an Octane and your PC? The Octane works and your PC is a stinking shithole. Not to mention IRIX, see, the guys using this (then-top-notch) architecture were actually busy creating Toy Story and other shit in the whole nineties, while you are doing what, ten years later? Busy waking up from nightmares where application writers don't explicitly fsync their data onto your latest work of "code" before their hardware burns away or their latest "enterprise" Bubuntu Kernel panics away in dying screams. All thanks to the swarm of politican coders like you, that were mass-injected during the late days of kernel 2.4, commiting
Stable-API-Nonsense.txt and other crap.Fourth, jumping on the usual bandwagon by saying "apparently some Ubuntu users are happy using proprietary Nvidia drivers". So yeah, Shitty Teddy argues that it's mostly "Ubuntu gamers", whose linuz box freezes and dies unexpectedly. Look, Ted, in case you haven't been paying attention to the gamer scene's latest years, linux is just not a gaming platform, you realize that?.. If you wanna game, you don't choose Linux. So the guys who you've just belittled are the sweet little linux switchers who would actually play something just a little more 2009-ish than Soko-Ban or Nethack while they are not possibly busy with deploying linux servers... using your filesystem? Well, from now on, probably not. Ideally, you should be hoping for these guys, but since you actually code linux for big money from big companies, you don't actually care about this "community" bullshit. That's actually understandable and logical. Proprietary display driver and whatnot: since the super-high-quality FOSS-bazaar replacements envisioned by Richard "Rot beard" Stallman are obviously just not gonna happen, the situation is at hand. Do something about it instead of whining like a pussy.
So what does he propose? Oh yeah, simple:
#1: don't use the "delayed allocation" ext4 feature which is causing you problems. But of course this is just a theoretical possibility, we're the stable-api-nonsense guys so we just don't consider sane things viable! On to number two!
#2: "fix" everyone else! Yes! This is the really recommended Tso method. It's simply being: "application writers who are worried about file getting lost... really should use fsync." See, Ted, you've just sucked cock again. People who worry about their files don't use ext4 or even Linux anymore! You've just lost them, and you'll keep on losing. Who the fuck do you think you are? You've coded a filesystem that will never make it to any other mainstream UNIX due to its restrictive GPL license (see the fatal decision of AdvFS' GPL release), and you expect the whole world to adjust and recompile to your new little shitty feature? The only reason you've written your shitload of "code" in the first place is that Linux people did not use XFS in masses, precisely because the LinuxPC world is just not stable for this! Power outage? Get a grip on reality instead of FUD'ing, the world has tons more unexpected software halts than power loss!
People who are "actually worried about files" have been using VMS for the latest twenty-some years, and they are fully intent on keeping doing so. You'd love that, every I/O is synchronous and guaranteed. Painfully slow, but guaranteed. Plus it doesn't panic. Ever. (unnecessarily compared to the horse manure that is called Linux)
Idiot.
¹: if I didn't have 10 ocassions where a linux kernel panic+reboot zeroed out the
O_RDONLY opened /sbin/cardmgr file on my PC laptop which I used before finally getting a Mac, then I had none at all. Reiserfs just rules all, huh.