Slashdot Mirror
2.6.13 Linux Kernel Released
LynuxFre@k writes "Linux Torvalds announced the release of the 2.6.13 Linux kernel. He noted that there was a major change to the x86 PCI code, and that while all bugs from the change were believed to be found during the release candidate phase, it's possible that some devices may have problems. From this release on, it is intended that major changes only be merged into the kernel within two weeks after a major release. The rest of the time will be spent fixing bugs, with the goal of both increasing overall stability and decreasing the amount of time between major releases. Download the latest Linux kernel from a kernel.org mirror."
I can't really understand why so many people have a problem with the current policy and the policy changes.
What exactly is wrong with refining the development process?
It's not THAT bad. Revision control system has been changed twice during the lifetime of the kernel. Developement-method has been changed once, and now that method is simply being tweaked a bit. And what do you care how they develop the kernel? Are you are kernel-developer?
Or how would you like them to do it? "We will do things this way, and by god, we will do it like this untill the end of time! Even if better ways of doing this comes along, we will not change our ways!"
Lesbian Nazi Hookers Abducted by UFOs and Forced Into Weight Loss Programs - -all next week on Town Talk.
Industry standard 3d, compositing and editing tools all run under linux which is the natural progression because of their IRIX legacy.
I've also done some DTP under linux but that probably wasn't professional, since I didn't just bang a series of poorly masked raster images together like most 'professional' agencies we dealt with.
Does this make me a linux fanatic?
A) Fine. So we'll distribute one binary version of the kernel. It won't work as fast. It'll be hundreds of megs in size. It will take ages to load as it checks for every single known piece of hardware. You've just lost all the speed/memory advantage of having a tailored kernel. Alternatively, it'll come with a hundred modules. It'll also never be x64 optimised or, for that matter, able to work on every machine (some options crash some machines, while the opposite options may crash others) and APM/ACPI will NEVER work on some machines.
B) Fine. You come up with a GUI that can allow me to find files modified on the second Tuesday of every month between May 1, 1946 and June 27, 1978, which contain the words 'secret' and 'report' within 26 characters of each other, sort them by date, and replace any occurence of the word 'anchovie' by 'dead bug'. Some things GUI's just cannot do, some things GUI's do that are just command line interfaces in a fancy coloured textbox, some things GUI's can do once in the time that someone who knows the command can do twenty times.
Secondly, how do you expect a GUI to be able to do stuff like modify computer internals safely? Windows answer to this is usually that settings won't take effect until the next reboot, which makes your computer *stop all it's work* until it's done. X can be restarted with a single keystroke to have the same effect. Maybe a couple of command line edits in between but meanwhile none of your users have been disconnected, no programs have stopped doing what they were supposed to be doing.
Command-lines are not for the faint-of-heart. Then again, last time I touched the command line on my own Linux desktop (not counting other machines that are cmd-line only via SSH) was to run LILO - not something that a "desktop doughnut" should be doing. You obviously have either different ideas of what you should be doing on a normal desktop machine or have not found out how to do them GUI-wise. By the same token, Windows should never expect me to recover in safe mode, or via recovery console, or by running any batch commands ever. Fine for the ordinary desktop user because it very rarely does. Not fine for a power user. An ordinary desktop user wouldn't even notice if you ran a Windows GUI on a Linux machine.
C) Man pages can be a pain in the arse (make it compulsory to include enough examples to demonstrate every option!). HOWTO's are not always up-to-date. Forums are, pretty much, for people who want to know how to install this Linux thing they downloaded. Then again... how much documentation do you get with Windows?
A small booklet showing you how to use a mouse to point at the various icons. An online help system that, even with it's wizard-style help for some items, is next to useless if you don't know the terms to look for (I work support for six schools... that's about 60-100 staff and a few thousand pupils. I have NEVER seen or heard of anyone even bother to try using Windows Help or Help inside ANY program because it's never been useful to them). Annoying dogs, wizards, paperclips that people want me to TURN OFF for them because they can't figure out how.
That's surely Linux 0-0 Windows in terms of help.
If you're an advanced user, you've got to be comfortable with the command-line. I carry a USB key full of cmd-line utils and use them almost every day on Windows and Linux. It's amazing how much quicker "Start, Run, Cmd, ipconfig" is than navigating that poxy GUI network settings. And while I'm there, doing "route print" is the ONLY way to discover Windows network routes.
Anyone who's not going to set up networks or advanced stuff (i.e. users), or home users shouldn't ever NEED to worry about the command line on either OS. And they don't. They pick a distro like Lindows and once the installation is complete, they never see it again. Or they have a decent desktop set up and then never see the command-line again. You, however, are on the border. You are trying to do stuff that NEEDS a command line, stuff that's beyond a GUI point-and-click.
Yes, unfortunately I have the same experience. I used to rely on the vanilla Linux kernel tree being rock solid. Now I feel I must stay many months behind in order to avoid potentially catastrophic problems. Considering the number of bug fixes, particularly w/ regard to security, that can show up in new kernel releases, it's not an ideal situation: you're damned if you do, and you're damned if you don't.
I've recently had networking go south when packets were being written to localhost. Some adaptec scsi stuff was recently messed up, apparently now fixed in 2.6.13 - but no way I'm going to try it until it's been out for a while. I've seen problems with quotas in combination with ext3. I recently started experiencing connection tracking weirdnesses with an iptables setup I've used at home for probably a couple of years. I've seen versions where network latencies would grow ever so slowly until they reached a critical threshold that sent my server(s) spiraling into oblivion. Yes, I file bug reports. Yes, problems get fixed. But at the same time, new ones show up. Sometimes bad ones.
I've become accustomed to rebooting Windows to fix problems, but that's exactly why I use Linux - because it was rock solid. I won't say that anymore, and it bums me out big time. I like new shiny objects too, but not at the expense of stability. Especially not on servers, which is where Linux has made the most headway.
The problem with the current versioning system is that even if there is a bug-fix only decimal release, and even if there is only a two week window to introduce new features, the bug fixing won't get done. Why? Because new features are more fun than fixing bugs. Even if I can't submit a new feature until several months from now, that doesn't mean I won't work on it in leiu of fixing bugs.
Linus should freeze the 2.6 kernel series against *any* new features at all, for a period of about a year. All work should be on increasing stability, ironing out bugs, improving device drivers, and other such menial housekeeping. The kernel contributers who really buck up, get to work, and help with this effort should get big karma bonuses from Linus. Those who hang back and work on their own thing should be pushed down a level in future kernel submission evaluations.
Sorry to be so negative, but I really hope this gets better. I'm a huge fan, but I have been wasting way too much time lately dealing with problems that end up being way beyond my control. When there is a problem with my systems, I want it to be my fault, because then I can do something about it.
Not me.
But it's very hard to generalize from one person's experience to any general "recent trend of unstability." Most of the bugs are in drivers, so people's experiences tend to be highly dependent on exactly which hardware they have.
--Bruce Fields
These discussions along with final reason for pushing udev were concentrated on devfs weaknesses rather than udev superiority.
udev pushes all the device naming policy to userspace. Moving policy stuff to userspace is something that linux developers (and hackers of other OSes too) love because it's a much better design. That was the main reason for udev.