However, you are right, the Linux OOM killer BLOWS.
I have encountered the same problem as you with spamassassin.
Fortunately, it is possible to disable the OOM killer. The tradeoff is that it is then possible for poorly-written programs (such as spamassassin) to exhaust memory and then not terminate; depending on the programmer's level of ineptness, they will either continue blithely on after a failed malloc producing unpredictable results, or sit there and spin forever waiting for the malloc to succeed.
I guess the only real solution is to run programs which are well written and nice to your system by design. I had an idea once about taking priority into account when in a OOM situation, but never wrote any code. The idea would be that lower priority processes would be killed in preference to higher priority ones. That way, if you know you have a misbehaving app like spamassassin, you can just nice it to 10 or something - that way it is harder to DoS the system, and it becomes the first choice for the OOM killer.
Pissed me off too, the night that it happened I had been drinking quite heavily at the local bar and wound up coming in to work at 11 pm totally plowed.
Yeah, I know that feeling. Fortunately I'm not employed as a sysadmin right now;)
Yeah, the statistic the article left out is that 6% of Mac users also decided to switch to PC citing lower hardware costs and more universal application support, rendering a net effect of zero.
The problem with Knoppix is that it installs its own custom set of packages, but them points the user to the Debian repositories for updates. This is just asking for trouble, and erases one of the biggest benefits of Debian's repository - a huge set of packages which have all been built and verified against each other.
So the tradeoff becomes do you want to have it in user space (where it would still vulnerable to DoS in this case) and sacrifice some speed
It's not just vulnerable to a DoS in terms of execution speed. The Linux kernel really loves killing inappropriate processes at inappropriate times when the OOM killer goes crazy. In a real micro-kernelish design, these processes would be granted exemption from the OOM killer.
Yay, pretty much like any distribution kernel - aside from drivers which have non-free components.
When a piece of hardware is installed, for non-hot-swapable hardware, it will automatically be recognized at the next boot and just work.
Wow. Just like PCI, USB, PnP, etc bus scans that Linux employs. Rocket science!
For hot-swapable hardware, a deamon is running that gets kicked whenever new hardware is attached (e.g. a USB storage device) and runs a user supplied script for that device.
Wow. Just like hotplug. You have utterly failed to amaze me with your post.
He's referring to the lack of a database abstraction layer like DBI for Perl, the point of which would be to prevent having to recode anything to switch to another RDBMS. ODBC is a bad suggestion because its performance sucks and there are many advanced features which it is too generic to expose.
A better suggestion would be to use the community written database abstraction classes for PHP. (I wrote one of these myself, disgusted with the PHP database APIs.)
All you need to do is keep your wallet in your pocket
That's a nice sentiment, but it's only part of the picture. If you do that, then the company doesn't even know that you were a potential customer, so nothing has been lost to them from their perspective.
The best approach is to keep your wallet in your pocket or buy from a competitor, and then contact a human at the company that you didn't buy from and give them a detailed explanation why you chose their competitor instead. This way you actually get attention, because from their perspective the sale came straight out of their pocket into a competitor's. If they won't listen to that sort of reasoning, they're going to sink anyway.
WHich bring up another angle. It's hardware. I doubt Intel has any 'obligations to others' as far as making a detailed description of the hardware workings available. This would allow someone to write GPL firmware.
You're being stupid. Even though that would be a good thing, if you would RTFA, that isn't what this is about. OpenBSD wants _freely distributable_ firmware, _not_ source code or anything else related to the firmware architecture.
So, that sounds more reasonable. What could keep Intel from doing this?
Third party patent licenses restricting free distribution
Third party software licenses restricting free distribution on the derived binary code
Fear of hardware cloners "dropping in" the firmware and selling a knockoff product
etc...
In short, there are a myriad of reasons why Intel would say no. If this is a problem for you, reverse the hardware and produce a free firmware, or make noise && vote with your feet.
On the one hand, Sun says their Java development process is the best way to go, because they maintain strict top-down control over it, and forking is disallowed. Then they paint this same picture of Linus's kernel tree to misrepresent all of Linux development, and somehow conclude that approach is bad.
This isn't an initiative aimed toward individuals. It is aimed towards academia and research institutions so that they can break down barriers between each other and cooperate, in response to a perceived growth in corporate control of biology research (through patents, closed source EULA'd software, and researcher mindshare).
Insightful my ass. This isn't about consumer or business software - this is about software produced by the scientific community, for use by the scientific community. By definition, this software has to be open to peer review and not subject to insidious licensing terms in order to be useful. What good is a piece of research software, for example, whose EULA assigns all patents derived from its use to the author's university?
Re:Common knowledge?
on
Hacking Vodka
·
· Score: 4, Funny
I carry a little flask around with me. Sometimes I take a discreet sip when I think nobody is looking.
A lot of people are expressing their frustration with doctors and their apparent "incompetence" or "arrogance" with respect to following established medical procedure instead of accepting a patient's self-diagnosis at face value.
The thing you have to remember about this is that doctors are being barraged with malpractice suits these days. The reason they follow established practices even when the patient becomes frustrated and insults them for it, is because if they deviate from that practice, then they have less defense in the case that the patient in question turns around later and files a malpractice suit when things don't turn out the right way, and if they are declared guilty of malpractice, their insurance company won't cover the losses if they were found to be deviant. Why would a MD invite trouble that way?
Of course, if all the sue-happy buffoons would chill out a little bit, maybe MDs would be more willing to go out on a limb, but unfortunately that's not the state of things these days.
I would justify a knockoff console solely on the basis of competition in a free market, besides the fact that it's completely legal to do so (since you didn't bother to cite a patent, I'll assume none actually exists outside your rhetoric, and see Sega v. Accolade to see how a clone console would be justified in potentially ripping parts of the BIOS).
Now if they are misrepresenting the item as a Nintendo product, I have a problem with that, because as a consumer, I am not able to make an informed decision about the quality of the item if its nature is withheld.
I don't find the arguments about Opera == evil because it is not free software very convincing for a few reasons.
You have alternatives were they to ever turn evil, so you're not supporting a monopoly
It uses standard network protocols and toolkits/APIs
It adheres closely to web standards (W3C recommendations and RFCs), any divergence thereof is regarded as a bug
It does not try to hide its configuration or storage in obfuscated file formats, so I can choose to use any other browser tomorrow and export all my bookmarks and such easily
It is cross-platform
It is reasonably well performing
It is well supported through newsgroups and a bug tracker
Upgrade and student prices are reasonable
I mean, doesn't Opera satisfy everything that a proprietary software company should do (Quality, Price, Support, interoperability/no lock-in)? What more could they possibly offer besides open sourcing it?
Re:What's wrong with freezing a drive?
on
Creative Data Loss
·
· Score: 1
How exactly would a trace which is contracting from the cold be more likely to make contact? It's _smaller_... perhaps you meant to say that cold tends to bring marginal IC's back into line temporarily.
Some of the Game Cube knock-off hardware is incredible. They make the game console smaller, prettier, more see-through-ish, higher quality and cheaper. I know that piracy is the issue here, but it's interesting to see what other designers can do to improve the console.
Exactly why would a Gamecube knock-off console be illegal or "piracy", provided:
- It was not referred to using any of Nintendo's trademarks
- Nintendo does not have patents covering the Gamecube
- They are not using Nintendo's BIOS except as permitted by the doctrine of minimum interoperability
Slashdot Mirror
Fortunately, it is possible to disable the OOM killer. The tradeoff is that it is then possible for poorly-written programs (such as spamassassin) to exhaust memory and then not terminate; depending on the programmer's level of ineptness, they will either continue blithely on after a failed malloc producing unpredictable results, or sit there and spin forever waiting for the malloc to succeed.
I guess the only real solution is to run programs which are well written and nice to your system by design. I had an idea once about taking priority into account when in a OOM situation, but never wrote any code. The idea would be that lower priority processes would be killed in preference to higher priority ones. That way, if you know you have a misbehaving app like spamassassin, you can just nice it to 10 or something - that way it is harder to DoS the system, and it becomes the first choice for the OOM killer.
Yeah, I know that feeling. Fortunately I'm not employed as a sysadmin right nowYeah, the statistic the article left out is that 6% of Mac users also decided to switch to PC citing lower hardware costs and more universal application support, rendering a net effect of zero.
The problem with Knoppix is that it installs its own custom set of packages, but them points the user to the Debian repositories for updates. This is just asking for trouble, and erases one of the biggest benefits of Debian's repository - a huge set of packages which have all been built and verified against each other.
You could always use grsecurity to disable running untrusted executables instead of trusting filesystem permissions to accomplish that.
2. MySQL has an entire community dedicated to helping people with the problems caused by its defects.
Sounds like such a book would be a perfect match for MySQL.
A better suggestion would be to use the community written database abstraction classes for PHP. (I wrote one of these myself, disgusted with the PHP database APIs.)
The best approach is to keep your wallet in your pocket or buy from a competitor, and then contact a human at the company that you didn't buy from and give them a detailed explanation why you chose their competitor instead. This way you actually get attention, because from their perspective the sale came straight out of their pocket into a competitor's. If they won't listen to that sort of reasoning, they're going to sink anyway.
So, that sounds more reasonable. What could keep Intel from doing this?
- Third party patent licenses restricting free distribution
- Third party software licenses restricting free distribution on the derived binary code
- Fear of hardware cloners "dropping in" the firmware and selling a knockoff product
- etc...
In short, there are a myriad of reasons why Intel would say no. If this is a problem for you, reverse the hardware and produce a free firmware, or make noise && vote with your feet.Which is it, guys? You can't have it both ways.
This isn't an initiative aimed toward individuals. It is aimed towards academia and research institutions so that they can break down barriers between each other and cooperate, in response to a perceived growth in corporate control of biology research (through patents, closed source EULA'd software, and researcher mindshare).
I appoint Darwin.
Insightful my ass. This isn't about consumer or business software - this is about software produced by the scientific community, for use by the scientific community. By definition, this software has to be open to peer review and not subject to insidious licensing terms in order to be useful. What good is a piece of research software, for example, whose EULA assigns all patents derived from its use to the author's university?
I carry a little flask around with me. Sometimes I take a discreet sip when I think nobody is looking.
HL1 had an OpenGL renderer. Why would it be so far-fetched for Valve to develop one for HL2?
Dunno. I've never run into a dry/cracked joint for which poking and prodding didn't suffice. I guess it depends on what you work on.
The thing you have to remember about this is that doctors are being barraged with malpractice suits these days. The reason they follow established practices even when the patient becomes frustrated and insults them for it, is because if they deviate from that practice, then they have less defense in the case that the patient in question turns around later and files a malpractice suit when things don't turn out the right way, and if they are declared guilty of malpractice, their insurance company won't cover the losses if they were found to be deviant. Why would a MD invite trouble that way?
Of course, if all the sue-happy buffoons would chill out a little bit, maybe MDs would be more willing to go out on a limb, but unfortunately that's not the state of things these days.
I spray freeze spray on suspect IC's. If there's a suspect joint, I just reflow it. Why bother with the spray?
Now if they are misrepresenting the item as a Nintendo product, I have a problem with that, because as a consumer, I am not able to make an informed decision about the quality of the item if its nature is withheld.
I mean, doesn't Opera satisfy everything that a proprietary software company should do (Quality, Price, Support, interoperability/no lock-in)? What more could they possibly offer besides open sourcing it?
How exactly would a trace which is contracting from the cold be more likely to make contact? It's _smaller_... perhaps you meant to say that cold tends to bring marginal IC's back into line temporarily.
- It was not referred to using any of Nintendo's trademarks
- Nintendo does not have patents covering the Gamecube
- They are not using Nintendo's BIOS except as permitted by the doctrine of minimum interoperability