I put together a dead simple driver that only translates interrupts to blocking reads and poll/select. This allows me to eliminate hard polling loops and/or high latency sleepy loops. Beyond that, all of the I/O is done in user mode. This is for a 16-bit, I/O based ISA card, so it doesn't exactly strain the system performance. In testing I've found if you can use the ins() and outs() system calls to perform large chunks of I/O in big whacks, the perfomance is indistiguishable from I/O in kernel mode. Of course, this may only be true for the ISA bus.
However, if I do end up writing a kernel mode device driver, I will have no problem licensing under the GPL. I think some companies are just too paranoid for their own good.
There is certainly a lot of engineering that would otherwise be very tedious without our design software. But when I'm coming up with new ideas I don't go right off and simulate it on my computer. I estimate whether it is realistic in a ballpark sense (using my mind). Anyone who doesn't is wasting time. There is no reason to put any work into a solution if you can convince yourself it won't work just by thinking about it.
This estimating capability requires one to have a good intuitive undestanding of the underlying physics/logic of what you are working on. It also helps to have a large memory of magnitudes and orders. So one checks the high order terms, and not some lowly constant factor.
It seems like every few months someone bemoans the death of "real" engineering. But hey, idiots are born everyday, we just don't do a good enough job weeding them out before they get published.
Let see, not only will a company be tied to Apple for the OS, but also for the hardware. At least when you sign you soul away to M$ you have a wide choice of affordable and/or high performance hardware from a number of competitors to run xp on. Or you can just take what Apple chooses to give you?
That was a different article. The "sponsored" article was definately very pro media mogul. The Miller and Feigenbaum article is a more interesting and more critical analysis of copyright law in the digital age.
A lot of people make the mistake of thinking that the security of biometric systems relies on keeping the biometric data secret. I can steal your finger prints much easier ways than breaking into a biometric database, ala latent prints.
For there to be real security the system to which you are authenticating needs to verify two things: one that the biometric data provided is that of the person trying to authenticate, and second that the data is coming from a valid tamperproof biometric scanner.
The second part is certainly very tricky, basically you need very good scanners (which apparantly aren't common), and an infrastructure where you have a certificate for the known good scanners, and you only accept biometric data signed by one of those certificates. To scale better a PKI system could be employed with a certificate authority to manage the certificates of scanners.
Slashdot Mirror
I put together a dead simple driver that only translates interrupts to blocking reads and poll/select. This allows me to eliminate hard polling loops and/or high latency sleepy loops. Beyond that, all of the I/O is done in user mode. This is for a 16-bit, I/O based ISA card, so it doesn't exactly strain the system performance. In testing I've found if you can use the ins() and outs() system calls to perform large chunks of I/O in big whacks, the perfomance is indistiguishable from I/O in kernel mode. Of course, this may only be true for the ISA bus.
However, if I do end up writing a kernel mode device driver, I will have no problem licensing under the GPL. I think some companies are just too paranoid for their own good.
First buy the toys.
Start a strip of magnesium burning, then place it into a hollowed out piece of dry ice, and cover.
Not only does the magnesium it keep burning it does so with a violent blue flame.
There is certainly a lot of engineering that would otherwise be very tedious without our design software. But when I'm coming up with new ideas I don't go right off and simulate it on my computer. I estimate whether it is realistic in a ballpark sense (using my mind). Anyone who doesn't is wasting time. There is no reason to put any work into a solution if you can convince yourself it won't work just by thinking about it.
This estimating capability requires one to have a good intuitive undestanding of the underlying physics/logic of what you are working on. It also helps to have a large memory of magnitudes and orders. So one checks the high order terms, and not some lowly constant factor.
It seems like every few months someone bemoans the death of "real" engineering. But hey, idiots are born everyday, we just don't do a good enough job weeding them out before they get published.
Let see, not only will a company be tied to Apple for the OS, but also for the hardware. At least when you sign you soul away to M$ you have a wide choice of affordable and/or high performance hardware from a number of competitors to run xp on. Or you can just take what Apple chooses to give you?
It's quite a few bits to be sure, but certainly some spendy bits.
That was a different article. The "sponsored" article was definately very pro media mogul. The Miller and Feigenbaum article is a more interesting and more critical analysis of copyright law in the digital age.
As opposed to skittering around on metaphorical legs?
Running out of oil just means we have to move to nuclear.
A lot of people make the mistake of thinking that the security of biometric systems relies on keeping the biometric data secret. I can steal your finger prints much easier ways than breaking into a biometric database, ala latent prints. For there to be real security the system to which you are authenticating needs to verify two things: one that the biometric data provided is that of the person trying to authenticate, and second that the data is coming from a valid tamperproof biometric scanner. The second part is certainly very tricky, basically you need very good scanners (which apparantly aren't common), and an infrastructure where you have a certificate for the known good scanners, and you only accept biometric data signed by one of those certificates. To scale better a PKI system could be employed with a certificate authority to manage the certificates of scanners.
But first we'll need androids, town crier types, shouting out: #1 Sim City 10,000 #2 Photoshop 9 #3 ...
And a queue of people following behind.