You don't see net viruses for Mac, Linux or BSD because you can't guarantee there's enough machines visible from an infected machine to substain growth.
With Windows, it's easy. You're more likely than not to find another machine like yours. Hence the effort of writing a virus is merited to gain control or create disruption.
With everybody else, it's more a case of "Oh, I found XYZ machine, I wonder how I can hack into it". It's an abberation. If you did write a virus, it'd probably not be useful except in certain corporate settings where they standardize on an OSX config or Linux distro. So it'd be for a private endeavor, you wouldn't see it on the Internet anyway or hear about it @Sophos or Symantec...
Thus you treat it singularly and special case it. The black-hat hacker puts the IP addy in his little black book.
does not suffer from intense negative feedback as does the stock market.
I think part of the ease of predictability may have a little to do with the kind of protocols used to collision detection/TDMA in congested 802.11 nets. If they are suffciently simple a single node could outmanuver the others.
Some questions... What is the behavior of this algorithm as the number of enabled clients increases and the bandwidth demand of the clients exceeds the channel capacity? Does it degrade gracefully? Does it unfairly compete with non-enabled clients?
Parents who hate kids who complain about schoolwork.
"This work is too hard! I have to help my son with it and I dont' even understand it. He's a genius that deserves better grades so he can go to college. I'm going to join the PTA and bitch and moan."
This kid who was told she couldn't read a book because it was too hard for the other kids... what's stopping her from reading the book anyway? Why does she need the teacher's permission, to what, get extra credit? That's what the teacher didn't want to do. Too lazy to rearrange his gradebook, I surmise.
Set an X-Forwarded-For: some.ip.add.ress header. I think some of the more advanced ad servers will honor that. Also make sure to find a URL that will give you a new session key if you pretend you don't have one, and use that for a bunch of ads on the same network. This makes it look like a "person" viewing those ads over time. Expire the session keys after a day or two using the script (maybe use the cookie's expiry instead)
change the associations in the registry to point at your new app. (Start with CLSID under HKLM, also under Internet Explorer in software settings to change your view source command. Even a blind search and replace is probably safe enough.)
That's bullshit. You do the exact same number of fetches, adds and stores. Just different coefficients and different strides. Your neighborhood of fetches will be about the same, so cache usage should be about similar.
Wavelets are slower mostly because people haven't had years to fiund all the clever tricks (SIMD instructions on various targets, DFT->FFT-like optimizations for popular wavelets)
it doesn't matter what you run it on. It needs to be multithreaded to work on a mainframe as well. And you still want to limit thread-to-thread communication to avoid scaling issues.
Not that midframes and mainframes are useless then. They are much less latent than a cluster and can potentially provide more throughput.
Memory is nice to be able to replace while the system is hot, especially when you can tell if it's going bad ahead of time (ECC).
Some x86-compatible systems _can_ do that, but they definitely don't save you money over Suns, that's for sure.
CPU, PCI cards? Not likely to need. If you fry a PCI card inserting a cable, chances are more than just the card is fried (and maybe you don't know it yet). Hotplug is nice though for a different reason, dynamic reconfiguration.
Really, it's the disk drives that you care about being redundant/swappable. They tend to fail more frequently as they have moving parts, and they store what is most valuable.
I have this feeling that Solaris would run bitching on a modern Power5 platform, especially w/Hypertransport providing the IO.
For all I know they've tested it on a G5 and gotten it to work. I bet that'd scrap the Ultra V in a heartbeat, requiring them to reassess where their going with their hardware line.
There isn't anything in the UltraSPARC chip that makes it particularly suited for massive SMP or not, other than a lot of L3 cache, and not running insanely hot.
A dual CPU motherboard from Sun is specifically designed for that scale, it's not like there's a few errant unused pinouts that can make it magically scale to 8 CPUs.
The 8-way systems and larger have special crossbars and infrastructure that make them scale, but are incredibly expensive.
The UltraSPARCs themselves don't really have much to do with it. It's all about massive internal bandwidth.
You could acheieve that scaling on x86, but it'd be hella expensive, and you'd probably have to throw out the Intel MP specs for something more ambitious to get the scale-up and cache consistency without sacrificing performance. At which point someone says:
"I thought using x86 was supposed to save us money?"
And so that's why no one does it. (Well, except Unisys EMC, but they're off in their own little world, catering to those Windows shops with tons of money)
The SHA-4 (well, SH7750) which powers the Dreamcast is _very much_ a 32-bit architecture. 32-bit memory/IO addresses, 32-bit words, 32-bit FPU, 32-bit aligned. However it can dispatch 2 integer instructions simultaneously (like the pentium), while chewing on up to 2 FPU ops as well. This is how 32 + 32 + 32 + 32 = 128 bits is claimed.
Nothing could be further from the truth. Still, the architecture is very nice and allows for high throughput without needing a lot of cache, deep pipelines, or high clock speed.
It uses javascript to check the ads to make sure the size of the image is not too small (as replacement images often are) and the display properties are maintained (not set hidden with CSS).
It can't actually detect if an ad was replaced by a blank image by a proxy server, but it'll know if the ad is the wrong size.
Solution? More sophisticated ad blockers should attempt to match image size to a URL pattern by fetching it a few times and seeing what it gets back. Then it should autogenerate the replacement content with that size.
is capable of considering "consequences" in as much as it's highly narrow world view is concerned. It is fully capable of considering the possible future states that a state transition could imply, thus allowing it to also consider future possibilities in it's decision making process. That's why it's called goal-directed reasoning. You pick the action or state that gives you a higher probability of reaching certain goal states (desirable consequences).
It is probably self aware as well. If it couldn't examine it's own internal state, that it'd probably do very poorly considering the limited sensory input it has to work with. Being able to determine it's approximate coordinates and heading is a kind of self-awareness, IMHO.
So then, what is "self-aware" other than a vague notion of being in control and being able to self and not-self?
that whatever situational evaluation framework that robot has would value the commands and goals of another robot over whoever it gets it's programming from.
If it knows it's not it's master, what incentive does it have to regard it, other than to be aware of it to help it further it's intended goals?
The ability to self-generate is difficult because manufacturing a single robot now takes an inordinate amount of communication and understanding between different disciplines. Robots would have to be in control of everything already before they could start taking initiative to designing and building new ones without intervention or assistance.
So either we let it happen (if it's at all possible to progress to that point), or we "put the brakes on things" we were start getting remotely close to being there.
It has nothing to do with the code, and everything to do with what responsibilities we figure out how to automate/self-regulate.
The Roomba most definitely interprets it's sensory input to determine it's best course of action. It's not a sense->react machine. t can determine what needs to be done to complete it's tasks with very little sensory input by comparing it to it's own internal world model.
Are you looking for a more specific definition of "interpret" that outclasses what we are currently capable of with simpe machines?
Because samba + windows nt can be a lot more complicated especially now that Samba is in 3.0. This is especially true in environments with a an active directory domain. The number of session negotiation security options between he two systems is _staggering_ ^_^;;;
They have it, but aren't pushing it because they don't consider it fully tested. Put an order in your saved list, then call up the sales staff and ask them to do it. They will happily oblige. (I know because I deal with them frequently)
Slashdot Mirror
You know you want to know all about Nigritude Ultramarine . Come check it out!!! (Hint, this is funny)
I was hoping somebody would chime in saying NetBSD stole it from somewhere else. Something unexpected, like AcornOS or some such.
You mean like BSD init + rcorder, which predates Gentoo.
You don't see net viruses for Mac, Linux or BSD because you can't guarantee there's enough machines visible from an infected machine to substain growth.
With Windows, it's easy. You're more likely than not to find another machine like yours. Hence the effort of writing a virus is merited to gain control or create disruption.
With everybody else, it's more a case of "Oh, I found XYZ machine, I wonder how I can hack into it". It's an abberation. If you did write a virus, it'd probably not be useful except in certain corporate settings where they standardize on an OSX config or Linux distro. So it'd be for a private endeavor, you wouldn't see it on the Internet anyway or hear about it @Sophos or Symantec...
Thus you treat it singularly and special case it. The black-hat hacker puts the IP addy in his little black book.
does not suffer from intense negative feedback as does the stock market.
I think part of the ease of predictability may have a little to do with the kind of protocols used to collision detection/TDMA in congested 802.11 nets. If they are suffciently simple a single node could outmanuver the others.
Some questions...
What is the behavior of this algorithm as the number of enabled clients increases and the bandwidth demand of the clients exceeds the channel capacity? Does it degrade gracefully? Does it unfairly compete with non-enabled clients?
Parents who hate kids who complain about schoolwork.
"This work is too hard! I have to help my son with it and I dont' even understand it. He's a genius that deserves better grades so he can go to college. I'm going to join the PTA and bitch and moan."
Blah blah blah. You know the deal.
This kid who was told she couldn't read a book because it was too hard for the other kids... what's stopping her from reading the book anyway? Why does she need the teacher's permission, to what, get extra credit? That's what the teacher didn't want to do. Too lazy to rearrange his gradebook, I surmise.
Set an X-Forwarded-For: some.ip.add.ress header. I think some of the more advanced ad servers will honor that.
Also make sure to find a URL that will give you a new session key if you pretend you don't have one, and use that for a bunch of ads on the same network. This makes it look like a "person" viewing those ads over time. Expire the session keys after a day or two using the script (maybe use the cookie's expiry instead)
change the associations in the registry to point at your new app. (Start with CLSID under HKLM, also under Internet Explorer in software settings to change your view source command. Even a blind search and replace is probably safe enough.)
That's bullshit.
You do the exact same number of fetches, adds and stores. Just different coefficients and different strides. Your neighborhood of fetches will be about the same, so cache usage should be about similar.
Wavelets are slower mostly because people haven't had years to fiund all the clever tricks (SIMD instructions on various targets, DFT->FFT-like optimizations for popular wavelets)
it doesn't matter what you run it on. It needs to be multithreaded to work on a mainframe as well. And you still want to limit thread-to-thread communication to avoid scaling issues.
Not that midframes and mainframes are useless then. They are much less latent than a cluster and can potentially provide more throughput.
Memory is nice to be able to replace while the system is hot, especially when you can tell if it's going bad ahead of time (ECC).
Some x86-compatible systems _can_ do that, but they definitely don't save you money over Suns, that's for sure.
CPU, PCI cards? Not likely to need. If you fry a PCI card inserting a cable, chances are more than just the card is fried (and maybe you don't know it yet). Hotplug is nice though for a different reason, dynamic reconfiguration.
Really, it's the disk drives that you care about being redundant/swappable. They tend to fail more frequently as they have moving parts, and they store what is most valuable.
But he wrongly corrected the person, telling him it was 64-bit (IIRC), not 32.
I have this feeling that Solaris would run bitching on a modern Power5 platform, especially w/Hypertransport providing the IO.
For all I know they've tested it on a G5 and gotten it to work. I bet that'd scrap the Ultra V in a heartbeat, requiring them to reassess where their going with their hardware line.
Just speculation, mind you.
There isn't anything in the UltraSPARC chip that makes it particularly suited for massive SMP or not, other than a lot of L3 cache, and not running insanely hot.
A dual CPU motherboard from Sun is specifically designed for that scale, it's not like there's a few errant unused pinouts that can make it magically scale to 8 CPUs.
The 8-way systems and larger have special crossbars and infrastructure that make them scale, but are incredibly expensive.
The UltraSPARCs themselves don't really have much to do with it. It's all about massive internal bandwidth.
You could acheieve that scaling on x86, but it'd be hella expensive, and you'd probably have to throw out the Intel MP specs for something more ambitious to get the scale-up and cache consistency without sacrificing performance. At which point someone says:
"I thought using x86 was supposed to save us money?"
And so that's why no one does it. (Well, except Unisys EMC, but they're off in their own little world, catering to those Windows shops with tons of money)
The SHA-4 (well, SH7750) which powers the Dreamcast is _very much_ a 32-bit architecture. 32-bit memory/IO addresses, 32-bit words, 32-bit FPU, 32-bit aligned. However it can dispatch 2 integer instructions simultaneously (like the pentium), while chewing on up to 2 FPU ops as well. This is how 32 + 32 + 32 + 32 = 128 bits is claimed.
Nothing could be further from the truth. Still, the architecture is very nice and allows for high throughput without needing a lot of cache, deep pipelines, or high clock speed.
It uses javascript to check the ads to make sure the size of the image is not too small (as replacement images often are) and the display properties are maintained (not set hidden with CSS).
It can't actually detect if an ad was replaced by a blank image by a proxy server, but it'll know if the ad is the wrong size.
Solution? More sophisticated ad blockers should attempt to match image size to a URL pattern by fetching it a few times and seeing what it gets back. Then it should autogenerate the replacement content with that size.
is capable of considering "consequences" in as much as it's highly narrow world view is concerned. It is fully capable of considering the possible future states that a state transition could imply, thus allowing it to also consider future possibilities in it's decision making process. That's why it's called goal-directed reasoning. You pick the action or state that gives you a higher probability of reaching certain goal states (desirable consequences).
It is probably self aware as well. If it couldn't examine it's own internal state, that it'd probably do very poorly considering the limited sensory input it has to work with. Being able to determine it's approximate coordinates and heading is a kind of self-awareness, IMHO.
So then, what is "self-aware" other than a vague notion of being in control and being able to self and not-self?
Its quite tasty. Im sure your like it.
that whatever situational evaluation framework that robot has would value the commands and goals of another robot over whoever it gets it's programming from.
If it knows it's not it's master, what incentive does it have to regard it, other than to be aware of it to help it further it's intended goals?
The ability to self-generate is difficult because manufacturing a single robot now takes an inordinate amount of communication and understanding between different disciplines. Robots would have to be in control of everything already before they could start taking initiative to designing and building new ones without intervention or assistance.
So either we let it happen (if it's at all possible to progress to that point), or we "put the brakes on things" we were start getting remotely close to being there.
It has nothing to do with the code, and everything to do with what responsibilities we figure out how to automate/self-regulate.
The Roomba most definitely interprets it's sensory input to determine it's best course of action. It's not a sense->react machine. t can determine what needs to be done to complete it's tasks with very little sensory input by comparing it to it's own internal world model.
Are you looking for a more specific definition of "interpret" that outclasses what we are currently capable of with simpe machines?
And what class did you write this paper for?
Because samba + windows nt can be a lot more complicated especially now that Samba is in 3.0. This is especially true in environments with a an active directory domain. The number of session negotiation security options between he two systems is _staggering_ ^_^;;;
They have it, but aren't pushing it because they don't consider it fully tested. Put an order in your saved list, then call up the sales staff and ask them to do it. They will happily oblige. (I know because I deal with them frequently)