"Testers" (Software Design Engineer in Test) at MS typically are coders, writing software that will test programs. This way a huge library of tests are built up over time so repetition is avoided. If you make changes to the software, you still have the old tests which can be rerun to verify things still work. I think it was in a channel 9 interview I heard where they said they were aiming for a 48 hour verification period which could completely put an application through its paces. Also, they aim for about 70% code coverage through automated testing. This number is likely to go up since some percentage of the 30% is legacy code which is mostly being left alone.
It'd still be nice to hear from the poster on his personal breakdown and thoughts.
Very true. It's hard. But clearly they're trying to split things or we wouldn't be having this conversation. I agree that you'd have to commit to doing something one way or the other and rebalancing might be slow. But I don't think a typical use case would be jumping all over the map. We have to assume that there's some level of predictability. I suppose it might be easier to envision of an effective load balancing approach if this scaled to a higher number of cores per chip and used and didn't balance between modes within a single core. Lets say in five or ten years we have a 16 core CPU. Maybe it'll make sense to put 15 cores against a CPU hogging task and assign a ton of IO bound threads to the last. This would avoid slowing down the main execution with context switching between processes, which is usually considered to be costly. Context switching might be even more of a pain with reverse multithreading. I couldn't really say.
See my other comment for a little more idea tossing on this.
I was citing where my idea came from. I certainly did NOT suggest that a GPU and CPU were the same thing. Thanks for that. I suppose I meant core rather than CPU as this whole article is about a multi-core processor, not two GPUs. But let me be a little more descriptive. You've got three cases in the simple GPU example which sparked this for me, although one isn't really practical.
1) 100% graphics 2) A balance between graphics and physics (unlikely to be 50:50) 3) 100% physics (seems unlikely in practice)
So we have two GPUs and one or two tasks. Take a 60:40 graphics:physics situation in case 2. Ignoring overhead, nVidia would aim to use 100% of one GPU on graphics and 20% graphics, 80% physics on the second GPU. I'm only considering a case where CPUs could offer a similar kind of load balancing. In case 1, they operate in a manor similar to how AMD is proposing, where the single task is split between two cores. In case 2, they operate independently.
Lets say it takes a full half second for a dual core CPU to switch from dual mode to reverse multithreading mode. But let's say that the first core can continue working without noticeable interruption. An OS's dispatcher could be implemented to pick modes. If you've got 50 idle processes/threads and one CPU-bound thread, use reverse multithreading. If you have a sufficient number of 'active enough' threads to bring each CPU to 100% independently, do that.
This would work under the reasonable (imho) assumption that a typical machine does not flip between many active processes and one single active process rapidly. Low load would probably run faster in a reverse multithreading mode. If multiple applications happen to spike at the same time, multiprocessing can handle that just as easily. Or just idle the 2nd core entirely for power consumption.
That's doing it without a smooth load balancing but would be effective in most situations where you're going to have a clear choice. Doing a smooth load balancer would certainly take more silicon and may or may not make sense. Kinda hard to say when we haven't heard how reverse multithreading is being implemented.
You completely failed to explain to me what is different between the two cases or why specifically this doesn't work. I'm throwing ideas around here so if you have anything productive to say, feel free to add constructively, sir anon coward...
It might be interesting if they took this idea in a slightly different direction. Set it up so the OS detects two CPUs. But, when the OS fails to utilize both CPUs effectively, allow the idle CPU to take some of the active CPU's load. I'm taking this idea from nVidia working on load balancing between graphics and physics in a SLI setup. So in this case the OS gets the best of both worlds, the ability to break tasks off to each CPU and a free boost when it's stuck with a single cpu-limited thread.
It was mentioned in a networking class I took that token ring is used in aircraft due to their predictability. So by extension, any real-time system would be a good candidate for a token ring setup. It lets you prove that you have adequate bandwidth for the situation. Ethernet is at heart still random, no matter how much bandwidth you have.
Not entirely true. It might be good to point out that Vista will only allow some updates to be made by Microsoft. So no amount of admin privileges will let you change certain aspects of the OS. Presumably the update has to be signed by them for it to be accepted. This would be for core portions of the kernel, etc. It's specifically designed to combat rootkit style attacks. At least, that's the theory as I understand it.
"Endowment: $28.8 billion Total grant commitments since inception: $9,259,952,552 Total 2004 grant payments: $1,255,762,783"
A lot of money for a few PR stunts most people never hear about... And CEOs, etc are held BY LAW to do what will make the most amount of money for the shareholders. Microsoft's money isn't Bill's money. If all of his empire away now, he loses Microsoft. That'd be a pretty big thing to ask of someone. And hell, Bill stepped down from the CEO role to focus more on the development end of things. Balmer seems to have more fun in the business end of things anyways. Hey, I'm not calling the guy a saint but he'll do his fair share in the end.
When I read the exploit description, it relies on specific memory offsets. One working case is running the vulnerability with just one IE window loaded. If you have other stuff open, the desired offset for injected code might not be accessible. I'm not sure how much stuff you actually need open before the exploit will be rendered useless... So load up a few dozen apps before browsing or just maybe don't use IE?:)
And no, I don't consider disabling JavaScript as any real solution. We're trying to advance the web, aren't we?
Rescues can be dangerous. Alerting a weak swimmer near to someone fighting for their life would essentially result in two people drowning. The weak swimmer would approach the drowning victim, be grabbed, and pulled underwater as the drowning victim pulls/pushes themselves upwards for air. That's how I see a conscious scenario working out.
In a scenario like this one, pulling them up improperly would likely result in a lot of extra water in the lungs. This makes resuscitation significantly more difficult. A proper rescue would cover the mouth and nose and tilt the face downwards as they're raised to the surface.
If the victim was injured in a such a way that a spinal injury was incurred, having an untrained patron grabbing them could result in paralysation.
Untrained patrons may also find themselves ill-prepared to deal with other conditions such as seizures.
Not to mention the legal ramifications of this. If a patron was at all injured or traumatized by being in a situation where the facility placed a moral obligation for them to help on their shoulders, there's the potential for an ugly law-suit.
All in all, I think alerting the lifeguards to these alerts is adequate. There should always be lifeguards available to respond to an emergency. When there is limited staffing available to respond to emergencies, the pool is closed. That's standard. Bring public into a sketchy situation is something I would, as a lifeguard, be very hesitant to see.
Just keep in mind not everyone can swim. Not everyone lives near a beach. Not everyone is from a part of the world where swimming is particularly common. Many aquatic dangers are not obvious if you haven't grown up around water. Work in Vancouver for a few years and you'll get a pretty good idea of how swimming abilities range in various countries. I'm not bashing them. I'm just saying swimming abilities and water safety skills range greatly.
Variable-depth pools have grated floors. There are small holes so as the floor is raised or lowered, the water passes through. So with less depth, there is more water beneath the floor. These pools are great since you can teach young kids in them, then a few minutes later, an older group for a fitness class. Many pools also have a variable depth portion at one end of a 50m length. In that case there is a bulkhead that raises and lowers to block off the variable-depth part when it is not flush with the rest of the pool. The bulkheads are typically heavy and raise to the surface by having air pumped into them.
Did anyone at microsoft actually say they were developing this into a product? Not that I'm aware of. People are calling it vapourware but was there ever any claim made stating it will be developed?
Some guy at Microsoft's UK research branch (iirc) got interested in file distribution and wrote a paper on how it could potentially be improved upon. Bram even said that this was one of the better papers, despite a few major flaws. Now the poor guy's being accused of being a part of a microsoft conspiracy to take down bit torrent? Give me a break. If microsoft wanted to take bit torrent's market share, they'd package a complete product, not release an academic style paper. Maybe one day there will be competition from microsoft but that day is not yet today.. And hey, even if they are developing a full product, isn't it good news that they're throwing ideas around and thinking them through in an intelligent fashion? So what if they're not real world yet? You have to start somewhere. It doesn't matter if you're banging out code for a shakey first version over the period of three days in your basement or if you're writing papers and doing simulations to get some ideas..
Alright, end rant... just calm the flame wars guys. You'd think writing a paper is a crime.. the media and these guys blogging are who's blowing this out of proportion.. keep that in mind.
Windows prefetches some data on bootup. The list of what is prefetched is based partially on set settings and partially on learned user behaviour. What processes you have running after bootup don't reflect the prefetching optimizations windows makes. You can get a vague idea of what's being prefetched by browsing to C:\WINDOWS\Prefetch.
How about bring along some of the videos from the DARPA Challenge and explain that computer science is about figuring out problems like path-finding and vision and software engineering is about how to build bigger and more complex systems out of all this smaller components. In this case you've got tons of different near-sci-fi parts of the vehicles you could tell the kids about as well as putting it all together into an AI that can drive a 175-mile course. "That's like having your car drive you from here to ____!" There's tons of neat stuff to inspire kids plus tons of video footage to base your presentation on.
Look around at various machine learning topics. There are many many extremely cool projects described on university webpages.
I've been interested in computers for a long time but machine learning has been one of few things that has actually made me go "damn! That's cool!" in the last couple years. Uni was getting a little boring until I took a course on it and was absolutely blown away with what we were taught to do.
Those topics are covered in machine learning courses actually. Pretty well every curriculum I've seen has this sort of material in it. I've taken one such course and found it extremely interesting. The material is so incredibly useful though that I don't think anyone would call it a course on anti-spam. The fact is that these learning techniques have far-reaching benefits beyond classifying email.
But the really interesting stuff I think is mostly left for grad students who specialize a little in the topic.
Safari has been mentioned (though I've never used it). But, for those who are curious or might find this helpful, there is an optional firefox spell checker extension that can be installed from here...
It has been around for a while and was recently accepted into the Mozilla trunk. Don't think it's just a cheap hack.
Alternatively, quick search turned this up for IE. With the right design, something as simple as a spell check should be quick and painless. Some IM clients even have a live spell check built in which can be of great use. Nothing can beat the live spell checker for helping you pick up on your own problem words. Or do what I recently did and parse through the last five years of IRC chat logs to find your most common mistakes. That was ugly, let me tell you! I can't make the rest of the world love my crappy spelling so fixing it is all I've got. But at least after doing that I felt a little better about myself.
"and even DirectX, are all, in my opinion, overly integrated and give hackers too much access to core PC functions."
Wasn't that the point of DirectX? To provide more direct access to hardware for purposes such as graphics? That's why you couldn't play games in windows 3.1 and had to use DOS; you couldn't get at the hardware. The trick is how to do it safely.
It sounds like this guy's taking one idea and applying it to everything here. There are some things that do need kernel integration for performance reasons. As for doing it with your browser, I don't really see the point. Integrate all the browser's IO by way of tcp/ip, win32, directx, etc and leave all the rendering engine out of kernel space. But microsoft is probably doing exactly that for the most part. It's hard to say what's part of the windows operating system (kernel) and what just ships with it. Sure there's a lot of stuff that you can't uninstall but that doesn't mean that stuff isn't bound by the same rules that an application like firefox is bound by. It's pretty hard to say how integrated IE really is or if most of these bugs are just there because MS ships when stuff is just "good enough.
Well, they were taking over marketshare like crazy (even before being bundled) and were inventing new technologies faster than the rest of the company could keep up. So they shut them down to ensure that web apps wouldn't take over the rest of their business. They certainly didn't stop developing IE because they couldn't afford it. I don't know about you, but grabbing a few guys that can eat away at the rest of Microsoft's business doesn't sound like such a bad idea to me;)
I wonder if this reason would have anything to do with back to school. The week ending at Aug 21st sounds about right. Are kids deciding more and more on AMD boxes whereas older users are sticking with Intel? I know a few people who've gone purely Intel after the days of the cyrix or K6 chips.... I can't really blame them either. Those were horrible gaming chips.
In any case I'm seeing a lot of shiny new AMD boxes around my campus and it's alright by me.
I've never been a fan of c++ add-ons. You can't build a newbie-friendly community when your mods are potentially unsafe. Unreal, for example, allows c++ add-ons but it's considered very poor form. The big MSU contest bans entries with native code. These days I don't see c++ as a pro. When I see that I think of it more as a "no scripting language" con. I'd hate to see mods begin to bundle in other software, for example. I'd just much rather keep things in a safe sandbox. I certainly don't have the time or desire to review source code before playing a mod.
If you can add support to mozilla, there's no reason why you can't add support to IE via plug-in. Well, that's probably not entirely true. Maybe someone who's familiar with what can and can't be done via IE plugins could comment?
Slashdot Mirror
"Testers" (Software Design Engineer in Test) at MS typically are coders, writing software that will test programs. This way a huge library of tests are built up over time so repetition is avoided. If you make changes to the software, you still have the old tests which can be rerun to verify things still work. I think it was in a channel 9 interview I heard where they said they were aiming for a 48 hour verification period which could completely put an application through its paces. Also, they aim for about 70% code coverage through automated testing. This number is likely to go up since some percentage of the 30% is legacy code which is mostly being left alone.
It'd still be nice to hear from the poster on his personal breakdown and thoughts.
Very true. It's hard. But clearly they're trying to split things or we wouldn't be having this conversation. I agree that you'd have to commit to doing something one way or the other and rebalancing might be slow. But I don't think a typical use case would be jumping all over the map. We have to assume that there's some level of predictability. I suppose it might be easier to envision of an effective load balancing approach if this scaled to a higher number of cores per chip and used and didn't balance between modes within a single core. Lets say in five or ten years we have a 16 core CPU. Maybe it'll make sense to put 15 cores against a CPU hogging task and assign a ton of IO bound threads to the last. This would avoid slowing down the main execution with context switching between processes, which is usually considered to be costly. Context switching might be even more of a pain with reverse multithreading. I couldn't really say.
See my other comment for a little more idea tossing on this.
I was citing where my idea came from. I certainly did NOT suggest that a GPU and CPU were the same thing. Thanks for that. I suppose I meant core rather than CPU as this whole article is about a multi-core processor, not two GPUs. But let me be a little more descriptive. You've got three cases in the simple GPU example which sparked this for me, although one isn't really practical.
1) 100% graphics
2) A balance between graphics and physics (unlikely to be 50:50)
3) 100% physics (seems unlikely in practice)
So we have two GPUs and one or two tasks. Take a 60:40 graphics:physics situation in case 2. Ignoring overhead, nVidia would aim to use 100% of one GPU on graphics and 20% graphics, 80% physics on the second GPU. I'm only considering a case where CPUs could offer a similar kind of load balancing. In case 1, they operate in a manor similar to how AMD is proposing, where the single task is split between two cores. In case 2, they operate independently.
Lets say it takes a full half second for a dual core CPU to switch from dual mode to reverse multithreading mode. But let's say that the first core can continue working without noticeable interruption. An OS's dispatcher could be implemented to pick modes. If you've got 50 idle processes/threads and one CPU-bound thread, use reverse multithreading. If you have a sufficient number of 'active enough' threads to bring each CPU to 100% independently, do that.
This would work under the reasonable (imho) assumption that a typical machine does not flip between many active processes and one single active process rapidly. Low load would probably run faster in a reverse multithreading mode. If multiple applications happen to spike at the same time, multiprocessing can handle that just as easily. Or just idle the 2nd core entirely for power consumption.
That's doing it without a smooth load balancing but would be effective in most situations where you're going to have a clear choice. Doing a smooth load balancer would certainly take more silicon and may or may not make sense. Kinda hard to say when we haven't heard how reverse multithreading is being implemented.
You completely failed to explain to me what is different between the two cases or why specifically this doesn't work. I'm throwing ideas around here so if you have anything productive to say, feel free to add constructively, sir anon coward...
It might be interesting if they took this idea in a slightly different direction. Set it up so the OS detects two CPUs. But, when the OS fails to utilize both CPUs effectively, allow the idle CPU to take some of the active CPU's load. I'm taking this idea from nVidia working on load balancing between graphics and physics in a SLI setup. So in this case the OS gets the best of both worlds, the ability to break tasks off to each CPU and a free boost when it's stuck with a single cpu-limited thread.
It was mentioned in a networking class I took that token ring is used in aircraft due to their predictability. So by extension, any real-time system would be a good candidate for a token ring setup. It lets you prove that you have adequate bandwidth for the situation. Ethernet is at heart still random, no matter how much bandwidth you have.
Not entirely true. It might be good to point out that Vista will only allow some updates to be made by Microsoft. So no amount of admin privileges will let you change certain aspects of the OS. Presumably the update has to be signed by them for it to be accepted. This would be for core portions of the kernel, etc. It's specifically designed to combat rootkit style attacks. At least, that's the theory as I understand it.
Another perspective maybe not based on specific single charities?
e et/default.htm
http://www.gatesfoundation.org/MediaCenter/FactSh
"Endowment: $28.8 billion
Total grant commitments since inception: $9,259,952,552
Total 2004 grant payments: $1,255,762,783"
A lot of money for a few PR stunts most people never hear about... And CEOs, etc are held BY LAW to do what will make the most amount of money for the shareholders. Microsoft's money isn't Bill's money. If all of his empire away now, he loses Microsoft. That'd be a pretty big thing to ask of someone. And hell, Bill stepped down from the CEO role to focus more on the development end of things. Balmer seems to have more fun in the business end of things anyways. Hey, I'm not calling the guy a saint but he'll do his fair share in the end.
I wish :)
Isn't Bill the guy who's giving away all his money to charities to fight diseases like aids or world hunger? hmmm :P
When I read the exploit description, it relies on specific memory offsets. One working case is running the vulnerability with just one IE window loaded. If you have other stuff open, the desired offset for injected code might not be accessible. I'm not sure how much stuff you actually need open before the exploit will be rendered useless... So load up a few dozen apps before browsing or just maybe don't use IE? :)
And no, I don't consider disabling JavaScript as any real solution. We're trying to advance the web, aren't we?
Rescues can be dangerous. Alerting a weak swimmer near to someone fighting for their life would essentially result in two people drowning. The weak swimmer would approach the drowning victim, be grabbed, and pulled underwater as the drowning victim pulls/pushes themselves upwards for air. That's how I see a conscious scenario working out.
In a scenario like this one, pulling them up improperly would likely result in a lot of extra water in the lungs. This makes resuscitation significantly more difficult. A proper rescue would cover the mouth and nose and tilt the face downwards as they're raised to the surface.
If the victim was injured in a such a way that a spinal injury was incurred, having an untrained patron grabbing them could result in paralysation.
Untrained patrons may also find themselves ill-prepared to deal with other conditions such as seizures.
Not to mention the legal ramifications of this. If a patron was at all injured or traumatized by being in a situation where the facility placed a moral obligation for them to help on their shoulders, there's the potential for an ugly law-suit.
All in all, I think alerting the lifeguards to these alerts is adequate. There should always be lifeguards available to respond to an emergency. When there is limited staffing available to respond to emergencies, the pool is closed. That's standard. Bring public into a sketchy situation is something I would, as a lifeguard, be very hesitant to see.
Just keep in mind not everyone can swim. Not everyone lives near a beach. Not everyone is from a part of the world where swimming is particularly common. Many aquatic dangers are not obvious if you haven't grown up around water. Work in Vancouver for a few years and you'll get a pretty good idea of how swimming abilities range in various countries. I'm not bashing them. I'm just saying swimming abilities and water safety skills range greatly.
Variable-depth pools have grated floors. There are small holes so as the floor is raised or lowered, the water passes through. So with less depth, there is more water beneath the floor. These pools are great since you can teach young kids in them, then a few minutes later, an older group for a fitness class. Many pools also have a variable depth portion at one end of a 50m length. In that case there is a bulkhead that raises and lowers to block off the variable-depth part when it is not flush with the rest of the pool. The bulkheads are typically heavy and raise to the surface by having air pumped into them.
Did anyone at microsoft actually say they were developing this into a product? Not that I'm aware of. People are calling it vapourware but was there ever any claim made stating it will be developed?
Some guy at Microsoft's UK research branch (iirc) got interested in file distribution and wrote a paper on how it could potentially be improved upon. Bram even said that this was one of the better papers, despite a few major flaws. Now the poor guy's being accused of being a part of a microsoft conspiracy to take down bit torrent? Give me a break. If microsoft wanted to take bit torrent's market share, they'd package a complete product, not release an academic style paper. Maybe one day there will be competition from microsoft but that day is not yet today.. And hey, even if they are developing a full product, isn't it good news that they're throwing ideas around and thinking them through in an intelligent fashion? So what if they're not real world yet? You have to start somewhere. It doesn't matter if you're banging out code for a shakey first version over the period of three days in your basement or if you're writing papers and doing simulations to get some ideas..
Alright, end rant... just calm the flame wars guys. You'd think writing a paper is a crime.. the media and these guys blogging are who's blowing this out of proportion.. keep that in mind.
Windows prefetches some data on bootup. The list of what is prefetched is based partially on set settings and partially on learned user behaviour. What processes you have running after bootup don't reflect the prefetching optimizations windows makes. You can get a vague idea of what's being prefetched by browsing to C:\WINDOWS\Prefetch.
How about bring along some of the videos from the DARPA Challenge and explain that computer science is about figuring out problems like path-finding and vision and software engineering is about how to build bigger and more complex systems out of all this smaller components. In this case you've got tons of different near-sci-fi parts of the vehicles you could tell the kids about as well as putting it all together into an AI that can drive a 175-mile course. "That's like having your car drive you from here to ____!" There's tons of neat stuff to inspire kids plus tons of video footage to base your presentation on.
Look around at various machine learning topics. There are many many extremely cool projects described on university webpages.
I've been interested in computers for a long time but machine learning has been one of few things that has actually made me go "damn! That's cool!" in the last couple years. Uni was getting a little boring until I took a course on it and was absolutely blown away with what we were taught to do.
Those topics are covered in machine learning courses actually. Pretty well every curriculum I've seen has this sort of material in it. I've taken one such course and found it extremely interesting. The material is so incredibly useful though that I don't think anyone would call it a course on anti-spam. The fact is that these learning techniques have far-reaching benefits beyond classifying email.
But the really interesting stuff I think is mostly left for grad students who specialize a little in the topic.
They were waiting for our dollar to catch up so they could save costs by marketting only 99 cent songs ;)
Report me? haha. Knowing my ISP, they'd probably increase my bandwidth.
I hope the guys who attacked Lycos are getting hit hard by their service. Keep it up Lycos! You're obviously hitting a nerve.
http://spellchecker.mozdev.org/
It has been around for a while and was recently accepted into the Mozilla trunk. Don't think it's just a cheap hack.
Alternatively, quick search turned this up for IE. With the right design, something as simple as a spell check should be quick and painless. Some IM clients even have a live spell check built in which can be of great use. Nothing can beat the live spell checker for helping you pick up on your own problem words. Or do what I recently did and parse through the last five years of IRC chat logs to find your most common mistakes. That was ugly, let me tell you! I can't make the rest of the world love my crappy spelling so fixing it is all I've got. But at least after doing that I felt a little better about myself.
But seriously, learn how to type ;)
"and even DirectX, are all, in my opinion, overly integrated and give hackers too much access to core PC functions."
Wasn't that the point of DirectX? To provide more direct access to hardware for purposes such as graphics? That's why you couldn't play games in windows 3.1 and had to use DOS; you couldn't get at the hardware. The trick is how to do it safely.
It sounds like this guy's taking one idea and applying it to everything here. There are some things that do need kernel integration for performance reasons. As for doing it with your browser, I don't really see the point. Integrate all the browser's IO by way of tcp/ip, win32, directx, etc and leave all the rendering engine out of kernel space. But microsoft is probably doing exactly that for the most part. It's hard to say what's part of the windows operating system (kernel) and what just ships with it. Sure there's a lot of stuff that you can't uninstall but that doesn't mean that stuff isn't bound by the same rules that an application like firefox is bound by. It's pretty hard to say how integrated IE really is or if most of these bugs are just there because MS ships when stuff is just "good enough.
I couldn't get a response from NASA's site but the Coral p2p mirror is working well for me. NASA's download page held up for me though.
Well, they were taking over marketshare like crazy (even before being bundled) and were inventing new technologies faster than the rest of the company could keep up. So they shut them down to ensure that web apps wouldn't take over the rest of their business. They certainly didn't stop developing IE because they couldn't afford it. I don't know about you, but grabbing a few guys that can eat away at the rest of Microsoft's business doesn't sound like such a bad idea to me ;)
I wonder if this reason would have anything to do with back to school. The week ending at Aug 21st sounds about right. Are kids deciding more and more on AMD boxes whereas older users are sticking with Intel? I know a few people who've gone purely Intel after the days of the cyrix or K6 chips.... I can't really blame them either. Those were horrible gaming chips.
In any case I'm seeing a lot of shiny new AMD boxes around my campus and it's alright by me.
I've never been a fan of c++ add-ons. You can't build a newbie-friendly community when your mods are potentially unsafe. Unreal, for example, allows c++ add-ons but it's considered very poor form. The big MSU contest bans entries with native code. These days I don't see c++ as a pro. When I see that I think of it more as a "no scripting language" con. I'd hate to see mods begin to bundle in other software, for example. I'd just much rather keep things in a safe sandbox. I certainly don't have the time or desire to review source code before playing a mod.
If you can add support to mozilla, there's no reason why you can't add support to IE via plug-in. Well, that's probably not entirely true. Maybe someone who's familiar with what can and can't be done via IE plugins could comment?