I like how OSX lets me run as a user, and then when it needs to do something as root, asks me for the password automatically. It's a wonderful approach to the problem. I know that it's now trying to do something as root, and I have a point where I can sit and think, "Is that ok?"
When they change the namespace conventions ever service-pack and release for apps that need access to shared memory blocks, MS doesn't make it easy at all.
We ended up writing a sizeable helper library just to make sure that our apps could properly exchange data via shared memory on about 6 different configurations of windows. Win 2K, Win XP with fast user switching and with domain security (with and without SP1 or two), and Win2k3 server.
It wasn't exactly fun. Especially when the daemon app was started by a service, and a client app was started by a user.
The security model's never really been coherent, and it's a continually moving target for developers to keep up with MS's bastardization of the API. Or rather, the API doesn't change, but they change the naming conventions for shared objects so that they fall into the appropriate security contexts.
A LOT of academic networks (even lab networks at universities) want to have the plaintext of each user's passwords accessible to the admins. I think because A) they think it's "more secure" for them, and B) they don't know how to setup admin access for those that need it to have full access to all of the account space.
My problem with it was that it required me to come up with a disposable password, which was hard to keep track of vs. my normal password set.
You went to the wrong store. Go to a hi-fi shop that doesn't specialize in tube eqiupment, and does home theatre stuff, and you'll find piles of high tech equipment. There's a lot of really good stuff out there now, and they're starting to ditch the black plastic for much more interesting designs.
There are a number of "minimalist" systems out there, integrated DVD player, receiver, decoder, and amp, with a decent bit of power, and decent numbers. Usually they look much more like a mac mini (lots of brushed aluminum in different colors), then like 80s stereo equipment.
I rarely see a southbridge that has a heatsink on it, and the hypertransport bus negates most of the need for a northbridge that's sucking down a lot of power. There just isn't the transistor count in those areas to worry about the power consumption as much as it matters with the CPU.
Technical Release 1? I'm not sure of the exact breakdown of the acronym. However, it's the upcoming revision to the language standard (C++ for 2005, or C++05).
Lots of very interesting things in it.
And while CUJ is a mostly paid site, if you're a C++ programmer, and want to keep up to speed on the new language developments, I can't think of a better place. The magazine contains a lot of very cool articles showing what people are doing out on the bleeding edge of C++ application design.
Have you looked at the power consumption of the 90nm process Athlon64s? They're pulling VERY low amounts of power at idle (11 watts, IIRC). At full burn, they use less than a P4 does at idle. And this is a dual-core chip like the announced 4800+ X2.
From poking on Intel's site, the Penitium M in my laptop runs about 28W at full load (1.6Ghz,1.484V, 21 Amps). The faster processors pull a bit more (edging up towards 30 watts).
The AMD mobile chips are in the same ballpark (25-35 watts, depending on the clock), with a better connection to the rest of the system (hypertransport).
One thing that WILL definitely be interesting from all of this, is that for apps that run on both hardware, which will be faster? And for the same hardware (depending on what Apple does with the mobo/chipset/bios, this could be hard), how do similar tasks between the OSs fare (Windows, Linux, OSX)?
Funny. I have a PentiumM laptop and a 1.5Ghz PowerBook sitting side by side on my desk here. Both with the same memory. Frankly, the powerbook runs faster, except in a few very small circumstances, which center around non-vectorized float-point operations.
Operations that I was looking at Tiger, and thinking that with some recompiles, I was going to fix (open-source applications compiled using Xcode's version of GCC).
But I just don't see the performance benefit of the PentiumM over the PPC (G4, even). And when you include the battery life of a G4 at full load vs. that PentiumM, then the G4 REALLY starts to shine.
I know that the PentiumM is the future of Intel, and that it's FAST, but for most things, it just doesn't add up.
The only thing I don't do is game, so if that's the place where the real gains are, then that's why I'm not seeing them. But it shrugs off photoshop just fine.
Yes, the FSB is limited on the G4s, but really, that's the only limit I've ever really ran into (and then only when loading very large files from the system cache).
And, if you did a half-baked job in college, did you fail to get a high grade?
Of course.
And did it really matter in the end?
Not a whit. Grades didn't seem to be that important to my interviewers for my job in real life. And frankly, now that I'm the one doing interviews, I rarely pay any attention to them, except to be suspect of anyone with too good of grades. What else did they do during school? why are their grades so high? It doesn't rule them out, just changes the questions asked a bit.
Main things I learned in college were how to solve problems, how to figure out what things can be copromised due to schedule and what can't (very much like triaging bugs for a release), and how to learn fast, and on my own, becuase the prof was a @!#%@ joke.
The phosphor's response time isn't irrelevant in a CRT, but it's usually enough to not be a problem.
One thing I've noticed on some very high contrast ratio CRTs is that there is a bit of streaking. Especially noticeable with the interlaced signals and an object moving quickly at a diagnol across the screen, if your eyes are fast enough to see single-chip DLP "rainbows", you'll probably catch the fade-out of the phosphors from a white object on a black background through a few paints of the screen.
Luckily, they tend to use faster-fading phosphors in monitors than in tv CRTs...
If they can ace every test without doing the HW, why should the HW be part of the grade? Sounds like the tests were too easy vs. the HW, or they just knew how to do everything they needed for the class.
Most important thing I learned in college, when it got like that my senior year (sleeping in classes, due to being up to 3-4am or later working on projects, and back at school at 8am), was to prioritize and figure out what I wasn't going to do. Which project would get shelved and done half-assed? the busywork one.
It was painfull at times (dropping from 98% to 78% in a class because you simply do not have the time to do the research for a paper that's 20% of your grade, and it's a liberal arts elective and you're in a technical degree).
Do some profs/teachers try to make up for being bad profs/teachers by assigning piles of homework? definitely. I had numerous profs like that in college. HS was the same way.
The second-best math teacher I ever had lectured for 30 minutes, and then assigned a series of problems on the lecture, due the next class. We had 30 minutes of class-time to work on those problems. I never had "homework" and always understood the concepts. I could be done in 15-20 minutes most days, and then would help friends that were having problems. Not letting them copy, but going over the concepts again with them, and showing them how it was done. You learn even more that way, when they ask "why?".
and a remarkably large number of (cheap) devices simply don't bother. You need to buffer and reclock the data on the receive end to get some immunity to the jitter of the sender, including able to slighly increase/decrease your playback speed to match (or you'll eventually get an over/underbuffer situation).
But it IS an easy problem to fix. But do they bother? not usually.
So, before I spend money on special cables, I spend money on decent hardware that doesn't need it.
Or trying to open a vpn connection from a computer on one NAT'd network into another NAT'd network, which use overlapping private address spaces. I've had to change my home subnet twice now do to changes in the subnets used at work and VPN.
When gateway(at either end) to the outside world is 192.168.0.1, you're just kinda screwed for actually getting that to work right.
They hacked out a bunch of plot. Mathilda's crush on Leon is kinda glossed over a bit in the Professional due to some cut scenes from the "Leon" version.
MS really screwed up the security model for this. Anything that needs to be accessed as a named object (blocks of shared memory, named pipes, etc.) requires that you very carefully prefix the names with the right token so that it knows how to deal with it.
And they changed this mid service pack, IIRC. And it's different on Win2K, WinXP, WinXP Home, Win2k3Server, and if you're running terminal services (fast user switching) or instead are part of a domain (or worse, both).
A developer I used to work with basically had to sit down with about 300 pages of documentation from MS to figure out how it all worked. And then our app took forever to get back to working the way it was meant to (with the various pieces uses queues in shared memory).
The APIs for proper registry and file permissions are overly complicated, when what you want to do is just have the OS "do the right thing".
I've watched many movies that used surround to good effect without screwing it up, I just haven't heard any DVD-As that could do it.
Multiple channel (more than 2.0/2.1) does actually make sense, unless you're using headphones. Because unless you're using headphones, you'll hear some of the left channel in the right ear, etc. So being able to place the audience in a circle of speakers allows for being able to properly place sounds behind a person, which are VERY difficult to do through digital manipulation, even when you can drive the ears separately with headphones. I've heard demos of it, but it wasn't nearly as accurately portrayed as a properly setup surround-sound system can do.
Reproducing the same using stereo speakers is next to impossible. With an anechoic chamber (or a VERY dead listening room), it gets closer, but it's still nothing like what headphones can do, which is better, but still not as realistic as just putting the sound behind you or to the side of you in the first place.
All the DVD-A discs I've heard used the side/rear speakers so poorly that it sounded like playtime and using the channels for the sake of being able to, not because it added anything.
But I've yet to hear a surround encoded DVD-A that didn't sound like the engineers just had no idea how to deal with the multiple channels. They didn't create a convincing listening evironment, instead it was just instruments randomly thrown into the back channels.
(and this isn't a playback system problem, it's a very well setup theatre system at a friends house).
A lot of modern engines are actually higher compression than they really should be, and even when run with high octane fuel, if they ran at stociastic fuel ratios at heavy load, they'd detonate heavily.
So instead they dump extra fuel into the cylinders to cool them. The extra fuel just doesn't get burned. EPA doesn't seem to care about full throttle operation.
The higher compression ratio means that at lower loads/rpms, the cylinder pressures (and therefore temps) are higher, and they get a more complete burn of what IS in the cylinders.
So it turns into a bit of a game to get the compression ratio right for constant speed mileage and for passing power.
If the networks enable it, then yes. All the current CDMA phones know where they are, using both GPS and the CDMA network to provide differential GPS capabilities (as each tower knows EXACTLY where it is).
The problem is that while the API exists in most phones, the carriers haven't "turned it on", and so apps downloaded to the phone don't have acccess to the data. The 911 mode is "special", and allowed acccess.
It's bunk, because I'd love my phone to tell me exactly where I am.
Due to the relatively low compression ratio of a turbo gasoline engine (like the Evo), running with no boost at all would kill the performance and efficiency. There just isn't enough compression to get a good burn of the fuel.
Some forced induction engines run at 8:1 compression instead of the 9:1 or higher compression that most performance engines that are naturally aspirated tend to be (and therefore require premium fuel). So when not spooled up, the engine can't run nearly as efficiently as it does when spooled.
Now, if the engine was designed with a higher compression ratio, and used higher octane fuel, you could probably eek more off-boost efficiency out of it, at the cost of less (or more difficult to manage) on-boost performance.
maybe, but a better solution is to use electricly spun turbo turbines. Still pretty rare, but when the engine is running slow enough to not get the turbo to spool up into an efficient speed range, an electric motor built into the turbo housing spins it instead. Either just enough to offset the lag of the turbo, or enough to act like a super-charger and force a bunch of air into the engine via the compressor side.
Or, you run dual turbos, in series. One that spools up early, and one that spools up late. The one that spools up early helps with getting the engine running well, and provides flow to spool up the late one when it's needed.
heavily modded (600hp) diesel trucks try this route, and the 3rd gen RX-7 used that style turbo setup.
A spooled up turbo is pushing in more air, and so requires more fuel. But below the point where the turbo is spooled up, it creates a parasitic loss on the engine. Addition backpressure in the turbine housing, without the benefit of the compressor side of the turbo feeding more air into the engine. Below that point, it actually hurts performance.
You often see that on import cars with large aftermarket turbos that need to be pushed pretty hard to get them to spool up, at which point they give back a lot more power than they are using to spool up the turbo. This is "turbo lag".
***
With the ability to turn off cylinders, the methods I'm aware of involve altering the valve timing so that the intake/exhaust vavles are open all the time, so that the cylinder never compresses and receives no fuel.
No fuel is better than a little, as the cylinder will just ask as an air compressor/spring. Add a little fuel, and enough heat to make it detonate, and very bad things will happen (melted pistons).
Slashdot Mirror
I like how OSX lets me run as a user, and then when it needs to do something as root, asks me for the password automatically. It's a wonderful approach to the problem. I know that it's now trying to do something as root, and I have a point where I can sit and think, "Is that ok?"
When they change the namespace conventions ever service-pack and release for apps that need access to shared memory blocks, MS doesn't make it easy at all.
We ended up writing a sizeable helper library just to make sure that our apps could properly exchange data via shared memory on about 6 different configurations of windows. Win 2K, Win XP with fast user switching and with domain security (with and without SP1 or two), and Win2k3 server.
It wasn't exactly fun. Especially when the daemon app was started by a service, and a client app was started by a user.
The security model's never really been coherent, and it's a continually moving target for developers to keep up with MS's bastardization of the API. Or rather, the API doesn't change, but they change the naming conventions for shared objects so that they fall into the appropriate security contexts.
A LOT of academic networks (even lab networks at universities) want to have the plaintext of each user's passwords accessible to the admins. I think because A) they think it's "more secure" for them, and B) they don't know how to setup admin access for those that need it to have full access to all of the account space.
My problem with it was that it required me to come up with a disposable password, which was hard to keep track of vs. my normal password set.
You went to the wrong store. Go to a hi-fi shop that doesn't specialize in tube eqiupment, and does home theatre stuff, and you'll find piles of high tech equipment. There's a lot of really good stuff out there now, and they're starting to ditch the black plastic for much more interesting designs.
There are a number of "minimalist" systems out there, integrated DVD player, receiver, decoder, and amp, with a decent bit of power, and decent numbers. Usually they look much more like a mac mini (lots of brushed aluminum in different colors), then like 80s stereo equipment.
I rarely see a southbridge that has a heatsink on it, and the hypertransport bus negates most of the need for a northbridge that's sucking down a lot of power. There just isn't the transistor count in those areas to worry about the power consumption as much as it matters with the CPU.
Technical Release 1? I'm not sure of the exact breakdown of the acronym. However, it's the upcoming revision to the language standard (C++ for 2005, or C++05).
Lots of very interesting things in it.
And while CUJ is a mostly paid site, if you're a C++ programmer, and want to keep up to speed on the new language developments, I can't think of a better place. The magazine contains a lot of very cool articles showing what people are doing out on the bleeding edge of C++ application design.
Have you looked at the power consumption of the 90nm process Athlon64s? They're pulling VERY low amounts of power at idle (11 watts, IIRC). At full burn, they use less than a P4 does at idle. And this is a dual-core chip like the announced 4800+ X2.
From poking on Intel's site, the Penitium M in my laptop runs about 28W at full load (1.6Ghz,1.484V, 21 Amps). The faster processors pull a bit more (edging up towards 30 watts).
The AMD mobile chips are in the same ballpark (25-35 watts, depending on the clock), with a better connection to the rest of the system (hypertransport).
One thing that WILL definitely be interesting from all of this, is that for apps that run on both hardware, which will be faster? And for the same hardware (depending on what Apple does with the mobo/chipset/bios, this could be hard), how do similar tasks between the OSs fare (Windows, Linux, OSX)?
Funny. I have a PentiumM laptop and a 1.5Ghz PowerBook sitting side by side on my desk here. Both with the same memory. Frankly, the powerbook runs faster, except in a few very small circumstances, which center around non-vectorized float-point operations.
Operations that I was looking at Tiger, and thinking that with some recompiles, I was going to fix (open-source applications compiled using Xcode's version of GCC).
But I just don't see the performance benefit of the PentiumM over the PPC (G4, even). And when you include the battery life of a G4 at full load vs. that PentiumM, then the G4 REALLY starts to shine.
I know that the PentiumM is the future of Intel, and that it's FAST, but for most things, it just doesn't add up.
The only thing I don't do is game, so if that's the place where the real gains are, then that's why I'm not seeing them. But it shrugs off photoshop just fine.
Yes, the FSB is limited on the G4s, but really, that's the only limit I've ever really ran into (and then only when loading very large files from the system cache).
And, if you did a half-baked job in college, did you fail to get a high grade?
Of course.
And did it really matter in the end?
Not a whit. Grades didn't seem to be that important to my interviewers for my job in real life. And frankly, now that I'm the one doing interviews, I rarely pay any attention to them, except to be suspect of anyone with too good of grades. What else did they do during school? why are their grades so high? It doesn't rule them out, just changes the questions asked a bit.
Main things I learned in college were how to solve problems, how to figure out what things can be copromised due to schedule and what can't (very much like triaging bugs for a release), and how to learn fast, and on my own, becuase the prof was a @!#%@ joke.
The phosphor's response time isn't irrelevant in a CRT, but it's usually enough to not be a problem.
One thing I've noticed on some very high contrast ratio CRTs is that there is a bit of streaking. Especially noticeable with the interlaced signals and an object moving quickly at a diagnol across the screen, if your eyes are fast enough to see single-chip DLP "rainbows", you'll probably catch the fade-out of the phosphors from a white object on a black background through a few paints of the screen.
Luckily, they tend to use faster-fading phosphors in monitors than in tv CRTs...
If they can ace every test without doing the HW, why should the HW be part of the grade? Sounds like the tests were too easy vs. the HW, or they just knew how to do everything they needed for the class.
Most important thing I learned in college, when it got like that my senior year (sleeping in classes, due to being up to 3-4am or later working on projects, and back at school at 8am), was to prioritize and figure out what I wasn't going to do. Which project would get shelved and done half-assed? the busywork one.
It was painfull at times (dropping from 98% to 78% in a class because you simply do not have the time to do the research for a paper that's 20% of your grade, and it's a liberal arts elective and you're in a technical degree).
Do some profs/teachers try to make up for being bad profs/teachers by assigning piles of homework? definitely. I had numerous profs like that in college. HS was the same way.
The second-best math teacher I ever had lectured for 30 minutes, and then assigned a series of problems on the lecture, due the next class. We had 30 minutes of class-time to work on those problems. I never had "homework" and always understood the concepts. I could be done in 15-20 minutes most days, and then would help friends that were having problems. Not letting them copy, but going over the concepts again with them, and showing them how it was done. You learn even more that way, when they ask "why?".
Good luck, man. BTW, what school you at?
and a remarkably large number of (cheap) devices simply don't bother. You need to buffer and reclock the data on the receive end to get some immunity to the jitter of the sender, including able to slighly increase/decrease your playback speed to match (or you'll eventually get an over/underbuffer situation).
But it IS an easy problem to fix. But do they bother? not usually.
So, before I spend money on special cables, I spend money on decent hardware that doesn't need it.
Or trying to open a vpn connection from a computer on one NAT'd network into another NAT'd network, which use overlapping private address spaces. I've had to change my home subnet twice now do to changes in the subnets used at work and VPN.
When gateway(at either end) to the outside world is 192.168.0.1, you're just kinda screwed for actually getting that to work right.
They hacked out a bunch of plot. Mathilda's crush on Leon is kinda glossed over a bit in the Professional due to some cut scenes from the "Leon" version.
They sanitized it a bit for US audiences...
MS really screwed up the security model for this. Anything that needs to be accessed as a named object (blocks of shared memory, named pipes, etc.) requires that you very carefully prefix the names with the right token so that it knows how to deal with it.
And they changed this mid service pack, IIRC. And it's different on Win2K, WinXP, WinXP Home, Win2k3Server, and if you're running terminal services (fast user switching) or instead are part of a domain (or worse, both).
A developer I used to work with basically had to sit down with about 300 pages of documentation from MS to figure out how it all worked. And then our app took forever to get back to working the way it was meant to (with the various pieces uses queues in shared memory).
The APIs for proper registry and file permissions are overly complicated, when what you want to do is just have the OS "do the right thing".
I've watched many movies that used surround to good effect without screwing it up, I just haven't heard any DVD-As that could do it.
Multiple channel (more than 2.0/2.1) does actually make sense, unless you're using headphones. Because unless you're using headphones, you'll hear some of the left channel in the right ear, etc. So being able to place the audience in a circle of speakers allows for being able to properly place sounds behind a person, which are VERY difficult to do through digital manipulation, even when you can drive the ears separately with headphones. I've heard demos of it, but it wasn't nearly as accurately portrayed as a properly setup surround-sound system can do.
Reproducing the same using stereo speakers is next to impossible. With an anechoic chamber (or a VERY dead listening room), it gets closer, but it's still nothing like what headphones can do, which is better, but still not as realistic as just putting the sound behind you or to the side of you in the first place.
Sounds like I need to check those out.
All the DVD-A discs I've heard used the side/rear speakers so poorly that it sounded like playtime and using the channels for the sake of being able to, not because it added anything.
But I've yet to hear a surround encoded DVD-A that didn't sound like the engineers just had no idea how to deal with the multiple channels. They didn't create a convincing listening evironment, instead it was just instruments randomly thrown into the back channels.
(and this isn't a playback system problem, it's a very well setup theatre system at a friends house).
Yes, it does.
A lot of modern engines are actually higher compression than they really should be, and even when run with high octane fuel, if they ran at stociastic fuel ratios at heavy load, they'd detonate heavily.
So instead they dump extra fuel into the cylinders to cool them. The extra fuel just doesn't get burned. EPA doesn't seem to care about full throttle operation.
The higher compression ratio means that at lower loads/rpms, the cylinder pressures (and therefore temps) are higher, and they get a more complete burn of what IS in the cylinders.
So it turns into a bit of a game to get the compression ratio right for constant speed mileage and for passing power.
If the networks enable it, then yes. All the current CDMA phones know where they are, using both GPS and the CDMA network to provide differential GPS capabilities (as each tower knows EXACTLY where it is).
The problem is that while the API exists in most phones, the carriers haven't "turned it on", and so apps downloaded to the phone don't have acccess to the data. The 911 mode is "special", and allowed acccess.
It's bunk, because I'd love my phone to tell me exactly where I am.
Due to the relatively low compression ratio of a turbo gasoline engine (like the Evo), running with no boost at all would kill the performance and efficiency. There just isn't enough compression to get a good burn of the fuel.
Some forced induction engines run at 8:1 compression instead of the 9:1 or higher compression that most performance engines that are naturally aspirated tend to be (and therefore require premium fuel). So when not spooled up, the engine can't run nearly as efficiently as it does when spooled.
Now, if the engine was designed with a higher compression ratio, and used higher octane fuel, you could probably eek more off-boost efficiency out of it, at the cost of less (or more difficult to manage) on-boost performance.
maybe, but a better solution is to use electricly spun turbo turbines. Still pretty rare, but when the engine is running slow enough to not get the turbo to spool up into an efficient speed range, an electric motor built into the turbo housing spins it instead. Either just enough to offset the lag of the turbo, or enough to act like a super-charger and force a bunch of air into the engine via the compressor side.
Or, you run dual turbos, in series. One that spools up early, and one that spools up late. The one that spools up early helps with getting the engine running well, and provides flow to spool up the late one when it's needed.
heavily modded (600hp) diesel trucks try this route, and the 3rd gen RX-7 used that style turbo setup.
A spooled up turbo is pushing in more air, and so requires more fuel. But below the point where the turbo is spooled up, it creates a parasitic loss on the engine. Addition backpressure in the turbine housing, without the benefit of the compressor side of the turbo feeding more air into the engine. Below that point, it actually hurts performance.
You often see that on import cars with large aftermarket turbos that need to be pushed pretty hard to get them to spool up, at which point they give back a lot more power than they are using to spool up the turbo. This is "turbo lag".
***
With the ability to turn off cylinders, the methods I'm aware of involve altering the valve timing so that the intake/exhaust vavles are open all the time, so that the cylinder never compresses and receives no fuel.
No fuel is better than a little, as the cylinder will just ask as an air compressor/spring. Add a little fuel, and enough heat to make it detonate, and very bad things will happen (melted pistons).