And you doing that would affect the results how? You won't prevent a good teacher from getting clearly good results - the whole rest of the class will more than overwhelm your purposeful glitch, especially when taken in context over a number of years. Plus, teachers do in fact have some influence over whether their students take tests seriously or not.
But hey, it's not a perfect system - there will be issues. They'll be fairly minor, but they will be there.
Holding up merit pay to a standard of perfection is a straw-man with no value. Currently, we don't measure performance, so that's what we get - no performance (not really, there is some altruism involved - but altruism only goes so far). Even an imperfect measurement of performance is much, much better than no measurement at all.
Although laser projection tech has been around for some time now, I'm really surprised that it's not used.
Laser projection is just getting started out at the very high end of the market (theater projection).
Sony is currently demoing their Laser Theatre Dream Display at the World Expo in Aichi, Japan. What they are showing is actually three displays with a slight overlap (which actually works reasonably well except for blacks). Each display consists of many red diode lasers, blue diode lasers, and green lasers piped through fiber through a 1-dimensional MEMS light shutter which is scanned across the display. The current vertical resolution is 1k, but by keeping the shutter 1-dimensional, it's easy to see how the resolution can quickly be pumped up. Horizontal resolution is controlled by the light gate speed and scanning speed. The displays are absolutely gorgeous - they blow current digital theatre projectors away. Color gamut is outstanding, no visible scanlines, no visible flicker, even when looking off to the side. I can't wait to be able to start watching movies this way - it'll be the thing that finally gets me back into theatres regularly.
Of course, it will be a while before you can have something like this in your house, much less as a DIY project. Maybe by then someone will have figured out how to get a solid state green laser going.
- 50% based on classroom performance improvement over the year. The second test of the kids should take place months before summer break, to prevent the pure teach-the-test problem.
- 30% based on school performance improvement over the year (to encourage sharing of lesson plans and cooperation). May be further subdivided into improvement relative to other schools in district, state, or nationwide. Lack of cooperation is one of the whining complaints always given as a reason for not having merit pay, and this is an easy solution.
- 20% based on parent and student feedback. This needs to be on a curve, probably within the district, since there will always be that percentage of crazy parents that dislike any teacher their kids have or who are upset when their kids don't always get the undeserved A.
For administrators:
- Replace the portion based classroom improvement with relative ratio of money under their control to money that makes it to the classroom, relative to other schools in the district/state/nation. Until you start measuring and negatively impacting administrator pay for a lack of efficiency, the current bloated eduocracy will continue to burn money inefficiently.
Other things:
- Stop this crazy extra long summer break thing. Yes, kids need a break to be kids. No, it doesn't have to be three months long, with the resultant loss of retention.
- Keep teachers with the same class longer (i.e., follow a class through grades 1, 2, and 3). Increases the accuracy of any measurement of improvement.
- Admit that some students learn differently than others, and put the students in classes/tracks based on that. Get those that learn visually together, etc.
- School vouchers. It's one sure-fire way of getting parents more involved, and one great measurement of parental feedback. If all the kids move to another school, you can bet you kinda suck. I have not heard one cogent argument against this (the typical one is that it takes money away from the schools, which is bull, because no voucher program ever had the voucher value anywhere near what the schools got per student - only if the administrative overhead is so ridiculously high that it's greater than the difference between per-student funding and voucher value is there any damage, and the solution then isn't to not use vouchers, but to fix the overhead!).
- Long or no tenure period. It's ridiculous that after just 3 years in some places, poor teachers can have a lock on their job. If you don't have the ability to get rid of the bottom 5% of performers, guess what you end up with?
As an "educational libertarian" (I believe that we should fund education through college - but only when a system is in place that creates efficient spending) I'm disgusted at the morons who think that we can solve the problem by throwing money at it. Guess what? Per-student funding in the U.S. is quite high. Efficiency of that money is extraordinarily low. And the "teachers" unions (esp. the CTA) is made up of mostly administrators! Their grab for additional funding is all about self-preserving their bloated bureaucracy (as an aggregate behavior in the face of no measurement of efficiency).
Until we start measuring what we want to see - improvement, efficiency - we will never see those things and we will continue to throw good money after bad.
Cheaper than flying, more efficient than flying, more comfortable than flying, and much, much more punctual than flying... and as a bonus, the schedules are linked up with the local trains.
You literally can walk right on (even with luggage), walk off on the other end, walk over to another platform and catch a local within about 10 minutes.
Now, how much extra time do you have to pad a one hour plane flight with because of all the garbage on either end?
Actually, I'd be willing to bet that among Photoshop users those using mice with Photoshop are a minority. Between tablets, trackballs, point sticks, drag pads and other alternatives...
(I have no hard data, I'm guessing).
Americans yell into cell phones and don't have any cell phone etiquette. I just got back from Japan, where everyone has a cell phone, and where everyone understands how to actually use them. You don't use them while on the train, or you go between cars where you won't bother anyone, and you talk into them, letting the noise-reduction actually work and not bothering those around you.
Only here, back at home, do people seem to be so stupid that they don't understand there's a proper place and method for using cellphones. And it's not while tailgating me on the freeway!
I never understood this fixation with mice. They just plain suck.
Trackballs are great for day-to-day pointer tasks - easy to whip the cursor across the screen, just as easy to precisely hit a button, all without doing the spastic slap-slap-slap mouse dance.
I always replaced the cheap Apple mice with the Kensington trackballs. Unfortunately, Kensington forgot about design somewhere along the way and fell behind Microsoft and Logitech.
I avoided the Logitech trackballs when they went with that awful thumb-ball design. Now Logitech has a decent finger trackball, but it's *wireless*?!? Wireless trackball sort of misses the point of a device that never has to move, don't you think?
Anyway, when Microsoft put out the Trackball Explorer, I grabbed a bunch. Finally, a trackball designed by someone who obviously used and liked trackballs.
And if a trackball doesn't work for you, get a tablet - or a point-stick, or...
64 bit UI, because it means you don't have to ship around memory just to display things, which isn't exactly cheap when memory bandwidth is so scarce. What? Just map shared memory in? Why, that's so cheap (not!) we should just do that. Just veneer malloc and keep all the UI assets in low memory? Ooh, yeah, that works so well (bzzt).
Yeah, separating out the UI from the business logic gets you *most* of the way there, but if what you are processing is what you need to display (e.g., video), most of the way isn't good enough and all the workaround schemes die in the 3rd ring of hell (OS overhead).
Sucks, but true.
Now, if you've got a kernel where mremap/madvise using shared memory is a small constant time operation in user mode no matter the chunk size being dealt with - well, things change at that point and doing a gemini-task application becomes interesting. I just don't know of one yet.
This is a cop-out. It means the data structures need to be chosen better so that they're not slinging around pointers as much, and integers don't automatically have to be 64 bits.
So while it may be true that the 64-bit version of Apple's UI frameworks were slower, it's not true of frameworks in general, and would be more an indication of a problem with the framework than with moving to 64 bit.
And let's face it; all the tricks with dealing with more than 2GB of data on a 32-bit system aren't nearly as clean or efficient as just having the address space. The Final Cut Pro and Shake teams were right to be ticked off, and Microsoft got the 64-bit answer right.
Go grab the Adjusted Refresh plug-in. It will cut down on the initial scratch sizes used. Certain smaller files with large numbers of layers were especially pessimitically affected. It will also result in somewhat slower processing in the cases where the image would fit into RAM.
Make sure the memory percentage is adjusted properly for your usage on the machine - if you're running WinAmp, IE, or other memory hogs, turn down the percentage, perhaps to 50% or 40%.
Make sure the scratch disk being used is NOT the same physical disk as the physical disk Windows is installed on. If you only have one disk and you are dealing with large images, think about getting a second one.
Well, it's also not Photoshop's fault since he's apparently talking about a very old version. Photoshop's right click and wheel functionality is all over the place. On both platforms. In fact, if you use Photoshop with a one button mouse and no scroll wheel, you're missing a lot.
I do, in fact, like Apple's approach. I agree with them that there are many computer users out there who, if they aren't confused by a two button mouse, won't ever use that second button for fear of what it might do.
But they leave the rest open for those willing to poke and prod. Sometimes not as nice as Windows. Keyboard layout, until Tiger, has been a pain to fix. The control key belongs next to the A key, darnnit, like Emacs intended - for those who still want it fixed, there are Info.plist entries you can tweak (http://gnufoo.org). If you don't know the registry entry to hack to do the same thing on Windows, I pity you. (Or map in a Windows key on IBM laptops;-).
But Mac applications get right click and wheel events, just like on the other platforms.
You just can't let your developers plod around with a regular old mouse if you ever expect them to actually implement things that use the extra features.
Me, I like tablets and trackballs. Can't stand mice and trackpads. For me, it's ironic - the bext mouse for my Macs is from Microsoft (Trackball Explorer). You should see people try and use my setup - figuring out which buttons are which. And then they try and use my keyboards (with the above CapsLock-is-Control mappings - Mac and Windows).
Naw, it's not Apple's fault that users don't actually try right clicking in recent versions of thier Windows programs.
Multi-core can, in some designs, share some cache, which can help with processor thread migration performance and in reducing cache misses in certain kinds of threaded apps (e.g., DVD decoding could benefit). Though that really begins to complicate your processor affinity algorithms...
Note that early multi-core designs don't usually have such cache sharing.
SMP designs are moving towards NUMA (non-uniform memory access) in general to help with memory bottlenecks, but it's not clear yet how that will be applied in multi-core situations.
But one of the most interesting aspects of multi core is how certain companies are considering it in terms of licensing.
Microsoft, for example, has stated that they will consider a single die as a processor for licensing terms, but will still treat each core appropriately. That means, if you use four-core chip, you could end up with eight processors (on two cores) while still using XP Professional.
That is, if you don't count the $1000 it'll cost you to replace the tubes due to burn-in every few years.
Or the $500 it'll cost you to have it calibrated every year to keep that "sharper" picture.
Yeah, I've gotten 6 years out of a CRT HDTV - on it's third tube set and it needs another recal. Sharper? Hah! And with no DRM interface, it's getting obsolete real fast.
But then, that's at least 3 more years than you'd get out of a plasma and doesn't suck like DLP with the screen doors and rainbows.
...than see entire teams get RIFed and have the projects sent overseas.
I code, I'm darn good at it, but guess what? Without a team around, how long would my job exist? The team I work on is a great cross section, and I think it's a better team because of it.
Not only that, when working with people locally, it takes far less time to teach them that (guess what) formatting *does* matter, that naked pointers with one character names are a *bad* idea, that it really *is* a good idea to write manageable code that you'll be able to read and modify in two years (not like a lot of the junk coming back from a lot of the overseas stuff - a lot of that stuff would keep lint output scrolling past for days).
It's unfortunate there are so many people who don't look past the color of someone's skin. I just want to work with people who are self-motivated, want to constantly learn, and who aren't just clocking in from 9-5 because they had Java in college and think that makes them hot stuff. Or worse, PhDs who's eyes glaze over whenever the phrase "type safety" is muttered.
What the networks will need to realize with the broadcast flag is that with the spread of PVRs, at some point abusing the flag will cost them more than it will gain them.
If we can't DVR a show, it won't get watched. Period.
_Then_ try and sell the use of the flag to advertisers.
Slashdot Mirror
But hey, it's not a perfect system - there will be issues. They'll be fairly minor, but they will be there.
Holding up merit pay to a standard of perfection is a straw-man with no value. Currently, we don't measure performance, so that's what we get - no performance (not really, there is some altruism involved - but altruism only goes so far). Even an imperfect measurement of performance is much, much better than no measurement at all.
You reap what you measure.
Laser projection is just getting started out at the very high end of the market (theater projection).
Sony is currently demoing their Laser Theatre Dream Display at the World Expo in Aichi, Japan. What they are showing is actually three displays with a slight overlap (which actually works reasonably well except for blacks). Each display consists of many red diode lasers, blue diode lasers, and green lasers piped through fiber through a 1-dimensional MEMS light shutter which is scanned across the display. The current vertical resolution is 1k, but by keeping the shutter 1-dimensional, it's easy to see how the resolution can quickly be pumped up. Horizontal resolution is controlled by the light gate speed and scanning speed. The displays are absolutely gorgeous - they blow current digital theatre projectors away. Color gamut is outstanding, no visible scanlines, no visible flicker, even when looking off to the side. I can't wait to be able to start watching movies this way - it'll be the thing that finally gets me back into theatres regularly.
Of course, it will be a while before you can have something like this in your house, much less as a DIY project. Maybe by then someone will have figured out how to get a solid state green laser going.
They do, and recognizing the good ones is a must. Implying that they don't is pretty disrespectful.
- 50% based on classroom performance improvement over the year. The second test of the kids should take place months before summer break, to prevent the pure teach-the-test problem.
- 30% based on school performance improvement over the year (to encourage sharing of lesson plans and cooperation). May be further subdivided into improvement relative to other schools in district, state, or nationwide. Lack of cooperation is one of the whining complaints always given as a reason for not having merit pay, and this is an easy solution.
- 20% based on parent and student feedback. This needs to be on a curve, probably within the district, since there will always be that percentage of crazy parents that dislike any teacher their kids have or who are upset when their kids don't always get the undeserved A.
For administrators:
- Replace the portion based classroom improvement with relative ratio of money under their control to money that makes it to the classroom, relative to other schools in the district/state/nation. Until you start measuring and negatively impacting administrator pay for a lack of efficiency, the current bloated eduocracy will continue to burn money inefficiently.
Other things:
- Stop this crazy extra long summer break thing. Yes, kids need a break to be kids. No, it doesn't have to be three months long, with the resultant loss of retention.
- Keep teachers with the same class longer (i.e., follow a class through grades 1, 2, and 3). Increases the accuracy of any measurement of improvement.
- Admit that some students learn differently than others, and put the students in classes/tracks based on that. Get those that learn visually together, etc.
- School vouchers. It's one sure-fire way of getting parents more involved, and one great measurement of parental feedback. If all the kids move to another school, you can bet you kinda suck. I have not heard one cogent argument against this (the typical one is that it takes money away from the schools, which is bull, because no voucher program ever had the voucher value anywhere near what the schools got per student - only if the administrative overhead is so ridiculously high that it's greater than the difference between per-student funding and voucher value is there any damage, and the solution then isn't to not use vouchers, but to fix the overhead!).
- Long or no tenure period. It's ridiculous that after just 3 years in some places, poor teachers can have a lock on their job. If you don't have the ability to get rid of the bottom 5% of performers, guess what you end up with?
As an "educational libertarian" (I believe that we should fund education through college - but only when a system is in place that creates efficient spending) I'm disgusted at the morons who think that we can solve the problem by throwing money at it. Guess what? Per-student funding in the U.S. is quite high. Efficiency of that money is extraordinarily low. And the "teachers" unions (esp. the CTA) is made up of mostly administrators! Their grab for additional funding is all about self-preserving their bloated bureaucracy (as an aggregate behavior in the face of no measurement of efficiency).
Until we start measuring what we want to see - improvement, efficiency - we will never see those things and we will continue to throw good money after bad.
Using phish for bait?
Cheaper than flying, more efficient than flying, more comfortable than flying, and much, much more punctual than flying... and as a bonus, the schedules are linked up with the local trains.
You literally can walk right on (even with luggage), walk off on the other end, walk over to another platform and catch a local within about 10 minutes.
Now, how much extra time do you have to pad a one hour plane flight with because of all the garbage on either end?
Actually, I'd be willing to bet that among Photoshop users those using mice with Photoshop are a minority. Between tablets, trackballs, point sticks, drag pads and other alternatives... (I have no hard data, I'm guessing).
Americans yell into cell phones and don't have any cell phone etiquette. I just got back from Japan, where everyone has a cell phone, and where everyone understands how to actually use them. You don't use them while on the train, or you go between cars where you won't bother anyone, and you talk into them, letting the noise-reduction actually work and not bothering those around you. Only here, back at home, do people seem to be so stupid that they don't understand there's a proper place and method for using cellphones. And it's not while tailgating me on the freeway!
I never understood this fixation with mice. They just plain suck.
Trackballs are great for day-to-day pointer tasks - easy to whip the cursor across the screen, just as easy to precisely hit a button, all without doing the spastic slap-slap-slap mouse dance.
I always replaced the cheap Apple mice with the Kensington trackballs. Unfortunately, Kensington forgot about design somewhere along the way and fell behind Microsoft and Logitech.
I avoided the Logitech trackballs when they went with that awful thumb-ball design. Now Logitech has a decent finger trackball, but it's *wireless*?!? Wireless trackball sort of misses the point of a device that never has to move, don't you think?
Anyway, when Microsoft put out the Trackball Explorer, I grabbed a bunch. Finally, a trackball designed by someone who obviously used and liked trackballs.
And if a trackball doesn't work for you, get a tablet - or a point-stick, or...
Anything but an RMI-causing mouse!
64 bit UI, because it means you don't have to ship around memory just to display things, which isn't exactly cheap when memory bandwidth is so scarce. What? Just map shared memory in? Why, that's so cheap (not!) we should just do that. Just veneer malloc and keep all the UI assets in low memory? Ooh, yeah, that works so well (bzzt).
Yeah, separating out the UI from the business logic gets you *most* of the way there, but if what you are processing is what you need to display (e.g., video), most of the way isn't good enough and all the workaround schemes die in the 3rd ring of hell (OS overhead).
Sucks, but true.
Now, if you've got a kernel where mremap/madvise using shared memory is a small constant time operation in user mode no matter the chunk size being dealt with - well, things change at that point and doing a gemini-task application becomes interesting. I just don't know of one yet.
This is a cop-out. It means the data structures need to be chosen better so that they're not slinging around pointers as much, and integers don't automatically have to be 64 bits.
So while it may be true that the 64-bit version of Apple's UI frameworks were slower, it's not true of frameworks in general, and would be more an indication of a problem with the framework than with moving to 64 bit.
And let's face it; all the tricks with dealing with more than 2GB of data on a 32-bit system aren't nearly as clean or efficient as just having the address space. The Final Cut Pro and Shake teams were right to be ticked off, and Microsoft got the 64-bit answer right.
Trust me, try the plug-in. Given the problem you describe, the Adjusted Refresh plug-in will make things better.
And, if it doesn't, removing it is really simple.
Go grab the Adjusted Refresh plug-in. It will cut down on the initial scratch sizes used. Certain smaller files with large numbers of layers were especially pessimitically affected. It will also result in somewhat slower processing in the cases where the image would fit into RAM.
Make sure the memory percentage is adjusted properly for your usage on the machine - if you're running WinAmp, IE, or other memory hogs, turn down the percentage, perhaps to 50% or 40%.
Make sure the scratch disk being used is NOT the same physical disk as the physical disk Windows is installed on. If you only have one disk and you are dealing with large images, think about getting a second one.
Well, it's also not Photoshop's fault since he's apparently talking about a very old version. Photoshop's right click and wheel functionality is all over the place. On both platforms. In fact, if you use Photoshop with a one button mouse and no scroll wheel, you're missing a lot.
;-).
I do, in fact, like Apple's approach. I agree with them that there are many computer users out there who, if they aren't confused by a two button mouse, won't ever use that second button for fear of what it might do.
But they leave the rest open for those willing to poke and prod. Sometimes not as nice as Windows. Keyboard layout, until Tiger, has been a pain to fix. The control key belongs next to the A key, darnnit, like Emacs intended - for those who still want it fixed, there are Info.plist entries you can tweak (http://gnufoo.org). If you don't know the registry entry to hack to do the same thing on Windows, I pity you. (Or map in a Windows key on IBM laptops
But Mac applications get right click and wheel events, just like on the other platforms.
You just can't let your developers plod around with a regular old mouse if you ever expect them to actually implement things that use the extra features.
Me, I like tablets and trackballs. Can't stand mice and trackpads. For me, it's ironic - the bext mouse for my Macs is from Microsoft (Trackball Explorer). You should see people try and use my setup - figuring out which buttons are which. And then they try and use my keyboards (with the above CapsLock-is-Control mappings - Mac and Windows).
Naw, it's not Apple's fault that users don't actually try right clicking in recent versions of thier Windows programs.
> ... ("Site Finder" and its blind ignorance ...
Blind ignorance?
Or open-eyed arrogance?
Multi-core can, in some designs, share some cache, which can help with processor thread migration performance and in reducing cache misses in certain kinds of threaded apps (e.g., DVD decoding could benefit). Though that really begins to complicate your processor affinity algorithms...
Note that early multi-core designs don't usually have such cache sharing.
SMP designs are moving towards NUMA (non-uniform memory access) in general to help with memory bottlenecks, but it's not clear yet how that will be applied in multi-core situations.
But one of the most interesting aspects of multi core is how certain companies are considering it in terms of licensing.
Microsoft, for example, has stated that they will consider a single die as a processor for licensing terms, but will still treat each core appropriately. That means, if you use four-core chip, you could end up with eight processors (on two cores) while still using XP Professional.
Yeah, CRTs are cost effective. Sure.
That is, if you don't count the $1000 it'll cost you to replace the tubes due to burn-in every few years.
Or the $500 it'll cost you to have it calibrated every year to keep that "sharper" picture.
Yeah, I've gotten 6 years out of a CRT HDTV - on it's third tube set and it needs another recal. Sharper? Hah! And with no DRM interface, it's getting obsolete real fast.
But then, that's at least 3 more years than you'd get out of a plasma and doesn't suck like DLP with the screen doors and rainbows.
LCOS where are you?!?
Lastly, if you have a hummer h2 ("girly hummer"), your manhood has already suffered too much.
And you probably think 4WD/AWD helps you stop faster.
...than see entire teams get RIFed and have the projects sent overseas.
I code, I'm darn good at it, but guess what? Without a team around, how long would my job exist? The team I work on is a great cross section, and I think it's a better team because of it.
Not only that, when working with people locally, it takes far less time to teach them that (guess what) formatting *does* matter, that naked pointers with one character names are a *bad* idea, that it really *is* a good idea to write manageable code that you'll be able to read and modify in two years (not like a lot of the junk coming back from a lot of the overseas stuff - a lot of that stuff would keep lint output scrolling past for days).
It's unfortunate there are so many people who don't look past the color of someone's skin. I just want to work with people who are self-motivated, want to constantly learn, and who aren't just clocking in from 9-5 because they had Java in college and think that makes them hot stuff. Or worse, PhDs who's eyes glaze over whenever the phrase "type safety" is muttered.
That's backwards. If I can't TiVO a show, I'm less likely to watch it.
What the networks will need to realize
with the broadcast flag is that with the
spread of PVRs, at some point abusing the
flag will cost them more than it will gain them.
If we can't DVR a show, it won't get watched.
Period.
_Then_ try and sell the use of the flag to advertisers.