Woz does not really have a good track record of innovation. I think it's the Steve Jobs type of person, the enthusiastic task master, cajoling, encouraging, criticizing, apologizing, and throwing tantrums, that can get 150% out of the brightest and sharpest hardware and software engineers.
I think Woz wants to "play" again. Nothing wrong with that. I've spent the last 22 years of my life trying to return to the mythical high-school computer playground of my youth (the PLATO computer system - a system that was copied by Xerox PARC, Apple, NetNews / Lotus Notes, etc., and a host of other "mee-too" companies..)
This is not too surprising. We all know that body cells have a "fixed" life, e.g. divide something like 200 times, stop dividing, and die off 50 years later.
However, the one really variable thing in a person is the amount of blood that they lose. In women, blood is converted to mother's milk, so you might expect that mothers of 20 children or more would die young... but they probably do not.
In men who fight wars, it's easy to lose gallons upon gallons of blood. Where do all those red blood cells come from ?? From the bone marrow.
For nearly all types of injuries, blood (and sometimes muscle) morphs into scar tissue. So whenever you break a limb, you lose quite a bit of blood even if there is no visible bruising. In fact, in this respect blood seems to behave a lot like stem cells.
If I had been looking for a source of stem cells - cells that can differentiate into almost any other type of cells, the bone marrow would be the first place i'd look. In fact, I'd be really surprised if there WEREN'T stem cells in the bone marrow.
We recently had children and had their stem cells (from cord blood) stored. What's the primary purpose of doing this ?? Bone marrow transplants !!!
Running two Durons is just as smart as drying your bald head with a hairdryer. In a multiprocessor system, it's the cache that makes all the difference in performance, and the Duron (and Celeron) have less cache than their brethren.
So stop wasting energy, and either upgrade to Athlon / Athlon MP, or throw out that second Duron because you're wasting Oil !!
Yes, if an engineer makes $50 an hour, if you can save the guy 6 hrs/yr over the course of a 3-year product cycle (0.3% of the year), you have paid for a $1000 chip. The economics are compelling.
Of course, once you install Windoze 2K01XP, you've blown all the savings that Intel sweated so hard to produce...
Here is the problem with design patterns. Once upon a time, people would read the AT&T UNIX and BSD Kernels in order to get ideas for how to write programs well. In order to be a good programmer, you would invariably find a better programmer and then read all his stuff. It was in this way that you found cool programming concoctions, such as this :
#define _CONCAT(X,Y) X/**/Y
I'll list three more examples to make the point clear. On the Xerox STAR workstation, the BITBLT instruction was used to perform database range queries - e.g. give me all the servers in the network that support file serving.
As a second example, the MESA signal facility was used during the expansion of mailing distribution lists to detect circular expansions (e.g., raise a "Have_I_Seen_This_One_Before?" signal, go up the stack looking for matches on the signal arg, if there is no match, fall through the signal handler and continue the recursive expansion).
A fourth example is the way that the "C" switch statement is the ideal construct for parsing command-line arguments in a UNIX shell command.
Now in the object-oriented world, people do not want to be bothered with the problem of reading others code, or thinking for themselves, so they go for these trendy "design patterns" books where other people do the heavy lifting or simply make up garbage to see if it will sell. This lets the average or below-average programmer off the hook and relieves them of the burden of forming their own opinions and developing their own tastes in programming.
This is evil in it purest form. It is the an example of the same laziness that has us receiving 95% opinion and 5% in every news story reported today. This laziness have moved outside the news room and is invading software engineering.
You may already have killed a small dinosaur by reading the on-line version, why go ahead and murder trees by purchasing a copy ?? If you really want to show your support, why not send bruce a check for the retail price, so that the author can profit from his own work ??
You could buy an Omnitracs terminal from Qualcomm, they run about $4k plus monthly fees. They provide 300 bits per second of throughput through a geostationary satellite, perfected by a wicked-good Eudora email client and a keyboard to rival the Sinclair ZX-80 of 1982. Not only that, but your fans could track your geolocation nationally to within a few kilometers to predict when you will arrive at your next gig. Not only that, but truly die-hard fans could tell if you were making unauthorized stops, like at the Chicken Ranch of Amarillo Texas.
Or, you could wait about 3 years until Qualcomm comes out with the next mobile satellite-based trucking system. It should be almost 10x faster. But, I think it will still provide mainly email and geolocation.
Alternately, you could buy a Globalstar phone and get 10 kbit/sec access for only $800 (phone) plus $1.50 a minute. However, it would be a race to see which ends first : your tour, or the operation of the Globalstar System.
As a consumer in the Greenspan Information economy, you have many choices !!!
I have taught a class in logic design / computer architecture from Hennesy & Patterson's book, and I feel that to teach it with SPARC assembly language is just silly. The degenerate SPARC CPU architecture was motivated by the fundamental question, "what can I fit into a 10K gate array in the year 1985 ???" if I remember correctly. I can understand the desire to fill the student's mind with stale and rotting information, as a way to taking revenge for your low academic salary, but teach SPARC assembly language is simply going too far...
Since the whole world is going embedded, and increasingly, battery-operated, the two CPU's that are most practical to teach about are (a) ARM (most mips per watt), and (b) 8051 (when you absolutely positively don't need 200 mips to start with...)
Actually, you have to give Windows-XP some credit (much as I hate to admit). Anyone can bang out a multitasking kernel in a couple of thousand lines of code. Adding 143 system calls maybe is perhaps 30 klocs of code (after all, original V7 unix ran on a 64 Kb PDP-11).
The real proof in the pudding is the drivers. When I try to install a Win2K printer, literally 1000 different printers are supported. If you look at my HP-85xi printer driver, with its 2-up and 4-up printing, scaling, preview, ad nauseum, it's probably 1.0M lines all by itself.
I'll be impressed when linux supports 1000 device drivers with an average size of 1000 bytes of compiled code per driver. And, it will have to support than a line-printer style of interface to printers. It will take some sort of new O-O architecture or "what's beyond O-O ??" to make this a reality.
IMHO, we ought to start a movement where software drivers are written for a kernel and ONLY THEN are hardware engineers allowed to build hardware. This would eliminate 15 million line kernels. Too often the software guys are treated like pooper scoopers following the hardware guys around in the public parks.
When I was an undergrad at MIT, there was a guy who modified a flatbed plotter so that he could print PCB's from any PC / workstation. The last issue of Technology Review contained an article called, Print your Next PC that describes researchers at MIT who are working on a device that could print hardware (chips) directly. It should not be hard if you can get an automatic etcher and a laser together. The graphics guys working on 3D device printing with a bed of liquid acrylic and a laser and an anvil that slowly submerges (the laser is used to solidify just the area you need) have already solved many of these problems.
A 32 Mb floppy is a stupid idea. First, reliability of both the media and the drive are poor. Second, a Minidisc holds 140 Mb and is faster and the drive is smaller. I'd much rather use minidiscs for cameras, PC's, etc. When the drive costs 1000x the cost of the media, somebody is being rather stupid.
This whole thread is a bogus idea. In fact, the entire web site is bogus. No, actually the whole web is bogus. I hate this bogus computer. I'm going to kick this computer in the screen. here i got right now... ready... set... #&@#(*^&#(*&$@#%&$(
At 1 Ghz, the speed of light in a vacuum moves 11.3 inches every clock cycle. To get to 10 Ghz, light can only move 1.13 inches. The diameter of these chips right now is something like 0.5 inches. So, I'll be really surprised if we get to 10 Ghz.
Slashdot Mirror
Woz does not really have a good track record of innovation. I think it's the Steve Jobs type of person, the enthusiastic task master, cajoling, encouraging, criticizing, apologizing, and throwing tantrums, that can get 150% out of the brightest and sharpest hardware and software engineers.
I think Woz wants to "play" again. Nothing wrong with that. I've spent the last 22 years of my life trying to return to the mythical high-school computer playground of my youth (the PLATO computer system - a system that was copied by Xerox PARC, Apple, NetNews / Lotus Notes, etc., and a host of other "mee-too" companies..)
This is not too surprising. We all know that body cells have a "fixed" life, e.g. divide something like 200 times, stop dividing, and die off 50 years later.
... but they probably do not.
However, the one really variable thing in a person is the amount of blood that they lose. In women, blood is converted to mother's milk, so you might expect that mothers of 20 children or more would die young
In men who fight wars, it's easy to lose gallons upon gallons of blood. Where do all those red blood cells come from ?? From the bone marrow.
For nearly all types of injuries, blood (and sometimes muscle) morphs into scar tissue. So whenever you break a limb, you lose quite a bit of blood even if there is no visible bruising. In fact, in this respect blood seems to behave a lot like stem cells.
If I had been looking for a source of stem cells - cells that can differentiate into almost any other type of cells, the bone marrow would be the first place i'd look. In fact, I'd be really surprised if there WEREN'T stem cells in the bone marrow.
We recently had children and had their stem cells (from cord blood) stored. What's the primary purpose of doing this ?? Bone marrow transplants !!!
Running two Durons is just as smart as drying your bald head with a hairdryer. In a multiprocessor system, it's the cache that makes all the difference in performance, and the Duron (and Celeron) have less cache than their brethren.
So stop wasting energy, and either upgrade to Athlon / Athlon MP, or throw out that second Duron because you're wasting Oil !!
Pre-ordering is ecologically sound. 'Nuff said.
Yes, if an engineer makes $50 an hour, if you can save the guy 6 hrs/yr over the course of a 3-year product cycle (0.3% of the year), you have paid for a $1000 chip. The economics are compelling.
...
Of course, once you install Windoze 2K01XP, you've blown all the savings that Intel sweated so hard to produce
Here is the problem with design patterns. Once upon a time, people would read the AT&T UNIX and BSD Kernels in order to get ideas for how to write programs well. In order to be a good programmer, you would invariably find a better programmer and then read all his stuff. It was in this way that you found cool programming concoctions, such as this :
#define _CONCAT(X,Y) X/**/Y
I'll list three more examples to make the point clear. On the Xerox STAR workstation, the BITBLT instruction was used to perform database range queries - e.g. give me all the servers in the network that support file serving.
As a second example, the MESA signal facility was used during the expansion of mailing distribution lists to detect circular expansions (e.g., raise a "Have_I_Seen_This_One_Before?" signal, go up the stack looking for matches on the signal arg, if there is no match, fall through the signal handler and continue the recursive expansion).
A fourth example is the way that the "C" switch statement is the ideal construct for parsing command-line arguments in a UNIX shell command.
Now in the object-oriented world, people do not want to be bothered with the problem of reading others code, or thinking for themselves, so they go for these trendy "design patterns" books where other people do the heavy lifting or simply make up garbage to see if it will sell. This lets the average or below-average programmer off the hook and relieves them of the burden of forming their own opinions and developing their own tastes in programming.
This is evil in it purest form. It is the an example of the same laziness that has us receiving 95% opinion and 5% in every news story reported today. This laziness have moved outside the news room and is invading software engineering.
We're all dumber as a result of this trend !!!
You may already have killed a small dinosaur by reading the on-line version, why go ahead and murder trees by purchasing a copy ?? If you really want to show your support, why not send bruce a check for the retail price, so that the author can profit from his own work ??
You could buy an Omnitracs terminal from Qualcomm, they run about $4k plus monthly fees. They provide 300 bits per second of throughput through a geostationary satellite, perfected by a wicked-good Eudora email client and a keyboard to rival the Sinclair ZX-80 of 1982. Not only that, but your fans could track your geolocation nationally to within a few kilometers to predict when you will arrive at your next gig. Not only that, but truly die-hard fans could tell if you were making unauthorized stops, like at the Chicken Ranch of Amarillo Texas.
Or, you could wait about 3 years until Qualcomm comes out with the next mobile satellite-based trucking system. It should be almost 10x faster. But, I think it will still provide mainly email and geolocation.
Alternately, you could buy a Globalstar phone and get 10 kbit/sec access for only $800 (phone) plus $1.50 a minute. However, it would be a race to see which ends first : your tour, or the operation of the Globalstar System.
As a consumer in the Greenspan Information economy, you have many choices !!!
I have taught a class in logic design / computer architecture from Hennesy & Patterson's book, and I feel that to teach it with SPARC assembly language is just silly. The degenerate SPARC CPU architecture was motivated by the fundamental question, "what can I fit into a 10K gate array in the year 1985 ???" if I remember correctly. I can understand the desire to fill the student's mind with stale and rotting information, as a way to taking revenge for your low academic salary, but teach SPARC assembly language is simply going too far ...
...)
Since the whole world is going embedded, and increasingly, battery-operated, the two CPU's that are most practical to teach about are (a) ARM (most mips per watt), and (b) 8051 (when you absolutely positively don't need 200 mips to start with
Actually, you have to give Windows-XP some credit (much as I hate to admit). Anyone can bang out a multitasking kernel in a couple of thousand lines of code. Adding 143 system calls maybe is perhaps 30 klocs of code (after all, original V7 unix ran on a 64 Kb PDP-11).
The real proof in the pudding is the drivers. When I try to install a Win2K printer, literally 1000 different printers are supported. If you look at my HP-85xi printer driver, with its 2-up and 4-up printing, scaling, preview, ad nauseum, it's probably 1.0M lines all by itself.
I'll be impressed when linux supports 1000 device drivers with an average size of 1000 bytes of compiled code per driver. And, it will have to support than a line-printer style of interface to printers. It will take some sort of new O-O architecture or "what's beyond O-O ??" to make this a reality.
IMHO, we ought to start a movement where software drivers are written for a kernel and ONLY THEN are hardware engineers allowed to build hardware. This would eliminate 15 million line kernels. Too often the software guys are treated like pooper scoopers following the hardware guys around in the public parks.
When I was an undergrad at MIT, there was a guy who modified a flatbed plotter so that he could print PCB's from any PC / workstation. The last issue of Technology Review contained an article called, Print your Next PC that describes researchers at MIT who are working on a device that could print hardware (chips) directly. It should not be hard if you can get an automatic etcher and a laser together. The graphics guys working on 3D device printing with a bed of liquid acrylic and a laser and an anvil that slowly submerges (the laser is used to solidify just the area you need) have already solved many of these problems.
A 32 Mb floppy is a stupid idea. First, reliability of both the media and the drive are poor. Second, a Minidisc holds 140 Mb and is faster and the drive is smaller. I'd much rather use minidiscs for cameras, PC's, etc. When the drive costs 1000x the cost of the media, somebody is being rather stupid. This whole thread is a bogus idea. In fact, the entire web site is bogus. No, actually the whole web is bogus. I hate this bogus computer. I'm going to kick this computer in the screen. here i got right now... ready... set... #&@#(*^&#(*&$@#%&$(
At 1 Ghz, the speed of light in a vacuum moves 11.3 inches every clock cycle. To get to 10 Ghz, light can only move 1.13 inches. The diameter of these chips right now is something like 0.5 inches. So, I'll be really surprised if we get to 10 Ghz.