Well, ogg is released and people are using it but it has't exploded in popularity. The difference here is in 10 years, there will be still millions of people who have VCDs. If you release a $19.95 program on even an obscure platform, you are likely to sell hundreds of copies and earn a nice vacation. Or, a few of the hundred will be programmers who decide to cooperate and write a free player. I know I am optimisitic when I am betting that there will be still countries which do not enforce software patents or require equal licensing - if you charge $5 for Sony to sell a DVD player, I can send you a check for $5 and use any software players of my choice.
So anyway, there will not be many VC3 videos sold in an asian bookstore. So mostly, it will be limited to free, self-produced or pirated content, only a small fraction of the number of users you would get if movies were consistently sold in that format. Now, look at OpenDivX. Most people who were using it will now forget all about it and jump to VC3. Now, if I made some important home videos in OpenDivX, I can only rely on maybe 0.01% as many users looking for a solution in 10 years. Who is to say VC3 will not be also abandoned in a short time because someone discovers better algorithms or because it turns out that someone did have a patent on a key piece.
So it is worth to give up some performance, image quality and money to be able to record your content in a durable format. I don't know if putting Java, C or some special VideoML code on each CD is feasable. But I do know that if someone can come up with a format that is guaranteed to be watchable by me, on my OS in 10 years I would spend some money to upgrade my hardware and software in order to use it.
How much would you bet that you will be able to watch a movie compressed with the beta version of VP3 in 10 years? That is, without spending weeks programming? Compared, to let's say a VCD? For me, this makes it a better format to record my TV episode, artifacts and all. Because I don't want to worry about which OS my next PC, or my friend's PC is going to run.
Of course, the reason is that so many people already have VCD movies that someone will come up with a way to watch them in future. Obviously open source is better than a single company that might just go belly up or drop the product without releasing the data format. But let's say only a few people start using VP3 before Xiph developers discover the "next big thing" and stop maintaining the old format. Are you really up to taking over?
Now look at it this way. Media files are getting bigger and bigger and VMs are getting faster and faster. Why not just put the decoder into every file? Say, each video starts with.class files that use regular Java APIs to render the movie. Then it's possible to make a software or hardware player that is compatible with many standards, including future ones.
Ok, maybe it's not practical for high-resolution video yet. But definitely, this aproach can be used today for high-end, AC-powered audio players. Or for streaming video - I bet you can use Java for anything you get over a 56K modem. Finally, high-resolution video can be archived using this approach. Today, you will still native code to play it, or you will be able to uncompress it overnight first and re-compress to your platform's native format. Then in a couple of years real-time decompression using the same Java code will be possible.
MacOSX is not exactly getting killed by Microsoft Office, Internet Explorer and Media Player. It's one reason people can switch without getting locked out of their documents, web sites and recorded TV programs. In fact, there is a healthy competition for each product. I suspect more people use Camino, Safari and Quicktime than the Microsoft counterparts.
I can't see as having another option available is a bad thing for open source, especially Linux itself. Before it's "if I want to watch this video, I need to boot windows". Now its "if I want to watch this video, I can run one Microsoft app under Linux, without having to boot/buy windows. This also got to create some pressure to make official, up-to-date versions of Quicktime and RealOne (hopefully without sneaky changes to/etc/rc.d or.xinitrc) available for both regular and embedded Linux.
What, like x86 instruction set?
on
Legacy-Free PCs
·
· Score: 3, Informative
It could be true that there is nothing wrong with the serial port to connect a modem. But the article said nothing about the most obsolete component in today's Intel PC - the processor itself. So we might see a PC with a fancy BIOS that comes with its own windowing system, but still has a processor with less than 10 general purpose registers. I know that Apple, Sun, AMD and so on probably underplay the significance of clock speed. But, 1GHZ PowerPC sure runs faster than 1GHZ P4.
What we need is a modern, legacy free instruction set specially designed to support modern programming languages like C++ and Java. Large number of registers and hardware stack ("register windows") support is a start, but I am sure there are new ideas developed after Sparc design. What would an ideal machine language of today look like if it doesn't have to be remotely understandable by a human, only by the optimizing compiler. For example, if Intel's branch prediction, load/store reordering, parallel execution and so on are already specified by instructions themselves.
And of course, this means starting anew with a single instruction set. No more emulating 8086, 80286, 80386 and so on in hardware. Software emulation, like 68K programs under MacOSX comes to mind, but I guess better legacy free all the way.
Which means that the start is probably not a desktop PC, but a cheap, high-performance server. If you can have a Linux port, database server and a J2EE application server available, you might not care about the rest for your online store server if you get a better price/performance ratio.
When the technology does come to desktop it will be probably covered with adapters, software emulation and even some bits of hardware emulation like a christmas tree and it will take years to whittle them away. Well, that's life.
Then you might write a quick and dirty function that calls sprintf to format a message (snprintf is not portable, so you might not have a simple fix). Then after a while you forget that it was quick and dirty and use it in a client that will only connect to your own server. I think its a very easy mistake to make.
It gets more interesting. Say you are reading a 1024 bit number that is supposed to be a product of two 512 bit primes. Your code has a hand-optimized assembler loop that will not violate bounds of a fixed-length array if the number is what it's supposed to be. But if it has small factors, the loop might blow away the memory. On the other hand, checking the bounds would make your performance-critical loop twice slower. Still think it's easy to validate the input?
How are you planning to benefit long term?
on
SCO Group Lawsuit Q&A
·
· Score: 4, Insightful
Many big OS vendors - IBM, Sun, Apple - have realized that if a decent free kernel is available people are not willing to pay $$$ for pretty much the same thing. Instead they are selling software that runs on top of Linux/*BSD, basic UNIX utilities and gcc.
It seams that SCO decided to hold out like Microsoft and try to compete from the kernel level up. What are you plans to make this approach succeed when even IBM, with more money to pay for development, is trying to hedge its bets on AIX.
It seams to me that a better strategy for SCO would be to take advantage of existing SCO binary compatibility Linux offers and release a professional quality Linux workstation and a low-end server using SCO applications, administation tools and so on.
But in this case, the lawsuit makes no sense. Sure, there is a chance of one-time payout from IBM. But nobody except law firms builds a successful business on lawsuits alone. If applications is what matters, you guys might as well release whatever "corporate secrets" you think will further increase Linux scalability, stability and so on and let the enormous number of willing hobbyists integrate them into the kernel.
There are a lot of things going for this approach. For one thing, nobody buys an OS itself. Customers buy an OS to be able to run certain programs. Nothing prevents SCO from selling those proprietary, closed-source programs under Linux. Just look at MacOS X.
On the other hands, lots of people are obviously willing to write, optimize and improve OS code just for the fun of it. SCO could just use their work to get a performance boost for both UI and server applications with no investment. Also, writting device drivers is a thankless work but thousands already exist on Linux, free for the taking. Is there an optimized NVIDIA or ATI 3D driver for SCO?? How about adding some rendering applications and shipping a serious alternative to SGI based on Linux?
I don't know if IBM used any knowlege of AT&T UNIX to make improvements in Linux. But I am pretty sure that trying to guard yesterday's technologies is not in the long-term interest of SCO and its shareholders. Spend effort where the new markets are today.
Otherwise, you'll just get a worthless puddle of protoplasm when you uncompress people on the other end of the teleport. Also, don't compress humans and insects together, altough you might get a better ratio that way.
I think both arguments - XML sucks and doesn't suck - are missing the point. For example, my company switched to ClearCase because of the touted features, but there are several meetings per week to solve some branch problems, or VOB problems, or wrong tag. The same group was using RCS before and people had to do some manual merge but didn't spend nearly so much time on it.
Something technology can be incredibly powerful, extensible, theoretically clean and yet incredibly unpleasant to use. I had to write a simple servlet the other day and was dismayed to see that JSDK 2.0 command line switches are gone and I have to write a multi-page jumble of brackets, columns and what not to get the thing running. Ended up searching for an old version of JSDK on co-workers machine. What wouldn't I give for a servlet engine that uses good old.ini files.
It's true that various libraries could hide the mess they use internally. But then when they break, I get to look at a particulary convoluted example of the mess. There are so many smart people in CS. Can't someone come up with a data format that is pleasent to read, edit and parse, not just powerful?
Suppose I send a request to withdraw some cash from my account. If two backends agree to withdraw $100 and the last one decides to donate $100 to world peace, what exactly happens to my account and how much cash do I get?
Clearly the systems that do real work, that need to be protected the most, can not be duplicated. Would work well for DNS/web servers though.
MNP5 got rid of the annoying line noise. In fact, there was a terminal program that implemented it in software, on regular 2400 modems. Would love to hear how it is possible. Worked like a charm, together with screen for completely dropped connections - no more {}_a(9*&& in Emacs.
As for speed, yes MNP5 was a little choppy and didn't make any difference for zmodem on a compressed file. V42 doesn't hurt performance much though.
How about that anti-smoking ad by a guy smoking through a hole in his neck? Or inmates coming to school to talk about abiding the law? I think Microsoft has quite a lot to talk about on the subject.
Wouldn't this be a good term when someone is thinking about programmers gender or race when reading a technical article?
I use she because women (and some men!) get all offended if I use "he" as a generic pronoun. I couldn't care less myself, so fine, have it their way. Besides... ladies first! At least when it's this easy.
I don't think most people have mental images of every word they say/hear. Otherwise, how can you stand walking around at work hearing things like "you screwed my PC", "this code kicks ass" and "I have to remove all the bugs from this piece of shit today"?
I want a replacement for Explorer that looks like WindowMaker. I know you can run GNUStep on Cygwin, but it doesn't manage native apps and Cygwin X server is pretty slow. I want to actually kill explorer, run the new program and add my windows apps to the dock. Ideally, it will also replace window borders with a leaner version.
I feel my P3/128M RAM desktop at work runs twice faster after switching XP to "Windows classic" UI. I figure it will go supersonic if I get rid of the rest of the UI bloat.
It does try share the code pages of DLLs, even though they are mapped into address space of multiple processes. The difference from Win3.1 is that the global data is, by default, per-process.
The problem is that the DLL can be loaded into different virtual address in different processes and VC++ doesn't have -fPIC. After the code is relocated, there is not much of it that is still shared.
I hope they either fix the phone to work on older CDMA network, or even better actually fix the "next generation" GPRS to have coverage outside downtowns of big cities. Here, we have a big Sprint PCS building and within a few blocks there is a coverage hole on a Sprint network. Only a couple of networks ("23 century" version of AT&T wireless and Cingular) are usable on the beach and for making calls while you keep driving. All the cute Palm and CE "smart" phones on the other hand, only work with Sprint and Verizon and are not very usable to actually make a call around here.
Non-US customers can ignore this post and might actually enjoy all the next generation gadgets. How is life in the 24th century?
A system made with junk hardware, for example all defective components that failed QA tests, but sufficient redundancy and error correction to run normally. Mass storage that can *maybe* store a terabyte on a CD-size disk, with astronomical error rate and error-correction software that gets 100G in reliably. Processors that run overclocked by design and have extra circuits to correct occasional failures. Any takers?
All the ways to earn or get food/money today call for people skills (interview, dress style, filling out forms), being able to show up to work on time and so on. A lot of people just don't have this kind of skills.
Say I am a big (but non-criminal) guy walking around city block in shabby dress, carrying a sign and asking some random people if I can do some work for them. If someone did ask me to wash their car, carry a heavy bag home etc, I would do it to the best of my ability and bring money to my family, if any. But something tells me I might get some handouts but no work. In an agricultural society, on the other hand, my neighbors would make good use of my muscles to plow a field, dig out a tree, move stones and so on.
I don't think it's always "refuse to take responsibility", more like "unable to find a common language with modern society". And the society should also take some responsibility for eliminating choices that once existed.
Either Cassini is really expensive for an unmanned research probe or poor people are being neglected. I mean, space exploration is great, but so is making sure that everyone has food to eat.
Private lines don't contribute much to security, as they still go through public phone companies, public land, public airwaves and so on. If your company has confidential data that could be worth a million bucks to someone, you shouldn't trust this kind of security. Let's not even talk about state secrets.
On the other hand, VPN over Internet can be very secure and far cheaper. Not VPN using OpenSSL on Linux boxes, because both OS and the relatively big library could have buffer overflows or some other low-level bugs. But it's easy to build a layered system that will be extremly secure. Say, hardware routers that decrypt and check signature on every incoming packet in hardware before looking at it otherwise. And then AFTER that, a Linux box that does a santity check on what comes through the router, just in case.
Should really bug users before allowing any type of remote access or automatic program download/execution. Like making one talk to a live customer support person to get an "advanced user activation code". Or at least make the user take a randomized multiple-choice quiz on security. Otherwise, you are restricted to outbound connection to WWW, DNS, POP and IMAP and the requests are filtered though a local Java proxy that check the line length and absense of suspicious control characters. Otherwise this kind of problems will just breed out of control.
Which always gave boost to interactive processes by raising internal priority of the ones that didn't take up the whole CPU slice. Remember how jerky Solaris 2.x felt right after Sun killed SunOS 4.x? My BSD-based aqua desktop on the other hand is very responsive, even when playing mp3s AND doing real time video compression in the background.
Anyway, I hope Linus and Ingo now have some spare time for real dreams of power users with Linux desktops - drivers! I used to run Linux at work and at home (used to, because I got rid of the Intel box at home). Between these two machines, I had an NVIDIA card, lucent WinModem, CLIE, Zaurus and an NTFS partition that wasn't recognized by default Redhat kernel.
After every kernel upgrade, I had to recompile 5 drivers. CLIE and Zaurus drivers came in the form of patches that usually refused to apply, or caused a hang when the device was attached! Once I tried a 2.5 kernel because it had some features, like suspend and resume that I could really use. While the default configuration built Ok, once I enabled the drivers I wanted, I couldn't get the thing compiled even before applying my patches.
Yes, you could just run "Redhat operating system", never upgrade the kernel and wait a few months to install a new release. Then you might find binary drivers to download for the kernel for your particular kind and number of CPUs. But the whole point of Linux (on a personal desktop) is to have some fun and try new stuff out easily.
Linux developers really need to stabalize driver interfaces. I should be able to go to kernel.org and download the latest kernel *binary*, then install a binary driver from the CD-ROM that came from my NVIDIA card.
USB and Firewire buses should be exposed by kernel as network interfaces, accessible to user programs through socket API. In this way, USB drivers will be both easier to write/debug AND will not contribute to Oops. For the ultimate of cool, Wine should support Windows USB drivers (my Virtual PC does!) and I should be able to just install Palm and Zaurus desktops and use them rather trying to feed ttyUSBNN to kpilot.
Having a stable system that doesn't have to be rebooted to Windows to use some unsupported USB device is far more important than raw performance. I wish any system's developers - Linux, *BSD, Darwin, BeOS, etc - would concentrate on this goal before going back to play with cool toys.
I bet it will cost the poor guy who just tried to share some pages with friends more than his tuition. Charging for bandwidth is a bad idea, because a lot of traffic - slashdot, worms, buggy research projects, DOS attacks, windows update, P2P and so on - is not something user can control or measure until the bill arrives. Limiting speed is better, especially if the limit only takes effect after you used up your daily quota of unlimitted-speed traffic. In any case, there should be liberal exceptions for students doing research projects (like Internet search, multimedia streaming or P2P protocols) that might use up a lot of bandwidth.
The whole point of subscription is to profit from insightful, funny etc comments that readers contribute for free. If paid users start posting 20 minutes earlier, "free" users will be discouraged from contributing, because their post is unlikely to get noticed, with lots of replies or a karma bonus, late into the picture. Or they might post, but spend less effort. With fewer posts, there will in turn be less appeal to subscribe.
On the other hand, slashdot could easily solve this problem by randomizing the comments while still preserving threads (randomize top level posts, then children of each parent and so on). As a bonus, discussions will no longer be fixated on a single catchy but off-topic thread.
Slashdot Mirror
Well, ogg is released and people are using it but it has't exploded in popularity. The difference here is in 10 years, there will be still millions of people who have VCDs. If you release a $19.95 program on even an obscure platform, you are likely to sell hundreds of copies and earn a nice vacation. Or, a few of the hundred will be programmers who decide to cooperate and write a free player. I know I am optimisitic when I am betting that there will be still countries which do not enforce software patents or require equal licensing - if you charge $5 for Sony to sell a DVD player, I can send you a check for $5 and use any software players of my choice.
So anyway, there will not be many VC3 videos sold in an asian bookstore. So mostly, it will be limited to free, self-produced or pirated content, only a small fraction of the number of users you would get if movies were consistently sold in that format. Now, look at OpenDivX. Most people who were using it will now forget all about it and jump to VC3. Now, if I made some important home videos in OpenDivX, I can only rely on maybe 0.01% as many users looking for a solution in 10 years. Who is to say VC3 will not be also abandoned in a short time because someone discovers better algorithms or because it turns out that someone did have a patent on a key piece.
So it is worth to give up some performance, image quality and money to be able to record your content in a durable format. I don't know if putting Java, C or some special VideoML code on each CD is feasable. But I do know that if someone can come up with a format that is guaranteed to be watchable by me, on my OS in 10 years I would spend some money to upgrade my hardware and software in order to use it.
How much would you bet that you will be able to watch a movie compressed with the beta version of VP3 in 10 years? That is, without spending weeks programming? Compared, to let's say a VCD? For me, this makes it a better format to record my TV episode, artifacts and all. Because I don't want to worry about which OS my next PC, or my friend's PC is going to run.
.class files that use regular Java APIs to render the movie. Then it's possible to make a software or hardware player that is compatible with many standards, including future ones.
Of course, the reason is that so many people already have VCD movies that someone will come up with a way to watch them in future. Obviously open source is better than a single company that might just go belly up or drop the product without releasing the data format. But let's say only a few people start using VP3 before Xiph developers discover the "next big thing" and stop maintaining the old format. Are you really up to taking over?
Now look at it this way. Media files are getting bigger and bigger and VMs are getting faster and faster. Why not just put the decoder into every file? Say, each video starts with
Ok, maybe it's not practical for high-resolution video yet. But definitely, this aproach can be used today for high-end, AC-powered audio players. Or for streaming video - I bet you can use Java for anything you get over a 56K modem. Finally, high-resolution video can be archived using this approach. Today, you will still native code to play it, or you will be able to uncompress it overnight first and re-compress to your platform's native format. Then in a couple of years real-time decompression using the same Java code will be possible.
MacOSX is not exactly getting killed by Microsoft Office, Internet Explorer and Media Player. It's one reason people can switch without getting locked out of their documents, web sites and recorded TV programs. In fact, there is a healthy competition for each product. I suspect more people use Camino, Safari and Quicktime than the Microsoft counterparts.
/etc/rc.d or .xinitrc) available for both regular and embedded Linux.
I can't see as having another option available is a bad thing for open source, especially Linux itself. Before it's "if I want to watch this video, I need to boot windows". Now its "if I want to watch this video, I can run one Microsoft app under Linux, without having to boot/buy windows. This also got to create some pressure to make official, up-to-date versions of Quicktime and RealOne (hopefully without sneaky changes to
It could be true that there is nothing wrong with the serial port to connect a modem. But the article said nothing about the most obsolete component in today's Intel PC - the processor itself. So we might see a PC with a fancy BIOS that comes with its own windowing system, but still has a processor with less than 10 general purpose registers. I know that Apple, Sun, AMD and so on probably underplay the significance of clock speed. But, 1GHZ PowerPC sure runs faster than 1GHZ P4. What we need is a modern, legacy free instruction set specially designed to support modern programming languages like C++ and Java. Large number of registers and hardware stack ("register windows") support is a start, but I am sure there are new ideas developed after Sparc design. What would an ideal machine language of today look like if it doesn't have to be remotely understandable by a human, only by the optimizing compiler. For example, if Intel's branch prediction, load/store reordering, parallel execution and so on are already specified by instructions themselves. And of course, this means starting anew with a single instruction set. No more emulating 8086, 80286, 80386 and so on in hardware. Software emulation, like 68K programs under MacOSX comes to mind, but I guess better legacy free all the way. Which means that the start is probably not a desktop PC, but a cheap, high-performance server. If you can have a Linux port, database server and a J2EE application server available, you might not care about the rest for your online store server if you get a better price/performance ratio. When the technology does come to desktop it will be probably covered with adapters, software emulation and even some bits of hardware emulation like a christmas tree and it will take years to whittle them away. Well, that's life.
Then you might write a quick and dirty function that calls sprintf to format a message (snprintf is not portable, so you might not have a simple fix). Then after a while you forget that it was quick and dirty and use it in a client that will only connect to your own server. I think its a very easy mistake to make. It gets more interesting. Say you are reading a 1024 bit number that is supposed to be a product of two 512 bit primes. Your code has a hand-optimized assembler loop that will not violate bounds of a fixed-length array if the number is what it's supposed to be. But if it has small factors, the loop might blow away the memory. On the other hand, checking the bounds would make your performance-critical loop twice slower. Still think it's easy to validate the input?
Many big OS vendors - IBM, Sun, Apple - have realized that if a decent free kernel is available people are not willing to pay $$$ for pretty much the same thing. Instead they are selling software that runs on top of Linux/*BSD, basic UNIX utilities and gcc.
It seams that SCO decided to hold out like Microsoft and try to compete from the kernel level up. What are you plans to make this approach succeed when even IBM, with more money to pay for development, is trying to hedge its bets on AIX.
It seams to me that a better strategy for SCO would be to take advantage of existing SCO binary compatibility Linux offers and release a professional quality Linux workstation and a low-end server using SCO applications, administation tools and so on.
But in this case, the lawsuit makes no sense. Sure, there is a chance of one-time payout from IBM. But nobody except law firms builds a successful business on lawsuits alone. If applications is what matters, you guys might as well release whatever "corporate secrets" you think will further increase Linux scalability, stability and so on and let the enormous number of willing hobbyists integrate them into the kernel.
There are a lot of things going for this approach. For one thing, nobody buys an OS itself. Customers buy an OS to be able to run certain programs. Nothing prevents SCO from selling those proprietary, closed-source programs under Linux. Just look at MacOS X.
On the other hands, lots of people are obviously willing to write, optimize and improve OS code just for the fun of it. SCO could just use their work to get a performance boost for both UI and server applications with no investment. Also, writting device drivers is a thankless work but thousands already exist on Linux, free for the taking. Is there an optimized NVIDIA or ATI 3D driver for SCO?? How about adding some rendering applications and shipping a serious alternative to SGI based on Linux?
I don't know if IBM used any knowlege of AT&T UNIX to make improvements in Linux. But I am pretty sure that trying to guard yesterday's technologies is not in the long-term interest of SCO and its shareholders. Spend effort where the new markets are today.
Otherwise, you'll just get a worthless puddle of protoplasm when you uncompress people on the other end of the teleport. Also, don't compress humans and insects together, altough you might get a better ratio that way.
1. Release a game with buffer overrun
2. Leak information to XBox-Linux community
3. Profit!!!!
They could even get the lindows.com award money if they did it right.
I think both arguments - XML sucks and doesn't suck - are missing the point. For example, my company switched to ClearCase because of the touted features, but there are several meetings per week to solve some branch problems, or VOB problems, or wrong tag. The same group was using RCS before and people had to do some manual merge but didn't spend nearly so much time on it.
.ini files.
Something technology can be incredibly powerful, extensible, theoretically clean and yet incredibly unpleasant to use. I had to write a simple servlet the other day and was dismayed to see that JSDK 2.0 command line switches are gone and I have to write a multi-page jumble of brackets, columns and what not to get the thing running. Ended up searching for an old version of JSDK on co-workers machine. What wouldn't I give for a servlet engine that uses good old
It's true that various libraries could hide the mess they use internally. But then when they break, I get to look at a particulary convoluted example of the mess. There are so many smart people in CS. Can't someone come up with a data format that is pleasent to read, edit and parse, not just powerful?
Wow, the other team had 2 2/3 games. Sounds nice and bloody. What were you playing?
Clearly the systems that do real work, that need to be protected the most, can not be duplicated. Would work well for DNS/web servers though.
As for speed, yes MNP5 was a little choppy and didn't make any difference for zmodem on a compressed file. V42 doesn't hurt performance much though.
How about that anti-smoking ad by a guy smoking through a hole in his neck? Or inmates coming to school to talk about abiding the law? I think Microsoft has quite a lot to talk about on the subject.
I use she because women (and some men!) get all offended if I use "he" as a generic pronoun. I couldn't care less myself, so fine, have it their way. Besides... ladies first! At least when it's this easy. I don't think most people have mental images of every word they say/hear. Otherwise, how can you stand walking around at work hearing things like "you screwed my PC", "this code kicks ass" and "I have to remove all the bugs from this piece of shit today"?
I feel my P3/128M RAM desktop at work runs twice faster after switching XP to "Windows classic" UI. I figure it will go supersonic if I get rid of the rest of the UI bloat.
The problem is that the DLL can be loaded into different virtual address in different processes and VC++ doesn't have -fPIC. After the code is relocated, there is not much of it that is still shared.
Non-US customers can ignore this post and might actually enjoy all the next generation gadgets. How is life in the 24th century?
A system made with junk hardware, for example all defective components that failed QA tests, but sufficient redundancy and error correction to run normally. Mass storage that can *maybe* store a terabyte on a CD-size disk, with astronomical error rate and error-correction software that gets 100G in reliably. Processors that run overclocked by design and have extra circuits to correct occasional failures. Any takers?
Say I am a big (but non-criminal) guy walking around city block in shabby dress, carrying a sign and asking some random people if I can do some work for them. If someone did ask me to wash their car, carry a heavy bag home etc, I would do it to the best of my ability and bring money to my family, if any. But something tells me I might get some handouts but no work. In an agricultural society, on the other hand, my neighbors would make good use of my muscles to plow a field, dig out a tree, move stones and so on.
I don't think it's always "refuse to take responsibility", more like "unable to find a common language with modern society". And the society should also take some responsibility for eliminating choices that once existed.
Either Cassini is really expensive for an unmanned research probe or poor people are being neglected. I mean, space exploration is great, but so is making sure that everyone has food to eat.
On the other hand, VPN over Internet can be very secure and far cheaper. Not VPN using OpenSSL on Linux boxes, because both OS and the relatively big library could have buffer overflows or some other low-level bugs. But it's easy to build a layered system that will be extremly secure. Say, hardware routers that decrypt and check signature on every incoming packet in hardware before looking at it otherwise. And then AFTER that, a Linux box that does a santity check on what comes through the router, just in case.
Anyway, I hope Linus and Ingo now have some spare time for real dreams of power users with Linux desktops - drivers! I used to run Linux at work and at home (used to, because I got rid of the Intel box at home). Between these two machines, I had an NVIDIA card, lucent WinModem, CLIE, Zaurus and an NTFS partition that wasn't recognized by default Redhat kernel.
After every kernel upgrade, I had to recompile 5 drivers. CLIE and Zaurus drivers came in the form of patches that usually refused to apply, or caused a hang when the device was attached! Once I tried a 2.5 kernel because it had some features, like suspend and resume that I could really use. While the default configuration built Ok, once I enabled the drivers I wanted, I couldn't get the thing compiled even before applying my patches.
Yes, you could just run "Redhat operating system", never upgrade the kernel and wait a few months to install a new release. Then you might find binary drivers to download for the kernel for your particular kind and number of CPUs. But the whole point of Linux (on a personal desktop) is to have some fun and try new stuff out easily.
Linux developers really need to stabalize driver interfaces. I should be able to go to kernel.org and download the latest kernel *binary*, then install a binary driver from the CD-ROM that came from my NVIDIA card.
USB and Firewire buses should be exposed by kernel as network interfaces, accessible to user programs through socket API. In this way, USB drivers will be both easier to write/debug AND will not contribute to Oops. For the ultimate of cool, Wine should support Windows USB drivers (my Virtual PC does!) and I should be able to just install Palm and Zaurus desktops and use them rather trying to feed ttyUSBNN to kpilot.
Having a stable system that doesn't have to be rebooted to Windows to use some unsupported USB device is far more important than raw performance. I wish any system's developers - Linux, *BSD, Darwin, BeOS, etc - would concentrate on this goal before going back to play with cool toys.
I bet it will cost the poor guy who just tried to share some pages with friends more than his tuition. Charging for bandwidth is a bad idea, because a lot of traffic - slashdot, worms, buggy research projects, DOS attacks, windows update, P2P and so on - is not something user can control or measure until the bill arrives. Limiting speed is better, especially if the limit only takes effect after you used up your daily quota of unlimitted-speed traffic. In any case, there should be liberal exceptions for students doing research projects (like Internet search, multimedia streaming or P2P protocols) that might use up a lot of bandwidth.
On the other hand, slashdot could easily solve this problem by randomizing the comments while still preserving threads (randomize top level posts, then children of each parent and so on). As a bonus, discussions will no longer be fixated on a single catchy but off-topic thread.