Oil prices at $75 a barrel and worries about global warming are driving the shift.
Aargh! Where do people get the idea that any alternative to petroleum will help reduce global warming?
Any process that generates energy by burning a hydrocarbon procudes CO2. That most certainly includes biofuels.
(In other news, unless you can find a place to mine hydrogen fuel cells, "hydrogen-powered cars" will also not necessarily reduce total CO2 emissions. Those fuel cells have to be charged up somehow.)
It's not quite so simple. Suppose a manufacturer were to build a computer that would only run an OS signed with Linus's key. That turns his "signing key" into an "embedded key". The problem here is that there is no fundamental distinction between the two kinds of keys; it's just a question of how they are used.
That is a nice little bit of math on the web page you link to, but you overlook an important aspect. That paper defines "error rate = false positives + false negatives", and shows that layered testing increases the total error rate. But the paper itself points out that layering can decrease one of those terms, and, depending on your application, the bad-ness of each term can be very different. For example, in return for preventing the very bad situation of a thief getting to my money (false positive id), I am willing to accept a lot of inconvenient situations where I have to repeatedly attempt to authenticate myself (false negative id).
All you are saying is that it is better to get a 10% return than a 1% return on your money, and I certainly don't disagree with that. But you will notice that your example has absolutely nothing to do with the CEO pay. You would want the invesment that returns 10% no matter what the CEO was paid, wouldn't you?
So let's introduce CEO pay into the equation. If you invest $10K an earn $100, a 1% return, you probably wouldn't be too happy, but you would accept that the person managing that money deserved something for his trouble, so perhaps you're willing to pay him $5. If you invest $10K and earn $1K, a 10% return, you're probably much happier; let's say you're willing to pay the person managing that investiment $50. If guy that returns 1% manages a $10B pool of $10K investments and each one pays him the same $5 fee, he earns $5M. If the guy that returned 10% was managing a $100M pool of $10K investiments and each one payed him $50, he only earns $500K. Notice that everyone here pays and gets payed in proportion to what he produces, but the guy who earned a lower percentage returns still earned more, because he managed more money. See how that works?
There is a problem with this study: it measures shareholder return as a percentage, but compensation as a dollar value. If a CEO grows a $10B company by 1%, he generates $100M for shareholders. If a CEO grows a $100M company by 10%, he generates only $10M for shareholders. The study implies that the second CEO deserves to be paid more, because his company had a larger percentage return. But one could certainly make a good argument that the first CEO deserves to be paid more, because he generated a larger absolute return to shareholders.
In fact, given the general trend that smaller companies tend to grow faster than large onces, you would expect the data to look like this even if there is no intrinsic correlation between CEO pay and corporate performance.
I don't write this because I believe that the market for CEO pay works. I write this only because this particular study doesn't prove that the market for CEO pay doesn't work.
I'd mod your reply to my post insightful myself, if I could. You are right, I don't know with any accuracy what percentage of slashdotters are web standards zealots (and what percentage of those are web standards zealots, as another poster implied, only because it allows them to bash MSFT). So I should have said: I wonder how the W3C zealots arround here will react to this depiction of their sacred cow.
The summary mis-represents the bulk of Bjoern's critique, which less about the lack of non-corporate participation and more about the fact that the organization just doesn't work.
I wonder how the bulk of slashdotters, for whom a W3C standard seems to be a sacred cow, will react to the message that these standards are all-too-often ambiguous, bone-headed, poorly supported, slow-moving, and lacking important features.
For a very simple reason: Buffett believes that Bill does a much better job of allocating capitol than your run-of-the-mill charity.
And he is not alone in this belief. The Gates Foundation has made big waves in the non-profit world by replacing a give-away-money-to-feel-better model with a run-a-charity-like-a-business model. That includes setting targets, measuring results, and providing incentives.
I am an (ex-) particle theorist. I worked on phenomenology, which is how particle physicists describe people try to work with actual data.
I don't think the rise of string theory has been the cause of the dearth of breakthroughs in particle physics in the last 30 years, but rather the effect. For all that time, nothing unexpected has come out of accelerator experiments -- just more confirmations of the predictions of the standard model developed in the 1970s, and more accurate measurements of its parameters. In an environment like that, it's no surprise that theoreticans turn to highly speculative and mathematically challenging models to keep their work interesting.
There are still some related fields generating new and interesting data for good young theorists to cut their teeth on -- cosmology, for example.
Like most slashdotters, I feel and instinctive affinity for net neutrality. And I think having a medium where all "content providers" are equal has been great plus, not only for internet culture, but also for the level of competition in internet commerce.
Still, the tremendously increased investment that can be conjured up by the profit motive is nothing to be sneezed at. I was using the internet as a graduate student before there was a web, and I remeber the ruckus over the first advertisment that appeared on usenet. Like most usenet denizens of the time, I was appalled, and I thought that commercialization would destroy our beloved cooperative internet. Obviously, I was dead wrong. So having been proved wrong once, I'm not inclined to dismiss the power of the profit motive to provide us with an infrastructure capable of doing things we haven't even dreamed of yet.
It's a wee bit disingenuous to blame browsers for the lack of strictly validating web pages out there. I'd venture that upwards of 90% of the issues you see when you validate pages against the HTML 4.0 schema are not there because the author had to violate the standard in order to achieve the effect in some non-compliant browser. They are there because the author achieved the effect he wanted and did bother to check whether he had, or could, achieve it in a standards-compliant way. From the beginning, browsers tried to degrade gracefully in the face of invalid input, and as long they do that there will continue to be a lot of invalid input out there.
I cited my source in the parent post: William Stallings, "Operating Systems Internals and Design Principals", Fourth Edition, pp. 86-87. It's a standard CS text.
Certainly a full-on microkernel, as Tannenbaum envisions it, isolates different system services in seperate processes with seperate address spaces. That is the only way you can get some of the advantages he touts, such as the ability to swap out subsystems without a restart and isolation of buffer overruns.
Think about what you have to do to engineer such a system. As Linus says, the big difficulty isn't how to do the message-passing; it's how to dis-entangle the state of your subsystem from the state of the rest of your kernel and define small, stable interfaces that allow the systems to coordinate their activities in real time. That is to say, the hard part of making a micro-kernel is getting the encapsulation and interfaces right. Once you have done that work, putting the systems in seperate processes is easy; it's just a matter of deciding to accept the performance hit of the message-passing.
NT is a "microkernel-inspired" kernel because they have already done the hard part. They have isolated the graphics, networking, file, driver, and other subsystems into silos that interact with the executive via small, well-defined interfaces. They are already at the point where deciding whether to put these subsystems into a seperate process is just a matter of deciding whether to accept the message-passing overhead. From NT 3.1 to NT 4.0, they decided to move the graphics subsystem in to avoid the message-passing overhead. From XP to Vista, they decided to move many serial device drivers out. These are easy changes for them because they have already done the hard work of isolating the state of each of the subsystems.
Linux has not done this work. (And Linus doesn't appear to believe that it would be worthwhile.) That is why, in Stalling's book, "monolithic" appears dozens of times in the Linux section and "microkernel" appears dozens of times in the NT section, even though everyone agrees that NT is not a "full-on" microkernel.
I won't argue over semantics, but obviously at least one author of a standard textbook on OS design does not share your strict definition of what can be called "microkernel-inspired".
But even if we lay aside the loaded term "microkernel" and compare and contrast Linux and NT on the grounds of modularity, encapsulation, and interfaces, we will find that NT adheres to these principals much better than Linux does. Linux's view of these principals is like Perl's: if you want to respect them, that's fine, but it won't enforce them in any way. And, in practice, they are much less respected in Linux than in NT.
So, did VMS have a graphics subsystem in the kernel as well?
No. In NT 3.1, the graphics subsystem ran in user-space. In NT 4.0, it was moved into kernel-mode to avoid the performance hit of the context switch. As this history suggests, the actual architectures of the executive and the graphics subsystem are not tightly coupled. They share an address space for performance reasons, not in order to share state. (To be clear, I don't think this is a great thing, and it is a violation of microkernel principals. But the programmers made the smallest possible departure from microkernel principals to achieve their performance requirement.)
I find it highly amusing that many of the same people who defend the monolithic Linux kernel architecture that tightly couples so many subsystems to the kernel also attack Windows for running the graphics subsystem in kernel mode.
Also can you provide some examples of kernel experts praising the NT kernel for its microkernel properties?
W2K does not have a pure microkernel architecture but what Microsoft refers to as a modified microkernel architecture. As with a pure microkernel architecture, W2K is highly modular. Each system function is managed by just one component of the operating system. The rest of the operating system and all applications access that function through the responsible component using a standard interface. Key system data can only be accessed through the appropriate function. In principle, any module can be removed, upgraded, or replaced without rewriting the entire system or its standard application program interfaces.
William Stallings, "Operating Systems Internals and Design Principals", Fourth Edition, pp. 86-87.
You are forgiven for being wrong, but not for spouting off nonsense despite knowing that you don't know what you're talking about, apparently applying the principal "if my argument involves M$ doing the wrong thing, it must be right".
While neither NT nor Mac OS X are true microkernels, the architecture of both is strongly inspired by microkernel ideas. Like Linus, the developers of these kernels recognized the practical difficulties involved in making full-on microkernels work, but unlike Linus, instead of throwing in the towel completely and doing full-on monolithic kernels, they created cleanly seperated layers interacting via well-defined interfaces whenever they practically could.
If you talk to kernel programmers, most will express a high degree of respect for the NT kernel, which is based on the DEC VMS kernel. It mostly the poor design of systems that sit on top of the kernel that has earned Windows its reputation.
It's not secure.
It's not efficient.
It's not stable.
I find it amusing that, once you found your "standards ueber alles" argument demolished, you proceeded to repeat precisely what I wrote in my orginal post, namely that MSFT, instead of turning off ActiveX, needs to work on making it more secure and less buggy (oh, and you added "efficient" -- fine by me). And yet you still seem to believe that you disagree with me, I guess because I didn't do the requisite "MSFT is evil" genuflections before making my point.
That was supposed to be funny, right? In case not, I'll be pedantic: why have a cross-platfrom method for running native binaries, when the native binaries you want to run aren't corss-platfrom? Or do you refuse to run Linux binaries under Linux because they won't run under Windows?
I suppose you could say that people just shouldn't be able to run native binaries via the web, even if they want to, because that wouldn't be cross-platform. But in doing so, you would be illustrating perfectly why a business would prefer dealing with Microsoft to dealing with OSS zealots.
Business: The web is great, but we want our web app to support drag-n-drop, and there isn't a Javascript standard for that yet. We have a lot of programmers who know how to code against your native GUI APIs, so why don't you give us a mechanism to call into those APIs via our web app?
Microsoft: Okay.
OSS Zealot: No! That's not standards-based! That's not cross-platform!
There are all sort of internal corporate tools for accounting, logistics, mapping, human resources and more that run via ActiveX. With better Javascript, web development frameworks, AJAX, etc. the architects of these applications would probably make different choices if they were to start over today. But 5 years ago ActiveX was the fastest, easiest way to deliver rich functionality via the web, and a lot of people used it. In these intranet scenarios, malicious code was less of a concern, which is probably why the original ActiveX design didn't give that scenario enough thought.
While from an engineering perspective it is often more interesting to start over from scratch, from a business perspective that's often not an option. Any browser manufacturer that cut features that big corporate customers had built against, even Microsoft, risks loosing a significant number of those customers.
The summary implies that the "right" engineering decision would be to eliminate ActiveX. This is complete bullshit.
ActiveX is a mechanism that allows compiled code delivered via the web to run on the client. This feature is an absolute must-have for many corporate environments.
Was Microsoft's ActiveX security framework insufficient? Absolutely. Were their implementation buggy? Yes. Were their security defaults too lax? Certainly. But with a feature as important to your customer base as this, the right solution isn't to cut the feature. It's to fix the problems.
Doubtlessly, MS already uses obfuscation extensively in every one of its published.NET assemblies.
Wrong. None of the.NET Framework library assemblies are obfuscated. (You can find them under C:\WINDOWS\Microsoft.NET if you install the framework.) Feel free to take a look using a decompiler. (My favorite is Lutz.NET Reflector.) When I need to understand a.NET API in more detail, I often find looking at the decompiled source to be quite instructive.
I find it quite telling that a fan of extreme programming, a methodology whoose central tenant is that programming should be done in pairs, is in fact the sole programmer in his organization. I wonder if any fans of XP have actually used it.
Slashdot Mirror
Oil prices at $75 a barrel and worries about global warming are driving the shift.
Aargh! Where do people get the idea that any alternative to petroleum will help reduce global warming?
Any process that generates energy by burning a hydrocarbon procudes CO2. That most certainly includes biofuels.
(In other news, unless you can find a place to mine hydrogen fuel cells, "hydrogen-powered cars" will also not necessarily reduce total CO2 emissions. Those fuel cells have to be charged up somehow.)
It's not quite so simple. Suppose a manufacturer were to build a computer that would only run an OS signed with Linus's key. That turns his "signing key" into an "embedded key". The problem here is that there is no fundamental distinction between the two kinds of keys; it's just a question of how they are used.
I wish slashdot had a +6 extra-brilliant mod. That is ingenious!
That is a nice little bit of math on the web page you link to, but you overlook an important aspect. That paper defines "error rate = false positives + false negatives", and shows that layered testing increases the total error rate. But the paper itself points out that layering can decrease one of those terms, and, depending on your application, the bad-ness of each term can be very different. For example, in return for preventing the very bad situation of a thief getting to my money (false positive id), I am willing to accept a lot of inconvenient situations where I have to repeatedly attempt to authenticate myself (false negative id).
All you are saying is that it is better to get a 10% return than a 1% return on your money, and I certainly don't disagree with that. But you will notice that your example has absolutely nothing to do with the CEO pay. You would want the invesment that returns 10% no matter what the CEO was paid, wouldn't you?
So let's introduce CEO pay into the equation. If you invest $10K an earn $100, a 1% return, you probably wouldn't be too happy, but you would accept that the person managing that money deserved something for his trouble, so perhaps you're willing to pay him $5. If you invest $10K and earn $1K, a 10% return, you're probably much happier; let's say you're willing to pay the person managing that investiment $50. If guy that returns 1% manages a $10B pool of $10K investments and each one pays him the same $5 fee, he earns $5M. If the guy that returned 10% was managing a $100M pool of $10K investiments and each one payed him $50, he only earns $500K. Notice that everyone here pays and gets payed in proportion to what he produces, but the guy who earned a lower percentage returns still earned more, because he managed more money. See how that works?
There is a problem with this study: it measures shareholder return as a percentage, but compensation as a dollar value. If a CEO grows a $10B company by 1%, he generates $100M for shareholders. If a CEO grows a $100M company by 10%, he generates only $10M for shareholders. The study implies that the second CEO deserves to be paid more, because his company had a larger percentage return. But one could certainly make a good argument that the first CEO deserves to be paid more, because he generated a larger absolute return to shareholders.
In fact, given the general trend that smaller companies tend to grow faster than large onces, you would expect the data to look like this even if there is no intrinsic correlation between CEO pay and corporate performance.
I don't write this because I believe that the market for CEO pay works. I write this only because this particular study doesn't prove that the market for CEO pay doesn't work.
I'd mod your reply to my post insightful myself, if I could. You are right, I don't know with any accuracy what percentage of slashdotters are web standards zealots (and what percentage of those are web standards zealots, as another poster implied, only because it allows them to bash MSFT). So I should have said: I wonder how the W3C zealots arround here will react to this depiction of their sacred cow.
The summary mis-represents the bulk of Bjoern's critique, which less about the lack of non-corporate participation and more about the fact that the organization just doesn't work.
I wonder how the bulk of slashdotters, for whom a W3C standard seems to be a sacred cow, will react to the message that these standards are all-too-often ambiguous, bone-headed, poorly supported, slow-moving, and lacking important features.
I left physics after two postdocs. I loved the research and the people, but I was getting too old to put up with low pay and moving every few years.
I am now a software engineer.
For a very simple reason: Buffett believes that Bill does a much better job of allocating capitol than your run-of-the-mill charity.
And he is not alone in this belief. The Gates Foundation has made big waves in the non-profit world by replacing a give-away-money-to-feel-better model with a run-a-charity-like-a-business model. That includes setting targets, measuring results, and providing incentives.
I am an (ex-) particle theorist. I worked on phenomenology, which is how particle physicists describe people try to work with actual data.
I don't think the rise of string theory has been the cause of the dearth of breakthroughs in particle physics in the last 30 years, but rather the effect. For all that time, nothing unexpected has come out of accelerator experiments -- just more confirmations of the predictions of the standard model developed in the 1970s, and more accurate measurements of its parameters. In an environment like that, it's no surprise that theoreticans turn to highly speculative and mathematically challenging models to keep their work interesting.
There are still some related fields generating new and interesting data for good young theorists to cut their teeth on -- cosmology, for example.
Like most slashdotters, I feel and instinctive affinity for net neutrality. And I think having a medium where all "content providers" are equal has been great plus, not only for internet culture, but also for the level of competition in internet commerce.
Still, the tremendously increased investment that can be conjured up by the profit motive is nothing to be sneezed at. I was using the internet as a graduate student before there was a web, and I remeber the ruckus over the first advertisment that appeared on usenet. Like most usenet denizens of the time, I was appalled, and I thought that commercialization would destroy our beloved cooperative internet. Obviously, I was dead wrong. So having been proved wrong once, I'm not inclined to dismiss the power of the profit motive to provide us with an infrastructure capable of doing things we haven't even dreamed of yet.
It's a wee bit disingenuous to blame browsers for the lack of strictly validating web pages out there. I'd venture that upwards of 90% of the issues you see when you validate pages against the HTML 4.0 schema are not there because the author had to violate the standard in order to achieve the effect in some non-compliant browser. They are there because the author achieved the effect he wanted and did bother to check whether he had, or could, achieve it in a standards-compliant way. From the beginning, browsers tried to degrade gracefully in the face of invalid input, and as long they do that there will continue to be a lot of invalid input out there.
I cited my source in the parent post: William Stallings, "Operating Systems Internals and Design Principals", Fourth Edition, pp. 86-87. It's a standard CS text.
Certainly a full-on microkernel, as Tannenbaum envisions it, isolates different system services in seperate processes with seperate address spaces. That is the only way you can get some of the advantages he touts, such as the ability to swap out subsystems without a restart and isolation of buffer overruns.
Think about what you have to do to engineer such a system. As Linus says, the big difficulty isn't how to do the message-passing; it's how to dis-entangle the state of your subsystem from the state of the rest of your kernel and define small, stable interfaces that allow the systems to coordinate their activities in real time. That is to say, the hard part of making a micro-kernel is getting the encapsulation and interfaces right. Once you have done that work, putting the systems in seperate processes is easy; it's just a matter of deciding to accept the performance hit of the message-passing.
NT is a "microkernel-inspired" kernel because they have already done the hard part. They have isolated the graphics, networking, file, driver, and other subsystems into silos that interact with the executive via small, well-defined interfaces. They are already at the point where deciding whether to put these subsystems into a seperate process is just a matter of deciding whether to accept the message-passing overhead. From NT 3.1 to NT 4.0, they decided to move the graphics subsystem in to avoid the message-passing overhead. From XP to Vista, they decided to move many serial device drivers out. These are easy changes for them because they have already done the hard work of isolating the state of each of the subsystems.
Linux has not done this work. (And Linus doesn't appear to believe that it would be worthwhile.) That is why, in Stalling's book, "monolithic" appears dozens of times in the Linux section and "microkernel" appears dozens of times in the NT section, even though everyone agrees that NT is not a "full-on" microkernel.
I won't argue over semantics, but obviously at least one author of a standard textbook on OS design does not share your strict definition of what can be called "microkernel-inspired".
But even if we lay aside the loaded term "microkernel" and compare and contrast Linux and NT on the grounds of modularity, encapsulation, and interfaces, we will find that NT adheres to these principals much better than Linux does. Linux's view of these principals is like Perl's: if you want to respect them, that's fine, but it won't enforce them in any way. And, in practice, they are much less respected in Linux than in NT.
So, did VMS have a graphics subsystem in the kernel as well?
No. In NT 3.1, the graphics subsystem ran in user-space. In NT 4.0, it was moved into kernel-mode to avoid the performance hit of the context switch. As this history suggests, the actual architectures of the executive and the graphics subsystem are not tightly coupled. They share an address space for performance reasons, not in order to share state. (To be clear, I don't think this is a great thing, and it is a violation of microkernel principals. But the programmers made the smallest possible departure from microkernel principals to achieve their performance requirement.)
I find it highly amusing that many of the same people who defend the monolithic Linux kernel architecture that tightly couples so many subsystems to the kernel also attack Windows for running the graphics subsystem in kernel mode.
Also can you provide some examples of kernel experts praising the NT kernel for its microkernel properties?
W2K does not have a pure microkernel architecture but what Microsoft refers to as a modified microkernel architecture. As with a pure microkernel architecture, W2K is highly modular. Each system function is managed by just one component of the operating system. The rest of the operating system and all applications access that function through the responsible component using a standard interface. Key system data can only be accessed through the appropriate function. In principle, any module can be removed, upgraded, or replaced without rewriting the entire system or its standard application program interfaces.
William Stallings, "Operating Systems Internals and Design Principals", Fourth Edition, pp. 86-87.
You are forgiven for being wrong, but not for spouting off nonsense despite knowing that you don't know what you're talking about, apparently applying the principal "if my argument involves M$ doing the wrong thing, it must be right".
While neither NT nor Mac OS X are true microkernels, the architecture of both is strongly inspired by microkernel ideas. Like Linus, the developers of these kernels recognized the practical difficulties involved in making full-on microkernels work, but unlike Linus, instead of throwing in the towel completely and doing full-on monolithic kernels, they created cleanly seperated layers interacting via well-defined interfaces whenever they practically could.
If you talk to kernel programmers, most will express a high degree of respect for the NT kernel, which is based on the DEC VMS kernel. It mostly the poor design of systems that sit on top of the kernel that has earned Windows its reputation.
...oh wait, I guess it really is!
I find it amusing that, once you found your "standards ueber alles" argument demolished, you proceeded to repeat precisely what I wrote in my orginal post, namely that MSFT, instead of turning off ActiveX, needs to work on making it more secure and less buggy (oh, and you added "efficient" -- fine by me). And yet you still seem to believe that you disagree with me, I guess because I didn't do the requisite "MSFT is evil" genuflections before making my point.
That was supposed to be funny, right? In case not, I'll be pedantic: why have a cross-platfrom method for running native binaries, when the native binaries you want to run aren't corss-platfrom? Or do you refuse to run Linux binaries under Linux because they won't run under Windows?
I suppose you could say that people just shouldn't be able to run native binaries via the web, even if they want to, because that wouldn't be cross-platform. But in doing so, you would be illustrating perfectly why a business would prefer dealing with Microsoft to dealing with OSS zealots.
Business: The web is great, but we want our web app to support drag-n-drop, and there isn't a Javascript standard for that yet. We have a lot of programmers who know how to code against your native GUI APIs, so why don't you give us a mechanism to call into those APIs via our web app?
Microsoft: Okay.
OSS Zealot: No! That's not standards-based! That's not cross-platform!
There are all sort of internal corporate tools for accounting, logistics, mapping, human resources and more that run via ActiveX. With better Javascript, web development frameworks, AJAX, etc. the architects of these applications would probably make different choices if they were to start over today. But 5 years ago ActiveX was the fastest, easiest way to deliver rich functionality via the web, and a lot of people used it. In these intranet scenarios, malicious code was less of a concern, which is probably why the original ActiveX design didn't give that scenario enough thought.
While from an engineering perspective it is often more interesting to start over from scratch, from a business perspective that's often not an option. Any browser manufacturer that cut features that big corporate customers had built against, even Microsoft, risks loosing a significant number of those customers.
The summary implies that the "right" engineering decision would be to eliminate ActiveX. This is complete bullshit.
ActiveX is a mechanism that allows compiled code delivered via the web to run on the client. This feature is an absolute must-have for many corporate environments.
Was Microsoft's ActiveX security framework insufficient? Absolutely. Were their implementation buggy? Yes. Were their security defaults too lax? Certainly. But with a feature as important to your customer base as this, the right solution isn't to cut the feature. It's to fix the problems.
Wrong. None of the .NET Framework library assemblies are obfuscated. (You can find them under C:\WINDOWS\Microsoft.NET if you install the framework.) Feel free to take a look using a decompiler. (My favorite is Lutz .NET Reflector.) When I need to understand a .NET API in more detail, I often find looking at the decompiled source to be quite instructive.
I find it quite telling that a fan of extreme programming, a methodology whoose central tenant is that programming should be done in pairs, is in fact the sole programmer in his organization. I wonder if any fans of XP have actually used it.