Are people buying Pentium 4s to run DOS in Real Mode?
Not far from it, even though it's not DOS. Every bootloader currently in existence for IBM PC compatibles uses real mode and BIOS calls, however.
Hopefully that will change with either Intel's plans for a new PC firmware, or (preferably) LinuxBIOS/OpenBIOS.
Also, you have to realize that virtually all "freeware" and "shareware" programs for Windows require binary compatibility. Not to mention Windows itself...
Mind you, you are completely wrong when you say that IA32 CPUs "are basically RISC". See my previous post from when Apple announced their switch to Intel for why that is true.
In other news, Microsoft's PR office apologize for the disability of some of their staff to spell. They also promise that, while their editor staff won't be updated until the release of Microsoft Windows Longhorn, it most certainly will make it into the company's version of Windows coming after Longhorn, currently code-names "Eventually".
That really wasn't the original design goal of Unix, though. The original design goal of Unix was for Ken Thompson to be able to play a really funny little game that he had developed while working on Multics.
The first few years, Unics, like all OSs of the day, was indeed written in assembly. Even after it was rewritten in C, it wasn't portable for another few years, since it still relied on stuff that only worked on PDP-11s. It was only when they decided to try and port it to another architecture (which I don't remember right now which one it was) that the portability issues were sorted.
Quite simply, Intel no longer uses CISC. Sure the instruction set is CISC, [...]
I don't know if you noticed, but the IS in CISC and RISC does stand for "Instruction Set". Thus, IA32 is still CISC by definition. The fact that the executional units are executing CISC as broken down into RISC is quite irrelevant, since it is the instruction set itself that matters. Microcode was invented long before the term RISC was coined. Even the IBM System/360 was using Microcode.
The reason RISC was "invented" was, therefore, not as some new, revolutionary architecture, since most CPUs even then were RISC "at the core" in the same way that the IA32 is now. RISC was invented since higher-level languages like C became more and more popular instead of writing assembly. Since almost noone was writing assembly anymore, there was little need for all the exotic instructions and addressing modes that were present in CISC computers for programmer convenience.
Because of that, CPU designers recognized that by cutting the microcode decoding logic and in essence making the C compiler output microcode directly, they could use the freed die area to implement more cache, larger TLBs, more registers, and also tune the RISC instruction more than the CISC instructions could be.
In fact, I believe that even the 80386 ran microcode, even though it couldn't be reprogrammed, like the Pentium II and upwards are capable of. Noone would design a CISC chip that doesn't run microcode -- it's just too inefficient.
So as you can see, IA32 chips are in no way RISC.
As for MMX, I'm not sure what you mean by "dead". It certainly exists in all modern IA32 designs, so in that sense it is not "replaced" by SSE, SSE2 (which you are calling SIMD and SIMD2), SSE3, MMX2, 3DNow or 3DNowExt. That insanely large variety of IA32 SIMD processing units is also, from what I've gathered, the reason the most programmers seem to prefer the PPC Altivec unit before any IA32 SIMD.
If you mean that MMX is dead in the sense of being unused, I can't honestly say that I really know, but I do know that the Linux kernel implements a CRC acceleration function with MMX, which seems to often be the fastest CRC function, so I can't say that it seems very dead in that regard either. Furthermore, I believe that MMX was invented not for games or video effects, but rather for video or sound processing in media players, and I know that MPlayer uses it.
While all which you say is doubtlessly true, the fact remains that MS Office in stand-alone mode may well be the primary reason for people to keep Windows.
If Microsoft starts using open file formats, there will be far less reason for people not to be using a Linux desktop instead.
Even though my 20" CRT is from 1996 and I have to knock a circuit board inside it every now and then to keep the voltage regulator for the cathode heater working properly, they're going to need one h*ll of a toolbox to pry it from those cold dead hands of mine.
Even though LCD technology has come a long way, I still have three major problems with them:
The color purity is no match for even a half-assed CRT.
Due to the angel dependency of LCD pixels, the colors look different at the edges.
It's really hard (although not impossible, admittedly) to find an LCD monitor that sports a reasonable resolution. My 20" CRT can run 1600x1200 (or 2048x1536, but only at 60 Hz). I find it hard to find even a 20" LCD monitor that's capable of more than 1400x1050.
And then there are dead pixels and the fact that they only work well at one resolution (even though I'm beginning to care less and less for the latter). Not to mention an LCD costing roughly twice as much as a CRT that it's still only half as good as.
Mind also, how virtually the only actual problems with CRT monitors versus LCD monitors are that they are huge and produce more heat. I'd say the issues of picture quality beat those problems any day -- hands tied.
Given all that, there's no way I'm giving up CRT monitors -- at least not until I'm beginning to see good, cheap OLED monitors.
Oh, and long live Trinitron tubes.
Re:The article sounded reasonable until:
on
A Gamer's Manifesto
·
· Score: 1
I'm not sure that I get this. I was very surprised when I read this in TFA -- aren't PPCs out-of-order by nature? Is out-of-order execution just some kind of "plugin" or similar for PPCs? Can someone shed some light on this?
In all else I agree with the parent -- why should in-order execution hinder unpredictability? It's not like it makes the CPU less Turing-complete...:-?
[S]ome of the security work in IE7 relies on operating system functionality in XPSP2 that is non-trivial to port back to Windows 2000.' While security fixes will still be available until 2010...
Isn't this quite interesting? What makes IE7 impossible to back-port is the security features in XPSP2, but even so, the security fixes is what will be available for Win2k. One would think that if it depended on the XPSP2 security features, it would be the functionality, but not the security, that would be avaiable to non-XPSP2 systems?
Is it only me, or is something strange along those lines?
Was it on the lower parts on the mountain, or do you just mean that we should get low to see it -- after all, it was near the grou... oh wait! I see, you meant "lo and behold"!
It would be interesting if releases like this significantly boost the IPv6 take-up rate but as far as I know, Earthlink doesn't supply end-to-end IPv6 yet.
Have you tried checking if they support the IPv4-to-IPv6 anycast router address 192.88.99.1? If they do, you can set up a 6to4 tunnel Real Easy (R).
Just set up an IPv6 tunnel (Linux SIT tunnels support this natively), and point it to 192.88.99.1 to send to non-6to4 addresses. Other 6to4 destinations will be auto-tunnelled with IPv6-over-IPv4, and any IPv6 packets sent to you will also be automatically routed over IPv6-over-IPv4 by the Internet. Therefore, there's no need to set up a tunnel with a third party if you're using 6to4.
Fedora Core supports 6to4 more or less out-of-the-box. All you need to do are two things:
1. Add these lines to/etc/sysconfig/network (why does Slashdot split the lines?):
Except it doesn't work like that. IPv6 uses a hierarchical routing model, much stricter even than IPv4 classful routing.
Ever heard of IPv6 Mobility? I haven't read the RFC myself, but from what I've gathered, when you are somewhere else in the world, you use the IPv6 address at that site to contact your "home agent" back home, and the two of you set up an IPSec/IPv6 tunnel between each other, and you get to use an IPv6 address from your subnet at home -- probably a static one. Remember, an IPv6 node is supposed to have multiple addresses. The major advantage is that you can travel between different wireless hotspots and actually keep your address. In other words, you don't even need the "worldID" card that the GP was talking about -- just an account on your home agent.
The other common misconception is that IPv6 has more addresses (2^128) than particules in the known universe. This isn't really true as the lower 64 bits are not routable.
Your point being...? This is, of course, the entire point of having so many addresses. 64-bits of routable addresses is still far more than will ever be enough to create a more than satisfactory routing model, while the link-local lower 64 bits will be more than will ever be enough to allow any number of devices on the uplink provided by the ISP. That was the point from the beginning, so I don't really know what you're trying to debunk. That there aren't more addresses than there are particles in the known universe? Now that would be a real loss -- now we only have as many as there are particles in our own planetary system!
Well, I'm running Gentoo 2005.0 over here, and I put "ipv6" in my USE flags. I can't think of a single program (that is at least a little significant) that does not support IPv6 on my system. Firefox does, apache does, MIT Kerberos does, OpenSSH does, netkit-telnet does, wget does, gnomemeeting does, xinetd does, etc.
Oh yeah, I forgot -- Most Windows programs don't.
However, I heard rumors that Longhorn will support IPv6 out of the box. So whenever Longhorn is released -- 2010?;-) -- we'll have it.
That's not strange or unexpected by any means, though -- How could a user appreciate something that (s)he doesn't even know what it is (that is, the O/S)?
The point that I believe that the GP wanted to make is this: It's far easier to find a user who is passionate about Linux/UNIX/OSX/WhatHaveYou, while it's very hard to find a user who is passionate about Windows. Even though there are far more users of Windows in total!
Oh.. and they all use IP. Which would make IP the one mainly responsible for the IP-theft! Yup. Sounds like double-A logic to me.
Sigh... Small-minded people like always miss the big picture.
It is obviously the physical universe that is responsible for disemanating Episode III. Without the physical universe, there would be no IP thieves and thus no IP theft.
The only solution is evidently to outlaw the physical universe. I'd like to officially propose to MPAA that we all work together to cause a total existence failure of the entire universe. I hope noone disagrees?
Another interesting implication of that analogy is that people usually don't just get into a car for the first time in his or her life and expect to be able to drive -- they go through rigorous training to get a driver's license first.
This is unlike computers, where O/S manufacturers like Microsoft more or less design the O/S so that people are to be able to use the computer the first time they see it. At least last I looked, that was one of their mission statements.
Imagine if car dealers had similar mission statements -- "People should be able to use our cars immediately, even if they've even been in a driver's seat before" -- how many more traffic accidents there would be.
Maybe there should be a nationally standardized education that corresponds to something like a "computer user's license". Of course, unlike cars, it shouldn't be mandatory just in order to use a computer, but it could be a good idea for ISPs and prospective bosses to check for it.
A Scottish company, Ocean Power Delivery (OPD) and it's Norwegian backer
Editor's: Im glad to see that youre capable of correcting the posters use of apostrophe's. Its too much to assume that the poster's would get thei'r grammars right anyway.
On the other hand, you could manually use UTF-8 in Ruby. PHP also does a (very) poor job at Unicode, and thus my solution is to simply encode all my strings with UTF-8. It has worked very well for me so far.
If you were to ask for my opinion on IRC, I think the entire protocol is so abysmally ugly that you don't just need to standardize on a charset for it, you need to rewrite it from scratch. Seriously, IRC is probably the ugliest protocol I've ever seen, rivalled only by the Direct Connect protocol.
Most of us probably won't have to "deal" with it, since most (all?) systems except Java handle it natively. I know for sure that GNU libc (and therefore all Linux distros) has had a 32-bit wchar_t type since... ever? I'd be exceedingly surprised if Mac OS X can't handle it since they claim Unicode 3 support, and although I'm dubious of everything that Microsoft does, I believe that Windows also supports more than the Basic Multilingual Plane.
So, unless you're using Java applications, this most likely won't be a problem, except, of course, for actually finding a font that contains all the glyphs you want to display.
Mind you that Unicode was never defined as 16-bit, and neither is it defined as 32-bit today. It just defines code points as natural numbers, with little or no regard to the maximum number of these. That's why wchar_t is defined as an opaque data type, which can be changed at will with only a recompilation required for any program that uses it correctly to work. At the time Java was developing, it just happened so that only the BMP was defined, and thus the Java devs thought that Unicode wouldn't extend beyond 16 bits.
Lastly, I should add that UTF-8 will still be used to encode Unicode, no matter how many bits it goes to. Mind you, you don't need UTF-16 to encode the BMP codepoints. Thus, you don't need UTF-32 to encode beyond the BMP. The entire purpose of UTF-8 is to encode longer-than-8-bit-codepoints to octet sequences (in an ASCII-compatible manner). What you are referring to is probably "Here comes UCS-4", but the fact is that it doesn't... it's been here for a very long time.
You know, I have to wonder -- why care about the security of Microsoft programs? There's so much more that is so fundamentally wrong with the architecture of any Microsoft program that security seems like a rather secondary matter.
To me, it seems almost like discussing the problems of intellectual property in communism. There are so many other, much more important, issues about communism -- it's founded on an absurd philophical model and a historical perspective that's outright wrong. It's pointless to begin discussing IP issues, since there are so much more fundamental problems to discuss before that even becomes relevant.
Who cares if the security model of Windows is absurd? I mean, seriously, what can you expect from a system that is not only absurdly abstracted, but even abstracted in 50 redundant ways, and which has a kernel-level windowing system, a monolithical kernel, etc. ad nauseam. Is security even an issue?
Slashdot Mirror
Hopefully that will change with either Intel's plans for a new PC firmware, or (preferably) LinuxBIOS/OpenBIOS.
Also, you have to realize that virtually all "freeware" and "shareware" programs for Windows require binary compatibility. Not to mention Windows itself...
Mind you, you are completely wrong when you say that IA32 CPUs "are basically RISC". See my previous post from when Apple announced their switch to Intel for why that is true.
What is it called then? I hope the intention came through, at least.
The first few years, Unics, like all OSs of the day, was indeed written in assembly. Even after it was rewritten in C, it wasn't portable for another few years, since it still relied on stuff that only worked on PDP-11s. It was only when they decided to try and port it to another architecture (which I don't remember right now which one it was) that the portability issues were sorted.
The reason RISC was "invented" was, therefore, not as some new, revolutionary architecture, since most CPUs even then were RISC "at the core" in the same way that the IA32 is now. RISC was invented since higher-level languages like C became more and more popular instead of writing assembly. Since almost noone was writing assembly anymore, there was little need for all the exotic instructions and addressing modes that were present in CISC computers for programmer convenience.
Because of that, CPU designers recognized that by cutting the microcode decoding logic and in essence making the C compiler output microcode directly, they could use the freed die area to implement more cache, larger TLBs, more registers, and also tune the RISC instruction more than the CISC instructions could be.
In fact, I believe that even the 80386 ran microcode, even though it couldn't be reprogrammed, like the Pentium II and upwards are capable of. Noone would design a CISC chip that doesn't run microcode -- it's just too inefficient.
So as you can see, IA32 chips are in no way RISC.
As for MMX, I'm not sure what you mean by "dead". It certainly exists in all modern IA32 designs, so in that sense it is not "replaced" by SSE, SSE2 (which you are calling SIMD and SIMD2), SSE3, MMX2, 3DNow or 3DNowExt. That insanely large variety of IA32 SIMD processing units is also, from what I've gathered, the reason the most programmers seem to prefer the PPC Altivec unit before any IA32 SIMD.
If you mean that MMX is dead in the sense of being unused, I can't honestly say that I really know, but I do know that the Linux kernel implements a CRC acceleration function with MMX, which seems to often be the fastest CRC function, so I can't say that it seems very dead in that regard either. Furthermore, I believe that MMX was invented not for games or video effects, but rather for video or sound processing in media players, and I know that MPlayer uses it.
For further reading:
If Microsoft starts using open file formats, there will be far less reason for people not to be using a Linux desktop instead.
Even though my 20" CRT is from 1996 and I have to knock a circuit board inside it every now and then to keep the voltage regulator for the cathode heater working properly, they're going to need one h*ll of a toolbox to pry it from those cold dead hands of mine.
Even though LCD technology has come a long way, I still have three major problems with them:
- The color purity is no match for even a half-assed CRT.
- Due to the angel dependency of LCD pixels, the colors look different at the edges.
- It's really hard (although not impossible, admittedly) to find an LCD monitor that sports a reasonable resolution. My 20" CRT can run 1600x1200 (or 2048x1536, but only at 60 Hz). I find it hard to find even a 20" LCD monitor that's capable of more than 1400x1050.
And then there are dead pixels and the fact that they only work well at one resolution (even though I'm beginning to care less and less for the latter). Not to mention an LCD costing roughly twice as much as a CRT that it's still only half as good as.Mind also, how virtually the only actual problems with CRT monitors versus LCD monitors are that they are huge and produce more heat. I'd say the issues of picture quality beat those problems any day -- hands tied.
Given all that, there's no way I'm giving up CRT monitors -- at least not until I'm beginning to see good, cheap OLED monitors.
Oh, and long live Trinitron tubes.
In all else I agree with the parent -- why should in-order execution hinder unpredictability? It's not like it makes the CPU less Turing-complete... :-?
No, it's because they're dull, you twit! It'll hurt more! *Spitting*
Is it only me, or is something strange along those lines?
It seems that the extra processors could be great for accelerating things like SSL encryption.
Just set up an IPv6 tunnel (Linux SIT tunnels support this natively), and point it to 192.88.99.1 to send to non-6to4 addresses. Other 6to4 destinations will be auto-tunnelled with IPv6-over-IPv4, and any IPv6 packets sent to you will also be automatically routed over IPv6-over-IPv4 by the Internet. Therefore, there's no need to set up a tunnel with a third party if you're using 6to4.
Fedora Core supports 6to4 more or less out-of-the-box. All you need to do are two things: /etc/sysconfig/network (why does Slashdot split the lines?):
2. Add these lines to the1. Add these lines to
Oh yeah, I forgot -- Most Windows programs don't.
However, I heard rumors that Longhorn will support IPv6 out of the box. So whenever Longhorn is released -- 2010? ;-) -- we'll have it.
The point that I believe that the GP wanted to make is this: It's far easier to find a user who is passionate about Linux/UNIX/OSX/WhatHaveYou, while it's very hard to find a user who is passionate about Windows. Even though there are far more users of Windows in total!
It is obviously the physical universe that is responsible for disemanating Episode III. Without the physical universe, there would be no IP thieves and thus no IP theft.
The only solution is evidently to outlaw the physical universe. I'd like to officially propose to MPAA that we all work together to cause a total existence failure of the entire universe. I hope noone disagrees?
This is unlike computers, where O/S manufacturers like Microsoft more or less design the O/S so that people are to be able to use the computer the first time they see it. At least last I looked, that was one of their mission statements.
Imagine if car dealers had similar mission statements -- "People should be able to use our cars immediately, even if they've even been in a driver's seat before" -- how many more traffic accidents there would be.
Maybe there should be a nationally standardized education that corresponds to something like a "computer user's license". Of course, unlike cars, it shouldn't be mandatory just in order to use a computer, but it could be a good idea for ISPs and prospective bosses to check for it.
If you were to ask for my opinion on IRC, I think the entire protocol is so abysmally ugly that you don't just need to standardize on a charset for it, you need to rewrite it from scratch. Seriously, IRC is probably the ugliest protocol I've ever seen, rivalled only by the Direct Connect protocol.
So, unless you're using Java applications, this most likely won't be a problem, except, of course, for actually finding a font that contains all the glyphs you want to display.
Mind you that Unicode was never defined as 16-bit, and neither is it defined as 32-bit today. It just defines code points as natural numbers, with little or no regard to the maximum number of these. That's why wchar_t is defined as an opaque data type, which can be changed at will with only a recompilation required for any program that uses it correctly to work. At the time Java was developing, it just happened so that only the BMP was defined, and thus the Java devs thought that Unicode wouldn't extend beyond 16 bits.
Lastly, I should add that UTF-8 will still be used to encode Unicode, no matter how many bits it goes to. Mind you, you don't need UTF-16 to encode the BMP codepoints. Thus, you don't need UTF-32 to encode beyond the BMP. The entire purpose of UTF-8 is to encode longer-than-8-bit-codepoints to octet sequences (in an ASCII-compatible manner). What you are referring to is probably "Here comes UCS-4", but the fact is that it doesn't... it's been here for a very long time.
To me, it seems almost like discussing the problems of intellectual property in communism. There are so many other, much more important, issues about communism -- it's founded on an absurd philophical model and a historical perspective that's outright wrong. It's pointless to begin discussing IP issues, since there are so much more fundamental problems to discuss before that even becomes relevant.
Who cares if the security model of Windows is absurd? I mean, seriously, what can you expect from a system that is not only absurdly abstracted, but even abstracted in 50 redundant ways, and which has a kernel-level windowing system, a monolithical kernel, etc. ad nauseam. Is security even an issue?