You see, by releasing it as a 6.0 (not beta, but just as a version), people will download it, and not download anything for a while...
I totally agree. However I really know I hate NS4.x even more and therefore I would want to see NS6.0 really soon now. Solution I would suggest is to make it nag after six months or so to force people to upgrade (it's annoying but it's effective). How do you think MS has gotten rid of IE3/4 so well? _________________________
You just are a Unix wannabe. Learn how to configure your fonts before you come back. There are all these documents on your system, all these hints on the web and you still don't have a fscking clue how to set up your system.
Sure, I'm able to setup font servers and build fontdirs but why must it be so difficult? It's only a few files or if you use TT fonts it's only one file for one font. Why it doesn't work like copy the file to/usr/X11/fonts/. and be done with it? I have yet to see good font handling in any system but IMO windows and mac os have usable ones. _________________________
The Unix answer to non-anti-aliased fonts is (and long has been) an insanely high resolution
Yeah. Unfortunately my CRT only supports up to 1600x1200 resolutions in usable refresh rates. Therefore I would really want to use anti-aliasing.
Anyhow, I fail to see how X the display protocol is any worse than the win32 GDI for building browsers.
Two words: font rendering. Do you have any idea how much faster GDI is in font rendering? Compare Mozilla rendering speeds between windows and linux in similar hardware. Windows version of Mozilla is much faster than any version of IE. linux version on the other hand is almost as slow as Netscape 4.x under windows. The cause? Font rendering takes time. Network transparency is cool feature and I even use it every now and then. But I would sacrifice it any day for decent font rendering. Oh well, I think I need to wait berlin. _________________________
Could somebody explain what's the point, please? Why would I want to code in assembler if it will still be executed by VM or compiled into native code? How is this supposed to compile into faster native code than java application compiled into native code? [Java is higher level language and therefore compiler has *more* space to make it's optimizations. If you code in virtual assembler that is then compiled into native assembler compiler doesn't have that many ways to optimize code.]
I think it's because of the example - graphics with alpha channel - people get so excited about this. And this is 'cool' only if you are limited to X11! Even in that simple example the code looks pretty hard to understand when compared to equivalent C(++) code. In the end, if all you are interested in is speed you have to use native assembler. Otherwise simply use C(++), Perl, Python or Java - all of those have "unlimited registers" also. _________________________
There's also the issue of using ethernet MAC addresses in the last section of the address, which would uniquely identify individual computers (and therefore attach your "fingerprint" to everything you do on the net).
How about I change my MAC address? Get root and type in ifconfig eth0 hw addr 01:02:03:04:05:06. Just got yourself another MAC address. Do this like once a minute and it can be quite hard to track you down. Of course it breaks many other things but I'm just trying to tell that MAC address is not hardwired address and therefore shouldn't be used like one. [I found this information here.]
_________________________
Links is text WWW browser, similar to Lynx. Links displays tables, downloads on background and uses HTTP/1.1 keepalive connections.
Get it from freshmeat. I can warmly recommend it over lynx any day [screenshot ]. w3m is worth a try also but I prefer Links to it. _________________________
It might be that I didn't understood correctly how 1T-SRAM works, but I think it's only SDRAM made look like SRAM for memory controller.
Therefore it has to be *really* fast internally with extra logic and cannot possibly be cheap. It sure is fast but far too expensive for system memory. Nintendo is using it for graphics adapter memory - I think this is the least valueable location you will [ever] see this type of memory.
If 1T-SRAM can really be done (without a tweak like this) it could really change situation. Then it would be like DRAM without refresh cycles and without capasitors. I doubt that's possible but even 2T-SRAM would be a dream come true. Remember that as long as we are talking about desktop computers it really doesn't mean how much power those are using but how fast memory you can get economically. It doesn't matter if it's 1T-SRAM (tweaked DRAM) or 6T-SRAM (real SRAM) if it costs the same [as long as speed is the same]. _________________________
I can imagine how hard coding programs for this can be. For example if we want to make top-like program for viewing what our processor is doing or killing netscape. Accoding to article (which could be wrong...):
The use of the process server is optional; a process need not register itself with the process server, in which case others cannot obtain information on it.
Oops. It's pretty hard to find "netscape". You didn't have process server running during netscape start or this version of netscape doesn't support process servers. Though luck - maybe we just kill some PIDs at random and hope for the best.
If all services are provided by other servers how can I code anything? Try to read file? Sorry fs server is not running. Try to setup alarm? Sorry time server is not running. Of course one can require POSIX compatible system but what's the point running micro kernel if you always have basic set of servers (everything required to be POSIX compatible).
Idea of translators described in the article sounds nice, but if you think about it it's only mounting in user space. I don't know how mounting is done internally, but I have often wondered why I have to be root to be able to do it. In the end I cannot imagine any good reason why this couldn't be done with Linux. Monolithic kernel with device drivers as modules as the current trend with linux seems to be sounds like the best solution for me. Though I would break some quirks inherited from UNIX. _________________________
couldn't they have borrowed the ILM/Illusion Arts team that did Dragonheart?
They *could have*, but because all nerds are going to view this anyway why bother. After all everyone knows that while ILM does really good work their work is also really expensive.
On the other hand it could be that people in charge have no clue what CGI should look like nowadays and this was not a low cost solution only. _________________________
the logical conclusion of the argument is that we should have one program that does everything.
Except that one-tool-does-all solution is seldom the best solution for a single problem. It's like using a pocket knife for felling instead of chainsaw - it works but it's not the best way. It's the same thing with emacs: you can play games, write mail and read slashdot with it. However, I find it more sensible to use special programs for all of those.
What I'm currently missing is a small cross-platform text editor with a good interface. Sorry, emacs is out - I want *small* editor with *good interface* you remember? The problem is that it's that hard to make such a thing that those who have tried have found it easier to accustom themselves to use emacs. Unfortunately.
_________________________
...will tell you that 128Kbps MP3 is "CD quality". Fact is, it's inferior to Minidisc, to FM radio and--in many respects--analog cassettes.
I'm surely happy I don't own "golden ears". I'm happy with 192kbps (or even 128kbps) MP3s. They are not as good as CDs but good enough. And to my ears those sound better than my Sony(R) Minidisc(tm). FM radio is actually pretty good also (I listen it through cable without antenna problems). IMHO any one of these will beat any analog cassette system. _________________________
The whole point of a certificate is to prove identity.
Yeah - but if I'm only interested in encrypted communication. Can I do this without a costly certification? And I wouldn't want to cause those scary warnings to user either. (I know that I lose some of the security this way because I cannot be sure that the other end is what I think it is.) _________________________
Wow - you must have a *really* good ad! 20e6 ppl is far from 100k pages per day. Back to topic: say those 100k pages are served during one hour - it makes 100k/60/60 or 28 pages per second. You still can easily run quite heavy scripting.
_________________________
The point is one shouldn't need to generate (almost) any code. Never. If your API needs exact the same code each time you want to make and MDI app then one should not make source code generator. Instead API should be extended so that you could simply type new application.MDI() instead of lots of generated code. Why do you think anyone ever come with an idea of functions? Because their copy-paste keys wasn't working (compare: code generator)?
_________________________
And what is more I think that this test should have tested DDR and SDRAM in the same system if this was supposed to test the speed increase from DDR.
OTOH if this was supposed to be MB test both boards should have used SDRAM. I'm afraid that current test has no real information whatsoever because pre-production MB is slower - but DDR memory is faster.
I would have wanted to see at least test where AMD760 was running SDRAM to have real results about DDR being faster.
_________________________
We already are off loading video processing to specialty built video hardware. Were not doing this very much with sound (yes, you can get you high end cards to add base and sound fields, but you cant do OpenGL like sound calls - play the sound I uploaded to you before like it was coming from (x,y,z) with this ambient sound....)
How about OpenAL. I think we should have also OpenIL (Input Library) and perhaps OpenFL (Feedback Library) for control devices - think about you could treat keyboards, joysticks and insert-your-favorites-here input devices as one from programming viewpoint.
_________________________
The current speed record for a digital flip-flop is 770GHz..
That's nice and all but it really doesn't matter how fast *single* transistor can work. The problem is that the whole circuit must work at given speed. And that speed is limited by the speed of light.
For example say our chip is 10mm x 10mm and we have to send signal from side to side during clock cycle: the time required for signal to get on the other side is distance/speed = (0.10m)/(299792458m/s) = 3.34e-10s. Now if we need to send signal like this every clock the maximum clock speed we can achieve is 1/(distance/speed) = 10*299792458 Hz which equals to less than 3GHz chip.
Of course chip designers are aware of this and design chips in a way no signal needs to be send across the whole chip, but even if greatest distance needed to send signal through - during one clock cycle - is 1mm (one 10th of the chip) we can only get 30GHz - and only if this is our only bottleneck. And also in this case average throughput time is at least 10 clock cycles (time for "operation" to go through the chip).
In the end one should notice that the problem of memory being too slow compared to processor isn't getting away in the future because we surely need those signals from our memory chips and yet again we are limited by the speed of light. Expect to see memory chips really near CPU in the future...
The only way to always double the computing power in the (not so distant) future is to invent a way to transfer *information* faster than light. If it's possible - I don't know. _________________________
but from 7' - 15' away (the average distance between a couch and TV)
The difference is you have to watch TV from distance because it has such a lousy picture because of it's loysy CRT. I'm using my 19" monitor to watch TV from about a 3' away and the only problem is that resolution of PAL systems is low (still higher than NTSC I think) and interlaced. Think about the future and HDTV. How much will HDTV set cost because it should be able to show resolutions equal to monitors and be huge in the same time?
If I had a choice I would use a headset with resolution of 1024x768 per eye or more - with stereo vision. Then simply place those tiny screens near enough and you have that big TV you ever need.
_________________________
If it helps, think of the Registry as a virtual file system, that happens to reside in a single file on disk
Exactly! But then why to do virtual file system on top of file system? One can control file access by file level too. Why not simply use each key as directory and each value as text file (key=filename, value=file contents)? Think about/proc. My guess is that MS tried this and noticed that their file system was that poor they had to try another (IMHO unsuccessful) way.
We could implement registry like system in unix(/Linux) if we want using way described above - just save that registry under ~/.registry and it's done. Works over NFS and all. And you have all bells and whistles of unix fs also - including symbolic links. No more wondering about different config file syntax (xf86config anyone?). Everything is a file...
[File not found: ~/.registry/mail/signature] _____________________ ____
If you get into someone's house by picking their locks, it's breaking and entering. If you do it by using their key, it's not. That's the difference between circumvention and plain use.
More like "if you get into your house by picking your locks" (after for example losing your key). Or even "if you get into your house by using key you have made yourself". After all you have already bought the DVD disk you are trying to use. IMO because CSS algorithm isn't patented decrypting it with your own program should be legal.
_________________________
I actually like or liked IE as a browser. I'm getting a bit pissed off with its non complience to standards though. Somthing that Netscape does very well (& is the way forward. web needs standards)
Am I replying to flamebait? Netscape has better compliance to standards than IE? I don't think so! Think about CSS, HTML 4.x, XHTML 1.0 and XML - all of those are usable with IE. With Netscape you can try to use part of HTML 4.x (and don't tell me Netscape has CSS support - it's so buggy I would be better without).
If you mean Mozilla when you say Netscape things are different but IE is still better nowadays - future may change that also.
_________________________
Why to make it so hard? How about if you press power button system toggles power on and off and keeps it's state while off. If your system (from Redmont?) crashes and cannot continue simply press reset button (there is two buttons in ATX cases...) and system is restarted with memory cleared.
This could be implemented with current systems with only BIOS update if we had MRAM.
_________________________
If the console can do proper anti-aliasing and GUARANTEE a 60 fields per second rendering rate
No console (or computer) can guarantee any framerate above zero because it depends greatly how do you use processing power.
Take for example piece of hardware A that can render 25Mpps (polygons per second) and hardware B that can render 5Mpps. Now if it takes 0.1M polygons to fill entire screen you'll get 250fps with A and 50 fps with B. But you have to render overlapping polygons practically always - if you have to render each pixel, for example, 5 times (average 5 polygons overlap) you only get 50 fps with hardware A and 10 fps with hardware B. How much fps would you guarantee for hardware A and B? 250, 50, 10 or 5 fps? I would guarantee 0 fps for both because it only depends how good your rendering engine is (less overlapping=better).
Hardware can only guarantee minimum fill rate in pixels/texels but it's up to developer how to use it [for fps].
Only if you are talking about something like DVD playback where you have guaranteed maximum input (10Mbps) there can be guaranteed minimum output (60 fps). _________________________
Slashdot Mirror
I totally agree. However I really know I hate NS4.x even more and therefore I would want to see NS6.0 really soon now. Solution I would suggest is to make it nag after six months or so to force people to upgrade (it's annoying but it's effective). How do you think MS has gotten rid of IE3/4 so well?
_________________________
Sure, I'm able to setup font servers and build fontdirs but why must it be so difficult? It's only a few files or if you use TT fonts it's only one file for one font. Why it doesn't work like copy the file to /usr/X11/fonts/. and be done with it? I have yet to see good font handling in any system but IMO windows and mac os have usable ones.
_________________________
Yeah. Unfortunately my CRT only supports up to 1600x1200 resolutions in usable refresh rates. Therefore I would really want to use anti-aliasing.
Anyhow, I fail to see how X the display protocol is any worse than the win32 GDI for building browsers.
Two words: font rendering. Do you have any idea how much faster GDI is in font rendering? Compare Mozilla rendering speeds between windows and linux in similar hardware. Windows version of Mozilla is much faster than any version of IE. linux version on the other hand is almost as slow as Netscape 4.x under windows. The cause? Font rendering takes time. Network transparency is cool feature and I even use it every now and then. But I would sacrifice it any day for decent font rendering. Oh well, I think I need to wait berlin.
_________________________
I think it's because of the example - graphics with alpha channel - people get so excited about this. And this is 'cool' only if you are limited to X11! Even in that simple example the code looks pretty hard to understand when compared to equivalent C(++) code. In the end, if all you are interested in is speed you have to use native assembler. Otherwise simply use C(++), Perl, Python or Java - all of those have "unlimited registers" also.
_________________________
How about I change my MAC address? Get root and type in ifconfig eth0 hw addr 01:02:03:04:05:06. Just got yourself another MAC address. Do this like once a minute and it can be quite hard to track you down. Of course it breaks many other things but I'm just trying to tell that MAC address is not hardwired address and therefore shouldn't be used like one. [I found this information here.]
_________________________
Links is text WWW browser, similar to Lynx. Links displays tables, downloads on background and uses HTTP/1.1 keepalive connections. Get it from freshmeat. I can warmly recommend it over lynx any day [screenshot ]. w3m is worth a try also but I prefer Links to it.
_________________________
Therefore it has to be *really* fast internally with extra logic and cannot possibly be cheap. It sure is fast but far too expensive for system memory. Nintendo is using it for graphics adapter memory - I think this is the least valueable location you will [ever] see this type of memory.
If 1T-SRAM can really be done (without a tweak like this) it could really change situation. Then it would be like DRAM without refresh cycles and without capasitors. I doubt that's possible but even 2T-SRAM would be a dream come true. Remember that as long as we are talking about desktop computers it really doesn't mean how much power those are using but how fast memory you can get economically. It doesn't matter if it's 1T-SRAM (tweaked DRAM) or 6T-SRAM (real SRAM) if it costs the same [as long as speed is the same].
_________________________
If all services are provided by other servers how can I code anything? Try to read file? Sorry fs server is not running. Try to setup alarm? Sorry time server is not running. Of course one can require POSIX compatible system but what's the point running micro kernel if you always have basic set of servers (everything required to be POSIX compatible).
Idea of translators described in the article sounds nice, but if you think about it it's only mounting in user space. I don't know how mounting is done internally, but I have often wondered why I have to be root to be able to do it. In the end I cannot imagine any good reason why this couldn't be done with Linux. Monolithic kernel with device drivers as modules as the current trend with linux seems to be sounds like the best solution for me. Though I would break some quirks inherited from UNIX.
_________________________
I'm afraid this is not the real source. Clearly windows source wouldn't use code blocks but goto statements.
_________________________
They *could have*, but because all nerds are going to view this anyway why bother. After all everyone knows that while ILM does really good work their work is also really expensive.
On the other hand it could be that people in charge have no clue what CGI should look like nowadays and this was not a low cost solution only.
_________________________
Just in case you want to know what it is...
_________________________
Except that one-tool-does-all solution is seldom the best solution for a single problem. It's like using a pocket knife for felling instead of chainsaw - it works but it's not the best way. It's the same thing with emacs: you can play games, write mail and read slashdot with it. However, I find it more sensible to use special programs for all of those.
What I'm currently missing is a small cross-platform text editor with a good interface. Sorry, emacs is out - I want *small* editor with *good interface* you remember? The problem is that it's that hard to make such a thing that those who have tried have found it easier to accustom themselves to use emacs. Unfortunately.
_________________________
I'm surely happy I don't own "golden ears". I'm happy with 192kbps (or even 128kbps) MP3s. They are not as good as CDs but good enough. And to my ears those sound better than my Sony(R) Minidisc(tm). FM radio is actually pretty good also (I listen it through cable without antenna problems). IMHO any one of these will beat any analog cassette system.
_________________________
Yeah - but if I'm only interested in encrypted communication. Can I do this without a costly certification? And I wouldn't want to cause those scary warnings to user either. (I know that I lose some of the security this way because I cannot be sure that the other end is what I think it is.)
_________________________
Wow - you must have a *really* good ad! 20e6 ppl is far from 100k pages per day. Back to topic: say those 100k pages are served during one hour - it makes 100k/60/60 or 28 pages per second. You still can easily run quite heavy scripting.
_________________________
The point is one shouldn't need to generate (almost) any code. Never. If your API needs exact the same code each time you want to make and MDI app then one should not make source code generator. Instead API should be extended so that you could simply type new application.MDI() instead of lots of generated code. Why do you think anyone ever come with an idea of functions? Because their copy-paste keys wasn't working (compare: code generator)?
_________________________
OTOH if this was supposed to be MB test both boards should have used SDRAM. I'm afraid that current test has no real information whatsoever because pre-production MB is slower - but DDR memory is faster.
I would have wanted to see at least test where AMD760 was running SDRAM to have real results about DDR being faster.
_________________________
How about OpenAL. I think we should have also OpenIL (Input Library) and perhaps OpenFL (Feedback Library) for control devices - think about you could treat keyboards, joysticks and insert-your-favorites-here input devices as one from programming viewpoint.
_________________________
That's nice and all but it really doesn't matter how fast *single* transistor can work. The problem is that the whole circuit must work at given speed. And that speed is limited by the speed of light.
For example say our chip is 10mm x 10mm and we have to send signal from side to side during clock cycle: the time required for signal to get on the other side is distance/speed = (0.10m)/(299792458m/s) = 3.34e-10s. Now if we need to send signal like this every clock the maximum clock speed we can achieve is 1/(distance/speed) = 10*299792458 Hz which equals to less than 3GHz chip.
Of course chip designers are aware of this and design chips in a way no signal needs to be send across the whole chip, but even if greatest distance needed to send signal through - during one clock cycle - is 1mm (one 10th of the chip) we can only get 30GHz - and only if this is our only bottleneck. And also in this case average throughput time is at least 10 clock cycles (time for "operation" to go through the chip).
In the end one should notice that the problem of memory being too slow compared to processor isn't getting away in the future because we surely need those signals from our memory chips and yet again we are limited by the speed of light. Expect to see memory chips really near CPU in the future...
The only way to always double the computing power in the (not so distant) future is to invent a way to transfer *information* faster than light. If it's possible - I don't know.
_________________________
The difference is you have to watch TV from distance because it has such a lousy picture because of it's loysy CRT. I'm using my 19" monitor to watch TV from about a 3' away and the only problem is that resolution of PAL systems is low (still higher than NTSC I think) and interlaced. Think about the future and HDTV. How much will HDTV set cost because it should be able to show resolutions equal to monitors and be huge in the same time?
If I had a choice I would use a headset with resolution of 1024x768 per eye or more - with stereo vision. Then simply place those tiny screens near enough and you have that big TV you ever need.
_________________________
Exactly! But then why to do virtual file system on top of file system? One can control file access by file level too. Why not simply use each key as directory and each value as text file (key=filename, value=file contents)? Think about /proc. My guess is that MS tried this and noticed that their file system was that poor they had to try another (IMHO unsuccessful) way.
We could implement registry like system in unix(/Linux) if we want using way described above - just save that registry under ~/.registry and it's done. Works over NFS and all. And you have all bells and whistles of unix fs also - including symbolic links. No more wondering about different config file syntax (xf86config anyone?). Everything is a file...
[File not found: ~/.registry/mail/signature]_ ____
____________________
More like "if you get into your house by picking your locks" (after for example losing your key). Or even "if you get into your house by using key you have made yourself". After all you have already bought the DVD disk you are trying to use. IMO because CSS algorithm isn't patented decrypting it with your own program should be legal.
_________________________
Am I replying to flamebait? Netscape has better compliance to standards than IE? I don't think so! Think about CSS, HTML 4.x, XHTML 1.0 and XML - all of those are usable with IE. With Netscape you can try to use part of HTML 4.x (and don't tell me Netscape has CSS support - it's so buggy I would be better without).
If you mean Mozilla when you say Netscape things are different but IE is still better nowadays - future may change that also.
_________________________
This could be implemented with current systems with only BIOS update if we had MRAM.
_________________________
No console (or computer) can guarantee any framerate above zero because it depends greatly how do you use processing power.
Take for example piece of hardware A that can render 25Mpps (polygons per second) and hardware B that can render 5Mpps. Now if it takes 0.1M polygons to fill entire screen you'll get 250fps with A and 50 fps with B. But you have to render overlapping polygons practically always - if you have to render each pixel, for example, 5 times (average 5 polygons overlap) you only get 50 fps with hardware A and 10 fps with hardware B. How much fps would you guarantee for hardware A and B? 250, 50, 10 or 5 fps? I would guarantee 0 fps for both because it only depends how good your rendering engine is (less overlapping=better).
Hardware can only guarantee minimum fill rate in pixels/texels but it's up to developer how to use it [for fps].
Only if you are talking about something like DVD playback where you have guaranteed maximum input (10Mbps) there can be guaranteed minimum output (60 fps).
_________________________