once you get rid of cross-thread marshalling, apartments, proxies (which are the same as marshalling), IDispatch (which is an interface, not a part of COM), and support for OOP servers (which require marshalling), COM is actually livable.
Sure COM does not support concepts like weak references and expirable objects (client assumes object is valid while it holds a reference to it, a proxy is required to get around this), but it is rather nice if you ditch all the complicated fluff they added to try to make IPC 'transparent'.
Yes, there are some people who have no real humanity. IT is best to learn to determine the 'worthwhile' people and hang around them. Luckily (at least in this part of the country) there are a lot more worthwhile people than heartless bozos.
yah, they've met, I don't know any details other than Carmack visited Transmeta and they talked about things like 3D on Linux (and things needed for the kernel)
1.0 was the release of the API, not of 'GNOME' as a package. Unfortunately it was taken this way, and many of the components that are considered a part of gnome were not at 1.0 then (and still aren't last time I checked!)
thats what prefetch instructions are for. The idea is that the programmer (or the compiler, hopefully soon) will know better what memory might be needed. So by increasing your memory bandwidth you get the ability to preload into cache the data needed by the branches coming up. Hence for an optimized program (and a large L2 cache), the 'observed' RDRAM latency drops to zero. This is a big feature of Merced (and a feature of SIMD, although that is both optional and P3-only)
Matrox has released all the specs they do not consider to be IP to the linux community. Meaning that we do not have the specs for the 'WARP engine', the triangle setup engine(s) on the G200/G400. Luckily there are some rather good-hearted and persistant engineers inside of Matrox, they got a binary-code version of the setup code out, and specs for loading it in.
3Dfx has released full 2D specs. It is a shame they do not release 3D specs, but it is definately not for lack of people asking:) 3Dfx was also the first to release, and the only people to actually keep their product up-to-date (glide on Linux is just as fast as glide on windows, I believe the entire thing is in assembly with some MMX and 3DNow optimizations for bandwidth reasons, since Glide does not do transformation).
Compare to Matrox's zero released drivers, or NVidia's driver release. Will they release another one? Who knows, hopefully. Maybe when XFree 4.0 comes out.
The only thing you could do with the glide source code is port it to another architecture. Really. Or inline with Mesa. There is absolutely no optimization that can be done to the glide API implementation left (that human beings can comprehend).
NVidia has released no register specs whatsoever. They did release a special version of the internal toolkit that they use to make the windows drivers (i.e. their version of glide, 'cept more object-oriented and all-encompassing, not just 3d but 2d and video). But they have been rather bad about promoting this, the binary drivers on their page for the API are for an old kernel, and the toolkit is not up-to-date with their own internal version. Because of this (And because the Riva XServer/GLX don't even use it) there haven't been that many people messing around with it.
But some people on the GLX mailing list (including me) have been trying to lobby for the release of the 'real' nVidia toolkit. If this gets released, things like the GeForce would be supported day of RTM. So keep your fingers crossed.:)
Detonator is Win9x and WinNT 4.0 only. All other drivers are written differently. The XFree support was not done by NVidia AFAIK, it was done by one of the people working for ST (Their chip producer, at least at the time). The GLX driver that is out was written by a third party I believe, and does not use any of the NVidia code at all, just #def'd register location/vars.
It would be nice to put one your ceiling and be able to surf or watch TV while laying in bed. But for that kind of money I think a projector would work just as well.
the problem shows up in some machines shipping now, but supposedly in great numbers of the 500 MHz chips. I imagine people were already complaining about reboots on the 1st wave of G4s
My former employer (worked for a small software engineering firm, which since I left got larger;-)) has said many times to me that he has never, ever had to show his degree to anyone. Nobody has ever asked him 'wait, you do have a college education don't you?' before hiring him. But he is self driven, willing to start his own company and put in the hours (80 wk? more?) to make it pull off. And it is hard work, I have seen him mess up some times.
So no, you don't need a degree if you are willing to get yourself on equal footing as those with the degree. Honestly (and my mum would kill me if she heard me say this), the degree isn't important as far as the piece of paper is concerned. What is more important is your willingness to learn and your knowledge of the fundamentals. If you don't have a degree, you better have real-world experience going into a job interview. And you will still run into problems where you need to know things like how serial ports work, how network protocols work and the fundamentals of servers, operating system theory and threads/concurrancy/deadlock , etc. If you do not know java, you can learn in two days, assuming you know the fundamentals and are sufficiently motivated.
So, like many people here have said, don't think computer science is knowing to program C. Or any of the other languages of the week. You must know the fundamentals, and that is what a computer science program will teach you. If you are willing to learn the fundamentals on your own time (if you are sufficiently motivated), then the only thing a degree buys you is a college saying 'yes, we think he really does know this stuff and is qualified to work'
Mine spends about 20%, and I don't run distributed.net:) (and I would love for someone to invent a technology so that I could save up that 20% for during compiles, etc)
if tomorrow they fucked with the code and made advertisements in, and somehow did it in such a way that people couldn't figure out how to remove them..
..then the source code from tonite would be the basis of a new independant browser project.
it wouldn't work due to perspective. You would need something similar to a hologram, where different angles give different images. And then comes the pain in getting the image data from the other side of you...
Am I the only one who thinks that they could meet this guarantee by saying 'well, it was up 999 milliseconds before crashing that one millisecond, right?'
I think having a server down for a day a year (8 hours in a working day, 365 days a year) is unacceptable. Lets work on 99.99% uptime;-)
I wonder how many lines of say, the first ten versions of Linux are still in 2.3
none? probably.
Don't knock design until you look at linux-kernel. Linus doesn't look at design (although you can see some rather heated arguments about it sometime).
He gets to look at things after implementation. You can easily tell if someone thought things through before starting to implement. If it is ugly or a hack, he won't take it 99% of the time (the other 1% I think happens when he knows other people will quickly make it right;-))
They license out their Postscript name and technologies. The reason Mac OS x is supposedly still delayed is that the drawing API was based on Display Postscript. Besides being slow (from what I heard), Adobe wanted a lot of money for them to use Display Postscript. So they went and started work on a new graphics API (which, funnily enough, sounds like Display PDF)
Slashdot Mirror
and the winner gets paid a new computer :)
once you get rid of cross-thread marshalling, apartments, proxies (which are the same as marshalling), IDispatch (which is an interface, not a part of COM), and support for OOP servers (which require marshalling), COM is actually livable.
Sure COM does not support concepts like weak references and expirable objects (client assumes object is valid while it holds a reference to it, a proxy is required to get around this), but it is rather nice if you ditch all the complicated fluff they added to try to make IPC 'transparent'.
Taco bell is the only one left (and they will probably sell out to reach their long time goal:
"In the future all restaurants are taco bell")
Yes, there are some people who have no real humanity. IT is best to learn to determine the 'worthwhile' people and hang around them. Luckily (at least in this part of the country) there are a lot more worthwhile people than heartless bozos.
yah, they've met, I don't know any details other than Carmack visited Transmeta and they talked about things like 3D on Linux (and things needed for the kernel)
a laptop running windows 98?
1.0 was the release of the API, not of 'GNOME' as a package. Unfortunately it was taken this way, and many of the components that are considered a part of gnome were not at 1.0 then (and still aren't last time I checked!)
They have said there are no plans to use Rambus in Merced? when/where they say this?
thats what prefetch instructions are for. The idea is that the programmer (or the compiler, hopefully soon) will know better what memory might be needed. So by increasing your memory bandwidth you get the ability to preload into cache the data needed by the branches coming up. Hence for an optimized program (and a large L2 cache), the 'observed' RDRAM latency drops to zero. This is a big feature of Merced (and a feature of SIMD, although that is both optional and P3-only)
That just goes to show you how bad the crack is..
Matrox has released all the specs they do not consider to be IP to the linux community. Meaning that we do not have the specs for the 'WARP engine', the triangle setup engine(s) on the G200/G400. Luckily there are some rather good-hearted and persistant engineers inside of Matrox, they got a binary-code version of the setup code out, and specs for loading it in.
:) 3Dfx was also the first to release, and the only people to actually keep their product up-to-date (glide on Linux is just as fast as glide on windows, I believe the entire thing is in assembly with some MMX and 3DNow optimizations for bandwidth reasons, since Glide does not do transformation).
:)
3Dfx has released full 2D specs. It is a shame they do not release 3D specs, but it is definately not for lack of people asking
Compare to Matrox's zero released drivers, or NVidia's driver release. Will they release another one? Who knows, hopefully. Maybe when XFree 4.0 comes out.
The only thing you could do with the glide source code is port it to another architecture. Really. Or inline with Mesa. There is absolutely no optimization that can be done to the glide API implementation left (that human beings can comprehend).
NVidia has released no register specs whatsoever. They did release a special version of the internal toolkit that they use to make the windows drivers (i.e. their version of glide, 'cept more object-oriented and all-encompassing, not just 3d but 2d and video). But they have been rather bad about promoting this, the binary drivers on their page for the API are for an old kernel, and the toolkit is not up-to-date with their own internal version. Because of this (And because the Riva XServer/GLX don't even use it) there haven't been that many people messing around with it.
But some people on the GLX mailing list (including me) have been trying to lobby for the release of the 'real' nVidia toolkit. If this gets released, things like the GeForce would be supported day of RTM. So keep your fingers crossed.
Detonator is Win9x and WinNT 4.0 only. All other drivers are written differently. The XFree support was not done by NVidia AFAIK, it was done by one of the people working for ST (Their chip producer, at least at the time). The GLX driver that is out was written by a third party I believe, and does not use any of the NVidia code at all, just #def'd register location/vars.
umm, *Fall*, **CRUSH
=)
the problem shows up in some machines shipping now, but supposedly in great
numbers of the 500 MHz chips. I imagine people were already complaining about reboots on the 1st wave of G4s
I've been there, I know people that go to school there. The whole place makes you feel like you are on drugs as you walk around.. very nice place :)
My former employer (worked for a small software engineering firm, which since I left got larger ;-))
has said many times to me that he has never, ever had to show his degree to anyone. Nobody has ever asked him 'wait, you do have a college education don't you?' before hiring him.
But he is self driven, willing to start his own company and put in the hours (80 wk? more?) to make it pull off. And it is hard work,
I have seen him mess up some times.
So no, you don't need a degree if you are willing to get yourself on equal footing as those with the degree. Honestly (and my mum would kill me if she heard me say this), the degree isn't important as far as the piece of paper is concerned. What is
more important is your willingness to learn and your knowledge of the fundamentals. If you don't have a degree, you better have real-world experience going into a job interview. And you will still run into problems where you need to know things like how serial ports work, how network protocols work and the fundamentals of servers, operating system theory and threads/concurrancy/deadlock , etc. If you do not know java, you can learn in two days, assuming you know the fundamentals and are sufficiently motivated.
So, like many people here have said, don't think computer science is knowing to program C. Or any of the other languages of the week. You must know the fundamentals, and that is what a computer science program will teach you. If you are willing to learn the fundamentals on your own time (if you are sufficiently motivated), then the only thing a degree buys you is a college saying 'yes, we think he really does know this stuff and is qualified to work'
Mine spends about 20%, and I don't run distributed.net :)
(and I would love for someone to invent a technology so that I could save up that 20% for during compiles, etc)
Actually that is Bill's law, and it is double every 12 months.
if tomorrow they fucked with the code and made advertisements in, and somehow did it in such a way that people couldn't figure out how to remove them..
..then the source code from tonite would be the basis of a new independant browser project.
it wouldn't work due to perspective. You would need something similar to a hologram, where different angles give different images. And then comes the pain in getting the image data from the other side of you...
;-)
Not saying it wouldn't kick butt though
"...damn NT! All my walls are BLUE again. I HATE blue! Ralph, go to the server room in the cellar, and reset the things AGAIN"
Am I the only one who thinks that they could meet this guarantee by saying 'well, it was up 999 milliseconds before crashing that one millisecond, right?'
;-)
I think having a server down for a day a year (8 hours in a working day, 365 days a year) is unacceptable. Lets work on 99.99% uptime
I wonder how many lines of say, the first ten versions of Linux are still in 2.3
;-))
none? probably.
Don't knock design until you look at linux-kernel. Linus doesn't look at design (although you can see some rather heated arguments about it sometime).
He gets to look at things after implementation. You can easily tell if someone thought things through before starting to implement. If it is ugly or a hack, he won't take it 99% of the time (the other 1% I think happens when he knows other people will quickly make it right
Go look at any of the FSF's projects. You can't get code in GCC without turning the rights over to the FSF
They license out their Postscript name and technologies. The reason Mac OS x is supposedly still delayed is that the drawing API was based on Display Postscript. Besides being slow (from what I heard), Adobe wanted a lot of money for them to use Display Postscript. So they went and started work on a new graphics API (which, funnily enough, sounds like Display PDF)