Code that is drawing the screen is a lot easier to write if it knows the data it is drawing will not change under it. The only way to do this is locking. Fine-grain locking introduces overhead for each and every lock, and is also really difficult to get right. If you instead just use one big lock, you will just make your event and drawing threads take turns, which will just add overhead compared to the single-threaded version.
FAR bigger savings are done by not drawing until all pending events are handled. All modern toolkits work this way. This does mean you can't do incremental update (at least not easily) and for some reason novice programmers often think the way to make the program fast is to do an incremental update after every event, but they do learn.
I also have the old Natural Keyboard (the off-white bent one).
I blew $60 on one of their new ones (all black, with weird layout of the arrows, a center scroll lever that Linux does not see, and a whole bunch of "media" buttons) (I bought it because I figured the extra buttons would be useful, and they do work with Ubuntu to control media playback and volume, etc). This piece of junk is now in my closet because it was definitely harder to type on, I am back on the old one.
This is sad, MicroSoft used to mean good hardware...
This is because the installer programs are impossible (or at least very difficult) to configure to not put up an "accept this license" screen. So a lot of people building the installer stick the GPL text in there. It would be better if it said "this software is free to use, you don't have to accept anything, but this stupid program requires you to hit the button down below that says Accept".
Though I don't know Java, I would think case insensitivity in a static-typed language would be a really huge pain. It would prevent you from making an instance of class "Foo" called "foo", in particular.
Perhaps the Java syntax is such that whether a symbol is a variable or a type can be determined at parse time. But C has this and it is a real problem for C++, leading to a lot of the strange syntax, including the double colons, and a lot of extra modifier keywords needed by templates.
Another problem is that "case insensitive" is pretty nebulous and undefined in Unicode. Do they mean only the ASCII characters are "case insensitive"?
My understanding is that network transparency will continue to work in various ways. As accurately as I can see here is what will happen:
1. X11 applications will work just like now, they can directly talk to a remote X server, and a remote X11 application can talk to the local X server that is running atop Wayland.
2. Although it is likely that Qt and KDE, etc, will be rewritten to talk to Wayland directly, they will retain the ability to talk to X11 in their code. In the worst case this means you have to run a version of the program linked with this alternative version. Much more likely the toolkit will automatically use the X11 backend if it detects a remote X11 server is in $DISPLAY, similar to how OpenGL right now uses XGL only for remote displays and direct hardware manipulation for local. Though I would worry about bitrot making this ability fail after a while if nobody uses it. Another possibility is that changing LD_LIBRARY_PATH will cause an X11 version of the toolkits to be used instead of the Wayland version.
3. Wayland itself does not have to "composite" all the windows onto one local screen. I suspect there will be a protocol to send the window changes to a remote Wayland server and thus the window appears on a remote screen and input comes from the remote. This is pretty much what RDP does with official Microsoft clients.
4. Arbitrary programs can pretend to be a Wayland server and thus receive the above window changes. This is pretty much what all non-Windows RDP clients do, and it works well.
5. The composited result can be screen-scraped and sent over the network to a client that displays it.
It looks to me like the one thing Wayland replaces is the window manager and the compositor. Therefore it IS a "panacea for poorly-written window managers or compositors" (on the assumption that it itself is not poorly-written which I know may be a problem).
Like the X11 rootless server on OS/X, only client applications will work. X window managers and compositors are not supported, and they are replaced completely.
It looks to me like the "compositing" step does not necessarily mean "composite all the windows onto a certain screen". It could instead composite windows onto several screens, including putting the same window on several. And it could composite onto a remove wayland compositor by some VNC-like protocol, sending the changes to the buffer over the network so the remote one can recreate the window.
The main reason for huge bills is to pay for unfunded ER visits.
I agree with tort reform, but it is not insurance costs or payouts that add up. The problem is (as you mention at the bottom) that threats of lawsuits make doctors choose as many tests as possible and choose the most expensive and safest treatements. I have heard that this adds up to costs many times the money paid to lawyers and lawsuits. Tort reform is needed though the real reason is not the knee-jerk one you first list.
I think the Republicans will shit a brick if anybody suggests that ER visits be paid directly by the government so that idea of yours is going to go nowhere. Some of those ER visits are going to be Illegal Aliens!
In theory (and I certainly agree there are arguments against it) it is not the *total* cost that goes down, but the per-person cost. The reason is that everybody is forced to buy sufficient insurance, so the total (perhaps larger) cost is divided amoung more people.
There would still be a huge bureaucracy because that is an enormous amount of money to distribute and enormous incentives for cheating. And, just like with welfare, I'm sure people of all political persuasions will start to put rules on the money (ie you can't spend it on gambling or on psychic advisors or abortions or on junk food, etc) which will bloat the bureaucracy even more.
Still I have to admit the idea sounds intriguing.
I was confused for a bit because it sounded a bit like the refund portion of the "flat tax". That idea has very serious problems, as the refund is distributed to *everybody*, not just the poor. It is thus something like 10 times the size of the governments budget and 30 or more times the size of all welfare and other redistribution programs. Access to this astronomical amount of money by the government wont lead to problems far worse than the worst things you could imagine the current government (D or R, whichever you hate) could do.
Lets just ignore the obvious solution of making the fine larger than the premium for the entire uninsured period, ok (ie it goes up the longer you are uninsured). That way you won't have to have pesky logic ruining your arguments.
Another huge cost is treating patients without health insurance which the government forced on hospitals but never funded. Fix these two things and health care comes back down to reasonable levels. I'd much rather the government fund ER visits by those that don't have insurance
I'd really like to know just what the f**k you are talking about. Last time I checked, ERs *were* in hospitals, and people without insurance had only one way to not pay for healthcare: go to the ER. So it seems it is already set up just the way you request.
I suppose in theory that currently the ER visits are paid for by overcharging others with insurance and thus get eventually paid for by the premiums for those with insurance, but I'm sure there is a lot of government direct subsidy in there already as well.
Could you explain *exactly* what you are requesting? Also keep in mind that ER is about 10x more expensive than normal doctor visits. If you don't want poor people to get health care, just say so. Otherwise you are going to have to come up with a scheme that is different than what is done now.
I think a conservative-moderate and/or a democrat-moderate would easily win.
Unfortunately both parties see the other one going extreme, and instead of saying "hey we can win easy by grabbing the middle" they say "that means we can go more extreme too". So there are no moderates any more.
the consistent way to do it since win2k days is: start, run, cmd, ipconfig/all.
So for Linux, you say press Alt+F2.
Then what next?
Well on SUSE you type "ifconfig -a" which I know is really hard compared to typing "ipconfig/all", since we all know the letter f is much harder and user-unfriendly to type than p. And don't get me started on how difficult and user-unfriendly it is to type a dash.
I think you have a point however. This works with KDE. Tried it on Gnome and discovered those *IDIOTS* broke it totally. It runs it (as typing a non-existent command produces an error) but I cannot get the output to appear in a terminal. There is a "run in terminal" checkmark (somehow the KDE guys figured out how to pop up a terminal only when there is text sent to stdout, but considering Microsoft has not figured that out yet (requiring you to decide at compile time if there is stdout!) I guess it was too much to hope for Gnome). But checking the mark does not make it work. After some testing I finally see a very fast flash which indicates they dismiss the terminal when the program exits.
So idiocy is what is preventing Linux from working, not "inconsistent".
No. You tell them to type ALT PLUS F2!!! How stupid are you? I might as well ask if the Windows user should be told to run command.com or cmd.exe, that question makes as much sense as your inanity.
Sorry I was pretty certain Alt+F2 was copied from the shortcut for the "run" command on Windows. This really does not work? I could have sworn it did.
Um, you can type Alt+F2 (the SAME SHORTCUT AS WINDOWS) on Ubuntu and get a command line, too. I also suspect that if customer support needs to get somebody to do this, they will use Alt+F2, which is a lot easier to tell over the phone than getting them to popup the start menu and type a shortcut letter that may have been redefined.
Another fun one is to try to add "not to be confused with ODF, the format produced by the software known as Open Office". This will get instantly reverted with the incredibly bogus excuse that the official name for Open Office is not "Open Office".
Of course I did not RTFA, but it does sound like it is misleading, perhaps yelling for open source when in fact what is needed is a paper trail.
A fully open-source machine that produces no physical result could still be reprogrammed at the last moment (well of course with closed source but that is not really the point) to throw the election.
A machine producing a voter-verified paper trail is what is needed. If the machine designers can't figure out how to draw text on the screen except by using Windows, well that is too bad and a waste of money, but it probably has nothing to do with the trustworthiness of the system. As long as there are a significant number of random sample recounts, and the machines are not producing obvious errors, then it is as safe as anything written open source that also uses a paper trail.
Some of the more complex crypto systems could also be closed or open source. The resulting crypto can be checked (though it is necessary that the *checking* software be open source so anybody can run it). The problems with the crypto is that there is no way for the general public to understand it.
The actual location for the shade would be at the L1 point, which is closer to Venus by just about the same ratio the Sun is closer to it. So the shade would be about the same size. The main reason it would have to be bigger (or really denser) is that you would have to do significant more work to fix Venus, as you state.
What I meant is that normal video compression relies on encoding differences in frames and throwing away differences that are considered invisible to the end user. They are not designed to correctly encode an image that is really two alternating images, where the objects are in slightly different positions in the alternating images.
So it would seem the actual video data stream is two interlaced video compressions, or a new type of compression with the ability to efficiently encode the differences between left and right images. There is certainly nothing wrong with this but I would think it would require modifications to hardware decoders, and therefore another reason for the side-by-side encoding is that existing hardware can be reused.
I would think another advantage of side-by-side is that existing video compression would work. Page flipped means that parts of the image are, in effect, vibrating at the frame rate, with the annoyance that any compression must not remove or distort this vibration. Surely the page-flipped versions must also have some alteration of how the video is compressed as well?
What they do is add a preservative that has a word similar to "ammonia" in it's chemical name and this got mangled (either on purpose or accident) into the post.
I agree with the grandparent that Huff Post has so far failed to correct the "ammonia" comment. However the grandparent has also contributed his own spin: read his article and see if he is not implying that the *picture* is wrong.
I understand the idea, and it is similar to Multics.
It would make sense to serialize tokens that had a little more meaning than just characters. For instance a number could remain a "number token" in the stream. I think your examples, and also your request for s-exprs, shows that you think serialization into a larger set of basic types, such as maps and numbers, would be better. I can certainly buy this, google serialization tries to do this also. However in all the work I have done, all attempts at this quickly devolve into a more JSON-like syntax with only text. Some of this may be due to the need to efficiently convert to/from text, however.
Believe me I would have thought that passing structured data would be better as well and if I had been working on Multics I probably would have also thought it was a brilliant and powerful idea. I think some of the brilliance of Unix was in their realization of what was not necessary. It is thus easy to claim that something is somehow better with a complete logical argument, while in reality it is an unnecessary complication.
Slashdot Mirror
Code that is drawing the screen is a lot easier to write if it knows the data it is drawing will not change under it. The only way to do this is locking. Fine-grain locking introduces overhead for each and every lock, and is also really difficult to get right. If you instead just use one big lock, you will just make your event and drawing threads take turns, which will just add overhead compared to the single-threaded version.
FAR bigger savings are done by not drawing until all pending events are handled. All modern toolkits work this way. This does mean you can't do incremental update (at least not easily) and for some reason novice programmers often think the way to make the program fast is to do an incremental update after every event, but they do learn.
Why does it have an accelerometer if the intended use is to place it on the table under the TV?
I also have the old Natural Keyboard (the off-white bent one).
I blew $60 on one of their new ones (all black, with weird layout of the arrows, a center scroll lever that Linux does not see, and a whole bunch of "media" buttons) (I bought it because I figured the extra buttons would be useful, and they do work with Ubuntu to control media playback and volume, etc). This piece of junk is now in my closet because it was definitely harder to type on, I am back on the old one.
This is sad, MicroSoft used to mean good hardware...
This is because the installer programs are impossible (or at least very difficult) to configure to not put up an "accept this license" screen. So a lot of people building the installer stick the GPL text in there. It would be better if it said "this software is free to use, you don't have to accept anything, but this stupid program requires you to hit the button down below that says Accept".
Though I don't know Java, I would think case insensitivity in a static-typed language would be a really huge pain. It would prevent you from making an instance of class "Foo" called "foo", in particular.
Perhaps the Java syntax is such that whether a symbol is a variable or a type can be determined at parse time. But C has this and it is a real problem for C++, leading to a lot of the strange syntax, including the double colons, and a lot of extra modifier keywords needed by templates.
Another problem is that "case insensitive" is pretty nebulous and undefined in Unicode. Do they mean only the ASCII characters are "case insensitive"?
My understanding is that network transparency will continue to work in various ways. As accurately as I can see here is what will happen:
1. X11 applications will work just like now, they can directly talk to a remote X server, and a remote X11 application can talk to the local X server that is running atop Wayland.
2. Although it is likely that Qt and KDE, etc, will be rewritten to talk to Wayland directly, they will retain the ability to talk to X11 in their code. In the worst case this means you have to run a version of the program linked with this alternative version. Much more likely the toolkit will automatically use the X11 backend if it detects a remote X11 server is in $DISPLAY, similar to how OpenGL right now uses XGL only for remote displays and direct hardware manipulation for local. Though I would worry about bitrot making this ability fail after a while if nobody uses it. Another possibility is that changing LD_LIBRARY_PATH will cause an X11 version of the toolkits to be used instead of the Wayland version.
3. Wayland itself does not have to "composite" all the windows onto one local screen. I suspect there will be a protocol to send the window changes to a remote Wayland server and thus the window appears on a remote screen and input comes from the remote. This is pretty much what RDP does with official Microsoft clients.
4. Arbitrary programs can pretend to be a Wayland server and thus receive the above window changes. This is pretty much what all non-Windows RDP clients do, and it works well.
5. The composited result can be screen-scraped and sent over the network to a client that displays it.
It looks to me like the one thing Wayland replaces is the window manager and the compositor. Therefore it IS a "panacea for poorly-written window managers or compositors" (on the assumption that it itself is not poorly-written which I know may be a problem).
Like the X11 rootless server on OS/X, only client applications will work. X window managers and compositors are not supported, and they are replaced completely.
It looks to me like the "compositing" step does not necessarily mean "composite all the windows onto a certain screen". It could instead composite windows onto several screens, including putting the same window on several. And it could composite onto a remove wayland compositor by some VNC-like protocol, sending the changes to the buffer over the network so the remote one can recreate the window.
GTK is using Cairo.
The main reason for huge bills is to pay for unfunded ER visits.
I agree with tort reform, but it is not insurance costs or payouts that add up. The problem is (as you mention at the bottom) that threats of lawsuits make doctors choose as many tests as possible and choose the most expensive and safest treatements. I have heard that this adds up to costs many times the money paid to lawyers and lawsuits. Tort reform is needed though the real reason is not the knee-jerk one you first list.
I think the Republicans will shit a brick if anybody suggests that ER visits be paid directly by the government so that idea of yours is going to go nowhere. Some of those ER visits are going to be Illegal Aliens!
In theory (and I certainly agree there are arguments against it) it is not the *total* cost that goes down, but the per-person cost. The reason is that everybody is forced to buy sufficient insurance, so the total (perhaps larger) cost is divided amoung more people.
There would still be a huge bureaucracy because that is an enormous amount of money to distribute and enormous incentives for cheating. And, just like with welfare, I'm sure people of all political persuasions will start to put rules on the money (ie you can't spend it on gambling or on psychic advisors or abortions or on junk food, etc) which will bloat the bureaucracy even more.
Still I have to admit the idea sounds intriguing.
I was confused for a bit because it sounded a bit like the refund portion of the "flat tax". That idea has very serious problems, as the refund is distributed to *everybody*, not just the poor. It is thus something like 10 times the size of the governments budget and 30 or more times the size of all welfare and other redistribution programs. Access to this astronomical amount of money by the government wont lead to problems far worse than the worst things you could imagine the current government (D or R, whichever you hate) could do.
Lets just ignore the obvious solution of making the fine larger than the premium for the entire uninsured period, ok (ie it goes up the longer you are uninsured). That way you won't have to have pesky logic ruining your arguments.
Another huge cost is treating patients without health insurance which the government forced on hospitals but never funded. Fix these two things and health care comes back down to reasonable levels. I'd much rather the government fund ER visits by those that don't have insurance
I'd really like to know just what the f**k you are talking about. Last time I checked, ERs *were* in hospitals, and people without insurance had only one way to not pay for healthcare: go to the ER. So it seems it is already set up just the way you request.
I suppose in theory that currently the ER visits are paid for by overcharging others with insurance and thus get eventually paid for by the premiums for those with insurance, but I'm sure there is a lot of government direct subsidy in there already as well.
Could you explain *exactly* what you are requesting? Also keep in mind that ER is about 10x more expensive than normal doctor visits. If you don't want poor people to get health care, just say so. Otherwise you are going to have to come up with a scheme that is different than what is done now.
I think a conservative-moderate and/or a democrat-moderate would easily win.
Unfortunately both parties see the other one going extreme, and instead of saying "hey we can win easy by grabbing the middle" they say "that means we can go more extreme too". So there are no moderates any more.
the consistent way to do it since win2k days is: start, run, cmd, ipconfig /all.
So for Linux, you say press Alt+F2.
Then what next?
Well on SUSE you type "ifconfig -a" which I know is really hard compared to typing "ipconfig /all", since we all know the letter f is much harder and user-unfriendly to type than p. And don't get me started on how difficult and user-unfriendly it is to type a dash.
I think you have a point however. This works with KDE. Tried it on Gnome and discovered those *IDIOTS* broke it totally. It runs it (as typing a non-existent command produces an error) but I cannot get the output to appear in a terminal. There is a "run in terminal" checkmark (somehow the KDE guys figured out how to pop up a terminal only when there is text sent to stdout, but considering Microsoft has not figured that out yet (requiring you to decide at compile time if there is stdout!) I guess it was too much to hope for Gnome). But checking the mark does not make it work. After some testing I finally see a very fast flash which indicates they dismiss the terminal when the program exits.
So idiocy is what is preventing Linux from working, not "inconsistent".
what do you tell them to type?
gnome-terminal? terminal? konsole? rxvt?
No. You tell them to type ALT PLUS F2!!! How stupid are you? I might as well ask if the Windows user should be told to run command.com or cmd.exe, that question makes as much sense as your inanity.
Sorry I was pretty certain Alt+F2 was copied from the shortcut for the "run" command on Windows. This really does not work? I could have sworn it did.
Um, you can type Alt+F2 (the SAME SHORTCUT AS WINDOWS) on Ubuntu and get a command line, too. I also suspect that if customer support needs to get somebody to do this, they will use Alt+F2, which is a lot easier to tell over the phone than getting them to popup the start menu and type a shortcut letter that may have been redefined.
Another fun one is to try to add "not to be confused with ODF, the format produced by the software known as Open Office". This will get instantly reverted with the incredibly bogus excuse that the official name for Open Office is not "Open Office".
Of course I did not RTFA, but it does sound like it is misleading, perhaps yelling for open source when in fact what is needed is a paper trail.
A fully open-source machine that produces no physical result could still be reprogrammed at the last moment (well of course with closed source but that is not really the point) to throw the election.
A machine producing a voter-verified paper trail is what is needed. If the machine designers can't figure out how to draw text on the screen except by using Windows, well that is too bad and a waste of money, but it probably has nothing to do with the trustworthiness of the system. As long as there are a significant number of random sample recounts, and the machines are not producing obvious errors, then it is as safe as anything written open source that also uses a paper trail.
Some of the more complex crypto systems could also be closed or open source. The resulting crypto can be checked (though it is necessary that the *checking* software be open source so anybody can run it). The problems with the crypto is that there is no way for the general public to understand it.
The actual location for the shade would be at the L1 point, which is closer to Venus by just about the same ratio the Sun is closer to it. So the shade would be about the same size. The main reason it would have to be bigger (or really denser) is that you would have to do significant more work to fix Venus, as you state.
He asked for "real time".
I would suspect the switch from hardware doubles to software arbitrary-precision produces many orders of magnitude of slowdown so the answer is no.
What I meant is that normal video compression relies on encoding differences in frames and throwing away differences that are considered invisible to the end user. They are not designed to correctly encode an image that is really two alternating images, where the objects are in slightly different positions in the alternating images.
So it would seem the actual video data stream is two interlaced video compressions, or a new type of compression with the ability to efficiently encode the differences between left and right images. There is certainly nothing wrong with this but I would think it would require modifications to hardware decoders, and therefore another reason for the side-by-side encoding is that existing hardware can be reused.
I would think another advantage of side-by-side is that existing video compression would work. Page flipped means that parts of the image are, in effect, vibrating at the frame rate, with the annoyance that any compression must not remove or distort this vibration. Surely the page-flipped versions must also have some alteration of how the video is compressed as well?
Soaking food in ammonia is illegal.
What they do is add a preservative that has a word similar to "ammonia" in it's chemical name and this got mangled (either on purpose or accident) into the post.
I agree with the grandparent that Huff Post has so far failed to correct the "ammonia" comment. However the grandparent has also contributed his own spin: read his article and see if he is not implying that the *picture* is wrong.
I understand the idea, and it is similar to Multics.
It would make sense to serialize tokens that had a little more meaning than just characters. For instance a number could remain a "number token" in the stream. I think your examples, and also your request for s-exprs, shows that you think serialization into a larger set of basic types, such as maps and numbers, would be better. I can certainly buy this, google serialization tries to do this also. However in all the work I have done, all attempts at this quickly devolve into a more JSON-like syntax with only text. Some of this may be due to the need to efficiently convert to/from text, however.
Believe me I would have thought that passing structured data would be better as well and if I had been working on Multics I probably would have also thought it was a brilliant and powerful idea. I think some of the brilliance of Unix was in their realization of what was not necessary. It is thus easy to claim that something is somehow better with a complete logical argument, while in reality it is an unnecessary complication.