I, for one, welcome our new 3d-printed overlords. I'd like to remind them that as an engineer I can be helpful in improving all aspects of their design from mechanical and electrical engineering to ASIC, FPGA and software development.
This is blatantly false. If server and client are on the same machine, rendering happens only once. Even if they are on different machines, rendering will still happen only once on the server if it is implemented through primitives supported by X and extensions (such as 3D in OpenGL). Popular GUI toolkits such as Qt and GTK+, render on the client and produce bitmaps, however those are details of their current implementation -- if their performance will become a problem, they can be easily moved into X extensions, and render on either client or server depending on available features and convenience.
That's irrelevant because in your example it's the minimum lag achievable with any protocol. Obviously, in reality network lag varies. Urelated to this, 83ms lag is acceptable for all purposes other than games and graphics/CAD editors -- web browsers, for example, have higher latency, Google Maps takes seconds to update, and they are perfectly usable applications.
I think we talked about this before and Microsoft in their UI training is finding that when it comes to touch more than 1ms is detectable by humans.
Desktops don't have touch interfaces.
Having worked on voice systems 83 ms is very confusing for humans,
Desktops don't have voice interfaces.
In terms of response to input the this should be about 10 ms on a mouse
If that was true, no remote access systems would ever be usable, as 10ms latency is almost never available over the Internet. In reality, I have perfectly usable X11 without wrappers over 25-30ms connection between my home and work.
and 1 ms on a touchscreen.
Absolutely irrelevant because desktops don't have touchscreens, and devices that do, revert to remote-cursor interface for remote access.
In any case, everything you said is irrelevant because no network connection outside a local network, and no protocol, existing or possible, would satisfy your requirements. If your argument is that remote access is useless, it contradicts the fact that huge numbers of people use various remote access protocols constantly. If your argument is, only access over local network is important, then X11 is perfect already because it already works better than everything else when latency is low.
Err.. The compositor already has a copy of the buffer that is sent to the GPU. Why would anyone be stupid enough to read the texture from VRAM.
Because some genius implemented 3D rendering on one host (possibly even in GPU with no display connected), and has to display it on another host. X11 model avoids that by moving rendering as close to display as possible.
The funny thing is after all this network transparency X is so shit that even VNC has it beat for performance (you don't even have to try rdesktop or spice).
No. X11 is faster than VNC on local network with low latency, and still faster over a high-latency network when wrapped in NX (NX does not change the nature of the protocol). The only situation where VNC is faster, is very high latency with no mitigation.
This is a classical viewpoint, generated before the end of the Cold War, in an attempt to explain some very puzzling aspects of the history of WW2.
Not only this predates Cold War, that was the primary disagreement in doctrine between Stalin and Trotsky (who was, naturally a hawk in that regard).
See, for example, Viktor Suvorov's book "Chief Culprit: Stalin's Grand Design to Start World War II".M
There are plentu of books about tin hat conspiracy theories, and more dishonest "interpreters" of well-known events.
Suvorov discusses a whole bunch of evidence that suggests Stalin was actually setting up an invasion of Germany. The movement of troops and equipment associated with setting up this attack made the Soviets extremely vulnerable to the German attack. For example, large numbers Soviet tanks were on trains being moved to the border when the attack occurred, making them very vulnerable to the Luftwaffe.
That's the evidence of the opposite -- that they were not expecting the war until it happened. Only a conspiracy theorist completely detached from reality would turn it into confirmation of something opposite.
As far as the "warm bodies" issue goes, some Soviet veterans have given interviews in which they described being sent into battle with one rifle between 30 untrained men,
I have already explained and refuted this as supposed evidence.
so this misconception is understandable if not particularly accurate.
Misconception can not possibly be accurate.
According to Suvorov, in some respects the Soviets were actually far ahead of the West in military equipment and tactics (as they demonstrated against the Japanese shortly before WW2), but much of their equipment was lost in the initial weeks of the invasion, and many of their best units were disorganized or destroyed, leading to a very difficult supply situation and a shortage of trained units.
This is entirely in line with what I have already explained, USSR was not prepared for war, and had minimal resources available when it started, there was no grand strategy to "throw warm bodies" anywhere, the enemy was simply stronger and better prepared, so there had to be significant losses and poorly trained militatry involved, no matter what military and civilian leaders did at that point. Once weapons and equipment production went into the "war mode", and military was recruited and trained, the real strategy shown up, and it was not "throw warm bodies".
As for the other point: If networks can shift insane amounts of data... THEN WHY ARE YOU USING THE X MODEL AND NOT SHIFTING FUCKING BITMAP IMAGES THAT HAVE BEEN ASSEMBLED REMOTELY. Congratulations you just torpedoed your own case. Genius.
Copying data that is already generated, creates additional access to memory that otherwise would not be accessed. More so if copying implementation has to guess which part of the screen actually updated. There is a reason, in 3D graphics most of the graphics buffer is never accessed by CPU. Now slow memory and local bus (synchronous mechanism) can have greater effect on performance than networking (asynchronous mechanism).
Display subsystem does not need short round trip times most of the time (when it doesn't react to input, moving windows, etc.), so all delays are lags and protocol works just fine. So fundamentally X design is right, and you are complaining about nothing.
Modern GPUs require huge amounts of data to move from the app to the card...
No. 1. Modern GPUs require huge amounts of data to move from rendering library to the card. This is not the same as application. 2. Networks are INSANELY GREAT at moving huge amounts of data now. What usually kills the performance is the latency. However latency is not a problem when application is constantly uploading things to draw, user won't see them until the frame redraw (that is still every 8-16 milliseconds) anyway.
All this means that X "chatty" protocol may be suboptimal but X extensions model, moving rendering to the server, is the right design decision. It would be eve better if not just OpenGL but UI toolkits and rendering engines had an option to be a part of the server. But that requires some careful design thinking when designing extensions and their protocol, a kind of thinking that it completely foreign for Wayland developers.
So you are claiming that everyone on Social Security is starving, and the US is unable to distribute food within its borders? You are simply wrong. SS works pretty well at distributing funds.
No, I am claiming that Social Security requires a massive effort to be maintained, and it deals with things that can be done very easily in established society. Now, imagine comparable number of people who are refugees, in a completely different country, with no connection to the society there. There is nothing Americans can contribute there that Chinese wouldn't already do better, and it's extremely arrogant of Americans even to think that their presence would produce anything but irritation.
For the same reason why streets are not being rebuilt with sidewalks in the center and traffic lanes around them. Some design decisions are actually good.
The issue isn't that an orderly distribution of North Koreans across China would be an issue, but they would be very concentrated in a more remote area of China.
Again, we are talking about the number of people far beyond one that American government can feed in their own country in organized fashion using well-established infrastructure and bureaucracy. Now, imagine half of American elderly -- but moved to China, living in tents. Then think, how Americans would be of any help there.
Your argument is that a bullet in your brain isn't an issue if the total amount of lead isn't deadly from lead poisoning. If you want to talk about what a stupid idea is being tossed about, that would be it.
My argument is, American resources would not be only useless there, they would be completely unnoticeable due to the enormous scale of the problem.
That happened after the war, mostly involved forces other than the former military of Iraq, and Iraq resources were not exhausted because it was just as full of oil as before the war. If anything, former government kept locals from fighting each others, something that Americans only understood years later.
Unix has it's problems, too. Plan9 proved that there was a much better way to do things. Most new features on Linux are half-assed because of the need to maintain some semblance of backwards compatibility. For example, namespaces---Plan9 executed namespaces perfectly; Linux namespaces are an abomination. And yet... they're good enough. Because reinventing the wheel just isn't cost effective.
Anti-Linux trolls are getting more sophisticated again.
Once Wayland components developers started trying to implement something practical, they discover, one by one, that they need those "unnecessary" X features after all, however there is no way to explain it to the rest of developers, who still believe that removing everything they don't immediately use in their narrow area is a great design practice.
so long as we promise to take care of the refugees.
China has FIVE TIMES MORE PEOPLE than you do, and they are ACROSS THE BORDER from Northern Korea. North Korea has 1/12 of US population, and that's half the number of Americans that now are receiving Social Security benefits.
If Chinese see refugees as a problem on their territory, Americans would make absolutely no difference. Words can't describe what a stupid idea that is.
Stalin's entire strategy was to throw warm bodies at the German guns
That's a talking point of American propaganda, created at the time of Cold War to denigrate Russians. Stop repeating it.
At the start of the war USSR was poorly prepared for it -- Stalin was convinced that Germany is not interested in attacking USSR because he seen it as a part of Western politics, and Stalin's own doctrine was that USSR must develop Socialism and Communism while keeping itself in isolation from Capitalist countries around it. All they had was a lot of people and some tanks, so with sufficiently dishonest commentary it's possible to present it as "throwing warm bodies" at Germans, however it was not a strategy and certainly not a winning one, as most of the Western Russia, Ukraine and everything to the West was occupied by Germans by the end of 1941, half a year after the initial attack. By then, however, USSR rebuilt large chunks of its industry in the Eastern part of the country beyond the Ural mountains, and organized the production of military equipment that was designed either before or during the war.
Once supported with steady production of weapons, USSR military became stronger while Germans became hobbled by logistics over a large still-hostile territory, and cold winter did not help them, either. The events after that include huge battle of Stalingrad and truly massive battle of Kursk (yes, that's thousands of tanks on both sides). While losses were still high, that's hardly "warm bodies" or "peasants" thrown at "sophisticated Germans". At that point Germans were pretty much doomed as USSR had rebuilt industry backed by natural resources of a huge territory on the East, and modern military while Germany was scrambling for oil, metals and other resources with dwindling size of their military.
Our big fear was that Japan would fight to the last man and that we'd have to invade.
And it was a completely baseless fear, created by American propaganda for the purpose of justifying the nuclear bombing. At no point, before or after, such idiotic thing ever happened -- when military is nearly destroyed, and resources are exhausted, they don't magically start fighting better.
It also has nothing to do with any practical application.
Says the person who thinks unix is good design. I think I will use this as a joke now. Thanks !
Your arrogance is misplaced, as you defend Microsoft design, the result of decades of ad-hock writing, managed by people who know nothing about software development.
Haha.. anger management problems as well. Man, you have a long list of mental defects. Sorry about that.:( Goodbye.. I mean.. fuck you.
I don't have anger management problems, I just hate you. People who do not hate scum like you and your Microsoft overlords, are themselves morally bankrupt.
Slashdot Mirror
My grandfather has spent years in prison during Stalin times without any reason
No, he did not.
Also fuck you.
I, for one, welcome our new 3d-printed overlords. I'd like to remind them that as an engineer I can be helpful in improving all aspects of their design from mechanical and electrical engineering to ASIC, FPGA and software development.
This is blatantly false. If server and client are on the same machine, rendering happens only once.
Even if they are on different machines, rendering will still happen only once on the server if it is implemented through primitives supported by X and extensions (such as 3D in OpenGL). Popular GUI toolkits such as Qt and GTK+, render on the client and produce bitmaps, however those are details of their current implementation -- if their performance will become a problem, they can be easily moved into X extensions, and render on either client or server depending on available features and convenience.
83 ms is totally unacceptable.
That's irrelevant because in your example it's the minimum lag achievable with any protocol. Obviously, in reality network lag varies.
Urelated to this, 83ms lag is acceptable for all purposes other than games and graphics/CAD editors -- web browsers, for example, have higher latency, Google Maps takes seconds to update, and they are perfectly usable applications.
I think we talked about this before and Microsoft in their UI training is finding that when it comes to touch more than 1ms is detectable by humans.
Desktops don't have touch interfaces.
Having worked on voice systems 83 ms is very confusing for humans,
Desktops don't have voice interfaces.
In terms of response to input the this should be about 10 ms on a mouse
If that was true, no remote access systems would ever be usable, as 10ms latency is almost never available over the Internet. In reality, I have perfectly usable X11 without wrappers over 25-30ms connection between my home and work.
and 1 ms on a touchscreen.
Absolutely irrelevant because desktops don't have touchscreens, and devices that do, revert to remote-cursor interface for remote access.
In any case, everything you said is irrelevant because no network connection outside a local network, and no protocol, existing or possible, would satisfy your requirements. If your argument is that remote access is useless, it contradicts the fact that huge numbers of people use various remote access protocols constantly. If your argument is, only access over local network is important, then X11 is perfect already because it already works better than everything else when latency is low.
Err.. The compositor already has a copy of the buffer that is sent to the GPU. Why would anyone be stupid enough to read the texture from VRAM.
Because some genius implemented 3D rendering on one host (possibly even in GPU with no display connected), and has to display it on another host. X11 model avoids that by moving rendering as close to display as possible.
The funny thing is after all this network transparency X is so shit that even VNC has it beat for performance (you don't even have to try rdesktop or spice).
No. X11 is faster than VNC on local network with low latency, and still faster over a high-latency network when wrapped in NX (NX does not change the nature of the protocol). The only situation where VNC is faster, is very high latency with no mitigation.
Does it matter?
This is a classical viewpoint, generated before the end of the Cold War, in an attempt to explain some very puzzling aspects of the history of WW2.
Not only this predates Cold War, that was the primary disagreement in doctrine between Stalin and Trotsky (who was, naturally a hawk in that regard).
See, for example, Viktor Suvorov's book "Chief Culprit: Stalin's Grand Design to Start World War II".M
There are plentu of books about tin hat conspiracy theories, and more dishonest "interpreters" of well-known events.
Suvorov discusses a whole bunch of evidence that suggests Stalin was actually setting up an invasion of Germany. The movement of troops and equipment associated with setting up this attack made the Soviets extremely vulnerable to the German attack. For example, large numbers Soviet tanks were on trains being moved to the border when the attack occurred, making them very vulnerable to the Luftwaffe.
That's the evidence of the opposite -- that they were not expecting the war until it happened. Only a conspiracy theorist completely detached from reality would turn it into confirmation of something opposite.
As far as the "warm bodies" issue goes, some Soviet veterans have given interviews in which they described being sent into battle with one rifle between 30 untrained men,
I have already explained and refuted this as supposed evidence.
so this misconception is understandable if not particularly accurate.
Misconception can not possibly be accurate.
According to Suvorov, in some respects the Soviets were actually far ahead of the West in military equipment and tactics (as they demonstrated against the Japanese shortly before WW2), but much of their equipment was lost in the initial weeks of the invasion, and many of their best units were disorganized or destroyed, leading to a very difficult supply situation and a shortage of trained units.
This is entirely in line with what I have already explained, USSR was not prepared for war, and had minimal resources available when it started, there was no grand strategy to "throw warm bodies" anywhere, the enemy was simply stronger and better prepared, so there had to be significant losses and poorly trained militatry involved, no matter what military and civilian leaders did at that point. Once weapons and equipment production went into the "war mode", and military was recruited and trained, the real strategy shown up, and it was not "throw warm bodies".
As for the other point: If networks can shift insane amounts of data... THEN WHY ARE YOU USING THE X MODEL AND NOT SHIFTING FUCKING BITMAP IMAGES THAT HAVE BEEN ASSEMBLED REMOTELY. Congratulations you just torpedoed your own case. Genius.
Copying data that is already generated, creates additional access to memory that otherwise would not be accessed. More so if copying implementation has to guess which part of the screen actually updated. There is a reason, in 3D graphics most of the graphics buffer is never accessed by CPU. Now slow memory and local bus (synchronous mechanism) can have greater effect on performance than networking (asynchronous mechanism).
Yes! You will have a lag of 83 ms! How horrible!
Display subsystem does not need short round trip times most of the time (when it doesn't react to input, moving windows, etc.), so all delays are lags and protocol works just fine. So fundamentally X design is right, and you are complaining about nothing.
Modern GPUs require huge amounts of data to move from the app to the card...
No.
1. Modern GPUs require huge amounts of data to move from rendering library to the card. This is not the same as application.
2. Networks are INSANELY GREAT at moving huge amounts of data now. What usually kills the performance is the latency. However latency is not a problem when application is constantly uploading things to draw, user won't see them until the frame redraw (that is still every 8-16 milliseconds) anyway.
All this means that X "chatty" protocol may be suboptimal but X extensions model, moving rendering to the server, is the right design decision. It would be eve better if not just OpenGL but UI toolkits and rendering engines had an option to be a part of the server. But that requires some careful design thinking when designing extensions and their protocol, a kind of thinking that it completely foreign for Wayland developers.
So you are claiming that everyone on Social Security is starving, and the US is unable to distribute food within its borders? You are simply wrong. SS works pretty well at distributing funds.
No, I am claiming that Social Security requires a massive effort to be maintained, and it deals with things that can be done very easily in established society. Now, imagine comparable number of people who are refugees, in a completely different country, with no connection to the society there. There is nothing Americans can contribute there that Chinese wouldn't already do better, and it's extremely arrogant of Americans even to think that their presence would produce anything but irritation.
For the same reason why streets are not being rebuilt with sidewalks in the center and traffic lanes around them. Some design decisions are actually good.
The issue isn't that an orderly distribution of North Koreans across China would be an issue, but they would be very concentrated in a more remote area of China.
Again, we are talking about the number of people far beyond one that American government can feed in their own country in organized fashion using well-established infrastructure and bureaucracy. Now, imagine half of American elderly -- but moved to China, living in tents. Then think, how Americans would be of any help there.
Your argument is that a bullet in your brain isn't an issue if the total amount of lead isn't deadly from lead poisoning. If you want to talk about what a stupid idea is being tossed about, that would be it.
My argument is, American resources would not be only useless there, they would be completely unnoticeable due to the enormous scale of the problem.
That happened after the war, mostly involved forces other than the former military of Iraq, and Iraq resources were not exhausted because it was just as full of oil as before the war. If anything, former government kept locals from fighting each others, something that Americans only understood years later.
Unix has it's problems, too. Plan9 proved that there was a much better way to do things. Most new features on Linux are half-assed because of the need to maintain some semblance of backwards compatibility. For example, namespaces---Plan9 executed namespaces perfectly; Linux namespaces are an abomination. And yet... they're good enough. Because reinventing the wheel just isn't cost effective.
Anti-Linux trolls are getting more sophisticated again.
Once Wayland components developers started trying to implement something practical, they discover, one by one, that they need those "unnecessary" X features after all, however there is no way to explain it to the rest of developers, who still believe that removing everything they don't immediately use in their narrow area is a great design practice.
It could be scary, for example, if your military claims it has detection technology for this and the technology doesn't actually work.
Except it's the other way around.
so long as we promise to take care of the refugees.
China has FIVE TIMES MORE PEOPLE than you do, and they are ACROSS THE BORDER from Northern Korea. North Korea has 1/12 of US population, and that's half the number of Americans that now are receiving Social Security benefits.
If Chinese see refugees as a problem on their territory, Americans would make absolutely no difference. Words can't describe what a stupid idea that is.
Stalin's entire strategy was to throw warm bodies at the German guns
That's a talking point of American propaganda, created at the time of Cold War to denigrate Russians. Stop repeating it.
At the start of the war USSR was poorly prepared for it -- Stalin was convinced that Germany is not interested in attacking USSR because he seen it as a part of Western politics, and Stalin's own doctrine was that USSR must develop Socialism and Communism while keeping itself in isolation from Capitalist countries around it. All they had was a lot of people and some tanks, so with sufficiently dishonest commentary it's possible to present it as "throwing warm bodies" at Germans, however it was not a strategy and certainly not a winning one, as most of the Western Russia, Ukraine and everything to the West was occupied by Germans by the end of 1941, half a year after the initial attack. By then, however, USSR rebuilt large chunks of its industry in the Eastern part of the country beyond the Ural mountains, and organized the production of military equipment that was designed either before or during the war.
Once supported with steady production of weapons, USSR military became stronger while Germans became hobbled by logistics over a large still-hostile territory, and cold winter did not help them, either. The events after that include huge battle of Stalingrad and truly massive battle of Kursk (yes, that's thousands of tanks on both sides). While losses were still high, that's hardly "warm bodies" or "peasants" thrown at "sophisticated Germans". At that point Germans were pretty much doomed as USSR had rebuilt industry backed by natural resources of a huge territory on the East, and modern military while Germany was scrambling for oil, metals and other resources with dwindling size of their military.
Our big fear was that Japan would fight to the last man and that we'd have to invade.
And it was a completely baseless fear, created by American propaganda for the purpose of justifying the nuclear bombing. At no point, before or after, such idiotic thing ever happened -- when military is nearly destroyed, and resources are exhausted, they don't magically start fighting better.
Congratulations, you are an idiot (and American patriot).
Go, try to run an application built for GTK over X with shared GTK library built for Wayland.
Say, Firefox.
No I don't. If you attend conferences and speak to people inside microsoft, you would know it. You deserve some kind of medal for being this stupid.
You believe Microsoft marketing, and you call others stupid?
What? RTX has nothing to do with unix. http://www.intervalzero.com/
It also has nothing to do with any practical application.
Says the person who thinks unix is good design. I think I will use this as a joke now. Thanks !
Your arrogance is misplaced, as you defend Microsoft design, the result of decades of ad-hock writing, managed by people who know nothing about software development.
Haha.. anger management problems as well. Man, you have a long list of mental defects. Sorry about that. :( Goodbye.. I mean.. fuck you.
I don't have anger management problems, I just hate you. People who do not hate scum like you and your Microsoft overlords, are themselves morally bankrupt.
What do venture capitalists and CEOs know about innovation?
Pronunciation.
P. S. Just like SOAP. Remember SOAP?