XWayland Adds "-Output" Option For Better Rootful Fullscreen Control

[mSparks]

your very well observation fits into another observation i did make and it has to do with power9 is fully opensource and power10 is closed source...

but the power10 ISA itself Power-ISA-3.1 is opensource and there is a chip company making such CPUs: https://www.solidsilicon.com/desktop/index.html

my observation is that all the opensource/free-software people did left IBM/RedHat the CPU hardware people did go to SolidSilicon the GPU people did go to RED-Semiconductor https://redsemiconductor.com/ and most of the Linux/software people did go to Suse-Linux and Fedora.

your observation is exactly the same as my observation any sane opensource/freesoftware person left IBM/REDHAT and they flee to Suse/Fedora and SolidSilicon and REDSemiconductor.

you say you did go to oracle linux "OL8" well this is a movement what looks like better than IBM/REDHAT to...

probably the first interesting post the whole thread.
Ive not had much contact with the ibm power series stuff, not sure if that is because Im Europe based and there isn't the same presence here. did cut my teeth on the cell processors in the early ps3s, but binned that when geohotz (iirc) did his hacking thing and sony pulled the "other OS" from new devices.

everything is in such a wierd place right now, like I mentioned those 10 dell r710s, with a combination of perf improvements and better cpus my m1 macbook air can complete jobs faster than the early runs of those... But we are totally not geared up for rolling that out to the consumer space.

nvidia, good grief nvidia, thats changed the game, theoretically, but for the longest time our bottleneck has been disk storage and networking rather than compute (especially with the latest AMD cpus), getting fast there is still bonkers expensive, wifi is now hitting faster speeds than wired.... wtf!

I think maybe the root of the problem is many of what were hard to solve problems are solved problems now (or maybe more the whole solutions looking for problems problem), and nobody really has any idea what to do next.

Anecdote
I still build any windows stuff that needs recompiling on a windows 7 VM.

My hopes for things to improve are still pinned on openELA, they have the A-team now, but I very much suspect the next big industry shift is going to be driven by "apple vision" - depending on if its as good as it looks.

Imagine where linux application development could be today if the effort wasted on wayland had gone into a decent X11 extension for XR, occulus sold more headsets last year than any other game console aiui, and certainly more than ms shipped copies of windows....

But nooo, instead we get a command line options replacement for xrandr, wahhooo, amazing... not.

[oiaohm]

My hopes for things to improve are still pinned on openELA, they have the A-team now, but I very much suspect the next big industry shift is going to be driven by "apple vision" - depending on if its as good as it looks.

https://www.infoworld.com/article/3705094/follow-the-cloud-money.html

Problem here is everyone who has gone looking as found that openELA does not have A-team.

https://aws.amazon.com/partners/redhat/

Yes Redhat is official partner with amazon so you have official support from Amazon and Redhat on AwS. Now you go Oracle or Rocky on AWS you will be lucky for those to be in market place without official support from Amazon..

https://aws.amazon.com/marketplace/seller-profile?id=e6a5002c-6dd0-4d1e-8196-0a1d1857229b

This is something you find with cloud providers. Redhat, Ubuntu, Suse and their own custom rolled.

SUSE use to have a-team but other companies have taken those personal.

[qarium]

Problem is from what I can see how the you abuse AGPL attempting to fix it will cause other legal problems. If what you put in the license is classed as "Coercion" its illegal license.
You cannot force a company or a person to-do business with anyone they don't have a business relationship with by contract/license. You cannot write a license in fact without an exit clause of some form. That why AGPL/GPL only has to give code to the user. This is as far as you can legally go and still have a valid license in any contract/license I have seen.
When you get into law there are limited sections of law that can force you todo business with people you don't want to. Anti-trust and anti-discrimination can.
Contract law and Copyright law sorry this not allowed. Contract and Copyright license law does not allow to use "Coercion". Making party like Google give you the source code and not cancel your account falls under needing "Coercion" as in the means to directly limit party like Google actions. Anti-trust to make sure business are competing with each other fairly is allowed to use Coercion to set what are acceptable business practices under law. .
Yes it would be nice if this problem was fixable by License change but there are legal limits on what you can legally do in a license.
Yes if the Coercion is written into some countries anti trust law and you have your license say that its enforcement is inside that country that countries anti-trust rules apply, So you cannot write this directly into the license but if the case has been done you can indirectly apply it.
Legally Linus Torvalds by himself cannot re-license the Linux kernel. There is no CLA or anything else to allow Linux kernel to change license without complex process being done. This means the only way to change the Linux kernel license is contact all authors who have submitted code and get their approve from them or their companies.
qarium how quick do you think google/facebook/....who have submit code to the Linux kernel legal department are going to respond with no we don't approve of the Linux kernel license change so detailing the process.
Like it or not there are many projects were changing license will be impossible because the license change you are trying to get will have to have approve from the very parties who are not doing what you want.

its not impossible to change the license of the linux kernel many developer could just agree to do so and for the rest any code of developers who do not want it could be replaces by new code with the new license. this could be a good reason to move the new code from C to rust.

"You cannot force a company or a person to-do business with anyone"

right this is something we maybe can not fix with a license but you did make many more examples what could be fixed by license.

"There is no CLA or anything else to allow Linux kernel to change license"

linus torvalds has something much better than CLA... he has control over upstream means he could only accept upsteam patches with the new license.

[qarium]

probably the first interesting post the whole thread.

thank you very much. these infos i share are infos hard to get because the truth is promoted nowhere because all the big money people only promote closed source and proprietary stuff.

Ive not had much contact with the ibm power series stuff, not sure if that is because Im Europe based and there isn't the same presence here. did cut my teeth on the cell processors in the early ps3s, but binned that when geohotz (iirc) did his hacking thing and sony pulled the "other OS" from new devices.

the cell processor was a flop yes but a fully opensource Power-ISA-3.1 cpu is not a flop. it does very well.

everything is in such a wierd place right now, like I mentioned those 10 dell r710s, with a combination of perf improvements and better cpus my m1 macbook air can complete jobs faster than the early runs of those... But we are totally not geared up for rolling that out to the consumer space.
nvidia, good grief nvidia, thats changed the game, theoretically, but for the longest time our bottleneck has been disk storage and networking rather than compute (especially with the latest AMD cpus), getting fast there is still bonkers expensive,

i do not believe in nvidia at all just to make you see what kind of person i am the last gpu i did buy was a AMD PRO w7900 48GB for 4000€
i do not believe that nvidia and cuda has any future in the opensource/linux world.
but ok you can have a different opinion.

these dell r710s, you talk about are "Quad-Core oder Six-Core Intel® Xeon® Prozessoren der Reihen 5500 und 5600"
i would say you better go with AMD cpus who support AVX512 on their double pump 256bit engine.

wifi is now hitting faster speeds than wired.... wtf!

this is not the case because wifi is a shared medium and for example 10GBPS RJ45 copper network cards are exclusive speed
yes the newest wifi standards go higher than 10GBPS but because it is a shared medium the more people are at the same place the less speed they get because it is a shared medium.

and remember this 10GBPS RJ45 copper network cards operate with CAT6/CAT7 twisted pair cables
but CAT8/8.1 cables can operate at 1600-2000mhz and 2000mhz result is then 40Gbps

also even without CAT8 you could simple do teaming with 2 lines what would result in 20Gbps and for example in my house we already did put in CAT7 double line in the wall means with teaming it is already 20Gbps

just see this 2000mhz 40Gbps as example: https://www.amazon.de/dp/B09YH81HZJ?...hals10-21&th=1

I think maybe the root of the problem is many of what were hard to solve problems are solved problems now (or maybe more the whole solutions looking for problems problem), and nobody really has any idea what to do next.
Anecdote
I still build any windows stuff that needs recompiling on a windows 7 VM.
My hopes for things to improve are still pinned on openELA, they have the A-team now, but I very much suspect the next big industry shift is going to be driven by "apple vision" - depending on if its as good as it looks.
Imagine where linux application development could be today if the effort wasted on wayland had gone into a decent X11 extension for XR, occulus sold more headsets last year than any other game console aiui, and certainly more than ms shipped copies of windows....
But nooo, instead we get a command line options replacement for xrandr, wahhooo, amazing... not.

i really do not unterstand your anti-wayland hate is is plain and simple not mature enough for you wait 1-2 years and you will see it becomes mature then.

i think in the future X11 De people will run their DE in a wayland rootful mode

[oiaohm]

"You cannot force a company or a person to-do business with anyone"

right this is something we maybe can not fix with a license but you did make many more examples what could be fixed by license.

The redhat method of malicious compliance because they really do give you everything should just cut of access if you don't do what they like is one of the ones you cannot deal with using license.

The highly costly media problem. What right do you have to define a companies trade dress in a software license? The answer is you don't have the right to. The highly expensive cd and what on it the company can class as trade dress. Trade dress is to prevent counterfeit products.. Anti-trust/business court ruling don't have this problem they can rule that this item has to be traded in plain packaging.

Even the disc being 650meg byte CD a company can define this as their trade dress standard.

Reality here is most of my examples in fact need legal "Coercion" to fix without causing major problems.

Something to remember don't understatement the data transfer rate of package with a stack of DVD/CD inside. There are still places around the world where it quicker to send stack of discs than use their internet.

its not impossible to change the license of the linux kernel many developer could just agree to do so and for the rest any code of developers who do not want it could be replaces by new code with the new license. this could be a good reason to move the new code from C to rust.

Replacing new with old by prior examples that have been done with license changes to and away from GPL have end up in court for possible infringement with the new vs the old code.

Please note I am not saying it impossible. Just the process absolutely will not be fast. It possible anti-trust case could complete before the Linux kernel could change it license to a new one. Basically smaller projects with multi developers the process has taken over 10 years to complete.

With the size of the Linux kernel 20-30 years to change license by agreement/replacement is absolutely not off the cards. Yes anit-trust cast last a max of 15 years if you go though every single court.

That my problem the fast way to fix this issue may be the anti-trust case.

linus torvalds has something much better than CLA... he has control over upstream means he could only accept upsteam patches with the new license.

Also he has to worry about it wage from the Linux foundation as well.

[mSparks]

these dell r710s, you talk about are "Quad-Core oder Six-Core Intel® Xeon® Prozessoren der Reihen 5500 und 5600"

6 cores, 12 threads, 2CPUs each for 24 vCPUs a pop, 240 in total, 4x UPS, 2 racks in the cluster drew roughly 10-15kW at full tilt, full run would take about a week.

Important part there is the 15kW... no one told me about that, getting rid of 15kW of heat is quite the interesting engineering challenge. Bought back in 2012, this is before data centers were a thing. mostly running LLMs, 3 years before openAI was founded.....

outperformed now by a single M1 chip drawing 10W,
AMD EPYC completely changed the landscape,
CUDA and Tensor cores make analysis we never dreamed of possible now.

also even without CAT8 you could simple do teaming with 2 lines what would result in 20Gbps and for example in my house we already did put in CAT7 double line in the wall means with teaming it is already 20Gbps

yeah, 20Gbps, is 2.5GBps, 1TB of disks filled in 400 seconds, call it 10 minutes..... That is still very much the bottleneck, storage basically the same price if not increased slightly from 2012 with the switch to SSDs... SAS drives not changed much afaics.

10 minutes, 3 hours, same difference tbh.

much cheaper and easier to solve by spinning up linnode or digital ocean instances now, but then the limit is public internet speeds..

i really do not unterstand your anti-wayland hate is is plain and simple not mature enough for you

Its nothing to do with its "maturity", everything to do with the utter hyperbole and hubris that surrounds it, X11 is simply a far, far superior design, that fits the "linux use case" perfectly. The world became more X11, Everyone now uses multiple devices on powerful home networks, wayland is the exact opposite of X11, designed for graphics/applications to be locked in to a single device and never leave it.

Yet "Wayland will replace X11".

Or to put it in its technical terms

multiple devices on powerful home networks are going to be replaced by graphics/applications locked in to a single device and never leave it.

How am I supposed to resist slapping those people with a wet fish for stupidity?

[oiaohm]

multiple devices on powerful home networks are going to be replaced by graphics/applications locked in to a single device and never leave it.

How am I supposed to resist slapping those people with a wet fish for stupidity?

Network transparency stopped working correctly with X11 a long time ago.

waypipe does show that output of wayland application can be done on a different computer.

mSparks you are the one that need to be slapped around with the wet fish for saying stupidity

RDP is better optimized for crossing network than X11.

https://silverweed.github.io/assets/docs/distributed_rendering_in_vulkan.pdf

Then you have stuff like this where opengl/vulkan themselves are directly altered to cross network. This has a lot in common with how waypipe works in fact.

[mSparks]

waypipe

Not something I will ever need to worry about

its taken waypipe 15 years to get to version 0.8, at that rate it will be production ready in another 75 years. I'll be long dead by then.

[qarium]

6 cores, 12 threads, 2CPUs each for 24 vCPUs a pop, 240 in total, 4x UPS, 2 racks in the cluster drew roughly 10-15kW at full tilt, full run would take about a week.
Important part there is the 15kW... no one told me about that, getting rid of 15kW of heat is quite the interesting engineering challenge. Bought back in 2012, this is before data centers were a thing. mostly running LLMs, 3 years before openAI was founded.....

this system is clearly legacy by now. i mean you could replace these 10 dell r710s with one single 192core ARM ampere altra system and you would have much more performance.

can you explain the rationality of keeping this system and burnm 10-15kw instead of having just 1 system and only burn 1kw ?

outperformed now by a single M1 chip drawing 10W,
AMD EPYC completely changed the landscape,
CUDA and Tensor cores make analysis we never dreamed of possible now.

right appel m1 is nice but it is my opinion that the Qualcomm Snapdragon X Elite is even better soc

i honestly do not unterstand why you are in favor of CUDA instead of open standards ?

yeah, 20Gbps, is 2.5GBps, 1TB of disks filled in 400 seconds, call it 10 minutes..... That is still very much the bottleneck, storage basically the same price if not increased slightly from 2012 with the switch to SSDs... SAS drives not changed much afaics.
10 minutes, 3 hours, same difference tbh.

with CAT8.1 and 2000mhz and teaming of 2 cables you have 80Gbps... but the best networkcard i ever see was a 10Gbps card
so we need to wait until new network cards for Cat8.1 cables emerge.
my house in germany did get fiber internet with 1Gbps and in some regions they already roll out 10Gbps fiber internet...

you call it a bottleneck but its the fastest we have for decades now. before that we had a 100/40 Mbps vector DSL line...

much cheaper and easier to solve by spinning up linnode or digital ocean instances now, but then the limit is public internet speeds..
Its nothing to do with its "maturity", everything to do with the utter hyperbole and hubris that surrounds it, X11 is simply a far, far superior design, that fits the "linux use case" perfectly. The world became more X11, Everyone now uses multiple devices on powerful home networks, wayland is the exact opposite of X11, designed for graphics/applications to be locked in to a single device and never leave it.
Yet "Wayland will replace X11".
Or to put it in its technical terms
multiple devices on powerful home networks are going to be replaced by graphics/applications locked in to a single device and never leave it.
How am I supposed to resist slapping those people with a wet fish for stupidity?

X11 hat over 40 years of development... if we give wayland the same timeframe then yes wayland is not yet mature.

"X11 is simply a far, far superior design,"

to be honest this is clearly not the case X11 is a legacy implementation with design goals no longer justified by reality and this for like 20 years now.

you are clearly no X11 or wayland developer because then you would admit that X11 can not reach government contract standards as ISO 27001
x11 plain and simple can not be as secure and can not garantee privacy to fullfill the ISO 27001

wayland instead by design can do this.

[mSparks]

this system is clearly legacy by now. i mean you could replace these 10 dell r710s with one single 192core ARM ampere altra system and you would have much more performance.

Very, it's not been powered up since 2018, maybe 2019. That earlier screenshot of it running is a very old one I keep as a momento from, oh, September 9, 2011,

https://web.archive.org/web/20130625031656/http://theborgmatrix.com/

But the software stack (all in house) that ran on it is very very much alive, last build/deploy was 4.232.6.0.2024/01/08, it did take four complete rewrites to hit something production ready.

running large international projects predominantly in healthcare, but also aviation and finance.

Aviation stuff is the closest the general public can get to it

https://www.youtube.com/watch?v=_GzhwJMFDHk

It can generate a 500GB file for subanalysis from a $20/m VPS faster than anyone can download it. The "toys" I ran on it now run quite happily 24/7 on an older intel mac mini. all it needs to deploy is an OS with ssh -X support.

i honestly do not unterstand why you are in favor of CUDA instead of open standards ?

I would absolutely love for open standards to be competitive with what nvidia have created, they just aren't, not even close. Not yet anyway.

to be honest this is clearly not the case X11 is a legacy implementation with design goals no longer justified by reality and this for like 20 years now.

AFAICS that's just redhat propaganda, xorg-server releases have run like clockwork for decades

https://www.x.org/releases/individual/xserver/

It works, it does the job, better than any of the alternatives (and there are a lot of alternatives) its no more a "legacy implementation" than TCP/IP or HTML, X11 itself is a very simple protocol, any complexity comes from extensions, like, HTML is a very simple protocol, an HTML5 canvas or SVG support, not so much.

Now, is it perfect, nope, a nicer security extension would be lovely, something like a multiple domain config would be awesome, I should be able to fire up my android phone and run all my linux workstation applications directly off it with the click of an icon, click the same icon on my macbook air and carry on where I left off. VR support should have gotten way more attention (monado guys are doing a great job of that, but a little slow for my liking).

None of that has happened, imho because the people being paid to do it are wasting their time trying to replace xrandr with command line options for a display protocol no more than a few thousand people will ever use, and will never see the light of day in production use - at least not in my lifetime.