Fedora 40 Eyes Dropping GNOME X11 Session Support

[Mez']

But we were talking specifically about the technological advancements Red Hat brought to Linux, both in kernel and in userspace (Systemd, dbus, Pulseaudio, Pipewire, etc). For example, do you know how painful was the audio experience on Linux before PulseAudio existed? (when you only had plain ALSA). And these days, who's bringing HDR support to Linux? That's right, it's Red Hat. That's what I'm talking about. These are the advancements that brought the desktop experience at least to par with Windows and MacOS.

I know what they brought, and I was also talking about the technologies. As I said, the only one of value in my opinion is pipewire, puleaudio has been a broken mess for years (a bit like wayland), which has brought me a lot of grey hair. And yes, I was using Linux much before Pulseaudio, so yes I know what it used to be.

It sounds like you simply don't like Gnome and its app ecosystem. That's fine, switch to KDE, or Cosmic when it's going to be released in the near future.

Again, switch to KDE, Cosmic, or any other DE. So I don't get your problem. Is someone forcing you to use Gnome apps or something?

Don't know why people keep giving those dumb advices to "switch to X if you don't like this or that". If so, you can basically be done with the 12 most popular Linux DEs in a few days and have nothing left to try. There's a broken tile in your kitchen, a bit of mold in the bathroom and a door unhinged in the attic, make sure to sell your house and go buy another one...
How about treating the problems instead?

But beyond that though (since they won't and stay with their head buried in the sand towards the many problem each of their technologies bring), it's not that I don't like Gnome (I just think their implementation of the vision is terrible, but the vision itself is good) but the mentality of Gnome devs in general and their bad design decisions due to being close-minded. And that is exactly why I think their technologies systematically disserve the Linux community at the end of the day, in addition to being bullies to other technologies of potential greater value.

[oiaohm]

absolute nonesense.
You cannot administrate, schedule and monitor graphics jobs going out to 20000 servers with gpu direct and cuda

https://www.nvidia.com/en-au/data-center/bright-cluster-manager/

You don't administrate or schedule a NVIDIA HPC using X11 that all done by web browser or command line. That right Nvidia bright cluster manager does not have a X11 or Wayland client.

As I said the writing is on the wall with Nvidia in their HPC the officially support solutions don't have X11 or Wayland.

that report their status from as simply as current FPS to as complex as displaying exactly what is being rendered.

Current FPS information of rendering that able to be collected by bright cluster management no X11 required. GPUDirect supports on windows client to see what is currently being rendered not supported when using X11 or Wayland..

kubernates is at the core of what Nvidia up to here. mSparks what trace of X11 does kubernates have.

mSparks what has been done for decades does not mean that will be done for ever. Like it or not Nvidia has not moved to Wayland and is moving away from X11 at the same time HPC.

https://developer.nvidia.com/hpc-sdk

This again no X11 required. Nvidia has systematically removed X11 usage from their HPC solutions like it or not mSparks. So Nvidia HPC market does not help Linux desktop users in any major way.

Yes one of the reasons why Nvidia X11 drivers are so bad is for their big HPC customers they don't use them anymore. Yes this started with the EGLdirect stuff quite a while ago. Why take a % hit running X11 server when you graphical application that running headless can go straight to the GPu instead with basically zero overhead.

In fact the way Nvidia HPC market kind of explains why Nvidia drags their feet updating off GLX to DRI3. Yes DRI3 Nvidia drivers would equal Wayland compositors working as well.

Something to consider Nvidia does not want your consumer GPUs to be usable in the high process HPC market. So it in Nvidia best interest to keep the Nvidia X11 driver being total garbage for Wayland and not support DRI3 as long as possible. DRI3 secure way to run multi applications on single GPU can you see problem for Nvidia HPC market here. Nvidia wants the X11 solution insecure so use in HPC is highly limited so they can sell other products like the bright cluster manager.

[mSparks]

mSparks what trace of X11 does kubernates have.

Just the meaningless token of "best practice".

https://www.alibabacloud.com/help/en/ack/ack-managed-and-ack-dedicated/user-guide/best-practice-for-linux-graphics-applications

[oiaohm]

Just the meaningless token of "best practice".

That is graphical applications as in remote desktop. Not for HPC.

https://hub.docker.com/r/alsadi/containerized-gui-wayland

The reality is that documentation is getting very old very quickly. Notice here wayland application access by webbrowser so you don't need anything client side. Also your application does not go splat because your internet connect gets disrupted.

Really you need to have used that on alibabacloud. The diagram is deceptive. X11 over network is in fact blocked on alibabacloud include over ssh -X.

nvidia-xconfig -a --virtual=1920x1080 --allow-empty-initial-configuration --no-use-display-device --enable-all-gpus --busid "PCI:0:7:0" -o xorg.conf

This is a Nvidia grid configuration remote.

https://resources.nvidia.com/en-us-v...-virtual-pc-vi scroll down and curse. Notice no CUDA/opencl support for Linux clients at all using the method alibabacloud documents. Hopefully you like paying windows licenses for your CUDA processing going that route.

This is the thing Nvidia is slowly but surely doing X11 in. Nvidia grid solutions X11 remote has been non functional for years.

[mSparks]

That is graphical applications as in remote desktop. Not for HPC.

https://hub.docker.com/r/alsadi/containerized-gui-wayland

The reality is that documentation is getting very old very quickly. Notice here wayland application access by webbrowser so you don't need anything client side. Also your application does not go splat because your internet connect gets disrupted.

Really you need to have used that on alibabacloud. The diagram is deceptive. X11 over network is in fact blocked on alibabacloud include over ssh -X.


This is a Nvidia grid configuration remote.

https://resources.nvidia.com/en-us-v...-virtual-pc-vi scroll down and curse. Notice no CUDA/opencl support for Linux clients at all using the method alibabacloud documents. Hopefully you like paying windows licenses for your CUDA processing going that route.

This is the thing Nvidia is slowly but surely doing X11 in. Nvidia grid solutions X11 remote has been non functional for years.

Good grief.
When you are running an app on 20,000 servers you don't log in to them one by one by remote desktop.
You start an X11 app on one of the cluster masters that sees the entire cluster as one homogenous CPU/GPU it can dispatch calls to.
99% of linux boxes in the wild dont even have a desktop environment or display server installed.

What they do have however is very expensive X11 clients and servers that actually do the work, and xwayland breaks.

[oiaohm]

Good grief.
When you are running an app on 20,000 servers you don't log in to them one by one by remote desktop.
You start an X11 app on one of the cluster masters that sees the entire cluster as one homogenous CPU/GPU it can dispatch calls to.
99% of linux boxes in the wild dont even have a desktop environment or display server installed.

Except what you just described does not work inside the "alibabacloud" system. There configuration you have to start x.org server on the machine with the X11 application. X11 traffic over their network will be blocked by their switchs and their ssh has the -X functionality disabled.

The X11 over network bit is dead in lots of setups. GRID provides the output over HTML.

The reality as soon as Nvidia drivers correctly support wayland you would just swap the X11 for Wayland and there will be no issue. Problem is currently Nvidia grid vGPU does not support Wayland.

Nvidia grid setups to run X11 application mean you have to run x11 server with X11 application.

Wayland breaks really mSparks with Wayland is nothing more than Nvidia has not updated their drivers to have functional Wayland support. Wayland lack of remote in Nvidia grid is not a problem because Nvidia grid setups don't allow X11 protocol to go over network anyhow.

Dispatch calls into the cluster you have started Nvidia Grid X11 server you have now disabled CUDA you now cannot use CUDA messaging between your clients. This is when you start cursing. Yes the GRID vGPU is CUDA compatible and opengl compatible and vulkan compatible as long as you use EGLDirect mode and never start X11 server.

Yes EGLDirect you can still draw to screen application and have control interfaces that Nvidia GRID sends over network.

Lets just say the use case you are trying to say X11 has to stay is kind of broken bit of mess.

[mSparks]

Except what you just described does not work inside the "alibabacloud" system. There configuration you have to start x.org server on the machine with the X11 application. X11 traffic over their network will be blocked by their switchs and their ssh has the -X functionality disabled.

Well elevating yourself to pretending to be a network guru, in an alternate world where switches have OSI layer 4 filtering had to happen eventually I guess. I just didnt expect you to fall so far so soon.

You almost had me convinced people who think wayland has a bright future aren't completely delusional, for at least a second there, but you should stick to learning about the unix philosophy and "everything is a file". There is no way you will learn networking this side of 2030, and its far to off topic for me to help you here.

[oiaohm]

Well elevating yourself to pretending to be a network guru, in an alternate world where switches have OSI layer 4 filtering had to happen eventually I guess.

Not pretending.

https://wiki.wireshark.org/X11.md

X11 protocol over network is really simple to kill just block all port connections '6000-6063'

https://www.planet.com.tw/en/products/lan-switches/layer-l2-l4-managed-ethernet-switches

Yes Layer 4 Switches exists. China had them for quite some time and the China government uses them in the great china firewall.

Who failed so soon mSparks you are only decade of switch technology out of date.

[jorgepl]

Redhat gave linux many things.
Unfortunately the only thing that remains of Redhat mid 2023 is the logo. IBM fired most of the staff

https://www.theregister.com/2023/04/28/red_hat_layoffs_union/

the rest are leaving of their own free will and IBM completely reoriented their remaining staff to try and be more like Microsoft, which while it may make sense at face value, is pretty much never going to work out for them.

I'd say Fedora dropping X11 for wayland is wayland's last chance (and to a certain degree RH). If it doesn't work out (and it wont), even more heads will roll, only this time in upper management.

You're imagining a crisis that doesn't exist.

[mSparks]

You're imagining a crisis that doesn't exist.

crises would imply immediacy. RHEL8 is fully supported until 2029, its only 2023.