Announcement

Collapse
No announcement yet.

XWayland 24.1 Planned For Release Next Month With Explicit Sync & Other Features

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • mSparks
    replied
    can confirm nothing depends on wayland, xwayland or their drivers, it is safe and easy to uninstall.
    Last edited by mSparks; 16 April 2024, 03:04 AM.

    Leave a comment:


  • Blademasterz
    replied
    Haters gonna hate, better do something productive

    Leave a comment:


  • oiaohm
    replied
    Originally posted by mSparks View Post
    DELETED
    Try again this is a repeated invalid argument disproved before.
    Phoronix: XWayland 24.1 Planned For Release Next Month With Explicit Sync & Other Features The XWayland 23.2 series was introduced last August while now release preparations have begun for releasing XWayland 24.1 as the next feature release for this X.Org Server derived code for allowing X11 clients (apps / games) to work


    I see you have run out of arguments now are attempting to replay.

    The reality I already covered that like it or not implicit sync is required in a display server. Without it your display server is unstable. This is 50 year old known stuff about memory management.

    You cannot do stable explicit sync without particular things being implicit sync. So it makes sense to get the implicit sync functionality of the protocol correct first since this need to be correct. Also do remember Wayland would have had explicit sync in 2016 if all GPU vendors had agreed. So its not 15 years that the Wayland developers were anti explicit sync in the display server.

    It took Nvidia 15+ years to work out that pure explicit sync does not work and then to decide to play ball with everyone else..
    Last edited by oiaohm; 15 April 2024, 10:54 PM.

    Leave a comment:


  • mSparks
    replied
    Originally posted by oiaohm View Post

    Being a network protocol mean it design should not be just depending on the host OS kernel for security.
    Entirely depends on if you -Y or -X

    Originally posted by oiaohm View Post
    the nightmare.
    The nightmare here seems to be you telling me wayland bypasses ~50 years of security research thinking it knows better... an application whose developers thought implicit sync was a good idea for a display server and took 15 years to realise their mistake thinks they can do security..... That is terrifying, so much so everything wayland will be uninstalled from all my machines by the end of the week, I am now going to recommend everyone else does the same.

    Leave a comment:


  • oiaohm
    replied
    Originally posted by mSparks View Post
    Same reason cat and ps are "historically at core insecure"
    Because applications dont enforce permissions in Linux, the kernel and apparmor/selinux does.
    Except these argument don't for X11 protocol. Being a network protocol mean it design should not be just depending on the host OS kernel for security.
    Also X11 core protocol predates LInux and this is where things get problematic..

    Originally posted by mSparks View Post
    Does wayland not abide by that most basic of software design principles or something?
    You need to wake up and ask these questions about X11 protocol.

    X11 protocol avoids using OS file handles and other OS controlled structures to be able to be so called cross platform to non Unix OS. This is why attempting to make XACE work with X11 was uphill battle that turned out to be impossible to complete. Wayland on the other hand uses resources the Linux/Unix platform can control and has been designed like this from basically day one.

    X11 protocol was not designed for Unix in fact was attempted to be designed to be generic. Wayland being design for Unix/Linux platforms does mean it can outsource successfully security things to the host Unix/Linux security frameworks in a very transparent way.

    There is a set of words about X11 that are at it core design that explain the nightmare. "create mechanism, not policy" yes these word are from the two founding authors of X11 being Bob Scheifler and Jim Gettys. Yes worry about creating mechanism to get stuff done don't worry about security/stability/policy stuff at all. X11 protocol problems come from its founding principle.

    X11 protocol was design for the 1987 time frame same time we were using msdos and using telnet when security and stability was not the highest fact when segfaults from buffer overflow, use after free and so on were normal day things.

    mSparks basically another attempt at a counter argument that does not hold up.

    Leave a comment:


  • mSparks
    replied
    Originally posted by oiaohm View Post
    Why was X11 historically at core insecure
    What...?
    Same reason cat and ps are "historically at core insecure"
    Because user applications dont enforce permissions in Linux, the kernel and apparmor/selinux does.

    Does wayland not abide by that most basic of software design principles or something?

    Ouch, it's worse than I thought, even worse than I thought. that kind of fuck up is almost certainly terminal.

    wow.
    Last edited by mSparks; 15 April 2024, 09:40 PM.

    Leave a comment:


  • oiaohm
    replied
    Originally posted by mSparks View Post
    this is the root if the issue tho, you dont seem to understand the design of X11.
    It you who don't know the X11 protocol and need to take those rose color glasses off.

    Originally posted by mSparks View Post
    X11 is absolutely unrivalled, and will remain absolutely unrivalled for what it is designed for, not least because none of the others even try.
    This is a problem you don't know what it was designed for.

    Originally posted by mSparks View Post
    X11 is designed for GUI when compute and presentation are not just isolated, but physically separated. although html5 and "remote desktop" based solutions have a fair bit of cross over into its domain they dont come close to the feature set, versatility, cross platform compatability, ease of use and performance.
    This is you not understanding the X11 core protocol. Physically separated and isolated don't equal the same thing. X11 core was design for private LAN where all applications could be trusted. This is critical this is why the X11 core protocol have so many race conditions and lack or resource protections baked straight into the protocol. X11 is design that everything explicitly does the right thing with each other. X11 core protocol has no consideration that application could do the wrong thing. X11 core protocol was not designed with any form of isolation.

    Originally posted by mSparks View Post
    ​That design has also proven itself to be VERY difficult to even match, let alone beat in a "desktop use" environment.
    This is not true Apple and Microsoft both proved this false. Reason why X11 has been so hard to replace on Unix/Linux is the reality is alternatives that were basically fully functional yes more functional than X11 or Wayland never got GPU vendor support.

    Originally posted by mSparks View Post
    Not least because many of the assumptions the push to wayland were based on turned out to simply be wrong.
    Start naming them. I will take two out straight away. Every frame is perfect is still active in Wayland only applications that ask for tearing don't get this. That assumptions were wrong is not case. Tear free out for majority of users that what they need.

    Implicit sync on resource management vulkan and Nvidia both these days see that as required. Not optional. Wayland core was going to have to be design to support implicit sync because wayland protocol has a lot of resource management.

    The reality is large number of the so called incorrect assumptions under closer look is over 90% correct.

    There is a fun point about this explicit sync protocol added to Wayland. Turns out not to have major issue you need implicit sync control over the explicit sync interface pass out to applications so that if applications share a buffer so you don't end up with split brain.
    https://en.wikipedia.org/wiki/Split-brain_(computing)
    Yes eglstreams did suffer from split brain events as well as race conditions, segfaults.... yes all the faults core X11 protocol suffer from as well. Those who say explicit sync for everything need to wake up. Implicit sync is required without it you have unstable mess. Explicit sync is right to be optional as it just a method to get high performance not improved stability. Fun right explicit sync you cannot make function correctly without implicit sync in the mix.

    Why was X11 historically at core insecure and unsafe explicit sync is because not caring about security or stability you are to get higher performance and for the limited computing power at the time this made sense and was important but we are not in the 1970s/1980s any more so it does not make sense any more to give up this amount of security and stability. Yes it the same as saying people should use MS dos instead of windows because it will be faster and it faster due to the lower security and stability.

    mSparks do list what were the trade offs that were done when Core X11 protocol was designed I can list those. That list is not a good read for those attempting to push X11.

    Originally posted by mSparks View Post
    how many linux boxes have 20+ year old gpus.
    This is lack of knowledge. You system has a aspeed BMC this is a brand new chip right. What is exposed to the OS from aspeed BMC is equal to a 1987 GPU. https://www.phoronix.com/news/GNOME-Wayland-ASpeed Turns out modern wayland compositors do have to work decently well with 20+ year old bits of hardware. Yes it the same reason why matrox kms driver for the Linux kernel has to be updated it used in BMC.

    20+ year design GPU turn up in brand new hardware mSparks that Linux with Wayland solutions support perfectly. So people in server world using 20 year old GPU designs is way more common than one would think.

    Leave a comment:


  • Noitatsidem
    replied
    Originally posted by mSparks View Post
    "almost as good as X11 for the desktop" is simply not good enough for a linux desktop. Linux users don't care about new shiny and cool, if its the best they are quite happy to - prefer to - keep using it for 20+ years unless something actually demonstrably better - that they need - is available. e.g. how many linux boxes have 20+ year old gpus.
    wait what's so good about the x11 desktop?

    Leave a comment:


  • mSparks
    replied
    Originally posted by oiaohm View Post

    X11 core protocol is very much like...
    this is the root if the issue tho, you dont seem to understand the design of X11.

    X11 is absolutely unrivalled, and will remain absolutely unrivalled for what it is designed for, not least because none of the others even try.

    X11 is designed for GUI when compute and presentation are not just isolated, but physically separated. although html5 and "remote desktop" based solutions have a fair bit of cross over into its domain they dont come close to the feature set, versatility, cross platform compatability, ease of use and performance.

    Additionaly, and tbh quite unexpectedly for everyone.

    That design has also proven itself to be VERY difficult to even match, let alone beat in a "desktop use" environment.

    Not least because many of the assumptions the push to wayland were based on turned out to simply be wrong.

    "almost as good as X11 for the desktop" is simply not good enough for a linux desktop. Linux users don't care about new shiny and cool, if its the best they are quite happy to - prefer to - keep using it for 20+ years unless something actually demonstrably better - that they need - is available. e.g. how many linux boxes have 20+ year old gpus.
    Last edited by mSparks; 15 April 2024, 10:00 AM.

    Leave a comment:


  • oiaohm
    replied
    Originally posted by mSparks View Post
    Hmmm, so I agree maybe I came across to strongly against implicit sync.
    This is the danger of pushing for something and not understanding it. X11 core protocol is very much like MS-Dos memory management both are explicit sync without correct protective items. Making them highly dangerous and unstable. XACE tried to fix X11 of this but never succeed.

    Wayland core is a lot more implicit sync this has advantage blocking Wayland from having a lot of exploits and segment fault risk. Like not being able to mess with another applications memory.

    Originally posted by mSparks View Post
    Both implicit/explicit have very meaningful differences, pros and cons,
    This is the next trap there is not just 1 form of implicit or explicit sync. It also possible to have a solution that is mixed.

    Originally posted by mSparks View Post
    But, afaik, the #1 Pro for explicit sync is performance - which is why from a display server pov, implicit is substandard.
    There is more to consider here. Vulkan did. Vulkan is not pure explicit sync. Everything that called immutable with vulkan when you create buffers is because the buffers you are creating are in fact managed by implicit sync. Yes you free a buffer under Vulkan this is managed by implicit sync. Vulkan and Wayland buffer/memory management in fact lines up.

    Wayland explicit sync is about GPU and other buffer processing operations not buffer create or destory operations just like Vulkan.

    Originally posted by mSparks View Post
    There is very similar arguments for opengl v vulkan, both have a place, because opengl is and will always be far more development friendly, due to most of the hard work being done by the driver, whereas there is no beating a well written vulkan application for performance - but its much easier to create a badly written, buggy, poorly performing vulkan application than an opengl one.
    This is where people straight up go wrong. People fail to notice that Vulkan is not pure explicit sync instead Vulkan is everything bare memory/buffer management explicit sync. with the memory/buffer management being implicit sync. Pure explicit sync is unstable. Multi process Memory management need to be implicit sync if it not you have problems..

    Examples of pure explicit sync are eglstreams and X11 core protocol and both of these items are unstable and suffer from race conditions and segfaults and other issues due to using too much explicit sync.

    Remember the Wayland explicit sync( linux-drm-syncobj-v1) is based on GBM/DMABUF sync files so does not cover creating or destroying a buffer instead leaves that to implicit sync where it makes sense to be..


    Also note the top copyright.
    Copyright 2016 The Chromium Authors.​
    This is in reality is updated version of the 2016 protocol that did not take off back then because Nvidia back in 2016 would not agree to support it. Yes 2016 Nvidia was sure they could get eglstreams to work. There is a older protocol extension proposed than even 2016 that would have seen explicit sync in the Wayland protocol that was shot down by Nvidia in 2014.

    The reality is stable form of explicit sync without Nvidia getting in the way would have been in Wayland protocol quite early on in Wayland protocol development.

    mSparks the thing with implicit sync and explicit sync you need moderation. Correct mix of both. Too much implicit sync the result is slow. Too much explicit sync the result is unstable. There is a Goldilocks​ valve of how much explicit and implicit sync you should have for good performance and stability.

    Yes if your agree that implicit sync is really good for resource management you should be anti core X11 protocol or you actions don't match your words mSparks.

    Early software development created some insanely dangerous things like MS Dos and core X11 protocol.

    Leave a comment:

Working...
X