On Friday 2022-05-20 18:16, Kent Overstreet wrote:

Problems with ioctls:

* There's no namespacing, and ioctl numbers clash

IOCTL v2 proposal:

* Namespacing

To solve the namespacing issue, I want to steal an approach I've seen from
OpenGL, where extensions are namespaced: an extension will be referenced by name
where the name is vendor.foo, and when an extension becomes standard it gets a
new name (arb.foo instead of nvidia.foo, I think? it's been awhile).
To do this we'll need to define ioctls by name, not by hardcoded number, [...]

https://www.khronos.org/registry/vulkan/specs/1.3-extensions/man/html/VK_NV_glsl_shader.html
Right there on the front matter: "Registered number #13".

https://www.khronos.org/registry/vulkan/specs/1.3/styleguide.html
"All extensions must be registered with Khronos."
"Each extension is assigned a range of values that can be used to create globally-unique enum
values"