FUSE?

[Jipok]

Will it be possible to implement fuse in the kernel in the future?
And does it make sense at all, i.e. will it be more efficient than userspace implementation?

[Jipok]

I understand that I was not very precise in terminology. Because fuse is about userspace. But I think it is clear that I want to create my own file system

[lneto]

hi @Jipok, it's actually on my long-term TODO list ;-).. we would need to create bindings for that.. if you are interested in implementing this, I could definitely help.. a good starting-point, IMHO, would be to study the in-kernel FUSE support and figure it out what kernel APIs we would need to provide to the scripts..

will it be more efficient than userspace implementation?

we would need to measure.. and it would be a good academic experiment, IMHO.. my gutting feeling is that we indeed might be more efficient for some use cases

[Jipok]

if you are interested in implementing this

Oh no. I never want to go back to C (or C++) again. If I do anything for the kernel, it will be on zig.

That was a theoretical question. I'm actually more interested in whether an overlayfs-like fuse might benefit from moving into the kernel. Your project could easily allow me to quickly try out the concept instead of suffering through kernel module development. Maybe I'll wait until you implement fuse and access to the file system.

[lneto]

Oh no. I never want to go back to C (or C++) again. If I do anything for the kernel, it will be on zig.

well.. it should be mostly boiler plate code and it will set you free from C ;-)

Maybe I'll wait until you implement fuse and access to the file system.

it might take a while then.. heh =)