Linux Looks Toward Dropping Very Old WiFi Drivers

[jeisom]

What about a new 'unstaging' area where things being retired go?

I like this idea. Put all the depreciated drivers/features there, preferably add to the info area why(lack of known users/unmaintained) and when it is likely to be removed. If it is lack of a maintainer and someone were wanting to find some way to contribute they could start in the unstaging area. And honestly they should have a policy on depreciating and how long till something is removed. ie, 1 year or 2 years, whatever and right after an lts release.

[rmfx]

Linux Hardware drivers should have included metadata with info such a release date of the targeted hardware.
At build time, we could disable some drivers per time period leading to some good optimization.

Otherwise, good choice to remove that. If you want to use a prehistoric machine, then don’t target the latest software.

[robimarko]

I like this idea. Put all the depreciated drivers/features there, preferably add to the info area why(lack of known users/unmaintained) and when it is likely to be removed. If it is lack of a maintainer and someone were wanting to find some way to contribute they could start in the unstaging area. And honestly they should have a policy on depreciating and how long till something is removed. ie, 1 year or 2 years, whatever and right after an lts release.

This won't work in practice as in many cases these drivers are the sole users of various deprecated APIs that get dropped as well with the driver.

[Gabbb]

I think this is not great. Using Linux on ancient machines is very good for the whole ecosystem. Maybe there should be a new kernel type that has all the unmaintained crap included for these special use cases. Reddit is full of people dusting off their P3 machines & then installing Linux on them, there is no reason why their whatever obscure sound card or S3 GPU shouldn't be supported, just because someone decided to drop unmaintained code.

[robimarko]

I think this is not great. Using Linux on ancient machines is very good for the whole ecosystem. Maybe there should be a new kernel type that has all the unmaintained crap included for these special use cases. Reddit is full of people dusting off their P3 machines & then installing Linux on them, there is no reason why their whatever obscure sound card or S3 GPU shouldn't be supported, just because someone decided to drop unmaintained code.

That is exactly the reason why it should be dropped.
If nobody can be bothered to take care of the driver, why should rest of the kernel keep API-s that have been deprecated for years just so driver XYZ works?

[billyswong]

I think this is not great. Using Linux on ancient machines is very good for the whole ecosystem. Maybe there should be a new kernel type that has all the unmaintained crap included for these special use cases. Reddit is full of people dusting off their P3 machines & then installing Linux on them, there is no reason why their whatever obscure sound card or S3 GPU shouldn't be supported, just because someone decided to drop unmaintained code.

A easier solution might be someone creating a distribution that let one stay with older kernels but update most other software timely.

[LightBit]

Both are modern and still maintained drivers.

It says:

some Broadcom B43xx

I don't know what are those.

[jeisom]

This won't work in practice as in many cases these drivers are the sole users of various deprecated APIs that get dropped as well with the driver.

I am not sure which part you are referring to, but there can be exceptions to general rules. If the reason a driver is being depreciated is listed and it is easy to find these drivers there is a higher chance of someone finding interest on updating a driver. Obviously not every driver will be saved or even should be. But some drivers are dropped just because of personpower and if someone wanted to start contributing an easy way would be getting a unmaintained, broken driver or api depreciated driver up to snuf. That would be a lot easier than writing a driver from scratch and would give useful kernel development experience.

[jeisom]

A easier solution might be someone creating a distribution that let one stay with older kernels but update most other software timely.

A kernel fork focusing on hardware over 10 or 15 years old would probably be the simplest approach and attempting to keep relevant userspace compatibility that mainline is using when practical and possible. There is a lot of hardware supported by the kernel that is more of a "it is cool that it works in latest kernel" , but a lot of times it isn't really practical for actual everyday use.

[avis]

Linux Hardware drivers should have included metadata with info such a release date of the targeted hardware.
At build time, we could disable some drivers per time period leading to some good optimization.

Otherwise, good choice to remove that. If you want to use a prehistoric machine, then don’t target the latest software.

Your average PC now has running drivers for HW which was first released decades ago.

Check whether you have the following modules loaded:

• ac97_bus (AC97 first released around ... 1997)
• bluetooth (exists for over two decades now)
• i2c_piix4 (at least three decades old)
• nvme (exists for over a decade)
• snd_hda_codec (almost two decades old)
• ata_generic (almost three decades old)

I continue to marvel at people opining everywhere often when they have zero expertise. Why? Why do you even care? And if you cared, why didn't you ... compile the kernel yourself which I've been doing for over 25 years now?