Hi,
I totally understand the workload to maintain a rolling distro is huge, especially for less common hardware.
Thus there is some ideas to reduce the workload:
- Drop all 32bit arch
AKA, no armv6/v7 support.
This may sound too aggressive, but trust me, those SoCs are really too slow for most modern works.
Just imagine try compile a distro kernel using RPI2.
And with less archs to support, we can focus on perfecting aarch64 support.
- Require the boards to have a dedicated boot flash (normally a SPI flash), except RPI
This is to reduce the requirement on compiling U-boot.
A lot of RK3399 boards are already having such SPI flash, like RockPro64, while newer
RK3588 boards all have SPI boot flash (some even has a full working EFI env).
And this would align us to x86_64, and let U-boot to be part of firmware.
This not only makes new boards support easier (only need to bother kernel), and in fact it's
going to be the only way.
For cases the vendor U-boot is not good enough, we can compile our own and provide it as
precompiled image, with docs on how to flash them.
And more and more SoCs has closed source ATF, making it much harder to package a working
U-boot.
The exception for RPI is for its popularity, but I really don't believe the boot sequence is RPI is
any good.
Considering ArchlinuxARM is really the core of a lot of downstream distos (Manjaro ARM, Void Linux ARM), and
a lot of people may even have their day job depending on an ARM machine (3 aarch64 hosts running VMs for my daily fs tests),
I really hope ArchlinuxARM to be better and better.