4 ms·

> Buy something else. I have a raspi running pretty much non-stop since 2011 and am still able to update its software 12 years after purchase. All "somethings

by voytec 3y ago

> Buy something else. I have a raspi running pretty much non-stop since 2011 and am still able to update its software 12 years after purchase. All "somethings elses" which I bought had 3-4 years of software support and then their usability diminished drastically.

TeMPOraL 3y ago
Why would the boards of this class need ongoing software support? Aren't you supposed to be able to run normal Linux on them, with few device-specific drivers (which don't need to change, since the hardware doesn't)?
- voytec 3y ago
  Some had very customized kernels which stopped being released. Since getting all the custom drivers to work was too much pita for these devices' vendors, why would I pay them and invest my time in getting their hardware to work? That these devices are useful for hobbyists, doesn't automatically mean that people want to periodically spend hours to have them working with up-to-date software.
- aleph_minus_one 3y ago
  > with few device-specific drivers (which don't need to change, since the hardware doesn't)? The reason is that Linux has no stable in-kernel driver interface, so the drivers often have to change with new kernel release. Feel free to complain on the LKML why you consider this to be a bad idea ...
- 908B64B197 3y ago
  > ongoing software support Part of the problem here is that SoC drivers are hacked on top of the kernel (and are not merged in-tree). So SoC vendors have this pile of legacy and hacked-on code they need to cherry-pick every time the Kernel changes an interface they used (which is a lot) [0]. The code is also, let's be honest here, low quality and full of workarounds for implementation details (some radio-modem talking over I2C to transmit voice to the SoC???) so there's no way it would get merged into the kernel. Of course, that costs and fortune and hampers the sales of newer chipsets so it’s not in the SoC vendor’s interest to do so. [0] https://en.wikipedia.org/wiki/Application_binary_interface https://en.wikipedia.org/wiki/Application_binary_interface
  - aleph_minus_one 3y ago
    > Of course, that costs and fortune and hampers the sales of newer chipsets so it’s not in the SoC vendor’s interest to do so. On the other hand, it would be in a lot of vendors' cost-saving interests to massively lobby for a (much more) stable in-kernel driver interface in the Linux kernel so that the vendors only rarely have to change their Linux drivers when a new kernel version gets released.

by voytec 3y ago

TeMPOraL 3y ago

Why would the boards of this class need ongoing software support? Aren't you supposed to be able to run normal Linux on them, with few device-specific drivers (which don't need to change, since the hardware doesn't)?

voytec 3y ago
Some had very customized kernels which stopped being released. Since getting all the custom drivers to work was too much pita for these devices' vendors, why would I pay them and invest my time in getting their hardware to work? That these devices are useful for hobbyists, doesn't automatically mean that people want to periodically spend hours to have them working with up-to-date software.
aleph_minus_one 3y ago
> with few device-specific drivers (which don't need to change, since the hardware doesn't)? The reason is that Linux has no stable in-kernel driver interface, so the drivers often have to change with new kernel release. Feel free to complain on the LKML why you consider this to be a bad idea ...
908B64B197 3y ago
> ongoing software support Part of the problem here is that SoC drivers are hacked on top of the kernel (and are not merged in-tree). So SoC vendors have this pile of legacy and hacked-on code they need to cherry-pick every time the Kernel changes an interface they used (which is a lot) [0]. The code is also, let's be honest here, low quality and full of workarounds for implementation details (some radio-modem talking over I2C to transmit voice to the SoC???) so there's no way it would get merged into the kernel. Of course, that costs and fortune and hampers the sales of newer chipsets so it’s not in the SoC vendor’s interest to do so. [0] https://en.wikipedia.org/wiki/Application_binary_interface https://en.wikipedia.org/wiki/Application_binary_interface
- aleph_minus_one 3y ago
  > Of course, that costs and fortune and hampers the sales of newer chipsets so it’s not in the SoC vendor’s interest to do so. On the other hand, it would be in a lot of vendors' cost-saving interests to massively lobby for a (much more) stable in-kernel driver interface in the Linux kernel so that the vendors only rarely have to change their Linux drivers when a new kernel version gets released.