[Sugar-devel] Compatibility report on latest Debian-based sugar live build - Xorg fails to load due to missing firmware

[Alex Perez]

Answers inline

James Cameron wrote on 1/17/2021 2:16 PM:
> On Sun, Jan 17, 2021 at 02:03:10PM -0800, Alex Perez wrote:
>> James,
>>
>> I booted up the latest Sugar-live-build image, which I'd downloaded from
>> http://people.sugarlabs.org/~quozl/sugar-live-build/ and written to a USB
>> stick, and booted it up in an HP-branded terminal from ~2012, which works
>> perfectly fine with the latest Fedora SoaS images.
> Thanks for testing.
>
>> Simply bundling the 'firmware-amd-graphics' package from the
>> firmware-nonfree repo when you build the Live image would mean the image
>> would work correctly on a vastly larger amount of hardware, out of the box.
>>
>> I would encourage you to take it one step further, and bundle the
>> firmware-linux-nonfree metapackage, which will include firmware for things
>> like Marvell wireless cards, Intel wireless cards, Atheros wireless cards
>> (both USB and integrated/PCI/PCIe)
> How will Sugar Labs comply with the licenses of these firmwares?
>
I'm afraid I don't understand what the concern is here. "Sugar" isn't 
subject to anything different from a licensing perspective, and 
therefore under no obligation to "comply" with anything:

https://opensource.stackexchange.com/questions/8849/does-distributing-gpl-software-along-with-binary-image-force-the-binary-image-to

All of the firmware images packaged by Debian in the non-free repo is 
freely redistributable, but not open-source.

Fedora packages them, and includes them by default. Their LiveCDs/images 
work out of the box with them. Debian packages them, but does not 
install them by default, presumably out of ideological reticence.

Since the goal of the Debian Sugar LiveCD should be to work, 
transparently, on as many computers out-of-the-box as is possible, this 
would seemingly be an obvious improvement. It's not possible to install 
from this LiveCD on a ton of "modern" hardware (the machine I'm using is 
from 2011) with the current state of bundled packages. If the goal is to 
only allow it to function fully on machines which are incapable of 
functioning fully without binary firmware blobs, I'd argue that this 
should be disclosed during the installation process.