Purism deserved some criticism between 2015 and 2017 because its initial crowdfunding promised that that the L15v1 would have a freed BIOS and it took Purism over 2 years to fulfill that promise. However, Purism was the first company to sell new PCs with the Intel Management Engine neutralized, and its example pushed System76, ThinkPenguin and TUXEDO Computers to start doing the same. Purism was the first company to sell new Linux laptops with Coreboot preinstalled which weren’t locked-down Chromebooks. Its example pushed System76 to start doing the same.
If you want an x86 processor built after 2008, Purism is the best company available in terms of free software. At this point, Purism has managed to replace UEFI with Coreboot, get rid of the proprietary VGA BIOS, and neutralize and zero out 90%-92% of the ME’s code. The only proprietary parts that remain are the Firmware Support Package and microcode, which are impossible to free without help from Intel. Purism was the second commercial company (after Libiquity) to create an FSF-endorsed distro, and PureOS is the highest ranked 100% free distro. DistroWatch ranks PureOS at #51 over the last 6 months compared to Trisquel at #72.
As for RYF certification of the L5, Purism is the first company to ever seek RYF for a mobile device and there is no indication that Purism won’t eventually get it. At this point, Purism still has to demonstrate that it can make the two cameras and the smartcard reader function with 100% free software, but the bug reports for the front (selfie) camera, back camera and the smartcard reader don’t indicate that proprietary code will be needed. It makes sense to me that the FSF will hold off on certifying the L5 until those features are ready. Given that there currently isn’t a single laptop, tablet or phone with RYF certification, the fact that the L5 is so close really is a huge accomplishment.
I don’t see a better path for Purism to pursue if the goal is to produce usable devices that respect user’s freedom.
As I see it, the Power ISA architecture is a dead end. IBM has demonstrated that it is only interested in making making high-powered chips for servers that cost an arm and a leg and suck huge amounts of power and NXP stopped designing new Power ISA chips back in 2012. RaptorCS has done the best that it can with IBM’s over-priced POWER9 chips, but I don’t see it as ever being a viable solution for normal people. At some point, powerpc-notebook.org may actually manage to finally start selling a laptop that runs on NXP’s antiquated PowerPC e6500 chip from 2012, but there is zero chance that NXP will ever produce another PowerPC chip, so I don’t see much point in pursuing a dead end that has no future.
As I see it, Allwinner is also a dead end, because of the company’s attitude toward the community. Amlogic doesn’t have any decent chips in the offing. Snapdragon and Exynos do eventually get mainline Linux support, but they will always require proprietary blobs for the WiFi, Bluetooth, cellular modem and GNSS. MediaTek has no interest in collaborating with the community. Rockchip and NXP’s i.MX 8M are the only realistic paths forward in the near future for mobile chips, and nobody has any idea when the ARM “Natty” GPU will be supported by Mesa, so you can’t plan on doing anything with the RK35xx in the future. Basically you are stuck with the RK3399 which can only barely boot with 100% free software or pouring tons of time and labor into making the i.MX 8M Quad work, which is what Purism is doing.
The only other viable strategy is create an FPGA that uses some experimental RISC-V core which is going to have horrible performance for the price or to pray that Alibaba or NXP produces something in the next year or two that you can use. SiFive seems to have no interest in running on 100% free software.
Installing free software in ancient hardware is a waste of time in my opinion. The first problem is that nobody is maintaining the Libreboot code and nobody is working on making Libreboot code work better on the X200, T400, T500 and W500, so it is always going to be a suboptimal solution. In order to be able to pay people to work on the code, you need to charge a significant markup, and none of the half dozen companies that sell old Thinkpads with Libreboot are able to charge enough to pay for programming work. Nobody is going to pay more than $500-$600 for laptops that are 13 years old, so there is no way to pay developers to work on Libreboot.
The second major problem is that buying used hardware from 2008 and installed free software does nothing to actually change the hardware industry. Tiny companies like Purism, OLIMEX or PINE64 that design NEW hardware have some hope of actually getting component manufacturers like NXP, Silicon Labs, Thales, BroadMobi, etc. to listen to them, so their hardware becomes more compatible with free software. Nonetheless, companies like Libiquity, RetroFreedom, TechnoEthical, Vikings, Zerocat and Nitrokey that sell old hardware with free software have zero influence over the component suppliers, so there is no chance of ever improving the situation in the long term.
Given the choices, it seems to me that Purism is already pursuing the best available path. If Alibaba or NXP actually start producing RISC-V chips that Purism can use, Purism should jump on them, but Purism can’t be in the business of doing chip design. Likewise, if Luke Leighton’s libre-soc.org ever produces anything, Purism should make a device based on it. However, Purism doesn’t have the resources to do original CPU design, so it is unrealistic to expect it to do cuttting-edge work with RISC-V.
