Re: early days of ACPI
It's not just Samsung, and it's not just Linux. Hardware vendors write a lot of stuff that doesn't quite meet published spec, but when released with a current popular OS works.
Back in the early Pentium days, before Linus had even started work on Linux there was a company that released a multiple CD drive that ran on the IDE channel. Beautiful part about was, under Windows 95, you could share each of the drives in the Caddy, and computers linked on a work group could read what was in the CD caddy. We had a client running some library software in a Windows workgroup configuration (connections to remote PCs were pretty much by modem in those days) and they depended upon this functionality. One day they had problem with an internal modem that fried the motherboard. CPU was good, so we tried to replace the MD and memory. Win95 would install just fine, find the device and initially you would see all the drives. Then you installed the Intel chipset for the MB to make all the devices accessible and all of a sudden the device stopped working. We spent hours trying to figure it out before calling the manufacturer. The driver didn't work with the new Intel chipset, and the company had no intention of updating the driver. They had even discontinued making the device, so we couldn't even just buy a replacement. Instead we had to install a SCSI card and a SCSI caddy, about tripling the bill to the customer (we discounted the labor since we should have called the vendor sooner than we did).
Years later I was in a different environment where they'd been using an IDE "tape drive" for ages to backup data from local PCs (don't ask what they were backing up, I thought it was epic fail at the time), but still in a 95 environment. But drive sizes were finally getting large enough to force us to install 98 on some systems, and the tape drives started failing on the 98 systems. Because apparently the manufacturer didn't fully meet the IDE spec for the drive.
Now, I recognize that sometimes specs come out with things that people don't really have a way to test at the time the spec was written, and that there are monetary limits on what a company can reasonably be expected to test. But cases like these always leave me scratching my head.