Yes, I am not sure why someone would roll their own protocol when ZMODEM was available and solved most of XMODEM's problems. I was forced to use XMODEM back in the day (because ZMODEM did not exist yet), but as soon as the ZMODEM spec was published in 1988, everyone moved to it and XMODEM became obsolete almost instantly.
If you're talking about this blog in particular, then my read between the lines was that 1) XMODEM is much easier to implement from scratch, and 2) the Linux Community recommends an XMODEM implementation for certain tasks, despite the fact that the implementation comes from a package that also supports ZMODEM. Lacking context for why that may be, that seems like a pretty dumb position. I am curious if anyone can dig up a defense of that rationale.
"First boot" protocols are often the simplest possible - and it may be the case that XMODEM is slightly "more available" in that realm than ZMODEM (often these boot protocols simply ignore checksums, etc on the client device to reduce code size needed).
I have modern (produced in the last couple of years) network switches at work which only support XMODEM (assuming you have to do it over serial, they also support TFTP or transfer from a USB storage device, but in the one occasion we had to use it the OS had trashed itself comprehensively enough that serial was our only option).
