"Ok so RO4 Select is more hardware independent. Then it should be a doddle to port to the Iyonix shouldn't it... Also the OSes being "very different internally" is irrelevant"
From what I understand of how modern ROS 4 works: It will be easier now to port RISC OS Select to the Iyonix, since the 32bit work is being wrapped up. The stumbling point is the actual hardware drivers. A hardware absraction layer doesn't take away the need for drivers, it merely allows the upper level of the OS to free itself of worrying about what the underlying hardware is doing. Picture it as.. OS <--> HAL <---> drivers <---> hardware.
So, for an Iyonix port, ROL will have to produce an Nvidia driver, USB driver, etc that will work with the Iyonix hardware and the OS 4 hardware abstraction. Ideally, RISC OS 4 could reuse the OS 5 driver modules. Except ROL don't know how the OS 5 modules communicate internally and how they collude to pass around information. For example, the OS 5 kernel may expect a particular line of communications to and from the graphics driver (such as a Vsync ticker) and won't work without such a heartbeat.
ROL have no way of knowing how the CTL Nvidia or USB modules work with other OS-level components. That's what ROL mean when they talk about differences internally, and it's also what ROL mean when they called for more information and help from CTL for the Iyonix port - the internal design of OS 5 is not documented publically. As you say, the internal workings of the OS will not affect third party programmers, but it's crucial for the OS-level developers.