>Thinking pie-in-the-sky .... I think that using it to teach elegant OS >programming under resource constraints would be a great niche. It's a >lost art, but one that has many "teachable moments" built in. I think this is a good niche, but it will require that ELKS diversify beyond the 8086. I see the low-end 32-bit Intel platforms as a good target to begin with; while one can build a nice small Linux for use on a 486, Pentium, or even a VIA C3 or Transmeta Crusoe, such systems are grossly underpowered and somewhat memory-starved for modern Linux, glibc, X.org, gcc, and so on. These systems are still relatively easy to run into and are regularly discarded. I have no working 16-bit PC-compatible hardware, but I have a closet slam full of 486 through P3 desktops and laptops. It would be fantastic for ELKS to run on these platforms. I don't see the 8086/286 as a viable target anymore. To be a "teachable" platform, people need to be able to use it without scraping up rare vintage hardware. Linux abandoned 386 support somewhat recently due to the nasty batch of workarounds required to keep it working, and I think ELKS must follow suit in its own way. I would like to have a kernel that can build and run on everything from 8-bit to 32-bit, with or without an MMU, and taking advantage of platform specific features where possible. Unfortunately, once again the problem comes back to the compiler; 8-bit processors tend to be very poor compiler targets with limited ability for position-independent code and often require bank switching to reach extra memory, 16-bit processors tend to have ugly segmented memory models, and finding a good compiler that targets even a fraction of them is no easy task; never mind generating quality code! So we come back to the original problem: we either need a compiler or a change of target... -- To unsubscribe from this list: send the line "unsubscribe linux-8086" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
