Re: Where does ELKS need to go?

[Harley Laue <losinggeneration@gmail.com>]

The TL;DR version of my reply below: the project likely needs to 
diversify into other platforms.

On 04/18/14 15:03, Jody Bruchon wrote:
> On 4/18/2014 3:26 PM, Edoardo Liverani wrote:
>> I'd really like to help this project as I'm going to use it a little,
>> and I would enjoy to help hosting a website with updated links, to
>> generate updated and working precompiled images, istructions etc.
>> I have a shared hosting plan I can use, whom should I ask to get
>> current website sources and the permission to update a little and
>> re-publish them?
>> My question is particularly for @Jody cause he is the latest who
>> managed to maintain the project sources.
>
> I am still here to maintain the project. Unfortunately, I've run into 
> a couple of major issues with it that make its future questionable.
>
> * I'll happily move and clean up the website to my own hosting if 
> there is renewed interest in the project. I'll overhaul it while I'm 
> at it.
>
> * The compiler we use, bcc, suffers from some serious limitations and 
> misbehavior. In particular, nothing can require more than 64K of code, 
> including the kernel. The compiler needs some work or we need a new 
> compiler. As I would like to see ELKS target other classic CPUs 
> (65816, 68000, maybe 6809, etc.) a compiler change may be the best 
> option. The compiler is the biggest obstacle.
I'm pretty sure it's been mentioned in the past, but SDCC might be an 
option (though, I don't think there is a code generator for 8086-80286). 
Edoardo's suggestion of TACK might be a good fit as well. I've had some 
experience with SDCC, but I haven't ever tried TACK. I will say, SDCC 
has a fairly dedicated community behind it and the project is still 
quite active. TACK, however, doesn't seem to have had many commits in 
the last few years.

In any case, it'd be a good idea to avoid compiler specific pieces of 
code as much as possible. Otherwise, it could be a BCC situation all 
over again (being stuck with a more-or-less dead compiler.) Obviously, 
for something like a kernel, it's nearly impossible, but most of that 
could be provided via headers and/or pulled in conditionally.

> * What can/does ELKS offer compared to other small OSes such as NuttX?
Running more than one process (I'm assuming at least.) IIRC, NuttX 
compiles the kernel into the application to run the single application 
on the hardware.

> * The hardware ELKS is made to work on is (to my limited knowledge) 
> becoming rare. Quite a few 8086/88 machines have by now suffered 
> capacitor failures that have rendered them inoperable and probably 
> junked. Most PCs going in the garbage now are Pentium II/III/4 
> systems, all of which enjoy Linux compatibility and are far more 
> capable under Linux than under ELKS, even if they only have 16MB of RAM.
>
> * The "E" in ELKS means "embedded" and yet the only platform it was 
> ever developed for was 8086/88 PCs and the Psion SIBO. The 808x target 
> made more sense 10 years ago, but embedded and low-power computers 
> today are dominated by 32-bit ARM and MIPS cores that happily run 
> Linux (if they have enough RAM, that is.) What should ELKS be 
> targeting today?
Might be kind of cool to see it running on other 8 & 16 bit processors. 
Not that it would be terribly practical (is anyone looking at this 
project /really/ that concerned with that though?), but an 8-bit AVR or 
16-bit MSP430 might be interesting target(s). In this case, it would be 
good to really try to appeal to the hobbiest and build a mini computer 
with the chip. Most all of these will need more RAM, USB (might be 
tricky), and a method to display video output. Audio would be a plus as 
well. This would obviously be a large to very large undertaking (porting 
the codebase alone would be challenging), but I think it /could/ really 
spark some interest in the project again.

> * The project has no active real-hardware testers to call upon. No one 
> has real hardware AND time for the project AND wants to test changes. 
> I personally have no 8086/80286 hardware but have a plethora of 
> functioning Compaq 486 and Toshiba Pentium laptops, all of which have 
> Linux on them. Without real hardware and a skilled, willing owner that 
> can test ELKS on it, there can be no proper development. I can use 
> emulators but they don't emulate the many various quirks and "just 
> non-standard enough to piss you off" hardware of the early IBM PC era 
> (I'm thinking about you, Tandy.)
I /really/ need to get my Tandy 1000TX (a 286 with a 20M harddrive) up 
and going. The power button, AFAIK, the only thing that doesn't work (it 
wore out and doesn't click into place anymore.) It's one of those easily 
fixable things that I just haven't bothered to do. It's also one of 
those, too many projects, not enough time.

> * I have a TRS-80 CoCo and an Apple IIgs. Maybe we should port ELKS to 
> those. ;-)
>
> I would like to hear what anyone reading thinks. Please reply either 
> to me OR the ELKS list since I (obviously) subscribe to it and prefer 
> not to receive duplicate messages.
>
> -Jody
> -- 
> To unsubscribe from this list: send the line "unsubscribe linux-8086" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html