With the AToF solution, I think I no longer need (or perhaps even care about) a 64-bit version of OS/2 ...
If you'll recall, AToF provides for a 64-bit Host OS (linux or windows), a virtualization program (Virtualbox), and OS/2 or ArcaOS is running in a VM. 64-bit apps are provided to OS/2 in a browser (Firefox) via Winflector; or, the entire AToF platform allows windowed access on the same PC to other os's (HostOS or other OS VM's) running 64-bit apps. OS/2 stays at 32-bits, but none of the current crop of hardware/software problems are there under AToF. More AToF details at:
os2world.com/forum/index.php/topic,3498.0.html
It remains fun to think about a 64-bit OS/2, though ...
Of course, getting to a 64-bit OS/2 "the old way" (writing it from the ground up) is near impossible, because of IBM "licensing issues" & such, where the source code for OS/2 itself most likely never gets opened up. Are there any alternatives, short of clean-slate rewrites of each piece of "existing" OS/2? With no money, this clean-slate approach also remains unavailable.
But, is a new way opening up for us? A thought exercise is that a "microkernel" and "workplace OS" (a new form of OS/2, kinda/sorta already done, but later abandoned by IBM) approach would get us there, and some of the pieces are getting tantalizingly closer ...
The microkernel part is complete, and it appears we can bring up a 64-bit microkernel (seL4) on pc hardware (and other platforms). The IBM Workplace OS microkernel (mach) is replaced by an *existing* seL4 microkernel (NOVA, based on seL4). OS/2 was up and running on the Mach microkernel. We can now swap out Mach for NOVA (seL4). To keep up with the interesting work in this area, look at Genode Lab's efforts on:
- the Genode OS Framework
- the NOVA microkernel (seL4)
- Sculpt OS (which we would want to replace with Workplace OS pieces)
The OS/2 part would be the other elements of IBM's Workplace OS, which hopefully doesn't have the same license restrictions, as IBM & partners were going a clean-slate route for development (hopefully, no Microsoft code remains in there). Instead of being layered on Mach, they are layered on NOVA.
I can't pursue this any further (even on paper) until I can get the original Workplace OS up and running in some form of virtualization or emulation ... most likely, it will be with QEMU. I believe I've found all the Workplace OS bits and pieces, but don't think I'll be ready to show something at this year's Warpstock. It looks like I can get there, though. Virtualization or emulation would enable more poking around on how Workplace OS did it's thing, which in turn would enable something more like a roadmap:
1. done (Genode solution): get a microkernel up and running on PC hardware
2. wip: get Workplace OS up and running under emulation ... enable more investigation of OS/2 in this environment
- I could find that I still don't have all the pieces, but, some progress
- don't know how to ask IBM for missing pieces; ignoring licensing, for the moment, unless it actually starts working.
3. determine how to rewrite the OS/2 pieces on top of an existing microkernel
- will IBM open up the Workplace OS parts? That would jumpstart things ...
- if they won't, can these piece be clean-slated?
Again, I'm fairly happy running OS/2 as a 32-bit OS in an AToF solution, because AToF takes care of all my 64-bit needs, on one PC. I also no longer need 16-bit DOS or 16/32-bit Windows, each under 32-bit OS/2, because they are available under AToF, and run with no problems in other VM's.
AToF does enable virtualization and emulation, so it's actually quite fun to attempt going down the Workplace OS rabbit hole, and to see how far I can actually get. It won't cost me anything but some emulation time & effort ... if successful, that would get item 2 on the above roadmap done.
Your thoughts?
