I agree. We know IBM will not open source OS/2. We should support the effort to clone OS/2; Voyager, OSFree or a port of K42.

I believe Voyager is the way to go. It can be developed on top of OS/2, but can be easily ported to another plataforms (this is one top priority objective of the project, at least). The objective is to isolate the system look and feel from the base operating system, in a way the user cannot even know what kernel is being used in daily operations.
This means you could replace top-OS/2 subsystems with a new one that keeps the OS/2 feeling (Voyager) and then use it as a easy-pass to another plataform, keeping the best thing of OS/2 (usability) on a modern kernel with better device driver support from vendors. Note that this doesn't mean this kernel will be Linux one. It can be even Windows' (ReactOS?), Haiku's or MacOS' (MACH?).
I spent several hours in the last few days reading every piece of document produced by the Voyager project and this small post presents why I believe this is the way to go. It is like recreating the idea of OS/2 for PowerPC's "personalities" ... And with some effort, these ideas can be applied to any kernel, not only uK.
It *is* important to notice that Voyager is really a *good* concept. If Voyager is implemented on top of Linux kernel, for example, the idea is you'll use Linux kernel but your feeling will be the same as using OS/2 Kernel. This seems to be impossible? Well, spend a time playing with MacOS X and tell me there is a *Un*x inside it. The Voyager experience *shall not* be dependent of the kernel being used. Even using Linux or ReactOS kernel, you'll have a WPS, a CONFIG.SYS (if this is the best way to go... /etc is, for sure, the worst way, though) and your prompt will be like OS/2 CMD.EXE, not that bizarre shells UNIX's users are used to. The way you feel the system will be the same of OS/2. It is the "everything is an object" approach, not the bizarre "everything is a file" approach.
In fact, nothing will prevent that someone creates a clone of OS/2 for PowerPC Kernel... and this could be used as Voyager kernel (the kernel part of the project is not defined!). OS/2 for PowerPC would be, for sure, a lot better than OS/2 for Intel kernel, which is very outdated in several aspects. But the problem remains the same: which device driver model to use in this kernel? Not OS/2's model, for sure: nobody wants to write hacked device drivers forever. The best would be Windows model, since Linux model is somewhat limited. But even if every hardware vendor builds Windows' drivers, this model is a moving target, since Microsoft mess with it at each Windows release.
This is the big question that needs an answer, but the point is: it has not to be answered now! Since we do not have a working and portable WPS, there is no reason to define what device driver model will be used on the kernel. Since this is a very complex question (and the answer may change until a working WPS clone is ready), Voyager decided to start from the top layer, the one which is the most important for the users: the WPS. Until we have a working WPS clone, there is a lot of time to discuss kernel issues, models and so on.
What was the error of the past (osFree)? Discuss moving targets in the beginning of the project (Kernel, Device Drivers etc) and forgetting about the "fixed targets", like WPS.
OS/2 Kernel for Intel will not survive forever without *really big* updates. And even if I wish IBM could release the source code for some parts of OS/2, I really doubt it. And this is even more difficult in the case of the Kernel (even OS/2 for PowerPC's one, I believe. Most parts of OS/2 for PowerPC were simply *ported* from Intel, AFAIK... All documents I read said that IBM have "enhanced" MACH kernel, and I believe this was done with patented OS/2 code). Even if Voyager project do not succeed in producing a new operating system with "OS/2 feeling", it can succeed in producing a working, portable WPS... in that case, it would be at least less painful to migrate to another OS, something we all will need to do some time from now.
The point is: we do not need "yet another operating system". We need to keep the "OS/2 way of use a computer" built on modern technologies and able to run on modern hardware. And this is "less difficult" to achieve than building a new operating system from scratch (which is a far more complex task than we are capable to accomplish with the man-power available into this community ).
Well, these are my thoughts.
My kindest regards!
PS: OpenBeOS clone could not use BeOS original code, *but* I was informed that some of OBOS developers had been able to peek in the original source code, besides the fact the inner works of BeOS are a lot more well documented that OS/2 internals (or so I was told). These are very important facts to take into account when comparing the speed of development of a new BeOS versus the development of a new OS/2.