IBM

oh hamei! could it be true?!

http://searchdatacenter.techtarget.com/news/article/0,289142,sid80_gci1508584,00.html

My two complaints of OS/2 nowadays are

(a) it's not a multi-user operating system
(b) it's shared library mapping is primitive, LIBPATH is a global variable, not a user variable

other than that it's all good. Shame it got crippled with Win-os2 clunky mapping support etc.

I wonder if they'll make it multi-platform this time, perhaps providing support for some of their own POWER/PPC hardware? (Preferably the workstations, even though most seem at their EOL by now).

It previously ran on PowerPC! I don't know if it was in big or little endian mode.

I didn't know.

(c) If your app thrashes the stack, it crashes the system, not the app.

I did s/w development back in the OS/2 days. We were so happy when Windows NT appeared. Windows NT (3.5.x) was slower, but more robust.

FWIW, it seems this redux thing is more about creating some sort of OS/2 personality / desktop on top of Linux.

Surely you can do WorkPlaceShell in Java?

IBM did have SOMobjects running on AIX and OS/400, and they originally had the intention to be able to sit PresentationManager on the top of X11.

But these days you'd want to do CORBA with Java to get all the IIOP over SSL plumbing for free.

Not gonna happen, OS/2 is 10+ years old. That would be like HP bringing back Tru64. Also, didn't eComStation buy the rights to OS/2 anyway?

Don't forget the REXX interpreter. They even had a Linux/i386 version at some point, under the name `Object REXX'. I wonder how much REXX code is still in use those days...

Not Tru64, Ultrix!

Yes, I just tried that email and it got bounced.

Could it be done..? Sure. Will it be done? Highly unlikely.

Just look at this:
http://www.ibm.com/developerworks/syste ... OS2toLinux

Although there is nothing stopping someone from building something like DOSCALL1.DLL, KBDCALLS.DLL, QUECALLS.DLL, SESMGR.DLL, OS2CHAR.DLL in POSIX giving some kind of layer to act like OS/2 ....

But then we run into the age old thing about there being no big apps...

And frankly Virtual PC & Qemu run OS/2 pretty darn good, why on earth would someone go through all that hell to port something to 'kind of work' when you can just emulate the whole thing?

That being said, I just scored a copy of Excel 3.0 for OS/2... It's neat, but it's not like MS is going to release the source anytime soon.... (but it was worth the $18 on ebay!)

I recently aquired the Borland C++ compiler for 32bit OS/2, it's great to have a true IDE on WARP.

Did you get the Open Watcom stuff ? I might have Visual Age around somewhere ....

Yes, I have OpenWatcom. I would really like to find a copy of CSet2/CSet++2.

Its probably not that much work, considering the resources IBM has, and if there really are customers actually wanting this type of thing.

An OS/2 personality for text only apps on NT was developed by MS/IBM using the native NT API and included in the initial NT releases. And IBM later developed a full OS/2 personality (including DOS emulation) to run above the IBM Workplace OS (i.e. a Mach variant) - this was released as the aforementioned PowerPC port. Creating an OS/2 personality above the more functional POSIX API would arguably be easier than either of these two previous microkernel style APIs they have targeted.

Running OS/2 apps on "proper" OS/2 in an x86 virtual machine still leaves you at the mercy of a 10 year old OS. Compared with OS/2 apps running natively as processes, it is more resource intensive, and it hinders you from taking full advantage of modern hardware features such as threads etc (it all depends on how the virtual machine is designed though). Also, apps are only as reliable as the underlying OS, and nowadays, some might be more comfortable with that being Linux on raw hardware, rather than the OS/2 kernel running on emulated hardware above another real kernel on real hardware.

I'm sure there are some out there somewhere.... Much like you can bet there are some mainframe people not ready to give up their SNA networks either...



Lots of people forget that Windows NT started out as NT OS/2. And they kept the 1.x personality, but threw away the "cruiser" 32bit personality and took that WILO port of Windows 3.0 to OS/2 as a starting point of porting Windows to OS/2... It's in some old book "showstopper".

I guess the question is, would you implement it on top of LIBC, or tie it to the kernel with syscalls to make it more 'native' to the kernel?

Without having the source to either, who knows... I'm just glad it's not my job.



Which is true, as long as you never programmed around that quirky OS... Just look at all the fun the Wine people have had, to the point they have to be bug for bug compatible.

Sometimes I'm surprised nobody has tried.. not everyone wants to run Unix, but ReactOS is still coming along in Alpha stages, but it's too fragile for day to day usage...

I guess the real thing is that unless you are XYZ corp with this internal app (which will never see the light of day for external testing) what great OS/2 programs are there to test with? Sure I've got Excel 3.0 for OS/2, but it's the 16bit stuff.... And OS/2 1.3 runs great in Virtual PC!

The structured exception handling would be "fun" to provide, very similar and a fore runner of Win32 SEH. Also OS/2 has a unique solution to thread local variables, they all share the same address which is switched to different real memory based on current thread.

Yeah, perhaps some kernel module trickery would be needed those; especially for that thread scheme - its a strange way to do things, as designing it like that means more potential TLB/cache work when switching between threads in the same address space, which kind of defeats the main benefit of threads.