MACON, Ga. A week into using OS/2 2.0, I'm still happy with IBM's new operating system.
When IBM introduced Version 2, it was called "a better DOS than DOS, a better Windows than Windows and a better version than Version 1.3." That sounds like hyperbole, but I've found it's pretty close to the mark.
Ignoring the "better version than Version 1.3" part as pretty obvious, let's look into IBM's claims.
First, "a better DOS than DOS."
One of DOS' major stumbling blocks is its 640-kilobyte memory limit. Although PCs can now be had with more memory than Ronald Reagan's entire National Security staff, DOS by itself will only access that first 640K chunk of conventional memory.
The amount of usable memory is further reduced by the space DOS itself takes up, as well as whatever terminate-and-stay-resident programs you may have loaded. With Version 4 of MS-DOS, all that could leave you with something around 400K of memory space, too little to run many serious programs (such as Lotus 1-2-3 Version 3, for example).
DOS 5 helped the memory situation by allowing DOS itself to be loaded "high" (into the memory range between 640K and one megabyte), and memory managers such as Quarterdeck's QEMM freed up more memory, but it required some computer expertise and a fair amount of time tweaking to get more than 600K of free conventional memory.
OS/2 does all this memory management for you. When you start a DOS session in OS/2, you get about 620K of conventional memory. If the application you're running uses expanded or extended memory, OS/2 provides it. If you don't have that much RAM available, OS/2 will use your hard disk as memory. If memory gets swapped to disk, of course, things slow down, but your programs keep running; under DOS, they'd stop and you'd probably have to reboot.
Speaking of rebooting, OS/2 makes sure you have to do that a lot less. If a program should hang, OS/2 will give you an error message telling you, more or less, why the program quit running, then will gracefully remove it from memory. If a program hangs under DOS, nine times out of 10, your only option is to repower or otherwise reboot the whole machine.
OS/2 additionally provides a couple more nifty reliability features, one of which will automatically respond to DOS' "Abort, Retry, Fail" prompt.
Multitasking is another of OS/2's DOS enhancements. DOS can multitask after a fashion using DESQview or Windows, by letting first one program and then another run.
Because OS/2 is a 32-bit system (DOS is an eight-bit system), it's able to take advantage of the 32-bit architecture of 386 and 486 CPUs to create multiple "virtual DOS machines." A VDM is a close, but faster, emulation of the 8086 chip -- as though you were running a PC with multiple CPUs. The VDM methodology also prevents multiple applications from interfering with one another, preventing the Unrecoverable Allocation Errors that bomb Windows.
Another OS/2 bonus is its clipboard, which lets you copy and paste data between the applications you're multitasking.
version of DOS does have some limitations because of the nature of the system itself. Programs that access the computer hardware directly instead of through DOS, mostly utility programs such as the Norton Utilities and PCTools, won't operate properly, because OS/2 is itself controlling the PC hardware.
To get around the limitations of its DOS, OS/2 offers three ways of switching to a real version of DOS, so you can still use these programs.
I'm not the best person in the world to comment on the "better Windows than Windows" claim, as running Windows and getting a root canal are about equal on my list of things to do. (I find Windows aggravatingly slow, and the applications just aren't good enough to put up with the aggravation, compared to their Macintosh counterparts).
To be fair, though, I tried PC Magazine Labs' Windows Benchmark, running first the real Windows and then OS/2's Windows. There was no significant difference in performance between the two on my machine.
But OS/2's version of Windows does eliminate Windows' annoying Unrecoverable Application Errors.