Win32 is growing like a weed, sprouting new extensions as Microsoft takes Windows places it has never gone before
Jon Udell
The millions that IBM sank into OS/2 bought a pearl of great value. Big Blue now owns an operating system that uniquely combines five key virtues: OS/2 is small, fast, modern, 80x86-tuned, and mature. Do you doubt the strategic value of such a thing? Microsoft doesn't. RISC machines running portable operating systems may be the future, but you can't wish away millions of 80x86 PCs.
Hence Chicago, which Microsoft claims will be small (4 MB), fast (Windows 3.1 or higher), modern (equipped with threads and preemptive multitasking), and 80x86-tuned. The ultimate OS/2 killer? Maybe not. What Chicago won't be, at least in 1995, is mature. Major parts of Chic
ago--including its file-system manager, memory manager, and scheduler--are, although inspired by Windows NT, otherwise brand new.
The comparable pieces of OS/2 have been solidly in place for years. While I believe that Chicago will eventually prove itself a great foundation for Win32 applications, I think OS/2 could be one, too. There's room in the world for both--and more as well. If Win32 dominates, as I think it will, more niches for compatible substrates will exist than any single company can occupy.
Sadly, OS/2 has a tragic flaw. It wears the ``wrong'' API. Presentation Manager, available since 1988, has been dead in the water since Windows 3.0 took off in 1990. That's a hard truth, especially since PM's imaging model was unsurpassed in the Windows realm until NT appeared, and it's a truth that continues to cloud OS/2's future.
I think IBM should swallow its pride, ditch PM, and license the Win32 API. OS/2 guru Michael Kogan, on the other hand, thinks Microsoft should swallow its pri
de and simply use the OS/2 kernel. Either way, we would not now be waiting for Microsoft to reinvent OS/2. Most Windows programmers would know how to build applications using thread synchronization, preemptive multitasking, and sparse virtual memory. Some of us would be using OS/2 to run those applications natively on mid- to high-end 80x86 PCs. Others would be using NT to run them on high-end RISC or 80x86 systems. Windows-on-DOS, minus some advanced APIs, could continue to serve the low-end 80x86 population.
In the real world, of course, the great divorce of 1991 did occur. Microsoft went on to build Chicago. IBM stuck with PM, failed to deliver credible Windows-to-PM conversion tools, and then embarked on a strategy to host Windows itself, rather than Windows applications, on OS/2. This bold hack has served more people far better than most outside IBM thought it could. It will grow bolder yet if, as promised, OS/2 gains mastery of the Windows VxD (virtual device driver) and, through it, Win32s.
But patching Windows in memory, as OS/2 for Windows does, won't take IBM far in the era of Chicago. Neither will a PowerPC-based Workplace operating system running an OS/2 personality that has few mainstream applications.
To survive as an operating-system vendor, IBM will need a credible Win32 strategy, both for OS/2 and Workplace. What to do? It has just two choices--convert Windows applications and components or assimilate them. Both strategies merit a second look.
Conversion tools like Microsoft's Windows Libraries for OS/2 and Micrografx's Mirrors and Oasis never amounted to much, but cross-platform toolkits for C, C++, and Smalltalk have come on like gangbusters in recent years. IBM's Visual Age, which is a Smalltalk-based development tool for OS/2 and (soon) Windows, exemplifies this new breed of product. Thanks to IBM's groundbreaking SOM (System Object Model), Visual Age will help build distributed systems made of reusable components. This class of tool could level the playing field a
s the first mainframe era and the twentieth century draw to a simultaneous close.
IBM doesn't need to fight in the PIM (personal information manager) and spreadsheet skirmishes. It needs to define a new era, the one in which PCs control the back room as well as the desktop. This is the company whose systems run banks and airlines. Its mission, often poorly articulated, is to move those systems onto modern hardware and software platforms and, while doing so, to improve and extend them. To play its intended role in that mission, OS/2 needs tools like SOM and Visual Age. But neither tool would be likely to flourish if it did not also support the volume platform, Windows.
The same holds true for OS/2 itself, and that's why cross-platform tools, while necessary, are not sufficient. IBM should bite the bullet and license Win32. Sure, that will cost something, but if Microsoft offers IBM terms less favorable than those offered to Insignia, the FTC and Justice Department might reopen their inquiries.
Win32 is growing like a weed, sprouting new extensions as Microsoft takes Windows places it's never gone before. Consider telephony. IBM's CallPath blazed the trail that TAPI (the Intel/Microsoft telephony API) and TSAPI (the AT&T/Novell telephony services API) follow. But today, IBM doesn't get to define these or other vital infrastructures. It should, given its experience with big, complex systems. And it could, with a Win32-equipped OS/2. An IBM/Microsoft detente seems farfetched, but so did the PowerPC alliance. I say go for it, IBM.
Jon Udell is a BYTE senior technical editor at large. You can contact him on the Internet or BIX at
judell@bix.com
.
Why IBM Should License Win32
