OS/2 is Alive and Well and Coming to a Windows Near You! by John D. Ruley, Associate Editor "Contrary to what you may have heard yesterday, I am not dead, and I have not been dropped or scrapped. Neither has OS/2!" -- Steve Ballmar, Senior VP of Systems Software at Microsoft, commenting on the Wall Street Journal's January 28 cover OS/2 story. The 1991 Microsoft strategy seminar had been planned for months as a way to "get the word out" about future developments. But this year was different -- because of a story in the Wall Street Journal ("Microsoft is dropping OS/2 and plans to build its future operating system software around Windows, page 1, January 28, 1991), the conference assumed an unintended importance, for Microsoft as well as the attendees. It's difficult to distill down all that was said, formally and informally, into just two pages -- but I'm going to try. One note up front: everything I'm going to say is "from the horses mouth", _but_ I'm talking about _plans_ rather than _products_ -- so this all may change! With that said, Microsoft's strategy for the next few years revolves around four major areas: New Technology (NT), Windows, Networking and Objects. New Technology _everywhere_ The first of these, NT, is the answer to the questions all of us have been asking about the future of OS/2. It's _not_ a product, and probably will not be -- rather it's a technology which is going to wind up in a lot of places. NT is a completely new Operating System kernel being developed by Microsoft'sadvanced operating systems division. In contrast to earlier operating systems, including both OS/2 and DOS, NT is modular, portable (98% written in C), and based on an object-oriented, client-server architecture. It's designed to run on high performance processors -- initially the Intel 386 and 486, probably later on one or more RISC machines -- which support a "flat", unsegmented, memory address space. The NT kernel is also designed to support symmetric multiprocessors, by dispatching execution threads or complete programs to different CPUs on a prioritized basis. At the moment, the single-tasking kernel is complete, and Microsoft sources tell me that it's fully operational with an OS/2 command-line interface _on both Intel and RISC based microprocessors_. Programmers can expect to see NT development tools by the end of this year. The second area of advancement is Windows, which is to become the standard user interface, _and_ the new mid-range operating system, when combined with the NT kernel. The new APIs for using Windows with that kernel are called 32-bit Windows, or "Win32", and include many of OS/2s features, such as preemptive multitasking, execution threads, and 32-bit "flat" memory addresses. Windows is also becoming modular, with many advanced feature sets, called "extensions" available to be added to the base system. Extension sets under development include Multimedia (covered elsewhere in this issue), Sound (a subset of multimedia) and Pen support (sometimes called Pen Windows). The Pen Extensions are particularly interesting, and were demonstrated at the seminar -- they permit a hand-held stylus to function as a mouse replacement, but go beyond that to provide real-time handwriting and gesture recognitiion. The result is that unmodified Windows applications (Excel was demonstrated) become completely operable without a keyboard or mouse. This is likely to produce a revolution in new, very small, personal computers. For the near future, before the NT kernel ships, Windows is being modified to coexist with OS/2 in two ways: first, by integrating the former system migration kit (SMK) and binary compatibility libraries (BCL) into a single new system called Windows Libraries for OS/2 (WLO). In contrast to previous attempts to marry Windows with OS/2, WLO must be used by the programmer -- it will not allow existing, non-WLO compatible Windows applications to work in OS/2 -- on the other hand, WLO has the great advantage that it actualy _works_. Microsoft announced that versions of Excel and Word created with WLO are in final beta now. To encourage developers to use WLO, Microsoft intends to integrate it directly into the Windows SDK sometime this year. Windows will also, of course, run with DOS 5.0, which will be introduced shortly. DOS 5.0 amounts to an enhancement of DOS to improve performance, and includes no new features of earth-shattering importance. Networks and Objects Networking is the third component of the overall strategy -- and is the one that assures OS/2s future. Microsoft sees networks as the majority of its future business user base, and support for these users as essential to their continued prosperity. Accordingly, a high-end Operating System based on the NT kernel, currently called OS/2 version 3.0, will be enhanced to function as the best possible server operating system. These enhancements will include moving much of the LAN Manager kernel into the operating system, and exploiting the installable file system (IFS) feature to provide new levels of functionality. One of these new file systems will be a distributed file system (DFS), which will permit any storage volume on any network server to be acessed by any user (security settings permitting, of course) -- in effect, "M:" will span many disks on many computers. In addition, greatly improved security features will be implemented at the operatingsystem level, allowing NT-based operating systems to be used at secure government facilities. This will be complemented by the incorperation of a POSIX-compliant API for government users. Windows will also be modified to work better on the network -- Microsoft's view is that the NT-based enhanced Windows product will be the preferred system for network _clients_. To this end, the Win32 API will include named pipes, mailslots, and so forth. The named pipe facility is particularly important, and will be enhanced to provide peer-to-peer operation, unlike the current system, where pipes have to be created by servers. _Some_ of these features _may_ be added to the 16-bit Windows API set, to permit users of less capable (286 based) PCs to take advantage of the network -- but this has not been decided at this time. Further networking enhancements will include combination of DFS with a powerful Remote Procedure Call (RPC) system, allowing networked NT to function as a true distributed operating system, with programs executing on various machines on the network. This will combine with multiprocessor support to make NT the ideal base for enterprise-wide information systems. The final component of the overall strategy is Objects. This is at the core of the "Information at your fingertips" concept introduced by Microsoft CEO William Gates III at last years COMDEX. Conceptually, it means that the user will no longer be aware of, or concerned with, details of the location of programs and data -- indeed, the distinction between the two will blur significantly. Gates forsees the elimination of the current program manager and file manager, and their replacement by a query-based interface -- rather than finding a document yourself, you will describe it to the computer, which will find it for you. In combination with the new user-interface technologies provided by multimedia and the pen; and the distributed computing architecture; Gates expects this to compltely revolutionize the personal computer as we know it. There's lots of other news, especially for developers -- but these are the high points of the overall strategy. What Does it all Mean? Let's talk about time scales: the only NT-based stuff we will see this year is some sort of NT toolkit -- you can forget 32-bit Windows and OS/2 3.0 in the immediate future. What we will see this year is a series of enhancements to Windows, especially in the area of network support. This will be followed by the debut of NT-based products sometime next year, probably beginning with an enhanced Windows product, followed by OS/2 3.0. The big_ question is: what to do in the meantime? If you're a Windows user or programmer, relax -- things are going to get better over time. Windows developers should be sure to stay on top of the developers kits, and make sure to get new SDKs as they become available, so that they can take advantage of the new APIs. Windows users can expect to see a reasonably smooth transition to NT-based applications over the next couple of years. Users with entry-level (8086 or 80286 based) hardware will want to think about upgrading in order to take advantage of the new features, but as they will have more than a year to do so, it should be a relatively painless process. The OS/2 situation is more complicated. OS/2 users can take heart at the imminent release of WLO, which should bring lots of Windows applications to OS/2 in the near future, and OS/2 version 2.0, when released, will actually be capable of running Windows itself in Standard Mode. Microsoft's comittment to an NT-based OS/2 as the LAN Manager server means that OS/2 remains a safe bet as a server platform, so network administrators can breathe a sigh of relief. Developers of network applications need to start looking hard at Windows. It's the developers of native-mode OS/2 programs, and users comitted to OS/2 and Presentation Manager (PM) who have the most confused situation right now. Microsoft did comitt to maintaining all the OS/2 APIs, _including_ PM, in OS/2 version 3.0 -- but they also made it clear that Win32 is the preferred API. Users and developers comitted to PM, especially those who've bought into IBM's Systems Application Architecture (SAA) must look to IBM for guidance as to what the future holds.... And IBM's not talking! All Rights Reserved - (C) - 1991, Windows and OS/2 Magazine