Stretching DOS.

What's the verdict these days on DOS extenders?

We met recently with Terry colligan, president of Rational Systems, who gave us a briefing on his own DOS 16/M product and the DOS extender market in general. We're usually skeptical about people who peddle raw

' technology, but colligan's presentation was an eye-opener. Bottom line: He convinced us that most of the key standard-setters in the marketplace have already accepted DOS extenders as an essential part of their nextgeneration development strategies.

Is this hyperbole? Consider who's on the bandwagon so far: Lotus (which just bought 10% of Colligan's company) has adopted Rational's DOS 16/M to shoehorn 1-2-3 Release 3 into DOS machines. Autodesk, TOPS, Informix, Sybase, Cognos, and a good many other developers have also become Rational customers. Microsoft has promised that the next version of Windows will contain a proprietary DOS extender. And Ashton-Tate now says that dBase IV will soon incorporate some (unnamed) extender technology. To us, this looks like a pretty powerful consensus.

Why are these companies suddenly so interested?

First, extender technology is fairly tricky stuff that only recently became robust enough for commercial use. Intel's 286 chip is essentially a dual-processor envirornment; extender developers had to find a way to run applications in one segment, the 16 megabyte 'protected mode," while preserving access to the "real mode" (which emulates the less-powerful 8086 chip) for I/O functions that DOS normally controls. That's not a trivial problem.

Second--and perhaps more importantly--DOS extenders until recently were overshadowed by the assumption that OS/2 would soon become the dominant PC envirornment; interim solutions looked like a waste of time. Now that OS/2 is turning out to be a hard product to sell, developers of oversized applications are taking a fresh look at ways to stretch the limits of the basic DOS environment.

What's the pain level in moving to DOS extenders?

For users, the changeover is virtually transparent. Applications look the same; the only requirement may be some additional memory chips (which PC users have been installing by themselves for years). moreover, our guess is that most standard 286-based PCs will soon come with at least two megabytes of RAM, so users won't even have to open the box to get some immediate benefits from extenders.

For developers, the question gets a little more complicated. Colligan says that protected mode and real mode share almost identical instruction sets, so recompiling a standard Dos application for protected mode operation "usually takes no more than two or three weeks.' But he admits that there are some quirky applications that will take much more effort.

Who's likely to emerge an the market leader in DOS extenders?

That's a tough call. So far, Rational has a decisive lead over competitors like Phar Lap and AI Architects (who are focusing almost exclusively on 386-based machines). once windows 3.0 ships, however, we're bound to hear some intensely persuasive arguments from microsoft that DOS extenderB should be treated as part of its system software turf. In turn, that claim will make Microsoft's competitors on the applications side even more nervous, so the end result will probably be several years of drift and confusion.