The Unix factor.

What a fascinating (and genuinely unexpected) piece of news: For a tidy S350 million, Novell gets to buy outright ownership of Unix, as well as the development resources of AT&T's Unix Systems Laboratories. At least in theory, Novell now becomes chief architect, sponsor, and ultimate source for one of the industry's oldest and most powerful development platforms.

Of course, there's no way to tell if Unix will turn out to be a huge and unruly white elephant or the most brilliant acquisition of Ray Noorda's career. But one consequence is already clear: Unix suddenly becomes a far more credible factor in the race to dominate the high end of the operating system marketplace. High-end workstations and server-based systems are where Windows 3.1 is still weakest, and (if PC history repeats itself) this is the market that will shape development standards for the majority of next-generation applications.

So far, Unix has played an ambiguous role in the high-end market. Unix technology underlies several advanced operating systems--including Solaris, NeXT, and SCO's Open Desktop--but there has really been no one to evangelize for Unix itself. Ironically, Unix already has thousands of third-party

applications, dozens of substantial hardware OEMs, and millions of dedicated users--all the usual success criteria--and yet somehow Unix remains an underdog in competition with Windows NT, an operating system which isn't even shipping. Clearly, Unix has an image problem that Novell needs to take seriously.

In fact, Novell does have a sexy technology story to tell. Even before the acquisition, Novell and USL were beginning to roll out a graphical environment called UnixWare, which integrates Unix and NetWare, two of the most entrenched standards for enterprise and workgroup applications. We have no idea whether Novell now plans to extend UnixWare integration to private-label Unix licensees (a tricky problem, since Novell is already having trouble with Microsoft over the licensing of NetWare protocols). But we expect that NetWare integration increasingly will become a reference point for most Unix application development. It's hard to imagine a NetWare or Unix user who wouldn't find such a combined platform incredibly compelling.

But the most important result of Novell's Unix acquisition probably isn't technology so much as distribution . Novell already has a formidable and mature channel of NetWare VARs and systems integrators; now it inherits thousands of Unix developers, resellers, and hardware OEMs. Almost all of these share what might be called a "server-based" distribution model: The customer is typically an MIS professional (or the equivalent) who supports enterprise and workgroup applications. Service, trust, and technical expertise are the perennial hot buttons; direct sales relationships and negotiated licensing fees are more common than buying through retail channels.

In contrast, desktop software companies have relied traditionally on a client-based distribution model. PC applications most often become corporate standards once individual users become convinced that a product is personally useful (or fun, or powerful). There's rarely any face-to-face contact between customers and developers in this world; instead, retail and direct mail are the prevailing channels, and the best marketers are companies that are masters of packaging, pricing, and merchandising skills.

In many ways, the battle for the high end of the market is beginning to take shape as a contest between these two distribution models. Microsoft has placed a solid bet on a client-based strategy: By proliferating Windows 3.1 (and now Windows for Workgroups) throughout the end-user community, Microsoft hopes to establish a de facto API standard that will force MIS managers to adopt Windows NT on servers and networks. Meanwhile, Novell is likely to continue fine-tuning its server-based channels, leveraging relationships with network managers and developers of mission-critical applications.

As Microsoft, Novell, IBM, Sun, NeXT, Taligent, and others try to carve out turf in the high-end market, it's inevitable that some of the sharpest differences between these two models will blur. Microsoft is already adding service and consulting programs aimed at enterprise customers. And Novell keeps trying to develop a low-end shrink-wrapped network so simple that it becomes a genuine retail purchase.

But mature companies don't easily transform their distribution channels. Novell and Microsoft are both very good at selling to their traditional customers, and--probably--not especially competent at winning over each other's buyers. (In fact, IBM's laughable effort to sell OS/2 as a consumer product suggest just how inept a company can be when it ventures into an unfamiliar channel.)

Meanwhile, a lot of application developers are likely to find that the choice between server-based distribution (the Unix model) and client-based distribution (the Windows NT model) is a tough call. Each model has implications for product design, pricing, marketing tactics, and even company structure. Aldus and Interleaf both sell publishing software, but Interleaf's product line (which happens to be Unix-based) targets enterprise customers, while Aldus's PC-based products are designed chiefly for individual users. The two companies are very different in large part because they start with different distribution models.

It is possible--as Borland has found with Interbase and Lotus with Notes--to support both distribution models simultaneously. Nevertheless (as Borland and Lotus have also found), it's neither easy nor inexpensive to fight a two-front war. Thus, most software companies in high-end markets probably will end up making a choice--to develop Windows NT applications that are sold through client-based channels, or Unix-style applications for server-based channels. Right now, the tide is running strongly in favor of Microsoft's model; to succeed, Novell probably has to find a way to convince enough developers and customers that the Unix model is a plausible alternative.