Microsoft has talked the enterprise talk for a long, long, long time but, with rare exceptions, it has failed to deliver. And even when Microsoft has delivered, what it's delivered has been questionable. For example, until now, SQL Server probably has not been part of the migration strategy for huge DB2 databases. But with NT 5.0, we can start to see how that might happen.

Exchange is perhaps the most widely deployed NT application. While the client software is great, I'm always floored when I hear about the sorts of servers used to deploy it. Dual-processor servers with tens of gigabytes of disk space and hundreds of megabytes of RAM--all to run a mail system for a few hundred people. That sort of bloat and limited scalability is what leads many to dismiss NT as a back-end enterprise server.

It's a Matter of Attitude There are a good many IT professionals who think Microsoft lacks the proper mind-set to turn out truly enterprise-capable applications. After all, the company has been built largely on products that fly in the face of prevailing IT methods and technology. Managers within Microsoft who made that success happen have been rewarded for not following the enterprise crowd; however, some applications need the careful attention to cost of ownership, manageability and similar features that make software right for the enterprise.

During some meetings with Microsoft over the past month, its presenters made a point of detailing their enterprise credentials, mentioning IBM, Digital, Unisys, Oracle and others as the companies where they earned their enterprise scars. Some of them even lamented that in the past, they'd been ignored when they suggested that things be done in any fashion other than the proven Microsoft way (whatever that might mean). But it's still hard to tell if the prevailing Microsoft culture is one that can turn out a business application.

The time we spent with NT 5.0 didn't convince me that Microsoft has this down. Active Directory needs to mature; performance technologies, such as clustering support, aren't going to arrive until sometime after NT 5.0's release, and the programming interfaces that Microsoft supplies, particularly in Visual Basic, result in bloated applications that clearly won't scale. We get way too much pretty point-and-click stuff and way too little serious performance tuning from Microsoft.

In our NT 5.0 tests, we often reached the breaking point whenever we went beyond trivial tests. That doesn't mean NT 5.0 won't be ready for action when Microsoft decides to release it, but it does seem likely that the operating system won't live up to Microsoft's unspoken message: that NT 5.0 will be a better NetWare than NetWare or a better Unix than Unix.

Not There Yet Thirty years of system management finesse and flexibility still leaves Unix with the significant advantage over NT. Unix is the Swiss Army knife of the application server game. NetWare also has distinct advantages. Performance remains Novell's trademark, and its real commitment to Java means NetWare 5.0 deserves a serious look.

What Microsoft doesn't yet get, I think, is that simply providing DFS or an NFS client or even an AFP server doesn't mean that NT is ready to be deployed in the enterprise. Management and flexibility are key. Creating a system that can be used elegantly in ways that no one at Microsoft ever imagined is what it takes. Testing your software with thousands of users beating on it on real network configurations--that's what it takes.

From what we've seen, Microsoft has yet to get it right. The less trivial the problem we threw at the NT 5.0 beta, the more miserable the response. Should you evaluate it? You bet. Should you deploy it? Maybe for selected applications or in small-to-medium-sized workgroups. Should you run your enterprise on it? Not just yet.

Send your comments on this column to Art Wittmann at awittmann@nwc.com.