Columns

The network is the computer, and the DBMS is the operating system

Late last year, Oracle and Sun announced a technology swap, giving Oracle access to Solaris components for its Raw Iron Products. What is Raw Iron? It is Oracle's DBMS-without-a-whole-operating-system "Internet appliance" offering. The premise is that a DBMS already provides OS-like functionality in areas such as memory management and general "processing" control. True, it is missing other OS elements, such as device drivers, but now that this deal is in place Oracle is licensing those from Sun.

And, as Oracle points out in connection with its "Aurora" Java Virtual Machine, the processing requirements of multi-user database applications are rather different from those of other kinds of computing. So Oracle's idea is this: If you're going to have a dedicated DBMS box anyway, who needs a general purpose OS? Can't we actually do a better job of giving you what you need? In theory, this makes perfect sense. Even so, in its first iteration, Oracle's Raw Iron is about as silly as, well, as a typical Microsoft Version 1 product.

There are two major flaws in the premise that haven't yet been addressed. First, maybe a general purpose OS isn't necessary, but how does having an OS on your computer actually hurt anything? Processing overhead? Administrative difficulty? Purchase price? Microsoft would find it easy to respond competitively in any or all of those areas.

Second, Oracle says that a major benefit of Raw Iron will be its simplicity, stability and administrative fit-and-finish. Already in Version 1. Yeah, right. I had hoped never to hear again about a Release 1 product that promised superior robustness after a short beta cycle. Indeed, the initial release will be Oracle 8.0-compatible, not 8i -- in other words the "Internet appliance" won't run the "Internet database."

Many silly Version 1 products eventually turn into Version 3 powerhouses, however -- and Raw Iron might just be one of them. To see a hint of why, let's consider more generally the future of computing.

OS vs. DBMS -- The end of Microsoft?

The Oracle/Sun deal is a no-money technology swap. Sun's side of the deal is that it gets to use the Oracle DBMS internally to support Solaris services, especially middleware. Now, it's not clear how important this is. After all, Sun wasn't in a position to charge a high price for the Solaris code, for several reasons:

  • Availability of low-cost alternatives (free Unixes).
  • Obvious benefits of integrating with Oracle's DBMS.
  • Obvious benefits of annoying Microsoft.

Suspect as it is, however, Sun's side of the agreement at least illustrates a key point -- operating systems will increasingly rely on DBMS or DBMS-like underpinnings. For a variety of reasons, there are going to be ever more "state," "context" or "user-profile" variables to track. And a DBMS will be needed to keep them all straight. Add in the DBMS' role in memory and process management, and it indeed seems that the OS of the future is going to look ever more like a DBMS.

By contrast, the central Microsoft value proposition -- industry-standard APIs for low-level C bit-twiddling -- will become less and less important. The Internet has brought us a whole new GUI API, and it's one that Microsoft doesn't own. And small, special-purpose smart appliances certainly won't house much of Microsoft's current technology. Microsoft Office will also become ever less important, even on standard desktop/laptop PCs. The Internet has changed document creation paradigms permanently. Traditional beautification is less important than it was, and it's not coming back soon. When complexity does get added back in, it will be database-oriented tagging (from XML on static documents to dynamic HTML generation). And while Microsoft Excel continues to be a crucial product, even it is sharing pride of place with an ever-growing universe of database-oriented business intelligence tools.

Microsoft's core server-side business model is to sell near-state-of-the-art products in high volume at low prices, while working to create intangibles such as ease-of-use that reinforce low total cost of ownership -- i.e., it's trying to commoditize OLTP operating systems and DBMS. It's succeeding, too. When NT 5.0 works, in whichever millennium that first occurs, only a few applications will need the traditional superior robustness and scalability features of, say, Oracle's DBMS or the HP-UX operating system.

But while Microsoft may be able to give fits to the current server platform leaders, we should recognize that Microsoft's own core technology strengths are equally subject to marginalization.

Interesting platform businesses --
The universe and me

While traditional platform businesses are looking increasingly commodity-like, other platform opportunities are emerging. Most of them have to do with networking, context-sensitivity and/or high-level logical APIs. Specific examples (some of them overlapping) include:

  • Distributed and mobile database topologies;
  • Intelligent appliances;
  • Speech recognition;
  • Document selection application paradigms;
  • Heterogeneous information access via "catalapps," a term I've coined for the new style of application represented by e-commerce, portals, knowledge management and other "catalogs" of information, goods and services;
  • User profiling and personalization; and
  • General context sensitivity.

Most of those opportunities, if you think about it, look a lot more like database management and synchronization, or else application frameworks, than they do like normal operating system services. The main exceptions would appear to be intelligent appliances, where the real issue is thin, fast Java and maybe speech recognition. But even those two categories have strong DBMS or application framework elements.

What Oracle should do

1Oracle has some interesting strategies for distributed database architectures. In fact, it has too many of them, including: Larry Ellison's "centralize everything" stump speech; the forthcoming Oracle Lite central administration via total database refreshing; and the Raw Iron "simple servers everywhere" strategy. These various good ideas should be integrated into one brilliant, interoperable whole.

2 Similarly, Oracle needs to get its data/ app server integration clarified. Everybody knows that the Oracle DBMS, the Oracle application server and the Oracle Java Virtual Machine should (and will eventually) be integrated into one coherent package. So c'mon guys, do it already.

3 One of Oracle's best "the database is the operating system" proof points is iFS, the "Internet File System" due in general availability some time in 1999. The idea is to simulate the Windows File System over an Oracle DBMS, thereby providing data integrity and searchability for all kinds of files. Taken together with interMedia, Oracle 8i's text/graphics datatype handling, this could make for a very interesting Internet content management platform.

However, iFS as currently explained is still an incomplete product. There are no tools that make it particularly useful. Oracle should invest in the tools, and, where appropriate, in getting third-party ISV support. And, since iFS seems to be a natural candidate for unfortunate performance overhead (a Java DBMS program replacing native OS services), Oracle should be careful about its performance tuning to minimize the problem.

4 As I keep pointing out, Oracle isn't trying hard enough in the ISP (Internet Service Provider) market. Certainly its never tried a sales push remotely as strong as Microsoft's and Sun's ongoing efforts. Of course, part of the problem is circular -- ISPs say "no" to Oracle's products, so it meekly goes away, instead of trying harder to win them over.

But maybe Raw Iron can be made into a product that appeals to ISPs. It's a long shot; a relatively new product will have trouble gearing up to the kind of industrial-strength that ISPs should, and someday soon will, demand. But if Oracle can pull it off, they could have a very nice way to bridge the networking/software skills gap. This is one long shot that' s worth a bet.

5 Oracle needs to push interMedia big-time. While Oracle has plenty of advantages over Microsoft among high-end users, it doesn't really have a lot of differentiation left in DBMS features for the mainstream market -- except interMedia. Assuming that interMedia finally works well in 8i -- which I think is likely -- it should be near the center of Oracle's product and marketing strategies.

6 Oracle should buy Nuance Communications. Speech recognition is the best shot at destroying Microsoft's monopoly. An acquisition of Nuance would undoubtedly be brutally expensive. But Microsoft, Cisco and AOL overpay for immature technologies all the time -- this is one case where Oracle should do the same thing.

7 (I'd rank this one a lot higher if I thought Oracle could actually pull it off.) Oracle should try for an across-the-board software partnership with Sun. Sun's problems in the enterprise software business are illustrated both by the ongoing JavaSoft fiasco and the botched NetDynamics merger. Now it has acquired a ton of enterprise software that it has little idea what to do with. In particular, via its role in the AOL/Netscape deal, Sun wound up with a second application server (Netscape/Kiva) that will never be coherently integrated with NetDynamics.

What's more, Sun is committed to ambitious strategies on AOL's behalf that it isn't at all qualified to carry out. Absent further interesting deals, the anti-Microsoft coalition will have most of its energy drained by the twin black holes of IBM and Sun. Sun has proved itself unable to manage a software business, and there's not a lot of time left for it to keep fooling around trying to learn.

I could keep on going with more ideas indefinitely -- but if Oracle really hopes to ascend to leadership of the software industry, these seven initiatives would be an excellent start.