NT Religious Wars
Why Are DARPA Researchers Afraid of Windows NT?
Mark Berman 1
BBN Technologies
10 Moulton Street
Cambridge, MA 02138
mberman@bbn.com
Submitted to DARPA NT Research Workshop
Seattle, WA
5 August 1998

Abstract: Researchers for the Defense Advanced Research Projects Agency (DARPA) are feeling increasing demand to conduct their computer science research on commodity hardware and software platforms, and specifically on Intel-type processors, running Windows or Windows NT. The forces behind this pressure are almost exclusively market-based rather than technical. However, many analyses of the desiderata behind OS selection are conducted from a fundamentally technical point of view. This paper presents a number of talking points that highlight the administrative and political challenges surrounding the decision to use Windows NT in leading-edge operating systems and networking research.

The Defense Advanced Research Projects Agency (DARPA) sponsors a significant amount of advanced research in computer science. As prices for commodity hardware have dropped over the past two decades, DARPA’s customers have become increasingly savvy about the nature of the research that is conducted, at least nominally, on their behalf. They have made increasingly strong demands for greater use of "Wintel"-type environments as the delivery platform for research products. Accordingly, DARPA has passed this demand on to the researchers who carry out DARPA’s sponsored research efforts.

In many cases, the research community has responded haltingly or even with hostility to the pressure placed upon it by its customers and its customers’ customers. Certainly, researchers have not rallied enthusiastically behind Microsoft’s operating system products as their preferred research environment. This fact should be surprising, given the uniformity of user preference. Even though customers have a legitimate need for Windows NT-based research products, the research community balks, stalls, and generally avoids embracing Windows and Windows NT. The result is an unfortunate tension between research sponsors and researchers. This paper visits some of the (mostly non-technical) motivations leading to this tension, with a focus on research in the areas of operating systems and networking.

I.    "Customer Pull": Customers want products that run under Windows (but will settle for products that run under Windows NT).

While DARPA is chartered to conduct "high-risk, high-potential-payoff" research for the U.S. Defense Department, DARPA managers are concerned, and for good reason, that this charter not turn into carte blanche for researchers to conduct too many "high-risk, no-payoff" projects. By keeping researchers’ attention focused on future technology users, DARPA increases the probability of the successful military or commercial application of the fruits of DARPA’s labors.

The customers whose needs DARPA must address fall into two categories. The first is the military users, people who fight on the front lines, work in military command centers, manage military logistics, and generally conduct the business of the nation’s largest organization. The second is the vaguely-defined future technology marketplace, individuals and businesses who stand to benefit from future DARPA-developed technology revolutions like the Internet. Customers in both of these groups, and particularly those in the military, need Windows (or Windows NT) solutions for two main reasons.

    i.    Microsoft has a monopoly on desktop operating systems.

    For all intents and purposes, it is reasonable to assume that all of DARPA’s customers already use computers in their work. Furthermore, they have been using computers for years. They have developed their own procedures and policies that are not conveniently switched to a different operating environment at the drop of a hat. And they are all using Windows. They are all using Windows because it is the superior choice for Intel-hardware operating systems for na├»ve users. The reason, however, is irrelevant, because the fact remains that users are using Windows. As a result, users who find themselves on the receiving end of DARPA systems developed on non-Windows platforms generally have no choice but to make room for another computer in their offices.

    Consider for example, the response of an Army officer assigned to the U.S. Atlantic Command (USACOM) Plans Office to the delivery of new software systems. Because USACOM is a frequent trial customer for DoD research projects, this officer sees many technology projects arrive in his office. So many, in fact, that he has found a lab where most of the delivered computer systems reside. While acknowledging the fact that these new systems provide ready access to information that is essential to his work, this officer nevertheless objects that "The last thing I need is another damn computer." Naturally, military users prefer research products that will integrate more smoothly into their current operating environments.

    ii.    Microsoft (almost) has a monopoly on desktop office suites.

    If the first step to improved customer acceptance is getting software to run on the same operating systems platform, the next step is to integrate seamlessly with application software. It is too easy to separate "special" applications from office suite-type applications. It is certainly true that DARPA research supports the development of real-time command and control systems that push the envelope of available OS support while they defend U.S. Navy ships from attack. However, it is equally true that the same person operating that system will probably use Microsoft Word to prepare an after-action report. That user will and should expect to transfer information from the operational application to the office application as easily as she can place a database table into a spreadsheet.

II.    "Go With the Flow": More and more researchers are already running Windows on their desktops.

Not too many years ago, very few self-respecting DARPA software researchers would admit to running a Microsoft operating system on their primary personal desktop machines. Although DARPA researchers are a persnickety bunch, they are giving in to the realities of the marketplace in greater numbers. DARPA researchers regularly exchange information in Microsoft Office formats. They use Windows or Windows NT for the same reasons the rest of the world does: the price is right; it gets the job done for basic office automation functions; and everybody else is using it.

Nevertheless, a lingering antipathy towards Microsoft OS and applications products lurks very close to the surface of DARPA researchers’ psyches. Frequently, a researcher will explain that she is using Microsoft products under some form of duress. She will explain that her organization’s IT department makes the choice for her. She will protest that her laptop computer dual-boots Windows NT and Linux, "and I can do almost everything I need to without booting the NT side."

A.    "Windows Sucks": Researchers have negative biases about conducting research on top of Microsoft OS products. (Sometimes these biases are well founded.)

Negative opinions about the technical merits of Windows NT put DARPA project principal investigators (PIs) in an awkward position. Customers are clamoring for Windows NT-based products, but PIs see two clear arguments against using Windows NT.

    a.    Developers think Windows sucks.

    Software developers who conduct DARPA research are interested in pushing the state of the art in computer science. They believe that they should be using state of the art tools in their work. Whether correctly or mistakenly, they believe that Windows and Windows NT do not fit the bill. Consider, for example, the following comment, addressing the less demanding IT arena.

    Yesterday's college students learned their UNIX expertise on Linux and FreeBSD. Today they're working in IT departments, and many of them are openly hostile to both Microsoft and Windows NT. As a result, Linux, BSD, Solaris, and other forms of UNIX are finding their way into IT departments, both overtly and on the sly.2

    Similar information comes from the informal "Operating Systems Sucks-Rules-O-Meter," which polls the Internet community’s zeitgeist by counting web pages that claim a particular operating system "rules" or "sucks" and plotting the results.3

    From one point of view, the correctness of developers’ opinions about Windows and Windows NT is not as important as the fact that such opinions are widely held. If the best and brightest software developers are proud to remain Windows-ignorant, PIs will continue to find it difficult to move to Windows NT and successfully staff their research projects. Furthermore, the top universities from which DARPA draws its research community are not producing Windows-literate graduates.

    b.    Windows really sucks.

    But even from a more purely technical viewpoint, PIs have some good reasons to be wary of working in a Windows NT environment. This paper is not intended to discuss in detail the relative technical merits of Windows NT and UNIX. However, many researchers accustomed to software development in a UNIX environment find themselves hamstrung by Windows NT’s development environment. Common complaints include the lack of a real remote login capability, poor support for scripting and utility languages, and frequent system crashes (the so-called "blue screen of death"). Developers often report on their efforts to make Windows NT more UNIX-like with packages like GNU Win32 and GNU Emacs, development tools based on the GNU (GNU’s not UNIX) project.

    This dilemma is particularly strong for researchers in the areas of operating systems and networking, two areas in which Windows NT is considered to be especially weak. With DARPA project lifetimes typically in the two- or three-year range, PIs must swallow hard before agreeing to take two giant steps back in order to begin their research on a Windows NT technology base.

B.    "Selling My Soul": Negotiating source license agreements for Windows NT can be hard.

 PIs conducting research projects in operating systems and closely related areas confront an additional impediment when choosing Windows NT as a technology base. Microsoft, understandably concerned about protecting its trade secrets, has traditionally been highly protective of the source code for its operating systems. The result has been highly restrictive non-disclosure agreements that some organizations have been hesitant or unable to sign. In at least some cases, university students with access to Microsoft source code have been required to agree not to accept employment with Microsoft competitors for a period after their graduation. While Microsoft’s motivation for strict source code control policies is clear, the policy is nonetheless at odds with researchers’ agendas.

Researchers do have the option of attempting to conduct their research without access to source code. While such an approach is theoretically possible, it is quite risky. Developing operating system software without access to source code is working without a net. If the project should run up against a problem that requires changes to the Windows NT software base, there is very real potential for substantial delay to acquire modified software or to develop a required work-around. This particular problem could be remedied by highly responsive Microsoft technical support, but the company’s track record in this area is not promising.

C.    "All Revved Up With No Place to Go": There is no clear technology transition path for OS and network research products built on Windows NT.

Finally, and perhaps of most concern to DARPA researchers is the challenge of effecting the transfer of sponsored research to the DARPA customer community. In the research areas of operating systems and networking, technology transition is optimally tightly bound to progress in the core operating system release. Ideally, researchers would hope to release their products either by independently re-releasing modified Windows NT products or by teaming with Microsoft to include enhancements in standard releases of Windows NT. Unfortunately, neither of these options seems particularly promising. The former option is likely to prove unacceptable to Microsoft, as it has the potential to place Microsoft into competition with DARPA research products. The latter option provides researchers with little control over their products. Furthermore, although DARPA research results often have the potential to offer widespread benefit several years in the future, the immediate results of research projects are frequently specialized, niche products in the present. It is unlikely that such specialized improvements will provide the mass market motivation needed to bring about changes in a mainstream Microsoft product.

This difficulty seems especially ironic because the promise of improved technology transfer is the chief motivator for DARPA’s initial desire to use Windows NT as a research platform. Solutions to this technology transition challenge are not immediately obvious. Close teamwork between DARPA and Microsoft will be needed to arrive at useful technology transition strategies.


Notes:

1.  Some research presented herein was conducted under DARPA contract number F30602-98-C-0187.  The views and conclusions contained in the document are those of the author and should not be interpreted as representing the official policies, either expressed or implied, of the Defense Advanced Research Projects Agency or the U.S. Government.

2.  Nicholas Petreley, The New UNIX Alters NT's, NC World, April 1998.

3.  Donald B. Marti Jr., “Operating System Sucks-Rules-O-Meter,” Electric Lichen L.L.C., http://electriclichen.com/linux/srom.html.