Market research house D H Brown Associates just released a Consumer Report-style comparison of Windows NT and Unix chillingly titled The Inevitability of NT. However its findings do not necessarily point to the inevitability of NT on sheer merit, but because of the marketing clout Microsoft Corp is credited with having. In fact the study’s […]
Market research house D H Brown Associates just released a Consumer Report-style comparison of Windows NT and Unix chillingly titled The Inevitability of NT. However its findings do not necessarily point to the inevitability of NT on sheer merit, but because of the marketing clout Microsoft Corp is credited with having. In fact the study’s summary basically says NT wouldn’t even be a contender if it didn’t have Microsoft’s name on it. Albeit a worthy technical effort, NT’s lack of functional and speed advantages would normally prove fatal for a product targeting the brutually competitive desktop market. To begin with, Brown has trouble finding a market for NT. The study discards current Window 3.1 users in the intermediate term because it neither improves their speed nor enriches their functionality. It throws out current MS-DOS users who don’t need a graphical user interface because there’s no incentive to move. And it figures MS-DOS users upgrading to a graphical user interface will go with it instead of Windows 3.1 only after the pricing comes down and the difference between the two is only $200 or less.
They will also need more stability and higher performance to make the jump. These findings presuppose Microsoft’s dominant strength rests with the desktop. Brown estimates that a loss on the server side would adversely impact NT’s position on the desktop. At the very least, it says, Microsoft must enhance its credibility on the server side to maintain its desktop position. However, Brown figures large networked installations will stick to Novell Inc NetWare or possibly move to Univel Inc’s UnixWare. The relatively small numbers of servers in comparison with desktops, and the challenges of robustly implementing the distinct product capabilities of NT raise serious doubts concerning its viability as a server operating system. That leaves Microsoft with small peer-to-peer networks as a possible market if users require only simple file- and printer-sharing and electronic mail exchange. NT is a more powerful alternative to Windows For Workgroups, Brown says, and this market niche gives Microsoft the opportunity to undermine Novell from the low end rather than take it on head-to-head. Unfortunately Brown provides no numbers estimating the size of this market. The study then compares NT with Unix on 80 functional areas falling into three categories: user environment, operating environment and development environment. The Unix systems used in the comparison include AIX 3.2.1, HP-UX 8.0X, Solaris 1.X, Ultrix 4.2a and DG/UX 5.4.2. Their scores are averages and broken out as Best Unix and Worst Unix. NT dominates in the user environment because of the overwhelming proliferation of applications running native mode. However, this claim to fame means NT really offers no new dekstop innovations. It relies instead on backward compatibility with current Windows and MS-DOS products. On top of which its MS-DOS compatibility falls short of perfect, essentially emulating an 80286 running in real mode. Therefore some MS-DOS applications will not run. Those that depend on MS-DOS extenders, or attempt to control input-output devices directly, fail. In addition MS-DOS and Windows 3.1 device drivers have to be reimplemented. Graphics applications run very slowly because they have to go through a filter to ensure they don’t crash the machine. As a result, software vendors will have to ensure their applications work under NT. Unix at its best outperforms the NT windowing system and even at its worst is equivalent. Brown says the lack of network transparency and support for Display PostScript compares unfavourably to the flexibility of the X Window System with Unix.
By Maureen O’Gara
Moreover, NT’s graphics performance takes a hit. Preliminary data indicates performance has been compromised in comparison with Windows 3.1. Even allowing for the fact that NT has not yet been tuned for speed or reached production release, Brown says there appears to be additional overhead incurred when executing Windows 3.1 applications
under NT. So developers will have to convert for optimal performance, reinforcing the potential for confusing application and operating environment choices for users. Brown also faults the NT desktop manager as being much less flexible than the best Unix has to offer. On co-ordination services, NT is far ahead of Unix thanks to Dynamic Data Linking and Object Linking and Embedding, for instance. On installation however, Unix is again ahead, and is at least the equal of NT in intra-family portability, an area in which Microsoft’s superiority would normally be assumed. NT has a slight edge over the average Unix on-line documentation but falls short of the best available and the introduction of the CD-ROM is a hurdle for the desktop crowd. Moving into the operating environment, Brown says NT’s suitability as a desktop environment depends on Microsoft’s ability to shrink current RAM and disk space requirement down to something reasonable. NT, it says, has bloated memory requirements with 16Mb needed for practical use. Brown doubts that Microsoft’s pledge to reduce it to 8Mb in the final release will produce a very usable system, certainly not one that could run multiple applications concurrently. Brown reasons that memory requirements will inhibit NT’s widespread acceptance. NT’s massive resource requirements diminish the benefits of upgrading to a high-powered system. A machine that would be considered fully loaded for Windows 3.1 becomes a minimal NT-suitable configuration. In addition NT has extreme disk space demands for a desktop environment, approximately 50Mb, a severe increase over the 20Mb needed for Windows 3.1 but better than a typical RISC-Unix set-up. Hence Brown figures it’s unsuitable on a notebook. If it could boot from a network or a CD, disk space would be less of an issue but it can’t. On the other hand, NT’s pre-emption matches the best Unix has to offer, typified by AIX 3.2.1 and followed closely by DG/UX. NT also has excellent basic networking support, a key Unix feature, by bundling the functionality of LAN Manager. Its personal computer network protocol support is good but not the equal of Unix at its best. However, NT has yet to define its compatibility with critical software like NetWare. Brown notes here that NT must interoperate with it to be taken seriously in the corporate community. Unix at its best outperforms NT on the file system front but at its worst hardly scores.
The NT File System has only adequate space efficiency, lacking many of the innovations some Unix file systems offer. Also it does not allow cluster fragments. As a result, Brown says, on the desktop, where the user creates many small files and receives mail messages, the fragmentation will consume a disproportionate amount of disk space. Having been designed from the ground up, the NT kernel has yet to mature and rates as only adequate for robustness. Its goals of supporting advanced features like multiprocessing and security while remaining very portable and scalable and at the same time backwards compatible with MS-DOS and 3.1 will inflict costs in reliability and efficiency while it experiences its share of growing pains. Unix, meanwhile, derives its greatest unequivocal advantage from kernel robustness, where its maturity earns it a high degree of reliability. However robustness is not as great a factor on the desktop as on the server. Brown says NT exhibits a superior memory management scheme and is almost but not quite the equal of Unix at its best. Unix schemes, however, vary widely and the worst have little memory management optimisation and grant only limited control precision. Brown finds NT has a weaker development environment, lacking the powerful character-oriented tools, shells and scripting ability of Unix. Microsoft can barely compete with the decades of innovation poured in by workstation manufacturers. Its graphical user interface libraries give a good showing, but it lacks an integrated graphical application for compiling and editing and Unix developers will be frustrated initially by N
T’s primitive tools. And while it supports common languages well, its compilers lack the maturity of Unix compilers. Microsoft is also pressing the new and less mature C++. Despite the advantages of the best Unix systems, and the enhancements coming with Solaris 2.0 and OSF/1, Brown reckons the window of opportunity for Unix on the commercial desktop will shut within two years.