This is the week that Unix International gets to take its shot at Microsoft Windows NT thanks to the publication of a detailed study it’s had done by Locus Computing Corp tellingly titled NT versus Unix (CI No 2,058). Taking potshots at NT is fast becoming an industry recreation despite it being a little unfair. NT makes such an easy target. This study, like others before it, finds it’s the Microsoft brand not the software that’s responsible for the momentum. In a nutshell, Locus says NT breaks little new ground technically. In addition, International Data Corp has come up with some projections that indicate not everybody is buying the hype. Despite a 95.2% compound annual growth rate, it says NT as a client system won’t pass desktop Unix any time soon. In 1996, despite the pundits tolling the bell in anticipation of Unix’s imminent demise, Unix desktop shipments will be close to 2.4m units versus only 855,000 for NT. NT will get even slower acceptance as a server with annual shipments reaching only 95,000 units in 1996 versus more than 800,000 for Unix. Locus, meanwhile, has dissected the preliminary release of NT that Microsoft is circulating for a functional comparison with desktop Unix systems like Santa Cruz Operation, Interactive, Solaris, Destiny and UnixWare. It warns users interested in open systems to be wary of NT because of Microsoft’s reluctance to implement standards or create something more than a limited proprietary system.
Ignores major standards
NT ignores Posix 1003.2, Posix 1002.4, Posix.2, Posix.4, XPG3, System V Interface Definition 3, Federal Information Processing Standards and X Window conformance. NT is, however, compliant with Posix 1003.1 base functions but it gets there via a subsystem that does not integrate with the Windows environment, and Windows applications are not Posix-compliant. NT claims to be portable but ignores most of the key open systems portability standards. Microsoft also continues to press for acceptance of its proprietary local network, electronic mail and other technologies as standards despite the existence of de facto and de jure standards supported by desktop Unix systems, a position that must give users pause. Of course just to run the thing, users will have to be prepared to buy more hardware than is typical on a desktop. Configurations need 16Mb or more to achieve acceptable performance. Once up, users run the risk of losing data if there’s a system incident because of the way NT’s file system is built. And for all the hoopla about NT, one would think the thing would be multi-user but it’s not. It supports only one user at a time. It can support multiple clients but only one actual user. Users hell-bent on transaction processing, the key downsizing element, had better look elsewhere. It’s not available under NT. It may also be hard for an NT user to communicate. NT supports basic terminal requirements through add-on programs and combinations of hardware and software. Limited direct-connect support for widely available TCP/IP capabilities are also included. However, the TCP/IP in the developer’s release supports only access from NT to other TCP/IP systems. Inbound TCP/IP sessions are not available and it offers only limited terminal emulation capabilities. File transfer functions, for instance, are not supported. A user also has to be prepared for only rudimentary back-up and restore features making NT a dicey choice for mission-critical and commercially sensitive environments. NT provides C2-level security only when the user selects the new NT File System option instead of the MS-DOS File System or OS/2’s High Performance File System. There is likely to be a compatibility issue moving data from either of these systems to NT File System and with applications that object to being moved between secure and insure files. While Microsoft is committed to a future custom version of NT that’s B2 secure, its plans for network security are not clear. Although it has announced an intention to provide Distributed Computing Environment conformance with the Windows Remote P
rocedure Call, it is not implementing all of the Distributed Environment’s functionality, only the Call. It ignores the security, directory services and the time services that are also part of the Environment’s secure core.
By Maureen O’Gara
This means it will be difficult to implement servers or distributed applications with a high-level of security or be compatible with systems that do and there will only be very limited interoperability between NT and Distributed Computing Environment applications. Multinationals meanwhile are going to find NT a bit parochial. The study finds that NT may be only monolingual with appropriate help files installed for local languages. The application vendor can clearly provide a language version of the binaries if desired, but there appear to be no facilities for multi-lingual binaries. The numerous messages hard-coded into NT will be difficult to find and translate. If this issue is not addressed, NT may come to be viewed as an American English system. NT’s greatest strength, on the other hand, is its compatibility with Windows. It’s a strength, however, that comes at a price. Users looking to develop distributed client-server applications, beware. NT does not provide remote windowing or networking services. Distributed graphical applications will be far more difficult to develop and will not integrate with existing applications. The addition of networking capabilities will mean a significant level of additional complexity and will require a proprietary and probably expensive support package. On the graphical user interface side, a user’s ability to manipulate system objects from the desktop is limited in NT as are drag-and-drop capabilities, a definite drawback. It does not generally support dragging-and-dropping files onto other applications nor can its drag-and-drop be tailored. Neither can file association be customised to the user. NT is not as easy to install as Windows. It requires the right adaptors and a CD-ROM. Network installation is particularly tricky and TCP/IP start-up is now something done from a MS-DOS session. The TCP/IP implementation is still poorly documented. It is not clear how much of the network administration will be done through the graphical user interface and how much through more primitive facilities. In general, when something doesn’t work right from the graphical user interface, the error messages require significant technical skills. The preliminary version is clearly not for the naive user. It remains to be seen how much Microsoft improves it. Compounding the situation, NT’s help facility is only minimally acceptable. There is no standard way of adding multimedia help. It offers no compaction format or help compiler to minimise system resource requirements. There is also no character-based interface for remote access, making it clumsy as a server. On the other hand, Microsoft has provided NT with a number of system admin tools. In the preliminary version, however, many of the options these tools provide are not fully automated. The administrator is required to update parameter files or use line-mode commands from an MS-DOS emulator session: what will change by NT’s release date is unknown. Locus believes Microsoft has underestimated the importance of open networking in the market. The emphasis on LAN Manager as its primary networking facility ignores the state of networking in the real world. NetWare support is not provided through LAN Manager. Novell has said it will make NetWare available on NT. Users that require to link to NetWare will be forced to buy an add-on package to integrate NT with NetWare networks. Worst, perhaps, Microsoft has neither provided nor announced support for Network File System or any other native Unix distributed file system. Providing network file services to Unix servers will require that LAN Manager for Unix be installed on each Unix file server. LAN Manager is available as an expensive add-on package for some Unix systems that will enable NT systems to have distributed file services with Unix systems only at significant
ly greater cost.
No native open network support
Add-on Network File Systems may materialise, but none have been announced. Without support, environments with large numbers of Unix systems already using Network File System will have to choose between changing all of these over to support the Distributed Computing Environment or being unable to integrate NT systems into these environments easily. Moreover, NT provides no native support for OSI, X25, SNA or DECNet. NT, as the study pointed out, is not a multi-user system, raising suspicions about Microsoft’s motives. While it can be used to share files and resources like printers, clients must be systems based on Windows or Windows NT. Environments not currently using MS-DOS micros will have to replace existing terminals and systems with Windows-based desktops. With no native capability for remote administration and maintenance, it will be difficult to set up NT servers in locations without a skilled administrator. On applications, the study says the MS-DOS and Windows emulation scheme Microsoft uses leaves developers facing a complete conversion to take advantage of any of the new advanced features. Each subsystem provides its own interface and applications can’t use more than one. For instance, it’s either the Windows Application Programming Interface or Posix. Applications using a mix of interfaces must be rewritten to native NT – those relying on Dynamically Linked Libraries written for the 16-bit Windows interface will have to be converted to the Win32 API to be implemented. Posix-compliant programs can run only in the Posix subsystem. More complex applications using other standardised facilities or interfaces will not be supported. Many programs running on Unix systems will have to be converted and hundreds of X Window applications are unavailable to NT. Hindering the conversion effort are the tools Microsoft provides, good for only compile and debug. Graphical compile and edit facilities are not included. The MS-DOS shell is the only shell environment available. Outside of MS-DOS and Windows, Microsoft has shown no success in providing advanced system software. Its departure from OS/2 and limited market share for LAN Manager and SQL Server cast doubt on its ability to deliver a complex advanced operating system, market-ready for business-critical applications.
