/xenix: Networking choices
©1986 - Richard A. Bilancia - All Rights Reserved
to be published first in the April 1987 issue of UNIX&reg/World

LAN’s, WAN’s, NFS, RFS, coax, twisted pair, file locking, record locking -- confused about your choices for networking your XENIX&reg system with other UNIX&reg systems and DOS PC’s? Perhaps you don’t even need a network, a larger multi-user computer would be better: or perhaps a network of a combination of a multiuser computers and PC’s would be ideal for your situation. Such dilemmas are not easy to resolve. No single answer is right for all situations. In fact, the many tradeoffs have even the experts confused.

It would be nice if someone could write a decision table to help us, but as far as I know it has not yet been done. Accordingly, with a focus from the XENIX&reg perspective, I thought that I might discuss this month some of the issues and alternatives currently available.

Some background.

For the last few years there has been an ongoing feud between the various factions of the multi-user system (including UNIX&reg and XENIX&reg) advocates and the MS-DOS world PC based network advocates. Certain multiuser advocates kept suggesting that UNIX&reg should replace MS-DOS as the PC operating system of choice, while PC networking advocates suggested that large multiuser systems were now outdated (compared to PC’s) and would go the way of the dinosaurs. Of course neither position was correct, but both did have considerable plausibility. In any event, this conflict benefited no one, especially end users who simply got more confused about their choices.

More recently, possibly recognizing the damage to industry credibility that has occurred, calmer more rational and cooperative positions have emerged. Now one can frequently hear, "Let’s build a network system across several operating systems that capitalizes upon the strengths of each." When speaking about UNIX&reg and DOS this usually means using each of the various operating systems for what they do best: namely PC’s & MS-DOS for personal productivity and user interface activities and UNIX&reg or XENIX&reg and multiuser systems for large frequently accessed and updated databases, sharing of expensive devices including printers, and automatic remote communications. This approach makes even more sense today with the ever declining price of PC’s that is converging with the cost of intelligent terminals.

Another issue is that of security. No one anticipating the installation of a network should minimize the security aspects of that network. My accounting background keeps me constantly aware of security issues, and quite frankly, most networking options open a Pandora’s box of potentially catastrophic security holes. A detailed discussion of such security issues is well beyond the scope of this column, but if you are planning a network of any type, be prepared to spend considerable time dealing with security issues and concerns.

An ideal network.

No ideal networking system yet exists, but if it did I’d like it to have several features. First, the system should be easy to use for even novice users, but extensive enough that even operating system wizards can quickly and easily accomplish the most complex tasks. Such ease of use is often described as "...being transparent to the user." While such transparency may be desirable, it must be coupled with a syntax that is both intuitive and obvious.

Second, the system should take advantage of the various hardware processing capabilities of the computers in the network and distribute or share the processing load without adding a great deal of additional processing overhead. For example, if a consumer system requires only a single record from a server system, the server system should not supply the consumer system the entire contents of a database so that the consumer system can then extract the single record. Similarly, software needs to be designed in such a way that consumer systems are not normally allowed to process a large database on a server system and then replace that database with an updated copy (as would be the case with sorting a database file).

Third, the system should be flexible enough to require expensive optional cabling and hardware, only when economically justifiable. All too often I’ve seen overly complex networks constructed when far more simple techniques would have been more than sufficient, simply because the simpler solutions were overlooked when their capabilities were prematurely discounted.

The Microsoft offerings.

In last month’s column I described ’micnet’, Microsoft’s answer to the simple network requirements described above. While ’micnet’ is a XENIX&reg specific network system that does not incorporate MS-DOS computers within its structure, PC’s running MS-DOS can easily become part of a ’micnet’ network by simply running terminal emulation software and then logging in as intelligent terminals to one of the computers in a ’micnet’ network.

Additionally, Microsoft has developed Microsoft Networks (also called XENIX&reg-NET when purchased from The Santa Cruz Operation (SCO, Santa Cruz, CA)), a local area network (LAN) for XENIX&reg computers connected to other XENIX&reg or MS-DOS systems. According to SCO Vice President Doug Michels, "With XENIX&reg-NET, XENIX&reg becomes a complete environment for shared information and shared resource computing on personal computers. It puts to rest any contention of ’multiuser versus LAN,’ or ’XENIX&reg versus DOS.’ XENIX&reg-NET allows the user to take advantage of those aspects of DOS and XENIX&reg for which each system is best suited, without having to give up the functionality of either."

Aside from an AT compatible running the latest version of SCO’s XENIX&reg System V/286, XENIX&reg-NET also requires certain hardware cards to allow for high speed, broad band communications between systems in the network. XENIX&reg-NET is designed to co-exist nicely with the ’uucp’ system including ’mail’, as well as ’micnet’. While not all XENIX&reg commands can be used to access files on shared MS-DOS resources, most shell level commands can be executed to access remote data files by simply pre-pending a ’//system_name’ to the file desired. For example, ’grep fast //bilanc/usr/dict/words’ will run the grep program on the system where the command was keyed, looking for the string "fast" in the "/usr/dict/words" file on the "bilanc" system.

The three main features that Microsoft Networks supplies that make it a superior networking product than ’micnet’ are: a 2MBPS broad band transmission speed rather than 9600 baud with ’micnet’; the ability to include MS-DOS computers within the network; and the extended file-naming syntax used to identify files on other systems. The only disadvantage, if you can call it one, is the additional hardware required.

Another approach.

At Uniforum in January, Unify Corporation (Lake Oswego, Oregon) announced a version of their database and fourth generation application generator, ACCELL IDS (integrated development system), that addresses the issue of accessing and updating a large central database within the confines of a network that uses a combination of terminals and PC’s connected to a single multiuser system containing the database. ACCELL CP (standing for cooperative processing) is a package of programs that can distribute the processing load for an ACCELL IDS application from a single central UNIX&reg or XENIX&reg multiuser computer to several PC’s running MS-DOS connected to it.

Using the philosophy of allowing each component in a network to perform those functions that it does best, ACCELL CP transfers the input/output interrupts generated by each terminal key stroke to a PC running an MS-DOS based program that communicates with the multiuser host running a database serving program. This approach allows the record locking functions required in any multiuser application to be managed by a single computer while distributing the other more intensive screen and keyboard processing. The communications between the PC’s and the multiuser computer take place over standard serial lines (twisted pairs) and therefore eliminates the often unnecessary additional expense of network cards.

Conclusions.

Choosing a networking system is not a simple matter primarily because of the large number of choices and the fact that no true standards yet exist. Each different solution has some advantages over its alternatives, but no single solution is, or can be expected to be, perfect. Therefore, each network installation should be carefully planned with the help of an individual who understands these trade-offs, but more importantly understands the implications for the application[s] being networked.

Lastly, be prepared to incur the operating overhead costs of network administration. No network, no matter how simple or complex, can operate for very long without a knowledgeable and dedicated network manager.