Life after the mainframe


WE'RE well beyond the mainframe now. The old world of monster Honeywell machines serving a captive community of slave terminals and slavish users is but a distant memory. The microchip revolution and the introduction of the PC saw to that.

With the help of a few highly paid consultants, end users could mount the computer show themselves and no longer needed to be in thrall to the sluggish central data processing staff the PC, for better or worse, was king.

Conor Sextons stimulating book deals with the consequences of the anarchy unleashed by the uncontrolled proliferation of PCs and small PC networks. He looks at the current trend towards open systems and at the hardware and software standards which underpin the desirable aims of portability and inter operability or, in English, getting disparate computer systems to talk to each other.

Most of the book simply explains the standards which the many international standards bodies have ratified. Sexton (a consultant and lecturer based in Dublin) succeeds in presenting a lucid and admirably broad account of standards for networking, the Unix operating system and programming languages. He presents his facts clearly and the book serves as a useful introductory reference, especially to networking standards.

Interpreting the facts, however, leads us to conclude that compared to the task of standardisation in the computer industry, Sisyphus had an easy job pushing his boulder to the top of the hill in Hades.

The Unix operating system, written mainly in the C language, is portable across a wide range of hardware platforms and offers the best hope for providing a standardised operating system environment, consistent with the ideals of open systems.

But there are as many as 20 versions of Unix Sexton identifies eight mainstreams, the two most important being Unix System and OSF/1. The treatment of standards for operating systems in the book is focused almost entirely on Unix it tells a complex tale of shifting alliances between the major players (AT&T, IBM, Digital, HewlettPackard, Novell) as the standards struggle to keep place with the ever evolving commercial versions.

There are Posix (Portable Operating System Interface Standard), X/Open and Cose (Common Operating System Environment) standards to grapple with. This least proprietary of operating systems is proving resistant to standardisation.

When we look, as this book does, at the microprocessors (Intel, Motorola), the computer languages (C, C++, Fortran, Basic, Cobol) and the operating systems (variants of Windows, 05/2, Unix, VMS) which dominate today, it is their diversity which is striking.

We can have a standard ANSI version of each individual language but the programming universe is still a veritable Tower of Babel. Standard SQL for relational databases and the de facto ODBC connectivity standard are honourable exceptions.

At bottom, the sheer diversity of operating systems, from the Apple Mac at one extreme (graphical user interface all the way) to Unix at the other (command driven by obscure commands) militates against the kind of standardisation which Sexton forecasts. Incidentally, the Mac and the rich VMS environment from Digital are barely mentioned in the book IBM systems and Unix dominate.

When it comes right down to it, open systems depend on standard combinations protocols with standard protocols, networks can talk to each other and client server or distributed systems can be developed. The 051, Reference Model devised in 1977, provides an elegant standard framework for communications protocols between open systems.

Even here, especially here, the real world intervened. TCP/IP, originating in the mid 1970s before the 051 model, was developed and heavily funded by the US Department of Defence (DoD) as Sexton explains, several large computer vendors, including IBM and Sperry were commissioned to provide."

TCP/IP with the systems for the US government. TCP/IP made the DoD's Arpanet network a success, becoming the backbone of the emerging Internet the rest is history.

The layers of TCP/IP can be mapped reasonably well conceptually to the 051 model, but it is, the TCP/IP that continues to, make the running in the Unix and open systems world.

In the book's Afterword, the author makes some forecasts he predicts a generic operating system similar to Unix, with a bit of GUI thrown in, which will hold sway after the millennium they will need to rush to make it in time for this coming millennium and then who will collect the royalties? The old war horse Cobol language will still be around, 20 years after the same complacent consultants who gave us "end user computing" forecast its imminent demise. Rear Admiral Grace Hopper can allow herself a contented puff on her cigar in the great software research lab in the sky.