It’s one thing to hear about the US Government wasting billions of dollars on the purchase, development and maintenance of incompatible systems and backing Unix as a way out of the problem. It’s quite another to hear a description of the chaos from the horse’s mouth. So it was a surprise to most of the audience when Gerald Riso, assistant secretary of the US Department of the Interior for Policy, Budget and Administration, stood up at Uniforum last month and reeled off not the expected platitudes about improving efficiency but a description of how vested interests, inadequate financial control, lack of communication and organisational problems lead to a situation where systems are not only costly, incompatible and antiquated but in many cases don’t even work. A horror story that has lost its impact through over-familiarity, you may think. But Riso was talking about a department with a 1987 hardware and software budget of $350m, and is only one of many. One example, he said, was that several years ago the department had no fewer than 13 different accounting systems running – and there are only 10 bureaux within the department. These systems were used to manage an overall budget of $5,000m a year, and themselves took 815 people and cost $27m a year to keep up and running.
You can’t run anything that way, said Riso, and the crunch came when the department called in a team of auditors who found that 11 of the 13 systems failed miserably to come up to approved accounting standards. Predictably, the problems come when you intrude and try to change the status quo. Riso noted that each Government agency and bureau has its own culture; the way they plan, set priorities and gear up to perform better are quite dissimilar. Among the factors that any advocate of change is fighting are the professional pride of different departments in acquiring and producing their own systems. And lack of communication between younger technical heads and bureau chiefs with an average of 55 gave a new meaning to the term generation gap. Economic considerations in purchasing systems are not overriding, he commented – the budgeting arrangements are so convoluted that it is possible to fund systems endlessly without it being possible to ever pin down exactly what was spent and the benefits received from it. And systems are bought on the basis of cost avoidance promised by a vendor – give me so much today, you’ll save x% in 1989 – but the budget accounting simply wasn’t capable of monitoring whether the pay-off ever came. Two kinds of action are being taken at the Department to counter these problems, he said. Tighter control over spending includes a top level review of all DP acquisitions of over $1m. And there is an overall drive to reduce the number of systems in operation, one result of which has been to reduce the four payroll systems in use two years ago to one today. But these efforts are fighting entrenched attitudes, are a direct threat to the plans and interests of staff in the Department bureaux and require substantial personal commitment; Riso was astonished by the number of people who react in a negative way. And the stakes are high, he pointed out; in one case, where it was decided that instead of diving straight into development, three months would be spent surveying the market for a suitable package, Within 48 hours, we were pulled up before a congressional committee to explain what we were doing. And in the US, where top departmental officials are political appointees, many of the vested interests reckon they can hold out until the likely end of the Reagan administration in a couple of years and the people at the top may be replaced by someone more malleable. That may work against Riso pushing through with his most ambitious proposals. Faced with the need to keep costs down and deliver efficient systems, while finding it difficult to attract and keep good people at the comparatively low government salaries, Riso is looking to contract out some of the Department’s computer operations altogether. But that
, as he remarked, is striking right at the heart of the problem where it will hurt most; the vested interests and futures of the staff. As for the multiple, inadequate accounting systems, by 1988 Riso hopes to reduce the number to one, and already, after meeting six vendors, Department officials were convinced that at least two would fill their needs or anyway come closer than those already in use, at a fraction of the cost. Riso managed to get through his entire speech without mentioning Unix once, but elsewhere at Uniforum several Government speakers described how they coped with installing Unix systems. At one presentation, Tolerant Systems’ Shirley Henry commented on the often-quoted figure of 70% of Government RFPs specifying Unix; she said that a minimum estimate was that 30% of government procurementc include a hard and fast Unix requirement; more like 60% suggest a Unix preference. The Internal Revenue Service, a body with 80,000 staff, is a long standing Unix user that started working on a definition for a Unix procurement back in 1981, when there were few vendors about. Now, according to the IRS’ Gene Barbato, Unix is standard within the IRS, and users wanting anything else have to go through a waiver process. By 1983, an initial RFP resulted in a contract awarded to Zilog and there are now some 430 Zilog supermicros distributed among IRS offices US-wide. That contract ran out in 1985, and since then offices have been buying a handful of various Unix systems at a time, so that Altos, Fortune, AT&T and Sequent machines are also in use, with a probable total of 600 systems altogether. Given the lack of software at the time, the IRS took what now seems an extraordinary approach of buying the hardware, throwing it out there and saying ‘see what you can do with it’. Back in due course came a large number of applications and the IRS was faced with the problem of spreading them around as much as possible while limiting the distribution of source code to prevent losing control completely. The best solution, Barbato suggested, is an arrangement whereby applications are mandated at central office for nationwide use and then source code distribution is prohibited. The developer is presumably then faced with maintaining the thing. Barbato concluded that Unix allows a Government entity like the IRS to manage resources better in a multi-vendor environment – and, as he pointed out, it has to remain a multi-vendor environment because major procurements are handled by the RFP and tender process. Therefore a major factor in the decision to go with Unix initially was that 100 major IRS offices had no DP staff at all; We didn’t know how we would support anything bigger than a PC without going for the Unix approach. Since then, the IRS has developed 400 to 500 DP staff.
Adopting a standard operating system allowed some uniformity in systems administration and made it easier to implement security standards – but on the security side, We had to crack down or they don’t use anything. The IRS has ended up being effectively locked in to Informix and C for development – portability is reduced because most development uses old versions of Informix without an SQL standard interface – and other predictable problems have been finding C programmers, and third party software that is uniform across the variety of kit in use. Nor has it been immune from self-generated portability problems: shifting software from the System III based Zilog systems to System V boxes 90% of software gets ported in an afternoon. 10% takes three months. Next, the IRS is looking towards a big and hopefully better new contract which is based on the assumption that many of the existing systems will be replaced. One of the problems is that some of the older systems in use, while inadequate in current configurations, are not designed to be easily or cheaply upgraded and IRS officials suggested that expandability – possibly by using multiprocessor machines – would be a strong emphasis. It has also been scouting around for integrated office s
oftware that has to be as sexy as IBM, DEC, and Wang. It will also be supplementing the current piecemeal system of ad hoc communications between sites with a nationwide packet-switched system. Future purchases, Barbato said, will be eased not only by the Unix experience the IRS has built up but by being able to coordinate efforts with other Government departments – we are now speaking common systems; he also considered the acceptance of the IEEE-defined Posix standard a necessary step.