TAURUS

~ the obit ~

TAURUS was cancelled five years before I first heard a modem handshake. It belongs to the part of computing history that happened on mainframes inside institutions, in a country I had not visited, in an industry I would not have been able to describe at the time. I am writing it down because it is one of the largest software failures anybody has ever had to publish a number for, and because the number is still shocking.

The name stood for Transfer and Automated Registration of Uncertified Stock. London still settled share trades on paper in the early 1990s. A certificate moved and a registrar wrote it down, and everything that happened inside a two-week account period was settled in one batch at the end of it. TAURUS was supposed to abolish both halves of that. Shares would live in an electronic register with no certificate behind them, and trades would settle three business days after they happened rather than whenever the fortnight ended. It was going to replace TALISMAN, which only handled trades between participating brokerage firms.

The idea went back to the 1980s. The version that actually got built dates from 1990, and it was built in an unusual way: rather than pick one design, the Exchange took seventeen proposals from the various interests in the City and merged them. Everybody who mattered had to see their own requirements in the result, so the result contained all of them. The legal and regulatory changes that electronic ownership required were still being negotiated while the software was being written.

~ the rap sheet ~

~ seventeen designs ~

The merged specification is the whole story of the overrun. Scope creep usually arrives as a series of small concessions, but here it arrived at the start, as a founding condition, and nobody involved had the standing to refuse it. Registrars, brokers, banks and the Exchange itself each wanted ownership records to work the way their existing systems already worked, and electronic ownership touched law that had not been written yet.

IBM built the security subsystem, which was one of the parts that slipped, and finished it in the summer of 1992. Acceptance testing followed. Installation across more than two hundred financial institutions was completed in February 1993. On paper the project was weeks from being finished.

~ the day it died ~

On Wednesday 11 March 1993 the Exchange board accepted the recommendation of its chief executive, Peter Rawlins, that TAURUS be cancelled. Rawlins resigned the same day. The public admission came on the 12th and the press worked through it over the following week. The Exchange wrote off £75 million of its own money. Counting what the banks and brokers had spent installing and integrating a system that was then switched off before use, the bill to the City came to an estimated £400 million.

The Sunday Times called the failure to build it the beginning of the end for the London Stock Exchange.

~ the inheritance ~

The replacement was built by the Bank of England rather than by the Exchange. It was called CREST and it was deliberately smaller: one central depository with a narrow remit, and no attempt to re-engineer the registrar business along the way. It went live in 1996 and it is still the plumbing under UK share settlement today, which is the quiet verdict on the seventeen-design approach.

The other inheritance is the case study. TAURUS turns up in project management courses and in risk digests, and it sits on the reading list for anyone being taught why requirements gathering is not a formality. Three decades later it is still the example people reach for when they need to explain that a project can be technically complete and still worthless, because the thing it was supposed to fit into never agreed to be fitted.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »