Information Technology Reference
In-Depth Information
delays and challenges of major programming projects. Project delays and cost
overruns were essentially the norm for this time period—it was a lesson that most
organizations learned the hard way, and a fundamental element to a perceived and
real “software crisis” throughout much of the 1960s and 1970s. When IBM's OS/360
project manager Frederick Brooks documented challenges from major programming
projects—including the lesson that adding additional programmers to a late project
can actually further delay a project's completion—in the
Mythical Man Month
, the
topic became an instant classic [26]. Brooks had outlined fundamental pitfalls in man-
aging software engineering that all too often victimized organizations. The topic,
which changed the landscape of software engineering, drew on lessons from the
1960s, but was not published until the mid-1970s.
In 1972, just after awarding CDC the contract, AFLC officials visited CDC's
Sunnyvale facility. Laccabue emphasized the substantial work yet to be done before
deployment. AFLC officials were surprised, but generally understood and took the
attitude that with substantial cooperative effort, these challenges could be met. Sev-
eral weeks later Laccabue met with General James Bailey, who in July 1971 had
become the deputy chief of staff, comptroller, Headquarters AFLC, overseeing ALS
for the Air Force [27]. He conveyed to Bailey that to be successful, the central control
system had to be written in machine code. He offered to submit a CDC proposal to
add this. General Bailey's view was that CDC and the AFLC continue to work closely
together, and as they got to next steps, the AFLC would entertain providing a contract
to CDC to write the central control system in machine code—but this never happened.
While CDC soon began to deliver Cyber 70 computers to Wright-Patterson and the
other AMAs (in the second half of 1972), the AFLC leaders, in spite of CDC efforts
to inform them, did not grasp the gravity of the software problem. Late in 1972, in an
early test at Kelly Air Force Base, Texas, a simple transaction to replenish a part got
caught in a loop and took 23 hours. This led to AFLC leaders' general recognition of
the severity of the problems with ALS. The system clearly did not work efficiently in
real-time. The AFLC leaders made the decision to redefine the project.
This redefinition of the ALS project occurred during renegotiations held between
the AFLC and CDC in January and February of 1973. Terms included giving CDC
later delivery dates in exchange for added work on the software and hardware [6].
While Elliott generally presents CDC as falling down on original specifications, this
is, at best, distorting. There were some issues with hardware reliability, but the great-
est problems were with the software, and system integration—namely integrating the
Air Force's central control system and CDC's ZODIAC. These problems extended
directly from what the CDC team had expressed concerns about to the Air Force from
the bidding process forward. At the renegotiation, the AFLC abandoned the concept
of a unified data base and real-time computing system in favor of multiple databases
and batch processing.
Both CDC's and AFLC's work on the project was significantly
redefined to build a workable batch processing system.
CDC had little choice but to
agree to the change and receive the extensions—but it was a bitter pill, as CDC had
delivered a system specifically designed to handle real-time processing efficiently that
would now be used for batch processing 90 percent of the time [28].
The redefinition
greatly reduced the sophistication of the ALS system [7]. Both Elliott's study, and a
GAO assessment report after ALS was shut down, stressed that CDC got the better of
the Air Force in this renegotiation [2][6]. CDC would provide a version of its SCOPE
Search WWH ::
Custom Search