What would you do in four minutes?
In the fifties Britain probably led
the world in real-time computers.
Computers were gigantic and those on
the drawing board even bigger.
Silicon devices weren't yet around,
at least in commercial quantities, and certainly not at economic
prices.
Integrated circuits hadn't surfaced
and people designed their mammoth processing equipments with
germanium transistors.
Before even those were available in
quantity, and more importantly in a reliable form, neon tube
devices were in use.
Early electronic computers developed
in WWII had used valves and later included gas-filled things
called thyratrons. 12SN7 tubes, which were double triodes had
been popular and could be used to make single bit memory cells
and eight of these, for example, were used to make an elementary
storage device called a register.
To reduce size and power consumption
a bank of neon devices could also be used to make a rudimentary
register but these were very unpredictable. An early computer
prototype, using these devices at a certain Government establishment
worked OK until someone opened the door of the rack. Then the
change to ambient conditions from room lighting affected the
triggering of the neons and we had what was probably the first
computer crash (predating Windows by over 40 years).
Using experience gained on this early
model the first germanium transistor based computer prototypes
were developed.
The first machines being designed and
built in a purpose built factory in the North West of England,
under the vague guise of telephone equipment, and later installed
near London to form the heart of the system designed to provide
the UK with a 4-minute warning of impending Soviet attack.
The initial computers called XL2, later
supplemented by the XL4, took their names from the building in
which they were born (Exchange Laboratories) and were manufactured
in the early 60s.
No more than 25 of these gigantic data
processors were installed in a large building which, whether
by plan or accident, or for some other reason, was located in
a housing estate next to a main railway line!
I use the term "gigantic"
because, including their interfacing hardware, the 25 machines
filled some 1000 racks, each standing taller than a man.
Interestingly, during the years of the
hardware production, silicon transistors could have been used.
New computers were of course being developed
and these, not only used discrete silicon devices but also very
early integrated circuits.
"The powers that be", were
unsure of the new technology and, it is said, because of the
unknown reliability of such things vetoed their introduction
except in a very minor role.
In fact as experience was gained, reliability
of silicon transistors proved to be far better than that of their
germanium counterparts and integrated circuits even better than
this. Not only would reliability have been improved but device
speed would have been dramatically increased and power requirements
much reduced. But that's the way things were.
Being made from germanium transistors
the computers were much more complex than equivalent designs
based on silicon.
Early transistors suffered from thermal
runaway, so biasing of the devices was critical, and the extra
circuitry demanded required more current.
Enormous power supplies, providing many
hundreds of amps, were built, and these, backed up by lead acid
accumulators, provided the several different voltages which were
distributed by large-section copper bus bars to minimise resistive
losses. When one considers a building of maybe 300 feet in length
and perhaps half of this in width, the complexity of the power
distribution problem can be imagined.
Cabling of the computers was also a
mammoth task, demanding a complete floor of the building specifically
for this purpose.
Cooling, another huge undertaking, calling
for trunking to each and every rack so that the germanium devices
did not cook.
Apart from the real time data processing
for providing signals to radar consoles there was also raw radar
processing which occupied another mammoth system in an adjacent
room of similar proportions and the cabling of signals between
this area and the radar console floor above holding scores of
giant displays, was in keeping with the scale of the system.
Probably the biggest challenge though,
was not so much the hardware, because once designed, could be
mass produced, and although reliability was finite, this could
be calculated and provision made for failures.
The biggest challenge was software.
In parallel with the hardware development
of the first real-time solid state computers was the specification
and design of the means of programming the machines.
At the lowest level one used machine
code through which the hardware could be directed with precision.
Above this was assembler level and above this a high level language
which provided a general way of communicating with the hardware
for which a programmer did not need to understand the intricate
operation of the computer.
A new real-time language, loosely based
on the contemporary "CORAL" language, called "MINI-CORAL"
was written for the XL computers and provided the first insight
into the difficulties of getting computers to perform their tasks.
Because huge teams of people had to
be involved there was clearly a need for intercommunication and
this demanded tons of documentation and a carefully structured
organisation.
Translating performance specifications
into computer tasks using systems analysis was embryonic and
nearly all the staff involved would have had no previous experience
and had to learn on-the-job.
Anyone that's worked in Industry must
be aware that everything that's ever been written down is subject
to errors. These include misunderstandings, incompetence, plain
mistakes and failure to advise or communicate with other people
involved with associated software routines to which a particular
program may interface. On top of this add typographical errors
and compilation problems due to the conversion of the high level
or assembler level code into machine code.
All these things made life for the first
systems programmers next to impossible and the science of Configuration
Mangement, or the rigorous control of changes, which commenced
following the first Apollo shots was unknown in the UK.
Input to computers was via paper tape
using crude electro-mechanical machines and involving huge rolls
of paper tape which had to manually respooled after loading.
Correcting errors usually required a
second reel of tape, which contained program patches, to be loaded.
This was fraught with problems as it was necessary to distribute
the patches to other teams so that everyone was working with
the same up to date information.
Sometimes of course patches were unofficial,
and didn't work, and merely one of several attempts to overcome
a problem.
Periodically patches were assimilated
into re-compilations when everyone resynchronised their program
build. This recompilation would often generate more problems
than had been solved.
All in all for an undertaking of this
magnitude it was virtually impossible to complete the job before
the whole system became obsolete.
Programmers would move on to better
things, losing the project valuable experience, and those remaining
would be overloaded and unable to cope and of course new recruits
would need to be trained and this would dilute the efforts of
the others.
Any major computer system will suffer
from these problems, and the faster does computer hardware evolve,
the more the difficulty in completing to a customer's satisfaction,
the task of getting everything to work properly.
If one is philosophical, one could say
that to a large extent, a major defence contract like the one
described above, need never be completed as long as it is never
actually required to be used for its prime purpose.
If it added to our overall deterrent,
and I'm sure it did, then even if it never worked properly, or
had to be scaled down to a shadow of its original promise, it
had at least been useful in that role.
My memories are fading fast but I do
remember the huge screen showing really useful information with
the illuminated legend above, fortunately not counting down,
with those chilling words "MINUTES TO IMPACT... 4".
