Talk:Harvard Mark I

From Wikipedia, the free encyclopedia

In 1955 I was a student at Harvard and spent many hours in laboratory time programming the Univac I prototype, which was located immediately adjacent to the Harvard Mark I computer, and the later Mark II and other early vacuum tube computers.

The Mark I was enormously large and complex, and, at the time when I first saw it, it was running a code for designing lenses for military equipment. This computational task ran for weeks and weeks, and produced all of its results on a set of 10 Flexowriters (really just computer driven typewriters). The programs were not in memory, they were on giant rolls of thick paper the width of the IBM 80 column card. And, there were reading stations for multiple rolls. After watching it for an instant, anyone could grasp the concept of loops since the controls tranferred from one reading station to the other and the execution could be watched while in process. The programmers often applied paper patches on the paper rolls to fix program errors -- these were real patches, and of course the word "patch" is used to this day for temporary changes to any software.

Originally, the memory was implemented with a set of disks the size of saucers mounted on shafts, each with a cam to engage the disk in one of ten positions. The number of disks seemed enormous to me at the time. There was also a very small vacuum tube memory added much later, but the Mark I was really an electromechanical machine.

My fondest memory of early computing was when, one day, the lens program was done and the Laboratory staff set up a new problem for the Mark I to solve.

I watched as two graduate students entered constants into the memory: One was on a stepladder in front of a huge panel with 10-position butterfly switches. The other student yelled out 4, 6, 7, 2, 1, etc., and the first student on the stepladder set the switches for each digit in each memory constant. This took many hours to accomplish because all the constants needed to checked and verified. Then the paper tape programs were mounted and sequenced to their start positions. The next step was the process of actually transferring the fixed constants to disk-on-shaft memory.

There was a very large lever mounted inside the machine on what looked like a transmission box. The lever was quite long, giving good mechanical leverage, and the lever was up in the disengaged position -- withdrawing the cams and letting the disks spin freely on the shafts. The shafts were put in motion, at first giving a low-pitched mechanical humming sound, quieting down as the bearings came up to operating temperature. Then, the student brought the lever down a short way and we heard the most enormous clatter of the cams engaging the disks. Of course the inertia of accelerating the disks caused a strong backlash on the lever and one man could barely hold it. The lever was brought down further and the clatter got even louder and it seemed like the computer was shaking and the lever oscillated up and down violently. Then, suddenly -- total silence -- and the lever dropped home to the engaged position when all the disks on all the shafts were syncronized to the settings on the butterfly switches.

The dynamic memory was implemented in a large array of relays which actually had a very small capacity. This memory was initialized and the tapes commanded to read and execute the codes contained there. This computer ran the problem for weeks.

But, this was the world's first fully automatic computer, and it ran many problems in its lifetime. It is a pity that it no longer exists, but I saw some parts in Museums in Boston and Sunnyvale, California.

Arthur Lemay Harvard Class of 1959

Does anyone know what sort of problems the Mark I was created to solve? I heard it was made to calculate ballistic firing tables for artillery during the war, but I'd like to read that from an authoritative source. I've also never seen a ballistic firing table and wonder what they looked like.