IBMWatson autobiographythoughts on 766

Eric Sosman

Well, your formula of course is right.

But I am surprised that the incremental cost of CPU time to pre-sort an input file would be greater than manual sorting, especially when the sorted file would be stored on a temporary disk file.

I'm thinking of the labor cost of an operator to bring the raw cards to the sorter, stand at the sorter, and sort the cards through the multiple pbuttes.

When I developed a job stream on our S-360-40, I used a CPU sort and it was far faster wall-clock wise than handling the cards on the sorter. IIRC, the whole system had several sorts for various reports, all done internally.

Spread throughout a working day, eliminating an operator who just sorted and the sorter itself in favor of CPU sorting ought to have saved some money. But I guess CPUs were SO expensive back then it wasn't the case.

Admittedly, by the 1970s the tab machines were old and had more than paid for themselves many times over. But being old meant they were maintenance intensive and being mechanical added to that. I know data centers didn't worry too much about tab equipment sitting idle since in most of them that's just what the machines did. But still a fairly send craftsman C-E had to come out regularly and adjust and repair the machines and that must have had a cost.