Comment Aperture Punch Cards in late 80s (and CICS) (Score 2) 620
Aperture cards are punch cards with a cutout for a piece of 35mm microfilm (picture) and about 50-60 characters of indexing data. They were used in the aircraft industry to handle blueprints, because they're fairly high density - a 747 can't even hold all of its blueprints on paper, much less take off with them, and almost all large aircraft back then were unique, with slightly different parts, shapes of metal pieces, etc., due to design and manufacturing changes that happen in parallel to construction, as well as to different end-user requirements.
My company had a contract to develop an aperture-card scanning system that would digitize the pictures and upload the index data to a CICS database. We were the low bidder, which back then usually meant that either we were bidding against system integrators who were even more expensive than we were, or else that the department that was doing the bidding didn't have a clue what they were doing. (Yup, it was the latter.) The contract was hopelessly underspecified, the end-users had pushed lots of scope-creep into it without changing the price, and the only things that were really specific were that it had to scan 1000 cards/hour (it was getting about 200) and the database had 5 unique key fields (the end-users had upped that to 6, which also meant the keys were no longer unique which the database needed), and the price and due date were fixed (they'd way exceeded both, but the database change gave them some negotiating room on schedule.)
My department got asked to help, because we did R&D on things like electronic publishing and Unix systems and system integration, but it wasn't as risky as it sounded, because we'd get lots of credit if we succeeded and wouldn't get the blame if we couldn't help them fix it. I got sent in to do the consultant thing, found many of the things we needed to find (mostly by asking lots of dumb questions about the right parts; I'd dealt with TSO about 5 years earlier and mainframes in college, but had never heard of CICS, and I was mainly a systems generalist and Unix hacker), and we borrowed some people who actually understood CICS to help. Fortunately, most of the problem turned out to be bottlenecks in the interaction between the Unix box driving the scanner and the CICS front-end to the database, which led to the scanners having to stop and wait and get up to speed again on each card, and once the communications got straightened out the scanners could run at full hardware speed, which was something like 1500-2000 cards/hour.