They're not merely replacing floppies, swapping in some emulators to take over. They're attempting to redesign and future-proof the entire system. That kind of a big deal. Oh, and it all has to run flawlessly during the transition period.
No, the $212 million includes the entire upgrade (and 20 years of support) of the automatic train control system. The full $700 million plus is for the overall modernization of multiple systems.
The idea was one computer on the LAN would hold the “talking stick” (the token) and transmit whatever data it needed to, then pass the token off to the next computer in the ring. If a computer received the token and didn’t have anything to transmit, it’d just pass on the token. The problem would be detecting when one of the computers in the loop had gone offline or crashed and taken the token with it. After some amount of time with no traffic, some system was responsible for generating a new token and an amended turn order. Similar problems existed when a new computer wanted to get added to the rotation.
It’s not quite as crazy as it seems. The older/larger floppy disk formats were more reliable due to their lower track density.
There was more surface area per byte of data. The old floppy disks could be written once and read for years in harsher environments. New floppy disks we more prone to failure after a few years.
Yeah if that's all that needed replacing. The entire system is ancient, not just the disks, like:
Much more critical than the dated use of floppy disks is the system's loop cable, which transmits data between the central servers and the trains and, according to Roccaforte, "has less bandwidth than an old AOL dial-up modem."
The SFMTA's website adds:
The loop cable is fragile and easily disturbed. This makes subway maintenance more difficult. This also means the system cannot be extended outside the subway, along surface rail, where currently we don’t have automatic train control.
All the deserved ribbing aside, if you had to design a removable, R/W, high-capacity, environmentally tolerant, secure, fault-tolerant, mission critical storage system that could last 25 years, starting NOW...
What would you pick?
That's a tough one, even if you design future hardware upgrades into the system.