ATM failure - No the other kind of ATM

According to today's Wall Street Journal, Citibank is explaining
the outage as "software", and declining to release further details.
But the article also noted that

  many of the computer systems on which banking networks
  are based are aging -- and trouble-shooters who can fix
  problems quickly are becoming rarer. Indeed, most banks
  and Wall Street firms run on a mishmash of systems slapped
  together after years of computer upgrades and bank mergers.
  hey are made all the more confusing by the profusion of
  computer languages that run the Internet. Most of them are
  incompatible with their mainframe predecessors.

i'll say. when we got into our little bank project, they were using
a Frame Relay WAN based on PCs with serial cards and MS-DOS based
software. it was 1999, they were never going to get a Y2K certificate
from the defunct integrator that built the PCs, they'd just gone public,
and were loudly proclaiming that they were Y2K ready on their website.
on top of that, their in house IT staff reminded all of us of Larry,
Moe and Curly.

when we suggested upgrading to cisco routers, their response was "isn't
that expensive?" we didn't say it out loud, but we were all thinking
"not nearly so expensive as the shareholder lawsuit you're risking
right this minute".

the situation was aggravated by the fact that NCR was deeply involved
in Y2K upgrades at the big customers, and didn't have time for 20
office small banks, and couldn't even be bothered to give ETAs for the
necessary ATM upgrades. we finally got the whole thing cut over
in december 1999.

the 9600 baud SNA links from the ATMs to the routers are
still unencrypted, and physical access to them is not hard at some
branch offices. it's a little frightening.