Another nighttime story ... In 1983 (*), I started working in a company developing an X.400 system. Testing interaction with other systems was essential, so those of us who kept up old contacts at our alma mater was welcome to keep in X.400 contact with them - the University ran a different X.400 implementation. The underlaying network wasn't perfectly stable, so every now and then, messages didn't make it. One day we had a huge rush of incoming messages, some of them months old. All the missing ones were there. Where had they hidden in the weeks before? It took a while before we found an explanation. When the University mail system failed delivery, it was configured to make a new try later - at midnight. At our company, a raw disc offline backup copy was made every day, or rather: every night, starting at midnight. So when the MTA at the U made another attempt at midnight, night after night, our mail machine was just taken down for backup. One evening, our mail machine had a total crash. As it was already down, the operators decided to make the backup a couple hours early. The machine was back on air before midnight, ready to receive all the failing messages for months from the university. It took yet a couple days before anyone connected the early backup to the rush of emails, but when someone suggested it, the connection was obvious. (*) Some people claim that while internet protocols are based on real experience, real testing, OSI standards are just paperwork that never works in practice. That is of course far from truth. This was in 1983, a year before the fist official X.400 standard was passed. There were (at least) two complete, independent implementations available for testing. But, being ahead of time can be costly. My company obviously based the implementation on working drafts. For a couple years, the drafts for the directory functions (then still part of X.400 - later to be split off as X.500) was quite stable, the implementation was based on that. A few months before the finalizing of the official standard, a major part of the directory drafts was ditched, and another alternative pulled in. My company had to do a huge crash job to implement the other alternative, the one in the final standard. (Obviously, proponents of this other alternative were from companies who had a running implementation of that alternative ready. In order to be competitive, my company had to offer the standard solution as soon as possible, not something of their own, before the competition took over the mar
