[ Date Index ] [ Thread Index ] [ <= Previous by date / thread ] [ Next by date / thread => ]
On 24/10/2015 19:53, Simon Waters wrote: > I'm busy telling my son statistics is a really interesting topic. > Statistics lets you spot really small effects in timing or deviations > from expected behaviour, and it can be really hard to prevent these. This is very true. >From a long time ago in my past. Much from little differences. A mathematics with 'some electronics-computing experience' lecturer was tasked with recording and allocating costs ( charged to decimals of a cent) for telephone department access dial up costs during early use of Arpanet at one USA University. His costs as billed by teleco were different from summation of department costs logged internal by a few cents per month. He argued this indicated some mal function, university bursars etc said quit talking! A few cents in thousands of dollars. Do not bother investigating. He decided to intensively study this 'few cents' costs. Eventually, it proved first experience of a foreign power infiltrating the USA military network. Only then were funds allocated to carry out a watch at every one of the few universities then on Arpanet. Reason a lot of data had been acquired from USA by others. Method: Gain access to university phone system, and then immediately dial out to a mil net site. Grab data, release call. Mil-net storage system saw a legitimate access from a permitted account so no investigation occurred or warning alarm set off. The very few parts of a second from access to university phone and then dial out, the statistically insignificance to university bursars few cents in a thousand per month, had proved the attach. At the time this caused a major change in Arpanet so that Universities could not get direct access to mil net storage. This was much complained about as mil-uni co-operation was foundation of many shared tasks. The uproar was great at Unis about the inconvenience. The data already lost was a great embarrassment to mil-net. It was kept quiet for a long time. -- regards Eion MacDonald -- The Mailing List for the Devon & Cornwall LUG http://mailman.dclug.org.uk/listinfo/list FAQ: http://www.dcglug.org.uk/listfaq