Datex-P

~ the last words ~

i never used Datex-P, and by the time it was switched off in January 2019 almost nobody in Germany thought they were using it either. That is the interesting part. A network can outlive its own visibility by twenty years if enough people installed something on it once and then stopped thinking about it.

The Deutsche Bundespost opened it in August 1980 as a trial and turned it into a regular service the following August. It ran X.25, the packet standard the CCITT had finished in 1976, which by then every European post office with any ambition was building on. The post office issued you a Network User Identification. What you dialed at the far end was a Network User Address, a number rather than a name, and the network billed for the packets moving between the two.

Speeds ran from 50 baud on an acoustic coupler at the bottom to 64 kbit/s at the top, which covers the entire distance between a man typing and a bank branch talking to a mainframe. Both of those were paying customers on the same wires.

~ the record ~

~ what ran on it ~

For German home users in the eighties, Datex-P was how you got out of Germany. CompuServe was reachable through it at German packet rates, which turned an American online service into a local phone call and an eye-watering bill at the end of the month. The Altos chat servers were on there too, and people dialed them for the same reason people dial anything. There was somebody on the other end.

The business side was where the actual revenue lived. Bank branches and point-of-sale terminals sat on X.25 links because X.25 was reliable in a way nothing cheaper was, and because the alternative was a leased line for every location. Once a shop had a working card terminal talking to a working host, the incentive to change anything dropped to zero for the next thirty years.

~ the fall ~

Frame relay took the corporate traffic first and IP took everything else afterward. By the late nineties nobody was designing anything new on X.25, and Datex-P stopped being a growth product and became a maintenance obligation with some revenue attached to it.

Deutsche Telekom handled it the way you handle a legacy service you cannot actually kill. In 2010 the existing connections were moved onto a new technical platform and no new ones were accepted. The operating company by then was I.T.E.N.O.S., a T-Systems subsidiary whose job was largely to keep the thing breathing for customers who could not leave.

~ the day it died ~

The end date was 7 January 2019. That is thirty-eight years after the regular service started and forty-three after the standard it ran on was written down.

The reason it took that long is the same reason it lasted. The remaining customers were security systems and payment terminals, installed by people who had since retired, in buildings whose owners did not know the equipment was there. Switching off a network in that condition is not an engineering problem. It is a hunt for whoever is still on the other end of it.

~ the legacy ~

Austria's version went first, in 2005, with a reduced service for point-of-sale kept alive over ISDN until 2013. The Austrian railways kept an internal packet network for signal control long after the public one was gone, which tells you what X.25 was actually good at. Not being fast. Being boringly correct on a bad line.

Every European country built one of these and the German one was among the last to go. Nothing about it was beautiful. It moved money and alarm signals for four decades on a protocol designed around modems, and the thing that finally killed it was not a better protocol. It was the last customer finally answering the phone.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »