DATAPAC

~ the eulogy ~

The year DATAPAC was switched off, i had broadband at home in Ankara and had been using it for years. That is the fact that makes this grave worth writing. A packet network specified in the mid seventies, running switches designed before i was born, was still moving money across Canada in 2009, and Bell had to publish a date to make it stop. Most of the networks it grew up with had been dead for a decade by then.

It came out of a boring problem. Bell Canada rented leased lines to banks and insurance companies who needed to move data between offices, and the customer had to supply all the equipment at both ends, usually from whoever sold them their mainframe. It was expensive and it excluded anybody small. The answer was a shared service, and even its own history is muddled: Wikipedia says the work started in 1972 and was announced by Bell as Dataroute in 1974, then says in the next section that it was commercially released in february 1973 under the name The Dataroute. The dates in the record do not line up and nobody has reconciled them.

DATAPAC was packet switching added on top of that. Announced in march 1976 and open the same year, it beat every other network to being purpose-built for X.25 rather than converted to it later. The traffic went through SL-10 switches designed by Bell Northern Research and built by Northern Telecom, with the first nodes in Calgary, Toronto, Ottawa and Montreal, tied together over Dataroute's existing 56k microwave links. A network operations center in Ottawa watched the whole thing over two 9600 bps lines. Average trip time for a packet was 0.36 seconds, or 0.13 if you paid for priority, which is a thing you could buy in 1976 and cannot buy now.

~ the dossier ~

~ what came before it ~

Dataroute was the parent and it deserves its own line. It ran across the Trans Canada Microwave system and did something leased lines had never done: it folded several customers into one channel using time division multiplexing, so a small office could buy a slice instead of a whole circuit. You could pay less for a lower speed or for a window of the day rather than all of it.

The weakness of that scheme is what packet switching fixed. A time slot belongs to its customer whether or not the customer sends anything, so a quiet line wastes the channel. Packets fill the gaps with somebody else's traffic. On a network where most users are bursty, that difference is the whole economics, and it is why DATAPAC could sell to a corner store when Dataroute could only sell to a bank.

~ what it actually carried ~

By the end, the load was almost entirely machines. Interac Direct Payment ran over it, which means the debit card transaction between a Canadian retailer and a Canadian bank spent the nineties traveling as X.25 packets. Some automatic teller machines used it. So did the lottery terminals, sending ticket numbers back to head office.

There were three service classes and the naming was as dry as everything else here. DATAPAC 3101 handled teletype connections over dial or leased lines, and DATAPAC 3000 was plain X.25. The one that outlived the others was 3201, a leased-line class written specifically for the financial industry, which is exactly the customer that never upgrades anything until it is forced to.

~ the long ending ~

The encyclopedia entry contains its own contradiction and leaves it standing. One sentence says DATAPAC services were largely replaced by TCP/IP in the 1990s and 2000s, the way every western X.25 network was. The next says Bell phased out the service on 31 december 2009. Both are accurate, and the fifteen years between them is where this network actually lived: commercially finished, technically finished, still switched on because somebody's till depended on it.

Even the ending is contested at the edges. A Quebec blogger marked the moment on 2 january 2010 with a note about the robust Datapac 3201 network finally taking its retirement. Meanwhile the same Wikipedia article files a security researcher's X.25 survey, compiled in april 2012, under the heading of evidence of DATAPAC usage as late as 2011 and 2012. Bell's date and the scanning record do not agree, and there is no source that settles which of them describes the last packet.

~ what got buried with it ~

The address format went. On DATAPAC you did not have a name, you had an NUA, a number under the X.121 scheme, and you dialed a machine the way you dialed a person. There was no directory in the sense the web has one. You either knew the number or you scanned for it, which is why the surviving folklore about this network is almost entirely in hacker text files rather than in Bell's documentation.

It never became a consumer product. Canada's videotex ambitions lived one layer up in Telidon and its offspring, and those died in the eighties while the network beneath them kept billing. That is the reason DATAPAC lasted so long. Nothing about it was loved, nobody had a home account on it, and no fashion could kill it. It ended when the terminals ended.

@}-->--

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

« previous  ·  random  ·  next »