Carrier Pigeon

One of the Internet's most obscure technologies came to life last weekend: transmitting network information by carrier pigeon.

In 1990, David Waitzman wrote RFC 1149, a tongue-in-cheek standard for using pigeons to transfer information using the Internet Protocol (IP). On Saturday, a group of Linux enthusiasts in Bergen, Norway, succeeded in exchanging some data using the Carrier Pigeon Internet Protocol (CPIP).

The group transmitted a "ping" command, among the most basic operations of the Internet, in which one computer sends a signal to another, which in turn signals that it is attached to the network. In the experiment, packets of network data were printed on paper then attached to pigeons' legs. Upon their arrival at the destination, the data was transferred to the computer using optical character recognition software.

The Bergen Linux Users Group had some assistance from the Vesta Brevdueforening carrier pigeon club and Alan Cox, a programmer at Linux leader Red Hat and top deputy of Linux founder Linus Torvalds.

The pigeon protocol didn't mean the fastest of networks, though. Taking an hour and 42 minutes to transfer a 64-byte packet of information makes the pigeon network about 5 trillion times slower than today's cutting-edge 40 gigabit-per-second optical fiber networks.

With a bit more luck than the Bergen group experienced, a basic Web page probably could be loaded in a couple of hours, participant Vegard Engen said in an e-mail interview.

Why bother with such a pokey protocol? "Because it could be done, and because no one had done it before," Engen said.

Indeed, Waitzman said the only other attempt he knew of faltered for lack of avian carriers.

"In February 1996, I got an e-mail from Sprint Communications South Africa asking if anyone has implemented it, and that they are thinking of doing it as a publicity stunt," Waitzman said in an e-mail interview. "I replied 'No, and good luck,' and they responded that they were having trouble finding a carrier pigeon owner. I never heard anything more."

Carrier pigeons, with their navigational abilities, have been used for communication in numerous forms. A black check cock carrier pigeon named Cher Ami was awarded the French Croix de Guerre medal for delivering 12 messages in World War I. His valorous duty included a message that helped the U.S. Army locate its "Lost Battalion," a message delivered despite the bird's being shot by enemy fire, according to the Smithsonian Institution.

There were some hitches with the Internet experiment as well. "The main reason for the long ping times was the neighbors' pigeons," Engen said. "They were out flying in the valley. When our pigeons were released, they did not want to go home. They would rather fly with the other pigeons."

And the flocking tendency caused another problem: The arrival en masse of a bunch of pigeons that had been released at 7.5-minute intervals caused a lot of confusion among the pigeon handlers, who left some cages open. As a result, some pigeons returned back to the starting point before they could be fitted with a datagram. But the packet loss wasn't bad enough to sink the experiment, Engen said.

And, as usual among Linux groupies, Microsoft took some of the heat. One pigeon bonked into a nearby window, spurring Cox to quip, "Oh, no! Windows causing problems again," Engen said.

Waitzman's avian carrier protocol is one of a number of "April Fool's Day" standards at the Internet Engineering Task Force. Waitzman puts the total count of joke standards at 29, including a sequel he wrote in 1999, RFC 2549, for ensuring quality of service on carrier pigeon networks.

Waitzman, an employee of Internet pioneer BBN Technologies, seasons his standards descriptions with networking jargon adapted to feathered formats.

For example, "the carriers have an intrinsic collision avoidance system," a problem that afflicts the transfer of data on crowded networks. And regarding patent considerations, "There is ongoing litigation about which is the prior art: carrier or egg."

The pigeon experiment may have been just a lark, but it actually served a serious function in setting an example of how helpful good documentation can be.

"If all protocol tests included such good documentation, interoperability in the Internet would be even better," Waitzman said.