Electronic Mail and the @ Sign

Who invented email depends on which email you mean. The first email to travel from one computer to another was sent in late 1971 by Ray Tomlinson, a programmer at Bolt Beranek and Newman (BBN), across the ARPANET. Mail between people sharing a single computer is older: at MIT, Tom Van Vleck and Noel Morris wrote a MAIL command for the Compatible Time-Sharing System (CTSS) in the summer of 1965. What Tomlinson added was the network — and, with it, the user@host form that every email address still uses.

No exact day was written down. Tomlinson's own account says only that the first messages were sent in late 1971.

Mail Before the Network: CTSS, 1965

Time-sharing let dozens of people use one large computer at once, each from a terminal, each with files stored on disk. Once people shared a machine, leaving messages for one another on it was an obvious next step. Van Vleck, who wrote the user-facing half of CTSS MAIL, describes the command being added to CTSS in August 1965 and announced in a Computation Center bulletin. A mailbox was a file in the recipient's directory; MAIL simply added the new message to it.

Van Vleck is careful about priority. He writes that computer mail has probably been invented independently many times, and that he has found no documented version older than the one he helped build. Tomlinson, for his part, wrote that single-computer electronic mail "had existed since at least the early 1960's". Neither man claimed to have invented email outright, and this article does not either.

SNDMSG, CPYNET and the ARPANET

By 1971 Tomlinson was part of a small BBN group building TENEX, a time-sharing system for Digital Equipment's PDP-10. TENEX already had a local mail program, SNDMSG, which let a user compose a message and append it to another user's mailbox — again just a file, one that others could add to but not read or overwrite.

Tomlinson had also written CPYNET, an experimental program for copying files between machines over the ARPANET, the research network that had carried its first message two years earlier. He noticed that CPYNET could append to a mailbox file just as easily as SNDMSG could. The CPYNET protocol lacked an "append" operation, so he added one, then folded the CPYNET code into SNDMSG. The result could deliver a message to a local mailbox or, through a network connection, to a mailbox on a different host.

Two PDP-10s, Side by Side

The first network messages went between two machines that, in Tomlinson's words, were literally side by side. Their only physical connection, apart from the floor they stood on, was the ARPANET. He sent a number of test messages to himself from one to the other.

Their content is lost. Tomlinson wrote that the tests were forgettable and that he had forgotten them; the first was "most likely" QWERTYUIOP or something similar. Once the program seemed to work, he sent a message to the rest of his group explaining how to send mail over the network — so the first real use of network email was to announce its own existence.

The new SNDMSG went out with the next release of TENEX in early 1972. Because many ARPANET sites ran TENEX, the feature spread with the operating system. Van Vleck notes that sending and receiving mail between Multics and other ARPANET hosts followed in early 1972 at MIT.

Why the @ Sign

SNDMSG needed a way to tell local mail from network mail. Tomlinson's answer was to append an at sign and a host name to the user's login name. Asked why, he answered that the sign simply made sense. In English, he explained, @ marked a unit price — "10 items @ $1.95" — and he used it to indicate that the user was "at" some other host rather than local.

The choice was not universal at first. Van Vleck records that on Multics the @ character was the line-kill key, so mail from Multics to other hosts used a -at control argument instead. The user@host order nonetheless became the convention and outlived the systems it was designed for.

From a Hack to a Standard

Network mail grew faster than its plumbing. The CPYNET method was soon replaced by mail handling inside a real file transfer protocol. In 1972 Tomlinson took part in a meeting on extending FTP to carry mail, according to the Internet Hall of Fame; that arrangement lasted until 1982, when SMTP replaced it.

Message format needed agreement too. RFC 561, "Standardizing Network Mail Headers," dated 5 September 1973 and co-authored by Abhay Bhushan, Ken Pogran, Tomlinson and Jim White, opened by noting that the FTP mail protocol of the day had no way to state the author, title and date of a message, and that every system sent that information differently. The Internet Hall of Fame describes it as the first standard for Internet email message formats, defining fields such as From, Subject and Date that are still in use.

How to State the "First"

The defensible version is narrow: Ray Tomlinson wrote the first program to send electronic mail between users on different computers connected to a network, and he chose the @ separator. "Inventor of email" without that qualifier overstates it, because shared-computer mail at MIT and elsewhere came first. Tomlinson was inducted into the Internet Hall of Fame in 2012 and died in March 2016.

For the network that made all this possible, see ARPANET's first message and the Internet and Web timeline.