History
The first email across a network
In 1971 two computers stood side by side in the same room in Cambridge, Massachusetts. A message was sent between them that took the detour through the network, even though you could have shouted it across. That was the point.
What already existed
Messages between people via a computer were nothing new in 1971. Large machines were used by many people at once, and there had long been programs that let you drop a note into a file for another user of the same machine. That was useful, but it stopped at the cabinet wall.
What was missing was addressing across machine boundaries. As long as everyone worked on the same computer, a username was enough. The moment there were two, the message had to be able to say where it was going — and there was no notation for that.
Two programs, merged
Ray Tomlinson worked at Bolt Beranek and Newman, the company building the ARPANET. He had two tools at hand: one that filed messages for users of the local machine, and one that copied files across the network to another machine. His idea was to replace the first with the second — don't file the message locally, write it across the network to its destination.
The setup consisted of two DEC PDP-10 machines standing right next to each other, connected only through the ARPANET. That is precisely what made it a valid test: the message had to take the network route, because there was no other.
What Tomlinson could not remember afterwards was the content. He described it as
something like QWERTYUIOP — the top row of the keyboard, which is what you
type when the text does not matter. That is not a footnote but the most honest
indication of how casual this was: there was no reason to keep the first sentence,
because nobody suspected it was a first sentence.
Dates should therefore be treated with care. The year 1971 is established, an exact date is not. Anyone reading a day-precise figure is reading a reconstruction.
Why this was the actual invention
Technically, combining two existing programs was not a great leap. Something else matters: from that moment on, a person was no longer defined by their machine. Before, it was "user such-and-such on this machine". Afterwards there was an address that combined both and could be handed out without explaining where you sat.
This separation of person and machine is the foundation to this day. It is in every email address, and it is why the separator between the two deserves a story of its own.
What came of it
Email was never planned for the ARPANET. The network was meant for shared access to remote computing capacity: expensive machines were to be usable from anywhere. That people would write to each other was a by-product — and within a few years it became the most-used application on the network.
This pattern recurs so often in the history of networking that it is almost a rule: the traffic that carries a network is rarely the traffic it was built for.