The @ Sign Nobel
John S. Quarterman
From Matrix News, 7(10), October 1997 A version of this article also appeared in SunExpert Magazine.
BBN (see ``BBN and Internet History Books'' in this issue) sent the first electronic mail message that travelled across a computer network, and in the process invented the @ sign syntax. To be more specific, Ray Tomlinson sent the first networked mail message in 1970, in collaboration with others at BBN and elsewhere.
This was never a secret, but nobody had taken much interest in it as a significant phenomenon, and few people remembered who did it, until I spent several months following leads to get to who I probably should have just asked in the first place: Ray. Peter followed up with interviews and published a writeup in Matrix News 502 (February 1995) ``Mail: The application that hadn't been thought of'' and in his book.
Please note that this was not the first electronic mail message. Electronic mail had existed since approximately 1965 on the Compatible Time Sharing System (CTSS) that had been developed by Fernando Corbato. It had probably been reinvented on almost every time sharing system since then. But BBN and Ray Tomlinson were the first to send electronic mail between two computers across a network. This seems appropriate, since BBN had in fact invented timesharing in 1962.
Although in hindsight networked electronic mail may seem an obvious application for the first geographically distributed packet switching network, it wasn't even part of the original plans for the ARPANET. It was a hack. A convenience. A time-saving device, to avoid carrying tapes around or talking on the telephone. Soon, it was an indispensible tool. This is the way many network applications have developed. Other examples include anonymous FTP, archie, Gopher, and WWW. But electronic mail was the first such example, the ur-example, the archetype of all the others.
The @ sign was very important for the development of computer networks. I don't mean the specific character @, although that character is a particularly appropriate choice, having as it does appropriate meaning already built in. But I want to concentrate on the syntax and the semantics of the use of the @ sign. One reason my first book was called The Matrix was that in the middle of it is a matrix, a two dimensional table, of methods for sending mail among networks with different addressing syntaxes. These ranged from UUCP a!b!c!d to CompuServe 1234.56789 to FidoNet 1:151/299.0 to mixed syntaxes like a!%c@d. Most of those syntaxes are now dead or clearly moribund. Even CompuServe has recently been sold, and the current leader in outfits most like old-style conferencing systems, AOL, now permits symbolic [email protected] syntax. The Internet DNS syntax of local_part@domain_part, which is merely an elaboration of Ray Tomlinson's ARPANET format of user@host, has won.
The beauty of local_part@domain_part is at least threefold. It is simple and often mnemonic. This is important, if people are to remember addresses and if they are to fit into common business conventions such as business cards. I couldn't even remember FidoNet syntax; I had to look it up for this article. Anybody can remember [email protected].
In addition, local_part@domain_part is complex enough to facilitate internetworking. The syntaxes used on conferencing systems such as CompuServe, Prodigy, EIES, etc. were not. Connecting such a system to a network required shoehorning its local addresses into a network mail address syntax. The @ sign syntax recognized from the beginning that the networked world would be diverse, and at least one level of visible hierarchy would be necessary to handle that diversity. Each host might be a universe unto itself (as conferencing systems tried to be), but a host that wants to talk to other host must recognize some sort of external address for other hosts together with a way to specify local users on other hosts. That is exactly what user@host does.
Less obviously, but equally importantly, local_part@domain_part leaves network routing to lower layers. Both UUCP and FidoNet spelled out routing in their visible electronic mail addresses. This was partly because neither of those networks had many applications other than electronic mail. (They had USENET news and Echomail, but that was about it.) The ARPANET and the Internet always did have several other applications, starting with FTP (file transfer) and TELNET (remote login). Those applications needed a host (or later a domain) address to connect to a remote host, but might not need a username or local_part at all. The utility of network routing common to numerous applications was more obvious on the ARPANET and the Internet.
Of course, UUCP and USENET were popular as a poor man's ARPANET for those who couldn't get access to the ARPANET, which at the time required governmental approval. FidoNet was invented as a poor man's USENET, for MS-DOS users. The inventors of these later networks all knew electronic mail as the most popular ARPANET application.
If there were Nobel prizes for internetworking, I would nominate Ray Tomlinson for one. Vint Cerf and Van Jacobson would be other obvious candidates.
Why isn't there an internetworking Nobel Prize, by the way? I think it is easy to argue that internetworking is changing the world, from academic to business to personal, in at least as many and profound ways as, for example, economics. There is at least as much academic rigor in internetworking as in, for example, economics. Electronic mail may have originated as a hack, but so did plenty of physics or medicine Nobel innovations.
Plenty of people in the computing industry have enough money to endow such a prize. For example Michael Dell, Andy Grove, Rick Adams, Bill Joy, Steve Jobs. I suppose one problem might be that some of the very people who could endow such a prize would be candidates for it. Another might be that many of the richest people in computing got their money from PCs, and don't really appreciate networking as more important than individual computers. Maybe George Soros could endow it.
It's a thought.
BBN (see ``BBN and Internet History Books'' in this issue) sent the first electronic mail message that travelled across a computer network, and in the process invented the @ sign syntax. To be more specific, Ray Tomlinson sent the first networked mail message in 1970, in collaboration with others at BBN and elsewhere.
This was never a secret, but nobody had taken much interest in it as a significant phenomenon, and few people remembered who did it, until I spent several months following leads to get to who I probably should have just asked in the first place: Ray. Peter followed up with interviews and published a writeup in Matrix News 502 (February 1995) ``Mail: The application that hadn't been thought of'' and in his book.
Please note that this was not the first electronic mail message. Electronic mail had existed since approximately 1965 on the Compatible Time Sharing System (CTSS) that had been developed by Fernando Corbato. It had probably been reinvented on almost every time sharing system since then. But BBN and Ray Tomlinson were the first to send electronic mail between two computers across a network. This seems appropriate, since BBN had in fact invented timesharing in 1962.
Although in hindsight networked electronic mail may seem an obvious application for the first geographically distributed packet switching network, it wasn't even part of the original plans for the ARPANET. It was a hack. A convenience. A time-saving device, to avoid carrying tapes around or talking on the telephone. Soon, it was an indispensible tool. This is the way many network applications have developed. Other examples include anonymous FTP, archie, Gopher, and WWW. But electronic mail was the first such example, the ur-example, the archetype of all the others.
The @ sign was very important for the development of computer networks. I don't mean the specific character @, although that character is a particularly appropriate choice, having as it does appropriate meaning already built in. But I want to concentrate on the syntax and the semantics of the use of the @ sign. One reason my first book was called The Matrix was that in the middle of it is a matrix, a two dimensional table, of methods for sending mail among networks with different addressing syntaxes. These ranged from UUCP a!b!c!d to CompuServe 1234.56789 to FidoNet 1:151/299.0 to mixed syntaxes like a!%c@d. Most of those syntaxes are now dead or clearly moribund. Even CompuServe has recently been sold, and the current leader in outfits most like old-style conferencing systems, AOL, now permits symbolic [email protected] syntax. The Internet DNS syntax of local_part@domain_part, which is merely an elaboration of Ray Tomlinson's ARPANET format of user@host, has won.
The beauty of local_part@domain_part is at least threefold. It is simple and often mnemonic. This is important, if people are to remember addresses and if they are to fit into common business conventions such as business cards. I couldn't even remember FidoNet syntax; I had to look it up for this article. Anybody can remember [email protected].
In addition, local_part@domain_part is complex enough to facilitate internetworking. The syntaxes used on conferencing systems such as CompuServe, Prodigy, EIES, etc. were not. Connecting such a system to a network required shoehorning its local addresses into a network mail address syntax. The @ sign syntax recognized from the beginning that the networked world would be diverse, and at least one level of visible hierarchy would be necessary to handle that diversity. Each host might be a universe unto itself (as conferencing systems tried to be), but a host that wants to talk to other host must recognize some sort of external address for other hosts together with a way to specify local users on other hosts. That is exactly what user@host does.
Less obviously, but equally importantly, local_part@domain_part leaves network routing to lower layers. Both UUCP and FidoNet spelled out routing in their visible electronic mail addresses. This was partly because neither of those networks had many applications other than electronic mail. (They had USENET news and Echomail, but that was about it.) The ARPANET and the Internet always did have several other applications, starting with FTP (file transfer) and TELNET (remote login). Those applications needed a host (or later a domain) address to connect to a remote host, but might not need a username or local_part at all. The utility of network routing common to numerous applications was more obvious on the ARPANET and the Internet.
Of course, UUCP and USENET were popular as a poor man's ARPANET for those who couldn't get access to the ARPANET, which at the time required governmental approval. FidoNet was invented as a poor man's USENET, for MS-DOS users. The inventors of these later networks all knew electronic mail as the most popular ARPANET application.
If there were Nobel prizes for internetworking, I would nominate Ray Tomlinson for one. Vint Cerf and Van Jacobson would be other obvious candidates.
Why isn't there an internetworking Nobel Prize, by the way? I think it is easy to argue that internetworking is changing the world, from academic to business to personal, in at least as many and profound ways as, for example, economics. There is at least as much academic rigor in internetworking as in, for example, economics. Electronic mail may have originated as a hack, but so did plenty of physics or medicine Nobel innovations.
Plenty of people in the computing industry have enough money to endow such a prize. For example Michael Dell, Andy Grove, Rick Adams, Bill Joy, Steve Jobs. I suppose one problem might be that some of the very people who could endow such a prize would be candidates for it. Another might be that many of the richest people in computing got their money from PCs, and don't really appreciate networking as more important than individual computers. Maybe George Soros could endow it.
It's a thought.
