Interview- Nathaniel Borenstein - The Father of MIME

I never set out to go into the business of email. I got to graduate school and Carnegie Mellon at the time, I think still, had a thing called the Lieberman queue. And grad students had to be in a queue, and when their number came up, they were assigned a sort of community service to the department. And mine was maintaining this ancient email system on Top’s Ten. It was such a crufty system that I got motivated to just rewrite a system on Unix. So I wrote a mail system there and then I thought, well, that’s enough of that and I did my thesis in user interface design. And then at, when I was looking for jobs post thesis after I’ve gone on all my interviews, my thesis advisor, who was also the head of the Andrew Project, said how would you like to stay here for a while and build the world’s greatest mail system? Email was a text thing in English only. They were extensions that had been made for other languages but they only worked within national or linguistic communities. So if a Japanese person went to France and used a computer there he wouldn’t be able to read Japanese email. So that was clearly not an optimal situation. Also back then we didn’t have just the Internet. We had a whole bunch of networks. We had Bitnet, we had Usenet, there were, Nanocef net. And people sent mail to each other across these networks. They were connected by gateways that were not always perfect because it wasn’t always possible to preserve all of the information exactly the way it was intended. So the way I like to put it is I’ve never met a gateway that made a message better, right? The best you could hope for is that it doesn’t make it worse. So, there were a number of problems. And I came from a research background where I had built a system that did multimedia mail, which meant you could have pictures and sounds and animations and all sorts of stuff that was relatively ahead of its time, as part of a team at Carnegie Mellon that did that. And what really got me thinking standards was this. One day Steve Jobs came to visit, and we demoed the Andrew message system to him and he immediately tried to hire our whole team and nobody went. I suspect that was one of his worst days ever in hiring. So he went home, this was when he was with Next Computers, and he put together a team to build a multimedia mail system. So NeXTMail is sort of an imitation of the Andrew mail system. And imitation is the sincerest form of flattery, that’s wonderful, I didn’t have any problem with it. But once it was built, it was hard to ignore the fact that while my users could send pictures to each other and his users could send pictures to each other, my users couldn’t send pictures to his users because we weren’t doing it the same way. So I started thinking, there really needs to be a standard here. Now simultaneously, there were a lot of people, almost all outside the US, worrying about the language problem. Worrying about getting non-English characters in other languages into email. And there was a third problem. There were people who worked on gateways. This gateway problem I described before. Who wanted standards for formatting that would make it possible to gateway a higher level of information to some of these other systems. So what MIME was, was a fortuitous coming together of those three needs. And the man who deserves credit for starting that process was a man named Einar Stefferud, who is one of the great unsung heroes of the Internet. He was on the Internet from the early 70s. He started the first Internet mailing list. So, he was a pioneer in many ways. He was also kind of ornery and I think didn’t get as much credit as he should have because of that. He passed away a few years ago. Anyway he introduced me to Ned Freed. Ned was working on gateway issues. We quickly met a lot of people who were interested in the language issue and I was interested in the multimedia issue. And quickly found that we could put together a solution for all of those things. And what’s good about that from a standard perspective is bandwagon. All those people who don’t care about multimedia mail get it because they want to get the multilingual mail. When you put together those three motivations I mentioned, multimedia, multilanguage, and gateway, you had a lot of people concerned. And it doesn’t matter how much goodwill you have, when you have that many people involved there’s going to be differences of opinion. I’d say there were about 70 people involved, not all of them as actively as others. And it took us about a year and a half I think to produce what we considered a stable first draft. The first one then was published was an RFC. And it was a roaring success almost overnight. One reason, I think, was that the whole time I was writing the draft, I had a program called Metamail that I had written and I was updating it to follow each version of the draft. And what Metamail did was it allowed you to make a very simple patch to existing text-only mail systems and it would call programs that could display the MIME parts. So all of a sudden these ancient systems with a very little amount of work started working with this cool stuff, languages, pictures. And so that made it, that made the adoption incredibly fast. In fact, we released the MIME standard and then I released the Metamail software, and I started getting patches for it. You know, people augmenting it, the next day. And within the first week, this was a Unix program, within the first week, I got the patches for the DOS port and the Amiga port. And at that point I was pretty sure I was on to something. [APPLAUSE] Art, can we pretty much trace what we’re doing today all the way back to MIME 1.0? It’s been very linear and it actually hasn’t evolved all that much, and I’m tempted to say that’s because we did a lot of things right But lest I sound arrogant, I also need to say that MIME is one of the ugliest things I’ve ever seen. It’s a tower of hacks. Honestly it is, and that’s one of the reasons it was successful. We had a tremendous problem, issue, of compatibility with the installed base. Even then there were probably millions of people using email around the world, and a lot of software that wasn’t going to be thrown out. And so, there were a lot of things that were constrained, and that led to MIME being ugly. But because it was designed with that in mind, it was ready for adoption. Another reason it didn’t have to evolve that much is that it was almost a version two of something else, in one sense, which is X.400. X.400 was the email standard from the OSI protocol stack, which was supposed to replace the whole TCP/IP stack, and never did. And a lot of us looked at X.400, which was tremendously complex, and would be very hard to integrate with existing mail systems. And so, our more simplified approach was a reaction to that. And that’s why you have some things in MIME that you would never design if you were designing from scratch. So for example, 80 column lines. The limit’s a little longer than that but basically limited line lengths. Why is that? It dates back to teletypes. Actually it goes further back than that. It goes back to punch card machines because the first video terminals had 80 columns so they could easily handle punch cards and this went on from them. So basically, your mail today is limited to 80 column lines more or less because punch card machines were 100 years ago. So we made this ugly thing to be backwards compatible and that’s why it took off like a rocket, I think. Another reason it took off like a rocket was that as we were finishing it up I got mail from a fellow who was working on this project in Switzerland I had never heard of. And he said I’m working on a project called the World Wide Web. And right now it’s text only and we’re looking for how to make it multimedia and we were thinking that MIME might be the way to do it. Do you have any opinions or ideas? And so having the Web pick up MIME did not hurt in having it take off like a rocket either. And what it really picks up is effectively the typing of various- Yeah. Really being able to mark in a common language between MIME and HTTP. Yeah, yeah. There are several parts of MIME and certainly they aren’t all used by the Web. At its core, the most important part of MIME, as you say, is the content type registry. So if I need to be able to send you some weird type of data that I’ve just invented, I have to label that somehow. And there’s a way to label it for experimental stuff but there’s a registry to go to for something that you want to be interoperable semi-permanently at least. And when we released the MIME standard, I think it had 17 content types it defined, something between 10 and 20. And last I looked, the IANA registry of MIME types was over 1,000, 1,200 or something. So that kind of extensibility did not hurt either, because a lot of people who had totally different motivations for what they wanted to do had an easy way to identify the data for the Web and for email.