An Interview with Internet Pioneer Radia Perlman

30/05/2019

Creator of the algorithm behind the Spanning-Tree Protocol (STP), an innovation developed in 1985 that made today’s Internet possible, American computer programmer and network engineer Radia Perlman does not consider herself to be the mother of the Internet, as many call her.

On the contrary, thirty-four years after an invention that is still used to surf the Internet, Perlman prefers to talk about developing “a small part” that contributed to the connection of multiple networks. In her own words, “I was in the right place at the right time.”

In her keynote presentation at LACNIC 31, Perlman noted that the STP was born contradicting the way of thinking at the time.

During her conversation with LACNIC News, she expressed her opinion against setting female participation quotas to reduce the gender gap in the digital world and observed that, in her opinion, it is a mistake to hold conferences or activities solely for women. “if you improve the environment for everybody, it will also be better for women,” she concluded.

We´ve read that in many places you are known as the “mother of Internet”. Why do you think that is?

I did happen to be at the right place, at the right time, and at the dawn of networking. And I did get to invent a lot of the very lower levels of the Internet. Somebody who was doing a magazine interview with me thought that it would be sort of cute to call me that because all sorts of men were saying “I invented the Internet”, and nobody was saying that any women had.

Actually, I would not have chosen to be called by that name, because no one person invented the Internet, there are so many different pieces of it. Furthermore, if I had not invented what I did, somebody else would have. I like to believe that, because it was me that did it, it´s simpler and cleaner, more usable and more robust than if somebody else had. But you can kind of make anything work. It is very flattering to be called that, but I myself would not claim that.

How did you come up with the idea of creating a protocol to allow computer networks to communicate with each other (STP)? 

Well, that particular thing was a very small piece of my career. The only reason we needed it was because people made a mistake and they thought the ethernet was the same as a network, when it was really only a link. Because they thought it was a network, they left out the piece of the networking software and the host that would put an “envelope” on the data in order to be able to send it across the network, because they were thinking that only nodes on the same ethernet would want to talk to each other. The ethernet, as originally invented, could only hold maybe a few hundred nodes and in a very small distance, so it was only a way of hooking together machines in one building. I said “don´t do that, the way you are building your application you still need to put this extra piece of software in there that would interact with a router,” which is by putting in the extra envelope that the router understands so it can be carried between links. But they ignored me, as it was usual at that point in my career. So they built these applications that were very good, but they would have been just as good if they had done it correctly, which would have meant putting in the extra envelope. So, years later, my manager proposed the problem to me: to figure out how to be able to move packets around even though the source and the destination don´t know anything about the extra header you need to carry packets across a network. They actually did have the basic concept of just move the packets around, meaning that you listen to every packet on every port, and you store it up, and then when the link technology allows you to transmit on another port, you forward that packet on to another port. And this is fine, but not if there´s any way for packets to get back to where they came, because you´ll never stop. In a protocol that is designed to be carried from link to link, you have an extra field that is called a hop count that changes every time it gets forwarded so eventually you´ll notice “wow, this has been forwarded too many times, there must be a loop” and you can drop the packet. But there was no ability to do that with ethernet because there was no such field in that header, not because the people who invented Ethernet didn´t know about it, it´s just that Ethernet was not intended to be carried from link to link. So it didn´t have that field.

So, my manager asked me to invent something that would not require changing the end nodes, and there were no spare fields in the Ethernet header, and there was a hard size limit. The funny story about that was that he asked me to do that on a Friday and then disappeared for a week. This was before people read email or had cellphones, so I had no way to reach him while he was on vacation. He furthermore made another ridiculous suggestion: he wanted the amount of memory to run the algorithm to not grow as the network grew, and nothing does that.

As it turns out, that night I realized I just knew how to do it, I could prove it worked and, furthermore, the amount of memory did not grow with the size of the network. I had never seen anything like that before. So I was all excited and I wrote the spec on Monday and Tuesday, and it was such a simple protocol and the spec was so complete that the implementers got it working very quickly without asking me a single question. So, that part was done but it was now Wednesday and I couldn´t concentrate on anything else because I needed to show off to my manager and he wasn´t around. So I spent the remainder of the week working on the poem, which is the abstract of the paper in which I published de algorithm.

I still thought it was the wrong thing to do, I wanted them to fix the computers attached to the Ethernet, to have the extra software in there to be able to put in the envelope to go through real routers that were designed to do that. But this was going to buy them a few months to get around to doing that.

Why do you think there are fewer women than men in ICT?

There are lots of ways to answer that. Other than being good at math and science and logic, I did not fit the stereotype that people think that an engineer is. This stereotype is actually very bad for the industry in lots of ways. The ideal team has people with different skills. Women think they wouldn´t be good at it, because they picture someone that is just diving around and building stuff, and that isn´t them. Or they think they wouldn´t like it because they picture an extreme introvert who is very uncomfortable talking with other people and just sits in his little cubicle and writes code. You don’t want a team that is just like that, you may want to have a few of them, but you also need people who can communicate, who are friendly enough to go ask questions, who have an artistic view of what can make the product exciting to users.

It´s a combination of women thinking that they wouldn´t like it because they don´t realize all the various dimensions of a career, women thinking they wouldn´t be good at it because they don’t happen to fit the stereotype.

And the person doing the hiring in a company thinks that the only type of engineer is someone who is just like themselves, they don´t realize how important these other skills are to the group.

This is perhaps a bit controversial, but a lot of things that the industry is trying to do to help, don’t really help. So, having conferences specifically for women in computer science definitely makes the stereotype that women are always getting special treatment, that all you have to do if you are a woman is walk up to any high-tech company and say “Hire me”.

I you have a conference that is only for women, that doesn’t help all that much, they can share each other’s consent about how you manage to have children and a career, but men have the same concerns.

Rather than focusing on what can we do for women, if you improve the environment for everybody, it will also be better for women.

Watch Radia Perlman´s presentation here