Command Line Heroes - Connecting the Dot-Com

Episode Date: March 23, 2021

The year is 1995. The internet starts going mainstream and the dot-com bubble begins its rapid inflation. But 10 years before all of this, a small team of systems administrators made a seemingly simple decision that would turn out to have a monumental impact on these events and would set the course of the internet for the foreseeable future. Dr. W. Joseph Campbell sets the stage for our season on the internet in 1995. Claire L. Evans explains how hard it was to find anything on the early internet. One team was charged with compiling that information in the early days of the ARPANET. Elizabeth “Jake” Feinler recounts being the internet’s sole librarian in those early days, and how she helped come up with the rules for future domain names. Paul Mockapetris describes designing the domain name system they later implemented as the internet went from a public network to a private business. And Ben Tarnoff explains the results of that increasingly privatized internet.If you want to read up on some of our research on the domain name system (DNS), you can check out all our bonus material over at redhat.com/commandlineheroes. Follow along with the episode transcript.

As we approach the turn of the century, communication methods are changing drastically. Internet technology uses computer networks to link people around the world. Millions of Americans have a personal computer in their home. These Americans can glimpse into the future online. In the year 2000, every business will be on the internet. 1995. That year, the world's imagination was sparked. A networking technology that few understood

had been developing for decades. And then it burst into the public sphere. Movies like The Net and Hackers and Johnny Mnemonic captured our new obsession. An arcane series of virtual tubes called the Internet had been brought to life by the creation of the World Wide Web. And tens of millions of people were jumping online. I'm Saranya Dbarik, and this is Season 7 of Command Line Heroes, an original podcast from Red Hat. This season, we're looking at that year, 1995, from every angle.

We're exploring how that moment in our tech history gave birth to the online world we know today. From e-commerce to web design to search engines and so much more. It certainly was a watershed year in the popular emergence of the internet. Professor W. Joseph Campbell is the author of 1995, The Year the Future Began. It was a moment when the internet goes from obscurity to near ubiquity. There was a critical mass of users, there was a critical mass of content, and there was a relatively easy way to get there. And that way to get there was the Netscape browser. And it really just took the web and the internet from obscurity to

prominence during that 12-month period. In fact, Netscape's IPO on August 9th of that year made it a multi-billion dollar success overnight. The IPO of Netscape really ignited a great deal of interest and made it very clear to lots of people that there was a lot of money that could be made online. The dot-com bubble was born. While moviegoers were lining up to see Sandra Bullock in The Net, investors were lining up to buy stocks in anything that had a dot-com attached to its name. The tech-dominated NASDAQ composite index would quadruple over the next five years.

That dot-com bubble transformed the tech landscape. And yet, few investors had any idea what those words dot-com really meant. So we're launching our new season with a little history lesson for those mid-90s traders. It's the story of the invention of dot-com itself. Before the dot-com boom, long before you could hop on GoDaddy to grab yourself a domain name, there was a woman you'd have to call the keeper of all domains. Her name was Elizabeth Jake Feinler. Back in 1972, the newly built ARPANET,

godfather of our internet, consisted of about 30 computers. Three, zero. And one of the major nodes of that fledgling network was at Stanford Research Institute. That's where information scientist Elizabeth Jake Feinler worked. And people like her, wizards of information management, indexing, and data organization were about to become incredibly useful because networks require order as they grow, or else they collapse. Douglas Engelbart, the engineer who became famous for his work on human-computer interactions,

was then running the Stanford lab. And he enlisted Feindler, giving her the task of writing a resource handbook for the first demo of the ARPANET. Basically, that was a contact list of technical liaisons and administrative liaisons for all the host sites. Organizing information for a network of 30 computers was easy enough. This was, after all, just supposed to be a way for researchers to bounce information back and forth between universities.

Not a ton of pressure organizing something like that. But, and you're probably ahead of me here, that was about to change. Try to imagine a few dozen host sites at a few dozen universities, military bases, and research centers. Each one with different resources. The military sites were way more secure, of course. The university sites were often run by students. There was no consistent organization.

There was no search engine to any of this. Claire L. Evans is the author of Broadband, the untold story of the women who made the internet. In order for it to be even remotely useful to anyone, you know, they had to know exactly what they were looking for, where that something was, when it was online, and who could give them access. And the only way they could really do that was by getting in touch with Jake Feinler and the Network Information Center,

who had all of that information on hand. That Network Information Center, also called the NIC, was run by the Stanford Research Institute. And that Jake Feinler is actually Elizabeth Feinler. Jake's a nickname she's had since she was a kid. Just two years after creating her resource handbook for the ARPANET, Feinler was the principal investigator planning and running the NIC. And like Evans just mentioned, the NIC was essentially serving as Google for the ARPANET.

They were the only people who knew where anything was. And that was okay at first. They had a job that got very complicated very, very quickly. So it went from, you know, a few dozen host sites to thousands in a matter of years. And Jake's project at the NIC went from being, you know, a two-person operation with one telephone and, you know, a file folder of index cards to like an $11 million military project with six phone lines ringing off the hook from 5 a.m. to midnight every day for 20 years.

With time, the flood of phone calls became a flood of emails, and Feindler's team was working hard to keep up. All of a sudden, it becomes a really complicated job to actually keep track of where all those host sites are and manually proofing that document, which was called the host table. It was something that was done twice a week. It was incredibly tedious kind of burnout job, and it became too big for some of the hosts on the network to even be able to download the file from the NIC regularly enough to have updated information. It became something that was kind of Sisyphean in nature. That host table Evans mentions, the original domain registry for the ARPANET,

was just a flat ASCII text file that folks could download directly from the NIC servers. It was literally a spreadsheet listing all the different hosts and their addresses. Yellow pages for the networked world. Soon, that rudimentary yellow pages became the original Whois, a dedicated server at the NIC where you could look up names and contact information for every authorized user on the network.

They were inching toward a new form of organization. I mean, she literally answered the phone for the internet and answered the email for the internet for a really long time. So she knew how explosive that growth was. And I think, you know, as an information scientist, as someone who came from library science, she knew that it was going to be something that could get really chaotic and impossible to manage if it wasn't structured correctly and early on with, you know,

thought in mind for how to preserve sort of core structures amid all this growth. And so, as these small steps were outpaced by the network's growth, it became increasingly clear. Big structural arrangements were going to be necessary, a way to move beyond the host table and the who is. And that would require an information scientist kind of imagination. Well, I always felt like Alice through the looking glass. It was just a whole different world. Feinler herself spoke with us from her home in California. She's now 90 years old. She still lights up when talking about her work on Douglas Engelbart's team. Doug's group, they're wearing, you know, Birkenstocks and jeans and no bras and,

you know, beards down to here. And because everybody at SRI was saying, what's with these people up on the second floor? You know, they were very strange for the time in the way they were dressed and the way they were acting. It was a lot of fun. It was more like a club in a way, the way everybody knew everybody and wanted to make this thing work

because we knew, we just had this feeling it was good stuff, you know. But just think what that club was demanding of Feinler and her team. They were providing the map, the manual, the address book that made the ARPANET usable. They were archiving and indexing its documentation, coordinating technical liaisons, managing the requests for comments. They were the hub and the call center for the whole enormous project. What our problem was, was running that database and, you know, upgrading it and doing the programming needed because nothing ran on the machines during the day. They were just so slow.

During the days, we were answering telephones and emails and people's questions and problems. And during the night, we were trying to run programs. The simple, manual approach of organizing all the network's users was growing more untenable by the day. Oh, it could be anything. They want a document. They want to know how to use something.

They want to know how to contact somebody. We had a hotline. So, you know, it was how do I get on the ARPANET? Who's at this site? How do I do this program? You know, where is this document? It didn't matter. That's anything you could think of a reference desk would get.

That's what we got. A yellow pages, a spreadsheet, a reference desk. However you want to describe the project, it was more than mere humans could manage. And that was the moment when Feinler truly organized those online crowds for the very first time. She didn't know it, but Feinler was preparing the internet for that moment Professor Campbell described. The moment in 1995 when the early internet would transform into an information superhighway.

It was all being prepared in the offices of Elizabeth Jake Feinler. So, that flat text file that indexed all domain names needed to be replaced. But with what? One computer scientist had a solution. Hi, I'm Paul Makapetras. Around 1982, Makapetras was asked to look at different ideas for replacing SRI's host table. He checked out five different proposals. I looked at those proposals and decided that something completely different would be a better idea. So I designed the domain name system,

and that's how it came to be. Macapetris is being a bit modest. The DNS delivered several advantages. For starters, it replaced all those numerical addresses with names that humans could remember. Navigating the internet no longer meant looking up a string of digits on some spreadsheet. By 1986, a few years after the DNS specs came out, the old host table had fallen to the

wayside. But that wasn't the only advantage to the new system. The really important things about the domain name system was that people could get their own domain and then manage the names under it. So that, for example, MIT came along and we gave them mit.edu. And whenever they wanted to change their network, they could do that. They didn't have to call up SRI and find out that it was only open from nine to five. They could manage their own space. That was one part of the magic was to let people separately manage their own names and addresses and so forth.

So the domain name system began to relieve what had become an impossible task for the people on Feindler's team. The second part of the magic was the system was set up so that it could itself tell a server at UCLA or anywhere else how to find the servers at MIT that had that information. Maka Petras had delivered an elegant form of automation. What he didn't know was that he was arranging a system that would define our lives. You know, there's a funny thing that runs through a lot of the Internet history, which is people talk about how they could have never done what they did if people had realized how important it was. Vince Cerf and Bob Kahn that invented the IP and TCP protocols say that in their Turing Award speech.

And people always ask me how I got the fabulous job of designing the DNS. And the answer was that the experts of the time, they thought it was a nice little problem for a recent graduate with a new PhD to work on. But it turned out to be a fundamental building block of internet that let us get to things like the web. That's a key point he just made. You can't imagine the World Wide Web that emerged in 1995 without something like the domain name system coming into place in the 80s.

It would be impossible to navigate. Most people would never bother trying. Preparing the ground for that new web experience required more than just the delivery of the DNS, though. Feinler's team had a revolutionary task of their own. That was how we got into naming the top-level domains. Feinler's team established.gov,.mil,.org, and the rest. But despite best intentions, choosing names for those top-level domains and deciding who gets what was sometimes fraught. Lots of large companies came knocking, saying they wanted top-level domains of their own.

But Feindler made the decision early on that the organization of the internet would be more neutral than that. I felt that the domains had to be generic and also the structure of the domains was going to be very different. Military structure was not going to be the same as the National Science Foundation. I mean, they just were different worlds. So the idea would be you came to the top level, you picked which one you wanted, and then after that, you agreed to whatever their terms for the structure were. Maka Petras remembers those conversations when that neutral sense of organization was anything but guaranteed. There were people that said that the top level domain should be phone companies

because they would naturally own the networks. There were people that said, the top-level domain should be phone companies because they would naturally own the networks. There were people that said, oh no, it should be network names, on and on and on. And I said, no, the names should be totally independent. But what we need to do is to think about having the country codes, which will allow us to placate all the different countries that want to do their own thing. And we should try some of these, what are now known as generic top-level domains like.com. At the time, Feinler worried that the U.S. government would decide, hey, DARPA isn't a standards body, so hand all these decisions over to the NBS, the National Bureau of Standards. But in the end, they were left alone, and the top level domain system took hold. Soon, as a kind of afterthought, Feinler realized they would need

another domain name, one for commercial entities. At the time, almost everyone online was associated with a government agency or university. So, making room for businesses didn't feel like a priority. We didn't know where they would land. They weren't quite org and they weren't quite government. And I was the one that said something like use BUS. Bus for business. But that sounded too close to pieces of hardware that had the same name. They went back and forth. Dot bus. Dot biz with a Z. And when a young system architect on her team, Ken Herenstein, was tasked with setting it up,

he chose dot com for commercial. He was the one that implemented the whole thing, so he just decided to make it dot com. And who knew? It was just an afterthought. It was that simple. Just a holding pen for certain contractors that didn't fit under.gov or.org. And just like that, one day in 1985, Dotcom was born. It would take 10 years for that seed to grow into maturity, but the ultimate year of internet optimism, 1995, was already taking shape.

Setting up the domain name system meant that, suddenly, all that organization and indexing was scalable. Naming responsibilities were dispersed to a number of servers across the network. The process wasn't centralized anymore. Different groups could manage and organize themselves as they pleased, all while remaining synchronized with the larger network. And that set the stage for some truly extraordinary growth. It also set the stage for a little formula that, today, we take for granted. Here's Claire L. Evans again. We still use the same addressing format that they came up with, which is, you know, user at host dot domain, right?

That's what email addresses look like now. User at host dot domain. That basic address grammar that we use every day to navigate the internet. Suddenly, we weren't a chaotic crowd manually organized by a team at Stanford. Suddenly, we were a self-sustaining system. A system not just of email addresses, but web addresses too. That explosive communications medium quality of the internet was not anticipated by its builders. And it was something that, you know, the administrators

of the internet had to kind of hustle to make sure it wasn't going to destroy the system that they had built. We know the internet didn't get destroyed. And yet, the process of managing all that growth was about to change once again. For a few years there, the new DNS registry could happily distribute IP address blocks and domain names. But eventually, even that level of automation was more than the U.S. government wanted to handle. In 1993, just as Internet use was really taken off, the National Science Foundation awarded a five-year contract to a company called Network Solutions. The task of registering about 7,500 addresses each year would now be theirs.

Of course, those numbers were growing fast. Two years later, Network Solutions processed 145,000 new addresses. Meanwhile, millions of users were flooding online. No single entity was going to be able to handle such exponential growth. And as the process of indexing, organizing, managing the internet grew larger and larger with each iteration, the fundamental nature of the internet's organization began to change too. I think the spirit of the early internet was about this kind of convivial sharing of resources, the collective building of something larger than the sum of its parts.

It was really mostly used and inhabited by its builders. So there was this spirit of kind of collective enterprise. As the user base grew, though, the potential for profits was growing, too. All that tech was about to be monetized. Soon, Network Solutions was charging 100 bucks for two years of registration. Meanwhile, in the early 90s, the National Science Foundation handed operation of its backbone to a consortium of Michigan universities that were working with companies like IBM and MCI. But we'll get into that story in our next episode. For now, what you need to know is that the NSF

believed that the private sector was a necessary partner if the internet was ever going to grow to its full potential. Ben Tarnoff, author of the book Internet for the People, told us about that moment when the NSF was overseeing privatization. He describes the bind that Stephen Wolf, the foundation's director, was in. The key problems that he was looking at was primarily a capacity problem. we going to find the money to continue to upgrade our backbone, to keep pace with the type of infrastructure investments that we will need to make in order to accommodate all of this demand as more and more people want to use the internet? And how are we going to broaden out to serve not just the academic community, not just the researcher community, but the wider public? As non-commercial as the internet may have been, there was general

agreement that privatization was the only way forward. Indeed, the political climate at the time in the early to mid-90s, which is an intensely deregulatory political climate, there were very few opposition figures who sounded the alarm or raised criticisms to these moves, at least in the mainstream political sphere. Government would continue to give away and delegate functions to the private sector. And that process has been rolling out for decades. The handoff hasn't only moved toward companies, though. Nonprofits have taken over from the government, too. So there's a nonprofit called ICANN, the Internet Corporation for Assigned Names and Numbers,

which, among other things, governs the domain name system. And they cut an arrangement in the late 90s, when ICANN is created, with the Commerce Department, which at the time still controls the domain name system nominally. But they reach an agreement where ICANN essentially will run it on the Commerce Department's behalf. And in fact, that arrangement was in place until 2016, at which point ICANN formally took sole control over the domain name system. The withdrawal of government and academic management, the withdrawal of people like Feinler, felt inevitable. But Tarnoff feels it may have been more ideological in the end. There was nothing inevitable, nothing technically mandated

about the path that the internet took. It could have taken a different path, and it could still. So I think the privatization of the internet was absolutely an ideological choice. The people who were involved in engineering it were not primarily technical. They were politicians, bureaucrats, policymakers who were participating in this broad ideological consensus around neoliberalism, around deregulation, around the need for the private sector to lead. When the people organizing and indexing the internet are no longer library scientists like Feinler, that changes the rules. For example, compare Feinler's decisions about

top-level domains with this more recent decision. ICANN has released these new top-level domains, and they released one, for instance, called.sucks. predatory for releasing this domain because every brand, every corporation in the world has to go out and buy their.sucks domain. The way we organize and manage the internet matters. And when you remember the pioneers who organized it in the first place, the way their decisions live with us still, you can't help but wonder how our decisions are organizing the future right now. When you follow the origin story of top-level domains, you eventually get a story about not just organization, but privatization.

And nowhere is that more obvious than in that crazy moment in 1995 when the dot-com bubble began. You had a capitalist feeding frenzy epitomized by two simple words that a public institution had set up. Dot. Com. And yet, people like Feindler,

who managed and organized the online arena, never got rich and famous. All those investors back in 1995 probably thought they were pretty smart for picking up any stock with the dot com in its name. And you can bet that when Network Solutions sold to VeriSign for $21 billion, they were pretty pleased with themselves, too. But the people who really gave dot-com its value were heroes like Jake Feinler, Paul Makopetras, and Ken Herenstein. You can check out loads of bonus material about Feindler and the invention of top-level domains

over at redhat.com slash commandlineheroes. Next time, we zoom in on that day in 1995 when the NSF net was shuttered and the internet's privatization took a giant leap forward. I'm Saran Yitbarek, and this is Command Line Heroes, an original podcast from Red Hat. Keep on coding. Hi, I'm Mike Ferris, Chief Strategy Officer. I've been a Red Hatter for about 25 years. And before your episode starts, I want to talk a bit about AI. The hot topic right now is

foundation models. And those are important, but at Red Hat, we see them as just a piece of the larger AI infrastructure. And here's what I mean by that. Enterprises are built of hundreds or even thousands of applications. It's not hard to imagine a future in which those applications are being served by hundreds or thousands of models. Without a common platform for your data scientists and developers, without a way to simplify some really complex workflows as you train, tune, serve, and monitor models, it can get overwhelming pretty quickly. And that's why we've built Red Hat OpenShift AI, a platform where everyone is working together on the same page to build and deploy AI models and applications

with transparency and control. Find out how at redhat.com.