Brad Templeton Home
The History of ClariNet.com
The Rise and Fall of the earliest "dot-com"
June 8, 2009 saw the 20th anniversary of the public launch of an electronic news publishing company I built named ClariNet. It seems odd to think of just 20 years as distant history, but the story of the internet is so accelerated that the world has changed in such a short time.
ClariNet has a claim on being the first "dot-com." Of course, how you judge that depends on your definition of what a dot-com company is, and there are of course other definitions and other companies with valid claims.
ClariNet still exists today, though only as a service of Yellowbrix. Some of the raft of companies created to use the internet as a platform still thrive, but many are gone or completely changed. Many of them have fascinating stories, this story is simply one of the oldest. This is mostly the story of the earliest times, my personal journey and some of the pre-history, and not of the boom and bust.
In here you'll also find some interesting stand-alone stories, such as building the most reliable computer system possible and early adventures in eBook publishing and how to trick and treat your staff at the same time.
The pre-history of the dot-com
I'm defining a dot-com as a company born to use the internet as its platform for business, either to sell primarily over it or to deliver over it.(see notes) However, before such companies could exist, there had to be companies selling that platform, what we call ISPs or the first of the "dot-nets". (It's now ancient history, but when we first laid it out, the .net domain was meant to be for network service providers.)
In particular, a dot-com, as I lay it out, is not simply a company doing some business using the net -- that activity goes back to the 70s. The many companies "born for the net" are of some signifigance to history. It's fair to say that for many people, the 1990s are thought of as "the dot-com decade" and within high tech, the early 2000s (the "noughties") are thought of as the "dot-com bust" period.
The company with the earliest "internet pioneer" claim, by far, is Bolt, Beranek and Newman or BBN. BBN was the contractor hired by ARPA to build the first Arpanet routers (IMPs) and networks. Many other companies would follow, though Digital Equipment Corp. (DEC) became the leading vendors of the computers that would be connected to the Arpanet.
BBN was bought by GTE, which later became Verizon, and more recently was spun out as its own company again. The domain bbn.com was the second one registered in the ".com" top level domain, symbolics.com being the first -- but this article isn't about the top level domain, but rather the world of internet based business.
While BBN's claim is quite solid, it was not born for the net. The other company which can lay claim to being the first "dot-net" company is UUNet. UUNet was formed in 1987 as a non-profit initially funded by USENIX, the Unix Users association. Rick Adams, who ran a major USENET node called seismo, wanted to build a company dedicated to being a network hub for USENET and UUCP and connecting those worlds to the Arpanet/Internet.
(I myself had wanted to do the same thing, and even debated the numbers with Rick, but he was smarter and actually pushed forward with it to great success.)
The cross-country backbone of the internet at the time was the NSFNet, funded, as the name suggests, by the National Science Foundation. The NSFNet had an "official use policy" that seemingly blocked commercial use.
At the same time, a number of non-commercial ISPs provided Arpanet/Internet connections in their regions. These tended to be formed as regional consortia of major internet users who would build one or more hubs so each university and internet using company in the area could connect. These hubs in turn were connected to the NSFNet.
UUNet was immediately popular. As interest in the internet grew, UUNet wanted to become the first commercial ISP. With the use of venture money, they formed a commercial venture which bought the business of the non-profit and redirected that shell to doing more non-commercial activity. The new for-profit version began offering internet connections for money.
The non-profit regionals were doing the same thing, and in fact in the coming years some of them would also turn commercial.
Other early ISPs included BBS like systems which prominently featured connectivity to USENET or the Arpanet as a selling point. These included John Little's "Portal" (Later a successful provider of ISP billing software, now a unit of Oracle), Barry Shein's "The World" and a variety of free sites that called themselves "public access unix" systems. These were effectively a Unix-based BBS with USENET access like all the other systems on USENET. Later they moved into dial-up internet.
Also worth noting in the early history of internet commercialization are the gateways that were made from the older commercial online systems and E-mail systems, including Compuserve, MCI-Mail and ATTMail. In some cases, big online systems like the news databases (Dialog and Mead) allowed telnet access as an alternative to dialup or X.25 networks.
Some might suggest that some underground businesses may have existed earlier, flying under the radar on the acceptable use policy, or just being illegal. One could argue that the various proto-spammers who posted ads on the net were trying to do an internet-based business, but I know of none who created companies for that purpose, and would be interested in reports.
Pre-History of ClariNet
The story of ClariNet begins with my own first encounter with the ARPAnet in 1979. (If you would prefer to get to the founding and not what brought it about, you can skip ahead.) Just a teen-ager, I had found work as the first employee of a new software company named Personal Software Inc. Personal Software (you know it by its later more famous name VisiCorp) aimed to sell applications software to the growing microcomputer market, in particular the Apple II, PET and TRS-80. Personal Software's first big product had been Microchess, a capable chess program that started out on the 1-kilobyte Kim 1 and moved to the Commodore Pet and Apple.
Microchess had been written by Peter R. Jennings (not the newscaster) who had a small software company named MicroWare in Toronto, where I lived. I met Peter through the late, great Jim Butterfield and made a deal for Personal Software to publish a game for the Pet I had written, and they asked me to come work for salary, first for MicroWare, and then Personal Software, the firm Peter founded together with Dan Fylstra. Dan was living in Cambridge, MA, and had met and was working with Bob Frankston and Dan Bricklin who had written Visicalc -- the first spreadsheet and the program that would effectively create the microcomputer applications business.
I was sent down to Cambridge to visit Frankston and Bricklin and help them port VisiCalc to the Commodore, which was my big area of expertise. That in itself was heady stuff for a teen-ager, but the real life-redirecting moment came when Bob gave me an Arpanet account at MIT so we could interact by E-mail about the work on VisiCalc.
2 years Later, during a summer at Personal Software -- now in California and having changed its name to VisiCorp -- I was given a more capable account in order to do software development on VisiPlot (a companion program) for the IBM-PC. The PC had not been released -- indeed I was not even permitted to know what machine I was coding for -- so I developed in C on a machine on the east coast, using an X.25 network known as Tymnet. (X.25 was a packet switched network primarily used for long distance computer terminal connections, with command line interfaces.)
On this machine I encountered USENET.
USENET is and was a multi-network bulletin board system (initially for Unix computers) which used the Arpanet and a modem-based network based on a program called UUCP. On the Arpanet, most online community was in mailing lists. The USENET software formalized this, and eventually became the seat of community for the network world, including the internet. USENET software divides discussions into topic areas called newsgroups, themselves in a hierarchy.
Most importantly, it's a bit like e-Mail. You subscribe to a newsgroup and you have a concept of messages that are "new" -- not yet seen by you, or unread in E-mail parlance -- in your areas of interest. There are also messages that you have seen or skipped, which you can access but are not normally shown. It was inspired by programs that existed to provide "news" to users of a timesharing system. They wanted to see what was new for them, and not what they had already read. (Today RSS feeds are more commonly used for this purpose.)
The Arpanet supported primarily E-mail, file transfer and remote terminal sessions. USENET created community, and was a fundamental step in the internet revolution. Bulletin Boards (BBSs) predated USENET by a year, but much of the action in online community in the 80s took place on USENET.
I was so intrigued by USENET (which also carried the older Arpanet mailing lists) that I worked to connect it to the University of Waterloo, in Ontario, where I was an undergrad during the regular school year. I started by downloading discussions from USENET and copying them up to the computers at the university using the X.25 networks.
When I returned, I pushed to get the university's computer systems themselves on USENET. I was working for the computer center part time, and due to the generosity of Digital Equipment (DEC) -- which paid the international phone bills in the interest of growing the network -- I was able to arrange a connection. Henry Spencer at the University of Toronto did the same thing, at first dialing in manually and later getting the same calls from DEC.
These links were the first international links on the network. Jim Ellis, one of the developers of the original USENET software, remarked to me that he was particularly proud to see it go international. It was the first hint at the network's power to change the world.
At Personal Software I also was lucky enough to meet with a number of people who would also change my life and play a role in the dawn of the PC and internet age. They included by 2nd boss there, Mitch Kapor (who would found Lotus and the EFF -- which I currently chair), Roy Folk (who would be GM of Ashton-Tate and later become my COO at ClariNet and eventually buy it), Lynn Brock (who would also run ClariNet after I sold it) and many others.
(Other famous fellow alumni from that era include Ed Esber, who would run Ashton-Tate and Creative Labs, Rich Melmon who co-founded Electronic Arts and Bill Coleman who would found BEA)
The Network Grows
I became immersed in USENET and to a lesser degree the Arpanet upon my return to Canada. I built a more traditional software company in Waterloo, doing programming tools for education, and later spreadsheet enhancement tools (that would be sold by Dan Fylstra, founder of Personal Software/VisiCorp.) This was named "Looking Glass Software," linked to its first product, an integrated programming environment for education about software called ALICE Pascal.
At the same time I spent more and more time online on the net, a story familiar to many of you. I began to make many friends there. I wrote a variety of free software tools to help make USENET run better, and participated in the discussions that "ran" the net. I put "ran" in quotes because it's hard to say just how the net was run in those days. It was a co-op of willing sites that exchanged messages, but most site administrators just went with the flow. The hottest topic was always what new newsgroups to create. It generated immense amounts of discussion, almost always more heat than light.
We often felt as the net kept growing exponentially that it would surely collapse under its own weight. The people paying the bills would realize that all this money was going for 90% chit-chat and 10% useful technical discussion, and pull the plug. It never happened.
An early proposed business was one, oddly, that tried to bypass the internet and UUCP, partly for reasons of cost, and partly to get around commercial use restrictions I will talk about later. This business, known as "Stargate," delivered a USENET feed via satellite. Even though satellite transmission was more efficient that use of the phone, it was still quite expensive at the time, so Stargate sought to send only a limited set of more technically useful newsgroups. Stargate never moved beyond the experimental phase, but also deserves a place in the history of early net business.
Like many, I was spending too much time on the net. While I had made my living writing software, I decided I either had to stop using the net so much, or start making my living out of it.
I had been foiled in my plans for starting an ISP, largely because Canada was not the place to do it. But I was convinced I should try to start a business on the net itself.
I would choose publishing both because it was obvious, and due to a project of mine which had grown quite popular. It should not be too surprising to learn that from the earliest days of the net, one of the most popular applications was sending around jokes. It was also one of the most controversial ones.
An early USENET group was devoted to jokes named net.jokes. In the mid-80s, the groups were renamed and it became "rec.humor." Rec.humor was a very active group, but much of the traffic was not jokes, but people commenting on them and complaining about them.
Some years earlier, a feature called a "moderated" newsgroup had been added to USENET. Normal newsgroups can be posted to by anybody. Moderated groups required all postings to be approved by a moderator. There was no security on this, it was just a simple technical and social convention. Today, completely open discussion areas have been seriously hurt by spam. At a minimum they have parties who are empowered to delete the spam.
One of the first moderated newsgroups also deserves the title of the world's first "blog." This was named mod.ber, and it was managed by Brian Redman of Digital Equipment Corp. Brian and his associates combed USENET for interesting things, and then posted pointers to them in the moderated newsgroup. You could read mod.ber in 1984 for a "best of the net." This should remind you of today's HTTP based "blogs" and it probably qualifies as the first.
I enjoyed reading the jokes, but like most people was frustrated with the noise that went along with them. Always proud of my sense of humour, I took the steps to create a moderated newsgroup devoted to jokes. Most moderated newsgroups of the day had a moderator who simply tried to keep down the off-topic noise and flamewars. My plan was more like that of an edited magazine -- to judge on quality and only accept the jokes I felt were funniest. I started by culling the original rec.humor newsgroup, but eventually people were submitting directly to me. The newsgroup was built by taking only 1/50th of what I saw, it wasn't hard for me to produce something that had a vastly higher quality level than what was on the unmoderated net. The name was sort of a joke, it being the funny subset of rec.humor.
Thanks to that, it quickly became popular, shooting up the readership charts fairly quickly. What made it #1, however, was getting banned.
About a year into the life of RHF, a random accident with a politically incorrect joke offended an MIT student immensely. He began a campaign to shut down the newsgroup. While largely laughed at by the USENET community, he found some sympathetic press willing to publish reports about the scandalous things these brand-new computer networks were really for. Front page stories appeared in the local newspaper where I lived, pushing the University of Waterloo, which was my gateway to the network, to cut me off. To the surprise of many, they did. Some time later, fear of political incorrectness led Stanford to also ban the newsgroup. In both places, and on the net, the bans generated much more protest than the original complaints. Over time, they were reversed.
However, as is often the case, banning something gives it even more attention, and RHF soared quickly to be the most widely read newsgroup on USENET, and thus the most widely read online publication in the world. It probably retained that title until Yahoo came along to unseat it.
I've written up the history of these events in a few other articles:
RHF and other net activity started diverting a lot of my time away from LGS, my software business. In the late 80s, my focus was on spreadsheet tools. Once again Dan Fylstra, founder of Personal Software, was publishing my tools through a company he had set up for these purposes. We made a deal with Intel to bundle my spreadsheet compiler with the 8087 math coprocessor line, which provided a small cash infusion that would be the money to start ClariNet.
The net was taking so much of my time that I resolved I would either have to find a way to make it pay or get off it so I could focus on what did pay. The problem was that the Internet backbone was the government funded NSFNet run by the National Science Foundation. The NSF had laid down a rule -- the backbone was not for commercial use. Only use in support of research or education was permitted. Selling net access to schools and labs as UUNet did was great, but using it for business was verboten.
As a publisher of comedy, I greatly admired the writings of syndicated columnist Dave Barry. What Barry wrote resonated with the computer geeks on the net, including me. He was so popular that people were typing in his column every week and posting it to the net or mailing it to friends. This was a pretty straightforward copyright violation, so net postings were discouraged, but the mailing lists grew to thousands of readers. (Understand, that was a big number in those days.)
I wondered if it might be possible to legitimize this, and go to the syndicate and get permission to publish the column just like a newspaper. Already some syndicated columns were appearing online on commercial online services like Compuserve.
I called Tribune Media Services, which handled Dave Barry and got nowhere. They largely were not very aware of the internet at the time, and didn't know how to deal with my request. It was unknown enough to not inspire them to take any risk, and I certainly wasn't big enough to offer them the money to make it worthwhile. So I let Dave Barry alone for a while.
In spite of being foiled in the comedy department, the idea of doing a business on the internet continued to intrigue me. The online services had online news services drawn from wireservice material. It seemed the obvious application for the internet might be a full online newspaper.
In 1988, a small Virginia company named Comtex provided the news services for Compuserve and some other online services. They had developed the software for the spook market, but they also aggregated news from a selection of newswires, notably UPI and the press release wires, and put it into a mixture for Compuserve, GEnie and others. I approached them for a deal to do it over the internet. Comtex's base was actually providing news and news classification software to the intelligence community. They used stripped down versions of their spook software to try to figure what stories were about.
I made a rough deal with Comtex and started to build the software to convert their news for reading in USENET newsgroup format. The USENET software had been designed as a tool to deliver computer system news to timesharing users, so it actually fit the bill well. USENET software lets you subscribe to a series of topic areas, and keeps track of what you have seen and not seen -- like read and unread e-mail. Articles can appear under more than one topic area (newsgroup) but you only see them once even if you subscribe to all the topic areas. It's a good fit.
In addition, USENET, designed as it was for the disconnected UUCP network, pre-feeds the news you will read to a local machine, and you then read it directly there, or over a local ethernet. The access is thus blindingly fast, even in comparison to the web over high speed DSL 30 years later.
Making the software was one thing, but the NSFNet "no commercial use" rule was another, until I found a way around it. The NSF net backbone was for use "in support of research and education" and so I put forward the position that if I sold a subscription to something like Stanford University, and they asked that it be delivered over the internet (which they were certainly entitled to use since everything they did was in support of education) then it fit within the rules.
To confirm this, I called Steve Wolff. Steve was the director of NSFNet, and if there was ever a person who was close to "in charge of the internet" it was him. It was reasonably easy to win him to my interpretation. The challenge was on.
For customers who were not schools or labs, I would need to find another way to reach them. Many of them had UUCP links with UUNet, which was not government funded. In addition, the way USENET worked provided another bypass. Also, a few sites were also starting to get direct fully commercial internet links to UUNet and the other nascent ISPs.
On USENET, sites with USENET feeds would pass their feeds on to other sites. Most links took place locally, and were cheap or free. Long distance links cost money, though some companies could easily afford them. USENET was designed to allow each site to connect to lots of other sites redundantly, while never sending the same article more than once to stay efficient.
It turned out that if one site in a city legitimately got my feed over the internet (like the example of Stanford) they could legally pass the feed to somebody local without using the NSFNet. With just a few customers in the big areas, I could legitimately sell to most prospects.
The deals seemed to be in place, and the basic software was working, so on June 8 of 1989, with a Comtex based system in beta testing, I announced to the world the new ClariNet service. (This is a link to Google's archive of my post. There was some controversy over the idea of a commercial service over the internet (you will be surprised how negative some posts are,) but I was able to convince most folks I had dotted my "i"s. The next step was to get those who said they were interested to place orders.
I use Stanford in my examples above because they were in fact our first customer. However, before I would deliver the first articles to them, Comtex changed the terms of our deal to ones too expensive for me to afford. I had been premature in my announcement, which also legitimately upset them. I started to look for something else while trying to knock down their price.
Most of Comtex's material came from UPI, because Comtex had bought a 20 year unlimited licence to that material during UPI's hard times in the early 80s. However, I contacted UPI's Toronto office to see if I could cut a better deal directly.
UPI, once one of the world's great news organizations, was starting to fall on hard times. Recent upheaval there, along with significant latitude given to Mike Barrett, the director of their Toronto office, made them much more flexible and willing to try new things than your typical news provider.
Mike was willing to cut a deal that ramped up my minimum royalties over time. In the early months, when I had few customers and could not afford to pay large fees out of my own pocket, I would pay very little. Even after things ramped up, my minimum royalties would only be $2,000 per month. This I could handle, and it was much better than Comtex's deal.
In addition, UPI had started an experiment where their reporters coded topic categories on each story. Comtex was throwing these away for their less accurate automated categorizations. If I could add other services to UPI, I could assemble my own great service without Comtex, for less money.
I knew my audience would be computer people, so I sought out computer news. Wendy Woods had built a small computer-oriented newswire called Newsbytes that she was selling to online services and BBSs. I added them to the package, and combined all the technology coverage with UPI to make a sub-product called TechWire.
To round out the service, I went in search of syndicated columns. I called Tribune Media Services, who 6 months prior had not known what to do with my request to publish Dave Barry.
"Let me connect you with our VP for online sales..." said the receptionist. This was quite a change.
Around that time, IBM and Sears had decided to enter the online world with a dial-up online service called Prodigy. Prodigy was intended to be highly polished and content focussed, instead of a more open forum like Compuserve. They decided to spend a lot to licence premium content. They had walked into Tribune, and asked if they could get star sportscaster Howard Cosell to write a custom column for them. The Tribune folks laughed and said, "Howard Cosell won't light his cigar for you for less than $10,000." The Prodigy folks said yes without blinking, thanks to their rich mommy and daddy. Thus a division to sell to online services was created.
I couldn't pay like Prodigy, and they realized it, but now they were willing to treat me like a newspaper. I would report my circulation and pay the fees a newspaper of that size would pay to run the column. Those fees are actually quite reasonable, so I was able to add a half-dozen prime syndicated features.
(Later I would even add comics, most famously bringing Dilbert to the internet. And then losing it, like Dave Barry, because they wanted to put it up free on their own web site.)
Writing software to handle the UPI feed turned out to be an interesting challenge. While they were adding the codes to categorize stories, they often made mistakes. The rest of the material was often free-format, not meant to be handled by a computer. It turned out a large part of the value we would be adding involved writing clever software to make that transformation. It was necessary that the software be easy to change, as the habits and styles of the wire stories changed every time they got a new reporter at some bureau.
One amusing example: The category codes were entered by the reporters with a 5 letter cryptic string, with 3 letters for category and 2 for location. For a long time "xx" was used as a code for "unknown" but the actual code-book declared "xx" to be the code for sex. So a lot of stories would go out under our sex topic that had absolutely nothing to do with it until we built heuristics to tell if the reporter really meant sex with their xx code. (The sex newsgroup was, for some reason, one of the most popular.)
Somehow I built all this code while still running the spreadsheet software business, trying to do sales and marketing and manage my small staff (who were mostly working on the other software tools not related to ClariNet.) However, soon, due to good fortune, ClariNet activities would take over and pay for themselves.
By September of 1989 we delivered our first news to Stanford. We cut them a good deal, and it was a wise choice -- 20 years later, Stanford is still a customer in spite of an effort to cancel it by bean-counters that was countered by ClariNet fans. Stanford could also feed customers in Silicon Valley for us without using NSFNet.
To encourage that sort of activity, we provided a discount to anybody who fed other customers. Unlike regular USENET, feeds could not be given to just anybody, but when you did give a feed, it got you a discount. It didn't really cause anybody to go out and try multi-level-marketing, since in the USENET community feeds were something you did happily as a favour, but it did help justify the activity on internal budgets.
Not long after I announced ClariNet, Geoff Goodfellow, who had been on the internet and USENET for a very long time, announced plans for another USENET based business named "In Moderation Network." IMN planned to sell a series of moderated USENET groups as a service, where their moderators would make existing discussion groups more manageable. While IMN never fully got off the ground, Geoff later had a bit more success (though again too early) with his RadioMail product, which offered internet E-mail to handheld devices over pager networks. My Waterloo classmates at RIM would later make a grand fortune doing just this when the time was right.
Another early product of ClariNet's that was premature in the market was the Street Price Report, produced for us exclusively by small outfit called Consumer's Database. They had people type in all the prices on technology products from the ads in the thick technology magazines like Computer Shopper and PC Magazine. You could then look through the lists and find the stores with the best prices on what you had. Later, online shopping-bot spiders which did this on the web would become a huge business. I recall one of the first ones selling for $35 million for no reason during the true dot-com craze.
Our prime marketing technique, which would become the staple of internet marketing, was to give some of the product away for free. (Nobody realized that eventually the rule would be to give all of the product away for free, then make it up in volume.) First of all, we offered a free trial to any site. The same way I had snuck USENET into the University of Waterloo, I knew a free trial would get people hooked, and it would be hard for the purse-string holders to take it away.
Secondly we created a free newsgroup called biz.clarinet.sample. In this newsgroup we provided one of our real newsfeeds on a given topic, changing the topic every week or so. Readers would stumble upon it, and if they were interested in the topic, get hooked.
We never bought much in the way of traditional advertising. There weren't really any magazines in which to advertise. The main marketing and sales focus was trade shows, notably Unix oriented trade shows like USENIX and Uniforum. Later, we would exhibit at networking oriented shows like Interop, and eventually Internet World the internet started to get public attention.
The USENIX community was our home. The shows were cheap to exhibit at, and most of the system admins who ran large USENET and Internet sites would go. People knew me, and soon new the company. I didn't directly use rec.humor.funny to market ClariNet, just a few subtle mentions, until much later when ClariNet hosted the popular web site for rec.humor.funny as comedy.clari.net.
At USENIX and other conferences, I hit upon a nice gimmick that was unusually successful. I purchased a button maker and press, and wrote some software that would let you type in any phrase and it would typeset it as a custom button, printed on a laser printer. We would give you a trade show button with our logo and number but your phrase. People loved to wear those. Most people weren't actually that creative, but we used all the best phrases, plus many I thought of myself, for those who just wanted a pre-created funny button. Of course we could and did make some of those up in advance, but often the most popular phrase would make fun of some event which had happened right there at the conference, perhaps in the keynote address. The topical nature of this reminded people of our product, and everybody would come to our booth to get one.
Growth and Travel
By the standards of later internet companies, ClariNet would start small, with about $100,000 of sales in the first year. However, unlike my past forays in the packaged software field, I quickly came to love the value of a recurring revenue business. Once I had sold a customer, if I kept them happy, I kept getting a cheque every month. Many customers happily prepaid a year in advance. New sales added not just to my bottom line for that month but every month. My churn rate was remarkably low. ClariNet would double in sales every year of its life until it was sold. The first year we lost a little bit of money, but by the 2nd year we were profitable and would be profitable until sold. (Our early start and decent growth made us the fastest growing internet company in the Inc 500 in 1995, and the 13th fastest growing company on the Silicon Valley Business Journal's list for 2 years in a row. However the Inc 500 honour is somewhat misleading. There were faster growing companies but they had not been around long enough to qualify for the list.)
The company ran fairly leanly in the early days. Costs in Canada were low, and I had written most of the software. One other programmer (Tim Tyhurst) and one office Manager (Grant Robinson) combined with one admin (Paula Hendsbee) were able to put it all together.
Costs in Canada might have been low, but it also wasn't a good place to run an internet company. Internet connections were slow and not cheap, and UUCP links over the phone were also too expensive. Having worked in Silicon Valley 10 years prior for Personal Software/VisiCorp, I knew that California was "the place I oughta be."
So we moved. Of course, it was far from that simple and required some careful legal wrangling to transfer a business from one country to another. Fortunately my brother, Michael Templeton, is one of the best lawyers in Toronto at handling international deals and taxes, and his selfless assistance made the difference. The procedure was complex enough that we were in his office the day before we finally flew out, signing the papers. My lady companion and I parked her car the street next to his office tower. When we came out, there was an expensive parking ticket on it, even though it was Sunday. Amusingly, we realized we were flying out the country within 24 hours, and the car would be sold within a week. With a smile, we tossed the ticket with impunity.
The move also involved immigration work. I used a new Visa under the Canada-US free trade agreement, but Grant Robinson, who was my majordomo in all this for many years, could not get that status. After debating with lawyers what to do, I asked if he had any ideas. "Well," he said, "my mother grew up in the USA." That turned out to get him not just the ability to go to the US, but an immediate U.S. passport.
Programmer Tim Tyhurst would not be making the trip. We left the satellite dishes and news processing engines with him, continuing to feed customers as we set up the new office in Sunnyvale, California. There, we brought in a duplicate set of equipment arranged through UPI's U.S. offices, using a different satellite. The move was made in November of 1991.
The nature of USENET feeds made this easy to do. As I noted, USENET was designed to have redundant feeds. Every site could get the news from multiple sources, and if offered an article that had already been delivered, it would simply decline it. As such, we were able to build a completely redundant news processing room, and have each system -- the one in Silicon Valley and the one in Waterloo, Ontario -- independently feed the news to the customer. They would accept it from whoever offered it first. We were able to switch from one country to another without a hiccup, and in fact now had a system in which either city could be nuked and the news would continue to flow. (More on this later.)
Indeed, in the early days, the system was so automatic that when we attended our first trade show after the move (50 miles away in San Francisco) the entire company went, leaving an empty office. Fortunately our customers largely used e-Mail to reach us. Anybody who phoned would get voice mail. Should the systems fail, the Waterloo system would act as a backup until I could drive back to Sunnyvale to fix it. I didn't have to.
The company grew and I quickly added new employees, including Editor Aaron Priven, who would oversee the quality of the news flow (making it less automatic but better) and Chris Lynch, VP of Sales. Chris was the son of Dan Lynch, who had created the popular Interop trade shows (at which I exhibited.) Dan is a real internet pioneer, having worked on the early Arpanet. We jokingly said that, since Chris had pounded on his father's keyboards as a child, that he had 18 years of Internet experience.
How we priced
Pricing (and getting any revenue at all) was a major question for all of the dot-com era. Traditionally newspapers sold advertising, but it was too early in the game to even try that. I sought to get paid subscriptions. Later, if advertising developed as a market, having a large audience would readily give you something you could convert to advertising supported -- or so I planned.
However, with remote reading, counting readers directly wasn't possible. We arranged to pay our sources according to how many readers we had but were allowed (as is common in lots of areas of media) to do statistical estimates of the number.
Thus when we came to a site to give them a price, we did it based on the number of people on their network. That was a number they generally knew. We gave them a very good price per user because we knew only some portion of them would actually read our material, but they could give access to all. At some sites we were actually able to measure precise readership from logs of the news servers, and we used those to estimate the total.
Some sites wanted to see if they could just pay a fee per actual reader, and charge it to the reader, but we resisted this strongly. Just as cable companies know they make more money bundling a whole package of channels together for a flat price even though most people only watch a few of them, we knew if people had to pay a much higher per-user price it would cut down on usage. For a flat fee, the news was distributed to everybody, and more people used it, and thus it was harder for the site to cancel its entire subscription.
As the internet grew, our total number of "paid subscribers" got very high -- up to about 1.7 million. These were people for whom we were getting some monthly fee, even if, in the case of Netcom, it was only around a dime. We liked to promote that number since it is more than the daily circulation of any other U.S. newspaper except the Wall Street Journal and USA Today. (In the case of USA Today, a lot of the copies also go unread.) Our real number was still respectable, at about a 6th of that.
During my ownership, we didn't switch to advertising. There were some sites with fancy ad sales forces able to run ads and get high CPMs for them for a while, but that was transitory. Only when Google and Overture came along did a really sustainable ad model arrive. I don't know if it was a right decision or not to wait so long. From the standpoint of crazy market valuations, it was a bad idea, since advertising gave you this nebulous but large upside, as long as people believed in it. In addition, people won't admit it, but the idea that you can switch business models quickly may be false. The more you invest in a business model, the harder it is to turn the ship even if you know the time is right.
Not content to just build the first internet newspaper, I turned my attention to a longer-term vision of online book publishing and a search for a general way to sell content in 1992 and 1993.
You can read about the plan to create an all-you-can-eat flat price library of books (called the Library of Tomorrow) and the giant 1993 Hugo and Nebula Anthology, which was published on the internet and on CD-ROM in a sidebar.
Tim O'Reilly deserves a special mention here, and not just for the early Viola browser his company developed. Tim is primarily a book publisher, with the best line of technical books in the unix, linux and internet space. However, he was also keen on internet eBook publishing and started GNN (Global Network Navigator) as the web was dawning. It was the first serious commercial web site, with a plan to be advertising supported. He sold it to AOL for a good price at the time. If your definition of the first dot-com depends on it being HTML based, O'Reilly and Associates and GNN.com may well merit the title.
We began before there was an internet industry, and so used trade shows oriented to the Unix world and networking world. Interop, a show devoted to networking products, was the pioneer on the infrastructure side, but when it comes to using the internet as a platform, Alan Meckler's Internet World show opened the doors. The first one was a tiny affair in New York's Javitz center. I had exhibited in Javitz before, and at first I could not find the event -- it was in a small basement exhibit hall under the main hall. There were just two short aisles of booths, with companies like Netcom, and sellers of TCP/IP software for Windows computers. One booth was stacked 6 feet high with empty boxes of their "Internet in a Box" product -- which wasn't quite ready, so the joke was that the product was just the box.
An interesting crowd came, including some earlier era luminaries like Nolan Bushnell (whose daughter Alissa would later do P.R. for ClariNet, a true 2nd generation resident of silicon valley.)
What surprised everybody was the enthusiasm of the crowd. The people poured in, trying to find out what this "Internet" thing was, and grabbing all the literature they could find. Many companies ran out of literature within 30 minutes, and ended up calling to get more overnighted for the next day of the show. This pattern would repeat for the next 3 or 4 shows, even for companies that now knew to bring extra literature. It was a taste of how interest would explode. Here, by the way, is the text of my talk at the 2nd Internet World show.
Meckler had an Internet World magazine as well, which did very well for him and sold for a high price. For those who believe the real way to get rich in the gold rush is to sell picks and shovels, he picked a good course -- though a number would do even better.
Another important trade show for us represented the wave of the past. Jack Rickard, who had created a successful magazine for the BBS industry, started a conference for BBS operators and suppliers. We had imagined BBSs turning into internet access points, and buying our news services to attract customers. We got some success from this, though primarily because some BBSs turned into full fledged ISPs. Rickard himself realized the BBS was a dying concept, to be replaced by the web, and later renamed the conference ISPCon and sold it.
The UPI shutdown and "M5" reliability
When ClariNet first began, we had to tell information suppliers what the internet was. Some were afraid of it, and wouldn't participate. Others were excited and were willing to go the distance to make things happen.
UPI had worked well for us in our early years, but it was a dying organization. Once one of the great newswires, it had fallen on hard times and been through a series of owners. Eventually it would lose all its newspaper clients and be sold to the Saudis and eventually the Moonies. In 1992 it still was enough of a serious news organization, but to expand ClariNet it was time to try to get the big boys -- the Associated Press and Reuters. I entered initial contract talks -- they were going to be a lot more expensive, but we had grown and could afford it. However, unlike UPI, Reuters did no editorial classification of stories at our level of product, and AP did a fairly meager classification in their stripped-down AP Online product.
To thrive, we had to do more than just be a conduit for the news. We had to add good value and make it really usable on the computer. I began work on a system to automatically classify the news. Others had tried this before, but not with great success. I had a small boost in that we had a large body of already-classified stories from UPI to train the system.
All plans changed, however, in the fall of 1993. I had renewed my UPI contract several times by this point, usually with minor negotiations and without incident. In this case, I even had a trip to the DC area where they are headquartered and planned to see them on the trip, but the Sheraton I was staying at lost the phone message from them about setting up the meeting, so I went home assuming all was well.
To manage my systems, I had built a fancy warning tool. Stories come down the wire at random, and at night there can be gaps of 20 to 30 minutes between stories. You want to be alerted if there are any equipment failures (and such failures were frequent.) My system would have a sense of how long it should be between stories on various wires, and start an escalating series of alerts as the gap got longer, first on the screen, then phoning and paging. Such systems are common in system administration today.
One morning, the alert went off -- no news from UPI on our California boxes for several hours. As was usual in these events, I phoned UPI's support desk to find the source of the trouble. It took them a while to figure out that the reason for my failure was my account had been disabled. UPI had cut me off.
This was a cause for panic -- it could destroy my business -- but in fact the news was still flowing. Our servers in Canada got their UPI from an older satellite system that had been the main one when we first set up our contract. The Canadian office had not shut off our news, and it flowed through there and to our customers automatically. (As I described above, each system worked independently and customers were fed each article twice, once from each source. They simply refused the second offer.)
I have said that this represents the highest level of computer reliability possible. I call it "M5 Reliability." In a famous episode of Star Trek, called The Ultimate Computer, the Enterprise has a new advanced AI computer named M5 placed on board and in control. Of course, M5 goes wrong, and takes over the ship and tries to destroy real ships. Kirk sends a red-shirted crewman to turn it off, and you know what happens next. M5 sends out a power beam and vapourizes the red-shirt. Thus M5 was a system so reliable that even those authorized to could not turn it off. (Science Fiction has a number of examples of this.)
ClariNet's news delivery system had survived an effort to turn it off by the people authorized to do so. However, this would not last forever, and of course there were questions on the legality. I remembered how Comtex had the right to resell UPI news, purchased during UPI's bankruptcy problems in the 80s. I immediately called Comtex and did a deal with them to buy UPI news. To stay legal, I asked them to backdate the contract to the time the UPI contract had ended. We arranged a modem dial-up channel to pick up the news from Comtex until a satellite feed could be arranged.
UPI was of course furious that their own effort to shut me off had failed, and to learn that Comtex had become my supplier. They had primarily hoped to show me who was boss to get lots more money out of me. As such, they started putting pressure on Comtex not to sell to me. Comtex had the right to make that sale, but the UPI relationship was much more important to them than my relationship was, and they told me they would have to pull back. I was forced to do a deal with UPI at several times my prior rate, and on a month to month business, to allow my business to survive.
AP and Reuters
It was a good thing negotiations were already underway with AP and Reuters, and those negotiations were sped up. However, the problem of putting out the highly-categorized news product remained. The automatic classifier showed promise, but I had just 2 months, and so quickly cobbled together a system that would allow human beings to review all the stories and classify them. Some automatic classification was possible, but we would now need a staff of human beings reviewing the news, ideally 24/7.
I got the software ready, and we rapidly expanded our editorial staff by hiring several junior employees (journalism grads were very cost effective) to categorize and organize the news coming from all our sources. We had to go live with almost no beta test, and we were worried how our customers would deal with a sudden, massive change in the structure of our news content.
To great relief, it went off largely without a hitch. The AP and Reuters brands were actually better of course, though UPI had given us everything they had, in other words more volume of news and more varied news. To UPI's surprise, in just 2 months we were able to tell them goodbye. Their gambit had failed. I swore not to do business with them again, and would always have at least 2 major sources so no single source could hold me over the barrel.
Loss of Dave Barry and Piracy
Another blow came, sometime later, when Tribune Media Services, which syndicated Dave Barry to us, decided in June of 1994 to pull Barry and Mike Royko from us.
When I had first started the business, it was, as I noted above, because people were distributing Dave Barry's column left and right on the internet without payment. It showed the market, and I hoped providing a legitimate channel, where some money got back to Dave, would actually be a positive step. And indeed, I think a lot of people switched to reading through us. But not everybody, and of course columns got forwarded in E-mail as well.
Somebody forwarded one eventually to an executive at Knight-Ridder/TMS, we were told, and they ordered us cut off. However, not long after, they started offering the column for free on their own web site -- so there is some reason to suspect it was more about going their own way. We were paying like a small newspaper -- not enough to give them much loyalty to us.
ClariNet never used any copy protection or DRM on what we sold. It would have just stood in the way of legitimate customers. Instead, we took the approach that while people could forward single articles to others, we were not in the business of selling single articles. We sold a complete news service, and it was a lot more difficult to forward all of that, and to not be found out.
I came to the conclusion that this would be a rule for copyrighted material in the future; sell an entire service, not individual items. We did sell individual subscriptions to Barry and other columns, but they never sold particularly well, and we wanted to discontinue them anyway.
By the time we lost Dave Barry, the service was much more established, and so we didn't lose much, if any business by losing him. However, since publishing him had been the impetus of the whole business, it was a psychological blow.
When I finally discussed this with Mr. Barry in person, he apologized for what had gone on. He hadn't wanted to cancel the service, and was coming to embrace the internet and enjoyed the readers we brought him there. It was pulled at higher levels.
A bigger company
The company kept growing, and was profitable. We started getting attention from VCs who were looking around the space for a place to invest. Here I made the misjudgement of not realizing just how big the internet was going to be. I knew it would be huge, but I was too close to it, and more aware of what it was really about. I knew some of the hype was true but that much of it wasn't. I felt with a profitable, fast growing (100% per year) company, that VCs weren't needed.
I didn't realize that all the new companies would have vast purses from VCs and public markets, and that they would run around looking for ways to waste millions as fast as they could. Sometimes in bizarre hopes of real business, and sometimes just to pass the investment on to a greater fool.
In hindsight, I should have taken their money, as the only way to compete against people who would give away product just to see what would happen. From there I could have moved into a number of other opportunities that became even larger. Most of us who were involved in the early internet probably considered, at one time or another, all of the ideas that later became spectacular business successes and failures, but you can only do one thing at a time.
By 1994 I knew I didn't want to do all the management any more, so I recruited Roy Folk to join the company as President and COO. I retained the title of CEO and Publisher. It's handy that the newspaper business includes several titles that all sound like they are in charge when you want to impress contacts.
As described earlier, I met Roy at Personal Software/VisiCorp when he was product manager for VisiCalc. He had come to visit ClariNet in his own investigations of how to get into the obviously hot internet business. We had a real business and he had the management background, including having been general manager of Ashton-Tate, one of the largest PC software companies in the world during the 80s.
By this point Netscape was about to go public and the dot-com boom was ready to enter full swing. ClariNet's story becomes less special, and the annals of business are full of stories of that crazy time. In fact, because ClariNet was self-funded and profitable, it was considerably less crazy and extravagant than many of the other firms out there.
We continued to grow, and hired a number of good people, including folks like Don Woods, the co-author of the popular "Adventure" computer game that had sucked so much of my time at school, and Wayne Davison, the developer of the popular trn USENET newsreading software which had indirectly sucked so much of my time later, and David Brandin, former president of the Association for Computing Machinery. There were many others -- I had about 45 staff when I sold the business, and they all deserve credit, but this is a story more about the beginnings than the middle.
One week I sent out a memo to all staff that there would be an important all-company meeting on Friday afternoon to discuss new hope for the company. They all gathered, somewhat perplexed. Was the company in trouble? Was it to be bought? Was there a big new deal? I talked for a minute and then flashed out the 40 tickets to the opening of Star Wars I had bought. The receptionist had gone down to get a place in line. We piled into cars and had a great time. Even after I left, this became a company tradition for all the remaining Star Wars re-releases and a few other new films.
We were profitable, but we knew that to some extent we were riding the dot-com wave. A good part of our growth came from the fact that our ISP customers were growing. We gave them better deals as they got bigger and worked hard to keep them, but we still enjoyed their growth. One company, Netcom, became almost 20% of our business which was a major concern. It was time to plan the next stage of the business.
Two visions emerged. My own I called ClariNetwork, and it was an extension of the "all you can eat" content pricing model I had first worked on in the Library of Tomorrow. I planned to build a single-payment, single sign-on system for a network of content sites. You would pay a monthly fee and get access to all their content. The money would be distributed according to which sites were more popular. Some sites would remain ad supported, of course, but I was correctly predicting the collapse of CPM banner ads as a means to support general web sites. (Google and Overture however later demonstrated how well keyword targeted CPC ads would be the wave of the advertising future.)
The other was Roy's, a project later to be named NaviLinks. Roy felt that linking was what the web was all about, and web sites full of rich links would be the winners. He wanted to develop tools to fill pages with all sorts of useful links, and start by doing it to news.
While I believed in ClariNetwork, I couldn't get the others to buy in, so I greenlighted Navilinks, but my heart wasn't in it as much. ClariNet needed more resources to survive in the news game, or a bolder plan than that. We had been spending more heavily on various items, including R&D for Navilinks, and starting to move a bit into the red.
(I should note that I still think the collective licencing pool approach still has merit in selling content, and many have tried variations of it. My latest version is called Microrefunds).
This led to the decision to sell the company. We retained Broadview, a boutique investment firm. They pumped us up with dot-com hype as well, pitching dreams of selling the firm for $75 million, based on valuations of companies not doing nearly as well as we were. Sadly, the problem was that they weren't doing as well -- by making a profit, we nailed a stake in the ground about how well we might do in the future. The other companies were much greater risks but had (in theory) uncapped upsides.
It became apparent that the best chance involved some of the larger competing companies in our own space. Individual Inc. and Desktop Data were two firms that were almost 10 times larger, and were similar in age, but were still largely pre-internet firms. They had formed to deliver tailored news at high speed in a specific business niche. Individual started via fax, and Desktop Data has a proprietary platform.
We weren't able to interest Don MacClagan, the CEO of Desktop, but we were a better match for Individual. Individual had been a high flyer, but after a dispute between its founder, Yosi Amram and the board over the acquisition of another internet oriented company for a very high price, Amram had been forced out and the stock tanked. I gambled that their stock had been depressed but the next bout of internet good news would soar it back up. I got along well with Michael Kolowich, the new CEO, and big plans seemed in the offing.
Individual had an internet offering called NewsPage. I had been working on a special tool to do custom web delivery of our news, in a private project independent of my own developers. However, this tool was able to make our news appear to look just like newspage with just a few hour's work, so as we were closing the deal, I amazingly found time to do it, and it made things look rosy.
However, all was not rosy. We had done most of the negotiations in secret from our own staff, as is the norm in such cases. If you don't keep such things secret, people can soon think of nothing else, and news is likely to leak. When the deal was announced all of a sudden, and for less than people had dreamed, there was disappointment, and resentment even of the program that had made the deal. The deal took place, with a value about 8.5 million dollars of depressed Individual Inc. stock.
The majority of corporate mergers and acquisitions fail, it is reported, and unfortunately I can't report great success for this one. 6 months after the deal was done, Individual merged with Desktop Data to form a new company called NewsEdge, which had been the name of Desktop Data's flagship product. Maclagan, who had passed on the purchase of ClariNet a few months before, did not have it in his product plan. Staff were laid off, and eventually the company was spun out to Roy Folk and some of the other managers to pursue the Navilinks linking system. NewsEdge kept equity in the company but wanted Roy to raise further money, which he was not able to do.
However, the core ClariNet business was actually still strong. With the staff cut back the company was able to become profitable again, though it was able to gain few new customers. When revenue dropped, Roy passed the reins to Lynn Brock, who had not been involved with ClariNet before (though I knew him well since I had reported to him in my last period at VisiCorp) but who had a long association with Roy.
Finally Lynn sold the business to Yellowbrix, a competitor, which has incorporated the products into its line and continues to serve many of the original customers. You can still read the free sample newsgroup biz.clarinet.sample. In some senses the company is all but gone, but in other senses it remains, still the oldest of the dot-coms. The clarinet.com domain got sold off, though http://www.clari.net still exists as a shell. Of course, the news business and online news business have changed tremendously, and printed newspapers are starting to drop like flies. More change is in the offing.
Sometimes it surprised me how much staying power it did have. As I saw most major newspapers spend more money giving away their news as we had for a total budget, as I saw companies like Yahoo give away products that contained news just as good as ours, I felt it would fade sooner.
Some of the strength came from a hidden strength in USENET. While millions still use USENET (often just by searching it in Google) many think of it as a relic of the internet's early days. And it has many flaws which have resulted in its failing to grow like wildfire the way many other internet applications did.
But USENET contains something very important which has been misunderstood by most of the browser world. There are at least two types of content we like to read -- stuff we browse, like a web site, and stuff we read serially -- like e-mail, newspapers and magazines. USENET (and E-mail) aim directly at the serial media. They track what areas you are interested in, and what you have read and haven't read. Each time you come to them, you are shown only what is new for you. What you glanced at before is not seen unless you ask for it.
USENET had another key advantage that it has gradually been losing over time. USENET was designed to be read off your local server, at the LAN speed. Instant access makes a huge difference. Most ISPs now outsource their USENET and don't always provide the truly local speed, but universities and companies usually keep it in-house. As you know, once you have truly instantaneous response time, you will give up a lot to avoid something with any lag and certainly with unpredictable lag.
Turns out that form is the only one that really works for large volumes of data. Many web sites have tried to kludge it using the browser interface. They have often presented nicer looking interfaces with a number of innovative features, but they have often ignored the serial.
Recently, the RSS format and the rise of the blog have recreated interest in the serial. They do what USENET does, if in a not particularly efficient way, but with much prettier user interface, and with less centralization. But for those addicted to the USENET way, the news on Yahoo and Google News are no substitute. There is no web interface that provides the essential combination of speed and features, which is remarkable 30 years after USENET was born.
That's why, even as late as 2005, I was pleased to learn that Stanford University had decided to keep their ClariNet subscription, after those holding the purse-strings had pushed to shut it off. The readers had revolted, and demanded it be kept.
I have a short blog post on serial vs. browsed and reader friendly vs. writer friendly with more thoughts in this area.
Other early business on the net
Naturally, when making a case, there is a tendency to write definitions that support the case. I welcome posts in the comments with other people's thoughts on what the "dot-com" concept means in a way different from my "company born for the internet" definition.
If you do expand the definition, there were of course companies that did some business over the internet. This was often in technical violation of the NSFNet acceptable use policy, though sometimes it was truly in support of research and education.
This is the abridged history of one man and one company. While I believe ClariNet was the first company built to use the internet as its platform, there were many other companies in all facets of the business, doing all sorts of innovative things. I certainly make no claim to inventing the idea of the dot-com -- lots of people were all expecting it to happen sooner or later, and wondering where the first business would come from. And of course, many larger companies did experiments, using the internet as an adjunct to their business, going back to the earliest days.
For myself, the main note of regret was my own inability to accept the hype that turned out to be true (to a point) surrounding my own industry. Very few could have planned for that explosion; if anybody did they probably became quite wealthy, though they also had to be one of those rare few who knew when to sell.
Nonetheless it remains one of the most heady and fascinating times in the history of business. To participate in the dawn of a new era for business and society and to know it's a new era while you are living it is a rare opportunity. Each day some bold new announcement of seeming insanity in the financial community would come. Every day and every product meant breaking new ground, getting people to think outside of their old rules. At first they wouldn't do it, later they were more eager to do it than the experts were. (Then of course after the crash they went back to too much skepticism.)
I consider myself very lucky to have played at the start of two revolutions, the personal computer and the internet. However, I also believe more revolutions are to come. Computer networking is good for another one (I'm personally working on reinventing the phone call using networking to change the very nature of how we interact with one another) and I predict even more spectacular revolutions from biotechnology, nanotech and A.I. I'm working on self-driving cars and I'm on the boards of both the Electronic Frontier Foundation and the Foresight Institute for Nanotechnology because I plan to be there. Along with you, I hope.