A long-forgotten Microsoft project from the late 1990s revolutionized how we saw the Earth and became the underpinning technology of Google Maps and Google Earth.
The Earth fit inside a 45-foot by 25-foot Compaq computer in an office building in suburban Seattle. As the East Coast woke up Monday mornings, it would roar to life.
"The temperature in the room would go up 5 to 6 degrees, things would start banging around," Tom Barclay, the man tasked by Microsoft with putting the Earth inside a database, remembers. "You'd really marvel at it."
Terraserver could have, should have been a product that ensured Microsoft would remain the world's most important internet company well into the 21st century. It was the first-ever publicly available interactive satellite map of the world. The world's first-ever terabyte-sized database. In fact, it was the world's largest database for several years, and that Compaq was—physically speaking—the world's largest computer. Terraserver was a functional and popular Google Earth predecessor that launched and worked well before Google even thought of the concept. It let you see your house, from space.
So why aren't we all using Terraserver on our smartphones right now?
Probably for the same reason Microsoft barely put up a fight as Google outpaced it with search, email, browser, and just about every other consumer service. Microsoft, the corporation, didn't seem to care very much about the people who actually used Terraserver, and it didn't care about the vast amount of data about consumers it was gleaning from how they used the service.
"It was something we did to show off our software could do this, but the company didn't care about the information," Barclay told me. "Google was an information company first. They saw the value of the information."
From the outset, the plan was to make a database. Microsoft didn't really care what information it contained, it just had to be big. The biggest in the world, something that would test the scalability of Microsoft's SQL database products.
"We had been asked to work on a very large database, to test this next-generation database product," Barclay told me. "It turns out that finding both an interesting and real terabyte of data that wasn't encumbered in some way, that we had the permission to [distribute legally], was a challenging problem."
According to a USA Today article from June 22, 1998, the initial plan with Terraserver was to list every single transaction in the history of the New York Stock Exchange online and make it searchable. But that was only a half terabyte of data. Microsoft needed something larger.
In 1997, the United States Geological Survey was in the process of uploading greyscale satellite photos and other aerial images from its archives onto the internet. Hedy Rossmeisl of the USGS met with famed Microsoft computer scientist Jim Gray, and they started brainstorming. Wouldn't it be interesting, and perhaps useful, they thought, if someone put searchable satellite images on the internet?
The timing was more-or-less perfect. The Cold War was over, which allowed spy satellite imagery to be declassified, no one was worried about terrorism in a pre-9/11 world, and, well, the average person was beginning to get the internet.
"We had imagery from maybe half of the country done digitally and we had some capabilities to deliver them, but not in a fast, accessible way," Rossmeisl told me. "I thought getting the data on the web was really important, and I wanted to help make it happen."
The images, along with some from recently declassified Russian military photos, totaled just over 2.3 terabytes. The idea for Terraserver was born.
Gray put Barclay, who Rossmeisl called "the brains of the project" in charge, and he got to coding. He was a database guy—Terraserver was the first website he'd ever made, and it was the first project he'd ever tried that had anything to do with mapping, which proved to be quite a challenge. Barclay quickly ran into an age-old cartography problem.
"It turns out that 'round Earth, flat monitor' is an enormous pain in the neck," Barclay said.
He decided that using a standard Mercator map projection, which is what you see in the image above, wouldn't work because it distorts the sizes of land masses as you move north and south on the projection. After trying a few things, Barclay came up with the idea of creating "mosaic" images that would be automatically generate based on where you're clicking on the map. Basically, the images given to Microsoft by USGS were stitched together but were then chopped into smaller images that could recenter themselves on cue.
"Originally, we hadn't done this. The very first demo we did, I chopped Bill Gates's house in half, which was not very good," he said. "We ended up with a progressive display that allowed people to drag and center the screen where they wanted it, and we computed zoomed out and zoomed in views."
These innovations proved to be revolutionary, and the "mosaic" strategy is now the "underpinning of Google Earth and Google Maps," Barclay said.
"I don't want to break my arm patting my back, but it's amazing how similar the current technologies are to what we did in 1998," he added. With the mapping problem solved, Terraserver went live, and the real fun began.
I was 10 years old when Terraserver launched, and if I used it, I don't remember. Unfortunately, there's no way of using it today. Terraserver went offline in 2007, and Barclay spent most of his time working on Bing Maps. Microsoft periodically revived Terraserver from time to time even after 2007, but it's offline forever now. Barclay attempted to bring it back on a separate server for the purposes of this article, but said that the project proved too time consuming.
So while I don't remember Terraserver, it does seem like it made quite a splash when it launched. In addition to the USA Today article (more of a blurb, really), Terraserver also scored early stories from the New York Times and Newsweek, which worried about the system's potential privacy-invading potential (headline: "Surveillance in the sky").
Microsoft held a launch event in New York City that Bill Gates attended. On the first day, 8 million people accessed the site, "millions more were rejected," according to a white paper published in 2000. By the end of the week, it was getting 30 million hits a day. Ultimately, the site settled down and served roughly 7 million people every day. It was more successful than anyone at Microsoft ever anticipated.
And that brings us to the dumbfounding thing about Terraserver, and about Microsoft. The reason, really, why I'm writing this article. In reading the white paper, it's astounding to see just how much information about general web behavior Microsoft was able to glean from the project, and it's astounding to see how it essentially blew it by looking at Terraserver as a novelty project rather than a potentially world-changing one.
Microsoft learned, maybe even before Google, that most search is local. If Terraserver didn't have images for people's hometowns, they got angry.
"In the first year, I got 20,000 emails, and the vast majority of them said one of two things," Barclay said. "It was either 'I love Terraserver, I saw my house' or 'I hate Terraserver, I didn't see my house' We learned that 85 percent of all geospacial queries are local. They're looking for local search—they want to find whatever dry cleaner is around the corner, or where they could get fast food."
The entirety of the New York Times article about Terraserver's launch focuses on its utility as a database and all but ignores the possibility that it could serve as a method of collecting information about user habits.
"The project not only marks the creation of one whopper of a digital scrapbook, it also says something very big about Microsoft's effort to enter the database business, using as an opener a venture that can capture the public imagination," the Times wrote. "Microsoft's strategy is to use Terraserver to prove that its software and operating system are suited to massive databases."
It wasn't just that basic information, however. Microsoft also gleaned that "the internet is busiest on Mondays and Tuesdays" and that there was a "steady slide [in volume] from Wednesday through Friday." Saturdays and Sundays were half as busy as Mondays were. The 45-foot by 25-foot Compaq computer that stored the images would roar to life on Monday mornings as the East Coast woke up.
"The temperature in the room would go up 5 to 6 degrees around 9:30 AM on the East Coast, things would start banging around," Barclay said. "By 8 PM pacific time, you didn't have any traffic, because we didn't have any imagery in the Pacific Ocean."
None of this information was used by Microsoft, except as a way to determine when to perform maintenance on its servers or when to staff the server rooms. The only revenue Microsoft made directly off of Terraserver was on the sale of some of the satellite images, which you could buy and have mailed to your house for $9.95.
"In the science community, this technology took off, but as a business I could never get anyone at Microsoft to latch onto it," Barclay said. "There's definitely a little bit of frustration there."
It's easy to look at Terraserver as a missed opportunity for Microsoft to dominate the next era of computing, and it's hard to say why, exactly, the company decided to stop pouring resources into it. Current Microsoft representative declined to be interviewed for this article, and Jim Gray, Barclay's boss, was lost at sea in 2007.
It may be as simple as Barclay suggested: Microsoft didn't see itself as an information company, and the media was skeptical of its intentions had it decided to become one. In addition to the Newsweek article, the Chicago Sun Times ran an opinion piece in 2000 that questioned the company's motives with Terraserver.
"Some people are paranoid enough about Microsoft," Andy Ihnatko wrote in an article I accessed using LexisNexis. "How would these people react to discovering a Microsoft web server with an aerial photo of their house that's so good it shows the kiddie pool in the backyard?"
Other groups weren't as skittish. The most notable was Keyhole, which launched "Earth Viewer" in 2003 and used Terraserver as some of the underpinning of their technology. It sold the license to its Earth Viewer software for upwards of $600 annually to businesses and charged consumers $79 annually for a stripped down version of it. Google bought Keyhole in 2004, rebranded Earth Viewer as Google Earth in 2005 and, well, you know the rest of the story.