I don’t remember how I wound up at the Internet Archive. Obviously I interviewed and all that but I don’t remember how I found the job or anything. Maybe through someone I met while I was volunteering as a sysadmin at the 24 Hours of Cyberspace event? Regardless, it was my next job after Sun Microsystems, which I left after the JavaStation work I’ve already written about. Fascinating place, no matter how I got there. There’s not going to be any central thread to this post, just a bunch of memories of a cool place to work and a bit of technology nostalgia.
I wasn’t really working for the Internet Archive. I was working for Alexa Internet, which was the for profit company Brewster Kahle started because he wanted to make sure the Internet Archive could stay a going concern. In practice at the time there wasn’t a ton of differentiation between the two companies. Alexa produced a browser add-on that suggested other sites you might like based on the site you were currently visiting, while the Internet Archive (at the time) was a digital archive of as much of the Web as we could crawl. They were both built on the same infrastructure, and de facto it felt like we were one company.
At the time, everything was located in a small building on the grounds of the Presidio in San Francisco. The Presidio had recently opened up leases to non-profits, which is how Brewster managed to get space. This was 1997, before the Presidio decided that maybe they should rent to for profit companies as well – Lucasarts is now a major tenant, in a complex I had reasons to visit a bunch when I was working at Disney. The Alexa office was really close to where I was living in the Sunset neighborhood and I loved going to work there every morning. It’s probably the most beautiful place I’ve ever worked, although Zillow’s offices in a skyscraper right on the edge of Puget Sound come pretty close.
I’d go down to the Presidio Burger King for lunch and sit near the huge window that looked directly onto the Golden Gate Bridge. It was a ridiculously amazing view and I found it delightful that it was benefitting a fast food restaurant. That place is long since gone; I think the visitor center replaced it.
We were in the old PX building, which I think is still standing just north of what is now the Walt Disney Family Museum. When I say everything was located in that building, I mean everything: the offices and the machine room. The most important thing I learned at the Internet Archive is that you shouldn’t put a machine room in a century-old structure with wildly uneven floors.
The servers themselves were fine. We were running Solaris on Intel servers, which I had not encountered before: Brewster was happy to get a little experimental in the interests of saving money. They worked quite well. The problem was the StorageTek tape robot we kept the archive on.
Why a tape robot? Well, hard drives were pretty expensive back then. We had just donated a couple of terabytes to the Library of Congress, which took the physical form of 44 digital tapes. The tape robot was much more practical.
A company with the descriptive name Large Storage Configurations supplied us with a filesystem called SAM-FS. This allowed us to mount a view of the files on the StorageTek robot onto our Solaris filesystems with a much smaller amount of hard drive space configured as a cache. Thus, when someone used the Wayback Machine back in 1997, the request for an archived page turned into a request to load a file from this special filesystem. If the page happened to have been accessed recently, it’d still be in the hard drive cache and it’d come up quickly. Otherwise you’d have to wait five minutes or so while SAM-FS told the tape robot to load the file in question. I believe the technology has improved somewhat since then.
The problem was that tape drives are sensitive to little things like hundred year old army base floors that aren’t level. There were three physical drives in this thing, and every couple of weeks one or more went out of alignment. I got to know the StorageTek guys pretty well: “Hey, this is Bryant down at Alexa. We have a drive out of alignment again.” “Oh, sure, we’ll send out a tech tomorrow.”
Cheaper than renting datacenter space, anyhow.
I also met some great people there. Tim Pozar was my boss. He and I used to carpool to work since he lived near me, cruising through Golden Gate Park in his Miata talking about the communal wireless network he was designing and implementing for our neighborhood. Tim would later go on to bring Internet to the Farallon Islands.
I reestablished contact with my old college pal Darcy Gibbons, who the politically aware may know as Darcy Burner, netroots candidate for the House of Representatives in Washington State. Her husband Mike was VP of Development at Alexa. We were both pretty surprised to discover that connection.
Oh, and Scott Hassan! You may know that name from Google’s early history; he was a key engineer on the original Google prototypes at Stanford, and left for Alexa before Google became a company. Scott was scary smart and offered me the biggest brass ring I’d ever see in my career. He had a deal where he was allowed to bring in a Sun server, plug it into the company Internet, and keep it under his desk. Cheaper than datacenter space. He had this little side business he ran on that server.
At one point he said “Hey, Bryant, you’re pretty good at what you do and my side gig needs a system administrator – want to join my company?” Like an idiot, I said, “Scott, I’m totally flattered, but I don’t think there’s really a future in offering email list management – surely anyone who wants a mailing list can just set one up themselves, like I’ve done myself?” He shrugged and said he understood.
In 2000, Yahoo bought eGroups.com for $432 million. Scott used his share to get into robotics. So it goes.
Not that I have anything to complain about. After I’d been at Alexa for about a year, Tim left, and I arrogantly decided that I should be the next manager. I told whoever I was reporting into at the time my conclusion, and they politely said “no,” which was the right answer. I shrugged, called my pal Ambar, and said “I hear this AltaVista place is fun?” A month or so later I was starting a new job back down in Palo Alto.
Amazon purchased Alexa less than a year later. That was before the deadline for exercising the stock options I’d gotten while working at Alexa, and even I realized it was a good idea to pick up a bunch of Amazon shares for pennies on the dollar. So while I missed very early retirement by turning down Scott, I still wound up making a significant chunk of my current retirement fund at Alexa.
More than a decade later, I wound up at Amazon. Coincidentally, I got that job through Ambar too. The role I took on the front end Amazon search team was open because almost everyone who’d previously worked on the front end infrastructure team had just moved over to work at… Alexa! That’d be the Alexa you know and love, the voice-based virtual assistant – nothing to do with my Alexa but the name. Funny coincidence nonetheless.
On the rare occasions when I let myself be tempted to play the what if game, Alexa’s one of the big ones. I wouldn’t have made as much money if I’d stuck around, but I’d have gotten to work on projects that materially benefit the world. They have real datacenters now too. I’m glad I played a tiny part in the Archive’s early days.