AlbSoft

In an era of information overload and ephemera, where an online sensation may last all of five minutes, word is spreading that the internet has a memory, and its name is not Google.
Even more surprising, it has a physical address: 300 Funston Avenue, Richmond, San Francisco. It is a sleepy, unremarkable street, until you come to an imposing, neo-classical building with Greek pillars, heavy metal doors and a flag of the world planted on the lawn.
Step inside, and the first thing you see at reception is a heap of newly-delivered boxes containing hard-disk drives, each capable of storing 4 trillion bytes of information. Follow the humming sound up a flight of stairs and you encounter rows of machines, lights blinking, methodically hoovering up humanity's knowledge. This is the Internet Archive.
"Our mission is universal access to all information all of the time," said Rick Prelinger, president of the board. "We are part of the infrastructure of the web. We are the web's memory."
The Internet Archive, a non-profit, is the digital equivalent of the ancient library of Alexandria, a burgeoning hoard of websites, video, film and music which could otherwise be lost. It currently holds 281bn webpages, or URLs, and each month adds billions more. It also captures and stores books, journals, YouTube clips and cable news. Long revered by scholars, techies and librarians, the Archive's fame is now spreading among ordinary people, drawing more than a thousand hits per second to its website. Many, however, remain unaware of its existence, and as he gave a tour to the Guardian, Prelinger said:

I'm tremendously surprised that there are not more internet archives. It's the medium of our time but there is an ethos of ahistoricity. We're trying to negate that.

The organisation, which was co-founded in 1996 by Brewster Kahle, an internet pioneer and entrepreneur, considers its mission to be increasingly urgent. Technological, economic and political pressures devour digital information, just as termites – once dubbed the "teeth of time" – chomped through ancient libraries. Disks fade and warp, destroying information. Businesses go bust, or evolve, and in the process shed much if not all of their digital archives. Governments and institutions like to delete information that becomes inconvenient or embarrassing, leaving 404 error messages where once were pages.
"During the Iraq war the [Bush] White House quietly took down some of its earlier press releases. But we had them," said Prelinger, whose speciality is archiving film. "Digital information is part of our cultural heritage but it's tremendously volatile. It's fragile." Storing it is not just an act of historical preservation, he says, but a means to hold institutions accountable. "We want to help keep the internet honest and safe and defend it from ignorance."

'Philosophical allies'

Aaron Swartz, the internet activist and developer of Reddit, who died in January. Photograph: Fred Benenson /creativecommoners/flickr Philosophical allies include www.wikimedia.org, Mozilla, the free software community, the Electronic Frontier Foundation, a digital rights advocacy group, and the internet activist Aaron Swartz, until his death in January.
Google is not on the list. It is a marvel, said Prelinger, but tilts search results. "Its algorithms are not public. We don't know why we're seeing what we're seeing and we don't know what we're not seeing. Google knows your profile and adjusts accordingly. They want to sell you ads. We're not Google. We're a library."
Staff held a party last October to celebrate a milestone: 10 petabytes – equivalent to about 10 billion books – archived. Librarians and scholars acclaim the Archive's workers as "heroes" and "rock stars", but staff members are likelier to call themselves geeks and nerds. They joke about kilowatt consumption and meta-data replication. Some take part of their pay in Bitcoin and have persuaded the neighbouring Chinese restaurant to accept the currency. There is a growing throng of half-size terracotta statues depicting workers with more than three years service. The Wayback Machine, a searchable online museum of billions of web pages dating from 1996, is named after a segment in The Rocky and Bullwinkle cartoon show.
Kahle, a computer scientist who made a fortune in the 1990s with tech ventures, including Alexa Internet, dreamed of a Great Library of Alexandria 2.0 since he studied at MIT. The archive's first headquarters was in the nearby Presidio district. In 2009 it moved into a former Christian Science church on Funston Avenue; its pillars and facade evoke antiquity.
About 50 staff work here and another 100 work elsewhere in the bay area and in 32 scanning centres, usually in libraries, around the world. The centres digitise books, microfilm and regular film. Automation proved imprecise so it is done manually, each worker processing 800 to 1000 pages per hour. This labour means material such as Boston's John Adams Library, the Hoover archive and the 1930 US census are now online and free. Institutions such as government agencies, libraries and universities, many outside the US, pay modest fees for special requests.
The archive has also stored 750,000 actual books at a nearby climate-controlled storage unit, a literary equivalent of the Svalbard global seed vault. There is space for another 780,000.
Engineers "crawl" the world's top million websites, capturing and storing pages which link to other pages which are captured and stored. Every three months they start over, because the list of top million sites constantly changes. An average web page lasts 75 days. In 2009, they raced against the clock to save as much as they could of the web-hosting service GeoCities, before Yahoo shut it down. If the owner of a defunct website prefers that the pages remain dead, he or she can ask the archive to remove them, requests that are almost always granted.
Engineers also collect news from more than 60 TV stations worldwide and YouTube videos, selecting the latter according to Twitter mentions. "It's not perfect but tweets give us an idea of what people consider important," said Alexis Rossi, the web collections manager. She estimated that the 10bn URLs saved each every three month cycle represented – very, very roughly – about a 10th of the internet's output:

It's a Sisyphean task. We know we'll never get it all. The web by its nature is infinite.

The archive's three bay area data centres use 180 kilowatts, the equivalent of 45 homes, to power servers and keep the lights on. New disks hold 4 trillion bytes, in contrast to earlier models which held 2 or 3 trillion, helping the archive keep pace.
"I'm proud that we're keeping all this going. We do it on a shoestring budget," said Jim Shankland, director of operations. "As long as we do our jobs, the bytes will live forever and ever."