AllPeers

» see allpeers.com as it lived, on the wayback machine

~ the eulogy ~

AllPeers was a Firefox extension, which in 2006 was an ambitious thing for a company to be. It let you build a list of people you actually knew, drag a file or a folder onto one of their names, and have it go to them without passing through anyone's server. BitTorrent handled the transfer and SSL handled the encryption, and the whole thing ran inside the browser on every platform Firefox ran on. At the beta it was around 200,000 lines of C++ and JavaScript.

The company was run out of Prague by Matthew Gertner and Cedric Maloux. TechCrunch wrote it up as the Firefox killer app on 3 january 2006, which was eight months before there was a public build, and the expectations set in that window are most of what happened next. The beta went out on 24 august 2006. Wired and PC World gave it awards. In march 2007 the client was released under a triple license, though the server stayed closed because the design needed one stable instance to work at all.

It closed on 2 march 2008 and the service stopped the next day. Gertner's account of it afterwards was unusually direct for a shutdown notice. The investors had expectations about how fast the user base would grow, and while the team was satisfied with how the product was being taken up, the investors were not. When the cash ran out nobody was willing to put more in. Mangrove Capital and Index Ventures had funded it. Eighteen months of public life is very short for a piece of software that took two years to build.

I was at university by 2006, on TTNet broadband, and file sharing at my level meant public trackers and whatever survived a direct transfer over MSN, which mostly meant nothing did. Both ends behind a router and the transfer would sit at zero percent until somebody gave up and found an upload host instead. AllPeers was aimed precisely at that failure, at the ordinary case of moving something large to one person you actually know, and it is genuinely strange that it took another decade and a half for that to become a solved problem.

~ the dossier ~

~ how it worked ~

The clever part was using BitTorrent for the wrong thing on purpose. Torrents were built to move one file to thousands of strangers efficiently, and AllPeers pointed that machinery at a friend list, so what you got was a private swarm of people you had approved. Nothing was published, nothing was indexed, and there was no search. The technology behind it was the most public distribution system on the internet and the product built on top of it was the most private.

Living inside Firefox was the other decision, and it is the one that dates it. In 2006 the browser extension was where new ideas went, because Firefox was taking real share from Internet Explorer and its extension model let you build something close to a full application without asking anyone for permission. That window closed within a few years, when browsers tightened their extension APIs and the interesting work moved to the phone.

~ the money ~

AllPeers had two serious European funds behind it and a product press coverage described as a killer app. What it did not have was a metric that grew the way venture funding requires. The people who installed it appear to have liked it. There were not enough of them, and there was no obvious mechanism to make there be more, because the product's whole value came from your friends installing it too and the growth had to travel one friendship at a time.

That is a slow curve by construction. A company funded on the expectation of a fast one has maybe two years to demonstrate otherwise, and AllPeers had eighteen months of public existence to do it in. The shutdown notice is worth reading for how plainly it separates the product working from the company surviving, which are two questions people still routinely confuse.

~ what got buried with it ~

The friend-to-friend model went into the ground with it and stayed there for a long time. What actually solved large file transfer for normal people was the opposite architecture: upload to somebody's server, get a link, send the link. Dropbox launched into private beta the year AllPeers died and WeTransfer arrived the year after. Both put a company in the middle of the transfer, which is exactly what AllPeers had been built to avoid.

The trade was convenience against custody, and convenience won so completely that a whole generation now assumes a file has to go through somebody else's machine to reach a friend. There is a straight line from here to Firefox Send, Mozilla's own attempt at the honest version of the same idea, which arrived eleven years later and died for entirely different reasons.

~ what came after ~

The code is still out there under its open licenses and it has not been runnable in any practical sense since the server went off, which is the recurring lesson of open sourcing a client and keeping the server proprietary. You are given the half that cannot survive on its own.

Both founders stayed in the Prague startup scene afterwards, which is a considerably softer landing than most entries in this graveyard get. It does not change the fact that the thing they built understood what people needed and had eighteen months to prove it, which was never going to be long enough for a product that spreads one friendship at a time.

@}-->--

              )
             (_)
            .-'-.
            |   |
            |   |
            |   |
            |   |
          __|   |__   .-.
       .-'  |   |  `-:   :
      :     `---'     :-'
       `-._       _.-'
           '""""""

« previous  ·  random  ·  next »