Millicent

» see millicent.org as it lived, on the wayback machine

~ the goodbye ~

millicent was a micropayment protocol that came out of digital equipment corporation's systems research center in palo alto. the founding paper, the millicent protocol for inexpensive electronic commerce, was presented at the fourth international world wide web conference in boston in december 1995. the authors were steve glassman, mark manasse, martin abadi, paul gauthier, and patrick sobalvarro. they were all at dec src. the design target was a single purchase costing less than a cent, with a transaction overhead small enough that the per-charge cost would not eat the charge itself.

the architecture used three parties. you had a broker that you trusted with a real balance, and you had vendors who sold content. between them flowed scrip, which was a small signed token good only at one specific vendor. you bought scrip from the broker in bulk, which amortized the expensive credit card transaction. you then spent scrip a fraction of a cent at a time at the vendor, which was cheap because scrip validation was just a hash check. this was the whole idea. it was clever. it required nothing from the existing banking system once you had the broker relationship set up.

dec ran a beta on its own website and signed up a small set of partner publishers. the most committed external deployment was a commercial trial with kddi in japan starting around 1999 to charge for ringtones and small content drops. compaq bought dec in 1998 and inherited the lab. the millicent group kept working for a while. then hp bought compaq in 2002, the research labs were reorganized, and the project went quiet. by 2004 the website was gone and nobody was answering email about commercial licensing.

i did not use millicent. by the time i would have been a paying web user, around 2003 with university broadband on ttnet, the question of whether to pay anything for web content had already been answered by advertising. publishers ran banners. readers blocked them. nobody got out a wallet to read a single article. the engineering question millicent was solving turned out to not be the question the consumer web was asking.

~ the chronicle ~

~ why brokers and scrip ~

the cleverness was the broker tier. credit card transactions had a fixed per-charge cost of about thirty cents in 1995. you could not charge a cent for anything without losing money. millicent's answer was to make the expensive transaction infrequent. you bought a few dollars of scrip from your broker once a month. after that, every individual content purchase was a token swap that cost essentially nothing to validate. the broker absorbed the credit card overhead and made it back on the spread.

vendor-specific scrip meant you could not spend leftover tokens elsewhere, which was the price of keeping validation cheap. each vendor minted and verified its own scrip with a shared secret. brokers maintained the trust relationship in both directions. this is the same shape that prepaid app store credit later took, and the same shape that subway transit cards take. millicent got there in 1995 and could not find the consumer market.

~ what really killed it ~

the dshr blog post about millicent makes a sharp argument. the technology was not the problem. the problem was that content providers preferred advertising over micropayments because advertising is passive on the user side. with banners the reader does nothing. with micropayments the reader has to decide, every time, whether this specific article is worth a fraction of a cent. that decision friction killed every micropayment scheme of the nineties and 2000s, regardless of how clever the protocol was.

the dec lab that built millicent also built altavista and the first quad-core workstations. when compaq absorbed dec and hp absorbed compaq, all of that research talent dispersed. millicent did not get a real shutdown notice. it just stopped being maintained, and after the project page went away the protocol moved to the same shelf as digicash and cybercoin.

@}-->--

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

« previous  ·  random  ·  next »