Danny Hillis: The Internet could crash. We need a Plan B


15 Responses to “Danny Hillis: The Internet could crash. We need a Plan B”

  1. jgs says:

    And the rebuttal from Vint Cerf:


  2. I carry some baggage. I bought Hillis’ book on The Connection Machine. It was his big first effort. As I remember it, the first couple hundred pages were on how massive fine grain parallelism with multiple instructions (MIMD) would win the day. Then there was a page saying that this was too hard, and so The Connection Machine was single instruction (SIMD) and fine grained. A cheat. It was apparent in a few years that the effort lost on two levels. SIMD lost and fine grained parallelism lost. Basically grain-size grew as fast as microprocessors, moving from 16 to 32 to 64 bits, and then to graphics processors and VLIW. Somehow through this though, Hillis kept his day job as a great thinker. It is just a Sausalito thing? Know the boys at Whole Earth and then Edge and keep your podium? I worry it is so.

    • Gatto says:

      I’m not sure you’re right thinking SIMD lost. The PS3/cell architecture is basically just a bunch of coordinated SIMD chips, and GPUs are composed of pixel and vertex shaders which are essentially computing the same ( though not single ) instructions on multiple data. Not sure about the rest though, haven’t read his book :}

      • “Fine grained SIMD” was key.  “The CM-1, depending on the configuration, had as many as 65,536 processors. The individual processors were extremely simple, processing one bit at a time.”  The argument in the book was that massively parallel integer units of very small size would make it.  The CM2 added shared floating point units, in a bit of a capitulation.  “In order to improve its commercial viability, the CM-2, launched in 1987, added Weitek 3132 floating-point numeric co-processors and more RAM to the system.”  

        I’d suggest that graphics processors have more Cray roots than CM.

  3. Navin_Johnson says:

    Can’t you just turn it off and on again?

  4. Funk Daddy says:

    It’s serious because if we can’t play our MMORPG’s reliably, we will get real with that shit.

    I have a sword and a spaceship, but only the sword works. Best to keep the internet up since I can’t leave this rock.

  5. Brainspore says:

    In my fantasy world some unstable vigilante has a directory like that, except it only lists the people who send out spam.

  6. Ramone says:

    Right, so asteroids, the Internet–anything else?

  7. JoeBuck says:

    The number of people who had email addresses in 1982 was far larger than Danny Hillis’s book would indicate.  However, most of those addresses were hybrid things, like joe%foovax@berkeley.edu (mail was first routed to Berkeley’s Internet node and then routed from there onto a local network). I actually had an address on a proper core ARPAnet machine, so maybe I’m in his book (I was a very young employee at Naval Research Laboratory in DC at the time).

    • PaulDavisTheFirst says:

      I remember you from the mid 1980s. You’re definitely on his list ;)

    • Ito Kagehisa says:

      I can’t actually remember if I had an email address in 1982.  I mean, I should have had, and it should have been some ungainly thing that had a UUCP link somewhere in it, but I honestly can’t remember any more… that was the year I reached legal drinking age in Delaware, so I didn’t have to drive to Maryland anymore.

  8. tylerkaraszewski says:

    Is amateur radio not already his “plan B”?

Leave a Reply