JAOO 2007: Joe Armstrong - On Erlang, OO, Concurrency, Shared State and the Future, Part 2

Sign in to queue

The Discussion

  • User profile image
    aL_

    FPGAs.. they sound alot like a  3d accelerator to me Smiley so in that way, most of us already has them Smiley
    id just like to point that out..

    perhaps thats the way things will be.. i doubt strongly that we'll be writing everything in erlang/ml/f#, after all not everything can be parallelized to 20 tasks, but perheps we'll use functional constructs in a library to witch be pass a task (aka ccr /plinq)
    thats what i think anyway Tongue Out

  • User profile image
    mpcm
  • User profile image
    TimP
    mpcm wrote:


    Watched it last night, definitely a must see. Smiley
  • User profile image
    sylvan
    aL_ wrote:
    

    FPGAs.. they sound alot like a  3d accelerator to me so in that way, most of us already has them



    Not really. A 3d accelerator still just interprets instructions from a fixed instruction set. An  FPGA is basically a grid of gates, and you litterally change how they're wired together (not physically obviously). It HAS NO instruction set, it's just a bunch of hardware gates that you can rewire to do whatever you want.

    That said, I'm skeptical that it's the way forward. I mean they've been around for a very long time, but mostly you just use them for prototyping, and then when you're done you build it "for real" and it goes a lot faster.
  • User profile image
    raymond
    A great series of Videos. Smiley

    Thank you again Charles and Microsoft.

    Cool
  • User profile image
    ivan_
    sorry for the silly question. When Joe was talking about fault tolerance, he said something that you could copy one machine on to the other, or whatever one machine does can be replicated on the other machine. 
    How would you create another copy in a secure fashion? How to make sure that only a certain entity/function/whatever can create a copy, especially when there is nothing to copy from when everything is crashed, wouldn't there be a need for a static state which can be replicated? What if that third entity crashes, wouldn't there be a need for every machine to be able to create another machine and exchange messages to see whether there is at least one machine running, somebody also needs to hold a security policy, that is static... right? no?...

Add Your 2 Cents