Linda -- A Cool Programming Idea

• The best link is http://www.sca.com/LINDA_overview.html
• The original developers are http://www.cs.yale.edu/Linda/linda.html

It's a cool parallel programming idea offering

• Seriously easy programming.
• Easy to achieve fault tollerence.
• Obvious hetergenaity in terms of both hardware and programming languages
• Can run in background on networks of workstations
So what's wrong with Linda?
• It is not the fastest solution possible.
• It does not scale perfectly well for LARGE problems.

There is a Tuple Space, which consists of an associative memory of tuples.  Each tuple is an ordered list of items, and each item has a value and a type.  For example, ("hello", 10, 20.3) and ({1,2,3,4,5,}, 5, "The Solution") are both tuples.
Some process places into tuple space the initial data needed to solve the problem.  Other processestake data from tuple space and refine it, producing data closer to the output.  When the last transformation has finished, the answer is in tuple space.

The basic data item shall be the tuple (board position, list of moves).
The initial process places (current_position, null) into tuple space.
Each worker
      reads in (board_position, moves)
      for each position that can be reached from board_position with one move
              If the new position is not a checkmate
                      add (new_position, moves + new_move)
               Else
                       out("Check Mate", moves + new_move)