From: ted.dunning@gmail.com   
      
   Actually, most serious implementations of what might be called that   
   have any traction by large audiences do not involve any motion of code   
   and involve only limited movement of data other than task   
   specifications.   
      
   The simple reason for this is data concentration.  If you have an agent   
   that you want to operate over a really large dataset, it is much easier   
   to keep all agents that operate over that large data set proximal to   
   the large data set.  If you are talking about the web, then it makes   
   very little sense to have very many spiders running, better to have the   
   agents live near the spiders (aka put them at Google).   The end result   
   is really that you tend to have server farms that have agent nests   
   alongside.   
      
   This fundamental issue applies almost any time you have agents that are   
   intended to do something non-trivial and particularly if you want   
   agents to do something social (if only for privacy reasons).   
      
   As such, most speculations about how agent code needs to move around is   
   really aiming at the wrong end of the problem entirely.  In fact, it is   
   probably better to search out problems where agents really make sense   
   rather than working on solutions to problems that nobody ever really   
   had (or will have).   
      
   [ comp.ai is moderated.  To submit, just post and be patient, or if ]   
   [ that fails mail your article to , and ]   
   [ ask your news administrator to fix the problems with your system. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)