From: grahn+nntp@snipabacken.se   
      
   On Wed, 2010-11-10, Morten Reistad wrote:   
   > In article ,   
   > Martijn Lievaart   wrote:   
   >>On Tue, 09 Nov 2010 13:09:16 +0100, Morten Reistad wrote:   
   >>   
   >>> If you are on Linux you can use tc(8) to simulate all kinds of networks;   
   >>> you can make drops, jitter, duplicates, delay etc; just as a real   
   >>> network will.   
   ...   
   > I strongly suggest to test programs with tc instead of local code-based   
   > simulations.   
   >   
   > You provoke out more bugs that way, and the object under   
   > test remains static.   
      
   What's a "local code-based simulation"?   
      
   If you mean disturbing the process on either end like I was talking   
   about upthread (SIGSTOP/CONT, or whatever it was) I'd argue that   
   - you need to do that too to know more about how the software behaves   
     in the real world   
   - it doesn't modify the object under test, if the object is the process   
     rather than the whole host   
      
   But if I only had time for one of them, I'd probably try your dirty   
   network simulation! (And document that I skipped another set of tests.)   
      
   /Jorgen   
      
   --   
     // Jorgen Grahn    O  o   .   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)