22cdb741   
   From: cpolish@nonesuch.com   
      
   On 2011-07-11, David Schwartz  wrote:   
   > On Jul 10, 9:33 pm, Robert Wessel  wrote:   
   >   
   >> >Thank about it logically -- it is impossible any hop to show worse   
   >> >than any hop after it. So if you think hop 3 is somehow worse than hop   
   >> >4, it is an artifact of how you are measuring. (Because you have to go   
   >> >through hop 3 both ways to get to hop 4.)   
   >   
   >> Or some of the devices are just slow in responding to the tracerts -   
   >> having that handled far off the packet forwarding fast path is far   
   >> from uncommon.   
   >   
   > That would be precisely the kind of thing I meant by "an artifact of   
   > how you are measuring".   
      
   I don't see how the data support your assertion. If variation in   
   hop 3's RTT affected data from subsequent hops, then how could   
   hop 9's points cluster tightly on a line when hop 3's points are   
   widely dispersed?   
      
   The measuring technique used was devised by Van Jacobson (author   
   of traceroute, RFC1144, co-author of RFC1072, RFC1185, RFC2117,   
   RFC1889, RFC2327, RFC3550, RFC1323 (defines the RTTM mechanism   
   for TCP), and others. I don't have the chops to critique his   
   work but I presume his methods are sound. The implementation uses   
   iteratively-weighted least squares fit on a large number of   
   samples.   
      
   Please take another look at http://imgur.com/a/tcQFJ   
   --   
   Charles Polisher   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)