Re: random testing...

Tom Anderson (tom@boris.cs.washington.edu)
Tue, 11 May 1999 02:11:01 -0700 (PDT)

or receivers could use a change in the TTL field to defer dupacks
for out of order packets.

tom

---
Content-Type: TEXT/PLAIN; charset=US-ASCII

Cool! As another application, presumably senders (like TCP Vegas) could use these as heuristics to provide clues as to when routes have changed.

Hmmm. Although... Do level 2 routes change much? And are there many asymmetries in these systems? (ATM,...)

neal

On Mon, 10 May 1999, Stefan Savage wrote:

> Here's a really stupid test for path symmetry. > > 1) do a traceroute from A to B, number of hops reported = n > 2) send a ping from A to B and look at the ttl field, most os's start ttl at > 64 or 255 (easy to figure out which). Subtract from ttl value. > > 3) now take the difference of these two numbers. If the difference is large > then there are alot more hops in one direction than the other. I've found a > couple doozies, but generally folks are pretty close. It'd be a one day > project for someone to built a random tester to measure this though. > > - Stefan > >