Article: 45354 of comp.protocols.tcp-ip
From: vern@daffy.ee.lbl.gov (Vern Paxson)
Newsgroups: comp.protocols.tcp-ip
Subject: Re: Traceroute and TTL
Date: 29 Sep 1995 19:38:54 GMT
Organization: Lawrence Berkeley Laboratory, Berkeley CA
Message-ID: <48407@dog.ee.lbl.gov>

In article <44c1a6$jtj@news.cuny.edu>,
Erik Van Riper <geek@imageek.york.cuny.edu> wrote:
> Even more bothersome is no matter where I go, as soon as I hit 
> SprintLink, I get the same "problem".  Anyone else have an idea what 
> traceroute is telling me here?

I'm just finishing up a draft of a study of end-to-end Internet routing
behavior.  One of the findings that surprised me is the prevalance of
routing asymmetry.  Of the ~600 different Internet paths I looked at
(i.e., source/destination pairs), about 30% exhibited a major asymmetry,
meaning that the routes in the two directions visited different cities.
I also found that such differences generally lead to different TTL's
(different number of hops in each direction), which explains the behavior
you're seeing.  Finally, due to the phenomenon of "hot potato" routing
between ISP's (explained in the paper), we can expect asymmetries to
increase in the future.

I'll be submitting a more polished version of the study to SIGCOMM '96.  If
you're interested in reading the draft, drop me a line.  (Those folks who
I've already communicated with about this study needn't send me additional
email, I'm already planning to let you know when the draft is ready.)

		Vern