[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]

Re: Curious about the way Tor makes cells

> I know that on many of my systems, I've increased the TTL to the maximum
> (255). It's probably overkill... but why not? The Internet's a different
> place than it used to be.

This is called cargo-cult science, I believe.

> I don't know if it really makes all that much of a difference for Tor,

It makes no difference whatsoever.

> (If someone presents a convincing argument as to why this is harmful,
> though, I may change it.)

It multiplies by four the amount of traffic you generate during
transient routing loops.  But even multiplied by four, this remains
a negligible amount of traffic.

Let me explain.  In a stable Internet, there are no routing loops;
every packet follows a path to the destination, or is blackholed early

However, the Internet is not stable: routers crash, transatlantic
links are nibbled by sharks.  So the routing protocols need to switch
routes when the previously selected route to a given destination changes.

Now most of the routing protocols used in the Internet do not
guarantee loop-freeness during reconvergence (OSPF does not, BGP
almost does, EIGRP always does).  For that reason, every IP packet is
equipped with a hop count (called a TTL for historical reasons) that
is decremented whenever the packet is forwarded.  Thus, an IP packet
stuck in a routing loop is forwarded at most MaxTTL hops.

You want MaxTTL to be as low as possible (so that such packets are
dropped as early as possible), but to be larger than the longest path
in the Internet.  The current default of 64 should be good enough to
cover the most extreme cases in today's Internet; setting it to 255
will uselessly forward your packets 255 times before deciding we're
stuck in a loop.