[Author Prev][Author Next][Thread Prev][Thread Next][Author Index][Thread Index]
Re: Proposal 144: enforce distinct providers
On Thu, 3 Jul 2008, M Fr wrote:
First, because this excellent paper
http://freehaven.net/anonbib/#feamster:wpes2004 as already
demonstrated that it's not necessary because of the BGP logic of
traffic repartition between peering and providers.
That paper doesn't address the question of how exchanges fit into the global
routing structure, and also makes naive assumptions about how traffic is
repartitioned.
The objective is to prevent people observing a circuit with all
routers in the same provider area.
Further, without correlation of leaf nodes that are partially or completely
subsumbed in a given AS, changing AS numbers isn't really indicative of any
useful characteristic. For that matter, without correlating all of the AS
numbers owned by a given entity (an interesting challenge, to be polite),
there's no guarantee at all that a changing AS reflects anything at all.
Please check the list referenced
http://as4jtw5gc6efb267.onion/IPListbyAS.txt if you find many
redundant providers. I've not found many cases in fact two (Road
Runner and one chinese operator)
For more than 90% of nodes it's efficient.
I'm looking at that file, and need some clarification. It looks like the
file consists of a list of tor nodes and their associated AS -- is that
correct? Also - over what time period was this information collected?
Perhaps as i've said the proposal could be improved making AS family
but cos i've no ideas on circuit build in china, i've not added any
restrictive proposal.
And also I don't want to increase network latency.
This proposal is supposed to be iso-latency.
To spell it out - even looking at the list that you've provided, I can
see multiple cases where you have a node that's on an AS which hangs off
of one or two entities already listed, so you've immediately got an overlap.
There's also a notable lack of some of the tier-1 ISPs -- and that's before
even considering the question of physical vs virtual circuits.
Beyond that, if you're still talking about classful address space in this
day
and age, I'd suggest that some consideration of modern networking might well
be in order...
Sorry English is not my mother langage, sometimes i use the more
simple language for me.
FYI: http://en.wikipedia.org/wiki/CIDR
How can you improve this proposal ?
I'm honestly not certain that I can think of a way that this proposal can be
improved enough to be a help, rather than a hinderance.
cheers!
On Tue, 1 Jul 2008, Nick Mathewson wrote:
Filename: 144-enforce-distinct-providers.txt
Title: Increase the diversity of circuits by detecting nodes belonging the
same provider
Author: Mfr
Created: 2008-06-15
Status: Draft
==========================================================================
"A cat spends her life conflicted between a deep, passionate and profound
desire for fish and an equally deep, passionate and profound desire to
avoid getting wet. This is the defining metaphor of my life right now."