[Cryptography] Apple's iCloud+ "VPN"

Ray Dillinger bear at sonic.net
Tue Jul 6 22:47:46 EDT 2021



On 6/30/21 2:39 AM, Bill Woodcock wrote:
>
> And that puts us right back to what I said in the last email.  There’s no magic here.  Adding two new segments to a path cannot decrease the length of the path.  There’s no such thing as a negative-length segment.
>
>
Just to point out a thing because it is true - without any real
knowledge of whether it is relevant in this case;  The length of a path
through a network is logarithmically proportional to the number of nodes
in the network - thus more nodes absent any other change always means
longer path. 

But the base of that logarithmic function is defined by the number of
network interfaces each machine has available.  For example if each
machine can be directly connected to fifteen other machines, then a
network of sixteen machines has diameter one.  But if each machine can
be directly connected to exactly two other machines, then a network of
sixteen machines has diameter fifteen. 

The point on which I don't really know whether this is relevant, is
this:  How does the connectivity between machines in the routing mesh
today compare to the connectivity from back when the net
was half this size?  If the connectivity per machine in the routing mesh
is much higher now then the network diameter may indeed be smaller.

                    Bear




More information about the cryptography mailing list