🚀 go-pugleaf

In article <3FCC5D1A.6ADAA453@welho.compromised.invalid>,
Lassi =?iso-8859-1?Q?Hippel=E4inen?= 
	 <lassi.hippelainen@welho.compromised.invalid> wrote:
:NAT is not a solution, it is a problem. With NAT boxes you can't have
:peer-to-peer connections. Only the client-server model works. Each time
:you want a direct connection, you have to use a bridging server provided
:by your ISP, at a fee. And that server may become a bottleneck and a
:single point of failure. The perfect target for DoS.

A non-trivial problem with peer-to-peer networks is the discovery
protocol: how does one find out the address and port of another
installation without using a central registry? Once you have solved
that, then if your NAT box allows the creation of translations
that are not restricted as to who connects to them, then you
have no more problem then before.

It isn't NAT itself that is the problem for peer-to-peer: the
problem is NAT implimentations that do not permit inside systems
to create connections that are accessible by random outside hosts.
And the work-around is not using bridging server "provided by
your ISP, at a fee": the work-around is to extend the discovery
protocol to dynamically find a machine that is willing to
act as a bridging server. Not -everything- is behind blocking
NAT implimentations.
-- 
   Cottleston, Cottleston, Cottleston pie.
   A bird can't whistle and neither can I.  -- Pooh