[Date Prev][Date Next][Thread Prev][Thread Next]
- Subject: Re: Send data using tcp/udp to devices behind NAT - UDP Hole Punching
- From: Sam Roberts <vieuxtech@...>
- Date: Sat, 25 Feb 2012 12:57:37 -0800
On Fri, Feb 24, 2012 at 7:12 PM, Pascal J. Bourguignon
> Ross Bencina <email@example.com> writes:
>> On 24/02/2012 6:44 PM, Stelios M. wrote:
>>> The trick is to be able to find out the remote end's IP. In order to do
>>> this, you will need a centralized server to which all peers will
>>> connect. This way the central server will know the public IP for each
>>> node. When a node request a connection to another peer the central
>>> server will pass the public IP and and port number to both peers for the
>>> target node.
>> I wonder is there any generic service that could be used for this.
> What about switching to IPv6 and forgetting NATs?
NAT is turning out to be pretty useful for IPv6. Not for
pseudo-security, or for dealing with address depletion, but because of
scalability problems in global IPv6 routing tables, and renumbering
problems inside corporate networks.
As I understand it, either you use a provider-allocated (PA) address,
in which case you get to renumber your entire corporate network when
you go to a different provider (and also, are totally hosed if you
want to have multiple providers), or you use a provider-independent
(PI) address, which means the global routing system needs to know
about all the PIs, which scales poorly, and the backbone's are pretty
nervous about that.
The more I learn about IPv6, the more the simplistic "its just IPv4
with a wider address" summary seems to be a gloss-over.... its really
different from a network management perspective.