Congratulations to Germany, Netherlands and Portugal ;-)

[Martin Millnert]

On Fri, 2012-12-14 at 00:57 +0100, Daniel Roesen wrote:
> On Fri, Dec 14, 2012 at 12:51:32AM +0100, Martin Millnert wrote:
> > You can however set up port forwarding on a port or two in a CGN
> > device's public IPv4s to allow incoming sessions to be established
> > towards a subscriber's private IPv4 address.
> 
> a _specific_ one, yes.

I failed to mention the use case was indeed statical mappings, user
managed, and not a UPnP replacement.

> > It could even be managed from some web interface/portal by customers
> > themselves.
> 
> Yes, that concept is called PCP - Port forwarding Control Protocol and
> being actively worked on. Still, every NAT pool IP only has one TCP port
> 80, so usefulness for forwarding to server services on "well-known" (not
> the traditional definition of that!) ports is limited. It's more useful
> to allow applications to establish an inbound path on some NAT-device
> determined semirandom port, that the application can then signal to
> remote folks as connect coordinates. P2P should e.g. work again using
> that concept.

PCP is largely equivalent to a two-hop UPnP v2, right? This will be
great for applications obviously, provided the CGN port allocator is
sufficiently well-configurable, to limit ports, etc.
But I was talking about human-managed mappings though.

Since IPv4 has run out (in RIPE+APNIC land), obviously (ipv4 address,
proto, port) for any well-known port has as well. Is implied by
"run-out" IMO.  You don't get well-known ports in IPv4 land behind CGN.
That's what IPv6's for.

/M
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 836 bytes
Desc: This is a digitally signed message part
Url : http://lists.cluenet.de/pipermail/ipv6-ops/attachments/20121214/293a1401/attachment.bin