Subj : Connection Tests
To : Michiel van der Vlist
From : Victor Sudakov
Date : Mon Apr 24 2023 01:20 am
Dear Michiel,
15 Apr 23 09:28, you wrote to me:
MV>>> In IPv6 avery device has a Unique Global Address, so one
MV>>> can simply create pinholes in advance as needed for the address
MV>>> in question.
VS>> Only when you know the IPv6 address and port beforehand.
MV> When runing servers you normally do...
P2P apps like Transmission are not really servers.
Well they are in the strict sense of the word, but people just start them up
and hope for them to work out of the box, and they are often configured by
default to randomize port numbers on each start.
VS>> Usually an IPv6 address on the home LAN is dynamic (SLAAC),
MV> No. SLAAC addresses are not dynamic. They are derived from the MAC
MV> address.
Not any more. AFAIK the recent implementation of SLAAC uses the privacy
extensions which do not use the MAC address but some random numbers to derive
the IPv6 host address.
VS>> and the port in peer-to-peer applications, VoIP applications etc
VS>> is often dynamic too.
MV> VOIP normally uses standard ports.
SIP (the signalling protocol) does, but the RTP uses random ports. A firewall
has no way to know the RTP dynamic port numbers unless it inspects the SIP
protocol.
VS>> The situation is different of course when you are hosting an IPv6
VS>> web-server or something like that. It would have a fixed IPv6
VS>> address and port anyway, so there is no need for punch-holing the
VS>> firewall.
MV> Indeed.
I don't really understand your point. If we decide that UPnP (think "automatic
firewall configuration from the inside") is desirable for IPv4, then it's
desirable for IPv6 too. If we decide that UPnP is not desirable, you can do
without it in IPv4: just configure a static RFC1918 address and port on your
internal "server" and create a static NAT/portmapping entry on the router.
