SSU RelayRequest via IPv6

[zzz]

ref: SSU spec, RelayRequest? section

from an IRC discussion w/ orignal

Spec allows it but unimplemented in Java or i2pd. When IPv6 was implemented, I had concerns about the security of supplying the IP in the RR, and so left it TODO. We could verify against the RI. See IntroductionManager?.receiveRelayRequest() and PacketBuilder?.buildRelayRequest() for the places to change.

Orignal also pointed out we didn't accept an IP or port via IPv4 even if it was the same as the conn. Spec sortof implied that but wasn't clear. Fixed in 0.9.23-6 to accept an identical IP/port. Fixed spec.

[zzz]

<zzz> after looking into it, i can say that the docs are basically correct, but you are right about finding some inconsistency about v6
<zzz> docs say that bob/charlie cannot be v6, but then alice would not publish a bob v6 introcuer, and then why would alice be connecting to bob via v6…
<zzz> ok. the reason that the bob/charlie connection must be v4, is that if it were v6, bob has no way to verify charlie's v4 address. He would have to get it from Charlie's RI and "trust" it
<zzz> that sounds like a possible security issue where bob could be used for amplification attacks, so we disallow it

<orignal> so, what's a possible scenario of ipv6?

<zzz> if we continue to disallow bob/charlie v6, then there's no scenario for alice/bob v6, unless (as you pointed out) alice does a reverse lookup of bob's v4 IP/port in the netdb, then looks for an active SSU connection, finds it, and it happens to be v6
<zzz> so there's not really a possible alice/bob v6 scenario right now

<orignal> thanks. that what we wanted to know
<orignal> actually ipv6 relay is worthless for now

<zzz> correct. But maybe my fears of amplification attacks are silly. Not sure. UDP/SSU has a -lot- of security issues to worry about from source address spoofing.