A ipv6 test net sounds great, but I agree getting out a test net on ipv4 is more important. For this first test net iteration, time is of the essence, imo.
Maybe the community test net can stick with ipv6 and see how that goes? Then we have both options running in parallel.
I do not have ipv6 here on the internet facing connection. We are still on ipv4 with our ISP, the block of IPv4 addresses they have will cover all their customers for a very long time, so there is no incentive for them to move us to IPv6.
Still cli 0.17.2 and node 0.25.39.
Last night an OS update of my second Safe network machine resulted in the OS crashing, so today I’m going to reinstall the OS from scratch and test the newer versions.
Until now all test networks were IPv4. Launching a test network with IPv6 would be interesting to get information under this condition and possibly identify new issues.
IPv6 is ready right now, but there are things that remain to be implemented to properly support IPv4 in cases needing NAT. You mentioned earlier:
These use cases are given for free with IPv6, but it is important to check that they can actually work in the real world and not only in the lab (independently of the protocol).
My thinking is that it is quite easily done as @folaht has shown. It does allow much simpler code (no NAT) and also something that’s valuable, clients would be connectable. So push notifications can be done with the network not having to handle state (caches of user data etc.) and also easy audio/video conferencing as we can have a simple data type that is a single value, key=UserPubKey and value = ipv6 address (personal DNS). Clearly though the coverage is not good enough, but will that improve faster than we can improve NAT traversal? It’s worth discussing and possibly testing.
Dual v4/v6 is a nightmare though. 
I just found this: https://test-ipv6.com/
I’m not sure how good/bad it is, but I got a 0/10 rating!
You appear to have no IPv6 at this time…
I’m not sure whether I could request one or not (I have TalkTalk broadband), but it’s a shame it isn’t just working out of the box.
EDIT: From September 2019 https://community.talktalk.co.uk/t5/Fibre-Broadband/IPv6-availability/td-p/2406789:
TalkTalk is committed to transitioning to IPv6. To minimise any potential impact to our customers and services, the transition will be phased over the next few years as we roll-out new products and services such as fibre to the premises.
We continue to work closely with our network and systems vendors to ensure all new equipment is fully IPv6 compliant. In fact, our network already operates as dual stack, using IPV4 compliant addresses while the major network components are IPV6 ready.
Blah, blah, excuse, excuse, blah… 
Edit: Another test here: https://ipv6-test.com/ (I get 3/20 - no dice!)
I wonder if it would be worth starting a poll to see who actually has ipv6 on this forum?
Works for me, interesting to see how many in the community it does work for? Then it’s can we route 4->6 or get widespread NAT handling (the plague of p2p network for years). Or take the hit after launch to dual the network?
NAT handling seems to cause all sorts of weird/inconsistent behaviour. I know my router doesn’t play nicely with it (said TalkTalk freebie) and from following the issues on Safe Network, it looks like I’m not alone.
I wonder whether IPv6 would be more consistent. Is it more likely to either work or not, rather than have inconsistent/weird issues that are hard to resolve? I wonder whether IPv6 may just cause fewer headaches, even if support is less complete?
This is what I am wondering right now, can we tunnel 4->6 with stuff like How to route IPv4 via IPv6 | ungleich.ch and go v6 and bite the bullet?
The issue is ip6 means client need to be v6 and that could be a killer. Folks running nodes might take some small steps, but clients wont.
I think it could be interesting to start a poll, no IPv6 for me.
Would it be possible to insist that nodes communicate with one another over IPv6, but clients can attempt to connect with either IPv4 or IPv6?
I’m not aware of how the communication logic differs between the two, but given a client needs to communicate with fewer nodes, is it feasible?
Kinda, this is where complexity starts though. Clients need to speak to Elders and with AT2 etc. really all Elders. So Elders prob need to be v4 capable. That means we had to fix the NAT for nodes anyway. If nodes could be v6 and also have available v4 public addresses via some tunnelling or similar then perhaps? It’s a bit of a jungle though.
Added a poll here. I’m not sure if I have done it right, but it seems to be ok!
Many places ISPs have only enabled ipv4 by default, but will allow technical users turn on ipv6 support manually. I would guess it’s like that in Sweden too.
In Sweden it is often electric power companies who owns the cables in cities, that goes to different internet boxes, before reaching the customers. Those boxes are not ipv6 ready so ISP’s can’t turne on ipv6 for customers.
With shortages of ipv4 addresses they put users behind a cgn and let’s users share ipv4 addresses. A switch to ipv6 might happen in 2022 for Sweden but who knows about the many other countries in the world.
Tested both sites @tfa provided and in my rural area no ipv6. Bummer. I have such slow speeds I would be a poorly ranked node anyways. That might be something to consider while the worlds ISPs catch up to providing ipv6 anyways?
Sounds really beneficial to development in a way.
I don’t want to be a stick in the mud but, is what we currently have in place so sketchy that pivoting to ipv6 on testnet eve is an actual consideration?
Or is the suggestion 2 separate test networks.
From the poll the vast majority of us dont have ipv6 so I am scratching my head.
I’ll admit that my understanding of it all is well below par so maybe an ELI5, if someone has a minute… please.
Not at all, running an IPv6 testnet is change a couple of config files and run. It’s not a big deal in itself. If we were free of NAT though it would make a big difference to code size etc. the counterpoint is ip6 coverage is as bad or worse than the folk NAT won’t work for.
Clients (apps) are a diff kettle of fish, they will all be fine with ip4 as they don’t use/need NAT traversal.
So interesting experiments atm, no more no less.
Absolutely, there’s no suggestion of ditch IPv4 here and switch, it’s only a test and separate network.
Thanks David!
Back to my cycle now… I was so distressed I had to stop and clear that up 
