[LoganDark]

> I will raise you the opposite point: why deprive people of their ability to have a globally addressable IP address?

I wouldn't. I just don't understand, if the alternative is having no internet access at all, why CG-NAT is so utterly deplorable.

> This kind of NAT is NOT hole-punchable. And because you don't control the NAT, you are simply SOL if one day your NAT decides to switch to it.

Can you clarify what you mean by hole-punchable? If all else fails, just use TCP, right? Does TCP also not work? I'm also not talking about connection between peers but connection to a server. Connection between peers has never been a 100% reliable strategy regardless of anything.

> You haven't faced an overloaded CGNAT gateway, have you? [1]

I have not, but that is not inherent to CG-NAT, is it? Any switch or other hop between you and your destination can be overloaded. The destination itself can be overloaded.

[orangeboats]

>Can you clarify what you mean by hole-punchable? If all else fails, just use TCP, right? Does TCP also not work?

I... uh, what? Please... learn more about hole punching before trying to engage in the topic.

Hole punching, in the context of NAT, is a technique where you establish peer-to-peer connection between hosts behind a NAT.

It does not matter which protocol you use, UDP or TCP or chuckles SCTP. If you want to establish P2P connection, you must hole punch.

The only alternative is to use relays.

>I have not, but that is not inherent to CG-NAT, is it? Any switch or other hop between you and your destination can be overloaded.

A typical hop does not need to maintain a huge dynamic state table. NAT, due to its very own temporal nature, must do so.

>destination itself can be overloaded.

Apples and oranges. Destination overload is a service problem. Hop overload is an infrastructural problem.

[LoganDark]

> Please... learn more about hole punching before trying to engage in the topic.

I'm not engaging in the topic of hole punching though? The topic is whether CG-NAT has drawbacks other than lack of port forwarding. As I've said many times, expecting P2P connectivity has never been viable. But you ignore that and keep talking about how hard hole punching is, as if it's indispensable. What makes it so indispensable? Why is it so critical?

> Hole punching, in the context of NAT, is a technique where you establish peer-to-peer connection between hosts behind a NAT.

Good, that confirms I was never talking about that. I even explicitly clarified I was not talking about that (though you may have loaded my comment before that edit.)

> It does not matter which protocol you use, UDP or TCP or chuckles SCTP. If you want to establish P2P connection, you must hole punch.

You don't need to establish P2P connection so I don't see why that's such a problem. Again, it has never been safe to assume P2P connection is possible. Period. It is merely a progressive enhancement.

[orangeboats]

>The topic is CG-NAT and port forwarding

You don't mention port forwarding without mentioning about hole punching.

Because what port forwarding is for, if not to ease the establishment of direct connections?

>You don't need to establish P2P connection

If you are seriously suggesting Server-Client Is All You Need (TM), I feel we might as well stop the discussion now. VoIP essentially requires P2P, WebRTC is much better with P2P. BitTorrent etc obviously runs on P2P.

Services that provide relays (for people who can't establish P2P connection) for free, can only do so because they expect most connections to NOT go through the relay, and so they could simply stomach the costs of running one small relay.