Comment by kstrauser
5 days ago
CGNAT is nowhere near the common case yet. And frankly, I’m horrified that anyone’s describing it as a good thing. CGNAT is the devil, even if it accidentally has one not-terrible feature, and especially when ISPs realize that they can sell those NAT logs to companies who still want to track end users.
For tracking purposes, an IPv6 address is 48 bits long. That’s what identifies a customer premise router, exactly like a IPv4 /32 identifies one. The remaining 80 random bits might as well be treated like longer source port numbers: they identify one particular connection but aren’t persistent and can’t map back to a particular device behind that router afterward.
>CGNAT is nowhere near the common case yet. And frankly, I’m horrified that anyone’s describing it as a good thing.
For some reason, "CGNAT == privacy" is a very common sentiment on Hacker News. Yeah, Hacker News. It's bewildering, and after my last comment [0] talking about it, I have kinda already given up trying to convince people that CGNAT is devilish and not at all a privacy protector.
[0]: https://news.ycombinator.com/item?id=40180058
It’s right up there with “NAT == security”, which is also disappointing for here. It’s not so much the sentiment, as how confidently it’s asserted.
Without NAT my computer isn't on the internet, because my ISP only affords me one IP which my router uses. If it's not on the internet, and adversary can't send my computer any packets.
With NAT, an adversary can't send my computer any packets either unless I explicitly set up port mappings.
So, if you can't send my computer any packets, how is it not providing security?
Of course, it doesn't provide full security like a firewall can do, since there's ways to punch holes in the NAT from the inside. But it seems just as incorrect to fully dismiss "NAT == security".
NAT provides some functional security. It is not a replacement for a proper firewall.
2 replies →