On Mon, Feb 19, 2024 at 10:22 AM William Herrin <bill@herrin.us> wrote:
Yes and no. The client application has to be programmed to understand link-local addresses or it can't use them at all. You can't just say "connect to fe80::1." Even if there's an fe80::1 on your network, it doesn't work. The client app has to also carry the interface identity into the stack (e.g. fe80::1%eth0) in order to use it.
Sure, you and I know this, as a network engineering fact. But, all over the US, thousands of taco trucks (Joe's or otherwise) are using Square and similar solutions, and I happen to know from pcaps that they are (at least some of the time) using the method I described. So everything else we discuss is kind of academic; Joe will continue printing receipts for taco orders over link local addresses just fine, since it works in production today. We can talk all day about how it's not optimal, has limitations if you have 4000 Chromebooks, etc., but Joe won't care, because he is selling tacos. Businesses (not enterprises) that need dual WAN will fall into this category 99.9% of the time. I guess the point I'm making is, the methods we are using today for v6 dual WAN, work fine for most people. There isn't really an advantage to using v4 NAT. That was the original topic I was responding to... as it is visible fuzzily in the rearview mirror currently.