Nick Young

Agreed, this _cannot_ be only at the HTTPRoute level, as it requires a HTTP3 listener at the Gateway level. At a _minimum_, this requires defining a `HTTP3` Protocol in the...

Having one listener listen on two ports (TCP and UDP) does not fit with our existing Listener semantics, which map one port/protocol combination to one Listener. "HTTPS" protocol current implies...

@lianglli, I understand how HTTP3 works, and the point you are making, and I disagree. I don't think that we should overload the "HTTPS" protocol to silently include HTTP3 behavior,...

We are very interested in providing a competitive Gateway API implementation, that's why we have one, and it already has a lot of supported features. However, we also have the...

To address particular feedback that doesn't need a rerun of the benchmark: > - 🚨 After the generated Envoy configuration reaches 1.5mb (etcd limits), the proxy configuration is unable to...

The only way that you _don't_ do at least some sharing of proxies between Gateways is if you have a completely separate Envoy Deployment per Gateway, which neither Contour or...

> I'd be happy to add a comment indicating Cilium's opinion on that matter to give users both viewpoints. Unfortunately, this category was the one purely subjective category which is...

Thanks for this @lianglli, this seems like another GEP, if a small one.

Hmm, you could conceivably do #2895 and this request together, maybe. @robscott, @shaneutt, thoughts?

The problem here is that, while ingress-nginx defines ImplementationSpecific path as meaning the same as Prefix, it doesn't have to, and Cilium's Ingress support has ImplementationSpecific mean "Regex match" instead....