Request Processing Model of HTTP Intermediaries

Does anyone know of an overview or comparision table of the common HTTP intermediaries (caches) such as Squid, Varnish, TrafficServer,...?

I am looking in particular for information about their request handling approach (sync vs async,开发者_StackOverflow中文版 multi-process, multi-threaded etc.)

(I am investigating options I have with regards to ESI. Where my ESI-approach might involve executing a substantial amount of logic during request handling)

Jan

One of the big complaints I've read so far is that Varnish processes (replaces) the ESI parts of a (cached) response sequentially. I'm not aware of a concurrent processing implementation, but it would be interesting to see a comparison of the solutions.