Proof by reputation doesn't cut it here. He needs to demonstrate either that it is impossible to reasonably configure Squid to do very little paging or, in the alternative, that a non-paging instance of Squid performs unimpressively compared to Varnish -- at least to support his stated thesis. This is not to say either Squid or Varnish is better, only that his criticisms of Squid appear to be unfair or naive or very poorly stated on their face.
Lots of people have run benchmarks that show exactly that - non-paging, in memory working set Varnish clobbers Squid. The difference is even more noticeable when the working set is larger than memory. Most of those benchmarks are micro-benchmarks that can be criticized. To the extent that we have data, Varnish is much faster than Squid. It's not just PHK's reputation.
In all seriousness, can you please point me ideally to a peer reviewed comparison from a decent forum or at least what you consider to be some of the best examples outside of the reviewed publication stuff?
Benchmarks are freaking hard to perform and interpret and, sorry, I don't buy the reputation of "lots of people" either. Not saying you're wrong, just that you haven't convinced.
