[AusNOG] IPv4 Exhaustion, APNIC EC, and James is a nice bloke ; -)

[Adrian Chadd]

On Fri, Aug 01, 2008, McDonald Richards wrote:
> Because by participating in the protocol you are indicating adherence to the
> specification where as NAT is just a giant hack.

I suggest looking at some of the "faster" HTTP proxies out there (read: the
ones much faster than Squid) and do some protocol adherence testing on them.

Some (eg pound?) don't even try to participate in the protocol - they do the bare
minimum required to get data shovelled down TCP based on some URL trickery
and thats it.

Others (eg varnish) implement enough of the protocol to suit their immediate
needs but don't handle everything (eg, varnish has/had a hard coded number of
headers to avoid allocating a seperate array for em; any further reply headers
are silently dropped.)

So maybe you're indicating adherence to the specification but there's certainly
no requirement for you to implement it.

(Dare I ask what you think about client implementations of stuff? Eg,
if you connect to a remote host on port 80, and issue what looks like a valid
HTTP request, would you expect a HTTP reply? Becuase if you do, and you think
proxies can safely assume this, let me introduce you to mr Shoutcast..)




Adrian