I have had some requests about what capabilities the squid3-ipv6 work has so here are some details for the testers and other interested parties.
Please keep in mind that this post describes a completely theoretical set of capabilities. While they are supposedly possible now. They still need testing to find the bugs that might stop it being so.
I have so far been concentrating on the client side of squid. Where it locates the pages requested.
Features page is at:
Combined the bits roughly make a request-path from a URI -> DNS -> ACL -> Remote Server and back to storage. (NOTE: this does NOT include Auth). It is not yet listening for requests on IPv6 from clients.
There are two important points to notice: Firstly clients still have to be in IPv4, Secondly the URI being in the path allows them to request domains or direct IPA that are within IPv6 netspace.
The goal and jackpot for testers is to see the dancing turtle at http://www.kame.net from a machine in IPv4 through squid.
Squid3-ipv6 needs presently (for IPv6 to work) to be running on a machine with both IPv4 and IPv6 addresses, v4 to receive requests and v6 to test how well it copes there. I won’t go into the setup of IPv6 on a machine here.
squid.conf can contain IPv6 addresses if an IPv4 was previously valid there AND if the configured feature is in the experimental list. IPv6 equivalent of ipv4:port is [ipv6]:port
debug_options ALL,1 14,9
Right now there are around 4-dozen points in the code where it unconditionally requests an IPv4 address to be used from that point on. Please report to me any info squid spits out about bad .conf options or stack backtraces of where it says “Cannot convert non-IPv4 to IPv4.” (its expected to crash there for now) along with any other bugs not in the squid bugzilla.