Also, you should take a look at this thread in the archives starting last September -- "IPv6 capability and name lookup cost" -- to see if somebody brought up some useful points that I can't remember at the present time.
I read the thread. I am beginning the implementation using IPLookups as the configuration parameter, and will be providing an API call for modules to use to access the mask information.
I am currently poking around the core_dir_config structure and wondering if I should be shy about adding another bitfield. Some of the options are specified as parts of bitfields, and I want to make sure I don't mess anything up. I'll go ahead and make it an unsigned int :4 for now unless there is some reason that I shouldn't.
Another note: while reading through the APR code and the aforementioned thread, I noticed that the suggestion was made to allow "IPv4Only" and "IPv6Only" options, but it was never really followed up on. I can add these easily enough, but they affect a different setting (family instead of flags), which in my opinion muddies the waters for module writers. I'm considering just leaving them out (and agreeing with the afterthought that they are really an error path optimization, so they don't really matter).
By the way, thanks for taking this on!!!!!! This was a missing bit of reality check in the current support for IPv6. The usual two round drips to the DNS really sucks.
No problem. I'm happy to help out, especially since it helps me, too. ;-)
C
