[rbldnsd] ipv6 support: beginnings of an ip6trie dataset

[John R Levine]

> With modern processors, their multi-level caches, and the increase in
> CPU speed relative to main memory access speed, I think cache
> efficiency is the main thing to keep in mind when looking at this sort
> of code.

That and page faults.  The fewer pages you use, the more time you can 
spend doing work and less time waiting for the disk.  I agree that smaller 
is better in this application.

> It is possible that my common_prefix() could be optimized to do
> comparisons by words rather than bytes, at least on architectures
> which can efficiently load non-aligned words.  Then again, I suspect
> that most compilers are probably doing that for us already (when
> optimization is enabled).

GCC optimizes memcmp() into inline code using whatever the target machine 
provides, e.g. repz cmpsb on x86.  I wouldn't try to do better than that.

R's,
John