Oh, and readers of source-changes may have observed that I *did* note that
the `unsigned long' was `wrong, but harmless'.

> 	1. match the existing `unsigned long' in arpa/inet.h
>
> 	2. change arpa/inet.h for susv2 compatibility, and track this change
> 	   in libkern.h, libkern/inet_addr.c, and various lib/libc spots. But
> 	   this is dangerous, as we have made several LP64 releases to date,
> 	   and it would have to be investigated carefully.