draft-ietf-dnsext-message-size-00.txt specifies that, for A6-capable
	DNS servers/resolvers, we must support EDNS0.  it seems that,
	from IETF48 meetings, it is strong consensus to make EDNS0 mandatory
	for IPv6 transport-ready DNS servers and resolvers.

	NetBSD's libc resolver code supports IPv6 transport DNS queries.

	I'd like to bring in partial EDNS0 support into netbsd libc resolver,
	for both -current and 1.5.
	"partial" EDNS0 support works like follows:
	- if there's "options edns0" in /etc/resolv.conf line, attach EDNS0
	  options to DNS queries.  the option should be specified if and only
	  if we know that the nameservers (on "nameserver" line) are EDNS0-
	  capable.
	- if there's no option, behave normally.
	is there any objection to the change?

	Full EDNS0 support requires negative caching (so that, if our server
	chokes with EDNS0, make resolver retry without ENDS0).  it is too much
	to ask for libc resolver, and it does not make sense to put it into
	libc as the cached result will go away when process terminates.
	I think it make sense to wait till BIND9 lwres daemon for full EDNS0
	support.

itojun