- type extensibility is not hard anymore since we worked out the
  unknown rr types doc

- existing lame implementations are indeed an issue, but:

- we're not talking about every dns implementation on the internet,
  only the authoritative name server infrastructure and the iterative
  resolvers associated with mtas (still a lot, but can ignore, eg,
  legacy stub resolvers).

- dnssec is going to push the envelope on this.  why do we care about
  dnssec?

- remember countermeasures by spammers?  attack the dns.  so we need
  dnssec (which we need anyway for other reasons), because this will
  happen.

- protocol use of txt rrs considered harmful.

- using leftmost label(s) in place of dns type codes is fragile when
  used with wildcards.

- dns tree follows administrative lines.  forward tree follows
  authority for names, reverse tree follows authority for address
  space.  forward tree authority tends to follow organizational lines,
  reverse tree authority tends to follow network topology.  imho,
  following organizational lines makes more sense, but one could make
  a case either way.

- a potential wg should spend time specifying the semantics, and
  not syntax. there are enough dns geeks around which can help
  with the syntax specification.