RE: Teredo spec has been published!

["Christian Huitema" <huitema@windows.microsoft.com>]

> Regarding the private usages, it will be probably so easy as to
describe
> them in a new draft, so implementers can decide if support only the
> existing
> RFC or also the private usages (I guess most of the open source
> implementations will be happy to do that).

The Teredo specification, as it stands, is largely "good enough". It
could certainly be made even better. But, before we work on a new
solution, it would be nice to understand the new requirements that we
try to address. 

The current specification is designed for the "capital I" Internet, and
the use of a common prefix pretty much derives from that requirement.
The 32 bit length derives from a requirement was to embed the server's
IPv4 address in the prefix. If you don't do that, you end up with a
reliance on a static "IPv4 anycast" address, which would have to be used
as source address for the packets sent by Teredo servers. This was in
fact the original "shipworm" design. It was rejected by the IESG because
using "anycast" as source causes hard management issues. In any case,
obtaining a /32 bit allocation from the IANA was very hard, and I don't
think we will see another one anytime soon.

-- Christian Huitema