It's only needed as the initial deployment option. +5 years
from now it
will be a mistake that we are stuck with for another +20 years.
I think it is reasonable to attempt to get deployment of DNSEXT as
collateral benefit from MARID. But don't push it any further.
If you have time to spend planning deployment strategies, concentrate
them on something that matters like DNSSEC or IPv6.
This conversation sounds remarkably similar to the constant discussion
on 'correcting' the spelling of the referer field. Like at this point
my spelling is the one that is used a trillion times a day, so the fact
is that the next version of the OED is going to use my spelling as a
valid alternative to the legacy one.