►
From YouTube: IETF95-HOMENET-20160405-1400
Description
HOMENET meeting session at IETF95
2016/04/05 1400
A
D
E
C
D
D
H
A
D
Okay,
let's
make
a
start!
Thank
you
all
for
coming
the
like
to
welcome
the
survivors
of
the
home
at
protocol,
routing
walls
to
Buenos
Aires
right,
so
we're
down
quite
a
few
of
our
reggin
appetitive
participants.
This
time,
I
think
this
must
be
the
quietest
home
it
session
ever
so,
let's
hope
that
they'll
be
more
of
us
in
Buenos
Aires.
So
let's
get
a
few
administrative
bits
out
of
the
way.
D
Thank
you
Tim
and
Tim
for
volunteering
to
take
the
notes
and
do
the
job
a
rillo
respectively,
Timmy
Shinsky
is
doing
the
notes
and
he
is
using
etherpad
and
he
would
welcome
other
people
contributing
live
edits
while
he's
doing
that
as
well,
especially
if
you
happen
to
get
your
name
wrong,
so
feel
free
to
go
in
and
fix
the
names
if
you've
spoken
and
you
didn't
get
your
right,
the
blue
sheets
are
already
out.
So
if
they
can
happen,
please
start
circulating
that'd
be
great
and
sorry
I'm.
D
D
So,
as
a
small
agenda,
bash
mark
I
will
do
an
update
on
what's
happening
with
the
various
working
group
documents
very
shortly.
We
have
Julius
to
talk
about
the
label
profile
for
home
notes,
Ted
lemons,
going
to
then
speak
to
his
helmet
name
and
architecture.
Draft
which
came
out
early
last
week
actually
slightly
off
to
deadlines,
but
hopefully
had
sufficient
time
to
read
it.
D
D
E
These
are
the
two
documents:
d
NCP
in
hnc,
p,
being
advanced
together
as
a
package,
we've
been
cc'd
on
the
various
emails
with
the
RFC
editor
and
the
authors,
etc.
We're
seeing
a
progression
happening.
I
imagine
this
will
come
to
the
club
come
to
close
very
very
soon.
E
You
all
see
also
see
an
update
to
the
routing
consensus
call
document
that
was
back
when
we
were
working
with
our
area
director,
terry
on
exactly
what
we
were
going
to
state
from
the
working
group
and
from
the
80s
in
terms
of
what
to
do
about
a
mandatory
to
experiment
with
routing
protocol.
We
emailed
the
list
with
some
carefully
worded
text
along
those
lines
we
discussed
it
at
the
last
meeting.
E
E
You
have
the
internet
draft
out
there,
let's,
let's
just
leave
it
at
that
and
we
were
fine,
but
we
wanted
to
update
the
document,
the
internet
draft
to
say
that,
okay,
to
say
that,
oh
by
the
way,
this
is
just
sitting
here
expired
because
nobody
cares
it's
because
we
explicitly
decided
to
just
not
advance
it
anymore,
because
the
the
email
to
the
list
and
the
discussion
at
the
meeting
last
time
was
sufficient
in
his
mom.
So
that's.
Why
that's
there?
E
We
have
some
new
drafts,
one
of
which
Ted
will
be
talking
about
later
today
and
Julius
will
be
talking
about.
Basically,
the
two
new
drafts
are
what
you're
going
to
hear
about
on
the
agenda
today,
Jeff's
hold
putting
other
documents,
we've
got
some
stuff,
that's
stuck
because
it
references
other
working
groups,
in
particular,
hybrid
proxy
stuff,
and
things
related
to
the
naming
architecture
that
clearly
ray
wants
to
say
something
about
yeah.
D
The
the
two
drafts
from
dinamico
&
Co
on
the
front
head
naming
delegation
and
naming
architecture-
one
of
those
has
in
fact
just
recently
expired,
but
position
we're
taking
on.
That
is
that
we
don't
see
any
point
in
advancing
those
Wiles.
The
overall
picture
of
the
whole
net
naming
so
a
discovery.
Architecture
is
being
rationalizing
that
there's
no
point
writing
something
up
in
those
documents
and
then
we
actually
decide
as
a
working
group
that
it
doesn't
fit
and
is
not
consistent
with
the
overall
architecture.
D
D
I
So
hello,
I'm
julia
scrubber
checks
next
slide.
Please
next
slide,
please
yeah.
So
there
are
two
issues
that
I'm
trying
to
solve
with
the
draft.
When
is
the
table
is
a
flexible
protocol
and
that
different
implementations
interoperate,
but
to
the
extent
possible.
We
then
do
the
I'm
going
to
explain
that
in
a
second
and
the
other
thing
is
that
there
are
some
interactions
between
H
n,
CP
and
label,
and
those
interactions
have
to
be
gotten
just
right.
I
Ok,
so
Babel
is
a
flexible
protocol
and
implementations
can
implement
very
different
things,
so
they
will
interoperate,
but
one
needs
to
understand
exactly
what
that
needs.
So,
to
give
you
an
example,
it
is
quite
legal
for
a
table
implementation
to
only
implement
ipv6
and
not
ipv4.
So
if
you
have
one
implementation,
a
that
supports
ipv6
and
ipv4
and
the
second
implementation
that
only
does
ipv6,
they
will
interoperate
in
the
sense
that
they
will
exchange
ipv6
routes,
but
obviously
they
will
not
exchange
I
key
forwards.
Let's
slide
and
the
bagel
RFC.
I
The
current
experimental
Eric
C
does
not
fix
the
values
of
metrics,
so
it
could
very
well
come
up
with
an
implementation
X
that
gives
a
metric
of
96
to
aillard
links,
which
is
the
case
of
all
implementations,
I,
think
that
exists
now
and
the
second
implementation
that
assigns
a
metric
of
one.
So
those
two
implementation
will
interoperate,
but
of
course
you
will
get
suboptimal
routing,
as
the
links
handled
by
implementation
y
will
be
preferred
to
those
handle
the
implementation
X,
that's
like.
I
So
what
I
have
written
down
is
a
short
document
that
describes
the
subset
of
the
protocol
that
a
helmet
implement.
Eructation
must
implement
as
well
as
suggested
values
that
the
helmet
implement
must
assigned
to
the
parameters.
It
has
seven
requirements,
so
I
used
the
numbered
requirements
style
with
rationale
after
each
five,
which
I
expect
to
be
non
controversial
and
people
are
already
laughing
in
the
audience
when
I
say
non
controversial.
I
That
I
think
has
been
somewhat
contentious
and
one
that
I
don't
understand.
Sir
I've
written
it,
but
I
need
clarification
because
I
I'll
come
back
to
that
next
slide.
Please
so
I'm
not
going
to
go
through
the
five
less
interesting
requirements.
Please
don't
take
my
works
or
that
check
them
in
the
draft,
but
the
one
that
proved
to
be
controversial
is
requirement.
Three,
which
says
that
ipv4
support
is
a
sugar,
and
one
person
expressed
the
notion
that
it
should
be
a
must.
I
My
take
on
that
is
that
I'm
finalists
should
I'm
fine
with
must,
because
people
will
implement
ipv4
for
as
long
as
the
market
needs
it,
but
I'm
fine
was
changing
that
to
master
people.
Think
that's
a
good
idea.
Next
slide,
please
so
I've
cooked
in
a
requirement
site
which
I
don't
understand.
It
says
that
a
helmet
implementation
of
Babel
must
implement.
J
E
What
feedback
did
you
just
get
where?
Where
are
we
I
I'm
hearing
from
lee
and
I'm
hearing
from
marcus?
That
should
is
fine
for
that,
but
your
slide
said
you
would
move
it
to.
Must
wait,
hang
on,
should
oh,
that's
because
it
says,
should
and
must
on
the
same
sentence.
Should
this
be
a
must,
I
think
consistent
with
this
charter
in
the
working
group
it
should
not
be
a
most,
it
should
be
issued.
Does
everybody
agree.
K
E
E
K
I
Okay,
so
as
always,
there
is
this
year,
security,
so
cuz
I
couldnt
in
requirement
five-
is
that
a
helmet
implementation
of
Babel
must
implement
h.
Mag
based
authentication
is
defined
in
RFC.
Seven
to
nine.
Eight
must
implement
the
two
men
retreat
into
implement
algorithm's
and
must
enable
and
require
authentication
when
instructed
to
do
so
by
H
NCP.
So
I
have
no
idea
whether
this
does
reflect
the
opinion
of
the
working
group
and
I
have
no
idea
how
that
is
supposed
to
work
since
I.
I
L
Might
speak
to
that
a
little
bit
after
my
presentation
on
names
tis
Ted
lemon,
but
I
think
that
the
musts
are
necessary
because
otherwise
you'll
have
implementations
that
don't
have
that
feature.
So,
whether
or
not
we're
able
to
turn
it
on
we
need
or
whether
or
not
we
turn
it
on
all
the
time.
We
need
to
be
able
to
turn
it
on
and
that's
what
that
must
says.
J
M
Michael
Abramson,
so
I
think
this
makes
the
perfect
perfect
sense,
because
I
expect
a
chance
if
you
turn
Babel
on
and
off
on
certain
interfaces
like
the
when
you
defy
when
HSP
decides
that
this
is
an
external
facing
interface.
I
would
expect
it
to
tell
Babel
do
not
talk
on
this
interface,
so
I
think
we
already
have
that
agency.
P
must
be
able
to
provision
or
instructive
able
to
do
stuff,
and
this
seems
like
a
perfectly
fine
thing,
for
a
chance'
p
to
instruct
able
to
do.
I.
J
E
K
M
K
Well,
I
would
think
the
way
would
need
something.
Some
part
I,
don't
think
it's
specified
an
agency
p
now,
but
somebody
who
knows
more
than
I
do
can
correct
me
that
we
would
need
a
common
way
for
H
NCP
to
signal
that
it
should
be
that
it
should
be
done,
and
then
the
Babel
implementation
can
decide
how
it's
going
to
implement,
how
to
interpret
that
instruction.
K
I
Shmc
p
right
now
doesn't
speak
to
Babel
at
all,
except
through
the
colonel,
and
we
have
recently
implemented
a
communication
on
the
Babel
side,
not
on
the
h
and
CP
side,
yet
over
a
UNIX,
the
main
socket
so
I
wouldn't
like
to
mandate
how
this
is
done,
because
we
have
different
implementations
and
we
want
to
keep
the
liberty
of
implementing
them
in
different
manners.
Yeah
I
think
this.
K
J
Sir
marcus
seems
to
agree
with
that
what
he
say
on
jabber
but
I'll
relay
it,
so
he
doesn't
object
the
text,
as
is
he's
fine,
with
that
hnc
p,
just
defines
the
shaky
generation
and
that
would
either
be
stuck
to
a
configuration
file
or
towed
by
IPC.
For
example,
bird
demon
implementation
might
support
that
one
day,
but
he
says
that's
just
an
implementation
detail
and
then
response.
J
D
E
D
I
So,
okay,
so
the
well
I
was
writing
that
I
started.
You
know
grabbing
for
the
HNC
p
specification,
and
I
realized
that
there
is
something
that
is
nowhere,
written
down,
which
seemed
to
us
reasonably
obvious,
which
is
that
a
chance
EP
uses
babel
to
announce
roots,
and
you
need
to
get
the
precise
rules
right.
So
if
you
have
received
a
dhcpv6
prefix
delegation,
you
announce
a
source
specific
default
route.
I
On
the
other
hand,
if
you
receive
a
dhcp
v
for
lease,
then
you
announce
a
nonspecific
default
route,
and
so
there
is
an
asymmetry
here
between
ipv6
and
ipv4,
and
there
are
very
good
technical
reasons
for
this
asymmetry.
However,
if
you
get
it
wrong,
you
are
going
to
get
suboptimal
routing
and
you
might
even
have
root
in
black
holes.
So
I
think
this
needs
to
be
written
down.
M
I
I
So
I
have
used
it
so
here
I've
been
lose.
This
is
stuff
that
does
not
written
down
yet
yeah
before
it
can
belong
in.
A
draft
I
need
to
check
very
carefully
both
implementations
to
check
whether
they
comply
and
if
they
don't
why.
So,
here
the
point
of
view
I'm
taking
is
that
hn
CP
is
using
be
able
to
announce
the
roots.
So
if
you
want
to
see
it
as
Babel
as
H
NCP
causes
Babel
to
announce
the
roots,
that's
fine
too.
Yes,.
M
J
N
I
I
I
I
What
I'm
saying
is
that,
if
some
nodes
distribute
is
so
specific
and
others
a
nonspecific
default
route,
you
will
have
root
and
failures
speak
about
dhcp
about
ipv4
only
so
in
ipv4.
If
you
have
some
nodes
that
announced
a
specific
route,
others
a
non
specific
route,
you're
going
to
have
issues,
so
everyone
has
to
agree
that
ipv4
is
nonspecific
or
yes.
E
I
M
E
And
to
be
clear,
this
seems
like
a
general
dual-stack,
source-specific
routing
issue,
we
may
come
up
with
a
solution
such
as
assume
nonspecific
default
route
with
an
ihome
net,
but
it's
a
general
issue,
I
wonder
if
we,
it
needs
to
be
stated
in
some
source
desk
routing
document
in
its
general
form.
I
Ok,
so
what
I'm
suggesting
here
is
that
we
write
it's
in
a
single
two-part
document
that
describes
the
Babel
profile
for
whole
net
or
have
based
on
the
current
draft
and
a
second
part,
that
is
the
interactions
between
H,
NCP
and
Abel,
something
that
is
not
written.
Yet
the
only
written
down
thing
is
slide.
8
of
this
presentation
and
I
would
like
to
suggest
that
this
this
should
be
a
working
group
document.
Thank
you
for
attention.
M
I
One
of
the
requirements
that
I
have
that
I
haven't
put
in
this
presentation:
one
of
the
five
requirements
speaks
about
metrics,
ok
and
what
it
says
is
there
is
a
must
and
eschewed.
The
must
says
that
the
metrics
must
be
no
less
than
96
for
why
or
no
less
than
256
for
wireless
links,
the
shooter
says
you
should
be
using
an
algorithm
on
wireless
on
wireless
links
that
determines
link
quality
mm-hmm.
Okay.
So
if
you
want
to
tighten
this
formulation,
that's
a
reasonable
thing
to
do.
Yeah.
M
That's
what
I
wanted
to
bring
it
bring
it
up
here
is
when
we're
doing
in
a
veil
profile.
It
sounds
like
that
would
be
beneficial
if
table
implementations
behaved.
You
know
similar
in
this
aspect
when
it
comes
to
setting
the
metric
and
not
just
it
should
be
in
the
same
in
order
of
magnitude,
but
it
should
be,
or
you
know
it
should
be
closer,
it
should
be
more
specified
somehow
of
it,
and
this
could
be
a
later
work.
I,
don't
know,
but
I
think
that
should
be
more
specific.
I
First,
we
are
speaking
about
how
networks
here,
which
I
accept
to
have
expect
to
have
reasonable
reasonable
amount
of
diversity,
you're
not
expecting
to
have
ten
different
paths
between
two
routers.
So
if
the
values
of
the
metrics
are
not
exactly
so,
if
one
router
has
one
and
196
you
run
into
trouble,
but
it
only
uses
96
and
another
104
you're
still
going
to
get
pretty
optimal
route.
At
the
same
time,
I
don't
want
to
prevent
people
from
experimenting
with
vetterling
quality
estimators.
So
I
wouldn't
like
this.
I
M
N
Darlin
harder,
so
I
would
just
like
to
voice
agreement,
as
in.
I
personally
think
that
it's
very
important
to
have
for
home
knit
for
the
application
space
of
Home
net
document
describing
what
the
metrics
are
and
what
algorithm
you're
using
I
think
this
is
absolutely
in
scope
for
home
net
and
not
for
the
Babel
of
working
group,
because
this
is
application
specific
to
the
home.
D
Okay,
okay,
so
from
the
chair,
what
will
probably
do
is
issue
a
working
group
called
for
adoption
to
the
mailing
list
just
to
get
the
formalities
out
of
the
way,
I
can't
see
had
personally
reason
why
this
won't
be
adopted.
I
will
respect
to
the
interactions
between
nature
and
CP
and
Babel
tziporah
question
for
Judas
is
just
going
to
sat
down.
Do
you
feel
jealous
that
you
could
do
with
additional
contributors
to
such
a
document?
D
Okay
and
presumably
the
ocean
and
CP
experts
would
probably
be
a
useful
pool
for
that.
So
yes,
so,
unfortunately,
mostly
and
CP
folks
aren't
here
what
for
a
factor
of
any
of
them,
are
here
but
yeah.
We
would
encourage
them
to
work
with
Julius
and
best
guess,
do
as
Julius
proposed
and
get
this
all
addressed
in
that
document.
D
E
L
D
Here
Maine
is
raised
like
that
I.
L
Can
stand
closer
to
it,
how's
that
okay
next
slide
assume
everybody
knows
what
what
I'm
gonna
be
talking
about?
How
many
people
have
read
the
president
this
the
document
by
the
way?
Thank
you
for
doing
that?
Okay.
L
So
I
forgot
to
document
the
multiple
name
space
case.
That's
the
case
where
you
might
want
to
have
more
than
one
sub
domain
and
you're
in
your
home
net.
The
whole
net
document
currently
just
talks
about
a
single
forward
domain,
but
that
was
just
a
an
oversight
and
I'll
add
that
in
and
a
new
version,
I
forgot
to
add
the
remote
device
registration
cases.
L
The
case
where
you
have
a
device
that
you're
using
in
the
home
that
you'd
like
to
be
reachable
in
the
home
that
even
when
you're
not
actually
there,
for
example,
you
might
have
a
phone
that
has
something
on
it,
that
you
want
some
device
in
your
home
to
be
able
to
get
to
reach
out
to.
And
it
would
be
nice
if
there
a
way
to
register
that.
So
that
will
be
added
in
a
future
version
of
the
document.
L
The
one
of
the
things
about
in
the
home
that
architecture
is
a
requirement
for
being
able
to
rename
devices
give
a
device.
That's
got
a
name
in
the
home
net
namespace
a
different
name
and
implicitly,
that's
without
actually
telling
the
device
that
it's
got
a
different
name
and
making
the
device
register
that
different
name,
that's
problematic,
I,
don't
think
it's
possible
with
mdns
there's
a
an
idea
for
how
to
make
dns
updates
work
in
the
document
and
I
think
it
could
be
done
with
that.
L
So
it's
possible
that
we
could
specify
that
and
that's
probably
how
I
would
address
that
and
then
otherwise.
I
went
through
the
home
that
architecture
requirements
again
this
morning
just
to
check
and
I'm
pretty
sure
that
I
hit
everything
so
I
think
it's
complete.
If
you
don't
think
it's
complete,
please
tell
me,
I
think
tim
is
going
to
tell
me
that
it's
not
complete.
J
Tin
shine
is
this:
okay,
I
was
set
up,
told
I
was
overdriving
the
mic
before.
Can
you
hear
me
like
this?
I
can
hear
you
so
I've
got
a
hearing
problem
cycle
yeah.
So
what
about
that?
There
was
the
Christians
talking
came
up
in
DNS
SD.
So
maybe
we
want
to
think
in
this
about
obfuscating
names
from
the
privacy
point
of
view
as
well.
Yeah
so
hit
that
in
somehow
right.
That
is
a
that
as
an
interesting.
L
L
Yeah,
the
problem
with
do
I
mean
I.
Think
it's
worth
it's
probably.
What
I
should
do
is
just
say
that
this
is
something
that
needs
to
be
done
in
its
future
work,
because
I
honestly,
don't
know
how
to
do
it
and
I
think
that
if
we
tried
to
hold
off
on
completing
this
document
until
I
know
how
to
do
it,
but
by
all
means,
if
you
have
a
suggestion,
send
text
so.
L
L
J
J
J
He's
waving
they're
waving
the
fact
church
and
the
other
thing
of
course,
is
the
discussion
of
a
hybrid
proxy
approach
against
the
DNS
update
approach,
where
I
think
in
DNS
SD.
We
agreed
to
take
that
to
a
separate,
maybe
short
document,
to
discuss.
It's
not
going
to
be
a
short
document,
but
ya
know.
J
L
Just
simply
can't
do
it
the
exact
same
way
that
mdns
SD
hybrid,
does
it
and
have
it
work
in
the
home
net,
which
is
fine,
but
what
what
I
think
we
can
do
it
I
talked
about
it
in
a
later
slide
by
the
way.
So
we
could
get
to
that,
then
what
we
can
do
is
come
up
with
a
way
to
do
it.
That's
it's
compatible
with
what
mdn
SSD
hybrid
does
and
that's
what
I'd
like
to
do.
J
It's
just
that
Andrew
Sullivan
on
Java
has
said.
Perhaps
we
should
for
number
of
people
haven't
read.
The
document
should
be
a
good
idea
to
get
the
overview
out
of
the
way
before
we
start
talking
about
things
that
are
missing
right.
L
O
P
O
L
L
L
O
L
O
L
Yeah
I
mean
personally
I
think
that
I
I
think
that
this
isn't
actually
yeah
it'd
be
nice
to
be
able
to
do
this.
But
this
is
not
a
core
piece
of
functionality.
So
yeah,
let's
go.
Let's
go
on
to
the
next
slide.
Sorry,
okay,
good
enough!
So
aspects
of
the
architecture,
so,
first
of
all,
just
on
a
very
high
level.
There
are
the
the
document
is
currently
written,
describes:
two
databases
of
names.
One
of
those
names
is
the
set
of
names
that
you
might
think
of
as
the
forward
zone.
L
L
Currently
it
does
not
support
the
model
for
naming
a
human-readable
sub
domain,
as
is
described
in
the
dns
SD
hybrid
document,
and
a
Stewart
and
I
had
a
conversation
about
whether
or
not
there's
a
better
way
to
do
that
and
Stewart.
So
if
you
go
think
about
it,
and
so
we
that
could
turn
out
in
a
variety
of
ways
but
anyway.
L
So
the
idea
is
to
basically
two
databases
to
two
places
where
data
about
names
are
stored
in
the
home
net
and
if
you
were
to
do
a
multi,
multi,
sub-domain,
homenet
architecture,
both
of
these
well
actually,
no
I,
guess
just
the
forward
would
be
duplicated
or
where
there
would
be
two
of
them.
There's
a
question
of
public
versus
local
public
is
stuff
that
would
be
visible
from
the
internet.
L
If
you
have
a
global
domain
name,
if
the
home
net
has
a
global
domain
name,
you
would
be
able
to
go
type
nslookup
or
dig,
and
then
some
name
and
in
principle
it
would
look
it
up.
Of
course,
we
don't
want
to
publish
necessarily
every
name
that
appears
in
the
whole
net
out
on
the
global
internet
so
that
the
public
don't
domain
of
the
public
namespace
is
kind
of
a
Babel
rised
version
of
the
private
domains
domain
space
with
not
as
much
stuff
in
it.
L
Private
domain
space
is
expected
to
contain
every
name,
that's
published
in
the
home
net.
So
that
way
you
can
discover
all
the
services
in
the
home
that
you
can
talk.
You
know
between
subnets.
You
can
talk
to
devices
in
the
home
that
so
forth,
there's
a
question
that
I
had
as
to
whether
we
also
need
a
guest
namespace.
The
reason
you
might
want
to
do
that
is
you
have
a
friend
come
over
your
friend.
You
want
to
play
some.
You
know
p2p
game
with
your
friend.
L
L
The
current
draft
doesn't
and
then
a
the
document
is
written
with
the
assumption
that
we
want
to
be
able
to
support
the
use
case
where
the
home
that
actually
has
a
publicly
reachable
domain
name
in
the
DNS,
but
also
it
has
to
work
if
there
it
doesn't,
and
so
it
talks
about
a
home
homenet,
a
special-use
TLD
could
be
dot
home
I
I
chose
got
home
that
because
I
was
pretty
sure
that
dot
home
had
conflicts
and
wasn't
a
good
idea
and
tim,
which
Hinske
is
nodding.
Yes,
so
I'll.
L
Take
that
as
having
been
a
good
idea
and
also,
I
actually
think
it
sounds
better
so
anyway.
So
that's
the
basic
high
level
view
of
what
what
names
look
like
in
the
in
the
home
net
and
then
next
slide
so
database.
So
I
talked
about
two
different
views
of
the
database
and
actually
there's
notionally.
Probably
I
was
really
thinking
that
the
private,
the
local
version
of
the
database,
would
be
the
ideal
view
of
the
database
with
all
of
the
data
in
it,
but
actually
I,
don't
think.
L
That's
correct
and
you'll
see
why,
in
a
minute,
so
I
think
there's
sort
of
an
ideal
version
that
has
everything
in
it.
That's
what
gets
updated
it's,
what
we
collect
from
mdns
or
from
DNS
updates
from
services
and
that's
maintained
as
a
database,
which
is
which
is
shared
either
using
an
authoritative
primary
name,
server
or
using
some
kind
of
other
mechanism,
but
I
assume
in
the
document
that
it's
just
enough
off,
NS
an
authoritative
name
server
and
then
the
the
public
and
private,
the
public
and
local
I
said
private
on
the
slide.
L
But
I
that's!
Actually
it's
not
really
private,
because
somebody
could
get
onto
the
network
that
isn't
you
and
they
can
see
it.
So
that's
not
private,
but
the
public
and
local
views
of
the
database
are
probably
also
they
could
be
DNS
zones
that
are
derived
from
Miss
DNS
own
programmatically
and
then
the
public
and
private
names.
Private
zones
have
the
same
external
name.
So
if
they
have
a
global
name,
that
name
is
the
same
and
what
which
zone
you
see
depends
on
whether
you're
at
home
or
whether
you're
away
from
home.
L
L
L
But
in
my
opinion
it's
it's
not
quite
right,
and
there
are
a
couple
of
reasons
that
are
described
in
the
document.
I'm
not
going
to
go
into
it
here
unless
people
want
to.
But
my
feeling
is
we
needed
to
find
something
new.
It
would
work
a
bit
like
mdns.
I
would,
but
not
exactly,
and
ultimately,
what
we
really
want
is
for
servers
on
the
home
net
to
use
dns
updates
with
a
with
a
secure
key
mechanism
that
uses
you
know,
sort
of
first
first
come
first
serve.
L
L
No
devices
currently
do
that
I,
don't
think
any
device
manufacturer
has,
even
you
know,
thought
of
doing
that
and
whether
they
would
do
that
if
we
wrote
a
spec
that
told
them
how
I
don't
know
it's
quite
possible,
they
would
not,
but
maybe
they
would
so
I
I'd
like
to
do
I'd
like
to
do
that,
just
because
I
think
it
would
be
good
to
have
that
option.
Tim
Tim.
J
J
The
other
thing
I've
got
ringing
in
my
ears,
sort
of
even
I
can't
hear
is
two
or
3i
ETFs
ago,
and
someone
gave
the
chairs
of
the
NSS
they're
good
slapping
for
assuming
a
sort
of
english-language
domain
name
for
for
this
you
might
want
to
in
the
document
use
el
q
DN,
whatever
the
language
we
used
in
the
architecture
document
is
rather
than
assuming
it's
home
or
home
net
or
whatever.
Perhaps.
L
L
J
L
Don't
know
how
to
do
it:
okay,
so
global
domain
names.
I
mentioned
that
that,
ideally
or
perhaps
not
ideally,
but
at
least
optionally
home,
that
should
have
a
name
that
the
names
that
are
that
are
known
on
the
home
that
should
be
knowable
from
the
internet
and
in
order
for
that
to
happen,
the
home
that
has
to
have
a
global
name,
a
name
that
appears
in
the
domain
name
system.
And
in
order
for
that
to
happen,
there
has
to
be
some
process
for
making
it
happen.
L
Well
I'll
talk
about
that
in
a
minute,
but
I
don't
think
that's
going
to
really
work
for
a
home
that,
because
you
know,
let's
face
it,
we're
all
much
better
at
dealing
with
registrar's
than
the
average
person
and
and-
and
I
don't
want
everybody
who
has
a
home
that
and
wants
to
have
a
public
name
to
have
to
learn
how
to
talk
to
a
registrar.
So
I
think
that
this
requires
a
new
service
which
I
have
called
you
know.
L
Global
name
registration
provider
and
the
idea
here
is
essentially
that
the
home
that
needs
to
be
able
to
have
some
kind
of
probably
restful
api
or
something
I
got
a
little
bit
slap
for
saying,
restful
api,
because
that
assumes
a
solution
which
I
haven't,
which
isn't
necessarily
the
right
thing.
But
some
kind
of
API
that
allows
the
user
to
pick
a
name.
L
L
L
If
you
try
to
try
to
do
secure
connections
to
the
to
the
web
api
for
the
for
the
for
the
home,
that
and
assuming
that
there
is
a
way
baby
I
for
the
home,
that
and
I
think
it's
extremely
bad
designed
to
set
things
up
so
that
people
see
those
security
warnings,
I'd
actually
rather
not
encrypt,
and
have
them
see
that
warning,
because
that
warning
trains
them
to
click
through
the
warning
and
it's
probably
a
battle
that
can't
be
won.
But
I,
don't
think
that
we
should
be
on
the
wrong
side
of
it.
L
I
haven't
actually
looked
too
closely
at
Daniel's
drafts
for
a
while,
and
so
I
need
to
rethink
in
this
was
something
that
the
chairs
were
talking
about
earlier.
I
need
to
rethink
whether
or
exactly
how
those
fit
into
what's
written
in
this
document
and
whether
I
need
to
change
this
document
or
suggest
changes,
the
Daniel
next
slide.
L
So
the
home
that
Navy
architecture
assumes
that
naming
is
secure
on
the
home
net
and
that's
a
not
that
hard
of
a
problem.
If
you
have
a
global
name,
it's
a
pretty
hard
problem,
though,
if
you
don't
have
a
global
name,
because
currently
a
validating
resolver
has
no
way
to
establish
trust
for
a
local
zone.
That
is
has
the
same
name
as
that
equivalent
zone
and
everybody
else's
house.
L
Number,
that's
big
enough
that
it's
very
unlikely
that
there
would
be
a
collision
and
then
the
dot
home
that
name
the
definite
domain
for
your
got
home
net
will
have
a
uuid
that
is
different
than
the
uuid
for
a
different
home
net
and
that
zone
will
be
signed
with
a
key.
That
is
that
that
proves
that
that
uuid
is
claimed
by
that
zone
it
doesn't
you
could
have
a
clash,
two
different
home.
That's
could
have
the
same
uid
just
by
random.
L
O
N
O
Of
those
home
nuts
have
a
machine
in
it
with
a
names,
a
banana
okay,
okay,
both
both
your
house
and
my
house
have
a
machine
in
it,
hosting
it
with
the
name
banana
I
have
my
laptop
I'm
currently
sitting
at
your
house
and
I'm
trying
to
resolve
banana
dot
homenet.
What
are
you
saying
you
want
to
happen
here
and
what
is
how
to
security
work?
That.
L
O
L
L
O
L
And
I
mean
I,
think
I
think
the
way
you
secure
the
only
way
that
I
can
think
of
to
secure.
That
is
basically
to
have
an
application
that
needs
to
have
a
trust
relationship
with
that
particular
device
to
cash
that
devices
key,
and
then
it
can
look
at
the
at
the
name,
entry
and
see
if
it's
got
the
same
key
and
if
it
doesn't,
then
it
doesn't
use
it,
but
that's
kind
of
kludgy
so
and
it
requires
applications
to
do
a
lot
of
work
that
I'd
rather
applications
didn't
have
to
know
how
to
do
so.
L
L
But
if
we
want
to
do
that,
we
have
to
have
a
standard
API
or
we're
going
to
have
a
million
different
apps
out.
There
always
do
something
different
and
use.
You
know
cobbled
together
ad
hoc
stuff
that
doesn't
really
work
and
we
get
lock
in
and
on
interop
and
that
sucks
and
then
for
the
central
management
case.
L
You
know
I,
think
I,
explain
the
use
case
there,
I
think
netconf,
yang
or
a
restful
api
is
probably
the
way
to
go.
Probably
not
comfy
hang
because
there
seems
to
be
a
certain
amount
of
energy
behind
that
model
in
the
in
the
industry.
Right
now,
and
obviously,
if
we
do
that,
we
need
a
way
to
provision
it,
and
you
know
corollary
to
all
this-
is
that
really
talking
about
something
that
belongs
in
the
home
net
management
architecture
document
which
does
not
exist?
L
So
that's
something
that
that
that
probably
merits
further
exploration
next
slide.
I.
D
Should
tell
her
question
from
the
chair
on
that
one
I'm,
a
particularly
with
respect
to
our
Charter
and
scope
of
home
net,
which
is
unmanaged
networks
when
you're
talking
about
the
central
management
on
this
slide?
Are
you
eviscerated
something
simply
where
the
centralized
system
is
providing
almost
rendezvous
service,
or
you
actually
talk
about
something
where
third
party
is
performing?
The
management
I
was
from
techNavio's.
L
O
L
So
if
we
don't,
if
we
don't
do
this,
we're
going
to
wind
up
with
all
these
routers
looking
like
nest
boxes
and
nest
boxes
are
great
they're
very
pretty
they
pervert
provide
good
functionality,
but
I,
don't
like
lock-in
and
I,
don't
think
lock
in
is
what
the
IETF
should
be
encouraging
and
in
order
for
us
to
discourage
it,
should
we
choose
to
do
so.
I
think
we
do
need
to
address
this
use
case,
so,
okay,
yeah
so
anyway,
next
slide.
L
L
D
L
You
want,
or
you
know,
if
you
want
to
come
up
with
a
competing
document
cuz,
you
think
I'm,
an
idiot
which
I'm
sure
people
do
that's.
Okay,
too
I
don't
have
any
attachment
to
that.
There's
just
something
that
I
think
needs
to
be
done
and
if
I
have
nobody
else
is
going
to
do
it
I'm
going
to
do
it
so
Lee
Lee,
Howard.
K
Sorry
I
was
a
little
too
loud,
Lee
Howard.
This
is
one
of
those
points,
of
course,
where
I
always
say
hey,
we
need
a
document
and
anybody
out
then
I
always
like
to
point
out
to
anybody
in
user
land.
When
somebody
says
we
need
a
document,
if
you've
always
wanted
to
author
an
internet
draft
that
can
become
an
RFC.
This
is
your
chance.
There's
a
dot,
there's
a
space
here,
waiting
to
put
your
name
on
it,
which
is
not
what
I
came
up
to
say.
K
There
are
two
things:
1i
didn't
stand
up
fast
enough
when
you're
talking
about
the
population
of
the
population
of
the
global
zones-
and
you
said,
and
the
isp
should
delegate
the
reverse
to
the
to
the
home
gateway.
Cam,
ok!
Well,
the
document
says
should
because
the
documents
issued
the
document
says,
should
to
the
ISP.
Yes,
actually,
it's
not
a
capital.
K
What
I
wanted
to
hear?
Yes
very
much
I
am
perfectly
content
to
use.
The
word
should
in
on
non
RFC
2119
contexts,
I'm
happy
with
that.
The
other
one
is
I
haven't
quite
spent
enough
time.
Looking
at
the
gunner,
that
seems
sit
seems
big.
Yes,
he's
like
you're
doing
a
lot
there,
yes,
which
I
think
we're
really
going
to
want
to
socialize
I'm,
not
even
sure,
that's
actually
a
charter.
It
may
be
yeah.
L
Q
Andrew
hi
I'm
Andrew
Sullivan.
So,
first
of
all,
thank
you
for
writing
this,
because
it's
valuable
I
think
that
there
is
a
fairly
serious
gap
in
the
document
unless
I
overlooked
it,
and
that
is,
if
you
use
the
sort
of
globally
ambiguous
name,
lots
of
stuff
doesn't
work
yes
and
and-
and
it
would
be
good
to
sort
of
state
that
innocent
single
sentence
up
front
because
it
actually
removes
a
lot
of
confusion
and,
after
the
third
reading,
I
realized.
Oh,
I
see
this
just
doesn't
work
in
that
case
right
so.
L
Q
A
very
good
point,
but-
and
then
I
guess
just
on
the
other
on
another
issue
on
because
I
don't
want
it.
I
don't
want
to
fully
explore
this
rat
hole,
but
we
started
to
talk
about
it
earlier
and
that
is
this
internationalization
yeah
there's
a
thing
that
came
up
in
the
image.
A
broom
and
I
think
this
is
an
important
difference
right.
The
the
names
in
mdns,
the
the
final
label
is,
is
hidden
in
the
UI
and
the
point
here
with
with
this
globally
ambiguous
name
is
that
it's
not
going
to
be
hidden.
Q
I
think
yeah,
unless
we
think
that
unless
we
think
that
that's
something
we
want
to
impose
on
user
interfaces,
which
I
think
would
be
an
interesting
adventure
for
the
ITF
to
start
yeah.
So
so
that's
a
really
important
difference
and
it
means
that
there
are
internationalization
considerations.
The
IETF
actually
has
a
requirement
for
internationalization
considerations,
which
is
honored
as
far
as
I
can
tell
entirely
by
ignoring
it.
But
but
it's
going
to
be
problematic
here-
and
this
is
a
fundamentally
unsolvable
problem,
so
so.
L
L
Well,
I
mean
actually
so
so
a
little
anecdote
from
from
ancient
days
at
deck.
We
had
this
problem
that
we
couldn't
sell
our
devices
in
certain,
I
guess
in
the
EU
or
whatever
predated
the
EU,
because
I'm
not
sure
that
you
it.
Actually.
I
don't
know
anyway,
so
we
changed
all
of
the
names,
although
all
of
the
words
that
were
used
in
the
bootrom,
which
is
place
where
this
was
an
issue,
all
those
words
were
changed
so
that
they
didn't
have
any
any
vowels
in
them,
so
they
weren't,
really
English
words
and
I.
L
Think
that
might
be
a
little
bit
too
too
little
for
this
problem,
but
it
may
be
that
what
we
want
I
see
the
problem
with
that
home.
That
is,
that
it
is
mnemonic
for
people
who
speak
english
and
it
would
be
nice
if
there
was
something
like
that
for
other
people
as
opposed
to
just
using
a
name
that
is,
you
know,
like
you
know,.
Q
Yeah
well,
so,
just
so
that
you
understand
the
number
of
thorns
that
you
have
just
grasp
yesterday
at
the
H&R
see
right
on
the
human
rights
research
group,
somebody
gave
a
presentation
about
a
programming
language
that
they
tried
to
write
in
arabic,
using
arabic
script,
including
like
never
using
left
to
write
scripts,
and
it
turns
out
not
to
work
very
well.
So
I
think
this
is
going
to
be
a
real
problem.
Thank.
H
H
If
we,
if
we
define
that
home
net
is
a
special
label
that
signifies
these
non-global
unique
names,
then
the
UI
layer
of
the
software
can
show
a
little
picture
of
your
house
or
something
in
italics
or,
however,
we
want
to
represent
that
are
in
a
different
language,
but
on
the
wire
it
has
to
be
the
same
string
because
I
don't
want
to
visit
Canada
and
find
my
iphone
doesn't
work
because
it's
not
speaking
French
on
the
wire
so
right
wire.
It
has
to
be
the
same
language
worldwide.
H
You
know,
and
it
is
a
little
bit
unfortunate,
that
we
tend
to
use
English
mnemonics
in
those
wire
formats,
because
it's
convenient
for
debugging,
but
that
damage
is
mitigated
if
the
user
never
sees
it.
When
I
tap
the
air
print
button
on
my
iPhone
and
it
says,
turn
printer
I'm
not
seeing
any
dot,
local
or
IPP
or
and
many.
L
L
And
actually
one
of
the
things
that's
missing
from
the
home
net
architecture
document
on
that
naming
architecture
document
is
currently
written.
Is
that
in
in
the
hybrid
DNS
document
you
talk
about
being
able
to
have
hierarchical
names,
so
building
one
printer
one-
and
you
know
you
might
want
to
have
something
like
that
for
your
home
net,
where
you
give
your
home
that
a
name
and
then
when
you're
somewhere
else.
Do
you
see
a
different
hierarchical
name?
L
So
when
you
see
predator
one
its
its
Twitter
one
at
this
other
house,
and
you
don't
think
it's
the
same
printer,
the
one
issue
with
the
using
using
something
other
than
what's
on
the
wire
is
that
it
has
to
be
something
that
definitely
would
not
clash
with
names
that
could
be
allocated
nice.
So
your
icon
of
a
house
is
a
good
idea,
but
that
then
doesn't
work
for
blind
users
and
doesn't
work.
J
So
relying
Marcus
Stenberg
from
a
few
moments
ago,
so
he
says
it
actually
depends
the
NSS
d
legacy,
browsers,
typically
hidden
one
flat
list
of
services.
I
think
this
is
repeating
wattage
to
it
so
yeah
and
normal
browse
one
with
domains
includes
domain
names,
also
in
service
names
in
the
user
interface.
J
Ironically
enough
legacy
grow,
seems
the
more
supportive
type
I
think
that
common
with
10
shown
my
my
own
comment
is:
it
was
energy
future
and
it
raised
this
first
about
six
or
seven
meetings
ago,
very
vocally
that
we
should
be
considering
this
somehow
and
I.
Do
remember
from
a
recent
discussion
on
some
list
that
in
the
gtld
allocations,
I
think
home
corp
and
another
domain
were
part
to
be
punted
back
to
the
ITF
or
not
to
be
discussed
here.
I'm
sure,
with
those
there
were
those
three
new
gTLDs
that
were
not
allocated
on
the
basis.
D
J
L
Losing
there's
a
little
bit
of
you
know:
man
I've
been
meaning
to
do
this
for
a
while
and
there's
the
deadline.
I'd
better
get
going
on
in
this
document,
so
so
yeah
there
needs
to
be
a
probably
a
dash.
There's
a
dash
0
1,
coming
with
everybody's
comments
and
the
things
that
I've
learned
since
then,
but
they're.
L
C
D
M
Michael
Abramson
earned
I
just
type
this
into
jebra.
My
throne
say
to
hear
why
don't
we
ask
for
something
you
know
they're,
like
X
M
dash
dash
has
been
reserved
by
for
the
IDN.
So
why
don't
we
ask
him
to
create
some
kind
of
hierarchy
for
this
kind
of
special
username
that
never
goes
into
actual
DNS.
Well.
L
D
30,
without
my
check
out
on
actually
there
may
be
some
mileage
in
that
other
exhale
hyphen
hyphen
is
allocated
czar
prefix,
specifically
for
ID
ends.
The
xx
dash
dash
is
a
nate
is
this
is
actually
a
name
spacing
there
and
you
can't
have
a
normal
name
with
the
dash
dash
in
it
like
that.
So
not
standing,
that's
not
check
the
user
friendly
that
might
mitigate
some
of
those
problems.
Yeah
without
my
hair,
chattel
yeah.
L
We
would
need
to
make
sure
that
the
name
is
not
going
to
be
used
in
some
UI
in
something
that
was
written
by
somebody
who
never
read
in
RFC,
which
is
a
very
common
thing,
given
that
it's
constrained
to
the
home
that
maybe
that's
safe.
But
the
problem
is
when
you
get
into
the
case
where
people
do
sanity,
checking
on
names
and
they're,
like
oh
dash,
that
you
know
X
M
dash
dash,
that's
not
a
valid
domain
name
will
just
reject
that.
L
Q
Andrew
Sullivan's,
so
so
this
is
the
the
very
adventure
that
I
suggested
earlier.
We
might
not
want
to
get
into
right,
because
the
the
reason
xn
minus
minus
ended
up
being
used
for
I
DNA
is
because
it's
internet
domain,
internationalized
domain
names
or
applications,
and
so
every
application
needs
to
become
I
DNA
aware,
and
if
what
you're
saying
is
in
order
to
use
this
mechanism,
you
also
have
to
have
home
Neto
air
applications.
I
think
we're
going
to
be
in
a
world
of
hurt,
yeah.
C
Q
I
think
that
doesn't
meet
the
requirements
that
we
set
for
ourselves
now
I.
It
would
solve
this
problem
in
that
you
could
just
scope
things
in
a
new.
You
know
a
through
z,
a
through
z,
minus
minus
patterned
name,
and
then
you
peel
all
scoped,
but
it
has
this
nasty
property
that
now
we're
starting
to
specify
stuff
at
the
wire
format
in
order
to
in
order
to
enable
certain
you
I
behaviors,
all
right.
Maybe
it's
the
right
thing
to
do,
but
I
I
would
certainly
want.
L
Q
D
Yeah
sorry
Andrew
before
you
walk
too
far
away
and
again
no
chair
hat
on
here,
I
mean
I.
Thinking
was
that
using
the
a-to-z
A
to
Z
dash
dash
mechanism
that
doesn't
it
I
mean
compared
to,
for
example,
that
homeless
or
at
home.
We
would
still
have
that
problem
of
having
a
home
little
where
application.
Potentially.
If
we
have
that
sort
of
special
reservation
but
I
think
possibly
their
scope
with
the
ACC
HTC
dash
dash
of
at
least
being
able
to
have
a
name
which
has
no
possibility
of
a
collision
in
the
normal
namespaces.
Q
So
I,
like
I,
don't
I,
don't
want
to
fully
build
out
the
rat
alcohol
and
like
build
a
basement
rec
room,
but
the
ibn
a
reserved
everything
of
the
pattern:
a
to
Zed,
a
to
z,
minus
minus,
a
chi
almost
autumn,
so
they're
reserved,
and
the
idea
was.
If
we
needed
to
do
like
nasty
hack
like
this
again,
then
we
could
do
it
again,
so
so
home.
That
is
in
the
position,
in
fact,
to
specify
that
the
problem
is
that
there
are
rules
out
there
that
currently
prevent
it.
Q
But
most
of
them,
of
course,
are
on
the
registration
side.
I,
don't
know
very
many
applications
that
do
validation
of
label
strings
at
look
up
time.
So
so
you
might,
you
might
be
able
to
get
away
with
that.
Of
course,
there
is
the
additional
problem
that
if
we
thought
that
home
net
was,
you
know
not
user-friendly
for
the
portion
of
the
population
that
doesn't
use
Latin
characters.
You
know
mq
minus
minus.
Something
is.
D
K
Lee
Howard
well,
at
least.
If
you
make
it
impossible
to
remember
and
use,
then
nobody
will
try
right.
That's
internationalized
actually,
and
we
try
to
do
it
from
the
jabber
room
which
fantastic
conversation,
the
jabber
room.
If
you're
not
there,
you
know
good
stuff
going
on
there
too.
Are
you
missing
out
I'm,
not
excluded?
If
I
know
sorry
their
logs?
You
can
read
it
later
that
I
feel
like
we
need
some
name
nerds
to
get
in
the
room.
K
I
know
we
tried
to
do
a
design
team
and
you
know
maybe
somebody
needs
to
buy
a
round
of
beers
and
and
see
if
we
can
hash
out
all
the
possibilities
I'm
not
qualified
to
to
weigh
in
it.
Just
don't
feel
like
we
need
to
move
this
one
board
somehow
and
I
feel
like
this
is
one
where
we
could
easily
get
stuck
for
three
years,
trying
to
figure
out
what
the
name
is.
K
D
A
D
D
D
L
D
So
say
I
would
be
good
to
get
at
least
another
clot
yeah.
I
know
we
already
have
some
proposed
changes
to
do
it
for
a
dash,
0
1,
quite
shortly
on
the
individual
draft,
but
I
think
it's
clearly
clearly
some
fairly
substantive
additional
contributions
before
we
can
look
at
something
that
we
could
take
for
the
work
and
route
document.
I
wish
we
can
gather
one
group
consensus,
yeah,
yeah
and.
D
D
Okay,
if
there
are
no
further
comments
and
then
unusually,
we've
finished
somewhat
early,
so
I
stretch
your
legs
and
for
the
next
break
and
for
the
next
our
obsession.
I
will.
Thank
you
all
if
we
could
place
how
the
other
blue
sheets
return
back
to
the
front
desk
and
we'll
see
you
folks
in
Berlin,
I
get.