►
From YouTube: IETF100-HOMENET-20171113-1550
Description
HOMENET meeting session at IETF100
2017/11/13 1550
https://datatracker.ietf.org/meeting/100/proceedings/
A
B
Anyway,
so
those
let's
see
Steven,
put
together
the
slides
and
told
me
to
talk
to
them
so
I
hope.
There's
no
surprises
here.
This
is
the
agenda.
Does
anyone
have
a
comment
on
the
agenda
and
so,
let's
see
4-minute
taking
by
the
way,
I
think
Tim
and
Michael
Abramson
is
doing
jabber
and
thank
you
very
much,
gentlemen.
That
is
much
appreciated.
B
C
Here,
for
one
reason,
and
that's
to
at
the
end
of
the
administrivia
thank
ray
for
all
of
his
efforts
over
the
last
many
years
for
HomeNet,
as
co-chair,
I,
really
appreciate
it,
and
thank
you
very
much
for
your
efforts.
If
everyone
doesn't
realize
this
is
going
to
be
raised
last
HomeNet
session.
After
this,
he
will
be
officially
stepping
down.
So
if
I
could
get
everyone
to
say,
thank
you
to
write
in
the
usual
manner.
Please.
B
Thank
You
Terry
appreciate
that
I
would
have
remembered
to
thank
right
yeah.
That
was
good
okay,
so
we
do
have
that
we
have
the
HomeNet
simple
home
net,
naming
which
we
just
adopted
as
a
working
group
draft
in
the
in
Prague,
and
then
there
is
the
home
net
dot
14,
which
hopefully
will
be
published
at
some
point
and
then
there's
all
these
expired
drafts
which,
since
they're
expired,
there's,
presumably
nothing
to
talk
about.
But
if
you
have
something
to
talk
about,
we
can
talk
about
it
at
the
end.
B
B
That's
all
the
slides,
that's
all
the
chair,
slides,
yes,
which
then
brings
us
to
our
first
presenter.
Geordi
I
do
believe
you're
up.
First.
D
Okay,
so
I
I,
somehow
voluntary,
to
to
work
a
little
bit
on
this
I
am
going
I,
think
it's
about
tennis
lights,
I'm
going
to
be,
or
try
to
be
very,
very
quick
with
the
slides,
probably
about
five
minutes.
So
we
have
more
time
for
discussion
which,
which
I
think
it's
good.
What
we
really
need
in
this
topic.
D
Let
me
explain
the
history
about
how
how
this
come
here.
I
I
submitted
a
draft
in
in
the
ITF
in
Chicago
for
updating
RFC
7080
for
my
basic
idea.
That
time
was
to
include
the
new
transition
protocols,
because
the
actual
document
only
supports
six
or
D
and
DS
Lite,
and
while
we
we
are
knitting
in
in
the
market,
we
are
knitting
support
for
ipv6,
only
mekinese
like
map
or
forces
for
ex-lap,
and
so
on.
So
I
I
really
thought
that
we
need
to
include
them
in
in
the
document.
D
D
Now
so
this
is.
This
is
what
what
I
think
will
happen.
It
will
depend
on
the
consensus
on
this
week
and
this
document,
which
is
FC
1784,
that's
B's
transition,
October
7
is
0-1.
I
am
still
calling
it
RFC
1784,
just
just
as
a
placeholder,
because
we
don't
know
if
the
working
group
adopts
its.
What
will
be
the
final
name,
but
basically
this
document
no
longer
contains
a
suit
or
mast
or
anything
like
that.
D
Support
for
a
change
and
CP
and
I
have
a
single
sentence
that
that
say,
would
you
have
in
the
screen
so
basically
saying
if
you
need
support
for
multiple
cascaded
routers,
you
need
to
look
into
each
NCP,
then
I
organized
at
the
panel
in
the
a
pinic
44
in
Taichung
in
September,
so
it
was
two
months
ago.
In
this
panel
we
got
I
invited
globally
to
all
the
sea
vendors
that
want
to
participate.
Unfortunately,
I
got
only
three
participants,
which
was
good
anyway,
because
we
could
not
handle
probably
eight
or
ten,
but
we
got
d-link.
D
D
D
Video
I
think
it's
quite
interesting,
but
I
am
trying
to
summarize-
and
this
is
light,
the
main
things
related,
especially
to
hominid
first
question
is:
there
are
seas
which
ipv6
support
that,
even
if
they
have
dhcpv6
prefix
delegation,
they
are
actually
no
don't
know
or
they
don't
do
very
well.
The
allocation
of
his
last
sista
for
two
different
lands
inside
the
customer
network.
We
know
that
and
it
happens
even
for
the
guess.
D
When
you
have
that
dynamic
routing
right.
Another
question
that
we
got
in
the
in
the
panel
was
maybes
too
early
to
have
support
or
to
implement
HomeNet.
Also
there
were
some
considerations
about
how
much
is
the
cost.
It
seems
that
in
regular
seas
today,
it's
possible
to
do
that.
So
it's
not
a
consideration
of.
We
don't
have
enough
hardware,
we
don't
have
enough
memory
or
anything
like
that.
The
code
is
there.
So
that's
that's
not
a
problem.
We
know
that
in
in
some
open
source
like
open,
wrt
or
really
it
has
been
implemented.
D
There
is
also
a
good
advantage
there
in
the
sense
that
the
price
that
customers
are
paying
for
better
wireless
when
they
had
hired
their
own,
see
your
wireless
access
point
or
whatever
is
growing
up,
so
it
means
also
the
margin
for
the
for
the
vendor.
Sis
is
going
up
and
the
main
issue
that,
according
to
the
people
in
the
panel
we
have
right
now,
is
when
we
have
a
network
span
across
different
hum
net
routers
the
device
discovery
okay,
so
it
it
seems
that
that
may
be
an
issue
that
that
we
need
to
to
resolve.
D
Then
I
try
to
compile
some
of
the
comments.
I
think
I
take
all
of
them.
So
excuse
me
if
I
miss
it
anything
in
the
list,
one
of
the
inputs
and
I
think
that
was
mine.
It
was
we
have
a
missing
opportunity
for
home
net.
If
we
don't
go
to
be
six
hops
and
say
in
the
case
that
document
for
transition
adopted,
because
we
are
not
going
to
modify
RFC
7080
for
we
are
not
actually
asking
the
vendors
to
support
devices,
including
ComNet.
D
Another
question
was:
if
they're
from
net
sees
behind
the
eye,
yet
ISP
see
that's
a
situation
which
is
not
very
fun,
and
then
there
were
somebody
said
there
are
two
possible
models.
One
is
friendly
ISP
model,
the
ESP
has
information
about
the
ComNet,
including
other
external
links,
and
then
we
have
my
home.
My
castle
model,
which
is
external
links,
are
considered
by
HCA
change
CP
as
untrusted,
and
the
C
has
no
home
net
topology
information.
D
D
Okay,
so
I
am
mixing
here
comments
in
the
list
with
my
own
open
questions
to
the
group
here
or
it's
because
is
piece,
don't
understand
that
the
deployment
model
for
ipv6
is
not
the
same
as
ipv4,
because
in
ipv4
you
have
multiple
Nats
and
it
works
and
nothing
happens
and
never
mind
how
the
user
is
connecting
the
root
is
among
them.
It
will
work,
but
in
AP
v6
is
not
going
to
work.
So
that's
something
that
we
need
to
consider.
D
D
D
D
You
will
use
one
of
those
writers
only
so,
for
example,
I
have
I
got
to
G
port
links.
Each
link
will
have
one
Ont,
and
maybe
both
operators
provide
me
a
real
C,
which
ComNet
I
will
use
only
one
of
them
or
the
other
one
I
will
use
inside
my
home,
but
to
connect
both
links.
I
will
use
only
one
good
about
hybrid
routers,
for
example.
D
D
Maybe
subnet
I
believe
not.
But
that's
that's
an
open
question.
Maybe
ComNet
is
only
to
be
in
retail
routers,
not
the
ones
that
provide
the
SP,
but
the
one
that
you
by
in
the
supermarket
I,
really
think
that
we
are
going
to
see
in
the
next
few
years.
More
and
more
people
having
multiple
links
or
even
multiple
eyes,
piece
for
many
reasons.
I
am
convinced
at
about
that,
so
that
that
should
be
part
of
our
discussion
and
I.
Think
that's
the
last
slide.
D
One
comment
in
the
list,
I
I,
think,
is
very
relevant.
Maybe
what
we
need
to
document
as
a
standard
is
something
that
supports
both
models,
because
we
cannot
depend
on
what
the
ISPs
are
going
to
do
and
yeah.
That
was
the
last
slide.
So
I
hope
this.
This
is
useful
for
just
starting
the
discussion
trying
to
collect
all
day
all
the
inputs.
E
So
what
I've
seen
people
talk
about
the
isp
model
versus
the
HomeNet
router
model?
What
I
generally
hear
is
that
they
want
to
actually
just
manage
the
home
net.
So
it's
not
really
a
home
that
anymore.
It's
you
know.
H
NCP
isn't
really
necessary,
it's
all
being
done
with
with
management.
You
know
OEM
stuff.
So
to
me
that
doesn't
seem
I
mean
I,
it's
not
necessarily
wrong
for
us
to
talk
about
it,
but
it's
not
clear
to
me
that
it's
that
it's
really
the
same
thing
as
a
home
net.
E
And
aside
from
that,
the
other
thing
that
my
other
observation
to
this
is
just
that
we
need
to
have.
We
need
to
get
working
on
actually
having
real
home
net
routers
that
we
can
demonstrate
because
there's
no
point
in
talking
about
any
of
this
stuff.
It's
all
theoretical
and
that's
kind
of
it
feels
like
that's
where
we
are
right
now.
D
F
Beauty
link
to
be
very
honest:
LED
we
don't
yet
have
any
homemade
deployment
and
I
know
there
are
a
lot
of
I
I
know:
I
can
I
can
get.
You
know
open
cells
from
open
wrt
for
the
works
for
our
marks
and
gentleman's.
F
However,
the
reason
why
we
didn't
put
the
H
be
put
in
later
us
because
led
some
house
here
today
we
still
have
service
Coast,
talking
about
very
very
fundamental
ipv6,
like
I
mentioned
in
AP,
need
44
like
even
after
so
many
years
since
2011
yeah,
when
we
first
had
less
v6
day
or
v6
day.
Since
then,
we
keep
getting
calls
asking
about
question
one
to
another.
So
I
think
over
reason.
Why
we
didn't
put
in
there
is
because
I
don't
really
see
it
is
about
time.
Yep.
G
H
H
So
in
France,
my
provider
gives
me
a
box
I,
don't
really
match
in
changing
that
box,
but
I
don't
see
why
that
box
couldn't
implement
her
net
I,
don't
see
the
collision
between
being
managed
and
supporting
HTTP
and,
in
my
opinion,
option.
2
is
indeed
killing
whole
net.
In
fact,
because
one
of
the
main
well
I
see
two
use
cases
main
use
cases
for
whole
net
I
have
two
providers.
They
provide
each
provider.
Give
me
one
box:
I
want
to
plug
the
two
boxes
together
and
I.
Want
it
to
work.
H
H
The
second
use
case
is
the
Wi-Fi
extension
same
thing:
I
I,
don't
want
to
buy
to
make
one
Wi-Fi
extension
I
don't
want
to
buy
three
routers
I,
don't
want
to
have
one
si
router
that
is
not
running
a
chain
CP
and
then
the
first
agency
P
and
then
the
second
agency
I,
want
to
buy
one
extension
and
both
of
them
should
work
with
agency
P.
So
really
I.
Don't
think
how
vendors
would
ever
deploy
it
in
CP
if
it
doesn't
become
C
feature
I.
Think.
I
Hi
this
is
Ryan
a
concern.
I
I
think
maybe
there's
a
bit
of
a
business
model
problem
with
the
belief
that
the
that
any
ISP
is
going
to
want
to
facilitate
a
residential
customer
using
more
than
one
upstream
now
in
certain
geographies.
There
are
regulatory
requirements
that
may
impede,
but
I
think
they
want
to.
They
won't
really
would
prefer
to
have
their
own
CPE
and
to
have
everything
behind
that
you
know,
and
part
of
it
is
that
it
reduces
customer
support.
Telephone
calls
and
I'm
told
that
it
in
North
America.
I
At
the
moment
it
runs
about
a
hundred
and
twenty
dollars
to
pick
up
the
phone
and
say
hello,
so
that's
more
than
the
cost
of
the
of
the
ISP
service.
So
they
really
want
to
control
the
the
customer
premise
box.
It's
not
that
you
can't
get
around
that
sometimes
depending
on
the
provider,
but
they
have
no
incentive
to
facilitate
no
business
incentive
to
facilitate
a
residential
customer
using
more
than
one
upstream.
At
the
same
time,.
G
Michael
Abramson
they
here
again
so,
depending
on
what
kind
of
residential
gateway
it
is.
If
it's
a
there
are
different
organizations,
doing
the
standardization
CableLabs
for
about
forum
and
so
on,
and
if
they
are
not
talking
about
this
at
all,
I
I
think
there
is
very
little
chance
of
getting
this
into
volume.
Mainstream.
That
devices
that
ice,
please
by
N
and
you
know,
put
in
people's
homes
and
have
heard
figures
of
you
can
put
the
the
user
can
put
the
modem
the
DOCSIS
modem
in
bridge
mode
and
run
their
own
devices.
G
How
many
do
this
less
than
one
percent?
So
if
we
want
this
as
a
volume
product,
it
needs
to
go
into
the
ISP
managed
device,
because
from
what
I
can
tell
more
and
more
users
the
day
you
just
want
or
worked
in
service
they
just
they
don't
want
to
go
to
a
store.
They
just
want
the
ISP
to
send
me
something
that
works.
I'll,
connect
it
and
everything
should
just
work.
It
should
always
be,
should
just
match
everything.
G
D
G
B
Barber
stark,
AT&T
and
I'm
just
here
actually
individually,
I
really
would
not
recommend
focusing
at
all
on
anything
called
an
isp
model.
I
would
only
recommend
focusing
on
something
called
an
individual
model.
That
is,
you
know
a
person's,
or
what
did
you
call
it?
My
home
is
my
castle,
the
reason
being.
B
B
So
there
is
this
firewall
between
my
home
network
and
my
ISP
router
and
I
do
want
to
make
absolutely
sure
that
use
case
is
fully
supported,
but
I.
You
know,
I
mean
I,
also
fully
support,
making
sure
the
ISP
for
the
people
who
don't
know
how
to
do
it.
I
do
can
support
them,
because
that
is
very
necessary,
and
so
but
anyway,
I
don't
think
I
ET
f
needs
to
specifically
target
the
ISPs.
I
don't
think
that's
your
target
audience.
I
think
really.
B
The
basic
question
is:
how
do
we
get
the
home
NetSuite
of
protocol
solutions
more
widely
known
about
and
understood
and
I
think
the
killer
use
case
is
going
to
be
mesh
networking.
The
problem
with
mesh
networking
is
right.
Now
it
is,
you
know
why
do
we
buy
that
additional
router,
the
second
router?
It's
not
because
we
want
a
second
router.
It's
because
I'm
trying
to
get
Wi-Fi
farther
into
my
house
and
there's
some
really
thing
interesting
things
going
on
at
other
organizations
which
some
of
you
are
probably
members
of
and
may
know
about.
B
But
if
you're
not
we're
not
supposed
to
talk
about
it,
but
it
is
going
on
that
I
think
will
really
get
mesh
networking
to
be
more
accessible
to
the
common
person,
because,
right
now
it's
just
like
I'm,
a
gamer
and
I
need
my
mesh.
You
know
or
something
like
that
and
for
the
rest
of
us
or
the
common
people.
Getting
that
second
router
installed
is
just
so
incredibly
painful,
but
it
really
is
for
coming
people,
it's
painful,
but
I,
think
there's
some
things
that
are
going
on
in
the
industry.
B
That
will
make
it
much
easier
to
get
that
second
router
installed
and
I
think
if
they
were
a
comprehensive
recommendation
of
here,
some
things
that
we
recommend
in
a
multiple
router
topology
in
its
recommendation
from
the
IETF
and
it
can
come
from
home
net.
That
I
think
that
might
be
useful
might
be
can't
predict
the
future.
But
that
would
be
my
recommendation
of
a
way
to
go.
Is
a
home
that
router
recommendation
since.
J
After
10
a
team
time,
dudes
can
Ozzie
Apple,
so
you
definitely
want
the
agree
with
Barbara
I.
Just
first
off
I
really
wanted
to
disagree
with
Michael
here.
Well,
I
definitely
agree
that
a
lot
of
users-
they
just
get
the
box
that
their
eyes
who
speaks
them
they
plug
it
in
and
then
they
forget
about
it
and
that's
fine
and
that
most
you
a
lot
of
users
are
that
way
and
we
want
that
to
still
work.
J
My
understanding
of
HomeNet
is
that
I
don't
have
the
charity
memorized,
but
the
two
problems
were
trying
to
solve
is
one
multihoming
and
two
multi
routers
in
that
use
case.
Hope
Ninh
is
completely
useless.
If
you
just
get
one
box
from
the
CP,
we're
not
providing
any
value.
Add
so
there's
no
point
in
doing
work,
and
that's
for
that
use
case.
That's
already
solved
and
that's
been
solved
for
over
the
less.
Unless.
J
There
is
no
such
thing,
I
know
like
the
moment.
You've
said
the
word
bridge
mode.
You've
lost
like
that's,
not
simple
anymore,
so
in
my
mind,
they're
really
like
the
power
of
home
net,
and
what
we're
trying
to
solve
here
is
that
you
have
your
ISP
box
and
you
go
by
the
second
router
and
you
plug
it
in
and
now
they
play
nicely
together
or
you
buy
these
three
things.
They
just
work,
that's
what
we're
trying
to
solve
here.
K
J
G
So
my
point
to
the
ISP
router
was:
if
I
have
a
DOCSIS
connection
and
a
fiber-to-the-home
connection.
At
the
same
time,
they
would
be
better
if,
if
I
can't
put
them
in
bridge
mode,
they
need
to
participate
in
the
home
net
for
this
to
work
decently.
So
did
this
was
what
I
was
saying
about
getting
into
the
ISP
router,
but
I
think
it
comes
to
our
comment
on
the
on
the
that
email
about
this
on
the
list
and
the
qualifier
element.
L
So
it's
in
tumor,
sinski
and
I
think
the
the
homie
Castle
I
mean
Barbara's
point
is
very
true:
we
we
buy
those
second
round
boxes
to
solve
a
problem,
and
90%
of
us
here
have
probably
solved
that
problem.
Much
like
I
solved
recently
with
my
brother.
It's
the
mom
problem
right
we
buy
stuff
just
to
send
to
our
parents
just
to
fix
their
problems.
Right
I
mean
come
on.
90%
of
us
here
have
probably
done
this
recently
right.
So
when
I
see
the
home
that
thing
I
exactly
see
it
for
that.
L
D
E
Stuart
Cheshire
has
kindly
agreed
to
join
us
as
a
co-author
and
has
been
contributing.
Documents
been
clarified,
based
on
a
bunch
of
comments
that
were
that
came
up
during
during
the
call
for
adoption.
It's
seems
like
it's
got.
Sort
of
the
outline
of
the
document
is,
is
probably
complete,
but
it's
not
detailed
enough.
So
I
don't
think
that
you
could
actually
do
an
implementation
just
using
the
document
as
it
stands,
so
it
still
needs
some
some
work.
E
E
Open
issues
there's
actually
there's
a
fair
amount
of
text
in
the
document
that
I'm
not
convinced,
actually
needs
to
be.
There
might
be
curious
to
know
what
people
think
about
that.
If
you've
read
it
so
document
there's
their
stuff
that
talks
about,
like
you
know
what
the
state
of
the
art
is
prior
to
this,
which
is
interesting
to
us
as
working
group
people
but
I'm
not
sure
it's
going
to
be
interesting
to
a
reader
of
the
document.
So
that's
the
question
there.
E
We
had
a
really
good
conversation
on
the
mailing
list
about
MP
VD
about
three
months
ago.
That
is
not
yet
been
reflected
in
a
new
document.
I'm,
not
convinced
that
that
belongs
in
this
document.
It
may
be
that
we
actually
need
a
special
document
for
npvd
and
there
are
some
issues
with
MP
VD
in
the
sense
that
we
don't
really
have
a
way
to
configure
it
right
now,
but
that's
being
worked
on
in
the
interior
working
group.
E
So
one
of
the
goals
is
to
try
and
give
us
at
least
the
degree
of
security
that
we
have
with
with
multicast
DNS,
which
is
that
you
can't
really
do
multicast
dns
off
the
wire.
So
one
of
the
things
that's
that's
mentioned
in
the
document,
but
not
yet
detailed,
is
how
we
do
how
we
detect
that
a
response
to
a
to
an
M,
DNS
query
actually
came
from
the
wire
that
needs
some
work
C
and
then
you
know,
as
hopefully,
all
of
you
know.
We.
We
punted
on
DNS,
SEC,
validation,
global
names
and
remote
use.
E
We
have
something
that
we
can
work
on
right
now.
I
have
kind
of
a
question
about
it,
which
is
it
seemed
like
there
was
a
ton
of
pushback
on
doing
a
fully
functional
naming
architecture
in
various
working
group
meetings
in
the
past,
and
yet
when
we
decided
to
go
with
the
simple
architecture,
a
lot
of
disappointment
was
expressed
by
other
people
who
didn't
say
anything
when
the
controversy
about
how
complicated
it
was
came
up.
So
so
I
still
don't
actually
feel
like.
E
That
issue
was
resolved
and
I'm
curious
to
know
what
people
in
the
room
think
and
also
you
know
what
people
who
aren't
in
the
room.
Think
and
also
you
know,
I,
don't
actually
know
if
the
people
who
were
against
doing
a
more
fully
featured
naming
architecture
actually
had
any
interest
in
doing
implementations
anyway.
So
I'm,
not
even
so.
It's
really
very
up
in
the
air
as
to
whether
whether
we're
actually
going
down
the
right
path
and.
E
You
know,
obviously
the
document
is
written,
doesn't
satisfy
the
home
that
naming
architecture.
So
that's
a
problem
that
said
the
document
has
some
useful
stuff
in
it.
The
document
isn't
a
waste
of
time,
because
even
if
we
do
add
more
features,
the
the
basic
features
in
the
document
are
still
features
that
we're
gonna
want
in
the
full
system.
So
it'd
be
nice.
E
M
Putting
people
on
the
spot
go
ahead:
Andrew
I,
my
name
is
Andrew
Sullivan
I
have
read
this,
and
so
first
of
all,
I
was
one
of
the
people.
Who's
complaining
about
the
the
lowered
expectation
but
I
I.
Think
I
also
said
on
the
list
as
well
that
if
this
is
the
expectation
that
we
could
actually
realistically
implement,
then
that's
probably
better
than
nothing.
What
I
like
about
the
current
draft
is
that
it
just
says
it
right.
It
just
says:
well,
yeah,
we
said
we're
gonna,
do
this
and
then
we're
not
gonna.
E
M
M
Yet
because
they're
not
polished,
the
I
I
think
that's
a
bad
idea
to
be
honest
with
you,
okay
Rama,
so
to
promise
a
later
document
that
you
haven't
written
yet
seems
like
a
bad
idea,
but
the
the
the
bigger
worry
that
I
have
about
the
draft
as
it
is,
is
you're
quite
right.
There
are
some
some
pretty
significant
gaps
in
the
things
that
it
it
proposes
right
that
there's,
for
instance,
you
meant
DNS
SEC
just
a
moment
ago,
so
I'm
gonna
pick
on
it
right.
It
says.
M
Well,
you
know
that's
beyond
the
scope,
but
here's
the
things
you
must
do
and
there's
the
you
know
the
2119
must
in
there
and
and
that's
that's
a
pretty
big
hand.
Wavey
must
and
I
think
that
the
working
group-
this
isn't
a
criticism
of
the
of
the
document
authors.
This
is
just
a
problem
for
the
working
group.
We're
either
gonna
have
to
decide
that
their
stuff
were
we're
gonna,
punt
and
by
punting.
What
I
mean
is
we're
just
not
going
to
do
it
rather
than
having
a
specification
that
says?
M
Well,
you
know
you
have
to
do
this
stuff,
but
we're
not
going
to
tell
you
anything
about
how
it
works,
because
that's
that's
actually
the
worst
kind
of
specification
right,
a
specification
that
says
you
must
do
it,
but
everybody
does
it
differently,
means
that
it
doesn't
really
interoperate
at
all,
and
the
goal
here
is
interoperation.
So
that's
that's
actually
where
my
biggest
worry
about
about
the
gaps
in
the
document
are
but
I'll
send
I'll
send
this
in
more
detail
to
the
list,
but
that's
a
sort
of
general
point
that
wouldn't
make
so.
E
M
You
know
the
the
problem
in
this
working
group
has
been
that
you
know
in
the
past.
We
we
really
really
wanted,
for
instance,
routing
things
and
that
we
said
well.
We
want
both
and
then
we
spent
like
a
year
and
change
fighting
about
it
and
what
what
seems
to
me
to
be
really
important
actually
is
to
get
things
shipping,
so
I'm
I'm
inclined
to
support.
M
You
know
lowered
expectations
that
that
actually
yield
real
results
that
people
can
live
with,
because
I
think
there
are
some
gnarly
problems
in
look,
but
let
me
step
up
even
one
more
level
of
abstraction.
Sorry
to
do
it
on
what
we're
trying
to
do
is
make
a
protocol
that
really
is
designed
for
geeks
and
a
you
know.
It's
got
a
bunch
of
assumptions
about
about
how
that
thing
is
going
to
be
operated,
the
kind
of
environment
it
was
gonna,
be
operated
in
we're
trying
to
make
that
a
totally
no
management
required
protocol.
M
That
works
just
fine,
no
matter
what
weird
stuff
you
do
in
your
network.
Well,
that's
that's!
A
pretty
seriously
hard
problem
and
I
think
that
the
the
chances
that
we
will
not
succeed
at
doing
it
completely
satisfactorily
for
everyone,
they're,
pretty
high
and
and
and
I'm
and
and
the
more
ambition
we
have
for
this,
the
bigger
the
danger
is
I
guess.
Maybe
I
was
more
hopeful
when
this
all
got
started
than
I
am
now,
but
you
know
I
I'm,
so
so
yes,
I'm
willing
to
work
on
it.
E
Frustrating
thing
for
me
is
that
actually
I
had
pretty
clear
ideas
for
how
to
solve
a
lot
of
these
problems
in
the
original
naming
architecture
document
and
was
told
that
nobody
wanted
to
do
them
by
a
small
minority
of
people.
But
nobody
argued
with
that
really
and
so,
for
example,
I
mean
if
you
look
at
acne
acne
solves
the
same
kind
of
problem
that
we
need
to
solve
to
make
DNS
SEC
work.
It's
not
that
hard.
E
It's
been
done,
I,
don't
I,
don't
think
that
we
I,
don't
think
that's
something
we
can't
do
and
so
I
don't
actually
buy
that
it's
as
hard
a
problem
as
people
think
it
is,
and
my
experience
like
you
know
working
so
so
you
know,
there's
there's
the
there
is.
The
you
know
best
is
the
enemy
of
good
enough
attitude
and
then
there's
the
you
know.
E
M
That
making
it
automatic
is
gonna
be
easier,
but
I
mean
maybe
it
is.
But
but
let
me
let
me
say
something
else,
and
and
and-
and
let
me
just
preface
this
by
saying-
I-
don't
think
that
I
this
is
not
to
try
to
cast
aspersions
on
anyone.
I
I
think
that
the
earlier
effort
that
you
made
suffer
a
little
bit
from
some
of
the
timing
of
the
drafts
right,
so
that
it
was
very
difficult
for
people
to
follow
all
of
the
drafts.
The
larger
draft
I.
M
You
know
there
were
some
sketches
in
it
of
things
that,
like
it
was
pretty
obvious.
You
understood
what
you
meant,
but
I'm,
not
sure
everybody
else
did
and
and
I
think
that
I
think
that
that
came
up
against
just
the
fact
of
a
certain
amount
of
energy
that
had
been
sapped
out
of
the
working
group
because
of
other
fights
that
had
gone
on
and
and
that's
an
unfortunate
fact.
But
it's
the
way
it's
the
way
the
world
is
right.
Sometimes
we
just
have
to
ship
things.
M
I
mean
there's,
there's
a
secure
delegation,
stuff,
there's
some
there's
some
stuff,
for
instance
in
there
about
about
the
registration
for
discovery
processes
and
so
on,
and
it's
this
has
got
to
be
on
the
local
link.
But
of
course
we
know
that
these
networks
are
not
only
local
link
things
and
then
there's
literally
a
parenthesis.
How
well
you
know
we
can.
E
E
The
reason
the
text
about
delegation
is
in
there
is
actually
that
Stuart
mentioned.
You
know:
okay,
great,
we
don't.
Actually
we
haven't
specified
a
mechanism
for
how
to
set
up
a
DNS
delegation,
but
it'll
be
really
stupid,
not
to
at
least
say
that
there
ought
to
be
some
place
in
the
UI
where
you
can
type
in
what
the
domain
name
of
the
net
is
and
have
that
kind
of
work
right.
E
N
I
B
E
It
would
be
nice
if
there's
anybody
in
the
working
group
who's
interested
in
implementing
this
stuff
if
they
would
look
at
the
document
and
see
if
they
believe
that
they
can
implement
it
and
and
I
think
the
answer
will
be
no
and
the
comments
that
come
out
of
that
exercise
will
be
useful.
So
that's
that's
the
the
main
thing
actually
and
then
also
like
you
know
these
questions
that
I
asked
like
in
the
previous
slide.
You
know
how
do
you
feel
about
the
text
in
the
document
like
like
there's
a
ton
of
text
about
that?
E
There's
a
ton
of
detailed
explanatory
text
about
why
things
are
the
way
they
are
and
that's
maybe
that's
good
and
maybe
that's
bad.
Maybe
we
should
just
specify
how
to
do
things
and
not
say
why
to
do
them
in
the
same
text.
So
you
know
we
could
put
that
in
an
appendix,
so
you
so
people
are
interested
in
why
it's
specified
that
way
they
can
find
out.
But
maybe
we
should
just
say
do
this
and
that
would
be
much
less
text
to
read
and
therefore
would
be
much
more
likely
to
actually
get
implemented.
E
B
A
Can
you
come
to
the
next
slide
so
so
am
I
right
is
that
the
plan
you
would
like
to
see
follow
it
is
that
you
or,
and
others
hopefully
implement
this,
and
when
you
get
done
with
that,
then
you
can
finish
the
document
and
then
we
could
publish
it.
That's
the
idea,
okay,
and
is
it
implicit
in
that
that
if
nobody
implements
it
at
some
point
we
say:
well,
we
tried
it
and
it
didn't
get
traction
or
what's
the
what's,
the
fallback
I
mean.
E
L
A
E
B
E
E
Maybe
the
working
group
will
agree
with
me,
maybe
maybe
everybody
will
be
bored
and
we've
actually
done
some
work,
so
we
actually
have
a
Babel
security
document
that
that's
an
individual
submission,
I
announce
it
on
the
mailing
list,
awhile
back
I.
Think
I,
don't
know.
If
people
have
read
it.
Has
anybody
seen
that
document.
J
E
E
E
E
That's
the
one
draft
lemon
HomeNet,
Babel
security
latest
and
basically
well
actually
I
have
some
text
on
it.
So
let's
talk
about
that,
so
the
we
submitted
to
there
was
zero
version
of
the
document
in
time
for
the
ITF
100
deadline.
It
provides
a
mechanism
for
for
distinguishing
between
babel
routers.
So
you
can
just
you
can
trust
some
and
not
others.
E
It
identifies
which
server
sent
a
packet.
It
works
for
multicast
and
unicast,
using
public
key
for
multicast
and
shared
secret
for
unicast,
and
it
relies
on
h
NCP
for
trust
establishment,
which
is
an
open
problem.
So
for
multicast
we
just
sign
the
packet
using
the
elliptical
curve.
Hopefully
there
are
fewer
multicast
than
then
unicast
packets.
But
that's
you
know,
that's
something
we
talked
about
in
previous
sessions,
so
the
the
shared
secret
from
for
unicast
has
established
this
using
DTLS.
E
We
don't
invent
a
new
method
for
doing
that,
but
then
the
shared
secret
is
just
used
to
sign
the
packet.
So
it's
not
quite
it's
not
entirely
DTLS
and
shared
secret
is
the
shared
secrets
are
pairwise,
so
in
other
words,
we
don't
share
a
secret
across
the
entire
network,
because
then
it's
not
a
secret.
We
we
use
diffie-hellman
to
produce
a
secret
that
that
is
shared
by
two
hosts
only
and
that
allows
us
to
identify
who
sent
a
packet.
E
E
J
G
E
Right
now,
the
the
security
of
Babel
on
the
home
net
is
essentially
just
you
know,
cross
your
fingers
and
hope
and
personally
I.
Don't
think
that
that's
a
good
idea
but
like
if
essentially
nobody
else
in
this
room,
thinks
that
having
a
secure
home
that
matters
that
I
guess
I
can't
really
do
anything
about
that.
But
that
would
be
a
very
disappointing
outcome,
I
think
so
and
on
to
more
detail
about
this.
E
So
we
don't
really
have
a
home
that
security
architecture
right
now,
I
think
we
need
to
have
one
and
I
think
it
needs
to
talk
about
a
bunch
of
things
needs
to
talk
about.
This
is,
and
by
the
way,
this
is
all
basically
cribbed
from
a
conversation
that
stephen
and
a
couple
of
other
folks
that
I
had
a
lot
of
about
a
month
ago.
I
think
so.
E
The
the
the
issues
that
we
identified
include:
how
do
we
establish
a
perimeter
meaning
which
hosts
are
in
or
which
routers
are
in
the
home,
then
in
which,
which
routers
or
not
what
kind
of
perimeter
security
we
have?
And
you
know
so,
you
know
with
a
regular
home
network
with
ipv4
and
that
you
have
perimeter
security
in
the
form
of
NAT
which
isn't
very
secure,
but
nevertheless
it's
it's
harder
to
get
through
than
nothing.
And
so
then
another
issue
we
need
to
discuss
is
how
we
establish
trust.
E
So
how
do
we
figure
out
if
the
new
router
that's
trying
to
join
the
HNC
p
network
is
allowed
to
join
it
or
not?
How
do
we
secure
DNS
lookups?
How
do
we
protect
service
publications
so
another
it's?
How
do
we
avoid
having
a
rogue
device
on
the
network
claimed
to
be
providing
a
service
that
it's
seen
be
provided
in
the
past
and
essentially,
like
you
know,
grant
like
so
you?
E
You
know
the
usual
sort
of
danger
case,
for
this
is
you're
printing
out
your
taxes
and
somebody
on
the
network
notices
that
you're
about
to
pray
your
taxes
somehow
and
masquerades
as
your
printer
and
gets
your
print
out
and
somehow
that
turns
out
to
be
bad
for
you.
So
you
know,
generally
speaking,
we
would
like.
E
We
would
like
it
that
when
we
published
a
service
that
some
other
device
can't
come
along
later
and
claim
to
be
that
same
service
or
you
know
either
for
denial,
a
surface
surface,
denial
of
service
purposes
or
to
steal
your
data
and
then
also
you
know,
most
home
nets
or
most
most
home
home
network
routers
have
the
ability
to
talk
to
a
web
UI
and
right
now.
If
you
try
to
talk
to
that
web
UI
you,
basically
you
have
two
options:
either
you
don't
secure
it
at
all.
E
So
now
you
have
all
of
those
packets,
including
passwords,
and
things
like
that,
going
over
the
network
and
the
clear
or
if
you
secure
with
HTTPS,
you
get
a
security
warning
when
you
try
to
connect
to
it,
so
we
don't
really
have
a
solution
for
that.
As
far
as
I
know,
it
would
be
nice
to
have
one.
So
these
are
things
that
I
think
ought
to
be
in
the
document.
E
E
For
perimeter
security
firewall
that
mimic
snap
behavior
is
sort
of
a
starting
point
on
the
current
state
of
the
art.
On
that
we
have
a
document
RFC
1692
that
talks
about
how
that
works.
So
another
question
to
pursue
is
whether
we
need
different
behavior
than
that,
and
also
whether
RFC
1692
gives
us
enough
information
to
implement
perimeter
security.
E
It
might
be
interesting
to
to
specify
that
that
needs
to
be
supported
and
also
mud
provides
a
way
of
saying
what
device,
what
a
devices
sort
of
network
activity
profile
should
look
like,
and
that
can
allow
us
both
to
block
things
that
it's
not
that
are
not
supposed
to
talk
to
it
and
to
enable
things
that
are
supposed
to
talk
to
it.
So
I
think
there's
possibly
a
pony
in
there
establishing
trust.
E
E
Establishment
ritual,
like
you
know,
sort
of
like
it
could
be
Bluetooth
style
pairing
where
you
have
where
you
punch
in
an
ID
number
or
a
pin
to
validate
that
you're
talking
to
the
right
thing
or
it
could
be,
you
know,
could
it
could
be
something
where
you,
where
you
plug
the
two
devices
together
using
a
wired
network?
And
then
you
press
a
button
or
something?
And
that
says:
okay,
you
can
trust
what
you're
talking
to
on
the
other
end
of
this
of
this
wire.
E
Securing
dns
lookups
nice
to
use
DSS
DNS,
but
we
don't
have
a
way
to
do
it.
As
as
discussed
in
the
previous
talk,
we
don't
have
a
way
to
do
a
DNS,
SEC
delegation
that
would
allow
us
to
have
a
trust
anchor.
So
one
option
is
to
come
up
with
a
way
to
have
a
delegation
that
allows
us
to
establish
trust
anchor
and
the
other
is
to
come
up
with
some
kind
of
hack
to
do
trust.
E
On
first
use,
validation
of
trust
anchors
for
the
home
for
home
DARPA
on
the
on
the
theory
that
that
you
know
home
DARPA
can't
have
a
trust
anchor
and
therefore,
therefore,
there's
no
opportunity
for
for
there
to
be
a
conflict
there
and
then
for
protecting
service
publication.
We
actually
have
two
ways
to
publish
services
on
the
home
that
one
of
them
is
using
mdns
and
the
other
is
using
the
dns
SD
over
dns.
E
We
can
use
the
service
registration
mechanism
that
that
Stuart
defined
and
obviously
we
have
to
support
em
DNS
for
the
long
run,
because
M
DNS
is
not
going
away
anytime
soon,
so
we're
always
going
to
have
it
for
the
for
the
foreseeable
future.
We're
gonna
have
both
of
these,
but
obviously
you
get
a
lot
more
security.
If
you
do
Service
registration
than
if
you
don't.
E
E
E
Trust
anchor,
but
it
doesn't
seem
to
be
supported
and
any
browser
that
you're
likely
to
be
able
to
download.
So
again,
we
that
doesn't
really
help
us,
so
that's
kind
of
an
open
issue,
if
we
had
it
seems
like
it
seems
like
a
lot
of
things
in
the
in
the
security
of
the
of
the
home
net,
would
be
better
if
we
could
get
a
global
name
automatically.
So
that's
why
I'm
so
interested
in
that
particular
problem.
E
E
Does
anybody
have
any
interest
in
commenting
on
that
sounds
like
no
so
and
then
there's
a
whole
pile
of
documents
too
right
so
I
mean
I.
Suppose
one
of
the
things
to
say
about
this
is
you
know,
as
you
were
talking
about
in
the
home,
that
architecture
there's
a
ton
of
work
to
do
here
and,
if
nobody's
interested
in
doing
the
work
and
if
nobody's
interested
in
implementing
the
work,
then
there's
no
point
in
talking
about
it.
E
A
L
E
N
N
N
Identifying
web
servers
you're
talking
to
not
by
the
name
or
the
address,
but
by
other
things
like
hash
at
the
public
key,
because
when
you're
configuring,
your
home
gateway,
192,
168,
0
1,
is
the
same
IP
address
in
many
different
places,
but
it's
not
the
same
device
and
when
your
web
browser
autofills
your
passwords
you've
just
sent
the
wrong
password
to
the
device
you
may
not
want
to
disclose
it
to.
Until
recently,
all
of
these
home
devices
were
clear
text
HTTP
in
the
last
couple
of
years.
N
They
seem
to
have
got
the
message
that
security
is
important,
so
they're
now
running
HTTPS,
which
means
you
get
a
certificate
error.
So
now
you
have
a
secure
connection
to
the
imposter
instead
of
a
clear
text
connection
and
but
it's
kind
of
not
their
fault,
because
there
isn't
a
good
way
to
do
that.
Also,
in
the
last
couple
of
years,
the
browser
vendors
have
got
stricter
for
many
many
Safari
windows,
a
certificate
error.
It
would
give
you
a
blue
throbbing
button
saying
go
ahead.
Click
me
now.
N
J
David's
Kazi
sorry
I
had
to
jump
up
when
you
said
Babel
people
don't
want
any
security,
so
I
definitely
can't
speak
on
behalf
of
all
Babel
P,
but
in
my
personal
opinion,
oh
yes,
we
absolutely
want
security,
especially
like,
in
my
opinion.
Homenet
without
security
is
just
a
non-starter
like
this
is
2017
it's
just.
It
will
never
get
deployed
unless
it
has
a
decent
security.
J
What
I
was
saying
on
the
babel
mailing
list
was
that
the
current
babel
RFC
is
not
necessarily
the
best
place
to
mandate
it,
because
the
only
way
we'll
get
any
kind
of
security
is
if
we
have
a
whole
security
architecture,
and
that
includes
the
trust,
establishment
and
maybe
the
trust
establishment
for
the
whole
net
isn't
necessarily
the
same
one.
As
for
all
of
the
papal
use
cases,
my.
B
J
C
B
Barbara
Starck
I
do
think
security
is
really
important
to
that
effect.
We
do
have
a
breakfast
on
Wednesday
morning,
where
people
are
meeting
at
the
IETF
registration
desk
and
going
to
find
someplace
to
go
talk
from
8:00
to
9:00,
and
if
anybody
wants
to
join,
you
know
come
join
us
Wednesday
morning,
8
o'clock
at
the
ietf
registration
desk
and
we're
going
to
try
and
have
a
little
sit-down
and
chat
a
little
more
about
security.
Another
thing
that
I
see
on
this
security
question
is
number
one.
B
There
is
no
certificate
authority
in
the
world,
not
Verisign,
not
any
other,
that
I
trust
to
say
whether
or
not
something
is
trusted
to
be
on
my
home
network,
not
a
one
of
them,
and
so
the
there
is
nothing
that
can
come
from
above.
Whatever
trust
has
is
that
is
going
to
be
in
the
home
network.
It
has
to
come
from
me,
the
home
network
owner.
If
it
doesn't
come
from
me,
then
I
don't
trust
it,
and
so
that's
always
been
my
problem
with
talking
about
you
know,
route
CAS
and
things
like
that.
E
E
M
This
is
Andrew
Sullivan
I,
the
the
the
thing
that
I'm
worried
about
is
that
we
have-
and
this
is
the
thing
that
has
been
the
bug-
bear
all
along
for
this
working
group
right.
We
have
a
bunch
of
people
in
this
room
who
are
experimenting
with
these
things,
who
have
complicated
stuff
that
they
want
to
do
and,
and
we
all
go
to
our
grandmother's
house
is
apparently
thinks
they're
out
of
fix
that
and
but
but
for
for
shipping
devices.
None
of
those
things
are
realistic
right.
M
Anything
that
involves
you
know,
step
one,
the
open
box
step
to
do
something.
This
is
not
any
longer
a
realistic,
think
right,
it's
you're,
just
gonna
plug
it
in
and
it's
got
to
work
and
that
that's
I
think
we've
heard
that
many
tons
at
the
microphone
so
I'm
really
concerned
about
anything
that
says.
Well,
it
can't
you
know,
I
can't
depend
on
this
or
that
certificate
authority
or
whatever
I
think.
Instead,
what
we've
got
to
do
is
is
set
up
the
requirements.
The
other
way
along
the
lines
of
you
know.
M
M
I
think
that's
probably
right,
but
and
maybe
maybe
those
are
the
questions,
maybe
that's
the
the
prerequisite
for
the
for
the
first
document
you
had
in
your
big
list
before
you
get
to
to
architecture
document.
D
G
I
think
that
it's
looking
at
these
boots
are
problems,
it's
basically
impossible
to
achieve
what
Barbara
said
and
then
having
just
plug
it
in
it.
Just
works.
I
mean
there
needs
to
be
some
set
for
your
blipping,
your
a
phone
and
with
NFC
and
provisioning.
Something
on
this
thing
in
order
to
say
I
trust
this
I
mean.
Do
we
think
that
there
there
is
a
magic
way
of
doing
this.
I
haven't
I.
E
Don't
think
there
is
a
single
way
we
talked
about
this
last
time
and
I
think
that
there
are
a
number
of
ways
of
doing
it.
I
can
tell
you
some
of
them
if
you're
interested
but
I
think
it's.
You
know
that
I
was
I
was
getting
the
sense
that
there
were
a
lot
of
glazed
over
eyes
during
the
previous
presentation
and
I
think
that
might
even
get
further
into
the
glaze
over
zone.
It
just.
G
G
J
David's
Kazi
Apple
definitely
agree
with
previous
comments.
I
think
it
might
be
worth
it
for
us
to
take
a
step
back
and
agree
on
at
least
if
not
a
solution,
requirements
for
the
trust
establishment,
because
it's
true,
they
asked
bra
ways.
You
can
have
security
and
you
can
have
the
best
usability
in
the
world.
Then
it's
hard
to
have
both-
and
in
this
case
the
as
simple
as
you
just
plug
it
in
and
you're
done,
there's
no
way
in
hell
the
fact
usually
not
physically
possible
to
make
that
secure.
J
Because
then
you
could
have
something
evil
that
just
plugs
in
and
just
works,
so
it
might
be
worth
it
first
to
decide.
My
personal
opinion
is
that
if
we
have
someone
said
the
word
optional
security,
let's
not
waste
our
time
on
that,
if
it's
optional,
no
one's
gonna
use
it
and
just
not
worth
anyone's
time
to
standardize.
J
That
could
be
a
good
solution,
but
a
lot
of
these
routers
are
really
cheap
and
they
probably
don't
support
and
FCU
other
things.
One
other
possible
solution,
plugging
in
a
wire
plus,
you
know
pressing
a
button.
That's
a
solution,
another
one
you
can
do
that
came
out
of
a
conversation
I
had
yesterday,
you
have
one
of
the
routers.
Just
have
an
LED
blink
with
a
certain
pattern
and
have
a
user
press
a
button
on
the
other
router
at
the
same
rhythm.
J
That
sounds
a
little
ridiculous,
but
if
it's
a
twenty
seconds
that
20-second
process
and
then
you've
exchanged
enough
bits
that
you
have
keys
but
anyway,
I
think
we
can't
make
progress
on
a
security
architecture
unless
we
have
a
trust
model,
because
that's
gonna
be
the
sticking
point.
The
other
parts
aren't
easy
for
from
it,
but
we'll
figure
those
out
right.
So.
J
A
J
E
J
That's
a
great
point,
if
anything
maybe
try
to
get
on
the
agenda
of
like
the
security
open
meeting,
because
judge
hearing
a
oh
yeah
we're
just
not
gonna
secure.
This
is
gonna
freak
them
out,
so
bad.
They
might
come
out
and
help
good.
E
B
Okay,
so
it
sounds
like
there
is
certain
interest:
I'm,
not
sure
that
there
was
enough
interest
to
go
ahead
and
adopt
today.
Would
that
be
accurate.
B
B
B
No,
no
okay!
So
again,
I
want
to
remind
people
want
to
remind
people
who
are
interested
in
further
discussing
the
security
topic
of
home
net
and
also
of
Babel
in
all
of
this
home
networking
stuff
in
general
and
I.
Think
it
also
gets
to
the
securing
of
your
name,
your
DNS
SD
services,
all
that
sort
of
thing.
It's
not
just
Babel
that
there
is
the
meeting
informal
on
Wednesday
morning,
where
we'll
probably
find
a
place
and
I
suggested.