►
From YouTube: IETF103-HOMENET-20181107-1350
Description
HOMENET meeting session at IETF103
2018/11/07 1350
https://datatracker.ietf.org/meeting/103/proceedings/
B
B
B
A
A
C
C
B
Okay,
Steven
stop
playing
with
the
remote
welcome
everybody,
it's
time
to
have
your
your
regular
dose
of
IETF
fun
at
the
home
net
and
here's
the
obligatory
first
slide
with
lots
of
information
which,
if
you
haven't
figured
this
out,
you're
really
lost,
because
you
are
here-
and
you
are
here
at
this
time
on
this
day
and
let's
see
we
have
who's,
doing
chatter,
mica
and
who's
doing
minutes
Stewart.
Thank
you.
D
B
B
E
Is
Ted
lemon
and
I
am
feeling
deeply
alarmed
so
when,
when
we
originally
set
up
the
agenda
the
agenda,
so
we've
got
15
minutes
on
the
agenda
for
simple
naming
I,
don't
know
that
it
necessarily
needs
to
take
15
minutes.
I
have
a
couple
of
other
slides
one
of
them
that
I
think
is
actually
kind
of
important
to
talk
about
is
like
marketing.
I
know
that
that's
a
term
that
most
of
us
here
at
the
IETF
consider
to
be
bad,
but
it's
actually
gonna
be
a
real
problem
for
us.
E
E
B
Great
and
then
Michael
you're
going
to
be
on
tap
to
do
all
righty
then,
and
then
we've
just
got
some
chair
led
discussion
points
here
about
where
we
had
about
securing
the
home
network.
Do
we
want
to
recharter?
You
know
those
standard
kind
of
things
that
we
hopefully
won't
spend
a
whole
lot
of
time
on
all
righty,
no
bashing.
F
B
E
E
So
so,
basically,
this
the
current
update,
I'll
just
give
you
a
little
a
little
brief
overview.
The
current
update
is
kind
of
a
long
attempt
to
a
long
overdue
attempt
to
actually
put
everything
in
the
document
that
needs
to
be
the
document,
with
a
sufficient
degree
of
clarity
that
we
actually
know
what
we're
talking
about
on
each
point
and
I
believe
that
it
currently
does
that
I
think
this
is
the
first
version
of
the
document
where,
where
there
is
no
real
hand-waving
going
on,
that
isn't
to
say,
there
isn't
more
work
to
do.
E
E
Second,
that's
a
better
solution
and
kind
of
as
a
result
of
the
work
that
Stuart
and
I
were
doing
in
in
DNS
SD
on
the
Service
registration
protocol.
I
realized
there's
actually
a
pretty
easy
way
to
set
up
the
delegation
automatically,
and
so
the
current
document
actually
just
proposes
to
do
that.
So
if
you
wanted
to
do
DNS
SEC,
you
need
a
global
name
delegated
the
DES
record
has
the
public
keys
of
all
of
the
HomeNet
rodders,
and
so
that
means
that
any
home
that
router
can
now
sign
the
zone.
E
The
if
there's
a
master,
if
they're
sorry
a
primary,
then
for
securely
delegate
his
own
it
signs
with
its
own
key.
If
there's
no
stateful
primary
primary
than
every
hnr
signs
all
of
the
zones
it
publishes,
which
every
H&R
would
be
publishing
the
the
home
darpa
domain
or
the
whatever
the
domain
is
and
also
any
linked
domains
that
it
supports,
which
which
are
separate
zones.
E
So,
in
order
for
this
to
work,
we
need
a
secure
delegation
for
every
per
every
per
link
internal
zone.
I
didn't
come
up
with
a
way
to
support
DNS
SEC
for
reverse
mappings,
because
the
reverse
mappings
are
the
interesting
reverse.
Mappings
are
all
going
to
be
locally
serves
owns,
and
that
means
that
you
have
the
same.
You
know
trust
establishment
problem
that
I
decided
was
too
hard
to
solve.
E
I,
don't
think
it's
that
important
anybody
wants
to
debate
that
there's
a
microphone
there
or
there's
a
mailing
list
and
I'm
genuinely
curious
to
hear
if
anybody's
thinks.
This
is
bad,
and
so
anyway,
I
think
that
the
current
description
of
how
to
do
DNS,
SEC
on
home
Nets
works,
and
it
would
be
great
if
people
could
look
at
it.
E
So
the
way
the
global
domain
name
works
is
similar
to
the
way
DNS
SD
SRP
works.
Basically
somebody
and
you
know
the
obvious
person
to
do
this
would
be
the
ISP,
but
somebody
has
to
provide
a
name
server.
That's
going
to
accept
a
sort
of
an
SRP
style
update.
If
you
don't
know
about
DNS
SD
SRP,
it
has
a
mode
where,
where
you
can
do
it,
basically,
you
send
an
update.
E
The
update
is
signed
with
its
own
key,
and
so
it's
kind
of
a
first-come,
first-serve
deal
where
once
you've
once
you've
done
the
update
successfully
the
first
time
which,
which
you
do
based
on
policy
rather
than
based
on
authentic
ation,
any
subsequent
updates
to
that
domain
are
secure,
because
you
now
have
a
key
that
you
can
valid
you.
So
you
can
validate
future
updates
based
on
the
key
that
you've
got
in
the
first
update.
E
So
this
allows
us
to
either
have
the
ISP
set
up
a
service
for
doing
this,
which
I
think
is
the
easiest
way
you
configure
the
hnr.
If
a
provider
of
an
hnr
wants
to
do
this,
they
could
set
up
a
domain
under
which
registrations
for
their
home
that
routers
can
be
delegated
or
the
end-user
can
just
get
a
the
domain
and
set
up
a
security.
E
Secure,
Attalla
Gatien
on
their
own
appointments
are
on
that
router,
so
the
delegated
names
have
to
be
generated
automatically,
because
there's
no
user
involvement
here
and
importantly
in
the
current
document
home
network
do
not
answer
queries
for
the
home
net
domain,
whatever
it
is
if
they
come
from
out
of
the
home
net.
So
we're
not.
Actually,
even
though
this
is
a
global
dns
name,
we're
not
actually
publishing
anything
to
the
internet
in
the
future,
we
could
describe
a
way
of
doing
that,
but
we
actually
have
a
nice
basis
for
doing
that.
E
In
this
document
a
nice
we
we've
established
all
the
things
that
we
need
in
order
to
do
that,
and
so
anything
else
can
just
be
an
extension.
We
don't
have
to
do
anything
differently.
This
is
a
solid
basis
for
for
doing
publication
in
the
global
DNS
if
we
want
to,
but
we're
not
doing
it
in
the
current
doc,
so
publication
of
names.
One
way
to
do
that
is
with
DNS
SD
discovery.
Proxy
discovery
proxies
for
those
of
you
who
haven't
been
to
any
of
my
previous
talks
on
this.
E
Our
proxies
that
basically
sit
on
the
home
intruders
and
listen
on
the
local
links
for
EM
DNS
and
send
em
DNS
queries
and
listen
on
port
53,
UDP
port
53
for
DNS
queries
and
translate
those
into
em
DNS
queries.
So
you
can
then,
from
anywhere
on
the
home
net.
You
can
query
for
the
availability
of
services
on
any
link
on
the
home
net.
Using
these
proxies
so
discovery
proxies
are
each
each
discovery.
Proxy
is
authoritative
for
its
own
link
and.
E
That
slide
bullet
doesn't
actually
make
sense,
but
so
so
there's
there's
basically
several
different
layers
here.
One
is
the
one:
is
the
poor
link
stuff,
which
has
always
done
with
discovery
proxies
and
is
stateless,
and
then
there
is
a
pro
home
net
domain,
home,
DARPA
or
if
you've
got
a
delegated
domain,
it's
that
domain,
whatever
it
is
and
that
domain
can
either
be,
can
either
be
state.
Fuller,
stateless.
E
If
you
have
sort
of
really
simple
hum
net
writers
that
don't
have
stateful
DNS
servers
in
them,
then
it's
just
stateless
and-
and
you
agree
with
using
H
NCP
on,
what's
going
to
be
in
the
zone
and
then
every
home
that
router
publishes
its
own
zone,
its
own
copy
of
the
zone.
So
you
could
query
any
home
that
router
and
get
answers
in
that
zone
so
and
then
so
the
stateful
dns
resolution
is
not
mandatory
implement,
but
we
specify
how
to
do
it
and
how
we
make
it
work.
E
So
if
there
are
home
that
way,
if
you
buy
a
home
router
that
supports
stateful
DNS,
then
you
can
also
have
several
home
net
rotors
that
don't
and
it
all
works.
In
principle.
We
haven't
actually
tested
this,
so
let's
see
there's
now
a
detailed
specification
for
how
link
names
are
generated,
which
I
think
will
result
in
Lake
names
being
generated,
that
that
have
some
hope
of
making
sense
to
the
end
user.
Obviously,
there's
no
guarantee
of
that,
but
at
least
these
names
will
not
just
be
gibberish.
They
won't
just
be
like.
E
E
If
it's
stateless,
then
then
every
hnr
answers
for
the
top-level
domain.
So
so
you
got
these
kind
of
two
states
that
the
home
that
can
be
in
either
either
stateful,
TLD
or
stateless
TLD,
and
how
the
how
the
queries
are
answered
depends
on
which
of
those
you're.
In
so
who
manages
the
pearling
delegations
is
determined
using
h,
NCP
and
I'll
talk
about
that
later
and.
E
How
to
do
the
the
resolution
so
that
you,
so
you
can
have
a
stub
resolver
that
doesn't
doesn't
recurse,
is
actually
a
little
bit
of
an
open
question.
Queries
for
the
internet
obviously
can
go
just
to
the
ISPs
recursive
resolver,
but
queries
for
the
home
that
can't
so
either
we
have
to
have
a
recursive
resolver
or
we
have
to
have
something
that
approximates
in
a
cursory
resolver
like
a
discovery
broker,
which
is
a
document
Stuart's
working
on
in
the
DNS
SD
working
group.
So
that
is
actually
an
open
question.
E
I
only
noticed
this
issue
when
I
was
writing
these
slides,
but
I
think
it's
pretty
easy
to
resolve.
Having
a
full-service
resolver
on
the
home
net.
Maybe
is
not
great,
but
there
are
a
number
of
them
that
work
quite
well.
I
mean
there's,
not
there's
I,
think
I
think
DNS
mask.
So
it's
not
out
of
the
question
that
we
could
just
have
a
full
service
resolver
on
the
home
net,
and
that
might
be
the
easiest
way
to
solve
this
problem
because
it
doesn't
involve
inventing
anything
new.
E
So
the
way
that
that
works
is
that
when
host
sends
a
DNS,
lookup
query
to
port
53
on
whichever
router
they're
configured
to
send
it.
To
that
router
looks
at
the
domain
to
see
how
to
resolve
it.
If
it's
local,
then
it
has
to
resolve
it
locally.
If
it's
non-local,
then
it
just
forwards
it
to
the
ISPs
resolver.
E
It
doesn't
try
to
do
anything
fancy,
and
so
we
need
to
have,
as
I
was
mentioning
the
previous
slide,
some
kind
of
way
of
doing
recursion
on
the
local
domain
so
and
then
for
multiple
provisioning
domains
for
multiple
provisioning
domains.
I
think
that
we
can
use
the
the
option,
the
PVD
our
adoption,
that
they're
working
on
in
int
area,
which
is
next
and
then
I-
think
we
need
an
additional
IDI
and
s0
option
to
specify
which
PVD
were
doing.
E
The
query
in
I
looked
at
a
bunch
of
different
ways
to
solve
the
which
PVD
to
use
problem
and
I
think
this
is
the
cleanest.
It's
it's
really
simple.
We
have
to
write
a
spec
for
it.
There
isn't
a
spec
for
it
yet,
but
I
think
it's
dead
easy.
So
that's
what
I
propose
to
do
and
I
think
it
was
Tokyo
was
saying
that
that
he
is
not
willing
to
require
that
hosts.
Do
this,
which
of
course
is
correct,
because
it's
not
in
the
Charter
that
we
can
require
a
host
to
do
things.
E
So
if
you've
got
a
host
that
doesn't
support
the
PVD
option,
then
that
will
be
signified
by
a
query:
come
to
the
resolver
that
doesn't
have
that
option
in
it,
and
so
if
the
resolver
gets
a
query
for
something,
that's
not
on
the
home
net,
that
is
for
a
moat
for
a
host
that
doesn't
support
pvd's,
then
it'll
just
try.
You
know
if
you've
got
if
you're
multi-home
they'll
try
ISP
number
one
and
then
the
next
time
it
tries
to
send
a
query
out.
E
So
if
that
query
fails,
it
will
try
ISP
to
and
then
it
will,
try
is
p3
and
then
it'll
try.
The
next
resolver
from
is
p1
and
the
next
resolver
from
is
p2,
and
the
next
resolver
from
is
p3
so
that
we
get
reasonably
light.
A
reasonable
likelihood
that
we'll
succeed
in
getting
an
answer
to
the
query
and
the
host
will
actually
be
able
to
connect
and
that,
if
the
hose
connects
it
will
connect
to
the
right
thing.
So.
E
E
Review
acts.
Ok,
so
it's
pretty
it's
pretty
simple:
it
used
to
be
used
to
be
like
standards,
action,
and
that
was
not
so
simple,
but
it's
not
that
anymore
so,
and
whether
to
do
that
in
this
document
or
in
some
other
document
I,
don't
know
it
seems
like
it's
kind
of
a
general
purpose
thing,
so
it
probably
doesn't
belong
in
this
document.
E
So
remaining
issues
as
I
said:
I
think
it's
mostly
complete
need
to
write
a
section
that
just
has
like
the
list
of
things.
You
have
to
do
to
implement
a
home
net
router,
because
right
now
the
architecture
document
just
says
how
the
architecture
works.
It
doesn't
doesn't
have
any
normative
language.
I
didn't
wanted
to
have
normative
language.
It
just
describes
how
the
architecture
works.
There's
a
lot
of
new
stuff
in
here
that
nobody
else
is
doing
right
now.
E
E
The
reality
is
that
I
spent
most
of
my
time
this
year,
working
on
DNS,
SD
and
issues
relating
to
DNS
SD
and
have
not
had
time
to
spend
to
put
much
effort
into
HomeNet
other
than
you
know,
working
on
the
document.
It
would
be
really
nice
if
there
were
somebody
else
who
were
interested
in
working
on
this,
and
so,
if
you're
sitting
in
the
room
and
you're
thinking
to
yourself
gosh,
this
would
be
nice
to
have.
E
B
B
G
When,
if
you
look
at
the
successful
consumer
products,
they're
ones
that
tend
to
remove
the
complex,
knobs
and
features
and
make
it
easy
to
use
the
technology
seamlessly,
there's
a
reason
why
a
lot
of
people
do
cloud,
centered
players
and
recording
for
their
devices?
It's
because
they
don't
want
to
be
responsible
for
maintaining
infrastructure
in
their
home.
G
To
do
these
things
they're
loud,
you
know
I
personally,
have
gotten
rid
of
most
of
the
data
center
type
things
from
my
house,
because
they're
loud
and
they're
noisy,
and
when
you
talk
about
doing
some
of
these
complex
things
here,
like
sending
dynamic,
dns
updates
for
the
devices
that
you
know
for
the
CPE
for
my
home,
when
I
don't
have
any
broadband
choices
and
I
sit
behind
double
NAT
behind
a
wireless
isp.
These
things
aren't
really
applicable
to
me.
G
You
know
for
my
home
network
because
I
don't
have
a
lot
of
market
choice
for
these
things.
Sorry,
so
this
stuff
is
interesting,
but
it's
also
designing
the
expert
solution
for
the
expert
people
and
I'm
not
sure
about
the
consumer
applicability
of
many
of
the
things
that
are
going
on
just
just
to
prepare
myself
for
all
the
tomatoes.
I'll
sit
close
now
interesting.
H
Michael
Richardson,
so
what
I
just
heard
I,
don't
know
your
name,
sorry,
Jared
Jared,
so
so
hi
I
forgot
my
badge
in
my
room
yesterday
too,
and
so
what
I
heard
is
that
the
document,
where
defying
something
super
complex
for
super
expert
people,
and
so
what
I
heard
actually
is
that
our
documents
are
really
bad,
then,
because
we,
our
goal
was
to
specify
automated
systems
for
people
that
don't
know
anything
what
they're
doing
specifically
his
spouse
and
children.
Okay,
that
would
operate
the
network
when
he's
not
there.
H
G
Yes,
because
there
are
some
telecoms
that
have
fiber
within
twelve
hundred
feet
of
my
house,
but
because
there's
no
commercial
viability
or
there's
no
way
to
talk
to
them
about
it.
I
actually
have
to
build
my
own
fibre
to
my
own
house
to
span
two
miles
to
go
and
get
get
places,
and
so
some
of
these
things
are
really
complex
and
when
I
talk
to
the
Wireless
ISP,
whom
I
sometimes
provide
consulting
resources
for
any
of
this
complexity
in
the
home,
he's
just
like
what.
Why
do
people
need
all
this
bandwidth
and
these
resources?
A
G
A
So
I
actually
heard
just
to
respond
to
what
Michael
said.
I
heard
I
heard
two
different
senses:
criticisms
one
was
that
we
might
be
designing
something
for
that
requires
an
expert
in
the
home
and
I
press
think
wait
well,
I
thought
sure.
That's
true,
but
the
other
thing
I
heard
that
I
think
may
be
issues
that
there's
too
many
moving
parts.
Even
if
they're,
all
automated
and
I
think
that
right
we
do
have
a
lot
of
moving
parts.
Yeah.
E
G
E
You
yeah,
no,
the
reason
I
was
asking
that
question
is
because
I
felt
like
your
answer.
A
little
bit
was
was
not
talking
about
so
our
job
as
ITF
people
is
to
is
to
is
to
build
things
that
produce
the
result
that
you
want
right
and,
and
and
that's
what
I
want
to
produce
is
something
that
you
would
feel
comfortable
putting
in
your
home
I'm,
not
convinced
that
we
can
do
that
for
just
the
reasons
you've
stated,
but
that's
that's
our
goal.
So
if
we
can't
do
that,
that's
actually
useful
knowledge.
E
G
I
think
those
are
the
things
that
make
some
of
the
other
products
that
people
try
to
use
in
their
home,
not
work
well
because
they
don't
share
the
same
broadcast
domain
because
of
the
assumptions
made
about
when
the
when
these
other
people
are
designing
some
simple
consumer
device
to
do
things
and
so
the
more
complex
we
make
it,
the
more
likely
we
are
to
break
the
consumer
products
that
people
are
expecting
to
use.
Whether
or
not
those
products
were
designed
correctly.
I
think
is
a
different
discussion.
Yeah.
J
K
Abram
said
yeah:
it
struck
me.
The
this
question
was
good,
so
when
I
try
to
get
budget
to
do
this
kind
of
stuff,
my
management
had
hard
time
motivating
doing
this.
Instead
of
like
the
customer
complaints
about
bad
Wi-Fi,
and
how
do
we
help
the
customers
to
fix
their
Wi-Fi
and
home
network
and
so
on?
And
how
do
we
help
them?
Do
that
yeah?
We
do
that
by
letting
our
customer
service
get
more
insight
into
what's
going
on
in
the
home.
K
So
that's
one
of
the
reasons
I
have
been
you
know
not
so
active
in
here
is
that
have
been.
You
know,
limited
funds
whatever.
What
do
we
prioritize,
and
so
it's
been
more
about
managing
that
the
devices
the
device
in
the
home
then
I
do
think
that
the
the
basic
premise
here
as
a
as
a
engineer,
I
I,
think
this
is
cool
stuff.
It's
just
that
for
the
ninety-eight
percent
of
the
population.
K
The
use
case
is
kind
of
hard,
and
also
some
of
the
like
HomeNet
and
the
routing,
and
so
on
relies
on
the
kind
of
devices
that
we
actually
don't
have
in
the
field
that
doesn't
have
a
bunch
of
routing
interface
it
it
has
a
one
and
a
lan
interface
and
the
LAN
interfaces
are
switched.
It's
this
seems
to
be
at
least
historically
there's
been
a
bit
of
a
mismatch,
but
what's
actually
available
what
I
can
buy
and
that
I
can
build
on
and
what
we
have
been
talking
about
here.
C
B
A
E
E
E
So
I
believe
that
our
market,
our
the
following
people-
managers
at
ISPs
managers
at
router
vendors,
early
adopters,
who
are
willing
to
run
open
wrt,
although
even
they
can't
run
this
and
developers
who
are
doing
it
because
we're
developing
those
are
those
are
the
four
people
who
potentially
might
act
on
what
we
were
doing
here.
So
our
pictures
of
friendly
people
isn't
working
as
we
have
a
product.
E
So
if
you,
if
somebody
tells
you
oh
just,
go
download
open
wrt,
while
I
tried
that
doesn't
work,
it
also
doesn't
work
if
your
network
is
in
any
way
weird
like
it
sort
of
works.
If
you
only
have
open,
wrt
and
nothing
else,
but
even
then
it
doesn't
work
very
well.
So
we
really
don't
have
a
product
to
offer
even
to
early
adopters
right
now,
so,
basically,
and
speaking
as
a
developer
running
this
stuff
on
my
network
is
very
difficult
and
in
fact
I'm
not
doing.
I
E
Right
now,
so
that's
kind
of
a
problem
right,
if
I'm
not
doing
it,
how
can
I
expect
anybody
else
to
do
so?
I'd
be
curious.
If
there's
anybody
in
this
room
who
is
running
this
on
their
network
and
I,
actually
pulled
some
people
who
are
doing
development
like
the
Serie,
wrt
people
and
they're,
not
using
this,
so
who's
zero
wrt.
This
is
like
they've
Tate
and
you.
E
H
n
CP
they're,
not
using
you,
know
any
of
the
stuff
that
we're
working
on
here.
They're,
the
you
know,
dave
tate
is
not
using
Babylon
his
network,
so
that's
kind
of
bad
news,
and
then
you
know
when
I
think
about
going
to
management
at
a
large
company
and
saying
to
a
manager.
I
think
that
you
should
spend
money
on
this.
What
do
I
tell
them?
What
do
I?
What
is
my
sales
pitch
right?
E
Because
if
I,
if
I,
don't
know
what
to
tell
them,
then
then
it's
hopeless
right,
I'm,
not
gonna,
go
into
a
meeting
with
somebody
like
that,
and
and
and
not
have
an
answer
to
this
question.
So
so,
let's
look
at
what
we've
got
out
here
right
now.
Our
competition
we've
got
home
net.
We
wish,
which
is
a
routed
mesh
with
lots
of
services.
We've
got
layer
two
mesh
and
there
are
several
vendors
out
there
right
now
that
are
selling
layer
to
mesh
products.
E
We've
got
wired
to
access
point
infrastructure,
all
right,
so
that's
like
ubiquity,
and
so
all
of
the
all
of
the
options
that
we
have
right
now
that
sort
of
solve
the
home
net
problem
solve
it
using
some
form
of
layer
to
not
layer.
Three,
none
of
them
have
fancy
service
discovery.
Stuff
service
discovery
doesn't
necessarily
work
on
them
because
of
multicast
issues,
but
but
that's
what
people
are
selling
if
people
are
buying
these
like
people
are
going
out.
E
So
naturally
at
ITF
we
prefer
a
routed
mesh
of
some
sort.
But
how
do
we
talk
people
into
it?
So,
let's
talk
about
like
what
you
get
if
you
get
a
routed
mesh
versus,
if
you
have
some
kind
of
layer,
2
network,
if
you
have
a
router
bit
mesh,
you
get
whenever
the
host
switches
access
points,
it
has
to
renumber
any
connections
it
has
open,
have
to
be
restarted.
If
I
have
a
wife,
I
call
going
on.
It's
probably
going
to
glitch
out
or
even
drop
and
latency
is
quite
large
for
this.
E
A
E
So
if
you
have
a
home
that
this
is
the
this
is
the
experience
you'll
get
like.
So,
for
example,
I
have
I,
actually
have
a
wired
infrastructure
out
to
my
garage.
So
if
I
walk
out
to
my
garage
from
inside
my
house,
I'm
gonna
have
to
switch
access
points
and
when
I
switch
access
points,
if
my,
if
the
application
layer
isn't
taking
care
of
that
switch,
I'm
gonna
lose
my
connection.
Fortunately,
application
layers
tend
to
yeah
yeah.
G
And
if
the
my
client
isn't
Romijn
properly,
I
will
get
poor
behavior
for
a
Wi-Fi
call.
So
I
I
don't
know
if
anybody
here
still
talks
on
the
phone
or
if
I'm,
old,
school
and
I'm
one
of
those
weird
people
who
hasn't
converted
entirely
to
texting.
But
when
I
walk
when
I
talk
on
the
phone,
I
tend
to
walk
around
as
well,
and
that's
part
of
my
getup
for
my
desk
and
stuff
and
do
exercise.
And
so
some
of
these
properties
of
HomeNet
are
going
to
make
it
unattractive
to
me
right.
F
G
E
That's
a
good
question:
I
mean
you
know
my
experience
that
this
is
an
area
director
was
that
I
was
an
area
director
for
a
bunch
of
int
area
working
groups,
including
trill,
and
it
was
amazing
how
much
I
got
about
trill,
like
just
amazing
how
many
people
were
like.
Why
haven't
you
closed
them
down
yet
they're
doing
layer?
Two!
That's
crazy!
We
don't
do
layer
two
at
the
IETF
layer.
Two
is
bad,
so
I'm
not
I'm
not
going
to
justify
that
position.
E
Your
your
phone,
if
it's
properly
implemented,
will
automatically
switch
from
one
to
another,
and
your
call
will
will
continue
without
any
issues
at
all,
it's
possible
that
there
could
be
some
some
bad
things
that
might
happen,
but
relatively
speaking,
they're
going
to
be
less
bad
than
what
you
would
experience.
If
you
were
on
a
home
net.
K
E
A
H
H
Ok,
if
you
add
VLANs,
then
you're
screwed,
ok,
so
there's
a
whole
bunch
of
things
and
actually,
just
to
your
point,
that's
actually
one
of
the
reasons
why
we
did
trill
is
because
all
of
this
was
broken
and
we
needed
actual
real
routing
in
data
centers
to
make
that
happen
right,
but
to
the
point
of
it
walking
around
and
home
net.
So
what
we're?
Really?
What
you're
really
talking
about
is
the
fact
that
we
made
a
decision
that
the
Wi-Fi
would
not
be
bridged
for
reasons
of
congestion
with
the
link.
H
Yes
and
there's
lots
of
companies
out
there
selling
you
systems
that
are
mostly
vertically
proprietary
that
do
things
to
make
that
all
better,
ok
and
we
haven't
specified
that,
but
we
haven't,
we
have
what
the
problem
is,
that
we
haven't
told
people
you
have
to
do
that
in
your
home
net
right.
It's
just
that's
the
kind
of
default
that
open,
wrt,
sh
NCP
comes
up
with
right,
so
I
think
it's
important
when
we
distinguish
what
our
specs
say
and
what,
but
the
reference
implementations
does
be.
E
H
L
M
F
J
A
K
K
I
specifically
got
to
you
bhiku
to
a
piece
that
are
on
the
same
layer
too.
So
I
can
walk
from
one
side
of
my
domicile
to
the
other.
All
doing
avoid
fall
right
into
running
this
and
home
that,
with
the
current
way,
that
agency
piece
sets
up
saying
it
would
be
a
complete
nono
would
be
worse
than
whatever
you.
K
Or
something
100%,
yes
or
basically,
you'd,
rather
have
it
be
bridged.
Somehow
I
would
rather
have
one
AP
with
really
bad
2.4
gigahertz,
covering
the
entire
domicile.
Then
these
2
5
gigahertz
that
give
me
perfect
Wi-Fi
everywhere
in
my
domicile.
This
is
this
handover
is
like
a
complete.
No,
no
none
of
the
applications
or
layer
3
handle
any
kind
of
you
know,
handover
of
anything
between
there,
it's
oval.
We
could
talk
about
this
at
length.
Do
we
actually
have
something
new
to
say.
H
Richardson
so
I
have
something
new
to
say.
One
of
the
advantages
about
finding
a
solution
that
works
at
layer
3
is
that
it
makes
it
much
more
doable
in
the
future
to
be
able
to
separate
different
categories
of
devices,
specifically
refrigerators
and
light
bulbs
and
whatnot
to
be
able
to
separate
them
from
other
traffic.
And
when
you
do,
you
realize
that
you're
gonna
have
to
do
that.
Then
it
suddenly
becomes
very
useful
that
you
can
solve
a
layer,
3
isolation
or
mobility
problem
at
layer,
3
right
right.
E
Yeah
I
think
the
one
of
the
best
stories
for
home
that
is
actually
IOT
anyway.
So
I'm
gonna
just
talk
about
some
more
comparison,
so
service
discovery.
We
need
a
complicated
name
resolution
infrastructure
which
we
are
defining
and
I
think
will
work
really
well,
probably
works
better
than
them
doing,
service
discovery
or
multicast
for
a
variety
of
reasons
having
to
do
with
the
way
multicast
is
handled
on
Wi-Fi
networks.
E
E
So
on
a
layer,
2
network,
you
just
use
em
DNS,
which
means
using
multicast,
that's
not
great,
but
you
can
probably
make
it
work
and
by
the
way
there
are
ways
to
mitigate
the
multicast
traffic
issue
like,
for
example,
just
unicasting
to
every
station
I.
Don't
know
why
that's
not
more
common,
but
it's
not
ours.
I
know.
L
L
Say
just
use
DNS,
it
says
just
use
em,
DNS
I
know
this
light,
says
I
said
it
incorrectly
out
loud.
Sorry,
the
point
is
that
in
home
networks,
people
run
arbitrary
applications.
Arbitrary
applications
often
use
service
discovery
protocols
that
they
implement
themselves
like
built
into
the
application
that
might
be
20,
30
years
old
or
whatever,
because
the
application
was
written
ten
years
ago
and
they're
still
running
that
thing,
so
people
that
do
homeland
play
games
or
whatever
else
and
so
yeah.
L
L
E
L
Yeah
and
so
of
layer,
2
networks,
the
bottom
slide
as
long
as
everything
is
on
the
same
subnet
and
it
is
broadcast
capable
of
multicast
capable
for
the
link.
Then
all
the
old-style
stuff
works.
Okay
on
the
top
half
once
they're,
not
on
the
same
link,
then
all
the
old-style
stuff
starts
breaking,
and
so
some
apps
work,
the
ones
that
might
be
using.
L
You
know
the
complicated
stuff
for
the
mdns,
the
DNS
proxies
or
whatever,
and
all
the
ones
using
legacy
protocols
stopped
working,
but
kids
and
people
that
are
not
us
actually
notice
that
things
break,
and
so
that
causes
them
problems.
And
so
this
should
you
say,
just
use
in
DNS
or
any
other
service
discovery
protocol
that
exists,
that
uses
link,
scope,
traffic,
sure
enough.
E
N
E
N
N
And
the
points
that
you
make
there
are
applicable
to
anything,
that's
using
lots
of
multicast
and
broadcast.
It
can
be
expensive
on
Wi-Fi
yeah.
Some
Wi-Fi
access
points
block
it
completely.
This
is
common
at
universities
and
companies.
They
just
block
multicast,
so
discovery
doesn't
work,
so
we're
kind
of
stuck
between
a
rock
and
a
hard
place
between
two
things
that
don't
work
very
well
right.
M
Hi
Steve
McCowan
Blaine
brewery
whoops
bit
louder,
I'm
completely
new
to
this
group,
but
I
thought
you
might
like
to
know
that
$82
levin
has
just
finished
an
amendment
11
a
queue
called
sirs
discovery
and
there
are
some
aspects
of
mdns,
although
it's
not
explicitly
called
out
within
that
amendment
operating
over
the
Wi-Fi
owned
to
face.
So
some
of
you
guys
might
want
to
look
at
it.
There
may
be
some
interesting
things
in
there
are
they
doing.
L
Dave,
since
this
is
the
marketing
problem,
part
of
the
other
audience
would
be
the
Gateway
vendors
or
the
the
vendors
of
the
boxes
that
span.
You
know
things
within
the
home
now
right
right,
and
so,
if
you
think
about
their
marketing
problem
between
choosing
between
which
one
of
these,
even
if
they
support
both
right,
which
is
the
default
or
whatever
right
they
look
at
this
and
say
well,
if
the
customer
puts
something
in
and
is
app,
stop
breaking
what
happens,
they
return
it
to
the
store
and
so
they're,
driven
by
things
like.
H
E
Okay,
so
let's
go
onto
routing
on
home
net
we
have
a
routing
fabric
which
maybe
works,
as
I
say
said
earlier,
I've
heard
some
discouraging
reports
from
de
tape
and
one
thing
that's
really
interesting
about
having
a
routing
fabric
that
you
don't
get.
If
you
don't
have
one
is
that
you
can
join
it
with
things
that
are
not
Wi-Fi
routers
and
get
routing.
Instead
of
bridging
that's
kind
of
nice,
it
keeps
traffic
isolated.
E
So
the
your
IOT
network
is
isolated
from
your
from
your
your
Wi-Fi
network
and
doesn't
see
all
of
the
multicast
traffic
on
your
Wi-Fi
network
and
doesn't
see
you
know
any
other
traffic
that
might
leak
so
layer.
Two
mesh.
We
have
a
proprietary
or
I
Triple
E
layer
to
mesh
protocol.
I.
Think
the
I
Triple
E
layer
to
mesh
protocol
doesn't
actually
work,
but
maybe
that's
changed
over
time.
I,
don't
know
the
last
time.
I
oh
yea,
Leben
s.
E
E
It
doesn't
entirely
work
for
neighbor
discovery
either
because
neighbor
discovery,
packets
have
to
be
seen
by
the
whole
network
in
some
cases
so
doing
isolation
on
an
on
a
layer,
2
infrastructure
is
not
easy.
On
the
other
hand,
you
know-
maybe
that's
not
important,
but
these
are
the
things
that
I'm
just
going
through
the
sort
of
the
bullet
points
that
you
would
go
through
in
a
marketing
presentation,
isolation.
We
can
have
separate
subnets
that
our
firewall
from
each
other,
we
can
have
a
DMZ.
We
can
do
service
discovery
across
the
DMZ
on
a
home
net.
E
You
can't
do
that
on
layer,
2
we
can
on
layer
2.
We
could
use
VLANs
for
isolation,
but
then
we
need
routing
between
the
VLANs.
We
need
to
be
able
to
do
service
discovery
across
the
VLANs,
so
we
wind
up
actually
replicating
a
bunch
of
the
technologies.
We're
inventing
here,
standardization
home
that
can
in
principle
be
standardized
we're
not
quite
there
yet
Hey
and
then
l2
hub-and-spoke
is
pretty
straightforward.
We
already
know
how
to
do
that.
E
So
that's
actually
like
a
feature
if
you're
a
router
vendor,
because
it
means
that
once
you've
sold
somebody
one
router,
they
have
to
buy
all
of
their
routers
from
you
and
if
you're
a
host
vendor,
you
don't
care
because
you
know
it
just
works,
hopefully
doesn't
work.
Then
the
customer
probably
returns
the
router
staple
name
service.
E
But
so,
if
you
have
a
flat
layer
too,
and
you
want
to
do
stateful
name
service
or
wanted
a
DNS
SEC,
then
you're
gonna
be
adding
a
lot
of
stuff
to
your
router
that
you
don't
currently
need,
but
they
could
also
just
say,
install
an
appliance
or
do
it
in
the
cloud.
So
not
sure.
That's,
actually
an
important
distinction
code
complexity,
I
think
that
that
kind
of
losses.
Here,
it's
not
necessarily
the
case
that
the
layer
to
mesh
implementation
is
less
complex
than
Babel,
but
they
are.
E
You
know
the
vendors
who
are
selling
this
already
have
it
so
that
doesn't
really
buy
us
a
whole
lot,
and
you
know
we
do
have
a
whole
bunch
of
moving
parts
here
that
that
need
to
work
right
and
you
know,
discovery
proxy
full
service,
resolvers
Babel.
These
are
all
significant
piles
of
code,
whereas
like
on
a
layer,
2
gateway.
All
you
really
need
is
a
dumb
DNS
proxy.
Maybe
you
need
a
DHCP
server.
It's
pretty
simple,
so
I
think
we
kind
of
lose
on
code
complexity,
multihoming
home.
That
does
pretty
nicely.
E
If
you
have
a
layer,
2
network,
you
could
probably
do
it.
You're
probably
gonna
have
multiple
Ras
per
ISP.
The
host
is
gonna
have
to
figure
out
how
to
do
how
to
deal
with
this,
but
it's
pretty
simple
to
specify
I'm,
not
convinced
that
we
actually
win
here.
I
mean
there's
a
lot
of
complexity
in
the
way
that
home
networks
to
make
this
work.
K
E
J
E
But
yeah
and-
and
in
fact
I
mean
that's-
you
know
the
the
the
features
that
I'm
talking
about
for
the
resolver
are
actually
offering
a
service
that
that
I
think
a
layer
to
Gateway
probably
would
never
implement.
So
what
is
up
with?
Is
there
a
way
to
make
that
title
bar
go
away?
So
we
can
read
it's
really
weird.
O
E
E
H
E
That
yeah,
so
a
layer,
2
device,
can
also
do
acne.
It's
pretty
straightforward
again
requires
a
DNS
delegation,
so
in
that
sense,
you'd
have
to
add
some
features
to
the
home.
That
router,
let's
see
so,
and
why
this
is
important
is
because
it
means
that
you
can
have
a
browser
that
doesn't
have
a
bypass
button.
Basically,
a
browser
that
doesn't
ask
you
to
violate
your
own
security,
private
yeah.
You
know
to
do
something
really
stupid
security,
wise
and
train
you
to
do
that.
E
F
H
H
That
would
be
the
good
name
that
you
could
get
right
and
get
it
into
acne,
but
that's
work,
but
actually
one
of
the
things
that
that
that
is
relevant
is
that
you
really
do
want
somehow
to
be
able
to
get
a
TLS
or
or
a
name
that
has
a
ula
underneath
it
so
that
you
can
talk
to
your
router
to
reconfigure
it.
When
you
have
lost
your
connectivity
and
your
browser
has
removed
all
the
bypasses
all
right.
E
H
Implement
link
local,
addressing
with
the
the
percent
blah
blah
blah,
which
they've
refused
to
do,
and
then
they
can.
We
can,
you
know,
say
you
can
remove
the
bypass,
because,
but
part
of
that
is,
though
we
need
to-
and
this
is
this
is
a
potential
real,
actually
work
item
that
we
may
actually
need
to
push
over
into
I.
Don't
know
if
it's
lamps
or
something
else
it
would
be
really
nice
if
we
could
have
a
TLS
cert.
That
was
bound
to.
A
H
Been
a
bit
disconcerting,
but
okay,
that
was
bound
to
an
Ethernet
address.
So
specifically,
if,
if
you
could
get
a
tea,
Leicester
was
bound
to
a
layer
to
address.
First
of
all,
it
would
be
bound
Herbal
obvious
to
a
link
local
address
that
was
derived
from
it,
whether
that's
a
stable,
whether
it's
a
randomized,
Macker
or
not.
But
the
point
is
that
the
vendor
would
be
able
to
ship
it
with
a
certificate
validated
by
the
vendor,
which
says
this
is
really
router
number
a
:,
2
3,
:,
4,
7,
8
right.
H
That
would
be
really
useful,
because
that
would
get
rid
of
this
problem
and
and
I
really
think
we
need
to
somehow
have
this,
and
if
we
don't
do
that,
we
don't
have
some
mechanism,
then
somehow
some
browser
is
going
to
always
have
to
have
that
that
exception,
so
that
you
can
talk
to
things
in
the
beginning
and
it's
a
real
problem.
Ok,
we
did
something
in
our
secure
home
gateway
work
that
I
can
talk
about
later.
H
G
That's
that
that's
not
gonna,
work
and
I
think
that
if
you
see
the
way,
this
is
currently
being
solved
for
the
consumers.
If
you
buy
certain
routers,
they
have
a
you
know
my
router
login
dot
whatever
and
when
they
set
themselves
up
to
be
the
DHCP
server,
the
DHCP
server
a
DNS
server.
They
just
answer
for
all
the
clients
and
they
give
it
they
give
the
correct
answer
for
them
for
the
service
level
answer
for
their
device
with
an
actual
domain
name.
G
N
I'll
be
very
brief,
but
I
want
to
answer
this
because
it's
something
that
I've
witnessed
changing
over
the
last
decade
as
browsers
have
become
more
strict.
This
is
becoming
almost
unusable.
I've
got
a
bunch
of
stuff
on
my
home
network
that
has
embedded
web
servers.
Much
of
that
has
moved
to
HTTPS,
which,
on
the
face
of
it,
is
a
good
thing,
because
it's
more
secure
the
inverters,
my
solar
panels,
when
I
connect
to
them
Safari
says
this
is
a
bogus
certificate.
N
I
can't
trust
this
certificate
and
and
then
it
needs
my
admin
password
to
override
the
security
which
is
very
discouraging
for
an
end
user.
This
very
clear
message
that
you
do
not
want
to
do
this.
This
is
a
bad
idea
and
I
know.
Firefox
did
some
experimental
work
on
how
you
have
TLS
certificates
for
things
that
don't
have
global
unique
DNS
names?
I,
don't
think
it
ever
shipped
in
the
main
version.
But
this
is
a
very
real
problem.
B
E
So
I
ot
support,
so
this
is
actually
to
me
kind
of
the
killer
app
for
home
net
because
you
can
have
your
IOT
edge
routers,
which
is
to
say
the
outside
edge
of
the
IOT,
pointing
at
the
home
net,
join
the
routing
fabric
of
the
home
net,
and
you
can
use
staple
service
discovery
on
the
home
net
and
everything
just
kind
of
works.
So
this
is
actually
a
pretty
sweet
thing
with
layer.
Two.
E
There
are
ways
to
make
this
work,
but
they're
all
kind
of
kludgy
and
they
require
the
host
to
do
things
that
are
questionable
safety
and
getting
transit
between
IOT
networks
can
be
challenging
and
it
actually,
when
I
was
writing
a
so
I
was
thinking.
You
know
if
I
were
an
IOT
edge,
router
provider
I
would
want
to
implement
home
that
in
my
IOT
router,
so
that,
even
if
the
main
router
on
the
link
doesn't
implement
home
net,
my
to
IOT
routers
can
have
a
routing
mesh
between
them
so
interesting.
E
So
what
am
I
missing
I
think
the
strongest
pitch
here
is
IOT
problem
is
that,
unfortunately,
the
people
who
are
most
likely
to
make
routers
are
not
making
them
for
the
IOT
space,
and
so
they
don't
they're
not
going
to
care
about
that
pitch
personally.
I
think
having
a
home
net
services
would
be
much
better
than
having
a
layer
2
network
in
many
ways.
E
If
we
could
solve
the
the
the
roaming
problem,
which
I
think
is
a
real
problem,
so
the
problem
that
I
have
with
this
slide
deck
the
reason
I
composed
this
slide
deck
is
because
I
actually
don't
know.
Based
on
what
I
just
said
to
you,
what
my
pitch
would
be
if
I
went
to.
You
know
a
VP
of
router
vendor
company
to
try
to
pitch
them
on
and
on
implementing
home
net
I.
Don't
know
what
I
tell
them.
I
don't
have
a
clear
message
that
I
think
would
be
convincing.
I,
don't
feel
confident.
E
I,
don't
have
belief
in
this
message,
and
so
even
though
I
personally
as
a
hobbyist
think
home
net
is
great
and
might
be
willing
to
continue
working
out
on
that
basis,
that's
not
going
to
get
us
over
the
finish
line.
If
we
really
think
home
that
is
worth
doing,
we
need
to
answer
these
questions.
We
need
to
seriously
think
about
this,
and
not
just
be
like.
Oh
I'll
come
to
home
net,
because
it's
kind
of
interesting
right.
B
F
B
C
F
B
B
First
I
want
to
just
go
back
to
the
simple
naming:
I've
got
some
more
I've
sent
one
email
on
simple
naming
I'm
going
to
be
sending
some
more
I'd
suggest
that
anyone
else
who
has
comments-
and
some
of
them
do
get
to
some
of
the
things
you
brought
up
in
your
slides,
so
I
think
we'll
have
some
of
that
discussion
on
the
list
as
well
great.
You
said
you
had
some
comments.
B
A
N
E
Well
and
another
outcome
is,
we
could
say,
look
you
know
this
is
really
cool
stuff,
but
actually
it
turns
out
that
having
a
flat
layer,
two
routing
infrastructure
is
a
hell
of
a
lot
easier
and
more
likely
to
ship
than
having
Babel
and
all
that
stuff.
But
you
know
what
we're
also
doing
some
really
interesting
stuff
on
service
discovery
and
naming.
So
let's
just
do
that.
C
F
B
I
Sonoco,
what
a
few
people
have
said
already-
this
has
been
a
fantastic
discussion.
I
really
appreciate
the
very
collegiate
way
everyone
has
come
to
the
microphone
good
on
you.
Yes,
there's
been
some
interesting
stuff
here.
What
I
really
want
to
call
out
is
that
RF,
sieves,
39:35
or
something
like
that
says
the
IETF
goal.
Its
mission
is
to
make
the
internet
work
better.
What
Ted
is
identified
here
is
situations
that
it
doesn't
make
the
internet
work
better.
I
In
fact,
the
exact
opposite
so
I'm
really
keen
to
see
if
it
put
in
to
this
to
either
address
those
issues
or
as
Steven
mentioned
go.
You
know
what.
Maybe
we
can't
I'd
hate
to
think.
That's
the
outcome
and
I
think
Stephens
right
that
we
need
to
put
in
a
lot
of
effort
not
to
convince
ourselves
but
to
have
the
answers
as
to
how
we
make
it
better.
So,
yes,
I,
think,
there's
interesting
work
here
and
I'm
going
to
let
it
go.
It
has
to
keep
going
for
those
reasons.
I
I,
don't
I
really
do
dislike
unanswered,
States
and
people
throw
their
hands
up
and
go
it's
all
too
hard.
That
said,
I'm
also
taking
on
board
some
of
the
aspects
with
this
of
it
takes
effort.
It
takes
a
lot
of
effort,
it
takes
people
to
be
invested,
it
takes
time
and
it
takes
money,
and
so
I'm
going
to
be
watching
that
as
well.
Please
continue
Thanks.
I
B
G
E
Sorry,
yes,
okay,
so
a
new
agency
few
work:
this
is
the
overview
publishing.
We
need
to
be
able
to
publish
keys,
I,
don't
writers
need
to
have
public
keys
and
private
keys.
We
need
to
be
able
to
name
links.
We
need
to
be
able
to
elect
HomeNet
routers
to
do
service
discovery
for
each
link,
so
there
might
be
more
than
one
router.
We
need
to
be
able
to
maintain
a
list
of
stateful,
authoritative
services
which
may
be
empty.
We
need
to
be
able
to
elect
a
stateful
primary
and
we
need
to
fix
the
ipv4.
E
Addressing
fail
and
I
will
now
explain
what
those
are
in
less
than
five
minutes.
So
each
opponent
Robert
needs
to
generate
its
own
public/private
key
pair
I
can
publish
the
public
key
using
HNC
key
now,
every
hnr
knows
every
other.
Hnr
is
public
key.
We
can
use
this
from
mutual
authentication.
This
does
not
help
us
to
establish
trust,
but
if
we
can,
if
we
find
a
way
to
establish
trust,
we
can
use
it
on
mutual
authentication
and
we
can
use
it
for
DNS
x,
signing
and
validation
link
naming.
E
We
need
to
be
able
to
come
up
with
names
for
links,
and
links
may
be
connected
to
multiple
H
in
ours.
The
link
name
is
somewhat
dependent
on
the
name
of
the
hnr
than
the
link
name.
Mine
might
change
if
an
H
n
R
goes
away.
So
if
you
read
the
naming
architecture
which
two
of
you
have
done
or
three,
maybe
then
you'll
know
how
that
all
works
and
understand
why
this
is
important
per
link.
Discovery
proxy.
E
So,
as
I
was
saying
earlier,
every
link
has
potentially
one
well
has
one
or
more
at
least
one,
but
possibly
more.
Each
ni
was
connected
to
it.
Only
one
hn
r
can
be
authoritative
for
service
discovery
on
the
link,
so
we
need
to
be
able
to
elect
which
one
that
is
list
of
state
servers
just
maintain
the
list
pretty
straightforward.
Basically
any
hnr,
that's
willing
to
do
stateful
says
hi
I'm
willing
to
do
stateful.
We
might
want
to
be
able
to
say
I'm
willing
to
do
stable,
but
I
prefer
not
to
so.
E
There
might
be
that
slight
additional
amount
of
complexity
and
then
election
of
the
DNS
primary.
You
know
it
is
the
time
of
the
gathering.
There
can
be
only
one
home,
it
routers
can
come
and
go.
So
if
the
primary
goes,
then
we
need
to
elect
new
primary.
If
some
new
router
shows
up,
then
we
need
to
join
it.
I
E
The
failure
mode,
the
worst
failure
mode
is
if,
if
there
are
two
hnr
is
one
of
which
hasn't
yet
been
configured
and
it
comes
on
to
the
network,
just
as
the
primary
goes
away
and
now
you
have
lost
all
of
your
state,
but
we
need
to
figure
that
out.
I
think
we
don't
really
need
to
solve
that
problem,
because
I
think
it's
unlikely,
but
and
and.
E
We
heal
so,
but
we
need
to
do
that
and
then
the
ipv4
address
fail,
which
is
it
right
now,
if
you
have
ipv4
addressing
on
your
internal
network,
will
be
numbered
and
have
routing
for
ipv4
and
then,
if
your
upstream
provider
ipv4
address,
expires,
ipv4
addressing
will
just
die,
and
this
is
very
surprising,
I
know
because
I've
had
it
happen
to
me
if
service
discovery
and
ipv6
writing
were
working.
Maybe
that
would
be
okay,
but
in
my
case
that
was
not
because
I
was
developing
on
it.
E
I
E
At
least
be
it
should
there
should
at
least
be
a
delay,
but
anyway
that's
it
Oh
implementations.
This
is
a
kind
of
a
problem
as
far
as
I
know,
there's
nobody
who
both
Grox
and
H
an
existing
H
NCP
implementation
and
has
time
to
work
on
it.
That
includes
me.
We
need
to
either
get
one
or
more
of
those
people
to
spend
time
on
it,
or
we
need
to
put
some
effort
into
making
an
existing
H
NCP
implementation,
more
Gras
Keable.
There
are
I,
think
three
of
them
out
there.
E
Maybe
one
of
them
is,
is
Gras
cabal
I'm,
not
that
fond
of
them
so
or
we
need
a
new
one.
That's
more
Gras
cabal!
That's
a
little
bit
better!
Architected
I
think
you
know
it
might
be
taken
personally
that
I
say
this,
but
the
existing
implementations
are
kind
of
piles
and
there
isn't
a
lot
of
layer
separation,
and
so
it's
very
difficult
to
understand
it's
very
difficult
to
modify,
and
so
it
might
be
worth
doing
something.
That's
a
little
bit
more
general.
So
that's
it
I
think
I.
B
B
E
Well,
I
think
we
need
to
have
discussion
on
the
mailing
list
of
people.
I
mean
I.
You
know
we
keep
seeing
rooms
with
like
actually
kind
of
a
nice
number
of
people
in
them
and
I.
Don't
know
whether
everybody's
just
in
here,
because
it's
convenient
place
to
read
mail
or
whether
it's
sort
of
like
watching
a
train
wreck
in
slow
motion
or
whether
people
are
actually
interested.
But
if
the
answer
is
that
they're
actually
interested,
then
even
a
little
bit
of
engagement
would
be
better
than
the
amount
of
engagement
we
have
right
now.
Okay,.
H
H
That's
a
good
spot
there,
yeah
I've
been
working
since
May
with
Sierra
labs,
chocolate,
litter,
I,
think
talked
to
you
in
July,
when
I
didn't
come
to
the
meeting.
Surprisingly,
and
basically,
what
we've
been
working
on
is
we
call
it
a
secure
home
gateway
and
a
couple
of
the
goals
are.
One
of
them
is
that
every
device
comes
with
a
name
and
zero
is
the
Canadian
internet
registration
authority,
so
dot
CA,
and
they
obviously
have
an
interest
in
more
people
having
dot
CA
names.
B
H
There's
audio
here,
but
if
the
audio
doesn't
work,
then
all
those
explained
years
ago
anyway.
So
so
the
point
was
is
that
we
are
building
a
secure
home
gateway.
It
processes,
mud
files.
You
could
have
heard
about
that
in
a
couple
other
working
groups
and
the
general
effect
is
that
it
basically
restricts
devices
that
are
are
that
you
can
describe.
There
are
purpose,
specific
purpose
devices
rather
than
general
competing
laptops
or
whatever
to
doing
just
what
they
need
to
do.
H
Okay,
so
this
was
actually
presented
at
ripe
and
I
can
in
the
last
couple
weeks,
and
we
were
gonna,
do
a
demo
on
site
and
we
realized
that
was
impractical
because
ripe,
wouldn't
let
us
have
a
higher
drop,
and
so
we
did
a
video
and
we
actually
had
built
this
little
little
box
that
you
know
it
was
a
three
hour
effort
to
basically
push
one
button
and
it
rhymes.
Normally
you
push
the
other
button
and
it
apparently
attacks
something
right
and
then
we
could
show
it
and
that
didn't
work.
P
P
J
P
P
It
says
a
fine
mud
profile,
so
it
just
finished.
What
it
did
here
is
that
essentially
downloaded
the
mod
file
it
applied
all
the
rules.
You
know
we're
in
our
viral
and
now
the
device
is
secure
and
it
acts
as
only
the
it
has
only
traffic
that
it
can
actually
access
and
everything
else
should
be
blocked
and
secure
and.
H
H
We
would
much
prefer
to
be
able
to
do
things
among
different
networks
that
HomeNet
would
let
us
have,
and
in
particular
we
have
no
ability
to
isolate
devices
that
are
not
hair
pinned
through
the
router.
So
if
you
plug
in
a
switch
which
many
of
us
might
do
and
then
plug
in
some
wired
devices,
well,
of
course,
we
can't
isolate
those
de
choir
devices,
because
we
don't
it's
not
a
switch
that
we
can
manage
directly.
So
it
would
be
better
if
that
was
a
home
net
router.
H
That
just
was
in
pass-through
mode
because
then
it
could
do
that
right.
So
if
you
needed
a
4
port
or
8
port
router
somewhere
and
you
just
got
one
and
it
happened
to
be
a
home
net
rotor,
that
would
be
better
because
now
it
would
be
smart
and
we
would
have
HN
CP
and
we
could
actually
do
things
among
different
things.
H
L
Dave
they
were
cool
demo.
Thank
you
for
doing
this,
I
think
so
Mike.
Why
I
had
a
couple
questions,
but
I
think
that
they
might
have
been
answered.
So
let
me
verify
that
I
have
your
right
answer.
The
first
question
I
was
gonna:
ask
is
where
do
they
get
the
mud
file
from
and
the
answer
was
it's
encoded
in
that
QR
code
right
and
so
because
that
was
when
we
finally
got
the
audio
working.
That's
what
was
being
explained
right,
I'd
scan
that
that's
the
or
ID
to
go
and
grab
the
mud
file
right.
F
L
H
The
answer
is
that
if
you
have
a
device
to
which
you
want
to
let
to
do
anything
you
want
it
wants,
then
that
would
be
essentially
swipe
up.
Okay
and
that
might
because
you
don't
have
a
mud
file
or
you
haven't,
found
a
mud
file
or
you
haven't
built
one
yet
or
because
it'll
be
inappropriate
for
your
laptop
to
be
restricted
and
where
it
went.
H
But
that's
the
discussion
that
we're
having
so
we're
actually
considering,
for
instance,
things
like
looking
at
the
first
three
bytes
of
the
of
the
MAC
address
and
having
a
database
that
may
actually
return
interesting
or
useful
things.
We
strongly
believe
that
we
need
to
have
some
kind
of
a
curated
crowdsource
database
of
this.
It
could
be
as
simple
as
a
github
that
you
know
accepts
pull
requests
and
does
some
thinking
before
it
accepts
them.
H
It
could
be
much
more
complicated
like
Stack,
Exchange
or
something
like
this,
but
we
think
that
people
are
going
to
be
able
to
do
this
and
so
that
QR
code
doesn't
have
to
come
from
the
packaging.
It
could
come
from
a
screen
write
or
in
this
case
movie
you
don't
need
to
screen,
goes
directly
from
one
app
to
the
other,
but
the
point
is
that
that
that
could
come
from
a
variety
of
sources
and
it
could
become
the
public
on
people
and
the
other
part
of
it
is
that
we
realized.
H
L
There
the
question
I
was
gonna,
ask
which
I
think
was
answered
was
what
work.
If
any
does
home
that
need
to
do
and
I
think
you'd
said
that
right
now,
all
of
your
apples
and
stuff
is
based
on
the
MAC
address,
which
may
be
gotten
from
the
QR
code
or
something
to
you.
I,
don't
know,
and
maybe
there's
something
for
home
had
to
do
there.
That
was
why
I
took
away,
and
then
my
let
me
know
if
that's
not
correct,
but
I.
L
H
H
We
are
looking
at
the
question
of
how
do
we
onboard
the
router
itself,
given
that
it
has
no
internet
when
we
onboard
it,
I
had
only
three
minutes
that
anama
to
talk
about
that
document,
but
my
slides
are,
and
that
are
kind
of
and
I
would
love
your
feedback
and
related
to,
but
not
identical
the
DPP,
because
DPP
is
not
deployable
on
current
iOS
or
Android
devices.
There
are
no
API,
so
send
the
packets
that
we
like,
and
so
I
would
prefer
to.
H
B
K
Michael
Abramson,
so
I
just
installed
okay,
so
there
has
been
some
talk
about
different.
You
know
not
getting
a
TV
6
and
so
on.
I'm.
In
the
fortunate
situation,
I
have
one
DOCSIS
connection:
I
want
to
fibre
to
the
home
connection,
they're
both
dual
stack
that
both
get
slashed
56
is
through
them.
So
I
said:
okay,
I'm
gonna,
try
a
home.
It
so
took
opened
up
here
to
1806
point
one
audience
to
try
to
install
it.
The
the
home
that
suite
of
packages
I
got
some
conflicts.
I
had
to
remove
the
existing
HPV
6
server.
K
Then
home
net
installed
correctly
I
had
to
fight
a
little
bit
with
the
configuration
and
losing
connectivity
and
I
would
muck
around.
But
after
a
while
I
got
it
working
and
I
had
to
lock
up
links
and
and
get
both
prefixes
and
it
was
announced
on
my
land
and
so
on.
I
took
another
one:
I
did
the
same
thing
hooked
it
up
to
to
that.
First,
one
and
Babel
seemed
to
kick
around.
Everything
was
working
and
it
didn't
do
double
mat,
but
single
mapped
out
on
the
external
interface
and
so
on.
K
That
seemed
to
work,
then
I
noticed
that
I
was
getting
I
had
one
week
of
least
time
from
one
is
be,
and
one
now
with
12
hours
from
the
other
one
and
the
12
hour,
one
was
at
most
one
I
really
wanted
to
use
as
a
primary.
So
all
my
host
chose
to
to
the
seven-day
one,
of
course,
because
of
the
lifetime
of
the
prefix.
So
that's
one
thing
we
should
do.
We
should
put
in
the
ability
to
cap
the
lifetime.
B
K
B
K
E
K
K
When
my
hosts
are
modified
and
they
were
smarter
and
they
run
in
PTC
and
so
on,
they
will
use
both
so
but
I
want
to
be
able
to
influence
this
right.
Okay,
because
right
now,
with
my
unmodified
host,
they
just
went
100%
to
the
one.
I
didn't
want
to
use
this,
which
was
not
desirable
so
either
by
default
name,
taking
the
lowest
value
and
saying
this
is
both
and
then
another
thing
is
the
way
I
had
set
it
up.
K
I
hack
Chile
had
I,
did
I
wouldn't
get
a
link
down
event,
so
there
is
some
other
work
on
detecting
I
think
I'm,
proposing
that
in
v6
up
to
prefer
to
check
their
life
liveliness
of
the
upstream,
because
here,
in
this
case
we
actually
want
to
fail
over
to
the
second
one,
because
for
me
this
was
no
more
of
a
redundancy
case.
I
wanted
to
be
able
to
if
one
went
down,
I
wanted
to
use
the
other
one,
so
that
doesn't
work
it
currently
out.
K
But
apart
from
that,
the
code
seems
to
know
my
I
ran
this
for
a
couple
of
hours
before
I
said:
okay,
I
don't
want
to
use
this
from
a
live
with
network
at
home,
the
same
wife.
You
know
observation
that
Darren
had
so,
but
from
that
part
it
still
seems
to
work
in
1806
or
one
that
the
code
seems
to
be
I.
Don't
know
how
much
it's
being
maintained.