►
From YouTube: IETF104-HOMENET-20190326-0900
Description
HOMENET meeting session at IETF104
2019/03/26 0900
https://datatracker.ietf.org/meeting/104/proceedings/
A
B
B
B
A
That's
just
push
here,
just
keep
pushing
it.
It
works.
C
A
A
A
Welcome
Eric
and
then
we
have
basically
one
person
tell
you
stack
it
here.
No,
so
we
have
a
couple
of
drafts.
Pettalia
may
choose
at
some
point
to
go
to
the
mic
and
talk
a
bit
and
other
than
that.
It's
the
Ted
show,
and
then
we
want
to
discuss
where
we
want
to
go
from
here,
because
we
tried
to
have
this
discussion
on
the
list
and
the
the
the
lack
of
activity
was
underwhelming.
B
D
D
You
know
that
well,
at
the
last
at
the
last
ITF
in
Bangkok,
I
presented
a
topic
about
marketing
HomeNet
like
what
what
is
the
market-
and
you
know
what
is
the
market
for
things
like
home
net
routers
right
now,
and
how
does
that
compare
to
what
we've
been
doing
here?
Is
there
anything
that
we've
been
doing
here?
That
would
be
useful
on
those
routers?
Is
there
anything
we
can
learn
from
from?
D
What's
actually
succeeding
in
the
market
right
now,
and
so
I
wrote
up
a
document
that
actually
talks
about
that
problem
and
in
quite
a
bit
of
detail
and
tries
to
systematically
compare
what's
available
on
various
multi
router
home
networks,
to
what
we've
done
and
so
I.
The
document
is
draft
lemon
HomeNet
review
I
couldn't
really
come
up
with
a
good
title.
I
didn't
want
to
call
it
home
net
marketing
document
because
I
figured
nobody
would
read
it,
but
but
it's
basically
a
market
analysis
and
I.
D
Think
it's
pretty
important,
because
there's
no
point
in
us
doing
this
work.
If
there's
no
market
for
it
right
doesn't
have
to
be
a
huge
market,
but
but
somebody
ought
to
be
benefiting
from
this
or
else
we're
just
kind
of
standing
around
wasting
our
time.
So
I'd
really
appreciate.
If
people
could
take
a
look
at
that
document
critique,
it
suggest
things
to
add
to
it
and
at
some
point
we
should
probably
even
consider
drawing
some
conclusions
from
it.
I'm
not
going
to
present
the
document
in
detail
here,
because
I
already
did
that
in
bankok.
D
Okay,
sorry
about
that,
I
should
have
actually
asked
the
chairs
to
put
it
on
the
agenda
by
name
so
anyway.
So
so,
please
take
a
look
at
that
and
let's
discuss
it
on
the
mailing
list.
I
think
there's
some
good
stuff
to
discuss
in
there.
So
hopefully
that's,
hopefully
that
will
generate
some
discussion.
D
The
next
thing
that
I
want
to
talk
about
is
home
net
and
open
wrt.
Let
me
just
go
back
to
my
sub
agenda.
Oh
yeah,
right,
okay,
yeah,
so
I
haven't
done
any
work
on
the
simple
naming
document
since
the
last
IETF,
because
I've
spent
all
of
my
time
working
on
the
stuff
that
we
need
to
implement
simple
naming
instead
seemed
like
a
good
trade
off,
so
so
what
I
did
for
for
today's
presentation
was
to
just
go
through
the
simple
naming
document
and
see
like
you
know.
Where
are
we?
What
do
we
have?
D
And
you
know
what
do
we
still
need
to
do,
and
this
was
actually
really
instructive.
I
think
as
a
result
of
doing
that,
I
have
quite
a
few
changes
that
I'll
be
making
to
the
document,
which
I
think
will
improve
it
substantially
and
make
it
a
lot
easier
for
people
to
read
it
and
review
it.
So
it
was
all
good,
so
I'm,
just
gonna
go
through
all
the
sort
of
headings
of
the
document
talk
about
what
what's
in
there.
D
E
D
D
There
are
a
few
things
that
need
to
be
enhanced
about
how
H
NCP
derives
that
information,
but
for
the
most
part
it's
there,
but
we
don't
have
a
way
to
extract
that
information
from
the
H
NCP's
state
and
put
it
into
an
authoritative,
DNS,
server
and
I.
Don't
think
we
want
to
make
H
NCP
itself
be
an
authoritative
server.
We
could
do
that,
but
that
just
doesn't
seem
like
the
right
solution
to
me.
So
that's
that's
a
bit.
We
have
to
figure
that
out.
D
D
F
D
Of
the
problems
that
I
have
with
HNC
PD
is
that
is
that
all
of
the
things
that
it
knows
are
contained
within
it
and
so
actually
seeing
what
it
knows
is
not
easy,
and
so
so
basically,
what
I'm
talking
about
is
that
I
would
like
to
be
able
to
to
have
that
I
like
to
be
able
to
have
something:
that's
sort
of
equivalent
to
you,
boss
or
d-bus,
for
talking
to
HN
CPD
and
getting
the
information
that
it
has.
And
if
we
had
that,
then
that
could
be
used
to
populate
the
authoritative
server.
So
we.
F
D
About
the
implementation,
this
is
all
about
implementation.
This
is
not
about
like
the
protocol
has
available
yeah.
No,
it's
yeah.
It
really
is
a
matter
of
taste
like
like
I'm,
not
saying
it.
The
way
that
it's
currently
implemented
is
wrong.
I'm,
just
saying
this
is
not
to
my
taste.
So
so
we
can
fake
up
Authority
for
a
single,
broader
use
case
and
a
lot
of
a
lot
of
the
stuff.
D
D
Trying
to
figure
out
what
this
is
on
the
slide,
all
right
yeah.
So
one
of
the
things
that
that
we
need
to
be
able
to
do
is
choose
an
authoritative
server.
If
there's
an
authoritative
server
and
there
needs
to
be
a
primary
and
that
is
to
say,
if
there's
a
stateful,
authoritative
server,
then
there
needs
to
be
a
primary.
If
it's
all
stateless,
then
we
don't
really
care
we're.
Just
getting
the
information
out
of
H
n
CP,
every
router
will
have
it.
They
can
all
be
authoritative
for
the
same
zone.
We
don't
care.
D
If
there's
state,
then
we
need
the
state
to
be
in
a
particular
place.
So
if
we're
gonna
do
that,
then
we
don't
currently
have
support
for
electing
a
stateful
router
NH
NCP
I,
don't
anticipate
that
being
difficult,
but
we
need
to
do
it
and
then
for
the
stateless
case.
We
need
to
actually
we're
missing
some
of
the
stuff
that
we
need
to
name
the
the
purling
subdomains
that
we
would
delegate
so
reachability.
This
is
the
ability
to
query
things
on
the
home
net
from
outside
of
the
home
net.
D
D
Link
naming
so
there
I
came
up
with
a
way
to
name
links
that
I
think
is
sort
of
human,
readable
and,
and
it's
described
in
the
simple
naming
architecture.
Basically,
it
takes
the
name
of
the
router
and
the
name
of
the
links
on
the
router,
and
if
there
are
two
routers
with
the
same
name,
it
gives
one
of
them
a
different
number
than
the
other,
and
that
lets
the
user
sort
of
see
a
name
that
maybe
they
could
figure
out
what
it's
referring
to.
D
If
they
can't
it's
not
the
end
of
the
world,
but
it's
better
than
just
like
you
know:
hex
vomit,
as
Stewart
likes
to
call
it
so
that
needs
to
be
added
to
H
NCP
again,
it's
it's
exactly
the
kind
of
thing
that
H
NCP
knows
how
to
do
so.
I,
don't
think
it's
hard,
but
we
haven't
done
it
so
and
of
course
we
can
fake
this
up
for
the
single
link,
as
I
was
saying
previously,
but
yeah
there's
a
lot
of
new
work
to
do
here.
D
We
need
link
names
because
we
need
to
list
the
browsing
domains.
We
need
to
list
the
registration
domain.
So
if
you
want
to
do
service
discovery,
you
need
to
know
where
to
do
it,
and
so
we
need
to
have
a
list
of
all
of
the
domains
in
which
a
host
on
the
home
net
needs
to
do
queries
in
order
to
discover
services
throughout
the
home
net.
D
G
Ted
yeah
Robin
Wilton.
To
what
extent
does
this
depend
on
good
behavior
by
the
vendors,
because
so,
for
example,
if
I,
if
I
look
at
the
device
names
on
my
own
normal
home,
routers
DHCP
list,
some
of
them
are
ones
that
I've
been
able
to
set.
Some
of
the
things
are
things
like
an
HP
printer
that
has
HP
at
the
beginning,
yep.
G
D
Specifies
what
the
correct
behavior
is
and
then
whether
the
vendor
follows
it
is
up
to
them,
but
certainly
certainly
I
mean
one
of
the
reasons.
I
think
that
you
know
your
HP
printer
is
just
hex.
Vomit
is
because
they
don't
have
a
better
thing
to
do,
and
so
what
I
tried
to
do
here
was
suggest
a
better
thing
to
do.
H
Do
I
think
because
we're
so
danyoung
we
go
so
I
think
because
it's
like,
if
you're
a
router,
you
plug
it
and
you
say
well
if
it
is
not
routing
properly
so
we're,
because
we
have
a
lot
of
functions
within
the
same
box.
We
expect
it
to
have
a
different
length.
So
I
mean
this
impacts
of
that
implementation
should
be
limited,
I
guess,
yeah,
it's
a
it's
a
bundle
of
functions
so
I
guess
that's.
D
So
per
link
Authority
again
just
going
down
the
subheadings
in
the
in
the
document
requires
that
that
we
choose
a
dns
SD
proxy
for
each
link.
So
remember,
home
routers
are
talking
to
each
other
over
shared
links,
which
means
that
for
many
of
the
links
in
the
home
net
there
will
be
more
than
one
router.
That
could
act
as
the
DNS
is
d
proxy,
and
at
present
we
don't
actually
have
a
way
to
decide
which
one
of
those
two
routers
are
three.
However,
many
it
is,
is
actually
going
to
do.
D
Dns
SD
server
service
on
the
shared
link.
So
we
need
to
add
that
to
H
NCP,
that's
not
done
or
even
specified.
Yet
we
need
the
authoritative
server
or
servers
to
delegate
to
each
link
zone.
So
if,
if
H
NCP
knows
which
which
router
is
going
to
do
service
for
which
link
and
we're
doing
stateless
authoritative
service,
then
we
would
actually
have
all
the
information
we
needed
at
that
point
to
set
it
up.
D
This
is
not
hard,
but
we
haven't
done
anything
of
it.
None
of
this
stuff
is
done.
It's
not
really
that
important,
because
in
order
to
in
order
for
reverse
mapping
to
work,
hosts
need
to
do
something
special
which
zero
hosts
currently
will
be
doing.
So
this
is
not
a
high
priority.
That's
why
I
said
not
very
import.
It's
not
very
important
for
two
reasons:
one:
it's
not
a
high
priority
in
two
nothing's
gonna
break.
If
it's
not
working,
probably
we
don't
so.
D
D
A
D
I
mean
that's
just
a
now
you're,
just
using
a
different
court
and
a
different
transport
right.
It's
not
there's
nothing!
Nothing!
None
of
the
plumbing
changes,
I
think
other
than
you
might.
If
this.
So
so.
If
this
we're
gonna
do
doe
upstream,
then
we
would
need
different
plumbing
upstream.
If
it's
gonna
do
doe
downstream,
then
we
need
different
plumbing
downstream,
but
the
actual
wiring
inside
of
the
thing
I,
don't
think
changes,
but
but
we're,
but.
A
D
So
so
I
mean
one
of
the
one
of
the
issues
with
with
doe
or
with
anything
where
you're
tunneling
DNS
is
that
you
need
to
have
some
kind
of
exception
process,
or
else
you're
gonna
tunnel,
things
that
can't
be
answered
by
by
an
external
resolver.
So
yeah,
good
point
and
I
think
that
the
document
actually
says
that
but
you're
right
it
and
I
didn't
mention
that
here.
So
that
is
another
loose
end
that
we
should
talk
about
or
figure
out.
D
D
Yeah
I
mean
you
know
personally,
I
would
be
very
tempted
to
do
to
have
my
home
net
rod
or
do
dot
to
some
external
service
and
then
just
use
the
name
resolution
on
the
local
link
when
I'm
on
that
local
link,
because
you
know
I
trust
myself
more
to
to
set
that
up
than
I
trust.
My
random
IOT
device
vendor,
but
yeah
so
DNS
push
so
DNS
push
exists
to
give
us
feature:
parity
with
multicast
DNS
when
doing
service
discovery
over
DNS
with
multicast
DNS.
D
You
send
out
a
query
and
you
get
back
answers
as
they
come
in
and
you
retry,
and
maybe
you
get
more
answers,
and
so,
as
answers
come
in
the
multi,
the
EM
DNS
client
will
add
and
will
add
information
to
the
user
interface.
If
there's
a
user
interface
or
will
potentially
establish
connections
to
new
services
or,
however,
it's
using
this
information,
the
point
is
that
updates
will
come
in
as
new
information
arrives
with
DNS.
D
Typically,
you
just
do
a
query
and
you
get
a
response
and
that's
the
end
of
it,
but
which
is
actually
not
ideal
like
it
actually
causes
problems
in
real
operation.
So
so
we
added
the
ability
to
just
say:
hi
I'd,
like
you
to
tell
me
whenever
the
meaning
of
this
name
changes
and
that's
what
DNS
push
doesn't
so
our
discovery
proxy
does
DNS
push
as
a
client,
sorry
as
a
server,
and
so
if
you
have
a
client
on
the
on
the
local
wire
that
knows
how
to
do
DNS
push,
it
will
do
it.
Bring.
I
Those
you
know
that
all
these
pieces
are
great,
but
the
question
is:
is
how
do
you
make
sure
they
all
sort
of
come
up
automatically
right?
Because
if
the
DNS
discovery
proxy
stuff
and
the
relay,
you
got
a
configure
stuff
right,
the
security
associations
and
stuff
like
that.
So
if
you
looked
at
that
part,
you
know
how
you
might
do
that
in
this
sort
of
more
automated
home
that
environment
well,.
D
I
E
D
Yeah,
so
the
answer
is
right:
now
we
don't.
Actually
we
haven't
actually
solved
that
problem
to
do
TLS.
If
we
have
ipv6
in
principle,
we
can
just
go
use
Acme
to
get
TLS,
certs
and
and
I
think
I
actually
mentioned
that
in
the
document
but
yeah.
This
is
a
somewhat
hard
problem,
yeah,
yeah
and-
and
it
would
be
nice,
for
example,
if
browsers
had
a
way
to
to
use
DNS
DNS
SEC
to
sign
assert
whether
than
using
PKI,
maybe
with
different
security
properties.
D
But
we
don't
have
that
yet
far
as
I
know,
I
think
it
was
attempted
at
one
point
and
then
Google
pulled
it.
So
oh
whoa
round-robin
hang.
So
this
is
really
technical,
trivial
stuff.
But
one
of
the
things
in
the
document
specifies
is
that
when
you're
doing
DNS
resolution,
if
you
have
multiple
servers,
you
want
to
you
know
if
one
of
them
isn't
working,
try
the
others
and
that's
basically
working
because
we're
just
in
our
implementation,
we're
just
using
mdns
responder
to
resolve
queries
and
endianness.
D
D
Any
of
the
PVD
authors
are
in
the
room-
oh
hi,
yeah,
of
course,
because
you're,
your
now
official
or
semi
official,
so
yeah.
So
so
we
need
the
PVD
are
a
option.
In
order
to
do
this,
we
don't
have
it
yet
it's
not
it's
not!
Is
it
like
close
to
last
call
or
what's
going
on
with
that?
Okay
yeah
so
hopefully
close
to
last
call,
but
not
currently,
so
so
we're
basically
waiting
on
that
and
there
would
be
there.
That's
gonna
require
a
fair
amount
of
work
to
implement.
F
Trouble
check
again,
could
you
perhaps
describe
how
you
envisioned
that
to
work?
So
one
thing
that
I'm
worried
about
is
that
we've
done
a
lot
of
work
in
at
the
agency
P
at
Babel
levels
to
make
sure
that
home
networks
well
with
multiple
simultaneous
ISPs
and
all
the
naming
stuff
is
a
little
bit
worrying,
because
there
are
some
unwritten
assumptions
and
what
you've
been
presenting
that
there
is
only
one
is
P.
D
That
that's
a
great
question
so
so
there
are
basically
two
approaches
to
handling
multiple
ISPs.
One
of
them
is
to
use
provisioning
domains.
Ideally,
we
use
provisioning
domain.
It's
like
if
a
host
supports
provisioning
domains,
we
use
it.
Of
course,
the
number
of
hosts
that
currently
supports
the
provisioning
domain
are
a
option
to
the
best
of
my
knowledge
is
zero.
So
so
we
can't
rely
on
that,
and
so
all
we
can
really
do.
We
don't
really
have
a
way
to
to
know
how
a
host
is
gonna
use
the
information
that
it
gets
out
of
DNS.
D
We
do
have
some
degree
of
trust
that
most
modern
hosts
do
happy
eyeball
style
stuff.
So
if
a
particular
answer
from
DNS
doesn't
work
with
the
source
address
that
they
choose,
they'll,
try
a
different
source
address,
and
so
essentially
the
way
we
solve.
This
problem
is
by
deliberately
round-robin
amongst
provisioning
domains.
If
we
don't
have
a
tag
on
the
happy.
D
Write
the
source
address
is
kind
of
random.
We
don't
have
any
control
over
source
address
selection,
so
whatever
the
host
does
is
what
it
does.
So
all
we
can
do
is
hope
that,
just
by
trying
a
bunch
of
different
things,
eventually
the
host
chooses
a
combination
of
source
address
and
destination
that
work.
We
we,
we
don't
have
a
way
to
tell
it.
What
to
do
there,
I
mean
I,
guess
it's
possible,
we
could
use
the
DHCP
source
address
selection
stuff,
but
this
is
good.
We've
got
a
lot.
F
Perhaps
you've
misunderstood
each
other,
so
what
I'm
worried
about
so
this
the
layer-3
stuff
I'm
familiar
with,
but
I'm
would
like
to
understand
this.
What
happens
if
you
have
multiple
ISPs,
with
multiple
DNS
servers
right
and,
of
course,
you're
telling
them
to
not
do
to
have
global
DNS
only,
but
of
course
the
DNS
servers
are
not
equivalent
because
I
speeds
are
what
they
are
yeah.
J
F
D
D
So
so,
in
other
words,
you
would
you
would
try
IP
address
number
one
and
that
would
get
you
PVD
number
one,
and
then
you
try
IP
address
number
two,
and
that
would
give
you
PVD
number
two
it,
and
that
way
you
would
get
more
than
one
answer
to
the
same
query,
and
so
if
one
of
them
didn't
work,
you
get
some
from
other
ISPs,
so
you're.
Assuming
that.
F
D
K
K
Exactly
and
it's
even
more
than
myth
was
working
on
because
we
have
this
huge
complication
know
it
do.
H
is
coming
along
as
well,
so
it
did
I
mean
we
already
have
the
problem
with
a
device
for
having
many
VPNs
having
multiple
solvers
and
now,
all
of
a
sudden,
you
might
even
connect
this
VPN
to
your
home
router.
So
the
home
router
now
has
five
different
up
links.
Five
different
resolver
sold
begging,
mirror
attention
to
please
use
me
and,
and
we
had,
we
don't
have
a
solution
of
this
I
mean
at
home.
K
I
have
I,
have
five
four
different
VMs
running,
and
each
of
these
are
sitting
in
a
separate
VPN
partner
of
the
problem
is
solve
this
problem
by
attaching
the
one
by
one,
because
I
don't
have
a
PVD
solution
to
to
do
what
any
so
yeah.
This
is
something
we
should
in
which
it
doesn't
necessarily
need
to
be
worked
on
here
because
I,
don't
think
homeless
consultant,
but
it
needs
to
be
worked
on
elsewhere.
K
D
D
It's
the
problem
is
lack
of
focus
in
both
places.
I
mean
honestly
I
think
we
have
more
focus
here,
except
that
this
room
is
too
small:
they're,
not
of
people
showing
up
so
I.
Don't
think
that's
good
either!
I,
don't
really
know
what
the
right
solution
I
mean.
We
could
try
doing
a
new
v6
ops
and
then
see
what
happens
and
if
there's
sufficient
interest
that
we
may
want
to
to
do
a
short-lived
working
group
that
just
solves
this
problem.
Eric.
L
Yeah
everything
about
the
PVD,
just
clarification
with
I'm,
pretty
sure
you
know
it
anyway,
we've
DVD
with
zero
implementation.
Indios,
that's
a
bit
part
of
it.
You
get
a
complete
bender
of
the
address,
the
next
stop
and
the
DNS
resolver
right.
So
you
will
always
ask
the
right:
I
mean
you
can
ask
always
the
resolver
with
the
right
source
address.
So
it's
sort
of
a
lot
of
problem
yeah,
but
mostly
you
know
it
already.
D
So
I
was
being
a
little
bit
hypothetical
there
and
I'm
realizing
that.
Actually,
what
I
proposed,
what
I
wanted
proposed
in
the
document
as
a
single
RA,
as
opposed
to
so
so?
In
other
words,
the
the
host
wouldn't
be
doing
multiple
provisioning
domains
unless
it
supports
it
with
them,
and
if
it
is
supporting
it,
then
it's
not
round
robin
ik.
It's
just
like
doing
queries,
though
the.
M
D
D
First-Serve
naming
requires
there
to
be
a
stateful,
authoritative
server,
so
I
have
code
that
will
take
an
SRP
update,
their
sort
of
specially
formed
updates,
validated
and
then
in
principle,
do
a
DNS
update
based
on
it,
but
that's
not
actually
done
and
I've
also
been
talking
with
Mark
Andrews
about
maybe
getting
Service
registration
protocol
into
by
9:00,
which
would
be
very
cool.
We
could
also
do
this
in
MV,
NS
responder
and
just
have
the
updates
go
into
the
cache,
but
the
bottom
line
is
that
we
don't
have
code
to
do
this
right
now.
D
We
don't
have
working
code.
Do
this
right
now
we're
not
super
far
from
that,
but
we
don't
have
it
so
that's
work
that
needs
to
be
done.
If
we're
gonna
support
SRP
on
the
home,
node
and
I,
think
it's
worth
supporting
SRP
on
the
home
net
from
the
beginning,
because
if
we
don't,
then
vendors
aren't
going
to
implement
it.
You
know
why
implement
something
if
nothing
is
gonna
hear
it.
D
D
D
Thanks
alright,
so
then
I
wanted
to
talk
a
little
bit
about
home
net
and
IOT,
because
I
think
that
that
this
is
an
important
problem,
I
think
being
able
to
have.
If
you
have
a
home,
that's
got
a
sophisticated
network
in
it,
and
your
doing
things
like
operating
switches,
operating
lights,
using
switches
over
an
IOT
internet,
then
having
a
good,
solid
solution
for
how
you
make
that
work
in
a
home
that
is
larger
than
a
single
Wi-Fi.
Router
is
important.
D
So
and
there's
two
things
that
I
think
are
key
to
making
this
work.
One
of
them
is
making
it
making
sure
that
that
IOT
devices
are
reachable
by
the
by
the
other
devices
that
need
to
reach
them
and
by
no
other
devices
so
and
by
the
way
when
I
say
IOT,
there's
there's
two
different
kind
of
instances
of
IOT.
D
That
I'm
talking
about
one
is
things
like
we
Mo's
that
are
connected
to
your
Wi-Fi
or
you
know:
homekit
mostly
uses
Wi-Fi,
and
so
so
there's
you're
not
actually
using
constrained
network
and
the
devices
are
particularly
constrained
because
they
have
power
they're
they're.
Only
the
only
sense
in
which
they're
constrained
is
that
they
have
relatively
small
CPUs,
but
for
those
of
us
that
grew
up
programming,
Apple
twos,
they
actually
are
kind
of
big.
So.
D
Nevertheless,
like
you
know,
all
of
my
IOT
devices
are
currently
isolated
on
a
separate
link,
and
even
so
they
can
at
least
they
can
still
cross
infect
each
other
I'd,
really
like
it.
If
the
devices,
if
my
i2
devices
could
only
talk
to
the
things
that
I
want
them
to
talk
to
I,
think
that
there's
an
opportunity
to
do
some
work
there
and
I
think
that
would
be
good.
D
O
E
O
E
And
Elliot
Lear
and
Sarah
in
Si,
DN
and
mark
plowshare.
So
there's
there's
a
lot
of
us
working
on
this
and
the
goal
is
to
build
the
prototype
that
it's
based
on
mud.
So
we
have
a
func
semi-functional
mud
controller
where
we
can
provision
IOT
device
and
assign
access
control
to
just
control
where
it's
supposed
to
go.
D
D
D
P
I'm
Stuart
Cheshire
from
Apple
I'm
interested
hearing
about
this
I
had
not
heard
about
this
before.
That's
partly
why
we
come
to
these
meetings
to
learn
new
things.
Is
this
a
working
group?
Is
it
an
open
source
project,
so
the
documents
describing
this?
What,
where
can
I
go
to
learn
more
about
this.
D
K
D
So
so
that's
actually
kind
of
like
a
thing
that
some
of
the
IOT
vendors
are
trying
to
get
away
from
because
controllers
suck
you
wind
up
with
like
a
giant
pile
of
controllers
with
one
IOT
device
behind
each
one
of
them
and
that's
really
stupid
so
or
maybe
three
IOT
devices
behind
it.
So
using
you
know,
general-purpose
IP
networking
is
a
better
approach,
but
you
know
requires
more
interoperability
work,
so
so
yeah
I'm
talking
about
that
use
case,
the
the
the
actual
IP
networking
use
case
anyway,
just.
K
N
D
K
D
D
Well,
though,
the
homemade
architecture
actually
talks
about
this
issue.
I,
don't
know
if
the
document,
if
the
Charter
talks
about
it
specifically-
but
it
certainly
is
in
the
architecture,
so
right
I
mean
the
architecture
basically
says
you
know
there
will
be.
There
will
be
constraint,
networks
on
the
edge
and
we
need
to
solve
that.
We
need
to
make
that
work.
Okay,
okay,
yeah
anyway,
so
you
have
a
next
steps
thing
on
the
agenda.
I
also
put
down
my
next
steps.
This
is
just
my
pitch
about
what
I
what
I
want?
D
It's,
not
necessarily
what
home
that
should
do.
I
want
to
see
home
that
become
real.
We've
been
working
at
this
for
a
long
time.
We've
got
some
pretty
cool
stuff,
who'd
be
nice
to
actually
be
able
to
demo.
It
I
really,
don't
think
that
Linksys
our
hero
or
ubiquity
or
any
of
those
big
vendors
is
going
to
talk
this
anytime
soon,
although
if
there's
anybody
here
from
those
organizations
that
wants
to
talk,
please
talk
to
me,
but
we
can
definitely
make
this
work
in
open,
wrt
and
I.
D
Ik
makes
the
Taurus
omnia
and
that's
a
kind
of
a
high
feature,
full
router
that
could
also
be
doing
home
net.
That
would
be
really
cool.
So
if
we
could
get
some
grassroots
support
for
this
and
then
also
ISPs,
if
there
is
peas
that
are
interested
in
this,
we
could
potentially
create
a
groundswell
of
support
for
this
functionality.
That
would
then
result
in
it
actually
spreading
to
some
of
the
bigger
vendors,
and
so
that's
kind
of
what
I
would
like
to
see
happen.
K
Michael
Abraham's
owning
it
so
I
I
concur
with
that
and
just
to
say
that
there
are
commercial
vendors
that
are
shipping
home
routers
to
ISPs,
where
this
is
based
on
open
wrt.
So
doing
this
in
open
diversity
would
make
it
extremely
easy
for
them
to
include
that
package
as
well
and
to
do
this
ode,
I
mean
and
I
know:
I
mean
ubiquity.
If
you
look
under
the
hood,
that's
also
a
lot
of
open
source
on
under
there.
So
it
would
be
definitely
possible
for
them
to
take
that
as
well.
Right,
yeah.
D
Q
Q
We
go
to
very
cheap
Chinese
vendors,
which
have
the
same,
however,
as
more
expensive
ones,
because,
basically
it's
a
single
on
chip
and
that's
it
and
we
use
custom,
open
wrt
and
that's
it
and
I
know
by
fact
that
without
naming
specific
big
vendors,
like
some
of
those
that
you
have
in
the
in
the
slide,
but
some
others
as
well,
they
have
in
fact,
as
the
base
code
for
their
products
open
wrt
and
they
just
changed
the
face.
Okay,
so
I
think
having
that
in
open,
wrt
will
have
impact
in
the
market
and
I.
Q
F
Julia
scrubber
check,
you
know
what
I'm
going
to
say:
no,
no,
okay,
so
a
whole
net.
All
of
the
home.
That
stack,
I
mean
the
parts
of
the
home.
Let's
tag
that
exists
are
far
too
open,
wrt
yeah
and
they
work
yep,
Marcus,
Steven
and
Pierre
have
done
some
amazing
work
of
integration
of
home
that
into
open
wrt.
The
complaint
is
not
that
it's
not
an
open,
EE
aarti,
but
Java
is
not
enabled
by
default.
Exactly.
F
A
little
bit
of
tweaking
of
configuration
files
I
have
no
idea
whether
you
can
do
it
for
the
web
interface,
because
it's
simpler
to
do
or
the
configuration
files
now
I
happen
to
disagree
with
you
I
think.
That's
not
where
the
problem
lies
that
there
were
interest.
The
fact
that's
not
enabled
by
default
would
not
be
a
problem.
It's
only
a
problem
because
there
is
no
interest
in
home
that,
for
other
reasons,
if
actually
people
think
that
is
the
problem.
F
Well,
it
might
be
politically
different,
difficult
to
get
home
not
enabled
by
default
in
open
wrt,
but
it's
very
easy
to
fork
open
wrt
and
to
build
images
that
will
actually
enable
home
the
home
net
for
open
ability
for
home
net
by
default
in
those
images.
So
that
should
not
be
a
problem.
If
actually
there
is
the
motivation
to
get
the
work
done,
yeah.
D
So
so
Julius,
what
the
approach
that
I'm
suggesting
is
is
that
we
do
just
that
right,
because
that's
obvious
right.
It's
open
source!
We
can
do
whatever
we
want
with
it,
so
make
it
easy
for
for
a
certain
for
us
for
people
who
want
to
do
it
to
do
it.
But
then
the
other
bit
is
that
I
really
actually
would
like
to
pitch
it
to
some
of
the
some
of
the
vendors.
That
I
think
would
actually
get
market
benefit
from
doing
it
because
they're
small,
but
they
don't
have
the
advantage
of
owning
the
market,
but.
F
K
D
K
I
guess
a
profile
because
I,
wouldn't
even
you
I
mean
using
the
word
fork,
that's
a
good
point,
so
you
create
the
custom
feed
and
you
might
need
to
disable
some
stuff
that
it
is
default
on.
You
enable
some
other
stuff
and
that's
what
I
did
when
I
test
at
home.
That's
a
half
year
ago
or
whatever
it
was.
So
this
is
perfectly
possible
yeah.
So.
D
The
other
thing
to
say
about
what
Julia
said
is
that,
yes,
we
do
have
support
for
agency
P.
We
have
support
for
routing
it's
it's
pretty
easy
to
get
working.
We
don't
have
support
for
naming
that's
kind
of
the
missing
link.
So
so
that's
really
what
I'm
focused
on
I
don't
actually
see
myself
having
to
do
a
whole
bunch
of
work
on
H
n
CP.
D
There
are
a
couple
things
that
I
wanted
h:
n
CP,
that
it
doesn't
currently
provide,
but
it's
relatively
minor
and
I
think
they're
all
very
much
in
keeping
with
what
h
n
CP
is
quite
capable
of
doing
so.
I'm
not
too
worried
about
getting
that
working,
I'm,
hoping
that
I
can
get
some
help
from
the
H
n
CP
experts
on
it,
but
yeah.
D
So
that's
what
I
have
I'd
like
to
see
this
I'd
like
to
try
and
make
it
a
goal
that
we
actually
have
some
code
that
we
can
try
out
at
hackathon
in
Montreal,
if
anybody's
interested
in
participating
in
that
either
as
people
who
are
helping
me
to
get
that
done
or
as
people
who
just
want
to
bring
some
routers
to
Montreal
and
see
if
they
can
interoperate.
That
would
be
awesome
and
that's
all
I
got.
R
B
Thank
you
Ted
really
appreciate
that
we
kind
of
skipped
over
the
first
item
on
well
one
of
the
first
items
on
the
agenda,
which
is
just
to
get
an
update
on
some
of
the
expired
drafts,
and
maybe
what
thoughts
and
plans
are
with
those,
and
we
asked
if
Daniel
would
just
maybe
come
to
the
mic.
You
can
pick
your
choice
of
mic
and
tell
us
a
bit.
H
H
H
Another
draft
is
about
how
you
configure
this
mechanism
and
based
on
the
gscp
options
so
which
means,
in
our
case
it's
it
would
be
most
likely
the
ISP
providing
the
necessary
para
mirrors
that
could
be
used
then,
to
configure
the
outsourcing
our
architecture
I'm,
not
sure
this
draft
I
don't
know,
that's
a
it's
a
it's
a
way
to
configure
something:
I
mean
if
it's
never
going
to
be
implemented.
It
may
be
not
worth
doing
that
one.
H
H
A
Okay,
thank
you
so
I
guess
that
these
are.
Where
do
we
go
from
here
discussion
and
I
can't
remember:
Barbara
was
I
supposed
to
be
the
bad
cop
or
the
good
cop
I.
Think
I
was
the
bad
cop,
so
so
I
sent
mail
to
the
list
and
there
was
like
dribbles
of
response.
A
few
people
I
mean
a
few
people,
obviously
doing
good
work
here,
it's
as
chairs,
it's
absolutely
impossible
to
know
what
the
working
good
think
of
it
is.
Nobody
saying
anything.
A
So
that's
kind
of
puts
us
in
a
position
where
we're
not
gonna
proceed
to
produce
RFC's,
as
is
so
Ted,
is
suggesting
a
change,
slight
change
of
direction.
I
think
so
we
need
to
know
what
people
think
of
us,
or
if
people
aren't
willing
to
say
what
they
think
of
that
and
comment
on
the
list,
then
we
need
to
just
talk
to
our
aid,
our
new
esteemed
ad,
about
getting
out
getting
out
of
people's
way.
D
Lemon
so
I
have
a
theory
which
I
hope
is
true,
which
is
that
if
there
were
actually
like
hope
of
progress,
there
would
be
more
feedback.
I,
don't
know
if
that's
true,
but
it
would
be
nice
to
see
what
happens
over
the
course
of
the
next
few
months
if
I
actually
get
time
to
do
some
code
and
push
some
stuff
out.
Maybe
we
get
some
more
comment.
E
S
Agree
with
the
stuff
Ted
put
up
there
earlier,
particularly
with
respect
to
open
work.
One
thing
I
noticed
looking
at
the
the
working
groups.
Data
tracker
material
is
that
it
seems
the
Charter
is
unchanged
since
July
2011,
when
the
working
group
was
formed,
it
talks
about
forming
an
architecture
blah
blah
blah.
Maybe
this
is
really
a
reach.
Our
touring
discussion
as
well,
and
to
maybe
look
at
what
we
put
in
the
Charter
and
whether
we've
achieved
it
or
not,
might
be
useful
as.
A
Part
of
the
exercise,
so
just
as
a
clarifying
question,
given
that
I
asked
the
mailing
list
is
one
of
the
things
we
should
do.
You
know
shut
down
or
change
direction
and
got
bugger-all
response.
Then
do
you
think
we
reach?
Our
doing?
Discussion
is
likely
to
energize
people.
It
might
help
people
remember.
S
S
What
open
works
in
particular
and
then
hope
that
it
gains
traction
from
there,
but
yeah
it's
difficult
when
there's
so
little
impetus
and
I
think
it's
been
something
that's
sort
of
looking
more
recently
from
the
outside
of
the
working
group,
it
does
seem
to
have
stalled
for
the
last
couple
of
years
in
terms
of
producing
things.
It's
not,
obviously
not
your
fault.
You
can
only
work
with
what
you've
got,
but
at
the
moment
it
seems
pretty
pretty
sad,
really
there's
some
good
work
happening.
G
Stephen,
probably
Wilson
again,
I
think
for
me
at
least
the
clue
lies
in
the
two
sections
of
what
had
presented.
There
was
one
section
on
essentially
how
to
structure
the
architecture
if
people
have
got
multiple
routers
in
their
home,
and
then
there
was
another
section
about
functionality
for
routers
and
home
controllers
in
a
in
an
IOT
environment.
G
How
do
we
make
multiple
routers
work
in
a
single
house?
It
may
be
very
worthy,
but
it
feels
to
me
like
that's
not
what
the
market
is.
Gonna
drive.
What
the
market
is
more
likely
to
drive
is
you've
got
one
router
in
your
house,
and
you've
got
maybe
multiple
IOT
device
controllers
in
your
house.
That
seems
to
me
the
use
case.
That's
most
likely
to
try
for
the
market
and
vendor
base
yeah
I.
E
Shocked
late,
Oh
totally
agree
with
forgot
your
name
so
I
I'm
in
the
same
boat.
So
when
I
started,
this
secure
home
IOT
project
at
Sara,
I
looked
at
old
net
and
all
that-
and
this
addresses
the
use
case
of
maybe
point
zero.
One
percent
of
the
people
that
know
how
to
do
home
gateways
with
two
links
in
your
home
and
the
work
associated
with
that
is
not
really
relevant
with
what
people
are
asking
for
at
all.
E
So
when
we
did
our
quick
survey,
people
at
home
want
to
have
a
secure
home
network.
When
you
put
an
IOT
device
that
they
have
no
clue
what
it
does,
they
want
to
make
sure
that
it
only
does
what
it's
supposed
to
do
and
it
behaves
according
to
normal
roles.
So
what
we
started
to
look
at
is
because,
if
you're
gonna
be
adding
a
bunch
of
device
at
home,
you're
gonna
create
a
your
home
network.
Is
gonna,
be
like
an
enterprise
network
need
to
have
zones.
You
need
to
have
security
zones.
E
You
need
to
have
outbound
security
control
to
protect
your
traffic
going
out.
You
want
your
fridge
wrongly
to
talk
to
your
vendors
fridge
and
that's
where
mud
came
in
to
manufacturer
user
description,
so
certain
rules
need
to
be
applied
to
outbound
traffic
and
haven't
seen
any
work.
I
relate
related
to
that
and
this
is
a
gap.
We
need
to
build
a
secure
home
security
framework
to
define
how
home
network
actually
works
and
that's
what
we're
trying
to
implement
today
we're
defining
a
framework
for
that,
but
I
have
no
clue
where
the
works
fit.
E
We're
building
you
API
is
to
provision
to
provision
IOT
device
to
provision
the
router
itself
and
I,
don't
know
where
that
it
doesn't
fit
in
this
group,
because
this
is
more
focused
yeah.
That's
a
lot
yeah!
Yes,
but
there's
there's
a
gap
for
the
home
network
to
make
them
secure,
so
we're
working
ahead,
but
we're
writing
drafts
and
we'd
on
the
way.
If
it-
and
so
that's
a
challenge,
we
need
to
figure
out.
D
Actually
you
know,
if
you
don't
know
what
working
group
to
do
this
and
I
think
you
know
to
me
this
is
you
know,
I
we've
done
a
bunch
of
work,
I
think
we,
you
know
it
would
be
nice
to
finish
the
simple
eminent
naming
stuff
that's
been
going
on
a
while.
I
really
feel
like
implementation
needs
to
happen
before
that
gets
finished.
But
to
me
the
one
of
the
very
interesting
problems
that
we
face
is
how
do
we
do
you
know?
D
So
how
do
we
do
exactly
what
he
was
talking
about
and
that
actually
seems
to
me
to
fit
in
to
what
the
home
networking
group
ought
to
be
doing?
I,
don't
know
if
you
know,
if
you
guys
would
want
to
do
it
in
the
home
networking
group
if
we
made
a
place
for
you
to
do
it
here,
if
you
would
I
think
that
might
be
an
interesting
thing
to
pursue
it's.
It
is
a
bit
of
a
change
of
direction.
D
It's
consistent
with
what
we
started
out
doing,
but
it
would
probably
require
a
recharter
and
it's
it's
definitely
a
topic
that's
near
and
dear
to
my
heart
and
that
I've
been
interested
in
having
us
work
on
for
a
long
time
and
that
we
haven't
really
worked
on
because
there
hasn't
been
energy
in
the
working
group
to
work
on
it.
So
so
I
think
it
would
be
like
if
that
were
a
way
to
get
some
new
churn
happening
on
the
mailing
list
and
and
so
forth.
I
think
that
would
be
really
good.
A
F
Do
you
recycle
so
julia
scrubber,
track,
I,
think
I'm
agreeing
with
Robin
and
Jack
earlier
on,
but
a
little
bit
differently.
I
think
that
in
this
group
we
are
absolutely
hopeless
at
marketing
and
we
are
hopeless
at
marketing
for
two
reasons.
One
is
that
we
communicate
badly
so
I
don't
know.
If
you
remember
a
couple
of
weeks
ago,
somebody
said
that
the
problem
of
hope
net
on
the
mainland
lives.
F
That's
the
problem
of
hope
net
is
that
it
only
supports
ipv6
and
then
I
explained
that
no,
we
have
full
feature
parity
for
ipv6
and
ipv4,
and
people
were
surprised
that
we
actually
implement
everything
for
ipv4
and
that's
the
people
who
are
on
the
hamlet
mailing
list
right,
that's
frightening.
It
shows
you
how
badly
we've
managed
to
communicate
what
we
are
doing
and
I
think
there
is
a
more
fundamental
reason
to
us.
Communicate
in
badly
is
that
we
have
failed
to
find
out
to
have
a
clear
message
on
what
feature
we're
giving
to
people.
F
So
we
know
everyone
in
this
room.
Everyone
on
the
jabber
channel
knows
that
doing
things
that
layer-3
rather
than
switching,
is
good,
it's
good
in
itself.
It
doesn't
need
any
justification,
but
you
don't
actually
sell
things
to
people
by
telling
them
look.
It's
a
good
thing.
It's
the
light
side
of
the
force
you
actually
need
to
tell
them
look.
I
can
do
that,
which
you
cannot
do
in
your
home
network
right
now
and
we
have
not
been
concentrating
on
communicating
over
the
fact
that
we
have
actual
features.
F
Actually,
if
I
ask
you,
what
are
the
features
we're
giving
people?
How
do
you
sell
helmet
to
someone
whom
you
meet
in
the
metro
and
have
just
three
minutes
to
explain
it
to
I?
Think
many
people,
even
in
this
room,
would
have
to
think
about
it
for
a
while.
So
when
I
was
right
and
so
I
think
that
the
main
feature
good
support
for
shortest
paths,
routing
in
complex
topologies,
but
that's
not
the
way
you
explain
it
to
normal
people.
F
Most
people
when
they
install
Wi-Fi
at
home
are
very
surprised
to
discover
that
they
cannot
have
the
access
points
communicate
over
wireless
links.
They
expect
to
be
able
to
put
an
extra
access
point
in
their
kitchen
and
they're
very
surprised.
When
you
tell
them,
you
actually
will
need
to
have
to
wire
your
kitchen
over
Ethernet
home
that
has
perfect
support
for
topologies
in
which
wireless
links
are
used
on
are
used
for
transit.
K
K
K
K
So
now,
I'm
Michael
again
so
frankly,
the
whole
not
being
able
to
transport
layer
to
over
home
net
I
mean
I
would
never
expose
myself
and
my
family
to
her
to
my
two
access
points
being
in
two
different
broadcast
domains
within
trying
to
do
handover
between
them.
The
way
it's
currently
implemented,
I
mean
my
requirement
is
I
need
to
sit
on
a
some
kind
of
IP
real-time
communication
channel
and
walk
through
my
kitchen
to
get
tea
and
walk
back,
and
my
calls
should
stay
up
the
bay
the
homeowner
today
doesn't
provide
this
premise.
K
So
I
need
multi
need
a
broadcast
domain
from
a
two
Wi-Fi
access
points.
Now
we
have
this
IOT
thingy,
popping
up
and
and
so
on.
It
has
the
same
requirement.
I
I,
don't
think
this
can
succeed
without
us
being
able
to
provide
broad
Kasinga,
broad
couples
broadcast
domains
over
this
infrastructure.
I
mean
we're
not
gonna
win
that
war.
K
We
want
to
have
incremental
deployment
towards
halted,
domains
and
doing
a
service
discovery
between
rotting
domains
dissolved,
but
saying
now
that,
in
order
to
use
home
net
need
to
be
able
to
do
seamless,
you
know
Island
PE
or
something
like
that
handover
between
domains.
The
host
hunter
I
mean
this
is
a
miserable
bad
user
experience.
We
can't
drive
this.
We
need
to
be
able
to
do
it
to
be
part
of
an
incremental
deployment
in
doing
this.
So
I
think
that
is
one
of
the
major
problems,
so
I
I
actually
tried.
K
This
I
mean
I
I
spent
I
actually
set
up
homeland
with
three
routers
in
my
home
and
tried
to
use
it,
and
then
I
searched
it
off
again
because
I
doing
this
for
Wi-Fi,
so
yes
network
and
the
current
implementation,
it
needs
more
knobs
I.
Try
to
use
it
because
I
had
to
I
to
Jules
back
links
that
were
pretty
good
at
home.
K
I
couldn't
use
some
of
it
because
it
didn't
allow
me
to
tune
it,
so
it
shows
he
prioritized
my
host
to
use
one
of
them
and
the
second
one
for
fall
back,
because
the
one
that
II
didn't
wanna
use
had
one
week,
HP
v6
prefix
delegation
time
and
there
was
a
lifetime
of
there
in
the
RAS
and
that's
what
the
host
shows
and
I
didn't
want
them
to
prioritize
that
and
and
I
mean,
if
I,
unplug
or
reboot,
the
router
in
the
current
implementation
of
home
NIT
next
time,
I
plug
in
the
roadie
the
link
it
has
new
addresses
if
I
reboot
it
it
held
it
all.
K
A
D
D
Think
there's
there's
a
there's
a
nugget
in
in
what
Mikkel
just
said,
though,
which
is
I,
don't
remember,
actually
doing
any
debugging
of
this
problem
on
the
mailing
list
and
if
we're
gonna
actually
have
a
functioning
working
group.
Part
of
what
we
need
to
be
doing
is
actually
just
that,
like,
if
you're
having
these
kinds
of
problems,
we
should
be
talking
about
them
on
the
mailing
list.
I'm
gonna,
let
Julius.
F
So
concerning
the
roaming
issue,
you're
extremely
right:
we
needed
to
have
roaming
without
remember
in
between
access
points
and
we're.
I
don't
agree
with.
You
is
that
that
requires
a
single
broadcast
domain
and
the
proof
that
you
don't
need
a
single
broadcast
domain
is
4G
in
4G.
You
have
seamless
roaming
with
no
remembering
and
you
don't
have
a
single
broadcast
domain
and
what
they
use
is
a
complex
infrastructure
based
on
the
whole
series
of
acronyms
I.
Don't
know
if
you
try
to
read
any
4G
standards.
We
at
ITF
are
actually
good.
F
We
don't
use
a
lot
of
acronyms
okay,
oh
they
use
numbers
too.
They
will
use
cheese,
50
59
and
s
43
and
try
to
follow
the
standards,
but
they
do
actually
use
some
of
our
protocol.
They
use
pmipv6
proxy
mobile
ipv6,
so
there's
a
whole
infrastructure,
but
it
can
be
done
now.
Question
is,
can
it
be
done
in
a
simple
distributed
way
with
no
infrastructure?
And
the
answer
is
yes,
but
it's
tricky.
It
took
me
two
years
to
get
the
implementation
done
and
it
has
been
running
for
the
last
two
weeks.
F
K
G
K
A
D
Ted
lemon
again,
I
think
I'll
just
point
out
that
that
this
discussion
I
just
heard
a
bunch
of
things
about
work,
that's
being
done
at
home
net
and
problems
that
are
being
had
at
home
net
that
never
made
it
to
the
mailing
list
and
I
think
that
may
be
part
of
our
marketing
problem,
because
so
and
and
I
don't
claim
to
be
immune
to
this
I've
been
thinking
about
this
exact
same
problem
too.
How
do
we
make
this
work?
Because
you
know
after
the
presentation
in
Bangkok
I
was
like?
D
Well,
ok,
do
we
do
we
actually
have
a
problem
that
we
need
to
solve
here?
I
think
we
do
and
if
we
do
have
a
problem
we
need
to
solve
here.
What
are
the
things
that
we're
not
doing
that
we
need
to
do,
and
one
of
them
is
this
exact
thing
that
Julius
was
just
talking
about
whether
we
do
it
with
a
flat
layer
or
a
or
a
routed
layer.
Three,
we
need
to
address
that
particular
use
case.
D
The
the
my
phone
should
continue
to
work
as
I
walk
from
my
house
into
my
garage
and
switched
to
a
different
AP.
So
I
think
that
we
have
interesting
work
to
do
and
part
of
the
problem
that
we
have
is
that
we
haven't
actually
been
talking
about
it
on
the
mailing
list,
and
so
I
would
suggest
that
part
of
the
way
forward
for
the
working
group
is
to
actually
you
know
this
actually
kind
of
goes
back
to
I.
Think
it
was
Tim
was
saying
we
need
to
reach,
hurt
her.
D
My
initial
reaction
to
that
was,
oh
god,
I
hate
reach
are
during
that
sucks,
but
but
actually
I
think
he's
right.
I
think
that
we
should
have
a
discussion
about
chartering,
because
it
may
be
that
part
of
the
reason
we're
not
discussing
these
things
on
the
mailing
list
is
because
we
think
they're
out
of
scope
and
they're,
not
out
of
scope.
So
maybe
the
way
forward
is
actually
to
open
a
reach
are
during
discussion.
D
A
K
Michael
Abramson,
and
so
this
was
discussed,
but
it
was
discussed
in
like
2016
or
something,
and
then
it
was
then
it
hasn't
been
discussed
since
but
yeah.
We
should
resurrect
these
discussions
and
figure
out
where
the
where
the
gaps
are
and
what
you
need
to
do
when
actually
making
this
useful
and
yeah.
We
need
configuration
knobs
and
we
need
to
fix
some
things.
That
really
is
not
working,
so
sounded
good.
What
Julis
described
and
all
the
gladly
testa.
G
Robin
will
turn
again
just
one
quick
one
to
pick
up
on.
Do
races,
point
about
communication,
I
think.
On
the
one
hand,
if
a
working
group
exists
to
develop
specs,
then
it's
developing
things
that
aren't
immediately
useful
as
marketing
documents
and
that's
normal.
On
the
other
hand,
the
problem
that
we
seem
to
be
faced
with
is
we've
got
vendors,
who
are,
as
usual,
creating
islands
of
automation
and
it's
in
their
interest
to
create
walled,
gardens
and
we're
doing
a
piece
of
work
aimed
at
giving
them
something.
G
That's
not
a
walled
garden
that
we
would
like
them
to
do
instead,
and
that
doesn't
seem
to
me
to
be
a
new
problem
for
this
working
group.
It
seems
to
me
to
be
a
generic
problem
about
open
standards,
so
I
can't
I
can't
believe
that
issue
hasn't
been
addressed
and
fix
elsewhere.
Maybe
what
we
should
be
doing
is
looking
at
how,
in
general,
open
standards
enter
the
market
and
seeing,
if
there's
anything,
that
we
could
be
doing
that
we're
not
doing.
D
Ted
lemon
yeah,
so
I
totally
agree
with
that.
I
think
that.
So,
if
you
look
at,
if
you
go
to
home
net
dot
or
sorry,
what
is
it
home
that
doc
work
right?
Now,
it's
actually
a
blank
page
like
it
doesn't.
Even
it
doesn't
even
produce
you
don't
get
a
web
page.
We
could
probably
do
something
about
that,
but
yeah
I
think
there.
You
know
another
thing
that
that
that
I
think
would
be
interesting.
To
think
about
is
there's
the
in
in
the
t2
TRG
meeting
that
we
had
on
Friday.
D
There
was
discussion
about
you
know:
do
we
want
to
be
able
to
do
I
can't
remember
what
the
acronym
is.
It
was
I
think
a
four-letter
acronym,
but
basically
the
idea
is:
is
the
ability
to
to
store
data
in
the
infrastructure
or
to
have
work
done
in
the
infrastructure,
and
you
know
for
relatively
confident
home
net
routers.
The
ability
to
do
work
in
the
infrastructure
would
actually
be
an
enabling
technology
for
things
that
that
end
users
might
real
want
to
be
able
to
do,
but
currently
can't
do
right
now
as
an
end-user.
D
If
you
want
to
have
a
server,
you
pretty
much
have
to
go
to
some
kind
of
cloud
right
like
or
you
have
to
set
up
a
Mac
or
something
like
that
on
your
home
network
and
people.
Don't
do
that
and-
and
it's
not
automatically,
there's
no
way
that
I
could
pitch
to
an
end
user
who
doesn't
already
believe
in
having
a
server
at
home.
Why
they
should
spend
a
thousand
dollars
to
set
up
a
server
at
home
so
that
they
can
do
chat
with
their
friends
or
something
like
that.
D
So
there's
a
whole
bunch
of
stuff
there's
a
whole
bunch
of
services
that
aren't
being
enabled
by
our
current
stop
by
our
current
infrastructure.
That
could
be
enabled
by
something
like
a
home
net.
Where
we
don't
have.
We
don't
need
a
lot
of
computes,
but
we
need
a
rendezvous
point
so
that
so
that
we
can
communicate
and
that's
actually
something
that
I've
been
wanting
to
do.
D
I
remember
proposing
this
at
at
a
now
defunct
company
as
a
product
like
five
or
six
years
ago,
and
it's
something
that
home
that
could,
in
principle,
work
on
and
and
I
think
it
would
be
valuable
to
work
on.
So
you
know
there's
a
lot
of
interesting
stuff.
We
could
do
if
there's
interest
in
doing
it.
People
keep
showing
up
for
the
working
group,
so
there
must
be
some
reason
why
everybody
keeps
showing
up.
K
A
K
I'd
be
really
interested
in
opinions
in
that
space,
yep
Michael
Abraham's.
Is
it
so
I
think
that
this
is
an
interesting
question
about
the
small
and
medium
businesses
use
case
as
well?
I
mean
yeah.
I
can
also
see
people
with
they
have
like
of
summer
house,
and
you
know
of
normal
house
sure
they
have
I.
You.
A
A
K
A
A
K
A
L
D
Ted
lemon,
so
I
actually
thought
about
this
particular
question
that
you
asked
a
bit
the
other
day
and
the
conclusion
that
I
came
to
is
that
if
we
continue
the
way
we're
going,
I
think
that
there's
interesting
work
in
there
are
people
who
are
interested
in
it.
But,
on
the
other
hand,
we're
consuming
a
bunch
of
time
at
the
IETF
and
so
one
additional
alternative.
You
propose
three
alternatives.
D
I
would
say,
there's
a
fourth
alternative,
which
is
continue
to
have
the
working
group
open,
but
just
accept
that
it's
a
long-term
thing
and
we
don't
need
to
meet
at
every
IETF
I,
don't
think
I'm
not
proposing
that
we
not
meet
at
Montreal,
because
I
think
that
we
have
reason
to
believe
that
there
might
be
some
things
to
talk
about
in
Montreal.
But
and-
and
you
know,
obviously,
you
might
prefer,
as
working
group
chair,
not
to
have
this
responsibility
if
nothing's
happening.
D
But
but
it
has
been
the
case
in
the
past
that
working
groups
have
kind
of
submarine
for
a
while
I
would
I
would
hate
to
lose
this
community.
If,
if
there
are
people
who
are
interested
in
this,
but
but
you
know
now
isn't
the
time
for
them
to
be
hacking
on
it,
fine
I'd
still
like
to
be
able
to
have
a
place
where
we
can
do
this
work.
D
Even
if
it's,
even
if
it
winds
up
going
very
slowly
so
so
I
personally
I
think
retiring
is,
is
something
we
should
try,
because
it
seems
like
there's
some
energy
to
do
some
stuff.
That's
not
really
clearly
in
the
Charter
and
if
we
recharter
that
energy
might
wind
up
landing
in
the
working
group,
which
would
be
good,
I,
think
we've
got
a
bunch
of
stuff
on
the
agenda
to
get
done
for
Montreal
and
if
that
stuff
gets
done,
then
that
might
create
some
new
energy.
D
So
that's
the
thing
to
try
and
then
you
know
if
we
find
that
we
don't
have
the
energy
that
we
need
to
continue
being
a
working
group
that
meets
every
IETF.
I,
still
think
that
it's
worth
having
the
having
the
working
group
around
in
some
form,
because
clearly
there
is
interest
in
this
topic
in
the
long
term,
so
sure.
A
G
G
Vendors
aren't
generating
demand
for
this
spec
yet,
but
the
way
to
convince
vendors
to
do
that
is
and
I
think
this
comes
back
to
Julian's
point
about
communication.
Again
the
way
to
convince
vendors
to
do
that
is
not
only
to
write
specs
it's
to
write,
specs
and
do
other
stuff.
As
well
and
I,
don't
know
whether
that
other
stuff
is
the
legitimate
work
of
this
working
group
or
not.
A
G
A
We
just
established
so
again,
I
think
we
established
there's
a
baseline
here
that
nobody's
gonna
suddenly
get
active.
If
we
don't
change
something
we
can
go
quiescent,
we
can
pause
a
while
wait
for
the
implementation
work
to
continue
and
reevaluate.
You
can
start
retiring
now.
So
those
are
the
kind
of
options
that
we're
trying
to
get
feedback
on
specifically
right.
G
But
so,
okay,
so
I'm
speaking
at
individual
capacity,
I'm,
not
rec,
not
representing
your
particular
organization.
But
if
an
organization
like
mine
were
working
on,
for
example,
things
that
increase
trust
in
the
infrastructure
for
IOT.
This
might
be
the
kind
of
thing
that
they
might
want
to
do
or
support
or
enable
or
facilitate
okay
and.
S
S
A
S
S
So
if
we
did
go
back
and
look
at
you
know,
Thomas
is
five
points
is
that
we
need
a
clear
problem
that
needs
solving
and
that
the
IHF
is
the
right
place
to
do
it.
Second,
critical
mass
participants
willing
to
work
on
it,
which
I
think's
the
point
you're
clearly
making.
Thirdly,
the
scope
of
the
problem
is
well
defined
and
understood,
I'm,
not
sure
at
the
moment
it
is
forth.
S
There
is
agreement
that
the
specific
deliverables
are
the
right
set,
I'm,
not
sure
what
our
plan
deliverables
even
are
at
the
moment
and
fifth,
it's
believed
that
the
working
group
has
a
reasonable
probability
of
success,
but
it's
very
hard
to
define
success
when
you
don't
know
necessarily
what
the
problem
that
you're
working
on
is
I.
Think
if
we
viewed
it
is
trying
to
answer
those
five
questions,
maybe
or
particularly
the
the
first
one
defining
the
problem
statement.
Maybe
that
is
about
recharging,
maybe
that
is.
S
B
B
S
Think
it
can
possibly
be
helpful
that
we
have
an
eight-year-old
charter
I.
Think
part
of
the
reason
not.
However,
here
I
think
Julia
said
that
people
think
this
is
a
v6.
Only
group
is
because
the
Charter
is
written
eight
years
ago,
yeah
when
point
zero,
something
percent
of
the
internet
on
the
of
the
Internet
traffic
was
v6.
It
was
written
with
a
sort
of
an
optimistic
v6
will
make
home
networking
better
perspective,
and
obviously
we
could
have
a
long
discussion
about
that.
S
H
Dan
so
I
think
there
are
two
issues,
one
which
is
the
adoption
by
the
industry
which
we
produced
and
the
one
will
we
do.
We
have
enough
energy
in
the
working
group
to
provide
some
feedbacks
even
and
we
can
provide
documents
without
knowing
whether
it's
going
to
be
adopted
or
not.
So
what
is
important
is
that
I
think
that
that
level
is
that
we
don't
lose
all
the
things
we've
done.
I
see
in
the
at
least
in
there
I
think
we're
making
progress
in
in
the
way
we
implement
things.
H
Don't
think
recharging
will
be
helpful
because
we
will
end
up
is
ok.
We
have
the
launch
sending
draft
to
and
they're
like
a
head
of
looking
in
TCP.
So
these
are
the
one
and
maybe
you're
gonna
have
some
others,
but
we
will
need
the
one
that
are
here.
First,
so
I,
don't
think
the
recharge
I
think
recharging
discussion
will
lose
some
of
the
energy,
but
rather
than
being
constructive.
H
B
E
B
B
Okay,
so
my
summary
is,
there
is
desire
to
start
a
reach
are
during
discussion
where
it
will
go,
nobody
knows,
but
will
try
to
direct
it
and
channel
it
so
that
it
has
the
potential
to
become
something.
Successful.
There's
also
does
not
seem
to
be
a
strong
desire
for
I
guess
the
chairs
to
do
sort
of
more
setting
light
to
fires,
and
things
like
that.
Necessarily
it's
kind
of
people
need
to
get
their
work
done.
B
Maybe
there's
not
a
meeting
next
time,
we'll
keep
list
open,
even
if
we
don't
successfully
recharter
will
keep
things
open.
We
may
or
may
not
have
a
meeting
in
Montreal,
but
in
any
case
the
list
will
be
there
and
it
sounds
like
there
may
be
hackathon
progress
now.
I
do
have
a
question
given
that
there's
been
experimentation
with
sort
of
these
open
sessions.
D
So
Ted
lemon
you've
probably
heard
me
Grich
about
the
tendency
in
IETF
for
us
to
all
just
get
up
and
do
presentations
and
a
few
questions
get
asked,
and
then
we
all
wander
off
and
no
actual
work
gets
done.
I
think.
If
we're
gonna
have
sessions,
it
might
be
cool
to
have
sessions
where
we
actually
do
just
what
you
said
hack,
because
really
the
work
that
we're
doing
is
development
work.
D
We're
not
like
the
the
documenting
of
what
we've
developed
is
important,
but
you
know,
as
I
discovered
when
I
went
through
and
thought
about
what
I
needed
to
do
to
implement
the
HomeNet
architecture.
That
is
the
first
thing
we
have
to
do
like
and
all
of
the
stuff
that
I've
been
doing
the
last
year
on
on
home
net
and
on
DNS
SD
has
been
implementation.
Focused,
it's
been
very
effective.
It's
a
great
way
to
generate
energy.
You
know
we
have
people
show
up
for
hackathon.
D
If
there
are
people
who
are
interested
in
working
on
this
stuff,
it
will
be
awfully
nice
if
we
could
just
get
together
in
a
room
and
work
on
it
together,
and
you
know
and
share
the
information
that
we
have
like
I
mean,
for
example,
you
know
not
to
put
you
on
a
spot
Julius,
but
you
have
a
great
deal
of
knowledge
about
H,
n
CP.
That
I
don't
have,
and
you
know
also
you
have
a
great
deal
of
knowledge
about
Babel
I
mean
I.
D
Think
Babel
is
fairly
easy
to
set
up
so,
but
you
know
I
I
think
it's
still.
It
would
be
useful
for
us
all
to
be
in
a
room
whether
it's
during
the
hackathon
or
whether
it's
a
session
in
IETF,
like
you're,
describing
where
we
just
get
a
room,
that
the
people
who
actually
want
to
hack
can
be
in
I.
Think
that
would
be
great
I.
D
B
A
S
A
S
Question
is
how
we
engage
with
projects
or
other
communities
that
we
think
bring
useful
things
to
this
working
group.
So,
for
example,
we
heard
of
one
today
I've
not
heard
of
your
project
until
today
and
when
I
certainly
be
looking
at
it.
How
do
we
actually
find
the
others?
How
do
we
attract
them
to
come
here?
Maybe
part
of
that
is
defining
a
clear
problem
statement
and
advertising
it.
The
other
thing
that's
different
from
where
we
charted
in
2011
now,
is
that
we
have
in
theory
a
lot
more
sophisticated
home
networks.
S
We
have
25-30
percent
of
the
home
networks
on
the
planet,
now
running
v6
and
so
on,
sat
quite
close
to
me,
who
has
five
million
customers
I
think
with
the
v6
in
their
homes,
a
quite
a
few
anyway,
you
know
and
they're
deploying
smart
devices
in
the
homes.
How
do
we
get
their
input
here
and
comment
and
input
from
those?
What
is
it
that
we
can
do
that
might
make
it
attractive
for
them
to
actually
use
this
stuff,
where
at
the
moment
they
don't.
B
K
Okay,
so
Dave
that
they've
tat
wants
to
say
that
he
doesn't
like
hackathons
and
he
would
like
stuff
to
be
done
continuously
over
time.
Every
week
get
builds
done.
Do
continuous
integration
gather
keyway
on
testers,
iterate
iterate
iterate,
so
he
has
a
point
of
having
an
IRC
channel
or
some
kind
of
forum
to
continue
the
do
stuff
in
instead
of
saying.
Oh,
let's
have
a
hackathon
every
three
months
it.
There
is
a
point
to
that.
D
D
Well,
it
sounds
like
I
mean
I,
think
you,
if
you
want
to
do
some
prodding
I,
don't
think
that
would
be
a
bad
thing.
It
sounds
like
Dave's
got
some
energy
to
actually
try
to
to
create
momentum,
and
so,
if
Dave's
willing
to
drive
that,
then
that
kind
of
takes
you
off
the
hook.
I
think
somebody
needs
to
drive
it.
Otherwise,
you
know,
realistically
it's
not
going
to
happen.
So
if
that's
Dave,
that's
great,
if
that's
you
that's
great.
D
This
is
more
on
my
agenda
now
than
it
was
prior
to
this
ITF,
because,
prior
to
this
IETF
I
was
mainly
doing
DNS
SD
implementation
work.
That's
mostly
done
at
this
point,
so
actually
very
much
on
my
agenda,
for
this
break
is
to
do
work
on
open
wrt.
So
there
should
be
some
self-sustaining
momentum
on
my
part
with
that
as
well.
But
yeah,
you
know,
if
you
don't
see
us
talking
about
it
on
the
mailing
list,
I
think
I
think
it's
entirely
appropriate
for
you
to
say
what
the
hell
are.
D
You
guys
doing-
and
why
aren't
you
talking
about
this
on
the
mailing
list,
because
otherwise
you
know
we
definitely
have
a
pattern
of
not
talking
about
this
on
the
mailing
list
and
then
not
talking
about
it
on
the
mailing
list
means
that
those
people
who
are
on
the
mailing
lists
aren't
seeing
the
the
work
being
done
and
they
aren't
they
aren't
receiving
the
sort
of
energy
that
that
is
that
is
going
on
in
the
background.
So.
U
U
Yes,
absolutely
I
do
I
think
the
working
group
does
need
a
air
quotes,
project
manager
of
sorts
to
keep
people
along
whether
that's
you,
the
working
group
chairs
or
you
wish
to
assign
somebody
to
take
on
that
role
to
be
the
ringmaster
to
to
use
that
whip.
Then
I'll
leave
that
to
to
yours
in
the
working
groups,
discretion
for
another
thirty
hours,
I
think
that's
that
the
only
direction
I
have
for
you
I,
think
I
think
this
means
some
fantastic
work
and
done
in
honet.
U
A
Okay,
so
there's
some
challenge
in
the
jabber
room,
but
mechanics
of
care
discussion
with
you
use
the
list
and
or
from
github
repo
I,
think
I
think
we're
gonna
suggest
not
trying
to
dive
into
that
charity,
especially
right,
but
brother.
Take
it
to
the
list
because
I
think
part
of
the
test
is.
Does
it
happen
on
the
list?
B
K
It's
all
what
you
used
to
I
can
do
either,
but
I
mean
if
I
have
to
choose.
We.