►
From YouTube: IETF105-ROLL-20190724-1000
Description
ROLL meeting session at IETF105
2019/07/24 1000
https://datatracker.ietf.org/meeting/105/proceedings/
A
B
C
B
A
A
Okay,
this
is
the
MIDI
material.
Thank
you
again
to
Dominique
and
Michael
for
healthy
nuts.
Please
sign
the
blue
sheets.
Okay,
this
is
our
agenda.
Do
we
have
comments?
We
are
going
to
have
Rao
remotely,
but
he
is
not
going
to
be
able
to
use
video,
so
we
are
going
to
have
it
only
in
audio
mode,
but
he
is
going
to
be
online.
A
E
C
A
E
Especially
for
the
young
model,
I'm
still
looking
forward
to
some
people
who
may
comment
on
that,
and
once
that
is
done,
we
can
put
it
forward
because
not
really
uncompressed
document
and
reading.
It
is
not
too
complex
to
take
you
in
quarter
of
an
hour,
you
point
out
to
most
mistakes,
and
that
will
be
it.
The
young
model
of
nipple
MPL.
A
A
A
C
So
the
the
recent
changes
are
related
to
our
the
encoding,
the
RFC
eight
ones
create
encoding
and
things
like
that
to
to
elf
implementation.
So
we
give
examples,
and
there
was
discovered
that
we
got
that
when
we
want
to
ever
to
tell
the
route
that
there
is
an
error
on
projected
route,
then,
instead
of
use
from
the
packet
as
it
stood
to
differentiate
a
source
route
packet
which
no
more
resources
run
packet
from
a
project
it's
or
a
packet.
So
because
there
is
no
state
from
the
projected
route.
C
In
the
intermediate
note
in
the
source
route
model,
remember
we
started
with
norms
with
storing
mode
and
we
added
non
story
mode
projected
routes
in
a
non
storing
mode
projected
route.
You
have
a
source
routing
it
now
you
follow
it,
but
you
have
no
state
in
ultimate,
not
note.
So
when
you
see
the
packet
Oh
should
I
report
to
the
soul.
Should
I
report
to
the
route,
or
is
there
any
other
case
right
now?
It
would
be
the
route
the
route
anyway.
I
can
ask
that
in
the
future
support
of
so
swatting.
C
It
could
be
anything
so
we
decided
about
the
mailing
list
that,
having
a
flag
to
say,
a
it's,
this
type
of
route,
so
the
euro
message
should
be.
This
was
was
interesting,
and
so
what
we
did
with
the
last
version
is
add
a
flag
in
the
API,
the
hop-by-hop
option
and
what
that
was
fine.
But
that
means
changing
dating
RFC.
C
So
we
went
through
the
exercise
of
actually
defining
a
new
format
for
our
state,
one
three:
eight
for
compressing
the
API
when
it's
being
used
in
the
context
of
projected
route.
So
it's
actually
smaller
right
and
it
can
contain
this
flag
that
that
we're
about
so
that's
the
the
the
biggest
change
we
made.
So
here
is
the
flag
in
the
non
compressed
format.
Basically
extending
our
see
six
five,
five
three,
they
see
my
call
to
my
question
right
now,
my
college,
just
my
good.
F
F
C
It's
it's
going
to
be
the
wrong
gear
of
code,
but
you
still
will
go
to
the
root
and
you
see
will
get
an
error,
meaning
that
it's
okay,
if
it
ignores
and
the
way
we
built
the
elective
and
and
monetary
compression
formats
makes
it
so
that
an
old
implementation
will
actually
let
any
API
through
which
will
just
ignore
it.
So
we
are
backward
compatible,
that's
kind
of
cool.
C
C
Well,
first,
the
format
is
like
this:
we
pick
the
same
type.
If
you
looked
at
the
way
we
did
at
one
three
eight.
We
actually
avoided
overlapping
the
types
in
elective
and
mandatory,
meaning
that
the
number
five,
which
is
the
RPI
in
modern
critical,
can
be
used
also
in
elective
or
also
a
RPI
right,
so
not
type.
Five
in
elective
and
critical
will
we'll
both
be
RPI,
which
is
kind
of
cool.
So
it's
pretty
much
why
we
did
not
match
at
the
time,
because
we
thought
that
maybe
this
would
happen.
It
actually
does
now.
C
The
formats
of
doing
of
the
packets
being
sauce
rot
from
a
source
which
is
not
the
route
to
a
destination
which
is
not
the
route
either
was
never
used
before,
because
for
now
till
now
the
source
routing
was
always
coming
from
the
route.
So
there
is
an
optimization
in
the
IP
to
IP
when
the
route
is
the
source
of
the
destination.
In
this
case,
it's
neither
but
eight
one.
Three
eight
was
sought
for
it,
even
if
we
never
used
it
before
so.
C
G
C
Are
you
saying
that
the
compression
just
is
it
is
because
you,
the
path
that
the
storing
mode
injects,
doesn't
have
to
be
a
longer
tier
that
it
can
be
any
path
right,
I
mean.
Usually
we
give
example
where
the
storing
path
is
a
compression
of
a
non
storing
path.
Just
to
reduce
the
number
of
countries
in
the
rotting
either
doesn't
have
to
be
so
if
the
route
is
capable
of
injecting
a
non
storing
path
in
the
network,
it's
also
capable
of
injecting
a
story
path
in
the
network.
C
G
C
C
C
C
So
here
is
the
format
actually
that
we
end
up
with
if
you
have
a
23-8
in
place
between
a
source
and
destination.
So
what's
interesting,
you
know
8
1
3/8
is
supposed
to
be
equivalent
to
ipv6,
it's
just
another
encoding
of
the
same
thing.
Now
the
weights
encoded
is
usually
that
you
place
all
the
destination
information
in
the
front.
C
So
if
you
have
one
destination
is
the
first
thing
that
you
have
if
you
are
watching
a
Dominic,
that
you
have
a
series
of
destinations,
their
whole
place
in
the
front
one
after
the
other,
which
is
very
different
from
normal
way
of
ipv6,
where
you
have
the
first
destination
in
the
IP
header
and
then
the
saucer
out
there
somewhere
else,
and
you
have
to
do
swapping
games
etc.
All
this
is
gone
when
you
use
the
eight
one,
three
eight
format,
so
you
start
with
the
routing
at
the
destinations.
C
Usually
you
have
either
the
destinations,
which
means
you
come
from
the
route
and
you
go
down
or
or
you
have
the
junk
apps
alator,
which
then
you
go
up
and
the
encapsulation
is
not
the
route.
But
if
junk
of
territory
is
a
route,
it's
elided,
when
these
guys
you
have
both
the
destination
of
the
IP
and
IP
and
the
source,
because
neither
neither
is
delighted
right.
So
it's
it's.
C
It's
completely
compatible
with
a
23-8
issues
that
maybe
people
never
coded
for
that,
because
if
you
did
the
minimum
code
for
exactly
what
you
need
it
for
classical
repo,
where
the
sauce
rotator
is
always
coming
from
the
root,
then
you
never
had
an
encapsulated
or
that
they
have
in
blue
here
in
your
packets.
Now
you
do
because
the
source
is
not
the
root,
so
we
still
have
a
number
of
discussions.
That's
where
I'm
I'm
wondering
what's
the
well
now
is
the
time
for
Walker
plastic
or
not.
C
I
really
want
to
close
this
document
that
we
wanted
to
finish
it.
We
want
to
implement
it,
but
but
there's
one
big
question:
is
we
never
really
soft
is
how
does
the
root
know
about
the
topology?
What
do
we
do?
We
said?
Oh
it's
not
this
back
right,
but
we
don't
have
nobody
started
the
other
spec
which
says
how
that
happens
so
right
now
we
know
how
to
project
a
lot.
C
We
don't
know
how
to
compute
it
and
I
I
mean
either
we
start
a
draft
and
then
we're
in
for
two
three
years,
or
we
just
add
an
option
in
that
one
so
that
the
root
knows
so
maybe
this
one
could
be
the
vessel
for
if
we
can
find
a
minimal
way
to
provide
topology
information
to
the
root,
maybe
we
could
package
that
option
into
this
document.
That's
the
question
I'm
asking
to
the
group
here
agree
with
that
you're
greedy.
It
was
basically
what
we
have
today
right
in.
C
We
have
story
mode,
sorry,
non
storing
mode,
a
non
storing
mode.
The
root
has
most
of
the
information
already.
It
has
all
the
do
that
right,
but
if
you
want
to
do
a
route
east-west
well,
maybe
it's
gonna
be
not
to
optimize,
because
you
only
see
the
do
dark,
so
we
only
see
links
in
the
do
dark,
the
parent
to
child.
What
you
would
like
to
see
is
a
bit
more
sibling
rot
sibling.
So
so
what
you
really
need
is
something
like
the
target
option.
Well,
you
have
the
target
option.
C
That's
me,
and
then
you
have
the
transit
option,
which
is
my
parents
and
I
have
a
list
of
transit
option
so
so
I
tell
the
routes
here
is
here.
Are
my
parents,
but
I
would
like
to
tell
the
route
is
here?
Are
my
siblings
right?
So
we
could
define
a
sibling
option
in
like
30
lines
of
text
in
this
pack
or
we
try
all
new
RC,
which
will
take
two
years
if
you're
inclined
of
doing
that.
If
the
group
says
I,
do
it,
let's
just
propose
an
option.
Put
it
in
that
stack
proposed
the
option.
C
C
A
A
A
I
C
Second
thing
is:
we
still
need
more
capabilities,
so
the
node
can
express
that
you
can
do
either
of
those
things
so
the
for
instance,
how
much
memory
do
I
have
how
many
storing,
not
storing
rods,
can
I
store
in
my
node
right.
So
the
root
does
not
overload
I
note
with
too
many
routes.
We
don't
have
right
now,
any
way
of
doing
it.
The
logical
way
is
to
provide
a
map
using
Howells
draft.
The
only
problem
with
that
is
that
round
draft
is
not
yet
an
RFC,
and
so
we
will
have
a
dependency
on
it.
C
So
what
I
want
to
do
is
use.
Ronald's
raft
refer
to
it,
and
so
I
will
ask
you
to
progress
files
draft
as
soon
as
you
can
right.
That's
pretty
much
what
goes
with
that,
because
we
wit
will
have
to
hold
this
one
this
one
or
if
you
have
a
better
way,
you
can
also
put
some
options
in
there,
which
would
be
very
specific
to
this
and
not
choose
Ross
draft.
That's
y'all
tonight,
I
think.
C
Yes,
then,
the
draft
goes
through
something
very
complex,
which
is
what
happens
if
you
want
to
suppose
to
have
a
full
path
that
is
completely
established
in
one
shot.
How
do
you
stitch
path?
Well,
that's
still,
okay,
but
how
do
you
find
sense
of
a
loose
or
out
with
storing
in
between
right,
because
tall
told
you
if
you
start
using
projected
rod,
you
cannot
go
away
and
use
the
normal
dag,
because
it
could
take
you
back
earlier
in
the
projected
rod
you
may
you
may
create
clips
right,
that's
where
our
loops
come
from.
C
So
if
you
start
entering
this
topology
this
virtual
topology,
that
is
the
projected
rod.
You
must
follow
it
to
the
end.
That's
how
you
avoid
loops.
Now
the
route
can
say:
hey
I
can
do
an
on
storing
path
from
A
to
B
and
then
down
from
B
to
C
or
B
to
D.
I
can
have
two
other
projected
rods
and
you
concatenate
them
as
long
as
you're
doing
that
you
probably
look
lies,
but
if
you
start
doing
things
like
oh
I'm,
stitching,
storing
and
non
storing,
rot,
etc.
C
Then,
if
some
nodes
in
the
middle
lose
some
state,
you
know
that
so
how?
How
far
do
we
want
to
go
with
us?
Do
we
want
to
say
I
place
some
rules
already
to
limit
the
way
you
could
play
with
those
stacks
I
would
like
people
to
review
that
I
mean
I.
Think
that's.
The
most
critical
piece
in
this
draft
is
loop
avoidance,
all
right
with
which
protection,
which
limitation?
How
do
we
make
things
simple,
that
we
know
that
won't
the
raw
loops,
but
then
same
time
will
limit
the
capabilities
of
the
protocol
right.
C
C
A
C
We're
meeting
our
draft
as
well,
we
need
we
need
anyway,
a
draft
and
rip
okuu
pressure.
I,
don't
understand
that
we
don't
have
it
I
thought
that
we
would
do
it
in
the
coming
years,
right
after
repo
in
Tim.
Okay,
now
now
we
have
eight
one
three
eight,
so
we're
in
a
very
different
situation
than
we
were
in
2007
2010.
We
have
eight
one
three
eight,
so
we
already
have
the
format
for
the
compressing,
the
the
rods.
C
So
why
don't
we
just
Express
the
via
option
already
in
the
eight
one,
three
eight
format,
meaning
that
when
you
build
a
package
natively
in
eight
one,
three
eight,
you
just
have
to
take
this
and
put
it
in
your
packet,
so
it
would
make
the
implementation
much
easier
if
the
format
that
we
have
here
is
the
native
formatting
format
in
which
they
will
actually
write.
The
packet
makes
our
control
smaller
and
makes
the
implementation
easier,
so
I
mean
right
now.
C
E
This
is
Peter
fantastic
I
really
want
to
draft
I
mean
you
have
been
talking
about
this.
You
have
changed
it.
We
changed
it,
I
mean,
and
now
we
seem
to
converge,
which
is
very
nice,
but
I
really
want
to
get
it
fixed.
So
if
you're,
if
there
are
additions
which
can
be
actually
exported
to
other
documents
and
if
it
can
be
simplified
and
such
such
that
additions
can
be
done
in
other
documents,
I
would
certainly
prefer
so
that
this
document
is
the
standard
way
where
you
could.
E
This
is
the
base
document
in
which
you
can
show
that
we
have
these
alternative
routes
and
if
there
any
additions
or
optimizations,
then
this
document
is
the
reference
and
you
can
go
forward
so
them
you
can
write
up
with
this
document
without
because,
if
you
look
at
this
list,
we
are
not
going
to
have
very
group
last
call
before
before
November
noise.
We
will
make
a
pause.
Oh
I,.
C
F
F
F
C
F
C
So
guess
what
Peter
I'll
try
to
put
some
more
energy
on
this
one
to
get
those
things
soft
on
this
one
I
did
not
mention
I
mentioned
stitching
right,
but
think
about
it
for
just
a
second.
We
are
not
if
we
just
push
okay,
if
we
just
push
Cyril
path,
ABCD
in
the
network
and
that's
it,
it
will
never
work
because
it's
very
the
route
can
be
very
far
away
from
the
route
and
the
route
cannot
react
instantly
to
any
problem
along
that
path.
C
C
I'm
I
have
two
very
validate
that
the
way
we
express
things
today
allows
us
to
stitch
things
in
the
right
fashion,
because
it's
not
variable
otherwise.
Linear
of
path
is
okay.
In
the
TV
world,
where
you
have
wires
and
very
reliable
links,
but
it's
not
viable
in
a
wireless
mesh
which
is
quite
large,
we
build
meshes
of
two
thousand
three
thousand
nodes
and
sometimes
10
to
20
hops,
and
we
deploy
them
and
they
work.
C
But
if
you
have
to
push
a
storing
mode
or
non
storing
mode
project,
Arad,
then
20
hops
away,
and
then
there
is
a
glitch
on
one
of
those
links.
You
can't
go
to
other
route.
Please
solve
this
problem
for
me,
you
have
to
separate
the
forwarding
from
the
rotting.
What
belongs
to
the
routing
is
those
gray
lines
that
it
monitors
statistically
now
to
understand.
This
gray
line
is
a
bit
more
great,
and
this
one's
with
more
blacks
ones
will
go
white.
Let
me
be
able
to
use
some
redundancy,
so
you
have
enough.
C
We
doesn't
see
at
all
times
to
get
packet
from
A
to
B
and
the
rest
is
not
my
problem.
That's
what
the
writing
will
do.
That's
what
the
Dow
projection
will
do.
What
the
forwarding
will
do
is
use
the
lip
of
things
like
that
to
understand
what
the
actual
state
are
of
those
links
is
to
use
this
neo
dag
between
a
and
B.
C
It's
usually
two
three
halves
right,
it's
not
four,
but
but
it's
multiple
possibilities
and
in
run
time,
I
have
to
look
at
my
packet
and
say
hey
what
kind
of
redundancy
do
I
want
to
use
in
this
packet
and
to
do
that?
I
need
to
very
well
understand
the
state
of
all
those
hops,
which
means
they
have
something
in
the
following
plane,
which
is
just
happening
in
that
little
deal,
dag
that
the
rot
projection
was
built
for
me
and
that
lets
me
use
it
optimally.
C
So
I
get
my
packets
wrong
if
I
flood
the
Geo
tag,
I'm
wasting
energy,
so
I
need
to
use
it
wisely
and
that's
when
all
the
information
I
can
get
actually
is
good.
That
you're
in
the
runs
that,
but
it's
actually,
this
kind
of
information
package
know,
am
feedback
to
source
to
source
routing
about
the
Geo
tag.
That's
what
worth
will
save
the
day
right
so
so
we
basically
need
the
source
in
storing.
In
on
storing
we
need
to
source
to
be
able
to
know
exactly
what
are
the
possibilities
in
this?
C
Do
that
so
it's
it's
about
getting
through
information
and
stitching
it.
So
so
you
need
to
be
able
to
correlate
multiple
segments
and
make
that
a
Geo
tag
right
now.
The
draft
just
gives
you
segments
not
the
full
deck,
so
we
need
to
make
sure
that
we
can
express
that
I
hope.
I
was
clear,
maybe
not
okay,
and
that's
that's
it
for
me.
So,
yes,
we
have
suggestions
here.
So
roll
roll
is
the
place
to
do
the
slow
piece
right.
How
do
you
do
this
great?
C
Do
tag
and
I'm
trying
to
create
a
new
working
group
called
row
for
reliable,
viable
Wireless,
which
is
the
other
piece
of
that
story?
I
have
this
local
to
your
deck?
How
do
I
use
it?
It's
not
a
writing.
Prime
anymore.
It's
a
falling
prom!
It's
about
OAM!
It's
about
link,
information,
instant
reaction
to
two
events.
So
that's
what
Roe
is
about
ranch
was
in
Roe.
Then
we
have
a
meeting
on
Thursday
morning.
What's
gone,
that's
pretty
much!
It
I'm!
Sorry.
C
K
C
C
Basically,
what
we
call
the
traffic
specifications
in
did
not,
and
in
the
one
hand
it
could
be
some
application
somewhere
in
the
cloud,
but
on
the
other
hand
it
could
be
a
node
in
the
network
which
decides
it
wants
to
talk
to
another
node,
in
which
case
we
need
signaling
from
that
node
to
tell
the
route
please
I'm
a
and
I
want
a
projected
route
to
be
yes
right
and,
and
so
we
don't
have
signaling
for
that
it
would
be.
It
would
be
like
a
multi
hub
this
or
something
well.
C
Well,
you
know
the
node
could
tell
a
mister
route
I
want
to
project
it
down
to
to
be.
Is
it
something
want
to
include
in
this
document?
So
it
was
not
in
malleus.
That's
just
lease
proposal,
this
question
so
I
guess
it
comes
from
operational
requirements
right,
yes,
something
we
want
in
the
field,
so
it
seems
there
is
a
need
for
support
for
the
host
there's
a
route
I
want
to
talk
to
mr.
B,
so
I'm,
sorry,
but
no
can
we
are
pushing
some
new
staffing,
which
is.
C
A
A
E
C
Yes,
I
just
want
to
know,
maybe
Li
you
can
post
to
the
mailing
list,
a
proposal
like
to
use
a
disk
or
something
like
that.
You
know
it's
probably
something
like
it's.
A
message
has
to
go
all
the
way
to
the
root
and
and
ask
for
a
path
with
certain
characteristics,
and
you
need
to
be
able
to
express
those
characteristics.
What
I
call
T
spec
is
three:
what
in
that
net
we.
C
So
we
say
we
want
this
reliability,
so
the
route
has
to
translate
that
oh
do
I
can
be
able
to
dag,
which
is
all
those
shades
of
gray.
If
I
want
to
reliability
of
five
nines,
that
mean
I
need
five
paths
or
something
like
that.
I,
don't
know
with
so
much
prae,
so
packet
ribbons
really
be
replication,
elimination,
type
of
games,
you
know
or
cross
three
depends
on
the
quality
that
you
want
to
obtain
and
the
route
cannot
complete
the
route.
C
Rotting
to
the
leaf,
so
this
one
is
easier
because
I
mean
you
need
to
understand.
The
PIO
is
a
big
thing
right.
It's
okay!
It's
a
lot
of
rockets
taken
time.
It's
a
big
thing!
It's
just
it's
like
a
ogv
right.
It's
it's
a
new
new
way
in
repo.
We
are
using
Sdn.
So
that's!
Well,
it's
no
more
that
it
takes
time
and
we
have
to
think
it
through,
because
it's
a
big
thing
now,
the
the
leaves
the
unaware
leaf
is
a
small
thing.
We
we
have
created
at
five
or
five.
H
A
A
H
C
We
made
sure
that
what
ripple
needs
can
be
actually
expressed
with
our
five
five
right,
but
so
now
I'd
five
four
five
gives
let's
the
host
tell
the
Reaper.
Router
is
what
I
want
in
a
way
that
the
river
Rother
can
turn
it
into
ripple.
The
cool
thing
with
that
is,
if
you're
low
power
device,
you
don't
have
to
do
six
low,
pile
and
D,
which
goes
all
the
way
to
the
6l
BR
and
then
Dow,
which
goes
all
the
way
to
the
root.
It's
just
one
of
those
two
flow
that
happens.
C
So
you
divide
the
amount
of
traffic
between
the
low
power
node
and
the
router.
You
divide
it
by
two
and
you
also
/
to
the
traffic
along
the
mesh,
because
you
don't
have
the
diode
ack
on
the
one
hand,
and
the
dow
in
the
other.
Actually
the
draft
allows
to
to
make
it
so
that
just
sending
the
devil
regenerates
the
DAC
at
the
root
dividing
the
traffic
by
two,
so
good
for
low
power
good
for
for
the
master.
Okay,
so
that
that's
pretty
much
what
this
graph
does
in
the
future.
C
As
you
can
see,
we
also
have
the
address
protection
and
D,
which
allows
us
to
prove
the
ownership
of
a
Russ
writes
now
right.
Now,
it's
done
on
the
nd
side.
It's
not
done
on
the
repo
side,
meaning
that,
if
you
have
access
to
the
repo
network,
you
don't
have
zero
touch
properties
well
by
you
could
not
inject
an
address
if
you're
on
it
right
now.
As
soon
as
you
get
access
to
the
ripple
Network,
you
can
send
it
out,
you
can
send
any
Dao,
meaning
that
you
can
claim
your
anyone
right.
C
It's
the
old
wait
before
zero
trust,
but
now
we're
in
the
zero
trust
world
a
PNG
gives
you
a
trust
on
the
nd
level
and
in
the
future,
I
hope.
We
extend
it
to
the
repo
level
that
we
cannot
inject
an
address
that
is
not
owned.
Basically,
so,
let's
follow
future
form
now
anywhere
leaf
is
very
simple.
Just
covers
eight
five,
four
five
translate
that
into
repos
saves
off
over
the
traffic
and
all
those
leaves
that
don't
support
ripple
at
all
they'd
have
no
knowledge
of
repo.
That's
what
it
is
for.
C
So
that's
pretty
much
the
expectation
that
we
are
working
at
so
like
I,
said
the
the
EDA
exchange
their
deck
is
now
originated
at
the
root,
so
we
don't
have
to
have
both
flows
across
network
thou
is
used
as
keep
alive,
and
the
draft
enables
to
separate
the
Ripper
root
and
the
6lv
are
most
of
the
case.
They
will
be
the
same,
but
we
actually
I
thought
at
first
I
thought:
I
would
remove
the
section
which
allows
it
and
then
I
got
use
cases
right
actually
happens,
so
so
I.
C
There
is
a
limitation
to
it,
and
the
limitation
is
that
in
the
Dow
that
is
used
to
regenerate
the
dyrdek,
there
is
no
concept
of
the
the
token
that
what
week
another
Rover
in
at
5:05,
which
used
to
be
the
MAC
address,
so
normally
when
you
build
a
doubt
that
you
need
to
place
all
the
fields
which
were
in
the
ear
option,
including
this
unique
ID,
that
is
the
Rover.
But
since
we
don't
have
it
in
repo,
we
cannot
actually
originate
it
at
the
root.
So
so
right
now
we
say
something
we
aren't
like
it.
C
C
And
Pierre
would
merge,
we
said
it
would,
but
then
since
then
there
were
use
cases
that
I
was
told
about
use
cases
and
just
found
some
where
we
we
don't
want
to
merge.
It's
actually
separate
an
example
of
that
is,
if
you
have
a
backbone
and
the
6lv
is
common
to
all
all
the
mesh.
So
the
6lv
are
now
is
a
central
guy
covering
the
whole
backbone
and
the
root
is
still
the
root
of
one
t
oday
now
separate.
C
So
if
you
actually
have
a
DHCP
server,
you
could
you
could
use
that
to
to
at
least
refresh
the
DHCP.
So
what
you
do
is
you
get
your
address
from
DHCP
first,
but
all
the
renew
is
actually
done
this
way.
So,
instead
of
doing
it,
I
could
do
a
DHCP
renew
and
actually
there
are
still
deployments
which
use
DHCP
like
ours.
To
be
honest,
so
it's
no
it's
it's
out
of
scope
of
this
back,
but
you
can
actually
do
a
DHCP
renew
as
opposed
to
a
data.
C
So
these
are
two
use
cases
where
I
used
to
me
to
split
so
so
kept
it.
You
know
it
did
not.
It
did
not
remove
it,
but
for
the
straight
operation
of
that,
if
we
really
want
to
do
it
well,
it
would
be
better
to
place
the
rather
field
in
repo,
and
that
will
be
useful.
I
just
told
you
a
soon
we
will
want
to
do
zero
touch.
We
would
want
to
be
able
to
authenticate
that
the
address
is
owned,
as
we
inject
it
in
ripple.
C
To
do
that,
you
will
need
the
rover
field
anyway,
right
so
I'm
tempted
to
ask
a
why?
Don't
we
already
put
the
rover
field
in
the
dowel
right
now,
so
next
we'll
be
able
to
secure
the
injection
of
it
out,
because
that's
the
tool
to
secure
it,
I'm
open
to
it
away.
Today
the
refresher
of
the
dalbergia
has
the
field
set
to
zero
I'm,
let's
sing
in
your
opinion,
I.
Don't
think
people
have
time
to
think
about
that.
So,
but
that's
pretty
much
the
only
improvement
they
can
think
of
in
this.
C
And
there
is
nothing
much
I
did
text
which
guy
was
cast
on
kind
of
an
object
that
if
the
leaf
does
not
support
it,
1
3
8
and
it's
not
supposed
to
then
the
IP
an
IP
goes
to
the
parent.
How
do
you
know
that
the
IP
IP
goes
to
the
parents?
It's
because
the
rod
is
injected
with
a
EP
8,
which
means
X
star
in
ripple.
So
basically,
if
the
roots
is
the
e
bit
set
in
the
dowel,
then
it
needs
to
to
I
pin
IP
to
the
parent,
not
to
the
note.
C
A
C
So
we
get
this
question
on
the
mailing
list,
whether
we
should
support
meepo
as
well
and
well
I'm,
not
too
keen
to
be
honest,
because
that's
completely
a
news
back
but
I'm
open
to
two
arguments
on
that,
because
the
clients
could
also
Peter.
They
could
also
be
people
unaware
kind
of
right.
So
do
we
need
to
do
the
same
game
from
evil
and
aware
leaves
what
is
there
to
be
done
and
is
it
the
same
document
part
of
the
middle
prime?
Is
the
hop-by-hop
cannot
be
ignored
either
all
right.
C
F
I
feel
like
maple,
is
so
much
less
well
understood
in
as
a
group
and
as
implementers
that
I
wouldn't
want
to
put
it
in
the
same
document.
For
that
reason,
because
I
think
we
won't
get
what
we
need
actually
sure
that
it's
as
big
a
problem,
that's
reunit
cast,
because
the
biggest
problem
we
have
in
unicast
is
that
we,
with
the
unaware
leaves,
is
that
we
don't
know
where
they.
What
what
the
pot
before
them
is
right.
F
If
we
could
know
what
the
hop
would
reliably,
what
the
hop
before
them
was
they
and
we
would
be
able
to
put
all
these
artifacts
addressed
to
them
and
they
could
be
removed
in
the
unaware.
Leave
just
works
right
in
unicast
right
the
problems.
We
don't
know
that,
which
is
the
the
issue
and
and
and
learning
that
is
hard
and
compensating
for
it
is
hard,
but.
C
C
F
So
so
I'm
thinking
about
some
really
sleepy
node
window
break
sensor
or
something
like
that
that
is
using
nipple
to
I,
don't
know
what.
But
the
point
is
that
if
it's
a
really
sleepy
node,
something
else
is
going
to
have
to
door
store.
So
since
that
thing
is
storing,
and
since
the
the
middle
packet
is
a
new
packet
does
being.
C
F
Can
send
it
unicast,
if
necessary,
with
whatever
else,
because
there
has
to
be
some
kind
of
thing
happening
there.
So
maybe
that
there's
a
bunch
of
cases
that
where
we
think
that
it's
nipple
to
the
to
the
leaf
it
doesn't
actually
have
to
be,
it
can
be,
the
last
hop
can
be
unicast
and
therefore
we
do
whatever
we
like,
because
it
has
to
be
stored
anyway.
F
So
I
suggest
that
I
don't
want
to
discourage
you
from
writing
a
document,
but
I
would
suggest
that
we
need
to
actually
kind,
of
wit,
not
wait
for
someone
to
come
up
with
a
real
use
case
or
a
real
situation
where
this
happens
or
as
I
feel
in
the
unicast
case.
We
kind
of
do
know
what
is
going
on,
maybe
Carsten
and
and
the
core
and
those
people
have
a
real
use
case.
That's
going
to
happen
and
that
that
would
be
great.
Then
we
would
know
what
we're
dealing
with
yeah.
C
But
but
okay,
what
you're
telling
us
so
far
is
there
are
a
lot
of
unknowns.
We
need
a
good
description
of
the
promised
and
then
it's
probably
a
separate
document.
You
don't
bother
the
idea
of
doing
it.
Just
say
not
this
back.
Let's
publish
this
one.
Yes
pretty
much.
Do
you
agree
with
that?
Peter
yeah
me.
C
C
K
Anita
a
requirement
from
the
parsnip
note
which
need
to
support
our
way
and
where
we
meepo,
we
cannot
use
unicast
er
to
send
the
multicast
packet,
because
it
we
are
support
image
that
if
a
thousand
of
people
deep
asleep
node
connect
to
one
power
node
that
you
need
to
up
to
firm
your
operator
for
the
tips
of
Noda,
we
suppose
to
use
multicast.
So
it's
better
to
support
your.
K
F
Michael
Richards
can
so
that's
wonderful,
that's
great
use
case
in
and
I
get
what
you're
saying.
But
what
I'm?
What
I'm
trying
to
say
is
that,
because
the
packet
going
out
is
different
than
the
packet
coming
in
and
none
of
those
sleepy
noes
are
expected
to
relay
the
note,
are
they
no
because
they're
too
sleepy
to
do
that?
So
it's
it
seems
acceptable
that
we
send
can
send
a
different
set
of
headers
on
that
multicast
cuz.
We're
not
expecting
that
that
that
packet
to
get
relayed
again
right.
F
F
We
make
every
change
it
change,
nipple,
but
but
I'm
saying
what
I'm
trying
to
say
is
that
if,
if
you've
got
the
synchronization,
it's
happening
and
none
of
those
nodes
that
are
expected
to
receive
that
message
during
that
wake-up
period
are
expected
to
relay.
But
then
then,
there's
no
point
in
inputting
any
headers
on
it's
just
a
late,
just
a
local
multicast.
If
some
of
them
are
going
to
relay
and
some
of
them
will
not
because
they're,
sleepy
and
they're
all
waking
up
at
that
time,
then
it
makes
sense
to
do
that
in.
C
E
C
E
C
The
trickle
you
have
to
track,
oh
okay,
this
document
I'm
asking
do
you
want
me
to
change
this
document,
say
something
about
trickle
roof.
I
was
afraid
you
wanted
me
to
do
to
say
it's
something
about
trickle
here.
So
what
so,
true?
You
want
to
talk
about
trickle
in
another
document
to
solve
what
not.
E
C
A
C
C
A
J
This
is
the
update
from
the
version
zero
one.
So
we
have
this
equation.
Zero
for
I
will
go
to
the
next
slide.
One
of
the
details,
so,
first
of
all,
many
thanks
to
Dominique,
Diego,
Arun,
Abbas
Carla,
so
to
Dominican
to
Diego
for
the
thorough
review.
So
we
got
a
bunch
of
comments
and
then
are
all
named
Pascal
for
the
discussions
we
had
the
mailing
list,
so
the
main
changes
that
we
got
to
zero
for
is,
but
now
we
are
having
the
common
ancestors
define
a
subjective
function.
It
is
draft
in
three
different
patterns.
J
J
So
the
common
ancestor
pattern
is
not
anymore
as
an
example
in
the
document,
but
it
is
as
objective
function.
So
it
is
a
prescriptive.
Prescriptive
next
thing
is
the
next
big
thing
that
we
changed
is
that
we
removed
any
reference
that
we
had
about
six
Laura
or
six
law.
I,
don't
know
how
to
pronounce
this
yeah,
yes,
which
lures
load/unload
where
H?
Ok,
so
we
remove
so
there's
no
conflict.
This
based
on
the
discussions
we
had
with
Pascal
with
Dominic
about
this
about
this
point
and
the
third
yeah.
J
There
were
like
many
comments
on
the
ability
improvement.
So
we
address
these
ports
as
well,
but
sort
of
the
main
changes,
and
so
before
asking
for
the
w
GLS
if
there
will
be
any
other,
if
it's
necessary
to
have
any
other
review
round
of
reviews
to
see
about
the
structure
or
the
new
version
that
we
have
now
a
subjective
function.
E
J
L
Everyone,
so
if
something
would
be
helpful,
if
you,
if
someone
would
take
a
look,
is
actually
a
bit,
maybe
an
editorial
issue.
So
the
way
we
defined
the
objective
function
is
what
would
be
the
equivalent
of
an
object-oriented,
the
extension
of
mr.
Hoff.
So
we
took
basically
every
section
in
the
mr.
Hoff
RFC
and
we
said
how
our
objective
function
is
different,
because,
basically
it's
an
extension
of
mr.
Hoff,
so
I'm
not
sure
I
haven't
done
this
before
I'm,
not
sure
this
is
the
correct
or
incorrect
way
to
do
it.
A
Thank
your
is
who's
being
to
review
the
that.
C
When
you
say
I
update
and
when
you
say
you
don't
in
this
case
you
are
not
updating.
Mr.
half,
because
mr.
half
steel
stands
as
it
is
for
the
use
case,
you're
just
in
everything
from
it
in
inject
that
you're
using
what
you
don't
modify
the
way
it
already
exists,
and
then
there
are
things
you
magnify
so
I
think
it's
perfectly
okay
to
express
it
this
way,
and
you
don't
need
to
say
you
update
mr.
half,
but
you
can
say
you
should
have
a
section
saying,
I'm
extending
mr.
half
and
your
option
is
I'm
doing.
C
L
I
I
J
Maybe
a
small
remark
is
on
what
Pascal
was
saying
before
so
since
we
removed
all
the
references
about
the
compression
at
the
da-jung
for
the
sixth
floor,
so
I
binned,
it
was
a
good
idea
of
having
a
new
document,
maybe
how
to
compress
the
control
plane
can
be
another
document
to
to
all
together.
Potentially
thank.
D
J
H
J
A
E
F
I
just
have
a
question
for
some
of
the
people
that
have
just
presented
some
things
mentioned:
you're
teaching
them
yeah,
so
I
have
a
question:
I'm
not
exactly
related
to
the
group,
but
is
ietf
interest.
The
documents
that
you're
writing
are
they're
being
counted
by
your
professional
tenure
review
committees;
no,
okay,
just
curious!
If
that
was
something
something
better
in
your
world,
that
I.
C
So
much
concur
with
Michael.
There
is
a
big
problem
with
our
Asian
IGF
and
academia
is
that
the
RFC
should
be
counted.
Look
a
look
at
Google
Scholar
right,
it
counts
patents,
it
counts
documents,
normal
publications
and
it
counts
RFC's.
All
that
are
valid
documents.
I
mean
we
really
need
to
work
hard
with
the
academia
to
get
our
our
public
she's
recognized.
A
I
G
G
So
the
talk
is
about
I'm,
going
to
talk
about
emoji
extinction,
which
is
a
pretty
straightforward
thing
to
do.
The
proposition
is
very
straightforward
and
the
other
thing
is
the
capabilities
handshake,
which,
in
my
opinion,
is
a
little
bit
complex,
is
definitely
more
complex
than
M.
Okay
extension,
both
of
the
things
are
handled
in
the
same
document
now
well
the
need
for
a
mopey
extension.
The
mode
of
operation
extension
is
quite
straightforward.
We
already
have
existing
drafts,
which
are
eager
to
use
new
mo
peas,
and
we
have
already
exhausted
the
existing
mo
fees.
G
What
is
an
MLP?
It's?
It
specifies
it
mandates
the
primitives
to
be
supported
by
the
6lr.
If
a
node
has
to
join
as
a
6lr
in
the
network,
it
has
to
support
the
mode
of
operation.
If
it
doesn't
support
the
mode
of
operation,
then
it
has
to
it
can
it
can
join
as
a
leaf
node,
it's
three
bits
in
size
and
it's
already
exhausted.
You
can
see
in
the
bottom
right-hand
side
that
there
are
existing
documents
waiting
to
be
using
those
course.
G
Like
aiya
so
moppy
extension
option,
this
is
a
new
option
that
is
introduced.
It's
a
pretty
straightforward
option
now,
one
one
one
question
here
would
be
the
extended
mo
p-value.
If
you
see
the
number
of
bits
that
have
allocated
for
extended
mo
p-value
is
quite
high
and
I
am
not
sure
if
those
many
extended
mo
P
values
would
indeed
be
required
in
the
future.
The
reason
why
I
ask
is
this
directly
impacts
the
control
over
head
of
the
TI
o,
so
so
I
am
considering
maybe
shortening
the
size
of
the
six
extended
mo
p-value
here.
H
C
Like
you
know,
you
could
have
both
like
your.
These
are
your
these.
Are
your
that
I
want
to
make
sure
that
we
express
things
in
a
fashion
that
you
can
have
these
these
this
and
that,
so
you
set
multiple
bits,
so
there
is
a
main
thing
and
then
some
options.
So
just
what
offer
we
express
it,
but
today,
right
all
the
maps
are
exclusive.
That's
why
I'm
saying
you
got
a
mix
of
throwing
a
non-story,
but
what
we
want
to
extend
in
the
extension
want
to
be
able
to
do.
C
G
G
G
C
C
So
I'd
like
to
use
the
map
for
what
the
map
is
and
say
it's
an
on
storing
and
it's
using
projected
Road,
storing
a
non-story
I,
don't
know,
and
if
the
rules
are,
if
you
can't,
you
know,
do
those
things,
then
you
don't
join
this
network
as
a
router.
You
have
a
leaf
at
best.
The
rules
are
located,
but
much
this
thing
capability
is
different.
C
That
I
can
do
this,
that's
nice
use
it
or
not,
and
then
there
is
another
thing
which
which
which
is
needed
in
the
in
the
in
your
draft
is
I
can
do
it
and
it's
on
all
rights
off,
for
instance,
you're
using
use
of
ripple
info
okay,
there
is
a
hop-by-hop
that
well
there's
a
new
up.
I
hope
there
is
this
new
flag
in
the
DIA,
which
says
a
it's
all,
no
itself,
it's
controlled
on/off
by
the
route,
but
what
did
the
node
receive?
C
It
did
not
apply
it,
so
the
node
is
asked
to
be
capable
of
supporting
the
new
hub,
I
hope.
That's
one
thing,
so
it's
a
capability
and
then
it
has
okay
I'm
using
it
right
now,
so
I
understood
what
the
route
told
me
and
I
switched
it
on
which
is
next
to
capability,
but
it's
actually
actually
how
I
use
it
right
now.
So
also,
all
this
is
more
right.
It's
a
bit.
G
More
one
example:
just
I
want
to
go
here,
so
if
you
see
the
existing
mo
piece,
we
have
storing
mode
of
operation,
and
then
we
have
storing
mode
of
operation
with
multicast
support
those
use
two
different
mo
piece
and
I,
don't
see
any
other
way
of
using
it.
So
something
if
something
is
mandatory.
It
will
eventually
end
up
taking
a
new
mode
of
operation.
G
Maybe
I'm
not
have
not
fully
got
just
of
what
you
are
trying
to
say.
Maybe
you
can
discuss
this
offline,
but
if
there
are
things
which
have
to
be
manually
supported
by
the
6l
ours,
then
they
go
in
as
a
new
mo
p
value
altogether.
The
problem
with
this
approach
is,
you
have
multiple
set.
For
example,
you
have
multicast
support
and
you
have
storing
mode
of
operation
and
all
the
permutations,
all
the
combinations
will
have
to
become
a
new
mo
p
value.
C
And
what
goes
in
the
MRP
in
the
contribution
option?
You
can
turn
something
or
not
in
the
ARP.
You
are
capable
of
doing
it
or
you
don't
join
the
network.
So
if
we
want
to
build
a
network
and
we
need
projected
route,
then
you
need
to
have
the
bits
which
you
are
used
to
for
the
storing
and
on
storing.
You
also
need
to
have
the
bits
which
they
would
say.
C
Oh
we're
using
Project
instance
projection,
storing
and
polishing
non
storing,
and
that
gives
you
an
array
of
bits
in
the
map
here
and
if
you
want
join
us
a
router,
you
need
to
support
all
those
bits.
Otherwise
you
don't
join
or
leave
kind
of
now
the
configuration
option.
Some
time
you
don't
even
know
you
don't
know
right,
because
those
bits
are
reserved
until
I
used.
So
that's
part
of
the
progress.
If
you
don't
know
that
you
don't
this
bit
is
now
meaningful,
then
you
cannot
know
you
don't
support
it.
G
So
the
next
thing
in
this
in
this
draft
is
about
capabilities
which,
which
essentially
indicate
the
set
of
features
which
could
possibly
mandatory
or
optional.
So
so,
the
new
specs,
which
finds
the
new
capability,
indicate
whether
it
is
mandatory
or
optional,
and
that
the
the
existing
craft
would
provide
some
some
recommendations,
or
at
least
some
some
some
basic
rules
as
to
what
have
to
be
minimally
specified
for
every
new
capability.
G
Now
why
a
mu
P
by
itself
is
not
sufficient.
Well,
it
is
pretty
much
clear
that
mo
P
mandates
the
primitives,
while
capabilities
are
not
mandated
in
those
can
be
optional
features
and,
unlike
a
mu,
P
capabilities
can
be
negotiated.
We
have
a
three-way
handshake
in
a
form
of
D
iota
and
our
work
and
capabilities
can
be
sent
as
part
of
the
Felicity's
annex
like
this.
C
G
C
G
G
This
this
draft
only
specifies
the
capability
options,
as
in
the
number
of
bits
that
have
to
be
carried.
The
next
next
light
weighs
now
the
use
case
for
route
using
the
capabilities.
The
route
can
signal
all
the
nodes
in
the
in
the
network.
What
capabilities
route
support
in
the
response?
You
have
no
node
sending
the
dow
responding
with
its
own
capabilities
and
in
the
Dow
I
possibly
route
can
tell
the
node
whether
it
is
okay
to
continue
with
such
a
capability.
G
The
three-way
handshake
again
has
to
be
has
to
be
mentioned
as
part
of
the
spec,
which
defines
a
new
capability,
whether
it
requires
with
whether
it
essentially
requires
such
a
three-way,
handshake
or
or
whether
it
is
okay
just
to
say
that
I
support
this
and
then
stop
it
I.
It
completely
depends
upon
the
new
capability
that
has
to
be
support.
Next,
like
this
now
now
this
this
slide
is
interesting
because
it
allows
so
what
does
like
tells
you
is.
G
There
could
be
six
LR
nodes
in
between
who
are
also
sending,
which
are
also
sending
the
iOS,
and
there
has
been.
There
are
capabilities
from
the
root,
node
and
and
sister
nodes,
which
the
6lr
receives
is.
Would
it
be
okay
for
this
6lr
to
mask
some
of
the
capabilities,
because
so
that
is
something
that
is
not
there
in
the
draft,
as
of
now
has
not
been
discussed
in
the
draft
as
of
now,
but
that
certainly
seems
to
be
possible.
G
G
I
C
Configuration
option
cannot
be
masked,
draw
cannot
be
modified
at
all,
because
that's
the
configuration
of
the
do
dang
now
this
one
gives
capabilities
and,
for
instance,
a
parent
can
could
have
maximum
100
children,
I,
don't
know,
and
it
might
be
useful,
and
we
know
it's
useful
to
be
able
to
express
number
on.
You
have
or
no
more
frauds
that
you
have
in
storing
mode
so
that
people
would
prefer
to
join
someone
else
because
you're
reaching
your
maximum.
C
G
G
Next
next
like
this,
so
some
of
the
other
points
to
work
upon
is-
or
some
of
the
other
points
that
I
would
like
to
mention
here
is
the
capabilities
and
the
mobis,
even
though
they
are
as
part
of
the
same
drafts
are
completely
exclusive
and
they
are
not
dependent
upon
each
other.
The
capabilities
can
be
made
use
of
with
existing
mo
keys
as
well.
It
is,
it
is
technically
possible
to
use
existing
mo
pieces.
G
G
This
is
just
something
that
that
I
want
to
float,
and
this
is
something
that
I
thought
about
next
like
this,
and
thank
you.
Thank
you.
George's
for
this
for
the
review
we
had
made
several
updates
post
that
one
of
the
things
that
is
updated
is
the
clarification
that
mopix,
if
it
is
absent,
and
if
the
base
mo
p
suggest
that
my
mopix
should
be
present,
then
how
it
should
be
handled
by
the
implementation.
Secondly,
it
has
been
made
explicit
whether
cap
and
more
pics
are
are
mutually
exclusive
and
the
detail.
I
G
G
F
The
wrong
words,
because
I
think
that
would
be
confusing,
but
so
I
have
some
comments.
Michael
Richardson
here
about
the
sites
not
avoiding
to
the
end
I
did
scan
through
your
draft
as
he
started
speaking
I
hadn't
read
it
before,
and
I
have
I.
First
of
all,
I
think
it's
a
very
good
idea.
I
think
we
should
immediately
adopt
it.
I
don't
know
if
we
should
split
it
into
a
cap
document
and
a
mop
packs
document
there.
The
pushback
is,
you
know
too
many
documents
right.
F
As
far
as
I
can
see,
the
two
concepts
are
are
completely
independent
right
and
maybe
one
will
take
longer
than
the
other
or
you
know,
I,
don't
know,
I,
don't
know
the
answer
that
question.
Maybe
our
ad
can
tell
us
at
some
point:
I
I
read
about
the
MA
packs
that
if
the
ma
packs
is
seven
sorry,
if
the
moth
is
seven,
then
you
add
the
number
that
you
find
in
the
mop
X
to
that
value
so
effectively.
F
I
F
If
this
number
is
seven,
then
the
real
number
is
here
I'm
using
two
hands
that
you
can't
see:
okay,
I
think
that'll
be
easier
and
less
confusing
to
code,
as
opposed
to
adding
the
two
things
and
someone
will
get
something
wrong
by
putting
negative
numbers
or
something
like
that
or
fine
extending
something.
Okay.
I
too,
do
you
want
to
respond
to
that
Pascal
yeah.
C
I
would
like
to
react
on
that.
One
please
I,
wasn't
with
person,
concur
with
mic
on
that.
What
I
wanted
to
wait
for
and
raise
the
might
say
that,
and
also
to
answer
you
on.
The
other
thing
you
said
is
hug
by
compress
things.
Well,
because
seven
says
that
the
map
is
in
the
MOU
text.
It
also
means
that
if
you
don't
have
a
map
extent,
you
have
to
wait
for
next
year.
You
to
know
what
the
map
is,
so
you
compress
it
right.
C
It's
kind
of
obvious,
once
you
have
adapted
Michael's
proposal
that
you
can
compress
the
map
back,
so
I
lied
it!
No.
When
you
want
to
alight
it
just
by
saying
eh,
it's
not
there
in
seven.
If
the
map
is
not
there
and
you
have
seven,
that
means
wait
for
the
next
yeah
you
to
know
what
the
map
is.
That
comes
with
it.
It's.
F
F
Maybe
I,
don't
think
the
mop
X
should
be
bit
encoded.
I
think
Pascal
was
thinking.
You
know
there
would
be
a
bit
here
and
a
bit
there.
I
think
that,
yes,
we
have
an
a
cross
product
of
different
kinds
of
things
that
we
need
to
do.
Not
all
those
cross
products
will
make
sense.
Okay,
having
a
bit
field
is
hard
to
compress
compressing
it
to
an
integer
for
the
ones
that
we
care
about
is
probably
better
for
compression.
F
So
if
you
want
to
use
fewer
bits,
then,
and
but
give
us
at
least
I
would
say,
7
to
128
combinations,
it's
probably
good,
okay
and
because
I
don't
think
that
we're
I,
don't
think
that
we
want
to
have
the
the
testing
complexity
of
having
to
test
devices
in
to
to
the
number
of
bits
possibilities
of
things
we
want
to
say
well.
Well,
there
was
two
to
the
you
know:
twelve
bits
possibilities.
Only
twelve
of
them
actually
never
makes
sense
in
the
field,
so
we
have
twelve
test
cases,
not
2048.
C
Yes,
okay,
I
can
sorry
I
can
buy
that,
but
then
I
would
like
to
express
it
not
as
but
that's
2.5,
meaning
meaning
the
two
is
what's
military
and
the
dot
five
is
what's
up
Shauna
or
something
like
that.
Just
I
think
my
I
was
not
very
clear
what
I
wanted
to
do
with
this
discussion,
but
what
I
really
want
to
do
is
right.
Now
the
map
is
all
of
nothing
and
I
was
thinking
in
the
projection.
If
I
ever
knowed
that
doesn't
do
projection.
What
do
I
do
with
this
guy
can
I?
C
F
Feel
that
that
needs
more
further
discussion,
so
I
understand
that
I'm
not
going
to
go
into
it.
I
I
wonders
if
the
DAO
projection
can
be
expressed
as
a
capability,
I
dot
sure
it
can.
I
always
thought
it
had
to
be
a
mop.
I
always
thought
that
we
maybe
can
make
heterogeneous
networks
by
somehow
reporting
ability
to
support
dial
projection
in
the
doubt
and
I
can't
remember
if
we
did
that
in
the
document
now.
C
That's
another
interesting
discussion
because
I
wouldn't
want,
if
you
don't
mind,
if
we
have
time
to
come
back
to
the
previous
slide,
why
it's
actually
about
the
eight
one
three
eighteen
compression
also
easy
to
capability,
or
you
know-
and
it's
not
a
matter
of
map
of
configuration
options
right
I'm
concerned
that
we
take
things
where
they
belong,
which
could
be
map
and
configuration
option
and
try
to
put
everything
in
capability.
Well,
the
capability
is
just
I
can
do
that.
It's
not
how
it
works
right.
C
F
F
I,
don't
have
an
answer
for
what
you
just
said,
but
I
like
it
I
like
Raoul,
very
much
that
you've
introduced
these
capabilities.
It
seems
like
that
should
be
a
bit
field
and
I
to
remember
something
about
the
containerization
of
it,
whether
or
not
we
can
extend
the
bit
filled
up
arbitrarily
by
just
making
it
longer,
but
I
can't
remember
both
the
container
of
that
part,
but
so
that
would
be
great
and
I
think
that
we
should
adopt
this.
F
And
if
and
I
further
say
if,
if
this
takes,
if
the
discussion
takes
the
entire
tyre
T
of
the
next
IETF
role
session,
I'm.
Okay
with
that,
because
I
think
this
is
a
really
valuable
thing
to
get
past,
because
and
I
don't
and
I
think
we
actually
need
to
delay.
Maybe
we
want
a
virtual
interim
I
think
we
need
to
delayed
out
projection
until
we
figure
this
out,
because
we
can't
publish
that
projection
without
having
figured
out
how
we
figure
out
whether
it's
on
or
not,
okay,
I.
C
Had
the
similar
discussion
very
recently,
this
is
infrastructure
work
right.
How
can
you
build
a
house
if
you
don't
have
a
solid
infrastructure,
so
we
need
to
refine
what
those
things
are.
We
need
to
define
what
goes
where
and
explain
it
correctly,
and
once
we've
got
this
infrastructure
work
in
place
now
the
turn-on
and
projection
can
all
use
it.
So
I
will
work
on
projection,
don't
worry,
but
we
cannot
ship
projection
until
this
is
shipped
anyway.
So
there
is
no
point
in
erring
100%
on
projection.
C
C
C
Here
you
seem
to
send
a
Dao
from
the
route
to
a
6-0
and
another
Dao
from
the
route
to
another.
6-0
and
I.
Don't
know
how
you
plan
to
do
that,
because
the
Jo
is
not
unicast
right.
It's
spread
and
the
route
capabilities
are
who
warned
that
the
wood
sets
and
it's
a
bit
like
the
configuration.
It
cannot
be
changed.
It
will
be
the
same
for
everybody.
You
means
the
root
capability.
C
C
C
A
A
A
A
K
Lena
from
Cisco
I
introduced
this
after
about
enabling
eighty
one
thirty
eight
in
brownfield
and,
as
you
know,
eighty
one
thirty
eight
is
used
to
compress
up.
He
people
are
the
factors
such
as
the
API
header
and
sorting
handle
and
is
fine
in
a
new
deployment,
but
we
find
that
there
is
a
capability
issue
when
we
need
to
upgrade
the
network
to
support
eighty
one.
K
Thirty,
eight,
you
know
the
all
the
FC
to
support
sorting,
sorting
header
and
the
RPI
option
is
fired
about
eight
years
ago,
so
we
have
deprived
the
millions
of
device
which
turned
to
support
eighty,
one
thirty,
eight
and
in
81
so
narrator
in
a
thousand
covered
colic
extent
and
and
migration
scenario.
So
that's
the
requirement
of
the
chapter
we
need
to
support
it
issue
and
the
Incans
can
contrast,
search
after
use
of
reporting
for
its
assets
fighter.
Three
in
to
data
configuration
which
indicates,
it's
used
twenty
three
four
up
here
header.
K
It
works
yeah.
It
works
for
this
use
of
reporting
for
chapter
and
we
add
another
flag
T
in
the
to
tell
configure
option.
Please
note
it's
a
configuration
of
option:
it's
not
a
capability
and
it
can
avoid
Flag
Day,
which
is
a
flag
day.
It
means
in
currently
deployment
if
we
need
to
under
upgrade
the
network
from
now.
Eighty
one,
thirty
eight
to
eighty
one,
thirty,
eight,
we
need
to
make
sure
or
node
in
this
network
I
mean
upgrade
the
firmware
and
we
need
to
Newport
Network.
K
K
That
means
we
can,
we
can
disable
the
t
flag
and
they'll
upgrade
the
firmware,
and
after
we
Apple
Apple
creator
or
the
firmware,
we
can
open
the
flag
to
enable
the
eighty
one
thirty
eight
and
you
know
if
no,
that
cannot
support
eighty
one
thirty
eight
they
can
stay
as
as
leaves
if
they
stay
at
leaves.
We
there's
no
crab
the
issue.
K
Okay,
this
is
the
operation
and
know
that
that
support
destructor
should
sauce
packets.
In
the
compress,
the
form
use
81
38
when
the
t
flag
is
set
and
another
support.
This
chapter
should
refrain
from
sorting
packets
in
the
compressor
form,
with
81
38
with
t
flag
reset.
This
is
the
this
is
how
to
use
the
t
flag
and
the
regardless
of
the
setting
of
the
pit,
the
loader
master
for
the
packet
in
the
form
it
is
received,
whether
it's
compressed
or
uncompressed.
K
K
Operators
can
refresh
over
the
air
and
then
restart
the
device
sonorously
keeping
the
network
globally
life
I
think
this
is
the
most
important
advantage
for
teens
draft.
That
is,
we
can
make
the
network
keep
life
when
we
upgraded
to
support
it
once
at
the
8th
and
network
can
only
be
migrated
to
T's
specification
if
all
notes
support
81
38
or
the
remaining
node
stay,
as
the
leaves.
K
If
we
fail
to
observe
that
the
drug
may
cause
the
remaining
node
received,
the
compressed
packet
in
81
38
for
more
format,
they'll
take
him
neither
uncompressed
nor
for
what.
So,
we
need
to
make
sure
the
node
that
had
done
to
support
81
38
stay
as
leaves,
and
we
have
some
method
to
keep
them
as
leaves
we
can
use
a
new
map
or
we
can
use
a
new
object
function
that
can
keep
the
latency
nodes
to
stay,
as
leaves,
if
all
possible
roads
are
not
compatible
of
81
38.
K
The
zone,
maybe
is
loaded
and
we
propose
two
scenarios
to
enforce
leaf.
Enforce
the
node,
as
leaves
we
can
use
single
instance
or
doubling
instance.
The
single
instance
means
we
should
keep
the
flag
off
and
we
use
the
new
board
or
new
object
function
to
upgrade
all
the
network
and
after
we
are
up
with
the
upgraded,
we
should
enable
the
T
flag
and
let
legends
in
know
the
two
P
dips
and
the
double
instance.
K
This
is
the
same
as
the
ship
in
night
case
in
6
5,
5
0,
which
we
can
find
it
the
last
one
is
we
need
to
encapsulate
the
parent?
That
means,
if
some
know
the
stay,
as
leaves
it
means
it's
a
doesn't
support
anyone's
at
the
8,
so
it'll
stay
at
leaves
say
in
this
case.
If
we
want
to
send
a
packet
to
this
leaf,
node
we
need
to
in
have
you
later,
the
packet
who
is
parent,
the
impaired
endure,
came
decompressed
packet
and
forward
to
the
leaf
nodes.
K
F
Michael
Richardson
I,
don't
think
we
want
to
I
agree
with
what
you're
doing
it's
good
I.
Don't
think
that
amok
would
help,
because
we
would
have
to
double
the
number
of
mops,
because
we'd
have
to
have
one
for
on
and
off
for
each
one
of
them.
So
I
don't
think
we
ever
would
a
mop.
Would
ever
use
I,
don't
know
the
venom.
Oh
F
would
work,
but
maybe
it
would.
If
a
capability
will
work,
then
we
should
try
that,
because
I
think
this
is
a
a
shove
whether
the
capability
is
useful.
F
F
F
So
I
don't
know
that
we
need
to
do
that
there,
which
makes
things
a
lot
easier.
They
still
have
to
be
leaves
because
they
can't
understand
other
things,
but
the
critical
part
is
their
parent
has
to
know
that
there
they
are
incapable
they
could
actually
operate
as
routers.
If
all
of
the
nodes
were
willing
to
do
that,
to
keep
track
of
parents
and
children
to
keep
track
of
what
they
can
do,
but
I
grew.
F
K
C
One
works
I'm,
not
saying
it's
a
good
idea,
but
I
agree
with
me.
It
works
because
the
new
nodes
which
become
configured
with
the
new,
wealthier
ones,
will
just
ignore
it
yeah,
and
so
we
would
build
that
Earthship
in
night,
so
so
it
would
work.
Another
thing
it's
durable
was
just
proposed.
All
the
options
and
I
tend
to
disagree.
It's
a
capability
again.
We
have
to
define
what
we
mean
by
the
world
capabilities
and
what
we
want
to
place
in
that
options.
I
think
it's
really
just
like
use
of
ripple
info.
C
It's
the
same
thing
and
because
the
same
thing,
it's
much
more
logical
to
use
the
same
means
you
use
b3.
We
want
to
use
b2
right,
it's
Semco
same
effect.
We
are
very
logical.
We
have
the
same
approach
to
everything.
So
that's
why
we
started
with
this
bit,
regardless
of
the
progress
of
capabilities,
because
I
want
to
progress
capabilities
anyway.
C
C
C
So
two
things
on
this
draft:
it's
it's
like
a
two
pager
right.
It's
it's
actually
my
fault,
because
when,
when
I
read
your
rewrite
reviews
of
reporting
for
itself,
I
realized,
hey
I
forgot
to
do
that
with
eight
one.
Three,
eight,
it's
my
fault
and
because
because
I
forgot
to
do
that
now,
we've
got
to
work
on
all
those
transition
scenarios
in
all
things
right,
but
it
also
means
it's
a
very
simple
draft,
like
it's
one
bit
in
one
configuration
option,
so
we
already
need
it
in
the
field.
C
We
would
like
to
deploy
it
once
created
weekend.
So
the
more
we
delayed
this
work,
the
longer
it's
going
to
take
to
deploy
it
once
right.
So
all
we're
asking
is
just
a
bit
which
is
very,
very
similar
in
a
sense
as
the
one
we
just
ship
in
use
of
report
info,
so
I,
don't
think
it
would
should
take
two
years
for
this
group
to
do
the
work
through
that
bit
I
mean
we
have
a
longest
three
at
the
HF
of
making
FCS
in
four
to
ten
years.
C
E
My
personal
opinion
cut
off.
Is
that
convinced
not
well
understood
at
this
moment?
So
for
me
personally,
I'm
not
clear
if
this
is
a
special
salute,
special
solution
to
such
a
problem,
or
that
we
should
have
in
general
solution
which
we
can
apply
it
more
often
and
just
use
in
general
decision
algorithm,
which
can
be
adapted
for
these
kinds
of
things.
Actually.
C
It's
replicating
the
solution
that
we
used
for
use
of
ripple
information,
so
it's
just
replicating
it
and
we
hope
that
we
can
replicate
again.
So
it's
not
a
first,
it's
actually
a
second
it's
just
because
I
did
not
do
in
this
draft
art.
We
didn't
use
of
ripple
info,
which
is
provided
a
way
to
do
the
brown
field.
I
just
wrote
it
in
a
green
field
without
thinking
of
Brahms
field
use
of
ripple
info
thought
about
brown
field
at
the
bit.
So
now
I'm
adding
the
bit
as
the
second
start
means
dump.
G
C
There
is
nice
to
have
which
we
don't
have
in
news
of
ripple
info
and
I
would
like
you
to
add
in
actually
to
just
as
a
starter,
for
the
capabilities
draft
I
told
you.
There
are
two
bits
really.
It
is
true
for
use
of
ribbon
for
it's
true
for
this
one
is
one
bit
which
is
I
can
do
it
and
the
other
bit,
which
is
I,
started
it
it's
running.
C
C
So
it's
just
for
a
better
automation
or
something
like
that
that
you
would
like
to
know
whether
it
has
actually
be
turned
on,
but
the
weights
written.
If
some
notes,
if
all
notes,
support
it
and
some
notes
turn
it
on
its
gonna
work.
It's
not
like
you
need
everybody
to
learn
done
because
you
have
to
for
the
packets
the
way
they
are
written
right.
C
So
if
the
administrator
knows
the
current
level
of
software
of
his
notes-
and
you
knows
that
they
are
all
now
graded
and
use
this
bit
to
turn
it
down-
it's
gonna
work
for
meat
enough
to
ship.
This
draft
was
enough
to
ship
use
of
ripple
info
now.
I.
Would
love
to
for
this
draft
and
for
use
of
ripple
info
to
have
the
capabilities
by
the
node
I
can
support
it
as
I've
switched
it
up
whether
it
goes
in
this
bag?
Fine,
yours
back,
I,
don't
mind
they
will
ship
together.
C
E
N
E
E
E
We
should
be
recharter,
I
mean
that's
almost
the
promises
we
made
at
the
end
of
this
year
of
other
stop
well
now,
I've
seen
that
there's
a
certain
set
of
new
documents
are
going
forward
in
this.
In
this
group
we
have
also
an
outgoing
documents
going
for
which
so,
which
means
that
other
healthy
group
at
this
moment-
maybe
you
don't
see
the
number
of,
but
we
all
use,
work
and
have
very
good
results,
so
I,
don't
think
that
we
need
to
change
the
Charter
and
that
without
changing
the
church.
Are
we
just
continue?
P
I've
liked
very
much
how
you
guys
have
managed
the
working
group
in
the
you
periodically
recharter
to
add
things
in
scope
and
go
from
there
now
I
honestly
didn't
read
the
Charter
in
the
last
seconds,
so
I
remember
exactly
all
the
details
of
the
Charter,
but
the
last
few
times
that
we
have
recharter.
It
has
been
to
include
specific
work
items.
So
I
have
been
happy
with
that.
P
I'm
sure
that
the
Charter
says
something
around
work
on
Ripple
and
you
know
something
that
so
in
theory,
there
are
other
things
that
are
in
the
Charter
right.
The
risk
of
putting
specific
work
items
is
that
the
assumption
is
that
others
are
not.
Let's
go
right,
so
you
know
for
one
point
of
view:
I
want
to
I
guess
my
answer
to
read
the
Charter
and
think
of
it
about
a
little
bit.
P
Think
they're
working
group
being
a
relatively
small
community,
is
really
good
in
not
just
diverging
and
bringing
all
kinds
of
of
other
proposals,
so
so
yeah
I
would
be
opening
with
that
ping
me.
Maybe
later
we
can
talk
about,
you
know
specific
sort
of
Charter
and
we
can
go
from
there
and
a
half.
The
actual
I
can
actually
really
answer.
E
Those
the
church
have
also
really
formulated,
in
general
terms,
there's
a
few
examples
of
work
that
you
wanted
to
do
key
my
reformulate.
Those
may
emphasize
actually
additional
things.
What
you
want
to
do
and
then
cite
some
of
the
examples
which
have
been
done
and
examples
we
want
to
be
do
and
then
possibly
continue
like
that.
Maybe.