►
From YouTube: IETF106-ROLL-20191119-1000
Description
ROLL meeting session at IETF106
2019/11/19 1000
https://datatracker.ietf.org/meeting/106/proceedings/
B
B
We
have
our
first
meeting
yesterday.
This
is
the
not
well
please
be
aware
that
this
meeting
is
aligned
with
that.
We
will
not
read
it,
but
it
applies
to
alight
the
of
activities.
B
These
are
the
meeting
materials
check
out
that
eater
pad
inside
the
same
done
just
read
if
we
put
them
Monday.
Thank
you.
Thank
you.
Thank
you.
Michael
Michael's
country,
this
I
prescribed
and
mineus.
Taken,
please,
okay.
Thank
you
very
much,
Alex
I'm
Dominique.
Thank
you
very
much
guys.
I
appreciate
yeah.
The
slides
are
complete
said.
This
is
the
second
part.
Please
sign
the
blue
sheets.
The
blue
sheets.
Are
there
okay?
Thank
you,
okay.
B
Thank
you
very
much,
Pascal
good
teamwork.
Here
this
is
the
agenda
for
today,
so
we
will
start
with
this
introduction.
Then
we
go
for
Pascal
with
an
hour
live,
then
we
will
keep
some
update
and
use
of
ripple
info,
then
Pascal
or
the
Minich.
They
are
going
to
talk
about
the
light
India
information
to
keep
the
semi
status
between
the
child
and
the
father.
Then
Raul
is
going
to
provide
his
update
for
MP
do
and
then
Pascal
is
going
to
talk
about
the
turn
of
a
d138.
B
C
Basically
we're
from
Cisco,
so
this
is
about
the
ripple
unaware
leaf.
So
there
was
a
lot
of
activity
on
this
document
and
the
reason
is
that
we
actually
found
that
it
was
a
normative
reference
to
other
documents
which
were
well
advanced,
and
so
now
the
ripple
unaware
leaf
is
kind
of
holding
other
documents
in
miss
rough
state,
and
we
want
to
avoid
keeping
those
documents
in
Miss
rf2
too
long.
So
we
ended
up
spitting
some
work
on
unaware
leaf.
It
is
not
a
very
complex
document.
C
It
is
the
application
of
RFC
8
5
4
5,
which
is
agnostic
to
the
routing
protocol,
but
still
is
a
registration
mechanism
by
a
host
to
a
router,
and
in
this
case
we
explained
how
you
can
use
RFC
8
505
to
register
to
ripple
services.
There
are
other
documents
at
the
ITF,
for
instance,
in
rift,
400,
proxy,
etc
that
also
leverage
RFC
8
5
for
5
at
storage
as
a
registration
mechanism.
So
this
particular
document
is
how
a
leaf
will
use
at
5:05
to
to
register
to
ripple
services.
C
C
Because,
basically,
there
is
this:
this
concept,
that
the
DCO
is
the
only
in
report
that
goes
down
as
synchronously
from
the
root
to
the
devices,
and
there
are
a
number
of
flows,
including
the
backbone
router,
which
require
possibly
a
miss
a
that
goes
through
the
repo
Network
and
synchronously
down
to
the
6lr.
So
the
router
can
actually
tell
the
leaf
that
there
is
a
problem
whatever,
and
so
we
found
that
it
was
actually
useful
and
we
did
it
to
carry
an
ND
status
into
the
repo
status
and
the
way
we
initially
did.
C
That
was
not
satisfactory.
It
was
kind
of
a
mapping
table,
and
now
we
replace
that
by
your
next
Blissett
signaling
inside
the
ripple
status,
so
that
if
you
carry
an
ND
status
in
the
repo
status,
then
there's
this
flag
telling
you
right
so
initially
the
repo
status,
the
left-most
built
was
a
knee
reverse
is
warning
that
if
the
status
was
above
128,
then
it
was
a
neuro
below
it
was
a
warning.
C
Now
we
have
the
second
bit
which
signals
whether
you
actually
have
a
native
report
errors
or
if
you
are
transporting
and
ND
status,
we
are
doing
all
this
game
because
actually
ng
and
repo
repo
is
kind
of
an
extension
to
nd
for
multi
hub.
You
can
see
it
as
that
and
that's
how
we're
using
it
in
in
many
use
cases
it's
just
ND
fun
and
BMA
networks
and
at
the
edge
of
the
ripple
Network.
C
What
you
see
is
with
the
real
draft
is
pure
handy
and
that's
true,
also
at
the
northern
edge
with
if
you
have
a
backbone,
with
the
backbone
router,
it's
pure
nd.
So
at
some
point
there
is
this
need
to
map
the
way
ripple,
expresses
thing
and
the
way
and
the
Express
things
and
you'll
find
the
same
kind
of
prime,
not
prime,
but
but
things
with
the
PIO.
For
instance,
the
prefix
information
option
in
rapport
is
the
only
one.
So
you
see
there
is
a
tight
integration
between
Andy
and
ripple
in
the
case
of
unaware
leaf.
C
So
we
said
a
with
the
real
draft
will
kind
of
package
them
into
one,
which
means
that
at
some
point
you
have
to
re
explode
that
and
so
the
integrating
the
status
as
we
did
is
part
of
that
game.
We
use
repo
down
the
geotag
to
signal
to
a
six
on
our
the
six
are
our
fixed
information.
It
needs
sends
it
to
the
to
the
leaf
so
that
that's
that's
one
big
thing
we
did
is
this
status
game?
C
The
other
big
thing
we
we
did
was
clean
up
the
the
way
we
integrate
with
address
protection,
so
in
ng
you've
got
this
concept
of
with
five
eight
five,
four
five
and
a
PMD
you've
got
this
concept
of
a
rover.
The
rover
is
a
capability
by
the
host
to
actually
prove
that
he
owns
the
address
right
now.
We
are
not
using
that
in
repo
to
protect
repo,
I
hope
in
the
future.
C
We
will
be
able
to
do
that,
but
for
now
we
have
not
using
it
to
protect
repo,
but
the
rover
was
needed
at
the
north
of
the
geotag.
To
actually
signal
to
the
6lv
are
in
case
the
6l
dri
separates
it
from
the
root
we
had
this
long
discussion.
Do
we
want
to
force
that
the
6
lv
r
is
the
same
box
as
the
root
or
not
at
some
time?
C
C
The
way
to
prove
the
ownership
of
the
target
is
now
piggybacked
in
to
report
so
that
that's
another
big
change
that
we
made
in
the
road
raft
so,
as
of
today,
is
just
to
enable
us
to
build
the
correct
exchange
with
the
6
lvl.
So
the
root
can
do
a
duck
with
a
6
lb
are
as
supposed
to
see
settle
down
there.
C
So
we
save
all
this
thing,
but
in
the
future,
that
also
gives
us
the
opportunity
to
actually
validate
a
ripple
rot
because
now
we'll
be
able,
since
we
have
the
rover
in
as
part
of
the
ripple
signaling
we'll
be
able
to
use
that
to
prove
the
ownership
of
the
address.
If
we
get
to
because
right
now,
anybody
can
inject
the
down.
C
C
Last
but
not
least,
we
so
we
pulled
kind
of
DCO
from
the
editor
cube
because
of
the
changes
which
are
related
to
to
what
I
just
said,
the
status
and
stuff.
We
also
pulled
the
use
of
ripple
info
from
from
the
editor
cube,
and
that
was
more
like
a
problem
of
definition.
What
is
a
rule?
In
particular,
there
was
a
discrepancy
between
how
the
use
of
ripple
info
define
the
real
repair
in
a
relief
and
how
this
draft
defined
rapport
anomalies.
C
The
end
of
that
long
story
to
make
it
very
short,
is
that
a
leaf
with
first
we
had
to
define
a
leaf.
You
realize
that
the
leaf
was
not
redefined
since
the
days
of
early
repo.
So
now,
with
the
unaware
thing,
the
leaf
definition
had
to
be
updated.
So
relief
now
is
a
ipv6
host,
so
it
has
to
comply
with
a
200.
So
it's
an
ipv6
host
which
is
attached
to
a
ripple
Network
right.
C
It's
a
host
supports
ipv6
attached
to
report
that
we
are
getting
routing
through
ripple
and
with
that
simple
definition,
you'll
find
that
the
way
the
rule
to
the
I
know
the
way
use
of
ripple
info
describe
forwarding
to
the
leaf
is
completely
consistent
because
it
doesn't
care
how
the
rod
was
set
up
right
rule.
The
rule
draft
will
be
one
way
of
setting
that
that
rod.
You
could
imagine
that
there
are
other
ways
or
other
sorts
of
leaf,
so
the
leaf
is
a
plain
ipv6
host.
C
The
only
thing
you
want
from
him
is
that
it
supports
a
hop
by
hop
it
out.
That's
fully
consumed
and
that's
part
of
8,200,
and
the
other
thing
you
want
from
em
is
that
if
there
is
this
hop
by
hop
header,
I
can
ignore
it,
which
is
also
compliance
to
8200.
So
it's
an
ipv6
host
with
that
we
can
use
the
use
of
ripple
info
to
wait.
What
is
right
for
a
long
time,
so
we
didn't
have
to
change
that
that
piece,
okay,
so
that's
a
leaf!
C
A
leaf
is
a
ipv6
host
attached
to
ripple
Network
I
mean
that's
kind
of
easy.
Now
ripple
aware
will
mean
that
he
understands
RFC,
6550
and
ripple
unaware.
He
doesn't
understand
their
FCC
specialty
simple
as
that.
So
now
we
have
a
very
simple
concept
of
what
a
ripple
unaware
leaf
is.
Is
that
enough
to
play
with
this
draft?
No,
because
this
draft
explains
how
a
ripple
unaware
leaf
will
use
eight
five,
four
five.
It
has
to
use
H
five,
four
five
in
a
certain
manner,
so
this
draft
adds
dependency
on
the
host
describes.
C
How
the
host
that
supports
at
5:05
will
use
H
five
four
five,
so
that
the
router
in
turn
can
provide
ripple
rotting
for
it,
so
that
so
this
is
kind
of
not
initially
like
four
versions
ago.
It
was
presented
as
the
definition
of
a
row.
Now
it's
no
more
the
definition
of
a
row,
the
rule
is
this
lifts
it
which
is
not
aware
of
ripple,
but
then
we've
got
this
additional
support.
That's
needed
on
a
crew
to
make
it
a
rule
that
does
this
raft
basically
hopes
clear.
C
B
D
C
C
What
alternate
would
we
have
like
define
the
second
field
for
and
the
status
separate
from
repost
errors?
And
it's
one
more
bite
right
I
mean:
do
we
need
more,
so
the
new
format
leaves
us
64
values
for
pure
ripple
status
and
64
values
for
nd
status
that
are
carried
along
and
I
mean
we're
very
very
far
from
needing
all
those
statuses
right,
so
I
think
yet
another
byte,
which
is
you,
know,
8
bits
in
year,
4
something
that
we
are
very
very
far
from
consuming
for.
C
We
did
not
even
define
a
single
arak
code
so
far
right.
We
just
said
if
the
bit
on
the
left
is
on
it's
an
error.
It's
a
rejection
and
we
expected
that
new
specs
which
come
in
with
examples
of
rejection
and
so
far
the
only
examples
of
rejection
we
have
are
the
one
that
are
carried
from
an
T
actually
and.
D
C
Exactly
so,
so
that's
why
the
bit
that
we
used
to
signal
and
D
is
actually
the
second
bit
when
it's
on.
So
if
people
started
using
1
0
and
1
2
3
4
5
6,
whatever
we
don't
impact
those
specifications
or
those
implementations,
but
really
if,
if
people
want
to
start
using
status
code
in
repo,
they
should
go
to
IANA
and
now
it's
something
we
can
look
at
actually
as
part
of
the
review
for
the
real
draft.
Since
we
are
actually
telling
IANA,
we
are
doing
those
things.
C
C
Where
do
you
expect
that
the
ripple
status
will
be
coming
from?
Would
you
like
some
some
companies
to
propose
directly
status
to
aiya
now
and
be
granted,
or
you
know
so,
I
mean
under
how
I
see
or
anything
interest?
But
but
that's
that's
a
big
question.
Iana
I
mean
if
you,
if
you're
telling
me
our
implementations,
which
may
be
starting
to
use
some
code
and
they
never
declare
them
anywhere,
it's
very
risky
for
them,
because
at
some
point
the
main
spec
may
decide
to
use
the
statuses.
That's
why
there
is
a
high
honor.
C
So
so
what
is
the
policy
for
allocating
at
IANA?
Do
we
open
it
so
that
you
know
some
implementation
can
go
to
IANA
and
say:
oh
I
want
this
code?
Yes,
you
get
it
or
do
we
want
to
close
it
a
little
bit,
so
it
has
to
be
a
standard
from
any
any
solid
body,
or
does
it
need
to
be
an
expert
with
you
or
never
a
CEO,
so
there
are
worlds
and
I
just
picked
one
at
the
moment.
So
please
please
consider
this
carefully
is
part
of
your
review.
D
The
other
question
that
I
had
was
with
regards
to
the
capabilities.
Now
that
we
have
the
6l
are
essentially
signalling
on
behalf
of
the
unaware
leave.
What
does
it
mean?
For
example,
if
let's
say
the
six
LR
is
capable
of
doing
a
23-8
and
the
never
leave
is
clearly
not
so
so
it
would
mean
that
somehow
the
route
should
know
that
the
six
LR
is
in
charge
of
the
unaware
leaf
and
has
to
do
IP
in
IP
I
would.
C
Say
that
way,
because
the
compression
is
a
subway
I
mean
it's
under
layer,
three
right,
which
means
that
you
could
you
could,
in
theory,
decompress
at
every
hop.
If
you
like
to
it's,
it's
it's
it's
the
way
you
place
the
packet
on
the
wire
I
mean
the
compressed
form
is
equivalent
to
to
the
uncompressed
form,
meaning
that
the
6rr,
if
it
if
it
doesn't,
know
better,
it
should
just
uncompress,
but
the
route
doesn't
have
to
care
whether
the
6r
makes
the
compression
decision.
It's
a
one,
hop
decision,
okay,.
D
C
B
B
B
Like
Pascal
said,
we
had
the
terminology,
be
fine
once
a
ripple
leaf,
so
he's
like
any
PVC
host,
but
we
can't
have
when
the
topology
is
firm.
We
can
have
a
router
as
a
leaf
as
well
right
Pascal
when
so,
but
actually
we
define
a
ripple
if
I
proceed
host
as
functionality,
but
whether
when
we
formed
at
the
apology,
a
router
when
join
a
topology
is
going
to
be
a
leaf
for
for
some
temporal
time,
but
we
are
in
here
kind
of
functionality.
So
then
we
have
the
ipv6
host
aligned
with
the
8200.
B
B
C
B
Yes,
we
we
say
in
the
document
that
we
don't
expect
that
rule
to
support
IP,
IP,
encapsulation
and
as
well.
We
don't
expect
the
rule
to
support
eighty
one
three,
eight
compression,
so
the
traffic
is
encapsulated
in
the
road.
If
it's
going
to
be
destination
to
read
the
rule
when
it's
encapsulated
into
the
route
the
destination
is
the
6lr
father
of
the
rule.
So
therefore,
two
use
cases
are
modified
in
our
use
cases
and
as
well.
We
update
the
RFC
editor
comments.
B
Basically
are
some
needs,
so
those
are
the
four
three
big
changes,
the
terminology,
this
section
about
external
targets,
and
then
we
update
the
story
mode
use
cases
where
the
rule
is
the
destination.
So
this
is
the
document.
So
please,
if
you
don't
agree
as
our
comment
or
otherwise
we
can
proceed,
we
will
send
an
email
to
the
mini
list
as
one
kind
of
last
call,
so
you
can
state
if
you
don't
agree
based
on
comments.
C
B
E
With
our
routing
ID,
yes,
what's
gonna
happen
is
you're
gonna
review
this
from
the
main
list.
You
instance
me,
I'm
gonna,
look
at
it
most
likely.
Looking
at
the
changes
that
have
been
done,
it
will
be
fine
because
we're
reviewed
it
I,
don't
think
we
initially
need
to
go
back
to
a
AG
evaluation
and
ITA.
Let's
call
and
all
this
stuff,
so
I'll
probably
just
tell
the
RC
editor
that
we're
ready
to
go.
Okay.
B
C
Thank
You
eNOS.
Yes,
when
we
did
repo,
we
we
thought,
as
usual,
on
the
immediate
prom
and
not
necessarily
on
how
the
protocol
would
live
on
meaning.
You
know
the
expectation
was
network
will
have
a
single
and
simple
configuration
and
that's
gonna
be
at
fiver,
and
now
the
kind
of
the
protocol
becomes
more
successful
and
you
see
all
sort
of
variation
of
how
it's
being
deployed.
C
And
so,
if
we
take
the
particular
case
of
what
we
call
the
configuration
option,
this
is
an
interesting
thing,
for
instance
animal
likes
repo
because
it
has
this
autonomic
thing
in
it,
which
is
that
basically
the
CLI,
the
configuration
of
the
routers
is
learned
as
part
of
the
protocol
itself.
So
you
just
have
to
configure
code
and
code
the
root
and
the
root
will
distribute
the
configuration
as
part
of
the
signaling
which
which
creates
the
rotting
in
the
first
place.
So
the
rod,
the
neural
network
is
self
configured.
C
This
way
just
well
not
really
fully
self.
You
need
to
configure
the
root,
but
then
the
whole
network
will
will
learn
the
configuration
from
the
repo
signaling
itself
as
opposed
to
what
you
find.
Usually
you
have
to
configure
all
the
routers
say
you
run
this.
This
protocol
on
that
interface
and
other
parameters,
and
you
need
to
be
imagine,
is
in
all
the
routers
for
this
to
work.
C
When
everybody
understands
the
configuration,
there's
no
point
repeating
repeating
repeating
and
so
ripple
says
you
can
alight
it
after
a
while,
but
that
really
implies
that
everybody
is
aware
of
that
configuration,
and
mostly
it
tells
you
that
you
don't
expect
the
configuration
to
change
after
that.
Well,
that
was
cool
for
the
only
repo,
but
now
we
want
to
be
able
to
place
modifications
and
if
we
just
send
a
Gao
with
a
new
configuration-
and
we
know
some
nodes
will
miss
it.
So
how
long
should
we
be
repeating
and
repeating
that
configuration
till?
C
We
are
sure
that
everybody
knows
every
everyone.
Every
node
knows
that
the
configuration
was
updated,
it's
kind
of
a
prom,
so
we
discussed
that
number
of
times
at
the
end
trim
and
we
ended
up
saying
a.
We
want
a
simple
but
very
explicit
signaling.
That
tells
us
what
is
the
version
of
those
fields
that
we
want
to
alight.
So
the
configuration
is
one:
the
capability
could
be
one.
The
prefix
information
can
be.
One
ripple
says
you
can
alight
them,
but
it
never
says
a
if
you
like
them.
How
do
the
nodes
know
that
it's
updated?
C
It's
not
complicated,
it's
all
about
using
a
sequence
number
for
every
option
and
being
able
to
signal
all
the
sequence
numbers
the
leak
of
like
city
is
that
the
sequence
number
is
in
the
main
Dao,
because
you
relied
the
options
and
so
basically
it's
kind
of
one
sequence
number
for
all
the
option.
If
you
don't
like
that
approach,
then
we
need
to
rewrite
number
of
things
in
this
draft,
but
it's
important.
It's.
The
only
kind
of
weird
thing
in
this
draft
is
the
number
is
the
number
of
the
DAO
all
options
considered.
C
You
change
one
option.
You
change
this
number,
meaning
that
you
have
to
figure
out
when
the
number
of
changes
which
options
are
impacted
and
when
you're
back
level
from
that
number
well,
you
need
to
dig
which
were
the
options
that
were
changed.
So,
in
order
to
be
able
to
do
that
smoothly,
we
introduced
a
new
option
which
is
kind
of
just
giving
you
option
by
option
the
option
number
and
the
sequence
at
which
it
was
last
updated.
So
you
can
very
rapidly
see
the
sequence
of
each
particular
option.
C
What's
the
last
change,
you
know
where
you
are
so
you
know
which
one
has
change.
You
get
those
ones.
So
basically,
if
you're
back
level
with
this
number,
you
send
this
message
to
to
your
parent
and
the
parent
of
the
parent
will
come
back
to
you
with
everything
that
changes
since
the
seconds
that
you
saw
giving
you
the
abbreviated
version
for
things
which
need
not
change.
So
you
have
the
real
number
and
the
unabbreviated
version
for
those
options
which
did
change
so
now
you
re
synchronized.
C
So
it's
really
a
synchronization
protocol
based
on
a
single
sequence
number
for
all
options,
since
this
equal
number
wraps.
That
has
the
consequence
that
you
need
to
increase
the
sequence
number
of
each
option
before
the
thing
wraps.
So
it's
not
perfect.
We
might
like
to
do
it
slightly
differently.
So
it's
just
a
proposal.
That's
on
the
paper.
If
we
want
to
have
one
sequence
number
per
option,
then
we
will
have
to
signal
the
abbreviated
version
of
all
options
in
every
packet,
which
is
a
lot
more
costly.
C
So
it's
a
choice.
The
choice
right
now
in
the
draft
was
to
put
one
single
sequence
number
for
our
options,
so
you
have
to
put
even
the
abbreviated
versions
that
saves
you
a
lot
of
mites
I
mean
like
4
bytes
per
option,
and
if
we
want
to
use
this
mechanism
for
more
and
more
stuff,
we
talked
about
having
the
SLA
option.
We
talked
about
I,
think
I,
don't
know
number
of
options
from
ng-include
into
the
DAO
and
you
want
to
abbreviate
them
all
well.
C
I
think
the
list
of
all
the
abbreviation
with
the
sequence
number
will
start
to
be
big
already
so
I
kind
of
like
the
idea
that
we
have
a
single
sequence
for
everybody,
so
we
don't
put
all
those
things
but
to
be
these
guys
right.
I
think
that's
the
most
discuss
about
item.
Apart
from
that,
it's
a
very
simple
spectrum.
I
mean
straight
to
the
point.
I
hope
I
mean
very
simple
problem
to
solve
very
simple
solution.
C
Apart
from
this
common
sequence
than
nothing
so
please
review
bla,
bla,
bla
I
mean
the
problem
is
still
fresh
in
the
mind
of
those
party
will
participate
it
to
the
entrance.
So
please
you
asked
me
to
write
it
review
it.
Let's
move
forward,
because
that's
that's
something
we
can
forget
soon
as
we've
done
it
so.
C
C
C
We
also
there
was
a
question
on
the
mailing
list
and
what
do
we
do
for
Dao?
And
you
know
mostly
in
storing
mode
I,
mean
if
nothing
changes.
That's
a
lot
of
now
information.
You
would
like
to
refresh
just
to
refresh
the
lifetime
or
something
like
that.
So
now
there
is
a
message
which
is
navigated
down,
which
tells
you
same
as
last
time,
just
refresh
eliezer
the
right
time.
C
C
So
we
have
a
new
object
to
request
the
the
last
version
of
the
options
in
the
desk.
So
you
say
what
what
is
your
view
of
the
last
sequence?
So
the
parent
can
decide.
There
is
also
a
discussion
like
if
two
parents
don't
have
the
same
sequence.
That
means
one
of
them
is
kind
of
obsolete.
You
might
have
a
parent
which
is
not
aware
of
the
changes
and
the
other
that
yes,
so
obviously
the
text
tells
you
use
the
parent,
which
is
aware
of
the
most
recent
change
right.
Rahul.
D
C
D
Right,
you
have
to
update
that
yeah,
but
but
but
the
AO
ability
adoption
is
is
the
responsibility
of
the
of
the
node
to
be
a
node
sending
it,
and
that
node
is
not
aware
that
something
in
between
has
changed.
The
path
is
going
is
not
going
to
be
the
same.
So
if
it's
a
in
sense
an
abbreviated
option,
all
the
all
the
all
the
intermediates
Excel
ours
are
not
available
on
aware
of
the
latest
latest
information.
C
If
we
can
discuss
that
on
the
mailing
list,
but
my
feeling
is,
if
a
parent
of
so
the
child
sends
the
abbreviated
saying:
oh
no
change,
the
parent
gets
that,
but
she
has
a
new
parent
for
the
parent,
which
was
old
yeah.
You
can
send
the
abbreviated
for
the
parent,
which
is
new.
He
sends
some
based
on
its
own
state,
but
you
need
again
to
be
able
to
reveal
the
state.
That's
what
your
discussion
yesterday
right.
The
assumption
is,
you
can
always
reveal
the
down
from
your
state.
C
So
if
you
have
a
new
parent,
which
is
not
aware
of
the
last
variation,
you
need
to
tell
him
makes
sense.
So
a
BC
right
a
stands
in
abbreviate
it
B
knows
everybody
say
else.
Not
challenge
D
knew
that
so
you
can
send
the
abbreviated.
Ii
did
not
know
it
new
parent.
You
have
to
send
a
full
version.
Yet.
D
The
only
problem
that
I
see
is
now
with,
with,
with
with
the
rich
information
that
we
are,
we
which
were
propagating
with
the
capabilities.
Are
we
expecting
every
node
to
keep
the
capability
in
the
routing
table
in
context
to
each
of
the
node
below
it,
because
that's
a
big
information
and
we
may
not
be
able
to
do
aggregated
now?
In
that
sense,.
C
You
know
the
capability
is
an
interesting
thing,
because
if
we,
if
we
say
we'll,
define
this
new
messaging
for
gathering
capabilities
that
it's
not
part
of
the
DAO,
so
it's
not
really
in
our
discussion.
It's
only
for
whatever
we
would
like
to
place
in
down
messages.
I,
don't
know
if
we'll
place
capabilities
into
our
messages,
so
I
think
it's
gonna.
Be
all
this
new
message.
You
have
to
define
right
so.
D
C
C
Bla
bla
bla
is
that
we'll
be
able
to
have
more
options
in
repo,
so
put
more
of
nd
into
repo
like
a
cellular
option
and
whatever
else
because
doesn't
cost
as
long
as
it
can
be
lighted,
it
can
be
safely
United
now,
so
that
that's
kind
of
the
big
thing
is
this
draft
is
simple,
but
it
really
opens
up
for
a
future
where
a
live
network
evolves.
Basically,
the
configuration
challenges.
C
The
prefix
is
updated
things
like
that.
Now
we
can
safely
say:
oh
everybody
will
would
be
aware
as
soon
as
he
gets
a
dial
that
the
prefix
has
changed.
So
even
if
it's
a
prefix
that
they
will
be
able
to
pull
it,
and
now
we
will
know
so
that
makes
you
know
our
network
much
more
manageable
or
can
live
for
much
longer
time
in
a
lot
of
what
we
are
doing
with
this
ripple.
V2
is
to
do
go
around
broad
networks
and
live
networks
right
and
keep
them
alive
all
the
time.
C
So
this
is
the
abbreviated
option
that
I
discussed,
which
replaced
the
full
option.
So
so,
basically
all
it
says
is
it's
so
it's
there
is
only
one
abbreviated
option
right:
it's
not
one
per
option,
so
the
abbreviated
option
is
the
abbreviated
option
and
what
it
tells
you
is
the
option
value
that
I'm
carrying.
C
Is
the
option
that's
being
abbreviated
and
next
to
it?
Is
the
sequence
number
associated?
So
you
will?
You
can
find
five
times
this
thing
with
five
different
values
of
the
abbreviated
option,
meaning
pio
configuration
blah
blah
blah
five
of
them
and
for
each
one
you
will
sindelle
the
last
LCSs
for
which
the
assumption
was
modified.
C
So
we
we
can
we
updated
this
message
and
this
please
challenge
this,
because
it's
not
necessarily
needed
think
about
it.
We
use
some
of
the
flags
to
say
I
want
this.
I
want
this
I
want
this
I
want
that.
But
at
the
same
time
we
also
say
here
is
the
SES
s
so
on
the
child
I'm.
Sending
this
this
and
I'm
saying
eh
here
is
the
last
sequence
to
which
I
was
sequence.
I
was
synchronized,
so
the
parent
could
decide
everything
that
challenge
and
push
it
to
the
child
on
this
I'm.
C
On
that
the
reason
for
the
flag
is
like
okay,
this
information
has
changed,
but
they
don't
care
I'm,
not
using
it.
I,
don't
know
which
configuration
that
would
be,
but
if
one
of
those
five
things
is
of
no
interest
to
the
child,
then
you
would
not
set
the
bet.
Is
it
a
valuable
thing
to
do?
I'm,
not
sure
it's
just
a
proposal
on
the
table.
C
C
If
you
have
a
long
time,
then
we
can
read
this
I
won't
read
it,
you
can
read
it.
So
that's
the
mechanism,
so
part
of
the
discussion
is,
can
you
detect
a
reboot
of
your
parent
or
actually
it's
a
route
that
cites
the
RCSS?
Can
we
can
we
find
reboot?
And
the
idea
is
yes,
because
the
straight
part
is
very
short
this
time
and
it
should
be
incremented
rapidly.
So
the
time
during
which
a
reboot
cannot
be
identified
is
actually
very
short.
C
I,
don't
remember
what
initial
value
I'd
put
for
a
CSS,
but
baby
is
250
for
something
which
means
that
one
or
two
changes
of
the
our
CSS
and
Bank
during
the
run
part
and
you
will
detect
a
reboot.
So
it's
it's
still
imperfect
because
of
the
straight
part
of
the
lollipop.
But
it's
the
windows
will
be
very,
very
small
where
you
don't
detect
a
reboot
Aled,
something
you
didn't
say:
Raul
yesterday
is
even
if
okay,
so
this
you
are
here
in
the
straight
path
somewhere
and
you
reboot.
So
you
go
down
the
straight
path.
C
C
Oh
okay,
explain
the
data
offline,
but
basically,
if
you
go
left
to
go
back
in
time
on
the
straight
part,
but
you
increment
rapidly,
there
is
a
point
where
you
will
pass
the
old
value
and
at
this
time
people
will
detect
your
you
know
where
you
are
so
that
they
will
at
least
learn
the
new
configuration
you're,
not
stuck
with
the
old
configuration
forever
I'm
clear,
I'm,
not
clear.
Okay,
we.
C
Okay,
it's
just
that
I!
Don't
expect
that
the
lollipop
in
the
CIO
is
an
efficient
tool
use.
You
said
it:
it's
not
an
efficient
tool
for
detecting
reboot
I'm,
trying
to
make
a
CSS
a
more
efficient
tool
to
detect
reboot.
It's
not
perfect,
but
it's
already
tuned
for
two
reasons
to
be
much
more
efficient.
First
reason,
too
much
shorter,
a
straight
path;
second,
much
faster
updates.
You
can
update
it
and
update
it
and
update
it
doesn't
cost
much,
and
it's
not
like
having
to
reveal
the
diag
or
anything
right.
D
But
so
my
what
I
want
to
say
here
is
the
you,
and
you
mentioned
it
already,
that
you
know
the
straight
part.
How
long
is
that
straight
part
going
to
be,
and
you
said
it
should
be
shorter,
a
shorter
as
possible.
It
has
both
implications,
it
has
advantages
disadvantages,
so
those
have
to
be
considered.
Yes,
I
think
that
we
should
have
a
different,
separate
discussion
for
the
server.
C
C
You
have
your
horn
on
sequence,
for
for
each
of
the
options
and
implicitly
so
each
option
bears
the
sequence
of
the
last
time
it
was
presented
info,
and
so
you
know
why
you
are
elastic
once
I
saw
was
20
and
I
see
that
the
GAO
has
23,
so
that
must
have
been
three
changes
that
I'm
not
aware
of,
but
I
don't
know
what
those
changes
are.
So
I
said
this
to
the
parent
with
20,
and
the
crown
should
be
should
send
me
back.
Everything
I
need
all
the
way
to
23.
C
It
might
need
more
than
one
message
to
get
there
because
of
MTU,
okay,
so
to
make
sure
that
I
have
all
the
information
all
the
way
to
23
I
need
to
see
all
the
options
either
in
the
abbreviated
form
or
in
the
full
form.
If
the
option
does
change
the
route
the
parent
will
set,
it
me
in
the
full
form,
with
the
newer
CSS
I'm
done.
If
the
option
was
not
changed
between
20
and
23,
the
parent
will
send
me
the
option
in
the
abbreviated
form,
saying
19
or
whatever
so
so.
C
I
know
that
even
if
I'm
at
23
deception
was
still
in
19
and
they
did
not
need
to
get
a
new
value
for
it,
so
that's
how
it
works.
Now,
like
I
said,
there
are
those
five
bits
which
are
the
five
option
we
can
alight
and
the
child
can
actually
right
now
most
indicate
which
one
it
wants
and
if
it's
always
all
once
then,
maybe
we
can
avoid
those
flags.
C
C
C
C
D
C
That's
true
I
mean
so.
Yes,
let
me
give
you
some
context
here.
So
there
was,
there
was
a
draft
about
this
I
hope
we
are
talking
about
the
same
thing.
There
was
a
draft
about
providing
some
options
to
Adeus
to
decide
whether
the
answer
should
be
multicast
or
unicast,
and
things
like
that,
and
so
the
discussion
was
I
just
like
two
bits.
Couldn't
we
package
that
discussion
into
this
draft
and
I'm
completely
positive
with
that
gate
and
we
want
feedback
and
yes.
F
It's
only
github
but
not
published,
yet
we've
started
merging
those
two
draft,
so
in
the
current
most
updated
version,
I'm
proposing
that
the
two
bits
we
had
in
AD
is
modification
draft,
be
the
two
bits
that
are
left
in
Pascal's
format,
and
there
are
more
questions
that
we
had
in
it.
Yes,
modification
dress,
which,
which
word
we
reset
the
trickle
timer.
F
C
I
have
some
comments
about
those
two
bids.
So
if
we
have
tied
and
then
we
can
start
the
work
on
this
piece-
yes,
we
have
time.
So
if
we
can
have
agree
that
it's
a
good
idea
to
just
put
them
in
here
as
opposed
to
having
a
draft
for
that,
the
first
thing
is
the
original
repo
gives
you,
oh.
If
the
message
was
unicast
or
multicast,
then
you
should
reset
this
or
not
all
you
should
respond
unicast
or
multicast.
C
C
Yes,
yes,
so
the
bits
should
not
signal
whether
you
use
trigger
or
well
use
multicast.
In
my
mind,
they
should
signal
whether
you
reverse
the
repo
behavior,
because
otherwise
you
need
a
third
bit.
If
you
want
to
be
able
to
signal,
because
if
you
basically
want
to
send
a
multicast,
but
you
want
to
reverse
the
behavior
of
repo
to
respond,
multicast
I,
don't
know,
then
then
the
bid
should
signal
a
do.
C
F
F
I
I
hear
what
you're
saying
I
think
it's
been
some
time
since
we've
updated
the
das
modification
draft,
but
I
think
they
current
ripple.
Behavior
is
too
complex
so
that
you
can
reverse
it
with
just
one
bit,
because
it
has
all
these
cases.
So
what
we
did
is
we
had
one
bit,
which
was
a
no
inconsistency
bed
which
basically,
when
it
was
said
to
zero,
which
was
a
reserved
value
anyway
in
the
original
draft,
would
not
change
current
ripple
behavior,
so
we
had
backward
compatibility,
and
so
we
need
to.
C
C
The
other
piece
is,
is
I
talked
to
the
team
implementing
that
at
Cisco,
and
they
are
this
interesting
observation
like
they
would
like
to
be
able
to
sound
the
disks
in
the
form
of
multicast,
so
has
to
cancel
the
emission
of
ADIZ
by
other
nodes.
So
if,
for
instance,
that
is
this
case,
what
that
we
call
power
outage
where
all
the
nodes
die
together
right
and
now
they
restart
and
guess
what
they
all
want
to
send
it
is,
but
if
they
trickle
out
this
thing,
one
will
send
first.
C
If,
if
the
deceased
sends
unique,
ask
to
the
old
parrot
to
a
previous
part,
then
the
other
children
won't
hear
it.
So
they
won't
be
able
to
refrain
from
sending
ideas,
and
so
what
they
really
want
to
achieve
is
I
want
to
send
these
to
my
previous
parent,
because
probably
I'm
rebuilding
the
same
network,
so
I
want
to
send
this
unicast,
but
still
I
want
the
other
children
I
want
the
unicast
behavior
of
ripple.
C
For
for
this,
like
not
resetting
the
well,
we
have
to
look
at
that,
but
the
bottom
line
is
I,
want
the
other
guys
to
hear
that
I
sent
is
dis
dis,
so
they
don't
sell
the
dis.
They
just
wait
for
the
DI
or
multicast
from
the
parent,
so,
basically
out.
If
I'm
the
winner
I
sanded,
this
first
I
won't
set
the
bid,
so
the
response
is
multicast
I,
don't
necessarily
want
to
to
reset
Treecko.
That's
another
discussion.
C
Well,
maybe
I
do
because
I
generate
yeah
I
need
to
reset
trick-or,
because
that's
how
it
generates
a
multicast.
So,
at
the
end
of
the
day,
yes,
I
would
set
the
multicast
bit,
but
I
want
that
even
the
dis
is
heard
by
the
other
guys.
So
in
the
meantime,
between
ice
and
the
DS
and
the
multicast
response
is
sound.
The
other
guys
don't
also
send
this,
which
could
interfere
with
the
DI.
Oh
right,
because.
F
B
F
C
Look
at
we
looked
at
that
and
we
thought
that
we
could
have
a
multicast
address
which
is
derived
from
the
parent,
so
we
already
in
v6
of
our
SC
2630
I,
don't
remember
what,
which
is
how
you
derive
a
multicast
address
from
a
prefix
right.
This
exists
there
is,
there
is
an
automatic
formation
of
an
ipv6
multicast
group
per
prefix.
Here
everybody
has
the
same
prefix.
C
We
don't
care
about
the
prefix,
but
we
would
love
to
be
able
to
configure
a
multicast
address
within
the
current
prefix
of
this
ripple
domain
for
the
surfing
store,
the
ID
of
the
parent,
meaning
that
if
you
have
a
charge
of
this
parent,
you
would
always
listen
to
that
multicast
address
and
obviously
the
parent
himself
would
listen
to
that
multicast
address.
So
when
you
sell
ad
I
get
to
that
multicast
address,
only
the
parent
can
reply
with
the
multicast
Dao,
but
the
other
guys
hear
it
so
now
they
refrain
from
sending
the
DES.
C
So
the
trick
we
had
in
mind
was
to
to
create
this
new
multicast
address.
That
would
just
be
this
parent
and
this
children,
and
there
is
a
side
effect
to
this,
which
is
quite
interesting,
is
the
children
will
do
MLD
to
their
parent
to
say:
I
am
listener,
and
now
that
gives
you
a
way
for
a
child
to
declare
to
the
parent
and
your
child,
because
the
way
we
did
non-story
the
child
sends
a
packet
to
the
root,
but
the
parent
doesn't
Reno.
C
If
he
has
children,
I
mean
it
can
infer
it
from
traffic
or
what?
But
there
is
no
explicit
signaling
say
I'm
your
child.
It
exists
for
the
rule,
but
does
not
exist
for
normal
non-story
and
so
I
think
this
multicast
group
and
registration
to
the
multicast
group
gives
the
parents
an
idea
of
the
children
that
it
has,
because
that's
how
the
member
of
the
multicast
group
fights
or
us.
So
we
were
looking
at
that
just
chef
I,
judge,
commands
on
this.
D
Roger
I
just
want
to
comment
on
the
whole
utility
of
know
of
we
having
a
way
for
telling
all
the
other
nodes
to
overhear
the
disc
and
not
send
it.
What
what
I'm
trying
to
say
is
the
scenario
where
a
node
sense
at
this,
where
mu
lot
of
nodes
are
sending
this
is,
is
is
a
one-off
scenario
where
all
the
nodes
gets
rebooted,
let's
say
even
if
they
send
all
that
all
of
them
send
the
discs.
D
What
the
problem
that
we
have
currently,
which
is
already
addressed
by
the
draft,
is
that
the
DI
electrical
timer
won't
get
reseted.
Once
we
have
that
solution,
then
the
DIA
is
sent
to
everyone,
and
everyone
stops
at
this
anyways,
so
I
mean
if
we
don't
go
the
other
way
we
can
avoid
a
lot
of
complication
is
what
I'm
trying
to
say:
let
let
the
node
sent
is
anyways.
C
Okay,
so
yes,
it's
two
things.
You
said
two
things.
The
first
one
is
I,
think
a
misunderstanding
of
what
it
means
to
reset
the
trickle
timer.
There
is
a
difference
between
resetting
a
trickle,
timer
and
restarting
a
t-coil
timer,
and
the
spec
fails
to
discuss
that
correctly.
So,
basically,
we
should
respect
I
what
it
means
to
reset
the
trickle
timer.
So
you
don't
have
this
problem.
Basically,
if
you're
triggered
timer
is
very
long
because
the
network
was
very
stable.
If
you
reset
the
trickle,
timer
I
mean
I
equalized,
zero,
reset.
C
Need
to
restart
it
because
otherwise
it
will,
it
will
timeout
in
a
long
future.
So
you
need
not
only
to
reset
I
to
i0,
which
is
really
what
reset
the
trickle
timer
means,
but
you
need
also
to
restart
the
trickle
timer.
So
now
you
fall
into
this
between
I
and
I
zero,
but
if
you
reset
it
again-
and
it's
already
started
for
this
short
value,
there
is
no
point
in
restarting
the
timer.
C
That's
a
mistake:
if
you
restart
the
timer,
it
will
never
her
labs,
you
should
not
restart,
so
you
should
let
it
elapsed
because
it's
already
in
the
right
region.
So
we
need
to
specify
that
so
you
don't
have
your
prime
but
I,
don't
think
we
need
signalling.
We
need
to
better
specify
what
it
needs
to
reset
the
trickle
timer.
So.
D
Just
a
clarification
here
when,
when
we
as
when
I
said,
resetting
I
mean
going
back
to
the
I
mean
okay,
so
there
is
no
use
case
where
we
restart
at
the
current
current
I
I
value,
I
interval,
so
that
there
is
there
is.
There
is
no
reason
why
a
node
can
reset
the
Dro
timer
to
the
current
I
value.
It
always
resets
it
back
to
the
I
mean.
C
If
some
origin
I
mean
then
then
well
yeah,
but
the
prime
is
at
that
time
you
should
not
restart
the
timer.
If
you're
already
in
I
mean
you
should
not
restart
the
time
or
if
you
restart,
you
will
keep
restarting
it
for
every
days
and
then
it
will
never
elapsed.
So
so
the
mistake
here,
it's
an
implementation
discussion
for
your
auto-draft.
The
observation
you
should
observe
the
way
trickle
is
implement
and
the
observation
is,
if
I
more
than
and
then
I
mean
I
zero,
then
resetting
means
I.
C
Equal
I
mean
my
zero
and
because
now
you
will
elapse
further
than
that.
You
need
to
restart
your
timer,
but
if
I
is
already
equal
to
I
mean
then
you're
already
elapsed,
saying
in
the
expected
window,
let
it
run
don't
restart
the
timer,
that's
what
it
means.
So
so
this
discussion
is
not
in
report.
It's
not
well
enough
explained.
I
have
to
go
to
the
trigger
RFC
because
it
should
have
been
written
there.
But
if
it's
not,
then
we
need
a
section
in
your
document
explaining
exactly
that.
C
If
I
equal
I
mean
don't
restart
the
timer
boom,
this
problem
goes
away.
Now.
The
other
prime,
you
say
it's
one
your
case,
but
it's
a
real
prime.
We
have
it
in
the
field.
The
real
parameters
go
to
Canada,
some
of
those
are
from
Canada
and
go
down
the
basement
of
a
big
building,
and
you
will
find
all
the
we
call
matters
on
the
wall
and
that
can
be
200
of
them
now.
C
Outraged
in
this
building,
200
notes
got
out
of
power
bank,
so
they
have
to
notify
the
utility
that
they
lost
power.
They
have
to
send
this
parallel
Taoiseach
notification,
bank
and,
as
we
have
all
sorts
of
primary
being
able
to
do
that,
but
the
worst
case
is
now
when
you
power
back
up
again
guess
what
all
those
notes
start
sending
this
at
the
same
time,
on
the
same
frequencies,
blah
blah
blah
the
interfere
with
one
another,
the
interfere
with
the
DDI.
Oh,
it's
a
cacophony.
Nobody
hears
anything
right.
C
So
what
you
really
want
to
do
is
either
the
router
is
like
keys
and
cdiu
before,
or
one
of
them
send
the
first
disk
and
the
router
sends
the
DAO,
but
if
all
the
others,
200
plus
notes
and
this
at
the
same
time
it's
really
a
cacophony.
So
for
us
it's
a
real
feel,
prime,
that
we
have
right.
Then
it's
it's
not
unusual.
D
But
this
is
solved
with
the
current
text.
Isn't
it
I
mean
if
we
stop
resetting
the
trickle
timer
and
our
definition
of
resetting
the
article
timer
is
going
back
to
I
mean
if
it?
If
we
stop
that
and
we
letter
the
node
send
this.
The
DI,
Oracle
timer
won't
get
restated
and
other
nodes
will
hear
the
DI
oh
and
they
will
stop
sending
this
and
start
yeah.
D
C
Soon,
as
the
giu
multicast
is
sent
with
the
bits,
thanks
to
the
bits
that
you
have
then,
yes,
the
cacophony
stops
if
the
nodes
here
of
the
DI-
oh,
but
if
they
are
sending
this
at
that
time,
they
won't
even
hear
that
yeah
you
and
they
will
keep
sending
this
so
so
you
can't
receive
and
transmit.
At
the
same
time,
right
I
mean
there
is
a
weird
moment
where
everybody
wants
to
talk
and
nobody
listens
just
like
in
this
room.
No,
no!
No
joking.
C
So
that
needs
to
be
an
trickle
is
good
at
that.
That
needs
to
be
a
winner,
East
out
of
the
parent,
sending
the
DI.
Oh
that's
perfect,
or
a
child
sending
ideas,
but
then
everybody
should
silence,
be
silent
chill
because
the
DAO
is
not
instantaneous.
The
whole
principle
of
Treecko
is
the
response
is
always
delight
by
something
akin
to
admin
right
in
the
order
of
I
mean,
and
so
during
that
time
today,
with
the
current
implementations
in
the
field,
everybody
shouts
nobody,
he
listens
right
and
the
Jo
is
lost
and
even
the
DS.
F
C
F
Yeah
we
I'm
saying
we
can
have
an
option
that
helps,
but
we
should
pretend
it
and
pretend
we
resolve
in
by
ourselves
same
thing
in
in
our
draft.
We
had
this
spreading
option
as
well
for
the
responses
when
you
you
know,
if
it's
your
posit
situation
in
unit
that
turns
the
network
and
Kroy
has
no
idea
what
brought
us
around.
So
it's
sending
a
case.
C
F
C
C
C
F
F
C
B
D
B
D
D
B
But
as
for
example,
each
document
has
like
a
security
consideration,
we
may
you
can
include
like
a
section
what
happens
when
a
note
restart.
So
we
can
feel
that
because
they're
always
question
so
we
can
kind
of
have
a
mandatory
section
well
mandatory
like
like
it.
What
happened
and
that
would
be
nice
in
each
document.
Have
that
section
I,
don't
know
what
do
you
think
I.
C
Believe
that
house
document
on
the
observations
is
so
useful
and
at
least
verbally
we
should
capture
what
was
said
to
the
in
rolls
document.
So
at
least
we
have
a
status
of
that.
We
know
this
problem,
we
discussed
that
and-
and
there
is
where
we
are
at
least
I
agree
with
what
you
said-
I
mean
having
a
section
on
exactly
but
already
writing
down
a
for
instance
to
trickle
time
we
are
resetting
trickle.
Does
it
means
we
starting
the
time
on
this
le?
C
Oh,
that's
important
must
be
written
somewhere
and
and
so
all
this
discussion
about
the
use
of
the
bits
etc.
So
we
need
to
complete
on
the
mailing
list
in
the
agreement
on
the
bits
that
are
inherited
from
the
other
thing,
which,
which
also
relates
to
the
fact
I'm
using
five
of
those
bits
and
it's
not
awfully
necessary.
So
if
we
start
using
those
bits,
remember
I'm
using
five
bits,
and
so
that
means
you
know
kind
of
a
waste.
C
D
Just
one
more
on
each
other:
well
just
one
more
thing
that
I
would
like
to
mention
in
this
context
is
that
we
should
expect
future
specification
to
generate
new
options
which
should
be
readily
or
it
should
be
possible
to
elect
them
as
well.
And
we
give
some
sort
of
considerations
in
this
document
saying
that
if
you
have
a
new
option
and
it
rarely
changes,
then
you
should
follow
so
and
so
things
I
mean
you
have
to
define
a
sequence
number
for
every
option.
B
C
B
C
B
For
adoption,
I
hope
there's
a
if
you
are.
If
you
agree,
we
can
publish
your
marriage
documents
and
then
we
go
for
adoptions.
Is
that
okay
for
you,
yep?
Okay,
comments,
questions!
Okay,
thank
you
around.
C
D
Other
I'm
going
to
talk
about
the
updates
to
the
NP
Dow.
It
went
into
the
IG
last
call.
We
received
all
the
IG
reviews
and
then
we
had
an
update
when
one
important
update.
Rather
we
are
now
carrying
and
ripple
status
as
part
of
DC,
oh
well.
This
is
very
generate
and
it
is
important
for
DC.
Oh,
the
node,
generating
the
DCO
to
sort
of
give
a
reason.
Why
is
it
generating
that
ECU
in
the
first
place,
so
it
is
regardless
of
the
unaware,
leaves
proposition.
D
It
is
important
for
DC
or
to
give
this
reason,
we
should
have
identified
that
in
the
first
place,
but
now
we
have
it,
but
so
this
slide
explains
you.
The
impact
of
unaware
leaves
on
on
on
this
particular
draft,
and
we
have
seen
a
Pascal
has
already
explained
it,
but
what
we
have
here
is
we
have
the
RPL
status
format
changed.
We
have
another
bit
which
is
which
is
mentioning
whether
it
is
rather
nd
status.
So
now
this
text
I've
picked
up
from
unaware
leaves,
but
it
is
quite
important
to
be
noted
in
this
context.
D
D
C
So
sorry
I
missed
that
one.
We
just
discovered
it
by
reviewing
with
how
like
yesterday
or
something,
but
we
are
the
more
interesting
discussion
if
I
may
say,
is
that,
since
now
we
are
carrying
the
stairs
and
the
status
can
be
one
to
127
like
the
EBIT
is
not
set,
then
we
could
really
make
the
future
of
this
yo
lot
brighter.
C
C
This
message
for
new
cases-
and
actually
we
had
one
yesterday
so
it
looks
like
I-
would
like
to
see
this
additional
change
to
say
a
don't
discard
the
path
if
the
idiot
is
reset
and
if
you
understand
the
value
and
the
value
tells
you
to
do
something
to
it.
Otherwise
just
pass
the
DCU
on
down
the
path,
and
this
way
we
can
signal
nan
your
information
down
the
path,
for
instance,
I,
have
too
many
children
I
need
to
rebalance
the
tree,
which
is
the
use
case
we
had
yesterday.
C
So
it's
an
encouragement
of
people
down
the
path
to
repair
in
somewhere.
So
we
rebalance
the
graph,
it's
just
one
case,
but
we
thought
eh.
That's
the
only
message
we
have
that
flies
I
synchronously
down
from
the
root
and
if
it's,
if
I
mean
just
using
it,
for
one
purpose,
is
kind
of
a
waste
right
having
the
capability
to
say
a,
it
can
be
more
than
a
rejection.
C
E
Hey
al
Loredana,
that's
a
big
change,
very
big
change
for
several
reasons.
Not
only
are
you
using
the
mister
something
else
which
is
great,
I
love
we
use.
If
that's
what
works?
Do
that's
perfect
the
drafts
a
couple
things
besides,
that
it
says
something
about
not
expecting
the
nodes
to
do
anything
based
on
status
now,
you're,
changing
that
right
and
and
the
other
part
that
really
concerns
me
is
the
fact
that
you
say
that
the
status
I
forgot,
the
exact
words
must
not
change
or
something
on
the
path.
E
So
you
know
that's
I
it's
just
a
it's
just
a
way
that
was
written
that
maybe
was
written
in
draft
that
because
we
can't
enforce
that,
then
we
can't
really
say
you
must
not
change
it
right.
So
so
we
need
to
to
be
clear
about
those
those
cases
and
be
clear
about
the
fact
that
I
can
now
tell
someone
hey.
You
need
to
repair
it
for
no
apparent
racist
right
just
because
I
want
to.
Of
course,
this
may
be
better
than
telling
them
the
path
is
not
available
at
all
correct.
E
C
Please
my
bottom
line
is:
if
we
keep
the
text
as
it
is,
the
nodes
will
effectively
not
because
of
this
mast
will
not
look
at
the
stairs,
which
means
that
this
creates
a
backward
compatibility
issue
whereby,
even
if
we
want
to
extend
it
in
the
future,
if
we
have
a
legacy
node,
it
will
kill
the
route,
which
is
not
what
who
I
mean?
That's
sad.
If
the
life
I
mean
having
this
bit,
which
sighs,
if
you
don't
understand
it,
but
at
least
you
know,
destroy,
not
destroy
something.
C
That's
proven
very
useful
in
ng
and
many
other
use
use
cases
right
what
you
do
when
you
don't
understand
it.
So,
although
all
the
idea
is
okay,
we
did
not
expect
the
nodes
to
understand
it
and
they
would
destroy
the
route.
Now,
at
least
we
say
eh,
if
you
underst,
you
don't
understand
it,
but
it's
a
warning.
You
don't
destroy
the
route
you
pass
it
down.
That
was
so
useful
in
other
protocols,
I
mean
I
can
wait
for
the
RFC
number.
C
D
I
just
want
to
say
the
statement
at
the
the
enforcement
policy
that
Anwar
you
mentioned.
You
know
there
is
no
way
for
the
nodes
intermediate
nodes.
I
mean
intermediate.
Nodes
can
essentially
change
it
and
there's
no
way
to
enforce
it.
But
the
same
is
true
even
for
the
existing
down
messages,
so
I
mean
if
we
completely
depend
upon
Delta
security
or
the
security
or
mechanisms
in
our
field.
For
that
so.
D
A
B
D
H
C
C
But
the
whole
network
needs
to
use
the
same
functions.
So
as
long
as
you
get
but
level
devices,
you
cannot
reuse
it.
So
if
the
new
software
said
just
use
it
and
you
all
software
who
is
not
using
it,
then
your
network
would
be
inconsistent.
So
there
is
the
need
to
replace
the
Flag
Day
by
a
process
whereby
your
greater
nodes
at
the
speed
you
can
and
when
they
are
all
graded,
then
you
fire
a
bit
somewhere
and
everybody
works
with
the
new
operation.
C
So
that's
the
way
that
we
solve
the
problem
in
the
use
of
repo
information
for
the
new
value
of
the
hub
by
option
of
API,
and
we
have
the
exact
same
problem
for
another
repo
specific
operation
which
is
the
use
of
compression.
So
it's
not
the
first
discussion
the
meaningless
like.
Oh,
it's,
not
the
shape
of
the
geotag.
Well,
the
configuration
is
not
just
the
shape
of
the
geotag,
it's
everything
that
people
does
and
the
compression
is
one
of
those
things
that
ripple
does.
That
needs
to
be
homogeneous
within
a
particular
do
DAC.
C
If
you
have
a
multi-tier,
that
configuration
may
be
one
deal
diagnoses
it
and
the
other
does
not,
but
at
least
under
a
certain
route
as
part
of
your
subnet,
but
not
necessarily
the
host
at
that.
You
need
to
know
if
it's
on
or
it's
off
and
you
need
to
be
able
to
deploy
the
nodes
with
a
new
software
and
this
it's
off,
so
they
don't
choose
it,
and
when
all
the
software
is
installed
everywhere,
then
you
can
decide
to
turn
it
up.
So
this
is
a
configuration
thing.
It's
not
the
same
thing
as
the
capability.
C
We
can
live
without
it
because
we
have
other
means
to
manage
the
devices
which
are
not
based
on
a
capability
right.
So
does
this
need
to
wait
for
the
capability
draft
now?
Does
this
need
a
normative
reference
to
a
variability
draft?
No,
is
it
something
that's
useful
in
standalone?
Yes,
just
like
it's
actually
a
Mis
use
a
free
pollen
fo
as
the
bed.
We
did
not
miss
that
time,
but
when
we
rotate
one
three
eight,
we
are
so
concerned
by
the
compression
that
we
forgot
the
bran
field
and
we
forgot
to
put
this
bit.
C
So
it's
just
catching
up
with
a
mess.
That's
no
more
than
that
draft
has
been
very,
very,
very
stable.
There
were
discussions
about.
Is
there
are
other
ways
to
use
to
do
that,
rather
than
the
compression
the
six
CIO
is
nd.
It
talks
about
a
subnet,
not
a
geotag.
We
should
not
make
ng
aware
of
our
setback,
configuration
or
ripple
specific
stuff,
so
it
doesn't
look
like
a
good
idea
in
this
case
and
now
it
if
we
did
in
6ew
it
would
be
harder
to
tie
with
the
capability.
C
I
would
like
to
do
the
same
thing
for
using
the
hub
by
hop
and
for
using
the
compression,
because
it's
a
very
similar
problem,
so
it's
cool
to
be
consistent.
The
way
we
approach
those
Prime's
somebody
said
we
could
use
a
new
protocol.
So,
yes,
let's
define
a
new
protocol,
that's
priced
through
the
network
and
also
new
messages
just
to
carry
one
bit.
Why
not
well
I'm,
not
too
good
for
that,
so
I
think
we
are
we're
kind
of
okay
and
ready
for
let's
call
for
an
option.
Yeah
thanks
for
adoption,
I'm!
Sorry!
H
They're
all
just
bits,
just
one
at
a
time
Michael
here
so
Pascal,
explained
to
me
the
disconnect
between
I
ween
capabilities
and
this
document,
I
I
was
not
previously
happy
with
turn
on
not
using
capabilities
to
enable
itself,
but
now
I
understand
what
he's
proposing
that
that
later
on,
we
may
have
a
capability
that
asks
where
we
can
discover
if
the
network
is
81.
Thirty-Eight
ready
in
Pascal's
case
and
I
believe
this
is
a
metering
networks.
H
So
they've
done
them
all
I
guess
it
takes
two
years
or
something
to
visit
every
home,
and
then
they
turn
on
81
38
right
and
if,
if
they
lose
some
devices,
I
guess
they
they
screwed
up
and
missed
some,
and
that's
all
so
I
support
doing
this
I
understand
the
the
somewhat
urgency
of
it
and
okay
right.
It
would
be
nice
again.
H
D
Roger,
though,
I
completely
agree
with
that,
we
need
this
kind
of
stuff,
but
the
one
important
one
thing
that
I
want
to
mention
is
that
basically,
what
it
avoids
is
what
we
are
we
are
sort
of
avoiding
through.
This
is
a
flag
day
operations.
Where
we
now
know
the
route
can
signal
all
the
nodes
that
something
is
now
can
be
turned
on
by.
So
what
I'm
trying
to
say
is
in
eventually,
we
will
have
other
other
mechanisms
which
have
to
be
turned
on
by
the
route
after
then,
after
the
discovery
of
that
mechanism.
D
So
is
it
possible
for
us
to
generalize
this
so
that
it
can
be
used
for
other
purposes
as
as
well,
and
not
only
for
80
138?
So
this
is
a
generic
problem
statement
where
the
route
needs
to
tell
all
the
nodes
in
the
network
and
the
reason
why
I'm
concerned
is
because
now
we
are
making
use
of
the
DAO
base
objects,
I
mean
the
configuration
options,
flag,
which
are
limited
and
if
we
use
for
specific
purpose
as
eight
one,
three
eight
and
I
see
a
reason
why
we
are
doing
it.
D
But
eventually
we
will
need
this.
For
other
things
as
well,
so
what?
What
so?
Is
it
possible
for
us
to
do
it
generally,
such
that
in
the
future,
if
the
route
has
to
say
to
the
all,
to
all
other
nodes
that
you
can
yes
start
using
this,
not
only
for
eight
one,
three
eight
I
mean
that's
a
great
question.
C
And
I
think
it's
realized
a
lot
with
them
up.
We
had
a
similar
Prime
with
them
up
and
I
expect
that
when
we
need
more
room
and
we
have
to
configure
ripple
configuration
option
right,
we'll
we'll
extend
the
ripple
configuration
option
for
more
bit
space
or
whatever
so
so.
For
now
we
don't
need
it.
Let's
use
the
space
we
have.
The
very
cool
thing
is
because
of
the
other
draught
eliding.
C
Think
right.
Yes,
you
can
start
thinking
about
this
mechanism
and
if
you
want
to
piggyback
that
with
the
map
ax,
why
not
because
it's
infrastructure
I'm
not
really
using
my
packs
right.
We
are
just
saying
a
the
value:
seven
will
mean
ripple
v2,
but
we
are
not
defining
new
Maps
at
the
moment
same
thing,
we
could
already
start
for
providing
the
mechanism
to
put
new
options
in
the
configuration
option
and
some
things
missing,
and
so
we
have
the
infrared
II
so
around.
C
Document
which
does
not
recreate
a
new
map,
it
just
nabel's
us
to
extend
them
up
space
in
the
same
fashion.
Maybe
if
we
need
to
do
something
to
be
able
to
extend
the
configuration
space
with
new
options,
and
you
want
to
give
formats
or
something
like
that
to
ignore
them
or
understand
them
or
things
like
that,
then
you,
you
could
package
that
with
more
tags,
if
you
don't
define
anything
new,
but
just
a
a
a
generic
transport
or
signal
for
those
just
like
map
exaggerate
transport
from
new
maps.
Do
the
maps
don't
exist
right?
C
D
C
But
by
then
we
will
have
the
configuration
option
there
with
extensions
and
whatever
yeah
I
see
the
prime
right,
but
it's
true
for
the
hop-by-hop.
Also
I
mean
you
can
view
that
in
two
years
from
now
everybody
we
use
will
be
using
X
23
and
then
signaling
it
will
become
useless,
uh-huh.
Okay,
so
same
prime
I,
don't
have
an
answer.
Well,.
H
C
H
It
would
be
easier
if,
if
we,
it
would
be
easier
if
we
had
split
that
off
into
its
separate
document
and
then
we
could
mark
it
experimental
as
well,
and
so
instead
we
would
have
to
write
another
document
in
the
future.
If
we
did
a
65
50
bits
when
we
would,
we
would
then
mark
that
bit
as
as
deprecated
right.
H
H
C
B
C
C
H
Well,
so
what
I'm
trying
to
say
is
that
that
so
so,
ok,
so
next
year
you
deploy
new
software
and
it
takes
a
year
for
an
operator
to
actually
deploy
it
to
all
the
household.
So
that's
that's
I
would
say
realistic.
You
know
we're
talking
about
probably
a
truck
wool
guy
rolling
yeah
walking
down
the
street
updating
cousin.
A
lot
of
these
meters
don't
have
don't
necessarily
have
out
there
a
field
upgradeable,
but
not
in
band
field.
Upgradeable,
over-the-air,
okay,
at
least
the
one
in
my
house
is
not
is
not
it's
not
upgradable.
H
There
there
they
they
have
in
times
come
removed.
The
meter
put
it
in
a
truck
well
they're,
probably
through
a
new
one
on
it,
but
put
it
in
a
truck
upgraded.
It
put
it
back
and
I've
had
no
power
for
10
minutes
right.
That's
how
it's
worked,
but
I'm
told
that
they
don't
have
to
remove
the
meter.
They
just
are
become
overly
cautious
about
electrocuting
themselves.
H
But
the
point
is
that
it
takes
some
time.
As
you
saying
it
takes
some
time
before
the
network
is
upgraded
and
then
they
turn
the
bid
on
okay,
and
then
they
run
for
a
number
of
years
during
which
there's
probably
other
other
software
patches
that
come
out
and
they
they
apply
them.
At
some
point,
you
you
make
the
decision,
you
know
what
we
are
no
longer
ship
support
for
having
the
bit
off
right.
If
the
bit
is
turned
off,
the
meter
won't
run.
H
Right
I
mean
it's
full
I
mean
in
this
case
it's
a
little
bit
weird
because
you
can
always
not.
You
can
always
not
compress
as
well
like
eighty
one.
Thirty
one
lets
you
compress
better
right
and
if
you
just
don't
compress
so
well,
well,
that's
just
the
networks
just
copes
with
it.
So
in
this
case
it's
a
little
bit
weird
because
you
can,
you
can
always
run
you
can
always.
If
you
have
81
thirty-eight,
you
can
always
not
use
it
as
I
believe
it's
still
the
case
right,
III.
C
Don't
see
that
I
mean
the
mechanism
looks
good.
The
way
I
describe
it,
but
in
in
in
the
real
world
I
don't
see
that
the
world
will
care
if
this
bit
was
experimental
and
some
Islanders
may
so
you
know
it's
most
like
at
some
point.
You
plan
to
reuse
it
and,
and
that's
the
day.
The
nightmare
starts
because.
C
With
an
old
implementation
that
sees
that
this
bit
means
that
and
that
guy
was
not
reflashed
and
and
now
you
say,
oh
I
have
to
refresh
my
own
network,
and
maybe
there
is
this
thing
in
particular
deployment
which
is
not
so
much
in
use.
That
still
speaks
your
language
of
that
bit.
I'm
really
really
concerned
that
when
we
reused
a
bit
for
something
else,
it
will
create
a
nightmare,
but.
H
C
So
so
we
need
to
define
the
method.
For
that
like
say,
there
is
another
or
something
next
to
it,
which
tells
you
the
version
of
those
bits
version:
zero,
its
use
of
ripple
info
plus
RFC,
eight
one,
three
eight,
and
when
we
go
version
one
this
bit
will
meet
something
completely
different,
but
that
is
the
current
software
can
already
know.
If
those
bits
mean
that
or
not-
and
if
it
means
it's
future
thing,
then
it
was.
C
H
C
H
It
might
not
it
might,
it
might
not
be
mop.
Equal
seven
makes
it
it
might
be
mop,
equals
7
plus
mop
X
values,
we'll
say
what
what
what's
going
on
right,
but
that
I'm
saying
is
that
something
in
that
space
sounds
like
the
way
that
we're
getting
to
ripple,
v2,
okay
and
and
we
get
there
quickly
or
slowly,
I,
don't
know.
But
but
that
sounds
like
the
process
by
which
we
should
essentially
obsolete
some
of
the
compatibility
options.
H
It's
at
that
stage
that
we
should
say
that
whether
and
I'm
not
picky,
whether
at
which
particular
thing
is
the
clue,
the
the
the
statement
that
we're
at
quote
v2
right
and
that
doesn't
mean
that
we've
spawned
the
document
even
right,
because
that
may
be
a
lot
more
effort.
We
may
want
to
wait
for
a
little
while
and
then
we
spend
the
document
remove
stuff,
nobody
used
right,
but
they're.
So
oh
they're,
oh
you
probably
already
spoke
while
I
was
no.
C
It's
a
very
important
discussion.
I
was
wondering
why
we
we
needed
that
much
time
for
this
document,
but
now
I
realize
it's
just
four
uh-huh,
but
because
it's
not
really
this
document
right.
It's
really
those
bits
will
be
there
forever
and
it's
a
way.
H
C
H
So
so
maybe
maybe
the
document
should
say
this
document
applies
to
mop
non,
storing,
I,
don't
know
if
you
want
to
apply
to
storing
as
well
I'm
open
to
that
statement.
Okay,
but
you
may
not,
but
but
then
you
may
want
to
say
that
this
bit
is
undefined
for
other
mop
values
until
such
time
as
someone
else
decides
whether
it's
defined
or
not.
H
H
H
No
I'm
saying
in
use
of
ripple
we
would
we
would
qualify.
We
also
qualify
our
use
of
the
bit
to
say
it's.
The
three
values
of
mop
that
I
think
are
the
relevant
three
that
matter
and
I.
Don't
think,
there's
any
other
ones,
there's
one
that
no
one
uses
or
something
right
right.
I
can't
remember
what
the
list
is,
but
so.
C
B
C
C
C
E
E
Because
in
reality,
what
it
sounds
that
you're
doing
is
that
you
were
creating
sort
of
a
different
registry.
I
mean
right,
because
the
way
I
ain't
had
those
things
right,
there's
the
registry-
and
it
says
this
bit-
is
whatever
right
today.
It
doesn't
differentiate
between
the
moms
right.
So
basically,
by
saying
this
only
applies
to
some
mops
and
not
others,
you're
sort
of
adding
columns
to
the
registry.
We
are
to
say
this
bit
applies,
you
know
one
four,
six
or
whatever
this
other
bit
applies
to
everything
or
you
know
whatever.
C
E
E
C
E
Exactly
so,
that's
what
I
mean:
there's
no
column
today
right
so
someone
somewhere,
and
one
of
these
documents
has
to
update
the
registry
so
that
now
and
in
the
future,
for
every
other
bit,
you
have
to
indicate
what
mops
the
bit
is
valid.
For
so
I
mean
it
gives
you
a
lot
more
flexibility,
and
it
actually
gives
you
a
lot
more
bits
right,
because
now
you
can
have
all
the
bits
supply
to
one
specific
one
or
not
apply
to
something
I.
C
Completely
agree
for
one
and
I
was
thinking
as
we
went
on,
which
document
should
be
doing.
This
I
was
thinking
after
the
case
have
in
mind
is
6lowpan
in
sixth
up
and
initially
we
started
with
with
the
bits
in
just
one
table,
and
then
we
did
arrive.
C85
28
and
28
0
25
its
0
25,
basically,
is
mostly
about
saying
that
the
sixth
applied
compression
now
is
in
tables,
so
you
have
like
16
tables.
C
So
we
are
did
this
table
column,
which
is
exactly
the
same
as
the
mop
column
now
well,
initially,
6
up
and
even
without
it,
and
then
there
was
8
525
8
0
25,
which
was
mostly
about
creating
that
column.
In
our
case,
the
inference
which
an
infrastructure
thing
it
does
not
really
a
new
function.
It
please
new
infra
and
we
have
a
mock
draft
which
is
coming
map
X,
which
will
bring
new
in
fraud,
but
not
not
new
feature,
and
this
new
infra
will
be
when
the
new
cap
maps
will
be
introduced.
C
They
won't
because
the
problem
is,
if
we
did
it
in
use
of
repo
or
if
we
did
it
in
this
one,
then
it
would
not
be
balanced
kind
of
and
right
now
we
don't
need
them,
because
we
say
it
applies
to
all
the
that
exists.
So
so
we
could
say
we
can
leave
for
those
two
drafts.
The
way
use
of
ripple
is
already
written
and
the
way
this
one's
already
written,
but
we
need
to
do
what
you
say
very
soon
in
an
infra
draft
which
is
mopix.
Maybe
that's
just
my
suggestion.
Right.
E
So
yes,
so
saying
that
it
applies
to
all
the
mops
that
exist
really
says
that
it
applies
to
all
the
moms,
even
the
future
ones
that
are
not
defined
right
because,
again,
right
now,
there's
nothing
that
says
that
whatever
bit
has
already
been
defined,
would
not
apply
to
some
future
thing,
because
think
of
the
registry
is
flat
right,
it's
just
one
column.
So
do.
C
E
C
We
know
in
our
two
drafts
will
say,
applies
to
mob
bla,
bla
bla,
and
for
the
rest
we
can
say
that's
what
we'll
say
right,
but
we
don't.
We
won't
be
doing
inr
games
we'll
just
declare
the
base
to
the
IANA
as
if
there
is
a
single
set
of
bits
and
in
map
X.
We
say,
oh
by
the
way
for
future.
Like
you
say,
it's.
H
H
E
E
D
Raja,
though,
can
we
go
for
another,
buy
one
more
question
in
the
same
context,
not
in
the
same
context
with
regards
to
the
use
of
a
23-8.
There
is,
and
now
we
have
a
dependency
of
this
on
P
Dow,
because
we
might
have
non
storing
mode
segments
in
projected
Dow,
and
we
would
like
the
node
nodes
to
know
that
a
23-8
is
on.
C
In
storing
I
mean
even
in
even
if
the
whole
network
is
not
fully
at
once
rate
capable
it
might
be
that
a
given
path.
Yes,
and
so
you
could
you,
if
you
are
the
capabilities
to
know
that
this
path
is
8,
1,
3,
8
compliant,
you
could
signal
in
the
PIO
use
compression
for
that
path
right,
so
it's
not
just
on
store
and
I
mean.
D
C
Necessarily
so
dot
no
I,
don't
think
so.
First
thing
is
by
default
today:
P
doubt
that's
what
the
duck
does.
So
if,
if
down
on
I
stand
down,
that's
true
for
died,
Opie,
doubt
everybody,
so
the
dog
is
is
doing
it
or
the
dag
is
not
doing
it
and
P
da
is
just
another
lie
inside
this
big
tag.
So
so
for
now,
because
we
didn't
specify
anything,
that's
probably
the
way
it
works.
C
It
needs
words
in
P
Dow
to
just
say
that,
but
it
would
be
the
implicit
way
now
if
we
want
to
do
something
more
fancy
inside
the
P
Dow
subtag,
not
just
narrating
from
the
underlay
but
having
its
own
capabilities.
It's
not
because
of
this
signaling
anymore.
It's
because
of
a
bit
in
the
P
Dow
telling
you,
oh
by
the
way
you
can
compress,
even
if
the
rest
of
the
tag
does
not
do
we
want
that.
Well,
you
tell
me
on
the
mailing
list:
I.
C
Hope
you
know
my
thing
is
I'm.
That's
why
what
you
said
only
I
resonated
to
me
is
like
I
really
hope
that
in
the
near
future
everybody
does
it
and
does
not
ask
the
question,
which
is
what
you
said:
the
bit
becomes
useless.
It's
kind
of
the
same
thing
for
using
23
right
I
mean
the
big
becomes
useless.
Okay,
if
the
bit
is
useless,
then
don't
spend
too
much
effort.
Doing
the
PIO
different
from
the
rest.
Just
think
that
everybody
will
use
80
once
right
very
soon
and
we
all
set.