►
From YouTube: ROLL WG Interim Meeting, 2020-04-29
Description
ROLL WG Interim Meeting, 2020-04-29
A
A
Welcome
to
the
virtual
ITF,
once
it
was
even
role
interim
meeting,
we're
happy
that
you
are
here.
Please
be
aware
that
this
meeting
is
aligned
with
not
well
pretty
carefully
they're
not
going
to
read
here
our
meeting
materials.
Please
put
your
name
and
affiliation
into
the
other
path,
is
the
link
there.
Can
we
have
the
slides
as
well
and
the
Java
Java
Michael?
Is
the
a
Pascal
seen
the
linking
to
the
mailing
list?
Sorry
for
not
green
here,
I
stroll
at
Java
dot,
ITF,
dot,
org
and
in
as
well
in
the
ether
pad.
B
A
B
A
A
So
this
is
our
agenda,
so
we
have
pass
card
first
explaining
the
updates
in
unaware
list
then
draw
all
the
updates
same
video
a
little
with
the
appropriate
triple
status
then
Pascale
talking
about
which
is
about
the
cluster
that
is
dealing
with
a
lot
of
documents.
Then
Rahul
is
going
to
explain
the
updating,
capabilities
and
more
pegs.
Then
Eris
is
going
to
play.
Explain
us
tablets
in
NC
extension,
then
try
old
is
going
to
give
out
the
rebel
ripple
of
observation.
Then
we
are
going
to
go
to
a
new
document.the.
A
A
A
A
That's
why
we
already
have
the
no
path
yeah
yo
about
the
use
of
rappelling,
for
we
are
with
the
last
updates,
and
the
Latian
39
should
be
soon
ready
for
shipping
but
yeah.
We
are
like
in
some
reviews
so
which
served
the
status
of
our
in
the
internet
job.
So
how
you
see
we
are
going
to
have
a
lot
of
of
this
today.
Discussion
was
saying:
two
of
them
are
in
the
is
G,
and
there
are
some
others
that
they
are
a
standby
from
the
related
internet
job.
A
We
are
going
to
have
two
of
them
today,
so
we
have
thank
you
very
much
to
everyone
that
open
the
tickets.
So
we
have
documents.
We
have
opened
this
because
until
the
well
finalized
I
think
it's
fine
to
have
it
open
related
to
scold.
One-30
is
already
changed
195.
So
then
we
have
some
for
ripple
observations
and
from
their
capabilities
as
well.
We
have
for
100
a
parity.
Probably
they're
fixed
now
have
to
check
that
and
then
the
double
prediction
we
as
well.
A
We
have
some
all
tickets
that
we
keep
ball
system
but
priority
into
github.
Does
it
ask
for
anyone
to
look
and
at
that
they
can
see
as
well
into
the
idea
hey
tool?
Ok,
so
we
have
kind
of
connections
for
the
documents.
So
we
have
the
ripple
observation
state
with
issues
should
be
work
on
then
capability
draft
I
have
no
connection,
would
add
and
with
different
documents,
with
a
leading
the
information
with
more
pics,
the
status
are
mutually
exclusive.
A
A
A
D
So
there
was
a
lot
of
effort
between
IETF,
106
and
IETF
107
on
the
Reaper
and
the
well
draft,
and
there
are
reasons
for
that.
One
big
reason
is
that
the
inner
wall
draft
is
the
last
draft
in
in
order
there
was
the
least
progress
draft
holding
cluster
310.
So
I
will
tell
you
later
about
clusters
returned,
and
you
see
why
the
other
thing
is.
D
There
is
actual
demand
from
the
field
about
providing
the
unaware
leaf,
because
people
sense
that
the
low-power
devices
at
the
edge
of
the
network
that
don't
participate
to
ripple
would
spend
a
lot
of
energy
if
they
would
have
to
use
ripple
so
enabling
just
the
error,
c88
505
as
the
interface
to
the
river
Rother
and
removing
ripple
from
the
edge
device
from
the
low
power
leaf.
That
is
actually
very
useful
in
the
field,
so
there
is
demand
for
that.
D
Last
but
not
least,
eight
five
four
five
is
used
as
the
host
to
router
interface
in
a
number
of
of
drafts.
Actually,
it's
just
ask
you
by
ripple
it's
with
underling,
if
it's
used
by
rift,
it's
also
used
in
the
backbone
water
draft
to
to
trigger
proxy
operation.
So
it's
a
a
useful
feature,
but
there
was
no
draft
which
really
explained
for
any
of
them
grant
details.
So
so
it's
a
good
example
for
somebody
who
wants
to
implement
the
functioning
references.
D
So
next
slide
please
so
what
went
in
well
I
should
make
the
diff
across
all
the
versions
of
the
steps
we
made
since
last
IGF.
There
are
many
many
details
and
small
changes,
so
I
didn't
go
through
any
as
the
long
list.
I
just
picked
the
most
important
changes
so
for
one
we
did
improve
quite
a
lot.
The
description
of
our
associate
505
works,
and
we
did
that
to
explain
how
that
much
is
because
it
505
is
the
host
behavior,
how
that
matches
the
router
behavior
that
we
have
in
this.
D
In
this,
there
was
also
a
lot
of
work
with
the
Ennis
and
Michael
on
use
of
ripple
info
and
all
the
flows
and
you'll
see
that
some
of
my
slides
after
this
about
that
you
know
so
we
synchronize
the
two
specification.
This
I
did
the
normative
relationship
between
the
two,
and
so
we
ended
up
with
this
duck,
I
mean
normative
reference
to
use
of
propellant
flow
and
placing
a
news
of
repo
info
of
the
text
about
external
rods
and
the
repair.
I
know
a
leaf
is
a
form
of
external
rot.
D
It's
it's
a
host
external
rod,
so
we
also
included
text
about
the
minimum
functionality
that
can.
That
needs
to
be
supported
by
the
role
to
be
able
to
qualify
as
a
role
and
be
usable
in
this
trust,
even
though
we
we
document
the
rotor
behavioral
test
young
testee
dependency
on
on
the
hospira,
in
particular,
being
a
ipv6
compliant
for
the
way
it
consumes
hub,
either
as
a
host
and
the
way
it
ignores
a
sauce
rotator
that
is
actually
consumed.
D
So
we
also
added
text
on
6ew,
since
we
use
it,
it's
just
descriptive
because
it's
the
behavior
the
background.
That
is
needed
for
this
draft,
so
we
don't
change
our
a/c
7500,
but
we
you.
We
show
how
it
plays
in
this
story
and
we
also
added
a
flag
and
that's
important
one
in
the
reaper
config,
to
option
to
be
able
to
signal
that
the
route
will
do
the
DA,
a
dark
exchange
with
the
repo
yeah.
So
you
can
actually
save
this
packet
across
the
network.
D
So
the
periodic
a
DA
attack,
extremes
that
reaches
the
state
at
the
sixth
yeah-
is
actually
prompted
by
the
route.
So
so
you
don't
have
to
packets,
the
dow
and
then
the
dart.
That
goes
all
also
in
the
network
and
that's
again
because
world
safe
power,
so
there
was
a
house
request
to
to
actually
document
the
termination
flows.
So
we
have
that,
like
I,
said
many
details,
the
next
slide,
please
so
I
don't
know.
D
If
we
have
time
to
chest,
tell
me
what
I
did
here
is
a
cross
like
the
four
five
next
slides
I
provided
illustrations
on
the
headers
that
have
to
be
installed
in
case
of
a
ripple.
I
know
a
leaf.
If
you
have
time,
if
we
have
time,
then
we
can
spend
the
five
minutes.
If
we
have,
then
please
save
that
that
can
be
useful.
D
Typically,
what
happens
is
you
have
to
encapsulate
IP
an
IP
because
the
API
and
since
we
don't
actually
insert
we
have
to
encapsulate
so
we
got
to
be
able
to
add
the
API
to
the
packet,
a
packet
coming
from
the
internet
without
this
API,
so
it's
kind
of
typical
inside
a
ripple
Network
to
do
some
or
capsulation
next
slide.
Please.
D
And
in
non
storing
the
the
we
have
more
information
to
add
in
the
outer
either
because
we
have
to
put
this
off
hot
bath
and
I've
indicated
the
case
here
of
this
node
56,
for
which
we
have
to
add
not
only
the
API,
but
this
was
route
info
next,
and
it's
just
to
be
able
to
compare
the
typical
writing
stretch
for
installing
mode
in
in
a
for
repair,
aware.
Node
would
be
that
the
packet
have
to
go
to
the
current
parent.
D
D
D
The
route
will
will
turn
on
the
packet,
so
it
will
have
to
encapsulate
the
even
if
it's
coming
from
inside
the
ripple
domain,
and
you
see
the
difference
now
we
have
this
double
encapsulation
and
I
hope
it's
clear.
Then
again,
there
is
nothing
new
to
the
real.
It's
just
a
way
of
river
works,
but
just
to
be
clear
that
that's
the
situation
we
have
today.
D
D
So
at
the
root
we
get
the
packet,
which
is
just
so
51
this
nation,
52
and
stuff,
which
is
the
content
of
the
packet,
but
the
API
is
go,
which
means
that
the
packet,
all
the
way
up
is
bigger,
but
the
packet
on
the
way
down
is
smaller.
Basically,
so
using
the
format
on
the
left
is
what
the
run
can
do
if
it
wants
to
send
a
packet
to
the
Internet,
and
it
wants
to
make
sure
that
the
API
will
not
fly
there.
D
Okay
and
then
that
goes
to
new
discussion,
and
these
are
the
slides
which
are
new
and
I
hope.
I
mean
us
check
with
me
please,
but
I
hope
I
was
correct.
So
we
again,
we
start
from
a
packet
which
is
a
generated
by
your
hole,
which
is
51,
so
51
is
not
aware
of
ripple.
So
by
definition,
there
is
no
API,
so
it
just
sends
this
package
51
to
whatever
this
nation.
D
D
If
what
the
packet,
which
is
the
root,
then
the
root
will
de
capsulate
and
that
the
thing
that
you
see
on
the
top
right
so
stick
to
our
destination
internet
stuff.
So
if
the
packet
goes
to
the
internet,
fine,
that's
exactly
what
you
expect
on
the
internet,
there's
no
API
and
if,
on
the
other
hand,
the
packet
is
for
this
nation,
which
is
like
56
on
the
bottom
right,
then
the
root
will
have
to
our
capsulate
so
wearing
in
the
gift.
D
D
So
now
if
the
destination
is
a
real,
so
the
way
up
is
exactly,
as
we
said,
I'm
not
doing
it
again,
I'm
just
focusing
on
the
on
the
rightmost
piece
now
and
they
have
two
cases
of
back
yet
and
and
the
packet
can
be
fine
coming
from
the
internet.
So
that's
the
the
C
on
the
top,
so
you
get
to
source
internet
business
in
56
and
whatever
content,
and
you
see
that
the
the
root
will
are
capsulate
destination,
the
6lr
that
serves
not
56,
because
56
is
a
row.
D
So
remember:
we
are
using
non
storage
mode
signaling
for
X
or
rods,
meaning
that
the
the
root
knows
the
parent
of
56
and
that's
46,
so
the
root
will
turn
out
to
46
2
at
the
API.
That's
true!
If
the
packet
comes
from
the
Internet,
that's
true!
If
the
package
comes
from
the
rule
and
as
we
saw
it's
also
true
if
the
packets
come
comes
from
the
rule
on
the
bottom
left,
because
it
would
have
been
encapsulated
by
4
challan
to
the
root
and
the
root
would
have
2
capsulated
it.
D
D
So
in
story
mode
it
doesn't
make
any
difference,
because
the
root
to
the
route
256
is
not
known
to
the
network.
So
following
the
default
route
to
the
command
parent,
gives
you
all
the
way
to
the
root
anyway,
so
whether
the
source
is
around
or
the
source
is
a
real,
the
packet
will
always
reach
the
root.
If
the
source
is
a
real,
then
we
saw
that
the
parent
41
would
have
a
capitulated
to
the
root
anyway.
D
But
if
the
source
is
a
run
like,
for
instance,
illustrated
on
the
left
words,
not
even
what's
illustrated
without
really
not
doesn't
don't
care,
the
packet
will
reach
the
root
anyway.
So
what
the
root
gets
is
a
packet
with
possibly
an
LP
I
if
it
was
coming
from
Iran,
but
also
otherwise
it's
a
plane
packet
to
mr.
56.
D
Sauce
rotator
to
the
parent,
as
opposed
to
using
an
AA
capsulation
to
the
product.
If
it
is
the
source
of
the
packet,
it
was
also
the
packet
is
something
else
than
that
cannot
happen,
but
I
guess,
but
I
really
wanted
to
illustrate
to
this
slide
is
the
root?
Is
the
source
it
has.
This
alternate
capability.
D
So
this
is
the
more
normal,
so
I'm
just
surprised
that
this
slide
comes
after
the
other,
because
it
should
I
think
before
so
I
just
so.
This
is
the
most
normal
case
for
non
storing
mode
to
a
rule,
so
the
packet
is
where
that's
coming
from
the
Internet
or
coming
from
the
root
in
most
cases
or
even
the
network
anyway
reaches
the
root.
D
Anyway,
the
root
the
capsulate
said
if
it
was
encapsulated
from
the
inside
or
just
supreme
packet
from
the
internet
and
the
root
will
have
two
or
capsulate
the
packet
to
the
to
the
parent,
and
since
we
are
in
on
storing,
then
that
will
be
all
the
intermediate
hops.
I
give
you
one
format,
which
is
the
black
source
source
route
destination,
46
that
it
doesn't
really
matter
if
the
inner
packet
is
the
one
from
the
road
from
the
internet
or
from
front
from
51,
which
was
a
rule.
D
In
this
case,
you
see
that
at
the
end
of
the
day,
the
route
has
this
packet,
which
is
three
levels,
but
it's
all
are
capsulated
and
the
way
it
arrives.
So
what
you
see
the
straighted
with
this
big
ladder
is
what
46
see.
So
what
46
sees
is
a
packet?
Well,
35
has
done
this
work,
so
35
is
the
last
conscious
and
consumed
on
tree
in
the
road.
Neither
46
is
the
destination,
so
46
gets
a
packet
from
which
the
external
header
is
fully
consumed.
D
External
rotting,
it
is
we
consume,
it
is
the
destination,
so
it
will
die'
capsulate
and
the
inner
pocket,
which
is
the
original
packet
realist
of
its
destination,
will
enrich
56.
So
in
this
case
you
see
that
56
gets
a
clean
packet
without
an
app.
The
only
case
right
that
would
be
an
API
is
the
case
of
a
source
inside
the
network,
which
is
a
wrap.
D
D
F
E
So
this
is
going
to
be
a
quick
update
about
the
DC
of
status.
The
document
the
draft
has
been
waiting
for
a
particular
value,
the
sparkle
status
value.
We
finally
have
it
again,
thanks
to
Donna,
believe
strap.
We
I
think
are
in
a
final
position
to
control.
This
conclude
this
chart.
Canny
next
looks
like
this,
so
so
this
is
the
final
volume
that
we
arrived
at
195.
The
reason
being
this
value
eventually
is
the
same
value
which
six
slow
as
well
as
RPL,
will
be
making
use
of.
E
So
it
makes
an
implementation
slightly
easier,
such
that,
even
if
a
RPL
anywhere
leave
or
rarp
aware
node
moves
from
one
parent
to
another.
The
move
status
would
be
same
for
both
the
cases,
so
it
actually
simplifies
not
only
the
implementation,
but
I
am
is
hoping
that
in
the
future,
this
would
aid
debugging
as
well
yeah,
so
that
this
is
the
only
update
that
the
draft
has
next
slide.
Please.
E
E
What
is
the
status
that
should
make
you,
when
the
node
at
the
success
XLR
would
node
moves
from
one
one
parent
to
another,
which
is
which
is
the
same
value
which
are
peeled,
we'll
also
be
making
use
of
four
RPO
Parent
Program
are
pure
movement
across
parents,
the
E
and
a
bit
has
been
newly
defined
and
this
definition
has
been
provided
by
another
leaf.
So
that
is,
that
is
what
the
slides
talks
about.
That's
that's
pretty
much
it.
You
know.
The
whole
point
about
waiting
for
this
valley
was
to
make
sure
that
you
know
this.
E
E
A
D
Mdrc
editor
tends
to
group
drafts
based
on
those
on
this
graph
into
clusters.
So
basically
interdependent
drafts
are
grouped
into
a
cluster
and
there
is
usually
like
one
draft
of
two
drugs
which
kind
of
are
not
yet
published
and
they
are
looking
every
other
draft
that
as
the
normative
dependency
on
them
and
the
cluster
is
kind
of
locked
until
the
the
dependent
draft
are
finished,
the
documents
were
walking
on
after
all
and
at
six
lower
60s.
They
cannot
have
this
sort
of
interdependency
yeah.
D
The
cluster
that
holds
them
is
East
Street,
damn
just
L
Street
down,
so
I
gave
you
the
link
here
or
estas
freedom.
Can
you
please
move
on
okay?
So
there
are
in
the
cluster
of
irises,
which
already
published
industry
up
here
in
310
as
published.
So
so
some
some
progress
can
be
done,
SATA
cluster
some
drafts
can
be
published
and
somebody
else
can
be
waiting,
but
there
are
other
cases
next
slide.
Please,
and
what
we
are
interested
in
is
fine.
D
I
just
give
one
example
of
a
draft
in
inside
this
cluster
and
it's
the
efficient,
then
P
Dow,
and
we
see
that
the
efficient
NP
Dow
is
not
published
as
an
RFC.
And
why?
Because
it's
ready
well
because
it
does
this
miss
fraph
that
you
see
on
the
right
missing
nomads,
that's
what
it
means
and
the
one
g1j
means
that
it's
just
one
hop,
meaning
that
the
graph
is
not
deep.
It's
just.
D
We
are
waiting
for
draft
IGF
raw
and
a
relief
to
be
published,
and
it's
not
see
and
then
immediately
and
usually
with
almost
consecutive
numbers,
the
role
and
if
an
efficient,
NP
Dow
would
be
published.
Now
you
see
that
the
draft
Ana
well,
if
he's
marked
as
not
received,
and
that's
because
the
RSC
editor
only
sees
DRC
once
it
has
passed
your
oscilloscope,
the
ITF
last
call
at
the
ILC
reviews.
So
we
just
are
making
the
the
publication
request
now
and
that's
why
the
resonator
works
it
works.
It
does
not
received.
D
So
it's
marked
as
iesg
other
draft
like
the
IP
and
ER
Marcus,
yes,
G,
because
the
users
had
to
make
some
addition
and
the
IES
GS
to
confirm
that
these
additions
are
correct
before
giving
back
to
the
RFC
editor.
So,
on
the
twist
of
pages
from
the
RFC
letter,
you'll
find
the
explanation
for
all
those
terms
like
miss
Roffe
is
G,
but
that's
basically
what
it
means.
Next.
D
Can
you
see
that
some
draft
maybe
may
have
a
high
number
of
dependencies
and
you
even
see
a
net
2g
there,
because
the
architecture
needs
a
nun,
speakin
and
then
speak
and
I?
Don't
remember
which
one
so
that's
two
degrees
now
of
debts
in
the
graph.
But
when
you
look
at
all
of
them,
there
are,
most
of
them
are
in
the
city,
talk
to
the
only
one,
the
only
ones
which
were
not
there.
I
know
a
leaf
and
myself
and
we
were
to
do
it
well,
I
guess
that
was
ahead
of
nowhere
leaf.
D
So
in
this
cluster,
all
trade
are
only
street
rods
which
were
marked
as
not
received
when
I
made
this
slide.
So
I
gave
the
status
at
the
time
I
made
the
slice.
There
was
only
one
years
missing
and
one
discussed
to
solve
I've
not
seen
that
solved
for
myself,
but
pretty
much
there,
and
the
cost
at
last
was
on
the
HGH
at
agenda
for
four
days
ago.
I
have
not
seen
the
result,
but
creature
hurry
up
for
that.
Casta
classes
is
now
down.
You
just
have
to
check.
D
D
E
I'm
going
to
talk
about
ladies
and
mopix,
so
finally,
we
had
separated
these
two
topics
into
into
two
different
drafts
capabilities
draft
had
a
major
update.
We
have
made
major
new
sections.
One
of
them
are
primary
section
was
the
recommendations
for
adding
new
capabilities.
So
what
should
anyone
who
is
trying
to
add
a
new
capability?
What
should
they
take
care
of?
You
know
there
are
several
beds
how
each
of
those
bits
have
to
be
handled.
So
that
is
what
is
been
explained
in
that
section.
E
New
capability
instances
have
been
added,
so
this
this
draft
not
only
uses
the
syntax
about
the
capabilities
but
add
specific
instances
now
which
could
be
used
for
six
low
RH.
Having
said
that,
the
capability
instances
are
not
specifically
for
successful
Harwich.
It
is
specific
to
six
flourish,
but
for
projected
now
it
can
be
made
use
of
for
any
other
any
other
any
other
purpose
as
well
and
I'll
get
to
that
in
subsequent
slides
security.
Consideration
I
think
it
still
needs
more
work.
E
E
Having
said
that,
there
is
one
primal.
What
there
is
one
point
that
I
would
like
to
discuss
with
the
working
group
which,
which
came
up
later
in
which
came
up
in
a
discussion
last
week,
anyways
that
that
that
that
problem
statement
has
been
added
to
the
observations
raft,
and
we
are
wondering
whether
that
problem
statement
can
be
taken
care
of
in
the
more
big
stuff.
But
we
will
come
to
that
point
later.
This
is
a
minimal
document
with
a
very
clear
motivation
and
proposition.
We
hope
to
take
it
very
fast
to
the
next
level.
E
E
This
is
this
is
one
of
the
one
of
the
thing
that
that
was
that
what
that
was
realized
that
we
realized
that
anytime,
we
are
adding
a
new
option.
We
have
this
backward
compatibility
issues,
so
this
is.
This
is
not
only
true
for
enrollment
priority
lighting
towers
extension,
so
what
happens
is
a
node
which
supports
the
new
option,
sends
an
option
downstream
peers
or
the
upstream
peers
who
doesn't
understand
that
option
are
going
to
strip
that
option.
So
what
happens
so?
E
So
we
are
hoping
that
we
can
handle
this
in
the
morphix,
essentially
telling
that
whenever
you
define
a
new
option,
if
the
most
significant
bit
of
that
option
type
well,
the
option
type
is
set,
it
would
mean
that
the
intermediate
node
has
to
copy
forward.
Of
course,
this
means
that
all
the
nodes
have
to
be
at
least
supporting
this
chart.
But
once
this
raft
is
supported
by
the
nodes,
then
it
would
result
much
less
of
issues
with
backward
compatibility.
Problems
is
whatever
is
we're
thinking.
E
D
How
you
agree
that
was
the
sort
of
reason
why
I
bought
it
at
some
point
to
say:
okay,
we
we
are
talking
about
332,
because
we
we
will
place
actually
requirements
on
any
repair,
v2,
node
and
being
able
to
understand
this
particular
specification
is
certainly
a
cornerstone
of
what
it
means
to
be
a
ripple
v2
node.
So
we
know
that
those
options
will
be
processed
as
expected.
So
yes,
it's
completely
important.
The
other
thing
that
we
we
usually
signal
is
what
you
do
is
an
accident
that
you
don't
understand.
Do
you
ignore
the
full
pipe?
E
E
But
one
thing
is:
we
don't
have
any
option
bags
as
such
specifically
for
every
option,
so
all
we
have
is
option
time
and
the
length
and
the
option
data
so
the
only
place
where
we
can
make
change.
We
can
depend
upon
this.
This
kind
of
depends
in
case
of
options,
is
the
type
first
view,
so
it
would
reduce
the
entropy
it
will
reduce
because.
E
But
I
got
this
point:
okay,
let's,
let's
take
this
up
on
the
mailing
list
as
well.
Thank
you.
Thank
you,
sir
next
slide,
please
so
capabilities
of
update.
One
of
the
primary
thing
that
we
wanted
to
do
before
was
compare
it
with
existing
mo
P
configuration
option
and
routing
metrics
and
constraint.
I
can
see
that
a
lot
of
people
are
getting
confused
between.
How
is
it
different
from
these
three
things?
E
So
there
is
explicit
section
added
on
what
you
should
do,
what
you
should
consider
for
defining
a
new
thing
as
the
capability,
rather
than
anything
else
which
is
mentioned
here.
There
are
some
guidelines
that
have
been
defined
for
new
capabilities:
how
to
set
the
total
flag
in
flight
join
us
Lilly
Flags.
How
should
the
node
handle
the
capability
kept
at
Pt
if
it
has
not
supported
before
or
after
joining
the
instance?
E
What
happens
if
the
node
joins
the
instance
and
later
it
finds
out
at
F
ability
which
says
that
I
am
mandatory
to
support
and
it
has
already
joined
the
extra
source
things
like
this.
The
corner
cases
have
been
handled
and
again
when
to
use
and
when
not
to
use
capabilities.
Of
course,
there
is
a
different
circuit
section
on
that.
Next,
like
this
global
and
local
capabilities.
E
So
this
is
one
thing
that
was
discussed
on
the
mailing
list
before
we
need
to
have
a
bed
which
says
that
root
has
sent
this
capability
and
the
one
should
change
it.
This
way,
the
root
will
be
able
to
say
that
I
am
supporting
so-and-so
capability
without
any
of
the
intermediate
node
touching
it.
So
there
is
a
flag
for
that.
Now,
it's
it's
a
separate
flag.
If
you
can
see
that
the
G
flag
is
then.
E
E
Next
slide
is
so
the
our
tuning
capability
is
two
types
of
capabilities.
Basically,
that
I
had
headphones
so
that
two
types
of
capabilities-
one
are
a
node
sense
that
I
support
a
certain
feature
and
other
type
of
capability
is
that
I
support
a
feature
with
so-and-so
parameters
or
so-and-so
kept
a.
There
is
an
additional
information
so
far
for
any
capability
with
supports
and
additional
information.
We
already
have
a
definition
for
that.
E
We're
in
a
capability
can
have
additional
capability
information,
but
for
all
the
capabilities
who
require
who
are
justin
support
or
does
not
support
a
flag
like
indicator.
We
are
trying
to
put
all
those
together
in
a
single
option
in
a
single
type.
What
it
allows
us
to
do
is
it
allows
us
to
retain
the
control
overhead,
because
if
you
keep
having
a
separate
option
capability
at
time
for
each
of
these
indicators,
then
it
would
unnecessarily
increase
the
control
over
it.
For
this
message
is
carrying
this
capabilities,
so
we
call
it.
This
capability
indicator
screw
again.
E
B
H
E
E
So
we
are
hoping
that
we
should
be
able
to
reuse
one
two
three
or
any
value
in
the
future
right
now,
the
draft
seek
mandates
into
I
mean
the
right
now.
The
draft
says
that
it
is
length
equal
to
three
but,
for
example,
the
example
teen
the
six
large
flag,
which
is
specified
here
at
t.
It
shows
that
the
leftmost
the
rightmost
bit
is
set
I,
don't
know
how
to
put
it
too.
How
can
I
phrase
it
properly?
What
I
want
to
send?
It
says
it
is
the
location
of
the
pet
in
that
area.
E
G
E
E
G
E
E
Next
week,
okay,
so
this
is
another
capability
that
has
been
added,
so
this
is
a
different
capability.
We
want
to
read.
The
note
should
be
able
to
set
rerouted
resource
capability.
It
has
how?
What
is
the
total
capacity
it
has
again?
The
capability
length
here
is
is
is
flexible,
so
if,
at
some
point
of
time,
if
it
goes
beyond
256,
they
can
simply
used
kept
a
little
length
of
2
and
half
65536
has
the
maximum
value
for
a
route.
If
this
was
capability,
this
capability
will
specifically
be
useful
for
projection.
E
Tau
is
what
we
are
thinking.
We
are
wondering
if
we
should
do
something
similar
for
neighbor
cache
as
well,
and
if
we
want
to
do
it
for
neighbor
cache
as
well,
shall
we
nee
if
it
is
the
same
same
capability
or
have
a
separate
one,
I
think
routing
house
routing
resources
more
a
more
more
required
by
P
now,
but
maybe
the
neighbor
cache
might
also
have
an
implication
on
on
the
route
segment
selection
for
prop
relation
to
project
to
town.
So.
I
It's
been
sometime
in
the
meantime
about
four
versions
have
have
passed.
We
have
addressed
a
lot
of
feedback
from
Dominique
Rahul,
Fabrice
Pascal
in
Diego,
I,
think
you're.
So
very
much
so
as
part
of
addressing
of
addressing
this
feedback,
we
did
a
lot
of
editorial
work.
We
fix
typos,
we
rephrase
some
sentences
and
we
find
some
part
wasn't
clearly
defined
and
one
of
the
things
we
also
did
was
reorder
the
structure
of
the
paper.
I
So
initially
we
were
describing
the
objective
function
that
is
contained
in
this
word
and
afterwards
we
gave
some
variations
on
how
the
object
function
work
and
now
we
reversed
it.
So
we
give
more
examples
up
front
and
then
we
go
into
the
details
of
the
specification
next
slide,
please,
okay,
so
one
the
first
changes
is
regarding
the
objective
function
described
in
this.
We
describe
an
objective
function
based
on
the
common
ancestor
in
order
to
mark
a
replication
and
elimination,
and
initially
we
have
defined
three
objective
functions
for
three
different
variations
of
this.
I
I
We
also
explained
how
these
policies,
the
three
policies
that
we
describe,
how
they
are
ordered
in
terms
of
how
restrictive
they
are.
So
we
start
from
the
most
restrictive
one
going
to
the
most
relaxed
one,
and
we
also
explained
that
there's
a
trade-off
in
terms
of
energy
consumption
and
reliably
to
the
strictest
one,
consumes
less
energy,
but
provides
lower
reliability
and
the
more
relaxed
one
consumes
more
energy,
but
is
a
less
reliable.
I
One
of
the
changes
we
also
did
is
regarding
this
change
from
multiple
objective
functions,
just
one.
So
these
objective
functions
are
local
to
each
node.
So
the
only
four
they
note
the
decision
is
made
at
the
node
based
on
information
received
from
other
nodes,
but
the
decision
at
the
end
of
the
day
only
regard
this
node.
It
decides
how
to
work,
replicate
a
packet,
so
there's
no
need
for
multiple
objective
functions.
Each
node
can
make
the
decision
in
their
own
way.
I
I
We
also
provided
more
details
about
the
information
that
is
required
for
this
objective
function
to
work.
So
we
extend
the
dieö
control
message
and
we
put
some
extra
information
in
the
da
G
metric
container
in
in
the
part
of
the
node
state
attributes
object,
so
we
define
a
new
TLV
which
is
inserted
there
and
before
it
was
a
bit
low
on
details.
We
have
not
specified
exactly
how
the
addresses
that
are
card
inside
are
represented,
how
they're
stored
in
the
field.
How
many
can
derive,
how
many
addresses
are
present
and
what
their
order
signifies.
I
I
Thank
you
for
that
by
the
way
was
that
we
identified
that
we
initially
described
this
information
in
the
NSA
object
as
a
constraint.
We
use
this
information
in
order
to
restrict
which
nodes
can
be
used
for
replication.
However,
Dominic
pointed
out
that
if
you
have
a
constraint
in
the
I/o
racket,
you
also
need
to
have
a
corresponding
metric
so
because
there
isn't,
as
make
sense,
also
have
a
metric.
I
We
changed
the
definition
from
a
constraint
to
a
metric,
and
we
also
set
the
flags
for
this
method
to
be
carried
in
recording
mode
and
with
the
partial
flag
enabled,
because
this
is
information
sent
from
a
parent
with
children
doesn't
go
further
away.
So
it's
always
partial
information.
You
don't
get
the
information
all
the
way
from
the
root
of
the.
I
Finally,
we
also
highlight
we
clarify
that
this
metric
is
not
actually
used
to
modify
the
rank
of
a
parent
of
what
rank
a
node
would
get
through
a
specific
one,
so
it
own
it's
only
used
to
filter
out
potential
parents,
nothing
else.
It
basically
operates
in
a
way
as
a
constraint,
but
it
card
as
a
metric
next
slide.
Please.
I
We
also
described
finally
some
considerations
regarding
the
security
aspects
of
draft,
so
basically
the
main
addition
is
that
carry
inside
the
DAO,
the
addresses
of
a
node
neighbors,
so
a
node
can
potentially
have
a
small
privacy.
Small
network
discovery
issue,
so
a
malicious
node
can
see
one
hop
beyond
its
immediate
neighbourhood,
which
it
is
an
issue.
But
for
me
it's
not
like
super
bad
issue,
but
we
did
highlight
is
in
the
draft
and,
secondly,
potential
problem
with
rerouting
so
a
malicious.
I
The
sender
can
modify
what
the
parent
that
it
reports
and
that
way
it
can
take
other
nodes
into
routing
through
itself
and
taking
over
the
routing.
However,
this
is
already
possible
by
just
reporting
a
fake
rank,
so
this
is
not
really
a
problem.
At
least
it's
it's
no
more
problem
than
it
already
is
in
these
terms
next
slide-
and
this
is
basically,
we
have
tried
to
address
every
issue
that
was
raised.
The
some
issues
will
require
discussion.
We
did
it
in
the
mailing
list.
I
B
One
point
that
last
meeting
we
discussed,
we
want
it
to
compress
those
addresses,
because
we
know
you
know
this
is
going
to
carry
ipv6
addresses
so
I
think
with
my
cell
phone
Pascal,
we
said:
let's
shape
this
draft
first
and
then
see
how
we
do
the
compression.
We
just
need
to
make
sure
that
we
have
a
way
forward,
that
we.
B
I
So
actually
I
don't
know
if
you
remember,
but
initially
we
did
implement
compression
right
in
our
draft.
So
we
took
the
example
provided
by
six
Laura,
the
the
source,
routing
header,
if
I
remember,
correct,
Michael
proposed
that
and
we
implemented
that
for
the
set
of
parent
addresses
of
that
world.
We
actually
have
code
of
that,
but
after
discussion
with
Pascal,
he
said
that
an
idea
would
be
to
have
a
separate
draft
that
just
handles
compression
for
all.
I
B
B
I
I
would
assume
that
the
fact
that
you
can
deduce
both,
which
are
the
addresses
in
the
field
and
how
many
there
are
just
from
the
size
is
enough.
This
is
a
very
homogenous
formality
of
the
type
your
length
and
just
a
multiple
of
16
bytes,
one
for
each
56
address.
Now,
when
you
do
six
lower
compression.
Typically,
you
you
use
the
same
compression
level
for
each
IP
address
in
a
set
of
IP
addresses.
I
So
if
you
let's
say
compress,
it
shift
invite
others
to
8
bytes,
then
all
of
them
in
a
set
will
have
the
same
compression
level.
If
you
want
different
compression
levels
because
say
one
is
the
root,
for
example,
and
the
others
are
not
so
the
root
can
be
compressed
further
than
the
other.
Then
you
need
great
to
set
of,
let's
say
compression
sets,
but
I
don't
see
why
this
would
affect
the
form,
the
uncompressed
format,
in
any
way,
you
still
maintain
the
same
order,
so
I
guess
this
is
completely
up
to
the
compressor
to
solve.
I
B
I
So,
ok,
that's
the
thing
if
you
want
to
be
able
to
compress
so
my
thinking
for
this
was
that
you
either
do
compression
or
you
don't,
and
if
you
do
compression,
then
the
compressor
does
need
to
figure
out
the
format
of
signifying
which
part
of
the
control
packet
it
has
compressed
or
not,
and
that
would
go
directly
into
the
work
of
the
compressor.
It
wouldn't
have
anything
to
do
the
uncompressed
format.
If
you
would
like
to
do
a
piecemeal
part,
so
some
parts
of
the
control
packet
are
compressed.
I
E
Have
an
opinion
about
the
compression
part,
so
I
thought
what
we
discussed
previously
was
that
we
should
have
a
randomized
compression
which
would
work
across
not
a
minimal
set
of
analysis,
but
at
the
set
of
addresses
as
well,
for
example,
multiple
targets
and
that
loud
message,
I,
don't
know
whether
the
same
should
be
used
or
whether
it
cannot
be
used
regardless.
It
would
be
really
this.
These
addresses
are
sent
as
part
of
the
atomization.
The
dianna
message
is
a
multicast
message.
We
talk,
we
know
that
we
we
cannot
afford
to.
E
You
know
predict
any
further,
so
it
would
be
really
nice
to
have,
in
my
opinion
again,
I
know.
We
have
already
discussed
this
on
the
mailing
list
and
we
have
already
decided
not
to
compression.
In
fact,
it
is
my
personal
opinion
that
you
know
compression
is
a
must
for
this
because
it
without
it
might
not
be
feasible.
In
my
opinion,
you
know.
So,
even
if
I
have
two
addresses,
it
is
like
32
bytes,
and
it's
it's
to
my
party,
a
message.
So
what
I've
tried
to
say
it?
E
B
G
I
did
some
work
on
this.
In
the
early
days
of
that
before
65
50
was
shipped,
there
was
an
appeal
to
rip
out
all
the
options,
in
fact
to
rip
out
all
the
TLV
processing
and
basically
have
a
static
header
that
would
contain
all
the
important
information,
so
he'd
never
extend
right
and
I
thought
that
was
dumb
and
so
I
wrote
about
about
20
lines
of
C
code
that
implemented
generalized
header
compression.
G
That's
what
I
really
think
we
should
do,
because
that
would
compress
any
ipv6
address
in
any
option
wherever
we
put
it
fairly.
Fairly
ease
well,
didn't
use
a
dictionary
and
so
stateless,
but
we
could
do
better.
I
mean
there's
ways
of
doing
better
and
GHC
lets
us
do
better.
Incrementally,
like
we
can
build
it's
up
to
the
compressor
to
decide
how
smart
he's
going
to
be.
G
I
Think
this
also
gives
rise
to
opportunities
for
further
compression.
For
example,
you
have
the
maybe
have
a
target
option
as
well,
but
if
you
have
the
same
ipv6
address
in
our
extension
in
the
target,
for
example,
then
you
can
eliminate
it
from
both
places.
If
you
have
like
a
global
compressor,
if
you
just
have
specific
compressors
for
each
small
part,
then
you
can
take
advantage
of
things
like
this.
So
yeah.
There.
D
D
There
are
two
other
angles
that
that
we
need
to
keep
in
mind,
so
Jason
may
not
be
enabled
on
every
node
and
that
what
preclude
those
nodes
from
participating
to
repo.
So
if
we
really
think
JC's
direction,
maybe
it
should
be
one
of
those
things
that
must
be
present
in
every
node
in
v2.
So
that's
ripple,
v2.
That's
one
thing
to
keep
in
mind
the
compression
itself.
You
look
at
projection
I'm,
not
using
the
full
traces.
We
had
that
discussion
long
time
ago
as
well
and
since,
in
the
meantime
we
made
eight
one
three
eight.
D
So
we
have
a
compression
for
ripple
data
packet
for
photos,
source
routing
interferences,
so
the
whole
projection,
in
particular,
when
it
chooses
source
routing,
it
made
sense
to
use
the
exact
same
compression.
It's
a
really
simple
compression
right,
mostly
lines
to
zeros.
You
know
so
virtually
sighs.
What
is
the
significant
part
and
what
is
narrated
from
a
reference?
If
you
have
a
reference
which
has
like
in
bytes,
which
are
always
the
same,
then
you
just
have
to
write
two
bytes
per
address
in
the
compressed
form.
Exactly.
D
What's
important.
Is
you
don't
necessarily
have
to
place
all
the
options?
So
if
you,
you
may
have
to
put
some
2d
I
use
in
a
row,
and
that
has
to
be
very
very
welcome.
Consider
the
options
that
monetary
or
some
time,
because
imagine
that
a
guy
gets
did
one
year
you
and
that
the
other,
the
military
options
are
gets
lost.
So
that's
something
to
consider
in
the
capability
the
map
X
drafted.
D
K
B
I
A
Thank
you
very
much.
Brown.
E
A
E
Observations
again,
so
we
have
an
update
next
next,
please
so
you've
been
had
the
discussion
about
adding
a
section
about
triple
time,
what
recent
edits
and
again,
if
it's
there
now
again,
the
backward
compatibility
issue
that
I
mentioned
before
is
mentioned
as
a
problem
statement
in
this
draft,
the
same
problem
that
I
mentioned,
we
add
a
new
option.
How
do
we
ensure
the
backward
compatibility
is
the
same
thing
that
is
mentioned
in
next
site
this,
so
one
thing
one
of
the
major,
so
we
have
kept
discussing
about
tower
so
many
times
before,
and
we
finally.
H
E
To
do
something
about
it,
we
have
a
new
draft
which
takes
which
sort
of
tries
to
handle
the
problem
statements
towards
this
this
section,
so
just
to
briefly
in
turn
what
the
working
group
it
is
about.
I
know.
We
have
already
discussed
this
a
lot
of
times
before
Tarkus
is
a
local
signaling
in
case
of
storing
mode
of
operation.
E
Get
the
primary
problem
statement.
Is
that
a
tiger
not
aware
of
the
into
it
past
establishments
whether
to
start
an
application
with
her
to
start
an
application
traffic?
It's
that
the
target
needs
to
know
when
should
get
to
it,
if
not
knowing
it
in
case
of
multiple
hops.
If
the
node
is
by
hop
steep-
and
it
really
has
major
implications
on
the.
E
Another
problem
I
think
we
don't
need
to
go
here,
but
the
six
enter
the
its
if
an
intermediate
six
and
now
for
some
reason,
indicate
induces
and
negative
now
status.
How
should
it
be
handle
it?
One
of
the
point
was
that
node
that
the
child
know
who
has
generated
the
Tao
packet
has
to
in
turn
take
care
of
the
handling,
but
there
are
some
issues
with
it
and
the
observations
Tao
draft
has
clarified
it.
This
was
discussed
again
you
the
previous
session,
so
I
won't
go
into
the
details.
E
That
I
would
like
to
present
the
solution.
Next
slide.
Please
so:
okay,
forget
the
right
hand,
side
graph
for
a
moment,
just
concentrate
on
the
left
hand
side.
So
what
what
we
are
suggesting
is
the
root
root
node,
since
my
cat
now
act
directly
to
the
node
the
target
load
with
the
TI
option,
which
has
the
path
sequence
so
in
this
way
in
this
way,
basically,
the
root
node
can
inform
the
target
node
whatever
to
whatever,
where
whether
the
end-to-end
path
has
been
established.
There.
E
This
and
I'll
come
to
those
points
in
the
subsequent
slides,
so
this
is
this
is
the
fundamental
position
that
in
place
of
so
we
are
not
replicating
the
existing
tower
act
mechanism
at
all.
It
is
possible
that
the
Hopis
have
existing
open
out
our
magma
charism
is
implemented
at
the
same
time.
This
did
this.
This
mechanism
is
also
implemented.
The
only
difference
between
a
half
and
this
tower
work
is
that
No
we'll
send
back
the
transit
information
option
to
the
target
Road.
E
Why
transit
information
option
has
to
be
sent
back
to
the
front
to
the
target
node?
It's
because
the
target
node
needs
to
know
for
which
tau
is
this
tower
sent
for
every
every
target
has
is
attributed
by
a
part
sequence
which
is
present
in
the
transparent
information
option.
So
this
graph
basically
shows
the
different
attributes
which
have
to
be
passed.
So
one
of
the
things
that
it
that
this
mechanism
also
enables
is
that
the
border
router
can
actually
the
target
know
as
to
what
kind
of
route
lifetime
it
should
use.
E
For
example,
the
node
decided
to
use
the
target
node
decided
to
use
the
router
iframe
of
3600
seconds
the.
Finally,
the
Dow
reached
the
border
router,
the
router,
the
power
the
route
might
say
that,
okay,
you
can't
use
3600
life
time.
You
have
to
use
1800
seconds
now.
This
is
possible.
This
was
not
possible
before
so.
There
are
some
side
effects
positive
side
effects
to
promoting
such
a
coup.
In
such
a
kind
of
do.
D
D
D
So
this
is
not
needed.
I
work
for
the
periodic
down
the
periodic
down
could
be
set
by
every
note
on
the
way
aggregates
it
on
their
own
time,
and
since
the
path
is
already
enabled,
the
Refresh
doesn't
act
very
precisely
when
and
21
cents
the
dowry.
So
these
down
work
could
wait
a
long
time.
It
does
not
matter
when
you
just
refresh
as
long
as
everything
is
done
with
it
much
time
so
absolutely.
D
E
So
that's
cool!
That's
that's!
That's!
That's
a
very
important
point
that
you
mentioned
so
essentially.
The
first
tau
needs
to
be
taken
care
off
with
the
start
with
the
signaling,
even
to
call
it
a
Dow
ad.
We
also
thought
about
it.
You
know
I
was
thinking
about
it.
Whether
I
should
be
calling
it
a
tower
because
it's
it's,
it's
basically
an
asynchronous
message
which
gets
sent
to
the
target.
So
that
is
originally
your
reachable.
E
So
it's
we
should
not
even
be
calling
a
tower,
but
for
the
sake
of
this
discussion,
let's
call
it
tau
I
am
open
to
changing
this.
This
term
int,
you
know
this
is
mean
I,
think
we
should
change
it.
Essentially.
What
it
will
allow
us
to
do
is
this.
A
synchronous
message
can
take
care
of
certain
other
problems
as
well
in
the
future,
because
this
is
the
first
time
in
storing
a
movie.
We
have
a
message
from
the
route
directly
sent
to
the
target
target.
E
Your
second
question,
the
Refresh
Tao.
So
basically
that
is
what
that
is
what
I
was
coming
to.
So
the
the
only
change,
the
only
change
that
is
required
in
the
transit
information
option
is
that
the
setting
of
this
game
flag,
which
says
that
acknowledgement
is
sought
or
some
sort
of
responses,
source
or
sort
you
are
reachable
to
sponsor.
So
now,
when
I'm
in
typical
6lr
sends
a
Tau
or
refresh
tau,
it
won't
set
this
key
flag
or
it
should
not
set
this
gay
flag.
It
does
not
make
sense
for
it
to
set
the
scale
flag.
E
H
E
L
E
Target
node
thinks
that
it
needs
to
know,
then
it
can
set
the
k
flag.
So
what
I'm
trying
to
say
is
that
the
target
no
use
in
charge
of
setting
this
flag
and
letting
itself
know
whether
the
complete
end
to
enter
outfit
routers
will
set
or
not.
So
in
this
example,
if
n
2
n
2
1
decided
to
switch
to
N
2
2,
it
can
set
the
Cape
like
a
key
or
not
set
the
Cape
racket.
I
mean
it's
completely.
E
D
Concert
in
piggybacking
this
information
it
now
because
the
Dow
is
maybe,
as
you
said,
the
first
down
is
just
copied
us
received
all
the
way
up,
but
normally
the
dowels
can
be
completely
regenerated
by
the
intermediate
nodes.
It
all
has
to
be
stored
and
forwarded,
meaning
that
should
we
store
this
k
flag
and
then,
if
we
should
be
put
it
on
every
copy
we
set
up
until
we
get
down
from
below
which
doesn't
have
you
know
we
can
at
some
point
there
are
cases
where,
by
changing
the.
D
But
it
is
flag
in
the
Jo.
The
parent
can
ask
the
child
to
give
all
the
list
of
all
the
dowels
that
hands
so
basically
empty
its
its.
What,
if
all
the
rods
that
it
has
just
to
to
resync
so
should
should
keep
like
this
be
used
in
this
case,
you
know,
is
it
something
that
that
nodes
need
to
learn
and
remember
when
they
reproduce
the
tau,
so.
E
D
D
So
I
know
that
somehow
it
cannot
arrive
to
the
route
before
the
Dow
arrives
to
the
road,
because
how
the
route
send
the
packet
back.
On
the
other
hand,
it's
hard
to
signal
it
in
the
dow,
unless
we
say
a
if
you
thought
the
dow
immediately,
then
you
leave
take
a
bite.
They
came
yet.
But
if
you
store
it
in
your
database
in
your
logical
database,
read,
then
you
don't
store
the
key
flag
or
something
it's
not
something
that
you
can
reproduce
a
synchronously.
D
So
so
it's
it's
kind
of
weird
for
me
to
use
that
it's
not
like
there
is
a
design
self.
What's
the
best
way,
what
I
was
thinking
about?
Exactly
that
would
a
student
at
port
earlier
and
and
my
conclusion
was
there-
should
be
a
setting
in
the
route.
Basically
that
says
the
setting.
Is
there
each
time
you
get
a
new
route
like
in
20
mm
salsa?
D
Damn
you
said
if
I
get
to
20
21,
but
the
prime
is
if
in
21
moves
route,
may
not
see
anything
new
and
maybe
and
21
wants
that
information.
So
it's
it's
kind
of
you
know.
Basically
the
way
I
see
it
is
that
could
be
a
toggle
setting
in
the
route
which
says
on
the
first
appearance
of
an
21.
You
saw
the
synchronously.
This
message
says:
okay,
I
know
you're
here
and
I
synchronously
when
in
21
moves
it
may
ask
reachable
unicast
to
the
root
and
the
root
outside
unicast.
D
D
G
D
The
ripple
ping
is
the
first
thing
right.
What
triggers
the
first
thing,
because
you
can
21
cents
a
dollar
on
the
ping,
maybe
that
the
ping
passes
goes
faster
than
the
Dow.
Another
roots
receives
the
first
thing,
but
does
not
so
what
does
it
do?
So?
That's
why
I
mean
I,
guess
it
was
piggybacked
inside
the
Dow
kind
of
to
say
a
I.
G
That's
that's
the
that's.
The
oh
you're
new,
welcome
to
the
welcome
to
the
thing
and
actually
that's
kind
of
opportunity
actually
even
to
do
some
capabilities
exchange.
What
are
you
able
to
do
then?
Afterwards,
then?
Maybe
there's
something
in
there
and
that
thing
that
enables
dates
later
on
where
the
route
says.
Oh
and
who's,
your
parent.
Today
and
it's
just
it's
not
a
it's
a
survey,
it's
not
an
exhaustive
survey,
it's
just
a
regular
survey
and
we
talked
about
this
in
capabilities
that
you
know
you
might
have
to
do
that.
D
So
well,
yeah,
yes,
you
are
establishing
a
connection,
connection
kind
of
between
an
21
and
the
route.
The
first
time
and
one
shows
up
the
route,
tells
a
based
on
the
configuration
that
it
has.
It
should
basically
tell
and
21a
I
know
you're
here
and
when
n
21
moves,
it
can
ping
the
route
and
check
that
the
movement
was
actually
done.
E
Next
slide
things
I
mean
I
will
move
faster,
so
a
lot
of
points
to
be
so.
The
lot
of
points
Frank
is
speaking.
All
these
are
considered
already.
I
hope
we
can
get
some
reviews
and
one
of
the
primary
point
that
Pascal
mentioned
I
think
it
Baskerville
Michael
mentioned.
Michaels
also
has
mentioned
that
point
of
all
using
Bing
conform
like
reckon
ISM.
So
one
thing
you
change
the
name
of
the
message
from
the
Hyksos
difference.
E
Okay,
next
right,
so
I
wanted
to
discuss
about
other
things
as
well.
It
looks
like
we
don't
have
time
I
wanted
to
understand.
How
do
we
handle
the
observation
start
because
to
that
there
is
some
other
work
that
so
all
the
other
points
also,
since
we
are
not
going
to
take
it,
for
we
are
not
going
to
publish
ripple
observation
there's
some
other
aspects.
Also
lot
of
aspects
does
not
require
the
whole
new
draft,
but
it
is
still
not
an
errata
so
how
to
handle
those
things.
A
K
K
K
K
If
the
meter
will
send
the
D
is
white,
it
will
send
in
multicast
because
has
no
information
about
the
network
is
running.
So
in
this
case
we
will
have
the
note
but
will
receive
this.
D
is
a
multicast
to
reset
the
timers,
and
thus
we
will
have
bunch
of
packets.
Back-To-Back
thing
is
that
this
D?
Are
you
responses
with
being
multicast
which
will
trigger
to
consume
more
energy?
In
this
note,.
K
India
is
notification.
Draft
a
potential
solution
would
be
to
add
few
flux
and
options
inside
the
ideas,
control,
packets,
so,
for
instance,
with
no
inconsistency
flag
receiving
notes
will
not
reset
the
timers
with
the
I
you
type,
the
notes
will
send
back
in
unicast
at
the
I/o
response
spreading
option.
K
The
nose
but
received
e
is
really
not
response
at
the
same
time
and
thus
will
avoid
the
collisions
by
sending,
at
the
same
time,
the
idea.
Packets,
yes,
control
packets,
furthermore,
may
include
metal
container,
we
list
rata
constraints
and
then
the
nodes
that
receive
this
DRS
would
respond
only
if
they
satisfy
these
constraints.
K
K
K
Tackle
with
this
problem
again
di
di
yes,
missus
can
be
sent
with
flag,
such
as
inconsistency.
No
inconsistency
in
order
to
not
reset
the
Terkel
timers
from
the
other
notes
include
the
solicited
information
option
with
I
Indy
flux.
In
order
to
identify
the
dog
respond
spreading
option
in
order
to
specify
the
suitable
time
interval
over
which
the
DAO
responses
may
arrive.
K
So
the
next
slide
comes
from
the
draft
of
role
of
ripple
observations.
The
adjustments
is
probing
in
order
to
reduces
the
control
traffic
overhead
ripple
with
the
triple
time,
an
algorithm
to
update
the
configuration
parameters,
but
the
problem
is
that
when
we
do
not
have
actual
traffic,
regular
traffic
or
linear
to
traffic
its
operations,
the
enhance,
because
that,
since
it
cannot
be
tasted
and
therefore
cannot
be
repaired,
if
they're
broken
now,
6550
comes
with
mechanism
in
the
fog
of
unicast
DNS
to
trigger
vio
from
the
other
node.
K
K
K
After
a
few
slides
to
show
you
some
preliminary
results
that
we
got
by
implementing
the
ideas
fortification
and
to
see
the
benefits
we
are
working
on
this
with
iris
and
our
new
intern
Kentucky
energy,
we
have
a
network
small
network
of
10
nodes
in
cryptology,
with
a
ripple
in
60
minimum
and
belayer
talk
next
slide.
Please.
K
So
here
we
have
the
default
version
of
6550.
There
are
no
flex
and
no
options
and,
as
you
can
see
in
the
note,
6
is
the
one
that
is
willing
to
join
the
network,
but
is
already
running
the
background
of
the
node
6
in
great
color.
It
means
that,
because
no
radio,
so
it's
radio
is
off,
and
so
it
means
that
it
got
a
parent
around
at
30
minutes
the
node
6
it's
reveal
and
since
DIDS
multicast
control
packet
according
to
6550
the.
K
What
do
you
see
here
it
will
happen,
basically
means
that
all
neighborhood
nodes
will
respond
with
the
I/o
back-to-back
in
broadcast
and
they
reset
the
typical
timer
and
that's
why
you
see
this
burst
of
di
us
want
next
to
each
other.
Now,
if
you
go
to
the
next
slide
by
including
and
I
will
go
quickly
here
by
having,
for
instance,
the
N
flag,
only
option
note
that
will
receive
this
das,
the
multicast
is
control
packet.
They
will
not
reset
the
typical
timer,
so
they
will
send
only
one
here.
K
K
K
K
K
So
this
is
the
fun
with
having
the
two
flux
and
the
two
options.
Basically,
we
have
n
flock
for
not
present
resetting
the
timer.
We
have
the
T
flag
in
other,
the
devices
will
transmit
back
in
unicast.
So
this
is
the
green
color
that
you
see
in
X.
We
do
have
the
matter
container.
That
include
constraints.
Basically,
it
says,
for
instance,
if
the
rank
node
receives
this
D
is,
is
higher
than
x
and
y
value.
K
If
it
satisfies,
then,
yes,
you
can
send
back
your
Dao
and
rs
option
in
order
to
give
some
spreading
so
that
we
do
not
receive
these
responses
at
the
same
time.
So
we
do
not
have
a
collision.
So,
as
you
can
see
in
the
bottom
figure
the
only
four
green
responses,
so
basically
we
reduce
by
self
the
nodes
that
will
respond.
They
do
respond
in
unicast
because
we
have
only
the
green
colors
do
not
send
at
the
same
time.
So
we
do
have.
We
do
avoid
the
collisions.
K
K
A
E
B
E
B
B
D
We
we
had
this
a
number
of
discussions
for
the
DS
magnification
and
you
know
I'm
a
big
supporter
of
this
work.
I
think
they
just
gave
the
important
in
one
use
case.
That
really
hurts
us
in
the
past
in
the
field
is
the
use
case
where
you
have
a
power
down.
No,
we
have
Smart
Grid
applications
of
repo.
So
we
have
this.
D
This
area,
which
is
power
down
like
it's
got
and
then
all
of
a
sudden
there
is
a
power
restore
and
so
all
the
matters
of
Isis
they
cannot
on
at
the
very
same
time
and
that
interesting
use
case,
because
if
we
just
do
the
disk
in
a
fashion
that
the
devices
don't
know
that
the
other
devices
are
suffering
from
the
same
thing,
then
it
might
actually
be
sending
a
lot
of
traffic
and
dystrophy,
which
would
possibly
even
pretty
much
the
reception
of
the
iOS.
And
in
that
case
that's
exactly
the
opposite.
D
B
B
D
D
B
G
Good
gun
here
can
good
alright,
so
III
made
the
slide.
I
realized
the
title
of
the
of
the
document.
Maybe
is
a
little
bit
wordy
and
welcome
changes
to
that
next
slide
please.
G
So
we
split
this
document
off
from
six
dish,
enhanced
beacon
in
2016,
and
it
will
have
the
name
join
in
the
title,
which
was
confusing,
because
people
thought
it
was
about
doe
tag
joining
and
so
as
much
consistently
as
possible.
It's
now
says
enroll
or
enrollment
rather
than
join.
It
says
there's
no
to
but
I
meant
over
three
and
I've
removed
all
the
remaining
references
that
were
joined,
except
for
joint
proxy
from
the
document.
To
be
clear,
this
is
not
about
joining
a
finding.
A
parent
next
slide.
G
G
So
in
this
case
note,
13
is
gonna,
sending
an
enhanced
beacon,
we'll
probably
all
of
them
aren't
sending
beacons,
but
this
one's
only
one
we
care
about-
and
let's
say
it
contains,
the
enhanced
beacon
contains
the
value
22,
which
is
what
it
would
have
broadcast
it
had
it
been
or
23
what
it
broadcast
had
it
been
there
next
slide.
So
in
the
case
of
a
deeper
one,
it
continues
down
next
slide.
The
color
gets
darker
as
the
number
gets
higher
and
what
happens
is
at
the
end.
You
get
a
number
in
this
case.
G
7251
sees
this
value
72
next
slide,
and
it
goes
another
track
next
slide
I'm.
So
in
this
case,
32
is
discovered
that
he's
full.
He
has
no.
Maybe
he
has
no
more
neighbor
entries
or
his
storing
node
and
the
the
routing
tables
full
exact
reasons
are
not
important
or
specified,
but
the
important
thing
is
that
he
sends
he's
full.
G
So
he
sends
us
127,
7f
value
down
to
42
next
slide,
and
so
42
still
sends
his
enhanced
beacon,
but
he
sends
it
was
the
7f
which
tells
51
and
52
that
this
is
really
a
bad
place
to
join
or
enroll
next
slide.
So
it
comes
down
33
and
a
lower
value.
Next
slide.
Next
slide
and
43
announces,
there's
much
better
value
of
52,
and
so
coincidentally,
node
52
is
going
to
enroll
to
his
right
to
a
node
43
rather
than
to
his
left
to
node
42
next
slide.
G
They
changed.
Oh,
it's
an
impaired
note,
okay,
so
in
this
case
we're
gonna
have
an
impaired
node,
which
is
node
number
24.
Okay,
so
that's
the
one
two
layers
down,
so
it
receives
the
23,
but
it
doesn't
know
what
to
do
with
it,
so
it
deletes
it.
The
next
slide
does
not
support
this
option
next
slide
and
so
number
35
when
it
receives
the
D
IO,
and
it
doesn't
see
this
value,
it
assumes
the
value
of
both
of
40
of
so
four
zero.
So
that's
a
64.
G
So
that's
the
halfway
value
between
zero
and
one
21
27,
and
this
number
is
certainly
open
to
discussion
next
slide.
So
it
adds
some
value
to
this,
based
on
its
own.
Local
condition
announces
that
next
slide
46
happens
to
announce
another
value
51,
and
so
that
means
there's
still
a
value
still
has
something
to
say,
but
it
may
be
inaccurate.
There
may
be
no
real
connection
between
35
and
the
conditions
at
24.
24
could
be
completely
full.
We
don't
know
next
slide
and
obviously
there's
other
announcements
there
of
a
another
value.
G
Next
slide,
that's
really
yet,
as
I
said,
I
added
the
section
on
what
to
do.
If
there's
no
value
and
there's
some
discussion
about
what
would
happen
if
the
option
wasn't
present
and
if
the
option
wasn't
present,
then
we
would
still
have
the
problem
at
24
that
it
was
suggested
and
no
one
knew.
But
at
least
we
have
a
little
bit
more
management
in
the
sub
dodeca
of
what's
happening
and,
as
I
said,
maybe
there's
some
other
better
default
value.
We
can
pick.
A
D
C
D
Is
for
some
of
you
it's
a
new
work
for
some
of
you,
it's
more
another
work.
So
this
is
a
draft
it
we
kind
of
decided
collectively
to
do
during
one
of
the
role
in
term
eating's.
What
we
saw
is
that
there
is
more
duration,
information
that
is
in
exchange
within
the
repo
protocol
and
that's
very
important
to
us,
because
we
care
about
what
we
call
autonomic
properties,
meaning
that
the
device
can
boot
with
almost
zero
configuration
and
they
will
learn
from
the
protocol
how
to
operate
the
protocol.
D
It's
like
you,
don't
have
CLI
in
the
IOT
devices,
all
the
cotton
concealer
as
to
come
from
from
the
root,
but
then
the
more
and
more
information
we
place
there.
The
biggest
US,
which
has
said
earlier,
the
D
I
use,
will
become
to
the
point
that
it's
done
variable.
If
you
have
to
provide
this
information
all
the
time,
we
already
foresaw
that
with
the
ripple
configuration
option,
and
so
we
said:
hey,
you
don't
have
to
place
the
ripple
configuration
option
in
every
di
o.
But
yes,
what?
D
So
we
discuss
that
and
together
at
this
increment,
it
is
added
to
to
have
some
form
of
synchronization
information
like
database.
Synchronization
is
like
a
bit
like
when
you
synchronize
that
as
DP
is
in
the
state
just
just
to
to
to
not
configure
synchronize
the
rods
but
synchronize
the
configuration
make
sure
that
every
node
has
the
latest
information
about
configuration
and
as
we
go,
we
realize
that
it's
not
just
the
river
configuration.
There
are
actually
five
options
that
are
concerned
with
this
problem.
D
So
the
idea
is,
let's
put
something
Jo
so
that
nodes
can
know
whether
they
have
the
latest
or
not,
and
if
they
don't
have
the
latest.
Let's
do
something
with
the
disk
and
that's
what
Dominica
was
telling
us
a
few
minutes
ago
that
this
is
actually
used
in
this
document
to
be
able
to
pull
a
new
version
of
the
configuration
or
the
other
options
that
we
cover.
D
Can
we
go
down
so
I
have
not
been
working
on
this
draft
I
mean
we
did
not
discuss
it
much
on
the
mailing
list
either
since
ITF
106
and
the
reason
is,
you
know
the
record
I
know
well
the
turn
on
eight
one,
three
eight
well
already
a
lot
of
work
with
the
chairs
will
be
workers
and
Michael.
We
worked
on
user
free
info.
So
all
this
was
a
lot
of
work
just
so
we
caught
every
cart
at
the
same
speed.
D
So
so
this
one
was
kind
of
in
bad
brother,
but
it's
probably
cool
to
look
at
it
again.
Quite
it's
quite
soon.
Now
it's
also
something
that
we
have
to
probably
probably
integrate
as
part
of
this
a
bunch
of
RFC's
that
we
want
to
see
a
score
for
repo
v2
I'm
asking.
Is
it
a
good
time
now
to
start
working
on
this
again
and
the
same
question
would
apply
to
NP,
dark,
Valley
and
so
I?
D
So,
basically,
the
goal
is
not
to
have
to
place
every
option
and
we'll
see.
There's
a
slight
list
of
options
in
every
possibility
is
also
an
able
to
place
options
in
not
just
one,
the
iOS
but
the
sequence
of
the
iOS,
and
to
be
able
to
assure
that,
then
we
need
to
do
something
so
that
all
the
dials,
which
represent
the
same
bunch
of
information,
look
like
a
single
di
Yahoo
that
split
in
multiple
messages.
So
that's
also
part
of
what
we
want
to
achieve
here
and
do
that.
We
have
added
security
I.
D
Basically
that
per
option
it's
it's.
Basically
the
sequence
of
the
DAO.
If
you
sequence
the
deer
you,
then
you
can
do
two
things.
The
first
thing
is:
if
I
need
to
split
this
thing
in
sweetie
are
your
messages,
then
they
will
have
the
same
sequence
and
this.
The
second
thing
is:
if
I
want
to
refer
to
when
the
configuration
option
was
changed
for
the
last
time.
D
State
sequence,
and
we
haven't
you
yet
adoption
option.
So
it's
you
have
tools
and
there
is
a
reason
it's
an
option,
and
it
tells
you
that
it's
abbreviating
another
option.
So
basically,
this
is
a
very
small
vector
which
size
which
option
it's
abbreviating
and
basically
the
only
thing
it
does
give
you
the
last
FCSS,
the
last
sequence,
when
this
particular
option
was
updated.
D
So
instead
of
putting
the
full
option,
if
it
wasn't
a
change,
since
yes,
as
you
can
just
say
here,
is
this
a
Oh
which
is
a
replacement
if
it
matches
what
the
node
that
received
the
DI?
You
knows
as
the
most
recent
steps
on
reception,
then
the
node
knows
it's
it's
up
to
date.
If
it
doesn't
match,
then
the
nodes
needs
to
know
to
pull
so
I
needed
to
change
the
base
objects,
as
we
weren't
that
so
I
had
to
change
the
GI
you
to
put
its
own
sequence.
H
D
We
keep
sending
down
over
and
over
with
a
for
information,
same
thing.
That
could
be
a
very
rigid.
If
you
are
issues
on
the
exact
same
list
of
hosts
as
before,
then
you
could
aggregate
them
so
there's
also
the
previous
version
of
down,
and
since
now
we
turn
the
decent
to
a
query
message.
Please
give
me
the
latest
option.
Blah
then
done
need
to
be
also
something
changed,
that
this
base
object,
basically,
flags
indicating
which
option
is
being
queried
next.
D
So,
as
I
said,
there
are
actually
five
options
which
are
considered
in
this
draft
and
this
global
capability.
The
map
map
X
the
PIO,
the
dot
that
configuration.
So
if
I
come
in
an
option
and
the
right
information
option,
which
which
is
basically
how
you
know
most
more
specific
routes,
this
particular
route
can
reach.
D
So
if
you
have
to
select
the
DAC
based
on
where
it
can
actually
reach
on
the
outside
and
that's
the
route
information
option,
all
them
or
information
that
I
useful
to
the
node
that
are,
that
should
could
and
could
be
ready
on
you.
But
you
know
the
prefix
of
all
these
are
odds
can
really
reach
through
the
route.
They
don't
tend
to
change
a
lot,
so
there
is
really
no
point
to
expose
them,
but
unless
there
is
a
new
node
or
something
you
know
that
does
not
know
that.
D
D
With
the
other
slide,
that
explains
you
what's
happening,
so
if
the
child
wants
to
resynchronize,
then
you
can
use
those
five
bits
and
indicate
you
know
what
it
knows
and
what
the
latest
knows
and
then
the
parent
can
say.
Oh
this
has
changed.
Let
me
give
it
to
him.
This
has
not
changed.
Let
me
give
him
the
option.
D
A
You
very
much
thank
you
very
much
guys,
questions.
D
A
D
B
Exactly
we
could
have
one
that
focuses
on
lighting,
Dao
and
sequence
numbers
and
and
have
one
focus.
The
new
D
is
cementation
format,
but
at
least
we
want
them
to
be
consistent,
or
we
could
do
just
one
big
draft,
but
it's
usually
a
bit
more
difficult
to
handle
strong
opinion
on
this
as
long
as
they
work
together.
D
G
Stuff.
I
think
we
need
to
get
some
cycle
of
feedback
from
implementers
to
to
continue
before
we
can
step
on
to
this
part,
for
that
and
as
I
wrote
in
the
chat
earlier
half-hour
ago,
I
think
we're
getting
very
close
to
6550
this
thinking
and
and
I
think
that
that
this
kind
of
thing
actually
may
be
what
pushes
us
in
over
the
over
the
hill
and
just
saying.
Okay,
this
is
really
a
new,
actually
just
a
new
mop,
and
we
need
to
do
this
right.
That
way.
A
B
I
was
just
saying
at
least
we
have
the
need
for
modification
in
the
air
clearly
identified
the
way
that
solution
is
engineered
for
lighting.
The
IO
is
probably
maybe
not
fully
casting
in
Aaron,
but
at
least
we
we
know
we
want
to
do
something
with
GIS
to
address
that
need
on
outside.
We
can
start
working
on
the
ice
modifications,
taking
that
just
case
into
account.
H
A
So
we
will
organize.
Thank
you
very
much.
So
next
steps
is
that
we
are
going
to
send
a
summary
of
this
meeting
with
the
action
action
points
and
the
link
with
that
recording.
So
and
if
someone
have
further
questions,
we
are
yeah,
so
I
think
we
can
conclude
the
meeting.
We
are
four
minutes
about
the
limits
so,
but
thank
you
very
much
for
today
the
conversations
we
keep
in
the
mailing
list
and
in
the
next
meeting,
it's
okay
for
you
to
hours,
right
or
2.5,
I
think
Joe.
What
is
fine,
so
we
don't
do.