►
From YouTube: IETF104-ROLL-20190325-1610
Description
ROLL meeting session at IETF104
2019/03/25 1610
https://datatracker.ietf.org/meeting/104/proceedings/
A
Good
afternoon,
West
I
withdraw
meeting
welcome
to
everyone.
This
meeting,
please
note
that
this
meeting
is
a
touchstone,
not
well
be
aware
of
that
and
the
meeting
materials.
Thank
you
very
much
to
Dominique
for
being
the
meanest
taker
and
diverse,
cried
Raoul.
Thank
you
very
much
guys.
Thank
you
very
much
unequal.
There
is
someone
else
that
want
to
please
help
those
guys
be
kind
and
help
us
please,
and
there
are
the
links,
please
sign
the
blue
sheets
yeah
the
bruises
are
they
are
so
yeah,
so
this
is
the
agenda.
A
B
Like
to
keep
the
milestones
as
the
I
am
not
updating
and
the
less
that's
a
very
good
reason
for
it,
so
it
shows
how
fast
our
progress
is,
and
if
you
keep
on
updating
the
milestones,
we
all
think
your
own
time
and
you're.
Not
that
mean
it
just
slow,
so
I
wanted
to
keep
the
mouse
saying
date
system
if
there
any
objections
like
this
shows
badly
off
on
the
reputation
of
the
working
group.
Please
tell
us
now
otherwise,
I'll
just
come
to
keep
them
yep.
C
A
A
A
A
Ok,
the
use
of
rippling
info
M.
We
believe
that
we
add,
like
some
operational
considerations,
for
the
new
option
and
the
security
section
as
well
was
modified.
We
believe
that
address
all
the
tickets.
A
please
read
if
you
have
comments
into
the
main
list,
because
now
she's
in
the
last
call
charlie
is
not
here,
so
we
can
pass
to
Raoul.
D
Hello,
hello,
this
working,
ok,
thank
you.
I'm
technologists,
I'll
be
talking
about
two
things.
Today,
one
is
the
mo
P
or
the
mode
of
operation
extension,
which
is
rather
an
easy
subject,
but
the
other
thing,
the
capabilities.
This
is
something
that
we
are
trying
to
introduce
new
into
our
peel
and
I'll.
Try
to
reason
as
to
why
this
is
required,
but
before
going
into
the
into
the
details
about
capabilities,
let
me
talk
about
mode
of
operation.
D
Mode
of
operation
is
something
which
is
defined
in
our
peel,
such
that
all
the
6l
R's
in
the
network
have
to
mandatorily
support
all
the
primitives
that
are
required
by
this
mode
of
operation,
so
6
l
ends
doesn't
need
to
support
this
primitive.
This
is
only
mandated
for
6
hours.
Now
it's
3
bits
in
size
and
it's
already
exhausted
in
the
sense
that
if
we
finish
all
our
work
in
the
existing
on
the
existing
documents,
it
will
be
exhausted.
So
you
can
see
on
the
left-hand
side
here
the
mo
P's.
D
There
are
five
six
seven
which
is
unused,
but
there
are
already
contending
drafts
which
would
like
to
use
those.
So
unless
and
until
we
have
this
extended,
we
won't
be
able
to
proceed
with
some
of
these
jobs.
So
well
it's
a
very
straightforward
extension.
What
we
are
doing
is
we
are
adding
a
new
article
control
message:
option:
it's
essentially
we
have
the
existing
three
bit
value.
If
it
is
seven,
then
it
serves
as
an
extension.
In
that
case,
an
implementation
should
go
and
look
for
this
extended.
D
The
article
control
message
option
and
the
final
mo
P
should
be
calculated
as
the
base
mo
P
plus
the
move.
The
new
map
X
that
so
in
the
table
here
shows
how
should
you
calculate
the
final
mo
P?
It
gives
you
a
nice.
So
the
reason
why
we
did
what
this
was
so
that
in
the
future
we
should
be
able
to
refer
mo
p,
0,
1
2
3,
in
the
same
way
that
we
do
refer
right
now
and
all
the
new
movies
that
we
add
in
the
future.
D
We
should
be
able
to
incremental
e
refer
from
8
onwards
or
7
onwards.
Now
this
is
something
new.
So
the
primary
rationale
for
supporting
this
or
for
introducing
this
capabilities
was
there
are
certain
options
or
certain
features
in
ripple
or
on
the
nodes
which
might
be
optionally
supported.
Now,
I,
like
I
mentioned
before
the
mode
of
operation,
is
something
which
all
the
6l
arts
have
to
mandatory
support,
for
example,
source
outing
with
multicast
mode.
D
All
the
six
alerts
in
the
network
have
to
Manderley
support,
so
it
is
required,
but
there
are
certain
features
which
might
be
optionally
supported
and
by
the
border
router
or
by
the
nodes
in
the
network.
In
that
case,
it
would
be
good
for
the
border
router
to
signal
such
an
extension
such
that
nodes
can
optionally
enable
it
if
they
require
or
sixel
ends
of
the
6l
ARS
can
signal
their
own
capabilities
such
that
border
router
will
know
whether
six
Allen's,
what
are
the
capabilities
supported
by
the
six
islands?
D
So
this
capability
is
a
new
option
again
it
will
carried
in
Dao
and
our
message
and
the
primary
difference
with
a
movie
is
that
it
could
be
optional.
So
the
capabilities
that
are
supported
well,
it
could
be
mandatory
as
well,
but
but
it
could
be
optional
as
well.
That's
that's
what
the
point
is
here.
So
an
example
use
case
I
was
just
thinking.
This
is
not
something
that
I've
really
thought
about
in
detail,
but
I
was
thinking
six
LRH,
which
is
which
right
now
doesn't
have
a
dispatch
mechanism.
D
There
is
no
way
of
signaling
in
the
network
that
the
whole
all
the
nodes
in
the
network
support
six
lower
edge.
This
could
be
signal
using
such
kind
of
an
extreme
cells
such
kind,
such
capabilities,
so
most
of
the
most
of
the
the
capabilities
field
is
modeled
upon
existing
Aikido
2.11
already
have
such
kind
of
mechanism,
so
we
try
to
model
based
on
that
that
kind
of
information.
D
So
it's
a
very
straightforward
option.
We
have
just
the
bits
specifying
new
capability
options
and
these
these
capability
flags
are
what
have
to
be
difficulty
finding
the
new
specifications
well
points
to
ponder,
so
capabilities
can
essentially
be
made
used
over
by
existing
mo
peas
as
well.
Let
me
clarify
here
that
there
is
no
cross
relation
cross
coupling
between
mo
peas
and
the
capabilities,
their
movie
extension
and
the
capabilities.
There
is
absolutely
none,
so
you
can
use
individually.
The
movie
extension
leave
out
the
capabilities
or
you
can
use
capabilities
with
existing
mo
peas
as
well.
D
The
both
these
possibilities
exist
now,
caps
capabilities
have
to
be
sent
in
every
dau.
Well,
that
is
something
that
needs
further
thoughts.
In
fact,
the
same
applies
for
di
o
as
well,
and
the
same
applies
for
MA
packs
or
the
movie
extension.
We
would
like
to
Elia
elide
it
in
the
future
messages
its
subsequent
messaging
I,
don't
know
whether
we
can
we
can
make
use
of
the
same.
D
We
had
some
time
back
the
I/o
configuration
option
which
could
be
lighted
so
I'm,
hoping
some
similar
mechanism
can
be
made
use
of,
but
that
drop
is
currently
expired.
You
hope
to
we
hope
we
can
revive
that,
and
we
have
the
author's
here
in
another
thing
is
the
impact
on
the
memory
requirement.
Well,
this
is
a
critical
factor
that
I'm
thinking
of
so
now
every
6lr
can
generate
a
dowel
for
any
of
the
sub
tree
nodes
by
itself.
Please
don't
routing
into
it
now.
D
D
F
Sketchy
were
here
when
we
thought
the
map.
Initially,
it
was
exclusive
like
mutually
exclusive.
You
were
either
storing
mode
or
non
storing
mode,
etc.
When
we
look
at
the
Dow
projection,
it
is
actually
in
addition
to
storing
mode,
often
on
storing
mode
right,
so
we
cannot
easily
signal
it
in
replacement
of
the
original
mode.
It's
more,
it's
an
extension,
but
it's
on
top
of
right
right.
G
F
Mode
projection
or
non
storing
mode.
It
looks
like
an
enumeration
of
things
right.
So
right,
I
don't
know
exactly.
We
are
not
done
yet
about
how
we
express
them,
but
we
need
to
consider,
but
in
the
future
we
may
have
a
main
mode
of
operation
and
then
sub
mode
of
operation
which
are
not
necessarily
supported
by
everybody.
Right.
H
F
A
A
A
D
A
Yeah,
but
he
won't
anyway
in
doubt,
projection.
There
are
tickets
in
security
open.
So
we
why
these
kind
of
mode
of
operation
draft
is
going
on.
We
can
address
the
ticket
from
our
projection,
so
I
mean
the
ticket
opens
in
doubt
protection
from
security
considerations.
So
we
can
address
all
the
the
issues.
B
F
Depends
where
we
want
to
take
this
document?
Well,
one
one
kill
dependencies:
how
to
signal
the
support
of
doubt
projection
right,
so
that
one
it's
it's
you
know
it's
easy
to
to,
but
will
need
this
to
be
an
RFC
to
be
able
to
build
a
product
which
express,
oh
I,
support,
doubt
projection
as
to
storing
us
a
non-story,
etc.
So
that's
the
easy
one.
The
less
easy
is
I
have
questions
still
on
the
doubt
projection
about
things.
F
We
don't
do
in
this
draft,
but
you
need
for
the
global
solution,
so
that
does
not
prevent
the
draft
from
progressing,
but
that
prevents
from
building
a
solution
like
how
does
the
route
know
how
much
memory
is
available
on
this
device?
So
I
can
establish
10
rods
for
him.
How
does
the
route
know
the
topology?
So
this
one
could
not
address
the
GG
thing
probably,
but
it
could
address
the
amount
of
memory
has
sort
of
those
capabilities
as
well.
B
D
For
me
so
now
the
question
that
I
would
like
to
ask
is
now:
there
are
two
components
here:
one
is
their
mobile
extension
and
the
capabilities.
If
you
really
want
to
expedite
your
mochi
extension,
we
can
leave
capabilities
in
a
different
raft.
Do
you
think
we
can
proceed
it
in
the
same
capabilities
will
definitely
require
more
deliberations.
If
we
introduce
henge,
that
would
be
more,
it
will
add
to
the
complexity.
D
A
F
Need
to
discuss
that
in
more
detail,
spec
precisely
the
minimum,
he
has
to
be
able
for
node
2
to
tell
the
route
I
support
out
projection,
storing
non-story,
which
is
very
close
to
the
map.
It's
just
a
map,
the
other
way,
because
you
know
you
have
a
main
map,
which
is
the
one
which
is
in
the
GI
unit
non-story,
and
then
you
get
this
thing
in
the
tell
you
which,
in
the
extension
which
size
oh
by
the
way,
we
also
do
non-story
projected.
So
the
route
does
okay.
Now
the
device
need
to
respond.
F
Yes,
I,
also
like
it
right
and
as
well.
I
would
package
that
as
a
bare
minimum
with
this
draft,
no
the
extension
to
the
caps
where
I
say
I
get
this
much
memory
and
whatever
else
the
route
would
need
to
know
can
be
a
later
of
C,
because
it's
not
you
can
can
be
a
down
the
minimum
expected
memory
or
anything
like
that
right
up
configured
the
route
could
be
configured
to
know
that
this
device
is
that
at
least
that
much
memories
throughout
the
network
whatever.
F
A
Okay,
it
seems
we
need
more
discussion
on
need.
We
can
put
that
information
in
ripple
observation,
strap
so
a
section
with
that
with
all
their
requirements,
so
because
it's
a
trap
that
we
use
yeah
because
in
a
drug
that
we
used
to
so
we
add
in
this
is
easier
to
discuss,
because
we
have
like
writing
right
if
you
agree
with
that,
so
we
can
separate
the
work
and
depend
up
or
merge
against
needed
agree
ready.
Okay,
thank
you
very
much.
A
I
I
I'm
Charles,
Perkins
and
I'm
here
to
review
the
modifications
to
a
ODB
ripple
draft
in
response
to
a
commis
received
during
last
call
so
oops
wrong
direction.
Turn
it
around
about
that
okay.
So
the
main
thing
that
was
done
to
the
draft
was
to
add
us
expanded
or
greatly
expanded
security
considerations
section
and
that
ended
up
being
a
lot
of
text
for
a
reason,
I'll
explain
momentarily
and
then
it
was
also
requested
that
a
ODB
riffle
actually
does
two
things.
I
It
does
peer-to-peer
router
discovery
and
it
also
does
Isis
Isis
bi-directional
asymmetric
route
discovery,
and
then
there
were
some
significant
editorial
improvements
that
were
also
incorporated
into
the
draft
so
about
the
security
considerations.
The
previous
section
just
said:
hey
ODB
ripple
can
use
the
same
security
model
as
RFC
6550,
and
it
was
requested
for
us
to
be
more
explicit
about
that.
I
So
then,
the
the
new
security
consideration
section
actually
talks
about
the
three
modes
that
are
defined
in
RFC
6550,
which
are
unsecured,
was
basically
I,
think
should
be
interpreted
to
mean
secured
by
means
other
than
ripple
and
pre-installed.
Basically,
just
as
you
install
some
keys
and
use
those
pre-configured
shared
keys
to
establish
the
security
of
the
messages
and
then
there's
an
other
mode
is
called
authenticated,
which
presumably
pre-installed
also
in
authenticates
but
using
the
same
terminology
using
the
pre-configured
shared
keys.
The
devices
obtain
another
key
that
they
use
for
purposes
of
authentication.
I
Now
it
was
on
purpose,
not
none
of
the
provisions
and
discussion
in
this
new
security
consideration
sections
none
of
that's
really
different
than
what
was
already
present
before,
but
we
were
able
to
apply
this
specifically
to
the
route
request
and
without
replying
messages
from
a
ODB
ripple.
So
I
think
that
we
did
not
introduce
any
new
bugs
or
anything
like
that,
and
it's
pretty
much
as
before,
but
much
more
explicit.
I
So
there's
a
to
specific
parts
of
a
OTV
ripple
that
provides
advantages.
One
is
that
that
you
can
find
peer-to-peer
routes
so
that
they
don't
have
to
go
through
the
route
they
don't
have
to
do.
The
up/down
routing
ripple
does,
and
so
that
has
two
benefits.
One
is
that
there's
less
interference
and
congestion
nearby
the
route
which
can
be
important
than
a
lot
of
applications
and
then
also
a
peer-to-peer,
allows
you
to
find
also
be
shorter
paths.
Yes,
Pascal.
F
Yes,
I
want
to
insist
on
what
you're
just
saying
she
on
the
shorter
path
becomes
sure
off
the
short.
Actually,
you
can
really
think
different
metrics,
because
you
don't
have
to
depend
on
the
fact
that
the
main
repo
back
you
see
is
whatever
ITX.
Here
you
might
want
to
establish
a
more
deterministic
path
and
yeah
a
very
different
criteria
than
the
main
deck
right.
I
I
That's
right
so
I
guess
I
was
a
bit
hasty
when
I
grew
up
these
slides.
It
really
is
that
the
metric
is
more
favorable
between.
So,
if
you
want
to
measure
something
else,
perhaps
a
reliability
or
whatever
yeah,
it
depends
on
the
objective
function.
So
I
didn't
say
that
here
but
you're
exactly
right
and
then
the
other
thing
is
about
the
bi-directional
asymmetric
routes.
Sometimes
you
can
find
routes
are
perhaps
even
off
and
you
can
find
routes.
It
wouldn't
be
discoverable
by
normal.
Our
triple
means.
I
So
that's
almost
all
of
it.
It
was
pointed
out
that
we
were
using
a
target
and
introduction
when
we
didn't
define
what
it
meant.
So
we
just
expanded
that
out
to
be
target
note
and
some
of
the
references
were
not
normative
and
they
were
moved
to
informational,
and
then
there
was
some
verbiage
around
sequence
number
implementation.
That
was
not
clear,
so
I
made
that
clear
and
then
it's
hard,
the
normal
number
of
spelling
and
grammatical
syntax
Corrections
amazing.
I
A
Well,
we
believe
that
the
document
is
ready,
so
we
are
going
to
do
after
the
idea.
One
week
last
call
like
so
to
confirm
that
if
you
please
read
the
last
version
and
and
yes
special
security
considerations
and
let
let
us
update
us,
know
your
comments,
but
I
mean
for
us,
the
East,
Grady,
okay,
well,.
J
F
Okay,
so
back
to
the
Dow
projection
draft,
so
number
of
changes
first,
is
we
asked
Matt
to
actually
contribute
from
these
operational
experience
and
implementation
experience?
So
that's
a
new
thing
on
this
and
the
latest
publication.
We
have
added
text
on
which
conditions
could
actually
cause
rots.
It's
really
superposition
of
different
routes
installed
by
different
persons
like
the
domain
rod
installed
by
the
background
repo
and
then
the
storing,
node,
storing
mode
projected
route
and
then
a
non
storing
run,
and
you
could
actually,
in
some
cases
I
mean
we're
two
loops.
F
F
My
feeling
is,
you
should
not
basically
try
to
have
too
many
things
at
the
same
time
on
the
same
network,
I
think
a
main
mode
and
then
one
projection
mode
should
be
enough
for
what
for
a
given
use
case.
For
instance,
if
you
have
a
long
lines
a
long
line
of
meters,
then
you
would
use
a
non
storing
mode
for
the
network
and
then
for
this
long
line.
You
would
just
do
storing
mode
along
the
line
to
just
save
the
raw
together
along
that
line.
F
But
then,
at
the
same
time,
would
you
be
using
storing
mode
projected
route
in
that
area?
I
have
trouble
believing
it
so
to
be
a
dozen
natural.
It
makes
sense
to
try
to
put
too
many
things
on
the
same
network.
So
maybe
we
can
have
recommendations
like
that
as
a
main
mode
and
then
project
in
mode.
One
of
them
then
you're
not
creating
the
slips
I
guess.
F
And
then
we're
missing,
maybe
it's
better
now
happy
tears
and
people
could
review
it.
We
are
missing
a
good
text
on
what
happens
if
there
is
a
kitchen,
so
we
have
added
a
new
ICMP
type
actually
to
re-explain.
Instead
of
the
traditional
repo
ICMP
row
that
you
had
in
no
story
node,
we
actually
created
a
new
type
so
that
you
can
describe
that.
The
problem
is
not
jus
funston's
to
the
underneath
non
storing
mode,
but
tree
to
the
projected
route
itself
know
the
route.
F
Can
you
know
if
you
are
not
storing
the
route
will
get
if
there
is
a
problem
along
the
path
right,
you're
following
a
routing
inner
and
the
next
hop
is
not
there.
So
you
have
to
send
an
ICMP
error
which
is
defined
in
RFC
sixty
sixty
five.
Fifty
that
tells
you
I,
have
this
error
on
my
project
on
mine
lost
on
path.
F
F
That's
the
change,
but
then
again
is
new
tags.
So
if
people
want
to
have
a
closer
look
at
basically
what
happens
when
there
is
a
breakage?
Do
we
signal
it
correctly?
Can
we
repair
it
correctly,
etcetera?
Let's,
please,
review
that,
and
discussions
are
pretty
much
outside
the
scope
of
this
draft.
That's
the
map
discussion
that
we
just
heard
switch
listed
somewhere
on
this
Friday.
Yes,
yes,
and
there
is
the
first
problem-
is
how
does
the
route
even
know
the
topology
in
non
storing
mode?
F
The
route
knows
kind
of
a
tree,
our
dag,
but
it
doesn't
know
the
neighbors
the
peers
at
the
same
rank.
So
that's
that's
one
thing
that
that
not
propagated
to
route
and
in
storing
mode
there's,
nothing
pretty
much
from
topologically
speaking.
It
does
nothing.
It's
a
distance
vector,
it
doesn't
see
thing
so
should
we
have
a
draft
somewhere
in
this
group,
so
my
house
is
it
something
which
which
should
be
in
another
group?
How
do
we
express
the
topology
to
do
that?
One
thing
is
missing,
so
they
kept
ability.
F
Yes,
we
have
the
the
map
extension
draft.
We
could
actually
have
options
in
the
capabilities
to
say
amount
of
memory
amount
of
laborers,
whatever
house
could
be
needed,
throughput
capability
battery
level,
I,
don't
know
so
so
matrix
pretty
much
very
close
to
the
metric
container
that
we
have,
but
it
would
be
a
capability
container
or
something
that
we
could
sell
to
to
the
route.
And
so,
if
you
have
to
ship
the
map
X
rapidly,
then
this
should
be
a
different
draft.
Radiator
and
after
that
there
is
a
traditional
recall
thing.
F
Is
we
never
compressed
repo
itself?
Now
we've
FCAT
138,
we
compressed
the
data
path,
but
we
never
comprised
the
contraband.
We
said,
let's
express
things
naturally
and
then
we'll
do
a
compression
of
it
later,
but
we
never
did,
and
so
here
again,
if
you
have
a
via
option
which
is
an
on
storing.
So
it's
going,
you
will
have
a
number
of
hops
in
there.
It's
gonna
take
someone
right.
It's
going
to
be
a
big
packet.
So
should
we
already
think
about
compression?
F
F
So
you
already
hear
people
tell
you
a
I
can't
sound
my
Dow
MIT
I/o,
with
all
the
options
I
want
it
to,
because
I
don't
have
em
to
you.
For
so
I
have
to
send
multiple
messages,
so
sometimes
compression
will
alleviate
that
created
it,
and
so
it's
all
questions
which
are
kind
of
raised
by
this
draft
which
are
not
inherent
to
the
draft
itself.
But
it's
it's
questions
for
the
group,
so
the
most
urgent
we
really
need
is
the
capability,
because
if
the
route
doesn't
know
I
can
do
story,
I
can
do
read
projection
I.
F
Have
that
much
memory,
you
can
install
five
routes
for
me
or
ten
routes
for
me
know
if
the
route
doesn't
know
that
what
can
you
do
and
I
don't
expect
that
from
a
couple
of
GE
expression
protocol
that
expressed
apology,
it's
really
inside
of
the
of
the
device
I'm
pretty
sure
that
belongs
to
us.
Now
all
the
topology
descriptor
and
stuff,
like
that
I
mean
there
are
STIs
and
on
OC
count
those
groups.
K
Yasmani
and
some
into
the
microphone,
but
I,
don't
know
about
what.
So,
of
course,
it's
always
a
problem
in
the
design
of
a
particular
whether
you
design
it
in
uncompressed
form
and
trust
somebody
else
during
the
compression
for
you.
So,
for
instance,
if
there
were
something
like
six
know,
our
age
that
might
do
the
compression
for
you.
K
F
K
F
You
gave
a
perfect
example
right.
A
RFC
8
138
allows
you
to
compress
the
next
at
rise,
because
that's
the
same
prefix
in
part
as
the
previous
address
and
things
like
that
and
that
that's
how
we
comprise
the
rotten
either
in
6rh
or
the
data
path,
and
it
makes
a
lot
of
sense
to
compress
the
rotten
leader
that
we
have
in
the
projected
route
in
the
exact
same
fashion
right.
This
is
insane,
but
when
we
did
8130
aching
this
book,
not
there
is
no
watch
me
the
thing
back,
but
they
ate.
K
F
A
G
F
Okay,
so
I
presented
those
slides
at
six
slow.
A
bit
were
like
just
before
this
meeting
so
so
this
is
the
context
in
which
the
unaware
leaf
operates.
This
context
is
a
set
of
four
RFC's
which
basically
enable
or
extend
RFC
6775
and
provide
a
trusted
registration
whereby
you
can
register
on
the
trials,
obtain
routing
services,
and
that
can
be
proxy
ng
that
can
be
rapport.
That
can
be
anything
and
then
there
is
a
way
to
prove
that
you're,
the
owner
or
the
address
that
was
registered
and
for
which
your
services
is
being
granted.
F
So
this
is
a
set
of
of
trash.
Actually
one
of
them
became
our
Shi
for
five
and
and
the
other
are
following.
So
that's
pretty
much.
The
story
that
was
standing
here
at
505
is
a
registration.
It's
a
layer,
3
version
of
the
Wi-Fi
Association.
You
tell
your
router
in
advance
at
your
share,
so
that
we
won't
have
to
answer
to
broadcast
or
do
not
speak
at
first
off
or
that
the
backbone
router
is
the
proxy
Andy.
F
So
if
the
wireless
guys
a
Wi-Fi
device
or
cheating
for
device
registers,
the
backbone,
router
or
proxy,
indeed,
the
back
on
water
condition
the
device
on
these
on
that
side,
even
if
the
device
is
slipping
and
IP
Indies
the
same
the
security
variation
for
that
model
well
by
the
address,
that's
being
registered
associated
with
a
cryptographic
token,
so
we
can
prove
later
the
ownership.
So
the
device
moves
and
wants
to
to
say
all
these
IP
addresses
there
was
here
on
out
there.
Oh,
are
you
the
same
guy?
F
F
That's
what
I
can
with
that,
and
there
is
this
new
work
that
sits
low,
which
is
about
using
the
method
that
we
have
in
our
state
five
of
I've
inherited
from
6775.
Well,
we
have
a
registrar,
a
social
racetrack
called
a
6lv.
Are
they
we
would
have
it
on
the
backbone,
so
you
can
do
the
traditional
labor
discovery
in
unicast.
F
So
when
you
have
a
new
address,
you
just
don't
do
that.
You
also
populate
this
registrar
so
next
time
somebody
does
that
you
can
ask
the
transfer
registrar
and-
and
he
knows
that
that
your
drives
are
in
taken,
doesn't
do
a
broadcast
as
I
need
to
write
someone
second
et
cetera.
It
already
knows
the
transistor.
F
Ok,
so
the
ripple
counter
part
of
this
is
remember:
I
told
you
are
CH
505
enable
a
host
a
sit
Ln
which
doesn't
know
that
the
network
is
being
a
repo
network
or
she
is
attaching
to
a
switch
infrastructure
or
anything
else.
It
just
enables
this
device
to
say
mr.
rotter
you're,
my
router
you're.
Protecting
me
I
can
sleep
and
we'll
store
my
packet,
whatever
I
trust
you
to
do
this
routing
service?
For
me,
and
now
I
can
go
to
see,
routing
service
could
be
and
the
proxy
with
the
backbone
router
another
one
is
ripple.
F
So
when
we
rotate
505,
we
write
care
to
make
sure
what
was
needed
for
repo
was
actually
there
and
there
is
actually
no
back
field,
which
must
be
populated
a
certain
way
when
you
want
ripple
services,
but
which
much
on
the
fields
that
needs
to
be
injected
via
dowel
by
the
router
are
present
in
this
association
process.
So
the
association
registration
is
lethal
agnostic,
but
it
has
all
that
ripple
needs
that's
pretty
much
what
this
drug
does.
So
it's
been
around
for
a
little
while
now
we
have
added
multicast
address
which
match
the
way.
F
K
F
Really
about
what
a
repo
rotterdam
when
the
host,
when
the
host
uses
out
eight
five,
four
five,
so
ok,
h
505
is
a
host
or
router
operation.
Yes,
but
it
doesn't
say
what
the
router
does.
If
the
router
is
a
ripple
rata,
this
one
is
only
inside
the
repo
router.
What
do
you
do
if
you
get
a
registration
which
is
repo
agnostic?
F
K
F
The
design
that
affects
505,
so
we
just
published
at
505
with
in
mind,
to
be
able
to
do
this.
Now
there
is
one
field
which
I
just
mentioned,
which
is
the
OPAC
field
in
which
a
node,
which
is
vaguely
aware
that
there
is
some
repo
layer
or
something
similar
to
repo,
will
pass
the
instance
ID,
because
repo
has
this
concept
of
instanceid.
So
there
is
some
vague
text
in
H
5,
4
5,
which
says
you
have
this
topic
field.
F
You
can
pass
some
information
for
the
routing
protocol
and,
if
you
face
it,
the
Roger
protocol
would
be
happy
and
in
that
traffic
we
are
saying
hey
by
the
way,
if
your
host-
but
you
know
what
you're
doing
and
you're
attaching
to
a
repo
pass.
For
instance,
I
give
you
get
one,
but
the
only
field
which
is
but
but
if
have
a
purely
6lowpan
age,
five
or
five
devices,
it
will
join
the
instance
zero.
It's
going
to
be
very
happy.
K
F
It's
in
the
official
at
5:05,
independently
of
people,
so
it
505
specifies
the
Chi
ID.
We
specifies
it
and
size
job
because
we
need
a
sequence
counter
when
we
have
mobility,
so
we
needed
one.
Oh,
we
copied
the
text
in
red
ball,
so
it's
perfectly
compatible
fine.
We
could
have
picked
just
a
normal
sequence
counter
increment,
oh,
but
we
pick
the
one
which
is
easier
to
convert
into
ripple
because
we
knew
wanted
to
do
it,
but
it's
the
we
H
five
four
five
tells
any
lambda
random
6ln
to
use
to
our
GID.
K
K
F
D
F
D
F
A
But
anyway,
will
you
add,
because
if
you
mention
in
the
document,
I
know
that
is
I
read
it
Anna
is
like
a
leaf
is
mentioned
in
the
in
the
description
now
in
terminology
we
want
to,
if
you
could
add,
please
in
the
terminology
section
like
what
means
host
for
the
drop
from
the
point
of
view
of
the
draft.
It's
like
a
Rooter
yeah
like
a
rope
router
without.
A
F
F
A
K
F
K
K
F
K
The
question
I
mean
people
who
are
thinking
about
how
the
router
interfaces
can
they
use
this
opaque
thing
for
for
interesting
things,
and
if
yes,
is
there
any
requirement
following
from
this,
that
should
go
into
this
design
and
this
this,
but
yeah
I
would
prefer
to
have
it
in
but
I'm
just
saying
when
we
design
something
like
this,
you
look
a
little
bit
on
the
envelope
and
see
who
else
might
want
to
use
this,
and
if
yes
do,
they
have
any
requirements
that
it
I
can
easily
accommodate
by
design
even
in
a
certain
way.
So.
F
A
F
A
surprise
to
the
slice
that
was
one
which
we
did
not
read
this
cascade.
Can
you
go
back
one
slide
for
a
second?
Yes,
the
ups?
Yes,
this
working,
because
there
is
something
in
this
draft
which
I,
don't
think,
is
a
good
idea.
I
just
made
exercise
of
enabling
it,
but
the
sixties
architecture
says
it
should
not
be
that
way
and
I
think
when
I
read
67
75,
it's
also
the
same
way
so:
okay,
67
75,
introduced
to
the
Beast,
which
it
called
6lv
on
thy
name,
it's
a
god
or
otter.
F
That's
what
the
B
eyes
for
my
function.
It's
a
registrar!
That's
when
you
put
the
list
of
IP
addresses
that
are
in
your
network
and
should
at
least
be
hosted
by
a
border.
Router
I,
don't
know
I
mean
maybe
yes,
maybe
no.
Maybe
it
can
be
somewhere
in
the
cloud.
And
so
we
have
this
big
confusion
of
name
and
function
and
now
we'd
be
of
this
aggregated
saying,
with
with
the
backbone
backbone,
routers,
etc.
And
the
question
comes
whether
the
6lv
are
which
my
name
is
about.
F
F
If
it
is,
then
you
don't
need
this
weird
stuff
that
you
have
here
with
people
I've
dark
etcetera,
because
if
there
are
a
different
beast
and
you
must
have
a
mechanic,
some
mechanics
between
the
route,
which
knows
you
know
through
the
down
process,
the
route
knows
who's
there
and
the
6lv
armies
to
know
who's
still
alive,
because
it
will
time
out
registration
as
a
time
out.
If
you
want
to
maintain
the
status
six
and
VR,
you
must
periodically
tell
the
six
MVR
I'm
still
there
but
part
of
what
we
do
in
this
draft.
F
Is
we
use
the
down
for
that?
So
you
don't
have
to
send
it
down
and
then
sending
two
messages.
That
of
one
is
a
waste
throughout
the
whole
network.
So
this
draft
saves
that,
because
it
says,
if
the
route
gets,
the
the
down
and
the
route
is
co-located
with
the
six
lvr,
then
by
the
way,
I
think
about
refreshing.
The
states
in
the
six
and
y
are
the
guy's
not
dead
right.
B
F
A
F
F
A
J
So
hi
I'm,
iris
I
think
the
continuation
of
our
work
on
an
extension
to
the
dog
metric
container
for
storing
a
parent
set
to
enable
PRA.
So
we
have
updated
the
draft
we
review
from
Diego
and
now
has
given
us
some
extra
comments.
So
he
pointed
out
that
we
have.
We
had
a
strict
dependence
on
6
days
and
we
removed
that
none
of
this
world
depends
on
6
dish.
J
We
also
added
the
explanation
of
how
a
turret
parrot
selection
can
work
one
of
the
methods
which
we
have
been
using
internally,
but
it's
just
an
option
and
for
some
some
sections
of
the
draft
we
have
made
the
prescriptions
well
on
what
to
use
a
bit
more
flexible.
So,
for
example,
you
can
use
any
alternative.
Parrot
selection
with
this
extension
and
the
different
nodes
can
use
different
alternative
selections.
We
don't
prescribe
some
that
all
the
nodes
must
use
the
exact
same
method.
There's
no
reason.
J
J
As
a
brief
overview.
I'll
go
over
this
quite
quickly.
We
discussed
this
a
lot
of
times,
so
the
idea
here
for
dystrophy
is
to
try
to
have
some
form
of
determinism
to
have
higher
and
have
higher
reliability
with
low
jitter
and
bounded
delay,
and
the
idea
is
to
do
PR
a
to
send
multiple
copies
of
a
packet
over
different
paths
so
that
we
can
achieve
that.
J
J
The
idea
looks
like
this
more
or
less
so
you
have
a
source
node
that
sent
some
information.
It
has
the
default
path,
which
is
a
solid
line,
and
it
might
try
to
create
an
alternative
path
through
the
dashed
line.
So
it
replicates
information
to
both
a
and
B,
but
you
want
to
send
the
information
to
be
finally
right
to
be
where
one
of
the
copy
will
be
eliminated.
J
If
it
doesn't
go
to
the
end,
go
to
some
other
node
to
the
side,
then
you
want
to
packet,
elimination,
and
this
leads
to
flooding
in
order
to
select
where
you
will,
which
will
be
the
alternate
opponent,
which
you
will
send
information.
You
need
some
extra
information
to
guide
you.
So
in
this
sense,
what
we
are
sending
is
the
parent
set
of
the
potential
parents
to
which
you
will
replicate.
So
you
know,
which
are
the
parents
in
the
parent
set
of
a
and
B,
and
that
helps
us
choose
who
it
will
replicate
to.
J
J
J
This
had
some
overhead.
It
adds
one
byte
of
overhead
in
this
new
version
of
the
draft.
We
prescribe
that
you
must
use
six
lower
compression
on
the
addresses,
but
in
the
previous
version,
who
said
it's
optional,
but
in
okay
and
yeah,
this
is
just
normal.
Six
Laura
compression
you
decide,
which
is
the
most
common
denominator
prefix
and
is
based
on
this.
There
is
a
limitation
here
that
all
the
addresses
are
compressed
to
the
exact
same
prefix.
J
So,
if,
for
whatever
reason
that
a
bigger
difference
between
the
others
is
you
want
such
an
efficient
compression
level-
and
here
we
have
an
example
of
how
this
would
work,
so
one
of
the
proposed
methods
of
parted
parent
selection
is
called
see
a
medium
common
ancestor
medium.
So
in
this
case
one
would.
The
default
path
is
from
s
to
a
and
wet,
and
you
want
to
evaluate
B,
1,
B,
2
and
B
3,
who,
who
will
choose
to
be
the
alternative
parent.
J
So
in
this
case,
B
1
will
be
rejected
because
its
parent
set
D
and
it
does
not
contain
F
the
default
grandparent
and
we
can
choose
either
b2
or
b3,
because
both
their
parents
hence
contain
F.
There's
other
ways
of
doing
this
in
the
draft.
Depending
on
how
you
do
it,
you
might
have
more
or
less
flooding
and
more
or
less
energy
consumption
and
more
or
less
related.
J
J
So
there
is
issue
whether
we
would
like
to
discuss
the
scope
of
this
draft.
So
do
we
keep
this
simple,
just
a
small
extension
to
an
optional
field
in
the
diag
metal
container,
which
contains
this
information,
or
do
you
want
to
actually
prescribe
specific
methods
of
implementing
with
this
alternative
para
insurrection
that
uses
this
information
and
in
discussions
we've
had
with
my
impression
polls?
That
would
prefer
to
keep
it
simple,
but
someone
has
a
different
opinion.
G
D
Right,
so
it
is
a
complicated
thing.
In
my
opinion,
you
know
there
are
that
there
are
two
things
you
mentioned,
that
the
parent
selection
logic
yeah.
That
is
that
that
is
something
very
specific
to
implementation.
Those
are
implementation
guidelines,
yeah
it's
great
to
have
in
the
document,
in
my
opinion,
but
as
an.
J
Example,
you
mean
here:
okay,
okay,
so
the
one
thing
is
the
information
itself
carried
and
how
you
use
it.
Maybe
you
can
give
an
example,
but
actually
it's
up
to
you,
it's
an
implementation
detail,
or
maybe
there
can
be
another
draft
something
specific.
But
at
this
point
this
is
just
an
idea,
and
then
you
have
the
third
point
here,
which
discusses
aiding
correctly
enrollment.
Was
that
it
or
not
the
unknown?
Oh
sorry,
so
it
was
the
idea
that
Rahul
had
was
in
addition
to
the
parent
set
itself
to
add
for
each
parent
priority
level.
J
So
it
will
quantify
and
you
don't
just
have
the
order
of
preference
between
the
parent
which
parents
in
the
parents
that
are
ordered
based
on
rank.
You
also
have
a
quantification
of
how
much
preference
you
have
for
each
of
those
parents,
so
maybe
you
can
use
it
for
whatever
you
might
want
to
use
it
and
actually
doing.
That
is
not
that
complicated.
We
can
reserve
a
bit
from
the
six
Laura
type
which
and
if
that
bit
is
present,
that
you
know
that
you
also
have
an
optional
bite
for
the
priority
body.
D
J
This
is
actually
a
table
from
the
form
RFC
eight
138
I
just
copied
it
exactly
as
it
is.
So
we
are
just
reusing.
The
definition
of
there
was
no
recent
torture.
We
create
something
over
own,
but
this
thing
actually
is
the
holy
right.
So
you
can
use
one
bit
to
to
indicate
that
this
optional
preference
value
is
present
or
not.
J
J
M
G
A
N
D
J
F
J
Want
three
introduce
myself?
Okay,
so
this
is
another
draft
regarding
load
balancing.
So
this
is
an
objective
function
that
tries
to
load
balance,
notes
based
on
the
traffic
load.
They
experience.
So
again,
thanks
to
Rahul
for
his
review,
we
did
a
small
rename
to
what
properly
reflect
what
this
value
is.
There
is
a
value
which
decides
the
monitoring
for
throughput
over
which
window
is
being
performed
on
original.
It
was
called
the
period
it's
renamed
to
a
window
which
is
more
accurate
because
it's
actually
a
sliding
window.
J
So
there
are
two
things
that
a
node
will
have
to
do:
it
will
need
to
monitor
traffic
over
a
throughput
window
and
that
loop
on
window
size
in
terms
of
time
duration.
It
has
a
time
unit
and
in
the
original
draft
that
was
hard-coded,
and
now
we
added
some
tlvs
that
allow
you
to
configure
that
so
the
route
can
say
you
know
what
I
want
you
to
want
on
traffic
over
five
minutes
or
five
seconds
or
five
hours.
J
You
can
also
have
the
card
coded
or
you
can
use
whatever
external
configuration
you
might
want,
coop
or
whatever
else,
to
also
convey
this
information.
There's
also
mention
in
the
new
version
to
the
oscillation
issues
that
Pascal
mentioned.
So
he
said
that
normally
what
will
happen
is
that
as
to
parent,
one
gets
a
bit
lighter
and
load,
all
other
nodes
will
switch
to
it
and
it
will
get
overloaded
and
then
once
it
get
overloaded
it,
the
traffic
will
switch
to
the
other
one,
which
now
is
lighter
in
traffic
you.
J
So
we
have
one
oscillation
between
multiple
parents
as
an
issue,
and
we
have
mentioned
it.
But
at
this
point
we
don't
have
a
solution
for
that
and
following
Rahul's
reviews
also,
we
have
defined
how
the
bits
in
the
metric
the
PCO
are
a
bits
are
used,
should
be
used.
So
I
mean
this
objective
function.
Works
with
I
did
some
antics
all
that
anyway.
J
J
The
context
of
this
is
that
have
different
objective
functions
like
all-zero
and
mr.
Huff,
which
use
typically
TX
or
generally
additive
matrix,
but
none
of
them
don't
actually
perform
something
that
can
load
balance
based
on
traffic
and
there
is
a
proposal
for
the
load
balanced,
Oh
F
that
you
the
chart
kind
of
apparent.
But
if
you
have
children,
if
you
have
note
which
have
different
traffic
requirements,
that
won't
work
so
we
are
trying
to.
J
J
So
as
an
example
here,
if
you
have
two
potential
daughter
to
connect
to-
and
you
have
that
node
see
it
should
connect
to
dock
to,
since
it
has
more
available
bandwidth,
more
available
resources
for
forward
the
traffic,
but
since
normal
object,
the
conscience
don't
know
about
that,
it
might
collect
to
the
other
one
and
experience
packet
loss.
Now.
Etx
won't
correct
that
over
time,
but
it
will
take
some
time
for
that
to
be
corrected
and
we're
trying
to
help
nodes
connect
directly
to
a
note
that
already
has
the
capacity
to
spare.
J
So
we
are
proposing
a
metric
called
remaining
throughput
and
when
it's
used
is
used
aggregated
with
a
minimum
function,
so
a
parent
node
sense
I
have,
for
example,
the
ability
to
send
five
packets
per
second
and
it
it's
received
to
a
child
node
that
child
node
has
a
capacity
for
word
only
three
packets
per
second.
So
when
that
child
will
report
its
own
remaining
food
metric,
the
value
will
be
three
so
goes
down.
J
Obviously,
as
we
go
further
down,
you
are
using
a
sliding
window
and
that
sliding
window
tracks
to
put
available
to
put
capacity,
and
that's
basically
the
the
capacity
that
you
have
what
you
have
been
using.
So
you
need
to
implement
some
tracking
in
the
code
to
check
what
spare
capacity
hub.
The
idea
is
that
the
objective
function
will
pick
the
as
a
preferred
parent
one
of
the
potential
parents
which
has
the
highest
remaining
flow
hood,
and
you
can
use
it
with
a
hysteresis
function
like
in
mr.
J
The
structure
of
the
metric
is
like
this:
you
have
the
typical
header,
coral
matrix
and
the
remaining
throughput
is
two
byte
unsigned
integer
and
you
can
use
it
as
a
constraint
as
well.
So
you
can
say
that
I
own
to
connect
to
note
that
have
at
least
five
packets
per
second
remaining
report,
and
we
also
added
these
two
optional
DLP's
one
is
to
define
how
big
the
window
is.
J
So
that's
it
in
bits
and
allows
you
to
define
the
window
size
in
time
and
the
time
units
use
for
this
are
also
LV,
which
expresses
the
window
time
unit
in
multiples
of
a
millisecond
its
but
8-bit.
And
it's
the
power
it's
a
power
of
two.
So
if
it's
zero,
then
you
have
one
millisecond
time
unit
it
255.
Then
you
have
until
the
thermal
death
of
the
universe.
J
Finally,
we'd
like
to
we
are
addressing
the
issue
of
enrollment,
so
this
information
can
be
used
when
you're
trying
to
join
a
network
through
a
media
player
to-
and
you
can
convert
this
remaining
throughput
into
a
one
priority
value
like
this,
and
you
can
also
use
it
as
normal
whenever
you're,
initially
turning
a
ripple
Network
to
pick
the
dog
directly
with
which
has
the
maximum
remaining
throughput.
Basically,
if
you're
using
the
aggregation
of
minimum,
whichever
note
you're
connecting
to
it
already
knows,
which
is
the
available
remaining
throughput
through
that
path,
all
the
way
to
learn.
J
F
Told
it
then,
you
know
so
you're
using
the
tree
that
smell
okay.
So
now
say:
the
bandwidth
of
the
link
is
typically
100
right
and
my
my
dad
says:
I
50
left
so
I
see
all
50.
Nice
I
need
five.
Let
me
jump
to
him
and
then
I
children,
so
I
disposed
to
55
right,
because
I
used
five
out
of
the
50
well
left
is
45.
At
least
this.
F
So
he's
post,
45
and
no
children
come
in
come
in
come
in
and
they
join
me
and
they
use
this
45
yeah.
So
far,
so
good
and
now
I
need
to
move
to
another
parent.
They
see
a
parent
which
exposes
that
20
is
left,
I
needed
me,
farms,
yeah,
so
I'm.
Fine!
Let
me
move
to
him,
but
then
the
children
will
be
stopped,
not.
J
Necessarily
so
you
will
be
tracking
your
throughput,
including
what
you're
forwarding,
so
you
will
know
that
it's
not
just
5.
So
when
you,
when
you're
calculating
your
used
throughput
what
you
have
been
using
out
of
your
capacity
you're,
not
counting
just
what
you
are
using
what
is
originating
at
that
host,
but
also
what
you
have
been
forwarding,
at
least
that's
how
we
implemented
it.
So
you
won't
have
that
from
your
right.
If
you
are
just
recording
your
own,
then
you
will
not
have
that
issue.
J
A
O
I
I'm
I'm
working
in
my
PhD
thesis
and
the
routing
protocol
and
I
used
comprising
four
mice,
mine
and
I
rip
out.
There
are
some
interesting
scenarios
where
I
I
think
something
to
say
about
Twitter,
so
we
we
are
in
so
context
or
essence.
We
we
shown
three
scenarios.
The
first
one
is
a
simple
scenario
where
we
just
put
some
notes
together
and
then
after
well,
because
I
am
working
on
a
similar
thing.
O
Okay,
so
first
point
is
that
a
symmetric
and
I'm
not
completely
stupid
there,
it's
mainly
to
theoretical
explanation
of
them.
The
first
one
is
environment
itself.
You
can
have
intestines
that
focus
on
one
up
and
not
on
the
other,
and
the
second
point
is
that
we
have
different
hard
whales
and
it
is
a
ranch
between
the
emission
from
the
transmission
power
and
the
reception
threshold
is
different
between
notes.
O
O
First
now
you
use
it
for
tech
giant
notes.
We
just
put
out.
You
can
note
together
put
one
thing
and
we'll
record
what
happens
doing
to
house.
We
have
a
deputation
layer,
one
UDP
query
response
every
five:
it's
traffic
we've
decided.
We
use
the
feature,
our
utilities,
then
so
it's
a
French,
that's
been
deployed
in
the
labs
in
some
laboratory,
Suites
I
think,
and
we
use
the
notes
from
zr2
to
test
how
thanks
we
use
the
kontiki
limitation
of
rivers
of
we
don't
do
not
tweak
slots,
we
should
choose
the
default
parameters
and
use
mr.
O
O
And
what
we
are
is
with
recording,
so
there
is
a
time
you
know,
isn't
all
you
gots?
What
is
the
number
of
packets
recorded
by
minute
every
minute
per
minute,
and
we
have
at
both?
We
have
the
production,
tremendous
traffic
and
video.
We
have
the
host
shoots,
maintenance
traffic,
so
we
have
at
the
beginning.
Of
course
we
have
many
trial
and
in
many
traffic
to
build
the
tree
and
after
it
is
more
stable
and
we
notice
mainly
three
kind
of
traffic.
First
one
is
the
broadcast
via
traffic.
It
is
trickled
okay,
no
program.
O
We
have
opportunities
three
packets
per
minute
in
the
water
network,
then
we
are
also
hosted
maintenance
for
daos.
We
have
recorded
approximately
ten
packets
per
minute
for
also
network,
so
little
less
hawkish
note
across,
and
then
we
have
noticed.
Czar
says
that
there
is
many
unique
at
the
I/o
packets
in
in
this
situation,
which
is
in
fact
a
pro
being
where
each
node
needs
to
probe
its
neighbors
to
keep
an
estimation
of
the
link
between
between
them,
and
we
have
recorded
that
there
is
approximately
2000
mute
for
this
kind
of
traffic.
O
So,
in
fact,
doesn't
define
explicitly
the
the
usage
of
unicast
di
of
it
expects
an
external
mechanic
to
be
triggered
to
verify
link
properties.
Okay,
also
mr.
function
says
that
we
should
compute
the
path
cost
to
to
each
candidate
neighbor
okay,
but
we
do
not
know
how
to
do
that,
and
so
kontiki
decided
to
implementation
decided
to
use
unique
at
CI
o--'s
for
that,
but
it
is
not
well
defined.
In
fact,
so
on
one
side,
we
need
to
a
way
to
estimate
a
link
listening
quality
between
each
node
and
observable
society.
O
It
is
for
now
left
alone
by
variable
itself,
because
it's
not
really
a
routing
protocol
problem.
It's
more
link
local
problem.
So
the
question
is
there
that
I
rise?
I
I've
raised
questions
money
I
give
themselves
is
that
can
can
go
as
no
as
a
ripple
node
can
I
can
I
receive
a
unique
idea?
What
should
I,
what
should
should
be
my
behavior
to
that?
O
For
example?
Do
they
count
as
redundancy
for
trigger
on
one
side?
Yes,
she
should,
because
this
is
improv,
those
who
knowledge
of
the
topology,
but
another
sides
are
non
broadcast
packets.
So
the
doesn't
count
in
konschak
interventions
accounts,
but
we
don't
really
know
what
to
do
on
that.
Also
what
about
traffic
event?
We
have
seen
that
it
is
20
packets
per
minute,
so
it's
bigger
than
than
what's
when
ripple
yeah
reset.
Does
it's.
F
About
nature
should
counted.
This
is
not
in
the
protocol.
Right
I
mean
they
are
sending
by
using
this
message
as
a
probe
for
the
mink.
It
appears
because
we
expected
neighbor
discovery,
sorry
for
doing
that,
or
a
click
layout
nourishment,
but
basically
trickle
is
about
eliminating
broadcast
di
ohms
right,
because
you
don't
need
too
many
you're
sure
that
this
guy
will
receive
the
broadcast
da
yo
if
you've
seen
four
or
five
or
around
you
sure.
F
It's
a
big
well
because
at
the
edge
of
the
network's,
not
true
but
inside
the
network,
if
you've
heard
like
5
di
O's,
you
say:
oh,
my
neighbor's
got
it.
Let's
do
5
so
I
have
no
points
anyone,
but
if
they
were
a
unicast,
the
I
use.
My
neighbors
did
not
get
them
because
they
didn't
care.
They
were
unique
as
for
me,
so
they
don't
count
us
redundancy.
They
are
not
redundant
the
dirty
dozen
seeds
for
me
not
to
send
my
broadcast,
because
this
guy
already
got
it
from
this.
D
D
O
O
O
M
Why
not?
Okay
he's
really
nice,
so
so
he
it's
a
real
strange
thing
that
he
did
that
he
doesn't
Kentucky,
and
so
what
you
measure
is
not
really
representative
six
five
five
zero
and
the
tricks
that
he
plays
are
compliant
but
they're
tricks
that
are
not
explained
exactly
here.
So
you
will
not
find
in
six
five
five
zero.
M
He
starts
by
unicasting
the
iOS
all
of
his
neighbor's.
So
he
gets
an
acknowledgment.
So
he
is
sure
that
his
neighbors
haven't
updated
the
information
I'm,
that's
very
fuzzy
as
I'm
saying,
because
I
don't
I,
don't
really
remember,
but
I'm,
just
telling
you
you
know.
Okay
will
not
find
the
answer
in
c50
and
the
fact
that
you
have
so
many
UCSD
iOS
is
not
representative.
Six
five
zero.
It's
representative
of
the
tricks
that
Seema
had
to
play
to
have
to
get
through
those
five
9s
download
reliability
paper
at
your
setting.
Okay,.
O
D
Right,
thank
you
so,
based
on
my
understanding
of
what
Connecticut
does
is,
it
tries
to
prove
whether
the
preferred
parent
is
still
there
and
the
probing
duration
is
all
configurable
so,
but
it
depends
upon
your
scenario.
So
if
you
actually
have
unique
edge
traffic,
then
it
should
not
come
into
which
so,
but
in
your
case
the
UNIX
traffic
is
every
five
once
every
five
minutes,
or
something
like
that.
So
maybe
that
is
why
it
is
coming
in
the
picture.
D
O
O
O
Where
is
the
question?
We
expect
the
link
layer,
detects,
asymmetry
and
blacklist
go
kind
of
like
it
interrupts
the
communication
between
them.
But
the
point
is
that
the
link
layer
cannot
detect
that
the
link
is
a
symmetric
because
the
packets
are
broadcasted.
So
no
not
acknowledged
so
only
a
ripple
knows
that's
the
link,
it's
a
guess
that
the
link
is
that
symmetric,
because
we
send
you
messages,
and
so
now
so
we
mod
know
continue
to
send
GIS.
O
D
Shadow,
so
this
is
a
very
interesting
topic
action.
So
you've
mentioned
black
testing
here,
but
black
listing
is
difficult
to
do
because
it
requires
number.
So
you
once
you
need
even
to
start
keeping
the
MAC
address
and
the
flag.
It
means
a
lot
of
memory
based
on
the
density
of
the
nodes.
It's
not
very
easy.
D
F
Basically,
the
point
is
the
node
should
Swatow
it's
the
iOS,
so
asked
not
window
sensor
Mayo's
and
that's
exactly
what
it
does,
because
we
have
Treecko
and
you
have
a
nice
hero
and
I
mean
Facebook,
and
and
so
basically,
if
you,
if
you
have
network,
while
you
think
you
can
have
this
sort
of
prime,
don't
put
the
minimum
trick
on
too
low
and
that's
exactly
what
it
does.
It's
wattles
the
initial
FDR
used
to
the
minimum
the
I
mean
now.
F
Can
this
be
improved?
Yes,
some
people
were
working
on
it
and
the
fact
to
reset
the
I
I
zero
since
Robin.
Is
it
a
good
idea?
Well,
it
can
be
debated
that
point,
but
I
don't
think
we
need
to
blacklist
anything
like
it's
a
way
to
get
more
attacks
against
our
devices.
Should
we
throttle
yes
I've?
Actually,
we
do
with
triple
strategies.
So
it's
all
a
matter
of
to
do
should
be
reduced
at
Rico
and
each
situation
and
there
is
a
draft
in
there.
So
it's
an
interesting
observation.
O
F
F
Isn't
it
I,
don't
know
it's
steep
as
much
as
we
could
think
about
no
protections?
Our
rotor
can
protect
itself,
there's
a
point
where,
if
there
is
a
babbling
idiot,
which
is
what
we
call
this
thing
which
keeps
babbling
idiot
with
the
Geisha,
some
network
management
of
any
form
should
realize
something's
wrong.
Take
the
device
out.
We
can't
program
into
small
devices,
all
the
situations
and
now
at
some
point
in
this
network
management,
there
is.
F
Whether
we
should
know
about
the
symmetric
links
and
do
something
about
them,
I
think
yes,
I
have
a
draft
and
others
on
the
shelf
somewhere,
but
actually
Charlie
found
at
some
point.
That's
why
we
call
devious
means
yes
to
do
things
in
this
matrix
and,
yes,
we
know
what
to
do
with
that.
Now
purely
mute,
no
account
because
you
cannot
well
maybe
sounded
this,
but
you
cannot
do
anything
else
and
until
you
have
a
three-wide,
and
so
until
you
have
asserted
that
you
can
talk
to
the
water,
but
maybe
Sunday
this
you
can
yes,
yeah.
F
Right,
you
have
not.
The
same
range
now
understand
that
this
ice,
which
we
can
happen
with
pure
mute.
You
very
well
case
as
I
thought.
Don't
you
know
how
repo
we
watch
it
in
recovered
that
there
is
an
assert
that
the
link
is
be
directional
before
you
can
use
it,
and
they
all
say
you
telling
me
yeah
you're.
G
F
G
F
F
F
O
Completely
mutant
know
these
doesn't
think
the
problem
is
the
network
at
all.
We
just
approximately
knew
it
that
it's
this
no
data,
muted
node
at
the
beginning
of
the
experiment,
exam
and
after
15
minutes
we
just
silent
it
and
it
just
only
can
only
communicate
with
33,
which
is
it's
produced
successor,
so
we
just
have
a
canary
occupy
with
other
and
what
we
with
the
same
scenario
as
a
Death
Note,
we
just
kind
of
traffic.
O
So
we
we
see
that
we
have
almost
no
losses,
except
from
the
to
concern,
notes
during
10
minutes
during
the
local
right
bank
but
yeah,
and
we
have
a
little
more
rigid
control
traffic.
But
such
a
detective,
and
so
the
only
point
is
that
it
is
a
real
limited,
topology
change.
So
we
can
just
check
if
we
have
no
too
many
packets,
but
we
I
I,
think
you
just
confirm
me
I!
Think
that's
because
presentation,
I
think
only
the
two
notes,
the
trigger
of
the
tuners,
are
accelerated.
O
O
O
F
Pascal,
there
are
two
questions
really
which
cannot
fall
out
of
this
presentation.
The
first
one
is:
what
do
we
do
with
this
document,
but
I
actually
a
completely
agree
with
that
document.
There
is
something
to
be
done
about
this.
We
understand
I
want
to
sell
it
when
to
receptacle
timer
all
these
things,
and
we
have
kind
of
discussed
around
the
draft
and
then
it
kind
of
stale,
but
it's
something
we
really
need
to
understand
how
to
really
use
this
correctly.
This
is
why
you
should
sell
it.
This
is
why
you
should
not.
F
F
No
it's
a
lot
of
that
is
on
paper
like
if
you
have
a
blackout
in
your
meters,
it's
50
meters,
not
great.
We
black
out
and
now
the
power
comes
back
and
all
the
matters
in
the
city
start
sanding
this
and
your
Diaz
won't
progress.
They
will
be
interfered
with
by
all
this.
Yes,
very,
very
bad
idea:
don't
do
it,
on
the
other
hand,
I'm
putting
this
new
meter
on
my
world,
it's
the
only
one,
that's
new
in
the
network,
I
mean
and
the
guy
is
putting
it
and
waiting
for
it
to
join
the
network.
F
Should
he
stay
in
my
home
for
four
hours
waiting
for
the
next
day?
Oh,
probably
not
so
that's
when
you
should
press
the
button
on
the
matter
so
that
this
is
sound,
the
metal
joins
the
network.
The
guy
can
leave
so
so
there
is
not
too
serious
cases.
What's
valuable
in
cases
right
should
not
be
done.
We
did
not
determine
I'd.
B
Is
more
than
that
actually
yeah,
we
have
this
repo
beer,
and
that
was
quite
a
lot
of
enthusiasm
on
it
and
I
think
we
should
like
to
continue-
and
this
is
the
moment
to
see
if
people
are
interested
to
continue
and
that
you
can
actually
make
some
decisions
on
how
to
go
forward.
That
will
be
lovely
so.
C
C
So
we
had
a
few
meetings
and
and
and
and
the
minutes
are
on
the
ether
pet,
but
we
we
hadn't,
really
come
to
conclusions
of
document
wise,
what
to
do
as
the
next
steps
and
and
and
maybe
that's
exactly
what
we
should
get
back
to
I.
Think
we
had
a
very
good
think
upon
a
lot
of
understanding
of
the
technical
details.
G
C
As
it
seems
to
be
somewhat,
can
you
nudge
this
to
go
full
picture
so
last
time
we
we
said
these
were
the
relevant
documents
that
we've
basically
been
involved
in
the
process.
I
think
we
can
go
back
to
those
and
figure
out
if
that's
still
true
and
I
think
Pascal
would
know
more
about
this
than
I,
because
they're
pretty
much
most
of
of
most
of
all
of
them.
So.
C
Right-
and
maybe
you
know
as
a
starting
point,
if
we
can
just
say,
let's,
we
have
a
better
understanding
of
all
the
other
things,
but
we're
not
trying
to
bring
them
into
this
draft.
But
let's
figure
out
how,
in
this
draft,
we
can
merge
all
the
necessary
pieces
where
I
think
it's
primarily
the
B,
the
C
cast
pieces
right.
You
have
a
bunch
of
sections
that
are
empty
there,
that
are
just
stakeholders,
and
you
know
so
will
primarily
in
question
to
Carsten.
If
he
thinks
it
makes
sense
that
we
create
out
of
drafts.
C
Word
Roll
beer
emerge
document
where
we
also
have
the
you
know
non
IP
multicast.
They
adjust
the
role
layer,
functionality
that
you
also
have
in
the
Seekers
proposal
right
so
that
we
basically
then
have
one
document
that
describes
the
the
functional.
You
know,
definitions
right,
so
the
the
specification
part
there
was
the
other
part
right.
So
if
in
the
document
may
we
should
try
to
work
out
which
sections
are
informational
which
are
normative
right
so
because
there
is
some
encapsulation
defined.
C
But
then
basically
we
get
to
a
point
of
knowing
what
is
really
the
things
that
we
want
to
specify
and
once
we're
complete
with
the
specification
we
can
figure
out
if
all
the
explanations
should
stay
there
or
if
we,
for
you
know
further
explanation
of
into
some
form
of
architectural
informational
document
and
strip
this
down
even
further.
But
in
the
first
place,
let's
maybe
try
to
work
on
code
complete
or
the
normative
pieces
of
the
role
pieces
with
the
with
the
normal
bit
strings
and
then
with
the
heuristic
bit
strings.
But
what
do
you
think.
C
Wait
a
second
can
I
just
interrupt
it.
Yes,
would
it
be
fair
to
say
there
is
really
nothing
substantial
about
bloom
filters.
Acceptance
proof
existence
in
your
document,
I
mean
no
real
technical
details
about
them.
Is
that
fair.
C
C
K
K
B
C
K
K
I'm
not
talking
about
the
the
IETF
I'm
talking
about
we
are.
This
document
originated
instructive
in
this
is
like
six
years
old
right
and,
of
course,
people
can
might
still
have
petted
a
part
of
it
since
steam
published
for
a
long
time.
I
don't
care
that
much
about
that,
but
at
the
time
it
took,
but
again,
I
don't
have
a
feeling
for
the
risky
level
that
is
energy
when
you
I.
F
Agree
Pascal
to
be
here,
I
agree
with
Gaston
that
we
need
to
look
at
IPO
and
in
particular
for
the
normative.
So
maybe
the
question
you
asked
her
earlier
was
having
a
a
document
which
gives
you
the
big
picture
on
how
things
work
together.
I
keep
that
informative,
even
if
all
you
describe
as
I
care
on
it,
I
think
I'm
perfectly
happy
with
that
and
then
probably
when
we
start
describing
or
what
you
do,
if
you
use
bluem
or
what
you
do,
use
the
explicit
bits,
etc.
F
So
if
people
wants
to
want
to
do
only
a
fourth
like
our
Foursquare
business
story,
if
you
just
want
to
implement
one
of
these
four
boxes,
it
happens
that
this
box
doesn't
have
a
cure.
If
we
package
the
four
boxes
in
one
normative
specification,
we
will
be
implementing
the
document,
which
cleverly
has
a
lot
of
regularity,
and
so
I
was
just
wondering
whether
we
should
look
at
which
squares.
We
want
to
start
with
and
and
limit
a
document
to
one
of
those
squares
that
will
immediately
Albert's
attacks,
because
there
will
be
a
cottage
so.
C
So
this
is,
this
is
orthogonal
to
what
I
was
thinking.
My
thinking
was
that
through
the
last
year
month,
at
least,
my
understanding
is
that
there
is
a
humongous
amount
of
explanation.
If
you
want
to
explain
the
stuff
to
people
who
don't,
for
example,
have
all
the
background
of
ripple
and
don't
want
to
read
a
thousand
pages
of
ripple
stuff,
for
example
right
so,
which
is
something
I'd
love
to
do,
but
I'm
not
sure
if
I
have
the
time.
So
that's
when
basically
we
were
visiting
what
what
could
we
do?
C
Is
the
next
step
saying
that
you
know
we
may
not
be.
You
know
code
complete
on
good
explanations
anytime
soon,
but
we
could
maybe
be
much
easier
code
complete
on
the
specification
part,
even
if
that
just
means
the
specifications
only
understood
by
the
people
who
been
involved
in
the
process
right.
So
that's
why
we're
saying?
Can
we
first
focus
on
the
specification
parts
and
now
you're
saying
the
specification
part
still
can
come
in
four
quadrants,
and
maybe
we
don't
capture
all
the
quadrants
of
the
specification.
C
Look
I,
don't
have
a
commercial
head
on
for
for
this
stuff
here
right,
so
that's,
oh,
no
understood,
but
I
think
that
I
mean
unless
you
want
to
create
ultimately
four
four
quadrants
for
different
documents
right,
but
even
then
some
somebody
who's
nasty
can
simply
claim
you
know
IPR
against
all
four
of
them
and
leave
it
up
to
you
know
your
company's
lawyers
that
you
want
to
implement
you
to
figure
out
now.
Sorry,
this
guy
is
full
of
you
know.
Basically,
his
patent
only
applies
to
this
one
out
of
four
documents.
C
So
but
but
fine
I
mean
it's
it's
it's,
hopefully
a
design
team
right
and
we
hopefully
make
this
up
if,
if
we
start
specifying
one
particular
quadrant
and
then
basically,
either
run
out
of
steam
or
enough
member
of
the
team
say:
okay,
let's
for
IPR
reason,
others
not
do
the
other
quadrant
and
we
stop
there
right
and
call
that
data
first
win
and
then
try
to
format
the
document
accordingly.
I'm
just
saying
what
is
because
we've
been
doing
a
lot
of
high-level
all
over
the
place,
analysis
and
details,
and
how
can
we
get
focused?