►
From YouTube: IETF94-ROLL-20151105-1740.webm
Description
ROLL meeting session at IETF94
2015/11/05 1740
B
B
B
C
A
A
We
found
in
a
number
of
use
case,
but
growing
industrial,
that
that
something
like
ripple
is
very
important
to
form
an
initial
initial
writing
topology,
so
that
you
can
actually
talk
between
the
devices
and
a
controller.
But
then,
when
you
want
to
create
control
loops
in
your
factory
or
optimize
path
for
any
sort
of
reason,
they
are
having
the
capability
to
compute.
Those
routes
from
the
standpoint
of
an
external
resource
like
a
PC
is
actually
very,
very
valuable.
A
So
next
slide
please.
So
we
came
up
with
this
draft,
which
basically
uses
the
ripple
mechanism
of
the
dowel
to
establish
a
rod
which
could
be,
in
one
hand,
shortening
a
non
storing
path
and
that's
kind
of
easy,
because
in
non
stirring
the
topology
is
known
to
the
root,
but
with
the
new
version
of
the
document
we
can
also
establish
push
rod
which
can
go
in
storing
an
on
stirring
across
the
network
so
as
to
create
a
rod,
optimized
path
between
two
nodes.
A
A
Yes,
so
there
is
work
which
is
starting
at
that
net
for
deterministic
networking
in
the
rotting
area,
and
this
work
will
enable
to
expose
a
topology
of
devices
to
a
central
controller
at
the
central
controller,
compute
a
path
and
then
push
that
path
into
the
field
devices.
In
the
case
of
a
ripple
network
like
a
sixties,
networker
or
any
other
repo
network,
it
is
not
clear
yet
what
the
mechanism
could
be
to
actually
install
those
routes.
A
But
if
you
have
a
constraint
device
that
already
odds,
though
messaging,
then
it
would
make
sense
to
actually
use
the
same
Dow
exactly
to
populate
this
path
into
into
the
routers
oops
fast.
So,
just
to
tell
you
tease
is
this
working
group
where
the
topology
model
is
actually
being
developed
for
the
HF,
so
one
of
the
goals
that
sixty
has
is
to
help
cheese
document
a
topology
model
which
would
include
the
needs
of
our
low
policy
network
devices.
A
Yes,
we
already
so
so
here
is
so
I
hope
people
knew
the
previous
proposal.
If
he
goes
through
the
slides,
you
will
see
the
flow
for
non
storing,
which
were
which
was
presented
last
time
and
I
won't
come
back
on
that
because
it
was
already
presented,
is
still
the
core
of
this
document,
but
this
new
release
of
document
as
this
new
generic
route
optimization
as
well,
please.
A
So
it's
a
compliment.
What
we
had
so
so
we'll
get
this
this
bag,
which
is
tree
in
this
case
that
that
was
built
for
repo
next
slide.
So
if
we
are
using
non
starring
mode
and
by
luck
we
want
to
stay
on
the
packet
from
this
guy
here
to
this
guy
here,
the
packet
will
first
follow
the
default
route,
all
the
way
to
the
root,
and
then
the
route
will
add
a
source
routing
header
which
will
enable
to
flow
down
to
this
node
actually
missing.
Yes,
it's
actually
missing
a
little
bit.
A
A
A
A
11
here,
which
means
that
if
this
guy
thinks
this
guy
to
go
all
the
way
to
the
common
parent
and
down-
and
you
see
that's
quite
a
road
stretch,
so
the
right
strategy
is
important
in
ripple,
because
that's
how
we
stay
routing
state,
that's
how
we
avoid
knowing
all
the
movement
which
can
happen
in
this
appalling
topology,
which
mostly
care
for
the
default
route
throughout
the
route
and
and
the
dowel
rods
towards
the
children.
That's
all
we
care
for,
but
the
price
to
pay
is
that
we've
got
this
extensive
pass
in
some
particular
use.
A
Cases
when
you
want
to
do
node
to
node
ripple
is
not
advisable
now
to
node
next
night.
So
what
the
proposal
says
basically,
is
you
can
project
a
dau
I
supposed
to
inject
now
in
topology?
You
can
project
it
and
projecting
the
doll.
What
goes
right
away
to
the
egress
of
the
tea
path
that
you
want
to
inject
next
night
and
I
said
ow.
A
It
will
go
on
the
reverse
direction
of
the
path
that
you
want
to
set
between
this
guy
and
this
guy
next
like
and
the
Dow
flows
to
the
increase
of
the
path
and
sets
up
rotting
state,
as
it
does
normally
in
ripple
it's
light
and
the
dialog
could
basically
flow
all
the
way
to
the
de
Groot
next
slide.
So
with
this,
we've
basically
installed
a
route
in
41.
We
are
42
to
this
guy
here,
each
time
I
don't
see
the
numbers
can
use.
A
So
again,
we
are
talking
about
writing
to
mr.
53
here,
and
so
what
the
doused
has
done
here
is.
It
has
installed
a
state
in
42
to
route
253,
VI,
43
and
stating
41
to
route
253
via
42,
and
this
basically
enables
a
shortcut
between
451
and
maybe
other
children,
the
41
53,
so
to
enable
this
sort
of
installation.
It's
important
that
the
root
or
PC
which
talks
to
the
root
is
aware
of
the
capability
of
those
devices
of
the
neighborhood
information
between
those
devices
etc.
A
So
it
has
enough
information
to
to
build
this
particular
path,
and
that's
why
I
mentioned
that
the
world
has
done
a
death
neck.
That
at
ease
can
be
actually
a
necessary
component
to
the
whole
solution
and
that's
basically
the
addition
to
to
the
work
that
was
proposed
last
time
for
memory
and
next
slide.
Please
for
memory.
Last
time
we
are
just
following
the
nonce
touring
mode
and
enabling
a
shortcut
like
this
along
the
non
starring
path
and
the
reason
was
to
make
the
Surat
either
shorter
in
case
of
a
long
path.
A
We
are
using
the
exact
same
method.
What
we
have
added
here
is
the
generic
capability
to
install
as
a
dhow
path
from
any
node
20
node
under
the
control
of
a
remote
control,
and
that's
it
you'll
find
you'll
find
the
next
slide.
The
slides
I
presented
this
time.
It's
just
for
me.
Maria
one
goes
through
them.
Is
that
so
I
think
I
have
questioned
the
very
young
you
can
scroll
to
the
very
end
of
my
slide.
So
this
is
what
I
won't
present
again,
but
please
feel
free
to
to
go
snow.
A
You
don't
have
to
do
for
that.
Now
appreciate
free
to
go
through
them
and
come
back
on
the
mailing
list,
but
you
know
the
questions
that
we
had
since
last
time.
The
first
question
was
terminology,
and
the
the
answer
so
far
is
that
the
term
projected
route
is
winner,
but
if
you
have
other
ideas
about
how
you
could
call
this
thing,
second
question
was:
do
we
need
a
new
mode
of
operation?
A
What
actually
went
when
you,
you
install
a
state
in
ripple
normally
storing
a
non
storing?
This
is
something
in
the
middle
because
it
could
operate
in
a
non
stirring
mode
to
do
a
shortcut
or
it
could
operate
in
story
mode.
Is
it
a
new
mode
or
is
it
something
that
everybody
should
support.
A
Yeah,
the
last
thing
was
the
one:
once
the
path
is
established.
There
is
a
message
that
flows
from
the
ingress
of
the
path
to
the
route,
to
say:
okay
or
down.
Should
that
be
a
dark?
Is
it
the
semantics
that
we
expect,
or
should
that
be
the
continuation
of
the
Dow
or
something
like
that?
What
message
and
right
now
the
drive
size?
A
The
is
the
right
way
following
us:
okay,
so
so
this
is
basically
a
summary
of
what's
new
in
this
draft,
and
what
I
wanted
to
do
here
is
for
this
group
for
interest
in
this
work.
I
mean:
do
people
think
that
storing
and
not
stirring
a
sufficient
or
do
people
think
that
going
to
a
controller
based
model
is
of
interest
in
some
of
the
topologies
they
care
for
Fayetteville.
D
D
C
D
D
A
Completely
agree,
I
mean
so
far,
I've
put
can
doubt
the
dependency
under
the
cheese
kind
of
information,
so
cheese
is
able
to
describe
a
topology
and
we
would
extend
that
to
to
put
more
capability
into
that.
So
we
would
know
how
many
buffers
are
viable,
this
device,
how
many
rot
and
trees
I
vibe
on
this
device.
So
now
we
can
know
how
many
of
those
optimization
we
can
actually
populate
into
any
particular
device.
That's
part
of
what
we
want
to
do
for
60s
at
Beth,
net
and
I.
Think
it's
wrong,
actually
a
complete
story.
C
Any
anything
else
from
the
room
about
this
would
would
love
to
have
comments
on
the
list
as
to,
if
you
think
this
deserves
a
recharter.
If
so,
then
we
will
write
one
and
if
not,
it
would
be
good
to
know.
Also,
this
is
important
enough.
I
will
implement
it.
Things
like
that
would
be
very
useful
to
know
coming
on
the
list
to
gauge
whether
we
should
engage
in
this,
because
it
would
essentially
involve
keeping
this
working
group
open
for
at
least
I
think
another
year.
Otherwise
we
would
we
would
close
very
soon.
A
If
I
may
react
on
that,
there
are
actually
a
number
of
very
exciting
things
which
could
be
done
in
this
working
group.
I
would
be
very
sorry
if
it
closes
so
for
me.
Obviously,
this
worker
is
important,
but
the
work
are
having
non
ripple
leaves
as
well
as
a
very
important
work.
So
I
don't
know
if
it's
feasible,
but
at
least
that
looking
at
that
is
quite
an
important
work
documenting
when
to
use
an
capsulation
and
everything.
The
draft
that
you
have
produced
is
also
a
very
important
work
for
this.
D
So
I
don't
know
if
we're
in
the
middle
of
a
charter
discussion
right
now,
I
just
wanted
to
mention
that
the
year
ago
we
submitted
this
internet-draft
hard
constraint
cast
and
we
haven't
really
pushed
for
for
it
to
go
forward,
but
it
would
fit
well
with
a
new
road
working
group
that
is
focusing
on
one
story.
Mode
I
did.
C
C
C
So
partly
lacking
anybody
else
to
complete
this
work,
no
pink
skirt
after
staying
I
guess,
Inez
and
I
undertook
to
take
this
on
and
we
did
have
a
virtual
interim
meeting,
which
was
it
not
well
attended,
but
the
point
of
it
was
actually
to
get
the
the
the
work
done,
so
that
was
in
September
29th.
I
think
it
was
and
we
went
through
a
bunch
of
things,
so
this
was
a
reference.
C
This
is
a
reference
diagram
that
I
wanted
to
make
sure
I,
explain
the
room,
understood
and
was
clear
because
I
basically
just
copy
and
pasted
this
about
12
times
afterwards.
So
in
this
diagram
I
nevered
the
nodes
I
like
Pascal's
numbering
so
I
use
that
and
so
I
have
a
6lb
are
at
the
top
help
sighs
you're
the
right
way
up
here.
I
have
some
kind
of
a
backbone
here,
I'm
I'm
a
little
I'm
going
to
be
a
little
bit
vague
as
to
whether
it
runs
ripple
or
not.
C
But
over
here
I
put
a
note
over
here,
which
perhaps
is
ripple
aware.
So,
for
instance,
it
make
might
speak.
You
might
be
crippled
across
across
this
this
network,
as
well,
which
would
allow
this
guy
is
to
see
that
essentially
appear
to
be
a
leaf,
and
then
I
stuck
the
internet.
Here,
we've
all
heard
of
that
and
some
generic,
no
doubt
there
on
the
internet
on
at
the
bottom.
Here
I
have
two
nodes:
q,
&
N,
and
they
are
6lowpan
only
ND
nodes,
so
yeah.
So
it's
trying
to
chase
me
this
camera.
C
C
So
what
we
try
to
do
is
walk
through
the
rules
that
we
know
that
six
pan
has
given
us,
but
when
we
can
add
and
subtract
extension
headers
and
when
we
need
to
use
IP
IP
encapsulation
to
do
that,
and
so
this
basically
summarizes
it
that,
basically,
if
you
can
only
add
or
subtract
headers,
if
you
are
either
the
originator
of
the
packet
or
the
destination
of
the
pack,
and
if
you
want
to
do
with
something
else,
you
need
to
put
an
IP
IP
header.
On
top
of
that,
it
seems
pretty
obvious
and
simple.
C
You
know
it's
a
pretty
simple
rule,
and
yet
it
turned
out,
as
we
went
through
the
list,
that
it
wasn't
as
obvious
as
we
thought
and
that
the
ramifications
were
not
as
obvious
as
we
thought
and
furthermore,
that
the
number
of
times
when
you
could
have
traffic
in
a
lln
without
an
IP
IP
header
was
remarkably
slim,
and
that
was
quite
disappointing
because
it
sounds
stupid
to
design
a
network
where
you
always
put
IP
IP
headers
in
all
over
the
place
when
you
probably
shouldn't
so
next
slide.
D
C
D
C
So
I
would
love
to
do
that.
Kirsten
I,
hope,
you're
volunteering
to
to
think
about
this
a
little
bit
or
write
something
up
because
to
first
order.
We
did
do
a
little
bit
of
that.
Didn't
we
a
year
ago
or
more,
when
we
thought
we
were
going
to
get
we're
going
to
deal
with
our
well
compression.
You
know
issue
right
and
sorry.
C
Flo
label
flow
label
as
instanceid
issue
right
when
we're
going
to
try-
and
we
thought
we
were
sat-
was
going
to
deal
with
the
problem
because
that
gets
rid
of
an
IP,
IP
header,
and
so
we
ran
up
against
a
well
and-
and
so,
if
you
want
to
reopen
that,
if
the
working
group
feels
that
that's
a
viable
means
because
we
gone
down
another
route
at
this
point,
we've
gone
down
the
other
route
of
doing
it
properly
and
we
get
to
a
relatively
unacceptable
answer.
And
so
maybe
it's
more
reasonable.
Go
back.
C
So
I
want
to
make
sure
they're.
Working
with
the
purpose
of
these
slides
is
to
make
sure
the
working
group
understands
the
answer
that
we
got
to
so
we
went
through
the
set
of
storing
and
non
storing
ripple,
aware
leaf,
non
ripple,
aware,
should
say,
leaf,
root,
internet,
that's
us
sources
and
then
the
same
thing
as
destinations,
but
we
mitad
all
the
cases
that
were
internet
to
internet
because
they're
not
really
interesting
right
because
they
don't
go
through
the
Ln
right.
They
just
do
the
comet
oryx.
C
So
you
get
basically
four
times
three
times
two
up
here
and
we
get
two
cases
non-store,
storing
and
on
storing
mode.
So
you'll
find
in
the
document,
inez
authored
all
of
these
cases
and
we
went
through
them
in
the
virtual
interim.
We
went
through
all
of
these
cases
and
wrote
them
all
in
ok.
What
does
it
happen?
C
What
headers
go
where
what
happens
here
next
slide
Inez,
so
I'm
just
going
to
couple
a
couple
things
and
I
colored
them
in
the
previous
slide
to
indicate
which
one
we
went
through
and
we
could
spend
more
time
going
through
more
of
them,
but
I
color
code
them
a
little
bit
green.
So
there's
no
problem
this
one.
So
let's
talk
about
story
mode
from
a
ripple
aware
leaf
to
root.
This
would
be
a
simple
case
right,
because
this
is
your
metering
case
right.
You
have
a
traffic
from
your.
C
Your
electrical
meter
goes
back
to
headquarters
to
the
root,
and
that's
why
I
kind
of
stuck
here
this
here
because
maybe
verge
that's,
really
the
the
root
or
really
the
destination
of
the
traffic.
Ok,
that's
the
collection
box,
how
it
gets
there.
It's
not
so
important.
What
matters
is
the
traffic
comes
up
so
simple
case
IP
the
rpi,
because
you
need
that
in
story
mode
to
avoid
to
discover
loops
and
you're
up
early
earlier
payload
boom
boom
goes
through
this
guy's
RPL
aware.
The
important
thing
about
that
is.
C
C
Don't
know
why
we
didn't
do
that
at
the
beginning,
because
to
me,
I
can't
think
of
a
reason
why
you
would
never
want
that
if
we
change
that
today
we
would
have
a
flag
day
unless
we
also
changed
it
in
the
context
of
compressive
compression
I
here
so
I
see
some
nodding
from
Robert
and
I
really
want
to
hear.
Why
tell
me
what
you
what
you
this
is.
What
their
point
of
here
is
this
our
time
here
today
is
so.
C
D
C
But
the
rules
in
the
rules
in
24
succeeds
that
tell
us
exactly
how
to
do
that.
If
you're
telling
me
it
will
get
dropped
because
someone
ignored
those
rules-
okay,
that's
fine,
but
they
also
could
drop
it
because
it
wasn't
TCP
right,
just
the
whole
world
that
runs
on
TCP,
so
it
could
wrap
it
too.
So
I
agree
with
you
on
that
point.
Okay,
let's
continue
on
next
slide,
okay,
so
a
few
problems,
so
we
go
from
story
mode,
RPL
leaf
to
internet.
So
there's
your
case.
Okay.
C
To
get
that
to
work
today,
you
need
to
put
an
IP
IP
header
with
an
rpi.
No
22
needs
to
know
that
the
traffic
is
out
of
the
rip
out
of
the
lln.
How
do
we
know
that?
Well,
I
thought
it
was
obvious
writing
code.
It
doesn't
match
the
pio.
The
destination
doesn't
match
the
pio
that
I
was
given.
I
have
to
use
my
default
route.
Therefore,
it's
not
in
my
all
n.
Therefore
I
have
to
do
this
sounds
reasonable,
Mike
break
if
you
have
more
complicated,
addressing
schemes
or
stuff
like
that.
C
Okay,
so
it
goes
up.
Goes
up
this
guy,
the
6
lb
are
removes.
The
blue
part
relieves
the
red
part
and
goes
out
there
red
part.
Generic
internet
node
we're
good.
Okay.
That
works
does
require
awareness
here.
Okay,
that
he's
sending
traffic
outside
of
the
ll
n
next
slide.
Please
so
non
story
mode
internet
tune
on
our
PL
o--
air
leaf.
So
this
is
one
of
the
first
cases
it
becomes
interesting.
Very
interesting,
so
traffic
comes
from
here
goes
to
the
root
in
non
story
mode.
C
D
Other
ones,
youth.
B
C
Right,
fair,
fair
enough,
good,
good,
just
good
point
I'll
try
to
do
that,
so
6lp
are
realizing.
The
host
is
attached
to
26
puts
an
IP
IP.
It
has
to
put
a
route
route
header
in
to
get
it
there.
The
RP
is,
we
realized,
is
are
sometimes
optional.
If
you've
only
got
one
instance
ID
arm,
then
in
your
in
on
storing
mode
and
you're
giving
an
explicit
route,
you
can't
have
a
loop
with
an
explicit
route:
it's
not
a
loose
source
route.
It's
an
it's
a
fully
enumerated
route.
C
So
when
you're
out
of
right
items,
if
it
somehow
doesn't
get
there
you're
right
a
lot
right,
so
I'm
told
are
some
implementations
that
in
context
would
would
omit
the
rpi
or
could
or
omit
the
rpi.
But
I've
put
it
here
anyway.
So
and
the
roadheader
says
basically
goes
to
13
and
then
26
here
26,
the
IP
IP
packet
is
addressed
to
node
2
leaf
26
leave,
26,
says
packet
for
me,
I
see
it
has
some
stuff
in
it.
I
will
remove
that
stuff.
C
It'll,
look
at
this
IP
header
and
go
aha
I'm,
sending
it
to
node
host
in
and
it's
done
right
great
next
slide.
Please,
let's
try
the
same
thing
in
story
mode:
hey,
let's
get
rid
of
story
mode.
This
is
the
solution
because
it
solves
the
problem.
Actually,
so
same
thing
comes
up
here
we
have
to
put
an
rpi
in.
We
don't
need
an
rh
three,
because
it's
storing
mode.
C
D
C
D
C
C
So
so
so
I
said,
there's
two
so
there's
two
two
answers
here:
there's
a
third
pause,
which
is
that
if
we
change
the
rpi
extension
to
be
23
instead
of
63,
then
that
would
that
would
correspond
to
the
extension
header,
which
is
to
be
skipped.
If
you
don't
understand
it,
okay.
D
C
C
C
D
Prabhakar
you
so
we
went
through
this
one
and
it
is
a
long
time
ago.
So
I,
don't
remember
all
the
details.
We
did
all
this
for
Zeke
VIP
and
one
of
the
ways
of
doing
is
in
the
IP
IP
header
is
just
you:
do
it
singlehop
effectively
and
your
kind
of
reconstructing
it
every
time
and
you
so
you
apply
an
IP
IP
header,
so.
D
B
B
C
C
Right
so,
but
I
just
wanted
to
be
restate
what
I
think
you
just
said
to
as
well,
so
that
means
0
sends
an
IP
IP
header
addressed
the
13
and
13
says:
oh
wait:
I
have
to
add
an
IP
I
have
to
remove
it
then
I
have
to
add
it
ok
strip
it
off
at
it
and
now
I
address
it
to
26,
yeah,
ok
and
now
26
is
happy
to
remove
it.
Yeah.
Ok
that
may
solve
the
problem.
Okay,
I!
Don't
think
that's
written
down
anywhere!
Okay!
C
C
C
13
does
its
stuff
again
that
IP
IP
header
as
you,
this
is
non
story
mode,
so
it
doesn't
matter
it
goes
through.
13
goes
to
26,
oh
I'm,
sorry,
you
know
what
this
is.
This
arrow
is
superfluous.
Sorry
I'm,
confused
with
this
last
green
arrow,
is
copy
and
paste
error
error.
So
the
packet
arrives.
20
it's
addressed
to
26
26
is
happy.
It
removes
the
IP.
Ip
header
gets
rid
of
those
two
extra
rpi
in
ours.
Three,
it
sees
the
IP
packet,
there's
an
rpi
in
there,
that's
cool
it
steps
over
it.
C
C
Think
that's
a
duplicate
slide!
Oh!
Oh!
Oh,
oh,
oh!
This
is
the
other
way
that
we
could
do
is
if
we
add
it.
Instead,
if
we
added
an
IP
IP
header
here
with
the
IP
there,
we
at
course
this
IP
IP
would
be
a
dad,
would
be
added
so
that
it
can
be
removed
by
the
route
the
route
would
remove.
C
C
C
C
This
is
a
very
useful.
Whatever
half-hour
me,
we
had
I
think
because
I
think
that
was
really
useful
to
say
to
say
this
now,
one
of
the
three
of
those
things
that
we
said
was
getting
rid
of
story
mode
solves
a
lot
of
problems.
Okay,
hot
by
story
mode
has
to
use
hop
by
hop
IP
IP.
That
needs
to
be
written
down.
Okay
and
possibly
we
need
attributes
on
routes
so
that
when
you
send
things
up,
do
you
need
them?
Well,
you
don't
need
that
if
you
do
hop-by-hop
story
mode,
IP
IP
rather
go
ahead.
A
C
Sure
no
one's
saying
that
you
have
to
turn
your
equipment
off.
It
just
means
that
we
look
at
the
spec
may
say
in
the
future.
You
don't
need
it
because
you'll
have
dau
production,
which
you
would
but
I
don't
know
I'm.
Is
it
always
better
I,
don't
know
if
it's
always
better
I
think
we
have
a
yes
there's
some
other
presentations
at
the
end
here.
Isn't
there
I
just
don't
want
to
run
at
a
time
here.
So
then
we're
past
gallon
yeah,
yeah,
yeahs.
B
C
Anything
else
to
add
here,
because
I
just
want
to
make
sure
we
capture
what
we're
having
here
and
we
know
what
we
need
to
cook.
If
we
need
to
compress
new
things,
we
know
what
we
need
to
do.
Okay,
that's
the
critical
thing
of
this
effort
was
to
say
what
are
the
patterns
we
absolutely
see.
All
the
time
and
need
to
come,
therefore
need
to
compress
are
the
things
that
that
we
missed
so.
A
Let's
get
again,
there
was
the
question
at
six
low,
whether
the
sixth
floor
erage
work
would
support,
the
non
ripple
leaves,
and
the
answer
is
no,
because
we
don't
know
what
to
do
yet
right.
That's
this
work,
which
would
define
much
right
on
so
it
cannot
be
granted
that
the
6ths
it's
there
who'd
support
work.
This
group
will
find
as
the
solution
for
this
problem,
if
well.
A
C
D
So
it
is
the
designer
of
the
encapsulation,
of
course,
what
I
really
would
like
to
have
is
a
corpus
of
packets,
where
I
can
run
code
on
that
tells
me
what
the
compression
performance
is
for
this
very
end
and
what
the
compression
from
it
and
performances
for
that
variant
and
so
on
yeah.
So
if
we
could
come
up
with
a
list
of
packets
and
and
probability
for
that
for
them,
we
could
actually
at
the
end
say
this.
This
approach
is
better
than
that
approach.
Is
that.
C
D
C
B
C
D
C
D
C
D
C
A
Let's
get
the
six
letter,
H
is
actually
compressing
the
outer
IP
very
much
because
we
know
that
one
of
the
hand
points
is
is
usually
the
root.
So
with
one
bit
you
can
signal
which
one
point
is
the
root,
and
we
know
that
the
other
side
is
usually
either
the
last
router,
which
would
d
capsulate
or
the
final
destination,
which
is
the
one
inside
the
iph,
see
just
like
robot
set
right.
A
But
did
the
D
capsule
later
at
the
end
of
the
tunnel,
is
today
in
the
six
LaRoche
is
never
indicated
because
well
today,
it's
expected
to
be
the
destination
ending
our
backyard.
But
if
we
define
that
it's
like
the
last
row,
router
before
the
lobe
and
leaf
or
node,
then
we
don't
even
know
that
guy
is
so
fulfilling.
We
don't
see.
No,
it
says.
A
A
Yes,
it
was
with
young
ingress
of
this
tunnel,
but
your
belly
most
of
the
time
saying
that
you
don't
know
would
they
capture
later
so
it's
either
the
final
destination
than
you
know,
always
the
last
router
before
it
and,
like
you
said
right,
you
don't
know
that
guy
is
just
the
router
which
will
be
the
last
hop
before
so
it's
good
and
collided
because
the
guy,
no,
we
knows
who
he
is,
but
you
don't
so
so
you
have
to
say
it's
like
many
guess.
The
drives
of
the
last
router
or
eyes.
C
B
C
C
A
C
Okay,
so
so
the
only
problem
I
have
with
that
process.
Is
that
imagine
a
scenario
where
we
aren't
running
over
a
6
lo
pan
and
we
don't
actually
compress,
but
then
we
don't
know
how
to
do
it.
Okay,
we
may
in
some
cases,
not
know
how
to
do
it,
so
so
it
would
be
better
if
it
was
well
defined
before
the
compression,
but.
A
There's
also
the
Internet
of
saying
what
what
you
just
said
before
I
found
this
to
do
well,
which
is
which
is
implicitly
the
destination
of
the
outer
other.
It's
the
next
hub,
which
you
receive
the
packet
virtually
digitally
it
actually
originally
to
its
next
up
towards
destination,
which
is
the
inner
pocket.
If
you,
if
you
express
it
this
way.
Implicitly,
this
nation
is
always
the
guy
you
receive
that
next
up
yeah,
but
that
works
too
I
mean
that's.
That's
just
the
format
in
the
way
or
India
would
really
be
what
we
describe
to
do.
A
A
C
B
B
E
Meeting
is
this
morning
and
and
its
new
draft
on
an
very
old
work.
Actually
we
did
that
I
think
two
or
four
years
ago,
and
we
just
we
rewrote
it
and
we're
going
to
produce
experimental
data
because
at
the
time,
a
few
years
ago,
the
idea
was
not
accepted
in
the
group,
because
we
didn't
have
back
it
with
power
consumption
data
that
we're
going
to
produce
in
the
next
few
months.
E
So
basically,
what
we
want
to
do
is
make
the
GIS
message
is
more
selective,
and
by
that
we
add
a
flag,
we
propose
to
add
a
flag
in
2d
is
so
that
we
can
add
a
metric
container
into
the
GIS,
which
will
make
only
the
routers
that
man,
the
constraint
in
the
DMV
to
be
allowed
to
respond
to
the
das.
This
way
we
can
control
the
population
of
responders,
saying
only
results
that
are
of
interest
to
me.
I
allowed
to
respond,
so
we
have
less
responses
from
an
unknown
population
of
routers
around.
E
Typically,
it's
for
a
note
that
wants
to
join
the
network.
That
know
anything
about
the
network,
and
so
this
way
you
can
say
only
the
routers
that
are
of
interest
to
me
a
lot
to
respondent.
If
I
don't
find
any,
then
we
lower
the
expectations
and
eventually
you
find
somebody
somewhere.
Does
it
not
meet
my
request?
That's
one
thing.
The
other
thing
is.
We
also
want
the
routers
to
respond
using
a
unicast
and
multicast,
because
I'm
the
only
note
joining
the
network,
so
I
don't
need
all.
B
E
We
propose
to
add
an
option
so
that
the
rotters
will
do
some
time
deterring
time
spreading
for
the
responses
that
we
don't
generate,
collisions
on
a
systematic
basis,
so
free
options,
I
mean
free
changes
into
the
D,
is
one
flag
and
allowing
allowing
the
DNT
option,
which
previously
was
only
allowed
in
the
iOS,
allow
it
in
D
is
and
as
this
new
option
and
if
the
flag
is
off
and
the
options
are
not
present
in
in
the
eyes
everything
works
like
before.
So
we
don't
have
to
change
anything.
The
notes
I
didn't
know
about
that.
C
C
E
C
C
C
Ok,
so
so
it's
even
newer
than
I
thought
yeah.
So
it's
very
new,
no
one's
read
it,
but
it
was
good
now,
just
we
had
three
minutes
just
stuff
it
in.
So
please,
please
bring
some
discussion
up
on
the
mailing
list
about
you,
know
this
and
how
it
could
be
used
and
and
I'll
good
to
figure
out
how
it
fits
into
our
schedule
and
things
like
this
and
whether
we
can
weather
before
we
even
consider
whether
we
could
adopt
it.
We
need
to
figure
out
whether
we're
even
charger
to
exist
much
longer.