►
From YouTube: IETF114 ROLL 20220726 1730
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
It's
a
it's
a
kind
of
we
all
know
it's
a
failed
opportunity
on
our
part,
but
you
know.
A
D
Get
yeah.
Thank
you
welcome
everyone.
This
is
the
ata
514
pronunciation,
routine
overlapping
networks.
So
I
please
be
aware
that
this
meeting
is
under
not
well
and
we
have.
We
will
not
read
it,
but
you
have.
If
you
participate,
you
have
to
already
with
atf
process
and
policies.
D
Okay,
the
meeting
tips,
so
we
request
you
to
sign
mytico,
so
we
can
have
the
blue
sheets
and
to
know.
What's
in
the
meeting
use
mitigo
to
join
the
meet
you
that's
as
well
required
keep
audio
and
video
off
if
yeah,
it's
not
used
in
the
onset
version
for
more
participants,
well
mute
unless
you
are
speaking
and
the
use
of
headset
is
recommended.
D
Okay,
our
agenda.
We
are
going
through
the
working
of
the
status
right
now
then
we
have
pascal
depicting
the
status
of
the
projections
and
then
the
six,
no
multicast
registrations,
and
we
will
have
two
minutes
for
open
floor
for
everyone.
D
D
All
the
variable
is
back
to
the
working
group
and
that's
where
we
are
addressing
the
open
issues
to
be
able
to
send
it
back
to
this
team,
though,
projections
are
going
to
be
discussed
today,
as
mentioned
previously,
and
noticed
that
this
document
is
in
working
group.
Last
call
for
several
weeks
already
and
we
haven't
received
any
objections.
So
today
we
want
to
know
their
opinion
and
if
you
agree
with
the
document,
we
can
just
close
the
investment
and
then
the
enderman
priority.
D
Now
those
are
working
in
addressing
open
issues.
Mopeds
and
capabilities
are
waiting
for
attention.
They
were
expired
in
november
last
year,
so
we
need
to
return
to
work
on
those
as
well.
We
want
to
form
ribbon
steering
root
act.
We
would
like
to
again
issue
the
working
group
adoption
for
this
document
for
nfp.
D
D
F
D
D
D
So
we
would
like
you
to
participate
and
yes,
there
is
a
agenda
basically
present
multicast
in
barbell
house
multicast.
In
the
end
we
manage
as
well.
So
this
is
as
well
kind
of
a
brainstorming
how
to
address
multicast,
but
I
think
it
will
be
interesting
and.
F
D
B
B
So
this
is
basically
what
we
do.
We
build
instances
within
instances
and
you
can
you
can
call
them
overlays
and
the
overlays
take
the
usual
shape
of
a
traditional
ripple
instance,
meaning
that
it's
a
geodag
and
and
it
reaches
some
targets
which
are
considered
as
external
destination
in
rfc
9008
and
basically
to
reach
them
from
an
ingress
point
which
acts
as
root.
You
need
to
build
a
deodag
and
we
use
typically
non-storing
mode
to
to
have
the
dotted
lines
and
storing
mode
to
join
the
dots.
Basically.
B
So
this
is
the
things
that
we
are
building.
We
call
track
a
complex,
dotted
line,
basically
a
geodag,
a
complex
expression
of
a
of
a
potential
for
a
path
between
an
ingress
and
an
egress,
and
we
call
a
segment
a
serial
direct
sequence
of
nodes
which
we
can
install
using
storing
mode,
for
instance,
in
black
on
the
top,
you
see
a
story
mode
pedal
which
installs
a
segment
between
a
relay.
B
A
and
a
e,
so
so
basically,
you
can
also
see
that
as
a
way
to
use
ripple
to
do
traffic
engineering
and
really
to
install
what
row
will
be
using
to
make
wireless
a
lot
more
reliable.
B
So
summary
of
the
recent
changes,
so
we
had
a
number
of
reviews
we
had
lee.
It
took
two
round
three
to
to
go
through
the
the
edit.
The
second
round
included
michael's
editions
about
amman's
and
extends,
and
since
then
I've
tried-
and
thanks
for
dominique
and
dinners
to
remind
me,
but
I've
tried
to
to
use
ameld
when
it
was
needed.
B
25
included
a
rewarding
of
terminology
to
make
sure
the
terminology
was
more
clear
on
the
concept
of
legs,
which
is
one
possible
path
along
the
track
source
right
path
along
the
track
and
and
possibly
a
sequence
of
segments
to
get
there.
Then
26
was
remove
series.
So
we
we,
the
chairs,
are
screamers
to
to
do
a
deep
review
as
well.
So
it
was.
B
It
was
quite
instant
intense,
but
I
think
there
was
one
bug
that
he
found
and
for
the
most
piece
it
was
a
lot
of
cleaning
up
and
so
his
review.
The
second
step
of
his
review
went
in
27
with
dominic's
review.
B
B
So
that
was
neat
and
then
dominic
had
these
comments
in
line
in
the
in
the
pro
request.
So
so
I
handled
them
in
line
that
we
actually
use
get
to
discuss,
but
I
I
copy
the
main
results
on
the
mailing
list
as
well.
So
all
the
gory
details
can
be
seen
on
github
and
the
summary
of
the
changes
in
the
mailing
list.
B
So
there
were
a
number
of
interesting
issues
that
were
raised
directly
or
indirectly
by
dominic's
review.
First
thing
is,
and
I
think
it's
quite
important-
the
the
sibling
information
option
basically
gives
you.
B
So
a
node
it
sends
a
daw
with
an
sio
in
it
and
basically,
what
it
says
is
I've
got
all
those
siblings.
So
these
are
the
nodes
which
are
one
hop
away
from
me,
which
I
can
directly
route
to
so
basically
that
means,
for
instance,
if
the
root
wants
to
get
one
of
those
notes
along
the
path
I
can
forward
to
him
he's
mine.
It
can
be
a
next
stop
or
if
he's
a
target,
then
you
can.
B
B
I
guess
in
many
they've
got
variations
around
that,
I'm
pretty
sure
you
can
basically
say
hey
if
I'm
one
hop
away
from
the
destination.
Well,
two
hops!
I
have
one
neighbor,
which
is
a
neighbor.
Then
we
don't
need
rating
multicast
now
in
ripple
was
initially
without
siblings.
Just
for
direct
connectivity,
I'm
sure
I
broadcast
the
multicast
everybody
around
me
knows.
I'm
here
no
need
to
use
a
ripple
path
to
get
to
me.
We
are
neighbors
with
the
sio.
We
extend
this
to
two
halves.
B
It's
one
of
the
consequences
of
having
an
sio
is
we
can.
We
can
do
two
halves
away
and
it's
loopless
it's
just
having
a
common
neighbor
that
can
rely,
so
it
was
there,
but
it
was
not
clearly
stated
and
doing
the
changes
for
dominic.
I
realized
that
it
was
important
to
make
it
more
clear
and
in
particular
what
was
missing
is
we
were
not
amending
rc6550
on
the
multicast
now.
The
multicast,
though,
was
explicitly
for
one
half
and
now
because
we
have
cio,
we
can
make
it
to
hops.
B
So
yeah,
and
by
the
way
there
is
one
api
which
was
related,
so
it's
not
fully
that
it's
never
fully
that,
but
just
in
case
I
asked
the
lawyers
to
to
add
that
ipr
we
already
had
cisco
apr
on
this
draft.
So
this
is
the
second
cisco
apr
on
the
draft,
which
will
I
mean
the
guy
promised
me
that
he
will
push
it
just
now
so
before
before
we
do
the
shepherd
call,
I
will
make
sure
that
the
second
apr,
which
really
relates
well
at
some
point
it
uses
those
two
hop
things.
B
Oh,
it's
me.
I
forgot
hey
yeah,
so
another
thing
which
came
up
as
part
of
the
discussion
with
dominic
and
that
one
also
is
important
forever.
We
had
that
discussion
about
loop
avoidance,
I
mean
we
have.
We
are
in
a
situation
that
we
used
to
call
multi-topology
writing
where
you
have
the
main
topology,
which
is
the
main
report
diode
x
or
the
underlay.
B
If
you
like,
and
you've
got
the
track,
which
is
which
is
an
overlay
and
when
you
do
multi-topology
routing,
if
you
let
a
packet
bounce
between
topologies,
it
can
look
just
depends
on
the
way
the
topologies
are
organized,
so
it's
very
important
that
the
topologies
you're,
manipulating
or
basically
the
rules
that
you're
using
for
writing
have
a
strict
order.
So
you
can
go
from
this
to
that
and
this
to
that
and
this
to
that,
but
never
the
other
way
around.
B
So
we
we
are
the
the
text
that
we
had
was
clear
that
in
the
cons
context
of
forwarding
along
the
loose
track
so
to
the
loose
hops
that
you
could
use
either
the
next
up,
which
is
a
segment
to
get
to
the
next
loose
hub,
you
could
use
a
direct
neighbor
or
via
a
command
neighbor
using
the
you
know,
that's
when
we
found
that
the
sio
discussion
was
missing,
but
basically,
if,
if
you
have
a
packet,
you're
you're
a
logger
truck
you're
one
of
those
loose
hops,
if
the
destination
is
direct,
one
hop
then
pass
it
to
him.
B
If
it's
not,
but
you
have
a
common
neighbor,
then
pass
it
to
him.
If
not,
but
you
have
a
segment
route.
Basically,
a
story
mode
dao
which
reaches
the
next
loose
hop
use
that
else
it
could
be
encapsulating
inside
another
track
which
effectively
gets
to
the
next
loose
help.
Do
that
which
creates
a
new
encapsulation,
but
that's
still,
okay
else
drop.
So
you,
you
add
all
those
methods,
but
what
you
didn't
have
is
a
strict
order
of
them,
so
we
we
clarified
that
to
avoid
loops
basically
do
there
is.
B
There
is
not
only
a
preference
to
use
direct
top
if
you
have
else
common,
neighbor,
etc,
but
it's
also
a
precedence,
meaning
that
you,
you
can
never
take
a
packet
off
a
segment
and
put
it
in
a
different
track.
B
B
There
are
two
things,
though:
one
of
them
is
a
very
old,
a
pull
request
which
really
does
not
apply
to
the
text
at
all
anymore,
but
maybe
there
is
something
to
dig
into
it.
They
don't
know
if
halle
is
with
us,
but
howell,
basically,
is
the
one
who
made
that
poll
request
and
I
would
have
liked
raul
to
shine
in
and
and
look
at.
You
know
the
new,
the
new
way
we
do
a
group
of
items,
basically,
if
it
matches
his
intention
in
this
poor
request.
B
The
second
thing
is
this:
preference
of
you
know
using
a
direct
harp
and
then
a
common
neighbor
since
segment
is,
is
on
the
right.
You
know
it's
less
preferred
forwarding
a
longer
segment
can
use
a
direct
hub,
but
it
could
also
use
a
common
neighbor.
So
today
the
way
the
segment
is
is
written
is
you
have
to
be
in
direct
site
to
the
next
step
in
the
segment?
It
is
a
straight
thing,
but
we
could
have
made
it
one
hop
loose.
Basically,
the
segment
could
be.
B
We
want
to
to
to
take
a
big
risk
and
change
that
inside
the
segment,
but
when
the
segment
you
you
you're
done
with
the
segment
that
you
have
to
deliver
at
least
the
destination
could
be
via
common
neighbor,
but
anyway,
the
way
it's
written
right
now
is
when
you
forward
the
longer
pedal,
which
is
a
story
mode,
pda
or
segment.
B
It
has
to
be
a
hub
by
hop
strictly
when
you
forward
along
the
track.
It
can,
and
you
have
to
deliver
a
packet
that
you
know
just
you
know,
lose
heart
and
you
look
at
how
to
forward
it.
Then
you
can
use
the
one
hop
two
halves
other
than
the
segment.
B
So
there
is
a
slight
inconsistency.
If
you
look
at
the
strict
order,
two
common
neighbor
is
acceptable
is
has
more
precedence
than
segment,
so
common
neighbor
will
not
loop,
even
if
you
forward
inside
the
segment
at
this
time.
B
It
might
not
be
very
clear
the
way
I
said,
but
basically
the
question
is:
should
we
allow
to
forward
the
outcome
and
neighbor
as
we
write
along
the
segment
and
mostly
when
we
deliver
the
aggressor
segment?
It's
not
fully
consistent.
The
fact
that
we
we
allow
it
for
tracks,
but
not
for
segments
just
because
I
didn't
change
the
text
and
did
not
want
to.
I
wanted
discussion
on
this,
so
I
hope
we
come
back
to
that
and
that's
pretty
much
it
by
the
way.
B
There
is
a
28
on
the
works,
because
I
got
the
new
feedback
from
dominique
after
27
and
we
found
that
something
else
of
interest
was
was
missing,
so
it
is,
let
me
find
maybe
I
can
share
the
divs.
B
Yes,
so
the
the
bottom
line
is,
we
didn't
clarify
this
document,
something
which
is
very
clear
in
the
main
ripple
spec,
which
is
that
each
ripple
instance
it's
on
rib.
It's
it's
a
vrf.
If
you
like
it's
it's
it's
it's
on
rip.
So
when
a
ripple
node
runs
two
instances
already
with
650
when
you
forward
along
a
a
a
a
an
instance,
you
really
use
one
verb
and
when,
if
you
can
never
go
from
one
instance
to
the
other,
it
was
already
there.
It
was
important.
B
They
are
shipping
the
night
and
that's
why
we
tagged
the
packet
with
the
hub
by
hub
option
the
ripple
up
the
api.
Basically,
the
rpi
tells
us
what
instance
that
is
that
allows
us
to
look
at
which
rib
we
forward
against,
and
we
can
only
use
this
rep.
That's
why
again
we
avoid
loops
in
this
multi
topology
routing
thing.
It
was
already
there
with
ripple,
but
it
was
not
really
said
again
in
this
document
that
yeah,
even
if
you
do
an
overlay
inside
the
main
dodge,
it's
still
another
dag.
B
So
it's
still
another
instance
and
still
it's
a
different
trap
and
you
have
to
forward
along
that
trip.
You
know
unless,
unless
you
have
a
common
neighbor,
unless
you
have
a
one
hub,
so
I
I
added
some
text
to
just
clarify
that
it
doesn't
change
anything.
It
just
clarifies
that
hey
each
instance
is
a
rip
for
the
for
the
router,
multiple
instances,
multiple
reps
and
then
forward
along.
C
C
B
B
It's
it's!
Basically
the
fact
that
if
you
have
a
common
neighbor,
then
it
will
relay
without
the
loop
and
without
a
writing
protocol.
I
mean
that's,
that's
in
money,
that's
a
usual
thing!
Just
that
triple
was
not
using
it
because
we
didn't
have
the
sibling,
but
now
we
have
it
so
we
need
to
average
it
in
full.
B
You
expect
no
sights,
no
stars,
we
cannot
do
because.
So
if
you
go
that's
when
we
refer
to
the
raw
architecture,
so
that's
that's
another
interesting
discussion
and
alvaro
is
in
the
room.
So
I
will
like
just
to
say:
hey.
I
will
run
for
you
here.
Basically
one
one
consumer
for
what
we
are
doing
here
is
raw
wood
needs.
B
Those
tracks
and
row
has
defined
them,
so
I
I
refrained
to
do
too
much
definition.
This
document
wanted
to
keep
the
two
completely
consistent
and
so
raw
as
a
concept
of
track,
which
is
a
little
bit
wider
than
what
ripple
can
build,
because
in
repo
we
have
a
dodec,
meaning
that
all
the
links
are
oriented.
B
So
you
go
east-west
from
source
to
destination
with
a
row
there
is
also
north
south
north
south
is
bi-directional:
you
can
use
it
up
down
or
down
up
just
don't
want
to
loop,
so
you
must
have
something
which
tells
you
how
not
to
look,
for
instance,
a
source
route,
but
the
link
could
be
used
in
both
directions
and
the
reason
for
that
is
in
a
row.
Sometimes
you
want
to
do
checkpoints
to
make
sure
that
the
packet,
for
instance,
if
you
have
a
north
path
and
a
south
path,
having
a
north-south
link
to
connect.
B
The
two
allows
you
to
check.
If
both
hands
have
received
the
packet
and
if
they
don't,
then
they
share
it.
So
now
both
hands
have
the
packet
and
you
can
fall
so
you
can
basically
replicate
over
the
north
south
in
either
direction
to
make
sure
that
the
packet
is
is
there
and
then
you
can
continue
to
to
the
west.
B
This
capability
is
not
present
with
ripple,
because
ripple
does
not
signal
sibling
rods,
basically,
rats
which
would
be
between
notes
of
the
same
cost.
B
It
was
proposed
very
early
in
the
design
of
ripple
because
of
one
big
drawback
of
geodex,
which
is
basically
that
the
node
which
is
closest
to
the
root,
has
no
plan
b,
because
every
link
is
oriented
at
least
the
node,
which
has
half
of
the
load
under
his
heart
on
his
arms.
You
know,
because
it's
close
to
the
root,
it's
just
one
half
away
from
the
root.
Well,
because
he's
the
closest
he
cannot.
B
So
that's
why
a
row
has
a
little
bit
more
than
what
ripple
can
build.
We
can
always
discuss
in
the
future
how
we
do
nerf
cells,
but
the
the
use
case
is
basically
the
scene
as
if
you
want
to
replicate
a
packet
on
two
path,
and
you
want
to
verify
at
some
point
in
the
middle
that
both
sides,
the
packet
is
still
there
was
not
lost.
Then
you
can
reconcile.
You
can
think
about
this
north
south.
B
And
basically,
the
first
one
which
has
it
passed
it
into
the
other
universe.
Is
that
oh,
I
got
it
from
him.
It
won't
pass
him,
but
if
I
didn't
get
it
from
him,
I
will
pass
him
and
you
see
this
kind
of
rule
and
so
you're
sure
that
both
sides
of
the
segment
now
they
have
the
copy
and
it
can
progress
against
to
the
west.
D
Okay,
thank
you,
and
can
you
mainly,
how
can
we
differentiate
from
a
track
from
a
complex
drug?
Basically,
there
is
a
metric
or
basically,
if
they
have
a
lot
of.
B
Segments
the
term
complete
track
was
there
to
just
explain
that:
it's
not
serial.
You
know
if
you
express
something
f,
a
dot
b
that
dot
c,
and
that
would
be
your
simple
track.
Basically
serial
track,
and
now
you
would
have
segments
between
a
and
b
and
b
and
c,
which
would
be
the
the
one
hop
storing
mode
thing
now.
If
with
this,
if
you
lose
the
packet,
there
is
no
way
to
to
get
it
to
destination.
B
If
you
have
more
than
one
possible
path,
then
you
have
a
complex
track.
The
track
is,
is
a
potential.
You
can
send
your
packet
along
this
diode,
you
know
and
when
it
forks,
you
can
choose
to
send
it
up
through
the
up
path
or
the
down
path.
I
mean
each
time
it
works
or
you
can
decide
to
replicate
or
anything,
that's
raw,
so
how
you
use
it
is
really
wrong,
but
basically
it's
no
opportunity
see
the
track
as
an
opportunity.
B
If
a
packet
is
not
replicated,
then
when
you
forward,
then
you
have
more
than
one
next
step.
You
choose
one,
that's
what
you
typically
do
with
repo
you
choose
one.
Typically,
for
instance,
you
use
your
preferred
parent
to
go
to
the
route,
but
you
have
an
alternate
parent.
You
could
have
passed
the
packet
to
him.
It
would
have
gotten
there.
B
So
so
the
dag
is
a
potential.
A
certain
packet
will
experience
one
of
those
potentials.
It's
gonna
be
the
path.
So
the
path
is
the
actual
experience
of
one
packet.
A
different
packet
may
experience
a
different
path
between
the
same
a
and
the
same
b.
If
you
have
a
diode
between
the
two
and
and
you
load
balance
or
do
whatever
so
so,
the
path
is
different
from
the
track.
The
track,
the
potential
it's
the
global
set
of
everything
that's
possible
and
the
path
is
the
experience
of
one
given
packet.
C
B
For
instance,
what
I
just
said
today,
you
know
reply
to
you
on
on
the
email
is
something
that
has
always
been
there.
Each
instance
is
a
red,
but
it
was
not
written
so
so
I
think
it's
all
going
the
right
direction,
the
only
thing
which-
and
it's
nothing
new,
the
the
thing
that
which
the
most
surprising
thing
might
be
this
this
sio
thing,
which
maybe
people
did
not
realize
it-
was
there.
So
so
again,
it's
clarification,
so
I
just
voiced
it
on
the
mailing
list.
B
So
I
would
be
surprised
know
that
somebody
shines
in,
but
if
somebody
shines
in
you
know
today
or
on
the
mailing
list
happy
to
address
that,
but
no
it's
I
mean
you've
seen
the
fixes
that
you've
proposed
right.
You
you
made
the
document
a
lot
more
readable
and
the
guys
before
you
I
mean
tourists.
For
instance,
I
spent
a
lot
of
time
asking
me
to
to
reorganize
to
help
the
first
reader.
Then
we
had
all
the
reviews
that
I
showed
I
mean
we
had
a
good
number
of
reviews.
B
If
you
look
at
my
slides
for
the
previous
ietf
and
the
one
before
I
always,
you
know
retrofit
the
last
four
or
five,
the
items
you've
seen
it's
a
lot
of
of
clarification,
a
lot
of
typos
as
well.
Sorry
for
that,
but
it's
all
going
in
one
direction.
That's
basically
my
answer
to
you.
We
have
not
fought.
We
have
not.
We
have
refined
clarified,
but
I
I
don't
think
we
it's
all
stabilization.
It's
not
forking
to
something
completely
new
at
the
last
minute.
B
And
I'm
sorry
because
I
was
I
was
telling
alvaro.
I
have
a
point
for
you.
The
point
for
you
alvaro
was
that
I'm
referring
the
raw
architecture
for
the
terminology
of
trucks
and
because
it's
terminology
I
made
it
the
normative
reference
problem
is
the
raw
architecture
is
still
in
the
works.
I
mean
it's
in
lascaux
type
of
situation,
but
it's
still
on
the
works
and
so
and
and
by
the
way,
I'm
pretty
sure
it's
going
to
end
up
as
an
informational.
B
So
we
will
be
in
this
double
situation
of
of
a
done
ref
to
informational,
and
basically
this
document
will
be
held
by
the
raw
architecture.
The
alternate
was
to
redefine
the
terminology
from
the
raw
architecture
into
this
document.
B
F
Hey
pascal
thunder
rolling
id,
so
is
this.
F
So
I
I
think,
referring
to
another
document
for
terminology
versus
redefining,
is
the
right
thing
to
do
right.
We
don't
want
to
define
the
same
thing
in
many
places
and
then
have
the
definitions,
maybe
not
be
in
sync
or
you
know,
etc.
So,
that's
a
good
thing.
F
Don't
worry
about
the
reference
and,
if
there's
a
downrive
or
not,
first
of
all,
knowing
that
it's
an
architectural
document,
I'm
sure
that
it
will
get
referenced
later
by
some
other
things
as
well,
so
yeah
that
shouldn't
be
a
problem
and
I'm
going
to
say,
don't
worry
about
the
sequencing
either
thinking
that
the
raw
architecture
is
a
little
bit
behind,
because
in
any
case
my
publication
queue
is
relatively
deep.
So,
by
the
time
I
get
to
this,
you
know
we
won't
have
to
worry
about
it
anymore.
Yes,.
B
Okay,
that's
good,
and
if,
if
for
some
reason
we
completely
abort
row,
then
I
change
that.
Obviously,
but
I
I
mean
the
architecture
is
on
the
right
track,
so
I'm
not
just
to
let
you
know
that
it
will
it
will.
It
may
hold
this
one
that
depends
on
the
different
cues.
I
mean
john's
versus
servers,
okay
and
that's
pretty
much
it
for
me.
C
C
B
B
C
B
Oh
ho
ho
ho,
okay,
so
well
I
mean
I'm
lost.
Why
do
you
pass
me.
C
C
B
A
B
B
Yes,
there
we
go
so
this
is
a
work
which
started
at
six
low,
but
really
the
way
it's
written
right
now.
It
impacts
six
low,
six
man
and
raw,
and
the
reason
for
that
is.
We
wanted
to
to
give
a
complete
story
across
work
groups
and
that's
kind
of
based
upon
some
other
histories,
for
instance,
the
sixth
opponent
versus
report
stories,
which
has
been
bouncing
back
back
and
forth
between
six
low
and
raw.
B
So
next
next
slides
me.
So
that's
why
we
ended
up
with
this
single
document.
The
document
showed
up
at
six
low
because
it
starts
with
what
you
register.
In
this
case,
we
are
talking
about
extending
eight
five,
four
five
for
any
cast
and
multicast
registration,
and
why
are
we
doing
this?
We
are
doing
this
because
we
got
some
users
which
came
to
us
and
came
to
the
atf
effectively
and
said:
hey.
B
We
we
we're
good
for
nd,
we
live
without
multicast
and
that's
really
great,
but
we
still
need
multicast
and
for
multicast
we
have
the
same
prime
as
we
had
with
nd
right.
So
the
the
report
creates
a
pool,
a
multicast
pool
by
the
router,
which
is
exactly
the
model
that
you
don't
want
in
an
iot
network,
because
it
means
that
the
devices
has
to
have
to
be
awake
waiting
for
this
pool
and
reply.
B
B
Now
there
will
be
no
mld
at
all,
even
for
multicast.
You
can
register
to
your
router
and
the
router
will
do
what's
necessary.
So
we
we
have
the
same
benefits
as
we
already
have
with
h505.
We
have
a
basically
a
host
to
router
interface,
which
is
completely
abstract
to
how
the
router
routes
or
now
does
multicast.
B
So
that
was
that
was
where
we
started
from,
but
then
we
said
hey
by
the
way.
What
is
the
multicast
for
this?
Oh,
it
could
be
meepo
and
that's
what
weissan
is
doing,
but
most
of
the
time
people
are
using
non-storing
mode
and
guess
what
we
have
multicast
only
in
storing
mode.
B
B
So,
typically,
a
device
that
doesn't
want
to
do
mld
is
a
low
power
device
that
doesn't
want
to
do
ripple
either.
So
it
will
do
it
505,
it
will
be
seen
as
an
external
destination
and
then
it
there
will
be
a
tunnel
to
the
router
for
him
so
for
non-storing
mode.
Multicast,
why
don't
we
leverage
this
tunnel?
B
And
so
we
said
hey?
Basically,
the
multicast
addresses
are
registered
using
non-storing
mode,
but
basically
the
way
it
percolates
will
be
very
close
to
to
the
way
we
do
unicast.
With
only
one
thing,
the
tid
will
be
omitted
that
will
that
won't
be
considered?
That's
gonna
be
next
slide,
because
multiple
sources
can
can
ask
the
same
thing
and
the
the
root
will
do
ingress
replication.
B
So
there
is
no
multicast
tree
inside
the
non-storing
ripple.
There
is
just
a
set
of
tunnels
between
the
root
and
the
routers
which
serve
the
final
destinations,
and
so
we
we
added
this
into
into
this
document
and
we
called
it
mod5.
So
that
was
the
discussion
on
on
odv
repo
with
map
five,
and
we
agreed
that
aodv
repo
is
map
four,
so
we
can
beam
up
five.
B
So
it's
still
question
mark,
but
basically
we
we
are
there.
So
we
have
new
flags
in
the
arrow
and
those
flags
are
propagated
in
the
dow.
So
now
that
that's
basically
consistent
the
changes
we
do
in
the
arrow
and
the
changes
within
in
the
dial,
so
that's
the
the
cool
thing
about
having
a
single
document.
B
B
So
we
clarified
that
it's
not
an
a2mld
where
the
device
pushes
that
he
listens
to
an
address
as
opposed
to
to
getting
a
pull
by
the
router.
We
clarified
that
the
tid
freshness
assertion
is
not
done,
because
there
will
be
multiple
devices,
not
synchronized.
That
will
do
the
same
request
for
the
same
multicast
or
any
cast
address
so
ripple
should
not
decide
that
the
only
good
ones
are
the
freshest
based
on
tid,
because
the
tid
won't
be
synchronized.
B
B
So
what
we
did
is
we
created
a
new
ro
status,
which
is
registration,
refresh
request
that
the
router
can
sound
in
an
asynchronous
n,
a
which
already
exists
in
a
545,
but
this
status
is
new
and
the
asynchronous
na
can
be
unicast
or
multicast.
So,
for
instance,
if
if
the
router
gets
a
packet
from
a
1600
that
it
doesn't
know,
it's
not
a
normal
situation,
because
you
need
to
register
to
your
router
before
you
use
them.
I
mean
for
savvy
or
stuff
like
that,
so
the
writer
should
know
about
you.
B
B
It
can
also
be
an
essence
multicast
in
the
case
of
ysun,
for
instance,
they
have
effectively
a
multicast
channel,
it's
slow,
it's
periodic,
but
they
have
a
time
where
the
devices
are
synchronized
and
will
listen
to
a
multicast.
So
the
router
will
have
an
opportunity
to
send
a
multicast.
B
B
Basically,
the
the
we
used
to
call
things.
Subregistries
and
registries
wasn't
clear
what
was
a
registry
and
sub-registry?
There
is
no
sub-registry,
so
so
all
the
tables
individual
tables
now
are
registries
and
the
the
basically
the
web
page
which
aggregates
the
number
of
registries
now
is,
is
kind
of
a
registry
group,
or
you
can
say
it's
under
the
name
blah
of
the
page.
But
the
page
itself
is
not
a
registry,
it's
a
group
or
just
a
page,
the
registries
or
the
things
inside.
B
So
so
this
this
item
I
just
said,
is
next
slide:
let's
go
to
it,
so
the
second
thing
that
the
router
could
do
when
it
reboots
is
effectively
sent
in
a
second
array
now.
Sadly,
this
does
not
exist
at
six
men
and
that's
the
reason
why
I'm
saying
this
document,
in
fact,
six
men
alvaro
again.
This
one
is
one
that
could
deserve
your
attention.
B
What
we
do
with
this
option-
and
there
are
two
things
we
can
do-
we
can
take
the
text
out
of
this
document
and
publish
without
it
or
we
can
get
the
documents
through
six
man
to
say
to
get
our
approval
and
the
recline
yesterday
was
more
into
let's
make
it
a
separate
document
which
means
that
either
we
we
of
block
this
document
to
make
the
other
one
a
normative
reference
or
we
live
without
it,
and
when
we
publish
the
other
one,
we
say,
oh
by
the
way
we
extend
this
one
with
you
know
as
a
potential
user
of
this
option
and
basically
what
the
option
is
is
just
the
time
that
the
router
has
been
awake.
B
B
So
that's
where
we
were,
and
so
basically
eric
has
a
task
to
to
discuss
it
with
the
other
ids.
You
know,
basically
you
and
eric
wink
for
six
man
and
probably
the
six-man
chairs
as
well-
and
you
know,
come
back
to
me
and
whether
I
should
take
that
particular
option
and
it's
processing
out
of
this
document
make
it
a
separate
document.
B
So
it's
for
me,
it's
either
way
is
okay.
We
can
live
without
the
array
option
because
we
have
dna,
so
we
still
have
a
solution.
It's
good
to
have
the
array,
the
toolbox,
adding
it
later.
The
only
problem
I
have
is
is
the
implementers
of
this.
Rfc
might
not
implement
the
array,
and
when
we
publish
the
array
they
won't
even
care
or
know
that
it's
there
so
so
that's
they
will
have
to
leave.
Probably
with
dna.
B
Now
going
through
six
man
having
its
own
document,
I
guess
we'll
have
a
much
better
result,
because
six
man
will
hopefully
pass
it,
but
they
will
message
it
make
it
more
useful.
Hopefully
so
that's
where
we
are
so
so
don't
be
surprised,
although
if
you,
if
you
see
rick
coming
to
you
and
and
asking
you
know
how
we
proceed
with
this
thing,.
F
Hey
I
love
her
though
yeah
I
didn't
know.
If
you
wanted
me
to
say
something:
yes,
it
makes
sense
to
me:
okay
and
I'm
neither
six
eleven
or
six
man
or
my
working
group.
So
yes,
I'm
fine
with
this
when
you're
finished,
I
do
have
a
couple
of
other
questions
for
you,
but.
F
Okay,
so
what
is
the
status
of
this
in
six
law?.
B
Oh,
we
are
done.
I
mean
the
the
expectation
that
what
was
discussed
at
six
low
was
to
trigger
work
group
last
goal
and
I
guess
the
chairs,
maybe
I
we
should
leave
dominica.
I
guess
discuss
this
right
because
you
already
agreed
with
them.
I
guess
munich,
so
I
will
leave
you.
C
F
Perfect
that
works
perfect.
For
me,
the
reason
I
was
asking
was
was
not
that,
but
was
because
of
the
relationship
with
pim,
and
I
know
that
I
had
asked
that
we
discussed
this
in
pim.
I
know
that
he
sent
an
email
to
the
print
working
group.
I
refreshed
the
email
a
month
or
so
ago
to
see.
If
you
know
there
was
something
and-
and
there
was
nothing.
B
Yeah
and
I'm
actually
that's
that's
thank
you
alvaro
for
bringing
that
up.
Tim
also
mentioned
that
yesterday
and
the
answer
timothy,
winters
and
and
the
the
the
answer
I
made
to
him
was
like
we
are
not
really
doing
multicast.
We
are
doing
grass
replication.
F
Is
that
pim
does
other
things
besides
pim?
The
protocol
like
mld,
which
is
something
that
you're
placing
here
and
the
pim
working
group
right
now,
is
working
on
moving
mld
to
be
an
internet
standard.
F
So
it
would
be
nice
for
them
to
consider
if
there's
other
work
needed
if
the
signaling
doesn't
come
through
mld,
but
the
signaling
comes
through
the
mechanism
you're
specifying
here,
so
I'm
gonna,
so
that
you're
not
surprised
when
people
come
running
behind
you,
I'm
going
to
go
talk
to
the
pip
chairs,
because
I
think
this
is
important
that
the
working
group
discuss
that.
F
No
problem
type
response
from
them,
but
but
I
would
really
like
to
see
a
response,
because
there's
a
lot
of
processes
that
are
dependent
on,
as
you
mentioned
in
the
draft
mld
snooping,
and
you
know
some
other
stuff
that
would
have
to
change.
So
if
a
host
now
starts
using
the
signaling
in
a
network
that
would
otherwise
be
using
mlv,
you
know
someone
has
to
do
something
different,
so
just
so
that
they're
at
least
aware
that
we
get
some
positive
feedback
from
them.
That
that
someone
read
the
draft
and
that
yeah.
B
That's
neat
alvaro.
I
really
appreciate
it
the
the
way
I
I
I
would
like
what
I'm
mostly
interested
in
on
my
side
is
whether
there
is
something
we
would
need
to
add
in
our
spec
here,
because,
basically,
the
the
whole
the
whole
thing
about
545,
the
philosophy
behind
it
is.
We
have
an
agnostic
way
for
the
host
to
talk
to
the
router.
B
That's
why
the
arrow
has
the
sequence
counter,
because
when
you
move
the
routers
cannot
know,
what's
the
most
recent
location,
you
have
to
provide
it
to
them.
So
so
the
sequence
counter
the
proof
of
ownership,
those
things
the
router
cannot
infer,
but
everything
else
like
deciding
to
inject
it
and
how
to
inject
it
cetera
long
as
you
know.
So
if,
if
mld
needs,
because
the
router
could
today
what
we
do
with
this
pack,
is
we
say
hey
if
you,
if
you
have
non-storing,
you
can
use
this
new
mob
5..
B
F
As
as
we've
talked
about
before,
8505,
and
even
the
extension
that
you
mentioned
to
this
about
restarting,
have
broader
applicability,
and
so
it's
not
only
going
to
be
in
networks
where
ripple
is
running,
it
may
be
in
the
network
exactly.
F
So
I'll
it's
too
bad
that
the
pym
meeting
was
this
morning,
but
I'll
have
mike
and
stigg
chase
it
down
at
some
point
and
and
try
and
have
some
discussion
with
the
pin
working
group
that
you
know
they
push
them
to
actually
say
something
well,.
B
If,
if
someone
could
contribute
to
the
worker
plus
call
and
a
supplement
saying,
oh,
you
know
I'm
good,
but
I'm
I
mean.
If,
if
you
know
he
comes
and
say,
please
add
this
flag
to
the
to
the
arrow,
because
we
need
it
or
this
field.
We
have
an
opaque
field
where
we
can
put
one
bite
by
the
way.
So.
B
A
D
B
B
This
is
basically
traditional
ripple
now
it
was
not
said
enough
in
the
ripple
multicast
support
in
6550,
which
is
only
story
mode.
It
was
not
said
enough
that
the
tid
does
not
apply
to
multicast,
because
if
multiple
people
register
to
the
same
address
I
mean
they
don't
know
what
tid
the
other
guys
are
using.
They
all
want
the
multicast.
B
So
it's
not.
The
tid
was
meant
to
keep
only
one
route,
only
the
most
recent
for
unicast
moving
for
any
cast.
You
want
to
keep
them
all.
So
basically,
the
only
way
to
retract
the
multicast
route
is
to
time
it
out
to
let
the
lifetime
time
out
and
the
tid
cannot
be
used
to
say.
Oh,
this
one
is
absolute
because
it
has
an
order
tid
now
for
the
rover,
the
rover
is
normally
a
proof
of
ownership
and
then
again
we
will
need
to
do
something
like
8
9
28.
B
B
Now
in
the
future,
we
may
have
a
spec
which
synchronizes
the
rover,
values
and
synchronizes
the
tids
between
the
different
requesters,
in
which
case
the
rover
would
would
have
effectively
a
significance.
But
for
now
I
guess
I
need,
as
I
take
as
a
workplace,
call
command
that
I
have
to
say
what
I
say
to
about
the
tid.
I
would
say
about
the
rover,
since
they
are
not
expected
to
be
synchronized,
there
can
not
be
one
value
of
the
rover
for
this
multicast
group.
B
E
B
E
It's
okay,
it's
okay!
From
my
side!
If
I
I
let
me
think,
but
if
I
have
some
question,
I
will
email
on
the
group.
B
D
D
Okay,
we
had
four,
so
it's
good
okay!
Thank
you
very
much.
Okay,
five!
Good!
Okay!
Thank
you
very
much.
I
think
we
can
close
the
meeting.
Thank
you
very
much,
pascal
for
the
presentations,
and
we
continue
in
the
mailing
list.
Have
a
nice
itf
week.
C
C
Since
I'm
in
a
room,
we
have
a
few
new
comments
with
a
new
attendee
ribbon,
so
maybe
we
jumped
quickly
into
detailed
stuff
if
you
are
interested
to
learn
a
bit
more
about
what
we
are
doing,
what
this
is
all
about.
You
can
come
forward
to
me
and
we
can
have
a
five
minute
chat.
So
I'll
give
you
the
the
context
of
what
is
being
done
and
the
history
of
this
working
group.
If
you
are
interested.
Thank
you.