►
From YouTube: IETF95-ROLL-20160405-1620
Description
ROLL meeting session at IETF95
2016/04/05 1620
A
A
B
A
A
C
A
Ok
and
we're
going
to
see
the
use
of
real
information
graph
and
after
Pascal's
want
to
give
us
the
rotten
spots
financial,
a
remotely
we
are
going
to
have
Raul
and
after
period
hunt
display
the
data
model
proposal
and
we
are
going
to
have
some
discussion
about
the
role
it
ins
for
reach,
after
if
we
are
going
to
reach
out
there
and
after
fina.
Finally,
we
are
going
to
have
open
floor
discussion.
A
A
Okay,
about
this
day
for
the
interactive
drive,
we
have
some
newer
pcs
for
him
and
okay,
a
day
Amy
a
draft.
We
got
a
security
review
with
just
some
neat
son.
He
says,
isn't
Hg
and
thank
you
nancy
for
the
work
of
that
and
okay.
Next,
okay,
we
have
some
related
internet
draft
that
are
not
part
of
the
charted,
but
we
can.
We
can
include
that
for
in
the
future
turtle.
A
A
A
Okay
for
the
cases
in
a
story
mode?
Basically,
we
have
12
cases,
a
rough
mean
ripple
where
leaf,
no,
no
Raph.
What
no
ripping
a
web
leaf
and
all
this
case
in
story
mode-
should
have
the
API
that
is,
the
fec
65
53
yeah
LPL
option
in
the
hot
back
home,
hidden
hill
and,
like
I,
said
story
mode.
We
don't
need
the
h3,
the
65
54
okay.
For
these
cases
we
will
need
encapsulation,
in
which
cases,
in
the
cases
that
we
have
note
that
doesn't
belong
to
air
pure
domain,
be
following
our
rule.
A
Obviously,
ok,
so
the
encapsulation
for,
for
example,
for
this
case,
the
destination
of
the
encapsulation.
Ok,
we
are
world.
We
were
in
having
some
problems
and
how
to
know
to
pass
the
information
to
any
story
mode
to,
for
example,
form
through
to
a
note
that
doesn't
belong
to
ripple.
The
solution
that
we
propose
is
to
use
hope,
I,
hope
encapsulation.
A
So
in
this
case
the
encapsulation
is
going
to
end
in
the
last
CSL
air
before
going
to
the
no
that
doesn't
belong
to
ripple.
So
this
is
an
air
is
going
to
remove
the
FBI
encapsulate
field
and
pass
information
to
the
no
so
for
it
for,
for
that,
one,
the
dead,
for
we
have
her
behalf
encapsulation
destination.
In
that
case
no
Raph
to
route
the
destination
of
the
cancellation
is
the
route
and,
for
example,
in
this
case
from
the
internet
to
raf.
Ok,
the
internet.
A
They
the
package
arrived
to
the
6
lb
air,
so
this
is
al.
Ba
encapsulated
the
PA
api
in
the
IPV
Sirius
hither
and
pass
information
and
with
the
destination
of
the
graph
and
the
same
for
the
trace
cases
from
run
story
mode
to
go
downward.
We
need
to
specify
the
whole
route,
so
we
use
the
LG
and
h3
feel
sore
throat
in
hitter
and
in
some
cases
we
have
FBI
on
optional.
Yes,
because,
like
the
road
is
specified,
we
are
not
going
to
have
loop,
and
so
we
will
need
to
have
api
and
okay.
D
Get
you
bail,
Serena's,
a
quick
note
is:
if
we
use
the
six
leverage
compression,
then
the
FBI
becomes
useful
again,
because
that's
how
you
find
the
route
of
the
geotag,
which
is
actually
used
as
a
reference
for
the
compression.
So
if
you
may,
you
know
belong
to
multiple
do
tags.
You
need
to
know
which
route
of
which
do
tag
is
actually
used
to
compress
everything.
So
you
find
that
even
if,
in
theory
you
could
live
without
the
FBI
for
one
thing,
LC
6550
doesn't
tell
you
you:
can
you
light
it?
A
A
Thank
you.
We
are
going
to
qualify
the
job
with
it
well
as
well
for
the
destination,
for
example,
in
this
case,
a
from
no,
no
that
does
it
belong
to
rebuild
of
mine
too.
I
know
that
doesn't
belong
to
rebuild
a
mine.
First
encapsulation
go
to
the
road
with
the
destination
after
to
the
CSL
air,
in
this
case,
in
story
mode.
This
is
lb.
Air
has
a
complete
knowledge
of
the
topology,
so
they
put
knows
which
is
the
seas.
A
El
aire
that
have
to
send
a
packet-
and
we
don't
have
this
case
with
story
mode
right,
so
one
next
time
so
well
in
story
mode,
it
has
I
say
the
center
has
to
know,
which
is
the
receiver
reset
receiver.
That
is
not
a
area
where
and
well.
The
solution
that
we
propose
is
say
hope,
I,
hope
in
absolution.
A
Slide:
okay:
we
are
in
the
new
version
as
well,
see
slow,
orig,
compressions
cases
and
well.
The
document
is
almost
to
last
column
is
missing.
Some
information
from
sis
low
air
eh,
but
we
have
some
bites
to
determinate
still,
but
would
be
nice
to
have
your
opinion
on
the
mainland
is
about
this
document
about
the
use
cases.
B
I
like
to
have
Peter
fan
of
stock
I
like
to
invite
more
comments
actually
in
review,
because
it's
very
complex
document
I
think
it's
very
useful
because
actually
it
is
quite
puzzling
way
to
find
all
these
ways
and
to
see
how
he
musta,
capsulate
or
not,
and
must
look
carefully
at
it.
So
I
would
like
to
invite
what
are
two
reviews
of
this
document
and
that's
the
camellia
and
sent
in
on
mailing
list
whom
will
be
interested
ooh.
B
D
Okay,
so
six
voyage
has
been
around
for,
like
your
honor
half
I
mean
the
work
has
been
mostly
done
at
six
low
62
pure
Brantley
is
happy
with
it
now,
but
since
it's
about
compressing
ripple
artifacts
in
data
packets,
SiC
slow,
you
know
suggested
that
role
does
the
final
review
and
goes
to
HD
with
it.
So
that's
what
happening
now,
changes
that
happen
in
this
document
before.
Actually
we
moved
it
to
roll,
so
we
have
improved
the
documentation
and
how
things
work.
D
There's
a
lot
more
text
about
how
addresses
are
coalesced,
which
was
kind
of
encase
it
in
the
document.
Initially.
So
now
it's
much
clearer!
You
do
this
step!
You
do
this,
that
there
are
number
of
stepwise
operations
which
retail,
how
you
do
things
and
we
went
with
that
to
the
HC
plug
test,
but
the
document
was
too
late,
so
there
was
only
one
implementation
so
really
for
this
HC
blood
test,
we
could
not
validate
the
the
draft
that
will
happen
in
Berland
for
the
next
hf.
Well.
D
Another
interesting
outcome
of
the
HC
blood
test
is
this
rupture.
We
found
that
actually,
with
six
lower
age,
60
and
six
up
and
HCM
sorry
with
RFC
6282,
one
cannot
use
the
outer
header
to
compress
the
inner
ater.
So
if
you
have
some
something
like
a
prefix
which
is
present
in
the
header
outside
encapsulated
header,
well,
you
cannot
inherit
that
knowledge
to
compress
the
inside
the
inner
header.
So
considering
the
use
cases
we
had
with
a
lot
of
packets
coming
from
the
root
and
having
the
prefix
of
the
root
compressed
with
six
well
write
without.
D
Oh,
that
would
be
very
useful
to
say
that
the
inner
header
is
also
compressed
with
this,
while
this
inheritance
does
not
exist
in
6282.
So
we
wrote
this
this
very
small
draft
to
say
if
you're
in
the
pure
ripple
world-
and
you
don't
want
to
have
to
create
an
additional
six
dope
and
context
just
to
reflect
the
same
prefix
that
you
already
compressed
with
ripple,
then
why
don't
we
create
this
inheritance?
D
Basically,
what
what
you
get
with
this
is
that
you
can
consider
that
6lowpan
as
its
own
context,
which
is
used
by
6282
and
in
the
header
compression,
is
except
a
new
indicator
context.
Numbered
say
where
you
take
this
text
phone
information.
Well,
you
can
consider
that
now
repo
is
a
secondary
source
of
context,
the
only
context
being
the
prefix
of
the
route,
but
once
you
have
it,
then
you
can
inherit
that
to
compress
the
whole
frame,
inner
and
outer
header-
that's
what
is
this
new
work
brings
it
thanks
to
the
HC
black?
D
So,
like
I
said,
like
we
said
like
we
all
set
the
six
th
initial
draft
work
split
into
a
piece
which
read
deals
with
indexing
content,
which
is
left
at
six
low
and
is
currently
in
let's
go.
The
last
call
review
is
extended
just
like
the
six
leverage
Rosco
reduced
extended,
because
there
are
not
enough
comments.
So
if
your
interest
in
this
space
please
go
to
six
low,
the
draft
is
a
few
pages
just
to
say
now
you
can
index
things,
but
they
need
your
review.
They
need
your
comments
to
pass.
D
D
Okay,
so
so
there
are
a
number
of
issues
which
were
discussed
and
it
dries
by
this
document.
You
will
find
in
particular
that
the
what
was
called
the
air
h3
loh
is
now
called
s.
Rh
rh
there
was
a
discrepancy
between
the
granularity
that
we
had
above
the
rpi
cigarette
I
am
the
one
below
with
your
address
is
very
right,
so
we
thought
to
keep
things
consistent
at
the
same
level.
D
Now
the
we
keep
things
abstract
and
we
don't
say
it
specifically
for
archery,
we
say
it's
for
a
rotten
either,
so
we
call
the
option
SRH
supposed
to
rh,
so
they
must
still
be
in
the
slide
somewhere
to
mention
their
age
36
garage.
But
it's
now
in
the
document.
It's
it's
called
s,
LH,
six
large
and
then
I've
got
a
number
of
examples
of
how
things
are
comprised.
The
first
thing
that
that
just
consider
is
we
wanted
to
make
the
forwarding
plane
easy.
D
But
what
we
did
was
six
or
a
cheese
is
order
things
in
a
way
that
the
code
in
the
intermediate
router
net
can
be
very
fast
and
very
simple.
So
inshallah,
if
you
decompress
things,
you
like
a
normal
ipv6
packet,
but
in
the
compressed
form,
things
are
other
differently
so
that
the
router
operation
can
be
very
quick
and
very
okay.
D
The
original
format,
noncom
price
would
be
IP,
then
hub
a
harp
option
and
the
rest-
and
you
see
that
now
that
the
Reaper
option
is
before
okay,
but
but
the
compressed
form
of
that
is
well,
the
non
compressed
form
of
that
is
really
hot
by
up
after
the
IP
address.
Okay,
so
that
allows
us
two
things.
One
of
the
things
is
just
to
process
the
ripple
thing
without
looking
at
the
rest
and-
and
the
second
thing
that
it
allows
us
to
do
is
keep
the
code
that
was
handling
the
6282
packet
without
a
church.
D
So
you
can
really
add
this
fast
path,
operation
without
changing
your
6282
passer
next
line.
Okay,
just
to
show
that
the
fragmentation
as
defined
for
the
944
is
not
change.
So,
whether
you
six
or
Rachel,
not
six
leverage.
You
can
steal
fragment
things
exactly
the
way
you
did.
The
60
reg,
just
like
the
air
I
phc
are
after
the
fragment,
so
the
fragmentation
is
independent
and
whether
you
have
sees
no
Rachel,
not
look
slide.
D
Next,
oh
so,
basically,
this
is
more
words
to
say
why
and
where
we
are
in
the
final
order
of
things.
Basically,
what
it
boils
down
to
say
is
that
this
is
an
example
of
iph
encapsulation.
That's
the
first
IP
IP
header
that
the
second
IP
either
you
find
that
we
always
start
with
a
sorcerer
header.
So
because,
when
you
have
to
to
rub
this
packet,
what
you
really
need
to
do
is
pop
the
first
entry,
the
first
next
up
in
this
rotten
either.
So
by
placing
this
first,
it
makes
the
forwarding
plane
easy.
D
D
If
you
don't
do
this
new
draft,
then
you
have
to
have
one
a
ripple
context
based
on
the
root
and
to
a
6lowpan
context,
which
will
have
the
exact
same
prefix
and
you
will
have
to
distribute
both
repo
distributes
the
roots
with
the
DI
use
and
through
a
race
will
distribute
the
6
lépine
context.
Distributing
the
12
them
for
a
single
result
seems
like
an
evacuee.
So
that's
why
we
introduced
this
in
your
head
or
thing
we
are
pretty
there.
So
thanks
Michael,
thanks
a
lot
for
being
the
Shepherd
for
this
work
and
guys.
G
So
Tomas
Medina,
just
let
you
know
about
the
blood
test
in
Berlin
that
will
take
place
right
before
the
IDF
meeting.
I
think
the
dates
are
16,
17,
I,
believe
that's
July,
and
so
this
this
will
most
probably
I,
cannot
say
it.
Definitely
because
we
haven't
formed
a
group
yet,
but
what
is
most
probably
going
to
be
one
of
the
things
that
are
tested,
so
we
were
going
to
do
this
or,
if
you're,
interested
contact
us
and
participated
like
this.
Thank
you.
Ok.
Well,.
B
I
Hi
I'm
Megan,
hi
hi,
so
so
honest,
you
know
one
of
the
things
I
would
wonder
if
you
could
speculate
for
us
is.
Is
there
a
cost
in
six
low
rh
for
the
fact
that
we
have
to
remove
that
our
pie
before
we
get
to
the
non
RPL
node?
Would
60
rh
be
would
be,
would
would
there
be
any
efficiencies
if
we
would
make
the
option
type
change
in
the
thing
you
speculate
on
that
I.
D
D
Should
probably
die
or
drop
the
back
head
to
something
wrong,
so
that
would
not
help
much
whether
we
call
it
a
drop,
a
ball
or
not
drop
a
ball,
because
just
to
give
some
context,
maybe
for
the
room
on
this
discussion.
There
is
a
discussion
at
six
men
and
probably
ate
some
slight
somewhere
about
that.
This
is
discussion
at
six
man
about
the
death
of
hop
by
hop
header
and
you
know
ripple,
is
using
heavily
a
bio
better.
I
But
but
the
point
is,
is
that
we
are
doing
some
effort
to
be
able
to
essentially
say
that
we
are
doing
encapsulation
and
D
capsulation
and
IP
IP
headers
on
a
hop
by
hop
basis
which
changes
the
addresses
in
the
outside
that
we're
putting
in
which
means
which
may
make
them
less
compressible,
which
is
one
concern.
I
have
such
that
we
can
remove
if
necessary.
I
If
you've
pointed
out,
we
can
remove
all
the
compression
and
we
can
deliver
the
packet
to
an
end
hosts
that
does
not
speak
six,
lower,
h
or
ripple,
and
we
can
do
that.
But
if
we
didn't
have
to
officially
room,
have
that
IP
IP
header
there,
we
wouldn't
have
to
compress
it
and
we
could
just
deliver
a
packet
directly
to
the
the
end
hosts
with
the
rpi.
Still
there.
D
Yes,
I
mean
we
could
leave
it.
I
mean
that's.
The
question
is
really
for
the
last
six
or
six
lr,
knowing
whether
it
should
remove
it
or
not,
which
is
because
I'd
living
it
or
removing
it
is
not
so
complex.
The
question
is
to
know
whether
you
have
to
do
it
or
not,
because
you
don't
know
what
kind
of
device
you
have
to
deliver
the
packet
to
which
is
the
tree.
D
The
problem
that
we
have
in
this
face
all
I'm
saying,
is
the
weights
formatted
right
now,
the
packet
from
from
the
root
with
just
an
LP
I
is
formatted
like
you
get
a
six
dollar
edged
with
the
rpi
easy
to
remove,
and
then
what
comes
after
is
the
60
to
80
2
pi
get
the
way
a
plane
host
expects
that.
So
that's
we,
the
design
point
now
the
decision
of
leaving
it
or
not,
leaving
it
is
kind
of
orthogonal.
It's
just
easy
to
remove
it.
That's
the
only
design
the.
G
I
A
bunch
of
stuff
there
that
we're
doing
to
keep
the
that
IP
IP
header
intact
because
of
the
sixth
IP
and
the
six-man
rules
about
adding
and
subtracting
headers.
So
as
a
bunch
of
stuff
that
we're
doing
so
that
we
are
able
to
reconstruct
the
uncompressed
packet
and
in
such
a
way
that
we
it's
addressed
to
that
host
so
that
we
can
remove
the
RPI
right.
D
Don't
see
that
would
be
able
to
compress
better
I,
see
that
we
would
be
able
to
leave
to
to.
It
would
be
easier
to
encapsulate,
because
you
would
be
able
to
say
what's
node,
which
will
be
capsulate.
That
would
be
the
final
destination
right
now
promise.
We
need
to
be
able
to
indicate
that
the
six,
a
law
that
serves
this
final
destination
is
the
one
that
should
the
gaps
light,
and
we
don't
know
how
to
do
that.
D
D
I
So
I'm
just
trying
to
figure
out
what
the
case
for
changing
the
type
code
is.
If
that,
if
the
result
is
you
know
to
bite
savings
on
every
hop,
then
that
would
seem
to
be
a
really
big
argument
for
why
we
should
change
it
immediately,
because
we
have
it.
We
have
with
this
document.
We
have
essentially
a
flag
day
right
there.
We
do
not.
I
We
do
not
propose
at
this
point
any
kind
of
signaling
to
say
we
support
this
new
new
protocol
or
we
don't
it's
going
to
be
the
operator
network
installer
who
either
turns
this
on,
because
all
the
devices
support
it
or
leaves
it
off,
because
some
devices
don't
support
it
right.
So
we
have
an
opportunity
for
a
flag
day
here.
Yeah.
D
D
J
F
J
A
J
F
F
J
Okay,
all
right,
okay,
so
nobody
just
to
just
a
little
recap
on
what
nope
at
dow
is.
The
no
path
now
is
essentially
in
the
down
with
a
lifetime
of
0.
It's
used
for
out
in
validation,
a
channel
road
when
it
switches
the
parent
node
its
with
it,
sends
lopat,
dow
whites,
why
it's
old
previous
part
and
a
regular
dau?
Why
its
new
path,
so
it
traverses
in
all
the
cases
all
the
towels.
J
Likewise,
the
regular
doubt
the
name
feed
also
traverses
upwards,
along
the
topology
in
the
upstream
direction,
so
you
I
new
bat
dau
might
be
important
for
us.
So
as
you
might,
you
might
already
know
that
routing
entries
are
the
biggest
memory
Hawking
component,
especially
in
the
bigger
story
mode,
article
networks
so
does.
J
If
there
is
a
contention,
it
will
be
nice
for
the
routers
to
know
which,
which
of
the
entries
are
non
active
and
can
be
removed
or
can
be
replaced
with
this
when
the
another
important
point
here
is
when
the
nodes,
which
is
a
parent,
it's
not
only
that
know
which
is
just
the
parent,
but
it's
the
complete
dependent
know
it
all
the
all.
The
other
different
nodes
along
along
beneath
that
give
that
child
node
also
switches
the
current.
J
J
And
can
you
please
up
on
to
the
next
all
right,
so
this
is
the
sample
topology
I
am
considering.
Let's
for
the
sea
of
this
discussion
considered
this
topology
no
D
might
switch
from
parent
v2c
for
certain
reasons,
such
as
transient
or
permanent
link
failure.
Another
reason
would
be
that
D
finds
a
better
parent
in
the
form
of
C
which
satisfies
better,
which
which
has
better
metrics
as
compared
to
X
previous
path
through,
be.
J
So
these
are
the
cases
that
have
highlighted
where
node
B
switches
from
Lodi
switches
from
node
B
to
C
for
various
reasons
such
as
link
breakages
because
of
transient
or
permanent
green
failures,
or
a
simple
reason
such
that
you,
the
new
metric
metric
from
from
the
previous
parent,
has
went
bad
and
there
is
an
alternative
path
available
through
C.
So
this
is
another
reason
when
NP
doubt
can
be
generated
in
all
the
cases.
Currently,
the
NP
doubt
can
be
generated
through
the
previous
path.
J
J
So
this
this
is.
This
essentially
is,
are
all
the
is
the
list
of
problems
that
are
associated
with
the
NP
doubt
mechanism
currently
specified
in
the
article
specifications,
so
the
first,
the
first
biggest
problem
is
the
NP
down
messaging
depends
on
the
link
which
the
node
had
decided
already
know.
Log
or
decided
to
no
longer
use
because
of
whatever
reason,
especially
if
the
reason
is
because
of
the
link
failure,
then
the
NP
dome
messaging
will
definitely
suffer
it's
a
guaranteed
failure
of
NP
double
messaging
there.
J
Another
important
point,
the
point
of
a
to
hear
is
that
went
wrong.
The
invalidation
happens
only
for
the
load
hood
channel
rates,
NP
thomas's.
Now
they
are
dependent
nodes
on
behalf
of
whom
there
is
no
NP
de,
which
is
generated
all
together.
So
on
the
previous
part,
there
is
no
way
of
generating
off
route
invalidation
along
the
previous
path.
J
D
If
you
lose
your
connectivity
to
that
particular
child,
then
you
know
that
this
child
is
lost
for
you,
but
you
also
know
that
if
he
was
the
only
way
to
get
to
all
this
served
AG,
then
all
the
nodes
in
this
set
dark
are
lost
as
well.
So
you're
entitled
to
send
a
note
out
to
all
of
them
is
no
double
its
name.
Pete
yeah.
J
F
D
Up
no,
the
subject
that
you
lose
because
if
you
know
you
lose
a
child,
then
because
you
get
the
route
via
this
child
to
all
the
rest
of
that
tag,
and
you
also
know
that
you
lose
this
step
deck,
so
you
can
found
a
new
path
dau
for
all
the
children
of
this
of
your
children,
all
the
family,
if
you
like
those,
are
you
guys
on
the
bottom
of
your
drawing?
They
will
get
a
no
doubt
if
this
guy
moves
from
left
to
right.
D
A
nice
good,
but
I'm
seein
names
from
where
I
am
and
that
there
is
ABCD,
but
I
don't
see
them
from
where
I
am.
It
means
ok,
ok,
so
you
seem
to
be
saying
that
there
is,
if
d
goes
away
from
be
ok
d
goes
away
from
be
so
this
is
broken.
So
I
understand
you
are
saying
that
there
is
no
no
path
battle
from
for
E
or
F.
All.
H
D
J
Yeah
first
year,
the
assumption
is
that
the
no
part
now
has
actually
reached
reached
from
the
previous
panel.
So
the
first
point
here
takes
takes
care
of
a
I
mean
such
is
the
problem
that
the
NPD
of
itself
might
not
reaching
all
the
cases
I
mean
along
the
previous
path.
There
is
a
possibility
that
infidel
might
not
reach
at
all
because
of
the
permanent
link
failure.
J
Let's
say,
for
example,
let's
take
the
scenario
of
permanent
link
failure,
so
there
is
a
there
is
a
chance
that
you
know
NP
da
might
not
reach
at
all,
so
there
is
no
that
there
is
a
guaranteed.
Not
working
of
NP
de
might
not
work
in
a
guaranteed
way
if
there
is
a
permanent
ink
failure
along
the
previous
time,
if
you're
saying,
if
there
is
a
way
if
there
is
an
alternative
way
of
GU,
realizing
that
there
is
a
link
breakage,
then
it
can
generate
to
this
in
the
subsequent
slide.
J
I
will
just
talk
about
you
know
what
are
the
problems
that
you
know
all
the
other
nodes
kept
the
there
is
a
problem
in
the
in
the
sense
that
and
another
node
cannot
generate
a
low
fat
now
on
behalf
of
the
target
node.
So
especially
because
there
is
a
there
is
a
dependency
in
terms
of
state
information
such
as
Dow
sequence.
J
So
it's
I
mean
there.
Is
there
ism?
There
is
a
sentence.
There
is
a
line
that
is
a
statement
mentioned
in
the
artists,
pay
expect,
which
talks
about
sending
a
packet
a
packet
to
invalidate
the
routes
and
clear
the
states
clear,
the
Argos
state.
But
it
is
easier
said
than
done.
You
know,
especially
because
every
town
has
state
information.
F
J
J
Here,
can
you
please
move
on
to
the
previous
slide
again
the
yeah,
so
here
the
other
problems
that
I
mentioned
is
the
possibility
of
crowd
type
route
down
time
because
of
the
asynchronous
operations
of
entry
tower
and
down
so
NP
da
is
sent
through
another
path
and
now
is
sent
through
another
another
path.
So
there
is
a
possibility
that
NP
de
actually
reaches
and
out
never
reaches,
because
even
though
I
even
do
it,
the
the
load
has
sent
the
doubt
that
are,
there
is
no
guarantee
that
now
might
actually
reach
the
water
out.
J
So
there
is
a
possibility
that,
because
of
this
synchronous
mode
of
operation,
that
there
is
a
downtime.
Secondly,
in
terms
of,
let's
assume
for
the
sake
of
discussion,
that
there
is
no
way
of
you
know
clearing
clearing
the
routing
table
if
there
is
no
impede
out
reaching
the
previous
part,
in
which
case
there
is
an
impact
on
p2p
traffic,
the
example.
The
point
of
a
foo
here
shows
an
example.
If
G
originates
up
traffic
for
e
and
f
and
if
G
has
invalidated
invalid
routes,
then
it
will
cause
traffic
failure.
J
B
B
J
Million
means
I
can
just
at
the
point
that
I
want
you
to
make
is
from
the
requirements
point
of
view.
What
can
be
an
optimal
solution
for
an
MP
now
salute
bit
of
a
problem
statement,
so
the
the
next
couple
of
slides
talks
about
the
possible
existing
solutions
and
which
we
discussed
on
the
mailing
list
thanks
to
some
kind
salmon
and
then
maybe
on
the
this
particular
slide
talks
about
some
specific
scenario
that
can
be
handled.
You
know
using
douek
mechanism.
J
J
The
network
throughout
formation
is
hundred
articles,
the
routing
validation
else,
which
can
be
handed
quantity
by
a
ripple.
Secondly,
this
the
route
down
time
that
it
should
be
possible
from
the
requirements
point
of
view
to
avoid
route
down
time
for
n
fit
out
because
of
mt
down
and
our
fresh.
So
so
this
this
is
from
the
requirements
point
of
him
yeah.
That's
the.
B
B
Okay,
this
is
quite
new
subject.
It
is
something
which
has
been
an
old
fish
in
the
old
working
group
that
there
will
be
a
management
and
MEB
to
define
the
parameters
and
to
look
at
the
operation
of
nipple
and
ripple.
So
what
you
have
done
now
is
that
you
didn't
make
a
nip
in
this
my
language,
but
we
have
used
young,
which
is
a
more
convenient
language
here,
to
describe
good
model.
So
what
I
want
to
do
here
is
just
described
to
model
and
see
if
there
is
any
interest
that
we
continue.
B
This
work
in
this
working
group
I
have
separated
the
model
in
four
parts.
As
far
as
I
remember,
one
is
about
how
the
multicast
addresses
spin
up
a
domain
because
maple
knows
domains
which
has
spent
up
by
multicast
addresses
next
slide.
Please
then,
I
have
a
set
of
operational
parameters
which
are
the
setting
actually
of
the
people,
such
as
the
lifetime.
The
domain
eight,
the
amine
in
the
Emacs,
which
other
operational
parameters
used
to
start
the
maple
operation
of
the
next
section.
B
That
is
the
actual
execution
of
the
maple
spare
seat
and
ending
the
domain
so
that,
well
you
look
at
it.
You
can
see
how
the
parameters
perform.
For
example,
this
here
you
have
the
pocket
that
current
interval,
the
seat
count
on
how
many
packets
has
arrived
and
the
t
that
an
expected
has
to
be
sent
and
the.
B
Finally,
the
statistics
which
gives
you
from
how
many
messages
have
been
received,
forwarded
to
read
a
rejected,
not
received,
etc.
So
these
are
the
four
parts.
I
would
welcome
some
some
comments
on
the
mailing
list,
because
the
domain
is
rather
complex
with
lots
of
a
race.
You
may
say
that
for
a
small
device,
that's
not
wanted
at
all
and
that
the
move,
for
example,
the
domain
definition
is
too
complex
or
that
you
say
the
different
item.
B
J
A
A
A
A
Ok,
we
have
ok,
we
have
like
a
some
items
for
maintenance.
The
protocols
like
well
a
proposed
from
raul.
Oh,
we
have
as
well
some
other
draft
that
we
don't
have
so
much
and
make
me
about
this
modification.
So
will
be
nice
if
you
can
review,
they
know
the
working
group
graph
and
send
your
comments
and
pascal
order
tell
projection.
Do
you
want
like
we
have
a
some
story
mode
in
non
stirring?
You
want
to
say
something
about
this.
D
A
A
D
B
B
Then
the
automatic
selection,
okay.
This
is
another
point
which
came
for
me.
If
you
look
at
the
nipple
and
you
have
been
very
dense
network
with
lots
of
maple
forwarders,
then,
but
you
get
to
too
many
much
too
many
a
slightly
a
package
which
are
sensory,
catan,
so
no
pocus
almost
so
what
I
propose
is
that
as
an
automatic
selection
of
maple
forwarders
dependent
on
the
density
of
the
network,
so
that
you
reduce
the
number
of
verbal
forbearers
and
you
haven't
possibly
minimum
the
most
maple
forwarders.
B
B
B
H
Thank
you,
a
better,
not
touch
it.
So,
38
years
ago,
I
had
a
short
presentation,
I
think
at
six
low
on
this
idea
of
using
ripple
for
source
voltage
multicast.
This
is
essentially
taking
the
ideas
of
the
beer
group
and
applying
them
to
too
constrained
environments.
So
I
haven't
asked
for
presentation
time
here,
because
time
is
short,
so
you
can
look
up
the
slides
from
a
year
ago
and
this
has
been
implemented.
H
B
H
It's
not
just
the
performance
issues,
it's
also
an
issue.
What
roles
do
you
give
the
nodes
in
the
network,
so
nipple
makes
relatively
high
requirements
on
all
the
non
root
routers
in
the
network.
This
constraint
cast
really
is
non
storing
mode
Saudi,
everybody
else,
but
the
root
has
no
idea.
What's
going
on
and
just
forwards
packets,
according
to
what's
in
the
package.
D
Let's
catch
email,
I
one
thing,
I
support
this
work
and
the
other
thing
is
this:
it
seems
tint
in
tainted
off
of
flag
multicast,
but
the
use
of
beer
for
repo
can
be
of
interest
for
unicast
as
well
as
multicast,
and
the
value
would
be
a
better
compression
of
what
we
are
doing.
Even
for
unicast
simplification
of
what
you
put
in
the
dough
you
just
with
a
simple
bitmaps.
You
can
aggregate
all
the
dowel
rods
and
that
basically
gives
us
with
a
limited
state.
D
The
capability
to
do
a
story
mode
in
modes
which
would
not
be
capable
of
doing
story
mode.
Is
there
to
store
the
full
of
dresses,
so
you
would
basically
express
each
router
with
a
bit.
So
not
only
is
this
very
interesting
that
the
work
that
Gaston
is
proposing
us,
but
the
fact
of
using
beer
in
repo,
I
think,
is
of
high
interest.