►
From YouTube: IETF101-ROLL-20180323-0930
Description
ROLL meeting session at IETF101
2018/03/23 0930
https://datatracker.ietf.org/meeting/101/proceedings/
A
B
B
C
C
A
C
This
desire
to
have
a
simple
IOT
device
in
a
very
big
room
like
this
with
many
quote:
unquote,
access
points
on
the
ceiling
or
something
and
the
device
would
just
like
a
Wi-Fi
device
today,
move
across
this
room
and
without
having
to
change
it
security
associations
without
having
to
form
new
addresses
and
renumber
and
everything.
Charlie
I
took
your
slot
because
you
were
not
there,
but
they
promised
you
would
you
know.
I
took
my
risk
so
you'll,
be
you
just
keep
this
long.
C
D
C
So,
basically,
we
are
talking
about
a
device
which
attaches
in
our
case
to
a
ripple
network.
So
we
call
it
a
leaf,
but
it's
not
even
ripple
aware
so
for
the
sake
of
this
document,
we
will
call
it
a
ripple
unaware
leaf.
So
there
is
this
sad
acronym
of
rule.
We
can
change
it
by
name,
but
it
is
a
ripple
unaware
leaf
some
reports.
C
The
concept
of
leaf
leaf
can
be
should
be
fundamentally
aware
that
there
is
repo,
even
if
it's
not
forwarding
ripple
unaware
leaf,
is
sitting
at
the
edge
of
a
ripple
Network
and
doesn't
have
to
know
about
it,
which
means
that
the
simple
interaction
it
will
have
with
the
6r
are
using
6lowpan.
Indy
should
be
enough
for
repo
to
Ford
the
packet
all
the
way
to
the
root
and
should
be
enough
for
the
route
to
do
nd
proxy
for
the
device
if
there
is
a
backbone.
C
So
actually
that
doesn't
take
a
lot
on
the
repo
side
to
achieve,
but
it
took
a
lot
on
the
6lowpan
to
achieve
so
my
my
long
story,
my
friday
story
will
be
all
the
steps
that
take
us
to
this
situation,
which
looks
so
easy.
Well,
you've
got
this
simple
IOT
device,
doing
simple,
simply
lay
or
to
access,
and
then
6lowpan
ng
roaming
across
a
very
big
room,
with
many
routers,
many,
whatever
meshes,
etc.
Without
caring
about
all
this
complexity,
the
story
starts.
So
you
may
you
may
want
to
read
that
one
of
those
days.
C
So
please
read
the
sort
of
things
this
is.
This
is
kind
of
the
reasons
why
we
ended
up
doing
all
this
work,
but
but
they
won't
have
time
today,
so
I
will
start
with
the
work
that
we
had
to
do.
An
ND
I
won't
be
as
complete
as
I
was
yesterday
at
6.
Lo,
but
I
would
will
show
you
what
we
had
to
do
in
nd,
so
that
we
can
make
ripples
work
possible
to
actually
allow
this
device
to
roam
across
multiple
ripple.
Do
docks
without
even
knowing
about
so.
C
The
first
big
thing
we
had
to
do
is
change
6lowpan
ND
and
the
reason
why
we
had
to
change
the
slapping
andy
is
that
a
device
that
moves
across
the
fabric
must
provide.
There
must
be
enough
info
information
for
the
fabric
to
figure
out
what
is
the
latest
location
of
that
device.
There
are
two
ways
to
classical
ways
of
doing
this.
C
The
first
one
is
the
fabric
captures
a
very,
very
precise
off
when
the
device
is
seen
and
then
compare
us,
you
know,
is
our
multiple
places
where
the
device
has
been
seen
recently
compares
the
most
recent
finds
the
most
recent
timestamp.
The
prime
with
that
approach
is
that
the
mobility
of
the
device
can
be
very
slow
or
very
fast,
and
the
amount
of
space
and
the
precision
of
the
timespan
of
the
time
stamp
can
vary
a
lot.
C
So
it's
very
hard
in
a
UT
situation
to
find
the
right
unit
for
a
timestamp
and
the
right
precision
of
the
timer
and
exposed
that
so
what
ripple
does
and
what
now
6lowpan
ng
we'll
be
doing
with
this
change
is
that
the
source
keeps
a
sequence.
So
that's
the
AO
t,
v--'s
destination
sequence.
Number,
if
you
like.
That's
that's
a
number
that
the
IOT
device
the
6ln
has
to
maintain,
because
the
routing
will
need
it
to
figure
out
the
most
recent
location.
C
So
now,
when
this
new
work
that
we
are
doing
here,
repair
in
our
lives
acts
on
behalf
of
this
device,
it
will
have
a
sequence
number
which
will
allow
to
find
the
freshest
location
of
the
device
in
the
network
and
set
the
routing
appropriately.
So
that's
a
big
chance.
We
made
in
six
the
patent
and
either
otherwise.
I
will
show
you
second
thing
we
are
changing
in
6lowpan
and
E
is
source
address
validation.
C
There
are
two
aspects
in
Savi,
so
first
aspect
years
when
somebody
forms
an
address,
if
is
the
first
in
the
network,
it
should
hone
that
address.
That's
handy,
but
the
question
is:
can
somebody
steal
that
address
with
the
current
and
if
loose,
that's
very
easy?
You
just
exposed
the
same
address.
Now
you
get
the
traffic
so
on
on
powerful
devices.
The
IDF
has
defined
CGI
AMC
and
which
RFC,
Sutton
and
71
22,
and
that
enables
to
protect
the
address
against
theft.
But
it's
a
very
complex
protocol.
C
If
you
take
six
up,
and
indeed
the
first
document
and
the
backbone
router,
what
you
get
really
is
the
lay
of
three
equivalent
of
dots
11,
a
total
11
Association,
meaning,
if
you
figure
out
what
really
dot
11
does
with
with
an
association
on
wires
when
you're
looking
up
somebody,
you
just
send
a
broadcast
about
spending
3,
while
the
old
fabrics
right
you
send
a
broadcast
of
spanning
tree
and
that
that
feels
the
transparent,
bridging
state,
the
source
of
this
broadcast.
We
know
areas
along
the
spanning
tree
now,
that's
ok
on
wires!
C
C
And
now
the
AP
will
know
whether
to
copy
over
a
multicast,
Avella,
radio
or
not,
because
he
already
has
proactively
a
state
which
they
assume
you
need
to
do
it
or
not,
and
we
had
the
exact
same
prime
on
our
robot
devices.
We
didn't
want
to
have
all
those
broadcasts,
so
what
we
end
up
doing
with
the
RFC
67
and
the
earth
diet
is
the
registration
process
in
six
leper
in
six
of
energy,
which
is
very
much
equivalent
to
the
Association
and
the
equivalent
to
the
transplant.
C
Bridging
operation
is
done
by
the
backbone
router,
so
this
is
the
guy
which
will
do
ND
proxy
so
that
you
look
like
a
normal
and
d
device
over
the
Ethernet
backbone,
but
still
like
an
association,
the
backbone
will
know
when
to
copy
you
and
the
rest
of
the
time
you
don't.
You
won't
get
all
those
multicast
so
that
that's
what
this
series
of
draft
do
and
I
will
just
tell
you
the
most.
C
The
changes
with
it
in
the
update
so
has
to
serve
ripple,
but
it's
not
just
to
us
to
serve
ripple
is
so
as
to
serve
any
routing
protocol
which
behaves
a
bit
like
ripple
supports
mobility.
So
you
need
time
stamps
of
sequence.
Country
will
see
that
we
have
a
sequence
counter
and
you
need
a
lifetime
for
getting
a
state
in
the
network.
So
we
see
how
we
did
that
on
the
side
we
changed
from
six.
C
We
can
only
protect
it
on
the
first
hop
with
the
AP
and
you
drafted
I
just
discussed
right,
so
there
might
be
interests
in
future
work
so
that
you're
sure
that
what
you're
exposing
in
your
dowels
is
actually
true
right
now
we
just
know
it
in
the
first
half.
So
if
your
first
hop
is
rich
rusted,
then
you're,
okay,
now,
if
the
first
hub
could
be
a
trick,
then
anything
still
could
leak
in
the
repo
fabric
and
that
would
disrupt
rotting.
So
right
now,
we've
done
it
on
6lowpan.
We've
done,
we've
not
done
internal
report.
C
C
This
bit
is
is
of
utmost
importance
and
that
was
discussed
on
the
raw
mailing
list,
the
orbit
so
in
six
open,
nd
this
field,
which
is
now
the
orbit,
was
reserved
and
current
ripple
implementation
which
to
ripple
and
6lowpan,
and
you
as
the
shirt
well
the
idea.
Since
it's
a
result
field,
they
will
set
it
to
zero.
C
So
the
orbit
is
not
set
by
routers
and
it
is
set
by
hosts
implementing
this
new
spec
and
what
it
means
is:
oh,
if
there
is
a
rotting
fabric
or
if
there
is
some
proxy
activity
or
anything
else,
please
do
it
for
me,
make
sure
to
get
my
rich
ability,
because
I
want
to
be
doing
it
myself.
So
a
repo
rata
will
always
set
the
orbit
to
zero
and
the
six
eleven
which
doesn't
participate
to
the
writing
doesn't
even
care
if
there
is
writing
mesh
under
just
proxy.
C
C
So
this
TIG
as
the
exact
same
description
as
the
path
sequence,
you
can
take
a
tid
coming
from
a
ripple
on
our
leaf
and
stuff
that
natal
as
the
path
sequence
now
will
change
slightly
the
text
to
say
that,
but
people
did
not
want
direct
reference
because
the
6lu
people
are
very
keen
that
this
spec
is
not
dependent
on
ripple
right.
So
it's
an
exact
copy
of
the
text.
We
can
use
it
with
the
current
ripple
specs.
That's.
E
C
C
C
Repo
was
a
bit
more
farsighted
than
this,
because
we
we
have
a
configuration
option
where
we
tell
what
the
time
unit
is
so
now
in
the
dowel.
We
can
express
the
time
in
with
less
bits,
because
there
is
a
multiplier
of
some
sort,
which
is
the
time
unit.
So
since
it's
not
the
same
unit
when
you
want
to
match
a
dowel
I'm
sorry,
when
you
want
to
match
the
registration
coming
from
a
6ln
and
turn
that
into
a
dowel
to
get
relative
right,
you
will
have
to
do
the
unit
mapping.
C
E
B
C
C
E
E
The
math,
but
it
should
be
assured
the
reason
why
I
know
about
this
is
because,
when
I
experiment
with
most
of
the
open
sources,
none
of
the
ones
is
currently
implement,
use
part
sequence:
I
mean
they
don't
use
part
sequence,
the
same
sequence,
but
they
don't
really
act
on
the
path
sequence
and
we
had
a
discussion
on
the
mailing
list.
I
remember
you
said
that
the
porn
sequence
is
used
to
clear
the
downwards
rap.
No,
that's.
C
The
yes
M
Oh,
what
are
you
talking
about?
I
know
the
basic
one
says
he's
supposed
to
not
to
clean
the
downwards
it's
supposed
to
update.
Well,
okay,
there
are
two
actions
when
you
have
a
chance
right,
one
of
them
is
easy
set
up
the
new
path.
That's
what
the
paths
take
on
stairs
and
on
the
current
parrot,
you
will
clear
the
olds
everything
related
to
the
old
sequence,
but
then
there
is
this
action
of
pruning,
which
is
new
to
your
draft
pruning
all
the
paths
and
that's
that's
new
job,
but
yeah.
E
I
mean
keeping
the
other
side,
because
in
my
draft
at
least
I,
don't
talk
about
part
sequence
at
all,
because
it
doesn't
I
talked
about
part
sequence,
but
it's
just
a
part.
So
if
I
take
the
base
ripple
specification,
if
I
take
the
base
level
specification,
you
you're
saying
it
has.
If
transit
information
is
not
mandatory,
it
has
to
be
mandatory
for
this
or.
C
B
C
E
C
Look
at
it
that's
a
good
point
that
let's
not
spend
the
missing
like
transit
information
is
there
and
if
the
implement,
okay,
very
big
message
here,
there
are
some
quick
and
dirty
implementations.
Are
around
I,
won't
give
names,
but
don't
use
that
as
a
reference
of
what
triple
A's
don't
write.
A
paper
saying
ripple
doesn't
work
in
this
case
because
that
particular
open
source
implementation,
which
was
done
by
somebody,
was
not.
There
doesn't
work
right,
there's
more
to
report
and
then
a
particular
I
would
say
quick
and
dirty
implementation
funnel
open
source.
C
C
But,
as
I
said,
we
have
to
change
the
unit,
and
this
is
just
the
Xu
I-64
field
ripple
doesn't
need
it
for
now,
but
if
we
define
a
way
to
trust
hop-by-hop
to
Dowe
information
you're
getting
in
not
in
storing
mode
or
if
the
OPR
and
route
talk
together
in
non
storing
mode,
which
is
hopefully
what
they
do,
then
we
can
actually
trust
that.
What
we're
seeing
here
is
true,
like
the
dowel
I
think,
is
the
same.
C
Guy
can
correlate
the
nan,
storing
dowel
with
a
direct
message
so
just
to
give
you
a
hint
of
normal
boot
of
a
network,
so
so
here
I
get
a
network
with
a
backbone
router.
So
this
is
a
tradition
Ethernet,
whatever
switched
networks,
a
running
traditional
neighbor
discovery.
This
is
the
backbone
router
doing
in
the
proxy,
and
this
is
all
one
big
fat
subnet.
So
this
room,
like
I,
said
this
factory
floor.
This
huge
thing
is
one
big
fat,
multi-link
subnet,
one
single,
slash,
64.
Why?
C
Because
you
want
seamless
mobility,
you
don't
want
to
form
a
new
trace
going
some
data
center
and
if
it's
a
different
band,
changing
all
your
security
tokens
and
everything
we're
joining
so
you've
got
this
big
fat
backbone
and
what
your
FC
67
35
slightly
updated,
do
is
well.
The
array
will
carry
some
information
like
the
prefix
you're
using
in
this
a
64.
It
will
also
carry
information
about
the
the
border.
Router.
That's
about
the
albion.
B
C
O
is
the
area,
and
this
is
new-
this-
the
new
spec
mandates
a
6e
option,
which
is
our
c7
7000
blah
blah,
and
that
tells
you
some
bits
that
can
exposing
the
array
about
capabilities
and
the
capability
that
we'll
have
here
is
support
for
by
the
6lb
awed
by
the
6ro
of
this
news
bag
right.
So
when
we
build
a
network
which
will
take
lives
and
and
do
the
ripple
for
them,
everybody
must
support
this
bag.
C
So
you
must
have
a
six
CIO
and
the
six
year
you
must
say:
hey
I'm,
supporting
this
back
at
the
6
lb,
are
under
two
six
zero
and
that's
how
we
know
we
can
play
from
the
earth.
There
will
be
the
six
up
and
registration.
A
few
things
changed
the
the
biggest
changes
in
traditional
six
dependency.
The
registration
is
for
the
source
address
of
DNS,
but
that
was
a
problem
when
we
want
to
do
the
proxy
registration
by
6lb.
C
Our
who
happens
to
be
the
root
normally
we'll
discuss
that
a
on
the
sixth
video,
because
the
source
of
this
semester
has
to
be
the
six
L
Bureau.
So
if
that
was
what
your
registers
you're
registering,
you
would
be
screwed.
So
what
we
are
now
registering
with
the
update
is
the
target
you
know
in
DNS
any
messages
there
is
a
field
called
target.
Usually
that's
what
the
message
talks
about.
Well,
we
made
that
consistent
with
the
old
or
other
hand,
and
these
facts.
What
the
message
talks
about
is
the
target.
C
It
talks
to
with
his
address
and
reciprocally,
this
guy
has
only
one
note
it
talks
to
with
this
hurly
cockatrice,
so
the
link
is
really
just
two
guys
talking
together.
That's
what
the
link
becomes
with
a
scope
for
little
Contras.
If
that,
so
then,
the
uniqueness
of
the
link
local
of
this
guide
is
what
this
guy
sees.
So
you
don't
have
to
do
any
big
flow
like
like
the
attack
right,
it's
a
local
locally
managed.
C
You
could
say
the
same
thing
if
this
guy
here
as
a
sledge
64.
So
in
some
deployments
you
could
say,
hey
I
have
a
big
/
64
for
my
net
worth,
but
they
also
have
a
specific
tool:
a
486
Allah.
In
this
case
you
would
not
have
to
go
all
the
way
to
this
LVL
because
it
would
be
the
6
lb.
R
4
is
on
prefix,
okay,
so
that
there's
no
need
to
have
one
appear
for
all
the
prefixes
which
are
around
each
prefix
as
its
own
lb
o.
G
C
E
C
The
spec
says,
if
you
know
by
your
6
cio,
if
the
6
lb
or
supports
it,
the
6
LVL
does
not
support
it.
You
come
to
any
of
the
security
thing
right,
because
you
better
have
large
of
keys,
even
in
so
of
the
key
I
mean
it's
easier
to
attack,
if
is
shorter
and
longer.
So
that's
why
we
want
to
extend
the
size.
So
the
security
aspect
is
really
weaker.
C
If
you
don't
support
the
new
spec
now,
if
so,
so
this
the
dark
side,
if
you
see
both
attached
to
something
which
about
the
new
spec,
because
you'll
get
more
services.
If
you
have
no
choice,
then
you
can
attach
to
a
6-0.
But
you
know
by
the
60
IO
that
it's
a
6-0,
the
classical
one,
then
you
know
you
cannot
do
any
of
this
right.
So
so,
basically,
that's
it!
C
Now,
if
you
attach
to
a
6-0
that
supports
it
and
on
the
ripple
path,
there
is
a
six
L
out
that
does
not
as
long
as
the
6lr
attach
to
and
the
6lb
are,
both
supported,
you're
all
set.
You
can
have
no
more
routers
on
the
way,
but
you
should
be
able
to
see
a
six
era
which
does
this
otherwise
you're
screwed.
C
Okay,
so
basically
the
response
of
this
flow,
which
is
you
know,
one
half,
multi
hot,
damn
SH
duplicate,
address,
request,
mapped
into
by
the
CSLB
R
into
an
SRO
to
feed
the
60
B
I'll
get
a
proxy
state
in
place
and
back
well.
This
is
now
described.
Okay,
it's
a
combination
of
operation
by
the
RFC
update
around
the
backbone
router,
which
explains
how
the
proxy
works.
So
the
signaling
is
in
the
update
document.
How
you
do
the
proxy
operation
is
in
the
background
proxies
back
and
this
isn't
the
register
review.
C
There
were
already
many
many
commands,
so
we
are
going
through
them
and
there
is
nothing
that
hurts
too
much
and
we
really
thank
all
the
reviews
really
well
that
we
got
Adrian
did
it
for
the
rotting
area.
So
we
had
some
exchanges,
it's
not
closed
yet,
but
we
are
very
close
to
closure
ideas.
So
this
the
RFC
6725
update
as
far
as
I
see
it
is
very,
very
close
to
our
C
editor
now.
C
So
I
won't
talk
about
a
P
and
E,
but
if
you
want
to
look
in
more
detail,
that's
how
we
do
the
validation
of
the
address
so
that
basically,
as
I
said,
gives
you
two
things
that
gives
you
the
capability
to
to
form
any
address.
You
want
and
get
it
protected
later,
so
nobody
can
steal
your
drawers
you're
still
the
onion
are.
C
The
second
thing
is
that
the
6lr
now
can
see
packets
coming
in
and
if
the
source
is
not
the
owner
of
the
address,
it
can
filter
out
this
packet
because
it's
not
topologically
correct
question
right
so
so
that
protects
the
owner
of
the
address
against
somebody
else,
impersonating
him
and
that
protects
the
network
and
the
rest
of
the
world
against
somebody
sending
a
package
which
is
sourced
a
televised
which
is
not
him.
That's
traditional
sorcerers
validation,
stuff,
savvy.
C
Okay,
so
we
are
working
on
the
crypto
bla
bla
bla,
like
I,
said
the
most
important
piece
that
we
need
to
know
right
now
is
only
the
first
harp
is
protected.
Ripple
is
not
then
the
backbone
router
again,
I
won't
speak
too
much
just
show
you
this
flow.
So
there
was
a
discussion
yesterday
about
whether
the
dialog
is
local
or
global,
and
whether
or
not
let
me
be
very
clear-
it's
local
right
has
always
been
local.
It's
then
against
open-source
guys
can
play
with
our
implementation
do
what
they
like,
but
it's
not
a
protocol.
C
If
it's
not
clear
in
the
protocol,
we
talk
to
our
world,
we
need
to
clarify
the
protocol.
That's
how
spec
goes
into
the
internet
level
does
not
level
right
now.
Repo
is
not
internet
standard
level.
We
need
to
refine
this
sort
of
things
so
that
it
becomes.
It
does
double
level.
So
we'll
do
meaningless
exchanges
to
explain
things
about
roles
documents.
Maybe
we
will
turn
that
into
some
small
documents,
which
explains
better
of
things
like
that.
C
But,
yes,
the
dead
work
is
a
local
acknowledgement,
whether
you
use
the
link,
the
lower
layer,
acknowledgement-
or
you
said,
the
real
dialog
as
a
significance.
If
you
use
a
lower
layer
acknowledgement,
then
it
really
means
that
you
trust
that
you
will
never
lose
it
between
the
match
and
repo.
If
you
do
real
dialog
from
the
repo
level,
then
the
sender
of
the
Dower
knows
that
it
was
received.
C
Now
you
must
be
very
aware
when
you
are
doing
a
ripple
implementation,
that
if
you
earned
out
and
storing
more
down
the
guy
who
acknowledges
it,
takes
responsibility.
It's
like
a
letter
from
the
king,
that's
being
transmitted
hop-by-hop
right.
If
you
accept
the
letter,
you
can
reuse
it
you
can.
So
you
can
reuse
it
down
all
right.
You
can
say
down.
Och
I
mean
I,
don't
want
it,
but
if
you
accept
it,
it
means
you
will
handle
it
now,
which
means
that
you
take
the
commitment
to
send
it
to
your
parents.
C
If,
eventually
it
fails,
you
can
not
send
it
to
a
parent
after
retries
and
whatever
are
trying
on
your
parents.
It
really
needs
that
you
lost
connectivity.
If
you
lost
connectivity,
then
you
need
to
poison,
and
your
children
would
took
responsibility
of
this
will
advertise
it
to
another
parent,
but
the
case
where
a
dau
stay
stuck
in
the
middle
of
the
network
is
not
supposed
to
exist
right.
It's
it's!
It's
somebody
accepting!
Oh
I,
will
do
that
for
you.
I
put
it
in
my
pocket:
I
promise
I
will
give
it.
C
Somebody
goes
to
somebody's
home
somebody's,
not
there.
Okay,
scratch
you
later
now.
You
cannot
do
that
right.
So
you
you
accepted
a
you,
take
responsibility.
You
have
to
follow
it
up
or
you
have
to
say:
I
cannot
be
a
router
anymore
either
way,
but
by
the
way.
So
that
was
fine
yesterday
discussion,
so
so
so
so
there
we
go
the
dow
so
that
in
story
mode,
that's
multiple
hops
doing
this
in
non
stirring
is
just
end
to
end,
so
the
ripple
route
gets.
Yet.
C
There
is
a
big
discussion
and
we
live
it
on
the
leaf
document.
Without
a
ripple
root
is
the
Albia
that's
an
interesting
topic
in
in
in
the
leaf
document.
I
showed
how
we
could
do
it
if
they
are
separate
in
my
mind,
it
makes
no
sense
to
separate
them.
Actually,
if
you
look
at
the
definition
of
a
6
lb,
Rho
BR
means
border
rod
right
now.
The
way
6lowpan
using
the
6l
Bureau
is
not
as
a
border,
router
or
6lowpan
uses.
The
6
lb
are
as
a
registry,
which
could
be
anywhere
in
the
universe.
C
Long
as
you
can
talk
to
it,
but
it's
called
a
6
lb
are.
It
makes
a
lot
of
sense
that
this
registrar
is
actually
the
bottle
rod
for
many
reasons.
So
even
if
I
specify
a
show
that
we
could
separate
them,
it
might
be
that
this
specification
say
says
they
must
be
the
same
right.
We
laugh
that's
one
of
the
big
discussion.
C
There
are
not
many
big
discussion
on
my
document,
but
one
of
the
big
discussions
we
will
have
is
do
we
must
that
the
river
route
and
the
6
will
be
our
artisan
guy,
which
is
logic
from
the
name
that
6mm
gave
to
this
device.
That's
why
I
can
say,
and
otherwise
there
are
additional
throws,
which
are
really
waste
of
time.
So
I
will
show
this
flows,
but
where
we
go
so
that's
why
we
are
the
the
cool
thing.
Is
you
know
this?
Is
this
is
done
repeatedly
every
lifetime?
C
You
want
to
maintain
the
stage
you
want
to
maintain
the
the
registration,
the
six
or
now
you
want
to
maintain
your
down
state
in
the
route
and
in
story
mode
along
the
path,
and
then
you
want
to
maintain
the
nd
proxy
state.
All
this
requires
that
periodically.
This
is
set
now
remember.
If
normally,
when
you
have
an
SRO,
you
must
have
a
DA
deck.
C
All
the
way
to
the
6lb
are
across
the
mesh,
that's
a
cost,
but
a
since
we
have
ripple
doing,
though
we
can
actually
use
the
fact
that
we
got
to
down
at
the
route
which
can
coexist
with
the
six
lvl
to
say
a
mr.
Cecil
B.
Are
we
okay?
We
got
to
Dow
that's
equivalent
to
the
DAC
right,
in
which
case
you
don't
have
to
do
both.
C
So
in
the
the
draft
suggests
that
in
the
real
life
you
know
at
the
beginning,
you
have
to
do
the
data
for
the
phone
with
address
detection,
duplicator,
eyes
detection.
But
after
that
we
say
a
the
six
or
the
river
route
can
act
as
a
proxy
for
doing
the
DAC
with
a
6
lb
or
if
they
are
co-located
is
just
some
internal
function
call
without
not
co-located.
That
must
be
a
nice,
oh
I'm
signaling,
but
at
least
it's
local
to
here
doesn't
have
to
go
across
the
whole
night.
C
Let's
part
of
the
discussion
in
the
draft
10
minutes,
yeah
I'm
good,
no
worries
we
another
thing
which
is
expressed
here.
People
can
not
do
that
because
we
we
cannot
pass
the
Rover
all
the
way
across.
So
we
cannot
secure
anything,
and
we
cannot
do
anything
like
that.
So
if
free
poll
shows,
with
the
same
tid
same
guy,
coming
from
two
different
direction,
ripple
will
keep.
The
two
possible
paths
will
be
all
do
tax
right,
so
we're
happy
to
have
two
paths
to
get
somewhere
actually.
C
Nearest
maybe
the
meat
of
the
current
document.
So
now
you
see
the
background
right.
We
have
changed
6lowpan
ng,
so
that,
in
a
fashion
which
is
abstract
agnostic
to
ripple,
we
have
all
the
information
we
need
in
6lowpan
ng
to
actually
populate
the
state
in
the
six
so
that
the
6lr
can
do
ripple
on
the
ALF
of
the
device.
That's
what
we
did.
We
also
made
it
so
that
this
can,
at
the
other
edge
at
the
root.
C
Now
we
have
the
means
to
have
what
they
said
earlier
in
this
room
of
many
many
Badwan
routers,
a
big
Ethernet
fabric,
collecting
all
the
backbone
routers
forming
many
many
do
dogs
and
the
6lowpan
device
completely
unaware
of
ripple
will
be
able
to
run
well
just
to
a
6lowpan
and
the
update
with
this
sequence,
contour
in
it
there
will
be
a
new
dowel
in
the
same
dog
or
in
a
different
day.
You
see
the
power
of
it
very
simple,
host
and
yet
the
whole
power
of
ripple
by
quadrotor
everything
to
solve
it.
C
What
do
we
need
to
do
on
the
ripple
side
to
get
it
well
ripple
as
a
bit,
which
is
called
the
e
bit,
which
allows
to
inform
I?
Think
it's
also
in
the
target
option.
I
mean
you
can't
live
without
it,
which
basically
says
this
is
an
internal
tries
on
external
drives?
If
it's
Nexon
tries
it
remains.
Ok,
it's
something
that
was
redistributed
redistributed
into
ripple.
You
don't
really
expect
report
to
be
the
backbone
of
something
it
could,
but
it's
really
good,
but
but
but
really
what
it's
meant
for
is
exactly
for
those
lives.
C
So
we
knew
about
lives
when
we
design
repo.
It's
just
that.
The
only
thing
we
did
for
them
is
an
idiot.
We
didn't
risk
write
this
draft
because
we
are
missing
the
6lowpan
side
of
it.
So
now,
if
we
have
it
and
what
really
happens
is
okay
I'm
getting
this
registration,
there
is
the
orbit
which
tells
me
I
have
to
work
for
him
and
I
will
send
it
down
with
a
EBIT
saying:
okay,
I
did
Sun.
C
C
We
can't-
and
we
don't
mandate
anything
on
the
leaf,
because
there's
supposed
to
be
a
ripple
anywhere
still,
it
won't
work
for
any
host.
There
must
be
a
minimal
support
which
is
not
rapport,
but
it
is
something
else.
So
we
mentioned
three
things
in
the
draft
in
your
C
version
of
four,
which
was
just
published.
The
first
thing
is
the
leaf:
must
support
RFC
sixty
seventy
five,
a
date
because
I
need
all
those
bits
and
all
this
information
right.
C
There
are
two
other
things
which
are
not
linked
to
the
control
plane,
which
are
linked
to
the
falling
plane.
We
need
to
be
able
to
dekap
sai
P
nightly
and
we
need
to
be
able
to
ignore
the
RPI,
because
if
you
look
at
all
the
ripple
and
use
of
ripple
info
draft,
you
will
find
that
at
some
point
some
packet
we
arrived
to
a
river
and
our
leaf,
which
still
have
some
ripple
information
in
them.
The
RPI,
the
IP
Nyepi
since
ipv6,
does
not
mandate
that
the
host
can
t
capture
a
type
in
IP.
C
So
this
more
text
about
the
air
flag,
so,
basically,
like
I,
said
the
air
flag
is
designed
to
be
used
set
by
a
ripple
unaware
leaf
which
supports
767
75.
So
if
you
see
this
bit
arm,
it
means
this
is
a
horse
that
does
everything
I
want.
I
mean
it's
it's
about
67
75
because
he
could
set
the
bit
alt
I
have
to
do
the
repo
or
the
NT
proxy
for
it.
C
If
there
is
no
rapport,
if
there
is
just
a
backbone
and
you're
talking
already
to
a
Badman
router,
since
this
is
the
same
NS
message
that
the
products
that
is
proxy
by
6lb
are
or
that
is
not,
does
it
just
works
the
same
so
you
can
deploy
this
exact
same
spec
was
just
a
Wi-Fi
type
of
ESS
information,
topology,
okay,
so
the
spec
has
asked:
must
there
I
mean
if,
if
the
whole
sets
the
air
bit,
we
must
do
the
repo
foil
if
the
host
does
not
set
there?
How
based?
C
So
I
took
care
of
showing
what
happens
when
the
6l
bureau
is
actually
just
a
registrar
and
it's
anywhere
in
the
universe
and
the
six
the
river
route
is
separate
from
them.
Then
again,
we
have
to
make
this
decision
whether
we
have
to
support
that,
but
I
made
it
the
problem
with.
It
is
probably
that
this
node,
that
must
be
one
heart
away
from
the
bbr,
is
the
river
route,
not
the
six
lvr.
C
So
if
we
decide
that
the
river
route
was
not
the
six
lbr,
then
three,
because
we
want
to
pull
the
6
lb
are
far
away
in
the
network
and
the
node,
which
is
doing
the
proxy
registration,
is
really
the
route.
So
in
all
my
previous
slides,
when
you
saw
the
6l
Bureau
doing
the
registration
on
behalf
well,
that
would
be
the
route
really
doing
it,
which
means
on
the
very
initial
flow.
It
does
not
happen
so
or
differently.
C
So
the
way
it
would
work,
then,
is
the
the
ripple
route
would
do
a
keepalive
data
exchange
to
the
6
lb,
r
and
the
health
of
the
node.
But
since
it
doesn't
have
the
rover
option,
it
would
be
more
constrained,
one
which,
which
does
not
have
a
proof
of
itself
in
it,
so
not
to
show
it's
a
good
idea,
but
that's
how
it
would
work
and.
C
Like
I
said,
the
node,
which
is
one
hot
away
from
the
six,
be
BR,
has
to
be
the
root.
So
it's
the
one
that
has
to
do
the
proxy
registration,
because
the
root
will
pass
the
packet
to
him,
so
he
can
then
check
the
packets
all
the
way
down.
So
you
see
that
basically,
in
order
to
get
the
rover
to
make
this
proxy
registration,
what
would
really
happen?
Is
you
send
a
keep
enough
either,
which
doesn't
have
the
rover
in
it?
C
But
the
6lv
are,
if
it's
happy
with
it
would
respond
with
the
rover
and
now
the
route
has
it.
You
can
do
the
products,
it's
kind
of
weird,
to
do
all
this.
We
could,
if
we
have
a
radius
guys,
but
I
would
like
the
group
to
actually
at
least
recommend
strongly
that
those
two
are
the
same
guy,
because
the
wrist
accel
border
router
is
him
anyway
right
so
meaningless,
yeah,
and
so
the
document
also
us
text
about
what
I
just
said.
You
know
how
you
map
the
Chi
ID
into
a
Dow.
C
Six
dishes
in
architecture
which
puts
everything
together
so
it
comes
from
us
as
the
bigger
view
architectures
yes,
six
dish
uses
will
use
that,
but
there
is
no
requirement
coming
from
six
tier
saying:
do
this
right,
it's
more!
It's
more
6lowpan
actually
which
which
told
us
a.
We
want
a
6lowpan
net6
island
not
to
have
to
do
anything
special
because
it's
ripple
and
because
they
want
about
right
over
in
general
yeah
at
least
we
have
an
example
now,
but
what
we
did
is
what
I
represent
agnostic
triple
it's.
C
C
I
mean
this
is
a
very,
very
simple
draft.
I
mean
do
all
the
work
was
done
at
six
load.
That's
why
yeah,
if
you
have
waited
what
I
presented
I
presented
the
work
at
all
in
two
minutes,
I
presented
the
work
at
six
slow
in
like
30
minutes,
because
that
was
where
all
the
works
up.
Well,
it's
not
finished,
so
she
want
to
work
on
that.
Come
to
seek
slow,
but
backbone.
Water,
in
particular,
is
not
in
worker
plus
call
yet.
C
But
what
we
have
to
do
here
is
so
simple
and
so
important
I
mean
know.
You
have
a
stupid
note
very
simple:
now
we
can
have
a
big
route
of
run
network
and
the
very
simple
nodes
roam
into
it
transplant.
You
know,
seamless,
mobility,
that's
the
dream
we
had
at
the
beginning,
so
we
could
not
do
it
with
repo.
As
initially
now
we
can
act,
it's
the
terminology
compliant.
C
I
C
D
D
I
see
right,
I
know,
I
can
make
this
work
there
you
go,
so
the
idea
is
basically
that
we're
going
to
set
up
possibly
between
a
source
and
a
target.
We're
gonna
set
up,
possibly
asymmetric
routes
by
using
similar
to
a
odv
messages,
a
route
request
and
around
reply,
and
although
there
can
be
asymmetric,
the
original
intention
was
that
there
to
actually
be
connectivity
and
the
reason
for
the
asymmetry
would
be
because
of
the
objective
function
or
the
metric
would
be
evaluated
differently
in
each
direction.
D
But,
let's
see
well
I
think
I
already
said
all
of
these
things.
So
some
of
the
comments
during
the
last
call
was
it.
It
was
inappropriate
to
have
constraints
on
the
expected
transmission
time
without
prior
knowledge
of
the
link
quality.
Well,
that's
kind
of
difficult
for
IOT
V
ripple
to
actually
solve
within
the
protocol
itself
and
the
way
that
device
is
in
the
network
know
about
their
link.
Quality
is
I
think
it
should
be
considered
out
of
scope,
but
we
could
add
some
neighborhood
messages
to
a
ODB
ripple
to
actually
make
these
estimations.
D
If
that
was
some
undesirable,
and
it
would
be
not
closely
related
to
the
existing
a
message
structure,
but
the
most
I
guess
you
would
say
extensive
comment
was
that
IOT
V
ripple
was
written
more
or
less
from
a
fresh
start,
given
the
Charter
on
him
to
do
peer-to-peer,
networking
and
didn't
allowing
for
asymmetry
as
well.
But
we
didn't
at
that
time
go
in
two
previous
experimental
RFC,
Peter,
Peter,
P
ripple,
and
so
some
of
the
features
in
the
I
think
it's
6997
removes.
D
Some
of
the
features
were
were
not
included
in
any
ot
V
riffle,
and
so
the
comment
was
made
that
that's
a
bad
idea.
We
should
actually
not
lose
features
so
that
people
wouldn't,
if
they
wanted
to
feature
they
would
have
to
implement
the
experimental
protocol
and
a
oddity
ripple.
So
the
attempt
was
made
to
actually
put
in
the
relevant
features
from
this
feat
of
P
ripple
and
I
have
a
lot
more
to
say
about
that
shortly.
D
C
I
may
so
there
are
already
two
sequence
counters
going
down
just
for
the
sake
of
just
I.
Think
it's
worth
this
world.
There
are
two
sequence
number
coming
down
from
the
root
well
and
I
mentioned
that
yesterday.
Somehow
one
of
them
is
when
you
build
a
new
geo
tag,
and
that
remember
this
is
all
form
by
diodes
right.
It's
not
from
bite
out
so
far,
so
so
the
sequence
number
you
have
order,
the
TSN
and
the
the
version
number.
C
If
you
have
a
new
version,
it
really
means
that
you're
painting
and
you
do,
but
the
route
could
also
issue
a
new
DTS
mm,
in
which
case
you
would
keep
the
do
tag
that
you
have
right
now,
your
parenting
or
anything
on
your
deck
formation,
but
you
would
be
making
sure
that
the
prefixes
are
available.
So
it's
really
a
matter
of
what
exactly
the
protocol
wants
to
do.
C
Anyways
it's
from
the
root
right,
because
that's
one
forming
this
sending
this
reply,
but
whether
you
want
to
use
an
existing,
do
diag
and
just
you
know
refresh
it
or
if
you
want
to
form
a
new
do
dikes,
but
the
first
rut
reply.
It's
obviously
using
versions,
I
mean
that's
right.
Yes,
now,
if
there
is
a
case
where
you
just
want
to
do
something
keep
alive,
then
you
could
say:
yes,
I
could
just
increment
the
DTS
N
and
keep
my
structure
other
ideas,
and
that's
that's
why
I
asked
you
when
we
talked?
C
D
Okay,
so
this
is
actually
sort
of
an
expansion
of
what
some
of
the
small
points
on
the
small
font
on
the
previous
slide,
but
so,
for
instance,
some.
We
don't
currently
have
a
way
to
exactly
tell
you
what
is
asymmetric.
The
link
is,
but
to
presume
that
the
devices
already
have
that
knowledge,
and
we
can
revisit
that
assumption.
D
We
ended
up
coming
up
against
the
submission
deadline
and
felt
that
we
had
plenty
of
new
description
in
the
document,
but
there
are
some
things
that
I'll
mention
later
that
you
would
think
would
indicate
that
it
wasn't
completely
updated
to
all
these
points
and
for
that
I
apologize,
but
I
fully
expect
to
get
a
more
consistent
version
of
the
specification
out.
I
was
going
to
try
to
do
it
this
way,
but
it's
been
pretty
intense.
C
D
C
That
implies
that,
even
if
the
links
are
not
symmetrical
at
least
they
there
is
a
bi-directional
connectivity
and
that's
really
a
limitation
of
p2p
report.
That's
also
a
limitation
of
ribbon
general.
You
need
at
least
at
least
a
degree
of
bi-directional
connectivity
and
I
looked
up
everything
I
could
about
what
causes
you
know:
asymmetry
in
those
radio
links
and
really
the
main
reason.
They're.
Probably
the
only
reason
is
poor
hardware,
where
the
exact
setting
on
the
radio
on
this
of
both
endpoints
are
not
exactly
aligned,
meaning
that
shows
you.
C
C
Double
nickel.
Tell
me
no
okay,
that's
what
I
found
in
my
research,
so
I'm
happy
to
learn
more,
but
where
I
was
getting
at.
Is
this
document
be
very
very
aware
of
the
big
big
differences
that
it's
not
using
Dao
Dao,
it's
using
2d
iOS,
both
directions,
the
and
well
machining
until
Dominic.
Fixes
me
is
that
a
symmetrical
length
was
not
such
a
big
deal.
That's
why
we
live
without
you
know,
but
you
need
directional
link.
I
mean
that's
a
big
big
difference,
so
you
say
yeah,
it
looks
good
I
mean
I
own
faces,
I!
C
Think
it's
wrong.
It's
wonderful,
I
think
it's
actually
very
very
important
because
there
are
such
a
thing
as
in
directional
links
and
ripple
could
never
cover
those
yeah.
So
this
is
the
first
draft
that
can
actually
do
something
for
repo
for
unidirectional
links.
And
yes,
it's
not
easy.
It's
not
hard
to
do
because
you're
doing
this
in
both
directions.
Yes,
you
must.
You
should
support
it
express
how
it
works,
because
it's
much
much
more
important
mathematical,
in
my
view,
right
well.
D
I
don't
so
when
I
joined
the
author
team
for
this
document
there
was
already
2d
iOS,
and
so
we
never
changed
that
and
it
never
came
up
in
the
working
group
until
this
week,
so
I'm,
okay,
with
trying
to
refashion
it,
especially
since
we
were
using
a
different
mode
of
operation,
we
can
redefine
what
we
need
to
make
the
asymmetric.
For
instance,
if
you
want
to
design
it
as
di
oh
and
then
read,
and
then
our
reply
would
be
a
Dao
that
we
can
try
to
do
something
with
that.
But.
I
C
I'm
just
correcting
what
was
said
because
saying
because
I
think
I
was
misunderstood,
I
was
saying:
yes,
keep
it
as
2d
iOS,
because
you
have
the
support
of
directional
links
and
I
think
you're.
The
first
work
ever
to
support
directional
links
in
repo.
If
you
went
back
to
Dow
like
Peter
vehicle
does,
then
you
would
lose
this
direction
of
thing
isometric
or,
in
my
view,
is
less
important,
which
we've
never
really
raised
up
in
network
side
touching,
but
directional
I
think
is
a
great
plus,
so
you
should
make
sure
it
works.
C
H
Thank
you.
That's
a
good
clarification.
Yes,
Dominic
Dominic
button,
just
as
I
came
to
Pascal's
comment.
Another
common
reason
for
asymmetry
in
radio
links
is
interference,
because
if
you
have
somebody
blasting
in
your
area
very
loudly,
the
new
desensitizer
receiver,
because
the
overall
level
rises
to
the
and
you
know
to
have
less
sensitivity,
and
so
your
reception
is
bad.
While
your
transmission
going
a
long
way,
is
at
a
lower
layer,
a
lower
level,
sorry
and
doesn't
desensitize
the
others
receiver.
So
that's
another
common
cause
for
asymmetry,.
D
D
One
of
the
major
features
from
RFC
6997
was
a
ability
to
have
a
request
for
this
for
multiple
targets,
and
that
was
we'll
do
that
when
we
already
read
in
toad
and
current
specification
provides
for
that,
but
there's
a
source
of
some
bit
of
inconsistency
which
I'd
like
to
go
into
here.
So
well,
there's
a
minor
plan.
D
We
removed
the
s
bit,
but
it's
now
in
the
round
request
option
and
the
option
link
was
added
to
enable
oh
I
forgot,
to
mention
an
important
thing
that
in
the
new
version
of
a
OTV
rippable,
we
also
allow
for
source
routing
in
addition
to
a
story
mode.
So
that
is
a
source
of
a
number
of
changes
in
the
document.
D
But
eventually,
according
to
discussions
that
we
had
it
just
didn't
make
sense
to
require
people
to
go
to
a
whole
nother
specification
if
we
could
provide
for
source
routing
solution
as
a
relatively
minor
modification
to
the
existing
specification
text,
so
that
that,
in
combination
with
the
requirement
for
multiple
targets
has
led
to
some
some
additional
stuff
in
the
specification.
So
other
features
from
Art
Loosed
keep
pushing
the
Run
button.
D
So
we
also
want
to
have
this
come
CLM
compression
field
and
talk
about
how
long
the
router
can
be
existing
in
the
dough
deck
and
then
also
allow
for
this
match
for
rank
field.
So
these
were
features
that
were
imported
from
6997
and
we're
still,
you
might
say,
verifying
that
it's
consistent
with
the
rest
of
the
specification
and
then
again
for
source
routing,
allow
for
an
address
vector
and
then
to
allow
for
multiple
targets.
We
had
this
iotv
ripple
target
option
which
is
modeled
on
the
one
in
our
FCC
997.
D
D
That
turned
out
to
be
somewhat
less
convenient
and
I'm,
not
sure
how
often
that
would
be
used.
So
my
current
suggestion
and
still
under
discussion,
it's
when
you
have
multiple
targets
and
the
target
option,
just
to
not
have
the
sequence
number
and
let's
see
how
that
works,
because
then
it
just
means
that
each
target
individually
has
to
respond
with
the
route
reply,
and
my
current
intuition
is,
but
that
will
be
okay
to
do
then.
D
Alright,
so,
and
then
we
had
discussion
about
how
it
is
that
the
instance
IDs
would
be
paired
an
original
document.
Our
earlier
document,
I
should
say
just
said.
Well,
we
have,
if
you
add
one
to
the
instance,
idea
the
route
request.
You
get
a
route
reply
and
then
occasionally
there
might
be
a
collision
or
even
rarely
there
might
be
a
collision
with
the
SSID
already
in
use
at
the
target.
D
So
now
what
this
a
proposal
is
to
have
another
field
and
that
you
usually
the
ID,
would
be
well
first
of
all,
there's
a
bit
here
that
says
whether
or
not
is
from
the
route
request,
the
rat
replied
and,
second
of
all,
if
the
route
replay
ID
our
instance,
ID
is
already
in
use
by
the
target,
then
there's
another
field
called
shift.
That
basically
says:
okay
well,
that
one
was
already
and
used,
but
you
can
get
the
instance
ID
by
shifting
it
to
another
number.
D
So
next
steps.
Well,
in
my
opinion,
it's
we
cannot
cite
RFC
6997
in
any
normative
fashion
in
this
document,
because
that's
an
experimental
document
and
so
any
things
that
we
import
from
RFC
6997.
We
basically
just
have
to
cut
and
paste
the
text
or
copy
and
paste
the
text.
So
that's
pretty
much
straightforward
to
do
that
and
then
that'll
be
part
of
the
new
submission
in
a
week
or
two
there.
So
we
want
to
make
sure
I
I'm,
not
actually
there's
no
suggestion
was
made.
D
That
is
possible
if
you
have
multiple
originating
nose,
in
other
words,
the
device
that
sends
out
the
route
request
that
they
could
accidentally
use
the
same
ripple
instanceid
for
their
own
purposes
and
then
what
happens
if
the
target
or
an
intermediate
node
has
to
deal
with
that.
So
that's
going
to
be
a
case
for
us
to
sit
down
and
actually
analyze
it
in
in
some
detail
and
though
that's
remains
to
be
done.
D
I'm
hearing
here
in
this
meeting
that
it'd
be
good
idea
for
us
to
have
this
black
hole,
Lane
detection
and
so
make
a
proposal
for
that
on
the
on
the
mailing
list
and
then
well.
Whatever
else
we
discuss
here
and
so
think,
I
think
I'm
done
and
they
delighted
to
take
questions.
If
anybody
has
there
will
be
short
questions.
E
So
you
mentioned
that
the
asymmetric
link,
detection
will
is
out
of
scope
or
the
unidirectional
link.
Detection
is
out
of
scope,
I
think
that
is,
and
you
also
try
to
mention
that
maybe
the
working
group
can
work
on
it,
but
yeah,
like
you,
mentioned
it's
better
to
keep
it
out
of
scope,
it's
difficult
to
get
a
universally
acceptable
solution
for
that
particular
problem
statement
in
this
particular
draft
I
feel
what
do
you
think
so?
What
I'm
trying
to
say
is
the
actual
detection
of
asymmetry
or
unidirectional
links.
You.
E
B
D
C
Get
email
I,
don't
even
think
we
need
to
do
anything,
because
what
we
already
doing
is
we're
building
another
do
dag
based
on
the
metrics
on
the
other
directions.
So
we,
the
Reggie
metrics
like
the
lq,
I
was
things
together-
are
what
you
see
right.
So
if
a
link
is
really
a
symmetrical
because
we
are
building
another
deal
DAC
with
with
another
dao,
I
mean
we,
we
might
not
be
going
through
the
those
links
at
all.
If
they
are
it's
not
like,
when
you're
sending
a
dao,
you
want
that
work.
C
D
That's
right
when
you
were
when
you're
spending
all
the
resource
to
do
flooding
and
both
directions.
That
gives
you
a
lot
of
power.
I
understand
that
fully,
but
we
just
need
to
be
careful
about
it,
because
this
is
a
good
place
to
put
bugs
in.
In
addition,
the
symmetric
bit
turns
the
reverse
of
di
o
into
a
unicast
and
make
sure
that
there
can
handle
that
case
as
well.
Okay,.
G
Fine,
but
I
should
like
is
this:
this
document
gets
concluded,
that's
not
of
the
from
this
the
work
to
be
concluded.
So
if
you
can
have
this
discussion
on
the
mailing
list,
whether
we
add
it
or
not
to
the
document-
and
you
can
conclude
within
two
weeks,
if
it's
a
lots
of
work
or
not,
then
she
have
the
court
document
concludes.
Okay.
I
would.
E
All
right
so
I'll
be
presenting
the
performance
report
for
the
route
invalidation
update.
So
briefly,
it's
the
current
ripple
specification.
This
applies
specifically
for
storing
mode
of
operation.
The
current
route
invalidation
that
happens
using
NP
Dow,
has
a
design.
We
had
made
a
design
choice
of
sending
and
pit
out
through
the
previously
through
the
previous
parent.
So
this
is
this
is
the
existing.
This
is
the
existing
mechanism
I'm
not
going
to
go
into
the
details
of
this
this
this
mechanism.
E
Here
it
was
already
explained
a
lot
of
times
before
so
there
is
a
new
message
that
has
come
into
existence:
the
dqo,
which
is
essentially
invalidating,
invalidating
the
route
downstream.
So
all
the
down
messages
till
date
except
for
our
projection,
were
going
upstream.
This
is
the
first
time,
and
apart
from
this
now,
projection
is
the
other
place
where
cows
are
now
sent
downstream,
okay,
so
the
implementation
report.
E
This
is
the
statistics
we
have,
so
we
used
eco
only
in
context
when
the
parent
switching
happens,
but
there
are
the
cases
where
NP
dows
are
sent
and
those
are,
for
example,
they're
out
lifetime
expiry,
and
we
don't
really.
We
haven't
really
changed
behavior
of
that,
because
it
does
still
make
sense
to
use
NP
Dow
in
that
particular
context.
So
essentially,
what
I'm
trying
to
say
is
currently,
in
the
current
in
the
example,
implementation
that
we
have.
We
have
deco
and
NP
now
operating
at
the
same
time
in
the
same
network.
E
But
having
said
that
pattern,
switching
is
the
most
barren.
Switching
is
what
happens
most
of
the
time
and
the
route
lifetime
expiry
also
is
there,
but
in
that
case
it's
just
a
now
so
in
case
it
has
to
update
the
route.
If
it
knows
that
there
is
a
the
previous
path,
the
existing
path
is
already
sufficient
or
it
is
already
there
to
handle
the
route
invalidation.
Then
it
will
be
made
use
of.
If
not,
then,
the
new
path
which,
whichever
it
has
been
shown,
whichever
it
has
chosen,
will
be
used
for
out
invalidation.
E
E
The
performance
report
was
sent
a
month
and
a
half
before
on
the
mailing
list.
I
hope
someone
has
looked
at
it.
So
basically
we
had
performance
taken
performance
on
two
scenarios.
One
is
the
metric
deterioration,
regular
parent
switching
and
in
second
case
we
had
deliberately
introduced
some
connectivity
impairment.
So
we
had
used
quantity
where
all
the
code
changes
were
done
in
the
kontiki.
The
framework
that
we
had
used
was
rightful,
so
essentially
it
is.
E
E
This
is
a
snapshot
of
the
topology
that
you
can
see
that
wait,
for
you
is
eventually
the
ripple
routing
protocol
makes
up,
and
it's
it's
part
of
the
Whitefield
framework
which
shows
up
this
technology.
So
this
is
an
interesting
point,
so
in
the
regular
case.
So
when
the
network
starts
up
the
network
boots
up,
all
the
nodes
are
started
up
together.
There
is
a
lot
of
parent
switching
that
is
happening.
So
during
that
time
there
is
a
lot
of
control
over
right
off
or
impede
out,
as
well
as
for
other
people
control
messaging.
E
So
we
have
found
that
these
particular
numbers
were
exceptionally
surprising.
For
us,
we
found
that
there
is
a
big
difference
in
the
overhead
with
decoy
with
compared
to
NPM,
and
this
is
not
what
we
had
thought
would
result
in.
It's
actually
a
big
improvement,
but
there
is
one
reason
why
this
has
happened
in
the
subsequent
slide
and
try
to
explain
our
primary
intention
was
to
work
on
reducing
the
steal
entries
count,
and
you
can
see
that
there
is
a
steep
gradient
for
Beco
in
case
of
decoy.
E
If
you
see
the
number
of
stale
entries
are
much
much
lesser
than
when
NP
dog
is
used,
it's
been
it's
it's
it's
natural
that
you
know.
Np
now
makes
use
of
a
path
which
is
suboptimal,
so
it's
going
to
be
bad
by
design.
It's
it's
it's
bad
by
design,
so
decoy
improves
on
it
and
we
can
see
that
the
number
of
the
amount
of
time
that
the
steal
interests
linger
is
much
much
lesser.
E
E
With
regards
to
NP
Dow
versus
D
Co,
the
parent,
switching
the
steel
entry
statistics
have
improved,
but
not
as
much
as
you
know,
as
you
can
see,
as
in
the
previous
context,
one
reason
is
the
steal,
so
what
we
had
done
is
the
one
of
the
problem
that
we
had
faced
in
this
context
is
that
the
subdue
deck
has
to
be
updated
when
a
when
a
parent
switch
is
the
path
in
of
storing
mode
of
operation,
the
subunit
has
to
be
updated.
The
subdued
a
routing
interest
out
to
be
updated.
That
doesn't
happen.
E
E
Okay,
the
important
observation
that
I
wanted
to
mention
here
is
why
the
difference
in
the
controller
over
L.
So
if
you
see
the
current
implementation
or
the
current
mechanism,
this
is
this
is
I,
don't
think
it
is
implementation,
specific
thing,
but
you
can
have
a
better
implementation,
but
currently
the
diorite
and
Quantic
implementation.
Does
it
this
way?
E
What
it
does
is
on
parent
switching,
it
invalidates
the
current
route,
so
it
sends
it
out
on
the
previous
path
and
it
schedule
a
regular
down
to
updates
to
update
the
current
new
path
or
the
new
on
the
new
path,
of
course.
So
what
happens
is
since
NP
Dow
is
sent
first,
it
reaches
all
the
way
to
the
border
router,
and
there
is
a
lot
of
control
over.
C
E
Okay,
so
what
I
want
to
what
I
try
to
reason
is:
is
it
possible
to
handle
the
implementation
in
forever?
It
is
possible,
but
it
requires
some
most
state
information.
So
what
implementations
are
doing
is
what
implementations
are
doing
is
title
time
trying
to
take
a
easy
way
out.
It's
not
necessarily
bad.
It
is
saving
some
state
information
for
them.
So
essentially,
what
an
implementation
could
have
done
is
they
could
have
scheduled
a
scheduled
NP
down.
I
did
a
timer
here
and
then
it
should
do
it.
E
Even
if
you
sew
this,
this
has
to
be
sent
after
a
delay.
There
is
a
delay
now
involved
here,
so
if
this
has
to
be
sent
after
a
time
well,
but
this
is
sent
as
soon
as
you
do,
a
parent
switching
unless
and
until
you
have
a
timer
and
that
time
were
somehow
expires
after
this.
This
is
a
randomized
time.
So
again,
there
are
a
lot
of
ifs
and
ifs
and
buts
there.
It's
actually
not
very
trivial,
to
implement
this
in
efficient
way,
but
anyways
coming
back
to
the
point.
E
So
this
is
what
the
route
goes,
that
we
found
out.
Why
the
difference
is
such
large.
We
do.
We
were
not
expecting
such
a
big
difference
in
the
control
over
so
but
this
is
the
primary
reason.
Maybe
what
I'm
trying
to
say
is
it's
an
implementation,
specific
issue?
Also
here,
so
maybe
the
difference.
It's
not
good
for
us
to
get
excited
that
you
know
you'll,
be
saving
a
lot
of
route
invalidation
traffic,
so
it's
an
improvement.
So
how
does
D
Co
help
here?
So
what
I'm
trying
to
say
is
in
case
of
D
Co.
E
The
implementation
is
much
more
at
ease
because
you
now
don't
have
to
initiate
or
out
invalidation
from
the
target
loaded,
so
the
common
ancestor
take
care.
Zaki
cares
of
it.
It's
a
synchronized
operation,
one
more
bigger
problem
with
the
previous
point
was
that
the
NP
da
was
sent.
It
clears
out
the
route
in
the
previous
path,
and
then
there
is
it's
an
a
synchronous
operation.
So
there
is
a
scoffer
out
downtime
there
yeah,
but.
C
There
is
a
neat,
though,
if,
if
you're
spending
your
your
dau
like
forming
this
new
Dao
but
I
want
to
have
multiple
problems
right,
so
I'm
actually
sending
the
same
path:
sequence
to
two
parents,
yet
one
is
slower
than
the
other
and
what
you
may
be
doing
and
beware
because
that's
tricky
you
may
be
sending
those
to
dowels
and
one
is
faster,
faster
and
now
you
trigger
additio
too
quickly
to
quickly
do
you
don't
have
to
worry?
But
if
you
do
it
too
quickly,
we
will
cross
yeah
and
hopefully
the
sequence
number
game.
C
Yes,
sorry
out,
but
I
mean
you,
you
clean
up
a
path
which
is
still
valid
till
it's
reformed.
All
I'm
saying
is
you
know
having
this
this
broken
path
for
a
little
while
doesn't
hurt
so
much
okay,
so
don't
be
too
quick
just
it
was
too
quick
to
send
the
down
the
nope
after
that
were
traditional.
No
doubt
up,
don't
be
too
quick
in
sending
the
new
part
out
down,
because
you
might
similar
issues
actually.
E
This
this
there
is
a
section
in
the
raft
which
also
talks
about
this.
There
is
a
possibility
that
the
same
target
node
might
send
the
same
dow
from
two
different
paths
and
how
that
be
handled.
Essentially,
the
path
sequence
takes
care
of
it.
So
yeah
it's
through
this
light.
What
I
wanted
to
bring
out
is
that
for
the
target
node
at
least
or
the
overall
operation
of
the
DCO
versus
the
regular
dow
is
much
more
synchronized
and.
C
Then
there's
there's
another
thing
which
troubles
me
forever,
because
maybe
we
will
set
our
life
here,
but
this
is
a
traditional
thing
in
in
inks
tied
to
to
kind
of
well,
some
people
have
tried
in
the
past
to
clean
up.
You
know
the
bad
state
that's
left
over
for
some
reason.
In
due
to
mobility,
for
instance,
you'll
get
to
may
I
say
its
network
and
you've
got
a
guy
moving
from
here
to
here
and
you've
got
this
old
path,
which
stays
there
for
a
while,
and
people
want
to
go
and
clean
it
up.
C
There
I
forgot
the
name
for
printing
this
and
I
just
know
from
here
site,
because
I
don't
have
experience
in
that,
but
it's
it
always
proved
to
be
quite
complex,
at
least
at
least
in
their
networks.
Our
network
is
be
different,
with
no
sales
etc.
But
I
would
like
to
make
some
research
about
why
printing
in
some
other
types
of
networks,
doing
this
exact
operation
has
proved
to
be
difficult.
Part
of
what
I
heard
is
that
how
that
intermixed?
C
With
with
newer,
you
know,
information
things,
change,
back,
etc,
and
then
you
end
up
printing,
something
which
is
actually
good.
I,
think
your
explanation,
the
sequence,
number
etc
I
think
we're
safe
I'm
saying
is.
There
is
a
very,
very
bad
history
of
trying
to
do
that
in
other
routing
protocols
and
I.
Don't
have
all
this
history.
J
F
F
J
C
J
C
E
Just
one
point
to
add
here,
so
you
mentioned
that
the
target
node
actually
owns
its
own
routing
entry.
So
what
one
there's
one
section
in
the
draft
which
talks
about
it?
So
what
it
does
is
it
says
so,
the
regular
doubt
that
is
sent
here.
It
sends
a
plaque
saying:
okay,
you
invalidate
so
if
you
see
this
flag
set
only
then
the
common
ancestor
will
go
ahead
and
do
an
invalidation.
So
it's
sort
of
still
is
controlling
in
some
way
so
yeah
it
was
not
needed,
but
we
still
kept
that
option
of
having
that
yeah.
E
C
Anyway,
I
solve
a
whole
set,
I
mean
it's,
not
somebody
making
the
decision
you're,
not
there
I
mean
we
have
that
as
well,
but
it's
much
later,
so
all
things
are
fixed
before
it's
the
parent
times
out,
but
still
it's
still
message
from
from
the
real
guy
which
flows
up
and
back
it's
it's
still
the
message
from
the
guy,
so
it
still
all
this
state,
even
even
on
the
way
back.
Okay,
I,
think
I'll,
say:
okay,
all.
E
Right,
so
that's
all
I
think
we
have
not
made
any
semantic
changes
for
a
long
time.
In
fact,
since
the
version
1,
we
have
made
a
lot
of
synthetic
changes.
We
have
changed
the
message,
types
and
all
message
codes.
Earlier.
We
were
trying
to
use
a
flag
in
an
NP
des
message,
but
then
we
changed
to
a
new
message
type.
All
together,
it's
called
eco
destination,
cleanup
object,
and
that
was
the
major
change.
E
G
E
G
E
C
Is
good,
I
always
live
up
an
X
like
that
me
right.
Two
things
Cisco
as
I
appear
on
this
I
think
my
way
as
well,
yeah
right,
so
just
how
you
to
know
second
thing,
how
invited
me
so
I
said
yes,
yeah
yeah
under
which
actually
should
which
describes
all
this
actually
a
bit
more.
So
so,
in
the
light
of
that,
we
decided
to
work
together,
who's.
C
So
good,
yes,
the
del
projection,
good
thing:
we
we
are
out
of
time
because
I
don't
have
much
to
say,
apart
from
the
fact
that
some
people
actually
implementing
yet
I'm
aware
of
two
implementations
on
the
way,
I
actually
invited
the
earthers
of
those
two
implementations
to
participate
to
a
draft,
because
I
was
well
James
and
I,
also
in
implementer.
We
were
kind
of
alone
on
this
and
we
wanted
you
know
a
new
I
and
noted
more
review
and
practical
implementation,
feedback,
etc.
C
So
those
two
implementations
that
I've
heard
of
is
one
is
coming
from
away
and
that's
led
by
our
usual
suspect
here,
and
so
apart
from
how
there
is
also
mad
machinima
for
might
run,
and
so
it
was
there
yesterday.
Somehow
we
are
missing
him
today,
but
I
also
invited
him
because
of
his
experience
and
all
that
happens
in
the
real
world.
So
what's
what's
very
good
with
this
news
is
that
you
know
we
have
real
implementations
even
two
of
them,
which
is
one
of
those
thing,
want
to
check
so
very
happy
with
that.
C
C
We
have
one
problem
that
we'll
have
to
discuss
at
all
at
some
point
is
the
size
of
the
mode
of
operation.
The
mode
of
operation
is
3
bits
and
we've
got
0.
That's
the
collation
tree.
We've
got
1
2
for
storing
non
storing
and
we
get
3
for
storing
with
multicast
support,
and
then
we've
get
four
for
the
RFC
69
997,
right,
p2p
and
I.
Guess
a
OTV.
You
are
using
a
new
mode
of
operation
as
well,
so
we
will
be
short
on
them
very
soon
and
so
for
the
Bell
projection.
C
Is
it
the
right
way
of
signaling
things?
Should
we
I
mean?
How
do
we
tell
the
network
that
we
are
doing
the
owl
projection
over
whatever
mode?
It's
doing?
Is
it
through
bits
in
the
mode
of
operation?
Again
we
end
up
consuming
them
all.
What's
the
way
it
could
be
a
new
option
or
something
you
know,
and
basically
the
ripple
Network
works
a
certain
way,
storing
non
storing
and
then
we
have
projected
rods
which
can
be
storing
a
non
storing
both
of
them
in
any
of
the
main
modes.
C
D
This
non
Charlie
Perkins
sub
from
Huawei
this
that
came
up
in
the
discussion
for
ATV
ripple
as
well,
should
a
odv
rippled
be
using
a
different
mode
of
up
mode
of
operation,
or
should
we
just
rely
on
using
different
message,
numbers
that
message
option
numbers
further,
our
our
operation-
and
you
know
if
it's
really
getting
to
the
point
where
we
should
conserve
mode
of
operation,
bits
the
net.
We
should
discuss
exactly
how
the
information
is
organized.
C
Now
we
are
very
sure
that
we
have
very
few
mobs
right.
We
awaited
them
Toto
and
we
have
consumed
hard
consume
for
them
with
the
main
spec
one
of
them
is
consumed
by
an
experiment.
Also,
we
could
get
it
back
some
point
because
now
we've
got
the
non
experimental
version
of
it,
but
we
should
you
know,
listen
chance.
We
do
not
steal
from
the
experimental,
but
we
should
recycle
it
someday.
C
So
that
gives
us
at
least
five
with
iotv
then
find
for
Givi
to
consume
one
Bob,
because
it's
so
important,
but
not
for
projection,
arrived
I
hope
we
could
find
something
else.
But
that's
that's
that's
on
the
table
right
we've!
Yet
we've
got
this
stupid.
Quite
a
good
non-technical
problem
on
our
to
signal
the
projection.
Well,.
G
C
C
Text
just
to
explain
know
if
the
message
looks
like
that,
it's
a
storing
if
it
looks
like
that,
it's
non
storing.
You
know
what
you
put
in
a
target,
what
you
which,
how
many
addresses
you
put
there
I,
don't
like
that,
it's
indirect,
that's
free
when
we
added
a
non-story
piece.
So
then
again
it's
all
about
signalling.
What's
the
best
way
to
see
how
we're
doing
this
and
then
I
wanted
to
share
this
this
remaining
question
and
I
guess
we
know
the
answers.
C
How
is
the
topology
known
to
the
root
in
non
storing
mode
the
root
as
the
dyrdek?
It
doesn't
have
more
than
that
doesn't
have
sibling
links
constants,
but
it
tells
did
you
die,
so
we
can
already
work
like
that
in
story
mode.
There
is
not
a
protocol
information
but
I.
Think
the
big
point
is
whether
the
root
is
something
or
not.
It's
not
this
documents
problem.
If
we
want
to
talk
about
topology
and
it's
chair,
exchange
things
with
the
root,
let
you
know
what's
going
on.
That's
interesting!
That's
this
kind
of
work,
but
it's
it's!
C
It's
it's
not
this
back!
This
back
is
just
about
that.
So
it
has
to
be
done
mainly,
but
it's
to
be
done
elsewhere.
That's
pretty
much!
What
the
answer
would
be.
Please
react
if
you
don't
agree
same
for
not
capabilities,
we
could.
We
could
extend
it
out,
for
instance,
to
play
some
memory
and
there
are
other
trusts
which
already
extending
the
Dow
to
say
things
about
the
notes.
So
that's
interesting
that
you
so
far
this
draft
is
it
this
draft
to
specify.
Maybe
not.
C
E
Rahul
Jadhav,
so,
regarding
the
topology
known
to
the
route,
you
mentioned
that
in
case
of
non
storing
mode
only
it
is.
It
is
known.
We
noted
here
that
you
know
the
well.
It
is
possible
to
know
even
in
storing
mode
of
operation,
because
the
transit
information
contains
the
parent
address
it
uses,
which
also
can
be
populated
in
case
of
storing
mode
of
operation.
Yes,
and
it
can
still
be
no
that's
right.
I
mean
with
a
current.
C
C
Well,
you
you
would
have
to
you
know,
since
it's
a
story
mode
thing,
I
would
not
do
it
that
way,
I
would
basically
send
it
down
to
the
root
and
non
storing
down
to
the
root
in
the
same
network,
because
if
you
do
it
to
the
parity,
you
double
the
memory
on
each
heart
because
they
have
to
remember
this
to
set
it
up.
Remember
you
cannot
just
flush
it
out
that
you
got
from
bottoms,
I
need
to
top
and
forget
about
it.
You
need
to
keep
a
state
didn't
be
able
to
reveal
it.
C
We
had
this
discussion
understand
constraint
out
of
travel
with
that.
That's
why
they
don't
do
story
mode.
If
you
do
story
mode,
do
it?
Well,
if
you
do
it
well,
you
must
be
able
to
reconstruct
the
dowel
on
behalf
of
the
guy,
because
you
are
a
parent
and
you
want
to
tell
the
new
parent
about
it
or
anything
else.
C
C
C
That's
pretty
much
it
so.
We've
we've
got
these
questions
like
I
said
this
is
probably
a
different
document.
The
topology
should
be
inferred
by
some
other
work.
The
note
capabilities
are
probably
another
work
as
well.
It's
interesting
right
how
many
rats
can
I
put
in
that
node?
Can
I
push
in
that
node?
We
are
working
and
putting
new
matrix.
C
You
know,
remember
yesterday's
discussion
so
as
we
build
information
so
that
the
geoduck
root
knows
to
do
load
of
his
on
tier
dag,
there
could
be
information
like
that
coming
from
the
nodes,
but
I
think
any
information
that
we
need
sure
could
be
in
those
documents
because
that
they
will
really
be
talking
about
metrics
about
the
node
so
that
the
root
knows
better
eyes
network.
So
it's
not
this
document
either
that's
what
I'm
getting
at
mixed
mode.
Is
it
too
complex
on
that?
C
Where
we
are,
we
have
two
implementers
now
co-author
of
the
document,
so
I
hope
we
get
more
feedback
on
that
more
saturation,
I
raise
the
prime
because
that's
the
biggest
issue
and
then
compression
the
answer
has
always
been
will
do
the
compression
someday
for
even
many
people.
We
never
did
it
so
we'll
have
to
figure
when
they.
C
If
we
want
to
compress
the
repo
signaling
I,
think
we
should
do
it,
and
then
there
was
a
question
by
how
about
loop,
adroit,
ons
and
actually
part
of
the
things
that
how
accept
it
to
take
over
in
this
document
is
to
to
have
a
new
section
about
local
variants.
Make
sure
you
know,
because
now
we
are
building
multiple
topologies.
We
have
the
main
report
apology
and
then
we
have
all
those
overlays
which
are
the
projected
route.
There
are
real
overlays
over
the
main
road
topology,
and
could
you
be
building
loops?
C
C
C
G
C
Mean
even
if
when
so
so,
remove
the
original
thing,
which
was
storing
mode.
Well
because
are
you
talking
about
the
overlay?
Are
you
talking
about
the
underlay?
The
underlay
is
the
normal
ripple,
the
big
thing,
and
then
the
overlay
is
the
projected
rod.
The
overlay
is
either
storing
a
non-story
you're
talking
about
that
right
and
you're,
suggesting
that
only
the
non-story
makes
sense.
You
can.
G
C
G
Like
like
the
right
I,
do
there,
the
question
was:
do
you
actually
want
everything
in
the
route
I
mean
it's
not
enough
information,
all
the
other
roads
to
handle
the
problem,
because
that
was
one
of
your
questions.
Why
do
how
could
I
get
all
the
information
to
the
route?
Thinking
that?
Maybe
you
don't
need
it
all
in
the
hood,
but
it
could
be
at
a
lower
level
things
good.
You
could
have
enough
information.
C
Yeah
I
mean
so
far.
You
know
there
are
only
two
types
of
node
in
the
ripple
Network.
There
is
the
root
and
there
is
the
others
and
the
others
are
all
the
same.
So
they
can
intermix
not
some
fashion.
Now,
if
you
either,
you
have
a
third
type
of
node,
which
is
like
something
in
between
which
could
have
some
of
those
capabilities
and
or
you
you,
you
will
have
every
node
capable
of
doing
much
si
and
that's
a
good
question,
but
I
mean
pushing
you
know
we
always
wanted
to
have
repo
minimal.
C
Almost
you
know
we
can
discuss
what
minimal
is
I
did
this
morning
with
Charlie,
but
we
wanted
some
denotes
to
12
a
minimum
thing.
So
if
you
give
them
all
this
capability,
yes
it's
technically
doable,
but
it
kind
of
means
that
every
node
in
my
network
can
do
it.
Yeah
or
you
know,
I
enter
this
cell
type
of
thing,
which
and
I
have
to
deploy
the
right
place
and
yeah.
Maybe,
but
you
see
you
see
what
that
carries.
G
C
C
Where
there
was
another
thing
right,
I
don't
remember,
but
I
there
are
two
things
that
you
have
to
look
at.
You
don't
try
that
so
we're
good
I
also
would
like
to
maybe
look
at
the
NEPA
appendix
which
is
already
a
result
from
you.
Our
implementation
and
early
testing
like
something
which
would
tell
us
what
we're
missing
in
report,
because
we
had
nothing
at
the
time
which
would
tell
us
more
details
and
how
to
use
the
applicability
statements
so
so
some
beginning
of
a
peak
ability
statement.
When
do
we
see
that
we'll
be
using
this?
C
If
somebody
already
tried
it,
you
know
what
kind
of
finding
he
got.
So
yes,
because
we
will
have
implementers
now
I
would
like
some
written
on
where
and
why
and
how,
and
that
would
be
dressed
appendix
text.
I
guess.
But
we
saw
with
report
that
having
no
information
at
all
like
that,
actually
made
the
implementation
much
harder,
and
people
now
complain
that
always
hard
I
mean
in
my
mind,
is
so
simple.
So
but
three
because
we
did
not
explain
enough.
C
I
H
Gis
modifications
the
draft
hasn't
change
for
maybe
a
year.
I
just
want
to
give
you
people
were
asking
and
we
had
an
exchange
on
the
mailing
since
yesterday,
I
think
or
this
morning,
people
we
got
comments.
People
were
wondering
what
happened
with
this
book,
so
yeah,
and
just
let
you
know
originally
I-
think
I
proposed
this
modification
first
at
the
Maastricht
meeting.
H
In
the
working
group,
you
remember
probably
JP
Vassar,
and
he
told
me
yeah,
you
know
right
now,
ripple
is
trying
to
establish
itself
as
a
as
a
routing
protocol
in
this
area
and
which
was
a
yeah
I.
Keep
it
stable
and
don't
introduce
changes.
Maybe
don't
say
you
can
make
it
better,
because
people
would
understand
it's
not
good
enough.
H
Interested
in
that
okay
sounds
good,
so
we
have
a
chance
right
so
so
this
has
always
been
in
the
back
of
my
mind
that
you
know
we
had
this
stuff
on
a
shelf
and
I
figured
if
nobody
is
interested
boy.
Buzzer
I
have
to
do
that
on
my
spare
time
and
hide
away
from
my
manager,
but
now
we
get
a
sense
from
a
few
of
you
that
this
is
really
needed.
If
this
is
needed,
I'm
more
than
happy
to
revive
it.
B
C
Here
so
yes,
the
world
leader
on
the
mailing
list,
because
a
lot
of
the
discussions
we
had
was
kind
of
ATF
type
corridor
discussions,
but
we
we
actually
discussed
these
drafts
of
all
time
since
right
and
at
least
you
and
I
and
I,
remember
I,
don't
have
all
the
details.
I
have
to
rebuild
them,
because
we
did
not
capture
that
with
corridor
thing,
but
there
are
cases
right.
C
It's
a
very
good
improvement
and
cases
where
it
will
actually
work
the
other
way-
and
you
know,
depending
on
how
the
podcast,
how
many
people
are
on
your
broadcast
domain
cetera.
So
it's
a
mechanism
and
it's
useful
when
it's
useful
and
not
when
it's
not
so
so
the
primaries.
It's
not
like.
It's
always
useful.
It's
at
some
points.
It
can
become
detrimental
detriment
or
we
discuss
that.
C
If
you
remember
this
discussion
and,
and
so
really
I
think
I
mean
the
bits
are
racing
for
everything
very
simple:
it
could
slice
through
this
working
rock,
but
what's
really
needed
and
missing
is
there
is
why
it
works
series
what
it
does
use
it
in
that
case,
but
if
you're
in
that
case,
don't
use
a
torch
any
differently.
That's
a
three
to
tricky
piece
of
this
draft.
I.
Think
and
I
agree
with
history
at
some
point
at
the
beginning,
people
want
it
very,
very
stable.
C
Let's
learn
from
what
we
have,
which
we
did
over
the
years
and
now
since
we
learned
it's
the
time
to
do
version
to
clarify
the
text
which
is
not
clarified,
which
is
what
house
drive
yesterday.
Does
it
set
red
straw?
So
at
some
point
we
can
turn
it
to
an
internet
standard.
So
this
would
be
needed
for
the
internet
solid,
but
with
all
the
right
words,
which
explains
why
and
when
you
do
it
and
when
you
don't
do
it
right.
H
So
the
intention
is
to
introduce
the
flags
and
bits
so
that
you
can
tweak
the
way
they
are
yes
and
the
I/o
that
you
get
in
response
triggered
now.
So
yes,
we're
opening
ways
by
which
you
can
make
the
network
really
bad,
and
hopefully
people
will
learn
how
to
use
that
I
mean
how
to
use
it.
The
wise
way
such
that
it
proves
it
makes
the
network
better.
So.
C
H
H
Draft
by
a
Mook
wall,
which
was
a
different
situation
and
they
wanted
things
different,
but
we
figured
you
know
it's
just
a
matter
of
specifying
what
we
want
in
the
GIS,
using
a
few
flags
and
constraints,
and
so
now
I
don't
have
the
deployment
situation
to
test
all
those
cases,
so
I
will
not
be
able
to
justify
why
it
makes
it
better,
and
this
in
that
situation,
I
need
your
use.
Cases
I
mean.
C
We
just
really
have
to
rebuild
this
information,
though
I
got
a
bit
out
of
my
hat,
but
I
really
remember
that,
depending
on
intensity
and
what
and
how
things
are
built.
This
could
actually
cause
a
lot
more
traffic
than
doing
the
normal
way.
It
can
really
be
useful,
like
you're
not
coming
in
next
in
network,
etc.
It
can
be
a
really
determent
or
maybe
it
forming
a
very
dense
network.
C
I,
don't
remove
all
the
use
cases
in
other
cases,
but
really
we
are
to
look
at
what
we
discussed
that
day
and
make
sure
that
that
we
use
it
when
it's
good
and
not
way,
it's
not
good
and
we
provide
information
to
know
when
it's
good.
It's
not
good,
because
the
basic
node
will
be
out
coded
setting
those
bits
just
for
the
pleasure
of
it
and
depending
on
where
you
place
it,
and
when
it
does
that
that
could
be
good
or
bad
and
I.
Don't
think
that's
right,
I
think
we
we
need
those
bits.
C
C
H
H
The
use
cases
that
we
had
in
mind
I'm
no
longer
in
the
situation
where
I
can
actually
run
it.
I
may
become
again
in
the
future
and
I
know
there
were
people
which
had
different
use
cases.
Different
topologies,
which
I'm
not
too
familiar
with
so
I
can
design
the
mechanics
that
has
a
bits
that
you
can
turn
on
and
off,
but
telling
which
is
better
best
in
which
situation
this
is
kind
of
out
of
my
reach
right
now,
I
can
tell
for
the
one
situation
that
I
know.
H
C
H
So
yeah
we
promised
with
chunk,
will
allocate
some
time
on
that
same
thing.
Putting
actually
does
not
work
on
this
anymore,
I
understand
and
all
those
other
people
who
have
left
to
field
as
well
in
manual
and
Mukul
and
all
those
guys.
So
we
promised
to
allocate
some
time
to
revive
this,
have
a
discussion
going
on
the
mailing
list
and
come
with
some
draft
for
next
meeting.
I'm,
not
sure
what
will
be
in
the
draft,
but
at
least
we'll
raise
awareness
again
did.
H
C
And,
as
you
know,
if
we
are
not
too
clear,
I
mean
we
can
always
make
it
in
two
steps.
Right
of
this
experiment.
Also
people
implemented
return,
and
now
we
could
ship
it
very
very
quickly
and
then,
as
a
second
stage,
just
like
we
did
for
the
p2p,
then
do
the
real
thing
and
it
could
still
be
you,
but
that
would
be
first
in
experimental
and
then
start
truck
once
Reno
says
we
carefully
a
stylet
track
because
we
put
it
back
in
the
Internet
at
this
very
moment.
C
We
are
not
close
to
an
interest
or
not
with
ripples,
so
you
have
a
window
of
time
to
learn
and
and
get
more
feedback.
So
so
the
answer
that
way,
if
we
want
you
know
not
to
have
too
many
things
in
power.
This
one
is
a
great
experimental
RFC
publishing
tomorrow
morning,
because
it's
been
the
favorite
in
trial
grounds,
if
we
just
said,
send
bits
and
use
them
and
tell
us
if
it
works
and
what
it
does
for.
You
then
publish
this
incremental
tomorrow
morning,
but.