►
From YouTube: ROLL WG Interim Meeting, 2021-01-29
Description
ROLL WG Interim Meeting, 2021-01-29
A
A
So
please
be
aware
that
this
meeting
is
a
line
alonso
aligned
with
the
not
well
guidelines.
We
don't
want
to
read
them,
but
you
have
to
be
familiar
so
the
meeting
the
enterprise
they
go
for
the
md,
please
say
add
your
name:
if
it's
not
there
and
the
slides
are
the
intermediate
materials
side
of
the
idea,
so
the
agenda
for
today
we
are
going
to
have
dow
topics.
A
So
the
milestones
where
we
have
a
achieve
recently
with
milestones,
you
said
ripple
info
efficient,
mpdio
turn
on
the
word
in
the
rfc.
Editor
queue
so
like
you
can
see
so
well.
Use
of
ripple
info
now
as
well
is
in
the
rfc
editor
queue
so
very
nice
that
we
have
this
work
done
so
now
we
can
continue
with
next
work,
so
we
will
have
from
these
documents.
We
are
going
to
discuss
about
this
dow
projection
and
please
we.
A
We
got
this
one
review
from
enrollment
priority,
but
will
be
very
nice
if
we
can
get
more
reviews.
We
are
missing
that
to
progress
with
the
document.
Okay,
so
I'm
from
capabilities
once
I
say
work
in
progress
at
mopex
and
as
well.
Well,
I
would
be
ripple
is
a
with
a
discussion
with
alvaro,
yeah,
basically
and
then
okay
inactive
internet
draft.
They
are
in
the
standby.
A
A
A
Okay,
these
are
the
open
tickets
from
github
page,
basically
as
well.
Well,
these
are
tickets
are
open
still,
so,
particularly
if
it's
when
they
are
closed,
just
close
when
they
are,
if
the
version
16
address
them
and
they
as
well,
we
put
the
open
tickets
from
the
itf
web
page,
so
we
mean,
if
you
want
to
continue,
please
about
the.
B
High
hi
hi,
this
is
lee.
This
is
lee
and
I
will
appreciate
the
chapter
for
me.
I
think
he
is
get
sick,
okay,
okay,
how
do
I
everyone?
This
is
lee
from
cisco.
Today
I
will
introduce
the
draft
about
a
new,
the
deck
magic
used
for
the
deck
selection.
B
B
B
B
B
So
in
this
page
we
propose
a
new
type
to
carry
the
to
that
size.
As
in
the
table,
we
use
two
bytes
to
indicate
the
to
that
size.
Meanwhile,
we
use
the
original
flags
defined
in
65,
50
51..
B
B
C
Yes,
I
do
have
questions.
This
is
dominic
speaking
as
an
individual.
You
decided
to
use
two
bytes
for
the
dark
sides.
Does
that
mean
you
don't
expect
to
need
more
than
that?
No
do
that
with
more
than
65
000
notes,.
B
Ask
yeah
two
best
can
has
60
hundred,
so
is
it
maybe
it's
enough
or
maybe
you
can
consider
to
use
more
bytes,
maybe
three
or
four.
C
D
E
Yes,
in
in
ripple,
we
at
least
use
a
variable
measurement
for
a
time
we
we
provide
the
time
unit
and
then
in
the
diy,
and
then
we
we
can
every
time
it's
expressed
in
that
time
unit.
So
that
allows
to
keep
a
very
compressed
size
expression
of
time
and
still
express
any
any
duration.
I
mean
if
a
dodge
is
living
on
a
very
low
pace,
then
this
late,
this
duration
is
very
long.
The
unit
is
very
long.
E
On
the
other
hand,
if
we
have,
if
you
are
living
in
a
fast
pace,
then
that
duration
can
be
much
smaller,
but
we
avoid
having
to
place
many
many
bytes
for
time
in
every
packet.
By
having
the
units
separate
from
from
the
count,
you
should
look
for
instances
in
the
dio
in
the
dio,
the
the
path,
the
lifetime
it's
expressed
in
that
in
that
unit.
E
D
I
I
I
so
the
heuristic
about
when
to
send
the
dio
if
the
routing
table
size
on
the
route
has
increased.
Basically,
I
believe
this
new
metric
this
metric
will
be
updated,
but
I'm
just
trying
to
wonder
here
whether
there
are
other
ways
to
handle
the
situation
currently
and
how
how
much
insufficient
are
they,
for
example,
whether
rank
is
is
how
much
of
rank
can
help
in
this
particular
situation.
D
G
Actually
in
why
some
formula
two
for
the
notes,
this
satisfied,
which
pen
to
to
join
it,
consider
both
the
rank
and
the
pen
size
and
also
a
different
weight.
G
So
actually
that
can
be
yeah
that
that
can
be
the
way
it
can
be
decided
through
experiments
in
the
real
deployment.
E
H
E
You
know
one
way
of
deciding
what
the
priority
will
be
remember
just
for
those
who
did
not
read
it
recently.
The
environment
priority
is
an
information
given
by
the
root
disseminated
in
dio,
and
that
gives
basically
the
floor
of
the
priority
that
a
node
can
express
in
a
beacon,
so
a
node
as
a
joint
proxy
will
send
beacons
and
those
becomes,
it
will
say,
a
priority
and
that
priority,
if
it's
a
very
small
number,
it
means
I
really
want
to
be
a
joint
proxy.
If
it's
a
large
number
like
200,
I
don't.
E
I
don't
want
to
be
a
joint
proxy.
So
if,
if
globally,
the
geoduck
is
kind
of
full,
the
root
can
express
this
floor
as
a
high
value,
meaning
that
no
node
in
the
network
can
be
expressing
a
priority
lower
than
this
floor.
So
no,
no,
no
node
in
the
geoduck
will
be
attractive
and,
and
one
way
of
setting
this
priority
is
to
take
kind
of
you
know
the
exponent,
as
we
said,
or
the
I
order
number
of
of
geodex
size.
E
So
there
is
possibly
a
relationship
between
this
priority
and
and
the
geodex
size
which
is
kind
of
how
what
you
were
asking
for.
If
I
understand
well
now,
it
doesn't
seem
to
me
that
we
must
we
can
hide
the
geoduck
side
within
the
priority,
because
it
might
be
that
the
priority
is
built
out
of
other
reasons.
So
you
cannot
really
guess
from
the
priority
what
the
geoduck
size
is.
E
But
my
question
is
really
do.
Do
we
want
to
have
a
new
option?
I
mean
to
have
to
to
put
a
container
on
everything,
or
could
we
could
we
look
at
this
enrollment
priority
draft
which
just
at
the
moment
passes
a
single
byte
and
that's
a
full
draft
which
has
to
go
through
isg
blah
blah
blah
for
one
byte,
which
is
this
priority?
E
Now
I
was
saying:
could
we
merge
the
document,
the
work
that
lee
and
women
are
doing
with
the
the
environment?
Priority
work
make
it
a
bit
more
rich
or
something
so
they're
just
a
suggestion.
I
don't
know
what
the
authors
think.
D
Yeah,
that's
an
interesting
point,
but
here
my
assumption
is
that
the
dudac
size
is
a
static
value
which
is
disseminated
by
the
root
and
all
the
intermediate
nodes
are
simply
passing
it
on.
E
E
E
F
E
H
E
C
Can
you
can
you
provide
an
example
of
we
understand
the
concept,
but
do
you
have
a
use
case
that
you
would
use
that
and
what
would
be
the
threshold.
A
G
Yeah
less
news
advertisement
from
both
tank
to
tag
and
it
need
to
decide
which
one
to
join
and
that's
since
in
the
poll
all
the
at
the
root.
So
today's
the
root
can
manage
limited
number
of
nodes.
So
it's
better
that
the
two.
If
there
are
two
or
multiple
dotages
and
they
can
be,
the
number
of
nodes-
can
be
balanced
between
them.
C
C
Okay,
but
I
mean
you're
from
cisco:
do
you
see
a
business
case
that
needs
that?
Can
you
give
us
some
details.
D
I
I
would
just
like
to
you
know
remind
one
more
thing
here
that
there
was
a
proposal,
some
time
back,
which
basically
gave
out
the
new
deck
size
from
so
every
parent
node
who
is
disseminating
the
dio,
will
fill
in
its
own
the
dodge
route
rooted
at
its
position
to
the
downstream
peers.
There
was
one
proposal,
I
don't
remember
the
name,
but
it
was
from
some
uk
university.
D
This
proposal
is,
is
having
a
state
static
value
directly
sent
from
the
root
to
all
the
nodes.
That
proposal
was
additionally
that
every
node
will
disseminate
the
the
size
of
the
dodge
rotate
at
its
own
point.
C
D
E
And
we
discussed
it
at
length
and
we
found
that
as
a
matter
of
balancing,
it
was
not
necessarily
a
good
information
because
it
does
not
relate
to
the
to
the
traffic
which
is
going
through
this
node.
I
don't
think
it
was
the
size
of
the
geoduck
beneath
it.
I
think
it
was
the
number
of
children
and
that's
really
what
why
there
was
a
problem.
It
was
number
of
children.
D
I
I
think
that
was
yet
another
proposal
pasco,
and
I
believe
that
was
from.
F
George's
yeah,
we
worked
on
that.
So
maybe
you
have
two
resources
here.
One
is
the
memory
available
on
your
border
router
for
route
and
that
is
related
to
the
dodge
size,
and
then
you
have
the
capacity
for
forwarding
traffic
and
that's
related
to
the
the
traffic
per
node
per
side,
etc.
So
it's
it's
two
different
aspects
right,
true,
okay,.
D
No
so
it
is,
it
is
indeed
two
different
aspects,
but
if
the
use
cases
you
know
and
the
use
cases
are
also
different,
but
if
we
are,
you
know
okay,
anyways
yeah.
I
think,
let's
keep
that
discussion
separate.
I
don't
want
anyone.
C
E
Different
anyway,
my
my
proposal
stands
that
we
should
consider
not
doing
five
rfcs
with
one
two
bytes
each
mostly
if
they
are
highly
related,
like
the
environment,
priority
and
the
size
of
the
geodec,
because
the
use
case
that
I'm
aware
of
for
the
size
of
the
geoduck
is
to
help
what
we
have
in
this
slide,
which
is
which
dodge
to
to
join
and
then
in
the
life
of
the
geodex.
Should
I
move
to
the
next
geoduck,
because
it's
less
loaded
than
the
one
I'm
in.
D
E
E
Reporting
the
same
thing,
each
one
they
are
equal
now
it
could
be
that
it's
just
a
right,
so
it
could
be
that
I
would
also
suggest
if
we
did
do
this
merge
that
we
also
have
a
metric
of
load
in
terms
of
bandwidth
at
the
root,
because
the
usually
in
a
network
like
this
most
of
the
traffic
is
from
or
to
the
root.
If
the
network
is
a
radio
network,
then
the
cons,
the
most
constrained
resource
will
be
the
radio
bandwidth
next
to
the
route.
E
So
if
we
can
also
have
an
information
about
that,
that
certainly
helps
deciding
if
you
move
to
the
pen
on
the
left
or
the
pad
on
the
right,
but
it's
it's
for
the
case
of
wyson
that
on
which
women
is
working,
understand
that
the
geodex
size
is
enough,
because
there
are
electrical
matters
for
today,
that's
what
they
have
so
short
term.
That's
enough
longer
term
and
with
you
we
also
need
a
kind
of
a
load
metric.
E
I
So
we
did
a
bunch
of
this
stuff
in
sextish,
with
enhanced
beacon
to
determine
the
to
provide
the
information
outside
of
the
encrypted
and
the
ripple
etx,
and
my
understanding
is
that
this
is
a
metric
that
would
could
be
used
to
drive
that
in
a
six-dish
environment
which
ysun
is
not
so
I'm
wondering.
Does
this
work
imply
that
this
node
is
able
to
heal,
hear
dios
from
both
networks.
I
G
Yeah,
that's
the
same
ysan.
So
in
the
first
step
when
when
the
new
node
has
no
information
about
any
any
diagonal
and
the
the
enhanced
speaker,
the
first
speaking
is
not
encrypted.
E
It's
just,
I
guess
that
we
don't
they
don't
use
the
60ish
soon-to-be
rfc
document
right
here,
not
speaking,
but
they
want
something
equivalent
kind
of
as
it
goes.
We
are
thinking
about
triple
which
is
used
in
common
between
the
two.
So
we
want
to
provide
information
for
anybody.
I
Okay,
so
so,
but
is
this
about
the
the
pan
advertise
beacon,
which
I
don't
think
it
is
it's
about
the
information
that's
carried
down
through
the
dodag
to
the?
If,
if
I
you
can't,
I
can't
point
with
my
mouse,
but
you
know
so.
There
was
there's
that
red
arrow
points
to
one
node
in
the
left,
which
is
closest
to
which
is
presumably
what
it's
hearing
and
maybe
to
the
right.
G
E
E
I
To
raise
the
thing
there
there's
a
document
that
rule
and
I
wrote
called
role,
enrollment
priority,
which
is
not
a
metric,
but
it's
a
dio
option
and
where
the
root
says.
Essentially
this
a
similar
thing.
But
it's
abstracted
as
a
a
priority.
I
E
Michael,
I
actually
said
that
for
some
people
it
might
be
that
the
priority
is
exactly
the
most
significant
bite
of
those
two
bytes
for
some
environments.
They
can
do
that,
but
some
environments
will
have
the
priority
more
complex
than
just
the
geoduck
size
or
independent
of
the
geodex
size.
So
just
that,
why
don't
we
put
both
information
in
your
draft
actually
and
why
don't
you
guys
merge
the
work
in
this
draft
and
and
the
environment?
E
E
I
E
I
E
The
priority
kind
of
summarizes
the
door.
So
if
we
had
only
one
thing,
we
could
hide
everything
inside
the
priority.
Now
it
might
be
the
the
the
edge
routers.
The
joint
proxies
wish
to
have
more
details
about,
what's
going
on,
to
make
a
more
intelligent.
I
Well,
the
the
issue
is,
I
think,
and
this
was
brought
up
by
alvaro
and-
and
I
think
others
is
when
we,
when
we
had
talked
about
the
six
dish
enhanced
beacon
was.
The
concern
was
that
unless
we
were
detailed,
the
algorithm
and
made
everyone
run
the
same
algorithm,
that
the
tuning
of
the
parameters
would
not
result
in
a
deterministic
result.
I
It
would
not
be
deterministic
and
that's
why
I
kind
of
prefer
a
single
number,
keep
the
math
or
the
formula
within
the
network
and
and
have
the
the
joining
nodes
be
more
most
more
dumb
and
not
try
to
be
smart.
That's
smart
with
the
risk
of
stampeding
elephants,
but
that
that
I
think,
was
the
suggestion.
But
I'm
I
don't
have
really
the
routing
math
clue
to
say.
I
So
you're
saying
that
that
calculation
it
needs
to
be
within
the
the
pan,
the
lln
needs
to
be
standardized
and
route
1
and
route
2
may
actually
need
to
communicate
to
to
discuss
their
relative
importance.
I
don't
disagree
with
that
concept.
E
E
H
A
A
E
E
Well,
it's
certainly
big
and
when
I
have
to
explain
I
remember
last
last
intro
and
and
after
that
michael's
reaction,
the
mailing
list
I
have
to
reduce
the
slightly
tetra
tells
me
that
this
there's
too
much
complexity
and
we
need
to
to
to
figure
out
what
we
want
to
keep
or
divide
and
conquer,
like
we
did
in
the
main
repo,
the
main
repo.
We
did
story
non-story
as
a
way
to
simplify
the
completely
implementations
which
could
be
doing
either
mode,
but
not
necessarily
all
same
fashion
without
projection.
E
I
started
with
something
very
simple
which
now
I
call
the
first
profile
profile,
one
and
we'll
see
that
we
go
all
the
way
to
six
profiles
now,
with
with
everything
we've
been
building
over
the
years.
E
My
recent
efforts
was
to
simplify
the
explanation
and
provide
examples
and
divide
into
profiles,
because
that's
what
the
mailing
list
has
been
asking
me
so
so
this
is
what
I
will
be
presenting
to
you
today
and
quote
unquote
the
good
news
for
michael
or
bad
news
or
whatever.
Is
that,
whatever
you
understood
at
the
last
instrument?
E
E
The
route
encapsulates
to
place
the
srh,
so
the
idea
is
is
if
you
are
within
that
that
main
geodag
and
you
want
to
to
to
have
a
track
inside
that
main
teardog.
The
enclave
submission
node,
should
be
just
like
the
root
above
it
should.
It
should
be
the
root
of
this
small
diode,
which
was
not
what
I
had
before.
E
E
E
E
But
because
now
the
ingress
is,
the
encapsulator
is
the
one
which
has
the
the
srh.
You
have
two
real
caps,
because
you
change
the
source
of
the
packet
and
and
when
you
change
the
source
of
the
packet,
that's
the
big
reason
that
six
men,
while
they
forced
you
to
rear
cap
as
opposed
to
just
insert
the
srh.
They
say
they
said.
If
you
insert
an
srh-
and
you
don't
tell
who
you
are,
then,
if
there
is
a
problem,
we
can't
come
back
to
you
and
say
you
they.
E
You
did
a
mistake,
so
six
man
was
very
clear.
If
you,
if
you
add
an
srh,
the
source
of
the
packet
must
be
the
guy
you
did
at
the
srh
and
now
since
the
srh
is
added
by
the
ingress
which
is
root
and
which
becomes
the
source
of
the
packet,
then
that
comes
with
well,
the
the
addition
of
the
srh
complies
to
six
men,
but
that
means
encapsulation.
E
E
Actually,
it
can
be
left
to
right.
So
if
you
have
a
leaf
talking
to
another
leaf
and
you
build
a
deer
dag
between
those
two,
the
ingress
leaf
will
be
just
the
root
of
that
small
deodag
and
the
egress
leaf
of
that
diode.
The
destination
will
decapsulate,
but
you
can
see
it
as
michael
said,
recursive
and
and
all
the
things
we
have
explained
in
user
freeport
info
like
if
the
packet
is
coming
from
outside
the
geoduck,
then
the
route
has
to
encapsulate
or
if
it's
coming
from
an
external
route.
E
You
know
something
coming
from
the
outside
of
the
geoduck
but
below
the
this
xlr,
which
which
exposes
the
external
rod
has
to
to
encapsulate
as
well.
You
know
to
to
well
all
this
applies
exactly
the
same.
Now
it's
just
a
geoduck
within
a
geodec.
E
So
let's
I
will
have
slides.
I
mean,
don't
try
to
understand
everything.
From
my
words,
I
spent
a
lot
of
time
on
the
slides
that
you,
you
will
see
now
and
I
I
hope
that
you'll
find
your
answers.
So
first
thing
is:
I
push
15
and
16.
15
says
the
root.
Is
that
the
root
of
the
diode
is:
is
the
ingress
not
the
egressed
anymore,
so
you
can
have
a
geodet
which
which,
which
is
entered
on
the
left
by
a
node
or
the
root.
The
knight
can
have
multiple
extents.
E
So
so
it's
like
a
like
a
tree.
If
you
like
out
of
this
root
and
16,
is
where
I
cut
out
of
the
comments
that
I
had
on
15.
E
I
created
all
those
profiles,
and
I
also
placed
a
lot
of
examples
like
at
least
four
six
examples
which
which
are
which
you
will
see
not
really
in
that
form.
But
in
the
slide
the
examples
I
also
posted
on
the
building
list,
but
they
didn't
get
much
feedback,
but
they
still
stand.
If
you
dig
in
the
mailing
list,
you
will
find
examples
and
the
examples
say
what
goes
in
the
in
the
dow
what
goes
in
the
main
object.
What
goes
in
the
target?
E
What
goes
in
in
the
vio
and
what
sorry
just
thought
somebody
was
speaking
so
so
I
I
just
one
small
clue
which
I
wrote
in
this
this
this
parchment
on
the
right
is,
you
should
see
the
vio
as
a
transit
information
option
for
if
you're
used
to
non-storing
mode,
a
transit
information
option
is
how
you
do
you
tell
who
your
parent
is
in
non-storing
mode
in
the
transit
information
option.
I
say
here
is
the
parent
for
the
address
which
is
in
the
target.
E
E
If
you
see
it
like
that,
it
helps
a
lot.
So
you
want
to
reach
the
target
and
the
viu
is
the
multi-hub
sequence
to
reach
those
targets.
Just
like
a
tio
is
a
one-half
sequence.
The
vio
is
a
multi-hub
sequence
and,
and
most
of
it
stays
as
you
know
it
so
another
thing
to
simplify.
I
said
it
at
last
in
trim.
E
If
you
are
using
storing
mode,
then
the
topology
of
the
geoduck
is
not
known
by
the
root,
and
then
you
you
would
need
something
to
tell
the
root
about
it.
So
we
we
can
build
stuff
just
along
the
main
geodag
just
because
it's
non-storing
and
you
will
see
that
profile,
one
and
profile
two
are
there:
they
are
just.
They
don't
need
any
sibling
information.
They
are
just
building
along
the
main
geodes.
Now,
if
you
want
to
build
transversal
routes
left
to
right
east
west,
then
you
will
need
the
sibling
information
option.
E
So
a
track
now
is
a
non-storing
node
deal
dag,
which
works
exactly
like
the
main
non-storing
mode.
Diode
the
route
is
a
track
ingress.
You
you,
you
have
to
encapsulate
the
packets
coming
from
the
outside.
Just
like
user
free
power
info
tells
you
to
do.
E
You
have
to
place
an
srh
and
you
you
will
see
that
there
is
this
case
of
an
implicit
when
it's
fully
storing,
but
normally
for
real
signal
track.
It's
signaled
with
non-stirring
pidao.
So
what
we
call
an
srvio,
basically
source
related
vio
that
will
signal
a
track
and.
E
A
E
E
Oh
okay,
anyway,
please
please
send
back
the
slides.
E
Okay,
so
what
I
was
saying
is
we
make
now
we
make
the
difference
between
what
we
call
a
track
and
what
we
call
a
segment.
A
track
is
when
you
have
to
encapsulate
the
packet
that
ingress,
for
instance,
to
place
an
srh,
but
also
to
place
an
api
and
a
segment
is
the
the
the
hops
between
two
loose
source
rod
hubs.
E
E
So
we
said
hey.
We
can
reduce
that
if
we
place
a
story
mode
state
between,
for
instance,
if
you
have
a
b
c
d,
e,
f
g,
h,
block
to
z,
you
could
say
between
h
and
m,
I
can
have
a
story
mode
and
so
in
in
that
case,
I
can
have
a
source
rather
which
says,
h
and
m
right
away
and
skips
all
the
intermediate
steps,
because
there
is
restoring
mode
there.
So
so
that's
the
initial
case
for
what
projection
before
I
went
into
non-storing
and
everything
I
still
don't
see
a
slide.
A
E
Okay,
otherwise
I
shall
mind,
I
mean
no
worries.
Okay,
yeah,
can
you
go
back
to
slide.
A
Okay,
I
have
to
take
the
permissions
to
you
first,
but
it
seems
like
I
got
stuck.
It's
not
responding.
My
browser.
A
E
E
What
I
also
changed
recently
is
how
we
signal
the
the
vio
in
particular.
What
I
wanted
is
to
avoid
that
an
intermediate
node
like
this
ingress
has
to
compute
the
compressed
form
of
a
source
rotator
in
with
with
eight
one
three
eight
remember,
I
tried
three
eight,
the
srh
is
always
compressed
by
the
root.
E
It's
it's
not
a
very
complex
operation,
but
still
it's
some
work,
and
so
I
I
wanted
to
avoid.
You
know
the
bugs
and
everything
which
may
come
by
having
a
node
compute
a
compressed
watching
either,
and
so
I
said-
and
we
discussed
that
on
the
mailing
list-
that
we
wanted,
that
the
vio
is
expressed
in
the
compressed
form
and
in
particular,
if
it's
an
srvio,
it's
expressed
exactly
as
will
be
in
the
packet.
E
You
also
have
to
see,
know
the
ingress
somewhere,
so
you
have
to
also
have
ip90,
but
for
all
the
the
hops
you
just
copy
verbatim
the
option.
So
there
was
there
was
an
effort
there.
E
So
that
was
one
thing,
but
when
then,
if
you
look
at
8138
for
the
rpi,
you
realize
that
the
way
it
compresses
it
was
not
a
good
fit
for
what
we
need
here,
because
for
this
draft
we
we
don't
need
the
flags
that
are
present
in
the
normal
fpi,
which
say
basically
the
geodex
structure
has
changed
or
needs
to.
You
need
to
be
able
to
rebuild
that
all
those
flags
go
away.
E
So,
basically,
what
I
did
is
I
I
added
some
formats
to
rfc8138
for
exactly
the
rpi
optimized
for
what
we
need
in
this
draft,
which
is
basically
make
it
elective
or
not
elective.
So
you
can
you
can
ignore
it.
If
you
don't
understand
it,
if
you
have
source
route
anyway
and
put
just
the
ripple
instance
id,
because
that's
the
only
thing
we
care
for
so
it's
a
new
six
layer
edge
type,
seven,
which
would
be
the
one
we
use
for
tracks.
E
So
so,
basically,
that's
why
this
draft
now
updates
or
extends
actually
six
five,
five,
three
and
eight
one
three
as
well,
so
encapsulation
rules
have
changed
since
last
time.
Now
the
the
tracking
grass
is
the
root.
Is
the
source,
the
the
source
of
the
outer
header,
and
then
you
encapsulate
if
you
need
to
place
either
an
api
or
to
place
an
srh
just
like
before
the
repair
instance.
Id
must
be
the
track
id
so
we'll
see
in
the
profile
2
that
use
non-storing
mode
between
to
lose
hops.
E
So
you
need
to
re-encapsulate,
and
then
you
need
a
track
id
for
it.
So
we
have
two
types
of
yios
at
some
point,
there's
another
name,
but
that
I
got
people
confused
with
that.
So
now,
both
storing
and
non-storing.
They
are
both
called
vios,
but
there
is
the
source,
routing,
vio
and
the
state
for
vio,
so
that
that
removed
that
simplified
the
draft
by
removing
extra
names.
E
E
It
goes
from
the
ingress
to
digress
if
the
ingress
of
the
egress
or
lose
hops,
for
instance,
from
the
root
in
a
source
right
header
from
the
root
and
the
rpi
in
the
packet
is
already
zero
and
the
sfvio
was
signaled
also
for
for
reaper
instance
id
0,
then
everything
is
correct.
We
don't
need
to
do
any
encapsulation
to
do
anything
we
just
forward
along
the
sfvio
which
allows
you
to
have
a
loose
rotting
header
in
the
from
the
root.
That's
basically
the
trick
that
we
wanted
to
do
initially
with
this
draft.
E
What
we
have
not
done
yet
is
matching
rules,
flow
information,
option
which
I
think
how
asked
at
some
point,
which
is
okay,
I'm
getting
this
packet
at
the
tracking
rest
and
the
tracking
grass
is
a
possible
way
to
get
there.
Should
I
put
the
packet
on
the
track?
Should
I
give
it?
You
know
let
the
package
follow
the
default
route.
E
E
Now
I
also
changed
not
much.
It
was
before
actually,
but
I'm
still
giving
them
now
the
the
construction
of
the
pidao.
E
E
So
all
those
guys
you
want
to
reach
across
this
tunnel.
Basically
you
place
them
in
the
target
option
so
so
that
the
one
dial
can
apply
to
multiple
targets.
It
basically
says
you
have
the
route.
There
is
a
sequence
of
nodes
to
reach
all
those
targets,
but
that's
what
the
dao
says
now
to
simplify.
We
said
last
time
there
can
only
be
one
vio
in
the
case
of
srv
io.
At
some
point
we
had
multiple
of
them
but
say
for
some
reason:
you
can
only
install
two
of
them
and
not
the
third.
E
E
Basically,
we
said
a
single
one
for
the
stateful
vio,
because
it
has
to
go
from
the
egress
to
the
ingress,
hop
by
hop
by
hop
two
things:
one,
you
have
to
make
it
sequential
you
can
you
it
must
be
strict,
always
because
you
have
to
to
tell
to
each
each
node,
which
is
this
parent
to
which
is
going
to
give
this
backhand
and
the
the
other
thing
is.
E
You
cannot
cannot
have
two
of
them
because
decrease
passes
to
the
previous
app
which
passes
to
the
previous
app,
which
passes
the
previous
up,
which
says
ack.
Now,
if
you
have
two
sfvios,
that
means
that
would
be
two
ingress
and
and
which
one
does
the
ack?
How
does
he
know
what
happened
on
the
other
side?
That
means
two
acts
now,
so
it
becomes
kind
of
complex.
So
we
said
single
vio
in
a
dial,
so
the
dow
has
one
vio
multiple
targets.
If
you
like
zero
targets,
if
you
like
as
well
so
pascal.
I
Yep,
I
I
think
that
one
thing
maybe
to
emphasize
here
or
explain.
I
think
that
in
the
original
projected
deo
that
we
were
essentially
adding
source
road
headers
to
storing
mode
right.
I
E
I
Right
so,
but
now
we
have
both
storing
mode
and
non-storing
modes.
For
these.
D
E
Some
profiles
would
just
implement
one,
but
that
was
my
answer
to
this,
because
some
profiles,
I
don't
believe
we
can
live
with
just
one
of
the
two,
but
I
believe
one
particular
use
case
can
thus
just
like
storing
my
non-stoic
mode,
those
profiles
that
will
present,
but
but
it's
the
beginning
of
this
discussion
once
you've
seen
the
profiles,
you
can
come
back
to
me
and
say
maybe
provide
three.
We
don't
want
it.
E
E
Almost
always.
I
will
show
you
the
case
where
the
track
is
implicit
and
then
again,
that's
exactly
a
case
where
you
will
tell
me
well,
I
don't
want
that
one,
this
extra
complexity,
let's
drop
it.
I
don't
know
you
will
tell
me
so
this
challenge
is
not
from
this
time.
It's
it's
a
bit
some
time
ago,
but
I
just
want
to
remind
you
a
few
hfs
ago
when
we
still
still
met.
E
Now
there
are
weird
conditions
where
you
can
compress
to
two
bytes
and
then
at
some
point
you
need
to
compress
to
four
bytes
and
back
to
two
bytes
or
stay
at
four
bytes.
I
don't
know
in
that
case
you
need
to
have
multiple
srh,
six
layer,
headers
and
and
the
addresses
which
are
all
the
same
size
for
one
another.
E
So
it
might
be
that
you
end
up
having
exactly
this
exact
situation,
in
which
case
only
all
you
see
here
may
repeat:
it
is
probably
rare,
but
it
may
repeat
the
only
case
where
the
spec
must
that
you
you
do
it
to
the
most
optimized
way
is
if
the
geoduck
is
operated
with
8138
on
and
it's
non-storing
mode,
and
oh,
it's
it's
well,
it's
an
answering
mode
pedal,
meaning
that
it's
it
will
be
inserted
by
the
ingress.
So
we
want
the
format
that
you
see
here
to
be
exactly
what
the
ingress
will
insert.
E
E
E
It
was
how
to
scope
for
a
long
time.
Then
we
said,
let's
have
non-storing
mode
so
that
at
least
the
geoduck
is
known,
and
then
we
said,
let's
have
sibling
information
options,
so
we
can
also
know
who's
left
and
who's
right
by
the
way.
As
written
now
we
have.
I
just
found
that
we
have
an
acronym
conflict,
because
6550
has
a
solid
statement,
information
option
in
the
disk
which
has
the
same
acronym
as
sibling
information
option.
So
I
will
have
to
change
that
and
any
any
idea
is
welcome.
E
But
you
see,
I
cannot
have
sio
it's
already
taken
now
this
this
information
about
the
siblings
or
the
neighbors
is
is
needed
for
profiles
more
than
three.
So
so
that's
why,
for
instance,
if
you,
if
you
just
want
to
compress
the
srh
of
the
mendiodag,
you
never
want
to
to
do
horizontal,
east-west
routes,
then
you
just
implement
profile
one
or
two
above
but
develop
nevermore.
E
Now
we
provide
a
way
to
to
say
who
is
east
and
west,
but
if
you
have
1
000
neighbors
right
now
the
current
draft
doesn't
say
which
ones
you
pick
and
actually
it's
a
complex
problem.
I
mean
we
have
some
thoughts.
We
even
have
ipr
about
doing
that
selection.
E
I
would
be
ready
to
share
that
ipr
and
share
designers,
but
I
don't
think
it's
the
same
draft
finding
out.
You
know
which
notes
from
a
dominating
set
or
something
could
be
complex
and
and
should
be
a
separate
discussion.
E
E
But
but
yes,
it's
a
complex
discussion
and
there
we
go
for
the
profiles.
So
now,
if
you
want
to
take
some
spraying,
some
sugar,
a
break
coffee
now
is
a
good
time.
E
E
E
So
you
see
already
in
ripple,
you
have
the
choice
between
passing
tunneling,
the
packet
to
e
or
tunneling
the
packet
to
f
and
those
two
possibilities.
You
will
find
that
you
always
have
them
for
profile.
One
and
two.
You
have
the
same
case
of
you
end
the
tunnel
at
the
final
station
or
you
end
the
tunnel.
I
did
not
just
before
it,
which
encapsulates
and
finds
that
this
stock,
the
packet
is
for
his
neighbor
and
and
in
this
picture
you
have
an
example
of
both
actually
pedal
ward
and
pitaro
too.
E
So
what
we
want
to
do
in
profile
1
is
provide
a
storing
mode
state
4
routes
in
the
intermediate
nodes,
so
that
the
srh
from
the
main
geodag
can
just
say,
can
be
shorter.
Instead
of
saying
you
know
in
the
compressed
form,
basically
it
would
say
abcdef
normally,
if
it's
not
an
external
rod,
it's
if
you
turn
all
the
way
to
f
it
would,
it
would
say,
abcdef.
E
E
Is
it
very
clear
if
somebody
doesn't
understand
what
case
I'm
so
I'm
talking
about?
I
can
repeat,
I
want
to
be
very
clear
cases.
We
we
make
the
multi-hop
storage
smaller
by
removing
some
hops,
because
we
do
a
storing
mode
pdow
to
replace
that,
and
so.
In
this
example,
I
have
two
pid
outs
of
one
which
allows
to
go
from
a
to
c
and
one
which
goes
from
c
to
e.
That's
why
the
srh
says
acef.
E
You
realize
that
to
do
the
exact
same
thing
for
abc,
I
could
have
sent
p
o
2
to
c,
which
would
have
passed
it
to
b,
which
would
pass
it
to
I,
but
I
didn't
do
that
in
this
example.
It
would
have
worked.
I've
been
exact
same
thing
as
the
other
one.
What
I
did
is
I
passed
it
to
b.
Only
what
that
means
is.
E
I
don't
use
the
dow
to
tell
b
that
how
to
go
to
c,
basically,
because
I
know
that
c
is
already
b's
neighbor,
so
I
know
that
even
if
I
don't
put
a
c
in
the
dow
b
will
still
be
able
to
reach
c.
E
So
what
I
do
is
I
stand
in
this
example:
control
a
to
the
one
on
the
right
to
b,
one
on
p
o
two.
I
start
a
stateful
tao,
which
says
just
a
and
b,
so
it
means
I
send
it
to
b,
which
follows
the
path
all
the
way
to
ace.
That's
a
single
hop
in
this
case
and
a
will
answer
and
a
will
be.
The
ingress
quote:
unquote
of
this
segment.
E
I
Pascal
and
so
for
the
case
for
c
d
e,
when
the
packet
arrives
at
c,
it
has
a
source
header
that
says
d
e
on
it.
That
gets
it
added
by
c
no.
B
E
That
exactly
what
happens
is
with
the
way
it
spell
right
now
in
the
document
and
we'll
see
there
are
little
things
we
could
we
decided
to
do,
but
the
weight
spelled
right
now
when
this
packet
reaches
e
right
e
says:
okay,
here
is
a
destination
e.
Well
very
interesting!
I
already
know
myself
but
this,
but
this
creates
a
route
towards
e.
I
don't
care
for
it,
but
that's
what
it
does
and
the
route
is
cde.
E
Okay.
So,
as
e,
I
will
not
install
anything
because
I'm
not
interested
in
this
target
and
I
will
pass
the
same
packet,
the
packet.
I
will
pass
it
to
to
d.
So
d
gets
the
packet.
What
does
it
say?
It
says?
Oh
with
this
route
I
can
reach
e
and
that's
via
e,
well
very
interesting.
I
already
edited
my
neighbor
table.
I
won't
do
much
out
of
that,
but
I
will
pass
it
to
c
c.
Gets
the
packet
and
say
oh
to
reach
target
e.
My
next
hop
is
d.
Fine.
E
I
will
install
route
destination
d
decision
e,
I'm
sorry
next
stop
d,
and
since
I'm
the
first
in
this
list,
I
will
do
the
the
I'm
quoting
what
the
ingress-
and
I
will.
I
will
do-
the
dow
act
to
the
root.
So
the
end
result
of
all
this
and-
and
there
is
like
multiple
tables
in
the
draft,
which
shows
all
this-
I
spent
a
lot
of
time
in
the
draft.
Giving
all
these
tables
here
is.
The
rib
here
is
what
you
do
blah
blah
blah.
E
E
E
I
just
passed
the
packet
to
b,
and
you
realize
that
in
that
I
still
have,
I
always
have
the
egress
as
target
it's
one
thing
that
could
be
implicit
at
the
moment.
The
way
it's
written,
it's
not,
but
we
could
decide
that
I
don't
have
to
write
b
as
target
because,
implicitly
it's
oh,
it's
it's
there.
So
why
do
I
need
to
write
it
here
at
the
moment
for
stage
four
vio?
I
write
it,
but
it's
a
decision
we
can
make
on
the
list.
It
could
be
implicit
actually
when
you
have
a
long
list.
E
E
Well,
it's
you
make
your
own
decision
based
on
the
memory
you
have,
but
I
think
the
draft
should
be
more
precise
on
whether,
as
of
today,
we
just
install
rods
to
the
target
or
if,
for
instance,
if
I'm,
if
there
is
cdefg
blah
and
then
d,
should
I
install
a
route
to
h,
blah
via
e
or
not
if
they
are
in
the
sfv
I
o.
I
could
because
I
know
how
to
reach
them,
but
it's
more
memory
so
so.
This
is
why
today,
I
just
installed
rods
to
e.
E
E
Can
I
reach
c?
Yes
c,
my
neighbor
is
my
neighbor
okay,
so
it
looks
fine.
I
can
reach
both
targets,
I'm
a
valid
ingress
egress,
I'm
sorry,
I'm
a
valley
degrees
for
those
targets,
no
prob.
So
let
me
pass
it
to
a
and
a
will
see.
Oh
to
reach,
b
and
c
I
can
use
b
so
b
is
fine,
it's
already
my
neighbor.
So
let
me
install
a
rod
to
see
via
b,
so
now
a
has
a
route
to
c,
via
b
and
c
as
a
route
to
e
via
d.
Is
it
clear.
E
E
E
E
2
c
was
a
target
and
the
next
stop
was
b
the
guy
after
me
in
this
list,
so
I'm
going
to
pass
the
packet
to
b,
so
b
gets
a
packet
and
that's
your
question,
michael:
what's
the
shape
of
the
packet
when
b
gets
it,
the
packet
is
source
the
root
arc,
id
0
destination
c,
because
a
has
consumed
the
writing
editor
and
has
placed
the
destination
to
c.
So
a
gets
a
packet
basically
root
to
c
c.
Is
my
neighbor.
Let
me
pass
him.
The
packet
c
gets
the
packet.
E
So
c
is
the
next
hop
in
the
loose
srh,
so
it
will
consume
that
hop
and
now
it
will
place
e
as
destination
same
thing.
Now
he
has
a
packet
to
e.
It
will
look
in
his
red
the
rib
size
from
this
dial
here
to
get
to
target
e.
My
next
stop
after
c
that's
d.
So
my
next
stop
is
d.
Let
me
pass
the
packet
to
d
same
thing
d,
as
a
packet
2
from
root
to
e
is
my
neighbor.
E
Let
me
give
him
the
packet,
so
you
see
that
the
two
ways
of
sending
the
dows
all
the
way
to
e,
or
just
to
be
the
note
before
that
gave
you
the
exact
same
result
same
thing.
The
loser
srh
that
the
root
has
done
could
have
been
the
ace
with
the
inner
packet
to
f,
in
which
case
it
encapsulates
the
packet
from
the
root.
E
That's
what
we
do
for
external
routes
or
the
route
could
encapsulate
all
the
way
to
f,
which
allows
to
to
leave
the
rpi
in
the
packet
and
maybe
e,
does
something
when
it
transmits
out
of
this
rpi
like
quartz.
I
don't
know
so
it's
useful
in
this
case
to
maintain
it
in
the
packet
and
then
that's
what
the
route
did
in
this
case.
So
we
can
it's
not
an
external
route
for
for
ripple,
and
so
basically
he
consumes
the
srh.
E
E
So
what
you've
seen
here
is
so
this
profile.
One
is
exactly
my
original
intention
for
this
draft
being
able
to
put
one
or
more
storing
mode
segments
in
a
non-storing
mode,
dear
dag,
to
compress
the
srh,
make
it
smaller
and
get
the
best
of
both
worlds
like
a
bit
of
stirring
and
a
bit
of
non-stirring.
E
The
root
maintains
how
many
epidiols
it
stands
so
that
any
any
node
will
not
have
too
much
state
beyond
its
memory
capabilities.
Remember
the
capability
draft,
which
will
come
later
and
may
help
the
root
figure
out
how
many
rods
can
install
in
each
node,
but
basically
the
idea
is
compared
to
full
non
to
full
staring
mode.
We
know
that
we'll
never
push
too
many
routes
in
any
of
those
nodes.
H
And
I
have
questions
sorry
if
it
does,
and
I
have
questions
sorry
if
it
is
too
naive-
the
communication
from
dota
group
to
any
other
node,
it's
clear
to
me.
I
mean
the
lock
has
the
view,
but
from
a
p2p
communication,
let's
say
leaf
node
to
another.
E
H
E
That's
the
that's
the
point
right
now
doing
story
mode.
You
would
exactly
that's
that's
exactly
the
point.
So
thanks
for
the
question,
if
we
are
doing
full
storing
mode,
we
would
end
up
in
a
situation
where
we
saturate
the
memory
of
nodes
near
the
root
at
some
point
and
that's
why
most
people
don't
implement
storing
mode
yeah
they
end
up
with.
If
you
have
a
long
line,
many
many
hubs
in
the
store
in
the
source
route.
E
H
E
E
That's
when
we
need
the
sibling
information
and
and
what
I
told
you
a
bit
earlier
and
we
can
come
back
to
that
is
you
may
have
a
thousand
sibling
which
of
the
sibling.
Do
you
tell
the
root
and
that's
what
I
said
we
don't
discuss
in
this
draft,
which
we
provide
you,
this
big
information
option
that
we
need
to
rename
because
of
the
sio
which
already
exists,
but
they
also
tell
you
if
there
is
more
than
two
three
siblings.
E
E
Okay,
so
that
was
just
profile,
one
guess
what
we
have
six
profiles,
and
that
was
the
easy
one.
Okay,
so
good
news
with
profile.
2
is
the
context
the
same.
We
have
the
menu
dial.
We
have
the
root
on
the
left.
We
have
the
destination
on
the
right.
The
only
thing
is,
we
are
using
estor
vio
instead
of
sfvio,
meaning
we
are
actually
adding
a
source
rotator
to
go
to
to
do
abc.
E
Now,
it's
no
more
stateful
using
storing
mode,
it's
stateless,
using
non-storing
mode,
meaning
that
this
is
implementing
what
howell
was
asking
me
very
early
in
this
design.
How
last
me,
I
think
it's
your
hour
right.
You
said
hey.
Instead
of
having
a
story
mode
between
cba,
why
don't
you
don't
you
pass
a
p
down
to
a
and
it
has
a
source
route
to
go
to
c?
E
Well,
that's
profile,
two!
Now
because,
because
this
source
route,
header
has
to
be
added
by
node
a
to
go
via
b,
a
has
to
encapsulate
remember
my
discussion.
Now
the
root
is
the
ingress,
so
I'm
changing
the
source
of
the
packet
and
inserting
an
srh.
E
E
You
you
stay
on
the
same
track
if
the
track
is
the
egress,
but
now
that
we
made
the
track
the
ingress,
if
you
change
the
source,
you
change
the
track
and
so
that
that
basically
preventing
prevented
the
the
games,
the
games
we
are
doing
before
so
so,
basically,
we
are
back
into
the
normal
mode.
Like
repo
user
free
pulse
says
you
are
inserting
an
srh,
you
must
change
the
source
in
the
sources
a
and
there
will
be
an
rpi
which
says,
for
instance,
129
in
the
ripple
instance
id
in
a's
namespace.
E
E
The
srh
is
inserted
by
the
root,
so
the
source
now
is
a
rule,
track
id,
say
zero
and
again
has
the
loose
srh
exactly
the
same
as
before,
because
the
root
is
not
aware
of
where
it
is,
but
it
doesn't
care
how
the
loose
hops
are
implemented,
whether
it's
vio,
srvio
or
rest
of
the
io
and
as
before.
You
have
two
ways
of
doing
it:
either
the
tunnel
ends
at
the
not
before
the
last
top
or
it
ends
at
the
last.
E
So
actually
it's
it's
just
here
to
reach
e.
Not
it's
not
aware
of
f
it's
the
inner
header,
which
will
be
aware
of
f.
So
what
d3
says
is
you
have
this
obd
e
and
it
it
in
this
store?
The
io
could
try
to
the
normal
vio
to
the
sfvio.
E
E
So
in
the
rib
there
is
a
rod,
destination,
d
and
source
rotator
d
e,
and
basically
that
that
makes
the
packet
go
source
routed
all
the
way
to
e,
which
means
that
e
is
the
node
which
will
decapsulate.
E
Now
you
see
that
the
other
possibility
is,
like,
I
said,
to
end
the
tunnel
at
the
node
before
the
destination,
so
going
all
the
way
to
c
the
node,
which
is
just
before
him
is
b
and
that's
what
the
the
the
pdo2
does
the
ingress.
So
the
root
of
this
track
is
a
track.
Id
is
say,
129
and
just
pick
that
number
out
of
a's
namespace
and
the
srv
I
o
just
says
b
for
target
c,
so
you
realize
we
are
signaling
the
exact
same
thing.
I
E
Right
because
we
install
a
source
rotator
at
the
ingress
and
the
rest
of
the
network
doesn't
need
to
know
exactly
that's
the
big
difference.
The
flow
is
is
simpler
with
the
source
rod.
That
was,
I
guess,
the
motivation,
if
I
remember
the
mails
from
howell,
so
the
motivation
is
the
signaling
is
simpler,
because
you
just
tell
it's
a
question
response
with
the
root
of
the
track,
so
message
to
the
root
of
the
track
answers
by
the
root
of
the
track.
Here
we
have
to
do
what
the
dao
does.
E
The
tao
is
sent
by
the
child
to
the
parent
to
the
parent
for
establishing
a
route
the
other
way,
so
the
dialogue
is
sent
by
a
node
which
is
different
yeah.
That's
that's
a
big
difference.
The
other
big
difference
is
here.
You
don't
re-encapsulate.
The
packet
is
the
same
as
was
written
by
the
root.
It's
just
that
the
srh
is
consumed
by
a
c
and
e,
but
the
packet
is
not
re-encapsulated.
E
So
you'll
see
that
so
you've
seen
that
one
was
storing.
Two
was
known
stirring,
we'll
keep
the
pace
where
three
will
be
stirring
and
five
will
be.
Storing
and
four
will
be
non-storing
and
six
would
be
long
story.
There
is
always
one
way
or
the
other
way,
one
way
or
the
other
way
one
way
or
the
other
way.
E
E
E
It's
a
packet
which
is
you
know
it
could
have
an
rpi
of
zero
for
all
or
I
care
you
know
it
could
be
anything,
but
it
just
goes
that
if
you
want
to
to
forward
it
along
the
track,
you
need
to
signal
a
track.
So
three
is
like
one:
it
uses
stateful
vio,
so
it's
storing
mode.
E
So
if
it
was
along
the
main
diode
or
if
you
know
the
track
id
in
the
packet
was
already
correct
and
the
source
was
already
correct
like
if
the
source
was
the
root
and
the
rpi
was
the
right
rpi,
then
a
would
not
need
to
re-encapsulate
because
it's
what
it's
configured
with,
but
in
this
case
we
have
an
rpi,
which
is
say,
zero
and
the
source
which
is
f,
but
in
the
p
dow
I've
seen
that
the
ingress
is
a
it's
not
f,
and
I
see
that
the
track
id
should
be
129.
It's
zero!
E
Oh,
so
it's
not
my
track.
If
I
want
to
forward
it
along
my
track,
even
if
the
track
was
even
if
it's
it's
just
a
segment
that
was
signal,
remember
a
pdo,
storing
mode,
it's
just
a
segment.
Yes,
it's
a
segment,
but
there
was
there
is
no
track.
I
don't
see
the
track,
which
will
tell
me
how
to
encapsulate.
E
E
Just
to
make
a
the
source
and
follow
the
path
abcde,
which
is
called
track
id129.
E
E
But
here
it's
it's,
it's
not
there.
So
it's
simplicity
there,
it's
like
if
I
had
sent
this
dowel
and
also
a
dial
which
says
source
route,
a
e
which
is
of
great
interest
and
track
id
129..
E
E
So
what
happens?
Is
a
gets
a
packet
to
f?
He
has
this
diode.
This
dow
says
for
ta
gets
enough.
So
in
this
case
it's
f.
My
next
stop
is
b
because
I'm
a
so
a
passes.
The
packet
to
b
b
has
seen
the
same
dow,
the
dow
size
for
target
e
and
f.
So
in
this
case
the
destination
is
f.
My
next
stop
is
c,
so
I
pass
the
packet
to
c
c.
Will
do
the
same
thing
pass
the
packet
to
e.
E
E
Once
you
reach
e
is
the
last
entry.
What
is
the
destination?
I'm
sorry
in
the
outer
header,
so
the
outer
either
has
a
source,
a
destination
e
or
the
all
the
way
through
so
e
being
the
destination
it
will
decapsulate,
because
there
is
no
srh,
it
will
just
encapsulate
oops
I
moved
and
now
we
get
back
the
same
packet
as
before,
and
the
packet
is
passed
to
a
question.
E
The
difference
between
three
and
one
is:
is
that
the
instead
of
being
the
root
which
has
already
placed
the
right
rpi
and
everything
and
is
actually
what
was
signal
here?
So
if
you,
if
you
look
at
the
dow
in
one,
the
ingress
was
signaled
as
being
root
and
the
track
at?
U
was
zero.
So
when
a
got
the
packet
a
did
not
need
to
re-encapsulate,
because
the
root
was
the
one
from
the
dow
and
the
track
id
was
the
one
from
the
tao,
so
the
packet
was
already
on
track.
There.
I
D
D
I
E
Well,
you
could,
if
you,
if
you
do
the
code
for
three,
it
would
work
for
forward.
But
if
your
code
just
works
for
one,
maybe
it's
not
capable
of
supporting
more
than
one
track.
Okay,.
E
Need
to
have
a
table
which
says
you
know,
I
look
up
the
ingress
and
the
track
id
and-
and
I
find
my
next
step
based
on
that,
like
the
the
c
could
get
a
packet
for
ingress
y
track,
id
130
and
so
depending
if
the
packet
is
coming
from
y
or
coming
from
a
the
next
up
will
not
be
the
same
thing.
E
E
A
complex
track
is
when
you
make
you
make
it
comes
later,
when
you
make
multiple
segments,
so
this
is
a
complex
track.
E
E
You
see
just
like
in
row,
because
now
you
have
multiple
segments,
so
you
can.
You
can
have
a
whole
graph,
whatever
you
dag,
really
and
just
like
a
classical
geodag
with
the
root
in
a
and
you
forward
down
the
diode,
but
you
establish
it
up,
so
you
could
have
many
many
segments,
many
nodes
as
long
as
you
have
global
your
geodag.
E
And
for
this
one
we
can
decide
to
make
it
serial
or
to
to
say
for
the
same
track
id
same
ingress
you
can
have.
You
can
send
multiple
dials
with
different
srvios.
As
long
as
it's
a
the
root
you
could
say,
I
can
do
like
a
wheel.
If
I
like,
where
a
is
the
hub
and
e
is
a
spoke,
we
could
have
multiple
spoke
and
reuse,
the
same
track
id
and
some
ingress,
in
which
case
it
would
not
be
a
serial
track.
E
It
would
be
a
hub
and
spoke
so
the
draft
does
not
prevent
the
hub
and
spoke,
but
really
the
thinking
is
to
do
the
exact
same
thing
as
three
by
choosing
a
tunnel
and
I'm
actually
missing
the
encapsulation.
So
let
me
fix
my
track.
E
So
well
well,
it's
represented
by
by
the
blue
thing.
Okay,
so
it's
the
other
thing.
So
basically,
at
this
on
this
slide
we'll
be
doing
the
exact
same
thing:
we'll
be
encapsulating
blue,
so
the
packet
is
gray.
We
encapsulate
it
in
a
and
we
encapsulate
in
e.
The
only
difference
is
it's
an
srvio,
so
the
srvio
gives
you
the
next
hops.
So
what
ac's
access?
Rvio
is
b
c
d
e
with
the
v
io.
E
E
So
we
have
this
double
implicit
that
the
last
entry
point
is
a
possible
target
and
that
the
first
entry
is
the.
Why
do
we
do
this?
It's
because
now
the
srvio,
as
presented
here
expressed
in
rfc8138,
is
exactly
what
goes
in
the
packet,
because
the
source
is
not
part
of
the
routing
either
only
the
destination,
and
then
the
next
hops.
So
formatted
like
that,
you
get
exactly
the
routing
either
as
you
wanted
it,
whether
you
want
to
reach
e
or
you
want
to
reach
f.
We
only
signal.
E
Stargate
e
is
also
implicitly
a
target,
so
so
s
passes
the
packet
to
a
a
adds,
an
srh
which
says
destination
b
and
srhcde,
or
in
the
complex
form
bcd
passes.
The
packet
to
b
b
consumes
the
destination,
makes
the
destination
c
and
the
routing
header,
the
e,
which
is
what
I
wrote
here.
So
that's
the
non-compressed
form
and
et
cetera,
et
cetera.
The
writing
header
gets
consumed
till
it
reaches
e,
which
is
the
last
hop
in
the
writing
error
the
capsulates
then
to
f.
E
So
it's
exactly
the
same
thing
as
if
we
are
talking
to
a
dexter
destination,
but
instead
of
being
the
root
which
does
the
encapsulation
now
it's
I.
E
E
I
So
we're
effectively
picking
an
instance
id
that
is
is
renamed
it
the
track
id
because
you've
reused
the
rpl
instance
id.
E
Is
a
local
instance
id
and
the
elective
thing
says:
basically,
if
you
don't
understand
this
thing
ignore
so
basically
you
could
decide
to
if
you're,
using
only
non-storing
mode
like
this,
this
profile
4.,
you
could
decide
not
to
implement
this
at
all.
It
might
be
some
nodes
in
the
network.
They
do
implement
it
well
because
of
the
ebit
set
to
elective
you're.
Okay,
to
ignore
completely
the
it's
like
the
two
and
the
six
you
know
in
in
the
bit
which
change
with
use
of
3.4.
E
E
E
Maybe
that's
an
answer
profile.
E
E
Well
oops!
This
is
a
typo.
E
So
so
you
don't
need
to
re-encapsulate
the
packet,
which
is
a
129
and
if
you're
a
that
means
you're
the
ingress,
and
you
realize
that
we
have
one
additional
versus
before
the
track
is
not
implicit.
E
In
this
case,
the
track
is
in
is
rooted
in
a
so.
A
will
have
two
dao.
E
First
dao
is
a
non-starring
mode
which
gives
you
this
to
lose
ac
e,
and
that
will
cause
the
first
imp
that
will
cause
the
encapsulation
so
instead
of
profile,
one
where
the
encapsulation
was
done
by
the
root
and
a
did
not
need
to
encapsulate
now
you
realize
that
a
is
the
ingress
of
the
track.
E
E
But
then
this
explicit
track
has
a
loose
sequence
of
hubs
and
it
will
go
across
the
segments
which
are
done
by
pedal
one
and
two.
If
you
look
at
the
definition
of
of
the
segments,
it's
exact
same
as
in
one.
So
if
you
understood
one,
you
understand
them
the
only
difference
and
back
to
michael
asked
about
if
the
implementation
is
the
same
as
before,
mostly
yes,
if
they
support
any
track,
you
know
it's
it's
basically
very
similar
to
what
they
have
to
do
here.
They
have
to
understand
the
track
and
they
follow.
E
It's
just
that,
if
you're,
mr
r
and
mr
c
that's
kind
of
interesting,
because
mr
c
operates
a
bit
like
a
in
in
the
original
flow
here,
because
the
packet
is
already
encapsulated,
it
already
has
the
right
ingress.
A
the
source
of
the
packet
is
a
when
c
receives
it
is
the
destination,
so
it
will.
It
will
process
the
rotting
header,
which
c
has
insert
which
a
has
inserted
and
it
will
make
the
destination
c,
but
but
basically
the
packet
that
c
reply
receives
as
the
right
source.
E
E
So
the
difference
between
a
and
c
is
that
for
a
it's,
not
an
implicit
track
compared
to
profile
three
profile:
three,
you
would
not
in
a
you
would
not
have
gotten
p
o3,
so
a
would
have
encapsulated.
But
actually,
since
f
is
not
a
target,
you
would
not
even
have
resolved
it.
The
only
thing
that
resolves
for
a
is
because
you
received
this
pdo3,
for
which
a
target
is
f.
E
So
if
you
look
at
the
rib
in
a
the
only
way
to
to
reach
f
is
actually
via
this
track,
a129,
which
implies
encapsulating
with
an
ester
vioce
so
and
himself
source,
so
it
does
and
once
the
process
is
packet.
Oh,
I
need
to
reach
c
that's
when
I'm
going
to
use
the
state
coming
from
p
02,
which
says
your
next
step
is
b.
E
So
there
is
kind
of
a
recursion
that
happened
in
a
c
just
found
that
the
packet
was
already
encapsulated
correctly,
but
a
good
source,
good
rpi.
So
I
have
a
route
to
e
in
that
red
next
up
is
delete.
E
I
know
it's
hard
right.
I
mean
I've
been
thinking
about
that
for
four
months.
You
just
had
10
minutes,
but
you'll
find
this
in
the
mailing
list.
You
also
find
it
in
the
draft
in
the
draft.
I
I
put
many
tables
with
the
ribs
with
what
goes
in
the
bow.
What
goes
into
target
everything-
and
I
also
explain-
what's
in
the
packet
at
this
stage,
which
is
kind
of
summarized
here,
but
if
you
take
this
and
you
take
the
draft,
the
slides
are
a
bit
of
a
compliment
to
the
draft.
E
So
all
these
together,
you
should
be
able
to
to
follow
and
profile.
Six
is
the
same
as
profile
five,
so
you
see
that
five
and
six
are
the
profiles
which
enable
you
to
do
complex
tracks
because
they
do
segment
routing.
Basically,
what
I
call
segment
routing
is
very
much
what
the
atf
calls.
I'm
not
writing.
Anyway,
all
those
things
are
segments
and
by
placing
a
source
rotator,
you
can
go
across
segments,
so
it's
segment,
routing
for
ripple
and
the
segments
can
be
can
be
a
longer
complex
track.
E
So
six
is
just
like
five,
but
it's
all
source
rotted.
So
so
what
that
will
cause
is
a
re-encapsulation
because
first
based
on
p03,
when,
when
you
know
this
packet
from
s
to
f,
reaches
a
a
looks
at
its
rib,
the
only
way
the
only
target
f
is
here.
So
that
means
I
need
to
encapsulate
source
myself
track
id
141.
E
And
I'll
put
together
edge,
saying
c
and
e,
meaning
in
non-compressed
form,
destination,
c
and
srh,
which
is
what
you
can
see
here
now
now
a
ends
up
with
a
packet
for
instead
of
packet
for
f.
Now
I
have
an
encapsulated
packet,
but
for
c
c
is
still
not
my
neighbor.
So
how
do
I
get
to
c?
I
have
c
as
a
target
here
for
which
my
next
up
is
b.
E
E
So
I
will
re-encapsulate
the
packet,
and
now
we
have
a
packet,
an
outer
packet,
which
is
from
a
to
b
track
id
129.
So
it
looks
a
bit
stupid
because
there's
just
one
hub,
but
you
think
about
three
halves
between
a
and
b
and
you
realize
that
you
need
an
srh
etc.
So
you
need
an
encapsulation,
so
the
packet
is
decapsulated
by
b
and
it's
very
similar
to
what
you
had
here.
It's
just
that
the
inner
packet,
the
blue
packet,
is
not
the
one
which
came
from
the
left.
E
The
difference
between
this
profile
2,
and
what
we
have
here
is
that
the
packet
from
the
left
was
encapsulated
the
first
time
by
a
to
to
add
the
the
loose
source
via
ce,
and
then
we
do
the
exact
same
thing
as
we
did
before.
So
so
you
end
up
having
a
doing
two
level
of
encapsulation
b,
removing
one
level
of
encapsulation,
and
so
what
you
have
is
the
the
inner
packet.
A
to
c
c
is
my
neighbor.
I
pass
it
to
c
blah
blah
blah
same
discussion
as
for
profile.
E
2
reaches
destination,
which
is
c
c,
looks
at
the
next
step.
In
the
storage,
which
is
e
o
for
e,
my
only
target
is
the
implicit
target
here.
So
I
need
to
encapsulate
with
an
srviode.
E
And
so
so
that
creates
a
encapsulation
that
e
decapsulates
and
plug.
E
That
was
most
of
it,
so
I
thank
you
for
bearing
with
me
ready
for
any
questions.
No,
no.
There
are
still
a
number
of
points,
but
but
I
told
you
the
main
changes
it's
it's.
Basically,
the
root
is
the
ingress
to
signal
the
track.
You're
in
you,
you,
you
have
to
change
the
the
source
and
to
encapsulate
in
non-storing
mode,
which
makes
so
that
you
end
up
with
this
tracking
track
that
I
just
described
here.
So
you
have
more
encapsulation
than
in
the
previous
iterations
of
the
draft.
A
E
E
So
the
first
thing
is
the
lifetime
unit.
Remember
we
discussed
it
earlier
ripple
instead
of
having
a
long
unit
or
an
expansion
unique
for
for
time,
we
provide
in
the
dodag
configuration
option.
We
provide
the
lifetime
unit,
which
is
two
bytes.
E
That
tells
you
what
one
is
when
we
express
time
whether
it's
one
second
one
minute,
one
something,
and
so
it's
just
like
a
multiplier.
If
you
like,
and
so
every
time
you
find
something
like
a
lifetime,
for
instance
in
a
transit
information
option,
or
something
like
that.
E
E
E
So
he
has
some
questions
about
that,
but
that's
one
thing
we
could
discuss
on
the
mailing
list.
Second
is
track
id
and
ripple
instance
id
so
yes,
a
lot
of
the
of
the
work
that
you
have
seen
is
to
make
a
track
be
exactly
the
same
thing
as
the
main
diode.
It's
just
another
non-storing
geodag,
just
that
it's
a
local
ripple
instance.
E
So
it
means
you
need
the
api,
but
you
also
need
either
the
source
of
the
destination
and
now
it's
the
source
and
and
so
when
you
do
a
an
insertion,
you
have
to
change
the
source.
That's
what
sigma
told
us,
because
we
change
the
source.
We
change
the
track,
meaning
that's
a
different
track
and
it's
in
the
re-encapsulation
blocks.
That's
what
I
discussed
already.
E
And
so
the
question
was:
how
does
a
node
understand
whether
a
message
is
a
normal
pidao
or
well
a
normal
ripple
message,
so
he
wanted
basically
to
make
not
just
in
the
pdow
but
also
in
the
data
packet.
He
wanted
to
to
make
it
clear
whether
we're
talking
about
a
packet
which
is
forwarded
along
a
track
versus
a
normal
ripple
instance
operated
normally
by
ripple
and
in
this
it
is
code.
Actually
that
was
a
different
cut
path,
so
it
was
useful
for
for
implementation
to
make
a
difference.
E
E
We
actually
update
it
to
to
use
a
bit
saying
this
is
forwarded
along
a
track
a
longer
projected
now.
So
that's
why
we
extend
six
five
feet
three
and
we
actually
create
a
new
six
lower
range
where
this
flag,
which
says
I'm
going
along
a
track,
is
implicit,
but
it's
there.
So
if
you're,
a
simplicity
set,
meaning
that,
if
you,
if
you
use
six
large
type
seven,
you
must
be
forwarding
along
the
pirate,
the
bit,
which
says
forwarding
a
longer
period
is
set.
E
E
Then
it
was
the
question
about
directionality
and
for
now
we've
dropped
that.
But
it's
it's
open
to
the
group
right.
It
was
asked
on
the
mailing
list
because
we
are
doing
a
audio
deck,
so
you
could
say:
hey.
Could
I
forward
the
packets
back
to
the
root
like
we
do
in
normal
in
the
main
diode?
Well,
the
thing
is
in
this
thing:
there
is
no
dio.
If
there
was
a
dio,
it
would
be
iodv
ripple
or
something
there
is
no
gio.
E
So,
if
you're
using
non-storing,
how
do
you
know
how
to
get
there
if
you're
using
the
storing
mode
segments,
you
could
assume
that
you
have
a
way
back,
but
with
non-storing
you
don't
even
know
so
for
now
we
we
said
if
you
want
to
go
the
other
way.
You
also
build
a
track
the
other
way,
but
there
is
no
return
path
and-
and
we
are
looking
for
simplicity,
if
we
we're,
if
we're
trying
to
do
bit
directional,
I'm
afraid
we
are
adding
a
lot
of
complexity.
E
E
E
So
for
the
for
the
bi-directionality,
you
realize
there
is
no
dio,
so
we
don't
know
the
way
back.
We
could
follow
the
reverse
epido
in
storing
mode,
but
it
does
not
always
take
you
all
the
way,
and
you
never
know
I
mean
it's
not
meant
for
it
for
non-storing
it
it's
just.
There
is
absolutely
no
state,
we
don't
know
so.
E
If
you
have
clues,
but
for
now
the
answer
is
for
now,
the
answer
is
no.
Everybody
has
the
same
lifetime
unless
you
tell
me
something
different,
but
the
draft
right
now
and
I
have
no
plan
to
change
it
unless
I'm
being
pushed
so
so
if
nobody
raises
a
hand,
nothing
will
happen.
E
E
The
cases
of
loops
will
be
this
and
in
particular,
when
you
do,
you
can
re-enter
this
several
times,
because
even
this
track
too
can
be
actually
loose
leading
to
re-encapsulation,
and
I
don't
know
if
at
some
point
you
may
create
loops.
I
believe
you
can.
That
would
be
crazy,
but
I
believe
you
can
should
we
have
all
the
safeguards
against
that.
E
E
It's
always
the
guy
who
will
be
the
ingress
of
the
truck,
but
it
could
be
an
application.
It
could
be
a
third
party,
you
know,
maybe
I'm
a
neighbor
of
a
and
I
want
to
reach
d,
and
I
just
tell
the
root
give
me
a
path
from
a
to
z.
You
know,
and
I
will
pass
my
packets
to
a
so
open
to
suggestion
on
this.
For
now
I
have
no
action.
E
E
This
I
did
so
what
was
called
aprpos.
That's
the
name
that
went
away,
so
it's
the
vios,
whether
it's
rviu
or
sfviu,
they're,
all
vios.
Now
and
at
the
moment,
what
what
is
is
ins?
I
have
made
that
more
specific.
I
changed
the
text
to
make
sure.
Basically
what
we
do
is
in
non-storing
mode.
We
start
at
the
node
after
ingress
all
the
way
to
ingress
included,
and
that's
because
we
want
this
to
be
the
exact
test
range
that
will
be
placed
in
the
packet
in
x138
and
in
storing
mode.
E
E
Okay,
so
I
think
we're
pretty
close
to
have
everything.
It
is
a
good
path
of
cleaning
to
be
done,
and
if
you
I
need
to
look
again
at
the
tickets.
I
understand
this.
Yes,.
A
E
E
Yes,
the
problem
is
not
really.
The
terminology
is
the
fact
that
sibling
starts
with
an
s
so
right
now
we
in
the
draft.
We
call
it
an
sio
and
right
from
some
reason,
the
this-
the
discussion
in
ripple
on
the
this,
and
you
find
that
the
solicit
information
option
that
you
may
pay
for
this
is
sio
as
well.
E
J
E
E
Okay,
I
hope
that
my
co-authors,
that
would
work
on
that,
but
I
have
no
news
from
from
some
of
them
for
a
very
long
time.
So
I
just
don't
know
what
they
are
doing.
A
Okay,
thank
you
very
much,
so
we
will
collect
the
action
points
and
hopefully
the
recording
is
available
soon
as
well,
and
then
we
will
send
the
coding
with
the
action
points
to
the
mailing
list.
Oh
that's,
fantastic.
J
E
E
The
end-
and
let
me
know
things
like,
for
instance,
if
you
want
me
to
to
make
things
implicit,
I
told
you
for
now
this
this
looks
like
a
duplication,
see
this
is
always
there
and
there
okay
enough,
looks
like
a
duplication,
could
make
it
implicit
easy
and
whether
we
we
say
should
must,
for
instance,
in
c
should,
should
it
say,
with
this
non-storing
m,
with
the
storing
mod,
should
it
also
place
d
in
its
routing
table
as
a
destination
right
now
it
does
not
just
place
e
in
his
destination,
his
routing
table.
E
E
If
you
want
me
to
install
the
whole
list,
everything
which
is
after
you
in
this
list,
I
can
do
text,
but
that
could
be
a
discussion.