►
From YouTube: IETF101-LISP-20180319-1330
Description
LISP meeting session at IETF101
2018/03/19 1330
https://datatracker.ietf.org/meeting/101/proceedings/
A
C
A
E
A
F
G
E
H
I
A
E
A
C
A
A
A
C
A
L
Hello,
let's
start
welcome
everybody.
This
is
the
least
working
group
you
can
see
on
the
mean
scream
that
I
was
not
well,
which
you
accepted
when
registering
for
this
ITF.
So
basically
you
agree
that
whether
whatever
you
say
today
here
is
a
contribution
to
the
ETF,
okay
and
issue
by
any
chance,
you
know
an
IPR
that
applies
to
to
whatever
we
discuss
here.
We
are
supposed
to
tell
us:
okay,
the
usual
pointers
to
the
various
material
I'm
Luigi
anona,
here's
Joel
Halpern
we
had
the
cultures.
Was
him
in
the
front
row
at
mound?
L
Quick
update
on
the
status
of
the
documents,
so
we
have
the
usual
introduction
document
has
been
def
a
few
years.
We
missing
just
a
reference
will
come
later
on.
This
is
actually
the
document
that
came
back
from
the
is
G
in
order
to
switch
it
to
to
stand
attack.
We
have
the
signal
free
document
which
is
now
in
a
RFC
editor
queue
so
making
progress,
and
we
have
the
the
truth
to
B's
drafts
that
we
hope
we
are
close
to
to
working
group
last
call.
We
will
talk
right
after
all,
that
will
give
a
presentation.
L
L
K
L
L
So
their
agenda
first
today
we
will
start
with
appointed
about
their
these
documents.
Then
we
have
this
GP
the
young
model
and
vendor-specific
al-kahf.
Okay,
these
are
working
to
defense
and
then
we
have
of
you
know,
working
group
items
we
will
discuss
with
publishers
subscribe.
We
we
have
to
switch
to
presentation.
L
Actually
the
map
server
originated
SMR
and
the
lease
control
plane
for
ila,
okay,
so
to
let
Tom
Canberra
to
attend
a
presentation,
and
then
we
will
continue
with
two
presentation
about
the
Graham
based
Lisp
analytical
telecommunication
network
and
the
corresponding
BGP
routing,
and
then
Dina
will
give
us
two
presentation
about
a
decentralized
list,
mapping
system
and
an
update
on
the
mobile
network.
Okay,
any
comment
on
the.
F
L
L
We
discussed
this
already
last
time
when
we
took
the
document
back.
The
basic
idea
is
we
switch
it
to
standard
track
what
this
is
pretty,
but
we
make
sure
that
is
coherent
with
the
B's
documents.
So
once
we
last
call
the
beast
documents,
we
do
a
proof,
read
the
security
document
and
we
ship
it.
Our
associate.
L
F
H
Eh,
the
problem
is
pulling
something
from
the
RFC,
any
our
cue
to
say,
change
the
text
so
that
there
was
a
reference
to
these.
It's
not.
We
can't
just
change
the
references.
We'd
have
to
change
the
text
and
I'm
pulling
something
from
the
RFC
in
here
to
change.
The
text
means
that
has
to
go
back
to
me.
Working
group
last
call
IETF
last
call
is
cheap,
so.
M
So,
first
for
the
data
plane,
so
we
had
many
revisions,
so
it's
5
seems
last
IDF
in
Singapore.
Actually,
revision
number
12
was
posted
today,
like
a
few
hours
ago
and
I'm
summarizing
all
the
changes,
including
number
12,
somehow
I'm,
trying
to
start
from
more
important
to
to
less
important.
So
the
first
one
is
that
we
move
services,
metric
map
request
from
data
plane
to
the
control
plane.
The
second.
Why
is
that?
M
Then
we
also,
they
did
the
section
on
router
locator
selection,
stating
just
that
that
a
plane
must
follow
what
is
stored
on
the
map
guides,
meaning
that
you're
taking
the
contrary,
different
weights,
and
you
follow
that
and
how
we
set
priorities
and
weight
is
up
to
the
control
plane.
Then
we
remove
section
16,
17
and
18,
which
were
mobility,
considerations,
the
primary
considerations
and
treasurer
considerations,
and
this
text
must
be
placed
in
a
new
OEM
document.
M
And
then
there
is
a
bunch
of
other
comments,
so
we
remove
for
reference
to
research
because
they
don't
make
sense
anymore,
and
this
research
has
been
abraded
on.
We
made
it
compliant
with
21,
19
and
90
needs.
We
went
through
many
many
many
little
changes.
You
can
change.
You
can
check
all
the
divs.
Then
we
move
the
control
import
from
the
data
plane
to
the
control
plane,
because
this
was
really
wrong.
M
We
also
many
commenters
commented
on
the
definitions
and
we
ablated
that
we
also
dated
the
text
regarding
encapsulation
and
decapsulation
regarding
the
dscp
field,
and
this
is
the
last
change
for
today.
This
was
followed
following
and
Regis
comment.
We
provide
that
an
immanent
is
required
forever
lock-in
set
of
per
destination.
We
remove
the
clock,
sweep
section
which
is
related
to
how
to
update
the
ad
to
error
of
mappings,
and
this
text
again
should
be
placed
on
a
new
OEM
document,
and
then
some
reference
went
from
normative
to
informal.
M
So
regarding
the
Contra
plane,
three
versions
since
last
idea,
our
last
one
was
posted
yesterday,
so
we
added
the
control
plane
poured
into
Ayana,
because
this
is
a
control
import.
We
other
told
the
tags
regarding
routing
locator
authority
from
the
data
plane
that
I
just
mentioned
that
we
removed
same
with
search
it
map,
request
and
same
with
air
lock
glory,
so
pretty
much.
M
F
L
M
L
M
M
L
We
have
the
right
context.
Ization,
do
you
need
one
what
I
mean
for
me?
It's
pretty
pretty
simple
in
you
use
more
you,
freelance
introduction
that
says
that
Lisp
is
defined
in
data
plane
with
30
B's,
console,
plane,
33
B's
and
in
this
document,
discussed
all
the
things
that
are
in
the
middle.
Let's
say
a
management
administration,
etcetera
till
it
and
the
you
add
if
you,
what
I
really
want
to
be
picky,
the
reader
is
supposed
to
be
familiar
with
the
other
joints.
M
Then
I
have
a
problem
with
the
expectations
are
regarding
the
reader,
considering
the
title.
So,
if
I
entitle
a
document
list
over
again,
I
expect
to
read
about
the
whole
spectrum
of
what
means
operation
and
management
of
list
and
then
I.
What
I
found
out
are
three
sections
which
are
very
narrow
regarding
three
different
aspects:
let's
say
that
without
an
and
the
other
one
and
then
I
feel
a
little
bit,
but
this
so
I
have
a
problem
with
the
title
really
and
what
people
should
expect
in.
M
M
F
L
F
F
F
You
that's
people
should
read
the
section
and
make
this
decision,
because
there
was
a
lot
of
Scotland
put
a
lot
of
thought
into
writing
that
text
and
I,
don't
think
we
should
lose
it
because
people
have
different
expectations
of
what
mobility
is.
Changing.
Service
providers
is
a
mobility
event
that
happens
once
a
year,
possibly
versus
being
on
a
high-speed
train,
changing
your
outlooks
and
that
needs
to
be
set
up
somewhere.
F
L
P
L
L
We
can
do
is
this
afternoon.
I
sent
an
email
on
the
mailing
list
stating
that
we
wish
there
are
concerns
about
the
name.
The
one
discussed
right
now,
something
where
our
mobility
deployment
and
trace
consideration
anybody
complaints,
adding
body,
has
a
different
kinda
suggestion
and
by
the
end
of
the
week,
having
the
decision.
What
do
you
think?
A
I
P
Problem
I
know
presenting
about
Lee's,
GP
and
I.
Get
this
to
be
easy
with
you.
So
this
is
the
status
very
quick
status
update.
There
is
a
list
of
the
toilet
changes
that
were
saying
thanks
by
Luigi
and
those
are
being
introduced
it
in
version
0
1.
There
is
one
technical
comment:
opening
that
also
came
from
Luigi
and
I.
P
Think
Dino's
comment
about
backward
compatibility,
so
I
have
two
slides
the
next
to
slide,
proposing
out
to
address
that,
and
the
idea
is
to
use
the
same
logic
that
are
used
for
cat6
t1,
the
descriptor
and
also
on
status.
Update
I
just
wanted
to
you
know
just
let
the
working
group
know
that
there
are
2
implementation
of
the
GP
that
are
open
source
in
FDI,
yo,
+
depois,
so.
E
P
Is
the
idea
on
the
proposal?
So
the
problem
is
how
to
handle
the
case
where
there
is
a
mixed
deployment
and
some
of
the
router
are
released.
Router
and
some
of
the
router
level
is
GPU
now
at
least
GP
capable
drought
so
looking
at
it,
it
seems
to
me
that
this
is
exactly
the
same
problem
that
Li
script
is
addressing
right,
so
when
we
basically
send
the
least
encrypted
packet
in
a
network,
the
receiving
router
may
or
may
not
be
capable
to
understand
the
packet,
so
English
crypto.
P
P
This
is
the
text
that
we
have
today
that
we
want
to
basically
change-
and
here
I,
don't
know
some
comments.
I
heard
this
morning
when
I
share
the
slides,
we
thought
that
he
was
pointing
out
that
this
may
be
even
little
bit
too
tight,
because
really
you
know
here
what
I'm
saying
is
that
the
idea
has
to
explicitly
acknowledge
the
capability
and
I'm
saying
that
the
more
people
that
are
playing
that
calf
type
has
to
be
used.
There
may
be
deployments
where
everyone
can
do
static,
configuration
I
think
so.
Maybe.
P
Q
Algorithm
just
clapping:
what
are
you
saying
right
now?
My
point
here
was
that
we
should
definitely
use
the
multiple
determinants
have
to
to
detect
or
to
be
sure
if
they're,
if
they
XT
on
the
other
end.
Suppose
this
or
not,
but
that's
something
that
we
don't
need
to
use
if
we
know
for
sure
be
a
configuration
or
is
that
that
the
GP
is
reported.
So
four
cases
are
when
we
are
not
sure
we
can
use
the
ELCA
for
other
cases.
We
can
avoid
that
completely.
L
Have
a
clarification
question
now
with
this
G
bit:
okay,
you
have
the
a
way
to
understand
whether
or
not
another
unlocks
course
GP.
Now,
assuming
you
support,
if
it's
not
okay,
you
you
use,
let's
say
legacy
least
vanilla
least,
if
it
supports
GP.
How
do
you
deal
with
the
nonce
and
versioning?
You
want
to
keep
it
the
way.
It
is
that,
due
to
shin.
F
Add
to
a
comment
based
on
what
you
just
said:
if
the
encapsulation
types
returned
by
the
et
are
all
the
data
planes
that's
supported
by
it?
Well,
all
those
bits
will
be
set,
so
you
don't
have
to
say
if
the
G
bit
is
is
not
set,
and
if
that
encapsulation
type
comes
any
other
data
planes,
you
don't
have
to
say
use
the
default.
The
only
time
you
use
the
defaults
from
a
1630
is
when
this
elk
calf
has
not
returned,
because
you
have
no
information
about
any
data.
Planes
accept
the
default
one
right.
That's
all.
F
F
Well,
good
point
that
might
we
should
maybe
check
to
see
if
the
text
says
that
yeah
so
related
to
the
same
point,
since
we
have
to
change
it
to
add
the
G
bit.
Should
we
take
an
inventory
of
all
the
popular
data
planes,
like
maybe
right
like
add
the
I
bet
for
aisle
a
since
it's
becoming
fashionable?
No,
no,
no,
it's
not
there
and
then
the
question
is
show
your
little
ones.
We
think.
P
F
F
L
B
L
L
K
L
The
last
call
we
should
have
the
the
new
version
of
the
document,
but
we
could
sense
the
room
if
there
is
out,
but
then
we
will
proceed
with
a
last
call
on
the
mailing
list.
Okay,
so
for
we
assuming
all
these
changes
will
be
done.
Okay,
if
you
agree,
stand
to
go
for
working
the
platter
last
call
and
disturb
Mandi
stone.
Please
hum
now.
L
N
N
We
need
to
work
on.
We
need
to
do
quite
a
few
more
changes,
so
the
security
considerations
section
is
pretty
empty
right.
Now,
that's
very
important.
We
need
to
change
who
yang
version
1.1.
The
yang
itself
will
not
change
because
of
that,
and
also
we're
going
to
be
adding
a
few
XML
examples
which
will
validate
using
yang
lint.
N
Also
right
now,
the
way
the
model
is
the
list
is
at
the
top
of
the
tree,
and
the
guidelines
is
to
have
this
as
a
control
plane
protocol,
so
we'll
have
a
new
identity
and
we're
going
to
augment
the
control
plane
protocols
existing
model
which
is
being
modified
in
880
22
bits
right
now,
and
also
we've
done
a
comparison
with
some
of
the
data
which
is
in
the
list.
Nib
7th,
RFC,
70,
52
and
we've
noticed
that
there
are
a
few
things
which
are
missing
in
the
yang
model.
N
The
major
things
are
the
three
things
identified
here.
For
example,
we
don't
have
anything
to
map
a
VN
I
to
a
vrf.
The
Mable
has
an
iid
to
vrf
table.
We
don't
have
any
leaf
nodes
to
explain
what
them
up
to
identify
what
the
map
cache
limit
and
map
cache
size
are,
and
also
we're
missing
x
times
for
stuff,
like
what
was
the
first
registration
in
the
last
registration
any
idea.
So
those
three
things
are
things
which
are
in
the
memorize,
though,.
F
F
N
F
F
N
F
F
N
Q
So
this
is
one
data
has
been
promising
why
we
TPD
quite
working
group
with
his
surprise
for
me.
So
this
is
a
story
of
the
of
the
drop.
Basically,
we
were
submitted
in
Eurasia.
We
adopted
no
document.
No
ghost
and
many
people
commented
on
the
lease
support
in
the
document
and
it's
a
quite
straightforward
extension
to
work.
So
no,
no
technical
changes
have
been
interesting
documents,
innocent,
very
innocent
missing.
We
have
some
internal
testing
on
this
at
Cisco.
E
Q
L
Just
to
clarify
you're
on
my
review,
there
are
very
strict
comments.
Iii
think
the
totally
editorial
okay,
so
there
is,
there
will
be
a
new
version
of
the
document,
but
technically
zero
changes
now
I'd
like
for
the
other
document,
we
can
test
today
to
vote
for
look
for
consensus
and
then
make
the
last
call
on
the
mailing
list
with
the
new
version
of
the
document.
Okay,
so
concerning
window
specific
karakov,
if
you
think
that
this
document
is
ready
so
to
be
published,
okay,
please
are
now.
L
I
Q
In
this
one
we
have
gone
through
a
few
different
changes,
so
it's
not
that
a
safe
harbor,
but
almost
we
submitted
in
August
last
year.
In
total
we
did
a
revision
with
some
comments
that
we
got
from
people
comment:
nothing,
nothing
major.
In
the
yesterday
we
will
be
send
this.
We
were
coming
from
ignominy,
known.
Q
P
L
F
Comment
for
support-
that's
seen.
A
free,
multicast
stuff
can
make
use
of
pops
up
as
well,
especially
when
we're
using
star
comma
G
trees,
because
what
we
do
now
is
we
notify
s
G
by
looking
up
the
s
in
the
mapping
system
to
find
the
ITRs.
But
if
who
we
need
to
notify
is
not
where
s
is,
and
so
that
I
think
Victor
that's
going
to
make
Starcom
G
work
better.
L
L
Q
A
lot
of
presentation
from
me
today,
okay,
so
this
one
is
a
bit
more
interesting,
this
a
draft
that
we
did
a
long
time
ago
before
the
departure
document,
and
now,
as
we
saw
it,
seems
that
there's
some
contention
on
the
working
group
to
move
forward
with.
What's
up
this
document,
MSI
you
need
to
this
mouse
is
the
commenting
what
this
implemented
right
now
in
open
daylight
in
an
open
source
controller
that
isn't
went
in
this,
an
issue
we'll
use
in
in
civilized
in
diplomas.
Q
So
the
question
now
is
we
we
want
to
pass
up,
but
this
is
being
used
actively
right
now.
So
this
is
what
it
has
been
in
running
code
and
in
production
for
some
time
now
and
it's
been
used
to
kind
of
have
a
absol
ike
support,
we've
been
Alexia's,
so
we
don't.
We
don't
think
that
this
will
move
to
the
working
group,
because
pops
up
is
what
I
focus
on
I
believe
but
sorry
handle
this,
because
he
is
documenting
this
running
code.
So
maybe
one
way
to
to
close
document
is
to
move
it
to
informational.
P
Just
one
comment:
the
the
point
of
this
drafted
that
it
doesn't
require
changes
to
XD
ours
as
defined
by
lease
/
FC
6830,
so
that
was
why
was
implemented
in
this
way.
I
don't
know
three
years
ago
and
the
documentation
was
done
because
there
is
a
public
implementation
of
a
daylight.
So
people
were
asking
a.
How
do
you
implement
this
feature
so
yeah
I?
Definitely
support
with
Alberto
is
saying
right:
we
we
want
to
move
forward.
Europe,
sub
and
future
implementation
will
deal
with
app
server
specified
in
the
other
draft.
P
L
L
Then
you
push
it
as
an
individual
submission
to
the
RFC
editor
and
the
editor
will
ask
us
the
the
chairs,
what
we
think
about
the
document
and
then
it
can
be
published
and
we
stay
around
as
this
with
the
reason
is,
there
is
implementation,
so
we
won't
just
mean
from
the
community.
There
is
this
implementation
in
this
way,
but
the
way
to
go
is.
H
So
there's
not
separate,
there's
two
different
things:
there's
informational
and
individual
as
informational,
there's,
some
technical
things
that
need
to
be
tweaked
and
put
in
the
ref
word,
the
other
pointer
and
the
working
with
could
publish
an
informational
or
document
that
clearly
says
this
is
not
the
preferred
solution,
but
it
is
the
more
compatible
one
so
for
transition.
You
can
do
this
for
individual.
The
working
room
has
no
saying
they'll
be
asked.
Is
this
a
competitor
and
as
long
as
it
says
explicitly,
this
is
the
not
preferred
option
but
is
published
for
information.
H
Q
H
Q
Is
the
last
one
for
me
I
promise,
so
this
is
the
most
interesting
one
out
of
this
document
that
we
published
for
the
idea
that
talks
about
how
you
can
use
the
least
control
plane
with
an
ila
plane,
as
you
will
now,
with
the
split
with
the
PCs
and
so
on.
Now
the
control
plane
they
deplane
are
separated
and
nothing
prevents
us.
Happily.
It's
been.
It's
been
used
for
the
data
planes
in
there
in
German
condition.
Nothing
prevents
us
from
using
the
least
under
play
with
any
data
plane
of
our
ship.
Q
So
this
is
not
introducing
any
modifications
in
either
the
release
for
the
ila
architecture,
but
rather
discussing
how
you
would
use
the
distance
of
painting
when
you
have
an
ila
later
plane
in
place.
You
don't
see
thumb
here,
but
it
would
be
good
if
we
can
there
before
we
finished
off
and
comment
as
well,
so
basically
and
I'm
not
working
actively
and
in
LA
besides
this
de
stuff.
So
so,
if
you
have
questions
on
ila,
you
can
refer
to
Tom
Paris
Teresa,
both
on
Thursday
evening.
You
can
attend
if
you
have
interest.
Q
So
for
those
of
you
that
I'm
not
aware
is
a
data
plane
and
the
follows:
they
ideal,
obviously
philosophy,
but
instead
of
doing
encapsulation
is
doing
other
stuff
formation.
So
you
take
an
ipv6
address
and
you
split
into
one
part
is
gonna,
be
the
locator.
The
other
parts
may
be
identified
well,
the
locator
may
change
the
file
is
gonna
is
gonna
remain
so
the
same
ideas
we
further
ideal
of
space,
but
this
time
use
maybe
BC
services
and
doing
everything
inside
the
single
activity
a
service.
E
Q
Q
Q
Again,
if
you
have
any
particular
question,
you
want
to
learn
more
about
la.
You
can
refer
to
took
him,
but
you
can
see
here
the
concept
of
keeping
the
identifying
aesthetic
and
then
so
in
Al
Andalus
consul
of
seer
a
standard
identify
representation,
which
is
what
endpoints
would
use
so
in
in
our
lives
world.
This
will
be
kind
of
in
the
area
space,
and
this
would
be
in
the
airlock
aspects,
so
devices
on
the
ad
space
will
be
used
in
the
desert.
I.
Think
again,
this
is
our
case.
In
LA.
Q
There
are
some
exceptions
on
different
use.
Cases
may
use
different
different
ways
of
talking
about
the
energy
use
case
and
then
when
they,
when
the
packet
has
to
leave
the
space
and
traverse
that
workers
face
the
surfer,
they
would
be
replaced
with
the
locator
part
that
this
row
table
on
the
on
the
undulation
and
so
on.
Well,
a
Q&A
in
annihilators
of
the
legatus
other
story.
Q
This
is
a
topology
that
Tom
used
to
represent
the
de
la
Sol.
There
are
components
that
we
could
try
to
map,
to
least
in
diversity,
which
basically
you
have
the
Daniel,
a
note
that
is
where
they
host
at
us
to
use
the
earlier
horse
is
running
an
acquaintance
of
knows
but
itself
within
itself.
So
let's
say
that
this
would
be
our
being
at
the
center
has
some
containers
that
are
on
the
S
phase
and
before
they
led
the
horse,
they
are
really
versatile,
employee
lane
traffic.
Q
They
a
lead
to
a
locus
which,
if
no
one
has
done
that
before
in
the
past,
so
now
talking
about
of
the
roles
of
the
different
these
devices.
In
this
scenario
with
what
you
have
both
elegant
list,
this
is
using
some
of
the
ila
terminology,
but
he's
not.
The
proportion
of
the
finally
made
sense
of
unleased,
so
anima
somatic
cell
is
MSM
are
nothing
nothing
new
that
needs
to
be
done
for
for
ila,
but
basically
to
put
things
in
context
and
where
you
have
both
things
which
device
as
well.
Q
So
the
missamma,
as
you
can
imagine,
is
restoring
the
the
mappings
they
identified
rockateur
mappings
for
the
domain.
Oh
that's
in
a
later
safer
for
the
south,
which
is
basically
have
subsets
or
USB
the
the
mapping
sector
of
different
devices
and
then
is
supposed
to
have
a
complete
map
pass
for
all
the
mappings
in
the
domain
and
in
the
case
that
you
use
in.
Q
You
can
say
that,
if
ascribe
to
the
MST
map
to
get
all
those
mappings
or
to
complete
your
map
cache
or
they
mr.
Mallen
la
route,
there
are
located
so
they
pass
access
to
to
a
limitless
another
asset,
and
this
is
something
from
from
from
ila
in
general.
Regulation
fleet
is
that
it
announces
or
they
announced
the
prefix
into
the
underlay.
Q
So
if
I
lean
now
is
not
able
to
translate
a
packet
into
into
violet,
it
can
just
for
were
the
sir
packet
into
the
under
lid
and
it
will
be
attracted
to
an
area
route
that
will
it
serve
in
a
study
for
path
for
the
debacles
ever
say.
So,
if
you
don't
have
a
mapping,
you
can
always
for
whether
they
not
just
like
the
packet
into
the
underlay
and
I,
then
they
know
for
the
proposal
is
laughs.
We
know
that
has
an
incomplete
mouth
cast
and
abilities
on
the
man.
Q
The
summarize
not
much
but
just
a
clear
picture
daily
note
is
supposed
to
be
the
one
having
the
10
points
and
for
this
now
we
use
I
only
know
or
I'll
a
horse
simple
suitable,
so
we
say
I
Ellen.
No,
so
it
has
a
much
not
much
for
us
if
the
end
points
are
directly
within
the
I'll,
a
node
or
a
attach
outside,
and
we
assume
that
the
eyelid
other
will
not
have
end
points.
But
again,
this
is
just
for
for
I,
find
explanation
and
is
not
a
constant
architecture
or
other
protocols.
E
Q
With
inattentive
would
have
three
subtypes
to
go
for
free
service
and
the
reason
we
use
split
seat
aircraft
type,
for
this
is
to
be
explicit
on
the
length
because
in
in
LA,
is
not
identified
on
locator
length
is
not
fixed,
so
usually
you
have
SP
by
the
middle,
so
64
64,
but
that's
not
that's,
not
a
requirement
of
the
connector,
so
you
can
have
different
lands
in
the
imitation
they
on
the
right.
I
put
an
identified
on
a
locator
like
in
the
most
common
for,
but
the
length
again
is
no
fix.
Q
So
you
don't
need
to
to
look
at
the
at
the
bits
here
because
it's
just
to
represent
this
particular
example.
But
this
can
be
any
number
of
bits
same
here,
so
we
use
a
specific
app
attached
to
the
resident
and
also
to
give
this
explicit
meaning
on
the
mapping
system.
So
if
you
have
this
circle
scenario,
you
know
that
this
is
this
to
be
use
in
devices
and
also
to
add
some
metadata
read
that
on
the
on
the
drug
we
use
both.
E
Q
Directive,
sigh
I'm
in
the
in
the
locator
site.
One
comment
that
we
got
under
this,
for
instance,
was
that
we
could
use
I
just
abused
for
this,
which
I
completely
agree.
The
question,
then,
is:
you
may
run
along
the
way
into
some
Travis
if
you
start
having
deployments
where
you
may
have
different
look
at
the
plans
or
when
I
lay,
maybe
only
part
of
the
first
act,
but
you
have
some
other
data
planes
and
we
are
using
the
dis
control
plane
to
amal
or
if
you
have
the
zombies
that
you
require
and
lemon
needs.
Q
So
you
can
not
allocate
those
him
directly
in
ipv6,
also
in
using
ipv6
other
set
without
any
ELCA.
Then
we
need
to
discuss
what
do
we
return
in
the
locator,
because
the
locator
is
only
the
upper
part,
so
you
can.
You
can
send
the
lesser
address
to
the
mapping
system
and
then
return
the
locator
we
with
the
either
identifier
or
there
are
other
options
ICD,
no
there,
because
he
already
implemented
this
on
his
map.
Insistent.
You
want
to
comment
on
this
part
in
particular.
Okay,
so.
F
You're,
saying
being
explicit
that
it's
an
Iowa
address
is
important
because
you
may
have
multiple
uses
an
ipv6
address,
but
I
would
say:
annihilate
annihilate
system
is
going
to
look
up
something
and
probably
be
configured
with
a
seraph
Excel.
It's
going
to
know
what
it
gets
back
is
going
to
match
that
there's
a
good
chance
that
a
lisp
xtr
may
want
to
look
up
any
either
a
so
prefix
or
an
ipv6
to
get
a
locator,
because
maybe
ila
is
running
over
lisp
overlay
as
well.
F
So
I
think,
depending
on
who
does
the
lookup,
they
know
how
they're
going
to
use
the
information
and
it
comes
from
and
the
lookup
comes
from
different
sources.
So
I
don't
know
if
you
need
to
distinguish
it
unless
there's
explicit
ila
functionality
that
has
to
be
conveyed
with
the
address
and
and
that's
what
I
thought.
Your
comment
on
the
mailing
list
was:
is
that
there's
specific
information
but
I,
don't
know
what
it
is.
Q
Okay,
first
I
do
agree
with
what
you're
saying
so
I'm,
not
against
you
see
in
ipv6
here.
My
point
was:
if
we
start
deploying
this
and
we
start
a
disclosure
of
an
equally
as
it
goes
along
the
line
we
may
stir
seen
places
what
it's
not
clear,
and
even
even
if
someone
is
asking-
and
you
know
who
is
asking-
maybe
those
guys
are
seven
different
deployments
at
same
time.
So
that
could
be
a
reason
for
concern
on
the
road
and
regarding
the
explicit
ila
metadata.
Let's
say
in
the
draft.
Q
We
have
two
by
two
bits
that
others
may
it
may
appear
is
to
distinguish
identify
when
you
are
doing
ideal
or
come
up
in
the
solution
on
when
you
are
trying
to
to
resolve
the
endpoint
address,
to
identify,
because
there
are
some
cases
in
ala
where
the
others
that
you
use
the
Senate
and
I'm
sure
that
you
are
aware
of
that,
the
others
that
we
use
as
an
identifier
through
the
network.
It's
not
the
address
that
the
endpoint
use.
Q
So,
in
this
case,
in
this
case
the
endpoint
is
using
the
standard
identifier,
presentation,
others
mm-hmm,
and
this
is
the
one
that
ila
understand.
Second
Thursday.
In
other
cases,
you
may
have
different
others
and
then
need
to
be
mapped
in
some
sort
of
these.
So
this
kind
of
a
twisted
process
that
you
can
combine
into
one
and
to
distinguish
those
lookups.
You
use
just
one
example
and
in
the
locator
and
every
day
we
use
there
is
a
zoom
in
ila-
is
doing
some
smart
League
regarding
the
checksum.
Q
So
the
remote
STR
contained
the
other
guy
look
I,
don't
need
you
to
to
real
computer
through
the
adjustment
of
the
checksum,
because
I
know
that
in
the
underlay
the
no
one's
gonna
drop,
this
packet
and
I
am
being
able
to
understand
it
right.
So
you
may
prefer
to
say
that
guys,
don't
waste
your
resources
on
doing
their
time,
because
they.
Q
Q
L
I
didn't
lead
yet
the
death,
but
I
mean
you
seem
cool,
could
be
documented
and
with
discussion
like
you
are
doing
that
says,
look
if
you
are
in
some
scenarios.
Ipv6
is
okay.
It's
enough!
You
don't
need
more,
but
you
need
more
for
a
list
of
reason
or
because
you
had
a
certain
class
of
scenarios.
Jerry
it's
what
you
can
do
with
al
Kahf.
F
Agree
with
that,
because
if
you
give
too
many
options
to
the
implementer
they're
going
to
wonder
why
it's
needed
and
if
they
don't
do
this
special
use
case,
which
is
not
that
you
have
to
find
they're
gonna
say,
do
I
need
it
and
then
you're
gonna
have
interoperability
problems,
because
some
vendors
are
gonna
support
only
ipv6,
so
we're
going
to
support
the
other
ones.
Go
support
both
and
we're
gonna
have
to
add
another
encapsulation
type
to
the.
F
E
F
F
Q
F
F
So
an
ipv6
doesn't
ever
check
some
minutes.
Header,
so
hop
by
hop,
doesn't
mean
you
have
to
consider
this
as
an
option.
It's
only
it's
only
before
you
hand
the
packet
to
tcp
on
the
desk
or
and
to
the
transport
layer
on
the
destination.
That's
all
right.
So
it's
a
pseudo
header
checksum
for
the
transport
layer.
That's
all.
Q
Q
As
I
was
saying
before,
we
see
two
places
where,
where
this
can
be
helped
can
be,
can
be
used
and
held
in
the
error
solution.
The
basic
one
is
when
you
need
to
resolve
identifiers,
so
you
need
to
find
out
either
for
identifiers
and
that's
the
on
the
left
and
then
our
case
that
that
we
see
is
that
sometimes
were
saying
disk
is
melee
where
dude
on
me.
Well,
you
don't
have
a
directly.
Q
They
they
identify
that
they're
gonna
use
to
yourself
and
to
get
the
rocket,
so
you
may
need
to
do
a
prayer
translation
on
the
further
identified.
So
in
those
cases
you
can
do
this
as
well
to
ask
for
further
in
the
fire
and
I
wonder
what
you
can
get
the
locator.
So
you
either
do
this
as
a
twisted
process,
or
you
collide
everything
into
a
single
step.
Q
Then
how
the
signaling
looks
like
when
using
the
the
daily
data
play
and
I
need
staff
here
too,
this
is
the
standard
list.
The
next
one
is
something
that
we
have
is
passing
on
on
the
list.
That
is
how
I,
in
a
also
considers
doing
things,
but
basically,
how
would
you
do?
Is
you
have
an
island
now
that
is
same
person
traffic
to
another
island?
No,
you
need
to
resolve
the
mapping
for
that.
So
you
see
no
more
requests.
Q
You
get
a
my
reply
now,
meanwhile,
and
that's
why
here
no
can
send
the
traffic
to
the
underlay
they're
out.
There
is
not
seen
the
Senecas,
they
share
profits,
so
it
can
track
the
traffic
and
translate
it
while
the
island
knows
on
the
left
is
retrieving
the
map.
Another
notion
is
to
present
here
to
drop
it
as
we
so
different
options
on
how
to
do
wide
visa
kill
yourself.
Now,
why
is
the
napping?
Is
some
daily
know
that
it
can
direct
itself
and
carefully
there?
The
traffic
from
one
area
now
to
to
the
other?
Okay.
Q
Nothing
to
do
with
each
other,
besides
that
they
are
now
in
this
case,
which
is
something
the
thumb
is
the
specimen
on
the
list
with
us
is
that
do
do
prefer
that
instead
of
resolving
the
the
map
into
there
and
not
by,
but
you
directly
send
the
traffic
to
the
Underland,
so
the
nav
route
router
will
catch
it
and
then,
when
it
sees
a
pocket
that
is
not
being
trusted,
it
sends
a
message
back
to
the
LM.
Not
in
this.
This
message
will
be
modified
and
again
this
PC.
L
Q
D
E
L
F
C
F
Q
Is
the
UDP
participe,
as
you
guys
know,
leased
by
default
goes
over
UDP,
but
nothing
to
run
it
over
TCP,
and
actually
there
is
a
split
in
the
specification
on
reliable
transfer
for
lease
that,
if
you
have
had
about
just
in
place,
you
can
leverage
some
features
that
we
found
on
that
rock.
The
another
point
is
is
what
we
were
discussing
a
bit
before
on
the
I/o
rather
I'm
after
work
location.
Q
Options
for
this
custom
is
that
you
can
post
to
destination.
Ila
note
the
mapping.
When
you
see
someone
request
in
the
mapping
ourselves,
I
mean
the
the
source
mapping,
so
there's
available
when
the
difficult
in
the
same
map
server
and
under
the
place
to
the
discussion
would
have
no
way
of
the
share,
prefix
and
so
on
and
underneath
as
well.
If
you
want
to
have
this
locator
is
of
faces
man
that
I
chose.
Q
R
Sorry
I'm,
like
at
a
question
a
couple
questions
so
TCP
versus
UDP,
mm-hmm
I'm
not
entirely
familiar
with
with
how
that
works
in
a
list,
but
I
have
some
ideas
for
ila
when
you're
using
UDP
is
there
some
sort
of
state
that
says
which
map
servers?
You
might
send
you
a
map,
notify
or
would
you
it
would
you
accept
I'm
not
notified
for
many
maps,
or
there
must
be
some
security
there
right.
So
there's
a
50
CPI
would
just
have
connections
open
up
to
whichever
map-servers
I
wish
to
use
right
right.
R
F
Q
R
Q
R
So
so,
like
do,
you
know,
was
saying
FRA
la
if
we
go
with
a
like
a
lot
of,
we
have
a
lot
of
redirects,
it
does
become
the
issue
of
security
and
having
that
TCP
connection
gives
us
that
much
benefit
the
other
part
about
TCP.
That
I,
suggested
and
I
know
there
may
be
some
implementation
unless
but
using
something
like
rest
and
just
upping.
R
The
whole
control
plane
look
more
like
a
data
center
type
of
API
controlled
protocol,
as
opposed
to
a
low-level
routing
protocol,
Smita
sings
to
to
think
about
no
I,
don't
think
any
of
that
affects
the
primary
packet
format,
or
things
like
that.
But
I
am
concerned
about
the
security,
especially
of
UDP.
I
will
say
that
no.
Q
F
Q
P
The
point
is
that
it
will
certainly
help
adoption
right
in
the
sense
that,
if
you
are
the
arrest
API
for
the
mapping
system,
I
guess
that,
for
the
typical
developer
of
ILA
application
it
will
be
user
easily
to
to
use
right.
I
mean
they
will
be
drawn
to
probably
use
the
REST
API
rather
than
having
to
deal
with
yeah.
R
R
Q
P
P
He
was
saying
I
think
that
would
be
I.
Think
that
is
emerging
from
the
conversation
in
the
DMM
working
group.
With
regard
to
support
for
mobility.
Is
that-
and
there
was
one
of
the
goal
of
writing
these-
this
draft
right
so
I
think,
and
it
was
clear,
I
think
inside
the
working
room,
because
we
did
the
separation
of
68
1368
30
beats
exactly
with
the
intention
of
having
a
control,
plane
and
values
at
a
plane.
So
what
I
am
observing?
P
Is
there
sir
I
think
now
times
are
more
mature
and
and
people
even
I'm,
working
at
least,
but
to
play
the
clock
right
and
EMM.
The
DMM
work
for
3gpp
I
think
it's
a
great
place,
because
it's
basically
showing
that
you
can
have
at
least
control
playing
that
that
can
support
mobility
for
ila
one
example,
and
that
you
know
we
just
speak
a
lady
because
you
know
we
like
Tom
and
is
very
vocal
in
supporting
its
I
is
a
very
good
source
for
learning
about
it.
P
N
F
E
T
Hi
I
had
a
comment
about
you
know
your
being
back
on.
What
can
you
say
about
the
placement
of
the
mapping
system
in
the
5g
I
will
encourage
you
guys
to
up.
There
is
a
document
that
we
have
published
in
another
SDO,
which
actually
explained
very
clearly
all
their
interactions
between
the
Mme
and
older
5g
architecture,
and
actually
I
would
like
to
discuss
with
you
guys
and
actually
make
sure
that
we
ensure
that
all
this
work
can
be
leveraged.
Q
R
Someone
comment
on
the
applicability
I
agree
that
it
would
be
nice
to
have
one
mapping
system
that
could
cover
all
use
cases.
One
thing
that
I
would
point
out
if,
for
instance,
you're
using
a
mapping
system
within
a
closed
network
versus
using
a
mapping
system
and
a
public
mobile
network
privacy.
Security
in
our
service
are
very,
very
different
and
I.
Think
I
raised
that
issue
on
a
list,
especially
denial
of
service.
R
That's
going
to
be
a
tricky
one,
so
we
know
that
whenever
you
have
a
cache,
for
instance,
so
we
know
we
need
to
cache
to
scale
in
most
of
these
mapping
systems.
That's
going
to
be
a
target
of
denial
of
service
attack
at
some
point.
So
these
are
things
that
I
think
we
just
have
to
consider
for
the
broad
use
cases
I'm,
not
saying
it
anything
specific
to
ila
in
this
case,
but
definitely
want
to
see
some
really
solid
explanations
for
why?
Then,
our
service
isn't
mitigated,
for
instance,
yeah.
Q
M
This
is
one
when
we
got
rhythm
denial,
denial
of
service
attack
to
the
control
pain,
we're
working
on.
We
have
work
in
the
past
on
solution
for
that
we're
planning
to
to
release
a
document
explaining
how
you
can
solve
that,
and
you
will
see
all
the
details.
But
to
me
the
main
point
is
that
this
kind
of
infrastructure,
where
you
can
sakura-chan
a
control,
plane
they're
very
common
in
NT,
and
there
are
main
stations
around.
So
it's
a
very,
very
important
with
many
well-known
solutions.
F
Since
you
might
subscribe
to
a
lot
of
different
ideas
that
are
spread
across
thousands
of
map
servers,
they
don't
need
to
send
you
a
notification
itself.
Something
changes
at
that
time
that
the
TCP
connection
is
established.
You
have
a
three-way
handshake
delay
that
will
cause
convergence
problems
and
you
can't
do
the
same
priority,
because
that
means
every
map
server.
We
have
that
with
TCP
connections,
everybody
together
in
the
Internet.
This
is
not
a
scalable
solution,
even
though
the
big
data
centers
of
the
world
have
thousands
of
TCP
connections.
This
is
over
what
orders
of
magnitude.
B
N
Okay,
so
this
is
just
an
update
on
the
ground-based
live
document
which
we
presented
in
las
ITF
in
singapore.
The
drawers
renamed
by
the
way,
so
just
a
reminder
about
the
where
this
work
is
coming
from
the
ICAO
is
proposing
to
replace
the
existing
services
with
a
next-generation,
ipv6
based
infrastructure
for
air
traffic
management.
N
Okay,
sorry
so
first
point
was
the:
there
was
some
evaluation
done
by
a
third
party
who
were
comparing
a
list
based
solution
using
open
overlay,
router
and
the
Eroica
solution.
The
main
difference
which
they
found
was
that
there's
initial
packet
loss
with
less,
but
what
Alberta
Argos
was
referring
to.
When
you
don't
have
your
map
cache
and
there
was
none
with
Europe,
so
that
was
one
of
the
major
major
things
at
the
last
ICAO
key
meeting
and
also
the
co-authors
of
the
draft
from
frequent
is
they
presented
the
solution.
So
we.
P
E
N
N
The
next
ICO
meeting
is
going
to
be
in
May
we're
going
to
give
feedback
the
feedback
that
Fabio
has
GSK
described
about
how
the
initial
packet
loss
can
be
avoided,
using
an
RTR
for
example,
and
then
the
feedback
the
frequencies
co-authors
got.
Was
they
really
want
to
understand
how
this
supports
multicast
and
somebody
named
TBD
is
going
to
be
presenting
there
to
see
how
that
actually
works
and
we'd
also
like
to
get
some
comments
from
the
lisp
or
group.
So
just
a
reminder
that
draft
itself
is
a
it
doesn't.
L
F
L
K
L
Just
for
the
plug,
so
in
this
is
not
a
public
network.
It's
like
very
me,
control
traffic
or
something
closed,
and
so
the
point
is,
can
we
have
a
dos
attack
coming
from
the
outside?
That's
the
little
point
that,
because
why
we
drop
so
is
one
reason,
is
not
to
fill
the
buffer
of
the
queuing
of
there
now
if
this
cannot
happen
so
that
you've
really
safe
to
it
too.
So.
N
As
I'm
going
to
be
a
clue,
suki
network
but
Fabian,
remember
the
details
more
than
I
do
but
there's
also
it's
not
just
one
okay
network.
There
is
maybe
the
North
American
power,
there's
a
European
part,
and
how
do
you
connect
those
get
together
and
those
kind
of
things?
So
it's
not
just
a
simple
slab
networks,
but
the
security
thing
has
definitely
come
up
with
few
times
in
discussions.
J
U
Viktor
Murano,
so
just
to
build
on
on
the
conversation,
that's
going
on
in
the
room,
I'll
encourage
the
group
to
to
really
have
a
read.
It's
it's
an
easy
read
and
it's
a
lot
of
food
for
thought.
You
will
find
implications
also
on
the
applicability
of
the
pub
sub
that
has
been
proposed.
You
will
find
implications
in
this
multi
regional
network
or
mobility
thought
on
how
you
would
do
that
best
or
some
proposal
on
the
draft,
but
it's
not
necessarily
immediately
using
things
like
DDT.
U
L
L
V
V
The
airplane
is
a
mobile
internet
of
things,
a
BGP
based
routing
system
in
a
hub
and
spokes
arrangement,
with
a
core
autonomous
system
in
the
middle
and
hub
autonomous
systems
out
at
the
datalink
service
provider,
sub
networks,
mobility
of
traffic
engineering
and
the
stubs
stable
routes
in
the
core,
and
it's
multi
link
capable.
So
where
you
see
the
airplanes
having
multiple
links
into
the
aviation
data
like
sub
networks,
that's
the
multi
link
aspect
of
it
next
chart.
V
I'm
still
waiting
to
see
the
next
chart
there,
it
is
so
in
the
edges.
We
have
what's
known
as
the
Aero
ipv6
mobile
networking
services.
Again,
the
airplane
is
a
mobile
ipv6
network.
I
could
have
millions
of
addressable
Internet
of
Things
entities
onboard
and
it
can
connect
to
multiple
data
like
simultaneously
like
SATCOM,
L,
Dax,
AeroMax,
4G,
5g
and
others.
We
have
traffic
engineering
for
inbound
and
outbound
datalink
selection.
V
L
V
And
the
the
main
architectural
change
since
the
last
version
that
I
I
gave
in
in
Prague
is
that
we
developed
a
new
architectural,
a
construct
called
the
arrow
proxy.
The
arrow
proxy
sits
at
the
datalink
subnetwork
border
router,
and
it
acts
the
same
as
for
an
enterprise
network
web
proxy.
So
inside
the
datalink
sub
Network,
the
client
airplane
interacts
with
the
proxy
in
the
same
way
that
it
wouldn't
interact
with
the
server
in
the
outside
world
outside
the
sub
Network.
V
V
So
that's
that's
the
main
architectural
change
and
it's
it's.
It's
a
big
component
of
the
architecture
now
the
diera
proxy
and
and
that's
all
I
have
for
presentation
materials
as
as
Rashard
just
mentioned.
We
did
have
this
discussion
that
the
the
last
ICAO
meeting
and
there's
going
to
be
another
meeting
coming
up
in
May
yeah
in
in
Montreal
to
talk
more
about
the
arrow
and
the
Lisp
approaches.
B
B
B
F
E
F
I
think
you
guys
will
so
Colin
and
myself
got
together
and
said:
how
can
we
bring
applications?
These
centralized
applications
together
with
a
network
layer
that
can
be
decentralized
as
well,
and
that's
the
basis
of
this
talk
we're
going
to
talk
about
how
we
can
be
centralized
the
mapping
system,
and
it
turns
out
that
we
have
the
ideas
are
relatively
simple
and
most
people
will
get
it
and
we
didn't
have
to
make
any
protocol
changes.
You
have
to
make
some
implementation
changes
depending
on
what
set
of
protocols
you
use.
F
Okay,
so
the
problem
statement
is
what,
if
lists
best,
X
TRS
didn't
have
to
depend
on
a
third
party,
in
other
words,
go
out
and
talk
to
map
servers
in
a
mapping
system.
What
if
this
affects
the
hours,
could
multi-home
in
Rome
to
inform
each
other
about?
There
are
look
changes
rather
than
being
notified
by
the
mapping
system
and
what,
if
xt
hours,
could
be
their
own
mapping
system?
F
F
So
what
can
happen,
though,
is
that
maybe
these
sites
have
better
connectivity
and
usually
it
is
via
wireless,
and
so
maybe
the
south
and
the
east
on
the
same
Wi-Fi
network,
or
they
can
communicate
through
local
towers
because
they're
geographically
close
or
have
some
reference
of
locality.
But
what
happens
if
we
have
a
situation
where
they
really
don't
have
internet
access?
So
how
could
they
keep
roaming?
F
So,
let's
see
if
we
could
decentralize
the
map
server?
Okay,
what
if
each
xti
was
a
map
server
themselves
and
what,
if
each
xtr
could
map
register
to
each
xtr?
The
mapping
system
would
then
always
be
synchronized,
because
they'd
be
all
telling
each
other
all
of
their
mappings,
and
then
the
xtr
could
be
a
map
resolver
for
itself.
Since
everybody
has
synchronized
mappings,
you
don't
have
to
go
to
an
external
map
resolve
you
just
have
to
basically
send
a
map
request
to
yourself
or
do
something
internally.
If
you
want
to
make
it
faster.
F
So,
let's
first
define
a
mapping
system
and
what
a
consolidated
or
decentralized
mapping
system
could
look
like
a
consolidated
mapping
system
is
identified
by
a
multicast
group.
Let's
say,
let's
just
make
that
a
fundamental
definition
and
then
the
the
XT
hours
that
are
part
of
the
same
mapping
system
will
join
the
same.
Multicast
group
there'll
be
receivers
of
the
group
because
they
join
it
and
they'll
also
send
to
the
group.
So
it's
not
all
one-to-many,
it's
a
many
from
any
sort
of
multicast
relationship.
F
The
map
registers
are
then
sent
to
the
group
in
that.
Therefore,
all
xt
hours
that
are
joining
to
that
group
will
receive
the
mappings,
and
then
we
get
efficient
distribution
of
these
map
registers,
because
if
the
underlay
supports
multicast,
the
XT
I
was
only
sending
out
one
map
register
and
the
underlay
is
replicating
or
kidding
to
all
the
places.
If
the
underlay
doesn't
support
multicast,
we
could
do
head
end
replication
by
using
signal
free,
ignore
thei
cast.
F
So
what
happens
there
is
that
the
map
register
from
the
control
plane
would
be
sent
down
and
the
data
plane
would
know
if
there's
underlay,
multicast
or
not,
and
when
it's
not
it
would
do
an
SG
lookup.
According
to
the
signal
free
spec
to
find
out
who
all
the
etrs
are
and
it
would
head
and
replicate
the
whole
okay.
This.
H
Seems
to
be
a
slight
difference
between
what's
in
a
draft
and
what
you
just
said,
and
that
this
seems
to
say
there
is
a
multicast
group
that
is
all
the
XT.
Ours
are
just
very
analyst
mapping
system.
Yes,
the
draft
seems
to
talk
about
the
XT
RS,
forming
peer
groups
and
for
a
multicast
among
those
peer
groups,
and
this
is
the
peer
group-
is
all
the
X
TRS.
Okay,.
H
F
F
So
here
we
have
three
sites
that
are,
on
the
left
hand,
side
that
are
part
of
the
same
mapping
system
there,
one
three
and
five
and
the
registering
there
eid-prefixes
based
they
are
all
join
to
multicast
group
224,
one
one
one:
they
send
their
map
registers
2
to
4
to
24
for
401
image
mapping
in
the
mapping
system.
The
right-hand
side
is
just
another
set
of
XT
I
was
doing
ipv6
that
are
building
their
own
mapping
system.
Okay,
so
that's
fundamentally
how
it
works.
So
in
this
case
mapping
it
to
the
spec.
F
We
would
say
that
1,
3
and
5
are
all
part
of
one
peer
group.
If
that
wasn't
the
case-
and
we
wanted
one
in
3
to
be
only
in
part
of
that
peer
group
and
5
would
still
have
connectivity,
the
XTS,
1
and
3,
then
all
we
would
have
to
do
is
send
map
request
messages
to
the
multicast
group
and
when
you
send
man
3
quest
messages
to
the
multicast
group,
what
happens
is
1
&
3
would
receive
the
request
and
they
both
would
respond.
F
So
there's
ways
Joel
that
you
can
not
be
part
of
it
and
have
the
peer
group
just
be
a
set
of
map
servers.
So
we
specify
in
the
spec
that
a
map
because
can
send
to
a
multicast
group
what
we
haven't
specified
is
n
messages
come
back.
Is
that
a
good
bad
thing?
Can
we
scale
it
better
because
there's
no
election
mechanism
between
1
&
2
me
to
say
I'll
answer
the
map
request
or
the
other
one.
So.
U
F
F
The
the
problem
is,
if
you
are
in
a
situation
where
you
have
the
internet
and
you're
using
it.
What
we
wanted
to
work
for
also
is
in
neighborhoods
that
maybe
you
can
all
your
connectivity
is
lost
to
the
system,
and
so
you
need
to
auto-discovery
each
other,
because
XT
hours
are
going
to
be
added
and
deleted,
and
so
by
using
any
unicast
address.
There's
no
way
you
can
auto
discover
things,
don't
forget,
normally
how
we
auto
discover
things
is
using
multicast
later
use
directory
services.
F
Well,
if
you
use
directory
services
you're
depending
on
the
third
party
right.
So
if
I
am
always
configured
that
you're
my
next-door
neighbor
and
along
he's
going
to
use
you
as
a
map
server,
then
we
can
use
our
and
then
Padma
has
another
neighbor.
That
is
a
map
server
and
we
use
the
exact
same
IP
address
and
therefore
it's
anycast
I'll
get
you
a
key,
we'll
get
your
neighbor
right,
and
so
we
could
do
that.
But
but
that
means
you'd
have
to
reconfigure,
coordinate.
E
F
We
want
this
to
work
just
like
a
cryptocurrency
where
work
of
a
peer-to-peer
network
where
everything
is
pretty
self
discovering,
like
mesh
networks,
that's
the
whole
point,
we're
basically
buying
the
applications
from
the
cryptocurrency
world
and
trying
to
build
it
here
now.
It
kind
of
looks
like
a
routing
protocol,
but
it's
really
not
because
it's
done
over
the
top
for
just
multi
casting
things
with
efficient
distribution
right.
R
F
R
F
R
F
F
F
H
But
you
start
piling
on
contacts.
I
can
understand
this
in
a
small
case
of
a
moderately
small
set
of
things
which
want
to
use
lists
which
are
all
on
a
common
fabric
which
supports
multicast
so
that
all
the
scaling
properties
working
all
the
benefits
work,
but
as
a
general
mechanism.
This
seems
like
it's
going
to
need
so
many
complexities
by
the
time
you're
done
for
the
general
deployment
case.
Well,.
F
L
L
L
F
If
you
pick,
if
you
always
pick
somebody,
then
you
might
as
well
just
statically
configure
all
the
are
looks
and
the
map
cache
entries
and
don't
even
run
a
mapping
system,
and
that
can
be
possible
too.
But
if
you
configure
static
mappings
with
our
looks
and
they
want
to
move
around
and
change,
there
are
loops.
You
have
a
problem,
so
that's
why
you
need
multicast
for
the
dynamic
discovery.
F
H
E
F
F
You're,
allowing
what
XT
hours
are
allowed
to
your
you're,
basically
not
creating
an
input
filter
on
saying
I'm
not
going
to
receive
something
from
that
neighbor
over
there.
Even
though
I
don't
know
if
it's
Joel
or
Luigi
right,
the
thing
is,
is
you
will
register
by
signing
the
map
register
and
then
I
will
look
up
in
our
database.
The
public
key
and
I'll
verify
the
signature
to
make
sure
that
it's
Joel
that's
allowed
to
join
this
particular
multicast
group
in
this
instance
ID.
This.
L
F
F
Complex
understand
so
in
this
case,
there's
no
third
party
trust
or
dependency
that
exists
other
than
and
each
other
and
the
map-request
is
the
map-request
lookup
is
pretty
low.
Lane
sees
because
it's
a
local
operation,
you
don't
have
to
send
it
to
a
third
party,
could
I
finish
a
slider
or
do
you
have.
M
A
question
yeah
it's
on
the
second
are:
these
are
better
than
the
second
bullet
point,
nor
throw
party
the
way,
I
understand
it,
and
maybe
I'm
wrong.
Is
that
actually
you
have
full
trust
right?
You
are
trusting
all
the
participants
on
your
multicast
group,
there's
no
way
to
verify
what
they
are
telling
you
all.
They
are
right
or
not.
Well,.
M
F
M
F
So
let
me
explain
how
so
our
role
in
a
neighborhood
we
lost
Internet
connectivity.
We
want
to
talk
on
our
radios
to
each
other,
where
in
some
signal
range,
so
we
know
that
we
can
get
packets
to
each
other.
So
what
happens?
Is
we
all
register
our
public
keys
to
this
multicast
group
address,
and
so
they
what
we?
What
the?
What
we're
registering
is
a
hash
of
a
public
key
to
the
mapping
system
with
public-key,
and
these
are
the
nets
and
that's
happening
just
without
any
signature.
This.
M
F
Create
key
pairs
up
front,
yeah
and
then
part
of
running
in
the
ECDSA
authentication
draft
that
Eric
and
I
published
you.
You
sign
it
using
the
private
key
or
some
signature
data
that
includes
the
hash
of
the
public
key
and
the
instance
ID,
and
then
that
map
register
goes
to
the
other
everybody
and
they
verify
it
by
looking
up
the
public
keys
that
they
have
local
in
their
database
because
it's
been
distributed.
P
F
P
F
P
F
E
F
You
guys
are
saying
that
if
I
was
registering
an
ipv4
address,
it
would
be
a
problem.
Okay,
okay,
XT
ours
then
build
and
send
one
map
register
for
NXT
is
so
that's
important
because
the
distribution
is
being
done
by
a
routing
layer
versus
building
separate
messages,
so
really
the
control
plane.
It
just
has
to
build
a
single
message.
Even
though
it's
going
to
many
places,
the
distribution
is
being
done
at
the
network
layer.
A
management
is
pretty
simplified
because
you
have
all
it
by
accessing
one
XTL.
F
You
can
get
all
mappings
very
similar
to
link
state
protocol,
so
you
don't
have
to
you
know
rest
into
all
the
so
because
today
the
mapping,
the
third
party
mapping
system-
is
a
shard
where
each
map
server
is
doing
horizontal
partitioning
of
the
of
the
mappings.
So
if
you
wanted
all
the
mappings
in
the
mapping
system
today,
you'd
have
to
go
access.
All
the
map
servers
I,
don't.
F
F
So
here
the
high
level
use
cases.
We
believe
that
this
can
go
with
cryptocurrencies,
where
maybe
mining
pools
could
be
associated
with
a
multicast
group
and-
and
we
want
the
miners
to
be
in
their
own
mapping
system
and
it's
closed
and
authenticated.
Maybe
we
want
certain
sets
of
wallets
to
talk
to
certain
miners
or
certain
peer-to-peer
networks,
so
we
think
we
could
do
all
this
without
relying
on
a
third
party.
Emergency
networking
is
the
use
case.
F
Among
those
things,
we
think
there's
going
to
be
some
use
cases
for
deploying
routers
and
stuff
in
space,
where
we
think
there
will
be
some
software-defined
satellite
type
applications
that
we
could
use
this
as
as
well,
and
it
just
kind
of
falls
in
with
the
general
shareable
economy
apps,
where
you
want
to
pay
for
renting
out
resources
to
people
either
bandwidth
CPU,
whatever
routing
frequency
or
whatever.
So
we
did
this
in
a
brief
demo,
we
used
three
thousand
containers,
Joely
Joely.
H
F
Xt
is
register
okay,
so
the
demo,
where
the
demos
doing
is
they
the
XT
arthritis,
tur,
ipv4,
eid-prefixes
and
a
name
as
the
div
okay.
So
here's
an
example
of
the
first
container
and
n1
does
one
on
1/32
and
to
dusty,
so
you
see
they're
all
sitting
there,
so
this
is
showing
the
result.
Okay
and
I'll
explain
how
the
result
is
achieved.
This
is
the
second
one
that
obviously
has
the
same
set
of
entries,
and
then
this
is
the
third
one.
F
Well,
what
happens
to
twenty
four
one,
one
one,
because
that
was
the
only
thing
that
was
configured
was
the
multicast
group
and
so
what
they
did
is
they
started
sending
their
map
registers
to
224
one
one,
one
now
how
you
would
strap
it
is
you
first
send
it
to
yourself
and
then
that
gets
added
to
this
and
then
once
that
happens,
all
this
guy
needs
to
know
is
about
one
other
guy
and
that's.
We
have.
F
You
need
you're,
saying,
there's
a
circular
dependency
yeah.
He
got
out
of
the
circular
dependency
I
believe
you
did
yes,
but
I
can't
guess
how
so
the
way
I
did.
It
first
was
I
first,
so
the
map
was
configured
to
be
224,
101,
27001
rawhide
kay.
So
we
sent
the
register
to
ourselves
and
we
got
the
first
and
then
I
can't
remember
the
configuration
because
I
evolved
a
couple
times.
I'm.
H
F
We
were
trying
to
show
where
we
were
using
the
list
beta
plane
to
solve
the
problem
as
well,
but
but
yeah.
Clearly,
if
two
24:01,
when
one
was
available
on
the
on
the
local
area
network,
which
it
can
be,
you
know
a
lot
of
times-
these
radios
were
in
layer
2.
So
maybe
in
the
emergency
situation,
you
just
defend
them,
but
you
don't
want
to
really
necessarily
depend
on
the
layer
to
overlay
that.