►
From YouTube: IETF100-LISP-20171117-0930
Description
LISP meeting session at IETF100
2017/11/17 0930
https://datatracker.ietf.org/meeting/100/proceedings/
A
Okay,
good
morning,
everyone,
let's
start,
we
have
a
busy
schedule
today.
Okay,
welcome
to
the
least
working
group,
I'm
Luigi
denied
by
the
oil.
We
are
the
coaches
of
this
working
group,
the
first
order
we
have
admah.
Our
secretary
will
take
notes
today,
Waseem
Thank,
You
Padma.
What's
him
could
make
it
this
time,
a
bet
that
he
had
to
leave
yesterday
for
other
things,
so
as
usual,
they're
not
well
stating
that
whatever
you
say
today
here
is
a
contribution
today,
ietf
the
usual
pointers.
Blue
sheets
are
running.
A
We
have
links
to
the
charter,
audio
stream,
jabber
room
agenda,
slides
quick
status
updates.
So
we
will.
We
had
the
least
security
draft
that
was
in
ITF.
Last
call
came
back
for
the
simple
reason
we
are
chartered
to
do:
standard
track.
Work,
SEC
was
still
a
experimental,
so
that
fully
comes
back.
We
will
is
already
submitted
as
a
standard
track
and
we'll
work
on
it
make
sure,
is
coherent
with
the
these
documents
and
then
move
it
forward.
A
A
So
a
couple
of
months,
more
is
not
a
big
issue,
would
say:
okay
and
then
there
are
the
two
beasts
documents
we
have
to
move
forward:
I
plan,
at
least
to
finish:
the
30
beasts,
bye,
bye,
Christmas
guys,
so
I
will
push
on
the
mailing
list
to
make
programs
okay
we'll
come
to
that
later.
So
the
agenda
is
an
update
on
the
beast
documents,
security
and
the
yang
model.
A
These
are
the
working
groups,
items
okay
and
we
have
a
bunch
of
non-working
repair
items,
so
the
reliable
transport
damn
baseless
for
the
avionics,
deluxe,
GP
pub,
subscribe
document,
traffic
engineering,
a
demo
with
Android
and
iOS
implementation,
exciting
and
and
then
control,
plane,
authentication
and
authorization
and
mobile.
Okay.
B
On
the
security,
it
wasn't
so
much
that
it
was
experimental
and
when
standard
it
wasn't
so
much
that
it
was
experimental
versus
standards
track.
I
just
felt
that,
with
the
this
work
going
on
that
we
will
be
doing
standards
track
what
we've
I
think
it's
there's
a
lot
of
emphasis
on
security,
as
you
all
know
now,
here
at
IETF
and
I
felt
that,
if
to
ensure
that
we
make
get
our
best
documents
through,
it's
going
to
be
really
important
to
have
the
security
document
and
they're
they're.
B
C
To
give
an
update
on
on
the
beast
documents
so
explain,
which
are
the
changes
in
the
last
ones,
both
for
the
data
plane
and
control
plane
so
next
night,
please
so
for
the
data
plane
document
seems
last
IDF
Chicago.
We
had
four
versions
of
the
document,
mainly
they
are
the
result
of
some
discussions
on
the
mailing
list,
so
some
discussions
that
then
propose
some
changes
on
the
documents
which
I
will
say
that
they
are
mostly
editorial.
So
next
slide
please.
C
So
there
are
many
aspects
that
have
been
clarified
because
at
the
time
of
the
writing
of
Iraq
security,
some
aspects
were
still
under
debate,
for
instance,
what
to
do
with
UDP
checksum
for
ipv6.
So
this
has
been
clarified.
We
have
clarified
the
definition
of
the
RTR
s,
which
is
basically
an
editorial
work.
C
We
have
removed
the
a
statement
that
said
that
this
must
not
be
used
for
air
lock
space,
since
this
is
actually
relative
to
the
deployment.
That's
something
that
we'll
learn
from
the
original
document
again
and
for
the
same
reason
state
that
their
logs
are
wrote
over
in
their
local
space
and
not
globally
routable,
as
it
was
stated
before,
and
the
reason
is
the
same
because
that's
relative
to
the
deployment
there
are
some
deployments
which
this
might
not
be
strictly
true.
C
Another
similar
statement
state
that
the
map
cache
is
generally
shortleaf
as
opposed
to
shore
lift,
which
again
it's
it's.
It's
a
fact
then
state
that
V
pertains
to
the
data
plane
because
before
we
were
saying
that
it
was
either
an
ipv4
ipv6
addresses
and
the
payments
have
shown
that
this
is
not
always
true.
So
next
slide.
C
Again,
because
of
other
protocols
theta/dt
ours
may
not
may
not,
will
send
all
with
the
map
replies.
We
are
changing
a
Monday
to
recommend
in
the
maximum
number
of
list
headers
that
one
can
prevent
again,
for
because
of
some
deployments
are
actually
doing
that
we
have
added
a
reference
to
VPN.
In
the
instanceid
section
we
have
clarified
the
user,
the
EBIT
also
the
use
of
some
private
addresses
and
I'm
going
to
use
him
sent
ID
and
remove
the
reference
to
nineteen
eighty
eighty
so
next
slide.
Regarding
the
control
plane.
C
E
This
is,
do
you
know
good,
can
hear
test
test.
Yeah,
okay,
swallow
it
okay.
So
we
had
one
comment
from
you:
Albert
about.
We
made
a
reference
in
6832,
some
research,
that's
going
on
that
research
has
finished
a
long
time
ago
many
years
ago.
So
the
recommendation
by
Albert
was
to
remove
that.
So
we
could
put
a
new
update
out
and
send
it
to
the
list
and
see
what
they
think
on
sixty
eight
thirty,
three
bits:
alberto
just
came
out
with
a
set
of
clarifying
comments.
E
E
A
D
A
A
A
F
So
this
is
a
leasetech
and
Papa
Alvina,
Albert
and
diamine.
Some
of
them
are
here
next
slide,
so
yeah.
This
is
what
Debra
was
introducing
at
the
beginning
of
the
meeting.
This
is
the
reason
why
leasetech
was
moved
back
from
last
call
I
think
she
explained
very
nicely
the
reason.
So
this
is
just
a
record
of
what
she
was
saying
that
next
slide.
F
We
have
done
a
few
changes
due
to
comments
that
have
been
made.
Those
were
my
oh
sorry,
oh
I
can
watch
here,
Co
easy,
so
minor
changes.
One
was
about
the
recommendation
to
refresh
the
leaf
section
keys
to
address
key
edging,
and
now
there
is
a
small
section
that
is
basically
recommending
that
key
aging
and
key
compromise.
F
Mainly
preventing
over
claiming
attack
is
the
the
the
big
portion
of
the
world
at
least
sectors
next
slide,
and
this
is
actually
the
text
changes
that
are
that
have
been
done
since
the
last
version
next
slide,
yeah
I,
don't
know
exactly
what
what
is
the
processor,
so
I
don't
know.
Should
we
move
to
to
ask
I'll
wait
whatever
needs
to
be
done
there
I.
A
D
A
A
F
H
We
need
to
couple
it,
but
that's
there's
a
message
there
for
the
working
group
folks
to
make
time
to
look
at
these
three
documents
now,
because
there
were
gonna,
be
less
calling
all
three
of
them
and
you
don't
want
to
be
trying
to
fully
read
from
scratch.
If
you
haven't
looked
at
them
all
three
documents,
so
read
them.
They're
close
enough.
Now
that
you're
gonna
get
almost
everything
you
need
from
your
review.
You'll
want
to
look
again
when
we
last
call
them,
but
please
look
at
them
now
get
comments
to
the
list.
I
I
Rashaad
ramen
I'm
going
to
be
presenting
the
can
you
guys
hear
me
yeah,
okay,
so
a
young
model
update
next
slide,
please.
So
the
current
yang
model
was
updated
right
before
a
proud
meeting.
I
was
supposed
to
present.
It
had
some
slight
issues.
It
has
not
been
updated
since
we've
got
no
comments
on
the
mailing
list,
I'm
hoping
that
ball
gets
some
comments
on
the
mailing
list.
After
this
idea,
some
of
you
may
be
aware
of
the
nmda
guidelines,
which
has
been
going
on
all
over
the
place.
I
Knight
EF,
so
that's
part
of
the
main
change
we
made
in
the
latest
rev
in
July,
is
to
basically
not
have
separate
containers
for
configuration
data
and
for
operational
data.
What
the
nmda
does
is.
It
gives
you
a
new
data
store
which
is
for
the
operational
data
store
and
there's
metadata,
which
annotates,
which
gives
you
what's
the
source
of
the
data,
whether
its
intended
meaning
from
configuration
or
whether
it's
learned
dynamically
from
your
protocol,
for
example,
next
slightly
yeah,
so
we
remove
the
duplicate
containers
across
the
models.
I
For
example,
now
we
have
a
single
map
cache
an
idea
and
piti
offer
both
static
and
dynamic
same
stuff
for
the
map
server.
We
had
lots
of
references
to
state,
config,
CFG,
etc
and
descriptions
that's
been
removed.
Next
slide,
one
change:
we
did,
we
added
the
xtr
ID
type,
so
in
the
this
is
not
related
to
the
nmda
guidelines,
but
previously
we
had
the
xtr
ID
and
the
site
ID
replicated
in
the
ITR
model,
DT
remodel.
So
that
implied
that,
depending
on
your
role,
you
could
have
a
different
different
ID.
I
So
we've
moved
that
back
in
the
Lisp
router
instance,
which
is
in
the
main
model
the
ITF
Lisp
model,
but
on
the
MS
on
the
map
server
in
the
map-resolver
you
can
handle
multiple
sites
like
slide.
Please
mapserver,
previously,
the
authentication
key
was
per
mapping,
I
believe
with
doing
that
per
site.
Id.
Now
we've
added
the
site
X
to
our
IDs
to
the
mapping
records
and
we've
also
reordered
the
tree.
The
data
tree
a
little
bit
that
the
vni
is
the
parent
of
the
mappings
now
as
opposed
to
being
an
attribute
of
a
mapping.
I
Okay,
so
discussion
we
had
was,
as
far
as
we
know,
the
implementations,
the
most
common
implementation
is
to
have
it
per
site
ID,
and
the
discussion
is,
if
it's
uncommon,
to
have
it
per
prefix.
That
could
be
done
as
an
augment,
but
not
in
the
base.
The
implementations
I'm
aware
of
have
it
per
site
ID.
So.
I
E
I
Our
opinion
yes,
next
slide.
Please
we've
merged.
We
had
separate
models
for
P,
ITR
and
PTR
and
for
Peter
sorry,
so
we
move
with
merge
those
with
the
ITR
and
it
your
models,
we're
using
a
one
statement
and
the
present
statement
for
ITR.
It
enables
both
ITR
and
piq
operation.
We've
done
the
same
Tory
TR
next.
I
I
Also,
now
the
PTR
configuration
is
per
ITR.
It
was
per
mapping
earlier
and
we've
updated
all
the
versions
except
the
Lisp
address
types
which
hasn't
changed
next,
please
so
the
we
would
like
some
comments
from
the
workgroup,
but
I
guess
the
comments
we're
looking
for
is
I
mean
on
the
configuration
too,
but
we
feel
that
maybe
there's
more
operational
data
which
might
be
needed
counters
stuff,
like
that
and
a
private
comment
we
got,
was
to
take
a
look
at
the
list.
I
A
I
A
B
H
Work
was
started
before
we
started
putting
this
other
the
base
work
on
standard
tracks,
I
think
it
would
make
very
good
sense
to
move
this
to
be
proposed
standard
as
well
and
we'll
advance
it
behind,
but
not
very
far
behind.
We
hope
the
other
documents,
so
yes,
it
should
moving
at
the
standards
track,
would
make
very
good
sense.
Okay,.
F
Yeah
just
wanted
to
mention
that
yeah
there's
been
UDI.
No,
you
did
quite
a
thorough
review
with
the
your
coworker,
and
that
was
about
a
few
implementations
right
on
that.
There
is
also
an
implementation
on
WR,
an
open
source
implementation
and
data's.
Also,
you
know,
contributed
to
a
review
and
you
know
making
sure
that
would
that
was
aligned,
so
I
think
we
have
quite
a
few
implementation
available
and
yeah
I
totally.
A
I
I
Itf
91,
it
replaced
I
mean
the
previous
name
of
that
draft
is
written
there
on
the
first
key
bullet,
the
main
change
I'm
just
going
to
give
a
quick
repot
recap
what
was
presented
in
91
I'm
sure
some
of
you
might
have
forgotten,
and
the
main
change
in
the
documents
inside
EF
91
is
adding
of
the
scope
field
in
the
refresh
procedure.
Go
over
that
soon
next
slide
so
currently
with
UDP
registration.
This
is
your
typical
soft
state
problem,
which
has
happened
in
RSVP
and
other
protocols
where
you
need
to
continuously
refresh
your
state.
I
You
know
the
constant
communication
puts
a
load
on
your
system
and
we
feel
that,
because
of
use
cases
for
high
scale
that
we
need
something
which
addresses
the
scaling
issues
with
soft
state
refresh
next,
please
so,
what's
the
alternative?
The
alternative
is
you
use,
TCP
us
CDP
to
do
a
reliable
session
between
your
ETR
and
your
map
server.
You
use
that
session
to
communicate,
to
do
all
your
registrations,
it's
being
proposed
as
an
alt,
optional
alternative.
So
it's
optional
to
the
current
udp-based
registration.
You
have
to
support
UDP.
I
I
That's
such
format,
nothing
much
to
say
it's
very
straightforward.
Next,
okay,
so
the
we've
reused,
the
registration
message
from
UDP.
Just
for
simplicity,
you,
so
that's
your
registration
message
is
one
okay
message
type.
You
can
also
have
a
registration
acknowledgement
or
reject
from
the
map
server
to
the
ETR
and
the
third
bullet.
That's
the
scope
thing
we're
talking
about
in
the
earlier
slide.
Where
you
can
ask
the
map
server
can
ask
for
a
selective
refresh.
I
You
can
ask
for
global
scope,
which
is
everything
it
can
ask
for
a
certain
address
family
or
it
can
even
ask
for
certain
prefixes.
The
mapping
notifications
has
their
own
UDP,
not
really
needed
for
registration
acknowledgements.
So
we
have
a
new
message
in
TCP
which
shows
the
merged
map
server
view
of
the
mapping.
When
merging
is
a
news.
Next,
please
Oh
map
server
operation.
So
after
the
registration
is
obtained
via
TCP,
there
is
no
time
out
of
that
registration.
Unless
the
TCP
session
goes
down,
then
we
fall
back
to
timer
base
expiration.
I
The
registrations
can
be
rejected
for
one
of
the
reasons
listed
here,
so
the
refresh
refresh
request
can
be
issued
to
the
TR
to
obtain
initial
state
so,
for
example,
after
a
TCP
session
lab
and
already
spoken
about
the
notification
message
next,
so
basically
the
ETR
there's
regular
UDP
registration
until
the
reliable
transport
session
is
established
and
there
is
no
refresh
in
the
reliable
state
unless
the
map
server
asks
for
it.
Next,
please
so
beasley
conclusion
we're
using
TCP
or
any
other
reliable
transport
between
DTR
and
the
map
server.
I
It
eliminates
the
need
for
a
periodic
fresh
uses.
Glv
could
we
use
the
session
and
there
are
existing
implementations
of
this
and
they've
been
very
effective.
They
are
deployed,
they've,
been
very
effective
in
large-scale
deployments,
and
next
one
is
the
last
one
I
believe.
So
we
would
looking
for
comments
and
we'd
like
to
ask
for
adoption
by
the
list
working
group.
E
E
You
haven't
said
if
you
use
TLS
here
and
I'm,
seeing
requirements
for
having
the
interaction
between
the
mapping
system
be
encrypted,
so
I'm
just
wondering
if
we
should
maybe
spend
some
time
there
get
some
folks
together
and
see
how
Lisp
can
work
over
quick
to
make
it
reliable.
So
I
don't
know
if
you
think
that's
radical
well,.
I
A
D
I
Okay,
that's
the
last
one
run
baseless
next
slightly,
so
this
is
a
new
draft
which
was
submitted
in
September.
There
is
various
authors
in
a
way
you
can
take
it.
So
the
background
of
this
I
mean
actually
the
background
for
you
who
are
there
in
Prague
I,
believe
Fred
who's
sitting
here.
I
gave
representation
for
a
BGP
solution
for
exactly
this,
so
the
ICAO
is
proposing
a
new
network
which
would
be
ipv6
based
and
they
call
it
a
tea
and
IPs,
and
that
new
network
would
be.
I
You
know
for
the
air
traffic
systems,
the
air
traffic
controllers,
the
airline
operation
controllers
and
all
that
we
presented
the
draft
at
the
ICAO
meeting
in
September
and
basically
there's
I
mean
there's
no
protocol,
there's
no
new
protocol
extensions
as
part
of
this.
It's
informational
and
it
builds
on
mechanisms
which
exist
in
lists
already
to
show
how
we
can
do
this
next,
please
so
you're
going
to
hear
a
bunch
of
acronyms
on
the
left.
I
Okay,
so
that's
not
showing
us
okay,
they're
supposed
to
be
two
airplanes
on
on
on
the
left
there
you
basically
see
to
raid
your
regions
that
could
be
SATCOM.
Vhf
and
basically
the
plane
is
moving
around
and
gets
connected
to
the
core
or
interworking
region.
Here,
based
on
what
radio
region's
connects
to
you
see
the
three
three
circles
in
blue
those
are
supposed
to
be
X
TRS
one
is
an
Air
Ground.
The
two
on
the
left
are
a
ground
routers,
so
between
the
air
regions
and
NOC
underground
and
the
one
on
the
right.
I
I
I
Okay,
so
yeah
I
didn't
know
if
you
were
going
to
present
the
PowerPoint
or
no
it's
okay,
so
you
see
the
nice
little
plane
so
come
the
left
there,
and
so
it
basically
shows
so.
Each
airplane
gets
a
prefix
we're
showing
two
airplanes.
Here
we
know
prefix
X
can
prefix
why
there's
routers,
so
what
we
call
it
ground-based
Lisp
is
because
there's
no
list
mobility
is
not
running
on
on
the
airplanes.
I
I
You
see
that
the
aircraft
is
attaching
to
the
access
router
on
the
top,
and
this
gets
that
information
gets
transported
somehow,
for
example,
via
nigp
to
their
ground
router,
which
does
the
registration
on
behalf
of
of
the
Accra
I
mean
for
the
aircraft,
a
ID
with
the
msmr
and
step.
Two
is
exactly
the
same
thing,
except
that
it's
on
the
radio
radio
region.
Two.
The
important
thing
is:
there's
preference.
I
So
different
radio
regions
are
not
a
radio
expert
but
different
where
your
region
have
different
characteristics,
so
the
aircraft
might
decide
that
the
prefer
is
certain
radio
link
to
another
and
that
gets
signaled
over
to
the
access
router
and
that
gets
reflected
into
the
list
priority.
When
the
registration
happens
step
three
here
we
see
an
a
TSN
system
trying
to
send
traffic
to
the
airplane
the
ground
ground
router
does
a
map-request
with
the
msmr.
The
MSM
are
knows
that
the
Negron
router
of
region
two
has
better
priority.
I
So
the
map
request
is
forwarded
to
that
air-ground
router,
the
Aggron
router
responds
with
a
map
reply
and
then
the
ground
gran
router
sends
the
traffic
over
the
over
the
tunnel
next
slide.
Please
so
default
forwarding
path.
Typically
in
Lisp
implementations
when
you're
sending
a
map
request-
and
you
don't
have
a
map
cache
entry,
the
you
know,
many
implementations
will
drop.
Those
drop
drop.
The
packets
in
terms
of
the
requirement
were
told
that
that
was
something
which
was
undesirable.
I
So
what
we're
proposing
is
to
have
some
form
of
configuration
on
the
ground
gone
router
for
his
default.
Forwarding,
forwarding
path
and
the
default
forwarding
path
would
go
to
an
RTR,
so
the
traffic
would
be
encapsulated
to
the
RTR
and
the
RTR
would,
by
some
means
or
the
other.
It
could
be
co-located
with
the
msmr.
It
could
be
using
pub/sub.
It
would
have
all
the
older
mappings
in
the
network,
so
the
x-ers
are
still
doing
the
pull,
but
the
RTR
is
using
pub/sub.
So
we
see
after
the
map
request
is
sent.
I
I
Ok,
mobility,
so
this
call
it
optimize
because
it
works
around
the
issues
we
have
with
map
cache,
TTL
and
stuff
like
that.
So
here
in
the
slide
one,
you
can
see
that
the
traffic
from
the
end
system
to
the
aircraft
is
going
over
radio
region.
Two,
that's
the
connection.
The
airplane
has
right
now
using
the
procedures
we
described
previously
next
slide.
Please!
I
So
now
the
aircraft
connects
to
Radio
region,
one
on
the
on
the
other
access
router
and
the
aircraft
prefers
to
go
over
radio
region
one
so
that
again,
the
Akron
router
gets
that
information,
for
example
via
the
IGP
registration
happens
and
with
a
better
priority.
Now,
typically,
the
map
cache
would
have
to
expire
on
the
ground
ground
router.
What
we're
proposing
here
is
we
can
use
pub/sub
and
the
ground
ground.
Router
would
be
notified,
I
mean
for
that
aircraft,
prefix,
that
there
is
now
a
better
path.
I
Next
slide
and
the
switchover
happens
before
the
map.
Goshawk
expires
each
a
typical
stuff,
redundant
map
map
servers,
ground,
ground
routers.
They
are
drawn
routers
and
all
that
you
know
you
can
have
multipathing
in
the
underlay
and
also
the
aircraft
is
not
attached
to
an
anchor
point.
So
it's
as
if
the
anchor
point
fails,
the
fate
of
the
aircraft
traffic
to
the
aircraft
is
not
impacted,
so
you
don't
have
the
triangular
where
it
always
has
to
go
through
the
same
xxt.
Our
next
slide
security.
So
yeah
you
can
protect
the
control
plane
using
Lisp
sac.
I
You
can
encrypt
the
traffic
if
needed,
using
8061
and
your
dust
mitigation.
Is
you
know
the
usual
control
plane
limiting?
You
could
also
do
stuff
like
ACLs,
where
you
say
certain,
because
the
IP
addresses
are
fixed.
The
prefix
of
the
aircraft
is
fixed,
that
this
belongs
to
a
domestic
aircraft
I'm
not
expecting
that
to
register
over
a
radio
region
which
is
international
flights
and
stuff
like
that.
You
could
do
that
next.
K
L
L
Just
just
another
wing
through
the
ground
beast,
a
gate
link
link
is
just
another
link,
but
I
had
another
comment.
This
is
Fred
Zeppelin
from
Boeing.
Icao
is
actually
meeting
this
week
up
in
Bangkok
and
I
was
there
on
Monday
and
ground-based.
Lisp
is
one
of
three
proposals
that
they're
looking
at
for
what
they
call
the
mobility
solution.
A
second
proposal
is
the
simple
BGP
method
that
I
talked
about
last
time
that
Rashad
mentioned
earlier
and
now
they've
also
brought
proxy
mobile
ipv6
into
the
into
the
picture
to
so3
solution.
E
This
is
Dena
I
just
wanted
to
make
a
comment
about
the
when
you're
parked
at
the
gate,
and
you
have
a
wire
connected
that
could
just
be
another
mobility
event
right
and
I.
Think
that's,
maybe
what
you're
getting
to
use
the
same
machinery
and
it
doesn't
matter
if
it's
wireless
or
wireline
there
will
be
a
ground-based
xtr
there.
That
would
have
a
new
locator
at
the
airport,
I.
D
A
I
F
Yes,
yeah
just
a
note,
it's
a
very
articulated
use
case.
It's
interesting
to
know
that
this
pretty
much
basically
sprite
there
isn't
anything
new,
also
I.
Think
it's
interesting
to
note.
One
of
the
few
optimization
that
can
be
done
is
using
pub/sub.
That
is
the
direction
where
we
are
trying
to
move
the
protocol,
so
I
think
it's
a
nice
use
case
articulating
the
value
the
lease
brings,
and
also
it's
probably
confirming
that
some
of
the
direction
that
the
working
group
has
been
going
on
to
evolve
lease
per
hour.
Okay,
this.
E
Deal
yeah
I,
agree,
Fabio
what
some
I'm
kind
of
excited,
because
the
idea
mobility
draft
is
showing
that
this
one
mechanism
could
be
used
for
all
these
use
cases
we
found
via
mobility,
uses
the
ID
mobility,
this
airplane
or
aeronautical
application,
and
we'll
show
that
the
5g
proposal
we're
making
is
using
the
exact
same
thing
and
what
they
all
have
in
common
is
that
the
Eid
in
the
AR
lok
aren't
co-located
in
the
same
device.
They're
actually
separate.
So
we're
finding
that
this.
M
So
you
can
see
there's
a
lot
of
authors.
Let's
do
the
next
slide
please.
So
what
is
IO
am
basically
it's
a
mechanism
to
gather
a
bunch
of
telemetry
and
OAM
information
along
the
path
and
we're
doing
that
in
the
data
packets
that
you
normally
send.
So
you
don't
need
to
send
additional
probe
traffic
or
so
on.
You
just
appending
information.
M
The
data
packets
on
one
end
of
your
network,
we're
moving
it
from
the
other
end
of
your
network
and
then
reporting
it,
and
you
can
see
kind
of
in
real-time
all
this
stuff
about
what's
happening
in
your
network.
It
can
be
node
IDs,
it
can
be
interface
IDs,
it
can
be
transit
delays.
It
can
be
timestamps,
so
cue
depths
all
kinds
of
stuff
about
what
the
way
your
network
is
actually
behaving
and
what's
going
on
in
there.
M
M
So,
basically,
what
the
way
it
would
look
is
you
would
have
all
your
outer
headers.
Then
you
have
your
VX
line
GPE
and
after
the
VX
line
GPE,
you
would
put
all
this
I
owe
em
information.
So
there's
my
an
IOM
trace
header
that
tells
you
what
kind
of
information
you're
capturing
and
then
a
bunch
of
data
fields
showing
what's
going
as
what
you've
collected
as
you
go
across
the
network.
M
M
If
we
go
to
the
next
slide,
one
of
the
questions
in
the
design
is
whether
that
should
be
one
code
point
which
is
what
you
have
on
the
right,
but
that
kind
of
leads
to
nested
tlvs,
because
then
your
one
thing
is
a
length
and
you
have
multiple
things
with
lengths
inside
it
or
if
we
have
a
few
different
code
points
and
the
reason
for
the
few
different
code
points
is
we
have
a
few
different
things
that
we're
doing.
We
have
a
hop-by-hop
collection
of
information.
M
We
have
an
edge-to-edge,
so
on
one
edge,
you
insert
information,
and
then
you
remove
that
at
the
desk
at
the
the
other
edge
of
your
network,
when
you're
reporting
all
the
information,
we
have
a
proof
of
transit.
So
if
you're
doing
an
NFV
type
of
thing
and
you're
using
lips
lisp
to
go
between
different
middle
boxes,
then
you
can
actually
prove
that
you're
going
through
the
middle
boxes
you
wanted,
which
is
a
major
security
issue.
M
If
the
point
of
the
middle
boxes
is
to
deliver
your
your
a
security
functionality,
so
the
the
figure
on
the
left
is
showing
that
if
we
give
a
different
next
protocol
code
point
for
each
one
of
those
for
the
hop-by-hop
for
the
edge
to
edge
for
the
proof
of
transit,
then
it's
easier
for
the
hardware.
If
you're
doing
a
hardware
solution,
it
does
mean
that
you
may
have
to
go
through
you
know
one
or
more
ii.
It
could
be
more
than
one
next
protocols.
M
You
have
to
jump
through
each
of
them
having
a
length,
so
it's
not
hard
to
jump
through.
So
we
were
requesting
some
feedback
on
the
Left
versus
the
right.
The
expectation
is
if
we
go
on
the
left,
you
may
get
more
implementations
that
can
support
this
in
hardware,
but
both
options
are
things
that
that
we
are
willing
to
discuss
and
get
feedback
on.
M
H
M
H
F
H
Be
done
in
other
groups,
we're
not
going
to
get
into
defining
paulus
works
with
other
groups.
Well,
we
will
happily
consult
to
some
other
group.
If
the
nvo
three
guys
said
they
would
like
to
be
able
to
use
Lisp
as
a
control,
we
would
work
with
NGO
three,
but
I'm
not
gonna.
I
specifically
do
not
want
to
get
into
any
risk
of
this
working
group
trying
to
standardize
in
capsules
that
are
not
standardized
somewhere
else.
That
is
not
our
remit,
but
and
that's
why
I'm
trying
to
be
careful
here
sure.
F
We
are
specifying
a
protocol
right
that
can
be
using
doing
certain
things
right,
like
layer,
two
overlays,
and
we
are
basically
not
providing
a
question
on
how
that
can
be
done.
One
thing
that
I
would
suggest
from
my
point
of
view
is:
can
we
look
at
NFC,
sixty-eight,
33,
right
and
say?
Okay,
if
you
want
to
use
NFC
68
33
with
a
different
data
plane
that
allows
you
to
use
some
of
the
features
that
you
know
might
not
be
available
with
the
Lisp
encapsulation.
This
is
how
you
would
do,
and
that
could
be
a
draft.
F
F
H
H
F
H
E
G
F
Joel,
my
point
is
that
I
think
we
are
a
few
step
away,
so
I
think
the
question
on
Fuji
adoption
for
for
this
particular
draft
I,
don't
think
is
on
the
table
today.
I
think
the
working
group
should
think
and
I
you
know:
I
can
bring
up
a
draft
where
I
try
to
specify
how
you
can
use
68
33
with
another
overlay,
and
you
know
in
that
I
think,
is
a
discussion
that
we
should
have
in
the
working
group.
Then
you
know
the
working
group
will
decide
what
to
do
with
it.
N
Frank
brokers,
one
additional
piece
of
contacts
which
is
a
little
off
argan
oil,
but
we
have
running
code
for
this
very
encapsulation
in
VPP
Fido.
There
is
running
code
of
this
very
encapsulation
in
hardware,
with
a
bunch
of
chipset
vendors,
we're
using
Lisp
implemented
in
open
daylight
in
the
Maps
mapping
system
to
go
and
control
the
ving
solanum
encapsulation,
so
very
much
the
solution
that
that
value
was
referring
to.
N
So
we
are
already
using
the
fact
that
control
and
data
plan
have
been
decoupled,
and
this
overall
thing
is
integrated
in
OPN
avi,
fast
data
stacks
to
build
a
full
stack
solution
where
you
basically
set
up
tenant
networks
which
are
VX
land
based
and
list
controlled
through
OpenStack.
So
the
thing
is
happening
as
we
speak
and
we
can
say:
okay,
we
don't
want
to
have
it
here
and
but
it's
it's
there
in
the
industry
and
it's
happening
so
go
back
to
the
talk
that
they
fought
it
the
other
day
where
he
said
well.
N
A
Just
a
comment
on
this:
you
mentioned
two
things
about
the
implementation:
one
is
the
the
deluxe
GP
the
solution
implemented,
and
then
you
mentioned
that
you
are
using
the
least
control
plane
to
manage
this
now.
The
second
part,
the
working
group
could
be
interested
because
we
have
charted
at
two
on
the
control
plane
to
support
multiple
encapsulation.
Now
the
encapsulation
itself.
A
O
E
M
H
M
E
E
I
E
M
P
Padma
huawei
I
know
even
though
I
really
like
the
OEM
I
am
worried
about.
One
thing
is
that
how
we're
going
to
actually
reconsider
it
doubt
with
having
to
get
it
encrypted
and
I,
hear
what
you
said,
who
is
reading
their
information?
How
about
East
robbers,
who
actually
get
back
in
in
the
middle
I
know,
there's
been
so
much
of
discussion
both
in
SFC
and
other
places
about
monitoring
and
the
abuse
of
monitoring?
Have
you
thought
about
that?.
E
This
Deana
well
funny
you
bring
that
up
Padma,
because
if
this
has
to
work
with
list
crypto,
then
maybe
the
data
plane
did
just
come
in
charter
right.
So
the
big
question
is:
is
an
ITR
pre
pens?
This
puts
this
packet
data
in
the
packet
encrypts.
The
packet
sends
it
to
the
RTR
or
ETR,
let's
say
in
RTR,
because
it
seemed
more
interesting
there.
It
D
crips,
D
caps,
then
D
Crips
then
can
read
the
data.
P
So
before
you
respond,
I
want
to
say
something
else
is
also
about
where
the
data
is
stored.
What
I'm
worried
is
when
you
said
this
is
stored
in
the
map
server.
Maybe
that's
not
the
way
to
go
it's
depending
on
what
we
want
to
do
so
I
think
we
need
to
think
carefully
about
what
to
do
with
this
information
where
it
should
be.
Who
can
read
it
I
think,
there's
a
lot
of
things
to
do
there.
We
actually
make
a
decision.
There's.
E
P
You
know
just
to
say
some
of
my
some
of
our
latest
woes
is
the
fact
that
this
encryption
is
considered
not
enough
by
some
people
and
that's
why
I'm
kind
of
erring
on
the
side
of
caution
before
we
actually
say
we
will
do
this
or
we
will
do
that.
I
think
we
have
to
look
at
the
implications
of
what
kind
of
comments
is
going
to
come
from
other
areas.
F
Okay,
so
this
is
a
draft
froggies
rodriguez,
natal,
lease
pub/sub
version,
one
list
of
photos,
the
last
three
at
the
result
of
a
review
and
the
conversation
that's
been
going
on
in
the
working
group.
There
were
other
proposals
that
were
kind
of
adding
pub/sub.
We
worked
together
and
this
is
basically
a
version
that
summarize
they've
all
done
by
May.
F
F
F
So
when
the
map
server
is
sending
the
map
reply
directly
to
the
idea,
rather
than
sending
it
to
the
EDR,
the
reason
is
that
the
agents
for
pub/sub
comes
from
Sdn
like
applications
where
the
mapping
is
not
change
anymore
at
the
ideas
or
the
ideas,
but
is
changed
more
in
the
mapping
database
VI
northbound
API.
So
that's
when
you
need
to
work
date,
dxtrs
about
the
new
state
of
the
mapping
system,
so
the
messaging
used
is
not
introducing
a
new
message
but
Israel
rather
piggybacking
on
on
the
map
request.
F
There
is
a
notification
bit
that
is
basically
asking
for
subscription
during
a
map
request,
and
if
the
map
server
supports
the
then
the
publish/subscribe
or
if
it
has
resources
for
public
subscribe,
will
return
him
up
notify,
rather
than
a
dynamap
reply
to
notify
that
the
subscription
has
been
accepted.
If,
for
whatever
reason
the
subscription
is
not
supported
or
is
rejected,
it
will
be
returned
a
regular
map
reply,
so
in
this
way
even
X
tears
that
are
not
aware
of
pub/sub.
F
They
can,
you
know,
keep
using
regular
map
replies
and
map
request
with
the
WIDA
map.
Server
capable
pops
up
capable
the
publication
of
the
actual
updates
of
the
mappings
is
done
via
map
notify
and
there
is
a
map
notify
ACK
that
is
basically
confirming
to
the
map
server
that
the
mapping
update
has
been
received
next
next.
F
So
these
are
the
subscription
work
right.
So
this
is
the
initial
state.
If
you
look
up
in
the
in
the
mapping
system
in
a
mapping
server,
you
you
see
that
you
know
there
are
mappings
for
two
prefixes
in
in
dark
grey
1,
1,
1,
/,
24
and
2
2,
2,
/,
24
and
the
corresponding
are
lots
in.
In
light
gray,
there
is
a
DX
er
that
is
basically
going
and
look
up
for
one
of
these
prefixes
next
slide.
F
F
If
you
note
the
color
code,
the
map
request
will
contain
an
X
driv,
that
is
the
identifier,
unique
identifier
of
dxtr,
and
that
will
be
useful
to
identify
the
subscribers
at
the
mapping
system
and
then
that
map
request
will
have
the
notify
bit
set
to
one
specifying
that
this
is
actually
a
subscriber
request.
The
part
that
are
in
black
are,
you
know
basic
RFC
6830,
so
there
are
no
changes
needed.
The
parting
blowers
are
already
incorporated
in
RFC
68
that
it
will
be
so
this
is
aligned
with
what
is
happening
in
68
33
weeks.
F
Let
us
than
the
subscription,
so
this
table
is
the
one
that
will
basically
point
to
all
the
unlocks
that
are
subscribed
to
a
particular
to
a
particular
mapping
according
to
this
table
up
here
next,
so
the
map
server
is
capable,
is
a
pub/sub
capable
so
will
reply
with
a
map
notify,
rather
than
with
a
map
reply
and
the
mum
notify
will
contain
the
mapping
with
the
Eid
prefix
requested
and
the
corresponding
and
lock
and
also
announced
to
identify.
You
know
that
this
was
this.
F
The
xtr
that
are
subscribed
to
these
mapping
should
be
right,
and
so
the
map
server
can
look
up
at
the
at
the
of
subscribers
and
basically
find
out
the
airlock
of
the
xtr
that
needs
to
be
notified.
Next,
the
map
notify
is
sent
so
it's
sent
to
the
airlock
of
the
ideas
that
are
subscribed
and
basically
the
xtr
can
update
the
local
cache
that
now
is
synchronized
for
two
to
two
to
the
location.
F
Next
and
the
as
we
described
right,
the
XDR
we
send
a
map
notify
acknowledge
acknowledging
that
he
has
received
the
same
notification
of
the
of
the
change
mapping,
so
we
have
an
effective
way
to
basically
push
mapping
updates
to
xt
ours
that
have
subscribed
to
a
certain
mapping,
and
then
we
saw
in
the
icon
or
use
case.
This
is
something
that
you
know
is
useful
for
that
sort
of
application.
Q
F
Q
Q
E
F
And
you
will
see
exactly
that.
That's
I'm
not
sure
we
need
to
go
back.
It's
a
good
comment,
let's
check
on
the
next
I'm
subscription.
So
how
does
a
subscriber
I'm
subscribed
for
a
mapping
just
sending
a
map
request
with
the
n-bit
set
by
the
TTL,
with
no
idea
of
lock?
So
that
is
basically
telling
that
the
subscriber
is
not
interested
anymore
to
to
the
to
the
mapping
and
what
happened
if
a
prefix
is
unregistered
from
the
map
server,
so
a
prefix
particular
mapping
is
removed
to
the
map
server.
F
In
this
case,
the
subscriber
will
get
a
map
notify
with
ETL
equal
zero.
So
that's
the
message
that
is
basically
telling
the
subscriber
that
that
mapping
is
no
longer
it's
no
longer
active
and
there
will
be
a
map.
Notify
occur
as
an
acknowledge
back
next
slide.
Okay,
so
this
is
the
demo
that
dino
put
together
next
slide
on
lispers
may.
F
A
F
E
A
F
So
this
is
the
demo
that
Dino
put
together
on
lease
path,
so
there
is
a
leak
line
that
is
actually
sending
a
subscription
map
request.
So
the
example
we
show
that
the
map
server
creates
a
subscription
state
and
the
map
server
does
the
right
thing
by
storing
the
knowns
as
it
was
discussed,
and
then
we
we
see
that
there
is
a
change
in
the
airlock
set.
F
Okay.
So
here
is
example.
The
first
example
on
the
on
the
top
is
basically
showing
a
regular
map
request
why
the
map
request
is
going
out
from
the
league
client
for
Eid,
3,
3
3,
and
the
map
reply
he's
coming
back
here
with
this
particular
one.
So
these
are
my
ProQuest
just
provided
as
an
example.
What
happened
when
the
client,
the
X
yeah?
The
idea
here
want
to
subscribe
to
these
particular
projects?
The
map
request
is
sent,
but
in
this
case
the
the
NB
is
set.
F
F
So
this
is
the
state
up
in
the
in
the
map
server.
So
you
can
see
that
you
know
there
is
the
Eid
prefix,
the
TTL
The
Associated
arlok
of
the
subscriber,
and
then
this
is
the
xtr
ID.
It
is
basically
a
subscriber
table
that
we
were
showing
before
the
way
it's
displayed
in
the
in
the
list
box
implementation-
and
this
is
the
nouns
of
the
map
request.
Lisbon's
dotnet-
is
also
taking
track
of
the
map
of
device
that
have
been
sent
in
this
column
here.
F
The
map
request
is
the
regular
map
request
sending
up
the
oh
okay,
you
are
showing
the
registration
updates
yeah,
so
the
palette
we
are
showing
that
is
showing
up
here
is
that
we
need
to
update
the
mapping
right.
So
this
is
done
through
a
traditional
map
register.
So
this
is
just
adding
the
mapping
dr
lock,
join
and
ouija
to
the
mapping
of
tree
to
tree.
This
is
just
basically
spur
nothing
and
nothing
new
here,
and
then
you
see
the
result
of
the
the
map
notify
right.
F
G
F
E
O
R
Jung-Hoon
Bloomberg
LP,
this
actually
I
mean
this
may
be
a
lot,
but
this
actually,
ultimately,
is
the
enhancement
to
all
is
T
and,
interestingly
enough,
what
I've
heard
from
IO
am
could
actually
nicely
play.
Was
you
know
it
was
what
we're
what
we're
trying
to
do
anyway,
so
actually
the
bulk
of
the
work.
Well,
you
I've
proposed
the
use
case,
but
the
bulk
of
the
work
has
actually
been
done
by
Dino
in
this
effort
and
yeah
I'm,
ultimately
anyways
presenting
on
Gina's
behalf
next,
so
I
mean
we
did.
R
R
R
R
There
are
numerous
alternative
solutions
to
it:
they're
quiet,
you
know
the
again,
there's
there's,
certainly
various
people
who
have
come
to
IETF
and
solve
this
problem.
In
fact,
actually,
when
we've
looked
at
what
folks
that
did
the
Fitbit
project,
did
we
actually
thought
it
wasn't
of
great
interest
to
us?
R
We
just
couldn't
find
something
that
would
work
the
same
way
for
ISAs,
but
the
biggest
thing
that
we
didn't
like
about
all
the
different
like
crowd,
control
solutions
is
that
it
created
it
created
the
relationship
in
which
we
catch
you
create
an
eID
and
and
that
the
idea
would
capture
correspond
to
a
particular
path.
So
if
we
so
if
we
maintain
a
rather
dense
fabric-
and
we
have
numerous-
we
and
we
have
hundreds-
maybe
thousands-
of
different
tests
through
that
fabric,
we
would
have
to
be
creating
so
many
IDs.
So
that's
kind.
R
So
that's
why
we
wanted
to
go
and
explore
Lisp
as
an
option
next,
all
right.
So
why
did
we
want
to
do
SPE?
We've
we've
realized
that
you
know
we
could.
Actually
you
know
that
in
the
end
we
could
completely
adapted
to
other
micro
service
architectures.
We
we
thought
that
list.
He
could
easily
be
adapted
to
be
excellent,
GPE,
which
I've
just
heard
today,
so
that's
actually
pretty
nicely
playing
with
what
we're
trying
to
do.
We're
trying
to
you
know
to
us.
You
know
to.
D
R
R
What
generates
that
ELP
path
could
be
any
other
application
that
is
not
a
function
of
Lisp
that
is
a
function
of
as
a
function
of
IGP
does
a
function
of
some
intent
based
system.
You
know
whatever
that
may
be
so.
Yo
peace
would
come
from
topology
aware
systems
yo
peace
would
be,
you
know,
would
be
an
input
to
the
mapping
system
so
we're
able
to
what
we're
actually
doing
so.
What
we
would
effectively
do
is
we
would
get
the
that
I.
R
We
would
get
the
ITR
to
do
to
do
complete,
ELP,
probing
and
to
end
this
could
actually
be
combined
with
all.
We
could
effectively
take
everything
that
has
been
done
by
IOM
and
and
use
and
use
all
of
that
functionality
to
report
on
the
past
that
could
be
reported
to
the
mapping
system
that
could
be
reported
to
other
applications.
R
We
would
obviously
have
to
have
mechanisms
to
prune
those
paths
for
that
we
would
use
whereas
graph
algorithms,
but
in
the
end
of
the
day,
if
we
had
multiple
paths
to
different
a
ideas,
we
wanted
to
make
sure
that
we
wanted
to
make
sure
that
the
switch
over
to
any
other
paths
with
the
current
in
the
middle
in
any
of
the
Artie
ours.
So
so
the
idea
was
trying
to
create,
as
part
of
the
Lisp
header
either
was
to
create
a
path
ID
and
use
that
path.
R
R
R
It
independently
of
you,
know,
bgp
fabric.
If
we
wanted
to
convert
an
experiment
again,
yeah
IDs,
you
know
can
be
selectively
registered
was
the
mapping
system.
That
is,
we
could
selectively.
We
could
selectively
traffic
engineer
in
our
fabrics.
In
the
same
way,
people
traditionally
have
selectively
traffic
engineers
was
protocols
like
rsvp-te
we've
again
we
we
were
thinking
that
the
mapping
system
could
be
a
controller
base
or
it
could
actually
be
integrated
into
the
fabric.
R
Again
we
were
thinking
of
using
the
link
state
database
and
graph
algorithms
to
build
elts.
You
know
I.
Obviously
we
we
Pro,
we
we
can
support
ie,
both
monster
on
and,
and
you
know,
Antion
relationships.
Then
we
could.
Then,
probably
we
should
look
at
some
other
Oh
amp
protocol
and
ultimately,
the
way
we
steer
the
traffic's
through
those
paths
or
whether
we're
probing
or
actually,
traffic
engineering
is
through
the
us
through
the
use
of
Lisp
header
and
either
the
in
either
the
nonce
or
LSB,
and
the
draft
further
goes
into
that.
R
You
know
into
that
description
next,
so
we're
certainly
kind
of
like
looking
again.
It's
a
you
know:
it's
a
project
that
we're
thinking
of
exploring
you
kind
of
started
at
some
of
the
work,
but
ultimately
what
we
we
want
is
we
could
certainly
operate
both.
You
know:
data
center,
bgp,
topology
and
listy
topology
and
selectively
move
certain.
You
know
certain
workloads
and
did
you
see
on
there
a
lisp
tea?
R
We
can
you
know
we
can
do
some
interesting
manipulations
to
you
know
to
influence
forwarding
behavior
we're
we're,
certainly
interested
in
exploring
the
ability
to
use
this
on
a
number
of
interfaces
in
being
able
to
use
ipv6
link
local,
addressing
as
we
as
we
actually
under
Lisp
T.
So
that's
more
exploration,
work
on
that
again,
I
GP
for
our
logs
and
list
for
service
nodes,
again
process
parts
contain
there's.
You
know
kernels
whatever,
whatever
ways
to
split
things
into
micro
services.
R
Even
those
vendors
that
do
suddenly
realize
that
there
are
limitations
of
you
know,
certain
tables
can
be
so
we're,
certainly
trying
to
see
if
this
could
be
if
this
could
be
adapted
to
to
build
multicast
trees
and
reliably
maintain
them.
And
you
know,
and
and
potentially
you
know,
I
understand
that
this
discussed
that
this
working.
I
R
We
don't
necessarily
want
to
extend
you
know,
seamless,
BFD,
but
a
seamless
being
Duke
has
the
capability
to
probe
in
to
probe
individual
paths.
We
were
thinking
of
capitalizing
of
using
list
as
a
way
for
as
a
way
for
is
a
way
for
seamless
BFD
for
the
seamless
BFD
to
be
encapsulated
in
the
list
packet
and
use
Lisp
T
as
a
capability
to
probe.
She
protests.
R
Serious
we're
certainly
I
mean
there's.
Certainly
some
possibly
you
know
room
I
mean
the
discriminators
could
still
be
lets
say:
exchange
using
using
various
IG
peas
are.
There
are
traps
and
approaches
that
describe
that,
but
you
know
there
may
be.
There
may
be
an
opportunity
and
we
covered
exactly
so
that's
through
that.
Maybe
discriminators
could
be
register.
It
was
the
mapping
system,
but
that
I,
don't
think.
We've
thought
this
through
enough
to
really
speak
on
that.
H
T
T
Lower
it's
an
open-source
implementation
to
create
a
program
model
overlay
networks,
it's
written
in
C
and
it's
available
for
Linux
Android
and
open
the
well
via
RT.
It
will
be
available
for
iOS
and
Mac
Oh
soon.
We
hope
so
it
does
to
support
this
week's
LAN
and
it's
integrated
with
open
daylight.
So
what
about
I
use?
T
I
use
a
lot
of
people
things,
including
myself,
that
I
use
its
unique
space.
It's
unique,
it's
UNIX
base
it,
but
it's
in
rarity.
It's
not
UNIX
wasted
its
raise
it
on
Darwin
that
it's
X
and
you
extend.
You
means
it's
not
Unix
and
X.
Nu
is
a
mixture
of
Mac
and
BSD
and
also
iOS
is
not
posted.
This
means
that
if
you
write
some
code
that
it's
UNIX
compatible
or
POSIX
standard,
we
are
going
to
need
to
write
some
lines
of
code
to
make
it
work
on
iOS.
T
Also,
the
main
languages
to
deploy
applications
for
iOS
are
Objective,
C
and
Swift.
So
what
differences
we
have
found
compared
to
the
whole
of
our
Linux
implementation?
Here
you
can
see
a
little
some
examples.
So
Timmers
have
written
implementation.
We
miss
need
net
link
protocol.
A
very
sweet,
this
Liberty,
the
we
use
this
liberally
in
Linux
implementation
to
be
able
to
detect
the
network
changes
on
the
interfaces.
Some
signals
are
different.
Neural
Nexus
in
the
areas
were
in
the
iOS
version.
We
need
to
replace
some
system
signals
for
user-defined
signals.
T
We
are
unable
to
create
interfaces
and
raw
sockets
and
we're
for
for
widen
access
to
file
a
script
to
interface
file
descriptors.
This
means
that
we
cannot
send
raw
packets
even
that
we
can
say
that
the
80%
of
the
wall
of
our
core
code-
it's
reusable
in
that
case,
so
network
extension
Fri
wore
this.
This
is
the
tool
that
apple
provides
us
to
create
VPN
based
applications,
so
we
use
it
it
to
create
the
tuner
provider
week,
Allatoona
provider,
the
tool
which
creates
the
link
between
Swift
and
the
World
War
code.
T
So
it
allows
us
to
manage
VPN
connections
on
system.
It
allows
us
to
create
tuned
interface,
to
route
traffic,
to
the
tune,
interface
and
run
jet
and
send
traffic
to
it.
An
interface
and
create
and
manage
UDP
connections
are
so
TCP
connection,
but
it's
not
the
case.
Also.
It
allows
us
to
protect
the
wall
over
sockets
from
loops
into
the
system,
because
we
are.
T
All
the
traffic
to
the
tool,
so
we
need
some
mechanism
to
avoid
to
enter
on
in
a
loop.
It
was
available
since
iOS
8
before
was
necessary
to
tell
with
Apple
and
seeing
a
nondisclosure
agreement
to
develop
an
application
with
DPN
or
low-level
network
requirements.
That's
the
case
of
Open
VPN
application,
for
example,
and
until
I
use
them.
We
need
to
request
an
integer
and
entitlement
with
Apple
know.
We
need
to
request
this
environment
for
some
functions.
That's
not
the
case
of
of
VPN.
T
No
it's
without
entitlement
and
within
that
that
entitlement
was
necessary
to
check
some
some
things
when
they
pass
the
test
of
Apple
Store.
So
here
we
can
see
that
get
a
tour
of
the
application.
Ok,
so
when
the
user
launches
the
application
of
the
World
War
II
and
it
touches
the
vote
on
of
connect,
the
wall
of
our
application
runs
it
on
the
provider.
The
tuna
provider
is
ruining
the
background
always
and
when
the
tuna
prover,
when
the
tuna
provider
stars
it
creates
a
tool
and
stars
at
the
wall
of
our
core
process.
T
The
old
workgroup
core
process
opens
some
sockets
and
then,
after
that,
tuner
provider
start
shooting
all
this
application
ceased
application,
traffic
and
system
traffic
to
the
tool,
interface
tool
provider
from
the
tunnel
interface
and
sends
it
to
the
core
to
the
world
war
ii,
the
to
the
double
war
process
and
the
order
process
sends
it
to
the
internet
for
the
incoming
traffic.
The
procedure
is
the
same,
but
reverse
said
okay,
so
no,
we
are
going
to
the
demo
to
show
the
demo.
T
We
have
this
mug
book
that
has
were
shocked
to
install
it
and
we
have
an
iPhone
that
it's
ruined
all
of
our
application.
The
iphone
is
connected
through
USB
to
the
macbook
and
we
are
mirroring
the
screen
and
the
internet
were
interfaces.
The
network
interface
is
mirroring,
it's
like
when
you
mirror,
when
you'll
do
a
poor
mirroring
in
US
fiscal
switch.
T
T
Yeah,
it's
okay!
So
now
we
are
going
to
start.
You
know,
first
of
all,
they
lose
that
when
we
are
mirroring
the
screen
on
a
QuickTime,
Apple
activate
the
demo
mode.
This
means
that
the
old
information
on
the
status
bar
only
the
status
bar
its
demo.
So
it's
not
real.
As
you
can
see,
I
I.
We
cannot
see
the
courier
so
2
into
to
avoid
this
or
to
solve
this
decision
here
in
this
terminal
we
are
going
to
start
pinging
to
the
Wi-Fi
error
lock
of
my
iPhone.
T
T
T
Okay,
as
you
can
see,
we
have
the
an
IP
of
the
IDF
ranch.
So
now
we
are
going
to
open
the
lower
application,
the
first
time
that
the
user
opens
the
wall
of
our
application.
The
system
request
application
requests
to
the
user
to
add
a
VPN
profile
to
the
system
that
this
only
happens
the
first
time.
Ok,
so
we
recently
installed
application.
That's
the
fifth
time
that
I'm
going
to
open
the
application.
Ok.
T
T
T
T
Okay,
this
is
working
so
here
you
can
see
how
we
have
or
EAD
okay.
So
now,
let's
make
a
handover
from
Wi-Fi
to
cellular,
so
we
can
we're
going
to
see
the
ping
so
how
to
see
how
it's
not
failing
or
how
it
recognized
fastly.
So,
let's
disconnect
with
Wi-Fi
just
wait.
Okay
and
now
we
can
see
how
here
we
can,
the
the
Wi-Fi
airlock
is
stupid,
responding.
We
have
another
I
for
request,
an
IMAP
register,
and
now
we
can
see
how
the
packets
are
encapsulated
and
what
the
sort
interlock
interface
is
a
server
interface.
T
T
U
T
What
trigger
trigger?
Yes,
we
are
using
a
framework
of
up.
We
are
discussing
it
with
Apple
engineers,
because
I
don't
know
if
it's
the
most
effective
way,
because
Apple
give
use
a
system
configuration
framework
that
we
and
we
are.
We
can
register
a
notification
system
that
the
system,
when
detects
some
changes
on
the
networking
that
doesn't
mean
that
the
text,
for
example
it's
something
like
more
higher
level,
so
it
sends
a
a
trigger
to
the
system
that
tells
okay.
You
are
reachable
your
with
Wi-Fi
interface,
it's
working
or
it's
not
working.
Your
server.
T
U
T
E
E
Raining
you
know
event,
okay
and
the.
The
next
question
is
your
sending
info
requests
and
info
replies.
That
means
it's
can
support
going
through
Nats
as
well
and
I
mean
I
know
in
this
demo.
That's
here
at
the
ITF
is
we're
not
going
through
a
NAT,
but
does
the
implementation
work
going
through
NAT?
We
thought
not
NAT
devices.
Yes,.
T
E
E
T
Yes,
you,
you
can
always
I
us
always
tells
to
double
our
core
procedure
that
the
systems
that
the
interfaces
are
up
so
I
think
that
you
can
but
I'm
not
sure
if
you
can
use
it
at
the
same
time,
both
interfaces,
because
I
think
that
the
your
voice
are
working
on
a
sandbox
or
something
similar,
so
the
final
bind
or
the
final
decision,
or
of
where
what
interface
used
to
send
some
packet
I
think
that
it's
the
decision,
it's
part
from
it's,
it's
excited
for
the
iOS
kernel.
So
I
don't
know
here.
E
Unsanded
I
mean
if,
if
we
don't
get
send
and
receive
on
both
if
we
could
receive
on
either,
in
other
words
at
the
remote,
our
ITR
can
our
log
probe
and
find
the
better
path
and
come
in
on
either
direction.
That
would
be
more
useful
than
if
you
just
if
the
framework
only
allows
you
to
send
on
one.
That's
not
the
end
of
the
world.
If
you
could
receive
on
both.
That
would
be
cool.
Okay,.
I
E
T
T
E
E
Okay,
we're
gonna
present
list
the
Joule
signatures,
it's
a
draft
that
we
sent
out
in
July
the
ITF
in
July,
but
we
didn't
have
time
to.
We
didn't
have
time
on
the
agenda
in
July,
so
it's
had
the
original
version
and
one
revision
to
it
and
what
the
draft
covers
is
authentication
and
authorization
of
ITRs
using
the
mapping
system
describes
details
on
how
to
sign
map
registers
and
how
to
sign
map
requests
and
how
to
store
public
keys
in
the
mapping
system.
E
This
is
how
it
works.
Ex
TRS
are
assigned
public
private
key
pairs,
one
or
more,
depending.
If
you
want
to
have
multiple
IDs,
we
introduced
a
concept
called
crypto
key
ID
and
a
crypto
key
ID
is
an
ipv6
address.
That's
divided
into
two
parts,
a
prefix
and
the
hash
of
that
X
TRS
private
public
key
and
it
can
be
used
as
a
source,
Eid
or
simply
a
signature
Eid.
So
the
system
doesn't
have
to
use
this
particular
Eid
to
source
packets.
E
E
E
Both
the
shared
key
and
the
signature
verification
is
required
to
accept
a
map
register.
So
the
reason
this
is
interesting
is
if
I
look
at
Joel
and
he's,
and
we
want
Joel
to
register
the
mapping
system.
We
want
Joel
to
identify
himself
that
he's
Joel,
but
just
because
he's
Joel
doesn't
mean
he's
allowed
to
join
this
particular
instance
ID,
so
the
shared
key,
that's
associated
with
the
instance
ID.
Any
idea
that
he's
registering
also
has
to
match
so
there's
kind
of
two
layers
of
authentication
going
on
here.
E
Xt
RS
also
sign
map
requests
with
the
nonce
and
the
instance
ID
and
Eid,
so
maybe
I'm
not
registered
the
same
instance.
Ideas
Joel,
but
I
want
to
look
him
up
because
I
want
to
talk
to
him.
Well,
I
have
to
sign
my
map
request
and
I
may
not
be
allowed
to
get
mappings
for
him.
So
this
not
only
authenticates
me
but
also
authorizes
me
if
I'm
allowed
to
get
the
mappings
for
him
and
if
you
don't
get
if
you're,
not.
E
E
So
some
of
the
benefits
of
this
scheme
is
that
we
have
strong
elliptic
curve
cryptography
because
we're
using
ECDC
si
versus
all
the
traditional
diffie-hellman
RSA.
We
can
verify
an
invalidate
a
single
XT
R,
which
is
good
previously.
If
you
registered
a
bunch
of
e
IDs
to
an
instance
ID
and
one
of
them
had
to
be
invalidated.
You'd
have
to
change
the
key,
and
everybody
would
have
to
be
updated
with
the
new
key.
So
now
you
could
invalidate
a
single
entity.
E
We
can
use
the
signature
ID
for
re-registering
other
iid
types,
so
maybe
I
have
some
to
your
coordinates.
Maybe
I
have
an
ipv4
address
a
MAC
address.
These
are
things
I
want
to
register,
but
I
use
this
ipv6
crypto
e
ID
as
a
signature
e
ID
only
to
register
those
things.
So
it's
just
not
limited
to
registering
ipv6
EE
IDs,
and
now
that
we
have
access
to
the
public
key
of
these
XT
ARS,
we
could
actually
encrypt
the
results
going
back.
If
we
have
the
public,
the
map
server
could
actually
send.
E
Maybe
with
pub/sub
could
actually
send
map
notifies
and
encrypt
the
map
notifies
with
the
public
key,
so
the
only
the
xtr
could
decrypt
with
their
private
key.
So
this
provides
some
identity
privacy
as
well,
because
multiple
key
pairs
can
be
used.
So
you
can
actually
combine
this
scheme
with
the
ephemerally
eysies
that
we
have
in
another
draft.
A
family
ids
is
just
picking
a
random
number
and
constantly
changing
them.
As
you
connect
new
TCP
connections,
but
here
now
you
can
have
multiple
key
pairs
and
you
can
authenticate
for
each
one.
If
you
want.
E
So
I,
let's
show
how
let's
show
you
how
this
works
with
some
running
code.
So
this
is
a
little
program
that
actually
will
allocate
the
key
pairs
and
create
the
hash.
So
if
you
want
to
provision
a
new
XTR,
what
you
would
do
is
you
would
give
it
the
prefix
here
the
prefix
is
used
is
using
my
right
base,
dianna
thing
with
s1
FES
that
stands
for
safe.
So
it's
a
/
64
prefix,
that's
being
used.
E
If
you
go
a
little
bit
further
down,
you'll
see
that
the
crypto
hashed
a
ID,
that
is,
that
prefix
and
then
the
lower
door.
64
bits
is
the
hash
of
the
public
key
that
public
key
that
you
see
towards
the
bottom
is
what's
put
registered
with
the
map
server
and
then
there's
the
private
key
that's
stored
either
in
software
or
Hardware
on
the
xtr.
Oh
I
pointed
to
it
I
forgot,
so
that's
the
that's
the
prefix
for
the
hash,
that's
the
crypto
ID
and
those
that's
the
private
and
public
key
pair.
E
E
You
can
see
there
is
the
signature
in
a
JSON
string.
Okay,
then,
if
you
look
at
the
database
mapping,
that's
the
actual
ipv6,
a
ID
that
you
want
to
register
it.
We
tagged
it
as
a
signature.
E
ID
equals
yes,
and
then
we
would
register
this
particular
mapping
that
our
look
for
this
mapping
is
on
that
interface
and
an
angle
brackets.
E
Okay,
here's
the
map
register
comes
to
the
to
the
map
server
and
if
the
public
key
is
not
registered
to
the
mapping
system,
the
verification
will
fail,
and
you
see
that
right,
you
see
it
at
the
top
of
the
circle,
says
public
key
not
found
for
this
signature
e
ID,
and
so
the
signature
must
fail,
because
you
don't
have
a
public
key
to
verify.
You
do
not
accept
the
map
register
and
you
return
a
negative
map.
Reply
back
with
authentication
failure.
E
Okay,
this
is
a
map
register
that
has
a
good
signature
on
it.
In
this
case,
the
the
authentication
passed
for
the
shared
key
allowing
him
for
this
to
register
this
particular
EE
ID.
With
this
particular
instance,
ID,
the
lookup
of
the
crypto
hashed,
a
ID
was
successful.
You
see
found
I
can't
point
to
it
found
and
then
we
do
the
verification
and
it
worked,
and
then
we
accept
the
the
register.
Okay
on
the
bottom,
we're
showing
how
the
signature
is
being
good
for
a
map
request.
E
Now,
I
want
to
look
up
this
guy
and
the
signature.
Verification
passed
same
thing
that
the
hash,
the
public
key
mappings
looked
up
in
the
mapping
database
returned
to
the
map
server
or
the
map.
This
could
be
a
map-resolver
only
system
that
would
do
a
lookup
of
the
hash
to
public
key.
So
we
can
verify
the
signature.
E
Draft
status
were
not
requesting
for
a
working
group
document
this
time
we
want
to
get
more
implementation
experience
for
this,
but
we
think
this
is
possibly
a
good
feature.
Some
of
the
things
we
could
do
in
the
in
the
in
the
future
with
this
is
actually
encrypt
the
exchange
between
the
XT
RS
and
the
mapping
system.
We
also
may
want
to
figure
out
ways
of
publishing
hash
to
public
key
mappings
for
map
server,
so
encryption
could
go
in
the
opposite
direction,
from
X
TRS
to
map
resolvers
and
map-servers
questions
comments.
F
F
E
S
E
F
F
E
O
E
O
C
C
E
E
E
H
E
A
good
sanity
check
when
somebody
actually
registers
that
the
hash
actually
maps
that
you
perform
the
same
hash
algorithm
to
make
sure
that
the
hash
and
the
public
key
actually
map
right
and
there's
some
other
data.
That's
in
there
too.
So
it's
just
not
the
public
key.
It's
your
Eid,
a
KitKat
needed
with
the
instance
ID
as
well.
H
E
O
A
comment
to
this
semester:
this
is
very
good
for
Lisp.
Actually,
it's
giving
basic
security
properties
and
what
is
important
is
also
anonymous
idea
is
that
if
Emma
Larry's
generation
it's
one-way
hash,
we
cannot
maybe,
with
this
approach,
you
cannot
do
multiple
ephemeral.
Ids
can
be
generated.
This
I'm,
not
sure
about
that
I
think
Bob
was
working
on
this,
but
I
can
compare
to.
This
is
basically
keep
high
and
hit
it's
like
a
big
hippo
at
key
pace
and
yeah.
E
I
thought
about
this
quite
a
bit
is:
is
the
ephemerally
IDs
from
the
other
spec
as
general
or
random,
as
hashes
and
I
think
the
wisdom,
the
conventional
wisdom
is
that
hatches
are
pretty
random
in
their
own
right
as
well,
so
I
mean
I.
Think
these
cryptic
IDs
are
pretty
anonymous
and
I.
Think
if
you
continue
to
change
them
at
the
expense
of
provisioning,
provisioning.
Keys
key
management
is
not
a
trivial
matter,
so
you
have
to
be.
You
have
to
consider
that
you
know
this.
E
O
E
E
So
what
are
we
trying
to
do
at
the
highest
level
we're
trying
to
simplify
the
mobile
network?
This
proposal
is
called
list
mobile
network,
which
is
not
the
same
as
lists
mobile
node
lists.
Mobile
node
assumes
that
the
Eid
and
our
Lokar
co-located
in
one
device.
This
is
running
Lisp
in
the
mobile
carrier
network
to
simplify
things
and
enhance
mobility,
and
so
there
are
goals
of
the
new
5g
network.
E
That's
coming
out
in
2018
release
15
and
released
16
coming
out
in
2020
is
to
be
able
to
meet
these
new
latency
and
bandwidth
demands
for
VR,
AR
type
applications,
one
millisecond
latency
and
one
gigabit
worth
of
bandwidth
very
tough
requirements
to
meet,
but
we're
gonna
try
to
do
that
and
address
more
demanding
applications
like
IOT.
How
can
we
scale
billions
of
devices,
and
you
know,
orders
of
magnitude
more
than
the
number
of
cell
phones
that
are
in
the
world
today?
E
Of
course
we
want
to
do
dynamic.
Encapsulating
overlay
switch
gtp.
The
current
tunneling
protocol
inside
the
mobile
network.
Isn't
it
is
a
statically
provisioned
sort
of
network,
but
we
maybe
think
we
could
use
gtp
or
Lisp
as
a
data
plane,
but
we
just
make
these
dynamic,
encapsulating
overlays,
which
means
the
outer
header
changes
depending
on
where
the
host
or
the
Yui
is
right
now.
So
this
is
going
to
create
fast
mobility,
handoffs
and
I'll.
E
E
E
We
just
put
these
functions
in
the
UPF
UPF
for
both
physical
and
virtual
and
could
be
put
anywhere
in
the
mobile
network,
but
I
like
to
describe
a
node,
B's
and
P
gateways,
because
they're,
specifically
physical
they're,
at
a
specific
physical
location
in
the
mobile
network,
so
those
Ino,
B's
and
P
gateways
are
lisp
xt
ARS
that
could
be
running
any
type
of
data
plane
and
they're
assigned
our
locus.
Those
our
local
addresses
are
already
assigned
to
those
devices
in
the
network.
E
Today
they
come
out
of
the
space
that's
being
allocated
by
the
EPC
and
there
could
be
routable
by
the
underlay
and
their
aggregate
able
the
encapsulation
occurs
over
the
EPC.
That's
the
Evolve
packet
core,
the
IP
part
of
the
global
of
the
network
and
the
encapsulation
does
not
happen
over
the
ran.
Everybody
is
sensitive
on
the
number
of
bytes
that
go
in
a
packet
over
the
ran
to
get
this
one
millisecond
latency,
so
we're
not
doing
encapsulation
over
the
RAM.
The
way
packets
are
sent
over
the
ran
today
will
continue
for
4G
and
5g
applications.
E
E
Show
you
in
the
next
slide
the
underlay
is
the
existing
or
next-gen
EPC
or
NGC
NGC
core
IP
network,
and
the
overlay
runs
over
the
EPC
okay,
so
here's
a
sample
packet
flow
will
show
packet
flows
for
various
source
destinations
and
then
we'll
show
that
the
Yui
moves
and
we'll
show
you
how
it's
done.
The
way
this
slide
is
is
notated
is
that
the
white
line
is
where
the
TCP
connections
happening.
So
we
want
Yui
to
Yui
traffic
to
happen
here.
All
the
nodes
that
are
in
green
are
nodes
that
are
not
modified
at
all.
E
O
E
In
this
example,
when
the
Yui
on
the
left-hand
side
wants
to
originate
a
packet,
it
sends
it
like
it
does.
Today
it
could
be
an
ipv6
packet,
ipv4
packet.
It
has
no
encapsulation,
it
gets
sent
over
the
ran
just
like
it
does.
Today,
the
Ino
B
receives
the
packet,
does
a
lookup
for
the
rightmost
Yui
Yui
in
the
mapping
system,
the
mapping
system
that
is
managed
and
deployed
by
the
mobile
carrier,
and
you
see
it's
there
in
the
EPC
Network
and
then
the
left,
Ino
B
knows
to
encapsulate
the
packet.
E
That's
why
you
see
the
red
arrow,
it
means
encapsulated
to
the
right,
motes
Ino
bees,
our
locus
it
strips
the
outer
header
and
then
delivers
it
uninflated
to
the
UE
right
now.
Packets
have
two
hairpin
packets
that
go
from
you
te
to
you.
You
have
two
hairpin
now
through
the
P
gateways,
so
we're
gonna
eliminate
all
the
packet
traversal
over
the
multiple
hops
from
you
know,
B
to
pj
way
back
to
Ino
B.
E
This
is
an
example
where
the
UE
now
wants
to
talk
to
a
server
in
the
internet,
a
server
that
is
not
in
a
list
of
site.
So
it
doesn't
have
an
our
Lok
and
it's
not
an
eID.
It's
just
the
routable
address
the
packet
from
the
UE
to
Dino
be
would
would
move
just
like
it
did
in
the
last
slide.
The
e
node
B
would
look
up
that
IP
address
of
the
server
and
the
mapping
database.
It
would
not
find
it
because
it's
not
in
the
list
site,
so
there's,
no,
our
look.
E
So
what
it
does.
Is
it
defaults
to
one
of
many
P
gateways?
Those
P
gateways
can
be
clustered,
so
he
can
load
split.
All
the
you
know.
Bees
are
that
are
doing.
This
could
load
split
this
across
all
the
P
gateways
that
are
available.
P
gateway
gets
the
encapsulated
packet
strips
the
outer
header
and
delivers
it
natively
to
the
server.
E
This
is
a
situation
where
the
ue
wants
to
talk
to
a
server,
that's
behind
an
ex
TR,
which
means
it's
in
a
list
site
that
could
be
a
server
that's
sitting
at
a
data
center
where
that
ex
TR
is
a
top-of-rack
switch.
This
is
quite
interesting
because
this
is
the
first
time
we
can
get
the
shortest
path
from
the
e
node
B.
All
the
way
to
where
the
packet
has
to
go
and
the
P
gateway
is
out
of
the
loop.
It's
not
part
of
the
data
path.
E
This
could
be
a
feature
bug
depending
on
who
you
are
and
where
you
sit
in
the
standards
organizations.
But
basically
what
happens
is
a
packet
goes
from
the
UE
to
the
you
know,
B
just
like
it
did
on
the
last
two
slides
the
you
know,
B
does
a
lookup
and
finds
that
the
ex
TR
is
all
the
way
over
there.
In
the
data
center
encapsulate
sit
directly
there
that
packet
flows.
E
E
E
Okay,
what
happens
is
well?
It
starts.
It
either
originates
a
packet
where
the
Eno
B
recognizes
that
2001
:
:
one
is
a
new
Eid
that
it's
attached
to
it
or
when
the
security,
Association
and
provisioning
parameters
are
set
up
in
the
3G
network
or
4G
4G
5g
network.
To
tell
the
Ino
be
that
somebody
has
just
attached.
That's
probably
a
good
time
to
register
the
information,
because
you
could
minimize
packet
loss
if
you
do
it
before
the
real
IP
packets
start
coming.
So
what
happens?
E
Is
the
Ino
B
now
registers
itself
as
the
our
look
for
the
2001
:
:
1e
ID
to
the
map
system,
and
then
it
tells
the
P
gateways
or
any
other
Nino
B's
that
have
the
old
outlook
cashed
that
they
need
to
do
a
new
lookup?
We
have
three
mechanisms
and
list
today
that
we
can
do
our
loka's.
We
can
use
small
TTLs.
We
can
use
SMRs,
which
is
the
same
procedure
that
IOP
uses
to
update
locators
or
we
can
use
pub/sub.
E
These
are
the
three
mechanisms
we
can
use
and
they
all
have
different
levels
of
scalability
and
convergence
properties.
We
could
also
use
predictive
our
looks.
There
is
actually
a
fourth
one
if
we
know
that
this
Yui
is
moving
back
and
forth
in
a
metropolitan
city
area
between
these
Oh
Dino
B's
may
be,
the
remote
ITR
can
encapsulate
to
both
of
them.
E
So
this
specification
is
describing
how
Lisp
the
list
by
architecture
and
protocols
could
be
used.
In
an
LTE
5g
mobile
network
to
support
session
survivable
AI
D
mobility.
We
will
have
TCP
and
UDP
sockets
stay
up
during
all
these
moves.
We're
gonna
find
with
millimeter
wave
that
these
cells
are
going
to
be
smaller
and
that
people
will
move
across
subnets
more
often
and
the
need
to
keep
the
same
IP
address
is
going
to
become
increasingly
important,
so
we
need
to
have
the
mobility
at
the
IP
network
layer.
E
E
Organizations
we're
going
to
make
all
standards
organizations
be
part
of
the
solution,
so
we
don't
have
the
non
invented
here:
problem:
okay,
so
we've
already
presented
at
ITU
our
sorry
Xen
GP
and
we're
we're
planning
on
I'm
going
to
three-beat
GPP.
To
present
this,
the
only
reason
this
is
an
Internet
dreth
is
because
we
wanted
to
publish
the
information
somewhere,
so
we're
going
to
try
to
have
the
the
personalities
and
people
there
involved
in
these
other
standards.
Group
help
evolve
this
solution,
but
this
is
the
beginning
of
it
now
note.
E
A
A
A
E
Mean
if
the
adjure,
if
the
ID
address,
is
provision
either
by
the
manufacturer,
statically
like
MAC,
addresses
and
IMEI
czar,
and
it's
you
know,
Luigi
I,
annoying
that
the
ID
has
never
changed
for
you.
You
were
born
with
it.
You
use
it
all
the
time
right.
So
this
is
so.
The
question
about
address
allocation
will
be
simplified,
but
it
will
have
to
be
addressed
either
by
the
manufacturer
or
by
the
cell
provider.
On
giving
you
static
addresses
now
our
static
addresses
good
things.
Are
they
trackable?
E
I
E
Day
Luigi
has
this
phone
and
I
know
it's
Luigi
and
then
I
find
out
Luigi's
doing
stuff
in
Virginia.
It
might
be
Joel
that
has
the
phone
now
right.
You.
P
P
U
E
E
E
U
Can
do
the
marking,
but
my
point
is
yet
the
big
it
has
to
allocate
the
resources
right.
Initially
there
was
P
gateway
1.
It
has
allocated
the
network
resources
so
now,
after
you
hand
over
or
you
route
the
flows
through
a
different
path,
you're
no
longer
anchored
through
the
same
P
gateway.
So
now
I'm
wondering
how
you
move
the
Kostich
from
Piguet.
Do
you
want
to
be
gateway
to
that
part
is
well.
E
I
mean,
like
Padma,
said:
there's
a
lot
of
other
specifications
and
there's
a
lot
of
other
machinery.
That's
operating
here,
what
we
that
that
makes
3gp
even
the
whole
mobile
network,
working.
What
we're
trying
to
show
you
here
is
a
way
to
get
shortest
paths
on
the
data
plane
and
pulling
state
from
the
mapping
database
there's
a
lot
of
other
stuff
that
needs
to
be
worked
out
and
that's
why
we
want
to
go
to
all
these
standards
groups
that
have
expertise
in
this
area.
No.
U
O
O
P
Just
want
to
add
something:
there
are
other
documents
who
actually
show
how
the
actual
registration
involves
the
mapping
system
in
the
5g.
These
these
documents
are
currently
under
review
and
so
they're
not
ready
yet,
but
there
are
much
more
machinery
behind
this,
but
the
part
that
Gino
is
presenting
today
is
just
a
data
plane,
and
we
know
that
the
GDP
control
plane
there's
a
lot
of
discussions
happening
right
now
and
there's
a
lot
of
things
in
flux.
R
E
I
mean,
as
you
know,
Yann
our
TRS
can
be
placed
anywhere
in
this
in
the
EPC
to
hop
through
different
things,
and
since
we
know,
there's
going
to
be
a
lot
of
virtual
functions
that
are
going
to
be
in
the
5g
Network.
We
could
hop
through
them
where
necessary.
Now
it
got
to
be
a
little
bit
careful
about
that,
because
we
can
suggest
all
this
functionality.
E
I
E
The
mobiles
okay,
so
what
I'm
trying
to
scale
is
IOT,
which
is
order
magnitudes
more
than
the
mobile
system,
so
I'm
gonna
start
there
and
then
we're
scale
it
down
to
a
mobile
node.
So
I
know
that's
a
kind
of
a
ton
in
cheek
and
hand
waving
answer,
but
there's
bigger
fish
to
fry
in
IOT.
Now,
having
said
that,
in
the
IOT,
we
think
the
majority
of
devices
aren't
going
to
move
as
fast
as
mobile
phones
are
going
to
move.
E
I
have
some
slides
of
how
to
deploy
a
mapping
system
with
a
billion
nodes
in
it,
and
if
you
just
do
some
back
of
the
envelope
calculations,
if
you
only
store
one
million
entries
in
a
single
map,
server
and
cluster,
a
thousand
of
them
you're
at
a
billion
I,
don't
think
the
state
is
the
scaling
problem.
I!
K
Peter
Smith
huawei,
thanks
for
the
plug
just
on
the
scale
issue,
we've
been
doing
some
experimentation
with
pub/sub
in
conjunction
with
Lisp,
like
data
plane,
because
we
think
that
the
mapping
and
the
scale
of
the
mapping
is
actually
the
larger
part
of
the
problem.
The
data
plane,
sort
of
just
works
and
part
of
our
experiments.
We've
been
using
Google
pub/sub
we've
been
using
Microsoft
d'azyr
we've
been
using
different
massive
scale,
already
deployed
already
working
on
the
billions
scale,
pub/sub
systems
and
we're
seeing
one
or
two
hun
millisecond
response
times
just
using
those
existing
systems.
E
It's
an
interesting
comment
because
will
application
level
databases
scale
better
than
a
DNS
like
structure
like
DDT
and
I?
Think
we
need
to
do
experiments
in
this
area
to
see
what
is
better
now,
every
all
these
database
schemes
all
have
the
same
sort
of
problems
they
have
to
do
it.
They
have
to
deal
with
so
I'm,
not
sure
what
what
I
think
some
of
the
advantages
of
Lisp
DDT
is
is
that
the
same
database
does
not
have
to
be
stored
everywhere
with
a
lot
of
the
DHT
based
systems
do.
E
But
the
cost
of
that
is
that
you
have
to
go,
find
the
information
when
you
need
it,
and
do
you
want
to
pay
the
cost
at
that
time.
So
you
know
really
interesting
stuff,
but
a
lot
of
people
have
said
why?
Don't
you
connect
a
bunch
of
map
servers
together
with
Cassandra
or
the
blockchain
I?
Don't
know
why
people
think
the
blockchain
could
scale
it's
a
really
long
linked
list
right,
but
I
mean
you
know
the
UPC
guys
are
doing
a
lot
of
research
and
trying
to
figure
this
stuff
out
and
so
keep
doing
it.
O
One
clarification
for
this
question.
It
also
depends
on
where
you
are
putting
the
mapping
system
scale.
Question
doesn't
come
into
picture
at
all
if
it
is
if
it
is
a
4G
system
today,
the
you
PHP
Gateway
region
is
thousand
kilometres,
so
in
the
thousand
kilometer
some
cells
and
I/o
T's
are
connecting.
If,
if
your
mapping.
O
In
the
sitting
in
the
control
plane
and
if
it
comes
to
Phi
G,
the
thousand
column,
it
is
shrinking,
200
kilometers.
That's
why
your
you,
P
of
mobility,
comes
into
picture.
That's
well,
lispy.
Solving
that
that's
the
crucial
problem,
Phi
J's,
continuing
with
that
is
the
key
solution
here,
and
also
one
more
clarification
is
for
Louie
Lewis
Lewis
question.
It
is.
We
need
to
be
a
little
bit
careful
here
if
it
is
roaming
between
providers.
O
Q
Pandora's
box:
no,
this
is
very
good,
but
I
was
wondering
if
it's
this
goes
forward.
Do
you
think
that
there
will
be
additional
security
requirements
that
show
up
I
mean
I'm
just
trying
to
figure
out.
You
said
they
right
now
they
what
you're
showed
there
is
no
changes
to
the
protocol,
but
you
think
that
there
will
be
more
requirements
down
the
road.
E
I
think
it's
probably
gonna
be
inevitable.
I
mean
maybe
Pam
iruma
could
comment
to
that.
I
asked
the
question
yesterday,
a
five
gang
IP
if
less
crypto
would
be
required
and
how
much
you
know
how
much
encryption
versus
monitoring
of
flows
people
need,
because
that
pendulum
has
to
go
back
and
forth.
So
yeah
I'm
not
sure
how
much
security.
O
Actually,
that's
a
very
valid
question:
the
mappings
are,
or
if
it
is
put
it
into
the
providers,
control
plane
it
has
to.
It
has
to
have
an
optional
interfaces
where
it
can
attach
to
the
authentication
and
all
the
stuff.
So
it
has
to
be
mapping.
Cyril
interfaces
has
to
be
defined
so
that
it
can
be
plugged
into
their
system.