►
From YouTube: IETF101-ICNRG-20180320-1550
Description
ICNRG meeting session at IETF101
2018/03/20 1550
https://datatracker.ietf.org/meeting/101/proceedings/
A
A
C
E
F
D
D
D
G
D
D
D
Is
there
any
need
to
bash
the
agenda,
and
all
right
so
mark
is
on
line
perfect,
quick
update
of
what
has
been
going
on
so
so
mark
is
going
to
talk
about
these
hidden
X
documents
in
detail
later,
so
they
have
been
in
the
SE
IRS
she
review
process
and
that
led
to
some
comments
that
have
been
addressed
so
more
on
that
by
mark
they
has
also
been
updated
on
the
flick
draft.
So
flick
is
so
file
like
ICN
collection,
so
a
manifest
specification.
D
D
Okay,
okay,
yeah,
sorry
for
spoiling
the
surprise
so
M.
We
recently
believe
we
finished
the
disaster
scenarios
draft.
So
we
did
a
last
call
on
analyst
and
as
far
as
we
have
seen,
there
were
no
requests
for
changes,
so
we
think
we
kind
of
move
this
on
towards
publication.
After
the
meeting
we
recently
adopted
a
few
drafts,
so
the
terminology
draft
and
also
the
deployments
conservation
draft.
So
these
are
now
research
items,
so
that
means
we
will
continue
spending
time
on
on
them
and
so
that
we
also
you
know
given
by
change
control.
D
So
if
you,
if
you
want
to
participate
in
that
work
or
if
like
really
comments
or
consulted
comments,
please
check
that
out.
We
have
other
draft
that
have
been
progressing
so
for
G.
We
are
gonna
hear
later
this
also
an
update
on
I,
applying
I
siento.
I
io
t
see
the
main
list
for
latest
update
on
that.
D
So
something
is
okay.
This
there's
a
couple
of
new
things
that
have
been
submitted.
So
thank
you
very
much
for
doing
that.
So,
first
of
all,
Luca
and
colleagues
submitted
Russian
map
nice,
so
managing
incur
less
mobility
in
CCN.
There
was
a
presentation
on
this
on
Sunday,
so
unfortunately
we
didn't
missed
it.
Mistake
was
the
name,
so
this
is
currently
called
draft.
Int
fic
energy
met
me
so
that
it's
not
an
adopted
group
item,
so
we
kind
of
need
to
fix
that,
but
we
cannot
really
delete
drafts
easily.
D
D
D
Right
so
click
on
view
or
not,
but
the
new
things
perhaps
that
you
might
may
have
missed
Sao
Paulo
had
an
draft
about
ICN
routing
in
opportunistic
wireless
networks
and
geun-hye
was
talking
about
architecture.
Conservations
for
ICN
was
named
solution.
We
see
an
info
I
mentioned
and
we
also
talked
about
key
management
for
CCN
Indian
and
now
it's
time
for
the
first
presentation
and
that
should
be
Greg.
E
H
H
That's
what
I've
been
lower
a
year
ago,
so
the
title
here
IP
over
CCN
for
seamless
by
jig
mobility
work
at
the
time
when
I
give
it
the
earlier
presentation
on
this,
had
not
written
up
the
protocol
draft
and
had
not
done
as
much
simulation
work
at
the
so
now
I've
been
working
in
coordination
with
a
couple
of
PhD
candidates
at
Colorado,
State
University
in
the
US
and
they've
implemented,
ipok
india
and
sim
simulator
platform,
and
so
we've
got
some.
Some
eight
results.
I
H
So
background
here
is
the
ICN
see,
is
pretty
attractive
for
mobile
networking.
Elegant
built-in
support
for
consumer
mobility
via
the
safe,
stable
boarding.
Multipath
connectivity
is
also
pretty
interesting,
particularly
making
around
IG
networks
and
networking
scenarios,
as
well
as
heterogeneous
networks
with
Wi-Fi
and
hygiene
and
LTE
radios.
H
The
network,
caching
and
processing
for
edge
computing,
as
well
as
efficient
content
distribution.
But
how
do
we
get
there
like
so
network
slicing
is
a
key
component
of
Phi
G
or
a
lot
of
different
potential
applications.
That's
one
way
to
potentially
deploy
ICN
in
a
5g
network
would
be
to
run
it
in
parallel
with
IP
but
effectively.
That
means
for
the
network
operator,
you're
running
to
network
and
parallel
to
networking
protocol
parallel,
you
can
run
ICN
over
IP.
That's
a
lot
of
the
existing
employments
of
ICN
or
layered
on
top
of
IP
and
TCP.
H
But
then
you
lose
the
benefits
of
they're
listed
bug
because
the
mobility
benefits
are
difficult
to
achieve
over
running
on
top
IP.
So
you
want
to
take
a
different
approach
than
that
and
just
for
the
idea
of
using
CC
n
as
deporting
plane
for
high
G
and
we're
not
the
first
to
observe
that
running.
Natively,
ICN,
abre
mobile
network,
good
advantages
and-
and
so,
if
we
store
off
of
that
as
a
reporting
plane
kill
me,
you
support
all
existing
applications
on
top
of
that
and
use
effectively.
H
H
So
here's
how
it
looks
in
the
network
to
the
user
equipment
supporting
TCP
and
IP
applications,
the
top
of
its
stack
have
a
shim
layer.
We
call
ipok
IP
/
CCM
CC,
underneath
that
and
then
on
the
radio
link.
It's
running
CC
n,
actually
through
the
e
node
B
or
G
node
B
in
the
case
of
IG
over
Wi-Fi
IP
is
natively
forward
in
CCM,
and
then
we
have
at
the
edge
of
the
mobile
network,
a
gateway
device
that
interfaces
with
the
rest,
the
IP
world.
H
This
initial
work,
the
no
change
IP
stacks,
leverages
consumer
mobility
aspects
of
CCM
that
work
so
well
support
multipath
next
activity,
and
do
this
all
with
high
performance,
low
overhead
or
by
this
compelling
of
things,
but
for
RDP.
So
you
think
I've
become
a
simplified
network
architecture
of
the
client
and
a
gateway
within
the
mobile
network.
It's
all
you
need,
according
and
regardless
of
how
many
access
points
or
radio
links.
The
client
is
able
to
connect
to
they're
all
named
according
CCN
and
then
bridging
into
an
IP
network
via
the
Gateway.
H
So
I
mentioned
on
a
previous
like
leveraging
the
consumer
mobility
aspects
of
CCN,
so
that's
kind
of
key
piece
of
approach
we
chose
and
that
the
client
is
only
sending
in
Chris
messages,
so
all
of
its
traffic.
It
in
towards
a
network
is
composed
of
interests,
and
then
the
Gateway
is
only
setting
content
objects
in
return.
H
H
Each
IP
packet
is
encapsulated
into
a
CCN
interest
using
the
the
interest
payload
functionality.
Just
since
you
see
n
the
source
IP
address
goes
into
the
name
in
interest
message,
along
with
the
sequence
number
and
that
gets
routed
across
the
network
to
the
gateway
gateway,
dean
caps,
laced
the
payload
IP
packet
that
goes
into
a
packet
me
sequencer,
not
so
important
in
the
case
of
a
single
path,
connection,
but
the
case,
a
multi
path
connectivity
and
we
weren't
delivering
packets
in
order.
H
H
Interest
table
business
capsulation
actually
is
not
just
one
to
one
as
many
packets
as
to
fit
into
the
CCN
halo,
given
the
MTU
of
the
network
can
be
packed
into
that
that
content
object,
but
actually
consumes
one
interest
name
out
of
that
kind
of
interest
able
to
deliver
the
content
object
across
network
back
to
defined
again
we
have
a
pathway
sequencer
similar
to
upstream
in
man
in
the
Gateway,
and
so
after
the
encapsulation
of
IP
packets.
They
three
sequence
running
out
to
the
stack
of
device.
H
So
naming
convention
that
is
that
we
were
chosen,
is
to
encode
the
IP
address
the
source
IP
address
of
the
client
in
in
the
name,
along
with
this
sequence
number,
and
there
may
be
other
ways
to
do
naming.
But
this
there.
The
format
that
we've
chosen
effectively
implies
that
in
the
CCN
network,
that
the
Gateway
is
advertising
a
route
to
prefixes
IP
prefixes,
that
it
supports
that
map
to
where
routes
are
are
directing
traffic
that,
on
the
IP
network,.
H
This
gets
into
now
the
interest
content,
object,
exchange
and
make
sure
that
there
are
enough
interest
names
in
flight
and
that
that
find
interest
table
is
sufficient
to
support
accounting
traffic.
So
to
recall,
when
a
packet
arrives,
the
client
it
can
just
send
it
it
just
encapsulated
into
an
interest
and
sends
it.
The
gateway
is
a
little
bit
more
complicated.
It
has
to
make
sure
that
there
is
a
name
in
this
client
interest
table
in
order
to
build,
afford
anything
in
the
elfin
direction
towards
the
plant.
H
So
we
want
to
make
sure
that
they're,
the
steady
flow
of
interest
arriving
that
can
support
the
down
from
traffic,
and
so
in
order
to
do
that,
we
have
two
pieces
of
snap.
One
is
that
the
gateway
has
a
target
number
of
entries
that
it
wants
to
maintain
during
idle
conditions,
and
the
next
I'll
talk
about
how
it
manages
that
and
then
on
the
client
side.
Every
time
a
content
object
arrives
it
increments.
H
H
So
a
header
field
we
defined
in
the
ipod
better
of
the
idea
or
into
an
interest
deficit
report
and
allows
the
Gateway
to
effectively
request
for
an
increase
in
the
number
of
interest
that
are
in
fly
or
a
decrease
in
that
or
poor,
keep
the
same
so
value.
Zero
means
effectively
that
contra
object
contains
a
value
0
from
the
IDR.
H
What
that
means
is
that
the
client
is
going
to
respond
when
it
received
that
Kelantan
object
effectively
with
an
interest
heading
back
towards
the
Gateway,
so
it
keeps
that
will
keep
the
number
of
in
flights
interests
concept
if
Gateway
wants
to
increase
the
number
of
employ
interested
just
puts
one
in
the
IDR
that
effectively
results
in
the
client,
sending
two
interests
and
responds
to
that
content.
Object
in
research.
H
If
the
city
is
full
in
the
Gateway,
when
an
interest
arrives,
and
it
has
no
place
to
store
this
new
interest,
it
will
DQ
the
name
at
the
head
of
the
client
interest
table
and
send
an
empty
automatic
with
IDR
minus
one
to
decrease
the
amount
and,
if
number
of
said
entries
ever
falls
below
its
minimum
watermark.
This
min
value
and
it's
going
to
send
the
IDR.
H
H
The
the
main
graph
and
the
bottom
you
can
see
in
red
is
the
plot
of
legacy
and
there's
a
there's,
a
short
spike
at
the
very
beginning
which
is
enlarged
in
the
zoom.
Above
you
see
the
nominal
path
delay
crossed
this
I
networking
to
Gateway.
The
client
is
10
milliseconds.
We
see
a
nominal
here,
10
milliseconds
for
most
of
the
traffic,
from
the
gate
which
the
client
and
you
can
see.
The
initial
packet
achieves
that
ten
milliseconds
of
play,
and
this
use
case
initial
window
is
at
16.
H
So
we
have
a
floater
packets,
some
of
them
the
roasted
entries
for
and
they
can
go
directly
across
the
network
to
the
client
some
soon
those
that
enter
you
get
exhausted
and
if
that's
more
packets
that
are
sitting
there
waiting
to
go
across
the
link
and
they
have
to
wait
for
a
full
round-trip
time
or
that's
it.
I
just
get
refreshed,
but
then
pretty
quickly.
The
assistant
of
Auto
scales
the
number
of
interest
in
flight.
G
H
One
other
aspect
that
we
deal
with
his
PMP
lifetimes
expectation
is
that
in
intermediate
routers
between
the
client
and
the
gateway
that
it
empties
would
have
a
finite
lifetime
and
if
they
would
expire
in
an
intermediate
node,
then
the
name
that
sitting
in
that
that
client
interest
table
on
the
Gateway
effectively
useless.
It
would
not
be
able
to
carry
a
packet
across
a
network
to
the
client,
so
we
implement
a
lifetime
effectively.
As
a
interest
name
arrives.
It
goes
into
that
client
interest
table
also
as
for
within
an
expiration
time.
L
H
I
pocket
and
CCN
intended
to
show
complexity
and
simplicity
rather
than
going
message
by
message,
but
it's
really
your
much
much
simpler
operation,
there's
no
tunnel
manipulation,
there's
no
control,
plane
messaging
between
ease
and
II,
no
fees
and
sgw
and
PGW
is
to
setup
and
teardown
tunnels
and
hairpin
traffic
through
the
old
II
node
B
over
2,
nu
e
gonna
be,
and
all
that
kind
of
goes
away.
Exactly.
H
What
happens
when
the
ue
attaches
to
a
new
attachment
point
in
the
network
is
if
it
has
pit
entries
that
that
that's
waiting
for
until
office,
for
it
just
resented
those
identities
on
on
the
new
lake
and
in
some
cases,
those
pit
entries.
Those
interest
names
effectively
correspond
to
entries
that
are
in
the
sit
at
the
gateway,
and
so
when
they're
resent
across
the
network,
they
just
populate
a
forwarding
state
in
the
network
so
that
when
the
those
entries
are
utilized
at
the
gateway,
the
packets
are
out
of
control
objects,
arrived
client.
H
In
other
cases,
they
may
the
interest
names
that
have
been
used
by
the
Gateway
and
I've
already
sent
across
the
network
if
their
pad.
If
the
those
content
objects
are
passed
in
the
network,
then
the
UE
is
just
pulling
them
for
cash,
something
a
media
network
effectively.
The
intermediate
knows
the
network
hands
for
a
FIFO
cache
or
a
circular
buffer,
that's
equivalent
to
a
handover.
Iteration,
then
I
think
you
have
I,
think
no
packet
lost
during
handover.
The
things
are
simple:
to
support
our
soft
hand
over
so
adding
a
new
network
connection.
H
Youi
can
just
start
sending
its
new
interests
on
the
new
link
and
wait
for
new
objects
to
arrive
on
the
old
link
and
then,
if
you
tear
that
down
once
they
will
arrive
and
multipath
connectivity
is
also
pretty
simple,
striping
interests
across
different
links,
using
whatever
logic
you
chooses
to
do
purpose
and
heterogeneous
networks
with
different
radio
types.
Different
network
technologies,
handovers
and
multipath
negativity
can
be
support.
Raising.
H
So,
okay,
so
showing
the
hard
hand
over
in
red
and
the
soft
and
over
in
blue
perspective,
with
the
soft
hand,
nor
we
see
a
spike
in
throughput
because
for
a
short
period
of
time
we
have
two
radio
links
and
so
we're
using
both
of
those
for
a
short
period
of
time.
And
then
we
discard
the
old
link
and
only
use
a
new
one
from
a
blade
perspective.
H
The
officer
yeah
a
drop
in
delay
with
a
soft
hand
over
or
for
an
instant,
and
that
is
and
that
broth
is
not
something
that
the
end
applications
would
necessarily
see
the
physicist.
This
was
actually
measured
at
the
input
the
resequencer
Beckley
what's
happening.
There
is
the
new
link,
doesn't
have
the
TCP
buffer
build
up
that
the
old
Lake
did,
and
so
it's
much
lower
latency
connection
for
a
short
period
of
time.
That's
what
some
packets
in
the
TCP
connection
are
arriving.
H
So
we
have
two
implementations
so
of
12
2/3
implementations,
so
one
a
Linux
implementation
of
both
the
client
and
the
Gateway,
and
we
also
poured
the
client
side
of
that
to
Mac
OS.
Those
were
written,
the
top
of
the
Metis
for
that
implementation.
Couple
years
back,
then,
we
have
a
more
recent
engagement
in
DNC
M,
which
I'm
allowing
us
to
do
some
more
more
sophisticated
spirits,
I
guess
with
different.
H
M
Too
much
meat,
avi,
I'm,
Boog
I
have
first
maybe
a
more
remark
in
the
Sikh
Indus
as
a
Christian.
So
for
the
remark
I
mean
when
I
saw
your
intermingling
of
addressing
and
names.
I
was
thinking
of
remember
when
the
people
in
in
Stanford
in
the
tribe
project
came
up
with
the
name
based
routing
actually,
with
the
main
motivation
was
to
eliminate
the
binding
between
no
addresses
started
names
and
directly
go
on
the
name
so
to
eliminate
the
Alice.
M
M
H
M
I
understand
there
slowly,
I,
don't
but
still
I
was
I
had
no
no
particular
solution
in
mind
like
that
with
Christian,
and
we
do
this
sibling
the
car
second
Chris.
What
we
related
to
this
and
what
you
mentioned
in,
like
the
aggregation
effects
like
multicast
and
blah
blah
and
typically
interests,
are
you
can
effectively
do
they
still
work?
If
you
put
a
note
and
addresses
the
names
so.
M
H
That
those
names,
the
case
where
intersect
again
does
come
to
play,
is
in
the
handover
case,
where
a
interest
name
is
recent
north.
Now
that
reporting
path,
you
know
we
used
to
go
as
far
as
the
where
the
where
the
pals
join
right
doesn't
doesn't
need
to
go
all
the
way
back
to
the
Gateway
and
so
interest.
Aggregation,
helps
that
and
just
create
the
forwarding
path
to
the
new
location
of
the
network,
where
the
client
devices.
N
H
So
I
POC
handover
is
simple
in
both
case,
but
whether
too
hard
handover
or
a
soft
hand
over
it
simple
case
is
very
simple:
it's
either
just
do
disconnecting
one
radio
link
and
and
reconnecting
to
Nick
to
a
new
new
radio
link
and
then
resending
the
interest
in
the
front
of
it
or
it's
import,
handover
or
a
soft
hand
over
just
adding
a
new
link
and
starting
to
use
that
for
any
future
interest.
My
fourth
person
and
the.
N
Other
come
and
I
had
was,
but
like
so
you
put
the
on
the
IP
side.
It
looks
like
at
the
IP
address
of
the
source,
but
you
still
have
to
pin
your
route
to
that
gateway
which
case
it
has
to
come.
I
mean
it
has
to
come
back
to
it.
Yes
from
the
server,
so
I
guess,
maybe
you
should
put
the
I
mean
they
have
to
get
rest.
You
should
rewrite
the
IP
address.
It
goes
back
to
that
server
or
something
like
that
or
that
gateway
I
mean
yeah.
H
H
So
the
network
could
support
multiple
gateways,
each
supporting
a
different
subnet,
and
so
you
can
scale
that
way
right
that
any
client
is
going
to
be
interacting
with
a
single
gateway.
Any
point
in
time
for
any
IP
address
it
that
it
has,
but
a
network
would
have
dozens
or
hundreds
of
gateways
that
you
need
to
for
scaling.
O
Hi
Alec
Alec
I'll
just
follow
on
from
that
last
point.
First,
if
you
have
multiple
gateways,
that
means
that
the
client
presume
would
they
each
have
different
IP
address
ranges,
so
you
would
be
HCP
to
each
one
individually,
so
then
you're
an
it
would
have
to
be
farther
down.
So
it
was
far
enough
down
so
that
you
knew,
which
IP
address
the
client
reviews,
depending
on
which
destination
address
it
was
going
to
another
question.
I
had
was
your
pin
time
at
issue.
H
Directly-
and
it
all
depends
on
what
the
lifetime
is
right
and
I,
don't
know
that
there's
been
much
consensus
in
terms
of
what
is
a
realistic
lifetime
for
entries
in
CCA
networks.
If
it's
on
the
order
of
seconds,
then
we're
talking
about
a
fairly
low
packet
rate,
I
think
in
our
experiments
we've
been
using
something
like
300
milliseconds
as
a
timeout.
O
G
G
O
O
H
F
I
F
D
E
F
F
F
I,
especially
thank
the
chair
is
because
I
really
mean
honestly
within
planned
to
come
here.
We
got
so
busy
at
home
with
so
many
other
sins
what
other,
since
they
won
many
things
to
offer
the
option
to
activities
that
open
to
everyone
and
I
hope
some
of
you
will
consider
participate
in
some
of
those
activities.
This
one.
This
one
is
a
drive
for
the
ACM
information-centric
networking
conference,
so
you
can
just
hover
whenever
flag
to
go
past
and
collect
after
unified
work
has
been
so
you
know
when
you
write
a
paper.
F
The
first
thing
you
have
to
consider
is:
what
is
the
motivation
for
that
paper
for
answer?
We
were
lucky
because
Amazon
code
I
saw
was
a
motivation
by
creating
events.
It's
right
at
the
time
which
didn't
hear
that
at
with
Trento
funnel
and
the
paper,
something
big
happens
in
Amazon,
a
huge
failure.
I,
don't
know
if
everyone
heard
about
that
big
filter
or
how
big
is
big,
so
listen
for
the
fact
of
it.
I
want
to
utilize
some
ions,
our
own
words,
to
describe
that
you
can
see
in
case
you
haven't
heard
about
it.
F
It's
very
interesting,
so
this
thing
happened
it's
not
just
about
a
year
ago
and
exactly
what
happened
is
during
the
some
regular
maintenance.
Our
maker
ob9
errors
yeah
this
ever
just
reload
larger
site
observers.
Then
they
have
intended-
and
you
know,
Murphy's
Law
always
work
so
because
it
rewards
some
more
servers
as
they
you
you
intended,
and
what
are
those
extra
servers?
Those
extra
servers
support
to
a
powder
as
three
systems
as
three
days
and
some
replicated.
F
Yet
those
two
other
as
three
subsystems
down
in
turn:
how
do
the
index
that
manages
the
metadata
and
the
locations?
Information
for
all
the
s3
objects
in
the
region?
I,
don't
have
to
say
more.
You
can't
disappoint
here,
understand
the
complexity
of
today's
cloud
services
and
you
think
about
this.
Is
a
mass
figure
pick
again,
so
that's
really
get
us
to
look
into
how
we
can
divide
local
systems.
F
F
The
first
question,
but
you
Samson
looking
to
you
why
there's
such
a
high
dependency
on
the
cloud
and
the
way
I
understand
what
a
cloud
the
base
that
LT
actually
works,
so
we
subtracted
those
functionalities
of
those
existing
systems
and
I
didn't
do
it.
The
student
did
it
so
you
can
see.
Why
is
it
critical
to
have
the
cloud
to
be
app
for
your
phone?
F
Like
than
the
cat
feeding
machine
work,
let's
do
that
the
identity
management,
the
authentications
under
cloud,
the
device
management
and
the
pops
up
focus
and
together
were
in
the
storage
and
machine
learning.
For
these
analytics,
I
think
from
the
storage
machine
learning
stuff,
they
probably
always
will
pull
on
to
the
given
on
your
home,
maybe
in
the
far
future.
Well
so
far,
those
things
but
a
quarter
will
be
useful,
but
the
fundamentally
is
whether,
though
the
others
is
the
trust,
the
management
in
the
application
semantics,
whether
that
can
be
localized.
F
So
after
many
is
caching,
the
analysis
or
I
should
mention
that
lately
some
of
the
functions
has
been
downloaded
into
the
local
servers.
I
think
I
thought
probably
have
your
local
that
your
phone
can
do
some
of
the
functions,
but
that's
the
only
after
the
initial
configurations
like,
for
example,
I.
Don't
think
you
can
add
a
new
device
into
your
home
without
the
internet
connection.
But
after
how
configure
to
have
all
the
transitive
everything,
then
you
can
have
some
basic
functions
mention
the
locally.
So
some
people
ask
this:
why
it's
a
nutty
enough?
F
Well,
this
is
not
enough.
So
you
know
the
issues
I
think
the
people
can
understand.
Clearly
you
need
remote
services
to
achieve
local
functions.
That
can
can
hurt
really
the
real-time
performance.
I'm
gonna
give
you
an
example
that
we
used
to
exercise
our
design
and
then
you
really
subject
lock
operation
operations
to
the
remote
fillers.
That's
Hawkeye
dimension,
light
that
you
get
in
a
failure.
It's
only
like
marked
the
unwell
ability.
F
I
mean
only
showed
the
availability
problem,
but
the
bigger
issue
is
even
privacy,
so
opinions
as
a
very
recent
that
in
the
last
couple
of
days
there
is
a
exposure
about
the
Facebook
data
being
an
only
utilized
for
other
purposes,
so
so
think
again
about
you,
think
your
data
will
be
kind
of
our
shirt,
how
they
say
providers
in
a
cloud.
I
think
this
was
like
that
that
data
leakage
was
knowledge.
F
So
we
are
actually
in
the
indian
work
when
we
started
in
2010,
we
immediately
started
looking
into
how
we
can
utilize
in
the
end,
to
enable
local
iot
systems.
It's
largely
able
to
give
the
credit
to
professor
jack
work.
He
really
came
here.
I
think
they
showed
up
once
probably,
but
he
is
a
leading
person
in
developing
the
indian
applications
and
because
of
her
face
background,
you
know
he
he
didn't
do
home
LT,
but
the
his
in
the
UCLA
TFT,
the
theater
film
and
television
school.
F
So
therefore,
his
expertise
was
in
stage
control
before
we
had.
In
the
end,
he
questioned
all
of
the
people
who
do
networking
to
see
how,
if
and
so
kind
of
a
cat
your
head
around
so
many
times
when
I
want
to
do
simple
things
of
stage
control
to
say
the
left,
side,
lights
off
and
the
right
side
less
on,
and
you
have
got
me
goes
through,
like
many
steps
are
for
gymnastic
jumps,
I'd
need
to
get
IP
addresses
and
to
figure
out
how
to
run
up.
F
You
know
you
understand
it's
not
in
pure
science,
you
have
to
help
DNS
servers
and
all
of
that.
Why
can't
I
just
use
my
name's
to
control
the
systems
and
so
therefore,
when
we
started
putting
together
the
in
team,
when
Jacobson
wants
him
to
me
to
say
we
must
have
this
person,
but
that
time,
even
though
we
were
both
of
you,
Chevy
I
hadn't,
yet
long
story
short.
So,
but
you
know
over
the
years
we
decided
that
we
need
to
write
in
saga
about
what
we
have
learned
to
share
with
the
community.
F
So
the
work
on
protecting
the
year
actually
split
into
two
papers.
The
first
one
is
published
in
the
same
conference,
but
the
year
before,
to
sketch
out
to
the
basic
amount
of
to
say
where
and
the
in-house
yesterday
when
they
give
it
talk
about
technology
for
the
IOT
security
someone.
They
asked
me
the
question
to
say
what
about
the
energy
I
forgot
to
mention
that
think
about
all
the
severe
reduction
when
you
can
directly
on
top
of
the
physical
transmission
that
contrasted
at
aerial,
saves
tons
of
your
energy
as
well.
F
Now
they
remove
your
energy
so
based
on
that
idea
to
say
that
we
can
just
the
name,
the
data
collected
way
of
sensors
and
build
the
security.
Now
the
big
stompy
are
acts,
but
the
pace,
sound,
the
signature
or
otherwise
the
hash
of
the
data
generated
by
each
producer.
Then
you
can
build
a
pretty
simple
for
the
driver,
but
also
very
secure
system.
F
I
will
save
all
the
analysis
and
the
tests
show.
You
conclusion
would
believe
that,
given
that
we're
gonna
use
name
the
data
networking
for
the
home
LP,
so
we're
helped
will
well
already
name
individual
components
and
name
the
controllers,
then
the
functionality
we
need
just
to
why
the
laundry
wall,
so
all
the
devices
upon
each
other
or
the
different
components
of
application,
can
find
each
other
and
then
the
trust
management
that
is
the
ISIS
of
all
the
functionalities.
F
The
cloud
actually
provides
because
if
you
have
that
you
or
the
other
things
can
be
built
on
top
of
this
too,
so
therefore
a
focus
mostly
how
we
can
divide
the
local
trust
management.
The
other
day
I
was
talking
to
Dirk
and
he
said.
There's
one
thing
that
the
secure
sunny
you
buy
a
device
from
home
people,
I
think
about
that
the
Grand
Moff
either
device
from
Home
Depot.
How
could
she
just
plug
in
the
system
in
a
secure
way?
F
F
It's
all
based
on
naming
that
the
fundamental
difference
I
believe
between
what
we
have
there
that
compared
it
to
the
work
that
has
been
going
on
saying
in
the
IDF
efforts,
because
there's
a
phone
that
activities
they
have
done
a
lot
of
work,
but
it's
I
feel
fundamentally
different
because
of
a
name
home
light.
Have
this
in
the
whole
night
at
the
TLD
to
say
you
can
use
that
for
discovery
and
other
things,
but
that
is
not
the
name
for
your
home.
F
You
know
your
whole
night,
color
home,
that
my
home
that
night,
when
the
I
burn
my
laptop
shell
from
night,
my
laptop
cannot
have
which
on
that
is
real
phone
and
which
one
that
is
not
named
me.
Okay,
it's
a
unique
naming
for
individual
entities
in
the
server
space.
That
is
the
essential
of
the
starting
points
for
to
build
this
kind
of
a
local
autonomous
systems,
to
name
things
within
the
local
context.
That
really
requires
each
of
home
has
its
own
unique
name,
uniqueness
and
the
global
reach.
F
F
Beauty
final
rules,
assuming
that
I
think
many
more
a
wife
actually
is
a
monster
so
that
sums
up
in
the
master
you
defendant
knows.
You
use
your
key
to
sign
to
say
this
is
the
security
policy
in
my
home
dishes
pop
and
then
nobody
will
alter
it,
and
once
we
have
those
two
defined
in
a
way
that
the
devices
understand
your
technician.
We
have
this
thing
called
the
trusses
in
MA.
You
use
some
kind
of
a
formal
language
to
write
on
your
trusted.
Policies
and
device
can
really
execute.
F
That's
how
we
can
automate
the
whole
LT
system
so
that
your
grandma
can,
in
the
bad
device
from
default
of
the
year,
have
been
to
a
secure
system.
Now
the
next
question
is
really
about
a
long
table.
Will
you
buy
a
device
in
how
the
other
device
is
funded
so
that
you
can
integrate
the
new
device
into
a
computer
system?
F
Having
said
that,
so
oh
yeah,
where
we
mentioned
the
other
services
applications
can
be
bootstrap
with
the
three
things
and
also
I.
Just
repeat
myself
again:
kunal
service
is
definitely
it's
important,
we'll
be
there.
They
decided
we
should
now
that
depend
on
tight
to
run
on
the
life
well
cloud
cloud
service.
Fails
you
still
home,
you
can
open
your
home
door
rather
than
anything
locked
up
again,
I
like
to
inject
all
kind
of
editor,
those
being
locked
out
and
habit
and
number
of
times
and
the
helm.
F
My
necklace
there
the
hand,
experience
I
hope
to
experience
it
in
a
wonderful
start
out:
Thousand
Eyes
in
San
Francisco,
so
the
the
least
there's
a
commercial
buildings
and
then
one
day
they
couldn't
get
into
the
onyx
because
some
anthem,
server,
okay,
that's
a
real
story:
deponent
yeah
I'm,
saying
enough
of
that.
So
a
now
after
we
define
the
basic
concept.
How
do
we
know
it
actually
works?
F
We
are
doing
science,
you
know
where
cut
ourselves
scientist
the
only
way
to
improve
your
design
actually
works
in
the
store,
experimentation,
real
experimentation,
so
Jeff
work
had
the
students,
that's
the
second
author
to
how
he
actually
b14
the
DISA
I
would
say:
it's
really
demo,
where
game
thing
about
the
color
flow.
I
strongly
objected
to
that
I
said
throughout
the
positive
meaning
in
the
IT
environment
and
you
can
misleading
people.
F
Nevertheless,
the
exactly
the
workflow
that
it's
kind
of
an
exploration
game,
you're
carried,
is
that
your
scope
out
when
you
are
around
in
your
home
and
so
there's
the
location,
sensors
and
not
where
you
are
and
the
propensity
you
go
into
something
true
something
whatever
and
the
game
engine
to
show
that
this
way.
So
through
this
experimentation,
we
actually
kind
of
empirically
prove
that
the
basic
concept
actually
work.
So
this
is
other
Hardware
they
put
together
to
make
that
hope.
You
can
work.
F
I'm
gonna
skip
the
details
about
how
we
are.
Let's
talk
about
naming
because
that's
a
like
I
mentioned.
The
name
name
is
the
fundamental
starting
point
for
building
on
the
limb,
the
data
networking
or
any
applications.
On
top
of
that,
this
is
a
half
and
flow
application
exercise.
This
is
naming
design
you
can
see.
That
is
the
three
parts
of
the
naming
when
you
file
device
other
people,
how
you
kind
of
interact
the
key
into
the
device
and
how
would
your
home
system
accept
the
foreign
components?
F
That
is
her
own,
that
the
initial
is
the
trust
the
bootstrapping
started
with
something
of
your
arm
manufacturer
secret.
You
know
you
buy
a
device
with
a
label
with
you
in
the
car
code.
If
you
can
scan
that
in
that
will
be
considered
the
shared
secret
between
us
and
your
home
system,
and
so
there
you
can
install
that
share
the
secret
with
the
device
after
you
integrate.
So
that's
going
to
trap
the
trust
between
the
device
and,
after
that,
that
way,
they're
not
because
they
live
in
the
floor
implementation.
We
recognize
the
need.
F
You
still
have
two
suckling
species,
a
wise
in
terms
of
devices
that
you
want
to
mend
in
the
device
where
it
is
other
things
we
want
to
do
with
the
device,
but
then
the
namespace.
That's
the
application
link
space.
That's
a
how
you'll
control!
You
know,
man,
you
start
playing
games
where
all
the
signals
go.
The
request
of
dinner
goes:
that
is
application
names.
You
don't
wanna
cry.
Oh,
this
is
a
sensor
Amanda
you
know,
and
that
will
make
the
application
environment
difficult.
F
F
Doors
in
a
local
environment-
you
just
do
a
local
broadcast,
that's
enough
to
to
like
finally
hug
each
other
yeah
separate
namespaces.
Allow
you
to
limit
which
devices
received
which
can
occur
on
their
own
signals,
who
know
we
form
about
each
other,
not
come
at
it
from
a
parentless
to
a
quick
comparison
between
the
existing.
F
Our
two
system
is
on
earth
and
the
and
then
stuff.
So
you
can
see
that
all
this
trust
the
application
logic
it's
in
the
cloud
data
makes
missing
in
New,
Zealand
cloud
will
shut
down
your
application
in
the
in
the
end,
which
is
for
you
to
add
about
where
we,
what
is
so
great
just
like
what
a
father's
got
unique.
You
can
stick
with
anyone
on
your
own.
That's
how
much
really!
First,
you
have
this
local
trust
and
the
local
parks.
F
It
really
depends
on
your
local
naming
from
the
name
you
can
have
certificate
and
and
from
there
you
building
block
this
year,
and
that's
the
signal
I
mean:
that's
it
exactly
information,
don't
I
tell
you
I,
think
that's
also
the
difference
between
what
we
are
doing
name
the
data
compared
it
to.
We
just
think
about
how
it
tries
to
be
the
it's
the
other
established
in
the
relations
or
the
trust.
Often
you
meet
your
components
this
morning.
F
I
had
a
conversation
with
a
colleague
in
fact
was
a
difference
between
you
know
the
the
word
ICN
and
the
name
that
they
did.
Networking
I
don't
want
to
make
soy
sauce,
but
that
carrier
the
fundamental
of
this
new
architecture.
It's
not
about
context,
so
he
has
profession.
So
there
was
a
conflict
between
the
Indian
and
each
contribute
because
isn't
Indian,
so
other
cases
I
would
read
abstract
no
and
he
is
named
the
beta.
It's
an
old
model
names.
Ok,
the
contents
is
merely
the
site.
F
F
F
What
I'm
we
depend
so
much
on
the
cluttered
it
and
I
tell
you
the
market
always
pick.
The
chip
is
a
way
to
move
forward.
So
we
thinking
that
it's
a
the
so
called
the
old
is
a
cloud
centric
implementation.
Today,
it's
really
just
the
market
pick
the
least
resistant
of
has
to
reach
the
immediately
result
from
your
home
from
any
device.
There
is
cloud,
then
it's
just
mattresses,
a
20
point
of
delivery
are
for
the
Akeem
are
those
oh
wow.
F
You
know
to
cover
the
eyes
from
the
cluster
and
connect
it
and
that's
it,
and
also
there
is
a
tremendous,
economical
game
by
the
whether
you
come
me
on
scale.
You
know
you
put
lots
of
servers
out
there,
but
it's
cheaper
than
if
I
run
them
hang
themselves.
More
than
that,
the
IT
expertise,
it's
probably
the
most
scared
resources
and
not
many
people
can
write
a
color
service
so
great
so
great,
but
that
get
us
where
we
are
out,
but
we
cannot
stay
here.
We
cannot
stay
here,
a
cloud
dependency.
F
D
E
C
Okay,
good
afternoon,
everyone,
my
name
is
Guillaume
Dwyer
I'm,
an
associate
professor
at
100
city
of
Technology
in
France,
and
just
waiting
for
the
slides
I
will
press
yes,
this
is
what
I
see.
Okay
I
will
present
you
I
will
walk
to
LA
the
design
and
implementation
of
orchestration
solution
for
the
deployment
of
information,
centric
networking.
C
If
we
look
back
at
ICM
research,
we
have
about
one
decade
of
research
and
development
and
fundamental
topics
have
been
covered
regarding
architecture,
packet,
specification
and
so
on,
and
now
we
are
more
moving
toward
advanced
topics
such
as,
for
instance,
security
or
traffic
engineering,
and
we
also
have
a
set
of
operational
implementation,
which
are
available.
I
mean
we
have,
for
instance,
Ingenix.
We
have
CC
annex
and
we
also
have
some
testbed
0.40
research
such
as
the
Indian
testbed
off
of
idiot,
is
bed
whoops.
Sorry
I
tried
to
okay,
it's
very
sensitive.
C
I
mean
symmetric
traffic,
routing
and
also
caching,
but
there
are
also
some
other
challenges
to
address
and
I
do
identify
two
of
them
here.
The
first
one
is
related
to
management
and
security.
If
we
want
tell
to
adopt
this
networking
stack,
we
now
have
to
provide
efficient
solution
regarding
the
management
we
have.
C
There
have
been
a
lot
of
efforts
regarding
the
data
plane
and
control
plane
for
ICN,
but
it's
also
now
time
to
design
and
implement
a
management
plane
or
ICN,
and
the
last
point
is
the
core
Edition
with
IP
that
must
be
handled
as
we
have
seen
in
previous
presentations,
and
this
is
the
position
of
our
research
project,
which
is
founded
by
the
French
national
research
agency.
It
started
in
2014
and
it
will
end
this
year
at
the
end
of
September
and
we
have
selected
MDM
as
a
core
technology
for
all
our
design
and
deployments.
C
So
when
we
talk
about
deployments,
the
first
question
is
where
an
a
licensed
a
key
can
be
deployed,
and
there
are
two
different
solution
to
class
of
solution.
The
first
one
consists
in
coupling
I
seen
an
IP,
so
we
can
mix
protocol
stacks
that
this
is,
for
instance,
the
case
of
Cisco
hybrid
ICN.
We
have
seen
last
Sunday
all
we
can
leverage
software-defined
networking
and
there
is
a
set
of
references
I
provide
at
the
end
of
the
presentation
and
beyond.
C
We
could
also
think
about
that
applying
programmability
for
ICN
pipelines
through,
for
instance,
before
and
another
type
of
solution,
consists
in
isolating
ICN
an
IP,
and
if
we
want
to
isolate
these
two
network
in
stacks,
we
have
two
solutions
which
can
be
either
also
parallel
combination
with
dual
stacks
nodes
and
an
host
or
the
serial
combination
with
dedicated
gateways.
In
the
case
of
the
doctor
project,
we
have
been
ambitious
and
we
have
tried
to
combine
both
of
these
solutions
by
leveraging
network
function
virtualization.
C
So
the
promise
of
nav
is
that,
by
leveraging
the
isolation
property
of
virtualization,
a
Nicean
stack
can
be
deployed
independently
from
any
other
networking
stack.
We
have
a
tenant
domain
and
an
infrastructure
domain
that
are
deep
and
ICN
is
a
pelham
domain
protocol
stack,
which
is
just
offer
a
virtual
linear
to
network
and
at
the
infrastructure
domain
we
still
a
by
NP,
which
remains
as
a
networking
substrate
carrying
all
the
Internet
traffic,
and
that
way
we
are
able
to
either
either
later
traffic,
but
also
combine
them
to
carry
the
traffic
or
whatever
the
destination
is.
C
C
Okay,
so
you
can
see
on
this
architecture.
Our
proposal,
perhaps
if
you
analyse
this
architecture,
is
familiar
to
you,
since
this
is
actually
exactly
the
reference
architecture
provided
by
Etsy
for
nlv.
We
did
not
change
anything
regarding
the
role
of
components,
but
we
just
have
adapted
on
redesigning
them
in
order
to
take
into
account
the
specific
cities
of
ICN.
So
on
the
left
part,
we
still
have
the
authorization
layer
and
we
have
chosen
docker
as
a
core
technology
offering
virtual
computing
storage
and
networking.
C
And
now,
if
we
look
at
the
right
part
here
at
the
virtualized
infrastructure
manager,
level,
which
is
in
charge
of
controlling
or
the
network
function,
virtualization
infrastructure
itself
docker
here
we
have
straightly
use
docker
swan
since
it
offers
requirements.
We
we
had,
we
we
need,
but
we
have
design
level
components,
especially
an
adapted
version
of
manager
who
is
in
charge
of
pushing
of
the
configuration
and
into
Vienna
fees
and,
generally
speaking,
managing
the
lifecycle
of
virtual
network
functions
and
on
top
of
that,
an
Orchestrator
that
take
as
an
input
a
Tosca
specification.
C
Okay,
now,
let's
go
deeper
into
the
element
manager.
We
have
extended
the
work
of
the
Indian
management
protocol,
which
already
defines
a
set
of
metrics
to
monitor
and
provide
a
first
step
toward
a
full
instrumentation
of
nfg,
and
we
have
extended
it
in
order
to
fully
features
NFD
data
playing.
So
you
can
find
here.
Metrics
reading,
t2
faces
content
store
and
pending
interest
table,
and
there
is
especially
here
one
logic
which
is
a
bit
size
that
will
monitor
to
evaluate
there's
a
load
of
MDM
rotor
for
all
of
these
metrics.
C
We
also
have
design
some
micro
detectors,
one
micro
detector,
for
one
metric
who
is
in
charge
of
so
lowly,
detecting
that
a
particular
matrix
is
moving
away
from
a
normal
behavior
and
in
that
case,
raising
alarm
to
that
aim.
We
model
the
distribution
of
expected
behavior
for
one
metric
and
once
the
distribution
moves
away
in
terms
of,
for
instance,
mean
value
also
dispersion
of
data.
Then
we
raise
an
alarm
with
this
basic
detector.
C
Okay,
now,
let's
move
to
the
orchestration
part,
as
I
said
before,
we
have
extended
the
Tosca
format
for
ICN.
If
you
look
at
the
Tosca
specification,
you
will
see
that
there
are
six
different
definitions,
definition
for
the
actual
Department
units,
virtual
link
and
connection
point
for
this
three
definition.
C
Ok
regarding
the
orchestration
component,
we
have
a
code
written
in
Python
plus
rest
api
s.
We
have
implemented
all
of
the
component
from
scratch.
Apart
from
the
task
our
parser,
which
has
been
reused
from
an
existing
implementation,
we
have
so
three
components.
The
nav
orchestration
core
is
in
charge
of
the
initial
deployment
of
an
intent
service,
but
also
enforcing
the
dynamic
policies
and
for
the
initial
deployments
it's
sequentially
deploy
virtual
networks,
virtual
UNIX.
C
It
connects
a
virtual
units
to
networks,
and
then
there
is
a
second
engine
which
is
an
Indian
engine
that
take
parts
of
the
second
step,
which
consists
in
retrieving
videos
and
network
configuration
and
engage
V&H
configurations
and
to
that
aim.
Actually
we
have
to
take
into
account
NBN
forwarding
paths
extract
from
the
Tosca
specification,
plus
some
information
provided
dynamically
by
the
network
network.
C
C
C
We
have
selected
this
topology
because
the
different
links
are
available
through
the
internet
due
to
policy
data
set
and
in
each
point
of
presence
we
are
instantiated
engine
rotors,
a
signature,
verification
module
and
the
ND
and
firewall
for
all
these
network.
One
point
of
presence
actually
is
deployed
into
one
OpenStack
virtual
machine
means
that
under
the
virtual
network
you
see
here,
we
also
have
a
second
layer
of
the
actualization.
C
You
can
see
two
peering
rotors
on
the
left
on
the
right
of
the
topology
and
right
after
the
first
entry
point
after
the
peering
water
is
a
firewall
that
can
be
dynamically
configured
and
we
have
here
route
four
and
five
here
that
are
route
that
can
enforce
dynamically
a
signature,
verification,
and
then
we
have
the
core
of
the
network
right
here.
We
also
have
a
good
provider
on
the
left
that
provide
good
content
to
consumers
at
the
bottom
and
also
a
bad
provider.
We
try
to
send
corrected.
C
C
Last,
an
extract
of
the
policies
that
feature
this
topology.
You
can
see
here
the
water
sky,
recognition
for
the
signature,
verification
enforcement.
If
a
CPA
alert
is
raised
by
the
element
manager-
and
you
can
see
that
it
targets
rotor,
4
and
5-
and
for
both
this
route
or
when
such
an
alert
is
raised
by
rotor.
Two,
then
four
out
of
four,
we
will
update
the
rotor
mode
dynamically
to
enable
the
signature
verification.
C
We
have
similar
policies
for
the
update
of
the
failure
mode
blacklist
with
prefixes
that
have
been
detected
as
corrupted
by
the
second
by
the
signature,
verification
modules,
and
we
also
have
a
last
policy
which
allow
us
to
spawn
dynamically
some
n
generators
to
cope
with
the
resource
consumption
due
to
the
signature,
verification
of
waterfall
and
water.
Five.
C
Okay,
so
now
some
results
we
have
obtained
on
such
a
topology.
The
first
result
I
display
on
the
left
here-
is
the
performance
of
a
world
detection
solution,
and
especially
our
micro
detectors.
Remember
I
said
there
is
one
micro
detector,
one
for
each
metric,
which
is
monitored
in
an
FD
node,
and
actually
we
have
leveraged
the
hypothesis
testing
salary
to
be
able
to
analytically
and
selectively
anticipate
the
performance
of
our
detector
and
what
you
can
see
on
the
dotted
sure
here
is
the
theoretical
expectation
of
our
detector
in
terms
of
false
alarm
probability.
C
According
to
the
Shui
sod,
we
provide
to
a
lot
a
doctor,
and
the
main
result
here
is
that
the
set
of
lines
is
completely
consistent
with
this
erratic,
on
line,
meaning
that
for
three
metrics
for
monitor
here,
we
are
in
blue
font,
install
it's
monitoring
the
in
interest
heat
in
common
interest,
heat
and
pit
number
size.
We
are
able
to
accurately
feature
them
through
our
micro
detector
here
now.
C
This
is
an
extract
of
result
for
the
Bayesian
network
classifier,
and
you
can
see
here
the
data
set
used
to
train
this
network
and
you
can
see
that
starting
from
250
samples,
we
start
adding
a
estimation
of
I
would
say
abnormal
behavior
in
terms
of
security.
That
goes,
that
is
very
performant,
since
the
percentage
of
euro
is
about
two
percent
right
here,.
C
Sure
I'm
right
on
it
finish
for
wintering
part.
Now
the
idea
how
to
use
it
to
mitigate
the
policy
to
mitigate
the
economic
attack.
I
show
you
here.
Two
results
for
weak
attack
footprints.
You
can
see
here
with
blue
lines
the
set
of
good
data
received
by
a
good
consumer
and
in
red
lines,
amount
of
data
that
is
present,
and
you
can
see
that
for
weak
attack
footprints.
Our
orchestration
system
requires
more
than
300
seconds
to
mitigate
the
attack,
but
when
the
attack
footprint
is
stronger,
I
mean
we
have
a
rate
of
128
interest
per.
C
C
Okay,
so
this
is
the
end
of
my
book
about
the
way
we
could
liberate
Indian
and
especially
my
new
component
for
the
deployment
of
MDM.
We
have
currently
a
proof
of
concept
for
all
the
component
of
the
architecture.
Part
of
code
availability
is
available
here
following
this
link,
since
some
component
are
still
under
publication
and
not
fully
implemented.
Yet
they
are
still
not
available,
but
it
will
be
the
case
in
two
weeks
or
months.
C
One
open
question,
or
just
one
motivation
for
margin
here.
First
to
present
to
you
the
work
we
have
done
in
this
area,
but
the
consortium
is
also
open
to
serve
ice
energy
efforts
and
to
push
forward
the
deployment
it's
solid,
ization
of
this
neutral
parody,
and
what
about
pushing
it
forward?
Standardized
management
plan
for
ICN
the
track
for
future
works.
We
are
consists
in
evaluating
our
solution
with
real
users,
and
since
we
have
HTTP
gateways,
we
are
currently
opening
our
test
bed
to
students
from
12
ersity
of
technology
and
a
non
serial
university
of
rohan.
C
We
also
want
to
go
one
step
beyond
regarding
the
content
orchestration
saying
that
in
realistic
perspective,
we
could
a
situation
in
which
a
Tosca
specification
would
not
specify
any
from
any
sort
the
transport
infrastructure,
and
we
can
decide
that
for
content.
The
infrastructure
should
have
to
decide
automatically
if
ICN
is
the
best
opportunity
or
if
we
could
stay
on
HTTP
over
IP
solution.
D
D
P
Ok,
so
we'll
just
start
so
my
name
is
Jay
Jenk
from
akw,
hamburg
and
I
will
just
revisit
our
graph
that
we
have
on
the
yeah
as
an
adaptation
layer
for
CCN
and
India
and
Nepal
constraint,
networks
the
first
version
or
this
draft,
so
we
uploaded
the
world
into
a
meeting
last
year.
Since
then
we
have
2
version
bumps
and
there's
a
third
one
coming
so
just
before.
I
get
technical,
just
a
little
recap
on
on
what
actually
happens
here.
P
1504
radios
are
actually
widely
used
in
the
IOT,
for
example
the
Philips
hue.
Probably
everyone
knows
this
Philips,
you
Orakei
at
you
know,
and
even
the
Amazon
echo
plus
now
has
uses
1504
beneath,
so
they
call
it
zippy.
And
but
the
problem
is
that
using
1504,
you
have
basically
two
main
problems
here,
and
these
problems
are
in
the
protocol
header.
We
don't
have
protocol
identifier,
so
actually,
as
a
packet
puzzle,
you
don't
know
what
you
have
in
the
payload
there's
an
IP
packet
and
the
integrator
CCN
or
something
else.
P
The
other
problem
is
the
small
packet
size.
So
on
the
physical
layer
you
can
only
transmit
170,
byte
and
each
packet
and
and
it's
especially
for
Indiana
and
CCN
big
problem,
because
we
have
signatures
in
the
data
packets
which
of
course
can
get
like
to
200
and
and
more
bytes.
So
what
the
IP
world
did
is
to
define
a
sixth
open.
They
they
called
a
layer.
P
Let's
call
the
sixth
open
layer
and
the
sixth
open
layer
basically
handles
link
fragmentation,
and
they
have
had
a
compression,
didn't
everything
abstracted
so
that
you
can
use
IP
over
1504
and
what
people
and
with
this
raft
is
to
reuse
these
ideas
for
MVN
and
CCN,
so
that
we
have
ICN
okay.
So
we
just
reflect
out
what
what
was
done
in
these
six
token
groups
and
used
it
for
for
ICN.
P
So
we
did
some
restructuring
of
The
Dispatch
fields
and
picture
thing
and
then
and
we
considered
on
how
we
compress
they
had
us
and
they
also
did
yeah
in
the
elaborate
more
on
the
name
encoding
for
for
the
IOT.
So
we
I
will
show
this
later
and
also
we
did
some
number
crunching
and
put
some
numbers
in
the
raft
in
the
appendix.
So
let's
get
technical
now
this
is
actually
what
or
how
the
ndn
naked
specification
for
an
interest
looks
like
now.
P
P
We
have
a
name
which
is
that
has
the
same
the
same
structure,
so
you
have
the
name
TLV,
and
then
you
have
components
of
a
series
of
name
components
and
at
the
end
you
can
either
digest
so
and
then
you
have
a
bunch
of
optional
parameters
which
you
can
see
you
so
what
we
have
now.
If
look
on
the
bottom,
there
you
see
like
a
dispatch
octet.
These
are
eight
bits
that
we
use
as
a
dispatch
byte.
P
So
this
is
the
same
language
as
four
six
open,
and
this
is
this
is
basically
this
one
byte
is
basically
sitting
between
the
1504
header
and
the
actual
payload,
so
ndn
OCC
n.
So
we
allocate
one
bit
of
this
byte
to
just
notify
packet
Pazza.
This
is
a
compression
of
compressed
in
the
N
or
not
of
compressed
in
the
N
packet
and
from
there
we
go
further
and
we
can
basically
just
use
one
bit
to
encode
the
same
as
the
interest
packet
at
the
interest
type
here.
P
So
we
are
getting
rid
of
one
byte
and
trading
it
for
perform
bit.
So
we
do
the
same
with
the
name.
I
will
talk
about
the
name
encoding
after
this
slide,
so
I
will
just
skip
this
note,
and
then
there
are
some
specialties,
for
example.
In
the
end
we
have
boolean
values,
for
example,
can
be
prefixed
and
must
be
fresh
and
in
ndnd
we
use
two
bytes
for
this,
which
can
basically,
you
can
encode
a
boolean
value
on
one
bit.
This
is
what
we're
doing
here.
P
P
So
if
you
can
see
this,
the
length
field
of
a
lifetime
is
actually
variable,
so
the
lifetime
can
be
like
1,
2,
4
or
8
byte.
What
we
do
for
the
IOT
is,
we
are
constrained
at
work
and
we
limit
this
to
2
bytes.
This
means
we
can
encode
65
seconds
into
this
field.
So
we
can
say
the
lifetime
of
the
pit
is
65
seconds
at
max
and
1.
Once
we
have
the
structure
you
can
see,
the
nonce
is
limited
to
four.
P
This
is
given
by
the
N,
and
now
we
had
lifetime
limited
to
two
bytes,
and
now
we
have
the
hop
limit
limited
to
one
one
button
here.
You
can
see
the
binary
pattern,
so
it's
a
binary
encoding,
and
so
we
move
the
parameters
up
and
now
from
the
perspective
of
a
pocket
passer.
You
have
you
see
the
length
field
here
and
if
this
length
field
is,
let's
assume
five,
then
you
then
the
Parthenos
okay.
There
must
be
a
nonce
and
hop
limit
because
they
both
add
up
to
five.
So
this
is
no
trick.
P
We
use
here
to
implicitly
remove
the
type
field
and
the
length
field
without
using
any
bits
here
or
The
Dispatch
field,
and
after
this
there's
also
CID.
But
I
don't
talk
about
this
today.
This
is
for
the
stateful
compression.
This
is
still
the
future
work,
but
basically
we
would
use.
This
is
just
a
pointer
to
a
further
byte
that
contains
State
for
the
compression,
so
I
promised
to
talk
about
the
name
encoding.
P
So
a
name
component
length
is
15
bytes
long,
and
we
say
that
we
in
this
case
we
say
you
have
used
the
uniform
component
types
of
each
each
component
actually
can
be
composed
off
of
different
component
types,
and
we
say
we
just
use
a
generic
component
for
indian
names
in
this
case
and
when
we
have
this,
we
can
alight
the
type
type
field.
So
we
move
out
that
the
type
field.
So
that's
the
second
picture
here
in
the
middle
and
after
that
we
can
rearrange
the
length
fields
and
reduce
the
length
to
14
bits.
P
Sorry
to
4
bits,
because
I
said
we
can
add
max,
have
15
bytes
for
for
name
component,
and
then
we
end
up
with
the
last
picture
in
the
bottom.
So
we
see
this
is
just
we
use
just
one
byte
to
encode
the
length
for
the
next
two
name:
components
and
we
use
a
stop
symbol,
and
by
using
this
you
can
even
get
rid
of
the
name
of
the
outermost
name.
Tlv
yeah
length
field,
which
is
also
one
byte
or
more
I,
wish
you
is
the
last
picture
in
the
bit
level.
P
It
looks
like
this
so
because
here's
a
trick
now
and
M
so
by
by
using
by
rearranging
the
length
fields
into
into
one
bite,
we
have
an
advantage
and
with
an
odd
number
of
name
components,
because
now
we
can.
You
can
see
the
the
green
pattern
here
now
we
can
encode
the
stop
symbol
into
this,
the
same
byte
that
would
use
yeah.
P
That
would
be
used
for
the
second
M
component
and
by
using
this,
the
same
crew
and
the
Alice
the
same
encoding
where
we
can
see
at
the
formula
I
don't
so
so
we
we
win
like
a
factor
of
four
of
bytes
because
of
removing
all
the
type
gears
and
in
the
length
fields
and
here's
an
example.
So,
let's
just
assume
we
have
a
name
ITF,
101
sense
or
temp
2018
and
in
the
upper
part
me
up
a
picture.
P
That
is,
why
is
a
green,
particularly
for
even
number
of
components.
You
can
see
that
they
said
the
stop.
Symbol
is
using
a
full
byte
here,
where
we
decided
for
this
actually
to
to
rearrange
the
binds,
because
we
keep
the
byte
alignment,
which
yeah
I
mean,
which
would
make
sense
when
decoding
and
coding
course,
if
you're
by
the
lines
you
don't
have
to
fiddle
around
with
the
names
later,
when
n
decoding
and
here's
a
a
real
Indian
interest
would
look
like.
So
at
the
upper
part,
this
is
in
a
none,
encourage.
P
You
can
see
the
the
1504
header
is
using
in
this
scenario,
9
bytes,
and
this
is
actually
pretty
reduced
terror
size.
You
can
also
have
higher
sizes
by
longer
sizes,
which
are
double
the
size
by
using
long
addresses
long
make
addresses
here.
We
are
just
using
yes
short
address
mode,
it's
just
a
two
by
four
source
and
destination
address
and
that
we
have
at
the
end
two
bytes
for
the
frame
check
sequence.
So
you
can
see
that
the
1504
headers
is
already
taking
up
a
lot
of
space
of
this
170
127
part.
P
And
for
me
you
can
see
that
we
have
yeah.
The
outer
most
type
length
field.
That
I
set
is
the
interest,
and
then
the
whole
name
is
which
is
blue
green.
We
have
here
that
can
be
prefixed,
boolean
value.
Let's
say
it
here
and
then
there
must
be
fresh
boolean
value
and
then
the
nonce
interest
lifetime
table
limit
and
that's
it
and
from
there
you
can
see
the
encoded
interest
below.
P
P
And
if
you
also
look
at
the
just
look
at
the
type
and
length
of
the
meta
info,
then
we
can
see
that
we
almost
we
su
removed
80
percent
of
type
and
off
of
meta
information
by
using
missing
from
encryption.
So
with
no
encryption,
compressing
scheme.
So
yeah.
Let's
look
at
the
implementation.
What
we
have
currently
well
as
a
write,
develop
I
am
biased
here
so
I'm
using
write
as
a
lower
stack
and
Esalen.
P
What
happened
a
1504
excess
and
what
we
did
is
or
what
County
actually
currently
doing
is
refactoring
the
sixth
open.
We
have
6
open
and
write,
but
we
this
highly
highly
monolithic
and
we
try
to
pull
out
the
refactor,
the
presentation
and
the
dispatch
dispatch
in
framework
so
that
I
can
have
easy
access
from
the
ICN
world
and
use
CC
and
light.
In
this
example,
here
over
a
binary
which
is
here
called
lip,
ICML
I
have
dope
sauce.
Well,
the
open
source
get
up
here.
P
The
next
steps
we
have,
or
some
colleagues
are
working
on,
Bluetooth
Low,
Energy,
bringing
this
into
into
right
or,
and
also
the
Laura
yeah
elaborate
more
on
this,
and
once
we
have
this,
I
want
to
focus
on
yeah,
bring,
also
ICN
and
or
in
the
end,
CCN
meditation
sound
onto
these
layers
so
or
to
these
technologies
for
ble
and
fedora,
and
after
that
they
are
also
other
state
listed
compression
schemes.
Currently
we
just
looked
at
removing
the
type
and
length
fields
which
is
pretty
simple,
but
also
you
can
do
like
here.
P
Lentils,
if
encoding
or
you
can
use
happen
encoding
on
top
of
this,
to
reduce
maybe
more
to
reduce
packet
side
more
and
then,
of
course,
which,
where
the
most
game
is
in
the
safe
of
compression.
So
we
are
also
looking
into
this,
and
also
we
worked
on
this
for
the
ICN
right
Agathon
and
at
the
weekend,
so
that
we
can
assume
pre,
name
prefix
and
just
remove
it
or
write
the
name
prefix
and
the
IOT
nodes
are
aware.
So
this
is
the
same
concepts.
P
Concept
is
done
by
vi
open
and
where
they
just
remove
an
IP
prefix
and
they
assume
it
and
also
there's
we
can
leverage
the
pit
table
because
we
have.
We
already
have
state
that
we
have
when
we
send
interest
out
and
receive
the
data,
so
we
can
probably
also
yet
cut
apart
form
from
the
name
yeah,
the
name
in
the
data,
because
it's
already
stored
in
the
tables.
P
This
is
also
what
we
are
looking
into
currently
and
a
couple
of
days
ago,
I
got
I,
got
a
point
that
mock
Moscow
also
had
to
talk
about
the
head
of
compression
in
bonus
iris,
so
I
try
to
get
in
touch
with
him.
I
see
him
here
in
the
miracle
yeah
and
we
emailed
and
let's
see
what
happens
there,
yeah
and
that's
the
last
slide.
So
any
questions.
F
Q
Yes,
yes,
okay,
great
all
right!
Well,
thank
you
very
much
for
the
timeslot.
Today
we
got
some
feedback
on
the
oh
six
drafts
of
the
semantics
and
messages
documents.
So
we
made
a
few
editorial
changes
to
the
drafts
to
address
those,
and
this
slide
can
summarizes
the
high
level.
What
happened?
The
security
considerations
you
know
got
a
significant
rewrite
based
on
the
feedback.
G
Q
Rtf
documents
and
some
mandatory
text
that
wasn't
in
there
so
that's
been
added
in
and
then
on
the
content
side.
There
were
some
comments
that
the
drafts
didn't
provide
any
background
or
motivation
for
what
they're
trying
to
address.
So
we
added
some
new
language
around
this
and
then
fixed
a
few
technical
style,
spelling
and
technical
content
questions.
Q
Q
So
this
table
goes
over
the
detailed
changes
it
says
kind
of
what
the
changes
are,
but
it
doesn't
say
you
know
the
the
text
of
the
changes.
You
know
that
that's
in
the
document
which
is
uploaded
to
the
data
tracker
now
so
in
the
semantics
document,
the
abstract
got
some
mandatory
text
about
being
an
icy
NRG
document
and
the
same
for
the
messages
draft.
Q
Q
We
cited
mdn
as
a
related
ICN
protocol,
there's
about
one
paragraph
that
that
tries
to
nail
down
what
the
underlying
differences
between
CC,
n
and
n
DN
and
view,
which
is
basically
discovery
at
the
network
layer,
and
we
also
cited
the
selectors
draft
as
an
example
of
doing
an
upper
layer
protocol
for
discovery
in
CC,
n
1.0
and
a
paragraph
on
that.
This
draft
is
a
consensus
document
from
the
research
group
and
we
added
a
four
pair
and
four
paragraphs
with
a
high-level
walkthrough
about
kind
of
how
the
protocol
operates
in
the
messages
document.
Q
We
added
just
a
quick
background
and
cite
the
semantics
document
for
everyone
to
look
at,
and
we
also
cited
the
79
27
as
context.
We
added
a
paragraph
about
the
two
to
TLV
choice
and
the
reason
why
we
chose
that
TLB
encoding
and
we
also
cited
work
on
header
compression
as
a
way
to
address
the
overhead
for
links
that
really
need
that
and
then
the
mandatory
paragraph
I
have
being
a
consensus
document
in
the
semantics
document.
We
added
a
new
section
on
architecture,
so
this
describes
what
a
typical
network
would
look
like.
Q
Yep,
okay,
it
just
takes
my
screen
a
little
moment
to
refresh
so
then
the
security
section
as
I
mentioned,
is
a
complete
rewrite,
and
so
we
added
some
text
about
the
these
drafts
are
for
a
layer,
3
protocol
with
authentication
and
not
encryption.
So
you
know
just
to
kind
of
scope
it
so
you'll,
because
there
were
some
questions
about
well.
Q
What
about
encryption,
the
and
then
we
provide
some
usage
guidelines
or
integrity
checks
and
authentication
codes
and
signatures,
especially
in
interest
where
those
could
be
an
attack
vector
a
computational
attack,
vector
we
mentioned
that
we
don't
talk
about
how
one
arrives
at
keys
or
trust
keys.
We
do
cite
the
key
exchange
protocol
is
one
way
to
exchange
keys
and.
Q
We
discussed
some
of
the
security
issues
around
TLV
encodings
of
data
packets,
such
as
aliasing
values
that
need
for
schema
value
at
Val,
adacia
n--
and
the
potential
issue
of
per
processing
of
sorry
per
hop
latency
due
to
TLB
processing,
which
is
you
know,
most
most
protocols
when
they
use
TLD
it's
in
an
end-to-end
and
and
not
per
the
there's.
Also
an
extended
discussion
of
caching
and
cache
poisoning
and
the
motivation
for
the
rules
on
cache
behavior,
which
are
in
the.
N
Q
Q
Q
So
please
read
these
updated
sections
and
get
me
feedback,
and
you
know
if
there's
other
citations
that
should
be
in
there
or
other
background
that
should
be
in
there.
Please,
you
know,
recommend
it
and
I'll
put
it
in
as
appropriate,
and
you
know
I'd
appreciate
any
feedback.
If
the
ndn
comparison,
paragraph
is
appropriate
and
sufficient.
You
know
I
tried
to
give.
What
I
saw
is
the.
I
Q
Q
Okay
next,
so
what
I
would
like
to
do
is
either
resubmit
to
IRS
G
or
do
an
OE
draft.
So
you
know
if
over
the
next
two
weeks,
you
could
get
me
feedback,
and
then
you
know
in
coordination
with
the
chairs,
we'll
figure
out.
You
know
either
resubmit
or
you
know,
bump
it
of
revision
as
appropriate,
and
if
you
need
more
time
to
read
the
things
just
let
me
know
you
know
we
can
give
you
more
time
to
read
it.
Q
D
O
G
To
the
summary
of
updates
update
to
this
draft
so
we're
we
are
looking
at
the
ICN,
so
basically
the
the
3gpp
is
5
GC
architecture
and
see
how
ICN
can
be
enabled
in
in
some
native
form
in
it
in
this
architecture.
So
I
will
not
go
because
this
the
main
difference
and
they
I
will
just
go
through
the
draft
updates
year.
So
main
draft
updates
is
that
we
had
who
is
contributed.
G
His
IP
over
ICN
work
into
this
draft
to
basically
show
that
ok,
you
can
basically
enable
new
kinds
of
different
types
of
icing
architectures
into
the
fisa
framework.
So
the
one
thing
is
that
I
changed
this
ICN.
We
we
had
this
I
see
an
anchor
point.
We
call
it
isin
gateway
to
avoid
any
misinterpretation,
basically
saying,
there's
no
real
anchor
point
from
the
ICN
perspective,
discussion
yeah.
So
we,
the
main
difference,
is
that
we
added
more
details
on
how
we
could
leverage
the
non
IP
p
do
provision
for
I,
CN,
n
5g.
G
Again
we
have
discussion
on
this
one
on
what
I
call
with
Dirk's
contribution
on
IP
on
over
ICN
realization
of
ITC
and
there's
been
update
to
some
of
the
use
case
scenarios.
So
here
we
have
one
use
case.
We
have
the
use
case,
considering
a
typical
ICN
kind
of
assisi,
an
Indian
kind
of
architecture,
but
we
have
an
extension
to
also
see
how
IP
over
ICN
can
basically
do
the
same
thing
for
smart
mobility,
and
there
was
another
contribution
on
multi-user.
G
We
are
used
case,
so
this
is
again
saying
that
you
could
fundamentally
use
some
of
the
characteristics
of
the
ER
application
to
that
could
fundamentally
benefit
from
a
nice
en
deployment,
so
the
mean
just
to
recap,
the
main
traffic
objectives
is
is
to
see
how
I,
CN
n
able
to
basically
see
how
I
different
I
features
of
ICN
can
be
used
to
address
a
lot
of
the
challenges
of
the
5g
application.
So
we
probably
know
people
know
this.
G
You
have
EMB
BMPs
in
your
LC,
so
a
lot
of
challenging
requirements
there
and
see
how
I,
see
and
I
see
in
Franz
mode
can
help
that.
So
the
first
thing
was
to
just
because
you
know
you,
you
have
to
look
into
this
architecture
and
see
how
IC
and
fits
in
so
that
main
thing
was
to
look
into
the
architecture
and
and
basically
see
how
ICN
can
be
plugged
in
and
the
main
flexibility
comes
from
the
different
principles
that
v
GC
has
followed
in
this
in
their
architecture.
G
G
Today
we
have
like
we
added
some
new
control
functions
to
fundamentally
allow
I,
see
a
new
ways
to
connect
to
the
network
and
finally,
do
some
ICN
session
management
to
the
ICN
gateway,
so
that
you
allow
this
IC
and
PDUs
to
flow
between
the
ICN
Yui
and
and
to
the
ICN
gateway.
And
then,
of
course,
since
we
do
its,
we
don't
want
it
simply,
not
lower
lace.
So
you
could
transport
those
ICN
as
basically
IP
based
PDUs.
H
G
G
Native
format,
too
right
so
so
coming
to
this
non
IP
p
do
considerations
so
I
mean
advantage
here
is
that
you
can
enable
I?
See
in
you
is
you
know
whether
they're
smart
devices
or
any
kind
of
an
IOT
devices
without
any
IP
configuration
right
so
and
the
maximum
efficiency
comes
by
enabling
the
ICN
gateway
in
the
ran
itself
right.
So
this
allows
you
to
fundamentally
maximize
efficiency,
what
you
get
from
from
caching
and
mobility
and
so
far
within
the
5g
network.
So
we
added
a
lot
of
considerations
here.
G
This
ends,
like
probably
they're,
more
about
unanswered
question
of
how
you
can.
How
are
the
different
things
that
needs
to
be
handled
or
to
be
addressed
for
this
to
work,
so
one
is
from
the
from
the
attach
procedures
and
the
control
plane
for
for
utter
pain
for
the
Yui,
with
non
IP,
PD
unknown,
IP
video.
So
here
you
could
I
mean
this.
Probably
you
know
you
could
use,
have
existing
mechanisms
to
do
it
Superfund
only
from
the
from
the
device
authentication
perspective,
but
again
with
ICN
operator
is
actually
running
an
IC
o----
network.
G
Fundamentally,
it's
not
about
transporting
bits.
He
is
actually
providing
accessibility,
certain
services,
so
probably
you
have
to
go
down
a
level
lower
to
fundamentally
when
authorized
to
basically
authentic
services
and
fundamentally,
if
you
are
doing,
services
then
probably
are
dealing
with
naming
and
and
basically
other
things
like
the
test
mechanisms
and
so
on
so
forth.
So
there
is
from
the
control
plane
management.
There
has
been
some
work
from
the
LTE
to
fundamentally
handle
this
kind
of
scenarios
to
handle
non
IP
PD.
G
You
can
leverage
some
of
those
features
here
too,
from
the
user
plain
perspective,
as
I
said
like,
if
you
really
bring
in
the
ICN
gateway
route
towards
that
you
know
to
be,
you
don't
have
any
more
of
that
anchor
point
kind
of
architectures
where
you
can
do
the
centralized
policy
or
charging
or
legal
interception
kind
of
things
which
are
fundamental
requirements
in
these
networks,
so
that
has
to
be
handled
in.
Basically,
how
do
you
do
when
you
are
actually
having
a
more
distributed?
G
G
Then
again,
you
have
probably
you
know
there
probably
I
think
these
are
all
different
ways.
You
can
identify
the
IC
and
PDUs,
so
you
have
other
identifiers,
and
if
you
read
this
draft
there
are
so
many
identify
as
you
can
use
in
different
contexts
to
do
it.
So
the
other
one
comes
from
the
slice
ID
itself
exactly.
How
do
you
take
dice
in
people
use
and
give
it
to
different
eyes
in
slices,
so
the
other
other
consideration
is
with
respect
to
mobility
handling.
G
So
again
it
doesn't
make
any
sense.
Basically,
you
bring
the
non
IPP
deal
ICN
in
its
native
format
to
the
you
know,
D
and
then
put
it
on
a
basic
I
I
see
tunnel.
Rather
you
you
probably
want
to
handle
everything
without
any
of
any
kind
of
anchor
points
in
the
network.
So
so,
as
I
said,
it's
efficient
to
look
co-locate
the
icin
gateway
with
iran,
but
again
with
iran,
you
you
have
other
than
the
other
control
plane.
If
you
want
to
handle
mobility,
you
might
need
some
ice
in
base
control,
plane
support.
G
So
then
you
have
to
ensure
that
that
control
planes
kind
of
integrate
with
the
five
g's
control
plane
to
ensure
the
ice
in
mobility
works.
Routing
considerations.
So
typically,
you
know
you
know
today
is
5g
networks,
pretty
much
do
kind
of
centralization
management
and
routing
right.
So
again,
when
you're
having
this
kind
of
flat,
I
seen
realization,
then
you
do
scalability
problems.
So
then
you
so
something
like
you
know
to
explore
the
feasibility
of
how
distributed
routing
protocols
can
work
in
this
scenario
is
something
to
consider
so
I
think
here
at
somewhere,
I.
G
What
I
understand
is
a
3gpp
is
going
in
this
phase,
with
allowing
more
kind
of
mesh
connectivity
with
some
current
segment,
routing
and
stuff
like
that
without
gtp
and
stuff.
So
probably
this
I
mean
this
would
be
another,
so
it's
probably
going
to
ICN.
You
could
take
some
of
those
principles
to
see
how
those
control
planes
are
supported.
Okay,
so
this.
G
G
Yeah,
okay,
so
the
main
thing
here
is
that
I
mean
to
come
to
this
point
on
next
steps
is
how
do
you,
so
some
of
the
things
that
we
have
outlined
here
is
I
mean
is
I.
Think
one
of
the
benefits
that
we
had
a
presentation
during
the
interim
was
to
see
how
IC
and
multi
casting
can
benefit
on
radio
access.
So
something
like
we
could
investigate
more
put
more
details
around
that
into
this
draft
or
again
non
IPP.
Do
as
I
said,
there
are
more
open
questions
there.
G
Those
can
be
further
addressed
separate
so
again.
The
other
thing
is
that
how
do
you
handle
non
IP,
PDUs,
particularly
saying
for
IOT
or
smart
devices?
Those
can
be
handled
in
a
more
detailed
manner,
mobility
handling
over
ICN.
In
fact,
you
see
we
have
a
seamless
probability
kind
of
section
discussing
that
managing
queues
for
heterogeneous
applications.
Again,
you
are
talking
about
flat
IC,
and
then
you
have
to
think
about
us
from
the
ICN
perspective
and
distributed
policy
charging,
and
this.
G
To
be
extracted
as
well
so
anyway,
so
the
main
idea
is
that
I
mean
the
key
thing
is
that
this
particular
work
could
influence
how
how
fight
or
3gpp
could
view
ICN
as
most
must
be
a
potential
kind
of
a
network
that
can
be
enabled
in
the
future
releases,
considering
that
five
years
till
four
five
years
out,
and
so
that's
a
kind
of
objective.
We
want
to
have
kind
of
going
further
with
this
trap,
so.
G
So
the
main
thing
is
that
it
would
have
to
have
good
to
have
some
kind
of
concrete
feedback
on
unless
there
is
any
particular
set
of
problems
that
you
see
would
be
more
beneficial,
I
mean
from
the
people
who
are
more
active
from
the
TCP
perspective,
or
also
who
is
following
this
particular
set
of
candidates.
Okay,.
D
D
R
R
Snots,
first
of
all,
I
would
say:
I
agree,
not
the
plantation
from
Professor
Jung
and
in
her
presentation
about
the
home.
Iot
I
would
like
to
call
eight
and
computing
devices
in
my
transition,
and
so
I
can
agree
with
the
transition
of
the
consideration.
I'll
apply,
ICM
to
and
computing
and
in
this
partition
is
related
to
the
draft
to
us
on
material.
Recently,
the
iPod
ending
of
this
drop
during
the
master
page
of
the
transition.
Okay,.
R
Okay,
okay,
first
of
all,
let's
look
at
this
post.
Well,
information,
static,
networking,
I'm
sure
there
are
not
experts
here,
so
I'm
not
going
to
take
none
time
to
talk
about
it.
So
yeah,
it's
named
based
routing,
distributed
local
communication
information
based
security
and
multicast
attention
forwarding
system.
So
all
that
to
say,
ICN
takes
very
important,
a
technological
revolution
and
a
fundamental
change
in
communication
and
networking,
and
they
will
we
look
back
another
side
over
the
add
computing,
it's
close
to
users.
R
Of
course
it
gave
us
no
latency,
it
has
data
pre-processing
and
it
is
decentralized
system
non,
con
computing
storage,
connectivity
and
one
of
the
interest,
and
maybe
the
promising
directional
rates,
maybe
the
intelligence
on
edge,
so
edge
computing
is
also
broadly
divided
among
all
TR,
the
some
other
industrial.
What
cost
such
as
a
smart,
great
sewing
way
just
the
kind
of
force.
I
me
feel:
okay,
post
technology,
they
have
on
studying
performers
and
we
have
the
feeling,
maybe
there's
some
relative
or
seminar
properties
of
books,
or
maybe
there
any
chance.
R
We
can
use
icing
to
benefit
and
computing,
so
okay,
opportunities
of
applying
I,
say
and
to
add
computing.
First
of
all,
I
see
an
enable
to
have
fake
converters
because
in
a
computing
there
are
not
of
devices
deployed
so
ATO
a
course.
A
considerable
fun
asari,
considerable
uplink
traffic
on
edge
and
this
kind
of
a
traffic
also
requires
some
flattening
service.
R
So
we
lead
the
traffic
is
more
convergent
and,
of
course,
the
functionality
complimentary
part
were
post
and
computing
and
ICN
they
have
some
relative
of
similar
properties,
but,
however,
they
also
have
some
different
properties
can
be
complimentary
night.
We
put
night
computing
on
edge,
which
means
we
give
it
intelligence.
R
Maybe
this
can
help
I
see
an
ad,
give
us
win-win
approach
and
benefit
force,
so
he
gets
the
maximum
performance,
and
when
we
talk
about
I,
say
this.
One
thing
we
always
need
to
talk
is
its
deployment,
so
is
8
worthy
or
is
it
a
realistic
to
deploy?
Oh
I
see
another
homeland
walk.
Well,
maybe
we
don't
need
to
deploy
a
tinhorn
Network.
You
all
only
needed
to
deploy
it
on
act
which
can
reduce
the
whole
network.
Deployers
cost
and
complexity
and
makes
the
network
more
flexible.
R
R
R
The
data
I
collected
starless
data
here
and
when
this
kind
of
data
storage
in
the
Android
devices
or
endnotes,
it
can
also
cause
issue
of
a
storage
limitation
of
the
unknown
and
devices
and
the
at
computing
traffic
congestion,
and
we
said
before
there
are
a
huge
uplink
flow
traffic
and
is
correct.
Raffe
ik
can
cause
congestion
and
the
security
part.
R
Yes,
although
it's
and
devices
it
is
also
connected
with
other
devices,
all
the
cloud
well,
it
has
also
made
to
face
the
risk
from
a
host
attack
and
just
like
the
the
temperature
control
issues,
we've
set
it
up
in
the
previous
transition
before
I
think
because
the
devices
are
edge,
ie
expose
the
two
users,
so
we
may
lead
a
very
strong
Authority
system
and
well
that's
icing
on,
say,
sailor.
Radio
venues
addresses
and
the
course
over
the
whole
network,
I
say
and
deployment.
Maybe
is
too
much.
R
R
R
Well,
because
I
see
a
network,
it's
name
based
a
security.
So
in
this
kind
of
case
way,
I
don't
need
to
worry
about
the
hosts
attack
attacked
and
because
it's
self-certify
of
certain
model,
this
can
help
strengthen
the
authority
model
of
edge
devices
or
not.
Well,
that's
icing
or
seasonal
venues
addresses.
R
R
Ok,
this
is
my
email
address
and
this
is
I
think
of
the
draft
and
we
need
more
reviews
and
feedback
and
input
and
contributions
for
draft
really
appreciated
our
welcome.
Ok,
thank
you
guys
still
here
and
my
presentation.
I
know
it's
the
pie
tonight
so
I
give
you
this
smile
for
you
guys
the
boughs
of
this
card.
It's
my
count.
It
has
nothing
to
do
with
that.
I
say
I.
Just
now,
catchy
not.
D
K
What
is
the
definition
of
a
computing
here
and
also
in
a
draft?
What
is
the
definition
of
ICN
here,
because
none
of
them
are
defined
in
the
draft?
So
it's
difficult
to
understand
exactly
what
you
mean
with
ICN,
what
you
mean
exactly
that
computing,
so
it's
difficult
to
understand
what
problems.
So,
what
is
what
kind
of
a
she
and
technology
is
bringing
what
kind
of
advantages
to
a
to
computing,
because
you
don't
know
what
what
you're
talking
about
I
mean
what
it's
not
the
time,
I
can't!
K
R
R
D
D
S
Hi,
my
name
is
Ben
stone,
each
and
one
of
the
co-authors
on
the
draft
for
native
deployment
of
ICN
in
4G
LTE
mobile
networks,
so
as
an
introduction
trying
to
gain
some
time
here
as
an
introduction,
the
draft
was
about
holistic
view
of
native
IC
and
deployment
specific
to
the
4G
network.
Some
of
its
aspects
would
apply
to
patch
in
the
future,
hopefully,
but
this
is
dealing
really
with
the
different
scenarios
of
what
happens
if
ICN
is
enabled
or
is
not
enabled
in
certain
particular
parts
of
the
network.
S
So
we
are
looking
at
the
options
here.
What?
If
what
if
UE,
is
a
dual
capable?
So
if
you
have
to
deckhand,
has
the
choice
to
make
and
none
of
the
other
nodes
in
the
network
are
ICN
enabled
on
another
hand,
what
if
I
see
an
enable
through
that
and
the
Yui
is
only
IP
capable.
So
the
idea
of
the
draft
is
to
go
through
different
scenarios
and
explain
what
it
was
doing.
What
was
what
what
needed
to
be
changed
with
some
specific
proposals
of
the
introduction
of
the
transport
kamuran
Slayer?
S
S
Major
changes
I
will
show
on
the
next
slide,
there's
a
three
of
them,
I
believe
and
as
far
as
minor
changes
in
addition
to
Dirk,
who
was
present
from
before
we
rally
was
listed
as
a
co-author
on
the
draft
and
then
also
we
have
a
pair
of
Fry's
some
of
the
text
and
updated
new
references
after
some
comments
like
that,
for
example,
the
QoS
section
has
been
clarified.
We
have
defined.
We
have
identified
that
certain
parity
needed
to
be
present
between
the
IP
and
ICN.
S
Now,
will
it
be
changing
principle,
or
will
it
be
or
or
not
this
beyond
the
point?
The
thing
is
that
the
the
features
such
as
prioritization,
such
as
policing
or
shaping
in
some
shape
or
form
would
need
to
be
present.
Also,
there
was
some
minor
changes
in
the
security
section.
Security
section
has
not
really
has
listed
the
security
concerns.
Now
we
have
that
done.
However,
a
certain
more
research
would
need
to
be
done
and
I
also
seen
in
one
of
the
previous
presentations
that
were
marks.
I
believe,
would
be
dealing
in
details
about
security.
S
As
far
as
the
major
changes
to
the
document,
so
we
have
added
a
reference.
There
was
a
comment
that
there
is
no
distinction
between
live
content
like
broadcast
services
and
and
on
the
main
content.
So
we
have
added
reference
to
the
broadcast
and
multicast
technologies,
e/m
BMS,
specifically,
so
that
we
would
handle
the
multicast
deliver
in
a
pre-planned
haunt
and
so
with
a
light
content
versus
that
one
that
would
be
on
demand.
In
this
example
also,
there
were
many
references
in
the
document
prior
to
this
revision
and
the
end.
S
S
Now
this
expect
to
be
normalized
once
the
whole
harmonization
effort
has
been
completed,
but
this
is
on
our
side.
We're
not
dealing
more
on
the
CC
NX,
as
opposed
to
mixing
and
also
the
change
was
that
after
one
comment,
we
he
removed
the
transport
convergence
layer
which,
as
a
reminder,
just
is
responsible
for
dealing
with
specifics
of
choosing
the
right
protocol,
depending
on
the
author
of
its
presence
in
the
next
hop
or
in
the
transport.
S
So,
for
example,
if
the
if
the
UE
is
dual
capable,
the
transport
is
not
you,
it
would
need
this
player
would
make
a
decision
which
protocol
to
use
and
would
pick
the
right
protocol
based
on
the
presence
on
the
next
one.
So
this
has
been
removed
from
the
protocol
step
in
a
PG
8
way,
and
the
reason
for
that
is
in
this
case,
with
dual
stack
ICN
deployment
in
Yui.
We
used
to
have
transport
conversions
there,
so
the
right
thing
would
be
to
have
transport
convergence
layer
between
Yui
and
on
the
CDN
side.
S
S
Cisco
Cisco
Systems,
we
will
be
using
CC
annex
in
the
code
and
that
is
being
in
the
process
of
being
set
up.
So
the
future
revisions
should
have
some
results
from
from
the
trials
from
empiric
results,
and
also
we
need
to
now.
This
is
not
necessarily
the
next
section
or
area
for
future.
Graph
updates
could
also
be
new
experimental
read,
so
not
that
this
won't
all
necessarily
be
in
this
draft,
but
are
certainly
some
questions
that
need
to
be
answered
in
the
future,
such
as
investing
you
have
to
meet
the
IP
QoS
requirements.
S
Just
like
I
mentioned.
It
has
to
be
met.
Some
way
in
ICN,
then
also
in
investigate
the
impact
of
cups,
the
control
and
user
plane
separation
and
traffic
offloading
at
the
edge,
so
that
we
can
optimize
the
user
playing
traffic
paid
paths
using
ICN,
which
is
the
whole
purpose
of
this,
with
UPF
being
closer
to
the
edge
to
investigate
the
realization
of
transport
convergence
layer
which
we
are
introducing
here
as
a
concept,
and
our
realization
needs
to
be
further
investigated,
then
impact
analysis,
ICN
and
mobility
management
messages,
structures
and
flows.
S
This
goes
from
the
fact
that
in
this
draft,
we
suggest
that
there
will
be
no
immediate,
significant
benefit
and
control
plane.
This
needs
to
be
further
investigated
to
see
if
there
would
be
possibly
in
the
future,
some
benefit
or
further
benefit
from
ICN
in
the
control
plane.
The
draft
deals
mostly
with
user
plan
messages.
S
We
want
to
investigate
how
I
see
an
IP,
intro
working
gateway
function
would
perform
the
conversion
between
ICN
and
IP
primitives
for
data
retrieval
in
the
draft.
We
are
dealing
with
this
at
the
end
when,
when
a
p
gateway
is
IC
incapable,
but
CDN
is
not
in
that
case,
some
level
of
interworking
function
would
need
to
be
introduced
and
we
need
to
investigate.
How
would
that
work
here
and
then
certain
aspects
very
important
aspects.
S
The
mobility
networks
need
to
also
be
examined,
how
they
would
react
to
ICN
and
how
they
would
perform
I'm
talking
about
lawful
in
action
billing
mediation
also
Network,
slicing
and
provisioning,
a
PR
api's
requirements.
So
these
are
all
concept
that
would
need
to
be
investigated
in
this
draft
or
in
some
other.
Thank.
M
D
D
D
Ok,
so
I
hope
you
haven't,
met
any
plans
for
a
September
yet
because
there
will
be
many
cool
events
there.
So
there's
would
be
the
riot
summit.
September
13
14
in
Amsterdam,
so
right
is
the
IOT
of
open-source
operating
system.
That
is
also
working
on
CC
and
night,
for
example,
and
it's
also
I
some
ndn
and
riot.
B
We'll
probably
continue
our
tradition
of
scheduling,
an
IC
n
RG
meeting
the
day
after
the
conference,
because
that's
worked
very
well
in
the
past
and
researchers
who
come
to
the
conference,
who
normally
don't
participate
directly
in
this
community,
have
an
opportunity
to
present
their
research
and
interact
with
the
rest
of
the
community.
So
well,
we
don't
have
that
nailed
down.
People
should
plan
one
on
that.
It
would
be
after
the
conference,
as
opposed
to
before.