►
From YouTube: IETF99-ThursdayLunchSpeakerSeries-20170720-1230
Description
THURSDAYLUNCHSPEAKERSERIES meeting session at IETF99
2017/07/20 1230
https://datatracker.ietf.org/meeting/99/proceedings/
A
Takes
that
have
some
really
cool
and
stuff,
and
you
have
some
sites
that
are
still
in
the
old
days
and
where
eighties
finally
have
legacy
vendors
from
STI
world
or
in
outputting
IP
interfaces
in
their
by
plants,
and
so
it's
an
interesting
time
and
it's
a
time
I
think
that
you're
going
to
start
seeing
more
and
more
people
come
from
that
traditional
media
entertainment,
world
and
their
devices,
their
hardware
and
their
protocols.
They're
gonna
start
coming
here
to
the
ITF
and
I
think
this
is.
This
trend
has
already
begun
with
a
few
people.
A
People
like
myself,
and
so
we
thought
we'd
give
a
bit
of
a
talk
and
a
bit
of
background
and
perspective
from
a
couple
different
views
on
where
we're
at
right
now
what's
been
going
on
in
that
transition
and
what
the
transition
looks
like
to
hold
for
us
in
the
future.
So
for
that
perspective,
let
me
introduce
my
panelists,
the
first
one
he
is
garage
nake.
He
is
a
professor
at
the
directory
university.
A
His
interests
span
both
computer
networking
and
cyber
security,
and
much
of
his
work
has
been
on
exploring
the
design
of
network
applications
protocols
and
suffer
Defined.
Networking
he's
also
a
veteran
of
the
ITF.
He
works
in
segment
brownie
and
a
few
things
in
bits
and
bytes
and
salsa,
or
our
GG
demos.
Groff,
has
been
heavily
involved
with
myself
and
others
on
working
on
some
of
the
GG
stuff
we've
been
playing
with.
A
We
also
have
Craig
Taylor
Craig
is
the
lead
Technical
Architect
for
the
BBC's
online
technology
group,
they're
responsible
for
the
design
and
performance
of
BBC's
media
delivery
platform
and
the
internet
edge
of
the
BBC
web,
often
working
its
lays
out
between
research,
colleagues
and
engineering.
Craig
contributes
to
video
audio
and
internet
standards
and
transitions
them
into
production
platforms.
Craig
provides
development,
details
for
transport
applications,
security,
core
protocols
across
the
DBC
and
finally,
but
not
least,
we
have
mark
mark-
is
a
cisco
fellow
he's.
B
Hello,
everybody.
Can
everyone
hear
Marvis,
yeah,
hi
I'm?
Don't
do
this
very
often
so
Paula
geez
for
the
obvious
anxiety
but
I'm
Craig
Taylor
work
for
BBC
online
technology
group.
You've
just
heard
I
don't
have
any
slides,
they
thought
I'd
spare
you
all
of
that.
So
I
do
have
a
short
video
and
and
a
lot
of
words,
but
will
then
plow
through
them
quickly.
So
here
the
ITF
I've
been
coming
since
ITF
89
you'll
typically
find
me
in
working
groups
such
as
HTTP
in
quick,
but
also
some
of
the
transport
areas.
B
My
unique
specialism
is
typically
that
I
don't
have
one
I'm,
a
massive
generalist,
so
I
apologize
for
stomping
all
over
anyone's
area
with
light
level
detail
at
the
BBC,
we're
a
public
service
broadcaster,
so
we're
based
in
the
UK,
which
is
obvious.
We
deliver
our
contents
worldwide
and
we
have
a
strong
relationship
with
our
audiences,
which
is
based
primarily
on
trust,
so
being
a
public
service
or
an
broadcaster.
We
have
different
biases
than
perhaps
some
of
the
attendees
that
may
come
here
regularly.
B
B
A
A
B
There
audio,
in
which
case
I'll
talk
over
it,
and
so
this
demonstrates
captured
atoms
which
are
being
rendered
in
real
time
on
an
edit
suite
we're
able
to
identify
individual
presenters
to
provide
things
like
accessibility
benefits.
B
The
presenter
is
just
one
video
atom
that
can
be
interchanged,
just
as
the
tiling
at
the
back
is
where
you
can
provide
a
high
contrast,
display
a
the
visually-impaired
subtitling
and
when
moved
on
to
the
screen
can
actually
interact
with
other
video
dogs
so
that
you
can
render
either
as
part
of
the
edit
process
or
even
on
the
playback
device
itself
the
appearance
and
actually
have
a
longer
version
of
this
as
well,
which
shows
a
portrait
style
viewport
where
the
high-value
objects
I,
the
presenter
can
be
rendered,
whilst
the
viewport
is
changed
to
meet
the
the
target
device
so
think
of
it
as
almost
what's
the
word
sorry,
words
foaming,
so
it's
similar
to
yet
sort
of
the
way
that
we
render
web
pages
these
days,
you
can
render
them
to
the
specific
viewport
retaining
those
assets.
B
B
B
So
this
is
where
I
get
back
onto
some
of
my
more
but
things
to
largely
do
distribution,
if
I'm
honest
the
distribution
challenges,
we
have
already
are
as
a
again
a
public
service
broadcaster
and
are
that
you,
you
can't
always
guarantee
that
the
audience
in
certainly
the
UK
has
a
data
line.
That
is
real
that
has
enough
bandwidth
to
receive
all
of
the
media
over
a
unidirectional
flow.
B
So
we
require
to
be
flexible
on
the
type
of
access
networks
that
we
might
use
in
the
future
that
something
like
LTE
broadcast
may
be
essential
to
reach
our
reach,
our
audiences
and
so
we're
trying
to
be
both
flexible
with
the
distribution
mechanism,
but
also
to
improve
upon
the
way
that
we're
doing
that.
So
some
of
our
work
here
at
the
ITF
is
to
try
and
engineer
a
common
layer
7
which
is
independent
of
the
access
method,
which
we've
had
some
conversations
here
and
we
hope
to
have
a
draft
refreshed
on
soon.
B
We
also
have
very
different
client
failure,
constraints
and,
in
the
broadcast
world,
compared
to
regular
sort
of
web
delivery
it
on
a
large
screen
device.
People
are
very
intolerant
of
video
rebuffering
or
am
stuttering
more
so
than
perhaps
they
are
in
a
web-based
world,
so
that
consistent
nature
is
is
difficult
to
achieve
in
a
way
that
is
sort
of
homogeneous
across
devices.
B
If
you
scale
those
numbers
up
to
UHD
at
24,
megabits,
you've
got
that
works
out
at
about
588
terabits
concurrence
from
one
content
service
provider.
If
you
translate
that
into
commodity
servers,
40
gig
attached
in
a
truly
cached
model,
that's
close
to
15,000
servers
if
they
are
deployed
with
near
100%,
efficiency
and
UHD
is
where
we
are
today.
B
A
C
So
Glen
I
think
as
a
media
production
shifts
where's
IP,
there's
kind
of
an
opportunity
to
go
IP
end-to-end
and
in
media
and
entertainment.
End-To-End
means
from
the
lens
of
the
camera
all
the
way
to
the
viewer
and
in
their
screen
and
Mark
said
that
ipv6
is
a
platform
for
innovation
and
I
completely
agree.
For
example,
today,
there's
a
very
large
content
provider.
That's
assigning
ipv6
addresses
to
containers
in
their
data
center.
What
we're
thinking
about
is
what,
if
those
containers
were
actually
pieces
of
data,
video
or
content
right.
C
So
today
we
take
a
video,
we
break
it
down
into
individual
files
and
we'll
put
them
on
a
CDN.
What
if,
instead
we
take
the
video
we
break
it
down
into
individual
chunks
that
can
be
addressed
using
ipv6,
and
if
we
do
that,
we
can
take
advantage
of
the
speed
the
simplicity
and
that
ipv6
routing
gives
you.
We
also
increase
this
sort
of
visibility
across
the
entire
media
production
distribution
chain
by
having
that
metadata
at
the
IP
layer.
C
So
later
today,
we'll
be
showing
kind
of
a
live
demo
of
this
concept
and
you're
seeing
a
screenshot
of
the
live
demo.
Above
what
we
did
will
be
done
is
taken
three
movies:
we've
broken
them
down
into
individual
chunks
and
prepared
them
with
ipv6
in
mind.
We
deployed
them
to
two
CD
ends
and
the
two
CD
ends
are
on
the
right
side
of
the
screen.
C
Cd
N,
1,
n
2
and
the
both
CD
ends
are
using
BGP
to
signal
their
reach,
ability
to
that
individual
content
and
we'll
show
you
how
and
we'll
demonstrate
how
that
content
can
be
addressed
and
access
directly,
using
ipv6
by
the
three
viewers,
the
three
screens
on
the
left
side
and
the
home
and
we'll
show
that
how
using
context
missing
content
this
way
and
combining
it
with
anycast.
You
can
get
smooth
kind
of
uninterrupted
Phil
over
in
a
stateless
way.
So
I
encourage
everybody
that
tend
to
come.
D
Happy
thank
you
very
much
for
paying
for
this
panel.
Thank
you
very
much
for
coming
here
and
having
some
food
and
listening
to
us
talk
and
one
of
the
best
parts
is
I
already
knew
Grove
pretty
well,
but
one
of
the
best
parts
for
me
for
this
panel
was
getting
to
meet.
Craig
he's
got
a
lot
of
really.
You
know
great
insights
across.
D
E
D
So
employed
introduced
me
mentioned
that
I
am
a
co-founder
of
the
Cisco
Paris
innovation
and
research
lab.
He
also
said
that,
since
he's
paying
for
this
I'm
allowed
to
do
marketing,
so
this
is
my
marketing
just
telling
you
that
we
exist.
It's
a
really
fun
group
of
people,
those
of
you
that
know
us
here,
Eric
Venky
here
and
Pierre
Feaster,
and
some
of
the
others
that
have
been
coming
to
the
IETF.
D
We
do
lots
of
bits
and
bytes
and
lots
of
hackathons
and
things
like
that,
and
we
work
a
lot
in
ipv6
and
do
a
lot
of
networking
kinds
of
things,
but
this
lab
also
is
full
stack
all
the
way
up
to
the
application.
We've
got
a
video,
immersive
lab
and
I'm
about
to
be
talking
about
some
of
the
media
production
stuff.
That's
a
little
bit
more
down
Craig's
line
here
in
just
a
moment.
Next
one
so
shift
happens,
as
has
been
mentioned
up
here
in
the
topic
of
the.
D
D
Also,
this
is
moving
up
into
the
media
production
side
and
the
problems
look
different
there
than
they
do
in
distribution.
The
we
don't
just
have
video.
We
have
raw
uncompressed
video
that
has
certain
timing,
constraints
and
I.
Think
Craig
covered
some
of
that.
So
next
slide
in
a
media
production
center.
There's
a
lot
of
different
vendors
that
you
might
not
have
heard
of
unless
you
play
around
in
this
area.
This
is
an
eye
chart
here,
but
if
you
go
had
a
you
know
binoculars,
you
would
see
lots
of
different
company
names.
D
It's
it's
a
it's
a
boutique
industry
in
that
sense
and
they're
all
new
to
IP.
Not
all
many
are
new
to
IP.
As
you
say
all
so.
This
is
a
something
we
did
in
Paris
back
in
October
with
a
number
of
different
partners
is
not
just
Cisco
working
with
France
television,
where
we
had
their
headquarters
where
they
do
all
their
production
and
they're
used
to
doing
it
all
on
site.
Right
for
you
know,
like
live
broadcasts,
breaking
news
kinds
of
stuff.
Well,
that
little
orange
line
is
a
fiber
connection
across
the
across
Paris.
D
That's
in
there
in
the
middle
to
a
remote
production
facility,
and
we
did
an
in
a
big
operability
test
to
see.
Can
we
do
you
know,
live
video
production
remotely
right.
It
might
seem
trivial,
of
course,
there's
an
Internet.
Of
course
we
can
do
that,
but
it's
actually
a
really
big
deal,
especially
with
all
the
interoperability
involved.
So
standards
are
important.
Next
slide.
D
So
again,
I
get
to
do
some
marketing
in
our
lab
or
working
on
various
new
things.
We
have
orchestrated
live
media
production
data
center,
okay,
it's
a
platform
to
model
the
various
workloads,
and
it's
great
that
Craig
just
told
me
that
he's
defining
his
IP
workflows
right
now,
because
that's
what
this
is
all
about.
It's
orchestrating
those
workflows
from
a
media
production
standpoint:
we've
got
fairly.
You
know
several
demos
and
proof
of
concepts
and
things
that
we've
delivered
to
various
partners,
including,
like
native
4k,
video
production.
D
A
video
pipeline
now
we're
using
a
container
based
infrastructure
to
do
the
workload
placement.
This
is
kubernetes,
etc,
but
the
video
pipeline
itself
is
ipv6
segment
routing.
Some
of
the
you
know
the
the
latest
new
stuff
that
the
IETF
is
working
on
in
order
to
get
those
efficiencies
to
be
able
to
push
the
bits
as
fast
as
possible
through
that
production
data
center
in
a
most
predictable
way
that
we
possibly
can.
We
have
other
things,
video,
newsrooms,
virtualized
for
shows,
etc,
but
in
the
interest
of
time
you
can
ask
me
later
next
slide.
D
This
is
a
little
video
and
it's
the
dashboard
of
this
orchestrated
media
production
pipeline
system.
We
can
go
in
and
you
see
you
know
various
container
running
and
various
different
servers,
these
being
probably
more
to
craig
than
they
do
to
me.
But
we
can.
This
is
basically
modeling
a
newsroom
working
closely
with
Qur'an
ego,
which
is
one
of
these
boutique
companies.
They
were
Chiron
and
hey
go
and
they
combined
and
they
became
Quran
ego
and
they
make
this
kind
of
stuff.
Now
that
kind
of
stuff
traditionally
runs
on
windows.
D
Sized
machines
connected
together
with
SDI
coax
cables.
Okay,
now
the
shift
that's
happening
is
first,
those
coax
cables
become
IP.
Then
they
become
virtualized.
The
lift
and
shift
virtualized
workloads.
That's
more
of
what
you're
seeing
there
right,
but
ultimately,
once
you
have
that
flexibility,
you
can
reimagine
everything
and
just
your
the
the
work,
the
windows,
sized
components
of
workflows,
lined
up
and
and
in
in
in
you
know,
redundant
fashion
so
that,
if
anything
happens,
you
can
flip
over
and
you
have
a
fully
redundant
set
up
all
that
kind
of
goes
away.
D
Eventually,
when
you
have
the
power
to
psych
spin
up,
you
know
hundreds
of
thousands
of
containers
when
you
can
burst
into
public
cloud
and
all
that
kind
of
stuff.
So
that's
sort
of
the
next
phase
and
that's
what
we're
working
on
one
of
the
things
we're
working
on
at
our
research
and
innovation
lab
Thanks.
Thank.
A
You
Mark
so
we've
heard
sort
of
about
what's
been
going
on
and
and
what
people
are
doing.
So,
let's
talk
a
little
bit
about
the
future
here
and
and
stuff.
That's
maybe
on
the
near
term,
horizon
growth
you
may
be
so
v6
is
great
right.
We've
all
heard:
v6
v6
we'd
love
you
six
to
the
ITF.
It's
a
bigger
address
space.
So
what
other
features
beyond
more
addresses?
Are
we
gonna
get
if
we
can
actually
do
video
over
over
v6
yeah?
So.
C
C
A
series
of
120
bit
structures
or
segments
are
essentially
ipv6
addresses
into
an
extension
header
called
SRH,
and
each
segment
represents
a
topological
instruction
may
be
a
wink
or
anode,
and
you
can
use
the
SRH
to
steer
a
packet
through
a
path
in
in
a
in
a
network,
and
the
idea
is
that
this
list
of
segments
sort
of
in
the
SH
represents
a
network
policy
that
you're
kind
of
pushing
to
the
edges
of
the
network.
Now,
if
we
go
back
to
our
fully
ipv6
media
network,
can
you
bring
up
the
thing
so
with
SR
v6?
C
We
can
describe
paths
that
media
takes
through
the
network
so
on
the
distribution
side.
Of
course,
it'd
be
very
useful
if
you
could
cache.
Let's
say
a
video
stream
in
the
path
of
delivery
from
one
of
the
CD
ends
to
the
screen.
But
what,
if
you
wanted
to
cash
out
of
the
path
of
delivery
and
SR
v6
lets
you
do
this?
C
So
in
our
demo
we
have
a
home
cache,
so
the
home
cache
is
there
in
the
top
middle
and
the
purple
and
using
SR
v6,
and,
as
you
see
it's
not
necessarily
in
the
delivery
path
to
the
screens,
but
using
SR
e6,
we
can
now
cache
data
anywhere.
V6
can
essentially
be
rerouted
to.
There
are
a
few
other
examples
that
s
our
v6
enables
such
as
load
balancing
allowing
you
to
put
state
information
into
into
a
packet
or
flow
and
pinning
clients
to
specific
servers.
C
There's
some
really
neat
things
you
can
do
with
multicast
type
of
structures.
If
you
need
to
do
replication,
where
you
don't
have
multicast
available
or
you
necessarily
can't
be
deployed,
you
know
I
think
overall,
SR
v6
is
probably
one
of
the
most
exciting
things.
That's
happening
in
for
ipv6
right
now,
and
it's
going
to
make
really
media
workflows
for
more
flexible,
agile,
cost-effective
and
in
on
the
academic
research
side.
I.
Think
SR
e6
has
enabled
new
research
really
involving
how
the
applications
are
participating.
Their
network
traffic.
A
D
Well,
you'd
said:
okay
ipv6,
we
love
it
want
more
addresses.
It's
a
lot
more
addresses,
not
just
a
lot
more
dresses
a
lot,
a
lot,
a
lot
more
dresses,
30
orders
of
magnitude,
more
that's
a
lot,
and
so
one
of
the
things
that
as
part
of
the
ipv6
centric
program
in
at
Pearl
is
we
are
looking
at.
You
know
like
what
does
that
mean,
and
it
can
reverberate
all
the
way
up
and
down
to
the
stack
right.
D
I
can
now
address,
at
least
if
not
route,
2
and
use
things
like
segment
routing
to
to
a
named
object.
That
is
a
v6
address
and
you
know,
Craig
talked
a
lot
about
having
objects
and
the
metadata
associated
with
them
move
all
the
way
from
media
production,
through
distribution
and
eventually
to
the
end
user,
which
I
think
is
the
whole
glass-to-glass
dream
right,
and
I
think
that
you
know
having
a
looking
at
it
from
from
top
down
Craig's
looking
at
it.
D
D
The
use
case
the
garage
mentioned
of
being
able
to
steer
a
path
through.
You
know
a
video
oriented
path
like
because
it's
8k,
because
whatever
is
a
great
example,
you
can
also
use
segment
routing
to
search
for
a
particular
object
so
that
you
don't
have
to
have.
You
know
perfect
synchronization
of
all
your
routing
tables
everywhere
you
can
sort
of
have
a
directed
anycast
if
you
will
towards
an
object.
D
That's
something
that
if
you
want
to
go,
if
everybody
should
have
ipv6
here
and
at
least
in
the
building,
and
so
if
you
want
to
go
to
6cn
dot
I/o,
it's
a
website
set
up
by
eric
vinci
and
some
of
our
students
and
researchers
at
the
lab
in
paris,
and
you
can
see
a
lot
of
this
kind
of
stuff
with
doing
distribution
to
v6
addressed
objects.
You
can
even
upload
your
own
video
and
you
will
have
your
you
globally.
Unique
v6
address
addresses
on
every.
You
know.
D
A
Mark
you
know,
we've
talked
a
lot
about
v6
and,
and
so
it's
pretty
clear-
the
panel
kind
of
likes
b6
but
yeah,
but
this
isn't
just
about
to
be
sinks
right.
There's
a
lot
of
other
aspects
of
things
that
we
work
on
here
at
the
ITF
there's
a
lot
of
things
that
we
are
actively
agent
developing
and
then
things
we're
talking
about
doing
the
future,
and
so
so
Craig
I'm
gonna,
ask
you
so
I
know
you're
active
in
the
quick
group,
because
I
see
your
email
on
the
list.
You
don't.
A
Could
you
maybe
tell
me
a
little
bit
explore
what
other
groups
you
go
to
you
here
at
the
ITF?
What
what
do
you
find
is
a
video
production
person,
a
video
distribution
person?
What
do
you
find
relevant
here
during
your
week
to
go
visit
and
participate
in
that?
You
think,
as
we
bring
more
video
people
in
here,
where
should
they
be
going?
I
guess
is
part
of
the
question
and
that's.
B
A
It's
a
bill
you
to
move
people
and
visualize
where
the
work
is
done
and
do
it
at
will
it's
very
powerful
and
it's
the
same
stuff
we've
seen
in
other
industries
and
smooth
transitions.
So
let
me
first
do
two
pitches
and
then
we're
gonna.
Take
some
questions
number
one.
If
you
want
to
talk
more
about
the
talk,
because
we
have
a
very
limited
time,
we
have
to
be
out
here
by
115
to
get
the
next
session
in
so
up
this
evening.
A
We
grabbed
one
of
the
breakout
rooms,
the
trial,
I,
can't
even
get
anybody
speak
check
the
troca
room,
it's
up
on
the
mezzanine
level,
if
from
6:00
to
7:00
p.m.
just
before
bits
and
bytes-
and
this
is
after
quick.
So
if
you
go
too
quick,
you're
good
and
they
come
to
this
and
then
go
have
a
beer
at
bits
and
bytes
come
talk.
We're
just
we're
seeing
their
talk.
It's
not
it's!
Not!
It's
sometimes
we're
out
specifically
talking
with
Gigi
we're
not
specific
to
help,
but
anything.
A
We
tablet
the
specific
video
at
the
ITF
and
stuff
like
this.
Carry
on
the
conversation
we
like
to
hear
what
you're
doing
and
what
are
you
thinking
and,
of
course,
don't
miss
bits
and
bytes
tonight
I
mean
come
on
free
beer
and
food
and
demos.
You
can't
beat
that
okay.
So
let's
take
some
questions
from
the
audience.
F
So
I'm
very
particular
from
very
matrix.
This
question
is
for
Craig,
but
pretty
much
for
all
of
you.
The
television
industry
started
with
broadcast
right
and
then,
when
we
started
moving
to
IP
it
was
IP
multicast
and
now
everything
seems
to
be
point-to-point
and
Craig
was
talking
about
15,000
servers
to
just
deliver.
No
television
are
we
ever
gonna,
go
back
to
multicast
or
are
we
staying
with
unicast
forever
so.
A
So
I'm
gonna
use
hosts
moderator,
discretion
first
and
jump
in
give
my
two
cents,
and
then
you
can
answer
why
not
I
think
multicast
has
a
big
part
of
the
potential
future
for
doing
stuff,
but
I
think
we
should
also
always
keep
in
mind
that
we
have
people
watch
very
long
tail
stuff.
So
there
are
still
people
right
now
somewhere
in
the
world,
somebody's
watching,
Nightrider,
okay
and-
and
that's
that's
something
that
five
other
people
are
doing.
It's
about,
one
guy
watching
that
one
episode
and
so
multicast
is
a
solution
in
many
places.
B
B
You
can't
deny
the
growth
in
on
des
and
context,
especially
even
with
life,
with
sort
of
you
know,
live
or
instant,
rewind,
etc.
Starting
back
at
the
you
know
the
head
of
a
program,
and
so
for
me
the
future
is
flexibility.
This
is
why,
I'm
after
looking
into
concepts
like
having
a
common
layer,
seven,
it's
actually
about
reusing.
Those
object,
objects,
no
matter
how
they
were
delivered
and
in
what
time
base.
B
So,
yes,
IP
multicast
is
a
thing,
but
will
it
be
pervasive,
probably
not,
and
also
you
still
need
to
try
and
discover
your
media
in
some
way,
and
actually
we
have
a
whole
raft
of
how
do
you
discover
your
multicast
and
is
it
encapsulated
in
the
same
way?
And
now
do
I
have
to
do
a
lot
of
heavy
lifting
in
my
application
and
it's
a
it
yet.
So
flexibility
flexibility
is
important
and.
A
Craig
miss
an
opportunity
to
market
something
you
did
that's
very
cool
in
Chicago
at
the
bits
and
bytes
and
I
think
at
the
hackathon.
You
guys
demonstrated
a
thing
you
had
worked
on,
which
were
led
you
to
merge
into
a
multicast
stream
and
then
break
out
when
you're
doing
a
trick
play
stuff
and
they
merge
back
in
I
thought
it
was
pretty
awesome.
So
talk
to
him
about
that.
It's
very
awesome,
so
I.
B
Mean
actually
Lucas
who's
just
over.
There
is
one
of
the
key
engineers
and
that
bit
yes,
absolutely
we're
we're
looking
at
some
effectively
a
multipath
with
quick,
where
you
have
a
bi-directional
unicast
flow
with
an
effectively
using
multicast
as
not
to
mystique
upgrade
on
that,
but
the
layer,
seven
semantics,
the
same.
That's.
G
D
E
Brian
I'm
so
first
a
lesson
from
history.
Many
years
ago,
people
thought
the
48-bit
MAC
addresses
was
an
infinity
of
MAC
addresses
and
we're
now
having
a
crisis
because
there
aren't
any.
So
when
we
go
to
the
fine
grained
addressing
of
media
content
through
IP
addresses.
When
do
we
expect
to
run
out
and
to
go
with
that?
What's
the
consequences
of
this
for
the
routing
system
itself,
with
very
large
quantities
of
disaggregated
addresses?
Well
so
I'll.
A
A
What
I
predict
you
to
be
all
of
YouTube
for
decades
to
come
and
YouTube's
huge
in
terms
of
number
of
videos
they
have
and
I
think
it
was
like
to
the
32
bits
to
actually
do
the
at
the
addressing
of
their
other
content.
So
the
secret
is
a
also
to
the
routing
tables
the
secret
allocate
addresses
intelligently
to
content,
just
don't
blast
them
on
it
and
leave
the
whole
thing
sparse.
If
you
aggregate
and
you
do
it
smartly,
it
works
mark
graph.
Any
other
comments.
D
D
D
We
use
them
for
years
right
and
the
micro
services
are
coming
up
and
down
on
an
average
of
like
one
or
two
seconds,
and
these
are
all
just
v6
addresses
and
it
sounds
to
our
minds
that
oh
you're
gonna
run
out
because
we're
so
used
to
being
careful
what
the
v4
addresses.
But
you
know,
30
orders
of
magnitude.
Bigger
is
a
lot
bigger
and
I.
Think
we'll
be
okay.
I,
don't
wanna,
be
scared
and.
A
A
D
B
It's
a
logical
concept
more
than
anything,
it's
whatever
you
want
it
to
be,
and
it's
what
it's
whatever
you
can,
uniquely
you
can
or
want
to
uniquely
describe
about
a
editorial
experience.
So
in
the
example
we
showed
I
mean
it
was
using
a
sort
of
traditional
green
screen,
so
the
presenter
was
an
atom.
The
backdrop
was
an
atom,
but
there
was
an
alternative
to
both
of
those
additional
atoms.
The
dog
on
the
screen,
atoms,
subtitles,
atoms,
audio
atoms.
So
it's
it's
effectively,
abstracting
any
one
object
that
you
may
wish
to
change
in
real
terms.
B
It's
more
difficult
to
go
beyond
those
individual
things
that
you
capture,
but
our
research
team
is
already
working
on
taking
things
that
are
not
uniquely
captured
and
atomizing.
Those
as
well
so
so
actually
take
a
regular
2d
presentation.
It's
relatively
easy
to
isolate
the
presenter
and
actually
take
them
out
of
the
shop
and
reuse
that
elsewhere.
So
so
you
know
anything
that
you
can
capture,
but
anything
that
you
can
extract
from
that
which
you
captured.
B
B
Anything
that
you
would
like
to
describe
with
some
kind
of
metadata
I
mean
this.
Is
that
the
point?
The
opportunity
here
is
an
extra
degree
of
editorial
control,
an
extra
degree
of
flexibility.
You
know
this
conceptually
and
is
in
what
we're
talking
about
here
is
almost
a
simple
engineering
exercise.
It's
not
until
you
release
that
to
the
the
editorial
creatives
that
you
understand
what
they
really
want
to
do
with
it.
I'm
here
in
the
video
is
an
analog.
H
Name's
muslin,
yes
about
privacy.
This
system,
it's
I,
mean
I,
mean
I.
We
discussed
with
you
before
it's
set
around
the
kind
of
prototype
so
far
that
they
leak
massively
about
what
you're
actually
viewing
on
the
network
layer.
You
can
see
all
the
doing
will
use
this,
but
it's
what
exactly
what
you're
watching
and
where
it
cetera
and
I
hope
we
will
have
a
better
privacy
stories
going
forward
when
continuing
to
evolve
this
so
I.
A
D
You
you
could
you
could
change
like
privacy
addresses
on
steroids?
If
you
will
right
so
and
I'm
sure
Glenn's
going
to
show
you
some
interesting
things
along
those
lines?
It's
it
ends
up
being
you
know
we
build
the
technology
here
and
at
some
point
you
do
have
to
put
something
on
the
wire.
That
means
something
to
be
able
to
route
it.
But
it's
you
know
a
matter
of
some
combination
of
regulation.
What
the
consumer
will
accept,
what
the
business
will
accept
your
terms
and
conditions,
whatever
that
I
think
will
ultimately
dictate.