►
From YouTube: IETF114-BIER-20220725-1400
Description
BIER meeting session at IETF114
2022/07/25 1400
https://datatracker.ietf.org/meeting/114/proceedings/
A
C
A
A
G
E
Let's
move
this
bike
over
here:
we've
done
the
test
test
test,
we
think
we're
functional.
It
appears
from
what
we
can
tell.
We
can
hear
each
other
on
both
ends,
and
I
think
I've
got
this
thing
running
like
it's
supposed
to
that
doesn't
mean
I'm
correct.
E
I
only
have
a
delusion
of
fat
correct
right
now,
so
considering
that
time
is,
we
should
probably
get
this
rolling.
I
have
not
certain
how
the
blues
work.
I
think
you
need
to
scan
outside
or
scan
this,
and
then
that
opens
up.
Does
the
auto
blues
like
before
so
we
get
both
remote
and
local,
auto
blues,
which
is
pretty
darn
nice?
E
This
was
supposed
to
oh,
I'm
not.
I
have
to
take
over
okay,
I
have
to
take
over
as
display.
I
have
to
share
screen
video.
D
E
G
C
E
Okay,
welcome
to
the
beer
working
group.
We
are
oh
12
and
a
half
minutes
delayed,
not
bad.
Considering
this
the
first
session
in
the
morning,
we
are
not
the
only
ones
going
through
this
confusion.
Any
working
group
chairs
associated
the
laptop
in
the
room
is
no
longer
controlling
the
slides.
E
E
E
I
E
E
Go
to
the
note
taking
tool
just
automatically
puts
it
up
there.
So
do
you
see
where
I'm
showing
that
yeah
there
all
right?
We
are
there
all
right.
So,
let's
look
at
the
agenda.
We've
got
status
going
through
the
drafts.
What's
in
the
last
call,
what's
actually
in
the
queue?
What
is
expired?
I
was
already
called
to
my
attention
that
there
was
another
expired
graph.
That's
actually
ready
and
just
need
to
be
refreshed.
E
I
E
J
With
no
torque
yeah
tallest
eckerd
just
send
email
to
you,
folks
that
we've
got
the
brte
in
auth48
yeah
yeah,
but
then
two
weeks
were
really
painful
for
all
type
of
other
stuff,
so
only
become
rfc
after
beer
to
e
draft
yeah
dear
earth
48
so
will
become
rfc,
but
only
after
the
itf
okay,
authors
didn't
get
around
to
spending
time.
E
We've
got
two
here
in
the
queue
I'm
not
certain.
I
think
these
been
around
a
while
there's
your
terkas
in
the
cube
right.
That's
who
you're
talking
about
doorless,
yeah,
yeah,
okay
and
bar
ipa
has
been
there
a
while
too,
I
don't
know
we're
waiting
for
a
d
at
the
time.
Last
I
heard
we
also
have
some
80
comments
outstanding
for
v3.
E
In
last
call
hbls,
we
need
well
need
some
review
from
what
I
understand.
I
think
we
had
to
go
to
idr
for
review
on
that
too.
Didn't
we.
I
think
we
already
had
that
in
parallel.
I
E
Then,
from
there
on
it
moves
into
a
theme
we
need
shepherds.
We
need
document
shepherds.
We
have
our
own
man-man
man
here,
who
is
in
one
of
the
reviews,
we're
waiting
for
his
to
be
complete,
php
beer
and
six
and
prefix
redistribution
all
need
shepherds.
It's
a
fairly
straightforward
process.
There's
a
template!
You
go
through
q.
A
you
know.
Do
some
hunting
around
to
make
sure
you're
consistent
with
what
the
questions
are
list
response
and
things?
E
C
E
E
I
You
so
much,
we
need
beer
in
six,
because
the
babel
and
the
national
guys
are
picking
up
the
issue
again.
E
E
Let's
see,
then
we've
got
yeah,
so
the
ping
is,
I
guess,
getting
an
update.
Some
of
that
stuff
has
expired
and
come
back
and
refreshed
here
and
there,
and
then
we
have
one
list
expired
draft,
the
idr
extensions.
L
L
I
did
post
an
update
before
the
meeting
in
march
and
sent
an
email.
I
believe
I'm
done
with
the
document
and
I
sent
an
email
a
couple
of
weeks
back
as
well.
Yeah
all
right,
but.
L
E
E
I
E
Response
from
okay
man
apparently
has
the
hendra's
leash
and
we'll
be
dragging
him
back
into
this.
Thank
you.
We
have
someone
at
the
mic.
O
I
N
So
idr
extension
drafts,
I
talked
to
the
main
author
and
editor
xiaohu
xu
before
and
he
said
he
is
no
longer
focused
on
this
work,
so
we
need
some
other
co-authors
in
the
draft.
To
pick
up
the
remaining
the
editing
work
to
address
those
comments.
E
N
Hello,
so
the
beer,
or
markus
as
a
service
draft,
was
recently
adopted.
The
working
group
documents
today,
I'm
going
to
talk
about
a
new
section
that
I
added
to
the
to
the
zero
one
version:
it's
about
the
supporting
beer
in
vpn
and
it's
actually.
N
N
So
this
involves
both
overlay,
signaling
and
underlay
work
customers
bear
signaling,
then
they
can
do
the
igp
bgp
signaling
in
their
own
sites.
But
when
it
comes
to
over
this
provider
core,
we
would
use
the
bgp
signaling
the
existing
idr
extensions
for
beer
signaling.
N
I
believe
it
can
be
used
for
vpn
ip
sapphire
as
well.
So
you
just
attach
your
beer
information
to
the
vpn,
ips
and
rois,
and
then
everything
should
should
work
for
the
overlay
signaling
part.
Now
it
comes
when
it
comes
to
underlay.
How
do
you
get
the
customer
beer
traffic
across
the
provider
network?
There
are
two
solutions
here.
One
is
that
you
just
use
ingress
replication.
N
So
when
you
do
that,
it's
nothing
different
from
regular
beer
at
all.
It's
now
the
your
vr
apps
become
the
customers,
beer
or
routers,
and
it's
you
just
do
the
tunneling
between
the
bfrs,
nothing
special
at
all.
Now
another
option
is
that
you
forward
customers
beer
traffic
across
the
provider
network
using
beer
itself.
N
The
reason
to
do
that
is
that
it
gives
you
efficient
replication
across
provider
core.
You
don't
have
to
do
the
ingress
replication
anymore.
The
the
one
issue
with
that
is
that
you
to
do
that.
You
need
to
maintain
the
customer
beer
state
in
underlay,
and
that
is
what
this
presentation
is
about
next
slide.
Please.
N
Before
we
talk
about
the
underlay
signaling
and
stage
for
the
blvpn,
let's
review
the
general
spear
signaling
and
pift
calculation
in
the
non-vpn
case,
so
the
bfr's
a
signal,
beer
information,
trvs
attached
to
the
beer
prefixes
and
and
for
the
egress
router
bf
ir
or
bfer.
They
would
include
their
bfr
ids
in
the
signaling
bftr,
which
stands
for
beer,
folding
transit
router.
They
don't
include
the
bfr
ids.
N
So
for
each
beer
folding
table
there
is
a
beef
id.
For
example,
it
could
be
a
beer
label.
N
Now
how
do
we
calculate
that
bift?
We
basically
for
each
fifth
entry
is
corresponding
to
a
bfer
brs
router.
We
find
out
how
to
reach
that
particular
bfer.
N
We
see
that
okay,
through
the
unicast
it's
going
to
be
going
through
a
next
hop,
bfr
neighbor
and
that
neighbor
advertise
a
particular
label
for
this
subdomain
comma
set.
So
then
we
put
the
entry
in
the
bits
and
the
key
for
that
entry
is
the
bfrd
modulo,
the
p-string
lens.
The
next
top
information
for
the
entry
is
basically
the
beer
label
from
their
neighbor
and
a
unix
unique.
Has
four
information
on
how
to
reach
their
neighbor
notice
that
the
bf
er's
prefix
here
is
only
used
to
find
how
to
reach
it.
O
N
N
You
will
have
a
per
customer
subdomains.
There
notice
that
your
different
beer
vpn
customers,
they
could
have
overlapped,
subdomain,
ids
or
even
bfr
ids.
So
now
we
need
to
identify
each
beer
folding
table
using
using
a
additional
piece
of
information.
Here
we
call
it.
We
use
edit
raw
distinguisher
in
the
control
plane,
but
in
the
for
in
the
following
plane.
N
It
still
just
use
the
pixel
id
as
before,
like
a
beer
label
and
the
beer
architecture
initially
intentionally
made
sure
that
beef
id
itself
is
just
opaque
number.
It
does
not
carry
any
other
semantics
and
that
design
allow
us
to
do
this.
Now
you
can
extend
the
sub
domain
to
includes
customer
information
whatever,
but
when
the
folding
plane,
the
b5d,
is
just
opaque
number
that
maps
to
whatever
tuple
you
give
it
to
give
to
it.
N
So
that's
a
very
good
design
that
we
are
making
use
of
now
now
when
it
comes
to
signaling
now
the
pe
routers
and
pe
routers
in
the
underlay,
they
will
add
it
has
beer
information
trv
for
their
underlying
loopbacks.
Only
and
now
inside
the
brt
info
trv
they
use
they
add
raw
distinguisher
to
the
subdomain
id
and
name
for
the
pes.
N
That
information
is
now
used
to
calculate
per
customer
or
gifts.
J
Can
I
ask
an
understanding,
question
sure
I'm
lost
at
the
point.
Oh
tell
a
second
behind
a
in
in
the
service
provider
domain.
Let's
assume
a
situation
egress
pe
with
the
two
or
three
egress
ce
of
the
customer.
Each
of
the
egress
ce
has
has
a
bit
in
the
bit
string.
Is
that
mapped
one
to
one
into
the
the
beer
bit
string
in
the
core
when
you're
right?
When
you
do
run
it.
N
So
for
each
for
each
customer
biffs
in
the
on
the
on
on
the
ce
side,
then
there
will
be
corresponding
biffs
in
the
in
the
provider
network
itself.
N
N
N
E
P
Huawei
for
the
beer
whipping-
I
think
english
replication
is
the
best
solution
because,
as
you
said,
it
has
a
low
air
state
on
the
key
motors
to
nick
the
client
pfid
range
to
provide
a
network
is
not
a
hierarchical
solution.
I
think
it
somewhat
violates
the
traditional
weapon
architecture,
because
either
of
the
clan
bfr,
prefix,
clan,
afrid
or
client
bfid
range
is
the
root
information
of
the
client
network
leaking
them
into
provider
network.
This
is
a
scalable
problems.
N
I
have
some.
I
have
one
slide
regarding
the
scaling
considerations.
N
If
that
should
not
be
the
last
slide
or
or
is
it
can
you
yeah
yeah?
So
indeed,
now
we
have
customer
specific,
specific
information
in
the
provider
network.
N
Now,
as
for
the
additional
states,
I
want
to
point
out
today
in
the
mvpn
or
evpn,
when
you
do
selective
tunneling,
you
are
essentially
adding
additional
state
in
the
provider
network
for
corresponding
the
customer
flows.
You
for
you,
for
example,
you
could
have
one
p2mp
tunnel
for
every
customer
flow
in
in
the
provider
network.
N
That
is
the
precedence
and,
of
course,
the
provider
has
to
make
a
choice
to
decide.
Do
am
I
going
to
use
ingress
replication
or
using
the
using
the
inclusive
tunnel
for
all
my
customer
traffic?
N
Sometimes
that's
that
may
not
work
well,
because,
just
because
the
the
amount
of
additional
replication
and
in
you
know,
provider
networks,
that's
not
going
to
work
out.
It's
not
it's,
not
economical.
So
here
is
the
similar
situation.
N
Yes,
ingress
replication
among
the
vrfs
will
work,
it
does
not
have
the
states,
but
if
you
do
the
introduce
additional
customer
gift
state
in
the
provider
network,
it
provides
a
trade,
efficient
replication
and
you,
as
a
provider,
you
can
control.
N
How
many
customer
base
to
be
supported
in
your
provider
network
is
compromised,
and
my
thinking
is
that
if
the
multicast
or
beer
demand
picks
up,
this
solution
is
acceptable
and
it's
actually
feasible
and
it's
worthy.
That's
my
view.
J
Thomas
eckhardt
again,
I
think
the
the
key
difference
is
not
even
starting
from
the
number
of
states
that
you
have,
but
how
you
can
plan
it
right.
It
moves
it
from
application
created
state
which
even
the
operator
of
the
customer
site
may
not
be
able
to
well
control
to
operator
control
right,
so
the
number
of
bits
are
controlled
by
the
customer
operator,
so
that
can
be
brought
to
be
a
no
number.
Obviously
right.
So
that's
that
that's
that's
the
main
benefit.
I
think
it's
a
lot
more
controllable.
N
Here
is
the
the
the
provider
can
still
has
control
right.
So.
N
J
The
numbers
you
cannot
plan,
you
know
when
applications
are
starting
to
to
to
create
flows
right.
E
J
You
still
have
this
date
on
the
pe
of
the
service
provider
when
you're
running
traditional
beer
only
in
the
service
provider,
you
still
have
that
state
there
now
you're
pushing
it
back
to
wherever
the
customer
wants
to
have
it
and
from
the
customer
you're
only
getting
a
planable
accountable,
chargeable
number
of
bits,
kind
of
distributed
via
the
bit
string.
So
it's
a
totally
different
model.
It's
not
kind
of
compare
sizes.
It's
called,
I
think,
yeah.
J
I
Now
I
have
a
technical
comment
on
the
igp
redistribution.
So
if
we
extend
the
beer
information
with
the
rd,
then
the
computation
is
basically
different.
Primary
key
and
the
whole
igp
will
have
to
support
it
to
work
correctly,
which
can
be
done,
but
what
I
suggest
is
to
add
to
igp
kind
of
a
capability
indication
in
the
router
description.
So
you
know
that
you're
dealing
with
people
who
can
actually
support
this
rd
or
otherwise,
whatever
you
have
it
partitioned
or
you
know
you
have
to
deal
operationally
with
the
issue.
Yeah
thanks.
E
N
Okay,
so
this
is
a
new
draft
from
co-authors
in
juniper,
zt,
channel,
mobile
and
nokia.
Next
slide.
N
So
there
is
a
illustration
of
the
header
definition
at
the
very
top
is
original
beer
header
it
would
have.
Its
protofield
will
have
a
value
to
say
that
the
beer
extension
header
follows,
and
at
the
beginning
of
that
there
is
a
header
of
extension
header
and
then
followed
by
the
extension.
Headers
could
be
multiple
of
them
next
slide.
N
Now
the
header
of
extension,
headers,
the
there
is
the
first
four
bits:
the
reserve
field.
It's
really
just
for
it's
a
enable
because
this
could
be
used
for
npos
as
well.
We
want
to
be
able
to
see
that
don't
treat
this
as
an
ip
packet
and
next
one
is
a
counter
of
the
extension
header,
followed
by
the
extension
header
total
lens,
followed
by
the
original
upper
layer
protocol
type.
That's
the
real
payload
after
extension.
N
N
Therefore,
we
want
to
use
this
address
the
number
space
from
the
ip
protocol
numbers
and
that
way,
if
there
is
a
ip
extension
header
for
a
function
that
can
be
applicable,
can
be
applied
to
beer
or
mpos,
we
will
just
use
reuse,
the
ipv6
extension
header
now,
the
the
key,
the
the
real
that
there
are
two
main
fields
here
are
critical
here,
total
I
shouldn't
say
critical,
but
it's
important
that
total
extension
header
lens
gives
you
the
ability
to
to
directly
locate
where
the
payload
is,
for
whatever
reason,
without
going
through
the
headers
one
by
one,
the
next
hop
header
or
value
using
the
ip
protocol.
N
Number
space
allows
you
to
do
the
use,
reuse,
ipv6
extension,
headers
number
of
headers
in
this
one.
I've
heard
from
some
people
saying
that
you
may
not
need
that
or
you
may
not
want
there.
I
want
it
there,
for
example,
if
you
have
to
remove
a
header-
and
you
may
need
to
update
this
field,
why
bother
including
it
there?
Just
as
long
as
you
have
the
total
lens?
Each
everything
should
work.
Fine,
that's
yeah!
We
can
discuss
that
with
mpos
the
co-authors
as
well.
Yeah.
H
N
Oh,
I
already
talked
about
that.
Why
do
we
want
this?
One
beer,
header,
protofield
itself,
is
a
six
bits
with
its
own
space.
Now
we,
when
we
use
this
one,
we
can
reuse
the
ipv6
extension
headers
for
whatever
function.
Now
I
do
want
to
point
out
using
the
internet
protocol
number
space
does
not
mean
that
every
time
we
design
a
new
beer
extension
header,
we
have
to
allocate
a
number
from
that
ipv6
registry.
No,
we
only
need
one
number.
N
N
If,
if,
if
you
know
that
you're
dealing
with
a
beer
or
encapsulation,
you
know
that's,
that
would
be
beer
specific.
If
you
know,
if
you're
doing
the
mpos,
you
know
it's
mpl
specific
next
slide.
N
So
now
extension
header,
it's
just
like
ipv6,
we
added
one
optional
extension
field
there.
Basically,
let's
say
that
we
have
to
define
five
mpos
appear:
specific
extension
headers
for
beer
functions.
We
get
one
number
from
ipv6
ip
protocol
number
registry
and
then
we
can
define.
N
We
can
allocate
five
extension
and
value
for
that
field
in
in
this
picture
and
so
that
they
can
represent
different
functions
with
different
beer
extension
headers.
We
could
also
have
one
extension
value
assigned
to
indicate
that
for
this
sphere
extension
header,
there
are
multiple
tlbs
for
different
functions,
so
they
could.
There
are
different
ways
to
do
things.
You
can
have
one
extension
value
for
each
spear,
specific
extension
function
or
you
can
have
one
extension
value
to
using
different
tlvs
to
encode
different
information
for
different
functions.
N
So
seeking
comments
and
suggestions,
we
try
to
try
to
reach
consensus
from
major
vendors
and
operators.
I
mentioned
that
we
already
have
quite
some
major
vendors
here.
That
draft
song
is
draft
itself
is
from
huawei.
So
I
assume
well
that
that's
the
implication
that
implies
that
we
we
do
want
to
work
with
all
the
mentors,
including
huawei,
to
to
to
support
this.
D
F
E
F
F
Okay.
Okay,
thank
you
should,
I
repeat
the
first
page
again:
oh
okay,
let's
move
the
this
background.
Okay
from
the
brt
architecture.
We
know
that
drt
uses
explicit
passes
to
forward
multicast
packet
in
vr
domain.
The
path
can
be
calculated
by
the
controller.
F
F
F
All
the
advertisements
are
similar
with
fc
8401
and
the
draft
lsr
mpls
extension
in
case
both
of
the
mpos
and
the
non-mpos
encapsulation
sub
sub
tovs
are
advertised
by
one
node.
The
label
in
mpls,
encapsulation
sub
sub
to
v,
and
the
fifth
id
in
non-ampus
encapsulation
sub-subtle
v
should
not
be
overlapped.
J
Tallest
eckerd
yeah-
sorry,
I
still
haven't
didn't,
have
the
time
to
come
out
to
conclusion
with
this
I'll
I'll
certainly
be
happy
to
support
it
from
the
r
for
from
the
brte
side.
We,
you
know
the
soon
to
be
rfc
has
is
describing
you
know
at
least
two
high-level
models
of
how
to
share
the
si
and
sd
space
with
beer,
and
I
think,
depending
on
what
we
want
to
do
there,
that
may
have
an
impact
on
what
we
do
in
the
signaling
plane.
N
Jeffrey
from
juniper,
so
I
had
a
comment
for
another
draft
from
why
more.
I
had
the
same
comment
applied
here.
I
saw
a
response
from
from
primo,
but
I
sorry
I
haven't
got
a
chance
to
reply,
but
here
my
understanding
is
when
it
comes
to
brte.
N
In
case
of
a
regular
beer,
each
bfr
will
calculate
the
contents
of
the
fifths,
but
when
it
comes
to
brte
my
understanding
that
each
bfr
only
needs
to
care
about
its
local
or
the
big
positions
for
its
local
adjacencies,
whether
it's
direct
adjacency
or
router
agencies,
so,
and
that
that
information
is
better
populated
from
a
controller
and
instead
of
using
igp
to
signal,
and
if
you
do
use
the
controller
to
signal
to
pro
to
provision
those
content
of
the
bits,
then
then
the
bit
id
itself
can
also
be
signaled
as
part
of
that
provisioning
instead
of
using
ige
signaling.
N
E
N
Assigning
the
bits
yes
well
so
assigning
the
bits
is
a
part
of
a
planning
process,
and
then
how
do
you
get
the
space
to
the
router
and
it's
in
the
old
days?
You
may
you,
for
example,
in
the
old
days
you
allo
you
assign
a
loopback
address,
and
then
you
write
down
a
piece
of
paper
and
include
the
router
and
enter
it
here
is
similar.
You
assign
the
bits
and
then
you
you
need
to
put
it
into
the
router,
how
to
put
it
in
the
router.
N
Nowadays
you
can
use
the
controller
or
something
once
you
do
that,
then
you
you
as
part
of
that
as
a
provisioning,
you,
you
could
use
to
do
the
big
signal
to
assign
the
fifth
ids
as
well
thanks.
A
E
F
Okay,
we
know
that
there
is
a
working
group
document
about
jabir
t,
young
and
and
in
that
draft
the
the
ftid
or
the
agency's
bp
can
also
be
assigned
by
the
controller
and
can
be
downloaded
to
the
devices
to
run
for
it.
So
so,
if
we
do
this
work
or
by
controller,
the
brt
young
is
enough
to
do
this
work
in
case.
We
want
to
use
igp
for
signaling
the
bit
operation
of
agencies.
E
E
I
would
call
this
by
far
the
quickest
beer
working
group
meeting-
we've
had
probably
the
least
animated
we've
had
if
we
went
like
yes,
even
though
we
started
a
good
point.
Actually,
if
I
even
docked
that
time
off
we're
about
45
minutes,
then
of
actual
operator
operation
here,
normally
we're
pushing
our
two
hour
window
and
people
are
chasing
us
out
the
room.
So
at
giving
the
point
give
some
okay,
we
have
an
a.d
who's
going
to
give
us
some
wisdom.
G
I've
actually
just
been
there's
been
a
bit
of
discussion
on
the
iesg
and
I
have
been
thought
that
we
would
remind
people
just
while
you're
in
the
meetings,
if
you're
not
at
the
mic.
Please
wear
your
masks.
We
there
are
people
testing,
positive,
that
aren't
wearing
masks
and
the
masks
are
mandatory.
So
I
just
needed
to
please
say
if
you're
in
the
session
and
you're
not
at
the
microphone
talking,
please
wear
the
masks,
but
other
than
that.
M
Ahead,
yeah:
well,
it's
not
a
question.
It's
a
comment.
Being
nokia
should
be
in
the
next
ietf
come
up
with
some
beer
success
stories.
I
don't
know
whether
that's
part
of
the
itf
or
not,
but
just
you
know,
show
some
of
the
deployments
for.
C
E
But
he's
here
so
isn't
there
some
work
in
i2,
where
european
i2
they've
got
some
p4
implementations
of
beer
and
they're
doing
some
actual
live
testing
now.
Q
Q
Oh
yeah,
they
have
a
network
called
rare
net
and
it
is
a
network
of
tofino
switches,
p4
switches
that
they
implemented
beer
on
and
it
is
a
beer
network.
It's
a
global
network,
rna
network
that
includes
that
that
does
have
beer
running
and
they
added
an
amt
relay
and
a
unicast
multicast
translator.
So
so
yeah
come
to
mbondi
or
check
out.
Q
We
would
love
to
you
know,
there's
there's
a
mailing
list
that
talks
about
this
deployment,
but
it's
a
pretty
neat
deployment
and
it's
it's
exciting
because
it
it
has
the
confluence
of
both
beer
and
amt,
which
is
kind
of
a
interesting.
You
know
very
promising
mix
so
yeah,
that's
a
that's
a
cool
network
to
take
a
look
at.
E
Q
I
I
can
also
add
that,
since,
since
you
asked
greg,
we
are
or
we'd
love
to
have
in
mbondi
your
other
working
group.
Don't
forget
about
us.
If
there
are
operational,
where
are
we
with
beer?
Do
you
are
we
ready
for
operational
considerations
and
are
we
ready
for
ops
documents
and
drafts
in
mbondi
pertaining
to
beer,
because
I
think
that
would
you
know
if
we're
at
that
point
now?
We
very
much
welcome
that
in
bundy.
N
Jeffrey
from
juniper,
so
as
a
multicast
engineer,
I've
always
said
the
beer
is
the
best
mouse
multicast
technology
I
have
since,
but
this
deployment
has
been
kind
of
disappointing,
mainly
because
it's
new
encapsulation
new
forwarding
algorithm
and
therefore
you
neither
need
a
programmable
chips
or
you
need
brand
new
chips,
and
that
gives
us
a
chicken
egg
problem.
N
There
have
been
many
customers
who
wanted
to
do
beer.
They
talked
to
us
to
many
vendors
and
they
said.
Oh,
you
cannot
do
this
yet,
oh
you
cannot
do
this
on
all
platforms.
Therefore
they
back
off
when
they
back
off
the
vendors.
Also
back
up,
they
say:
oh
no,
not
many
customers
wanted.
So
this
kind
of
chicken
and
egg
problem
is
what
prevents
white
deployment.
N
But
prime
pioneer
vendors
and
pioneer
operators
will
break
the
chicken
negative
active
dilemma,
and
I
am
optimistic
that
it
will
happen
and
in
the
near
future,.
J
No,
no
sorry,
I
have
a
lot
of
back
pain.
I
can't
stand
a
lot.
What
was
it
three
points
so
the
first
one
just
to
retaliate
on
your
great
recommendation,
the
beginning
for
for
chairs
me
for
the
people
in
the
room.
Please
make
sure
that
you
are
locked
in
at
least
once
into
meet
echo,
because
that's
how
you're
getting
counted
right.
So
you
want
to
know
that
for
the
room
size
next
time
and
sorry,
it
was
too
late
for
this
meeting
here.
But
that's
so
there
is
the
on-site
tool.
J
J
J
Right
there
was
something
in
the
middle.
I
forgot
okay,
h
and
then
yeah
I
mean
hope.
You
don't
find
this
offensive,
but
really
you
know
come
positively
mood
it
to
the
msr
six
path
right
and
consider
that
you
know
this.
J
R
Greg
mursky
erickson.
I
really
appreciate
the
ask
for
more
ops,
related
information
and
I
think
that
that's
what
oem
is
for.
So
it's
basically
using
us
our
alphabet
soup,
it's
a
operation,
administration
maintenance.
M
My
big
goal
in
nokia
just
to
comment
on
the
chicken
and
egg
issue.
I
do
agree
on
that,
but
I
guess
there's
a
lot
of
great
work
in
this
working
group.
There's
a
lot
of
drafts
that
are
trying
to
address
different
parts
of
the
beer
technology,
but
for
the
providers
well,
not
providers
for
the
main
companies
that
are
implementing
beer.
M
If
we
don't
have
the
deployments,
you
know
there
are
other
higher
priority
stuff
that
you
know,
companies
will
start
working
on.
So
that's
all
I'm
trying
to
point
out
is
that
maybe
you
know
in
the
next
meeting
we
can
talk
about
some
of
the
main
pieces
of
beer
work
that
we
need
to
implement
on
all
these
companies.
You
know
the
tier
one
companies
that
that
we
are
so
we
can
get
this
thing
up
and
deployed
in
the
customer
side
a
little
bit
more
again.
M
J
Just
a
quick
reply-
and
maybe
you
know
I
told
the
secret
again,
maybe
looking
into
what
what
we
all
came
here
for
service
provider
and
the
wide
area.
Isn't
the
only
you
know
thing
we
can
attack
right,
so
data
centers,
we've
seen
michael
menth
and
and
companies.
You
know
p4
implementation,
so
data
center
may
be
more
ripe,
even
though
we
may
not
be
the
ones
that
have
the
best
connections
there,
but
maybe
just
talking
to
the
right
people
right.
E
And
j2
broadcom
has
support.
Now
that's
a
huge
step.
K
Okay,
it's
vancouver
from
deutsche
telekom,
so
from
the
network
operator
side,
the
chicken
and
egg
problem
which
have
been
from
mentioned
from
jeffrey
and
human,
and
also
the
operations
management
side
mentioned
from
greg,
are
very
relevant
to
us
because
for
us
it's
always
the
question.
How
interoperable
is
is
this
stuff?
Because
it's
not
a
secret?
We
are
not
on
a
single
vendor
strategy
so
that
point
one
is
very
important,
but
the
other
point
as
well
is
that
we
are
getting
challenged
internally,
also
a
lot,
how
many
others
of
our.
K
Of
other
telcos
are
using
it,
are
they
using
it
in
the
backbone,
or
is
that
a
data
center
technology,
because
this
determines
for
for
us?
Also,
how
mature
is
the
technology
to
use,
or
are
we
the
first
one
or
are
we
the
only
one
demanding
it,
because
this
has
also,
obviously,
in
the
long
run
implications
on
how
well
is
it
maintained?
How
well
is
it
tested?
K
How
well
is
it
developed
further
for
on
for
upcoming
feature,
sets
so
having
somewhere
a
source
other
than
personal
contacts,
to
get
a
grasp
of
how
many
providers
are
interested
and
how
many
providers
are
using
it
and
for
which
cases
would
be
really
beneficial?
Also
for
me
and
my
arguments
to
use
it
internally
to
develop
it
further
and
yeah
to
to
give
it
also
some
more
relevance
in
order,
in
contrast
of
implementation,
excellent.
E
So
if
I
can
just
wrap
this
up
with
my
experience
in
this
as
well,
I
haven't
spoken
to
a
single
vendor
who
looked
at
this
and
said:
oh,
I
don't
like
it
it's
always.
When
can
I
get
it?
What
platform
is
it
on?
You
know
how
many
platforms
support
it,
but
then
you
get
to
that.
Who
else
is
doing
it?
There's
that
coward
state
right
which
I
understand,
but
this
is
forwarding
plane.
This
is
just
not
another
routing
protocol.
This
is
the
forwarding
plane.
So
it's
the
narrow
part
of
the
hourglass
is
the
big
challenge.
E
It's
new
hardware,
and
just
to
your
point,
jeffrey
too.
Not
all
programmable
chips
give
you
that
option,
because
it
requires
header
manipulation,
post
replication,
if
there's
not
a
pipeline
to
do
that,
you
have
no
programmable
path
to
do
it.
So
even
programmable
chips
aren't
necessarily
ready
for
this,
so
it
it's
it's
a
heavy
lift.
G
J
On
on
on
your
rant,
just
just
going
back,
you
know,
original
multicast
was
also
pushed
in
the
market
by
software
implementations
right.
So
that's
the
question:
what
is
the
strategies
that
we
that
we,
you
know,
find
marketplaces
where
that
happens?
And
then
you
know
the
big
elephants
are
getting
moved
last
right.
So
that's
that's
always
been
the
problem
when
we
primarily
selling
from
you
know.
The
big
elephant
devices
with
with
expensive
hardware
need
to
want
to
make
a
jump
start
for
a
new
technology.
O
Lampotter,
so
there
is
also
chicken
and
egg
with
software
implementations.
Speaking
as
someone
who
would
do
one
of
the
software
implementations,
we
are
stuck
because
the
use
cases
from
data
centers
and
telcos
in
this,
and
they
want
the
hardware
implementations
and
they
don't
see
the
point
of
start
with
the
software
implementations
and
there's
like
yeah.
J
So
there
is
a
joint
session
from
babel,
monet
and
roll
on
which
with
which
are
looking
into
multicast,
and
so
you
know,
we've
tried
already
to
to
discuss
beer
type,
beer
itself
or
beer
type
technologies
with
role,
and
then
we
just
didn't
have
the
time
and
critical
mass
and
so
on.
So
maybe
we
can
get
more
critical
mass
into
that.
You
know
software
space
because
I
think
there
are
a
lot
of
use
cases
all
around
but
yeah.
Obviously,
people
haven't
worked
enough.
J
G
E
H
Hi,
this
is
ron
enfeld,
the
chair
of
the
monet
working
group
and
the
session
that
just
mentioned
this
on
friday
morning,
first
session,
okay,
friday
and
yeah.
My
question
is
indeed,
these
are
mostly
software-based
routing
solutions
that
we're
looking
at
there
and
would
be
or
be
applicable.
C
H
At
least
sandy
is
giving
a
presentation
on
how
to
to
carry
beer
information
in
the
label
product,
but
how
this
works
on
the
boarding
plane
is
not
yet
clear.
So
I
have
to
find
that
out
before
friday,.
A
E
F
Yes,
I
just
want
to
add
some
discussion
for
for
for
the,
because
in
china
we
had
talked
beer
technology
to
operators
years
again
years
ago.
So
I
think
that
in
china
many
operators
know
that
beer
is
a
very
good
technology
and
they
want
to
use
it,
but
the
another
driven
for
the
beer
deployment
is
that
how
to
use
it
for
the
application.
So
we
just
we
had
talked
the
vr
technology
with
the
cdn
and
cdn
content
delivery
so
and
some
media
delivery
application
to
work
with
it.
F
So
I
think
that,
maybe
later
there
can
be
used
in
china
not
very
long
yeah,
okay,.
E
E
I
was
kind
of
surprised
because
he's
pretty
busy,
but
I
think
from
what
I
understand
they
may
be
coming
out
with
some
info
like
this
summer,
so
yeah
we
might
hear
what
they're
doing
I
mean
it's
pretty
clear
as
nasty
land
play,
knowing
the
players
but
yeah.
What
structure
were
we're
really
doing
we'll
find
out.
E
All
right
yeah
this
is
this-
has
been
fun.
You
know
it's
been
two
years
of
remote,
I'm
just
kind
of
doing
my
own
thing
for
a.
I
remember
this
last
night
talking
to
smart
people.
You
know
talking
about
heart.
Problems
like
this
is
fun.