►
From YouTube: IETF106-RTGWG-20191122-1000
Description
RTGWG meeting session at IETF106
2019/11/22 1000
https://datatracker.ietf.org/meeting/106/proceedings/
C
So
this
is
an
update
on
the
work.
What
we
were
discussing
in
the
working
group
and,
in
course,
of
the
presentation
I'll,
try
to
explain
why
I
think
that
it
might
be
interesting
for
the
RT
g
VG
group,
not
only
for
BFD
group,
so
we
are
all
well
familiar
with
the
BFD
and
the
specification.
It's
very
successful
development
design
of
the
protocol
that
being
deployed
in
many
networks
over
different
transports,
IP
MPLS
to
the
wire.
C
So
we
are
people
are
working
on
applique
it
to
overlay
networks,
but
now
and
then,
especially
in
a
B
of
D
group
since
we're
the
core
for
their
beauty
protocol.
There
are
proposals
and
discussions
about
some
extensions
to
BFD,
for
example,
to
monitor
the
quality
of
beef.
This
session,
because
BFD
as
a
fault
failure,
detection
protocol,
has
some
leverage
of
how
many
messages
in
row
it
has
has
to
be
missed
in
order
to
detect
failure.
C
But
what,
as
might
be
of
interest,
is
that
to
know,
for
example,
if
the
text
multiplier
is
3,
so
how
many
really
messages
did
we
miss?
We
are
receiving
constantly
each
in
every
message
or
we
start
missing
one
or
two
messages,
but
then
third
message
arrives
and
oh
ok,
we
keep
the
session
up
again.
There
are
some
proposals
just
to
evaluate
the
performance
and
path
MTU,
because
BFD
designed
to
be
lightweight
and
efficient.
It
is
on
a
very
small
control
message,
but
there
is
interest
to
see
if
we
can
do
something
more
with
it.
C
So
here's
we
came
up
with
there
some
ideas
of
how
we
can
make
BFD
extended
extensible
and
another
motivation
was
that
we
have
very
interesting
discussion
on
authentication
and
BFD
dedication
and
BFD
is
sort
of
you
configure
it,
and
you
must
do
it
on
issue
in
every
packet.
So
if
you
can
imagine
if
B,
if
the
interval
is
3.3
milliseconds
or
at
that
rate,
so
that
doing
F
indication
and
Asian
every
packet
in
both
directions-
that's
kind
of
heavy
load
on
a
system.
C
There
is
a
proposal
that
working
group
completed
discussing
and
progressing
further,
that
to
do
intermittent
authentication
for
busy
session.
But
there
are
some
other
issues
that
really
doesn't
make
it
always
at
will.
So
there
are
some
constraints
to
that,
especially
in
a
synchronous
mode.
So
let's
move
on
and
get
to
the
technical
details.
So
the
proposal
is
that,
because
BFD
control
message
does
not
specify
how
its
encapsulated
in
a
transport,
it
only
defines
the
control
message
and
the
size,
as
I
mentioned,
it's
a
fixed,
whether
it's
unallocated
mode
or
authenticated
mode
for
authenticated
mode.
C
It's
predictable.
So
our
proposal
is
this
to
have
VFD
control
message
for
proud
idea
of
a
guard
Ward
just
for
safety
to
separate
it
and
then
followed
by
govt
until
we
can
have
enclosed
sub
t
Obi's.
So
that
makes
it
extensible
well
at
the
same
time,
of
course,
might
be
not
as
Hardware
friendly
so
to
understand
whether
it's
regular
BFD
control,
message
or
extended,
of
course,
because
BFD,
at
least
in
IP
network
works
over
UDP.
C
E
C
E
C
Capability
negotiation,
so
the
active
side
announces
its
interest.
The
other
side
sends
well,
it
agrees
to
and
supports,
and
then
out
of
that
they
can
conclude
what
will
be
used
performance
measurement.
The
proposal
is
to
reuse
encoding
defined
in
RFC.
63
74
was
until
a
measurement
over
MPLS
networks,
and
the
nice
part
about
it
is
that
these
messages
allowed
to
do
separate
was
delay
measurement
direct
measurements
by
collecting
counters
and
combined
was
delay
measurement
and
because
of
BFD.
C
So
this
was
just
an
illustration
of
how,
in
coatings
from
RFC,
63
74
can
be
used
in
this
integrated
OEM
or
extended
BFD.
So
we
just
using
TLV
to
encode
their
message
that
already
defined
in
RFC
63
74
for
us
measurement
for
delay
measurement
and
for
combined
measurement
and
again
it
doesn't
have
to
be
sent
on
high
frequency
or,
for
example,
BFD
messages
that
monitor
path
continuity,
but
they
can
be
sent
as
part
of
their
poor
final
sequence.
C
So
then
the
load
on
the
system-
and
it
would
be
not
that
challenging
core
path,
MTU
monitoring,
it
introduces
basically
padding
type
of
TLV,
which
can
have
arbitrary
length
centered,
and
it
can
be
used,
as
in
the
proposal,
be
if
the
large
packets
set
to
certain
value
and
again
used
in
the
PO
final
sequence
to
casually
and
time-to-time.
Monitor
the
path,
and
one
of
the
benefits
of
using
poor
final
sequence
is
that
it
does
not
interfere
with
their
more
frequent
exchange
or
transmission
of
periodic
messages
by
each
side.
C
C
Paths
between
two
systems
and
the
proposal
here
is
that
to
give
the
Cape
capability
of
authenticating
whether
periodic
messages
for
final
messages
or
combination
of
both
and
again
using
4-pole
final
messages
like
an
MTU.
It
gives
a
local
policy
opportunity
to
define
the
rules
when
they
declare
that
session
is
invalid.
C
F
C
C
So
BD
/
lag
spans
single
hop
sessions
over
each
constituent
link.
It's
a
good
question.
I
think
that
might
be
applicability
again,
because
this
draft
allows
you.
This
mechanism
allows
you
to
use
only
this
extensions
over
poor
final
sequence,
okay
and
and
I
think
that
poor
final
sequence
will
go
only
one
not
on
each
individual
leg
right.
E
F
C
G
C
C
C
C
And
again
to
your
question
of
Tony
question:
I
want
to
stress
that
there
is
no
requirement
to
do
it
in
the
periodic
messages.
All
these
extensions
can
work
in
a
poll,
final
sequence
and
the
poll
final
sequence
usually
is
processed
well
at
least
as
bump
in
the
wire.
So
if
not
in
control,
plane
then
bump
in
the
water.
So.
C
I
K
Good
morning
everybody,
my
name
is
June
sky
and
I'm,
a
researcher
of
entity
laboratories.
This
draft
was
originally
proposed
in
calm
spot
and
its
first
time
to
present
this
graphic
in
this
working
group.
So
I
will
talk
about
digest
or
content
of
this
draft
and
introduce
our
related
walk
in
this
presentation.
J
K
Okay,
so
okay,
so
firstly,
either
I
will
talk
about
the
background
of
this
document.
So,
as
you
know,
devices
and
services
die.
Bussing
and
network
will
be
required
to
apply
to
such
diversity
and
networks.
Rising
is
emerging
approach
to
apply
networks
to
such
diversity
so
and
several
SDOs
discussing
about
it,
including
CDC
VP
so
and
ITF,
had
started
to
discuss
transports
right
so
instead
works
rising,
end-to-end
networks
right,
a
quick
realization
of
end-to-end
networks
rise
is
very
important.
Therefore,
provide
assured
communication
quality
for
each
service
and
3gpp
has
three
subnet
concept.
K
K
Answer
for
creating
and
to
enterprise
stitching
of
subnets
will
be
required
in
a
nation
to
teaching
subnet.
Several
functionality
would
be
liquid
for
printing
and
provide
network
services,
for
example,
resurrection
us
control,
encapsulation,
decapsulation,
etc,
and
thrust
Gateway
provides
such
functionality
at
each
boundary
of
domains.
K
This
slide
shows
a
requirements
for
thrust
gateway.
Service
gateway
will
provide
mainly
two
types:
roll,
a
faster
is
handling
underlying
infrastructure
to
create
slices,
secondly,
is
provide
providing
controllability
of
user
traffic
to
tenant.
We
are
assuming
that
prices
will
be
provided
to
sort
of
parties
or
tenant
and
they
use
network
services
as
part
of
their
own
service.
K
So
some
tenant
may
need
to
control
the
user
traffic
on
thrice.
So
let's
get
away,
provides
such
functionalities,
so
this
matrix
reached
the
requirements
for
thrust
gateway
so,
and
this
figure
shows
the
overview
of
the
thrust
gateway
structure
as
resgate
aware
is
composed
of
thrust
gateway,
function,
transfer,
gateway,
controller
and
several
data
rank
entity,
data,
plane,
entities
and
rest
gateway.
Controller
has
two
types
of
api's.
First
two
is
four
managing
slices
and
it's
corrected
to
the
higher
operation
system.
K
So
into
MF
we
proceed.
We
are
proceeding
the
average
aberration
test
feasibility,
test
of
creation
of
M
to
n
slice
across
multiple
administrative
domains.
If
this
figure
shows
a
scenario
of
POC,
MF
and
oak,
several
orchestrators,
who
played
each
other
and
create
thrice
subnet
in
each
domain
and
first
gate
away,
connect
the
subnets
and
create
end-to-end
fries.
K
We
comment
on
communication
so,
for
example,
player
players,
people
and
the
people
cause
high
speed
or
re
so
low
latency
communication
is
very
important
and
people
game
is
the
most
important
factor
in
this
content,
and
so
the
traffic
must
be
protected.
On
the
other
hand,
because
dreaming
doesn't
require
real-time
communication,
it
requires
just
broadband,
so
we
prepared
the
several
types
of
slices
and
game
application,
select
appropriate
slices.
K
So,
for
example,
players
that
the
pimp
on
earth,
the
game
and
game
Saba
recognized
the
device
has
players
loan,
so
request
a
central
request
to
allocate
such
traffic
to
the
players
thrice
and
this
content
can
walk.
Even
if
traffic
congestion
happens,
we
okay.
So
this
is
just
example
of
API,
so
a
game
application
sends
requests,
including
flow
identifier
and
trans
rights,
ID,
and
so
as
gateway
controller.
Translates
the
API
to
the
concrete
network,
configuration
and
enforce
it
to
the
discipline
entity,
so
game
application
don't
require
to
understand
concrete
network
topology
or
conflation
in
this
model.
K
So
finally,
I
would
talk
about
next
step
of
this
document,
so
with
pruning
to
breaking
down
a
nose
one
interface
with
leaf
or
link
definition
and
sophistication
of
n
s.
Dt
in
this
working
group.
Also,
we
provide
a
use
case
is
where
SSD
or
subnet
concept
will
be
beneficial
anyway.
So
we
have
very
few
feedback,
and
so
please
read
this
draft
and
send
your
feedback.
Thank
you
very
much.
D
Speaking
of
the
design
team
member,
not
as
working
group
chair,
I'm,
very
happy
with
your
next
steps,
the
step
number
one
we
are
allowing
terminology
so,
for
example,
it
would
be
highly
unstable
to
say
it's
a
subnet.
We
know
that
sublet
is
not
what
they
meant
right,
so
I
would
really
expect
you
to
use
terminology,
and
the
coverage
that
we
have
established
in
design
team
knew
your
document
and
fide
right
place,
and
here
are
key
of
controllers.
If
you
wish,
where
it
really
belongs.
K
M
L
N
Hello,
my
name
is
Angus
Holly,
police
and
closer.
This
is
better
great,
so
my
name
is
evangelist
Holly.
Please,
and
on
behalf
of
my
colleagues
I'd
like
to
present
you
our
take
off
force
based
PNG.
By
the
way
my
activities
are
carried
with
funding
provided
by
the
sand.
Iced
tea
initiative
from
you,
with
78
0
for
39
program,
grant
agreement
under
the
horizon
2020
program,
so
I,
don't
know
how
many
of
you
saw
the
mail
yesterday
that
the
IDF
is
not
going
to
continue
activity
on
the
PNG.
J
Ok
I
mean,
since
you
brought
the
topic
up,
there
was
a
liaison
yesterday
from
the
iesg
to
broadband
forum.
That
basically
said
you
know,
work
on
disaggregated,
B
and
G
had
been
suspended,
the
IHF
in
March
and
that
suspension
will
continue
until
you
know
a
decision,
a
change
in
decision
is
made
by
request
from
BBF
or
something
so
like
moving
forward.
I
guess
you're,
saying,
like
forces
is
a
technology
that
could
be
reused
right
for.
J
Based
on
you
know
the
IES
G's
feedback
in
the
future,
like
we,
we
probably
won't
be
allocating
time
to
discussion
of
this,
but
you
know
so
sorry,
if,
like
BBF,
wanted
to
look
at
forces
and
say:
oh,
that's,
really
good
and
then
request.
Oh
hey
forces
is
great.
That's
fine,
but
we're
not
going
to
be
allocating
time
until
there's
any
change
in
that.
J
O
Yeah
so
Davison,
crop'
speaking,
is
the
IETF
liaison
manager
to
BBF.
So
yes,
that
liaison
went
out
from
the
is
G
to
be
BFF
yesterday,
and
it
did
you
that
was
a
great
summary
Chris.
Thank
you.
It
basically
said
we're
gonna,
suspend
all
work
on
the
disaggregated
bng
here
forces
I
can
tell
you
was
a
topic
for
discussion
within
the
broadband
forum
and
while
there
has
been
a
protocol
selection
done
for
the
control
point
to
user
playing
protocol,
which
is
PFC
P,
there's
been
two
liaisons
from
BB
F
to
IETF
say
stating
such
you
know.
N
Q
R
Developers,
telefónica
not
in
the
front
page
for
the
hem
I'd,
be
I,
feel
myself
one
of
the
accomplices
of
evangelist
for
preparing
this
and
our
original
intent
when
we
started
talking
about
this
was
precisely
it's
not
only
about
VG's,
and
this
is
why
I
believe
that
this
current
title
current
focus
is
somehow
connected
to
this
statement
for
the
AG.
But
our
intention
was
precisely
to
explore
the
idea
of
a
unified
mechanisms
for
access.
Yes,
call
it
being
G's
call
it
whatever
the
3gpp
call
its
on
the
in
in
5g
call
it.
R
C
Greg
music,
is
it
II
pretty
much
concur
with
the
previous
statement
that
there
are
a
lot
of
things
that
can
benefit
from
their
control
user
played
separation,
so
I
think
that
the
right
way
thing
would
be
is
just
cut
them
whether
it's
access
network
specifically
and
then
give
the
requirements,
agree
on
the
requirements
and
see
what
the
generality
we
can
produce
again
without
any
specifics
of
talking
about
distributed.
Vengi.
S
T
T
J
T
T
T
J
O
Okay,
thank
you,
so
I
have
one
less
comment
to
make
I
will
say
is
a
quite
kind
of
a
public
service
announcement
which
I
didn't
cover.
In
the
last
time,
I
was
at
this
mic.
The
broadband
forum
has
actually
changed.
Their
operating
I
would
say
operating
procedures,
but
it's
not
really
procedures
they've
changed
their
bylaws
such
that
now.
Anyone
anyone
can
approach
the
broadband
forum,
pay
a
subscription
fee
or
an
access
fee
and
get
access
to
all
of
their
work
in
progress.
They
are
now
nearly
as
open
as
the
IETF
for
a
small
fee.
J
U
We
came
from
a
kind
of
mobile
you'll
know
that
there
are
about
3,000
to
PGCE
all
up
in
the
network.
The
protocol
patron
can
opener.
Are
you
definitely
the
wedding
hall
scene
for
us
when
Muhammad
the
to
see
that
they
are
not
just
rooting
for
this?
Is
it
face,
may
be
isolated.
Si
Dakota
direction,
I,
don't
know
that
the
the
solution
stressing
from
the
PBS,
but
there
could
they
could
contain.
There
are
other
kramitz
alpha
channel
mobile,
so
example
the
street
the
end
in
lotta.
U
U
O
Hi
so
Dave
syncope,
again
liaison
managers
BBF.
If
there
are
requirements,
I
mean
the
document
was
sent
over
in
one
of
the
leaves
on
the
latest
relays
on
I
would
encourage
people
to
take
a
look
at
it
review
if
there
are
requirements
that
need
to
be
covered
or
should
people
feel
are
missing,
that
process
is
still
open
in
the
BBF,
so
feel
free
to
approach
them.
Thank.
V
P
P
N
Things
to
go
for
bringing
go,
so
the
beans
is
actually
a
huge
case
that
we
want
to
to
showcase
that
we
can
support
mobile
blacks
types
and
multiple
scenarios,
and
so
how
many
of
you
know
forces
can
I
see
a
show
of
hands?
Ok,
ok,
so
I'll
just
do
a
very
brief,
intro,
so
forces
in
ITF
solution.
For
some
time
ago
there
are
a
couple
of
RFC's
already
published.
We
had
two
successful
interoperability
tests
and
for
this
has
a
number
of
features
like
availability,
publish,
subscribes
and
request
response.
N
I
think
if
the
most
important
most
important
is
that
the
protocol
is
independent
of
the
model,
so
Forces
models
the
the
functions
of
the
data
plane,
using
what
we,
what
we
call
the
logical
functional
block.
So
it's
an
abstraction
of
the
functions
of
the
data
plane.
The
forces
model
is
object-oriented
and
has
several
advantages
of
this.
Like
we
have
inheritance,
we
have
augmentations
and
so
developer
can
define
the
components,
capabilities
and
events
of
of
a
data
plane
function
and
he
can
manipulated
and
which
the
component
and
which
actually
change
the
behavior
of
the
can
change.
N
The
behavior
of
the
data
buff
program
and
so
packets
come
in
on
optional
metadata
and
they
are
being
altered
by
the
data
program,
possibly
and
exit
as
additional
metadata
or
changed
and
proceed
to
the
next
lfp
class.
So
we
have
lfb
classes
that
are
instantiated
and
connecting
it
into
a
graph
which
actually
creates
creates
a
service.
So
why
force
it's
an
existing
ietf
solution?
It
has
been
proven.
It
has
been
used.
N
It
allows
you
to
define,
among
other
things,
capabilities
hanging,
friends,
lfb
graphs
can
be
dynamic,
so
a
control
plane
application
can
actually
change
the
behavior
or
and
change
the
services
that
exists
in
the
uniform
plane.
It
can
natively
support.
Any
type
of
access,
foreign
will
show
and
I
will
showcase.
How
and
for
any
new
lfb,
including
the
into
the
into
the
foreign
plane,
does
not
have
any
impact
on
the
protocol
right.
So
we
can
add
ends
many
new
existing
or
renewal
fps
and
the
protocol
is
going
to
stay
the
same.
N
N
So
when
has
been
authenticated
and
authorized
to
use
the
to
proceed
to
the
network,
then
the
control
plane
is
going
to
program
the
classify
in
order
to
allow
this
subscriber
traffic
to
pass
through.
It
goes
through
the
classifier
through
the
people
of
B,
in
order
to
end
caps
to
capsulate
the
PPE
pppoe
hiddenness
IP,
very
routing,
to
discover
the
next
next
hope
and
then
goes
out
into
the
network
on
the
way
back.
E
N
Destination
IP
address
goes
to
the
IP
v.
Routing
LF
being
goes
to.
The
pppoe
will
be
to
be
encapsulated
in
pppoe
packets,
out
of
the
port
LF
in
towards
the
subscriber
side
right.
So
once
traffic
goes
through
the
vb
ng,
we
need
to
be
able
to
to
monitor
traffic
in
order
to
to
find
you
know
its
autistics
and
possibly
usage
usage
measurements.
In
order
to
do
accounting.
N
N
Protocol
has
specific
verbs
said
yet
delete
and
it
may
it
is
going
it
is
it
manipulates
components
inside
the
the
lfb
class
whatever,
as
long
as
it
is
model
using
the
first
model,
the
first
protocol
is
able
to
handle
that
so
now
that
was
a
brief
view
of
simply
connectivity
service.
So
what
happens?
If
you
want
to
to
have
new
new
services
into
the,
so
we
can
simply
add
new
LF
B's
into
the
graph
right,
for
example,
for
a
bad
news.
Bandwidth
management
service.
N
That's
going
to
be
a
next
graph
of
LFP
that
the
specific
user
is
going
to
be
handled.
So
you
can
have
specific
services
for
specific
users
right
and
for
the
for.
Similarly,
for
the
for
the
downstream
and
upstream
traffic
right
and
so
again,
having
a
data
model
is
quite
useful
in
order
to
be
able
to
support
this
kind
of
kind
of
activities
so
that
the
protocol
is
not
going
to
be
changed.
It's
not
going
to
be
changed
by
any
changes
of
new
services
or
new
function
that
we
want.
H
To
do
add,
yes,
thank
you
guys,
a
question.
Louis
change
from
Geneva,
yes
to
standardize
a
protocol
in
between
is
a
very
big
challenge.
The
reason
why
I
say
that
is,
if
you
look
at
so
many
radius
Archbold
for
each
vendor,
it's
over
100
radius
attributes.
It
means
different
functions
due
to
different
hardware
implementations,
and
you
want
to
standardize
it.
How
do
you
do
it
because
radius
already,
you
have
four
different
radius
kind
of
attributes,
meaning
different
implementations
in
different
vendors.
Only
a
very
small
set
of
standard
attributed
a
in
ITF
for
the
radius
sections.
H
All
the
rest
is
a
VP
or
special,
a
special
attributes
vendor
passive,
a
attributes
VSA
how
it
standardizes
right.
The
second
question,
actually
is
you
want
to
start
at
something
which
is
also
difficult
to
see,
is
like
a
line
cut.
We
put
I
like
a
crash.
Instead
of
I
mean
dropping
dousing
of
user.
You
still
in
send
message
one
message
to
the
control
plane,
how
you
digest
those
messages,
standardize
that
one
is
another
problem
for
me,
I
mean
for
me
to
see
a
criticism.
Okay,
so
for
the.
H
Or
a
signature,
because
today
radius
is
the
major
one
using
right.
Yes
and
already.
If
amanda
has
a
very
big
dictionary
VSA,
which
is
very
venta
pacific
kind
of
things,
only
a
very
small
portion
is
standardized
right.
Now
you
can't
standardized
or
distinguishing
one
protocol.
That
means
every
Vantage's
would
use
the
same
dictionary,
which
is.
H
N
H
T
T
H
Q
Know
this
okay.
First
of
all,
one
other
thing
which
my
brain
is
not
completely
fried
out
having
a
fixed
set
of
basic
primitives
and
then
adding
vendor
extensions
on
top
of
it
makes,
in
general
the
community
life
easier,
because
they're
saying
we
have
a
certain
semantics
and
syntax
that
we
can
all
use,
and
then
we
can
do
vendor
extensions.
On
top
of
that.
Q
On
that
part
I,
you
know,
I,
agree
and
Lewis
said
that
you
know
this
thing
with
the
radius
is
a
good
example,
because
there
are
a
few
common
ones
and
then
many
vendor
specific
ones.
That's
a
pretty
good
approach
and
I.
Personally,
like
that
approach
to
be
move
forward,
there
are
two
things
what
they
have
with
the
you
know
with
the
forces
I
like
the
idea.
Their
question
is:
is
there
a
implementation?
Because
there
are
other
frameworks
like
that?
Where
you
can
model
the
data
plane
and
put
it
into?
Q
There
are
much
more
used
and
if
you
could
explain
why
forces
is
better
than
the
other
existing
frameworks
that
are
getting
a
lot
of
traction
within
the
industry,
then
we
can
evaluate.
Then
we
can
evaluate
that
so
I
can
you
know,
I
can
mark
I,
can
use
cap
and
proto
to
model
the
data
plane
in
x-men
and
expose
it
up
north
string.
So
why
should
I
care
idea
plenty
of
other
ones
and
I
can
say
I?
Could
let
me
finish
now,
then
he
I'm
asking
him
not
asking
you
this
part.
Q
C
Q
J
N
N
It
is
transport
layer
ignosi,
you
can
have
augmentations
of
models.
You
can
do
all
these
kind
of
things,
so
it's
not
about
serialization
deserialization
right
so
from
what
I
understand
for
cap
and
proto
or
any
other
civilization
is
a
laser.
You
have
to
develop
all
these
kind
of
features
from
you
have
to
actually
develop
this
right.
So
this
inherently
supports
all
this.
Q
T
30
seconds,
okay,
so
they're
not
ITF
standards.
This
is
an
idea
of
standard.
It
has
been
vetted
here
for
many
years.
Sure
I
couldn't
go
and
implement
your
own
thing
and
you
can
get
a
hundred
of
your
friends
to
work
on
it.
This
is
a
standard.
It's
been
deployed.
It's
well
understood.
It's
got
a
lot
of
scrutiny
here.
The
idea.
T
J
We
we
don't
yeah,
we
don't
need
to
continue
the
discussion
thanks
a
lot.
It
was
very
thought-provoking
and
timely.
So
so
next
up
we
have
Chun
Fang
Zhi,
presenting.
D
X
X
This
draft
is
about
us,
obviously
considerations.
Actually,
it
has
relationship
with
other.
What
are
the
traffic's,
the
service
diploma
stators?
If
you
teachers
progressed,
this
draft
need
to
use
progress
as
a
v6
industry,
including
deployments
implementations,
academic
contributors
and
in
operating
tests
and
those
as
you,
maybe
you
may
be.
X
We
all
know
that
as
a
v6
has.
The
following
advantage
is
that
first,
is
that
a
PVC
throat
application?
This
is
compared
with
the
amperes
as
our
MPs,
because,
yes
amperes,
if
you
adopt
32
Pete's
segment,
which
will
be
at
the
vet,
has
across
all
the
dominant
work,
but
in
yes,
a
v6
because
it
it
in
use
the
native
ipv6
features.
So
the
road
can
be
aggregated
to
reduce
the
issue
of
the
routing
issues,
skeptical
of
any
issues.
X
So
we
also
can
say
that
as
a
basic
provide
end-to-end
auto
end
to
end
service,
auto
start,
which
means
service
agility,
because
because
well
as
I'm,
curious,
I
need
to
upgrade
other
node
to
support
support
capability,
but,
yes,
a
v6.
We
only
need
to
have
be
deployed
based
on
actual
requirements
of
the
operators.
X
We
all
know
that
a
services
can
be
untrue
on
demand
upgrade
which
reduce
service
proven
in
hand,
and
so
we
can
status
of
several
advantages.
This
draft
also
gives
some
deployment
academy
guidance
for
a
service
deployment.
We
also
have
two
options
for
this
network
transition.
The
first
one
is
from
IBM
ps/2
ipv6,
then
as
a
v6
and
as
a
second
one
is
from
IP
and
ps2
SMPS
and
12
is
6.
So
from
our
experience,
we
think
that
the
first
one
may
be
more
natural
and
3
forward.
So
will
you
come
back
so
right
now?
X
Will
you
comment
the
option
option
1
it
is.
It
is
very
simple
that
you
should
make
also
network
to
be
ipv6
capable
as
what
we
have
done
in
right
now,
so
then.
Secondly,
we
need
to
upgrade
the
some
at
routers
some
edge
device
to
be
a
service
keepo
based
on
the
actual
demand
of
the
service.
Then
next
we
upgrade
the
network
to
either
immediately
know
some
of
the
media
to
know
to
support,
to
support
ipv6,
to
support
as
a
basics,
to
meet
to
enable
traffic
engineering
except
a
dinner
and
as
a
SFC
etc.
X
X
So
in
this
case
this
case
we
depend
on
the
architecture
of
the
whole
network
in
China
telecom.
We
have
two
backbones
IP
backbones.
One
is
one
six
three,
which
is
a
lady
IP
network.
Of
course
it
is
ipv4
and
ipv6
to
stack
which
provide
many
internet
access
service
and
a
DC
service,
and
also,
and
also
some
mobile,
Internet
services
etc
and
another
the
other
backbone
is
seemed
to
which
provide
which,
which
is
unpierced
based.
8-Minute,
provides
some
kinds
of
to
improvise
users,
so
in
a
traditional
approach.
X
Maybe
this
is
the
traditional
approach
the
Enterprise
Service
carries
into
but
differently,
in
this
case
the
video
traffic
for
for
the
interconnection
with
the
video
platform,
the
connection
is
carried
in
1
6
3,
because
1
6
3
provide
abundant
bandwidth
but
which
can
meet
the
requirements
of
the
video
video
service.
So,
in
order
to
the
process
of
is
quickly
with
the
way
implement
P
routers
to
be
activist
to
be
SR
basics
people,
so
we
can
provide
service
quickly
in
Shoreham,
which
improves
the
service
service
of
Romanians
providers,
improve
the
speed
of
a
service
bell
ringing.
X
The
second
prominent
case
is
China
Unicom.
The
backbone
of
charity
comes
is
1.
6
9
is
once
called
169.
Actually
is
a
number
in
this
case
it
many
years
as
a
basic
has
been
adopted.
240
the
collection
of
cloud
Deepika's
centers
it
collects
data
center,
actually
is
similar
to
to
tell
it
had
come
to
some
extent.
So
in
this
case,
because
China
tan
the
backbone
Network,
has
PPP's
rider
over
the
country,
if
loading
the
metal
in
the
work,
so
we
has,
we
have
some
end-to-end
ipv6
foundation
at
work
for
a
service
department.
Y
X
Y
X
This
speed
right
in
time
in
half
no
seconds
seconds,
hundreds
of
milliseconds-
it's
not
a
second
actually,
because
this
is
the
service
for
enterprise
users.
A
traditional
would
maybe
take
more
than
one
months,
because
we
need
to
enter
some
kind
of
procedure.
It's
a
long
time,
but
now,
maybe
several
days
this
service
can
be
prevented
to
customer
is.
This
is
not
for
ordinary
users.
Ok,.
G
X
X
G
R
Z
X
So
also
in
China,
Unicom
set
up
the
key
equipment
to
be
a
savvy
skip
post
so
that
between
traffic
can
be
traversed
from
one
ottoman
system
to
another,
which
means
from
before
example,
from
Beijing
to
hong
kong
province,
and
also,
we
all
know
that
the
new
province
has
also
implemented
as
a
basic
space
VPN
service
in
in
guangzhou,
shinden
and
Bloomquist.
It
is
so
please
our
experience
as
a
v6
has
shown
showing
advantage
internal
service,
agility
and
incremental
migration.
X
So
we
think
that
a
service
is
very
attractive
and
well
variable
for
the
future
network.
The
prominent
of
course
it
needs
more
multi,
prominent
cases
for
for
the
further
evaluation,
and
maybe
we
need
to
for
the
maturity
of
the
network
in
the
future.
It
can
also
be
used
for
favortie
transport
and
the
people
signals,
etc.
X
C
C
Y
Z
Y
AA
David
I
found
say
ICT
Kenya
super
Banga
to
this
light
with
the
incremental
deployment
yeah.
This
one
in
is
that
one
isn't
from
your
site
to
our
career
to
ipv6
and
to
my
stand
at
my
on
the
Trenton
they
operate
from
ipv4
to
ipv6,
maybe
from
the
corner
was
first
and
your
step.
Two
is
a
bhagavate
to
the
edge
device
first
so
and
microwaves.
It
will
charge
up
day
today.
X
My
understanding
is
that
that
each
in
our
case,
because
ipv6
has
been
realized
everywhere,
a
PVC
has
been
capable
has
been
variety
I
refer
to
network,
so
we
things
that
maybe
this
transition
from
ipv6
and
ipv6
we
appear
will
be
better
choice
then,
as
our
MPs,
but
maybe
it
was
other,
maybe
for
other
operators.
We
may
have
a
real
specific
suppose
we
requirements
for
network
equipment,
maybe
as
ice
army
MPs,
will
be
a
bad
choice
as
Mohonasen.
Thank.
AA
E
AB
You
would
take
this
comparison
into
account.
All
your
things
are
chillie,
not
true.
What
is
here
so
you're
you're,
comparing
just
I
wanting
from
the
other,
but
there
is
other
ways
to
achieve
that,
in
which
case
this,
this
analysis
is
not
honest
and
up-to-date,
with
all
the
capabilities
which
we
have
available
in
the
idea.
X
AB
V
True
from
Softbank
thank
you
for
sharing.
Are
you
experienced
a
deployment
in
detail?
I
just
make
some
coin
comment:
I'm
really
happy
to
see
you're
talking
about
deployment,
not
only
a
Greenfield,
so
many
people
claim
that
hey
it's
our
basic
require
Rudy
Greenfield
different
case,
but
the
you
show
a
good
example
that
that
s,
our
basic,
doesn't
require
Greenfield
to
be
deployed
and
also,
if
you
have
some
chance
to
introduce
te
future
I.
E
V
This
some
discussion
here,
but
if
you
consider
that
some
controllin
technotrekker
flex
argue
or
something
to
obstruct
as
intent,
topology
or
path,
so
it
would
be
a
good
idea
to
reduce
the
size
of
the
seat
and
site
the
number
of
seat
in
the
pocket.
I
think
you
can
still
keep
the
one
seat
in
just
our
pp6
distinction
address
in
the
ipv6
header.
That
would
be
fine
to
do
the
traffic
engine
future.
Even
for
the
inter-domain
case,
if
the
if
inter-domain
flex
I'll
go,
is
possible.
Thank.
V
AC
X
D
So
thanks
and
all
deployment
cases
are
very
useful.
Please
keep
on
bringing
in
and
personally
I
would
be
really
interested
to
see
migration
from
T
based
networks
or
CPT,
where,
for
example,
iterations
done
in
Burns
versus
moving
to
computations
done
of
and
how
people
solve
problems
with
regards
to
bend
with
resource
reservation
and
stuff
to
be
really
interesting.
Use
cases
to
discuss
here.
AD
Hello,
everyone
I'm
climate
change
from
shoot
away
today,
I'm
going
to
talk
about
a
SRT
pass
meter
meter
point
of
us
protection,
so
first,
let's
give
an
overview,
so
we
have
existing
faster.
For
me
to
point,
however,
we
meet
a
point
of
a
surpass.
Fails
the
traffic
where
we
are
lost
for
some
time,
even
though
the
existing
fr4
midpoint
is
used.
So
our
solution
is
provide
a
protection
for
that
clear
of
a
time
that
means
that,
with
our
solution,
the
failure
of
a
meter
point
of
a
surplus
will
be
fully
protected.
AD
AD
So
we
can
see
that
the
traffic
get
lost
from
the
point
at
which
the
edge
we
convert
converges
on
Q,
the
new
and
the
trend
path
is
the
installed
so
for
that
for
those
period
of
time
which
marked
by
the
red
red
line
that
during
that
period
time,
the
traffic
get
lost.
So
our
solution
is
provide
faster
protections
against
the
midpoint
of
failures
that
fear
time.
So
with
our
protection,
the
full
layer
of
a
be.
The
point
is
fully
protected.
AD
So
basically,
we
revision
of
several
versions
so
operates
to
the
previous
versions
that
we
changed
the
title
as
a
suggested.
So
originally
we
use
a
segment
routing
proxy
forwarding.
So
right
now
we
use
a
srt
past
mutable
inner
protection,
so
this
is
more
is
better
than
the
previous
one.
In
addition
to
that,
so
originally
we
use
a
router
capability
trv
to
distribute
the
capabilities
for
proxy
forwarding,
so
we
change
that
one
to
router
functional
capability
of
ease,
so
this
one
I
think
this
one
is
much
better
than
previous
one.
H
Is
change
from
a
Juniper
I
do
foresee
that
actually
is
a
good
thing
to
have
a
peer
out
for
these
operations,
but
there's
a
issue
actually
I'm,
not
sure
with
the
diagram
is
a
it's
a
very
simple
one.
It's
be
possible
this
because
I
GP
the
synchronization
kind
of
timing
sequence,
it
would
possible
there's
some
kind
of
our
drops
actually
deterring
the
switching.
So
even
your
pilau,
you
still
have
trouble
of
the
one
we
say
Michael
your
problem,
Michael
looping
avoidance
problem.
Actually
it's
another
little
I
can
tell
you
actually
offline,
but
this
is
possible.
H
AD
H
Motor
be
correct,
I
am
saying
actually
when
I
GB
progress,
there
were
a
different
update
of
the
stage
and
due
to
synchronization
of
the
control
plane
and
synchronizing
of
the
after
14
play
and
control
playing,
there
will
be
a
little
bit
easier
there.
I
will
show
you
later
on.
Actually
the
minimum
is
a
minimum.
AD
AD
AE
AF
AG
AE
AD
AG
J
One
so
I
think
there's
a
miscommunication
here.
What
Stefan
is
asking
is
he
understands
what
the
solution
you're
proposing
he's
asking?
Why
not
instead
use
a
solution,
a
different
solution
that
delays
the
installation
of
basically
of
the
black
hole
or
of
the
black
hole
in
the
traffic?
When
you
have
no,
no
entry
basically
keep
keep
the
old
notes.
It.
Entry.
E
AD
AG
AD
AD
AG
AD
AF
That's
true
it
so
this
this
is
an
incredibly
common
scenario.
Us
protection
like
this
is
what
you're
trying
to
do.
I'm
still
puzzled
as
to
why
there
isn't
a
more
elegant
solution
involving
any
casts.
So
in
that
particular
scenario,
presumably,
you
should
use
in
any
car
seat.
The
N,
in
which
case
ordinary
fast
reroute,
would
sort
the
whole
problem
out.
No.
AF
Let
me
explain
it
then
right,
so
he
discovers
that
it
can't
get
to
any
anymore.
Alright,
so
P
is
going
to
does
a
a
fast
reroute
to
n
prime,
which
is
well
that
ends
anycast
address,
which
should
get
it
to
end
in
one
once
it's
in
1
it
will
go
to
C,
because
that's
the
next
sit
so
I
would
have
thought
you
could
have
built
an
anycast
solution
for
this
standard,
very
common
scenario,
and
it
would
have
just
worked
so
I
think.
AD
Y
Yeah
so
I.
Certainly
this
is
Dino.
I,
certainly
support
the
idea
and
there's
a
lot
of
reasons.
Why?
Because
if
the
CID
is
an
anycast
address
and
you
have
to
reroute,
the
headend
doesn't
have
to
change
the
SRH.
This
is
very
useful
and
if
you're
always
routing
to
end
the
any
cast
address
and
you
the
route
to
the
top
and
you
never
have
to
delete
anything
from
the
FIB
or
any
forwarding
table,
because
M
will
just
stay
there.
The
next
hop
just
changes.
AD
AF
Other
concern
I've
got
and
it's
a
concern.
I've
always
got
with
these
sorts
of
solutions,
is
that
we
reduce
it
to
a
sort
of
trivial
cartoon
case.
When
you
apply
this
to
a
real
network
with
multiple
hops
in
there,
then
it
very
quickly
breaks
down.
But
if
you
were,
if
you
could
anycast
to
the
two
exits,
then
ordinary
fast
reroute,
followed
by
ordinary
conversions,
would
mean
you
could
deal
with
any
failure
in
getting
from
a
to
the
a
serial.
AD
AF
AD
AF
I
Actually,
I
don't
mean
to
defend
this,
but
what
is
a
I
wrote
a
loop,
but
there
isn't
any
case
it
not
so
this
change
that
eternal
from
Siena
before
an
Manan
and
you
need
banana
casts
to
go
to
see
if
you
are
going
to
do
you
have
another
and
and
two
an
advocate
for
that.
Basically,
where
you
go
will
change
the
hominin
because
you
will
need
just
will
not
scale
so.
D
You
see,
the
working
group
is
clearly
demanding
comparisons
between
any
tacit
feature
proposal
and
the
best
way
to
address
it
is
to
dress
in
the
draft
not
sending
emails
to
mailing
lists.
Please
have
comparison
section
Ares,
where
you
compare
what
working
group
is
proposing
versus
what
you're
proposing
with.
AD
W
Yeah
I'm
still
allowed
to
make
a
comment
so
just
for
the
any
cards,
maybe
I
just
discussed
with
Stuart,
so
the
anycast
would
work
which
we
are
doing.
Egress
protection,
but
we
are
what
zebu
is
trying
to
do
is
trying
to
protect
the
city
in
the
middle
of
the
that
the
city
which
is
the
middle
of
the
SRT
stack
right.
So
it
can
be
like
a
P
note,
it
is
in
the
stack
now
that
fails.
It's
not
the
egress
known
anymore.
So
that's
why
you
can't
have
like.
Why
would
you
do
any
cards?
W
W
Y
If
you
don't
want
to
design
a
good
network,
we
could
do
it
here.
If
you
want
you
know,
the
thing
is,
is
if
you're
using
ipv6
and
you
have
link
local,
you
don't
need
unique
global
addresses,
so
use
link
local
addresses
to
realize
your
IGP
and
then
all
the
addresses
have
unicast
at
each
hierarchy.
So.
AC
I
have
one
comment
here,
so
there
is
an
alternate
proposal
in
spring
working
group
for
node
protection
for
SRT
paths
that
talks
about
the
any
cost
to
solution,
as
well
as
the
keeping
the
forwarding
plane
for
a
longer
time.
The
the
proposal
that
stiffened
talked
about
is
explained
in
the
draft
I
would
request.
I
do
want
to
take
a
look.
Thank
you.
Yes,
thank
you.
We
can
look
in
that.
Nobody
know
that
those.
AD
D
So
acceleration
here
first,
we
need
to
decide
with
link
working
group
chairs
where
we
do
this
work,
because
it's
partially
belong
here,
because
the
first
year
out
partially
in
spring,
because
seventh
routing
and
then
potentially,
we
should
be
able
to
conversion
solutions
and
decide
whether
we
need
both
or
we
agree
on
something
that
addressing
everything
and
so
to
your
next
command.
For
the
largest
reduction.
We
do
need
to
discuss
with
spring
working
group
what
we
do
and
where
the
swap
purpose.
AH
AH
We
removed
up
specific
this
community
timestamp,
editing
the
dr2
version,
and
so,
but
we
added
description
to
refer
to
the
interface
one
and
we
added
processing
up
and
crushes
up
example,
and
we
also
removed
up
dynamic
learning
example,
because
that
was
quite
obvious
one,
and
so
this
is
all
the
changes
and-
and
we
think
that
we
already
solved
all
the
remaining
issues.
So
we
we
think
that
the
next
step
is
that
we
services
it
more
comments
on
this
draft
and
also
waiting
further
working
group.
Last
call.
E
Jeff
house
I've
read
multiple
versions
of
this
draft.
We
are
doing
a
study
squirrel
through
simply
addressing
small
issues.
I
would
like
to
suggest
to
the
chairs
that
this
was
ready
to
go
probably
three
versions
ago
and
that
a
working
group
last
call
would
probably
push
these
to
completion
a
lot
faster.