►
From YouTube: IETF115-SPRING-20221108-0930
Description
SPRING meeting session at IETF115
2022/11/08 0930
https://datatracker.ietf.org/meeting/115/proceedings/
A
Confirm
that
you
are
hearing
RS,
it
would
be
appreciated.
Thank
you.
A
E
A
A
For
ITF
115,
so
if
you
are
in-person
participant,
you
need
to
to
sign
in
using
mythical
multicolite.
A
This
is
being
used
as
blue
sheet,
and
also
this
thing
is
being
used
to
to
queue
on
the
bike,
so
you
have
to
enable
queue
on
meticulite
before
going
to
the
mic
so
that
we
can
under
a
single
queue
for
both
Olo
code
and
remote
attendees,
but
obviously
keep
your
audio
and
video
off.
If
you
are
not
using
the
light
version,
I
saw
something
very
important:
everyone
need
to
wear
a
mask
anytime,
except
if
you
are
actively
drinking
eating
or
speaking
at
the
mic.
A
This
is
very
important.
This
is
going
to
be
checked,
including
during
the
meeting.
So
please,
please
make
sure
you
have
a
mask
if
you're
a
remote
make
sure
your
audio
and
video
are
off
unless
you
are
actually
presenting
on
the
usually
your
headset
is
a
is
a
better
audio
quality.
A
A
Ihf
is
moving
to
a
new
Wiki
system,
so
spring
working
up
has
moved
to
the
new
Wiki
it's
available
at
the
following
URL
the
week
has
been
slightly
reorganized.
It
has
also
been
updated.
So
thank
you
for
being
about
that.
You're.
Welcome
to
to
check
whether
the
content
is
correct
during
the
migration,
especially,
you
can
check
whether
your
request
for
last
call
or
adoption
code
is
still
there
and
still
correct.
A
Okay,
so
since
July
we
have
adopted
a
new
policy
in
the
working
group,
so
we
now
among
data
an
implementation
section
in
the
draft,
at
least
before
working
up
last
caller.
A
A
You
can
put
not
known
it's
fine,
but
just
to
check
that
is
not
lost,
it's
not
blocking
for
progressing
to
ASG,
but
we
do
request
the
best
effort,
so
you
can
feel
non-reporter
does
not
apply,
but
something
needs
to
be
written.
A
We
also
ask
for
a
statement
that
the
implementation
is
compliant
with
all
mercen
should
in
the
specification
and
if
not,
which
one
are
not
followed.
A
A
So
please
review
and
comment
it's
important
for
specification
to
be
to
be
supported
by
the
group
and
to
be
reviewed
from
a
technical
standpoint.
So
it's
very
important.
Otherwise
the
graph
will
not
progress.
A
With
nsh
integration
with
with
segment
protein.
A
C
Good
morning,
everyone,
my
name,
is
Rakesh
Gandhi
from
Cisco
Systems
I'm,
presenting
the
mpls
network
action,
header
and
coding.
Solutions
draft
on
behalf
of
the
co-authors
listed
many
thanks
to
all
the
authors
for
working
together,
collaborating
and
producing
this
converge.
Unified,
Solutions
I,
really
appreciate
it.
Okay
next
slide,
please.
C
Yeah
many
thanks
to
all
the
contributors
they've
been
co-authors
as
well,
but
with
the
limit
of
five
authors.
So
many
many
thanks
for
all
your
inputs
and
contributions
next
slide.
Please
and
next
slide,
please
so
in
agenda.
We
have
this
being
presented
in
past
tomorrow.
There
is
more
time
allocated
for
this
presentation
in
past.
There
will
be
good
discussion
on
it
as
well
and
we're
just
presenting
in
spring
because
of
some
Sr
mpls
implications
on
it.
The
scope
is
we
look
at
the
scope,
the
solution
that's
presented
in
the
draft.
C
Some
we
have
reserved
our
codes.
There
is
some
ordering
there's
some
good
discussions
happening
on
the
action
ordering
backwards,
compatibility
and
the
next
steps
next
slide.
Please.
C
So
the
scope
of
this
draft
is
for
the
mpls
encoding
Solution.
That's
in
the
label
stack.
There
is
a
another
draft
for
the
Post
stack
highest
draft,
so
the
both
of
them
together
form
the
complete
mne
solution
and
the
scope
is
to
meet
them
in
the
requirements
and
framework.
That's
already
adopted
by
the
mpls.
A
working
group
next
slide
please.
C
So
there
are
two
parts
to
the
solution:
one
is
the
network
action,
the
sub
stack
header
and
the
second
is
the
action
encoding.
The
header
is
basically
a
bspl
label
indicate
indicating
that
there
is
a
substance
there,
which
is
encoding
parameters
like
the
scope,
the
ordering
length
and
whatnot,
and
the
action
is
basically
a
TLP,
so
opcode
length
and
the
ancillary
data
is
the
value
next
slide.
Please
so.
C
Test
mentioned:
there's
a
bspl.
There
is
a
length
of
the
header
for
the,
so
a
lot
of
work
has
been
gone
in
defining
this
encoding
for
easier
parsing,
especially
in
Asic.
So
there
is
a
land
that
the
scope
is
there.
If
you
want
to
do
on
the
end
points
or
hop
by
hop
or
on
select
nodes,
there
is
some
discussion
ordering
there's
a
beat.
This
is,
it
must
be
executed
or
the
order
is
specified
or
it's
an
optional
next
slide.
C
Please,
and
the
second
part,
is
the
X
and
encoding
part,
which
is
a
tlv,
so
you
have
op
code,
the
data
and
the
length
of
it.
So
it's
not
recognized
using
the
land.
It
can
be
skipped.
There
is
a
known
awkward
handling.
So
if
you
don't
recognize
it,
there
is
a
way
to
specify
either
a
Skip
and
go
to
the
next
one
based
on
the
land
or
drop
the
packet
or
just
forget
about
M,
A
and
just
forward
the
packet.
C
So
this
is
basically,
if
you
can
see
this
as
a
toolkit
and
how
application
will
use
this
toolkit,
it
will
be
up
to
the
applications.
So
the
the
programming
aspects
of
it
and
all
the
parameters
are
in
the
solution.
C
Next
slide,
please
so,
for
there
is
a
substack
per
scope
so
for
the
hubby
Hopper
selector
a2n
scope,
there's
a
sub
stack.
This
is
defined
to
make
the
parsing
and
processing
easier
on
the
midpoint
nodes
next
slide.
Please.
C
So
we
have
reserved
some,
of
course
there
are
four
or
five
of
them.
First,
one
is
to
say
where
the
offset
is
for
the
Post
Tech
data.
So
this
is
to
allow
the
the
gsch
kind
of
existing
data
or
header
to
be
carried
another
one
is
an
OP
code
to
just
if
you
just
want
to
carry
the
flag
based
accents.
There
is
a
third
one
is
for
the
flag,
based
Excellence
with
data,
and
then
there
is
a
another
one
to
carry
the
post
stack
network
access
as
well.
C
So
that
was
that's
on
the
next
slide
yeah.
So
we
can
skip
this
as
well
yeah,
and
there
are
two
special
of
course
one
is
the
upper
127.
So
it's
a
seven
bit
of
course.
In
case
we
run
out
of
it
unlikely,
but
there
is
a
way
to
extend
it,
and
126
is
if
there
is
no
aim
stack
action
and
there
is
only
a
post-tech
accents.
126
is
a
reserved
for
it
next
slide
please.
C
So
there
is
some
discussions
about
ordering
how
it
should
be
executed.
So
there's
an
OB
Define.
If
it's
set,
then
it
must
be
executed
in
the
order.
That's
encoded:
if
not,
then
it's
based
on
how
the
application
Define
it
there
is
some
discussions
about
what
should
be
done
in
stack
or
post
stack,
so
there
are
some
flag
base
so
how
that
should
be
executed
as
well.
So
there
are
some
examples
in
the
draft
on
what
the
authors
think
can
be
done.
Welcome
your
comments
on
that
next
slide,
please.
G
C
So
there
is
a
lot
of
brainstorming
and
thinking
went
in
to
make
sure
the
solution
is
backwards
compatible,
especially
if
you're
only
going
to
do
the
end-to-end
processing
or
end-to-end
m
a
we
don't
want
to
upgrade
the
penultimate,
node
or
midpoint
nodes.
It
should
be
transparent
to
it
make
sure
the
ecmp
is
not
adverse
affected,
make
sure
that
it
doesn't
Alias
when
existing
reserve
bspl
it
can
coexist
with
gsh.
So
a
lot
of
the
thinking
has
gone
in,
but
there
could
be
something.
C
We'll
skip
the
advantages
so
at
this
point,
I
welcome
your
review
comments
and
feedbacks.
As
mentioned,
this
is
a
converge,
unified
solution.
Among
all
the
authors,
we
receive
lots
of
good
review
comments
and
we
are
addressing
these
comments.
Some
cleanup
is
required.
There's
some
discussions
on
the
scope
at
ordering,
so
on
all
the
aspects
that
we
presented
today.
Let
us
know
if
you
have
feedback
comments
and
that's
all
I
had.
H
Grand
whiskey
Erickson,
so
you
mentioned
that
the
offers
paid
special
attention
to
end
to
end
operation
of
MMA,
so
basically,
when
anime
is
expected
to
be
operated
only
at
egress,
but
my
understanding
is
that
there
are
other
options.
There
is
a
hubby
Hub
and
there
is
what's
called
select.
H
C
Yes,
definitely
so
there
is
a
sub
stack
for
hubby
hop,
and
that
is
a
substack
for
end-to-end.
Only
requirement
for
backwards
compatibility
is
that
the
m,
a
bspl
if
it's
exposed
or
under
received
packet
at
the
no
receiving
it,
and
that
does
not
understand
the
bspl.
It
may
drop
the
packet.
C
So
that's
the
only
constraint
for
the
backwards
compatibility,
so
the
Ingress
and
cap
node
needs
to
make
sure
that
along
the
path
is
never
received.
Vspl
is
not
at
the
top
on
the
received
node
that
you
cannot
understand.
H
Well,
that
was
not
really
my
question.
Let
me
rephrase
so,
for
example,
imagine
that,
because
of
maximum
stack
that
along
the
path
we
can
put
Hub
by
heart
and
end
to
end
at
the
bottom
of
the
stack
so
to
do
hard
by
hop
and
end
to
end
Network
actions
with
the
current
encapsulation,
do
you
see
it's
possible
to
do
it
on
the
one
special
purpose
label,
or
that
will
require
two
special
purpose
labels,
so
two
Network
action
sub
stacks
so.
C
The
the
one
special
purpose
label
we
can
have
multiple
copies
of
the
sub
stack
and
we
can
have
copies
with
different
scope,
so
you
would
have
one
sub
stack
with
hop
by
hopscope
and
another
sub
stack
with
I2,
Esco
and
hubby.
Hops
substack
will
contain
the
iom,
and
the
second
sub
stack
with
the
same
bspl
will
have
the
flow
identifier
for
I2.
Okay,
thank.
H
A
A
And
now
we'll
we'll
move
to
next
presentation
for
interlayer
Network
programming
presented
being
U1.
I
Good
morning,
everyone
this
chapter
is
on
the
topic
of
slv6
foreign
Network
programming
and
the
authors
are
from
Huawei
and
China
mobile
next
slide.
Please.
I
First
I
will
introduce
the
background
operator,
usually
have
a
multi-layered
network.
The
layer
3
is
normally
IP
based
technology,
where
different
technology
could
be
used
for
the
underlying
layers.
For
example,
there
are
two
another
one:
the
closed
layer,
Network
programming
and
the
optimization
is
expected.
It
is
considered
more
efficient
compared
with
the
independent
planning
for
each
layer,
but
it's
also
considered
more
complicated
as
the
way
6
enabled
Network
programming
by
encoding
Network
instructions
in
IP
Handler,
but
currently
only
the
network
instructions
related
to
ipd
are
defined.
I
So
this
document
describes
the
use
cases
for
the
internal
Network
programming
and
we
propose
a
new
srv6
function
for
this
purpose.
Next
slide,
please
foreign
use
cases
for
the
inter-layer
programming.
These
three
scenarios
are
very
common
and
widely
used
in
The
Operators.
Network
first
case
is
IP
and
Optical
Network
integration
in
many
scenarios
in
the
network.
The
underlay
of
IP
network
is
the
optical
network,
but
in
some
cases
Optical
parts
and
IP
parts
may
not
be
one
by
one
mapping,
so
the
reduction
to
Optical
process
may
not
be
fully
used
by
IP
layer.
I
Secondly,
the
optical
process
may
exist
between
adjacent
to
IP
nodes,
so
it's
not
too
reasonable
in
the
layers
rate
polity.
The
second
case
is
the
IP
and
MTN
integration.
The
MTN
is
Metro
transport
Network,
it's
a
new
technology
and
has
been
divided
in
itog.8310,
the
m-team
parts
layer
uses
TDM
switching
the
intermediate
nodes
to
avoid
Hub
to
hope,
I
hope,
IP
packet
display
team,
so
mq
nodes
can
support
both
per
hope,
ip40
and
it
also
supports
MTM,
Parts
cross
connect.
I
The
traffic
can
be
carried
using
IP
parts
and
team
parts
or
the
combination
of
the
two,
but
currently
srv6
only
can
be
used
for
the
IP
pass
programming
for
the
MTN
Network.
The
MTN
Parts
is
not
included.
The
third
use
case
is
traffic
steering
to
a
layer,
2
random
member
link.
Each
member
link
is
a
layer,
2
connection
without
layer
3
and
this
and
decency.
So
if
we
want
to
choose
one
specific
Layer
Two
member
link,
we
need
a
new
function
next
slide.
Please.
I
Also,
we
propose
a
new
slv6
which
called
end
XU
in
order
to
solve
the
inter-layer
program
in
the
use
cases.
This
endpoint
with
underlying
cross
connect
is
a
variance
of
the
and
X
Behavior.
It
may
propose
it
for
the
interlayer
programming
and
traffic
engineering.
The
end
actually
is
associated
with
an
underlay
interface,
which
connects
to
one
or
more
underlyings
or
connections,
and
the
light
and
15
from
the
slv6
and
progressing
in
the
current
RFC
8986
is
replaced
by
the
following
forward.
The
package,
through
the
underlying
interface
associate
with
seeds
next
slide.
Please.
I
In
this
slide,
we
give
application
example
to
use
the
index
using
IP
Optical
integration.
In
this
figure
we
can
see
there
is
a
IP
networks.
Only
the
optical
Network,
the
node
P1
to
pH
is
IP
nodes
and
the
node
01206
is
Optical
nodes.
If
we
want
to
set
up
a
package
transmission
connection
from
p7
to
the
right
pH,
with
normal
segment
routine
solution,
we
can
find
the
seedless
in
IP
network
is
p7
P1,
P2,
P3,
p8,
but
assume
there
is
Optical
parts
of
102
203.
I
I
I
We
give
another
example
in
this
slide:
it's
a
application
of
NX
during
ipmq
integration.
We
can
see
in
the
figure
in
the
figure
the
operator
earns
both
the
MTN
Network
in
the
metro
area
and
also
IP
network
in
the
backbone
area,
but
for
the
empty
nodes.
It
builds
two
separate
Network
layers,
the
layer,
3
function
in
the
topology
and
the
MTN
Parts
topology.
I
I
J
Hi
I'm
social
Einstein
from
ribbon.
Can
you
please
show
again
the
the
slide
with
the
no
one
one
little
bit?
Yes,
this
one
I
have
a
question
as
explained
the
optical
path
between
P1
and
P3
is
not
an
IP
an
iPad.
J
My
question
is:
when
P3
receives
an
IP
packet
from
p7
I
think
how
does
it
know
which
layer
two
encapsulation
should
be
pushed
on
this
packet
before
it
is
sent
by
the
optical
path,
because
without
this
encapsulation
P3,
most
probably
wouldn't
recognize
the
packet
as
an
IP
packet
and
wouldn't
be
able
to
forward
it
via
the
IP
pass
to
p8.
J
We
normally
we
ate
if
the
parts
were
an
IP
interface.
P1
would
try
to
resolve
battery
IP
address
or
something
like
that
using
our
Pro
Naval
Discovery.
But
since
the
Optical
path
is
not
an
hyper
pass,
such
resolution
seems
impossible.
E
How
well
ran
ran
Chen
you're
next.
K
Yeah,
okay,
hi.
We
agree
with
the
useless
it's
a
lion,
Network
programming
at
the
requirements,
and
thank
you
very
much
for
your
visitors
to
work,
and
we
also
have
similar
proposals
by
the
way
things
but
new.
As
always.
Six
and
the
bxc
we
have
here
for
interlier,
Network
programming,
the
srv6
and
the
pxc
function
mechanism
has
been
implemented
by
the
key
and
the
China
mobile
has
successfully
completed
the
basic
verification
of
this
function.
Hope
it
can
catch
you
and
the
working
group
attention.
Thank
you.
B
Go
kind
of
follow
on
to
snasha's
comment
that
the
whether
the
optical
path
is
IP
or
not,
is
I
think
not
clear,
but
n
dot
X
said,
which
is
in
RFC
8986
Network
programming
does
can
be
used
for
layer
to
bundle
member
states
and
therefore
I
mean
I.
Don't
see
any
reason
why
we
need
a
new
Behavior,
it
can
be
used
for
this
Optical
link
as
well
for
steering
into
Optical
path,
n
dot,
X
existing
one.
L
From
Huawei
and
firstly,
I'd
like
to
respond
to
Sasha's
comments
on
the
whether
a
P3
can
correctly
receive
the
packet
and
forward
it
I
think
it
is
a
good
question
rather
how
to
resolve
the
like
the
IRP,
all
the
layer
tool
information,
since
this
draft
currently
does
not
cover
that
details,
but
perhaps
we
can
consider
to
distribute
some
kind
of
information
through
some
control
plane.
That
is
a
reply
and
the
truth
comes
from
Curtin.
L
Our
understanding
is
the
current
and
Dot
X
behavior
is
very,
is
correct,
explicitly
defined
for
the
layer,
3
adjacency,
and
that
kind
of
information
are
used
for
the
to
build
the
leadership
topology
and
the
forwarding
pass
in
the
leadership
network.
We
Define
this
node.xu
to
separate
it
from
the
leadership.
Apologies
so
that
we
for
some
Network
behaviors.
We
don't
want
to
mix
the
literacy
adjacency
with
the
entirely
links
so
that
we
can
have
different
fast
computation
with
four
different
kind
of
service.
So
this
is
the
purpose
to
distinguish
these
two
food
behaviors.
L
E
M
Yes,
thank
you
I'm,
using
Leo
from
China
mobile
and
today
I
will
present
the
encapsulation
of
BFD
for
The
Advocate
policy.
This
is
the
first
presentation
of
this
draft.
M
Okay,
we,
when
we
use
the
BFD
session
to
monitor
the
srs6
policy
and,
as
we
know,
sos6
policy
has
three
levels
and
the
brt
session
runs
associated
with
the
second
list
level
and
when
we
have
decision
done,
the
set
the
associated
a
second
list
fail
and
the
traffic
will
stop
forwarding
and
with
this
specific
segment
list.
While
all
of
the
signals,
the
failed
of
the
active
candidate
parts
and
the
traffic
will
switch
over
to
the
backup
candidate
pass
and
all
of
the
candid
paths
fail
and
the
policy
field.
M
Why?
We?
The
second,
is
that
what
the
choice
of
the
address
for
the
BFD-
and
here
we
give
the
two
types
of
the
BFD
as
DMV
and
the
upfd
ubfd
means
that
we
use
the
bfpi
code
function
and
you
turn
on
the
reflector.
M
And
when
we
use
the
spfd
the
head
and
address,
we
can
use
the
iv6
address
of
the
head
end
and
the
and
the
two
and
the
address
we
use
the
endpoint
of
the
SV
rv6
policy
or
specified
by
the
local
configuration
and
for
the
ubft,
the
head
and
adjust.
We
use
the
x36
address
of
the
head
and
we'll
use
the
root
ball
or
non-root
ball.
For
the
tail
end,
we
must
use
the
root
Pro
ip6
address
of
the
head
end
next
slide.
Please.
M
And
we
have
the
two
modes
of
the
encapsulation
of
the
BFD
packet
and
the
the
the
difference
between
the
the
SRH
how
to
in
Captivate
the
SRH
and
for
the
track
mode
transport
mode.
The
SRH
is
inserted
after
the
ip6
header
and
the
tunnel
mode.
We
have
the
outer
ip6
header
with
the
SRH
and
the
inner
it's
for
the
BFD
packet,
like
the
plan
IPv6
pack
package
still
into
the
srv6
policy.
M
And
next
slides,
we
I
will
give
the
details
of
this
two
modes
for
both
the
fbd
and
ubfd
and
when
the
spit
in
chess
mode
transport
mode,
we
are
encapsulated.
Only
one
IPv6
header
and
the
SRH
and
for
the
sl0
position
should
be
the
athletic
feed
or
the
ip6
address
of
the
tail
end
node,
but
sometimes
the
last
segment
of
the
second
list
in
the
sr6
policy
does
not
belong
to
the
tail
end.
It
can
be
a
n
dot,
X
segment
of
the
parliament
Hub
or
a
bunny
seed.
M
M
When
the
sap
internal
mode,
we
are
encapsulated
the
inner
IPv6
header,
with
the
ability
packet
and
the
author,
with
the
IPv6
header,
with
the
SRH
and
at
the
same
so
zero
position
should
be
the
s36
seed
or
the
ip6
address
of
the
tail
and
node.
And
if
the
last
segment
of
the
second
listing
sr6
policy
does
not
belong
the
tail
end
and
maybe
a
n
dot
X.
But
did
this
I
know
that
without
the
USD
flavor,
if
if
he
he
has,
if
it
has
a
USD
flavor,
it
can
decapitulate
one
in
at
the
punishment
hub.
M
M
The
next
is
two
slides
about
the
ubfd
when
the
ubfd
in
transport
mode
that
it
is.
That
is
a
simple
thing
and
only
one
optimistic
header
is
used
to
encapsulate
the
echo
packet
and
for
the
SL
zero
partition.
We
use
the
SRA
6
seed
or
the
ip6
suggests
of
the
head
end
and
the
one
the
echo
package
arrived,
the
tail
end
it
it
will
use
the
destination
address
at
the
SL
0
petition.
M
One,
the
usbsp
in
the
tunnel
mode,
we
encapsulate
the
inner
Acoustics
header
and
the
outer
s36
encapsulation
and
the
destination
of
the
inner
fps6
is
the
address
of
the
head
and
the
node
and
the
the
Outer
Outer
ip6
header.
We
use
the
SL
zero
partition
should
be.
The
isab
is
six
seed
or
the
IPv6
address
of
the
tail
end,
and
that
is
the
same
scenario
that
if
the
last
segment
of
the
srv6
policy.
M
H
E
H
That
you
explain
different
modes
of
a
certain
encapsulation
in
the
SAR
modes,
but
I
don't
see
that
it's
different
from
encapsulating
payload
and
services
that
to
be
monitored
using
BFD
and
I.
I
cannot
find
anything
that
is
specific
to
BFD
being
used
over
this
a
service,
6
modes.
M
We
just
work
on
the
how
to
encapsulate
the
BFD
packet
in
the
atheristics
policy
and
it
may
be
have
a
major
difference
from
for
the
normal
date
package.
H
But
if
you,
okay,
if
you
suggest
that
DFD
should
be
encapsulated
differently
from
the
payload,
then
how
you
guarantee
that
PFD
will
follow
the
same
path
and
will
be
treated
the
same
as
a
service
that
you're
trying
to
monitor.
H
H
But
then
the
question
is:
if
that's
the
case,
then
what
are
you
monitoring?
Because
if
you
encapsulate
differently
on
a
service
six
level,
then
the
packets
would
not
follow
the
same
path.
E
O
A
O
I,
yes,
my
question
is
the
the:
why
can't
the
sspfd
packets
can
take
the
use
the
standard,
srv6
programming,
or
is
there
anything
special,
that's
being
done
in
this
graph
other
than
using
the
standard,
srv6
programming.
O
Okay,
so
what
what
is
what
would
be
the
benefit?
I
guess
of
this
graph?
What
is
this
graph
adding
outside
of
srv6
programming?
O
That's
that
would
need
to
be
standardized,
I,
guess
good,
special
outside
of
normal
behavior
of
a
packet
I.
Guess
it's
being
either
to
be
monitored
from
the
ahead,
and
you
know
to
the
tail
end.
P
E
P
P
P
P
Please
consider
taking
one
of
the
documents
and
simply
having
an
individual
encapsulation
example
for
each
of
them,
since
the
procedures
are
largely
the
same,
this
will
probably
save
yourselves
a
lot
of
grief
and
when
things
come
around
to
being
reviewed
by
BFD,
you
know
for
your
last
call.
Process
we'll
simplify
things.
E
Q
I'm
going
to
talk
about
the
topic
of
segment
routine,
that's
the
solution
for
hierarchical
ITF,
Network
license
on
behalf
of
all
the
co-authors.
Q
The
hierarchical
HF
Network
slices
described
the
networks
like
this,
which
can
be
further
sliced
or
convent
hierarchically.
For
example,
one
of
the
possible
scenarios
is
that
an
industry
slides
needs
to
be
divided
into
multi-customers
like
this.
So
generally,
the
idea
of
the
draft
is
to
propose
a
true
level.
Q
The
network
resources
is
also
partitioned
in
a
hierarchical
manner,
kicking
bandwidth
resources
as
an
example.
The
level
one
and
wise
is
guaranteed
by
a
little
series.
Sub
interface
with
dedicated
bandwidth
and
the
prefix
state
or
srv6
locators
is
used
as
a
datably
identifier.
Q
The
level
2
bandwidth
is
guaranteed
by
kills
with
dedicated
bandwidths
under
the
little
three
step
interface
and
the
NRP
ID
is
used
as
the
data
plan.
Identifier.
Q
Q
But
here
for
the
example
when
we
actually
deployment
theater,
it's
much
more
complicated
for
low
to
NRP
to
be
mapped
to
multi
slices,
so
in
actors
in
ISO-
and
we
think
it
makes
little
sense-
is
the
resource
is
already
finally
divided
as
previously
described
so
here.
For
the
example,
the
level
2
NRP
to
slice
is
shown
as
one-to-one.
Q
This
is
one
example
for
the
whole
solution
from
the
bottom.
Up
to
the
physical
network
is
divided
into
two
inner
pieces
through
the
flex
elbow
and
the
hnrp
is
mapped
to
two
one
slice
one
is
for
education
and
the
other
is
for
the
healthcare
and
the
corresponding
black
alcohol.
Id
is
128
and
129
fold
the
education
slides.
There
are
two
customers
University
one
University
two:
they
have
different
exercise
points
and
interactions.
Q
Based
on
the
topology
previous,
the
resource
partition
is
like
this
slide
shows
for
pe1
the
physical
interface
one.
There
will
be
two
sub
interfaces
and
the
under
which
there
will
be
two
or
one
kill
respectively.
Q
Q
The
final
result
is
that
for
the
normal
education
traffic-
it's
it
is
encapsulated
with
the
understated
and
also
associated
with
the
specific
flux,
alcohol
and
it's
to
forward
data
through
the
layer
series
of
interfaces
and
photos,
specific
traffic
from
University
one
and
the
university
to
accept
the
activistics
either
SRH.
S
E
T
T
T
In
your
University
example,
you
might
have
the
overall
set
of
universities,
those
resources
in
a
slice
that
was
being
managed
by
Sr
and
some,
but
not
all,
of
the
slices
for
specific
universities
being
in
either
Sr
or
mpls,
or
something
else
and
I
wanted
to
confirm
whether
you
thought
that
that
was
possible
in
in
the
approach
you're
taking
or
whether
the
linkage
to
the
underlying
nrps
made
that
impractical
and
the
last
question
was.
This
is
obviously
very
early
work.
T
A
zero
zero
I
didn't
see
any
IPR
disclosures
for
it,
but
I
wondered
if
the
authors
knew
of
any
that
were
in
preparation.
Thank
you
again
for
the
interesting
draft.
Q
Hi,
okay,
thank
you
for
your
comments.
I
will
try
to
answer
your
question.
If
I
cannot
answer,
it
clearly
welcome
to
move
it
to
the
main
list.
First
of
all
about
the
solution.
I
think
it's
an
informational
solution.
Yes,
you
are
right
and
there
is
no
other
extensions
on
practicals.
T
Q
Q
The
second
one
about
the
true
level,
hierarchical,
slides,
I,
think
the
first
level
we
use
the
flex
algo
is
for
the
control
plan
and
because
of
the
limitation
of
the
control
plan,
it's
very
difficult
to
realize
the
multiple.
Q
Slice,
so
we
proposed
that
the
second
level
slice,
which
is
from
the
forwarding
data
plate
data
plan,
so
the
second
level
I
think
they
are
also
based
on
the
srv6
segment
routine,
because
we
combined
it
with
the
SR
policy
of
the
services
package.
Folding
an
rpid
such
as
so
I
think
the
total
solution
is
related
with
the
s36.
T
Q
T
Was
just
a
quick
question
of
whether
you
knew
of
any
IPR
disclosures
in
process.
R
R
O
Is
there
in
your
examples,
you
had
two
layers
of
hierarchy:
is
it
possible
to
have
more
than
two
layers
hierarchy
as
well?
I
guess,
with
with
the
the
semantics
of
the
draft,
is
there?
Is
there
anything
special,
that's
being
done
other
than
what's
defined,
and
you
know
that's
part
of
segment
where
I
think
the
standard
is
there?
Are
there
any
a
special
Diana
requirements
related?
Q
Thank
you.
Thank
you
very
much.
It's
a
good
question.
We
also
have
thought
about
it
later
from
now
on
the
full
Channel
mobile,
it's
a
real
natural
it.
It's
just
the
true
level
slice,
for
example,
I
think
a
different
business
units
with
the
operator
government,
Affairs
departments
or
Enterprises
of
Affairs
Department.
Q
Under
this
there
will
be
different
users
and
another
example
for
the
operator
we
think
for
China
mobile.
There
are
several
sub
operators,
several
sub
operators
located
at
different
provinces.
They
also
want
to
manage
their
own
provinces
with
business
independently.
So
I
think
this
financial
is
very
special
for
China
mobile
operator.
E
A
O
You,
okay,
give
me
one
second
I'm
just
going
to
try
to
make
an
adjustment
to
my
Nick,
and
just
if
you
could,
let
me
know
if
it
sounds
any
better.
O
O
So
a
recap:
some
ministry
as
to
how
this
draft
came
to
spring.
So
during
the
adoption
of
draft
PCP
set,
pmtu
a
need
for
a
document
and
spring
was
requested
and
confirmed
that
both
the
pce
and
spring
working
groups,
so
an
IDR
document,
was
adopted
in
April
2020.
O
A
discussion
was
brought
up
by
Caitlin
that
the
concept
of
path
MTU
assess
our
policy
and
its
applicability
applicability
should
be
first
defined
in
the
spring
working
year
before
we
introduce
signaling
aspects
in
bgp
as
a
result
of
the
PC
PMG
extension
adoption
called
the
draft
piece
of
pmtu
was
maintained
only
protocol
specific
extensions
of
details,
while
the
SR
policy
path,
M2
definition
framework,
was
to
be
developed
in
the
spring
working
group
as
a
standards
track
document
to
ensure
vendor
interoperability
related
to
srpmtu,
Concepts
and
computation.
Details.
O
This
topic
came
up
is
a
critical
issue
to
be
addressed
during
the
IDR
working
group,
2020
to
GPS
our
policy
to
MTU
adoption
call
and
now
again
during
PC
workload.
2022
pm2
adoption
call,
as
mentioned
above
and
due
to
the
criticality
of
solving
this
issue,
related
to
handling
fragmentation
and
fragmentation
avoidance
procedures
using
SRP
MTU
next
slide
updates
from
revision
zero
to
revision,
one
so
mostly
cleanup.
There
was
an
update
as
well
on
the
Securities
consideration
section.
O
So
the
addition
of
the
SRP
MCU
constraint
information
can
be
sensitive
in
some
defined
and
can
be
used
to
influence
the
SR
path,
setup
and
selection
results
in
it
and
which
results
in
an
adverse
effect.
The
protocol
extension
that
includes
the
SRP
MTU
must
be
take
must
take
this
into
consideration.
This
document
does
not
define
any
new
protocol.
Extension
and
test
does
not
introduce
any
further
security
considerations
and
then
as
well.
Another
update
was
the
section
section.
Three,
the
SRP
empty
enforcement
reference
to
separate
extension
to
handle
pm2
draft
result
was
also
obtained
next
slide.
O
O
So
if
anyone
can
think
of
after
reading,
through
the
draft
of
any
caveats
related
to
tilfa
a
computation
at
the
plr
node
to
the
to
the
merge
point
in
any
any
any
any
of
that
bypass
Loop
and
any
fragmentation
crappy
outside
related
there,
another
question
that
we
have
and
we
want
to
get
feedback
from
them
from
the
working
group
as
well
as
srpmq
srv60
encapsulation,
and
any
fragmentation
caveats
there.
O
When
that
happened,
and
then
the
last
question
is
The:
srpmt
Binding,
Sid
path,
computation
caveat
so
on
the
source
node,
where
the
as
our
policy
candidate,
Panthers,
bouncing
forwarding
plate,
then
The
Binding
sit
is
popped
and
then
the
labels
are
pushed
onto
the
label
stack
or
in
the
srv6
case,
the
SRH
header
and
the
srv6
package
is
pushed
onto
the
label
stack.
O
Are
there
any
computation
caveats
that
were
missing?
We
just
want
to
make
sure
that
we
we
recover
all
our
bases
in
the
draft
and
any
any
questions
or
comments.
We
really
appreciate
any
feedback
on
the
graph
next
slide,
and
so
this
traffic
is
fairly
straightforward.
O
What
would
like
to
take
any
comments
to
the
mailing
list,
but
we
would
like
to
ask
for
work
group
adoption
on
this
job.
Thank
you.
R
A
So
we
are
now
moving
to
7600ps
interworking
from
Square.
U
Hello,
everyone
I,
am
swadesh
Agarwal
from
Cisco
Systems.
This
document
describes
the
srv6
and
mpls
interworking
procedures
and
I
present
this
on
the
behalf
of
all
my
co-authors.
Next
slide,
please
yeah.
This
is
slides,
shows
the
simple
view
of
the
interworking
scenarios
in
the
green.
We
are
showing
the
srv6
domain
in
the
orange
we
are
showing
mpls
domain
and
the
tasked
overall
is
for
the
interworking
functions
has
to
be
executed.
U
The
document
divides
the
problem
into
two
set
of
things:
one
is
a
transport
interworking
and
the
service
interworking
in
the
transport
interworking.
What
happens
is
and
from
the
end,
to
end
from
the
service
point
of
view
it's
running
either
srv6
vpns
or
the
mpls
vpns
and
since
end-to-end
is
an
srv6
VPN.
We
need
to
provide
a
transport
reachability
to
locators
and
that's
what
we
call
srv6
over
mpls.
We
we
do
interworking
functions
for
the
locators,
on
these
Dash
oval
boxes
and
for
the
when
we
have
a
mpls
vpn's
edge
service
is
running
end
to
end.
U
We
need
to
provide
a
interworking
procedures
on
those
test
boxes,
ovals
on
whenever
the
data
plate
changes
and
that
happens
for
mpls
or
an
srv6
domain
and
the
and
the
last
case
we
cover
is
service
intervoking.
In
this
case,
we
have
to
provide
connectivity
between
srv6,
VPN
and
mpls
VPN,
and
you
need
to
do
a
intervoking
function
on
some
data
working
box.
Next
slide,
please
so
for
interworking
functions.
The
document
describes
new
behaviors
one
is
the
n
dot
DTM?
What
this
Behavior
does?
U
U
So
these
are
the
and
hidden
behaviors
we
have
defined.
So
what
happens
is
when
the
H
dot
encaps
is
applied
to
an
mpls
label,
stack,
that's
what
we
call
H
Dot
and
caps
dot
m,
and
there
is
both
the
regular
Behavior
and
the
reduced
behavior
for
it
next
slide.
Please
so
Sr
policy
is
bound
to
an
binding
set.
So
in
the
procedures,
if
the
SRM
pillars,
label
can
be
bound
to
an
srv6
policy
as
well
as
sr66
can
be
bound
to
an
Sr
mpls
policy.
U
So
we
call
these
in
the
draft
as
interconnecting
binding
set
and
we'll
we'll
see.
This
concept
is
used
to
provide
those
interworking
functions
in
the
next
slide.
Next
next
slide,
please,
as
already
mentioned
when
the
services
are
of
the
same
type
end
to
end,
and
we
need
to
provide
a
reachability
either
to
her
if
it's
srv6
VP
into
a
locator
or
if
it
is
a
mpls
VPN
services
to
a
p
loopback.
So
what
it
is
draft
does
is.
It
extends
the
two
well-known
transport
procedures
to
provide
this
interworking
to
create
this
interworking
control
plane.
U
One
is
a
srpc
base
which
can
provide
both
the
best
effort,
as
well
as
the
intend
away.
Another
one
is
the
bgp
interdomain
routing,
which
is
right
now
described
for
best
effort,
but
we
will
Define
it
for
the
intent
aware,
based
on
bgp,
also
in
the
future
versions.
Next
slide,
please
so
just
to
go
over
the
procedures
the
Legends
used
are
the
green
boxes
are
sizes
capable,
and
it
shows
how
the
locators
have
been
referred
in.
U
U
So
we
see
these
are
srpc
based
procedures.
So,
let's
say
between
1
and
10,
we
are
running
the
srv6
VPN
Services,
it's
been
advertised
by
the
10
learned
by
the
one,
a
VPN
service
and
what
the
endpoint
or
locator
we
want
to
reach
is
of
a
node
10
B10
is
a
locator.
We
want
to
reach
it's
a
multi-domain
network.
One
doesn't
know
how
to
reach
one
request:
an
srpc
for
the
a
certain
objective
or
a
constraint
for
a
path
to
10.
Srpcs.
Look
at
the
constraint.
U
It
figures
out
the
let's
say:
node
2,
Note,
5
and
node
8
are
in
the
path
to
provide
that
objective
or
a
constraint,
but
since
srpc
is
having
the
bgp
ls
feed
from
each
other
domain,
srpc
sees
there's
an
inconsistent
data
plane
happening
on
node,
four
and
seven.
So
what
it
does
is
whatever
interconnecting
sets.
We
use
the
the
procedures
we
Define
of
interconnecting
sets.
It
use
that
behavior.
U
What
it
does
is
on
the
Note
4,
since
we
have
to
take
the
policy
200
fabulous
domain,
it
creates
an
mpls
policy,
but
the
The
Binding
said
it's
bound
to
is
an
srb6
Beast,
SRB
behavior
of
the
BM
type
and
that's
and
that
srb6
BC
is
being
once
srpc
creates
that
policy
on
the
four
and
it
and
that's
the
one
which
is
programmed
at
the
one
with
one
of
the
segments
which
takes
to
the
node
4
and
which
will
get
expanded
on
node
4
to
the
policy
MPS
policy
and
that's
where
the
packet
would
Transit
transition
through
the
mpls
domain.
U
The
objective
was
to
reach
node
4
node
5
from
4,
and
that's
what
the
prefix
it
we
have
put
in
the
packet
and
then
the
last
segment
we
have
put
is
a
note
7
where
on
the
Note
7,
and
it
would
be
a
regular
sr6
domain
and
it
will
go
to
the
next
segment,
which
is
node,
8
and,
of
course,
the
VPN
segment
which
will
take
to
the
note
10..
So
that's
the
way
interconnecting
by
index.
It
is
used
to
provide
that
6
over
M
reachability
through
srpc
next
slide,
please.
U
So
it
is
similar
Behavior.
It's
a
similar
operations
that
I
have
already
explained
for
6
over
M.
The
only
difference
is
here:
the
edge
services
are
mpls
VPN
services,
so
we
need
to
create
this.
Srpc
request
is
for
the
P
loop
back.
It's
it's
called
for
a
certain
C,
comma
e,
which
is
an
mpls
in
the
mpls
case
here
on
the
Note
4.
U
What
we
are
doing
is
we
are
putting
an
srb6
policy,
because
middle
domain
is
an
srb6
domain
that
policy,
but
it
is
bound
to
an
mpls
B
said
and
that
mpls
BC
that
is
programmed
for
the
policy
on
node
4
is
programmed
as
one
of
the
segments
on
the
Node
one.
So
now
what
happens
is
when
the
packet
leaves
one
it
is
going
through
the
segments
in
the
left
domain
on
the
Note
4.
U
It
is
expanded
to
a
srv6
policy
with
the
the
next
header
being
137,
because
getting
an
MPS
payload
below
it,
and
then
the
packet
will
be
reaching
the
Note
7
and
the
regular
mpls
operations
will
happen
from
there
onwards.
Next
slide,
please
so
these
are.
We
have
covered
the
srbc
PC
based
procedures,
where
PC
was
aware
of
this
interworking
procedures
and
was
creating
those
using
the
enter
connecting
sets
to
create
those
pass
through
to
the
different
type
of
data
plane,
but
similarly
bgp
interdomain
procedures
can
be
used.
So
in
this
again
it's
over.
U
It's
a
scenario
of
6
over
M,
where
the
services
are
Sr,
V6,
VPN
services.
So
now
what
we
need
to
do
is
when
10
advertise
the
service
to
node
one
then
we'll
also
advertise
the
locator
through
a
locator
to
node
7
from
Note
7
to
Note
4
to
node
one
when
it's
advertised
between
7
to
4.
What
you
need
to
do
is
it's
a
it's,
a
regular
IPv6
prefix
over
an
elude
mpls
transport.
U
So
this
is
the
case
of
mplus
being
in
the
edge,
and
that
means
it's
an
mpls
VPN
services.
So
we
need
to
provide
a
reachability
to
node
10
that
has
to
be
learned
by
the
node
one,
which
is
learning
the
services
from
the
10.
to
provide
that
reachability.
Today
we
know
bgp3107
procedures.
Seamless
mpls
is
used.
Where
what
will
happen
is
on
on
the
7.
U
U
So
these
are
the
cases
of
Transport
interworking
where
we
were
trying
to
provide
a
reachability
towards
the
service
endpoint.
But
in
this
case,
what
we
want
to
in
this.
This
is
a
service
interworking,
where
we
want
to
provide
a
connectivity
between
mpls,
VPN
and
srv6
VPN
running
on
the
Node
1
and
node
7..
So
what
happens
in
this
case?
Is
there?
Are
these?
Are
these
are
working
on
the
existing
Gateway
procedures?
What
happens
is
on
node
4?
U
We
have
a
service
awareness
because
we
want
to
terminate
both
the
services
as
the
as
well
as
the
underlay.
So
since
we
want
to
terminate
both
the
what
we
can
do
is
we
can
have
a
service
Awareness
on
node
four.
So
whatever
is
destinations,
VPN
destinations
are
learned
from
the
sr6
domain.
They
are
put
into
the
work
on
the
service
on
the
service
box,
Note
4
and
then
reordered
it
towards
the
mpls
side.
U
So
there
is
MPS,
VPN
and
s56
VPN
and
on
the
service
box
there
is
a
verb
IP
forwarding,
which
is
when
you
receive
a
traffic
from
one
side,
you
Decap
you
look
at,
you,
do
an
iplooker,
and
then
you
get
the
encapsulation
for
the
other
side
and
that's
the
way
as
we
provide
the
interworking
solution
for
this
scenario.
Next
slide
please.
U
So
this
is
the
last
case
where,
instead
of
having
a
service
Awareness
on
the
Note
4,
what
we
can
do
is
on
the
interworking
box.
We
can
do
a
translation.
That
means
a
VPN
label
could
be
translated
to
an
vpn's
srv6
service
set
or
the
SR
physics
service
set
can
be
translated
to
a
VPN
label,
using
a
procedure
similar
to
interest
option
b,
doing
the
next
12
cell
phone
dot
box
next
slide.
U
Please
yeah
somebody
is
the
staff
was
presented
first
published
in
2018,
and
it
has
the
data
plane
and
the
associated
control
plane
procedures.
It
defines
a
new
Behavior,
DTM
and
DPM
as
well
as
use
the
introduce
the
concept
of
interconnecting
binding,
sits
and
it
uses
how
to
pass
through
a
data
plane
which
is
not
consistent.
U
It
relies
both
on
the
srpc
and
bgp
based
Solutions
next
slide.
Please
next
slide,
please
yeah.
So
one
thing
we
have
done
is
from
the
last
ITF.
We
have
moved
the
bgp
protocol
extensions
that
we
have
to
a
separate
draft,
so
it
can
be
reviewed
in
the
best
working
group,
and
this
is
done
for
both
it
for
the
independent
review,
as
well
as
future
application
applicability
for
the
other
use
cases
as
well
as
we
have
expanded
on
the
interest
option
b,
like
procedures
for
the
service
interworking
next
slide.
Please
yeah.
U
We
would
like
the
working
group
to
review
as
well
as
adopt
this
draft.
Also
one
thing
more:
there's
another
transport
interworking
draft
pin
there
in
the
spring.
We
are
working
with
the
co-authors
of
those
of
that
draft,
because
there's
a
similar
one,
similar
Behavior
being
defined
to
see
how
to
move
forward
on
that.
U
S
Got
to
put
yourself
on
the
list:
hi
a
Swedish,
very
good
presentation.
Thank
you.
One,
quick
question
about
the
PC
part.
Currently,
in
the
document
you
are
focusing
only
on
a
single
PC
case
in
peace
up
working
group.
We
have
one
more
document
called
stateful
interdomain,
where
how
the
PCS
can
cooperate
with
each
other
to
set
up
this
kind
of
path.
So
just
have
a
look
at
it.
I
hope
that
the
solution
can
work
for
inter-domain
cases
like
where
multi
multiple
PC
managing
multiple
domain
cases
as
well
sure.
V
Just
Japanese
or
Microsoft,
so
the
drop
notably
means
a
separational
part.
How
do
you
deploy
the
stuff?
How
do
you
operate
it
end
to
end
and
I
mean
amount
of
Buffs
when
you're
going
to
deploy
this?
It's
just
scary
right.
So
it
would
be
really
good
to
see
sections
that
identifies
how
do
I
trace
when
it
Traverse
different
calculation,
different
domains
when
I
replace
seats
by
labels
and
so
forth.
Right,
it's
completely
missing
in
the
draft.
So
if
you're
saying
the
operational.
U
E
You
should
put
yourself
on
using
the
tool,
but
just
state
your
name
at
the
microphone.
Please.
D
Hey
so
you
said
that
there
is
another
drafting
base
for
the
pure
bgp
procedures
for
the
for.
U
The
signaling
aspect
to
because
we
are
extending
to
carry
these
new
behaviors
in
the
bgp
in
the
bgp
transport
layer,
like
pgplu
or
in
bgp
IPv6.
So
we
have
put
a
earlier
that
was
in
this
draft,
but
to
have
a
future
applicability
as
well
as
to
a
category
Group
by
the
best
we
have
put
in
the
those
signaling
aspects
we
have
put
in
sector
draft
and
the
architecture
and
the
procedures
are
still
keptical
in
this
draft.
But.
D
Also
for
the
option
b
section,
because
at
the
moment
it's
only
covering
the
ldu
part
right,
no.
U
No
option
b
is
for
the
services
right.
We
can
do
next,
stop
self.
On
the
interworking
box,
where
our
srv6
service,
it
can
be
mapped
to
a
VPN
label
and
the
outgoing
LSP
and
same
on
the
reverse
direction
right.
So
it's
using
a
DPM
Behavior
to
achieve
it
right.
So
it's
a
next
stop
sell
for
the
services
not
for
the
transport.
U
We
can
look
into
it,
but
it's
just
because
option
b
is
already
defined
in
certain
aspects.
This
is
more
of
using
those
procedures
to
create
this,
but
the
behavior
is
a
new
Behavior
DPM.
That's
the
reason
we
have
put
it
out
so
that
everybody
knows
we
have
to
use
a
DPM
Behavior.
We
have
to
push
our
labels
that
stack
of
labels
when
I
receive
a
subjected.
That's
the
reason
we
have
put
it
out.
Otherwise,
from
control
plane,
I
still
feel
existing
procedures
should
work.
It's
just
some
data
plane
and
update
to
that.
We.
J
Q
Taking
the
left
picture
as
an
example,
there
are
two
completely
isolated,
VPN
tenants
within
one
and
vaping
2.
They
are
both
multipoint
to
multi-pond
excise
and
the
whipping
one
has
three
different
types
of
sources
within
two
has
two
different
types
of
sources:
each
sources
has
different
quality
requirements.
Q
Q
So
here
we
still
the
the
traffic
into
one
is
a
policy
group,
so
the
traffic
belongs
to
the
SR
policy
group
who
have
different
sources
and
the
different
destination
addresses
and
the
next
different
type
of
sources.
The
traffic
under
the
VPN
can
be
steered
into
different
policies,
which
is
based
on
the
parent
policy.
The
traffic
has
a
specific
source
and
destination
address.
Q
Q
Q
The
next
sentence
is
also
talking
about
the
mapping
relationship
between
the
flow
characteristics
and
the
color
value
the
same
with
the
previous
three
slides.
Q
Q
First,
it
will
match
an
Sr
policy
group
using
the
color,
extended
Community
attribute
in
the
BP
route
and
the
next
it
will
match
apparent
as
a
policy
under
the
previous
policy
group
using
the
next
hope
in
BCP
root,
and
then
it
will
match
as
a
policy
using
the
flow
characteristics
through
the
three
steps.
Finally,
the
flow
is
still
into
one
specific
as
a
policy.
Q
From
this
picture,
you
can
see
that
there
is
one
ISR
policy
group
for
pt1,
which
is
with
the
color
one,
and
it
contains
the
two
parents
as
a
policy
pp1
and
pp2,
and
they
have
the
same
color
one
with
the
policy
group
pt1,
but
different
endpoint,
U1
and
E2,
and
each
parent
as
a
policy,
contains
three
threes
corresponding
to
three
different
policies,
which
is
with
the
color
100
to
2300.
Q
Okay
next
place,
that's
all,
and
this
is
the
the
real
version
of
the
draft,
so
welcome
any
question
or
comments.
Thank
you
very
much.
B
If
you
could
go
back
to
the
slide,
Bruno
yeah,
so
one
question
Sr
policy
Group,
which
is
a
new
new
thing
that
has
been
introduced
here.
Does
this
belong
on
the
head
end,
or
is
this
something
that's
on
the
controller.
B
So
it's
at
the
head
end.
Okay,
so
I
am
not
able
to
see
what
is
the
benefit
that
this
group
brings?
It
would
be
good
if
the
draft
could
be
updated
to
clarify
what
cannot
be
done
or
achieved
with
what
we
have
today
before.
R
W
W
Also,
the
cause
of
this
draft
I
would
like
you
to
answer
the
question
so,
firstly,
I
think
the
here
the
policy
group
will
not
introduce
any
new
Hider
or
new
change
to
the
data
plan
and
the
the
second
one
we
use
that,
based
on
maybe
a
controller
or
some
other
network
management
system,
so
that
that
is
logically
module
to
organize
the
hierarchical
services.
W
So
we
just
talked
talked
about
the
hierarchical
network
slides.
So
when
we
use
the
a
services
policy
to
implement
the
hierarchical
slicing,
Zuca
solution
can
be
used
at
that
application
scenario.
Foreign.
B
Er,
so
maybe
some
clarity,
more
more
clarity
in
the
draft.
W
W
So
if
you
can
use
those
concept,
then
you
can
very
easy
to
set
up
the
hierarchical
Services
based
on
a
services
policy.
B
A
very
very
last
bit
for
taking
the
time,
but
we
have
all
kinds
of
template
and
hierarchies
in
you
know
of
all
the
vendors
config
models
right.
The
question
is
here:
why
standardize
it?
You
know
you
can
do
whatever
you
want
in
your
controller
or
ahead
and
implementation.
B
If
we
are
standardizing,
it
should
bring
some
functional
value,
and
that's
not
clear
to
me.
Thank
you.
I'll
end
here
now.
X
W
I
think
I
added
the
same
Library.
We
just
use
a
one
color
so
such
as
at
the
isopers
group.
We
just
use
the
same
color
for
one
parent
policy,
but
within
the
ASAP
parent.
As
a
policy,
the
different
Services
may
have
a
different
color
so
that
it's
similar
to
the
hierarchical
cues.
D
Y
Z
Z
They,
the
consent
of
network
resource
petition,
is
introduced
in
draft
ITF
test
iitf
Network
slices
as
this
script
in
section
6.1
of
that
chart.
Nrp
is
a
collection
of
resource
in
the
underlying
Network
NLP
consent.
Consents
of
a
subset
of
the
buffer
queuing
scheduling,
resources
on
each
of
the
connected
set
of
links
in
the
underlay
Network
and
RFC
8986
has
defined
a
set
of
well-known
srv6
and
point
behaviors,
but
those
behaviors
are
not
associated
with
the
set
of
NRP
of
the
links
for
slices
or
slice
aggregate.
Z
Z
The
uses
of
the
end
of
the
seed
this
end
touch
and
now
PC
is
mainly
used
to
steer
traffic
forwarding
through
the
associated
NRP
by
using
the
end.chana
PC
to
build
a
segment
list.
The
srv6
policy
have
the
capability
to
program,
network
resources
and
restricted
strict
test.
Sla
guarantees
for
one
HP
link,
multiple
and
Dot
Channel
BC
should
be
allocated,
Each
of
which
is
associated
with
a
subset
of
Link
resources
in
such
as
dedicated
queues,
layer,
2,
logical
sub
interfaces
and
so
on.
Z
Also,
two
and
Dot
NRP
seat
s
are
allocated
to
those
two
nrps.
Z
Z
Z
Z
Any
state
instance
of
earned.nrp
behavior
is
associated
with
two
sets.
J1
and
J2.
G1
is
one
or
more
Elson.
Adjacencies
G2
is
NRP
of
J1
when
and
receives
a
packet.
That's
that's
changed
to
S
and
the
S
is
a
local
and
dotted
nappy
seed.
The
line
is
15
of
the
enter
text.
Processing
defined
in
FC
8986
is
replaced
by
the
following
by
the
following,
submits
the
package
to
the
IPv6
module
for
transmission
to
the
new
destination,
where
a
member
of
J1
using
the
NLP
identified
by
J2.
Z
This
is
a
use
case
of
the
end
of
the
nappy
behavior
in
the
srv6
list
line
celero.
In
this
use
case,
there
are
two
customers
with
different
list
line
requirements
from
P1
to
P2.
Z
Z
Secondly,
the
controller
cooperates
with
the
network
nodes
to
complete
NRP
reservation
and
the
end.net
PC
allocation
along
the
SRV
6
policy
paths
also
now
take
note
P1
as
an
example.
The
two
different
tender
piece,
such
as
two
dedicated
cues,
are
partitioned
from
the
network.
Resources
of
the
physical
link,
X1
and
corresponding
and
Dot
nappy
seat
is
also
assigned
to
each
of
those
two
and
a
piece.
Z
The
results,
as
shown
in
the
table,
one,
the
seat,
a
column,
one
column,
column
11,
is
associated
to
an
RPG
one,
zero,
zero,
Dot
q1
and
the
seat.
The
column
column
a
column,
one
color
22
is
associated
with
lrpg
one
zero,
zero
dot.
Q2
next
slide.
Please
please.
Z
Finally,
the
controller
uses
the
corresponding
and
Totten
rpsc's
on
link
X1,
X2
and
X3
to
build
the
seed
list
of
srv6
policy
paths,
as
shown
in
the
picture
too.
The
pass
for
SRV
6
policy
1
is
a
colon
1,
colon,
colon,
11.
Z
Next
slide,
please:
okay,
in
this
use
case,
the
traffic
from
customer
one
will
be
forwarded
to
P2,
while
the
NRP
previously
resolved
for
the
per
hoplink
on
the
srv6
policy.
One
pass
and
the
the
pie
providing
customer
one
with
and
and
head
to
end
1gbenchwise
resource
for
list
line
services
with
strict
SLA
guarantee.
O
Obviously,
I
was
wondering:
how
do
you,
how
are
you
napping
the
it
seems
like
there's
a
mapping
between
the
NRP
I
guess
when
the
when
the
controller
is
provisioning,
the
flow
I
guess
to
qos,
because
you
mentioned
queuing
so.
O
Mapping
that's
happening
as
well
as
if
you
have,
if
you
have
that,
let's
say
on
the
box,
you
probably
have
your
standard
queuing
set
up,
but
now
there's
a
another
layer
of
queuing.
That's
set
up
with
the
with
this
programming
and
I
just
how
that
works,
because
Ian
are
working
today.
Z
Okay,
that
I
I
try
to
answer
your
question
so.
W
I
would
like
to
answer
the
question.
I
I
think
it,
as
you
just
mentioned,
the
NRP,
maybe
existing
in
the
controller
for
the
provisioning,
as
well
as
how
to
attach
the
NRP
to
the
forwarding
leader
is
a
question,
and
it's
very
hard
for
controller
to
manage
the
dedicated
resource
in
the
forwarding
leader
so
combine
the
NRP
in
controller
modeling
and
the
the
and
the
NRP
Behavior.
W
It
will
to
set
up
some
good
solution.
Very
simple.
You
can
use
such
as
a
service,
6
policy
and
as
well.
The
dedicated
resources
in
the
forwarding
leader
can
be
attached.
It
will
help
operators
to
support
the
high
quality
list.
The
line,
something
like
that.
G
W
Y
Y
Yes,
we
proposed
this
draft
to
define
a
new
segment.
We
call
it
endpoint
and
VPN
to
contain
the
VPN
custom
informations,
and
this
segment
can
be
used
to
distinguish
traffic
from
different
customers
and
it
can
unleash
this
segment
of
the
VPN
multi-group
multicast
Group
address
allocation
can
align
with
the
unicast
ipvpn
address
allocation
and
this
the
the
implementation
of
the
solution
is
similar
to
realtor
and
VPN,
and
it
can
also
come
compatible,
always
a
beer
and
other
Technologies
and
next
slide.
Please.
Y
Yes,
one
a
customer
wants
to
join
the
multicast
group.
The
ecosp
is
connected
to
will
generate
an
endpoint
and
VPN
according
to
the
VPN
custom
information
and
send
the
segment
to
Ingress
PE
and
increase
P.
It
will
maintain
a
mapping
table
of
endpoint
and
VPN,
and
the
European
custom
informations
and
the
generated
endpoint
and
VPN
will
be
inserted
into
the
table
and
when
CE
sends
a
multicast
packet
to
Ingress
P,
the
Ingress
P
will
encapsulate
the
package
with
a
header.
Y
According
to
the
endpoint
NYPD
mapping
table
and
the
destination
address
will
be
set
to
the
endpoint
I'm
VPN.
But
then
the
packet
will
be
sent
to
the
the
responding
request
piece
and
the
behavior
on
P
routers
depends
on
the
generation
method
of
the
multicast
tree
and
finally,
the
egress
P
will
decapsulate
the
package
and
transmit
it
to
the
corresponding
C
is
a
Nestle
at
least
oh.
Yes,
this
is
the
encoding
of
endpoint
mopn
and
it
will
come
first
to
the
IP
wave,
6
multicast
address
and
the
way
defined
an
away
bit
over.
Y
It
defines
a
highest
speed
in
ff1
as
the
way
bit,
which
means
the
VPN
multicast
bit
and
when
the
speed
is
set,
it
means
the
multicast
group
address
is
and
a
VPN
segment.
Otherwise
it
is
a
common
multicast
address
and
the
network
network
prefix
field
will
carry
the
customer's
Rd
value
when
the
web
it
is
Set
addressed.
Y
When
an
Ingress
P
generates
the
endpoint
and
VPN
segment,
it
will
firstly
extract
Rd
and
multicast
Group
address
of
the
customer
and
set
the
value
of
vb21
and
set
the
value
of
a
net
Network
prefix
to
the
customer's
Rd
and
attend
the
Ingress.
P
will
set
the
value
of
group
ID
field
according
to
the
customers
multicast
Group
address.
If
the
address
is
iPhone,
6
ipv4,
the
the
value
of
group
ID,
will
be
set
to
the
ipu
for
a
multicast
group
address
and
if
the
customers
multi-group
address
is
an
ipoa
6
address.
Y
Multicast
Group
address
a
NASA,
please,
oh,
yes,
this
is
a
the
behavior
on
the
increase
P
when
it
received
an
happy
way.
6
package
waste,
the
destination
address,
D
and
the
D
is
associated
with
an
endpoint
of
mvpn
segment
on
the
egress
PE.
So
the
englishp
will
check
the
way
bit
in
endpoint
and
VPN.
If
we
beat
is
equal
to
one,
it
will
look
at
the
endpoint
MP
mopn
mapping
table
according
to
the
and
and
point
and
VPN
value
and
find
out
the
associated
Rd
and
the
related
unfield
table
T.
Y
Then
it
will
remove
the
outer
ipa6
header
with
all
its
extension
headers
and
set
the
packets
Associated
and
flip
table
2T.
Finally,
it
will
submit
the
package
to
the
egress
MVP
lookup
for
transmission
to
the
new
multicast
Downstream,
and
if
the
wave
a
bit
is
equal
to
zero,
the
Ingress
P
will
set
the
is
associated
I've
made
table
to
Global
fmib
and
submit
the
package
to
the
egress
MF
me
lookup
for
transmission
to
the
new
multicast
Downstream
and
next
slide.
Y
Please,
oh
yes,
and
since,
since
this
is
the
first
time
for
us
to
give
the
presentation
of
this
draft
Subway,
we
hope
we
can
get
some
comments
and
or
suggestions
about
it,
and
we
will
reverse
this
draft
accordingly.
Thank
you.
Y
Yes,
maybe
we
will
presentation
it
in
next
ITF
meeting.
Thank
you.
J
Hi
I'm
Sasha
Weinstein
from
riba
I,
had
looked
at
the
draft
before
the
president
sure
before
the
presentation
and
send
an
email
to
the
authors
and
the
work
group.
It
seems
that
the
solution
proposed
in
the
draft
only
can
work
with
Ingress
replication
as
the
provider
tunneling
technology
for
mvpn,
because
they
provide
the
structure
of
the
end
dot.
Nvpn
Sid
includes
the
route
distinguisher
of
the
multicast,
where,
if
in
the
airspe
rail
distinguishers
in
menu,
if
not
most
cases
are
specific
to
each
address
P.
J
For
instance,
if
you
use
type
1
route
distinguishers,
they
would
include
the
something
like
the
router
idea
of
the
recipe
is
part
of
it
and
therefore,
in
order
to
work,
this
proposal
would
require
the
Ingress
fee
to
send
a
dedicated
packet
to
each
address
fee,
where
the
specific
mvpn
service
is
instantiated.
J
AA
Yeah
I
I
I
think
your
actually
for
for
a
podcast
scenario,
the
Rd
of
an
artist
value
for
all
of
the
Eagles
okay.
AA
Then
the
you
know
our
job
just
described
for
for
matter
of
the
engineer,
VPN
for
the
multicast
services
and
my
customers,
the
marketplace
audition
must
be
the
same
for
the
market.
Services,
okay,.
J
The
text
is
definitely
not
what
the
normal
mvpn
Behavior
suggests
and
in
any
case,
if
you
started
from
the
unicast
a
VPN
service
that
you
wish
to
upgrade
to
multicast
VPN,
you
don't
change
the
route
around
around
distinguishers.
But
in
any
case,
if
this
is
the
Assumption
of
the
draft,
it
should
be
stated
explicitly
because
I
maybe
I
have
missed
it.
But
I
have
not
seen
anything
like
that
in
the
text
normally
in
layer
3
in
and
that
that's.
Why?
J
I
think
that
if
you
present
is
drafted
at
the
best
worker,
it
definitely
should
be
presented
because
it
includes
the
bgp
parts.
This
the
same
question
would
be
raised.
AA
O
I
sure
I
did
just
a
quick
question.
Does
this
draft?
Does
it
follow
like
the
mvpn
procedures,
RFC,
6513
and.
G
6514
using
like
Leaf
ad
to
discover
the
endpoints.
O
Like
the
auto
Discovery
piece-
or
this
is
completely
separate
from
any
nvpn
procedures,
also
I
I
do
agree
that
this
seems
like
it's.
G
More
in
line
with
like
Ingress
replication
and
not
and
not
any
other
ptas.
Thank
you.
AA
Hey
John,
I
think
the
kind
of
solution
for
the
foreign.