►
From YouTube: IETF103-PCE-20181105-1350
Description
PCE meeting session at IETF103
2018/11/05 1350
https://datatracker.ietf.org/meeting/103/proceedings/
A
Welcome
to
PCE
that's
around
about
10
to
2,
so
we
should
start
here,
I
just
passed
around
the
two
blue
sheets
as
they're
coming
around
the
room.
Please
key
to
make
sure
you
put
your
name
on
there
so,
but
we
have
a
record
of
who
attended
the
meeting.
I
will
start
with
the
ITF
note.
Well,
this
is
something
that
most
people
will
have
seen
before.
It's
really
just
a
reminder
of
the
ITF
policies
which
govern
things
like
intellectual
property
rights.
Please
be
aware.
A
A
Okay,
so
we'll
go
through
the
usual
administrivia
first,
please
sign
the
blue
sheets.
As
I
said
we
have.
Our
usual
way
of
taking
notes
is
to
use
the
ether
pad
on
the
you'll,
find
the
link
to
the
ether
pad
from
the
IETF
meeting
agenda
site.
Who
would
like
to
help
me
with
the
notes
today,
Dhruv?
Is
there
anyone
else,
Stan
yeah
Harman?
Thank
you.
Okay,
that's
that's
fine!
You
know
it
should
be
enough
of
this
anyone
else
who,
particularly
if
you
come
to
the
mic,
can
you
make
a
comment
you
might
want
to
go
check.
A
The
notes
make
sure
we
recorded
your
comment
correctly
and
your
name
correctly.
That
would
be
helpful.
Thank
you
and
a
job
ascribe
I.
Don't
know
if
we
have
anyone
in
the
jabber
room,
but
it
would
be
helpful
if
someone
is
in
there
down
is
in
there
thanks
very
much
down.
If
you
wouldn't
mind
keeping
an
eye
on
the
jabber
room.
A
Okay,
we
are
streaming
the
audio
and
video
of
this
meeting
on
to
meet
echo
I
see
we
do
have
some
participants
on
meet
echo,
including
some
people
who
are
not
in
this
room.
So
please,
when
you
go
to
the
mic,
please
only
speak
in
the
meeting
at
the
mic.
Please
state
your
name
while
you're
at
the
mic
and
there's
actually
not
a
pink
box.
A
Should
have
said
I'm
John
Hardwick
for
those
who
don't
know
me,
my
co-chairs
Julian
couldn't
travel
to
Bangkok
this
time.
So
sorry
about
that
it
would
be
a
good
idea.
Maybe
he
will
be
be
joining
us
I
think
it's
still
fairly
early
on
in
the
day
and
in
France,
but
then
maybe
I'll
be
joining
us
later
on
meet
echo.
A
Okay,
so
a
little
announcement
about
working
group
personnel.
This
will
be
my
last
meeting
as
PCE
chair
I'm
stepping
down.
Basically
the
reasons
for
that
are
because
I
need
to
free
up
more
cycles
of
my
time
to
do
other
things,
but
I
very
much
appreciate
the
time
of
being
able
to
spend
hours.
Chair
of
a
working
group
really
enjoyable
working
with
you,
guys
and
I
think
you
know,
we've
done
some
good
things
in
PCE
and
I'm.
A
B
A
Now,
on
to
the
traditional
admonishment
to
use
the
mailing
list,
please
actually
Lou
Berger
did
a
really
good
job
of
this
and
T's
I'll
see
if
I
can
do
remotely
the
same
job
here,
particularly
where
we
have
a
working
group
draft.
The
draft
is
owned
by
the
working
group,
not
by
the
individual
authors.
So
it's
really
helpful
when
making
changes
to
your
documents,
particularly
when
they're
a
working
group
draft.
A
So
if
you
could
please
using
the
mailing
list
both
to
discuss
changes,
you're
making
and
also
to
raise
comments,
it's
really
unhelpful
when
things
happen
off
the
list,
because
then
we
as
chairs
don't
get
to
see
what's
going
on
the
other
participants,
don't
get
to
see
what's
going
on
and
it
kind
of
reduces
the
amount
of
collaboration
that
we
can
do.
So
please
do
the
work
on
the
list
that
is
not
aimed
at
any
one
particular
thing
or
anything,
but
it's
just
the
general
trend
and
I
have
to
keep
reminding
people.
A
Today
you
will
listen
to
me,
hopefully
for
less
than
twenty
minutes,
and
then
we
will
speak
of
several
topics.
The
first
one
segment,
routing
there's
a
lot
of
work
in
the
pipeline,
we're
only
just
beginning,
it
seems
of
segment
routing.
So
we
have
quite
a
few
drafts
to
be
discussed
today.
Chang
lead,
Ruth,
Jack
Jeff
will
be
discussing
those
next,
there
will
be
new
optical
work.
Jung
is
going
to
discuss
piece
of
the
flex
grid.
A
Then
we
would
go
back
to
Dhruv
who's
going
to
talk
to
us
about
pce
as
a
central
controller.
Now
we
have
just
recently
adopted
the
base
protocol
extension
stripe
about
we'll
be
talking
about
further
extensions
for
p2
MP
and
using
SR
b6
and
then
finally,
Kwon
ji-yong
will
told
us
about
the
multi-layer
Association
draft.
Would
anyone
like
to
bash
the
agenda
suggest
any
changes
or
making
the
objections
going
going
gone?
Okay.
Thank
you
very
much.
So
that's
what
we
will
go
with
and
so
working
group
status.
A
So
we'll
start
with
things
which
are
not
working
group
dress
anymore,
so
one
urs
see
since
Montreal
as
RFC
80
408.
This
is
the
path
setup
type.
Rfc
is
really
good
to
get
that
published
that
generalizes
PCE,
so
that
you
can,
you
know,
rather
than
specializing,
to
RSVP
setup
types.
Now
we
have
a
general
mechanism
for
saying
what
kind
of
plaque
setup
type
and
this
was
really
needed
in
order
to
do
all
the
segment
routing
work
which
is
now
following
on,
so
it
was
really
kind
of
foundational
work.
A
So
it's
great
to
get
that
out
of
the
way
and
published
we
have
free
dress
right
now
with
the
IES
g2
of
which
the
gmpls
extensions
and
the
segment
routing
base
draft
both
been
through
IETF
last
call.
I
think
that
we
have
Directorate.
Reviews
opposes
well
Debra
or
I.
Think
we're
not
waiting
for
anything,
so
maybe
they're
they're
ready
for
the
next
step.
A
I
think
then
we
also
have
the
ws
on
draft,
which
kind
of
follows
on
from
the
gmpls
draft,
where
we
haven't
done
a
nice
yet
last
call
about
yet
we
we
did
a
Directorate
review
I.
Think
last
few
days,
young
you've
updated
the
draft
yeah.
So
that's
that's
addressed
the
director
of
comments,
so
I
think
now
the
ITF
has
calls
probably
ready
to
go
okay,
so
those
are
moving
forward,
which
is
good.
A
We
have
early
code
point
allocations,
I'm
just
using
this
moment
to
remind
people
that
we
have
allocated
early
code
points
for
their
drafts
and
that
those
early
code
points
don't
last
forever.
So
the
PC
segment
routing
draft
most
code
points
expire
in
August,
hopefully
gets
published
before
then,
because
you
know
we're
nearly
there,
but
then
the
other
two
stateful
PC
P,
2
MP
and
the
Association
group
drafts
I,
know
they're
kind
of
or
will
come
to
them
in
a
moment.
But
we
have
code
points
for
those
we'd
lose
him
in
October.
A
A
A
A
A
But
I'm
really
crying
on
the
inside
the
the
other
document
I
did
post
this
morning
was
the
the
stateless,
hierarchical
PC
extensions
document,
so
that
that
has
been
posted
have
sent
a
update
to
the
list,
just
summarizing
what
the
changes
are
and
if
Adrian
Powell's
in
the
room
I'll.
Let
him
know
that
I
will
email
him
on
the
list
and
respond
to
each
of
his
comments
that
he
made
when
he
did
his
review
thanks,
Dan.
A
Okay,
so
I
should
remind
everyone.
The
the
inter
is
update,
Dan's,
referring
to
us
in
response
to
my
Shepard
comments
from
a
while
ago
and
I
think
when
it
gets
published
and
I
probably
need
to
take
another
scan
of
that,
and
hopefully
we
move
it
forward
hierarchy
extensions.
The
updates
were
from
Adrian's
comments
at
last.
Call
I.
Think
the
next
step
for
that
is
that
the
document
Shepard
needs
to
get
hold
a
bit
and
do
the
write-up.
A
Staple
PC,
a
point-to-multipoint,
that's
sitting,
basically
with
me
waiting
for
me
to
do
the
shepherding
I
have
to
review
that
one
last
time
and
then
we
can
push
it
forward.
Apologies
for
not
getting
to
bat
until
now,
but
I'm
still
intending
to
do
it.
Maybe
I'll
do
it
later
this
week
and
then
lastly,
the
Association
group
draft
has
been
free
working
group
last
call
there
were
some
comments,
drew
view
of
update
of
the
draft
and
it's
basically
now
waiting
to
be
shepherded,
so
push
it
forward.
A
A
Staple
PCE,
Auto
bandwidth
number
one:
a
CT
and
applicability
number
two
piece
up:
stateful
extensions
with
GCM
PLA
gmpls,
the
free
and
peace
of
stateful
extensions
for
hierarchy,
where
I'd
really
kind
of
like
to
get
the
higher
but
base
hierarchy.
Extensions
out
of
a
working
group
on
with
the
isg
before
we
do
working
group
last
call
on
that,
but
I
feel
like
they're,
pretty
close
to
that
now.
So
I
think
that's
good
to
go
see.
A
A
D
Adrian
Farrell
just
to
talk
about
the
flow
spec
document.
We
think
that
we've
pretty
much
got
everything
we
need
to
have
in
there.
It
needs
a
bit
of
a
bash
on
there
on
the
language
and
I
really
want
anybody,
who's,
implementing
or
looking
at
implementing
to
say
something
soon,
because
otherwise
it's
going
to
move
forward
relatively
soon.
I
expect
the
chairs
together
hope
from
us.
Probably
after
the
next
revision
saying
we
think
we're
done.
E
From
Holly,
this
is
the
to
update
to
the
draft
elves
enhanced
era,
and
we
have
updated
this
draft
mainly
to
align
with
other
part
of
the
extension
and
the
more
important
thing
is.
We
also
propose
some
future
requirement
for
Averell
handling
sections
for
the
future,
PC
extension,
so
we
assumed
this
worker
is
now
ready
and
we
can
move
forward
and
you
are,
you
are
all
invited
to
review
it
to
make
sure
that
it's
done
thanks.
A
But
that's
a
really
important
point.
Actually,
if
this
draft
is
proposing
putting
requirements
on
future
drafts
for
error
handling
sections,
then
that's
probably
and
everyone's
interest
to
take
a
look
at
a
review
quite
closely
because
it's
gonna
come
back
to
bite
you
when
you
come
to
write
your
next
draft.
F
Hi
spending
on
me
as
well
I'm
supposed
to
review
oh
by
the
way.
This
is
truth
from
huawei.
Apart
from
that,
let
me
talk
about
Association
diversity
and
policy.
We
have
not
made
an
update,
I
think
the
documents
is
stable
is
we
are
just
waiting
for
the
Association
group
to
move,
and
then
we
can
move
these
as
well.
There
are
no
pending
open
issues
there.
With
the
end
we
did
make
an
update.
F
We
sent
for
the
early
doctor
review
I
had
a
chat
with
Mahesh
who,
who
is
who's
in
midst
of
preparing
his
answer,
but
I
got
the
gist
of
what
really?
What
he's
thinking?
Okay
and
I'm
continue
to
chat
with
him,
and
then
Tom
also
gave
some
comments
on
the
list.
So
once
I
have
the
yang
doctor
review
will
do
a
reprint
and
LSP
shed
Yuling
no
update
done,
but
more
reviews
would
be
then
I
useful.
Specially
for
that
document,
yeah
I
did
for
LSP
control
request.
F
No
update
documents
is
pretty
straightforward,
very
simple
extension
and
I
think
it
it's
in
my
mind.
It's
kind
of
ready
for
last
call
as
well,
because
it's
a
very
simple
extension
protection.
I
think
we
did
make
an
update,
but
that
was
just
because
of
the
hurries
affiliation
change.
It's
also
basically
waiting
for
Association
group
to
move
and
I
think
I'm.
A
Okay,
I
for
I'd
share
with
you
this
slide,
which
is
the
current
queue
of
drafts
sitting
in
the
adoption
holding
queue.
We
got
a
little
bit
behind
on
adopting
drafts
and
or
there
was
so
much
so
many
drafts
coming
forward
to
be
adopted.
That
is
generated
a
bit
of
a
backlog,
so
we've
been
working
on
clearing
that
we
adopted
a
few
of
the
last
period.
A
At
the
moment
we
have
these
four
still
sitting
in
the
queue
and
I
think
the
queue
is
in
this
order.
This.
Basically
the
order
have
they
requested.
There
was
the
PC
ridicule
bandwidth
draft,
which
I
assume
is
I,
haven't
heard
about
from
the
office
on
that
for
a
while,
but
I
assume
is
still
alive
and
wanting
to
be
adopted.
Apologize
for
the
delay
authors
we
will.
We
will
get
to
you
very
very
soon.
A
A
G
Online,
did
he
finish?
Yes,
absolutely
he
goes
come
into.
Us
VN
Association
can
be
also
applied
for
Turner,
said
Turner
said
binding
or
I
forgot
that
words,
so
we're
gonna
add
text
that
VN
Association
each
VN
member
response
to
tunnel
when
you
find
associated
with
VN
that,
in
effect,
satisfies
egos
desire
to
have
Turner
set,
bundling
yeah.
A
A
A
I
said
hands
up
if
you
like
it
and
I
think
about
12
people
put
their
hands
up
and
then
I
did
a
poll
on
the
mailing
list
and
I
think
three
people
answered
so,
but
that
was
if
I
was
a
strange,
disconnect.
I,
don't
know
if
that's
because
people
weren't
looking
at
the
mailing
list
or
if
actually
there's
just
been
a
genuine
reversal
on
that
draft,
but
it
didn't
seem
that
there
was
enough
momentum
behind
it
at
that
point
to
say
yes,
so
I
was
just
thinking.
This
is
what
happened.
E
Holly
speaking
as
the
one
of
the
authors
for
this
draft
I
think
that
now
it
is
a
slightly
different
way,
the
very
beginning,
so
we
are
switching
from
defining
new
objects
to
using
the
association
object
specified
here
as
a
document
or
so
so.
This
is
a
part
of
the
Association
work,
but
the
only
the
name
is
society
anyway.
It's
still,
we
would
like
to
steal
you
white
people
to
review
and
and
provide
your
opinion
on
this.
Okay.
A
A
C
So
the
motivation
of
why
we
are
doing
this
is
like
in
some
use
cases
like
intend,
pass
protection
and
bi-directional
pass
correlation
or
performance
measurement
if
they
require
the
ability
to
implement
a
segment
identification
in
essa
networks.
So
we
propose
that
to
draft
the
first
one
is
draft
showing
spring
amperes
per
segment.
You
define
a
new
path
segments,
a
new
segment
code
per
segment
to
identify
in
an
esse
pass
in,
as
I
am
Harris
Network.
C
Similarly,
we
proposed
a
new
draft
coded
draft
really
spring
as
IV
six-part
segment,
identifier,
Inessa
of
a
six-part
in
SM
v6
network,
so
for
configurating
all
like
allocating
pass
segments
in
as
a
network.
We
need
the
extension
in
P
step,
so
we
propose
the
draft
of
how
to
allocate
the
past
segment
and
as
well.
We
also
propose
and
draft
of
how
to
allocate
how
to
combine
to
as
a
pass
into
and
bi-directional
pus
in
a
sub
Network.
C
This
draft,
who
opposed
have
been,
have
been
posted
in
ITF
102
and
today,
I'm
gonna,
I
work.
This
I
will
introduce
the
updated
of
this
draft.
First
of
all,
we
we
update
the
droplet
c-segment
and
right
now
it
has
been
the
it
has
been
the
version
0
3
&
4,
the
draft
in
pcs
by
Russia
my
past.
It
has
been
the
version
2.
C
C
We
used
as
a
we
used
the
past
segment
instead
of
pass
ID
carry
in
the
site
and
also
we
deleted
the
ingress
allocation
mechanisms
and
to
label
solutions,
because
when
we
deploying
or
implement
implementing
the
per
segment,
the
customers
and
vendors,
we
have
make
a
decision
to
choose
the
eager
egress
allocation.
It's
the
only
way
to
allocate
the
power
segment
and
you
can
see
this
update
in
channel
spring
and
pairs
per
second
and
draft
also
as
well.
C
We
we
delete
the
two
labels,
the
solution,
new
things
we
add
new
orders
and
contributors
from
CMC,
see
China
Mobile,
Chang,
Chang
and
Rakesh
Dafa
from
Cisco.
Also,
we
add
the
part
of
a
Ana
consideration
and
data
plan,
consideration
and
error
handling.
So
for
implementation.
How
is
implementing
this
solution
in
PC
and
PC
products?
I
suppose
that
some
vendors
doing
the
same
thing
as
well.
C
So,
let's
go
to
the
second
part,
the
the
PCP
extension
for
bi-directional
Association
associated
policies,
so
major
updates
are
we
also
update
the
post
ID
to
per
segment
in
a
4
s,
a
B
6,
and
we
delete
the
stateless
PC
port
and
new
things
are
new
others
and
contributors
from
China
Mobile
and
Cisco,
and
IANA
consideration,
secure
security
considerations
and
error
handling
at
at
it.
So
for
implementation
we
we
also
were
also
implementing
this
solution
in
our
products.
C
So,
since
the
draft
the
contacts
of
the
draft
are
stable
and
commercial
implementation
is
going
on,
and
this
solution
and
mechanism
are
supported
by
several
vendors
and
operators,
so
we
decided
to
post
a
working
group
adoption
request
for
this
draft
Inc.
In
order
to
we,
we
can
request
for
earlier
I
in
a
allocation.
So
we
we
post
working
group
adoption
request
for
draft
Lee.
Pcs
are
segments
in
PCs
are
bi-directional
pass
and
your
Commons
and
contribution
are
very
welcome.
Thank
you.
A
A
A
I
think
the
process
I
propose
we
follow
here
is
that
I
don't
think
we
can
really
move
in
PCE
until
we
see
in
the
spring
working
group
is
adults
of
the
path
segment
draft
I
think
that
we
need
this
bring
working
group
to
say
yes,
pass
segments
are
a
thing
and
that's
how
we're
going
to
solve
these
requirements
once
we
get
to
that
point,
then,
yes,
we,
we
discuss
doing
an
adoption
poll
for
for
this
work.
Yes,
so
I
don't
see
any
impediments
as
far
as
the
early
iron
allocation
request.
I
A
C
A
F
Okay,
hi
everyone
so
I'll
be
talking
about
SR
e6,
the
first
draft
that's
being
presented
for
the
second
time.
We
had
the
early
discussion
of
this
couple
of
five
years
ago,
and
then
we
have
the
young
extension
for
sr
v6
as
well,
which
is
a
zero
zero
document.
So,
let's
start
with
just
a
quick
introduction
where
we
are
coming
from,
you
guys
have
seen
the
SR
in
piece.
F
is
already
gone
to
the
SG,
so
we
had
the
discussion
that
can
be
used
piece
F,
also
to
compute
s
our
v6
path.
F
It
kind
of
made
sense
that,
yes,
we
could
use
the
same
protocol
to
do
both
there's
nothing
that
stops
us
at
the
end.
We
have
a
good
PST
as
well
part
setup
type
or
using
that
we
can
do
both
SR
MPLS
and
SR
v6
Park
imputation,
and
we
be
kind
of
thought
that
whether
piece
F
is
the
right
way
to
do
it,
and
if
you
do
it,
what's
the
what's
the,
what
are
the
extension
would
look
like.
So
basically
what
is
happening
in
ipv6
phase?
F
You
have
the
path
segment,
which
is
an
ipv6
address,
so
we
need
to
have
a
good
way
to
store
the
list
of
ipv6
address,
which
are
sr
e
ro
can
do,
and
once
you
have
that
path,
the
ingress
has
to
convert
that
into
sr
h,
which
is
the
ipv6
extension
header
for
segment
routing.
So
some
conversion
of
s
ID
list
into
the
sr
h,
is
something
that
is
needed.
So
from
the
protocol
point
of
view,
it
kind
of
fits
quite
well.
You
have
an
asari
ro.
F
We
need
to
extend
it
so
that
we
can
do
s
our
v6
information.
There
same
holds
good
for
sr
r
ro
or
we
need
a
good
capability
mechanism,
because
this
is
something
new.
So
how
can
we
advertise
that?
We
support
sr
v6
in
pisa
and
the
new
parts
are
hot
setup
type
4
sr
v6.
So
that's,
basically
the
gist
of
what
this
document
has.
Just
a
recap,
I
think
everybody
kind
of
knows
this.
F
You
have
an
asana,
cheddar
or
included
in
the
in
this,
which
is
each
Sid
list
is
made
up
of
this
120
128
bits,
ipv6
address
with
location
and
function,
and
you
have
a
list
of
segment
lists
that
we
need
to
put
in
the
SRH.
So
basically
in
Pisa
protocol
we
need
to
have
a
way
to
carry
the
set
list
and
that
CID
list
needs
to
be
converted
into
SRH
at
the
ingress.
So
from
the
update
points
of
view,
not
big
update,
what
we
did
was
there
were
some
changes
in
the
PC
segment
routing
base
draft.
F
So
we
need
to
basically
align
this
work
to
that.
So
we
have
done
that.
We
have
SR
PC
capability,
so
here
we
have
an
SR
v6
PC
capability
for
to
carry
SR
v6
specific
information,
especially
what
is
the
maximum
segment
length
you
can
carry
and
whether
you
support
conversion
etcetera
and
then
we
have
a
new
PST,
a
new
part
setup
type
for
SR
v6,
and
what
you
see
in
the
bottom
is
an
extension
of
SR
ero
sub-object,
that's
defined
in
the
segment
routing
draft.
F
Now,
if
you
remember
there,
we
had
something
called
as
si
si
D
and
Nai.
So
we
kind
of
did
extension
that,
since
said
list
for
us
is
just
32-bit
our
s,
our
v6
thing
will
not
fit
into
it.
So
we
had
to
put
this
in
something
like
what
you
see
below,
which
is
a
SR
v6
identifier,
so
in
case
of
SR
v6
we
will
not
be
using
Nai.
We
will
be
using
something
called
as
an
SR
v6
identifier,
there's
something
that
you
see
at
the
bottom
and
all
the
information
related
to
SR.
F
V6
is
kind
of
put
there
and
we
don't
use
the
S
ID
32
bits
that
was
there
in
the
in
the
existing
SR
ero
sub-object.
Good
thing
is
that
that's
of
object
was
on
anyway,
allowing
that
s
ID
is
optional
in
some
cases,
so
it
kind
of
fits
in
you
just
don't
have
to
encode
s.
Id
just
include
the
SR
v6
identifier
and
put
all
the
information
right
there
and
the
rest
of
the
bits
that
you
have.
F
You
can
simply
reuse,
and
if
you
read
the
document,
you
will
see
that
it
kind
of
fits
the
extension
of
a
pc
segment
routing.
We
don't
have
to
make
too
the
changes
and
we
could
reuse
a
lot
of
things
which
were
already
there.
One
question
that
came
came
up,
I
think
I
have
that
in
the
next
slide.
I'll
discuss
it
at
the
end,
then
so
so
this
is
basically
what
we
updated
in
the
our
basic
SR
v6
document
for
stateful
PC,
a
simple
extension
for
supporting
yank.
F
We
definitely
don't
want
to
add
this
in
our
PC
yang
document.
We
want
to
move
that,
so
we
have
done
a
small
augmentation
and
making
sure
that
capability
is
handled
max
SLE
is
handle,
SLE
limit
is
handled
at
the
various
places
and
in
case
we
are
maintaining
a
list
of
information.
The
segment
list
that
we
use
from
the
SR
v6
said
this
is
from
the
SR
v6
yang
module,
which
has
been
done
in
the
spring
working
group
and
that's
something
it's
a
simple
extension
for
a
v6
yang
and
we
have
written
a
0/0
document.
F
So
please
review!
So,
what's
the
question
and
next
step
for
the
working
group
last
time
when
we
discuss
they
were
support
for
this
that,
yes,
we
could
do
sr
v,
6n
p
sep.
It
fits
in
within
our
scope.
Then
the
question
came
in
the
extension
that
we
have,
that
is
extension
of
an
SR
er.
Oh,
is
that
the
right
approach
should
we
come
up
with
a
new
sub
object?
Do
we
need
to
just
extend
them
is
a
better
idea
or
something
else.
F
So
that's
something
that
I
really
want
to
get
feedback
from
the
group
and
special
implementers.
They
do
they'll
like
extending
MPLS
SR
with
a
little
bit
changes
or
they
would
they
prefer
a
different
approach
there,
and
basically
we
feel
that
this
could
be
ready
for
adoption
as
well,
so
that
the
working
group
can
look
at
it
and
work
on
it.
Keeping
the
SR
part,
this
I
think
I
had
one
question,
but
I
might
fall.
I
have
oh
I
will
discuss
this
in
though
I
have
one
more
SR
v6.
F
That's
why
I
got
confused,
which
is
about
ECCC
for
SR
v6,
where
I
have
an
open
question
with
respect
to
the
last
point
that
is
in
s
our
v6,
they
have
the
node,
the
locator
included
SRV
success.
Id
is
made
up
of
the
node
and
the
function
so
sometime.
This
node
information
is
a
part
of
s
ID,
so
continue
to
use
the
NAI
functionality
where
I
used
to
encode
that
who's,
the
owner
of
this
SR
v6n.
Since
that
information
is
already
there
in
the
SRV
six
identifiers.
F
I
F
A
A
Mean
it
was
a
throwaway
question,
but
yeah
I
don't
think,
there's
anything
particularly
stopping
us
from
doing
a
poll
for
there
something
lost.
Last
time
we
asked
there
was
plenty
of
interest
I,
don't
through
some
comments
at
you
last
time,
I
think
they've
all
been
addressed
now
in
terms
of
aligning
with
the
base
segment
routing,
so
yeah
I,
don't
think,
there's
anything
to
stop
that.
I
Hi
everyone,
it's
just
I'm
sure
I'll
be
talking
about
the
binding
seat
setup
in
PC
networks,
so
binding
City
has
been
defined
in
80
402
and
it's
used
a
stable
anchor
local
to
the
knot
itself.
When
not
looks
up
binding
seat
label,
it
yields
a
an
action
or
another
sir
pass
or
pretty
much
anything
he
can
attached
it.
It's
essentially
use
another
application,
being
optical,
being
reduced
depth
of
label
stuff.
So
there's
a
long
time
to
you
adoption
so
binding
seed
could
be
reported
true,
except
from
PC
to
PC
PC.
I
Could
you
a
species
to
allocate
a
specific
binding
seed,
so
it
worked
what
direction
and
it
defined
that
was
be
objects
very
straightforward,
he'll,
be
I
would
say
it's
a
type
and
value.
So
far.
We
have
defined
two
values.
One
of
them
is
simplest
label
or
label
stacking
in
one
case,
it's
to
any
bit
another
case
32-bit.
I
So
again,
it's
been
around
for
quite
some
time.
Unfortunately
got
stuck.
The
number
of
implementations
been
used
in
many
other
solutions
that
rely
on
it.
So
we
would
like
to
request
working
reproduction
and
there's
number
of
things.
We
have
some
ideas
about.
We
would
like
to
consult
working
group
so
today
it's
generic
could
be
used
in
bottles,
PT
and
SR
case.
There's
no
really
use
case
for
a
CP,
so
the
question
to
grow.
Do
you
think
we
should
just
say
it's
a
sir
only
be
done,
ideally,
format,
I
think
everybody's
happy
with
it
again.
I
Many
people
used
it
in
other
solutions.
So
if
you
have
any
comments
be
happy,
there
are
some
questions
about
using
index
value
again:
I'm
not
familiar
with
an
implementation
that
would
do
banging
seat
as
none
absolute
value.
However,
if
you
have
any,
please
let
us
know
and
I
think
we
have
already
agreed
to
move
DCCC
section
to
p
CC
bc,
CC
dropped
rather
than
keeping
fear
questions.
I.
A
Me
just
on
this
previous
slide.
It's
I
think
we
probably
don't
ask
this
other
question
enough.
That
is,
it
would
be
really
good
to
get
feedback
on
things,
like
is
very
use
case
for
feature
X,
and
this
draft
before
we
just
go
ahead
and
blindly
assume
there
is
and
publish
it.
It's
really
good
to
be
able
to
simplify
the
drafts
as
much
as
possible,
just
from
speaking
as
an
implementer,
rather
than
as
chair
a
guy
who
has
to
write
the
code
as
well.
A
So
yeah
I
think
it
would
be
nice
if
we
can
get
feedback
that
there
is
a
use
case,
probably
the
right
thing,
Jeff
or
Dhruv
if
you're
driving.
This
is
to
pop
the
question
to
the
mailing
list
and
try
and
flush
out
an
answer
that
way
and
if
no
one
answers
then
I
think
we
can
just
assume
this.
We
can
simplify.
A
Yeah,
as
Jeff
Sosa
says,
this
has
been
around
for
quite
a
long
time
and
it's
been
quiet
and
there's
now
come
back.
It's
ideal
for
working
group.
Adoption,
I'm
I'm
a
co-author
on
this,
so
I
can't
do
the
adoption
poll
I
would
be
the
Juliane
to
do
that.
I
am
sort
of
interested
to
just
see
a
show
of
hands
for
the
numbers
just
to
record
in
the
minutes.
A
G
G
Think
there
are
some
changes
in
PCE,
related
work
for
flux,
grid,
RFC,
769
eight
has
been
published
as
a
framework
and
an
OSPF
to
work
has
been
RFC
3066
three
and
RS
Piketty
as
well
for
flux,
grid
networks
and,
as
we
have
seen
in
the
earlier
part
of
this
session
now,
pizza
for
gmpls
and
w.zahn
also
has
had
moved
to
RFC
publication
process
to
be
technically
aw
son.
It's
last
call
close
to
last
call,
but
a
simple
as
it
passed
idea.
Last
call
so
I
think
it's
time
to
move.
G
G
Maybe
this
one
is
good
example
in
Tablas,
on
DWDM
grid
for
the
flux
grid
allows
frequency
slots
that
have
a
nominal
central
frequency
and
the
slot
ways
so
that
the
bandwidth
can
be
a
flexibly
increased,
as
opposed
to
double
idiom
channel
being
fixed
with
a
10,
gig
or
2.5
gig
of
only
100
gig.
This
allows
you
to
create
up
to
one
terabyte
and
beyond.
G
So
this
is
a
very
important
capability
in
optical
networks,
so
we
basically
added
one
object:
si
object
in
the
existing
piece
network
spectrum
assignment
object;
basically,
it
allows
for
explicit
label
control
years.he
mode
or
not,
so
that
reply
can
reflect
what
that
means
and
estrella
object.
Basically
has
frequency
slot
selection,
TLB
and
frequencies
restriction,
constraint,
TLB
and
other
optional
TLB
and
frequencies.
G
Last
election
TLD
looks
like
this:
we
have
frequency
slot
assignment,
FSA
method,
first
fail
or
random,
or
anything
else
we
can
define
here
and
frequency
slug
restriction
is
basically
what
are
the
restriction
or
allowed
wavelengths
or
spec
in
those
reflects
great
terminology,
so
that
you
can
associate
with
the
link
with
what
other
you
know,
MN
and
value
available,
so
that
tunnel
can
grab
those
channel
and
in
peace
reply
frequencies,
la
decisión,
TLD
or
it
depends
on
years
similar
or
not.
So
we
haven't
defined
a
detail
on
this
part.
G
But
I
think
our
properties
are
quite
in
cachet
in
parallel
with
the
table
Sun
draft
and
we
need
to
have
some
work
to
return,
but
I
think
we
presented
this
before
and
what
was
holding
bag
was
gentle
s
and
W
sand
draft
now
they're
moving
down
the
pipe,
so
hopefully
we
can
move
this
faster.
Actually,
this
is
more
important
than
actually
aw
son
because
of
the
marketplace
is
now
changing,
so
I
just
want
to
ask
for
working
robot
option
this
time.
A
A
F
Yeah,
so
these
are
some
new
extension
0/0
documents
for
p2
MP
and
SR
v6
I
just
wanted
to
present
them
together.
So
we
know
what
PCE
CC
is.
We
have
the
t's
Architecture
document
well
published
some
time
ago.
We
have
the
base.
Pc
CC
need
to
change.
The
draft
aim
now,
because
it's
adopted,
and
then
we
have
the
PC
CCS,
are
also
just
waiting
for
adoption
pending,
basically
and
both
were
presented
in
ITF
I-10
too,
as
well
will
be
made
the
bigger
update.
F
That's
why
we
kind
of
explain
what
are
the
big
changes
that
we
did,
especially
with
respect
to
reusing
of
existing
messages,
rather
than
defining
something
new?
So
first,
let's
talk
about
p2
MP.
So
before
we
do
P
2
MP,
let's
just
have
a
quick
recap
of
what
we
basically
do
in
the
basic
PC
CC.
Your
aim
is
to
provision
all
the
notes
you
assign
the
labels
directly
from
PCE
to
the
notes,
and
this
we
do.
Why
are
something
that
we
call
as
a
CCI,
which
is
a
central
controller
instruction.
F
We
have
an
object,
type
1,
which
is
for
MPLS
label
and
wire,
that
you
can
simply
push
the
central
controller
instruction
to
all
the
nodes
after
the
path,
computation
and
all
the
things
related
to
that.
So
what
basically
changes
in
case
of
B
2
MP,
the
only
new
functionality
that
you
need
to
worry
about
is
the
branch
node
capability
more
and
this
diagram.
If
you
see
at
the
node
n
3,
you
have
incoming
label
B,
but
you
need
to
replicate
and
send
with
label
C
towards
n
4
and
label
D
towards
n
5.
F
So
this
is
the
normal
branch
and
replication
capability
that
P
2
MP
requires.
So
all
we
need
to
changes
may
be
program.
The
branch
node
with
this
thing,
pretty
straightforward,
good
thing
that
our
message,
which
we
had
in
the
base
document,
already
had
something
called
as
a
CCI
list.
So
using
that
list
we
could
have
multiple
brand
or
outgoing
branch
notes
and
one
incoming
notes
so
no
need
to
change
too
much
in
the
protocol.
Just
this
one
extra
functionality
will
take
care
of
our
basic
P
to
MP
in
case
of
cccc.
F
So
what
are
the
changes?
In
the
protocol
we
had
to
do
one
bit
that
says
that
you
support
P
to
MP
reuse.
The
same
CCI
object,
no
need
to
change
that,
and
then
it
is
just
some
error
codes
with
respect
to
make
sure
that
there
is
only
one
incoming.
You
can
have
multiple
outgoing
all
that
stuff
but
simple
stuff.
So
very
simple
extension,
just
the
branch
node
capability
in
the
existing
PC
CC
takes
care
of
our
multicast
use
case
rest
of
the
functionality,
like
all
the
other
use
cases
that
you
have
make.
F
How
do
you
take
care
of
everything
else,
like
a
synchronization
request?
Initiate
nothing
changes
all
remains
the
same.
This
is
one
new
functionality
that
you
need
to
take
care
for
p2
empty
rest
of
the
things
whether
it
is
a
p2p,
LSP
or
p2.
Mpls
fee
doesn't
change.
Stateful,
PC
and
PC
CC
takes
care
of
all
of
all
of
that.
So
this
is
about
p2
MP.
F
So
the
extension
here
was
that
we
had
a
new
type
for
CCI
or
new
object
type,
which
was
used
for
MPLS
SR,
and
then
you
could
use
that
to
do
node,
prefix
adjacency,
all
that
stuff.
So
what
changed
in
case
of
PCC
csr
v6
is
just
that.
Instead
of
that
s,
ID
being
an
MPLS
label
or
index,
it
could
now
be
an
SR
v6
segment
and
all
the
rest
of
the
things
are
fine.
F
So
what
is
another
new
thing
that
we
need
to
worry
that
once
we
get
this
instruction?
What
is
the
job
that
the
node
needs
to
do,
which
is
described
in
the
last
three
two
bullets
which
is
on
receiving
SRV,
six
node
or
prophet
society?
We
need
to
update
the
my
local
s,
ID
table,
which
is
described
in
the
SR
v6
document
and
also
do
some
local
next
hop,
computation,
etc.
All
what
is
as
per
the
SR
v6
document,
we
said
just
need
to
refer
those
and
same
whole
boots
for
the
adjacency
as
well.
F
So
how
does
the
CCI
object?
Look?
This
is
the
new
object
type.
The
first
two
things
are
very
similar
to
what
we
had
for
normal
SR.
You
have
an
ID,
you
have
an
empty
I,
D,
algorithm
Flags.
All
that
things
is
exactly
the
same.
The
new
thing
is
the
functions
which
is
the
SR
v6
endpoint
functions
added
their
mapping
to
SR
v6
s,
a
v6
network
programming
draft.
Then
you
have
the
128-bit
SRV
six
segment
and
some
optional
TTL.
F
We
so
the
question
that
I
had
which
I
was
thinking
also
valid
here,
which
is
we
have
locator
and
function
together
as
one
s
ID,
but
in
Isis
document
and
other
document.
They
also
allow
locator
as
a
separate
ele
and
function
as
a
separate
TL
way,
I'm,
not
particularly
sure
that
it
might
work
for
ISS
where
you
are
flooding
all
the
information.
But
here
it
is
it
better
to
just
keep
everything
together
and
program
together.
F
My
personal
opinion
is
I
like
this,
but
I
really
want
the
feedback
from
others
if
they
feel
that
something
needs
to
change
with
respect
to
the
CCI
object
encoding.
So
this
is
this
is
again
just
a
new
object.
Type
rest.
All
the
procedure
are
exactly
as
the
pc
c
c
sr.
We
don't
have
to
do
much
change
here.
So
most
important
are
these
features
useful?
Would
we
use
pc
c
c
to
do
p
to
MP
+
sr
v6?
F
F
A
A
H
A
F
The
base
piece
ECC
document
assumed
that
this
is
done
out
of
band,
that
why
are
some
configuration?
You
can
do
it,
but
at
the
same
time,
if
you
remember
last
IETF
Chang
Li
presented
something
called
as
a
PC
control
space,
so
using
open
messages
they
could
exchange
that
this
is
the
kind
of
label
space
that
I
would
like
my
PC
to
control.
So
there
is
an
inbound
solution
as
well,
which
is
in
a
separate
document,
but
this
document
assumes
that
that
has
been
done.
F
A
J
Matty
layup
association
in
this
document,
multi-layer
association,
draft
the
daryl
version
provided
a
march
2017
and
I
have
updated
the
days
to
the
second
version
and
this
Marquis
layer
as
LCP,
including
the
Apple
AOSP
and
related
the
lower
layer
SP,
and
they
can
be
associated
to
your
group
to
improve
the
service
provision
under
the
peace,
peace,
app,
extensions
to
Association
object,
including
their
multi-layer
cessation
type
and
is
to
be
assigned
and
the
multi-layer
cessation
group.
It
is
log4
short
under
a
multi-layer
session
TR.
J
We
it
including
H
page
and
the
L
Pete
the
Asia
page
is
that
indicates
that
it
is
the
Atlas
V
and
the
air
page
is
that
indicates
that
it
is
lower
layer,
SP
and
Adam.
There
may
be
many
more
than
one
air
a
lower
layer,
SP
but
owning
one
hour
per
layer
SP
under
this.
This
version
updates
sensor
their
own
version
and
there
are
two
updates
and
the
first
one
is
we
updated
their
operator
configured
under
the
cessation
a
type
list
here
we
as
the
generic
Association
tracked
and
ER.
J
We
add
a
use
case
for
their
inter
domain
stitching
and
though
we
will
make
detailed
discussion
in
for
estefany
under.
There
are
three
use
cases
provided
to
perform
their
multi-layer,
Association
SP
and
the
first
one
is
top
and
widest
adjustment
and,
like
the
figures,
show
their
lower
layer,
sp1
and
Lola
sp2
and
either
they
are
revealed
as
they're
forwarding
the
chanson
see
else:
pf8
SP,
aster
r
/
leo
SP
and
enter
oh.
We
if
we
want
to
update
an
inert
adjuster
the
band
widest
of
the
outer
layer
SP.
J
We
lead
to
first
to
energy,
to
lower
layer,
SP
and
then
nurture
a
player
speed.
There
will
be
two
messages
sent
from
their
pc,
but
if
we
use
the
cessation
group
we
will
lead
only.
We
just
owning
senator
message
to
announce
the
amp
relay
OSP
and
they're
the
PCC
one,
we'll
search
the
law
related
to
lower
SB
and
there
announcer
relates
the
null
else
piece
first
and
then
announce
that
they're
up
one.
So
this
is
the
first
or
use
case
and
asked
their
bo
t
scenario.
J
We
are
used
to
the
same
third
Marty
Leo
Association
LP
under
the
use
case
to
is
30
links
optimization.
This
is
the
same
visual,
but
if
they're
lower
layer,
SP
has
a
failure.
If
the
else
p1
failures
fares
and
it
relate
to
in
from
there
a
player
espy
because
they're
low
low
else,
P
and
field
as
their
T
links
of
their
a
player
espy.
So
if
they're
low,
loyal
else
P
fails,
it
leads
to
in
found
a
player
espy
and
to
to
perform
the
t
link
optimization.
J
So
next
night
stator
use
case
three,
it's
their
inter
domain
stitching
and
there
are
several
domains
and
they're.
Each
domain
contains
a
single
case
each
you,
which
is
which
be
responsible
for
total
for
its
own
domain,
and
each
domain
has
set
up
a
SP
under
the
lowly
OSP
under
a
blow
as
we
we
build
them
as
a
lasting
SP
under
the
lower
layers.
Be
is
field
as
far
SP
and
for
the
a
play
or
SP.
J
F
This
is
true,
my
suggestion
would
be
that
if
you
look
at
the
use
cases,
we
have
RFC
five
six
to
three,
which
was
the
interlayer
framework
in
the
PCE
from
some
time
ago.
Where
have
we
had
the
various
ways
the
PCs
work
with
each
other
to
do
interlayer
path,
computation
and
if
you
use
those
as
the
reference
to
define
your
use
cases,
things
would
be
much
much
clearer
right
now.
I
can
imagine
where
this
could
be
used,
but
when
I
see
the
use
cases,
it
gets
lost
a
little
bit.
F
For
instance,
let's
take
this
case
where
each
PCE
is
aware
of
only
one
of
the
layers,
so
even
if
Association
exists,
the
other
Ellis
fees
are
not
there
in
its
database.
So
how
can
we
associate
so
this?
There
is
a
use
case,
if
we
put
say
the
parent
PC
involvement
or
if
we
say
where
is
the
higher
layer
PC,
which
which
has
the
information
about
both
the
layers,
then
the
Association
object
would
be
very,
very
useful.
F
A
A
It
obviously
knows
that
LSP
a
runs
over
FAL
SBB
because
it
has
to
encapsulate
packets
and
send
them
down
FAL
SPB.
So
if
you
tell
it
to
increase
the
bandwidth
on
LSP
a
it
already
knows
that
it's
running
over
LSP
B,
because
that's
how
it's
been
programmed
it
doesn't
seem
to
me,
like
you,
need
an
extra
Association
to
do
that,
because
they're
already
explicitly
associated
in
the
data
playing
but
to
the
PCE.
A
It
does
make
a
difference
like
because
it
has
VL
espy's
in
its
database.
Did
these
to
track.
Some
kind
of
association
for
PCC
is
reporting
that
information
up
to
the
PCE,
and
it's
a
matter
of
telling
the
PCE
hey.
This
LSP
is
fusing
bad
LSP
is
the
fa
adjacency
and,
and
then
it
makes
more
sense.
So
I
think
my
I
would
I
would
have
had
a
less
confusion
when
I
read
this
I
think
if
you
had
presented
lea
use
cases
from
the
PCE
point
of
view.