►
From YouTube: IETF102-PCE-20180718-1330
Description
PCE meeting session at IETF102
2018/07/18 1330
https://datatracker.ietf.org/meeting/102/proceedings/
A
A
A
A
All
right
so
I'm
also
logged
into
Java
Druce
logged
into
jabber.
If
anyone
talks
a
question
at
each
other,
surely
one
of
us
will
will
see
that
so
as
usual,
this
meeting
is
not
just
in
this
room.
We
are
streaming
live
and
direct
to
the
Internet.
We
have
some
participants
on
meet
echo,
some
of
whom
are
not
even
here.
So
when
you
are
speaking,
please
speak
only
using
the
microphone.
A
The
usual
reminders
he's
the
mailing
list.
Sometimes
a
mailing
list
goes
quiet.
Sometimes
things
happen
to
drafts,
which
in
hindsight
we
we
couldn't
understand
because
nothing
got
discussed
on
the
list
beforehand.
So
please
use
the
list.
Raise
any
issues.
Ask
any
questions,
that's
what
it's
there
for
two
unsure
about
anything
all
you
need
to
clear
anything
up.
Then
please
use
the
list,
and
that
would
be
great
things.
A
A
A
We
had
quite
a
few
issues
raised
by
the
iesg
during
last
call,
and
we
had
to
address
quite
a
few
of
those,
but
we
finally
got
about
done
now.
We
are
just
sitting
in
all
48.
Our
offers
have
48
hours
to
respond
to
VRC.
Editor
I.
Think
we're
still
about
three
weeks
into
that
process.
Now
free
offers
have
replied
to
office
of
not
so
it
would
be
marvelous
that
both
two
authors
could
please
reply,
and
we
can
move
that
one
forward.
A
Besides,
that
we
have
early
code
point
allocations
for
the
staple
PC
point-to-multipoint
draft
and
also
the
segment
routing
draft
more
on
that
in
just
a
second,
some
errata
were
raised
for
some
of
our
already
published
our
FCS
I.
Think
these
all
amounted
to
editorial
clarifications.
All
the
via
writer
were
accepted
on
our
base.
P
SEP
RFC
54:40.
We
updated
a
reference
and
RC
a
bt
231,
we
corrected
a
typo
there
and
we
corrected
a
typo
in
8233
as
well.
A
I
jus,
I
think,
the
third
bullet
there,
w
w
saw
an
r
wi
extension
does
not
waiting
for
an
update.
That's
actually
waiting
for
b
gmpls
piece
of
extensions
updates
to
to
be
done.
If
anyone
here
want
to
talk
about
either
inter
area
is
applicability
or
gmpls
piece
of
extensions,
daniel,
hello,
yes,
Daniel,
King,
so
first
kind
of
a
mere
culpa,
I
think
for
inter
area
a
s
ability,
because
it's
it's
I
think
years
now,
since
you
might
have
done
a
review,
feels
like
very
long
time
actually.
A
A
I
think
there
were
significant
editorial
comments
in
terms
of
improving
the
document
as
a
document
and
I
think
that
the
major
problem
was
because
it
was
such
a
venerable
document.
It
didn't
include
anything
on
staple
PCA,
yeah
and
the
thing
to
know
the
question
is:
is
it?
Is
this
a
useful
document
to
write
and
should
we
commies,
you
know
because
it's
gonna
take
some
work
to
do
those
editorial
changes
and
to
add
a
staple
PC
which
I
think
we
don't.
A
Am
inheritant
inherent
inherently
lazy?
Yes,
that's
that's
what
I
was
trying
to
get
to
I,
don't
mind,
taking
the
lead
and
then
having
the
co-authors
review
and
and
see
if
I
can
delegate
some
of
those
new
sections,
it's
just
a
significant
task
and
I
just
want
to
know
if
it's
really
worthwhile
and
if
it's
going
to
be
a
useful
document.
A
A
A
E
He
provided
Shepherd
review
comment
to
serial
and
I
talked
of
cereal,
and
he
said
he
has
time
to
update,
but
I
haven't
seen
any
update
so,
but
if
he
had
doesn't
have
time,
I
might
help
him
to
reply
to
Julian's
comment.
But
I
have
another
question:
is
W
sound,
RW
and
also
staple
pieces
gmpls?
Can
you
make
Shepherd
review?
You
know
being
progress
in
parallel
sort
of
wait
just
waiting
for
the
previous
one
to
be
completed
because
it
takes
some
effort
right.
A
E
A
E
A
So
we
have
a
queue
for
working
group.
Last
call,
and
now
is
my
mayor
culpa
for
not
having
last
called
more
of
them
in
the
last
period.
I'm.
Sorry
about
that
we
are
a
little
bit
struggling
for
bandwidth
for
shepherds.
So
if,
if
anyone
out
there
is
interested
in
being
a
shepherd,
then
please
talk
to
me.
Offline
I
very,
be
very
grateful
for
the
help.
If
you
could,
just
you
know,
devote
a
few
hours
to
reviewing
one
of
these
drafts
that
I'm
helping
it
through
the
IHG
process.
A
A
F
Drew
from
who
are
we
so
I'll?
Let
me
pick
a
few
PCB
on.
We
did
the
update
recently
added
more
examples
cleared
some
ID
Nets
made
sure
that
some
examples
can
have
ipv6
and
other
things
that
usually
blocks
us
I.
Think
now
would
be
the
good
time
for
young
doctor
review,
especially
when
the
D
topology
and
those
things
are
moving
along.
F
So
it
would
be
really
good
if
the
chairs
can
request
for
yang
doctor
review
the
great
apart
from
that
I
think
the
rest
that
I
am
I'm
editor
of
are
moving
pretty
well
I,
don't
see
any
open
issues
listed
on
the
list
or
by
offline
by
any
anyone
in
the
working
group.
So
if
you
care
about
these
documents,
please
read:
please
come
and
find
me
or
any
other
authors
for
LSP
control
for
flowspec,
for
enhanced
errors,
enhanced
errors
anywhere
who
I'm
on
will
cover
LSP
shelling,
etc.
B
Holly
regarding
the
draft
apology
has
ever
us.
Actually,
we
did
receive
some
comments
of
flying
and
I
have
to
work
only
by
myself,
because
I
didn't
fun,
I
fail
to
connect
with
other
causes,
because
the
draft
has
been
quite
a
long
time
and
well.
My
understanding
is,
after
my
next
update
after
this
meeting,
the
I
think
the
crustal
would
be
ready
for
the
last
call.
G
Okay,
thanks,
it's
Adrian
Farrell,
the
flowspec
work
is
rumbling
along
quite
nicely
we're
sort
of
pushing
the
implementation
side
of
it.
Now.
The
only
thing
that's
really
sticking
in
my
head
is
to
what
extent
this
should
harmonize
with
the
SRT
policy
work,
because
they
are
in
a
sense
similar,
describing
what
traffic
should
go
on.
What
path
I'd
like
opinions.
A
A
F
Truth
I'll
be
talking
about
all
the
set
of
documents
that
we
have
regarding
the
pset
Association
and
let's
go.
This
is
our
Association
landscape.
You
have
the
base
Association
group
draft,
which
describes
the
Association
object,
and
then
we
have
these
bunch
of
other
documents
that
creates
different
types
of
associations.
We
have
on
the
left.
You
have
for
working
group
documents
dealing
with
diversity,
policy
protection
and
bi-directional
Association.
You
have
some
other
individual
documents
which
are
currently
on
agenda
separately,
so
I
will
not
go
in-depth
for
that.
F
Apart
from
that,
there
are
some
individual
IDs
still
dealing
with
resource
sharing
and
make
before
break,
as
well
as
a
new
document
published
for
segment
routing
how
the
candidate
paths
can
be
associated
into
an
SRT
policy.
So
this
is
just
a
0-0
over
shin
recently
posted
it's
right
now,
not
on
the
agenda
separately.
So
this
is
where
we
are.
So,
as
you
can
see
there,
a
bunch
of
documents
dealing
with
association
with
the
base
Association,
we
already
did
working
group
last
call.
F
There
was
one
substantial
comment
pending
from,
which
is
regarding
adding
the
capability
advertisement
for
various
associations
that
are
supported.
We
kind
of
discussed
that
how
to
handle
it
in
a
backward
compatible
way
for
the
things
which
are
already
deployed
versus
something
that
could
be
useful
for
various
different
associations
that
are
being
described
anyway.
So
we
came
up
with
this
approach
that
we
will
have
Association
type
lists
similar
to
how
we
do
of'
lists
in
an
open
message.
F
So
basically,
this
was
the
only
thing
pending
I
think
now
we
are
ready
for
the
next
step,
and
anyway
we
discussed
that.
That's
what
the
chairs
wants
to
do.
As
well
now
the
requirements
for
any
of
the
documents
that
specify
a
new
Association
type,
they
need
to
clearly
state
whether
they
think
that
the
advertisement
during
open
message
should
be
mandatory
or
not.
So
that's
a
requirement
on
them
and
also,
if
you
remember
in
the
Association
draft,
we
have
two
types:
whether
it
is
dynamic,
association
or
operator,
configured
Association
or
an
association
type.
F
That
could
be
both.
So
this
is
a
requirement
that
any
Association
should
clearly
state
that
what
type
they
belong
to
and
if
there
is
especially
the
issue
of
association
range
handling
that
needs
to
be
clearly
stated.
So
this
is
a
that
swai
I'm,
covering
them
together,
so
that
anybody
who
is
thinking
of
this
work,
please
do
take
care
of
these
things
as
well
in
your
document.
F
So
let's
go
through
the
updates
that
we
have
done
to
the
working
of
documents.
The
updates
have
been
pretty
small,
because
most
of
the
contents
have
been
stable.
Main
changes
have
been
because
of
the
recent
change
in
the
association
group
draft.
So
for
diversity
we
kind
of
made
the
capability
advertisement
as
mandatory.
F
Next
for
path
protection:
this
was
recently
adopted
after
the
last
last
ITF.
So
here
also,
we
made
two
updates.
A
one
is
capability
advertisement
mentioning
that
this
is
dynamic
and
one
requirement
that
came
in
that
to
clearly
state
what
is
the
protection
type.
So,
for
this
thing
we
added
a
protection
type
field
and
the
the
handling
is
as
per
RFC.
Four,
eight
seven,
two
so
same
bits
and
same
description
as
described
in
four
eight.
Seven
two
has
been
added
so
now
you
can
tell
what
type
of
protection
is
this
when
you
create
this
association?
F
Regarding
bi-directional
draft
Rakesh
made
the
update
and
he
added
some
more
descriptions
based
on
whatever
feedback
that
he
got
during
the
working
group
adoption
call
and
which
dealt
with
adding
more
figures
and
descriptions.
Basically,
because
he
describes
two
type
of
initiation
mechanism,
whether
it
is
a
single
single
sided,
bi-directional
association
and
double
sided,
bi-directional
association,
so
he
has
added
those
descriptions
and
some
error
handling,
so
that
covers
the
status
of
all
the
working
group
documents.
F
Now
some
other
documents
I
just
wanted
to
give
an
update
to
the
working
group
that
what
people
are
working
on.
So
please
give
comments
to
these
on
the
list
and
later
we
can
think
about
how
to
get
them
adopted
as
well.
So
one
is
resource
sharing.
Currently,
there
is
no
update
to
this
document,
but
I
think
an
update
should
be
coming
based
on
whatever
we
discussed
regarding
making
the
how
to
handle
the
capability
advertisement
and
other
things
as
well.
There
is
an
a
document.
F
This
is
pretty
old,
about
explicit,
make
before
break
and
an
association
object
earlier.
This
was
being
done
in
a
non
association
way,
so
the
authors
have
changed
and
used
an
association
type
and
that
mechanism
to
do
explicit
make
before
break,
and
apart
from
that,
there
has
not
been
any
big
change,
but
I
think
the
document
is
technically
up
to
date
and
it's
fine,
the
the
last
one
is
the
new
document.
F
This
document
describes
that
if
SRT
policy
has
been
created-
and
you
have
a
set
of
candidate
paths,
how
do
you
associate
those
candidate
paths
into
an
SRT
policy
so
good
that
we
got
some
comments
on
this
yesterday
on
the
mailing
list?
So
we
will
be
working
with
the
authors
and
trying
to
resolve
this
pretty
soon.
F
Up
with
a
new
version
as
well,
okay,
so
I
think
the
main
main
thing
is
that
we
have
to
do
is
get
the
Association
group
document
out
of
the
way
and
that's
already
happening,
which
is
good,
and
then
we
can
decide
all
order
in
which
we
want
to
handle
each
of
the
various
Association
types
document
in
the
working
group.
So
please
review
and
especially
for
all
the
other
Association
types.
F
E
Alignment
other
progress,
especially
since
101
IETF
meeting.
He
updated,
including
diversity
requirement
to
indicate
what
is
the
domain
ID
for
exceeding
domain
and
so
on,
and
this
update
include
referenced,
write
references
and
creating
some
reference
and
section
numbers
and
just
to
make
it
ready
for
working
group
last
call
so
I
think
the
authors
believe
this
is
ready
for
working
with
law
school.
Yes
in
McHugh,
okay,
so
again,
so
next
one
is:
we
apply,
stateful
a
PC
to
hierarchy.
A
E
Is
long,
okay,
queue
is
long.
Okay,
so
I
don't
have
to
go
through
yeah
and
PC
applicability.
Actn
also
presented
last
IETF,
basically
aligned
with
a
CTN
framework,
which
now
is
in
RFC
EQ
during
iesg
reviews.
There
are
some
changes,
so
we
try
to
with
that,
and
we
actually
think
that
this
is
also
working
group.
That's
already
so
all
ceramic
you,
okay.
A
F
I
stroke
again
and
the
help
we
will
be
talking
about
be
CCC
extensions,
so
these
documents
have
been
around
for
a
while,
but
they
have
gone
through
a
major
update
this
time.
So
that's
why
I've
asked
the
chase
for
a
little
longer
time
to
go
over
these
documents
a
little
bit
more
in
detail.
So
a
little
bit
history.
We
already
have
our
s
RFC
8,
to
8
3
published,
so
we
have
the
PCE
PC
as
a
central
controller
architecture
and
framework
published.
F
What
can
what
kind
of
applicability,
what
what
kind
of
scalability
and
other
considerations
that
we
need
to
take
if
we
use
PC
as
a
central
controller
and
kind
an
update
we
set
protocol
for
this
purpose
and
during
the
last
discussion
in
the
meeting
we
had
the
agreement
within
the
working
group
that,
yes,
we
are
going
in
this
direction,
anyways
and
let's
get
with
get
on
with
the
actual
protocol
work
and
the
solution
work
and
that's.
Why
that's?
What
I'm?
Here
for
ok,
so
let's
have
a
quick
description.
F
So
basically
we
have
two
documents.
The
first
documents
describes
what
we
call
as
a
basic
mode
and
the
other
documents
talks
about
the
segments
routing,
but
both
documents
are
quite
well
aligned.
They
use
the
same
extension
and
the
same
objects
and
same
mechanisms
as
much
as
possible
in
the
basic
mode
as
it
describes
in
the
RFC.
This
mode
is
nothing
but
static.
Lsp
configuration
you
have
a
controller.
He
calculates
the
path
once
the
path
calculation
is
done
along
the
path
he
instructs
the
node
to
take
forwarding
actions.
F
So
what
we
call
these
forwarding
actions
are
nothing
but
central
controller
instructions,
and
we
have
kept
the
name
little
bit.
Generic,
for
the
reason
that
this
central
controller
instructions
could
be
mean
to
do
multiple
things.
For
example,
it
in
this
case
is
just
an
MPLS
label
download.
It
could
also
be
SR
s,
ID,
SR
v6,
so
multiple
things
are
happening
so
keeping
it
gender
ik
as
a
central
controller
instructions
is
quite
useful,
but
in
this
mode
we
have
described
about
MPLS
label
only
so
once
the
LSP
once
the
labels
are
provision
along
the
path.
F
It's
same,
you
have
for
this
LSP.
This
is
your
out
label.
This
is
the
in
label
on
the
transit
node,
and
this
is
your
out
label.
This
is
your
next
stop
all
that
information
will
be
pushed
from
the
controller
right
to
the
device,
just
like
any
other
SBI
s,
DN
SBI
kind
of
protocol
it.
This
will
change
things
that
you
have
a
piece
obsession
across
all
the
notes
along
the
path,
and
you
are
using
that
session
to
download
information.
F
So
that's
the
basically
what
PCC
C
is.
It's
not
rocket.
Science
is
pretty
straightforward.
So
how
do
we
do
this
via
P
SEP?
First
thing
that
we
need
is:
we
need
a
new
path,
setup
type,
the
good
that
we
have
that
already
we
will
define
a
new
path,
setup
type,
that
this
will
be
PCC
C.
So
you
get,
let's
start
an
ingress
since
a
report
message
says
that
this
is
a
PCC,
C,
LSP
and
I'm
delegating
it
to
you,
handle
it.
F
F
We
listened
to
the
feedback
and
we
kind
of
think
that
it
is
actually
nothing
but
an
initiate
message,
but
with
a
new
object,
which
is
the
CC
ID,
which
is
the
central
controller
instruction,
so
you
identify
that
for
this
particular
LSP
I
am
downloading
this
instruction
on
this
device
and,
of
course,
we
start
with
egress,
then
transit
that
ingress
all
the
labels
are
downloaded
along.
So
this
central
controller
instruction
could
include
just
the
out
label
in
label
and
out
label
just
the
in
label.
F
Any
combination
is
possible
and
that
would
be
supported
and
identified
by
a
unique
ID.
Once
the
labels
are
all
downloaded,
we
update
it
to
the
ingress
or
by
the
way
your
LSP
is
up.
I
downloaded.
You
can
start
sending
the
traffic
pretty
straightforward,
so
that's
what's
happening,
so
your
basic
report
update
message
from
ingress.
They
don't
change
much
this.
They
remain
the
same.
Only
thing
is
they
have
a
path,
setup
type
that
clearly
tell
that
the
setup
type
is
PCC,
see
the
other
thing
that
would
be
changing.
F
Is
these
any
shape
message
going
through
all
the
notes
along
the
path
and
downloading?
Whatever
is
the
label
instructions
that
is
needed
for
PC
initiate
not
much
changes
again?
The
only
difference
would
be
that
you,
when
you
initiate
to
the
to
the
PC,
that
I
intend
to
create
a
PC
cc
LSB
and
get
the
LSP
ID
and
other
things
from
it.
The
main
reason
that
we
are
doing
that
is
so
that
we
can
get
those
identifiers,
because
we
want
to
put
the
same
identifier
later
on
when
we
are
downloading
the
instructions.
F
This
thing
this
is
only
done
for
manageability
reasons
so
that
we
can
equate.
Oh,
this
central
instructions
was
actually
done
for
this
particular
LSP,
so
that
helps
us
with
figuring
things
out.
Rest
of
the
things
kind
of
remains
the
same,
and
you
have
the
our
flag
for
cleanup.
If
you
want
to
ever
remove
it,
is
it
a
clarification
question?
Yes,
you.
H
F
Usually
it's
up
to
you,
but
the
one
that
we
usually
follow
is
start
from
the
end
towards
the
front,
and
then
you
good
thing
is.
We
have
report
messages
anyway
to
clearly
say
that
oh
I
have
a
feedback,
an
acknowledgment
coming
from
all
the
devices
that,
yes,
the
labels
have
been
downloaded,
so
I
am
pretty
sure
that
I
can
now
send
the
traffic
and.
I
You
understand
from
Nokia
the
POS
PID
being
associated
to
the
transit.
Is
that
an
optional
characteristic
of
this,
or
is
it
a
mandatory
characteristic
right
now,
it's
mandatory
so
in
the
case
of
PC
ECC?
Obviously
it's
a
generic
for
any
type
of
label
or
you
want
to
use,
but
if
you're
using
with
srte
it
kind
of
defeats
the
purpose.
That's.
F
Why
I
will
talk
about
this
little
later
and
that
will
be
the?
This
is
the
basic
mode
when
I
have
the
SR
mode
I
will.
So
that's
why
the
CCI
object.
We
kept
it
generic
and
it
LSP
is
mandatory
only
in
this
mode.
Ok,
ok!
So
now,
let's
talk
about
what?
If
I
need
to
make
a
change
into
my
LSP?
What
will
be
my
make
before
break
procedure
in
this
mode?
Well,
because
you
are
changing
the
path
you
must
have
new
set
of
labels.
F
F
You
do
this
operation
first,
once
that
operation
is
over,
you
send
an
update
message
to
the
ingress
asking
him.
Oh
by
the
way.
Now
you
have
a
new
path,
so
we
will
still
tell
him
the
part.
He
will
not
do
anything
with
it
that
ero
he
will
kind
of
it's
only
for
his
information.
He
doesn't
have
to
do
anything.
The
label
signee
wait
downloaded.
H
So
we're
gonna
have
another
clarification
question:
what
exactly
you're
making
in
this
case.
So,
basically,
is
it
just
at
the
tool
or
it's
the
new
path
end
to
end.
If
you
would
not
aa
new
button,
then
type
up
and
a
part.
So
why
do
you
need
the
entire
path
you
need
it
for
RSVP,
because
in
RSVP
you
can
do
only
end
to
end
signal.
In
fact,
but
in
this
case
you
you
just
can
do
the
tauren
and
then
just
walk
yes,.
F
F
We
will
talk
about
the
terminology.
Part
I
know
this
works.
Let's
fix
the
terminology
in
the
document.
I
will
discuss
with
you,
okay
moving
on
now,
we
will
come
to
the
segment
routing
part.
So
what
is
happening
in
segment
routing
is
that
we
are
using.
We
already
have
PC
for
segment
routing
where
the
label
stack
can
be
calculated
by
a
pc
e
and
push
to
the
ingress.
So
what's
new
here,
the
new
part
is
even
the
SRS.
F
Id
assignment
can
also
be
done
by
the
controller,
and
this
is
also
discussed
in
the
RFC
in
the
t's
working
group.
So
basically,
the
s
ID
allocation
can
be
done
by
PCE
and
pushed
to
all
the
nodes
in
the
network
or
the
notes
that
actually
needs
it
in
whichever
way
that
can
be
figured
out
by
the
PC
ECC.
F
So
what
kind
of
what
must
happen
that
there
must
be
some
label
space
clearly
allocated
to
PC,
so
the
PC
will
make
this
allocation,
so
that
means
to
be
done
and
then
from
that
the
node
s
ID
or
prefixes,
ID
or
Acheson
CSID
can
be
allocated
by
pc
e
and
pushed
on
to
the
device.
The
rest
of
the
procedure
to
talking
to
the
ingress
doesn't
change
you
just
prepare.
The
label
stack,
give
it
to
the
ingress
and
the
rest
of
the
segment
routing
works
without
any
change.
F
There
would
be
a
few
things
that
we
need
to
take
care
things
like
identifying.
Clearly
what
the
node
is.
So
for
that
purpose,
a
piece
of
speaker
currently
in
P
sub,
we
simply
use
whatever
is
the
TCP
IP
address
to
figure
out?
How
do
we
reach
the
PCC
or
identify
a
PCC?
So
we
wanted
to
make
that
very
clear
that
it
should
be
better
if
the
node
we
used
in
traffic
engineering
is
advertised
at
the
time
when
the
session
is
getting
formed
from
a
PC,
so
so
that
I
can
map
them
very
clearly.
F
So
that's
something
that
we
have
added
in
the
document
as
well.
So
what
happens
and
what
are
the
new
things
that
you
need?
First,
no
LSP.
In
this
case
we
have
a
new
object,
which
we
call
feck,
which
kind
of
describe.
Basically,
what
is
this
central
SRS
see.
Cid
belongs
to
so
in
this
case,
since
it's
about
downloading
node
information,
the
effect
is
points
to
the
node
ID
and
all
that
details
related
to
that.
So,
basically
think
of
this
as
simple
node
distribution
via
IGP
things
remains
exactly
the
same,
only
differences,
it
is
the
PCE.
F
That's
allocating
this
ID
and
talking
to
all
the
nodes
and
downloading
that
information.
But
what
should
be
then
what
should
be
actually
downloaded
on
to
the
device?
That's
still
done
by
the
node
Spacely,
then
how
to
figure
out
what
the
next
stop
should
be.
That
depends
on
the
IGP
information
running
in
your
node
and
based
on
just
the
s
ID
and
the
and
the
fact
is
coming
from
the
controller
yeah.
I
Just
the
claret
and
Andrew
from
Nokia
just
a
clarification
for
the
previous
slide
about
doing
the
correlation
of
the
TCP
session
to
the
router
ID,
that's
being
embedded
in
this
draft.
Would
it
make
sense
to
export
that
into
its
own
standalone
draft
so
that
way
for
all
piece
up
integration?
You
can
correlate
the
P
sub
session
to
what's
been
advertised
into
the
Ted.
We
can
talk
about.
F
Okay,
now
Jason
see
not
much
change
here.
Only
difference
is
you
have
a
different
fact,
which
kind
of
identifies
both
both
the
end
points
of
that
edges
in
C,
but
rest
of
the
procedure.
Kinda
remains
the
same
and
then
let's
move
on.
So
these
are
some
of
the
common
procedures,
irrespective
of
whether
we
do
basic
mode
or
SR.
We
need
to
worry
about
what
if
VC
goes
down
what,
if
there's
a
session
down?
How
do
I
handle
this?
F
When
would
I
clean
up
instructions
that
I've
received
from
PCE
most
of
the
things
we
are
piggybacking
on
the
the
mechanism
that
we
have
for
LSP
anyway,
so
not
much
difference
whether
it
is
an
LSP
cleanup
or
it
is
a
cc
central
controller
instructions
clean
up.
We
rely
on
the
same
procedure
that
we
already
put
in
in
our
state
for
PC
drafts,
so
using
an
initiate
message
instead
of
putting
a
new
message,
also
helped
us
with
this
procedure.
F
Quite
a
bit
that
now
we
can
piggyback
on
the
state
timeout
in
the
world
on
all
the
procedures
that
we
already
have
clearly
listed
out
in
our
of
C
at
231,
and
we
don't
differentiate
too
much
whether
it
is
a
SR,
LSP
or
RSVP
TLS
bu.
The
way
we
clean
up,
LSP
State
on
the
device.
It's
the
same
way,
we
can
clean
up
the
central
control
central
control
instructions.
Now
we
also
need
to
worry
about
who
the
delegation
procedure
a
little
bit.
F
That
is,
if
the
PCE
that
senses
this
instruction
is
that
B,
if
that
EC
is
no
longer
there,
what
do
we
do
with
this?
Is
there
a
way
as
the
LSP
get
Sri
delegated
to
a
new
PCE,
the
central
controller
instructions
which
were
used
for
that
LSP
also
needs
to
move
along
with
that.
So
in
the
document
we
have
kind
of
explained
the
procedure
for
that.
B
Me
uh-huh
absolutely
I,
didn't
really
very
carefully
all
this
kind
of
read
allocation
part,
but
I
would
like
to
extend
a
little
bit
because
pcs
actually
covers
a
larger
scopes
and
they
are
sweepy
as
it
has
its
own
range
of
nulls
and
corresponding
earth
peace.
So
I'm
wondering
besides
the
case
at
one
piece:
it
comes
shut
down
or
or
other
kind
of
failure
stuff.
Is
it
possible
to
also
consider
the
overloading
of
one
single
PC
and
so
that
some
of
the
errors
piece
or
note
can
come
over
to
another
piece
of
pcs
domain?
F
H
Risk
and
just
one
more
comment:
when
failure
happens
on
the
network,
sometimes
it
requires
to
reroute
numerous
Willis
piece
right.
So
the
vertical
signal
and
the
way
you
describe
could
in
theoretical
antek
by
a
provision
in
a
revision
of
multiple
recipes
at
the
same
time,
by
combining
the
instructions
sent
to
the
same
node
with
respect
to
all
LS
B's.
Have
you
considered
that,
yes,.
F
Yes,
ok
good
thing
is
initiate
message:
allows
you
to
club
a
lot
of
things
together,
so
we
can
definitely
do
that.
So
next
is
the
synchronization
part.
So
if
the
our
session
temporarily
goes
down,
what
will
we
do
again?
Lsp
state
synchronization
mechanism
helps
us
quite
well
here,
so
we
don't
differentiate
too
much.
These
are
just
CCI.
When
the
session
comes
up
during
the
LSP
DB
synchronization
state,
we
also
synchronize
all
the
central
controller
instructions
that
we
have
received.
F
The
difference
would
be
that,
in
case
of
in
case
of
LSP,
it
is
the
PCC
that's
sense.
All
the
LSP
is
that
it
has,
and
the
PCE
takes
the
action
in
this
case
also
we
want
to
do
the
same
thing.
We
say
that
PCC
will
report.
These
are
all
the
central
controller
instructions
that
you
gave
me
before
you
went
down
now.
Tell
me:
what
do
I
need
to
do?
Is
there
any
change?
F
So
it's
the
job
of
the
PCE
to
maybe
use,
initiate
messes
to
remove
some
add
some
and
handle
and
unhandled
it
so
kind
of
very
similar
to
PC,
initiated
LSP.
Just
think
of
if
a
PC
initiated
LS
peas
were
sent
from
a
stateful
PC
to
a
node,
and
their
session
goes
down.
Whatever
thinks
that
the
PCC
has
to
do
for
State
PC
initiated
LSP
the
same
thing.
It
needs
to
do
for
central
controller
instructions,
because
at
the
end,
that's
what
it
it
is.
F
It's
just
a
one,
half
LSB,
as
as
we
were
discussing
earlier,
we
have
lots
of
optimizations
that
we
did
for
LSP
state
synchronization,
all
of
them
RFC
8,
2,
3,
3
and
as
well
as
is
a
document
that
we
presented
the
last
time,
which
is
about
state
synchronization
across
species.
All
of
them
continue
to
work
with
EC
CC
mode
as
well,
so,
which
is
a
good
thing
that
we
don't
need
brand
new
procedure.
We
can
piggyback
on
things
that's
already.
They
are
there,
of
course,
to
support
this.
F
We
need
to
do
if
ability
advertisement,
so
we
will
have
a
capability
TLB
and
we
follow
the
same
same
changes
that
we
did
to
path
setup
type
and
make
it
a
sub
TLV
to
the
to
the
capability.
Tlvs,
etcetera
and
s
bit
is
for
PCC,
see
a
PCC
csr
mode
and
the
basic
is
for
PCC
see
the
changes
to
P
sub
message.
Basically,
inside
your
PC
initiated
message,
you
added
a
central
control.
F
So
this
is
for
initiate
message,
and
this
is
for
report
message.
The
report
message
helps
us
both
to
acknowledge
what
is
the
controller
asking
is
to
do
as
well
as
during
state
synchronization
when
I
need
to
report
or
what?
What
are
the
set
of
central
controller
instructions
that
are
received
earlier?
So
that's
why
we
modify
both
initiate
message
and
the
report
message.
This
is
what
the
CCI
object
looks
like.
It
has
a
type
1
for
basic
MPLS
label
and
type
2
which
matches
closely
to
how
the
SR
s
ID
descriptions.
F
F
A
A
F
A
Yeah
I
think
they're
independent
yeah
if
I'm
like
yeah
right.
Ok,
so,
and
then
that
brings
us
to
you
the
question.
So
we
have
now
two
drafts
talking
about
the
PC
CC
protocol
extensions,
one
for
what
you
might
call
regular
Bay
mode,
one
for
segment
routing
mode,
so
I
will
ask
separately
first,
the
basic
mode.
Please
raise
your
hand
if
you
have
read
the
strut.
A
A
A
Ok,
so
it's
about
8
people
and
please
raise
your
hand
if
you
would
like
to
see
but
adopted
into
the
PCE
working
group.
It's
the
same.
That's
about
8
people.
Ok,
so
it
looks
like
we
have
some
reasonable
support
within
the
room.
I
think
we
can
take
that
simplest
body
all
right
thanks
troops
over
next
presenter
is
Cheng
Li.
K
Hello,
everyone
I'm
Charlie
from
Huawei
Beijing,
so
my
topic
today
is
per
segment
in
he
said
so.
Basically,
we,
the
motivation
is
in
several
huge
cases,
such
chests
and
chain,
one
plus
one
like
pass
protection
by
the
rational
post
correlation
or
like
performance
measurement
like
we
need
the
ability
to
implement,
pass
like
identification
in
s
s
up
networks,
because
you
know,
especially
in
SMPS
networks,
the
the
label
will
be
pop-up
hop-by-hop.
K
So
when
the
package
reached
the
egress
note
the
aggression
or
can
not
identify
which
the
post
the
packet
belongs
to
right,
so
we
proposed
the
path
segments
in
draft
chun
spring
amperes
pass
segment
and
also
similarly,
we
proposed
s
a
v6
pass
ID
in
s,
a
v6
network,
so
which,
which
defines
a
pass
ID
to
identify
nan
as
a
as
a
v6
post.
So
for
configuring
or
allocating
the
the
path
ID
to
enhance
a
path.
So
extension
in
P
step
is
a
knitted
right.
K
So
so,
basically,
we
make
several
things
such
as
base
pass,
ID
allocation
and
conveying
it
with
with
in
peace
app
and
then
the
the
PC
controlled
ID
space
advertisement
and
whether
the
PCC
should
inform
the
PC
the
the
ID
space
that
should
be
controlled
by
the
PC
so
based
on
the
past
segments.
The
bi-directional
passed
correlation
extension
is
defined
as
well,
so
we
propose
three
drafts.
The
first
one
is
roughly
pcs
upper
segment
which
specifies
extension
to
the
receptive
support
past
segment.
K
Identifier
allocation
between
the
speakers,
so
the
second
one
is
Lee.
Pces
are
by
the
restaurant
pass.
So
in
this
draft
it
defines
the
precept:
is
tension
to
grouping
to
reserve
the
reverse
you
uni-directional
as
a
passes
into
a
teapot
by
the
rational
process.
Last
one
is
about
the
mechanism
for
PCC
to
inform
the
receipt
of
the
I
ID
space
and
it's
controlled
by
precept.
So
let's
take
a
look
up.
The
first
draft
in
this
graph
that
we
specifies
a
magazine
mechanism
to
carry
the
s
up
past
identification
information
in
P
step.
K
So
we
support
multiple,
multiple
models
to
to
allocate
the
pass
ID.
For
example,
the
power
ID.
Can
be
allocated
by
the
inverse
PCC
itself
and
then
informed
to
the
PC,
so
the
PC
should
then
inform
the
icarus
PCC
passed
identifier,
information
and
then
the
the
next
one
is.
The
PCC
could
also
request
the
piece
each
allocated
to
allocate
power
I
T.
So
in
this
case
the
PC
should
educate
and
inform
the
assigned
it
pass.
K
Id
through
ingress
and
egress
PCC
using
the
piece
AB
messages
and
the
third
one
is
that,
like
the
PC
can
allocate
the
pass
ID
directly
on
his
own
accounts
and
inform
the
ingress
and
egress
PCC
so,
and
we
also
like
plan
to
add
some
text
to
clarify
to
describe
the
the
another
mechanism,
another
mode
of
allocating
the
pass
or
ID
in
this
mode.
The
possibility
can
be
allocated
by
the
the
in
eager
PCC.
K
So
the
PC
should
request
the
the
eager
PCC
to
allocate
and
as
poss
ID,
and
then
it
should
inform
the
ingress
and
egress
PCC
about
the
post,
ID
education,
so
so
for
support.
Supposing
the
the
asset
allocation.
So
we
meet
some
competitive
tracks
in
s
RPC
capability
and
SRV
6pc
capability
theories.
So
we
basically
define
two
flex
post
and
them
the
same.
We
call
it
pass
post
identification
bit,
so
it
should
be
sent
it
should
be
set
to
when
we
need
to
like
require
the
pass
ID.
K
We
need
to
support
like
it
should
be
said
when
the
P
piece
P
step
speaker
support,
which
we
supposed
the
the
piece
outside
the
allocation.
So
for
indicating
past
idea,
log
request
PFLAG
should
be
like
added
in
the
air.
Speed
object
as
well,
so
in
the
PFLAG
must
be
set
in
pc
request
and
peace
report
message
when
the
pcc
requests
the
pass
ID
allocation,
so
in
a
system
is
similar,
so
the
the
P
Flags
must
be
set
in
the
PC
reports
and
PC
update
and
his
initiate
message.
K
When
PC
replied
to
the
power
side
allocation
requirement,
yep
so
for
carrying
the
post
ID
information,
we
create
a
new
tier.
We
coded
personality
in
in
the
errors,
P
objects,
so
Sara
fields
are
defined.
You
that
the
first
one
is
ID
t
ID
tag
field
which
defines
the
type
of
a
post,
ID
and
zero
means
emperor's
pasa
segment,
which
is
a
amperes
label,
and
one
means
s.
K
Rv6
post
ID,
which
defined
in
the
drafts
at
least
bring
passive
P,
m4
s
f6
and
P,
so
the
fraks
we
currently,
we
define
three
flags
error
flags
for
local
global
pitch,
see
for
PC,
CPC
and
E
for
egress
and
ingress
imposter
ID
field,
which
can
the
the
the
value
of
the
past
of
the
pass
ID.
So
this
this
field
that
the
pass
ID
time
is
like
indicated
by
ITT
ITT
right
so
as
as
we
introduced
before,
like
the
PC
should
inform
the
egress.
K
The
usage
or
the
information
of
the
personality
allocation
are
abiding
mapping
information.
So
we
used
instant
the
procedures
of
ID
child
PC,
t-cell
extension,
PC
controller,
so
by
defining
a
new
like
FEC
object
to
inform
the
post,
ID
identification
information
to
the
icarus
PCC,
so
one
or
more
foreign
terrible
earth
divided
for
for
for,
like
into
describing
the
outside
information,
such
as
a
symbolic
post
name
and
our
sphere
identifier
and
a
speaker
entity,
ID
Tara
wish
so
the
past
information.
K
Possibly
information
can
be
encoded
in
this
like
ERV,
so
that
the
PC
can
inform
the
icarus
PCC
by
the
FEC
object
and
also
the
CSI
object.
So
this
is
easy
instant.
How
to
initiate
to
introduce
the
PC
allocated
pass
IT
mode
on
its
own
accord.
So,
as
we
can
see
from
the
picture,
so
the
PC
could
like.
Maybe
it
can
like
use
the
PC
in
initiate
message
to
increase,
to
initiate
our
CP
an
SP
o
own
as
a
pass
with
the
possibility
allocation
by
the
post.
K
Id
govt
within
the
air
speed
object
and
then
the
e
inverse
P
PCC
would
report
the
error
cos
a
post
installation
if
it
is
successful
and
reports
the
PC
by
the
PC
reports
message
which
the
defrag
is
set
right.
So
after
a
confirmation,
the
PC
should
like
inform
the
egress
PCC
with
his
initiate
message
carrying
the
EFI's
object
because
to
pass
and
a
corresponding
CCI
object.
So
in
this
way
the
the
egress
PCC
could
understand
and
lend
the
mapping
information
of
the
past
ID
and
and
the
detailed
post
information.
K
So
the
second
draft
is
about
the
by
direction
of
heart
by
directional
I.
Suppose
Association
group
is
tension
in
in
peace
app.
So
so
for
associating
to
like,
as
a
passage,
this
document
is
defined.
A
new
Association
group
called
the
double
side
bi-directional
as
a
past
Association
group,
so
with
within
this
Association
group,
the
US
s,
hypothesis,
information
can
be
like
describe
or
conveyed
by
this
object.
K
So
this
is
this
is
the
example
example
to
describe
the
PC
is
initiated
by
the
rational,
rational
pass,
and
in
this
example,
we
can
see
that
the
he
can
like
use.
The
piece
PC
update
of
his
initiate
message
to
install
the
bi-directional
by
the
rational
pass
into
the
ingress
and
egress
PCC,
so
yep,
the
third
one.
K
But
in
these
two
drafts
they
assume
that
the
label
range
to
be
used
by
the
PC
is
known
and
set
on
both
pieces
appears.
So
for
solve
this
problem.
We
specify
the
extension
to
support
advertisement
of
the
various
ad
space,
such
as
amperes
label
and
pass
I
as
a
v6
pass
ID
space
to
the
PC
so
that
the
PC
can
allocate
the
post
ID
directory
and
then
dynamically
yep.
K
So
this
is
the
procedure
of
how
the
PCC
and
the
PC
is
changed.
The
ID
of
the
PC
showed
ID
space,
as
we
can
see
from
the
picture
like
we
need
in
include
the
SR
with
six
post
ID
control
space,
all
the
label
control
space
which
defined
in
this
document
currently
in
the
open
message
to
inform
or
exchange
each
other
like
how
which
range
that
the
PC
can
be
used
actively
yep.
K
So
this
is
the
example:
the
format
of
the
label
control
space
TV,
as
we
can
see
that
the
fraks
we
only
define
on
one
flag
called
a
flag
means
all
space
flat,
oh
space
of
the
ID
space,
and
we
maybe
we
we
have
like
several
blocks
of
the
ID
space
that
can
be
used
by
the
PC,
so
this
kind
of
information
can
be
described
by
the
products
which
combining
the
of
the
stars
and
wrench
information.
So
this
is
the
the
format
of
the
label
control
space,
govt
and
the
SF
is
expose
ID.
L
K
L
K
L
K
L
You
mean
this
one,
it's
a
generic
question
because,
as
you
start
assigning
path,
IDs,
which
we
potentially
could
use
to
match
SR
policies
to
to
do
some
analytical
analysis,
because
there's
no
context,
it
means
that
you
would
have
to
push
its
information
every
device
in
the
network.
That's
the
intention.
No.
K
L
L
L
So
if
you
want
to
do
account,
inclusion
pass
ID,
which
is
the
main
application.
It
has
to
be
known
on
every
device
where
packets,
with
potentially
Traverse
true
cuz,
it
would
trigger
increase,
encounter
and
some
reporting
back
right.
So
you
wouldn't
know
in
advance
where
your
packet
would
go
to
so
meaning
you
will
have
to
push
the
permission
every
device
through
this
obsession.
K
K
F
On
Swing
to
what
Jeff
was
saying
so
I
agree
that
accounting
is
one
of
the
use
case,
but
in
this
document,
as
of
now,
we
are
not
talking
about
it.
Yeah
we
are
talking
about
just
setting
up
a
bi-directional
LSP
and
note
noting
the
path
ID
allocated,
either
by
the
ingress
or
by
the
PC
to
the
egress,
but
I
agree
with
you
when
we
do
path
accounting
by
this,
this
is
a
problem
that
we
need
to
handle.
Yes,
this.
F
A
The
last
speaker
is
Danny
le,
while
we're
waiting
for
dummy
alley
tickets.
My
front,
where
are
the
blue
sheets,
is
anyone
sitting
on
the
blue
sheets?
Is
anyone
not
signed
the
blue
sheets?
Please
raise
your
hands.
Okay,
I.
Think
they'll
they'll
make
my
way
forward,
and
then
you
can.
You
can
sign.
Thank
you.
M
M
M
How
do
we
do
that?
Didya
is
to
use
the
Association
object.
Dissociation
object
simply
takes
a
number
of
LSPs
and
it
creates
an
association
between
them.
We
need
a
new
type
of
our
substation
to
explicitly
say
that
this
association
means
these
rsps
belong
to
the
same
user
network.
There
are
also
two
new
TL
visa.
The
file
defined
in
the
in
the
draft
one
is
the
virtual
network,
TLB
and
another
one
to
carry
vendor
specific
information
changes
since
last
version.
Since
it
was
up
for
meetings
ago,
there
are
quite
quite
a
lot
of
them.
M
M
Basically,
is
the
specification
of
the
details
for
the
virtual
network
TLB
and
the
Ayana
part,
which
was
a
little
bit
poor
in
the
past
rationale
for
doing
this
work.
Why
do
we
need
the
be
an
association?
As
I
said
this?
This
is
needed
between
the
parent
and
the
child
PC
to
share
information
about
the
virtual
network.
A
given
LSP
belongs
to.
M
Does
the
child
PC
needs
to
be
aware
needed
to
be
aware
of
this
information?
Yes,
in
most
cases,
it's
useful.
Why?
First
of
all
the
child
PC
is
a
responsible
for
peer
domain,
LSPs
and
sometimes
also
the
the
child.
Pc
of
the
ingress
domain
is
responsible
for
the
entire
and
to
and
path
it
is
possible
that
in
the
same
domain,
controlled
by
the
same
child
PC,
there
are
several
else
piece:
B
belonging
to
the
same
virtual
network
and
having
a
different
English
and
egress
nodes
within
within
that
that
domain.
M
So
why
there
is
the
need
that
we
have.
This
information
optimisation-
it
could
be
that
a
request
against
the
LSP
is
of
a
virtual
network
is
I,
don't
know,
share
as
many
resources
as
possible.
So
knowing
that
those
LSPs
belonged
to
the
same
virtual
network
allows
to
do
the
party
computation,
which
maximizes
their
usage
of
the
same
resources
or
we
need
the
diversity,
optimizing,
the
diversity
and
so
on.
M
Second,
finger
is
again
knowing
that
different
LSPs
belong
to
the
same
virtual
network.
Allows
us
for
relaxation
of
constraints
sometimes
is
not
possible
within
the
same
domain,
to
provide
super
strict
constraints
that
are
requested.
The
two
against
the
virtual
network,
the
most
common,
is
a
total
diversity.
It
might
be
possible
that
there's
no
way
to
accommodate
such
such
such
requirement,
knowing
that
those
LSPs
belong
to
the
visual
network,
it's
possible
to
relax
the
constraint
against
the
virtual
network
and
maybe
go
for
partial
partial
diversity.
M
I
think
this
is
basically
a
visualization
of
what
I
said
so
far.
Probably
the
only
addition
is
the
fact
that
local
policies
which
are
usually
implemented
within
a
single
domain
or
PR
LSP
can
be,
can
be
applied
at
all
this.
All
the
LSP
is
belonging
to
the
same
virtual
network.
Next
steps.
Well,
the
document
has
been
even
if
not
presented
not
to
discuss
recently
has
been
around
for
a
while
is
pretty
stable.
I
would
say
that
the
latest
changes
significantly
improved
it.
M
H
You
are
bursting
from
hallway,
so
the
question
is
I
understand
what
you
are
saying:
it's
it's
a
good
idea,
but
maybe
it
could
be
done
in
a
more
generic
way,
basically
to
say
that
this
is
association
with
a
particular
mercial
Phyllis
piece
or
a
set
of
tunnels,
rather
than
to
bind
it
necessarily
to
the
virtual
network.
There
could
be
applications
wider
than
virtual
network
for
using
measures
or
philosophies.
For
example,
I
do
understand
when
you
do
say
modification
and
you
can
do
make
with
a
break
for
entire
mesh
right.
M
This
is
just
a
new
one,
which
could
be
specific,
or
generic
I
mean
the
way
the
Association
is
defined
is
already
generic
in
the
sense
that
you
are
free
to
define
as
many
tea
also
be
Yogi's,
as
you
want.
So.
H
What
I'm
trying
to
say
that
say
in
tears
when
we
are
doing
t
tunnel
model-
and
we
are
considering
things
like
t
tunnel-
sets
right
so
and
we
do
not
envision
that
this
T
a
tunnel
set,
would
be
only
you
say
for
virtual
networks.
Okay,
so
that
you,
obviously
all
the
tunnels
that
will
do
on
the
chi
tunnel
set,
will
have
the
association
you're
talking
about.
H
E
E
E
M
A
Okay,
so
we've
got
some
good
feedback
from
Eagle
But,
notwithstanding
Eagles
feedback.
I
think
this
has
been
around
for
a
while
some
for
first
time
we
talked
about
it.
It
looks
like
it
sold
some
necessary
part
of
a
CTN,
so
I'd
be
fairly
comfortable.
Asking
the
question
who
and
who
in
the
room,
has
read
the
draft.