►
From YouTube: IETF105-PCE-20190725-1000
Description
PCE meeting session at IETF105
2019/07/25 1000
https://datatracker.ietf.org/meeting/105/proceedings/
A
A
B
A
We're
cells
they're
already
so
even
if
you
come,
are
supposed
to
be
aware
of
the
IPR
policy
within
the
ETF.
You
supposed
to
be
aware
of
that,
even
when
you
subscribe
and
restart
to
the
unit
of
meeting,
so
please
take
some
time
to
be
aware
of
the
wounds
on
make
sure
you
abide
by
them.
When
you
contribute
to
the
hf
so
usual
stuff,
the
blue
sheet
should
be
running
across
the
room,
make
sure
you
have
signed
them
down
before
during
the
meeting.
A
A
We
also
have
you
nothing
golf
connected
to
trouble
parental.
Please
make
sure
that
you
come
to
the
mic
on
the
pink
cross,
make
sure
when
you
have
question
to
you
the
mic,
because
we
have
streaming,
we
are
recording,
so
everything
should
be
hold
by
the
mics
and
make
sure
you
everyone
can
understand
what
you
say
remember
to
stay
tuned
and
when
you
come
at
the
mic.
Also,
this
is
your
football
sore
hari
taking
the
minutes
and
people
remote,
especially
we
have
people
attending
remotely
by
the
way
one
of
them
will
be
presenting
later.
A
So
this
is
not
just
Siri.
There
are
people
connected
from
remote,
so
how
I
said
before
how
he
has
been
appointed
as
the
PCE
working
group
secretary.
So
thanks
again
Harry
for
agreeing
to
take
that
position,
Harry
M
is
to
among
us
they
succeeded
in
helping
flush
the
backlog
of
the
PC
document
on
other
stuff,
PC
working
group.
So
thank
you
for
the
help
agree
on
it's
likely
that
it's
your
last
meeting
as
a
PCE
co-chair.
So
thanks
a
lot.
Thank
you
for
everything.
A
The
meaning
is,
as
usual,
the
many
documents
that
will
be
presented
here
that
haven't
been
discussed
on
the
list.
Yet
this
is
a
pity.
You
don't
have
to
wait
for
a
presentation
to
trigger
discussion
on
the
list.
Many
things
can
happen
between
the
meetings,
so
the
list
is
open.
24/7,
please
keep
on
trying
triggering
discussion
there.
A
That's
a
way
to
make
it
all
progress,
even
between
the
meetings,
so
it's
very
useful
to
force
yourself
to
trigger
discussion
over
there,
because
that's
one
of
the
tools
on
the
PC
working,
we'll
probably
under
use
that
tool
the
week
it's
been
a
while
that
we've
been
set
up
that
week
in
place
seconds
so
correctly,
up-to-date
and
it
allows
us
to
track
various
cues
for
document
adoption
in
the
walk
you
noob
walking
with
Blasco
sharing
responsibilities
and
so
on.
So
this
is
useful
tool.
A
You
may
be
interesting
also
to
bring
in
your
a
test
there
if
it
makes
sense
it's
open
to
anyone
with
an
account.
So
if
you
feel
something
isn't
accurate,
you
may
also
add
from
there
the
link
is
included
as
right.
So
have
a
look
at
it.
It's
useful
info
that
goes
beyond
what
the
data
tracker
is
already
providing,
so
feel
free
to
look
at
and
maybe
comment.
A
A
Decided
that
we
should
abide
by
this
rule
by
including
an
implementation
status
section
in
each
document,
so
unless
specific
reason
to
be
discussed
means
the
the
chair
is
identified,
we
expect
that
all
PC
document
include
that
kind
of
section,
and
it
is
especially
true
for
document
occur
close
to
the
walk-in
Blasco,
be
aware
that
we
won't
start
working
with
last
goal
without
addressing
this
issue.
It's
been
already
pointed
out
on
the
list
review
them
in
this
situation.
So
this
will
be
a
breaking
point,
starting
from
now,
which
is
already
a
breaking
point.
A
A
B
We
we've
been,
as
Jillian
said:
we've
been
working
quite
hard
to
push
the
backlog
out
of
the
working
group,
and
the
result
of
that
is.
We've
now
got
a
lot
of
documents
that
are
beyond
the
working
group.
Since
Park
two
of
them
came
out
as
RSC's
the
ability
to
actn
and
the
state
for
point-to-multipoint,
both
of
which
have
been
a
little
while
in
the
in
the
production,
so
I
thought,
but
it's
good
to
have
them
out.
B
Then
we've
got
three
more
that
are
in
the
RFC.
Editor
queue
so
they're
shortly
to
become
RFC's.
The
first
of
those
the
segment
routine
document
is
part
of
cluster
340.
A
cluster
is
a
a
group
of
mutually
dependent
documents
that
need
to
be
progressed
through
the
RFC
editor,
as
a
block
I
believe
that
all
of
those
documents
are
now
with
the
RFC,
editor
and
and
340
is
going
to
come
out
soon,
so
that
RFC
will
come
the
second
one,
w
son
RW
a
is
dependent
on
the
gmpls
document.
B
C
B
B
Okay:
two
documents
are
with
the
ASG
they're,
both
actually
really
sort
of
back
with
the
authors,
after
iesg
review
and
pending
updates,
I
think,
if
there's
no
well
updates,
they
need
to
make.
But
if
any
of
the
author's
is
confused,
please
talk
to
the
chairs
and
the
Shepherd
to
move
that
forward,
and
then
we've
got
we've.
We've
scared,
our
ad,
which
is
quite
fun
because
part
of
this
moving
the
backlog
means
we've
suddenly
pushed
rather
a
lot
of
documents
through
there
are
now
six
in
various
states
with
with
the
ad
with
the
ASG
I.
B
B
D
We
have
some
documents
which
the
authors
in
the
past
have
claimed
that
there
are
nearing
working
group
last
call
and,
as
Julian
mentioned,
the
main
gating
point
where
they
are
stuck
is
the
implementation
policy
that,
as
a
working
group,
you
all
have
agreed
to,
and
please
talk
to
us
regarding
how
to
progress
this.
Since
maybe
this
is
the
first
time
you
are
adding
this.
So
is
there
any
questions?
We
are
here
to
help
you
out,
and
some
of
the
documents
have
already
added,
so
you
can
also
refer
those
documents
on
the
amount
of
detail.
D
Note
that
that
this
is
your.
This
doesn't
mean
that
you
must
have
an
implementation.
It
just
means
that
you
must
talk
about
it
in
the
document
and
even
if
you
have
implementations
that
are
coming
or
you
don't
plan
to,
you
can
put
that
information
as
well
in
the
dis
section,
so
start
talking
about
it,
and
then
we
can
start
progressing
these
documents.
The
last
document
in
the
set
is
the
gmpls
document
that
got
merged.
During
the
last
last
meeting,
a
refresh
was
promised,
which
is
still
pending,
which
was
supposed
to
deal
with
more
editorial
changes.
D
E
D
So
now
one
document
which
got
updated
so
let's
look
at
the
status
of
that
document,
which
is
our
PCP
and
one
technical
change,
was
with
respect
to
key
where
LSP
ID
key
also
got
added
into
the
LSP
dB.
This
is
useful
and
make
before
break
cases
where
the
you
may
have
for
a
temporary,
multiple
LS
B's,
so
using
the
LSP
ID
as
a
key
is
helpful.
So,
based
on
that
comment,
this
change
was
done
and
rest
of
the
all.
D
The
comments
are
related
to
the
RFC
style
guide
and
how
to
properly
use
the
references
or
droughts
and
etc,
which
were
discussed
on
the
list
and
all
those
things
have
been
done
as
a
status.
The
early
young
doctor
review
of
this
document
is
also
done,
but
we
do
not
have
any
implementations
of
this
young
model
yet
so,
if
you
are
aware
of
this,
or
are
you
planning
to
implement,
please
reach
out
to
the
authors
or
on
the
list
or
the
chairs?
D
So
that's
about
the
yang
model,
some
other
documents
which
are
not
yet
fully
cooked,
but
but
we
need
to
discuss
one.
Is
the
PCC
see
the
PCC
C
document?
We
did
Maidan
of
the
update
was
made
and
it
was
posted
and
they
have
added
the
implementation
status
discussing
about
what's
what's
implemented.
With
respect
to
PCC
see.
The
second
document
is
the
native
IP.
This
was
presented
last
time.
D
There
is
no
update
and
I
think
mostly
dependent
on
the
movement
that
we
see
in
the
t's
working
group
of
the
related
documents
we
have,
the
LSP
scheduling
refresh
was
made,
but
more
technical
changes
again.
I.
Think
implementation
of
this
document
is
something
that
the
authors
would
like.
So
if
you
have
plans
for
implementing
this,
please
teach
out
to
the
working
group
to
documents
which
were
recently
adopted
or
the
flex
grid,
and
the
SRV
six
supporting
piece
up
and
the
one
document
that
we
wanted
to
explicitly
discuss
was
the
enhanced
errors.
D
This
was
an
agenda
in
IT
f-104
and
it
was
presented
after
that.
We
decided
that
there
should
be
a
guideline
of
all
the
future
documents,
how
they
can
use
the
enhanced
errors
and
notification.
So
the
guideline
is,
as
you
can
see
in
the
slide,
it
requests
that
any
new
document
if
they
are
defining
new
error,
type
or
notification,
they
should
think
about
whether
this
error
needs
to
be
propagated
in
multiple
domains
and
what
kind
of
error
criticality
condition
they
should
refer.
D
So
this
is
the
die
line
that
at
least
on
the
mailing
list,
who
kind
was
agreed
on.
So
if
you
have
any
thoughts
on
this
guideline
still
reach
out
to
us
and
we
plan
to
either
put
this
guideline
on
wiki
or
in
the
draft
itself
or
both,
and
the
request
would
be
that,
let's,
since
this
is
our
working
of
document-
and
if
we
have
consensus
on
this,
let's
start
making
sure
that
all
the
documents
which
define
new
errors
also
refer
to
the
enhanced
error
techniques
available.
D
D
D
C
D
So
wiki's
open
when
you
make
some
decisions,
and
you
want
to
remove
the
document
from
the
adoption
pool
cue
and
go
back
and
work
on
it
or
whatever
decision
that
you
make
come
back
to
us
and
let's
keep
the
wiki
up
to
date
so
that
all
in
the
working
group
are
aware
of
what's
happening
when
daughters,
okay,
thank
you.
Let's
start
with
the
first
presentation.
A
G
So
these
are
both
stored
on
both
the
PCC
and
the
PC
and
they
have
identical
structure
on
the
PCC
on
the
PC
and
one
thing
we
do
is
we
update
the
RFC
a23
one
by
saying
that
PC
request
is
not
mandatory
for
the
PCC
to
send
so
this
simple
files,
basically
OSP
bring
out
so
first
we
introduce
this
LSP
database.
So
we
kind
of
have
these
five
basic
rules
about
what
it
is
and
so
on
the
PCC.
G
G
So
this
one
has
is
more.
It
can
be
updated
kind
of
both
by
messages
and
configuration
on
both
the
PCC
and
the
PC,
and
so
each
we,
we
kind
of
say
that
each
a
so
is
a
container
for
LS
B's.
It's
not
a
container
of
tunnels
so
because
if
we
try
to
make
it
a
container
of
tunnels,
then
it's
not
possible
to
do
a
make
before
break
from
one
association
to
another,
because
you
can't
say
that
you,
my
re-up,
tell
us.
P
is
in
a
different
Association.
G
G
H
Okay,
so
one
thing
that
simple
comment:
if
you
want
to
have
a
different
name
for
LSP
I,
would
suggest
connection
I,
always
like
this
better
than
LSP.
For
the
reasons
you
just
described.
Okay
and
basically,
it
is
also
better
than
path
because,
but
you
would
confuse
passes
computed
paths
which
are
not
provisioned
so
allow
the
connection
database
would
be
a
better
term.
Okay,.
I
So
this
is
more
firm,
I
guess.
A
question
for
the
group
was
like
understand:
Nokia
I
think
this
is
more
of
a
question
for
the
group.
We're
diving
board
point
to
the
fact
that
PC
request
is
not
supposed
to
modify
state.
There
are
maybe
implementations
out
there
today,
which
are
doing
reservation
with
given,
of
course,
if
it's
a
PC
request,
you
might
not
expect
a
report
coming
back,
so
there
could
be
time
of
mechanisms
and
other
implementations
to
still
hold
reservations,
primarily
in
the
use
case
of
bandwidth.
I
B
It's
three
things
and
be
really
careful
about
specifying
the
format
of
databases
in
implementations
that
makes
no
difference
to
interoperability
and
what
we
do
as
protocols
for
the
wire
so
I
only
and
say
what
state
what
information
needs
to
be
stored
about?
How
it's
stored
is
up
to
your
competitors
to
screw
up.
You
want
to
encourage
them
to
screw
it
up.
B
Think
it's
too
strong
to
say
a
PC
is
not
supposed
to
create
any
state,
a
piece
he
probably
doesn't
need
to,
but
there's
this
kind
of
concept
of
sticking
state
where
you
you
make
a
request,
and
you
don't
know
whether
the
networks
going
to
use
that
information,
but
you
kind
of
want
to
get
ahead
of
the
IGP
just
in
case.
They
do
so
think
about
that
and
the
third
one
was
82.
31
updates
I've
got
this
little
draft
clarifying
the
use
of
bits.
B
G
Want
to
quickly
just
say
about
the
peace
request,
not
modifying
the
LSP
TV,
so
I
think
there
may
be
a
use
case
where
you
kind
of
use
the
PC
for
OAM.
So
say
you
want
to
ping
like
an
inter
domain
path
right
and
you
just
want
to
basically
say
you
know,
give
me
some
path
right,
but
you
don't
want
to
touch
the
network.
You
don't
want
to
modify
any
state
in
the
network.
Based
on
that,
like
you
just
it's,
it's
like
you
know,
you
just
want
to
probe
the
network
right.
J
Burger
I
agree
with
Adrienne's
point
about
the
databases
you
don't
want
to
specify
them
as
formal
requirements,
but
as
a
conceptual
talking
point
as
a
way
of
characterizing,
the
information
that's
being
stored,
I
think
that's
really
useful.
It's
good
to
have
some
common
terminology,
so
I
actually
like
that
and
just
don't
make
it
that
you
actually
have
to
build
it.
That
way,
the
have
you
thought
about
protection
paths
and
now
you're
gonna
represent
them
protection.
J
You
can
there's
there's
different
ways
to
model
if
you're
thinking
about
it
from
a
database
representation
and
you
could
choose
just
to
leave
them
uncorrelated
and
then
you
have
to
walk
the
whole
database
to
find
the
Association.
You
know
which
ones
are
associated
or
you
might
do
some
formal
linkages.
Okay,.
G
Cool
one
thing,
though,
about
the
the
database
I
think
we
should
specify
the
key
of
the
database
right
because,
like
if
different
vendors
use
like
if
one
vendor
uses
the
LSP
ID
as
the
key
and
another
vendor
uses
the
tunnel
ideas,
the
key,
like
you
end
up
with
a
broken.
You
know
they're
not
gonna,
interrupt
and
then
nobody
wins
in
that
situation.
So.
J
I'd
say
that's
an
appropriate
thing
to
think
about
if
you
are
defining
data
models
like
yang
models,
if
you're
talking
about
just
sort
of
informative
concepts,
it's
fine
to
say
that
you
think
of
it
that
way,
I,
don't
think
you
can
specify
it.
If
you
want
to
specify
it
I
think
you
have
to
end
up
with
some
foremost
Asian
language,
which
we
end
up
with
doing
yang
I.
J
K
D
K
G
So
this
is
basically,
you
can
have
association
between
tunnels.
If
every
I
was
peeing
in
in
tano
is
associated
to
every
RSP
and
tunnel
B,
then
the
two
tunnels
are
associated
but
making
it
purl
SP.
It
gives
some
more
use
cases
such
as,
if
you're
doing
bi-directional,
for
example,
then
the
two
LS
B's
there
may
be.
You
know
a
real
plus
B,
which
is
associated
to
the
real
optimist,
be
on
the
other
end,
it's
not
associated
to
the
current
of
us
B
on
the
other
end,
so.
B
Services,
agent,
again
I,
think
we
have
a
disconnect
on
terminology
that,
when
you're
down
in
the
transport
world
a
there's,
a
double
disconnect
on
terminology
and
not
not
the
layer,
4
transport,
the
lion
transport
when
you're
down
in
the
late
in
the
underlying
transport
world
a
tunnel
is
a
pipe
in
rsvp-te
world.
A
tunnel.
Is
this
conceptual
collection
of
potential
connections?
We
call
LSPs
right,
so
it
tells
term
tunnel.
Is
your
turn
LSP.
K
G
K
H
So,
basically,
just
a
clarification
when
we
model
T
tunnel,
we
basically
saying
that
Alice
piece
that
go
on
to
the
same
tunnel
like
walking
and
protection
Alice
B's.
They
do
not
meet
any
association,
they
already
I
children
of
this
tunnel
and
associated
native
lives.
For
that,
when
we
say
associated
tunnels,
it
means
that
there
are
tunnels,
basically
all
their
connections
or
only
working
connections.
It
doesn't
matter,
but,
for
example,
there
is
an
association
one
supports
one
direction
in
data
plane
and
another
in
opposite
direction
and
so
forth.
H
This
is
like
a
higher
level
of
Association,
so,
basically
putting
everything
into
connection
Association
or
I
lost
association.
It's
actually
confuses
a
lot
of
things:
okay,
so
again,
LSPs
that
belong
to
the
same
tunnel,
they
associated
by
definition,
their
association
object.
The
tunnel
ideas,
association,
object,
yeah.
H
G
H
N
N
There's
carotid
Ness,
you're,
so
cold
out.
Goodness
is
one
example
where
you
want
the
LSPs
of
the
same
tunnel
or
possibly
forward
and
reverse
to
be
associated
right
right,
I.
Don't
think
we
should
dictate
that
you
cannot
do
it
per
tunnel
or
you
cannot
I
mean
LSP.
Granularity
is
fine
by
me,
which
is
the
fine
granularity,
but
one
can
choose
to
do
it
on
per
tunnel
and
in
fact
you
know
if
you
have
diversity
you
want
to
always
be.
N
G
C
G
D
Thanks
Mike
Mike,
as
you
progress,
this
work
on
I
think
a
suggestion
that
I
would
give
you
for
this.
Like
you
know,
right
now,
it's
a
mix
of
clarifying
things
mix
off
making
changes
into
RFC
eight
two
three
one.
So
we
need
to
scope
it
a
little
bit
better
and
maybe
the
things
that
needs
to
update
eight
to
three
one
work
with
Adrian
and
we
we
scope
it
as
a
standard
strat
thing
and
the
things
which
are
clarifying
there.
D
We
and
again
when
you
are
clarifying
I,
think
currently,
you
clarify
from
the
database
point
of
view
and
you
think
of
what
is
the
database
at
see,
see
what
is
the
database
at
PCE,
but
in
ITF
and
in
we
should
worry
about
what's
on
the
wire.
So
if
there
is
any
confusion
between
what
people
are
encoding
in
LSP
identify,
TLD
or
people
are
making
mistakes
with
symbolic
paths,
name
and
some
correction
needed
there.
D
I
Agree
with
you
that
you
know
we
should
I
understand,
okay,
ideally
that
we
should
be
concerned
about
protocol
on
the
wire
stuff.
However,
in
this
case,
what
Mike
is
trying
to
do
here
is
clarify
interpretation
of
those
values
on
the
wire,
so
during
in
araba,
ran
into
that,
where
the
meaning
of
a
certain
number
of
the
meaning
of
a
certain
name
or
the
expectation
of
a
certain
name
or
number
to
exist
or
not
exist
and
how
it
was
used.
There
was
differences
of
opinion
there
and.
G
My
kind
of
comment
to
that
is
it:
it
doesn't
matter
what
the
message
is
so
I
mean.
Okay,
maybe
we
are
interrupting.
Message
is
more
important.
So,
for
example,
what
I'm
saying
here
in
the
first
point
is
that
the
PCC
DB
is
mod
of
PCC.
Lsp
DB
is
modified
by
any
trigger
it
could
be
PCI
update.
There
could
be
a
local
computation,
it
could
be
configuration
so,
regardless
of
what
triggered
it
like,
I,
don't
need
to
specify
it.
G
O
J
O
Going
to
add
some
recommendations
for
what
the
Association
DB
needs
to
look
like
and
proposed
terminology
for
tunnels,
and
things
like
that
I
would
encourage
you
to
look
at
the
T
tunnel.
Modeling's
work,
that's
been
done.
There's
a
fair
bit
of
work,
that's
gone
into
how
the
associations
are
specified
for
connections
belonging
to
various
Colonels
across
various
technology.
So
take
a
look
at
that
before
you
make
any
recommendations
in
this
topic.
Okay,.
G
P
The
first
worker
is
state
synchronization
between
states
for
PCs
and,
firstly,
let's
briefly
recap
what
has
been
done
in
these
drafts.
So
basically
it's
a
yeah
original
motivation
is
to
describe
the
procedure
for
the
sync
up
function
between
the
states
of
PCU
communications
and
the
objective
is
to
make
the
deployment
of
pce
in
a
more
resilient
way
and
we
can
solve
the
computation
and
loop
issues
for
the
dependent
task
computation
like
the
diversity.
P
So
since
the
present,
her
has
a
bad
memory
and
I'd
like
to
reintroduce
some
of
the
key
concept
in
the
draft,
so,
firstly,
with
har
get
all
set
up
at
this
obsession
between
the
state
for
PCs,
which
is
named
as
state
sync
session,
and
the
second
object
would
be
say
to
synchronize
the
RSP
state
between
this.
The
state
OPC
is
using
this
session
and
this
kind
of
say
synchronization
can
be
applied
to
in
both
directions
and
only
the
incremental
state
would
be
required
for
the
state
single
synchronization.
P
So
that
would
be
also
risky,
because
if
you
miss
one
of
them
or
you
duplicated
some
of
the
wrong
state
information,
that
would
be
a
kind
of
inconsistent
data
base
between
the
PC
NPCs.
So
we
also
introduced
the
master
and
the
slave
relationship
between
the
pcs.
So
for
multiple
errors,
piece
and
one
PC,
you
can
become
the
master
and
the
other
can
be
the
slave
and
all
the
RSPCA
is
a
kind
of
dedicated
to
the
master.
P
Pcs,
and
we
also
have
some
minor
contributed
have
to
describe,
say
the
use,
a
draft
database
portion
to
maintain
the
latest
change
and
also
we
need
to
maintain
the
earth
piece
data
from
multiple
sources,
and
we
use
an
example
to
describe
a
a
kind
of
how
this
state
of
synchronization
is
done.
So
we
assume
in
the
figure.
This
is
a
time
order
of
the
three
phases,
and
we
have
one
PCC
connecting
with
two
different
pcs,
and
there
is
a
kind
of
synchronization
request
is
between
the
two
pcs,
so
basically
are
as.
Q
P
P
So
this
problem
would
be
somehow
solved
by
using
the
states
of
synchronization
sessions
and,
finally,
the
PC
to
will
recorder
the
latest
RSP
state,
Korea
I,
received
from
the
PCE
one
before
receives
a
responding
message
from
PCC,
and
then
we
suppose
there
is
another
changer.
That
market
has
a
yellow
one
on
the
slides
and
PCC
also
need
to
trigger
the
changes
to
both
of
the
PCs
and
again
it
reach
the
PCE
number
one
first
or
so
out
great.
Thank
you.
P
Not
synchronized
the
state
but
to
give
him
the
mackensen
specified
in
this
document
we
allow
the
PC
number
2
to
figure
out,
which
is
the
latest
RSP
state
with
different
value
and
then
the
PC
number
to
understand.
He
need
to
ignore
the
the
Korean
War,
which
is
the
old
one.
So
and
finally,
there
is
a
kind
of
late
message
from
the
PCC
who,
if
you
see
number
two,
but
that
would
not
affect
any
changes
in
the
in
the
RSP
database.
P
P
We
okay,
we
looked
back
to
what
has
been
discussed
in
IETF
100
and
all
comments
has
been
clarified.
Oh,
how
the
PC
is
interacting
with
each
other,
and
we
have
addressed
this
kind
of
problem
by
aligned
with
other
state
for
PCs
and
hierarchical
PC
of
work.
So
this
would
be
an
important
mechanism
you
for
the
Mokpo
PCU
session.
So
we
request
to
the
working
group
to
adopt
this
draft
or,
and
it's
the
protein
to
the
adoption
queue.
Yes,.
D
I
personally
feel
that
this
is
important.
Some
some
of
our
recent
publishing
work
like
stateful,
HPC
and
actn
work
also
referred
to
this
document,
so
this
is,
if
I
have
feeling
that
has
not
been
seen
by
the
right
set
of
eyes
yet
so
should
we
ask
how
many
people
have
read
how
many
people
have
read
this
document?
Okay,
that's
good!
Should
we
check
the
interest
so
how
many
people
feel
that
this
work
is
whether
we
should
be
working
on
this
document
as
a
inside
a
working
group?
D
P
P
This
is
how
me
again-
and
this
time
I
marked
my
name
with
a
color,
so
the
next
presentation
is
about
the
extension
or
say
for
PCE
to
allow
optional
processing
of
the
PC
PC
objects.
So
basically
we
see
this
kind
of
discussion.
All
the
layers
are
talking
about
the
processing
of
flags
and
it's
kind
of
Seabury
is
a
kind
of
mist
when
we
migrated
from
City,
less
PC
to
late
fall
and
this
graph
to
specify
the
usage
of
PFLAG
and
I
flag
in
a
piece
of
message.
P
P
So
what
has
been
done
in
the
States
for
PCE
RFC,
82?
Thirty
one
is:
they
have
specified
the
state
for
PCE,
saying
that
the
usage
of
P&I
flag
must
be
set
to
zero
in
the
transmission
and
should
be
ignored
on
the
received.
So
the
problem
is
the
behavior
on
the
pni
flag.
In
this
object,
this
object
is
somehow
stated
but
see
in
the
other
object.
The
specified
so
the
specification
is
missing
and
we
finally
find
that
some
of
the
object
would
be
useful
if
they
can
have
the
pni
flag
set
or
not
set.
P
So
that's
the
draft
that
that's
about
the
draft,
so
we
clarify
how
the
pni
flaps
are
used
and
to
identify
the
optional
objects,
and
this
is
applicable
to
all
the
organs.
A
PCE
report
update
and
initiate
message
off,
semi
interstates
for
style.
So
we
also
update
what
would
be
done
if
there
are
some
unknown
objects
based
on
this
kind
of
flats.
P
So
so
far
we
have
categorized
all
the
PCE
into
two
different
one
can
proceed
as
a
peon
I
flag,
and
the
other
cannot
do
this.
So
it
would
be
necessary
to
introduce
a
new
flapper
to
indicate
the
capability
and
we
use
the
state
for
PC
capability.
Are
we
and
introduce
relaxed
format?
So
if
this
speed
is
set,
then
the
PC?
You
can
process
the
P&I
flag
according
to
this
craft,
and
if
this
is
now
set,
then
it
will
always
ignore
the
pni
flag.
P
N
P
P
D
Me
clarify
I
think
our
C
5
4
4
0.
When
we
set
the
P
flag
in
a
request
message,
we
basically
meant
that
PC
is
free
to
ignore
it,
but
it
is
allowed
to
process
it
as
well.
So
if
so,
ideally
what
we
usually
do
in
this
case,
if
the
path
is
found
meeting
that
constraint,
we
will
give
that
path.
But
if
we
are
not
able
to
meet
the
constraint,
then
we
will
ignore
the
object.
Yeah.
N
My
point
so
this
is
best
effort
in
my
mind,
I
don't
want
to
ignore
it.
I
want
to
still
consider
the
object
but
and
on
a
best-effort.
So
so
so,
let's
say
diversity
is.
You
know
computing
to
LSP,
two
paths
that
I
diverse
I
cannot
find
fully
diverse
paths.
They
could
be
neat
at
one
place,
I'm.
Okay
with
that,
is
it
possible.
It's.
D
H
My
question
is:
what
happens
is
that
we
see
while
doing
the
path
computation?
Okay,
though
there
could
be
two
situations,
it
is
possible
to
wanna,
say
inclusion
and
come
up
with
the
path
which
is
longer
in
delay
that
if
you
would
orbit
this
inclusion
allo
and
and
get
it
a
better
delay,
okay,
so
how
do
you
handle
the
compromise
between
two
optimizations?
One
is
following
not
mandatory
inclusions
and
another
provided
like
a
birthday
way.
I.
H
H
So
what
the
modeling
that
we
did
in
T
right
so
basically
the
way
we
addressed
this
problem,
which
is
a
big
problem
right
away,
the
resid
as
honoring
this
non-monetary
inclusion
as
a
base,
optimization,
okay,
the
same
as
optimization
like
this
delay,
and
we
provide
basically
the
weight
to
the
optimization.
Then
PC
knows
exactly
what
to
do.
What
would
be
the
best
optimization
overall.
H
P
A
P
So,
okay,
let's
move
back
to
slides
and
we
given
this
our
flap,
we
can
indicate
the
capability
of
the
PC
to
preceding
or
these
two
flags,
and
then
we
keep
the
definition
on
how
to
use
a
P
flag
and
I
flag
so
for
the
P
flag.
Firstly,
the
semantic
meaning
is
to
indicate
whether
the
object
so
with
this
flag
is
mandatory
if
the
flap
is
set
or
it's
just
the
option.
Now,
if
the
flap,
it's
not
set
so
different
message,
rule
requests.
P
The
flag
is
indicated
apart
by
different
entity
in
different
messages,
so
in
PCI
update
and
initiate
message
with
PC
is
entity
who
should
indicate
the
PFLAG
processing
and
PCE
report
manager.
This
this
black
rupee
indicate
by
by
PCC
and
P
equals
one
means
the
object
must
be
taking
into
account
and
P
equals.
Zero
means
the
object
that
can
be
ignored
in
the
discussion
from
the
etheric
or
the
best.
Air
filter
is
a
sub
case
for
P
equals
to
zero
and
four
regarding
was
a
flag.
P
P
So
if
there
is
unknown,
if
there
is
unknown
objects,
the
precept
should
reject
the
request,
always
I
error
message
and
if
everything
is
known,
that
would
be
a
kind
of
normal
procedure
processing.
So,
in
this
working
document
we
reuse
the
RSP
error
code
TR,
we
define
the
instead
for
PC
RFC
and
we
have
not
introduced
NC
new,
so
the
status
okay,
the
status
is,
we,
we
updated
the
the
text
by
adding
the
dedication
scenario
and
the
state.
P
B
A
Okay
thing
so,
but
this
one
I
think
we
don't.
You
have
some
problem
socialization
to
which,
even
that
on
I
think
it
is
linked
to
a
moral
issue
that
has
been
raised.
Monday
comin
from
turrican
ego
so
glad
to
see
some
parallel
discussion,
not
necessarily
linked
to
that
graph
itself,
but
maybe
also
to
diversity
graft,
which
is
also
tackling
some
interesting
mechanism
linked
to
tax
question.
So
please
make
sure
that
your
use
case
is
fine
and
we'll
take
that
to
the
list
of
progress.
Those
may
be
in
all
the
document
on
this
one.
Thank
you.
F
Chun
Li
from
Maui
my
topic
today
is
convey:
okay
thanks,
conveying
okay,
the
bank
vendor
specific
information
in
peace,
app
extension
for
a
state
for
PCE.
So
basically,
what
is
than
the
specification
information?
So
the
vendor
information
can
be
arbitrary
or
appropriate.
Information
such
as
than
the
specific
constraint
between
during
the
post
into
computation
and
the
price
number
uniquely
identifies
the
organization
responsible
for
the
definition
of
the
content
and
meaning
of
the
vendor
specific
information.
F
Support
in
his
PC
requires
request,
NPC
reply
message,
but
they,
but
this
RFC
didn't
mention
anything
of
than
the
information
in
for
state
of
all
piece
set.
So
that
is
what
we
do
in
this
draft.
We
stands
the
extension
for
supports,
supporting
the
state
for
PC
by
including
vendor
information
object
in
a
piece
app
report
and
PC
update
and
as
well
as
this
initiated
message
so
and
as
well.
The
TV
was
already
support.
F
F
A
No
one
so
question
of
the
usefulness
is
legitimate,
so
begin
to
have
some
feedback
on
that
correct,
I.
Think
so.
You're
using
some
more
different
mechanism
make
sure
it's
properly
supported
all
existing
device.
It
so
correctness
is
okay
for
me,
a
usefulness,
interesting
to
know
about
from
the
working
group.
So,
if
not
today
from
the
list,
that's
not
a
complicated
Pokemon,
so
ready!
Please
do.
We
will
consider
it's
useful
to
qualify
your
limitations
or
specification.
F
G
For
example,
not
sometimes
some
vendors,
you
know
they
they
kind
of
jump
ahead
of
the
of
what
the
working
group
is
doing,
and
sometimes
it's
useful
to
just
do
something
in
a
in
the
backward
compatible
way
to
signal
something-
and
there
is
also
some
things
that
you
know
may
be
very
specific
to
a
vendor
that
you
know
so
on
a
signal.
So
I
definitely
think
it's
useful
thanks.
R
Q
Triumphant
I'll
come
here
under
now
easier,
any
initiation
in
the
previous
document
that
limited
the
unity
of
the
theory,
harder
message.
You
know
you
have
Mason
and
his
tiara.
We
had
been
put
into
you
to
to
PMS
or
message
and
your
uncle
and
they
you
a
lot
of
Sri
message.
So
I
think
I
just
want
to
you
know
any
leave
shakes
me
in
the
kind
of
see
reverse
right,
yeah
yeah,
here
easy
Ernie,
you
something
you
now
see
7
for
17
to
restricted
the
tyranny
POV
you
in
other.
D
Let
me
try
if
I
understood
your
question
correctly.
The
7
for
7
0
said
that
this
theory
can
be
carried
in
any
object
that
allow
optional
TLV.
So
that's
why,
when
we
define
stateful
objects
like
SRP
and
LSP
object,
we
were
able
to
carry
the
TLV,
but
we
were
not
able
to
carry
the
of
judges.
Yes,
so
this
document
simply
says
that
you
can
also
carry
the
vendor.
Information
object,
not
just
the
TLB
okay.
Q
R
G
G
G
So
doing
kind
of
our
brainstorming,
we
realize
that
there
is
actually
two
use
cases
here,
so
one
the
first
use
case
is
when
you
have
a
single
optimization
objective,
and
this
was
the
solution
to
that
objective
is
multiple
ecmp
paths
and
the
second
use
case
is
when
you
have
multiple
optimization
objectives
and
you
want
ecmp
among
those
objectives.
So
you
want
a
signal
both
these
are
valid
use
cases
and,
in
fact,
they're
kind
of
orthogonal,
which
means
that
they
can
exist
independently
of
each
other.
So
here
I'll
give
an
example:
LS,
p1
and
LSB.
G
You
have
different
optimization
objectives
and
the
PCC
wants
to
have
a
CMP
among
LS,
p1
and
LSB,
so
LS
p1
gets
single
path,
X
and
speak
to
get
to
ecmp
paths,
y&z,
and
so,
as
a
result,
50%
of
the
traffic
goes
on
tax,
25,
1
y
and
25
on
that.
So
we'll
see
later
that
how
we
propose
to
represent
this
in
Pisa.
G
So
for
use
case
1,
which
is
the
single
optimization
objective,
yielding
multiple
paths.
We
we
basically
propose
three
ways
to
or
sir
two
ways
to
encode
this,
and
we
would
need
to
extend
Pisa
to
specify
the
maximum
number
of
paths
that
PCC
can
handle.
So,
for
example,
if
the
if
the
router
can
can
install
like
32
outgoing
paths
with
UC
and
PU,
and
it
would
just
need
to
signal
that
number
32
and
then
the
PC
would
reply
with
32
or
less
unequal
cost.
G
So
basically,
it
needs
to
carry
the
weight
and
needs
to
carry
the
upper
status
for
each
using
P
path
and
can
carry
optional
T
of
ease
and
then
option.
B
is
something
that
drew
suggested.
Basically,
this
is
to
reuse
something
from
T
to
MP,
and
here
we
we
can
use
already
existing
objects
to
separate
the
arrows.
G
So
there's
already
this,
this
type
of
us
to
our
us
object
that
is
actually
very
similar
to
what
we're
proposing.
So,
if
we
go
with
this
approach,
we
we
just
need
to
define
one
new
TLD
for
it
to
carry
the
weight,
and
basically
this
is
what
the
the
encoding
would
look
like.
So
you
see
arrow
1,
+,
0,
2
they're,
both
encoded
under
the
same
LSP
and
they're,
separated
by
these
endpoints
in
STL
a
subject,
and
basically
the
advantage
of
this
approach
of
option.
G
B
is
that
you
can
reuse
the
RB
n
F,
that's
already
there
in
the
RFC,
and
you
just
need
to
define
one
single
TOB
to
use
so
use
case.
2
is
a
is
a
completely
different
use
case.
So
it's
basically,
if
you
have,
if
the
PCC
knows
in
advance,
how
many
objectives
it
has.
So
if
it
wants
like
one
path
on
red
and
one
path
on
blue
and
once
ecmp
among
those
two
paths,
then
basically
the
LS.
G
The
PCC
will
allocate
different
piece
of
tunnel
and
tunnel
in
the
sense
of
my
previous
presentation
for
each
objective
and
it
would
create
some
new
Association
type.
For
example,
code
EMP
Association
by
buying
these
tunnels
together,
so
here
I
give
an
example
of
how
these
two
use
cases
are
orthogonal
and
they
exist
at
the
same
time.
So
I
already
gave
this
example.
I
mean
in
the
previous
slide,
so
I'm
just
gonna
give
the
piece
up
representation.
G
O
S
O
We
ever
extend
piece
up
to
signal
a
multipath.
Why
a
dag
you
could
reuse
the
same
sub
TLV
so
and
then
that
way
you
could
still
use
the
existing
BNF.
Today
we
do
have
precedence
for
a
CRO
very
inside
a
production
object
as
a
second
sub
object.
So
if
you
think
of
the
load
balance
specifier
as
an
english
action,
you
can
just
put
it
as
a
sub.
Clearly,
so
that's
one
comment
like:
can
you
go
further
up.
G
D
O
So
Liz,
you
have
two
options
that
one
is
today.
The
semantics
of
ero
would
say
that
you
process
the
first
one
and
if
you
see
more
than
one
year
or
you
ignore
the
rest
or
you
could
use
a
CRO
as
for
the
additional
ones
and
this,
and
that
fits
your
en
of
a
arrow
/
ero,
followed
by
C
arrows
and
all
of
them
carry
the
same.
I.
Think
our.
B
G
B
F
G
F
G
So,
basically,
there's
two
use
cases
right.
So
in
the
first
year's
case
you
don't
know
in
advance
how
many
ecmp
paths
you
will
get
you
just
simply
say:
I
can
install
up
to
thirty
two
paths:
okay
and
the
PC
then
may
give
you
one
path
or
it
may
give
you
two
paths
or
it
may
give
you
sixteen
paths
or
thirty,
two
paths
in
the
second
use
case.
That's
where
you
use
the
optimization
objective.
That's
why
I
use
the
Association
object
to
bind
the
different
Association
objectives.
G
N
To
me
the
second
example
option:
second
option
is
a
specific
case
of
you
know.
The
first
option
is
a
specific
case
of
the
second
meaning
one
LSP.
One
optimization
objective
is
a
specific
case
of
multiple
SPS
and
each
have
a
different,
optimization
objective.
Do
you
see
it
differently?
Yes,
okay.
I
am
not
clear
if
you
can
clarify
I.
G
N
N
H
G
Well,
you
can
say:
I
want
like
a
hundred
gig
path
and
with
certain
constraint,
like
you
know,
going
on
some
certain
set
of
links
and
I
also
want
like
another
hundred
gig
path
on
some
other
set
of
links,
and
that
is
use
case
too.
But
the
original
use
case
that
you
gave
is
a
basically
if
you,
if
you
request
that
you
want
100
gigs,
but
the
PC
cannot
give
you
one
path
that
satisfies
100
gigs.
So
it
would
give
you
two
paths
back
right
and
say
you
cmp
among
them.
Can.
H
H
G
R
So
to
adjust
audience
question,
the
binding
of
factor
T
know
is
coming
in
BGP
update
as
color
and
point.
So
this
is
how
you
know
which
still
associated
with
which
fat
here
we
are
mixing
a
bunch
of
use
cases
and
I
think
we
should
decouple
them.
One
of
them.
Sin
get
when
you
use
PC
to
compute
to
know
our
number
of
tools
associated
with
the
policy
rather
than
trying
to
provide
segmented.
G
T
So
my
name
is
human
wit
Kohli
from
Nokia
I'm
new
to
this
working
group.
Please
be
kind
as
the
French
say
in
Mexico
good,
so
this
graph
is
about
point-to-multipoint
policy,
so
just
to
give
an
idea
why
we
are
introducing
this
graph
and
I
think
on
the
unicast
side.
We
have
done
a
really
good
job
for
the
next
generation
networks
to
be
simplified
when
I
say,
simplify
I
mean
we
are
a
controller,
removing
some
of
the
MPLS
protocols,
but
multicast
is
kind
of
lagging
behind.
T
We
do
have
protocols
like
B,
which
uses
the
underlay
IGP,
but
when
it
comes
to
an
MPLS
transport,
as
of
now,
we
really
need
rsvp-te,
point-to-multipoint,
MLBPA,
etc,
etc.
So
just
to
be
in
par
with
unicorns.
We
try
to
introduce
this
draft
that
uses
a
controller,
very
thin
layer
of
communication
between
the
controller
and
the
PC
see
we
are
starting
with
Pisa.
Eventually
we
probably
going
to
move
on
to
other
protocol
su,
but
basically
the
controller
figures
out
the
tree
using
bgp
LS.
T
T
So
to
do
this,
there
is
a
couple
of
objects
or
segments
that
we
are
introducing
in
spring,
and
one
of
them
is
the
point
to
multi-point
segment.
So,
basically
the
simplest
way
to
explain
the
point
to
multi-point
segment.
It's
it's
a
container
that
contains
the
route
and
a
set
of
Leafs
and
it
needs
to
be
identified.
T
A
set
of
parameters
which
you
can
read
the
draft
and
figure
that
out
how
the
root
and
the
leaf
are
discovered.
There
is
nothing
new
we
can
discover
them
via
NP
BGP
procedures
was
at
the
entry
and
VPN
procedures,
the
RFC
6514
6513.
All
those
discoveries
can
become
indicated
up
to
the
PCE.
We
are
the
mechanism
that
you
guys
already
have
and
invented
previously,
as
you
will
read
it
in
the
draft
or
later
on
in
these
slides.
So
there
is
nothing
new
here.
T
The
only
thing
you
gotta
keep
in
mind
is
you
need
to
understand
what
is
a
point-to-multipoint
segment,
so
the
next
part
is
as
I
mentioned.
Point-To-Multipoint
segments
actually
contains
the
rule
and
a
set
of
leaves
obviously
between
the
root
and
the
Leafs.
There's
a
replication
points
and
that's
what
the
replication
segment
comes
in
in
a
simplest
way
to
think
about
sitting
a
replication
segment.
Think
of
it
as
a
forwarding
instruction
at
each
new
at
each.
T
No
that's
a
simplest
way
of
putting
it
incoming
seed
incoming
label
for
a
segment
label
for
a
replication
segment
that
identifies
the
resource,
and
then
you
have
the
outgoing
interfaces
with
outgoing
labels
that
will
connect
you
to
the
next
replication
segment.
That's
it
that's
all.
One
thing
you
need
to
keep
in
mind
is
pure
replication
segments.
They
can
be
disjoint
via
a
unicast
SR
segment.
If
you
will,
in
that
case,
you
guys
have
already
invented
the
seedless.
T
We
use
a
Sibley's
the
the
unique,
as
is
our
policy,
to
connect
to
replication
segments
together,
and
our
important
thing
to
keep
in
mind
here
is
replication
segments.
They
cannot
be
stacked.
Why
is
that?
The
reason
for
that
is
when
you
get
to
a
replication
point,
that
you
have
multiple
OFS,
outgoing
or
if's.
If
you
start
stacking
stacking
segments,
multicast
segments
where
you
really
cannot
go
through
every
single
one
of
those
labels
to
figure
out,
which
is
the
leaf
disconnected
on
beach
of'
in
malta
in
in
multicast.
T
You
know,
if
you
look
at
only
the
top
of
the
the
list,
if
you
will,
it
will
not
give
you
the
entire
picture
on
that
replication
segment
for
all
outgoing
interfaces.
So
this
is
why,
when
it
comes
to
the
replication
segment,
there
is
only
a
single
label
that
single
label
on
that
note
will
identify
the
outgoing
interfaces
and
the
labels
or
the
seats
that
are
being
used
to
the
next
one
again.
This
is
only
true
for
replication
segments.
T
You
can
connect
to
replication
segments
via
a
seedless
and
unique
ass
as
our
policy,
so
just
going
through
the
data
path
very
quickly.
The
source
sends
a
payload
that
payload
is
the
IP
gets
through
outer
a
router.
A
has
a
replication
policy
that
says
your
to
get
to
router,
D
and
E,
which
are
the
Leafs.
Your
next
hop
is
a
replication
segment
c.
B.Push
c
payload
is
on
the
bottom.
We
send
it.
The
the
packet
to
see
C
looks
at
the
see
the
label.
T
C
identifies
the
replication
policy
or
the
replication
segment
segment
looks
at
their
outgoing
instructions.
The
outgoing
OFS
outgoing
interfaces
figures
out.
What
is
the
label
that
it
needs
to
push
on
each
interface
to
D?
It
pushes
D
to
e,
it
pushes
E
and
it
forwards
the
packet
through
the
according
interface
to
those
routers.
The
next
one
is
fast
reroute.
The
PC
can
camp
compute
a
fast
reroute
that
bypasses
the
link
between
a
and
C.
That's
via
router
G
that
fast
reroute
could
be
TI
l
fa.
It
could
be
RSVP
T.
T
It
could
be
anything
it's
beyond
the
scope
of
these
draft,
but
you
have
the
fast
reroute
that
is
sitting
on
the
data
path
and
you
can
achieve
very
fast,
fast
readouts.
Here's
an
example
when
there
is
a
node
that
does
not
support
replication
segment
on
the
on
the
network.
That's
not
B
and
F
as
an
example
here.
So
in
this
case
is
because
PC
that
note
B
does
not
support
replication
segment.
T
They
actually
push
on
a
label
stack
saying
that
C,
which
is
the
next
replication
segment,
is
at
the
bottom
of
the
stack
pushes
B
for
unique
hash,
unique
SSR,
C
and
forwards
that
packet
through
B
to
C,
B
pops,
a
label
C
gets
the
label
label
C
figures
out.
What
is
the
replication
segment
and
forwarding
information
on
C
and
forwards
the
packet?
That
is
what,
as
it
was
in
the
previous
slide
going
forward.
You
can
have
all
the
good
stuff
that
you
guys
have
in
unicast,
canada,
tad,
let's
speed,
redundancy,
fast,
reroute,
etc,
etc.
T
We
are
using
all
the
draft
all
the
tools
that
you
guys
have
already
invented.
We
are
not
trying
to
come
up
with
new
mechanism.
I
think
you
guys
have
done
a
great
job
previously,
so
we're
just
using
all
those
tools
all
over
again.
Here's
that
high-level
bird's-eye
view
of
how
the
objects
loop
again.
Some
of
the
objects
that
are
already
there,
Canada
pads,
you
know
nothing
new.
The
only
part
that
I
as
I
explained
previously
is
you
need
to
understand
the
point
to
multi-point
policy
and
the
replication
policy.
T
The
replication
policy
can
contain
multiple
canada
paths
and
this
candidate
paths
can
be
used
as
LSP
redundancy.
They
could
be
end
to
end
or
aka
global
optimization,
if
you
will
throughout
each
one
of
these
candidate
paths
and
for
the
sake
of
time,
I'm
not
going
to
go
through
the
details
of
what
are
the
bits
and
bytes
in
this
graph.
You
can
go
with
the
draft
one
of
the
things
that
I
want
to
specify.
It
is
that
the
RFC's
that
I
put
in
there,
the
first
you
RFC's,
they
probably
don't
make
sense
to
anybody.
T
There
should
be
82,
31
I,
believe
it's
all
right,
yes,
and
the
second
one
is
86
23,
which
is
the
new
RFC
that
you
just
gave
out.
So
those
are
some
of
the
current
other
sins
that
we
are
using.
We
did
try
to
change
and
introduce
some
very
minor
objects
which
were
necessary
as
an
example.
We
are
using
the
CCI
object,
but
currently
in
the
CCI
object,
you
can
send
down
the
single
label
going
forward.
T
We
are
using
or
CCI
objects
on
multiple
labels
for
the
seed
list
for
the
outgoing
interfaces,
because
there
could
be
multiple
outgoing
interfaces.
Hopefully
all
business
stuff
are
explained
in
the
draft.
If
they
are
not
clear,
please
send
the
email
under
a
mailing
list
and
the
workflow
I'm
not
going
to
go
through
it
too
much.
For
the
sake
of
time,
I
want
to
leave
the
the
rest
of
the
time
open
for
any
question
comments
that
are
there.
T
T
N
T
It
right
there
and
that
that's
a
very
good
question
and
and
and
I'm
I
know
you
we
discussed
this,
but
for
the
sake
of
working
group
I
want
to
bring
up.
There
are
some
differences
between
multicast
and
unicast.
One
of
those
differences
is
if
everybody
here
remembers
the
old
way
of
way.
We
were
doing
the
unique
acid
for
VP
RN.
Even
there
was
a
subnet
that
was
in
a
verse
that
was
connected
to
a
tunnel,
so
a
transport
tunnel
back
in
the
days
was
really
identifying
a
subnet
itself.
T
We
went
away
on
the
unique
aside
from
this
mechanism.
We
actually
have
now
a
VC
label
or
a
rough
label
whatever
you
guys
want
to
call
it
that
the
transport
tunnel
gets
you
to
the
end,
and
then
you
pop
the
transport
tunnel,
you
look
at
the
valve
label
or
the
the
service
label
gets
you
to
the
to
the
service,
and
then
you
do
a
lookup
in
that
v
RF
and
you
figure
out.
Where
is
your
IP
going
on
multicast
on
fortune
of
behaviors
the
stock
in
the
18th
centuries?
T
So
it
a
transport
tunnel,
a
point-to-multipoint
LSP
can
identify
a
specific
SG.
So
you're
absolutely
right.
In
some
cases
we
could
use
replication
and
reuse
it,
but
for
the
cases
that
that
point
to
multi-point
LSP
identifies
a
specific
SG
within
the
worth,
we
need
to
have
a
unique
identification
of
that
point.
Moil
multi-point
lsv
throughout
the
entire
path.
That's
it!
Yes,
we
will
work
on
it,
try
to
figure
out
whether
we
can
make
it
reusable.
Thank
you,
woman.
Thank
you
very
much.
T
O
O
Yes,
it
could
be
feasible
that
two
different
or
multiple,
yes,
it
will
be
a
sub.
All
is
this
code
in
a
certain
part
of
the
topology
share,
the
same,
interrupt
location
state
and
therefore
you,
you
know,
might
the
replication
policy,
but
they
might
diverge.
You
know
later
down
the
tree
and
so
to
do
what
you're
suggesting
you
probably
have
a
carrier
tree
or
it
can
be
policy
context
in
the
data
plane
engineer.
So
my
point
is,
it
is
feasible,
but
it
comes
at
a
cost.
O
A
S
Model
today,
I'm
going
to
pretend
that
you
snitching
to
PC
for
SR
pass
protection.
So,
first
of
introduction,
we
know
that
some
of
a
critical
real-time
at
the
live
traffic
may
be
transported
by
SR
pass.
For
example,
we
may
have
a
surpass
from
where
Punk
to
New
York
Stock
Exchange,
which
is
to
be
used
to
send
requests
for
buying
or
selling
stocks.
S
We
may
also
have
SR
past
connection
between
operational
role
and
polkas
office,
which
is
used
by
doctors
to
do
remote
surgery.
We
may
also
have
some
kind
of
as
a
pass
to
support
life
Ethan's
life
broadcast.
So
for
all
these
esta
pass,
we
need
a
fast
protections,
so
without
fast
essa
protection
we
may
have
big
loss,
for
example,
for
real-time
stock
exchange
or
trade.
S
If
we
have
broken
answer
a
pass,
then
we
may
lost
lots
of
money,
for
example,
in
the
peer
market,
the
stock
away
going
down
very
fast,
so
bank
may
send
a
request
to
sell
quick
as
quick
as
possible
to
sell
stock
before
the
mock
market
going
down
going
down.
So
in
so
without
fast
protection,
if
SF
has
broken
and
then
those
requests
were
not
executed
faster,
so
we
we
almost
not
so
money
so
for
what
this
case
we
will
have
big
loss
result
faster
protection
for
SR
pass
so
night
speech.
S
S
S
Provide
fast
protections
for
the
ingress
note
of
SR
pass
so
next
page,
so
the
extension
for
PCC
for
support
has
equal
protection
is
a
very
good
feat,
because
right
now
we
have
PC
extensions
for
essa
passport
on.
Oh,
so
he's
very
literate
to
extend
the
PCE
to
support
SR
pass
English
protections.
So
next
page.
S
S
Primary
universe,
node
address
in
some
case,
and
also
the
description
of
the
traffic
which
is
transported
by
the
SA
pass.
In
addition,
so
essa
pass
may
transport
some
service,
so
those
IDE
of
that
surface,
which
is
transported
by
the
s
epatha,
may
also
need
to
send
solitude
the
backup
equals
note.
So
for
this.
S
For
some
of
this
information,
the
encoding
we
already
have,
for
example,
for
the
SR
passed
primary
of
us,
because
we
have
extensions
for
the
SR
sake
policy
in
that
exchanger,
where
we
already
have
extensions
for
support
as
a
passing
coding
so
for
pike
harbors,
a
passing
coding,
which
we
can
reuse
those
encoding
for
SR
primary
path.
So
for
the
rest
of
them,
we
may
have
some
kind
of
simple
extensions
to
PCE
to
support
backup
of
support
image,
ingress
protection.
S
D
S
D
So
I
think
III
raise
some
questions
on
the
list
as
well.
So
I
hope,
like
you,
can
answer
them
a
little
bit
more
clearly
and
explain
why
we
would
want
the
backup
ingress
to
do
this
protection
and
when
the
source
is
anyway
doing
the
affection
the.
Why
should
backup
ingress,
detect
and
I
got
the
not
a
direct
answer,
but
I
thought
the
your
answer
would
be
that
you
don't
want
to
make
the
backup
path
active
and
only
when
the
fail
you
detect
the
failure.
At
that
time,
you
make
the
backup
path
active.
D
So
these
kind
of
use
case,
if
you
think
it
exists,
maybe
we
can
generalize
them
and
we
already
do
flowspec
for
identifying
of
what
flow
to
go
into
a
particular
particular
tunnel.
So
if
you
have
a
particular
use
case,
maybe
we
can
think
of
doing
solving
this
in
a
more
general
way,
rather
than
just
as
a
backup
ingress.
That's
just
my
personal.
Without
my
chair
hat
on
opinion.
Thank
you.
U
S
U
D
For
let
me
try
to
answer
from
what
I
gather
the
differences
with
respect
to
detection
like
primary,
getting
failed
that
detection.
We
never
talked
about
it.
An
association
draft
and
this
document
somehow
says
that
it's
the
job
of
the
backup
ingress
to
somehow
detect
the
primary,
has
failed
and
now
switch
the
traffic.
That's
what
I
gather
I'm
not
wasn't
sure,
but
just
we
can
move
the
conversation
along.
W
To
give
a
present
for
my
colleague
on
yeah,
all
of
them
are
not
able
to
join
a
session
and
therefore
this
contribute
for
this
project
in
so
it's
about
as
beautiful
she
see
for
assigning
to
demand
for
the
over
real.
There
are
some
more
passing
has
been
done.
They
are
being
dying,
I
have
in
different
working
group,
so
I
just
what
I
could
to
give
an
overview
of
this,
but
I
want
to
redesign
that's
try
to
save
time
and
then
next,
okay,.
W
Okay
for
this
so
I,
do
it
just
give
some
analysis
about
what
we
could
do?
The
first
amazing
is
a
very
direct
and
a
possible
association
and
signalizing
is
inter
demand,
pass
corrilesa
or
we
can
call
it
association.
Now.
This
is
a
different
kind
of
a
location
of
a
past
segmental.
It
could
be
located
by
the
past
prior
in
the
PCE
or
could
be
by
child
oficial.
She
said,
and
the
english
instructor
we
consider
two
kinds
of
scenario:
the
facility
in
Terraria
ii
know,
I
think
I
yes,
it
could
be
pieces
in
unity
throughout.
W
It
could
be
stated
for
Casilla
hierarchic
a
PCE
scenario
next,
for
this
slides
away,
just
gave
us
some
analysis
of
the
case.
That
is
the
inter
area
case.
How
it
was
this.
This
I
was
kidding
in
the
story
area
together.
We
just
try
to
use
an
association
work.
What
was
this
sorry?
I
was
a
few
second
in
the
together
to
a
former
and
and
I
was
he
passed
and
next
day
some
into
a
solution,
and
the
in
this
case
is
a
different
from
the
previous
line.
W
In
this
case,
I
would
also
try
to
propose
to
use
a
stitching
liberals
who
try
to
form
a
end-to-end
pass
with
the
included
into
a
s,
pass
segment
and
the
int
sorry
intro,
yes
and
the
inter.
Yes,
it's
a
consisted
of
a
five
five
of
a
segment
in
this
scenario,
and
this
is
a
this
draft-
raises
a
some
give
some
over
real
over
the
absolution
that
we
want
to
do
by
the
way.
Currently,
we
did
not
give
our
extension
that
may
be
needed.
They
also
another
companion.
W
D
A
similar
question
was
asked
in
the
SR
when
you
presented
this
in
the
MPLS
working
group
that
why
the
path
segment
is
being
used
for
stitching.
So
I.
Don't
want
you
to
answer
right
now,
but
this
is
something
that
I
hope
you
can
discuss
in
spring
or
MPLS.
Whichever
is
the
right
place,
and
then
we
can
think
about
doing
this
work
NPC,
because
our
main
concern
is
whether
this
segment
should
be
used
for
stitching
itself.
That
needs
to
be
clarified
before
and.
W
X
Good
morning,
everyone,
my
name,
is
Ching
I'm
here
to
discuss
RGB
extension
for
PC
EP
security
capability
support
in
the
PC
discovered
this
job
is
Aris
are
working
to
adapt
and
the
basic
idea
is
so
now
we
have
dcpl
suppose
we
have
a
TRS
supporter.
We
have
many
different
security
capability:
how
to
reproduce
these
different
ability
to
allow
ECE
to
company
with
his
PCC.
So
a
quick
update.
Actually
this
job
actually
is
a
DB
extension.
X
So
that's
why
it
discussed
in
the
Arisaka
kukaku,
but
it
also
references
our
GP
extension
for
PC,
a
discovery
that
we
developed
by
the
PCE
working
cool.
So
for
for
this
job.
Actually
we
discussed
in
a
Paris
our
working
group
when,
during
the
carpet
option
actually
summer
situations
is
to
consider
to
add
the
tier
theory
for
the
key
ID
and
a
key
name,
but
to
support
that.
X
Actually,
we
required
to
update
the
obviously
fake
50,
88
and
fifty
eighty
nine,
and
so
that's
actually
we
make
this
change
actually
and
this
get
adopted
in
last
night
in
November
ITA
meeting
in
Tucson
18.
So
here
what
we
like
arrays
is.
We
would
really
want
to
get
a
feedback
from
PC
would
go.
It
was
when
we
first
bring
up
this
draft
that
didn't
get
a
lot
of
Attraction,
so
we
do
to
the
concerns
in
iOS
are,
but
that
we
didn't
get
feedback
from
the
PC
so
for
the
channel
will
make.
X
Actually
we
require
to
update
of
C
58
89
to
allow
the
advertisement
that
Tara's
teacher
several
theory
related
to
the
security
parameter.
So
we
have
some
feedback
from
Aris
our
Minister
some.
Some
people
really
some
concern
whether
we
should
advertise
this
and
long
RG
period.
Information
is
may
impact
the
performance,
but
the
the
benefited
to
allow
advertisers
kinda
capability
is
actually
you
can
really.
You
know
piggyback
some
security
community
to
simplify
the
operational.
Also
we
from
also
perspective.
X
So
we
we.
We
are
flexible
for
these
issues
and
we
want
to
hear
your
feedback
on
this
and
in
a
minute
I'm.
Actually,
you
can
come
out
and
right
now
we
can,
you
can
come
on.
It
is
so
wait.
We
also
can
follow
up
how
to
deal
with
these
issues.
So
for
next
step,
we
will
look
at
all
your
comments
and
and
also
to
chair.
We
want
to
make
sure
whether
these
right,
a
placing
errors,
are
or
in
PC.
We
can
go
yeah
comments.
A
Thank
you
so
so
far
working
on
the
external
GPS,
so
you
make
sense,
show
me
from
Eastern
with
the
PC
working
group
with
also
for
you
event.
So
it's
thank
for
that
back
to
the
issue
quickly,
because
we
won't
have
much
time
to
go
over
them.
But
second
issue
is
for
me
under
a
set
of
rate
equation.
Either
graph
remains
in
the
working
group.
A
The
way
women
are
progressing
is
up
to
the
reserve
rule
punchers
and
for
the
first
question,
as
far
as
I
remember,
the
limitation
about
further
extension
was
imposed
by
the
IGP,
your
eighty
working
groups
at
that
time.
So
if
now,
okay,
to
move
this
limitation,
a
scene
from
the
PC
standpoint,
it's
pero
to
have
this
restriction
removed
Thanks.