►
From YouTube: IETF100-PALS-20171113-1740
Description
PALS meeting session at IETF100
2017/11/13 1740
https://datatracker.ietf.org/meeting/100/proceedings/
B
B
B
D
B
Dave
just
pointed
out
that
we
need
to
send
out
a
liaison
about
the
control,
would
draft
to
some
other
stos,
so
we'll
just
very
quickly
go
through
the
list
of
organizations
that
we
plan
to
send
the
liaison
to
okay.
So
we've
had
some
new
RFC's
since
the
last
eight
F
RSC
8220
on
him
over
virtual
private
land
services,
a
VPLS
and
RC
282
37
on
MPLS
LSP
pseudo
wire
status,
refresh
reduction,
we've
had
16
RCS,
published
in
the
last
two
years,
so
we've
been
quite
productive.
B
B
B
And
we
I'm
sorry
I
was
talking
about
the
wrong
draft.
That's
on
the
next
one,
this
one,
this
one,
that's
on
the
RFC
editors
queue
is
waiting
for
the
next
draft.
I
were
talking
about
so
that.
So
this
is
the
draft
that
just
went
through
iesg
processing
and
it
was
just
revised
today
based
upon
the
the
security
ad
review
and
the
security
Directorate
reviews
the
the
issue-
and
this
is
something
that
I'd
like
to
make
sure
you're
all
aware
of,
because
it
affects
work
in
this
group
and
work
in
the
MPLS
working
group.
B
The
issue
was
at
LDP,
which
of
course,
we
use
heavily
for
pseudo
wires
is,
as
the
control
plane
currently
relies
on
md5
for
cryptographic,
security
of
its
messages,
but,
as
we
all
know,
md5
no
longer
meets
current
security
requirements
in
the
IETF.
So
this
issue
really
needs
to
be
solved
in
the
MPLS
working
group,
not
in
this
work,
because
it's
the
MPLS
working
group
that
has
ownership
of
LDP
and
it
affects
all
dress,
use
LVP,
not
just
the,
but
any
future
work
that
uses
LDP
will
run
up
against
this.
B
So
the
mpls
and
pals
working
group,
chairs
and
deborah
who
just
walked
in
hi
Deborah
are
working
together
on
a
solution
which
would
be
a
new
draft
in
the
MPLS
working
group
on
LTP
security
to
replace
md5
there's
been
a
discussion
currently
ongoing
on
the
PALS
email
list.
So
so
you
may
have
seen
it
already
if
you've
been
following
the
email
either
on
this
list
or
on
the
MPLS
list.
E
So
this
is
not
just
a
penalty
people
of
them.
This
is
actually
a
problem
for
all
of
the
TCP
based
routing
protocols
and
I
think
they
won't.
They
have
two
problems.
One
is
that
there
isn't
a
default
widely
agreed
a
crypto
function.
Secondly,
there's
no
negotiation
function
that
anyone
has
actually
built.
What
will
TC
Bao?
Let
you
do
is
to
say
what
the
encryption
function
you're
going
to
use.
E
B
E
So
before
I
start,
who
didn't
hear
the
plea,
the
previous
version
of
this
presentation
last
time,
because
if
you
want
okay,
all
right
so
I
will
go
through
the
the
introduction
slides
which
briefly
introduce
the
problem.
Okay,
next
slide,
please
Andy
so
story.
So
far,
a
number
of
operators
have
experienced
problems
with
miss
ordering
in
their
network
of
Ethernet
packets,
and
this
was
attributed
to
some
of
the
Pease,
not
using
the
pseudo
wire
control
word
for
Ethernet
and
the
I
Triple
E.
Now,
quite
legitimately,
although
I
you
issuing.
E
Mac
addresses
that
start
with
0,
4
or
0
6,
which
look
a
bit
like
IP
addresses,
IP
header
starts
now
I.
Suppose
I
should
have
added
something
that
hat
said
last
time,
which
is
that
they
were.
There
are
other
criteria
as
well.
That
can
cause
that
to
happen
the
one.
What
this
is,
the
one
we're
all
familiar
with,
that
there
are
more
subtle
reasons
why
it
can
happen.
E
They
say
I
see
a
four,
so
it
must
be
an
ipv4
packet
and
then
they
go
and
do
the
standard
5
topple
hash
calculation,
which
is
a
bit
of
a
problem
if
it's
some
random
bits
inside
an
Ethernet
header
and
the
packet
goes
off
some
way
other
than
the
way
they
expected
it
to
so
the
result
is
a
miss
sequence.
The
result
is
a
lot
of
time
and
money
spent
by
the
operator
community
trying
to
figure
out.
Why
some
custom
wire
customer
has
complained
that
the
packets
aren't
arriving
in
order.
E
E
We've
got
a
an
adopted
draft
and
you
will
talk
about
who
we're
going
to
liaise
it.
We
should
have
liaised
it
to
some
other
third
parties
now
it
turns
out
and
that
some
vendors
are
in
some
circumstances,
using
zero,
X,
zero
nibble
zero.
The
start
of
the
control
word
as
a
as
a
hint
that
what
follows
is
an
Ethernet
pseudo
wire
packet,
so
they
skip
I,
presume,
14,
bytes,
but
I,
don't
know
the
exact
number
and
they
say
we'll
do
the
five
couple
hash
on
the
IP
packet
that
we
assume
follows.
E
We
didn't
know
about
this
at
the
last
IDs,
so
this
is
an
extra
piece
of
material
next
slide,
please.
So
what
I
would
like
to
do
at
this
meeting?
The
working
group
agrees
is
for
us
to
look
at
the
some
critical
pieces
of
text
that
are
in
the
document,
the
core
of
our
recommendation
for
two
scenario:
the
two
scenarios
and
for
us
to
agree.
E
E
What
the
the
other,
what
some
vendors
have
got,
a
switch
that
they
can
turn
on
when
you
turn
the
switch
on
it's
a
it
looks.
It
looks
down
the
MPLS
label
stack
just
like
we
do
at
the
moment,
and
it
says
if
there's
a
four
or
a
six,
then
we'll
do
the
standard
four
or
six
thing
it,
but
if
there's
a
zero,
it
must
be
an
Ethernet
pseudo
wire,
so
skip
14
bytes
on
we
should
I
believe
is
the
the
right
number.
E
E
B
E
And
it
could
be
that
there's
a
legitimate
packet
be
following
the
MPLS
label
stack
that
starts
with
zero
zero.
That
could
be
anything
remember,
there's
no
definition
of
what
that
foot.
You
know
what
happens
on
that.
First.
First
nibble
right,
you're
supposed
to
just
operate
on
the
MPLS
label,
stack
according
to
a
strict
interpretation
of
the
MPLS
architecture.
These
other
tricks
and
techniques
are
tactical
things
that
people
do
and
what
happens
is
they
do
them
and
it
doesn't
always
work
out
as
they
expect,
and
some
operators
and
Ignace
has
spent
a
lot
of
time.
E
Talking
to
operators
are
complaining
that
these
things
get
turned
on
in
their
network,
presumably
because
people
don't
really
understand
the
problem
or
because
someone
who
did
understand
the
problem-
I
understood
it
correctly
and
there
was
only
Ethernet
in
their
track
when
they
turned
it
on.
But
you
know
time
moves
on
the
use
of
the
network
changes.
People
had
other
things,
people
had
other
protocols
and
then
you
get
the
miss
ordering,
and
then
they
spend
hours
trying
to
sort
of
go
through
with
Wireshark
or
whatever
the
tool
they're
using
is
and
find
out.
E
What
on
earth
is
happening,
so
it's
that
community
that
have
asked
us
to
publish
material
that
helps
them
resolve
the
problem
and
and
the
only
way
we
can
resolve
the
problem
is
to
provide
advice
or
right,
but
at
least
they
can
point
to
an
RFC
and
say
it
says:
look
it
says
here,
don't
do
this
or
they
can
specify
a
dierence
to
the
RFC
in
there
sort
of
request
for
service
and
stuff.
So,
but
we
it's
it's
we're
not
doing
this
because
we
think
it's
a
random.
E
You
know
a
good
thing
to
do,
even
if
it
is
a
good
thing
to
do
we're
doing
it,
because
people
have
come
to
us
and
say
this
is
happening.
It's
costing
us
time
and
money
managing
our
networks.
Please
could
you
figure
out
what's
going
on
and
fix
it
for
us
and
I'm
sure
if
Ignace
wants
to
go
to
the
microphone,
he
can
explain
because
he's
been
the
person
who's
been
talking
really
to
the
operator
community.
C
It
was
ton
of
thickness
booked
on
his
Equinix
indeed,
and
some
statistics
so
since
last
meeting
I
was
speaking
about
this,
at
least
at
nano
gripe
euro
X,
some
other
smaller
events
and
statistics
that
I
gathered
unguarded
I
see
that
I
had
44
different
discussions
with
different
operators.
Two
out
of
those
were
aware
of
this
problem
to
a
bigger
portion
of
them.
This
does
not
apply.
C
The
basically
the
case
has
where
this
problem
surfaces
they
show
up
regularly
on
on
on
various
operator
groups,
and
certainly
it's
a
question
of
Education
saying
that,
while
the
architecture
allows
you
to
do
this,
this
is
not
a
right
thing
to
do,
because
you
will
run
into
the
problems,
and
here
is
the
reference.
If
you
went
for
more
details,
ok.
E
So
the
problems
that
that
I
want
to
address
I'd
like
to
address
at
this
meeting
are
for
us
to
look
at
the
additional
that
the
text
that
we've
put
in
this
draft
and
what
we'd
like
to
do,
is
to
try
and
agree
the
text
on
those
two
recommendations.
So
I'm
going
to
explain
the
material
I've
got
in
here.
The
last
two
slides
are
a
repeat
of
the
key
text
and
I'd
like
to
do
live
editing
in
this
meeting.
E
E
The
first
is
on
the
use
of
the
control
word,
so
in
section
4,
which
is
the
recommendations
text
I've
said
that
this
document
updates
RFC
four
four,
four,
eight
to
state
that,
where
both
the
ingress
and
the
egress
pe
support,
the
ethernet
suitor
wire
control,
word,
then
the
control
word
must
be
used.
So
this
tries
to
steer
people
away
from
ever
having
this.
This
problem.
E
F
E
F
E
E
So
also
in
the
operational
considerations
section
and
protection
I'm
proposing
is
to
remove
this
problem
in
the
long
term
and
hence
to
reduce
the
operational
costs
of
investigating
problems
associated
with
the
incorrect
forwarding
of
Ethernet
packets
over
suitor
wires,
not
using
the
control
word.
It
is
recommended
that
equipment
that
does
not
support
the
control
word
be
phased
out
of
operational
use.
We're
going
to
discuss
this
at
the
end,
because
I
know
that
Himanshu
has
got
some
opinion
on
that
all
right.
So
next
next
slide
please.
So
we
met
that
text
before.
E
So
dpi
beyond
the
deep
packet
inspection
beyond
the
control
word
so
section
5
explains
that
there
are
two
methods
that
want
to
provide
safe
methods
of
doing
pseudo,
wire,
ecmp
and
they're,
the
they're,
the
fat
label,
which
is
where
a
label
that
goes
in
the
suit
in
the
stack
right
at
the
bottom
of
stack.
We
where
people
can
provide
entropy
alright,
and
this
predates
the
later
solution,
which
is
that
we
use
the
an
LS
P
entropy
label
in
the
stack
either.
Those
are
acceptable
approaches.
One
is
entirely
within
the
domain
of
the
pseudo
wire
application.
E
There
is
a
caveat,
however:
RFC
67
90
is
soft
on
how
ecmp
is
carried
out
when
the
e
Li
is
present.
If
you
look
at
the
text,
it
says
if
a
transit
LSR
recognizes
the
LSE,
the
LSI,
rather
it
may
choose
to
load
balance
solely
on
the
following
label:
the
entropy
label.
Now
this
is
a
wider
problem
because
it
actually
upsets
it
upsets
the
synonymous
labor
work
and
it
upsets
the
any
any
attempt
we
may
make
to
do.
E
Io
am
in
an
MPLS
context,
so
anything
that
we're
doing
in
the
OEM
world
that
wishes
to
change.
Thus,
the
the
stack
should
that
those
changes
should
be
neutralized
by
the
presence
of
the
e
Li,
but
because
of
this
wording
it
is
not
absolutely
the
case
that
it
will
be
now.
It's
not
the
work
of
this
job
of
this
working
group
to
change
the
Li
definition,
but
I
think
that
we
should
take
work
to
MPLS
and
request
the
tightening
of
those
criteria
such
that.
E
If
anyone
follows
this
new
RFC,
the
our
se
foo
that
if
you
see
the
e
Li,
then
you
load
balance
on
the
e
Li
and
or
the
entropy
label,
and
you
don't
and
and
you're
very
careful
about
anything
else.
You
take
into
account
and
I
think
that
would
allow
us
to
solve
a
problem.
We
got
on
synonymous
labels
and
another
problem.
E
E
E
Rfc
60
790
are
the
e
Li
and
both
ingress
and
egress
pease
support
the
fat
pseudo
area,
RFC,
6391
and
ecmp
of
the
ethernet
pseudo
wire
traffic
is
required.
Then
one
of
these
methods
and
we
can
discuss
whether
it's
a
should
or
a
must.
But
the
point
of
me,
the
the
the
thing
here
is
to
say
that
if
the
two,
if
the
two
of
them
support
fact
or
one
of
them-
supports
the
Li,
that
is
the
preferred
method
and
that's
what
we
should
be
should
be
doing.
E
There's
an
operational
considerations
section
as
well
as
well
to
remove
any
ecmp
passed
based
on
dpi
beyond
the
pseudo
wire
control
word.
It
is
recommended
that
pease
supporting
Ethernet
pseudo
wires,
implement
RFC,
6391
and
RFC
60
790
I've
said
both
because
I
want
to
make
sure
that
we
can
do
interoperability
and
there
will
be
some
kids
out
there.
That
can
do
one
some
kid
out
there
that
can
do
the
other,
and
so
this
is
they're.
E
Actually,
the
minimum
changes
to
do
both
you
may
have
an
opinion
on
that
is
further
recommended
that
the
use
of
dpi
based
ecmp
that
looks
beyond
the
suitable
control
word,
is
disabled
by
default
and
only
enabled,
if
both
RFC
6391
that's
the
fact
and
RSC
sixty
seven.
Ninety
a
li
are
unavailable
and
that
a
risk
assessment
has
been
taken
on
the
long-term
operational
costs
and
other
implicit
implications,
such
as
possible
out
of
order.
Ethernet
frame
delivery,
in
other
words,
we're
not
saying
people
can't
do
it.
People
will
want
to
do
it.
E
There
are
all
kinds
of
circumstances:
will
they
will
they
will?
But
this
is
a
strong
recommendation
that,
if
you,
if
it
really
is
your
only
choice
or
there
is
some
other
reason
why
you
want
to
do
it,
you
really
have
to
take
a
serious
look
at
the
long-term
implications
for
your
operations
team,
so
I
think
the
next
to
all
the
text
we're
going
to
edit.
Aren't
they.
G
H
C
H
E
H
E
E
E
E
C
F
C
F
E
Recognized
sorry
with
which
paragraph
you
I'm
talking
about
section
7
right,
okay,
so
this
is
this
is
so
so
I
guess
the
question
for
the
groupies.
What
advice
do
we
give
the
operations
team
I
mean
I?
Think
it's
important
that
they
understand.
There
is
a
problem
that
I
mean
sure
that
the
the
configuration
guy
for
the
PA
won't
control.
E
What
the
the
PE
the
Elat
the
peerages
are
doing,
but
they
should
read
this
and
say:
I
better,
go
and
talk
to
my
friends
in
the
P
route
of
space
and
make
sure
that
they're
not
gonna
screw
up
these
packets
so
that
the
two
of
us
don't
spend
a
bunch
of
time
trying
to
sort
out
a
an
incorrectly
delivered
SLA
for
up
for
our
customer.
No.
F
I
F
I
mean
so
you
are
a
peer
outer.
There
are
thousands
of
PE,
routers
and
pseudo
wires
going
through
you,
yes
and
some
are
doing
something
and
others
are
doing
other
things.
How
are
you
going
to
on
that
peer
outer,
decide
or
disable
or
enable
the
behavior
of
whether
we're
to
do
and
how
to
do
a
CMP
right?
Do
you
see
what
I'm
trying
to
say
here.
G
C
G
E
So
I
think
the
sort
of
steps
I'm
trying
to
sort
of
emphasis
our
if
you
need
to
do.
If
you
need
to
do
ecmp,
then
there
are
other
think
about
trying
to
use
the
other
techniques
and
only
if
you
can't
do
that
think
about
the
Cassavetes
method
but
think
very
carefully
about
the
long-term
consequences.
So
do
it
with
your
eyes,
open
I!
Think
that's
what
Gregg
was
trying
to
yeah
so
I
think
they
both
maybe.
E
J
E
No
okay
because
the
original
intention
was,
but
they
during
the
discussion
on
the
list
and
operator
said
actually
the
problem
is
much
worse
than
you
think
this
is
happening,
so
we
need
to
address
that
they
did
suggest
a
solution.
One
of
the
solution
that
one
of
their
solutions
was
that
we
turn
on
can
a
suit
that
control,
word
sequence,
number
checking,
but
I,
don't
think
there
would
be
much
consensus
for
it.
I'd
not
sure
it's
supported
in
any
of
the
equipment.
That's
in
the
field,
maybe.
E
J
You
aware
that
ecmp
performed
at
P
routers
happens
in
a
very
wide
variety
of
way,
some
of
which
your
proposals
would
break,
and
would
it
not
be
better
to
write
something
that
describes
what
behavior
is
required
of
ecmp,
rather
than
try
to
describe
ways
that
the
cmp
must
or
must
not
be
performed?
Given
that
you
ain't
gonna
change,
silicon.
E
K
K
E
Here
is
the
well
I'm,
assuming
that
these
are
the
key
pieces
of
text
in
the
document.
I
assume
that
we
will
agree
on
those
and
then
the
authors
will
make
the
rest
of
the
document
align
with
whatever
we
agree
here
and
then
we'll
go
through
the
standard
process.
But
I
was
hoping
to
get
some
active
sort
of
discussion
on
what
this
needs
to
be
to
try
and
sort
of
get
there
a
bit
faster.
E
B
B
F
F
E
B
E
F
E
E
E
F
M
M
B
J
In
the
en
I
can
be
placed
midway
in
the
stack
yes
and
it
can
be
placed
midway
in
the
stack
for
the
reason
that
somebody
using
it
might
not
be
able
to
get
all
the
way
down
to
the
bottom
of
the
stack.
Yes,
this
language
implies
that
you're
kind
of
free
to
choose
which
one
you
use
yes
without,
which
gives
the
implication
that,
given
a
free
choice,
I
might
as
well
just
use
fat
which
is
not
going
to
help
you
get
through
this.
The
P
routers
that
need
the
e
Li
halfway
up
the
static.
So.
J
There
is
a
problem
that
needs
to
be
solved
as
believed
that
this
approach
is
the
right
approach
to
solve
the
problem.
I
believe
that
described
what
behavior
you
must
not
create
when
you
move
packets
around
is
a
better
way
of
doing
that
than
pointing
out
well,
you
could
do
this
thing
here
and
this
widget
here
and
and
change
these
defaults
just
tell
people
what
they
must
not
screw
up
in
the
network
and
allow
people
then
to
go
and
implement.
J
N
F
Guess
you
know
what
I
did
is
saying
is
somewhat
true,
but
the
the
draft
six
sixty
seven.
Ninety
says
that
it
also
applies
for
the
pseudo
wires
as
well,
which
means
that
to
identify
a
micro
flow
within
the
pseudo
wire,
you
can
still
either
use
the
e
Li
L
or
you
can
use
the
fed
to
the
wire
right
and
I
view
this
within
the
context
of
that
right.
F
Obviously,
when
you
have
hierarchical,
LSPs
or
IP
over
MPLS
and
so
on
and
so
forth,
you
can
use
e
Li
there
as
well
yeah,
but
this
specific
text
is
at
ingress,
PE
doing
the
pseudo
wire
and
doing
identifying
the
flow
within
the
pseudo
wires
yeah.
My
assumption
is,
you
may
want
to
if
you
put
that
in
context
in
that
context,
it's
not
hierarchical
LSP,
it's
not
IP
or
MPLS.
It
is
pseudo
wire
over
edie
ingress
B
and
between
the
two
fetch
wood
or
wire
or
Li.
Yes,.
E
F
C
F
So
I
think
that
what
they
rien
was
saying.
This
is
because
of
that
confusion
right.
If
you
just
look
at
this
text
without
the
context
that
it
is
the
ingress
P
that
is
doing
through
a
wire
and
a
micro
flow
within
the
pseudo
wire,
you
have
two
schemes
available.
One
is
to
just
push
the
fetch
or
a
wire
variable.
If,
if
you
ingress
any
or
you
can
do
e
Li
and
yell
right
and
you
choose
one.
I
I
N
C
N
N
E
O
E
F
E
G
L
E
G
O
N
N
C
C
C
N
N
E
E
N
E
K
H
C
H
E
The
fat
label
is
the
bottom
of
stack.
It's
defined
to
immediately
follow
the
pseudo
wire
label,
because
what
happens
in
when
you're
handling
the
fat
label
is
that
you
do
a
CMP
on
the
whole
stack
or
at
least
the
bottom
label,
in
the
stack
and
then
when
you're
at
the
egress
PE.
You
say
this
is
a
pseudo
wire.
E
O
Movie
from
always
speaking
just
suggest,
maybe
can
we
make
it
easier
saying
that
we
require
that
at
least
one
of
these
Tomatoes
should
be
used
and
let
evil
implement,
as
it
decides
what
to
do?
If
one
or
both
you
say
at
least
one
of
these
tomatoes
has
to
be
used
and
whether
you
use
one
which
one
and
when
you
use
both
depends
on
on
on
you.
O
C
E
C
E
J
E
N
N
E
N
P
N
B
E
N
So
I'm
not
confident
I'm,
not
sure
that
this
should
be
in
this
draft.
Okay,
because
this
is
talking
about
P
Reuters
right
I
have
no
idea
what
service
is.
What
I've
you
know,
services
I've
been
transported
enter
and
they
could
be
in
transit
providers.
They
could
be
carrying
the
majority,
the
majority
the
traffic
could
be
just
IP
right
or
something
so
if
you're
gonna
make
recommendations
about
how
to
configure
or
do
load
balancing
on
LS
ours.
Yes,
I
think
that
probably
should
have
a
spin
a
draft
in
the
MPLS
working,
but
not
here.
Okay,.
I
E
C
E
So
here's
a
problem
I
have
that
when
I
looked
at
the
the
example
text
from
the
from
the
sort
of
release,
notes
track
user
manual
or
whatever
it
was
that
was
circulated
on
the
studio
I
list,
it
was
actually
stated
in
the
context
of
pseudo
wires.
So
this
is
not
someone
randomly
doing
this
thing.
It's
a
feature
for
use
in
conjunction
with
pseudo
wires.
E
The
only
reason
it
ever
looks
to
check
for
zero
is
just
to
to
detect
the
presence
of
anything
that
pseudo
I,
so
it
is
sort
of
in
it
is
really
in
the
scope
of
this
working
group
comment
on
features
associated
with
pseudo-ops.
Now
you
could
argue
that
we
should
also
talk
about
it
in
MPLS.
Iii
will
absolutely
support
that,
but
this
is
a
pseudo
wire
feature
that
appears
on
one
I
am
told
more
than
one
operating
network
operating
system
so
but.
I
I
E
N
E
E
E
B
N
E
E
E
Well,
there
are
some
circumstances
in
limited
networks
where
it
might
make
sense
right:
I,
don't
iiiiii,
I,
think
I'm,
it's
not
one.
That
I
would
recommend
anyone
implement,
but
I
can
understand,
because
there
are
better
ways
of
doing
it,
but
I
can
understand
why
people
might
have
done
this.
The
issue
is
to
make
sure
that
they
remember
praise
themselves
of
the
risks
associated
with
doing
it,
and
that's
all
I
really
wanted
to
put
in
I.
We
could
say
less,
but
I
do
think
that.
E
Maybe
we
need
yeah
I,
think
there's
a
legitimate
case
for
us,
as
the
pseudo
wire
experts
saying
an
interesting
feature.
But
you
know
do
your
risk
assessment
because-
and
we
say
that
on
the
basis
that
an
operator
complained
to
us,
that
our
pseudo
eyes
didn't
work
properly,
because
this
feature
got
turned
on.
E
F
F
I
mean
it's
not
within
within
the
scope
of
the
pause
working
group
to
specify
what
the
behavior
of
an
LS
re
cmp
should
be
right.
I
think
there
is
a
confusion.
You
know
some
pseudo
wires.
Have
this
control
word
there
has
the
has
it
enable
other
pseudo
bias.
May
not.
There
is
also
IP
over
MPLS
is
going
on
and
so
on
and
so
forth.
You
know
what
should
be
the
right
behavior
at
the
LSR
right
right,
so
it's
very
problematic
with
respect
to
what
should
be
the
right.
F
H
If
I
understand
this
correctly,
this
is
the
operational
consideration
section.
Yes,
so
I
don't
have
a
problem
with
the
past
working
group.
Actually,
if
describing
a
behavior
that
actually
caused
by
the
design
on
the
pseudo
eyes
right.
So
it's
fine
to
describe
it
here.
Yes,
maybe
when
it
comes
to
actually
writing
down
the
more
specification
like
text
that
actually
should
go
in
the
one
of
the
drafts
we
talked
about
earlier
in
the
MPLS
working
group,
so
I'm
kind
of
splitting
it
up
describe.
E
E
I
Saying
that
we
removing
that
pseudo
wire
control
will
replace
DPI
is
naive,
going
to
and
the
recommendation
is
naive
getting
into
the
DPI
and
how
we
can
solve
it
or
not
solve
it.
This
is
seriously
guys
it's
like
like
this
tight,
like
just
don't
do
this
or
if
you
want
to
do
this,
think
well
what
you're
trying
to
do
and
work
with
your
vendor
to
make
sure
this
will
work,
because,
by
definition
it
will
not
like
I
think
that
week,
that
we
can
say
is
just
like
bending
and
writing
text
so.
E
I
E
E
G
L
E
L
G
Should
clarify
for
the
record?
Actually
that's
my
son,
his
door
that
those
who
enter
will
be
shots
will
be
shot
again.
I
should
clarify
for
the
right
Reggie's
only
joking
yeah,
but
on
the
serious
note
and
the
first
paragraph
that
edited
the
text
take
a
look
whether
its
capital
must
or
it's
just
masked,
as
not
normative.
G
A
I
D
E
E
F
F
E
E
F
E
K
K
L
K
E
F
Which
is
what
he's
saying
he's
saying
that,
because
it's
not
recognizable
fetch
wood
or
wire
label
is
not
recognized
a
buzz
and
if
they
go
below
the
last
bottom
of
the
stack
label
and
look
at
the
control
word
and
then
do
a
DPI
be
below
that
they
not
unless
they
disable
that
or
not
stop
doing
that.
It's
gonna
continue
right.
Q
I
E
P
I
E
E
E
E
B
E
B
D
D
Who
should
get?
Who
should
this
go
to
and
I
Triple
E,
definitely
eight
or
2.1
right
and
then
the
other
one
was
math,
and
then
the
the
liaison
would
briefly
state
the
issue,
point
to
the
recommendations
and
the
updated
RFC
or
the
the
draft
and
asked
that
the
draft
be
can
taken
into
consideration
in
either
previous
or
future
work.
And
at
this
point,
based
on
the
maturity
of
the
text,
I
would
hold
off
on
this
for
at
least
a
couple
weeks.
Okay,.
M
D
M
And
I
think
yeah
I
think
probably
I
would
say
in
I
would
say
the
rack
and
probably
probably
802
dot
one
I
Triple
E,
8
or
2.1,
because
I
think
that's
that's.
Who
cares
most
at
this
level
either
that
or
you
certainly
can
send
it
to
802
as
a
whole,
and
we
won't
figure
out
where
in
802
to
go
but
I
would
copy
802
that
one
to
a
chair,
Glenn
you.