►
From YouTube: IETF93-IDR-20150724-1150
Description
IDR meeting session at IETF93
2015/07/24 1150
B
A
A
A
A
A
B
C
C
It
should
not
be
included,
introduce
our
new,
our
selfie.
A
second
one
is
sure
to
consider
most
scenarios
like
a
king
Google
s
name
at
the
door.
In
the
second
diversion
we
make
provisions
for
the
comments
surfin,
the
surfin
Ferrante
than
the
34
is
the
beautifying
for
the
team
photo
specification
rules
rather
than
the
old
Earl,
a
three-week
inflows,
pacifica
room
rules
and
the
additional
component
types
for
kinky,
o.h,
name,
etc
and
closest
to
1970
I,
careful
meeting
it
is
adopted
as
walking
crowd.
C
C
This
is
a
villa
action.
You
know
tweaking
a
scenario
where
we
can
use
full
spec
to
configure
the
policy
for
the
access
network.
The
real
and
should
be
the
villa
should
be
supported
supposed
to
pop
push
survive
and
other
operations
in
the
villa
action.
With
support
a
pub
push,
a
sweat
are
the
right
age,
ino
will
action
and
the
auto
villain
action.
We
also
support
mapping
in
a
cause
action
and
the
outer
cause
action.
C
D
So
there's
similar
reasons,
I
think
for
wanting
to
distribute
trolling
state
is
for
is,
is
and
ospf
in
fact,
role
uses
is
is,
but
it
uses
it
essentially
at
a
different
level,
doesn't
normally
interconnect
so
to
speak
with
layer.
3
is,
is
SDN
controller,
a
typical
kind
of
example,
where
you
want
to
have
the
link
state
information
available
to
some
box
that
can
be
gotten
to
buy
bgp.
D
D
D
D
They
put
that
lengths
the
labels
a
little
different
in
decades,
there's
a
whole
bunch
of
subtleties
and
things
that
are
in
true
older
defined
in
RFC,
7176
trill
use
of
is
is,
and
these
are
some
of
them.
I
think
is
actually
most
all
of
them
actually
they're
mostly
sent
as
part
of
the
router
capabilities
or
multi
topology
router
capabilities
povs
or
when
it's
talking
about
group
address
subscriptions
group,
either
Mac
or
IP
address
subscriptions.
D
D
D
You
can
do
that
in
trill,
but
you
would
want
only
if
you
do
that
if
you
were
sure
your
links
could
accommodate
such
large
packets
and
there
is,
in
this
draft
suggestion,
allocated
an
additional
bit
currently
reserved
in
the
node
flag
bits
to
indicate
that
a
particular
shrill
note
is
a
layer,
3
gateway
and
I.
Think
that's
about
it.
A
C
Hello
good
morning,
everyone
are
we
go
from
poly
on.
This
topic
is
about
a
distillation
of
full
specifications.
Those
four
were
suing
and
country
of
flow
spike,
kim
HP.
Full
spec
can
be
used
to
disseminate
traffic
filtering
rules
to
remove
the
piece.
The
typical
use
case
is
for
the
kiddos
mitigation,
the
PTP
flows
back
country.
C
C
C
This
is
the
requirements
summary
for
the
beautiful
Uzbek
for
email
screen
and
work.
The
first
one
is
the
match,
apart
through
the
include
in
the
air
to
Leslie
header
and
outer
layer,
a
layer,
the
less
we
headed
information
and
Emmas.
We
had
the
traffic
filtering
actions
supposed
to
read
up
with
regard
to
chi
tunnel
or
immense
methanol.
C
This
is
the
full
specs
extension
to
new
component
types
for
MCB
header.
Are
you
introduced
when
is
for
buying?
Is
a
vein
ID
to
identify
each
tenant?
It
is
a
24-bit
tending
to
identification,
I
identification.
The
second
type
is
the
MLC
a
protocol
protocol
type
and
deserve
so
the
one
we
want
to
introduce
a
layer,
flag
yeah,
because
emily
is
a
new
technology
which
the
future
and
the
traffic
of
a
terrain
Jews
can
be
based
on
outer
layer
of
inner
layer.
C
C
We
use
we
introduced
to
the
left
leg
in
traffic
action
yeah
once
this
beat
it
when
the
layer
flag
of
it
is
said,
the
corresponding
filtering
rules
will
be
applied
on
the
emissary
inner
layer.
If
not
a
set,
the
corresponding
filtering
rules
will
be
alive.
Applied
on
the
I
was
way
up
there,
but
maybe
there
are
other
solutions
for
the
layer
flag.
I
also
want
to
solicit
suggestions
from
the
community
about
the
layer,
flag,
okay,
I'm
questions.
F
F
Folds
back
extension
for
routing
policy
distribution.
Okay,
let's
go
you
all
know
from
x,
24,
different
kinds
of
reason:
providers
going
to
change
its
traffic
pass
entrance
exit
and,
for
example,
the
reason
could
be
linked
congestion
or
load,
especially
in
holidays
or
just
because
of
the
transmission
quality
is
not
good
packet
delay,
Lords
increase
or
just
because
of
the
crap
Taylor
packing
in
Christ,
or
even
just
a
money
money
issues
that
we
choose
to
lower
price,
and
this
one
is
the
one
real
requirement
for
my
customer.
F
I
was
going
to
a
simple
explain
here
so
as
here
my
customers,
AS
IS,
SP
100,
and
that
there's
one
p
device,
which
is
how
to
access
a
prefix
wine
different
isp
and
the
requirement
is
for
the
reason
before
we
said
you
want
a
hope:
the
AXI
that
the
traffic
axis
this
darling
through
l3
link
and
for
folks
for
explain
this
is
I
EGP
peering
between
this
HW
and
a
different
I
transit
provider.
So
this
is
for
out
buying
the
traffic
for,
for
my
customer
and
next
slide
is.
F
This
is
a
opposite
scenario
for
inbound
traffic.
So
for
a
foot
for
this
prefix
prefix,
to
which
p2
device
was
hope
to
access,
my
customer
was
hoping
the
traffic
can
enter
through
this
links,
three
between
sp10
and
I
should
w2
and
his
requirement
is
not
owning
to
just
transfer
the
traffic
from
previously
before,
for
example,
I'll
22
l,
three
two
out
to
transfer
from
l2,
but
his
requirement
is
required.
His
requirement
is
we
just
didn't
want
my
traffic
just
to
get
out
of
the
link.
I
wanted
the
traffic
to
enter
through
some
designated.
F
F
This
is
true,
too
big
burden
and
it's
easy
to
make
a
mistake
and
the
other
wise,
and
since
this
kind
of
traffic
adjustment
is
frequently
happened
during
a
day
for
different
kinds
of
reason,
but
he
so
he
hope
we
can
provide
some
dynamic
of
provisioning
mechanical
instead
of
steady,
configure
one
and,
of
course
this
purse
elution
need
to
can
be
saved
to
undo
or
roll
back
without
any
damage
and,
of
course,
as
as
time
going
this.
This
is
a
policy.
F
How
about
how
the
end
I
was
three
more
three,
more
slides
that
okay?
Thank
you.
Okay.
This
is
the
proposed
solution
and
we're
calling
this
routine
policy
distribution.
The
key
point
is
the
way
taking
effect
on
the
control
plane
set
up
forwarding
plane
and
the
impact
decision
is
executed
on
the
remote
site
and
we
are
using
which
befalls
back
as
a
vertical
to
carry
these
things
we
introduce
one
new
beat.
We
call
a
beat,
for
example,
to
add
in
the
traffic
action
community
if
this
beat
was
set.
F
That
means
we
have
more
information
character
being
in
another
new
attribute.
This
is
attributed
talking
about.
It
is
a
new
policy
attribute.
This
is
a
structure
to
his
primary
to
part.
1
match
viewed
the
others
action
field
so
for
the
match,
match
type
with
going
to
a
true
value
for
a
positive
match
or
negative
match.
Positive
match
means
when
you,
when
you
match,
is
you
have
to
take
some
action
and
that
the
negative
means,
if
you
match
you,
do
not
take
some
action
and
the
action
part
is
like
this,
and
this
is
some
subject
I'm.
F
I
was
going
to
explain
its
narrow
house
talking
our
proposed
solution.
The
next
slide
so
I'm
here
for
the
outbound
traffic
control
with
match
type
is
the
meet
and
we
use
the
user
sub.
Sorry
subtype
here
for
ipv4
neighbor
we're
going
to
precisely
designate
which
link
are
going
to
album,
so
the
action
type
is
a
route
preference
so
so
that
the
HW
to
maybe
prefer
one.
F
The
l3
link
the
route,
so
the
traffic
can
get
out
from
that
link
and
in
the
inbound
traffic
scenario
the
match
type
is
nigh
and
the
sub
T
0
is
the
for
the
souls
and
peer
address
so
designated
one
specific
link
and
that
action
type
is
prevent.
Prepend
current
s,
for
example,
5
times
that
that
means
what
advertise
the
ruta
to
the
remote
site,
I
was
going
to
I
was
not
going
to
prepend,
yes,
but
the
other.
The
other
interface
were
well
prepared.
F
G
Well,
okay,
the
question
was
in
the
context
you
were
saying
you
have.
You
have
a
customer
that
has
this
requirement
yeah
and
my
stupid
question
was
well
okay
and
you
were
talking
about
to
a
SS
and
you
wanted
to
control
the
traffic
between
of
a
simple
question
would
have
been
which
of
the
a
SS.
Does
your
customer
control?
I'm?
Sorry,
I,
didn't
get
the
last
word
well
well!
Well
there
well,
ok,
which
of
the
a
SS.
Does
your
customer
control
like
operate?
Configure
the
Reuters
CR.
F
A
F
H
H
F
A
John's
gutter,
but
before
I
make
my
comment,
do
you
I
saw
you
had
one
last
slide,
contrasting
this
to
conventional
policy?
Did
you
want
to
talk
to?
That
was
why
now
you
keep
going
to
the
end
of
your
your
deck,
keep
going
yeah.
Is
that
your
last
slide,
and
do
you
want
to
say
anything
about
it
before
I?
Ask
my
question:
okay,.
F
You,
let
me
finish:
first
yeah,
okay,
this
is
a
what
people
may
be
asked.
The
primary
concern
is
the
why
why
not
to
manual
configuration
because,
based
on
the
main
tank
experience,
my
customers
hated
to
do
these
things,
because
in
the
long
run,
is
too
much
policy
on
device
basis?
They
are
not
going
to
easy
to
maintain
and
a
very
easy
to
mistake,
and-
and
one
more
thing
is,
is
this:
this
manually
configured
party
is
going
to
change
the
on
daily
basis,
so
so
lately
cannot
take
it.
It's
not
easy
for
them.
F
B
F
C
from
thus
the
sea
traffic
is
going
to
our
visit
ip5,
but
my
customers
domains
inside
here.
He
has
no
way
he
can
do.
Pbr
base
flows
back
to
redirect
the
traffic
from
l2
to
al
wat.
This
is
legend
is
ur
l,
22,
l1
bucks.
Actually,
the
Seas
traffic
is
going
to
take
this
way
down
so
no
way
to
to
effect
that
the
device
outside
his
control.
So
this
is
a
meaning
here
and
next
slide
is.
F
We
all
know
the
PBR
basic
respect
is
changing
the
14
decision,
but
sometimes
it
may
be
causing
14
loop,
for
example,
is
see
here
he
is
going
to
visit
visit
to
the
prefix
one
and
okay
original
passes
for
em
too.
Yes,
I
can
change
that
it
directed
to
m3,
but
the
m3
actually
was
preferred
the
next
hop
for
I
gw2,
so
Luke
can
help
here.
So
that's
the
reason
why
we
promote
to
proposed
under
the
solution,
I'm
Don
John.
A
Thanks
John
Scudder
speaking
as
a
working
group,
member
in
case
anyone
cares
so
flow
spec,
historically
up
until
this
moment
has
been
about
affecting
the
forwarding
plane,
it's
about
programming,
boarding,
plane
filters
and
so
on,
yeah
you're
proposing
to
use
it
to
affect
the
control
plane.
Oh
now,.
B
A
Made
an
argument
that
says
yep
the
control
plane
is
the
only
way
that
I
can
affect
a
forwarding
plane
on
my
peer.
So
it's
in
the
interesting
argument
you
know
well
played
but
I
think
it's
a
bit
of
a
stretch
and
that
the
second
point
is
standardizing
policy
turns
out
to
be
pretty
hard.
We
already
have
an
effort
underway
to
do
that
with
a
yang
model.
A
We
have
an
effort
inside
the
IETF
that
exactly
has
taken
several
of
your
requirements,
and
you
know
in
particular
the
timeliness
requirement
like
I
need
to
be
able
to
change
this
data
quickly
that
one's
called
I
to
RS.
To
me,
it
seems
like
you
should
be
pursuing
those
and
either
making
them
do
it.
You
want
using
them.
If
they
do
it,
you
want
complaining.
A
If
they
don't
do
it,
you
want
rather
then
trying
to
fit
policy
into
sort
of
an
unnatural
container
for
it,
which
is
flow
spec
and
in
particular,
I
really
would
hate
to
see
the
working
group
trying
to
standardize
two
different
ways
of
configuring
policy.
At
the
same
time,
when
it
turns
out
that
we've
always
had
a
hard
time
doing
even
one
okay.
F
Sexy
on
okay,
my
answer:
what
maybe
was?
What
do
you
said
about
it?
Yeah
yeah
get
in
first
of
all,
I
didn't
say
this
is
the
only
way
we
can
do
that.
Sorry,
we
didn't
say
this
is
the
only
one
we
can
do
that,
but
yes
I
to
us
or
yeah,
maybe
the
other
past,
the
okay
I.
Maybe
I
do
not
want
to
say
too
much
about
this
section
if
they
need.
J
Us
right,
they
say
I
will
try
to
use
God
new
hobby
I,
try
to
explain
opinion
about
this
policy,
in
fact
that
we
also
thanks
to
careful
and
other
people.
They
propose
concern
about
the
policy
walk,
because
not
a
lot
of
pieces
of
work
is
being
done
in
the
IDF
with
a
different
working
group.
So
they
try
to
to
how
to
organize
this
work
together.
So
but
I
think
that
in
fact
that
we
truly
it
is
cut
with
cat
immerse.
J
We
had
the
other
opinion
to
you
some
days
that
dynamic
energies,
the
control
protocol
to
deploy
the
policy
so
from
our
phone
who
were
widely
finalists
one.
So
this
because
it
now
that
yonder
because
they're
the
policy
you
to
come
to
flexible
so
that
use
a
little
difficult,
it
would
be
masked
to
master
so
that
we
we
defend
this
of
the
PDP,
a
cloth
bag
of
or
the
routing
policy.
Our
the
measure
that
we
discussed
with
Latimer,
we
get
there
the
feedback
so
that
we
tried
to
set
hoppity's
the
water
from
the
beginning.
H
For
having
Eric
a
cylinder,
I
think
what
one
thing
you
need
to
do
is
document
your
assumptions
as
to
who's
as
administrative
control
of
the
a
esas
and
I
think
you
know
in
this
situation.
You'd
want
the
policy
controller
to
also
be
getting
all
the
routes
to
so
because
if
it's
just
blindly
setting
policies,
it's
got
to
know,
it's
got
to
be
co-resident
with
the
route
reflectors
I
mean
it's
gotta
know
whether
there's
a
path
before
it
does
Paul
it
making
sure
the
path
is
valid.
H
You
know
I
mean
you
should
just
document
that
so
it
just
doesn't.
You
know
if
Paul
policy
based
routing
the
way
it
typically
is,
and
you
just
configure
it
on
the
router
it's
under
its
in
its
intraday
nits
under
a
single
and
somebody
says:
okay,
I,
don't
want
to
use
any
of
these
complicated
solutions
like
multi,
topology,
routing
or
anything
like
this.
I
just
want
to
force
this
type
of
traffic.
This
way,
whereas
in
this
case
it's
the
external
where
you're
under
you
know,
presumably
both
a
s
is
under
a
different
administrative
domain.
H
I
can
I
well
do
that
yeah
I
just
say
I
just
said:
the
policy
controller
needs
to
only
set
these
flow
specs
for
reachable
destinations,
so
it
has
to
have
it
has
to
have
that
the
controller
has
to
have
that
information,
which
is
an
assumption
for
the
egress
pier
engineering
way
of
solving
this
problem
with
segments.
Ronnie
short.
K
L
Say
after
that,
Jeff
has
nope,
so
we
have
discussed
that
I
do
have
concerns
yes,
oh
those
so
I
repeat
at
least
some
of
these
things
that
have
been
mentioned,
you
for
I
IDR,
but
many
of
them
have
already
been
covered.
I
do
share
John's
opinion
that
flow
spec
is
an
awkward
container.
Okay.
I
do
also
have
the
concern
that
flow
spec
has
implicit
rule
order
ring
for
its
own
or
I.
Yes
and
I.
Don't
think
that
is
probably
a
natural
fit
for
this
as
well.
L
I
also
have
concern
that
bgp
being
used
to
distribute
this,
even
if
you
took
care
of
this
in
a
different
format,
different
nlr,
I
BJP
is
used
to
send
this
to
many
things.
It's
not
really
a
good
point
to
point
distribution
to
send
it
from
one
router
to
the
router
you're
looking
to
influence.
It
would
be
better
for
a
set
of
routers
and,
if
you're,
using
your
route,
reflector
infrastructure
to
send
it
to
many
routers
in
the
network
you're
causing
potentially
those
routers
the
carry
information
they
don't
care
about.
L
L
My
suggestion
to
you
is
joining
that
to
make
sure
that
your
policy
for
your
customer
can
be
represented
in
the
model
that
comes
from
this
I
also
suggest
joining
work
in
I
to
RS,
specifically
for
signaling
this
no
from
what
router
to
another,
and
if
you
don't
believe
I
dress,
is
the
right
signaling
method.
No,
is
a
distribution
mechanism
pursue
that
potentially
is
a
separate
item?
I
do
still
think
that
the
modeling
work
probably
belongs
in
yang.
Ok,.
A
J
J
M
On
this
is
Jenji
from
Huawei
I'm
going
to
present
level
information
for
bgp
philosophy
philosophy
is
an
interpol,
consider
several
matching
criteria
that
can
be
applied
to
IP
traffic
and
a
flaw.
Speckle
role
can
be
mapped
to
a
level
in
mpls
network
level.
Best
imagine
is
more
efficient
on
at
hand
and
I've
c55
on
75
defines
the
requirement
for
traffic
aesthetics,
Buffalo,
espectro,
sofla
flows.
Becca
role
could
use
a
level
found
with
the
best
match,
unique
culture
out
for
the
destination
prefix
embedded,
the
ending
of
lost
patrol.
M
M
Assume
assume
a
philosopher
channel
need
needed
from
ip12
IP
tube
needed
to
be
established
and
their
philosophy
bro
is
injected
in
p2,
then
p
to
distribute
to
the
philosophy,
control
boundaries,
level,
302,
sdr2,
sbr,
to
distribute
to
the
philosopher,
go
with
level
202
svr,
one
and
sv
r.
1
is
tributes
flow,
spec
row
with
level
100
to
pee,
when,
when
he
was
received,
traffic
from
ip12
IP
to
the
packet
will
be
40
the
best
on
levels,
seven
at
the
same
time,
the
corresponding
actions
or
will
be
enforced.
The.
M
The
second
way
to
associate
fellow
spectro
with
the
level
is
using
I've
c
2,
107
and
each
philosopher
control
is
associated
with
level
map
in
t.o.
We,
the
prefix
field
in
the
level
map
until
we
needed
to
use,
tell
me
prefix,
and
we
are
wondering
which
way
is
better
I'm
here
to
solicit
your
comments
and
suggestions.
Thank
you.
E
L
L
The
intention
of
the
authors
is
that
this
next
stop,
maybe
not
directly
connected,
maybe
far
away
in
the
network
and
that
the
device
that
is
installing
the
flow
spec
filter
may
need
to
tunnel
it.
You
know,
tunnels
may
in
some
cases
include
3107
no
tunneling
as
well.
Our
request
is,
if
you
think
that
a
3107
style
label
for
redirecting
things
should
be
explicitly
part
of
the
signaling,
for
something
like
redirect
IP.
E
B
B
B
So
this
draft
has
a
little
background
at
the
draft
was
submitted
in
grouting
working
group
in
march
this
year.
We
thought
that
most
of
the
work
has
been
done
for
bdp
flow
specification.
So
we
resubmitted
the
draft
in
IDR
working
group
after
taking
care
of
lot
of
configuration
and
state
parameters,
changes
and
filled
up
some
of
the
gaps
which
were
lacking
in
the
first
initial
version
of
the
draft.
B
So
here
I
am
going
to
talk
about
the
overview
of
the
young
model.
So
essentially
you
know
what
we
are
trying
to
cover
is
configuration
and
state
model.
Only
we
haven't
covered
the
RTC,
a
notification
as
part
of
this.
This
model,
essentially
because
typically
advocacy
and
education
are
part
of
the
vendors
with
extensions,
and
so
we
have
focused
currently
on
the
qualifications.
To
date,
only
the
flow
spec
rule,
as
defined
in
this
model,
is
based
on
Alta,
finding
RFC
5005.
B
For
state
information,
it
modest
route
information
flow
counters,
so
this
draft
is
compliant
by
RFC
687.
What
does
that
mean?
That
essentially
means
that
the
draft
has
been
formatted
based
on
the
suggest
given
in
RFC
60
x,
7,
and
that
also
means
that
the
the
model
is
fully
extractable
from
the
from
the
draft
and
it
compiles
without
any
warning.
B
B
Conflict
container
contains
a
set
of
policies
in
each
policy,
contain
a
set
of
flow
spec
rules
and,
at
the
same
time,
flows
back
state
container,
container
flows
back
rip
information
which
contains
our
out
information
for
the
prospect,
and
it
contains
a
full
spec
status.
The
counters,
which
is
/,
/,
/
flow,
spectral
counters
flow,
suspect
group
remain
tails.
You
know
close
by
crowds
and
flow.
B
Spec
statistics
maintain
a
set
of
counters
okay,
so
this
is
an
overview
of
the
model
where
you
can
see
that
in
this
tree
structure,
you
can
see
that
there
is
a
flow
spec
top-level
container,
then
we
have
a
conflict
container
which
is
treat
right
and
then
we
have
a
state
contain
state
container,
which
is
read-only.
A
conflict
container
contain
a
set
of
policies,
and
each
policy
contain
a
set
of
dual
state
container.
Contain
a
rib
container
in
the
stats
container
different
in
a
contained
set
of
routes
and
starts
contain
contain
a
set
of
hunters.
B
B
B
Ok,
so
for
configuration
data,
you
can
see
that
in
this
fee
structure
their
set
of
policies,
policy
associate
with
the
vrf
naman
adjust
family.
Then
we
have
set
of
rules
associated
with
contains
a
set
of
components,
and
each
component
is
effort
by
type
and
then
merely
so
component
types.
Similarly,
for
the
actions
we
have
set
of
actions
referred
by
action
type
and
as
explained.
D
B
B
Now
respect
route
information
is
contained
in
the
reef
container
and
contains
a
generic
as
well
as
protocol.
Specific
parameters
of
respect
rules
are
part
of
the
protocol.
Special
parameters,
data
container
and
all
the
components
and
actions
are
part
of
the
protocol
specification
flow
spec
statistics
contain
flow
counters,
which
is
/.
We
are
of
/
address
family
and
is
a
set
of
prospect
rules
and
each
rule
will
contain
its
own
counters.
The
counters,
which
we
support,
is
packet.
Bytes
number
of
packets
number
of
bytes,
then
drop
packets
count
in
top
fights
now.
B
B
Similarly,
in
the
next
slide,
I
will
show
you
for
the
flow
spec
stated
how
we
have
flow
spec
statistics
as
a
as
a
container
contains
a
list
of
respect
stats
each
floor.
Specs
at
sentry
contains
a
vrf
name
and
address
family
for
each
were
a
frame,
address-family
combination,
we
have
set
of
rules
and
each
will
contains
a
list
of
components,
a
list
of
action
and
set
of
counters.
L
Jeff
has
I
think
you
did
a
very
nice
clean
job
of
your
configuration.
So
no
congratulations
on
that.
I
will
be
sending
some
specific
comments
about
maintenance,
including
separating
some
of
the
stuff
for
I
Anna
registries
rather
than
inside
the
document,
because
otherwise
you
can't
maintain
the
code
points
separately.
Mm-Hmm
two
specific
comments
went
to
went
to
the
chairs.
The
first
one
is
for
the
operational
state.
L
B
If
you
know
different
vendors
have
different
existing
implementation,
many
times
you
know
in
the
yang
model,
we
have
to
map
it
actually,
and
this
is
the
fact
and
is
the
fact:
no
deletion
but
other
yang
models
also,
and
this
and
no
flow
yang
model
will
actually
fit
as
a
case
for
all
the
vendors.
So
people
have
to
map
it.
They
have
to
maintain
some
kind
of
layering
to
maintain
it,
and
that
is
a
fact.
Xp
right.
L
It
the
main
reason
to
make
the
observation
is
that
for
several
of
those
vendors,
it's
the
case
that
the
firewall
is
implementing
the
flow
spec
and
if
you
have
any
intersecting
rules
that
are
in
the
firewall
that
are
not
covered
by
the
flow
spec
filter.
Packets
are
still
being
counted
in
the
same
buckets
and
it's
not
possible
to
distinguish
them
so
I
think
not
so
much
that
it
can't
be
mapped,
but
I
think
that
it
will
never
be
clean
in
such
circumstances.
So
it
may
be
misleading
operational
state
mmhmm.
B
Yeah
those
are
interesting
comments.
Actually,
you
know.
As
far
as
our
implementation
is
concerned,
we
have
a
firewall
rules
and
after
federal
rules,
we
specifically
configured
prospect
rules,
so
anything
which
passes
of
firewall
hits
the
floor
spec
rules,
and
that
is
the
order
we
maintain
right
but
yeah.
Obviously,
a
many
vendors
will
have
their
own
implementation
and
so
to
struggle
a
little
bit
to
handle
voltage.
So.
L
The
question
partially
for
you
partially
for
the
chairs.
You
know
this
is
actually
a
nice
model
of
flow
spec,
but
it's
written
effectively
as
a
service
model.
It's
a
way
of
saying
here's.
What
flow
spec
is
capable
of
supporting
in
terms
of
firewalling
behaviors.
So
with
that
respect,
it's
good
bgp
mech
might
be
a
consumer
of
the
state.
L
Potentially
a
individual
router
might
be
a
consumer
of
the
state.
We
have
to
figure
out
which
those
which
of
these
scenarios
were
specifically
servicing,
and
how
does
this
interact
with
other
efforts
in
the
ITF
to
do
filters
about
that,
like
so
I
to
NSF?
As
an
example,
is
a
potential
consumer
of
this
or
potentially
would
have
input
on
this
as
a
service
model
again
realizing
that
this
is
targeted
towards
an
existing
protocol
mechanism?
L
The
protocol
mechanism-
there's
not
a
clean
fit
in
the
ITF
modeling
stuff
right
now,
for
this,
so
I'm
going
to
suggest
that,
though,
there's
gonna
be
some
of
the
little
coordination
that
everybody
involved
will
need
to
do
to
figure
out
where
this
fits
into
the
architecture.
So
we
can
please,
as
chairs,
raise
this
with
the
architecture
so.
E
E
L
E
L
So
the
question
being
flow
spec
is
was
invented
specifically
because
was
a
way
to
allow
the
distribution
of
junipers
no
firewall
rules
for
some
specific
cases
into
bgp.
So
it's
the
flow
spec
is
less
a
firewall
language
as
more
of
a
serialization
format
that
allow
me
to
keep
pick
this
up
as
written.
This
is
one
part
could
be
used
for
here's
how
you
actually
put
this
stuff
in
there.
L
Here's
a
way
to
represent
your
static
routes,
a
driven
know
each
of
those
cases,
but
it
also
sort
of
looks
like
it's
designed
to
be
able
to
be
used
as
sending
these
filters
directly
to
a
router
for
local
installation
with
no
bgp
necessarily
being
involved.
That's
actually
an
interesting
case.
You
can
say
it's
a
nitrous
type
thing.
This.
L
You
could
do
this
bgp
or,
if
you
have
it
through
no
yang.
The
point
being,
is
that
if
you're
going
to
take
that
use
case
into
account,
other
pieces
of
IETF
are
going
to
be
doing
firewalling
group
work.
This
is
a
in
that
context.
A
way
of
doing
firewalling
work
rather
than
a
control,
plane,
protocol
yang
model
right.
E
B
Right
on
that,
yeah
I
would
like
to.
You
know,
put
one
comment
that
you
know
the.
We
thought
that
it
is
important
that
we
covered
specifically
for
frost
education,
which
will
cover
all
the
flow
specification,
current
model
drafts
and
the
future
droughts.
It
will
focus
on
that,
rather
than
a
very
specific
firewall
model,
and
so
that
is
the
reason
we
focus
this
as
part
of
the
document
right
and.
L
If
your
intention
is
to
try
to
make
this
just
the
interface
for
bgp
d
use,
I
may
either
be
miss
reading
it
or
some
additional
clarification
could
go
in.
So
a
focusing
a
tightening
of
the
focusing
would
mean
that
it's
the
applicable
only
the
eye,
dr
as
an
example,
but
it
is
actually
an
interesting
thing
for
other
work
right.
L
This
is
actually
a
side
point.
The
eye,
dr,
should
be
where,
if
people
are
not
necessarily
paying
attention
that
the
IGP
groups
are
actually
doing
flow,
spec
redistribution
within
their
protocols,
the
ospf
group
has
actually
adopted.
This
already
had
not
been
previously
aware
that,
and
a
similar
draft
is
actually
present
in
is
is
so.
Our
technology
is
getting
used
in
other
places
in
context.
That
I
personally
find
a
little
weird
but
I'm
trying
to
communicate
with
the
authors
to
understand
your
skates
better.
N
B
B
N
B
C
C
C
C
C
The
fullest
backup
policy
in
this
argument
is
a
figure.
The
philosopher
co
policy
will
be
used,
spl1
a
spi
for
don't
need
the
doesn
t,
needles
of
low
spec
of
policy,
so
SPF
o
can
use
oh
I,
former
Kenny's
to
of
a
to
filter
the
of
La
Fleur
speculoos
SPF,
for
you
send
us
a
message
to
ask
yes
ray
I:
don't
need
that
I!
Don't
need!
The
love
follows
buckaroos
letter
s,
pra,
leão
notice,
standards
of
respective
used
to
as
therefore.
C
C
L
A
J
O
Hello,
everyone
I'm
sandy
from
the
key,
it's
my
presentation
for
upstream
assigned
label
collision
solution
and
at
first
let
me
see
an
example.
This
boy
is
simple:
p1
and
p2
send
different
MVP
and
flowed
through
ps3,
but
they
use
the
same
empty
2
and
p
colonel
to
send
the
different
flow.
So
ps3
will
be
confused
because
the
inner
label
and
outer
label
all
the
same,
so
ps3
will
not
distinguish
the
two
different
flow.
O
Because
the
employment
Network
Earth's
revision,
we
was
used
for
many
years
and
the
network
is
developing
the
MVP
and
the
TC
will
consume
at
least
number
of
upstream
assign
labels
for
endure
for
you,
nerd
label
or
after
label.
So
the
label
space
is
reserved
on
a
burpee.
We
can
have
you
girl,
h,
but
the
space,
if
we
resolve
as
large
many
levels,
will
be
waste.
O
O
So
all
the
time
stream
p
will
choose
one
the
same
one,
although
these
include
upstream
and
downstream
piece
will
receive
the
rules
so
also
the
upstream
piece
that
at
work
highest
roots
with
later
time,
step
will
at
just
he
can
just
the
label
and
also
it's
a
hybrid,
because
the
time
stable
maybe
same
from
different
p.
So
we
can
use
the
IP
address
of
upstream
ps2
kappa
rip
a
little
choice.
O
I
May
refer
shot
ski,
why
not
use
ruther
ID,
which
is
commonly
used
tiebreaker
and
it's
absolutely
deterministic
over
the
time
its.
If
you
have
time
sync
across
the
network,
fine,
the
area
see
areas,
but
then,
if
if
noone
dote
comes
in
or
somebody
don't
have
it
time
synchronization,
then
you
can
have
a
game
chair
and
expect
it,
because
I
can
the
Naruto
I
will
advertise
earlier
time
because
of
Miss
configuration
or
whatever
and
then
just
make
it
simple
and
just
need
this
simple,
tiebreaker
right.
O
O
J
J
O
J
O
J
So
that's
it
my
my
question:
why?
Why
why
you?
They
are
the
label
confliction
because
of
the
accordion.
My
understanding
about
this
is
a
I'm
reviewing
drafted
that
the
label
sabe
SI,
you,
the
contacts,
the
base,
that
is
the
contact
the
base.
A
label
is
a
related
with
the
PM
kitano,
because
the
will
they
are
two
different,
the
p1
katana,
so
the
label
spacer
should
be
different
way.
They
are
the
label
confliction.
J
O
A
I
just
like
to
mention
that
if
I
understood
robbins
point
correctly,
that
was
also
the
point
that
Eric
Rosen
made
in.
He
sent
an
email
to
the
eye,
dr
list
and
to
you
I,
think
on
july
twenty,
and
I
think
he
was
making
pretty
much
the
same
point
about
using
the
the
context
and
I.
If
you
hit
I,
didn't
see
a
reply
to
you
from
you
too
Eric
yet,
but
I
would
recommend
you
read
his
email
and
run.
N
A
Yeah-
and
this
was
going
to
be
my
second
point,
which
was
a
this-
is
probably
this
work
is
probably
mostly
in
the
mpls
area
and
so
I'm.
Sorry,
we
did
not
catch
that
before
and
suggest
that
you
present
to
them,
but
at
least
you
should
send
your
document
to
the
mpls
mailing
list
as
well,
so
that
they
can
discuss
it.
Okay,.