►
From YouTube: IETF101-GROW-20180319-1740
Description
GROW meeting session at IETF101
2018/03/19 1740
https://datatracker.ietf.org/meeting/101/proceedings/
A
Good
afternoon
welcome
to
ATF
100.
This
is
the
Gro
working
group
beating
what
Oh
101
even
says
on
the
slide
in
front
of
me,
the
101
thigh
ETF
has
commenced
you're
in
the
grow
meeting.
If
you're
not
here
for
grow,
you
could
stay
and
enjoy
it
where
you
could
go
to
where
you're
supposed
to
be
and
not
be
late
too
late.
A
A
Somebody
told
us
probably
three
meetings
ago
at
this
point
that
we're
supposed
to
remind
you,
if
there's
IPR,
on
the
draft
that
you're
working
on
that
you
should
keep
that
in
mind
and
then
tell
the
chairs
when
it
goes
up
work.
Your
last
call,
so
we
can
inform
the
right
people,
whoever
those
are
that
there's
IPR
claims.
Okay,
there's
a
whole
bunch
of
resources,
again
I'll
make
that
smaller
next
time.
So
it's
more
visible,
the
jabber
room,
where
we
need
a
jabber
scribe
who's,
the
jabber
scribe.
A
A
They're
blue
sheets,
hopefully
everybody
has
had
a
blue
sheet
go
by
them,
if
not
make
sure
to
sign
it.
When
you
go
through
it
make
sure
we
get
the
right
size
room
and
that
when
legal
problems
happen,
they
can
tell
you
were
in
here,
not
colluding
or
something
Deborah.
Scribe
is
John
who's.
The
minutes
taker,
it's
thicker,
who
wants
to
take
minutes,
can't
really
go
any
further
without
minister,
yes,
I
have.
A
Excellent,
so
you
for
the
minutes
taker
you
can
take
them
in
your
own
thing.
If
you
like,
you
can
use
the
crazy
web
enabled
cloud
thing
on
on
the
I
recommend
notepad.
Do
that
then
just
email,
yo,
abhorred
grotesque
chairs
at
the
end
as
much
as
you
can
get
down
as
good,
don't
kill
yourself
have
an
agenda.
Oh
I
forgot
before
we
go
to
the
agenda.
We
have
three
chairs
up
here,
because
one
decided
to
leave
Peter's
leaving.
A
C
D
E
E
F
F
F
So
the
first
thing
is
that,
as
you
see,
we
are
we
added
a
text
saying
that
we
have
a
new
pure
type,
so
that
we
are
saying
that
the
new
pure
type
is
la
crêpe,
and
this
is
a
clarification
only.
It
was
already
in
the
zero
draft
and
we
are.
We
have
added
just
this
little
text
to
you
know
in
green
here
to
say
that
we
also,
let's
say
added
this
other.
You
know
clarification
here,
saying
that
you
know
we
have
an
F
flag
in
so
a
little
bit
of
clarification
about.
F
Then.
Essentially,
we
have
you
know
third
modification
here
in
which
we
have
you
know
a
TLV.
So
we
were
already
saying
that
there
is
in
the
peer
of
an
optional
TLV,
in
which
we
could
say
what
is
the
VAR
f,
so
the
name
of
arrive
for
example,
and
essentially
what
we
are
saying
here
is
that
you
know
to
in
order.
You
know
we
can
add
the
same
information
to
the
peer
down.
The
motivation
for
this
is
just
to
try
to
reduce
the
state,
so
we
have
some
implementation.
F
We
also
add
three
modifications.
Also
this
one
is,
you
know,
just
a
clarification
only
essentially
here
we
have.
We
are
just
saying
that
we
have
you
know
the
El
flag
should
be
sad.
You
know
always
in
order
to
indicate
that
you
know
information
is
post
policy
and
plus
the
thing
that
you
know
if
we
have
a
post
policy
kind,
so
we
are
other
about
post
policy.
We
should,
you
know,
represent
how
things
are
actually
sent
to
the
pier
right.
F
How
to
say
that,
so
again,
so
we
have
the
El
flag
to
say
that
you
know
we
should
indicate
it
should
be
set
to
zero,
because
you
know
it's
a
to
indicate
that
it's
pre
policy
and
then
you
know
we
have
all
that
thing
over
there
and
and
then
the
last
you
know
here.
The
last
modification
to
the
draft
is
that
we
have
this
new
optional
element.
F
Only
you
know
the
peer
group
and
not
the
different
updates,
for
you
know
that
are
sent
to
the
single
individual.
You
know
bgp
speakers,
so
essentially
yeah.
What
we
want
to
represent
here
is
just
labeled
for,
for
you
know
that
group
of
update
messages-
and
this
is
some
feedback
that
we
received
in
chicago.
You
know
to
try
to
limit
the
chatting
a
sort
of
verbosity
of
you
know
of
BMP
on
our
about,
and
essentially
we
took
care
of
that.
You
know
with
this.
F
You
know
with
that
with
this
countermeasure
and
that's
it
so
essentially
should
I
summarize,
you
know
beyond
just
the
changes
to
the
draft.
So
what
we
have
here
is
that
we
got
two
set
of
comments
so
to
the
draft,
so
once
that
it's
a
you
know
need
at
work,
so
it
seems
there
is
a
kind
of
you
know,
feedback
that
this
is.
F
You
know
something
useful
that
we
are
doing,
and
the
second
is
that
we,
you
know,
received
little
feedback
in
sense
of
you
know
what
needs
to
be
done,
whether
you
know
these
things
that
are
missing
and
whatever.
So
we
feel
that
we
addressed
all
the
comments
we
received
so
far,
but
you
know
I
would
like
to
encourage
people
if
there
is
any
further
feedback.
G
Randy
Bush,
while
like
and
support
these
intellectually,
we
have
a
serious
problem
in
DMP,
both
in
implementation
from
the
vendor's,
with
massive
suckage
and
with
weak
deployment.
Probably
do
significant
part
to
the
offenders,
but
weak
deployment
in
the
community.
I'm
worried
that
adding
more
for
the
weak
vendors
will
not
improve
things.
H
F
I
I
I
The
information
is
completely
lost
and
kind
of
adding
to
the
cover
to
the
policy
languages,
an
optional
exit
or
drop
code.
What
we
would
seem
to
be
very
easy
and
having
one
drop
reason
recorded
and
available
easily,
but
I
think
helped
a
lot,
because
it
would
be
really
nice.
If
my
ops
people
have
an
easy
way,
Malka
to
report
to
my
customers
and
peers,
what
specific
kind
of
crap
they
are
sending.
F
Thanks
very
much
for
the
comment,
so
I
I
I
I
tell
you
this
I
mean
I
was.
I
have
been
doing
quite
some
work
with
the
pre
and
post
policy
myself,
really
using
it
in
practice.
Right
and,
of
course,
the
maximum
I
could
go
with
the
BMP
was
just
doing
the
validation
exercise,
which
is
like
I,
have
pre
policy.
I
have
expectations
and
I
have
post
policy
and
then
I
see
whether
the
post
policy
matches
my
expectations
or
not
so
I
do.
F
What
you
are
saying
is
to
go
the
next
level
and
with
the
already
getting
in
BMP
some
code
or
something
like
you
know
what
is
dropped
by
what
or
what
is
passing
due
to
which
policy
or
whatever,
to
be
honest
with
you,
I
I
understand
that
I
feel
this
is
a
little
bit
out
of
scope
of
this
work
and
especially
because
I
think
that
we
go
a
little
bit
in
Greenland.
Already,
policies
are
dreamland
for
what
is
standardization
and
but
yeah
I.
Take
your
comment.
Yeah
I
understand.
C
I
F
See
my
a
my
concern
and
I
agree
with
you
yup
that
whatever
we
should
do
in
this
area
should
be
something
separate
and
out
of
these
drafts.
My
concern
is
just
that
you
know
when
you
go
into
the
policies
you
know
area
and
trying
to
determine
you
know
what
is
going
on,
or
you
try
to
find
a
common
language
across
vendors
to
describe
what
what
is
going
on,
like
even
trying
to
find
a
single
code
for
different
policy.
Language
kind.
I
Of
kind
of
kind
of
very
simple,
just
just
give
me
like
in
the
usual
programming
and
shell
script,
shell
script
languages
just
give
me
an
optional.
An
optional
programmer
defined
exit
code
and
I,
make
I
figure
out
I
figure
out
how
I
do
the
code
assignments
and
interpret
them.
That's
that's
that
looks
fairly
easy
and
we
don't
need
more
than
32
bits.
Probably
16
would
be
fine.
C
K
Jeffires,
so
I
I
did
have
some
other
conversation
with
Ruettiger
about
this.
You
know
and
exactly
the
same
responses
you
do,
the
the
the
thing
I
do
sort
of
great
after
our
discussions
is,
we
probably
can't
standardize
everything
and
it
doesn't
even
make
sense
to
try
what
would
be
useful
potentially
is
allowing
for
you
know
that
extra
32-bit
annotation
something
else
it
in
most
implementations.
You
know
the
cost
is
understood.
K
This
is
IETF
LAN.
We
can
take
a
piece
of
the
code
space
and
carve
out
a
well-known,
no
set
of
exit
fields
that
people
can
put
their
policy
languages
for
very
common
situations.
No,
this
is
a
good
thing
and
we
could
also
leave
plenty
of
space
for
people
to
put
in
their
own
exit
codes
and
annotations
for
their
own
internal
purposes,
and
they
would
be
internal
and
sadly,
based
on
how
this
stuff
would
likely
be
now
per
router.
That
means
that
provisioning
has
to
ensure
the
consistency.
K
G
G
G
These
are
just
the
well-known
and
the
green
ones
they
delete
and
the
pink
ones
they
don't
okay,
so
you
have
to
know
that
and
take
a
tweezers
and
remove
those
if
you
want
them
God,
okay,
so
where
are
we
now
and
what
do
we
do?
We
try
to
figure
out
how
to
push
a
button?
Okay.
So
if
you
don't
know
this
is
happening,
you
maybe
not
advertising
the
communities
that
you
mean
to
advertise
and
you
may
be
advertising
communities
that
you
didn't
want
to
advertise.
G
So
what
would
we
recommend?
Should
we
tell
the
vendor
in
particular,
oh
by
the
way,
we've
now
got
four
or
five
vendors
there's
only
one
so
far,
who
has
this
unique
behavior?
So
should
we
tell
that
vendor
that's
a
bug
or
have
operator
their
operators
out
there
who
have
become
billing,
Li
or
unknowingly,
dependent
on
it?
G
G
I
In
German
in
German
and
German,
there's
a
nice
way
of
saying
yes
and
no
in
the
same
word,
I
was
surprised
when,
at
the
Chicago
IETF
at
a
nice
dinner
vendor
engineer
was
telling
me
about
very
surprising
behavior.
But
let
me
fro,
let
me
throw
one
question
at
you:
you
are,
you
are
talking
about
a
different
treatment
of
well-known
communities
and
differently,
colored,
well-known
communities.
I
Actually,
the
repertoire
of
well-known
communities
actually
actually
is
an
interesting
thing.
At
the
Irish
countryside,
I
ATF.
We
then
ID.
Our
chair
declared
that
thinking
about
how
well-known
communities
were
defined
in
the
first
bgp
RFC's,
there
cannot
be
more
and
recently
our
test
steam
came
back
to
me,
reporting
about
strange
behaviors
from
of
well-known
communities
that
I
wasn't
aware
and.
I
G
I
G
I
I
I
C
G
This
went
off
so
can
I
just
I
think
check.
This
is
where
are
we
interested
in
this,
or
is
this
just
a
bug
of
one
vendor
beyond
that?
If
we're
going
to
start
redesigning
communities
and
the
semantics,
we
attach
two
commands
of
them,
I'm,
not
sure
this
is
the
place,
and
this
probably
is
not
the
Internet
draft
to
do
it,
and
now
other
people
are
welcome
to
burn
the
rest
of
our
time.
K
Jeff
has
so
I
think
you
need
to
ask
I
Anna
why
it's
register
is
well-known.
I
think
this
is
one
of
those
cases
where
they
named
a
registry
based
off
of
a
piece
of
the
document
and
there's
not
a
proprietary
from
the
document.
So
vendors
did
this
partially
because
I
Anna
had
this
Ayanna
had
this
because
of
lack
of
paperwork.
G
Let's
not
go
down
the
subject
of,
what's
well-known,
that
says,
I'm
trying
to
isolate
this
from
a
long
primrose
path
who
he
is
there.
Are
there
people
here
that
feel
that
considering
this,
a
bug
is
inappropriate,
in
other
words,
that
significant
people
will
be
hurt
if
the
behavior
is
changed
for
that
deletion
of
some
well-known
communities?
I
L
G
C
C
If
we
improve
them
that
it
will
make
my
life
easier
in
terms
of
provisioning,
and
it
will
enable
a
certain
region
of
our
world
where
IRR
never
took
off,
but
our
I
definitely
is
taking
off
in
South
America
and
operationally
speaking.
We
are
struggling
in
that
region
of
the
world
because
of
the
lack
of
a
SS
equivalents
in
RPG
I'ma.
C
So,
let's
quickly
go
through
what
IR
s
and
then
we'll
build
comparisons
to
what
RPG
is
and
then
we
can
go
from
there.
The
Adam
after
IRR
is
the
route
object
and
the
route
object
is
basically
a
tuple
of
a
prefix
and
an
authorized
origin.
I
won't
go
into
how
we
measure
authorized
because
there's
many
different
databases
with
many
different
rules,
but
in
a
nutshell,
these
two
data
elements
are
the
most
important
ones.
C
Then
I
can
query
a
so-called
IRR
D
IRR
daemon
I
can
issue
it
a
command
say
give
me
the
list
of
prefixes
where
this
originated
n
appears
and
then
I
get
back
data
and
that
data
can
either
be
represented
in
the
raw
form.
As
you
see
here,
or
it
can
be
easily
converted
into
a
vendor
specific
CLI
construct.
And
basically
this
is
what
goes
into
the
provisioning
system
and
it's
deployed
on
routers
on
the
ebgp
sessions.
C
So
in
comparison,
you
can
inspect
the
role
as
the
repositories
you'll
see
lists
of
a
essence
and
lists
of
prefixes,
and
that
reminds
us
of
the
route
object
and
you
can
very
easily
use
simple
one-liners
to
obtain
similar
data
and
transpose
that
into
your
vendor
specific
routing
convicts.
So
so
far
so
good,
we
have
the
route
object
in
IR
we
have
Roja
in
our
PGI.
C
What
we
also
have
in
IR
is
a
thing
called
a
SS
not
to
be
confused
with
a
s
underscore
sets
entirely
different
thing,
and
you
can
see
that
two
very
different
groups
of
people
worked
on
this
material
and
did
not
communicate
with
each
other.
But
what
we
see
here
is
an
example
of
my
own
asset.
We
can
query
the
the
Whois
server
at
our
that
entity
donates.
We
we
ask
it
for
s
on
5,
5,
6,
2,
:,
air
Snyder's,
and
we
get
back
this
block
of
text,
and
this
is
a
bit
grind
to
parse.
C
Here
we
have
to
easy
to
remember:
commands
exclamation,
mark
is1,
5,
5,
6,
2,
:,
Snyder's,
comma
1
for
recursion,
and
it
gives
back
a
list
of
all
the
assets
that,
in
this
particular
context,
mean
my
downstreams
to
further
visualize.
What
an
IAS
set
looks
like
I've
wrote
a
small,
open
source
tool,
IRR
tree,
you
can
find
it
on
get
up
and
it
will
visualize
how
a
s
sets
expands
and
how
the
inclusion
or
recursive
inclusion
is
visualized
so
to
build
filters.
C
Today
we
take
an
AS
set,
we
expand
the
s
set
into
a
s
numbers
and
then,
for
each
a
s
number,
we
do
a
reverse
key
lookup
to
get
the
list
of
prefixes,
then
the
list
of
prefixes,
that
is
the
totality
of
all
those
commands,
is
deduplicated,
perhaps
aggregated
in
some
way.
Perhaps
we
prune
bogans
or
other
things
from
it
and
that's
installed
into
the
routers
now.
C
A
challenge
with
a
s
sets
today
is
what,
if
the
same
AR
set
name
exists
in
multiple
IRR
databases
and
a
great
example
is
death
stealth,
which
exists
in
both
the
right
database
and
the
RA
DB
database.
But
those
objects
with
the
same
name
are
not
managed
by
the
same
organization.
In
fact,
they
have
absolutely
nothing
to
do
with
each
other.
C
So
a
lot
of
Route
servers
will
query
appearing
to
be
per
ASM
and
then
per
ASM.
Look
at
a
fuel
called
IRR
records,
and
then
they
hope
something
possible.
Is
there
and
and
it's
it's
all
of
it
is
really
a
best
effort
thing
and
with
peering
to
be.
We
have
informal
guarantees
that
the
AAS
holder
is
the
one
that
put
that
data
there,
but
I,
say
informal.
There
is
no
cryptographic
verification
that
the
owner
actually
put
that
data
there.
C
So
this
is
my
wishlist,
and
this
is
the
starting
point
of
what
I
hope
will
become
an
extension
to
our
PGI
I
want
something
that
is
that
will
make
my
discovery
life
easier,
given
a
ASM
I
should
be
able
to
programmatically
find
the
appropriate
list
of
a
essence
of
downstreams
that
that
ASM
wants
to
announce
to
me
in
in
context
of
provisioning
or
perhaps
peering
I
want
guarantees
that
the
owner
of
the
ASM
is
actually
the
one
that
put
that
data
there
with
the
IR.
We
have
no
such
guarantees.
C
Anybody
can
put
anything
there,
preferably
this
should
be
unilateral
declarations.
There
has
been
work
in
the
past
about
adjacency
attestations,
where
both
sides
can
confirm
that
they
have
a
BGP
session
with
each
other
and
I.
Think
one
of
the
beauties
with
air
sets
is
that,
and
this
this
is
sometimes
worse,
is
better,
but
organizations
that
try
to
do
their
best
try
to
document
appropriately
in
in
a
public
registry.
C
What
they
intend
to
announce
should
that
life
should
be
made
easy
and
if
we
make
a
system
that
strictly
depends
on
all
parties
involved,
acknowledging
an
attestation
that
may
lead
to
operational
complexity
and
I
won't
/
adjacent
ASM
granularity,
because
the
list
of
downstream
areas
ends
155
62
may
announce
to
mgt
can
be
different
than
what
I
announced
to
say,
GTT
or
Tilia.
So
in
the
discovery
mechanism,
entity
should
be
able
to
find
a
list
that
is
appropriate
for
entity
and
likewise
for
our
competitors.
C
H
Don't
play
a
router
configurator
either
in
real
life,
all
my
dreams,
so
I
have
no
ability
to
talk
to
the
semantics
intent
of
what
you
are
doing.
You
are
trying
to
say
what
the
meaning
would
be
on
provisioning
or
routing.
I
can't
speak
to
that,
because
it's
not
my
competency
but
I
do
think.
There
is
a
moment
in
what
you
are
saying
here.
That
is
interesting.
You
said
provisioning,
so
that's
a
clear
statement.
You
have
a
not
an
intent
in
your
mind
and
I
like
that.
H
I
think
when
you
talk
about
rpki-
and
you
talked
about
modifying
I,
have
said
that
you
are
looking
to
register,
and
this
is
my
belief.
I
believe
that
you
are
looking
to
register
a
defined
object,
type
to
get
the
OID
assigned
to
legitimate
use
of
rpki,
because
admission
in
rpki
signing
demands
an
OID
be
given
and
understood
to
be
in
the
litany
of
objects
expected
in
registry,
so
they
can
be
constrained
and
understood
so
you're
not
asking
to
modify
our
PKI
you're
asking
for
an
OID
there.
H
Our
definition
of
an
object-
and
the
last
observation
is
the
critical
moment
in
a
I
believe
the
critical
moment
in
the
definition
of
signed
objects
in
a
cryptosystem
is
who
signs
and
what
should
be
in
the
certificate,
so
the
intentionality
of
which
a
s
which
a
s
is
plural,
have
to
assert
ownership
and
permission
has
to
be
explicit
in
what
you're
saying,
and
that
is
the
end
of
what
I
wanted
to
say.
Thank
you
for
your
presentation.
I
thought
that
was
great.
Thank
you
for
the
feedback.
I
Your
folk
doctor,
Telecom
before
I,
get
to
the
really
negative
side.
Let
me
let
me
let
me
say:
yes,
it's
absolutely
correct
to
point
out
that
a
assets
are
currently
very
much
used
and
they
are
a
concept
that
does
not
appear
within
the
rpki
context,
and
it
appears
obviously,
and
the
old
are
PSL
and
yes
for
making
progress,
identifying
the
gaps
in
functionality
that
are
with
the
probably
future
minded
systems,
of
course,
is
important
so
that
the
gaps
can
be
filled.
I
Amongst
the
design
decisions
there
was
that
well,
okay,
authorization
is
not
really
part
of
it
for
those
parties.
For
those
databases
that
wanted
to
do
it,
there
was
an
arc.
There
was
an
authorization
scheme
and
only
a
few
of
the
wretches
routing
registries
actually
deployed
that
it
is
I,
think
a
really
bad
idea
to
promote
now
to
see
the
IRR
as
an
authorization
database
which
it
isn't
for
the
large
majority
of
the
servers
and
the
data
sources.
I
I
I
K
K
Oh,
I
D
for
a
given
revision
that
I'm,
you
know
choosing
to
include
in
this
set
which
may
have
expired.
You
know,
so
you
could
do
these
things
with
a
lot
more
force
than
you
could
have
done
in
our
PSL,
but
I.
Don't
think
you
truly
mean
that
so
it
is
useful,
you
could
do
it
and,
to
some
extent,
I'll
point
out
the
same
thing
that
you're
looking
for
and
one
of
your
points
about
what
I'm
choosing
to
announce,
which
is
partially
what
you're
trying
to
use
for
is
very
much.
K
C
Just
it's
one
of
those
I
would
harder
than
even
that's
terrible
aspect
is
still
better
than
nothing,
and
so
what
I'm
arguing
is
that
if
we
end
up
with
a
same
thing
as
a
SS-
and
we
only
addressed
that
we
know
the
owner
of
the
SN
published
a
list
and
that
there
are
globally
unique,
that's
already
would
be
a
huge
win
for
me.
I
find
that
useful
I.
K
K
This
that'd,
be
great
I
would
similarly
suggest
that
the
export
list
that
you
have
out
of
other
RPS
objects
does
not
get
included
in
that
object,
so
keep
the
ideas
separate
you're
good,
but
the
last
related
thing
is
even
in
our
PSL.
You
know
the
headache
you
had
of
you
distinct
naming
you
can
actually,
as
part
of
your
exclamation
mark
syntax,
restrict
the
sets
of
databases
that
are
used
in
your
query
expansion.
This
is
going
to
be
very
similar
rpki
because
the
implicit
behavior
here
that
your
trust
anchor
set
is
what's
defining
it
correct.
G
G
B
G
I
Well,
since
the
a
via
sets
and
by
the
way,
actually,
there
is
a
very
limited
way
of
having
authorized
a
SL,
but
that's
extremely
limited,
and
obviously
only
works
were
obvious,
as
is
deployed
since
the
a
s
sets
actually
as
far
as
we
can
tell
have
no
anchoring
in
an
authorization
it
doesn't
matter
in
which
I
are.
Are
we
advise
the
Latin
American
folks
to
register
the
SS
kind
of
when
I
have
a
customer
out
of
that
region?
I
Ok,
the
least
least
common
denominator
that
usually
shows
up
is
our
a
DB
and
I'm
happy
I'm,
happy
I'm
happy
to
use
that
thing
as
that
is
available
and
about
everyone
outside
of
Europe.
I
A
B
A
B
E
M
M
M
M
G
M
M
I
I
Let
me
also,
since
we
were
talking
about
communities,
classified
graceful
shutdown,
that
is
a
next-generation,
well-known
community,
which
is
where
you
cannot
tell
whether
the
recipient
actually
will
honor
the
semantics.
It
is
just
a
hint
okay,
still
it's
a
technical
I'm.
While
no
export
is
first-generation
and
you
are
supposed
to
actually
honor
than
semantics.
Okay,.
M
Still,
this
is
the
technical
opportunity
to
fix
a
significant
problem,
and
if
you're,
we
remember
the
recent
DDoS
attacks,
including
memcache,
okay,
part
of
it,
a
part
of
these
attack
was
corrupted
services
and
part
of
it
was
about
unity
to
spoof
traffic.
If
we
are
able
to
fix
both
of
them,
it's
okay,
if
we
were
able
to
fix
one
part
of
this
front
of
the
problem,
it's
also
ok.
C
C
My
main
challenge
remains
finding
equipment
that
can
do
you
RPF
at
a
reasonable
speed
at
all,
and
this
strikes
me
as
solving
like
two
or
three
percent
of
an
issue,
whereas
my
90
percent
of
the
issue
is
packet
circulation
through
the
fabric
and
Europe
EF,
cutting
that
in
half
or
similarly
that
that
I
cannot
skill
access
lists
large
enough
for
all
the
potential
source
addresses.
So
until
that
is
solved
somehow,
where
my
PPS
count
is
no
longer
chopped
in
half,
I
cannot
I'm
not
even
looking
at
this.
J
Yeah
Warren
Kumari,
with
no
hats
other
than
like,
maybe
a
dunce
one,
because
I'm
feeling
really
feeling
really
stupid.
So
the
graceful
shutdown
thing
says
that
blah
blah
blah
this
policy
tags
are
parts
received
other
session
of
the
next
one
down
yeah
and
sets
local
press
to
a
low
value.
So
basically
you're
just
announcing
the
prefix
and
saying
please
set
local
pressure
on
here,
so
you're
just
announcing
the
prefix
that
way
right.
Yes,.
J
M
J
G
J
M
I
M
M
M
B
M
M
Okay,
this
is
just
if
I
go
to
your
question
right,
so
this,
just
as
the
assumption
that,
if
somebody
was
not
previously
using
your
pay,
if
it's
the
transit,
you
will
not
enable
your
benefit,
that
is
the
level
of
transit
so
for
for
transit.
I
speech
nothing's
check
the
connection
between
transits.
Nothing
is
changing
other
otherwise
is
if
we
should
so.
M
Basically,
as
far
as
I
understand,
the
your
fifth
problem
is
that
BGP
is
a
routing
protocol
was
the
idea
was
to
use
it
as
in
delivering
informational
messages.
This
sound
network
is
available
through
these
interface.
It
was
not
working
and
it
in
general
way
way.
It
is
not
work
still
not
working.
The
the
other
option
is
to,
for
example,
create
some
other
Safi
and
propagate
the
information
about
availability,
but
it
this
this
way
is
is
a
deadlock.
Maybe
it's
a
mathematical
perfect
solution,
but
it
will
never
be
shipped
anywhere
so.
C
I'm
gonna
cut
in
here
we're
out
of
time
for
this
session.
What
I
would
suggest
is
that
we
further
explore
the
upsides
and
downsides
of
this
idea
via
demanding
lists.
Thank
you
for
your
presentation.
Thank
you
for
giving
me
this
love
and
dear
grow
attendees.
Thank
you
very
much
for
your
time,
see
you
in
the
next
ITF.