►
From YouTube: IETF106-SIDROPS-20191120-1520
Description
SIDROPS meeting session at IETF106
2019/11/20 1520
https://datatracker.ietf.org/meeting/106/proceedings/
A
A
It's
searching,
20,
we're
gonna
start
the
meeting,
The
Cider
ops
meeting,
if
you're
not
here
for
cider
ops
you're
in
the
wrong
room
for
those
presenting,
please
make
sure
to
talk
into
the
mic
like
this
hold
the
mic.
Don't
use
the
stand!
If
you
do,
if
it's
in
the
standard
you're
doing
this
I
can't
see
here,
you
either
there's
a
nice
thing
in
front
of
the
presenters.
They
can
read
that
instead
of
doing
this
also
very
nice,
we
have
a
yep.
A
A
B
A
C
D
B
B
B
A
D
A
Good,
we'll
just
move
ahead
blue
sheets
for
those
who
didn't
get
them
earlier.
We
could
pass
into
the
back.
There's
a
note
well
for
the
meeting.
I'm
sure
that
by
this
day
of
the
meeting
you've
seen
this
a
few
times
it's
in
small
print
intentionally,
it
doesn't
have
all
the
resources
that's
further
on
okay.
So
today
we
have
six
presentations,
not
counting
this
one,
hopefully
will
be
done
quickly
with
our
part.
A
Those
times
are
approximate,
they
probably
don't
add
up
to
the
right
time.
I
didn't
pay
attention,
so
strat
draft
statuses,
there's
some
stuff
that
we
have
to
push
into
the
editors
queue
out
of
working
group.
Last
call
there's
at
least
two
things,
I
think
here:
yeah
yeah,
maybe
a
couple
of
more
the
last
thing
in
iest
processing
is
LTI
use
cases
I
think
that's
stuck
for
a
long
time
and
I.
A
Don't
exactly
remember
why
it'll
pop
out
eventually
and
the
things
that
are
probably
the
ones
that
need
to
go
into
the
editors
queue
are
things
like
OB,
egress,
I,
think
somebody's
complained
about
that
I
mean
sorry.
I
sent
a
message
about
that.
A
couple
the
others
here
at
will
as
well
so
we'll
hear
about
those
I'm
sure
at
the
mic
and
afterwards
we'll
scrape
through
the
land
list
and
figure
it
out.
Okay,
first
person
up
is
Tim.
E
Why
well,
first
off,
as
you
know,
our
sink
is
used
in
the
RPI.
So
is
another
protocol
that
was
invented
here,
RDP,
so
the
RPI
repository
Delta
protocol
for
friends
and
what
I'm
proposing
here
I
I
made
a
draft
just
in
my
own
name.
It
doesn't
have
to
stay
like
that.
I'm
actually,
looking
for
some
people,
I
mentioned
to
me
that
they
want
to
co-author
as
well,
but
I
wanted
to
get
out.
So
they
get
something
out
quickly.
E
So
I
did
that,
but
before
going
into
all
of
that,
a
quick
reminder
why
why
did
we
invent
the
Delta
protocol
there?
You
can
talk
about
this
for
longer,
but
mainly
in
order
to
run
an
R
sync
repository
reliably.
It's
not
that
easy,
and
with
RDP
you
can
leverage
CDMS,
so
the
job
becomes
a
bit
more
yeah,
less
challenging,
I
would
say,
and
also
from
the
relying
party
point
of
view.
Most
software
ends
up
calling
the
arcing
binary
on
a
system
which
may
have
different
behaviors
on
different
systems.
E
E
Pragmatically
speaking,
I
propose
to
keep
the
rsync
your
eyes,
even
though
a
repository
might
not
be
available
at
that
location,
and
the
reason
is
that
if
we
update
the
certificate
profile,
this
can
lead
to
things
be
becoming
rejected
and
that
will
make
deployment
much
harder.
There
is
some
prior
art
in
an
XML
where
namespaces
are
HTTP
your
eyes
quite
often,
and
their
main
be
anything
there.
E
Also,
we
could
instruct
relying
parties
to
prefer
our
DP
if
it's
available
and
well
ultimately,
phase
out
our
sync,
but
when
well
there's
a
bit
of
a
chicken
and
egg
problem
here.
I
didn't
find
good
pictures
of
chicken
and
eggs.
Maybe
I
didn't
look
well
enough,
but
essentially
this
is
the
situation.
You
have
many
repositories,
you
have
many
relying
parties
and
the
relying
parties
talk
to
all
the
repositories
and
repositories
will
need
to
support
our
thing
as
long
as.
E
Can
only
drop
support
for
our
sink
if
all
the
relying
parties
support
the
alternative
and
vice
versa,
our
relying
parties
can
only
drop
support
for
our
sinker
and
all
the
repositories
support
the
alternative
and
I.
Think
measurement
should
turn
into
that,
because
if
you
look
at
the
wrote
there
then
Randy
wrote
a
proposal.
Actually
that
can
work
but
I'm
not
sure
that
we
have
to
tie
it
on
to
that.
But
to
repeat
what
you
said:
Randy-
and
you
can
correct
me
later
if
I'm
wrong
is
that?
Well,
you
can
start
from
the
repository
side.
E
E
People
actually
deploy
these
new
versions
and
at
some
point
you
can
now
say:
ok
now
the
repositories
can
phase
out
parsing
but
in
my
view,
I'm
not
sure
that
this
should
be
a
heavy
segment
of
the
document.
I
think
in
the
eventual
document
that
final
document
that
you
probably
not
be
there
at
all,
in
my
view,
because
I
think
it
should
just
talk
about
what
it
updates.
E
But,
of
course
this
is
an
important
discussion
to
have,
in
my
view,
I
think
we
aim
for
having
everybody
move
forward
in
parallel,
because
I
don't
see
why
that
cannot
happen,
meaning
that
we
would
ask
relying
party
software
implementers
to
it
also
support
RDP,
even
if
not
all
the
repositories
already
have
it
so,
but
first
questions
and
well.
This
is
not
even
a
working
group
document,
so
that's
also
another
question
that
it
might
go
to
the
list.
F
Jeff
Houston
there
were
many
mistakes
and
bad
ideas
in
the
original
work
that
developed
this,
and
one
of
them
was
a
sink.
I
actually
think
that
the
problem
that
we
got
given,
which
was
the
entire
x.509
distributed
repository
mess
and
the
inability
to
sort
of
smash
them
all
together
was
was,
was
a
really
terrible
problem
to
get
given
and
opportunistic
pool
was
a
solution
clutched
out
of
desperation
rather
than
a
hard
design.
F
F
Maybe
we
should
look
further
afield
than
just
replacing
the
underlying
mechanism
of
this
particular
protocol
and
look
at
an
architectural
sense
and
go
other
alternatives.
We
can
piggyback
then
actually
do
do
reliable
flooding
so
that
when
I
publish
in
my
repository
it
just
flows
through
everyone
else.
F
E
To
that
because
well,
in
my
view,
it's
it's
quite
interesting
to
look
at
BGP
as
a
plotting
protocol
for
this,
but
I
think
there
will
be
many
things
that
need
to
be
figured
out
and
it's
going
to
be
a
while
before
that's
a
reality
at
this
moment,
I
think
supporting
our
sink
is
impacting
people
and
it's
low-hanging
fruit
to
do
that.
First
of
all,
so
in
parallel
you
can
look
at
alternatives.
At
least
that's
my
my
take
on
it.
G
Some
problems
with
arsenic
are
known.
The
RDP
has
been
developed
as
a
probably
working
solution.
We
do
not
have
the
proof
in
actual
scaled
operations,
I
think
I
think
we
need
some
something
there.
Some
document
that
explains
the
roadmap
for
actually
getting
forward
for
the
question
of
how
eggs
and
poultry
are
bred.
G
My
take
is
actually
actually
the
repositories
have
to
move
first,
that
does
not
preclude
getting
the
implementation
for
clients.
Quite
obviously,
if
the
road
map
says,
the
repositories
actually
should
work
on
providing
a
reliable
service
before
clients
start
to
rely
on
them
or
even
start
to
do
large-scale
operational
testing.
G
H
G
I
D
E
There's
some
experience
there
and
I
think
we're
all
new
software
there's
an
initial
phase
of
getting
it
right.
I
think
the
tricky
part
of
that
protocol
is
generating
those
files
reliably,
but
serving
them,
on
the
other
hand,
is
quite
easy,
but
I
agree
also
at
what
routing
a
said
that
we
should
build
a
operational
experience
with
running
this,
but
some
people
have
been
doing
that
so
particularly
the
ripe,
NCC
and
Edenic
and
others
are
in
the
process
of
developing
and
employing
as
well.
Thank.
I
H
Wendy
Bush
are
kissing,
I
oj
I,
don't
think
Jeff
went
far
enough
and
I
and
I
don't
think.
Jeff
went
far
enough
and
I
have
and
I
have
a
t-shirt
that
says
BGP
awarding
PhD
theses
since
1994
and
I
think
other
bullish
males
pushed
BGP
down
our
throats
when
we
really
should
have
considered
idrp
more
seriously.
I.
H
H
E
H
H
E
Into
mind
about
including
a
sequence
I
think
is
useful
during
this
discussion
and
the
coming
period.
But
if
eventually,
you
have
a
document
that
updates
the
existing
standard
documents,
I'm,
not
sure
that
it
should
still
include
a
road
map,
but
you
can
have
a
road
map
whilst
you're
talking
about
this
stuff
and
maybe
keep
it
in
a
document
because
it's
a
convenient
place
to
put
it,
but
once
it
becomes
our
of
seed
and
I'm,
not
sure
that
a
road
map
is
going
to
get
outdated,
I'm
afraid.
But
maybe
you
should
take
this
offline.
E
B
H
J
Lost
Arkus
and
Dragon
research
labs
a
couple
of
comments
with
a
claisen
said:
they
were
running
out
of
time.
First
of
all,
two
people
who
didn't
quite
get
Tim's
comment
about
why
we
want
to
keep
the
arson
to
our
eyes.
The
problems
would
definitely
basically
have
us
hashes,
because
we
very
very
carefully
removed
all
identifying
information
from
the
RPK.
Oh
and
the
only
thing
that
was
left
as
the
authorized.
J
If
you
take
your
eyes
away
to
them,
what
you
have
is
hash,
it's
lots
and
lots
and
lots
of
hashes
which
gets
kind
of
confusing
pretty
quickly
on.
Secondly,
I
slightly
disagree
about
the
marshalling
the
data
into
our
DP
being
difficult.
Actually,
it's
not
yeah.
We
hit
limit
of
this.
What
four
years
ago
we
had
to
reimplementation
zuv
our
see
a
code.
That
is
because
our
CA
code
was
in
retrospect
designed
badly.
It
should
have
been
doing
basically
a
transactional
model
to
begin
with.
J
As
soon
as
you
go
to
a
transactional
model,
this
is
just
you
generated
Delta
a
for
every
transaction
done,
it's
trivial.
There
are
a
few
interesting
characteristics
on
the
RP
side,
though
in
particular
when
you're
dealing
with
a
very,
very
large
collection
like
the
ripe
ncc
collection.
The
validation
behavior
is
different.
J
Okay,
if
you've
got
I'll
just
use
this
poster
children
throwing
rocks
of
anybody
poster
children,
you
have
a
pea
neck,
as
the
you
know,
holdouts
for
the
least
efficient
possible
way
of
doing
arson,
and
you
have
Reitman
CC
as
the
early
adopters
for
a
big
one
of
our
DP.
J
The
one
nice
thing
about
the
rather
strange
think
the
de
pinna
cause
is,
you
can
kind
of
do
all
the
crypto
in
parallel
right,
fetch,
it
fetch
d
fetch
and
the
fetch
is
so
slow
that
by
the
time
time,
you're
done
fetching
you've
also
done
all
the
validation
with
the
right
model.
It's
a
little
stranger
than
that
with
the
right
model.
You
have
to
wait,
wait
Rance
fir
and
process
and
suck
into
the
database,
which,
unfortunately,
is
slow.
J
All
of
that
data
before
you
can
start
validating
any
of
it
really
and
it
maybe
there's
some
efficient
way
to
do
it.
I
haven't
figured
out
yet,
but
to
first
approximation.
You
have
to
eat
the
whole
update,
apply
all
the
deltas,
and
unfortunately,
databases
tend
to
be
bad
about
doing
this
quickly
and
then
you
have
to
do
all
the
validation.
Is
this
massive
chunk?
So
the
behavior
is
a
little
different
I
think
this
protocol
is
actually
pretty
much
ready,
whether
or
not
we
are
ready
to
deprecate.
A
J
K
A
K
A
D
A
A
H
We
drafted
this
document
and
essentially
it's
transfer
of
resources
between
two
registries,
known
asses
and
IP
numbers.
By
registering
the
rest
of
this
discussion,
we
mean
ISPs,
our
IRS,
NIRS,
etc.
We're
not
going
to
transfer
between
two
I
Anna's,
it
should
be
doable
and
we
kind
of
have
to
do
it
fairly
often.
Actually,
how
often
do
you
do
transfers?
Nathalie
often
she
says
okay
and
just
to
get
some
terminology.
H
There's
the
seller
and
the
buyer
and
the
swing
point
is
the
I
are
at
the
lowest
point
in
the
hierarchy
which
the
seller
and
buyer
has
a
common
parent
and
which
is
agreed
to
act
as
the
agent.
If
this
one
doesn't
agree,
their
parent
could
be
the
swing
point
is
what
subtly
mentioned
there,
and
there
was
actually
an
internet
draft
I'm,
not
sure
whether
it
became
an
I
or
I've
seen
or
not.
That
talked
about
parents,
grandparents,
acting
with
authority
I,
don't
remember
that
that
it
didn't
make
it
to
RFC
yeah
I.
H
Remember
writing
it
because
of
this
problem,
but
anyway,
I'm
gonna
breeze
through
this
fairly
quickly,
because
the
point
of
this
exercise
is
really
not
transfer.
Okay,
simple
transfer,
is
the
seller,
separates
out
the
resources
and
tells
the
swing
point
the
swing
point
without
deleting
these
resources
without
the
leading
this
delegation
creates
and
gives
them
to
the
buying.
I
are
okay,
the
swing
point
delegates
to
the
buyer
and
then
the
swing
point
withdrawals
from
the
seller
when
agreed,
okay
and
Steve
Kent
said
he's
going
to
write
up
something.
H
H
H
The
point
of
my
story
today
is
the
our
IRS
have
are
driven
by
underlying
data
which
I'll
call
stats
files,
but
you
know
some
of
them
call
them
something
else.
Some
of
them
have
their
authoritative
data
hidden
and
the
stats
files
merely
a
realization
of
it,
but
it
makes
the
difference.
This
is
a
sit.
The
problem
is
these
files,
I
believe
in
all
the
RI.
Ours
are
one-to-one.
Mappings,
this
resource
may
only
be
owned
by
one
are
up.
One
I
are
at
a
time
which
means
break
before
me.
H
H
What
we
have
today
is
not
operationally
viable.
There.
I
was
brought
to
my
attention
by
an
LIRR
who
actually
broke
okay.
So
where
do
we
go
with
the
transfer
document?
Shall
we
move
it
forward
or
not?
I
don't
know,
and
how
do
we
help
the
are?
The
our
IRS
become
unstuck
from
this
problem
and
I
think
that's
my
last
slide.
I.
G
Dutch
Telecom
I
seem
to
remember
that
Steve
Kent
at
some
point
did
draft
that
allowed
some
outside
of
rpki
signalling
of
something
like
this
being
in
process
and,
first
of
all,
criteria
for
relying
parties
that
wants
to
check
whether
the
are
PKI
certificate
system
actually
looks.
Reliable,
quite
obviously
include
include
the
criterion
that
it
looks
very
suspicious
if
a
resource
is
claimed
by
two
parties.
In
parallel.
G
G
H
H
H
H
E
E
Yeah
no
I'll
keep
it
really
sure
I
think
we
should
talk
about
this
offline.
I
read
the
document.
I,
never
get
scared
of
the
mechanics
that
are
proposed
and
I
think
we
can
aim
for
simpler
solutions
and
I'm,
not
yet
convinced
that
you
need
to
signal
these
things
to
anybody.
Looking
at
the
rpi,
maybe
you
can
leverage
the
provisioning
protocol.
Let's
have
a
talk.
This.
G
For
an
attempt
at
code,
I
get
the
nasty
idea.
The
RI
hours
presented
us
with
an
applicability
statement
as
an
internet
draft.
That
pointed
to
a
certain
file
for
disambiguating
potential
conflicts
in
the
rpki.
It
would
be
very
simple,
I
guess
to
add
in
that
file
another
column
that
says
okay,
this
resource
is
moving
somewhere.
E
Right:
okay,
on
the
break
side,
for
example,
provisioning
protocol-
you
ask
a
question
to
your
parent.
You
say
what
are
my
entitlements.
It
tells
you
what
they
are.
It
might
be
less
than
what
you
have
there's
no
grace
period.
For
example,
that's
one
thing
you
might
consider
that
you
can
communicate
to
a
child
that
something
will
be
shrunk,
go
clean
up.
E
But
if
you
do
then,
yes,
currently,
the
proposal
is
to
create
a
separate
certificate
with
all
the
resources
that
are
going
to
be
transferred
on
it,
I'm,
not
sure
that
that's
the
way
to
do
it,
you
can
also
invent
a
signed
object
that
says
these
resources
are
going
to
be
transferred.
I
would
think,
that's
already
easier.
L
H
L
H
C
Hello:
everyone,
my
name,
is
Alexander
Asimov
I
work
for
Yandex,
and
this
will
be
a
rather
quick
update
thanks
about
a
spade
rats
during
this
year,
Randy
with
fighting
with
me
to
make
the
draft
less
ambitious,
to
have
clear
description
about
the
procedures
that
are
present
in
this
draft,
and
these
slides
is
another
tenth.
That's
failed,
so.
K
C
An
oversimplified
description
of
the
solution
without
no
feelings
and
unknowns,
and
so
on.
So
there
is
a
drafts,
suggest
two
procedures:
first
procedure
for
prefixes
that
I
received
from
customers
and
peers
and
the
second
procedure
for
prefixes
that
I
received
from
providers
and
for
if
we
are
receiving
prefixes
from
the
customers
and
peers.
The
only
pair
so
that
may
exist
in
the
ice
bath
is
custom
to
provide
a
custom
to
provide
a
custom
to
provider.
C
D
C
Ambiguity
of
these
slides
its
invalid
path,
the
downstream
path
it
may
have
both
upstream
path
and
downstream
sub
path,
and
but
each
fit
is
followed
again
by
an
upstream
path.
It's
a
route
lik!
So
don't
look
at
this
light,
its
listen!
What
about
what
I'm
speaking
about
it's
very
simple
and
very
hard
to
describe
not
in
gold
but
with
words
and
slides.
So
there
were
also
questions
about
deployment
process
I
and
it's
very
important
to
bring
immediate
benefit
for
the
ice,
be
who
will
be
the
first
one
to
sign?
C
First,
a
spay
record
and
a
speed
desert,
because
as
soon
as
it
signs
a
spay
record,
it
gives
a
chance
to
other
parties
to
detect
mistakes
around
leaks.
So
it's
very
poor
to
know
islands.
You
know
cooperation
which
just
works
for
malicious
activity.
It's
slightly
a
harder
process,
but
it's
not
that
hard.
So
you
need
to
create
royal
records.
So
it's
a
it's
also
up
to
you
and
your
security
also
relies
on
your
upstream
providers.
If
you
have
happy
to
be
t1,
you
are
not
relying
to
or
to
anybody,
so
it
uses
an
is
period.
C
C
After
the
suggestions
in
the
mailing
list,
we
changed
a
spay
profile
from
players
that
were
customer
and
provider
to
pairs
where
we
have
customer
and
set
of
providers
to
solve
Possible's
you
musician
issues.
We
also
added
another
state
for
verification
procedure,
which
is
unspecified,
so
we
have
valid
invalid
unspecified.
It
will
not
be
that's
useful
at
the
state
of
owner
adoption,
but
can
give
your
additional
debug
opportunities
at
the
state
of
middle
adoption
or
high
adoption
rate
I've
added
some
beautiful
drawings.
That
shows
how
the
state
between
valid
in
validity
unknown
are
changing.
C
C
C
So
if
you
want
to
at
least
work
to
proceed
quickly,
please
give
us
feedback
help
us
to
work
on
the
text
another
on
outside.
We
will
receive
your
feedback
work
on
it
and
we'll
try
to
finish
working
on
the
text
until
the
next
ITF
meeting
rain.
They
also
promised
that
he
will
work
on
archaea
protocol,
and
so
this
is
the
last
missing
part.
So
we
will
have
three
drafts
one
profile,
one
verification
and
a
data
that
is
supposed
to
go
together
to
the
working
group
Lascaux.
So
that's
all
Randy.
H
Object
was
clearly
designed
and
clearly
not
very
significantly
changes.
They
all.
E
All
right
profile
drafts
in
particular
that's
what
any
other
one
and
I
left
a
comment
on
the
list.
I
wanna
eat
here.
So
this
is
not
a
big
deal,
but
only
the
customer
is
actually
needs
to
be
present.
Only
and
entity
certificates
in
the
signed,
object
and
I
would
think
it
would
be
good
to
recommend
people
that
they
don't
include
anything
else,
because
if
you
do,
then
you
run
an
unnecessary
risk
because
of
any
of
those
resources
are
no
longer
held
by
the
certificate.
E
E
Think
the
CAC,
the
text
currently
says
that
the
object
is
valid
as
long
as
the
customer,
a
s
is
in
the
set
that
is
on
the
EE
certificate,
and
I
would
suggest
that
you
keep
that
EE
certificates,
small
in
terms
of
resources,
because
you
only
need
to
have
that
a
s
that
customer
s
there.
You
don't
need
anything
else.
So
I
wouldn't
include
anything
else,
because
if
you
do
that's
just
a
avoidable
risk
of
things
that
can
break.
C
This
is
my
second
set
of
slides
and
they
are
not
related
to
any
draft,
but
I
believe
they
can
bring
variable
considerations
for
this
working
group
so
and
we
will
discuss,
as
you
see
much
length
the
default
behavior
normally
when
we
are
speaking
about
default,
behavior
or
default
settings.
By
expecting
that
these
kind
of
settings
are
working
in
most
of
cases
and
for
a
current
advocare
object,
roll
the
default
behavior
suggested
to
be
drop
invalids
now
rejected
methods,
don't
accept
in
minutes.
C
So
when
route
becomes
invalid,
there
are
two
options:
first
option:
if
the
reason
is
match
in
the
outer
own
sister
numbers-
and
the
second
option
is
that
the
prefix
length
is
bigger
than
the
length
specified
in
the
row
object.
So,
as
you
have
already
guessed
so
we
will
spend
next
five
minutes
discussing
I
if
max
length
applicable
for
with
the
default
suggestion
that
such
with
such
rounds
should
give
it
a
rejected,
and
it
can
a
move
forward.
C
Yeah
there
will
be
another
type
us
on
the
next
slide.
So
I'm.
Sorry,
it's
not
a
thick.
It's
AfriNIC,
so
I've
tried
to
contact
all
current
reuse
to
find
out
what
are
their
default
settings
when
people
are
creating
much
length
because,
as
you
know,
the
the
hosted
our
creation
of
rows
is
the
current.
So
nearly
all
of
them,
while
answering
just
seems
to
think
they're
looking
at
the
prefixes
detective
at
the
moment
and
creating
marks
length
that
is
equal
to
the
prefix
length
look
from
now,
but
there
are
okay.
C
But
there
are
a
couple
of
problems:
first,
if
we
all
took
the
site
of
ingress
filtering,
you
cannot
drop
invalid
in
case
of
custom.
Sending
your
air
black
hole
from
egress
site
is
just
the
same.
So
if
you
are
sending
a
black
hole
route,
you
cannot
just
blindly
drop
invalids
and
it
is
related
to
Rudi
girls
draft
and
she
is
already
in
line.
C
But
there
is
another
issue:
let's
say:
let's
take
a
look
from
the
security
perspective
from
the
security
perspective
in
a
perfect
world,
where
is
a
pacaya
is
globally
deployed?
We
will
have
raw
and
I
hope
will
also
have
a
spare
objects
and
to
stop
both
malicious
activity
and
mistakes.
We
don't
need
max
length
and
there
is
another
point
and
it
is
more
crucial
at
the
state
because
it
happens
at
the
state
of
partial
adoption.
C
Imagine
you
are
the
possessor
of
Network
/
23
and
you
have
created
rows
with
help
of
some
web
interface,
and
your
supervisor
is
a
good
fellow
and
here
already
using
a
rotation
procedure,
but
somewhere
another
guy
makes
it
typos
and
he
hijacks
your
prefix
and
it's
going
to
be
propagated
and
there
is
it's
not
specific
and
you
cannot
fight
it
because
you
first
need
to
update
roles.
After
that,
you
need
to
call
your
action
provider.
C
C
There
are
networks
that
have
already
put
to
this
issue
into
consideration,
and
some
of
them
are
sitting
max
length
to
the
L
to
the
maximum
values
as
it's
possible.
This
way
I
said:
yes,
it
turned
into
to
fix
this
issue,
because
I
was
thinking
about
this
risk
a
couple
of
years
ago,
but
recently,
I've
learned
from
several
players
for
some
of
them
have
very
big
names
this.
C
So
they
are
already
hacking
on
top
of
apical
cache,
so
the
increase
in
max
length
on
their
side
to
accept
invalids
from
their
peers
and
from
their
customers,
and
of
course
it
is
all
connected.
What
is
set
by
default
in
hosted
CAS
in
routing
registries,
I'm,
not
giving
you
any
kind
of
solution
at
the
moment,
but
I'm
asking
this
working
group.
Do
you
consider
this
as
a
problem
that
should
be
taken
into
account?
If
and
if
you
think
we
should
work
on
this
issue?
Let's
discuss
possible
solutions
right.
G
Again,
okay,
there
is,
in
my
opinion,
absolutely
no
question.
The
correct
default
for
brewers
is
a
simple
row
with
max
length
equal
to
V
traffic's
length.
That's
the
basic,
that's
the
basic
rule,
the
max
length
the
max
length
longer
might
be
something
that
should
be
only
available
in
expert
mode.
The
user
interface
after
after
examination
that
people
have
a
rough
idea,
what
they
are
doing
may.
C
Enter
at
you
and
ask
you
is
a
co-author
of
egress
filtering.
How
do
you
see
deployment
process
of
these
draft
if
the
default
policy
and
drop-in
reddits
is
not
working
for
the
leaves,
because
I'm
not
I,
don't
think
that
the
target
audience
of
this
draft
is
dr.
chili
con
I
when
I
first
saw
this
draft
I
said
I
was
thinking
all
great,
so
they're
stupid
guys
will
protect
themselves
from
making
mistakes,
but
if,
if
it's
not
working
by
default,
I'm
not
thinking
they
will
deploy
it.
I.
G
C
G
G
G
It
is
quite
clear
over
egress
will
help
me
to
shortcut
stupid
misconfigurations
that
my
colleagues
are
doing
for
customers
for
customers.
A
kind
of
the
ingress
policy
will
ensure
that
I
will
not
accept
invalids.
They
are
in,
except
for
the
cases
where
we
have
a
special
agreement
and
that
special
agreement-
quite
certainly
you
have-
will
have
implications
for
special
handling
on
the
egress
site.
If
there
is
actually
a
egress
side,
yeah.
D
C
Okay,
okay,
okay,
I
will
not
least
each
company
that
is
at
present
in
this
room
but
anyway,
but
take
into
account
the
idea
that
there,
as
far
as
I,
understand
the
target
audience
for
egress
draft
is
Leafs
and
to
make
it
working
at
least
site.
You
make
you
need
to
make
it
as
simple
as
possible.
The
principle
keep
it
simple.
Stupid
is
the
only
way
to
make
it
and
to
make
it
work
to
make
it
global
or
not
global.
N
Jeff
has
trying
to
keep
this
short.
The
max
length
stuff
is
obviously
advice
to
the
global
internet
at
large.
This
is
what
the
global
Internet
should
be
seeing
for
these
things,
the
misconfigurations
we
tend
to
see
here
are
often
things
that
you
want
to
have
no
longers
show
up
for
some
scoped
purpose.
N
In
many
cases
now
the
black
holes
typically
need
to
go
1a
s
over
same
thing
is
true
in
many
cases
of
various
types
of
te
routes,
no
something
that
might
be
a
little
bit
longer
than
what
you're
really
good
announce
to
the
world
at
large.
The
two
things
that
can
be
done
here
is
that
the
parties
that
are
next
to
each
other
that
are
cooperating
can
just
simply
configure
something
into
their
system.
That
says
that
we're
willing
to
ignore
the
rpki
answer
for
this,
because
we've
otherwise
done
validation.
N
We
know
that
this
is
a
valid
origin.
It's
following
the
paths
for
my
filters
and
I'm
willing
to
accept
it,
because
it's
a
little
bit
longer.
The
other
thing
that
could
be
done
is
that
maybe
provide
some
sort
of
hint
into
the
RO
objects
themselves.
For
these
narrow
scopes
of
things
that
it's
okay,
if
you
are
here,
but
for
your
more
general
case
of
you're,
seeing
something
much
longer
and
trying
to
mitigate
an
attack,
I
think
that's
a
much
harder
problem
to
solve.
B
Yep
job
Snyder's
ntt-
it
is
quite
tricky
to
at
this
stage
of
the
game
fiddle
with
the
semantics
of
next
length,
because
we're
deploying
the
snowball
is
rolling
down
the
hill
and
for
better
or
for
worse,
some
things
kind
of
are
what
they
are,
what
they
are
right
now
you
identified
a
specific
case:
black
holing
for
deaths.
I
brought
up
a
different
approach
where
we
make
black
hole.
Route
requests
rely
on
less
specific
routes,
which
then
can
be
pumped
through
the
original
validation
procedure.
B
L
B
D
K
K
I
Folks,
I'm
Deema
from
zds
and
I'm,
also
working
for
mr.
open-source
project
for
RP
Caroline
party,
so
I'm
here
to
report
to
the
working
group,
I
can
say
if
it
was
about
the
update
after
this
informational
document.
These
things
is
the
working
group
last
call
and
during
the
working
group
last
call,
we
had
got
some
comments.
We
got
some
supports
and
Allah
is
out.
I
We
also
made
update
accordingly,
especially
by
adding
double
control
to
reference
IFC
84
16,
which
you
suggested
by
leaking
out
and
making
some
wording
improvements
as
such
as
by
oleg
komarov
ski
and
then
we
made
a
further
update
because
some
reference
to
the
draft
has
been
partially
of
the
documents.
Okay
at
the
present,
we
also
believe
this
document
is
ready
to
move
forward
and
the
request
for
chess
actions,
so
next
step,
I'm
ready
for
actions
and
I
suggest.
You
owe
me.
A
A
N
O
I
Okay,
it's
Dima
again
I
like
to
take
this
opportunity
to
brief
implementation
of
synchronizing
rpki
in
the
cache
by
distributing
cache
data
in
JSON
over
HTTP.
So
before
I
touch
upon
the
very
topic
of
my
presentation,
lesson
review
the
rpki
cache
deployment
scenarios
as
discounting,
obviously
attitude
when
they
are
ooh.
Another
background
for
the
following
discussions.
I
As
far
as
I
comprehend,
as
ARP
cache
deployments
in
arif
are
describing
I
see
a
tattoo
one.
Arrow
concurrently
could
be
made
that
a
lot
older
dogs
have
to
set
up
its
own
RPC
system
to
do
our
PTI
data,
synchronization
and
verification
and
RB
k
value
than
the
cash
is
not
necessarily
transported
around
us,
but
between
two
intermediate
italo's
arranged
in
a
hierarchical
scheme.
I
I
will
reassure
that
the
conclusions
that
I
just
made
in
my
last
slide
was
a
reasonable
and
it's
desirable
for
us
to
have
a
tool
to
distribute
our
PK
value
data
cache
to
among
different
organizations
over
network
the
university
disease
figure.
The
architects
implemented
ways
Avista
the
open
source
software
is
not
in
service.
Our
Praveen
cache
data
to
some
organizations
in
China
by
distributing
cache
data
in
JSON
format
over
HTTP
to
our
servers
deployment
in
lead
hooks
of
ISP
or
ICPs
in
China.
I
So
moving
along
to
this
slide,
I
am
now
presenting
our
consideration
of
the
implementations,
so
Alif
all
the
way
of
formatting
data.
We
prefer
to
facilitate
data
paths
for
this
violation
and
applications
other
than
router
origin
validation,
as
well
as
to
take
advantage
of
slum
to
do
local
control.
And
then
we
choose
HTTP
l
transport,
because
it's
a
widely
supported.
It
is
with
security
enhancement
to
offer
interpreter
protection,
and
it
is
independent
from
protocol
than
you
need.
I
So
if
you
are
already
familiar
with
our
C
84
16
slum,
you
would
have
finally
easy
to
do.
Rp
cache
that
update
by
the
virtue
of
slum
filters
and
some
assertions
which
are
responsible
for
deleting
and
adding
cache
data
items
respectively.
So
that's
a
reason
why
we
choose
JSON
as
a
format
to
be
in
alignment.
We're
still
on
fire
for
mate.
In
other
word
will
slim
file
to
describe
cache
data,
and
it's
update
so
I'd
like
to
reiterate
our
motivation
to
do
these
so
aren't
here
either
over
there.
Why
bother
to
do
these
on
here?
I
I
designed,
excuse
me
for
preventing
RP
cache
their
cache
data
for
routers,
so
the
RTR
PDU
is
bounded
to
transport,
which
is
difficult
to
as
some
extensions
and
some
applications
well,
which
is
not
supported.
Rtr
cannot
read
the
binary
data
offered
by
RT
r
and
we
will
also
like
to
take
the
adventure
of
just
you
know
why
API
to
do
business
in
correlation
and
local
control,
because
the
Salam
file
is
in
JSON
format,
and
so
that's
the
reason
why
we'll
do
these?
I
So
here
is
a
desire,
HTTP
header,
for
it's
used
in
transporting
cache
data,
so
it's
inevitable
for
us
to
have
got
some
had
to
be
her
extension
to
signal
what
is
piggyback
in
a
message
body,
so
the
we
look
added
we
can
find
inside
head.
That
indicates
some
control
information,
such
as
operating
model,
time
stem
version,
information
and
a
Sigler
ship,
which
is
the
for
integrity,
production
and
also
about
a
inside
body,
which
is
actually
a
slim
file.
I
I
I
I
Also
note
the
CloudFlare
has
a
made
us
similar
efforts,
yet
in
a
slightly
different
way.
Anyway,
we'll
keep
an
open
mind
to
any
inflammation
details
and
anyway,
if
any
in
Hawai
I
interested
in
this
walk
and
can
contribute
to
this,
let's
work
together
and
I
hope
it
working
group
will
consider
whether
this
work
is.
It
is
of
our
efforts
to
seek
deny
elation.
Thank
you.
N
I
N
G
G
I
H
K
K
H
Just
want
to
interject
something
in
the
transfer
presentation
and
make
before
break
issue.
I
forgot
to
give
credit
where
dude
this
issue
was
raised
by
Nicki
alert
I
did
not
invent
it.
It's
Nick.
P
Hello,
my
name
is
Alma
Boyett
I'm
from
NIST,
so
today,
I'm
talking
about
the
draft,
be
trapeze
act
like
validation,
signaling,
which
we
proposed
either
last
ITF
or
the
ITF.
Before
give
you
an
update.
So
since
the
last
time
we
talked
about
that,
we
had
Burton
group
adoption
call
and
which
was
successful
but
had
some
interesting
discussions.
P
P
P
Furthermore,
after
the
draft
got
accepted,
there
came
some
discussion
about
the
validating,
BG
speaker.
Another
draft
out
there
that
deals
with
origin
validation
on
ebgp
and
has
second
part
second
focus
on
the
operational
modes
that
can
be
configured
and
they
came
the
part
of.
Why
don't
we
merge
all
these
work
together
in
one
thing,
so
we,
the
whole
thing,
is
a
little
bit
more
organized.
So
then
you
cup
from
the
e6
and
I.
We
had
some
conversation
about
that
and
said:
okay,
the
his
history.
P
Basically,
this
this
is
two
sections
of
ebgp,
as
well
as
the
operational
part.
Half
of
that
also
applies
to
what
we
want
to
do
and
the
other
half
the
operational
mode
could
be
taken
out
as
as
a
BCP.
So
that
was
what
we
decided
so
far,
and
maybe
some
input
from
the
working
group
would
be
would
be
good,
especially
in
regarding
to
splitting
the
validation.
P
We,
we
add
some
additional
language
or
a
little
bit
more
clarifying
language
and
regarding
to
ebgp
usage
and
clarify
error
handling,
so
the
top
part
is
currently
RFC
1897
there.
You
have
the
long
block
of
reserved
and
at
the
end
we
have
the
validation
state
and
we
update
that
by
using
the
second
to
the
last
octet
for
pass
validation
and
we
rename
the
while
relations
state
for
you
it
into
origin,
validation
to
clarify
which
validation
state
is
in
which
octet.
P
P
P
P
P
From
which
one
was
called
validating,
b2b
speaker
yeah,
so
we
propose
the
change
in
this
wording.
To
say,
implementations
must
provide
a
configuration
mechanism
to
allow
the
use
of
the
community
in
both
sending
and
receiving,
and
the
implementers
should
enable
the
use
for
this
community
on
ib
tree
on
all
ibgp
sessions
and
short
by
default
disabled
on
ebgp
sessions,
to
make
it
more
clear
and
yeah
then
error-handling
currently
1897
is
has
an
indirect
error-handling
and
they're
mainly
deals
with.
If
the
attribute
occurs
more
than
once
in
the
BGP
update.
P
P
P
B
Job
Snyder's
NCT,
unfortunately,
I
am
not
entirely
sure
you're
heading
in
the
right
direction
by
merging
the
allegedly
called
validating
Beach
P
speaker
draft
into
this
I'm,
not
sure.
If
you're
aware
of
the
extreme
controversy
that
that
draft
has
generated
and
absolutely
failed
to
reach
consensus,
it
is
very
disappointing.
You
see
it
resurfaced
yet
again.
Well,
there
are
tens
of
comments
unaddressed
and
with
an
ITF
process.
It
is
not
the
best
quality
of
work,
so
I
think
when
you
set
out
to
specify
extended
communities
to
signal
bgp,
SEC
validation,
states.
B
That
is
a
lot
about
goal,
but
then
perhaps
you
accidentally
stepped
into
something
that
doesn't
smell
as
nice
and
that
may
drag
down
the
effort
that
you
set
out
to
originally
accomplish.
In
other
words,
if
you
fade
share
with
a
draft
that
was
not
well
received,
it
will
complicate
your
original
goal
and.
P
D
P
B
Maybe
if
you,
if
I,
can
research
for
us
I
understand
where
you're
coming
from,
but
the
cider
ops
working
group
does
not
get
to
redefine
what
transiti
means
in
bgp
attributes
that
is
outside
our
scope
and
a
non-transitive
extended
community
will
not
in
this
working
group,
be
redefined
as
well.
It's
now
subtly
transitive,
and
it
is
a
mistake-
that's
in
the
eight
zero
nine
seven
draft
that
it
suggested
that
or
perhaps
it
could
be
transitive
know.
B
L
B
So
this
this
is
a
technical
issue,
because
BGP
implementations
do
not
behave
that
way.
Putting
it
in
a
draft
will
will
will
not
help.
Okay,
so
I
would
recommend
to
not
proceed
with
the
merch
and
and
proceed
with
what
you
originally
set
out
to
do.
They
will
make
the
chances
of
success
higher.
Okay,
thanks.
Q
Q
That
looks
like
a
downgrade
attack
to
me,
because
if
you
pick
the
numerically
highest
one,
that's
sort
of
the
least
desirable
validation
state
and
that's
going
to
be
less
desirable
than
the
lack
of
signaling
of
any
validation
state,
which
is
what
the
bottom
box
does,
so
that
we
can
take
it
to
the
list.
But
that
doesn't
seem
like
a
desirable
strategy.
P
To
me,
I
think
we
should
bring
it
to
the
list,
because
the
only
problem
is
what
what
I
see
currently
now
we
have
two
validation
state
in
there
and,
let's
say,
for
example,
we
have
fun
attributes,
it
says:
Origin,
validation,
invalid
and
pass
validation,
valid
and
the
other
one
says:
pass.
Validation,
invalid
and
origin
validation
valid
so
now,
I
know
I,
have
the
problem?
Do
I
start
cherry-picking
the
values
out
of
them
and
takes
the
highest
integer
value
and
I
take
information
of
multiple
attributes?
P
What
I
don't
know
if
this
is
the
right
way
to
go
as
well
or
which
one
which
one
do
I
know
choose:
I,
prefer
origin,
overpass
or
path
of
origin,
so
I
think
there's.
Definitely
some
discussion
needed
that.
That's
why
I
said
okay,
the
easiest
way
in
this
regard
to
deal
with
that
is,
if
I
have
these
problem,
what
I
have
to
solve?
Somehow,
if
I
say:
okay,
if
I
shouldn't
see
more
than
one
Ennio,
that
I
mean
that's
the
first
thing
why?
H
H
N
Jeff
has
I
own
an
implementation
with
a
policy
engine
that
has
to
look
at
this
stuff.
You
do
not
want
to
put
these
things
both
into
the
same
community.
The
primary
reason
for
this
is
there's
a
lot
of.
If
you
flip
back
to
the
community
description,
a
lot
of
people
that
have
policy
out
there
that
is
matching
against
you
know.
Basically,
a
scroll
going
go,
go,
go,
go.
E
N
More
thank
you.
There
you
go
there's
an
awful
lot
of
people
that
have
policy
that
says
no
match
community
for
three
zero,
zero
bunch
of
zeros
and
then
number
one
two
or
three,
and
it
is
done
as
an
exact
match
type
thing,
despite
the
fact
that
no
it
says
reserved
most
people
know
if
they're
doing
something
slightly
smarter
for
match
purposes
might
do
regular
expressions,
regular
expressions
suck
for
performance.
So
a
lot
of
people
don't
do
that.
N
N
P
N
L
M
L
M
Q
Q
Q
Their
handling
that
was
prepared
proposed
on
that
slide
is
sort
of
less
stringent
than
what
was
it
that
the
you've
changed.
The
semantics
you've
changed
the
semantics
to
something
that
will
accept
routes
that
were
previously
rejected,
and
that
made
my
knee-jerk.
You
know
if
you
subsequently
analyze
the
you
know
the
the
case
and
sort
of
convinced
me
that,
and
you
know,
John,
even
though
we're
letting
in
routes
that
you
were
previously
gonna
drop
on
the
floor.
It's
okay,
hey,
okay,
but
I.