►
From YouTube: IETF99-I2NSF-20170718-1330
Description
I2NSF meeting session at IETF99
2017/07/18 1330
https://datatracker.ietf.org/meeting/99/proceedings/
A
Beautifully
on
my
own
display
of
the
slide,
it
says
interface
to
network
security,
foo
and
then
on
the
next
line.
It
says:
auctions,
I'm,
Adrienne
Linder
as
Linda.
This
is
I
to
NSF.
We
have
a
full
agenda.
Blue
sheets
have
been
around
once,
but
I'll
push
them
around
a
second
time,
so
you
can
fill
in
if
you're,
missing
and
one
thing
to
add
the.
If
all
things
go
well.
This
is
my
last
session
chairing
you
I'd
like
to
say
it's
been
a
joy
and
hopefully
you'll
be
successful
going
forward.
B
Thank
you
very
much.
Let's
get
started,
ok,
first
of
all,
is
it
notes?
Well,
so,
hopefully
everybody
really
well.
This
is
the
new
version
with
so
please
read
it.
So
here
is
some
trivia,
so
charter
I
think
I.
Don't
have
to
do
repeat
that
mailing
list
for
newcomers
and
we
have
minute
taker
and
Frank
and
Sue.
Thank
you
very,
very
much
and
for
everybody
else.
Since
we
have
etherpad
right
there,
there's
a
link
from
the
IETF
99
agenda.
Please
just
go
there.
B
Add
your
comments
or
name
especially
when
people
can
now
record
a
name
correctly
and
if
they
mistaken
your
notes
and
please
correct
that
online.
So
we
can
all
contribute
to
the
minutes
and
blue
sheet.
Okay,
so
here's
agenda
and
being
posted
and
we're
gonna
get
started
Wow.
So
so
here's
agenda
agenda,
passion
and
if
there
any
questions
anything
is
agenda.
Changing
if
not
we'll
just
go
as
this
plan.
Okay,.
B
So
congratulations
to
the
problem
statement
and
use
case
document.
Authors
may
finally
get
one
RFC.
Thank
you
and
the
second
one
to
be
out
is
the
framework
we
have
completed.
The
working
group
last
call
and
going
to
be
sent
to
iesg
for
review
so
that
two
other
working
group
documents,
the
working
group
have
decided
to
leave
them
current
in
the
working
group
until
the
working
group
charter
either
close
or
each
other,
so
we'll
leave
them
open,
as
we
can
move
forward
that
that's
the
terminology
and
the
gap
analysis.
B
So
there's
another
working
group
document
on
the
client
interface
requirement
and
we're
going
to
decide
if
we
want
to
move
that
forward
to
the
working
group.
Last
call.
So
here
are
five
four
or
five
drafts
ready
for
working
group
adoption.
We
can't
discuss
that
further
on
the
mailing
list.
They
are
primarily
information
models
and
data
models,
and
we
have
quite
a
few
contributions
on
the
information
models
and
data
models,
and
so
initially
there
was
some
inconsistency
among
them.
B
This
morning,
thanks
to
all
the
ulcers
we
got
together,
we
had
a
really
productive
discussion
in
the
morning,
align
the
models
and
decided
on
the
relationship
among
all
the
information
model,
drafts
and
data
model
drafts
later
on
sue,
Harris,
I'm
gonna
talk
about
those
relationship
and
how
we
move
forward
with
all
those
drafts.
So.
B
A
And
just
to
remind
people
that,
as
a
working
group,
we
looked
at
whether
or
not
to
have
information
models
and
made
a
decision
that
they
were
going
to
be
useful
for
us.
That
is
a
decision
we
could
revisit
if
we
want
to,
but
where
we
are
at
the
moment,
is:
let's
push
forward
and
have
information
models
to
to
drive
our
data
models,
and
so.
C
C
First
of
all,
information
models,
we
treated
as
blueprints
for
data
models;
okay,
and
there
are
two
types
of
alignment,
exact
door,
a
subset-
and
we
also
worked
on
this
based
on
which
is
your
core
concept
or
which
is
your
key
model.
And
then
you
group
your
models
around
the
core
concept
and
we
want
it
to
be
easy
to
understand
so
that
we
can
provide
models
which
can
go
to
then
vendors
and
people
who
are
trying
it
in
academia.
Like
our
hackathon
presentation,
which
is
next
with
Paul
Shaw,
we
want
easy
to
understand
models.
C
That's
why
yang
was
written
as
it
is
go
ahead
to
the
next
one.
So
let
me
talk
first
about
why
blueprints
are
the
basis
for
conflict
stripes,
as
you
see
we're
in
Prague,
and
this
is
one
of
the
largest
structures.
But
when
people
build
a
building
like
this,
they
start
with
a
portion
of
the
building
and
then
they
build
up
the
structure,
and
the
same
is
true
of
an
information
model.
We
may
use
the
where
you
were
treating
the
information
model
like
a
blueprint
for
this
large
building.
C
It's
it's
got
a
lot
of
structure,
but
maybe
we
only
do
so
much
in
the
data
model.
Some
working
groups
that
are
doing
yang
models
think
that
the
information
model
should
be
exact.
There
are
pros
and
cons.
That's
just
the
way
we
were
sort
of
talking
about
how
our
capability
models
and
our
capability
model
our
customer
facing
model,
our
NSF
final
facing
model,
so
I'm,
just
providing
some
thoughts
that
we
had
from
the
authors
go
ahead.
Then
it's
very
important.
C
You
invite
verify
the
blueprint,
because
if
you
don't
like
this
building
doesn't
work
so
well
together.
Maybe
you
have
a
floor
sitting
out.
So
if
you
have
the
data
model
equal
the
information
model,
then
you
can
check
the
data
model
out
with
tools.
If
you
do
like,
we
do
we're
suggesting
and
you
do
the
information
model
as
a
future
blueprint
which
will
guide
you
in
your
data
models.
You
must
actually
take
a
different
approach.
C
There
is
a
downside
if
you
use
the
case
where
you
have
the
data
models,
equal,
the
information
model
and
you
trust
and
checking
it.
You
go
all
the
way
down
to
the
data
model
with
lots
of
work
before
you
actually
check
it,
and
you
may
have
a
large
cost
like
this
building.
That's
not
quite
aligned.
So
that's
why
we
thought
okay,
we're
going
to
use
the
information
like
a
like
a
blueprint,
but
we
have
to
carefully
hand
check
that
all
the
information
models
work
together.
This
is
just
one
approach.
C
As
I
said,
you
can
use
the
information
model,
equals
a
data
model
and
then
trust
after
you've
got
all
the
way
through
that
the
data
model
checks,
checkers
will
work
or
you
can
do
what
we're
suggesting,
which
is
to
hand-checked
synchronize.
Look
at
the
information
models,
use
them
as
a
blueprint
and
implement
subsets.
That's
how
people
build
large
buildings
and
we
thought
that
might
work
go
ahead.
So
that's
the
recommendation
verify
the
complete
information
model.
C
That
means,
if
you
have
five
information
models,
you
must
verify
them
by
hand
and
then
build
a
portion
of
the
model
in
stages.
Simple
than
more
complex,
that's
the
approach.
Our
teams
and
development
teams
have
been
taking
well
well,
the
well
the
team
from
Korea.
All
the
people
were
here
at
the
hackathon
stand
up
and
be
recognized
at
this
point
for
their
work
on
the
hackathon
code,
so
I
guess
especially
the
students.
C
So
we
also
have
people
who
are
working
on
product
code.
We
think
that,
with
simple
more
complex,
we
get
an
understanding.
Now
this
is
possible
because
of
the
good
work
to
base
everything
on
the
Keen
model
of
capabilities
go
ahead.
Linda
capabilities
is
like
our
key
structure
that
that
you
may
see
in
Prague,
in
of
the
Cathedral
and
some
of
the
castle
here
that
around
which
was
built
all
the
family
models
and
the
structures
in
Prague.
That's
how
the
cast
was
been
hi
Hank
glad
you're
here.
C
So
why
and
how
the
data
model
I,
think
I've
gone
through
that
simple
set
sets
help
you
get
miles.
Accept
it
now,
there's
a
real
benefit
in
this
that
this
approach
fits
right
in
with
the
net
compressed
comp
work
in
their
data,
catalog
and
registry.
What
you
can
do
based
on
their
stuff.
Is
you
can
say,
I
want
this
subset
to
go
in
my
data
store
for
configuration?
Ok,
and
these
are
the
model
drafts
in
that
come
that
will
allow
this.
C
This
is
why
we
think
this
approach
may
be
successful,
for
people
actually
deploying
now
again
the
data
mantel
catalog
and
the
open
config.
If
I
remember
a
right,
Hank
and
those
in
the
hackathon,
they
finished
their
their
hackathon
version
of
it.
So
again,
there's
hackathon
early
code,
we're
following
early
code
movements,
go
ahead
to
the
next
one.
Now
that
was
a
long
explanation
to
go
to
this
slide
at
the
top.
We
have
an
information
model
that
goes
around
and
here
I'm,
going
to
dance
outside
of
the
box.
I
apologize
with
a
pointer.
C
C
You,
okay,
so
over
here
is
our
key
model
that
key
structure
like
the
castle
with
the
the
capability
model,
then
we
have
a
client
facing
information
model,
a
high-end
high-end
edge,
astray,
ssin
model,
the
monitoring
model
and
then
below.
The
capability
model
is
some
detail
from
Frank
on
the
security
policy.
This
creates
an
information
model.
C
Now,
as
you
look
and
you
review
this,
please
tell
us
what
you
think
now
the
bills
back
here
for
Kumar,
but
this
one
and
our
group
thinks
that
that
was
discussing
things
this
morning
really
well
thinks
that
the
client
facing
model
ought
to
take
an
approach
of
intent.
Okay,
how
many
people
know
what
declarative
and
intent
models
are?
Okay,
very
few
pardon
me,
while
I
I
give
the
rest
of
you
a
moment.
C
The
clarity
of
verses,
descriptive
is,
if
your
descriptive,
you
describe
how
to
go
to
the
store,
I
need
to
go
to
the
store
to
get
aspirin.
I
need
you
to
go
down
the
street
over.
There
find
the
drugstore
go
inside
pick
up.
The
aspirin
bottle
pick
up
the
excedrin
bottle
and
come
back
declarative
says:
please
go
to
the
store
pick
up:
aspirin.
C
Okay,
the
declarative
model
in
a
client
is
much
more
powerful
because
the
client
doesn't
have
to
know
what
the
interior
of
the
security
model
it
says,
secure
a
path
from
these
two
to
these
two
or
engage
in
a
threat
and
a
levin
or
some
complete
capability,
so
Nabil
I,
hope,
I
gave
a
good
description.
We
all
felt
like
this
client
model
should
be
declarative,
we're
working
on
that
these
models
are
still
in
the
info
model.
Here
a
work
in
progress,
but
we've
made
some
progress
along
that
and
that's
what's
different.
C
All
the
rest
are
probably
what
you
expect
for
info
model
now.
The
reason
we
take
this
is
some
of
this
is
new
thoughts,
and
this
blueprint
will
help
us
build
a
good
castle
down.
Here.
We
took
the
build
a
little,
so
this
capability
model-
that's
underneath
under
the
same
point,
is
a
subset.
Now
we
think
this
is
a
workable
subset,
based
on
the
the
hackathon
code
that
works
underneath
in
the
base
model.
C
It
may
be
an
incorrect
subset,
but
this
is
for
you
to
review
and
to
work
on.
There
is
hackathon
code,
so
try
it.
You
know
this
is
a
build
and
try.
There
is
the
interface
data
model
which
is
trying
to
take
it's
just,
not
a
hundred
percent
aligned
with
Kumar,
but
was
enough
to
do
the
hackathon
stuff.
So
you
have
to
look
at
the
Kumar
model
and
the
Kim
NSF
facing
models
to
see
what
you
think
is
a
declarative
thing
here.
C
It's
secure
connection
for
this
time,
this
location,
it's
young
customer-
excuse
me
I,
pointed
it
the
wrong
it's
Jung's
customer
facing
model
that
is
equivalent
and
all
the
comments,
I
just
said
should
have
said
come
where
I
pointed
it.
Kim
Kim
is
at
the
NSF
facing
model,
which
is
the
thing
that
goes
between
the
controller
and
the
actual
NSF
discussion.
I.
Think
we've
gone
through
this
type
of
discussion
quite
a
bit,
so
I
won't
repeat
it:
how
or
hands
anybody
want
me
to
repeat
the
substance
of.
C
C
We
think
that
the
capability
model
from
the
team
is
ready
to
go
along
with
this
and
as
our
key
model,
just
like
the
key
building,
we're
going
to
ask
the
working
group
from
the
authors
to
adopt
both
then
we'll
ask
you
to
adopt
other
pieces
from
the
models,
questions
on
our
little
review
and
and
these
models
and
structure
is
it
after
lunch.
Or
am
I
confusing?
Okay,
no
questions
I'm
impressed.
B
E
So
what
we
have
is
precisely
working
here
is
about
an
open
model
in
which
you
can
deal
with
different
set
of
capabilities,
so
in
the
sense
that
we
are
open
to
registering
different
kind
of
stuff-
and
this
is
very
important,
because
what
we
have
is
a
placeholder
for
the
diff
for
the
other
other
different
models,
as
the
ideas
are
not
that
people
have
to
take
their
capabilities
and
go
for
the
capabilities
and
and
specialize
them
or
extend
them,
is
about
that.
The
capabilities
mobile,
provide
the
this
support
for
declaring
the
capabilities
and
manipulating
I
mean.
B
E
Both
of
them
should
be
based
on
the
I
mean,
if
you
ask
me,
I
think
that
they
should
be
based
on
the
capability
model:
okay,
okay,
and
whether
whether
we
have
whether
we,
what
we
make
is
a
general
model
that
is
more
detailing
one
case
of
the
other.
This
is
up
to
the
group
and
that's
something
that
remains
to
be
decide
so
far.
Okay,.
B
B
H
So
the
objective
of
this
hackathon
is
to
show
the
validity
our
design
of
their
attitude
driven
security
management.
So
this
time
seven
credit
student
from
my
university,
a
sketchy
you
we
prepare
for
less
than
three
months
us
and
then
this
last
the
weekend
that
we
tried
to
tune
our
cold
and
prepare
for
the
presentation.
So
next
slide.
H
So
this
hackathon
is
collaboration.
Our
work
with
the
university
and
Korea
Telecom
is
I,
see
a
vendor
major
I
spend
in
South
Korea
and
also
a
tiara,
a
each
government
supported
Research
Institute.
So
three
organizations
are
working,
a
wait
for
the
hackathon,
so
the
correct
telecom
they
provide
in
the
year
the
applications
for
I
to
NSF
services.
Next
slide,
please.
H
So
this
portal
shows
last
week
an
student
and
Korea
telecom
researcher
and
I
are
working.
Also
children
also
trophy
our
customers.
So
we
work
together
and
also
we
presented
our
work
and
then
this
time
we
are
at
first
University
a
world
again
so
that
year,
soul
meeting
we
get
a
best
so
indicates
our
working
group
are
working
very
hard
next
slide.
So
next
you
already
know
the
concept
of
I-10.
H
Instead,
so
the
goal
of
this
hackathon
is,
we
tried
to
show
the
first
application
is
if
I
were
the
second
one
is
a
content-based
security
such
as
a
web
filter
and
then
a
week
combined
using
a
service
function
chaining
as
I
proceed,
we
showed
the
according
to
the
frank
information
model
case,
the
investigate
the
maquette
using
the
fiber
and
then,
if
this
content
should
be
investigated
in
detail
in
that
case,
that
factor
should
be
poor
word
to
the
DVI
deep
packet
inspection.
So
in
order
to
that,
we
implemented
the
opposite.
H
This
time
we
try
to
follow
the
separation
working
groups.
Solution
next
slide,
so
contribution
orbit
is
hecka,
so
we
showed
that
the
POC,
the
promo
concept
for
I
to
NSF
using
open
sources
and
also
we
showed
that
the
validity
of
our
simple
data
models.
So
even
though
the
data
model
is
a
simple
but
I
looked
like
at
a
powerful
and
then
we
show
the
peas
ability
of
a
data
driven
approach
next
slide,
so
those
are
a
bunch
of
open
sources.
H
So
we
especially
this
time
we
use
the
x8
XSLT
extensible,
stylesheet
languages
and
transformations,
which
means
high-level
the
policy
specified
I
too
and
I
step
user,
that
high-level
policy
translated
into
low
level
policy
will
be
delivered
to
NSF
backslide.
So
this
figure
showed
another
computation
so
left
hand.
Side
is
enterprise
network
right
hand,
side
is
service
website,
Facebook
web
YouTube.
So
this
time
we
implemented
the
fiber
and
the
web
filter,
which
means
the
paste
on
the
Porter
port.
Information
such
as
a
web
port
is
80
right.
H
So
if
traffic
is
from
some
clients,
such
as
employee,
the
traffic
is
aware
we
can
compute
out
the
second
one
is:
if
the
content
of
the
web,
the
packet
contain
a
YouTube
or
Facebook,
we
can
peer
table,
which
means
we
can
perform
our
web
filter
next
right.
So
this
figure
shows
the
computation
next,
so
we
can
show
the
fire
next,
the.
Secondly,
this
one
is
repeater.
H
Remember
we
implemented
the
using
I
suppose
she
so
service
function
forward
and
then
original.
Only
the
the
foil
packet
is
forward
to
the
web
filter
NSF
using
SFCC.
Next,
please
so
we
unloaded
our
document
and
the
open
source,
our
code
into
github,
and
also
we
unloaded
our
the
video
clip
the
explaining
how
to
install
and
how
to
test
next
slide,
so
this
figure
should
heat
up
next
light.
The
finally
are,
we
cut
are
many
lessons
from
the
last
two
hackathons
and
we
showed
a
POC.
H
B
Thank
you
very
much.
They
also
got
an
award
for
this
hackathon.
Thank
you.
Just
a
little
extra
note
from
hackers
are
so
hackathon
is
I,
Kiev
work
and
just
recently,
not
too
long
ago,
Linux
Foundation
start
a
new
project
called
security
controller,
open
source
project
they're,
like
five
companies
currently
in
there
like
Intel,
our
Pato,
Huawei
I,
think
that's
two
other
I
don't
know
Kia,
maybe
in
there
so
I've
approached
them.
They
were
supposed
to
come
here,
gave
a
presentation
but
kind
wise,
it's
too
early
for
them
to
really.
B
We
come
to
do
the
presentation
so
next
time,
if
we
meet
they'd,
probably
be
able
to
give
us
an
update
what
they
are
doing
and
they
are
hoping
to
use
iTunes
SF
interface
as
the
API
is
to
to
guide
their
work.
Their
work
is
mainly
for
the
controller
part,
just
like
the
the
logistics,
like
the
logics
of
getting
the
interface
from
client,
be
able
to
translate
into
and
as
they're
facing,
to
face,
Kathleen.
B
J
Have
a
question
power
to
how
future
plane
to
maybe
in
the
next
I
TS
meeting
to
two
more
excellent
test
about
new
interfaces
that
are
at
the
graduating
interface,
although
our
clients
facing
the
faces,
because
we
I
think
after
this
meeting
we
have,
we
have
we
output
song
a
new
new
content
about
this
part.
So
I
think
if
we
can
do
heck
song,
it
will
be
very
helpful.
Yeah.
H
Basically
are
currently
we
already
implemented
arm
the
monitoring
monitoring
on
the
information
model
so
and
also
this
time
we
implemented
the
content
security
so
such
as
a
picture
so
so
information
our
model
are
getting
the
more
detailed
information,
I'll
really
plaque
the
dead
one
into
era
model,
and
then
we
can
implement
it.
Of
course.
Next
the
100
and
the
singapore
meeting,
and
so
we
implemented
that
once
okay
tiago
have
it's.
E
More
related
to
yeah,
open
source
stuff,
when
I
mean
normally
open
source
projects,
are
very
reluctant
to
to
making
anything
that
sounds
like
they
come
will
tell
us
what
what
they
are
doing
and
we
can
try
to
influence
them.
In
fact,
I
knew
about
this
project
some
time
ago,
because
world
it
was
in
the
minds
of
some
people
that
are
working
with
us.
Well,
it's
rather
interesting
and
we
I
try
to
push
them
to
use
it.
E
They
join
us
at
five
years
and
and
well
in
confident
they
will
be
coming
here
and
we
will
have
even
managed
to
convince
that
needs
a
couple
of
guys
in
Intel
to
participate
in
the
India
to
NSF.
Fortunately,
I
am
on
it,
so
it's
I
hope
we
will.
We
will
manage
to
have
them
and
well.
It
would
not
be
a
bad
idea
if
you
bring
some
of
your
results
there
as
well
could
be
because
this
is
this
is
what
we're
running
cold,
and
this
is
what
they're
looking
for.
Okay.
B
J
Hello,
everyone
friendship
are
we
okay,
I
will
give
a
reduction
of
our
latest
I
honestly
have
capability
information,
motor
shaft,
since
it's
a
it's
a
basic
information
motor
for
a
lot
of
vitamins,
safety
to
motor
work,
so
I
warmly
welcome.
Everybody
can
give
us
put
feedbacks
okay.
Unfortunately,
let
me
let
me
say
that
I
think
the
whole
point
of
the
key
point
of
this
information.
Motor
draft
is
that
wait
if
I
are
where
we
proposed
an
information
model
for
capability
for
the
security
capability.
So
that's
that's
a
beginning
of
of
the
hosting.
J
J
It
is
not
dependent
on
what
is
is
device
motor
or
its
visualized
or
it's
physically.
Yes,
they
are
there
or
not
relevant,
but
we
want
to
define
the
logical
function
which,
which
you
are
gonna
safe,
can
provide,
and
so
it
is
abstract
concept
and
they
it
should
be
very
flexible
to
to
represent
the
functionality
of
the
security
functions.
J
Okay,
so,
and
in
addition
to
these
particular
concept,
of
course,
we
also
need
to
to
define
how
to
operate,
how
to
manager
this
capability,
because
our
following
work
is
based
on
how
how
can
we
defined
and
then
we?
How
can
we
manage
it?
So
then,
I
think
if
we
have
the
capability
information
motor,
then.
K
J
J
J
It's
really
helpful
because
because
it
photo
that
you
can
see
that
in
we
have
the
so
based
on
this
motor,
it
can
be,
for
example,
it
can
be
the
super
information
model
or
it
can
be
another
good
information
model,
but
it
did
provide
us
some
very
useful
policy
information
element
that
we
can
use
to
construct
our
security
policy
motor
and
that's
a
that's
a
that's
a
step,
but
it
but
that's
otoscope
of
I.
Don't
a
safe
work,
that's
a
yeah!
We
can
use
it
and
then
we
have
to
to
branch
to
inferior
footprint.
J
These
very
basic
or
information
motor.
The
left
branch
is
that
we
we
try
to
define
what
is
the
security
policy,
so
the
the
security
policy
that
we
could
say
that
Calgary
we
use
that
you
see
a
motor.
So
it's
it's
called
as
a
event,
condition
and
action.
We
use
this
motor
to
to
to
describe
the
any
kind
of
security
policy,
so
so
it
means
that
it
means
that
we
we
use
different
kinds
of
we
went,
object,
condition,
object
and
the
actual
object
to
compose
our
security
policy.
J
That's
a
physical
way
and-
and
if
you
can
see
from
this
picture
it's
very
similar,
like
some
class
definition
we
can
based
on
where
is
very
an
initial
class,
and
then
we
inherited
we
extended
to
define
our
our
model
and
in
the
in
the
in
the
in
the
right
branch.
It's
our
capability
information
motor
definition.
It's
a
little
bit
different
from
the
security
policy
actuary
that
hability,
whom
each
model
is
about
what?
What
is
the
function
of
these
and
SS?
J
J
J
It's
it's!
How
is
when,
because
before
I
use,
any
security
capability
I
need
to
know
what
we
have
so
I
need
to
some
you
can
you
learn,
sign
information
model
to
express
that
which
functionality
or
which
feature
that
unless
it
has
so
that's,
we
need
another
way
to
express
it.
So
you
so
that's
another
information
model,
yeah,
and
but
if
you
see
our
title
actually
the
security
policy
and
the
capability
information
model,
they
are
the
two
purity
they
ensure
they
are
the
same
thing,
but
the
use
in
the
different
objective.
J
In
the
later,
you
will
see
how
we
define
them
respectively.
Okay,
so
let
me
emphasize
our
way
to
define
the
whole
information
model
actually
with
to
achieve
the
scalability
and
to
achieve
a
modern,
driven
approach.
So,
of
course,
we
cannot
defy
certifying
a
lot
of
security
policy,
motor
and
capability
model
for
every
kind
of
network
security
function,
so
other
ways
that
we
foster
a
way
to
find
several
several
kind
of
capabilities.
J
They
are
the
very
physical
element
and
then
based
on
this,
based
on
our
define
the
capabilities
we
we
use
our
security
policy
information
motor
to
use
them
to
to
meet
different,
require
it
requirements.
So
everything
starts
from
the
capability
information
model.
Ok,
ok,
so
I
I
think
I
have
I
have
already
introduced.
What
is
our
current
model?
Is
the
ECA
policy
motor
so
event
the
condition
action
policy
rule.
But
if
you
of
course,
it's
a
container
to
contain
all
these
elements,
but
in
addition
to
ECA,
we
also
think
that
we
need
another.
J
We
can
even
to
make
our
policy
more
flexible
and
more
useful.
So
we
we
added,
we
added
the
resolution
strategy,
so
it's
used
to,
for
example,
I
have
several
security
policy
rule,
so
we
we
can
use
this
resolution
strategy
to
to
decide
how
to
use
this
rule,
how
to
combine
the
area
that
they
are
met
with
that
to
decide.
My
final
result,
my
final
action,
sorry
so
so
that
is
very
useful
to
construct
a
very
complex
or
very
intelligent
policy,
and
also
we
have
the
metadata.
J
It
is
also
used
by
the
by
the
by
the
resolution
strategy.
2
T
Phi,
also
also
to
define
how
to
map
to
the
final
actions
and
also
we
have
default
actions,
which
means
that
if
I
cannot
match
all
the
conditions,
what
can
I
do
so
we
can
do
the
default
action.
So
that's
our
country,
all
the
element
of
the
security
policy.
J
Ok
and
after
introducing
the
security
policy
I
want
to
introduce
either
I
want
you
to
use
the
capability
information.
Modal,
so
so
so,
if
you
can
follow
my
introduced
introduction
before
so,
the
capability
information
order
is,
is
I.
Think
it's
obvious
to
you.
So
so
it's
it's
just
another
expression
way
another
expressing
way
to
define
the
capability
model.
So
here
you
can
see
that
we
have
those
basic
element.
We
have
the
action
set.
We
have
the
condition
set.
We
have
the
event
set
and
the
resolution
set
and
default
action
set.
J
So
we
use
some
capability
algebra.
You
can
consider
it
as
a
set
theory.
We
we
listed
all
these
all
these
information,
all
these
capabilities
as
a
whole
to
express
what
I
can
provide
in
the
security
perspective
to
our
user,
for
example,
I,
have
the
our
action
set,
which
we,
which
include,
deny
and
pass
so
that
means
that
this
this
network
network
security
function
can
can
and
confer
to
the
traffic.
Oh
it
and
it
can
also
pass
a
traffic
order
or
other
actions.
J
Okay,
so
so,
by
this
way,
network
security
function
can
very
easily
to
express
what
you
can
do
to
the
security
controller
or
to
the
client.
So
so,
okay,
and
then,
if
we
have
this
kind
of
definition,
we
can
do
some
more
some
more
calculation
or
some
more
operation
on
then
down
below.
We
can
do
the
addition
and
the
subtraction
we
can.
We
have
to
tablet,
have
a
period
here.
They
are
two
to
two
functions
set
and
we
can
do
the
addition
of
them.
J
So
that
means
that
we
can
get
a
more
powerful
capabilities
and
we
can
do
subtraction,
which
means
that
I
don't
need
some
of
the
capability
in
that
motor,
so
I
can
I
can
subtract
it
in
the
future.
I
think
that
maybe
we
can
add
more
on
this
operation,
so
I
think
I'm,
always
I'm
almost
finished
about
this
inhumation
mode
of
interaction
so
about
future
work.
J
I
just
want
to
say
that
for
for
make
our
document
content
is
short,
so
we
move
a
lot
of,
for
example,
or
specific
information
motor
at
attribute
to
to
the
appendix
part.
So
they
are
trusted
for
the
example
to
show
how
to
use
it,
but
of
course
they
they
will
never
beat
a
complete
set
of
the
security
policy
information
model
and
the
capability
information
motor.
So
our
question
to
the
chair
or
working
group
that
I
think
I
think
maybe
it's
not
a
question,
but
we
just
want
to
say
that
we
don't
want
to
do
I.
J
Think
that
cannot
finish,
which
after
leaves
the
some
typical
example
in
the
appendix
and
that's
that's
current
way.
We
do,
and
and
the
next
next
page,
please
and
I
think
we
still
think
about
another
enhancement
to
current
information
motor,
for
example.
We
want
to
do
in
addition
to
current
poor
operations.
Maybe
we
can
do
more
on
the
condition
and
the
event
clause.
J
We
can
do
more
things
on
our
event
and
the
condition
clause
such
as
we
can
for
action.
We
can
define
different
policies
on
how
to
implement
the
action
we
can.
We
can
actually
the
first
action
or
Exeter
execute
that
the
last
action,
or
depending
on
the
priority,
execute
which
action
so
I
think
there
are
all
the
details,
technical
details
that
we
may
be
needed
to
add
in
the
future
version.
We
are
considering
this
and
the
way
where
we
can
anybody
can
give
us
more
otherwise
or
good
suggestions.
J
J
For
example,
it
can
be
loaded
and
wrong
and
in
England
one
time
so
so
it
is,
it
is
managing
in
the
management
aspect
you
it
is
very
efficient
and
it
can
save
a
lot
of
manual
configuration
and
the
menu
modification,
but
it's
more
complex
than
current
static,
inheritance
class
design.
So
we
are
where
we
are
not
yet
decided
if
we
should
choose
this
way
to
to
make
some
modification
to
our
current
model
design.
So
yeah
we
like
to
hear
more
suggestions.
B
J
A
H
Hello,
this
is
a
person,
so
destruct
motivation
is
applicability.
Is
the
one
of
our
milestone
item
so
the
based
on
our
the
hackathon
experience?
We
show
out
I
2
msf
can
provide
security
services.
The
one
thing
to
note
is
the
I
to
NSF,
provide
empirical
TNF
and
also
a
physical
data
security
function.
Also
virtual
trip
function
so
in
turned
out
using
the
Sdn
Software
Defined
Networking
can
leverage
the
security
service
next
slide.
H
So
the
combination
of
AI
to
NSF
Sdn
case
s,
TN
s
which
can
perform
perform
the
simple
to
firewall
services
such
as
a
polar
drop
mirroring,
and
then
the
complicated
security
services
can
be
done
using
the
more
complicated
NSF
learning
on
top
of
an
Apple
II
or
physical
yeah
devices
so,
and
also
I
to
another
policy
rule
enforcement
case
we
can
are
using
Estelle
suite
and
an
Apple
II
or
physical
devices,
the
net
a
function.
We
can
using
two
things
to
improve
the
performance
next
slide.
H
So
today,
why
are
we
have
augmented
our
framework,
which
means
the
the
bottom?
We
can
have
the
STL
a
network
or
having
a
stern
switch.
So
at
the
resort,
security
controller
and
a
switch
controller,
usually
a
switch
controller
arm.
We
can
say:
STNG
controller
can
talk
to
each
other
using
and
as
a
facing
interface
next
slide.
H
H
Information
should
to
be
enhanced
in
terms
of
ECA
up
the
paradigm
and
NSF
a
face
in
case
I,
think
at
the
mature.
So
we
I
think
at
least
security
that
all
security
functions,
such
as
the
fire,
is
fully
implemented
and
also
restoration
interface
case.
We
can
register
NSF
and
NSF
capability,
so
we
can.
We
are
planning
to
implement
next
hackathon,
so
this
applicability
case
we
can
show
these
three
interfaces
defined
I
to
NSF
can
be
used
next
slide,
so
this
figure
shows
that
the
flow,
so
you
can
aware
right.
H
H
So
this
draft
explain
the
three
use
cases
if
I
were
for
your
IP,
dpi
and
DDoS
attack
mitigation
services.
So
due
to
the
time
limitation
can
skip
you
can
our
leader
the
slide
over
trapped
next,
so
you
can
skip
yeah
so
use
cases.
I
mentioned
the
three
user
cases
next,
so
we
can
skip
the
use
cases
the
next
next
next
next
next,
so
our
next
step,
we
can
also
include
more
user
cases
such
as
SFC
piss,
the
spirit
of
function.
H
Chaining
will
be
added,
such
as
the
fiber
and
the
web
filter
and
fiber
and
DDoS
attack
mitigation.
We
can
combine
using
SFC,
so
a
separation
is
also
one
of
our
groups
trapped.
So
also
we
can
reflect
the
okay
Frank
information
model,
so
we
can
improve
our
NSF
or
facing
interface,
considering
content,
security
and
authentication
next
time.
Okay,
thank
you
for
attention.
If
you
have
questions
or
comments,
please
go
ahead.
L
M
Hi,
this
is
Hank
I'm
apologizing
for
the
mix-up
of
these
lights
here,
okay,
so
we
are
starting
with
the
information
model
for
monitoring
of
network
security,
functions
and
yeah,
which
does
give
the
most
important
part
here.
So
it
is
ultimately
intended
to
deal
with
the
huge
amount
of
information
that
will
be
emitted
by
functions
and
that
we
don't
only
mean
the
I
to
NSF
functions
that
will
be
standardized,
but
also
by
the
legacy
interfaces
that
are
already
pushing
a
lot
of
information
as
notifications
today.
M
So
the
content
here
is
not
only
the
I
to
NFS
events
that
are
of
interest
because
they
are
already
processed
by
a
rule
or
some
are
outputted
by
the
I
to
NSF
system
itself,
but
also
raw
data
that
may
be
of
interest
and
was
not
assessed.
Yet
the
policy
here
or
the
general
rule
here
is
that
the
consumer
of
those
notifications
decides
so
we
again
reuse
the
existing
model
of
capabilities.
Are
you
see
a
policy
rules
and
just
allow
them
to
consume
notifications
that
are
coming
from
outside
the
scope
of
the
I
to
an
SF?
M
There
is
rough
consensus
in
the
terminology
team
John
stress
na
who
was
basically
in
charge
of
the
ECA
logic
agreed
to
this,
but
we
have
no
text
for
this
because
John,
unfortunately,
I
was
unavailable
in
the
last
few
weeks.
So
unfortunately,
is
also
not
on
site.
We
skipped
the
drop,
the
stuff
of
the
alerts
and
the
alarms.
There
was
too
much
discussion
about
an
arbitrary
definition
of
classes.
M
Here
there
will
be
an
implicit
definition
due
to
the
chaining
of
policy
rules,
though
so
this
escalation
of
severity
or
confidence
will
still
occur,
but
they
will
just
be
named
events
at
the
first
time
and
be
annotated
with
metadata,
basically
telling
what
others
we
at
the
hackathon
were
exploring.
Next,
at
least
the
potential
solution
drafts
there
are
existing
in
the
scope
of
the
Netcom
realm,
so
we
basically
implement
its
cutting-edge
firmware
was
deployed
on
friday
and
software,
the
NCC
and
the
NC
client
open
source
office.
Everyone
github
on
Wednesday
this
weekend
at
the
hackathon.
M
We
were
able
to
do
periodic
updates
and
on
change
updates
from
the
USA
and
the
PRC,
so
globally
spending
creating
a
notification
flow
that
was
then
consumed
by
a
component.
This
was
first
intention
was
to
build
a
component,
and
then
we
realized
that
a
second
component
in
this
scope
of
first
proof
of
concept
is
pretty
much
looking
like
AI
to
NSF
component
and
also
the
function
capability.
M
Semantics
are
quite
similar
so
that
we
little
bit
saw
this
coming
due
to
the
convergence
of
the
two
terminology
drafts,
but
now,
whether
running
code,
we
can
actually
understand
this
better
and
the
rule
used
here
to
create
AI
to
NSF
events
was
the
filter
expression
for
a
subscription
on
to
a
or
multiple
yang
modules
and
su
who
was
in
the
room
where
you
probably
love
to
hear
that
we
did
III
as
unchanged,
updates
and
also
always
bf
table
at
its
edges.
Okay,
so
talking
about
legacy
interfaces
next
slide
yeah
this
pretty
much
already
spoiled.
M
So
we
will
use
the
notification
system
because
notification
is
this
thing
and
yang
also
a
life
data
we
will
push
via
yang.
Probably
at
least
this
is
our
route
that
we
pursue
at
the
moment.
Poor
mechanisms
has
very
much
more
of
a
legacy
thing,
as
are
the
log
files
that
are
created
or
in
retain.
We
call
those
records
because
it's
not
only
our
blocks,
it's
also
databases,
so
a
specific
agent
with
specific
rules.
We
create
events
from
those
quick.
M
We
also
have
an
early
proposal
of
a
categorization
system
for
information
monitoring,
because
how
you
get
it
like.
If
it's
a
query,
it
was
it's
a
subscription.
What
it
is
about
is
his
life
is
a
record.
Is
it
declarative
guide
and
how
you
should
be
imperative,
guys
how
it
should
do
things
or
an
assessment
already
that
the
outcome
of
a
rule?
Is
it
a
raw
aggregation
or
already
a
complex
set
of
rule
result?
And
then
again,
what
about
the
function
is
irrelevant
here?
Its
context,
as
introduced
by
John
stress,
know
he
would
elaborate
on
this.
M
Unfortunately,
he
is
not
one
side,
so
the
next
set
is
we
investigated,
of
course,
next
to
yang
subscription
other
ways
to
create
telemetry,
there's
an
effort
going
on
with
co-op
pops
up.
Unfortunately,
authors
meet
tomorrow
for
their
breakout
meeting,
so
I
couldn't
get
updates
on
that
today.
Excellent
hypocrite
is
under
severe
restructure.
It
will
be
fully
compliant
to
XMPP
and
jabs,
and
the
XMPP
protocol
console
is
at
the
moment
confirming
that
this
is
a
standard
or,
as
paconne
can
could
be
a
standard,
as
was
basically
not
violating
any
XMPP
rules.
M
So
there
are
alternatives
in
the
way
to
get
telemetry
not
for
only
for
monitoring
but
also
between
functions,
but
this
is
again
future
work.
So
next
steps
are
deciding
on
here
what
to
do
with
the
categorization
model.
You
saw
a
property
here
on
the
sides
and
we
will
also
have
to
focus
on
intervals
of
freshness
of
information,
because
that
might
be
of
relevance
to
security.
So
this
was
how
fast
enough
so.
C
Thank
you
very
much.
One
of
the
challenges
we
had
in
routing
hat
was
going
through
the
categorization.
The
fact
that
you've
actually
done
that
in
the
in
the
security
piece
is
impressive
in
routing
we
do
it
in
order
to
support
high
availability
and
sub-second
flow
and
failover
security,
it's
really
good
to
see
the
same
categorization.
Thank
you
very
much.
You're.
M
B
C
Again,
we're
taking
the
approach
of
high-level
thought
for
the
room
case
you've
coming
into
this
new
we're
taking
the
approach
of
implementing
like
I,
said
a
piece
of
the
the
large
blueprint
and
working
on
it
step
by
step
through
hackathons.
If
you're
interested
in
playing
with
us,
please,
please
download
the
code
from
that
that
has
been
created
from
the
Korean
team
and
please
I'm
sure,
they'd
love
comments.
C
This
is
to
help
get
early
review
early
fail
on
anything,
but
we're
starting
to
see.
We
think
this
is
close
for
some
of
the
product
groups
that
we've
talked
to
okay
and
also
the
Linux
theme.
The
groupings
are
what
you
would
think
from
the
from
the
original
capability
information
model,
we're
grouping
based
on
network
and
security
control,
the
content,
control,
the
mitigation
control
and
the
capabilities
information
remember
this
is
just
should
look
like
Frank's
original
thing
go
ahead
to
the
next
now,
there's
a
difference
between
the
NSF
facing
yang
data
model
and
the
capability
model.
C
The
NSF
is
used
to
configure
the
rules
of
policy
into
the
NSF,
and
the
capability
data
model
is
used
to
retrieve
it.
We're
taking
the
idea
that
you
want
to
know
what
you're
supposed
to
do
and
then
you
put
it
down.
This
provides
what
we
had
agreed
in
the
Charter,
which
is
to
write
what
we're
doing
and
then
allow
flexibilities
for
different
people
with
NSF's
to
download
things
the
benefit
of
having
the
good
open
source
that
the
team
did.
As
you
can
see
it
based
on
the
open
source
filters.
Please
go
ahead.
C
This
I
think
is
just
showing
the
same
thing
in
a
picture
keep
going,
and
this
shows
that
you're
getting
a
directory
list
interface,
and
this
shows
that
we've
actually
been
doing
the
fact
that
we've
used
this
for
SFC.
Okay.
In
case
you
missed
the
last
hackathon
in
this
hackathon
report.
Let
me
give
you
a
little
background.
There's
two
cases
that
we're
looking
at
that
are
real
important
phones
and
forwarding
with
the
classifier
using
the
SFC
stuff
really
good,
and
this
isn't
on
the
slide.
C
From
a
centralized
pointer,
if
you
distribute
it
what
they
found
in
their
initial
implementation,
is
you
really
ought
to
distribute
these
type
of
security
functions
throughout
the
network
because
it's
more
efficient?
So
with
that
high-level?
Let
me
go
show
you
that
you
know
the
ideas
the
packet
comes
in.
You
have
a
query
based
a
capability,
you
download
the
information
and
then
you're
able
to
forward
it.
C
This
allows
you
if
you
have
a
distributed
model
to
get
very
efficient
with
the
SFC
forwarder,
either
for
specific
functionality
such
as
mail
cleaning
or
filtering
or
deep
packet
inspection
or
other
things
within
the
network.
So
please
we're
giving
you
this
as
a
model,
but
we're
trying
to
keep
now
improving
the
contents
of
this
for
the
IT
resource
again
work
on
a
little
piece
build
it
test.
It
provide
you
models.
What
we
really
would
like.
C
A
C
I
did
take,
I
did
take
a
time
and
tell
the
chair
about
what
I
thought
was
the
high
level
order.
I
will
make
sure
that
I
also
go
back
and
correlate
that
with
that
working
group,
but
I
told
him
what
I
just
told
him
that
the
important
thing
for
their
working
group
is
the
fact
that
it
looks
like
in
our
early
thing.
The
distributed
filtering
for
security
functions
was
more
effective
and
I'm.
Sure
he'll
share
that
with
his
team,
as
that,
as
he
wishes,
that's
Joel
Halpern
by
the
way
for
people
as
the
people.
I
Chairs,
yes,
Kathleen
Moriarty
ad,
so
a
couple
of
general
comments,
a
few
of
the
drafts
some
not
adopted
yet
seem
like
they
are
overlapping
enough
with
other
ones
that
if
they
move
towards
adoption
like
the
applicability
statement
she
folded
into
something
else,
terminology
we
may
want
to
fold
into
one
of
the
protocol.
Drafts
it'll
have
an
easier
time
with
these
formational
type
drafts
with
the
iesg
and
the
other
one
we've
been
hearing
about
development
work,
which
is
really
exciting.
Can
it
also
get
plugged
into
code
stands
so
that
we
could
track
what
drafts
have
info?
I
C
M
This
is
Hank,
and
this
is
about
collapsing
terminology.
You
of
course
know
that
I
have
not
to
speak
for
John
and
me
at
the
same
time.
I
think
John
would
not
really
like
the
dependency
onto
a
protocol
or
a
specific
model,
because
there's
an
umbrella,
they
would
always
include
terms
for
other
drafts
and
they
they
are
not
at
the
top
of
the
heap,
so
it
would
never
align.
So
if
there
is
no
extra
charge.
A
I
I
So
if
it's
I'm
suggesting
a
protocol
document
because
there's
something
more
support
for
those
and
those
do
require
the
terminology
right
moving
forward.
So
if
it's
like
a
core
document,
that's
pivotal
to
the
working
group
that
you're
going
to
be
referencing
anyway
from
other
documents,
then
it's
a
good
one
for
the
terminology
to
be
referencing.
N
C
B
H
H
So
this
time
are
we
updated
the
event
component
and
the
action
component
using
the
more
concrete
data
types
next
slide,
so
this
is
showed
that
we
replace
the
integer
time
unsigned
integer
type
using
the
immigration
time
next
right
also,
we
previously
use
the
case,
but
this
division
we
use
enumeration
for
our
ingress
action
next
slide,
so
also
the
previously
the
year
apply,
some
apply
a
profile
axiom
or
domestic
action
context
based
action.
For
example,
we
try
to
simplify
the
data
structure
next
slide
so
next
step.
H
H
O
O
Yeah,
so
the
major
objective
of
the
registration
interface
is
registering
on
NSF
and
this
capability
into
the
system,
so
our
information
model
and
data
model
focus
on
that
function
and
also
in
some
cases.
The
security
controller
may
want
to
request
the
developers
management
system
to
create
a
new
instance
or
district
on
existing
instance,
which
is
underutilized.
O
So
in
this
to
document,
we
designed
the
information
model
and
data
model
to
support
the
following
functions,
which
is
registering
a
new
NSF
instance,
and
these
capabilities
into
the
security
controller
and,
additionally,
to
request
dynamic,
instantiation
or
the
instantiation
next
one
yeah.
So
this
slide
is
a
summary
of
the
update
and
in
the
in
this
new
version
of
the
information
model
draft,
we
added
the
portion
of
performance
capability
and
we
also
update
the
updated
young
data
model
to
align
with
the
information
model
update
next
one
please.
O
So
this
figure
is
briefly
describes
the
registering
process
through
the
legislation
interface.
So,
after
creating
the
new
instance
by
the
developers,
management
system,
developers,
management
system
registers,
the
creative
you
NSF
instance,
and
his
capability
to
the
security
controller,
so
the
existing
information
model
and
data
model
are
used
to
describe
the
security
capability
upon
NSF
instance,
next,
1/3
and-
and
we
thought
in
some
cases,
are
in
the
following
cases.
O
The
security
controller
may
need
to
request
to
create
a
new
instance
or
destruct
an
existing
instance.
So
this
is
the
possible
cases,
so
the
first
one
is
in
the
I-20.
A
popper.
A
mock
on
NSF
can
trigger
on
advance
the
security
action
which
requires
another
type
of
NSF
rapport
for
further
security
inspection.
In
this
case.
For
some
reason,
the
the
triggered
NSF
instance
may
be
unavailable
in
the
system,
in
this
case
the
security
controller
to
serve
disadvantage
security
action.
O
O
This
is
the
summary
of
the
version
update
and
we
have
added
an
additional
portion
of
the
performance
capability
to
describe
several
type
to
attributes
of
the
type
of
resources.
Next
one
please,
yes,
so
in
the
future,
we
will
extend
our
hackathon
implementation
to
to
support
this
registry
registration
process
via
the
registration
interface,
and
we
will
also
do
some
investigation
on
the
feasibility
of
dynamic
instantiation
or
these,
the
instantiation
yeah.
Thank
you.
E
E
E
Capabilities
and
or
push
or
pull
capabilities
from
or
to
the
NSF
seconds,
looking
at
the
atom
of
the
elements
that
appear
in
the
in
on
the
screen,
I
see
a
I
mean
is
somehow
very
also
how
it's
very
much
connected
with
the
hand
we're
talking
about
stages.
You
are
talking
about
memory
or
the
processing,
the
idea,
processor,
loads
and
things
like
that.
That
has
to
some
extent
there.
Well,
there
are
events
as
well
or
some
kind
of
telemetry,
so
so
I.
What
I
would
encourage
is
that
to
try
to
convert
with
what
hank
was
presenting.
E
So
we
have
a
single
set
of
telemetry
data
in
general,
whatever
they
are
events
or
measurements
or
whatever.
So
it
would
encourage
that
you
converge
with
them
when
it
comes
to
the
for
the
data
model
and
the
protocol,
because
this
set
of
the
of
protocol
choices
at
hand
was
showing
business,
it
would
be
very
good
to
experiment
with
them
in
the
next
seven
hackathon.
That's
thank.
A
P
B
P
It
looks
like
well,
first
of
all
of
with
reminder
of
the
main
of
the
dates
of
this
draft.
The
first
one
is
to
describe
the
architecture
for
the
basic
management
of
security
Association
by
means
of
the
HD
and
party,
and
also
to
define,
in
this
case,
the
network
ID
function,
facial
interface,
security,
controller
to
you,
managed
and
monitored
security
association
between
our
modern
networks,
objective
function.
P
P
Okay
in
this
drug,
we
propose
to
use
two
cases
in
case
one.
The
network
security
function
implements
both
the
IPSec
can
start
and
also
the
management
protocol.
This
case
I
key
in
this
case.
The
security
controller,
is
in
charge
of
provide
the
negotiated
function
with
the
required
configuration
information
for
IP
or
the
a
speeded-up
security
policy
database
and
the
policy
of
penetration
database,
but
not
the
security
Association
database
with
the
security
Association
are
created
in
run
time
by
means
of
a
akin
oscillation
the
table
length
in
case
two.
P
We
suppose
that
the
network
security
function
only
implements
the
NSA
data
bases,
the
security
policy
database
and
the
security
Association
database,
but
not
run
the
in
this
case,
the
key
management
protocol.
So
in
this
case,
the
city
of
the
controller
has
to
provide
the
configuration
for
both
policies
and
the
security
Association
intercept.
P
P
Okay,
changes
from
the
last
produced
version
we
are
talking
about
this
drug
is
person
0
3.
We
have
focused
this
niversity
in
the
Bodleian
hasta
los
scenarios.
The
idea
is
to
define
the
model
and
to
deploy
a
test
bed
to
test
the
Dawson
scenarios
and
the
work
for
the
hospital.
In
this
case,
the
Rottweiler
scenario
is
worn
to
be
done
for
next
session.
P
We
also
have
improved
the
discussion
among
this
one
versus
case
Chile,
following
the
comments
received
to
land
in
the
middle
list
and
concern
in
the
previous
presentation
rolling
the
main
contribution
is
the
John
configuration
data
model.
This
case
we
will
provide
any
providing
the
drug
just
one
junk
file
providing
the
model.
Sorry.
P
It
is
up
to
the
security
controller,
knowing
the
kind
of
scenario
to
be
applying
in
order
to
decide
the
part
of
this
model
to
be
used.
For
example,
if
we
are
talking
about
the
security
controller
is
deploy
in
a
case
one
scenario,
then
it
has
two
main
years
of
the
AIT
SPD
and
in
AD
the
models.
But
ether
controller
knows
that
has
to
deploy
case
scenario.
Then
he
has
to
make
use
of
the
SPG
and
basically
on.
In
order
to
define
this
John
certification.
P
We
have
negative
analysis
or
of
the
RFC
4301
to
describe
the
requirements
of
the
security
policy
with
the
Association
and
pushing
the
ways
we
are
also
analyzer
physicians,
t26,
to
to
analyze
the
configuration
required
for
the
key
management
protocol.
The
also
we
have
analyzed
open
source
implementation
like
strong
one
and
this
one.
But
again
here
we
call
for
of
the
experts
to
try
to
help
rule
the
land
and
turning
out
of
this
mode.
Next.
P
P
Transport
alternate
mode
replication,
Turkish,
illumination
of
educational
options
for
case
one
in
case
of
the
network
security
function.
We
have
menus
of
train
of
strong
one
for
the
implementation
and
the
API
provided
by
strong.
In
this
case,
I
said
in
order
to
allow
runtime
configuration
of
IP
and
for
case
two
in
the
case
of
Manitoba
security
functional.
We
have
made
you
soft
beefy,
Kim
Basinger
for
the
SAT
configuration
in
the
network
security
function,
the
came
station
for
the
SPT
configuration
and
also
the
XRF
M
for
a
CDN
speaking
in
sixth
system.
R
Who
so
say
quickly?
This
is
good
work
IPSec
or
the
flow
protection
is
good.
However,
in
practical
cases,
many
cases
it
doesn't
work
when
your
NSF
is
with
inside
the
corporate
boundary,
and
there
are
policy
rules
which
which
won't
allow
it.
Das
has
been
dealing
with
this
the
same
issue
in
dots.
It
needs
to
be
covered
and
be
taken
care
of,
but
it's
not
the
sole
solution,
because
it
doesn't
always
be.
It's
not
always
employable
that.
That's
all.
We
need
to
have
some
caveat
and
discussion
with
that
that
point
in
the
draft.
Yes,.
P
Yes,
well,
it's
true
that
we
have
not
analyzed
all
the
all
the
cases,
but
that's
why
we
reach
for
this
time
scenario
that
could
write
some
problem
model
just
to
say
that
we
have
follow
the
RFP
see.
We
have
tried
to
be
very
accurate,
the
in
the
definition.
So
in
theory
we
are
basing
on
their
feces.
So
of
course
those
the
scenario
appears
and
we
want
to
try
to
analyze
these
cases.
P
S
You
have
mirror
might
set
it
on
the
mitt
on
the
IPSec
mailing
list,
not
and
I
train
a
semi
long
list.
Well,
the
problems
I
see
with
this
model
is
that
in
a
lot
of
time
in
VPN
scenarios,
your
NSF
is
not
in
the
corporate
bounds,
but
in
some
kiosk
at
the
mall
somewhere
and
the
network,
there
is
not
known
to
the
administrators,
so
the
information
about
what
is
behind
what
what
the
address
the
NSF
has
and
what
addresses
are
protected
behind.
S
That
gateway
are
known
only
to
the
NSF
at
the
start,
so
the
information
about
all
that
has
to
flow
from
the
NSF
to
the
controller.
Now,
if
we
build
the
model
of
where
the
information
only
flows
from
the
controller
to
thee
to
the
NSF,
then
we
have
to
have
some
side
channels,
some
Bachelor
of
that
where
people
read
off
the
routing
table
from
the
configuration
of
the
NSF
and
then
input
it
to
the
administrator,
and
that's
not
what
we
want
want
something
automated.
S
P
P
That's
a
the
current
high-level
policy
to
the
controller
in
order
to
allow
the
controller
to
specify
the
correct
network
security
function.
But
we
have
to
assume
that
the
security
controller
has
the
world
view
of
the
network
and
I
think
that
is
something
that
where
we
are
now
see
that
is
required
thanks
to
the
deployment
of
the
HD
1
solution,
and
of
course
there
are
a
scenario,
were
it
could
be
more
difficult
to
apply,
but
there
are
other
scenarios
were
work
problem.
S
P
K
Wondrousness
love,
Alice,
+,
I'm,
very
much
concerned
with
I
Clay's
piece
and
option
number
2,
because
first,
a
security
controller
became
a
really
very
attractive
target
for
attack.
Once
you
execute
you
control,
you
get
all
the
keys,
and
the
second
concern
is
that:
well
it's
a
could.
You
clarify
please
how
you
securities,
that
are
distributed
from
security
control
to
endpoint,
because
I
hope
you,
you
are
not
send
it
in
clear.
K
Yes,
you,
if
you,
if
you
have
some
security
channel
between
security,
controller
and
endpoint,
that
you've
already
implemented
some
kind
of
TLS
icon,
something
some
security
protocol.
So
I,
don't
think
you
you
get
some
again
when
you
remove
likely
to
because
you
you
need
to
have
some
security,
some
standard
security
protocol
to
to
secure
transport
keys.
So
I
don't
think
it's
a
good
solution
to
remove
by
well.
P
Unless
the
end
are
they
all
the
network,
devices
that
say
have
to
have
to
have
a
trust
relationship
with
the
shiggity
controller.
That's
one
of
the
implication
and
that's
why
the
security
controller
hospital
to
have
the
wall
of
view
of
the
network
you
know
file.
Those
network
security
function
has
to
be
proficient
with
this
authentication
encryption
act
area
even
but
in
this
file
name.
A
S
S
Provide
the
ability
to
further
for
a
user
or
application
to
express
the
security
policy
in
a
fashion
for
the
security
controller,
which
turns
decide
where
to
apply
those
security
policies
and
how
to
apply
them.
Then
the
the
enforcement
point
on
which
bouchard
NSF's,
on
which
these
security
policies
and
the
being
confident
by
the
security
controller,
applied
the
policies
to
the
traffic.
As
with
the
end
points,
10
points
could
be
applications
or
could
be
users
or
could
be
applications
and
user,
and
any
combination
of
these.
S
They
based
on
requirements
coming
from
the
owner.
What's--What
stands
for
the
open
network
user
group
feedback
from
them,
skillett
policies
that
were
added
related
to
segmentation,
what
people
know
sometimes
as
micro
segmentation
policies
that
track
policies,
as
well
as
compliance
and
audit
policies,
and
there
are
further
kind
of
finally
great
defiant
of
different
policy
blocks
that
pertain
to
source
policy.
Endpoint
group,
the
direction
in
which
the
policy
needs
to
apply
throughout
the
group.
S
S
So
what
are
the
design
principles
behind
behind
the
the
client-facing
interface?
Really,
the
the
goal
in
here
is
to
enable
again
a
user
or
application
to
express
the
policies
in
an
abstract
way
and
very
declarative,
a
slash,
descriptive
way,
rather
than
being
imperative
and
prescriptive,
meaning
asserting,
really
the
what,
rather
than
the
how
things
need
to
unravel
and
when
they
need
to
apply
and
that
fashion.
That
means
that
the
user
express
them
without
necessarily
knowledge
about
the
network
topology.
S
How
then
I
saw
his
eyes
within
elaborate,
how
they're,
connected
and
so
forth,
because
that's
really
for
the
for
the
security
controller
to
decide
about
where
to
apply
the
things.
Having
said
that,
that
could
be
cases
where
the
user
or
the
application
needs
to
express,
also
a
policy
of
certain
affinity
by
which
the
policy
need
to
apply
to
a
practical
device,
particular
in
particular
a
particular
anis
a
function
and
that
exists
as
well.
S
Can
we
go
to
the
next
slide
please?
So
there
are
many
requirements.
There
is
no
point
in
going
through
one
by
one
in
here
really
what
has
been
covered,
but
suffice
to
say
really
this.
These
policies
are
intended
to
be
applied
in
multi
domain
multi-tenant
ways,
so
there
is
there.
Are
these
objects
that
there
are
requirements
for
object,
that
that
identify
tenants
at
five
domains,
ways
in
which
the
policies
could
be
triggered
in
/?
That
could
be
event
digger
that
could
be
finally
triggered.
S
S
That
is,
the
terminology
used
in
this
left
would
die
to
an
ACEF
terminology
laughs
as
well
as
just
some
comments
on
the
on
the
mailing
list.
I
think
there
is
related
to
here
really
can't
separate
what
side
of
a
requirement
this
is.
What
are
the
must
require
meant
as
well
as
to
get
clarification
on.
What
is
the
difference
between
what
we
call
in
the
draft
as
construct
based
policy
definition
versus
intent
based
the
Tahlia
truth?
C
Helpful,
the
indentity
that
you're
the
groups
that
you're
using
it
would
also
be
good
to
clarify
whether
using
how
you're
looking
at
identity
concepts
is
identification,
etc.
S
C
S
Let
me
I
think
I'll
give
an
example
in
here,
just
a
brief
one.
So
the
idea
I
mean
before
there
people
used
to
define
policies
that
says
block
traffic
from
this
IP
address
to
that
IP
address,
block
user
I
acts
from
accessing
application,
be.
The
idea
in
here
is
to
really
group
things,
and
rather
than
define
them
based
on
information
as
such,
unless
you
need
to,
but
you
could
say,
block
traffic
between
application,
a
and
application
B.
How
many
instances
of
that
application
exists?
Really?
S
Is
a
Robin
that's
up
for
the
really
controller
to
arrival
death
and
expand
that
policy
to
the
application
instances.
That's
that's
one
example.
So
an
application
is
that
the
case
could
be
identified
by
a
name
or
attack,
for
example,
and
how
that
could
the
same
thing
they
use
it.
You
could
put
a
user
with
user
group
and
obtain
that
tag
or
event,
event
of
that
user
group
from
an
Active,
Directory
or
LDAP,
or
something
along
that
line.
S
K
S
The
idea
I
mean
as
the
information
on
that
model
and
John
what
what
it
is
is
really
I
think
so
referred
to
its
early
and
her
presentation
has
a
blueprint
to
define
a
data
model
or
basically
finding
what
needs
to
be
contained
in
the
model,
in
other
words,
as
really
as
there
is
one
our
C
that
says
how
it
should
be.
Fine,
it's
about
defining,
really
information
objects.
S
As
I
said
earlier,
there
have
been
few
extensions
that
we
added
to
the
requirements
related
to
different
policy
categories,
though
some
of
them
are
already
in
the
information,
while
others
will
need
to
be
extended,
in
particular
things
pertinent
to
the
to
their
competitive
to
the
to
the
audit,
for
example,
category,
as
well
as
to
the
segmentation
category,
it
has
to
be
explicitly
mentioned,
but
Jean
drill.
The
information
model
is
consistent
with
the
requirement
left
that.
S
The
functional
requirements
really
is
applied
to
multi-tenancy
related
objects
so
that
we
have
a
tenant
domain
object,
object,
the
main
object
and
so
forth,
and
there
are
just
kind
of
defined
in
the
requirement
what
we
call
the
user
group
and
point
the
application,
endpoint
and
so
forth.
So
there
are
object
models
for
these.
S
S
So
really
the
way
you
define
the
policy
instances.
Is
he
a
composite
of
the
different
policy
marks
or
policy
construct
that
we
talked
about
earlier
and
in
this
presentation
next
slide
please.
So
the
draft
is
has
been
updated
if
there
is
any
input
from
the
working
group.
That
would
like
the
work
group
would
like
to
hear
about
that
on
the
manifest
and
will
discuss
and
update
according.
E
To
command
so
this
first
of
all
again
I
see
that
there
is
some
overlap
with
other
end
drafts,
etc.
So
we
will
need
to
have
an.
In
fact,
when
it
comes
to
these,
the
customer
facing
interface
I
was
wondering
whether
can
move
it
a
little
bit.
I
was
wondering
whether
it
would
make
sense
I
mean
looking
at
the
text.
The
text
to
me
looks
more.
A
set
of
one
hand
is
extended
requirements
that
are
better
formalized
the
one
hand,
and,
on
the
other
hand
they
are
more
about
most
about
metadata.
E
I
mean
we're
talking
about
the
issue,
the
source,
dates,
etc.
What
looks
to
me
as
an
envelope
for
a
general
statement
and
probably
I
mean,
and
it's
something
is
something
that
we
will
have
to
go
through.
This
Linda
was
proposing
this
design
team,
in
which
we
are
going
to
try
to
steer
a
little
bit
all
the
drafts
we
have
at
hand.
I'm
probably
I
mean
my
recommendation
looking
at
it.
E
Looking
at
them
looks
to
me
that
this
is
justified
in
the
case
of
the
capability
model,
because
a
capability
model
is
something
complicated,
so
to
say
to
express-
and
it's
a
new
thing
that
we
are
bringing
here
in
terms
of
generalizing
the
whole
thing,
but
probably
in
other
cases,
just
having
data
mobile
I
mean
a
data
model
document
with
a
justification
talking
about
why
that
data
model
is
built,
probably
should
be
nothing
so
we
reduce
a
number
of
entities
and
we
apply
these
outcomes.
Racer.
S
K
S
Been
in
other
working
groups
where
people
had
really
forgot
didn't
see,
they
need
to
have
an
informational
model
documents.
They've
done
in
some
working
groups
were
done
done
and
then
we
dismiss
some
data
after
the
date
of
all
that
have
been
find,
so
they
were
just
used
as
guidance.
If
you
will
to
define
the
data
model,
I
think.
A
S
Coming
back
to
your
comment
that
that's
really
where
the
information
model
is,
it's
really
very
kind
of
separated
from
the
requirement
the
requirements
talk
about.
What's
driving
things,
what
needs
to
be
done,
what
the
information
model
is
supposed
to
be.
They
talk
about
what
information
is
too
deeply
complain
and
an
information
model?
How
things
relate
to
each
other,
and
that's
really?
The
information
model
is
so
it's
not
really
requirements.
We,
we
have
no
requirements.
E
M
Really
quick,
there's
a
sink
height,
the
term
that
I
heard
already
and
that
I
also
wrote
on
a
lot
chart
on
the
first
day
of
the
athan
is
telemetry.
So
my
assumption
is
that
the
solution
on
data
model
level
will
be
a
yang
based
posh
subscription
subscribe.
But
if
occation
based
data
model,
a
datum,
also
say
they've
been
stream
telemetry
stream.
So
maybe
we
can
converge
all
the
information
items
in
a
corresponding
graph.
That
would
make
use
of
this
one
method.
S
T
Hello
Jessica
from
Huawei
I'm,
going
to
present
you
the
latest
changes.
I
was
a
policy
object
draft
first,
as
a
background
recap,
and
the
framework
draft
I
at
USF
has
selected
the
save
a
model
for
for
policy
rules,
and
the
capability
draft
has
defined
many
attributes
for
different
for
different
conditions.
Subclasses
the
attribute
base,
the
policy
rules
and
phases
to
possible
problems,
repeatable
configurations
for
the
creation
and
the
maintenance
of
the
policy
rules
is
time-consuming
and
tedious.
T
T
The
way
we
provide
the
reducer
bureau
chief
for
the
creation
of
policy.
Object
can
be
referenced
for
multiple
times
and
also
brings
simplicity
for
the
maintenance
of
the
policy
rules.
So
next
slide
please
there
are.
There
are
two
main
changes
from
last
meeting.
First,
we
provide
a
minimal
set
of
policy
objects
and
attributes.
Instead
of
defining
every
possible
object.
It's
even
needed
as
a
new
one
will
be
included
in
this
draft.
T
We
sing
again
go
to
the
examples
the
example
a
slide
and
yes,
so
we
were
added
an
example
of
the
application
scenario
for
the
policy
objectives.
This
is
a
typical
enterprise
network
technology
which
the
fair
wall,
which
is
an
example
of
the
NSF
located
at
the
boundary
of
the
network,
and
we
have
Chu
department,
shook
hands
out
departments,
the
marketing
department
and
the
R&D
department
next
slide.
Please,
then
we
can
make
different
policy
control
a
code
by
using
by
defining
and
using
many
policy
object,
and
in
this
slide
the
event.
T
The
event
patch
is
missing.
We
were
added
in
the
in
the
next
version,
and
so
so
next
slide
please.
So
we
have
come
up
with
some
questions
and
us
sausage
for
comments
such
as
jus
event
and
action.
Also
need
objects
and
we
can
build
policy
object
for
post
client
facing
interface
and
an
NSL
facing
interface
so
and
other
questions
I
think
we
can
talk
about
each
in
the
mini
list.
Thank
you.
That's
all.
H
H
Last
one,
yet
so
the
next
the
hackathon
which
we
tried
to
okay,
this
time
we
have
made
our
improve
the
East
a
model,
so
hopefully
the
momentum
of
the
conspiracy
probably
improved
next
slide.
So
next,
please,
so
we
tried
to
synchronize
with
the
super
information
model
next
time
and
also
the
flag
to
make
is
consumer
taste.
Thank
you.
Thank.
M
Yes,
hi
so
I'm
Hank
again
last
meeting,
we
made
a
very
small,
very
tiny
introduction
of
the
structure
moment,
which
is
in
our
way
later
document
to
the
HR
NSF.
What
group
it
is
about
remote
attestation,
so
we
are
doing
the
same
thing
as
Diego.
Well,
really,
of
course,
we
are
doing
remote
system
entity,
characters,
authentication.
This
is
using
RFC
4949
terminology,
so
this
is
basically
for
remote
attestation
is
this
is
composed,
recording
it.
M
Wretch
three
activities
exists
so
at
the
station,
actually
is
only
the
creation
of
the
evidence
by
having
one
or
more
claims
about
the
characteristics
of
the
business
with
system
entity,
and
so
this
claims
can
be
used
as
evidence.
Then
we
convey
second
activity.
This
evidence
to
the
verifier.
The
an
interconnect.
Interconnect
is
a
strange
term
here.
It
is
typically
used
by
people
of
USB
or
bluetooth.
When
you
have
hard
times
finding
the
the
the
message
pass
were
using,
it
can
be,
GI
opens
and
it
can
be
up
to
the
internet.
M
So
everything
that
is
able
to
transport
data
can
be
interconnect.
It
could
be
in
a
small,
form-factor
SFP
interface
talking
for
the
module
or
it
could
be
a
cluster
of
super
high
performance
computers
talking
to
another
cluster.
So
the
interconnect
is
basically
just
the
thing
you
convey.
The
evidence
with
verification
then
happens
on
the
other
side.
It's
the
appraiser
of
the
evidence
against
the
electric
cloud
of
guidance
defined
by
second
terminology
next
slide.
M
So
typically,
this
is
a
challenge
response
act,
interaction
model,
because
you
want
to
know
that
the
evidence
you
get
is
fresh.
It's
just
the
thing
you
want
it
and
the
HST
just
created
it.
So
at
creating
a
nonce
and
the
action
model
is
tightly
coupled
timewise
Radhika
bird's
this
by
using
trust
time
some
tokens
and
set
of
a
month
and
therefore
allows
for
a
unidirectional
interaction
that
does
not
need
a
return.
Channel
next
slide.
M
So
there
are
a
lot
of
objectives.
I
think
I
present
them
last
time.
Basically,
unidirectional
is
great
because
of
rest
call
me,
and
it
is
basically
used
for
constraint,
environments
and
it
will
be
able
to
be
used
with
loss
of
connectivity,
so
you
can
basically
create
audit
logs
of
the
attestation
basis.
The
evidence
and
then
transfer
later
next
slide
we're
introducing
a
third
party
for
this,
because
these
term
tokens
have
to
come
from
somewhere.
M
They
are
provided
by
the
trusted
by
a
timestamp
Authority,
there's
an
RC
for
that
I
forgot
to
put
the
number
in
by
Google,
so
the
Tuda
protocol
has
a
nested
sync
protocol,
which
is
basically
encapsulating
the
time
scepter
into
two
other
times
Thames,
and
now
it
comes
that
are
provided
by
a
hardware
root
of
trust,
they're,
calling
that
rod.
So
in
essence,
true,
there
is
a
read
based
on
Hardware
rod.
The
TSA
is
required
to
create
this
evidence,
basically
every
scenario.
M
So
every
thing
that
would
create
a
secure
audit,
fire
audit
block
or
something
that
hasn't
in
the
past
and
you
want
to
prove
it
really
happened-
requires
a
TSA.
This
is
not
unique
to
to
DES
we're
just
making
use
of
it
in
the
terms
in
scope
of
agitation
next
slide.
So
we
had
a
lot
of
feedback
for
this
already
coming
in
from
a
lot
of
participants
of
it--when
SF
I
got
some
of
them
more
past
slide
previous
slide,
please
because
it's
the
second
question,
so
why?
M
Using
tudi
if
it
uses
a
sync
protocol
and
makes
things
more
complex,
yeah.
Well,
there
was
a
very
first
question,
but
I
think
some
are
answered.
There
are
please
look
at
RFC
4949
the
bill.
La
patrulla
and
the
Biba
model.
Both
are
specific
models
that
focus
on
confidentiality
and
integrity.
If
you
change
trust
level
zones
talking
about
a
nuclear
plant
and
critical
infrastructure,
that's
kind
of
important
to
adjust
the
things
inside
without
being
able
to
compromise
them
from
the
outside.
M
Just
read
that
up
it's
it's
kind
of
yes,
then,
second
slide,
please
a
question
I'm
running
through
these
items
because
of
time
so
which
environments
would
be
a
benefit
from
the
more
complex
solutions.
So
we
have
a
lot
of
IOT
stuff
going
on
in
the
ITF
space,
and
most
of
it
is
coop.
Coop
is
restful.
There
is
no
state,
so
you
kind
of
faciliate
a
challenge
response
protocol
without
state
on
the
server
side.
Also,
there
is
the
idea
of
aggregating
a
lot
of
requests
for
two
stations
with
one
aggregated
nonce.
M
This
is
an
idea
that
was
published
by
our
lead
author
actually
10
years
ago,
and
it
does
not
address
restful,
coop
IOT
problem.
We
think
next
time.
Sorry
so
doesn't
it.
He
is
a
introduced
new
attack
vectors
here.
It
will,
of
course,
as
well.
Every
other
component
that
is
vampire
require
to
create
security.
It
would
create
another
service.
M
In
fact,
that
is,
of
course,
an
attack
vector,
for
example,
there's
probably
a
night
when
s
of
component
that
takes
charge
of
the
security
of
the
service
of
the
software
itself
or,
for
example,
prevents
resigning
of
anything
you
create
into
a
night
renders
F
domain,
so
there
will
be
security
functions.
Of
course,
the
trade
secured
you
into
their
I
to
an
SF
domain
and
the
TSA
will
be
just
another
function
like
this
next
slide,
so
yeah
we
can
broadcast.
Of
course
you
can
beacon
it
out
unsolicited
and
they
are
not
encrypted.
M
So
if
we
just
become
this
out,
yeah
those
would
compromise
privacy,
but
that
is
not
a
problem
unique
to
tutor.
That
is
a
problem
that
you
chose
to
do
it,
but
the
broadcast-
and
there
please
just
choose
a
secure
transport
for
that.
So
next
slide
next
session
was:
does
the
state
let
nature
support
incremental
data
stations
and
most
certainly
it
does.
It
is
basically
the
one
point
of
creating
a
partial
order,
locks
that
you
can
create
as
measurements
company
with
the
evidence.
So
the
map
shows
you
a
very
good
way.
M
M
So
how
and
up
again
and
up
again
please
so
how
would
hardware
root
of
trust
and
virtualized
environment
work
because
virtual
machine
does
not
have
hardware?
This
is
called
a
virtual
hub
of
trust.
A
lot
of
stos
are
working
hard
on
this
for,
like
I,
know
six
years
and
I
think
much
at
least
two
SDOs
that
are
not
the
ITF
have
solutions
for
those.
Unfortunately,
they
do
not
do
public
review
on
a
regular
basis,
but
I
basically
know
for
certain
that
they
will
do
public
review,
that
one
of
them
will
do
public
review
very
soon.
I
M
This
was
about
the
terminology
for
agitation,
because
everybody
thinks
at
a
session
is
something
else
or
doesn't
know
what
attestation
is
at
all.
So
it
is
about
talking
about
the
right
things
there.
There
is
a
draft,
it's
an
individual
draft
with
General,
Electric's
and
arm
it
is
posted.
I
can
post
this
graph
to
the
I
to
NSF
list.
If
you
want
I
was.