►
From YouTube: IETF100-I2RS-20171113-1550
Description
I2RS meeting session at IETF100
2017/11/13 1550
https://datatracker.ietf.org/meeting/100/proceedings/
A
A
B
A
C
D
D
We
may
put
you
at
the
end:
if
that's
okay,
so
I
can
switch.
Okay,
then
we'll
do
the
CU
separation
info
model
and
the
ephemeral
state
and
come
back
to
the
data
fabric
and
again
I
apologize
for
the
places
that
we've
missed.
Okay,
we
are
down
to
iOS
is
having
is
in
its
last
phases
of
closing
out
work
that
was
in
its
original
Charter
and
we
hope
to
be
completing
that
work
in
going
into
hiatus
or
closing
the
working
group
in
March.
So
adoption
that
we
do
at
this
point
must
have
some
very.
D
Strict
requirements
because
we've
got
to
go
quickly
from
any
adoption
to
working
group
last
call,
so
that
will
be
true
of
the
two
drafts
we're
talking
about
today.
One
is
the
ephemeral
datastore
and
one
is
the
informational
draft
and
both
of
those
have
a
need.
Now.
This
is
a
little
picture
about
what
your
chairs
life
has
been
like.
Both
Russ
and
I've
been
a
little
swamped
by
our
life
outside.
D
So
thank
you
for
helping
us
authors,
we're
sorry
we're
not
catching
up
with
you,
the
good
news,
and
hence
here
with
the
adoption
of
that
confident
lot
of
the
revised
datastore
models
and
everything
iOS
can
complete
its
task.
We
have
to
complete
the
ephemeral,
datastore
model
and
submit
all
models
to
the
is
G
and
then
go
into
hiatus,
and
we
hope
to
do
this
by
March.
Our
models
are
the
two
topology
drafts
which
were
handling
editorial.
It's
the
rib
data
model,
which
their
authors
tell
me
they're
doing
the
same
thing.
D
D
The
other
pieces
in
the
Charter
that
I'd
like
to
remind
you,
is
that
we
had
l2
topology,
fib
rib
and
the
data
center
fabric,
we'll
get
a
discussion
of
the
data
center
fabric
since
I
know
they've
implemented
that
we'll
probably
move
that
quickly
to
working
group
last
call,
which
is
why
thanks
young,
we'll
go
through
that
second,
but
we
have
to
go
through
the
ephemeral
data
store.
So
there
are
two
models
that
we're
going
to
discuss
here.
D
One
is
an
ephemeral
data
store
and
wanted
to
see
you
separation
with
an
info
model
that
a
couple
service
providers
one
is
implemented.
Two
are
working
on
and
I'm
just
going
to
say
this
one
more
tone.
Anything
we
adopt
now
must
be
ready
to
roll
and
discuss
shortly
so
that
we
can
start
adoptions
and
working
group
last
calls
within
a
month.
Okay,
so
that's
that
the
chair,
slides,
let's
see
Donald.
If
I
can
do
this.
D
D
E
E
Next
slide
is
about
the
concept
of
the
PNG
device,
who
is
controlling
and
use
a
plane,
suppression
and,
as
introduced
in
in
this
morning,
in
a
rowdy
work
group
on
the
BMG
device
is,
is
presented
with
a
new
architecture
with
the
way,
seize
control,
plan
and
use
of
land
suppression.
Here,
for
short
and
in
the
traditional
byungji
device.
E
The
objective
of
this
draft
is
to
present
information
model,
to
present
the
interaction
interface
between
the
control
and
and
user
plans,
because
now,
in
the
new
architecture
of
the
bng
device,
there
are
two
paths.
One
is
the
control
plan
and
another
is
the
user
plan.
There
are
many
action
in
the
face
between
them,
and
so
we
want.
So.
The
aim
of
this
draft
is
to
present
a
standard
information
model
under
the
standard
information
model.
E
D
E
Good
since
last
idea
of
meeting,
we
have
several
updates.
The
first
is
that
we
specify
the
information
model
in
nineteen
that
has
no
form.
That
means
our
information
model
obeys
the
that
has
no
grandma.
It
make
us
easier
to
understood,
understand
the
information
model
and
the
second
one
is,
we
add,
a
new
section
to
introduce
the
usage
of
the
information
model,
and
then
we
delete
the
appendix
of
the
young
model.
According
to
some
comments,
since
we
think
our
information
model
is
protocol
independent.
E
E
Next
slide
is
the
overview
of
the
information
model
in
our
information
model
of
the
control
and
use
of
land
separation,
energy
device.
It
is
divided
into
two
parts:
Oh
one
is
the
contributor
information
model
and
another
is
a
user
plain
information
model.
The
contributing
information
model
is
used
by
the
control
and
device,
while
the
user
plain
information
model
is
used
by
the
user
plan.
Why
are
in
the
control
information
model?
It
divides
into
three
parts.
E
Information
on
such
as
the
users
ID,
the
users
IP
address,
and
the
user's
on
quality
of
service
rules,
while
the
interface
related
information
model
contains
a
lot
of
attributes
related
with
the
service
interface,
while
the
device
related
information
model
contains
a
lot
of
attributes
related
with
the
device,
such
as
the
address
tech
segment
distributing
to
the
user
plan
by
the
control
plan.
While
in
the
user
plan
information
model,
there
are
two
parts,
one
is
the
party
resources
information
model
and
another
is
the
traffic
statistics
in
our
party
resource
information
model.
E
E
Next
next
slide
is
about
our
future
steps.
The
first
is,
we
want
to
ever
want
to
to
give
us
more
comments,
and
after
now
we
have
update
these
drafts
for
many
times,
so
we
think
the
information
model,
maybe
metro,
and
maybe
this
draft
is
ready
for
the
workgroup
adoption.
So
let
us
leave
the
question
to
our
chairs
on
what
do
you
recommend
about
this
draft
and
the
last
slide?
D
F
E
This
is
slide
way
present
last
IGF,
it's
the
usage
example
of
this
information
model.
Here
we
can
see
a
user
Pub
access
into
the
internet
and
the
u.s.
C
in
C
in
blue
character.
The
CP
on
the
controller
sends
the
table
to
the
U
key,
and
the
tables
includes
the
QC
rules.
Here
are
ways
that
the
sensitive
prevent
bandwidth,
such
as
the
six
or
eight,
that
is
the
Cure's.
E
E
H
G
Is
the
question
is
basically
for
those
who
are
interested
in
the
work?
Would
you
also
come
to
our
to
gwg
to
review
the
work
and
comment
there
or
do
you
feel
like
it
needs
to
the
smaller
and
rapid
speed
that
it
would
require?
If
it
were
an
I
to
RS,
I
mean
that's,
that's
really
the
question
question
and.
G
At
this
point,
I'm
very
reluctant
to
take
things
that
aren't
explicitly
it
required
frightful
or
us
into
the
working
group.
Even
though
this
looks
like
really
interesting
work
and
I
know
you
have
to
play
man
experience
this,
isn't
it
shouldn't
happen
and
it's
just
the
management
of
which
working
group?
It
should
happen
in
so
I.
Think
that
talk
with
sue
but
probably
should
have
a
discussion
with
you
and
Jeff
can
Sarah
and
Chris
powers
as
well
to
see
if
we
can
migrate
to
work
over
there.
G
Right
and
I
want
to
be
clear,
I'm
not
concerned,
so
it
is
clearly
you
have
done
work.
There's
deployment,
pieces
and
I
appreciate
they've,
gotten
private
feedback
on
it.
I
would
encourage
people
to
do
more
public
feedback
on
the
mailing
list
as
well,
because
that
helps
provide
a
record
of
interest
and
the
discussion
and
improvement.
E
D
We
will
and
yeah
and
I,
so
we
can
provide
a
good
bridge
on
that,
because
I
think
there
it
has
shown
significant
interest
here
and
people
who
are
willing
to
comment.
So,
let's
see
if
we
can
try
to
move
it
in
that
direction,
with
some
guidance
from
me
to
Jeff,
okay.
Thank
you
all
thank
you
and
thank
you
for
this
interesting
work.
I
mean
this
is
really
fascinating.
Work
I
wish
we
could
go
forward
with
it
here.
Okay,.
D
D
D
D
Easier
because
90%
of
5%
of
the
work
has
already
been
done
and
ups,
then
that
mod
revised
data
store
has
been
done
in
ops,
the
net
man,
the
revised
data
store,
Netcom
4s
comp
additions,
and
we
just
have
to
have
that
last
5%
to
define
a
standard.
There
are
only
three
things
to
discuss
regarding
this
model:
one,
the
neat
whether
we
need
an
ephemeral
flag.
We
did
this
so
that
you
could
have
a
flag
in
the
data
model.
That's
in
all
the
requirements.
Could
this
be
done
in
a
combination
of
dynamic,
plus
configuration?
D
I
D
D
D
Certain
models
may
be
sent
over
rest
come
over
HTTP
and
TCP.
There
is
a
datastore
flag
that
says
this
and
that's
again
part
of
our
requirements.
I
recommend
just
minute.
I
recommend
that
you,
it's
always
been
recommended
that
there's
a
user
flag
that
turns
off
or
turns
off
the
capability
to
do
this,
the
flag
in
the
data
store
that
does.
D
This
is
just
a
value
that
says
this
date
is
not
sensitive
again
for
a
portion
of
the
model
or
for
the
bottle
most
likely
a
portion
hi
this
web
server
is
up,
and
the
last
thing
that's
in
the
current
model.
It's
very
short.
It
looks
like
the
revised
data
for
a
store.
Example.
Is
the
list
of
models
just
simply
in
a
comment:
hi.
These
are
the
ones,
but
it
is
anticipated.
You
would
then
put
comments.
What's
left
out,
any
flags
on
validation
or
validation
by
implementation.
D
D
D
D
Is
part
of
the
weed
combination
into
the
higher
2's
into
the
into
the
data
model,
and
that
is
actually
something
that
needs
to
be
part
of
the
implementation.
So
that's
part
of
the
architecture
and
the
data
store.
Now,
if
you
need
that
in
the
data
store
model
to
specify
it,
that
would
be
good
feedback.
So
we
put
it
in
the
data
store
model
and.
J
This
is
Alex
and
one
also
one
concern
area.
We
got
you.
The
second
item
is
what
I'm
wondering
is
Betty.
Is
this
actually
unique,
ephemera
latest
or
believe
the
sensitive
making?
Therefore,
basically,
from
from
this
perspective,
I'm,
not
exactly
sure.
Actually,
if
this
is
the
right
match
to
put
this
here,
I
think
everything
else
is
this.
One
I
have
some
just
just
some
concerns
in
my
mind
as.
J
I
D
Yeah,
okay,
so
non-sensitive
define
the
stream
I'll
come
back
to
the
first
question
that
the
structure
makes
ephemeral
fairly
easy
right
before
a
model
base.
It
even
makes
it
easy
if
you're
overlaying
bgp
peer
as
a
ephemeral,
because
if
you
have
your
bgp
peer
added
to
the
ephemeral,
when
you
add
the
bgp
to
the
operational
piece
you're,
just
overloading
that
they
dynamic
data
store
into
the
intended
again,
I'm
tempted
to
leave
it
in
until
we
have
a
few
examples
just
because
I
think
it
might
be
working
within
it.
D
It's
just
we'll
have
to
pay
attention
when
we
do
the
first
bgp
peer
edition
and
that
coming
after
the
IDR
seems
to
be
strongly
in
favor
of
the
rice
datastore
for
BGP,
that
we
they'll
be
a
on
our
s,
BGP
picture,
because
that
was
one
of
them.
The
other
one
is
the
rib
case,
which
was
our
original
case,
and
that
fits
within
a
data
store,
so
I
think
it's
the
BGP
ephemeral
case
where
we'll
test
this
out.
I
anticipate
that
particular
draft
to
resurface.
D
Okay,
take
a
look
at
it,
I'm
a
I
think
Alex
and
I
will
maybe
make
another
pass
on
the
comment
on
notification.
Stream
alia
did.
Nor
did
we
restrict
the
non
secure
stuff
to
notification,
a
dit
to
notification
and
read-only,
so
we
need
to
put
both
notification
read-only.
This
is
an
incredibly
short
draft,
so
just
take
a
peek
at
it.
Okay
give
me
one
moment
while
I
is
there
any
other
questions?
D
Any
other
concerns
about
this
revised
this
ephemeral
data
store
any
I
mean
this
basically
is
fairly
easy
again
because
of
the
excellent
work
done
in
in
net
mountain
rest,
come
okay,
it
is
Alex
and
Mies
desire
that
we
put
this
to
working
group
last
call
will
make
a
couple
revision
and
then
the
working
group
last
call
will
be
fairly
quickly.
Okay,
give
me
one
moment,
while
I.
D
A
D
Okay,
is
there
well
I'm
afraid
I
lost
the
winner,
I
switch
the
screen.
K
Right
to
help
you
yeah,
okay,
thank
you.
This
is
there
something
about
our
fabric,
based
management
for
this
internet
world,
and
this
night
is
being
prepared
presented
several
times
in
this
group.
So
can
you
who
knows
yeah?
This
is
some
recap
and
the
concept
of
this
design
is
what
is
in
the
network.
We
want
to
simple
eyes
our
taste
and
they
sent
a
network
of
management,
so
we
divided
it
layers
into
three.
K
The
moister
one
is
the
physical
layer
and
the
middle
one
is
the
fabric
layer,
which
is
for
fehb,
Network,
ok,
street
or
in
a
network
layer,
and
the
layer
is
for
users.
As
you
can
see,
there
are
many
users
slice
over
that
in
their
layers,
so
this
is
their
general
concept
Thanks,
and
this
is
what
we
provided
in
our
job,
sending
our
modules.
K
Actually,
we
are,
we
have
done
some
working
I
twice,
which
provided
the
fabric
based
a
topology
many
men
as
this
layer
which
Esther
obstructed
the
physical
layer
into
the
network
provider
and
provided
how
the
physical
layer
network
looks
like,
and
the
network
provider
can
looks
the
whole
network
as
composers
to
offer
several
network
fabrics.
So
at
the
higher
layer,
which
is
the
service
layer,
we
provided
a
service
interface
for
users
who
defines
their
networks.
So
next,
so
the
previous
nice
just
give
you
a
recap
of
this.
K
What
what
kind
of
work
we
are
doing
in
height
wise-
and
this
is
the
progress
since
last
right,
ITF
meeting
and
thanks
for
everyone's
work,
we
get
consensus
on
the
fabric
apology
layer,
which
is
the
second
layer
which
provided
for
the
network
provider
to
manage
their
networks.
So
this
is
a
top
by
the
working
group
and
all
the
modules
and
we
remove
some
models.
K
Duplicated
definitions
between
these
kind
of
two
modules
and
also
we
provide
the
nmda
structure
and
to
pastured
to
resolve
some
errors
in
the
young
validation
and
currently
the
two
modules
don't
have
any
arrows,
and
there
are
some
warnings
and
I'm
trying
to
fix
in
the
next
version.
Yeah,
that's
the
progress
and
updates,
and
the
next
yeah
then
access.
We
welcome
further
feedbacks
and
to
help
you
improve
these
modules
and
we
were
prepared
next
version
with
no
errors
and
no
warnings
do.