►
From YouTube: IETF95-I2RS-20160407-1620
Description
I2RS meeting session at IETF95
2016/04/07 1620
D
A
E
E
C
F
A
Okay,
there
must
be
a
someone
left,
their
house
they're
going
to
be
unhappy.
Some
pie
or
sister
know
anything,
that's
correct,
that's
a
max!
All
I
mean
hey,
let's
go
if
it's
just
I'll
just
put
it
here,
I'm
sure
someone
will
come
back,
hurt
and
there's
another
one.
What
we
really
rock
okay.
H
F
G
I
B
F
B
F
F
H
E
E
A
I
H
G
A
A
So,
thank
you
for
coming.
There's
blue
sheets
going
around
this
is
going
to
be
a
session
as
Jeff
said
that
we're
going
to
go
through
the
summary
and
then
we're
going
to
pull
up
chairs
and
have
a
different
type
of
session.
We're
going
to
draw
pictures
up
here
ask
questions
and
talk
through
issues,
so
don't
feel
embarrassed
if
you
want
to
see
the
summary
and
don't
want
to
go
the
Deep
Diver.
We're
happy
with
you
being
here
just
to
get
a
background.
A
Well,
not
work!
You
don't
have
something.
Okay,
our
whole
goal
in
this
next
session
is
jeff
and
I
have
are
trying
to
push
version
one
out
we're
trying
to
get
a
set
of
requirements
for
this
protocol
to
the
is
cheap,
I'm,
a
fifth
the
protocol
to
the
SG
and
that
come
and
then
do
hackathons,
because
we've
specified
everything
we
can
and
I
think
we're
tripping
over.
You
don't
have
to
run
it
you're
tripping
over
things.
A
And
those
are
ephemeral,
state
and
data
flow
and
protocol
straumann.
They
all
do
have
to
do
with
how
the
data
flow
works.
What
sort
of
extra
requirements
we
have
do
we
need
xml?
Do
we
need
jason?
Do
we
need
to
worry
about
how
much
it
overloads
and
then
validation?
So
this
is
what
the
data
flow
is
that
protocol
strong
man
is
trying
to
give
you
the
examples
in
rest.
A
Company
come
go
ahead:
Jeff
the
T,
the
rib,
the
filter,
the
topology
models
are
all
working,
fine
go
ahead
and
we're
going
to
try
in
terms
just
to
push
this
over
and
we'd
appreciate
your
help
for
review.
We're
just
trying
to
in
this
face
to
get
the
requirements
hand
it
off
to
net
come
and
see.
If
we
can
start
the
net,
come
back
and
forth,
go
ahead.
Okay,.
A
Okay
r2s:
this
is
my
hackathon
picture
and
keep
going
out
one
more
diverse,
okay,.
J
A
A
Wrong
way,
that's
cool!
Ok,
so
we
originally
started
a
hackathon
and
we
thought
we
would
participate,
went
kind.
Liam
did
the
delete
subscription
to
the
gang
push,
and
then
we
try
to
just
get
an
environment
because
in
it
in
hey
the
hackathon
94,
we
had
much
problem
trying
to
get
a
no
dl
environment
up
this
time,
I
pre-loaded
the
OTL
environment-
and
it
took
me
about
20
minutes
downloaded.
That
was
a
long
time
and
then
run
it.
It
was
much
better.
I,
of
course
ran
it.
A
Oh
the
odl
environment
on
the
ubuntu,
well,
Edwin
Kordell
Oh
had
done
that
and
set
it
up
for
us,
and
he
couldn't
be
here
so
he
sent
it.
Our
problem
is,
he
was
probably
running
on
a
fairly
large
server
with
multi
course.
So,
just
kept
pouring
on
our
laptops
going
back
and
forth,
so
being
hackers,
we
decided
we're
going
to
try
a
new
goal.
Maybe
it
was
vertical
and
our
goal
was
to
set
up
an
environment
with
you,
bunt
to
native
and
a
virtual
box.
Try
to
get
min
net
min
x.
A
Net
set
up
routers
use
the
Yuma
pro,
which
is
a
downloadable
free
net
Kumpf
CLI
CLI
to
net
come
use
quagga
as
a
router,
so
we
can
add
and
delete
routes
and
use
cisco,
tell
f
as
our
net
conf
server.
That
would
act
like
the
arduous
agent
and
try
to
add
a
model
and
myself
thon,
thetic,
Jason,
Stern,
Lucy
and
Mamadou
and
Mamadou.
If
you
hear
I'm
so
sorry,
if
I've
mispronounced
it
again
at
0
and
hits
it's
really
cool
because
he's
a
first-time.
A
Participant
from
French
Ghana,
so
I
was
real
thrilled
about
that,
so
it
actually
worked.
We
got
further.
This
is
what
we
did
in
about
six
hours.
We
got
the
builds
working.
We
figured
out
the
CLI,
we
got
downloaded
the
company,
we
got
it
up
and
going
men
that
took
a
little
bit
more
to
put
in
the
virtual
environment.
It
was
without
the
virtual
environment
in
the
regular
environment.
The
stuff
came
up
and
I
was
in
the
middle
of
hacking.
A
This
RPC
now
the
way
that
I
thought
to
do
it
in
the
hackathon
was
simply
to
make
it
as
another
protocol
like
ospf
and
when
it
wrote
over
the
configuration
it
wrote
over
it,
which
means
it
sits
with
a
higher
priority,
seems
really
simple.
Okay,
keep
in
mind,
this
was
a
really
simple,
so
it
just
had
a
set
of
glue
and
it
looked
like
another
routing
demon
and
it
wrote
routes
into
the
route
table
should
be
that
simple,
hopefully
with
lots
of
routes.
Okay,
that
gives
you
all
the
details.
You
can
read
it
online
now.
A
The
wreath
the
reason
the
push
subscription
is
the
pub
sub
was
required
from
ITRs
and
so
Eric
was
is
gotten
code
in
od
Allen,
Oh
dl's
data
model
that
they
use
for
topology
is
the
IRS
data
model.
So
you,
basically,
if
you're,
using
that
you're
using
an
ir,
2
s
data
model
already
and
that
data
models,
basic
concepts
are
what
T's
is
using
for
traffic
engineering
and
they're
doing
tease.
So
this
is
a
fairly
simple
piece:
okay,
so
they
had
a
net
comp
API.
A
They
pushed
it
through
subscriber
to
net,
come
to
the
publisher
where
they
published
interface,
States,
ACL
and
routes
sound
familiar,
and
they
also
had
a
subscriber
that
was
just
a
terminal
and
that
would
they
were
working
on
the
delete
code.
Delete
subscription
and
they've
already
demonstrated
the
create
okay,
so
they've
been
working
along
and
the
reason
we
finally
got
you'll
find
out
later
on.
Why
we
said
we've
just
started
implementing
it,
because
it
makes
a
whole
lot
more
sense.
A
If
you
start
playing
with
the
code
and
there's
bits
and
bytes
in
this
one
tonight
go
ahead,
we
don't
have
our
router
up
and
glued.
So
we'll
post
the
source
code
when
we
get
it
done
and
we
hope
to
show
this
and
the
another
demo
at
ITF
96.
We
hope
to
show
both
the
filter
based
rib
and
the
route
rib
and
the.
A
A
Okay,
this
is
where
we
had
a
long
discussion
on
Tuesday
and
what's
the
background
for
this
one
of
the
fun
things
that
chair
gets
to
do
before
they
send
things
off.
Let
me
take
one
step
back
I
to
asses
a
reused
protocol,
as
I
said
earlier,
be
reusing
net
compn,
we're
we're
using
rest
comp
as
our
chosen
protocols
and
we're
telling
the
net
conf
working
group,
which
is
handling
that
conf
andres
comp.
What
changes
they
should
make.
This
is
a
new
style.
The
is
G's
encouraged.
A
So,
as
you
see
it's
already
borne
fruit,
there's
the
push
model
coming
it
that
comp
for
lots
of
things
but
I
r2s
is
coming
in
with
a
femoral
state
and
some
of
this
bush
and
traceability.
So
what
we
have
to
do
is
look
at
it
and
say
before
we
send
them
version
one.
What's
left
that
we
have
to
work
on
data
requirements,
so
the
chair,
it's
the
joyful
thing
of
finding
someone
or
doing
it
themselves
to
actually
look
at
the
data
flow
requirements
and
I
took
a
crack
at
it
myself.
A
But
please,
if
you
have,
are
better
at
this,
don't
feel
shy.
So
the
first
thing
we
have
and
we
need
really
some
implementation
notice-
how
we're
beginning
to
turn
from
this
is
wonderful
Theory
to
what
works
one
of
the
things
than
I.
We
started
this
process
a
year
ago,
maybe
more
18
months
ago,
we
went
out
and
said
to
people
who
were
doing
odl,
who
were
doing
other
open
source
who
wanted
this
I
to
us
in
the
face.
So
what
do
you
want?
A
A
So
that's
one,
then
someone
said
well.
Maybe
I
went
a
little
bit
more.
I
really
want
referential
checks.
I,
don't
want
the
musts
I,
don't
want
the
leaf
breaths,
because
that'll
take
more
time.
Ok,
that's
another
option!
The
standing
option
is
the
full
netcom
function.
Given
no
real
strong
thing
for
version
1,
we
will
do
the
net,
come
full
validation,
which
may
take
time
if
we
get
six
months
and
we
get
a
lot
of
inflammation
feedback
will
send
back
a
comment
saying
we
need
to
have
the
minimum
referential
now
I'm
going
to
pause
here.
A
A
Ok
now
the
second
one
we've
agreed
that,
although
some
people
are
using
proto,
buffs
and
IP
fixin
some
of
their
prototype
code,
they
really
didn't
provide
us
enough
information
or
input
to
start
using
proto,
buffs
and
IP
fix
in
the
first
version.
We'd
be
glad
to
go
on
to
it,
we'd
be
glad
to
change
to
it.
Not
at
this
point,
that's
all
we
have
is
a
standard
netcom
press
comp,
XML
JSON
formats,
which
are
defined
I'll
pause
here,
any
other
implementers
in
the
house
who
want
to
say
something.
A
K
Right,
Jason,
sterner
I
know
this
one's
come
up
a
few
times.
I
understood
you
guys
are
converging
on
making
a
couple
of
mandatory
implement.
This
is
mandatory
to
implement,
but
also
in
the
preliminary
spec
you're
going
to
try
to
have
a
hook
that
would
allow
other
proprietary
extensions
or
whatever
that
made.
You
start
off
the
proprietary
and
may
get
standard
later,
but
that's
that's
too.
E
The
experience
we're
seeing
it
a
couple
of
the
vendors
that
do
have
this
stuff
in
flight.
Is
that
not
everybody's
starting
with
Metcalfe,
not
everybody,
starting
with
you
know
any
particular
external
transport,
such
as
your
pc
or
thrift,
and
basically,
what
we're
likely
to
see
is
a
rendezvous
towards
common
data
models,
which
is
something
we're
seeing
IETF
why'd.
You
know
some
level
of
common
transports
and
at
some
point
we're
all
going
to
have
to
do
something
to
interoperate
right
now.
All
I
TF
has
is
a
common
set
of
things
that
are
IETF
specifications.
E
A
Is
no
reason
after
we
shipped
the
first
protocol
requirements
over
the
net
count
that
we
can't
take
a
new
set
of
requirements
and
finalized
as
Jeff
and
I
have
found
it
takes
a
longer
time
than
we
thought.
So,
if
you,
if
we
close
on
this
and
next
week,
you
say,
or
in
a
couple
weeks-
and
you
say-
oops
I
missed
it
just
start
in
with
proposing
something.
Okay
now
the
next
one
is.
We
need
to
have
resource
constraints.
Ok,
so
the
archer
s
actions
were
overload
the
router
or
the
switch.
A
Now
this
may
be
plugged
into
some
of
the
OEM
that
helps
you
track
it.
But
this
is
a
piece
where
we're
going
to
say:
we
can't
overload
the
router
with
our
2s
and
we've
got
to
have
some
ability
to
to
bring
it
back,
that's
sort
of
where
we
got
on
Tuesday.
Other
people
have
asked
right
now
we
have
a
single
pane
of
glass
for
config
in
a
single
pane
of
glass.
That's
been
on
the
list
that
and
I,
and
you
have
to
think
of
that.
A
Just
like
what
I
had
in
the
demo,
the
config
writes
a
value.
There's
a
priority
between
the
two
pans.
Excuse
me,
they're,
two
panes
can
fake
an
hour
to
s.
Both
panes
have
a
value.
Let's
say
for
the
moment
and
r
2
s
is
10
and
configures.
One
then
I
urge
us
would
win,
and
if
you
wrote
a
route
with
config,
our
choice
would
right
over
it
and
I
will
go
through
some
examples.
A
But
where
we
are
running
in
is
this
used
to
be
the
world
of
config
before
the
OP
state
changes
to
net
mod?
So
some
of
what
we're
struggling
over
is
the
difference
terms
that
really
don't
sort
of
match
what
we're
doing
in
the
demos.
That's
how
we
come.
We
went
to
the
demos
you
may
laugh.
This
is
probably
something
you've
seen
before.
So
this
is
the
previous
world.
This
is
the
RFC
that
says
you
have
metadata
about
config
state
data
notification
operation.
In
your
a
model,
you
have
an
artist,
client
and
server.
E
So
if
I
could
stop
you
here
who,
here
in
the
room,
has
seen
some
form
of
this
app
state
type,
stuff
and
other
presentations,
including
net
mod
and
net
conf
now
this
is
important,
because
a
lot
of
the
places
were
hung
up
on
has
to
do
number
one
with
consistent
terminology
used
in
net
com
and
number
two,
the
impact
of
how
ephemeral
fits
in
with
the
app
states
terminology
that
has
been
introduced
for
open
config.
If
you
don't
understand
this
you're
going
to
be
lost.
So
if
you
have
questions,
please
come
to
the
microphone.
Please.
A
S
now,
because
I
assure
you
the
reason
we're
going
over.
This
is,
if
you
try
to
read
it,
and
then
you
see
it
change,
because
we've
tried
to
adapt
to
the
changes.
That's
hat
that
are
happening
through
the
OP
state
draft.
From
that
mod.
We
just
want
you
to
be
informed.
It
is
a
frustration
to
the
authors
and
to
other
places,
but
it
is
where
we
are
in
the
IDF
now
this
is
the
RFC
6244
and
the
femoral
State
dropped
into
that.
A
K
A
H
A
So
over
here,
I've
got
a
great
hope.
You
hear
me
in
the
back
without
the
night:
yeah.
Okay,
if
federal
stay
is
all
this
year
as
the
applied
config
is
transform
it
coming
from
the
attendance
in
the
ephemeral
day
to
the
five
and
then
when
you
get
to
drive
state,
it's
like!
Oh
that's
right,
I
should
have
done
that
bgp
and
ospf
formulas
do
that,
and
that
was
fun.
Ok,
let's
go
on
so
the
way
we
if
you
see
the
documentation,
we
turned
that
into
intended.
A
A
Ok
memorize,
this
little
document.
If
you
read
documents-
or
just
put
it
next
to
you,
because
this
took
us
a
little
bit
to
dig
out
of
the
Upstate
stuff
now
for
a
very
simple
temperature,
let's
assume
we're
in
an
energy-efficient
world
and
we're
sitting
in
one
of
these
rooms
like
I
was
upstairs
and
the
default
is
to
let
it
go
to
80
and
I'm
getting
sleepier
and
sleepier
and
I
think
we
should
do
something
and
turn
it
back
to
72.
A
So
again,
you
have
the
desired
temperature
degrees,
Celsius,
the
actual
temperature
for
operational
state,
a
very
simple
one.
So
again
you
have
your
desired
temperature
that
goes
from
a
running
to
an
intended,
config
and
then
gets
applied,
and
then
you
have
an
actual
state.
So
what
happens?
If
you
have
this
online,
please
reset
my
temperature
I'm
having
a
hard
time
falling
asleep
in
I
us
well,
the
our
truest
client
would
overwrite
it
over
late.
Excuse
me,
thank
you.
The
trimmers
overlay,
the
80
degrees,
on
top
of
the.
A
On
top
of
the
80
degrees
would
lay
the
ephemeral
72
degrees
and
because
it
for
our
example,
it
has
higher
priority
for
me
to
be
cool,
it
will
turn
to
72
and
that
will
get
applied
and
then
hopefully
sometime
the
derive
state,
which
is
the
temperature
you're
reading,
will
get
down
to
72
and
I
will
be
comfortable
and
not
fall.
Asleep.
Actually,
sue
go
back.
D
Sorry,
hi
Bill,
Fenner
I
have
a
question
which
I've
had
since
I
saw
this
slide
before
so
since
we're
now
talking
about.
Let's
everybody
understand
it,
the
thing
that
I
don't
quite
get
is:
what's
a
femoral
state,
you
only
have
one
state
right.
What
is
ephemeral
state
so
because
I
think
there's
only
one
temperature
in
the
room,
the
current
temperature,
if
I'm
reading
the
temperature
but.
D
A
E
E
So
one
of
the
things
that's
really
causing
us
problems
is
a
lot
of
these
names
are
coming
from
things
that
are
stuck
metcalf
materials.
You
look
at
things
like
candidate
and
that
probably
makes
sense
apologize
to
the
people
online
when
I
kind
of
stain
to
the
camera,
so
candidate
is
specifically
the
store
on
a
Metcalfe
server
when
you're
allowed
to
actually
edit
the
configuration
on
a
temporary
scratch
copy.
That
is
not
in
effect,
so
candidate
has
a
name
that
everybody
thinks
they
know
what
they
mean.
E
It
has
an
explicit
meaning
in
the
net
conf
datastore
protocol.
The
same
thing
is
true
for
start-up.
You
think
you
know
what
that
is.
It
is
specifically
the
data
store
that
is
present
on
some
systems,
because
not
every
system
needs
support,
a
start-up
datos
or
running
configures.
One
of
the
things
that
case
is
the
hell
out
of
most
people,
because
you
look
at
this
new
say.
This
is
obviously
what
the
box
is
running.
No
in
the
net
kind
of
sense
of
things.
It
is
the
data
store
that
reflects
a
config
that
has
actually
been
committed.
E
They
also
have
a
form
of
the
Davis
store
that
is
writable
running.
So,
if
you've
used
the
old
iOS,
the
routers,
maybe
change
something
on
the
fan
line.
It
took
effect
immediately.
That
was
a
writable
running
type
of
count
of
criteria.
So
those
are
despite
the
fact
that
they
have
put
very
simple,
very
English,
normative
names,
don't
mean
exactly
what
you
think
they
are
there,
specifically
data
stores
within
a
net
of
context.
E
Now
we
start
getting
into
things
that
are
not
in
the
existing
that
count
documentation.
Some
of
these
things
are
verbiage
that
have
introduced
as
part
of
the
open
config
efforts
to
document
operational
state,
no
operational
state
in
net
conf
today,
just
as
simply
a
config
false
note
allows
you
to
read
the
current
state
of
what
is
actually
in
effect.
E
E
Config
is
basically
the
state
that
happens
once
you've
applied
a
running
config
and
you
say
go
the
thing
is
there
could
be
some
level
of
latency
from
the
point
that
you
apply
a
commit
for
a
running
config
to
the
point
where
the
running
config
has
an
impact
on
the
state
of
a
box,
and
this
can
vary
for
a
lot
of
different
reasons,
though
of
his
local
box.
It
probably
commits
almost
instantly
very
large-scale
box.
E
You
do
something,
and
maybe
it
takes
it
a
while
before
it
actually
goes
into
effect,
now,
maybe
ten
seconds,
maybe
10
minutes
or,
if
you're,
in
a
massive
distributed
system.
You're
like
we're
looking
at
for
some
of
the
remote
mount
stuff
in
net
com.
This
may
be
something
you
change
to
a
local
data
store
and
it
may
take.
You
know,
half
an
hour
before
supply
to
the
network
we
get
into
the
femoral
dick
Adele
candidate
stuff.
This
is
where
we
start
introducing
hrs
concept.
So
what
do
we
mean?
Is
I
dress
by
ephemeral?
E
E
Now
we
now
have
ephemeral
state
as
a
layer,
so
you
have
running
you
have.
On
top
of
that.
You
know
some
of
your
priority
is
to
set
up
that
way.
You
know
some
amount
of
state
that
is
being
overlaid
or
overridden
and
that
state
you
know
if
you
haven't
clear
the
I
trust
piece
of
it
would
flush
out
and
you
would
still
have
the
original
statically
configured
stuff
question.
J
E
E
Top
left
so
I'm
going
to
draw
analogy
to
iOS,
XR
or
no
juniper
boxes.
When
you
do
edit
config
make
a
change
and
to
the
point
where
you
before
you
type
commit,
so
you
can
actually
show
what
you've
changed.
You
can
actually
do
things.
You
can
see
what
you're
you
know.
Basically,
what
is
the
scratch
add
before
you
write
it
out?
That's
the
candidate.
L
Either
briskin
are
the
Jeff.
Can
you
elaborate
a
little
bit
on
the
difference
between
intended?
Config
can
apply
that
config,
but
in
particular,
is
it
just
a
timing
issue
or
maybe
the
sewer
basically
decided
to
apply,
say
most
of
the
intent
intended
configuration,
but
not
all
of
them.
So
it
will
be
like
a
permanent
difference
between
internet
and
applied
I.
E
Will
make
my
best
attempt?
You
know
this
is
one
of
the
points
where
you
know
the
upper
config
guys
have
been
iterating
with
the
net
camp
group.
Trying
to
get
this
stuff
nailed
down.
It's
tricky.
They
have
obviously
one
thing
in
mind
and
I
TF
is
taking
ty
sort
of
fully
converge
on
it.
To
the
best
of
my
understanding,
it
is
exactly
that
lately,
so
I
do
make
know
again.
Sticking
to
a
Juniper.
Router
is
an
example.
E
Since
that's
what
I
know
I
edit
my
config,
you
know
that
would
normally
show
up
in
the
canada
config
I
type
commit
it.
Moves
to
the
running
data
store
that
runnings
datastore
is
now
effectively
the
intended
configuration
that
will
take
some
amount
of
time
to
percolate
through
the
system
come
apply
and
I'm
about
to
be
corrected.
I
think
Jason's.
K
E
Time,
thank
you.
That
was
actually
the
detail
that
I
was
missing.
So
in
some
cases,
though,
you
may
say,
go
prepare
for
something
that
could
actually
exist
at
a
later
point
in
time
and
then
we'll
get
to
the
PI
config.
You
know
when
such
things
happen,
for
example
you
plug
in
that
line
card.
It
comes
up.
We
knows
what
to
do
because
of
that
configuration
that
was
intended
for
the
thing
and
then
sort
of
at
the
bottom
of
the
stack
is
once
we've
actually
done
all
this
stuff.
We
actually
have
the
drive
state
now.
J
E
Was
actually
one
of
the
discussions
that
we
were
having
just
before,
we
left
the
room
from
that
camp
and
no
Kent
Watson
was
suggesting
to
us
that
perhaps
a
better
way
to
draw
that
picture
is
that
you
know
in
the
intended.
Config
of
the
device
is
probably
closer
to
the
running
config,
which
is
coming
from
what
we
tended
to
call
static
configuration
in
hrs
sense
with
the
ephemeral
config
applied
to
that
as
well,
either
as
overlay
or
override
mechanism.
So
we
intended
probably,
as
the
union
of
these
two
I.
J
E
E
D
I
should
let
you
go
down
there.
I
still,
don't
understand
what
ephemerals
to
you.
It's
like
I,
I,
understood
everything
you
said,
but
there's
still
just
one
state
like
the
it's,
not
that
the
normal
temperature
is
72
in
the
in
the
room
and
the
ephemerals
temperature
is
something
else
that
that's
the
roomie
72.
The
desired
temperature
is
different
between
the
intent
of
Keith,
Engel
and
femoral
config,
but
the
state
there's
only
one
state
state
is
what
is
happening
right
now.
D
A
When
you
write
the
yang
yeah,
when
you
write
the
yang,
you
actually
write
the
end
just
the
way
you
write
the
regular
configuration
except
you
use
the
word
ephemeral.
So
you
have
the
femoral
config.
That's
that's
a
femoral
config,
but
in
your
config
area
you
can
have
notifications
right
when
you
write
your
yang.
There's
a
notification
or
you
can
have
an
RPC
operation
or
you
can
have
operational
state
actually.
E
I
think
I
see
where
your
question
is
going
from
so
part
of
the
problem
here
is
that
if
we
go
back
to
the
actual
yang
for
the
thermometer
model
yeah,
so
the
the
thing
that
is
somewhat
confusing
here
is
we
provided
an
example
no
with
a
desired
state.
This
is
no
piece
of
configuration
and
we
have
no
a
piece
of
information
that
also
say
what
this
thing
actually
is
right.
Now,
that's
what
we
see
in
the
screen
in
front
of
us
and
I
think
next
slide.
Has
the
ephemeral,
config
part.
E
From
role,
which
is
your
exam,
there
we
go
so
this
is
what
happens
when
we
actually
apply
ability
to
do
if
from
roll
type
of
stuff
and
we'll
come
back
to
the
full
check.
You
know
conversation
to
the
second.
This
is
an
example
of
a
data
model
that
allows
for
static
configuration
to
be
overlaid
with
a
temporary
value
that
does
not
persist.
E
They
can
wonder
or
poke
a
couple
buttons
it's
going
to
take
effect
for
a
period
of
time
and
if
the
thermostat
resets
it
goes
back
to
20c
that
state
of
putting
in
the
temporary
value
that
is
on
top
of
you
know
these
static
configuration.
Is
the
ephemeral
configuration
state
right?
What's
the
ephemeral
state
ephemeral,
operational
state?
Yes,
okay,
so
this
is
where
we
start
to
know
that
this
example
is
not
the
best
example
of
that,
and
that's
probably
why
you're
confused.
E
So
this
is
an
example
in
a
configuration
schema
where
we
want
to
override
static
things.
This
is
actually
nice,
although
it's
several
I
etfs
back,
we
did
these
nice
Venn
diagrams
about
how
no
static
configuration
and
ephemeral
configuration
could
be
either
fully
destroyed.
They
never
touch
each
other
or
we
may
allow
for
some
amount
of
ephemeral
configuration
to
override
you
know
these
data
configuration.
So
in
this
case
we
have
no
ephemeral
configuration
state
where
the
node,
that
is
the
operational
state,
happens
to
be
shared.
Obviously
it's
the
same
thing
that
we're
looking
at.
E
We
want
to
know
what
the
effective
temperature
actually
is.
Other
modules,
especially
the
ones
that
are
fully
disjoint,
may
have.
We
know,
for
example,
a
complete
subtree
know
for
like
ephemeral,
static
routes
that
do
not
share
the
same
schema
as
the
system
static
routes.
If
that
is
a
fully
disjoint
model.
In
that
case,
the
operational
state
is
ephemeral.
It
may
not
either
there
may
be
no
nodes
there
whatsoever,
even
if
the
only
container
know
is
expected
to
be
generally
present
its.
Is
it
fully
ephemeral?
A
D
A
May
bill
I
may
take
it
out
for
future
things
because
it
seems
to
have
caused
more
configuration
and
more
confusion
than
aided.
Those
are
just
logical,
diagrams,
yeah.
E
And
I
think
the
thing
that's
confusing
is
that
you
know
ephemeral
state.
No,
it's
just
its
standard,
no
config
falls
as
far
as
know
the
system
and
the
consumers
are
concerned.
The
big
difference
is,
you
know
in
terms
of
labeling
the
stuff
and
making
it
present
in
the
yang
model.
The
model
is
when
the
entire
model
is,
you
know,
ephemeral.
True,
no,
there's,
no
static
configuration
see
it's
not
guaranteed
to
be
present
on
the
system.
That's
that's
really.
The
subtle
difference,
questions
please.
This
is
the
time
we
want
to
walk
through
this
target.
M
E
M
E
L
E
In
the
case
of
default
values,
especially
in
the
context
of
ephemeral,
configuration
that
does
not
overlap,
static
configuration
mm-hmm,
it's
really
no
different,
that
the
difference
really
being
led
different
default,
much
like.
Instead,
a
configuration
only
gets
instantiated.
When
a
related
note,
it's
not
done
so
like
you
know,
for
example,
a
leaf
node
that
has
got
a
default
value,
may
only
instantiate
when
the
container
that
contains
it,
your
AG
leafless
that
contains
it
becomes
instantiated
itself.
So,
if
you
instantiate
a
ephemeral,
isa
configuration
the
f
the
related
default.
They
also
instantiate.
So.
L
E
To
put
I
think
in
the
concept
of
what
the
open
config
guys
are
looking
for,
it
is
part
of
the
intended
config
and
it
may
actually
show
up
as
a
exclusive
node
in
Indian
is
definitely
a
show
up
as
part
of
the
operational
state.
There
may
not
be
a
note
present
in
the
configuration
tree,
but
it's
effectively
there.
So
it's
a
portion
of
the
configuration
it's
I
did
ephemeral,
stuff.
Alright,.
L
So
basically,
its
intended
right
and-
and
the
other
question
is
that,
let's
say
I
was
asking
for
10,
but
because
of
like
a
local
policy
limitation,
I
can
go
up
to
nine.
Is
it
possible
that
the
server
will
do
line
and
it
will
be
applied
configuration
and
they
will
be
different
or
it
should
reject
the
configuration
response?
As
are
you
getting
validation
and
we're.
E
E
So,
given
this,
given
what
these
things
mean,
there
are
impacts
on
the
net
com
protocol
and
the
rest
conf
wrote
of
all
is
but
stick
mostly
with
net
comp,
because
that's
a
slightly
more
complicated
version
again.
The
stuff
that's
across
the
top
are
things
that
are
netcom
data
stores,
and
you
know
the
net
com
protocol
was
very
much
set
up
to
serve
the
net.
Have
data
stores
you
do
an
edit
config
by
default.
E
If
you
don't
specify
the
data
store,
your
editing
candidate,
if
you
are
doing
a
get
me
on
by
default,
you're
getting
you
know,
operational
state
that
happens
to
include
running
the
ephemeral
stuff
does
not
exist,
and
even
if
isset
I
tourists
aside,
you
know
the
whole
point
that
the
open,
the
guys
after
their
operational
state
stuff-
and
you
know,
they're
the
sort
of
disjoint
behavior
between
now.
What's
intended
versus
what
supplied?
That's,
not
something
that's
an
accord
netcom
protocol.
So
this
is
something
that
net
comp
is
having
to
deal
with
completely
sighs.
E
My
dress,
I
tres,
is
really
adding
just
an
extra
layer
of
subtlety
there
and
you
know,
maybe
helping
to
suggest,
go
part
of
the
solution
space,
so
edit
config,
when
you
go
to
actually
change
something
once
a
target
data
store,
what
you're
doing
you
can
specify
you
know
in
a
system?
That's
right,
well
running
you
can
edit
config
directly
on
running
or
you
can
do
candidate.
It
also
allows
for
other
data
stores,
so
I
can
actually
say.
E
Eddakath
faked
target
equals
ephemeral,
for
example,
and
whether
this
is
louder
nut
will
be
determined,
know
whether
or
not
be
actually
have
actually
go
back
to
the
angley's,
whether
that
we
have
a
femoral
true.
So
basically,
what
one
would
expect
is
that
if
you
do
it
edit
config
and
a
federal
true
note
is
not
present
as
part
of
the
relevant
configuration
tree.
No,
it
shouldn't
take
effect,
it's
probably
a
validation
or
some
sort.
E
Okay,
go
back
to
the
other
side.
Look
one
more
so
same
thing
is
true
of
yet
get
has
two
purposes
inside
of
net
cough.
It
gives
you,
when
you're
acquiring
a
subtree.
What
the
configuration
that
actually
is
running
is
it's,
actually,
they
applied
configuration.
Sorry,
the
intended
configuration
from
the
context
of
the
open
config
guides
now
mostly
running,
but
it's
effectively
know
what
the
system
actually
has.
No
in
terms
of
its
applied
configuration.
E
So
that's
that's!
Actually
it
is
down
to
the
blue
box.
No,
this
is
know
what
the
now
this
is.
The
running
state
of
the
configuration
is
stick
to
the
net
comp.
Only
for
a
second,
you
know
with
us
without
I
dress,
when
you
do
a
get
you're
going
to
get.
For
example,
you
know
if
you're
talking,
bgp
here's,
this
PGP
configuration
here's
the
things
that
have
actually
been
configured
and
have
been
conveyed
and
they're
part
of
running.
You
will
also
get
config
false
notes.
Now
this
is
outright
operational
state.
E
It
doesn't
exist
in
something
that
you
can
edit
the
thing
that
changes
when
you
apply
I
to
RS.
No,
to
this
general
picture
is
your
it
now
able
to
do
a
get
that
will
include
any
how
ephemeral,
II
instituted
state,
which
might
be
how
they
config.
So
if
you're
doing
in
the
thermostat
example,
a
overlay
of
the
intended
temperature,
no
so
you're,
sending
it
down
to
18
c,
has
your
little
bit
chilly
you're
a
little
bit
too.
When
you
want
to
cool
down
a
little
bit
more?
No,
that's
a
temporary
value!
E
You
would
expect
that
the
show
up
in
your
get
because
you've
actually
done
that.
As
you
know,
what
is
my
intended
temperature
right
now?
It
is
no
longer
what
the
nose
static
configuration
says.
It's
funny.
I
should
be
able
to
get
the
18
there.
Does
that?
Does
that
piece
make
sense,
I'm
saying
some
heads
nodding
those
that's
good
same.
K
E
We're
getting
there
so
if
you
do
a
get
config
and
which
I
hadn't
actually
gone
to,
if
you
do
a
get,
you
should
get
the
configured
ephemeral
value
because
that's
what's
the
applied
one,
the
system
level?
Your
question
goes
into
the
next
observation,
which
is,
let
me
finish
the
other
piece
again.
The
other
piece
of
get
is
you're
going
to
get
any
config
false
notes
that
are
present.
That's
just
a
normal
thing
that
doesn't
change
with
hrs
the
fact
that
something's
ephemeral,
febrile
operational
state
religious,
where
it
lives
in
the
module
Jeff.
E
Get
off,
okay,
get
get
you.
Both
configuration
and
config
fall
state
config,
good,
configure.
We
get
victory,
and
now
we
get
to
the
get
config-
and
this
is
this-
is
where
we
start
getting
to
these
subtleties
know.
So
you
know
I'm,
and
this
is
really
where
our
discussion
is
with
Metcalfe
know.
How
do
we
actually
figure
out
when
we
have
these
things
that
are
overlays
of
some
sort?
We
now
have
to
figure
out
what
thing
do
I
really
mean
to
get
when
I
do
get
most
user
applications
want
to
see.
E
You
know,
for
example,
what
is
by
no
applied
configuration
state
when
you
have
an
ephemeral
thing,
you
want
to
see
that
thermostats
been
set
18
and
not
the
20
same
thing
for
operations
that
you
see
want
to
see
what's
actually
running.
But
how
do
you
actually
see
where
the
things
coming
from?
Do
you
know
how
do
I
actually
asked
the
system?
What
is
the
actual
static
configuration
that
you're
going
to
return
to?
If
I'm,
actually
writing
a
system
to
manage
the
static
configuration
I
want
to
do
a
git
config?
What
does
that
return?
E
Does
it
return
the
static
configuration
which
is
running?
Does
it
return?
No,
the
ephemeral
configuration
the
get
can
think.
Rpc
does
specify
a
source,
no
operation.
You
could
say
source
running.
That
is
one
hundred
percent
clear.
That's
always
the
static
and
fake
you
could
say
source
ephemeral.
That
is
one
hundred
percent
clear.
The
thing
that
we
don't
really
have
is
sort
of
a
logical
view,
a
union
that
says
show
me
the
thing
that
is
on
top
in
terms
of
the
configuration
I
went
into
what
the
config
is
worthy
static
and
the
ephemeral
are
interacting.
E
K
E
Where
the
interaction
with
the
open
config
stuff-
because
we
were
talking
about
this
ephemeral
stuff
before
open,
they
made
our
lives
even
more
complicated.
So
we
do
have
to
reconcile
that
sort
of
big
view,
and
this
is
part
of
what
they
look.
If
you
guys
are
asking
for
what
might
be
the
protocol
configuration
changes.
The
RPC
changes
to
allow
you
to
see
each
of
these
either
logical
views
or
operating.
E
So
the
reason
why
we're
actually
having
this
discussion
here
is
because
know
when
you
ask
any
three:
no
people
inside
a
net
conf
note
for
help
on
this.
There
would
be
get
slightly
three
different
answers.
No,
this
is
more
now.
We
now
have
four
now
so
the
part
of
the
goal-
the
goal
of
this
session
here,
because
we've
had
these
discussions
several
times
with
the
guys
and
netcom-
is
to
get
majority
the
art
urs
working
group
behind
you.
What
the
ideas
mean.
E
That's
not
fuss
over
what
the
terminology
is
in
terms
of
the
functional
component.
So
what
are
the
functional
requirements?
Understanding
the
functional
requirements?
Let's
refine
the
functional
requirements,
so
they
can
be
implemented
against
what
net
conf
is
providing
in
their
protocol
inside
of
their
RPC
operations
and
inside
of
the
interaction
Natalie
with
ephemeral,
config
and
Static
config,
but
also
the
Oakland
config
operational
state
stuff
ants
are
both.
E
Again,
it
was
like
the
protocol
operations
in
dec-
calm,
not
really
hard,
but
the
trouble
is
no.
We
we
have
strong
context
of
when
you
do
a
get
gets
normally
give
you
config
plus,
know
your
operational
state.
Now
you
can
fake
fall
stuff.
Maybe
we
want
to
have
a
filter
attached
to
get.
This
is
one
of
the
things
that
the
ephemeral
protocol
requirements
document
that
we
put
together
for
to
help
to
try
to
help
nick
have
understand
this.
Maybe
sometimes
we
going
to
get
the
actual
state
that
was
present
and
configured.
E
I
want
to
retrieve
the
fact
that
this
was
20
because
get
can
give
you
the
running
state
as
well.
Maybe
I
want
to
get
only
the
ephemeral
pieces
of
configuration
and
one
of
the
requests
from
the
app
the
open,
config
guys
for
the
app
state
stuff.
I
want
to
be
able
to
get
the
intended
versus
the
actor
I've
state.
No,
it
intended
apply
your
I've.
You
know
the
sort
of
lag
between
these
four
different
things
or
the
fact
that
you
may
have
like
no
an
interface
configured
duplex
or
you
may
have
interface
configured
auto.
D
Are
three
things
here
that
you
might
want
to
get
right?
You
might
want
to
get
the
light
blue
box
or
the
green
box
or
the
dark
blue
box
exactly
okay.
The
next
session
starts
in
seven
minutes,
I'm,
sorry
to
say
this
is
a
great
discussion.
I
love
that
we're
having
it
but
I
think
we
might
be
out
of
time.
I
think.