►
From YouTube: IETF103-RTGWG-20181108-0900
Description
RTGWG meeting session at IETF103
2018/11/08 0900
https://datatracker.ietf.org/meeting/103/proceedings/
A
B
C
The
service
model
is
assumed
to
be
used
by
service
provider
to
provide
two
provisioning.
The
C
devices
on
behalf
of
the
enterprise
networks,
enterprise
customers
and
the
sd1
way
pian,
is
assumed
to
be
provide
a
secure
connection
over
multiple
enterprise
sites
and
over
multiple
when
transfer
connection,
and
is
assumed
to
be
to
simplify
the
management
further.
C
The
whole
sty
management
and
this
draft
was
presented
in
las
IDF
meeting
by
China
Telecom
and
based
on
their
development
experience
and
at
the
same
ATF
meeting
in
las
meeting
o
nog
also
proposed
there
open,
as
the
one
exchange
API
based
a
service
model
to
try
to
solve
their
problem
in
the
pride
in
the
enterprise
community.
So
we
think
these
two
model
have
some
relationships,
so
we
we
try
to
work
on
this
and
I
also
presented
this
draft
being
Tuesday.
C
Ops,
working
group
and
I
also
got
some
feedback
from
there
working
the
working
UPS
working
group,
and
they
feel
that,
as
the
one
is
the
concept
that
proposed
in
owner
an
owner
enterprise
community-
and
they
have
they
are
the
end
users
of
sd1
technology.
So
their
wish
our
model
work
should
be
consistent
with
the
Oneg
requirements
of
sd1.
So
that's
the
current
status
of
this
of
the
background
of
this
sd1
service
model
and.
C
C
It
seems
for
now,
as
the
one
has
more
functionality
than
the
sea
bass
Rypien.
For
example,
in
in
spring
working
group
there's
a
draft
is
I
saw
for
sd1.
In
that
draft,
it
is
assumed
the
CEO
jiwon
could
be
attached
to
internet
or
MCAS
Network,
and
the
city
can
also
make
out
three
to
our
seven
classification
and
based
on
the
classification.
They
can
steal
the
flow
of
different
SLA
to
different
paths
and
also
in
the
past
working
group.
C
They
are
working
on
the
CPE
based
city-based
I
was
really
pionship,
so
the
city
can
offer
the
fine
granularity
of
virtual
network
separation
and
and
also
your
insecurity.
Working
group
attack
maintenance
I
to
NSF.
They
are
working
on
a
new
controller
based
Epis
equipment
roaches
to
try
to
solve
the
in
a
lot
like
AB
SEC.
We
can
apply
it
to
a
large
scale.
C
So
our
proposal,
yes,
the
the
goal
of
our
service
model.
We
we
try
to
define
a
a
general
sd1
between
service
model
that
not
only
based
on
the
the
original
already
defined
sea
bass
Vivian,
but
also
has
other
functionality
that
just
a
mentioned
in
earlier.
In
the
best
and
a
spring
working
group,
they
are
working
on
the
the
functionality
of
sd-1.
C
Cannot
work
so
to
our
sites?
Each
side
could
connect
to
MPR's
or
internet
and
and
in
one
particular,
and
in
this
VPN
there
could
be
one
or
more
subway
via
that's
fine
granularity
or
worth
network
inside
an
enterprise
and
Cyprian
can
has
its
own
topology
addressing
in
the
policy,
and
these
are
similar
concept
like
insecure
layer,
3
VPN
and
the
see
can
do
the
layer
3
to
layer,
a
seventh
row
classification
and
to
do
the
multi
path
theory.
So
that's
also
mentioned
in
this
sr4
sd-1.
C
This
this
is
a
like
to
clarify
what
the
onyx
service
model
and
our
models
difference.
There
are
two
major
differences
of
this
to
service
model
of
the
first
one.
Is
the
onek
try
to
define
the
the
service
model
used
in
enterprise
network?
It
means
that
all
the
under,
like
all
the
infrastructure-
oh
here,
oh.
C
Yes,
so
this
infrastructure
belongs
to
enterprise
network,
but
in
the
our
service
model,
the
CPE
they
belongs
to
service
provider.
So
this
is
such
the
major
difference
and
the
second
major
difference
is
the
owner:
try
to
define
the
interoperability
part
of
this
one
as
the
one,
the
two
inter-domain,
how
to
how
this
to
inter
vendor
domain
can
interoperate
and
inter
connotation,
based
on
the
the
service
models
request,
so
that
the
operator
will
send
the
OSD
gateway
service
to
the
to
wonder
controller,
to
try
to
initiate
the
the
connection
the
during
these
two
network,
their
gateway.
C
So
but
our
model
is,
is
defined
for
a
service
provider
network
and
there
also
could
have
some
in
some
case.
There
could
be
also
to
a
network
have
some
like
try
to
provide
an
interoperability
between
the
two
domain.
So
so
this
is
a
two
major
difference,
but
for
now
on,
AG
just
gave
the
the
to
give
their
service
model.
C
They
try
to
define
the
reachability
model
and
the
palace
management
model,
and
but
I
think
there
is
a
gap
here
because
there's
we
don't
know
like
in
a
1:1
or
domain
what
kind
of
functionality
that
sd1
has.
So
it's
not
easy
to
define
the
interoperability
model,
so
I
think
the
model
we're
trying
to
define
will
help
to
clarify
that
part
and
I
think.
Based
on
this
to
include
input,
we
can
try
to
define
a
generic
model.
C
D
Mrs.
Linda
Dunbar.
Thank
you
very
much
for
the
presentation
is
really
good,
so
I
invest
lots
of
timing.
Oh
and
Yugi,
and
Oh
narc
and
I'd
share
one
of
the
working
group.
So
for
the
st1
part
they
basically
have
abandoned
the
effort
because
ona
slightly
different
from
IETF.
They
want
to
catch
the
enterprise,
communities,
need
and
desire,
and
then
initially
five
years
ago
sd1.
So
they
had
sd1
into
working
group
because
just
like
IKEA,
whenever
we
do
any
work
interoperability,
it
takes
very
long
time.
D
So
the
patient
seems
to
run
out
so
now
they
are
all
moving
to
security
because
for
connecting
to
the
cloud
they
find
out
that
enterprise
today
is
more
about
data
portability
and
app
portability,
and
so
sty.
It
was
started
as
the
lower
cost,
but
they
find
out
that
they
need
to
reach
cloud.
So
now
this
kind
of
swing
into
security.
Now
so
this
work
in
oh
and
Eugene,
they
have
to
find
the
api's
yeah.
So
I
talked
to
Steve
wood
and
asked
the
October
meeting.
D
B
D
E
E
Eg
and-
and
it
will
probably
that
would
be-
the
group
I-
think
that
you're
at
this
point,
you
still
you
want
feedback
from
this
group.
Yes,
because
you
know
we're
continuing
sort
of
the
focus
many
sessions
on
st
man
that
were
started
in
last
time
in
montreal.
Continuing
that-
and
you
know
so-
Linda's
presentation
coming
up
next
is
also
about
SD,
WAN
and,
and
so
anyway,
it's
just
just
providing
the
context
that
you're
not
necessarily
trying
to
adopt
this
in
this
working
group.
But
but
it's
related
to
this
focus
that
were
initiating.
D
Sorry
late
and
you
start
right
on
time,
so
unfashionable,
okay,
so
where's
my
presentation,
okay,
so
we
present
this
at
ITM
one
or
two,
and
it's
mainly
about
SD
when
the
problem
statement,
the
motivation
and
the
gaps,
and
we
had
lots
of
lots
of
feedbacks
for
this
problem
state
and
the
cap
analysis
and
today
I'm
concave,
some
update
since
last
time.
Oh.
D
You
just
your
old
one:
is
it
ok?
Ok,
so
so
that
I
just
start
with
some
key
characteristics
of
the
SD
one
which
we
didn't
describe
very
clearly
at
the
last
ITF
I
think
this
is
the
old
one.
D
Ok,
that's
ok,
I'll,
explain
it
so,
okay,
so
my
update
is
besides
CP
1
CP
I
have
a
CP
is
sweet
in
there
to
show
that
the
key
characteristics
is
like
between
CP
1
and
CP
2.
On
this
picture
you
may
have
two
different
interfaces,
or
maybe
more,
but
unlike
the
content
network,
where
an
interface
you
advertise
your
property,
and
here
some
of
the
transporter
interface
is
through
public
Internet
right.
D
So
it
is
local
decision
on
a
CPU
one
to
decide
which
service
to
go
to
where,
like
maybe
at
day,
one
voice
service
goes
through
and
POS
network,
and
maybe
a
day
to
some
other
service
goes
to
the
transport
network
and
then
not
showing
on
this
picture
is
I.
Have
a
CPS
3
and
between
CP
is
3
2
CP
2
I
may
also
have
multiple
networks
right.
Some
may
be
provider
Verizon's
that
maybe
the
18
do
some,
maybe
china
mobile.
D
Ok,
yeah,
ok,
so
so
I
just
start
talking,
so
that
there's
like
another
difference
between
pure
overlay
is
in
SD.
One
will
have.
Some
interfaces
goes
through
the
underlay
like
traditional
and
Kos.
Now
some
interfaces
goes
through
the
overlay
like
a
public
internet
and
for
the
overlay.
Then
you
need
to
set
up
IPSec
tunnel
and
in
the
example
I
show
there
like
CPU
one
CPT,
you
say
ps3,
you
could
have
CP
1
CP
2.
D
D
Thank
you
so
so
in
this
example
like
CP
3
and
CP
1,
they
also
can
have
direct
IPSec
tunnel,
but
because
of
the
time
of
the
day
or
the
particular
traffic
or
maybe
a
regulatory
purposes,
the
some
of
the
traffic
has
to
traverse
to
CP
2
before
you
go
to
CPS
3
and
then
that
one
key
thing
is,
the
decision
is
local
and
for
the
remote
side
like
same
traffic,
for
example,
voice
traffic
between
CP,
1
and
CP
2.
They
can
go
through
transport
network
2
and
then
between
CP
3,
2
and
CP
2.
D
They
can
go
so
transport
network
1,
so
they
can
go
different
network,
and
so
the
encryption
can
have
an
option
of
end-to-end.
Cp
1
to
CP
3
can
have
intermediate
like
goes
to
the
CP
2
through
the
the
provider
network
and
then
from
CP
2
to
CP
3
through
the
public
and
that's
through
the
encrypted
traffic.
So
that
is
key
difference
between
SD
1
and
our
traditional
underlay
network,
and
also
a
little
bit
different
from
pure
overlay.
D
Second
part
is
IPSec
management.
Episode
has
been
running
over
20
plus
years
running
very
well
from
your
PC
from
your
phone.
So,
but
when
the
network
get
bigger,
forget
you
hundred
or
two
hundred
or
thousands
those
pair
wise
as
IPSec
key
management
becomes
the
issue
yeah
the
I
to
herself.
Yesterday
we
had
very
lengthy
discussion.
D
How
do
we
distribute
the
key
between
different
pairs
so
and
also
the
key
can
be
set
up
her
tenon
of
her
service,
but
those
fine
grant
IPSec
tonneau
can
be
increased
complexity
in
management
on
a
team
management,
because
not
only
you
need
to
distribute
the
key,
you
have
to
manage
the
rekey.
You
have
to
manage
frequency
difference,
so
it's
quite
complicated
so
so
for
the
large-scale
sd1
management,
the
IPSec
tunnel.
We
want
to
limit
it
to
the
transport
layer,
basically
only
through
the
public
network
on
trusted
network.
You
need
to
do
that
IPSec.
D
So
the
underlay
network,
you
don't
so
you
could
have
a
traffic
between
CP,
1
and
CP
3.
One
part
is
encrypted.
Another
part
is
not
ok,
so
this
picture
was
from
last
ITA,
one
or
two.
Just
to
recap:
hey
people
were
not
there.
It's
basically
from
the
CPE
based
VPN.
We
can
have
a
VPN
between
CPEs
and
that's
through
the
MPLS
network
and
the
same
CPE
can
also
have
a
pass
through
the
public
internet.
So
that
means
them.
D
The
BGP
can
control
for
the
CPE
based
VPN
has
to
be
able
to
control
both
the
MPLS
Network
and
another
instance
to
manage
them
the
pass
through
them.
The
overlay
and
there's
another
case
of
you
have
to
see
bees.
They
are
interconnected
by
the
public
network.
You
said
of
IPSec
tunnel.
It
works
fine,
but
then,
when
this
CPE
can
zoom
get
changed
from
the
AWS
availability
zone,
one
to
two
or
two
different
region,
suddenly
the
performance
getting
much
much
worse.
D
So
with
that
case,
that's
when
the
drug
we
have
in
the
spring
working
group
to
show
how
do
we
steer
the
traffic
to
the
closest
PE,
so
the
packet
can
traverse
the
provider
secured
and
good
quality
network
as
soon
as
possible.
So
on
the
right,
we
added
another
figure
to
show
that
the
the
Popolo
between
Nakano
the
network
from
particular
CPE
to
another
CP
in
different
region.
They
have
different
application
attached
and
from
same
tenant.
Thus
some
application
can
go
through
one
tunnel
and
another
application
go
to
another.
D
D
We
have
to
tell
them
this
so
there's
some
some
gap
analysis.
We
have
this
tunnel
in
cap
in
idea.
That
definitely
is
a
very
good
solution
to
hop
along
to
to
do
the
route
distribution
associate
out.
Who
is
tunnels
and
with
that
draft
we
find
there's
still
something
lacking
like,
for
example,
the
end
point:
when
they
power
power
down
I,
see
peeping
power
down,
they
need
to
be
able
to
tell
the
route
reflector
or
the
local
properties
just
like
when
I
sum
and
I
NH
RP
is
doing
advertising
what
kind
of
episode
property
your
support.
D
What
kind
of
algorithm
you
support?
What
kind
of
duration
exchange
the
property
so
that
one
need
to
be
added
to
the
tunneling
cap
and
another
one?
Is
them
I
call
the
net
property
is
really
the
private
port,
so
the
CPE
in
the
sd1
case
could
have
dynamic
address
because
pop
up
being
a
shopping
mall
at
my
home,
so
I
could
need
to
tell
the
remote
and
how
to
establish
tunnel
to
me.
So
that's
I
called
in
that
property,
propagation.
D
And
this
draft
there's
a
rosin
bat,
secure
and
sweepy
and
draft
okay,
so
that
one
describes
that
we
have
a
Alice
tree,
bpn,
suddenly
I
put
a
PE
into
the
remote
in
and
but
not
wrapped
assume
the
P
in
the
remote.
They
call
CPE
is
pre-configured
with
IPSec
ahead
of
time
manually
and
then
can
carry
the
to
like
a
hotel
room.
So
everything
pre-configured
they
care
a
lot
about
pgp
two
instances
inter
working
with
each
other,
so
that
part
will
also
work
for
us.
D
But
in
terms
of
scalability
we
need
to
have
a
CPE
to
be
able
to
plug
in
and
play
so
in
Sdn
industry.
The
term
zero
touch
provisioning
is
a
must
feature,
must
requirement,
meaning
when
CPE
powered
up
in
a
shopping
mall
in
a
hotel
room,
it
has
to
be
able
to
connect
it
to
their
peers
automatically.
We
thought
anything
so
that
gap,
that's
the
specific
gap
and
also
the
number
of
CPS.
D
Okay,
so
ITF
is
a
protocols
right
so
here
the
draft
is
mainly
talked
about
the
problem
statement
and
gap
analysis
and
since
IETF
102,
we
put
together
drawers
for
solutions.
There'll
be
different
aspect
of
solutions,
as
we
see
in
the
earlier
right,
the
SD
one
is
really
hybrid
of
overlay
underlay,
some
port
interface
is
underlay.
Some
interface,
overlay
and
overlay
part
need
a
piece
a
tunnel
and
need
to
do.
D
The
key
exchange
need
to
manage
the
three
key
process
and
the
underlay
you
can
utilize
existing
and
can
network
so
for
the
robbery
route
advertisement
for
the
transport
network
establishment,
all
those
need
work,
so
their
internal
protocol
and
besides
the
first
track,
talked
about
data
model
rights.
As
about
service
model.
On
the
protocol
side,
there
are
already
two
drafts:
why
is
it
an
idea
and
once
in
the
bass
and
idea,
I
didn't
have
meaning
than
this
time?
So
I
want
to
just
give
a
brief
overview
of
that.
D
It's
mainly
to
stimulate
interest,
discussion
and
see
this
kind
of
implementation
have
any
flaws
or
stimulate
other
people
to
come
up
with
alternative
solutions,
because
today
in
SD,
one
industry
is
pretty
much
single
vendor
and
the
reason
for
that
is
everybody
claim,
there's
no
industry
standards.
So,
let's
just
forget
about
interoperability
that
has
created
lots
of
headache
paying
for
the
managed
service
provider
managed
service
provider,
not
necessary
network
service
provider.
D
D
My
port
number
three
may
be
LTE,
and
here
are
the
IPSec
property
and
support
from
for
number
one,
and
also
my
local
and
public,
address
mapping,
and
then
it
is
up
to
the
Rob
reflector
to
determine
where
to
forward
that
propagate
that
advertisement
to
so
in
traditional
IP
sack
like
Ikey
v2.
There
is
a
policy
database,
so
each
node
is
provision
with
who
you
can
talk
to
right.
What
traffic
is
to
be
encrypted?
What
traffic
is
now
to
be
encrypted?
D
And
here
the
sd1
is
slightly
different,
because
the
traffic
itself
can
be
one
moment
need
to
be
encrypted.
Another
moment
doesn't
and
this
policy
has
to
be
coming
down
from
the
controller
and
who
they
can
talk
to
they
can
the
the
configuration
of
the
policy
can
be.
It
can
be
propagated
to
the
CPE,
but
if
the
CP
is
a
low
cost
like
$1,000
switch
and
managing
those
policy
can
be
overwhelming,
especially
you
want
to
do
the
little
touch
provisioning
you
power
up,
the
CPE.
You
want
the
CPE
to
be
able
to
talk
to
his
peer.
D
They
have
to
leverage
the
controller
so
so,
basically
CP
1,
for
example,
advertise
his
property
here
and
I'm
Linda
I'm.
You
talk
to
my
peer
and
then
the
rock
reflector
has
the
configuration
ok
for
Linda
I
can
only
talk
to
Jeff,
but
nobody
else,
so
he
will
propagate
that
property
to
Jeff
and
them
and
then
for
the
10
and
2.
They
did
a
similar
thing.
D
So
that's
the
tunnel
advertisement
policy
being
determined
by
the
rock
reflector
and
here's
the
extension
for
the
tunnel
Inca.
So
one
thing
is
about
a
pea
sack,
so
the
episode
itself.
We
leverage
the
information
needed
but
I
key,
but
the
transport
route
is
different.
Instead
of
going
to
my
peer
who
I
need
to
talk
to
I
sent
to
my
controller
and
the
controller
will
help
to
send
it
to
the
tenon
there's.
Another
extension
needed
is
for
the
net
property,
so
as
a
CPE
I
can
request
from
a
local
stone
server
to
find
out.
D
Who
is
my
net
provider
and
what
kind
of
net
property
my
net
device
provide
or
and
then
be
able
to?
Tell
my
remote
in
I'm
a
private
address,
here's
my
net
property
and
here's,
my
public
interface
and
public
port,
so
those
TLV
are
needed
for
the
tunnel
in
cap
and
and
on
top
of
that,
because
sd1
itself,
it's
kind
of
hybrid
overlay
underlay.
Those
transport
network
need
to
be
established
prior
to
data
arrived
prior
to
associate
with
any
particular
service
or
raus.
So
the
idea
we
requested
a
new
Safi
is
really
for
to
advertise.
D
This
is
my
SD
one
and
here
my
different
type
of
transport
network,
and
so
that's
the
extension
we
have
done
so
next
step.
We
want
working
group
adoption
and
since
we
have
got
lots
of
feedback
and
main
reason,
main
reason
for
working
group
adoption
so
to
tell
the
world
IETF
has
this
work.
We
have
identified
the
problem
statement.
We
have
identified
a
gap
and
here
are
some
of
the
protocol
extension.
D
We
can
do
to
enable
interoperability
to
be
the
first
one
to
break
this
chicken-and-egg
problem
because
industry,
when
you
go
to
the
managed
service
provider,
then
they
would
tell
them
there's
nothing
to
about
interoperability.
Just
use
my
solution.
Nobody
else
can
plug
in
and
there's
no
industry
reference
on
that
either.
Oh
and
Yugi
is
trying
to
tap
on
that,
but
they
don't
have
the
expertise,
look
seriously.
They
don't
have
the
expertise,
they
talk
about
use
cases,
they
talk
about
requirement
and
they
talk
about
API.
D
G
You
yeah
this
is
you
know
your
comment
about
oh
nug,
so,
yes,
they
want
to
come
to
the
IETF
for
solutions,
and
a
lot
of
people
have
done
a
lot
of
hard
work
to
make
that
happen.
However,
if
the
IETF
comes
up
with
too
complex
of
salu,
they're
gonna
go
elsewhere
or
the
vendors
are
just
going
to
do
it
themselves,
so
you
have
to
be
able
to
articulate
extremely
clearly
to
these
users.
What
the
solution
is
gonna,
be
that's.
H
D
I
took
that
out
because
after
I
tape,
one
or
two
many
people
come
to
the
common
saying
that
has
been
obsoleted.
We
shouldn't
use
that
as
a
reference
and
with
that
I'm
taking
people's
opinion.
So
we
take
that
out
we're
using
the
tunneling
cap
and
we
want
to.
However,
we
do
want
to
be
able
to
further
for
the
robbery
firm
for
the
controller
to
be
aware
of
all
the
possible
transport
networks.
There
are
some
of
that
underlays
I'm,
not
overly
they're
different
from
tunnel
okay
Thomas
end
to
end.
H
So
I
encourage
you
to
look
at
that
and
if
it
meets
your
use
case,
use
it.
If
it
doesn't,
you
know
maybe
look
at
what
we
ended
up,
replacing
it
with,
which
was
just
the
way
everyone
effectively
used
it,
because
it's
a
lot
easier,
which
is
the
endcap
attribute
figure
out
how
to
extend
that
or
use
that
and
I
think
I.
D
Mean
so
glad
you
give
that
comment
I'm
so
glad
you
gave
that
comment,
I
being
puzzled
by
this
I'm
puzzled
by
this
and
ITM
can
create
the
IFC
and
a
few
days
later,
a
few
years
later
saying:
that's
not
valid
because
nobody's
using
it.
I
am
myself
I've
been
puzzled
by
that,
but
everybody's
telling
me
don't
fight.
H
We
updated
it
based
on
implementation
experience.
The
experience
was
is
that
we
could
use
it
in
two
different
ways
and
one
of
them
we
found
to
be
overly
complex
for
our
use
cases
and
given
them
that
was,
there
was
two
ways
and
one
was
overly
complex.
We
went
with
the
one
that
was
simpler
now
this
is
a
different
use
case.
Okay,.
D
So
for
the
paneling
for
them
5512,
I
want
to
separate
between
tunnels
and
the
transport
right.
So
in
st
one
we're
talking
about
transport,
the
interface,
the
transport,
the
tunnel
can
be
between
c
CP,
1
and
CP
3,
but
the
traversing
can
go
through
different
networks
so
that
part,
even
5512,
is
not
coming
that
so
that
can
be
extended.
We
can
talk
about
that
offline
they'll
be
great.
If
you
can
join
the
work
and.
H
I
D
Are
the
ones
I
wrote
in
my
draft
and
I'm
looking
forward
to
more
feedbacks,
more
gaps?
Actually
in
the
drive
itself?
I
probably
didn't
describe
very
well
and
has
some
course
of
comments
that
this
particular
part
wasn't
described
very
well.
So
that's
something
we
you
up
to
the
working
group
draft
and
they'll
get
more
feedback
on
them.
I.
I
As
an
SD
one,
so
you
actually,
there
is
way
more
kept
than
what
you
actually
have
identified,
which
is
I.
Think
I
do
agree,
so
wait
for
IPSec
that
could
be
some
soup
I.
It
would
be
good
to
help
support
in
setting
up
the
full
mesh
and
stuff
like
that.
I
think
we
don't
necessarily
have
to
reinvent
Ike.
In
my
personal
view,
I
think
it's
good
that
PGP
could
help
in
making
available,
which
endpoints
needs
to
be
set
up
with
a
full
measure
with
open
spoke
type
of
topology.
I
We
don't
have
to
reinvent
I,
ignore
celli
to
do
that,
because
you
can
assist
basically
the
IPSec
process
to
say,
okay,
to
which
end
point
you
need
to
address
and
then
still
use
Ike
to
do
the
IPSec
establishment,
because
then
you
get
all
the
rekeying
and
everything
available
from
those
protocols
for
a
while.
I
do
believe
that
for
knots,
that
could
be
some
extensions,
which
are
useful,
I'm,
not
sure
whether
defining
a
new
of
Tuffy
is
actually
the
right
way
forward,
but
that
will
be
an.
I
But
I
think
the
I,
where
there
is
definitely
another
area
which
I
potentially
we
have
to
focus
on,
but
it
will
be
a
very
challenging
one,
which
is
everything
to
do
with
load.
Balancing
some
people
use
packet
based
hashing.
Some
people
use
flow
based,
hashing.
They
use
latency
measurements
to
actually
figure
out
what
is
going
on
and
how
to
optimally
use
the
available
transports
which
are
available
to
that
and
that's
another
huge
area
which
is
not
such
so
easy,
I
think
to
to
address.
Okay.
D
Thank
you
for
the
icon.
We
had
that
more
than
our
discussion
in
iTunes
and
staff
working
group
yesterday,
so
for
the
AG
itself,
the
content,
we
don't
change,
but
how
do
we
set
it
up
so
in
iTunes,
except
we
end
any
by
three
cases.
The
first
case
is
controlled
or
facilitated
configuration
because
for
Ike
has
lots
of
configuration
right
policies
and
SP
des
ad
and
SPD.
D
Second
is
a
second
case
which
has
long
controversial
is
actually
controller
for
for
the
CPE,
which
is
low,
cost
low
power
CPE.
They
don't
want
to
deal
with
communicating
with
so
many
peers,
so
they
depend
on
the
controller
to
help
them
to
do
them
a
key
propagation.
Now
one
created
lots
of
issues
and
third
case
and
Davies
on
the
in
the
room
as
well.
F
Patel
arcus
I
just
wanted
to
remind
you
basic.
We
you
and
I
had
conversation,
but
was
just
updating
you.
It
is
his
RFC
5566
that
we
are
updating
and
ensuring
that
it
goes
over
tunneling
caps
as
an
author
of
a
tunneling
caps,
I'm
updating,
both
irises,
so
I
just
wanted
to.
Let
you
know
that
work
is
going
on
happy
to
work
with
you
and
cover
this
okay.
D
B
To
and
for
you
nevertheless,
so
I
think
when
you
have
BGP,
when
you
have
a
hammer,
everything
looks
like
a
nail
right:
we
really
need
to
find
right
boundaries
between
management,
plain
interruptions
versus
control
interruptions.
You
shouldn't
be
doing
everything
bgp
just
because
which
appeals
are
ridiculous
for
the
tracks.
Okay,.
D
So
that
that's
another
issue
really
for
us.
We
want
something
simple
right
in
our
deployment
today,
we
first
release
was
using
a
HR,
p,
TS
VPN.
We
deploy
that
for
sd1,
but
we
start
a
fun
problem
when
the
number
of
nodes
start
to
increase
more
than
100
is
that
choking
doesn't
perform
well
and
since
bgp
everybody
here
right,
it's
the
white
lady
point:
everybody
has
it
even
go
to
an
AWS,
they
don't
implement
any
protocol,
they
support
bgp.
So
because
a
widespread
just
like
ipv4
right,
everybody
saying
doing
improve
ipv4.
D
D
E
D
F
Patel
arcus
again,
if
you're
looking
for
the
solution
outside
BGP,
there
was
a
draft
which
expired
in
caught
working
group.
That
was
a
generic
draft
that
allowed
two
endpoints
to
exchange
King
materials.
Maybe
you
can
use
that,
but
you
know
they
say,
and
you
kind
when
you
can't
fight
them,
you
have
to
embrace
them
or
join
them.
G
D
With
all
key
okay,
like
SSL
right
I
can
reach
my
I
clicked
it's
not.
Okay.
We
can
talk
offline
on
that,
okay,
Elizabeth,
okay,
okay,
I
said
this
was
proposed
as
another
alternative
solution:
robert
Rosic.
He
said
why
don't
you
should
list
I
said?
Well,
maybe
you
should
have
a
drag
on
this.
How
do
you
do
this?
Well,
oh,
oh.
D
E
So
so,
we'll
we'll
cut
the
specific
discussion
about
the
draft
right
now,
unless
you
had
another
comment
on
on
this
aspect,
how
many
of
you
have
read
the
draft.
B
E
Thank
you
max
yeah,
so
well,
let's
continue
that
discussion
a
little
bit
since
we
have
a
pretty
open
agenda
today
you
can
go
and
yeah
I
wanted
to
phrase
it
also.
So
this
topic
seems
within
scope
of
the
the
broad
scope
of
RT
gwg,
especially
these
two
documents
are
informational.
It
seems
like
based
on
the
feedback
that
we're
getting
at
least
it's
the
right
venue
right
that
people
are
interested
in
this
topic.
E
J
7701
old
Eastlake,
with
Huawei
Technologies.
This
should
be
a
pretty
quick
presentation,
just
wanted
to
make
people
aware
of
something
which
was
specified
in
RFC
83
77,
so
I
usually
use
very
simple
slides,
but
I
was
getting
a
little
tired
of
all
this
horizontal
text.
So
we
have
some
angled
text
in
this
one.
J
So
I
assume
people
understand
multi
topology.
The
idea
is
to
specify
a
really
a
physical
subset
of
the
resources,
so
the
links
and
nodes
are
marked
as
to
what
topologies
they
can
handle,
and
that
implies
also
that
the
the
packets
here
trying
to
route
are
datura
are
classified
into
those
topologies
and
so
a
router
when
it
receives
a
packet
has
to
be
able
to
figure
out
which
topology
it's
in.
J
You
might
want
to
be
able
to
explicitly
labeled
packets
as
to
what
the
quality
they're
in
so
RC
83
77
specifies
a
way
to
do
that
using
an
ether
net
tag.
So
there
has
been
a
either
type
assigned
for
this,
because
it's
nine
a
twenty
two
more
hard.
We
see
and
they
the
initial
specification
with
this
there's
a
version
and
some
reserved
bits
and
then
twelve
bits
for
the
multi
topology
ID
using
the
ID
space
from
the
is
is
multi
topology
RFC,
that's
probably
fine
for
is,
is
a
no
SPF.
J
The
OSPF
has
I
believe
the
smaller
multi
a
topology
ID,
but
you
can
just
leave
the
upper
bits
zero
for
that.
If
somebody
wanted
to
specify
some
use
for
those
reserved
bits,
they
could
I.
Guess
the
RZ
83
77
explicitly
says
that
this
tag
can
be
used
for
other
protocols.
It
was
defined
in
connection
with
trail,
but
it's
available
for
people
to
use
and
if,
for
some
reason
you
need
something
even
more
different,
you
could
use
the
version
field
which
is
zero
in
as
defined
in
83
77
and
have
some
other
value.
K
K
Done,
okay,
so
here
why
we
here?
So
this
is
the
idea
to
in
Hmong
shield
as
a
we
propose.
The
one
protocol
uses
a
net
for
the
network
monitoring
protocol.
Here
we
propose
the
the
prod
code
who
monitor
ipv4
network
or
troubleshooting.
After
the
submission
of
this
structure,
we
received
a
lot
of
comments
and
have
a
lot
of
discussion
through
the
mailing
list
and
also
in
the
and
also
in
the
presentation.
You
know
idea
well
to
include
working
group
and
the
obvious
Authority
working
group
so
based
on
this,
the
feedback
you
know
after
idea.
K
For
so
we
go
on
to
identify
the
possible
user
cases
and
also
the
to
the
cap
analysis
on
the
existing
horse
and
inter
investigated,
the
existing
controlling
elementary
tours
and
the
messers.
This
and
after
that,
we
the
possible,
therefore
any
Walker
for
the
control
plane
elementary
for
Network
protocol
monitoring.
This
is
for
the
user.
Most
of
the
other
cases
is
for
the
network
troubleshooting
and
the
network
planning.
K
We
are
probably
suffering
walk
instead
of
directly
apropos
the
defining
the
protocol.
So
this
in
order
to
collect
the
feedback
and
they
have
to
say
if
we
can
reach
some
agreement
or
some
consensus,
so
IETF
four
one:
three:
we
organize
a
meeting
to
discuss
the
problems
used
cases
and
the
requirements
that
usually
interest
and
potential
solutions.
So
now
that
they
see
the
for
the
summary
of
this,
that
said,
the
meeting
okay.
K
They
said
a
medium,
we
is
honors
who
invited
many
routine
experts
and
the
operation
experts,
and
so
that's
almost
near
twenty
attendees
for
the
center
median,
and
also
a
special
thanks
to
the
individual
for
feedback
from
open,
wah-pah
Paulo,
a
CHF
who
can
not
get
the
time
to
for
the
said
meeting.
Okay,
so
I
think
after
the
set
a
meeting
from
my
pond
wheel,
we
receive
the
extensive
feedback
on
this
walk:
okay,
Sankyo
time
and
the
valuable
feedback
to
help
on
this
walk.
K
That's
the
you
know,
past
years,
that
manager,
clean,
a
lot
of
a
young
model
is
being
defined
under
the
extension
of
the
Netcom
or
the
GRP.
C
is
a
being
proposed
that
in
the
industry,
user
has
been
well
accepted
and
for
the
data
plane,
this
is
at
elementary,
so
the
IPP
I'm
working
group.
So
the
impact
you
see
to
OEM
Walker
use
a
pimp
riposte,
so
we're
seeing
that's
used
for
the
data
print
elementary
is
Anka
we're
so
now
we
see
it
uses
at
for
the
control
pony
for
in
order
for
the
network
monitoring.
K
K
Okay,
so
this
problems
so
I
think
this
saying
we
are
very
familiar
with
our
IP
network.
So
this
is
the
father
network,
a
troubleshooting.
So
that's
a
lot
of
what
these
issues
use
the
cost
by
the
failure
of
the
routing
protocol.
So
this
reason
now
that's
a
very
I
use,
a
take
a
long
time
for
some
that
you
import
for
some
of
these,
of
the
usual
network
of
failures
may
take
a
long
time
and
the
master,
the
very
its
appearance,
the
engineer
to
locate
the
route
course.
So
this
is
a
very
this.
K
Is
you
the
one
cause
for
this
walk
and
another
user
network
of
planning?
So
now,
that's
you
the
further
5et
or
say
some,
these
new
applications.
They
needed
the
real
time
on
network
up
Ronnie,
so
these
are
needed
to
collect
the
networker
space
in
the
live
and
network
under
to
the
network
planning
in
a
real
time.
So
they
say
is
the
cause
for
this
walk.
K
This
are
we
identified
a
possible
use
the
pieces?
So
we
you?
This
is
the
typical
cases.
That's
you
know
harder
to
be
located
for
the
network
of
troubleshooting
this,
including
the
Rooter
loop
and
the
route
flagging,
and
also
the
RSVP
synchronization
failure,
and
also
some
of
these
Kitano
failure.
So
this
is
the
so
this
is
the
user
cases.
There
are
general
the
characteristics.
So
that's.
This
is
a
network.
K
Why
the
failure,
instead
of
a
specific
device,
a
failure,
because
this
is
a
failure
they
mastered,
who
comparing
the
Peter
between
devices
among
the
internet,
work
instead
of
a
directory
log,
one
devices
to
located
the
good
course.
So
this
is
the
typical
case
for
the
networker
troubleshooting
under
then
they
say
your
network
planning
cases
this,
including
the
root
of
Haas,
Automation
and
also
the
root
of
policy
value
Asia.
K
So
this
means
we
collect
the
networker
Rio
date
her
for
the
simulator,
so
that
the
user
can
validated
the
new
through
the
policy
to
see
if
it
doesn't
work
and
say
what
a
effect
on
the
existing
network.
So
this
is
the
user
cases.
Okay,
are
this
your
the
cases
that
we
have
tracked
the
general
requirements
so
the
first
day
they
say
either?
We
need
a
tunnel
further
controlling
data
export
as
first
so
these
we
need
a
rally.
K
The
way
to
explore
the
require
control
playing,
theater
and
also
the
exporter
performance
muster
satisfy
the
requirement
of
the
of
the
collection.
So
this
is
the
first
one.
A
second
one
is
adequate.
The
protocol
data
collection-
in
fact
they
are
already
some
existing
torso
to
collect
the
protocol
data
such
as
a
Wii,
U
assemblies,
the
net
confer
or
us
a
meaty
RPC
to
collect
the
protocol,
running
state
hers
and
also
the
operation
data,
and
also
we
had
PDP
Lingus
data
to
collect
the
network
at
holiday.
So
this
is
can
be
used
for
the
protocol.
K
K
This
is
the
messages
of
the
routing
protocols
and
also
we
need
a
sound
reason
network,
a
wider,
the
ribs,
ribs
and,
and
then
this
is
the
policy,
and
also
this
is
a
correlate
here.
The
record
of
a
policy
and
the
route
so
means
we
not
only
to
exported
the
route,
but
we
also
a
proper
this
of
the
the
route
experience.
What's
the
process
has
been
defined
by
the
policies,
so
the
correlation
should
be
exported
so
that
if
we
get
this
information,
so
we
found
that
this
failure-
fear
of
happens.
K
We
can
know
what
a
policy
called
due
to
failure.
So
this
is
the
correlation.
We
think
the
important
data
should
be
collected.
Okay
under
then
they
say
the
NADA
is
a
network
coverage,
so
they
say
you
the
means.
The
density
is
needed
to
collector
multiple
devices
for
the
duty
in
theater.
So
that's
a
you
said:
hopper
Network,
a
wider,
the
protocol
date
her
yeah
okay.
So
these
are
you
the
summary
of
the
Sena
meeting,
so
this
is.
This
is
a
summary
of
the
like
this
one.
The
use
pieces
identified
real
problem
in
the
lam
networks.
K
Yes,
so
that's
what
the
second
one,
the
of
rehearsed,
use
a
bare
reality
of
what
you
consistent
horse
to
extract
the
data
they
needed.
They
means,
especially
when
the
failure
happens.
They
will
try
all
possible
that
horse
to
act
right,
the
possible
data
for
the
network
of
troubleshooting
yeah,
not
only
the
not
only
the
UGA's
interest,
such
as
Netcom
for
Sarah
and
the
people
in
estate
yeah,
but
maybe
the
for
example,
even
though
the
procured
at
her
to
be
a
further
proko
data
to
be
exported,
the
hi
you
Assam.
K
This
is
the
mirror
that
means
replicated
the
promoted
her
and
it
will
collected
from
the
devices
to
the
network
analytics.
So
this
is
also
some
this
the
tours.
So
that's
a
user
thanks
for
the
attendees,
so
they
proposed
this.
The
modulation
horse
for
help
our
earth
understanding
the
UGA's
Asian
Network.
K
K
But
after
the
Sena
meeting,
the
discussion
is
saying
that
that's
the
maybe
the
PM
key
is
already
adopted
that
you
know
you
can
see
a
network,
so
they
say
the
reality.
So
these
are
you
the
only
father,
so
this
is
only
used
for
the
PDP
for
the
PDP.
If
we
go
on
to
collect
a
sample
at
8
her
for
the
IDP
for
the
proto,
whatever
they
are,
all
the
prefer
that
that
surimi
entire
country
is
a
top
theater
for
as
the
generic
solution.
K
So
this
is
the
consensus
and
sensors
and
another
greatest
user
because
of
the
the
network
of
operation
when
the
feeder
happens,
so
they
must
use
almost
all
possible
tools
to
collect
all
possible
data.
Saying
that
they
cannot
with
enough
time
for
the
standardization,
so
this
is
the
Sims,
then
his
standards
always
a
lag
behind
of
the
operational
applications,
but
in
order
for
accelerating
the
process
they
sing,
the
open
source
will
be
much
helpful,
so
they
say
the
Rafah
consensus
of
this
list
and
a
meeting
so
after
this
other
center
median.
K
So
we
think
that
as
a
possible
worker
to
do
the
next,
so
that's
the
more
user
cases
identified.
That's
a
you
know,
hard
whoo
detect
our
location
low,
allocate.
So
that's
a
user
base
on
this
one
to
identify
these
the
requirements
on
the
new
data
collection
for
for
the
different
user
cases
and
last.
Why
so
that
you
who
agree
under
stream
attainment
reuse
the
generic
generic
a
solution
for
the
future?
So
maybe
this
is
the
new
young
models.
K
Standardization
Walker
will
be
in
proposed
in
IETF
and
the
open
source.
Ok,
so
my
point
will
I
think
that's
the
we,
after
that
we
achieve
our
cost
to
collect
these.
The
extensive
opinions
of
the
actors
in
the
industry
I
also
would
like
to
say
that
today's
the
raffle
consensus,
if
you
have
any
comments,
I
also
I,
have
an
opinion
on
this.
One
I
would
like
to
collect
the
more
comments
based
on
the
yeah.
Ok,
take
this
chance.
E
So
I
think
one
of
the
the
main
consensus
that
isn't
directly
stated
here
is
that
of
this.
This
side
meeting
that
you
referred
to
as
that
Chris
Bowers
speaking,
is
that
we
don't
necessarily
need
more
protocols
to
do
this
so
that
the
name
of
this
project
that
you're
working
on
you
know,
network
monitoring
protocol
is
a
little
bit
lead
you
to
think
that
it's
a
foregone
conclusion
that
we
need
more
protocols
to
do
this
and
I
think
the
general
consensus.
Is
we
don't
need
more
protocols
to
do
this?
I
guess
you
refer
to.
E
So,
aside
from
that,
though,
I
was
thinking
that
a
you
know
the
the
network
operators
represented
in
in
the
IETF
and
in
in
that
discussion
are
you
know
the
the
operators
running
the
largest
most
complex
networks
in
the
world
and
there's
a
whole
class
of
other
operators
that
are
at
represented
at
the
IETF,
because
they
don't
have
the
resources
to
to
come
here,
but
it
might
actually
be
useful
work
and
I.
Don't
know
where
that
work
would
be
done
to
sort
of
define
for
your
use
cases.
E
What
what
actual
subset
of
data
do
you
need
to
troubleshoot
them
so
that
you
know?
Maybe
those
operators
could
take
that
I've
smelt,
a
template
or
whatever
you
want
to
call
it,
that
subset
of
data
and
and
at
least
request
that
their
vendors
provide
that
subset
of
data.
So
they
can
then
either
build
a
tool
or
buy
a
tool.
E
That
does
that,
because
I
think
you
know,
a
mid-sized
operator
looks
at
the
thousands
of
pages
of
young
models
and
has
no
idea
what
to
request,
support
for
and
they're
just
going
to
be
sort
of
at
the
mercy
of
what
they
get
so
that
that
might
be
sort
of
useful
work
to
help
address
these
use
cases
for
the
midsize
operators-
and
you
know
larger
operators
would
also
benefit
from
that
as
well.
So
that's
just
mine
I'll
be
back.
B
Just
answer
abstract,
so
the
fact
that
there
are
three
things:
data
transport,
testable
synchronous,
binary,
encoding,
JP,
Singh
young
push
is
there
data
consumptions
are
how
you
model
it
and
young
helps
us
data
correlation.
This
is
more
interesting
topic
and
they're
still
right
area
of
the
table.
How
do
you
relate
events,
so
this
is
interesting
space
to
work
on
there's
a
lot
of
stuff
out
there
that
you
didn't
mention
after
up
will
mention
some
of
them.
I,
don't
want
us
to
reinvent
the
wheel.
B
We
need
to
focus
on
what
needs
to
be
solved
rather
than
reiterating
what
has
already
been
solved.
So
you
mentioned
some
tools.
He
did
mention
some
others
I
think
we
need
to
look
wider
what's
available
in
the
industry.
What
open
configure
is
doing
has
been
done
in
that
month,
net
conf
and
actually
be
GPS
on
a
GPS
working
groups
where
we
define
data
model
that
also
provide
operational
stage,
an
allotrope
okay,.
L
Rob
here,
Google
I
have
some
experience
in
this
area,
as
Jeff
just
told
me
up
to
talk
about,
and
so
we've
been
working
on
this
for
a
bunch
of
time.
I
think
I
think
the
observation
that
we
don't
need.
Another
protocol
is
key.
Yeah
we've,
we've
kind
of
defined
some
transports.
As
Jeff
said,
we've
we've
been
working
on
that
as
G
RPC,
based
with
a
way
to
be
able
to
have
model
data
that
doesn't
need
to
be
modeled
in
yang
pushed
over
it.
L
So
that
I
think
that
this
moving
towards
a
single
solution
is,
is
the
right
thing
and
then
there's
two
bits
of
work
that
I
think
the
worth
commenting
on.
One
is:
what
set
of
data
do
we
need
I?
Think
it's
an
interesting
problem
to
go
and
say:
okay,
what
what
are
the
things
that
we
can
export?
What's
the
performance?
What's
a
performant
way
to
get
it
out
of
the
device?
What's
the
encoding
is
needed,
we've
done
a
bunch
of
work.
L
There
are
shipping
implementations
of
LSD
be
streaming,
for
example,
over
G,
RPC,
mova
genome
eyes,
specifically
so
we've
kind
of
started
to
solve
some
of
these
problems.
Spaces
where
they're,
not
traditional
monitoring
data,
sets
I,
think
you'll
find
the
most
most
large
implementations
or
you
know,
implementations
that
they're
widely
adopted
have
some
roadmap
to
getting
a
decent
amount
of
the
operational
state
that
we
current
currently
collect.
Already.
There's
to
me
a
problem
space
around
what
additional
data
don't
haven't
we
identified
I
think
some
of
that
was
was
what
the
point
you
were
raising.
L
Basically,
what
instrumentation
do
we
need
in
in
a
in
an
implementation
to
be
able
to
debug
it
I
would
encourage
a
conversation
with
operators
rather
than
a
specification
to
operators,
because
I
would
say
that
the
expertise
for
operating-
but
let's
say
a
box
is
plugged
together-
is
probably
not
in
this,
not
mainly
in
this
room.
Let's
go
and
talk
with
folks
and
figure
out
what
they
currently
do,
how
they
currently
debug
things,
because
this
isn't
you
know,
debugging
network
problems,
there's
kind
of
bread
and
butter
and
the
other
area
that
I
think
is
is
interesting.
L
Is
this
medium
operators
problem
right,
so
yeah
I
have
a
software
engineering
team
dime
part
of
the
we're
writing
implementations?
We
can.
We
have
a
bunch
of
resource
to
put
on
this
other
operators.
Maybe
don't
have
the
same
expertise.
The
challenge
there
is
that
I,
don't
think
you
know
blueprints
are
the
right
thing:
it's
it's
got
to
be
running
code
like
folk.
We
need
to
get
the
the
this
kind
of
modern
telemetry
data
at
the
stage
of
like
mrtg
is
or
I
can
download.
L
So,
if
we've
open-source
to
collect
so
we've
open
source
test
frameworks
around
the
data
from
the
device,
I
think
there's
then
the
next
layer
up
and
so
Yahoo
oath
recently
open
source
panop
sees,
which
is
their
kind
of
Nick's
layer
up
I,
think
we
should
concentrate
as
an
industry
of
getting
this
stack
kind
of
stood
up
and
show
how
easy
that
how
to
adopt
this
stuff
for
those
operators,
that's
the
thing
that
will
make
things
go
quicker.
It's
not
necessarily
standards
documents
or
deployment
documents.
L
B
B
K
B
N
O
N
Ok,
ok!
So
when
you
are
looking
at
the
network
management,
you
need
essentially
a
transport.
You
need
a
essentially
a
data,
modeling
language,
ideally
that
you
can
use
for
representing
the
schemas
that
you
want
to
use
as
well.
As
you
know,
you
need
some
central
database
and
there
has
been
on
and
off
in
the
past
and
well.
A
10-12
years
have
been
different,
open
source.
N
Now
I
need
the
sunglasses,
you
know
from
Rob
Chuck
here
so
yeah
so
and
there
has
been
on
and
off
the
open
source
of
track.
You
know
tries
that
come
out,
but
finally,
there
is
a
solid,
open
source
implementation
out
available,
and
this
has
been
you
know,
driven
by
in
a
group
from
Czech
Republic
around
in
Alexis
net.
N
They
have
developed
lib
net
conf
and
Lib
net
come
to
they
are
you
know,
right
now
we
can
say
they
are
mature
implementations
that
you
are
in
you're,
getting
a
library
that
you
need
to
build
a
no
client
and
the
server,
so
the
net
appear
gars,
no
guy
started
it
and
they
started.
In
said
in
two
thousand
six
or
seven,
and
since
then
two
implementations,
you
know
came
out,
one
of
them
is
lib
net
comp
and
the
other
one
is
the
lib
not
come
to.
N
N
If
you,
if
you,
if
you
would
have
used
both,
you
would
see
a
significant
difference.
You
know
in
the
performance
as
well
as
the
resources
that
are
needed
in
order
to
run
it.
You
know
in
your
application,
so
here's
a
quick,
you
know
overview
to
see
what
is
supported
and
it's
a
I
said
it's
a
it's,
a
mature
implementation.
It
is
well
used,
we
have
tested
it
in.
You
know
in
different
environments
against
different
vendors,
and
it
works
quite
well,
and
the
community
has
said,
is
you
know
actively
working
on
that?
So
the
next
one?
N
What
you
need
is
a
data
modeling
language,
because
you
don't
want
to
code
directly
and
represent
your
essentially
schema
in
the
code.
So
with
the
yank,
you
can
essentially
compile
the
schema
that
you
need
and
you
can
update
it
much
quicker
in
order
to
advance
your
project
and
Lee
being
is
a
very
useful
library,
especially
if
you're
planning
to
implement
some
of
the
existing
models
that
are
available
out
there
being
there,
either
from
the
ITF
or
from
other
organizations
or
open
source
projects.
The
library
is
again,
you
will
see
on
the
next
page.
N
Is
a
it's
a
flow
mature
and
it
allows
you
to
load
and
validate
the
young
schemas
and
data
into
your
demon
that
you?
You
know
that
where
you
want
to
use
it,
it
supports,
you
know
Yang
and
Yin
format,
and
you
can
do
the
XML
and
JSON
representation,
as
you
can
see
here.
As
you
know,
as
you
can
see
here,
most
of
the
RFC's
that
are
well
used
and
that
you
need
today
are
supported
and
the
you
know
the
parsing
of
the
schemas
is
based
on
our
experience.
N
L
To
the
previous
buttons
so
sure
I'm
Rob
here
from
Google,
and
you
should
be
tied
up
my
voice
by
now
and
the
so.
This
is
useful.
If
you
want
to
output,
see
right,
there's
not.
We've
done
some
other
work
in
this
area,
just
to
add
to
the
to
the
collective
open
source
and
this
area.
We
have
a
library
that
does
from
yang
to
Python
and
class
bindings,
we'll
do
a
similar
set
of
functionality.
L
Here
we
also
have
written
a
Varela
to
be
fully
featured
yang
to
go,
go
strut,
library,
that's
also
open
source
and
open
config.
It's
called
wycott
that
library
also
does
yang
to
protobuf,
which
then
gives
you
a
bunch
of
other
language
bindings,
think
it
doesn't
have
the
same
validation
but
at
least
lets
you
start
dealing
with
the
schema
and
we've
been
using
these
with
various
other
implementations
that
run
again
and
brought
a
buff
to
young
yang
to
proto
and
no,
not
the
other
around.
L
N
So
this
is
so
once
you
have
the
the
transport
to
communicate
to
your
application
as
well.
You
have
the
language
to
represent
the
application.
You
need
a
centralized
management
data
store,
and
this
is
one
area
where
many
vendors
they
have
their
internal
management
demons.
That
are,
you
know,
connecting
the
internal
demons
to
the
external
world,
and
this
is
where
you
can
use
sis
repo,
sis
repo
is
a
again
mature
and
we
have
the
representative
in-house
here
for
the
SIS
repo.
It's
an
open-source
one.
N
There
is
an
open
binary
to
be
used
that
I
will
be
mentioning
later,
but
with
the
SIS
repo
you're,
getting
the
full
young
support
that
you
need,
and
essentially
you
have
the
ability
to
set
to
store
and
retrieve
your
young
model
data
from
all
your
different.
You
know
demons
that
are
using
it
and
you
can
use
it
essentially
as
a
centralized
configuration
store.
N
You
know
it's
just
provides
you.
The
regular
management
features
that
many
people
take
for
granted,
but
you
don't
have
to
do
a
lot
of
implementation.
You
can
take
it
and
be
pretty
quick
to
a
start
over.
You
can
have
you
know
the
startup
running
and
the
candidate
data
store
support.
I
was
trying
to
find
out
the
plants
on
the
nmda
implementation,
so.
P
Michael
Abraham's
yeah
we're
looking
into
that.
One
and
I
also
want
to
add
I,
don't
know
if
you're
saying
this,
but
this
is
says
net
that
also
does
Liv,
Yang
and
net
appear
is
also
involved
in
this
project.
So
we
consider
that,
like
a
sweet,
so
yes,
history,
purse,
a
configuration
store
and
that
appears
in
that
conserver
go
together.
If
they're
publicly
very
well
tutorial
how
to
bring.
P
N
And
again,
you
know:
sister
epub
provides
the
data,
consistency
and
constraint
enforcement.
According
you
know,
to
the
young
models,
which
is
you
know,
an
important
function
and
it
allows
you
know
the
validation
of
the
of
your
high
level
configuration
across
you
know
the
system,
so
there
are
some
nice.
You
know
implementations
about
the
details
of
the
synth
of
the
sis,
repo
and
one
other
thing
which
I
like
to
point
out
at
the
beginning.
It's
a
no
single
point
of
failure.
Design.
N
You
can
really
know
have
it,
you
know
you
can
restart
it
and
you
can
just
continue
where
you
left
off.
If
your
transaction
was
completed,
you
know
from
the
previous
time
and
the
also
it
has
the
full
transaction
and
concurrency
support.
You
can
find
out
more
about
it,
but
you
know
it
has
support
for
several
languages
and
the
it
also
supports
which
the
Nakamoto
maqam
is.
If
you
want
to
enable
multi
administrative
multi-user
system,
this
is
a
one
area
where
it
comes
in
useful.
N
Is
that,
instead
of
writing
a
bunch
of
code
to
represent
your
schema
from
your
demon
and
then
every
time
when
you
add
some
new
functions,
you
have
to
write.
You
know
this.
The
code
in
order
to
generate
the
schema,
the
young
gives
you
a
higher-level
language
that
can
be
represented
and
then
translated
into
the
code.
This
will
increase
your
development
speed.
Now
you
don't
have
to
use
yank
per
se.
N
There
are
other
data
modeling
languages
that
have
that
are
out
there
to
be
used
and,
to
be
honest
in
the
FRR
we
are
using
yang,
but
we
were
also
using
you
know
the
Google
products
at
etc.
In
order
to
provide
you
know
the
needed,
a
variety
and
with
this,
but
you
need
a
management
entity
and
a
central
data
store.
N
This
is
a
tremendous
input
to
the
community
in
order
to
continue
developing
the
code,
because
it
gives
you
know
the
corrective
actions
from
whether
from
a
wider
perspective,
you
also
need
a
protocol
to
communicate
via
the
external
systems.
Web
Netcom,
you
know
is,
it
is,
as
I
said,
is
a
stable
one.
It
gets
connected
to
multiple
existing
commercial
and
open
source
implementations
as
well.
You
know,
then
you
can
export
your
data
model
pretty
easily
and
let
and
make
the
integration
in
with
the
other
systems
on
the
network
much
more
easier
before.
N
N
N
So,
oh
sorry,
so
in
the
example
of
the
free
range
for
routing
how
the
demons
were
managed
before
there
was
a
CLI
vty
shell,
which
was
a
centralized
interface
into
the
demons
that
were
part
of
the
F
arc.
So
you
would
have
a
central
place,
but
you
were
communicating
from
there
to
each
demon
independently
and
each
demon
had
its
own
schema.
N
The
putting
service
level
features
service
level
applications
on
top
of
that
was
hard
to
do,
because
there
was
some
cross
daemon
verification
missing,
and
it
wasn't.
It
wasn't
an
easy
way
to
be
done.
We
realized
that
it
has
to
change,
and
the
project
has
been
started
to
create
a
new
northbound
architecture
and
the
FRR
team
decided
to
do
it
in
a
few
ways
to
show
the
different
possibilities
and
how
can
be
used?
It
was,
you
know,
some
open
source,
as
well
as
some
open,
binaries
or
free
binaries,
and
show
how
this
can
be
integrated.
N
So,
in
this
case
another
tation
layer,
well,
it
had
to
be
created
between
the
label
yank
and
the
existing
and
in
the
existing
management
layer,
because
that
was
needed
in
order
to
be
able
that's
an
in-between
SAP,
but
some
of
that
other
patient
layers
that,
if
you're,
if
you're
signing
from
scratch,
you
don't
have
to
worry
about
the
adaptation
layer.
But
if
you
have
an
existing
system,
you
might
have
to
look
into
it
and
see.
How
can
you
create
that?
Essentially,
translation
from
your
existing
schema
into
the
yang
schema
and
then
export
it?
N
N
But
the
idea
there
is
to
show
also
how
you
can
use
a
new
management
daemon
like
this
repo
in
in
essentially
connected
into
you
know
into
that,
and
for
that
the
Google
protocol
buffers
were
used
in
order
to
connect
and
a
Google,
protobuf
or
interface
was
created
for
each
I'm,
not
sure.
If
for
every
interface
is
already
created,
but
you
know
there
are
bindings
in
place
and
available
in
order
to
do
that,
the
communication,
through
the
internal
systems,
between
your
management
and
the
demons
in
place.
N
Now,
you
have
to
use
a
transfer
protocol
to
communicate
the
internal
essentially
to
communicate
from
the
internal
demons
to
the
external
work
world
through
that,
and
this
is,
for
example,
in
the
sis
repo.
You
would
use
something
like
lip
net
Kampf
and
in
the
demons.
You
would
do
something
like
lib
yank
and
in
the
next
slide,
you
are
seeing
essentially
the
internal
view
of
an
internal
demon
and
how
the
implementation
was
done.
N
So
the
idea
there
is
that
you
will
have
a
young
module
for
the
for
the
demon
that
provides
you
certain
functionality
and
those
functionalities
are
being
described
in
the
model
and
you
are
then
using
some
kind
of
and
other
in
this
case
IPC
mechanism
in
order
to
communicate
to
the
management
demon.
Now,
if
you
want
to
add
some
new
functionalities,
the
update
of
a
data
model
is
much
more
simpler
because
you
get
the
code
automatically
generated
and
is
being
then
update
it
and
is
being
tied
into
the
new
functionalities
that
are
being
done.
N
It
also
has
a
support
for
rollback.
So
if
you
are
interested
in,
you
know
making
sure
that
once
your
transaction
is
finished,
then
you
can
retrieve
it
in
the
future,
because
some
of
your
future
transactions
might
receive
an
error
might
end
up
in
error,
and
you
want
to
make
sure
that
you
can
go
back
to
the
working
State.
There
is
a
option
to
connect
to
a
volatile
memory
or
a
persistent
memory
that
will
allow
allow
you
to
do
that.
N
So
I
said
there
are
some
well-known
tools
out
there
that
you
know
are
ready
to
be
used
and
it
would
be
good
to
hear
back
from
you
know
the
community.
What
what
is
your
opinion
on
that?
How
useful
they
are
I've,
seen
here
some
people
that
are
using
them
that
are
involved
with
that.
But
you
know
getting
some
more
information
on
on
the
tools
would
be
helpful.
You
know
for
the
future
direction.
A
N
Open
source
is
viewed
as
a
Kickstarter
for
the
project
and
there
is
a
quite
healthy
community
that
is
interested
that
has
a
common
interest
in
maintaining
this
project.
This
is
not
a
ready
product
that,
if
this
is
these,
are
and
there's
a
difference
between
a
system,
product
and
building
blocks
that
you
can
use
in
your
product
and
if
you're,
using
this
to
build
your
product.
N
We
will
then
expect
that
you
will
contribute
in
one
way
or
the
other
back
to
the
community,
because
you're
using
that
to
build
your
own
product-
and
this
is
where
the
open
source
is
valuable,
to
kick-start
certain
things
to
be
used
by
other
people
and
then
through
their
commercial
success.
They
would
give
back
to
the
open
source
community,
because
it's
part
of
the
overall
solution,
I
know.
N
And
the
question
is:
is
there
a
critical
mass
of
people
and
companies
that
have
joint
interest
in
supporting
something
like
this
and
based
on
the
few
last
years?
There
is
increasing
interest
and
support
for
that.
So
this
is
just
to
raise
the
awareness
in
how
it's
been
done
and
how
his
system
been
built.
Is
it
the
system
for
production
deployment?
No,
it's
an
example,
but
can
they
use
those
libraries
to
build
a
production
ready
system?
Yes,
I
can
okay.
H
Lou
burger,
one
thing
that
Dan
didn't
say
is:
is
that
for
most
of
what
was
talked
about
here?
Is
there
up
on
github
you?
Can
anyone
can
go
fork?
The
repo
all
the
projects
take
pull
requests
I'm
a
maintainer
on
FRR.
We
take
pull
requests
from
anyone
all
the
time
we
when
in
when
FRR
was
doing
this
particular
work.
I
found
a
bunch
of
problems
with
live
yang
and
the
person
doing
the
work,
sent,
pull
requests
and
those
folks
accepted
it.
So
it's
sort
of
the
normal.
P
Michael
Abraham's
here,
yes,
so
we're
good
torchy
telecom.
So
we
we
use
this
for
managing
both
services
and
the
home
gateway,
for
instance,
that
I'm
responsible,
for
there
are
other
vendors
that
are
using
this
in
shipping
products
as
far
as
I've
been
told
because
they
needed
to
make
their
device
manageable
using
metal.
So
they
took
this
and
implement
that
all
their
management
needs
in
it
and
are
now
you
know
they're
at,
but
they
were
doing
their
or
their
own
support
for
it
and
it's
not
as
much
difference
from
Cisco
or
anyone
else.
P
You
know
the
taking
taking
a
bunch
of
tools
and
putting
it
in
their
products
and
then
they
sell
support
towards
the
the
end
customer.
There
are
part
of
people
involving
development.
This
that
are
consultants
that
you
can
probably
I,
can
direct
you
to
them,
and
if
you
want
to
get
new
features
implemented
and
so
on,
they
they
can
help
you
with
that.
So
yeah
I
don't
think
we're
developing
a
healthy
community
around
all
these
different
projects
that
are
tied
together
in
creating
something
that
is
useful
as
a
whole.
P
E
I
want
to
sort
of
maybe
clarify
something
both
for
myself
for
and
for
others,
or
get
you
to
clarify.
The
first
part
of
your
presentation
talked
about
using
open-source
as
sort
of
to
do
Network
management's
as
a
network
operator.
This
is
more
how
you
used
open
source
in
FRR,
the
internals
of
a
router,
basically,
okay,
so
just
there's
a
distinction
there
that
you
know
probably
the
right.
N
So
in
the
beginning,
I
said:
I
will
give
you
overview
of
the
open-source
tools
that
you
can
use
in
building
a
system
and
then
that's
used
FRR
as
an
example
how
to
use
them
because
I
could
have
just
left
it.
Here's
the
overview
of
them
fine.
But
this
gives
you
an
additional
way
how
they
were
used
in
an
open-source
implementation,
and
you
can
go
in
there
and
take
a
look
by
yourself
how
it
was
exactly
implemented
right.
E
P
Okay,
so
so
we
were,
for
instance,
in
the
process
of
packaging,
up
the
plug-ins
that
you
use
with
history,
but
to
implement
the
ITF
interfaces,
an
ITF,
IP
and
ITF
systems
model.
So
we
use
those
standardized
models.
It
talks
to
the
kernel,
it
talks
to
the
open,
wrt
configuration
subsystem,
and
then
you
can
configure
this
open,
wrt
box.
You
can
also
take
this
repo
in
the
same
plugin
and
for
the
stuff
that
it
talks
to
the
kernel
net
link
interface.
P
You
can
run
this
on
a
server
as
well
and
you
can
change
the
IP
address
of
the
server
interface
or
whatever.
So
there
you
didn't
mention
much
about
the
plugins
that
you
also
need
here.
But
this
is
what
implements
the
model
you
load
the
model,
you
load
the
plug-in,
and
now
you
can
use
the
model
that
then
runs
the
code
in
the
plug-in
that
actually
does
something
there.
N
There
are
guidelines
how
to
develop
your
own
plugins
and
yes
on
the
previous
slide.
If
you
can
just
go
back,
please
on
the
previous
slide.
I
forgot
to
mention
that
each
one
of
this
daemons
has
its
own
plugin.
That
is
being
used
to
connect
to
a
different,
centralized
management
daemon,
and
if
you
want
to
use
a
another
one,
you
can
develop
that
plugin
buyer
by
yourself
and
just
use
it
there.
B
K
Okay,
in
fact,
that
I
go
beyond
from
hobby.
In
fact,
I
have
seen
this.
We
use
a
very
useful
for
hearse
yeah
butter,
because
I
have
a
one
since
I
wanted
to
clarify
my
experiment,
this
one,
because
my
pointer
I
know
that
the
open
daylight
and
the
honors,
and
also
the
open
configure
they
also
had
some
these
the
young
horse.
K
So
here
the
this
mentioned
this
horse
because
of
the
high,
because
you
introduce
this
horse
because
you
can
satisfy
the
requirement
who
to
some
this,
the
worker
plug-in
in
the
devices
because
of
this,
but
because
my
my
Anna,
because
I
think
other
open
sources,
Lika
also
supported
the
young
Walker,
but
you
introduce
these
tours
I'm
used
because
of
this
open
source
at
work
and
used
for
this
purpose
to
implement
a
plugin
in
the
devices,
but
other
tours
cannot
achieve
this
purpose.
Anderson
is
right
or
not,
I
mean
you
mentioned
open
daylight
yeah,.
N
Is
a
controller,
and
in
this
case
it's
a
different
layer.
These
are
your
tools
that
your
application
can
control
to
the
controller
and
be
able
to
export
the
description
of
the
functionalities
that
you
want,
your
controller
to
control
in
that
device.
So,
if
you're
writing
some
new
application
and
you
want
that
application
to
be
managed
by
a
controller
like
open
daylight
in
that
case,
these
are
the
tools
that
make
it
easier
for
you
in
order
to
get
integrated
into
such
a
system
and.
K
K
P
Michael
Abramson:
this
makes
the
device
manageable,
it
not
manage
the
device.
So
that's
a
distinction-
and
this
is
all
run
in
this-
is
all
created
in
C,
but
you
can
make
create
the
plugins
in
a
lot
of
different
languages
and
it's
it's
made
to
be
fairly
small
footprint
because
we're
running
it
on
home
gateways.
That
was
one
of
these
cases
initially,
and
it
is
a
very
permissive
license.
So
there
it's
not
GPL,
but
Apache
version,
2.
L
Rubbish
k,
kugel
I,
guess:
we've
been
looking
more
the
we
have
implementations
of
the
tooling
that
I
talked
about
earlier
they're,
both
for
on
device
or
on
datasource,
but
also
for
the
tooling,
like
the
management
system,
so
I
think.
Last
year,
ons
we
gave
a
couple
of
presentations
I,
a
variety
of
open
config
presentations
are
either
myself
or
Annie
shake
of
given
talking
about
building
a
management
stack.
There's
examples
in
our
tooling
of
you
know:
building
HP
model
data
instances
validating
them,
applying
them
to
a
device.
L
B
P
B
H
It's
gonna
make
one
comment
following
up
on
what
the
Dan
was
talking
about,
that
this
is
can
be
used
by
others
in
their
products.
That's
that's!
Well,
that's
interesting
for
the
IETF
I
think
it's
really
interesting
that
this
can
serve
as
a
reference
implementation
of
a
reference,
early
implementation
for
emerging
work
that
we're
doing
here
and
it
can
really
inform
at
in
inform
our
work.