►
From YouTube: IETF102-RTGWG-20180716-1550
Description
RTGWG meeting session at IETF102
2018/07/16 1550
https://datatracker.ietf.org/meeting/102/proceedings/
B
A
A
A
A
C
A
So
we've
got
busy
agenda
for
birthdays
this
agenda
for
today,
so
agenda
for
tomorrow
should
be
different
than
usual.
We've
got
two
large
slots
dedicated
to
particular
topics.
One
of
them
is
a
control
user,
plain
separation.
Another
one
is
Devon,
so
we'll
be
focusing
on
these
particular
topics.
With
some
other
presentation
between
we've
got
three
new
RFC's
Wow,
so
well
done
working
group.
A
Thank
you
so
after
Schappert
review
its
past
working
group
last
calls
will
proceed.
Osg
number
of
existing
working
group
documents
in
worked
on
upcoming
adopted
call
the
RPM
model.
It's
been
reviewed
by
young
doctors
and
we
are
expecting
review
from
routing
directorate.
Same
goes
for
draft
tempering
java
chandi.
We
will
adopt
as
working
group
document,
which
doesn't
necessarily
mean
that
we
are
happiest
working
group
shares
with
the
way
editor
have
responded
to
our
comment,
so
we'll
do
something
about
it,
but
we
decided
to
adopt
document.
It's
important.
There
are
seven
implementations.
So
after
idea.
D
D
We
have
two
groups
to
look
at
one
of
them,
a
lot
of
metadata
in
encapsulation
in
the
header
another
is
that
doesn't
allow
them
and
to
that
I
refer
as
a
fixed
size.
Header
might
be
not
the
best
term,
because
they're
first
mentioned
above
mention
of
the
genetics
here
that
stage
and
GUI
they
can
have
fixed
size
what
their
base
header,
but
the
difference
is
whether
you
have
metadata
as
an
option
or
you
don't
next
slide.
Please.
D
Type
field,
so
if
the
CBB
said
then
the
second
field
interpreted
as
a
command,
and
that
gives
separate
namespace
or
an
application,
and
that
includes
OEM.
It's
not
clear.
What
was
the
goal
of
designers
of
this
encapsulation,
because
om
is
not
necessarily
something
that
is
always
done
in
a
control
plane
and
one
of
the
example
could
be
via
feet.
So
let's
go
next
slide.
D
D
D
D
Rfc
83
that
describe
the
use
of
an
SH
with
the
next
protocol
field,
as
none
of
no
payload
suggests
that
öbut
combination
of
open
set
and
next
protocol,
none
can
be
used
as
active
om.
That's
very
much
true,
but
the
problem
is
that
TL
leave
more
metadata
type
to
is
limited
to
512
octet
and
genève
has
even
smaller
a
limit
for
till
the
leg
in
metadata
to
128
offsets
so
which
limits
ability
to
use
this
approach
to
generate
synthetic
traffic
of
arbitrary
length.
D
So
again
the
document
I'm
not
clear
about
who
I
am
kind
of
information
that
follows
their
encapsulation
header
next
slide.
Please,
the
recommendation
I
would
suggest
proposed
in
this
draft
the
following.
If
the
encapsulation
allows
metadata,
then
it
can
use,
it
may
use
obit
to
indicate
the
presence
of
om
information
in
metadata.
D
D
Implementations
in
current
approach
we
use
Martian
address
with
an
additional
IP
defeating
encapsulation,
to
identify
the
OEM
protocol.
Alternative
is
to
use
galle
label
and
yesh
header,
so
generic
associated
channel
header
to
identify
the
type
of
the
payload
that
creates
some
issue
that,
with
the
using
gal
gas,
their
source
identifier
needs
to
be
explicitly
stated.
D
It's
either
to
trigger
performance
measurement
or
to
do
our
collection
and
transport
of
measurement
results
and
telemetry.
Information
example
of
the
protocol
is
alternate.
Marking
method
described
are
discussed
in
RFC
8321
in
situ
OEM
combines
both
it
can
be
used
as
a
measurement
trigger
and
the
same
time
has
a
provisions
to
do.
Collection
and
transport
of
information
hybrid
to
stab
is
a
proposal
to
do
collection
and
transport
of
informations
of
arbitrary
volume.
D
A
A
E
Okay,
so
this
draft
is
titled
a
simple
BGP
based
mobile
routing
system
for
the
aeronautical
telecommunications
network.
So
it's
some
background.
The
International
Civil
Aviation
Organization,
which,
by
the
way
the
headquarters
is
within
walking
distance
to
this
building,
they're
building
an
ip-based
networking
system
for
worldwide
air
traffic
management,
they're
developing
what
they
call
the
aeronautical
telecommunications
network
with
Internet
Protocol
services
or
etienne
IPs,
and
again,
as
under
developing
an
ICAO
communications
panel
working
group
I.
E
It's
going
to
be
an
ipv6
meet
based
network
with
mobility,
capabilities,
quality
of
service
capabilities
and
multi
linking
capabilities,
and
it
accommodates
the
aircraft
with
multiple
data
links.
So
some
of
the
data
links
used
by
aircraft
are
SATCOM
links.
L
Dax
is
a
new
cellular
link,
that's
being
developed
VHF
radio,
and
there
are
many
others
like
arrow,
Max
and
Inmarsat
and
nuridium.
E
So
we've
developed
a
draft
that
we
call
a
simple
BGP
based
mobile
routing
system
for
the
aeronautical
telecommunications
network.
It's
a
BGP
overlay
network,
that's
separate
from
and
removed
from
the
global
public
internet
BGP
routing
system,
so
there's
no
routing
exchanges
between
this
overlay
network
and
the
public
beach.
If
you're
on
it
system,
it's
based
on
a
hub
and
spokes
arrangement
where
we
have
regionally
distributed
sub
autonomous
system,
border
routers
and
centrally
located
core
autonomous
system
border
routers,
the
sub
autonomous
system,
border
routers
advertise
and
withdraw
airplane
mobile
network
prefixes.
E
E
The
core
autonomous
system
border
are
in
a
hub
autonomous
system
and
they
forward
packets
between
the
sub
autonomous
system,
border
routers
and
then
proxies
connect.
Datalink
sub
networks
to
the
overlay.
Clients
are
aircraft
that
may
connect
to
multiple
sub
links,
sub
sub
networks.
If
they
have
multiple
data
links
and
route,
optimization
can
be
used
to
remove
the
autonomous
system
border
routers
from
the
paths
so
in
in
terms
of
a
diagram.
This
is
what
it
looks
like.
E
We
have
the
aircraft
out
on
sub
networks
and
you
notice
that
some
aircraft
have
multiple
connections
to
sub
networks
because
they
have
multiple
data
links.
The
internet
work
connects
the
sub
networks
to
the
the
the
the
the
autonomous
systems,
there's
native
ipv6
within
the
sub
Network,
and
there's
tunnels
across
the
ipv6
internet
work.
So
in
this
diagram,
the
the
routers
that
appear
in
green
and
purple,
the
green
are
the
sub
autonomous
system,
border
routers
and
the
purple
are
the
core
autonomous
system
border
routers
and
between
them
they
run
a
BGP
overlay
over
the
internet
work.
E
So,
in
terms
of
BGP
details,
each
sub
autonomous
system,
border
router
is
a
sub
autonomous
system
unto
itself,
while
all
core
autonomous
system
border
routers
are
members
of
the
same
core
autonomous
system,
the
stuff
autonomous
system
border
routers
advertise
their
associated
airplane
mobile
network
prefixes
to
this
core
autonomous
system,
border
routers,
using
eg
BGP
and
the
core
autonomous
system
border
routers
originate
default,
but
did
not
advertise
any
mobile
network
prefixes
to
the
stub
autonomous
system
border
routers.
So
the
spokes
only
know
about
their
Associated
aircraft,
while
the
core
knows
about
all
aircraft
spoke
associations.
E
So
this
core
autonomous
system
border
arrows
discover
all
the
mobile
network
prefixes
in
the
system
using
AI
BGP,
and
they
can
connect
the
overlay
to
the
global
public
Internet,
in
which
case
they
would
advertise
a
short
and
unchanging
prefix
that
aggregates
all
the
M
MPs.
For
example,
they
would
advertise
a
slash
thirty-two
instead
of
the
dynamically
changing
more
specific
and
M
MPs.
E
E
Since
the
last
version,
we've
had
quite
a
few
comments
on
the
draft
we
removed
the
discussion
of
route,
optimization
and
multi
link
mobility.
These
items
are
covered
under
your
other
drafts.
We've
focused
the
document
on
the
BGP
topology
and
routing
interactions,
and
we've
emphasized
that
entered
the
autonomous
system,
ability,
events
and
quality
of
service
updates
are
not
propagated
into
the
BGP
routing
system.
E
Here's
a
pointer
to
the
draft,
it
was
presented,
ICAO,
a
working
group
by
mobility
subgroup
and
his
act
under
active
consideration
there
as
a
mobility
solution
candidate,
it's
been
presented
in
previous
routing
working
group
meetings
and
also
at
the
lists
working
group
in
IETF,
101
and
100,
and
now
we're
offering.
That
is
a
routing
working
group
item.
So
this
is
a
question
as
to
whether
we
can
have
this
draft
is
a
working
group
item.
A
So
I
would
like
you
to
think
here
encounters
on
working
closely
with
chairs
to
bring
it
to
this
state
that
we
feel
it's
really
ready
to
be
adopted
in
routing
working
group
done
for
a
question
and
we
request
a
review
and
I
believe
Stuart
you've
been
asked
to
provide
early
review
to
it
and
after
ATF
will
also
start
working
group.
Thank
you
again.
A
F
G
Wright
San
Cisco,
so
I'm
presenting
the
yang
data
model
for
ARP.
It's
just
been
presented
in
the
routing
working
group
released
once
before.
Maybe
twice
I.
Remember,
I'm,
one
of
the
one
of
the
authors
on
here:
I'm,
not
the
editor
here
then
I
have
it
today,
some
presenting
on
their
behalf.
We
try
and
keep
this
very
short.
So
this
is
really
quite
simple.
G
It's
our
plea
is
that
we're
really
requesting
Wratten
we're
here,
adoption
this
up
yang
model
who
hopefully
will
then
flow
through
the
working
group
in
a
fairly
expedient
process,
but
mister
settle
for
working
production
is
what
we're
after
why
this
needed.
So
obviously,
some
of
the
are
yanking
shot
is
covered
in
ITF
IP,
so
that
covers
all
the
basic
stuff
with
night
dynamic
and
Static
ARP
entries,
and
the
purpose
of
this
draft
is
to
cover
the
extra
bits
of
ARP
the
lots
of
vendors
implement
but
not
covered
by
those.
It's
the
salad,
yeah
model
there.
G
So
things
like
ARP
statistics,
global
art,
entries
and
proxy
are
so
it's
not
particularly
excited
draft.
It's
not
a
new
protocol.
It's
not
great
to
us.
Save
the
world
by
having
a
standardized
model
here,
does
make
life
easier
for
operators.
So
if
everyone
can
talk
in
the
same
way,
then
it
just
makes
life
easy
for
everyone
and
we're
not
looking
for
perfection
here.
G
There's
been
various
yeah
models
are
taken
years
and
years
to
go
through
ITF
we're
hoping
this
is
a
fairly
basic
young
model
that
should
be
able
to
go
through
these
quickly,
if
I
stuff
enemies
will
design
patterns,
meaning
that
functionality
that
not
everyone's
like
to
implement
is
made
optional.
Putting
on
the
if
keywords,
name
such
a
thing
and
we're
on
to
the
third
revision,
so
we
think
we're
at
the
stage
we're
ready
for
adoption.
G
So
really
is
questions.
Can
you
help?
Can
you
either
show
support
that?
You
think
that
this
is
a
good
thing,
that
ITF
is
working
on
and
are
you
willing
to
do
that?
Will
you
also
having
a
few
more
people
to
review
the
document
and
provide
feedback
and
input
would
be
great
but,
as
I
said,
I'm
hoping
that
it's
just
adopted,
we
can
move
reasonably
quickly
with
it
and
that's
it.
I
mean
I
have
got
slides,
showing
the
current
model
and
I
have
a
slide
showing
what's
changed,
but
don't
think
that's
really
particularly
relevant
I.
A
H
A
A
I
So
what
have
changed
since
last
ITF
here
is
the
list
of
things
that
will
dance
away
make.
It
may
look
like
ITF
model,
but
because
this
one
was
originally
bump,
the
open,
config
models
away
and
now
exporting
the
young
have
set.
The
routing
young
have
sus
defined,
that's
already
an
RFC
and
we
also
reduce
mouth
grouping
and
we
I
did
so
I
would
go
to
the
detail
for
the
action
we
added.
Another
thing:
is
we
updated
the
PCP
example?
I
I
I
So
the
the
bottom
two
function,
actions
as
highlighted
are
the
new
one.
We
added
so
once
your
policy
match
certain
conditions,
what
you
can
do
with
the
the
actions
you
can
perform
you,
how
only
you
either
accept
it
or
out
already
reject
God.
So
we
added
you
can
also
set
a
magic
and
I
also
said,
or
maybe
said,
preference
shall
preference.
So
if
you
use
how
some
other
actions
you
want
to
add
our
policies,
please
let
us
know.
I
J
As
if
you
could
go
back,
one
slide,
I
apologize
I
do
owe
you
feedback
it's
unlike
you,
it's
been
there
for
a
while.
So
you
give
examples
of
said
metric
and
set
preference
being
this
case
16
and
it's
wide
respectively.
A
lot
of
implementations
will
vary
and
internally
what
they
use
such
things
for
is
there
a
reason
why
you
didn't
go
for
allowing
for
like
a
you
at
32
and
then
allowing
for
implementations
to
put
modifications
on
and
tell
why
they're
gonna
allow
those
to
be.
I
I
J
Again,
my
point
being
rather
than-
and
this
is
general
advice
for
the
module-
is
whole-
go
for
a
larger
number
space,
but
allow
for
implementations
to
provide
variations
that
they
could
discover
in
yang
know,
through
extensions
that
say,
I'm
going
to
define
this
as
event
32,
this
implementation,
they
type
def.
This
knows
you're
the
255.
I
H
Asa
Linda
I
believe
yeah
I
believe
metrics
most
people
have
bigger
metrics
than
that
in
16,
but
but
for
the
preference
I
agree
with
you,
there
I
didn't
know
there
were
more
than
one,
but
the
way
to
do
that
I
guess
would
be
to
assign
I
have
a
deviation
into
a
refine
statement
that
refines
the
range
on
it.
Yeah.
Oh.
A
So
policies
are
fundamental
for
any
routing
protocol
and
we
are
in
routing
working
group.
So
please
do
review
this
graph.
You
would
like
it
to
progress.
There
are
a
number
of
things
in
protocols
that
are
pending
because
redistribution,
some
other
things
and
just
answer
your
comments.
Please
reveal
whether
it
works
or
your
implementation
and
we
will
try
to
find
a
common
way
of
treating
deviations
and
working
with
it.
But
please
do
review
it's
important.
A
K
L
L
L
L
L
Based
on
the
discussion
we
had
and
the
comment
be
caught
in
the
last
idea,
we
have
worked
hard
to
define
queues,
I'm
scheduler
see.
The
way
it
works
is
that
we
have
defined
two
different
modules:
cues
and
scheduler
policies,
cues
policies,
nothing
but
set
of
cues
with
classification
based
on
traffic
group
and
hour
traffic
group
is
nothing
but
metadata.
Information
which
is
carried
from
input
to
the
egress
of
the
network
device,
and
the
cue
policy
is
nothing
but
the
classification
of
the
traffic
group.
L
L
L
L
So
as
a
next
step,
you
know
we
got
some
comment
in
the
last
idea
from
this
community.
I
will
suggest
that
you
know
if
you
have
any
further
comments,
please
provide
us
and
should
be
able
to
take
care
of
it.
Also,
you
know
some
of
the
vendors
are
trying
to
prototype
this
model.
Some
of
the
vendors
are
actually
waiting
this
model
to
be
adopted
by
working
group
and
once
it
is
adopted
by
working
group,
then
they
will
be
ready
to
implement
that.
L
A
So
a
model
cup
qsr
is
probably
one
of
the
most
complicated
happens.
There
are
so
many
here
implementation
of
their
routers.
Do
you
feel
guys
we're
converging?
Do
you
feel
the
drop
in
the
current
state
addresses
your
implementations,
or
at
least
gives
able
to
argument,
but
had
been
defined
so
far
are
converging.
A
L
Essentially,
on
this
draft
was
earlier
part
of
the
net
mode
working
group.
We
had
got
certain
set
of
comments
from
them.
We
had
incorporated
those,
then
it
was
moved
to
routing
working
group.
We
have
got
certain
set
of
comments
from
routing
working
group.
Now
we
have
incorporated
those
so
I
would
say
if
there
is
no
further
comments
or
something
we
should
actually.
L
L
Okay,
so
in
the
earlier
version
of
the
model,
we
had
a
set
of
Q
statistics
as
classifier
statistics,
queue,
statistics
meter
statistics.
All
these
set
of
statistics
bought
directly
augmented
to
the
interface.
What
does
that
cause
is
that,
if
somebody
is
looking
for
the
whole
queue
statistics
for
one
interface,
they
have
to
execute
three
different
commands.
L
So
we
found
that
this
is
not
good
and
we
can
have
a
better
framework,
and
so
we
created
a
better
structure
with
that
and
that
cause
that
you
know
the
whole
statistics
can
be
retrieved
per
interface
basis
and
also
it
is
better
augmented.
So,
for
example,
if
somebody
has
to
augment
the
whole
Q
s
interface,
that
structure
it
can
be
augmented
to
a
particular
container
rather
than
augmenting
to
the
interface,
and
then
the
interface
augmentation
will
be
later
to
the
interface
rather
than
oqs.
L
So,
essentially,
what
we
have
done
that
we
have
created
a
separate
container
which
will
have
per
Direction
list
entry
per
directionless
entry
will
have
a
classifier
sets
meter
sets
Q
sets
and
aggregated
named
stats.
So
if
somebody
wants
now,
fewer
stats
per
interface
can
be
retrieved
a
Q
stats
per
Direction
per
interface.
It
can
be
retrieved
and
then
Q
s,
stats
per
classifier
action,
/
interface
or
/
cues
/
direction,
/
interface.
Similarly,
for
meter
and
named
stats,
it
can
be
diffused
separately.
So
that
way
it
is
pretty
buddy
pretty
much
flexible.
L
Also
I
will
again
try
to
allow
operate.
That
aggregate
you
name
sets
is
nothing
but
the
classifier
sets
which
are
aggregated
based
on
a
customer
requirement.
So,
for
example,
if
a
customer
want
aggregated
set
set
of
stats
class
first
sets
so
then
it
can
create
a
name
across
it
and
it
will
get
an
aggregated
sets.
L
As
a
next
step,
yeah
I
would
again,
you
know,
request
this
community
to
provide
any
further
comments
on
the
statistics
part
and
also
these
Q
s.
Telemetry
requirements
draft
which
elaborate
the
requirements
from
the
telemetry
perspective
fork
us.
So
we
are
looking
closely
into
their
draft
and
see
what
X
a
changes
which
we
have
to
bring
into
this
model
to
suffice,
some
of
the
requirement
listed
in
the
draft.
A
L
Means
it
is
actually
not
about
elementary,
but
how
we
can
create
data
set
which
will
suffice
for
the
purpose
of
telemetry.
For
example,
we
have
timestamp
information
which
we
have
to
add
when
the
particular
counter
was
actually
retrieve
from
the
hardware.
So
that
is
one
aspect.
The
other
aspect
could
be
because
in
telemetry
people
actually
draw
the
various
graphs
and
all
that,
so
it
will
actually
be
very,
very
helpful
in
that
case.
L
L
You
know
as
a
threshold
now,
typically,
if
somebody
wants
statistics
for
interface
for
the
Q
s,
so
they
will
just
execute
the
command
and
get
per
interface,
but
when
it
is
telemetry
data
which
is
pushing
the
stats
on
a
regular
basis,
customers
don't
want
to
look
into
all
the
statistics
in
one
shot,
so
they
want
to
see
the
status
which
is
critical
to
them,
and
that
is
the
reason
this
such
holding
of
the
counters
is
very,
very
important.
So.
A
The
timing
kiss
now,
because
implementations
are
going,
there's
number
of
what
we
call
native
models
supposed
to
do
this,
and
there
are
somewhat
different
than
what
we
do
here
so
again
would
be
really
great
to
converge
on
how
we
represents
data
and
progress
on
this
work
that
could
address
our
common
implementations
exams.
Thank.
M
This
is
application
applicability
of
a
CTN
to
enhanced
VPN
we've
been
working
with
Daniel
and
Jeff
and
chin,
and
any
early
on
this
idea
for
a
while,
but
we
didn't
know
where
to
present
and
also
a
CTN
was
in
progress.
So
maybe
right
now
is
a
good
time
to
present.
In
the
wider
audience,
a
CTN
is
actually
progressed
in
teeth,
working
group
and
also
see
chant
and
PCE,
but
PPN
plus
enhance
the
pianist,
will
be
larger
trafficked
and
traffic
engineering.
So
we
believe
that
this
working
group
would
be
a
good
basis
to
start
with.
A
M
Is
that
enhance
PP
and
some
instances
require
customized
control
plane
to
be
able
to
create
delete,
modify
as
they
want
to
dynamically,
and
they
want
to
have
control
over
PP
n.
So
this
is
quite
challenging
requirement
and
also
seamless
integration,
with
both
physical
and
virtual
resources,
and
so
this
is
also
another
complexity,
so
these
are
kind
of
main
requirements
from
VPN,
+
and
I'm,
going
to
show
how
a
CTN
and
an
extended
a
CTN
org
progressed
in
T's,
working
group
and
also
service
model
that
progressed
in
operations
area
can
provide
this
requirement.
Next,
please.
M
M
So
so,
basically,
there
are
several
models
developed
in
layer,
one
layer,
two
layers
with
service
model
in
ops
area
and
also
secant,
and
also
virtual
network
model,
which
is
a
tease
working
guru,
model,
I'm,
gonna
present
shortly
and
another
model.
We
called
T
and
service
mapping
model,
which
is
very
important
model
that
enable
services
to
a
traffic
engineering,
because
there's
some
gap
between
services
definition
and
the
traffic
engineering.
M
So
we
develop
independent,
T
and
service
mapping
model
that
map's
different
services
to
our
T
primitives,
so
that
in
action
can
take
take
place
so
that
services
people
can
understand.
What's
going
on,
there
were
a
virtual
network
into
the
traffic
engineering
networks
and
vice
versa,
an
MPI
model
which
is
a
very
important
model.
We
have
base
model
developed
in
T's,
working
group,
T,
topology
and
Tito,
know
and
then
also
augmentation
of
different
technologies,
starting
from
layer,
0,
1,
2,
3
and
then
SPI
can
be
applicable
with
n
model.
Next
quiz.
M
Yes,
so
this
is
kind
of
recap
of
what
is
taking
place
in
IETF,
ASA,
TN,
t--'s,
working
group
and
ICT
on
work
and
in
other
areas.
As
you
see,
on
the
left
hand
side,
we
have
topology
model
from
layer,
0,
2
layer,
3,
starting
with
higher
to
us
I
to
RS
a
generic
model
and
T
topology
model
is
worth
2,
ninties
working
group
for
layer,
2
and
1
and
0,
and
then
we
have
a
specific
layer,
topology
augmentation
layer,
3
layer,
2,
o
TN,
+,
SS,
ER
and
slacks
great
internal
model.
M
M
We
have
another
model,
called
PM
performance
monitoring
telemetry,
which
is
not
a
generic
telemetry
model,
but
for
iterative,
specific
telemetry
mother,
so
that
customer
can
subscribe
to
what
they
want
to
monitor
and
then
be
able
to
utilize
that
on
their
behalf,
so
it's
a
per
virtual
network
so
which
has
a
little
bit
different
connotation
I,
represent
that
in
a
little
bit.
Next,
please.
M
So
ICT
an
architecture,
principal
input.
One
thing
is
the
virtual
network.
We
allow
customer
to
create
virtual
networks,
as
they
wanna
see
at
their
endpoints.
It
can
be
just
a
endpoint
connectivity
across
multiple
domains
and
multiple
layers,
but
they
don't
want
to
see.
They
don't
want
to
bog
down
details
of
network
layer
technology,
but
they
want
to
create
services
connecting,
for
instance,
or
three
datacenter
locations
with
certain
SLA,
but
on
the
other
hand
they
want
to
control.
You
be
more
than
just
endpoint
connectivity.
M
They
want
to
own
virtual
networks,
which
we
call
tied
to
to
be
able
to
put
their
control
plan
on
the
virtual
networks,
which
is
one
of
the
important
requirements
of
VPN
enhanced
vision.
So
this
virtual
network
was
conceptualized
in
the
architecture
principle
and
then
with
young
model,
for
that
next,
please.
M
So
this
is
a
IETF
ICT,
envy
and
young
model
which
is
adopted
in
this
working
group.
Basically,
this
list
one
needs
to
be
provided
to
for
customer
to
be
able
to
create
their
own
virtual
networks.
As
you
see
here,
we
have
X
access,
point,
end
point
requirements
and
how
does
end
point
relate
to
T
system?
T
we
call
Delta
P,
which
is
what
it
says,
the
piston
so
actually
link.
Termination
point:
that's
good!
Thank
you!
M
Yes,
so
we
kind
of
connected
how
our
customer
end
is
connected
into
T
language,
and
then
we
define
a
VN
instance
and
each
VN
has
a
VN
member
and
via
number,
is
basically
an
internal
connectivity,
but
you
can
have
actually
a
more
than
just
connectivity
model.
If
you
want
to
have
more
details,
a
provisioning
of
each
ero
over
those
virtual
networks,
you
can
do
that.
M
Yeah
three
service
mapping
model
which
we
are
presenting
in
the
t's
working
group
tomorrow,
I
believe
and
as
I
mentioned
before,
it's
very
important
to
bind
layer
n.
So
this
model
with
a
1-2-3
to
tea
system
and
the
service
model,
we
provide
what
service
intents
are
basically
and
some
policy,
but
not
all
policies
specified
in
a
layer
specific
service
model.
So
we
added
a
few
requirements
like
how
my
virtual
network
need
to
be
created
in
the
tea
system,
for
instance
fortunate
of
tea
selection
policy
or
requirement
per
se
and
some
virtual
network
instance.
M
They
don't
want
to
share
with
anybody
else.
I
want
to
have
dedicated
isolated
instance
of
my
resources
so
that
if
other
people
fail
I,
don't
I
don't
want
to
get
affected
by
the
worse
thing.
So
we
have
a
hard
isolation
with
deterministic
characteristic
means.
I
want
my
latency
budget,
but
also
I,
don't
want
a
variation
of
the
latency.
So
in
order
to
accommodate
that
service
model,
we
have
to
work
with
Packer
optical
networks
together
and
optical
network
could
be
transport
TSN
networks
for
their
net.
M
If
we
have
a
test
and
network
underlay
I
think
that
can
be
used
in
order
to
reduce
packet
delay,
for
instance,
so
we
have
to
work
with
a
packet
level
and
an
optical
level
and
any
other
transfer
level
very
closely
in
order
to
create
the
right
traffic
engineering
tunnel
for
a
meeting
a
certain.
The
VPN
instance
requirement.
So
basically,
we
provide
this
selection
criteria
in
the
young
model
and
also
availability
is
a
new
new
item
that
we
added,
because
it's
not
specified
in
the
layer.
M
M
M
Yeah
another
one
that
I
mentioned
Orion
was
performance,
monitoring
or
telemetry,
and
network
autonomics.
This
telemetry
is
not
saying
telemetry
there
we
are
used
to,
but
rather
or
customer
subscribed
to
what
they
want
to
see
and
collect
for
their
virtual
network,
because
virtual
network
has
certain
characteristic
and
I
want
to
look
at
my
latency
and
utilization
and
combination
of
the
two.
For
instance,
question.
N
H
A
H
A
Protocol
models
without
understanding
service
requirements
is
pretty
much
worthless
unless
it
can
build
full
vertical
from
service
delivery
up
to
Iranian
Network.
Our
our
work
here
is,
you
know
unnecessary.
So
the
intention
here
is
to
show
workflow
of
from
service
request
coming
from
business
logic
or
services
there
up
to
the
network
and
eventually
what
we
do
here
should
be
blocked
in
there.
H
M
Yeah,
absolutely
the
actor
young
model
is
progressed
in
his
working
group,
but
this
is
a
PPN
applicability,
pn+
applicability,
which
we
believe
it
fits
in
here
as
an
information
model,
not
as
a
standard
standard
draft,
but
introducing
how
a
system
can
be
applied
for
PP,
n,
plus
and
I
believe
Louis
coming.
So
he
might
say
this
is
actually
this
working
group
item
but
which
I
didn't
know
what
I
didn't
hear.
First,
yeah
I'm.
O
Okay
with
it
being
sort
of
the
visibility
of
the
work
being
socialized
here,
so
you
know
whether
it's
here
or
somewhere
else
I
think
I
think
it's
helpful
for
more
people
to
understand
it.
At
the
last
meeting
we
talked
about
Stuart's
document,
which
was
aiming
to
be
a
little
more
looking
at
heading
towards
the
protocol
extensions,
even
though
it
doesn't
have
it
yet.
O
G
M
P
M
Performance
measurement
protocols,
I,
don't
know
what
that
means,
performance
measurement
protocol
or
import
ago,
and
we
are
independent
of
that.
Okay.
This
is
actually
the
right
slide,
which
I
haven't
explained.
Basically,
customer
has
subscribed
to
the
performance
metric.
They
are
interested
in
for
their
service.
M
For
instance,
I
want
to
have
latency,
strict
latency
50
millisecond
from
endpoint
one
to
end
point
to
but
endpoint
one
ten
point:
three,
maybe
I
have
a
different
requirement
desk
or
virtual
network
okay,
so
they
subscribe
it
and
then
network
basically
are
collect
from
various
level
to
drive
that
you
know
the
performance
matrix
that
is
meaningful
to
the
customer,
so
customer
only
see
end-to-end
measure
and
whether
it
is
met
they
are
satisfied.
If
not,
they
have
ways
to
Australian
scale
out
depending
on
the
performance
metrics.
So
next
one
actually
shows
some
examples.
M
Produces
this
metrics
actually
devised
a
link
has
the
produce,
link,
level
latency,
but
customers
interested
end-to-end
connection
level.
So
you
have
to
stitch
it
together
at
the
orchestration
layer
or
network
control
layer
collecting
from
domain
level
link
statistics
for
instance,
and
then
give
the
information
to
the
customer.
Ok,
yeah,
so
basically
yeah.
So
this
model
is
here
so
take
a
look
and
tomorrow
actually
in
T's
working
group,
we
are
going
to
present
another
presentation
on
this
detail
for
10
to
15
minutes.
So
please
show
up
okay
yeah.
So
this
is
a
last
slide.
M
Basically,
the
point
here
is
that
all
those
are
four
important
requirements
for
VPN,
plus
or
enhanced.
Apn
is
basically
made
with
service
model
and
actn
models:
VN
&,
NT
service
mapping
and
T
telemetry
in
conjunction
with
T,
to
follow
the
connectivity
matrix
model,
and
we
also
have
a
self
enabled
to
follow
the
model
in
this
working
group,
which
will
be
also
presented
tomorrow,
which
is
not
original
ICT
and
scope.
M
Q
Okay,
hello,
everyone,
so
this
is
the
lobby,
so
there
my
presentation
topic
uses
the
network
Elementary
for
a
walk.
Okay,
in
fact
that
this
had
to
be
immoral
present
either
in
the
last
idea
meeting
in
both
RTG
working
group
and
also
the
OPI
CWD
working
group.
So
we
would
like
to
introduce
the
latest
progress
about
this
work.
Q
Okay,
so
here
is
the
addition
of
this
standard
walk.
So
that's
the
after
the
presentation.
We
get
many
these
comments
and
also
get
a
more
cooperation
from
this
ti.
I
know
so
that
China
mobile
and
also
the
ICT
and
the
Nokia
and
as
a
cellist
can
calm
so
that,
based
on
the
comments
and
the
corporation's,
we
update
the
draft
and
that
is
to
show
the
progress
of
the
elementary
walk.
So
the
first
one,
that's
the
clear
definition
and
the
character
risk
summary
of
the
network
at
elementary
because
the
in
the
last
presentation.
Q
So
that's
the
many,
had
a
concern
about
the
scope
of
the
network
Elementary.
So
we
also
would
not
have
wanted
to
boil
the
ocean
so
that,
in
order
to
clarify
the
scope
so
that
we
in
the
draft,
we
determine
the
proper
this
the
clear
definition
and
a
characteristics
about
the
network
at
elementary
based
on
this.
So
the
user
will
have
to
understand
the
clear
distinction
between
the
conventional
om
solutions
under
the
net
work
elementary
solutions,
the
second
way
there's
a
new
module
for
the
framework,
so
that's
the
accelerator
and
the
human
Elementary.
Q
So
that's
the
we
get
this
in
for
comments
from
these.
The
co-operators
says
they
said
that
not
only
the
management
plain
elementary
control
in
elementary
and
that
a
Tamilian
elementary
in
order
for
the
elementary
worker,
so
that
we
need
a
some
of
the
activator
and
even
had
elementary
beyond
the
network
herself.
So
that's
this
add
a
new
module
for
the
free
walk.
Okay,
any
comments
there.
P
Was
a
question
from
Greg
mirskiy
when
you
refer
to
telemetry,
do
you
mean
telemetry
information
as
network
state
information
or
telemetry
protocol?
Is
protocols
to
collect
and/or
transport
network
state
information,
pulse
okay
and
another
question?
Would
you
agree
that
the
machine
readable
is
what
you
refer
to
as
telemetry?
Sorry?
P
Q
Thanks,
another
satellite,
that's
the
user,
and
you
can
a
new
content
for
the
control
pauline
callum
entry.
So
that's
the
user.
We
as
identified
the
requirement
as
a
challenges
of
the
control
plane
for
the
elementary
walk,
because
now
that
much
worker
user
related
with
management,
her
plane
and
also
the
data
plane,
so
that
a
user
after
we
use
at
Academy
the
requirements
as
a
challenging
of
the
control
panel
elementary.
So
we
propose
that's
the
network
monitoring
protocol.
Q
So
that's
the
ghost
of
all
the
control
valiant
elementary,
so
they
either
had
seminar
triggered
by
the
signal
source
from
the
BMP.
Okay
and
the
last
points.
Let's
use
a
new
content
for
the
theta
Flint
Elementary.
So
here
that
is
it
because
we
have
the
many
aware
mechanisms
or
the
elementary
mechanisms.
So
here
we
have
this
terminology
for
the
technical
classification,
this
classification,
including
the
active
and
the
passive
in
band
and
out,
opened
e
to
e.
That
means
the
host.
We
were
aware
of
this
elementary
or
this
in
network,
and
also
that
user
identifier.
Q
There
is
a
fellow
level
pass
level
or
the
Noda
level
so
that
the
based
on
this
is
easy
to
classify
the
different
technologies.
Okay,
as
well,
also
there's
a
new
technologies
introduced.
This
is
the
important
work
either
IP
of
p.m.
so
that
the
used
as
the
alternative
amok
for
the
inbound
traffic
for
the
point
to
point
and
the
multi-point
to
multi-point
service.
Q
Q
Okay,
so
this
is
the
we
reveal
this.
The
walk
the
first
day
is
a
challenging
of
the
today's
network.
So
now
that's
the
network
because
more
and
more
complex,
in
fact,
it
has
to
support
reach
for
the
applications.
This,
including
cloud
5g
I,
would
he
overlay
underlay
and
the
VPN
network,
slicing,
etc,
and
now
that's
the
in
order
for
the
byte
network
application
our
censored
he'll
and
also
become
there's
a
sensor
table
to
the
network
of
performance.
Q
So
this,
including
the
parameter
such
as
pan
wise,
latency,
Jade,
her
package,
the
law
or
Java
ratio
and
network
her
true
like
this
one.
So
that's
the
network
availability
it
become
more
and
more
important
for
the
purpose
of
the
network,
om
for
region
planning
security
and
in
the
troubleshooting
so,
and
also
this
the
our
auditors
for
the
network
availability,
we
think
is
not
enough.
So
the
first
drawback
of
the
auditors
is
the
lack
of
the
application
level
visibility.
So
you
so
always
some.
This
is
the
network.
Q
P
Comment
from
Gregg
Murphy,
who
just
wants
to
make
sure
that
you're
familiar
with
the
work
of
the
L
map
working
group
in
particular,
informational
and
yang
models,
I
mean
so
you
trusted
s.
What
yl
map
working
group
work
wanted
to
make
sure
that
you're
familiar
with
that
work
and,
in
particular
the
information
and
the
young
models?
Q
Q
Okay-
and
this
is
the
next
one.
So
this
is
also
we
face
a
challenge,
the
future
network.
So
now
that's
the
use
of
the
network
management
and
service
evolved
to
become
a
unit
and
retrieval
and
automatic,
so
they
say
either
we
have
identified
the
possible
characteristics
so
the
first
way
user
to
reduce
to
the
man
labor
a
human
labor.
The
second
way
is
the
improved
aichi
eyetality
and
the
performance,
and
so
the
way
to
optimize
resource
the
efficiency,
yeah
and
also
I
won.
So
they
said.
No,
that's
it.
Q
Q
A
Q
Ok,
so
I
will.
This
is
in
order
for
this
work,
so
we
have
this
in
order
for
button
all
data
as
a
half
we
of
the
virtual
and
the
cloud-based
the
device,
so
that
a
user
we
have
the
future
most
of
the
devices
in
the
network.
So
we
think
this,
the
network
of
visibility,
either
become
a
very,
very
challenging
work.
So
here
we
think
that
we
hope
to
identify
this.
The
network
visibility
based
on
the
network
at
elementary
to
set.
However,
this
closed
loop
than
the
to
improve
the
intelligence.
Q
Ok,
so
this
is
a
current
resolution.
We
will
skip
hated,
but
this
may
be
family
with
this
one
SMP
syslog
SMP,
ok,
so
here
this
user,
we
identify
this
the
distinction
between
the
conventional
network
aware
and
the
network
at
elementary,
so
for
the
yellow
one.
We
seen
that
as
the
characteristics
of
the
conventional
network
of
air
and
as
a
photo
per
my.
So
this
is
the
characteristics
of
the
network
at
elementary.
So
here
we
can
say
that
this
is
the
traditional
usage
history,
Beauty
based
and
now
we
must
take
into
account
of
the
SD.
Q
That
means
the
central
based
and
although
I
use
the
pusher,
a
pool
based,
but
now
you,
the
pool
based
I,
also
know,
there's
a
dream,
indeed
her
and
also
the
bigger
date.
Her
and
I
said
now
that
you,
the
human
consumer,
but
now
that
I
use
a
machine
in
consumer
that
for
the
automated,
under
than
their
traditional,
we
use
sa
RI,
but
Aneta
uses
the
more
program
and
also
that's
a
user.
All
the
way
users
react
you,
but
now
that
emphasize
on
the
proactive
and
as
the
user
troubleshooting
and
now
that
also
emphasizes
the
prediction.
Q
I
also
just
use
the
independent
horse.
Now
that
we
want
the
unified
and
correlate
heat
tours.
Okay,
based
on
this
walk,
we
can
see
here.
That's
the
network
operation
applications.
We
think
these
control
in
elementary
management
elementary
a
taste
of
template
elementary
now
that
we
also
identify
external
data
and
the
human
health
elementary.
So,
and
also
here,
that's
the
whole
picture
of
this
different
plane
and
also
this
is
the
different
components
from
the
five
aspects.
Q
Sorry
I
must
have
finished
my
presentation,
okay,
and
this
is
also
the
I
usually
based
on
this-
is
a
picture
also.
We
identify
the
Challenger
of
that
network
at
elementary
for
the
top-down
this,
the
design
for
the
new
technologies
and
solutions
or
improvement
of
the
existing
technologies.
So
this
here,
including
the
dynamics
as
a
multiple
resource
and
the
performance,
the
import
impact
on
the
network,
telemetry
solutions,
so
this
will
also
together
our
this
is
the
design
and
the
improvement.
Ok.
Q
So
here
we
would
like
to
this
is
the
conclusion
so
the
first
year,
so
that
we
wanted
to
promote
the
significance
of
a
telemetry
work
in
IETF.
We
think
of
this
IDF
uses
the
read
home
for
the
elementary,
because
this
is
the
proud
co-relate
hitter
with
the
device,
and
the
controller
is
not
only
the
high-level
walk
second
away.
We
also
wanted
to
formalize
the
elementary
related
terms
and
Technology
classification
in
IETF,
so
this
user
also
to
consolidate
existing
work
and
also
to
get
a
future
network,
a
future
work.
Q
Okay,
and
as
here
the
user
we
I
just
mentioned
briefly
fear
about
a
network
monitoring
protocol
work.
We
produce
the
draft.
So
this
draft
is
a
two
for
the
control,
plane
monitoring
and
we
also
use
a
further
control
for
a
control
protocol
elementary.
So
this
is,
after
this
draft
used
to
being
proposed.
We
get
a
lot
of
comments
and
the
so
here
that
we
are,
we
would
in
order
to
discuss
Nemo
in
the
association
and
welcome
to
a
cool
working
group
under
the
OPR,
see
working
group
for
more
discussion.
Q
R
Q
Q
So
the
first
way,
so
these
are
the
in
fact
this
is
the
some
token
you
consolidated
the
existing
worker
or
in
the
IETF
this
you
cruising
the
Netcom
improvement
and
the
GRP
see,
and
also
that's
you,
the
Iowa
doc
this
one,
but
also
that's
the
user
Sam.
This
is
the
p.m.
he
and
the
days
of
the
MP,
so
they
say,
uses
a
protocol
right.
R
Q
I
think
that's
a
you
think
that
this
is
you
the?
Maybe
this
is
a
user.
You
think
this
is
relatively.
You
think
I
use
a
new
protocol
or
they
say
user.
You
can
see
same
protocol
from
out
on
the
wheel,
maybe
the
TR
PC,
comparing
with
the
traditional
there's
a
SNP.
We
also
seem
to
use
a
new
protocol,
so
we
synced
as
the
user.
You
can
see
that
we
have
some
the
existing
work.
We
try
to
consolidate.
Q
That's
what,
by
the
way,
you
think
I,
usually
maybe
lack
assembly
so
that
we
can
something
that
is
the
new
requirement
so
that
we
enhance
these
of
the
protocols.
So
that's
what
that's
what
we
also
think
somebody's
a
new
protocol.
You
extension
we
will
not
have
would
like
to
data
user,
create
a
new
protocol
or
like
this
one.
That's
not
the
purpose.
Q
We
would
like
to
this
way,
so
this
is
also
you
can
see.
We
do
a
lot
of
work,
so
that's
a
user.
We
define
these
that
terminology
and
dear
to
the
to
the
clarify
the
distinction.
So
they
see
the
castle
want
to
identify
this
as
a
scope
carefully
and
also
they
had
the
scope
of
the
different
different
solutions.
So
that
is
also
the
purpose
of
this
and
he
have
work.
Okay,
Thank
You.
A
Q
S
You
are
asked
question,
yes,
question
actually
so
ignores
McDonough's.
Did
you
try
to
socialize
this
topic
with
the
potential
users
of
the
solutions
that
you
are
proposing?
That
is
the
actual
operators
and
other
problems
that
you
are
stating
as
problems,
in
fact,
real
problems,
realistic
problems
that
are
happening
that
are
still
current
and
whether
the
solutions
that
you're
proposing
in
fact
fit
into
the
operational
model
of
the
existing
operators
at
all.
S
Q
Definitely
we
think
the
operator
and
these
the
carriers
and
OTS
are
very
important
for
us
that
it
is
not
only
this
network,
but
also
we
can
follow
their
source
of
the
existing
IT
technologies.
They
used
the
use
of
the
maintenance
technology
for
the
computer
resource
and
the
story
resource.
In
fact,
we
had
a
lot
of
discussion
with
this.
The
OTS
also
we
presented
this
work
in
the
MPR's
Congress
for
the
you
know,
Paris
Texas
one,
so
that
also
we
can
see
there.
These
are
many.
Q
A
So
Robin
colors
I
mean
it
sounds
like
we
should
jump
higher
better
for
the
next
steps.
I
would
really
like
to
see
proper
comparison
between
existing
tools.
What
could
be
done
and
what
you
think
is
missing
before
going
and
tell
everybody,
we
need
something
new,
let's
assess
what
their,
how
it
could
work
and
indeed
talk
to
the
consumers
of
the
technology
asking
them
what's
missing,
Robin
talking
to
you.
S
Right
so
a
quick
overview
from
the
Hobson
Management
Area
on
how
yang
is
doing
overall
and
what
can
be
done
about
that
next
slide.
Please
actually
I
think
I
have
that.
Thank
you.
Well,
it
appears
that
yang
is
doing
not
so
that
well,
and
this
is
rev
alarming
ietf
as
an
organization
and
we
as
a
community
have
spent
15
years
working
on
the
new
paradigm
of
manageability.
S
But
the
big
question
is
whether
that
solves
sexual
problems,
whether
that
is
actually
getting
deployed,
whether
that
is
getting
deployed
in
a
way
that
we
as
as
IDF
understand
and
whether
we
understand
what
we
are
trying
to
solve
the
problem,
which
is,
in
fact,
the
actual
problem.
The
way
how
the
operations
community
sees
that
wonderful
arm,
I
wouldn't
use
the
word
alarming,
but
it's
probably
not
right
one.
S
Yes,
if
we
take
it,
if
you
take
a
look
at
the
overall
look
of
the
young
base,
manageability
vision,
it's
really
beautiful
if
I
want
to
use
that
today
for
practical
purposes
can
I.
Do
that
not
really
if
we
look
at
the
practical
aspects
of
that,
we
do
have
fragments
and
I
ETFs
organization
of
the
role
is
good
at
developing
fragments
protocol,
fragments,
young
fragments
and,
and
the
like
IETF
is
historical
and
not
good
at
dealing
with
systems.
S
If
we
look
from
operational
point
of
the
world,
what
matters
is
a
system,
how
the
components
and
the
fragments
bind
together.
How
do
they
work
as
an
operator
I
do
not
care
whether
the
protocol
is
this
or
that
or
whether
I
use
this
mechanism
to
get
something
or
that
mechanism.
What
is
important
for
me
is
that
I
treat
that
as
an
end-to-end
system
and
I
deal
with
the
actual
tasks
that
I
needed
for
the
from
a
service
perspective.
The
element
does
not
matter.
The
protocol
does
not
matter
that
much
so
practical
completeness
of
the
models.
S
You
can
do
quite
a
lot
today
with
based
on
the
models
that
are
being
developed
for
different
protocols
for
different
components,
but
very
likely
you
can
do
80%
of
what
is
needed
practically.
Yet
the
model
covers
much
more
functionality
than
that,
but
it's
80%,
it's
not
a
full
functionality
that
is
needed
for
majority
of
a
practical
deployment.
S
The
other
important
aspect,
practical
aspect
is
the
state
of
the
tooling
developing
a
model
might
seem
to
be
the
end
goal.
No,
it
is
not
the
fact
that
the
model
exists,
but
I
cannot
use
in
a
practical
way
its
equivalent
to
the
fact
that
the
model
cannot
be
used.
There
are
two
sides
to
this:
one
is
the
education
operations.
S
Community
does
not
necessarily
have
the
level
of
understand
ending
of
what
model-based
manageability
is
for
the
role
you
hear
all
spent
decades
talking
about
young
and
developing
the
things
you
know
the
intricacies,
the
language,
the
modeling
process
and
the
like.
It's
not
necessary
that
the
actual
operations
community
has
that
insight.
It's
a
question
of
education.
The
other
aspect
is
tooling
and
workflow
processes
and
how
that
integrates
into
the
existing
to
link
and
workflow
processes
of
the
operators.
S
It's
unrealistic
to
expect
that
Greenfield
deployments
are
widespread
and
do
exist.
You
have
to
deal
with
things
which
exist
in
college.
You
need
to
integrate
into
that.
If
you
have
a
beautiful
solution
which
cannot
work
with
what
you
have
already,
it's
just
a
reality
that
investments
done
into
building
working
but
obsolete
system
are
not
going
to
be
withdrawn.
Just
for
for
the
reason
that
ITF
made
yet
another
shiny
young
model.
That
is
not
going
to
happen.
S
S
They
are
large
enough
and
and
resourceful
enough
to
build
what
they
need
themselves
and
in
fact
they
don't
need.
The
idea
for
that
and
the
remaining
80,000,
not
just
80,
are
the
ones
which
rely
on
what
vendors
deliver
as
a
common
set
of
the
requirements
which
are
coming
from
all
8,000
of
them,
and
that
is
the
group
which
requires
ITF
and
ITF
to
come
to
the
practical
solution
which
are
usable
by
majority
of
them,
not
absolutely
by
them
by
all
of
them
and
at
the
same
time,
not
selectively
by
only
a
small
fragment.
S
It's
probably
more
practical.
That
idea
focuses
on
the
requirements
of
the
80
side
than
frequently
more
hard
and
more
visible
requirements
of
the
20
side,
so
I'm
not
saying
anything
negative
about
any
of
those
20
or
so
Tia
ones,
but
the
solutions
which
are
applicable
in
their
environment
most
probably
are
not
applicable
in
the
environment
of
the
majority
of
a
average
network
operators,
and
the
open
question
is
where
I
ETA
F
stands
in,
that
which
direction
it
follows
and
and
to
which
voice.
S
S
We
do
not
seem
to
have
a
channel
with
the
operations
community
about
how
things
are
being
used.
What
are
the
actual
problems,
and
we
also
do
not
have
a
channel
to
sense
the
feedback
of
those
mechanisms
which
ATF
develops,
whether
that
fits
into
the
environment
which
operators
are
using.
So
it's
a
bi-directional
communication
channel
to
get
the
requirements.
What
what
are
the
problems
and
how
to
solve
that?
The
alarming
point
for
me
is
that
we
are
in
a
situation
where
vendors
are
telling
operators
how
to
build
their
networks.
S
That
just
does
not
seem
to
be
right
when
they're
certainly
have
insight
and
how
to
build
platforms.
But
the
platform
is
not
a
designing
component
and
it's
just
not
right
overall,
from
a
personal
point
of
view,
to
try
to
build
a
network
based
on
the
platform.
The
platform
is
just
a
brick
which
provides
some
function,
but
the
overall
end-to-end
design
is
not
something
that
is
governed
by
the
particular
functionality
of
the
platform.
Yes,
it
limits
assets,
but
that
should
not
be
this
way
around.
S
S
But
routing
is
a
fragment
in
if
I
want
to
run
a
platform.
If
I
want
to
run
a
service,
there
needs
to
be
much
more
things
than
just
routing,
so
routing
design
team
took
a
problem
and
try
to
work
on
that
trying
to
align
with
the
community
the
good
thing
that
many
of
the
many,
but
at
least
the,
where
operators,
where
the
target
uses
for
that.
If
we
are
talking
about
a
broader
scope
of
usability,
certainly
we
don't
have
enough
of
a
participation.
S
S
Maybe
it
would
be
good
to
try
an
experiment
with
having
a
joint
working
group
across
to
of
research
organizations
this
from
ITF
sides.
This
working
group
would
produce
informational
documents
at
most
and
maybe
not
even
out
of
season,
but
a
temporary
working
document
which
guide
the
development
of
other
particularly
protocol
level
models,
both
alone
on
element
level
and
on
the
service
level.
S
That
to
me
personally,
that
appears
a
problem.
Itf
is
historically
good
at
developing
solutions,
spending
enormous
amount
of
time
on
that,
and
eventually
those
solutions,
just
don't
get
deployed
for
various
reasons,
sometimes
good.
Sometimes
bad
I'm
not
saying
that
yank
overall
is
moving
into
that
direction,
but
there
is
a
risk,
one
of
the
reasons
for
that
and
that
the
check
point
of
the
model-based
manageability
was
around
15
years
ago.
The
workshop
that
mentioned
IP
workshop
on
what
are
the
problems
and
then
what
needs
to
be
done
since
that
time.
S
T
I
was
part
of
the
routing
young
design
team
working.
You
know
like
design
team,
and
my
issue
is
mainly
that
we
discuss
about
a
lot
of
models
without
actual
implementation.
Unless
people
start
writing
code
for
what
they're
trying
to
model,
nothing
will
change,
because
until
then
we
will
have
experience
how
how
things
are
going.
The
other
part
is
what
you
said
about
the
other
80,000.
T
T
So
I
want
to
start
talking
to
the
operators.
This
is
where
many
of
them
are
still
there.
Yes,
they
are
early
adopters
they're
putting
into
it,
but
many
of
them
are
saying,
but
we
only
know
how
to
debug
with
the
CLI,
and
one
of
the
issues
is
also
that
you
have
to
think
about.
More
is
how
we
will
create
debugging
tools.
It's
not
just
a
modelling,
it's
how,
once
you
put
it
out
there,
how
will
you
solve
how
we
will
figure
out
when
there
is
a
problem
introduced
into
the
network?
T
Which
way
will
it
this
B?
We
are
a
lot
of
focused
on
the
configure
configuration
part,
but
we
have
to
be
also
much
more
focused
on
the
on
the
operational
and
one
of
the
things
that
the
existing
vendors
have
is
the
translation
between
the
config
models
and
the
operational
models,
these
two
to
figure
out
what
this
configurational
model
means.
You
know
to
this
operation
model
and
the
translation
between
them.
That
is
a
major
problem
and
we
are
still
building
models
in
the
ITF.
S
T
On
the
last
part,
on
the
tooling
I
think
there
are
good
tools
out
there.
They
are
proprietary,
there
isn't
much
open-source
there's
actually
there's
almost
no
open-source
tooling
out
there
available.
That
is,
that
is
worth
putting
together.
So
unless
there
is
again
a
group
willing
to
put
something
to
get
in
to
put
energy
into
it
and
write,
we
will
have
to
rely
on
commercial
tools
that
are
either
built
in-house
proprietary
or
you
can
get
it
from
third-party
vendors.
T
S
H
Sealing
of
Cisco
Systems
I
I'd
like
dancer,
you
are
our
models
that
we're
putting
forward
right
now
in
eius,
eius
and
OSPF,
and
see
us
there
in
the
sad
state
of
having
the
ops
at
Upstate
and
config
split,
because
I
think
we've
done
a
good
job
of
putting
them
together,
but
anyway,
I.
Aside
from
what
you're
saying
do
you
have
some
strong
people
from
operations
from
the
providers
to
lead
this
right
now
they
have
not
only
strong,
but
they
have
the
amount
of
time
to
devote
to
it.
S
Answer
is
probably
yes,
and
that
is
the
reason
why
this
experiment
is
being
tried.
That'd
be
great,
I
think
it's
a
good
idea.
This
is
not
something
that
happened
overnight
today.
This
has
been
socialized
for
quite
long
with
that
community
and
yes,
I
have
been
trying
to
pull
in
some
people
and
then
twist
their
arms
in
trying
to
do
that.
If
this
goes
as
seen
as
a
working
group,
one
equivalent
of
working
group,
certainly
the
chest
would
not
be
from
IT
of
community.
R
R
We
keep
telling
that
people
need
yang.
Actually
yang
is
just
a
means
to
an
end.
What
we
have
to
do
is
show
the
operators
how
they
will
be
able
to
reduce
your
packs,
and
then,
though,
those
80,000
or
many
people,
there
would
see
what
they
could
save,
but
if
you've
got
the
yang
and
the
tuning
and
everything
plus
you
dication,
that's
a
big
thing
that
we
need
to
do
right.
R
A
A
Q
I
mean
from
Holly
in
fact,
regarding
this
I
think
they
say
you
know,
could
a
topic
for
that
is
caution.
This
way,
I
have
a
Marshall
one
to
see,
but
I
tried
it
who
tried
to
catch
some
of
the
key
points,
so
the
first
I
wanted
to
from
these
the
some
aspects,
the
first
of
all
I
think
we
in
the
past
eight
years.
We
do
a
lot
to
the
fusion
work
about
this.
The
young
model
standard,
as
seen
as
almost
the
university,
the
samp
you
host
the
efforts.
Q
I
think
the
Desi's
should
be
appreciated,
but
the
second
one,
that's
the
the
in
fact,
I
think
that
that,
based
on
my
aspiration
for
the
young
model,
that's
used
in
the
industry,
I
think
it
was
a
beast
and
I
I
see
I
press
a
person,
mystic
press
a
mystic
about
it.
You
seem
suggest
now
that
we
either
enter
into
a
negative
a
circle
about
this
walk,
because
this
is
the
young
model
uses
not
any
novel
for
the
application
or
cannot
catch
our.
This
is
the
new
application,
so
that
but
a
user.
They
want
this.
Q
Well,
so
that's
we
have
a
su
defender.
Somebody's
the
private
young
models,
but
after
this
is
the
standard,
the
young
models,
the
so
that
they
would
not
like
to
change
the
existing
the
private
young
models
so
that
a
user
to
accessibility,
user
again
and
also
some
this
is
the
based
on
the
open
source
or
the
practical
completeness,
but
the
DES,
the
user,
only
two
can
be
applied
for
limited
scenarios,
can
other
satisfy
all
the
requirements
so
there.
So
that's
the
user
also
cannot
have
solve
this
other
issues,
so
I
think
III
need
to
persevere.
Q
Mystic
about
this
one,
so
I
also
the
further
one.
We
also
thought
about
this:
the
possible
solutions,
so
I
think
the
first
uses
the
yes
we
shoulda
tried
to.
This
is
a
set
up
the
channel
with
operation
community,
in
fact,
in
my
presentation,
I
also
user
mentioned,
and
also
a
people,
a
lot
of
people
as
a
reserve
over
there
about
this.
The
OTP
also
this
IT
companies
for
this.
For
this
walk.
In
fact,
they
also
faced
a
lot
of
challenges
about
this
operation,
but
they
have.
Q
This
is
a
rich
full
reach
of
your
tools
for
this
one.
This
is
not
only
the
standards.
These
other
operations,
sorry,
but
also
this
is
the
operation
community
and
also
a
lot
of
these
Assam
eat,
and
also
this
is
the
open
source.
So
this
is
active
and
they
use
a
lot
of
information
and
their
tours
can
be
shared,
but
until
now
button
further
for
the
force
and
they
used
for
the
traditional
conventional
vendors.
So
this
is
a
so
they
say
either
can
be.
Q
O
Q
Okay,
I
would
show
so
that
means
that
you
know
young
model,
you
don't
force
are
the
in
fact
that
we
can
change
now
that
we
changes
configuration
existing
configuration
to
some
control
protocols,
for
example
the
PCE,
for
the
study
obviously
doesn't
mean,
say
you,
the
PCE
control
protocol
for
the
pro
venue
for
the
condo,
so
I
think
this
is
the
advantage
of
the
IETF.
So
also
we
can
say
that
you
should
also
think
about
that.
We
change
to
to
some.
This
Walker
uses
this
method
because
they
see
you
the
more
increment
holding
yeah.
S
Yes,
quite
many
points,
I
will
respond
in
summary,
with
with
one
ants,
are
not
certain
how
much
you
like
that?
Just
the
fact
that
something
can
be
done
does
not
necessarily
mean
that
it
should
be
done,
and
we
have
plenty
of
a
selection
of
various
protocols
and
and
other
things,
and
we
better
use
what
is
already
there
for
the
simple
reason:
it's
much
higher
probability
that
Operations
community
is
aware
of
that
and
know
how
to
deal
with
that
than
to
try
to
invent
something
completely
new,
which
does
mostly
the
same
words.
T
T
We've
been
starting
to
add
support
for
young
models
to
various
routing
demons
and
that's
a
pretty
active
in
a
project.
We
are
starting
to
develop
code.
You
know
adding
things
to
it.
Coming
up
with
different,
you
know
abstraction
proposal.
Some
of
them
is
okay,
let's
do
just
like
a
standard
or
less.
What
is
a
CLI
and
the
other
one
is:
let's
try
to
see
some
of
the
abstractions
enough
of
us.
T
Vendors
here
have
experience
not
first
ten
per
second
hand
through
communicating
with
the
operators
what
they
are
looking
for
and
we
might
run
a
project
that
lets
put.
You
know
but
walk
the
walk,
not
just
talk
the
talk
where
we
can
come
up
with
some
of
the
proposals,
try
to
implement
them
and
see
where
it
will
take
us,
and
we
will
have
a
first-hand
experience,
because
what
will
will
be
first
implementing
and
after
that
who
will
be
discussing
about
the
actual
model.
Great.
A
K
T
K
B
T
Is
there
is
a
group
we
are
starting
to
meet
at
the
ITF
quite
regularly
and
instead
of
just
talking
come
up,
you
know
and
there
we
can
have
the
project.
We
can
do
it
in
parallel
and
see
what
works
for
a
dozen
there's
a
reason
why
we
drew
myself
from
several.
You
know
drafts
because
they
started
to
do
and
work
on
things
that
weren't
implemented
and
I.
Warned
you
not
happy
with
that.
S
T
S
A
U
T
Q
Okay,
sequence:
okay!
So
that's
a
regarding
this.
A
young
Walker
I
wanted
to
the
possible
suggestion
for
the
for
this
one.
So
that's
the
in
fact.
For
the
past
years,
in
fact,
we
have
do
a
lot
of
work
to
some
is
on
the
pieces
of
the
young
model.
We
have
the
IDP
and
the
PDP,
also
the
I'm
cars.
So
now
I
think
that
maybe
this
year,
the
right
time
we
center
this
the
design
team.
That
is
the
base
on
some
this.
Q
The
useful
cases
to
examine
the
existing
pieces
of
the
young
model
is
the
can
satisfy
the
obviously
user.
Can
the
requirements
of
this
is
a
solution
based
on
these
of
the
u2
kissing's.
So
that's
what
I
think
that
may
be
easy
easier
for
us
to
communicate
with
the
operation
team,
because
they
always
appear
about
this,
the
user
cases
and
the
solution,
but
you're
always
talking
with
them
about
some
of
this,
the
IDP
our
model,
this
may
be,
they
cannot
a
set
a
picture.
Q
So
that's
a
DES
is
also
in
this
is
may
suggest,
because
now
I
always
encounter
this
situation.
Do
you
even
for
the
over
adversity,
just
icky
buddy
see
you
the
young
this
year,
the
RC
or
the
trough
the
list
so
a
lot
of
young
models
at
worst,
but
in
fact,
I
talk
to
them.
What
you
want,
they
can't
think
I.
Also
not
because
I
mean
yeah,
you
clean.
What
do
they
want?
They
shall
say
this
is
the
working
who
drafted
this
have
a
seat.
Please
I
think
this
is
the.
S
It
seems
that
there
is
a
simple
way
to
get
the
sense
of
the
operators
community
than
trying
to
design
a
design
team
of
when
ITF,
which
will
work
on
our
problems.
It's
much
simpler
to
go
out
to
the
operators
and
ask
what
are
their
problems
and
then,
based
on
those
answers.
Try
to
develop
the
solutions.
A
So
we'll
start
talking
immediately
after
idea
how
to
form
the
group
we've
got
number
of
volunteers
and
I
know.
All
of
us
have
our
primary
job.
If
some
of
you
could
spend
five
percent
of
your
time,
contributing
I,
see,
Donald
smiling
would
be
great,
and
you
know
it
would
be
so
useful
for
the
community
and
people
who
pay
our
bills.
A
cent
right.