►
From YouTube: IETF100-L2SM-20171116-1810
Description
L2SM meeting session at IETF100
2017/11/16 1810
https://datatracker.ietf.org/meeting/100/proceedings/
A
A
A
A
A
Right,
so
our
agenda
is
it's
five
minutes
of
waffle,
five
minutes
of
us
just
describing
where
we
are
with
the
work
and
what
we
plan
to
do
and
then
handing
over
to
zippy
for
45
minutes
to
talk
about
work
in
progress.
The
draft.
What
changes
have
happened,
an
open
issue
that
we've
got
and
then
to
hear
if
there's
any
more
open
issues
from
from
the
room
and
then
we'll
just
wrap
up
with
five
minutes
worth
of
any
other
business.
A
So,
where
we
are
with
the
working
group,
we
met
ITF
97.
So
it's
not
so
long
ago
managed
to
adopt
the
draft
this
year
haven't
had
face-to-face
meetings
at
IDF,
98
and
99
part.
The
design
team,
which
is
really
the
authors
of
the
working
group
document
in
the
chairs,
met
at
both
those
times
have
a
chat,
and
then
we
had
two
virtual
interns
online
to
progress:
the
work
to
discuss
the
issues
that
have
been
raised
and
to
come
up
with
resolutions.
A
It's
it's
a
fairly
ambitious
plan,
which
means
it
it's
quite
tight,
but
we
think
it's
achievable.
So
women
built
the
plan.
It
was
to
start
off
with
the
interim
in
September
which
we
did
and
then
we
took
those
decisions
that
were
made
and
reported
them
on
the
mailing
list.
That
asked
if
there
were
any
issues
and
and
then
Giuseppe
posted
an
update
before
this
meeting
to
fold
in
those
issues.
So
that's
great
and
then
our
plan
was
meet
here,
which
we
seem
to
be
achieving
okay
and
what
I?
A
What
we
want
to
happen
in
this
meeting
is
to
capture
a
sort
of
final
set
of
issues.
So
this
is
not
a
last
call,
but
if
you've
been
sitting
on
some
concern,
this
is
a
really
good
time
to
bring
it
up
so
that
we
get
it
out
and
getting
into
the
version
that
will
go
for
last
call.
So
we
hope
to
see
that
new
version
by
the
middle
of
December
and
then
ask
for
what's
effectively
an
early
young
doctor.
A
So
yon
gave
us
an
l-3,
sm1
and
I
think
as
a
result
he's
on
the
hook
for
ELQ
USM
and
then
we'll
update
and
have
this
ready
by
the
end
of
January,
so
that
we
can
start
and
move
into
a
working
group
last
call
in
February
and
hopefully
be
able
to
pass
onto
the
ad
before
the
end
of
February.
Giving
him
the
opportunity
to
do
an
ITF
last
call
and
get
it
onto
an
agenda,
possibly
before
he
steps
down.
So
he
can
declare
success,
at
least
in
one
payment
of
life.
A
B
B
E
E
E
Okay,
as
a
reminder,
why
mean?
Why
do
we
need
to
define
an
earth
us,
then
so
why
we
need
to
talk
about
young
data
model
or
layer,
2,
VPN,
service,
yeah
work,
as
in
the
know,
so
for
a
tree
with
the
end
services,
though
so
we
have
to
do
that
and
in
particular
this
is
not
a
configuration
model,
but
it's
just
the
communication
interface
between
customers
and
network
operators.
So
in
this
case
we
provide
input
to
automation
and
control
for
the
Sdn,
orchestrate
or
needle
and
is
not
bound
interface
for
espionage.
E
F
C
E
F
E
B
F
E
E
F
D
G
F
E
There
is
the
support
for
Intuit
en
creation,
the
support
on
one
side
belonging
to
one
with
the
N
one
side
belonging
to
multi
VPN.
This
can
be
done,
for
example,
before
you
can
navigate
the
tree
under
side
network
access,
I
access
diversity.
Here
we
can
define
the
group
ID
and
the
diversity,
so
we
can
define
one
side
that
belonging
to
one
VPN
or
to
multi.
At
the
end,
there
is
the
support
for
one
side
with
multiple
logical
access,
also,
the
same
way:
network
access
on
their
access
diversity.
E
We
provide
LG
VPN
with
put
public
cloud
access,
so
under
services
we
can
navigate,
and
so
this
is
a
give
you
a
very
simple
view
of
over
the
lqs
and
service
model
just
to
understand
what
are
the
parameter
and
what
are
the
main
container,
and
that
shall
be
that
can
be
used.
There
is
also
the
support
for
us
requirement
for
me
and
choice.
Policy.
E
F
C
F
E
E
So
under
rich
site
network
access,
we
can
define
an
Ethernet
connection
in
the
top
lane
support,
so
that
is
the
physical
interface
mode,
dot
dot
tunic,
who
don't
want
to
interfere
mode
II
to
the
definition
of
the
Villa
mode,
otherwise
twin
to
twin
end,
and
we
can
define
a
stag
and
sila
seated
on
PE
decays
on
to
entry
mode.
Otherwise
only
a
start
on
PE
for
the
case
of
twin
animal
and
lock,
the
lager
specific
parameters
are
addressed
in
occupy.
E
F
E
We
need
to
specify
also
a
stag
or
only
dot
you
know
to,
and
in
particular
a
stag
is
provided
by
the
network
operator
could
be
this
an
information
that
can
be
considered
for
the
customer,
so
in
case
of
an
and
I
stuck
is
needed,
so
it
can
be
considered,
for
example,
optional.
This
can
be
a
proposal.
If
you
have
another
idea,
please
comment
and
we
can.
C
E
This
is
about
the
multicast
support
in
case
of
multicast
support.
We
need
to
describe
the
multicast
ring.
We
can
see
all
the
parameter
that
we
can
complete.
We
can
set
the
octopus
retired
prototype,
they
even
loaded
Gupta
port
mapping,
so
there
is
the
support
of
both
to
any
source
moving
any
source.
Multicast
was
also
specific
multicast.
Also,
the
idea
is
no
big.
Looping
is
enabled,
and
in
support
in
the
big
domain
and
for
the
Mac
multicast
group
at
side
middle.
There
is
VLAN
ID
that
we
will
an
idea
of
the
multi-touch
group
to
make
address.
F
E
These
right
explain
how
we
can
use
the
single
ohm
to
alone
and
multihoming
support.
There
are
parameters
under
site,
ID
container
that
the
thing
finds
the
way
we
can
achieve
this.
This
is
there
is
the
site.
Vpn
flavor
single
set
aside,
belongs
to
only
one
VPN
and
the
set
of
site.
Vpn,
plural
multi
to
site
belongs
to
Monte
bpm,
and
in
this
case
we
have
the
use
of
access,
diversity
parameter
and
thanks
to
the
access
diversity
parameter,
group,
ID
and
constrain
type.
We
can
support
the
do.
Lumen
movement
question.
B
B
E
Thank
you.
This
is
the
support
for
multi
provider
and
multi
domain
use
cases
particularly
make
a
section.
We
don't
make.
No,
no
consideration
about
the
general
architecture
in
a
multi
provider-
multi
domain
in
this
case,
because
it's
not
in
scope
of
this
work,
but
we
just
say
that,
in
case
of
an
L
and
nine
to
end
our
services
can
be
done
and
can
be
initial
network.
F
E
E
Segment,
one
segment
we
a
segment
for
alpha,
for
example,
if
we
look
at
the
connection
between
site,
1
and
side
3.
In
this
case,
the
assumption
is
that
service
orchestration
layer
have
the
visibility
of
complex,
a
star
topology
and
the
can
compute.
The
separate
can
set
up
the
separate
segment
in
case
of
knowledge.
F
E
Okay,
this
is
another
issue
in
a
USM.
There
are
some
parameter
that
define
signaling
option
and,
for
example,
this
label
control
world
that
are
used
between
the
two
P
or
between
two
sides
of
good.
But
we
are
also
some
parameters
that
are
defined
between
C
and
P,
for
example,
Mac
learning
of
oppression,
and
so
so
here
there
is
an
open
question:
do
we
need
to
specify
signaling
option
parameters,
I
operate
or
phasing
model,
so
what
kind
of
parameter?
E
So
we
asked
the
working
group
to
make
a
review
about
that,
because
we
need
to
check
what
are
the
parameter,
that
network
operator
can
or
sing-a-ling
option
communicate
with
the
customer,
because
the
question
is:
what
are
the
parameter
that
customer
care
about
the
PE
of
the
network
operator?
So
we
have
a.
We
have
an
idea,
but
the
way
we
and
we've
made
a
proposal.
You
have
to
check
all
the
parameters
that
should
be
considered.
B
Yeah
I
guess,
since
all
the
operators
are
the
authors
of
this
draft,
you
guys
probably
have
a
better
understanding
of
what
kind
of
how
you
provision
LT
VPN
service
and
what
is,
as
far
as
the
you
know,
what
is
required
as
in
terms
of
whether
you
need
to
specify
anything
network
related
stuff,
because
there
is
already
lot
of
stuff
in
that
you
device
model.
Thank
you.
B
Needs
to
be
some
translation
or
something
that
needs
to
be
at
the
top
level,
and
then
the
controller
can
give
those
tell
you
some
information
to
provision
the
network,
some
other
questions,
so
I'm.
Just
thinking
aloud
with
respect
to
the
you
know,
the
high
bandwidth
like
500g
pseudo
wire
right
I
mean
whether
would
you
you
know
you.
Could
you
specify
something
like
that
which
would
require
a
CMP
and
all
that
stuff.
E
F
B
E
B
B
E
Thank
you
yeah.
Regarding
your
point
now,
of
course,
as
a
service
provider,
I
can
say
that
the
signaling
option
between
ease
and
peace
maybe
can
be
omitted.
Just
we
can
consider
some
option
and
we
can
just
say
that
the
customer
could
say
about
the
technology
and
the
could
know
just
about
the
technology
and
which
kind
of
singing
option
can
be
used
between
Pease
and
P.
One.
B
More
thing,
just
like
a
sorry
to
interrupt
calendar
service,
I.
Think
I,
usually
you
know
l2
VPNs
are
you
know
you
provision
ones
and
it
stays
there
and
all
that.
But
with
all
this
advent
of
all
this
network
programmability
and
all
that
stuff
there
is,
you
can
support
calendar
service
in
your
service
model
where
you
can
just
schedule
learn
to
eat.
B
The
answer
is
that
you
know
there
where
that's
when
it
will
tell
the
device
to
controller,
to
establish
them
to
eat
the
end
service
and
then
tear
it
down,
and
then
you
know
when
all
the
calendar
based
service,
you
know
what
I'm
talking
about
right.
It
doesn't
have
to
be
permanent
that
once
you
provision
and
it
stays
there
forever
and
ever
it
can
be
any
instantiated
at
any
time
when
user
wants
to
use
it
and
then
tear
it
down.
So
you
just
you
just
pay
for
the
service
used
for
the
time
being,.
A
It's
not
a
lifecycle
management
question,
so
I
think
what
we
tried
to
limit
ourselves
to
was
the
tight
description
of
the
service,
not,
for
example,
how
it
would
be
billed
or
or
when
it
would
be
invoke,
or
even
what
its
operational
status
is,
and
that's
that's
also
what
we
did
in
l3
sm
now,
if
we're
open
to
have
a
wider
discussion
that
we're
probably
not
experts
on
lifecycle
management,
here,
we're
much
more
expert
on
on
protocol
stuff
and
the
service
that
can
be
built
on
those
protocol.
Great.
B
I'm,
just
I'm
thinking
that
the
service
provisioning
aspects
and
combined
with
the
calendar
type
of
services
would
make
it
such
a
lot
more
value.
Add
for
this
type
of
work
right
because
before
it,
you
know,
you
configure
the
device
and
it
just
remained
the
service
forever,
whether
you
are
using
the
resource
or
not,
but
now
that
we
have
the
controller
and
the
network
programmability.
E
C
H
No
cura
so
I
regarding
calendaring.
So
to
me
this
is
a
generic
functionality
that
you
would
apply
to
all
service
models,
or
not
only
the
air
twist,
and
so
it's
a
very
valuable
functionality.
But
it's
not
clear
to
me
whitehead.
We
have
to
be
in
this
model
here
to
me,
this
is
something
that
can
easily
be
separate.
A
C
F
Different
I
want
to
come
back
to
what
was
said
about
scheduling.
You
know.
Yes,
it's
a
it's
an
independent
yang
module.
Yes,
it's
interesting
for
all
the
services
now
here
is
where
we
started
an
issue,
because
that
makes
sense.
But
then
we
go
into
the
service
lifecycle
management
that
discussed
already,
and
then
you
know.
We
always
think
that
we
need
to
make
it
very
easy,
it's
up
and
down
and
then
maybe
not
because
we
start
to
have
like.
Oh
it's
pre-configure,
it's
dormant.
F
Then
you
start
to
have
like
dependency
and
the
type
of
technology
you
have.
If
it's
a
mesh
or
hub-and-spoke
etcetera.
Is
this
like
up
and
down
while
it
depends
of
the
spoke
etcetera
so
somehow
the
school
lifecycle
management
start
to
be
service
specific,
so
trying
to
get
into
a
standard
where
we
try
to
do
the
full
lifecycle
management
of
we
stayed
that
are
service.
Specific
is
mid
year
stretch
and
I'm
fiddling.
Maybe
is
interesting.
Yeah
we'd
like
to
have
a
distilling
again.
Maybe
this
is
in
the
same
type
of
area.
A
A
F
And
by
the
way,
one
more
thing
I
want
to
say
is:
this
is
typically
what
m
EF
is
working
on
with
the
lifecycle
orchestration
management,
so
they
don't
publish
all
of
all
of
them
models
why
they
are
working
on
it.
But
you
know
this
is
such
a
basic
function
in
a
service
lifecycle
that
I'm
pretty
sure
they're
working
on
it.
I.
F
B
So,
just
just
for
my
understanding
purposes,
maybe
I
understood
it
clearly
or
not.
I
will
repeat
so.
The
consensus
is
that
it
is
a
separate
thing
other
than
the
services,
but
aren't
there
any
service
related
implications
with
respect
to
reserving
the
resources
for
us,
so
that
when
you
are
ready
to
deploy
or
aq8
that
you
have,
you
know
you
have
a
greater
chance
of
doing
this
and
so
on
and
so
forth,
but
I
you
know
I
I
just
mentioned,
because
I
think
that's
a
huge
value.
Add
for
doing
this,
but.
B
B
A
Yeah,
so
this
this
was
the
spot
at
which
you
can
say:
we've
been
reading
your
draft
really
carefully
and
we
we're
an
operator
and
we
sell
the
service
and
it
would
be
really
useful
if
we
could
also
do
something
or
it's
not
clear.
What
they've
got
me
in
tears
or
you've
got
a
piece
of
broken
yang
or
or
anything
in
the
issue.
I
tell
you
what
Johanns,
who
have
who
has
read
either
of
the
last
two
versions.
G
Because
this
model
is
from
customer
perspective,
now
I'm
customer
of
operator
I
want
to
create
a
theory
with
some
specific
objective
and
constraint,
but
that
the
customer
do
I,
really
care
that
I'm
creating
a
service
which
is
the
pseudo
wire
base
or
EBP
M'baye.
I
want
to
say
I
want
to
create
a
service
with
bandwidth
of
50
Meg
I
need
redundancy,
I
need
Dualla
homie,
you
go
and
do
it
and
do
I
care
did
service
created,
you
think
VP
a
VPN
or
to
the
wire
or
whatever
you
are
going
to
do
in
the
network.
G
E
But
yeah
I
think
that
customer
may
be
interested
in
technology.
For
example,
in
your
home
you
have
a
fiber
connection
or
you
have
a
standard
standard
connection.
Our
ISDN
connection,
you
care
about
technology
who
who
can
be
used?
Maybe
some
customer
are
interested
also
to
which
kind
of
technology
I
agree
with
you.
Some
customer
could
not
be
interested
to
technology
to
VPN
or
the
PWS,
and
some
other
customers
may
be
interested
II.
G
E
So
it
depends
by
the
customer
in
my
money,
but
I
agree.
We
cannot
dispose
every
parameter.
This
is
the
question
that
I
have
done
before
we.
We
cannot
expose
every
parameter,
good
customer,
because
just
some
technology,
some
few
technology
specific
attributes
can
be
shared
with
the
captain,
because
the
other
attributes
has
to
be
specified
by
the
operator
because
the
core
network
is
owned
by
the
operator.
So
this.
E
G
B
We
this
when
last
time
it
was
presented,
we
had
the
same
common
right
because
there
was
a
lot
of
overlap.
I
haven't
read
the
letters
up,
but
I
see
that
in
your
this
thing
you
have
removed
a
lot
of
stuff,
so
we
do
have
to
read
your:
did
you
guys
post
cross
posted
it
to
the
best
and
the
past
working
group
and
all
that
stuff
for
people
to
review
this
draft,
because
I
don't
remember
seeing
let's
come
across,
maybe
I
missed
it.
I
think.
A
And
said
that
this
latest
couple
of
revisions
haven't
been
cross
posted
that
when
we
set
up
the
working
group
and
when
we
adopted
the
draft
we
did
cross
price,
then
the
working
great
last
call
will
get
cross
posted.
But
you
know
the
point
in
having
a
separate
mailing
list
and
working
group
is
to
do
the
work
in
the
right
place:
occasional
notification,
okay,.
B
Well
now
that
I
know
your
friends,
I,
never
tell
my
thing
all
right.
We
should
just
send
out
an
email
on
saying,
hey.
We
have
all
the
latest
L
to
SM
yang
model
draft,
please
reviewed
in
and
have
a
link
to
that.
So
that
would
help
that.
Will
you
get
lot
more
eyeballs
looking
at
what
is
there
and
you
get
more
comments
right?
B
So
the
other
thing
you
know
what
what
commenter
said.
The
same
thing
we
had
said
before,
because
there
was
a
lot
of
them
and
I
think
you
want
to
see
from
the
perspective
of
a
customer
doing
with
an
online
tool
so
that
it
doesn't
even
talk
to
an
operator
or
anything.
You
provide
some
tool
he
sits
down,
and
this
stuff
is
all
the
thing
in,
on
the
other,
on
the
windows
and
far
away
memorable,
you
know
the
services
provision
no
need
to
do
anything
extra
right,
every
order.
Everything
is
automated
right.
D
Some
comments
about
the
technology
and
tweaking,
as
far
as
you
know,
that
activity
many
folk
songs,
a
larger
scale
of
in
price
customer
generally,
they
have
strong
IT
staff,
there's
an
IT
staff,
so
they
they
care
about
specific
technology
to
the
way
I
am
so
I
think
that
you,
this
would
be
completely
this
model.
Thank.
C
A
So
well,
it's
just
restate
that
and
we've
got
a
couple
of
actions
and
one
issue
that
has
got
a
resolution.
So
there
will
be
an
update
but
and
I
will
send
out
a
ping
now
to
the
Bess.
But
now
is
the
good
time
to
take
a
look
at
this
work
so
that
when
we
get
to
last
call
on
the
next
revision,
we're
not
raising
big
issues.
F
E
F
I'm
wondering
how
do
we
get
this
nice
comment
right,
so
we
have
like
operators
doing
this
work
night
much
ticketed.
Would
you
have
in
your
surroundings,
other
operators
who
could
help
and
review
this?
You
know
I
could
have
like
trying
to
get
some
vendors
to
review
this.
But
in
the
end,
you,
you
know
what
you're
talking
about
right,
your
peers.
Also
there
a
way
you
could
contact
them.
No
don't
what
cast
everything
to
nano:
Nanak,
Mill,
English
right
but
I
personally.
E
And
are
working
on
young
data
model
for
specific
service
and
I.
Think
in
my
personal
opinion,
is
that
there
is
a
need
of
an
activity
and
service
model,
because
for
now
the
vendor
develop
specific
service
model
for
specific
service
of
its
service
provider.
So
it's
important
to
have
a
common
at
you
with
the
answer
his
mother.
So
there
is
a
good
baseline
to
and
also
the
vendor
can
have
a
little
effort
more
one
have
to
to
give
specific
service
enabled
for
a
service
provider,
for
example,
one
under
self
and.
F
E
B
A
A
F
So
actually
I
was
asking
because
whenever
we
did
it
falling
right,
then
we
comes
if
it
great
and
I'm.
What's
asking
because
you
know
exactly
what
you're
doing.
Maybe
you
know
your
peers
directly
right
and
if
you
do,
this
is
great,
send
those
emails.
Now,
if
you
don't
find
people
willing
to
leave,
you
may
be.
F
A
F
A
E
Or
not
is
not
a
service
l1
with
Ian
and
in
particular,
if
we
consider
a
simple
monolithic
architecture
in
which
we
have
the
customer
service,
Orchestrator
and
the
services,
the
n
controller,
while
at
USM
and
address
a
model
northbound
interface
over
the
Sdn
Orchestrator,
the
a1
GSM
is
modeling.
The
southbound
interface,
in
particular
use
case
that
we
want
to
mention
here,
is
the
multi
service
for
internal
customers.
So,
for
example,.
E
Same
service
provider
network,
we
have
a
common
transport
layer
that
shared
the
resources
with
the
multi
service,
for
example,
we
can
have
the
two
service
department
and
we
service
department
that
can
ask
for
resources
to
l1
connectivity
department.
So
this
is
very
useful
for
a
network
operator
point
of
view,
because
in
general
service
operator
and
the
organization
of
service
operator
can
be
done
by
different
department
and
in
particular
for
an
SDN
oriented
architecture.
We
have
also
different
network
controller
that
and
the
SDN.