►
From YouTube: IETF101-NETMOD-20180321-1330
Description
NETMOD meeting session at IETF101
2018/03/21 1330
https://datatracker.ietf.org/meeting/101/proceedings/
A
A
A
A
The
online
agenda,
the
agenda
has
been
posted
online
and,
of
course,
the
it's
also
a
data
tracker
and
tools
page
as
well,
oh,
and
when
you,
if
you
do
approach
the
microphone
to
say
something
be
sure
to
come
to
microphones
to
speak,
the
individuals
that
are
on
Medeco
would
cannot
hear
the
comments.
Unless
there's
you
know
spoken
clearly
into
the
microphones
and
also
be
sure
to
state
your
name
at
the
beginning
of
your
comment.
A
Again,
this
is
session
two,
so
the
chairs
will
spend
ten
minutes
we're
halfway
through
that,
probably
setting
it
up
doing
a
session
intro
or
into
and
then
we'll
have
a
presentation
from
Martin
on
the.
How
were
resolved
in
last
call
comments
for
the
scheme,
amount,
draft
and
then
presentations
from
both
Lada
and
Martin
on
to
proposals
for
long-term
solutions,
and
there
will
be
time
at
the
end
for
open
discussion.
B
A
Okay,
so
quickly
from
a
process
perspective,
the
publication
of
the
scheme
amount
drafts
is
moving
forward.
I
think
everyone
saw
benoit's
email
posted
to
the
net
mod.
Knowing
list
was
it
yesterday.
There
is
a
new
proposal
for
how
to
address
the
last
call
comments.
As
you've
probably
seen
there
was
some
discussion
around
the
zero,
a
draft
and
and
how
to
progress
it
on
what
was
it?
A
And
so
that's
that's
moving
one
quickly,
and
so
there
was
a
0-9
draft
that
was
opposed
to
it.
I
was
it
Tuesday
night
and
that
draft
will
go
through
another
last
call
or
group
last
call
so
that
everyone
has
an
opportunity
to
discuss
it
on
list.
We
hope
everyone
will
be
satisfied
with
the
current
draft
and
I'll
go
quickly
and
then
for
the
long-term
solution
were,
of
course,
in
this
session,
we're
initiating
that
discussion
and
you
met
a
preferred
approach
for
how
to
move
forward
long
term.
A
B
But
if
there's
questions
were
happy
to
entertain
questions
on
process
now
Martin's
going
to
go
through
what
the
changes
are
yeah,
that
should
be
a
technical
conversation.
So
now
is
the
opportunity,
if
you
want
to
have
any
process
discussion,
we're
hoping
there's
none,
but
you
know
we're
open
to
it.
If
anyone
has
any
questions
or
wants
make
any
comments-
and
we
have
the
AR
ad
here
and
we're
ready
to
talk
if
there's
anything
so
we.
A
Have
our
plan-
okay,
good
man-
maybe
the
last
point
I
could
make
is
I
mean.
Of
course
we
are
Act,
we're
good.
The
two-week
last
call,
but
we
also
hope
to
expedite
the
shepherd
right
up
and
getting
into
the
RFC
editor
as
quickly
as
possible,
and
so
that
way
these
drafts
have
been
waiting
in
the
miss
ref.
A
state
can
be
published
as
quickly
as
possible.
C
Okay,
so
this
is
the
presentation
of
the
draft
that
was
posted
earlier
this
week.
It's
all
nine
draft,
so
the
changes
from
the
previous
version
that
was
last
call
is
that
essentially,
we
removed
the
top-level
schema
structure
that
came
from
the
old
young
library
version
and
instead
say
that
every
time
I
have
a
schema,
it's
going
to
be
available
under
the
mouth
point.
So
the
details
are
that
we
remove
the
schema
sub
tree.
We
remove
the
youth
schema
method,
so
we
had
in
line
a
new
schema
before
in
line
works.
As
it
didn't
know,
it.
C
So
in
both
the
inline
and
shared
schema
case,
the
yang
libraries
instantiate
that
are
under
the
mount
point
in
the
shirt
schema
case.
The
same
instance
is
available
under
each
the
same
instance
of
young
libraries
available
under
each
month,
for
instance,
and
the
parent
reference
that
we
used
to
have
in
your
schema.
It
works
just
like
it
did
in
your
schema,
for
the
new
shared
schema
is.
C
So
this
is
a
picture
of
the
new
tree.
We
have
the
prefix
to
namespace
mapping
as
before
the
global
mount
point
list
as
before
the
the
last
case
down
here
is
the
shared
schema
case,
which
just
has
a
container
charge
scheme
and
then
the
parent
references.
So
this
tree
is
very
similar
to
the
money
in
await,
except
that
the
top-level
scheme
is
removed.
C
Okay,
so
one
important
point
here
is
that
this
document
now
allow
allows
any
version
of
the
yang
library
to
be
instantiated
under
the
mount
point.
So
if
you
have
a
pre,
nmda
server,
you're
allowed
to
use
the
old
young
library,
of
course,
for
your
top
level
stuff,
but
also
for
what's
mounted
under
an
amount
point
where,
as
an
Emily
server
can
use
the
new
memory.
C
Ok,
small
point,
but
the
inline
HR
schema
cases
are
presents
containers
rather
than
beliefs,
and
this
is
to
have
some
future
enhancements
possible.
We
make
that
possible
to
future
enhancements
and
another
important
point
that
also
you
mentioned
is
that
Ellen
Ellen
E
and
an
I
documents
they
do
not
have
to
change
significantly.
C
A
Aren't
you
taking
questions
now
sure
I
can't
as
a
contributor,
with
the
shared
scheme
of
being
a
presence
containers
that
necessary
when
it
has
a
leaf
underneath
it
to
non-mandatory
leaf.
So
we've
seen.
C
Because
the
leaf,
it's
actually
leaf
list
and
it's
not
mandatory,
so
you
don't
have
to
have
any
parent
references,
and
so,
if
charge
demos
just
container
you
couldn't
you
couldn't
create
it.
Unless
you
also
created
a
parent
reference-
and
we
don't
want
to
do
that-
ok
or
we
don't
want
to
require
that
there's.
C
Inline
point
containers
that
are
just
empty
leaves:
yes,
Oh
mate.
So
now
we
have
in
the
shared
schema
case,
we
have
the
print.
Reference
may
be
turns
out
that
pant
reference
might
be
a
good
thing,
also
in
the
inline
case.
So
by
having
the
containers,
we
ensure
that
we
could
add
that
in
the
future,
without
breaking
the
module
or
anything
else
and.
C
G
B
We
can
do
a
one
of
the
chairs
there'll
be
two
weeks
to
do
a
full
review
and
any
normal
comments
that
come
up
can
come
up.
Certainly,
your
comment
could
be
considered
a
last
call
comment,
and
it
you
know
we
can
discuss
it
more
on
list
will
be
a
another
version
if
last
call
demands
another
version.
If
lascall
doesn't
demand
another
version
we'll
go
forward
with
zero
night.
It's
really
based
on
the
comments
that
are
received.
B
There
is
a
question
from
the
floor
from
jabber.
The
first
one
is.
We
are
projected
all
this
material
is
already
uploaded
and
what
is
being
projected
is
the
same
as
data
tracker,
but
go
to
session
to
not
session
one
and
then
Joe
last
question
is
their
approach
described
in
Section
six
sufficient
to
implement
the
nmda
and
non
NMDA
case
is
what
is
0-9.
The
document
0
9
sufficient
to
cover
NMDA
and
not
an
MBA
I.
B
So
Joel,
if
it's
not
sufficiently
clear,
you
know,
send
a
comment
unless
it
will
try
to
authors
when
we
try
to
make
it
more.
So
one
of
the
things
we
didn't
say
before,
just
because
I
think
we're
all
just
so
in
the
mix
of
it,
you
may
have
forgotten
to
say
it
on
the
mic
or
in
the
session
is
why
we
are
having
the
second
part
of
the
discussion.
B
The
solution
that
we're
talking
about
here
was
talked
about
before
the
sort
of
breakthrough
that
happened
in
net
conf
in
December,
and
so
we're
trying
to
do
something
why
it
was
discussed
in
the
first
net
mod
session,
which
is
moving
to
a
point
where
we
can
publish
something
that
maybe
isn't
perfect
and
then
immediately
start
working
on
the
next
version
and
by
having
something
published
it
gives
us
room.
It
gives
us
time
to
work
on
the
next
version,
which
we
think
will
be
a
better
long-term
solution.
B
G
Been
one
again
so
for
most
of
the
images
lately
we
clearly
wrote
somewhere
in
the
intro.
This
is
an
MVA
compliance
right.
So,
based
on
your
answer,
that's
what
we
want
to
say
and
I
was
just
checking
that
it's
both
people
and
at
the
end,
MMD
that's
a
kind
of
a
key
statement
right
because
it's
exactly
what
you've
been
writing
in
67
bits
that
we
must
be
and
they're
compliant.
It's
not
the
best
solution.
We
know
that
I'm
going
to
discuss
it,
but
it's
the
key
message.
G
A
G
C
H
F
Good
afternoon,
can
you
hear
me
okay,
so
I'd
like
to
present
quite
a
concrete
puzzle
for
the
future,
but
to
start
with
I'd
like
to
I'd
like
to
present
some
kind
of
general
objective,
but
I
hope
we
can
all
agree
upon.
It
is
basically
what
the
pre
onine
relation
of
the
schema
on
draft
draft
contained.
That
then
became
point
of
contention
during
the
last
few
weeks.
F
So
the
idea
basically
is
to
do
what
the
use
key
mechanism
of
schema
mounted
in
the
generate
narration,
but
to
do
it
in
a
way
that's
integrated
with
young
library
base
so
that
it's
it's
them
much
more
compact
and
hopefully
easier
to
understand.
So
the
idea
basically
to
have
some
possibility
in
yang.
Why
bury
an
optional
possibility?
F
I
must
say
so:
it's
not
necessary
that
everybody
uses
it
so
that
we
want
to
augment
ink
library
I
mean
the
angle
where
based
the
new
version
and
MDA
compatible,
and
so
that
we
can
then
embed
a
schema
anywhere
in
in
the
parents
tree,
which
is
what
we
are
trying
to
do
it
in
the.
U
schema
approach
of
scheme
amount
that
has
been
abundant
in
in
the
version
that
Martin
just
presented,
so
I
hope
that
Martin's
proposal
later
on
will
basically
be
along
these
lines
as
well.
F
Many
aspects
of
my
solutions
are
intentionally
similar
to
yin.
Augments
I
believe
that
this
is
quite
important
and
useful
in
that
it
helps
to
understand
the
concept.
So
if
somebody
only
knows
how
opens
work
in
yang,
it
shouldn't
be
very
difficult
to
understand
how
this
embedded
schema
mechanism
works
and
for
to
makers.
It's
also
important
that
tools
can
reuse
most
of
the
functions
that
must
already
be
there
in
order
to
support
augments
I
checked
my
code,
so
this
would
be
very,
very
easy
expansion.
F
But
this
addition,
this
augment
of
the
new
yang
library,
will
be
its
implementation,
would
be
rather
trivial,
so
in
particular
what
what
it
means
for
this
embedded
schema,
which
can
be
also
called
external
augment,
perhaps
so
what
it
means
in
in
particular,
it
means
that,
as
you
will
see,
the
target
note
of
the
parent
schema,
which
means
it's
the
root
node
under
which
the
embedded
schema
resides,
is
in
this
case
identified
using
a
schema
note,
identifier,
that's
included,
along
with
the
specification
of
the
embedded
schema.
I
will
show
some
examples
later.
F
F
Perhaps
the
schema
mount
solution
presented
earlier
is
that
the
overall
schema
can
be
described
in
a
compact
form,
so
we
needn't
care
about
any
data
stores
like
operation
of
a
the
schema
mount
VM
that
it
had,
however,
is
supposed
to
be
found.
So
it's
really.
It
can
be
also
saved
in
a
file.
So
this
might
really
help
in
the
case
that
Barash
presented
yesterday
about
young
instance,
data
services
may
be
much
more
suitable
for
this
purpose.
We
can.
D
F
Publish
such
augmented,
hang
in
library
data
in
in
an
RFC
in
order
to
describe
more
complex
data
model
and
force
alone
proposes
so
I
tried
to
work
out
two
examples,
one
of
them:
okay,
sorry!
This
is
just
this
shows
what
the
augment
that
I'm
talking
about
really
means.
So
you
can
see
it
here
on
on
the
ask
a
tree.
Basically,
what's
what's
in
both
face
is
the
augment
and
the
rest
is
just
the
normal
young
library.
This
schema
and
I
am
only
showing
the
part.
F
That's
really
augmented,
and
this
is
the
schema
container
in
library
base.
So,
as
you
can
see,
apart
from
the
stuff
that
was
already
there,
it's
the
name
of
the
schema
and
reference
to
a
module
set
for
that
schema.
There's
a
new
container
called
embedded
schemas
and
again,
you
can
see
it's
quite
similar
to
what
we
had
in
schema
mount
before.
F
So,
first
of
all,
we
can
define
some
prefix
to
namespace
mappings
that
we
need
that
we
will
need
for
specifying
the
target
using
the
schema,
no
identifier,
but
also
for
specifying
the
parent
reference
and
that
when,
when
condition,
if
necessary,
these
are
both
some
of
the
target
leaf
is
mandatory.
So
what
this
means
is
that,
apart
from
the
previous
solution
from
scheme
amount,
the
you
schema
I
thought
that
had
Martin
explained
that
separate
list
of
schemas
that
was
removed
now
in
the
0-9
version
of
scheme
amount.
F
D
F
F
D
F
F
D
F
D
F
C
F
C
F
One
that
we
have
now,
yes,
it's
quite
possible
to
work
on
the
existing
schema
month
to
improve
it
somehow
for
the
future.
That's
fine,
but
that's
the
in
line.
Basically,
both
of
the
methods
are
the
former
inline
method,
so
I
believe
that
will
be
one
on
disability
that
beta
model
developers
can
use,
and
this
will
be
another
possibility,
that's
really
much
more
alike,
some
kind
of
an
external
augment
or
an
augment
with
an
external
specified
target
note.
So
so
yes,
so
this
in
this
case,
we
are
just
manipulating
the
schema.
F
F
And
so
the
first
example
I
have
I,
read
it
away
new
schema
that
we
now
have
in
the
schema
one
graft.
So
basically,
the
main
difference
is
here
in
the
bold
face
again.
So,
instead
of
using
some
reference
to
two
young
extension
like
mount
point,
what
we
have
in
schema
mount
it's
explicitly
written
using
some.
So
here's
an
example,
I
just
came
a
note
identifier
again,
we
are
using
the.
F
F
F
The
instance
data,
as
you
probably
remember
from
yesterday
and
the
container
will
be
empty
in
this
case,
because
this
is
the
place
where
we
want
to
embed
other
schemas
and
then
the
schema
part
of
en
this
would
be
like
this.
So
the
parent
module
in
this
case
will
be
that
previous
travel
module
that
contains
only
the
instance
data
container,
and
then
we
define
oh
sorry.
This
is
the
schema
that
we
want
to
embed.
F
D
F
This
trivial
welcome
just
the
container
instance
data
and
then
we
in
an
but
the
the
schema.
That's
called
arbitrary
schema
here
and
the
target
for
this
embedding
will
be
exactly
that
instance
data
container.
So
this
way
we
can
really
achieve
what
mirage
showed
us
today
in
his
examples
and
then,
if
our
to
support
being
all
abilities
and
this
mechanism,
we
can
easily
validate
the
documents,
the
instant
later
container
and
that's
some
basic
arbitrary
data.
I
I
So
in
the
enemy
and
then
I
addressed
today,
you
have
a
mount
point
that
says
this
is
where
we're
going
to
mount
this
information,
whereas
in
this
solution
you
don't
have
that
mount
point
that
disappears
and
instead
you
have
a
young
library
file
that
describes
what
a
server
was
doing.
This
equivalent
thing
would
also
have
seed.
You'd
have
to
put
that
in
the
draft
as
an
example
or
what
is
example,
text
I.
B
Actually
was
going
to
bring
up
a
comment
that
was
made
in
one
of
the
interims
which
is
directly
related
to
that
is
in
this
case
he
still
would
be
a
you
would
have
a
node
that
becomes
a
mount
point.
So
you'd
have
let's
say
a
container.
That
becomes
a
mountain
point.
So
it's
not
that
the
the
base
model
that
it's
getting
augmented
with
with
amount
still
has
it
have
a
node
correct,
or
are
you
gonna
actually
great
to.
B
F
B
B
Make
the
comment
that
was
made
in
the
intro,
where
there's
concern
about
I
think
what
that
would
allow
is,
is
mixing
of
that
node
that
you're
augmenting
could
have
other
children,
and
so
you
could
be
mixing
schemas
together
the
the
base,
one
plus,
the
one,
that's
being
augmented,
and
there
was
some
concern
that
that
might
not.
That
could
lead
to
a
confusion
and
to
added
complexity
and
I
believe
the
net.
The
response
to
that
is
well.
B
That
could
be
the
designers
choice
on
how
to
do
that
and
they
and
I
think
the
counter
to
that
was
in
rather
than
have
a
convention.
It
would
despi
put
better
to
put
it
in
a
restriction
and
that's
why
we
had
enough
where
we
were
today
but
I'm,
not
saying
we
have
to
end
up
at
the
same
point.
But
it's
good
just
to
remember
the
comments
that
were
business.
F
J
F
F
First
sentence
just
summarizes
the
difference
here.
We
use
schema
old,
identifies
where
a
scheme
amount
uses
amount
extension,
and
indeed
there
was
such
concern
that
this
schema
not
identify
us
just
a
low
for
having
the
root
of
the
embedded
schema
anywhere
as
the
schema
mount
or
mount
point
extension
basically
shows
all
the
places
rather
than
this
can
happen.
However,
I
have
three
replies
to
this.
First
of
all,
augments
as
we
know
them
and
love
them,
also
use
keema,
note
identifiers.
F
We
could
possibly
design
augments
in
the
same
way
to
use
some
extension
or
whatever
to
indicate
the
locations
that
can
be
augmented.
We
didn't
do
that,
and
in
Maya
doesn't
seem
necessary
to
formally
indicate
all
places
in
the
the
Augmented
module
that
that
that
may
receive
some
external
data,
as
some
kind
of
augment
points,
so
I
believe,
if
it's
not
necessary
here
either
and
the
situation
is
pretty
much
the
same
and
and
then
regarding
that
extension
or
some
other
tag
that
would
just
label
some
modes
as
candidates.
F
F
In
in
the
scheme,
among
draft-
but
we
are
not
doing
this
here-
this
really
requires
changes
to
to
yang
itself.
So
what
we
are
doing
here
is
really
some
augment
that
with
the
target
not
specified
externally-
and
in
this
case,
even
if
we
had
some
some
extension
indicating
the
candidates
for
schema,
embedding
it
doesn't
mean
that
something
really
ends
up
there,
because
it's
up
to
the
implementer
that
composes
the
young
library
data.
They
tell
whether
this
mount
point
or
whatever
is
used
or
not.
So
it
doesn't
guarantee
that
something
of
is
there
and
or
the.
F
On
the
other
hand,
it
also
doesn't
guarantee
that
some
external
stuff
does
not
end
up
in
other
places
that
are
not
tagged
with
this
extension
or
whatever,
because
we
have
the
good
old
augments
and
the
implementer
again
can
use
augments
to
stuff
an
so
I.
Don't
see
a
really
compelling
reason
to
not
to
permit
it
here.
I
believe
that
some
implementers
usually
want
to
do
something
that
will
work
for
further
devices,
and
it
means
they
will
not
be
doing
silly
things
like
putting
some
stuff
in
arbitrary
places.
F
C
A
comment
about
the
comparison
with
the
mind
movement
is
different,
because
when
you
have
meant
something
you
build
up
to
the
schema
that
you're
met
first,
in
this
case,
where
you
have
Mountain
schema,
they
work
in
some
kind
of
a
jail
situation
where
you
can't
do
references
to
the
rest
of
this
game
at
all.
You.
F
Course,
that's
true,
that's
true,
because
it's
the
modulo
author,
who
writes
an
augment
or
they
by
specifying
the
target
note
the
author
already
knows
where
it
will
be
placed,
so
it
can
just
write
XPath
expression
that
that
just
extend
this
target
note
path
and-
and
so
that's
right.
So
my
point
here
is
that,
from
from
the
point
of
view
of
the
recipient
module,
it
really
doesn't
matter
because
we
don't
have
any
indication
of
possible
places
for
augmenting
in
in
the
recipe
and
modules.
F
So
we
don't
have
it
here
either
for
sure
there
are
differences
like,
as
you
said,
because
here
we
had
to
have,
we
have
to
have
a
complete
schema.
You
can
have
our
pcs
in
the
modules
that
are
being
him,
but
it
so
we
will
have
to
take
some
care
and
maybe
apply
the
principles
that
we
already
had
in
in
scheme
among
so
it's
is
the
friend
book,
but
some
of
the
concepts
are
are
really
similar
to
in
Oakland,
but
I.
J
F
B
F
F
Requires
the
new
version
of
Iain
and
we
have
to
consider
I
think
the
system.
This
is
not
part
of
the
existing
yang
architecture
and
it
may
interfere
in
various
ways
in
other
mechanisms
like
augments
and
so
on.
So
if,
if
if
there
is
demand
for
this,
you
can
look
at
that,
but
this
really
requires
to
have
a
new,
probably
Eng,
2.0,
something,
and
this
this
will
be
much
more
difficult.
I
believe
that
than
the
schema
even
those
key
moment.
F
J
I
very
much
hope
that
by
defining
instance,
data
and
defining
instance,
data
for
a
young
library
that
will
cover
part
of
or
maybe
the
full
of,
design,
time
moths.
You
only
have
to
say
that
this
content
of
the
yang
library,
this
description
of
what
is
mounted.
Where
is
mandatory
for
my
node
or
standard
note
exactly.
F
F
B
F
I
F
B
Blue
is
contributor,
as
and
I
authored
those
documents,
I
don't
think
either
it
would
be
appropriate
in
either
of
those
documents.
Actually,
although
I
guess,
we
could
say
that
it
design
time
it
includes
yang
library
because
that's
a
requirement
of
schema
amount.
So
that's
one
in
one
form
of
design.
Time.
I
wouldn't
want
to
see
any
others
in
that
document,
but
there's
been
other
documents
that
have
come
up
now,
working
groups
that
people
have
said.
B
Well,
maybe
we
should
use
scheme
amount
here
and,
for
example,
if
you
have
a
service
module
and
you
want
to
be
able
to
get
down
to
describing
the
details
of
the
the
network
element
in
that
service
module,
it
might
make
more
sense
to
allow
for
amount
at
that
point
and
if
you
were
going
to
mount
there,
you
might
say
here
are
the
ones
you
have
come
out
and
here's
the
ones
you
may
and
then
the
others
may
be
implementation.
Specific,
okay,.
I
I
B
F
C
In
Oakland
I
mean
you
also
have
in
the
scheme,
I
have
things
like
deviations
and
features,
and
they
probably
know
those
to
the
instance
documenting
in
the
ER
seis,
and
then
we
would
have
to
say
that
the
instance
document
that
you
include
is
not
really
about
the
device
would
have.
It
would
have
something
similar
and
well.
My
real
comment
was
likely
about
what
you
said
earlier.
You
said
or
a
clarification
question.
You
said
this
would
require
the
inversion
of
yang.
F
What
what
you
know
I
mean
what
is
known,
not
this,
but
some
kind
of
implement
of
design,
time
scheme
amount
or
whatever,
because
we
had
in
the
scheme
amount.
We
have
these
three
phases,
so
this
is
about
the
first
that
we
claim
we
do
not
support
Indian
scheme
amount.
This
approach
doesn't
support
it
either
directly,
so
that
it's
not
possible
to
write
in
in
the
parent
module.
What
exactly
should
become
where
what
should
be
embedded,
what
we
use
should
be
pulled
in
so
what
this
approach
does
not
require
any
changes
to
react
so.
C
C
B
Comment
I
was
coming,
I
was
making,
as
you
said,
it
was
that
implied
that
it
might
be
out
of
scope
and
I'm.
Just
asking
the
question
is
just
the
working
group
wanted
in
scope
or
out
of
scope
and
I
think
Martin's.
Suggestion
of
you
know
if
someone
wants
to
come
up
with
proposal
and
they're
more
evaluated
in
context
makes
perfect
sense.
K
J
Actually,
with
instance,
data
design
time
is
so.
What
you
have
heard
here
is
the
argument,
but
do
we
really
want
to
prescribe
this
sorta?
Yes,
that's
a
decision,
probably
not
the
decision
of
the
parent
module,
but
the
decision
of
item
the
routing
work,
separate
RFC
the
instance
data
gives
you
the
tools
to
describe
to
prescribe
a
lot
a
bit,
nothing,
but
it
gives
you
the
tool.
J
I
B
C
It's
very
similar
to
what
Lara
just
described
with
the
maybe
major
change
difference
is
that
we
still
use
the
mount
point
extension
statements,
so
it's
not
open-ended.
You
can
only
do
this
where
the
mount
points
defined,
but
otherwise
and
and
yeah.
You
can
always
only
do
this
where
you
have
the
non
points
defined.
C
So
what
we
have
here
is
that
we
have
the
same
tree
as
we're
having
in
the
currents
it's
given
month
document,
but
then
we
also
have
this.
U
schema
case
that
has
a
reference
up
to
the
new
yang
library,
so
this
proposal
would
require
the
new
version
of
the
young
library
where
we
can
define
the
schemas
in
recursive
manner.
C
C
So
what
I
will
have
two
different
ways
of
doing
things,
and
essentially
what
they
solves
is
that
it
gives
you
a
way
to
have
when
you
know
that
the
scheme
is
the
same
for
all
instances
of
the
mount
point,
you
can
declare
it
here
and
not
use
the
inline
case,
but
instead
have
a
reference
up
to
the
to
this
schema
that
you
have
in
a
normal
young
library.
So,
in
a
way
it's
an
optimization
I.
F
Thought
so,
if
I
understand
it
correctly,
you
intend
it
as
a
replacement
for
the
current
schema.
Mount
yeah,
or
rather
update
to
the
current
I
must
say:
I
would
rather
prefer
to
to
PB
s--,
and
this
was
my
position
from
the
beginning,
as
you
know,
to
keep
this
in
line
and
whose
kemar,
whatever
approach
separate,
not
not
to
combine
it,
because
I
believe
these
are
two
different
two
different
approaches.
F
In
the
inline
case,
you
need
some
instance
data
to
find
the
yang
like
the
the
ability
and
library
in,
and
so
the
only
use
case,
I
can't.
Think
of.
Is
that
there's
a
mouth
point
that
can
be
used.
Alternatively,
for
for
the
in
line
or
in
scheme
approach,
but
yeah
we
can.
We
can
even
combine
that
in
this
approach
and
in
my
product,
I
can
use
the
same
mount
point
as
the
target
for
for
the
embedded
schema.
F
C
Oh
yeah,
so
I'm
just
I'm
a
bit
worried
that
we're
making
the
entire
thing
very
complex
and
complicated
by
having
similar
ways
of
doing
one
kind
of
the
same
thing
and
this
whole
scheme,
a
combination
thing
is
quite
complicated,
and
so
what
should
people
use?
It's
I
mean.
Why
do
we
have
many
ways
are
doing
almost
the
same
thing.
What.
F
F
Schema
only
manipulation
is
really
adds
to
this
confusion.
Confusion
I
believe
if
it's,
if
we
have
two
different
concepts,
it
better
to
have
them
separate
from
separation
of
here's
all
cells.
So
if
we
find
out
that
something
is
wrong
with
one
approach,
we
needn't
change
the
whole
thing
and
influence
people
that
may
be
interested
in
the
other
approach.
C
C
F
C
F
Different
in
the
sense
that
I'm
using
that
schema,
not
identifier
and
the
solution,
uses
still
some
some
expansion.
So
we
can
also
do
it
to
define
another
extension
for
for
this
embedded
schema
approach
and
so
that
these
two
cases
would
also
be
separate
and
and
then
we
can
do
it
basically
without
being
right,
so
that
only
won't
be
combined
under
one
the
an
extension
or
change.
It
would
be
a
good
thing,
but
as
I
try
to
explain
I,
don't
think
that
the
extension
is.
B
H
C
D
B
If
there's
no
more
specific
comments,
we've
been
bringing
a
lot
of
up,
also
and
see
where
it
goes.
I'm,
not
sure
if
we're
gonna
have
any
more
additional
discussion
or
if
we
had
it
already,
but
so
from
a
presentation
standpoint
Martin
is
done
and
now
we're
moved
on
to
the
open
discussion.
So
a
lotta.
If
you
want
to
come
up,
you
can,
if
you
got
to
stay
there,
that's
fine!
So.
I
So
my
comment
is
what
I
see
difference
between
the
two
solutions
is
really
the
question
of
how
heavyweight
is
a
mount
point.
So
if,
if
amount
points
are
heavy
weight
constructs-
and
you
don't
want
to
have
too
many
mount
points,
you
want
to
sort
of
limit
its
usage,
then
I
prefer
a
solution
where
you
have
an
explicit
mount
point
in
the
schema.
So
you
mark
the
places
the
unusual
places
where
you
have
these
map
points,
whereas
if
you
want
mount
points
to
be
lighter
weight
and
be
used,
a
lot
more
than
I
think
may
be.
I
The
solution.
That's
not
suggesting
seems
that
it
sort
of
suggests
that
path
that
you
say
that
the
mounts
have
not
a
heavyweight
thing.
You
expect
to
be
used
more
frequently
more
like
automates,
not
as
frequent
automates
and
then
having
that
flexibility,
maybe
seems
easier.
So
my
my
personal
view
is
I.
Think
mount
point
is
a
heavier
weight
mechanism.
I.
Think
that
often
you
end
up.
Writing
additional
code
potentially
support
those
places.
Where
you
see
the
mount
points.
I
don't
know,
that's
definitely
true,
but
that's
what
I
would
imagine
for
attorneys
and
ni
is
you've.
I
Got
some
code
there
that's
happening
that
and
hence
I
see
him
at
work,
said
she'd
be
a
good
thing.
It's
good
to
mark
that,
and
one
other
comment
that
I
think
you
could
do
here
is
I'm.
A
commission
see
proves
these.
You
can
have
additional
extensions
under
the
map
point
that
are
limited,
whether
you
could
see
information
either
inline
or
virus
schema.
Potentially.
So
in
the
ni
case
you
might
say
actually
I
only
ones
under
scheme
it
to
be
available
up
front
I,
don't
expect
it
to
be
available
dynamically.
So
that's
one
other
benefit.
I.
A
J
B
Think
of
it
this
way
we
we're
at
the
start
of
a
discussion
and
we're
at
a
brainstorming
level
in
it
and
when
we're
brainstorming,
it's
always
good
to
get
out
as
many
ideas
as
possible
and
that
then,
as
proposals
are
made
and
documented,
we'll
call
town
and
see
where
the
working
group
takes
us.
But
this
was
the
start
of
a
discussion.
So
from
my
standpoint,
the
more
ideas
the
better
at
this
point.
F
First
of
all,
that
I
that's
why
I
wanted
to
avoid
the
term
scheme
amount
so
that
it's
some
new
mechanism.
So
we
are
also
complaining
that
we
are
adding
some
new
mechanisms
apart,
that
that
may
be
competing
with
augments.
So
and
in
fact
we
had
all
these
mechanisms
already
in
the
scheme
amount
zero,
eight.
So
we
are
just.
We
are
just
trying
to
somehow
restructure
the
whole
thing
and
and
provide
it
in
a
way.
That's
easy
to
understand
for
for
the
users,
whether
we
succeed
or
not
will
be
seen
what
ya.
B
Know
I
think
if
this
is
the
experiment
that
Benoit
was
talking
about
yesterday
about
being
able
to
finish
publisher,
module
and
then
immediately
start
thinking
about
the
next
revision
of
it.
And
you
know
this
will
be
the
first.
The
existence
proof
of
that
of
see
how
that
works.
I
see
some
people
not
liking.
That
analogy,
but
sorry.
A
Right
so
the
queue
of
Alex's
presentation
it
was
presented
yesterday
in
an
account
working
group
and
at
the
end
of
that
presentation
we
pulled
the
work
group
to
see
if
who
you
know
who's
interested
in
the
workers.
You
know
I
think
of
interest
and
then
would
it
make
sense
for
it
to
be
in
the
neck.
Health
working
group
I,
don't
believe,
there's
any
and
were
raised,
and
then
would
it
make
sense
for
it
to
be
in
the
net
model.
A
J
Have
already
seen
it
and
also
would
have
seen
this
presentation,
so
I
guess
one
of
the
next.
If
you
do
decide
to
transfer,
this
would
be
to
resubmit
this
under
net
mob,
I
suppose,
and
given
that
it's
basically
the
reason
track
of
this
nipple
is
that
the
core
of
this
is
really
actually
yang
data
model
to
define
an
RPC
and
they're
such
basically.
J
So
basically,
what
this
graph
defines
is
in
our
PC
that
it
allows
to
compare
and
in
da
data
stores.
The
reason
is
that,
as
information
can
propagate
between
the
other
source,
there's
a
chance
for
our
differences
and
it's
to
exist
which
can
lead
to
potentially
to
various
a
situation.
It
basically
require
troubleshooting
and
support.
So
how
would
you
basically
detect
that
you
have
such
such
conditions
and
busy
having
this
having
such
an
operation
of
such
an
RPC
helps
with
that
to
detect
issues
anything
from
unexpected
failures
still
time
to
longer
than
expected.
J
J
J
Essentially
when
you
do
this
comparison,
a
wait
for
a
certain
amount
of
time
to
see
busy
if
it
comparison
is
only
fleeting
or
whether
it
persists
for
a
longer
period
of
time.
The
most
conditional
feature
before
about
based
on
the
discussion
that
we
had
on
the
net
contest.
There
is
a
thing
that
was
felt
to
help
that
review,
remove
poverty,
that
is
pretty
moved
to
a
possible
future
extension
extensions
and
the
younger
the
model
has
been
a
updated
accordingly,
and
likewise
they
were
also
suggestions
made
on
the
tool,
have
also
a
possibility.
J
The
possibility
of
an
additional
pre-filtering
step
to
exclude
data
from
the
comparison
that
would
not
be
within
the
scope
of
both
datastore
operational
data
and
preparing
in
conflict,
false
data.
We
preparing
a
configuration
later,
so
you
should
be
basic
excluded
from
the
comparison
so
waiting.
Your
next
steps
there
were
basically
one
thing
deafening
has
been-
has
been
moved
to
get
the
sense
that
this
is
the
right
decision.
Essentially
I
personally
happen
to
think
it's
an
interesting
feature,
but
sure
we
will
we
we
can.
J
We
will
simplify
this
and
the
second
aspect
concerns
that
the
original
discrepancy
should
be
also
added,
so
basically
that,
within
the
differences,
one
should
really
say
also
of
idea
the
original
that
the
data-
that
is
something
that
needs
to
be
added,
and
then
the
remaining
question
was
basically
asked
for
working
group
adoption,
which
is
average
snap
against
the
net
mob
natural.
In
the
den
looking
at
that's
Betty,
all
that
I
have
this.
Video
is
just
a
tree
of
the
of
this
data.
Mods
very
straightforward
comment:
I
can.
I
So
it's
a
question
actually
in
terms
of
the
comparison
you're
comparing
what's
in
intended
versus.
What's
an
operational
I
think
you
may
be
restricted
in
terms
of
how
you
can
make
a
comparison.
I
think
you
can
probably
ask
a
system
to
say:
is
everything
that's
in
intended?
Does
that
match?
What's
that
that
subset
conflict
or
force
confer
true
nodes?
Is
that
those
all
in
operational,
potentially
I?
I
Don't
think
you
necessarily
want
a
better
do
a
full
death
and
the
maybe
confer
the
straightest
to
actually
what
you
actually
what
you
really
want
to
ask
the
system,
so
it
could
be
that
this
configuration
that's
been
I've
been
over
in
some
way
through
system
configuration,
so
I
think
I
haven't
read
the
rates
credit
of
draft,
but
there
may
be
some
nuances
in
terms
of
exactly
what
that
comparison
is
and
what
it
would.
It
makes
sense
to
do.
That
needs
to
be
described
and
documented
who's,
not
ready.
Alright,.
J
J
I
J
C
Martin
Kirkland
I
was
also
going
to
suggest
that
you
actually
maybe
that's
her
intention
who,
this
from
a
future
announcement
that
actually
put
it
into
the
draft.
This
thing
with
just
comparing
the
notes
that
are
actually
in
the
scheme
able
to
a
playlist,
worse
I,
think
that's
something
that
you
put
into
the
draft.
Okay
and
I
also
would
like
to
see
what
what
the
thing
that
Rob
merchant,
where
is
I,
think
it's
more
like
except
subtraction
right.
So
we
wanna
see
what's
intended,
you
want
to
ensure
that
everything
is
intended
is
actually
also
in
operational.
D
C
C
J
That
computer
case
so
I
asked
you
back
I'm
busy
if
you
have
something
that,
if
you're
preparing
intended
and
operational
and
you're
expecting
for
is
that
the
operational?
The
original
of
this
indicates
that
it
came
from
attended
because
it
isn't
in
and
it,
however,
the
original
says
it
is
from
system
but
intended.
That
is
something
that
would
seem
interesting
in
the
condition
that
you
would
have
to
know.
J
C
I
C
J
J
I
So,
just
as
those
are
asking
you,
another
example
would
be,
for
example,
analysis
of
interfaces
where
some
of
them
are
explicitly
configured
and
some
will
create
a
bus
system
that
you
don't
necessarily
want
that
to
flag
up
as
being
a
problem.
The
fact
you've
got
a
loop
back,
zero
interface
created
or
all.
D
I
I
J
F
Thank
you,
I
thought.
I
also
support
this
worker
I
think
it's
it's
really
useful
and
regarding
the
filters
I
would
suggest
maybe
to
start
from
something
simpler
rather
than
inventing
some
complicated
filtering
mechanism.
Perhaps
it
would
be
sufficient
just
to
be
able
to
specify
the
route
both
of
the
trees
that
are
to
become
better,
maybe
depth
of
the
trees.
Just
something
really
simple,
and
then
the
client
can
somehow
combine
the
needs
and
and
get
the
information.
Then
that's
necessary
right.
J
C
C
C
F
And
use
XML,
so
if
somebody
wants
to
use
it
in
rest,
conf
and
in
Jason,
this
would
create
some,
maybe
some
some
complications,
I,
don't
know
as
I
see
it
the
important
news
cases.
It
seems
to
me
that
just
some
really
simple
functions
could
be
enough,
so
that
I
can
repeat.
This.
Compare
I
thought
several
times
to
get
the
information.
C
B
K
D
K
The
interactive
kind
of
paradigm
network
control,
when
the
client
asks
for
several
configurations
that
pulls
some
data
States
analyze,
that
cause
a
heart
disease,
make
some
analysis,
ask
the
more
configuration
so
forth.
So
there
are
many
important
use
cases
where
actually
always
get
data
edit,
config,
color,
PC
kind
of
request
needs
to
be
matched
together.
K
Push
to
the
server
as
the
server
to
trigger
on
a
particular
specified
by
the
client
event
and
and
all
of
them,
based
on
their
conditions,
which
involves
like
a
current
or
historical
data
space
and
the
whole
point
of
this
model
that
we
developed.
This
will
enable
configuration
of
such
ECA
seven
condition
actions
and
put
them
to
the
server
okay.
So
you
see,
basically,
they
have
four
components.
What
is
a
wearable?
It's
it's
a
memory.
K
It's
a
continuous
oil,
keep
intermediate
results,
while
your
conditions
and
reactions
that
which
could
be
used
for
immediate
or
future
condition
relation.
For
example,
if
you
want
to
condition
certain
reactions
based
on
say
medium,
we
know
a
particular
data.
Staple
is
available
the
place
where
to
keep
it
and
to
keep
updated
over
time
and
policy
variables
could
be
just
like
see
where
oppose
could
be
mobile,
shareable
between
multiple
CAS
or
local,
a
specific
only
to
a
particular
CA
and
local,
both
dynamic
and
Static
basic
dynamic,
common
goal
with
each
execution
and
Static
survive.
K
When
is
messes
around
what
what
kind
of
use
cases
you
want
to
consider?
Basically,
we
consider
that
what
we
call
closed-loop,
Network
automation,
say
there
are
events
where
it
is
very
clear
what
to
do
in
case
this
event
happened,
and
it
just
was
all
for
the
client
to
say
you
know.
Whenever
you
see
it
sounds
odd,
lacks
a
favor,
please,
you
know
check
this
data
States
and
do
these
actions
right.
Then
they
all
go.
So
it
is
basically
sometimes
there's
no
time
for
client
network
negotiation,
so
responsiveness
of
the
network.
K
So
you
do
better
if
you
start
doing
start
acting
as
fast
as
possible,
and
it
also
improves
skin
ability
for
the
client,
because,
as
soon
as
the
client
may
delegate
certain
decisions
to
the
network,
it
can
take
care
of
much
more
things.
And,
finally,
those
are
use
cases
when
it
just
free
program
of
logic
and
push
from
the
client
to
the
server
to.
But
if
kind
of
things
like,
for
example,
it
could
be
initiated
and
instructed
to
a
server
to
say
that
in
case
there
is
a
single
or
double
trellis
on
the
network.
K
K
So
the
way,
basically,
it
could
be
either
explicitly
defined
event
or
it
could
be
a
young
push
or
smart
filter
subscribe
to
trigger.
So
we
say
that
the
events
mostly
defined
by
the
client,
the
condition
it's
a
logical
expression
that
conditions
the
logical
expression
which
could
be
either
expressed
as
a
single
XPath
expression.
K
You
are
like
three
minutes
yes
and
the
actions.
Basically
that's
what
important
part
we
only
limit.
Our
actions
to
toe
is
already
impossible
to
do.
Interact
no
leverages
normal
net
confer
pieces
and
also
our
business
defined
by
the
model
partner
by
model
supported
by
the
sewer,
but
nothing
beyond
that.
So
there's
no
danger
to
basically
to
put
some
garbage
beyond
what
could
be
done
interactively
and
when
he
is
executed.
K
B
A
Here
that
would
be
make
sense
for
at
least
for
now.
I
think
continue
to
to
have
that
discussion.
How
large
an
overlap
is
there
what's
the
opportunity,
for
you
know,
bringing
these
two
ideas
together?
Maybe
maybe
there
is
maybe
there's
them,
but
I
think
the
jury's
still
out,
for
me
at
least,
do
you
want.
K
So,
basically,
first
of
all,
but
even
considering
that
we
have
some
overlap,
but,
for
example,
we
can
define
the
same
way
event.
What
event
needs
right.
So
in
either
case
we
can
produce,
for
example,
we
can
share,
but
it
would
be
very
desirable
to
continue
work
on
the
models
at
least
for
now
separately
and
then,
if
it
really
proves
to
be
that
it's
a
pic
of
elavil
combined.
But
we
want
to
avoid
this
in
your
ugly
children
that
sometimes
Parade
news.
When
you
know
decisions
marry
together
and.
A
A
A
B
You
very
much
for
two
very
good
sessions.
Thank
you,
everybody
and,
most
importantly,
let's,
let's
note
that
we
have
an
ad
who
is
has
been
incredibly
supportive
and
helpful
and
that
he
is
finishing
that
role
and
we
have
a
new
ad
sitting
farther
in
the
back
who
we
look
forward
to
continuing
with
that
trend.
So
thank
you
both
and
Benoit.
We
hope
you
will
continue
to
show
and
and
contribute
to
this
group
and
that
it's
just
the
change
of
hats
and
not
the
departure.