►
From YouTube: IETF105-NMRG-20190725-1740
Description
NMRG meeting session at IETF105
2019/07/25 1740
https://datatracker.ietf.org/meeting/105/proceedings/
A
We
efficient
in
our
presentation
I
also
ask
the
presenters
to
try
to
be
to
the
point,
and
also,
if
you
have
comments
or
questions,
make
it
quick,
we
try
to
keep
some
time
for
discussion
also
in
the
different
presentation
of
set
of
presentation.
So
this
is
a
animated
session.
One
session
for
this
IHF
week
and
gentleman
sitting
next
to
me
is
Pedro.
Material
Julia
from
NIH
in
Japan
is
one
of
the
energy
Secretary's.
He
was
appointed
earlier
this
year,
but
since
Jo
my
new
co-chair
is
not
here.
We
wanted
also
to
I,
like
that.
A
This
is
the
note
whale,
so
it
it
is
a
reminder
of
the
IETF
Follies
I
mean
the
Archaea.
Follow
the
IETF
policy
answer,
see
the
reminder
of
the
IETF
policies,
in
effect
on
various
topics
such
as
patterns
and
code
of
conduct,
so
I
invite
you
to
have
a
deeper
look
into
these
different
pointers
and
documents
that
you're
aware
about
the
IETF
policy
of
those
topics.
Thank
you
as
usual,
and
we
are
need
to
go
very
quick,
so
thank
speedor
will
be
a
minute.
Taker.
A
I
will
also
assist
him
and
if
you
want
help
it's
on
the
etherpad,
but
we
still
need
a
mythical
scribe.
Anyone
willing
to
just
relay
where
we
are
in
the
presentation-
and
if
there
are
any
questions
for
remote
participants,
is
built
with
which
I
appreciated
yeah
mythical
scribe.
Thanks
will
so
two
blue
sheets.
A
Please
kill
the
blue
sheets
circulate
the
blue
sheets.
If
there
are
new
participants
coming
in
the
room,
it's
very
important
order
to
know
who
is
participating
and
for
the
future
meetings
play
yeah
Jenna.
So
this
is
the
introduction.
I
will
go
quickly
through
a
bit
of
a
research
group
information
and
then
we
will
have
a
set
of
technical
presentation.
A
Essentially,
this
is
centered
on
intern
base,
networking
the
main
activity
of
this
research
group.
Currently
we
have
other
activities,
but
for
this
session
is
essentially
in
some
base
working,
so
we
will
have
Archer
which
will
be
removed,
presenting
on
with
finding
network
intents
for
self-driving
networks,
and
then
we
have
a
couple
of
a
set
of
presentation
which
are
more
on
the
core
in
turn-based
aspects
and
architecture
discussion.
So
an
update
on
some
draft
intern
classification
by
will
IBM
concept,
patch
packet
life
cycle
and
functionality
by
Alex,
which
will
be
removed,
I
hope
so.
A
Yeah
is
here
good
and
then
a
couple
of
presentation,
more
dedicated
to
the
architecture.
Discussion
that
we
are
starting
in
the
research
group
I
will
do
one
of
the
presentation
on
behalf
of
Marino's,
which
is
in
Europe.
So
it
is
right
for
him
and
we'll
we'll
also
do
a
presentation
on
behalf
of
young
I.
A
A
Okay,
we
will
have
to
check,
but
I'm,
not
sure
it
was
not
in
the
planning
that
she
will
do
it
remotely.
So,
a
presentation
on
intern
driven
management
framework,
which
is
a
there,
is
also
draft
on
that
and
then
for
this
set
of
for
a
presentation
to
have
a
H
own
discussion
to
keep
a
bit
of
time
for
discussion
for
this
aspect.
And
finally,
we
also
wanted
to
report
on
some
other
activity
that
are
still
ongoing
in
the
research
group.
Summary
of
AI
phone
network
management
activities.
A
Thank
you,
quick
recap
on
the
meeting
plans,
so
we
are
in
IETF
105.
We
are
a
regular
session
right
now
and
tomorrow,
photos
interested
and
I
think
you
are
aware
that
we
are.
We
have
a
one
day
interim
eating
colleague,
also
workshop,
which
will
be
dedicated
again
to
in
turn
based
it
working.
It
will
be
hosted
in
a
very
close
University
engineering
school.
So
it's
called
equal
the
technology
superior
or
the
memorial
and
to
biasted
fire,
dear
friends,
but
then
the
same
location
as
we
did
last
year.
A
So
we
wanted
to
repeat
the
experience
and
was
a
very
nice
very
nice
version.
In
October
we
will
be
having
an
in-person
interview
in
Roma
in
Italy.
It
will
be
co-located
with
the
network
of
the
future
hydrophilic
conference.
You
cannot
link
to
some
information
about
this
conference
in
the
slide.
The
the
topic
of
this
interim
meeting
will
be
to
go
a
bit
more
into
the
practical
activity
of
the
research
group
for
in
ten
basic
working,
so
to
focus
on
proof
of
concepts
and
tool
for
in
turn,
based
systems.
A
We
would
like
to
have
a
gathering
with
people
working
on
implementing
on
some
of
the
technical
aspect
of
intern
base
from
a
tool
and
I
would
say,
implementation
perspective
and
bring
together
and
see
what
what
can
worked
between
ghosted.
We
will
send
more
details
on
this
internal
meeting.
We
are
preparing
a
call
for
demo
and
it
will
be
absolutely
out.
We
will
provide
more
information
and
you
are
all
welcome
to
come
into
to
Roma
in
October.
A
First
week
of
October,
we've
been
to
be
in
IETF
106
in
Singapore,
and
also
based
on
the
I
would
say
the
output
of
the
preparation
in
October
for
the
intervene
to
include
also
participation,
the
a
curtain.
Again,
we
will
provide
more
information
as
soon
as
we
we
know
bit
more
how
we
would
like
to
participate
in
a
concern
we
are.
We
will
continue
to
run
our
monthly
virtual
meeting,
which
happens
to
be
quite
successful
and
quite
useful.
The
next
one
is
planned
for
meet
end
of
December.
A
The
other
topics
also
describe
their
even
if
they
are
good
less
with
the
detail.
There
is
a
draft
version
available
at
the
link-
it's
very
draft
at
the
moment,
and
we
are
still
working
on
that
with
your
home.
The
idea
is
to
stabilize
that
after
the
summit,
your
comments
and
feedback
are
most
welcome
on
this
new
chapter,
so
we
have
our
session
now.
The
interim
will
be
tomorrow.
B
B
Go
ahead.
Okay,
thank
you,
so
I'm
here
today
guys
to
present
our
work
on
refining
Africa
content
for
self-driving
networks.
This
work
was
actually
published
in
a
sitcom
workshop
last
year.
So
let
me
start
by
introducing
you
guys
on
what
a
self-driving
Network.
Actually
yes,
well,
basically
is
a
private
network.
I
would
ideally
run
itself,
but
to
do
that,
it
would
first
need
some
guidance
on
how
to
operate
and
how
to
behave.
Can
you
go
over
to
the
next
slide?
Please.
B
If
you
go
over
to
the
next
slide,
please
you
can
see
that
the
focus
of
our
work
in
this
is
the
first
part
right
here
that
is
marking
red,
which
consists
in
basically
translating
the
high-level
intent
on
how
the
network
should
behave
into
network
configuration.
So
you
can
go
over
in
our
next
slide.
Please
if
you
now,
if
you
wanted
to
do
that
nowadays,
you
would
have
to
rely
on
maybe
command-line
interfaces
for
interest
for
instances,
instances
or
configuration
using
files.
Maybe-
and
if
you
go
over
to
the
next
slide,
please.
A
B
Ahead,
please,
okay!
So
as
I'm
saying,
if
we
wanted
to
cut,
if
you
use
your
using
conventional
networks
and
you
wanted
to
deploy
an
intent,
you
would
have
to
break
it
down
to
multiple
command
line
interface
commanders.
But
if
you
went
a
higher
level
up
and
you
can
pass
the
next
slide,
you
may
be
used
programmable
networks,
you
would
have
open
flow
interfaces
or
before,
which
is
still,
which
is
very
useful,
but
still
very
hard
to
manage,
hair
to
master
and
very
prone
to
mistakes.
B
That
is
why
many
have
proposed
and
you
can
go
over
the
next
slide.
Many
have
proposed
several
network
programming
languages
configuration
abstraction
such
as
speciation
news,
netcat
and
many
others,
which
are
which
helped
a
lot
in
configurating
networks
and
entwine
intents,
but
still
require
network
operators
to
learn,
complicated
and
master
network
programming
languages.
That
are
very,
not
not
very
easy
to
do
so.
We
asked
ourselves
in
this
work.
Could
we
go
ahead
and
it
would
go
a
step
further
and
use
natural
language
to
deploy
an
intent?
You
can
go
over
to
the
next
slide.
Please!
B
Actually,
you
can
go
to
slides,
and
so
we
proposed
a
network
content
refinement
process
using
nyle
as
an
abstraction
layer.
Nyow
is
basically
a
network,
intent,
definition
language.
We
propose
which
very
closely
resembles
the
English
language.
It
is.
We
designed
it
to
be
very
easy
to
read,
and
we
do
that
because
we
use
nyle
as
an
abstraction
to
ask
for
confirmation
on
the
information
we
extract
from
natural
language.
So
there's
two
main
points.
Thinking
over
two
slides
two
main
points
in
this
network:
intent
refinement
sec,
networking
that
refinement
process.
B
B
Let
me
let
me
describe
this
experimental
scenario.
We
developed
using
our
prototype
and
sonata
NFV,
which
is
a
a
framework
for
deploying
virtual
network
functions
and
over
mininet,
and
so
in
this
scenario,
we
set
up
to
clients
and
iperf
clients
and
in
UDP
traffic
and
I
generally
HTTP
web
client
sending
traffic
to
servers
over
to
open
V
switches,
and
so
you
can
go
over
to
the
next
slide.
In
this
scenario,
if
a
network
operator
wanted
to
maybe
deploy
to
two
virtual
network
functions,
let's
say
a
firewall
and
an
intrusion
detection
system.
B
Okay
and
you
can
pass
one
and
once
again,
and
so
using
our
proposed
in
that
refinement
process,
an
operator
would
only
have
to
reach
a
client
that
could
be
either
a
a
cell
phone
or
a
computer
any
device
with
a
chat
bar
interface
such
as
Google
assistants,
which
we
use
to
develop
our
prototype
and
so
attached
to
Google
assistants.
We
developed
any
entities
extractor
using
dialogue
flow,
which
is
a
Google
proprietary.
B
B
The
possible
end
points
possible
middle
box
as
possible
QoS
requirements,
anything
that
an
operator
might
want
might
want
to
do
to
manage
its
networks,
and
so,
if
you
go
over,
which
in
this
case
right
here
is
a
pyro
and
ideas
and,
of
course,
the
origin
and
the
destination
of
the
intent,
those
extractive
entities
are
then
sent
over
to
a
intent.
Or
later
you
can
pass
one
flight,
which
is
which
is,
in
fact,
a
sequence
to
sequence.
The
learning
model
that
translates
the
entities
extracted
into
a
fully-fledged,
niall
intent.
B
You
can
pass
over
to
the
next
slide,
and
so
here
you
can
see
the
Nile
and
that
generated
from
that
from
that
extract.
The
text,
and
you
can
see
that
it's
very
easy
to
read.
Even
if
you
don't
know
the
syntax,
you
can
understand
what
it's
doing,
and
we
we
present
that
to
the
operator
for
him
to
him
or
her
to
confirm
that
you
can
pass
over
to
the
next
slides.
B
You
can,
and
so
the
upper
if
the
operator
confirms
it,
we
you
can
pass
over
next
slides.
You
can
just
go
passing
over
the
slice
breath
when
the
yeah
we
get
the
feedback
from
the
operator
and
add
it
into
the
our
training
database,
for
the
signal
sequence
learning
model
and
with
that
deployed.
If
that
confirmed
intent,
we
then
pass
it
over
to
an
intent
deployer
which
compiles
it
into
nmv
comments
in
the
final,
wasn't
it
into
the
network
you
can
pass
over
to
the
next
slides.
B
That
is
the
general
concept
of
our
idea
of
our
inter
a
froth
of
our
refinement
process
for
this,
for
this
I
cannot
see.
Okay,
and
you
can
see
there,
there
generated
nsnsv
comments,
and
so
those
two
commands
are
then
deployed
in
the
network
generating
or
final
result,
which
is
in
the
next
slide.
That
is
that
all
the
traffic
coming
from
the
iqrf
plant
now
goes
over
through
to
two
little
boxes
that
the
operator
requested.
B
B
So
we
as
future
steps
this
still
ongoing
research,
so
we
planned
to
develop
Nile
compilation
process
over
to
open
flow
and
p4
backends
and
further
evaluate
this
proposal
in
terms
of
NMR
GE,
maybe
I'm
not
sure
I'm,
not
very
familiar
with
the
process
of
NMR
G,
but
I
can
see
that
I
can
see.
There
are
at
least
three
open
drafts
and
I'm
sure
a
proposal
could
contribute
to
some
points
in
those
drafts.
Maybe,
but
that's
basically
it
thank
you.
D
D
A
E
Hello,
everyone,
this
is,
will
do
from
Poway.
So
basically
we
have
a
very
brief
short
slot
to
present
at
the
update
of
this
draft.
So
basically
we
are,
we
updated
it
to
zero
one
version,
and
next
please
so.
First
of
all,
as
some
of
you
might
not
be
here
in
last
meeting-
so
they
go
of
this
draft
days
to
I
to
achieve
agree,
intent
and
related
terminology
and
classification,
because,
as
you
might
know,
that
in
all
I
mean
there
is
a
lot
of
SDS
are
doing
similar
working
intent.
E
However,
the
content
term
itself
is
kind
of
vague
and
people
are
speaking
with
the
same
term,
but
with
different
meanings.
So
here
we
propose
the
several
ways
that
to
classify
this
intent
into,
for
example,
this,
whether
is
a
user
solution
or
when
to
activate
and
a
based
on
the
life
cycle,
requirement
or
granularity.
Something
like
that
and
the
scope
I
mean
in
this
draft
and
this
scope
is
related
to
the
army,
not
only
the
existing
network,
but
also
they
some
network
in
the
near
future.
So
basically
they
are.
E
For
example,
we
have
Sdn
controllers
and
also
the
network
management
systems
and
some
autonomic
system
and
nodes
next
piece.
So
this
is
the
history
we
proposed
the
very
beginning,
they've
draft
with
actor
in
there
is
a
small
father
in
the
name
and
a
0-0
version
at
last
last
October
and
in
this
era,
before
the
idea
of
meeting
we,
we
updated
it
to
another
0-0,
brilliant
sorry
for
that
and
before
this
meeting
we
update
to
zero
one,
and
so
the
main
update
are
listed
here.
So
we
got
some
feedback
and
comments
from
the
mailing
list.
E
We
addressed
some
of
them,
for
example,
Brian's
comments
about
it,
some
sentences
and
also
we
added
more
detail
the
materials.
For
example,
we
added
a
table
I
will
talk
about
a
little
and
some
more
contributions
were
collected
into
this
document.
Also,
we
did
some
small
modifications,
but
they
are
still
I
need
to
say
that
there
are
still
some
of
the
comments
with
that
not
addressed
by
this
update,
because
we
start
a
new
draft,
because
we
think
that
too
many
complicated
I
mean
contents
are
in.
The
same
document
may
lead
to
some
misunderstandings.
E
So
we
start
a
new
draft,
a
called
architecture
so
which
will
address
the
bottom
two
comments
from
brand
and
Iago
next,
please.
So
this
is
the
table
of
content.
So
basically
we
the
main
updates,
if
you
checked
it
on
the
differences
and
the
mini
updates,
are
in
the
three
main
sections
listed
in
the
red.
E
So
we,
basically,
we
have
three
main
areas
carry
on
PC
and
enterprises,
and
in
each
area
we
have
several
kinds
of
user
for
in
the
machine
carrier
Network
we
have
the
network
operator
and
the
service
designer
service
operators
and
customers
subscribers.
So
that's
it.
That's
the
I
mean
this
basically
show
you
that
in
this
category,
I
mean
me
in
this
way
of
classifying
intent.
How
do
we
do
that?
And
we
also
have
another
several
tables
and
the
lists
to
tell
you:
how
do
we
classify
the
intent
next
please?
E
E
I
mean
not
only
about
an
intent,
but
also
the
network
management,
so
that
you
will
see
that,
for
example,
in
which
I
mean
in
which
period
in
ten
days
are
I
mean
started
and
what
will
be
the
future
for
them,
the
intent,
as
well
as
the
network
management,
so
based
on
our
because
they
are
it's.
Someone
publications
recently
about
this
network
Hutton
made
autonomous,
for
example,
there
is
a
link
here,
is
about
a
white
paper
from
TMF
which
is
recently
published.
It
is
hourly
this
year
about
a
five
level.
Actually,
it's
zero
to
five.
E
There
are
six
levels
of
the
network
self-driving
just
like
what
I
mentioned
in
previous
presentation,
so
indeed
this
white
paper.
Actually,
there
are
several
levels
defined,
meaning
that
if
we
look
at
the
evolution
of
network
management,
so
what's
the
past
was
it
currently
and
what
will
be
the
future
so
in
each
levels,
the
the
way
we
can
show
network
is
different.
For
example,
now
we
can
have
our
centralized
way,
I
mean
to
issue
the
service
to
deploy
the
service,
and
if
we
have
this
intent,
work
realized.
E
So
in
future
we
can
have
this
song
intent
based
way,
alpha
control
and
managed
it's
the
system,
so
we
actually,
we
put,
for
example,
the
in
level
2
and
level
3.
Some
of
the
intent
functionality
can
be
realized
here.
So
next,
please,
due
to
the
time
limit,
I
cannot
explain
the
details,
but
those
are
all
written
in
our
draft
and
the
third
part
is
actually.
This
is
a
contribution
from
petrol.
E
So
basically
we
add
one
small
section
to
discuss
about
the
relationship
with
AI
and
the
intent.
So
if
we
considering
this
will
contain
AI
in
the
intent
based
system,
so
we
we
will
add
an
additional
dimension
of
how
do
we
classify
the
intense?
So,
basically,
here
what
petrol
proposed
is
that
the
degree
of
a
formality
is
an
important
dimension
to
classify
the
intent.
So
this
is
another
I
mean
way
of
classification,
the
next
piece.
E
He
MFM
yep
well,
which
kind
of
intent
what
kind
of
what
classic
class
a
class
of
intent
they
are
talking
about.
So
our
next
step
will
be.
We
want
to
refine
this
document
by
providing
some
specification
for
the
proposed
classification,
so
a
more
formal
definition
and
also
some
prioritization
and
discuss
between
the
overlap
and
also
we
as
suggested
by
Laurent.
We
will
probably
add
a
section
to
describing
the
state
of
ATO
of
intent.
E
For
example,
what
other
stos
are
working
and
which
class
are
they
working
on,
and
also
there's
some
of
the
comments
from
million
list
and
offline?
So
we
are
here.
We
want
to
ask
for
a
volunteer
to
help
to
review
this
document.
This
is
not
a
very
long
one
there's
only
about
less
than
20
pages
and
also
to
contribute
and
also
I
will
ask
for
I
mean
if
the
group
is
ready,
we
will
ask
for
a
adoption,
but
this
is
not
just
for
the
group
this
year.
So
that's
it.
That's
all
my
transition.
Thank.
A
Think
it's
an
important
work
and
I
really
invite
the
research
group
to
provide
videos
on
the
mailing
list
in
order
to
improve
the
quality
of
this
work
and
because
I
think
it's
important
that
we
we
bring
clarity
in
that
space
because,
as
Will
was
mentioning
other
activities
outside
up
in
many
definitions,
many
approaches
to
intern
based
networks
and
so
I
think
one
of
the
the
mission
of
an
emerging
that
space
would
be.
A
good
outcome
will
be
to
provide
clarity,
especially
statical
relation.
A
If
we
manage
to
do
it
well,
it
can
be
a
good
tool
for
others
in
nursery
in
the
research
and
other
standard
bodies
to
reuse.
Our
proposed
classification,
so
I
think
this
is
a
good
starting
point.
I
encourage
this
work
to
continue,
and
also,
with
other
exception,
from
the
aura
of
the
research
field.
Thank
you.
A
F
F
F
That
is
something
which
I
think
we
generally
have
been
struggling
with
a
little
bit
and
we
added
a
few
new
sections
month
in
busy
on
first
order
principles
of
intent
base
networking
a
new
sectional
intent
life
cycle,
the
goals
into
basically,
how
do
house
intent
being
measure
us
intent
can
come
into
being
how
how?
How
is
it
modified
and
what
happens?
It
hurts
a
fourth
and
then,
finally,
also
we
have
a
new
section
on
intent,
research
challenges,
since,
after
all,
we
are
a
research
group.
F
Next,
please
so
the
new
document
structure
or
the
update
document
structure
is
what
is
depicted
here.
Basically,
you
see,
building
we
have
well
really
see.
It
starts
with
the
introduction
of
the
concepts.
What
is
intent
base
the
intent
and
investment
interface
management
and
networking?
How
does
it
relate
to
other
concepts?
And
this
is
followed
by
the
fools
behind
intent-based
networking,
the
life's
a
key
consideration
and
then
the
functionality
that
is
associate
with
intent,
which
includes
not
just
intent,
fulfillment
but
also
intent
assurance
as
well,
and
this
whole
thing
is
energy
concluded
by
the
research
challenges.
F
Next,
please
so
we
have
only
very
limited
time,
so
I
will
just
make
a
skim
over
the
skills
capabilities
very
quickly
regarding
the
intent,
content,
clarifications
and
sending
one
of
the
things,
then
that
we
have
that
we
discussed
that
that
they
also
did
offside
offside,
really,
fundamentally,
what
we
need
to
bring
across
that
intent
is
fundamentally
outcome
oriented.
So
basically
it
is
about
what
outcomes
doesn't
ever
provide
a
expect,
not
how
those
outcomes
are
achieved,
which
is
basically
supposed
to
be
provided
by
the
intent
based
system
and
accordingly,
the
intent
system.
F
Not
the
user
is
responsible
for
determining
how
these
desired
outcomes
can
actually
be
achieved
in
how
should
and
how
they
should
be
translated
into
actions,
policies,
algorithms
and
so
forth,
and
so
from
that
definition
is
just
very
briefly
regarding
the
range
of
through
policy.
One
item
that
was
discussed
was
that
we
believe
it
really
actually
intent.
F
Based
relates
the
policy
based
in
similar
ways
how
a
I
in
machine
learning
relates
to
two
expert
systems
and
inference
engine
and
so
forth,
again,
basically,
meaning
that
it
n
is
defined
by
the
outcomes
not
so
much
how
to
achieve
them
versus
policies
to
define
rules.
The,
for
instance,
events,
conditions,
actions
which
say
what's
to
do
under
what
circumstance
so
yeah
next,
please.
F
So.
Regarding
the
first
order
set
of
principles
yeah,
we
did
affine
if
there's
a
new
section
in
document
on
this.
This
part
a
set
of
principles
define
which
are
certainly
subject
to
further
discussion,
so
here
just
very
briefly
mentioning
them.
So
the
first
principle
is
the
single
source
and
single
version
of
truth
regarding
intent,
this
is
important
to
capture
drift
the
system
and
ensure
the
system
converges,
and
it
here
still
what
was
actually
intended,
as
opposed
to
ability
to
change.
F
Over
time,
second
principal
concerns
one
touch
but
not
one
shot.
This
basically
indicates
that
it
may
actually
take
in
iterations
and
interactions
to
arrive
at
the
desired
attempt,
and
it's
not
necessary.
It
may
not
necessarily
be
specified
and
declared
once
like
you
see
line
or
so
would
be
amazing.
There
will
be
ambiguities
that
it
will
have
to
be
resolved,
unintended
consequences
that
have
to
be
bought
and
so
forth.
Third
principle
concerns
over
I
mean
oversight.
F
This
baby
has
to
do
with
that
violent
in
pain
basis
and
will
conduct
tasks
on
its
own
users
need
to
be
given
the
necessary
tools
so
that
the
to
retain
an
understanding
of
the
current
state
and
why
things
are
happening
that
are
happening
and
fourth
principle
concerns
learning
and
so,
which
means
it's
intrinsically
tied
to
AI.
This
basic
means
that
the
system
should
be
able
to
assess
the
effectiveness
of
its
own
actions
and
be
able
to
improve
in
order
to
optimize
those
outcomes
and
interact
with
dynamic
conditions
are
changing
context
on
the
on
its
own.
F
Regarding
this
with
off
pointing
to
the
details
here,
the
main
thing
is
busy
that
the
tenth
life
cycle
here
is
the
busy
distinguishes
between
three
different
phases.
One
thing
missing
where
the
Tempest
is
defined
or
articulated
by
users
from
the
second
phase,
where
they,
where
that
is
basically
translated
and
rendered
into
into
some
way
that
is
consumable
and
so
forth,
part
by
my
controllers,
operating
support
systems
and
so
forth,
and
then,
finally,
basically,
this
is
followed
by
the
operation
phase
next
slide,
please
so
second
view.
Part
of
a
second
perspective
is
slightly
different.
F
So,
finally,
just
a
small
set
of
intent,
research
challenges-
and
this
is
surely
will
be
an
open-ended
set
of
basically
the
starter
set.
Basically,
there
are
three
things
and
that
document
identifies
what
first
aspect
concerns
intent
interfaces
because
really,
ultimately,
the
the
goal
is
that
the
system
is
able
to
infer
the
intent
of
the
user
as
opposed
to
the
user
having
to
bating
defiance
of
special
language
and
support
that
the
system
would
not
have
said
so.
In
that
sense,
intent
is
different
from
a
yang.
F
F
Next,
one
and
yeah
I
think
the
most
last
step.
So
sorry,
that
was
a
very,
very
quick
run-through,
but
these
are
basically
the
additions
which
are
in
the
government
and
we
expect
a
lot
more
discussion
and
ultimately,
we
hope
that
we
can
at
some
point
request.
Adoption
of
this
is
a
working
group
item
Thanks.
G
A
E
E
Opened
a
couple
comments
to
the
Milanese
about
several
sections,
as
I
said,
I
also
have
some
comments
too,
though.
Those
two
life
cycle
figures,
so,
first
of
all,
as
we
mentioned
that
we
have
a
five
or
six
principles,
however,
in
the
first
figure,
some
of
the
principles
didn't
show
there.
For
example,
we
don't
have
an
urn
function
and
also
the
validation
part
is
kind
of
a
I'm,
not
sure
this
validation
is
what
kind
of
coming.
What?
F
Yes
also,
this
is
a
bigger
response.
First
of
all,
yeah
I'll
respond
to
them,
and
so
the
mailing
list
separately
of
this
didn't
have
the
chance
to
do
so
beforehand
and
we
gotta
get
our
items.
There
were
a
couple
of
items,
one
thing
missing
regarding
alignment
of
the
figures
and
regarding
some
things
that
require
adjustments.
Absolutely
this
is
very
much
basically
still
a
little
bit
in
RAW
format.
F
Further
discussions
about
unity
we
do
intend
to
who
either
align
these
figures
obviously
make
it
clear
that
they
are
really
different
viewpoints
and
showing
placing
different
aspects
regarding
the
validation
one
last
thing
of
the
validation
concerns
and
validating
that
the
network,
probably
that
the
actions
in
network
is
saying
that
they
do
actually
sill
it
here
and
confirm
to
what
was
intended.
So
maybe
it
is
related
to
this
assurance
aspect
of
this.
So
perhaps
this
then
perhaps
we
need
to
work
it
for
now,
but
this
is
just
a
very
quick
response
to
to
this
point.
A
Education
again,
this
is
working
progress,
two
figures:
it
really
reflects
the
fact
that,
among
the
authors
have
been
different
views
that
about
it
stylized.
Just
this,
the
validate
in
this
diagram
is
to
validate
I
would
say
the
correctness
of
the
intent
expression,
the
validated
shrugging.
The
second
life
cycle
is
more
as
Alex
was
mentioning
kind
of
validate
or
verify,
maybe
more
how
well
the
intent
was
actuated
on
the
network,
and
if
there
is
some
drift,
so
to
measure
is
the
difference
between
the
intended
configuration
and
the
actual
computers.
A
E
H
I
Ilx
binoculars
so
I
read
a
draft
this
afternoon.
It's
a
good
start
right,
but
it
seemed
that
there
is
a
focus
on
configuration
right
what
you
just
mentioned
with
the
derive
state.
So
this
is
what
you
call
in
that
not
like.
Okay,
I
want
to
configure
something.
I've
got
an
intended
state
and
I
see
the
applied
state,
which
is
pure
configuration.
I.
Think
like
that,
the
key
that
we
that
we
need
to
focus
on
also
is
the
Asscher
part
and
assure
parts
about
operational
data,
and
this
is
like
the
big
issue
in
intent
right.
I
How
do
we
make
sure
that
it's
not
about
making
sure
your
stuff
is
configured
right,
sure
it's
configured,
fine,
but
it
also
about
trying
to
find
the
operational
data
to
make
sure
it
works.
That
is
up
and
running.
It's
like
in
the
example
with
the
first
example
with
the
firewall
right
great
your
firewall
is
configure.
Is
this
your
intent?
H
Jeff,
so
I
would
like
to
disagree
with
the
derive
state.
It's
not
the
configuration
of
that
it's
the
result
of
configuration
so
if
I
have
lldp
or
BGP
or
any
other
way
to
derive
state
the
states,
that's
going
to
be
compared
to
intended,
not
what
they've
configured
it's
real
validation
of
operational
State,
plus
things
that
I've
learned
that
heaven
configures
its
basis.
F
So
quick
responses
over
I
completely
agree
regarding
the
importance
of
assurance.
This
is
something
that
is
often
missing.
We
will
need
to
make
this
clearer
and
actually,
regarding
the
assurance
I
think
there
are
more
aspects
and
just
making
sure
that
the
system
is
running.
It
is
also
busy
the
effect
to
assess
how
effective
is
a
translation
of
the
of
the
design
outcomes
into
the
system.
So
there
are,
there
are
several
dimensions
of
this,
but
anyway,
I
will
need
to
explain
all
aspects.
J
Maybe
not
so
about
the
outcome,
the
fact
that
its
outcome
oriented,
so
you
say
the
the
presentation
says:
what
outcomes
does
a
network
provider
expect
and
well
I
would
have
expected
that
the
out
that
it
would
be
what
outcome
does
the
user
of
the
intent
expect
because
intends
are
not
intended
only
for
operations?
I
would
rather
see
it
for
because
it's
so
extracted
for,
like
third
parties
ago
sure.
F
So
I
guess
it's
the
question
being
who
the
actors
are
right
when
you
say
the
intent
user,
there
are
different
roles.
I
completely
do
agree
there
that
there
are
different
parties
to
that
and
I
think
some
of
this
will
also
relate
to
the
terminology
or
taxonomy
in
terms
of
that
there
is,
but
one
a
certainly
the
network
or
a
network
provider,
and
but
they
are
going
to
be,
you
know
they
are
going
to
be
other
parties
as
well.
So
take
this
as
analytic
provider.
Okay,
is
it
slash,
slash
user
and
later
on?
F
J
I
have
a
second
question
regarding
the
the
validation,
although
the
user
feedback,
so
right
now
it
seems
to
be
focused
on
did
I
did
I
understand
you
correctly
well,
I
see
that
you
also
have
a
reporting
and
the
phase
do
you?
Do
you
intend
to
to
specify
different
levels
of
validation,
for
example?
First
step
would
be
well.
Did
I
compile
your
well
did
I
understand
you
correctly
validate
and
then
maybe
the
system
would
have
other
feedback
such
as?
Is
it
feasible
regarding
network
capabilities,
for
example?
Yes,.
F
Yes,
all
the
short
answer
is
yes,
there
is
a
section
in
the
document
that
talks
to
functions
or
functionality
that
will
need
to
provide,
if
I'm,
in
a
formal,
intent
based
system-
and
this
is
basic
where
we
would
yeah-
maybe
this
busy,
where
those
yeah,
where
those
things
would
be
accounted
for.
This
would
be
where
is
it
in
section
7,
I,
guess
which
we
didn't,
which
I,
which
I
didn't
go
into
here
in
this
presentation,
because
there
is
deposit,
it
hasn't
changed,
much
bother
versus
the
previous
version.
F
K
A
L
Then
Bogdanovic,
my
main
concern
is
about
the
assurance,
so
how
we
will
do
the
assurance?
Well,
let's
we
have
a
what
is
configurational
data,
what
is
provisional
data,
both
provisioning
data
and
then
when
how
we
will
do
the
translation
and
assurance
that
the
configuration
has
been
translated
in
the
right
operational
model,
because,
even
within
the
same
vendors,
we
have
different
models
that
where
they
have
multiple
different
configuration,
operational
models,
they
don't
they
do
not
match
and
how
how
you
will
solve
it.
L
I
sign
the
next
two
graphs
that
you
did
will
be
present
if
they
are
at
some
price
at
that,
but
that
is
becoming.
You
know
a
significant
problem:
how
to
make
sure
that
this
that
you
can
really
verify.
Oh,
this
was
my
configurational
data
that
change
in
the
provisional
data
because
I'm
getting
some
dynamic
inputs
from
other
systems
and
then
what
I'm
getting
from
the
rest
of
the
network?
F
Yeah,
thank
you
for
this
comment.
I
think
part
of
this
actually
is
actually
I
agree.
This
is
needed.
I
do
think
this
is
actually
where
some
of
the
research
challenges
lie
right,
because
this
relates
to
explaining
that
explain.
If
Isis,
what
you
did
to,
why
why
the
actions
as
about
Mita
are
supposed
to
meet
the
desired
outcomes
and
so
forth.
I
Alex,
but
not
here,
so
the
question
I'm
asking
is:
can
research
solve
that
one?
Because
whenever
you
said
oh,
hey,
we're
going
to
apply
a
is
for
it,
what
I'm
thinking?
Well,
no,
it's
I
mean
it's
something
to
work
right.
You
need,
like
the
domain
knowledge
expertise
to
be
able
to
assure
that
if
you're
covering
blah,
this
one
will
be
assured
by
itself
and
within
the
for
context.
I
F
I
think
the
first
point
is
to
see
if
it
can
be
solve
is
to
articulate
it
clearly
as
a
research
problem
and
then
see
when
then
charge
the
research
community
to
come
up
with
solutions
for
this,
but
I
guess
one
part.
So
so,
first
of
all
for
us
this,
the
challenges
well
not
simply
to
solve
it,
but
maybe
is
first
step
to
really
articulate
what
the
problem
is
and
get
people
working
on
that.
Let's.
A
So
I'm
going
to
x12
talk
about.
In
fact,
the
slides
were
prepared
by
in
Marino's
is
in
Europe,
so
it
could
be
difficult
for
intrude
to
be
up
in
present,
so
I
tried
to
do
the
presentation
on
his
behalf.
I've
been
working
with
him
on
some
of
those
consideration,
essentially
the
first
slide
and
then
model
the
functionality
are
from
Mario's.
A
Here.
What
I'm
going
to
highlight
is
more
some
consideration.
I
would
like
us
at
the
research
group
and
we
develop
our
work
on
the
architecture
for
intern
based
systems.
I
would
say
to
start
the
right
ray
and
to
scope
it
well
in
the
sense
that
we
are
already
having
some
proposal
for
architecture
on
intern
based
systems,
but
we
don't
have
a
clear
program
statement.
First
challenges.
We
start
to
see
challenges
in
some
of
the
draft
and
I
will
say
a
limited
set
of
requirements
will
be
good
as
starting
point.
A
What
is
specific
to
do
it
in
in
10
base
style,
Newton
based
approach
and
not
to
try
to
go
into
the
other
with
the
layers
of
the
system
in
terms
of
design,
principles
and
methodology.
We
are
still
discussing
this
between
Mariners
and
I.
Hope
tomorrow
will
touch
on
that.
A
bit
is
to
put
some
principles
in
design
principles.
A
If
we
think
there
are
new
approaches
or
new
functionality
that
could
be
useful
in
that
intent-based
framework
so
that
we
don't
have
to
redesign
all
the
connectivity
between
the
blocks.
To
just
add
some,
some
subpersonality
order
order
with
order
other
components
and
Integra
bility
or
integration
is
the
fact
that
I
really
believe
the
an
intern
based
system.
We
if
we
design
it
as
a
kind
of
stand-alone
architecture,
completely
out
of
connection
with
other
systems
that
we
are
also
developing
IETF,
but
also
in
other
groups
or
in
the
research
and
industry
products.
It
will
fail.
A
We
have
seen
that
for
other
initiatives.
I
think
we
should
really
that's
all
I
said
we
need
to
focus
on
the
intense
specifics
and
to
think
at
the
design
time
how
we
will
plug
this
into
I,
don't
know
with
nlv,
with
Sdn
with
5g
types
of
architecture
and
also
costs
or
other
types
of
network,
but
we
should
not
put
it
completely
standalone
completely
isolated.
A
Final
point:
consider
service
based
architecture,
approach
and
principles.
This
is
to
highlight
a
bit.
The
previous
point
on
simplicity,
extensibility
excetera,
is
that
service
based
architecture
could
be
a
nice
approach
to
consider
I'm,
not
putting
any
prerequisite
on
that,
but
we
have
seen
recently
more
specification
coming
from
the
standardization
organization
that
use
service
base
architecture.
It's
just
an
open
point.
A
We
want
to
go
this
way
or
not
not
so
critical
at
that
stage,
other
consideration
for
the
design
of
intent-based
network
in
turn
based
systems
stating
first
the
objectives,
simplify
the
configuration
of
complex
infrastructures,
so
a
couple
of
hours,
a
consideration
in
terms
would
be
agnostic
of
implementation
details,
so
the
importance
of
the
right
abstractions
at
the
design
time.
What
should
be
the
lower
boundary
of
an
Indian
architecture
watch
what
it
should
configure.
A
We
can
see
in
the
next
slide
proposal
for
that
and
I
think
this
relates
to
what
I
say,
bent
and
specific,
not
try
to
run
to
make
a
full
management
architecture,
how
Ibn
coexist
with
legacy
management
system
and
emerging
management
technologies,
automatic
decomposition
of
intense
and
continuous
and
firmness
of
intents
are
more
related
to
critical
functionality.
We
aim
for
for
this
system
so
determining
the
lower
Ibn
boundary
again.
A
L
The
am
bogdanovich
my
main
issue
here
is
with
the
management
system.
You
have
to
define
it
most
of
the
management
system
today
in
place
our
element
management
systems.
They
are
not
providing
you
what
you
need
on
the
underlying
you
know
from
the
underlying
management
system.
They
will
provide
you
implementation
level
api's,
which
will
be
very
hard
for
you
to
on
translate
your
Ibn
api's
into
the
device
you
are.
You
will
end
up
as
the
option,
one
and
all
the
issues
here.
So
for
really
to
be
determining
the
lower
lower
Ibn
boundary.
L
We
should
define
what
kind
of
management
system
we
would
expect
the
IBM
to
talk
the
IBM
to
talk
to
most
of
the
when
people
say
management
system
have
an
EMS
in
place,
I
want
to
talk
to
the
EMS
and
then
all
the
again
I
will
mention
again
all
the
implementation
details
are
being
exposed,
and
then
how
will
you
do
it
on
each
different
vendor
in
each
different
software
release
and
into
different
versions?
I
agree.
A
With
your
point
on
the
current
situation
and
I
think
it's
also
I'm
not
in
the
aid
of
of
Marino's,
but
I,
think
he
has
specific
absorptions
for
option
one
and
option
two
which
you
can
see
a
bit
here.
If
you
look
at
the
bottom
part
where
you
have
this
output
of
the
decomposition
logic,
configuring
management
function
when
it
is
aiming
at
management
system
management
function
is
not
thinking
about
an
EMS
type
of
thing,
but
more
abstractions
of
different
capabilities
in
the
network,
and
we
can
wrap
them
as
management
functions.
A
L
A
All
your
points
are
valid,
I
agree
and
we
need
to
meet
make
explicit
our
assumptions,
and
this
is
clearly
open
discussion.
But
if
you
are
referring
to
a
draft
on
architecture
discussion,
this
is
not
the
right
presentation.
This
is
the
next
presentation
from
Keyon,
which
has
a
draft
and
some
diagrams
and
mix
em
up
this.
C
A
For
the
sake
of
time,
I
will
not
go
into
too
much
detail
here.
This
is
a
general
diagram
to
highlight
main
the
main
functions.
Marinos
and
I
are
thinking
about
in
the
functionality
for
the
architecture
and
all
they
relate
a
bit
to
each
other.
The
important
aspect
is
more
on
the
next
slides
to
set
of
slide
to
slide
sorry,
what
are
the
different
functionality
providing
Saudi
offering
to
a
level
technical
requirement
that
drives
the
management
function,
so
this
is
to
capture
core
part
of
the
intent
base.
A
To
also
that
the
previous
comments
for
the
previous
presentation
considering
different
types
of
intense,
if
it's
an
user
or
customer
make
he
will
Express
service
types
of
intense.
What
I
want
the
provider
to
give
me
for
my
business
for
my
business
logic
and
operator,
type
or
infrastructure
type
of
intent,
which
is
more
I'm
a
network
operator
I,
have
to
satisfy
user
intent.
I
can
also
express
operationally
intent
to
say
this
is
all
I
want
my
network
to
react.
A
To
those
intents,
without
going
into
all
the
technical
details
of
provisioning
and
trade-off
on
cost
robustness,
ability
cetera
the
decomposition
logic
block
so
to
resolve
the
intent
to
the
relevant
service
up
to
to
resolve
intent
to
relevant
service
and
select
the
appropriate
management
function.
So
this
is
I
like
what
will
be
the
choice
of
the
mail
with
the
network
management
function,
computation
or
storage
management
function,
and
also
including
the
method
and
techniques
that
will
be
realizing
this.
A
This
decomposition
logic
but
I,
think
we
should
not
in
the
research
group
specification
or
work,
prescribe
what
those
specific
methods
should
be
to
leave
it
open,
and
we
can
make
examples
and
validation
on
to
prove
that
it's
working,
but
not
to
prescribe
a
specific
choices
of
measurement
techniques
or
implementation.
But
maybe
what's
working
well
and
unless
way
law
is
well-suited,
not
work
suit.
So,
in
order
to
go
from
very
simple
techniques
like
mapping
up
to
more
elaborated
reasoning
capabilities,
the
composition
is
facilitated
by
an
abstract
infrastructure
model,
resources
and
communities.
A
What
you
can
see
also
on
the
previous
slides,
to
expose
exposure,
an
exposure
framework
on
the
on
the
infrastructure,
service
and
management
function,
descriptors
to
represent
the
capabilities
of
your
management
functions
and
optionally,
a
knowledgebase
to
store
prior
knowledge
and
best
practices
on
how
intents
in
be
really
I've,
been
actuated,
and
the
performance
of
this
actuation,
going
on
with
the
other
elements
of
the
proposed
architecture
or
for
consideration
in
terms
repository
to
store
and
retrieve
intends
11
and
low
intent
representation.
Again.
A
If
we
in
the
draft
of
the
plantation
from
Alex,
we
have
this
single
source
of
truth.
We
need
to
have
artifacts
to
support
this
this
principle
and
to
be
able
or
unable
to
make
interesting
operation
to
me.
If
we
estimate
that
there
is
a
drift
between
the
intended
and
the
actual
configuration
of
the
actual
state
of
the
network,
this
is
typically
a
tool
that
we
can
use
to
it.
A
A
Consistency:
analysis
to
detect
and
resolve
the
inconsistency
at
the
expression
of
the
intent
and
tradition
to
the
to
the
configuration
specification
errors
competing
goals,
how
to
resolve
those,
sometimes
contradicting
questions
or
ask
to
the
network
and
to
carry
out
this
is
carry
out
during
the
decomposition
phase,
interface
to
the
management
systems
or
selected
management
function
and
execution
parameter.
This
is
typically
kind
of
so-called
software
interfaces
or
IP
is
to
call
the
right
controllers.
H
Just
Answer
a
sacrament
and
from
personal
experience,
the
intent
consumption
interface
seems
to
be
the
most
complicated
one:
she's,
not
something
southbound
of
intent
system
in
composition
of
the
consumption,
so
the
interface
between
business
logic
and
intend
by
system
I
serve
today,
I
think
everyone
who
is
trying
to
build
something
keeps
building
kind
of
number
of
workflows.
Trying
to
do
this,
there's
no
standardization
and
trying
to
connect,
and
then
by
system.
This
business
logic
is
an
extremely
complicated
exercise.
Pretty
much
you
cannot
do
anything
unless
you
try
to
just
hug
things
together.
H
A
Agree
and
I
don't
want
to
put
every
discussion
for
tomorrow,
but
what
I
exposed
to
the
meaning
is
us.
Something
I
would
like
us
to
discuss
tomorrow
is
using
examples
of
so-called
intent,
expressions
and
I,
like
what
other
typical
structural
part
of
this
intent,
capturing
the
the
functional
part.
For
instance.
What
is
the
business
express
in
the
intent
in
all
I
mean
what
supporting
languages
or
logic
we
need
to
apologies.
We
need
to
support
not
making
any
choice
on
languages,
but
what
will
be
key
properties
between?
A
M
C
Interface
to
the
two
to
the
devices
or
to
the
management
system,
I
think
it's
probably
more
complicated
than
that,
because
I
think
there's
certainly
a
third
option,
which
is
interfacing,
two
distinct
Service
agents,
for
example
a
service
agent.
That's
managing
well
address
prefixes,
you
know
so
you're
not
talking
the
device
you're
into
a
function,
which
is
somehow
this
out
might
be
part
of
the
management
system.
But
it's
not
the
management.
A
System
I
agree
with
you.
This
is
starting
point
and
again
I
don't
want
to
distract
myself,
but
this
is
more
proposal
from
from
Marino's.
We
are
still
working
on
that
I'm,
also
working
on
order,
for
instance,
it's
easy
lesson.
There
is
an
framework
that
is
proposed
that,
as
a
bit
more
intermediate
layers
to
to
in
this
kind
of
so-called
management
system,
I
think
ibn
could
be
a
good
addition
to
that.
But
I
agree
with
you
Brian.
Let
me
be
more
than
one
add
those
two
options.
N
Hello,
hello:
this
is
income
from
China
Telecom,
yes,
and
this
is
our
first
time
to
present
our
internship
in
management
framework.
Trapster
and
I
want
to
say
that
the
items
and
the
technologies
will
be
updated.
According
to
the
working
groups
decision
like,
for
example,
our
extractor
later,
we
will
update
all
the
terms
and
the
terminologies
and
that's
the
brief.
N
Exercise
please.
Yes,
thank
you.
Okay,
this
documentaries.
We
are
at
the
premium
to
describe
this
framework,
so
our
chapter,
zero
vision,
have
proposed
our
opera
from
framework
of
the
intangible
framework
and
we
also
defined
some
sample
operations,
including
the
intent
layer,
control,
layer
and
networks
there
and
later
in
the
next
version.
We
we
hope
to
add
a
use
case
of
example,
for
them
pose
as
the
one
to
show
how
to
work
with
this
framework.
Next,
please.
N
N
So
when
the
requester
friends
are
a
player,
including,
for
example,
operator
the
user
or
application
right,
the
cistern,
the
intended
management
module,
this
module
will
first
a
collect
the
intent
request
and
map
out
appropriate
policy,
including
using
the
translation,
module
and
also
verifying
and
also
verify
the
correctness
of
the
configuration
and
then
sorry
previous
the
last
night,
and
then
the
decision
module
with
deploys
a
configuration
through
the
control
layer
automatically.
So
when
the
configuration
is
excuted
in
the
network,
the
verification
module
should
verify
the
validity
of
the
configuration
complimentary
based
on
real-time
status.
N
So
the
effect
needs
of
the
configuration
can
be
constantly
monitored
to
ensure
the
user's
intent
ism
effectively
proceeded,
but
when
an
unexpected
street
using
encountered,
the
decision
module
can
provide
feedback
and
tangle-free
dynamically
to
the
user.
So
this
is
very
complete
a
life
cycle.
Yes,
this
is
the
management
module
in
this
intended
there.
N
The
management
module
will
mainly
have
three
functionalities.
First,
it
will,
it
will
receive
the
intent
of
friends,
a
player
and
it
will
verify
and
the
classify
the
intent.
So
if
the
Internet
is
valid,
it
doesn't
intend
will
be
passed
to
the
translation
module.
So
here
the
classification
function
will
will
do
something
like
the
classification
in
previous
draft
in
will
present
present.
N
Yet
also
the
man
from
the
module
will
need
to
interact
with
the
decision
module
and
may
receive
some
for
the
horn
active
feedback
from
the
teasing
module
when
the
configuration
security
did
not
work
correctly
and
the
management
module
will
also
need
to
regulate
the
life
cycle
of
the
intend
month.
So
this
is
a
closed
loop.
They
will
update
it
or
will
cross
the
loop.
Ok
next,
the
peace.
N
Thank
you.
The
priests
against
this
is
the
translation
module
the
tongue.
Translation
model
need
to
analyze
as
a
semantics
of
the
intent
and
the
formulate
of
the
corresponding
configuration
and
the
outcome
of
the
configuration
plan
to
the
verification
module.
So
the
configuration
module
also
need
to
translate
and
the
Spirit
is
a
somatic
continuing
in
the
intensity
month
and
also
maybe
a
I
can
also
help
to
reference
for
the
translation
module.
Okay.
Next,
please.
N
Thank
you
not
next
to
priests,
okay,
the
verification
module.
Yes,
the
verification
module
may
not
have
tool
functionalities.
The
first
is
to
verify
whether
the
configuration
can
be
executed
in
the
actual
network
environment.
It
worked
elector,
some
real-time
network
of
Staters
and
then
provide
relevant
information
to
the
decision.
Module
of
feedback
and
another
see
is
it
will
validate
a
verification
when
the
configuration
is
actually
actually
is
killed.
Yet
the
network
state
has
a
name,
may
not
change
as
expected,
so
it
is
necessary
to
verify
whether
the
execution
of
the
configuration
works
out
as
expected.
N
N
Next,
okay!
Next,
please!
Yes!
Yes!
Thank
you.
Next
is
a
t3
module
or,
as
I've
just
described,
the
district
module
can
proceed.
The
data
received
from
the
verification
mode
in
the
decide
whether
the
configuration
can
be
delivered
to
the
country
layer.
Actually
that
control
layer
is
a
traditional,
some
controllers
or
network
management
system.
It
can
deliver
the
configuration
to
the
network
element
automatically,
so
the
physical
module
will
decide
whether
to
configure
it
or
not,
and
when
the
users
internally
is
not
implemented
or
the
network
status
is
kaity
and
normal.
N
So
the
district
module
also
need
to
notify
the
intended
management
module
and
the
translation
module
to
make
response.
So
this
is
a
TCP
module.
Ok,
the
next
piece
say
laughter
modular
in
intend
more
intended
layer
is
intended
repository.
It
will
maintain
maintenance
or
mapping
between
the
intent
and
the
configuration,
and
so
it
will
also
interact
with
this
long
station
module
and
mapping
between
mapping
between
different
different
intent
and
the
country.
Read
means
told
it
to
have
translation
module
work
correctly.
Ok,
next
piece:
the
last
tour
is
the
control
layer
and
the
network
layer.
N
N
In
this
framework
and
also
we
have
gathered
some
offline
comments
and
we
were
updated
in
the
next
ability
and
we
also
hope
you
can
join
us
and
if
you
ask
more
feedback
and
more
comments
and
to
improve
the
document,
especially
I'm
intending
examples,
will
you
try
to
define
the
interfaces
modules
and
the
service
model,
the
internal
model
and
the
device
model
into
the
framework?
So
this
is
my
presentation.
Thank
you.
M
Just
give
summary
of
the
activities
we
are
doing
around
in
regard
to
AI
for
network
management,
so
in
the
main
animal
Jesus
unit,
ITF
100
in
Singapore,
we
began
this
activity.
This
is
in
initiative
to
involve
more
deeply
AI
into
network
management.
We
had
a
good
9
stone
at
IDF,
one
fallen
product
and
we
continued
that
with
the
20
meetings,
when
we
achieved
some
advancement,
the
incorporation
of
advanced
AI
external,
in
considerations
on
intelligent
reasoning
and
in
parallel
we
started
an
int
with
a
challenge
project.
It
is
really
different
approach.
M
We
are
trying
to
attract
different
committees
from
from
AI
and
network
management
in
order
to
exploit
as
much
as
possible
this
this
possibility.
So
mainly
in
this
challenge,
we
are
looking
for
right
now.
We
are
looking
for
data
sets
on
use
cases.
So
if
anybody
is
interested
in
participating,
the
million
or
even
the
execute,
you
know
that
the
assistant
to
this
challenge,
please
contact
us
mainly
list.
We
are
welcome.
M
So,
apart
from
that,
we
had
a
side
meeting
in
this
week
and
we
did
a
quick
overview
of
activities
and
it's
a
Eni
and
they
and
how
they
are
related
to
this
AI
enhancement.
So
we
have
this
evolution
from
the
human
based
to
a
a
based
operation,
so
it
is
being
real
right
now,
also
in
Etsy,
so
several
reference
points
were
being
specified,
so
the
net
knowledge
management
context,
management,
cognition
management
situation,
awareness
very
important
and
model
driven
engineering
and
policy
management,
of
course.
M
So
the
requirements
for
this
integration
are
mainly
the
normalization
and
the
normalization
of
this
information
on
intelligence
and
several
prefer
conscious
were
presented.
So
we
have,
in
general,
21
use
cases
in
five
categories,
including
management,
operation,
etc,
and
the
we
have
as
presented
before.
By
will.
We
have
different
categories:
Asia
Network,
intelligences,
similar
to
the
concept
of
ultimate
self
driving
and
in
general
we
are
also
working
toward
the
prediction
of
usual
computing.
That
is
very
important
for
IOT,
especially
when
the
things
know.
M
We
want
to
offload
some
work
from
the
device
to
the
H
or
cloud,
so
it's
very
important
to
involve
AI
here
and
in
general.
We
are
discussing
on
the
interface
and
scheme
of
AI
components
to
network
component,
so
we
are
in
discussion
with
model
training
and
collaboration
at
bend
or
an
operation
side.
I
think
that's
it
for
now.
M
This
is
that
ISA,
this
one
yeah
I
think
it's
next,
please
yeah
so
in
regarding
the
intelligent
reasoning
and
in
it
is
involvement
in
the
AI
for
nervous
management.
In
the
this
lab
last
update,
we
will
work
it
from
the
inner
application
of
AI
to
network
management,
so
we
are
researching
to
what
they.
How
can
we
step
forward
beyond
current
machine
learning
tools?
How
soon
AI
be
introduced
fully
introduced
into
network
management?
How
would
a
I
be
really
exploited
for
network
management?
M
That's
the
key
questions
we
are
trying
to
to
address
in
this
draft
and
also
we
remarked
the
importance
of
external
events,
so
we
have,
for
example,
with
the
definition
of
the
control
plane
that
is
focused
totally
on
the
network
itself,
but
we
can.
We
also
have
the
management
play,
but
it
done
apart
from
the
network,
it
can
also
watch
the
environment
of
the
network
in
order
to
see
what
events
are
important
for
the
behavior
of
the
network
and
react
to
them.
M
So,
in
order
to
do
that,
we
are
proposing
to
exploit
intelligent
reasoning
that
can
reason
they
required
management
actions.
From
this
kind
of
events,
I
am
esmi
business,
so
I,
in
this
update,
we
also
introduced
the
mingotts
and
issues
that
we
are
finding
here,
that
that
different
providers
and
vendors
must
be
able
to
echo
access.
So
we
need
the
methods
to
work
together
directly
or
by
means
of
translator,
but
we
use
the
same
concept.
So
we
need
to
agree
on
some
common
ontology.
M
They
come
said
that
that
will
be
working
together
in
this
reasoning
and
AI
application.
We
have
also
to
to
step
forward
in
the
information
retrieval
that
must
be
a
system
quality
for
for
the
output
from
their
reasoning,
so
it's
very
important
to
assess
and
to
ensure
that
some
company
is
using
some
solution
to
be
sure.
What's
the
quality
of
the
of
the
solutions
decision,
so
in
that
sense,
also,
ontological
conscience
must
be
consistent
to
the
types
and
qualities
of
information
that
I
mentioned
before,
and
the
protocols
used
to
communicate.
M
M
M
That
will
also
include
limited
information
together
with
this
AI
solution
and
in
coordination
with
the
formalization,
for
example,
of
the
formulation
of
intent
that
is
being
discussed
before
so
in
general.
What
I
we
want
to
call
to
the
animatic
community
in
order
to
support
an
enroll
with
us
to
the
design
and
validation
of
this
information,
including
the
protocols
of
interfaces
to
exchange
of
this
intelligence
information.
So
if
anybody
is
interested,
please
they
go
I
know
you
are
interested,
so
please
enjoy
nothing.
D
C
D
Reasoning
I
mean
if
I
have
understood
the
world.
What
you
were
saying.
This
is
something
that,
in
other
cases,
I
have
seen
referred
as
intelligibility
or
traceability
of
the
AI
I
mean
because
it
is
about
understanding
why
things
are
something
right.
For
example,
this
is
not
an
intelligent
reasoning.
You
don't
demand
that
the
thing
is
intelligent,
or
that
reasons
is
that
that
it
tells
you
why
things
are
tapping
it's
more
about
traceability
or
intelligibility
more
than
unreasoning.
The
idea
is
to
do
both
things
at
the
same
time,
work
with
use,
yeah,
yeah.
M
D
M
M
D
C
D
C
D
H
A
Participation,
especially
the
last
session
of
the
first
day,
I've
seen
from
the
presentation,
and
also
from
the
very
good
comments
that
we
are
a
lot
of
work
right
now.
First
I
will
really
invite
you
know
the
the
order
that
may
come
and
Jeff
bring
them
to
the
offers
or
to
the
meaning
list.
It
will
be
very
useful
if
you
can
write
it
down,
because
this
is
something
we
can
much
more
easier
easily
exploit
afterwards
and
also
I.
Invite
you
to
continue
being
active
in
the
research
group.
A
I
think
what
the
point
you
are
mentioning
can
really
make
us
concentrate
on
key
enablers
key
functionality.
We
need
to
deliver
to
really
make
this
in
turn
basic,
working,
more
or
really
key
that
it
is
today.
So,
thank
you
very
much
tomorrow.
Photos
I
want
to
follow
this
discussion.
There
will
be
a
very
nice
workshop
in
ETS
ten
minutes
by
walk
from
here.
Otherwise
we
will
meet
our
September
with
the
ritual
meeting
and
October
for
the
interview,
otherwise
see
you
in
soon.
Thank
you
very
much
nice
evening.
Bike.