►
From YouTube: IETF100-NMRG-20171113-1740
Description
NMRG meeting session at IETF100
2017/11/13 1740
https://datatracker.ietf.org/meeting/100/proceedings/
A
A
A
A
Okay,
so
I
think
we
have
Jefferson
for
Jabba's
cry
and
Giovanni
for
the
mid
ticker.
Thank
you
very
much
dog
eyes
next,
so
in
the
agenda
today,
there
will
be
three
into
two
blocks.
First,
one
is
a
topic
session
on
intern
based
network
management.
We
will
have
two
presentation.
We
have
recently
received
two
draft
discussing
intern
base.
A
Networking
Indian
emerge-ii,
the
first
one
will
be
presented
by
Alex
on
the
distinguishing
intent
policy
and
service
models,
and
the
second
one
will
be
presented
by
Molly
remotely
and
it's
called
concept
of
network
intent,
and
then
we
allocated
a
time
to
go
for
some
discussion
and
next
step
about
this
work.
The
second
part
of
the
meeting
will
be
to
discuss
what
has
been
done
in
2017
on
the
emoji
and
the
future
of
energy.
We
will
go
through
a
bit
of
history
and
get
your
feedback
about
what
to
do
next
year.
So
Alex.
C
C
So
essentially,
it's
basically
is
motivated
by
the
fact
that
they
intend
to
find
networking
is
a
is
increasingly
buzzword
and,
of
course,
that
the
concept
is
you
define
basically
what
you
want,
not
how
to
get
it,
and
it
sounds
of
course
great
I
mean
this.
Is
this
is
very
attractive
really
busy,
but
you
would
want
to
obtain
is
an
operator.
However,
the
question
is:
is
this
idea
really
new
and
actually
really
it
is
not?
There
are
several
things
that
suppose
that
came
before
it,
and
so
the
question
is
basically
so.
C
Is
it
just
a
new
term
for
the
same
concept
or
basically,
is
there
something
different,
and
how
can
we
busy
distinguish
between
between
these
different
items
and
see?
There
are
particularly
two
aspects
of
that
that
or
two
other
concepts
or
two
other
terms
that
send
out
there.
One
is
policy
based
management.
This,
of
course,
has
a
very
long
history
of
320,
plus
20,
plus
D
plus
plus
years
same
idea.
You
defined
as
you
think,
that
the
high
level
and
busy
leave
it
to
different
stages
and
so
busy
render
the
policy
to
do
the
rest.
C
So
you
don't
have
to
worry
about
the
low-level
details
and
the
second
aspect
is
service
models
and
service
provisioning.
We
have
also
similar
things
you're
defining
service
at
a
very
high
level
concept.
This
is
baby.
What
I
want
services
I
want
to
get
and
you
leave
it
to
service
provisioning
systems
and
what
have
you
to
figure
out?
What
are
the
low
level
details?
C
So,
therefore,
basically
you
get.
The
question
is:
is
this
policy
about
certain
amounts
synonyms?
Are
they
do
they
mean
different
things
and
by
all
those
some?
How
do
they
relate
and
is
busy
the
purpose
or
the
intent
of
this
draft,
and
this
falls
of
course,
discussion
that
we
had
actually
also
previously
in
Lisbon
and
then
progress?
C
Actually
this
came
out
of
in
MRP
I
believe
actually
as
well.
Intent
is
defined
as
an
abstract
high-level
policy
used
to
operate
the
network.
So
really.
Basically,
this
has
almost
that
it's
kind
of
like
actually
in
synonym,
and
then
you
have
service
models
which
are
models
that
represent
the
service
that
that
is
provided
by
a
network
to
a
user
or
set
of
users.
C
Okay,
so
anyway,
so
basically
concerning
these
different
terms
are
perhaps
a
few
other
aspects
that
are
worth
mentioning.
That
will
be
useful
to
explain
what
these
things
are
and
how
they
relate
their
basic
two
main
concepts
that
are
important
here
and
those
are
the
concert
and
the
the
concept
of
abstraction
and
the
concept
of
an
information
hierarchy
so
and
basically
relate
to
to
Allison.
C
What
it
means
is
the
lower
level
then
related
to
that
is
the
aspect
of
an
information
hierarchy,
some
of
the
flip
side
of
the
same
coin,
where
you
say
basically,
that
you
have
a
hierarchy
of
higher
of
model,
abstraction
that
basically
become
more
and
more
abstract
and
then
bcat
every
level
you
will
need
some
functionality,
some
logic
it
supposed
to
break
them
down
further
and
there's
the
for
those
that
remember
the
TM
n
reference
model
regarding
many
men
of
telecommunication
men
management
networks.
This
was
baby
built
on
that
right.
C
You
have
busy
at
the
Loews
devil,
late,
very
defining
you
have
the
device
or
the
network
element,
and
you
have
next
several
network,
topologies
and
so
forth.
Then,
on
top
of
that,
you
would
have
services
that
you
could
go
higher.
It's
about
still
some
other
aspects
that
are
perhaps
worth
mentioning.
It's
basically
how
you
define
these
abstractions,
so
you
define
it
in
a
declarative
way,
or
do
you
define
the
procedure
way
and
also,
of
course,
the
rendering
aspect.
I
think
it's
as
mentioned,
and
finally
also
basically
there
are.
C
These
concepts
also
don't
stand
by
themselves,
but
you
have
also
frameworks
that
implement
them
and
pedigreed
and
implement
the
information
hierarchy
in
the
breakdown
of
the
abstraction
and
framework.
Actually,
this
is
muli
is
going
to
present
after
this
Modi
and
Karthik
have
a
draft
that
allude
mostly
to
that
and
basically
taking
the
point
that
essentially,
the
intent
is
simply
what
you
would
be
provided
as
an
atomicity
and
controller
api,
which
is,
of
course,
one
way
of
defining
it
anyway.
So
basically
they
are
in
the
draft,
and
you
look
at
this.
C
There
there's
another
template
terminology
to
trying
to
distinguish
these
items.
So
intent
is
basically
they're,
referring
to
high-level
operational
goals,
whose
precise
mapping
is
non-deterministic
or
unknown
actually
defines
this
policy,
on
the
other
hand,
is
an
abstracted
rule
of
what
to
do
or
what
you
commit
so,
but
you
have
obligation
policies
or
permission
policies
and
also
using
this
page
again.
These
is
also
defined
given
a
set
of
well-defined
events,
conditions
and
actions.
C
So
basically,
it's
not
the
high
level
goal,
but
it
is
M
or
more
rule-based
and
then
finally
basic
service
models,
but
the
high
level
model,
abstraction
that
represents
services,
provide
to
the
end
user
and
defining
also
the
mappings
to
the
component
resources
that
are
needed
to
to
fulfill
our
services,
and
this
is
also
actually
where
you
would
for
which
you
would
use.
What's
not
so
defined.
Things
like
service
function
chains
alright
anyway,
so
so
the
structure
of
the
draft
as
a
pension
is.
D
Some
question
on
this
draft
yeah
told
us
Eckhart,
so
yeah
I
mean
I've
just
seen.
You
know
everybody
whom
I
asked
what
intent
is
come
up
with
a
different.
You
know
answer
and
I'd,
be
you
know
in
the
camp
with
Alex,
mostly
in
terms
of
that,
we
could,
you
know,
scope
it
to
be
smaller
and
which
I
think
other
people
would
would
also
like
I
mean
it
would
be
not
what
currently
I
think
in
the
animal
working
group
were
scoping
or
in
you
know,
7575
intend
to
be
so
that
might
be.
D
You
know
a
change,
then,
in
you
know,
with
respect
to
existing
documents,
but
I
am
not
sure
how
proactive
we
should
do
this
or
just
you
know,
wait
for
for
the
problem
to
boil
up
when
future
work.
You
know
it
gets
done
and
and-
and
the
question
comes
up
right-
so
maybe
keep
this
in
the
back
of
our
mind
and
proposed
to
give
these
answers
when
they're
actually
needed
right,
because
you
know
changing
from
you
know,
existing
ways
on
how
we
interpreted
words
to
tweet
to
use
them
differently
is
always
difficult.
E
E
In
that
case,
intent
was
as
well
focusing
on
connectivity,
but
using
an
intermediate
level
of
virtualization,
but
independently
on
that
it
was
already
focusing
on
how
to
design
an
intent
grammar
and
doing
this
implicitly.
It
was
like
defining
as
specifying
the
scope
of
what
intent
based
networking
is.
C
These
are
aspects
that
I
would
consider
actually
outside
the
scope
of
this.
This
will
also
be
needed
and
required,
but
I
think
there's
other
work
or
be
other
venues
in
suppose
that
will
take
it
up
reanima,
maybe
one
of
them
I'm
I'm
guessing,
but
this
would
be
autos
outside
the
scope
of
this
draft
here,
yeah.
D
If
I
remember
the
my
reading
of
the
draft
I,
don't
think
it
tries
to
captures
and
explain
all
the
different
forms
of
what
people
have
called
intent
yet
and
I
think
that,
for
example,
if
I
look
at
presentations
given
by
Gulen
about
their
intent
based
networking
right,
which
is
more
about
the
structure
of
you,
know
rendering
any
type
of
things
that
you
have
from
you
know
what
you
want
over.
What
you
get
I
think
that's
a
different
form
of
interpretation
that
I'm
not
sure
if
that
is
captured
in
the
draft.
D
C
True
so,
but
this
is
not
captured
there,
but
here
this
summary
is
captured
and
there
are
some
references
given
to
that,
because
I
mean
just
actually
alone
in
the
policy
area.
There
are
whole
treatises
and
look
at
that.
Actually
one
of
the
reference
is
the
one
by
what
was
it
Futaba
who
wrote,
I
think
it's
a
40-page
journalist,
type
of
thing,
elaborating
those
in
great
deal
to
it.
So
clearly,
basically
as
a
draft,
we
don't
want
to
compete
with
that.
But,
of
course
we
want
to
have
at
a
minimum.
G
G
We
are
all
so
excited
about
this
area
of
network
intent,
and
we
want
to
take
a
slightly
different
view
from
what
was
presented
by
Alex
former
colleague
good.
So
what
we
wanted
to
do
is
explain
some
of
the
concepts
and
at
least
provide
some
sort
of
you
to
our
thinking
in
terms
of
next
steps.
So
here
is
the
next
canopy
goes.
The
next
slide,
please
yeah.
These
are
all
the
things
that
I
want
to
be
one
of.
G
First
of
all,
this
is
joint
work
with
my
colleague
Karthik,
so
who
is
also
there
in
the
meeting
room,
so
we
wanted
to
define
what
is
Network
intent
at
least
what
we
are
thinking
and
as
the
Alex
was
pointing
out.
There
is
a
lot
of
confusion
in
terms
of
I
mean
the
lord
of
literature.
There
is
confusion
between
policy
configuration
for
all
that,
so
we
want
to
at
least
make
some
distinctions
between
what
those
are
and
then
we
have.
G
G
So
this
is
in
order
to
say
what
you
are.
The
slide
is
sort
of
explaining
what
we
are
not
for
example,
so
there
are
several
words
in
this
topic
of
network
intent.
One
is
as
Alex
residency
the
economic
Network
intent
that
talks
about
distributed
computation
on
network
devices
and
Paul
is
distributed
to
the
network
devices.
G
I
mean
then
I
mean
so
maybe
that
might
be
the
step
zero
or
something
but
I.
Think
from
a
research
perspective.
We
we
thought
it
should
be
important
to
have
a
much
higher
level
objective
than
the
last
one.
Let
me
go
to
the
next
step
slight
piece,
so
our
now
that
explained
what
we
are
not
here
is
what
we
are
thinking
in
terms
of
the
architecture.
G
So
an
administrator
can
articulate
what
is
the
desired
outcome
from
the
network
in
rough
terms,
and
it
does
not
have
to
be
very
specific,
and
that
can
be
ambiguity
in
that
in
the
desired
outcome,
and
so,
which
is
what
we
meant
is
the
network.
Intent
need
not
be
prescriptive,
it
can
be
declarative
and
a
given
intent
can
be
realized
in
multiple
ways.
There
need
not
be
to
the
point
that
I
was
saying:
if
an
intent
is
there
is.
G
If
there
is
a
one-to-one
mapping
between
configuration,
then
I
think
I
mean
there
is
possibility
to
do
lot
more
than
that.
It
becomes
a
very
restrictive
approach.
I
think
and
the
reason
why
there
can
be
multiple
approaches
is
that
the
functionality
and
the
capability
of
the
network
might
be
different
to
realize
and
devices.
G
So
that
can
be
seven
days
to
do
that
and
that
intelligence
has
to
be
somewhere
else
and
also
when
you
are
realized,
when
you
want
to
realize
intent
in
a
network
that
can
also
released
lead
to
conflicts
and
in
the
sense
that
suppose,
I
haven't
intended
number
seven
that
I
that
was
implemented
yesterday
and
today,
I
come
in
and
ask
something
about
in
1014,
but
the
question
is
to
realize:
14
in
ten
seven
has
to
be
deleted
or
something
so
there
are.
There
are
some
issues
in
terms
of
conflicts
and
how
do
we
adjust
them?
G
So
these
are
all
some
of
the
things
that
we
are
thinking
about,
how
you
know
I'm
just
this
this
whole
topic.
What
are
the
problems
are
not
solution
right
now,
that's
what
it
is
and
as
I
mentioned
in
the
slide,
there
is
a
confusion
everywhere
in
Cisco.
You
know
in
in
Cisco,
live
in
either
where
customers
are
talking
about
policy,
service
models
and
configuration
and
working
tend
to
be
the
same,
and
that's
at
least
not
our
at
least
the
goal
that
we
are
thinking
of.
G
G
A
network
administrator
sits
at
the
top
and
then
activates
an
intent,
and
let's
say
as
Alex
was
mentioning,
has
an
interface
with
is
the
in
controller
and
then
the
Israel
controller
sort
of
talks
to
an
intent
engine
and
the
intent
engine
resolves
that
intent
into
into
some
suitable
instructions
to
network
by
the
way,
I
am
not
using
the
word
policies
or
configurations,
etc,
and
those
who
are
realized
on
the
network
devices
eventually
too,
for
that
to
run
on
a
network
device.
It
has
to
be
configurations
right,
but
some
you
know
there
are.
G
There
are
things
that
would
have
calculated
at
the
intent
engine
and
then
trap
spirit
to
the
devices,
and
then
the
devices
run
those
configurations
and
then
and
the
point
that
I
made
in
terms
of
conflict
resolution.
All
those
things
need
to
be
started
out
before
realize
can
be
successfully
realized.
G
Can
we
go
to
the
next
so
I?
Had
there
a
couple
of
use
cases
and
I
made
that
pretty
vague
in
the
sense
of
you
know,
in
the
sense
that
there
is
enough
ambiguity
in
the
in
the
outcome.
For
example,
one
could
ask:
let's
say
you
have
a
large
network.
What
are
the
congested
in
the
network?
What
is
congestion
me?
Congestion
means?
Is
it
one
packet
loss
thousand
packet
loss
or
you
know,
link
utilization?
G
Eighty
percent,
so
I
mean
it's
not
a
I
mean
congestion
is
a
pretty
ambiguous
word
and
that's
one
intent
and
then
I
walk.
You
know
the
the
first
one
is
a
classical
thing
from
you
know,
for
some
folks
will
be
familiar.
I
want
to
be
is
a
call
from
user,
a
to
user
B
user.
A
has
some
privileges,
you
know
an
important
person
company,
so
in
that
sense,
they're
half
bandwidth.
G
That
needs
to
be
guaranteed
into
n,
so,
whereas
the
intent
is
very
small,
but
that
can
be
exploded
into
several
types
of
actions
that
will
happen
in
terms
of
bandwidth,
privileges
and
all
that
stuff
and
Inter
and
how
much
performance
the
network
has
to
supported
all
that.
These
are
all
some
simple
use
cases,
and
we
can
definitely
add
a
lot
more
to
this
and
that's
one
area
that
we
are.
We
could
potentially
seek
in
a
Margie
to
provide
some
guidance
in
terms
if
there
are
other
use
case
that
we
could
think
of.
G
Let
me
go
to
the
next
slide,
so
here
is
what
we
want.
What
what
are
we?
So
we
have
some
idea.
We
are
in
the
early
problem
definition
stage
and
there
are
some
ideas
in
terms
of
solution,
approaches
that
we
are
thinking
about
right
now,
as
we
mentioned
some
time
back.
There
are
multiple
approaches
that
can
be
pursued.
The
economic
networking
perspective.
That's
clearly
one
approach,
the
domain-specific
language
based
techniques,
those
are
also
there
are
folks
in
the
research
community
who
are
actively
pursuing
on
this
topic.
H
H
G
G
I'm
not
able
to
so
the
the
point
I
wanted
to
say
was
the
intent
is
very
ambiguous.
Let's
say
when
the
when
the
user
administrative
expresses
his
intent
congestion,
the
network
can
be
certain
things,
as
I
mentioned,
that
can
be
very
small
packet
loss
may
not
be
admissible
for
some
applications,
whereas
in
some
other
applications
you
know,
as
I
said,
thousand
five
thousand
packets,
etc.
That's
not
accept
or
link
utilization
are
various
things.
G
So
how
do
you
realize
that
intent
that,
since
there
is
a
little
bit
of
ambiguity
from
the
intent
to
what
sort
of
how
do
we?
How
does
the
network
react
to
this
of
integer?
So
that
was
the
point
I
wanted
to
make?
But
if,
if
the
question
from
a
network
management
point
of
view,
if
you
know
if
I
could
have
formulated
the
in
turn,
the
other
way
like
give
me
the
links
in
the
network,
if'
in
octet
is
greater
than
some
number
or
something
that
would
be
a
very
specific
problem.
Folks
have
solved.
G
Have
cases
some
of
the
ones
that
is
even
mentioned:
the
autonomic
networking
RFC
7-5
7-5
in
terms
of
V
M
moment,
and
all
that.
Actually
there
are
some
startup
companies
in
this
arena
and
what
Felicity
and
one
of
them
in
a
base
in
Bay
Area.
There
was
one
intent
in
terms
of
suppose
you
are
given
thousand
VMs.
G
Can
you
build
a
data
center
level
to
the
l2
design,
sort
of
data
center
I
thought
such
an
intent
possible
and
we
need
to
have
smaller
intents
before
we
can
actually
go
to
sort
of
having
a
much
larger
intent
in
terms
of
designing
a
data
set.
So
that
was
the
reason
why
we
thought
we'll
have
simple
initial
vanilla
use
cases
before
we
can
realize
much
large
at
one
thing.
I
Probably
due
to
my
sharing
neurons
and
the
fact
that
I
have
not
read
the
draft
but
from
the
presentations
and
not
these
two
documents,
somehow
addressing
the
same
issue
or
is
that
the
presentations
are
different.
I'm.
Sorry
about
that,
because
Virginia
was
about
the
previous
one
and
this
one
we
did.
You
seem
to
be
addressing
the
same
problem,
trying
to
define
what
intent
is
trying
to
put
it
in
the
framework
of
other
definitions
etc,
which
are
the
planning
of
the
authors
and
the
chairs
about
this
curious.
A
A
So
now
we
have
a
bit
of
time
to
actually
discuss
about
this
document,
but
majin
Larry.
We
asked
chairs
some
time
ago
initiated
this
topic
in
the
energy.
Ask
some
people
people
had
volunteered
to
address
some
aspect
of
developing
this
work,
so
it
lies
to
the
question
from
Diego
to
say
what
are
we
trying
to
achieve
with
those
documents
and
in
the
research
group
with
this
topic?
What
could
be
the
next
steps
in
terms
of
things
we
need
to
address
in
the
research
group
to
develop
this
work,
so
this
is
really
an
open
discussion.
A
We
have
currently
these
two
documents
that
are
very
I
mean
in
an
early
phase,
and
the
question
is
up
to
the
research
group
to
say
we
think
we
need
to
continue
on
this
aspect
or
these
aspects.
So
if
there
are
any
one
willing
to
comment
on
the
Mac,
we
will
receive
that.
We
have
also
a
bit
of
aspect
we
would
like
to
convey
with
Cassandra
we
like
to
give
you
the
floor.
First.
J
Giovani
aside
the
end,
it's
not
read
exactly
about
evolution.
There
is
more
a
philosophical
question,
there's
also
the
other
research
group
on
network
measurements
and
I
think
somehow
there's
another
life
also
of
the
network
management
working
group.
So
it's
something
like
maybe
just
work
groups
who
work
closely
in
the
future.
J
I
don't
know
I
mean
it
could
be
beneficial
for
both
sides,
I
guess
to
bring
more
measurements
into
this
working
group
as
well
to
helping
them
manage
management
approaches,
it's
just
an
idea
and
don't
try
I
get
your
point
you
were
talking
about,
so
that
there
is
the
map
RG,
the
measurements
analysis
for
a
core
research
work
can
group
in
they
have
some
sort
of
overlap.
This
to
grow.
Working
groups,
overlap
in
the
network,
commit
and
manage
one
and
I'm,
not
sure.
J
A
H
Dan
McDonald,
so
it
will
be
interesting
to
see
the
network
management
workflows
from
the
operators
and
essentially
saying
this
is
how
they
are
solving
certain
of
their
daily
operational
problems.
For
example,
what
would
be
the
workflow
for
peering,
Zoar
adjacencies,
or
you
know
how
do
they
translate
the
services
into
the
existing
logical
and
verge
and
logical
and
physical
topologies?
H
There's
a
lot
of
conversations
about
trying
to
replace
a
chassis
router
with
a
collection
of
physical
devices
connected
in
different
physical
topologies?
You
know
some
type
of
a
clause
how
to
calculate
the
specifying
I
need
the
capacity
or
the
throughput
and
then
calculate
based
on
the
service.
I
want
to
the
point.
What
kind
of
a
topology
physical
topology
should
be
deployed
with
with
logical
overlays
on
top
of
that?
F
Expedia,
so
we
come
and
hung
intent,
I
think
intent
based
network
management,
it's
a
very
good
idea,
but
it
has
to
define
intent.
Clearly
you
cannot
say
yeah
I,
don't
know
how
the
defines
congested.
I,
don't
know
what,
if
you
don't
know
how
to
define
the
detail,
define
the
above
level.
I
want
to
optimize.
I
want
to
look
balanced,
this
network
I,
don't
want
anywhere
have
more
than
80%
of
traffic
and
so
on.
You've
got
to
have
a
clear
definition
on
the
intent.
That's.
F
My
first
comment,
second,
is
just
to
say:
we're
going
to
use
a
I
in
machine
learning
to
to
do
nine
with
natural
language
recognition,
so
it
can
be
used,
but
it
does
not
help
if
you
don't
have
a
clear
description
on
the
problem.
For
example,
in
my
last
company
and
my
current
company
out
running,
ecommerce
so
sometimes
was
a
question
about
you
know
the
real
case.
If
in
the
networker
in
the
total
operation,
sometimes
you
run
into
the
problem,
people
just
writing
down
in
natural
language.
F
It
makes
problem
much
worse,
so
for
that
I
don't
think
it
start
with
a
natural
language
processing.
Everybody
right.
Therefore,
any
English
can
itself
can
be
ambiguous,
so
put
in
the
practice
right
how
people
do
I'm
just
bring
that
out.
So
people
say
well,
let's
have
a
common
form
fill
this
form.
Where
is
the
problem?
What
is
a
problem
in
certain
define
the
way
that
immediately
clear
all
the
problem?
Otherwise,
you
cannot
read
it
so
I.
Don't
think
that
part
you
need
to
have
AI,
but
what
AI
can
help
is
the
next
level.
F
D
Thank
You
Hollis
Eckhart
yeah,
so
to
go
back
to
you
know
the
comment
made
in
before
the
the
way
I
see
it
is
that
we
definitely
need
to
try
to
figure
out
terminology
first,
because
I
think
they
are
really
totally
different
things
that
people
understand
with
intent.
So,
for
example,
if
I
give
the
worst
case
example,
what
I
still
think
people
would
consider
to
be
intent,
networking
from
moley's
way,
which
is
I,
think
exactly
one
view
to
see
things
is
okay,
you
got
a
network
running.
There
is
nothing
cool
about
it.
D
It's
all
just
stupid,
CLI
that
you
have
and
now
I
basically
want
to
change.
My
service,
like
you
know,
a
lot
of
things
and
I
just
do
also
stupid
new
CLI,
but
now
the
intent-based
networking
is
carefully
take
the
new
bloody
CLI
and
try
to
make
the
network
adopt
to
this
without
breaking
down
and
going
up
in
flames
right,
and
that
sometimes
is
a
very
complex
operation
right.
The
fact
that
you
know
on
another
planet
people
are
thinking
about
intent
being
a
high
level
abstraction
of
that
crazy
configuration
that
comes
out
in
the
end.
D
That
is
I,
think
orthogonal
to
it
and
in
both
cases
we
the
word
intent
and
that
caused
a
lot
of
confusion,
so
I'm
not
sure
who
can
win
and
kind
of
keep.
The
word
intent
to
his
part,
I
think
we're
not
going
to
have
an
influence
on
the
big
players
in
the
market
that
go
around
and
say
that
the
whole
process
of
rendering
you
know
future
configuration
from
abstractions.
D
You
know
into
you,
know,
step
by
step
updates
to
devices
with
control
loops
and
then,
basically,
you
know
incrementally
make
sure
the
network
is
always
running
and
adopting
you
know
these
desire
level
things
the
whole
process
calling
that
intent.
That's
what
I've
seen
in
the
industry,
because
it's
very
abstract,
very
fuzzy
and
so
very
nice
for
marketing
to
use
the
word
intent
there
and
we're
not
going
to
get
rid
of
it
right.
D
So
maybe
that's
a
reason
to
say:
okay,
let's
keep
intent
being
that
and
for
anything
else,
we're
doing
we're
coming
up
with
more
priests
like
the
data
models
and
subdividing
it
in
the
way
that
Alex
said
maybe
trying
to
get
rid
of
the
word
intent
there
completely
and
instead
just
coming
up
with
different.
You
know,
subsets
of
policy.
You
know
with
the
taxonomy
of
these
things
and
just
leave
the
word
intent
to
the
marketing
people
describing
an
overall
system
just
one
idea
on
how
to
get
rid
of
it.
G
G
Can
you
tell
me
what
is
my
bank
balance
and
I
could
ask
that
question
and
somebody
else
can
ask
the
same
question
differently
and,
and
that-
and
there
is
an
answer
that
comes
back
saying-
your
bank
balance
is
so-and-so
and
if
your
question
is
not
phrase
correctly,
you
get
an
error
saying
us
is
not
a
good
question.
The
I
mean
there
are
approaches
how
to
solve
the
problem.
G
There
are
NLP
techniques
to
address
them,
so
the
question
is:
if,
if
that
called
Indies
and
Alexa
has
many
aces
and
all
that
and
most
of
you
might
be
from,
can
we
take
some
of
those
concepts
and
then
see
how
we
can
look
at
that
from
a
networking
perspective?
And
the
answer
is
I,
don't
know
I'm
Hughley,
as
a
researcher
I,
don't
know
the
answer.
Maybe
it
is
possible,
but
I'd
like
to
find.
H
The
amber
language,
I
think
you're
putting
the
cart
in
front
of
the
horse
and
I
would
support
tortoise.
You
know
tortoises
comments
on
that.
If
you
would
define
the
terminology
and
if
you
are
using
the
same
words
that
have
you
know
the
same
meaning
and
start
with
that
go
up,
you
know
to
some
use
cases
that
will
be
well
defined
by
the
operators
and
then
will
be
no
higher
because
right
now
we
really
have
to
know
that
we
are
when
that
you're,
not
overloading
the
terms.
That
would
be
the
good
start.
H
You
know
that
would
be
a
good
start
and
then
coming
into
the
use
cases,
because
I
know
that
often
what
operators
are
looking
for
is
the
SLA
with
the
functionality
and
there's
a
technical
SLA
in
the
business.
Smi
I
have
some
ideas
about
the
technical
isolates,
but
they
have
almost
no
idea
about.
You
know
the
business
as
allies
and
they
want
to
be
able
to
express
that
intercom
size.
You
know
essentially
description
what
they
want
to
get
from
the
network.
H
I
They
roll
opus,
apart
from
putting
the
cart
behind
the
horse,
I
think
I
see
not
the
cactus
has
to
be
behind
the
house,
but
what
what
was
yeah
we
all
set
yeah.
Apart
from
that,
the
idea
would
be
as
well
to
know
what
we
want
to
cover
in
the
cart
and
I
think
that
for
sure
we
need
a
terminology,
because
we
keep
talking
about
the
intent
and
we
keep
talking
about.
This
is
intend.
This
is
not
intent
and
we
all
we
all
have
a
different
I,
only
know
of
a
word.
I
That
is
that
much
of
what
loaded
recently,
that
is
slice
everything
is
a
slice
and
everything
and
when
we
could
do
slice
so
intent
of
slices
or
slicing
intent
or
something
like
that
seriously,
something
that
would
be
a
perform.
This
to
say,
I
support.
Definitely
that
we
need
I
mean
the
interpretation
of
this
group
that
hopefully,
will
end
up
being
the
interpretation
of
the
IRT
AFM,
even
the
IDF.
But
apart
from
that
something
that
would
be
interesting
as
well.
We
keep
talking
here
about
how
to
express
intent,
how
not
to
express
intent.
I
The
language
is
natural
language,
etcetera,
which
is
again,
is
a
it's
as
Trembley.
Why
the
set
of
possibilities,
and
probably
having
a
formal
description
of
not
I'm,
not
talking
about
the
language
talking
about
an
information
model,
a
schema
whatever,
so
how
the
intent
could
be
defined
will
be
an
extremely
interesting
results.
On
top
of
it,
we
could
build
the
formal
languages
to
build
intent.
We
could
consider
how
intent
can
be
translated
from
natural
language.
We
could
do
several
moving
several
directions.
A
Okay,
we
still
have
another
part
of
the
to
cover,
so
I
think
we
have
already
at
good
some
good
comments
on
the
mic.
So
just
to
give
a
rough
conclusion,
and
of
course,
in
the
minutes
and
on
the
meaningless,
we
will
try
to
precise
more
the
next
step.
But
what
I
get
from
the
from
the
comment
on
the
mic
is
that
there
is
a
need
to
fix
the
terminology.
First
I
mean
this
is
a
to
clarify
the
use
of
the
terms,
and
this
is
a
what
has
been
tried
to
be
done
in
the
initial
draft.
A
They
have
been
also
pointing
the
need
to
define
more
clearly
use
cases
and
your
personal
perspective,
I
mean
from
the
operator
or
from
configuration
of
networks
where
an
intern
based
approach
would
be
meaningful
so
which
are
also
to
affect
on
that
to
see
if
we
need
to
have
other
options
highlighting
use
cases
or
if
we
need
to
collect
requirements
or
input
from
the
operators,
they
have
in
also
the
comments
on
the
aspect
of
using
or
not
natural
language
processing
in
this
work
with
different.
Diverging
opinions,
so
we
will
see
what
we
can
do
with
that.
A
Just
a
bit
of
my
summary
I
would
like
to
propose
to
you
is
for
me
currently
that
multiple
definitions
and
multiple
domains
that
are
using
the
term
intense.
So
this
is
only
one
factor
that
explains
the
confusion
in
the
use
of
the
terms
we
have
been
Sdn
NFV,
but
also
in
network
management.
Everyone
is
using
intents
for
their
own
purpose,
usually
related
to
different
approaches
and
different
means
of
realizing
a
Nitin
based
system
and
also
targeting
different
levels
or
different
layers.
A
A
So
there
is
different
producer
different
consumer
of
intents,
the
I
think
most
common
aspect
of
intent
that
everyone
converging
is
that
it's
an
abstraction.
It's
not
it's
not
much
to
say
that,
but
at
least
everyone
converts
saying
that
we
need
to
work
on
abstractions.
Those
abstractions
can
be
at
different
levels
representing
different
things,
but
at
least
this
is
a
means.
People
want
to
use
a
neighbor,
I
would
say
more
powerful,
attract
management
at
a
configuration.
A
So
the
question
comes
now
what
to
do
next
and
I
think
we
will
also
explore
that
later
offline.
But
first
you
remind
that
we
are
in
a
research
group,
so
just
to
contradict
a
bit
Taurus
about
letting
I
mean
the
industry
or
the
things
stabilized
on
this
side.
I
think
we
are
here
also
to
a
bit
work
ahead
of
time.
I
mean
ahead
of
the
con
of
the
of
the
market,
so
we
need
to
investigate
this.
A
I
mean
if
the
research
group
agreed
on
that
this
is
I,
think
something
we
should
do
ahead
of
of
the
industry,
not
saying
that
the
industry
will
not
do
what
they
want,
but
at
least
from
a
research
perspective
we
need.
This
is
something
we
could
do
so
the
effort
will
surely
beyond
the
terminology
or
taxonomy.
There
would
surely
be
something
like
this
as
discussed
in
previous
meetings.
A
K
Do
like
just
to
try
to
also
summarize,
but
it
seems
to
me
data.
We
are
talking
at
least
about
three
different
aspects.
The
first
one
is
the
terminology
defining
what
what
an
intent
is
and
what
it
is.
Not
so
you
know
what's
the
relationships,
for
example
between
intent
and
policies.
The
second
aspect
is
how
to
express
them,
something
that
has
been
mentioned
too.
So
we
may
have
up
information
model
to
to
express
intent.
For
example,
as
Iago
mentioned,
that
right,
so
one
aspect
is,
is
the
terminology?
K
The
second
aspect
is
how
to
express
them
and
the
third
aspect:
it
seems
to
be
how
to
realize
it
over
the
network,
so
that
could
include,
for
example,
architectures
as
mentioned
before
or
components,
and
so
on,
so
in
in
use
cases
could
help
in
that
part
as
well.
So
I
would
summarize
in
three
aspects
the
different
terminology,
the
definition.
What
intent
is
the
how
to
express
it
and
how
to
realize
it.
A
A
This
is
just
to
reflect
on
what
has
been
done
this
year
in
energy
and
also
what
people
like
to
do
with
you
next
year.
So
we
had
four
meetings
in
2017,
including
this
one
in
March,
in
Chicago,
with
66
participants,
and
the
topic
was
beyond
autonomics
autonomics
2.0,
where
we
had
a
bunch
of
presentation
and
still
we
want
to
maybe
initiate
also
some
activity
on
this
revision
of
economic
networking.
A
second
meeting
was
in
the
am
conference
in
lisbon.
A
We
had
much
fewer
participant,
but
it
was
more
kind
of
interim
meeting
and
difficult
to
attract
people
in
the
this
conference,
but
we,
this
is
where
we
investigating
new
research
items
and
in
fact
this
is
where
we
initiated
more
concretely.
The
these
new
documents
on
base
networking
and
the
realization
is
today,
then
in
Prague,
you
HTF
99,
with
34,
334
and
85
pass
the
points
in
the
two
sessions.
A
The
first
session
was
the
general
meeting
on
onion
emoji,
with
general
topics
addressed
and
the
second
one
was
the
series
of
the
workshop
series
on
measurement
based
network
management,
who
is
always
successful
and
so
I
joined
Giovanna
in
the
fact
that
there
is
in
energy
one
of
the
pillar.
It's
a
monitoring
and
measurement
measurement
time
management
story,
and
this
has
been
here
in
the
research
group
for
a
long
time,
more
recently,
Mukherjee
as
emerge
and
maybe
collaboration
between
the
two
groups.
A
So,
concerning
the
progress
of
the
work,
we
have
one
active
research
group
document.
We
just
passed
the
IRS
report
and
the
next
step
will
be
to
go
to
iOS
review.
So
this
is
a
good
achievement,
especially
because
it
was
very
difficult
to
make
progress.
I
mean
on
the
approval
of
the
document
for
logistic
reason,
I
would
say,
but
this
comes
to
an
end
hope
we
will
be
able
to
speed
up
a
bit
the
process
to
a
publication
before
the
end
of
the
year.
A
In
parallel
to
that,
we
have
also
four
active
individual
documents,
so
there
are
two
on
the
intense.
The
third
one
is
on
reinforcement,
learning
techniques
and
the
fourth
one
is
on
Congrats,
and
we
also
had
this
your
decision
on
one
previous
research
group
document
that
was
on
IP
fix
that
was
not
accepted
to
be
published
after
an
is
G
complete
review
and
the
authors
are
investigating
if
they
would
like
to
propose
that
via
individual
submission
for
next
year.
A
This
is
really
a
shopping
list,
so
what
we
used
to
do
in
the
energy
is
to
have
at
least
one
meeting
with
one
of
the
I
Triple
E
a.m.
noms
conference.
Basically,
this
is
a
premium
network
management
conference
is
in
alpha
police.
So
we
would
like
to
keep
this
this
set
up,
so
this
will
make
us
meet
at
noms
entire
one
in
April
I,
beyond
that
we
also
used
to
have
co-located
meetings
with
IETF
meetings.
So
these
are
the
three
that
are
planned
for
next
year.
We
don't
need
to
meet
at
a
variety
F.
A
This
is
also
up
to
the
group
to
decide
where
it's
more
relevant
to
meet,
so
we
have
London
in
March
Morel
in
July
and
the
location
to
be
announced
in
November.
So
this
will
also
make
a
poll
on
the
mailing
list
to
get
the
feeling
of
the
research
group
where
it's
more
relevant
to
meet
in
parallel
to
that
through
either
through
discussion.
We
asked
with
different
groups
inside
and
outside
of
IETF
rgf.
A
We
also
trying
to
organize
order
interim
or
topical
meetings
on
specific
aspect,
so
also,
if
you
would
like
to
suggest,
suggest
specific
meetings
that
you
think
are
relevant
and
you
have
ideas
how
to
organize.
It
can
be
with
workshops
with
another
conference
on
order,
SEO
or
open
source
project
event.
This
is
something
we
can
organize.
So
if
you
have
proposal
ideas
where
you
think
the
energy
it's
important,
that
we
meet
other
people
with
other
views
during
other
stuff,
it's
also
something
we
can
organise
next
year.
A
Okay,
today
we
have
gone
through
the
themes
on
intern
based
network
management,
so
we've
increased
starting
to
have
a
bit
of
momentum,
winning
nests
of
people
to
write
documents
and
continued
effort
on
defining
work
for
the
research
group
on
this
topic,
we
also
had
discussion
in
the
past
about
autonomics
2.0
and
tomorrow,
we'll
investigate
a
ie
techniques
from
the
talk
management.
So
this
is
not
yet
settled.
This
is
things
we
want
to
discuss
with
the
research
group.
What
are
the
priorities?
What
are
the
energy?
A
We
can
collect
to
really
address
and
progress
the
work
in
those
themes.
Of
course,
if
you
have
other
topics,
share
them
on
the
mailing
list,
come
talk
with
us.
It's
really
open
network
management
is
it's
very
wide
topic
and
from
these
themes
we
want
also
to
be
able
to
define
a
bit
more
I
will
say
a
second-level
research
items
so,
as
we
discussed
on
the
IBM,
for
instance,
how
to
call
them
how
to
so
the
different
stage
of
doing
in
turn-based
networking,
but
also
also
to
reflect
on
your
global
work.
Man-
it's
not.
A
The
Charter
is
very
well
written
because
it
survived
on
up
to
now,
so
we
can
keep
it
if
you
think
it's
still
relevant,
but
networks
have
changed
a
lot
since
then,
and
if
we
think
that
the
missions
or
the
scope
are
not
well
reflected
in
the
Charter,
this
is
something
we
can
consider
to
revise
the
Charter
Morris.
So
it's
all
up
to
us
up
to
you
to
decide
what
you
want
to
do
with
this
research
group.
I
think
we
have
maybe
a
bit
of
time.
A
No
I,
don't
think
we
need
to
read
them.
I
would
prefer
to
have
a
feedback
from
the
room
if
you
think
we
are
going
the
right
way
to
restructure
the
the
research
group
and
not
restructure,
but
to
structure
the
research
group.
What
topics,
you
think
are
very
important.
We
need
to
prioritize
how
to
address
those
topics,
and
only
if
you
think
that
we
are
going
the
right
way.
This
will
encourage
us
to
continue
this
way.
Thank
you.
I
Okay,
so
know
just
develop
it.
I
I
like
the
idea
of
having
this
structure
around
a
few
main
goals
or
main
paths
to
go
so
I
and
I
think
that
even
related,
these
ivn
I
mean
the
intent
and
I
techniques.
Ai
techniques
really
relevant
autonomics
2:04
Mesa
is
a
little
bit
fussy
right
now,
but
probably
probably
somehow
connected
with
the
others,
and
this
is
something
that
we
had
to
I
mean
to
make
our
minds
up
to
to
be
clear,
whether
we're
referring
to
that
and
upon
that
I.
I
C
Yeah,
let's
trim
yeah
regarding
this
I
I.
Think
it's
a
good
idea
to
have
actually
some
themes
and
supposed
to
give
a
little
bit
structure
for
those
reasons
and
give
it
a
little
bit.
Direction
purpose
and
I.
Don't
think
it's
necessary
to
really
change
this
charter.
Father
I
think
the
Charter
has
served
very
well
the
faces
of
a
could.
C
There
be
some
terminology
that
could
be
modernized
if
you
will
have
the
control
reflected,
maybe
but
I,
but
I'm
I'm,
not
sure
at
the
same
time
it
has
served
us
well
so
far
so
might
as
well
also
be
left
on
untouched.
Maybe
looking
at
the
terminology
can
be
dusted
off,
but
otherwise
I
think
leave
it,
as
is.