►
From YouTube: IETF116-COINRG-20230327-0030
Description
COINRG meeting session at IETF116
2023/03/27 0030
https://datatracker.ietf.org/meeting/116/proceedings/
A
A
D
B
D
D
D
Unfortunately,
all
of
us
cannot
attend
this
meeting
in
person.
Just
really
I
know.
This
is
a
really
beautiful
season
in
Japan.
Very,
very
pretty
so
there's
a
note
well
for
ITF
meetings,
so
I
think
we
should.
You
should
pay
some
attention
to
this,
especially
for
the
rpr
part.
If
you
have
some
patent
or
something
related
to
rpr,
you
should
close
them.
D
And
this
is
some:
my
name
is
Chiefs
for
this
meeting.
You
have
already
I
guess
you
have
already
accessed
vitaco
and
we
have
some
links
for
live
minutes
and
the
meeting
materials.
So
if
you
need
want
to
read
them
the
slides
in
real
time,
you
can
access
that,
and
please
remember
this-
is
the
research
Loop.
We
are
not
a
typical
working
group
in
iitf,
we
are
basically
we
are
not
a
standard
developer
organization.
B
B
Sadly,
my
co-author
iconsa
is
a
sick
and
my
other
co-author
Dirk
Crossen
is
in
a
plane
because
there
were
problems
with
strikes,
but
I
see
that
we
have
some
of
the
other
authors
of
the
use
case,
analysis
and
the
terminology
that
are
there.
So
they
will
be
able
to
support
us.
We
also
have
a
number
of
new
drafts
going
into
as
the
end
a
framework
for
control
environment.
It's
a
very
interesting
problem
that
was
raised
and
actually
just
as
not
just
a
draft,
but
it's
also
more
of
a
technical
discussion.
B
Research
discussion
around
security
and
Coin,
and
that
is
that
was
actually
very
much
part
of
our
original
Charter.
So
it's
very
important
to
have
it
and
then
another
problem
in
data-driven
coordination.
Sadly,
because
Ike
is,
is
sick,
maybe
the
the
research
topic.
We
may
have
a
larger
discussion
and
we
kept
quite
a
long
time
at
the
end.
To
start,
maybe
addressing
a
larger
Tropic
topics
and
directions
and
have
time
for
once
to
answer
questions
next
slide.
D
B
D
G
F
E
F
You
so
I
mean
Luis
Contreras
from
telefonica.
I
will
present
this
drafting
on
behalf
of
Michael
authors
from
Orange
developer
from
telefonica
and
Carlos
bernardo's
from
uc3m
sorry.
So
what
we
will
present
today
is
an
evolution
of
a
layer
architecture
for
sdn
that
we
move
in
in
IDF
in
the
past,
so
the
idea
will
be
to
do
some
incremental
work
on
top
of
this
architectural
framework.
Next,
please,
thank
you.
F
So,
okay,
so
the
background
is
these:
are
these
architectures
original
idea
was
adopted
in
the
sdn
research
group,
but
at
the
time
that
this
research
group
was
dismantled,
we
moved
the
draft
to
the
independent
stream
editor
right,
so
the
the
that
becomes
an
RFC,
the
8597
and
essentially,
what
proposed
this
architecture
is
a
a
lighter,
competitive
interworking
between
the
control
of
the
services
and
the
control
of
the
transport
capabilities.
Next
slide,
please
yeah!
F
So
here
you
can
see
the
representation,
the
illustration
of
the
architecture
so
how
we
divide
the
control
responsibilities
in
with
regards
to
the
service
and
with
regards
to
the
to
the
transport
part.
So
we
Define
these
functional
strata.
There
is
a
Stratton
for
everything
related
to
the
to
the
service
itself
and
transport
stratum
and
within
each
of
the
Saturn
we
Define
different
planes,
the
control
plane
and
the
management
plane
that
are
surface
plane
and
the
resource
plane.
That
is
works
on
top
of
all
the
resources
for
each
of
the
of
the
strata.
F
F
F
The
first
one
is
that
the
networks
are
evolving
for
interconnecting
compute
environments,
so
the
idea
will
be
to
have
this
compute
component
as
part
of
the
architecture,
so
also
creating
considering
the
resource
playing,
control,
plane
and
management
plane
of
all
this
environment,
and
the
second
point
is
that
more
and
more
the
operations
are
leveraging
on
artificial
intelligence
machine
learning
capabilities
in
order
to
perform
this
closed,
loop
automation,
management,
Loops
right,
so
the
the
point
was
to
do
this
expansion
of
the
architecture,
considering
these
two
different
dimensions
next
slide.
Please.
F
So
then
we
well
oh
okay,
here
the
colors
probably
are
not
very
foreign,
so
essentially,
what
we
created
was
one
first,
we
renamed
the
transporter
stratum
that
was
the
previous
name
by
the
connectivity
of
stratum
just
for
reflecting
better.
What
about
with
the
aim
of
what
we
call
in
the
past
Transportation
so
essentially
working
with
the
networking
part
the
connectivity
part
of
the
of
the
architectural
framework.
F
Then
we
added
this
computer
Stratton
for
reflecting
the
introduction
and
the
management
capability
control
capabilities
for
the
compute
site
and
then
for
each
of
the
planes
we
added
the
learning
plane
for
each
of
the
strata.
Sorry,
we
are
the
the
learning
plane
with
the
idea
of
these
planes
to
work
with
the
data
analytics
and
the
the
the
the
processing
of
the
data
of
each
of
the
particular
strata.
F
As
you
can
see,
all
of
the
idea
is
to
integrate
all
of
them
so
either,
even
though
they
are
separated
so
how
they
need
to
interwork
and
collaborate
in
the
final
end-to-end
delivery.
So
this
is
why
you
can
see
the
different
lines:
integrating
the
different
planes
like
the
resource
plane,
the
control,
plane,
control
management,
plane
and
learning
plane
of
all
of
the
estrata.
So
the
idea
is
that
all
of
them
can
enter
work
for
the
learning
plans.
F
F
Potential
research
directions
that
we
identify
well
so,
essentially,
probably
we
can
propose
some
use
cases
that
probably
could
help
to
understand
better
the
architectural
capabilities.
This
is
something
that
we
could
consider
doing
include
in
the
in
the
draft,
also
to
work
in
different
aspects
of
the
architecture,
like
the
communication
interfaces
between
the
different
strata
and
the
planes
themselves.
The
deployment
scenario
so
also
understanding
how
this
architecture
could
fit
with
the
Legacy
scenarios,
as
the
ones
that
we
have
today
is
already
mentioned.
Also
the
link
with
ongoing
activities
in
nmrg
and
co-energy.
F
The
reference
here
to
nmrg
is
because
we
presented
this
architecture
in
the
past
in
an
MRG,
and
the
energy
chairs
recommended
us
to
present
to
you
here
in
Co
energy.
So
this
is
why
we
had
here
the
the
reference
to
an
energy
then
explore
explore
Noble
architectural
approaches,
probably,
for
instance,
one
way
to
follow
will
be
to
consider
Abbas
architecture
for
interconnecting
all
the
the
planes
service
based
service
bus
based
architecture
working
in
the
apis
between
the
different
strata
explore.
F
F
Yeah
next
steps
essentially
to
set
the
scope
of
the
draft
aligned
with
the
scope
of
co-energy
collect
feedback
from
the
research
group
to
see
if
this
is
of
interest
for
for
you
to
continuous
plotting
here
in
in
in
this
home,
and
prepare
a
new,
a
more
detailed
version
for
next
ITF
and
essentially
trying
to
collect
your
your
feedback.
Thank
you.
B
B
B
F
Yeah,
so
so
to
talk
what
we
had
in
mind
at
the
dino
defining.
This
was
the
idea
of
the
the
kind
of
situations
that
we
face
today
in
in
operational
networks
like
the
ones
for
telephonica.
So
essentially,
you
will
have
some
some
capability
manager
capabilities
handling.
The
computer
stops
you
can
imagine
openstack
or
kubernetes
or
whatever,
and
we
need
to
provide
a
kind
of
orchestration
between
that
part
and
also
the
traditional
SDM
part
in
the
transport
Network
and
so
and
coordinate
with
the
service.
The
Service
Partners.
F
So
imagine
that
we
deploy
some
service
function
in
a
computer
environment
and
we
need
to
provide
connectivity
to
that
service
functions.
Service
function
chain,
for
instance.
So
the
point
here
will
be
to
integrate
everything
together
to
ensure
that
the
end-to-end
delivery
is,
as
is
smooth
and
is
a
I
mean,
is
adequate
and
appropriate
to
the
expectations
of
the
of
the
those
service
functions.
F
B
B
I
could
I
would
suggest
that
you,
you
should
rearrange
this
to
be,
maybe
a
bit
more
clearer
in
terms
of
what
having
the
compute
stratum
would
actually
help
the
other
things
and
maybe
move
boxes
around
I'm,
not
saying
it's
not
interesting,
I'm.
Just
saying
that
the
way
it's
presented,
it's
not
clear
how
that
is
going
to
evolve,
and-
and
maybe
it's
me
but
yeah
I-
would
just
suggest
that
so
that
the
scope
is
is
clearer.
Actually.
F
G
H
Network
systems,
research
and
design,
I
I,
have
a
question.
There's
there
seems
to
be
an
implication
in
this
slide
that
I
hope
isn't
true,
which
is
that
there's
a
hierarchical
relationship
between
the
connectivity
and
compute
stratum
where
the
connectivity
stratum
is
somehow
at
a
higher
level
than
the
compute
stratum.
I
assume.
That's
not
intentional.
It's
an
artifact
of
the
way
you've
drawn
it
yeah.
B
J
J
In
the
past
five
to
ten
years,
we
have
seen
a
lot
of
research
on
programmable
Network,
so
programmable
network
devices
actually
have
a
really
inspired
a
lot
of
research
work
like
load
balancers
and
fire
firewalls
in
in
the
area
of
coin.
But
technically
we
think
that
they
are
not
strictly
Computing
in
the
network,
but
actually
Hardware
implementations
of
network
functions.
We
think
that
for
the
Computing
in
the
network,
they
are
very
closely
coupled
with
applications.
J
J
J
So
this
may
race
a
very
higher
high
bar
for
the
application
development.
So
we
think
that
we
need
to
find
a
very
application.
Development
I
would
think
of
a
way
to
make
it
easier
for
the
application
development
so
to
make
the
Apple
developers
to
better
leverage
the
network
capabilities
owned
by
The
Operators,
because
most
of
the
application
development
may
come
come
from
the
third
parties.
So
in
this
draft
we
think
that
we
want
to
to
increase
the
general
generality
and
of
the
coin
development
and
promote
the
the
standardization
of
the
coin.
J
So
in
this
drive
we
summarize
some
general
computing
Primitives
from
the
existing
research
work
and
also
that
we
also
want
to
hope
to
promote
them
to
be
standardized,
and
these
Primitives
can
be
implemented
inside
normal
network
devices
rather
than
the
programming
network
devices
so
and
also
we
propose
a
generic
framework
of
calling
the
controlled
environment.
Thank
you
next
slide.
Please.
J
I
here
is
the
the
overall
it's
a
very
initial
framework
of
coin.
It
contains
about
three
layers
on
the
on
the
top.
There
is
a
scheduling
layer
need
to
come
decompose
these
jobs
into
host
tasks
and
con
tasks,
because
when
they
think
that
to
make
the
coin
to
be
scaling,
some
of
the
some
of
the
tasks
may
be
distributed
across
multi-service
or
multi
racks
or
even
across
your
geographically
distributed
multi
sites.
J
So
the
the
scheduling
layer
needs
to
think
of
a
way
to
to
how
to
decompose
these
job
tasks,
and
there
is
also
a
control
layer
and
the
control
layer
may
contain
two
controllers
and
the
holes
on
the
whole
side
or
the
and
the
coin
side,
and
they
need
to
think
of
a
way
to
make
the
end
side
and
network
side
how
to
collaboratively
collaborate
more
effectively
and
also
on
the
top
on
the
bottom.
There
is
an
infrastructure
layer.
It
contains
different.
J
J
So,
in
the
as
I
mentioned
in
the
schedule
layer,
the
policies,
a
Shannon
policy-
vary
for
different
workflows,
for
example,
for
distributed
machine
learning,
training
and
mapreduce.
They
are
different
because
one
of
maybe
bash
processing
jobs-
another
might
be
streaming
percent
drops
so
these
different
tasks.
So
it
should
be
considered
differently.
So
we
need
to
think
of
about.
We
also
need
to
increase
and
make
them
all
all
down
in
the
scheduling
layer.
J
So,
and
also
with
the
addition
of
in
our
Computing
technology
is,
it
is
necessary
to
consider
not
only
the
host
resources
but
also
the
coin
resources.
So
next
slide,
please
so
the
control
layer,
we
think
the
core
of
the
entire
architecture
so
because
they
need
to
coordinate
the
host
and
the
coin
resources
and,
on
the
whole
side,
the
the
controller
need
to
make
the
host
application
cooperate
with
coin
processing
and
because
we
think
that
the
coin
processing
may
be
a
complementary
work
for
the
entire
Computing
Computing
work.
J
J
Foreign,
so
in
the
infrastructure
layer-
and
we
have
summarized
a
different
coin-
printers
that
we
hope
to
be
standardized
in
the
future,
not
exactly
in
the
research
so
like.
We
summarize
here
are
some
four
basic
Primitives
and
their
corresponding
dust
structures
in
the
existing
work
like
values
during
aggregation
to
these
are
usually
used
in
distributed
machine
learning,
training
and
also
asynchronous
value.
J
Aggregation
they're,
mostly
used
in
big
data,
analytics
applications
where
map
reduces
needed
and
also
key
value,
key
value
pair
that
you're
used
for
caching
and
also
consensus
they
use
for
synchronization
in
Distribution
Systems,
so
so
coin
transformation
of
applications
should
be
considered
and
in
on
the
whole
side.
So
in
this
way
the
flat
of
flexibly
process
of
the
data
can
be.
It
can
be
realized,
So
yeah.
Thank
you
next
slide.
Please.
J
So
the
the
since
the
framework
is
very
initial
and
we
will
update
it
in
update
the
details
in
next
meeting.
So
there
are
also
some
research
challenges
and
other
considerations
like
how
to
realize
efficient
collaboration
of
and
the
network,
for
example,
like
fallback
mechanisms
and
where
tasks
cannot
be
fully
accomplished
within
the
network.
This
will
be
finished
at
the
end
device,
because
the
Computing
capability
of
the
network
may
not
be
that
efficient
and
also
the
the
reliability
and
the
crackness
OR
at
dependency
need
to
be
solved
to
maintain
the
coin.
J
J
B
First,
you
mentioned
that
you
want
to
standardize
coin
I
think
we
mentioned
at
the
beginning.
Now
this
is
with
my
my
chair
hat.
We
mentioned
that
it
is
not
the
role
of
this
community
to
standardize
coin.
We
are
going
to
do
research,
and
so,
if
the
goal
of
your
of
your
efforts
is
to
come
with
a
standardization
effort,
well,
we
may
not
be
the
right
people
and.
B
B
My
text
to
to
Jeff
went,
went
on
on
this.
You've
been
very
critical,
but
at
the
same
time
the
three
articles
that
you
suggest
at
the
beginning
are
articles
that
are
previous
to
the
establishment
of
this
working
group
of
This
research
group.
Sorry,
so
those
are
old
now
and
I
agree
that
that
was
old.
There's
been
an
awful
lot
of
work.
That's
been
done
in
this
research
group
on.
B
Discovery
the
use
of
intelligent
devices-
it's
not
true
that
they're
all
network
devices,
you
know
that
P4
has
been
ported
to
Pi
devices,
so
you
can
do
these
things
very
much
at
the
edge
there's
been
some
work
on
device
Discovery,
so
I
think
I
I
welcome
the
fact
that
you're
doing
a
layered
architecture
that
I
think
is
valid.
As
you
know,
also,
there's
been
talk
of
doing
intelligence
slicing
for
6G,
for
example.
To
put
some
intelligence
into,
you
know
push
intelligence
and
Computing
into
rants.
B
J
Thank
you
for
a
question:
yeah
yeah
for
the
role
of
coin
RG.
It's
actually
a
research
group
and
also
like
an
open
Forum
to
sharing
them.
The
work
on
the
research,
yeah
I
think
it's
correct,
but
our
intention
to
prevent
this
working
coin
is,
to
you
know,
gather
some
information
or
discussion
for
the,
maybe
some
the
standardization
work
in
the
future,
but
not
to
to
change
the
role
of
coin.
But
you
know
I.
J
J
That
coin,
maybe
is
actually
standardization
group
for
standardizing
some
work
in
in,
like
yes,
so
and
but
we
want
to
gather
some
chance
to
promote
them
to
to
just
erase
this
question
to
to
share
some
our
understanding
or
not
just
to
standardize
this
this
working
coin,
but
maybe
somewhere
in
other
groups
or
in
other
areas
in
in
the
future.
So
we're
aware.
B
G
G
J
Actually,
they
are
based
on
some
actual
research
because
I
I,
sorry
I,
didn't
mention
all
of
these
cool
authors.
So
some
of
the
courses
they
actually
published
some
the
the
papers
in
in
major
conference,
so
Network,
Conference
and
also
I.
What
I
present
in
my
slides
are
actually
their
work.
The
summation
of
their
work,
so
I
mean
the
the
framework
framework
realization
or
the
the
Primitive
summation.
So
maybe
you
can
I
also
make
a
reference
in
my
drafts.
G
So
one
question
I
had
with
your
draft
was
so
you
the
title
is
somehow
generic
framework
for
coin
in
controlled
environments.
Do
you
think
the
conservations
would
change
a
lot
if
we
didn't
focus
on
controlled
environments,
but
just
focus
on
the
say,
technical
requirements
for
distributed
computing
in
general.
J
Thank
you.
It's
a
very
cool
question.
I
think
that
controlled
invariance
is
a
very
initial.
You
know
for
the
deployment
it's
very
initial,
like
a
current
research
focused
in
data
centers
like
they
are
controlled,
so
maybe
it's
easier
to
implement
so
maybe
to
make
it
scalable.
We
think
that
one
day
that
the
framework
May
extend
to
an
environment
and
not
control
or
be
shared
by
different
providers
or
like
that,
thank
you.
I
I
I
Can
you
say
more
about
how
those
compare
to
other
in-network
computer
operations,
P4
being
perhaps
in
network
compute
languages,
P4
being
perhaps
the
most
widely
known,
and
why
you
chose
those
Primitives
and
why
you
believe
that
set
of
Primitives
is
appropriate
compared
to
the
other
Primitives
yeah?
The
goal
here
is,
of
course,
to
explore
the
research
rather
than
to
to
just
present
the
ideas.
Okay,.
J
Okay,
so
I
and
as
you
mentioned,
that
coin
is
a
research
group
and
I
know
I
understand.
So
what
we
summarize
for
the
second
question
you
race
about
The
Primitives.
We
think
first,
like
we
summarize
them
from
most
of
the
existing
work,
and
also
these
Primitives
are
also
not
the
the
common
Primitives
like
Network
function
related,
but
also
the
I
think
we
want
to
emphasize
is
they
are
actually
application?
Specific
functions,
Primitives
right,
so
this
is
sure.
I
J
Your
research,
actually,
actually
we
don't
want
to
stand
out
here,
but
I
think
this
might
be.
You
know
one
day
in
maybe
in
you
know,
yeah
we
don't.
We
don't
want
to
standardize
here,
but
maybe
the
these
Primitives
can
be
used
in
in
in
common
industry
or
maybe
implemented
by
different
on
network
device.
Vendors,
not
exactly
by
programmable
network
devices
like
in
open,
open,
open
source,
Community
yeah.
J
D
So
my
turn
actually
I
have
the
this
is
Jeffrey
so
I
with
without
my
chair
hat,
so
we
it's
gonna,
have
the
same
question
similar
as
callings.
So
if
you
want
to
Define
some
Primitives,
it
seems
there's
some
contradiction
between
application,
specific
and
some
Primitives
abstract
some
abstraction
level
and
widely
used
by
many
applications.
So
it's
ambitious
I
think
it's
an
ambitious
objective
to
Define
this
Primitives.
D
D
I
To
follow
up
on
that
I
agree
that
it
would
be
that
this
is
an
interesting
idea
and
that
that's
understanding
those
challenges
and
understanding
why
this
is
the
right
set
of
Primitives.
It
would
be
a
very
interesting
thing
to
discuss.
I
would
like
to
see
that
discussion
happening
in
the
research
group.
D
Yeah
and
also
I
I
liked
mention
you
that
in
previous
meetings,
Ike
try
the
similar
ideas,
try
to
Define
some
generic
framework
or
genetic
protocol
for
this
so
I.
We
I
think
many
quite
a
few
people
Envision
this
their
legacy,
maybe
in
future,
but
real
at
this
stage
we
need
a
more
solid
research
to
justify
this
idea.
I
mean
maybe
a
lot
quite
a
lot
of
discussion
or
debate
on
this
kind
of
definition
of
Primitives.
G
H
B
B
So
there's
there
is
a
lot
of
research
into
moving
beyond
the
current
Primitives,
not
maybe
just
changing
them,
but
adding
to
the
the
richness
of
the
of
of
the
the
the
field
in
a
way
because
or
the
original
applications
were
all
you
know.
Data
center
were
very,
very
you,
know
their
line
rate
very
high
data
rate,
but
a
lot
of
people
who
do
research
on
this
and
I
will
come
back
to
what
Dirk
was
mentioning
in
terms
of
distributed
networks.
B
A
D
A
A
A
Starting
from
coin,
for
me,
coin
means:
okay,
Computing
the
networks
and
inside
my
deploying
pro
programs
based
on
the
Computing,
the
resources
still
in
a
programmable
Networks
devices.
That
now
refers
the
pnd
and
this
infrastructure
could
be
used
typically
in
Edge
Computing
or
that
kind
of
stuff,
and
the
program
works
with
several
pnd
exchanging
data
over
secure
communication.
A
A
Foreigntication
and
keying
mechanism
based
on
symmetric
or
asymmetric
secrets,
and
so
the
research
item
for
current
Security
in
my
mind
in
this
area
of
the
following.
First
is
the
security
architecture,
that's
to
say,
security
of
communication
or
and
second
pnd
Sacramento
model.
That's
to
say
how
much
security
is
there
in
pnd?
A
A
A
A
A
Sorry
next
slide,
please!
Yes,
thank
you.
So,
regarding
the
the
program
security,
the
program
security
is
as
may
have
some
security
requirements
so
through
the
security
of
the
infrastructure
itself,
the
programmer,
the
the
load
and
after
they
some
program
for
examples,
program
that
school
this
with
blockchain
and
whatever
could
need
to
understand
some
security
parameter
in
order
to
perform
some
operation,
for
example,
interaction
for
for
for
production
and
so
in
this
area.
A
A
May
have
some
requirement
in
order
to
to
have
some
confidence
in
the
fact
that
you
need
attestation
mean
transferring
the
control
by
the
infrastructures
or
say
the
trust
by
the
infrastructure
to
to
the
client
for
use,
and
this
implies,
for
example,
some
cryptographical
provisioning,
so
the
level
of
security
and
Trust,
as
you
say,
a
way
to
to
give
some
security
Assurance
to
the
client,
for
example,
in
model
like
common
criteria
in
others.
That's
you,
you
may
give
some
security
Insurance
to
to
your
customer
and
last
is
square
square
ability
and
performance.
A
So
it's
mean
it's
kind
of
ashesem,
implementing
key
system
management
procedures
and
that
is
can
be
rich
through
uniform
resource
identifier.
So
you
you
see,
there's
this
small
schematic
and
you
see
some
relationship
with
the
pnd
and
some
server
that
are
able
to
handle
on
the
model,
cryptographic,
services
and
I.
Believe
it's.
The
last
slide,
so
next
slide
I
believe
is,
is
no
slide.
Baby
I'm
down.
B
Thank
you.
We
wanted
to
take
the
opportunity
of
your
presentation
to
talk
a
little
bit
more
about
security
and
in
in
coin
I,
wouldn't
call
it
a
coin
infrastructure
I
would
call
it
a
coin
enabled
Network,
and
if
we
start
thinking
about
this,
you
know
you
have
your
your
coin.
You
know
your
network
is
a
bit
like
a
computer
board
and
you're
running
an
operating
system
on
it
and
your
operating
system
all
are
what
you
call
a
program,
but
I
would
call
it
they're
more
like
what
the
previous
talk
was
talking
about.
B
B
That
is
more
part
of
again
of
that
computer
network
and
has
storage
and
and
Computing
elements,
and
all
of
these
things
are
are
talking
together
and
and
with
the
with
the
evolution
of
programming.
You
know
these
things
are
becoming
more
and
more
white
boxes,
so
we
can
do
a
lot
of
things
on
them.
So
how
would
you
see
the
evolution
of
what
you're
doing
into
a
maybe
a
much
more.
B
Capability
instead
of
well
not
a
well
I,
think
capability
goes
back
to
what
people
are
saying.
We
need
running
code
at
one
point,
but
I
I
think
so
my
question
is:
how
do
you
see
that
evolution
of
the
security,
knowing
that,
if
we
want
all
these
nodes
to
talk
to
one
another,
we
have
to
be
able
to
have
reliable
and
and
configurable
at
the
same
time,
Network.
A
A
B
B
Map
into
into
like
a
more
of
a
of
a
core
Edge
or
actually
a
clouded
infrastructure,
because
what
we
see
us
people
working
on
edge
networks
is
that
we
cannot
just
use
things
that
were
done
for
data
centers.
We
don't
have
the
same.
You
know
we're
the
networks
in
the
small,
we
don't
have
the
same
capability.
So
how
does
that
map.
A
B
I
I
A
A
You
have
a
single
on
TT,
but
that's
try
to
deliver
what
you
said,
for
example,
if
I'm
singing
to
security,
I
think
to
others
secure
modules
and
in
fact
there
are
single
pin
on
device,
but
for
partitions,
I
try
to
create
a
multi
tenant
interface,
and
so
in
my
research
I
am
working
on
the
ID.
That's
Securities
is
multi-tenant,
that's
to
say
that
you
have
enough
physical
entity
and
logical
entity
in
order
to
manage
security
and
Trust
for
multiple
partition,
multiple
participants
that
can
Works
after.
A
Together,
so
my
vision
should
be.
That
is
the
idea
of
to
to
have
a
trust
for
some
some
program
in
the
in
in
the
coin
system,
and
so
it
means
the
security
is
multi-tenant.
It's
more
collaboration
of
multi-tenant,
and
the
idea
is
that
is
that
you
can
you
don't
rely
on
a
single
entity
like
a
mobile
operator
or
you
may
have
several
entity,
but
for
me
the
the
instance
of
this
the
facts
to
be
multi-tenant.
It
means
you
have
a
more
multi-purpose
source
of
trust
in
your
system.
H
H
Thought
I
put
myself
in
the
queue
all
right:
Dave,
Aran,
Network
systems,
research
and
design
in
MIT,
I'm
struggling
with
a
different
aspect
of
this
I'm
wondering
if
you
could
shed
some
light
on
this
I.
Take
these
slides
and
I
say
what's
different,
then
if
all
I
had
was
just
a
whole
bunch
of
hosts
on
a
broadcast
network
things
different,
it's
not
really
has
anything
to
do
with
networking.
It
has
to
do
with
how
you
do
authorization,
authentication
and
security
and
Trust
among
Computing
elements,
no
matter
where
they
are
or
how
they're
interconnected.
H
A
A
H
A
A
H
A
A
A
Because
part
of
your
data
will
be
managed,
for
example
by
equipment,
you
don't
know
the
the
level
of
trust
of
your
equipment,
the
ability
to
to
to
manage
the
Integrity
to
to
to
to
to
to
to
to
keep
the
secret
secret
you.
You
don't
know
that
if
you
have
a
single
tenant
system,
it's
been.
If
you
are
trusting
Microsoft
Azure,
you
put
your
data
in
Microsoft
Azure
and
you
say:
I
am
okay,
because
this
is
managed
by
by
Microsoft
Office
this
this
infrastructure.
A
E
Eric
nordmark
I
didn't
figure
out
how
to
click
on
the
right
button
in
on
the
web
page
to
put
myself
on
the
line,
maybe
maybe
I
I
can
try
to
throw
something
in
to
clarify
this.
So
I
think
that
there
is
a
difference,
as
you
say,
it's
not
about
multi-tenant
or
not
it's
about
who
has
physical
control
of
the
boxes
that
you're
running
on.
H
A
E
Don't
have
to
worry
about
it
and
there's
a
question
about
here
whether
you
do
have
to
be
more
explicit
about
that
I'm
having
different
sort
of
owners,
operators
of
network
equipment
and
and
how
to
can
actually
manage
the
different
levels
of
trust.
I
have
into
those
owners
of
the
of
the
hardware
that
can
potentially
Tinker
with
it
right
and
then
on
top.
B
B
Yeah
but
then
we
go
into
the
ipfs
and
the
and
the
blockchain
world
and
ipfs
I've
tried
to
do
things
like
that.
It's
another
working
group
here
that
is
the
pipfs,
don't
come
to
our
group,
but
it's
there
and
then
you
get
also
in
dlnrg
the
inrg,
where
they
had
a
presentation
on
blockchain,
so
I
I
think.
Maybe
that's
where
the
thing
is
going.
D
A
Mobile
operator
networks,
all
the
equipment
managed
by
the
mobile
operator,
and-
and
so
it's
mean
it's
a
kind
of
physical
Trust
and
or
the
Security
apply
both
for
hardware
and
software
or
the
security
applied
to
things
that
are
managed
by
this
single
tenant
and
entity
and
I.
Believe
in
my
opinion,
coin
is
more.
Some
things
could
is.
A
This
is
distributive
and
collaborative
so
it's
more
difficult
to
reach
a
level
of
operational
Trust,
and
this
also
part
of
the
security
model
of
that
the
security,
secure
communication
and
keying
in
order
to
establish
the
relationship
with
the
pnd.
But
it's
also,
it's
also
deals
with
the
issues
that
you
will
have.
If
you
have
a.
B
B
B
B
You
know
how
this
coin
is
different
from
you
know:
standard
key
management
systems,
all
the
issue
of
discovery
which
in
multi-tenant
and
also
I,
would
say
in
multitrust
domains,
is
going
to
be
a
big
issue.
So
if
I
need
dated
or
if
I
need
Computing,
that's
not
in
my
domain
can
actually
even
Discover.
It
so
I
think
there's
a
lot
of
stuff.
That's
interesting!
There.
A
K
K
Okay,
this
graph
is
about
the
network,
programming
and
I.
All
we
can't
think
it
is
probably
a
coin,
and
we
assume
that
this
network
devices
can
do
some
Computing
jobs.
Besides
just
forwarding
the
packet.
We
can
see
that
types
all
the
nodes,
including
the
Ingress
one
user,
one
one
three
four
and
egress
for
the
users:
two,
they
are
on
the
path
for
flow
and
they
can,
if
it
is
specified
to
do
some
work,
they
can
do
a
function.
K
Of
this
seed
list
in
the
packet
header
or
IPv6
in
the
left,
we
can
see
a
series
and
the
facilities
indicated
that
in
each
each
device
in
each
Network
device
in
the
past,
they
can
do
a
function.
But
we
still
created
the
strict
Network
program
maximum,
because
we
must
we
must.
The
function
must
take
place
in
a
specified
node.
K
K
K
K
This
this
page
is
about.
We
think
if
we
need
the
function
to
be
done
in
the
path,
but
some
coordination
is
needed
for
the
devices
along
the
path,
because
we
do
not
want
a
function,
be
implemented
twice
or
more.
So
a
coordination
working
domain
needed
here
everything
that
we
can
check
on
each
node
along
the
path
it
can
attack
on
whether
we
have
a
job
to
do.
K
For
example,
in
this
router
1
this
it
it
can
find
that
we
have
two
jobs
to
do
and
they
it
can
choose
to
finish
one
and
then
the
flags
change
to
all
this
router
3.
We
have
only
one
job
left
at
the
final.
They
reached
the
egos
tool,
we
kind
of
see
that
all
the
jobs
are
done
and
because
we
are
not
sure
which
device
will
do
the
job.
So
we
think
we
can
consider
to
add
the
this
and
these
functions
in
the
of
the
Hope
head
extension,
header
or
iPhone
6..
K
K
And
do
not
want
to
perform
any
functional
tasks,
the
it
will.
It
can
dust
forwarding
the
packet
and
some
other
nodes
have
the
high
availability
to
do
the
job.
They
will
do
the
job.
According
to
the
we
call
it
a
bitmap,
because
already
we
can,
then
we
can
set
a
one
in
the
speed
map
to
indicate
that
there
is
a
job
and
when
the
job
is
done,
it
can
change
the
one
to
zero
so
that
it
has
ended.
K
This
is
this
some
scenario:
we
have
a
thought
for
this
loose
Network
programming,
maximum,
for
example,
Network
nodes
can't
detect
this
attack
by
analyzing
the
characteristics
of
the
flows
and
drop
the
packet
of
the
suspect
flows,
but
a
node
can
only
do
it
when
it
is
not
building
okay,
so,
for
example,
before
is
not
really
I.
Can
you
can
do
some
details
attack
analysis
on
one
on
some
aspects
and
finish
the
job.
K
About
whether
to
do
the
job
according
to
his
leaders,
the
main
job
is
still
the
forwarding.
We
propose
a
job,
because
we
think
that
the
students
in
the
network
is
very
Dynamic,
so
we
want
to
enable
the
network
devices
to
do
this
local
decision
about
whether
to
do
the
job,
and
we
want
to
find
the
more
use
cases
and
have
some
comments
from
the
research
group.
I
think
that
is
actually
the
last
page
thanks
for
listening
and
a
welcome
for
Commerce.
Thank
you.
I
K
B
That's
amazing,
as
a
person
emergency
from
I
changed
in
universities,
but
it's
fine,
I'll
I'll
tell
people
later.
It's
I
I
also
have
this
question
of
what
is
the
research
in
this?
Where
do
you
feel
that
there
is
research,
I
I,
don't
think
I
I
really
understood
what
this
loose
Network
programming
is,
but,
like
Colin
said,
the
idea
of
invoking
some
entity
to
to
help
me
deal
with
my
my
network
issues.
You
know
there's
a
lot
of
stuff
like
that.
B
B
To
hear
how
this
is
is
different,
is
it
an
evolution
of
these
things
which
it
wouldn't
be
research,
so
it
would
be
interesting
to
see
what
is
like
Colin
said:
what
is
the
research
because
I
see
this,
as
you
know,
an
architecture
with
some
interesting
ideas?
Maybe,
but
where
where's
you
know
what
you're
doing
research,
what
else?
What
other
work?
Where's
your
state
of
the
art
and
how
are
you
going
beyond
that
state
of.
I
K
K
C
B
B
So
the
use
case
for
in-network
computing
has
been
something
that
we
started
very
much
at
the
beginning
of
the
research
group
just
to
have
a
number
of
good
applications
where
we
felt
that
in
network
computing
could
be
a
an
enabler
of
these
applications.
Actually
it's
very
funny
because
this
is
how
I
tell
my
students
is,
that
think
of
applications
and
think
of
what
would
happen
if
the
network
had
Computing
facilities
that
could
help
your
application.
So
this
is
a
little
bit.
B
B
Presenting
them,
so
this
use
case
structure
is
basically
what's
presenting
dire
there.
So
newborn
experience
on
things
that
are
already
existing,
so
mobile
application
extended
reality
and
immersive
media
and
interactive
Performing
Arts.
A
lot
of
these
things
are
currently
suffering
a
lot
of
times
from
having
a
strict.
B
We
want
to
put
new
coin
systems,
there's
a
lot
of
people
who
are
looking
at
adding
in-network
control
and
in-network
capabilities,
especially
for
time
sensitive
applications,
some
of
them
being
extended
reality
and
immersive
media,
but
also
everything
related
to
internet
of
things
and
Industrial
control.
Large
volume
applications.
Actually
there
was
there
is
interesting
work
also
in
ICM
RG
on
this.
But
what
do
you
do
when
you
have
applications
that
are
gathering
an
awful
lot
of
information,
be
it
Vision
in
Industrial,
Systems
to
all
kinds
of
iot?
B
You
may
want
to
have
in
network
at
that
point,
either
in
the
fog
or
in
the
edge
capabilities
to
do
the
native
the
data
reduction
there.
Industrial
safety
is
another
one
again
because
you
may
want
to
do
things
across
the
network
in
terms
of
where
some
of
the
safety
features
should
be
improving
existing
coin.
Obviously,.
B
The
the
existing
coin
were
very
much
into
content
delivery.
Network.
We
mentioned
all
the
data
centers
and
everything,
so
that
was
there.
We
want
to
make
that
better
compute
compute
fabric
as
a
service
I
mentioned
Noah
zilberman
before
she's,
the
one
who
came
up
with
this
ID
of
the
Computing
service
provider,
and
that
actually
could
be
something
that
could
be
improved
and
moved
into
even
more
type
of
services
and
again,
as
we
move
into
a
much
more
data-driven
Network.
All
of
these
things
become
very,
very
important.
B
Virtual
Network
programming,
Next
Generation
nfv
I,
was
told
it's
like
it's
10
years
of
nfv
this
year
and
a
lot
of
virtualization
make
start
using
in-network
capabilities
and
my
best
one
and
the
one
that
I
currently
be
involved
with
is
all
the
distributed
Ai
and
what
happens
when
you
start
having
Federation
of
AI
capabilities
and
that
you
need
not
only
distributed
networks
but
the
way
to
deal
with
these
incredible
amount
of
data
again.
So
we
worked
a
lot
on
that
next
slide.
B
So
the
current
status
is
I
think
we're
finished
we
have
and
we
invite
the
group
to
really
look
at
it
all
our
use
case
description
are
aligned
with
the
taxonomy,
which
is
the
next,
the
next
presentation,
because
we
had
in
order
to
describe
the
use
cases
we
had
to
agree
on
a
taxonomy.
So
we
did
that.
We
think
that
again,
from
our
point
of
view
and
again
talking
for
the
whole,
the
whole
team
we've
considered
that
this
is
finished.
B
B
So
now
that's
the
other
one
like
I
said
at
first
for
those
of
you
who
followed
or
or
saga,
we
had
everything
in
one
in
one
draft
and
we
realized
that
the
defining
of
a
taxonomy
and
a
terminology
and
a
way
of
the
semantically
describing
all
these
requirements.
That
is,
we
we
had
to
Define
so
much
terminology
that
it
was.
It
was
obvious
that
we
needed
another
document,
so
we
split
the
terminology
outside
of
the
use
cases.
Next
slide.
B
One
question
that
we
have
to
the
RG
is:
should
we
continue
managing
this
document,
or
should
it
be
taken
over
by
someone
else,
Ike
is
is
terminating,
is
is
working
on
his
PhD
and
obviously
there's
a
lot
of
there's
a
lot
of
authors,
but
we're
figuring
out
is
this:
are
we
the
right
group
to
continue
this?
We
should
should
we
collaborate
much
more
with
with
Dirk
and
and
York
on
the
other
coin,
RG
directions,
and
so
this
is
a
question
that
we
have
for
the
group.
B
We
don't
know-
or
we
are
not
sure
the
goal
of
the
document
I
think
we
would
well
personally
I
think
it
should
be
a
living
connect
collection
of
terminology,
but
are
we
going
to
ever
have
a
final
version
of
this?
That's
another
good
question
again,
and
it
goes
to
the
last
point,
which
is:
what
are
the
changes?
What
are
the
additions?
B
We
have
a
number
of
new
documents
of
new
approaches,
the
whole.
We
probably
need
a
whole
section
on
security,
type
terminology
and
other
related
topics,
so
how
we
again
it
goes
back
to
who's
going
to
take
over
the
management
and
and
what
should
be
the
scope
of
this
because,
of
course,
terminology
can
grow
like
crazy
and
I
can
tell
you
that
in
our
previous
research
group
and
network
coding,
I
took
us
a
number
of
years
to
figure
out
what
the
real
terminology
would
be.
B
And
we
would
like
these
to
include
network
interface
cards
and
switches,
but
maybe
at
one
point
other
type
of
Hardware
as
it
becomes
available
and
and
there's
a
lot
of
work
on
that
also,
and
we
use
P4
or
other
languages
again,
because
we
are
very
much
aware
of
the
limitations
of
B4
and
that
at
one
point
maybe
we
would
need
to
have
better
languages
to
do
it.
We
defined
a
execution
environment
where,
essentially,
where
we
could
execute
a
function.
B
B
All
these
devices
a
program
instance
well.
That
was
actually
something
very
simple.
It's
one
instance,
and
it
is
most
likely
communicating
with
another
instance
because
again
of
we're
talking
about
this
in
terms
of
a
network
and
we
Define
the
coin
experience
I
I'm,
not
I'm,
you
know,
I,
think
that's,
fine
and,
and
we
may
and
I
think
it.
It's
important
for
defining
what
an
application
gets
from
coin,
and
so
we
defined
it
as
something
new
and
probably
better
that
is
brought
when
we
have
coin
capabilities.
B
I
call
it
again
to
my
students:
I
call
it
coin
capabilities
on
the
life
of
a
packet
so
from
where
a
packet
is
being
generated
to
all
the
way
to
where
it's
going.
So
is
the
coin
experience
making
the
life
of
a
packet
and
the
life
of
the
application
that
needs
all
these
packets
better
next
slide,
but
that's
that's
just
a
a
a
an
overview.
B
So
so,
after
we
did
the
the
the
use
case
and
the
terminology
well,
the
next
one
is:
are
we
going
to
stop
at
describing
things
or
are
we
going
to
do
more
research,
because,
right
now,
this
stuff
is
what
we
have
is
is
essentially
opening
up
some
good
questions
and
that's
actually
that's
what
the
use
case
document
does
very
well
is
opening
up
directions
for
research,
so
we
we
started
thinking.
Should
we
do
another?
B
Well,
let's
put
it
this
way
when
we
did
the
work
at
the
beginning,
we
had
in
the
use
case
not
only
the
requirements
but
also
elements
of
analysis,
and
it
was
uneven.
Some
of
the
use
cases
had
already
more
in-depth
analysis
existing
some
less,
and
so
the
document
was
a
bit
of
was
a
bit
messy,
so
we
decided
to
take
the
analysis
again
out
of
the
use
case
document
and
making
making
itself
into
another
another
document,
and
that
one
is
is
fairly
new.
Next
slide.
B
So
so
this
this
is
to
go
beyond
the
descriptions
and
to
allow
to
analyze
opportunities
to
analyze,
research,
questions
and
requirements
that
com
that
actually
and
those
requirements
that
are
common
and
actually
could
give
in-depth
research
direction
for
coin
based
on
again
this,
this
Ensemble
of
of
different
use
cases,
so
it
started
with
and
actually
that's
the
status.
What
we
have
we.
B
Actually
we
found
that
there
are
good
research
questions,
but
we
would
like
to
go
into
deeper
analysis
and
the
little
ASCII
thing
here
that
who
described
this
Esky.
Well,
it's
the
ASCII
thing
again
so
that
there
are
application
area
of
applicability.
We
have
again
identified
things
related
to
transport
application,
design,
data
processing
that
includes
Network
elements,
routing
and
forwarding,
which
could
be
helped
by
adding
in-network,
Computing
I,
would
say
more
and
more
also,
probably
load
balancing
that
which
was
at
the
beginning
and
congestion
control
is
probably
there.
B
B
What
are
the
enabling
Technologies?
There's?
There's
new
hardware,
Nvidia
and
Intel
are
involved
in
new
hardware
for
some
form
of
intelligent
Computing
inside
the
network
and
that's
actually
part
of
the
enabling
that's
not
research,
but
there
is
research
and
enabling
Technologies
and
what
would
be
the
Visions
for
corn
and
I
think
divisions
for
coined.
We,
probably
each
one
of
us,
has
one,
but
it
would
be
interesting
to
investigate
which
of
these
Visions
could
lead
to
more
in-depth
research.
B
Next
draft
next
draft,
next
page,
so
again
questions
to
the
RG.
We
started
this
again,
the
same
type
of
people,
the
same
group
of
people,
so
it's
always
us
so
who
can
take
over
managing
the
document,
especially
with
Ike,
who
may
graduate
soon
but
I
think
it's
an
important
document,
and
we
feel
that
it
would
be
important
to
do
these
analysis.
I
think
this
whole
presentation
today
showed
that
there's
there's
a
lot
of
people
asking
some
interesting
questions
about
what
are
the
the
validity
of
the
research?
B
That's
been
done
and
is
it
going
in
the
right
direction?
Obviously,
people
who
are
doing
the
work
we
have
our
own
view
of
of
where
this
could
be
going
and
it
would
be
I
think
it's
important
to
do
the
analysis,
because
right
now,
it's
interesting
to
say.
Oh
great,
you
know
in
network
computing
is
perfect
for
Federated
learning
because
we
need
it.
B
But
it's
another
thing
of
saying
it's
perfect
for
Federated
learning,
because
we
did
it
and
it
is
much
better
than
what
is
happening
or
by
Distributing
the
Computing
from
the
data
center
into
the
edge
Network
or
going
back
to
the
net
Edge
Network
which,
like
some
people,
have
said
it's
more
eco-friendly
and
it
takes
away
a
lot
of
the
big
big
ecological
impacts
of
data
centers.
But
again
that
analysis
needs
to
be
done
and,
and
we
feel
that
this
is
something
important
for
the
group-
I'm
done,
I
think
I.
B
Anyway,
I'll
look
at
the
I'll
look
at
the
other
questions.
Are
there
any
questions
and
again
we
since
Dirk
and
Ike
are
not
there.
I
would
recommend
that
people
read
the
read
the
the
drafts
and
that
questions
could
go
on
the
list
so
that
everybody
can
can
participate
and
but
again,
I
think
there's
I
think
they
did
the
take
home
on
these
three
three
drafts
is
one.
We
would
like
the
use
case
to
go
to
research
group
last
call.
G
Yeah
hi,
quick
so
on
all
of
these
documents,
also
the
the
coin
directions
that
you
meant
mentioned
earlier,
that
we
didn't
update
for
some
time.
I.
Think
one
general
question
for
the
group
is
how
many
documents
and
of
which
nature
do
we
want
to
push
towards
the
RAC,
editor,
Direction
and
so,
of
course,
in
general.
G
So
personally,
my
preference
is
to
yeah
focus
on
experimental
specifications
mostly
and
then,
if,
if
there's
any
other
important
and
say
mature
enough,
say,
architecture,
work
or
some
analysis
work.
Of
course,
that
could
be
considered
as
well,
but
I
could
see
it
could
be
problematic
when,
if
there
is
a
it's
like
too
much
information
and
material
that
that
we
try
to
get
published.
B
This
is
this
is
agree,
but
with
draft
was
was
very
important
for
this
group.
So
you
know
I
I,
think
personally,
I
think
also
in
the
name
of
Jeff
and
and
Eve
who
were
expecting
this
to
go
further.
I
agree
that
we
shouldn't
you
know
you
shouldn't
spread
ourselves
to
thin,
but
for
example,
about
what
a
month
ago,
who
was
looking
for
information
about
coin
and
I
would
have
loved
to
be
able
to
direct
them
to
your
draft.
B
So
do
my
my
answer
is:
maybe
we
don't
need
to
make
that
decision
now?
Maybe
at
one
point
we
will
bring
back
the
terminology
in
the
use
cases,
and
maybe
the
use
case
analysis
will
go
back
in
the
use
case,
but
I
think
for
I.
Don't
know
if
we're
I
don't
know
if
we're
there.
You
know
because
I
I
think
part
of
the
discussion
is
that
we
have
a
number
of
interesting
things
that
were
started
and
Eve.
B
G
I
H
I
I
B
I
I
I
agree:
it's
very
useful
to
have
have
this
material
and
to
be
able
to
point
people
at
it.
I
I
will
just
encourage
you
to
be
creative
about
how
you
do
that
and
address
are
not
the
only
mechanism
for
doing
it,
and
sometimes
they're
very
appropriate
and
pushing
things
to
rfcs
is
very
appropriate
and
sometimes
for
a
research
group.
There
are
other
ways
which
achieve
the
goals
better.
D
C
Sure
can
you
hear
me
sure?
Yes,
thank
you
for
your
patience.
I
had
an
email
change
and
there
was
some
sort
of
a
large
authorization
problem
that
prohibited
me
from
logging
into
me,
Deco,
which
we
solved
in
the
background.
That
said
in
terms
of
going
forward
for
this
group,
one
of
the
things
that
we
have
talked
about
with
regularity
but
haven't
delivered
on
is
to
have
perhaps
an
interim
meeting
where
we
do
a
little
bit
of
this
retrospective.
C
Where
are
we
now
compared
to
when
we
first
began
and
I
I
think
it
actually?
In
light
of
this
past
discussion
or
dialogue
between
Colin
and
moon
Jose
about
how
and
even
Dirk
is
you
know
how
to
use
the
working
group
in
terms
of
creating
specs
versus
creating
an
organization
of
collections
of
research?
C
We
would
like
to
set
an
interim
between
now
and
July
to
do
that,
and
perhaps,
alongside
of
that,
we
try
to
gauge
what
are
the
best
ways
to
be
creative
about
achieving
that
state
of
the
art
and
pointers
for
those
who
come
into
the
coin,
Arena
looking
for
how
to
navigate
it
and
how
to
get
started
and
how
to
level
so
I
think.
That's
all
I'll
say
here
for
now,
and
thank
you
for
running
such
a
great
meeting,
Mauritius
and
Jeffrey
without
Mattel.
C
D
B
Inside
the
room
and
and
replacing
us,
and
thank
you
very
much
to
everybody
and
we're
over
time,
so
I
think
your
your
point.
Your
conclusion
Eve
was
perfect,
so
it's
great
so
so
have
a
good
rest
of
day
for
the
people
in
Japan
and
it's
10
30
p.m.
Here
so
I
guess
it's
a
start
of
the
evening
or
end
of
the
evening
and
for
people
in
California.
We'll
have
a
great
evening.