►
From YouTube: IETF115-CAN-20221107-1530
Description
CAN meeting session at IETF115
2022/11/07 1530
https://datatracker.ietf.org/meeting/115/proceedings/
A
Okay,
so
first
is
the
node
well
and
I
hope
that
you
have
seen
this
page
before
you're
in
this
room
participating,
be
aware
of
the
knows
well
identified
here
next
page
so
Computing
aware
networking
is
a
pretty
big
area
and
this
has
been
I'm.
B
A
We
had
a
we
had
a
bath
two
ietfs
ago,
and
this
time
the
proponents
have
narrowed
the
scope
of
the
work
and
they
make
this
to
be
the
home
to
discuss
how
the
availability
of
computing
related
resources.
This
is
not
just
Computing
but
also
storage
or
something
else
associated
with
the
places
or
sites
which
hosts
the
compute
servers,
how
they
impact
the
routing
decisions.
A
So
mainly
forwarding
mainly
focusing
on
the
routing
aspect
and
forwarding
focusing
on
service
instances
that,
with
one
address,
going
towards
multiple
destinations,
typically
known
as
any
cast
within
one
administrative
domain-
and
here
are
some
useful
information.
Our
mailing
list
can
at
ietf.org
encourage
everybody
here
to
sign
up
subscribe
to
the
mailing
list.
If
you
have
not
and
there's
also
the
information
about
what
it
is,
the
GitHub
to
which
has
information
about
the
scope
and
some
of
the
experiment
they
have
done
by
the
proponents.
A
Okay
next
page,
please
so
the
focus
of
this
bath
I
want
to
make
sure
that
this
is
not
for
discussing
Solutions
but
primarily
to
demonstrate
the
problem
space
and
to
make
sure
the
community
is
aware
of
what
problem
to
be
solved
and
also
get
the
agreement.
A
Another
important
thing
for
this,
above
is
to
make
sure
the
scope
is
very
well
defined
and
make
sure
we
don't
avoid
the
ocean
to
do
everything
under
the
sky
to
solve
this
problem.
Okay,
so
so,
for
this
make
sure
everybody's
clear
that
this
path
is
for
problems
in
the
routing
domain
next
page,
please
so
here's
a
brief
agenda.
We
have
proponents
present
the
use
cases,
scenarios
and
problem
statements
and
there's
also
another
one
for
the
Gap
analysis.
Yes,
can.
C
I
interrupt
for
just
a
minute.
There
are
a
lot
of
people
clustered
by
the
exit,
probably
in
contravention
of
various
fire
codes,
I'm
just
guessing
there's
a
lot
of
open
seats.
C
If
people
next
to
open
seats
can
either
raise
their
hand
to
indicate
or
shift
in
or
something
that
would
be
helpful.
Thank
you
very
much.
A
A
So
with
that
I
think
we
are
ready
for
the
use
case
and
the
requirement
discussion.
So
let's
bring
up
foreign.
E
E
Unfortunately,
we
got
a
pretty
good
result
about
the
use
case,
but
not
gonna
get
the
consensus
about
the
solutions
and
we
also
received
36
issues
and
we
are
responsible
to
response
to
them
to
the
Middle
East
GitHub
and
also
copied
to
the
question
nodes
and
for
the
last
idea
meeting.
We
present
the
progress
in
the
RDG
area
and
tsbwg
to
to
promote
small
consensus.
E
So
in
baseball
we
focus
on
the
problems
based
on
the
use
case
based
on
the
levels
of
consensus
on
the
on
them
have
been
reached,
and
so
many
operators
and
vendors
are
interested
in
this
works.
So
we
also
hope
that
no
solution
discussed
in
this
world
exercise.
E
So
now
we
focus
the
camp
work
loading
area,
it
considers
utilizing
the
Computing
related
results.
Conditions
in
traffic
steering
decisions,
especially
quickly,
can
focus
on
impacting
rotate
loading
decisions
through
propagating
the
competing
results
metrics
to
interested
nodes,
for
instance
the
Ingress
nodes
answer
the
police.
E
So
here's
a
quote
from
our
responsible
lady
John.
He
said
the
outcome
of
the
previous
buff
was
that
there
were
support
that
the
problem
is
legitimate,
but
there
was
no
consensus
around
the
approach
being
proposed
for
Solutions.
So
therefore
guidance
this
time
around
words
to
focus
on
the
great
part,
the
problem
and
going
forward.
People
with
Solutions
can
make
the
their
case.
E
If
a
WJ
is
chartered
and
here's
our
considerations,
there
are
many
downloading
based
Solutions
out
there
and
there
are
other
IDF
working
groups
working
on
the
downloading
based
Solutions,
but
those
Solutions
are
not
enough
for
some
scenarios
which
are
demonstrated
later
and
we
are
making
a
case
to
have
a
very
narrowly
scoped
WG
to
address
how
remote
metrics
impact
floating
decisions,
because
we
don't
want
to
boil
the
ocean
next
class.
Please.
E
It
shows
a
rapid
development
of
the
integrated
ICT
infrastructure
different
from
the
past.
The
operators
are
offering
the
integrated
competing
and
network
network
infrastructure,
some
Syrian
nodes
has
been
installed
and
this
node
can
be
upgraded
to
visiting
and
then
at
completing
infrastructure
and
difference
is
that
more
diverse
Computing
results
needed
to
be
provided,
comparing
to
the
original
content
providing
service,
and
there
will
be
more
Edge
Computing
nodes
demand
manner.
He
has
a
typical
analysis.
E
Result
shows
that
so
many
Edge
Computing
nodes
will
be
settled
in
the
country,
aggregation
access
application
and
on
site,
and
now
somewhere
has
around
1000
outside
in
more
than
20,
more
than
200
cities
and
200
percent
increasing
rate
comparing
to
the
last
year
and
in
China
there's
at
least
2585.
F
E
Suppose
so
why
does
Edge
infrastructure
develop
so
fast,
because
users
want
the
best
experience
such
as
low
latency
and
high
availability?
The
emerging
new
Services,
intelligent
Transportation
shows
that
and
the
user
also
one
stable
service
experience
when
moving
around
different
areas
and
in
times
of
changing
demand.
The
mobile
network
brings
more
comedians
for
for
the
users
to
use
a
lightweight
EOS,
get
more
service
and
how
to
make
user
comment.
E
One
point
is
to
deploy
instance
for
the
same
service
across
the
virus
outside
for
the
better
availability
to
provide
the
functional
equivalency
and
another
point
is
the
under
traffic
dynamically
to
the
best
service
instance,
because
traffic
is
delivered
to
the
optimal
ad
site.
Based
on
the
information
that.
E
Computing
information:
at
the
same
time,
the
definition
of
best
may
be
service,
specific,
maybe
the
appropriate,
but
exercise.
Please,
though
we
have
this
out
about
to
to
meet
the
users
demand.
E
The
reality
is
competing
has
advantage
of
closed,
but
in
some
cases
the
closest
is
not
the
best
for
the
service
because
of
the
virus
of
the
company
status
innate,
the
closest
site
may
not
have
enough
results,
particularly
when
load
flag
shades
and
the
closest
style
may
not
have
enough
specific
results,
for
example,
support
folders,
big,
specific
Hardware
or
software,
and
which
will
explain
in
the
later
yeah
in
exercise.
Please-
and
here
we
use
three
slides
to
to
show
that
the
Trap
steering
based
on
the.
E
Computing
results,
variety
yeah,
so
it
shows
a
common
deployment
deployment
of
the
ad
Computing
The
Edge
data
center
in
the
metro
area
with
high
convenient
results,
provide
the
service
to
many
UVS
in
material,
especially
in
the
working
time,
because
there
was
Office
Buildings
in
the
metro
area
and
at
the
same
time
there
was
also
The
Limited
company
results
in
the
remote
to
provide
a
service
to
the
to
the
field
areas
close
to
the
remote
outside,
and
in
this
case
the
service
request
and
the
committing
results
messed
up
well,
and
there
was
only
some
potential
traffic
steering
if
needed,
and
please
remember
the
location
of
the
US
which
will
move
in
next
slide.
E
E
So,
in
the
non-working
time,
for
example,
in
the
weekend
of
the
daily
night,
the
UVS
will
move
from
the
metro
area
to
remote
Edge
because
because
their
house
was
close
to
the
to
the
remote
Edge
and
there
will
be
more
service
required
close
to
the
remote
Edge.
But
the
company
yourself
is
not
enough.
So
some
traffic
need
to
be
there
back
to
the
Metro
Center
and
it
is
only
for
one
or
two
days
can
justify
adding
servers
to
the
remote
site
next
size.
E
And
then
activities
to
to
to
make
make
the
traffic
either
need
to
be
steering
among
the
different
ad
sites
because
it
also
unplanned
and
you
can't
find
the
ability
of
of
the
of
the
US
move.
So
this
shows
another
piece
for
the
traffic
steering
yeah.
G
G
G
G
H
Traffic,
so
we
will
try
to
illustrate
this
with
the
use
cases
where
the
overall
idea
is
to
take
into
consideration,
not
on
I
mean
to
going
to
enrich
the
decision
combining
information
from
the
application
site
from
the
resources
side
in
the
edge,
but
together
combined
with
the
information
in
the
network
site
as
well.
So
next,
please
so
so.
H
For
example,
we
can
consider
this
of
computing
aware
a
materiality
virtual
reality,
so
in
in
any
service,
we
will
have
for
sure
to
access
to
one
application
to
a
service
function
or
a
number
of
service
functions
delivered
in
some
service.
So
we
will
have
some
dependencies.
We
will
have
dependencies
on
the
kind
of
Terminals
and
processing
times
I
spent
on
the
user
side.
Similarly
to
the
application
side,
and
also
we
need
to
have
into
consideration
the
time
for
reaching
where
this
application
is
running.
H
In
order
to
come
to
consider
the
delayed
budget
that
we
have
available,
we
cannot
do
too
much
on
the
terminals
height
or
in
the
process
height
for
sure
there
could
be
ways
of
optimization
and
that
part,
but
we
can
consider
those
as
a
static
and
priority
out
of
the
control
of
the
of
the
operator.
So
the
only
thing
what
we
can
play
with
is
the
the
reachability
how
far
or
how,
how
much
time
we
can
spend
from
the
budget
delay
in
reaching
the
unspecific
service.
H
So
in
the
sample
that
you
can
see
there,
we
could
consider
that
we
have
the
the
different
applications
to
in
three
different
energy
sites,
one
of
them
imposing
a
delay
of
99
milliseconds,
another
one
impossible
delay
of
four
milliseconds
and
a
third
one
imposing
a
delay
of
five
milliseconds.
So
it
would
consider
as
fixed
the
time
employed
by
the
client
in
in
order
to
processing
and
and
do
their
job
for
consuming
the
application.
We
see
that
the
only
option,
at
least
in
this
particular
case,
would
be
one
of
the
obviously
the
three
edges.
H
So
if
we
decouple
or
or
if
we
take
the
decision,
all
the
application
takes,
the
decision
decoupled
from
the
information
on
the
network
could
be
the
case
that
the
application
could
decide
to
select
one
of
the
other
two
instead
of
the
the
optimal
one
in
order
to
fit
to
the
delayed
budget.
Next,
please.
H
This
will
be
another
example
where
it
is
important
to
take
me
to
consideration
the
the
combined
information
between
the
network
and
the
application,
as
I
mentioned
before
by
my
pen.
Probably
in
what
here
we
have
a
case
where
there
is
a
number
of
cars
moving
around
and
having
available
a
number
of
edge
sites
for
retrieving
information,
there
will
be
applications
that
could
require
low
latency.
So
now,
there's
a
good
record
headband
with
so
in
principle,
we
could
have
these
applications
could
be
served
from
the
different
edges.
H
However,
and
not
always,
the
closest
environment
would
be
the
best
for
the
service,
maybe
because
we
don't
have
enough
resources
for
processing
the
information,
and
maybe
we
could
affect
some
decision
of
the
systems.
And
so
so.
In
this
case,
we
try
to
simplify
that
that
point
that
the
idea
would
be
to
select
the
best
one
with
enough
resources,
enough
capabilities
for
satisfying
the
low
literacy
requirements,
but
also
the
high
value
requirements
in
such
a
way
that
we
don't
affect
the
application,
but
also
respecting
the
overall
delay
budget.
H
So,
finally,
having
this
combined
optimization
between
the
the
transport
side
of
the
network,
side,
sorry
and
the
application
side
as
well.
So
next,
please
so
the
essentially
the
the
point
and
the
key
message
from
this
kind
of
services
that
combine
resources
and
in
the
networks
in
the
network
side
and
also
in
the
computer
side,
is
to
have
a
combined
optimization
and
a
combined
decision
that
could
considering
a
simultaneously
the
network
and
the
compute
part.
So
moving
or
transitioning
from
the
decoupling
decision
that
we
have
today
to
an
integrated
decision.
Considering
both
parts.
H
Next,
please,
so
that
the
categories
from
the
use
cases
that
I
commented
so
essentially
the
traffic
can
be
still
among
different
sites.
So
we
have
somehow
the
possibility
of
Designing
web
so,
and
this
opens
the
possibility
of
optimizing
the
such
a
decision
and
for
that
decision,
essentially,
we
need
to
consider
the
the
impacts
on
the
compute
side
and
also
the
network
side
next,
please.
H
So
this
slide
summarized
some
so
another
interactions
that
we
we
had
in
in
the
previous
months,
so
that
there
are
other
activities
in
in
other
bodies.
So
how
the
idea
here
would
be
to
leverage
on
ITF
or
defining
the
Solutions
in
within
the
routing
area
also
is
important
to
highlight
the
fact
that
the
compute
environment
is
something
very
much.
Dynamic
is
diverse
in
terms
of
solutions,
but
also
is
varying
around
the
time,
because
the
different
applications
could
be
consumed
in
different
resources,
so
is
so
how
very
very
much
Dynamic.
H
Let
me
just
highlight
the
the
latest
point
in
the
lower
part
of
the
slide,
so
we
receive
a
number
of
an
important
number
of
comments
and
issues
in
the
GitHub
36
of
them,
so
we
have
tried
to
to
answer
all
of
them.
This
does
not
mean
that
the
issues
has
been
closed.
The
discussion
is,
is
open
and
yet
ongoing,
but
just
to
I
mean
to
highlight
the
point
that
we
are
generating
too
much
discussion
on
this
topic.
A
Thank
you
very
much
is
any
clarification
questions
any.
I
One
so
I'm
curious,
APN
was
canceled.
What's
the
difference
between
can
and
APN
number
one
number
two.
What
was
presented
here
is
more
or
less
policy-based
routing.
You
define
a
policy
based
on
the
policy.
The
traffic
is
being
routed.
What
is
part
of
that
policy
can
be
multiple
things,
not
just
latency
as
an
example.
So
that's
the
one
part
the
use
cases
that
you're
presenting
especially
one
with
the
self-driving
cars
too
much
into
the
future.
J
I
A
C
A
Yes,
yes,
yeah,
so
that
we
do
have
big
longer
discussion
at
the
end
with
the
charter,
and
maybe
the
APN
may
be
difference
to
APN
will
be
covered
right
now,
just
to
the
use
cases
they
presented
here,
any
specific
questions,
so
we
do
have
questions
on
the
on
the
Queue,
so
kitchen.
K
Cisco
my
question,
which
I
also
put
in
the
chat,
is:
is
this
all
about
anycast
destinations
or
is
there
something
else
as
well.
H
K
Okay,
so
I
understand
it's
not
limited
to
any
cars,
in
which
case
this
is
about
picking
of
our
selection
of
the
destination
addresses
as
well
right
for
the
for
the
application
or
use
case,
and
then
my
question
would
be
isn't
that
outside
of
routing
space.
A
He's
saying
that
if
it's
not
about
any
cousins
is
outside
the
routing,
so
let's
defer
that
to
the
end
of
the
show.
So.
K
Yeah
just
to
repeat:
if
it's
about
applications
using
destinations
based
on
some
metrics
or
some
other
criteria,
then
would
that
not
be
outside
the
routing
space?
That
was
so.
Let's.
A
Discuss
that
afterwards,
so
next
one
Chen
Lee
on
the
list
essential.
M
Yeah
actually,
like
I,
think
we
should
focus
on
the
use
case.
Firstly,
in
this
presentation,
lots
of
questions
related
to
the
solutions
should
be
Jose
deferred
to
the
next
meeting,
maybe
yep.
Thank
you.
Okay,.
D
Okay,
so
very
early
on
when
we
started
to
do
routing
protocols,
we're
you
know,
implementing
in
things
like
er
grp,
you
know
measured,
load,
aware
routing
and
then
the
consumption
of
the
path
you
know
changed
the
load
so
that
we
would
get
into
a
lot
of
fluctuating
Behavior.
G
I
want
to
ask
cutting
for
one
second,
so
we
have
online
participants,
and
so
everyone
in
the
room,
please
also
use
the
queue
on
online
tool
to
to
get
in
the
queue.
That's
signing
is
also
another
way
to
to
track
the
participation,
the
blue
sheets.
So
please
sign
in
even
if
you're
in
the
room.
E
Okay,
so
for
the
technical
issues,
we
also
have
the
government
analysis
in
the
next
advantage
and
for
the
use
case,
we
think
the
two
typical
use
cases
we
showed
is
not
really
far
away
from
now,
because
there
are
some
deployment
in
the
city
and
we
can,
we
can
see
them
and
and
for
the
nvi,
it's
really
a
strict
demand
for
the
20
million
seconds,
and
it's
just
because
there's
no
joint
consideration,
so
that
may
be
some
of
circle
for
the
relief
to
provide
at
this
time.
E
But
we
really
to
to
get
the
use
case
that
is
broadly
to
see
in
this
stage,
including
the
edge
Computing
yeah
thanks.
Okay,.
N
Can
we
just
live
this
discussion
until
I
finish
my
presentation
on
the
Gap
analysis
and
the
rack?
They
have
laid
the
foundation
about
use,
case
problem
statement
and
then
now,
if
I
can
talk
about
the
Gap
and
also
requirement,
maybe
people
have
more
ideal
comments.
Thank
you.
O
F
Yes,
I'm
online,
okay
I
think
it's
important
to
consider
the
Computing
and
network
delay
together
to
provide
an
Integrity
surveys
SP.
In
addition
to
the
use
cases
he
mentioned
there
I.
There
is
a
industry
control
scenario
in
the
industry
internet
in
which
Computing
tasks
need
to
be
performed
in
each
cloth
or
Regional
data
centers,
and
these
tasks
need
to
guarantee
a
certain
type
time
delayed
to
achieve
the
current
industry
control.
So
I
agree
with
the
use
case.
A
Thank
you
so
your
name
and.
P
Affiliation,
please
probably
a
pretty
fundamental
clarifying
question
very
brief
here,
but
maybe
it
has
been
answered
already,
but
it's
not
so
clear
to
me
when
you
talk
about
applications,
so
what's
the
expectation
of
an
application
from
the
network
to
take
a
decision
right.
So
this
is
plain
resources
in
terms
of
compute
storage,
or
is
there
an
expectation
on
the
availability
of
a
function
that
uses
that
storage?
P
So
is
it
kind
of
serverless
communication
that
we
assume
here
or
do
we
assume
something
existing
in
the
infrastructure
more
than
the
compute
storage
resources,
which
is
a
network
or
an
application
function
which
is
relevant
for
your
question
that
you
had
on
the
slide
before?
What's
what
is
a
decision
criteria
right.
H
However,
you
could
call
it.
So
it's
such
a
way
to
guarantee
that
we
can
accomplish
the
slos
that
are
requested
by
the
user
at
the
end.
So
in
some
cases
would
be
the
accessibility
to
the
to
the
function
disorder.
This
is
the
guarantee
of
the
resources
made
it
before
doing
some
processing
and
guaranteeing
maybe
an
improved
behavior
of
the
application,
so
in
this
kind
of
decisions
can
be
taken
Alone
by
the
application.
So
this
identiful
is
thinking
on
on
a
streaming.
P
H
A
So,
okay,
so
Dirk
is
the
next
one.
Dirk
trollsen.
O
Q
There
was
a
question
about
the
use
cases
before
and
and
and
the
reference
to
the
use
cases,
maybe
being
a
little
bit
far
out
as
one
of
the
examples
we
run,
you
know
about
25
different
trials
in
a
in
EU
funded
efforts
on
edge
Computing,
where
we
asked
application
developers
to
develop
use
cases
in
AR,
VR,
Mobile
gaming
of
smart
tourism
to
deploy
their
application
in
a
in
an
intelligent
environment
which
happened
in
three
different
places
in
Europe
and
and
it's
interesting
to
see
when,
when
Lewis
makes
a
statement
about
the
expectation
of
best
instance
to
be
chosen
is
quite
different
in
those
use
cases
and
a
lot
of
these
application
developers
come
from
a
an
application.
Q
Paradigm
where
just
pushing
this
into
a
central
Cloud
gives
them
an
infinite
capacity,
but
now
suddenly
they're
pushing
it
onto
a
base
station
which
has
12
cores
and
suddenly
they
really
have
to
calculate
very
well.
You
know
what
kind
of
service
functionality
gets
there,
so
that
material
is
available
and
I'm
very
happy
to
I
know.
We
didn't
have
enough
time
to
bring
this
properly
into
the
use
case
graph,
but
the
site
is
in
the
use
case
craft.
These
are
very.
Q
R
M
A
N
N
Okay,
you
know
I
well
from
a
pong
and
Lewis
the
presentation
on
the
use
case
and
from
statement.
There
are
a
lot
a
lot
of
questions
already
so
I
can
expect.
There
are
more
questions
after
I'm
done
with
my
presentation.
N
With
that
in
mind
here,
I
have
to
emphasize
in
the
previous
slide,
have
laid
a
very
solid
foundation
to
demonstrate
two
important
and
critical
points.
The
first
one
is
there
is
a
problem
that
is
the
the
first
takeaway
about.
Okay,
something
has
to
be
done
to
optimally
steer
traffic.
Okay,
based
on
you
know,
you
can
call
intelligent
or
or
kind
of
like
a
component
those
type
of
things,
the
second
they
have
lay
down
from
there.
N
So
with
that
one
here,
I
will
give
some
a
gap
analysis.
This
is
based
on
some
existing
Solutions,
so
basically
like
okay.
So,
let's
look
at
some
existing
solutions
to
see
whether
they
have
at
least
the
partially
addressed,
but
my
still
exists.
Some
deficiencies.
N
Next,
one
yeah
before
I
delve
into
the
details
and
the
categories
into
different
existing
Solutions
here
I
give
a
slide
to
do
a
pre-summary
about
the
things
here.
I
give
the
examples,
one
some
categories,
not
just
well.
You
can
assign
that
as
a
categorize
example
like
the
DNS,
like
a
low
balance
like
battle
like
a
client
base
more
because
some
people
keep
mentioning
about
application,
how
application
can
do
this
can
do
that
and
the
Paul
is
the
best.
So
I
will
put
that
also
in
one
category.
N
So,
but
if
you
look
at
all
the
things,
there
is
some
common
commonality
to
those
existing
Solutions
about
the
exposure
and
the
usage
of
networking
conditions.
And
but
the
thing
is,
if
you
put
the
you
know
this
thing
together:
wake
up
in
like
interaction
approach.
While
later
you
you're
going
to
say:
okay,
we
call
our
it's
not
solution,
but
the
cancer
is
like
on
path.
N
Was
the
indirection
means
there
are
some
pretty
intelligent
work
that
have
to
have
to
be
done
in
order
for
the
successive
work
to
proceed,
and
also
for
all
the
existing
Solutions
I'm
going
to
present
next
I
I
will
I've
concluded
them?
N
Lack
of
the
integration
of
the
Computing
resource
conditions
and
the
networking
conditions
together.
So
with
those
things
well
here,
I
just
give
some
implication
about
something
we
want
to
achieve
by
the
existing
Solutions
are
lacking
off
like
a
dynamics
of
service
relations,
expolder
of
holistic,
metrics
and
also
efficiency
and
complexity,
to
reduce
complex
of
costs.
N
Next
one
please
yeah.
This
is
the
first
category
about
the
called
the
DNS
another.
Is
the
globe
service
a
load
balance
so
for
this
category
there
is
the
first
one
is.
The
point
is
that
early
bending
the
client
has
to
resolve
IP
address.
First,
let
me
move
from
it
here.
Okay,
the
client
has
to
resolve
IP
address
first
and
then
only
after
that.
Well,
the
client
got
a
chance
to
send
the
traffic
so
here
well,
remember.
N
They
also
use
some
some
effect
effective
way
and
also
to
get
some
real-time
property
or
characteristic.
So
this
one
is
it's
not
the
ideal
for
this
case
and
and
also
remember
here,
we
call
the
interaction,
the
DNS
things,
it's
a
resolver
again
as
the
well
well,
not
not
on
the
traffic
apart.
So
it's
it's
going
to
cause
the
message:
signaling,
the
signaling
burden
and
also
message
distribution,
all
kind
of
things.
In
order
to
remember
here,
it's
not
just
like
the
the
time.
N
Granularity
is
very,
very
small,
so
for
our
objective
and
also
more
some
health
check
for
the
for
the
service
instances
has
has
to
be
done.
So
this
is
also
a
deficiency
for
this
type
of
things.
There's
like
a
load
balance
through
the
DNS
work
like
the
Guardians
SD,
multicast
DNS.
We
do
gives
credits
for
the
for
both,
but
we
concluded
them
all
on
the
line
in
the
land
environment.
Well,
here,
while
we
are
looking
at
one
scenario,
so
there's
also
got
the
DNS
still
cache
these
things.
N
This
is
the
first
category
about
existing
solution.
Next,
please,
the
next
one
is
about
a
loading
balancer.
It
got
like
no
matter
like
layers
really
or
four
layer,
7
and
all
kinds
of
things,
and
we
categorize.
We
put
them
into
two
buckets.
The
left.
One
is
the
single
load
balance
in
a
single
side
for
all
the
silver
instances
and
the
right
figure
give
the
the
mod
load
balancers
distributed
or
geographically
distributed
among
different
Edge
sites,
for
they
they
do
have
a
prongs
like
a
either.
N
They
are
either
they
are
easier
to
deploy
to
be
deployed,
but
they
do
have
accounts
for
the
left
for
the
single
lb
case,
and
then
you
can
once
one
thing
in
a
single
point
of
failure
of
the
other
part,
like
called
a
pass
stretch
like
you
put
out
the
load
balancer
at
one
Edge
site,
where
you
may
have
service
instance
on
the
other,
well
under
the
server
side
or
other
sites.
N
So
that's
why,
if
you
look
at
the
left
picture,
you've
got
a
red
dotted
path
for
the
for
the
for
the
right
picture.
It's
the
mate,
a
load
balancer
case,
and
then
the
thing
here
you
get
the
orange
the
red
and
the
green.
So
they
have
to
guide
some
way
to
do
this
type
of
the
distribution
of
the
load
out,
while
the
the
optimal
we
keep,
calling
the
optimal
service
instance.
N
But
here
is
like
okay
for
all
the
three
different
lbs:
they
have
to
find
some
way
to
exchange
a
message
to
go
through
some
signaling
protocols
in
order
to
achieve
the
the
target
next
one.
N
Yes,
this
is
the
third
category
about
the
auto.
It's
an
application
layer,
traffic
optimization,
so
this
is
called
The
Off
pass.
So
for
the
client
in
order
for
client
to
get
the
best
the
optimal
service
instance,
it
has
to
ask
a
scheduler
which
Alto
server
the
server
will
do
all
kind
of
intelligent
work
to
do
the
the
task
to
collect
information
from
the
the
end
and
the
service
instances
from
different
Edge
sites.
Okay,
so
here
is
going
to
involve
remember
for
the
takeaway.
N
The
second
part
we
want
to
get
an
effective
and
effective
and
efficient
way
in
in
a
real-time
mode,
so
that
is
important.
Part
for
the
article
is
like
the
off
pass.
So
it's
still
it
still
Bears
the
the
deficiency
of
two
beers,
something
we
our
can
are
targeting
for
is
targeting
for
next
one
please.
N
This
is
the
last
one
or
client-based
solution
as
people
here
will
got
some
questions
on
the
application
part.
You
can
consider
it's
like
the
the
client
side
is
a
sort
of
like
application,
but
the
thing
here
is,
like
you,
ask
the
client
to
do
all
kind
of
intelligent
work,
so
it
has
to
find
some
way
to
to
get
the
real-time
information
and
so
that
the
client
can
do
the
the
selection.
Of
course,
they
are
a
lot
of
challenges
here.
N
So
the
things
like
the
first
one
actually
well,
we
will
consider,
is
the
most
important
as
the
largest
mobile
operator.
Actually,
this
comment
for
all
the
service
providers,
like
you,
do
not
want
to
expose
your
domain
of
some
confidential
information
about
a
third
party.
You
is
well,
maybe
you
can
argue.
N
The
UE
has
already
subscribed
to
our
Network
to
subscribe
to
our
service,
but
the
thing
here
is
like
there
are
some
roaming,
the
scenes
like
okay,
you
know
there
are
confidential
information
and
this
is
a
very
valuable
asset
for
us,
so
that
event
cannot
be
exposed
and
the
other
thing
is
like
there
are
so
many
applications
or
Services.
You
do
not
want
the
client
itself
to
do
the
the
work.
So
it's
a
remember
here:
our
can
is
not
that
specific
to
any
particular
service.
N
It's
just
to
try
to
give
a
kind
of
the
framework
or
architecture
those
type
of
things
to
solve
the
two
takeaways
in
the
previous
slide
and
the
previous
line
has
laid
out.
Okay,
the
next
one,
please
okay,
so
here
I'm
here
now,
I
have
when
I
have
gone
through
four
different
categories
of
existing
Solutions.
The
first
one
is
like
a
DNS
based
that
the
second
one
is
a
load
balancer
based
the
third
one
is
the
auto.
N
You
can
conduct
applications,
sort
of
things
and
and
the
fourth
one
at
the
client,
and
also
consider
a
sort
of
including
both
applications
and
some
other
ways.
So
here
for
the
the
existing
solution.
Until
now,
we
are
considered
okay.
Well,
they
partially
they
partially
at
least
I
will
acknowledge
they
have
they
do
they
do
address
partially.
N
The
problem
scope
we
have
laid
out
here,
but
when
you
put
all
the
all
the
objectives
together
so
like
Dynamics,
like
efficiency
like
complex,
the
challenge
of
complexity
and
accuracy,
as
well
as
the
Matrix
exposure
and
usage.
So
when
you
put
everything
together
here
and
they
are
more
or
less
for
each
different
categories,
you
can
see
the
Gap
analysis.
You
can
see
the
the
Gap
the
Gap
does
exist
in
different
in
different,
existing
Solutions
or
okay,
the
next
one
yeah.
N
So
with
all
things
here,
I'm
going
to
lay
out
some
requirement
for
the
can
next
yeah
the
this
is
the
goal
we
want
to
achieve.
The
the
thing
is
like
okay,
so
remember
in
the
Pro
in
the
previous
use
the
case
and
prompt
scope
slide
well
layout.
Okay,
we
want
to
find
the
optimal
one,
so
the
first
goal
is
to
consider
to
access
one
of
many
possible
instances
deployed
across
multiple
sites.
The
second
one
is
high
and
Matrix,
so
it's
going
to
jointly
consider
both
the
networking
and
the
Computing
resource
status.
N
N
Okay,
a
graph
I
think
I'm
going
to
answer
the
question,
but
never
I
think
I
have
a
2003
model,
so
yeah.
So
this
is
the
the
five
bullies
here
with
some
sub
bullets.
Are
the
potential
main
requirements
we're
going
to
we
we
are
laying
out
here
for
the
camp,
so
the
first,
oh
you
know,
is
to
support
access
to
available
website
dynamically
and
and
then
the
next
one
will
to
jointly
provide
consider
the
network
computing
metrics.
N
The
third
one
is
the
to
support
Effective
computing
resource
representation
and
in
cap
you
can
consider
like
a
message.
Signaling
and
some
protocol.
Extensions
well
well,
I'm
not
talking
about
solution
here,
but
this
is
some
sort
of
like
a
signaling
since
in
mind
here
and
also
we
want
to
support
the
session
continuity
and
the
service.
N
Continuity
are
called
SSC
here
and
if
for
the
for
the
can-
and
if
you
we
want
to
to
select
the
best
or
optimal
service
instance
for
some
ndis,
with
the
with
the
prop
with
the
service,
with
the
offering
like
the
the
roaming
service,
and
then
here
will
SSC
service
continuity
and
the
standard
and
the
session
continuity
will
be
very
critical.
So
this
one
is
also
in
one
in
as
one
of
the
requirements.
N
The
last
one
I
have
I
have
to
talk
about
in
the
client-based
solution,
and
then
the
provider
want
to
product
the
confidentiality
about
the
the
you
can
call
the
other
service
or
the
valuable
assets
all
kind
of
things.
N
Next,
one
please
okay,
so
here
is
some
questions
that
have
been
asked
through
the
the
the
past,
the
boss,
the
same
meetings,
the
email
discussion
from
the
diecast
email
discussion
can
discussions
and
all
kinds
of
the
things
here:
I'm
not
going
to
go
through
the
the
details,
but
it
is
showing
okay
people
there
are.
There
are
a
kind
of
questions
that
have
been
asked
and
next
one
I
think
that's
gonna
be
the
last
okay.
This
is
the
the
slide.
N
Okay,
that
is
showing
some
people
who
have
contributed,
who
have
commented
or
who
have
worked
one
way
or
the
other
on
the
can
so
by
this
moment,
combine
this
the
two
presentations,
the
punks
Lewis
and
the
mind
here.
Unless
the
way
below
we
can
enter
some
questions.
The
first
question
is:
is
there
a
problem?
Yes,
the
problem
scope,
clear,
yes,
they
are
from
this
slide
is
the?
Is
there
a
massive,
a
critical
mass
of
participate
participants
who
have
showed
interest
one
way
or
the
other?
N
A
You
now
we
go
to
the
questions
and
first
one
on
the
Queue
is
Tom
Hill,
please
State
your
affiliation,
I.
S
Will
thank
you
very
much.
I'm
Tom
Hill
from
BT
I'm,
very
interested
in
VR
I
think
it's
actually
really
cool
I.
Everyone
wants
a
good
VR
experience,
so
I
can
entirely
understand
why
we
need
to
Signal
information
about
good
processing.
Good.
You
know.
Abundance
of
processing
resources
on
the
network
cool
completely
understand
that
if
we
are
going
to
Signal
metal,
metadata
or
meta
information
about
resources
on
the
network,
we
can
do
this
in
a
few
ways.
S
S
It
strikes
me
as
very
strange,
though,
when
we're
being
asked
not
to
discuss
solutions
that
this
is
being
formed
in
the
routing
working
group.
It
seems
particularly
ominous,
but
also
if
this
is
going
to
be
a
routing
based
solution.
How
do
we
actually
inform
handsets
such
that
the
applications
know
what
to
do
because
there's
a
disconnect
there
between
having
multiple
applications
for
VR
or
ar
and
a
network
understanding
that
this
company's
Computing
resource
is
different
to
this
company's
Computing
resources.
N
Okay,
I
I
think
you
asked
a
question
about
how
the
network,
in
our
scenario,
should
inform
the
UE
or
end
devices
about
this
intelligent.
Is
that
right,
yeah.
N
You
know
the
things
like
I
can
answer
your
question
here
for
us,
the
mobile
operator
here
there
are
a
different
ways
to
do
the
work,
but
the
first
part
is
still
is
like
under
looting
domain.
We
have
to
jointly
use
the
you
know.
We
proposed
Computing
and
networking
metrics
to
do
the
first
level
of
the
work
only
after
that.
Well
in
the
mobile
space
you
know,
I
can
I
can
say
like
the
3gpp
5G,
they
have
a
defined.
We
can
call
the
urfp
those
type
of
things
that
is
some
rule.
N
They
are
going
to
apply
at
the
UE
side.
The
things
like
you
may
people
may
argue
about
the
ursp
say:
okay,
this
is
like
a
pre-configured
rule
at
the
UE,
it's
not
dynamic
or
it's
configured
by
the
by
the
PCF.
It's
not
the
damage
either,
but
there
is
some
certain
ways
like
in
the
edge
Computing
like,
and
the
do
is
mentioned
for
the
in
the
Computing,
the
the
domain
there's
another
called
AF
guidance.
N
So
basically,
it
does
have
some
sort
of
intelligence
and
if
AF
can
provide
intelligence
to
aware
an
EF
and
to
a
PCF
and
then
to
the
UE
for
this
type
of,
although
it's
not
the
complete
the
real
time,
but
it
does
have
something
like
that.
But
since
for
AF
guidance
here
you
know
us
ASP
costs.
You
know
for
the
mobile
operator,
normally
they're
going
to
provide
the
service.
It's
also
like
a
ASP
or
maybe
contract
with
some
third
party
AFP.
So
this
the
cad.
That
actually
is
the
good,
very
good
point:
internet.
N
N
Right
but
the
thing
here
you
mentioned:
okay,
you're,
going
to
kill
the
UE
or
kill
some
other
device
about
the
things,
no
matter
what
something
some
intelligent
things
you
have
to
get:
first,
yeah,
okay
and
then
you
are
but
the
hotel,
okay,
you're,
not
you're,
not
going
to
ask
the
internet
device
or
you
eat.
Okay,
can
you
do
yourself
purely
no?
It's
not
it
cannot.
Do
it
remember
the
decline
based.
A
B
Thank
you,
Greg
Ericsson,
actually
interesting,
I.
Think
I
changed
my
question
standing
here
and
listening
to
your
discussion,
because
originally
what
I
was
about
to
ask
is
the
balance.
B
B
Okay,
so
that's
the
question
about
whether
you
consider
that
continuity
includes
their
stability
of
your
domain
in
second,
but
second
I.
Think
it's
a
larger
one.
Continuing
the
previous
discussion
is.
B
B
N
You,
okay,
the
the
fourth
question
regarding
about
the
the
SSC
service
at
the
session
continuity
here
that
seems
like
it's.
We
just
used
the
zooming
as
one
example
since
the
the
UE
drawing
will
be
very
common
case.
It
then
it
doesn't
matter
it's
a
5G
or
4G
or
even
a
model
3D
Park.
So
the
thing
is
here:
the
we
want
to
get
some
service
Affinity.
So
it's
like.
We
always
want
to
get
the
same
in-service
instance
that
is
going
to
offer
the
service
for
the
other
devices
for
the
device.
N
If,
if
you
talk
about
the
SSC
particular
for
5G,
you
think
you
mentioned
about
it
for
the
mobile
device
here
it
has
defined
three
different
SSC,
like
I,
say
one
two,
three
and
they're
going
to
have
a
make,
make
before
a
break
or
make
after
break
those
type
of
things.
So,
but
the
things
like,
if
you
use
the
different
SSD
mode
and
then
you
can
still
use
like
the
PS
PSA
ups
and
then
those
type
of
things,
but
no
matter
what
they're
going
to
voice
down
your
second
question.
This
is
not
3gpp
work.
N
Remember
I
keep
saying
okay,
that
is
just
the
one
example,
so
we
are
still
going
to
collect
information
from
Beyond
your
n6
interface
and
that
will
be
on
the
DNN,
so
that
is
still
going
to
include
the
collection
of
information,
the
singling
of
the
message
and
then
some
intelligent
that
has
that
has
to
be
applied
as
the
networking
node.
Only
after
that
we'll
be
able
to
consider
the
use
case
in
which
5G
of
3gp
can
be
one.
Thank
you.
G
Let
me
cut
in
one
second
change:
there
is
the
a
charter
presentation
and
then
we
have
the
okay
open
the
open
discussion,
so
you
want,
but
you
can
see
here
you
want
to.
Maybe
you
want
to
control
how
much
time
you
you
spend
on
answering
questions
or
comments,
or
maybe
say
that
bring
some
discussion
off
an
hour
to
later
discussion.
Just
a
reminder.
N
I
think,
let's
go
to
the
charters
in
the
charter
also
going
to
explain
something
about
like
the
the
problems
goal,
objective
background
and
then
the
timeline
of
those
things.
And
then,
after
that
we
still
have
a
sufficient
time
for
open
discussions.
M
Okay,
I
will
make
it
very
quick,
actually
from
Holly
I,
do
see
the
value
of
the
Gap
analysis.
Different
solutions
have
different
value
and
we
do
see
the
the
value
of
the
Gap
analysis
and
by
the
way
we
have
answer
again
again.
The
similar
questions
race
today,
no
matter
in
the
checkbox
or
here
right,
so
I
I,
encourage
people
to
read
the
draft
and
then
go
back
again
because
it's
really
hard
to
explain
in
the
short
term
it
waste
a
lot
of
time
for
us
right.
Thank
you,
okay.
Thank
you.
A
Okay,
so
next
one
is
Daniel
Daniel
Huang.
I
C
I
think
that
you
yeah
do
you?
Do
you
want
to
make
these
I
mean
there
are
people
in
the
queue
but
I
guess
I'm
gonna
request
that
people
take
themselves
out
of
the
queue
if
they
have
a
general
comment
and
leave
themselves
in
the
queue
if
they
have
a
clarification
question
about
what
you
just
presented
and
everything
else.
Okay,
now,
like
Jeffrey,
said,
there's
going
to
be
a
charter
discussion
and
a
general
discussion
sure,
let's.
N
A
Okay,
so
is
Daniel
still
there
to
ask
questions
and
Dirk
as
well.
T
Yeah
yeah,
the
general
one
from
ZTE
I'm
I'm,
trying
to
be
quick
and
when
we're
talking
about
the
off
path,
Solutions
like
TNG,
are
presenting
in
the
drafts,
and
it
seems
to
me
it.
It
makes
sense
it
why
it's
why
it
makes
sense
to
maintain
the
Computing
relative
resources
in
the
concentrate,
the
centered
data
centers,
and
to
make
it
they
concentrated
a
data
center
to
make
the
service
distance
to
selection
decisions.
T
But
when
it
comes
to
the
cilarious
the
world
the
Computer
Resources
has
been
distributed,
then
written
and
located
in
a
multiple
and
the
locations
in
the
quite
ubiquitous
way,
and
it
seems
to
me
it
will
be
more
better
and
reasonable
to
to
to
to
to
to
to
to
to
upload
the
the
the
decision
processed
into
the
routing
and
protocols
and
routing
systems.
Q
The
clarification
Dr
Austin
Huawei
I
can
also
encourage
before
please
read
the
reference
that
we
have
references
on
a
question
regarding
that
Tom
asked
about
what
needs
to
be
signaled
by
references
to
works
that
have
shown
that
signaling,
very
minimal
information
and
not
semantic
Rich
information
can
give
you
significant
performance
benefits.
We
have
a
site
meeting
tomorrow
on
Wednesday
showing
some
work,
that's
reference
in
the
Gap
analysis
craft
and
how
you
use
the
same
algorithm
layer,
7
or
layer
3,
and
what
are
the
differences?
Q
So
it
answer
some
of
the
questions
in
the
Gap
analysis.
If
you
do
even
compute
awareness
and
you
do
it
at
layer,
seven,
what's
the
performance
benefit
or
the
performance
penalty
compared
to,
if
you
do
it
a
lower
level,
there's
there's
it's
discussed
in
one
of
the
site
meetings
so
encourage
people
to
read
these
things
because
they
do
answer
to
large
extent
of
questions.
Thank
you.
Thank
you.
So.
G
Before
the
charter
discussion,
one
clarification
I
want
to
make
here
so
indeed
the
people,
some
people
believe
that
there
is
solution
outside
routing
and
indeed
we're
not
going
to
talk
about
Solutions
here
in
this
path.
But
indeed
this
path
is
focused
on
the
routing
side,
so
that
that's
what
what
you're
going
to
see
in
charter
is
about
the
routing
aspects.
O
E
Okay,
it's
the
scope
and
the
charter
statement
for
okay,
so
here
just
to
clarify
our
very
important
issues
that
is
about
Ingress
Solutions,
not
not
to
put
everything
in
the
underlying
loading
system.
So
the
can
working
group
is
a
home
to
discuss
how
Computing
related
results,
including
buds,
not
limited
to
Computing
and
storage
capabilities
may
impact
the
loading
system.
E
Therefore,
new
mechanisms
and
protocol
extensions
may
need
to
be
developed
to
adjust
the
requirement
and
then
find
in
the
WG
it's
overall
about
the
can
WG
and
next
is
about
the
background
and
the
motivations
and
some
brief
use
cases
here.
So
most
modern
service
create
multiple
service
instance
that
are
often
geographically
distributed
in
multi-stites,
including
on-premise
or
cloud
data
centers
to
scale
up
the
service
offering
for
improving
performance
as
any
of
those
sites
can
serve
a
service.
E
The
Computing
related
conditions
at
those
sites
are
just
as
important
as
Network
conditions
drive
by
floating
protocols
and
traffic
engineering
for
the
service
performance.
This
is
especially
true
in
Azure
Computing
environment
when
supporting
service
such
as
VR
and
AR,
and
the
intelligence
transportation,
whose
traffic
may
need
careful,
steering
due
to
latency
bandwidth
required,
convenient
results,
sensitivity
and
mobility
and
therefore
join
the
conservation
of
network
and
convenience
status
is
needed
for
runtime
traffic
steering
among
multiple
exercise,
as
well
as
potential
surveys
deploy
deployment
and
Resource
Management
to
meet
the
demand
for
optimal
user
experience.
E
So
there
are
existing
solutions
that
utilize
and
points
results.
Information
for
selecting
those
server
instance
such
as
DNS,
Auto
or
layer
4
and
the
layer
7
load
balancers
coming
to
all
those
Solutions
is
exposure
and
usage
of
Network
conditions
for
decision
making,
while
none
of
the
existing
Solutions
integrates
the
Computing
results.
Conditions
with
the
network
condition
for
deciding
on
the
optimal
pass
here,
I
want
to
emphasize
that
some
people
go
early
into
the
solutions
discussion.
So
please
to
to
think
more
about
the
curve
analysis,
parts
to
fight
cans.
E
You
use
case
and
it's
Delirious,
so
allows
Ingress
nodes
to
consider.
Computing
related
results
conditions,
including
decisions
as
such.
They
can
WG
May,
develop
an
informational
document
or
create
a
design
thing
to
Define
use
case
government
analysis,
requirement,
character,
ization
of
the
Computing
relative
results
and
how
those
conditions
impact
the
Ingress
nodes
loading
decision
process.
E
The
fighting
will
enables
the
can
WG
participants
to
also
study
the
suitability
of
existing
loading
Technologies
such
as
bgp
further
so
potential
for
extending
existing
Technologies
to
Signal
committing
information
in
addition
to
network
conditions
will
be
studied
and
the
first
time
existing
Technologies
such
as
Auto,
NETCOM,
OEM
and
emerging
Telemetry
proposals
will
be
investigated.
As
for
the
purpose
for
of
joint
Computing
Network
information,
collection
and
usage,
so
that
the
thought
configuration
accounting
performance
security
requirements
may
be
met
yeah
next,
please
so
the
can
W
is
chartered
to
work
on
the
following
items.
E
First,
the
Quran
work,
maybe
documented
by
a
design
team
or
via
a
set
of
informational
internet
jobs,
and
not
necessarily
for
publication,
as
an
obviously
includes
a
problem
statement
for
the
needs
for
considering
both
Network
and
Computing
result
status,
use
cases
for
using
the
network
and
the
computer
cells
starters
to
benefit
the
steering
of
traffic
from
applications
that
have
a
critical
SL.
That
would
require
the
integrated
conversions
of
network
paths,
conditions
and
Computing
results,
requirements
for
commonly
a
grade.
E
Using
existing
mechanism,
including
identifying
the
limitation
of
existing
tours
in
fulfilling
requirements,
study
potential,
new
approach
for
the
control
and
data
plan,
solutions
that
can
fill
the
identified
gaps
of
existing
tools
and
solutions
and
studies,
I,
say
AP
and
suitability
of
existing
message.
Protocols
such
as
Network
and
the
data
models
such
as
young
next
slides,
please.
E
So.
Here's
Milestones
if
the
WJ
is
chartered
and
plans
to
adopt
the
cam
problem
statement,
use
case
graph
analysis
under
common
documents
in
John
next
year,
July
next
year,
yeah
and
then
adopt
the
cam
framework
and
architecture
document
then
adopt
at
least
one
control
plan.
Protocol
document
and
the
data
plan
protocol
document,
if
there
is
I,
mean
if
the
work
refers
to
the
data
plan
and
then
Summit
the
framework.
V
E
Architecture
document
for
publication
and
adopt
the
can
OEM
Performance
Management
yarn
protection
document
and
then
submit
all
of
them
in
the
in
the
field
following
Maybe,
four
or
five
years.
All
of
this
it's
a
current
milestones,
yeah,
it's
all
about
the
purple,
the
charter
statement,
and
maybe
now
it's
open
discussion
or
some.
A
Okay,
I
do
want
to
say
a
few
words
before
we
start
an
open
discussion.
As
you
can
see,
there
are
22
people
on
the
Queue
okay.
So
if
you
have
expressed
your
opinions
already,
you
can
save
the
time
for
people
who
haven't
expressed
opinions.
We
want
to
hear
diverse
opinion
from
wider
range
of
people
to
express
okay.
So
just
keep
your
questions
short
and
don't
repeat
the
questions
that
you
already
asked.
Yeah.
B
I'm
sorry
curricular
for
your
suggestion,
so
that
those
who
already
spoke
at
the
microphone
will
be
punished
for
speaking
early.
No.
A
Okay
always
make
sure
you
can
send
emails,
express
your
opinions
on
the
mailing
list
as
well.
There's
a
can
at
ietf.com,
so
I
will
want
to
hear
more
opinions.
Okay,
so
let's
go
through
the
list.
First,
one
is
Tony
Lee.
W
W
A
Thank
you
very
much.
Next
one's
Greg.
Q
B
I
have
a
question,
so
you
point
to
Computing
and
storage
capabilities
in
the
terminology
of
iitf
Transport
Network
slicing.
Do
you
consider
to
be
a
service
level
objectives
or
service
level
expectations.
B
In
your
proposal
in
your
Charter
so
use
basically
recording
that
discuss
how
Computing
related
resources
include,
but
not
limited
to
Computing
and
storage
capabilities
impact
welding
system
yeah.
So
these
capabilities
or
requirements
that
set
forth
by
the
service.
B
E
Transportation
I
got
the
question.
In
fact,
we
have
some
work
to
discuss
it,
but
we
don't
put
the
draft
here
in
the
book
because
it's
not
determined
I
mean
in
in
the
trips
level.
Service
level
or
service
level
may
be
possible,
but
it
depends
on
the
the
solution
prioritizing.
B
Okay,
so
if
it's
measurable,
then
probably
most
likely
it's
a
service
level
objectives
and
then
why
you
believe
that
they
are
special
comparing
to
latency
metric
geometric
packet
was
metrics,
so
why
they
don't
lock
the
wrong
in
the
framework
of
ITF
transport,
Network,
slicing,
okay,.
A
So,
okay.
A
Okay,
thank
you.
So
maybe
you
guys
can
think
about
it
and
answer
on
the
mailing
list,
so
basically
they're
saying
why
it's
not
integrated
the
general
and
other
delayed
attributes
so
yeah,
okay,
so
Greg
also
make
sure
you
record
your
question
on
the
notes
so
for
them
to
you,
don't
have
to
speak
to
say
because.
B
A
You
can
do
it
now,
yeah,
you
can
put
in
your
prices
clearly
right,
because
quite
a
few
people
are
non-native
speakers.
Sometimes
they
cannot
react
very
quickly
or
click.
Listen
clearly
of
the
questions.
Okay,
next
one
is
Shara.
X
Yes,
hi.
Thank
you
for
the
presentations.
I
actually
agree
with
a
lot
with
what
was
presented.
The
challenges
are
real
and
concrete
right
now,
and
also
it's
not
just
a
challenge.
It's
an
opportunity
because
the
geo-distributed
compute
resources
increase
the
capacity
significantly
reduce
the
cost
and
they're
not
equivalent
because
they're
geo-distributed.
X
They
cannot
share
a
microsecond
database,
so
they
are
not
the
same
so
because
of
the
applications
and
the
changing
loads
and
because
of
the
Geo
distributions,
Services
actually
may
show
up
behind
any
PE
in
any
of
these
locations,
and
we've
seen
that
and
also
clients
will
show
up
behind
any
IP
anchor
because
of
breakout
and
cost
reduction
and
private
5G
and
Wi-Fi,
and
also
and
I,
also
agree
with
what
was
said
with
The
Gap
analysis
load,
balancers
they're,
not
a
solution
because
they
assume
some
concentration
and
we
don't
want
to
concentrate
DNS
is
a
big
caching
issue
in
the
client
and
the
resolution
break
service
and
any
client-based
solution
will
aggravate
any
coherency
challenge
that
we
still
have.
X
But
the
problem,
even
though
we're
only
discussing
the
problem,
you
can
State
the
problem
such
that
we
need
routing
in
network
to
be
able
to
handle
the
fact
that
clients
can
show
up
with
the
same
identity
behind
any
PE
and
so
do
Services.
They
can
show
up
with
their
identity
behind
one
PE
and
then
in
the
next
meeting
behind
another
p.
X
G
Y
The
tool
didn't
like
me,
so
he
basically
put
his
hand
up
for
me.
So
that's
why
I'm
in
the
cube
now
so
I
have
a
question
so
so
far
in
the
all,
the
use
cases
we
draw
basically
a
set
of
computes
and
one
network
is
the
scope
of
what
you're
trying
to
do
limited
to
a
single
operator,
Network
and
single
com,
a
single
owner
or
operational
guy.
Y
E
Well,
well
thanks.
First,
we
focus
on
the
use
case
that
the
both
of
the
network
and
the
committing
results
May
belong
to
the
same
undertale.
That
is
simple
because
I
think.
Y
Okay,
because
because
in
my
view,
if
you
look
to
the
problem
space
today,
yeah
it's
I
constraining
it
to
that
is
not
practical,
because
if
I'm,
a
user
of
an
app
I
will
use
multiple
Network
operators
and
I
will
use
multiple,
computes
and
I.
Think
that's
the
space
that
we
have
to
solve
this
problem.
And
then,
if
you
look
at
that,
routing
is
the
wrong
space.
Y
Yeah
and
I.
Think
that's
how
I
look
I'm,
not
saying
that
the
problem
doesn't
exist.
But
if
you
look
to
the
elements
that
are
involved
and
that
that
have
to
deliver
the
service,
you
cannot
constrain
this
to
a
limited,
a
single
Network
operator
and
a
single
owner
of
the
compute.
We
have
to
think
about
this
much
more
broader.
M
Like
a
person
personally,
I
don't
agree
with
that.
We
should
do
or
walk
from
like
multiple
operators
that
will
make
make
the
things
very
complicated
right.
So,
let's
start
with
a
single
operator,
a
very
easy
starting
point:
yeah
so
I
agree
with
the
charter
of
the
use
cases
and
then
and
and
and
the
work
items,
because
it
clearly
describe
what
we
should
do
in
the
potential
working
group
right.
M
So
thank
you
for
the
work
and
and
by
the
way,
I
need
to
mention
this,
like
in
this
book,
we're
following
the
the
instruction
that
we
should
not
discuss
the
solution
related
topic,
because
there
will
be
an
endless
discussion,
so
we
wouldn't
get
nothing
today.
So
so,
let's
focus
on
the
use
case.
Thank
you.
A
M
Q
Hi
Dirk
from
Huawei
again
and
I
wanted
to
be
sure,
because
I
was
already
in
the
microphones.
I'm
gonna
get
punished,
but
I
do
agree
with
Shane
I
mean
they.
We
put
a
lot
of
work
in
trying
to
make
a
case,
but
I
think
then
there
was
not
right
all
right
good.
Hopefully
we
note
this
properly
that
somebody
actually
disguised
himself
I.
Q
Q
It's
getting
boring,
but
there
may
be
use
cases
where
maybe
spanning
both
compute
service
providers
and
network
providers
may
be
interesting,
but
again
only
a
good
discussion
on
the
use
case
and
the
problems
will
bring
that
out
and
and
and
and
that's
for
me,
even
more
an
argument
to
do
this
working
group
moving
forward
and
work
on
this
properly.
So
I
I
certainly
support
this.
Thank
you.
N
Here,
I
try
to
answer
the
question
who
asked
about
the
use
case
and,
as
Paul
has
answered
already
so
far,
we
are
going
to
focus
on
a
single
domains
like
not
going
to
cross
some
operators,
but
the
thing
is:
if
you
look
at
the
work,
look
at
the
thing
we
have
led
here:
it's
it's
not
going
to
prevent
us
or
prevent
the
community
from
doing
the
cross
operator
things
so
that
was
actually
about.
N
A
So,
thank
you
so
what
our
closer
to,
because
too
many
people,
so
next
one
is.
Z
I,
yes,
chair
and
I
I
think
there
are
many
ongoing
work
related
to
the
computer
network
convergence
and
there
are
also
many
requirements,
and
we
think
that
the
use
cases
and
also
the
problem
statement
is
clear
and
it's
meaningful
to
working
on
that
and
also
the
charter
is
clear
for
us
and
we
think
that
we
need
a
home
to
discuss
the
related
issues.
So
a
new
working
group
May
better
to
promote
the
work
thanks.
L
J
L
AB
Well,
and
in
my
view,
the
chatter
makes
it
very
clear
and
the
establishment
of
the
new
working
group
is
of
great
significant
to
the
promotion
of
the
Computing
aware
related
work,
and
all
research
group
is
also
engaged
in
related
research
and
acting
Health
lists,
and
we
are
looking
forward
to
the
future
work
of
care
in
ietf
I
support
the
establishment
of
the
new
working
group.
Thank
you.
AB
R
Hi
I
I
think
the
use
case
support
access
to
the
available
Edge
computer
dynamically
and
the
scenario
is
valid
now.
I
think
the
charter
is
clear.
So
thank
you.
K
I
think
there
is
still
not
Clarity
in
at
least
in
my
view
whether
what
is
exactly
the
problem
to
be
solved
and
as
a
follow-up,
whether
it
the
solution,
is
at
the
routing
layer
for
that
so
things.
The
charter
talks
about
things
like
or
seems
to
talk
about
things
like
path,
selection
right,
using
some
additional
metrics,
perhaps
for
traffic
engineering
or
tunneling
right
to
the
right,
egress
Point.
K
Those
are
all
probably
things
to
be
done
at
the
routing
layer,
but
then
the
use
cases
and
other
things
and
the
comparisons
give
me
an
impression
that
that's
not
what
this
is
about.
This
is
more
about.
You
know
a
selection
of
the
address,
because
we
are
talking
about
load
balancers
and
you
know
Alto
and
all
those
things
so
I'm
to
me.
K
It's
not
clear
that
this
there
is
a
what
the
charter
is
and
what
what
the
problem
is
going
to
be
solved
and
whether
it's
appropriate
to
solve
a
routing
layer.
AC
AD
Hydro
web
regarding
the
discussion
about
how
to
handle
multi-domain
I
think
we
we
have
to
start
with
single
domain,
but
we
could
always
put
this
in
the
charter
that,
whatever
solution
that
we
come
up
with
should
be
able
to
handle
in
future
multi-domain.
We
have
done
that
in
the
past
as
well
for
various
things
and
I
think
we
have
to
start
small
but
have
a
way
of
like
you
know,
re-chartering
for
multi-domain
in
future
and
going
back
to
the
original
point
from
Tony.
AD
Like
you
know,
can
we
keep
the
charter
of
the
working
group
without
constraining
it
to
routing
I?
Think
it's
going
to
be
very
difficult.
We
have
gone
through
those
Cycles
over
and
over
it.
AD
We
need
a
starting
point
and
we
need
to
constrain
that,
because
of-
and
we
have
to
agree
based
on
the
Gap
analysis-
that's
the
key
whether
the
there
is
a
consensus
that
what
has
been
highlighted
in
the
Gap
analysis
and
the
solutions
which
could
be
Alto
based
load
balancer
base
are
not
going
to
meet
the
requirements
that
the
proponents
of
the
buffer
making
and
if
we
have
an
agreement
on
that,
then
we
can
definitely
say
that,
like
no,
we
need
a
routing
based
solution
and
looking
forward
to
that,
thank
you.
Thank.
AE
I
I
say
in
the
use
case,
and
the
list
chart
is
clear
and
I
see.
It
will
be
better
if
a
new
work
group
was
from
to
promote
the
work.
Thank
you.
AF
Okay:
okay,
in
fact,
either
is
IB
Shepherd
for
this,
both
that's
for
a
long
time.
So,
regarding
this
regarding
this
Charter
and
also
these
some
of
the
questions,
I
have
this
following
comments.
The
first
one
I
think
this
regarding
this
Charter
about
these
use
cases,
in
my
opinion,
I
think
that
the
edge
Computing
and
VR
and
AR
is
just
this
example,
I
think
maybe
in
the
charter
country
generalized
from
in
my
opinion,
I
think
this
is
the
consensus.
To
some
extent
is
to
advertise
this
Computing
information
to
the
network,
but
this
is
the
use.
AF
Cases
can
be,
can
be
Beyond.
This
is
the
arvr
and
this
other
computer,
because
in
my
opinion,
for
example,
in
the
isd1,
this
CPE
device,
maybe
can
also
use
this
information
for
the
traffickers
during
for
the
sd1
and
also
maybe
that's
the
SLC.
We
can,
according
to
this
Computing
informations
for
the
service
training.
This
is
the
I
think
this
is
the
past.
AF
So,
in
my
opinion,
but
because
our
this
many
discussion
is
about
always
about
load
balancer
and
about
this,
the
Computing,
AR
VR,
so
I
think
in
order
to
let's
use
for
our
data
consensus
and
generalization.
In
my
opinion,
I
think
the
use
cases
of
this
mentioned
in
the
charter
should
be
generalized.
This
first
P
first
comment.
Second,
one
I
think
this
is
when
Network
knows
so:
I
think
in
fact
the
load
balancer.
Maybe
this
is
a
load
balance.
AF
This
is
the
original
Ip
packet,
but
for
our
the
network
node
when
we
just
receive
this
is
the
packet
we
may
encapsulate
the
tunnels
for
the
traffic's
Theory.
So
this
means
the
network
node
can
take
user
computing
informations
for
the
load
balancer
for
SLC,
for
maybe
other
sd1
Etc.
So
this
is
why
we
network
node
can
learn
this
information.
Maybe
this
is
for
more
application
scenarios.
This
second
point
a
third
point
that
people
talk
about
this
application
Level
or
service
level.
This
is
this:
the
Matrix
information,
in
my
opinion,
so
this
is
just
some.
AF
This
is
the
work
item
we
should
take
into
account.
So
this
is
some
people.
This
talk
about
this
common
Matrix:
these
are
Matrix,
we
direct
user
delay
or
sum
this
or
some.
This
is
the
panels
or
some
this
like
this
one
or
we
use
the
application,
Level
or
service
level
or
directly
based
on
the
is
this
level.
So
I
think
this
is
just
open,
I
think
this
way
we
discussed,
we
set
up
some
of
these
common
Matrix.
So
this
is
the
third
point-
and
this
is
a
large
point.
AF
I
think
this
is
the
scope,
and
this
is
to
discussed
about.
In
my
opinion,
this
is
the
other
Computing.
This
is
the
typical
scenarios
as
in
the
cloud-
and
this
is
the
cloud
and
the
Computing
and
this
network
can
be
controlled
by
one
of
readers,
but
I
think
this
has
no
effect
to
define
the
common
Matrix.
So
from
my
opinion,
maybe-
and
also
this
can
find
this
scope
to
this-
the
Computing
and
this
network
in
one
operators
from
this
beginning
so
later,
maybe
think
about
more
complicated
scenarios.
AF
If
this,
if
this
involves
these
third
parties,
so
that
we
must
discuss
more
about
the
security
or
maybe
some
other,
this,
the
more
this
work
may
be
done,
but
I
think
the
current
space
I
think
if
we
consensus
to
learning
this
Computing
information
and
set
up
this
Matrix
and
I
think
this
is
I,
think
the
scope
can
be
confined.
Okay,
that's
all.
A
AF
A
T
I'm
from
City
I'm,
sorry
I'm
from
City.
The
first
thing
I'd
like
to
make
a
minor
clarification
about
the
charter,
which
is
the
statement
about,
is
the
exposure
to
networking
resources.
Actually,
it
occurs
to
me
in
the
industry.
There's
also
exposure
to
Computing
resources
when
it
comes
to
DNS
and
grp
solutions.
T
What
is
absent
in
the
industries
reaction
is
the
joint
optimization
of
the
optimization
and
exposure
to
networking,
Computing
resources
and
and
again
when
it
comes
to
the
the
situation
where
the
Computer
Resources
is
more
and
more
distributed
in
ubiquitous
I
I
strongly
believe
it
will.
It
should
justify
some
capability
enhancements
in
routing
assist
systems
to
make
it
adaptive
to
to
the
new
established
Aloma
which
which
in
which
the
Computer
Resources
is
more
and
more,
is
distributed.
AG
Yeah
I
think
the
the
network
has
been
run
through
the
Loopers
status
for
a
long
time
and
for
item.
We
have
event
many
tools
to
improve
the
network
performance,
but
in
Europe
and
that
increase
application
server
cannot
cannot
fill
the
fill.
Our
effort
I
think
it
is
time
to
open
the
network
and
incorporate
incorporate
the
hardware
information
to
improve
the
performance
or
of
the
application
server
and
all
the
user
and
your
exported
experience.
Okay
and
one
small
command
for
the
for
the
master.
Would
you
like
to
switch.
AG
AH
AH
AH
So
if
we
assume
that
Ingress
devices
providing
tunnels
or
in
caps
or
whatever
to
get
traffic
across
the
network
to
the
right
server
is
something
people
want
to
offer,
which
seems
sensible
and
desirable,
then
making
sure
there's
a
way
for
servers
to
provide
information
such
that
it
can
get
to
those
ingresses
devices
and
they
can
make
choices
that
use
the
service
information
and
the
network
information,
because
it's
an
Ingress
router.
They
see
the
routing
already.
AH
That
makes
sense.
But
when
you
talk
about
injecting
the
information
into
the
routing
protocols,
I
go
ballistic
because
it
doesn't
make
sense,
it
doesn't
work
and
there's
no
need
to
mandate
a
data
plane
protocol,
because
that's
not
what
we
need
to
do.
We
need
to
focus
on
getting
the
right
information
to
the
entities
that
can
make
the
decision
you've
said
on
the
list,
and
you
said
when
you
started
the
presentation.
This
is
about
the
Ingress
devices.
AH
Being
able
to
make
the
decision,
if
you
write
the
charter
around
that
I
think
we
have
something
we
can
work
with.
We
can
then
figure
out
what
area
it
belongs
in
I,
don't
care,
but
putting
saying
in
the
charter
we're
putting
the
information
into
the
routing
system
gives
me
serious
heartburn.
Thank
you.
Thank.
AH
E
E
Yeah
yeah
I,
don't
we
we
have
discussed
this
issues
and
I
think
it
is
the
most
important
issues
from
the
previous
pop
that
can
don't
really
expect
the
whole
routine
system.
It
is
a
focus
on
the
edge,
agile
router
and
if
the
current
version
of
the
each
other
will
lead
the
misunderstood
of
this,
this
direction.
We
can
work
together
to
refine
it.
I
think
yeah.
B
AI
Colin
Jennings
with
the
IEP
and
John's
comments
are
mostly
for
you
I.
Think
really
is
you
know
when
I
read
these
Milestones?
This
is
very
much
about
a
solution.
We
can't
be
having
this
boss,
where
we
say
we're
not
discussing
a
solution
when
the
charter
is
very
much
about
a
solution.
Okay,
so
no
do
not
cut
me
off.
AI
Thank
you
very
much
and
the
the
the
issue
here
is
so
that's
part
of
the
issue
and
the
issue
is
lots
of
people
are
saying
we
should
do
this
application
layer,
and
this
is
regularly
done
at
the
application.
Major
sd-wan
Solutions
major
of
VoIP
Solutions,
fine
through
DNS
or
other
mechanisms,
a
variety
of
endpoints.
Then
they
contact
each
one
of
those
endpoints
major
latency,
Jitter
and
stuff
to
them
and
also,
at
the
same
time
get
information
back
about
their
compute
memory
availability.
They
have
of
the
resources
they're
important.
AI
So
this
is
a
real
problem,
no
question
about
that.
But
there's
there's
solutions
to
this.
That
may
work
very
well
that
are
done
very
much
at
the
applications.
I,
don't
think
these
problem
statements
have
been
taken
over
to
the
dispatch
or
the
application
or
art
area
and
discuss
there.
You'd
get
some
very
interesting
feedback
on
how
people
think
we
should
solve
them
now.
I'm,
not
I'm,
not
arguing.
A
Thank
you,
so
maybe
the
you
guys
can
think
about
it
and
address
that
okay
yeah.
So
next
one
is
Dean.
I
If
we
want
to
set
up
a
working
group
to
define
the
problem
very
happy
to
do
so
resource
information
sharing
across
you
know,
let's
call
them.
I.T
systems
is
something
that
would
be
very
nice
to
solve.
Is
it
the
alpha
or
the
L7
layer
where
to
solve?
It
is
the
combination
of
that
we
don't
know
there
isn't
a
creole,
clear
problem
statement
what
has
been
presented
here,
what
has
been
presented
in
the
APN
and
the
closest
person
to
come
to
the
problem,
State
and
definitions
for
Colin
and
Joel
and
I
agree
with
that.
I
A
Okay,
thank
you.
So
next
one
Paul.
V
Hello,
this
is
Paul
Jung
from
University,
so
I
have
a
three
comments.
The
first
one
is
can
both
can
consider.
Energy
consumption
looks
like
basically
another
consumption
is
very
important.
The
Viewpoint
of
operators
right
so
currently
the
camp
of
don't
looks
like
don't
consider
energy
consumption,
so
that
is
a
command.
Second,
one
is
as
another
use
cases
as
you
can
see.
Urban
Air
Mobility
like
the
prime
cars
right,
so
the
100
automobiles
they
are
trying
to
commercializing
by
2018..
V
That
is
very
important,
so
you
mentioned
the
autonomous
vehicle.
In
addition
to
that,
also
we
can
see
the
prime
autonomous
vehicle,
or
something
like
that.
That
is
the
second.
So
in
this
case
is
the
third
one
is
some
proactive
at
the
server
assignment,
considering
the
users
or
moving
the
the
trajectory
right,
because
you
can
see
especially
UE
or
the
car
is
moving
according
to
a
lot
of
ways
that
we
can
proactively
assign
list
sources
in
for
active
alert
than
the
active
way
right.
V
So
that
is
the
one
thing
also
users
can
usually
their
prefer
some
the
applications.
The
application
requires
some
bandages
And
Delay,
so
that
kind
of
pro
Power
information
can
be
taken
advantage
for
our
camper
I.
Think
the
camp
pop
is
very
important.
I
think.
Thank
you.
Thank
you.
AA
U
A
G
Okay,
so
we
had
closed
the
queue
earlier
I
see.
We
still
have
many
people
in
the
queue
so
I
think
at
this
time
we
can
let
Jiang
or
ad
decide
when
he
wants
to
catching.
We
can
continue
the
queue
until
when
Zhang
decides
that
it's
time
to
for
him
to
to
step
in.
C
Well,
it
I
don't
have
time
for
monologuing
give
me
you
know
five
minutes
at
the
end.
I
think
you
guys
had
prepared
some
questions.
You
wanted
to
ask
so
make
sure
you
leave
yourself
enough
time
for
those
but
yeah
just
save
me.
You
know
five
minutes
to
close
with
sure.
Okay.
G
Yeah,
okay,
so
we
are
going
to
to
the
poll
at
certain
time.
So
if
you
are
in
the
room,
they
have
not.
You
signed
the
on
the
outside
tour.
Yet
please
do
so
and
we
will.
We
will
do
the
poor
like
in
probably
five
minutes
or
something
like
that.
Then
right
now,
we'll
just
continue
the
queue
yeah.
J
C
C
O
AA
Carrier
who
is
also
has
the
public
cloud
service
and
they
have
the
latency
requirement
on
a
network
in
the
service
of
Barcelona
and
I
even
made
this
kind
of
case
in
the
public
Cloud
carriers
service
in
China.
The
Practical
solution
was
under
the
Computing
and
storage
separation
solution,
so
I
think
the
requirement
of
network
to
provide
service
according
to
the
Computing
resource
allocation
of
red
region
and
even
the
service
calculus
is
start
to
appear
gradually
in
the
recent
years.
So
maybe
it's
the
right
time
to
start
doing
something
doing
the
research
on
each
okay.
A
AJ
Thank
you
for
hosting
that
can
Bob
I
just
wanted
to
mention
that,
like
the
use
cases
are
valid,
it's
absolutely
makes
sense.
The
service
metadata
need
to
be
distributed
to
the
network
age,
especially
on
the
advantage
for
better
decisions,
making
and
kind
of
having
a
better
performance
throughout
the
network.
So
the
use
case
stands
points.
It's
completely
makes
sense
and
we
should
do
a
single
operator
initially
the
how
we
do
that.
AJ
The
solution,
wise,
that's
to
be
kind
of
need
to
be
looked
like
if
it
is
involves
the
routing
and
kind
of
afraid
that,
maybe
that
there
are,
if
it's
kind
of
a
existing
work
group,
we
could
look
into
the
some
existing
solutions
for
that
and
the
routing
group,
so
that's
anyway,
I
would
be
kind
of
a
happy
if,
for
the
use
cases,
if
anything
needed
my
health.
AJ
AJ
AJ
Is
my
affiliation
is
with
the
Microsoft.
Y
J
AK
AF
I
missed
this,
the
Milestone
from
my
point
of
view,
I
think.
If
this
we
have
the
consensus,
so
I
think
the
first
I
think
the
work
we
should
defend
some.
This
is
the
common
common
meetings
about
this
Computing
info
I'm,
not
a
I'm,
not
sure
if
this
included
included
in
the
framework
draft
or
not,
but
in
my
opinion
so
I
think
it's
better.
This
can
be
identified
independently.
I
think
this.
This
is
the
one
one
thing.
AF
The
second
thing
I
think
here:
we'd
rather
talk
about
the
control
plane
and
the
data
plane,
I
think
maybe
we
after
this
The
Matrix
can
be
defined.
We
also
connect,
maybe
three
years
ago
we
know
that
it
will
I
forget
I
had
to
RS
yeah.
We
can
do
some
of
the
discussion
about
the
protocol
choice.
So
that's
after
that
we
can
just
use
computers,
use
the
consensus.
So
then
we
can
go
on
to
discuss
about
this
control
point
of
data
planning
solutions
that
might
comments
about
some
suggestion
about
this
Milestone.
Okay,.
AL
Really
joding
Joel
didn't
give
me
a
slot,
so
I'm
not
going
to
make
a
comment
about
if
this
should
be
made
in
the
working
group
or
if
it's
a
good
or
bad
idea.
But
I
just
want
the
group
to
think
about
this
architecturally
from
a
full
stack
perspective,
and
let
me
set
up
a
scenario:
first
I'm,
an
application
developer,
I'm
going
to
write
a
bunch
of
code.
I
have
a
lot
of
things.
I
have
to
worry
about
how
I
use
my
local
CPU
memory
management
data
structures
number
of
threads.
AL
What
apis
I
use
I
have
a
lot
to
worry
about
I,
so
I
have
my
local
compute
resources.
I
have
to
worry
about,
and
normally
I
didn't
have
to
worry
about.
Network
resources
now
what's
happening.
Is
there
is
a
burden
on
me
to
just
tell
the
network
what
to
do?
Let's
rewind
20
years
back
the
toss
bits
in
the
IP
header?
Were
there
so
the
application
you
could
see
how
the
packet
should
be
handled
right?
AL
We
could
have
back
then
said
the
network
should
tell
the
applications
how
to
set
the
bits,
but
we
never
did
that.
Thank
God.
So
now
what
happens
is
if
the
applications
would
have
set
the
bits
like
EF
and
AF
used
quite
a
bit
today
for
audio
and
video,
then
the
packets
could
be
handled
specially
for
these
specialized
applications.
AL
Maybe
AR
VR
has
to
have
different
bit
settings
than
audio
and
video
the
same,
or
we
have
to
figure
it
out,
but
I
have
a
feeling
that
if
this
group
builds
something
for
applications,
the
applications
are
just
going
to
ignore
it
because
they
don't.
In
fact
you
may
give
me
all
this
information
and
I'm
saying
oh,
this
is
great
now.
What
does
it
all
mean
if
I
tweak
this
and
I
tweak
that
how's
the
network
going
to
behave?
AL
Will
the
network
actually
deliver
what
I
say
and
will
it
deliver
it
at
every
single
hop
along
the
way
to
the
destination?
So
I
really
think
that
this
is
going
to
be
something
that's
going
to
have
complexity,
and
it's
going
to
be
fruitless
and
I'm
really
worried
that
we're
not
going
to
make
do
the
best
thing.
AL
You
can't
make
the
speed
of
light
go
faster,
so
the
Network's
going
to
do
the
best
it
can
and
it's
going
to
have
to
make
decisions
quickly,
based
on
the
resources
and
the
size
of
the
cues
of
what's
going
on
at
this
instance.
So
we
have
to
be
really
careful
what
you
build
here,
because
it
may
just
never
get
used
because
of
its
complexity,
Okay.
So.
G
I'm
going
to
cut
the
mic
line
here
now
we
are
going
to
do
the
poll
on
the
screen.
You
can
see
three
questions
I'm
going
to
skip,
skip
the
third
one,
because
we
know
there
will
be
many
people
interested
in
working
on
this,
so
we're
going
to
only
focus
on
first
two
questions
in
in
fact.
The
second
question
we're
going
to
change
it
to
is
routing
the
right
group
to
attempt
to
solve
it.
Okay.
So
the
second
question
is
the
routing
the
right
group
to
attempt
to
solve
it.
G
A
C
G
All
right,
I'm
going
to
end
it
now
so
that
we
can
get
to
the
second
question
for
the
record
we
have
148
participants
enter
a
little
choice.
G
Okay,
so
we
are
going
to
end
this
session
now,
for
the
second
question
is:
is
the
routing
group
the
right
group
to
solve
this
problem
in
the
starter
session?
Now
again,
if
you
raise
your
hand,
you're
agreeing
that
we
should
solve
it
in
a
routing
group,
and
if
you
click
on
do
not
return,
you
mean
mean
that
you
do
not
or
think
routing
group.
Is
there
a
rest
group?
Okay,.
C
Two
things
one
is
I
have
Greg
muttering
in
my
ear
that
he
thought
the
first
question
was
only
does
the
community
think
the
problem
statement
is
clear
and
he
thought
that
it
wasn't
that
well,
scoped
wasn't
part
of
that
question.
I
I
had
understood
it.
This
is
the
always
the
problem
with
these
polls.
Okay,
they're
approximate
and
you
know,
they're
not
votes,
and
if
you
didn't
understand
it
the
way
the
chairs
met
it.
C
We
should
probably
should
just
move
on
the
and
I,
but
on
on
the
question
that
you're
asking
right
now,
I
want
to
emphasize
or
to
to
clarify
we're
asking.
Is
the
routing
area,
the
right
area
to
be
solving
it
in?
Thank.
C
G
All
right,
I
think
we're
going
to
end
the
session
now
out
of
544
people
we
have
90
greatest
hands
and
55
does
not
raise
the
hand
okay.
So
now
we
leave
the
rest
of
four
minutes
to
jump.
Our
ad.
C
Thanks
everybody
for
spending
the
last
two
hours
in
in
a
room.
We
didn't
actually
intentionally
get
a
room
that
was
supposed
to
make
you
feel
like
you
were
being
stuffed
into
a
can,
but
I
guess
that's
what
we
got.
So
thank
you
for
your
your
patience
with
that.
C
So
you
know,
as
usual,
I
need
to
go
away
and
think
about
what
I've
seen
and
heard,
but
it
I
can't
ignore
the
fact
that
the
biggest
reaction
that
I
heard
in
the
entire
two
hours
was
to
Joel's
little
mini
speech
and
in
and
I
think
that
I
heard
that
a
bunch
of
times
and
that
his
two
points
and
you
can
yell
at
me
if
I
summarize
you
wrong,
Joel,
because
I'm
not
not
referring
to
notes,
but
the
two
things
I
heard
from
that
were
one
yep
seems
like
there's
a
I
mean
and
that
this
is
what
we
heard
last
time
too
by
the
way
seems
like
there's
a
legitimate
problem
that
people
legitimately
want
to
solve
in
a
you
know,
standardized
way,
but
his
point
number
two
was
but
maybe
not
in
the
routing
area
and
that's
what
I
heard
from
Colin
also
and
that's
what
I
heard
from
Dan.
C
Also
and
that's
what
I
heard
from
you
know
some
of
the
feedback
that
was
coming
through
on
zulup
also
so
there's
clearly
diversity
of
opinion
in
the
room.
C
The
you
know,
the
the
instrument
that
we
just
used
to
sort
of
you
know
take
a
home.
You
know
demonstrates
that
there
is
that
diversity
of
opinion
there's
right
now
going
out
of
this
room.
I,
don't
think
that
I.
Could
you
know
in
good
conscience,
say
we
have
anything
like
consensus
to
work.
This
problem
in
the
routing
area,
I
think
we've
got
some
very
good
advice
from
Colin
that
there's
a
wealth
of
expertise
in
applications
in
real
time
that,
if
we
can,
you
know,
engage
the
the
right
people
there.
C
We
may
be
able
to
get
some
extremely
interesting
feedback.
I
think
that
we
should
probably
pursue
that
and
I
think
that
I
would
like
to
find
some
time
to
talk
with
the
proponents
at
some
time
later
this
week,
but
right
now,
I'm
not
hearing
consensus
to
create
a
working
group
in
the
routing
area
to
pursue
this
Charter
as
we've.
Currently,
you
know
got
it
on
the
board.
C
I
am
hearing
consensus
that,
like
I,
said,
there's
there's
work
to
be
done,
but
we
we
need
to
nail
down
where
and
how
and
and
what
oh,
by
the
way,
the
other
thing
one
other
thing
that
I
heard
is
that
there
continues
to
be.
You
know
a
minority
of
people
that
really
would
love
to
actually
do
something
more
radical,
which
is
to
you
know,
work
the
problem
in
the
underlay
and
that
we
we
should
you
know.
C
Yes,
we
should
have
the
the
you
know
the
routing
system
trying
to
solve
this
problem
and
I
the
one
thing
that
I
heard
consensus
for
more
strongly
than
maybe
anything
else
is
that
no
we
shouldn't
do
that.
C
Sometimes
consensus
is
Right,
sometimes
it's
wrong,
but
that's
what
I
think
the
consensus
is.
So
that's
what
I
heard
in
the
last
two
hours.
Thank
you
all
for
your
work
on
this.
Thank
you
all
for
your
time
and
I
think
we're
concluded.