►
From YouTube: IETF106-RAW-20191120-1000
Description
RAW meeting session at IETF106
2019/11/20 1000
https://datatracker.ietf.org/meeting/106/proceedings/
A
A
You
are
here
for
the
reliable
and
available
Wireless
boss
and
if
you
don't
know,
there's
things
in
between
a
boss
in
a
working
group.
Please
go
look
it
up,
but
we're
trying
to
figure
out
if
there's
some
interesting
new
work
for
the
ITF
to
do
here.
So
we're
not
gonna
talk
necessarily
about
exactly
how
problems
will
be
solved
but
go
check
on
that.
I'm,
Eric,
Nord
mark
and
I'll.
Let
Rick
introduce
himself
good.
B
Morning,
everyone
I'm
Rick,
Taylor
I-
will
apologize
in
advance.
I
have
a
cough,
so
if
there's
sort
of
consumption
style
coughing
coming
from
the
chairs,
please
ignore
just
to
follow
up
with
what
Eric
said.
This
is
a
working
group
forming
buff,
so
there
has
already
been
a
raw
buff
to
discuss
topics
around
this
area
and
it
was
successful
and
well
represented.
So
this
is
now
an
opportunity
for
us,
as
ietf
attendees,
to
decide
whether
we
wish
to
form
a
working
group
to
work
on
this,
but
we'll
get
into
that
in
more
detail.
B
A
A
We
will
have
some
remote
presentations
as
well,
using
meet
echo
and
I
think
we
have
people
that
are
listening
in
our
meet
echo
as
well.
So
a
number
of
people
there
already
some
first
of
all,
we
need
a
minute
taker.
Do
I
have
a
volunteer
like
Ethan.
Should
we
second
volunteer
as
well
to
keep
dark
just
in
case
the
discussion
gets
heated
I've
been
thinking.
This
will.
A
A
Somebody
volunteered
orally
to
be
jabbar
scribe.
Yes,
thank
you.
So
we
have
that
taken
care
of
if
yeah
there's
mailing
list,
the
materials
are
all
uploaded.
I've
done
some
updates
as
of
this
morning
so
and
we
have
a
couple
of
presentations
so
so
this
is
the
agenda
for
today.
One
touch
a
bit
about
use
cases,
there's
a
use
case
document
that
talks
about
lots
more,
but
we
want
to
at
least
go
in
and
show
there's
something.
That's
a
bit
different
than
people
might
have
thought
about
before
in
the
space
and
problems
take
statement.
A
The
purpose
here
is
to
get
people
to
understand
what
the
working
group,
a
potential
working
group,
is
proposing
to
do
and
the
scope
of
that
work.
So
any
sort
of
agenda
bashing
at
this
point
in
time-
so
am
I
sorry
hinted
at
the
purpose
of
the
bar-
is
to
understand
the
problem
statement,
as
well
as
the
scope
of
the
proposed
work
and
not
start
discussing
necessarily
the
solutions.
Often
it's
useful
to
for
the
the
people's
assembled
here
to
understand
that.
A
Oh,
it's
possible
to
solve
this,
we're
not
trying
to
solve
some
unsolvable
problem,
but
then
the
other
part
is
sort
of
gauging
the
interest
of
the
people
are
here
as
well
as
on
the
mailing
list
to
do
work
in
this
space.
The
there's
likely
to
be
some
things
that
tie
in
more
with
radio
and
channel
aspect
here.
So
if
there's
people
that
are
sort
of
experts
and
working
in
that
field,
this
might
be
interesting
in
terms
of
sort
of
characterizing
things,
and
this
actually
applies
even
in
fixed
wireless
deployment.
A
So
that's
one
of
the
things
that
are
interested
in
interesting
in
this
space,
so
there
might
be
a
sort
of
a
slightly
different
community
of
people
that
we
can
encourage
to
come
and
join
us
to
help
with
this
work.
That
might
not
traditionally
come
to
the
idea
and
intent.
Is
you
know
to
leverage
existing
work?
Of
course
I
mean
from
other
working
groups
as
well
as
other
stos.
So
that's
it
for
the
introduction,
so
Carlos.
C
Okay,
can
you
hear
me
so
I'm
carlos
benardos
I'm
from
Jersey,
3m
and
I'm
gonna
present
this
draft
cysteine
rough?
Ok,
ok,
ok!
So
I
think
it's
better
now
so
I'm
going
to
present
this
cysteine
document
on
use
cases
that
is
caught
for
on
the
on
the
guys,
as
you
see
on
the
slide,
so
oops
the
typical
problem
with
this,
you
have
anything
it's
like
scrolling.
More
than.
C
So
in
the
drug
there
are
six
use
cases
the
commander,
for
the
sake
of
time,
I'll
try
to
be
very
briefly
as
for
allowing
for
discussion
on
the
actual
both.
So
there
are
six
use
cases,
I'm
gonna
just
talk
about
one
of
them
the
waters
for
integer
application
in
a
draft
for
itself
the
use
cases.
Basically,
we
describe
the
use
case,
the
generic
scenario.
C
We
justify
a
bit
the
need
for
waters,
which
is
pretty
obvious
in
most
of
the
cases,
then
we're
going
a
bit
into
the
specifics
of
the
each
of
these
cases
and
then,
based
on
that,
we
identify
some
of
the
actual
requirements
for
reliable
and
available
Wireless
from
each
of
the
use
cases.
So
next
one
please
so
for
the
one
person
for
in
neutral
application,
a
scenario
this
is
a
picture
I
will
not
go
into
the
witness,
of
course
that
tries
to
exemplify
what
this
case
is
about.
C
So
we
need
to
operate
those
in
a
way
that
ensures
that
each
of
the
types
of
traffic
gets
the
rare
ability
and
the
actual
conditions
that
they
required
to
operate
properly
and
next,
okay,
now
to
works
networks.
Thank
you,
so
gonna
be
did
more
into
the
specifics
of
this
unique
scenario.
So
what
are
the
specifics?
We
need
to
support,
or
we
have
very
nice
technologies,
most
likely
Wireless.
It
may
be
that
in
some
of
the
segments
of
the
connectivity
we
have
also
wire,
but
we
have
mostly
wireless
technologies.
C
Those
links
may
may
suffer
from
variation
that
we
need
to
cope
with
and
as
I
mention
before,
we
may
have
different
types
of
traffic
at
the
same
time
so,
for
example,
the
control
loop
kind
of
traffic
that,
of
course,
reliability
is
key
in
this
kind
of
a
scenario
plus
the
no
so
demanded
in
terms
of
reliability,
monetary
and
the
as
massive
traffic.
Those
are
two
examples
that
we
may
have
in
this
scenario,
so
based
on
that.
What
are
the
specific
requirements
for
or
the
main
requirement
for
role
in
this
for
this
use
case?
C
Well,
there
are
particularly,
although
there
are
a
few
others
that
are
the
first.
The
solutions
needs
to
be
backward
compatible
in
the
sense
that
we
need
to
support
both
this
traditional,
less
demanding
traffic
like
monitoring
and
the
agnostic,
plus
the
new
traffic
or
the
traffic
that
really
demands
reliable
and
predictable.
Wireless
behavior,
like
a
control,
loop
traffic
that
I
mentioned
before.
We
also
need
to
support
at
the
radius
what
as
technology's
operating
at
the
same
time
as
I
mentioned
before.
C
So
those
are
the
kind
of
key
to
new
requirements
that
demand
solutions
in
the
in
the
royal
space
and
I,
don't
want
to
spend
much
more
time.
So,
just
as
a
quick
summary,
so
we
have
different
use
cases
that
require
wireless
connectivity.
What
is
connectivity
with
reliable
and
available
behavior?
We
have
identify
already
six
in
the
draft.
We
are
working
already
in
a
couple
more,
for
example,
the
digital
optical
communications
and
a
smart
grid,
and
in
terms
of
next
steps
in
case
this
working
group
is
formed.
C
We
would
like,
of
course,
to
continue
identifying
potential
use
cases
for
ro
and
also
to
continue
working
on
existing
ones
in
terms
of
characteristics,
characterizing,
the
actual
requirements
that
we
need
for
the
potential
ro
technologies
to
be
developed,
and
that's
basically,
all
in
terms
of
very
short,
summary
of
use
cases.
So
you
have
any
question
comment.
D
C
Well,
the
the
scope
is
to
actually
I
mean
that
will
be
a
bit
solution,
space
of
it,
but
the
scope
is
to
support
multiple
technologies
and
how
to
properly
configure
and
use
those
multiple
cameras
in
order
to
support
or
to
provide
reliability
and
many
diabetic,
a
statistics
for
the
the
whole
operation
of
the
of
the
traffic
I.
Don't
know
if
I
mean
it
that
answers
your
question.
I,
don't
know
exactly.
A
A
Yeah
but
it,
but
it's
the
boxes
that
interconnect
that
so
you
can
provide
a
service
across
this
whole
network
that
that
is
reliable
and
available
using
different
underlying
technologies.
So
that's
the
key
thing
it's
illustrating
right:
it
is
about
layer.
Three,
just
like
that.
Net
is
about
layer,
three
using
different
layer,
two
technologies
underneath
okay,.
E
So
clearly,
one
of
the
main
discussions
we
need
to
have
today
is
to
establish
a
scope
for
the
world
that
want
to
Intel
to
we
tend
to
do
and
basically
what
kind
of
problem
we
we
want
to
solve,
and
so
we
we
started
a
draft
and
we've
been
working
quite
a
lot
on
that
draft.
Since
the
last
ATF,
where
we
had
this
informal
meeting
and
basically
here's
the
content
of
the
draft
for
a
minute,
so
you
will
see
that
the
trust
is
not
only
about
the
exact
crime.
E
E
So
basically,
we
expect
that
the
centralized
routing,
mostly
in
bed,
net
from
which
we
are
in
ur
it
will
beyond
rotting
with
enough
reliability,
viability
for
wire
wild
applications
and
pretty
much
the
pre
of
functions
that
better,
not
defined
and
the
routing
will
will
be
programming
the
fording
once
and
for
all,
and
the
forwarding
will
be
a
simple
application
of
what
the
routing
has
decided
and
that's
probably
enough
for
wires.
But
for
wireless.
E
E
Okay,
so
the
first
thing
that
the
draft
does
is
terminology,
because
when
we
start
those
discussions,
people
wonder
why
are
we
talking
about
the
same
thing?
So
it's
very
important
to
define
the
terms
we'll
be
dealing
with
the
first
term,
reliability
and
actually
by
abilities
like
that
as
well.
Both
come
from
automation
world,
so
it
nothing
really
specific
to
to
ATF
or
Internet.
E
Do
the
terms
that
you
find
for
everything
automation,
the
first
term
critical
term
is
reliability,
is
how
well
a
certain
system
will
be
able
to
perform
the
expected
function
and
in
our
case
the
expected
function
can
be
to
deliver
the
packet.
If
you
talk
about
determinism,
then
it
will
be
to
deliver
the
packet
timely
and
how
reliable
the
the
network
is.
E
Is
how
well
the
network
can
deliver
those
packets
whose
packets,
in
due
time,
so
you
have
to
establish
metrics
to
measure
the
reliability
usual
metrics
in
the
industrial
world
is
the
mean
time
between
photo
mean
mean
time
between
errors,
and
you
will
find
us,
as
you
are
Conroe,
that
we
might
be
interested
in
other
metrics
and
another
critical
metrics
is
the
number
of
consecutive
faults.
If
you
have
a
control
loop,
which
is
one
of
the
usual
applications
of
automation
networks,
you
can
actually
totally
it's
it's
very
acceptable
to
lose
a
packet
packets
get
lost.
E
But
if
you
lose
four
packets
in
a
row,
the
system
will
go
into
emergency.
Stop
and
that's
one
thing
that
is
not
acceptable.
So
the
number
of
consecutive
errors
becomes
a
creaky
metric
for
your
reliability,
okay,
so
the
first
phase,
one
one
of
the
phase
and
one
of
the
interesting
things
in
this
document-
is
to
actually
define
reliability
and
with
it
define
the
metrics
by
which
we'll
be
able
to
assess
the
reliability
that
our
networks
will
have
to
ensure.
E
E
And
finally,
there
is
this
third
term
which
will
be
kind
of
critical
to
us.
We
kind
pal
I
think
and
still
be
a
challenge,
but
it's
it's
kind
of
the
wireless
version
of
what
we
know
it
from
that
net,
which
is
pre
off
so
pre
off
in
that
net
means
packet,
replication,
elimination,
ordering
function.
It
remains.
Okay,
I
want
an
available
path.
I
want
to
reach
our
ten
ten
nines,
so
Ethernet
will
give
me
five
nines
if
I
have
two
parallel
paths
with
Ethernet
and
they
are
completely
independent,
completely
non
congruent
and
blah
blah.
E
Then,
statistically,
if
I
have
five
nines
on
the
Left
five
nines
on
the
right,
I
get
ten,
nine
okay
and
that's
the
replication,
typically
at
the
ingress
and
the
elimination
at
the
egress,
so
that
you
get
two
single
packet
out
and
guess
what,
as
you
do
this,
you
might
actually
put
the
packets
out
of
order.
So
that's
why
they
are
this.
E
Ordering
function
typically
add
egress,
and
that
model
is
also
typically
quite
enough
for
wire,
meaning
that
usually
you'll
get
two
parallel
paths,
you'll,
be
monitoring
the
activity
and
ensuring
the
still
up,
and
that
will
give
you
a
viability
and
ten
nines.
So
the
routing
will
actually
be
on
two
T
paths
set
that
up
install
a
flow
over
that
and
you'll
be
all
set
for
deterministic
networking
for
wireless.
E
It's
going
to
be
a
little
bit
more
complex
because
the
medium
itself
is
not
deterministic,
so
you
cannot
even
climb
oh
I'm
building
a
deterministic
Network,
and
you
will
see
that
our
specs,
if
ever
you
find
the
term
deterministic,
it
couldn't
be
caught
and
code
right.
We
try
to
avoid
using
the
term
deterministic
and
that's
why
I
see,
even
though
we
inner
it
from
that
net,
we
are
not
claiming
we're
going
to
bring
you
a
deterministic
Network.
E
E
The
prime
is,
as
you
do
this,
the
path
of
a
collection
of
paths
that
the
routing
will
have
to
device
for
for,
for
the
wireless
will
probably
be
oversized
for
a
particular
moment
of
time,
meaning
that,
if
you
really
want
so
many
nines,
that
means
you
have
probably
like
three
four
five
different
paths
that
you
could
use
in
power,
some
of
them
single
hard,
some
of
them
multiple
hubs,
etc.
Some
of
them
are
possibly
different
technologies
if
you
go
and
use
everything
all
the
time.
E
So
if
you
really
flood
this
graph
that
the
routing
has
provided
to
you
all
the
time,
that
means
other
use
of
energy
and
values
of
constraint
spectrum
in
many
applications,
we
can
not
waste
those
resources
on
wireless
again.
It
would
not
be
a
prime.
You
just
push
the
packets
on
both
paths
and
your
whole
set,
but
on
wireless
energy
and
spectrum
are
critically
important
and
guess
what,
since
the
links
are
not
so
so
viable,
you
really
need
to
have
a
lot
of
redundant
diversity
as
you
build
those
paths.
E
So
so
now
you
see
the
tension
right.
On
the
one
hand,
the
routing
has
to
give
me
a
very
diverse
path
of
every
diverse
possibilities
to
get
from
A
to
B,
but
on
the
other
hand,
if
I
use
all
these
diversity,
all
the
time,
I'm
wasting
my
spectrum
and
my
energy,
and
so
this
this
thin
pipe
that
exists
between
the
PCE
and
the
fording,
which
was
just
simple
programming
for
the
wire,
now
has
to
become
a
lot
more
intelligence
so
as
to
wisely
use
the
resources
that
the
routing
has
told
you.
E
E
So
we
have
a,
for
instance,
a
small,
a
small
graph
which,
which
is
here,
and
it's
a
control
loop
between
maybe
sensor
here
and
a
programmable
logic
controller.
So
it's
one,
the
first
half
of
a
control
loop
between
a
sensor
and
the
programmable
logic
controller,
and
we
want
that
all
packets
from
the
sensors
reach
the
PLC
with
in
due
time,
and
you
never
lose
four
packets
in
a
row.
Something
like
that.
So
on
the
one
hand,
you
realize
that
the
view
that
the
PC
has
on
the
network
is
a
very
large
network.
E
Also,
it's
a
very
large
view
and
solving
the
problem
of
many
paths
like
this
one
which
may
use
the
same
constrained
resources
is
an
np-complete
problem,
so
solving
this
problem
at
the
scale
of
this
large
network
with
many
paths
like
this
is
a
huge
computation.
It
does
not
happen
quickly.
It
happens
actually
rarely
meaning
that
when
finally,
the
computer
decides
oh
here
is
a
harm.
Gonna
lay
out
all
those
paths.
You
expect
that
the
paths
that
have
been
built
are
usable
for
a
large
amount
of
time,
so
the
computation
is
slow.
E
They
tend
to
to
work
most
of
the
time
like
80%
of
an
hour,
but
at
a
very
particle
of
time.
There
might
be
some
metallic
object
passing
through
between
the
sender
and
the
receiver
and
guess
what
for
half
a
second,
the
link
will
be
down.
Can
you
go
back
to
the
routing
and
say:
hey
I
need
to
to
recompute
my
words.
You
can't
that's
not
the
same
time
scale!
E
Okay!
So
now
we
realized
that
we
have
a
time
scale
and
some
particular
metrics,
which
are
what
we
tell
this
PCE
and
on
which
the
PC
will
be
on
this
path
for
a
long
time
scale
for
a
long
time,
but
at
the
same
time,
those
wireless
hubs
we
go
available
and
variable
available,
flapping
quote-unquote
for
much
smaller
amounts
of
time,
and
we
need
a
much
faster
reaction
and
a
much
more
narrow
and
fine-grained
understanding
of
this
particular
small
network
within
the
network.
But
we
need
this
information
all
the
time.
E
All
the
time
and
the
type
of
metric
that
that
you
will
use
now
is
no
more
a
statistical,
80
percent
availability
over
hours,
but
more
like
a
is
it
does
it
work.
Just
now
so
more
like
on/off
boolean
or
something
like
what
is
the
current
speed
of
my
file,
because
you
know
many
of
those
files.
It's
true
for
that
11!
It's
true
for
15!
For
now
you
can.
E
The
the
work
that
the
PSE
will
have
to
do
will
have
to
account
for
the
the
Richer
capabilities
that
radios
have
in
particular
in
the
Pajero
functions
we
have
intro.
We
introduced
the
capability
to
do
retries
and
maybe
the
retries
fail,
and
if
they
fail
twice,
then
you
try
another
hub
a
different
next
stop.
E
So
that's
part
of
what
we
call
a
RQ
retries,
the
capability
at
layer
3
to
decide
a
I
can
try
twice
on
the
left,
but
then
I
might
I
may
go
right
or
I
can
trigger
our
application
because
didn't
work
where
on
the
left,
so
maybe
I
feel
better
if
I
try
both
next
ups
at
the
same
time-
and
you
can
see
on
this
table-
the
PSC
is
really
different
kind
of
beast.
The
PSC
does
not
do
right.
Computer.
E
E
The
operation
for
the
PSC
happens
within
this
small
network.
This
dismal
overlay
that
that
the
rotting
has
decided
should
be
used
for
this
particular
variable
variable
flow.
The
decision
of
how
it's
going
to
be
used
could
be
sorted.
The
ingress
could
flag
the
packet
and
say
Oh
on
the
first
step
is
going
to
be
replicated,
but
then
there
will
be
no
more
replication
or
maybe
that
will
be
cross
copy
between
one
path
and
I.
E
Let's
talk
about
different
timescales.
We
are
talking
about
different
metrics.
We
are
talking
about
different
pace
right,
so
that's
really
the
key
difference
between
this
PSC,
which
we
live
somewhere
on
the
southbound
interface
and
the
traditional
PC
programs.
My
forwarding
that
we
have
in
that
net.
So
is
it
even
ok?
So
the
other
question
is
ok.
You
know
your
wireless
harp
is
not
deterministic
and
you
even
achieve
what
you
claim
you're
going
to
achieve.
E
Well,
we
claim
and
we'll
see
more
when
we
discuss
the
technologies
that
there
are
more
and
more
well
as
technologies
that
provides
provide
more
and
more
grantees
and
the
way
those
grantees
are
improved
widely
is
through
the
use
of
10-digit,
multiplex
and
scheduling.
As
you
schedule
radios,
you
will
get
the
usual
benefits
that
you
expect
on
wires.
E
So
for
those
of
you
participated
GSN
of
our
debt
net,
you,
you
know,
you
realize
that
scheduling
traffic
in
particular
scheduling
all
the
critical
traffic
allows
us
to
put
a
lot
more
I
cross
traffic
on
a
given
link,
usually
used
to
put
like
10
15
percent
of
ours
cross
traffic
on
the
link.
After
that,
you
start
getting
statistical
effects.
If
you
schedule
on
your
traffic-
and
you
know
exactly
how
much
is
going
from
everywhere,
like
you
do
in
a
plane
or
you
ESN
Network
now
you
can
load
the
network
a
lot
more.
E
One
critical
benefit
that
we
have
in
TSN
from
scheduling
is
that
we
ensure
that
we
never
get
too
many
packets,
India
outgoing
queues,
because
pretty
much
all
the
laws
that
you
can
find
on.
An
Ethernet
network
is
collision
last
collision
laws
being
I
have
ten
packets
at
my
outgoing
queue,
but
there
are
five
place:
I
saw
have
to
discard
packets
because
there's
just
not
enough
room
in
my
choose
on
wireless.
E
Most
of
the
last
will
be
in
the
air,
so
we'll
see
that
scheduling
also
allows
us
to
reduce
the
collisions,
not
in
the
outgoing
queue
of
the
Ethernet,
but
truly
in
the
air.
And
finally,
the
key
advantage
that
TSN
derives
from
scheduling
is,
if
you
know
exactly
what
the
packet
is
within
each
period
of
time,
then
you
know
how
many
period
of
time
it
takes
from
the
back
for
the
packet
to
go
from
A
to
B,
so
at
the
end
of
the
day,
TSN
and
that
net
over
it
will
give
you
granted
latency.
E
So
those
out
the
usual
benefits
that
you
can
derive
from
from
scheduling
in
wireless,
we
arrive
even
more
benefits
and
that's
kind
of
cool.
The
first
thing
is,
if
you
know,
and
for
instance,
we
mentioned
6000.
If
you
know
when
you
transmit
something,
if
every
transmission
is
scheduled
now
you
can
live
without
interframe
space.
You
can
live
without
back
off,
which
is
really
empty
space
in
your
transmissions
and
that
that
kind
of
coalition
battery
operated,
because
you
don't
want
to
be
listening
for
the
void.
E
E
Now,
if
you're
used
again
to
the
way
TSN
operates,
you'll
find
that
the
way
we
provide
the
row
properties
on
wireless
are
different
from
from
what
you
you
do
on
wires.
First
thing
is:
like
I
said
on:
why
you're
the
only
place
I
can
have
a
collision
is
my
own
outgoing
cue,
so
the
core
of
what
GSN
will
be
doing.
E
There
is
first,
you
schedule
to
make
sure
you
don't
have
too
many
packets
and
then,
if
you
have
a
non-deterministic
packet
being
sent
on
this
transmit
cue,
but
you've
got
this
urgent,
the
terroristic
packet
that
you
want
to
send
in
the
middle,
then
you
can
snip
snap,
the
current
packet.
We
put
the
half
that
you
didn't
sign
on
the
side,
send
your
deterministic
packet
and
resume
the
transmission
on
the
lower
priority
packet,
and
that's
that's
easy
because
you
hone
the
whole
thing
you're
on
the
silicon.
E
That's
transmitting
and
the
medium
is
the
the
unidirectional
transmission
of
your
Ethernet
was
the
switch
in
the
air.
It's
different.
This
is
a
shared
medium.
There
can
be
tons
of
things
which
are
happening
on
that
spectrum,
and
the
only
way
to
kind
of
avoid
all
forms
of
collisions
is
again
diversity.
If
you,
if
you,
if
you're
facing
a
white
noise,
maybe
sending
a
spike
of
energy
on
a
very
narrow
bound
channel,
is
the
way
to
defeat
that
noise.
E
On
the
other
hand,
if,
if
you
get
a
very
small
spectrum
with
a
lot
of
energy
in
it,
then
using
the
spread
pack
from
technology
is
the
way
to
go
around
it,
and
now
you
realize
a
if
I
use
a
single
radio
technology
I
have
less
probability
of
success
and
infirm
capable
of
using
two
different
technologies
which
have
different
properties
on
the
way
they
use
the
spectrum.
So
again,
diversity,
diversity,
diversity,
diversity
is
the
key
to
give
us
reliable.
The
variable
transmission
and
pareo
is
actually
the
term.
E
We
coined
to
express
all
those
forms
of
diversity
that
we
only
use
want
to
use.
So
the
key
one
that
does
not
exist
in
that
net
is
a
RQ
or
hybrid
a
RQ.
So
that's
the
retry
I
have
to
use
the
rich
hybrid
means
you
can
do
some
forward
or
a
correction
with
it,
meaning
that,
for
instance,
if
I,
if
I,
have
like
two
three
paths
and
I
do
some
network
coding
with
my
packet
size,
an
ABC.
E
These
four
segments
I
can
send
a
fifth
segment,
which
is
the
exhort
the
first
four,
and
if
I
do
that,
and
one
of
the
four
is
lost,
then
I
can
use
the
fifth
one
to
recompose
at
the
end.
What
I?
What
I
lost
the
amount
of
network
coding
that
you
apply
could
depend
on
your
appreciation
of
the
reliability
of
the
network
at
this
particular
moment
of
time.
E
So
you
see,
there's
there's
a
lot
of
intelligence
that
could
take
place
in
order
to
ensure,
like
the
number
of
nines,
that
I
want,
or
the
number
of
packets
loss
in
a
row
that
I
want
to
maximize
to
ensure
that
at
this
precise
time,
and
that
could
be
dynamically
tuning,
the
capabilities
that
are
there.
So
we
are
not
installing
the
routing,
but
we
are
using
only
capabilities
that
are
there
very
wisely
to
optimize
the
balance
between,
in
the
one
hand,
energy
and
spectrum
and
the
other
hand,
reliability
and
availability.
E
So
we
could
have
looked
at
how
we
interact
with
different
working
row.
Clearly,
we
narrate
us
from
the
many
architecture
will
have
the
central
computation,
because
you
need
this
good
view
of
the
network
to
be
on
the
routing
and
all
this
is
kind
of
the
same
right.
Maybe
our
small
network
we
talked
about.
We
call
that
the
track.
Maybe
this
truck,
is
a
bit
more
complex
than
what
the
date-nut
control
usually
builds,
but
it's
it's
cisco,
that's
what
it
does.
So
we
completely
narrate
this
this
model.
E
We
live
in
this
small
world
that
goes
overgrown
now
between
the
writing
as
computed
this
thing,
and
there
is
a
decision
for
this
given
packet.
In
our
case,
we
don't
use
all
the
possibilities
that
the
routing
gives,
because
there
are
too
many
of
them
and
we
have
to
find
two
now
we
use
them,
and
for
that
we
need
to
monitor
the
small
network
all
the
time,
all
the
time.
It's
a
very
fine
grain.
So
it's
it's
a
direct
inheritance
from
that
net.
E
We
also
inherit
from
MANET
and
we'll
really
hope
that
some
some
people
might
from
money
will
come
here
now.
We
are
not
mobile,
I
mean
there
is
a
big
tension
between
mobility
and
determinism,
because
it's
very
hard
to
solve
the
problem
when
it's
fixed.
Now,
if
everything
is
mobile,
it
becomes
even
harder.
So
there
are
people
who
can
actually
think
that
not
only
time
but
also
space,
our
dimensions
to
the
PC,
and
that
can
be
done.
E
If
you
know
deterministically,
where
the
objects
will
be
like
this
satellite
networks,
then
you
can
start
building
the
turistic
path
even
over
this
many
satellites
in
the
air.
Now,
that's
probably
not
the
most
usual
case,
but
when
the
movements
are
deterministic
you
can
you
can
build
a
deterministic
Network
when
the
movement
is
not
deterministic.
It's
just
like.
When
the
flow
is
not
deterministic,
then
you
cannot
build
anything
deterministic
over
it.
So
because
of
this
tension
with
my
name,
we
can
reuse.
Some
work
in
particular.
E
Things
like
dilip
looks
fundamental
and
very
interesting
for
us,
but
we
have
to
think
the
leap
of
our
multiple
Hopf
not
develop
just
between
one
router
and
it's
wireless
interface.
So
we
have
to
be
able
to
collect
all
the
dilip
information
of
our
multiple
hops
and
feed
that
back
to
where
the
PSC
decision
is
being
made.
E
Sika
obvious.
We
will
need
data
model
and
we'll
need
to
extend
the
wire
data
models
to
express
to
spelling
of
functions
that
now
we're
capable
of
ipm.
Probably
there
will
be
a
mix
of
in
burn
a
lot
of
burned,
om
to
dynamically
measure,
this
very
small
network,
again
Phi
space,
but
very
small
network.
That's
why
it's
doable
so
we're
not
talking
about
entire
domain
pareo
function.
We
are
looking
at
enabling
a
few
hops
over
one
or
few
technologies
and
bring
the
reliable
and
available
characteristics
that
we
after
so
we
are
talking
about
small
networks.
Here.
E
So
what
we
need
to
do?
We
first
need
to
check,
obviously
that
we
have
use
cases
and
technologies
and,
like
I
said
there
is
a
discussion
and
technology
draft
coming
next,
we'll
see
that
more
and
more
radio
networks
are
coming
with
capabilities
to
schedule,
meaning
that
we
know
that
once
through
us
completed
all
the
work,
we
can
actually
capo
multiple
hubs
of
those
different
technologies
and
produce
the
ultimate
goal,
which
is
you
know
the
number
of
nines
that
you
want
with
the
last
that
you
want
within
a
bounded
latency.
E
E
How
you
make
a
distributed
decision
of
using
this
redundancy
that
the
network
has
given
us,
so
we
need
to
tag
the
packets
with
this
decision
or
if
it's
allotted
or
helpers,
to
make
this
decision,
if
it's
not
so
swatting
and
on
the
other
hand,
we'll
have
to
do
this
BFD
or
om
Ohio
a.m.
or
whatever
else
extension
to
get
a
picture
of
the
current
state
of
this
graph
at
this
very
moment.
E
So
these
are
the
two
things
that
we
really
to
provide:
get
the
visibility
on
this
very
small
network
at
a
five
space
and
then
provide
the
right
signaling
in
the
packets
to
use
that
what's
what's
available
wisely
and
if
I
have
time.
I
have
a
few
experiments
that
I
could
share.
I
have
eight
minutes,
so
let
me
just
show
you
a
few
graphs
I
hoped
I
would
be
able
to
because
I
find
them
kind
of
interesting.
E
So
so
here
is,
there
is
a
small
network
and
these
are
real
world
measurements
right
and
we
just
we
just
artificially
created
some
loss
because
it's
those
networks
are
actually
quite
reliable.
So
it's
it's!
We
have
to
for
some
euros
to
get
there,
but
so
you
have
another
path
in
red
and
a
row,
a
path
in
blue
and
the
two
pictures
on
the
top
show
you
basically
the
number
of
transmissions
that
it
took
for
a
given
packet
to
gave
to
get
from
a
to
H.
E
So
on
the
first
two
we
do
normal
irq
over
top
left
up
right
over
a
single
path.
Meanings
you
try
to
send.
If
it
doesn't
work,
you
resend,
it
doesn't
work
every
sound
so
on
the
vertical
here,
it's
one
time,
so
you
see
that
this
packet,
the
packets
down
these
lines,
they
made
the
hole
they
made
the
whole
path.
Without
a
retry,
the
packets
which
are
seen
here,
they
made
the
whole
path
with
one
retry
doesn't
say
which
one
should
you
just
count
the
hops
that
it
took
for
a
given
packet
to
get
there?
E
And
you
see
that
as
you
go
more
and
more
packets,
there
are
less
and
less
packets,
which
add
more
and
more
retries
the
find.
Since
you
get
two
packets
here,
I
don't
see
if
you
can
see
the
blue
dot.
This
packet
experienced
15
transmissions.
The
more
transmissions,
the
more
spectrum,
the
Mohan
Archie
blah
and
now
does
that
translate
into
benefits
if
we
actually
do
replication
at
the
ingress,
but
no
retry,
which
is
this
curve
here
in
this
curve
we
have
scheduled
every
transmission,
that's
why
we
have
an
integral
a
number
of
transmissions.
Always
the
same.
E
So
there
are
ten
transmissions
to
cover
all
this
thing
with
replication
and
elimination,
so
that
was
a
schedule
with
Austin
transmissions
and
obviously
all
the
packets
that
made
it
made.
It
was
exactly
ten
transmission.
That's
what
we
call
bounded
latency
through
scheduling.
That's
exactly
that
now,
the
packets,
which
are
a
kind
of
minus
1
of
the
packet
which
never
made
it.
We
also
condemn
all
right,
so
you
see
that
the
number
of
packets
never
made
it
on
this
network,
in
particular
this
time,
because
we
placed
an
object
in
the
middle.
E
Only
we
put
away
do
ten
I,
don't
remember
what
we
did,
but
we
just
prevented
the
transmission
for
a
small
time,
a
small
amount
of
time.
So
basically,
what
does
that?
First?
This
tells
us
that
if
you
just
have
one
half,
you
don't
have
a
viability
right
because
you
lost
so
many
packets.
At
this
time
the
network
was
not
available.
You
cannot
build
a
safety
network
with
that
serial
design.
E
That's
one
thing,
second
thing
that
down
here
so
second
thing
that
this
tells
us
is
compare
those
two
curves,
so
the
red
path,
the
blue
path
and
the
deterministic
path.
You
realize
that
most
of
the
time,
hot
potato
does
better
latency
than
determinism
and
that's
actually
a
very
classical
result.
If
you
look
at
t-t-tech
the
old
presentation
they
made
forever,
they
showed
you
this
this
bell
curve
which
never
really
ends
compared
to
the
spike
of
the
deterministic
traffic.
E
So
when
you
schedule
stuff
to
deterministically
deliver
at
a
certain
time,
what
is
not
deterministic
is
the
amount
of
packets
that
are
that
are
delivered,
but
you
can
literally
say
the
packets
will
be
delivered
there
or
not,
and
so
you
see
that
actually,
the
packets
have
a
higher
latency,
which
is
not
too
good.
You
also
see
that
in
terms
of
flow,
so
this
is
the
count
of
-1.
So
this
is
the
count
of
loss.
E
So
if
you
just
cared
about
reliability
most
of
the
time
you
realize
that
a
single
path
with
a
RQ
is
actually
more
more
a
bit
of
latency
most
of
the
time
and
is
more
reliable.
So
we
are
not
necessarily
talking
when
we
build
this
return
path.
We
are
not
necessarily
talking
about
improving
the
latency
or
improving
the
reliability.
We
are
talking
about
a
viability
only.
We
are
talking
about
safety
applications
and
the
most
safe.
E
You
want
to
be
the
more
this
path:
you're
gonna
get
and
the
more
energy
you're
gonna
waste,
because
you
waste
10
energy
units
per
packet.
Sure
you
just
use
4.
So
you
see
that
even
in
terms
of
energy,
the
replication
elimination
game
is
not
a
sufficient
for
most
packet
as
just
standing
about
single
path
and
retries.
E
What
you
really
care
when
you
be
on
this
is
safety
right,
and
you
realize
that
you're
wasting
energy
and
you're
wasting
time
so
that
three
minutes,
that's
that's
when
you
want
to
do
better
than
that,
and
now
you
see
the
raw
prompt,
the
raw
Prime
Minister.
How
can
I?
What
can
I
do
to
get
this
green
curve
to
the
left,
to
get
this
blue
curve
down
so
that
it
matches
pretty
much
something
here?
E
How
to
avoid
the
waste
of
energy
and
a
waste
of
manual
that
you
can
see
here
and
still
provide
the
viability
that
you
care
about
and
I
will
not
have
time
for
both
slides,
but
I
can
at
least
tell
you
what
we
can
do
for
energy
saving
just
an
example.
So
what
we
we
do
here
is
so
that
we
have
a
draught
that
beer
so
feel
free
to
to
visit
that
draft
and
basically
the
bit
the
beer
Trappe
size.
E
You
put
some
bits
in
the
packet
to
say
which
of
those
segments
you're
going
to
use
and
when
you
do
elimination
when
you
to
transmission,
you
use
reset
the
bit
on
the
on
the
hop
that
that
means
the
hope
that
you
use
its
beauty.
On
the
other
hand,
when
you
do
elimination,
you
end
the
bit
mass
that
you
got
if
you
think
about
it.
That
leads
to
the
fact
that,
when
the
packet
arrives
in
the
end,
if
a
bit
is
not
reset,
it
means
that
transmission
failed.
E
So
now
you
can
observe
if
you
set
all
the
bits
at
the
ingress
tarika.
Theoretically,
the
packet
should
go
through
all
the
segments
and
all
the
bits
should
be
reset
if
the
packet
arrives
with
all
be
set
at
the
ingress.
If
the
packet
arrives
at
the
egress
with
some
bit
not
set
that's
a
map
of
what
didn't
work.
So
what
you
do
is
you
send
some
packets
with
the
full
bit
map
and
that's
the
packet
that
you
see
here.
All
those
packets
took
16
17
I
go
straight
from
here,
less
than
12
13.
E
Oh,
you
need
1100
units
of
energy
and
also
units
of
time
because
they
want
through
all
this
path,
but
you
see
that
most
packet
did
not
because
once
we've
done
that
we
have
a
map
that
we
can
use.
So
we
just
set
the
bit
of
the
segment's
that
we
really
want
to
use.
One
does
allows
there
is
a
signal
back
to
so
saying
a
in
order
to
avoid
follows
in
a
row.
E
Please
set
all
the
bits
against
which
increases
a
lot
of
reliability
of
your
packet,
and
by
doing
this
feedback
we
can
see
that
most
of
the
packets
take
a
minimum
number
of
hops,
sometimes
your
hands.
Here
we
went
around
a
certain
prime
by
rerouting
and
then
back
to
the
to
the
least
energy
path.
I
have
one
more
minute,
so
I
can
just
show
you
the
result
so
in
in
dark
green.
That's
when
you
do
for
replication,
you
realize
that
you
have
a
better
reliability.
E
You
get
more
packets
through,
but
the
cost
of
energies
is
twice
but
because
of
the
feedback
loop.
We
can
avoid
the
losses
in
a
row
because
you
see
there
is
never
for
us
in
a
row.
So
you
see
the
game
by.
This
is
just
one
method
of
tuning
the
use
of
the
network,
so
we
consume
less
energy
and
still
we
we
stay
within
the
the
bonded
latency,
because
we
are
still
little
mystically
scheduled
and
number
of
loss
in
the
ROM.
B
G
A
B
H
H
H
H
H
H
How
go
to
the
next
slide?
Please,
okay,
so
this
slide
shows
a
feature
of
a
two
2.11
called
a
two
2.11
s.
It's
a
mesh
network
that
actually
allows
multiple
a
to
2.11
nodes
to
be
connected
to
each
other
and
establish
pads,
sometimes
in
redundant
paths
to
enable
packets
originating
from
1-800
to
11
node
to
reach
the
destined
node
by
traversing
across
multiple
802
11
nodes.
H
H
H
Some
of
them
are
already
part
of
the
standard,
so
we
can
leverage
them
and
some
of
them
need
to
be
worked
on
in
future.
A
to
211
projects,
some
probably
in
coordination
with
the
IDF
raw
to
ensure
that
the
requirements
that
raw
comes
up
with
are
supplanted
with
features
in
802
dot.
11,
the
best
before
the
the
challenge
with
wireless
is
the
the
link
capacity
is
not
constant
and
changes
quite
frequently
because
of
interference
and
noise,
and
so
the
adjustments
for
reservations
made
end-to-end
to
cope
with
these
with
a
changing
capacity
of
the
link.
H
So
this
is
a
map
of
different
a
22.1
specifications
that
are
already
out
there
and
it
tries
to
show
you
where
a
to
2.11
needs
to
do
more
work.
For
example,
time
synchronization
a
to
2.11
already
has
a
couple
of
protocols
that
can
work
with
e
2.1
s
and
provide
time
synchronization
in
the
wireless
network,
and
the
same
is
true
of
interoperating,
with
8
or
2.1.
H
We
have
a
two
2.11
AAA
that
provides
the
stream
reservation
protocol
that
ensures
that
you
can
make
a
reservation
from
end
to
end
and
also
equal
to
11
a
k'
that
provides
bridging
feature
for
802
11
networks
for
bounded
low
latency.
We
have
some
features
in
lemon
ax,
but
that
they
may
not
satisfy
all
requirements,
and
some
extends
are
being
worked
on
in
11
b
e
that
provides
bounded
low
latency,
and
the
same
is
true
for
ultra
our
reliability.
H
H
We
have
this
notion
of
a
scheduled
operation
where
the
AP
has
total
control
over
the
medium
for
all
the
stations
that
are
associated
with
the
EP
and
if
the
stations
have
already
indicated
the
type
of
traffic
that
they
anticipate,
the
AP
can
schedule
transmission
times
for
each
of
the
stations,
thereby
reducing
or
eliminating
any
need
for
back
off,
and
if
you
can
eliminate
back
off,
you'll
eliminate
all
a
dead
that
time
in
the
air
and
maximize
the
use
of
the
spectrum
to
send
traffic.
That
is
critical.
In
addition
to
that,
we
also
have
other
mechanisms.
H
Qos
mechanisms
built
three
802
11,
where
we
can
prioritize
time-sensitive
traffic
over
non
time-sensitive
traffic
and
make
sure
that
any
queue
time
sense
in
traffic
gets
the
first
chance
to
get
out
of
the
device.
So
we
did
some
simulation
based
on
a
2.11
mechanisms
and
the
chart
on
the
right
hand.
Side
basically
shows
you
that
you
lower
the
latency
that
you
want.
The
less
number
of
stations
is
what
you
can
serve,
given
a
specific
given
a
constant
bandwidth.
H
H
But
this
is
definitely
achievable
with
20
megahertz
bandwidth
in
811,
you
can
serve
up
to
45
stations
if
you
use
the
811
X
protocol,
if
you're
bounded
latency
requirement
is
3
3,
second
3,
milliseconds
or
less.
If
you
want
title
latency,
then
you'll
have
to
sacrifice
the
number
of
stations
that
you
can
so
next
slide
please.
H
So
this
is
the
a2
2.11
Vee
is
a
new
project
that
we
are
currently
working
on
in
a
two
2.11
and
it's
building
on
a
two
2.11
ax
and
trying
to
bring
in
other
techniques
to
make
the
media
more
available
and
be
more
sensitive
to
traffic.
That's
time
critical
supposed
to
pence,
which
is
not
time
critical,
so
first
off
as
we
do
in
most
of
the
8
or
11
projects,
there
are
enhancements
in
better
coding
techniques,
using
most
patients
frames
using
wider
channels
and
thereby
achieving
wide
ab
a
higher
throughput.
H
So
this
is
just
sheer
throughput
that
again
we
can
go
up
to
gigabit
per
second
with
these
enhancements.
In
addition
to
that,
we
are
looking
at
techniques
where
the
same
station
can
operate
in
2.45
and
potentially
six
or
seven
gigahertz
spectrum,
so
that
you
have
multiple
links
between
two
devices
and
you
can
liberate
the
fact
that
if
one
link
becomes
noisy,
then
you
can
without
the
traffic
through
another
link.
They
are
all
able
to
got
11
links.
H
Some
mechanisms
for
using
hybrid,
you
are
being
talked
about
in
lemon,
be
as
well,
and
the
last
one
is
to
provide
a
new
mechanism
where
channel
access
to
time
I
where
traffic
is
prioritized
over
non
time
away,
non
non
time-sensitive
traffic,
thereby
ensuring
that
egress
of
time-sensitive
traffic
happens.
First,
if
it
happens
to
be
queued
at
the
same
time,
in
non
time-sensitive.
Perfect
is
also
cute
in
the
same
device.
So
all
these
mechanisms
are
being
debated
right
now
they
are
not
completed.
H
There
are
various
phases
of
being
specified,
but
our
focus
by
the
time
we
make
progress
in
draw.
We
can
bring
in
more
requirements
from
from
Raw
into
this
project
and
big
make
sure
that
a
2.11
satisfies
all
the
requirements
to
provide
a
network
wireless
network.
Then
it's
reliable
and
available
next
slide,
please,
okay.
H
So
this
is
a
summary
of
all
the
all
the
things
that
we
talked
about
already
mentioned,
that
811
X
provides
a
new
mechanism
where
latency
and
reliability
can
be
controlled
and
achieved
compared
to
what
we
used
to
do
in
leaven
in
the
features
that
were
before
11a
x11,
AC,
11,
n
and
lawyer,
1182,
1180,
X
and
11
B
will
also
operate
in
6
to
7
gigahertz
band.
Wherever
it's
available,
it's
not
you
know
still
available
yet
Wow
with
that.
H
H
Smaller
packets,
in
a
shorter
time
and
thereby
achieve
lower
latency
or
higher
reliability
and
a
to
211
B,
is
actually
a
natural
progression
from
where
we
are
where
we
complete
when
we
finished
in
Lebanon
X
and
where
we
want
to
go
beyond
that.
And
this
is
a
good
opportunity
for
us
to
inject
features
from
our
requirements
from
raw
into
a
total.
J
I
F
I
would
just
like
to
highlight
one
point:
every
iterate,
what
Ganesh
mentioned
already
that
there
is
a
work
ongoing
in
802
for
the
integration
of
TSN
and
EndNote
11,
and
if
we
do
a
good
job,
one
way
to
look
at
it
that
we
get
a
NATO
to
learn,
including
wireless
LAN
and
viola
line,
components
that
is
the
subnet
technology
or
can
be
used
as
a
subnet
for
that
net.
Like
a
seamless
fashion.
E
E
So
actually
we
we
have
at
least
four
technologies
which
already
discussed
in
the
draft.
Well,
actually
five
g's
a
bit
late,
but
we
so
we
have
text
from
Dave
and
and
Ganesh
about
the
extensions
of
Wi-Fi.
From
my
face,
six
beyond
we've
got
1540s
eh,
which
was
used
the
ATF
for
six,
and
we
are
also
taxed
on
L
Dax,
which
is
an
avionics
technology
for
air-to-air
and
air-to-ground
deterministic
transmissions.
E
J
Hello,
everyone-
this
is
a
new
tomorrow
from
the
German
Aerospace
Center.
So
let
me
briefly
talk
a
little
bit
about
what's
going
on
in
the
communication
section
for
aeronautical
communication,
so
everyone
using
a
plane
today
mainly
relies
on
analog
voice
communication
for
safety,
relevant
communication.
This
is
currently
changing.
So
there's
a
future
communication
infrastructure
coming
up
for
there
are
several
technologies
discussed,
such
as
air
omics
for
airport
communications,
LDX
for
a
terrestrial
air
ground
communication,
also
air-to-air
communication
and
satellite
communication
for
sonic
Poland
remote
regions.
So
we
have
a
multi-link
system
via
multiple
hops.
J
Data
needs
to
travel
finally
to
the
aircraft,
and
so
we
from
the
German
Aerospace
Center
husband,
have
been
working
for
15
years
now,
an
LD
X.
So
the
most
important
part
to
understand
here
is
data,
travels
via
different
technologies
via
different
hops
and
needs
to
reach
the
end
user.
So
the
aircraft
in
a
certain
amount
of
time
reliably-
and
this
is
why
we
think
role
is
very
important
here
for
our
work
and
we
can
contribute
with
that
technology.
J
E
Many
thanks
Niels,
so
we
have
actually
uploaded
a
set
of
slides
which
won't
be
presented
today.
That
came
from
nails,
and
so,
if
you
go
through
to
the
agenda
and
the
list
of
documents
now
available
with
the
agenda,
you
will
find
actually
some
slides
and
index.
There
is
also
a
separate
document
which
is
published
to
the
row
list
of
documents,
and
that
is
pacific-12
duck.
So
you
have
both
the
section.
Arnold
acts
in
the
technology
document
plus
a
separate
document
in
the
attacks
menu
thanks.
D
A
D
E
Okay,
so
just
to
destroy,
there's
quite
a
bit
of
work,
which
has
already
started
through
the
row
mailing
list,
and
so
we've
got
all
those
documents
which
are
already
available.
We've
got
so
carol
has
presented
the
use
cases.
There
is
a
full
draft
with
many
use
cases
in
there.
We've
got
the
technology,
trust
I,
just
discussed
about
and
we'll
publish
more
on
5g
very
soon,
I
don't
know.
Maybe
we'll
also
include
Bluetooth,
because
there
are
interesting
work.
E
E
No,
sir,
not
surprisingly,
the
work
at
row
is
expected
to
span
to
trigger
some
work.
Another
working
group
well
the
specialist
to
solve
a
particle
prime.
Oh,
we
already
have
a
draft
which
is
actually
in
the
solution
space
for
row,
which
is
being
pushed
at
row
for
enabling
a
complex
truck
using
ripple,
and
we
also
have
a
beer
Qi
draft
at
beer.
So
the
goal
is
yes:
we
inherit
from
man
a
winner
it
from
that
net.
E
We
understand
our
prime
because
we
have
allies
people,
but
decisions
are
not
necessarily
to
be
built
within
Roe.
Once
we
can
express
the
prime,
then
we
can
go
to
a
particular
working
group
and
work
with
the
experts
of
that
working
group
to
actually
solve
that
component.
That
Road
needs.
That's
how
we
see
the
work
and
it
has
already
started.
L
A
Of
us,
as
chairs,
haven't
been
involved
in
this
work,
so
we
sort
of
came
in
new
to
this
well
starting
about
a
month
ago,
and
we
said
okay,
what
this
is,
can
you
explain
it
to
us
and,
as
we
were
discussing,
we
sort
of
came
up
with
well?
This
was
the
most
complex
simplest
picture
that
we
could
come
up
with
saying.
This
is
how
actually
is
how
what
Rob
brings
to
the
table?
A
In
the
past,
we
assumed
that
you
can
compute
these
paths
and
they
say
stable
for
the
time
scale
of
routing.
So
this
was
what
Pascal
was
talking
about
earlier,
so
that
was
sort
of
us
as
chairs
trying
to
understand
you
know,
what's
the
big
picture
here,
of
how
this
stuff
fits
in
and
next
you
know
before
we
get
into
the
discussion.
I
will
pull
up
parts
of
the
proposed
charter,
so
you
can
go
read
this
stuff
yourself.
A
There's
three
pages
here
of
background
texts,
giving
the
context,
which
is
what
we
talked
about
during
the
ball
today
and
then.
The
key
thing
that
I
would
like
to
leave
up
here
is
the
set
of
proposed
charter
work
items.
So
now
the
floor
is
open
for
discussion,
whether
it's
about
the
work
items,
whether
it's
about
understanding
the
problem,
statement
and
scope
of
the
stuff
so
force
yours,
hi,.
M
Lou
Berger,
can
you
go
back?
One
slide
that
one
thanks
I
guess
you
skipped
over
a
bunch
I
jumped
up
when
I
heard
you
say:
we've
never
done
this
path,
selection
thing!
So,
as
I
understand
it,
there's
a
couple
of
different
aspects
of
the
problem.
One
of
the
things
you're
trying
to
do
is
to
do
a
fast
forwarding
decision
right.
Well,
one
problem
is:
is
a
better
informed
path,
computation
based
on
what's
happening
at
the
RF
layer?
That's
not
this
one!
M
A
M
Sir,
in
that
net
we
do
pre
off,
which
is
we
do
an
ale
DUP
version
in
the
debt
net.
Definition.
There's
allowance
for
other
forms
like
exactly
this
one,
whether
it
be
Network
coding
or
as
Pascal,
said,
or
something
that
we
previously
haven't,
talked
about,
which
is
maybe
doing
a
fording
decision
based
on
the
RF
layer
telling
you
which
of
your
available
links,
are
have
low,
have
higher
probability
of
delivery,
that
that
is
something
that
that's
new.
But
this
type
of
thing.
M
M
It's
it's
a
reasonable
question
in
terms
of
the
point
about
that
net
being
overloaded,
we're
wrapping
up
the
core
forwarding
plane
documents
by
the
way,
which
includes
the
protection
mechanism,
a
pre.
Often
you
know.
Maybe
we
have
a
new
protection
mechanism
that
comes
in
that
could
be
applicable
to
multiple
technologies,
because
there's
no
reason
why
you
have
to
limit
the
I'm.
Sorry,
harro
I,
don't
remember
the
new
acronym
you
don't
necessarily
have
to
limit
that
just
to
wireless
networks
that
can
see
applicability
elsewhere.
M
You
know
so
do
we
have
time
well
we're
wrapping
up
and
probably
gonna
do
reach
are
during
it
the
next
IETF.
So
maybe
there's
time
there
is
it
the
right
place.
That's
a
different
question
and
C
camp
has
historically
done
where
we
have
the
layered
network
model
and
doing
traffic
engineering
over
different
layers.
That's
C
campus,
nothing
in
the
home!
For
that
you
do
end
up
in
in
those
groups
having
a
set
of
people
that
sit
around
board.
M
B
E
Okay,
so
now
two
aspects
to
this.
The
first
aspect
is
we
care
to
receive
in
this
group
if
we
format
a
crowd
which
is
widely
interested
in
specialized
in
Wireless
proms,
to
make
sure
that
those
Prime's
are
already
considered
and
the
type
of
discussion
that
have
happened
so
far,
that
net
would
not
attract
that
kind
of
crowd.
E
So
if
we,
if
we
kind
of
plan
to
do
it
within
bed
net
and
would
want
to
avoid
boring
people
in
the
room
and
for
the
population,
the
wireless
all
the
way
of
health,
then
we
would
end
up
in
doing
like
two
sessions.
You
know
without
the
wireless
session
in
the
world
session,
and
so
in
the
end
I
mean
we.
We
could
also
do
a
single
working
room
for
the
Hawaii
effort.
E
I
think
that
the
idea
of
splitting
things
into
domains
and
areas
and
working
groups
are
supposed
to
having
one
big
workgroup
meeting
the
whole
week.
Although
it
is
due
all
right,
I
mean
I'm,
not
saying
it's
wrong
is
that
we
can
actually
focus
on
one
particular
thing
and,
and
then
what's
great
is
we
can
actually
interact
with
the
other
working
groups
and
andro
will
certainly
interact
a
lot
with
that
net
and
could
even
be
pushing
some.
E
You
know
extensions
to
some
models
that
did
met
or
see
Campa,
actually
defining,
so
that
we
are
sure
that
you
know
the
raw
requirements
are
being
served,
doesn't
mean
that
everything
has
to
be
handled
as
ro,
but
it
means
that
the
congregation
of
people
that
would
be
meeting
would
would
actually
have
its
own
space
to
discuss
and
and
within
a
group
of
people
who
understand
each
other
uses
some
term
and
talk
about
the
same
thing.
So
that's
just
to
make
the
meetings
more
useful.
E
The
other
thing
is
I
was
there
from
inception
in
that
net
and
we
helped
create
it.
I
hope-
and
it
was
my
goal
to
three
years
ago,
to
get
this
sort
of
discussion
within
that
net
and
I
pushed
and
pushed
and
been
there
and
tried
and
the
bottom
line
you
have
seen
the
world
wireless
come
in,
go
away,
come
in,
go
away
just
like
well
as
link
does,
because
that
was
never
really.
The
interest
of
the
people
in
the
room.
E
I
was
pretty
much
anyone
interested
I
mean
exaggerating,
but
never
managed
to
get
focus
or
attention,
and
wireless
and
I.
You
can
understand
why
right
now
very
interesting
problems
to
be
solved
in
that
not
in
many
areas
and
and
you
see
what
the
work
is
at
this
moment.
That
gives
you
an
idea
of
what
the
focus
and
what
the
people
are
who
the
people
are
in.
It
is
from
that's
perfectly
good.
M
B
M
From
a
an
interest
standpoint
in
debt
net,
we've
had
people
that
have
been
interested
in
in
wireless
from
the
start
and
you
example.
There
was
a
aircraft
folks
on
there's,
it's
been
mentioned
many
times
and
it's
always
been.
Let's
finish
the
core
technology
first
and
then
start
working
on
the
next
one.
So
the
core
technology
were
pretty
much
done.
We're
just
finished
last
call
the
first
layer
of
technology
adaptation,
so
mapping
debt
net
to
TSN.
That's
coming
along,
really
nicely.
M
F
Understood
so
first,
thank
you
for
the
problem
statement.
Draft
I
think
that
was
a
step
forward
compared
to
previous
ITF.
It
is
still
not
clear
to
me,
although
so
I'm,
trying
to
lay
out
train
of
thoughts
for
that.
This
comes
from
from
that
net
in
the
sense
of
what
is
the
ultimate
goal:
to
provide
deterministic
networking
in
in
combined
networks
of
wireless
and
wireline
deployments.
I
fully
agree
that
we
should
do
this
and
I
have
the
feeling
that
there
are
gaps
that
we
have
to
fill
in.
F
So
I
fully
agree
with
the
ultimate
goal
that
we
should
figure
out:
what
to
do.
What
are
the
gaps
and
fill
the
gaps
so
I,
my
understanding,
that's
the
ultimate
goal.
What
we
are
discussing
thinking
it
further
based
on
the
layouts
of
the
problem
statement.
It
goes
into
the
well
that
distinguishes
two
sub
layer
service
and
forwarding
sub
layer,
and
this
is
a
particular
lower
layer
in
the
forwarding
sub
layer.
In
my
understanding,
so
the
gap
analysis
could
be
as
we
started
in
in
the
core
technology
data
pane.
F
What
is
missing
in
the
data
for
running
server
here?
What
is
really
needed
to
be
done
in
support
of
various
wireless
technologies
and
with
that
top-down
approach
that
could
be
done
in
that
net
in
the
next
phase,
so
Wireless
has
been
in
scope
from
start,
it's
been
in
the
use
cases
and
we
want
to
solve
it.
Reading
the
problem
statement
draft
it's
it
to
me.
It
was
kind
of
an
extensions
to
60s
really
so
it
was
not
a
generic.
F
Let's
cover
many
wireless
technologies.
That
is
the
track.
This
directed
acyclic
graph
and
so
on
so
I.
My
understanding,
it
was
extension
of
60s
did
not
take
into
account
license
back
to
metal
which
has
better
scheduling
and
so
on
different
different
different
perspective
and
from
the
bottom
up.
Yes,
sure
something
needs
to
be
done
for
60s,
but
maybe
it's
different
for
the
different
different
technologies.
Furthermore,
the
kind
of
api's.
E
F
Read
as
necessary
from
the
problem
statement
are
not
really
supported
from
the
wireless
technologies.
This
is
where
D
lab
can
help,
and
this
was.
It
was
good
to
see
that
that
that
picture
to
to
expose
necessary
data,
but
my
understanding,
neither
dot,
eleven
or
or
far
G
provides
the
api's
that
immediate.
What
for
the
layout
of
the
max
of
selection
in
indeed
before
running
sub
layer,
I,
don't
see,
for
example,
how
from
layer
three
we
could
influence
the
hark
of
five
G.
F
So
this
is
a
gray
area
and
from
what
I
also
add,
this
is
the
last
sentence
and
I
finish
what
I
also
mentioned
before
from
802
LAN
technologies.
It
is
happening.
The
integration
is
happening,
so
we
will
have
some
kind
of
subnet
and
802
subnet
for
when
it
comes
to
5g
for
a
GTS,
an
integration
is
happening,
so
we
can.
F
If
somebody
wants
to
use
our
jazz
radio,
it
would
be
possible
to
thinking
of
the.net
architecture,
replace
the
TSN
subnet
with
the
heterogeneous
wireless
and
wireline
weed,
weed,
weed
weed
with
two
aspects,
and
you
had
in
our
Center.
Only
the
nice
picture
of
your
whiteboarding,
which
you
you
sent
to
the
list
nice
picture,
you
took,
you
drew
something
on
a
on
a
wire
board
and
sent
a
picture
which
actually
plays
through
as
a
some
math
technology.
B
B
B
N
But
it
could
be
that
the
other
half
Michael
here
Richardson
it
could
be
that
the
other
people
aren't
going
to
debt
net,
because
debt
Nets
not
working
on
the
on
the
topic
exactly
right
so,
but
but
that
that's
chicken
and
egg
problem
I
think
you
have
there.
I
mean
I,
think
that
I'm
not
going
to
detonate
cuz
it's
conflicted,
but.
A
I
mean
I
think
that
there's
two
things
right,
one
is
is,
you
know,
do
people
find?
Is
it
clear
what
what
bra
is
proposing
to
do
here?
In
terms
of
you
know
the
problem
statement,
the
proposed
work
items,
etcetera
right,
then
this
is
support
thing
about
and
do
people
want
to
work
on
that
etc,
which
is
questions
we
will
ask
shortly.
As
you
know,
we
have
discussed
more
about
this
stuff,
but
and
then
there's
a
separate
question
about
okay:
where
do
we
place
those
work
items
right?
A
Is
it
a
separate
working
group
where
we
have
a
separate
community
of
people,
potentially
new
people
coming
in
with
more
radio
expertise
or
or
does
it
get
folded
into
some
existing
working
group
and
I?
Think
it's
as
Lou
said,
I
think
it's
up
to
the
ADEs
to
decide
that
stuff,
but
I
the
key
thing
from
sort
of
understanding
what's
being
proposed
is,
is
you
know
the
clarity
about
this
stuff
right?
Do
people
understand
it?
Is
it
well
scoped
enough,
so
anyhow
continue
the
discussion.
Please.
O
It's
to
card
I've
been
doing
internet
protocol
based
airborne
networking
for
about
25
years
and
I
have
noticed
that
we
as
a
community
have
left
the
field
of
airborne
networking
to
the
builders
of
radios
who
know
at
best
how
to
build
data
links,
not
networks.
There
have
been
a
couple
of
notable
exceptions
where
the
IETF
has
indeed
contributed
to
airborne
networking
what
they
have
been
buying
large
notable
exceptions.
Airborne
networking
faces
unique
challenges
and
I
realize
this
is
not
all
about
airborne
networking,
but
the
LDX
example
was
was
listed.
P
I
named
John
from
Holly
I
think
good
work,
I
think
in
general.
I
think
this
is
interesting,
especially
on
the
problem
space,
but
I
still
would
like
to
see
architecture
to
highlight
how
this
work
actually
can
interact
with
as
a
component
like
that
net
and
things
more
detailed,
because
I
can
say
that
the
purpose
is
to
propose
the
based
on
charter
to
propose
some
l3
abstraction
to
enables
you
know
some.
P
You
know
to
work
and
how
to
you
know
how
to
do
such
kind
of
interaction
and
benefits
the
r3,
end-to-end
service
guarantee
or
something
pata
I
saw
some
text
by
a
but
I
haven't
never
seen,
architecture
which
I
think
that
maybe
make
it
more
clear
and
without
such
kind
of
detailed
architecture.
It's
really
difficult
for
me
to
you,
know,
agree
on
this
charter
or
to
some
new
work
in
a
new
working
group
or
in
existing
work.
So
I
don't
like
to
see
that.
Thank
you.
So.
A
P
M
Lou
Berger
I
think
we
could
narrow
the
discussion
a
little
bit
on
the
choice
for
the
room
for
the
for
the
80s.
You
know,
I
would
assume
we're
not
going
to
go
sort
of
take
on
the
work
of
the
radio
layer,
guys
who
own
those
thighs.
You
know
we're
not
going
to
do
that.
We
probably
don't
want
to
have
two
groups
doing
the
same
thing
like
defining
protection
layers
for
debt
net.
You
know
that'll
probably
go
over
to
that
net.
That's
pretty
clear!
There's
prop!
M
Maybe
some
things
that
deal
up
enhancements
always
gonna
continue
in
MANET,
wouldn't
come
into
this
group,
so
you
know
what's
really
left
it's
it's
the
architecture,
it's
sort
of
the
framework,
it's
sort
of
understanding
how
to
apply
it
to
a
different
environment.
You
know
that
that's
something
that
a
small
group
of
people
or
a
focus
group
of
people
can
go
do
and
you
know
whether
you
do
it
in
the
a
new
working
group
or
as
a
group
within
a
larger
working
group.
As
long
as
it
works
getting
done,
that's
the
most
important
thing.
M
B
So
Rick
again
as
as
I
understand
it
Pascal
and
the
other
guys
who've
been
working
on
this
and
got
it
as
a
buff
at
Montreal,
and
now
now
here,
I
see
as
I
understand
it.
They
are
asking
the
IETF
to
help
give
them
some
structure
to
and
some
more
formal
forum
in
order
to
have
this
discussion
so
yeah,
this
could
all
happen
inside
meetings.
B
Q
Tim
Costello
from
BT
the
good
good
presentations
and
support
the
work
earnest.
My
question
is
talking
about
down
streaming
or
putting
work
into
other
working
groups,
but
a
lot.
A
lot
of
that
was
referenced,
the
debt
net
group,
but
would
you
also
consider
work
going
into
working
groups
such
as
the
quick
and
that
to
adapt
the
protocols
to
make
them
more
wireless
friendly,
I.
B
Would
say,
as
the
one
of
the
great
difficulties
you
have
with
defining
working
group
charters
is
ensuring
that
the
work
is
achievable
and
scopa
ball,
and,
although
that
you
idea
is
admirable,
I
think
it's
so
far
out
on
the
periphery
of
of
what
we
should
attempt
to
do.
But
if
101
people
come
to
the
mic
and
say
we
want
to
go,
do
this,
then
that's
an
entirely
valid
discussion,
but
I
probably
don't
think
that's
raw.
F
R
Yeah-
and
this
is
the
fat
tied
down
from
our
way
so
I.
Firstly,
I
agreed
that
architectural
framework
stuff.
First,
it's
very
important.
So
could
you
move
back
to
the
page
four,
so
yeah?
Regarding
the
you
know,
technology
PSE,
I'm,
not
sure
if
this
terminology
a
crate
or
not,
but
you
know
it's
it's
a
little
bit
that
you
know
computing
about
the
relationship.
Tinder
PCE
and
you
know
RPS
year,
I'm,
not
sure
if
PCH
to
be
involved
in
the
law
or
not
actually
Islander,
you
know
our
problems,
presentation
slides.
R
It
was
mentioned
that
PCE,
it's
out
of
the
scope
of
you,
know
law.
So
I,
if
pc,
if
not
involved,
I
don't
know
how
they're
gonna
obviously
can
get
it
to
pass
information.
You
know
on
to
be
modified,
so
our
this
turn.
You
know
I'm
comment,
another
command
its
and
are
in
their
arm
program
down
pretty
prevalent
statements
are
presentation,
slides,
there's
a
comparison
table
so
unless
I
I,
don't
think
that
order.
You
know
comparison.
Our
assumption
are
correct,
for
example,
for
the
time
scale
it
says
that
you
know
I
for
the
PC.
R
B
You
so
a
little
bit
of
clarification
in
response.
The
this
diagram
is
talking
about
logical
blocks,
so
these
are
functions.
It
doesn't
mean
that
they're
not
co-located,
and
it
doesn't
strictly
define
responsibilities.
It's
more
of
a
a
whiteboard
rough
diagram
to
talk
about.
There
is
a
selection
function
that
must
occur
that
works
at
a
different
time
scale
to
the
global
path.
Computation
function,
how
that
is
actually
manifested
or
whether
it
actually
exists
as
a
thing
is
something
that
could
be
discussed
in
any
potential
work
that
comes
out
of
raw,
so.
A
Another
thing
I
was
good
at
is
that
the
time
scale
also
has
to
take
into
account.
What's
the
Equality
of
the
links
that
you're
running
over,
so
you
can
clearly
run
a
centralized
compute,
you
know
path,
computation
element,
if
you
have
your
lots
of
bandwidth
and-
and
there
are
little
drops
that
could
do
things
very
quickly
right,
because
it's
just
limited
by
compute,
if
you're
running
over
somewhat
unreliable
wireless
link
at
varying
speeds
over
a
large
network
that
might
not
be
feasible
at
all.
A
N
N
A
Guess
I
guess
one
question
for
the
room
me
not
being
a
dotnet
expert
is:
is
there's
a
notion
of
being
able
to
do
debt
net
extensions
that
don't
actually
change
the
debt
net
architecture
but
fill
in
some
pieces
and
being
able
to
do
that
in
a
focus
group?
That's
not
debt
net
itself,
right
I!
Suppose
it
needs
to
be
done
in
the
group
into
to
make
it
easier
to
invite
a
different
community
that
can
focus
on
on
a
device
problems,
but
did
that
question
make
sense?
I
was
rambling.
A
What
you
know
we
don't
want
to
have
a
different
architecture
here
right.
We
want
to
have
used
a
definite
architecture
but
figure
out
it.
If
you
add
the
missing
pieces
to
support
these
Wireless
use
cases,
and
is
it
feasible
to
do
that
open-ended
question
without
being
in
the.net
group
itself,
feeding
their
work
into
that
not
so
much.
M
That
that
box
goes
into
the
service
layer.
We
have
one
solution.
Today.
We've
talked
about
other
solutions
for
providing
reliability
over
a
different
network
technology,
so
the
that
last
piece.
What
is
that
number
for
doing
a
different
way
of
doing
redundancy
fits
right
into
that
architecture?
Perfectly
and
personally
I
think
it
fits
right
into
the
scope
of
the
working
group
and
you
can
go
do
it
there.
All
you
have
to
do
is
bring
us
a
brilliant
proposal.
Bring
a
draft
we'll
talk
about
it.
The
OEM
one
David
was
right
on
we're
in
the
process.
M
Oh
I
am
is
trailing
behind.
We
have
pretty
good
handle
on
what
we're
doing
for
MPLS
OEM.
We
don't
have
a
good
handle
on
IP
o.
Am
you
actually
could
help
us
solve
our
problem
and
by
the
way,
Dan
net
is
not
wireline?
Debt
net
is
debt
net.
Just
like
IP
is
not
what
you
know.
Wireless
or
wireline.
It
code
goes
over
any
network.
That's
what
we're
aiming
to
do
with
that
net.
M
B
So,
there's
a
quick
follow
up
from
that
wreck.
Again,
we've
never
had
the
intention
of
carving
out
our
own
little
Empire
or
you
know,
a
raw,
isolated
working
group.
It's
always
about
interaction
with
other
people,
definitely
debt
net
because,
as
has
been
repeated
many
times,
there's
so
many
similarities
and
there's
a
direct
inheritance.
B
My
question
to
Lou
is:
if
we
turned
up
with
all
those
drafts
and
I
won't
find
the
slide,
but
there's
about
nine
and
said:
okay,
we
want
to
work
on
all
of
these,
and
here
are
all
these
guys
and
I'm
looking
at
guys
and
girls
in
the
room
here.
If
we
all
turn
up
on
mass
into
debt
net
and
say
we
want
to
work
on
this,
how
disruptive
is
that
going
to
be
to
your
working
group
and
that's
a
genuine
question.
M
Sure
so
I
think
items
three
and
four
or
completely
complementary.
So
that's
that's
like
a
no-brainer
one
and
two.
You
know
if
there's
a
group
that
really
wants
to
talk
about
that
in
debt
net
well,
I
think
they'll
accommodate
them.
If
we
want
to
have
that
run
as
a
small,
focused
working
group,
I
think
that's.
Okay,
too.
You
know
I
as
I
said
earlier.
It's
about
getting
the
work
done
and
you
know
I
also
not
getting
competing
solutions.
M
M
But
in
terms
of
the
focus
of
the
discussions
you
know,
do
we
separate
together
we'll
have
to
coordinate
either
way,
I
think
there's
gonna
be
a
set
of
us
that
are
gonna,
go
to
both
meetings,
no
matter
what
so
you
know
the
fact
that
Pascal
was
talking
about
people
not
showing
up
and
people
having
to
go
to
multiple
means.
We're
gonna
do
that
anyway,
it's
not
a
big
deal.
That's
why
we
that's
why
we
still
have
meet
in
person
by
the
way
yeah
yeah.
K
Yeah
I've
heard
a
lot
about
this.
Like
you
know,
where
do
we
do
work,
and
you
know
existing
working
group
or
new
one
and
so
on?
You
really
come
to
talk
about
that,
but
rather
to
suggest
that
there
might
be
a
opportunity
to
change
the
topic
slightly
so
I'm,
mostly
interested
in
how
much
sort
of
market
demand
there
is
for
this
kind
of
suggested
functionality,
no
matter
where
that's
that's
being
done
and
I
think
that
would
be
a
worthwhile
thing
to
try
and
get
people
to
comment
on
and
just
thought
of
them.
K
My
personal
view
on
this
is
kind
of
like
mixed
a
little
bit,
because
clearly
there
seems
to
be
demand
for
this.
You
know
overall,
like
you
know,
low
latency,
critical
communications
type
of
functionality
in
the
world,
and
you
know
factories,
switching
to
new
types
of
communication
tools
and
so
forth,
and
no
evidence
for
that
is
like
the
work
at
I
Triple
E.
That
was
just
heard
about
the
work
at
3gpp
and
if
IT
and
and
many
other
places
this
sort
of
layer
to
capability
is
being
manufactured
to
to
do
that.
K
That
sort
of
service
but
I,
think
the
question
is
not
about
the
overall
market.
Really
it's
about
like
the
Delta
that
we
at
ITF
would
provide,
and
certainly
we
need
to
somehow
take
into
account.
You
know
what's
happening
layer
to,
but
the
the
question,
in
my
mind
at
least,
is
how
much
like
are
we,
you
know
happy?
Well,
we
have
happy
with
you
know
what
didn't
was
already
about
to
do.
Are
we
gonna
create
there
like
a
big
system
on
top
and
and
there
I
think
leister?
K
Maybe
the
question
mark
and
if
you
look
at
way
things
are
built
then
there's
a
you
know.
Of
course
the
question
of
we
do
things
at
layer,
3
or
or
layer
2,
and
it
seems
to
me
that
the
kinds
of
things
that
being
talked
about
here
are
most
suitable
for
so
the
extreme
situation
of
you
know
have
a
mesh
type
network
of
multi,
hope,
network
and
possibly
multiple
standards
and
then
sort
of
the
IP
is
the
only
way
to
get
that
done.
K
And
if
you
compare
that
sort
of
fundamental
things
that
you
can
do
at
layer,
3
like
well,
the
packet
can
cut
through
and
didn't
get
through,
whereas
in
radio
you
have
like
this
64
element
antennas
that
you
can
measure
like
every
beam
exactly
and
so
so
when
you
make
a
selection
at
layer
2,
you
sort
of
you
know
fundamentally
tweaking
the
physics
you
quite
aware
of.
What's
going
on
and
and
layer
3
in
comparison,
maybe
there's
a
bit
of
a
extreme
comparison.
But
but
it's
it's
like
you
know,
shooting
in
the
dark.
K
You
don't
really
know
it
like
it's
just
don't
off
kind
of
thing
and
to
me
that
speaks
you
know,
maybe
more
in
favor
of
relatively
small
amount
of
work
at
the
ITF,
and
you
know
make
sure
that
we
can
detect
things
that
are
broken
and
are
working
and
leaving
the
radio
layer
to
do
do
his
job,
because
it
does
have
all
the
tools
available
available
to
it
to
figure
out.
What's
actually
going
on.
K
B
B
S
Just
I'm
supporting
you
in
that
my
background
is
actually
upon
radio
networks,
so
I'm,
one
of
the
guys
building
the
radios
for
urban
networks
and
I
think
the
room
is
maybe
not
aware
of
that,
but
basically
I
live
in
a
parallel
universe.
We
have
a
whole
set
of
standardization
bodies
in
the
urban
industry
that
standardized
airborne,
internetworking
and
as
a
internetworking,
because
we
still
use
always
AI
protocols.
S
E
E
We
have
this
exact
problem
that
within
this
single
technology,
we
have
to
be
of
multi-hop
routing,
and
we
do
that
at
layer
3,
because
we
found
that
the
layer
3
technology
applied
quite
well
to
our
use
cases.
So
we
did
not
rely
on
layer,
2
meshes,
but
we
do
layer,
3
meshes
and
one
hop
can
do
the
best,
even
if
it
had
64
antennas
in
wanna
write
to
optimize
that
transition.
E
So
the
bottom
line
is,
we
need
this
technology
in
our
current
product
line
and
we
have
customer
requirement
that
are
associated
to
that
in
real
products
as
not
have
Unix
that
smart
grid,
so
just
to
enable
automation
within
a
smart
grid.
People
invest
a
lot
in
those
networks
right
now.
They
use
them
for
I
am
I
but
defined
that
considering
the
investment
they
want
more
services
from
this
network.
So
now
they
are
hiding.
They
want
more
capabilities
from
the
network
and
and
basically
those
capabilities
are
related
to
automation
and
they
want
to
grow.
J
Yeahjust
and
it's
mario
here
again,
so
one
really
quick
comment
about
bringing
new
people
to
the
IETF.
Why
wireless
is
important
to
do
that,
so
the
difference
is
here:
you
need
to
do
good
channel
measurements
in
order
to
understand
your
wireless
channel,
and
this
is
becoming
ever
more
important
with
the
advent
of
you,
these
coming
in
autonomous
flying
and
so
on.
J
D
D
It's
especially
for
the
point
for
you
guys
are
seeking
for
the
time
scale
issue
you
guys
are
seeking
to
put
in
band
signaling,
so
I
didn't
hear
which
data
plane
you
guys
like
icbt
is
one
data
plane,
but
it's
not
sure
we
are
seeking
to
create
new
data
planes,
but
I
see
this
is
a
new
thing
in
band
signalling,
there's
a
one
in
band
signaling
stuff
that
was
being
presented
in
TS.
We
working
group
using
ipv6
extension
headers,
but
I
see.
This
is
a
new
thing
which
is
not
done
in
deathmatch.
That's
my
comment.
A
So
that's:
let's
try
to
get
some
hum
so
not
about
we're.
Necessarily
where
we
placed
is
work
or
maybe
there's
a
right.
I
know
there
will
be
question
about
for
medium
working
group.
Sorry,
but
first
of
all
you
know
the
people
in
the
room,
please
hum
I,
will
ask
yes
or
no
will
ask
both
of
these
questions
as
one
question
and
we
can
split
them
apart
if
we
need
to.
A
T
Hello,
it's
Debra
the
ad.
Can
you
go
back
to
that
slide
here?
Okay,
so
I
have
heard
a
lot
of
concern.
There
were
four
items
right
for
the
work
items
and
the
first
two
were
informational.
The
third
and
fourth
were
warm
protocol
which
questionable,
if
it
we
should
be
dead
net
or
what
so,
if
it
was
only
those
first
two
items.
T
A
A
That
one
was
yes
quite
different
right,
so
it
was
not
completely
quiet
on
the
second
one.
What
no
I
said
it
was
not
completely
quiet
under
the
on
the
opposed
one,
but
it
was
not
volume
things
I
would
say
it
was
like
8020
90/10
type
thing
as
opposed
to
5050.
So
and
then
you
know
who,
in
this
room
are
planning.
If,
if
we
charted
that
yes.
K
E
E
Another
grown
piece
of
the
architecture
that
does
not
even
show
on
the
architecture
has
written
today
because
in
many
wire
deployment
doesn't
is
not
needy
at
all,
and
so
the
do
you
have
a
way
of
presenting
this
work
is
to
say
a
what,
if
that
net
delegates
this
work
to
role
and
does
not
try
to
do
the
competing
solution
just
since
it's
a
very
particular
side
effect
of
that
net,
which
only
appears
in
a
particular
types
of
use
cases.
Some
people
big,
not
easy.