►
From YouTube: ROLL WG Interim Meeting, 2019-10-14
Description
ROLL WG Interim Meeting, 2019-10-14
A
A
This
is
the
algebra,
you
can
say
a
pearpod,
please
put
your
name
in
vapor
path,
I'm,
contributing
with
the
opinion.
Okay
and
today,
please
be
aware
that
this
video
is
based
on
not
well,
and
we
are
going
to
discuss
about
the
capability
is
features
that
we
want
for
you
both
okay,
very
well.
We
are
not
going
to
repeat
it,
but
we
speed
it
up.
A
is
aligned
with
that
not
well
basic,
a
word
path,
but
this
is
the
credit
score.
A
B
D
B
Yeah
this
is
much
better.
There
is
no
sound
background
song
now,
okay,
so
let's,
let's
begin
just
a
quick,
just
a
quick
recap
of
what
we
discussed
last
time.
The
first
was
the
problem
of
lighting
the
options
we
discussed,
what
the
problem
was,
and
we
had
a
quick
solution.
We
had
this
discussion
back
on
mailing
list.
Having
said
that,
as
of
now,
we
haven't
so
last
time
we
discussed
about
using
the
lower
four
bits
of
lower
four
bits
in
the
reserved
section
in
the
base
di
of
message.
B
As
you
can
see
on
the
screen,
the
small
pink
color
box
on
the
right
hand,
side
of
the
message
shows
the
number
of
bits
that
we
are
going
to
make
use
of.
This
was
the
first.
This
is
the
first
point,
so
this
four
bets
one
of
the
problems
that
I
thought
about
was
if
we
are
going
to
just
make
use
of
four
lower
for
bets,
is
it
going
to
be
sufficient
enough
for
for
the
lollipop
sequence?
B
B
B
So
the
root
reboots
or
for
that
matter
any
other
node
as
well.
Yes,
yeah
in
this
case
the
root
has
to
reboot,
because
root
is
the
one
who
is
controlling
the
sequence
counter.
No
other
node
is
going
to
increment
the
sequence
counter.
So
so,
if
the
root
reboots
and
then
after
reboot,
it
comes
up
with
an
updated
set
of
information
and
if
ends
up
using
the
same
value
as
the
previous
counter
previous
value,
then
then
the
nodes
won't
get
updated.
Is
it
is
it?
Is
it
true.
C
B
C
B
D
F
C
We
have
like
16,
out
of
so,
we
have
a
large
place
where
we
can
detect
that
somebody's
out
of
sync.
So
we
only
only
a
four
bit
usable
in
the
current
sequence,
counter
which
has
like,
which
is
one
backed.
So
if
we
used
any
for
base
and
we
use
a
lollipop
and
that
we
just
set
to
like
two
bits
or
something
so
small.
B
E
G
D
A
C
C
E
C
C
B
E
B
Maybe
there
is
no
way
for
the
for
the
China
notes
to
know
that
the
root
is
rebooted
right
now,
even
if
the
so
e
e.
If
the
look
if
a
root
has
started
and
if
it
is
still
in
the
lollipop
I
mean
in
the
linear
part
of
the
lollipop
counter,
it
does
not
mean
I
mean
it
can
be
in
the
linear
part
of
the
lollipop
counter
for
a
substantial
time.
Frankly
speaking,
there
is
no
way
food
non
non
root.
Nodes
can
know
that
root
has
rebooted.
Oh.
B
G
E
Seems
to
me
that
we
should
have
one
of
these
numbers
be
resettable
to
zero,
which
indicates
that
you've
rebooted
we
never
roll
it
up.
So
no
one
uses
the
four
bits
that
we're
going
to
create
here
and
say
a
configuration
option.
Zero.
Maybe
one
is
in
the
linear,
stick
part
of
the
lollipop.
Then
that's
the
indication.
You've
rebooted,
but.
C
C
F
E
C
This
message
is
lost.
You
know,
assets
the
DI,
you
spread.
If
some
notes
music,
let
the
next
year
utility
will
it
will
not
be
in
the
linear
part,
so
we
will
never
see
that
the
root
as
we
put
it.
So
that's
the
reason
why
we
have
a
linear
part
is
to
make
sure
to
give
it
slack
retries,
if
you
like,
if
you
have
like
former.
D
C
C
C
C
F
C
B
B
G
C
I
agree
with
you
how
okay,
unless
we
give
some
rules,
you
know
to
to
exceed
this
version
quickly,
like
of
this,
the
linear
part,
very
small
in
the
version
number,
or
maybe
just
one
value
and
then
set
that
value
to
one
very
quickly,
so
that
you
know
you
can
use
it
to
detect
the
root
of
the
root,
but
it
will
take
a
different
draft
which
explains
how
that
work
current.
We
could
make
it
work
right,
but
it's
not
there.
It's
missing
right.
C
B
C
B
B
E
E
E
B
I
mean
I
I
mean
the
reboot
is
a
is,
is
an
issue
regardless
of
regardless
of
this
discussion.
It's
a
it's
a
it's
a
problem
that
I
have
even
mentioned
in
the
empathic
draft.
The
observation
strap
that
it
is,
is
impossible
for
a
node
to
power
for
other
nodes
to
discover
that
the
node
upstream
has
rebooted.
So
this
this
is
a
different
problem
altogether.
Having
said
that,
I
think
we
should
continue
with
our
discussion
with
regards
to
this
drop.
The
other
point
so
shall
we
shall
we
shall
we
continue
to
to
the
next
point
here?
B
C
B
B
B
The
the
reason
why
I
say
it
has
utility
outside
the
capabilities,
because
this
is
a
generate
counter
which
not
only
likes
capability
option
but
a
more
pecs
option
in
the
future,
as
well
as
existing
configuration
option
and
as
discussed
on
the
mailing
list.
This
could
as
well
be
made
use
of
for
other
options
such
as
prefix
information
option.
D
B
Each
piece,
so
basically
what
I
was
thinking
is
all
the
static
information
which
rarely
changes
and
prefix
information
option
is
one
thing
that
usually
doesn't
change.
So
in
this
case
I.
The
prefix
information
option
currently
is
a
different
option
altogether.
It's
not
part
of
any
other
container,
so
so
so
now,
if
we
have
and
I
feel
it
can
be
a
separate
container
altogether,
it
can
be.
We
don't
need
to.
We
don't
need
to.
We
don't
need
to
say
that
the
prefix
information
action
has
to
be
carried
inside
some
other
container
for
it
to
be
lighted.
B
B
B
Pascal
but
I'm
not
I'm,
not
saying
that
we
change
the
format
here.
We
just
say
that
whether
the
PIO
is
carried
or
not
carried
in
the
DI.
If
we
say
that
it
is
look
right
now,
we
need
an
option
which
says
that
the
information
that
I
have
is
stale
information.
With
this.
With
this
new
counter,
we
know
that
a
node
has
this
mechanism
of
of
detecting
whether
the
information
it
has
is
the
stale
information
or
it's
the
present
fresh
information.
B
C
B
C
E
C
E
C
Like
it
so
so
make
it
one
right,
because
many
things
will
change
now:
okay,
mate,
why
not
use
all
the
octet
and
define
it
exactly
like
the
the
pass
sequence?
You
know,
because
we
have
all
the
text
in
Section
seven.
Why
don't
we
just
say
this
is
analysis
because
counter
which
operates
just
like
the
others.
B
Iii
think
Pascal
that
that
problem
cannot
be
solved
with
simply
a
lollipop
counter
right.
I
mean
that
reboot
problem
is
the
different
problem
altogether.
If
it
will
become
so,
you
mentioned
that
the
root
when
it
reboots
it
will,
it
will
just
increment
the
sequence
counter
to
some
other
number.
The
problem
is:
how
soon
can
the
root,
eventually
increment
the
sequence
counter
with
the
a
lot
of
other
problems,
I
mean
a
lot
of
other
points
that
have
to
be
talked
for
handling
the
reboot
case.
B
Having
said
that,
if
you
forget
the
reboot
case,
if
you
simply
say
that
it
is
a
lollipop
counter,
the
moment
we
say
a
lollipop
counter,
we
it's
assumed
that
the
linear
part
has
to
be
stored
in
the
external
storage
for
the
node,
and
if
the
node
is
rebooted
it
starts
it
starts
from
the.
If
it
is
in
the
linear
portion,
then
it
starts
from
the
next
value
in
that
linear
portion
yeah.
B
B
But
I
mean
couple
of
ideas
back
I
did
I
mean
this.
This
problem
is
explained
in
detail.
In
fact,
this
problem
is
stated
in
detail.
In
the
observations
raft,
it's
there's
a
complete
section
on
the
saying
that,
if
it
reboots
in
the
linear
part,
then
the
linear
portion
has
to
be
kept
in
the
external
storage.
Now
this
is
a
statement
that
I
have
made
in
the
observation
draft.
Now,
if
you
now
you're
saying
if
it
is,
if
it
is
incorrect,
if
the
statement
is
incorrect,
then
I'm
not.
G
E
My
suggestion
here's
my
proposal
arm
that
we
write
a
document,
creates
this
sequence
number
one
byte
sequence,
number,
not
sure
I,
like
the
word
sequence
number,
but
configuration
number
thing,
and
we
also
include
in
it
clarifications
of
or
lollipop
counters.
So
it's
an
update,
6.5
500
anyway.
To
put
this
thing
in,
we
take
all
the
text
from
the
observation
drop
on.
E
C
Okay
I
mean
this
works
like
on
the
only
prime
with
it
is
the
is
G
doesn't
like
to
sweet,
very
small
drafts
these
days.
That's
why
we
do
we
reflect
write
this
text
anyway,
but
in
terms
of
packaging,
we
may
ask
the
D
ID,
because
it's
not
just
what
use
what
you're
proposing
is
cleaner
is
the
right
way
of
doing
it.
But
then
there
is
the
quotient
called
political
aspect
to
it,
where
the
agency
wants
to
have
less
glass,
I.
E
Understand
they
want
less
drafts,
but
then
they're
going
to
have
to
spend
a
lot
less
time
on
nitpicking
big
drafts
if
they
want
to
have
less
drafts
and
they're
going
to
get
big
drafts
to
do
many
different
things.
They're
going
to
spend
a
lot
less
lot
less
time
nit
making,
because
the
big
drafts
take
forever
to
process
so
yeah.
A
I
agree
with
Michael
that
time
documents
specify
the
counters,
so
we
can
go
further
with
it
with
a
work.
I
mean
if
I
don't
know
about
the
is
G,
but
I
think
is
that
I
wait
to
do
it?
We
just
do
it.
I
mean
we
have
their
uses.
I.
C
Said
Cheryl
I
have
to
discuss
a
serious
case
a
year
ago
or
two.
You
know
the
ASG
made
us
put
two
different
things
which
were
big
each
one
in
a
single
draft.
If
you
know
the
lt1
shake
trust,
it
has
fragmentation
and
compression
in
the
same
draft
and
I
oppose
that
as
a
chair,
exactly
like
you're
doing
and
won't
be
two
separate
drawers,
TV
ad
Raiders
with
that
in
one
big
document,
so
I'm
saying
is
I
agree
with
you
all.
Let's
just
ask
the
valve.
B
You,
okay,
so
continuing
right
right
now,
I'm
calling
this
counter
set
counter,
but
I
really
wanted
to
call
it
as
a
seal
counter,
but
I
couldn't
think
of
any
appropriate
full
form.
If
you
guys
can
think
of
it,
it
was
great.
Basically
if
we
call
if
we
can
call
it
a
seal
counter,
then
we
know
how
it
operates
like
it
seals
the
information
for
the
specified
unless
the
next
seal
is
applied.
B
B
Okay,
so
we
have
already
discussed
that
what
is
the
applicability
of
the
set
counter?
Currently
I
mean
we?
We
are
thinking
of
colliding
in
the
future.
We
might
take
like
capability
option,
the
more
pegs,
and
it
is
possible
that
in
the
future
we
might
collide
other
static
configuration
information
such
as
prefix
information
option.
Now
it
I
thought
about.
One
point
is
that
if
different
nodes
decide
to
like
different
information,
for
example,
if
there
is
a
node
downstream,
which
decides
to
Allied
prefix
information
option,
but
there
is
another
node
which
does
not
decide
to
like
that.
B
Will
that
be
a
problem?
And
as
of
now
I
couldn't
I?
Don't
think
that,
should
that
will
be
a
problem
since,
since
since
we
have,
we
would
have
a
clear
mechanism
of
telling
whether
the
information
is
fresh
or
stale
in
the
future,
based
on
the
sequence
counter
in
the
set
set
counter,
so
that
should
not
be
a
problem
net.
The
set
counter
is
going
to
be
controlled
by
the
root
node
alone,
and
only
the
root
node
can
change
the
counter
increment.
The
counter
in
both
in
all
the
stories
in
all
the
mode
of
operations.
B
C
In
part
of
it,
but
there
was
confused
because
what's
the
point
I
mean
either
of
people,
everybody
has
everything
or
they
culturally
upper
right
right.
We
are
talking
about
information
which
is
necessary
for
operation
like
the
configuration,
the
PIO
etc.
So
if
you
have
the
PIO,
but
the
set
is
not
good,
you
still
need
to
Center
this
anyway
to
see
the
rest,
because
you
can
not
operate.
It's.
B
B
No
I
Pascal
you're
right.
You
know
that
there
is
no
point,
but
having
said
that,
I
was
just
thinking
if
the
implementation
datian
for
some
reason
now,
I,
don't
know
the
reason,
but
for
the.
If
the
implementation
decides
to
not
light
the
particular
option,
then
what
happens
I
was
thinking
of
the
worst
is
here.
What
I'm
trying
to
say
is
even
if
this
happens,
there
won't
be
much
of
an
issue
because
we
have
yeah
yeah,
maybe
a
probably
good,
because
yeah.
B
F
C
F
C
Second
thing
is:
if
I
got
only
two
out
of
this
frame,
then
I
need
to
be
able
to
ask
again
or
the
third
and
today
daddy,
doesn't
say
well.
Thank.
G
B
Iii
think
partially
raised
another
four
important
point
that
you
know
the
right
now.
There
is
no
way
of
saying
whether
all
the
information
is
retrieved
or
not
correct.
The
node
has
been
linked,
all
the
information
or
not.
There
is
no
way
of
knowing
that,
but
having
said
the
as
of
now
I
didn't
I,
don't
think
we
have
I
mean
we
have
not
faced
such
a
problem,
but
as
the
DDI
was
growing
inside,
we
might
take
this
problem
so.
D
B
C
C
C
B
B
B
She'll
be
more
excited;
basically,
that's
that's
pretty
much
it
from
you.
Okay,
the
capability
is
option
syntax
now-now-now
this.
We
we
had
a
major
discussion
about
this
in
the
last
interim
call
earlier.
The
currently
the
draft
considered
caps
are
the
sequence
of
bits,
but
we
are
clearly
moving
away
to
work
from
from
this
sequence
of
bits
to
a
TLB
format.
Now,
in
the
last
meeting,
we
discussed
about
having
capybaras
having
several
cam
boots.
B
One
is
whether,
if
the
cap
kept
capability
bits
or
not,
if
the
capability
bit
or
if
the
capability
is
not
understood,
then
should
a
no
join
as
a
router
or
leaf.
This
can
be
indicated
as
one
bit
whether
the
information
has
to
be
copied
to
children
now,
I'm,
not
sure
I
mean
thinking
for
further
on
this
I'm,
not
sure.
If
we
really
need
this
this
bit
to
be
present.
If,
if
the
capability
is
not
understood,
then
is
there
any
reason
that
a
node
will
copy
the
capability
information
to
its
Stratton
box?
B
E
B
B
I
think
I
think
the
context
is
different
right
now.
What
you
are
saying
is
it's
kind
of
definitely
possible
with
the
current
capabilities
set,
but
what
we
are
discussing
right
now
is
is
a
different
problem
altogether,
it's
a
specific
point
about
whether
if
a
node
doesn't
understand
the
capabilities,
if
a
6ll
doesn't
understand
the
capabilities,
should
it
copy
this
capability
to
the
children
whether
there
needs
to
be
a
bit
specific
so
now.
G
E
To
do
here
is
we're
trying
to
future-proof
our
situation
so
that
we
can
find
out
who's
capable
what
nodes
are
capable
of
something,
and
in
some
cases
we
may
decide
to
enable
some
feature.
That
might
mean
the
network
is
going
to
reconfigure
itself,
because
some
devices
no
longer
can
operate
as
routers.
They
have
to
offer,
as
leaves,
but
we
don't
know
that
until
we've
actually
asked
them
all,
and
so
the
ones
that
don't
know
what
the
capability
is
they
may
become
leaves
and
some
of
their
children
could
become
parents
right
that
this
is
the
view.
F
E
Topology
could
be
different
as
a
result.
We
may
decide
that's
a
good
thing.
You
may
decide
afterwards
the
terrible
thing
in
our
apologies
crop
and
we
should
turn
that
capability
that
feature
off
to
leave
upgrading
more
nodes,
but
but
we
won't
know
that
unless
we
actually
can
interrogate
all
the
node,
that
means
we
need
to
be
able
to
ask
all
the
nodes.
Do
you
support
this,
and
so
that's
why
the
question
you
know
we
think
that
I
think
capabilities
need
to
be
copied
to
the
children.
E
G
G
B
B
B
B
B
B
Yeah
in
the
subsequent
slide,
I
will
just
come,
come
to
the
exact
format,
so
so
so
going
at
the
another
thing
that
we
discussed
in
the
last
interim
Watts,
we
need
additional
capability
information.
This
information
might
be
optional
and
again
on
perk
ability
basis.
So
we
need
a
map
sensor
using
which
extended
information
can
be
optionally
specified
as
or
on
per
capability
basis.
Can
we
go
on
to
the
next
slide?
Please?
The
next
slide
shows
the
format
of
the
capability
options.
The
first.
B
The
first
line
shows
that
the
all
the
capability
tlvs
will
be
containerized
into
a
single
option,
individual
tlvs,
then
in
the
second
second
part,
you
can
see
that
it's
an
individual,
capable
capability
type.
It
has
three
flags
right
now.
First,
is
a
join.
Only
as
leaf
is
capability
is
not
understood.
See
is
capability,
kept
copy
capabilities
to
children
and
I
a
flag.
Is
a
capability
information
option
present?
If
this
flag
is
not
set,
if
the
information
flag
is
not
set,
then
there
are
no
additional
extended
information
for
that
capability
option
now.
B
The
reason
why
we
are
doing
it
is
lot
of
capabilities
might
be
simple,
boolean
boolean
flag
on
or
off,
or
whether
it
is
supported
or
not,
but
certain
capability
options,
such
as
routing
information
and
neighbor
cache
information,
requires
additional
information
which
cannot
be
simply
stated
as
a
single
bit.
So
this
is
why
we
require
this
extended
information.
Now
the
extended
information
again
needs
the
length
and
then
the
actual
information
which
will
be
defined
by
the
individual
capability
specification.
Thus,
this
format
look
okay
arm,
don't.
F
E
C
E
How
do
you
distinguish
them
from
different
networks
and
ultra
stuff?
This
is
a
problem
but
and
I
say
as
I'm
saying
I
think
that
we
need
some
some
exactly
some
experimental
space
in
there.
Ok,
I
can
live
with
this
format.
I
just
I
would
say:
let's
go
back
and
look
at
the
other
TLV
formats
we're
already
using
in
and
see
if,
if
we
can
fix
this
of
a
one
byte
length,
maybe.
E
A
E
F
E
B
So
and
I
really
don't
I,
haven't
really
thought
in
detail
about
this
format.
The
only
thing
that
I
thought
was
it
is.
It
is
better
not
to
include
the
length
if
the
capability
information
is
not
to
be
present,
so
that
is
the
only
thought
that
came
to
my
mind
you
know
so
I
would
like
to
save
that
one
bite
additionally,
one
bite.
B
G
G
G
B
Anything
so
whatever
you
just
mention
is
exactly
same
as
the
existing
options
or
existing
existing
TLB
format
that
is
used
in
sixty
five
fifteen.
Only
problem
with
this
is
if
the
length
is
zero
and
there
will
be
lot
of
capabilities
which
will
happen
to
the
way
we
take
the
length
byte
is
useless.
So
what
I
was
trying
to
do
here?
Was
you
just
adjust
this
I
bit,
which
is
that
capability
poise
is
present
or
not,
and
then
avoid
ascending
length
altogether.
C
G
B
Already
have
a
use
case
where,
let's
say
this
way,
eight
one
three,
eight
one,
three
eight
turn
on
or
enable
or
supports
eight
one,
three,
eight
RS
political
to
eight.
So
that
is
just
a
judge.
Just
a
bar,
just
just
a
signaling
mechanism
which
which
doesn't
carry
any
value.
It's
just
a
flag
saying
that
it
supports
the
node
supports
it
would
be
eight.
There
is
no
value
0,
whether
you.
C
So
actually
mention
it.
138
I
was
thinking
whether
you
know
if
you
cannot
join.
If
the
capability
is
not
understood,
then
no
8
138
basically
says
that
the
types
that
you
have
to
recognize
you
don't
need
the
lungs,
because
it's
a
type
which
is
well
known
now,
I,
don't
think
it
you.
We
can
play
this
game
here
because
you
may
not
understand
one
capability.
You
may
still
join
us
a
leash
and
you
may
still
want
to
read
the
next
capability.
C
E
Let's,
let's
get
some
notion
as
to
what
the
capabilities
are
that
may
need
some
additional
data
and
and
then
let's,
whether
or
not
we
can
you
know,
maybe
we
only
need
to
have
four
kinds
of
lengths.
Like
you
know,
there's
one
bite,
there's
ten
bites,
there's
twenty
bites
right
and
we
don't
actually
need
a
whole
fight
of
the
length
we
can.
We
can
have
something
less
than
that
I,
don't
know.
Let's
get
some
of
the
notion
of
that.
First
I
think
and
then
let's
figure
out
what
what
we
need
for
the
lair.
Okay.
B
E
You
point
is
that
that
if
you
submit
the
capability,
I
support,
80
138
implies
a
whole
bunch
of
other
things
as
well.
We
don't
have
to
negotiate
them
separately
and
I.
Think
if
that's
what
you're
saying
when
I
completely
agree,
but
that
I
think
is
in
the
detail
of
of
the
capability
description
rather
than
the
framework
for
the
capabilities
with.
E
F
F
C
C
Now
you
could
have
one
bit
which
says
that
you
have
a
life
length
of
one
bite
and
then
whatever
info,
and
if
that
bit
is
not
so
with
one
bit,
you
could
signal
and
you
could
start
like
that
before
and
I
agree
with
Michael.
We
learn.
Let's
learn,
but
to
start
with,
we
can
say
Oh.
Either
we
have
data
or
we
don't
so
one
bit
to
say
we
have
data,
and
if
you
have
data,
then
you
have
one
byte
blogs
and
so
many
bytes
of
data.
B
C
Be
language
addresses
yes,
and
then
you
split
yeah,
you
do
well
I,
don't
see
the
lines
filled
right
on
a
I
would
like
the
bits
move
to
the
left,
okay
and
otherwise.
Yes,
we
need
to
decide
how
big
the
length
is,
but
if
we
are
formed,
if
we
have
four
flags
like
we,
we
may
need
a
better
flag.
We've
seen
it
with
our
for
one
flag,
another
and
then
twelve
cap
types.
C
C
B
Okay,
clearly,
one
thing
is
maybe
basically
right
now
we
had
both
the
discussions
that
using
length
as
part
of
the
base
capability
type
along
with
the
capability
that
we
also
have
length
or
we
have
I
bet.
The
only
thing
is
you
let
it
maybe
what
what
we
can
do
is
we
can.
We
can
explicitly
put
about
examples
and
then
decide.
Even
that
is
the
better
option
if.
B
C
B
C
E
F
F
E
Stuff,
that's
part
of
what
I
was
thinking,
maybe
the
case
I
don't
know
if
that's
a
good
idea,
but
but
also
you
know,
maybe
maybe
not
first
non
storing
and
storing
historic
ones,
but
maybe,
but
certainly
I
agree
for
a
new
mop.
So
we
want
to
do
this
and
I
and
I
strongly
think
we
need
a
new
mop
for
our
non
storing
with
down
projection.
B
So
one
of
the
things
that
we
were
trying
to
do
with
the
capabilities
was
to
check
whether
it
can
be
used
with
existing
mo
keys
with
the
existing
mods
but
looks
like
that.
Need
not
be
done,
or
maybe
that's
not
that's
much
of
a
hassle
or
maybe
it's
not
required.
Maybe
we
handle
it
only
with
anyone
mo
who
is
that
that
looks
like
that
is
the
position
well.
E
But
so
so
remember
so
we
had
some
conversation
about
how
we
might
have
to
go
and
pull
nodes
that
we
heard
about
right.
So
we
might
go
with.
We
might
start
off
with.
You,
know,
good
old-fashioned,
on
storing,
mop
and
put
some
cap
in
it
and
then
but
we
get
the
dowels.
So
we
know
who's
out
Shay
out
there
and
we
see
the
ones
that
did
not
reply
to
our
cap
poll
and
we
have
to
ask
them
directly
with
a
unicast.
E
B
B
B
So
one
of
the
important
point
that
we
we
discussed
but
haven't
really
concluded
is
the
new
messages
it's
not
mentioned
in
this,
but
new
messages.
So
we
discussed
last
time
that
maybe
the
I/o
and
out
is
not
the
right
place
to
carry
this
capability
options
now
I
was
thinking
what
are
the
repercussions?
What
are
the
waters?
What
are
the
side
effects
of
having
new
messages?
B
One
one
thing
that
came
to
my
mind
was:
if
a
load
decides
the
parent
set
based
on
this
additional
capabilities,
and
if
this
capabilities
change
keep
keep
changing,
there
are
certain
capabilities
which,
which
are
dynamic
capabilities.
What
happens
is
a
node
has
a
set
parent
set,
but
it
doesn't
know
the
current
capabilities,
in
which
case
after
it
chooses
the
parent
after
parent
selection
is
done
after
the
parents,
which
is
done.
It
has
to
again
query
further
for
that
parents
capabilities
set
using
additional
signalling.
B
B
Having
said
that,
we
don't
have
a
clear
set
of
I
mean
we
have
not
really
thought
about
this
in
detail
as
of
now.
But
the
only
point
that
is
mentioned
in
the
slide
is
what
happens.
Is
there
is
a
parent
switch
in
which
case
and
if
during
the
parents
which,
if
the
capabilities
are
required
to
make
the
decisions
decision,
which
in
that
case,
there
will
be
additional
signaling
that
will
be
required
by
the
new
doctor
parents,
which
one
is
which
yes.
C
G
C
B
B
One
thing
can
be
that
if
a
23-8
is
supported-
let's
say,
for
example,
I
mean
there
could
be
examples
for
K
leads,
for
example,
the
routing
table
size
that
we
are
trying
to
send
through
capabilities.
If
the
size
is
too
small,
then
there
is
no
point
in
using
that
parent
as
the
as
a
as
a
preferred
parent.
B
G
C
Addition,
the
the
capability
is
priced
like
you,
do
I
think
I
hear
that
the
capability
is
part
of
the
objective
function.
So
if
I
see
a
parent
with
without
a
certain
capability,
I
don't
know
yet.
Then
you
Iran
at
such
a
point
in
my
objective
function
and
then,
if
officer
then
I
get
the
capabilities
from
him.
Then
I
changes
is
value
in
my
objective
function.
C
So
not
not
the
right
that
I
captured
necessarily
but
the
fact
that
I
decide
to
use
them
the
chess
it
makes
it
a
two
step.
I
see
were
just
saying
how
it
makes
it
a
two
step
decision
specific
right.
So
whatever
saying
is
that
the
GAO
has
a
little
bit
of
a
concept
of
pulling
information
with
the
DTS
n,
but
it's
very
crude.
It's
not
reliable
and.
F
C
D
C
B
B
F
F
F
C
B
Right
right,
so
can
we
have
something
like
this
that
you
know
we
define
the
options?
These
options
can
be
carried
in
in
the
di
o
or
in
the
new
message.
Certain
capabilities
might
not
necessarily
might
not
need
to
be
carried
in
the
di
o,
where
maybe
those
those
capabilities
could
be
pairing.
I
mean
the
doubt
that
there
will
be
lot
of
capabilities
which
need
not
which
which
we
should
not
make
any
impact
on
parent
selection,
let's
capabilities,
so
these
capabilities
will
be
individually
the
when
the
capability
specification
for
this
individual
capabilities
are
defined.
C
G
G
B
G
Have
one
question:
maybe
it
is
not
related
to
this
topics?
We
are
discussing
the
mom.
The
time
I
was
asking
about
that
PC
right
path.
Computing
element,
I,
am
not
clear
how
that
he
liked
from
the
capabilities
how
he
will
going
to
get
the
network
topology
so
that
he
can
compute
a
best
path
using
some
path,
computation
algorithm,
unless
until
he
get
let's
say,
I
am
a
node.
If
don't
tell
who
are
my
neighbors
like
the
way?
Is
s
or
OS
with
works?
G
B
C
G
C
F
C
Neighbors
and
I
don't
know
how
much
of
that
will
go
in
the
RFC.
Probably
we
won't
say
much
because
we
don't
know
yet
to
take.
You
know
the
original
repo
as
a
lot
of
holes
it's
often
times
because
we
did
not
know,
and
so
we
said
people
will
try
and
get
good,
give
good
advice,
which
is
exactly
what
Howell
is
doing
with
this
document
with
observation
document,
and
so
we
will
improve
next
time
for
damming.
We
probably
will
just
say:
oh,
you
can
expose
neighbors
with
this
message.
Don't
put
too
many
of
something?
C
D
C
C
A
A
B
A
E
E
B
C
F
D
C
C
F
C
D
C
E
C
B
Amy,
hello,
hello,
yeah,
yeah,
yeah
yeah,
so
right
now,
as
I
understand,
there
are
four
documents
that
we
are
talking
about.
One
is
the
capex
okay
capability,
another
is
the
more
pecks.
Third
is
the
this
continuation
and
the
lollipop
count?
Fourth,
is
the
lollipop
counter
handling?
Is
that
what
we
are
saying.