►
From YouTube: IETF106-ROLL-20191118-1550
Description
ROLL meeting session at IETF106
2019/11/18 1550
https://datatracker.ietf.org/meeting/106/proceedings/
A
B
B
B
D
B
E
B
D
B
B
F
G
B
B
And
they
are
expired
in
internet
draft
that
we
should
work
on
in
the
near
future
to
be
aligned
with
milestones
and
that
they
are
as
well
related.
Introduct
are
driven
to
get
discussion
today
and
tomorrow
into
the
end
tomorrow.
Okay,
we
notice
that
they
did
have
repository
is
great
to
check
in
issues,
and
there
is
a
lot
of
people.
Updating,
including
issues
indeed
happen.
That
is
great.
B
This
to
list
is
more
useful
than
the
biggest
one
I
mean
I,
see
more
people
actively
participate
in
the
training
tickets,
indeed
hub,
but
the
still
so
we
are
going
to
track
these
geezers
and
as
well
them
all
tracking
system
tickets.
So
we
are
going
to
watch
pose,
but
it
is
that
the
community,
the
working
group
I
prefer
they
did
have
because
is
super
easy
right.
B
So
we
noticed
a
unaware
leaf
drop,
it's
aligned
with
user
people
in
Poland,
some
of
our
terminology,
then
officiant
and
video
relies
on
rebel
status
in
our
lives.
Then,
for
capability
mode
of
operation
graft
we
have
announced
three
relations
for
the
projection.
Sure
there
was
commenced
in
the
Middle
East
that
we
asked
to
the
working
group
should
be
used
a
new
mode
of
operation.
For
that
or
not.
B
B
B
B
So
this
is
that
enough
for
today,
so
I
want
to
start
for
ripple
observations
by
always
going
to
state
the
observation
of
which
issues
are
still
open
in
Ripon.
Some
of
them
then
Pascal
a
give
us
a
update
in
their
projections
and
which
the
next
steps
to
follow,
and
then
we
are
going
to
talk
about
capability
and
mode
of
operation
draft
and
then
we
have
our
backbone
open
floor.
H
I
So
what
is
this
draft
about?
So
we
have
been
documenting
most
of
the
observations
most
of
the
implementers
experiences
in
this
Jack
and
it's
it's.
Its
aim
is
to
clarify
6550
in
some
cases,
may
fix
it
and
provide
some
implementation
guidelines
and
so
in
the
book
in
the
course,
if
you
see
in,
if
you
have
read
the
draft,
you
might
see
that
there
are
questions.
So
every
every
section
has
a
questions
asking
it
should
we
do
it
in
this
way
or
other
way.
So
it's
essentially
from
the
implementation
perspective.
I
D
B
I
So
this
field
is
something
which
is
sent
into
D
IO,
which
informs
the
downstream
peers
to
update
the
DAO,
basically
update
the
destination
advertisement.
Now
it's
very
easy
to
handle
this
in
non
storing
mode
because
it
is
directly
controlled
by
the
route.
The
DTS
and
in
value
is
directly
controlled
by
the
route
in
case
of
storing
mode
of
operation.
It
becomes
a
little
bit
tricky
and
there
are
multiple
implementation
strategies
possible.
Now,
an
example
here
when
the
node
D
switches
from
B
to
C
now
no
B
needs
to
on
C's.
I
C
needs
to
have
not
only
the
information
about
no
D,
but
all
the
the
completes
do
subdue
that
routed
at
node
D
now
either
D
can
update
Big,
D,
Jason,
increment
the
DTS
and
then
the
and
reset
the
decay.
D
I
have
trickle
timer,
in
which
case
no
DS
the
Dow
can
go
to
see,
but
subdue
drag
also
has
to
be
updated.
If
that
is
one
of
the
way
to
handle
it
and
I've
seen
the
implement
some
implementations
doing
it,
the
other
way
is
node.
I
B
already
has
the
routing
information
on
behalf
of
all
the
nodes,
and
it
sends
an
aggregated
Dao
on
behalf
of
sub
Buddha.
Well,
this
is
a
great
option.
There
are
certain
issues
handling.
This
is
because,
if
the
subdue
dag
is
quite
big,
it
would
mean,
or
sending
a
lot
number
of
lot
of
lot
of
dowels
to
see.
Every
Dao
has
a
sequence
number
and
the
state
information
becomes
too
large
to
handle
implementation
wise.
So
it
has
a
direct
impact
on
the
RAM
and
things
like
that.
I
That
is
that
is
one
option,
but
then
we
have
seen
that
it.
It
has
a
very
I
mean
if
you
have
like
20
nodes
below
your
below
you,
then
the
Dow
object
is
really
really
big
object,
in
which
case
it
will
it
results
in
more
than
five
fragments,
which
case
the
performance
at
six.
Slow
fragmentation
goes
back.
Maybe
maybe
the
other
minimal
fragment
can
help
there,
but.
I
F
Pascal
we're
here
so
yes,
I,
see
that
the
specification
is
not
very
specific
about
exactly
what
you
say.
At
least.
What
was
in
mind
is
your
option.
That
is,
if
you're
in
norms.
If
answer,
if
you're
in
storing
mode
you
at
the
end
of
the
day,
you
need
to
have
this
state
about
all
the
children
and
children
of
children
all
that
they
send
out.
You
need
to
have
it.
F
So,
even
if
you
say,
if
you
do
your
option,
one
which
will
recursively
do
the
DTS
and
all
the
way
down,
you
will
end
up
in
Nazi
with
all
the
state
that
you
talked
about,
because
that's
that's
storing
mode.
So,
at
the
end
of
the
day,
you
need
a
state
for
all
the
children
and
you
need
to
know
the
latest
path
sequence
for
all
the
children.
You
need
that
state,
so
version
two
that
you
mentioned.
F
F
How
much
messaging
goes
through
and
I
would
I
would
say
that
in
terms
of
bytes
in
the
air,
if
you
have
to
go
the
way
to
G
4G,
to
do
the
dowel,
it's
more
bytes
in
the
air
than
if
D
reconstructs
it
I
would
also
argue
that
we
should
pack
many
and
we
can
many
targets
into
a
single
dowel,
make
the
Dow
as
big
as
possible,
but
reasonably
big,
because
after
that
the
transmission
has
some
chance
to
fail
right.
So
it's
not
one
that
I
prefer
yet.
F
F
I
You
brought
up
the
compression
issue
here.
You
know,
there's
one
more
thing
that
one
more
observation
that
that
that
we
have
in
context
of
this
compression
every
target
and
most
likely,
whatever
the
target
information
that
D
has
each
one
of
the
target,
has
a
different
path
sequence
and
be
a
different
life
path
life
time.
I
Having
said
that,
what
you
say
said,
this
C
needs
to
have
all
the
information
and
if
you
do
reset
trickle,
timer
and
increment
the
DTS,
and
it
is
definitely
going
to
cause
more
control
over
head,
whatever
we
do
as
compared
to
D
sending
the
aggregate
at
now.
That
is
very
clear
and
but
it
is,
it
is
condom.
Easy
to
implement
and
I
can
see
that.
J
Michael
Richardson,
so
I
I
have
I.
Think
three
questions
and
we've
talked
about
this.
What,
since
the
first
version
of
your
draft
right,
maybe
was
three
years
ago
right,
okay,
so
I
think
that
this
may
be
appropriate
for
I,
don't
know
if
we
want
to
put
this
in
a
separate
draft
or
just
writing
it
right
eat
down
as
an
errata
four
six,
five,
five
Oh
for
the
moment.
J
Once
we
have
working
group
consensus
on
the
right
answer,
I
mean
we
can
leave
it
in
your
draft,
but
I'm
just
I'm,
not
sure,
procedurally,
what
we
will
do
with
it
with
the
conclusion
I
think
we
can
make
it
I
think
we
can
make
a
conclusion.
Finally,
for
this,
that
we
should
yeah
sure
so
so
that
was
my
first
question:
I'm,
not
sure
what
the
procedural
next
step
once
we
decide
on
on
a
on
a
point.
The
second
question
is.
J
I
J
So
so,
going
with
the
second
option
doesn't
work,
because
the
handling
of
the
aggregated
Dow
is
not
clearly
articulated
in
the
document.
So
there's
an
errata
right
now
or
we
can
say:
thou
shalt
handle
multiple
dowels,
okay,
and
so
we
have
an
interoperability
issue
in
the
meantime,
incrementing
the
DTS
n
is
a
good
thing.
We
don't
know
whether
the
downstream
of
sorry
is
that
on
the
sender
or
the
receiver.
So
that's
one
on
see
that
has
the
issue
no
see
doesn't
have
an
issue.
So
it's.
I
J
D
knows
whether
he
handles
aggregated
dolls
or
not,
he
doesn't
know
if
C
handles.
I
aggregated
dolls
right.
So
so
we
do
have
an
interrupt.
It's
not
a
unilateral
decision.
I
know
I
implement
this
code,
therefore
I'm,
okay,
so
it
because
he
doesn't
know
the
best
he
can
do
is
send
whether
he's
got
whether
he
implements
increments,
DT,
SN
or
just
resend.
The
data
he's
already
got.
The
most
pessimistic
thing
he
has
to
do
is
send
single
towels,
single
single
prefixes
per
doubt
right.
So,
if
introduced
I,
what
I'm
saying
Pascal.
F
Just
saying
we
seem
to
be
mixing
two
problems
right
and
just
clarify
that
point
one,
which
is
whether
the
answers
for
all
the
children
and
prone
to
is
whether
D
package
them
as
few
messages
with
a
lot
of
targets
in
them
or
as
one
big
message
with
all
the
targets
in
two
prompts
right.
So
my
argument
was
sending
the
test
and
recursively
down
is
quote-unquote
a
bad
idea
and
to
which
house
said
yes,
but
it's
very
easy
to
do
something
like
that.
But
it's
it's
about
age
and
and.
J
J
F
J
So,
okay,
so
that's
a
good,
a
good
observation
so
that
the
the
multiple
dowel
we
have
a
problem
with
interoperability.
We
have
this
problem
break
it's
nothing
to
do
with
packet.
Switching.
We
just
have
that
problem
of
there
and
it
causes
an
issue
right.
Okay,
it
sounds
like
everyone
would
like
to
use
to
do
solution.
Be
it's
a
better
answer.
We
just
don't
know
whether
we
can
or
not
that's
the
point,
yeah.
Okay.
J
So,
having
said
that,
I
want
to
ask
not
that
many
people
in
the
room
how
many
people
are
have
storing
mop,
are
using
it
you're
using
it
and
you're
having
difficulty
across
devices.
Okay
anyway,
I
didn't
see
behind
me.
Anyone
put
their
hand
up
okay,
so
anima
ACP
is
using
storing
right,
and
we
could
probably
at
this
point
you
know
say
whether,
which
one
mattered
or
something
like
that
it's
early
days
enough
to
do
that,
I
feel
that
we
shouldn't
spend
to.
This
was
my
third
point.
J
Right,
find
it
if
all
the
nodes
support
packing
dows
and
multiple
dowels
in
a
single
message,
and
if
that's
the
case,
we
can
say
it's
enabled,
because
everyone
supports
it
right
would
be
an
example
of
using
the
capability
mechanism
right
to
fix
this
I'm,
not
sure
that's
worthwhile,
for
the
simple
reason
that
I
think
we
would
be
way
better
off
to
have
non
storing
mode
of
projected
Dao
as
the
other
replacement
regardless.
Okay,
and
we
may
just
want
to
write
the
errata
that
says
you
should
support
multiple
dowels.
J
If
you
don't,
you
have
bug,
and
you
should
you,
you
should
resend
them
when
you
change
parents
from
your
storage
rather
than
incrementing
the
DTS
n
right.
Okay,
we
could
write
those
two
things
down
and
not
really
worry
about
the
fact
that
we
may
have
a
backward
compatibility
issue
with
something
yeah,
so
weak,
yeah
right,
that's
my
suggestion
and
then
we
can
take
that
slide
off
your
deck
and
I'm.
J
Sorry,
you
know
go
on
to
the
next
place
right,
so
I
don't
know
if
you
want
to
that's
my
suggestion
and
we
could
monitor
or
whatever
we
want
to
do.
I
I
So
again
we
coming
to
another
topic
which
we
have
discussed
before,
but
there
are
some
updates
on
this,
thus
handling,
so
the
Dow
AK
handling.
Again.
This
is
not
much
of
an
issue
when
you
handle
it
in
non
storing
mo
P,
because,
as
you
can
see
here
in
the
diagram
that
the
root
node
is
going
to
send
a
now
add
back
to
the
D.
Now
that
Dow
AK
is
extremely
important
for
an
in
or
any
implementation,
because
that
is
the
signaling
which
gives
you
an
indication
that
the
application
traffic
can
begin.
I
That
is
the
only
indication
available
on
a
big
network
that
the
application
traffic
can
be
instantiated.
So
this
is
an
extremely
important
point,
so
sadawa
needs
to
be
handled
in
that
sense,
so
in
case
of
non
story
mode,
it
is
really
really
simple,
because
the
root
node,
if
it
informs
about
the
dirac.
Everyone
is
clear
in
case
of
storing
MLP.
I
Now
again,
what
what
we
have
is
every
the
Dow
AK
is
sent
on
every
hop
by
hop
basis
now
know
being
here,
I
know
where,
when
the
node
B
sends
the
DAO
to
node
B,
it
sends
a
Dao
ack
instantaneously
and
relies
on
node
B
to
reliably
send
send
now
all
the
way
or
to
the
to
its
upstream
peer.
Now
there
are
two
problems
with
this:
this
this
implementing
or
this
this
methodology,
one
is
if
the
node
B
is
going
to
instantly
respond
back
with
the
DAO
AK.
I
Now
X
is
no
more
an
indication
of
starting
instantiating,
a
traffic
instituting
the
traffic's,
so
you
can't
rather
the
the
node
here
which
might
be
like
ten
hops
deep,
should
not
really
instantiate
its
application
traffic
just
because
it
got
a
dowel
bag
back
from
an
immediate
upstream
peer.
That
is
one
issue
here.
I
The
second
issue
is
node
B,
for
whatever
reason
it
sends
its
upwards,
that
our
reason
that
node
a
might
report
a
failure,
now
act
failure
with
a
negative
status,
and
since
the
dirac
has
already
been
reported
downstream,
the
negative
status
cannot
be
reported
any
any
further
below.
So
that
is
another
issue,
so
in
storing
mo
P
we
really
have
problem
how
to.
F
I
F
Yes,
two
things
here.
First,
are
just
sorry
to
be
back
with
time
for
me
to
dig
the
DRF
see
there
is
a
sentence
in
repo
which
says
the
repo
target
option
may
be
repeated
as
necessary
to
indicate
multiple
targets.
I
was
surprised
by
your
comment
that
you
could
not
agregate
a
package
and
because
we
did
it
so
I
always
thought
we
should
be
doing
it.
I
was
surprised
that
the
tech
was
gone
from
repo,
but
it's
there.
It's
expected
that
you
can
package
multiple
targets,
sorry
that
was
fun.
F
Ok,
so
this
part
that's
not
yeah
yeah
yeah,
it's
it's
its
implementation
back.
So
so
it's
good
to
have
an
implementation
recommendation.
I
quickly
loved
the
idea,
and
we
so
we
should
have
like
the
side
document
of
shows
or
maybe
have
a
section
for
recommendations
and
that
one
it's
actually
to
implement.
F
I
J
J
So
the
issue
is
that
if
D
start
sending
traffic
up
to
be
traffic
will
not
come
back
right,
so
we
can
send
traffic
and
they
can
go
out
to
the
network.
If
he's
sending
you
know
unidirectional
traffic
cool
traffic
coming
back
doesn't
work.
So
if
you
tried
to
do
TCP
or
co-op
okay,
for
instance,
which
has
a
response,
then
you're
not
gonna
get
an
answer.
Okay!
J
Well,
that
happens
right
and
it's
okay.
That
happens.
That's
that's
part
of
the
world
right.
Okay,
if
B
does
have
a
route,
let's
say
that's
a
bad
example,
but
let's
lose
the
case
of
F
to
D.
Okay
and
Dino's
in
the
traffic
is
to
see
because
it's
destined
to
see
well.
D
knows
how
to
get
there
see
the
fact
that
B
and
8
don't
know
how
to
get
to
F,
yet
not
so
important,
because
the
traffic
is
going
to
see.
J
I
So
yeah,
so
it's
just
that
in
this
case,
all
the
nodes
downstream
starts
started
and
starts
their
application
traffic,
which
simply
aggravates
the
situation.
So
if
there
are
a
lot
of
nodes
below
so
it
not
only
impacts
the
application
traffic,
but
also
the
other
control
traffic
in
the
network,
so
the
situation
simply
goes
bad.
The
network
convergence
time
is
impacted
terribly.
I
The
next
line
yeah
yeah,
so
there
is
another
Dalek
interpretation
which
is
essentially
whatever
wire
you're
mentioning.
So
usually
this
is
the
case,
no
D
sense
the
doubt
would
be
and
it
simply
acknowledges
back.
In
other
case,
the
Dow
is
send
all
the
way
back
to
the
root
and
then
root
acknowledges
it
back
well.
This
is
great
I
mean
it
definitely
is
a
clear
indication
of
what
once
the
note
gets,
the
Dow
acknowledgement.
F
Basically,
I'm
just
to
give
a
little
context
on
on
the
design,
so
the
official
design
is
first
one
you
acknowledge
per
round
by
acknowledging
you
say:
hey
I,
have
the
resources,
you
don't
say:
I
have
the
path
right,
that's
part,
but
we
say
other
resources,
meaning
that
you
take
over
responsibility
to
propagate
watching
to
the
root,
meaning
also
that
if
you
fail
to
do
that,
then
you
will
actually
have
to
key
on
your
relationship
with
that
child,
because
you
can't
serve
him
as
a
parent
Sue's.
That's
how
recursively
today
it
will
reach
the
children.
F
Oh,
it
would
be
disconnected
if
it's
disconnected
you
poison
the
graph
that
blah
blah.
So
it's
not
like
you're
stuck.
There
is
something
which
will
happen
even
shall
I
like
either
you're
detached
or
the
the
packet
will
find
its
way.
So
the
only
real
problem
is
you
don't
know
how
much
to
wait
till
you've
got
your
routing
back.
That's
so
you
don't
have
any
assurance
coming
from
the
routes.
Okay,
now
I
see
you
right,
but
but
it's
not
like
we're
honest,
uation.
Well,
the
pot
all
be
stuck
in
a
state
everybody's
waiting,
nothing
happening.
F
J
F
The
other
scheme
does
not
exist,
I
mean
it's
easy,
so
this
scheme
that
that
the
world
is
talking
about
is
a
trick
that
was
implemented
by
Kentucky
because
they
wanted
to
do
something
to
work,
fine
for
them
forth,
because
they
wanted
to
make
a
product
with
power
plugs,
and
so
it's
that's
why
you
find
it
in
Kentucky,
but
it's
not
in
the
spike.
The
spec
was
never
meant
to
do
this.
Second
thing:
it's
not
like
that!
I,
don't
think
there
is
ambiguity
there.
F
You
did
well
okay!
Well,
that
was
not
known
to
be
because
the
department,
that
is
what
kind
of
time
out
do
you
have
and
how
long,
because
if
the
network
is
very
slow,
I
mean
the
dowel
work.
If
you
just
sell
it,
it
fails
and
you
don't
get
an
acknowledgement
and
you're
there,
and
so
so
and
it
can
be
mounted
how
what
would
be
do
your
time
out?
When
do
you
start
thinking,
I
didn't
work.
I
should
try
to
fire
another
one.
F
It's
like
we're
less
naughty
hope
we
do
acknowledgement
at
every
hops,
because
otherwise
it
can
be
very,
very
slow.
So
so,
what's
your
timeout?
If
do
you
know
standard
owl
which
is
supposed
to
be
propagated
and
then
come
back
and
anything
can
happen
at
every
step.
The
timeout
is
kind
of
long
now
and
the
rich
rise
of
the
all
the
paths
it's
much
more
costly
to
recover.
Then
you
should
do
it
right,
fails.
I
I
Yes,
what
about?
What
is
the
desired
way,
so
status
is
equal
to
zero
endow
act
should
serve
as
an
indication
that
the
path
is
established.
That
I
think
is
definitely
needed
in
story
mode
of
operation
in
any
mode
of
operation.
For
that
matter,
and
it's
we
hoping
that
there
should
not
be
an
explicit
state
that
should
be
required
to
be
maintained
on
intermediate
6l
ours
for
thou,
a
king
I
mean
if,
at
the
same
time
we
can
handle
Dawa
aggregation.
That
will
be
really
great.
I
So
signaling
resource
constraint,
so
the
point
here
is
that
mentioned:
that
is
the
let's
say:
node
B
note:
D
has
gone
through
node,
B
and
node.
B
has
limited
resources
and
some
of
the
nodes
in
the
sub
doodads
have
already
attached
to
node
D,
but
now
it
cannot
take
any
nodes
further.
So
we
don't
have
any
metric
system
to
tell
to
percolate
this
information
down
down
down
words.
Now
we
have
enrollment
priority,
which
is
coming
up.
E
J
And
Michael
Yin,
so
so,
if
I
recall
that
was
there
was
your
draft
that
had
a
whole
bunch
of
unbalanced
trees.
Yes,
ok,
so
I
don't
understand
the
last
point
but
needs
to
be
used
with
multiple
hops.
Maybe
I,
don't
maybe
maybe
that's
wrong
or
maybe
I
just
don't
understand
the
point.
What
why
is
it
not
more
usable
at
multiple
hops,
meaning
it's
not
a
can't,
tell
see
that
it's
out
of
space
or
no.
If.
I
B,
let's
say,
for
example,
this
is
the
routing
distant.
That's
the
same
example.
If
we
take,
if
note
edge,
is
going
to
send
it
down,
it
goes
through
F,
it
goes
to
the
OD
and
then
finally,
it
ends
up
at
B,
but
B
responds
back
with
a
negative
status.
Now
only
only
H
has
to
be
impacted,
not
D
and
F.
So
it's
there
is
no
point
in
these
which.
J
I
see
I
see.
The
point
is
that
that
that
D
would
switch
when
B
said,
I
haven't
got
any
space
that
would
cause
D
to
switch
to
C.
Well,
C
might
not
have
space
for
the
whole
thing
either
right,
and
the
answer
is
that
somehow
the
the
tree
has
to
be
split
across
B
and
C,
but
both
F
is
in
the
way
we
can't
in
this
network.
F
has
to
pick
one
parent
or
the
other
right,
so
it
can't
we
don't.
Maybe
there
is
no
solution.
Yeah.
J
So
so
we
can't
solve
the
problem
that
B
and
C
don't
have
enough
space
for
D,
F,
G
and
H
there.
If
neither
have
enough
space,
it
doesn't
matter
where
D
goes
your
stock
and
and
we
can
look
down
it
as
God
and
see
that
the
problems
not
solvable.
So
the
fact
that
we
can't
signal
it's
not,
it
is
maybe
not
so
important.
J
We
would
have
to
do
something
else.
F
would
have
to
somehow
be
able
to
have
two
parents
D
and
E
to
split
if
that
were
possible
and
I
guess.
F
somehow
has
enough
space
for
gene
H,
but
the
combination
of
G,
H
and
F
is
too
much
for
and
E
is
too
much
for
the
nodes
above
right
because
there's
other
we
could
assume
there's
other
part
of
the
tree
as
well.
If.
I
F
F
So
so
you
can
push
back
children
to
the
left
to
the
right,
whatever
you
do
in
the
distributed
fashion
like
this,
since
it's
an
np-complete
problem
at
the
end
of
the
day,
you
can
only
push
back
and
push
back
and
push
back
as
you
put
more
and
more
complex
stuff
in
the
solution,
space
you
and
signalling
and
stuff.
You
push
back
the
primer
little
bit
and
the
more
little
bit
and
another
little
bit
and
it's
expressional
e,
difficult
and
and
you
win
an
expansion
elation
s
whatever.
F
Whatever
got
almost
all
enough,
storing
load
space,
whichever
your
position
in
the
graph
and
if
you
want
to
shift
functions,
if
you
have
enough
memory
for
100
nodes
on
the
on
the
box,
you
should
say:
don't
be
able
to
network
with
more
than
100
nodes
with
this
device
in
storage
node,
that's
kind
of
what
the
conclusion
was.
Whenever
we
said
anything
else
is
just
incremental
and
to
to
complicate
yet
bla,
bla
bla.
F
I
So,
just
to
come
to
that
point,
I
mean
it's
it's.
It
is
completely
I
mean
I,
understand
that
we
take
a
decision
about
making
I
mean
taking
this
decision
that
node
fde.
All
this
node
should
have
enough
resources
to
hold
the
company,
but
this
this
this
this,
this
design
assumption
never
came
into
6550
and
people
who
are
deploying.
It
eventually
realize
at
a
later
stage
that
this
this
this
could
be
a
very
big
issue,
and
that
is
what
you
know
is
one
of
the
point
that
should
be
mentioned
somewhere.
Oh
yeah,
sure.
F
But
it's
the
early
point.
Is
it's
not
even
a
developer
recommendation?
It's
not
an
L
week
type
of
thing.
It's
a
type
of
thing
right.
When
you
build
product,
when
you
use
a
product,
you
should
you
should
not
choose
throwing
mode
in
the
network
speaker
than
this
product
can
do
it
even
I
notice.
It's
not
even
implementation
recommendation
like
a
lot
of
things
right.
It's
deeper
yeah.
I
F
Nodes
is
a
few
ten
thousand.
Well
maybe
100k
memory
for
some
box.
100K
is
nothing
for
some
others.
It's
impossible.
I
mean
you
can
have
a
power
grid
meter
which
is
always
powered
and
that
cost
thousands
of
dollars
anyway.
So
I
think
one
leg
of
memory
in
this
niche
is
no
problem.
So
this
is
a
layer
3
protocol.
It
doesn't
really
make
assumptions
on
what
the
hardware
or
even,
if
it's
a
ratio
right.
We
said
animal
uses
repo
on
Ethernet,
so
so
then
no
assumption
at
layer
3.
It's
just
that.
F
I
I
Ok,
going
to
the
next
point
here,
the
the
transit
information
option,
the
design
goal
of
transit
information
option
was
to
make
sure
that
we
map
one
or
more
transit
information
to
one
or
more
targets
provide
a
syntax
for
aggregating
common
transit
information
option,
but
then
what
whatever?
What
I
can
see
and
yeah?
I
Maybe
I
would
like
to
be
corrected
from
the
working
group.
The
certain
elements
of
the
transit
information
option
for
exam-
the
external
is
not
with
regards
to
the
transit
right.
It's
it's!
It's
a
target
specific
flag.
Similarly,
the
path,
sequence
and
path
lifetime-
these
these
these
characteristics
are
not
related
to.
It
is
about
specific
target.
Now
what
happens
is
once
you
have
this
target
specific
options,
information
going
into
transit
information
option,
the
aggregation
options
are
limited.
Now
you
can't
have
multiple
targets
pointing
to
the
same
transit
information
option.
F
Copy
theory
is
Kelly
here
again.
This
is
only
meant
for
if
I
have
this
device-
and
it
has
multiple
either
of
tethered
devices
like
leaves-
which
would
be
the
route
to
it
or
multiple
addresses
into
this
device,
then
you
can
aggregate
it
this
way,
because
you
will
notice
them
as
a
package
every
time,
but
that's
the
only
cares
pretty
much,
but.
I
F
Income
factor
this
is
factorization
with
about
so
factorizing.
A
transit
for
multiple
targets
works
if
all
those
targets
are
kind
of
co-located
location,
meaning,
for
instance,
a
reaper
unaware
leaf,
which
would
not
even
understand
the
rule
draft.
So
so
it
could
not
be
mobile
at
all.
It
would
be
physically
attached
to
the
river
water,
just
like
a
card
or
USB,
cable
or
whatever
else,
and
so
you
could
have
a
number
of
addresses
that
are
really
packed
together
from
where
they
are
attached
in
the
ripple
Network,
in
which
case
you
could
factorize
them
with
a
Jo.
I
So
just
one
question
so
in
case
of
unaware
news:
we
still
have
the
path
lifetime,
which
has
been
informed,
which
may
be
informed
from
them
from
the
unaware
leave
through
Andy
right
I
mean
it's
still.
The
bot
lifetime
might
still
eventually
be
derived
from
nd,
then
propagated
by
our
Phillips.
Yes,
in
this
case,
it's
not
factorizable
yeah
and
inverse.
F
Means
T:
they
are
in
really
like
completely
slave.
Okay,
the
road
raft
with
eight
five,
four
five
decouples
the
two.
So
the
rules
can
move
from
one
parent
to
another
and
that's
when
it
needs
to
be
able
to
have
its
own
path
sequence,
because
the
the
path
will
be
injected
other
than
different
paths
and
they
will
compare
to
common
parent
somewhere
with
the
basic
ones.
But
if
the
device
itself
cannot
provide
the
path
sequence,
it
cannot
move
and
then,
if
you
cannot
move
that's
what
I
mean
by
tethered
physically
attached.
I
F
I
F
There
is
no
protocol
for
that
right
at
five.
Four
five
is
the
protocol
for
both
so
I
expect
that
device
that
ste
third
is
an
ipv6
host
with
zero
zero,
but
nothing
like
at
five
four
five.
So
since
we
have
nothing
else,
it
means
that
it's
it's
just
doing,
no
more.
Even
normal
and
gee
I
don't
know
right,
and
so
so
the
six,
the
repo
router,
would
learn
that
the
device
is
sure
through
snooping
and
EEO
those
weird
stuff,
and
then
it
would
translate
that
into
repo.
F
If
you'd
like,
we
don't
have
a
spec
for
that,
it's
doable
I
don't
want
to
write
a
spec
for
that,
because
we
have
a
try
for
five,
but
what
in
mint
either
I
mean
this
could
be
the
water
that
does
the
river
water
could
be
a
big
router.
It
could
add
multiple
Ethernet
interfaces
on
which
you
could
plug
physically
some
devices
or
it
could
be
USB
ports,
and
it
would
just
report
them
together.
So
it's
a
possibility.
That's
when
you
factorize.
I
All
right
so
so
alighting
options,
maybe
I,
won't
talk
much
about
this.
The
only
thing
is
we
have
the
draft
updated
for
this.
There
is
a
scenario
where
you
can't
so
RFC
6550
already
says
that
configuration
option
can
be
lighted,
but
there
is
a
scenario
where
it
cannot
be,
so
we
had
updated
the
observations
out,
but
I
am
NOT
going
to
talk
into
detail
about
this,
because
you
already
have
another
draft
about
this,
so
so
we'll
discuss
about
this
yeah.
I
The
whole
purpose
of
having
that
draft
the
next
part
and
I
really
would
like
to
have
a
clear
conclusion
on
this,
that,
regarding
the
RPL
persistent
state,
that
has
to
be
maintained
now
for
every
lollipop
counter
for
the
linear
part,
we
need
the
persistent
purse,
persistent
memory
position.
Storage.
If
you
do
not
have
a
persistent
storage,
it
is
possible
that
there
are
lot
of
scenarios
where
a
node
might
not
be
able
to
join
in
the
network
for
a
long
long
time.
Now.
I
The
question
is
whether
a
so
we
had
discussed
this
back
and
we
have
a
clear
understanding
that
if
the
node
reboots
during
the
linear
part,
there
is
no
way
to
know
that
it
has
rebooted
or
the
other
nodes.
It
is
not
possible
for
it
to
know
that
it
is
a
rebooted,
and
this
is
a
very
important
problem
to
handle,
because
nodes
might
get
rebooted
a
lot
of
times.
F
Yes,
two
things
the
first
one
is:
you
have
to
adjust
the
linear
path,
so
you
avoid
this
progress.
Much
as
you
can.
Second
is
when
you
are
in
the
linear
part.
Why
do
we
have
a
linear
path
right?
Why
don't?
We
just
have
a
point:
zero
many
protocols,
just
at
the
point,
zero,
and
this
gives
zero
when
you
keep
turning.
So
we
have
a
linear
path
because
we
may
lose
some
messages,
so
we
want
to
have
a
chance
to
increase
and
still
detect.
F
Now,
if
I
send
one
of
this
message
every
year,
then
I
will
be
all
my
life
in
in
your
part,
if
I
use
sixteen
right
like
sixteen
years,
so
obviously
you
need
to
achieve
the
linear
path
with
the
frequency
at
which
you
send
those
messages
right
and
if
you
thought,
if
you
send
them
frequently,
at
least
when
you
reboot
use,
you
just
want
to
send
a
batch
on
them
to
make
sure
people
heard
it.
Then
you
can
keep
it
to
sixteen
now,
if
you
decide
to
be
very
slow
shrink.
I
It,
let's
take
an
example
of
DTS
n
here.
You
know,
let's
take
an
example
of
DTS
and
now,
if
we
follow
the
logic
where
DTS
and
doesn't
get
incremented
when
the
parent
switches-
and
there
it's
hardly
possible.
So
in
this
case,
what
would
I
mean
6550
says
recommends
the
sequence
window
of
sixteen.
In
this
case,
it's
very
difficult
to
recover
from.
F
That
we
should
basically
update
that,
depending
on
which
field
it
is
as
I'm
writing
the
P
Dow.
We
have
several
seconds
contours
there.
I
give
a
different
value
for
starting
the
relief
up.
I,
say
basically,
just
I,
don't
remember
about
something
like
254,
so
it
has
very,
very
short
enough,
but
when
yes,
we
just
put
16
and
twas
mostly
thought
for
the
ions
and
now
well,
it's
too
much
unless
you
really
send
them
a
rapid.
It's
too
much
I.
F
Point
is
you
should
be
I
mean
that's
what
I
was
trying
to
say.
If
you
use
a
linear
part,
something
should
be.
Basically,
there
is
a
tension
between
you.
Don't
want
to
stay
in
that
part
on
time.
So,
if
you
have
a
big
number,
you
should
have
a
short
interval.
The
problem
we're
having
DTS
and
in
short
interval.
Well,
the
cool
thing
about
it
is
that
if
some
child
missed
that
your
granny,
the
DTS
end
because
reason
reason
reason
encouraging
DTS
and
each
type
in
a
short
interval,
they
will
end
up
seeing
it.
F
F
So
run
children.
You
will
get
that
repeated
16
times
so
basically
I
think
in.
We
should
be
same
case-by-case
and
and
talk
about
how
often
as
you're
in
the
linear
case,
you
need
something
more
often
and
how
long
should
that
in
our
case,
be
in
for
DTS,
ed,
probably
almost
zero
understand
with
UD
TSN
is
a
bad
idea.
Dinner
parties
should
not
be
that
long,
so
so
it's
60
inches
like
the
default
for
the
new
normal
mechanism,
but
we
should
have
stated
over
this
flow
for
this
message.
F
F
So
what
what
should
we
do?
So?
Basically,
we
have
a
lot
of
lollipop
counters
and
you
should
look
at
all
of
them,
one
by
one
and
provide
a
different
increment
value
and
and
say
words
which
are
related
with
a
frequency
you
send
and
so
that
it
belongs
to
very,
like
limited
life
span.
After
that,
you
in
the
circular
part
once
you're
in
the
circular
part,
you
can
space
what
you
sell
much
much
more,
but
something
around
the
fact
that
the
trickle
timer
should
stay,
for
instance,
for
those
which
are
below
trickle.
F
I
Okay,
so
this
is
an
interesting
topic,
because
this
talks
about
part
control,
bits
and
I'm,
not
sure
if
there
is
any
implementation
which
uses
this
stuff
well,
it
occupies
a
lot
of
real
estate
in
6550,
but
no
one
uses
it
as
far
as
I
know
as
of
now.
Well,
it's
a
it's
a
very
it.
No,
it's
a
very
flexible
logic
for
having
multiple
download
routes.
You
can
have
multiple
download
routes
with
preferred
routing
over
these
routes,
using
path,
control
bits
now.
I
The
point
that
I'm
trying
to
make
here
is
path
control,
bits,
it's
not
an
easy
mechanism
to
implement
at
all.
It
completely
directly
implements
the
memory
size
and
the
program
size
everything,
but
it
might
be
useful
in
the
future
for
sixty
show
draw,
is
what
I,
because
it
allows
you
to
have
multiple
down
routes
and
allows
you
to
load
balance
and
do
things
like
that.
Having
said
that,
I
don't
know
if
there
is
any
implementation
which
is
compliant
with
the
use
of
pod
control
bits.
I
In
fact,
almost
every
implementation
is
non-compliant
with
6550
today,
because
60
65
50
says
that
a
node
must
not
unicast
a
down
message
that
has
no
active
bits
set
in
the
path
control
field
set.
I,
don't
know
if
anyone
any
implementation
which
is
setting
this
this
the
bit
correctly
or
handling
this
bit
at
all.
So
so,
essentially,
there
is
no
implementation
which
implement
65
52
appropriately.
I
F
And
jojos
would
come
to
a
mic
here,
but
this
is.
This
is
a
basic
thing,
even
if
it's
complex
I
mean
the
results
kind
of
basics,
because
you
know
your
picture
illustrates
that
you
can
end
up
going
through
the
same
parts
exactly
so.
There
are
new
proposals
which
are
a
bit
more
intelligent
than
this
okay,
and
so,
if
nobody
implemented
it,
it
could
be
something
we
duplicate
when
we
revisit
repo,
for
instance,
the
fact
that
you
look
at
common
home
parents
allows
you
to
ensure
that
that
you
can
do
you
talk
about
roll
right.
F
Roll
has
a
lot
to
do
with
having
redundant
to
parents
having
the
same
two
children,
so
you
can
crisscross,
and
so
this
is
not
capable
of
doing
this,
because
you
have
no
visibility
of
who's.
Your
grandparent,
so
so
I
think
the
new
algorithm,
such
as
Georgia's
draft,
how
much
more
powerful
than
this.
So
if
we
end
up
thinking
a
four
or
six
days,
you
want
to
do
something
like
that
now,
I
would
tend
to
think
that
we
need
to
look
at
the
newer
approaches
and
duplicate
this,
but.
I
I
think
the
new
approach
that
we
are
talking
about
I
mean
I'm,
assuming
you're
talking
about
the
NS
exchange
ins,
in
which
case
it
has
a
different
set
of
assumptions.
The
design
consideration,
including
the
security
considerations,
the
security
aspects
overhearing,
requires
you
to
have
all
the
nodes
using
the
same
network
key
while
di
or
requires.
But
in
this
case,
what
we
are
saying
is
the
data
path
uses
the
shared
key,
which
is
a
big,
very
big
implication.
In
my
hearing.
F
So
you
get
this
integration
between
two
paths
that
go
from
the
root
to
the
child
that
that's
one
of
the
benefits
of
the
draft.
So
it's
not
you're,
not
building
like
say
two
paths
which
are
far
away
from
one
another
or
the
three
converge:
you're
building
two
paths
which
are
integrated
with
one
another.
So
so
you
have
redundancy
at
every
hop
of
the
way
as
much
as
you
can.
Even
if
you
don't
over
here
so
ever
giving
is
the
same,
you
can
do
because
you've
got
this
interrogation.
I
Okay,
coming
to
the
last
I'm,
almost
done
the
miscellaneous
section
mandating
we
support
for
reception
of
aggregated
targets,
I
think
we
should
clearly
mention
this
in
the
raft
somewhere
or
somewhere
in
the
the
aggregated
now
should
be
handed
by
the
receiving.
No
is
it
possible?
Okay,
the
DAO,
AK
I.
Don't
know
why
Dow
AK
is
mentioned
as
I
showed
in
6550.
I
It
should
actually
be
must
so
the
the
receiver
node
might
on
saying
the
K
flag
might
still
decide
not
to
send
now
acts.
That
is
what
it
means
in
case
of
multi
do
dragon
multi
instance:
I
think
I
mean
we
have.
We
still
have
a
work
in
progress,
but
there
are
certain
observations.
Basically
one
one
important
observation
that
we
have
is:
should
the
routing
table
the
reset,
when
switching
from
one
do
that
to
another,
in
the
same
instance,
not
talking
about
different
instance.
I
So
there
are
certain
implementation
practices
which
I
mean
if
the
vendors
could
be
confused
about
when
implementing
multiple
two
dies
and
multiple
instances,
transit
information
option
is
an
OP
is
an
optional
element,
but
I
don't
see
how
it
can
be
optional
at
all.
In
case
of
stowed
non
storing
mode
as
well
as
in
storing
mode,
it
has
to
be
mandatory,
but
the
DAF
says
that
it
may
be
same.
I
The
transit
information
option
may
be
sin
I'm,
just
wondering
whether
is
it
better
to
have
for
us
to
have
some
mineral
PL
or
something
like
that
without
part
control,
bets
anyways,
most
most
of
the
folks,
are
not
using
path.
Control
beds
seek
condense,
the
whole
archaeologic
into
very
specific.
You
know
memory
optimized,
and
maybe
we
have
clear
assumption
saying
that
it
works
only
for
such
kind
of
devices,
and
you
know
the
general
art
pill
can
work
for
the
bigger
set
of
devices,
including
at
on
it,
and
things
like
that
for
a
CPUs
case.
I
I
He
was
not
in
time
for
this
idea,
but
he's
trying
to
prototype
and
try
to
get
some
numbers
as
in
the
control
overhead
and
all
that
so
Dow
AK,
that
there
is
one
more
option
that
is
trying
to
check
whether
just
like
non
storing
mode.
If
we
have
a
Dow
AK
reported
back
from
the
root
to
the
know
what
it
means,
what
what
will
it
result
into
so.
F
That's
K
once
you
have
the
DCI
and
you
have
statuses,
be
a
120
H,
which
you
know
you
to
do
anything
you
like,
which
is
positive.
You
could
have
a
state
either
this
yellow
value
below
128,
which,
which
is
triggered
by
a
new
bit
in
the
Dow
and
when
it,
which
is
the
root
order,
which
is
this
bit
just
fly
fire
on
this
year
with
status.
Warm.
I
Too
I'm
just
worried
about
the
other
case
that
we
talked
about.
You
know
about
instantiating
the
app
graphic,
so
you
need
now
act.
Do
you
know
for
the
node
to
know
whether
it
can
instantiate
the
app
traffic?
So
if
this
here
is
the
case
where
you
can
clean
up
the
route,
if
there
is
a
negative
failure,
I'm.
F
Just
thinking
of
this
you
it's
it's
a
very
powerful
thing.
It's
the
only
message
we
have
going
down
the
geotag
a.m.
so
for
now
it
was
meant
to
destroy
something
but
us
as
a
sweetened
it.
Now
it
carries
a
repo
status
when
it
was
very
recent,
but
does
the
report
status
in
it
and
for
now
it's
always
considered
as
a
newer
to
sell
it
the
CEO,
but
since
we
can
report
a
status
below
128
and
it's
not
yet
son
to
the
Aussie
editor,
you
could
still
say
a
by
the
way.
F
There
is
no
snooze
statuses,
which
means
those
new
things
and
one
status
disuse
and
by
the
roots
down,
and
you
see
I'm
going
I
guess
it
could
just
mean
either.
Art
is
open
fire,
but
if
you
get
the
DCO
from
the
word,
it
means
you
can
reach
you.
Now,
it's
not
telling
you
you
can
go.
You
can
do
anything,
it
just
tells
you.
The
route
can
reach
you,
because
the
rest
of
the
internet
may
be
brokered.
We
don't
want
you
office,
co-op.
I
I
F
F
It's
just
that
I
think
if
message
just
from
one
proposed
since,
since
the
only
message
that
goes
down
I
think
it's
a
bit
of
a
waste
now
we
see
more
value
of
using
this
yellow
for
and
we
can
forget
what
this
show
is
about.
Yeah
I'll
find
a
new
acronym,
something
which
stays
this
year,
which
is
more
generally.
Thank
you.
Thank
you
for
bringing.
I
Up
our
big
data,
yeah,
okay,
so
there's
still
some
other
points
which
are
uncovered
which
are
which
possibly
might
get
updated
in
the
future
or
version
of
the
draft
it's
about
the
multi
sink
and
water
outer
practices.
If
an
implementation
has
to
use
multiple
things,
multiple
border
routers,
multiple
doodads,
what
are
the
best
implementation
practices
that
one
should
follow?
I
The
multicast
operations
there
a
lot
of
configuration
items
in
multicast
operations
which
have
to
be
synchronized
appropriately
to
all
the
nodes.
Secondly,
the
SM
RF
+
sm
RF.
While
it
has
been
widely
used
in
the
networks,
it
has
not
been
standardized
for
some
reason,
so
it
seems
I
don't
know
if
there
has
been
any
attempt
of
standardizing
such
things.
I
The
next
part
is
the
partial
dependency
on
nd.
So
now
today
we
have
prefix
information
option,
which
is
which
is
part
of
our
PL,
but
there
is
a.
There
is
a
whole
lot
of
other
information,
for
example
the
DNS
configuration
which
needs
to
be
given
from
the
border
out
all
the
way
to
the
to
the
node.
This
information
still
relies
on
the
on
the
RS
are
a
kind
of
signaling
mechanism,
so
boobage.
I
J
My
question
I
would
like
to
say
that
if
I
would
be
very
happy
to
put
context
stuff
into
di
O's
and
be
okay
without
nd
I
think
that's,
okay,
the
the
the
corollary
is
that
you
know
we
made
some
effort
to
or
they
come
the
converses
we've
made.
Some
effort
to
support
unaware
leaves
so
we
actually
need
nd
for
all
of
that,
so
I
would
say.
That's
a
you
know.
J
Kind
of
here
know
there
and-
and
maybe
the
answer
is
we
should
put
it
in
the
DI
O's
for
networks
that
have
no
on
note
no
leaves,
and
if
we
have
routers
that
expect
to
6l
RS
that
expect
to
have
leaves,
because
that's
the
kind
of
network
there
are,
then
they
should
do
it.
That
way,
that
I,
don't
I,
don't.
I
J
F
I
F
Information
that
may
be
useful.
In
my
original
mind,
the
IO
was
adaption
in
there.
It
was
a
radio
array
with
Dao
and
people
meaning
list.
We
agreed
to
split
that
and
I
think
the
most
very
acceptable
reason
for
splitting
was
that
some
information
that
you
usually
find
narrow
is
sent
at
a
different
pace
and
for
different
reasons,
then
find
some
information
that
you
find
that
Dao
right.
So
you
can
expect
that
in
the
network
like
this
GRA
will
be
a
lot
less
frequent,
then
at
Yahoo.
F
Now
the
prime
is,
if
there
are
a,
are
not
frequent.
It
might
be
that
the
first
thing
that
note
sees
is
a
di.
Oh
right
well
become
internet
yo-yo,
so
you
don't
want
to
wait
for
our
array
after
you
got
the
jelly,
which
could
be
a
long
time.
So
now
you
end
up
broadcasting
in
our
eyes,
and
so
so
you
wonder
if
you
could
put
enough
information
that
you
don't
need
the
array
for
the
ripple
stuff,
which
is
I,
guess
your
point,
then
you
could
start
working
right
away
and
actually
the
way
we
did
ripple.
F
We
re
expected
to
put
in
the
options.
Just
change
their
minimum,
which
is
just
avoid
alignment
and
and
so
express
things
in
in
bytes
and
supposed
to
act,
a
2
to
4,
4
bytes.
So,
just
generally,
you
need
at
the
size
and
look
at
the
PIO.
That's
exactly
what
it
is
right.
So
it's
really
meant
to
package
more
options.
Roy!
It's
just
that
we
didn't
do
it!
Ok,
so
fully
supportive.
If
you
start
thank.
B
You
very
much,
please
say
voluntary.
If
you
want
to
do
some
implementation
about
what
the
topics
not
I'm
saying
could
be
very
useful,
or
if
you
want
to
write,
you
can
start
with
a
lollipop
contest
like
a
document.
How
can
I
paint
each
counter
that
we
have
in
in
britain?
You
can
start
writing
voluntary
and
contact
us
around.
How
can
I,
how
can
you
help
us?
I
mean
there
is
a
lot
of
more
to
do
and
we
are
happy
to
have
new
comers.
B
F
So
yes,
this
is
about
the
projected
down
document,
so
we
did
a
lot
of
work
in
the
last
month
since
last
IETF
on
Pio
and
on
the
rule.
So
last
HF
we
decided
number
of
things
and
I
will
go
through
them.
We
decided
to
add
more
capabilities
to
the
PDR
drafts
and
so
I
try
to
propose
things
on
email,
didn't
get
a
lot
of
responses,
so
I
thought.
Let
me
write
down
and
publish
starting
points,
so
they
are
not
fully
agreed
by
the
groups.
F
That's
why
I'm
presenting
them
today
and
I
really
hope
expect
that
I'll
get
feedback
from
you
guys.
Is
it
taking
the
right
path?
But
since
we
added
all
these
new
things,
I
expect
churn
in
the
document
count
till
all
of
those
things
settle
to
something
that
the
group
likes
and
that's
why
December
is
not
a
feasible
date
because
of
all
these
additions
we
made.
F
So
I
I
have
two
slides
just
for
the
changes,
I
light
and
it's
just
the
eyes.
So
yes,
I
mean
you
see
that
the
number
has
increased
as
well.
So
the
first
thing
we
said
we
should
be
doing,
and
this
one
was
easy
is
the
option
is
big
because
you
have
a
sequence
of
ipv6
addresses.
So
what
we
decided
to
do
is
compress
it
and
very
easy,
because
it
one
through
I
3:8,
says:
ok,
take
a
reference
address
and
you
just
have
to
put
the
bytes
which
are
different
from
that
address
thing.
F
Is
you
have
the
same
compression
for
all
the
addresses
in
the
in
the
P
Dow?
So
either
you
you
put
more
segments
or
you
just
have
to
use
the
same
number
of
bytes
five
row
address,
because
that's
what
the
compression
value
tells
you
so
as
written
right
now,
a
P
Dow
is
a
segment
and
the
segment
has
the
same
compression
for
all
the
addresses.
So
if
one
could
be
compressed
to
two
bytes,
but
the
next
is
to
four
bytes,
you
learned
everybody
to
four
bytes,
which
is
pretty
much
the
way
we
did
in
six
five.
F
Second
thing:
it's
related
question
that
Michael
placed
in
the
mailing
list
is:
how
do
you
manage
this
thing?
Creating
it
is
good,
but
you
need
to
delete
it.
You
need
to
update
it
and
for
that
you
need
to
identify
it
in
the
first
place.
So
we
were
missing
when
identifier
for
each
path
that
you
are
creating,
so
we
can
say:
hey
we
are
going
to
destroy
it
and
we
needed
the
usual
sequence
like
the
path
lifetime
and
the
path
sequence.
F
This
exact
same
thing
were
needed,
but
first
we
needed
a
path
ID,
and
if
you
read
the
60s
architecture,
you
will
find
that
60s
already
describes
that
actually,
a
path
that
we
built
in
60-ish
is
called
the
track.
Because
path
is
so
generic
that
you
never
know
what
you're
talking
about.
Is
it
a
multi
path,
a
single
path,
so
track
is
more
specific
to
what
we
are
talking
about.
It's
kind
of
the
directly
sik
graph
to
this
nation,
so
you
can
have
some
degree
of
redundancy
to
get
to
that
destination.
F
In
repo
add
a
model
for
index
this
thing,
which
is
the
local
instance
ID,
so
the
local
instance
ID,
is
a
number
which
is
taken
from
a
namespace
which
belongs
to
the
destination
or
the
source
of
the
packet.
Since
we
are
talking
about
a
traffic
engineer
path
to
a
certain
destination,
the
packets
always
go
to
that
destination,
so
it's
kind
of
logical
to
take
the
number
from
the
name
space
of
the
destination.
So
it's
just
like
your
use
today,
all
right,
you
send
a
packet
towards
the
route.
F
Well,
if
it's
a
global,
is
you
just
look
at
the
instance
ID?
But
if
it
was
a
local
instance,
the
route
the
packet
would
be
flagged
so
that
the
destination
plus
the
local
instance
ID
in
the
RPI
I,
tell
you
what
geotag?
That
is,
which
tells
you
what
routing
table
you
need
to
look
at
to
route
this
packet.
So
we
just
sixty
said
generalize
that
for
trucks-
and
that's
that's
now
replicated
in
this
draft,
so
we're
consistent
with
this
draft.
F
F
So
the
world
could
be
updated.
Basically,
now
you
have
a
track
ID
the
route
will
send
a
new
Pio
message
to
update
it
with
the
same
track:
ID
and
a
new
path,
sequence,
cutoff
so,
and
the
new
path
sequence
kind
of
remove
the
old
one
life
time
of
zero.
We
move
what
you
had.
If
you
have
the
same
sequence
number
it's
a
rich,
why
you
should
forward
it
the
way
it
is,
but
not
you
don't
have
to
change
anything
because
the
same
as
before,
so
maybe
it
was
lost
a
bit
later.
So
that's
installing.
F
F
We're
also
missing
a
segment
ID,
because
a
track
is
not
a
serial
path.
Why?
Because,
if
you
have
a
control,
a
controller
are,
they
are
outside
of
your
say,
say
typical
ripple
Network
is
a
fan
right
fill
our
network
smoked
with
you.
Do
that
I
do
that
the
control
is
far
away.
The
network
is
slow.
So
if
you
build
on
your
serial
path-
and
there
is
some
Prime
on
that
path-
you
need
to
go
back
to
controller
to
fix
it.
F
You
don't
have
the
time,
so
you
really
need
the
path
to
be
very
resilient
for
the
possessor
wrong
problem.
If
the
path
is
very
resilient,
it
means
it
has
multiple
segments.
It
has
one
way
north
one
myself
and
maybe
replication
elimination
between
those
and
maybe
several
even
north-south
middle
blah,
blah
blah
and
they
hold
the
same
track.
F
There
are
different
segments
on
the
same
track
when
you
put
all
the
segment's
together,
you
get
the
vision
of
the
track,
so
we
needed
not
only
a
track
ID
but
also
a
segment
ID,
so
you
could
modify
a
segment
without
repainting
the
whole
track,
so
the
red
best
version
that
was
missing
in
the
previous
version.
The
latest
version
that
I
published
this
morning
I
added
back
the
segment
ideas
which
were
missing.
F
F
Another
thing
we
said
we
would
be
doing
is
add
sibling
information
because
initially,
when
you
register
off,
we
said,
oh,
there
will
be
another
protocol
to
expose
the
the
logical
structure
to
the
route
and
that's
not
feed
out,
but
nothing
came
up
and
then
we
end
up
with
the
wrong
projection
mechanism,
but
no
way
to
expose
the
topology
to
the
route.
We
said
a
in
story
mode.
The
route
already
knows
the
dirt
dag
right.
So
the
only
thing
that's
real
missing
is
a
few
siblings
and
then
the
root
is
enough.
I
Other
father,
strangely
enough,
I
discovered
that
part
control
birds
can
actually
do
that.
So
in
in
transit
information
option,
it
is
possible
using
transmitters.
It
is
possible
for
one
target
to
send
multiple
transit
information
options
with
different
parents
to
the
route
and
I
discovered
this
very
recently
that
it
is
possible
for
the
node
to
sing
tell
the
route
that
it
is
attached
to
all
the
server
right,
but
they.
F
Are
parents
so
so
what?
And
so
we
needed
to
be
able
to
see,
know
something
which
was
not
apparent
and
the
poem
with
a
norm?
Parent
is
well,
you
want
it
to
be
able
to
signal
like
the
sub-matrix
and
so
so.
Basically,
this
the
sibling
information
can
have
complimentary
containers,
and
that
was
actually
you
Rabia,
yes,
Westers
I,
don't
remember
somebody
Estes
was
a
great
question
so
now,
since
it
was
this
case
on
the
mailing
list,
this
one-
it's
not
only
sibling,
but
also
some
metrics
about
you
know
between
you
and
the
sibling
very
important.
F
The
sibling
selection
is
out
of
scope.
This
is
just
signaling.
You
expressed
a
sibling.
The
route
manages
with
that
now
say:
I'm
in
a
very
dense
Network,
like
thousands
of
nodes
around
me,
I
won't
be
able
to
report
one
thousand
seedlings
to
the
route
worse
than
that,
it's
not
only
knowing
that
you
are
sibling,
it's
also
maintaining
bi-directional
connectivity
with
them,
or
at
least
having
some
packets
between
them.
F
So
you
know
the
signal
strength,
and
so
you
realize
that
something
will
have
to
be
done
to
select
siblings
in
all
directions,
so
I
can
get
anywhere
as
opposed
to
taking,
for
instance,
the
three
best
seed
neighbors,
which
are
all
on
my
right.
So
if
I
use
them,
I
can
never
go
left
so
that
there
is
a
prime
of
sibling
selection
which
is
not
addressed
by
this
draft
I.
F
Don't
I,
don't
see
that
we
should
address
it
because
some
networks,
they
don't
need
it
because
you
have
few
siblings
and
you
can
report
them
or
some
other
networks
have
different
sort
of
siblings
and
I
see
the
sibling
selection
as
an
objective
function.
Problem
just
like
gof
is
separated
from
a
pole.
The
sibling
selection
could
be
various
intelligences
based
on
various
networks
and
I
expect
that
we
would
see
documents
coming
to
them
to
row
just
about
objective
function,
to
select
siblings.
F
The
result
of
that
objective
function
will
be
I
said,
like
I,
said,
like
those
three
siblings
and
now
report
them
to
this
thing.
So
this
is
more
like
repo.
We
don't
give
the
objective
function.
We
say
there
is
this
plug-in,
and
now
we
expect
some
different
plugins
for
different
situations
to
come
up.
F
Ok,
so
that's
so
my
lights.
Okay,
yeah
disorder
request,
at
least
by
two
people,
I
see
that
the
Chinese
is
on
on
the
mythic.
Oh,
and
there
was
another
person
for
it
may
be
ready
again.
I,
don't
remember
so
that
was
oh
and
that
was
what's
your
name
well,
I'm,
sorry
well,
I,
remember
whom,
but
not
his
name
anyway.
So
so
number
of
people
asked
for
the
capability
for
a
node
in
the
repo
network
to
say
a
mr.
root.
I
want
a
path
to
this
other
guy.
F
So
now
we
have
a
PD
request
and
the
people
request,
acknowledgement
which
is
a
unique
aspect,
get
to
the
root
and
back
and
that's
the
response,
tells
you
I'm.
Mr.
wood
and
I
have
taken
you
your
problem
into
account,
and
then
you
expect
some
Pio
to
come
back
later,
but
at
least
an
acknowledgement
from
the
root
that
you
understand
that
you
asked
him
to
do
this,
and
the
root
will
come
back
with
a
track
ID
in
the
response.
F
So
when
you
see
that
the
the
PD
our
coming
with
this
track,
ID,
no,
it's
the
one
that
is
actually
built
for
your
request
and
same
thing.
You
have
lifetime,
so
you
can
also
ask
the
root
to
answer
it
by
putting
a
lifetime
of
zero
blah
blah
blah.
So
it's
a
simple
mechanism.
This
one
does
not
I
hope
this
did
not
need
a
lot
of
in
the
future
and
it's
missing
and
I
don't
know
if
you
wanted
to
define
it
here.
F
We
might
have
something
like
a
metric
container,
but
the
metrics
would
be
about
the
quality
of
the
path
that
you
want
to
build
like
if
I
want
to
say,
I
want
a
pass
with
three
nines,
which
means
that
the
route
will
have
to
compute
or
the
controller
somewhere.
I
will
have
to
compute,
maybe
more
paths,
because
with
two
paths,
I
can
only
have
two
nights,
so
you
know
being
able
to
express
the
quality
of
the
path
that
you
want
to
be
able
to
do
maximum
latency
or
whatever
else
that
needs
to
be
expressed
as
metrics.
F
F
If
you
want
to
be
able
to
indicate
that
that
we
need
to
improve
this
section
by
saying
here
is
container
of
some
form
with
metrics
which
indicate
the
quality
of
the
path
you
want
to
be
able
up
to
you
up
to
the
meaningless
to
say
we
needed
the
a
lot
of
work
on
the
iana,
etc.
We
still
need-
and
that's
one
of
the
issues
we
have
a
better
security
section
so
going
to
a
little
bit
more
details.
F
So
the
new
fields
are
here,
oh
by
the
way,
just
to
differentiate
what
we
do
for
from
the
path,
for
instance,
in
the
transit
information,
I
rename
the
same
track
and
segment,
because
you
have
the
two
components
right:
the
track
is
a
collection
of
segments.
Maybe
a
truck
is
just
one
segment
if
it's
a
serial
path,
but
the
truck
can
also
be
a
collection
of
segments.
F
So,
instead
of
cutting
something
paths
and
I
don't
know
if
it
would
be
this
thing
or
that
saying
that
I
should
call
paths,
I
put
a
different
name
track,
ID
at
segment
ID
the
track
ID
is
a
local
repo
instance.
You
will
find
it
in
the
packets
associated
with
the
destination,
and
so
in
the
PD.
All
you've
got
a
target
for
the
destination,
so
target
plus
truck
ID
gives
you
the
destination
that
I
get
with
the
FBI.
F
F
Surely,
to
build
it
right
so
because
you
might
have
a
VR
for
one
segment
and
then
as
Elvia
for
segment
and
lower
segment.
They
are
in
the
same
track,
but
when
you
forward
you
don't
negate
which
segment?
That's,
maybe
you
will
tell
me
you
need
to
and
then
with
James
thanks,
but
the
way
the
draft
is
written
right
now
we
say
put
it
on
this
track
and
it's
up
to
wow.
F
Wednesday
morning,
please
attend
Rome
Wednesday
morning,
it's
up
to
Rahl
to
make
the
decision
of
amusing
the
upper
segment,
the
lower
segment
I'm,
using
both
I'm
doing
application,
English
evaluating
constructive
interference
and
whatever
else,
and
so
for
each.
You
need
a
sequence
right,
so
both
track
sequence,
which
tell
you
when
you
flush
the
track.
You
have
and
build
a
new
one
and
something
for
each
segment
individually
within
the
same
track.
So
if
you
see
the
same
track
sequence
and
the
same
fragments
in
same
fragments
segment
sequence,
what
does
that
mean?
F
Well,
it
probably
means
that
you
have
a
storing
mode,
P
Dow,
which
was
sent
by
the
root
and
they
would
never
get
never
got
an
acknowledgement.
So
it's
just.
We
trying
it
so
the
idea
is
when
you
get
it,
if
you
already
got
it,
it
has
to
be
the
same.
So
if
you
took
actions
like
establishing
writing
state,
it
should
still
be
valid,
maybe
refresh
the
lifetime
if
it's
really
a
shot
lifetime,
but
it's
a
retry.
F
So
here
is
the
format
of
the
packet
by
the
words
I
indicated
so
d/c
new
text,
so
remember
a
local
WIPA
instance
iodine
in
repo
is
built
like
that.
The
first
bit
tells
you
if
it's
a
global
instance,
all
the
constants,
so
instant
zero
is
a
global
instance
instance.
128
is
a
local
instance
this
bit.
The
second
bit
in
the
in
the
API
is
for
the
data
path,
not
for
the
signaling.
It
tells
you
for
local
instance.
So
if
this
bit
says
lookö,
this
bit
tells
you
if
it's
local
to
the
source
or
the
destination.
F
So
if
you
use
a
deal
bag
in
both
direction,
the
the
address
which
is
the
namespace
for
that
instance,
local
instance-
is
on
the
way
up
the
destination
and
the
way
down
it's
the
source
right,
so
so
repo
as
this
notion
of
this
bit,
which
tells
you,
if
it's
a
source
out
of
this
nation,
which
is
a
namespace
associated
to
this
instance.
In
our
case,
the
truck
is
always
going
somewhere
else.
Never
coming
back
is
that
the
traffic
is
directional,
which
means
it's
always
the
destination.
F
F
So
sibling
information,
what
you
have
so
far
is
a
step
of
rank,
so
at
least
we
give
the
the
RAC
Express
by
the
objective
function
as
if
it
was
a
parent
for
which
you
compute
the
step
of
Frank
use
same
logic,
and
you
look
at
the
signal
with
this.
You
can
at
least
signal
the
step
of
Frank
to
the
root,
so
at
least
the
root
can
measure
everything
in
terms
of
Frank.
What
step
of
Frank?
F
Fidel
requests
so
asked
for
acknowledgment,
and
this
is
something
to
be
discussed,
but,
as
I
said,
we
are
missing
metrics
to
express
the
kind
of
quality
we
want
for.
The
builder
for
the
past
would
be
out
so
at
least
I
put
this
little
bit
to
remember
that
to
say,
I
want
a
redundant
path
or
not
like,
but
maybe
it's
not
the
right
for
signaling
it.
Maybe
we
need
options
or
something
to
really
say
what
kind
of
characteristics
you
want
for
this
path
that
you
want
to
create.
F
So
this
is
a
packet
that
the
node
sends
to
the
root
and
the
root
answers
with
that,
and
so
you
get
a
lifetime
which
is
actually
accepted
by
the
root.
You've
got
an
eco
of
the
sequence,
numbers
and
thing
for
diode
awak,
so
you
can
correlate
the
request
on
the
response
and
get
the
status
similar
to
the
repo
status
and
Odysseus
data.
So
tells
you
how
good
you
know
this
world
pending
issues,
we've
got
nice.
F
So
that's
part
of
the
game
and
oh
that's.
The
configuration
parameter
I
mentioned
earlier
so
can
do
want
to
push
some
parameters
on
the
quality
of
the
past
that
we
wanted
the
route
to
be
able.
So
this
is
basically
most
needy
thanks.
So
first
step
is
okay.
There
is
a
lot
of
change,
lead
document
right.
There's
everything
I
told
you
and
more
details.
So
it's
really
good.
Now
that
we
set
it
all
and
say:
okay,
let's
look
at
the
whole
this
proposal,
so
everything
we
said
with
the
dress.
F
F
Is
it
a
way
to
go,
and
then
we
can
progress?
This
thing
segment
information
was
published
this
morning,
so
it's
done
live
on
formats
to
optimize,
that's
something
we
could
look
at
you
now
or
in
the
future.
Is
it
because,
right
now
there
are
a
lot
of
reserved
space
you're
on
there,
some
of
the
format's
that
you've
seen
of
two
bytes
reserved
because
I
don't
know
what
we'll
discuss
now,
but
that
before
we
ship,
we
might
say
orders
to
be.
It's
us
two
bytes
won't
have
to
be
resolved.
We
can
just
not
send
them.
F
B
I
All
right,
so
there
has
been
doing
trims,
specially
input
to
discuss
the
capabilities
and
the
mopix
and
I'll
just
quickly
give
you
an
overview
of
what
capabilities
is
most
of
you
already
know.
It
allows
our
till
nodes
to
signal
their
capabilities.
Now,
one
of
the
things
that
we
discussed
last
time
was:
we
need
to
clearly
differentiate
between
what
we
already
have
mode
of
operation
and
the
DI
of
configuration
option.
How
is
it
different?
How
is
capability
is
going
to
be
different
now?
I
One
clear
difference
is
that
a
movie
in
the
configuration
option
is
strictly
route
control
in
case
of
capabilities.
This
could
be
sent
by
any
of
the
nodes,
so
that
is
one
of
the
most
most
important
difference.
You
know
that
should
not,
and
capabilities
can
emit
more
richer
information,
because
it
is
sort
of
a
format,
options,
TLB
format
by
itself,
then,
and
capabilities
may
be
more
dynamic
than
mo
PM
configuration.
It
might
change
over
a
period
of
time
and
design
goals.
Any
node
could
generate
the
capability.
I
An
option
could
be
sent
in
any
message,
so
the
K
+
1,
the
important
thing
is
it
should
be
able
to
it.
Should
it
should
be
possible
for
capabilities
to
work
with
the
existing
mode
of
operation?
Well,
this
is
this:
is
non-trivial
and
we'll
discuss
in
the
subsequent
slides
how
this
can
be
handled
and
capabilities
could
be
explicitly
query,
and
now.
This
is
something
that
we
have
not
handled.
I
How
it
looks
right
now.
This
is
how
it
looks
so
earlier.
If
you,
if
you
have
read
in
your
earlier
version
of
the
draft,
we
had
simply
a
combination
of
bits,
but
we
have
moved
away
from
the
combination
of
birds
to
individual
clearly
for
each
of
this
capability,
because
we
might
need
more
information
in
context
to
every
capability
that
to
be
sent.
So
there
is
one
more.
There
are
few
more
interesting
bits
in
context
to
each
of
the
capability,
for
example
the
che
bit,
which
is
whether
the
node
can
join
only
as
a
leaf
node.
I
If
the
capability
is
not
understood
or
and
their
capability
has
to
be
copied
downstream
and
then
the
third
bit
is
whether
the
capability
information
is
present
now
each
capability
might
be
optionally
associated
with
additional
set
of
information.
If
the
I
bit
has
said
that
information
is
present
and
if
it
is
present,
distr
has
to
be
in
this
format.
So
during
the
interim,
we
discussed
that
we
might
need
this
length
so
that
we
can
skip
the
whole
option
all
together,
so
we
avoided
this
length
in
the
draft
as
well.
I
Yeah
final
cap
unaware
nodes.
We
try
to.
We
try
to
reason
if
it
is
possible
to
handle
the
cap
and
aware
nodes
and
what
we
discussed
in
the
interim
was.
It
is
better
to
use
capabilities
only
with
them
new
mode
of
operation,
so
the
new
mopix
allows
you
to
send
the
old
mo
P
in
the
new
option.
So
if
the
new
option
is
used,
you
can
as
well
send
caps
with
that.
So
caps
is
only
supported
only
if
the
mo
pecks
is
used.
I
Now
this
is
something
which
working
over
it
needs
working
group
consideration
as
well,
so
the
capability
is
signaling.
The
capabilities
I
kept
one
thing
that
we
discussed
was
the
capabilities
could
be
carried
in
existing
signaling
messages,
the
d
io
douse,
or
it
could
be
explicitly
correct
signal
through
some
additional,
my
messaging
new
messaging.
This
new
messaging
is
yet
to
be
discussed.
I
This
would
be
the
right
time
if
working
group
has
a
specific
point
to
say
here
whether
it
is
not
good
to
have
a
new
message
all
together
for
querying
the
capability,
but
we
need
this
querying
mechanism
or
we
could
simply
rely
on
this
unicast.
This
di
o
to
query
this
such
scan
of
information,
which
I
feel,
is
not
a
good
idea,
because
the
idea
has
a
lot
of
other
information
apart
from
the
capabilities
which
has
to
be
given
as
well.
I
I
Knoppix,
so
one
of
the
things
that
has
changed
in
context
to
the
Malbec
says
is
the
final
mopix
value
calculation.
We
don't
use
the
complex,
it
was
not
complex,
but
it
was
still
the
multi-level
calculation
where
you
use
the
base
mo
P
and
then
you
mo
pick
add
them
together,
and
then
you
get
a
final
mo
P.
We
don't
do
that
anymore.
We!
F
F
The
configuration
is
when
the
route
imposes,
something
like
you
program,
the
CLI
in
a
router
or
something.
So
in
this
case
the
configuration
will
say
and
user
free
polyfill
says
it
for
for
its
own
use,
go
ahead
and
use
this
tank
set
it
up,
and
if
you
understand
it,
but
you
can't
set
it
up,
you
should
operate
as
a
leaf
or
something
so
you
mentioned
P
routes
and
six
leverage,
but
there's
also
the
use
of
the
new
FBI
in
user
free
/
info
right.
F
So
basically,
you
have
a
bit
in
the
configuration
option
which
size
use
it,
and
it
would
be
good
that
the
route
sets
this
when
it
knows
that
the
node
support
that's
when
the
capability
is
coming
to
play
right.
So
if
you
have
another
way
to
know
that
your
devices
support
this
right
like
functions
management,
you
don't
need
the
capability.
You
still
need
the
configuration
right.
F
You
use
the
management
to
know
that
the
node
support
and
once
you're
happy
with
that,
you
turn
the
switch
on
or
you
just
know
your
device
and
once
you've
uninstalled
the
last
one
which
did
not
support
or
we
flushed
it.
You
know
you
have
refreshed
everyone
that
you
can
use
the
bit
in
the
configuration
to
turn
the
function
on.
That's
true
for
the
new
API,
that's
true
for
using
projected
rods
well,
which
is
not
a
bit
special,
but
that's
true
fight,
one
three,
eight!
F
So
that's
why
we
have
those
bits
in
the
configuration
and
that's
one
thing
on
the
left.
The
thing
on
the
right
is:
how
does
the
would
know
and
that's
when
the
capability
come
into
play
with
this
draft-
will
be
capable
to
pull
from
the
network
what
the
network
device
do.
So
we
don't
need
to
configure
it.
The
management
thank
the
repo
manages
itself,
so
it's
completely
useful.
It
separates
it
from
the
configuration
we
already
did,
use
a
free
pollen
fo
with
a
bit
to
configure.
We
need
a
23-8
for
the
same
thing.
F
Pierot
is
more
like
the
wood
will
decide
for
just
this
rod.
So
maybe
it's
not
really.
You
don't
already
have
a
configuration
for
it.
The
P
Dow
is
the
configuration
file,
but
for
FC,
eight
one,
three,
eight
and
for
user
freepo.
It's
completely
separate
whether
you
can
pull
the
capability
and
when,
when
you
decide
to
configure
it
right,
so
the
you
do
not
drive
tonight
once
way.
It
does
not
need
this
one.
F
You
see
because
it
just
says
configure
it,
but
when
we
have
this
true
for
user
free,
/
info
true
for
for
eight
one
three,
eight,
then
repo
will
have
a
total
solution
to
manage
and
configure
okay.
So
that's
how
its
decouple.
So
there
is
no
the
idea
that
there
is
a
dependency
on
this
draft
in
use
of
a
well
cell.
I
F
F
J
So
Michael,
so
you
just
don't
really
really
don't
want
to
wait
for
us
to
do
capabilities
in
any
way
to
do
eight,
one
three,
eight,
and
what's
your
what
you're
saying
you
wanted
you
want
to
do
it.
We
could
do
eight
one.
Three,
eight
with
capabilities
is
the
point
right,
and
but
you
don't
want
to
do
that,
you
want
to
ship
it
the
way
that
it
is
I
want.
J
I
I
Clear
so
two
questions
mopix
currently
has
24
bits
now.
Do
we
really
want
to
use
those
many
bits
because,
as
I
can
see,
our
dependency
on
mo
PS
will
reduce
with
capabilities,
so
we
can
do
with
less
number
of
bits
for
more
packs,
but
again
I
don't
see
much
of
a
value
there,
because
once
we
start
alighting
such
fields,
then
it
might
not
even
be
important
to
worry
about
the
field
size
then,
but
I
would
like
to
have
any
opinion
here.
I
Second
question
is
whether
we
should
split
the
two
topics:
the
mopix
and
cap.
The
reason
being
looks
like
capabilities
might
take
some
time
to
mature
and
mopix
is
really
really
a
simple
thing
to
do.
We
might
as
well
ship
it
separately
and
there
is
a
clear
difference.
I
mean
there
is
no
overlap
of
information
between
I
mean
there
is
no
overlap
of
thing
between
mopix
and
capabilities,
so
any
any
any
views
there.
I.
F
Mean
I
would
I
would
really
push
that
question
12-hour,
because
basically
it
makes
more
sense
for
the
engineering
for
us
to
split.
So
we
have
one
document
for
one
thing
and
the
other
document
for
the
other
thing.
But
now
that
means
both
will
be
still
not
track.
Documents
to
go
through
the
iesg,
a
DHT
might
not
like
it
so
so
rope
X
would
be
a
very
simple
earth
and
we
could
flush
it.
Throw.
K
This
matter,
I
would
prefer
one
document
just
because
it's
easier
I
process
only
one
time
and
that's
it,
but
I
know
there
are
other
non
ihe
the
event,
a
middle
considerations
around
having
multiple
or
RCS
I've
heard
from
other
working
groups,
for
example
the
need
to
do
RF
piece
and
point
it
one
thing
or
another.
So
if
we
need
troops
through
the
functionality,
that's
perfectly
fine,
so
whatever
you
guys
want
to
do
work
for
me.
Okay,
thank.
A
J
Good
so
I
can
I
would
like
to
have
two
documents
I
asked
for
that
before
and
yours
I
so
I
think
your
diagram
is
incorrect
and
I
just
looked
at
the
document
and
it's
correct
for
the
document.
I
just
got
sixty
five
fifty
and
it
doesn't
match
6550.
So
if
I'm
not
mistaken,
this
is
a
a
our
pl
option,
lives
in
diyos
or
daos,
and
they
actually
are
in
the
format
of
a
type
length
and
stuff.
Oh
yeah,
yeah
yeah
right
that.
J
J
Part
is
that
that
means
that
we
can
decide
how
big
that
value
is,
and
it
can
be
as
long
or
as
short
as
we
like,
as
we
have
a
length
yes
like
64
bits
of
of
of
mop.
X,
probably
not
probably,
should
have
a
maximum,
but
the
point
is
that
if
we
have
only
one
byte
to
send-
and
we
say
it's
an
integer
and
it
fits
in
one
bite-
we
only
have
to
transmit
one
byte
at
the
point
where
we
have
to
find
256
mop,
mop
X
values.
We
can
add
a
second
bite.
J
We
just
need
to
make
sure
up
front
that
we
tell
people
please
pay
attention
to
the
length,
but
lengths
larger,
then
I
think
the
number
would
be
six,
because
that
we
four
bytes
are
probably
unreasonable.
So
you
don't
need
to
do
64-bit,
arithmetic
on
this
or
whatever,
okay,
so
I'm
pretty
sure
we
can
do
that
right.
You'd,
agree
to
Pascal
right.
K
F
Yes,
interesting
enough,
we
are
changing
the
first
bits
in
the
DI
you
pretty
much
with
a
new
map
x,
value
of
seven,
and
we
don't
have
a
version
for
people
all
right.
So
it
seems
to
me
that
somehow
we
are
packaging,
a
number
of
features
together
or
like
P
Dow,
which
I'll
really
rule
bring
all
these
capabilities.
We
are
really
making
a
version
to
a
free
poll
and
seems
to
me,
which
is
not
because
we
can
do.
We
can
formally
define
that
yeah
all
right.
That
really
needs
to
say
a
multi-core.
F
L
F
I
B
F
F
I
really
loved
his
arguments.
I
think
he
made
me
make
perfect
sense
what
he
said
and
I
kind
of
boat
everything.
So
no
it's
not
a
new
mode.
You
will.
You
will
have
a
main
gear
that
which
will
operate
in
the
normal
mode.
You
will
have
the
capability
to
project
projected
rods
and
you
will
install
them,
but
the
peel
the
peel
out
itself.
It
is
some
type
I
mean.
So
it's
not
like
a
mop.
F
The
mop
is
the
global
deal
dag
and
since
we
don't
use
the
yo-yo
to
build
the
P
Dow,
what
if
you
don't
have
anything
to
express
a
separate
mop,
but
the
big
dag
is
like
Bobby
said
it's.
What
it
is.
We
don't
need
a
new
mop
for
that.
So
if
there
is
still
text
simply
our
draft
I
need
to
remove
it.
I
think
there's
no,
but
if
there
is
you
to
clean
it
up,
I
mean
if
people
agree
right.
I
do
agree
with
Robby.