►
From YouTube: IETF108 BABEL 20200727 1300
Description
BABEL session at IETF108
2020/07/27 1300
A
A
Chat,
I
think
I've
actually
got
things
set
up,
so
I
can
keep
a
look
at
jabber
and
I
don't
know
if
there'll
be
very
much
of
the
jabber
going
on
in
any.
B
A
A
I
guess
I
will
go
ahead
with
the
administrivia
here.
This
is
the
note
12
statement
about
the
ipr
policies
in
the
ietf.
A
A
Anybody
can
look
at
this.
This
will
be
it's
all
being
recorded
be
available
later
people
should,
but
the
agenda
is
all
available
online.
You
can
download
these
slides
and
the
other
slides
for
this
meeting
and
you
should
take
a
look
at
them.
If
you
haven't
had
a
chance
to
do
so
so
far,.
A
A
This
is
the
tentative
agenda
which
is
currently
in
the
minus
trivia
and
agenda
bashing,
and
we
currently
have
three
presentations,
although
I
believe
mahesh
will
not
be
able
to
be
here
for
his,
so
I
guess
either
I'll
give
that
or
perhaps
barbara
who's
immediately
previous
presentation,
if
she's
willing
to
do
so.
A
We
also
need
somebody
to
take
some
minutes
to
be
highly
desirable.
So
barbara,
are
you
willing
to
take
minutes?
You
have
done
that.
A
A
All
right,
let's
see
a
little
bit.
A
I
can
stop
things
set
up
here,
so
I
can
actually
watch
the
list
of
participants
and
the
chat
at
the
same
time.
Okay,
so
barbara
is
willing
to
take
minutes.
A
A
A
A
In
any
case,
I,
the
question
is:
is
there
anybody
who
wants
to
change
the
agenda
if
the
first
presenter
has
continuing
difficulties
logging
in
we
can
change
the
order?
I
assume
that
would
be
okay
with
people
with
anybody
want
to
change,
do
anything
else
or
add
or
delete
items
on
the
agenda.
A
Not
so
the
next
thing
I
have
here
is
to
go
over
the
status
of
things
in
the
working
group,
so
this
is
the
document
status.
So
we
have
three
documents
that
are
past
the
iesg.
We
have
one
there's
a
applicability
statement
in
the
two
security
graphs.
A
A
The
yang
document
is
the
technically
in
working
group
last
call
which
has
been
running
for
a
long
time.
There's
been
some
delays
and
I'm
going
to
suggest
later
that
we
restart
that
working
with
last
call.
A
A
And
there's
the
whole
net
profile
which
is
currently
pending,
because
it
depended
on
bagel
documents
and
there
is
a
draft
on
ear,
which
is
a
type
of
technique
for
handling
multicast,
which
is
a
related
document.
A
Well,
let's
see
is
julius,
is
logged
in
I,
apparently
theophill
was
still
having
problems.
At
least
I
guess
that.
A
A
I'm
not
sure
why
that
would
be
true.
Why
don't
I
I
don't
know
because
people
object,
maybe
we
should
go
to
the
babel
information
model,
presentation
and
the
way
that
cfl's
to
see
if
that
problem
can
be
solved.
A
A
So
barbara,
would
you
like
to
request
the.
A
C
Okay,
so
before
we
start
the
actual
talks,
I
would
like
to
understand
what
is
the
stake?
What
is
the
status
so
I'm
understanding
that
the
isg
is
changing,
so
that
means
that
there
will
be
even
further
delay
for
the
documents.
Could
you
perhaps
explain
what
the
situation
is.
A
Okay,
chad
is
crank,
don't
like
well
I'm
here
you
will
faint
but
you're
asking
me
to
just
go
over
the
status
of
the
documents.
Again,
I
know
what
the
effect
is
of
the
issue.
C
The
question
is
that
six,
one,
two
six
bis
is
being
held
up
by
a
single
ad
and
has
been
held
up
by
just
two
ads
for
a
year.
Now,
okay,
we
have
managed
to
get
it
through
all
the
ads
save
one
right
now
and
it
is
my
understanding
day
of
jesus
has
changed
or
is
about
to
change.
It
changed
margins
actually,
so
what
does
that
mean
for
the
documents.
A
C
A
Know
how
to
get
that
objection
from
that
area
director
director
cleared.
So
I
don't
yes.
A
Well
already
one
of
them
already,
one
of
them
has
voted
no
objection
and
they
had
only
editorial
comments
which
are
non-blocking
so.
C
Okay,
so
donald,
it
has
been
a
lot
of
energy
a
lot
of
time
spent,
and
now
there
are
things
that
I
don't
quite
understand
that
implied
that
there
will
be
further
delay.
I'm
worried
about
how
much
more
energy
I
need
to
put
into
you
know
discussing
stuff
with
the
iesg.
I
mean
I've
been
discussing
stuff
with
the
asg
for
over
a
year.
Now
that
is
a
lot
of
energy
a
lot
of
time,
and
so
I
would
like
to
understand
what's
going
on
and
how
it's
going
to
end.
If
it's
going
to
end.
A
Well,
I
don't
know
if
martin
would
like
to
speak.
I
have
enabled
his
microphone.
D
D
So
this
is
david's
kanazzi
now
with
google,
just
so
to
comment
on
what
julius
was
saying.
As
the
co-author
on
rfc
6126
bis,
there
have
been
delays
yeah
with
this
process,
which
is
unfortunate,
but
luckily,
thanks
to
don
and
martin,
we
have
a
path
forward.
So
I
think
right
now.
The
action
item
julius
is
for
the
two
of
us
to
make
the
latest
edits
proposed
by
donald
and
then
post
a
revision.
D
So
we
could
post
it
this
week
now
that
they've
unblocked
submissions
and
then
that
should
allow
alvaro
to
lift
his
discus
and
once
we
have
that
it'll
be
very
swift
to
get
the
remaining
ads
on
board.
I
I
think
once
we
clear
the
discuss,
we
should
be
in
good
shape.
D
So
there
is
no
guarantee.
This
is
like
tcp.
If
you
have
100
packet
loss,
your
data
is
not
going
to
get
across
and
and
so
out
at
some
point.
D
Well,
let's
just
be
clear,
I
don't,
I
don't
think
there
everything
needs
to
have
a
timeout.
D
You
know
I,
if
I'm
underwater
and
I
can't
breathe,
I'm
going
to
keep
trying
no
matter
what,
and
so
in
this
case
the
main
reason
we
chartered
the
babel
working
group
was
to
produce
a
standards
track
document,
because
that
was
needed
for
other
groups
like
homenet
and
also
we
were
chartered
to
evolve
the
protocol.
D
So
we
I
don't
think
we
can
stop
until
we
have
a
standards
track
document.
I
don't
think.
C
They're,
giving
up
as
an
option
I'm
going
to
probably
retire
in
the
early
2040s,
so
there
is
a
default
timeout
at
some
point.
Okay,
so
at
some
point
we
need
to
give
up,
we
need
to
say
we
failed.
We've
gone
up,
we've
been
speaking
with
isg
for
over
a
year
now,
we've
been
speaking
with
the
isp
since
july
2019.
C
D
Absolutely
but
julius,
while
I
do
agree
that
some
members
of
the
ig
have
been
taking
a
long
time
to
get
back
to
us.
I
think
the
delay
is
also
on
the
side
of
the
authors
right
now.
Donald
has
has
may
and
martin
have
made
it
clear
to
us
what
alvaro
wants
and
we
haven't
made
those
edits.
Yet
so,
let's
make
those
edits
before
we
start
any
time
out,
because
right
now
it's.
C
C
A
C
D
Yeah
actually,
yes
same
question:
how
do
I
cut
my
microphone
and
leave
the
queue.
A
Well,
I
can
do
it
for
you
awesome,
thank
you,
zap
you're
gone
and
you
can.
You
can
ask
to
get
back
in
again
if
you
need
to
say
something.
So,
let's
see
if
theophil
is
still
having
problems
either
julia
should
present
or
we
should
just
go
right
ahead,
because
I
was
planning
with
barbara
stark
presenting
the
next
one
and
then
we
can
move
the
ipv4
over
ipv6
to
later.
A
So
what
I
was
heading
for
was
to
I
have
barbara
frazen
and
actually
barbara
is
currently
on
just
as
a
jabber
user.
Well,
that's
fascinating!
Maybe.
F
A
F
F
A
F
Do
I
need
to
get
in
the
queue?
Let's.
F
I
this
this
machine
doesn't
have
video,
so
okay,
okay,
so
the
information
model
yeah.
Let's
move
on
to
the
next
screen.
Thank
you
so
much
antana
for
doing
the
pr
with
those
fixes
that
was
a
big
help.
F
Those
are
the
things
that
antonio
fixed
and
I
merged
it
yesterday
into
the
editor's
version
I
haven't
actually
gone
through,
you
know,
github
is,
is
not
my
forte,
so
I'm
gonna
actually
go
through
and
make
sure
that
everything
looks
fine
and
hopefully
nobody
has
any
problem
with
this.
This
was
all
pretty
much
editorial,
but
then
we
also
had
discussion
if
you
can
go
to
the
next
screen
on
a
couple
of
other
issues.
Earlier
this
year,
clearing
versus
enabling
statistics,
we
discussed
that
the
two
booleans
we
have.
F
Four
statistics
do
need
to
be
separate
that
the
enabling
merely
enables
and
disables
the
collection
and
it
when
you
enable
it
just
picks
up
kind
of
where
it
left
off
and
it
doesn't
clear
out
anything
and
if
you
want
to
clear
out
the
statistics,
you
have
to
hit
the
reset
and
I
need
to
make
sure
that
the
description
is
properly
evokes
that
behavior
and
we,
I
do
believe
we
agreed
on
that.
So
I
do
need
to
do
a
change
just
to
make
sure
that
the
description
works
for
that.
F
But
now
the
next
thing
that
was
very
closely
related-
and
I
separated
this
into
two-
is
on
the
next
screen-
the
log,
the
packet
log.
F
So
this
one
was
just
a
little
bit
trickier
and
it
really
gets
trickier
in
and
given
what
mahesh
says
about
yang
and
I'd
really
like
some
input
from
implementers
on
some
of
this,
and
so
what
I
had
mentioned
was
okay,
so
we
have
the
babel
packet
log
enabled
we
don't
have
any
mechanism
to
reset,
and
so
it's
really
unclear
as
to
what
sorts
of
behaviors
you
can
or
should
get.
You
know
I
envisioned
four
different
possible
behaviors.
I
considered
two
to
be
a
bit
problematic.
F
F
F
All
it
can
do
is
clear
the
reference
to
the
log
file
memory,
which
is
a
bit
odd
and
not
really
that
great
of
a
thing,
because
I'd
hate
to
have
memory
just
filling
up
because
of
logging,
and
so
you
know
that
may
be
an
issue
with
yang,
but
I'd
like
to
get
some
feeling
of
other
implementers
if
anybody
has
implemented
any
of
the
packet
logging
and
what
your
thoughts
are
on,
how
to
clear
out
old
packet
logs.
So
you
can
start
new
packet
logs
and
I'd
like
to
pause
here
to
see.
A
Yeah
people
can
post
to
the
list.
I
know
you
can
post
a
question
there
also.
F
Okay,
so
not
sure
what
to
do
about
being
able
to
control
the
clearing
of
packet
logs
or
in
any
way,
control
that
behavior
or
know
what
the
behavior
would
be.
So
I'd
really
kind
of
appreciate
some
input
on
this,
and
it
may
be
that
nobody's
implementing
this,
in
which
case
we
definitely
need
to
keep
it
very
simple,
okay,
so,
let's
move
on,
which
is
just
a
screen
that
says
I'd
like
some
input
and
so
now
the
other
thing
you
know
this.
This
came
about.
F
Not
hearing
anything
or
saying
anything,
okay,
so
those
were
everything
I
had
and
my
question
was:
okay,
if
we
don't
do
anything
about
converge
fast
or
anything,
new
about
the
the
file
that
the
packet
logging
file.
C
Me
it's
so
good
to
hear
you
barbara,
I'm
wondering
so
that's
a
macro.
Essentially
the
convert
fast
thing:
that's
a
macro
for
a
set
of
parameters
right.
C
C
E
C
Be
in
favor
of
leaving
this
kind
of
convenient,
user-friendly
macros
to
the
user
interface
and
have
the
information
model
correspond
more
closely
to
what
is
going
on
in
the
implementation.
So
would
it
be
the
the
hello
interval
there's
the
hello
interval?
There
is
the
update
interval.
So
if
converge
fast
concerns,
what
is
I
think
in
a
previous
version?
I
don't
think
it's
in
the
last
version:
the
shorter
intervals.
There
are
quite
a
few
parameters
that
it
changes,
but
all
of
the
parameters
are
individually
exported
in
the
information
model.
F
Because,
at
the
time
you
you
were,
I
think,
suggesting
that
they're
I
mean
it's
certainly
very
complex
to
allow
a
user
to
directly
write
to
those,
and
I
think
it
was
desirable
to
have
the
user
know
what
they
were.
But
there
was
no
real
good
reason
for
a
user
to
change
them
at
the
time.
E
G
A
Yeah,
I
think
that
may
be
a
good
idea
because
we're
by
the
way,
the
the
way
this
way,
the
way
this
is
set
up
here,
for
these,
at
least
for
the
first
set
of
sessions
this
morning,
it
cuts
off
right
up
sharply
at
the
end.
F
Yeah,
okay,
so
the
last
slide
was
just
you
know.
If
we
do
changes,
I
do
need
to
post
another
draft
first.
I
think
I'd
like
to
have
this
discussion
just
to
see
what
to
do
about
that
and
then
do
we
need
to
do
another
wglc
or
these.
I.
F
And
some
of
this
is
is
because
of
the
the
comments
that
came
through
in
the
iesg
last
call
on
the
6126
bis.
So.
A
Okay,
yeah,
so
do
you
want
to
present
mahesh's
thing,
real,
quick,
barbara
sure.
A
F
A
And
so
go
ahead,
I
guess.
F
Okay,
so
mahesh
has
done
a
great
job
tracking
the
information
model
he'll
need
to
make
a
few
changes
because
of
the
pull
request
we
accepted
from
antonin,
but
he
has
no
problem
doing
that
and,
of
course,
if
we
make
other
changes,
he'll
need
to
change
that,
let's
move
on
so
there
was
the
you
know.
He
asked
several
times
if
we
needed
to
add
any
special
support,
other
than
default
routing
policy,
and
there
was
no
response.
So
he
closed
the
issue
with
the
assumption,
which
I
think
is
correct-
that
we
really
don't
want
anything
special.
F
F
And
it's
certainly
easier
to
do
nothing.
That's
always
just
the
easiest
thing
to
do.
Okay.
So
if
no
comments,
let's
move
on
it's
ready
for
working
group
last
call
so.
A
A
Yeah,
does
anybody
have
any
objection
of
doing
a
working
group
classical
which
is
kind
of
what
the
working
world
asks
anyway,
objections
to
the
document.
A
Okay,
I'm
not
hearing
nothing,
I
guess
we'll.
We
will
do
that
and
do
another
working
group
starting
at
we
start
with
a
working
group.
Last
call
on
the
yang
document.
A
So
see
in
the
chat,
there's
still
problems
with
cfl
getting
on,
but
julius
is
willing
to
present
for
him.
Actually
he
wanted
to
run
the
slides
himself,
but
I
don't
know
if
two
wants
to
do
that.
I
can
certainly
turn
over
the.
C
A
C
You
okay,
so
this
is
a
talk
given
by
tel
phil
bastian
and
it
was
joint
work
with
me
or
exactly
it
was
under
my
supervision,
so
to
feel
is
being
paid
by,
am
I
allowed
to
say
the
name
of
the
company
for
fun
still?
C
Okay,
it's
written
at
the
bottom
of
the
slide
yeah
so
anyway.
So
next
slide,
please
so
just
to
remind
everyone
about
what
we
are
doing
here.
This
whole
working
group
is
about
a
routing
protocol
and
the
goal
of
a
routing
protocol,
the
only
goal
of
a
routing
protocol.
Anything
else
is
just
useless
verbatim.
Technical
details
is
to
build
a
routing
table
now
it
will
surprise
nobody
that
the
routing
table
maps
network
prefixes
to
next
hubs.
Okay,
I'm
not
going.
C
C
Now,
how
is
this
rooting
table
used?
Well,
it
is
used
when
you
do
forward
and
packets
and
I'm
a
little
bit
shaky
on
this
slide,
because
it
is
due
to
tail
that
the
whole
point
is
that
the
next
hop
is
being
used
when
a
router
is
forwarding
a
packet,
so
it
receives
a
packet.
It
finds
the
right
entry
using
you
know
this
longest
prefer
shortest
longest.
Prefix
semantics
find
the
right
next
top
and
at
this
point
next
slide,
please
next
slide,
so
it
considers
in
the
routing
table
the
right
entry
next
slide.
C
Please
next
slide
please
next
slide
at
please
points
it
uses
the
next
hub
that
it
found
in
the
routing
table
to
find
out
the
mac
address
of
the
next
hub
interface.
So
it
says:
hey
using
the
ipv4
address
that
it
found
in
the
next
hop
entry
of
the
routing
table.
What's
your
mac
address
and
the
router
replies
next
slide,
please
next
slide
download
yeah.
It
says:
hey!
No,
no!
No!
Come
back!
Oh
sorry,
there's
a
time
to
work.
A
C
I
find
it
confusing
too,
and
it
says:
hey
my
mac
address
is
and
gives
the
mac
address.
That's
the
our
protocol
for
ipv4
the
neighbor
discovery
protocol
for
ipv6
next,
please,
but
the
whole
point
and
the
point
that
is
not
initially
obvious
when
you
start
looking
at
those
things,
is
that
the
next
hop
address
is
only
used
in
order
to
find
a
mac
address.
It
never
appears
on
the
wire
you
never
ever
use
it.
The
only
ik
addresses
that
actually
appear
on
the
wire
are
the
source
address
and
the
destination
address
next
slide.
C
The
only
thing
that
changes
is
that
the
next
hop
addresses
are
using
ipd6
next,
please,
okay,
so
what
you
end
up
doing
is
that
you
have
exactly
the
same
process
and
the
only
difference
is
that
in
the
routing
table
you
find
an
ipv6
next
hub
address.
Next,
please,
which
point
you
do
exactly
the
same
manipulation
as
you
did
with
the
northern
reipv4
table,
except
that,
instead
of
finding
out
the
mac
address
using
arp,
you
found
it
using
ipv6
neighbor
discovery.
Next,
please:
okay
and
the
forward
and
proceeds
as
normal.
C
Next,
please,
okay,
so
you
get
a
new
kind
of
entries
in
your
routing
tables.
In
a
routing
table,
you
have
usually
two
kinds
of
entries
in
a
double
stack
protocol,
such
as
babel.
You
have
v4
roots
and
v6
roots,
and
here
you
have
a
third
kind
of
route
which
we
call
the
v4
over
v6
route,
we're
not
very
happy
with
the
terminology.
So
if
anyone
has
better
suggestions
in
particular,
when
you
say
d4
over
v6,
people
tend
to
think
that
there
is
tunneling
involved,
that
there
is
absolutely
no
tunneling
involved
here.
C
We
found
the
idea
in
a
draft
that
applies
to
bgp
and
it
was
explained
to
me
by
a
guy
called
donald
at
an
itf
meeting,
not
this
donald,
a
different
guy
called
donald,
whose
name
I
forget,
but
it
is
in
a
draft
which
is
cited
on
this
slide.
Next,
please,
okay!
So
if
we
are
going
to
push
new
a
new
kind
of
routes
into
the
routing
table,
we
need
the
forwarding
plane
to
be
able
to
support
this
kind
of
route.
So
the
forwarding
plane
is
either
your
kernel.
C
If
you're
doing
software
routing-
or
it
is
your
hardware,
if
you're
doing
hardware
routing-
we
don't
know
about
hardware,
but
the
main
motivation
for
this
work
is
that
these
four
overv6
routes
have
been
supported
in
linux
since
july
2019,
and
we
even
have
on
this
slide.
The
exact
commit,
and
you
can
try
them
on
your
own-
you
can
do
use
them
using
manual
routing,
so
the
syntax
is
exactly
the
same
as
for
v4
roots,
the
only
difference
is
that
you
use
you
add
the
init
6
keyword
before
the
next
hop
address
in
the
ip
root.
C
C
Is
assigned
so
how
does
it
work?
Well,
if
your
interface
does
have
a
v4
address,
you
behave
just
like
an
extended
battle,
so
there
are
no
changes.
If
you
use
steels
in
the
code
in
the
network
in
which
all
interfaces
have
ipv4
addresses,
you
won't
see
any
difference
with
an
extended
battle.
On
the
other
hand,
if
your
interface
has
only
d6
addresses
well,
if
it
receives
a
v4
root,
it
installs
it
anyway.
C
C
C
So
there
are
various
ways
of
encoding
this
and
when
we
chose
the
encoding
we
decided
do
you
want
to
continue
to
heal
or
shall
I
do
it?
So
if
it
is
now
online?
C
Okay,
don't
hesitate
to
interrupt
me
if
you
want
to
take
over
so
one
of
the
things
that
we
absolutely
wanted
was
to
remain
compatible.
I
think
that's.
A
very
strong
sailing
selling
point
for
babel
is
that
we
have
remained
compatible
for
the
last
10
years.
C
So
what
we
wanted
is
that
and
the
next
time
that
that
you
could
have
in
a
single
network
an
extended
nodes
and
nodes
with
to's
extension,
so
a
node
needs
to
be
able
to
ignore
v4
over
v6
roots,
so
older
nodes
need
to
be
able
to
ignore
v4
over
v6
and
continue
to
root
correctly
pure
v4
and
v6,
and
there
were
various
encodings
considered
and
just
to
make
a
quick
reminder
in
babel.
We,
the
tag
that
is
used
to
tag
both
next
hops
and
prefixes
is
called
the
address
encoding.
C
There
are
three
of
those
there
are
ipv4
addresses,
ipv6
addresses
and
links.
Local
ipv6
addresses
that
use
a
compressed
representation.
Next,
please,
and
so
we
consider
various
encodings
and
the
two
encoders
that
seem
the
most
reasonable
are
the
one
that
we
call
the
toki
encoding
and
the
one
that
we
call
the
teufil
encoding.
So
number
one
is
what
was
suggested
by
toki
and
toki
said:
look,
don't
change
anything
to
the
protocol.
C
It
is
backwards
compatible,
but
it's
difficult
to
work
out
from
the
exact
wording
that
it
is
backwards
compatible
and
that's
something
that
implementers
might
get
wrong
and
it's
not
clear
that
an
extension
is
being
used.
There
is
no
tld.
There
is
no
way
just
by
examining
the
packets
all
the
tlvs
in
the
packet.
You
need
to
look
at
the
ordering
of
the
tlds
to
find
out
that
we're
using
the
inter
the
extension.
So
that
was
a
very
smart
encoding,
but
it
was
too
smart
for
my
taste.
C
What
we
suggest
instead
is
that
to
use
a
new
encoding
for
v4
over
v6
roots,
so
the
new
ae
is
used
only
for
v4
over
v6
routes
and
it's
and
it's
only
used
in
the
prep
in
the
update
tlv.
It
is
not
used
for
the
next
hub
and
a
d4
over
v6
roots
reuses,
the
v6
next
hub,
it's
obviously
backwards
compatible
since
new
encoding
is
going
to
be
cleared,
it's
more
obvious
from
the
package
what's
going
on.
C
C
There
are
no
tunnels
involved.
The
protocol
has
been
described
and
there
is
a
draft
that
has
been
there.
There
is
an
internet
draft,
we
have
a
production,
still
has
a
production,
ready
implementation
that
I
still
need
to
review
and
to
mark
that
I'm
fairly
confident
it
will
go
well.
We
think
that
it
should
be
experimental.
We
have
no
idea
how
to
proceed
from
here
and,
although
I
did
the
presentation,
the
work
is
almost
entirely
due
to
those
investing.
Thank
you
for
your
attention.
C
Maybe
not,
I
see
a
question
in
the
driver.
Let's
see
so
david
is
asking
for
direction.
D
Oh
all
right,
I
guess
I'll
just
finish
it
I'll
hide
it
back,
but-
and
I
forget
to
say
my
name-
this
is
david
snazzy,
so
just
to
fully
understand
the
the
the
options
we
have
here
so
question
for
julius
or
autofill.
Now
that
you're
here
with
the
first
encoder
to
refresh
my
memory,
I
apologize.
It
is
very
early
here
my
coffee
hasn't
kicked
in
yet,
but
on
the
tokyo
encoding.
D
Can
it
cause
backwards,
compatibility
issues
if,
with
an
implementation
that
hasn't
adopted
the
extension
or
just,
should
everything
just
work?
What
are
the
risks
of
being
too
clever?
So,
with
the
token
code.
G
Yeah,
if
I
maybe
I
might
explain,
yeah
thank
you
hi
by
the
way.
So,
if
you
look
at
the
rfc
for
babel
an
extended
level,
you
will
see
that
it
is
stated
that
if
a
root
has
no
obvious
next
hop
at
this
point
where
it
is
received,
the
node
should
just
drop
the
roots
and
not
use
it
at
all.
G
So
if
you
have
an
an
extended
bubble
which
receives
a
v4
root
with
no
v4
next
hop,
it
should
drop
the
roots
and
with
this
extension,
if
it
had
turkey's
encoding,
it
should
accept
the
root
as
v4
over
v6
sorry.
So
there
is
theoretically
no
problem
in
backwards
compatibility,
but
it
might
be
possible
that
someone
decides
that
maybe
there
is
a
smart
next
hop
for
v4
routes
with
no
v
for
an
x-hub
or
something,
and
it
might
break
it's
not
as
obvious
as
if
you
don't
know
an
ai
just
drop.
It.
C
C
Yeah,
so
here
we
are
assuming
that
people
deal
correctly
with
the
error
situation.
I
go
to
root
and
I
have
no
relevant
next
hop.
Okay,
it
is
ignore
the
because
people
implemented
the
error
handling
case
correctly.
I
expect
people
to
get
error
handling
wrong.
So,
oh
assuming
everybody
has
implemented
their
error
handling
just
right.
D
Okay
yeah,
because
so
I
can't
access
my
babel
implementation
anymore,
but
I'm
very
willing
to
bet
that
I
got
this
wrong
because
I'm
good
at
getting
things
wrong
like
that.
So
I
and
another
thing
I'll
add
is
so,
if
you
send,
let's
say
v6
net
hop,
then
a
v4
update
and
then
a
v4.
Next
hop
then
like
what
you're
describing
is
clear
because
you
haven't
had
a
v4
next
top.
But
what?
If
it's
in
the
other
order?
D
Let's
say
I
send
a
v4
next
top
and
then
a
v6
next
hop
and
then
an
update.
Would
that
was
the
current
draft?
Would
that
use
the
v4
next
top
or
or
would
I
need
to
only
send
it
without
having
center
before
next
top
tlv?
If
that
sentence.
C
Made
any
sense
so
the
next
the
so,
if
you
say,
send
an
update
in
which
them
in
which
ae.
C
D
Meant
with
the
token
coding,
the
the
the
one
with
the
ae
is
very
easy
to
reason
about,
I'm
just
trying
to
see
if
the
token
one,
what
are
the
problems
with
the
token.
C
One,
but
I've
definitely
noticed
the
token
coding,
so
what
we
do.
There
is
no
reason
to
send
a
single
tlv,
both
v4
and
v4,
over
v6.
Either
you
have
a
key
for
next
hop,
in
which
case
you
send
v4
or
you
don't,
in
which
case
you
send
g4
over
v6.
Yes,
that
makes
sense.
Okay,
so
the
case
just
doesn't
occur.
A
H
A
H
E
H
The
other
is
that
I
think
that
there's
an
issue
here
with
icmp
errors
coming
back
because
they
won't
the
nodes
that
are
on
the
ends
are
v4
but
they'll
be
getting
icmp
errors
back
from
an
ipv6
only
router,
and
I
don't
think
they'll
be
able
to
receive
them,
and
I
think
there
are
cases,
maybe
where
that
doesn't
matter,
or
you
want
to
argue
that
that
doesn't
matter,
but
I
think
that
should
at
least
be
discussed
in
the
draft
and-
and
there
should
be
some
explanation
of
what's
going
to
happen
in
that
case,
and
what,
therefore,
the
limitations
are
agreed
on.
A
C
A
Okay,
the
previous
session
ended
right
on
spot,
which
would
be
right
now,
but
I
think
maybe
there's
meat
echo
people
are
adjusting
these
things
in
real
time,
so
we
might
have
five
more
minutes,
so
we
could
continue
to
discuss
this
or
I
could
zip
through
my
presentation
quickly
whichever
people
want
to
do.
H
A
Oh,
yes,
you
can,
I
think,
but
I'll
see
I'll
do
it
there.
You
go
there's
some
way
to
take
yourself
out
of
the
queue
okay.
Well,
since
we
seem
to
have
some
more
time
here,.
A
A
Do
something
converted
to
implementation
experience,
but
it
specifically
says
it
doesn't
need
we
don't
have
to
prepare
a
document
reports
that
meetings
concerning
implementation
experience
would
be
good.
There
have
been
some
such
reports,
but
not
recently,
so
it
would
be
good
to
do
something
about
that
and
the
charter
current
charter
does
give
us.
They
have
an
optional
work
item
to
work
on
multicast
and
there's
been
at
least
one
draft
related
to
that.
That's
been
presented
a
couple
times
whether
that
would
work
or
not.
A
I
don't
know
the
question
of
course,
so
there
were
possible
new
lines.
This
ipv4,
via
or
whatever
v6.
There
was
a
type
of
service,
specific
routing
draft
a
while
ago,
which
expired
and
it's
the
possibility
of
doing
a
triple
a
211
mesh
path.
A
Selection
protocol
using
babel,
which
basically
is
mean
sort
of
integrating
it
with
ipv6,
and
that
was
a
sorry,
a
2
11
which
doesn't
require
any
changes
on
the
part
of
ieee
at
most
the
code
point
and
there's
precedent
where
they
did
previously
assign
a
code
point
for
a
ipf
draft
that
was
working
on
developing
a
a
pass
selection
protocol,
but
that's
kind
of
a
big
thing
and
might
be
a
little
hard
to
do
and
possibly
would
be
required
to
get
the
liaison
from
80
to
11,
saying
that
it
was
okay
for
us
to
do
that.
A
So
I'm
sort
of
rushing
through
this,
but
even
the
slides,
are
all
available.
So
it
changes
to
milestones.
I'm
suggesting
we
change
the
submission
of
the
yang
to
push
out
into
the
future
and
that
we
adopt
milestones
to
adopt
the
ipv4
over
ipv6
draft
and
to
adopt
a
multicast
draft
and
put
the
multicast
draft
deadline
pretty
far
into
the
future.
A
And
I
think
we
can
do
this
change
to
the
milestones
without
changing
the
charter,
because
the
multicast
is
already
in
the
charter
and
I
think
the
ipv4
via
ipv6
is
just
sort
of
an
extension
of
the
existing
protocol.
It's
not
too
like
a
big
change
or
anything
like
that.
A
So
if
anybody
has
any
comments
on
this,
that'll
be
fine,
otherwise
I'll
just
post
this
stuff
to
the
mailing
list
to
see
if
people
agree
and
doing
a
charter
update
is
a
big.
It's
a
big
deal.
So
it's
best
to
avoid
that
unless
we
need
to
do
that.
So
that
was
my
presentation
here.
A
And
I
think
we
now
have
one
minute
left
to
the
five
minute
end
of
the
five
minute
extension.
So,
unless
anybody
has
anything,
they
want
to
bring
up
or
say
thanks
to
everybody,
for
participating
and
discussing.
A
E
Hi
romaine
felt
here
yeah.
E
Pretty
well
by
the
way,
yeah,
I
have
a
question
on
the
expired
draft.
The
rtt
extension.
Is
there
any
plan
to
revive
that
one,
because
I
think
it's
interesting
and
by
the
way
way
back
in
singapore,
I
took
an
action
to
review
it.
I've
I've
read
it
now.
I
have
a
couple
of
minor
comments,
but
since
it's
expired,
I'll
hold
off
on
those
until
I
know
what.
A
I
I
would
say
we'd
welcome
your
comments,
even
if
the
draft
is
expired
and
you
should
probably
ping
the
authors
to
see
if
they
plan
to
update
it
and
once
it's
been
submitted
to
the
ietf.
You
know
it's
the
idf.
So
if
the
authors
aren't
willing
to
do
any
more
work
because
they've
moved
on,
but
you
want
to,
you
could
just
take
over
the
draft
so,
but
I
think
comments
on
it
would
be
useful
and
would
indicate
interest
in
that
kind
of
work.
A
Okay,
that's
good,
okay!
Well,
unless
there's
something
else,
I
guess
I
will
close
this
session.
Thank.
F
You
to
gabrielle
for
taking
the
notes,
did
an
excellent.
A
A
Great,
I
thank
you
also
and
not
hearing
any
other
comments,
not
to
figure
out
how
to
end
the
meetings.
A
I
guess
that
would
be
this
so
one
last
chance.
Anybody
want
to
say
anything,
okie
doke,
I
would
suspect
the
november
meeting
may
also
be
virtual,
but
we
shall
see
thanks
a
lot
bye.