►
From YouTube: IETF110-BIER-20210309-1200
Description
BIER meeting session at IETF110
2021/03/09 1200
https://datatracker.ietf.org/meeting/110/proceedings/
B
B
All
right,
we
are
just
a
minute
after
right
now,
what's
our
head
count,
look
like!
Oh,
it's
actually
looking
really
big
all
right
since
we
are
tight
and
I
think
they
probably
chop
us
off.
They
might
give
us
some
overlap,
but
we've
been
chopped
in
the
past
and
we
can
be
long-winded.
Let's
get
rolling
is
my
doc
still
sharing
it
is.
It
looks
like
right,
but
I'm
not
in
preview
mode
or.
B
B
That
was
slideshow
mode,
that's
the
one
that
makes
it
easy
all
right!
Welcome,
you're
in
the
beer
working
group
for
all
those
of
you
in
the
correct
time
zone.
Congratulations,
the
rest
of
us
don't
get
that
wow
slideshow
plays
on
its
own
all
right.
So
if
you're
here
you're
in
the
right
place,
no!
Well,
please
read:
please
note
note
it
well
any
questions.
B
Okay,
we're
going
to
need
some
note
takers,
so
this
stuff
gets
queued
up.
We
have
all
those
tools
available
right
here
in
our
meet
echo
ui.
So
can
we
have
someone
volunteer
to
manage
etherpad
and
take
notes?
B
B
B
Oh
fantastic
jeffrey
appreciate
it
okay,
so
we
can
move
excellent.
All
right,
noted
jabber
in
minutes
got
the
ether
pad
as
well.
Here's
a
current
agenda.
We
are
opening
wow,
that's
a
wrong.
Oh,
we
pushed
jeffrey
to
the
front
because
he's
got
to
take
off.
That's
right:
okay,
man,
I'm
sorry
jeffrey
you're,
not
here
for
the
the
next
one
too,
but
we'll
have
this
recorded
for
you,
I'm
sure,
you're.
Looking
forward
to
seeing
that
we
have
an
implementation,
we
give
them
a
good
time
slot.
B
B
B
Okay,
so
we
got
three
docs
in
the
queue
now
and
the
queue
is
gigantic.
I
think
the
one
has
been
in
there
for
months,
so
I
was
kind
of
waiting
to
see
things
go
through
to
stage
the
discussion,
but
it
looks
like
the
queue
so
thick
we're
gonna
start
pushing
some
other
stuff.
That's
been
prepared
as
well
just
waiting
for
some
feedback
from
other
groups.
Right
now,
we've
got
these
three
in
the
isdq,
which
is
pretty
exciting
status.
B
F
F
F
In
the
mld
extension
I'm
presenting
it
today,
but
it
hasn't
been,
you
know
last
call
yet.
B
Yeah
right
right,
this
is
just
status
right
now.
This
is
what's
happening
to
it
right,
it
says:
igmp
mld
extension
is
in
the
isdq
beer
mld
is
being
presented
in
the
meeting.
This
is
not
correct.
B
B
G
Sure,
I'm
not
sure
if
this
is
right
or
not,
can
you
guys
hear
me.
B
G
G
G
B
This
right,
this
one
is,
I
think,
what
she's
referring
to
is
an
actual
pim
draft
knight
gmp,
that's
in
there
that
I
guess
his
dependency
in
the
beer
mld.
B
Okay,
bls
extension.
We
have
again
we're
looking
for
that
to
be
revved
and
php
looks
like
it's
last
call
and
move
on.
It's
one,
that's
in
the
queue
that
we
got
response
from.
I
think
we
have
a
shepherd
for-
and
I
just
was
staging
them
out
so
we'll
start
pushing
those
on
through,
as
they
go
same
thing
with
evpn.
B
Possibly
next
call
and
we
definitely
need
shepherds
for
it.
As
we
move
we're
waiting
for
a
shepherd
review
of
beer
ping,
we
need
gosh.
Http
response
needs.
A
shepherd
and
v3
needs
a
shepherd.
Now
I've
asked
for
shepherds
after
each
one
of
the
meetings,
I
usually
get
two
or
three
and
it
moves
us
forward.
So
here's
my
pitch
once
again,
I
want
to
participate.
You
want
to
get
your
hands
dirty
involved
in
the
process.
This
is
a
great
way
to
simply
follow
instructions.
B
There's
a
draft
or
a
rfc
that
goes
through
the
process
just
to
confirm.
The
document
has
gone
through
all
the
necessary
procedures
in
the
itf
working
group.
Before
it's
moved
on
to
the
isdq
and
this
document,
then
you
know
it's
just
your
document,
you,
author,
confirming
all
those
questions
and
giving
any
comments
on
the
status
of
the
doc
and
that
moves
forward
with
the
queue
for
the
all.
I
have
to
review
before
getting
ratified
and
published.
B
So
please
I'll
I'll
jump
on
the
list
again
I'll
go
through
the
the
number
of
docs
and
which
ones
some
people
just
say
I'll.
Take
anything.
Some
people
have
some
passion.
I
want
to
grab
one
that's
great,
either
way,
whatever
works,
we'll
make
it
work.
So
there's
my
pitch
and
look
for
an
email
after
this,
and
probably
after
my
nap
and
then
I'll,
send
some
email
out.
A
Greg
to
thank
so
okay,
I
see
you're
putting
that
up,
I
I'll
be
presenting
tomorrow
in
best,
so
I
will
call
maybe
someone
in
best,
because
those
guys
have
dpdp
and
knowledge
will
volunteer
to
shepherd
actually
the
evp
and
extensions,
because
it's
no
hugely
vpn
specific,
I
don't
foresee
any
surprises,
but
you
know
it
would
be
good
to
maybe
get
them
involved
you
by
far
this.
B
Okay,
got
it
all
right,
idr
extension
has
actually
expired
kind
of
waiting
around.
We
need
authors
to
update
this
thing,
and
then
we
can
again
jump
on
a
shepherd
and
move
looks
like
lsr
extension
just
needs
an
update.
I
don't
know
what
the
status
is
off,
the
digging
that
to
see
but
they're
definitely
most
likely
connected
any
other
questions,
dot,
gaps.
A
B
B
B
C
So
in
this
zero
one
version,
I
have
updates
to
add
my
egress
protection.
So
today,
I'm
just
going
to
give
you
the
quicker
quick
review
of
this
egress
egress
protection.
C
First,
a
quick
overview
of
a
tethering
idea
itself,
basically,
a
router
that
does
not
supporting
beer,
which
we
call
a
non-bfr.
It
may
have
one
or
more
designated
helper
tether
to
it.
Typically,
it's
right
next
to
it
with
a
short,
a
fat
short
pipe.
C
D
C
Yeah
and
then
we,
the
section
6.9
procedure
in
the
architecture
rfc,
which
is
for
post
spf
handling
of
non-bfrs.
C
C
C
So
at
the
bottom
of
this
screen,
this
slides,
we
have
example
a
topology,
bfer1
and
bfbr2
c1
is
connected
to
both
bf
er1
and
bfbr2.
C
C2
is
not
only
connected
to
bfbr2,
so
bf
er2
is
the
designated
helper
for
bfer1
or
texting
a
protector
for
the
bfe
r1.
Now,
when
the
bfbr
one
fails
in
this
case,
it
could
be
either
the
real
node
fail,
failure
or
just
the
link
the
only
link
to
it
is
it's
down.
So
in
that
case
we
can
actually
consider
it
as
a
non-bfr.
C
So
with
that
we
actually
don't
need
additional
signaling.
Besides
existing
tethering
signaling,
the
bfier2
will
just
announce
that
it's
the
helper
of
bfer1
and
the
other
routers
will
see
that
and
they
realize
that
bfer1
itself
actually
advertiser
it
is
capable
of
beer.
Then
they
know
that
okay,
this
this
helper,
is
for
eager
protection.
Only
next
slide.
C
Yes,
yeah,
okay,
so
here
the
bfr1
is
the
point
of
local
repair
so
typically
or
in
the
workings
intuition.
Bfr1
send
the
package
to
that.
Has
the
bid
for
bfvr1
on
was
the
bfr1
was
in
the
packet
to
bf
er1
directly,
and
that
is
you
can
see
that
at
the
lower
left
corner,
which
has
the
bift
for
the
bf
r1,
in
that
it
has
an
entry
for
bfbr1?
C
C
So
that
backup
path
is
calculated
ahead
of
time,
based
on
that
advertising
on
from
bfer2
saying
that
it
is
the
helper
of
the
bfb
r1.
C
Now
bfo
er
two
has
its
own
beef
table
in
in
that
table.
There
is
a
bfb
r1
entry.
C
C
Because
when
when
bfer1
does
work,
we
do
want
the
package
go
to
bfer1
and
through
the
bfr1
now
the
backup
pass
is
that
to
decap
decapsulate
the
package
and
send
it
to
ce1.
C
C
It's
it's
worth
to
note
that
to
notice
that
both
ce1
and
c2,
if
if
they
both
need
to
receive
the
traffic
now
when,
when
the
when
there
be
a
a
package,
is
fired,
fastly,
routed
to
to
bfer2
and
that
if
that
packet
only
has
bfer1
bit
set,
then
the
the
bfdr2
will
only
send
one
ascended
to
c1
after
decapsulated.
C
That's
why,
in
this
scheme,
the
prr,
the
point
of
local
repair
will
not
flip
the
bits
that
is
they
would
set
for
bfer1.
You
will
not
flip
that
b2br2
bit
and
that's
what
what
was
done
on
another
egress
protection
method
now
also
in
the
bf
er
tools.
Now,
if
we
can,
we
go
back
to
the
earlier
slide
right
here,
yeah,
so
in
the
beer
table
for
bfbr2,
whether
it's
primary
paths
or
backup
pass,
I
did
not
draw
a
forwarding
bitmask
there.
C
C
And
that
is
actually
the
general
rule.
That's
any
encapsulation
and
and
branch
must
only
have
your
own
bits
that
one
particular
bit
set
that
well.
I
should
I
shouldn't
say
that
it's
a
general
rule
that
is
already
it
is
explicitly
spelled
out,
but
at
least
that's
my
understanding.
C
Next
next
slide,
please
two
slides
or
this
one
yeah,
just
this
one
yeah,
the
third,
the
third
notes
here.
Actually
I
I
need
to
go
back
to
the
earlier
slide
again.
Yes,
I
want
to
point
out
here
that
notice
steps
in
the
table
for
bfdr2
in
the
entry
for
bfbr1.
C
C
Now,
if
somehow
bfr1
detects
the
failure
of
bfer1
earlier
than
bfbr2
does,
then
we
will
have
a
transient
loop
here,
bfr1,
send
it
to
bfer2
and
bfbr2,
follow
the
primary
path
bfbr1
and
send
it
back
to
bfr1
so
that
transition
loop
can
be
avoided.
If
we
don't
use
that
primary
path.
Here
we
as
long
as
bfbr
one
gets
the
package
with
the
one
bit
set.
Then
you
will
always
just
do
the
decapsulation
and
send
it
to
the
downstream
receivers
in
the
over
in
the
overlay.
C
That's
to
avoid
that.
That's
to
avoid
that
transition
loop.
I
think.
That's,
basically
the
my
presentation
here.
It's
just
updates
to
include
the
egress
protection.
H
Yeah,
I
have
some
questions
so
last
year
we
post
the
equal
equals
protection
draft
as
an
individual
draft
under
this
draft,
so
jeffrey
proposed
a
new
protection
eagle
production.
So
this
new
proposal
is
added
into
a
working
group
and
document,
so
it
claims
that
either
we
use
the
helper,
but
the
organ
group
document
that
define
the
helper-
the
helper
is
different
from
the
protector
right
so
in
the
eagles
protection.
H
A
H
So
yeah
this
isn't
that,
because
it
is
this,
new
proposal
is
about
equals.
Production
is
a
different
topic,
so,
yes,
I
think
this
should
be
removed
and
then
placing
as
separate
individual
draft
right,
because
this
view
purpose
even
this
technically
technically,
this
new
proposal
have
some
issues,
so
I
will
submit
those
issues
in
at
least
I
would
because
I
don't
have
time
enough
time
to
discuss
those.
H
A
B
Yeah,
definitely,
if
we're
talking
about
capability
messages,
right
we'd,
like
something
consistent,
if
possible
and
consistency
as
well
in
terms
of
where
the
network
are
being
protected.
So
we
have
a
a
integrated
set
of
tools
instead
of
an
overlapping
set
of
tools.
C
So
procedure
wise,
if
we
want
to
separate
this
out
to
an
individual
draft
or
merge
with
another
draft
procedure-wise.
I
am
fine,
but
I
do
want
to
point
out.
C
This
is
very
well
integrated
with
the
with
the
tethering
solution,
because
in
the
tethering
solution,
a
helper
which
is
beer
capable
that
helps
a
non-vfr
and
when
it
is
used
in
the
egress
protection,
the
helper
becomes
a
protector
and
it
protects
the
the
protected
bfer.
C
And
in
fact,
when
we
did,
we
determine
when
to
time
to
tana
or
not
to
tunnel
that
is
basically
treating
as
as,
if
that's
that
helps
the
protected
bfvr
is
bfr.
Is
beer
incapable
as
a
non-bfr?
That's
how
we
decide
when
to
tunnel
when
not
a
tunnel,
so
it
is
indeed
extending
and
the
beer
tethering
procedure
to
the
egress
protection.
A
A
C
Oh
yeah,
so
I
I
didn't
realize
I
I
I
I
I.
D
B
Who
can
jump
up
here?
Jeffrey's
got
to
take
off
is
why
it
went
early
someone
I
think
meat
echo
needs
a
new
app
where
we
randomly
select
a
participant
and
they
become
the
notetaker.
A
Yeah
dice
throw
throwing
ice.
Yes,
gentlemen,
we
need
no
taking.
A
B
A
A
Okay,
so
we'll
call
us
all
these
notes
together,
fine,
so
I
will
put
in
so
that
was
presentation.
Okay,
I'm
on
it.
Yes,
thanks.
B
I
I
am,
I
am
here
is
the
presenter,
but
I
would
like
to
go
present
with
frederick
louis.
So
can
you
please
animate
him.
J
K
Can
you
see
this
so
thank
you.
So,
let's
start
and
thank
you
for
for
letting
us
present
our
for
our
letting
us
the
chance
to
present
our
work
here.
K
K
So
the
good
news
is
that
we
have
a
this
beer
implementation
on
free
data
plane
and
I'm
not
sure
if
you
are
familiar
with
p4,
but
we
are
using
p4v
and
v2
virtual
switch
data
plane
in
order
to
to
to
learn
to
to
do
our
homework,
and
then
we
transpose
the
work
all
to
all
the
whole
work
we
did
with
the
tofino
switch,
which
which
came
from
initially
barefoot,
now
called
intel,
and
we
also
progress
with
zpdk
that
are
playing
so
we
are.
K
I
will
be
co-presenting
this
this
this
20
minutes
with
shabba
mate,
which
is
the
rata,
lead
car
developer
and
I'm
just
the
the
the
r
project
working
for
for
about
kimball
forgero
on
behalf
of
and
I'm
the
technical
leader
of
the
the
project.
So
a
real
brief
agenda.
We
really
quickly
present
the
project,
brsc
draft
implementation
and
the
the
experience
the
experience,
the
br
implementation
experience.
K
We
we
we
lived
since
2017
a
little
a
word
about
interworking
and
one
focus
on
loop,
enrolling
topic
with
a
beer
application
and
a
very
small
conclusion.
So
the
haar
project
r
stands
for
router
for
academia,
research
and
education
and
started
in
2019,
and
the
group
request
was
to
simply
try
to
couple
a
control
plane
with
a
multiple
data
plane
and
by
and
by
interfacing
the
two.
K
We
can
have
a
fully
functional
router
working
at
different
data,
plane
rate,
but
basically
it
will
leverage
the
the
hardware
capacity
in
order
to
provide
some
interesting,
some
feature
that
are
that
that
are
specialized
per
use
case.
So
the
idea
is
to
re
the
the
real
idea
behind
a
rare
free
router
is
that
it's
a
one,
familiar
platform
that
can
be
used
to
create
multiple
solutions
and
each
solution
will
address
multiple
use
cases
in
the
research
and
education.
Of
course,
it's
not
only
about
research
education.
K
It
can
also
fit
your
own
use
case
as
soon
as
it
is
compatible.
When
I'm
talking
about
use
case,
it's
simply
because
we
are
using
one
control
plane,
but
multiple
data
planes
and
depending
the
use
case
you
would
like
to
address
the
data
plane,
will
be
different.
Whether
you
would
like
to
have
a
to
use
a
pure
label
switch
router
level
switch
router
as
a
p
router
with
mpls,
you
will
go
use
a
hardware
that
is
able
to
switch.
K
K
K
So
not
too
not
just
to
give
you
some
idea
about
what
we
we
did
with
the
the
the
the
doppler
plane
we
support
now
is
p
rp
for
vmv2
tofino
switch.
We
are
studying
the
possibility
also
to
to
run
our
code
on
fpgas.
K
K
We
are
very.
This
is
the
the
p4
lab
we
are
running
in
europe
since
yesterday
we
are
connected
to
chicago
in
the
us
and
from
there
you
have
we
have
already.
We
have
the
possibility
to
reach
the
whole
research
and
education
u.s
area,
so
we
are
working
with
10
tech,
tennessee,
university,
northern
university
and
caltech.
Also
in
order
to
to
provide
to
provide
this
live,
live
p4
lab.
K
We
are
using
this
lab
to
to
test
the
beer
implementation
where,
for
example,
some
nodes
are
hosting
some
bare
metal
server
that
produce
a
source
multicast
source.
Sorry-
and
we
are
we
we
can
also
pre
prevent.
We
can
also
connect
a
receiver
on
top
of
this
visa
green
box.
B
Chicago
that's
star,
tap
right.
Is
it
still
called
star
tap
starlight
starlight?
Now,
but
that's
that's
the
like
university
research,
education,
peering
point!
Yes,
exactly!
Yes,
okay,
excellent,
so
actually
you've
accessed
the
entire
i2
infrastructure,
but
those
are
the
universes
that
collectively.
I
I
I
know
about
the
others,
but
currently,
as
far
as
I
know,
at
least
for
us,
the
mpls
encompassion
is
the
target,
and
this
is
what
free
router
have
since
beer
became
rfc
in
software
that
plane,
that
is,
when
free
router
itself
handles
all
the
packets
itself
in
in
the
java
control
plane
and
and
do
the
the
the
forwarding
itself.
This
data
plane
supports
all
the
b
strings
written
in
the
in
the
beer
and
mpls
cerasi
and,
moreover,
it
supports
the
the
b
string
domain.
I
So
I
mean,
when
you
have
two
two
gear
domains:
one
running
on
on
a
on
the
b
string,
lengths
a
and
other
on
the
b
string
ranks
b
and
the
domain
border
router
have
two
igps
and
and
and
have
to
deconstruct.
And
if
it
goes
from
the
bigger
to
the
lower
missing
length,
then
it
have
to
emit
two
pockets
with
two
different
s:
eyes
on
on
the
lower
side
and
and
the
same
on
the
backward
direction.
But
obviously
it
won't
duplicate
the
packets.
I
I
It
has
no
inter
interoperability
tests
at
the
moment,
but
via
shrek
nowadays,
decodes
isis
lsps
with
bigger
extensions
very
well.
So
I
am
pretty
sure
that
if
one
of
you
have
something,
then
it
would
be
interesting
to
see
how
it
interworks.
I
Regarding
west
pf,
we
already
had
an
interrupt
interoperability
testing,
which
went
quite
well
and
thanks
guys,
tony
and
sorry,
I
cannot
pronounce
the
name,
but
I
had
to
so
thanks
to
ripper
guys
for
for
letting
this
happen,
I
mean
the
ospf
interrupt
regarding
the
idea
I
mean
bgp,
with
with
the
beer
attributes
attached
to
a
prefix
from
a
different
domain,
or
from
the
same
I
mean
as
or
the
same
as
we
already
have
it.
I
Obviously,
the
the
attribute
type
is
is
a
has
random
value
from
the
reserved
part,
but
the
population
of
the
data
plane
and
the
computation
is
already
there.
So
when
the
two
bids
get
populated,
it
will
be
the
final
version
from
our
side
and
more
interestingly,
so
our
primary
target
is
to
provide
an
iptv
like
experiment
in
the
field.
Finally,
in
our
enron
in
hungary,
so
we
went
for
basically
for
the
pim
beer
draft.
I
That
is,
I'm
not
only
concentrating
on
beer,
because
we
will
go
with
the
with
the
cps,
but
they
will
be
ps
in
this
case.
So
obviously
I
test
with
pin
but,
but
also
the
same,
is
true
from
this.
From
this
point,
when
I
mentioned
beam,
it
is
the
same
for
for
igmp.
That
is
from
freewriter's
point
of
view.
It's
unimportant
if
the
join
or
or
the
report
came
from,
pmore
igmp
frederick
excited
please
frederick
next
site,
please
thanks
expire
and
experiment.
So
two.
A
Quick
questions
louie
two
quick
questions,
so
one
pausing,
I
assume
he
you
are
attaching
poland
to
psnc
right.
Yes,.
D
I
So
the
thing
is
that
we
will
came
out
soon
with
with
the
public
only
order
image
and
the
data
plane
torsional
binary.
That
basically
did
the
forwarding
code
of
the
asic
is,
is
just
a
50,
kilobyte
debian
package
that
you
can,
that
you
can
field
replace,
not
everything
is
packed.
To
the
same
I
mean
not.
Every
feature
is
packed
to
the
same,
only
image,
because
we
have
too
many
features
to
fit
the
asic.
I
So
so
you
have
to
to
to
choose
between
the
between
the
features
but,
but
we
already
built
the
beer
feature
set.
That
is
an
mpls
features
that
choose
the
beer
forwarding
and
the
igmp
and
the
and
the
pim
interworking
from
the
data
plane
part.
The
software
I
mean
the
control
plane
part
is
animated,
so
it
is
the
same
free
router,
regardless
which
data
plane
you
are
using.
A
J
I
Answering
to
your
question
explicitly:
yes,
we
have.
We
have
a
specific
side
where
I
would
like
to
ask
what
vendors
here
that
please
please
find
us
or
or
let
us
find
you
and
let's
do
the
interrupt.
A
I
Okay,
so
basically
how
how
it
could
happen
when,
when
you
came
out
with
with
the
draft,
I
started
to
study
it
and
I
started
to
implement
it
on
my
home
network.
All
the
underlying
words
and
links
are
clicked
here.
You
will
see
that
it
is.
It
is
a
small
home
network,
but
about
10
or
more
routing
nodes
and
all
have
virtual
routers
within
in
in
some
basically
a
30
to
50
nodes,
depending
on
which
day
is
it
on
the
week.
I
Basically,
and
but
it's
an
es
esp
like
set
up
here,
that
is,
I
run
to
two
route
reflectors
and
and
these
mpls
vpns,
both
layer,
2
and
layer,
3
and
so
on.
So
so
people
with
a
very
small
one
in
in
a
in
a
small
household
and
obviously
have
my
own
needs
like
to
to
have
a
two
zone
music
here
and
it
quickly
turned
out
that
only
perfect
solution
on
over
internet
is
is
a
real
broadcast.
So
I
started
studying
initially
mldp
sorry,
initially
pinned
an
mldp
and
nowadays
beer.
I
Obviously,
when
when,
when
I
didn't
had
a
data
plane,
I
run
it
in
software
switching
and
I
and
I
saw
my
cpus
matting,
but
now
they
with
the
with
the
dpdk,
it's
gone
free
next
night.
Please
I
don't
know,
please
don't.
Please
are
back
please
back
and
yes
thanks
again
for
the
juniper
guys
who
let
it
happen
recently.
I
We
had
a
real,
the
first
real
interrupt
test
over
ospf
and
I
think
it
was
just
my
bad
that
I
didn't
proceed
toward
the
thing
is
that
it
really
interworked,
I
I
you
will
see
hopefully
in
the
next
slides
I've
got
an
access
to
to
an
intermediate
image.
I
I
had
some
screen
shots
from
it,
the
the
I
tanks
it
is.
It
is
not
one
but
but
but
somewhere
else
about
it.
It
is.
I
It
is
a
test,
it
is
a
test,
but
if
you
heard
about
dm42.net
it
is,
it
is
a
community.
I
It
is
a
community
movement
to
let
young
people
experiment
with
bgp
and
I
joined
it
so
early,
and
the
thing
is
that
anyone
can
peer
to
anybody
over
here
and
my
own
public
df-42
test
bed
got
an
a
beer
draft
idea.
That
is
a
bgp.
I
So
what
got
this
feature
about
two
years
ago,
when
I
implemented
the
the
bgp
part
for
the
for
the
beer.
Since
then
anybody
who
who
joins
the
tr-42
on
on
my
notes
or
or
pierce
to
be
over
a
gre
or
anything
it's
public
and
and
anonymous
you
can.
You
can
interrupt
test
basically
zero
to
24
hours
for
reckoning,
snipers.
I
And
yes,
it
is
the
first
screenshot
of
the
juniper
interrupt
test
on
the
upper
half
and
please
can
can
repair
guys
unmute.
You
probably
you
will
have
better
better
comments
than
me.
So
the
upper
half
on
the
screen,
you
can
see
how
how
the
vmx
decoded
decoded
the
ospf
lsas
that
I
I
sent
to
him.
Basically,
the
vmix,
as
you
can
see
on
the
bottom
of
the
screen,
was
surrounded
by
by
two
free
router
vrfs.
In
this
case,
from
the
left
and
from
the
right
I
will.
I
I
will
name
that
I
record
them
in
this
name
on
the
proceedings
frederick
next
night,
please
so
the
beer
forwarding
table
from
the
julia
burst
point
of
view.
Remember
the
left
and
right
were
on
the
g001
and
g002
and
from
the
left,
which
is
the
one
and
the
vmix
here,
the
222.111
and
it
picked
up
the
prefixes
perfectly
immediately
without
any
hacking
on
it
and
on
the
next
slide
you
can
see
that
the
bf
id
embedded,
I
mean
attached
to
the
preferred
please
back:
okay,
no
okay,
okay,
perfect!
I
So
on
on
the
bottom
of
the
screen,
you
can
see
that
how,
via
mix
attached
bfr
ids
from
the
prefixes.
I
To
the
next
hops,
basically,
but
yes,
I
think
it
is.
It
is
attached
to
the
next
host
from
my
from
free
router's
terminology,
but
I
think
juniper
have
a
different
terminology
for
it
never
mind
and
the
other
thing
you
note
that
my
implementation
advertises
random
labels
every
time
and
juniper
picked
it
perfectly
from
the
start
and
every
time
I
reloaded
my
side,
it
worked
fine
and
repopulated
the
tables
and
on
the
middle
of
the
screen,
the
one
two
three,
the
slash
60
is
the
juniper's
own
loopback
address
in
this
domain.
I
Obviously
he
didn't
allocate
it
for
himself
a
low
back
label
or
I
don't
know
how
to
cut
it
for
a
week
next
year.
Next
night,
please.
I
Yes,
detailing
the
the
forwarding
I
mean
the
beer
forwarding
table
with
the
next
hopes.
From
the
juniper's
point
of
view,
this
is
the
b
string
on
the
wire,
basically
dumped
you
in
binary.
I
It
is
a
very,
very
useful
thing,
because
when
you
open
the
packet
capture
and
you
convert
back
the
binary
to
eggs,
it
is
the
same
what
you
can
see
in
the
after
the
beer
header
I
mean
in
the
baseline
part
and
again,
the
next
hope
associated
to
it,
and
the
outgoing
interface
feel
like
next
time.
Please.
I
This
screenshot
is
my
buggy
sorry,
so
the
ospf
from
the
left
and
the
last
pair
from
the
right
basically
nothing
interesting.
I
I
used
the
1.11.2,
I
enabled
beer
on
the
area
and
the
traffic
engineering,
I'm
not
sure
if,
if
julia
checks
for
it
or
or
not,
I
nobody
just
for
sure,
because
because
I
actually
didn't
want
to
check
for
it
later,
I
I
did
anyway,
it's
just
when
the
screenshot
was
taken
yes
and
look
back
from
the
left
of
the
juniper
and
how
I
attached
the
beer
index
to
that
graphics.
I
Federic
next,
like
this
and
yes,
it
is,
it
is
with
the
with
the
set,
the
filter
from
the
juniper's
right
side,
but
still
from
free
router.
Basically,
what
you
can
spot
here
is
is
two
channels.
It's
it's
a
free
radar,
specific
thing
to
be
able
to
quickly
test.
The
idea
here
is
the
same
when
similar,
when
what
what
you
think
what
you
do
when
it
comes
to
point
to
multi-point
or
svpt,
and
basically,
you
set
a
dummy
destination
with
the
free
router
international.
I
Don't
care
about
it,
but
on
the
domain
name
part
you
can
list
more
eyepiece
and
freerupter
will
go
for
the
bf
ids
and
populate
the
header.
With
with
them
I
mean
the
bits
set
with
them.
So
basically,
I
I
can
do
a
static,
multicast
threading.
If
I
want
in
this
way
it
is.
It
is
here
for
just
for
this
or
two
testing
in
this
case
with
the
beer,
junipero
interoperability,
I
only
set
one
bit
from
the
juniper's
left
and
right
and
I
assigned
an
unicos
type,
as
you
can
see,
3c3
and
for
the
next
trip.
I
Please,
and
basically
I
did
nothing
but
pink
through.
No,
it's
it's
not
the
slide,
so
it's
this
slide
is
about
how
free
wrather
saw
from
juniper
advertisement
from
left
and
right
and
more.
Interestingly,
on
the
left
and
right,
I
heard
clearly
the
other
side
I
mean
on
the
left
side.
I
I
can
see
2
2
2
that
one
two
three,
which
is
truly
pairs
on
loopback
address
and
the
advertised
table.
I
One
two
three
which
I
configured
to
it
and
which,
which
I
wanted
to
see
here,
is
the
two
two
two
that
two
two
two,
which
is
the
juniper's
right
side.
So
juniper
advertised
I
mean
flooded
the
lsa's
through
itself
to
the
left,
which
is
which
is
basically
normal,
but
I
I
wanted
to
see
it
and
finally,
the
forwarding
table
from
free
router's
point
of
view
on
left
and
right
vrs.
No,
unfortunately,
I
added
the
mystic
and
the
left
the
visual
vf
for
the
next
site.
I
Please,
yes,
so
how
a
specific
label
looks
like
in
freewater's
point
of
view.
Basically,
is
the
same.
What
what
you?
What
you
have
so
already
on
on
the
beers
now
on
jupiter's
beer
forwarding
table
when
juniper
populated
its
its
b
string
and
us
assigned
it
to
an
s
op
and
an
interface
and
so
on,
but
it's
from
the
free
routers.
I
I
I
G001G002
was
connected
to
pv
at
l1
and
2,
and
the
thing
is
that
I
sent
70
kilobytes
and
I
received
70
kilobytes
and
the
backward
direction
is
the
same.
That
is,
I,
I
saw
no
a
single
bit
additional
deletion
from
the
pocket.
That
is,
the
forwarding
happened
perfectly
from
my
point
of
view
later.
Obviously,
I
did
some
other
testing
and
then
captures
on
the
interface
and
I
spot
nothing
really
nothing
floric
next
time.
Please.
I
Yes,
so,
finally,
the
last
sign
the
last
light
from
my
part
is
it's
a
bit
a
bit
of
poor
man's
poor
man's
beer
forwarding,
because
we
went
now
with
with
p
for
data
plane,
and
here
I
had
two
decision.
I
To
be
honest
I
didn't
like,
nor
of
them
the
the
fourth
decision.
From
my
point
of
view,
if
I
make
the
the
real
ending
of
the
b
string
with
the
calculated
with
the
calculated
b
string,
then
I
send
out
the
pocket
if
any
bit
remained
one,
and
I
I
do
a
note
end
on
the
pocket
and
recirculate
the
pocket
if
any
bit
remained
one.
I
But
this
thing
would
you
like
a
rest
circulation
which,
if
I
had
to
recirculate
to
three
piers,
I
I
mean
if
I
had
to
recirculate
three
times
for
three
very
different
peers,
then
it
would
divide
the
the
beer
performance.
I
I
don't
want
to
explain
two
more
or
only
if
you
are
very
interested
in
it.
The
thing
is
that
I
think
you
will
see
some
run
basics
in
the
future.
That
would
decide
if
they
go
with
the
r
welding
or
or
circulation.
I
If
you
have
any
clever
mathematics
over
there,
I
would
like
to
see
a
draft
and
the
proof
that
how
can
this
rolling
be
done?
Well,
I
mean
I
tried
some
things
and
finally,
I
have
now
an
idea
that
that,
if
I,
if
I
fully
disable,
equals
multi-passing
calculations
for
beer,
I
mean
the
the
prefixes
that
have
a
bfr
id
and
I
need
to
do
the
b,
the
beer
calculation,
because
the
operator
neighborhood,
etcetera,
etcetera,
and
then
I
do
a
second
calculation.
I
We
don't
need
only
the
first
possibilities,
populated
and
and
with
this
solution,
which
is
far
from
perfect.
Obviously
I
reached
the
state
that
the
last
hope
do
the
pocket
division.
I
mean
in
this
case
I
I
usually
saw
erwan
doing
replication
and
the
two
replicas
proceeded
on
air
1,
r2
f4
and
r1
s3
r5,
with
with
the
tricks
I
tried.
I
Finally,
I
reached
the
point
that
r1
does
not
duplicate
duplicate
the
pocket,
but
it
basically
four
bars
with
f2,
with
two
bits
set
and
air
to
basically
duplicates
the
pockets
to
air
four
and
there
five,
but
it's
it's
again
a
poor
man's
solution
for
just
to
to
overcame
the
the
forwarding
limitations.
A
Yeah
yeah
tony,
I
think
we
had
a
chat.
You
know
out
of
band
on
private
email
about
the
whole
thing.
There's
a
lot
of
traders
right.
The
beer
forwarding
pad,
allows
you
to
paralyze
or
recirculate
or
whatever
there's
a
lot
of
implementation.
Depending
you
know
what
you
optimize
for
cost
speed
fan
out.
You
can
do
all
kind
of
trade-offs
for
the
ecmp.
A
I
mean
I'm
not
saying
too
much
but
architecturally
what
we
did.
We
really
have
an
igp
next
stop
or
whatever
provides
the
next
stop
underneath
beer.
So
we
do
not
try
to
do
the
next
pcmp
in
beer.
We
actually
fall
back
on
the
igp
next
stop,
and
you
see
that
also
when
you
look
at
our
rips
in
the
cli
that
we
really
use
igp,
we
use
protection,
we
use
ecmp
and
so
on,
architecturally,
but
also
of
course,
it's
an
indirection
that
you
have
to
provide.
I
And
yes,
yes,
you
are
absolutely
right
and
this
problem,
I
think,
can
can
can
be,
can
can
cannot
put
without
ac
ecmp
easily.
I
I
agree
with
that.
The
the
forwarding
solution
we
are
using
can
result
sub-optimal
duplication
points
without
ac
mpl.
So
it's
it's
absolutely
true.
But
from
my
point
of
view,
what
I
am
planning
to
I
mean
the
the
forwarding
ac
code
is
tuned
to
at
the
moment
is
to
avoid
as
much
as
possible
recirculation
and
it's
and
it's
and
it's
true.
So
we
really
do
not
recirculate
for
beer
replication,
but
then
I
had
to
add
all
that's
it.
Yeah.
A
I
Yes,
absolutely
absolutely
absolutely,
and
the
only
thing
that
I
am
aware
of-
and
we
need
to
document
with
this
thing
is
here
in
my
test
bed.
I
I
run
it
over.
So
basically,
old
nodes
have
a
four
or
five
or
eight
maximum
gigabit
connections
to
a
common
layer.
Two
and
the
routing
nodes
here
are
rooted
on
the
stick
configuration
and
in
when
it
comes
to
the
hardware,
and
then
I
have
to
replicate
two
pockets
to
the
same
port.
I
Then
it
obviously
will
overload
it.
So
that's
that's
the
only
thing
we
spotted,
but
since
then
we
had
the
threat
we
have
the
chance
to
to
to
measure
it.
I
mean
not
fully
because
we
have
some.
We
have
some
limitations
with
the
visual
and
so
on,
but
I
saw
it
performing
at
50
gigabits
already,
which
is
not
bad
after
a
while.
A
I
K
Sorry
key
takeover
slide.
Well,
everything
everyone
is
is
doing
it,
but
we
just
wanted
to
to
to
mention
that
we
are
also
automating
all
the
testing
related
to
all
the
features
and
beer
is
included.
You
can
fill
the
the
whole
list
of
tests
in
the
project
page.
Essentially,
it's
roughly
2000
tests
that
are
a
daily,
no,
not
daily,
but
automatically
triggered
each
time.
The
new
code
is
committed
into
the
the
git.
K
In
the
end,
you
have
a
freshly
image
like
a
junos,
install
or
cisco,
or
whatever
firmware,
install
and
wineline
installer.
That
will
let
you
handle
the
network
that
lets
you
transform
your
p4
router
or
your
p4
switch
into
a
full-fledged
router
having
instruments.
K
See
almost-
and
one
thing
that
is
important-
is
that
we
have
a
prior
aspirant
that
help
us
to
where
you
are
using
rfc
conformance
test
next
step
is
to
provide
profile
p
for
profile
calibration.
That
example
mentioned,
which
means
that
each
image
will
have
a
profile.
One
maybe
may
will
be
dedicated
to
a
beer.
One
can
be
pure
mpls,
the
other
one
can
be
srv6
only
and
dpd.
We
have
a
p4dpdk
that
applies,
that
is
in
operation.
K
So
that
being
said,
this
is
you
can
find
the
useful
some
useful
link
which
are
the
the
project
page
halfwayrouter.net.
K
We
have
a
blog
where
you
can
find
a
lot
of
articles
interesting
article
that
we
that
will
help
you
install
your
home
router
with
our
product,
which
is
what
are
you
I'm
using
right
now
where,
during
this
vc,
and
we
have
also
docs
that
freewater.net,
where
we
are
trying
to
provide
a
a
control
plane,
documentation
related
to
that?
If
you
want
to
reach
us,
you
have
our
user
mailing
list.
K
If
you
want
to
contribute
and
add
your
own
yeah,
your
bits
and
bytes
into
into
the
control
plane,
and
also
the
before
that,
application,
where
you
are
very
welcome
to
to
contact
us
and.
I
An
iits
specific
thing
here
is:
if
any
one
of
you
have
something
that
works
or
badly
works.
Please
contact
us
and
we
are
more
than
happy
to
make
some
interrupts.
K
Yes,
thank
you
and
we,
you
can
follow
us
on
our
with
our
twitter
account
where
you
you
you.
We
are
pushing
frequently
new
new
progress
articles
and
an
experiment
and
special
thanks
to
all
of
this,
our
organization
we
which
make
everything
possible
because
we
just
we
really
did
not
reinvent
the
wheel,
but
we
assembled
a
lot
of
existing
working
technologies
coming
from
the
the
whole
this
whole
organization
and
sub
project.
So
thanks
any.
A
B
All
right
ken
thank
you
that
was
fantastic
appreciate.
It.
M
N
Yeah
go
for
it,
tauros
does
anyone
know
I
mean
I
forgot
the
name
of
it,
but
you
know:
isn't
there
also
an
equally
large
test
bed
in
the
u.s
is?
Are
you
folks?
I
mean
I'm
not
aware
much
of
what
goes
on
in
on
research
there,
but
have
you
folks
from
giant
heard
anything
from
from
the
u.s
side?
N
N
I
Perfectly
so
I
usually
made
a
guy
from
esnet
and
he
gave
me
an
access
to
the
bastion
server
where,
from
he
told
me
that
I
have
an
access
to
a
p4
switcher
somewhere
in
the
states
in
a
university
campus
and
that
that
site,
as
far
as
I
know,
somehow
connected
to
our
test.
But
that
is
I,
if
you
are
really
interested,
please
drop,
look
via
mia
and
I
will
find
a
guy
and
I
will
contact
connect
you
with
him
and
at
this
point
we
will
be.
Oh.
N
Oh,
no,
they
they're
not
allowed
to
to
to
talk
to
us
because
of
the
the
government.
I
I
I
Okay,
but
but
I
know
that
there
is
a
p4
switch
that
is
surrounded
by
by
juniper
and
cisco
boxes
and
probably
probably
others-
I
saw
them
in
ldp
and
if
you
can
somehow
provide
the
image
to
them
at
this
point
he
can
perform
a
very
good
experiment,
because
from
uw-site
we
regularly
can
inject
real.
I
I
mean
broadcast
from
my
house
that
if
you,
if
you
saw
on
the
presentations
link,
I
I
did
a
demonstration
where
I
I
had
the
music
broadcast
into
two
rooms,
but
it
is
available
in
all
the
pops
in
the
test,
but
I
think
it's
the
same
for
for
the
states.
So
basically,
when
I
press
the
play
button
here,
you
hear
the
same
at
the
states.
K
Just
be
more
factual,
we
we,
we
have
a
respect
to
u.s
area,
they
have
a
before
test
bed
and
we
which
and
we
plan
to
connect
to
this
testbed
via
chicago
fantastic.
I
Yes,
it
is,
it
is
another
yes
and
the
third
one
that
I
heard
about
the
essence.
Guys
are
cooking
something
on
top
of
smartnix,
but
I
I
didn't
went
too
far
to
to
ask
them
what
they
are
planning
to
do
with
that.
B
All
right,
we
are
crunched
time
wise.
I
wanted
to
give
you
guys
all
that
for
certain
some
of
these
connections-
and
you
know
getting
connected
and
understanding
the
links.
That's
all
there
in
the
presentation.
B
We're
there
excellent.
Thank
you
guys
that
was
fantastic,
stig
you're
up.
B
Yeah,
your
audio
is
working.
I
will
share
the
doc
here.
B
F
F
Okay,
so
I'll
just
go
ahead:
yeah
please
go
to
the
next
slide,
please,
okay,
so
yeah!
This
slide
talks
about
the
the
cur
it
says.
Currently,
the
source
address
is
assumed
to
be
a
beer
prefix,
but
that's
actually
the
previous
version
story
about
that.
So
so
we
used
to
get
the
beer
prefix
by
looking
at
the
source,
ip
address
and
then
somehow
knowing
the
subdomain
so
that
we
can
figure
out
which
b
friday
that
refers
to,
but
that's
kind
of
a
hack,
and
we
figure
later
that.
F
F
F
Okay,
this
is
just
from
that
pin
drop
that
shows
how
the
generic
mechanism
works,
so
there
is
a
restored
bit
indicating
that
this
mechanism
is
being
used
and
then
what
is
defined
as
the
auxiliary
data
section
in
in
igmp
version,
3
and
mld
version
2
rfcs
is
being
used
to
contain
this
extension.
F
F
F
F
You
can
fix
that,
if
you
like,
but
in
general
with
these
tlvs
there's
no
no
attempt
at
aligning.
So
even
if
this
one
was
aligned,
it
would
be
a
problem
if
the
previous
tlv
was
not
aligned
like
the
length
would
not
be
a
multiple
of
four
perhaps,
but
I
don't
see
a
reason
to
be
aligned
here
next
slide,
please
so
yeah
it's
a
little
bit
difficult.
F
B
A
Okay,
I'm
implementing
sorry
implemented.
You.
B
Have
hands
tool
there
we
go
yeah,
so
this
is
just
we
are.
This
is
the
beer
mld
dock?
Okay,
that's
what
I'll
call.
B
B
D
L
Just
like,
can
you
guys
hear
me
yeah,
we
can
hear
you
and
I'm
bringing
up
the
deck
here
and
I
can
hear
you
guys.
So
that's
a
good
sign,
yep.
G
G
G
So
what
do
you
see?
I
see
a
slide
number
one.
Okay,
thank
you
very
much.
So
just
a
little
background
why
this
draft
we
are
seeing
some
of
the
operators
they
want
to
upgrade
their
network
to
a
beer
infrastructure.
G
They
usually
start
with
upgrading
the
core
to
the
beer,
infrastructure
or
part
of
their
network,
and
basically
this
draft,
exactly
like
pim
signaling
over
beer
brings
a
mechanism
that
they
can
grab
a
part
of
the
network
and
replace
it
with
beer
without
doing
any
upgrade
on
the
other
part
of
the
network,
which
is
running
the
mldp
point
to
multi-point
tunnels.
G
G
Oh
yeah,
I
think,
there's
a
lag
okay.
So
basically
the
draft
is
very
simple:
it
uses
the
rfc
7060
that
rfc
has
been
used
to
tunnel
mldp
over
other
type
of
domains,
including
rsvpt
domain,
etc,
etc.
So
we
are
using
the
same
track
here.
Basically,
the
tldp
is
stitching.
If
you
will
the
two
access
mldp
domains
together,
we
need
to
find
the
ebbr
and
ibbr
just
a
refresher
for
the
folks
out
there.
G
The
ibbr,
the
ingress
pbr
is
the
router
that
is
closest
to
the
host
if
you
will
or
to
the
leaf-
and
this
is
exactly
the
same
as
the
pim
signaling
draft,
so
the
mechanism
to
find
the
ebbr
again
exactly
the
same
as
the
pim
signaling
draft.
There
are
multiple
mechanism
that
can
find
that
ebbr
router
when
the
ebbr
router
is
found
on
the
ibbr.
G
B
G
While
it's
loading,
I'm
just
gonna,
go
ahead.
So
from
the
datapath
perspective,
the
egress
boundary
b,
router
tracks
all
the
effects,
mldp
fix
and
it
actually
builds
a
table
if
you
will
of
all
the
ibbr
routers,
the
routers
that
are
beer,
boundary
routers
that
are
closest
to
the
leaf
that
are
interested
in
that
fake
and
it
will
actually
grab
that
signaling
information
and
send
it
up
stream
to
the
root
router.
G
And
when
the
traffic
comes
back,
the
ebbr
actually
encapsulates
the
traffic,
the
mldp,
the
the
mpls
traffic
over
a
beer
and
sets
all
the
bits
in
the
beer
header
appropriately
based
on
the
downstream
beer
boundary
routers
that
are
interested
and
it
shoots
out
the
the
traffic
over
the
actually.
I
think
sorry,
I
think
you
need
to
go
to
number
four.
O
G
G
Says
that
about
traffic
flow?
That's
the
good
aspect.
Thank
you.
Thank
you
very
much.
So
that's
the
slide.
I
just
explained
when
the
traffic
comes
back
from
the
route,
it
gets
encapsulated
into
the
beer
header
and
it
goes
to
all
the
egress
beer
forwarding
routers
over
beer
and
the
egress
routers,
remove
the
beer
header
and
just
swap
the
ldp
label
and
just
shoot
it
to
the
leaf
routers.
If
you
will
last
slide,
please.
G
Call
well
no,
and
that's
the
part
that
it
was
that's.
D
G
G
For
is,
if
there's
any
comments,
and
we
are
hoping
for
alaska
yeah.
This
is
the
wrong
version.
My
apologies
sandy
has
actually
uploaded
the
right
version.
G
Yes,
the
deck
there
is
a
newer
version
that.
B
I
just
pulled
it
off.
The
tools
did,
should
I
grab
another
deck
somewhere
or
are
we
cool.
B
G
B
P
I
don't
want
to
comment
for
this
draft.
I
just
want
to
say
the
materials
has
been
uploaded,
maybe
greg,
and
you
need
to
refresh
your
yeah.
P
B
B
Get
a
sense
of
the
room
to
see
where
we
are.
People
think
this
is
ready
for
last
call
and
we
can
take
to
the
list.
A
B
B
B
B
Q
Q
Okay,
so
the
ones
document
that
is
missing
from
this
list
is
the
recently
a
work
group
adopted
a
document
on
bfd,
and
I
really
appreciate
the
support
from
the
group
by
adopting
this
work
and
it's
progressing.
Q
What
I
wanted
to
bring
up
to
the
discussion
and
get
your
feedback
are
these
four
documents?
It
started
a
little
bit
backward.
We
started
working
on
performance
measurement
using
alternate
marking
method
and
thanks
to
alvaro
for
his
review
comments.
Q
We
worked
on
addressing
and
documents
posted
the
version
and
really
appreciate
the
feedback.
If
we
address
the
comments,
but
what's
important,
that
was
pointed
in
the
review
that
the
document
refers
to
the
requirements
document
and
it
doesn't
seem
very
logical
to
advance
the
solution
before
the
requirements
being
advanced
and
we
believe
that
oem
requirements
document
has
a
value
it
might
have.
Q
Well,
some
consider
it
to
be
only
just
as
a
historical
value
and
if
it's
informational
reference,
it
would
not
affect
progress
of
a
standard
track
to
my
understanding.
Q
So
there
is
a
this
way
forward
just
to
leave
this
document
oem
requirements
in
its
current
form
or
work
on
it.
If
there
are
any
comments,
recommendations
to
improve
it
and
then
in
the
meantime,
progress
the
solutions
to
publication.
Q
The
other
way
around
is
to
complete
the
work
on
oem
requirements,
the
publish
as
informational
document
and
then
as
it
progresses,
move
forward
with
their
solution.
Documents
such
as
listed
here
performance
measurement
using
alternate
marking
method,
which
is
a
hybrid
measurement
method
according
to
rfc7799
path,
mtu
discovery.
I
know
that
we
have
another
solution
and
we
discussed
it
and
we,
I
believe,
agreed
to
have
documents
separate,
not
to
merge
them
and
then
beer
ping.
Q
Q
So,
with
all
that
said,
I
looked
and
I
believe
that
requirements
in
performance
management
and
ping
they
do
have
volunteers,
who
agreed
to
be
shepherd
and
really
thank
you
guys
bath
until
you
discover,
is
still
looking
for
shepherd
and
if
somebody
volunteers,
I
promise,
I
will
owe
you
one
more
shepherd
document.
Q
All
right
thanks
greg,
yes,
and
actually
you
know
just
the
direction
so
what
we
we'll
do
with
the
documents,
how
we'll
progress
them
perfect.
O
H
H
So
for
the
blte
we
have
two
types
of
b
bit
positions:
one
type
of
p
position
is
for
equals
node,
so
they
call
the
local
d
cap
b
position.
Another
type
of
bit
position
is
for
adjacencies
for
forward
connected
adjacencies.
H
So
in
order
to
provide
faster
protection
for
any
transit
node,
so
we
just
extend
each
entry
in
the
normal
forwarding
table
with
a
fr
entry.
So
this
fr
entry
contains
a
flight.
We
call
the
fopa
which
is
sort
of
for
fr
protection
active
and
then
the
number
of
bicarbon
pass.
So
this
backup
path
is
for
the
when
the
one
one
maxwell
fails.
The
backup
pass
will
be
from
this
local
node
to
the
fairness
of
the
legacy,
like
hope.
H
H
So
here
we
have
a
more
details
through
example,
so
here
we
have
a
path
from
a
to
destination
d
and
h,
so
the
path
is
identified
with
the
green.
So
we
have
two
branches
in
this
path.
One
part
one
branch
is
from
a
to
g
to
h,
so
another
branch
is
from
a
to
b
to
c
and
to
d,
so
your
normal
operation,
so
that
any
package
with
this
path
will
send
it
to
these
two
branches.
H
So
that
means
from
along
the
upper
branch
from
a
to
g
to
h
and
then
the
practice
was
also
similar
to
along
the
second
branch,
which
is,
from
a
to
b,
to
c
m
to
d:
that's
normal
operation.
So
when
we
have
a
failure
of
a
transient
node,
for
example,
when
c
fails
in
this
case,
so
the
package
will
continue
send
along
the
upper
branch,
which
is
from
a
to
g
and
to
h,
but
for
the
second
branch.
H
So
when
node
b
detect
the
failure
of
c
the
flag
for
c
entry,
fr
entry
will
be
set
to
1..
So
in
this
case
the
fr
entry
will
be
used
to
forward
packages
to
around
the
phaser
load
c,
using
backup
paths
from
b
to
d.
So
here
by
composites,
from
b
to
g
to
h
and
to
d
before
sending
the
path
using
backup
paths,
the
node
b
will
remove
egress
bit
point
four,
which
is
for
for
h,
because
these
four
is
on
the
back
of
path.
H
But
this
four
is
loud
on
any
branch
plus
branch
starting
from
b.
So
we
need
to
remove
this
dispute
position
four
and
then
we
add
backup
bit
position
for
biker
bus.
So
with
this
modified
bit
string,
so
the
package
will
be
delivered
along
the
back
of
us
to
d
so
next
page.
H
So
this
is
the
updated
forwarding
procedure
for
support,
f,
b,
t
f
r,
so
for
microsoft,
node
n.
If
this
node
n
has
a
transient
node
of
a
path,
so
the
procedure
will
check
whether
this
node
n
is
failed.
If
node
n
is
failed,
so
the
node
b,
for
example,
b,
will
will
remove
b
position
for
the
adjacency
to
to
node
n,
and
then
it
will
remove
those
kind
of
eagle,
speed
positions.
If
that
bit
position
is
on
a
biker
path
and
then
also,
but
not
on
any
branches
from
from
plr.
H
So
after
removing
those
beat
position
for
each
node
and
then
the
prr
will
send
the
package
along
the
backer
path
to
the
right
leg:
scope
on
the
pass.
So
this
is
through
clean.
The
appeal
position
for
the
jc
from
n
to
n
slack,
hop
and
then
add
the
beta's
positions
for
the
biker
pass.
Then
we,
the
normal
procedure,
will
be
used
to
forward
the
package
along
the
back
of
path
to
the
leg,
stackers
hope
next
page.
H
So
this
this
pursuit
code
give
more
details
about
the
app
upgrades.
So
basically,
the
highlighted
part
is
the
part
for
supporting
the
bftfrs
next
page.
A
So
first
question:
why
I
understand
correctly:
this
is
updating
only
the
brte
path,
right,
40
pounds.
Yes,
so
my
comment
is
yeah
well
sold
out
and
since
we
have
to
study
the
architecture,
probably
necessary,
tolus
is
probably
the
best
judge
of
that.
Thanks
for
the
work.
B
B
I'll
just
queue
them
up
here
for
you
to
egress
patek.
H
H
So
basic
idea
is
that
first,
we
need
to
configure
the
bike
up
egress
for
protect
egress.
So
after
this
configuration
so
the
information
about
which
one
protector
which
one
will
be
distributed
to
the
director
adjacent
node.
So
those
adjacent
node
will
extend
the
forwarding
table
before
the
table
so
with
the
bicarb
entry
to
protect
primary
evos
node
using
backup
equals
node
and
then
in
the
normal
operation.
H
So
this
gives
more
details
about
the
extension
to
the
beautiful
ordering
table
or
equals
protection,
so
basically
for
a
node
x,
so
for
each
of
equals
node
connected
to
the
x.
So
we
just
extend
that
entry.
With
a
record
by
this
bike
happens,
you
will
have
a
flag
epa
which
is
standable,
equals
production,
active
and
then
the
backup
path
which
is
from
local
node
to
the
backup
equals
node.
H
So,
for
example,
here
we
have
the
extended
table
folding
table
on
node
c,
so
node
c
we
have
e
was
not
d,
so
d
is
protected
by
h,
so
for
the
forward
entry
with
lights
over
d
as
the
equals
node.
So
we
will
have
a
byte
path
from
c
to
this
bike
up
eagles
node
h,
so
backup
position
is
here
is:
is
there
so?
Similarly
for
node?
F?
H
If
load
f
is
the
use
node,
and
then
we
have
a
protective
node
for
node
f,
for
example,
is
the
e
and
then
we
have
a
backup
path
from
c
to
e,
so
your
normal
operation,
so
the
normal
part
of
this
forward
table
is
used
for
the
package.
So
when
note
onenote
as
the
equals
of
the
fails,
then
the
prr
will
use
the
backup
untree
to
forward
the
package
along
the
backup
path
to
the
by
copper
equals
node.
So
next
page
this
give
more
details
using
example.
H
For
example,
we
have
paths
from
from
a
to
equals
d
and
and
f
so
your
normal
operation,
so
the
package
will
be
divided
from
a
to
b
and
then
to
c
and
c
will
deliver
package
to
to
d
and
that
and
f
so
when
he
was
not
a
b.
Yes,
so
note
c
will
detect
this
failure
and
then
the
flag
epa
for
for
for
the
entry
d
will
be
set
to
one.
So
in
this
case
c
will
use
bicarbonate
entry
to
for
the
package
to
d
through
the
bike
up
pass.
H
The
bypass
is
from
c
to
h
and
then
the
package
will
send
it
from
c
to
h
and
then
h
will
deliver
the
particle
to
the
destination
ce
so
next
page.
So
this
gives
more
details
about
the
procedure
updated
for
supported,
equals
protection
so
basically
for
a
node
n
as
a
equals
node
in
the
package.
So
the
procedure
will
check
whether
this
node
n
failed.
H
If
this
node
n
failed
and
then
we
will
remove
the
old
path
from
the
bit
string
to
the
that
equals
node,
and
then
we
will
also
check
whether
this
by
half
equals
note
is
already
in
in
some
part
in
the
past.
So
if
it
is
already
in
the
past,
we
don't
do
anything
if
you
fight
over
use
those
beauty
positions
for
backup.
You
know
it's
not
in
the
past,
and
so
we
just
send
the
package
along
the
biker
path
to
this
backup
equals
node
next
page.
H
B
H
You're
on
so
this
is
about
the
pf
faster
route.
Nice
page.
H
So
I
would
like
to
thank
you
for
those
people
for
their
valuable
comments
and
suggestions,
jeff
and
the
daniel
and
the
shirsung
for
their
video
comments.
And
then
I
addressed
those
comments
and
I
made
the
updates
for
the
for
the
draft.
So
the
main
updates
is
that
we
change
the
status
of
a
draft
to
experimental
and
then
in
the
draft.
We
describe
it
in
details
about
the
difference
between
the
existing
existing
draft
and
our
draft,
and
then
we
add
some
text
talking
about
compressing.
H
So
the
difference
between
the
existing
one
and
the
our
draft,
so
the
existing
draft,
the
the
idea-
is
that
the
quotation
based
fr
so
basically
for
when
knight's
hope
fails.
So
the
prr
will
tunnel.
The
package
to
the
leg
stack
is
hulk.
H
H
H
So
before
this
new
computation
or
recomputation,
so
the
all
the
other
package,
some
of
all
the
aspect
packages,
maybe
get
lost
because
of
those
kind
of
inconsistency.
So
our
draft
resolved
these
issues.
So
the
idea
of
our
java
is
that
we
based
on
the
our
lf
fees.
So
when
next
fails,
we
use
the
rfa
by
carbonite's
hope.
In
addition
to
that,
we
use
multiple
14
tables.
We
call
that
fr
14
tables,
so
each
fr
4d
table
is
for
one
max
hop.
H
H
So
we're
also
talking
about
compressing
the
for
40
table.
So
before
confirmation
the
size
of
the
table
is
the
lambo
for
b
equals
equals
nodes.
So
after
the
comparison,
the
size
of
a
table
will
be
the
number
of
labels
minus
one.
So
after
the
compilation,
so
the
two
proposals
regarding
to
the
memory
use
will
be
in
the
same
level.
So
some
sometimes
this
one
will
use
more
and
then
sometimes
another
one
is
more.
A
So
as
a
workgroup
participant
no
chair
well,
I
originally
think
much
of
the
whole
idea
because
no
pretty
much
igp
protecting
the
fr,
but
I
see
the
value
of
that
I
mean
so.
Your
drop
is
pretty
much
formalizing.
What
igp
is
doing,
but
it
goes
natively
into
beer
and
yes,
yes,
value.
A
B
A
H
As
a
composer,
okay
and
then
we
can
go
because
this
is
disappointing.
B
All
right,
let's
turn
that
off
I'm
going
to
start
a
sort
of
session
there.
It
is
all
right-
and
this.
B
A
So
we
have
42.
Okay,
please
comment
on
the
mind.
People
have
two
hands
and
I
think
they're
confused
is
that
no,
you
cannot
push
both
buttons.
R
Okay,
can
I
go
yeah
go
for
it,
yeah.
Q
My
understanding
is
that
beer
that
one
of
the
idea
of
beer
is
that
does
not
create
a
state
in
a
transit
node.
Fr
requires
a
state
in
a
transit
node.
So
no.
Q
H
Q
You
cannot
protect
without
knowing
your
next
hop,
so
you're,
protecting
for
the
specific
flows
that
are
controlled
from
the
dfir
and.
H
Yeah
we
have
can
extended
forwarding
table,
we
call
the
fr
folding
table.
We
only
have
these
four
new
tables,
it's
a
limited
size,
and
then
we
don't
have
a
performance
space.
A
Q
G
Yes,
please
can
I
just
ask
a
quick
question
here
just
to
ensure
that
the
I'm
in
sync
my
understanding,
so
I
just
want
to
understand
you.
The
proposal
here
is
just
to
use
the
igp
won't
reroute.
Nothing
else
right.
H
Yes,
we
you
see,
we
have
a
fr
forging
table
for
each
lex,
hop
because
f
of
ipfr
using
ipr
ip
folding
table
here,
use
beer
folding
table,
so
that's
in
different
different
folding
tables.
So
for
beer
we
need
a
special
enhancement
for
the
bf40
table
to
provide
faster.
N
The
beer
taste
check
is
kind
of
the
the
stateless
check
right
that
something
does
does
something
really
smell
like
beer
is.
Does
the
state
in
the
transit
hop
changes
when
a
sender
becomes
active,
goes
offline
or
receivers
appear
or
go
away
right?
Does
that
change
the
state
on
the
transit
hops
and
I
think
when
it
does
not,
then
it
tastes
and
smells
like
beer
and
that's
why
this
tastes
and
smells
like
beer.
A
A
A
G
And-
and
I'm
I'm
good
with
all
the
my
entire
point
here-
is
that
it
would
be
beneficial
if
you
agree
on
one
type
of
fast
reroute
protection
right,
so
I
mean
in
igp
or
any
type
of
other
technology.
You
have
ecmp
as
one
type
of
protection
and
then
you
have
bypass
or
fast
rer.
That's
another
type
of
protection.
G
G
A
A
B
A
B
L
B
Okay,
gotcha,
so
chad,
I'm
trying
to
bring
up
the
next
egress
protect
draft,
your
presentation
and
it's
not
letting
me
share
it.
So
if
you
can
grab
it
yourself
from
the
tool
or
if
tony
can
bring
it
up
that
we
can
get
this
next
deck
going.
H
Page
so
we
made
some
updates
to
the
previous
version,
so.
H
So
we
just
more
details
about
differences
from
the
existing
one
and
some
identical
changes.
So
next
page.
H
So
here
the
difference
that
our
draft
just
for
provider
equals
production.
We
just
use
one
pass
to
from
source
to
eagle's
node,
but
for
existing
mofr
that
use
two
disjoint
paths
from
source
to
each
of
the
eagles
note,
and
then
the
eagles
node
will
select
one
of
the
traffic
because
two
same
traffic
to
from
source
to
the
same
destination,
so
the
destination
will
select
one
of
them.
So
in
this
sense,
so
our
equal
protection
will
use
less
network
resource
than
the
existing
one.
H
So
for
this
one,
I
think
we
will
welcome
more
comments
because
adoption,
because
we'll
have
more
options,
for
example,
jeffrey
proposed
another
solutions
for
equal
protection,
and
then
we
need
more
discussions,
maybe
merge
or
whatever
next
step.
Thank
you.
A
Yeah
from
my
side
sure,
I
think
you
should
have
it
together
with
jeffrey,
and
you
guys
should
talk
it
out
whether
you
want
one
single
draft
and
then
explaining
rest
protection
for
in
all
the
different
cases
like
helper,
and
this
case
you
know
and
whatever
you
have,
and
possibly
even.
D
A
P
P
This
draft
analysis,
the
issue
in
bureau
domain
and
in
this
version,
modi's
description
is
added
and
such
as
the
issue
analysis,
the
node
failure,
detection,
the
working
path,
failure
technician
the
following:
has
pages-
has
been
presented
in
several
meetings
such
as
1
0,
8
meeting.
So
I
will
skip
them
very
quickly.
P
P
P
B
Yeah
I'll
start
a
vote
on
this
where'd
it
go
yeah
there.
D
D
H
You
wonder
in
the
draft
you
mentioned
that
you
have
a
pfd
for
the
titan
failure
of
the
node
inside
the
via
domain.
Why
do
you
need
that
for
english
production.
P
H
The
you
describe
a
lot
of
things
for
detecting
the
failure
inside
the
via
domain,
so.
Q
Yes,
yes,
so
because
there
is
a
redundancy
in
bfir,
so
point
to
multipoint.
Bfd
can
be
used
by
bfers
to
detect
the
failure
of
their
active
source
and
then
change
it.
Q
So
if
you
can
suggest
any
other
way
to
detect
the
failure,
there
will
be
much
appreciated,
but
we
don't
have
any
other
itf
standard
mechanism
to
detect
the
failure
in
a
multicast
rather
than
point
to
multi-point
bfd.
H
Q
It's
the
same,
it's
the
same
because
for
the
bfer,
the
failure
is
indistinguishable.
Q
What
the
bfr
detects
is
the
segment
of
the
multicast
distribution
tree
is
not
working
so
because
how
point
to
multipoint
bfd
works
is
that
the
root
sense
periodically
control
packets
so
that
the
leaf
or
tell
in
terms
of
point-to-month
to
point
b
of
d
knows
from
the
previously
received
packet
and
calculates
their
failure.
Detection
interval.
Q
So
if
there
is
no
control
packets
received
with
the
control
packets
received
within
this
interval,
the
lead
declares
the
network
failure
and
that
equally
applies
either
to
failure
of
their
root
or
the
failure
of
their
distribution
path.
So
there
is
no
need
for
their
egress
to
differentiate
two
types,
because
if
there
is
no
repair-
because
in
this
use
case,
there
is
alternative
source
of
their
data
traffic.
Q
So
for
the
leaf,
it's
a
simple
to
switch
to
the
alternative
source.
H
Answer
a
question:
no,
I
I
mean
you,
I
think
you
have
description
about,
detect
failure
of
a
node
inside
the
be
a
domain.
I
think
that
may
be
not
unnecessary
right.
You
can
determine
the
failure.
H
Q
Why,
if
it's
a
transit,
node
essential
to
distributing
particular
flow,
we're
detecting
it
and
we
are
not
differentiating
again,
I
want
to
stress:
we
are
not
differentiating
the
failure
of
the
root,
the
head
of
the
multicast
tree
from
the
failure
of
any
segment
or
any
transit
node
from
the
standpoint
of
bfr.
Q
So
there
is
no
need
to
dis
differentiate
this
types
of
failures.
So
what
detection
is
that
it
stopped
receiving
for
certain
interval,
vfd
control,
packets
and
that's
interpreted
as
a
failure
of
their
distribution
tree
and
trigger
some
switch
over.
Q
B
Thank
you.
Thank
you
sandy
all
right.
That
is
all
we
have
on
the
agenda.
We
are
already
12
minutes
over
and
no
one
has
cut
us
off.
This
is
pretty
exciting.
Blue
sheets
don't
exist
because
you've
already
signed
up,
it's
been
done.
We
got
jeffrey
and
then
on
to
tony.
I
believe
they
finished
up
the
notes.
Was
it
tony
got
it
all
right,
we'll
see
that
in
email?
Any
other
comments
before
we
move
on
and
you
go
off
to
your
lovely
day
here
in
prague.