►
From YouTube: IETF100-MBONED_PIM-20171114-0930
Description
MBONED PIM meeting session at IETF100
2017/11/14 0930
https://datatracker.ietf.org/meeting/100/proceedings/
C
A
A
B
A
B
C
D
E
D
A
F
G
A
H
A
H
So
we
revived
the
draft
and
in
between
the
you
can
go
to
the
next
slide,
show
but
Oh
actually
yeah.
Let
me
let
me
step
back
and
before
I
get
into
that.
So
just
a
reminder
of
why
we're
doing
this
work
here
in
EMBO
Andy
is
this:
is
the
perfect
place
to
do
this
kind
of
work
because,
as
the
M
Bodi
Charter
says,
is
that
we
should
receive
regular
reports
on
the
current
state
of
deployment,
multicast
technology
and
create
practice
and
experience
documents
and
provide
feedback
to
other
relevant
working
groups?
H
So
we
feel
that
this
topic
is
a
good
thing
to
do
here
in
Ambon
d,
and
my
question
that
I'll
ask
at
the
end
of
my
time
here
is
that
we
should
adopt
this
draft
into
this
working
group,
and
so
in
the
when
we
get
to
the
next
slide.
I
can
talk
about
some
of
the
things
that
was
just
which
is
fine,
so
in
between
the
last
two
years
Charlie
my
co-author.
He
also
wrote
a
similar
draft.
He
proposed
it
to
the
into
area
working
group.
H
He
described
the
interior,
which
I'm
not
involved
in
as
kind
of
a
catch-all
for
these
kind
of
things,
and
he
and
I
discussed
now
that
we've
had
renewed
interest.
That
M
bode
is
the
appropriate
place
that
we
feel
that
this
should
be
done
in.
So
this
is
the
multicast
efforts,
experts
we're
not
the
Wi-Fi
experts,
but
we
are
multicast
effort.
H
So
that's
why
we
feel
that
it
should
be
done
here
and
he
in
fact,
in
this
other
draft
that
he
wrote,
is
asking
the
author's
to
take
that
information
and
put
it
into
the
draft
that
we
have
here.
So
just
a
reminder.
Some
of
the
issues
Michael
may
know
some
other
ones
that
I
may
have
forgotten
here,
but
there's
issues
with
using
multicast
over
Wi-Fi
and
part
of
it
is
that
there's
this
low
bandwidth
over
802
to
11
there's
increased
congestion,
which
causes
a
high
packet
error
rate
for
multicast
streams.
H
These
six
neighbor
discovery
usually
easily
saturates,
the
Wi-Fi
link,
because
it
uses
multicast
and
apps
like
bonjour,
for
that
does
a
service
discovery
for
things
like
printing
services
can
saturate
the
link
using
multicast
and
it
kind
of
was
agreed,
and
these
in
the
last
meeting
that
we
had
that
these
problems
will
not
be
fixed
any
time
soon.
Yeah
one
sec.
I
J
So
the
problem
is
that
it's
sent
that
the
lowest
common
denominator
right
asked
to
everybody.
So
typically,
even
if
you
have
like
1,300
megabits
of
unicast,
this
will
be
sent
at
like
2:00
and
everybody's
listening
to
it.
So
then
that
speed
is
not
really
going
up
that
lowest
speed
these
some
ways
on
the
edge.
This
will
be
going
at
that
rate,
so
it
can
be
one
megabit
than
it
has
been
for
Loza
right.
I
I
J
Is
another
issue
also:
is
that
so
let's
do
this,
why
this
problem
is
so
bad
that
a
lot
of
mobile
devices
actually
do
not
listen
to
these
packets
at
all,
when
they're
sleeping,
then
they
just
in
Hardware
just
profit,
they
don't
even
wake
up
the
CPU,
so
this
is
a
there.
So
there's
a
problem
where,
on
some
of
phones,
the,
if
you're,
if
you
have
anything
that
you
really
need
like
push
or
pull
over
v6,
it
just
stops
working
completely.
Ok,
but
that
is
that
is
the
fourth
problem.
H
I
J
H
L
M
A
K
Quick
push
with
my
hair,
so
I
didn't
realize
that
a
bandit
was
a
problem
but
yeah
there's
a
problem
that,
like
I,
found
whatever
it
doesn't
really
want
to
be
woken
up
every
single
time.
Someone
sensor
neighbor
discovery
packet,
so
what
they
often
do
is
to
not
listen
to
all
of
them,
and
then
you
drop
them.
The
other
problem
is
I,
think
the
reliability
you're
quite
likely
to
lose
some
neighbor
discovery,
packets,
which
is
pretty
bad.
D
N
So
yeah
Warren
quarry
kinda
related
to
the
last
thing
who
remembers
the
Prague
meeting,
not
the
last
one,
the
one
before
when,
although
I
Alice
was
kind
of
sucky,
a
significant
amount
of
the
reason
for
that
was
just
simply
there's
a
lot
of
our
traffic
and
you
know
it
turns
into
multicast
and
it
saturates,
the
link
so
yeah.
It
all
goes
at
the
lowest
possible
speed
which
makes
everything
sucky.
You
know
this
is
actually
a
problem
which
we
see
in
the
real
world
as
well.
K
O
Meddling
trouble
hearing
you
say
that
again,
Stig
Stig
was
Stig
was
wondering
why
the
minimum
speed
can't
go
up
joleigh
eglee,
so
it
has
over
time,
but
it's
several
orders
of
magnitude
slower
than
the
maximum
speed,
which
means
the
time
domain
in
which
the
minimum
speed
window
occurs
is
several
orders
of
magnitude
longer
than
say.
The
transmission
for
the
same
packet
at
a
different
rate
so,
like
other
things,
become
unavailable
to
you
like,
for
example,
beamforming
and
certain
kinds
of
time
domain.
O
Shifting
that
you
do
at
the
higher
rates
that
you
have
available,
so
this
is
actually
driven
by
backward
compatibility
with
the
minimum
speed
device
that
is
or
can
be
attached
to
than
that
particular
network
segment.
So
over
time
that
has
actually
gotten
faster.
But
it's
like
I
said
in
some
cases
three
orders
of
magnitude
difference
in
performance.
H
Okay,
thank
you
so
yeah,
let's
bounce
on
yeah.
So
let's,
let's
do
so
Moodle
x1!
So
again,
with
all
those
comments
again,
it's
our
feeling
that
it's
important
to
gather
that
information,
not
that
we're
gonna
solve
it,
but
it's
important
for
the
community
to
understand
the
problem.
So
there
are
some
workarounds.
This
is
a
problem
statement
draft,
not
a
solution,
strap.
We
felt
it
would
be
helpful
to
at
least
show
some
areas
where
there
have
been
some
workarounds.
H
Basically
the
notes
from
the
meeting,
and
some
of
those
comments
were
that
we
need
to
determine
what
the
problems
should
be
solved
by
the
ITF
and
what
problems
should
be
solved
by
the
I
Triple
E
as
much
as
we
can
in
a
somewhat
high
level.
Maybe
ITF
decides
that
broadcast.
Is
this
more
expensive,
so
multicast
should
be
sent
over
wired
I
mentioned
kind
of
added
a
class
of
service
to
multicast
packets.
H
H
There
is
a
mailing
list,
that's
not
being
used
very
much.
It
is
right
after
these
meetings,
but
then
it
just
kind
of
goes
quiet
because
a
lot
of
people
lose
interest
in
this,
but
there's
a
mailing
list.
You
can
join
to
discuss
this
and
it'd
be
interesting
to
see
if
this
group
full
of
multicast
experts
feel
that
this
is
an
area
that
we
should
be
working
on.
If
it
is,
the
authors
have
committed
to
actually
not
letting
I
go
and
to
actually
be
actively
working
on
it.
I
Quick
tour
lists,
quick
tour
lists,
we're
okay,
yeah
I
mean
I,
would
love
to
see
that
maybe
the
the
scope
is
not
kind
of
that
limited
because
you
mentioned
you
know
a
bunch
of
workarounds
impossible
ideas,
and
you
know
if
it's
just
you
know
kind
of
a
problem
statement
and
then
it
might
be
limiting
us
in
doing
you
know
the
text.
You
know
outlining
possible
ways
to
solve
the
problems
or
even
in
case
it's
just
you
know,
simple
attempts
to
provide.
I
D
A
A
H
A
That,
as
an
eye-opener,
I
mean
all
the
focus
we're
talking
about
here
is
really
with
the
most
part
is
the
fact
that
multicast
is
not
acknowledged
and
doesn't
get
retransmitted,
so
you
get
a
lost
problem,
we're
gonna
smogs
on
lost,
you
know
tolerance,
but
that
has
nothing
why
fie
that's
outside
the
scope?
That's
a
multicast
thing,
regardless
Wi-Fi
just
a
place
where
we
don't
get
that
transmission
acknowledgement.
The
point
that
came
out
to
me
that
I
was
not
prepared
for
at
all.
In
that
conversation
was
power.
A
We
had
people
from
Google
there
with
the
phone
we
have
people
Apple,
they
were
the
phone.
All
of
them
were
said.
No,
this
just
has
suck
in
our
battery-life
down
and
it
wasn't
because
of
multicast.
It
was
because
the
applications
that
use
the
multicast
so
it's
kind
of
like
we
have
two-tiered
problems
here
we
have
just
the
the
lossy
problem
with
lack
of
retransmission,
and
then
we
have
the
way
the
applications
have
to
just
always
be
listening
to
stuff
and
being
staying
alive
and
killing
the
battery
one
so
unfortunate.
A
P
Tyler
there's
some
crossover
here
with
some
of
the
work
that
MANET
has
reach
arted.
To
do
so.
Many's
got
some
kind
of
Charter
item
to
look
at
multicast
frameworks
for
meshing,
radio
systems
more
general
than
802
11,
but
everyone's
using
802
11
is
because
you
can
buy
them
rather
than
complicated
kit
I'm,
not
sure
how
much
traction
that
is
getting
in
MANET
and
I
would
suggest
that
you
guys
seem
to
be
better
organized
about
this.
Q
A
J
A
J
J
J
J
You
know
probably
so
yeah
so
this
is
so
Tim
had
a
whirl
at
this
document
and
and
and
changed
part
of
it
big
due
to
those
that
feedback.
So,
but
it's
all
at
least
for
me
and
for
a
lot
of
others,
what
does
it
mean
to
say
we
deprecate
something.
So
this
is
I.
Don't
know
we
don't
know
if
this
is
the
right
way
of
doing
it.
I
think
we
all
know
what
we
want
to
do,
but
like
what?
J
So
this
comes
from
simplifying
operational
practices
of
the
I
am
personally
myself
done
nature
domain
ASM
it
broke
every
two
years
and
yeah
just
fragile.
So
then
it
seems
I'm,
not
the
only
one,
so
yeah
the
the
more
people
want
to
duplicate
it
and
I.
Guess
to
tell
people
not
don't
do
this
if
you
can
avoid
it
and
it's
like
yes,
MSD
P
still
needed
for
anycast
RP.
So
why
not
want
to
deprecated
MSD
P
by
itself.
J
Okay
next
year,
so
so
we
Tim
cut
down
some
of
the
text
on
the
models
and
they
did
a
new
new
section.
We're
talking
about
these
different
parts
of
what
mechanisms
are
used,
phrase,
MSM
and
so
on,
and
describe
that
and
also
the
way
of
supporting
an
ASM
application
so
that
it
actually
does
and
SSM
join
from
the
edge.
If
this
is
a
way
of
supporting
older
applications,
for
instance
linear
TV
next
slide.
J
J
K
K
G
J
And-
and
it's
like
it's,
this
is
the
same
thing
to
a
lot
of
other
of
these
mechanisms.
It's
like,
we
don't
tell
people,
you
know,
filter
it.
You
cannot
do
it
blah
blah
blah,
and
so
that's
I
think
what
we're
going
to
do
is
have
a
strong
recommendation
and
back
it
up
by
it
by
advice,
I,
think,
I,
don't
know.
If
that
there
is
anyone
who
wants
to
do
anything
else,
but
yeah.
There
would
be
good
good
to
hear.
J
So
this
is
also
about
for
application
developers,
because
I
think
that
if
you
come
in,
you
want
to
develop
a
new
application.
They
had
the
whole
thing
about
having
a
SM
and
do
a
circle.
Majeed
Owens.
It
might
seem
right,
but
they
don't
understand
the
implications
of
doing
that
at
scale,
because,
if
you're,
just
looking
at
it
from
the
edge,
you
might
get
a
different
idea.
So.
A
Let
me
throw
a
question
out
there,
just
great
Shepherd
contributor
here,
it's
nice
to
think
that
application
vendors
will
look
at
this
for
guidance
when
have
they
ever
I
mean
says
it's
tragedy.
The
ietf
right
I
mean
we
have
people
who
just
randomly
pulled.
Class
D
addresses
out
of
their
tail
and
embed
them
in
devices
without
any
concept
of
how
it's
going
to
impact
the
network
or
direction
from
the
ITF.
So
this
is
kind
of
a
question
about
layer.
J
A
A
J
A
I
A
K
A
G
A
I
Sorry,
yep,
okay,
something
for
your.
You
know
sleepless
nights
to
read
up
so
after
Prague
we
basically
did
some
you
changes
to
the
document.
Then
it
went
to
iesg
review
and
we
got
a
humongous
amount
of
discusses
back,
at
which
point
in
time,
I
started
to
move
it
over
into
XML
and
github
so
that
it
can
easily
be
added
and
also,
hopefully,
you
know,
come
out
through
RFC
editor
and
then
I
started
doing
the
the
changes
listed
in
the
following
slides
next
slide.
So
that's
the
differ.
There
were
a
bunch
of
small
things.
I
I
There
is
now
obviously
a
section
about
congestion
control
because
that's
part
of
the
standard,
iesg
review
and
there's
also
pointing
to
be
CP
145
that
you
know
just
came
out
last
year,
which
has
all
the
multicast
congestion
control,
put
an
example
kind
of
what's
the
worst
typical
case.
What
happens
right?
You
have
something:
that's
controlled,
multicast
in
elastic
you're,
sending
it
to
another
ad
which
autonomic
domain.
Where
it's
you
know,
invest
effort
and
everything
goes
down.
I
I
Public
peering
editor
section
because
they
were
kind
of
you
know
on
and
off.
Reference
is
two
eyes
if
it's
not
a
private
private
peering,
so
I
described
the
problem.
You
know
the
you
have
a
land
between
more
than
two
a
SS
you're
running
multicast
over
it
and
you're
running
in
all
to
all
these
wonderful
problems
with
FEMA
cert
describe
the
three
possible
solution
cases
the
management
interactions.
So
I
try
to
introduce
a
clear
description
of
what
really
you
know
is
G
always
comes
back
with.
Ultimately,
what
ends
up
to
be
about
privacy?
I
What
are
you
revealing
about
the
customer,
and
so
the
model
that
were
adopting
here
in
the
draft
is
that
ultimately,
the
as1,
which
has
the
multicast
source,
is
one
that
is
also
responsible
for
your
content
relationship,
meaning
that
the
fact
that
it
knows
about
your
multicast
traffic
is
just
a
small
subset
of
what
it
knows
about
you
from
you
know,
being
the
content
source.
That
way,
you
know
the
problem
is
pushed
back
next
slide.
I
Troubleshooting
M
trace
trace
route
yeah.
The
one
thing
that
basically
had
to
add
there
is
that
the
most
important
troubleshooting
tool
M
trace,
doesn't
work
through
AMT
tunnels,
so
that
might
be
actually
something
of
interest
for
us
to
or
pin
to
fix,
because
that
really
is
that,
then
they
said
we
need
to
rewrite
the
security
considerations
section.
So
now
there
is
a
long
security
consideration
stuff.
I
One
big
bullet
point
is
dos
attacks,
so
that's
kind
of
creating
state
or
pulling
in
too
much
traffic
describing
these
problems
and
how
they
can
be
solved
next
slide:
the
content
security.
So
that's
discussing
the
alternative.
You
know
free-to-air
content
versus
content
that
has
DRM
and
then
the
main
challenge
that
we've
seen
in
service
providers
that
are
deploying
intradomain,
for
example,
traffic
filtering
for
the
multicast
traffic,
so
that,
even
though
it's
encrypted
you
can't
even
get
a
copy,
which
you
know
you
could
easily
decrypt.
I
So
that's
kind
of
a
management
issue
between
the
two,
a
SS
in
that
case
peering
encryption
details
about
the
operational
aspects,
they're,
also
talking
about
how
the
exchange
of
out-of-band
information,
like
you
know
the
s,
comma
G,
States
program,
channel
information,
how
they
can
be
encrypted
but
mean,
for
example,
just
passing
it
through
the
peering
point.
So
those
were
explicit
asked
by
the
IES
g2
to
comment
on
those
operational
challenges
next
slide.
So
of
course,
there
can
be
no
document
today,
anymore
without
privacy
considerations
and
the
previously
consideration
in
this
case,
I.
I
Think
for
inter
domain
multicast
peering
are
fairly
small.
So
the
main
challenge
really
is
a
s2
who
didn't
have
a
Content
relationship
with
you
can
see
what
s
comedies
you're
receiving
and
from
that
they
can
deduce
to
a
good
degree.
You
know
what
content
you're
consuming
and
that
is
you
know
the
whole
extent
of
change
of
privacy
and
yeah.
You
may
want
to
use
the
one
model
where
you're
a
MT
tunnel
across
a
s2.
I
A
A
S
The
first
page
is
still
the
motivation
of
this
job.
We
want
to
stand
a
high
level
to
leverages,
existed
separately
and
multicast
model
to
control
the
multicast
network.
Next
and
the
second
is
still
the
content
of
this
job.
We
provide
a
human
readability
of
the
whole
multicast
network
by
UML
class
diagram
and
also
we
provide
a
young
model
for
the
for
the
technology,
so
you
can
use
it
directly
and
you
can
diagnose
your
network
by
it
directly
and
the
updater
of
this
version.
S
We
had
operated
technology
in
the
UML
diagram,
as
has
been
missed
in
previous
version,
and
also
we
at
BRT
in
multicast,
the
transport
layer,
and
also
we
make
some
diversion
to
the
model
and
attracted
to
see
this
model
has
been
verified
in
open,
daylight,
their
projection
and
the
project
has
been
released
in
carbon
worship,
so
this
young
model
is
available
and
the
practical
this
page
is
a
brief
introduction
of
open
daylight
from
it.
You
can
see
that
open
the
right
project
is
driving
by
several
models,
so
this
is
the
pure
project
information
daylight
from
it.
S
S
So
the
doTERRA
update
of
the
diagram
and
in
the
overlay
layer,
we
ADA
the
overlay
part
of
this
ethnicity
in
traverse,
right,
previous
worship
and
also
in
trance
part.
We
add
PRT
next
and
the
little
more.
The
overview
is
a
remain
unchanged.
Now
it's
still
and
be
comprised
by
three
levels:
overlay,
a
transporter,
an
underlay
and
the
motive
has.
The
information
is
still
the
key
of
the
young
model,
as
it
remain
unchanged
and
the
in
overlay
layer,
via
the
PDP
and
mr
d
snooping
in
the
overlay
technology
and
also
in
transport.
S
P
Rick
Taylor
and
it's
a
sort
of
young
based
question
I
was
sat
in
one
of
the
I
think
it's
routing
area
meeting
first
thing
yesterday,
where
they
were
introducing
the
schema
mount
points.
It
occurs
to
me
that
when
you've
got
your
underlay
specific
I
can't
remember
the
exact
words,
but
your
your
underlay
points
for
BGP
or
Babel.
P
Yeah
that
slide
would
that
be
a
good
candidate
to
actually
have
a
schema
mount
point,
so
you
wouldn't
have
to
specify
a
particular
node
for
each
of
those
types
of
routing
protocols.
You
could
explicitly
say:
there's
a
mint
point
for
your
routing
routing
protocol,
specific
bits
and
leave
that
bit
of
the
yang
model
to
the
the
relevant
area,
meaning
less
work
for
you
just
a
question:
are
you
following
that
work?
The
main
points
key
moment.
S
So
maybe
I
should
edit
more
get
hair
about
it
because
you
know
SPF
or
EC,
as
we
know
that
it's
the
underlying
protocol
of
a
beer
Bureau
as
well
of
the
motive,
has
the
technology
and
also
PGP,
is
under
discussion
for
the
multicast
delivery,
information
delivery
and
also
PIM
as
original
the
the
technology.
So
as
it
has
been
here,
it
must
be
be
here,
so
the
Bible
is
the
newest
under
a
protocol
to
deploy
in
Hamlet
or
other
slow
Network.
So
so,
yes,
I,
can
add
more
detail
about
it.
I.
P
G
S
S
P
P
S
S
A
N
A
S
A
A
P
A
Right,
so
they,
this
is
the
kind
of
direction
this
groups
got
to
start
providing
anyway,
I
mean
clearly
at
some
point.
We
have
to
adopt
this
work
if
this,
if
not
this
something
like
this
and
we've
got
energy
in
traction,
so
some
sense
of
the
room
who's
read
it
pieces
of
it
versions
of
it.
Okay,
who
believes
who
should
adopt
this
work
for
the
group,
all
right
who
thinks
we
should
kick
this
to
the
curb
it's.
A
Right,
like
I,
said
this
work
has
to
come,
has
to
take
place
somewhere
anyway,
and
we
I
think
it
there's
nowhere
else.
We're
gonna
find
multicast
people
in
the
IETF
consolidated
in
this
scale.
We've
got
here.
This
is
where
this
work
should
take
place
and,
like
I
say,
even
if
it
doesn't
go
off
like
this,
it
will
evolve
with
some
support
from
the
yen
community
and
we'll
have
something
that
we
can
publish
all
right
next
or
next.
Aren't
you
next.
A
M
U
D
U
I'm
on
the
present
here
as
well,
so
the
first
slide
is
about
a
very
briefly
review.
How
MSTP
works
in
this
diagram?
We
have
six
RPS
ignored.
Other
routers
are
just
heartbeats
here.
There
is
a
first
hop
router
on
the
upper
right
corner,
sending
the
pin
register
messages
towards
our
p1,
and
then
up
here
on
were
sent
MSD
PSA
messages
throughout
the
network
to
other
peers
on
the
lower
right
corner.
We
have
a
last
hop
router
or
sending
a
starchy
join
and
it
arrives
at
our
p6.
A
U
Mep
n,
a
source
active,
is
very
similar
here.
I
have
those
four
PS
I
have
a
customer
RP
in
middle
of
the
diagram
and
and
I
have
a
CEO
outer
at
the
bottom
again.
I
have
first
hop
router
sending
as
she
joins
towards
the
customer
RP
at
the
top.
In
that
customer
RP
has
a
MSP
session
with
PE
one
and
because
of
once
the
p1
learns
that's
a
source
through
the
MSB
is
a
messages
message.
U
You
will
send
the
MSM
VPN
source
active
yeah
and
that
will
all
be
received
that
will
be
received
by
all
the
four
PS
and
notice
that
here
the
Cee
is
not
a
CRP
here.
The
Cee
device
gets
that
is
connected.
Ps3
that
gets
a
start
is
join
from
Mass
hop
router.
It
will
continue
to
send
a
star
Z
going
towards
the
RP
towards
the
ps3,
because
P
3
is
the
C's
upstream
router
towards
that
some
customer
RP.
U
Here
the
ps3
has
the
star
G
States
and
also
has
the
in
learns
the
source
because
of
the
MEP
and
si
routes,
then
it
will
send
MVP
and
join,
which
is
a
basic
MEP
America's
routes
towards
the
PE
2,
which
is
the
measure
out
our
upstream
router
for
this
particular
source.
So
you
can
see
that
some,
the
MS
MVP
interceptive
route
and
msg
vsauce
active
rows
are
very
similar
here,
the
Pease
I
just
liked,
and
our
P's
and
and
AHS.
U
U
So
a
a
a
PE,
it
will
advertise
MVPs
a
when
he
learns
the
source,
so
the
MSBs
in
messages,
but
it
will
not
do
the
opposite
direction.
So
we
received
the
MVP
NSA
routes.
It
will
not
convert
it
into
MST
PSA
routes.
So
in
this,
in
this
example,
imagine
that
the
the
see
at
the
botton
instead
of
regular,
see
we
have
a
customer
pete
shoe
there
and
then
ps3
and
CRP
to
the
between
the
ps3
and
CRP
to
there
is
the
MSTP
session
there,
because
that
CRP
to
there
it
will
terminate
that
starchy
join
there.
U
You
will
not
send
a
star
to
join
Pinscher
in
towards
the
ps3
anymore,
so
p31
when
learned
the
source
through
the
MVP
and
si
routes,
but
it
does.
Northern
the
star,
G
join
and
ps3
does
not.
Edward
has
MST
PSA
towards
the
CRP
and,
as
a
result,
CRP
does
not
run
the
source
and
ps3
doesn't
learn,
does
not
learn
the
starchy
join.
So
here
then
nobody
was
in
this
as
she
joined
anymore.
U
So
this
I
believe
this
is
just
pure
oversight
in
the
MVP
and
protocol.
It's
not
specified
there
now.
Because
of
that,
then
we
need
to
have
MS
DDP
session
either
between
P
1
or
P
3,
so
that
so
we
can.
We
can
have
a
continued
continuous
MSB
session
from
CRP
1,
all
the
way
towards
CRP
2,
or
we
can
have
MSB
session
between
p1
and
CRP
2
or
between
CRP
well
and
three.
Once
you
have
that,
then
you
have
the
continuous
MCB
session,
then
things
will
work
towards.
I
U
U
I
Obviously,
you
know
the
problem
exists
there
as
well.
That
should
be
what
we
would
be
primarily
concerned
with
and
sure
fix
it
for
him
MSTP
as
well.
I
hadn't,
you
know
remembered
all
these
MVP
n
specs
that
you
know
whenever
we're
writing
something
about
MSTP
I'm,
mostly
worried
about
what
stages
the
resulting
RFC
has
right.
If
kind
of
we
can
really
write
something
standards
track
mentioning
MSTP
right.
I
V
U
Solution
here,
at
least
in
MSD
be
case,
is
simple,
with
basic
is
specified
that
when
the
ps3
gets
the
Mme
PSA
and
it
will
treat
it
as
if
it's
received
a
MSD
PSA
messages,
so
that's
it
will
advertise
MSG
PSA
towards
a
CRP
too.
Once
that
happens,
you
can
see
that
the
CRP
two
now
learns
the
the
source,
the
sources
and
it
has
a
star
GGO,
and
then
he
was
in
PMS
he's
doing
towards
ps3,
and
everything
was
that
work.
In
addition
to
that
today,
the
MVPs,
a
Ross,
does
not
care,
carry
the
harpy
information.
U
U
I
A
K
U
U
K
That's
my
thinking,
I
think
it
might
be
done
more
invest,
but
it's
great
that
people
here
are
involved
and
I
guess
one
question
I
had
to
is:
is
it
a
common
problem
with
customers
that
they
check
they
face
this,
or
is
that
problem
to
move
our
piece
to
to
the
PE
or
instead
we've
kind
of
solve
this
problem
or
B
start
something?
This
script
would
know
maybe
more
about
the
deployment
challenges.
U
U
K
U
I
I'm,
not
sure
Matt,
but
maybe
the
the
question
was
this
deployments
in
area
right
that
you
both
have
non
PE
our
piece
and
the
PR
piece
right
and
if
I
understand
it
right,
then
the
PRP
is.
The
reason
is
because
the
service
provider
wants
to
run
MVP
in
a
particular
way.
That
requires
that
and
then
the
enterprise
customer
already
has
their
own
art
piece
avala.
That's
what
you
get
here
right
right,.
U
U
U
I
A
A
So
we
just
just
get
some
status
of
our
active
working
groups.
We've
got
em
trace,
b2
has
actually
been
submitted,
is
G
and
intermittent
pairing.
Now
it's
in
the
editor,
skew
lots
of
us
cos.
Taurus
dress
the
Rebbe's.
It's
our
responsibility
at
this
point
to
please
read
and
confirm,
and
what
their
working
group
last
call
I
understand,
but
reds
are
happening
while
it's
in
the
queue
and
the
sole
responsibility
to
keep
on
top
of
the
stuff
and
make
sure
the
document
represents
consensus
going
through
the
room,
all
right,
other
Doc's.
A
We
have
info
model,
we
have
the
Wi-Fi
doc.
You
talked
about
a
bit
and
I
guess:
Tim's
gonna
give
us
some
input
which
we
got
through
a
proxy
Tim,
but
thank
you
read
and
respond.
The
lists
will
take
the
votes
that
we
had
today
in
the
room.
As
for
adoption,
we'll
take
this
to
the
list
to
confirm.
So
please
pay
attention.
I
try
to
do
this
right
after
the
meeting.
A
While
you
know
people
are
still
engaged
and
hopefully
before
the
holiday
I
know,
we
disengage
here
in
a
few
weeks
and
vanish,
but
thanks
for
your
time-
and
it
is
now
the
pin
working
group
blue
sheets
for
M
Bundy-
have
they
been
signed.
Everyone
signed
the
blue
sheet
from
Bundy
this
guy
over
here.
Nothing
but
trouble.
K
K
L
K
H
K
We
also
have
this
source
discovery
BSR
that
you
requested
publication
for
I.
Think
couple
of
months
back
so
waiting
for
that
to
to
proceed
explicit
tracking
I
think
hasn't
had
any
work
recently.
I
don't
remember,
but
you
would
like
to
see
some
progress
with
the
explicit
tracking
draft
if
you
want
or
some
yeah,
if
you
want
to
sure
we
have
the
offer
here.
K
W
K
H
H
H
H
K
Passed
last
called
just
a
couple
of
weeks
ago,
so
just
been
waiting
for
the
offers
to
states
whether
there's
IPR
on
the
draft
them
and
I.
Guess
you
feel
like
I
might
be
the
Shepherd,
so
I
need
to
do
the
right
up,
but
so
it's
it's
basically
done.
That
will
also
go
to
ITF.
Last
call
at
some
point
in
time,
though,
and
and
if
you're
interested
please,
this
read
it
it's
still
not
too
late.
If
you
have
any
comments
or
issues
on
the
honor
draft.
Q
H
K
K
K
Community.
There's
a
sign
team.
You
know
it's
kind
of
happy
with
it.
At
least
we
need
to
get
additional
review
of
the
MSTP
draft
and
it's
great,
if
you
guys,
would
just
go
ahead
and
review
it,
but
I
feel
sometimes
we
need
to
do.
A
working
group
last
call
to
get
people
to
actually
review
their
drafts.
H
H
K
T
We
basically
are
trying
to
solve
a
specific
problem
where
you
could
have
ipv6
unique,
a
specific
says,
distributed
using
ipv4
sessions,
or
vice
versa,
and
this
is
using
RFC
549
and
if
you
want
to
deploy
multicast
with
it,
we
have
a
peculiar
problem
where
you
cannot
actually
identify
your
pim
upstream
neighbor.
If
you
have
more
than
one
neighbors
on
that
interface,
so.
T
T
Where
we
basically
zeroed
on
this
ipv6
sorry
Tim
secondary
list
option
and
that's
what
presented
and
then
later
on,
prove
accepted
by
the
working
group.
This
draft.
In
this
revision,
we
have
slightly
changed
a
model.
The
overall
as
I
would
say
we
have
just
rewritten
the
draft
in
the
way
that
we
want
to
use.
T
We
are
portraying
the
use
cases
for
secondary
list
option,
whereas
the
RFC
for
6:01
says
that
the
second
a
list
I
addresses
should
be
as
in
the
same
address
family
as
primary
address,
but
we
have
probably
used
cases
where
it
could
be
of
different
address
family,
and
one
of
such
use
cases
is
when
we
use
p.m.
with
RFC
four
for
FX
for
nine,
and
this
is
how
the
new
draft
is
written.
V
K
And
just
having
use
cases,
I,
remember,
tourists
last
time
said
something
about.
You
know
that
it
might
be.
This
might
be
useful
for
other
things,
maybe
also
you
listening
by
people,
six
addresses
for
ipv4,
let's
see
what's
right
now
and
anyway,
so
at
least
right
now
we
have
like
a
single
use
case
from
why
we
might
use
this
address
this
option
for
a
just
different
address,
family
and
I.
Think
if
the
working
group
has
any,
you
know
see
all
the
use
cases
it
you
know
you
could
consider
adding
those
to
the
document
as
well.
T
If
there
is
another
use
case,
people
can
think
about
I
think
that's
how
stick
I
know
I
wanted
to
put
it
out.
I
mean
we
know
one
I
think
what
Telus
last
time
said
that
the
networks
which
are
basically
trance
transitioning
over
to
ipv6,
how
we
could
connect
ipv4
networks
using
ipv6,
I,
think
in
on
something
on
those
lines
and
I'm,
not
sure.
If
that's
this
use
case
could
be
applicable
here,
but
yeah.
We
could
explore
those
options.
T
B
H
B
X
So
what
all
things
I
am
going
to
cover?
Basically
I
just
wanted
to
give
brief
history
about
this
draft
because
it
has
been
around
for
quite
long
and
last
last
time.
I
think
stick
did
speak
to
other
chairs
and
we
decided
that,
since
it
has
been
quite
long
for
out
of
so
it
was
not
updated
for
quite
long,
and
people
might
not
be
still
remembering
what
problem
exactly
we
are
solving.
So
we
wanted
to
go
through
what
problem
we
are
going
to
solve
and
what
all
the
kind
of
changes
we
are
doing
next.
X
So
this
this
draft
was
initially
presented
in
a
of
82
and
after
that
there
were
multiple
I
try,
son
and
it
was
it
got
adopted
by
83
and
in
2014
I
think
it
went
for
a
D
review
and
there
was
a
couple
of
comments
and
author,
most
probably
they
got
busy
and
they
never
updated
after
that,
and
we
basically
in
current
draft,
we
are
addressing
all
those
80s
coming.
I
can.
K
Comment
on
that,
so
did
you
didn't
have
last
colony
requested
publication,
the
draft
as
well,
but
an
earlier
that
was
the
ad
at
that
time.
She
had
some
comments
and
the
office
never
managed
to
update
the
document,
and
now
we
spoke
to
her
alvaro
about
this
current
ad
and
he
basically
said
that
you
know
it's
been
so
long,
so
it
makes
sense
to
take
it
back
to
the
working
group
and
see
if
everyone
are
still
okay.
X
So
the
problem
statement-
if
you
see
here
so
basically
what
exactly
we
are
going
to
solve
so
there
with
current,
if
you
see
PIM
PIM
protocol
itself
has
been
deployed
widely
and
it
has
like
it's
being
used.
So
so
definitely
question
comes
that.
What
exactly
is
the
problem
here,
which
we
are
trying
to
solve
so
in
this
picture?
If
we
have
three
routers
r1
r2
r3
in
same
LAN
segment,
and
only
one
of
them
is
going
to
be
PMD,
rs4
or
RFC,
four,
six
zero
one
and
we
are
basically
we
are.
X
For
example,
I
am
taking
that
the
current
multicast
flow
is
taking
around
500
MB,
so
we
have
one
more
request
which
again
it
is
for
500
MB
and
if
third
request
comes
so,
if
third
request
comes,
if
you
see
the
total
throughput,
what
are
one
can
provide
it
just
one
gig,
whereas
we
have
request
for
1.5
gig.
So,
basically,
what
we
are
going
to
do.
We
are
eventually
going
to
draw
500
MB
of
the
traffic,
but
in
this
whole
LAN
segment
there
are
two
routers
potentially
which
could
be
MDR.
X
X
So
if
we
see
if,
for
example,
if
there
are
1,000
multicast
flows
which
are
flowing,
why
are
one
which
is
currently
MDR
and
if
there
is
any
failure
happening
in
r1,
we
are
going
to
disturb
the
whole
1,000
whole
1000,
multicast
flows
and
they
are
going
to
sift
to
r2.
So
basically
we
are
so
we
are
going
to
relearn
our
coral,
so
r2
will
be
joining
the
code
tree
and
multicast
traffic
will
flow
via
our
so
what-what
exactly
so,
we
are
trying
to
solve
these
two
problems
in
this
draft
that
the
first
thing
was.
X
If
there
are
multiple
pim
routers,
which
can
be
dr,
why
can't
we
share
the
load
so
in
this
solution?
What
exactly
we
are
doing
now
with
dr?
We
are
having
group
dr
so
basically,
dr
election
will
happen
per
group.
There
are
few
groups
which
will
be
flowing
by
r1.
There
are
few
groups
which
will
be
flowing
via
r2,
and
so
now,
if
you
see
load
load,
our
load
is
going
to
get
shared
between
both
of
the
routers
and
the
second
problem
which
we
talked
about,
was
failover
scenarios.
X
If
you
see
here
if
r1
fails,
even
though
r2
is
going
to
take
over,
but
we
are
going
to
touch
upon
only
those
flows
which
were
which
we
are
flowing,
why
r1
and
the
flows
which
we
are
flowing
by
r2,
will
have
no
impact
and
which
all
places
it
is
applicable.
It
is
only
for
last
hop
router,
because
first
hopper
outer
doesn't
really
make
sense
to
do
any
such
thing,
and
it
is
for
SSM
s,
sm,
&,
TM
mode
only
and
by
D.
My
door
is
not
so
of
this
draft,
so
it
so.
X
This
Drive
actually
introduces
couple
of
protocol
changes,
but
the
these
protocol
change
is
not
between
the
version,
5
and
verse.
6.
This
protocol
chain
had
had
been
there
for
from
the
day
one,
and
what
exactly?
We
are
to
do
so
will
basically
now
router
with
their
p.m.
hello.
They
will
announce
themselves
as
a
probable,
dr
and
team.
Dr
is
going
to
elect
the
GDR.
Basically,
GDR
is
nothing
but
which
I'll
router
could
be
potentially
they
could
become.
X
Dr
and
the
pim
dr
is
going
to
provide
list
of
GTRs
to
all
of
the
pimp
routers
and
once
any
router.
It
knows
that
I
am
one
of
the
GDR,
so,
basically,
when
we
get
so
when
you
get
a
GMP
join
and
if
router
is
one
of
the
GDR,
basically
it
is
going
to
do
hash
and
see
that
if
I
am
dr
for
this
particular
flow
and
each
of
the
GDR,
they
are
going
to
run
this
hash
for
I
jumping,
join
and
create
forwarding
states
and
GDR
assert.
X
So
there
are
transition
time
when,
basically,
when
any
GDR
is
moving
from
GDR
to
Nandi
are
at
that
time
it
might
happen
that
out
outgoing
interface
has
not
been
removed
right
away.
In
that
case
there
there
could
be
assert
and
in
those
cases
we
would
always
like
to
consider
that
GDR
state
has
to
be
considered
for
insert
winner,
because
if
we
go
only
wire
I
IP
address,
it
might
so
happen
that
the
router,
which
is
going
from
GDR
tune
and
ER,
wins
the
assert
and
there
will
be
traffic
loss.
X
And
there
were
a
couple
of
minor
comments
and
there
there
is
a
six
on
six
point
to
where
more
of
the
comment
we're
in
the
section
six
point,
two
two
I
had
more
content
about
how
exactly
dr
GDR
will,
whichever
router,
who
is
supporting
this
specification,
how
they
are
going
to
behave
when
they
get
IGMP
join
so
in
this.
What
six
well
so,
basically
were
some
six
of
this
draft.
We
have
edited
more
of
sections
expand
to
and
there
is
a
link
which
provides
the
complete
list
of
comments
from
Eddie.
H
G
H
Anybody
who's
read
this
updated
draft,
so
we
have
just
a
couple:
hands:
okay,
okay,
so
we've
got
a
few
hands
all
right,
so
I
guess
the
requests
from
the
chairs.
We
need
to
have
more
participation
in
reading
this
draft.
If
you
have
comments,
please
ping,
the
authors,
our
intention
is,
to
you
know,
reissue
a
working
group
last
call,
so
we
just
need
to
make
sure
that
we're
in
agreement
as
a
working
group
that
that's
the
point
that
we're
at
it's
that's
unfair.
Okay,.
S
Good
morning,
everyone
I'm
Sandy's
up
from
the
tea.
This
presentation
is
for
the
updating
of
India
improvement,
drops
think
this
rushing.
We
have
a
cause
or
a
commoner
Francisco.
Thank
my
mother
for
his
contribution
to
this
charge
and
the
update
from
working
to
and
the
working
three.
We
had
more
statements
about
a
capability
who
is
not
the
improvement
of
capability
router
and
the
we
had
more
statements
about
the
election
algorithm
and
we
add
a
new
section
for
a
pre
capability
of
stander
side,
its
supplement
of
a
solution.
S
Well,
a
new
router
which
have
no
dia
improvement.
The
capability
comes
in
the
new
router
sends
the
hollow
piccata,
with
its
dia
priority
at
first,
it's
according
to
the
FC
7761
and
the
the
other
for
evasion
of
team
protocol
and
after
the
new
router
receives
hala
piccata
from
because
the
existed
yeah.
S
It
compares
the
dia
priority
if
the
new
router
Hilux
itself
to
the
dia,
it
sends
hello
packet
with
the
asset
itself,
so
the
exist
here,
dr
will
receive
this
packet
and
the
existed
idea
will
elect
the
new
router
to
be
the
according
to
thea
election
algorithm.
So
the
algorithm
is
compared.
Has
compared
compatibility
with
now
yeah
improvement
that
have
ability
wrote
her,
so
it
can
be
used
between
and
now
yeah.
S
K
So
I'm
curious,
so
many
people
have
read
that
your
improvement
draft
just
get
some
idea.
You
know
people
actually
are
looking
at
it's
this
every
couple
of
people,
but
there
we
we
can
definitely
need
more
people
to
look
at
this
and
yeah
I'm
planning
to
look
care
more
carefully
at
it
myself
as
well.
It's
it's
hard
to
make
progress.
If
we
don't
get
people
to
review
the
drafts.
Y
Y
H
Can
you
just
do
Mike?
Can
you
can
you
or
somebody
just
describe
to
the
working
group,
just
high-level,
what
an
MD
a?
What
is
it
asking
for
yang
met
models
to
do,
there's
an
update,
yang
models
right
and
MDA?
What
is
that
MDA?
Does
anybody
what
what
is
an
MDA?
What
did
it
what's
its?
Why
are
we?
Why
are
you
models
being
asked
to
conform
to
an
MDA
like
what's
the
purpose
of
nmda,
do
you
or
anybody
know
Rick.
P
Thailand
and
I'm,
not
a
young
specialist,
but
from
my
understanding
the
end.
Nmda
change
is
instead
of
having
a
single
yang
model,
which
has
your
configuration
parameters
and
your
running
state
all
blended
together
in
one
tree,
half
of
which
is
readwrite
off,
which
is
read-only.
The
nmda
and
I
can't
remember
what
the
acronym
stands
for
splits:
that
into
two
data
stores.
One
is
your
configurable
data
and
one
is
your
ephemeral
running
data,
but
they've
complicated
it
beyond
that
to
add
multiple
data
stores
or
multiple
names.
P
J
I
can
just
quote
from
that
there.
There
is
a
draft.
If
you,
sir,
for
young
and
nmda,
it
basically
says
that's
the
ability
to
inspect
the
current
operational
values
for
configuration,
allowing
clients
to
use
identical
paths
for
achieving
the
configured
value
and
the
operational
value.
So
it's
basically
a
way
of
figuring
out.
Okay,
I
configured
you
to
do
this.
What
is
your
current
state?
Did
you
actually
end
up
in
that
state,
because
I
said
you
were
going
there,
or
did
you
fail?
It's
a
that's!
J
Y
Data
we
merge
the
configuration
and
the
operation
of
state
to
do
one
for
the
updating.
According
to
the
MDA
Wersching,
we
still
have
three
level
hierarchy
and
and
the
global
level
and
the
interface
global
and
the
interface
basic
level.
So
after
the
updating
area
level
had
the
configuration
and
operational
stage
excluded
interface,
global.
Y
From
the
mailing
list
we
have
the
comments
and
the
Olympian
MRD
some
features.
A
one
is
to
two
features
we
add
to
the
Yamato.
One
feature
is
that
the
lightweight
item
here
is
real
or
a
multi
way
to
reference.
The
RFC
five,
seven
nine
or
that
is
means
the
simplified,
exclude
mode
in
the
a
protocol
for
the
standard.
Y
Another
new
features
for
the
explicit
in
tracking
and
that
that,
because
some
wonder,
specific
implementing
implementation
has
the
feature.
So
we
add
the
new
features
for
the
configuration
we
add.
A
country
configuration
in
the
interface
level
that
one
is
the
explicit
checking
enable
or
another
exclude
lead
excluded
for
the
lightweight,
as
in
this
way
3
or
MRD
way.
Y
K
Y
K
C
K
Are
ready
to
publish
this,
but
if
you
didn't
read
the
draft
during
the
last
call
yeah,
please
please
go
ahead
and
do
so
I'm
hoping
there's
no
outstanding
issues
but
have
happy
people
can
help
check,
but
at
least
yeah.
We
we
had
a
design
team
behind
this.
They
were
happy
with
the
draft
and
and
then
there
were
a
few
people
responding
for
the
last
call
saying
it
was
fine
but
making
us
a
couple
of
comments,
but
we
could
need
more
more
input.
K
G
Z
This
is
a
second
version
for
the
presentation
and
after
the
last
meeting
in
ITF
99,
we
we
have.
We
have
rich
reached
a
new
agreement
about
the
whole
host
culture
about
the
IgM
is
moving.
The
idea,
I'm,
assuming
instance,
is
defined
to
configure
the
parameters
and
shoulder
of
race
operator
original
state.
I
Z
Z
And
in
this
instance,
we
use
the
in
the
bridge
scenario.
Name
name
of
the
the
name
is
a
key
for
a
yummy
smoothie
instance.
We
also
define
the
IGMP
rotary
instance
and
the
out
to
monaco
ruling
group
could
be
configured
manually,
and
this
is
the
static
bridge,
monitor
interface
and
a
static
out
for
multicast
group.
Z
This
other
read
only
I
suppose
the
actual
show
the
average
original
state
data
now
and
there
are
three
Kangol
ocarina
face,
which
is
a
hint
of
is
Ralph
out.
Wavy
instance
AC
right
all
out
waking
incidence
pwf,
we
have
utilized
the
existing
idea
of
interfaces
and
the
ITF
out
to
a
few
model
to
you
negate
the
alchemist
for
another.
Z
Z
Z
And
now
we
have
reached
equipment
in
the
design
team
pause.
The
whole
structure
of
the
item
is
smoking.
Yamato
after
I
saw
miss.
The
latest
version
of
the
draft.
I
have
got
some
comments
from
the
middle-east,
and
these
are
unsolved
problems,
one
wise,
the
the
actual
forward
mode.
The
former
mode
may
be
reconsidered.
The
cause
is
only
fit,
for
the
instance
same
doesn't
fit
for
the
interface.
Z
Nitin
maximize
the
is
the
comes
from
the
table
from
the
Middle
East,
and
this
is
the
corner
she
asked
her.
He
asked
where
the
convert
for
them
non-member,
live,
will
also
be
useful.
I
have
some
confused
about
the
risk.
This
question:
that's
why
I
would
like
to
communicate
ways
with
him
on
this
honest
meeting,
a
coding,
freeze,
ice
cream,
all
polished
at
this
question,
I.
X
X
Z
H
H
K
H
O
P
Rick
taylor
again
I
sound
like
a
yang
fanboy
now
and
I'm,
not
really,
but
the
yang
models
are
really
important
because
people
have
got
to
be
able
to
configure
this
stuff
and
otherwise
we're
back
writing
MIBs
again
and-
and
you
need
to
kind
of
balance
that
somehow
we
got
to
model
this
stuff.
So
so
yang
is
worth
it
and
I
know
it's
hard
and
it's
tedious,
but
it's
worth
it.
J
Michael
Lorenzen
yeah.
You
just
want
to
reiterate
that
the
way
we
do
things
if
there
is
no
yang
model,
we
can't
configure
it
because
that
that's
how
we
do
things
so
either
it's
it's
vendor
proprietary
models
or
it's
the
ietf
models.
We
prefer
ietf
models
because
then
it
can
be
the
same
between
vendors.
So
yes,
I,
know
it's
tedious
work
and
feels
hard.
I
think
it's
hard,
but
we
need
it.
Yeah.
K
K
So
I
think
one
question
is:
do
you
think
it's
useful
to
have
a
model
for
a
GMP,
Emily
snooping
and
think
that
the
answer
is
fairly
obvious,
yeah
and
then
I
guess
that
the
next
thing,
which
is
important,
I
guess,
is
for
vendors
to
see?
Can
they
support
this
model
you
know?
Does
it
include
like
leaves
that
they
can't
support
or
when
or
operators
doesn't
include
the
leaves
you
need
to
actually
operate
this
box?
So
you
need
people
that
know
multicast
to
kind
of
look
if
Sinha
models
have
the
right
content,
Tony's.
I
Record
so
the
first
thing
I,
don't
understand
is
whether
or
not
it's
permissible
to
have
the
yang
model
as
a
standard
for
a
feature
that
cannot
be
standard
so
snooping.
We
had
this
wonderful
political
debate
that
it
must
be
informational,
because
snooping
is
so
bad,
and
so
we
have
an
ad
here.
Is
there
an
answer?
N
I
No,
no,
but
if
I'm
basically
doing
anything
else
that
you
know
a
derive,
you
know
protocol
solution
and
somehow
depends
on
igmp
snooping.
Everybody
comes
back
to
me
and
says
that
must
be
informational
because
you're,
depending
on
something
that
is
an
informational
thing
and
now
kind
of
the
operational
interface
is
I'm
just
pulling
because
I
just.
I
N
I
mean
it's
an
interesting
point:
I,
don't
think
that
there
is
a
you
know:
I,
don't
think
this
is
necessarily
being
discussed
before
I.
Don't
think
that
there
is
a
written
down
anything
but
I
mean
you
can
always
move
it
forward
and
if
it
ends
up
being
changed
to
informational,
you
know,
I,
don't
think
the
tracks
that
important
as
long
as
it's
actually
usable.
O
Yeah
I
mean
mechanically
there's
the
question
of
down
wraps
right
and
but
down.
Riffs
are
just
a
thing.
We
just
do
them
right.
We
we,
we,
don't
you
as
an
insurmountable
obstacle,
because
sometimes
we
come
along
with
a
standards
track
document
that
refers
to
an
informational
document.
Sometimes
we
even
do
that
normatively
and
that's
actually
just
a
thing
that
comes
up
in
the
last
call
and
we
choose
not
to
ignore
it
and
then
when
no
one
objects
we
proceed
or
if
there
are
objections,
then
we
address
them
at
that
time.
I
Interesting
that
the
question
hasn't
come
up
in
before
for
young
models.
Okay,
so
the
other
stuff
then
was
in
terms
of
we're
still
doing
new
features
without
having
yang
models
right.
So
there
were
just
a
bunch
on
dr
and
stuff,
which
I
think
have
configuration
is
assigned
with
them,
and
so
I
was
just
wondering
you
know.
Is
there
any
best
current
practice
off?
J
So
with
this
works
like
with
SNMP
but
but
I,
think
the
yang
is
now
getting
more
traction,
so
I
hope
it'll
be
better
I,
don't
I
mean
it's
all
about
this.
Is
the
ITF
the
right
venue
to
do
yang
models
under
zation
I
would
hope
so
I'm,
not
sure
it
is
I
mean
if
it's
all
down
to.
How
do
we
do
things?
I
would
like
the
models
to
be
developed
here.
I'm,
not
sure
the
process
is
great
for
it.
J
I
Yeah
I
was
I
was
even
thinking
about.
Is
it
even
possible
with
the
way
that
these
documents
do
one
thing
or
the
other?
To
that
say,
no,
sorry,
you
can't
bring
in
the
new
bloody
feature.
It
needs
to
be
configured
the
featured.
You
know
RFC.
The
protocol
RFC
itself
needs
to
have
the
yang
model,
so
that
always
avoids
this.
You
know
delay
with
the
yang
modeling,
it's
impossible,
yeah,
yeah,
big,
but
but
then
basically
yield.
Also,
you
know
create
these.
These
problem
that
they
will
come.
They
drove
so.
N
Warren
Kumari
again
just
I
relate.
We
hadn't
been
doing
that
subscribe,
so
I
mean
there
is
some
what
to
try
and
make
this
less
painful
and
move
stuff
faster.
Benoit
has
a
thing
on
semantic
versioning
so
that
you
know
it's
actually
easier
to
do
this
and
not
have
it
be
a
mini
year
thing
and
then
the
RFC
is
published
and
then
it's
set
in
stone.
N
J
I
do
but
it,
but
basically
I've
had
this
happen
where
oh,
we
need
to
fix
this,
then
the
vendor
just
goes
in
just
proprietor
changes.
We
just
take
their
model
and
then
hopefully,
a
year
or
two
later
out
comes
a
standardized
model
that
then
all
vendor
schemes.
But
this
is
like
reality.
That's
the
way
it
has
to
be,
but
I
think
the
working
group
all
working
groups
have
to
keep
track
of
what
stuff
do
we
have
that
we
don't
have
a
model
for
and
fix
that.
P
Rick
Taylor
again
just
following
up
from
that
point:
there's
also
a
reality
that
the
people
who
are
developing
the
best
practices
and
the
protocols
are
also
probably
the
best
people
to
define
the
data
models,
even
if
they
don't
do
it
within
the
ITF.
Those
are
probably
the
same
guys,
so
it
kind
of
happens
at
the
IETF.
Despite
the
pain
involved.
H
K
Also,
when
I
had
that
we,
so
we
have
an
design
team
that
anyone
can
join.
There
is
a
wiki
page,
there's
also
mailing
list
for
the
design
team
and
any
of
you.
If
you're
interested,
we
would
be
happy
to
have
more
vendors
or
operators
joining
the
effort.
Welcome
to
do
so,
and
just
ask
me
later
or
if
you
want
I,
can
send
an
email
to
the
mailing
list.
So
just
some
update
about
where
the
mailing
list
in
the
week
is,
and
so.
K
Yeah
yeah,
okay,
you
are
against
of
the
adopting
this
right
now
so
I
guess
I
should
do
an
adoption
color
on
the
list.
At
least
they
have
offers
from
a
few
different
vendors
that
probably
are
in
favor
of
adopting
it.
So
that's
also
helps
but
yeah
I
would
like
to
say
that
right
now,
it's
just
about
adopting
this,
and
hopefully
you
think
it's
useful
to
have
a
young
model
for
this.
You
can
always
change
the
content
later.
S
So
you
can
see
your
domain
one
two
three
here,
so
the
polar
route,
her
perversely
or
bada,
rotor
rotor,
must
convert
a
beer
encapsulation
such
as
mr3
and
Mar
fall.
We
want
before
so
multiple
PR,
encapsulation
and
decapsulation
function
will
run
in
the
border
rooters
and
at
will
and
increase
the
efficiency
of
be
rewarding
next.
S
So
we
found
that
maybe
we
can
merge
several
regions
into
one
beer
domain,
so
the
peer
encapsulation
that
the
consolidation
terms
will
be
decreased.
The
beer
for
the
invention
will
be
improved,
but
there's
a
new
problem:
how
to
build
the
be
rewarding
across
multiple
routing
regions,
routing
regions
I
mean
I'm.
Sorry
that
I
have
forgot
to
write
another
filter
with
the
state
configuration
so
because
they
are
stiffer
winter
IP
protocol
running
in
the
network
and
the
even
static
configuration
route
routing
routing
be
used.
S
So
maybe
we
can
give
you
some
conversation
in
border
routers,
though
it
in
more
complexity,
but
it
may
be
worked
in
some
regions,
but
with
the
static
configuration
regime
and
there's
no
IP
routing
protocol
running
in
the
region.
So
maybe
the
existed,
IDP
extension
cannot
be
used,
so
we
will
elect
to
form
our
solution
to
so
the
every
region.
So
we
use
a
protocol
independent
flooding
mechanism.
S
The
mechanism
is
defined,
young
team
sauce
is
Kara,
PS
are
trapped
in
human
group.
We
use
this
function
to
flood
PR,
know
the
information
in
order
to
build
your
forwarding
plane
and
the
extension
will
be
aligned
with
the
SPF.
Oh,
he
sees
a
BGP
extension
and
every
nodes
in
the
domain
will
build
appear
forwarding
playing
according
to
the
very
information
and
also
its
fever.
Pebble
is
forwarding
information
base
table,
so
we
can
build
the
pier
folding
playing
without
the
dependent
of
IDP
BGP
routing
computing.
S
Someone
will
be
confused
with
this
draft
and
the
other
a
TPP
or
BGP
exchanging
dropped,
so
this
structure
will
not
replace
the
IDP
or
PDP
very
extension
dropped.
This
function
is
just
a
supplement
of
routing
protocol
extension,
especially
in
hybrid
environment,
and
we
know
that
appear
eliminates
pim
protocol
states
in
cardamon,
but
the
destructor
is
not
backward.
We
just
use
the
flooding
function
to
delivery.
Er
knows
the
information
the
we
do
not
use.
Pin
protocol
to
building
Modi
has
the
trees.
S
So
there's
a
lot
of
discussion
in
the
mailing
list
and
I
have
talked
to
jeopardy
yesterday
and
found
that
maybe
in
IDP
in
her
option
it's
campy
are
way
too
so
the
IDP
interruption,
but
we
sink
the
pimple.
Adding
mechanism
may
be
our
general
function
for
this
network,
especially
the
networker,
who
has
which
has
the
static
configuration
area.
So
we
want
more
comments
about
the
function
and
the
IDP
interruption
function.
S
Maybe
maybe
this
function
can
be
used
the
default
generals,
and
maybe
there
can
be
two
drafts
fo
for
the
function
when
operator
can
select
the
right
way
to
use
for
their
networks.
If
the
network
hasn't
has
a
cos.
Theta
configuration
area,
we
can
use
the
IDP
interruption
function
and
if
they
had
some
static
or
configuration
region,
you
can
use
this
function
to
for
generic
solution.
S
AA
G
AA
S
AA
D
AA
S
K
A
committee
had
a
couple
of
comments,
so
one
thing
is
beer
is
doing
the
extension,
so
SPF
and
ice
ice,
and
so
they
are
done
in
the
dear
working
group.
So
it's
not
clear.
You
know
in
this
case
whether
this
would
be
maybe
done
in
beer
as
well.
I'm,
not
PIM,
but
that's
something
we
can
discuss
later,
but
the
misused
or.
G
K
AA
So
I
guess
the
only
comment
is
if
it's,
if
it's
not
pain-
and
you
guys
seem
to
be
a
little
bit
open
on
that
idea,
then
I
guess
this
whole
document
will
be
kind
of
pushing
to
the
beer
working
group,
which
makes
it
a
little
bit
different
like
it
seems
like
it's
kind
of
between
two
working
groups
right
now
so
I'm,
beginning
to
wonder
if
it's
cleaner
to
go
one
way
or
the
other.
That's
all.
Thank
you.
Thank.
S
V
Hey
I
saw
Cisco,
so
it's
not
a
bad
idea,
I
think
the
problem,
but
it's
not
very
fashionable
these
days
to
have
like
separate
protocol
to
distribute
this
information
as
a
lot
of
information.
These
days
is
put
into
the
IGP
all
right,
so
this
would
become
like
the
LDP
for
unicast.
It's
like
the
the
equivalents
in
full
form
for
beer,
so
that
sense
is
sort
of
going
in
the
opposite
direction,
yeah.
So,
but
it's
something
to
think
about,
but
that's
my
basic
concern.
N
So
Warren
Kumari,
without
any
hats,
I
mean
it
sort
of
feels
like
seeing
as
beer
is
doing
other
extensions
in
beer.
You
know,
maybe
it
makes
sense
if
this
goes
forward
but
to
be
discussed
there
and
just
to
be.
You
know,
strong
cross
working
group,
discussion
and
lost
cold
and
stuff
see
see
here,
just
a
thought.
U
Y
U
U
Another
thing
is
that
the
encapsulation
information,
for
example
the
NPS
labels
for
different
domains
and
subdomains
and
defends
us
pierre
sets.
Those
information
is
only
needed
in
one
domain.
It
really
does
not
need
to
go
into
other
domains
and
when
you
use
I
use
I
GP
border
routers
to
carry
information
for
for
those
PFA
ours
in
other
domains.
You
can
strip
those
encapsulation
area
information,
and
so
that
brings
you
more
optimization.
U
S
G
S
Problem,
it's
like
the
proxy
proxy
function
and
the
IDP
interruption
right,
yeah
yeah.
We
can,
you
think,
and
this
function
I
think
as
as
all
as
still
has
some
useful.
Maybe
maybe
someone
can
want
to
use
a
simple
way
to
achieve
it.
The
result
so
I
don't
know,
but
I
need
more
comments
about
yet,
and
the
operator
may
may
come
and
say,
and
the
week
have
got
the
thinking
about
the
operator
yeah.