►
From YouTube: IETF104-MBONED-20190328-1050
Description
MBONED meeting session at IETF104
2019/03/28 1050
https://datatracker.ietf.org/meeting/104/proceedings/
A
A
A
B
C
F
G
Okay,
all
right
in
terms
of
active
working
group
documents,
pretty
much
all
of
them
are
going
to
be
spoken
of
today.
So
since
Bangkok
we
adopted
the
Dryad
draft
Jake's
gonna
provide
updates
on
that
the
deploy
data
center
deploy
Draft
has
been
around
for
a
while
that
may
be
ready
for
a
last
call.
I
suspect
mike
is
going
to
ask
for
something
like
that
when,
but
he's
going
to
provide
updates,
there
is
the
Wi-Fi
draft
Charlie's
going
to
present
on
that.
Charlie
is
I
know
he
is
very
interested
in
working
group.
G
Last
call,
though
we
did
get
some
interesting
feedback
this
week
on
that
there's
the
deprecated
ASM
draft
for
inter-domain
we
do
feel
like
this
is
likely
for
ready
for
working
last
call,
but
toy
list
is
gonna,
provide
an
update
on
that
and
there's
the
yang
draft
sandy.
Do
you
want
to
have
any
thoughts?
What's
what's
the
latest
on
that
draft?.
H
G
I
J
:
yeah
sure
this
is
gonna,
come
up
again.
I
encourage
people
to
take
a
look
at
it.
It's
gonna
be
important
and
I'm
gonna
be
trying
to
make
a
viable,
try
it
and
implementation
for
like
browsers
of
a
web
UI,
and
this
is
going
to
be
a
critical
piece
of
it
as
I
understand
it,
because
the
browser
has
to
have
authentication
in
order
to
accept
it.
And
so
there
is
some
question
as
to
whether
this
will
come
through
and
bone-deep.
But
it's
going
to
require
some
review
by
multicast,
knowledgeable
people
I
think
so.
I
G
What
happened?
The
note
well
slide
what
happened
that
that
was
up
there.
It
went
by
so
fast,
I
put
them
together
and
yes,
yes,
I
forget
everything.
You
said
we're
gonna
have
to
start
over
now.
Would
anyone
like
me
to
read
the
note?
Well,
I
can
do
it
if
you'd
like
twice,
okay
yeah,
all
right,
please
note
it
well
so
yeah
when
I
hit
the
button,
it
went
I
hit
the
button
twice,
apparently
all
right.
So,
let's
start
again,
no
all
right.
J
So
basically
that
was
sort
of
a
maintenance
release
and
mostly
editorial
sort
of
comments.
As
you
can
see
here,
there's
not
any
deep
technical
additions
to
the
draft,
but
it
was
much
improved
anyway.
So
I
guess
I
don't
have
to
go
through
too
much
in
detail
on
that.
But
there
was
some
inconsistency
in
particular.
Using
sta
is
more
the
author
of
the
802
wireless
terminology,
so
I
put
that
in
instead
of
client
and
and
so
on.
So.
J
Jake
sent
me
some
email
about
how
there
are
commercial
Wi-Fi
solutions
that
do
a
better
job
than
what
were
you
described
in
the
draft
and,
and
you
can
see
what
it
says
here-
attract
some
members
of
the
group
in
them.
I
guess
the
local
attachment
to
an
access
point
and
figure
out
the
appropriate
rate,
that's
necessary
for
it
to
reach
each
one
of
those.
J
So
I
could
maybe
perhaps
characterize
this
as
a
tiered
approach,
instead
of
using
the
least
common
denominator
to
send
packet
out
to
the
subscribers
in
in
the
local
group
of
devices
attached
to
an
access
point,
the
access
point
does
a
certain
amount
of
record-keeping
to
figure
out
how
to
more
efficiently
send
this
packets
out
there
to
the
receivers
now.
I
think
this
actually
deserves
some
more
discussion
to
see,
for
instance,
if
you're
sending
it
out
at
the
most
weakest
tier.
J
K
K
J
J
L
D
L
Yeah
so
I
didn't
the
draft.
Have
the
kind
of
you
know,
conversion
to
unicast
already,
so
the
the
primary
new
thing
here
seems
to
be
a
little
bit
of
logic,
of
not
doing
it
at
higher
rates
right
or
was
the
other
stuff,
not
in
the
draft
in
there
before,
because
I
think
that
was
well
long,
known,
right,
Cisco
I
think
was
one
of
the
first
companies
to
do
that
hundred
years
ago.
Right.
K
Okay,
but
the
weekend
rathole,
what
he's
talking
about
a
solution?
My
concern
was
that
when
we
first
met
to
discuss
this
was
actually
last
Prague.
It
was
I
mean
we.
This
has
come
up
and
down
over
time,
but
we
sat
down
and
for
the
first
time
I
sat
with
engineers
from
Apple
Nick.
You
were
there
right
and
Google,
and
they
were
talking
about
whole
different
problem
that
we
weren't
thinking
about.
Basically,
the
fact
that
these
things
had
to
listen
all
the
time
and
killing
battery
power.
K
What
the
AP
does
in
this
case
didn't
solve
this
problem,
so
there's
a
whole
like
almost
siloed
problems,
we're
facing
and
I
think
when
we
make
sure
that,
because
you
fix
one,
we
didn't
solve
everything.
We
may
have
fixed
one
piece
that
were
most
concerned
with,
but
not
necessarily
what
the
vendors
are
concerned
with
Mike
and
named.
M
Are
you
with
Mike
McBride
wha
we
so
as
toyless
alluded
to?
We
do
in
this
draft.
Mention
non-standard
solutions
like
multicast
unicast,
something
that
Cisco's
done
for
many
many
years
and
maybe
and
probably
others
so
my
suggestion
here
would
be
to
just
I
was
originally
just
thinking.
Let's
just
not
deal
with
this
because
we
just
need
to
be
done
with
this
draft,
but
it
probably
would
be
a
worth
at
least
mentioning
this
not
getting
into
a
whole
two-page
discussion
about
it.
But
right
you
know,
maybe
so
yeah
and
then
be
done
well.
J
From
David
Lamm
producers
David
here,
okay,
well,
he
said
the
product
was
a
product
was
from
Aruba
and
so
I
I
want
I
asked
for
citation
because
of
very
detect.
It
be
nice
for
people
to
be
to
follow
up
a
little
bit.
On
the
other
hand,
I
don't
want
to
turn
into
a
product
advertisement
right,
so
so
I'm
still
not
quite
sure
what
to
do
about
that.
But
presumably
the
situation
will
clarify
and
I
can
make
a
modification
to
the
draft
printer.
N
So
I
think
the
important
thing
now
is
to
get
this
document.
It
doesn't
need
to
be
perfect
and
mention
all
the
nooks
and
crannies
everywhere.
We
have
I
still
think
we
have
a
problem
and
I
would
like
to
proceed
on
solutions
instead
of
working
on
the
problem
statement,
so
that
can
we
please
just
say
it
enough
point
out:
there
is
a
problem.
L
K
I
think
the
point
of
the
documents
are
focus
in
the
problem
statements
who
we
can
articulate
that
moving
forward,
and
then
people
start
working
on
solutions.
I,
don't
think
we
only
spend
time
in
the
document
on
what
the
various
solutions
are
and
get
get
into
the
niche
of
them
and
what
they're,
where
they're,
available.
K
K
L
E
N
J
Pressure:
okay,
the
other
comment
was
from
John.
Commissar
is
January,
so
basically
he
sent
some
good
comments
about
the
drought.
Amazing
editorial
problems
that
hadn't
been
fixed
yet
and
mentioned
a
couple
of
possible
additions
to
the
draft
one
is
to
get
away
from
stateless
auto-configuration
or
use
DHCP,
because
you
cut
down
the
number
of
multicast.
J
N
J
N
M
J
M
L
I'm
not
so
so
I
would
say:
I'd
be
happy
to
have
everything
taken
out.
That
kind
of
goes
beyond
the
problem
statement.
If
I
have
a
clear
understanding
that
we
have
a
good
plan
to
bring
them
back
into
an
appropriate
document
fairly
soon,
so
that
we
don't
lose
it
right.
So
what
would
be
that
plan?
I?
Think
that's.
A
O
Of
a
cut
and
paste
yeah
so
or
inquiry
Google
kind
of
following
on
for
something
Charlie
said
what
we
currently
have
is
some
sort
of
workarounds,
which
are
operational
changes
suggest
in
the
document.
They're
not
really
protocol
changes.
I,
don't
want
to
need
that
sort
of
stuff
and
not
just
rip
it
all
out
or
what
you
know.
Some
of
them
aren't
ready
they're.
Just
if
you're
having
this
problem,
you
can
work
around
it
for
now
by
turning
the
following
seven
knobs.
N
So
yeah
one
more
time,
Michael
here
again,
so
just
looking
quickly
through
here.
This
agree
with
that:
it's
not
proposing
stop
reducing
one
protocol
or
something
it's
like
partial
mitigation
and
documented
things
that
people
are
already
doing
so
I
think
leave
it
at
that.
I
think
the
document
is
in
pretty
good
shape.
I
think
we
should
just
try
to
you,
know
ship.
It.
I
J
G
This
is
great,
so
I
guess
you'll
address
these
last
few
comments.
When
you
think
it's
ready,
then
you
know
send
request
working
good
last
call
what
what
is
a
plan
for
the
next
step,
that
the
solutions
was
that
going
to
be
in
this
working
group
or
elsewhere?
Did
we
discuss
the
side
that
this
is
the
working
group
that
we
never
discussed.
K
Any
plans
yet
from
chair
to
chair
it's
in
our
Charter
that
we
don't
create
solutions
here.
We
don't
do
protocol
work.
That
doesn't
mean
the
ideas
can't
come
from
here.
Doesn't
it
Raph's
can't
come
from
here,
but
when
they
go
live
somewhere?
It
goes
to
a
working
group
that
actually
solves
problems.
We
just
point
out
problems
and
cry:
okay,.
F
K
N
J
Dangers
when
ragged
end
they
did
nothing
or
what
happened
I.
My
understanding
is
that
there's
been
some
more
deployment
for
I,
believe
DMT,
but
not
I'm,
not
aware
well,
first
of
all,
I'm
a
NATO
2.15,
which
is
logically
separate
from
dot,
11
and
so
I,
don't
know
all
the
stuff
that
goes
on
there,
but
I
try
to
pay
attention
anyway
and
I.
Don't
remember
any
additional
multicast
work
going
on.
G
Yeah
that
does
bring
up
a
good
point.
We
we
do
specify
some
protocols
and
tools
related
to
transitions
and
Inter
and
operational
troubleshooting
things
like
things
of
that
nature,
like
M,
trace,
AMT
protocol
routing
protocol
group
membership
protocol,
things
like
IGMP
MLD
protocols
of
that
and
PIM
those
go
in
the
other
working
groups
like
10
again.
N
I've
been
sitting
in
so
I
would
like
this,
you
know
shipped
and
then
I
think
we
need
to
go
to
the
IAB
or
something
because
this
is
way
way.
Bigger
than
this
group
I
mean
this
touches.
If
decision
is
Hawaii,
I,
Triple,
E
or
the
ietf
is
the
the
I
Triple
E
is
not
doing
much
so
then.
The
logical
conclusion
is
the
IHF
needs
the
useless
multicast,
and
this
is
now
all
over
the
place,
because
there
is
a
lot
of
assumption
here
that
you're
deploying
into
a
10base2
kind
of
network.
N
J
N
M
Is
Mike
McBride
and
I'm
Rick
I
am
in
agreement
with
Michael
Abrams
I.
Don't
think
we've
kind
of
I
wanted
to
see
if
we
can
resolve
it
right
now,
because
if
we
wait
till
I
don't
know
wake
up
Montreal.
So
are
we
in
agreement
that
you'll
add
a
line
or
two
about
what
was
just
discussed
with
that
Jake
brought
up
in
the
list
and
then
we're
done
or
do
we
not
even
include
that?
What
well
and
just
be
done
right
now.
J
J
K
K
Well,
okay,
then
do
a
list
so
who
actually
liked
the
proper
way
to
do
this
as
a
working
group
cheer
yourself
who's
read
the
drafts
in
its
current
form.
Okay,
who
thinks
that
you
know
we're
close
enough.
The
revs
of
the
notes
will
make
it
ready
for
work.
Your
blasts
call
for
into
a
same
set
of
hands
I'm
going
with
to
reading
it.
There's
not
commitment,
we're
not
gonna
hold
you
to
it.
Let
them
inutes
reflect
that
there's
room
support
in
the
direction
that's
taken
and
we'll
review
again
to
the
list.
G
O
M
Alright,
so
this
is
a
document
that
is
intended
to
be
a
practice
and
deployment
draft
coming
out
of
Ambo.
Indeed,
something
that's
in
our
Charter,
it's
giving
a
review
of
multicast
in
the
data
center.
It's
something
that
we've
presented
for
a
few
years
now
there
was
a
year
or
two
cap
in
there
with
we
kinda
lost
interest
and
I
had
some
of
the
things
that
I
was
working
on.
Besides
ITF
the
last
year,
we've
been
me
and
I,
honey
from
Arista
and
I
had
been
working
on
this
draft.
We
think
it's
ready
to
be
published.
M
We
have
given
a
decent
overview
of
multicast
use
in
an
honest
way
and
the
data
center.
We
did
include
a
few
things
of
forward-looking
things
like
what's
happening
within
segment,
routing
and
multicast
the
things
that
may
not
be
happening
today,
but
just
kind
of
giving
an
idea
of
some
of
the
things
that
were
working
on
in
the
idea.
So
again,
this
is
a
practice
and
deployment
draft
fairly
innocuous,
and
we
would
like
to
seek
for
booking
a
bus
call.
Maybe
I
should
ask
if
there's
any
comments
before
we
do
that.
P
M
P
M
I
I'm
taking
it's
Jake
I'm
taping
up
a
couple
of
editorial
suggestions:
okay,
there's
a
few
just
it's
but
I-
think
it's
mostly
fine.
Okay,.
M
K
M
G
L
Is
just
having
you
know,
editorial
edits
so
for
the
ones
who
don't
remember
what
this
was
about.
This
is
supposed
to
be
a
BCP
outlining.
Why
and
where
and
how
to
use
SSM
and
not
ASM
for
inter
domain
deployments
with
the
you
know,
goal
of
trying
to
retire
ASM
as
an
inter
domain,
multicast
solutions
and
yeah.
L
The
specific
editing
that
we
did
was
just
in
Section,
four,
two
and
four
three,
just
clarifying
the
verbiage
about
the
support
of
IGMP,
v3
and
ml
DB
support
in
terms
of
the
requirement
for
using
SSM
and
not
just
ASM,
so
I
think
we
really
are
close
to
you
know
being
finished
with
it.
So
I
think
we
would
like
to
do
a
working
group
last
call
for
it,
unless
somebody
sends
up
in
his
major
concerns
about
that.
But
I
think
we
finished
all
the
feedback
that
we've
gotten
over
the
time
on
the
list.
Oh.
K
Okay,
apparently
we're
drawing
straws,
that's
all
right,
so
who's
read
the
draft
in
its
current
form.
All
right,
you
got
four
who
thinks
is
ready
for
a
ringer
blast
call
burn
form
honest.
That's
some
solid
support,
though
okay
great.
Let's
the
note
show
that
we
hate
or
lists
Wendy
Praeger
minutes,
but
the
mini
show
that
we
saw
supporting
the
room
and
we'll
take
it
to
the
list.
Okay,.
L
L
Q
P
L
You
know
we
were
in
the
to
win
two
two-week
window,
so
we
couldn't
publish
it.
I
completely
got
backlogs.
I
was
doing
in
the
morning.
Don't
post
that
one
that
has
a
IETF
Tim
I
posted
one
under
Eckert
Tim
right
now,
once
we
go
through
the
official
working
group
which
is
PIM
right.
Basically,
we
can
do
another
rep
of
it.
Yes,.
P
L
For
PIM
are
better
because
I
was
trying
to
give
one
hour
before
this
very
backlog
right,
so
just
a
reminder
for
the
process
for
the
ones
who
haven't
followed
it.
So
we
started
to
discussing
how
we
can
get
to
conclusions
of
revisiting
the
status
of
the
IETF,
IGMP
and
MLV
documents
we're.
Basically,
the
only
full
internet
standard
is
the
9091
IGMP
version,
one
document
everything
else
is
at
best,
post-standard
or
so
so.
L
I
mean
kind
of
adoption
of
newer
versions
of
this
actually
has
shown
to
be
difficult
when
you
know
customers
and
implementers
see
that
status.
So
how
do
we
go
forward
with
this?
Well,
the
typical
process
with
which
we
are
trying
to
prove
that
you
know
the
newer
stuff
actually
is
adopted
and
can
progress
in
status
is
to
creating
a
questionnaire
so
that
we
know
what
the
industry
is
doing.
L
That's
what
we
started
around
IETF
103,
so
we
had
created
a
working
kind
of
a
small
team
and
and
those
those
were
all
the
folks
on
the
first
slide
that
contributed
very
actively
to
the
kind
of
irregular.
But
quite
a
few
meetings
here
that
we
had
so
I
think
a
very
wide
range
of
people
involved.
I.
Think
that
helped
us
a
lot
to
also
come
up
with
the
right
questions,
and
so
then
the
this
would
be
the
target
draft
name
right
now.
L
I
just
send
it
up
to
data
tracker
under
my
name
and
the
idea
would
be
to
have
I
think
officially
the
PIM
working
group
right.
So
this
is
just
informational,
showing
here
in
EMBO
Andy.
Of
course,
operational
feedback
is
equally
important.
As
from
the
actual
protocol
group,
a
proof,
the
questionnaire
is
good
to
be
sent
out
right.
I,
don't
think
that
we
want
to
publish
the
questionnaire
as
an
RFC,
but
that's
a
decision
to
be
made
by
team
working
group
and
then
basically
have
you
know
the
working
group
chairs.
L
That's
what
the
the
draft
says
or
working
group
members
find
appropriate
places
to
disseminate
it
through
the
industry
and
then
have
hopefully
Tim,
do
the
same
role
that
he
did
for
the
same
type
of
questionnaire
for
the
PIM
protocol,
collect
anonymously
the
feedback
and
create
statistics,
and
that
statistics
will
then
be
taken
to
continue
further
with
the
process.
So,
typically
that
would
be
written
up
as
a
draft
and
then
appropriate
conclusions
about
you
know.
L
The
change
of
the
status
of
the
IGMP
and
MLD
documents
would
occur
next
slide
so,
and
you
just
try
to
see
if
you
can
get
up
on
any
thing.
It's
it's
called
draft
Eckhart
Tim,
I
jump
zero-zero.
If
we
can
just
repeat
go
through
because
pretty
much
gather,
yada,
yada,
yada
and
instructions,
yeah
I'm
good
on
any
screen.
You.
O
P
L
L
Q
D
L
So
we
started
first
trying
to
create
different
questions
for
different
audiences,
which
would
have
made
it
easier
but
would
have
created
so
many
questions,
so
we're
trying
to
basically
keep
it
somewhat
pithy.
So
three
one
go
down
all
right.
Who
are
you?
Why
are
you
and
if
not
so
when
so
and
then,
basically
here,
it's
a
co-op
again
pic
affect
your
implementation
status
right.
So
this
is
basically
what
we're
really
interested
in,
because
ultimately
the
status
depends
on.
L
You
know
showing
some
good
amount
of
adoption,
so
I
GOP,
v1,
v2
v3,
then
also
the
lightweight
IGMP,
v3
I.
Think
that's
going
to
be
very
interesting
as
well:
ml,
DB,
1
ml
db2
like
well
MIT,
B,
2
and
then
basically
yeah.
So
what
we're
trying
to
stay
here
a
little
bit
open-ended,
which
is
which
is
kind
of
a
little
bit
dangerous,
but
it
was
so
easy
to
come
up
with
so
many
crazy,
detailed
things.
P
Stick
yeah
stick
yeah.
We
can
maybe
discuss
this
more
in
pmo
course,
but
I'm
a
little
worried
that
people
will
see
like
Oh,
which
I
can
pierce
and
three
features
and
they
will
be
like.
Oh
what's
a
feature
or
so
maybe
it
would
be
good
to
know
summer.
Call-Out
like
these
are
the
features
watching
PHP
and
I'll.
Give
it
to.
M
C
L
L
So
this
is,
you
know
already
just
for
implementers,
which
ambiguities
are
inconsistencies
right,
make
the
implementation
challenging,
and
what
suggestion
would
you
make
to
the
PM
working
group
as
it
seeks
to
update
these
documents
so
and
then
for
for
the
network
operators
for
don't
be
to
deployment
status?
So
this
is
pretty
much
duplication,
just
you
know
substitute
implement
with
deploy
so
that
we
felt
still
felt
that
that
was
sufficiently.
You
know
well
simple
and
good
for
us
to.
L
So
then
deployment
specifics
right.
So
here's
this
the
same
one
so
that
the
Xcode
mode,
obviously
is
the
primary
thing
we're
interested
in.
But
if
you
can
think
of
you
know
any
other
specific
features
we
may
want
to
mention
here
that
wouldn't
be
good
to
know
right
there.
The
most
important
one
is
number
three
right:
does
your
network
rely
on
fallback
mechanisms
between
different
IGMP
versions
like.
P
P
P
L
L
N
L
N
None
August
so
might
work
those
make
enemy
lose
I.
Just
had
one
thing
that
I
happened
to
so
I
required
to
Emily
me
too,
and
then
they
only
supported
stark
emoji
join
on
Emily
veto,
so
they
fulfill
the
letter
of
the
requirement.
But
not
the
intention
of
why
I
wanted
source-specific
to
be
supported.
Would
that
be
captured
here
that
those.
G
G
N
I
G
L
The
IGMP
implementation
of
those
set-top
boxes
because
I've
been
in
living
through
the
same
pain
of
course,
did
support
full
IGMP
b3,
but
the
application
wasn't
using
SSI.
So
now
here's
the
problem
a
little
bit
that
we
try
to
focus
this
questionnaire
on
the
actual.
You
know
ITF
standards,
as
opposed
to
the
actual
deployments
that
which
is
why
it's
inhuman
of
an
M
bone
D.
So
this
is
a
little.
L
N
Then
they
then
they,
then
they
said:
oh
we've
solved
the
problem
here
you
have
a
configuration
file
for
therefore
the
S
part,
and
that
was
like
okay,
but
I
have
to
have
two
different
sources
and
I
have
a
bunch
of
no
groups
here
and
a
good
bunch
of
groups
here
and
they're
like
oh,
we
only
implemented
a
single
source
configuration
line,
so
anything
that
you
would
join.
It
would
just
provide
the
same
s,
so
they
checked
me
again,
so
they
dead
okay.
So
we
will
skip
that
part
I'm.
Just
so.
G
M
N
N
J
I'm
Charlie
Perkins
I
just
have
a
simple
comment,
which
is:
if
the
depression
airs
long
you'll
get
a
few
responses.
If
the
questionnaire
is
short,
you'll
get
more
responses,
so
you
should
pick
out
your
top
four
features
and
then
you
can
have
one
at
the
end.
If
you're
interested
in
more
detailed
interaction
with
ITF,
please
yeah.
L
I
think
at
this
point
in
time
it
will
be
hard
to
cut
further
down,
given
how
we're
trying
to
figure
out
the
status
of
implementation
of
the
off
of
the
different
versions
and
the
only
thing
that
we
added
was
basically
this
exclude
mode
to
distinguish
whether
we
could
actually
promote
the
lightweight
version
as
the
sufficient
fall
standard
and
not
actually
IGMP
b3
and
ml
d
b2.
But
their
constraint
version
right.
So
I
think
we
already
are
trying
to
stay
very
focused
on.
Unfortunately,
only
it's
already
five
or
six
questions,
but
the.
J
I
So
I
brought
a
poem
concept.
I
had
that
kind
of
running
last
I
in
Bangkok,
there's
been
a
few
minor
refinements
but
still
been
running
since
November.
It's
reasonably
stable
is
still
doing
the
same,
be
kept
thing.
Yeah
tried
it
I
cleaned
this
up
a
little
bit.
So
it's
a
little
bit
better.
There
was
a
thing:
I
forgot
to
mention
the
crash
in
pindy
that
I
in
Fr
that
I
there's
a
patch
up
streaming.
Hopefully
at
some
point
but
yeah,
the
draft
has
cleaned
up.
I
The
RO
type
is
assigned
I,
think
I've
addressed
all
the
pending
review
comments
and
TV
DS
I
have
deployed
to
global
reverse
DNS
zones.
You
can
take
this
on
your
laptop
now
and
you
will
see
records
that
have
this.
This
type
260
field
in
it
that
point
to
some
of
my
AMT
relays
at
the
hackathon
104
I,
found
out
a
new
thing
and
I'm
gonna
be
updating
draft
slightly.
So
I'll
go
over
a
little
bit
of
what
I
found
out
and
explain.
My
plan
on
that
I
was
hoping
to
ask
for
last
call
now
now.
I
I
want
to
wait
until
after
I've
updated
it,
but
then
ask
for
a
last
call,
so
I'm
hoping
that
I
can
go
over
that.
So
what
I
did
at
this
hackathon
is
down
in
the
bottom
corner
here.
The
rest
of
this
is
basically
the
same
as
what
was
running
before
I
combined
the
ingest
router
with
the
border
router,
so
that
the
border
router
is
running
AMT
and
doing
dns
off
of
the
same
pim
join
but
I
added
then
on
the
bottom
of
the
access
router.
I
I
So
over
Wi-Fi
I'm
running
I
was
running
an
AMT
gateway
that
connects
to
a
AMT
relay
which
is
inside
the
home
router,
and
what
I
did
is
the
the
iucn
SSD,
and
this
was
the
the
activity
at
the
hackathon
I
added
DNS,
SD
service
discovery
or
sir
registration
to
the
AMT
relay
and
I
was
verifying
with
the
tools
that
do
DN
SSD
that
they
can
be
discovered.
It
turns
the
thing
I
found
out
is
that
it
relies
on
mdns
and
uses
just
the
one
hop
local
behavior.
I
That
was
an
added
and
so
I
couldn't
reach
the
thing,
and
so
I'm
still
finishing
that
proof
of
concept
as
far
as
being
directly
connected
with
a
joint
library,
but
I
do
believe,
based
on
the
DNS
SD
discovery
program
that
is
built
into
Apple
and
deployed
everywhere
already
that
that
thing
will
function
when
I
put
in
the
library
support
for
it.
So
the
interesting
part
is
that,
because
the
DNS
SD
discovery
is
broken,
there
was
another
different
kind
of
local
relay.
That
I
was
hoping
to
support
discovery
within
SSD
and
I.
I
Think
now
that
it's
not
going
to
work
because
the
access
router
cannot
push
into
the
home
router
a
search
domain
that
gets
propagated
to
the
home
in
general
in
deployment,
it
is
possible
to
coerce
or
a
home
router
into
doing
that,
but
the
default
home
router
configuration
does
not
tend
to
support
drugs.
So.
L
I'm
trying
to
remember
that
you
may
ask
Ralph
drones,
but
in
in
the
cabled
space
we
for
the
cabled
space
we
defined
in
the
ITF,
some
RFC,
which
was
basically
a
from
the
cmts
to
the
home
gateway
kind
of
a
DHCP
package
that
was
basically
carrying
DHCP
options
to
go
further
down
right
and
if,
basically,
that
discovery
is,
for
example,
including
something
like
DNS
server.
So
then
in
the
cable
space
there
might
be
a
way
to
get
these
things
through.
That's
the
only
thing
coming
to
yes,.
I
I
agree
that
is
their
and
I
agree,
I'm
pretty
sure
it's
possible
to
do
something
to
the
home
router
so
that
you
can
probably
get
it
to
happen.
What
I
believe
to
be
the
case
based
on
just
a
an
informal
poll
and
the
examination
of
the
default
configuration
on
the
home
router
that
I
brought
is
that
by
default
it
is
not
happening
at
present.
I
I
It
was
not
connected
to
a
cable
network.
No,
this
was
this
was
I
was
running
my
own
DHCP
on
a
yeah,
so
there
might
be
case
where
it
does
work,
but
my
claim
is
that
there
might
be
cases
where
it
does
not
work.
Also
now
for
cable
I
think
this
is
actually
less
important
than
for
like
dslam,
and
so,
if
there's
anything
specific
to
that,
then
we
can
go
over
it.
But
I
was
about
to
explain
the
change
that
I
proposed
to
make
to
make
a
difference
and
I
think
it
might.
I
I
Therefore
I'm
the
current
draft
says:
do
DNS
SD
n
do
Dryad
then
do
the
ne
the
well-known
anycast
address
from
the
AMT
spec
I,
want
to
change
the
order
on
that,
so
that
first
I
do
DNS
SD,
then
I
prefer
the
anycast
address,
and
you
should
put
your
anycast
address
on
that
thing.
When
you
want
to
be
discovered
by
devices
in
your
network
and
then
do
drive,
it
make
sense.
I
L
I
L
I
I
K
Actually,
the
process
is
last
call,
it
goes
to
list,
but
we
do
test
in
the
room.
Who's
read
the
the
doc
right.
Clearly,
I,
don't
think,
was
completely
ready
for
less
who
thinks
it's
closed,
rev
away,
ish!
Okay,
we
just
take
you
through
the
process,
read
from
notes,
listen
to
well
try
to
decipher
and
parse
Tauruses
words
I.
N
N
K
K
G
G
It
really
is
here
now
more
beyond
just
here's
something
that's
coming,
and
hopefully
maybe
we'll
have
something
done
by
next
IETF
and
next
so
to
illustrate
the
you
know,
it
really
is
here
now
you
can
watch.
This
is
a
actually
a
live
stream
from
the
mbone.
It
really
does
exist.
Is
a
live
multicast
stream
coming
into
here
in
my
laptop
on
the
ITF
land
that
is
not
multicast
enabled
to
demonstrate
how
all
this
works.
G
Yes,
using
a
mt.
So,
first
of
all,
let's
let's
talk
about
motivators.
Why?
Why
would
someone
want
to
use
multicast
over
the
Internet
just
just
especially
in
a
world
where
you
know
everything's
on
demand
and
who
cares
about
multicast?
Who
cares
about
live
so
first,
just
some
simple
facts
and
and
I
got
these
off
the
internet,
and
you
can
you
know,
depending
on
where
you
look,
you
can
get
different
numbers,
but
the
the
bitrate
for
high
definition
Netflix
recommends
five
Meg's
per
stream.
When
it
comes
to
4k
Netflix
recommends
25
Meg's
8k.
G
G
The
tip
this,
the
average
audience
of
an
NFL
game
is
15
million.
Simultaneous
viewers,
that's
average,
that's
a
typical
game
like
the
Superbowl
would
have
about
a
hundred
million,
but
just
your
typical
Sunday
game,
1
of
10
to
12
games
going
on
has
an
average
of
15
million
viewers.
If
you
were-
and
we
don't-
but
if
we,
if
you
were
to
move
all
15
million
of
those
viewers
to
watching
4k
over
the
Internet
again,
this
is
not
what
people
do
today.
G
They
watch
it
over
their
cable
systems
generally
and
there's
a
you
know,
smaller
subset,
who
watch
over
the
top.
But
if,
if
every
viewer
watched
over
the
top
at
4k
you'd
be
talking
about
375
terabits,
now
the
way
over-the-top
video
works
today
and
even
live
streaming,
its
unicast,
it's
all
unicast,
it's
what
I
would
call
brute
force
unicast.
It
works.
It's
fine,
but
the
question
is:
can
it
keep
up
with
these
kinds
of
numbers
at
4k
at
8k,
at
15
million?
G
Simultaneous
viewers
also
that
you
know,
we've
we've
been
going
through
a
cord
cutting
evolution
and
it
has
generally
been
people
have
been
cutting
the
cord
and
watching
movies,
and
things
like
that
on
Netflix,
but
kind
of
the
last
frontier
of
cord
cutting
is
live,
linear
TV
and
for
those
who
say
again,
no
no,
nobody
watches
live
linear
anymore.
It's
dead,
there's
a
there's,
a
long
answer,
but
the
short
answer
is
sports.
Sporting
events.
G
O
G
G
Yes,
so,
and
in
the
last
year,
or
so,
things
are
beginning
to
spring
up
so
getting
back
to
the
question
of
live
linear
whenever
somebody
says
nobody,
watches
live
linear.
The
answer
is
sports.
People
want
to
watch
sports
live
generally
speaking,
and
that
is
kind
of
the
last
frontier.
You
see
things
like
sling,
TV,
Hulu
TV
view
YouTube
TV.
These
are
growing
in
popularity.
More
recently,
so
live
linear
is
not
quite
Dead
Yet.
Maybe.
L
The
one
common
also
coming
to
mind
is
that
the
the
currently
available
TV
that
you
get
over
streaming
unicast
is
very
much
you
know
of
optimized
for
the
lowest
common
denominator,
like
you
know,
ADSL
to
16
megabits
or
so,
but
then
you
know
in
many
countries
most
cities,
you
even
get
you
know
VHDL
we
yeah
VDSL,
5200
megabits,
or
even
more
so
when
you
do
vectoring
and
then
the
question
is
okay,
why
the
heck
did
I
pay
40
more
bucks
for
this
high-speed
Internet
access?
What
better
video
quality
to
do
I
get?
L
Well,
you
could
get
all
these
better
video
qualities
but
yeah.
Sorry,
our
calculations
show
us
that
our
you
know,
IPTV
unicast
streaming
system
would
be
way
too
expensive
in
supporting
these
high
resolutions
right.
So
the
disjoint
between
ever
better
access
speeds
that
you
have,
and
you
know
the
low
quality
you
get
now
because
you
are
trying
to
make
it
work
with
unicast
I
think,
maybe
that
that
factor
of
the
excess
speed
is
actually
faster
than
what
these
systems
give
you.
If
you
don't
have
that.
Okay,.
I
K
C
Q
C
G
Another
interesting
trend-
that's
going
on
that
helps
motivate
this
is
what's
old,
is
new
turns
out,
live
streaming
is
actually
trending,
so
we
live
in
a
we
live
in
a
world
of
inauthenticity
of
of
fake
news
and
corporate
messaging.
That
could
be
ripped
straight
out
of
the
pages
of
Orwell
and
advertisers
have
begun
to
discover
that
the
the
the
sponsor
and
kind
of
uneditable
nature
of
live
streams
are
perceived
by
audiences
to
be
more
authentic
and
and
actually
drive
more
emotional
engagement.
G
Advertisers
have
found
that
viewers
spend
on
average,
and
these
are
some
numbers.
You
can
click
on
any
of
these
links
to
see
where
I
got
these,
that
that
viewers
spend
eight
times
more
longer
time
on
live
video,
then,
on
on-demand,
on
average
they've
found
that
you
know
two
thirds
of
live
audience.
Viewers
are
more
likely
to
make
a
purchase
and
more
than
half
of
marketing
professionals
say
they
see
the
best
ROI
on
live
video
and
in
the
and
and
recently
live
online.
G
G
G
G
We
all.
Have
the
people
in
this
room
know
what
went
wrong
so
I'll
skip
that
part.
So
let
me
talk
to
the
solution.
I'm
gonna
go
step
through.
What
is
what
we
can
do
about
this?
You
know:
we've
we've
talked
about
AMT,
we've
talked
about
solutions
and
more
recently,
this
is
basically
a
an
approach
that
actually
gets
multicast
out
there
and
available
for
people
to
watch
so
the
first.
The
first
step
is
getting
relays
out
there.
Well,
what
are
so
that
people
can
actually
view
this
stuff.
So
actually,
let's
go
through
each
of
these
steps.
G
G
That's
on
me
all
right,
so
relay
deployment,
so
we
started
with
this.
About
a
year
ago
we
had
there's
a
Thomas,
Jefferson,
High
School,
it's
a
magnet
school
in
Northern
Virginia.
They
had
a
10,
a
10
gig
pipe
to
i2,
which,
when
we
speak
of
the
M
bone,
you
know
the
the
M
bone
is
the
multicast
enabled
portion
of
the
Internet,
and
the
vast
majority
of
that
is
i2.
I2
is
pretty
much
to
my
knowledge,
fully
native
multicast
enabled
and
there's
some
content
out
there,
and
they
have
quite
a
bit
the
last
I
checked.
G
They
had
about
500
autonomous
systems
on
the
M
bone
that
were
multicast
enabled
which
ain't
nothing.
But
when
you
look
at
the
you
know,
I
don't
know
how
many
are
globally
routable,
it's
tens
of
thousands,
it's
a
small
number,
but
500
isn't
nothing.
It's
it's
a
significant
number
and
there's
a
fair
amount
of
content
out
there.
So
a
student
there,
a
high
school
student
William
Chang
deployed
the
first,
a
MT
relay
on
the
M
bone.
That
was
a
senior
project.
He
actually
also
to
find
all
right.
G
G
Since
then,
we've
added
a
second
relay
at
George
Washington
University,
thanks
to
Andrew
gallo,
we've
been
working
with
him
on
that
next
slide,
step
two
okay,
so
you
got
the
relays.
What
good
is
a
relay?
If
you
don't
have
a
gateway
and
the
Gateway
implementations
we
discovered
were
woefully
inadequate,
they
were
very
old
or
there
were
some
out
there
and
they
were
kind
of
proprietary.
G
But
for
the
content
we
were
trying
to
get
out.
The
the
implementations
were
woefully
inadequate
over
ten
to
twelve
years
old,
so
the
goal
was
so.
There
was
an
effort
to
build
a
modern
AMT
gateway
implementation,
not
something
that
required
you
to
hack
it
in
UNIX,
but
something
that
would
be
so
easy
that
that
your
grandmother
could
use
it,
preferably
in
the
browser
we
discovered.
So
it
an
effort
was
was
begun
by
Natalie
Landsberg
to
write
an
AMT
gateway
implementation,
she's
first
started
by
looking
at
in
browser.
G
You
know
what
could
be
done,
because
what
could
be
simpler
than
having
this
done
in
browser
in
brat,
but
we
we
discovered
quickly
that
in
browser
appears
to
not
be
possible.
Today
the
browser's
have
been
locked
down
very
significantly
for
security
purposes,
and
things
like
UDP
and
multicast
are
not
very
friendly
to
browsers
these
days.
So
she
settled
on
VLC.
G
You
know,
as
would
have
been
great
to
do
it
in
a
browser.
Couldn't
the
next
best
option
was
VLC.
Vlc
is
very
widely
deployed.
It
has
probably
the
widest
array
of
support
for
for
the
different
encodings,
a
video,
and
it's
also
great
because
it's
it's
available,
and
you
know
all
your
most
popular
Mac
Windows
iOS
Android.
There
are
available
images
for
each
of
these
things.
G
Natalie
presented
her
initial
implementation
in
Montreal
and
since
then,
her
work
has
been
taken
up
by
Wayne
Brazel,
who
has
added
some
enhancements.
Nathalie
initially
submitted
a
patch
to
VLC.
To
add
this
to
to
their
mainline
support.
The
VLC
folks
came
back
with
you
know.
It
looks
good
add
the
following:
Wayne
took
took
on
some
of
those
things.
He
also
added
some
capabilities.
One
of
the
things
that's
really
interesting
is
relay
resiliency,
so
Jake's
done
a
lot
of
work
on
relay
discovery
and
what
to
do
when
you
have
more
than
one
relay?
G
How
do
you
discover
them?
How
do
you
pick
one?
We
did
something
interesting,
Wayne
Twain
implemented
something
really
interesting
with
this.
In
that
we
essentially
took
the
two
relays.
I
gave
them
the
same,
give
them
the
same
fully
qualified
domain
name
and
two
different:
a
records
for
the
same
fully
qualified
domain
name.
When
you
do
that,
you
get
back
both
and
it's
up
to
the
application
to
do
something
with
it.
G
What
Wayne
did
was
he
receives
both
of
these
a
records
as
an
ordered
list,
and
he
joins
one
of
them
sends
the
IGMP
report
through
sets
up
the
tunnel
to
one
of
them
sends
an
IGMP
report.
If
he
doesn't
get
anything
back
after
a
second
or
so
second
or
two,
then
he
joins
the
next.
He
builds
another
tunnel
to
the
next.
One
sends
a
report
through
that
and
so
forth,
and
so
on.
So
this
gives
you
real
a
discovery.
It
gives
you
real.
G
G
A
yeah,
no,
they
don't
have
that
yeah.
So
you
can
so.
The
Wayne
has
passed
submitted
the
patch
and
he's
working
with
the
the
VLC
guys
each
time
he
submits
a
patch.
They
say
this
is
great.
Now
do
this
and
so
he's
going
through
that
iterative
process,
but
he
feels
confident
that
he's
getting
close
to
the
end
and
you
can
download
at
this
github
site,
you
can
download
VLC
binaries
to
run
on
Mac
or
Windows,
and
that's
what
we've
got
right
here
and
you
can
see
some
streams
to
try
out
and
that's
what's
been
running.
G
So
this
really
is
happening.
This
is
using
AMT
to
access
multicast
tunnels
on
the
mbone
which,
if
you
think
about
it,
you
know
think
back
15
20
years.
If
we
decide
look,
we
have
multicast
available
from
the
mbone
on
to
the
Internet
I
think
this
would
have
blown
minds
a
while
back
and
this
this
is
something
that
we've
been
working
towards
for
a.
I
G
K
L
So,
what's
what's
kind
of
the
exes
policy
on
on
the
relays
kind
of
you
know
if
more
important,
more
people
learn
about
that
and
become
overloaded.
What's
and.
G
L
G
L
G
So
that's
why
we're
using
DNS?
They
are.
We
created
a
domain
name,
m2i
cast
net
I,
think
it's
multicast
to
internet
cast
net
and
it's
real
a.m.
I
cast
net
and
the
goal
is,
as
we
add,
more
relays,
we'll
just
add
more
area
records
for
them.
Why
do
that
instead
of
the
global
anycast
address?
You'd,
say
because
that's
what
the
whole
spec
had
the
global
anycast
thing.
The
global
anycast
is
in
in
my
mind,
kind
of
a
non-starter
because
it's
basically
a
dos
attack
waiting
to
happen.
G
If
anybody,
if
anyone
can
just
advertise
and
say
hey,
I'm
a
relay,
then
you
know
that's
great,
because
anybody
can
rabbis
that
because
I'm
a
relay-
but
you
know
the
problem,
is
anyone
can
advertise
that
I'm
a
relay
and
they
may
not
be
in
relay
or
they
may
not
be
a
functioning
relay?
So
you
can
it's
a
black
hole
waiting
to
happen.
So
in
what
draft?
Which
draft
there
is
no
draft
there
yeah
we're
talking
about
an
effort
so.
G
Think
so,
there's
some
collaborative
work
that
I've
been
you
know.
Jake
and
I
have
been
talking
about,
I.
Think
it's
it
tangentially
touches
on
what
Dryad
is
trying
to
do
and
there's
there's.
You
know
different
approaches.
Yesterday,
Tom
posit
area
actually
gave
me
a
a
really
good
idea
to
handle
relay
discovery,
but
for
now
we're
just
using
plain
old
DNS,
because
it's
simple
and
it
works
and
it's
it
gives
us
gives
us
redundancy
load-balancing
and
it
works.
G
G
So
what
I
imagine
is
service
provider
a
will,
have
a
bunch
of
relays
and
they
use
one
like
anycast
address
service
brighter
B
will
have
a
bunch
of
relays
and
they
will
use
a
different
anycast
address
and
then
it's
a
matter
of
having
a
broker
or
whether
that's
if
there's
an
actual
broker
service
or
a
magical
new
protocol
or
just
DNS
picks
one
of
those
addresses,
and
some
of
them
happen
to
be
anycast
addresses
that's
what
I
think
is
the
way
things
will
go.
Mike.
M
Me
bright
so
first
kudos
to
you
for
making
this
happen.
I
love
that
you're
you've
been
using
high
school
students
and
college
students
and
they've
been
presenting
remotely
I
just
think.
That's
that's
awesome,
so
good
good
job
doing
that
at
the
beginning
of
the
presentation
you
mentioned
solutions
like
YouTube
TV,
and
things
like
that
and
I
had
one
of
these
guys
personally,
who
cut
the
cord
about
ten
years
ago,
but
just
this
year,
I
signed
up
for
YouTube,
TV
and
I.
M
Think
it's
pretty
cool,
so
I'm
kind
of
getting
back
into
the
game
to
watch
live
sports,
as
you
mentioned
so
I'm
confirming
that
so
the
reason
I
mentioned
that
is
is
Jim.
Have
you
thought
in
your
like
further
long-term,
like
with
regards
to
those
solutions?
Do
you
envision
productizing
this
through
your
company
to
make
this
available,
perhaps
to
those?
So
this
is
available.
G
Now,
and-
and
that's
kind
of
the
point
is
demonstrating
this
stuff
I,
so
let
me
let
me
get
to
actually
let
me
there's
only
a
couple
minutes,
and
hopefully
this
might
answer
your
question
step
three
and
step
four,
and
then
that
you'll
see
where
we're
going
with
this
so
step.
Three
is
a
Content
portal.
So,
if
you
remember,
when
William
wrote
that
script,
he
came
up
with
here's
the
40
multicast
groups
from
these
you
know
and
120
119
different
sources.
Let's
take
that
and
you
know
it's
a
text
file
right
now.
G
Let's
turn
that
into
some
kind
of
magical
portal,
where
you
can
click
on
each
of
these
things
and
it
launches
VLC.
This
VLC,
with
all
the
you
know,
requisite
information
for
VLC
to
do
it
needs
to
do
right
now.
You
have
to
you,
know,
cut
and
paste.
So,
rather
than
doing
that,
that's
kind
of
the
next
step
is
the
multicast
portal
next
slide
and
then
the
big
thing.
G
This
is
what
can
actually
be
impactful,
because
up
until
now
is
just
kind
of
a
you
know,
a
nice
little
proof
of
concept
and
you
know,
but
off
net
sourcing.
So
right
now
we
have
on
net
sourcing,
which
is
you've
got
to
be
on
I
to
you've,
got
to
be
on
the
embo
and
to
source
something.
What
if
I
could
take
my.
A
G
Start
recording
something
and
sending
it
to
some
magic
server
that
sits
on
the
mbone
and
it
could
take
your
unicast
stream
coming
directly
from
your
phone
and
turns
it
into
a
multicast
stream.
That
folks
could
then
watch
on
the
mbone
natively
or
off
the
mbone
via
a
mt
via
these
gateways,
that's
off
net
sourcing
and
that's
kind
of
when
we
can
get
this.
This
is
kind
of
the
big
Holy
Grail,
because
this
can
make
content
that
isn't
existing
on
the
internet
now
today.
G
So
if
anybody
here
is
interested
and
you
want
to
get
involved
first,
you
know
we'd
love
to
have
other
people
joining
the
effort,
content,
content,
content
content
is
king.
The
more
interesting
content
we've
got
right
now.
Is
these
just
Creative
Commons
open
sourced
content?
That's
out
that
andrew
is
streaming
plus
the
stuff,
that's
on
the
mbone,
but
if
anybody
has
interesting
content
that
they
want
to
get
out
on
the
mbone
we'd
love
to
we'd
love
to
have
it
further
development
of
VLC,
I
think
we're
converging
on.
G
You
know
most
of
what
we
need
in
VLC,
but
then
the
next
big
thing
is,
you
know
in
in
browser
implementation
and
speaking
with
Jake.
I
know
this
is
something
that's.
He
has
some
some
grand
visions
for
that.
There's
the
portal
that
I
talked
about
step,
3,
adding
more
relays
adding
relays
is
easy.
You
know
it's
like
one
line
of
config
on
a
router
if
it
supports
it,
that's
already
there
so
and
then
off
net
sourcing.
Those
are
the
big
things.
G
G
L
G
But
because
go
to
slides
from
now
here's
the
answer
to
your
question.
Okay,
next
slide
here,
here's
a
references,
so
you
can
find
all
the
information
that
you're
looking
for
this
great
github
link
has
the
has
the
am:
has
the
the
builds
for
a
MT
for
VLC
VLC
builds
for
both
Mac
and
Windows,
and
it
also
has
all
the
information
that
you
might
want
to
know
about
the
relays,
so
you
can
start
using
those
relays.
O
G
E
N
N
N
Set
up
a
relay
that
is
announced
everywhere:
mm-hmm
doesn't
work.
People
set
up
relays
on
ten
mega
overloaded
machines
that
just
doesn't
work.
It's
just.
You
can't
do
it
so
six
Rd
he
was
actually
deployable
because
then
you
use
provider
space
for
for
the
same
function
yeah
and
this
works,
because
now
you
had
you're
talking
to
a
specific
illness
or
a
specific
like
organization,
that's
maintaining
those
relays,
so
you
I
would
say.
Okay,
you
can
try,
no
put
the
zero.
G
G
G
Well,
if
you
help
us
with
the
portal,
then
you
can
find
it
yeah.
So
that's
it
any
other
comments.
We'd
love
to
have
folks,
oh,
go
to
the
last
this
previous
slide
to
this.
Okay,
if
you're
interested
in
any
of
this
I
can
be
reliably
found
in
the
bar
at
late
late
at
night
here
tonight.
So
come
see
me
if
you
want
to
get
involved,
but
in
the
meantime
try
it
out
check
it
out,
use
it
and
that's
it.