►
From YouTube: IETF105-MOPS-20190722-1330
Description
MOPS meeting session at IETF105
2019/07/22 1330
https://datatracker.ietf.org/meeting/105/proceedings/
A
D
We
go
all
right.
Thank
you
for
coming
to
the
media
operations
Boff.
My
name
is
Lesley
Daigle
co-chairing
with
blending.
The
blue
sheets
are
now
in
circulation.
Please
do
fill
them
out
and
in
the
meantime
please
note
well
standard
note
well
for
IET
have
activities
I'll,
give
you
a
moment
to
read
it
all.
I
will
not
sing
it.
D
Alright,
this
is
our
agenda.
For
today
it's
pretty
much
well,
it's
the
same
as
the
one
that
is
posted
in
the
meeting
materials.
We
have
a
jabber
scribe
that
stock
and
Kyle
Rose
has
volunteered
have
been
volunteered
just
to
take
notes:
Thank
You
Kyle.
Thank
you
Matt,
any
particular
bashes
to
the
agenda.
At
this
point,
everybody
ready
to
go
all
right
so
by
way
of
introduction.
D
Just
the
purpose
of
this
particular
Boff
is
to
highlight
that
there
are
many
existing
video
activities
ongoing
in
the
IETF,
as
well
as
many
activities
ongoing
outside
of
the
IETF
using
IETF
technologies.
To
do
the
stuff
and
aware
of
these
facts
and
having
seen
some
sort
of
cross-pollination
and
failure
to
cross
pollinate
over
the
years,
we
we
in
the
group
of
interested
parties
here
have
for
a
while
been
proposing
that
there
should
be
some
kind
of
an
IETF
activity
wearing.
D
D
There
is
no
specific
one
specific
area
where
all
the
work
gets
done:
it's
everything
from
transport
to
art,
to
int
and
everywhere
else,
and
so
there's
no
clear
single
home
for
it,
which
is
why
we've
landed
here
today
in
in
an
effort
to
try
to
coordinate
some
of
those
discussions
and
highlight
where
there
are
issues.
It
started
to
sound
a
little
bit,
like
maybe
an
operational
issue,
or
at
least
its
operational
issues
that
are
driving
this
discussion.
So
perhaps
there's
scope
for
an
Operations
working
group.
D
E
Hi
everybody
I'm
Glenn
Dean
from
Comcast
NBC,
Universal
I,
didn't
want
to
boom
in
your
ears
and
wake
you
guys
up.
I'm
Glenn
Dean
from
Comcast
NBC
Universal
we're
just
gonna,
pull
the
presentation
up
here,
closer
even
closer.
Oh,
my
gosh,
okay
I
will
be
loud,
I'm
Glenn
D
from
Comcast
NBC
Universal
once
again
and
we're
just
waiting,
no
pressure.
I
didn't
say:
I
did.
D
E
So
I'm
going
to
do
the
sort
of
a
historical
context
setting
for
some
of
this
stuff.
There's
two
things:
I'm
going
to
show
two
simple,
slides:
the
first
one.
Is
this
there's
a
draft
if
you
go
the
the
data
tracker
for
this
Boff,
there's
a
draft
that
sort
of
has
the
detail
here,
but
I
wanted
to
bring
a
tension
to
the
room
to
some
work
done
over
at
sim
tea.
Cynthia
is
a
society
of
motion
picture
and
television
engineers.
E
These
guys
do
professional
video
and
for
the
last,
while
they've
been
really
hard
at
work
on
defining
their
transition
plan
to
ID
networks
and,
of
course
they
sent
you.
What
that
means
is
that
they're
converting
all
the
stuff
they
do,
which
used
to
use
other
transport
technologies
into
IP
based
stuff,
that's
all
depending
upon
an
ITF
RFC's,
and
so
what
you're
gonna
find
there.
E
If
you
go
look
at
that
draft,
it's
just
it's
just
a
normative
list
of
the
top-level
stuff
they
depend
upon
and,
of
course,
below
all
of
those
references,
there's
references
and
references
or
references.
The
takeaway
here
is
that
this
is
the
tip
of
the
iceberg,
but
the
professional
video
guys
are
moving
to
being
built
in
Thapa
for
everything
they
do
on
top
of
IP
networks
and
as
such.
What
they're
discovering
is
that
there's
a
lot
of
things
that
work
really
well
with
that
we've
done
here,
the
ITF
and
there's
a
few
things.
E
Sometimes
that
are
little
rough
edges
and
so
part
of
the
historic
past
year
has
been
bringing
some
of
that
experience
operationally
from
those
rough
edges
back
into
the
ITF
as
lessons
learned
and
turning
them
back
into
feedback
and
comments
and
new
work
for
various
working
groups.
It's
a
very
broad
set
of
working
groups,
transport
routing
addressing
security,
has
been
stuff
that
they've
all
recently
been
interested
in
touching
so
part
of
the
the
thing
that
somewhat
motivated
getting
us
here
today
is
work
like
this.
E
F
E
So
the
blue
bars
here
are
the
number
of
people
that
went
to
the
to
these
sightings
directly
and
the
orange
bar
is
the
number
of
organizations
that
were
represented
at
that
particular
meeting.
There's
a
few
data
points
missing
areas
we
didn't
record
very
well,
but
the
takeaway
here
also
is
back
at
IDF
98.
We
decided
to
hold
a
little
before
he
said:
hey
if
you're,
just
an
Nvidia
with
the
ITF
come
meet
us
at
this
bar
over
in
Chicago,
while
we're
all
there
and
I
put
that
out
on
the
list
of
attendees
about
7:00
p.m.
E
on
Friday
by
midday,
Saturday
discovered,
there's,
so
many
people
have
stopped
me
in
the
hallway
said.
That's
an
important
topic
to
me:
I
want
to
talk
to
you
about
it
that
we
realized
we
wouldn't
be
able
to
go
to
the
bar.
We
were
gonna
go
to
so
we
ended
up
in
a
bigger
room,
the
bigger
room
held
about
70
people.
In
the
end
we
were
in
the
room
we
filled
them
in
to
capacity.
E
There
was
three
me
out
into
the
hallway
and
a
rough
head
count
of
the
times
that
we
had
about
10
percent
of
the
ITF
attendees
at
that
ITF
98
we're
trying
to
get
into
that
Boff
just
to
talk
about
the
topic
of
video,
so
a
lot
of
passion
there
in
the
end
of
62
people
successfully
actually
were
able
get
their
name
on
the
blue
sheet.
But
we
had
a
lot
more
going
on
one
of
the
lessons
we
took
it
away
from
that.
E
First,
attempt
to
sort
of
say
are
people
interested
was
that
there
was
a
lot
of
people
interested,
but
we
needed
to
sort
of
concentrate
the
conversation
down
a
little
bit
to
get
organized
and
get
our
heads
around
some
of
the
topic
space,
and
so
we
sold
the
people.
Well,
there's
this
other
male
in
this
3.
E
It
had
GG
at
ITF
if
you're
interested
go
sign
up
to
that,
and
that's
we're
gonna
put
out
the
call
for
more
ongoing
video
and
just
group
side
meetings,
and
so
we
did
that
and
we
ended
up
into
the
smaller
group
which
we
did
advertise
widely,
but
we
didn't
hide
it
either,
but
it
allowed
us
to
spend
time
talking
about.
You
know
where
we,
where
we
going
and
end
up
that
there's
a
series
of
presentations,
people
cadenced,
are
doing
as
a
regular
part
of
this
I
me
they
weren't
really.
Well.
Those
in
fact
are
archived.
E
E
G
All
right,
thank
you
good
afternoon.
My
name
is
Sanjay
Mishra.
So
I
start
this
presentation
talking
about
an
industry
consortium,
but
more
so
from
the
point
of
view
that
what
happens
in
IETF
does
not
stain
IETF.
It
actually
goes
out
in
a
real
world
for
some
real
world
implementations,
specifically
focusing
on
the
Internet
community
and,
by
the
same
token,
the
work
that
goes
out
into
the
real
world.
G
There
are
implementations
in
the
real
world
that
actually
may
have
feedback
into
the
IETF,
based
on
using
the
standardization
work
and
based
on
what
the
real
world
implementation
is
teaching,
and
some
of
that
may
actually
come
back
into
the
IETF.
So
the
the
intent
here
is
to
kind
of
give
you
an
example
of
one
such
symbiotic
relationship
of
where
we
take
the
work
from
IIT
F
and
then
potentially
look
to
feed
back
into
the
IETF
alright.
So
that
said,
let
me
just
kind
of
walk
you
through
with
one
such
example
of
what
I
just
said.
G
So
the
streaming
video
Alliance
is
really
is
an
ecosystem
really
focusing
on
single
aspect
of
addressing
the
over
the
top
streaming
at
scale
and
then
within
the
SVA.
There
are
several
different
working
groups,
but
I'm
just
talking
about
only
one
working
group
here
right
now,
which
is
the
open
caching
working
group
and
the
open
caching
work.
Working
group
is
primarily
focused
on
interoperability
amongst
various
players
that
includes
the
content
providers,
commercial
Sirians
technology,
vendors
and
the
Internet
service
providers.
G
G
So
I
talked
about
interoperability,
so
really
the
open
caching
compliance
solutions,
they're
not
only
based
on
the
CDN
I
RFC's,
but
based
on
the
real-world
implementation,
there's
also
a
feedback
into
the
CDN
I
working
group
that
hey
we've
done.
We've
done
implementation
based
on
CD
and
I
RFC's,
but
we
also
sees
see
areas
of
improvement
where
we
can
feed
back
into
the
IETF.
G
Now,
in
this
particular
case,
the
the
feedback
mechanism
was
fairly
straightforward
in
the
sense
that
we
have
taken
the
RFC's
from
CDN,
I
working
group
and
then
identified
areas
within
those
RFC's
where
we
think
there
are
improvements,
so
we
can
feed
them
back.
Oh,
there
was
a
very
straightforward
process
in
terms
of
knowing
where
to
go
back
to,
but
sometimes
that
may
not
be
the
case
as
to
not
knowing
where
we
need
to
go
back
to
so
really.
G
There
are
a
bunch
of
enhancements
that
make
the
protocol
much
more
agile
in
terms
of
what
functionality
you
can
add
to
it,
and
you
can
read
the
drafts
in
the
there
should
have
been
a
link
in
the
draft.
Anyway,
you
can
look
up
these
drafts
in
the
in
the
working
group.
So
having
said
that,
so
we
in
this
case,
we
clearly
see
interoperability
and
we
know
whether
whether
feedback
is,
but
sometimes
that
may
not
be
very
clear.
G
So,
within
the
open
caching
working
group,
we
have
a
very
clearly
defined
path
and
we
know
some
of
the
feedback
that
we
can
provide
to
and
we're
also
looking
at
additional
opportunities
where
we
know
we
can
leverage
the
work
which
is
going
on
in
the
IETF
in
certain
working
groups
HTTP.
This
is
of
one
of
the
working
group
that
we
we
see
a
couple
of
work
items
there
that
are
very
applicable
to
open
caching
and
we're
going
to
be
looking
at
lifting
and
reusing
as
much
work
as
we
can.
G
There's
also
issues
on
the
certificate
management,
and
we
are
closely
monitoring
some
of
the
work
that's
going
on
in
the
TLS
working
group
and
as
well
as
and
the
Acme
working
group
and
there's
a
draft
also
in
CDN
our
working
group,
that
sort
of
bridges.
What
are
the
specific
commands
you
would
need
to
use
in
order
to
use
either
the
sub
source
or
the
star
based
delegation
method
between
the
two
CDN
that
are
trying
to
use
one
of
those
mechanism
for
delegation.
G
G
There
are
things
that
we
are
looking
at,
so
basically
one
of
them
is
just
the
consistent
interpretation
of
the
video
players
on
the
browsers
or
any
other
way
where
the
browsers
do
not
necessarily
not
the
browser's,
but
the
video
players,
I
should
say,
do
not
necessarily
always
have
the
common
interpretation
of
what
has
been
laid
out
in
the
RFC
s.
So
some
of
the
work
is
really
more
practical
and
it's
not
something
that
the
necessarily
need
to
change,
but
maybe
there's
more
priority.
G
That
needs
to
be
added,
so
we're
looking
at
some
of
the
work
and
see
how
we
can
bring
that
back
as
to
what
clarity
is
that
we
think
is
needed.
There
are
other
issues
also
with
respect
to
bit
rate
shifts
and
the
retransmissions,
because
those
both
impact,
the
QE
and
then
the
360
virtual.
We
are.
We
are
looking
at
several
different
mechanism
of
how
best
we
can
address
the
latency
issue,
because
motion
to
photon
is
a
big
issue
and
latency
is
a
big
big
part
of
it.
G
So
so
that
actually
brings
to
our
intent
as
to
why
mops
and
we
think
that
mops
provide
provide
us
a
common
ground
where
we
can
have
a
soft
landing,
bring
some
issues
and
then
seek
direction
as
to
where
we
need
to
go.
Is
there
a
working
group
that
is
addressing
that
issue
or
or
is
there
a
working
group
that
may
be
willing
to
take
that
problem
that
we
bring
to
as
a
as
an
issue
that
they
want
to
work
on,
or
maybe
establish
a
new
working
group?
G
H
H
E
E
You
know
capture
it
in
IDs
that
eventually
may
fall
into
other
working
groups
at
the
ITF,
but
they
will
have
already
been
a
little
bit
more
mature
with
sort
of
a
video
technology,
expertise
added
onto
them,
because
in
the
past,
a
lot
of
times
that
when
people
bring
things
in,
they
sometimes
hit
a
group
and
the
group
doesn't
quickly
what
to
do
with
them
or
consume
with
because
they
aren't
quite
ready.
Yep
and
so
sort
of
the
thought
here
is
is
try
to
fight
away,
as
Sunday
said,
provide
a
soft
landing
place
for
them.
I
G
I
G
So
the
what
I'm
looking
here
is
that
there
are
areas
where
SVA
can
basically
forklift
work
that
already
exist.
So
these
we
see
as
opportunity
clearly
there's
some
work
that
we
can
reuse.
We
don't
have
to
reinvent
the
wheel
when
there's
work
that
already
exists,
so
just
like
the
HTTP
cache
had
a
response.
I
think
that
something
is
very
useful
for
SVA
members
to
be
able
to
use
that
work.
So
I
think
those
are
very
clearly
defined
paths.
G
To
be
able
to
partner
with
IETF
and
and
reuse
the
work,
and
then
the
challenges
are
some
other
things
that
you
know
may
or
may
not
have
a
clear
path
for
us
to
go
to.
We
don't
know
whether
it's
really
a
HTTP
working
group
issue,
the
the
RFC
3986
that
I
reflect.
We
have
had
some
conversations
on
that
with
the
HTTP
working
group
and
the
RFC
does
not
say
anything.
The
RFC
is
clear,
but
it's
it's
the
in
the
interpretation
of
what
you
and
what
the
implementation
is.
G
That
might
exactly
not
mimic
what
the
RFC
was
originally
intending
to
say.
So
those
are
like
a
little
bit
of
a
finer
issues
that
may
not
necessarily
have
that
requires
RFC
3986
to
be
revised,
but
maybe
they
can
be
in
a
helpful
document
that
sort
of
publishes
that
what
might
be
the
guidelines
to
interpret
that.
So
there
is
in
a
an
area
where
we
think
we
might
be
able
to
write
something
and
bring
it
back
to
the
IETF.
Because
that's
a
challenge,
that's
something!
That's
not
a
clearly
defined
solution
that
we
see
out
there.
E
That
this
is
good
for
ESPE
a
but
what's
in
it,
for
the
IGF,
if
you
put
this
together
with
the
slide,
I
showed
about
the
word
that
simply
is
also
doing
on
the
I'm
moving
to
ip-based
networks.
These
are
clearly
groups
that
are
making
use
of
the
work
we
do
in
here
at
the
ITF
and
instead
of
saying
to
them.
Well,
hey,
why
don't
you
go?
Do
IGF
networking
work
over
in
your
place
and
then
bring
it
back
here?
Potentially,
it
seems
to
me
please
to
me.
E
G
So
there
are
two
benefits
for
ITF.
You
know
why
one
is
clearly,
as
you
see
here,
that
there's
we
have
you
reuse,
the
RFC's
and
dividend
defied
areas
of
improvement,
so
I
think
that
really
feeds
back
into
the
IETF,
and
second
thing
is
that
we
still
have
issues
that
that
need
to
find
a
home,
and
it's
not
very
all
it's
not
always
easy
to
know
where
to
go
exactly.
In
this
case
it
was
very
straightforward
where
we
needed
to
go.
J
I'm
Justin
Uberti
from
Google
I've
worked
on
the
WebRTC
domain
for
quite
some
time,
but
Google
we
do
a
lot
of
video
streaming
and
some
of
the
video
streaming
we're
all
in
the
WebRTC
world
is
starting
to
come
together.
I'll
sort
of
talk
about
some
of
the
things
they're
happening
there.
We
just
talked
in
the
previous
presentation
about
some
of
the
challenges
and
one
of
them
is
how
to
deal
with
low
latency
environments.
J
It
doesn't
really
matter
that
much
what
the
actual
latency
is,
and
we
start
seeing
some
of
the
challenges
and
things
like
online
sports,
where,
like
people,
perhaps
in
an
apartment
complex,
you
know,
have
the
experience
of
hearing
their
neighbors
yelling
out.
You
know
cheering
before
they
actually
see
what's
happened,
and
this
can
happen
just
because
of
the
physics
of
like
okay
where's,
their
playhead
versus,
like
their
neighbors,
did
they
have
packing
loss
or
to
their
neighbors,
and
so
there's
some
desire
to
try
to
move
everything
toward
the
right.
You
know
not.
J
Every
situation
requires
this,
but
there
are
many
cases
where
in
low
latency
is
just
important,
is
actually
critical,
next
slide
or
right
here,
alright.
So
you
know
this
is
our
typical
streaming
and
for
those
who
have
you
have
been
not
asleep
the
past
few
years
like
this
is
now
like
the
vast
majority
of
all
traffic
on
the
Internet.
J
Is
you
know
online
video,
and
so
there's
some
technical
stuff
here
about
exactly
how
the
stuff
is
coded,
but
it's
gently
coded
for
like
maximal
picture
quality
and
for
easy
you
see,
though
they
jump
into
the
streamed
referred
to
BC
Keable.
You
know
really
at
any
point,
and
so
there's
this
GOP
group
of
picture,
sighs,
there's
a
frames
that
depend
on
previous
frames.
We
were
called
bi-directional
frames
and
the
way
this
generally
works
is
that
you
know
the
player.
The
stream
are
basically
sending
out
chunks
of
a
video.
J
The
player
then
downloads
them
ahead
of
time
once
it
feels
like
it.
Has
enough
video
source
stored
up
a
then
actually
can
start
playing
that
out
and
then
keeps
downloading
more
and
more
chunks
to
try
to
keep
you
know
its
buffer
full
and
make
sure
that
the
the
user
never
see
is
like
a
frozen
picture
because
it
has
a
head
and
underflow.
What
this
means,
though,
is
it
needs
to
build
up
a
pretty
comfortable
amount
of
data
to
make
sure
that
underflow
never
occurs?
J
Usually
this
stuff
is
being
transmitted
over
TCP,
which
means
that
there's
head-of-line
blocking
and
so
I
won
lost
packet
can
cause
the
stream
to
stall,
and
so
the
player
need
to
be
pretty
conservative
to
basically
build
up
this
buffer
to
make
sure
it
never
runs
out
of
data.
Now.
One
of
the
great
things
here
is
that
CD
ends
work
really
well
here,
there's
a
lot
of
things
and,
like
you
know,
an
ITF
there's
more
FCS
for
this,
where
basically,
the
streamer
can
just
start
copying.
J
Basically
chunks
to
the
CD
ends
or
CD
ends
can
be,
you
know,
pulling
them
from
from
the
streamer
and
there's
really
not
a
whole
lot
of
logic.
That's
involved
here,
you
know,
so
the
client
wants
a
block,
it
asks
for
from
the
CDN
C
and
goes
and
post
in
the
stream
or
caches
it.
Then.
The
next
client
that
comes
along
can
get
that
exact
same
block.
J
You
know
the
over
GOP
size,
the
buffer,
it
all
turns
into
multiple
second
stuff.
Up
and
I
can
see
you
know,
and
that's
even
in
the
good
case,
so
necks
are
moving
on
the
there's,
a
certain
growing
thing
of
like
these
low
latency
scenarios,
and
so
twitch
people
are
streaming
their
game.
They
want
to
talk
about
exactly
what's
happening
as
they
see
it.
They
don't
want
to
basically
say
you
know,
be
chatting
about
something
and
have
it
be
like
five
seconds
behind
real-time,
where
they're
talking
about
something
in
the
past.
J
J
You
know
this
is
you
know,
a
category
that
I
think
is
starting
to
ramp
up
and
there's
a
reason
that
number
reasons
why
it's
complicated
and
it's
taking
a
while
to
get
going,
but
I
expect
to
see
this
continue
to
increase.
There
are
a
lot
of
things
that
are
technically
different
here.
The
way
the
stream
is
constructed
is
different.
There's
no
typically,
no
group
of
pictures,
there's
no
bi-directional
frames,
everything's
designed
to
be
the
minimal
and
latency.
J
J
You
know
RT
M
a
few
actually
used
by
bi,
HQ
trivia,
and
these
things
all
you
know
they
can
basically
understand.
You
know
what
is
the
bandwidth
between
the
streamer
and
the
client
very
easily,
and
they
can
also
have
the
the
status
of
every
single
frame.
That's
being
delivered,
every
single
packet
that's
being
delivered
and
can
community
the
client
communicate
that
back
and
they
can
know
if
there's
losses
that
they
need
to
re-encode
a
new
frame
based
on
what
the
client
knows
that
has
received,
or
if
has
to
be
transmitted
or
PI,
error,
correction,
etc.
J
It
needs
to
be
a
little
more
complicated
than
the
typical
TCP
streaming.
There's
often
no
playout
buffer
here
that
you
know
if
it
runs
out
of
data,
then
it'll
just
sort
of
try
to
unlike
send
a
message
back
to
get
a
new
frame
and
continue
on
it
won't
try
to
build
up
this
big
buffer,
so
it
can
have
a
very
small
amount
of
latency,
but
the
main
thing
is
because
of
all
of
this
you
know
the
existing
CD
and
Technology
does
not
work
super.
J
Well
that
you
know
there
are
all
these
sort
of
protocols
that
are
different
than
what
CDNs
are
used
to
deal
with,
dealing
with
and
there's
no
simple
ability
just
copies
and
files
that
really
any
of
these
edge
streamers
need
to
be
able
to
estimate
the
bandwidth
down
to
the
client.
They
need
to
deal
with.
What
happens
that
that
client
is
now
missing
a
picture.
J
They
need
to
basically
be
more
interactive
than
today's
CTN
streamers,
but
what
you
can
do,
if
you
do
all
the
stuff
right,
is
you
get
sub-second,
latency
and
so
stadia
is
perhaps
the
most
extreme
example
of
some
of
this,
a
typical
PC
game.
You
know
in
the
best
configured
case,
the
delta
between
human
input
and
stuff
being
displayed
on.
The
screen
is
about
100
milliseconds
and
in
the
typical
deployments,
where
you
see
someone's
game
console
and
it's
hooked
up
to
some
TV.
J
J
You
know
to
the
device
which
then
plays
it
out
just
like
any
other
sort
of
low
latency
video
stream,
and
so
that
that
just
like
it
can
be
super
simple,
it
can
be
a
streaming
TV
stick,
it
can
be
a
mobile
device,
it
can
be
a
Chromebook.
It
can
really
be
anything
that
can
just
display
video.
It
doesn't
have
to
have
a
super
high
end
GPU,
which
has
been
like
you
know
the
requirement
for
high
end
gaming.
You
know
up
until
now,
I
just.
J
The
150
milliseconds
to
end
end
is
that
the
round-trip
time
essentially
like
put
and
then
the
time
between
you
hit
a
button
until
the
time
you
see
your
character,
jump
or
move
ever
so
right
now.
Sort
of
you
know,
a
lot
of
video
streaming
is
happening
in
the
sort
of
single-digit
megabit
range
you
see
like
720p
at
maybe
three
or
four
megabit.
You
see
1080p,
maybe
like
six,
eight,
maybe
even
10
megabit
on
the
minimum.
First
idea
right
now
is
15
megabit
and
they
sort
of
recommended
this
25
megabit
and
that's
for
1080p
30
streaming.
J
The
expect
expectation
is
that
this
will
get
to
4k
60
and
you
know
they
even
at
the
city
or
in
launch
meant
they
talked
about
doing
8k
120,
which
would
be
you
know,
a
200
megabit.
You
know
sort
of
range
so
on
this
is
kind
of
moving
things
and,
like
a
you,
know,
a
bit
further
down
the
sort
of
latency,
critical
and
high-end
performance
range.
So
how
does
stuff
all
work?
It's
pretty
complicated.
E
E
J
Think
it's
not
there's
some
lumpiness,
but
yeah
yeah.
It's
it's
not
like
you
get
it
all
at
once
and
then
plays
out.
It's
like
it's
continuous,
so
staging
runs
over
WebRTC
and
there's
a
bunch
of
different
things
that
happen.
Like
the
actual
app
you
know.
Does
our
pcs
over
HTTP
the
video
is
sent
over
typical
WebRTC,
which
is
SRTP
over
ice.
Then
the
game
on
input
is
sent
up
over
a
data
Channel.
This
also
kind
of
is
lost
tolerant.
J
So
it's
got
Iceland
DTS
SCTV,
which
is
basically
that
the
normal
web
RTC
set
of
protocols,
but
there's
not
a
lot
of
off-the-shelf
stuff
you
can
use
here.
Http
has
done
some
wonderful
things
about
all
sort
of
load,
balancers
and
like
front-end
gateways
and
other
things
that
you
know
for
basically
scaling
up
a
service
and
not
just
stadia,
but
really
everybody
is
doing
this
by
hand
yeah.
They
have
to
build
like
their
own
server
stacks
the
handle
WebRTC
and
it's
it's
not
a
simple
static.
People
are
having
to
doing
a
lot
of
work
here.
J
C
High
risk'
from
cable
labs
you've
focused
on
the
protocol
itself
getting
itself
low
delay,
but,
of
course
it's
there's
potentially
other
traffic
there
and
this
cueing
that
causes
delay
no
matter
how
good
the
protocol
is,
and
so
I
would
have
thought.
That's
more
sort
of
media
rock-type
problem
than
sure
I
mean
if
you
don't
have.
I
J
I
point:
some
of
these
things
are
structural,
that
even
if
your
pipe
is,
you
know,
you
know
you
don't
have
a
lot
of.
You
know
other
traffic
there.
Some
of
these
things
just
how
the
videos
been
coded,
how
it's
being
sent
down
you're,
just
not
gonna,
get
the
performance
you
want,
and
so,
even
in
the
best
case,
you're
kind
of
a
done
lock.
Certainly
if
things
are
not
ideal,
even
if
you
have
you
know
some
other
technology,
it's
still
not
gonna
work
right
either.
C
For
four
years
on
that-
and
it's
nearly
finished,
you
know
with
queuing
delay
for
all
traffic
in
the
sub
millisecond
range.
So
you
know
I'm
Matt,
I'm
from
cable,
apps,
cable
industry's
gonna
be
deploying
it
next
year.
There's
all
the
specs
are
written
and
everything
so
I'm.
Just
sort
of
surprised,
don't
know
about
that.
C
L
J
Mean
there's
a
few
things
that
are
all
combined
here.
One
is
that
for
the
Netflix
case,
your
Netflix
done
an
amazing
job,
basically
trying
to
get
the
most
of
every
bit
and
they
can
basically
go
through
and
recode
like
every
single
chunk.
You
know
with
a
bunch
of
different
quantization
other
parameters
to
get
like
you
know
the
ideal
setting,
because
it's
like
once
they
do
any
sort
of
VOD
thing
they
don't
need
to
do
it
again,
whereas
like
for
stadia
like
every
single
frame
is
like
the
first
time
is
probably
ever
been
seen
ever.
J
Also,
like
you
know,
there's
being
done
by
hardware
said
of
our
encoders.
The
typical
sort
of
psnr
of
hardware
encoders
tends
to
be
a
little
bit
more
less
then,
like
you
know,
the
best
offline
are
the
best
no
offline.
You
know
software
encoders
and
there's
also
just
very
low
tolerance
for
artifacts,
like
an
in
this
scenario
too,
and
so
the
bit
rates
are
a
little
bit
higher
for
the
same
resolution.
Just
given
like
some
of
those
limitations.
M
I
want
to
point
out
that,
in
order
for
these
applications
to
work
really
well,
we
have
to
have
an
l7
and
an
elf
for
that
and
below
that
all
agree
to
do
the
right
thing
and
having
an
awesome,
l4
and
an
l7
that
doesn't
adapt
or
that
doesn't
work
well
across.
The
multiple
different
surfaces
that
we
want
to
attach
to
is
still
going
to
cause
a
whole
lot
of
pain,
and
that's
one
of
the
slides
that
Justin
presented
earlier
right,
so
I
mean
these.
Things
are
whole
stack
things
because
they
are
whole
stack
things.
N
So
what
I'm
going
for
here
is
a
doc
I
wish
I
had
a
few
years
back,
and
the
point
is
to
try
to
kind
of
cover
the
video
issues
that
you
encounter
in
networking
and
to
give
people
a
kind
of
overview
into.
What's
what
we're
looking
for
and
to
to
be
a
when
you're,
making
a
new
solution
to
be
something
you
can
point
at
to
say,
I'm
solving
this
problem
in
this,
that
kind
of
thing,
I
thought
that
would
be
hopefully
useful.
N
So
as
an
overview.
The
the
the
current
draft
has
the
first
two
sections.
I.
Think
bandwidth
is
the
biggest
thing
that
bandwidth
requirements
of
video
and
media
in
general
on
the
Internet.
Today
you
know,
as
as
Justin
showed
it
it's
the
vast
mnDOT's,
the
vast
majority
of
current
traffic,
but
that's
only
the
tip
of
the
iceberg
for
the
demand
for
the
traffic
so
address
that
a
little
bit
and
I
talked
a
little
bit
about
the
congestion
avoidance
strategies.
N
I
think
there's
a
couple
other
sections
that
could
be
added:
I
love
feedback
on
you
know,
weather
conferencing
should
be
a
part
of
this,
and
and
maybe
the
kind
of
low
latency
use
cases
just
heard
about,
and
then
I'd
love
to
have
kind
of
a
pointer
to
a
page
where
we
could
maintain
a
set
of
implementations.
I
think
that
would
be
a
valuable
contribution
to
to
get
people
up
to
speed
when
they
try
to
enter
this
space
or
they
find
a
need
to
engage
with
it.
N
So
the
banner
with
requirements
I
go
over
just
the
kind
of
perceiver
what's
necessary.
I
have
a
formula
that
I
pull
that
out
a
thing
you
know
how
replicating
the
the
traffic
addresses
that
and
the
you
know
the
kind
of
different
ways
of
replicating
with
caching
and
multicast
some
of
the
effects
and
challenges
that
go
with
that
and
a
suggestion
from
oli.
N
That's
not
in
there
yet
is
the
live
versus
on-demand
and
also
covering
asymmetric
links
and
the
to
reprise
kind
of
the
motivation
behind
this,
a
few
slides
from
from
Prague
the
you
know
when
you
just
do
the
math
72
said
the
Akuma
is
one
of
the
you
know.
Major
CD
ends
on
the
internet.
They
published
in
December
they're.
Their
latest
record
was
72
terabytes
per
second
delivered
to
end-users,
and
when
you
just
take
the
40
megabits
required
to
send
a
4k
standard
traffic
and
you
do
the
division
you
get
like
1.8
million.
N
Simultaneous
viewers
is
what
can
be
supported
at
that
kind
of
traffic
level
to
set
a
new
record
if
the
entire
platform
is
dedicated
to
sending
that,
and
you
know,
if
you,
if
you
look
at
like
how
many
viewers
is
1.8
million
viewers-
that's
like
that's
the
audience
size
for
the
number
179
Nielsen
rated
program
from
from
last
year,
so
it's
just!
It
doesn't
really
touch
on
the
what's
needed
today
and
what's
needed
even
going
into
the
future.
You
know
this
is
this.
N
Is
these
are
the
audience
sizes
for
some
of
the
larger
events
in
the
last
few
years
and
the
actual
delivered
online
events
that
you
know
already
are
well
over.
This
sort
of
high
end
achievable
value
for
for
a
single
provider
for
for
4k.
So
obviously
these
these
users
are
not
receiving
their
traffic
in
4k.
N
How
long
that
will
last
is
an
exercise
for
the
reader,
I
guess
so
yeah,
but
there's
other
problems
also
that
are
kind
of
worth
going
over
and
addressing
you
know
some
of
the
challenges
that
you
encounter
when
you're
doing
a
PR
with
segment
transport,
how
you
know
he
run
into
how
the
network
interoperates
with
with
the
player
that's
trying
to
perform
bitrate
detection
and
the
kinds
of
effects
that
this
causes.
Some
of
some
of
the
may
be
insights
into
network
design.
N
You
know
how
to
address
conferencing
what
the
latency
requirements
are
the
multi-directional
options.
You
know
the
the
like
what
what
is
it
going
to
look
like
in
your
network
when
you
try
to
run
these
conferencing
things
and
when
they
start
to
scale
up
when
there's
more
people,
how
he
can
maybe
approach
how
to
estimate?
How
much
to
expect-
and
likewise
the
pointer
to
the
implementations
like
is
this
something
we
could
maintain
in
the
ITF?
Is
it?
N
Would
it
be
valued
I
think
it
would
be
valuable,
but
you
know
there's
some
questions
to
answer
as
we
do
that
I
think
it
would
be
a
valuable
contribution
for
a
working
group
to
to
address
and
then
just
where
to
go
from.
There
is
I'm
hoping
for
a
feedback.
I'm
hoping
for
I
feel
like
the
working
group
would
be
worthwhile,
but
you
know
that's
just
me
so
questions
comments.
D
H
N
This
is
the
this
is
one
of
the
documents.
I
think
I
hope
would
be
worthwhile.
You
know
what
problems
do
we
encounter?
I
think
it
would
be
I
asked
for
maybe
maybe
I'd
end
up
right,
I,
don't
know,
but
I
would
like
to
have
a
roadmap
of
of
the
various
different
video
solutions.
I
think
that
when
I,
whenever
I
try
to
figure
that
out,
I
just
get
confused,
I
can
spend
a
week
on
it
and
I
still
get
confused.
N
N
N
D
Like
we
just
had
a
little
bit
of
that
in
the
previous
presentation
from
Justin
right
and
they
articulating
on
some
of
the
different
requirements
of
that
kind
of
video
are
whether
it's
you
know,
differences
in
terms
of
the
impact
or
availability
of
options
for
compression
or
whether
it's
you
know,
differences
in
terms
of
how
you
know
how
much
details
matter
so
on
and
so
forth.
So
there
definitely
are
different
classes
of
video.
It's
not
just
like
videos,
one
big
chunk
of
it.
Yeah.
N
I'm
also
wondering
if
there's
any
solutions
that
don't
have
a
home,
yet
that
could
get
some
review
here
and
I
mean
I,
don't
know
like
I,
don't
know
if
HLS
is
in
scope
right,
but
that
is
a
informational
draft.
It's
almost
like
an
external
standards
body
doing
it
that
just
doesn't
exist,
but
is
posting
the
information
right
so
but
but
there's
features
that
would
be
maybe
valuable
to
have
or
not
have
in
HLS
and
to
get
some
review
on
that
right.
N
N
O
Tony
caliber,
Comcast
I,
don't
know
you
know,
I'm,
not
sure,
oh
well,
sidestep
where
this
fits
into
ITIF.
That
is
an
operator
I.
Think
it's
very
useful
to
have
some.
You
know,
well-defined
in
it
and
somewhat
agreed-upon
ideas
of
how
to
get
these
scaling
numbers
because
to
design
networks
that
we
need
to
do
this.
Everybody.
I
P
So
you
disappoint
a
unique,
unique
us
problem.
Many
have
millions
of
subscriber
and
4k
in
a
you
know,
be
a
stream
for
any
live
broadcast.
So
that
may
not
work
well.
We
could
consider
you
know
hub
and
spoke
model
and
I
think
some
of
the
operator.
They
seems
to
be
suggesting
that
you
know
they
won't
have
some
way
to
deploy.
So
look
like
it.
That
means
we
capped
it
somewhere.
That's
one
thing
so
then
another
fundamental
thing:
Steffan
multicast
and
4k
video.
P
P
N
D
D
D
I
don't
want
to
get
into
wordsmithing,
because
the
at
point
of
this
exercise
at
this
point
is
just
to
sort
of
say:
here's
a
straw
proposal
for
the
sort
of
thing
we
think
might
be
doable
as
a
working
group,
and
then
we
can
have
more
substantive
discussions
about
whether
or
not
people
feel
that
there
is
a
ETF
work
to
be
done
here
and
whether
there's
clarity
and
doing
it
and
whether
there
are
enough
people
involved
to
be
able
to
do
the
work.
That
seems
like
a
reasonable
approach
to
people
okay.
D
So
the
draft
charter
is
available
here.
You're
welcome
to
quickly
scribble
that
down
or
get
it
out
of
the
slides
and
the
meeting
materials
I
will
step
through
some
of
the
chunks
a
bit
by
bit.
Here.
First
observation
is
that
this
this
is
not
I
mean
charters
are
not
rocket
science.
We
basically
cobbled
this
together
from
previous
work
done
for
the
mbone
d--
charter
and
for
the
v6
ops.
Organ
group
charters
yeah.
So
this
is
the
overview.
D
I
think
this
is
pretty
much
what
we've
been
talking
about
in
general
in
general
round
numbers
here
today
that
there's
a
lot
of
video
going
on
in
the
internet.
There's
a
lot
of
video
being
worked
on.
You
know
here
as
well
as
in
other
organizations
it
should.
It
would
be
beneficial
if
we
could
get
better
coordination
between
those
activities,
as
well
as
awareness-raising
of
potential
solutions,
of
which
people
might
not
be
aware,
and
to
do
these
things
in
a
way,
that's
actually
sympathetic
to
the
Internet's
overall
design.
D
So
to
that
end,
the
purpose
of
the
working
group
would
be
to
solicit
input
on
operation
issues
and
practices
existing
and
proposed
technologies
related
to
the
deployment,
engineering
and
operation
of
media
streaming
and
manipulation,
protocols
and
procedures
in
the
global
Internet
inter
domain
and
single
domain
networks.
So
that's
kind
of
an
overview
of
where
we
think
this,
this
sort
of
thing
lands
I,
would
take
clarification,
questions
or
suggestions
at
this
point,
I
don't
really
want
to
get
into
hyper
wordsmithing.
At
this
point,.
E
So
one
of
the
common
things
we
like
to
ask
that
the
ITF
is
what
problems
it
solved,
because
we
shouldn't
do
things
if
there's
an
apartment
that
it's
neatly
gonna
contribute
you
one
of
the
takeaways
I
found
in
the
last
few
years.
Working
this
problem
space
within
the
IHF
and
in
other
places
like
the
SBA,
is
that
there
seems
to
be
is
recurring
problem
and
we
saw
it
a
little
bit
in
the
strata
stuff
right.
E
Chris
destroys
a
theater
where
there's
all
these
pieces
and
when
you're
trying
to
put
them
together
to
do
something
useful
and
you
keep
hitting
a
little
roadblocks
and
little
weird
things
in
that
particular
piece
as
especially
on
boundaries
between
one
piece
and
another
piece
are
where
another
layer
and
when
it
comes
to
video
expertise
of
well.
Why
is
that?
And
what
is
it
about
video
that
exercises
that
particular
boundary
problem?
It
needs
work,
it's
hard
to
find
places
to
have
that
conversation,
and
it's
also
hard
to
find
expertise
to
solve
that.
E
And
one
of
the
reasons
we
have
those
boundary
problems
and
they
get
really
exacerbated
by
video,
is
the
pure
volumes
of
traffic
that
are
sliding
over
our
networks
all
the
time.
So
these
little
inefficiencies
that
we're
willing
to
tolerate
in
other
situations
get
exacerbated
day
in
and
day
out
by
video
and,
as
we
start
doing
things
like
streaming
a
full-blown
4k
video
for
video
games
that
will
latency
across
the
network.
We
really
just
pound
the
heck
out
of
our
protocols
and
so
part
of
the
idea
that
I,
hear
and
I
see.
E
This
in
this
draft
overview
is
the
notion
of
it's
a
place
where
people
who
understand
that
problem
and
are
feeling
that's
that
scaling
issues
and
those
frictions
between
the
pieces
you're.
Putting
the
gallery
to
deliver
a
solution.
Can
that
knowledge
and
experience
back
into
the
IDF
find
like-minded
people
read
of
the
evil
drafts?
Call
us
that
those
ideas
and
those
leads
those
wants
and
ultimately
need
to
work
on
it
within
the
group
itself
or
feed
those
needs,
as
they
just
stated
and
become.
E
N
Q
Of
having
the
there's
enough
work
in
here
in
terms
of
the
I
know,
you
hope
you
don't
want
to
do
wordsmithing,
but
this
is
a
quick
question.
So
I
like
the
strictly
White's
is
that
it's
about
you
know
all
things
that
are
related
to
media,
which
include
video,
audio
and
other
objects,
so,
which
was
one
of
the
things
I
was
going
to
talk
about.
Is
that
there's
also
more
work
that
goes
just
beyond
video
in
terms
of
immersive,
media
or
AR?
We
are
kind
of
technologies
which
are
all
related
to
the
streaming
thing.
Q
D
J
D
D
This
group
wouldn't
be
doing
the
new
protocols
if
they
are
new
protocols,
but
it
might
do
some
legwork
say
you
know
there
really
are
no
protocols
that
currently
do
what
we
need,
there's,
not
something
that
we
can
adjust.
So
how
can
we
go
off
and
do
other
work
to
do
that
so
I
think
some
existing
protocols,
some
potential
for
spinning
up
Newark
or
at
least
characterizing
potential
for
new
work.
Okay,.
J
That
makes
sense
to
me
I
think
I
just
wanted
to
reinforce
the
point
that
Roberto
made
earlier,
which
is
that
will
basically
have
you
know,
probably
some
new
protocol
work
happening
and
perhaps
and
some
other
group,
but
then
like
that
will
actually
have
to
be
combined
with.
How
does
that
actually
work
with
your
network
right
sounds
good
cool.
R
Yeah
mess
TOC
limelight
networks,
so
I
mean
I
think
in
general.
This
is
this
is
interesting.
I
mean
one
of
the
values
that
we've
seen
with
stream,
video,
Alliance
and,
and
then
potentially,
this
is
the
you
know,
there's
the
technology
components
and,
as
Glenn
mentioned,
there's
all
these
different
pieces,
but
then
there's
everybody
learning
how
to
put
all
of
those
pieces
together
on
their
own
and
when
you're
dealing
with
scale
and
interoperability
at
scale.
R
You
know
everybody's
choices
when
you've
got
ambiguity,
has
impacts
and
ripples
with
everybody
else
and
and
often
you've
got
that
also
in
situations
where
we're
all
experts
in
a
in
a
slice.
But
you
need
to
be
able
to
have
an
opportunity
to
gauge
feedback
from
different
layers
in
the
text.
Stack
or
even
you
know,
silos
and
so
I
mean
that's
where
I
think
the
value
in
this
kind
of
a
function
is
you
know
really
it's
that
opportunity
to
say
yeah.
This
is
this
is
a
good
way
to
accomplish
it.
D
E
One
of
the
challenges
I
think
that
people
sometimes
don't
get
is
that
video
on
the
Internet
is
is
undergoing
a
constant
evolution
and,
and
they
look
at
something
like
you
know,
be
it.
We
all
be
able
look
stream
movies
now
for
a
few
years
and
do
it
pretty
well
and
that
works
pretty
well,
you
can
say
well,
there's
no
problems
here.
It's
his
job
was
done.
E
Chart
kind
of
showed,
you
know
he
was
like
way
down
the
Nielsen
tail
for
a
show
that
you
know
isn't
a
popular
show
and
it
was
consuming
a
tremendous
amount
of
bandwidth
for
just
doing
that.
The
number
one
shows
and
the
big
events
way
up
on
that
chart.
We
have
a
capacity
hall.
We
have
a
scalability
call
the
world
of
video
that
they're
transmitting
to
is
changing
right.
It's
evolving
from
just
being
playback
of
movies,
which
is
a
very
simple
experience,
is
very
linear.
I
push
the
button,
I
watch
it.
It
goes
back.
E
It
was
the
end
to
we're
doing
now.
Live
sports,
multi-camera
live
sports
and
not
just
professional
stuff.
It's
people
themselves
are
streaming
from
the
kids
soccer
game,
multiple,
multiple
camera
angles,
doctor
grandma,
grandpa,
they're,
doing
very
sophisticated
things
in
very
high
resolution
formats.
Right
people
are
doing,
4k
people
would
8k
on
the
Internet.
The
data
volumes
are
ridiculous
and
so
we're
sort
of
this
junction,
where
I
think
where
we've
gone
from
media
like
1.0,
which
was
sort
of
like
your
really
simple
basic
stuff.
E
And
that's
what
I
framed
I'm
very
excited
about
it,
because
I
think
it's
an
interesting
time
to
be
and
to
see
this
evolution
go
on.
But
it's
also
going
to
create
some
real
challenges
for
IP
networks
and
and
that's
why
I
think
this
work
is
kind
of
relevant
to
the
IGF
because
were
the
place
that
does
IP
networks?
Okay,.
D
S
D
M
Prepare
to
play
on
Google
sorry
now,
Facebook
been
a
long
time
anyway.
I
have
similar
fears.
This
is
very
broad
and
I
can
give
you
a
concrete
case
where
I
would
try
and
figure
out,
or
rather
I
would
have
difficulty
figuring
out
where
this
would
go.
So
let's
say
we
create
a
partially
reliable
HTTP
I
think
it's
obvious
that
you
can
extend,
etc
and
start
in
engine
upon
the
things
that
we're
currently
doing
with
really
low
latency,
WebRTC
etc
with
with
such
technology.
M
But
you
could
also
use
it
for
games
and
you
can
also
use
it
for
regular
object
transmission.
You
can
use
it
at
proxies,
for
you
know
eliminating
head-of-line
blocking,
but
not
necessarily
for
video.
So
I
think
that,
in
order
for
this
particular
thing
to
be
successful,
we
have
to
have
a
really
crystal
clear
idea
of
exactly
what
things
land
here
and
how
they
don't
impinge
upon
activities
in
other
places.
Right
so.
D
In
that
regard,
though,
I
think
again,
the
the
hope
and
expectation
is
that,
if
stuff
lands
here,
it's
largely
from
the
standpoint
of
getting
it
vectored
somewhere
else,
so
something
like
that
did
land
here
then.
Hopefully
there
would
be
expertise
there
and
say
you
really
should
take
it
to
this
working
group.
Yeah.
E
D
D
Okay,
so
yeah
goals,
as
we
have
them
listed
again,
soliciting
input
from
network
operators
and
users
to
identify
operational
issues
with
media
delivery
in
and
across
networks
to
solicit,
discussion
and
documentation,
always
important
of
documentation
of
the
issues
and
opportunities
for
media
acquisition
and
delivery
and
of
the
resulting
innovations.
So
I
mean
all
this
really
is
again
focused
on
the
kinds
of
work
that
we
envision
doing
here,
and
you
know
some
of
which
you've
seen
outlined
today.
D
D
Yeah,
the
only
real
potential
area
for
development
is
point
number
six
to
develop
tools,
extend
protocols
and
provide
operational
and
implementational
advice
that
assists
in
media
technology,
administration,
Diagnostics,
troubleshooting
and
deployment
between
and
within
native
and
non-native
environments.
What
a
mouthful
again
not
wordsmithing
here,
but
the
point
being
not
it's
not
about
developing
new
protocols.
So
much
as
the
best
current
practices,
and
you
know,
analysis.
K
Right
so
everyone's
talking
about
the
SVA
on
here
representing
the
SVA
yeah
I,
think
you
know
yeah
sorry
Jason
t-bo's
during
video
Alliance,
okay.
So
what
you
guys
are
doing
here
is
great
right.
So,
as
another
organization
that's
focused
on
streaming,
video
and
best
practices
and
guidelines
and
specs,
we
really
look
to
the
IETF
and
SEM
T.
For
things
like
this
right,
things
are
technology
technological
evoke,
astride.
K
There
are
lots
of
issues
to
resolve
technology
issues
in
the
workflow
like
encoding
and
things
like
that,
but
the
IETF
needs
or
is
doing
what
it's
focused
on,
which
is
what
I
would
say:
protocol
development
stack,
development,
operational
development
and
things
of
that
nature.
I
think
what
we
need
to
make
sure
is
that
you
know,
from
my
perspective
from
my
organization's
perspective,
that
the
IETF
doesn't
supplant
or
try
to
take
over
kind
of
what
we're
trying
to
do.
K
So
we
need
to
figure
out
how
to
work
together
to
make
sure
that
the
IETF
is
focused
on.
You
know
sort
of
the
technological
development
that
the
SVA
doesn't
really
focus
on
and
at
the
same
time
allows
us
to
remain
sort
of
flexible
to
approach
the
things
that
we
want
to
approach,
while
you
guys
are
doing
you
know
really
kind
of
the
in
the
weeds
development.
That
is,
you
know,
internet
engineering,
which
is
again
not
something
that
the
SVA
is
particularly
germane
to
Thanks.
H
Can
you
say
a
little
bit
more
so
I'm
hooking
on
the
developing
new
stuff
here,
a
little
bit?
Okay,
that
that
that
edge
of
it?
So
a
lot
of
that
is
done
with
extending
you
say
a
little
bit
more
about
the
BCPs
in
particular.
How
you'd
expect
those
to
constrain
other
work
that
happen
in
the
ITF,
because,
even
though
we
use
them
like
what
type
of
things
comes
up
there,
so.
D
I
think
the
the
notion
there
was
best
current
practice
in
sort
of
the
literal
and
not
in
the
literal
interpretation
of
the
words
and
not
sort
of
a
now
we're
doing
standards
track
stuff
so
for
operators
who
have
determined
that
here
is
a
useful
way
to
handle
this
type
of
traffic
in
our
network.
Here
is
our
current
practices,
for
instance,
right.
D
D
Yes,
no
choice,
yeah,
okay!
So
that's
that's
a
fair
point
and
definitely
bookmarked
for
for
clarifying
the
text
I.
It
was
not
at
least
in
my
mind,
intended
as
a
constraint
on
development
of
other
work.
So
much
as
a
capture
I
do
think
they
probably
are
best
and
they
probably
are
current
and
they
probably
are
practices
but
not
BC
piece.
I.
H
Take
any
one
of
those
three,
so
it's
still
be
worth
publishing,
so
okay
and
then
on
six
I
mean
when
we
start
talking
about
I
mean
most
of
our
work
is
actually
done
like
sending
protocols,
not
developing
new
protocols.
I
mean
III
think
that
that's
some
of
the
people's
uncomforta
most
comes
around
that
that
type
of
space.
H
R
D
E
To
that
point,
so
we
have,
we
have
JC
a
juice
himself
from
the
SVA,
something
who
was
trying
to
be
here
but
had
to
go
to
a
meeting
in
LA
on
some
video
stuff
was
the
director
Thomas
files
and
Mason
director
of
standards.
That
Cindy
was
also
going
to
be
here
today.
Unfortunately,
he
happy
go
that
way
instead
of
Montreal.
E
T
I
think
this
is
it's
a
helpful
list.
It
kind
of
mixes
a
few
different
things
and
I
think
when
the
community
goes
to
evaluate
the
Charter,
if
they
do,
it
might
be
useful
to
separate
them
out.
So
one
thing
is
like
goals
and
exactly
the
dictionary
definition
sense,
but
usually
a
Charter.
Would
it
be
a
little
bit
more
specific
on
like
actual
work
items
and
this
list
kind
of
intersperses?
T
These
right,
like
which
documents
are
you
gonna,
work
on
or
what
kinds
of
documents,
as
opposed
to
like
what
kind
of
discussion
you're
trying
to
to
stimulate
and
I?
Think
separating
those
into
like
a
specific
like
here's,
the
work
list
versus
like
here's,
the
here's,
the
loftier
goals,
might
help
ease
the
path
for
this
one
and
similarly
I.
Think
because
the
you
know,
the
mechanism
that
you
have
in
the
wharton
group
is
the
consensus
mechanism.
T
It
would
be
useful
to
understand,
like
which
of
these
things,
do
you
expect
to
be
working
group
consensus,
kinds
of
outputs
versus
like
number
five,
which
I
assume
is
more
like
something
that
people
who
have
been
together
here
will
do,
but
it's
not
something
where
you
need,
like
the
machinery
of
the
working
group
to
get
it
done,
so
I
think
making
that
distinction
can
be
helpful
and
then
I
would
just
echo
Cohen's
comment
on
on
point
number.
Six
like
I
want
to
know
which
protocol
is
you're.
T
Extending
because,
like
that's
kind
of
the
point
of
the
Charter,
is
to
scope
that
so,
if
there
are
specific
ones
in
mind
like
that's
great
and
we'll
make
sure,
there's
no
overlap
with
other
existing
groups,
but
if
we
don't
know
which
ones
those
are,
then
it's
probably
it's
not
really
a
scope
limitation.
So
it
gets
a
little
bit
trickier.
I
think
from
a
chartering
perspective,
sure.
U
D
L
Nathan
looks
okay,
the
number
one
I
was
wondering
the
users,
you
mean
content
provider,
so
you
mean
end
user.
What
was
the
meaning
of
users
here
and
because
I
think
both
of
them
are
applicable
and
I
think
that
for
number
six
I
would
expect
to
see
I
mean
the
Charter
we'll
have
to
see
which
other
for
forum
for
us
are
relevant
for
this
work
that
we
need
to
cooperate
with
stuff
that
it's
not
the
only
one
I
mean
there
are
others
that
are
doing
working
in
this
area.
E
One
of
the
growing
things,
especially
from
live
screaming
by
individuals
or
even
professionals-
well,
whatever
it
is,
we
were
doing
it.
It's
it's
something's,
really
growing
as
a
demand
on
the
network,
and
so
that's
why
acquisition
was
included
when
we
wrote
this
so
that
it
isn't
just
about
pumping
bits
from
a
server
to
a
viewer.
It's
also
about
the
capturing
of
the
bits
from
the
camera
into
the
new
system
and
science
people
or
in
the
case
of
like
the
gaming
space.
Its
actual.
S
E
Especially
the
live
stream,
which
is
incredibly
growing
on
the
internet
right.
We
is
that
the
two
things
go
hand
in
hand,
and
so
I
can
see
the
argument
for
you
separate
them
and
divide
and
conquer
the
two
of
them
have
to
work
really
well
together
and
so
putting
them
in
one
place.
So
they
can
work
well
together
and
and,
as
you
develop
the
documents
around
them,
it
makes
it
easier
to
work
and
coordinate
that
effort
instead
of
splitting
them
up
and
making
them
sort
of
fend
for
themselves
and
not
have
any
coordination.
E
V
On
kumari
google,
so
there
was
a
question
from
the
jabber
about
sort
of
how
many
people
here
are
not
regular
IETF
attendees
like
how
many
people
came
in
for
this
and
I,
don't
know
if
we
ever
actually
heard
an
answer
to
that,
could
we
do
something
a
show
of
hands
or
a
like
how
many
people
normally
participate?
How
many
don't
how
many,
if
this
were
to
actually
continue,
would
stop
participating,
etc?
V
D
You
want
to
do
that
right
now,
or
do
you
want
to
do
that
when
we
do
general
discussions
of
all
right,
so
so
I
think
the
list
of
questions
were
first
of
all,
how
many
people
here
are
not
usual
IETF,
attendees,
okay,
sun-ok!
Thank
you
and
then
of
those.
How
many
came
specifically
for
this
or
material
like
this?
How
many
people
came
because,
in
part
at
least
because
of
this
Boff
right,
so
smaller
smaller
show
of
hands,
then
the
on
the
on
word.
D
Questions
are
if
this,
if
something
like
this,
if
work
in
this
area
were
was
chartered,
how
many
would
expect
how
many
didn't
just
come
to
throw
bananas?
How
many
would
expect
to
continue
to
attend
and
participate
in
the
discussions,
and
not
just
so.
The
people
who
came
are
new
here:
okay,
so
I'm,
seeing
about
quarter
to
a
third
of
the
room,
are
putting
up
their
hands
dead
good
enough
for
now,.
L
D
L
D
M
Media
acquisition
is
or
is
not
the
same
thing
as
ingestion
or
contribution
acquisition
has
connotations
of
things.
I
can
pay
money
and
get
a
license
for
some
content
as
opposed
to
distributing
this
might
be
wordsmithing,
but
it
might
not
be
worth
smithing
depending
on
your
intent.
So
I
just
want
to
bring
this
up
that
a
definition
of
that
is
really
important
for
what
we're
agreeing
or
disagreeing
on
that's.
E
D
So
thanks
for
that
discussion
and
moving
on
to
the
next
chunk
of
the
the
draft
charter
and
for
all
of
those
points,
sort
of
is
a
global,
some
global
scoping
rules.
We
have
the
documents,
should
document
media
operational
experience,
including
the
global
internet
inter
domain
and
within
domain
operations,
and
that
media
operational
and
deployment
issues
with
specific
protocols
or
technologies
are
primarily
are
the
primary
responsibilities
of
the
groups
or
areas
responsible
for
those
protocols.
Ie
we're
not
trying
to
eat
anybody's
lunch,
so
it
might
be
useful
for
us
to
provide
input
to
those
other
groups.
D
W
Chris
lemons
from
Comcast
there
are
ship
it
off
to
another.
Working
group
is
particularly
interesting
to
me.
We've
got
whenever
you
put
together
a
bunch
of
a
bunch
of
protocols,
components.
What
have
you
a
lot
of
times?
You
discover
in
fact,
that
Sanjay
mentioned
it
in
his
presentation
from
the
SVA.
You
discover
places
where
you
need
extension
points
to
hook
them
together
and
so
making
sure
that
we
send
off
that.
Those
extension
point
works
to
the
appropriate
working
groups
is
I,
think
going
to
be
vital
to
the
functioning
we.
D
Thanks
and
then
the
additional
constraints,
just
that
this
is
stuff
the
ad
made
us
put
in
no,
that
yeah
there
is
some
hope
to
have
scope
and
and
tightness
of
scope
in
the
working
group
and
future.
You
know
extending
that
scope
is
a,
it
would
require
for
a
discussion,
and
if
we
don't
actually
continue
to
meet
and
discuss
this
stuff,
then
we
clearly
don't
have
any
work
to
do
and
we
should
just
stop
it.
F
F
We
actually
have
license
that
so
so,
overall,
the
the
proposed
text
for
each
of
the
items
is,
you
know
pretty
vague
and
covers
I
think
it
was
pretty
well,
so
it
could
cover
everything.
So
you
know
I
think,
would
be
good
to
have
a
little
more
specificity
because
it
just
seems
like
a
cover
between
the
Charter
and
the
goals
seems
like
it
covers
everything
and
as
I
read
it,
it's
still
not
clear
to
me
what
it
is.
F
If
I
didn't
know
that
I
was
in
here
and
saw
everything
else,
I
I
wouldn't
know
what
this
group
did.
But
then
you
know
it's
not
obvious
to
an
outsider.
What
exactly
and
specifically,
as
the
problems
we
trying
to
solve,
but
overall
I
would
express
a
lot
of
you
know.
I
think
this
is
a
great
effort,
a
very
useful
effort.
As
Jake
has
said,
there
seems
to
be
different
kind
of
a
Tower
of
Babel
problem
where
the
network
guys
don't
quite
understand
the
video
stuff.
F
The
video
guys
don't
quite
understand
the
network
stuff,
but
we
are
thinking
that
we
are
coming
up
with
requirements
or
solutions
for
each
other.
That
may
or
may
not
be
useful,
so
a
coordinating
role
by
this
working
group
would,
by
by
a
working
group
like
this,
would
be
very
valuable,
and
you
know
for
just,
for
example,
in
in
Bondi.
We
have
been
it's
it's,
not
your!
F
E
You
thanks
thanks.
The
point
you
raised
about
it
being
very
broad
is:
is
this
we
have
this
icon
of
me.
It's
the
the
challenges
that
it
touches.
Video
touches
everything
in
the
network.
It
touches
routing
transport
package
every
little
piece,
every
almost
every
workgroup
I
mean
recently,
people
were
coming
to
be
asking
for.
You
know,
help
with
time,
stuff
and
I
thought
well.
I
thought
time
had
been
sort
of
settled
years
ago
and
bam
it's
back
in
droves
and
so
far
the
channel
trash
and
we
try
to
face
and
writing
this
stuff.
E
Is
that
it's
how
to
capture
the
fact
that
it
really
does
package
all
the
pieces
that
we
do
with
the
IETF
together
to
do
the
work
it
does
so
the
answer
is
well:
what's
in
it,
everything
isn't
it
what's?
You
know,
but
where's
the
scoby
where's
the
limit
there.
Well,
it's
around
video,
it's
everything's
in
it,
but
it's
it's
specifically
limited
to
what
you
use
those
things
to
do
for
video.
So.
D
D
We
don't
have
one
specific
protocol
that
we're
talking
about
and
that's
a
challenge,
I
mean
initially
I
was
resistant
to
the
idea
of
a
proposal
of
having
an
ops
working
group
for
this
stuff,
because
I
think
it
really
is
and
remains
more
of
an
interest
group
type
thing
across
area
thing.
You
know
whatever
it's
not
really
a
working
group
per
se,
not
really
a
research
group
either,
but
but
the
IETF
just
doesn't
really
have
the
right
structure
for
that
kind
of
a
thing.
D
So
we're
trying
to
make
this
we're
trying
to
you
know
get
this
shaped
properly,
so
that
we
can
get
some
of
the
work
done
in
the
context
of
an
ops
working
group.
So
I
think
the
point
is
really
valid,
but
I'm
just
articulating
this
is
this
is
what
we've
been
struggling
with,
and
this
is
why
it
sort
of
looks
a
little
bit
like
a
dog's
breakfast,
because
it's
you
know,
we
can't
say
it's
specific
protocol
and
operational
issues
related
to
it.
It's
this
stuff
and
operational
relay
issues
related
to
it
so
help
appreciate
it
sure.
F
Well,
I
think
what
you
just
said:
it
compasses
everything
I
think
that's
that's
as
clear
of
you
know,
explanation
for
you
know
what
this
group
could
be
doing
and,
and
maybe
more
you
know
text
around
that
that
makes
it
clear
what
you're
trying
to
do
and
why
you
exist.
Looking
at
the
text
as
it
is
now
I.
If
I
wasn't
in
this
room
for
90
minutes,
I
wouldn't
know
yeah.
N
Jake
Allen,
just
a
clarifying
question
to
the
people
who
have
expressed
a
concern
over
the
breadth
of
scope
is:
would
it
satisfy
their
concerns
if
it
just
had
sort
of
a
it
must
be
tied
to
media
delivery?
I
mean
like
that
is
the.
That
is
the
key
point
that
we're
trying
to
address,
but,
like
Lesley
said
it,
you
know
video,
video
and
media.
If
you
think
of
point
cloud
or
something
right.
N
N
N
D
M
Roberto
one
of
my
problems
here
is
that
the
things
that
I
do
to
solve
video
are
exactly
the
things
I
do
to
solve
other
domains.
So
if
I
want
to
solve
gaming,
for
instance
and
I'm
not
talking
about
the
display
part
coming
back
right,
but
just
game
state,
it
has
exactly
the
same
properties
as
video.
It
is
potentially
stateful.
I
need
compression
I
need
to
distribute
it.
M
D
X
D
Well,
it's
not
just
you
can't
boil
the
ocean,
it's
it's
that
we
haven't,
found
the
right
cross-cutting
way
to
articulate
what
it
is.
It's
in
an
inner
out
and
I.
Don't
think
I'm
gonna
get
its
ending
here
at
the
mic
microphone
if
anybody
sees
a
lightning
bolt
come
down
from
the
sky
hallelujah,
but
in
the
meantime,
I
do
think
that
it's
worth
at
least
getting
some
sense
of
the
room.
These
are
the
questions
from
RC
54-34
that
are
guidance
for
whether
or
not
we
are
ready
to
propose
a
working
group.
D
I
certainly
understand
from
the
discussion
we've
had
so
far.
We
don't
have
a
charter
that
people
are
willing
to
hum
on,
but
I
think
that
there
are
two
separable
general
questions.
One
which
is
you
know,
do
people
think
there's
something
here
that
the
IDF
should
be
working
on
and
the
second
one
is,
you
know:
should
we
be
forming?
D
A
working
group
will
probably
say
primarily
focused
on
the
first
part
today,
because
there's
more
more
than
then
we're
tweaking
to
do
for
the
second
part,
so
just
in
general,
around
numbers
and
I
guess
we
can
start
with
a
hum
and
then
and
then
have
some
microphone
time.
Do
people
here
feel
that
there
that
there
is
a
problem
that
needs
solving
in
the
ITF?
Is
the
right
group
to
attempt
solving
it
so
go
for
a
hum
for
yes
and
go
for
hum
for
No
all
right.
That's
good
news.
H
Oh,
that
question
is
very
interesting.
I
know
I
can
prepare.
My
problem
is,
is
I,
don't
think
if
we
ask
these
people
what
the
problem
is
that
needs
to
be
solved,
they
can
articulate
it.
That's
that's
the
essence
of
every
presentation
today.
So
I
agree.
There's
a
problem
of
easily
solved
right,
yeah
I,
don't
know:
we've
articulated
it.
Yeah
yeah.
E
I
think
that's
a
small
to-do
list
sort
of
an
indicator
that
there's
a
problem
that
that
there's
a
lot
there's
a
sort
of
a
common
theme
here.
It's
it's
video,
video,
video,
I'm,
touching
all
these
things
and
I've
got
bandwidth
problems,
I've,
got
protocol,
frustrations
and
and
at
some
point
the
trying
to
bring
that
that
energy
together
to
do
productive
work
just.
H
E
H
Totally
say
let
the
problem
to
me
is
that
we
have
many
complicated
things.
There's
no
expert
single
one.
We
need
a
common
place.
That
brings
all
those
people
to
talk
about
their
different
technologies
together
and
they
don't
need
to
produce
anything
usually
talked
about
right.
That's
the
problem
to
be
sure,
that's
not
the
problem
to
everyone
else
in
the
room,
but
clearly
there's
a
problem.
I
mean
obviously
movie.
You
know
well.
D
D
The
next
standard
question
from
54-34
is:
do
we
believe
that
there
is
a
critical
mass
of
participants
willing
to
work
on
the
problem
to
write
drafts,
blah
blah
blah?
That's
a
little
harder
to
answer,
given
that
we
may
all
be
thinking
about
different
problems
at
this
moment,
but
I
think
a
legitimate
question
to
ask
at
this
point
is:
are
there
people
who
envision
that
they
would
be
willing
to
not
just
attend
the
meetings
but
read
the
mailing
list
and
help
write
drafts
as
they
came
along
in
Jake's?
D
Already
ahead
of
ahead
of
us,
he's
got
a
proposed
internet
draft
and
some
text,
so
the
question
to
back
it
up
is
and
I
think
this
will
be
a
show
of
hands
one
rather
than
a
hum.
Who
here
would
be
willing
to
participate
in
the
mailing
list
and
possibly
write
some
drafts
in
this
area,
all
right,
so
I'm
seeing
a
good
dozen
hands
anyway.
D
D
All
right
closer
to
20
hands,
so
I
think
that
there's
there's
there's
interest.
There's
willingness
and
I
think
the
next
thing
to
do,
because
I
don't
think
that
I
don't
think
anybody
here
is
gonna
agree
that
the
scope
of
the
problem
is
well
defined
and
understood.
Let
alone,
though
there
is
agreement
which
we've
articulated,
none
are
the
right
ones,
so
I
think
the
go
ahead.
Well,.
F
Q
D
D
I
E
D
S
This
is
really
simple:
I
think
I
I,
don't
want
to
wordsmith
and
give
give
names,
but
the
problem
is
not
recent
23.1
four
point
three
kilobits
per
second
speech:
codec
stream
and
the
problem
is
also
not
100
kilobits
per
second
AAC
audio
stream.
The
problem
are
these
multi
Multi
multi
make
up
eight
streets
yeah
and
those
are
so.
D
D
S
D
S
E
The
point
is
I
think
what
you're
trying
to
say
is
if
I
can
play
it
back.
Is
that
the
little
tiny
bits?
That's
all?
This
is
about
the
big
chunky
things
that
are
really
hard
to
manage,
that
are
stressing
the
infrastructure
and
stressing
the
protocols
and
to
get
back
to
the
video
immediate
question.
So
media
sometimes
is
just
lovely
catch-all
sort
of
refer
to
both
the
video
and
the
audio
and
all
the
little
bits
of
pieces
I
get
packaged
round.
Even
if
it's
data
objects,
it's
the
about
media
videos,
a
piece
of
the
media.
E
H
So
college
days,
I
just
say
not
I
was
arguing
scope
it
down
here
but
decide
to
boil
the
ocean.
So
all
the
things
though
Justin's
talking
about
may
be,
you
know
20
megabits
streams
and
right,
but
you
have
all
exactly
the
same
problems:
the
same
issues
when
you're
running
256,
kilobits
streams
of
video
in
a
stadium
over
a
wireless
network
and
40,000
people
all
simultaneously
stream
the
same
game.
H
N
D
D
Y
Higher
all
this
time,
Google
without
consulting
with
Justin
and
I,
think
we
shouldn't
think
about
defining
a
term.
Let's
call
it
foo,
we
might
actually
use
video
or
that
no
video
data
well,
which
is
that
we
want
to
transfer
stuff
that
contains
at
least
a
video
component,
may
contain
other
components
and
has
some
kind
of
timing
constraints
on
its
delivery.
Y
E
G
Mission
Verizon,
so
I
think
we've
already
hit
a
lot
of
points
about.
You
know,
sharpening
the
focus
so
I
I,
don't
need
to
say
anything
more
on
that,
but
really
it's
not
the
use
cases
that
keep
coming
up
and
something
that
I
know
within
the
SVA.
We
are
looking
at
virtual
reality
and
to
end
delivery,
and
there
are
issues
that
come
up,
whether
in
some
other
things
that
need
to
go
into
the
SRTP
right
about
the
measurements,
so
that
you
have
an
interim
sense
of
how
you're
delivering
me.
G
So
those
are
the
issues
that
would
come
up
so
we're
not
there
yet,
but
I
know
that
we're
going
to
be
heating
against
some
of
these
issues
because
but
that's
a
protocol,
you
know
SRTP
needs
to
be
looked
at,
whether
it's
really
how
do
I.
Where
do
I
go
with
my
measurements?
There
are
many
places
to
go
on
that
and
what
we're
measuring.
So
some
of
those
issues
are
the
ones
that
we
would
probably
be
bringing
up
here
and
latency
for
VRS.
Really
right
it
and
and
compression
is
a
whole
other
thing.
W
When
we're
framing
this
problem,
I'm
hearing
a
lot
of
people
articulating
ideas
about
what
we
think
it
should
be,
what
stuff
is
in
scope,
we're
all
trying
to
you
know
annex
some
scope.
It
might
be
most
useful
if
we
start
to
the
you
know
when
we
take
you
to
the
list
or
continue
a
conversation,
we
think
clearly
about
some
specific
things
that
we're
gonna
declare
out
of
scope.
W
For
example,
somebody
asked
me
when
we
came
here
so
so
is
this
about
telephony
right
is
a
telephone
call
in
scope
and
I
think
there
might
be
some
people
in
the
room
that
are
thinking
well,
yeah,
obviously,
clearly
it's
media
and
it's
being
transmitted
and
has
latency
requirements
and
other
people
are
saying.
No.
These
are
small
things
that
the
solved
probably
don't
need
to
include
that
and
so
clearly
defining.
What's
not
in
scope
taking
you
know,
scope
and
saying
that's
not
there.
E
M
To
pay
on
Facebook
I
just
wanted
to
throw
out
a
couple
things
that
I
haven't
heard,
but
I
think
are
implied,
which
is
that
what
we've
been
talking
about
our
stateful
things
that
they
are
consumed
in
real
time,
whether
or
not
that
real
time
is.
You
know
10
seconds
for
an
hour
whatever
it
doesn't
really
matter
and
their
session
based
when
I
watch
a
video,
it
doesn't
matter
if
it's
many
small
carts
or
anything
like
that,
it
tends
to
go
together.
M
D
Thank
you
for
that.
So
I
think
we
are
pretty
much
at
time.
So
I
have
one
action
item
to
ask
of
everybody
and
then
I
will
look
at
the
ADEs.
They
think
I'm
thinking
that
it
is
time
for
us
to
move
off
of
the
GG
list
and
make
a
new
mailing
list.
So
would
people
here
be
willing
to
supposing
the
eighties
support
creating
a
new
mailing?
This
will
people
here
be
willing
to
sign
up
to
a
new
mailing
list
to
pursue
this
discussion?
D
D
No,
you
look
good
alright.
So,
generally
speaking,
it
sounds
like
there's
there's
a
here
year,
we're
just
not
quite
sure
what
it
is
yet
we'll
work
on
that
on
the
mailing
list
and
hopefully
make
some
good
progress
that
way.
So
thank
you
very
much
to
our
jabber
scribe
and
our
note-taker
and
all
of
our
presenters,
and
thank
you
all
for
coming
and
having
a
good
discussion.
Thank
you.