►
From YouTube: IETF112-MOPS-20211111-1600
Description
MOPS meeting session at IETF112
2021/11/11 1600
https://datatracker.ietf.org/meeting/112/proceedings/
B
A
B
All
right
so
should
we
shall
we
get
started.
I
guess.
A
Yeah,
have
we
succeeded
in
finding
a
volunteer
for
the
taking
of
notes.
B
C
You've
found
it,
I
am
speaking
I
I
am
this
one
is
less
involved,
I
hope,
but
yeah.
Actually,
if
you
could
cover
notes,
while
if
there's
any
questions
that
come
my
way,
that
would
be
helpful.
Thank
you
and
mike
says
so
too.
A
Yeah
great.
A
Yes-
and
everyone
is
free
to
please
go
in
and
work
with,
the
oops
as
we
go
to
help
fill
in
the
blanks
as
useful
and
necessary
and
take
a
little
load
off
of
jake.
But
thank
you
jake
all
right.
A
So
I
guess
we
should
note
the
node
well
well,
remember
reminding
everyone
not
only
about
the
patents
and
the
the
requirement
to
declare
any
knowledge
of
ipr
and
encumbered
technologies,
which
you'll
I'd
like
to
end
up
covering
a
whole
lot
in
this
discussion.
A
But
that
is
still
an
iot
requirement
and
any
discussions
here
can
and
will
be
used
as
part
of
the
ietf
proceedings
and
in
particular,
I'd
like
to
draw
everyone's
attention
to
the
fact
that
we
do
have
a
code
of
conduct
and
the
url
is
on
the
slide
for
the
code
of
conduct
and
how
to
contact
the
ombuds
team.
A
If
you
leave
that
you've
had,
if
you
have,
if
you
feel
you've,
had
an
issue
with
improper
conduct
at
the
ietf,
I
think
that's
a
hopefully
that's
a
reasonable,
covering
of
the
note
well
kyle
anything
else.
You'd
like
to
surface
in
the.
B
A
Yep,
so
we
have
a
pretty
straightforward
agenda
this
morning,
we've
already
managed
to
get
through
all
of
the
admin
hooray.
We
do
have
some
extra
industry
updates
from
glendeen
and
sanjay
this
morning,
a
couple
of
updates
on
the
working
group
docs
and
then
some
interesting
other
work
going
on
within
the
ietf
and
then
it
seems
a
valuable
time
to
take
a
take.
A
look
back
over
the
last
two
years
of
the
mops
working
group.
A
My
goodness
has
it
been
two
years
already
and
talk
a
little
bit
more
about
where
we
have
been
and
where
we
might
go
from
here.
So
unless
anyone
would
like
to
bash
that
agenda,
I
think
we're
ready
to
go.
E
E
Bash
free
wow,
that's
awesome,
are
the
presentations
loaded
up
into
digital
or
into
meet
echo.
E
Oh,
I
love
it
just
in
time,
just
entire
presentations,
perfect.
The
question
is:
how
do
I
do
that
again.
B
Well,
no
I'll,
so
I
have
to
stop
sharing.
I
think.
E
B
But
yeah
so
europe,
so
it's
at
the
top.
It's
the
yeah
okay.
E
Here
we
go
and
then
let's
re-update
share,
because
sharing
is
caring
all
right.
It's
been
a
long
itf
week.
I
I
apologize
for
having
gone
there,
hey
I'm
glendeen
and
I'm
here
to
talk
about
a
presentation
that
sanjay
misha,
who
also
along
on
the
session
and
jason
thibodeau
from
the
stream
video
alliance,
have
put
together.
We've
done
a
series
of
these
and
sanjay
and
I
usually
hand
off
back
and
forth
on
who's
doing
it,
and
this
itif
was
my
turn.
E
E
And
I'll
feel
free.
If
you
want
to
jump
into
the
cube
and
take
questions,
you
know
what
we'll
clarify
questions
on
particular
slides
along
the
way,
if
not
I'll,
you
know
we
can
talk
at
the
end
of
this,
so
the
sva
has
been
pumping
a
lot
of
work
recently.
You
know
it's
been
a
very
productive
2021.
E
Despite
you
know
what
we've
all
been
facing
and
the
way
the
sva
is
organized
is
very
much
like
the
ietf
we're
divided
up
into
working
groups
that
focus
on
particular
areas.
We
have
focuses
on
you
know:
security,
privacy,
augmented
reality,
vr
live
streaming
and
caching
things
like
that,
and
so,
when
I
go
through
this
deck,
a
lot
of
this
can
be
sort
of
topic
focused
and
then
on
to
the
next
topic
and
it's
sort
of
organized
by
a
working
group.
E
So
the
first
thing
I
want
to
highlight
is
we
have
released
a
a
test
bed
now
for
the
open
caching
api,
so
those
who
are
familiar
with
open
caching
from
sva
there's
a
linkage
across
pollination
here
going
on
with
the
open
caching
at
itf,
there's
our
open
caching
sba
and
the
cdni
working
group
over
at
itf.
In
fact,
sanjay
you
know
is
involved
in
both,
especially
at
the
itf
as
one
of
the
co-chairs,
and
so
you
know,
we've
been
developing
these
these
apis
at
the
sva
around
open
caching.
E
For
several
years
now,
and
we've
now
put
together
a
test
bed
to
allow
people
who
are
trying
to
adopt
those
apis
to
actually
exercise
against
them
and
check
to
see
if
you
know
where
they
need
compliance.
E
So
that's
a
bit
up
and
rolling.
We
also
have
some
configuration
apis.
They
were
so
awesome.
We
had
to
split
them
into
three
documents,
so
there
they
are.
This
has
actually
been
a
considerable
amount
of
work
going
through
that
group
to
create
these
configuration
apis,
and
it's
really
meant
to
be
sort
of
a
companion
to
the
rest
of
this
work.
We're
doing
around
caching
and
availability.
E
Next
up
we
had
a
live
streaming.
Plc
and
a
couple
of
these
slides
are
going
to
talk
about
plc's
that
were
going
on.
You
know
one
of
the
things
the
sva
does
above
and
beyond
working
on
sort
of
apis
and
specs
and
best
practices
is,
you
know,
we're
sort
of.
I
guess
you
could
say
where
the
industry
meets
the
specifications,
because
in
the
room
we
have
a
lot
of
people
who
do
technical
specifications
and
architectures
and
and
product.
E
We
also
have
a
lot
of
people
in
the
room
who
are
actually
deployed
centric
and
content
delivery
centric.
So
in
the
end
you
know
we
have
this
great
opportunity
to
bring
together
the
people
doing
the
specifications
with
the
people
doing
the
deployments
and
allowing
us
to
sort
of
you
know.
Let
them
do
some
proof
of
concepts
and
say
well.
Does
this
architecture
we've
all
put
together
across
the
itf
across
cinti
across
other
wars
like
scudy
and
at
the
sva?
Does
all
that
come
together
in
a
meaningful
way?
E
And
what
can
we
learn
from
that
to
feed
back
into
all
the
various
organizations
that
are
contributing?
And
so
we
put
these
pocs
together?
We've
got
a
couple
that
will
come
up
in
the
decks
here,
but
typically
the
live
stream
plc
has
been
active.
They've
got
some
links
here
that
that
you
can
go
off
and
follow,
and
particularly
they've
been
working
on.
Measuring
live
our
latency
in
abr
streaming,
and
so
that's
a
very
much
a
great
example
of
you
know.
E
Putting
together
this
the
specifications
we're
all
been
working
on.
You
know,
abr
has
you
know
been
around
for
a
few
years,
but
you
know
latency
reducing
that,
especially
for
live
streaming
for
things
like
sports
are
huge
right
now
in
the
delivery
industry,
and
so
this
plc
brings
together
that
exercise
and
says
well.
Are
we
at
the
right
we're
the
right
way?
Are
we
going
the
right
direction?
Is
there
things
missing
that
we
need
to
take
back
into
our
development
work
into
our
specification
work
and
expand
on.
E
Another
plc
we
have
is
all
about
latency
in
streaming
vr,
so
you're
gonna
see
a
theme
here.
A
lot
of
these
things.
It's
latency,
latency
latency
and
a
big
part
of
that
is,
if
you,
you
know,
look
at
the
the
the
content
industry.
There's
been
this
shift
in
the
last
few
years.
You
know
when
we
started
mops
even
two
years
ago.
The
big
focus
was
very
much
on
delivering
of
pre-word
recorded
content,
or
you
know
stuff
that
was
from
files.
Often
in
the
industry.
E
We
call
the
spot
video
on
demand,
and
you
know
that's
what
we've
been
doing
when
we
watch
a
movie,
we're
watching
video
on
demand
and
we've
got
it
pretty
well
done,
but
one
of
the
things
that
you
can
be
very
tolerant
of
in
a
vaude
setting
is
latency.
You
don't
care.
If
you
know
the
movie
is
starting
up
one.
You
know
half
a
second
later
or
a
second
later
or
two
seconds
later,
you
you're
very
taller,
because
it's
a
recorded
film
another.
E
If
you're
watching
live
sports,
you
want
to
make
sure
that
the
the
events
you're
seeing
on
the
screen
are
as
close
to
live
as
possible,
and
this
is
sort
of
the
new
big
challenge
in
front
of
the
streaming
world.
It's.
How
do
we
reduce
that
latency
so
that
experience
is
as
close
to
live
as
possible,
and
so
that
applies
to
live?
Streamings
applies
to
vr.
Everything.
We're
doing
is
trying
to
like
bridge
that
that
gap
between
you
know
what
you
end
up
with
on
your
screen.
E
E
We
also
have
a
distributed
request
tracing.
This
is
a
another
way
for
us
to
tell
how
well
the
the
system
is
working
end
to
end
as
part
of
the
qoe
experience,
and,
of
course
you
know
you
can
make
things
very
fast
and
very
low
latency,
but
you
can
do
it
by
shaving
off
quality
and
that's
not
good,
especially
when
we're
talking
about
things
like
live
sports
equality
and
sports
is
something
that
viewers
demand
and
so
part
of
the
qe
experience
work
in
this
distributed.
Tracing
framework
is
actually
start.
Looking
at.
E
How
do
we
collaborate,
end-to-end
and
collect
all
that
data,
so
we
can
understand
when
it's
working
when
it
isn't
working
and
why
it
isn't
working,
and
so,
if
you're
interested
in
that
space-
and
I
think,
just
real
overlap
with
some
of
the
itf
stuff.
Please
go
check
out
this
link
and
you
can
see
what
that
project
is
doing
and
then
something
that's
probably
pretty
relevant
to
the
itf
quick
for
video
streaming.
You
know
now
that
we
have
quick
publishers
with
specification
yay.
E
You
know,
there's
been
a
lot
of
conversation
at
the
itf
around
quick
for
streaming.
You
know
we
have
mock,
which
is
you
know,
meeting
and
discussing
you
know
the
aspects
there
well,
the
sva
as
well
has
been
taking
a
look
at
quick,
and
you
know
looking
back
at
some
of
the
notes
when
we
did
an
update
from
the
svh
the
itf
a
couple
its
ago.
E
One
of
the
comments
we
made
was,
you
know
we're
looking
at
quick,
but
we
didn't
have
any
clear
plans
to
pick
up
and
start
deploying
it
at
scale
as
a
delivery
infrastructure.
You
know
as
an
industry
and
one
of
the
things
we
identified
in
asking
the
question:
why
is
that
was
that
you
know
there's
a
little
big
shift
going
on
here,
right,
obviously
shifting
to
quick,
udp
infrastructure
away
from
the
tcp
based
infrastructure
that
the
delivery,
guys
all
know
and
have
instrumented
really
well,
is
a
big
challenge
and
it's
retraining.
E
It's
developing
some
new
tools
and
new
ways
of
looking
at
the
problem
space,
and
it's
also
understanding
how
this
all
comes
together
with
expectations
of
the
all
the
expectations
of
the
new
and
what
tools
and
infrastructure
we
have
to
support
it.
E
And
so
one
of
these,
the
the
sva
has
decided
to
do
is
to
create
a
poc
that
is
going
to
build
both
a
reference
architecture
for
testing
the
environment,
but
try
to
tease
out
the
details
of
how
do
you
do
that
instrumentation
and
how
do
you
do
the
metrics
collection
from
end
to
end
so
that
you
can
actually
evaluate
quick
compared
to
your
current
delivery
environments,
evaluate
quick,
you
know
what
it
looks
like
compared
to
that,
but
also
how
you
can
then
start
exercising
some
of
the
more
advanced
features
of
quick
that
aren't
available
in
the
old
world,
and
so,
ultimately,
the
goal
here
is
to
build
a
reference
evaluation
platform
that
the
the
broader
industry
can
take
and
make
use
of.
E
So
they
don't
themselves
have
to
sort
through
things.
So,
in
practical
terms,
what
have
we
done?
So
one
of
our
first
things,
we've
been
working
on
is
looking
through
all
the
the
myriad
of
quick
server
implementations
that
are
out
there
and
trying
to
you
know,
create
a
set
of
criteria.
Well,
why
would
you
pick
one
over
the
other
and
what
features
are
you
looking
for
in
a
quick
server
implementation,
and
how
would
you
then
do
instrumentation
that
connects
in
with
the
tool
sets
you
already
have
as
part
of
your
delivery
environment?
E
Likewise
we're
looking
at
the
player's
side,
how
you
connect
with
a
player
that
supports
quick
and
how
you
instrument
properly
and
then
how
do
all
the
pieces
in
the
middle
come
together
in
a
coherent
way
and
the
idea
behind
this
plc
is,
it
will
start
publishing
at
each
of
its
five
phases,
outputs
that
the
sva
members
and
the
broader
industry
can
take
a
look
at
and
make
use
of
in
their
own
evaluation
quick.
So
hopefully
this
helps
bridge
that
gap
between
you
know.
E
I
want
to
play
with
it
and
I
now
have
touched
it
played
with.
I
now
understand
it.
I
can
start
figuring
how
it
fits
in
american
picture
and
also
one
of
the
big
goals
here
is
to
be
able
to
bring
back
to
the
itf
real
life
experiences
around
this
to
say:
well,
here's
what
worked
here
is
what
didn't
work.
Here's
some
things.
We
think
we
made
directions,
we
may
want
to
suggest
going
in
to
improve
things
and
it's
bringing
together
here.
E
E
All
right
kyle,
are
you
guys
bingeing
the
key
or
do
you
want
me
to
manage
the
queue.
B
C
C
E
Well,
you
know
the
official
answer,
as
we
both
know
is
always
you
must
ask
the
lawyers
for
everything,
because
you
know
engineers
aren't
allowed
to
make
decisions
like
that.
But
you
know
so.
The
work
of
the
sva
really
can
be
divided
into
a
couple
categories
right.
One
of
them
is
sort
of
technical
specifications.
E
Those
may
contain
ip
related
stuff
and,
if
you're,
an
sva
member
part
of
joining
the
organization
is
of
course,
agree
the
ip
licensing
rules
which
really
focus
on
providing
licenses
to
other
sva
members.
E
If
they
want
to
negotiate
with
you
yeah
and
it's
not
the
sv,
unlike
the
itf,
isn't
essentially
you
know
like
as
open
a
ip
policy,
it's
very
flexible
if
you're
a
member,
if
you're
an
outsider,
it's
it's
different
right,
but
that's
for
the
specifications
and
the
really
technical
stuff
or
the
things
like
plc's
best
practices
and
white
papers.
E
Those
are
things
which
do
not
typically
contain
any
ip
related
assertions
or
materials,
because
they're
they're,
looking
at
the
industry
level
and
they're
looking
at
the
technology
level,
that's
available
across
the
open
industry
like
so
quick,
would
be
a
good
example
right.
We're
not
building
a
quick
server
at
the
eye
at
the
sva
we're
not
putting
sda
member.
E
You
know
ip
into
that
quick
server,
we're
taking
stuff,
that's
already
out
there
that's
available
off
the
shelf,
typically
open
source
and
and
taking
it
in
and
making
use
of
it
and
documenting
how
we
make
use
of
it
so
that
stuff
is
all
available
and
will
be
available
when
published
to
the
outside
world.
You
can
just
go
to
the
sva
website.
E
Click
on
the
download.
You
know
it
does
the
same
thing
where
it
asks
you
to.
I
think,
register
your
email
address,
so
we
know
who
you
are
and
then
you
download
it
and
and
right
the
documents
you
know
copyrighted
by
the
sva.
But
of
course
you
know
you
can
go
out
and
read
and
use
it
and
stuff
like
that.
The
copyright
is
to
prevent
you
from
republishing
it
yourself
as
your
own
work,
but
not
to
consume
it.
E
E
I
kind
of
liken
it
sometimes
unto
like
what
analog
does
right
now
is
full
of
people
who
actually
take
the
technical
specs
for
networks,
deploy
them
and
make
them
work
and
the
itf
writes
those
technical
specs
and
that
will
feed
back
to
the
itf
where
things
work
and
where
things
don't
work
and
ask
for
changes-
and
I
think
that's
a
very
similar
relationship
between
the
sva
and
the
itf
for
video.
C
E
And-
and
the
other
thing
I
should
add
to
is
obviously
we
have
like
sanjay
and
myself
were
very
active
in
the
itf
and
so
part
of
our
you
know.
Role
here,
I
think,
is
to
act
as
that
that
bridge,
where
it's
appropriate
and
and
to
pull
things
out
of
one
and
bring
them
over
to
the
other
one
and
there's
other
people
at
the
itf.
That's
you
know,
engage
with
the
sva
on
a
regular
basis
in
various
written
groups,
and
so
I
think,
there's
a
you
know,
an
ongoing
relationship
of
back
and
forth
here.
E
E
A
Say
that
there's
a
sort
of
a
meta
purpose
of
sharing
that
this
work
is
going
on,
because
it's
related
to
ietf
work,
the
the
ins
and
outs
of
how
how
you
know
people
here
actually
engage
in
work
there,
and
vice
versa,
are
a
little
a
little
trickier.
Perhaps.
E
And
I'll
point
out
that
you
know
we
do
the
opposite
to
the
two
groups,
like
the
sva.
We
take
the
updates
that
happen
at
the
itf
in
places
like
mops
and
a
bunch
of
us
go
back
to
the
sva
in
their
working
groups,
and
we
talk
there
about,
what's
going
on
in
the
itf,
to
make
sure
that
we've
got
cross-pollination
going
on
actively
and
ongoing
between
the
two
organizations.
F
I
think
you
you've
covered
it
pretty
well,
so
I
don't
really
have
anything
more
to
add.
Unless
there
are
any
questions.
E
And
you
can
follow
up
with
either
one
of
us
or
both
of
us
afterwards.
You
know
we're
obviously
here
on
the
ietf
and
we're
around
other
places,
so
don't
be
shy
all
right.
Thank
you
great
great.
Thank
you.
Both.
A
All
right,
so
I
think
jake
you
are
up
next,
with
the
follow
up
from
working
group
last
call
on
the
ops
con
stock.
C
Yeah
all
right,
I
think
I
can't
click
on
the
video
while
it's
trying
to
wait.
Oh
what
happened.
There's
a
modal
that
pops
up.
Okay,
I
guess
I
have
to
beat
that
up.
That
makes
sense.
Sorry,
I
think
it's
my
first
time
trying
to
use
this
particular
tool.
Usually
I
shared
my
screen
all
right.
It's
kind
of
neat,
though,.
C
B
You
you
lose
the
opportunity
to,
you
know,
have
embarrassing
notifications
show
up
to
everyone
in
the
in
the
working
group.
C
Well,
I
can
still
do
that.
Like
on
my
screen,
you
can
see
my
reflection
off
the
car
out
the
window
there
it's
that
time
in
the
morning
for
me
out
here
on
the
west
coast,
all
right,
so
I
was
just
going
to
give
an
update
on
our
our
adopted
draft
that
tell
everybody
where
we
all
are
so
our
last
call,
you
know
quote
passed
without
without
objections.
C
There
was
one
comment
in
support
of
advancing
it,
but
this
is
not
sort
of
the
you
know
overwhelming
support
that
we
see
in
the
room
for
some
of
the
some
of
the
questions
we've
we've
asked
from
time
to
time,
so
we
wanted
to
just
kind
of
confirm
it
and
make
sure
that
there
really
is
working
group
consensus
to
to
move
this
forward,
and
I
was
going
to
go
over
kind
of
the
the
status
on
our
last
few.
C
C
We
had
to
open
issues
remaining
when
we
finished,
and
I
guess
I'd
characterize
these
as
like
more
running
out
of
steam
and
and
having
conflicts
that
necessarily
we
decided
not
to
like
we.
We
had
a
couple
of
leanings
on
this
in
different
directions
and
some
questions
as
to
like
really
should
we
put
this
in
but
know
so
there.
C
This
was
the
last
thing
where
I
think
we
just
dropped
the
ball
on
kind
of
getting
it
in
its
final
state
was
assigned
to
me.
So
this
one's
my
fault,
you
know,
there's
there's
some
reasonable
things
that
an
operator
can
do
if
dns
is,
is
coming
into
play
as
a
potential
performance
problem,
and
so
it's
not
an
unreasonable
thing
to
include.
C
We
do
not
have
this
listed
as
one
of
our
blockers
for
going
ahead
with
black,
with
with
last
call,
I
expect
there
are
other
things
that
can
affect
performance
that
we
did
not
really
cover
in
our
in
the
dock.
In
the
end,
so
you
know
this
is
perhaps
a
gap,
but
I'm
not
sure
it's
such
an
important
gap
that
we
should
update
it
and
do
another
last
call.
C
And
the
other
one
was
kind
of
there
was
an
interesting
talk
last
time
about
the
comparative
latency
under
load
from
from
comcast,
and
we
had
a
little
bit
of
discussion.
I
think
spencer
had
the
idea
that
that
there
were
a
few
worthwhile
sentences
to
to
suggest
on
this
topic,
and
you
know
I
was
a
little
bit-
I'm
not
sure
it
really
fits,
but
you
know
lawfully
and
regardless
the
sentences
never
quite
got
submitted.
I
guess
I'd
put
this
as
another
one.
C
That's
maybe
nice
to
have
but
not
necessary,
but
that's
really
all
we
had,
and
I
guess
my
big
question
is
like:
are
we
really
ready
to
move
forward?
Can
we
I
don't
know
we're
on
a
poll?
C
Does
anybody
have
concerns
about
this
that
didn't
make
it
to
the
list
etc?
So
that
that
is
all
I
have?
I
guess
really.
What
I
wanted
is
a
discussion,
maybe
a
confirmation,
something
like
this
section
on
mock.
Yes,
thanks
ali
that'll
have
to
be
its
own
working
group,
I'm
afraid
in
my
opinion,
but
please
feel
free
to
disagree.
Flourish,
floors,
open,
I'm
inviting
cues.
H
H
H
So
I
really
urge
other
participants
in
this
working
group
that
are
here
or
not
listening,
basically
to
review
the
document
and
saying
like
glenn,
it's
okay,
I
support
like
it
is
that's
enough
right,
so
you
don't
need
to
put
a
ten
thousand
issue,
the
sort
of
the
better
some
specifically
for
you,
jake
right,
yeah
yeah.
So
I'm
slightly
concerned,
if
only
one
review
done
by
glenn,
it
may
cause
problem.
When
you
go
later
in
the
process
right
actually
yeah,
I
have
to
review
it.
A
So
I
see
two
people
in
the
queue.
I
think
that
probably
where
we
want
to
close
out
today's
activity
is
to
take
names,
who's,
willing,
who's,
willing
to
review
or
and
or
just
send
a
note
to
the
list
saying
that
you
know
you
read
it
and
support
it's
going
forward,
but.
F
E
So
I
can't
volunteered
review
pieces
since
apparently
I
already
did
a
review,
but
I
will
take
this
over
to
the
working
groups
over
the
sva
and
make
them
aware
of
the
document
as
well,
so
they
get
some
eyeballs
from
that
organization
on
it
and
I'll
spread
it
around
some
social
contacts
I
have
in
the
industry,
because
I
think
this
is
some
important
stuff
to
eric.
E
I
want
to
point
out
that
you
know
because
we're
now
using
github
for
the
the
editor
process
for
this,
I
will
point
out
that
tremendous
amount
of
conversation
actually
took
place
on
that
github
and
I'm
wondering
if
somehow
we
it
would
be
appropriate
to
make
a
shepherd
aware
of
the
github
conversations
that
went
into
the
process
because
historically
it
would
have
been
on
the
list,
but
now
it's
sort
of
on
this
other
place.
So
I
don't
know
how
to
solve
that.
But
suggestions
would
be
welcome.
H
There
is
indeed
a
dichotomy
somehow
between
the
github
and
the
mailing
list,
but
until
now
it's
still
a
mailing
list
that
counts
and
if
you
open,
for
instance,
couple
of
issues
in
the
github,
no
things
block
you
to
send
in
the
working
group
last
call
nothing
prevents
you
to
say
on
the
meaningless.
I
have
open
three
four
five,
a
github
issue
and
that
counts
right.
Musically.
C
Yeah,
I
guess,
as
as
an
author,
I
I'd
also
like
to
point
out
that
contributing
to
the
document
by
you
know,
commenting
and
opening
issues
does
not
imply
support
for
getting
the
document
to
the
you
know
for
advancing
the
document
through
last
call.
This
is
these
are
two
separate
things
and
while
we
are
are
very
grateful,
we
we
need
a.
C
I
would
say
we
need
a
separate
confirmation
of
that.
Well,
what
regardless
of
the
venue
but
yeah
the
mailing
list,
as
eric
said,
is
the
is
the
normative
one.
There.
J
I
would
be
happy
to
review
the
document
as
well.
I
just
want
to
say
the
document
looks
really
good.
There
are
good,
comparing
and
contrasting
of
various
options
and
section
7
is
comprehensive,
but
I'll
I'll.
Definitely.
E
C
Great
thanks,
I
guess
I
have.
I
have
one
question
also
that
occurred
to
me
like
in
in
some
other
working
groups.
I've
seen
like
a
request
to
send
like
we're
not
aware
of
any
ipr
for
all
the
authors,
and
maybe
we
support
advancing
this
document.
Also,
I
I
didn't
do
that.
I
kind
of
forgot
that,
should
I,
I
guess.
A
Yeah,
well
I
mean
welcome
you're
you're,
our
first
document
for
this
working
group,
so
I
think
we're
kind
of
figuring
out
what
our
process
is
as
well.
As
you
know,
as
we'll
get
into
a
little
later,
dealing
with
the
fact
that
we're
not
actually
meeting
in
person
and
kind
of
struggling
to
the
dichotomy
is
not.
I
don't
think
the
dichotomy
is
between
github
and
the
mailing
list.
A
I
think
the
dichotomy
is
between
people
who
come
to
the
meetings
and
people
who
participate
on
the
mailing
list,
so
I
think
it
would
be
fine
and
helpful
for
the
for
the
the
authors
to
do
that.
I
think
that
I
mean
I,
as
I
said.
I
think
we
need
to
take
some
names
and
and
already.
K
A
I'm
I'm
yeah,
I'm
appreciative
of
glenn's
offer
to
take
this
to
a
wider
audience.
At
the
same
time,
as
you
know,
we
are
in
last
call
right.
This
is
not
a
what
document.
Would
you
write
if
you
were
writing
a
document
request?
This
is
a
is
this
document
in
reasonable
shape
to
meet
its
objectives?
Yes,
no.
What
can
we
do
to
fix
it
kind
of
call
ali.
L
So
to
jake's
question
at
least
when
I
was
a
working
group
chair.
I
you
know.
As
the
chair,
I
asked
the
group
whether
there
was
any
ipr
considerations
from
anybody,
not
just
authors.
Obviously
anybody
from
the
working
group
and
but
beyond
that,
since
this
is
an
informational
document,
I'm
not
sure
whether
the
ipr
declarations
do
apply
to
this
one.
L
A
L
C
All
right
that
is
it
then
I
guess.
If
nobody
else
has
comments,
I
don't
know
if
you
guys
want
to
run
a
poll
in
the
room
or
we'll
just
do
the
list,
but.
A
Yeah
and-
and
I
guess
kyle,
what
do
you
think
I
would
be
happy
to
to
get
more
names
from
people
here
who
are
willing
to
go
through
the
document
and
you
know
send
a
thumbs
up
or
questions
to
the
mailing
list.
I
don't
know
if
you
want
to
do
that
as
a
poll,
if
we
want
to
ask
people
to
volunteer
in
in
the
in
the
chat
or
if
we
just
want
to
take
the
mailing
the
attendance
list
and
send
mail
to
everybody,
I'm
not
above
mail
spam.
I
mean.
B
A
I
just
remember
the
very
clicky
keyboard
in
the
last
meeting,
the
okay.
So
how
about
this?
Let's
see
if
we
can
find
all
right
you've
got
one.
I
was
aiming
for
three:
let's
go
for
five
people
in
the
in
the
chat
who
are
willing
to
review
the
document
and
send
a
thumbs
up
or
a
question
to
the
mailing
list
for
last
working
last
call
within
the
next
two
weeks.
A
A
Great
you
jake
and
co-authors,
and
thank
you,
everyone
who
has
volunteered
all
right
so
now
on
to
the
media
operations
use
case
for
an
augmented
reality:
application
on
edge
computing
infrastructure.
L
Well,
leslie,
I
have
a
question
before
we
move
on
so,
as
you
will
remember,
we
decided
to
have
a
live
document
at
the
end
of
the
rfc
eventually,
and
we
haven't
really
received
anything
from
the
rfc
editor
at
this
point,
but
if
maybe
maybe
it
needs
to
be
taken
care
of
during
the
last
call
as
well,
maybe
ad
or
someone
else
needs
to
make
that
clear.
L
At
this
point
we
have
a
tiny
url
link
that
you
know
links
to
the
google
document,
but
eventually
we
want
to
have
a
base
in
the
you
know
somewhere
in
the
atf
domain.
So
if
anybody
has
any
idea
how
to
do
it
or
maybe
whether
we
should
take
an
action
at
this
point
to
to
do
that,
it
might
be
good
to
check
as
well
yeah.
A
I
suspect
it's
something
that
we'll
have
to
deal
with
either
in
working
group
last
call
or
ietf
wide
last
call
and
see
if
it's
get
us
to
a
point
where
everyone
is
comfortable,
that
we
have
a
usable
answer,
but
I
don't,
I
don't
think
we're
quite
there
today.
A
Cool
all
right
over
to
you,
raymond.
J
J
J
The
updates
are
in
the
form
of
reorganizing
section
3.1
around
the
notion
of
ar
data
processing
pipeline.
We
have
also
added
section
5,
where
we
summarize
the
network
characteristics
of
ar
application
traffic
as
well
as
tcp's
interaction
with
4g
lte
network
carrying
ar
traffic.
So
let
us
look
at
those
updates.
J
Firstly,
we
need
to
track
the
three-dimensional
coordinates
and
six-dimensional
pose.
That
includes
coordinates
and
orientation
of
objects
in
the
real
world.
These
track
natural
features
are
then
used
to
develop
an
annotated
point,
cloud-based
model.
Finally,
the
coordinate
systems,
brightness
and
color
of
virtual
and
real
objects
need
to
be
aligned
in
a
process
called
registration.
J
In
addition
to
the
requirements
for
abr
algorithms,
there
are
other
operational
issues
that
need
to
be
considered
for
ar
use
cases
such
as
the
one
described
in
this
draft.
In
a
study
conducted
to
characterize
multi-user
ar
or
cellular
networks,
the
following
issues
were
identified
and
have
been
added
in
section
5..
J
J
J
We
would
like
to
thank
spencer,
dawkins,
rohit,
abhishek
and
jake
holland
for
providing
very
helpful
feedback
on
the
mailing
list.
We
would
give
our
detailed
response
to
the
feedback
on
the
mailing
list.
We
would
like
to
take
this
opportunity
to
invite
reviewers
and
contributors
to
improve
the
draft.
The
draft
is
now
on
good
help.
Many
thanks
to
kyle
rose
for
helping
out
as
a
starting
point
for
our
discussion,
as
spencer
has
suggested
in
the
mailing
list
for
future
versions
of
the
draft.
J
A
M
Yeah,
so
I
read
the
draft,
it
looks
it
it's
an
excellent
start.
I
think
I
agree
with
a
lot
of
the
feedback
that's
already
been
given
on
the
list,
and
I
apologize
for
not
getting
my
feedback
on
the
list
before
the
meeting.
M
I
said
that
I
wouldn't
I
didn't,
but
I
will
get
you
some
feedback.
One
of
the
things
I
did
observe
is
that
the
there's
a
lot
of
discussion
about
generation
and
processing
on
the
device
itself
and
with
the
intended
audience
is,
you
know,
network
operators
to
me
it
was
in
concretely
it
was
not
concretely
tied
enough
to
what's
happening
on
the
network
and
where
those
latencies
are
coming
from
and
time
budget
being
spent
in
one
place
versus
another.
M
J
We
can
add
more
text
to
the
draft
clarifying
in
more
precise
terms,
what
are
the
various
steps
which
take
the
which
make
up
the
whole
time
budget?
I'd
also
like
to
add
here
that
the
description
of
the
various
ar
processes
are
just
a
brief
sketch.
They
are
not
meant
to
be
comprehensive.
If
someone
is
interested
in
a
comprehensive
discussion,
they
can
look
at
the
references,
so
the
the
discussion
of
er
is
is
deliberately
kept
very
brief,
but
yeah.
We
are
open
to
more
suggestions.
A
So
I
appreciate
that
you
will
go
through
the
detailed
responses
on
the
mailing
list,
which
is
absolutely
appropriate,
because
there
were
some
some
good,
in-depth
reviews,
but
I
did
think
it
would
be
worth
it
if
the
working
group
would
actually
yeah.
If
the
working
group
would
here
discuss
a
little
bit
more
about
one
of
the
points
that
jake
raised,
which
is
you
know
whether
whether
this
document
as
it
stands
is
straddling
at
least
two
worlds
and
best
pick
which
one
I
don't
know
reena.
J
No,
absolutely,
I
think,
I'll,
give
my
comments
on
that,
and
I'll
definitely
want
to
hear
what
the
working
group
wants
to
say.
So
I
think
both
spencer
and
jake
have
raised
this
issue
of
the
scope
and
the
intended
audience
of
the
document.
C
N
E
J
Vr
and
360
degree
streaming
applications
on
the
utility
of
edge
computing
resources,
so,
as
spencer
put
it,
I
think
he
said
that
reasons
to
deploy
edge
computing
with
the
reasons.
E
J
A
B
That
I
I
sort
of,
I
guess
one
of
my
fears
is
that
the
is
that
by
using
by
by
having
the
you
know
the
the
term
ar
in
the
title
and
having
it
be
focused
directly
on
that
use
case,
that
a
lot
of
people
would
be
like
augmented
reality.
I'm
not
doing
that,
even
though
the
concerns
would
be
would
be
shared
by
a
much
larger
group
of
people
anyway,
I
just
wanted
to
throw
that
out
there.
That's
something
that
I've
been
sitting
on
for
a
while.
J
Yeah
and
I
I
think
I
would
agree
with
that,
as
the
draft
progress
we
have
realized
that
its
scope
should
broaden
or
or
more
precisely,
you
should
abstract
out
of
particular
er
applications
into
more
general
issues
which
resources
on
the
edge
can
can
help
mitigate.
C
Right
so
that
that
was
one
of
the
options,
so
that's
good
that
we're
hearing
plus
ones
to
that
side
of
it.
I
guess
I
would
say
that
there
are
some
so
like,
like
renown
was
saying,
I
think,
there's
there
are
some
tighter
constraints
for
that.
Come
from
the
head
motion
kind
of
constraints,
then,
basically
anything
I've
heard
of
yet,
including
gaming
like
for
that
one
you're.
Looking
at,
maybe
I
think
last
I
saw
a
45
millisecond
total
budget,
but
this
one's
more
like
15
at
the
outside.
C
If
I
remember
right,
so
that's
a
significant
difference.
That's
that's
a
scope
constraint
potentially,
but
I
just
do
you.
You
know
you
said
there
are
many
applications
that
need
it
did.
You
have
others
in
mind
here:
kyle.
J
So,
besides
ar
we
are
360
degrees
streaming,
there
is
also
a
certain
vehicular
use
cases
that
would
require
edge
computing
resources.
Certain
iot
use
cases
as
well,
and
the
one
common
thread
across
all
of
those
applications
is
the
requirement
for
low
latency
and
also
the
fact
that
the
devices
are
typically
battery
driven
and
so
that
limits
the
amount
of
resources
that
is
available,
and
that
also
causes
this
heat
problem.
That
also
causes
the
battery
drainage
problem.
J
So
so
across
all
of
those
kinds
of
use
cases,
you
have
these
similar
requirements
and
they
can
be
mitigated
by
various
engineering
solutions
involving
edge
resources.
B
Yeah
I
mean,
I
guess
I
guess
it
almost
seems
like
there
are
like
there
are
two
levels
of
docs
that
are
that
are
wanted
here,
because
there
are
there's
a
unique
combination
of
problems
that
augmented
reality
and
and
virtual
reality
share.
B
But
the
individual
problems
are
then
then
overlap
with
other
areas.
So
it's
almost
like
you
want
the
augmented
reality
problem
statement
and
then
have
that
refer
to.
You
know
up
to
two
other
to
documents
and
other
areas
that
are
addressing
subsets
of
that
problem.
Does
that
make
sense.
J
Yeah,
what
I
would
say
is
that
the
at
the
operational
level,
as
spencer
says
in
his
mail,
there
are
a
lot
of
things
like
the
choice
of
the
transport
protocol
that
are
not
necessarily
decided
by
the
network
operators,
but
by
the
by
the
application
developers,
and
that
that
that
then
means
that,
at
the
network
operator
level,
there
are
these
these
paired
sort
of
traffic
patterns
that
they
need
to
deal
with,
to
provide
certain
say
qoe.
J
So
qci
comes
to
mind
which
are
provided
by
4g
and
5g
vendors.
So
so,
yes,
there
is
a
set
of,
let's
say
requirements
posed
by
the
by
the
ar
application,
which
are
also
then
the
requirements
for
other
kinds
of
applications
like
certain
iot
applications
or
autonomous
cycles,
etc,
etc.
J
So
I
think
the
fact
that
that
you
can
use
engineering
solutions
load,
balancing
deciding
where
to
offload
when
to
offload
those
sorts
of
things
in
combination
with
with
the
edge
devices
is,
is
sort
of
a
common
thread
across
all
these
applications.
I
think.
D
J
I
have
given
a
reference
to
that
and
and
typically
the.
N
J
That
the
application
designers
come
up
with
have
to
be
tied
to
actual
network
things
like
latency
and
bandwidth.
So
so
yes,
there.
M
I
J
Metrics
that
you
can,
you
can
use,
but
they
need
to
be
translated
into.
You
know
the
standard
network
performance
metrics
like
latency
and
memory.
A
That's
great,
thank
you.
I've
been
putting
myself
in
queue
and
coming
out
of
queue
and
in
queueing
out
of
cube,
because
the
the
around
the
question
of
you
know
is:
do
we
need
to
capture
sort
of
the
use
case
of
ar
and
and
and
tape
it
up
as
the
target?
A
Around
what
what
is
it,
what
is
special
about
it
that
requires
edge
computing,
you've
listed
a
lot
of
the
video
related
ones
and
which
is
appropriate
for
mob's
document.
But
it
strikes
me
too
that
it
might
be
worth
capturing
some
of
the
other
compute
challenges
that
are
make
it
not
possible
to
address
these
issues
by
traditional.
A
You
know
video
challenge
technologies
such
as
cdns
and,
in
particular,
I'm
thinking
about
whether
it's
ar
or
vr
or
the
metaverse,
the
whole
notion
of
synchronization
of
objects,
or
we
or
shared
reality.
If
you
will
and
how
that
impacts.
You
know
the
fact
that
you,
you
really
do
need
to
do
local.
You
know
you
need
to
break
the
problem
down
based
on
where
you,
what
you're
competing
locally
versus
what
you
are
yeah
achieving
through
communication.
A
J
J
Well,
for
definitely
for
ar
and
vr
applications,
the
the
latency
is
important,
and
one
of
the
reasons
why
that's
important
is
because
of
synchronization
issues
that
we
have
discussing
in
the
draft.
Also,
so
so
definitely,
whether
you
talk
of
your
vr
or
multiverse,
you
would
need
to
you
need
to
think
about
a
synchronization
and
a
way
to
to
deal
with
with
the
latency
issues.
J
Also
because
most
of
these
devices
are
on
on
wireless
networks
that
that
are
very
variable
in
their
quality.
So
so
so
yeah
definitely
synchronization.
A
And
I
think
part
of
my
point
was:
it
is
the
nature
of
the
problem
space
that
there
are
certain
things
that
you
can
split
across
nodes
and
there
are
certain
things
you
cannot
right.
So
I
mean
it's
the
age-old
problem
of
multi-parallel
processor
computing
and
shared
memory
and
or
messaging
right
when
you're
solving
a
problem.
What
are
the
things
that
actually
need
to
happen
locally
and
what
are
the
things
that
can
be
solved
by
synchronized
communication
elsewhere,
and
there
are
some
things
that
just
cannot
be
pushed
off
again.
A
I
didn't
want
a
rat
hole
there,
but
just
in
terms
of
there's
a
if
we
want
to
talk
about
the
problem
space
in
a
means
that
is
somewhat
agnostic
of
a
particular
application,
focusing
it
on.
You
know
the
level
of
breaking
that
breaking
the
problem,
the
compute
problem
apart
and
what
needs
to
be
done
locally
versus
not
locally,
and
therefore
what
video
processing
has
to
happen
locally
versus
not
locally,
and
one
of
the
implications
might
be
helpful
or,
as
I
said,
it
might
be,
a
complete
red
herring.
J
G
J
Of
mobility
and
sensing,
whereas
the
edge
deals
with
the
design,
consideration
of
network
proximity
and
the
cloud
deals
with
the
design.
Consideration
of,
I
think
compute,
elasticity,
storage,
permanence
and
hardware
consolidation,
and
within
these
design
constraints
you,
I
guess
you
need
to
consider
synchronization
and
how
it
is
affected
by
the
special
nature
of
arvr
applications.
A
Okay,
so
I
think
I
don't
see
anyone
else
in
queue,
so
I
think
that
the
next
steps
are
renin
you're
planning,
a
more
detailed
response
to
the
comments
received
on
the
mailing
list
on
the
mailing
list.
I
will
take
this
opportunity
to
observe
that
the
document
is
now
in
github
and
thanks
to
you
and
kyle
for
getting
that
sorted,
so
people
who
prefer
to
do
github-based
engagement
can
now
do
so,
and
we've
picked
up
at
least
one
more
person
offering
to
do
review.
A
A
Okay,
all
right,
thank
you
very
much
reena
and
we'll
head
to
the
mailing
list.
Kyle,
you
were
gonna,
say
something.
B
A
Okay,
so
I
think
that
we
probably
should
do
that
we'll
tell
in
the
chat.
We
will
tell
spencer
that
we've
reordered
the
agenda
and
if
you
can't
join
us,
then
I
think
we
can
probably
speak
from
his
notes
to
the
mailing
list.
O
A
Okay,
all
right
lee,
if
you
could
give
us
your
update
on
the
packet
differences.
Yes
thank
you.
K
A
K
B
If
you
want
to
share
slides,
you
can
click
on
the
share,
preloaded
slides
button
at
the
top
left
and
then
select
your
your
deck.
K
B
Yeah
we're
looking
at
your
screen
as
opposed
to
the
slides
that
were
that
were
in
the
that
we
uploaded
from
the
data
tracker,
but
it's
fine.
You
might
want
to
put
it
into
presentation
mode,
though,
so
we
don't
see
all
of
the.
K
Yeah
great,
so
I
wanna
have
a
summary
of
the
feedback
that
I
received
from
the
last
meeting.
So
the
comments
were,
I
think.
In
summary,
I
think
the
comments
are
mainly
related
to
how
to
review
the
characteristics
of
the
microblocks
containing
the
packets
to
the
network
and
implement
the
selective
packet
dropping
at
the
packet
level.
K
So
I
want
to
emphasize
on
the
packet
level,
so
I'm
very
glad
that
the
concept
is
very
well
received
in
my
mind,
and
we
think
it
is
beneficial
to
improve
the
quality
of
experience
of
video
streaming
by
reducing
the
package.
Dropping
granularity
from
the
current
traffic
class
level
to
the
individual
packet
level,
which
means
inside
a
flow
each
individual
packet
may
have
its
own
significance
level
and
the
router
could
treat
those
individual
packets
differently
or
even
within
the
within
the
packet
itself.
K
Yes
and
we
need
such
information
of
whether
this
packet
is
significant
or
not
to
be
reviewed,
network
nodes.
For
example,
api
could
be
implemented
to
input
such
information
or
metadata
from
the
application,
which
might
be
mapped
to
the
you
know:
ipv6
extension,
header
or
ipv4
options
or
a
dedicated
metadata
field
in
the
ip
header.
K
We
will
take
a
closer
look
at
your
your
comments,
because
we
haven't
gotten
much
time
in
the
morning
to
read
carefully
about
each
sentence
in
the
email
and
we
will
reply
you
later
offline,
but
for
the
for
the
div
serve,
I
want
to
say
is
the
traffic
management
is
at
the
class
level,
which
means
the
ad
router
will
look
into
some
fields
like
source
destination,
addresses
and
flow
id
to
map
traffic
to
map
list
packet
to
certain
traffic
class,
which
means
that
inside
of
flow
the
packets
would
have
the
same
significance.
K
Next,
I
want
to
talk
a
bit
more
on
the
major
characteristics
of
video
frames
that
could
affect
the
significance
of
a
single
packet.
The
first
one
is
the
frame
type
we
all
know.
Pack
4
has,
you
know,
have
three
types
of
frames:
iframe
is
self-contained,
so
which
means
it
will
have
no
references
to
other
frames
and
it
is
encoded
by
itself,
and
it
is
the
keyframe
that
provides
checkpoints
for
re-synchronized,
synchronization
or
re-entry
to
support
trig
modes
and
aerial
recovery.
K
We
we
can
think
this
is
most
important.
Type
of
frames
in
a
whole
stream
and
p
frame
is
the
predictive
frame,
which
means
it
can
allow
macroblocks
to
be
compressed
using
temporal
prediction.
In
addition
to
spatial
prediction,
so
for
a
motion
estimation
keyframes
could
use
frames
that
have
been
previously
encoded.
K
It
could
be
a
proof,
p
frame,
it
could
be
iframe
so
for
a
b
frame
it
stands
for
bi-directional
frames,
which
means
I
will
refer
to
the
frames
that
occur
both
before
and
after
it.
So,
each
macro
block
of
a
b
frame
can
be
predicted
using
backward
prediction
or
using
forward
prediction.
So
accordingly,
the
next
characteristic,
I
want
to
point
out,
is
the
whether
the
packet
contains
frame
that
are
referenced
by
other
frames.
K
So,
according
to
this
prediction,
relationship
among
different
frames,
so
each
frame
might
be
referenced
by
other
frames,
so
for
iframe
the
iframe
does
not
refer
to
any
other
frame,
so
is
is
at
least,
but
it
is
at
least
referenced
by
a
keyframe
after
it.
So
the
losing
the
first
iframe
in
the
group
of
pictures
would
cause
videos
picture
even
missing
a
few
seconds
because
p
and
b
frames
referencing
to
I
from
it
would
not
be
decoded
nor
displayed
either
for
a
p
frame
it
because
it
refers
to
a
picture
in
the
past.
K
However,
you
also
might
be
referenced
by
a
p
frame
after
it
or
a
b
frame
before
or
after
it,
and
for
a
b
frame
can
I
it
can
also
act
as
a
reference
in
the
real
case.
It's
not
a
common
case.
K
Another
important
characteristic
of
impact
video
packets
is
the
movement
level
of
the
video
sample
contained
in
the
packet,
the
video
things
with
the
low
level
movement
or
less
sensitive
to
both
b
frame
and
p
frame
packet
loss,
for
example.
If
right
now
we
are
having
a
conference
call
my
background
or
my
body.
K
If
I
did
not
move
much,
it
has
very
low
level
movement
contained
in
the
video,
so
the
the
video
things
with
you
know
were
not
very
sensitive
to
the
b
frame
or
p
frame
packet
loss,
but
on
the
other
hand,
if
you're
watching
football
video
on
live
video,
it
is
you
know,
the
movement
level
is
very
high
and
you
will
be
very
sensitive
or
more
sensitive
to
both
b
frame
and
p
frame
packet
loss
so
for
loss.
K
The
p
frame
can
impact
the
remaining
part
of
the
gop
or,
while
a
lost
b
frame
has
only
local
effects.
You
know
slowly
moving
content
or
very
large
stack
background.
Like
you
know
my
video
right
now,
if
I'm
sharing
so
you
know,
think
of
dynamically
moving
content,
loose
b
frame
has
more
true
magic
impact
and
scale
can
be
as
far
reaching
as
a
p
frame
loss.
K
So
if
we
can
have
some
level
of
estimation
or
indication
of
the
movement
level
with
of
the
video
sample
containing
packets,
we
can,
you
know,
combine
with
other
knowledge
of
the
packet.
We
can
have
more
information
regarding
the
significance
of
the
packet.
K
K
For
example,
we
are
watching
a
video,
you
are
more,
you
are
paying
more
attention
to
certain
object,
or
certain
person
in
the
video
than
other
microblocks
that
are
representing
other
objects
or
background
is
not
are
not
so
important.
K
Compared
compared
to
the
macroblocks
of
the
roi
so
based
on
those
characteristics
of
the
mempac
video
packets,
you,
we
probably
could
use
some
some
way
of
calculating
the
import
importance
of
those
of
those
of
the
individual
packets.
K
I
did
not
show
here
in
the
in
the
slides,
but
what
I
can
think
of
is
you
know,
based
on
the
frame
type.
We
think
that
iframe
is
the
most
important,
so
we
probably
give
more
weight
to
the
eye
frame
frame
and
for
those
frames
that
are
referenced
by
others
are,
would
be
given
more
weight
and
for
the
movement
based
on
the
movement
level.
K
You
know
if
the
movement
level
is
high,
then
the
b
frame
and
p
frame
would
would
have
more
importance
or
more
weight
than
b
frame
and
p
frame
with
less
movement.
K
And
then
you
know,
if
would
think
the
the
pack
leaf
has
roi
object.
Then
maybe
we
have
higher
weights,
so
we
can
combine
all
those
factors
into
a
certain
calculation
method.
I
I
personally,
I
don't
have
a
good
answer,
but
we
can
think
of
one
light
by
putting
different
weights
on
those
important
things.
Then
maybe
we
can
come
up
of
a
significant
level
force
or
individual
packets.
K
So
here
is
the
summary
of
the
requirements
of
network
and
applications.
So
we
think
the
application
needs
to
in
order
to
make
this
work
we
need.
It
needs
to
reveal
some
information
to
network
to
enable
is
selective
packet.
Dropping
you
know,
examples
could
be,
you
know,
receiving
any
users,
rough
preference
on
the
media
quality.
So
what
is
your
tolerance,
quality
degradation
regarding
to
the
example
resolution?
So
if
you
are
watching
before
you
know,
4k
video
is
it?
Is
it
okay
for
you
to
watch?
K
You
know
480
p,
video
or
not.
If
it
is
okay,
maybe
some
of
the
packets
does
not
need,
do
not
need
to
be
dropped,
but
those
you
know
like
the
iframes
should
be
maintained.
K
Other
frames
could
be
dropped
and
other
things
like
the
characteristic
of
the
media
content
I
presented
earlier
and
also
you
know
like-
are,
for
example,
the
ior,
the
frame
type,
whether
the
pack
contains
frames
that
are
referenced
by
others
and
correspondingly
the
network
should
be
able
to
leverage
above
information
reviewed
by
the
application
and
then
selectively,
drop
packets
or
potentially
parts
of
the
packets
from
com
from
competing
media
stream
flows
and
re-transmission.
K
We
can
think
in
that
way
re-transmission
could
be
maximally
eliminated,
because
some
of
the
packets
that
are
important
have
already
received
by
the
receiver
and
the
receiver
can
accept
this
qualitative
degree,
qualitative
degradation,
then
other
parts
of
the
packets
do
not
to
be
retransmitted
again,
so
the
receiving
end
user
is
able
to
consume
in
the
deliver
packets
as
many
as
possible
in
time
with
acceptable
quality.
K
And
this
is
the
slide
I
added
just
now.
Those
are
the
references
to
papers
we
have
published
and
the
very
first
one
which
we
will
publish
in
2018
and
the
recent
one
that
were
very
relevant
to
this
paper.
Oh
sorry,
to
this
contribution,
is
this
one?
The
the
coded
communication
for
emerging
network
applications
with
new
iep,
which
is
to
be
appeared
as
a
invited
paper
in
proceedings
of
the
17th
international
offers
on
mobility,
sensing
and
networking.
B
C
The
I
guess
I'd
like
to
to
clarify,
I
think
you
characterize
the
responses
as
positive.
So
far,
I'm
not
sure
I
would
call
my
view
on
this
positive
at
this
stage.
I
think
there
might
be
potential
here,
but
what
I'd
be
interested
to
see?
Is
it
so
first
off,
I
think
the
the
problem
of
differential
dropping
that
you
communicate
to
the
network
is
exactly
what
diffserv
can
do
and
that's
a
that's,
a
good
approach.
C
That's
a
much
better
approach
in
my
opinion
than
option
header
ideas
that
you
mentioned,
but
I
think
there
are
challenges
and
that
you
you'd
have
to
be
looking
at
a
controlled
at
a
controlled
network
either
way.
C
C
Really
given
a
lot
more
time,
but
that's
just
kind
of
my
view,
I
think
it's
it's
a
potentially
interesting
thing,
but
like
the
the
fact
that
iframes
are
different,
does
not
necessarily
imply
that
dropping
them
less
often
gives
you
a
better
qoe
without
the
sort
of
measurement
to
that
effect,
and
I
guess
I'd
like
to
you
know,
there's
probably
constraints
on
that.
I
guess
I'd
like
to
see
an
exploration
of
what
do
you
actually
get?
That's
that's
my
main
feedback
here,
but
thank
you.
K
Yeah,
no,
I
I
did
not
regard
regard
your
email
as
a
very
positive
feedback,
but
I
do
thank
you,
for
you
know
replying
me
through
the
email
list,
giving
me
comments
for
deep
serve.
I
want
to
point
out.
It
is
a
course-grained
class-based
mechanism
for
traffic
management,
which
means
at
the
edge
of
diffserv
domain.
K
The
edge
router
would
look
into
the
source
destination
addresses
and
the
flow
id
to
determine
the
you
know
the
dscp
value
for
that
packet.
So
it
does
not
mean
the
packet
within
the
flow
would
have
different
significance,
levels
or
dsap
assignment,
because
the
router
would
only
look
at
the
source
and
destination
and
the
flow
id.
We
will
not
look
at
anything
else
to
determine.
C
C
I
agree
if
you
use
the
router
to
set
the
value.
However,
I
think
you
can
set
the
value
at
the
host
and
have
it
come
in
with
the
disk
serv
id,
which
then
would
be
treated
differently
by
the
router.
So
I
guess
I'd
encourage
you
to
look
at
that.
C
Instead,
where
the
sender
sets
it
according
to
the
sock
up,
there's
the
iptos
api
in
in
linux,
for
example,
and
if
you
can
set
that
at
the
host,
then
I
think
you'll
be
able
to
perform
experiments
that
do
not
depend
on
router
understanding
the
class
and
just
use
the
class
just
based
on
the
incoming
diffserv
at
the
router
layer,
when
you
try
to
when
you
try
to
go
between
different
diffserv
domains,
but
just
to
just
declare
clarify
that
point,
but
good
luck
that
it
sounds
interesting
if
it
if
it
really
pans
out.
C
So
you
know.
K
P
Thanks
jake
magnus,
yes,.
Q
To
my
understanding,
you
need
to
ensure
that
you
can
actually
separate
out
the
data
you
want
to
drop
for
both
that
it
works
through
both
your
security
solutions
and
your
framing
solutions
and
formatting,
yet
still
get
readable
by
stream.
So
I
mean
I,
I
hope
you
have
some
qe
results
that
you
really
can
show
that,
because
before
you
can
show
it
that
even
get
that
with
the
framing
in
place
for
those
applications.
I
don't
see
a
point
of
doing
this
because
I
think
that's
usually
being
when
looking
at
this,
this
kind
of
barrier.
K
Yeah
yeah,
thank
you
for
your
comment.
Yeah,
yes,
so
be
honest.
I
I
my
background
is
not
video
coding
sites,
I'm
more.
You
know
networking
side
or
mostly
networking
side.
So
for
here
we
want
to
point
out.
The
you
know
there
there
is
could
be
significance,
differences
among
package
by
using
this
use
case.
K
We
did
some
research
on
what
are
the
characteristics
could
be
could
be
leveraged,
but
I
think
those
characteristics
could
be
you
know
does
do
not
need
to
be
revealed
to
the
network.
You
want
one
hand
because
you
know
if
we
could
have
the
help
from
the
codex
or
application
by
saying
okay
list
packet,
because
it
has
those
kinds
of
characteristics.
So
I
would
indicate
it
is
more
important
than
the
other
packets
in
the
stream,
so
we
will
have
the
application
or
codecs
to
do
that.
Q
Sure,
but
I
mean
the
the
the
issue
here
is
that
you
have
a
very
interactive
system,
is
that
if
you
drop
a
certain
amount
of
packets,
your
video
quality
will
be
subjected
to
different.
It
will
really
get
different
results
and
that
depends
on
how
your
system,
so
you
actually
need
to
have
it,
do
both
parts
here
to
figure
out
and
see
what
results
you
actually
can
get,
and
that,
I
think,
is,
is
the
kind
of
homework
you
need
to
go
do
here,
and
you
also
say
no
there's
some
properties
here.
Q
K
Yeah
yeah,
I
understand
yes,
I
think
your
comments
are
very
legitimate,
so
yeah
the
application
yeah.
We
definitely
need
to
support
front
application
side,
because
in
order
to
have
this
information
or
the
significance
of
the
packet
known
to
the
to
the
you
know,
network
list
information
needs
to
be
calculated
or
some
kind
of
a
revealing
front
application
to
the
network
and,
as
you
said,
dropping
some
parts
some
packets
may
result
in
the
quality
of
quality
degradation.
K
But
if
the
any
user
is
able
to
tolerate
that,
then
maybe
it
is
okay
right.
So,
as
I
said,
if
you
know
I'm
watching
4k
at
certain
point,
I'm
only
watching
480p,
I'm
okay
with
that,
then
the
retransmission
of
those
other
frames
may
not
need
it,
but
I
think
what
you
commented
earlier
very
legitimate
that
the
application
does
need
needs
to
support
less
kind
of
operation
in
the
network.
A
Follow
up
from
the
specific
discussion
points
as
this
as
outlined,
I
think
the
thought
people
probably
want
to
be
putting
putting
on
the
table
is.
If,
if
this
is
going
to
go
forward
at
the
itf,
what
is
the
appropriate
place
for
it
to
go?
I
don't
think
the
mops
working
group
would
be
the
right
place,
but
perhaps
we
can
help
shape
it
up
to
the
point
where
it
is
something
that
can
go
to
another
working
group.
A
A
All
right,
so
spencer
is
still
busy
elsewhere.
I
don't
think
we
are
going
to
skip
right
over
the
mock
item.
I
think
that
we
need
to
at
least
speak
a
little
bit
about
the
notes
that
spencer
sent
to
the
mailing
list,
so
the
media
over
quick
side
meeting
occurred
earlier
this
week.
I
was
not
there,
I'm
probably
not
the
best
person
to
to
present
the
report
out
from
it.
A
B
A
All
right
so
spencer,
as
I
said,
forwarded
the
mailing
list
to
point
her
to
his
notes
from
the
meeting.
B
Luke
is
in
the
is
in
the
queue
now.
N
N
Really
it
comes
down
to
there's
a
lot
of
different
media
use
cases
right,
there's
a
lot
of
different
protocols
out
there.
You
guys,
you
guys
know
that
there's
this
whole
swap
and
there's
this
new
transport
layer,
quick
and
really.
This
is
just
an
exploration
of
what
we
could
use
quick
for.
N
There
are
some
companies
and
some
individuals
that
are
using
it
already
to
to
deliver
media,
but
most
of
the
meeting
was
about
what
use
cases
exist,
people
speaking
up
saying
like
oh,
we
need
seeking
support
or
oh
we
need
you
know
I
want
peer-to-peer
or
all
this
is
for
real
time.
This
is
for
gaming
and
try
to
trying
to
explore
what
existing
standards
exist
at
ietf
and
and
whatnot,
and
also
what
would
be
a
good
use
case
for
for
quick
and
what
changes
would
be
required
for
quick.
N
But
the
meeting
didn't
really
cover
that
too
much
so
right.
Now
it's
still
in
an
exploratory
phase,
it's
still
trying
to
figure
out
what
working
group
or
groups
could
be
created
to
deal
with
this
if
they're
even
necessary
and
yeah.
It's
it's
it's
it's!
It's
a
really
rough
state
right
now,
just
trying
to
come
up
with
requirements
and
kind
of
come
up
with
what
desired
target
use
cases.
There
are
okay,.
A
Thank
you
for
that,
and
thank
you
for
leaping
in
the
last
moment.
My
understanding
from
spencer's
notes
is
that
a
large
part
of
the
conclusion
is
that
it's
important
to
sort
of
stake
out.
You
know
what
use
cases
people
want
to
pursue,
or
at
least
be
clear
about
what
use
cases
are
in
play.
E
Yeah
either
yeah
so-
and
I
can't
remember
which
lists
my
comments
ended
up
on
you
last
night,
but
I
made
the
observation
that
you
know
this
is
a
pretty
big
ocean
and
there's
lots
of
different
applications
that
something
we
we
you
know
chewed
through
when
we
created
mops
in
the
first
place
right
just
because
there's
so
much
there's
some
stuff,
that's
really
well
developed
with
some
stuff,
that's
sort
of
in
the
midpoint
development
use
case
wise
and
some
stuff,
that's
sort
of
like
well
hey.
This
is
really
cool.
E
Now
that
we
have
this,
we
can
do
these
new
things,
let's
explore
and
it
seemed
like
you
know,
and
I,
by
the
way
I
should
say
I'm
very
supportive
of
the
mock
stuff.
So
I'm
I'm
coming
from
a
friendly
voice
here,
I
think
there's
some
cases
in
particular,
I
I
suggested
things
like
ingest
median
just
because
we
already
have
existing
reliable
protocols
like
srt
that
have
said
that
they're
playing
with
you
know
going
over
quick
and
there's
other.
E
You
know
applications
and
we've
seen
postings
from
parties
that
are
using
quick
experimentally
for
ingest,
but
the
use
cases
for
ingest
are
fairly
well
developed,
so
I
suggested
hey
for
the
very
well-developed
stuff.
Maybe
that's
what
mock
could
build
a
charter
around
and
expand
upon
and
focus
on
and
then
for
other
things,
maybe
it's
time
to
come
back
into
other
groups
like
like
mops
mock
mobs.
E
We
name
these
things
badly,
but
come
back
into
mops
for
discussion,
because
I
think
what
mops
is
very
good
at
is
allowing
things
to
sort
of
evolve
to
a
more
mature
state
and
then
find
the
appropriate
home
to
go
to,
and
that
was
sort
of
my
take
on
on
the
two
things
and
again,
very
supportive
of
of
the
mock
stuff.
So.
N
What
exists
today,
I
can
help
illuminate
so
first
off
my
cat's
on
the
mouse.
Hopefully
it
doesn't
meet
me,
I'm.
Actually
I
work
at
twitch
and
we're
pushing
releasing
a
quick
based
video
delivery
protocol
using
web
transport.
N
I
know
facebook
is
using
quick
for
ingest
already
with
their
player
and
in
this
case
we've
got
two
kind
of
production,
ready,
implementations
and
then
people
have
been
experimenting
with
rtp
over
quick
and
also
srt.
As
you
mentioned,
srt
is
trying
to
proxy
over
quick
as
well
so.
N
Few
different
areas,
there's
a
few
different
implementations
out
there.
It's
really
too
early
to
tell
what
works
well
and
what
maps
will
as
well
too
quick
because
sometimes
it's
it's
like
you
mentioned.
O
You
actually
wanted
to
ask
to
luke,
because
he's
mentioned
his
implementations.
What
he's
found
you
know
what
he's
found
so
far
like
you
know
what
what
is
quick,
giving
him
that
you
know
existing
like,
say,
tcp
or
udp
based
things
wouldn't
do
for
him
in
particular.
Is
he
taking
advantage
of
some
of
the
more
complicated
reliability
and
ordering?
N
Reading
more
and
more
over
time
on
the
the
mock
mailing
list,
I
already
I
did
post
one
thread
about
the
problems
that
we've
seen:
we're
an
hls
based
stack
and
with
some
modifications,
but
it
really
comes
down
to
headline
blocking,
like
you're,
forced
to
download
a
single
segment,
you're
downloading
it
sequentially
and
when
there's
congestion,
what
do
you
do?
N
Quick
gives
you
the
option
to
have
things
in
parallel,
so
kind
of
coordinate
them
together
to
cancel
them
better
and
stuff
like
that,
and
that's
really
what
we're
utilizing
and
we're,
also
in
a
position
that
we
have
our
own
cdn.
So
it's
very
easy
to
just
push
data
over
web
transport,
but
you
could
totally
do
this
with
hdb3
as
well.
It's
just
harder.
A
So
I
think,
I
think
that
we
have
achieved
the
ambition
of
raising
awareness
and
doing
some
cross-pollination
between
the
mock
effort
and
mops.
I
think
there
are
some
suggestions
on
the
table
that
mops
might
be
able
to
help
base
some
of
the
discussions
going
forward
in
terms
of
taking
some
use
cases,
but
I
think
that
we'll
have
to
see
if
that
transpires
on
the
mailing
list,
and
apart
from
that,
I
guess
we
would
like
to
see
how
the
mock
work
continues
going
forward.
A
A
All
right
so
also
keep
your
eyes
peeled
from
for.
Q
B
I'll
assume,
that's
yes,
so
this
is
the
the
mops
two-year
retrospective.
Can
you
believe
it's
been
slightly
over
two
years
since
the
working
group
was
formed,
you
can
almost
smell
the
benzene
of
the
retrospecticus.
B
That's
a
simpson's
reference
if
nobody
gets
it
anyway,
so
I'm
I'm
afraid
that
the
that
the
first
part
of
this
deck
is
going
to
be
a
bit
of
a
downer.
But
you
know
it
gets
better.
I
promise
so
so
here's
how
it
started
basically,
two
years
ago,
well
up
until
two
years
ago
there
were
some
side
meetings,
the
the
ggie
meetings
and
then
the
video
interest
group
side
meetings
that
went
on
for
for
several
years
in
montreal
in
2019
is
when
we
had
the
mops
buff.
B
Where
we
got
you
know,
we
essentially
had
a
discussion
about
about
whether
it
was
reasonable
to
make
a
working
group
and
what
the
scope
of
the
charter
would
be
and
what
not,
and
so.
Finally,
in
november
it
was
formed
just
before
the
singapore
ietf,
and
so
we
had
our
first
in-person
session
at
in
singapore.
B
And
then
in
march
the
the
vancouver
meeting
was
cancelled,
and
so
here's
how
it's
going
now
note
the
the
the
highlighted
virtuals
there,
which
is
kind
of
it's.
So
I
think
it's
been.
It's
been
kind
of
a
downer
for
a
lot
of
working
groups,
but
I
think
it's
especially
been
a
it.
It's
been
a
special
challenge
for
for
mops
because
of
the
nature
of
a
lot
of
the
work
that
we
do
right.
So
I
just
took
a
screenshot
of
the
milestones
quickly.
There's
a
you
know
like
we
we've
this
is.
B
This
was
a
snapshot
that
I
took
earlier
this
week
and
I
know
that
we've
bumped
some
of
these
dates
up
or
back.
I
guess
several
times
as
we
have
kind
of
missed
milestones.
We
have
you
know.
Obviously
we
have
a
few
done,
but
we
still
have
a
lot
of
work
that
we
intended
to
do
at
the
beginning.
That
is
not
done
yet.
B
These
are
the
two
documents
we've
adopted.
It
seems,
you
know
it
seems
a
little
meager,
but
you
know,
on
the
other
hand,
that
you
know
publishing
documents
was
not
the
main
purpose
of
mobs
to
begin
with.
It
never
was
what
is
our
main
purpose?
Well,
if
you
look
at
damn
the
this
is
in
the
wrong
place
in
the
slide
all
right
anyway,
there's
some
we
I'll
come
back
to
the
slide.
B
So
if
you
look
at
our
charter
right,
I've
highlighted
kind
of
the
the
verbs,
the
actions
that
that
mops
was
chartered
to
take
on
solicit,
regular
updates,
solicit
input,
solicit
discussion
and
documentation.
B
B
You
know
figure
out
where
there
are
gaps
in
existing
standards
and
direct
work
to
to
other
working
groups
within
the
itf
and
and
outside
sdos,
and
so
without
those
without
those
hallway
conversations
without
the
you
know,
like
the
you
know,
having
a
bunch
of
people
in
the
hackathon
room
like
working
on
the
same
thing
or
doing
demos
without
a
lot
of
tourists
coming
to
mops,
because
you
know
because
they
just
they're,
they
happen
to
have
a
free
slot
and
they're
like
all
right.
B
Let
me
go,
you
know,
go
into
this
room,
it
looks
interesting
without
the
you
know,
a
lot
of
the
cross-pollination
of
having
everyone
in
the
same
physical
space
talking
about
problems
and
seeing
where
there
are,
where
there's
overlap
and
where
there
are.
You
know,
potential
synergies
between
work
and
seeing
the
the
reduction
in
new
work.
B
Overall
I
mean
that's
been
a
real
drag
from
virtual
meetings
and
I
think
it's
it's
impacted,
mops
more
than
a
lot
of
other
working
groups
which
had
which
had
explicit
work
items
in
progress
and
moving,
but
it
probably
has
also
impacted
working
groups
that
that
have
been
trying
to
find
new
work
as
well.
I'm
sure
it
differs
from
group
to
group,
but
it's
just
this
is
just
has
been
my
observation
based
on
talking
to
other
people,
so
that,
having
been
said,
we've
we've
solicited
a
bunch
of
presentations.
B
We've
had
a
bunch
of
interesting
talks.
Will
law
talked
a
few
meetings
ago?
New
participants
have
come
to
us,
they've
actually
sought
out.
Mops
talked
to
us
and
we've
hooked
them
up
with
people
in
in
the
working
group
and
have
them
come
in
and
give
give
presentations.
B
And,
moreover,
there
have
been
a
lot
of
intangibles
that
mobs
have
provided
right,
we're
responsible
for
a
bunch
of
conversations
happening,
even
though
they're
not
having
it
happening
in
a
physical,
hallway,
they're
happening
on
the
internet
in
virtual
hallways
and
so
like.
I
think
that
a
lot
of
that
things
like
things
like
like
mock
things
like
the
the
renaissance
multicast
work
on
reducing
latency,
are
all
things
that
mops
has
assisted
in
in
helping
get
more
attention
paid
to
these
things
and
assisted
in
bootstrapping
some
of
this
work.
B
And
so
I
think
that's
that's
you
know
that's
a
lot
of
what
we've
done
now
going
back
to
to
the
earlier
slide,
some
other
things
that
we've
done,
that
that
mops
has
done.
In
addition
to
the
you
know,
in
addition
to
kind
of
the
the
the
work
internal
to
the
working
group,
is
providing
external
community
engagement
right,
so
the
smt
smpte
updates
sympty
updates
streaming
video
alliance
updates.
B
We
have
a
paper
that
we
submitted
to
simpty
earlier
this
year
that
we're
going
to
be
presenting
on
in.
I
think
it's
in
december,
but
don't
quote
me
on
that.
So
there's
there's
a
bunch
of
stuff
that
we've
done
and
I'm
you
know
I'm
pleased
with
how
the
with
how
the
working
group
has
been
has
been
going.
B
But
I
mean
there
are
questions
about
how
we
can
increase
the
efficacy
of
the
working
group
like
what
can
we
do
to
make
ourselves
more
effective
right?
I
mean
meeting
in
person
yeah.
That
would
be
great.
That
would,
you
know,
solve
a
lot
of
the
problems
that
you
know
or
or
at
least
eliminate
many
of
the
challenges
that
we
see
today,
but
unfortunately,
that's
not
actionable
we're
either
going
to
have
an
in-person
meeting
or
we're
not,
and
we
don't
have
any
control
over
that.
B
That's
going
to
be
a
you
know,
that's
a
kind
of
a
shared
decision
among
all
the
itf
participants.
Do
we
just
carry
on,
as
is?
Are
things
going
well
enough
that
we
think
that
we're?
You
know
that
we're
on
the
right
track.
You
know
we
should
just
make
incremental
changes
from
here
on
in
do
we
broaden
draft
scope,
or
we
talked
before
about
looking
at
the
augmented
reality
draft
and
saying?
Well,
maybe
this
is
maybe
the
the
use
case
is
much
more
specific
than
the
problems
described.
B
So
maybe
we
ought
to
broaden
the
scope
there.
It's
another
option.
Some
things
that
we
probably
should
do
is
figure
out
how
to
unblock
some
of
the
unmet
milestones.
Right
I
mean
we
have
milestones
that
we
added
to
the
data
tracker
back
at
the
very
beginning
and
they
haven't.
You
know
some
of
them
haven't
been
finished
yet
right.
So
what
can
we
do
to
get
those
moving
along?
B
Should
we
solicit
feedback
from
non-participants?
Are
there
people
who
have?
You
know,
thought
about
participating
in
mops
and
decided
not
to
or
are
there
people
who
don't
know
about
mops,
and
maybe
we
ought
to
figure
out,
maybe
why
they
don't
why
they
don't
know
about
it.
Should
we
coordinate
with
in-progress
standards
activity
going
on
elsewhere
in
in
the
ietf
right,
so
so
we
could
have.
B
We
could
do
something
experimental,
like
designate
point
participants
for
interaction
with
other
with
other
working
groups,
both
within
the
ietf
and
other
standards
organizations
and
get
you
know
either
written
reports
or
or
just
you
know,
presentations
at
meetings
so
that
we
can
kind
of
spread
information
about
these
things
in
a
more
efficient
manner.
B
Having
some
forecasts
about
where
we
think
the
industry
is
going,
so
we
can
foresee
what
what
areas
of
work
might
need
doing
and
then
kind
of
bring
that
information
to
relevant
working
groups.
That's
another!
That's
another
option:
do
we
have
do
we
wanna
have
some
sort
of
vision
or
road
map
for
itf
contributions
to
the
media
ecosystem
right,
so
the
media
ecosystem
is
vast
and
confusing.
B
Ietf
participants
may
only
see
a
small
part
of
it
and
not
understand
how
it
fits
into
sort
of
the
the
larger
media
ecosystem,
and
so
maybe
we
could
develop
some
problem
statements
for
internet
media
operations
to
clarify
some
of
the
challenges
that
we're
seeing
and
maybe
to
use
the
the
challenges
to
then
suggest
directions
for
future
work.
B
And
finally,
you
know:
should
we
solicit
input
from
other
working
groups
facing
similar
challenges
right,
I
I
think
one
of
the
things
I
don't
have
a
good
a
good
beat
on
is
to
what
extent
other
working
groups
are
seeing
engagement
challenges
resulting
from
the
lack
of
in-person
meetings
and
I'd
you
know,
and
and
whether
you
know,
if
some
other
working
groups
have
seen
similar
challenges,
what
have
they
done
over
the
past?
You
know
20
months
to
to
to
help
work
on
those.
B
So
now
I'm
going
to
throw
it
open
to
to
the
floor
for
thoughts,
suggestions
looks
like
glenn
wants
to
make
some
comments.
A
A
We
get
there
before
we
get
there.
I
think.
H
To
request
speaking
so
thank
you
for
inviting
me
before
I
actually
request.
H
Indeed,
so
you
know
the
mops
is
a
kind
of
a
time
bomb
attached
to
it
right.
We
need
to
kind
of
renew
the
working
group
for
me,
it's
okay.
It's
doing.
I
was
really
concerned
the
beginning
of
this
meeting
right
when
I
saw
at
the
top
of
the
hour
only
11
people
attending
the
meeting.
It
was
kind
of
scary
because
mobs
is
basically
to
attract
people
sharing
the
same
common
interests
right
so
now
we
we
peak
at
40.
We
are
33,
so
that's,
okay.
H
I
think
I've
also
heard
a
very
interesting
discussion
in
the
last
two
hours
or
an
hour
and
a
half,
so
I'm
pretty
satisfied.
H
I
know
that
warren,
which
is
kind
of
the
ops
area
director
right
so
I'm
interior
director,
but
not
using
in
synopse,
is
also
happy
with
the
functioning
and
how
mops
works
he's
unable
to
join
today,
because
I
guess
it's
a
dns
office.
At
the
same
time,
it
is,
of
course,
more
important
for
him
than
mobs
the
issue
of
other
working
group
with
these,
not
in
person
meetings
is
everywhere
right.
H
It's
more
critical
here,
like
you,
said
guy,
because
it
was
really
met
mean
as
a
place
to
gather
people
with
the
shared
interest,
but
the
lack
of
work
is
eating
mostly
everywhere,
specifically
the
lack
of
new
work.
So
we
have
a
measure
at
the
isg
about
the
number
of
dash
zero
zero
draft,
and
this
one
is
lower
than
two
years
ago
for
all
the
working
groups.
H
We
don't
do
it
per
area,
so
maybe
some
areas
are
more
impacted
than
others,
but
it's
everywhere,
okay
and
as
leslie
we
talked
also
before
when
god
and
myself
and
and
lastly,
we
were
preparing
this
meeting-
that
you
think
I
had
no
time
to
investigate
further,
but
you
think
that
the
new
zero
zero
only
coming
from
very
interested
parties
right
so
with
a
kind
of
a
lot
of
small
party
trying
to
do
new
work.
H
So
for
me,
I'm
pretty
fine.
We
already
discussed
the
mops
at
the
previous
iesg
meeting
and
it
was
continuing
right.
So
everyone
says
okay
go
on,
and
I
remember
what
I
said
last
time.
Iot
ops
is
very
similar
to
mobs
right
and
more
and
more
when
you
do
both
for
creating
new
work,
including
sometimes
say:
oh,
is
it
ala
right?
Is
it
a
mobs
like
so
mobs
now
is
a
an
adjective
somehow
right
in
the
isg
slang.
So
for
me?
Yes,
it
is
it's.
Okay,.
H
It's
a
good
job
and
good
job
summary
kind
by
the
way.
A
Thanks
eric
and
just
to
close
the
loop,
my
understanding
is
that
the
we
were
reviewed
after
two
years
and
we
are
not
dead
yet
so.
H
Yeah
two
years:
it's
maybe
it's
basically,
you
know
right.
There
is
a
debrief
of
the
isg
and
the
iab
next
week.
I
will
talk
about
mobs
there
and
say.
C
B
E
I
was
wrong
in
that.
I
don't
want
to
correct
anything.
He
said
I
thought
you
did
it
really
well
thanks
kyle
for
running
through
that
I
I
just
want
to
add.
You
know
I've
been
a
long-time
proponent
for
mops.
You
know
going
way
back
and-
and
you
know,
I
think,
we've
done
a
lot
of
good
stuff.
You
know
the
core
problem.
That
was
a
lot
of
my
motivation
in
the
past
is
that
we
have
the
number
one
use
the
internet
by
you
know,
data
and
scaling
and
deriving
to
add
those
bigger
pipes.
E
Is
video
and
we've
always
had
a
sort
of
a
historic,
slight
disconnect
with
the
representation
of
voices
trying
to
do
video
stuff.
When
we
talked
about
things
at
the
itf
and
being
a
video
guy,
that
was
always
a
thing.
Well,
we
need
to
fix
this
and
you
know
the
reason
why
other
groups
of
the
itf,
you
know
are
very
active.
I
think
you
produce
a
lot
of
zero
zero
drafts,
as
eric
points
out
is
that
we
have
a
real
established
community.
E
E
You
know
we
have
a
number
of
parties
from
different
organizations
that
all
know
each
other
have
known
each
other
for
decades
that
come
together
and
work
on
problems,
and
I
think
that
we're
sort
of
the
early
days
of
reaching
that
gap
between
the
itf
and
these
video
guys
out
there,
which
are
producing
all
these
documents.
E
But
I'm
seeing
a
very
positive,
you
know
developments.
You
know,
I'm
seeing
that
we
have.
You
know
when
places
like
cd
and
I
needed
working
group
chairs.
You
know
sanjay
was
around
to
step
up
and
become
a
co-chair
of
cdni
which
connects
to,
like
you,
know
the
real
world
and
the
real
delivery
folks
versus
just
people,
developing
technology.
E
You
know,
and
we
have
other
things
that
are
being
discussed
and
thrown
around,
that
is
creating
a
welcoming
home
for
the
people
from
that
video
world
that
are
doing
all
these
video
things
on
the
internet
and
driving
all
the
traffic
and
all
the
scalability
to
come
into
the
itf
and
have
conversations,
and
I
think
that
mobs
has
shown
through
the
parties
that
are
coming
here
and
I
believe
the
parties
that
will
still
come
that
you
know
when
people
ask
the
question
you
know
I
come
from
a
video
world,
I
I
need
to
engage
with
the
itf.
E
I
believe
I
want
to
engage
back
to
you.
How
do
I
do
that?
But
that
now
that
box
exists?
It's
the
place
to
say
well,
you
can
talk
to
the
box
guys.
First,
they
may
not
be
the
right
final
destination,
but
they're
the
right
first
destination,
and
they
can,
you
know,
help
you
bring
in.
They
can
be
a
friendly
voice,
a
friendly,
a
friend,
face
and
help
things
mature
to
a
point
where
they're
ready
then
connect
with
other
itf
working
groups
and
build
that
bigger
community.
E
So
it's
early
days,
but
I
see
that
community
developing
nicely
and
and
I'm
very
helpful
hopeful
that
we're
going
to
see
you
know
this,
continue
on
and
and
and
the
relationships
grow,
and
so
that
you
know
10
years
now.
We're
gonna
have
this
whole
room.
Full
of
you
know
itf
video
people
that
have
known
each
other
for
10
years
and
when
there's
a
new
problem,
a
new
thing
to
work
on
this
is
where
they
come
to
do
it.
So
that's
just
my
two
cents.
I
want
to
throw
in.
B
All
right
thanks,
glenn
sanjay.
F
Yeah
yeah,
I
I
just
wanted
to
echo
what
glenn
just
said
so
well,
I
think
there's
a
lot
of
issues
that
are
not
necessarily
well
defined.
In
terms
of
that
you
know
I
know
I
know
I
need
some
work
on
protocols,
so
I
I
itf
working
groups
are
the
right
place.
There
are
many
cases
operationally
where
mob
seems
like
a
good.
You
know
first
place
to
start
and
bring
the
question
that
that
may
not
be.
F
You
know
wanting
to
change
the
protocol,
but
maybe
you
know
the
making
sure
that
you
understand
it
and
you
implement
it
right.
So
some
of
those
types
of
areas
really
fit
well
within.
You
know
what
this
working
group
can
offer.
So
I
think
there's
a
great
deal
of
value
and
obviously
the
as
kyle
was
saying
that
you
know
we
don't
have
the
luxury
right
now
of
meeting
in
person.
So
a
lot
of
those
hallway
conversations
that
may
happen
and
that
may
lead
to
people
that
are
attending
a
specific.
F
You
know
working
group,
but
they
may
show
up
here
because
oh
yeah,
that's
really
an
operational
issue
that
at
least
we
should
bring
it
back
into
the
mob's
working
group.
So
we
we
kind
of
don't
have
that
right
now,
but
regardless
of
that,
I
think
this
is
a
good
venue
where
I
know
sva
can
benefit
a
lot.
You
know
streaming
video
alliance.
F
B
Yeah
yeah,
I
agree
and
I
I
think
it's
I
think
it's
clear
that
it's
it's
been
useful
to
a
lot
of
folks.
So
I'm
I'm
optimistic
and
I'm
definitely
supportive
of
continuing
the
working
group.
A
From
the
chair
standpoint,
I
think
the
only
thing
that's
really
troubling
me
in
all
of
this
is
how
quiet
the
mailing
list
is,
and,
as
I
alluded
to
earlier,
the
sort
of
bifurcation
between
the
people
who
come
to
meetings
and
the
discussion
that
doesn't
happen
on
the
mailing
list,
and
I
don't
entirely
know
how
to
address
that.
A
I
mean,
I
think
the
reality
is
a
lot
of
mailing
lists,
have
gotten
quiet
and
just
because
mail
is
so
last
century
and
it
isn't
helped
by
the
fact
that
we
don't
have
the
sort
of
vibrant
people
know
each
other
community
thing
going
on.
But
I'm
I'm
saying
it
here
in
my
out
loud
voice
just
because
I
think
it
is
an
area
that
we're
going
to
have
to
find
answers
to.
A
B
N
N
Say
from
my
personal
experience,
I
haven't
known
when
to
interact
with
this
working
group
like
even
just
a
little
blurb
about
media
over
quick
earlier,
I
I
wasn't
sure
if
it's
appropriate
to
like
say
what
I've
been
working
on
in
this
working
group
or
if
I
should
just
go
straight
to
like
a
more
specific
one
like
media
over
quick.
So
it
kind
of
means,
at
least
in
my
mind.
I
discussions
are
either
scoped
correctly
already
or
end
up
in,
like
the
catch-all
and
mops.
A
Right
well-
and
I
think
it's
it's
a
good
question
and
and
maybe
partly
because,
even
if
you
review
the
history
of
discussions
on
the
on
the
mailing
list,
you're
not
going
to
see
a
lot
of
exemplars,
but
I
mean
I
would
say
at
this
point:
the
mailing
list
is
going
to
be
what
we
make
it
and
speaking
of
this
one
one
chair,
I
would
rather
see
more
discussion
there
more
ideas
tossed
out
there
than
fewer.
So
certainly,
please
don't
feel
shy.
A
Yeah,
I
think
that's
a
particular
area
where
you
know
the
ietf
is
trying
to
figure
out
what
it's
doing
so
I
wouldn't
want
to
get
in
the
way
of
an
organized
activity
of
media
over
quick,
but
if
there
are
things
that
we
can
do
to
help
in
discussion,
we'd
be
happy
to
be
see.
It
happen
on
the
mailing
list.
B
All
right,
oh
jake,.
C
One
last
not
not
complaint,
I
think
you
guys
are
doing
great,
but
regarding
the
mailing
list
being
last
century,
it
made
me
wonder
there
is
a
ietf
slack
room.
Should
we
I'm
not
sure
it'll
be
any
better,
but
and
it's
I'm
in
a
couple
of
rooms
there.
I
think
it's
usually
pretty
quiet,
but
I'm
not
sure
if
if
anybody
would
prefer
that-
and
I
thought
I'd
just
ask
instead
of
leaving
it
in
chat-
or
I
think
you
got
missed.
A
Yeah,
that's
that's
a
good
question
and
I
would
suggest,
if
you
don't
mind,
dropping
it
in
the
notes.
That
would
be
great
too.
I
think
that
we
may
want
to
do
that.
We're
still
going
to
wind
up
with
the
same
question
we
had
about
the
working
group
last
call,
which
is,
if
it
doesn't
happen
on
the
mailing
list.
It
didn't
happen
so
it
maybe.
We
need
a
slack
channel
and
more
comments
on
the
mailing
list.
E
One
observation,
I
wonder,
if
part
of
you
know
part
of
the
the
work
we're
doing
here,
is
to
bridge
in
culturally
these
folks
that
aren't
traditionally
itf
people
but
coming
from
the
video
world,
and
I'm
wondering
if
we
need
to
help
them
adapt
to
what's
appropriate
on
the
itf
family
lists.
E
You
know,
if
I'm
new
to
an
area,
I
I
probably
want
to
adapt
whatever
the
local
culture
is
and
since
we're
fairly
low
volume
and
fairly
new,
we
don't
have
necessarily
like
a
demonstratable
culture
that
they
can
plug
right
into
and
I'm
not
sure
how
to
solve
it.
Other
than
say
you
know
we're
very
welcoming
and
please
post.
Maybe
those
of
us
who
are
here
already
just
need
to
take
it
upon
ourselves
to
be
a
little
more
chatty
on
the
list
to
demonstrate
sort
of
a
culture
of
openness
and
engagement.
A
A
That's
a
good
suggestion
thanks
all
right.
Well,
unless
anybody
has
any
other
other
business,
I
think
we
are
at.
We
are
at
time
and
we
are
done
kyle
anything
else.