►
From YouTube: IETF93-PAYLOAD-20150720-1740
Description
PAYLOAD meeting session at IETF93
2015/07/20 1740
A
B
A
A
A
C
C
E
A
F
F
G
H
It's
no
joking
good
afternoon.
This
is
the
payload
working
group.
We
have
quite
a
busy
schedule
for
some
reason,
so
we
try
to
keep
the
time.
So
this
is
the
note
well
and
you've
seen
that
before
today,
I
hope
without
pits
at
your
first
session
for
the
day.
But
you
should
be
aware
that
you
have
to
be
clear
on
any
IP.
Are
you
are
aware
of
the
agenda?
For
today
we
have
the
RTP
payload
phone
I
format
for
the
vp9
video.
It's
not
just
just
who's
who's
going
to
present
it.
H
Okay
thanks
and
we
have
the
RTP
payload
format
for
the
no
need
to
leave
an
interleaved
parity,
the
Meo
PE
codec
and
for
interleaf
packets
are
the
document
status.
We
have
0
posts
are
TP
payload
published
as
RFC
75
87.
We
have
the
RTP
how
to
at
the
editor
q
we
have
g
7,
11
0
and
a
0
to
60
in
the
isg
processing.
We
have
a
265.
H
Also
in
the
with
SBC
is
g
for
expert
review
and
the
vp8
is
waiting
for
the
ad
others,
the
SDN,
silly
and
silly
read
data.
Is
there
we're
not
presenting
it,
but
we
probably
will
have
to
look
and
see
how
to
progress
that
one
so
there
there
was
an
open
issue
in
the
in
the
SDP
directorate
review
of
the
age
of
265
payload,
and
it
came
from
fleming
and
it's
the
the
issue
is
that
the
RTP
payload
format
in
hyssop
said
that
the
RTP
payload
format
should
include
an
overall
grammar
for
the
SDP
code
point.
H
These
are
the
parameters
that
go
to
the
FN
TP
line.
Currently
we
don't
do
any
formal
grammar
on
them
when
we
are
describing
them
in
the
federal
specification.
They
just
go
into
the
ftp
line
and
you
have
to
understand
the
parameters.
So
the
question
is:
should
we
require
have
a
full
grammar
for
the
f
MVP
parameters.
H
I
Last
one
do
the
recommendation
I
think
in
the
are
to
be
how
to
is
really
that
if
you
have
complex
parameters
that,
where
is
a
man
I
mean
the
actual
syntax
of
the
of
the
parameters
and
values
might
be
not
easily
deductible?
That's
where
you
need.
You
say
BNF,
but
otherwise,
if
you
have
like
parameter
simple
man,
you
like
a
bigot
or
a
string
or
whatever
that's
usually
find
you
that's
described
in
exton,
is
so
ugly,
hey
I!
J
So
Stephanie
not
just
to
remind
those
who
don't
have
the
hundred
pages
of
the
draft
by
heart.
Those
one
complex,
parameter,
sub
parameter
in
this
thing,
where
there
is
a
grammar
actually
included
in
this
very
draft,
but
it's
really
only
I
think
what
Fleming's
after
is
for
the
suburb's
all
other
payload
formats,
a
grammar
that
describes
all
possible
parameters
and
all
possible
values.
Bnf
for
that.
E
Perhaps
you,
sir,
are
defined
so
I
mean
for
sdp
itself.
There
should
be
an
overall
grammar
for
the
fmtb
attribute
I'm,
assuming
that
4566.
This
has
that,
if
not
it
should
the
parameters
themselves.
We
I
mean
recall
that
there
are
a
media
type
definition,
so
they
have
to
fit
the
definition
of
the
media
type
parameters
which
will
have
a
allowed
a
bamf
so
and
yeah
I
agree
that
we
should
have.
If
there
are
complex
parameters,
we
should
definitely
should
have
an
APB
NF
for
the
parameter.
E
E
So
we
can't
do
that,
because
this
is
a
main
line.
Registration.
We
can
give
grammar
for
individual
parameters
and
then
have
they
then
fit
the
usual
way
of
putting
the
mind
type
into
the
FM
TP
line.
But
we
shouldn't
be
giving
a
grammar
for
the
FM
speed
line
with
this
parameter
in
it.
Okay,
if
that
makes
sense.
E
D
Drive
I
just
want
to
say
that
you
know
if
the
RTP
how
to
says
what
Magnus
says
and
I
have
no
reason
not
to
believe
him.
Then
I'd
say
that
that's
the
thing
that
sort
of
has
the
idea,
if
it
senses
on
how
to
do
these
things,
and
you
know
it
sounds
like
from
what
Stefan
says
this:
that's
what
this
graph
is
already
doing,
and
we
should
do
what
the
RTP.
How
do
you
recommend
and
saying
you
want
something
more
than
that
you
know.
D
E
K
H
L
J
Then,
thank
you
very
much
for
the
statement.
What
a
move!
The
work
we
office
and
I
have
made
that
very
clear
to
climbing
and
really
but
what
I?
What
I
personally
worry
about
is
we
have
already
left
on.
They
burn
a
plate
in
these
rafts
right.
You
have,
let's
not
put
in
yet
another
one
that
is
just
considered
stuff
to
ignore.
H
Okay,
so
I
understand,
what's
that
and
maybe
I
can
talk
afterwards,
because
we
don't
have
the
time
to
discuss
it
now
in
this
session.
So
the
current
milestones
we
they
may
need
some
update
and
the
EVP
are
I.
Think
I
lost
any
reason
to
do
that.
Probably
it's
not
going
to
be
used
in
RTP,
so
we
probably
need
to
go
back
and
see
if
we
can
take
it
out.
H
H
M
Okay,
so
as
most,
you
can
see
I'm,
not
just
in
a
Magnus
problem
from
colleague
of
Justice,
see
where
I
can
see
the
slides.
My
first
time
at
IDF
have
to
be
here.
Dd
a
draft
was
adopted
at
the
last
I
def
meetings.
I'm
not
going
going
to
debates
an
exact
format
because
it
hasn't
changed
since
then.
I
will
not
like
to
talk
about
the
two
different
modes
we
have
in
the
paper
format
and
just
could
note
that
transport.
M
So
that's
what
about
next
like
this,
if
you
don't
use
spatial
or
temporal
age
for
vp9,
you
just
send
the
first
octet
of
the
pedal
header.
Nothing
to
do.
If
you
are
using
some
can
scalability,
you
can
either
send
flexible
mode
or
non
flexible
mode,
and
there
is
a
reason
for
me
to
have
these
two
modes
in
the
flexible
mode.
M
You
don't
specify
the
scalability
structure
in
advance,
but
of
course,
then
you
have
to
pay
with
specifying
what
each
frame
refer
to
so
there's
a
slight
increase
of
over
header.
We
also
have
a
known
flexible
mode
where
you
in
advance
for
another
keyframe.
You
send
a
scalability
structure
very
describe
that
this
is
the
pattern
I'm
going
to
send
with
for
the
coming
period
and
adjusts
for
each
frame
transmit
a
indexing
that
this
is
the
frame
number
excluding
these
group
of
frames
so
that
our
that's
pretty
much
the
difference
between
those
two
one.
M
I
want
to
give
you
two
quick
examples
on
when
to
use
these
tools
on
next
I
place.
The
North
flexible
load
is
what
you
normally
use
for
regular
video
there's.
Nothing
really
strange
Larry's,
send
a
scalability
structure
in
advance,
whatever
key
frame
as
I
said,
and
then
you
have
like
an
index
increasing
for
for
each
frame,
saying
that
this
is
very
number
three
in
this
group
of
frames
and
you
don't
have
to
signal
much
for
each
frame,
so
pretty
simple.
It
also
makes
it
simple
for
a
four
box
in
mail
to
thin
out
stream.
M
O
M
Mainly
when
do
we
screen
share
in
this
case,
we
don't
really
know
what
a
pattern
will
look
like
in
advance.
So
and
in
this
specific
example
I
have
here,
we
we
sent
as
much
as
we
can
in
in
layer
syrup
when
we
don't
have
enough
bits.
I
like
the
first
two
frames.
If
a
static
screen,
you
don't
have
to
send
that
mandates
is
quite
easy
to
encode
and
there's
just
overhead
to
send
multiple
layers
about
point.
I'm
so
just
some
one
layer.
M
M
So
this
is
why
we
have
this
flexible
mode
and
of
course
we
pay
with
in
this
case,
I
don't
have
a
bit
here,
but
we
for
the
gold
first
golden
color
frame
there.
We
actually
tell
that
we
are
referring
to
the
last
blue
one,
so
we
explicitly
for
the
frame
safety
system
frame
I
refer
to
and
we
can
do
it
in
for
up
to
three
frames
refer
to
three
different
way
to
the
same
time.
M
So
that's
it
for
those
two
modes
last
slide
is
the
transport
we've
discussed
this
a
lot
with
Jonathan
and
again,
and
we
are
currently
working
on
doing
this
similar
to
peace
tree
and
see.
You
need
a
transport,
a
couple
of
drawbacks
with,
as
of
course,
instead
of
having
multiple
RTP
streams,
but
the
parsing
is
simple
enough
that
we
thinks
that
this
is
the
way
to
go
anyway.
It's
easy
enough
to
do
parsing
it
anyway.
G
J
Mean
there
comes
off
of
both
application,
lower
standards
that
have
some
restrictions
like
that,
for
example,
and
pictorial,
contrast
to
what
some
people
believe
has
not
only
that
simply
I
know
bbp
BVI
protection
structures
on,
but
you
can
do
a
lot
of
other
things
there.
However,
when
you
go
to
some
deep
down,
cable.
J
Mbn
tell
you,
you
restrict
the
bitstream
to
something
like
you're
in
flexible
mode,
whereas
the
default
is,
you
can
do
whatever
you
want,
there's
something
very
similar
in
h.264.
So
so
the
concept
exists.
I
think
the
the
difference
here
is
that
it
is
specifically
called
out
as
such
right.
That's
a
good
thing.
I
agree.
Yeah.
G
Thanks
mogan
should
just
apply
I
like
that.
It's
a
RTP,
payload
hitter,
that's
the
part
they're
like
it
its
dynamic.
You
can
change
it
anytime
on
the
fly
and
doesn't
depend
on
any
signaling,
and
this
is
depend
on
any
codec
specific,
well,
pillow,
headers
or
I.
Guess
our
inheritance.
But
this
particular
pole
header
could
be
used
really
with
any
any
codec
right.
This
part
of
the.
J
I
apologize
but
I
think
I
personally
think
architectural
II
stuff,
like
that
belongs
in
to
the
codec
bitstream,
are
not
into
the
RTP
layer.
I
know
that
in
vp9
this
is
kind
of
not
bugged,
with
the
way
it
is
normally
dealt
with
it
with
our
critics
yeah,
because
you
guys
basically
stopped
at
what's
called
it.
The
slice
layer,
I
understand
right
and
everything
below
there
is
its.
You
call
that
RTP
payload
format,
so
there's
a
lot
more
functionality
living
in
the
RTP
payload
format
here
by
and
it
is
in
the
more
traditional
standards.
J
However,
keep
in
mind
when
making
this
type
of
architecture
decisions
that
they
such
a
thing
is
recording
and
playback,
and
you
basically
have
to
keep
this
information
somewhere.
So
you
have
to
record
an
RTP
stream
rather
than
and
then
and
then
the
what's
called
an
elementary
good
stream
and
in
the
older
standards
and
that's
you
know,
there's
a
reason
why
why
things
were
cut
up
the
way
they
were
cut
up
in
the
past
and
I'd
I
would
suggest
to
think
about
that
too,
and
not
only
think
about
real-time
transport.
Thanks.
D
So
John
client,
I
mean,
I
guess
the
reason
why
we're
mentioning
these
two
points
here.
So
if
you
feel
like
I,
think
these
are
the
two
biggest
architectural
cisions
were
making
in
this
palette
form
and
I'm
a
co-author
by
the
way
free
Pluto
pay
attention
such
things,
we
wanted
to
get
sort
of
get
at
least
a
not
a
formal
consensus.
D
Maybe
but
at
least
a
you
know,
a
lack
of
people
jumping
up
and
down
and
screaming
so
that
you
know
you
know
design
the
whole
thing
and
then,
when
we
get
to
work
last
call
we
get
people
saying
no.
This
is
terrible.
I
need
this
or
why
are
you
or
why
do
you
have
two
ways
of
doing
the
same
thing
or
whatever?
Just
so
that
we
can
so
be
reasonably
confident
that
we're
going
down
the
right
path
now
and
not
have
to
rewrite
everything?
And
so.
H
D
D
K
H
D
P
Mean
that
that
part
of
the
good
stream
is
but
but
how
its
packaged
informed
into
an
RTP
packet
is
solely
not
fixing
more
than
this
so
doing
the
type
of
traditional
I
mean
like,
so
I'm
not
very
fun,
not
arguing
against
thats
already
architectural
thing.
I'm
certainly
not
moved
by
the
argument.
We
couldn't
possibly
change
the
part
that
we
need
to
change
wiese
most
certainly
could
change.
Yeah
are
my
point.
Is
that
we're.
D
Which
is
being
defined
in
this
specification,
so
we
are
we're
not
saying
that
what
we're
defining
here
is
fixed,
it's
not
we're.
Defining
it.
We're
saying
is
the
choice
that
it's
in
the
payload
spec
and
not
Ian.
You
know
whatever
the
core
ep.9
bit
stream
has
been
decided.
We
have
to
do
it.
Well,
it's
okay!
I
guess.
F
D
Philosophical
which
group
is
I
mean
if
you
wanted
to
have
two
different
documents:
one
sort
of
a
ep.9
container
format
and
you
know,
and
the
other
there
are
to
be
a
load
and
take
put
half
of
this
document.
One
and
half
on
the
other
I
mean,
and
so
that
one
could
be
cited
by
on
Boris
or
whatever
I
mean
whatever
I
don't
find
that
pic
week,
I'm
telling
myself,
but
if
you
say
that
we
absolutely
need
that
I,
wouldn't
you
know,
rejected
out
of
hand,
but
I,
don't
personally
see
a
need
for
it.
D
K
D
Q
C
Q
Max
it's
fine
to
make
these
choices,
but
it's
there
are
certain
diagnostic
requirements
that
you
need
to
make
it
so
things
like
the
Russell's
and
different
layers,
stuff
like
that,
and
if
you
can
provide
all
of
that
and
not
do
more
transport
goo.
That's
fine,
but
just
recognize
that
that
some
laws
provide
in
the
current
artistic
Informer
provide
more
diagnostic
information.
H
Some
ugly,
flexible
fact,
I.
R
The
two
things
they
want
were
that
the
fact
that
we
would
remove
the
non
interleaved
in
the
interleaved
parts
from
the
SDP
so
that
it
would
become
totally
flexible.
He
would
not
have
to
indicate
that
in
stp
up
front
the
other
thing
was
to
make
the
if
you
were
using
two
dimensional
matrices
for
for
FEC,
and
you
did
it
so
originally
we
had
lnd
in
the
stp.
We
made
that
optional,
so
that,
if
you
are
using
a
flexible
mode,
you
don't
have
to
like.
R
Wyd
lmm,
and
we
thought
we
had
some
agreement
on
that-
was
that
we
needed
it,
identifier,
which
was
either
RS,
ID
or
SSRC
a
quick
recap
of
what
it
was.
What
it
meant
is
next
slide.
So
the
problem
is
that
you
have
these
effects
at
the
bottom,
and
you
have
maybe
RTP
streams
on
top,
and
you
wanted
to
know
VIP
to
which
the
twitch
stream
does
this
fec
apply
to.
So
there
are
two
ways
of
doing
it.
One
is
next
slide.
R
You
use
the
first
side,
okay,
that
was
SSRC
on
the
first
flight
and
the
second
slide
was
used
RS
ID.
So
we
didn't
make
much
progress
since
last
time
on
this,
but
I
think
we'll
look
at
the
payload
or
the
RTP
header
structures
to
figure
out
if,
if
either
is
sufficient,
because
I
think
SSRC
or
RS
ID
either
would
work
depending
on
the
use
case.
Q
Q
G
Yeah,
so
part
of
the
philosophy
when
we
were
coming
up
with
this
was
to
explicitly
not
have
anything
in
signaling
to
allow
flexibility
at
the
sender
to
be
able
to
change
it
dynamically.
So
I
think
what
you're
asking
for
is
really
a
sender
lock
in
that's
under
can
only
do
something
we
will
try
to
avoid
that
kind
of
lock
in
by
signaling.
If
group
feels
we
need
lock
in
or
restrictions,
we
could
explore
that,
but
I
think
that
philosophy
of
the
draft
so
far
has
been
to
avoid
those
kinds
of
restrictions
that
well
it.
G
The
philosophy
was
that
the
receivers
must
be
fully
flexible.
The
receivers
must
be
able
to
support
any
kind
of
effect
flows
anytime,
and
so
the
sender
has
the
discretion
to
apply
effect
whenever
it
wants
and
wherever
suppressor,
whoever
doesn't
want.
So
the
the
sender
never
has
to
actually
expressly
signal
that
it's
going
to
do
this,
it's
only
in
the
cases
where
you
think
that
that
you
want
to
have
a
constrained
receiver
that
you
want
a
receiver
that
cannot
handle
these
dynamic.
D
So
don't
think
so
I
mean
I,
think
I
guess
the
one
question
the
here,
which
is
not
probably
not
for
this
group
directly,
but
this
is
great
for
stp
used
for
offer
answer
it's
not
great
for
stp
used
in
face
up.
So
we
need
some
other
API
invasive
to
say
what
you
actually.
D
N
Justin
Brody
I
mean
we
already
have
this
problem
right
now
with
opus
FEC,
where
you
may
say:
I
want
to
get
open
cell
bc,
but
exactly
what
you
get
is
not
that
the
discretion
of
the
center
and
I.
Could
you
want
to
have
more
control
surface
then
we
need
something
else,
but
for
right
now
we
have
something
that
works
sensibly.
R
Yes,
sir,
the
big
mask
would
solve
that
problem
and,
let's
move
on
to
the
next
slide,
so
version
00.
I
had
this
so
currently.
This
is
in
the
draft
next
slide.
Please!
So
what
we
do
is
we
stop
out
the
SN
base
in
the
lens,
so
the
length
comes
to
the
top
and
the
SN
base
goes
below
the
timestamp
recovery,
so
that
is
hopefully
non-controversial
next
slide.
Then
we
add
in
the
SSRC
at
the
top,
so
that
you
know
what
it
relates
to
now.
R
D
That
I
see
I
mean
I
feel
especially
given
some
of
those
stuff
we're
talking
about
in
or
what
we
talked
about
in
perk,
putting
an
SSR
see
into
an
RTP
payload.
It
makes
me
scared
because
that's
end-to-end
pert,
so
you
know
that's
cuz,
that
looking
for
in
most
of
the
perc
architectures
I,
think
I've
seen
that
will
be
end-to-end
protected,
not
up.
I
have
protected
so
I
feel
like
putting
an
SRC
into
a
payload,
is
a
layer
violation.
G
D
So
job
thank
ya
I'm
so
it
sounds
like
yours
going
need
to
be
some
figuring
out
what
once
per
figures
out,
what
it's
doing,
figure
out,
how
to
do
effective
perk,
which
might
mean
feck
over
the
encrypted
payloads.
But
I'm
not
sure
you
can
solve
that
here,
but
it'll
need
to
be
solved
eventually
and
probably
relatively
soon.
R
A
G
R
Doing
xr's
and
through
the
mathematical
properties
to
loads,
ok,
then
next
site.
So
there
was
an
open
issue
that
some
people
indicated
that
there
is
this
comedy
stream
protection
and
we
thought
about
it
since
you're
already
putting
in
SSRC
already
putting
in
SF
base
there.
So
so.
This
is
the
reason
why
we
stopped
the
SN
base
of
the
length
next
slide.
So
what
we
did
was
we
steal
a
bit
of
a
few
bits
from
the
length
putting
in
SC,
which
is
SSRC
counters,
so
you
could
chain
together
the
the
SSR
season.
R
You
could
list
all
the
SSR
season,
the
SN
basis
that
you
want
to
protect.
So
if
you
add
mac
16
of
them-
and
what
would
happen
with
is
that
you
would
have
SC
as
a
counter
more
than
0
would
have
the
first
SSRC
then
followed
by
at
the
nexus,
RC
and
so
on
so
forth
and
and
then
you
would
have
the
payload
bits
after
that.
I
My
name
is
Ruslan
I'm
cutting
down
the
length
field
makes
you
not
be
able
to
support
in
the
IP
packet
sizes.
It's
I
guess
it's
4k.
Now,
even
if
the
most
used
packet
sizes
are
one
a
half
K
kind
of
leaf
nets,
we
do
have
larger,
including
like
9k
ethernet
frame,
something
like
that
bigger
I
like
you
were
I,
really
wonder
if
this
is
wise.
Okay,.
R
J
J
R
P
So
countries
uncompressed
broadcast
video
in
studios
is
often
sent
with
jump.
Jumbo
grams
I,
don't
know
if
that
would
be
in
this
case.
I
don't
know
enough
about
it,
but
you
might
ask
somebody
that
does
ask
not
with
faq
probably
I
mean
who
knows
how
it's
done
so
yeah,
ok,.
P
D
H
R
H
H
R
P
D
D
D
We
this
association
problem
is
this
is
not.
It
needs
to
be
solved
for
r-tx
as
well.
So
if
you
see
a
nobody
same
thing
r-tx,
so
that
problem
has
to
be
solved
with
somewhere
outside
the
payload
format
for
our
TX
anyway,
so
I
don't
see
a
benefit
to
solving
it
inside
the
payload
format
for
FEC.
Obviously,
our
TX
is
only
protecting.
Us
can
only
protect
a
single
stream
so
so
I'm
into
the
multi
extreme
cases.
R
A
D
So
having
listing
multiple
assets
are
seized,
isn't
sufficient
right
once
you're
protecting
multiple
services,
you
have
to
have
some
night
way
of
knowing
which
stream
any
recovered.
It
means
you
have
to
add.
Yes,
our
seas
to
the
protection,
don't
you
so
that
the
recovered
babik
know
which
stream
it's
for
ya.
R
R
G
G
R
G
D
D
H
E
H
K
D
R
D
Know
we
need
at
least
version
two
of
ya
would
come.
My
argument
before
was
that
I?
Don't
think
that
too,
without
three
is
useful.
I
understand
the
use
case
for
three
I
think
to
whatever
the
solution
we
do
for
our
TX
in
all
these
cases
will
also
apply
to
you
know:
FPC,
protecting
a
single
SSRC,
so
I,
don't
think
two
is
useful.
Three
is
useful,
potentially
though
it
might
be
art
effectively
watching,
but
two
I,
don't
think
is
one
of
them,
so
no
SSRC
as
the
SSRC.
D
E
E
A
R
O
E
O
D
Gentle
medics
I
mean
the
issue.
Is
that
fec
stream
one
is
protecting?
You
know,
I
mean
in
the
extreme
case,
fec
stream.
One
is
protecting
streams,
one
through
five,
every
scene,
2
is
broken
stream.
Six
through
ten
need
to
know
which
FEC
is
with
streams
or
chaos
erupts
and
so
II,
so,
and
so,
unless
you
so,
you
have
to
have
some
way
of
describing
that
association
and
either
end
I.
H
H
R
R
H
H
D
H
This
is
the
MLP
codec.
We
close.
C
D
H
H
H
H
The
issue
that
we
had
they
were
to:
how
do
they
did
only
that?
What
what
was
needed
to
make
the
document
look
correctly.
The
open
issue
that
was
discussed
in
the
previous
meeting
has
to
do.
Is
they
they
asked
for
having
parents
off
type
names
based
on
what's
being
used,
which
is
not
only
melt
but
also
map.
H
11
12
12
hundred
and
twenty
four
hundred,
and
there
was
some
people
saying
why
do
we
meet?
Why
shouldn't
they
go
back
and
do
just
one
and
do
this
as
a
as
a
parameters
of
course
that
the
the
betray
the
problem
is:
that's
already
being
used
with
those
subtype
names,
so
I
don't
see
any
reason
why
we
should
not
let
them
registered
all
those
names.
I
mean
it's
not
a
big
issue
for
us
to
stay
on
the
registration
and
be
done
with
them.
So
I
think
that
they
fix
the
document
based
on
the
corner
time.
H
I
reviewed
it
and
pointed
them
to
RTP.
How
to
and
I
want
to
do
is
that
I
think
they
reviewed
it
and,
in
my
view,
I
think
it's
it
can
be.
We
can
just
get
it
and
start
working
as
a
working
document,
and
ever
they
might
go
for
that,
so
I,
don't
think
didn't.
If
anyone
read
this
draft
except
for
me,
okay
yeah,
you
have
any
okay
sure.
J
H
So
I
want
to
take
it
home
to
you,
because
I
mean
if
you'll
say
that
I,
don't
think
people
who
didn't
create
it
should
say:
I
will
just
ask
them
the
least
about
it,
and
then
we
leave
a
milestone
and
we'll
ask
them
to
do
this
or
what
we
document.
Okay,
it's
just
a
pail
of
all
right,
I,
voice,
audio
failure,
forma,
okay,
good!
So
we're
done
with
this
one
with
short
one,
and
now
we
will
go
to
our
last
presentation.
H
K
Okay,
good
afternoon
alone,
this
is
about
home,
rd
pero
format
going
to
leave
application
space.
So
why
we
need?
Oh
sorry,
why
we
need
and
I
interleave
the
parent
element.
K
Q
K
K
K
K
This
kind
of
ganja
leaving
the
giveaway
help
useful
for
arm
or
for
some
applications,
which.
K
Packets
are
together
for
too
just
fate.
Another
one
is
wicked,
has
gathered
RTP
packets
in
the
living
where
each
are
giving
packets
is
evaluating
to
several
pots
and.
K
K
Others
we
consider
as
may
be
athletic
our
dps
soleta
mentally
being
could
hear
before
to
reduce
this
kind
of
delay,
which
will
be
introducing
the
later
in
the
later
later
slide.
Next,
please,
this
is
a
design
are
including
RTP
payload
format,
and
we
can
see
from
figure
is
that
it
says
the
RTP
fixed
RTP
header
also
introduced
a.
K
Rita
capsulate
RTP
pero,
the
type
that
used
on
the
original
pero
type,
and
we
have
two
kinds
of
capsule,
a
different
type,
which
can
distinguish
different
and
interleaving
skins,
and
also
we
have
a
second
number
offset,
is
to
use
it.
You
calculate
the
original
sequence,
number
and
Tom
stem
Bob's
asset
is
used
to
calculate
the
original
timestamp.
K
This
is
this.
Article
matt
is
a
blip
applicable
for
pores
of
the
interludes
games.
Introducing
the
previous
slide.
Yes,
please
I
heard
example
in
to
leave
rd
packets
only
including
only
one
of
capsule,
medium
frame
type,
and
for
this
this
example
a
cabeza
for
RTP
packets,
tying
two
NV
without
without
a
aggregation
yeah.
That's
please!
K
K
To
produce
the
dementia
is
choosing
a
package
carrying
some
important
information
like
I
brains
to
being
to
leave.
The
can
on
is
supposed
to
reduce,
reduce
costs,
a
delay
to
do
that.
The
we
must
allow
that
the
interleaved
RTP
packets
can
be
transmitted
together,
always
as
the
long
interleaved
RTP
packets,
so
that
the
the
change
lever
could
identify
the
interview
if
the
package.
Q
D
Add
junk
like
so
it
looks
like
you're
assuming
that
aggregation
never
changes
the
number
of
packets.
So
if
you're
aggravating
five
packets
together
everything
if
you're
in
early
ting
five
packets,
you
always
get
five
interleaf
packets,
you
won't
be
a
lot,
have
a
reduced
number.
Otherwise
the
sequence
number
offset
doesn't
work.
Yeah.
K
D
Okay,
so
you
so
you
can't
just
use
it
a
sort
of
a
generic
aggregation
mechanism.
I
mean
we
come
in
theory.
If
you
had
a
more
flexible
sequence
number
mechanism,
you
could
have
say
you
interleave
five
source
packets
as
to
literally
packets,
or
something
like
that.
But
you
don't
support
that
in
this
yeah.