►
From YouTube: IETF98-AVTCORE-20170328-0900
Description
AVTCORE meeting session at IETF98
2017/03/28 0900
B
A
B
A
So
all
right,
this
okay,
so
we
have
volunteers,
we
have
ball
and
baroda
spell
checkers
in
magna
from
the
upper
room
you
are
without.
If
you
can,
people
are,
please
use
the
music,
oh
he'll.
If
you
can
I'm
fund
up
the
attention,
everybody
yell
at
me
to
say
they
hit.
Pay,
has
a
little
to
impress
read
by
taking
notes.
Is
ball
and
Varun
photo.
A
E
A
A
So
yes,
a
blue
sheets
are
going
around
join
the
meeting.
You
know
if
your
remote,
you
do
really
a
sub
but
want
to
do
oops.
Yeah
I
have
the
wrong
URL
to
me
to
go
so,
hopefully
don't
try
to
go
to
the
medico
for
sole
one
of
a
good
time,
not
too
much
going
on
there
anymore.
All
right,
oh
no!
Well,
hopefully,
to
see
this
before.
If
not,
we
your
lawyer,
ok
working
group
status.
So
yes,
we
have
now.
A
You
might
think,
looking
at
the
two
of
us
up
here,
you
earn
a
dbx
but
no
you're
in
a
ticket
car
because
we
have
merged
the
working
groups.
So,
but
we
are
figuring
under
the
name
agency
far
because
that
was
the
easiest
thing
for
the
data
tracker.
Apparently
I
was
hoping
to
go
back
to
apt,
but
apparently,
if
we
did
that
we
would
the
data
cracker
would
decide
that
you
know
this
is
a
group
where
nothing
had
happened
in
five
years
with
the
toilet,
all
the
status
and
five
years
ago,
nothing
sits.
A
So
it
was
easier
to
keep
the
AVP
for
days.
So
thank
you
to
Magnus
Ronnie
for
carrying
a
BP
car
and
we
are
happily
too
noble
for
them
and
will
still
be
relying
on
them
for
helping
us
understand
what
happened
in
a
PC
to
our
property.
Sir.
Alright,
then
this
big
wii
virtual
start
with
our
then
the
back
and
status.
B
A
Which
what
we're
doing
right
now,
then
we
will
have
mo
doing
great
marketing
and
then
my
very
last
minute
addition
say:
I
decided
to
write
this
draft
there
at
the
welcome
reception
on
sunday
and
submitted
it
yesterday
afternoon.
Well
sure
you
all
right,
if
I
now,
it's
very
short
and
most
of
the
text
I
copied
it-
takes
it
from
the
graph
from
2007.
A
So
hopefully
it
straightforward,
but
by
what
could
be
talking
about
that
and
then
at
the
end,
if
anyone
has
any
other
business,
it
couldn't
be
a
very
cold
and
then
we
can
actually
against.
Then
we
should.
The
payload
is
in
this
room
after
that,
but
I
think
it's
not
scheduled
till
ten-thirty,
so
people
who
take
a
break
be
doing
all
right.
Since
we
heard
some
definition
of
lee
had
one
I
get
to
another
puzzle
as
well:
go
471,
RFC
published,
which
is
better
controlled
messages
to
their
codex.
A
F
A
A
Right
we
got
135
fists
and
the
other
three
unbundle,
so
I
think
now
great
rid
is
probably
also
the
qat
budget
that
doesn't
matter
anyway
they're
all
in
this
rest,
the
screens
are
basically
we're
done
unless
anything
here
comes
up
in
all
48,
which
hopefully
it
was,
and
even
then
that's
mostly
between
the
authors
in
this
and
the
RFC
editor.
So
that's
also
that
also
counts
as
productive
for
us
and
then
milestones
is
this
Corp.
You
still
have
to
do
so.
Let's
talk
about
that
so
aria,
it's.
G
A
bad
blogspot
too,
I
dropped
my
lap.
This
is
we
were
starting,
so
the
I
forget
the
name
of
it,
but
the
one
that
is
with
me
with
the
ietf
re
SG
is
right:
SRT,
p,
I
think
that
is
ready
to
go
cant
rember.
If
we
need
to
go
back
to
ETF
last
call
I
had
some
questions
for
the
Shepherd
or
chairs
that
whoever
it's
now
shepherding,
which
you
decide
that
and
that
I'd
like
to
get
cleared
up
first.
That
should
be
easy
enough.
G
A
Right,
okay,
so
yeah.
The
one
thing
to
go
through
the
side
is
what
should
happen.
I
mean
I
am,
of
course,
happy
for
Magnus
and
Ronnie
to
keep
Shepherd
and
graphically
we're
already
doing,
but
if
they
don't
want
to
do
the
work,
I
also
understand
that
so
any
comments
from
your
bio
Eckert
look
review
with
shepherding
this
one
of
it.
Ronnie.
Okay,.
B
H
H
I'll
have
to
go
back
because,
as
far
as
I
as
far
as
I
remember,
then
we
did
all
of
it.
It
just
for
the
SLPP
was
the
last
change.
Look
because
lady
was
just
taking
out
the
SS
registration
from
the
document,
because
that
was
the
comment
that
were
made
to
the
I
think
to
the
IDF
last
call
about
the
concerns
about
registering
doing
a
stunt,
attract
no
comment.
H
Don't
even
yes,
I
think
this
is
the
major
problem
in
the
moon,
at
least
when
I
was
looking.
If
I
was
it's
not
clear
why
the
Estes
is
requires
a
standard
trac
RFC,
while
the
SLPP
and
the
what's,
the
other
one,
the
Mikey
person
trick
doesn't
require
so
basically
asasta
pls.
This
should
have
the
same
registration
requirement,
but
that's
not
up
to
to
us
to
do
that.
So
that's
that's!
A
problem
of
the
documents
have
created
those
registry.
G
I'm
not
going
to
fend
defend
the
SS
policy.
We
have
lots
of
policies
that
are
inconsistent.
It
don't
make
sense.
I
would
have
to
look
back
at
that
one
to
make
any
any
a
judgment.
If
people
think
that's
the
problem,
one
can
always
start
an
effort
to
change
policies.
All
it
takes
is
a
RFC
updating,
some
other
RFC
I.
G
A
H
All
I
mean
it's
not.
I
mean
this.
This
document
doesn't
do
anything
about
the
it's
just
registered
this
security
profile
in
in
the
registry.
That's
all
it
does.
I
mean,
because
they
are
not
by
the
way.
All
right
surface
is,
I
think
it's
in
its
informational
or
something
that's
that's
there,
but
it's
it
just.
They
only
I
had
a
problem.
Also,
with
this
comment
saying
we
don't
want
to
have
a
standard
dr.
Eck
document
registering
something
that
is
not.
H
That
is
a
sum
that
it's
a
country-specific,
codec
or
something
but
I
mean
the
registry
should
be
more
easier
to
handle
on
this
case,
I
don't
understand.
What's
the
White
was
in
the
first
place
in
and
the
answers
I
got
to
a
not
releasing
the
answers
I
got
when
I
try
to
figure
out
why
this
like
that
they
were
saying
we
don't
remember
I
we
just
did
it
this
way,
there's
no
really
thought
behind
it.
I.
I
I
Right,
so
maybe
that's
wrong.
Do
we
care
right?
I
mean
you
know
it
was
it
was
they
just
were
formed
at
different
times
with
different,
like
by
the
time
we
were
on
dtls
srtp,
the
general
motivation
at
ITF
that
the
wind
had
shifted
to
the
we
should
let
anybody
register
anything
they
want.
Otherwise
they
just
sit
on
our
code
point,
so
they
should
be
free
for
everyone.
Unless
there's
not
enough
of
them
are
something
like
we're
discussing,
so
it
doesn't
really
matter.
I'm
is
there.
I
A
I
A
I
Corroded
the
great
the
security
ad
object,
no
but
just
know,
but
he
objects
to
area
being
a
standard
tracks.
Algorithm
write.
G
Okay,
originally
stewed
work
two
or
one
right,
and
the
objection
was
that
that
seemed
to
give
Aria
the
cachet
of
a
standard
strike
draft.
So
now
we've
split
them
right,
so
the
srtp
one
and
the
one
that's
describing
the
I
think
the
ones
describing
the
cipher
suite
is
is
informational
and
that's
what
we're
talking
about
today,
the
one
that's
the
one
that.
A
Base
what
I'm
trying
to
do
here?
Maybe
I
was
up
there.
You
know
what
I
mean
I,
you
know
personally,
I
had
the
problem
with.
You
know
the
security
descriptions,
registration
being
a
senator
document,
but
I
just
worried
that
trying
to
get
secor
that
we
want
to
do
a
ITF
consensus
on
a
cipher
double
cipher.
He
doesn't
like
using
a
key
mechanism.
He
doesn't
like
might
be
challenging.
I
You
know
sometimes
you
just
need
to
be
challenging
with
a
DS
I,
don't
mind
helping
with
that
look.
I,
think
that
you
know.
My
point
of
view
is:
if
the
working
group
decided
to
take
on
this
work
and
put
a
milestone
for
it
at
that
point
in
time
we
more
or
less
agreed
to
give
it
a
code
point.
You
know
like
that.
That's
ridiculous
to
go
back
on
all
those
decisions.
At
this
point
in
time
I
mean
we've
got
exactly
the
right
solution.
I
A
F
Mean
in
the
discussion
we
had
before,
I
think
he's
lost,
meeting
I
think
it
Chris
position
was
that
if
it
is
having
different
policy,
it
wouldn't
be
considered
vetted
by
ITF
and
therefore
not
really
approved.
So
that's
what
did
receive
a
shame
to
policy,
but
I
think
the
right
thing
you
asked
out
to
do
with
area.
It's
just
to
push
it
through.
K
Oh
I,
don't
think
so.
I
think
the
last
time
we
met
we
discussed
it
maybe
a
year
ago
and
the
output
of
that
was
that
we're
ready
the
documents
ready.
It
needs
some
work
on
the
Security
section,
but
apart
from
that,
it
needs
wider
review.
I
think
some
people
Wallin
tiered,
but
I
did
not
get
any
feedback.
So
I
guess
we
need
to
run
through
the
thing.
K
B
A
K
K
K
Powered
here,
yeah
I
think
it
would
be
nice
because
it's
being
implemented
and
by
various
people
and
I've
gotten
feedback
from
implementers,
so
it's
kind
of
ready.
From
that
perspective,
a
last
time
we
discussed
that
we
should
put
something
for
security
there
and
there's
some
discussion
of
like
with
what
way
we
want
to
go
with
this
I
believe
some
people
in
in
the
IETF
are
working
on
NPR
DP
with
srtp
do
TLS
thing
some.
We
don't
have
to
wait
for
that.
K
K
A
A
D
A
F
H
First
comment
was
about:
there
was
a
texture
about
that.
You
can
do
it
if
you,
because
it's
negotiated
support
for
the
tube,
a
mode
that
you
it
support,
both
or
one
bite
and
to
bite.
But
there
was
a
cop.
There
was
a
texture
saying
that
you
can
do
it
also,
if
even
if
you're
no
negotiation,
if
you
have
some
other
ways
to
know
that
it,
the
other
side,
supports
that,
and
that
was
the
comment
from
from
been
about
it.
H
The
reason
that
is
that
it's
because
it
mostly
was
because,
in
the
cases
where
you're
doing
a
declarative
sdp,
you
don't
get
a
response
in.
We
don't
know
if
the
other
side
supporters,
and
in
order
to
do
this
mix
of
one
back
to
do
by
then
you
have
to
know
that
that
is
supported
by
some
other
means.
That's
the
major
reason
for
that.
It
was
not
saying
that
negotiation
is
nothing.
A
Is
it
I
mean
probably
on
the
downer
I
mean
it
get
something
quick
to
say
otherwise,.
D
L
D
A
F
Okay,
recap
of
history
here
so
multiplexing
guidelines
in
all
those
one.
A
half
years
ago,
in
Yokohama
we
discussed
this
document
and
considered
okay.
We
need
to
scope
it
down,
focus
it
etc,
and
that
was
a
significant
editor
task
which
I
I
kind
of
took
on,
but
never
have
gotten
to
each
other
things.
So
and
and
the
question
is
there
anyone
who's
actually
willing
to
do
any
editing
on
it
now
so
I
I
might
suggest.
A
Mean
I
guess
the
question
is
this
document
you
know?
Is
it
like
a
needed
I
mean
I,
know
that
there
was
I
feel
like
it
was
sort
of
a
it's
all
such
a
position
paper
when
we
were
fearing
a
bundle,
but
now
that
bundle
is
in
consent,
as
we
have
more
or
less
consensus
on
what
that
looks
like
it
is
still
needed
yeah.
So.
F
I
mean
that's
opposed
to
kind
of,
refocus
and
say:
okay,
trying
to
be
clear
on
giving
actual
guidance
form
from
the
position
we
have
after
Bongo
a
little
all
other
things
come
into
place,
let's
say
so
it's
it
would
be
good.
I
still
think
you
can
figure
out
most
of
it
from
other
sources,
but
it's
just
that
you.
You
have
to
look
in
round
bits
to
figure
out
how
its
land
lies,
saying
what
to
think
about
it,
but
yeah
I
I
think
it's
there's
lack
of
energy
to
actually
complete
this.
H
A
F
F
F
A
H
A
M
A
A
We
were
not
going
to
say
that
so
I
think
I.
I
think
the
lrr
is
ready
and
then
he's
been
marking
relationships
to
it
and
you
figured
out
in
trademarking.
So
I
would
suggest
that
since
I'm
the
author
on
that,
you
want
to
go
ahead
and
do
a
booking
the
best
of
all
I'm,
totally
half
a
minute.
Okay,
buddy,
looking
the
left
column
at
and
free
marketing
is
on
so
only
15,
minutes
late,
but
no
execution
so
I
think
I'm
ready
to
start
trademarking.
N
On
things
happening
operatives
in
the
time
I
know
I
did
okay.
So
this
is
the
video
frame
marking
arts,
better
extension
version
for
slide.
Please
just
a
quick
review
of
why
we're
doing
this
work.
The
goal
is
to
have
a
narc
middle
box.
Will
call
here
are
two
piece
which
it
could
be
sfu
or
one
of
the
other
topologies.
But
the
key
point
is
that
it's
a
middle
box
that
wants
to
avoid
deep
payload
inspection
and
number
one
to
avoid
decryption
and
some
architectures
may
not
even
be
able
to
perform.
N
N
We
can
get
better
recovery
when
you
have
packet
loss
scenarios
identifying
where
the
start
and
end
of
frames
are
is
tricky
with
current
payload
floor
mats
and
this
unifies
it
and
makes
it
more
robust.
You
can
do
clean
video
switching
at
iframes,
so
you
can
identify
when
each
recipient
has
all
of
the
packets
necessary
for
intra
refresh
and
under
congestion.
Switches
can
make
decisions
about
the
relative
importance
of
packets.
N
So
if
it
has
the
draw
packets
towards
a
particular
endpoints
downlink,
it
can
make
an
intelligent
decision
about
that
drop
and
they
can
drop
entire
layers
if
it
wants
to
bring
down
the
the
quality
of
a
particular
stream
to
a
particular
receiver
and
endpoints
can
also
use
this
header
extension
for
some
of
the
same
benefits.
The
better
information
about
starting
in
frame
markers,
allow
better
recovery,
so
you
don
to
decode
and
display
dirty
artifacts
next
slide.
N
Okay,
so
this
is
the
extension
as
it
stands
right
now,
if
you
remember
in
the
last
version,
we
actually
forked
it
into
two
different
versions
of
it,
a
short
form
in
a
long
form.
So
this
is
the
full
long
form
it
has
the
layer
IDs
that
did
layer,
ID
and
teals
there
pickin
index
fields
and
that's
when
you
have
length
equals
2
and
the
50
285
length
is
11
plus
so
2
here
actually
means
three
bytes
following
it.
N
That's
the
long
form
and
the
short
form.
If
you
put
length
0,
then
you
only
have
one
bite
afterwards,
which
is
basically
the
flags.
The
frame
flags,
sei
DB
and
the
rest
of
the
course
of
the
fields
are
emitted
next
slide,
please.
So
the
changes
that
we
made
there
were
some
comments
about
what?
If
the
middle
box
doesn't
have
all
the
information
required
for
all
the
layer,
IDs
and
so
the
earlier
verte
in
version
3,
we
said
that
the
short
form
was
only
for
non
scalable
strains.
N
N
No
I
think
it's
right.
So,
okay,
it's
it's
contort
head.
It
was
for
the
nonce
Caleb
all
right.
The
short
header
was
for
the
non
scalable
and
it
was
only
for
the
non
scalable
in
version
3.
There
were
comments
about
what.
If
the
middle
box
knows,
it's
forwarding
scalable
streams,
but
it
doesn't
actually
have
all
the
scalability,
and
so
it's
allowing
the
middle
box
to
use
the
short
header
even
for
the
scalable
streams.
So
I
know,
I
stream
is
scalable,
but
I
really
don't
know
what
the
layer
IDs
of
it
are.
N
So
I
can
still
use
the
short
form.
So
the
text
is
there:
they
we
recommend
using
the
short
header
for
non
scalable
streams,
but
you
may
also
use
it
for
scalable
streams.
If
you
don't
have
information
about
the
scalability
and
then
the
flip
side
of
that
is
also
allowed.
So
you
can
use
the
long
version,
the
full
full
extension,
even
if
you're
forwarding
non
scalable
streams,
and
so
this
may
be
a
simplification
and
some
and
some
boxes
to
have
uniformity
of
the
the
headers
as
they
go
out
and
when
you
do
that,
you
must.
N
You
must
set
all
of
those
scalability
layer,
IDs
20
right
that
would
everything
was
fine
with
that
right.
Okay,
so
next
changes
a
lrr.
There
was
a
request
to
have
lrr
and
framework
harmonized
and
so
Jonathan
I
had
a
at
a
talk
and
we
look
at
the
different
fields
and
we
think
we
have
what
we
need.
It
was
just
a
message,
a
few,
a
few
minutes
before
the
meeting
started
about
vp9
payload
format,
the
VP,
not
scalability
fields
and
I
believe
that
that's
correct
it
needs
to
be
updated.
N
M
M
A
Obviously,
if
you
have
a
codec
where
your
layers
are
numbered,
sequentially
you
do
but
there's
some
that
are
particularly
to
the
tour
SVC,
I
think,
is
the
DQ
ID,
so
they're
actually
almost
always
not
going
to
be
sequential
arm,
in
which
case
no,
you
don't
right.
So
in
that
case
yeah
you
need
to
know
somehow
what
the
structure
of
you
know.
I
mean.
A
A
Good
competition
yeah.
So
basically,
so
you
know,
I
like
the
first
is
I
think
for
on
agencies
for
SVC
stream.
Your
LEDs
will
actually
be
numbered.
Something
like
how
many
bits
it
is,
but
it
is
only
like
0
8
16
24,
because
the
the
Q,
the
Q
layer
is
the
bottom
three
bits
and
the
others
and
the
D
layer
is
the
upper
bits.
So
you
like
so
a
box
which
is
ignorant
of
uses
for
SVC.
A
A
A
N
J
A
A
Be
layers
in
the
be
faced
our
sink,
which
likely
is
another
point
I
want
to
get
to.
Let
me
finish,
doesn't
miss
much.
It
is
but
I
mean
so,
but
I
guess
so
the
point
is
that
so
a
them
I
mean
if
your
situation,
where
spatial
layers
are
always
the
same
time
stamp,
then
you
can
tell
just
by
probably,
but
you
know
whether
you
are
to
be
sequence,
numbers
missing.
N
J
J
N
A
I'm,
so
we
want
to
pick
in
the
draft
saying
you
know
this
is,
for
you
know,
scalability
structures
that
follow
these,
don't
use
it
for
complicated
scalability
structures,
because
you'll
have
a
bad
time,
though
sort
of
basically
have
some
sort
of
applicability
statement
saying
these
are
the
sort
of
scalability
structures
you
should
use
this
for.
Don't
use
it
for
anything
else,
because
he'll
just
confuse
people
right,
okay,.
A
The
other
coming
to
heaven,
lrr
I,
think
this
is
good.
I
would
strengthen
so
to
the
one
thing
I'd
add
is
in
addition
to
saying
they
are
they
they
are
the
same.
I
would
say
that
when
you're
defining
this
for
a
new
codec
it
you
know
if
your
freedom
ideas
for
new
codec,
which
supports
both
lrr
and
frame
marking,
it
must
be
the
same.
A
sleaze
have
a
stronger
statement
saying
for
new
standards.
They
must.
We
must
encode
them
the
same
way.
Ok,
so
just
must
correspond
to
the
CI.
A
A
He's
gonna
mean
you
sendin
lrr
four,
you
know
al
ID,
you
know
four
to
forty
five
right,
you
get
a
packet
in.
Has
it
been
satisfied?
So
maybe
that's
just
you
know
you.
You
know
you
got
a
thing
with
LED
45
and
the
me
get
set
and
that
might
be
sufficient,
but
I
think
which
at
least
I've
got
have
statement
to
that
effect.
Okay,
okay
and
then
my
third
comment
about
the
vp9.
So
no.
A
J
A
Two
was
in
the
frame
marking
graphs,
it's
just
saying:
I
I
thought
I
need
to
you
know,
make
sure
this
is
correct.
Basically
saying
that
when
you
send
an
LR
are
the
way
you
tell
what
the
lr
has
been
satisfied.
I
think
is
that
you
get
a
receive
a
packet
or
maybe
a
start,
a
frame
packet.
This
is
the
detailed
who
didn't
work
out
with
the
be
bit
set
and
that
TI
d
and
li
d
in
the
packet
so.
A
A
N
H
Running
even
just
a
question
about
the
point
you
made
about
the
was
the
lrr.
Why
is
it
important
that,
when
you
sent
it
in
are
all
you
have
to
to
review
it
review
it
in
the
frame?
Art
I
mean
it's,
it's
mostly.
You
have
to
know
it
when
the
encoding,
but
I,
don't
understand
why
you
need
it
for
me
to
be
to
identify
it
in
the
frame
of.
A
N
H
N
Definitely
don't
plan
the
document
of
that
in
frame
working
I
was
just
document
the
when
you
receive
a
frame
marking.
What
does
it
mean
and
and
and
when
do
you
know
that
that
means
the
lower
refresh
happened?
So
we
can
put
that
part
of
this
in
the
spec
complex
things.
I
thick
should
go
on
the
lrr
little
box
considerations,
okay,
so
anything
else
in
lr.
Are
we
good
with
it?
Okay,.
N
Okay,
one
other
thing
that
Bernard
I
think
brought
up
was:
how
does
this
work
with
repair
streams
and
we
updated
the
document
to
show
that
this
only
applies
to
the
source
RTP
streams,
not
to
the
repair
streams
and
specifically
only
to
video
payloads.
So
if
you're
trying
to
use
this
for
audio
or
for
fact
or
retransmission
payloads,
that's
not
gonna
work,
it's
not
specified
here.
You
need
of
the
draft
for
that.
N
Now,
we'll
note,
though,
that
really
probably
what
you
want
is
to
recover
the
original
packets
markings
from
r-tx
or
effect,
and
the
RTX
effect
format
will
tell
you
usually
how
to
do
that,
how
to
recover
header
extensions
that
were
on
the
original
packet.
So
it
wouldn't
be
this
draft
showing
how
how
those
packets
get
marked
it
would
be.
The
repair
format
would
actually
show
how
to
preserve
the
header
extension
of
the
original
packet
and.
F
N
N
What
so
again,
Magnus's
point
if
you're
an
architecture
you
know
like
part,
one
of
the
goals
is
for
the
middle
box,
actually
participate
in
the
repair
and
and
not
be
blind
to
the
repair.
So
so,
in
that
case,
the
middle
box
can
receive
repair
and
generate
repair
if
it
wanted
to.
In
that
case,
it
would
be
able
to
to,
you,
know,
decide
whether
or
not
to
forward
lr
RS
or
whether
it
has
all
the
information
for
the
for
the
lair
anyway
and
despite
the
lrs.
N
So
there
was
a
also
a
previous
draft
from
Bernard
about
general
considerations
when
you're
building
in
SF
you.
It
has
a
lot
of
things
in
it,
not
directly
related
to
frame
markings,
but
some
things
do
directly
relate
to
frame
marking
into
the
previous
versions.
We
had
a
reference
to
it
and
I,
don't
think
the
work
is
going
to
progress,
so
I
think
the
decision
was
to
import
the
parts
that
are
relevant
to
frame
marking.
So
we
did
that
and
the
most
liberal
part
is
actually.
N
We
were
just
discussing
a
little
while
ago,
the
scalability
infrastructures,
when
you
have
fixed
fixed
scalability
formats,
those
scalability
structures,
autistics
for
male
impacts,
the
headers
or
actually
payload
units
and
265.
Now
in
Passy
info
those
those
things
are
pulled
over
from
the
from
the
other
aspect.
So
we
can.
We
could
let
it
die
if
I,
if
we
want
to
now
without
any
kind
of
file
reference
issue
and
I,
think
we
also
need
to
add
the
vp9
scalability
structure.
Oh,
we
will
remove
anything
about
vp9
wheel
at
the
vp9
spec.
N
N
The
next
the
next
point,
something
that
a
meal
had
brought
up
when
looking
at
looking
at
this
in
the
context
of
perc,
probably
longer
discussion
that
we
need
to
have
in
in
the
perk
session,
since
it's
going
to
presenting
something
about
about
it
there.
But
the
quick
higher-order
bit
here
is
that
when,
when
a
participant
or
middle
box
is
using
RTP
to
probe
the
link-
and
it
may
only
use
padding,
it's
not
actually
sending
any
media
payload
at
all,
just
using
RT
padding
to
probe
the
link
with
different,
you
know
bite
lengths.
N
The
the
problem
is
that
in
part
the
padding
is
encrypted,
the
padding
it
length
indicator
even
is
encrypted,
so
middle
box
wouldn't
know
whether
or
not
this
is
real
media
or
padding.
Only
media
so
may
not
know
whether
or
not
afforded
so
one
of
the
proposals
and
real
spec
is
to
extend
frame
markings
to
be
able
to
indicate
that
this
is
padding
only
one
of
the
things
that
we
considered
is.
N
Maybe
this
is
a
good
use
of
the
discardable
bit
if
the
padding
only
packets,
Marc,
Marquez,
discardable
and
middleboxes
could
drop
them
if
they
wanted
to,
and
then
also
another
option
is
don't
use
the
real
halo
type
of
media
stream.
It's
a
little
odd
to
say
this
is
you
know,
264
vp9
media
and
you
really
it's
just
padding
probe,
so
you
could
use
the
effect
or
retransmit
payload
type
or
or
even
rtcp,
something
that's
not
indicating
an
actual
real
milk
media
payload
type
to
do
that,
probing.
That
would
be
another
option.
J
J
A
N
A
J
N
J
A
B
A
J
J
No
taken
up
this
is
not
not
trying
to
defend
that
way
of
being
bandwidth.
Probing,
that's
how
bandwidth
probing
is
happening
today
in
chrome,
specifically,
I'm,
not
sure
about
Firefox.
Maybe
it's
the
same
and
the
third
bullet
there.
The
third
dash
there
is
a
very
noble
goal,
but
it's
not
how
things
work
today,
so
I
I
think
it
would
be.
If
you
compare
the
workload
necessary
to
get
these
implementations
to
change
with
the
workload
necessary
to
add
a
bit
that
says
you
can
drop.
J
This
I
think
it's
it's
clearly
easier
to
just
to
just
add
that
bit
not
not
to
mention
that
it's
I
mean
there's
nothing
wrong
with
it.
It's
a
great
so.
N
A
N
J
J
Yeah
sure,
but
we
reserve
ourselves
the
right
to
come
in
beg
again
for
this
to
be
included
in
frame
markings.
So.
N
If
anybody's
interested
there's
a
in
the
perk
session,
there'll
be
a
description
of
more
detailed
description
of
this
end
and
another
aspect,
RTX
formats
for
park
and
is
padding
only
formats,
so
we
can
follow-up
discussion
there.
That's
it.
I
think.
So
I
guess
we're
not
ready
Wes
long
as
we're
going
to
remove
the
vp9
stuff
out
of
the
draft.
B
A
My
surprise
last-minute
additions
of
draft
so
yeah.
This
came
up
when
you
know
I'm
trying
to
bully
various
implementations
and
to
finally
moving
away
from
scary
descriptions,
and
they
said,
but
we're
doing
you
know
aes-256
what
should
we
do?
And
I
said
I
looked
at
the
registry
and
said:
hey,
there's
no
registration
for
aes
256
counter
in
dtls
srtp,
that's
bad,
so
excellent,
so
well,
I
looked
over
the
I
did
some.
You
know,
draft
archaeology
and
discovered
that
you
know
the
thing
that
became
look
at
those
numbers
right,
I
think
I
got
the
numbers
wrong.
A
Whatever
the
detail
is
our
key
piece
back.
You
know
up
through,
like
draft
03
of
the
working
group.
Draft
had
definitions
for
these,
but
the
AES
56
graft
itself
wasn't
ready
yet
so
I
take
it
out,
but
then
those
rich
never
got
put
into
the
aes.
56
draft
be
sure
this
was
just
an
oversight
if
you're
not
going
to
Cabo
gawak
automotive
them
an
excellent.
That's.
I
J
I
Ago
was
that
we
were
trying
to
reduce
complexity
of
the
sure
number
of
security
options.
We
need
to
negotiate
and
no
one
had
any
argument
for
any
use
case
where
this
was
better
than
the
other
ones.
So
it
was
like
why,
at
we
were
trying
just
to
reduce
the
comp,
the
implementation
complexity
of
having
like
note
by
adding
this
registration
you're,
not
redoing,
removing
the
need
to
do
GCM,
you're,
just
adding
another
thing.
You
have
to
do
as
well
right.
So
that's
why
we
removed
it
back
then,
and
I
I'm.
I
It
was
too
because
reference
to
go
away,
it
wasn't
just
like.
Oh,
we
can
make
this
miss
Roth
killer,
yeah.
A
A
I
J
I
I
A
And
so,
and
so
what
and
but
you
know
these-
you
know
their
crypto
kernels,
you
know,
don't
have
you
know
gah
what
counter
mode
in
them,
and
you
know
it's
subjective
certified.
So
it's
not
gonna
happen
anytime
soon.
So,
basically,
if
I
can't
tell
them
yeah,
you
should
great
move
to
teach
a
lesson
straight.
They
said
great
I
say,
but
it
means
either.
You
have
to
cut
back
to
128-bit
keying
or
move
to
gamma
counter
mode.
They
say
no
I,
think
I'll
say
what
security
description
so.
I
It'd
be
really
interesting
to
understand
which
crypto
kernels
they're
using
that
don't
have
that
that
do
have
it
for
256
counter
mode,
because
that's
the
one
I'm
not
a
I'm,
willing
to
believe
there's
something
out
there
I'm
not
aware
of
any
that
don't
have
I
went.
This
was
a
year
of
my
life.
That's
why
I.
D
A
Yes,
yeah
and
the
other
point
is,
you
know:
Libous
rdp
actually
implemented
these
code
points
because,
as
far
as
I
can
tell,
nobody
noticed
that
it
got
taken
out
of
the
draft,
and
so
you
know
I
actually
I
submitted
a.
I
pull
request
on
removing
them
for
the
beta
that's
about
to
come
out
because
they
said:
hey
wait
a
second.
This
is
never
defined,
but
then
I
thought
about
it.
That
way.
This
is
a
bad
idea.
I
should
maybe
just
define
so
the
next
next
slide
again.
Why
not?
A
What
gives
you
a
few
cm
and
the
answer
is
you
want
to
make
it
easy
to
move
off
of
security
descriptions
and,
like
I
said
our
crypto
colonel?
Doesn't,
have
you
see
Evan?
That's
fun,
ok,
again
and
thoughts.
That
I
mean
so
I
guess
colin
is
saying
you
know
we
want
to
minimize
complexity,
and
you
know
I
understand
that.
That's
said,
like
I
said,
the
lib
srtp
has
had
this
in
the
code
for
a
while.
I
Mean
I
know
how
their
new
coach
I
mean
it
doesn't
matter
if
they're
already
in
the
code
or
whatever
I
know,
what's
negotiate
with
a
lot
of
that
stuff.
Is
we
generally
move
it
out
of
the
policy
level
higher
okay?
So
you
know
in
that,
what
we'd
offer
and
what's
generating
the
SDP
is
different
than
what
yeah.
A
I
A
I
I'm
clearly
with
things
so
stupid,
I
don't
want
to
explain
them,
III,
don't
know:
okay,
I,
don't
think
this
is
a
casual
decision.
I
think
that
this
actually
makes
interoperable
harder
on
the
long
term
and
I
and
I,
on
the
other
hand,
would
like
to
make
sure
that
we
can
make
things
easier
from
people
trying
to
actually
build
stuff.
So
I
don't
really
know
what
we
should
do
with
this.
But
I
do
not
think
this
is
a
trivial
discesa
and
I'm
a
little
I'm.
Certainly.
A
G
H
J
A
F
F
A
O
N
Mercenary
they're,
just
a
clear
for
one
thing
on
Collins
point:
I
think:
maybe
the
reasoning
would
be
that
you
know
we
allow
a
bunch
of
crap
into
stp,
because
we,
we
all
think
it's
crafty
anyway
and-
and
you
know,
love
you
want
to
move
away
from
it.
You
know
to
something
else
in
the
future.
N
We
typically
don't
have
that
philosophy
in
dtls
and
we
don't
want
to
pollute
detail
s
with
things,
and
so
maybe
the
pushback
is
that
this
actually
has
to
go
into
the
dtls,
a
certainty.
You
know
profiles,
you
know
in
those
structures,
and
so
that
does
add
you
know
crofton
to
dt
allows
something.
We
don't
expect
to
replace
something.
We
don't
want
to
grow
absurdly
complex,
like
STP,
so
maybe
there
is
a
difference
there
between
allowing
a
spec
to
go
forward
in
sdp
versus
allowing
it
to
go
forward
in
dtls,
fatal
structures.
G
I
Just
want
to
highlight:
look
I'm,
I,
don't
know
what
I
think
about
this
anymore:
okay,
okay,
because
all
good
points
made
it
too
much.
Okay,
but
I
do
worry
that
let's
say
we
go
down
this
path
and
these
people
do
exactly
what
you're
proposing
here
we
now
have
just
created:
dtls
srtp.
Your
implementation
will
not
interoperate
with
our
implementation.
I
Like
a
failure
of
the
ITF
and
I'm,
not
saying
that
means
you're
showing
the
one
code
point
I,
meaning
the
problem
is
you
do
not
have
a
detail
SSR
to
be
implementation.
Unless
you
implement
the
MTI
ciphers
witching,
which
has
just
one
GCM,
you
have
to
do
GCM
to
have
a
compliant
detail,
SS
department.
No,
I
think
that
the
client
is.
Why
differ?
Okay,
I
may
be
wrong
on
the
RFC's,
maybe,
but
you
have
to
do
that
to
have
it.
I
F
A
F
O
I
I'm
not
arguing
that
in
the
slightest
I
agreed
with
all
the
comments
are
made
up
here.
Okay,
I'm
saying
that
the
argument
to
do
this
that
the
proposal
you're
doing
for
adding
is
sort
of
thing
right.
You
know
I
find
a
motivational,
it's
good.
Maybe
we
should
just
go.
Do
it
but
I'm
I'm,
not
talking
about
the
iono
registration
process
that
doesn't
matter
that
is
irrelevant
here
right.
I
The
question
is
is
like
what
trying
path
do
we
try
and
encourage
implementers
to
go
down
to
get
us
to
what
we're
trying
to
accomplish
right
and
and
the
original
argument
why
we
took
removed
a
whole
bunch
of
profiles?
Was
we
thought
that
they
were
actually
causing
things
to
be
harder
for
implementers
not
easier
for
implementers?
And
he
is
you
know,
is
that
the
case
here
or
not?
It's
hard,
sell,
I,
don't
know
I'm
like
I'm,
not
objecting
to
this
draft
I,
don't
know
what
I
think
I
it's
just
brought
up
through.
I
A
Alright,
so
I'm
sure
what
the
best
way
forward
is
I
mean
I
could
do
it
as
a
I
mean
I
could
do
it
as
a
pure
specification
require,
but
I
still
feel
odd
about
this
or
something
that
feel
that
you
know
for
a
a
crypto.
You
know
that's
our
PP
spec,
which
is
itself
an
ietf
satyrs
track.
So
they're,
not
80
advice.
H
Known
even
I
think
you
you
should
pollute
it
and
register
it
here
among
its
I,
don't
understand
the
argument
about
inter
probability,
because
the
interoperability
is
not
as
in
this,
but
inter
bilities
in
the
implementations
and
and
both
the
RTC
weapon
I
had
to
do.
It
also
include
EP
mapping
to
specify
what
is
what
it
must
be
supported,
right,
yeah.
H
I
Will
certainly
try
to
consult
with
you
know,
I
mean
like
in
general,
like
a
plot.
Thank
you
for
bringing
it
and
considering
the
way
you
do
versus
just
going
in
filing
out
how
you
could
have
right,
you're
going
to
use
process
to
act
this,
but
me,
let's
just
like
obviously
the
bets
think
about.
What's
the
best
thing
we
can
do
to
help
you
guys
get
to
it
in
a
robber,
okay,.
F
A
G
I'm
going
to
take
back
my
statement
merely
go
about
having
no
particular
ad
advice.
I
will
point
out
that
making
this
a
working
group
by
them
takes
working
group
resources
a
time
for
something
that
can
be
done.
A
specification
required
so
getting
the
advice
of
the
working
group
is
perfectly
fine
going
through
the
glass
call
process
and
all
that
sort
of
thing
is
probably
overkill
I.
You
know
if
people
really
think
this
is
a
problem
that
we
need
to
relook
at
the
registration
policy.
A
A
D
A
F
J
F
I
A
All
right,
I'll
talk
to
I,
Anna
and
figure
out
what
they
actually
want
for
specification
and
see
what
and
probably
also
talk
to
security
people
and
see
what
they
think
they
think
with
us
all
right.
No.