►
From YouTube: WebRTC Interim Mar 1st 2017
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
B
C
B
D
B
B
Okay,
then
that
excess
you
will
see
in
the
ITR
policy
and
I
notice
that
we
have
new
members
joining
the
group
all
the
time.
But
if
you
haven't
read
the
ipi
policy
I
recommended
and
then
we
say
welcome
and,
as
you
know,
we'll
have
issues
in
media
capture
that
into
solving
not
medicine
and
if
using
webike,
a
CPC,
we're
still
hoping
to
get
to
the
CR
stage
real
soon,
but
so
that
requires
that
the
spec
stop
wiggling.
B
That
means
that
we
have
to
get
things
settled
and
if
you
and
when
we
have
bugs
traced,
they
have
to
either
say
here's
a
change
of
things
to
work.
All
that
we
have
to
say
this
is
the
bagra
16
before
CR.
That
might
mean
replace
it
in
the
icebox.
So
it
might
be
that
way
and
they
close
with
us
no
action
needed
of
this
land.
E
B
B
Talk
about
to
get
that
selectors,
but
you're,
so
anything
we
hear
a
lot
anything.
We
said
that
that
sender
and
receiver
make
much
more
sense
than
them
check
as
a
selector
and
as
I
felt
we
ship
with
shipping
chrome
without
the
selector
at
the
moment.
But
if
the
workgroups
want
something
special
keys,
come
Conan.
F
F
If
there's
no
selector
gather
everything
if
there
is
selector
if
the
selectors,
ascender
and
gather
stats
for
the
RTC
inbound,
RTP
stream
sets
object
corresponding
to
this
lecture
and
then
all
sets
objects
that
are
referenced
directly
or
indirectly,
by
a
metal
object
and
vice
versa.
For
receiver,
you
get
the
RTC
I've
found
strange
stats.
Sorry,
that's
all
isn't.
F
I
got
the
reversed
and
the
language
there
sorry
I
have
to
check
it,
that's
a
little
gay
as
well,
but
yeah.
So,
obviously,
if
you
provide
other
receivers,
you
get
the
inbounds
crisis
center,
you
get
the
outbound
and
you
get
all
sets.
Objects
are
reference
directly
or
indirectly
by
that
objects
and
correspond
to
means
the
corresponding
to
select
means
that
it's
SSRC
stats
at
agreed
matches
the
SSRC
in
the
selector
and.
B
F
F
B
F
A
F
F
That's
the
next
slide
and
speak
about
that
so
for
a
receiver
to
cassia
in
bone
RTP
stream
set,
which
would
include
which
has
references
to
associate
status
ID
for
the
rtcp
stuff,
so
that
their
share
would
be
included
and
also
the
show
the
track
track.
I
key
Kotick,
ID,
transportation
and
transportation
turn
will
include
ERCP
transfer,
ID,
selector,
canada
pair
Y
key,
which
is
only
as
I
understand.
F
The
main
value
is
still
the
selector
is
that
we
talk
to
customers
who
have
up
to
50
with
what
they're
doing
a
conversation
with
a
media
server,
and
they
have
up
to
50
tracks
in
their
peer
connection.
So
being
able
to
select
on
attract
would
would
substantially
narrow
down
the
amount
of
work
needed
to
pull
back
and
amount
of
data
fantastical
back
using
the
one
factor
that
is
the
multiplier
was
on.
The
road
tracks,
for
a
generous
receivers
has
been
language.
D
Twitter
lat-
this
is
adam,
just
just
a
question
about
number
kind
of
we're
redoing.
This
eight
guys
is,
has
a
rat
in
a
discussion
about
kind
of
deprecating.
This
select
their
singing
and
going
for
like
get
stats
on
this
understand
receiving
one
that
gets
that
from
the
pair
connection,
be
the
kind
of
the
knowledge
selector
getting
everything.
That's
too
big
of
a
change.
I.
D
Yeah
I
mean
the
select.
Your
thing
was
I
think
it
was
a
to
begin
with.
We
didn't
have.
We
didn't,
have
an
object
representing
a
track
being
sent.
So
we
kind
of
tied
a
pair
connection
and
the
track
together
right
this
electorate,
guy,
getting
that
such
rotation
I
mean
and
now
what
we
have
representations
of
objects,
beings,
antics,
lenders
and
receivers.
You
could
just
have
a
guest
at
called
on
them,
but
I,
don't
know
Greg.
That's
too
big
of
a
change.
A
D
A
D
B
I
know
another
thing
that
the
implementation
wise
it's
actually
easier
to
add
a
new
API
and
let
people
either,
however,
either
see
that
I
di
notte
than
it
is
to
add
this
kind
of
new
function.
New
use,
it
select
the
types
writing
the
fact
of
the
track
of
selector
is
that
none
of
the
implementations
do
what
we
want,
so
so
not
implementing
the
selector,
not
fessing
a
selector
for
for
the
day,
connection
stats
would
make
perfect
sense
right.
C
F
G
C
F
F
D
F
D
B
D
F
Well
can
I
suggest
one
approach
is
that
we
commit
this
PR
to
get
the
set
selection,
algorithm
and
I.
Guess
we
don't
want
to
sender
and
receiver
selectors.
Is
that
the
objection
and
we
could
work
on
a
one
way
to
do
it
would
be
to
commit
this
PR
and
then
work
on
the
moving,
get
stats
or
adding?
If
that's
the
center
here
as
a
separate,
we
are.
B
F
Therefore,
conference
that
we're
going
to
add
gets
that's
to
sender
and
receiver.
We
can
also
modify
this
PR.
To
not
add
the
extra
arguments
to
the
API
still
include
the
staff
selection
algorithm
using
the
media
stream
track.
There
will
be
a
compatible
way
to
get
the
sense
algorithm
in
which
I'm
not
hearing
any
objections
to
algorithms
Eric
hi.
C
B
Speak
speaking
it
with
it
with
a
chair
hat,
and
we
don't
form
to
that,
the
API
that
the
people
don't
have
come,
don't
have
a
need
for
hat
that
nobody's
stepping
up
to
say
they
have
a
need
for
a
year,
never
said
that
we
have
interest
lectures
because
it
reduces
the
amount
of
work
that
that
is
Peter.
When
you
do
many
checks.
F
B
D
D
F
F
F
A
A
Thing,
that's
really
I
mean
I
totally
agree
in
principle
where
it's
going,
but
it
seems
like
the
text
for
the
pr
still
needs
to
be
written
before
we
know
whether
it
works
or
not
before
we
decide
to
merge
it.
So
I
think
we're
a
little
bit
ahead
of
ourselves,
but
I
don't
care,
I'm
happy
to
do
it.
Whatever
we
want
I've.
E
So
this
PR
is
the
so-called
high
grade
or
loss
solution
that
which
we
read
every
the
last
forum
where,
if
you're
using
normal,
username
password
authentication,
then
his
name
is
a
username
express
credentials.
The
password
entries,
no
log
things
I
mean,
is
the
key
ID
and
credential.
Is
this
loft
credential
dictionary,
which
has
the
mat
key
and
access
token
so,
but
sound
right
so
far,
all
right
now,
the
next
slide.
E
So
it's
always
drawing
this
PR.
A
couple
issues
came
up.
I
just
want
to
make
sure
the
working
group
understands
them
and
now
that
there
are
no
outstanding
issues
with
the
precipice
approach.
So
I
believe
at
a
previous
virtual
interim,
we
decided
that
rather
than
the
OAuth
client
query
sighs
agent
for
its
hash
capabilities
it
we
just
always
request
a
56
bit
key
from
the
Olaf
server
and
use
that
to
generate
both
the
message:
integrity
and
the
message,
integrity,
sha256
attributes-
and
I
thought
I'd
believe
again.
E
The
decision
was
when
a
160-bit
key
is
needed.
The
256-bit
key
is
just
truncated,
but
RFC
76
35
says
that
an
appendix
we
go
150
bit
key
is
computed
by
taking
the
Shah
line
of
the
256-bit
key.
So
it
seems
like
there's
both
the
conflicts
between
our
c76
35
and
this
working
group's
decision,
and
also
it
seems
like
a
problem
that
this
is
only
mentioned
in
Appendix
B
and
not
defined
normatively
sense.
A
E
So
another
consequence
of
design
always
use
a
256-bit
key.
Is
that
essentially,
what
work
you
see
wouldn't
support
hatchability
then,
if,
if
the
client
would
want
transition
to
new
algorithm,
that
needs
something
more
than
256-bit
key,
so
is
the
problem,
or
is
that
I'm
not
just
working
to
understand
that
confidence
and
accept
it?
There's
angles
of.
A
H
Right
but
I
didn't
mean
10
year
was
that
service
different
sure
an
arbitrary
length
key
and
that
could
be
truncated,
whatever
the
kind
actually
support,
I'm
you're,
really
considering,
like
the
examples
being
so
I
wanna
cut
you
to
say,
but
there's
no
need
for
that
to
be.
You
know
restrictive
as
like.
Only
they
supported
now.
I
need
to
figure
out
like
this
thing
that
the
earlier
topic
about
like
to
be
a
shot
to
do
six
per
digest
always
which
produce
to
keep
replicates.
Then
we
are
kind
of
thinking
in
shock.
H
E
H
A
So
I
mean
I
I
think
we
should,
somewhere
on
the
ITF
side,
clearly
specify
that
you
truncate
and
whether
you
truncate
left
or
right,
because
that's
suspend
time
and
we
messed
that
up
in
srtp
and
and
then
your
whole
theory
for
how
we
have
crypto
agility
is
good
right.
But.
H
H
B
E
B
B
E
All
right,
so
this
issue
isn't
that
complicated.
When
you
first
create
a
channel
and
you
don't
pass
in
an
IV
explicitly,
then
ciebie
needs
be
assigned
based
on
the
negotiated
dtls
role.
But
if
you've
done
this
before
out
for
answers-
and
you
don't
have
a
Kiki
less
role
yet
so
currently,
chrome
and
firefox
just
returned
a
65
535
as
ID
in
this
state
and
I'm,
proposing
that
you
return
null
instead
and
then
once
the
DCF
soul
is
negotiated.
E
By
answer
the
ids
for
you,
they
channeled
upon
finding
ideas
are
signed
and
then,
in
the
rare
case
that
when
it
happens,
all
even
rod,
IDs
and
of
course,
binds
to
selected
dtls
role
are
taken.
Then
resource
engines
exception
would
be
from
for
looking
I'm,
not
not
thrown
from
the
most
description,
but
fighting
the
state.
A
channel
air
event.
I
This
seems
like
a
fairly
reasonable
way
living
we
based
in
acting
as
a
thin
shim
on
top
of
the
ATF
da
channel
stuff,
which
uses
that
that
idea
at
the
unassigned
ID
but
I
certainly
no
reason
in
the
JavaScript
layer.
We
have
to
retain
that
I'll
take
know.
Any
question
would
be.
Is
anyone
actually
cares
about
that
right
now,
probably
not.
I
D
B
A
A
A
Most
applications
would
don't
have
to
deal
with
us,
but
some
do
and
that
the
classic
case
is
sort
of
gate,
waiting
to
sit
with
the
you've
called
1
800
go
fedex
and
or
some
other
ivr,
which
plays
useful,
prompts
like
press
1
for
this
and
press
Q
for
that
as
part
of
the
ring
back
tone-
and
you
hear
that
before
they
actually
answer
and
connect-
and
this
is
all
done
to
scam
billing,
but
whatever
it
does
happen
in
yep,
she
loved
it.
So
we
have.
A
This
is
one
of
our
requirements
from
the
very
beginning
goes
back
to
this
74
78
and
the
proposal
here
is
that
the
the
default
option
with
these
types
of
stuff
that
arrives
before
you
have
the
signaling.
That
tells
you
the
fingerprint
of
the
dtls
connection.
So
you
know
you're
connected
to
the
right
person,
so
you
know
you're
receiving
media,
that
that
is
valid
media
in
some
sense.
A
Maybe
it
would
be
buffered
seat
in
those
eyes
reims
or
something
like
that,
but
that's
all
optional
and
up
up
to
the
implementation,
maybe
we'll
be
just
discarded,
but
if
the
flag
set
it
allows
this
data
to
to
actually
flow
while
you're
waiting
for
the
fingerprint
to
arrive
in
the
SDP,
so
it'd
only
be
used
by
applications
that
actually
wanted
this.
So
that's
I
think
that's
the
first
thing
to
understand
from
a
security
analysis.
Point
of
view
is
that
the
application
has
to
deliberately
say
I
want
this.
Behavior
and
I
know
what
I'm
doing
now.
A
The
it
still
needs
to
have
the
ice
fingerprints
to
connect
up
right.
So
it's
not
like
any
random
attack
or
on
the
internet
that
they
cannot
see.
The
signaling
can
spoof
this
media
and
cause
you
to
receive
it,
and
it's
also
the
application
knows
this.
Media
different
will
probably
handle
it
differently.
A
So
if
they
can
do
that,
you
know
most
scenarios
where
they
can
do
that.
They
can
also,
probably
just
you
know,
install
java
script
that
directly
plays
whatever
they
want
on
your
speaker
or
directly
connects
to
the
attacker.
Does
any
other
number
of
billion
things
so
I
find
that
the
security
the
this
introduces
a
security
argument.
You
know
super
weak
on
this
one
I,
don't
really
think
it
does.
That
said,
the
other
thing
was
done.
Sort
of
to
mitigate
that
scary
thing
is
some
people
suggest
as
well.
A
Let's
only
allow
this
to
happen
for
five
seconds
and
five
seconds
sort
of
arbitrarily.
Take
that's
long
enough
for
the
signaling,
it's
very
you
know
it
seems
like
in
most
systems
the
signaling
shouldn't
be
more
than
five
seconds
behind
the
media
and
getting
that
message
across
and
of
course,
people
can
melt
the
scenarios
where
it
is.
We
could
argue
about
whether
this
should
be
25
seconds
or
one
second,
but
it's
you
know
it's
in
that
order
of
magnitude.
Other
people
sort
of
pointed
out.
Well,
we
don't.
Why
bother
with
the
timer
at
all?
A
It
actually
really
doesn't
improve
the
security
situation,
much
at
all
and
I'm
sort
of
in
that
camp,
so
I'm
perfectly
happy
to
have
a
five
second
time
or
two
I
think
for
the
actually
use
case.
That's
driving!
Yes,
you
need
enough
time
that
you
can
raise
a
look
at
the
signaling.
So
that's
that's
roughly
what
the
pr
says.
A
I
This
is
randall,
I
generally
I'm
I
seems
reasonable.
I
mean
I,
certainly
there's
a
use
case
around
this
I.
Don't
remember
how
far
back
it
goes.
That
calls
us
back
to
the
beginning.
That's
behind
the
five-second
thing,
yeah
I,
don't
think
five
seconds
versus
longer
or
shorter.
You
know,
or
whatever
has
ads
much
security
level
here
either
doing
it
or
not,
and.
I
C
A
I
mean
I
think
in
those
environments
it
works
where
you're
connected
with
WebRTC
to
an
SBC
that
you're
connected
to
quickly.
That's
then
a
gate
waiting
on
but
trying
to
pass
the
dtls
connection
on
to
the
sip
gateway.
So,
like
look
I'm,
never
arguing.
This
isn't
basically
a
corner
case,
but
it's
unfortunately
being
able
to
have
go
fedex
work
as
an
important
corner
case.
A
H
Im
going
to
go
fedex
papers
actually
very
worthwhile,
though,
because,
like
somebody
is
to
turn
their
regional
call
center
or
whatever,
like
that,
you
know
for
the
Peugeot
fedex,
like
they're,
not
going
to
receive
that
and
coming
media
because
are
not
a
bunch,
the
biz
net
and
holes,
and
I
think
the
fundamental
question
we
should
be
considering
is
like
sure
we
could
add
this
flag.
But
every
other
stuff
adds
complexity
and
we
should
sort
of
have
this
region
decision
out
like
I
wanna
take
this
complexity,
because
we
think
that
I
topside
justified
I'm.
A
In
the
spc
you
know
the
platteville
panels
are
going
to
all
punch
because
they're
going
to
be
punched
to
the
SVC,
not
to
the
end
gateway,
which
is
cloud
I,
mean
it'll,
be
a
public
IP
address
them.
Oh,
that's!
That's
why
the
eye
stuff
is
going
to
even
come
up
before
this
right,
so
I
agree.
The
scenario
is
complicated,
like
that's.
The
only
way
to
make
it
work.
G
H
A
B
A
A
H
E
H
It
is
an
interesting
case.
I
think
you
need
to
see
the
actual
case
in
which
this
occurs
spelled
out
more
quickly
more
clearly,
because
I
think
we
now
ascertained
that
the
go
fedex
being
in
general
would
not
be
prohibited
by
this,
because
they
need
to
spend
it
to
your
answer
with
the
right
DG
on
tiger
print.
H
A
A
Don't
know
what
would
be
clear.
I
cannot
provide
a
full
architecture
of
like
Al.
The
whole
thing
like
if
I,
provide
an
example
that
works.
All
people
are
going
to
say
was
well
there'd,
be
other
ways
to
build
that
architecture.
I,
just
like,
like
I,
have
a
limited
amount
of
time
to
put
in
that's
what
I'm
not
quite
sure
what
I
should
go
do
well.
H
So
we
need
something
where
the
peer
connection
isn't
killed,
but
yet
media
somehow
arriving,
and
you
know
we
need
to
be
able
to
sort
of
still
play
that
out,
even
if
I
can
print
the
right
thing
if
it
hasn't
yet
arrived.
Okay,
that's
all,
but
you're
talking
about
its
base
where
no
answer
has
yet
been
received,
and
there
is
no
single
fingerprint
at
all
yet
and
you
need
to
do
with
you
with
media.
H
H
F
A
Through
it
really
closer
around
the
ice
happens
to
a
gateway,
and
you
get
what
you
need
from
the
Gateway
to
set
up
connections
some
connections,
not
all
connections
as
soon
as
you
can
actually
gateway.
Sorry,
as
soon
as
you
connect
the
SBC
right,
so
for
some
considering
a
cloud
device
that
doing
that
traversal
for
you
right.
So
that's
why
you?
How
you
got
the
ice
passwords
of
whatever
you're
going
to
talk
to
you
can
start
bringing
up
the
transport
connections
so.
H
A
A
joke
so
right
now
we're
just
we're
terminating
our
ice
on
the
first
endpoint.
Ok
and
then
the
dtls
is
going
to
be
relayed
in
point
from
the
first
cloud
device
right
the
spc
here,
then
we're
going
to
be
relayed
through
that
to
get
to
the
actual
gateway
which
will
do
the
dtls
will
will
be
from
the
real
a
connection
through
that.
Ok.
A
And
so
that's
why
we
end
up
with
getting
the
dtls
connections
information
you
know
later
and
you
can
try
and
avoid
all
of
this
by
decrypting,
all
of
the
not
having
in
the
end
media
secure
in
their
media.
You
know,
I
can
try
to
draw
something
more
crisper
than
that.
But
yes,.
B
E
B
G
E
E
A
H
So
in
a
32,
so
I
think
we
should
take
this.
As
you
know,
Conrad
checked
out
an
email.
You
know
how
you
sort
of
see
this
working
and
you
know
then
we
can
I
to
analyze
that
I.
Think
just
given
our
mission
here
is
closed
at
one
point
out,
we
really
want
to
be
sensitive
about,
like
nice,
Pamela
skipping
work
trying
to
do
here.
So
if
we
can
do
something,
it's
like
a
single
one
method-
Robin
roxbury
great,
but
it's
not
clear
that
would
actually
work.
A
G
A
A
You
know
when
you,
when
you
have
video
and
video
input,
that's
a
different
size
than
the
video
you're
trying
to
send
how
you
deal
with
the
scaling
and
cropping
of
that
such
that
we
don't
get
all
of
these
black
bars
on
the
tops
of
bottoms
or
letterboxing
or
whatever,
and
so
basically,
we
Center
the
video
we
scale
it
so
that
it
fully
covers.
So
there's
no,
you
know
it's
now
larger
than
what
you
need,
so
that
scale
it
so
on,
whichever
dimension
is
a
small
thing
and
then
you
crop
whatever
is
extended
over
the
edge.
A
A
H
A
H
A
This
says
when
you
do
need
scale
the
video
somehow,
but
let's,
let's
say
that
the
stp
is
demonstrated
to
640
by
480
and
your
input
source
suddenly
changed
resolution
this
set
or
or
even
on
the
initial
import
source.
But
the
name
of
the
original
bug
is
somewhat
misleading
from
what
the
actual
thing
is
being
tried
covered,
which
is
you
know
this
was
really
Martin
raised
the
issue
of
it's
not
really
clear
how
we
take
720p
video
and
put
it
into
640
by
480
sure,
okay
yeah.
That
certainly
is
everything
you
can't
change.
I
A
D
A
E
Actually,
keep
this
one
too
controversial,
so
we
decided
in
the
past
that
set
parameters
is
allowed
to
reward
you
order
these
codecs,
but
not
modify
them
or
add
them,
and
this
PR
is
just
specifying
that,
as
you
can
see,
light
up
the
text
there.
So
now
the
the
members
of
the
dictionary
can
be
more
fide
three
straight
forward
and
you
can't
remove
the
codec
and
then
add
it
back
later
and.
E
Next
slide
is
also
updating
set
codec
preferences,
because
I
made
sense
to
combine
the
two
and
I've
been
seeing.
Currently
that
specifies
what
codecs
are
allowed
in
cyclic
references.
So
now
it
says
that
they
only
contain
codecs
I
returned
by
the
sender
area
that
I
could
get
capabilities
with
the
kind
of
the
receiver
Tyga
capabilities
of
the
kind
of
the
relevant
transceiver.
E
C
So
I'm
in
the
process
of
talking
about
this
once
we
found
some
other
issues,
so
I
opened
the
new
issue
on
this.
So
there
are
examples
where
we
have
codex
with
multiple
clock
rates
so
like,
for
example,
comfort.
Noise
could
be
like
at
eight
thousand
and
sixteen
thousand
or
the
soap
code
that
can
operate
it
at
those
different
clock
rates
or
we
have
h
264
ABC
implementations
that
supports
accusation
modes
of
01
that
the
case
with
Firefox,
and
so
there
is
a
mental
module.
C
I
think
that
one
is
not
quite
as
bad,
though,
because
it
represents
kind
of
the
maximum
receiver
capability,
but
in
particular
we
have
get
capabilities
and
then
said
Kotick
preferences
and
when
you
call
get
capabilities,
the
question
is:
what
do
you
get
if
you
have?
If
you
have
an
implementation
that
supports
these
things?
Do
you
get
like,
for
example,
like
in
the
ftp?
I
believe
it's
true
for
firefox
you'll
see
the
h.264
AVC
codec
would
packetization
mode
0
and
there's
another
entry
for
packetization
mode
one
I
c'n.
C
I
think
you
would
see
an
8000
clock
rate
and
also
a
16,000.
So
the
question
is
whether
you're
going
to
get
multiple
codex
or
one
and
because
we
only
have
the
mime
type
so
far.
You
know,
if
you
had
multiple,
you
wouldn't
be
able
to
differentiate
them.
You
wouldn't
know
which
was
which
and
then
similarly
for
said,
Kotick
preferences
like
if
you
wanted
to
reorder
the
h.264
stuff
to
say
I
want
to
prefer
packetization
or
one
or
something.
How
would
you
do
that?
If
you
only
have
the
mind
type,
this.
C
C
G
B
E
Right
so
here
I
think
your
your
point
is
right
that
you
know
if
we
added
clock,
sprays
and
number
of
channels
super
ability,
then
you
get
multiple
dictionaries
can
get
capabilities
and
then
you
could
reorder
them,
but
for
the
parameters
you
know
you
wouldn't
get
a
codec
for
every
combination
of
variable
parameters
like
you
wouldn't
get
a
stereo
equal.
You
know
an
opus
with
stereo
8
peeples
one
and
that's
EPF
smtg
line,
and
what,
with
stereo
good
people's
key.
H
E
E
C
E
C
I
G
Well,
so
we
can,
we
can
separate
two
things.
One
is:
there
is
a
way
that
the
browser
extent
Elliott
application,
here's
what
I'm,
capable
of-
and
you
can
select
between
these
and
then
an
also
a
the
application-
can
jam
in
extra
codec
specific
parameters
that
are
inside
of
coded
capabilities,
necessarily
right.
C
C
A
Let
me
just
let
me
just
pawned
off
a
level
here
for
a
second.
Maybe
it
is
useful
with
just
the
mind
type
I
mean
what
your
thang,
but
the
packetization,
whether
it's
no
01
is
still
than
you
negotiated
with
the
offer
answer
by
the
browser
I
mean:
does
the
application
really
want
to
control
that,
and
it.
I
Seemed
totally
Cullens
point
is
a
good
one.
The
only
case
where
the
where
the
packers
ation
mode
would
matter
for
psychotic
preferences
would
be
if
you,
if
the
application
wanted
to
just
fiddle
with
the
preference
of
0
vs
1
in
the
negotiation,
otherwise
or
another
1i,
otherwise,
I
assume
that,
basically,
you
know
we
were
from
the
implementation
of
the
brother
side
of
things.
We
would
just
simply
consider
all
the
260
fours
to
be
a
group
and
moves
them
together
and
keep
the
ordering
of
zero
vs
one
that
the
browser
house.
I
A
I
C
A
I
A
G
C
Yeah
I
think
we
talked
about
the
tail.
Is
it
the
only
the
only
thing
that
you
need?
You
have
it's
only
the
stuff
that
makes
you
have
separate
stp
lines,
and
how
did
you
put
it
I
think
it
because
if
it
changes
the
encoding
cited
census
is
the
only
time
we
would
have
that
confident.
Oriole
explosion,
but
anyway,
I've
only
seen
a
codec
with
a
maximum
of
two
lines.
So
far,
so
good
I
mean
it's
theoretical,
but
I
don't
think
it
actually
happened.
H
E
H
I
E
I
The
guy
yeah
I
mean
I
can
see
in
return
the
rearward
situations
you
know
typically
I
could
I
would
expect
to
see
two
or
maybe
four,
perhaps
six,
you
don't
necessarily
application.
A
implementation
doesn't
necessarily
have
to
offer
all
the
combinations,
but
if
it
wants
to
support
the
different
profile
levels,
it
will
have
to
lose
support
some
version
of
each
of
those
and
you're
talking
on
the
order
of
what
you
know,
you're
talking
on
the
order
of
single
digits
generally
so.
C
E
G
C
H
B
B
B
C
G
G
C
C
C
C
C
A
C
A
C
E
C
E
E
G
C
A
A
E
E
C
E
A
E
All
right,
but
there's
we
agree,
that
is
everything
after
the
sailors,
I
I,
think
guy
named
it,
the
smtp
line
might
be
a
little
misleading.
Then,
since
mine
kind
of
implies
it's
the
whole
line,
but
I
think
we
just
need
clarify
that
and
in
the
description
attribute,
since
spp
smtp
comes
after
the
pace.
Okay
of
anyway,
those
other
I.
B
E
Fat
so
other
question,
which
I
think
it
shouldn't
be
very
controversial.
Sorry
I
actually
still
on
the
slide.
So
if
you
call
center
dot
gig
parameters,
the
ftp
different
Kiki
long
be
from
the
local
I'm.
Sorry,
the
remote
description
I
think
I
got
this
backwards.
Actually
sorry
about
that
sender
from
the
remote
description
receiver
from
the
local
description
is
nice.
Is
that
correct?
Sorry
that
was
live
servers
so.
D
G
B
E
This
is
I'm
just
proposing
a
consistent
terminology
when
errors
are
either
thrown
or
a
promise
is
rejected
with
them.
On
my
fairy
land
on
exceptions,
so
I'm
proposing,
we
just
we
link
to
something
well
I
dl,
but
subscribers
further
and
longer
use
it
just
throw
a
channel,
for
example,
and
valid
axis
air,
and
when
reject
you,
use
an
easily
created
now
that
is
their
leverage,
for
example,
reef
yeah,.
C
C
C
G
Think
it's
valuable
to
be
able
to
set
the
peak
time
without
XP
munting,
but
I
recently
learned
that
opus
and
in
particular,
live
ova
only
supports
certain
p
times.
So,
if
I
nice
to
let
application
know
which
P
times
it's
not
like
a
range,
it's
more
like
a
set
and
right
now
right
the
dopest
as
a
set
of
like
two
to
three
possible
happy
most.
I
F
A
Satellite
connection,
you
often
want
to
do
at
30,
milliseconds
or
even
higher
right
and,
if
you're
doing
really
high
quality
microphones
inside
a
Wi-Fi
conference
room,
it's
not
even
long
distance.
I
know
the
Opus.
People
wanted
to
get
like
five
milliseconds
or
some
people
even
want
two
milliseconds
or
something
as
ridiculous.
I
mean
I,
think
if
you're
doing
Wi-Fi
Mike's
inside
of
a
conference
room
or
a
studio,
or
something
like
that.
There's
a
good
argument
for
five
mil
thing
now.
I
C
G
I
Thing
as
a
time
is
advisory,
you
say:
I
would
like
okay
and
I
or
I
prefer
this
P
time.
Ok,
it
doesn't
necessarily
set
it
at
per
se,
and
so
the
list
becomes
less
less
interesting.
In
that
case,
you
can
say
two
milliseconds,
since
they
could
it
doesn't
sport,
a
tank
in
32
melting
it'll
use
whatever
at
best
it
will
do
whatever
is
closest.
Ok.
G
B
A
I
B
I
A
A
Good
yeah,
I
mean,
I
think,
I'm
going
to
say
the
same
user.
An
oil
is
just
like
look
I
think
it
seems
reasonable
to
allow
the
application
didn't
indicate
something
that
it
wishes.
The
P
time
might
be
right
and
not
a
few
times
is
sort.
Here's
my
desire
to
packetization
interval
as
a
hint
to
the
system,
and
then
the
system
chooses
something
reasonable
based
on
what
it
feels
like
doing.
That's.