►
From YouTube: WebRTC Working Group Virtual Interim May 22, 2018
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).
A
A
Dr.
Alex
and
his
team
will
provide
an
update
on
the
status
of
kite
and
interoperability
testing
and
then
we'll
try
to
make
progress
on
a
few
open
issues
in
WebRTC
and
then
our
goal
is
based
on
the
resolution
of
those
issues
and
PRS
to
recycle
web
urgency
when
I,
let's
ER
after
the
meeting
I
in
time
for
the
June
face
to
face
which
we'll
talk
about
in
a
minute.
A
So
that
meeting
is
scheduled
for
June,
19th
and
20th
at
the
Google
Stockholm
offices.
You
can,
but
you
can
find
very
easily
on
Google
Maps.
There
are
quite
a
few
hotels
nearby,
so
hopefully
everybody
who
wants
to
make
arrangements
can
we
will
be
doing
remote
participation,
which
I
think
is
very
well
supported
by
that
facility
and
google
hangouts.
We
also
have
a
will
be
meeting
at
keep
a
Camille
on
October
22nd
and
the
23rd
Monday
and
Tuesday.
So
that's
our
schedule
of
physical
meetings
and,
of
course,
we'll
continue
the
virtuals
on
a
roughly
monthly
basis.
A
So
a
little
bit
about
the
virtual
interim
I
think
you've
seen
a
post
on
the
proposed
agenda.
We're
still
working
that
through.
So,
if
you
have
things
you'd
like
to
see
addressed
or
comments
on
the
inner,
please
send
them.
But
this
is
more
or
less
the
outline
I
know.
Peters
provided
some
feedback,
but
we
would
love
to
have
that
from
everyone.
So
we're
going
to
start
off
the
discussion
with
use
cases
and
the
discussions
been
going
quite
well
on
the
mailing
list.
A
Thinking
of
them
talking
a
little
bit
about
ice
and
ways
that
that
can
be
done:
certificates
and
identity
data
transfer
and
we're
hoping
that
Leonard
and
we'll
develop
and
run
that
part
of
the
agenda
and
then
from
three
to
five.
On
the
first
day
that
will
be
turned
over
to
dr.
Alex
and
a
team
to
talk
about
testing
and
Interop,
and
then
the
second
day
we
have
Harold
talking
about
access
to
raw
media
surgery.
A
I
would
like
to
have
a
discussion
on
scalable
video
coding,
I'll
help
you
with
that,
and
then
talk
a
little
bit
more
about
sender.
Receiver.
Hopefully,
Peter
will
cover
that
and
then
the
topic
of
web
workers
we've
assigned
to
Harold,
and
then
we
have
two
open
slots,
but
we
can
assign
now
or
leave
them
open
and
assign
them
during
the
meeting
for
topics
that
need
follow-up
and
then
wrap
up.
So
that's
more
or
less.
A
The
agenda
feel
free
to
comment
on
it
or,
if
there's
other
topics
you'd
like
to
discuss,
send
them
off
to
the
man
list
and
we'll
continue.
Working
on
this
as
time
goes
on
all
right.
A
little
public
service
reminder
about
the
Charter,
the
it
is
out
for
review
AC
review
until
May
25th.
We
haven't
gotten
that
much
in
the
way
of
feedback
or
voting
on
it
yet,
but
we'd
like
to
make
sure
that
all
the
AC
reps
respond.
A
A
A
A
A
A
C
A
B
A
B
Right
thanks
Bernard,
so
basically
it's
going
to
be
on
a
date
with
all
the
latest
test
and
capacity
we
had
Vic
to
kite.
The
goal
is
twofold:
one:
to
make
you
aware
of
the
tests
we
already
running
every
day
and
what
to
to
give
everybody
the
opportunity
to
point
to
things
that
we
are
not
testing
and
we
should
so
next
line.
Bernard,
please
so
originally.
The
reason
we
aren't
come
very
quickly
is
that
we
were
trying
to
complex
if
I,
a
distinct
text
back
right.
B
B
Currently,
most
of
the
code
is
open
source
with
contribution
from
google
and
cosmo.
We're
looking
forward
to
more
people.
Jumping
in
Mozilla
gave
us
a
lot
of
pointer
to
things
that
we
should
do.
Val
goon
has
been
reaching
us
to
say.
Okay,
we
could
do
more
on
the
network
and
started
things.
We
would
you
welcome
that,
and
we
would
like
to
try
to
make
it
something
that
is
more
useful
for
people,
but
also
contributing
more
by
people
that
have
an
interesting
next
slide.
B
So
there
was
a
work
that
has
been
done
on
the
webdriver
itself,
but
if
we
get
outside
of
the
scope
of
WebRTC
but
is
very
crucial
to
actually
testing
web
RTC
cells,
Apple
has
been
doing
the
heavy
lifting
there
and
has
proposed
of
implant,
be
the
first
implementation
and
shift
it.
Maybe
UN
can
give
a
little
bit
more
details
here.
B
There
was
a
lot
of
discussion
specifically
on
the
permission
side
of
thing
that,
when
behind
getusermedia
that
involved
web
sockets
that
involve
geolocation
bluetooth
and
all
the
eleven
api's
that
are
defined
in
the
separate
permission
spec
as
far
as
I
know,
the
Apple
contribution
was
specific
to
get
user
media
only,
but
the
design
was
capable
of
and
being
extended
to
different
kind
of
permissions.
So
before
I
go
to
make
slide
UN
do
you
want
to
commit
or
to
comment
on
the
word
beauty.
D
B
Great
next
slide,
and
what
we
did
in
the
in
the
past
month
was
to
stabilize
the
browser,
console
browser
and
operating
system
configuration
we
had
in
the
path
and
add
one
we
support
most
of
the
Firefox
Chrome
and
Safari
and
edge
desktop
configuration.
There
is
including
Safari
tech
preview
and
edge
fast-track
I.
Never
remember
the
name
of
those
edge
distribution
that
our.
B
That's
the
one
burner,
the
the
one
that
give
me
most
work
today,
the
most
exciting
one.
We
also
thought
that
was
pretty
trivial
behind
the
engineering
problem
of
making
sure
to
keep
the
things
up
to
date
and
the
matching
between
the
webdriver
binaries,
the
webdriver
bindings
and
and
and
the
browser
binary
themselves
for
people
that
are
interested
in
in
the
background,
there's
all
sorts
of
limitation
in
a
number
of
copies
or
forgiving
browser
you
can
run
with
webdriver,
namely
edge
and
Safari
or
limited
to
one
per
operating
system.
B
So
when
you
do
an
interoperability
testing,
you
cannot
always
parallel
lines.
You
test
against
Safari
or
edge
unless
you
actually
have
different
physical
machines,
which
led
to
people
that
follow
my
tweets
in
pictures
of
me
buying
20
back,
you
need
to
be
able
to
run
backing
in
less
than
14
hours
a
day
so
one
month
ago,
the
oldest
test
or
the
interoperability
test
we're
reaching
the
bar
of
14
hours
a
day.
What
we
have
in
internal
configuration
that
was
a
little
bit
more
difficult
where
the
mobile
and
the
electron
configuration.
So
we.
E
B
Support
for
Chrome
and
Firefox
on
Android,
as
well
as
Safari,
on
iOS
as
part
of
the
testing
matrix
to
to
the
daily
runs,
and
last
week
we
added
support
for
electron
as
well,
since
a
lot
of
people
wanted
to
to
do.
That
is
not
shown.
Electron
results
are
not
shown
on
webrtc.org,
but
they
shown
on
or
public
Cosmo
dashboard
and
people
can
go
there
and
practice
next
slide.
B
One
of
the
the
biggest
problem
was
on
most
of
the
tests,
we're
supposing
and
given
back
end
and
one.
This
is
a
normal
approach
where
see
poor
flash,
for
example,
where
the
signaling
is
also
well-defined.
This
is
a
little
bit
more
challenging
with
web
RTC,
where
anybody
can
determine
their
own
signaling
and
you
still
need
a
complete
application
with
a
minima,
signaling
server
to
be
able
to
handle
the
handshake
and
the
discovery.
B
So
that
was
becoming
a
problem,
so
we
did
a
design
where
our
kite
does
not
suppose
anything
about
the
application
and
the
text
of
the
system
and
the
test.
So
you
can
run
any
kind
of
client
again
different
back-end
protection
on
production
stage
in
development.
You
can
run
it
locally
and
so
on.
All
these
cases
are
taken
care
of
next
next
slide,
so
that
that's
an
overview
right.
You
have
a
JSON
configuration
that
they
are
like
from
64
in
Windows,
10,
trees
or,
like
ovens,
browser
configuration.
B
You
have
an
engine
that
is
going
to
distribute
that
on
on
a
selenium
grid.
The
nodes
of
the
selenium
grid
are
to
be
going
to
run
the
test
against
the
system
and
the
test.
In
that
case,
the
system
and
the
test
was
to
see
video
glitches,
a
few
4u
serration
on
me
and
then
will
collect
and
aggregate
the
results
to
be
send
backs,
but
whatever
dashboard
or
common
line
you
you
want
to
see
it
right.
B
So,
by
separating
using
this
orange
line,
the
configuration,
the
dashboard,
the
grid
and
the
system
and
the
test
we
were
able
to
actually
track
the
problem
that
otherwise
was
becoming
difficult
to
you
to
handle
next
slide
please.
So
this
is
an
example
of
one
application
of
guide
to
the
number
of
browser
equal
one
use
case.
B
So
we
took
the
WPT
that
was
run
on
the
right
hand,
side
on
four
different
configuration
from
time
to
time,
and
we
show
that
we
could
run
it
on
the
18
configuration
that
kind
support
automatically
on
a
daily
basis,
and
that's
it.
We
took
one
day
of
work
on
me,
so
that
shows
all
you
can
take
an
existing
test
suite
and
basically,
you
just
have
a
little
bit
of
glue
to
make
to
make
it
run
on
top
of
guide,
which
is
going
to
separate
the
problem
of
running
the
test
and
having
the
browser
configuration.
B
B
So
this
is
another
example
of
support
for
the
start
integration.
So
if
you
look
at
the
lower
right-hand
drawing
with
the
wheel,
that's
the
result
of
all
the
one-to-one
tests
and
if
you
take
a
specific
just,
you
know
ice
connection
test.
You
know,
I'm,
just
checking
that
the
two
browser
can
connect
by
checking
that
is
connection
status
is
changing
to
complete
it
if
it's
chance
to
complete
it
on
both
rather
than
the
test,
is
okay.
Otherwise,
it's
fail.
B
I
can
take
a
specific
line,
so
you
see
the
one
that
is
chosen
and
his
game,
which
is
Android
with
chrome,
65,
against
Windows,
10,
age
16
and
if
I
click
on
the
left
hand,
icon.
That
looks
like
like
a
curve
right,
I
end
up
into
the
screen.
You
see
on
the
Left,
which
is
the
parsing
of
all
the
stats,
all
the
sdp
all
the
over
answer.
B
So
this
is
an
example
of
TTL.
So
this
is
an
example
of
a
case
where
you
have
three
browsers
participating
just
to
show
that
Mohamed
kite
can
handle
more
than
one
to
one
code,
given
the
right
system
and
the
test,
but
also
the
reason
why
we
tweaked
the
organ
ring
visualization
instead
of
the
grid,
is
to
be
able
to
accommodate
more
onion
rings
right,
more
and
more
browsers
next
slide.
B
So
the
really
new
thing
we
started
the
past
two
month.
What
testing
for
multi-party
then?
The
unified
plan,
multi
stream,
and
signal
cast
right
so
multi-party.
We
use
the
GT
test
so
before
there
is
no
limit
to
the
number
of
browsers
you
can
actually
use
in
type.
There
are
some
limiting
the
browsers
and
the
on
the
meeting
in
the
back
end.
B
I
will
not
tell
you
what
it
is
because
I'm
sure
everybody
want
to
have
the
joy
to
drag
that
thing
down
by
himself
and
see
how
much
how
many
peerconnection
day
they
can
rate
multi-stream
was
not
too
difficult
to
go.
There.
We
decided
to
test
against
unified
plan
and
not
to
test
against
plan.
D
and
Google
is
currently
using
these
to
help
fasten
the
support
of
her
unified
plan
a
little
bit
more
in
next
slide.
The
simulcast
was
a
little
bit
more.
No,
no
wait,
wait.
Wait.
B
Simulcast
was
a
little
bit
more
difficult
in
the
sense
that
in
web
RTC
don't
want
zero
simulcasting
only
define
on
the
satisfied,
so
you
need
to
send
from
a
browser
to
an
exception,
and
the
logic
was
switching
between
the
layers
is
not
standardized
not
defining
away.
So
we
reach
out
to
DC
Janice
current
Otis,
a
few
of
the
web
RTC
media
server
to
see
all
they
did.
That
and
the
logic
the
implement
is
actually
very
use
case
dependent
right
some
people,
let
the
receiver
client
decide
when
to
switch
the
layer.
B
Some
people,
let
the
SFU
until
we
decide
when
to
do
that
and
the
triggers
the
thresholding
self
or
are
different
from
one
place
to
the
other.
So
what
we
decided
to
implement
and
to
make
available
was
some
kind
of
a
loopback
test
where
the
sender
application
is
sending
signal,
a
stream
to
the
to
an
SFU
which
returned
one
of
the
one
of
the
layer
and
the
application
can
then
choose
between
the
layer
and
check
that
you
can
switch
between
one
and
the
other
validating
that.
B
Implicitly,
you
receiving
the
good
stream
information
about
which
one
is
I
resolution
and
low
resolution.
Then
you
can
switch
between
them,
so
I
know
it's
only
implemented
against
the
leaders
SFU,
because
we
were
more
familiar
with
the
code,
but
we
plan
to
extend
it
to
at
least
Janice
and
current
oak,
and
anybody
that
actually
want
to
have
a
test
for
their
own
SFU
is
welcome
to
contribute
as
long
as
they
make.
This
was
called
available.
So
the
test
is
10.
Long
can
be
run
so
next
line
specific
on
multi-stream.
B
It's
running
locally,
we
tried,
both
with
and
without
adapted
OTS,
to
see
if
attitude
OG
at
exactly
healthy,
taking
care
of
the
unify
plan
for
browsers
that
do
not
support
it
natively
and
we.
What
we
test
exactly
is
that
the
SDP
that
are
generated
are
compliant
with
the
with
the
format
on
the
set
aside
that,
on
the
SFU
side,
the
IDS
on
the
wire
are
marking
the
IDs
that
we
receive
on
the
signaling
path
and
that
eventually
the
media
is
flowing
right.
B
So
it's
not
I
mean
we
could
test
a
lot
of
things,
but
we
decided
that
these
were
free
condition
that
must
be
tested
and
would
be
enough
to
actually
have
a
good
quality.
We
expect
people
to
report
a
problem
if
the
test
is
not
sufficient
for
given
use
case
and
to
be
able
to
extend
the
case
in
the
future.
B
Sdp
format
is
not
exactly
up
to
date.
We
didn't,
we
didn't
change
it
to
a
lean
and
mean
compliance
mode.
So
we
for
us,
the
Firefox
testes
is
working
and
we
have
communicated
to
the
Mozilla
team
that
point
in
the
in
the
past.
Two
month,
I
think
we
discovered
30
or
40
bugs
you
know
all
the
browser
across
including
webdriver
implementation
and
so
on,
and
we
have
filled
up
ticket
with
all
the
group
as
we
were
going.
B
We
are
allowed
to
nodes
and
infrastructure
testing
mode
as
opposed
to
the
interoperability
testing
mode
we
had
so
far
and
the
ID
was
to
add
a
cron
light
time.
Frequency
notification
in
a
configuration
file
so
far.
The
most
interesting
is
the
bottom
one.
If
you
want
to
add
one
from
53
on
Windows
10,
every
20
seconds
to
connect
against
your
server.
B
B
The
the
biggest
problem
of
guide
so
far
was
how
to
set
up
a
grid.
We
set
up
the
grid
manually
and
we
maintaining.
We
were
setting
it
up
manually
and
we
were
maintaining
it
manually,
which
was
proven
to
be
difficult
and
very
time-consuming,
and
when
kite,
when
sky,
to
assume
that
the
grid
is
available
with
all
the
machine,
it
will
ever
mean.
So
that
was
not
very
user
friendly.
You
know,
if
you
had
to
reproduce
and
manually
set
that
they
give
us
help.
There
is
a
lot
of
bad
magic.
B
You
had
to
know,
and
that
was
just
not
practical
and
for
load
testing.
You
will
not
going
to
keep
1
million
not
wrong
to
do
the
log
testing,
so
we
need
to
have
something
a
bit
smarter
to
not
only
help
set
up
the
selenium
grid,
specifically
for
we
do
about
assisting
for
people
that
never
done
that
before,
but
also
to
automatically
on
the
flying
on
in
and
add
nodes.
When
you
were
doing
load
testing,
one
thing
that
we
wanted
to
do
and
we
actually
trying
to
collaborate
with
with
cost
at
Ohio.
B
There
is
to
have
network
instrumentation
to
be
able
to
test
the
algorithm
that
manage
the
bandwidth
estimation,
the
congestion
control,
the
bitrate
adaptation
and
autism
coupling
between
the
encoder
and
the
network
that
is
done
through
RTP
and
rtcp.
That's
for
the
bandwidth
situation
of
network
quality,
sister
packet
loss
and
things
like
that
Rhino
or
our
tests
do
not
allow
us
to
to
check
or
verify
the
ranking,
a
behavior
or
any
of
these
things.
B
B
We
added
a
lot
of
tests
to
be
able
to
test
the
latest
corner
of
the
web.
Rtc
1.0
specification,
namely
multi-party
multi
stream
and
civil,
casts
I.
Think
at
this
stage
we
have
tests
for
almost
everything
and
we
have
ongoing
development
to
make
skies
more
usable
for
everybody
that
we
don't
have
a
PhD
or
didn't
read
the
20
specs
in
the
last
week.
So
here
you
go.
B
F
B
D
Alex
just
to
know
that
WT
is
trying
nowadays
to
check
whether,
when
there's
an
issue,
whether
it's
shared
by
all
browsers
or
whether
it
just
an
issue
with
one
browser
and
each
way
they
were
trying
to
prioritize
in
top
issues.
Freedom
on
that
are
you
able
to
detect
issues
that
will
be
solely
targeted
in
one
browser
or
much
browser,
or
things
like
that
is
too
difficult
to
do
that
now,.
D
B
B
B
G
B
Something
failing
we
extract
it.
We
have
a
daily
analyzes
of
the
bird.
We
have
one
guy
through
part
that
there's
only
that
and
his
job
is
to
add
to
the
pinpoint
where
the
the
mismatch
is.
So
once
we
have
detective
beliefs
match,
because
that's
what
Interop
is
we
go
back
to
read
the
spec
and
we
check
which
one
is
not
doing.
What
is
in
the
spec.
So
now
is
the
one
running
the
test
and
he
analyzes
so
RS.
B
We
watch
all
the
tests
for
the
specs
and
know
the
the
specs
a
little
bit
better
is
the
one
that
is
checking
on
which
side
we
should
write
a
ticket.
So
eventually
we
try
to
connect
with
the
team
first
to
validate
with
them
and
ask
them
if
they
prefer
us
to
put
a
ticket
or
we
not
do
it
between
us.
Oh.
A
G
B
B
A
B
The
the
server
we
used
can
support
vp9
SBC
against
the
current
implementation
in
Chrome,
so
we
can
switching
the
layer
has
been
implementing
as
well,
so
we
could
repurpose
the
single
caps
test.
We
have
to
make
an
SDC
test,
but
the
thing
is
there
is
only
one
implementation
out
there
right,
so
we
thought
we
would
wait
until
someone
come
up
and
say
you
know:
I
have
another
use
case
with.
B
F
B
A
Thank
you.
Ok,
any
other
questions
for
dr.
Alex
all
right.
Well,
thank
you
very
much
and
just
as
a
reminder,
we
will
be
covering
testing
and
Interop
at
the
physical
face-to-face
in
june
and
stockholm,
so
that
would
be
another
opportunity
for
people
to
ask
questions
and
about
and
do
with
that
thing.
Thank
you.
A
C
E
You
hear
me
yeah
now,
I
was
a
bit
like
I
miss
setting
an
alarm
button.
Okay,
18:34
should
I
go
ahead
with
them
all
go
ahead.
Okay,
so
the
bug
is
about
checking
the
structure
that
we
feed
in
to
set
parameters.
We
currently
do
some
testing
of
this
scale
resolution
down
by.
So,
if
someone
is
at
entering
a
number,
that's
the
one.
We
throw
range
error,
but
there's
some
there's
a
book
about
two
more
checking
in
this.
In
this
function,
and
and
looking
at,
we
have
the
bug
that
what
we
have
is
like.
E
G
F
E
F
E
E
E
G
F
E
G
E
Mean
I
mean
these:
these
are
members
of
the
dictionary
spread
to
set
parameters
so
for
this
fiscal
of
this
issue
it's
only
set
parameters,
but
I,
don't
know
how
food
fence
is
good
course,
I
mean
it's.
The
option
is
that,
with
that,
we
let's
fell
on
set
some
something
a
max
bitrate,
and
we
need
to
explain
that.
Okay,
you're,
just
this-
is
just
a
hint,
but
the
word
max
kind
of
an
appointment
in
direction
that
this
it
will
be
not
more
nervous,
so
it
it
will
fail
somehow
otherwise.
E
G
I'm
I'm
all
for
adding
checks
on
a
pm-
specific
as
possible.
I
know
most
all
for
using
web
IDL
whenever
possible.
To
describe
that
and
I
know
there
we
go
extended.
Attributes
like
enforce
range
that
we
could
use
I,
don't
know
if
they
apply
to
dictionaries.
E
Thing
with
pea
time
is
like,
but
that
is
actually
specified
in
promise
to
to
respect
max
tea
time.
Mm-Hmm,
so
I
think
I,
don't
know
if
we
need
to
do
anything
thing
there,
but
number
of
it
should
should
allocate
invest
some
time
in
trying
to
investigate
in
the
lower
boundary
some
of
this
max
max
for
this
yeah.
E
C
A
F
F
F
The
current
text
seems
to
indicate
that
it
should
be
empty
and
the
and,
if,
if
it
isn't
and
and
that
can
that
seems
a
bit
odd,
I
mean
you
can
use
a
if
it
has
that
this
is
these
are
the
codex
you
can
expect
to
be
expected,
use
because
I
don't
know
they're
going
to
work,
then
it's
reasonable
to
have
them
empty
before
negotiation.
If
it's
differ,
the
codex
you
might
want
to
want
to
set.
F
That's
that
doesn't
make
sense,
and
if
you
want
to
use
get
permanence,
then
set
parameters
to
indicate
you
Steve
codex
and
doesn't
that
doesn't
make
sense.
But
on
the
other
hand,
you
don't
know
the
the
panel
pipes
for
the
codex
before
you
run
decoration,
which
is
one
of
the
parameters
so
now
I
found
I
asked
the
question
we
should.
We
should
at
least
have
have
a
clear
idea
what
we
want
to
do.
A
C
E
E
A
A
In
fact,
the
only
thing
in
it
is
the
rid
attribute,
and
then
it
states
in
5.3
that
so
the
currency
encodings,
with
the
rid
popular
populated
based
on
the
current
remote
description
and
whether
it
says
whether
the
media
is
actively
being
decoded
or
not,
and
then
you
also
have
header
extensions
and
codecs,
but
the
issue
is
that
you
look
at
the
encoding
for
the
receiver
side.
The
active
attribute,
isn't
there
anymore?
A
We
removed
it
because
we
didn't
think
it
had
any
use.
So
I
don't
know
what
the
text
is
for
and
my
proposal
would
be
to
delete
it
because
I
don't
understand
why
it's
there.
The
other
question
is
similar.
The
one
Harold
just
asked
about
sender
docket
parameters,
which
is
in
this
case
of
the
receiver,
is
the
header
extension
that
codex
populated
prior
to
completion
of
negotiation
or
ugly
after
in
this
case.
B
A
A
That's
the
next
one,
Harold
I,
don't
understand
why
the
writ
is
even
in
there,
but
we'll
get
to
that
in
the
next
issue.
The
answer
is,
you
wouldn't
know
what
hridya
until
you
have
the
remote
description,
but
I
don't
even
know
why
we
care
about
the
writ.
Even
then
so
we'll
talk
more
about
that
in
a
minute.
A
But
does
this
so
you
know:
Harold
talked
about
clarifying
the
text
for
the
sender's
side.
I
guess
in
this
case
was
a
clarification
probably
needed.
Is
it
needed?
Do
you
think,
and
also
the
other
question,
should
we
get
rid
of
the
red
text
or
not
I,
don't
think
it
matters
whether
the
media
effectively
being
decoded
and
I,
don't
even
know
how
to
indicate
that
I
think.
A
C
G
G
A
A
A
So
I
think
for
the
notes,
I
think
the
general
direction
proposed
in
1840
seems
amenable
and
we
should
try
to
develop
a
payoff
for
that.
Okay,
so
here's
the
next
one
1852.
So
this
is
where
we
get
into
the
reception
of
simulcast.
So
in
Section
5.1
it
says
that
we
don't
find
how
to
configure
create,
offer
to
receive
simulcast
but
conceivably
an
SF.
You
could
send
an
offer
to
a
browser
with
some.
A
You
know
this
if
you're
going
to
send
some
casts
to
the
browser
and
then
when
you
call
set
remotes
description,
those
that
would
somehow
be
reflected
in
what
you
get
out
of
receiver
dot
get
parameters.
That's
what
the
text
says,
but
there's
a
couple
of
questions
about
this.
One
is
whether
we
expect
any
browser
to
actually
implement
this.
So
do
we
really
expect
browsers
to
receive
simulcast
and
do
something
to
make
that
happen?
A
A
G
A
I
think
it's
the
use
case
in
Section
five
point,
one,
which
is
a
brown
SFU,
sending
an
offer
of
simulcast
to
a
browser.
I
mean
we're
still
not
I
mean
assuming
that
I
mean
there's
separable
questions.
One
is
whether
we
should
have
that
language
in
there
at
all,
but
I
think
Cullen
wanted
to
have
it
in
there
just
in
case
somebody
wanted
to
implement
it.
For
the
first
question
is:
is
anybody
actually
going
to
do
what
section
5.1
says
I'm
so.
F
The
nice
thing
about
this
is
that
you
can
discover
that
someone
sending
your
simulcast
that's
the
own.
If
you,
if
you
support
receiving
a
simulcast,
then
the
only
way
you
can
discover
whether
someone
is
sending
your
simulcast
or
not
just
by
by
looking
at
the
receive,
apart
from
math,
VP
parsing,
of
course,
if
by
looking
at
the
receivers
and
decoding
parameters
right.
G
A
F
G
F
A
A
Having
implemented
reception
of
simulcast,
the
first
thing
I
would
do
if
I
knew
I
was
receiving.
It
would
be
to
try
to
figure
out
the
stats,
because,
in
our
experience
it's
incredibly
problematic
and
very
bad
things
happen
when
you
do
that,
so
you're
very
likely
to
see
all
kinds
of
like
horrible
things
happen
like
lost
packets
delays,
bla
bla,
bla,
bla,
bla,
so
I.
You
know
I'd
question
whether
we
have
all
the
stats
necessary
to
diagnose
something
like
this.
It's
not
something
you
do
casually.
F
F
F
A
I
mean,
among
other
things,
right
if
you
don't
have
the
right
ntp
alignment,
because
what
will
happen
is
you'll
have
different
sequence,
numbers
that
you'll
have
to
align
internally
within
your
RGB
stack.
So
a
lot
of
really
ugly
things
can
happen
like
you
can
get
confused
about,
ruled
around
and
put
things
in
the
typewriter.
So.
G
What
I
heard
is
that
the
use
case
end
is
for
receiving
s-parameters
is
to
know
that
you
got
an
offer
with
final
constant
so
that
you
can
check
encoding
length
and
then
throw
an
error.
But
if
that
breaks
the
size
question
of
it
is
receiver,
get
parameters
will
not
return.
The
most
current
thing,
if
I
in
have
remote
offer.
A
G
A
A
F
A
F
A
G
A
I
was
assuming
that
you
would
get
something
just
after
SLV.
You
would
get
some
portions
of
this
back.
The
riffing
is
unique
right
because
it's
you're
only
going
to
get
this
is
only
going
to
be
turned
on
if
you
had
it
in
the
offer
and
called
set
remote
description,
so
that
portion
would
only
show
up
then
well,
if
you
know
what
it.
A
I
guess
my
bigger
question
is
my
first
question
is:
why
do
we?
Why
do
we
even
have
I
mean
based
on?
Should
we
keep
this
rid
thing
in
there
I
mean
Harold's
suggestion.
Was
it
tells
you
if
your
browser
supports
reception
of
simulcast,
but
do
we
have
any
browsers
that
are
actually
intending
to
implement
that.
G
A
G
A
A
G
G
A
A
G
Would
tell
me
if
you
received
simulcast,
don't
know
if
it's
a
remote
offer
at
simulcast
in
it.
G
A
F
G
F
F
A
C
Right
so,
basically,
just
due
to
limitations
of
the
bundle
spec,
an
answer
can't
reject
the
offer
bundle
tag
section
I
was
the
one
that
appears
first
in
the
bundle
group,
if
you
have
a
pre-existing
bundle
group
like
my
previous
offer
answer
or
if
the
offer
was
using
maple
bundle.
Only
so
what
happens?
C
If
you
you
know
if,
in
between
the
offer
and
answer
the
answer,
calls
transceiver
dot,
stop
and
then
create
answer,
and
should
it
reject
both
or
you
know,
all
bundle
them
sessions
is
because
it
has
to
which
that
may
be
surprising,
because
you
stop
one
transceiver,
but
then
all
of
them
in
just
off,
but
wasn't
what
you
wanted,
or
should
it
create
an
answer
with
that
M
sections
still
not
rejected,
with
the
direction
of
an
active
instead
and
then
rejected
in
a
subsequent
offer,
they'll
be
triggered
by
on
negotiation
needed
firing
right
after
cyber
both
description,
I.
F
F
G
C
F
F
G
G
C
C
C
G
F
F
A
A
F
F
How
many
is
encoding
is
chemist
and-
and
it
turns
turned
out
when
I
never
pointed
out
that
you,
when
I
read
already
fixed,
forbids
for
and
for
most
cases
that
it?
It
is
used
to
only
one
interesting
case,
which
was,
and
should
it
be
post
to
go
from
being
a
simulcast
Thunder
and
with
big
being
long
simulcast.
That
is
standing
only
one
encoding
to
having
multiple
encodings
after
you
create
the
transceiver
or
does
the
creating
the
transceiver
lock
in
the
number
of
simulcast
encodings
and.
F
What
I,
what
I
suggested
was
that
in
in
the
in
the
bang,
was
that
we
could
could
let
the
let
the
number
of
simulcast
encodings
be
set
up
to
the
maximum
supported
by
the
implementation
and
the
current
state
of
negotiation
by
saying
that,
but
the
capabilities
would
return
the
maximum
number
of
supporters
and
colleagues
at
all
times,
but
only
activate
one
of
them.
Now,
if
we're.
F
If
the
group
is
of
the
opinion
that
we
should
look
in
the
weather,
whether
we
are
sending
simulcast
or
unicast
at
the
beginning
and
create
the
transceiver,
then
I
should
just
kill
this
burg
and
say
we
find
with
a
we're
fine
with
the
text
process
so
working
group.
What
do
we
want
and
when
do
we
want
it.