►
From YouTube: WebRTC Virtual Interim 25 Jan 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
B
B
B
E
E
A
Bernard
you're
flipping
through
the
slides,
where
are
the
Welcome
range,
so
welcome
to
this
meeting
and,
as
usual,
we
hope
to
to
make
progress
on
both
the
media
capture
main
and
the
web.
Motogp
see
space
and
that
is
toast
will
get
key,
get
the
toaster
updated.
It
is
a
soft
after
the
meeting,
maybe
not
this
week,
but
pretty
soon
anyway.
F
G
Right
so
this
pure
just
attempt
to
step
fly
when
the
HTTP
transport
object
is
set
and
created
on
the
peer
connection
and
what
is
in
the
PR
is
that
is
created
when
applying
a
local
description
and
it's
destroyed
them
in
a
data.
Section
is
rejected
and
that
could
be
either
viable
for
description
in
the
answer
and
ended
on
a
roll
that,
if
example,
actually
the
may
not
be
issued
with
rollback.
G
Oh
nom,
if
you
know
where
to
send
offer
a
local
author
that
created
an
STD,
be
transport
and
then
rolled
it
back
then
now
it's
said
back
and
all
the
question
that
came
up
in
the
poll
request
is
whether
it
should
be
created
when
applying
a
local
description
or
just
went
to
find
the
answer,
because
there's
a
really
isn't
and
Association
until
there's
a
local
and
remote
port
and
also
is
there's
a
mass
message.
Size
attribute
and
tell
you
how
to
remote
description.
F
G
F
Yeah
I
guess
it
matters
what
them
at
max
message
size,
educators.
Therefore,
as
you
said,
if
it
represents
what's
actually
applied,
then
it
wouldn't
exist
until
you
set
the
remote
description.
You
know
what
you
see.
It's
like
a
it
is
both
a
capability
in
the
actual
applied
value.
So
this
kind
of
22
different
values
form.
F
I
So
the
sounds
like
the
question
is
whether
having
be
there,
anything
is
equal
before
it's
in
the
kind
of
permanent
state
rate
right,
because,
if
that's
not
useful
than
we
can
just
say
it's
not
there,
and
if
it
is
useful
than
we
need
to
figure
out
what
intermediary
state
is
useful.
When
it's
not
and
it's
both
date,
yeah
motherwell.
H
F
F
C
C
C
J
Hi
this
is
John
apart.
I
want
to
clarify
a
little
part
where
I
disagree
a
little
bit
that
this
would
include
everything
basically
I
think,
let's
just
back
up
a
little
historically,
I
think
we
added
tracks
of
selectors,
because
we
didn't
have
senders
and
receivers
at
that
time.
So,
but
as
far
as
and
as
far
as
filtering
filtering
out
exactly
what
you
need
in
javascript
is
super
trivial.
So
there
has
to
be
another
benefit
to
filtering
in
the
API
itself,
all
of
them
convenience
and
I.
J
J
It
also
make
sense
from
that
point
of
view
that
you
interested
in
the
video
or
you're
interested
in
the
audio.
If
you
have
one
of
each
mine,
a
multiple
video
tracks
with
a
multiple
audio
track,
I
think
that
makes
sense,
and
it
like
makes
it
easier
to
write
certain
JavaScript.
We
don't
have
to
iterate
on
stats
because
by
filtering
down
you
know
that
you
have
you,
don't
have
multiple
certain
objects.
You
have
singular
of
certain
objects.
J
Why
should
back
up
and
say
I?
Think
in
the
staff
structures
in
the
dictionaries
we
have
these
anything
to
insert
ID
is
a
foreign
key
reference,
so
to
speak
in
that
it's
a
reference
that
you
can
then
access.
Another
stats
record
in
that
same
snowball,
you
got,
and
so
in
a
sense,
I
argued
that
we
should,
if
there's
a
reference,
we
don't
want
to
leave
any
broken
references
in
data
we
return,
which
means,
I
would
argue
for
that.
J
We
return,
what's
through
the
relevant,
to
a
specific
track
of
video
track
and
every
any
data
that
it
references
and
I.
Don't
believe
that
includes
everything,
because
I
still
think
you
eliminate
everything
that
had
to
do
with
the
audio
in
a
simple
video
audio
snowball
and
in
fact,
I
think
we
could
specify
simple
recursive
algorithm
that
did
that
by
actually
starting
with
the
RTP
sender
or
receiver
object
for
a
given
track,
which
kind
of
makes
this
sense,
because
we
probably
should
have
made
senders
and
receivers
lectures
in
the
first
place.
J
So,
but
regardless
of
whether
we
do
that
now,
I
still
think
we
could
say
the
algorithm
would
be
forgiven
track,
find
its
sender
receiver
already,
specifically
rqp
media
stream
steps
or
outbound.
Whatever
the
outbound.
You
know,
in
phone
or
up
on
RTP
steps
include
that
for
that
track
in
every
reference
which
would
automatically
pick
up
the
immediate
stream
tracks
net
and
as
well
as
all
the
transports
for
them
for
the
track.
E
Because
this
is
one
I
just
want
to
add
my
two
cents
to
this
I.
Think
one
of
the
reasons
why
people
use
the
staff
selector.
It
was
specifically
figure
out
what
the
audio
and
the
video
track
specifically
might
be
doing
for
their
small
widget
or
similar
thing,
and
like
just
returning
the
media
stream.
Sacks
and
the
inbound
outbound
would
be
sufficient.
I
don't
know
I'm
not.
Your
team
use
cases
for
like
returning.
E
The
whole
block
wording
like
include
the
transport
and
such
so
I
would
caution
against
returning
everything
recursively,
because
then,
like
I,
am
with
howls
on
that
that
you
would
rather
just
to
get
stats
and
like
parse,
that
data
and
to
do
it
trivoli
from
there.
Some
people
will
not
do
not
want
to
do
that
and
they
want
to
just
like
get
some
settles.
E
That,
then,
might
just
be
in
the
past,
because
the
two
objects,
the
power
box,
ABI
and
Google,
or
the
chrome
stats
API-
were
returning
different
kind
of
objects
and
structures,
and
they
did
not
want
to
parse
all
of
that
and
figured
out
how
to
butte.
You
did
everything
into
like
a
standard
compliant
thing.
So
I
believe.
A
E
C
J
C
C
C
J
C
J
Sorry
I
think
we
should
follow
all
references,
I
don't
want
to
assign
up
or
down,
but
maybe
I
my
spender
still
URL.
What
I
meant
is
if
something
points
to
a
track.
It
doesn't
necessarily
mean
wishing
that
means.
We
should
not
include
it.
So,
whatever
data
you
find
in
the
returned
object,
if
there's
a
reference
to
something
a
better
point
to
real
data.
That's
that's
my
dividing
line
in
my
in
my
in
my
day
of
it
I'd.
C
C
C
F
Well,
I
I
have
a
common
Harold,
which
is
just
that
I
think
it's
important
that
we
not
have
a.
We
have
an
ability
to
get
just
a
specific,
more
or
less
of
specific
things
that
you
list,
because
I
think
this
will
be
called.
It
could
be
called
frequently
just
to
get
an
idea
of
you
know
overall
quality
and
packet
loss
and
stuff.
So
you
don't
want
to
pull
on
everything.
G
G
J
C
J
Because
you
know
there's,
if
you
have
a
track
as
a
track,
it
has
become
a
squirrely.
That's
the
word
selector
in
that
you
can
of
course
add
and
remove
a
track
from
career
connection.
You
can
have
the
same
track.
The
multiple
senders,
so
filtering
stats
on
I
track
can
give
you
some
surprising
results
in
educators
and
there's
some
I
know.
J
C
Not
the
condition
it's
it's
a
new
SKU
energy
extraction
and
operational
testing.
E
Because
I
guess
we
yes
I,
think
the
current
issue
was
about
black
person
and
I.
Wonder
if,
like
we're
going
to
move
forward
with
that
or
Gianna
you're
going
to
propose
an
alternative
because
they're
good
questions
and
we
need
to
like
perhaps
move
on
this
one
as
well.
In
addition
to
maybe
defenders,
I.
J
C
F
Yeah,
I
think
if
we
were
designing
this
from
scratch
today,
we
would
definitely
d
senders
and
receivers,
because
again
the
data
only
makes
sense
for
peer
connections
or
co
connection,
related
objects,
right
and
a
track
is
not
a
track,
doesn't
have
anything
to
do
with
the
peer
connection,
a
sender
and
receiver.
You
know
represents
the
track
being
sent
over
peer
connection
to
received
over
a
peer
connection.
So
those
are
the
right
objects,
so
I
would
be
in
favor
of
deprecating
the
track
as
well.
A
A
I
So
we've
had
this
argument
many
many
times
about
whether
when
you
do
a
reoffer,
it
should
include
more
codex
than
was
negotiated
and
so
I
thought
that
might
be
good
idea
to
just
let
the
application
decide
and
so
here's
an
example
of
how
it
works.
It
just
do
the
classic
great
offer
it
doesn't
preoperative
codec,
but
if
we
had
this
new
Rio
a
codex
offer
option,
then
if
you
pass
that
in
the
truth,
then
you
get
the
Codex
we
offered
am
is
a
book.
I
G
G
F
Yeah
Peter
from
my
only
my
only
question,
which
I
asked
virtually
every
time
now
we
have
any
kind
of
new
feature
that
goes
in
is:
are
you
sure
it
is
as
simple?
As
you
say,
it
is
now
because
almost
every
simple
feature,
we've
added,
has
ended
up
dramatically
more
complex
later,
and
you
know
we're
trying
to
reduce
the
number
of
things
we
need
to
repair
fix,
adjust,
etcetera.
I
G
K
Thank
you,
I
think.
If
you,
someone
should
think
about
the
other
option,
things
and
codecs
in
the
list,
like
you
mentioned
extensions
and
so
forth,
and
think
if
there's
any
possible
reason
to
reoffer
them.
If
they,
if
there
is
then
a
syntax
more
like
reoffer,
you
know
braids
codex
column,
true
extensions,
calm
tree
or
whatever
would
probably
have
more
appropriate
syntax,
but
and
more
general.
If
you
were
certain
that
there's
never
any
reason
to
go
beyond
that,
you
could
stick
with
this.
K
A
C
C
A
A
A
F
C
C
C
Spec-Wise,
we
can
do
things
well,
do
one
of
three
things:
do
nothing
which
leaves
them
out
in
the
dark
because
add
them
back
in
as
first
order
features
that
we
are
going
to
support
forever
poor.
We
could
get
back
into
the
compatibility
section
as
well.
Okay,
we
had
knit
that
these
things
exist,
but
we
don't
clear
he
liked
them.
So
I
prefer
to
do
not
do
nothing
because
keys
where
East
work
and
gets
us
to
trickiness
back,
but
Phillip
disagrees
with
me
and
what
we
always
think
I.
J
J
C
J
K
C
K
G
G
C
C
C
J
F
F
F
For
example,
if
you
have
transceiver
that
stopped
set
to
true
either
because
you
called
stop
or
because
you
add
another,
the
other
remote
peer
called
salt,
does
it
send
a
by?
Do
you
send
a
by
and
other
circumstances
such
as
calling
her
place
track
null
or
setting
track
that
enabled
false
or
calling
track
down
stock?
F
So
here
is
the
proposed
resolution.
The
general
principles
suggested
by
Harold
is
that
the
x
is
sent
only
if
the
SSS
he
won't
be
used
again
and
then,
if
you
get
a
bye
as
a
receiver,
the
receiver
track
that
on
mute
event
handler
would
fire,
and
the
implication
is
that,
if
transceiver
dot
stopped
is
set
to
true
that
you
would
send
a
buy
for
all
of
the
transceiver
dot.
Sanders
encoding
status,
SRC
and
coatings
that
FEC
SSC
and
then
encoding
starchy
status
se,
but
you
wouldn't
send
the
by
for
a
place
track.
F
No
because
you
could
call
replace
track
with
with
another
track.
That
would
make
the
essence
to
see
active
again.
Similarly
to
be
know
by
four
transceiver
that
Senator,
that
track
enables
our
default,
because
you
can
set
it
back
or
transceiver
that
Senator
track
got
stopped
because
you
could
call
replace
track
and
and
then
you
would
be
sending
again.
I
C
B
F
I
G
F
A
F
F
F
F
Then,
when
the
remote
description
is
applied,
then
transceiver
that
stopped
will
be
set
to
true.
So
he
stops
the
other
one.
And
the
question
was:
does
this
caused
the
unending
event
handler
to
fire
and
I
see?
The
proposed
resolution
is
that,
yes,
we
would
say
that
that
that's
what
would
happen
and
that's
that's
a
way
to
know
that
when
you
called
several
modus
shun
that
transceiver
that
stop
with
that's
true,
because
in
the
event
handler,
you
can
check
whether
in
fact,
transceiver
Don
stop.
This
is
true.
F
F
H
A
F
A
A
A
A
However,
the
situation
that
all
vendor
prefixes
were
removed
at
some
point
and
all
implementations
and
that
include
edge
and
WebKit,
now
support
create
on
your
land
and
we
have
countless
contained
the
usage
of
create
orbit
URL
with
the
using
source
object.
And
it
shows
that
the
former
is
much
more
common
and
next
slide.
A
So
I
mean
the
discussion
is
really
should
we
revive
this
create
orbit
URL,
since
we
know
less
implementations
will
continue
to
support
it,
or
should
we
try
to
convince
the
implementers
to
remove
support
of
it
that
would
break
applications,
but
we
have
a
polyfill.
That's
apparently
doesn't
work
in
all
cases
in
adopted
of
jas,
or
should
we
spec
it
again
to
put
it
back
into
our
spec
and
it's
not
the
thesis
pulling
it
back.
K
Would
absolutely
love
thee
to
get
rid
of
the
implementation
of
this
I?
Never
liked?
The
only
reason
we
input
it
in
the
first
place
is
because
chrome
didn't
have
source
object
and
everyone
was
writing
examples
that
you
sprayed
optic
URL
and
we
held
our
noses
and
implemented
it,
and
it
has
real
issues
and
I
would
love
to
see
it
go
away.
J
This
is
on
the
run
like
to
see
it
gone.
I
mean
there
are
side
effects
of
using
it
very
people,
I
think.
If
you
instrumented
revoke
object,
you
are
out
you'd
find
a
lot
of
people.
Don't
call
I'm
require
both
with
me.
You
can
actually
end
up
with
a
camera
being
on,
even
if
you
have
no
more
uses
so
in
cases
where
you
don't
call
stop
explicitly
so
this
is
obviously
a
much
worse.
K
A
K
E
G
C
C
G
K
Think
we
should
do
that.
We
should
make
a
posting
to
discuss
what
party
see
saying
it's
going
to
go
away
and
that
people
should
stop
should
start
moving
off
of
it
I
if
people,
if
you
were,
are
uneasy
with
pushing
deprecation
warnings
and
release
I
would
happily
push
deprecation
warnings
and
everything
up
through
release
as
a
start
and
and
so
that
we
can,
you
know
to
start
poking
at
people
to
start
moving
off
garba.
A
J
J
However,
the
way
we've
specified
it,
we
didn't
increase
implementation,
it's
just
a
shin
to
the
regular.
I
navigator
media
devices
getusermedia,
which
has
a
vault
which
brings
that
that,
over
the
last
year,
at
least
as
the
spec
is
involved,
we've
actually
made
changes
that
also
affect
the
legacy
API.
So
including
error
messages
like
permission
denied
error
became
secure,
which
later
became
not
allowed
air,
and
also
there
was
the
case
where
Firefox
was
actually
both.
J
If
you
call
get
using
medium
with
video,
true
and
audio
true-
and
you
can't
get
both-
let's
say,
the
system
doesn't
have
only
half
the
camera,
but
only
has
a
microphone
but
no
camera.
Then
we
should
fail
with
not
found
error
and
both
in
firefox
and
chrome.
In
that
particular
case,
where
the
hardware
is
missing,
it
would
actually
succeed
and
return
only
one
track.
I
got
fixed
and
firefox,
I
think
it's
still
broken
in
chrome
in
the
case
where
hardware
is
absent
and
so
I
think
they'll
be
too,
and
they
want
to
fix
that.
J
They
have
to
consider
whether
they
want
to
break
there's
a
choice
over
between
whether
just
breaking
the
spec
but
we're
also
breaking
the
legs
API,
which
kind
of
goes
against
the
reasons
one.
We
have
the
legacy
API.
So
to
begin
with
so-
and
this
is
hard
to
do
now-
because
if
I
just
come
online
as
well,
and
let's
get
user
media
supporting
the
legacy
api
we've
all
implemented
at
different
points
in
time,
different
things,
so
it's
probably
too
late
to
do
much.
J
F
So
yet
are
I'll.
Give
you
one
impression
that
I
have,
which
is
that
I
think
people
expected
that
error
handling,
would
change
and
be
updated
in
the
spec,
because
that
was
a
known,
weak
area,
essentially
by
every
developer
for
a
very
long
time
right.
So
the
examples
that
you've
given
are
all
has
to
do
with
error,
checking,
error,
handling,
etc.
I
think
that's
different
from
sort
of
the
promises
in
this
specific
case
that
we
had
made
about
our
the
API
for
the
get
user
media
call.
F
F
Think
it's
it's
quite
as
cut
and
dried,
you
know,
as
you
positioned
it,
where,
if
we're
going
to
support
legacy,
we're
going
to
one
hundred
percent
support
exactly
that
version
of
legacy,
I
mean
all
we're
really
trying
to
do
is
make
sure
that
you
know
people
who.
J
C
We
should
do
nothing,
we
specify
the
basically,
the
old
methods
are
a
shame
on
a
new
methods
and
the
behavior
is
specified
in
terms
of
the
new
methods.
That
means
that
this
that
we're
not
providing
the
old
specs
way
of
doing
it
with
we're
providing
a
shame
on
the
new
methods
that
looks.
Look
like
the
old
Becky
I.
J
J
Well,
it's
actually
Tyson
to
the
previous
slide
a
little
bit.
Doesn't
it
in
that
I'm
just
trying
to
shed
luggage,
share
or
baggage
from
a
legacy
baggage
and
that
we
used
to
have
advanced
and
then
we
iterated
on,
and
we
came
up
with
something
better
and
in
the
last
couple
of
years.
So
we've
had
this
new
API,
at
least
in
five
parts.
I've
not
seen
any
use
cases
where
you
had
to
pull
out
advance
to
get
we
wanted.
You
could
do
some
most
people
who
usually
get
what
they
wanted
with
min
max
an
ideal.
F
So
on,
look
again
again
again
on
principle:
edible
I
agree
with
you.
I
just
took
a
look
in
the
spec
to
see
what
it
would
require
to
move
advanced
out.
Advanced
is
really
strongly
embedded
in
all
of
the
explanations,
definitions,
etc.
There
are
at
least
five
to
ten
locations
where
it
is
mentioned
and
in
in
normative
way
in
the
constraint,
evil
pattern,
and
there
are
at
least
five
to
send
additional
places
where
their
descriptions
of
its
behavior
that
may
perhaps
not
be
normative,
but
are
still
sort
of
embedded
with
everything
else
right.
F
Well,
it
would
simplify
the
constraint,
will
pattern,
but
the
point
is:
if
we
have
to
describe
it
in
in
another
place
in
the
spec,
then
we
have
to
not
just
described
advanced.
We're
actually
saying
insert
these
lines
here
and
oh
by
the
way
in
the
description
of
the
algorithm
algorithms
behavior
the
general
description.
You
need
to
also
insert
this
as
well
and
watch
out
for
this
special
case
and
so
on.
So
I
think.
C
This
also
not
a
nice
way,
which
would
be
to
say
that
controllable
patent
algorithm
has
a
flag
which
says
with
our
advantage,
but
it
or
not,
and
wherever
advances
mention
you
say
he
spot
only
apply
advanced
flag
is
he
said
and
that
we
America
chokin.
We
use
a
patent
thing
that
is
running
this
without
the
advances
like
that,
just
just
information
about
the
image
capture.
C
J
E
D
D
And
I'm
just
okay,
so
the
main
question
is
still
wishes.
We
should
implement
and
separated
dictionaries
or
in
hybrid.
You
will
find
in
the
issue
7
14,
well,
that
the
DS
options
at
them,
Justin,
described
and
and
I
still
need
the
waiting
for
the
group
consensus
and
I
created
this
separated
the
option,
because
I
I
feel
it's
more
right,
because
it's
more
future
proof
and
so
on
and
more
clean
version.
I
got
some
feedback
about
my
nine
small
questions
from
Justin.
D
It
is
not
totally
finished.
I
didn't
edit.
The
audible
part
that
just
recently,
which
is
from
Justin
and
but
the
main
question
is
that
should
I
work
on
this
or
or
only
the
hybrid
is
acceptable
for
the
good,
because
because
it's
too
late
to
change
this
so
radically
so
deeply
and
I
would
like
to
do
here.
The
group
opinions
and
have
the
final
word
from
the
from
the
chairs.
G
Remember
they
also
have
our
perspective
on
this
I
think
this
enjoy
a
question
of
you
know
basically
code
that
he'd
ridden
today
the
axis
of
the
turn
server.
You
know
it's
going
to
specify
the
password
field
exists
today
and
that
code
would
need
to
change
if
we
went
with
the
separate
approach.
Number
two,
so
I,
don't
know
exactly
what
rubric
or
metric
reviews
here
in
this
working
group
is
I.
You
know,
what's
like
an
accessible
for
it
can
change
or
not,
but
I
think
this
is
marginally
not
a
lot
question.
G
C
B
G
D
But
it
was
a
very
difficult
of
you
if
you
had,
for
example,
the
idea
what
I
just
mentioned
to
Justin
about
just
add
the
real
something
to
this
to
this
hybrid
solution,
where
to
add
this,
this
optional
real
thing
for
a
week,
for
example.
So
if
you
want
to
do
add
something,
then
it
is
not
I
think
straight
for
our
goal
or
understandable,
maybe
for
developers
work
yet
I
think
option
to
use
is
a
suffix
playing,
and
so
this
is
why
I'm
backing
the
the
option
to.
G
D
C
G
Right
and
if
we
do
need
some
iteration
of
this
in
the
future,
you
know
we
always
have
like
this
grandfathered
way
of
doing
password
and
perhaps
have
some
more
thing
like
a
separate
way
in
the
future.
In
addition,
should
we
actually
need
that
the
non-negativity
closing
the
door
on
a
tomb
like
a
personal
future,
but
for
now
I
think
three
is
really
well
yeah.