►
From YouTube: TPAC 2019 WebRTC meeting @Thursday part 4
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
Usually
they
but
I
also
wanted
to
get
some
background.
Well,
first,
I
want
to
go
to
the
option.
So
it's
clear
what
we're
gonna
do
at
the
end
and
then
we
want
to
able
to
take
some
actions
and
then
so,
basically,
if
there
are
no
implementations
and
all
developer
interests,
remove
the
feature
and
rich
sorry
about
next
line.
A
A
A
C
A
Items
we
have
this
thing
just
in
case.
We
missed
something
in
the
process,
it's
better
to
have
all
eyes
on
it.
So
this
is
things
we
looked
at
in
no
particular
order
and
also
includes
things
that
are
not
features
at
risk.
So
the
legend
here
is
that,
in
the
yellow
column
on
the
right
to
choose
most
column
only
features
marked
with
a
red
X
admitted
loss
column.
Our
miss
are
both
missing
two
implementations
and
new
term
commitments
and
that
those
are
also
marked
in
bread,
so
you'll
see
and
the
red
X
is
not
implemented.
A
Green
is
implemented
and
the
working
arm
means
someone's
working
on
it,
a
writ
or
as
an
actual
developer,
assigned
Department
and
in
near
Karen
being
this
year.
So,
for
example,
the
first
line
that
is
rollback
which
are
naming
has
one
implementation
but
I
believe
we
have
commitments
from
from,
and
yes
so
that
Haymarket
as
as
commitment,
so
that's
good,
and
then
we
can
go
down
I'm
a
builder
so
quickly
to
this
list
in
case
and
someone
yell
out.
A
If
there's
a
problem
for
mice,
transport
and
DPS
transport,
it
looks
like
both
Chrome
and
edge
are
implementing
it.
There
might
be
educational
members
missing,
and
this
might
be
my
fault
as
a
from
our
first
perspective,
I
didn't
dive
into
individual
members.
So
much
so.
If
someone
has
concerns,
they
please
point
that
out
or
rtcp
transport.
A
A
So
we
farm
parts
think
that
we
can
add
them,
and
that
would
we
think
it's
low-hanging
fruit,
basically
because
and
we
haven't
implemented
yet
because
mostly
what
people
will
use
that,
for
probably
at
least
one
platform
test
uses,
because
it's
the
quickest
way
to
get
to
the
DBMS
transfer
luck
all
right
so
did
Monday.
We
would
fix
a
lot
of
ass,
a
lot
of
arm
tennis.
If
we
just
remember
that
it
doesn't
make
sense
to
do
that
until
we
implement
kikyo's
transported
vice-chancellor
set
parameters,
complications
same
problem.
A
E
F
D
E
A
F
J
A
E
N
Mean
basically
I,
don't
think
it's
gonna
have
a
backward
compatibility
problem,
because
it's
basically
extending
the
error
right,
and
so
it's
just
more
info.
You
might
not
look
for
it,
but
it
just
gives
you
more
detail
on
whatever
you
happen
to
implement
it
with.
So
it's
not
like
anything
to
do
should
have
already,
probably
from
in
there.
M
O
A
G
H
N
A
E
E
N
The
actual
failure
rates,
real
applications
are
actually
a
staggeringly
high
and
it's
it's.
You
know
if
you're
trying
to
build
an
operational
service
having
this
kind
of
date
is
actually
extremely
well.
I
really
I
mean
you
know
the
level
of
level
of
I,
don't
think
we'll
ever
get
to
the
level
of
perfection
where
every
browser
was
exactly
the
same
information
on
the
same
error
that.
M
E
D
N
M
O
N
I
E
Q
M
R
L
I
E
A
A
R
R
A
R
C
A
And
anything
was
separated
rather
than
just
ever
expect,
but
is
still
more
furtively
referenced
from
an
emergency
PC.
That
means
there's
still
a
couple
lines
of
language
in
limine.
See
you
see
that
talks
about
identity,
but
there's
only
one
implementation
identity,
so
I
was
brought
up
that
passage
of
need
discussed
here,
whether
or
not
is
okay
to
separate.
You
know.
M
A
Thanks
fine
there's,
also
20,
it
looks
like
there's
28.
Mandatory
stats
I
should
open
web
platform
tests,
it's
not
implemented,
but
we
believe
this
is
mostly
result
of
stats
time
for
summing
rules
in
the
spec
recently,
so
that
it
doesn't
mean
they
have
no
implementations.
There
are
limitations.
That
is
not
where
they
should
begin
sparing.
So
as
a
proof
of
concept,
they
work
and
I.
Think
that's
good
enough
for.
A
A
A
A
E
R
J
Q
E
A
E
M
J
A
O
A
I
T
R
T
E
T
A
O
A
J
A
A
E
Think
they
are
business
records
or
butter,
I
think
moving
moving
it
out.
As
surely
an
editorial
question,
acne
clock.
We
wanted
to
sign
it,
it's
this
carbon.
Let
me
see,
100
or
not,
answer
is
it's
not,
then
we
we
do
whatever
into
the
editorial
question
thing
is.
Let
us
say
to
make
sure
that
is
on
the
card
of
the
attention
span.
H
J
I
G
P
W
B
Some
video
conferencing
service
for
the
higher
education
and
research
and
I'm
working
for
a
distributed
service
for
this
community
and
what
I
have
here.
So
we
have
high
bandwidth,
Network
and
virtual
machines
around
the
globe,
and
we
want
to
have
a
global
terror
service
for
your
community
and
the
goal
is
to
keep
traffic
the
media
traffic
inside
our
network.
And
so
we
want
to
create
a
multi-tenant
turn
solution
and,
as
you
know,
it
is
not
possible
with
the
long
term
Prudential.
So
the
non-confidential
authentication.
B
What
is
the
basic
authentication
indeterminacy,
because
the
draft
origin
base
draft
is
rejected
by
relatives.
So
there
is
no
way
and
and
I
want
you
to
argue
to
keep
the
old
inside
spec,
because
this
is
the
only
one
provides
a
standard,
ITF
standard
that
is
suitable
for
the
web.
So
the
long
term,
prudential
username
password
is
not
really
for
the
four
browsers.
They
cannot
keep
it
in
secret
and
it
supports
it.
A
co-located
turn
and
and
co-located
turn
service.
B
So
we
could
have
multiple
authentication
that
the
bases
around
the
globe
and
four
different
communities
for
euro
for
I
don't
know
America
and
some.
But
we
could
operate
a
single
service,
so
this
is
the
goal.
What
we
want
to
achieve
and
I
started
by
working
on
the
cotton,
because
I
started
to
use
cotton
later
I
released
the
last
releases
I
made
so
I
am
developing
cotton
and
also
I
started
to
work
on
the
Firefox
side.
I
have
a
confirmation
from
buy
one
to
implement
the
old
stuff
in
five
routes.
B
I
am
working
in
the
nicer
level
and
made
some
so
I
have
some
progress
and
I
hope.
I
will
finish
it
in
this
year.
So
I
was
one
who
would
disarm
in
the
fire
foot.
So
it
was
not
a
anavar
mistake.
Sorry,
it
was
my
and
sorry
about
it,
but
I
think
it
will
happen,
and
this
year
and
I
I
don't
know
why
it
is
not
the
other
browsers.
Why
I
don't
want
to
implement
this?
It
is
not
I.
Think
too
big.
B
If
effort
to
to
put
these
two
parameters
in
the
old
end,
and
and
after
that,
in
turn,
you
need
to
implement
two
attributes.
So
a
little
change
in
the
in
the
logic
changing
the
MAC
address,
replacing
the
key
with
with
Becky,
but
I
think
it
is
not
to
be
forced
and
and
I
I
will
try
to
at
least
implement
it
in
Firefox
and
and
demonstrated
with
cotton,
because
cotton
has
the
implementation,
the
server-side,
so
everything
is
there
I
think
to
use
this,
and
this
is
the
only
one
that
supports
multiple
turn.
B
This
turn
service
for
for
the
different
in
different
services
that
we
are,
we
are
providing,
so
we
have
commercial
services
and
be
also
doing
some
open
source
development,
so
WebRTC
development
for
our
community
and
also
I
want
you
to
say
that
we
are
trying
to
use
the
identity
part.
So
the
identity
is
also
important
to
us.
B
We
are
still
not
really
are
a
small
group,
so
we
or
developments
are
going
very
slow,
but
we
want
to
keep
these
two
things
inside
the
spec,
if
possible
and
I.
Don't
know
why
why
the
what?
What
are
the
others
using
for
turn
authentication,
because
the
the
time
limited
thing
this
is
a
risk
expired
draft
and
this
is
expired
and
not
continued.
To
move
on
to
turn
old
as
far
as
I
understand
and
and
the
the
the
original
long
term
Prudential
is
is
is
not
designed
for
to
that.
B
M
It
will
make
it
in
the
same
release
cycle
of
the
specification.
We
got
most
big
gains
on
the
fact
that,
indeed,
currently,
that
has
a
big
enough
take
up
by
Blumenthal's
if
we
do
move
it
out
of
the
specs
of
plan,
an
extra
inspect
web
RTC
human
mg's,
and
if
so,
we
will
need
people
who
get
involved
to
actually
make
sure
that
this
extension
spec
is
maintained.
M
B
B
B
Actually
we
use
and
we
provide
order
to
authentication,
so
we
provide
long-term
Prudential
so
username
password
for
for
long
term,
but
we
also
providing
according
the
spec
the
rest.
So
we
have
a
REST
API
to
get
short
term
credentials,
unlimited
credentials,
but
I'm
not
happy
with
this.
The
turnout
will
be
much
better
way
and
more
standard
way,
because
the
time
limited
rest.
This
is
an
expired
threat.
So
it
is
not
a
standard
and
the
old
in
many
other
things.
B
It
is
better
because
it
also
supports
the
multiple
authentication
sources,
so
you
can
have
because
the
the
REST
API
use,
dreams
and
real
is
only
limiting
to
one
domain
and
if
we
can
have
multiple
out
an
authorization
servers
that
provide
tokens
for
the
service,
then,
if
we
could
sort
that
could
have
implies
Asian
server
for
Europe
and
America
and
so
on.
So
it
will
be
the
much
better
solution
for
it.
For
that,
so
I
am
arguing
to
keep
the
turnout
inside
aspect
still
I.
B
B
B
B
No,
no,
no!
No!
No
yeah!
You
can
use
the
same
from
those
client
sites
from
the
browser
side.
It
is
not
changed.
Anything
the
behavior
and
the
rest
only
on
the
server
side
needs
some
changes.
So
from
the
cotton
side
who
you
need
to
change
the
authentication
but
is
I
using
this
actually
in
in
their
service
and
I,
proposed
to
use
everywhere
if
possible,
but
it
has
a
delimited
that
I
can
have
only
one
authorization
server,
so
one
real
one
database
with
the
credentials.
B
X
Okay
cool,
so
a
lot
of
mine
come
from
a
very
used
very
use
case.
I
work
on
a
go
implementation
of
WebRTC,
and
so
these
are
the
some
of
the
things
that
come
up
increasingly
I'm,
seeing
what
the
ice
transport
a
lot
of
people
are
coming
to
me
with
complaints
about
not
about
knowing
what
kind
of
situations
they
want
their
ice
to
work
in
only
allowing
certain
port
ranges,
whitelisting
some
interfaces.
N
X
So
it
seems
it's
mostly
like
a
mostly
at
companies
where
it's
D,
it's
detached
from
the
developers
and
people
managing
their
networks
where
they
put
a
request
for
it,
and
they
say
we'll
give
you
this
certain
range
of
range
of
ports-
and
this
is
all
you're
allowed
to
use,
and
so
they're
stuck
in
this
case,
where
they
want
to
use
WebRTC.
But
they
have
to
end
up
choosing
other
things.
Because
of
because
of
this
constraint.
X
N
X
Are
mostly,
these
are
Mo's
yeah,
these
are
most
I
would
call
them
more
small
business.
It
seems
that
the
enterprise
level
people
are,
you
know
they
know
enough.
That's
going
on
that.
They
don't
run
in
these
issues.
I
see
a
lot
of
small
companies
that
are
trying
to
try
to
build
internal,
tooling
and
small
projects.
X
Yeah,
the
the
next
one
is,
is
probably
a
small
one.
A
lot
of
people
have
signaling
that
isn't
that
it's
somewhat
naive
and
just
sort
of
spews
messages,
and
they
come
out
of
order
and
stuff
like
that
and
I
feel
that
it's
maybe
it
feels
like
once
a
week.
I
have
someone
in
the
PI
on
slack
channel
that
complains
about
a
dice
candidate
before
a
set.
Remote
description
throws
an
exception.
I,
don't
know
if
it's
possible
to
just.
X
X
X
This
is
only
for
the
first
case.
If
you
add
candidate
at
candidate
and
then
then
you
get
an
off,
then
you
get
a
an
offer.
You
can
assume
that
these
candidates,
that
you
have
already
goes
into
the
first
offer
and
then
this
issue
doesn't
exist
anymore,
because
you
already
have
an
offer.
So
you'll
never
run
into
this
issue
again.
So
I
don't
think
renegotiation
applies.
A
You're
just
talking
about
an
initial
ago,
she
ation
but
does
have
to
deal
with.
There's
no
limitation
negotiate
once
and
I
can
come
in
for
a
long
time.
So,
if
you
renegotiate
and
on
the
other
side,
that's
sending
this
stuff,
usually
they
would
just
hook
up
JavaScript
or
nice
candidate.
It
is
system,
so
there's.
A
I
think
what
happen
might
be
happening
as
a
a
lot
of
people.
There's
the
race
potential
where,
when
something
comes
into
a
New
England
channel,
you
have
to
call
such
remote
description
first,
you
should
don't
wait
if
you,
if
you're
waiting
for
anything,
then
you
can
get
racist
like
if
you
try
to
get
you
to
medium,
for
example,
at
the
time,
the
light
on
that
before
calling
central
description
that,
if
that
is
then
that's
a
pilot
error
in
the
D
game,
uses
that's
a
problem.
They
yeah
I.
X
I
guess
I
agree
that
we,
it
would
be
great
if
we
didn't
have
these
races,
but
in
the
real
world
you
have
popular
WebRTC
shims
like
simple,
peer
and
stuff
like
that
that
cash,
the
ice
candidates
and
then
add
them
after
the
set
remote
description
is
called.
So
people
are
already
doing
this,
but
in
JavaScript
and
then
you
have
people
that
are
trying
to
build
their
first
applications
and
then
they're
hitting
this
case
and
it's
souring
their
experience.
A
X
A
E
X
So
the
next
one
and
I
guess
this
is
one
of
those
where
you
can
implement
this.
In
other
ways,
commonly
people
will
ask
if
they'll
come
from
using
WebSockets
and
they'll
use
the
close
codes
and
they'll
ask
for
the
equivalent
for
data
channels
which
we
can't
provide,
but
at
the
same
time
you
can
just
you're
finding
that
your
final
message
in
the
data
channel
can
be
your
closed
code.
If
you
want
not
sure
if
that's
worth,
but
people
asked
me
to
bring
it
up.
X
The
ability
to
deny
a
data
channel
someone
has
a
little
VPN
application,
that's
it
embedded
and
they
want
to
look
at
the
name
of
the
channel
that
you
know,
fires
the
open
and
immediately
close
it
again.
This
seems
like
something
that
you
can
accept
it
and
close
it
quickly,
not
sure
that
there's
value
here
are
to
complicate
the
spec
for
no
reason,
a
lot
of
a
lot
of
things
in
this
I
guess
will
fall
into
the
next
one
is
media
via
data
channel
is
becoming
incredibly
incredibly
popular.
I
have
one
user.
That's
there.
X
There
trying
to
you
know,
have
like
five
million
streams
or
whatever
all
these
users
and
the
amount
of
reduction
in
CPU
costs
because
of
the
cypher
choice
is
frustrating
them
a
lot
they're,
seeing
like
a
50
percent
jump
in
cost
and
media
encryption.
Because
of
that,
and
we
can
go
into
that
down
the
DTLS
transport.
So
the
RTP
transports
working
with
a
lot
of
embed
customers
that
are
doing
security
cameras.
X
They
want
to
be
able
to
control
those
latency
loss,
trade-offs,
I'm
hearing
from
some
people
that
they
would
prefer
to
have
packet
loss
or
they
would
prefer
to
have.
You
know
a
lot
of
keyframes,
but
have
you
know
400
milliseconds
of
latency
all
the
time,
because
that's
very
very
important
to
them
in
there
and
they
have
customers
that
are
frustrated
that
when
their
camera
latency
grows
or
shrinks
and
they're
confused
by
that.
X
Ya
know,
I
think
some,
it's
I
can't
think
the
one
I've
heard
that
I
don't
know
how
much
it
matters
is.
Some
people
have
SF
use
or
media
service
are
saving
service.
I've
been
there.
They
want
to
be
able
to
increase
the
latency,
but
I.
Don't
think
that
applies
here,
but
yeah
there's
some
places
where
there
or
someone's
like
doing
something
where
the
latency
is,
as
it
isn't
as
important
like
a
second
two
seconds.
It
might
be
okay,
but
they
don't
want
the
viewer
to
see
any
sort
of
corruption
or
packet
loss.
X
N
X
I
think
I
think
both
I
have
some
people
that
they're
very
concerned
about
head-of-line
blocking
and
they
want
to
have
really
low
latency
and
some
are
basically
just
sending.
You
know:
DRN,
like
they're,
just
sending
stuff
to
the
throw
in
to
the
media
source
extension
I
think
both
it's
people
just
want
more
flexibility
is
and
that
they're
not
able
to
get
with
the
current
media
api's.
H
X
That
50%
increase
is
because
the
four
SRTP,
the
only
cyprus
suite
that's
offered,
is
a
DSC
cm
with
an
authentication
tag
of
sha-1,
something
and
but
DTLS.
You
have
access
to
the
ATS
GCM,
which
you're
able
to
use
like
the
hardware
accelerated
the
ANS
and
I
or
whatever
it
is
for
neon,
and
there
is
a
email
of
the
mailing
list
about
this,
but
they're,
seeing
like
a
50%
jump
just
because
it's
not
hardware
accelerated.
J
X
X
Yeah,
so
those
are,
those
are
the
big
ones,
is
more
control
and
then
the
lack
of
hardware
acceleration-
and
it's
very,
very
simple.
Like
it's
already
available,
it's
just
not
turned
on,
and
then
the
next
big
one
and
and
really
I
realize
that
sorting
exists
is
the
second
set
codec
preferences
on
transceivers
aren't
available.
Yet
people
love
their
love.
Their
h.264,
which
I
know,
is
a
contentious
subject,
but
that's
I
feel
like
that's
the
first
thing,
people
so
and
I'm,
tired
of
seeing
people
throw
out
STP
munging
solutions.
X
The
DTLS
transport,
like
a
cipher,
suite
choice.
This
is
really
just
again:
people
want
hardware
acceleration
when
they
know
they
have
it
and
maybe
the
WebRTC
implementation.
The
client
needs
just
make
better
choices,
but
this
does
come
later
with
snowflake
and
then
for
the
peer
connection.
People
complain
about
this
all
the
time.
The
ad
stream
works
ad
track.
First,
ad
transceiver
users
unsure
which
to
use
and
I,
don't
know
what
the
answer
is
there.
X
Think,
literally,
my
only
concern
it's
ad
track
of
sort
of
this
leaky
abstraction,
it's
very
difficult
for
people
to
understand.
What's
actually
going
on
with
the
media
sections
until
they
have
the
ad
transceiver,
it
causes
a
lot
of
headache
for
people
when
they're
they
don't
understand
why
certain
things
are
working
and
then
also
you
have
the
multitrack
like
it
seems,
and
I
I
always
start
people
off
with
a
transceiver,
just
I,
don't
know
any
kind
of
professional
insulation
yeah.
Oh.
X
So
this
is
provisional
offers
and
answers.
Getting
rid
of
those
is
a
wishlist
item
for
me,
mostly
I,
don't
I,
don't
really
see
any
value
in
them,
and
on
top
of
that,
it's
like
a
anecdote
that
comes
up
like
every
week
or
so
that
someone
asks
me
what
a
PR
answer
is
because
they'll
see
them
in
random,
airs
or
work
like
this
leave.
You
know
and
I'm
like
I,
have
to
like
go
through
this
whole
conversation
and
like
I,
have
to
cover
it
in
PI
on
like
I.
X
Ok,
this.
X
This
is
sort
of
a
wishlist
item.
Probably
not
no
one
in
this
room
really
cares,
but
it
would
be
nice,
the
I
guess
the
v1
of
the
WebRTC
API.
It
never
really
did
any
multi
method
dispatch
off
of
types.
So
the
one
that
frustrates
me
is
the
ad
transceiver,
the
fact
that
it
can
take
a
Dom
string
or
it
can
take
a
track.
It's
difficult
to
implement
in
other
languages
that
don't
have.
This
I
have
a
C
implementation
that
models
after
the
w3c
and
then
the
go
one.
X
That's
public
I
get
that
doesn't
really
affect
with
this.
What
we're
concerned
about
here,
but
it
is
nice
for
others
and
I-
think
it's
nice
for
the
adoption
of
WebRTC,
because
there's
plenty
of
places
that
people
want
to
use
it.
It's
not
the
mouth
of
browser,
but
again
not
really
relevant
here,
but
just
a
nice
ask
and
then
tour
snowflake
is
like
a
is
a
censorship
circumvention.
X
Basically,
they
use
data
channels
to
do
the
Onion
Routing
and
it
would
be,
and
they've
spent
a
lot
of
time,
removing
the
ability
to
fingerprint
things
so
the
ciphers
that
are
provided
by
some
browsers.
The
fact
that
ice
is
authenticated
not
encrypted
little
things
like
that-
probably
not
a
definite,
not
a
high
priority
for
for
anyone,
but
it
would
be
nice
to
make
life
easier
for
that
and
who
knows
what
else
that
could
open
up.
H
H
H
X
H
X
Yeah
I
will
update
that
update
that
get
up
issue,
but
but
it
seemed
this
will
probably
be
require
a
lot
of
peer
connection
because
they
want
to
have
control
over
certain
attributes
that
most
people
don't
care
about.
So
I'll
update
that
right
now
and
that'll
that'll
be
available
on
the
the
NV
repo.
Y
Y
App
using
telemedicine
developer
focus,
so
we
were
telling
a
telemedicine
software
to
you,
healthcare
professionals
and
to
Clinton
to
our
Hospital,
and
we
have
to
you
know,
jump
through
hoops
to
make
something
interoperable
who,
in
our
support
in
Chrome,
Safari
and
Firefox
most
we
have
to
recommend
people
not
to
use
Firefox,
because
the
video
quality
is
not
good
enough
from
in
Safari.
Do
much
better
with
the
video
quality.
N
Y
Y
Tell
you
what
I
freely
admit
I'm
now
you
know
CEO
of
this
company
and
don't
actually
did
my
get
too
much
into
the
dips.
This
is
more
of
really
high
level
high
level
report
of
some
of
the
pains
we
have
to
go
through.
So
apologies.
If
I
can't
tell
you
the
reasons
why
something's
happened
so
I,
don't
know
why
the
new
quality
faculty
so
much
worse.
I
would
actually
love
to
use
Firefox
more,
but
we
just
haven't,
have
had
very,
very
poor
success
with
it.
Y
It's
particularly
I
think
when
we
do
screen
sharing,
as
well
as
when
you're
sharing,
so
we
have,
we
have
it
set
up
so
that
our
screen
sharing
is
actually
showing
just
like
like
these
on
the
side
of
running
the
videos,
it's
not
replaced
in
our
local
video.
It's
sharing
it
as
a
as
a
full-blown
piece
of
content,
and
so
we're
using
the
thing.
Y
About
what's
it
called
what
we
pray
right?
It's
me
and
I'm
sorry
we're
using
the
new
API
that
they
could
remove
from
from
the
legacy
to
the
to
the
nude.
You
know
when
you
can
save
multiple
video
streams
at
the
same
time
and
that's
working
better
in
chrome
and
not
so
well
in
in
Firefox
anyway,
one
of
the
one
of
the
key
problems
we're
having
is
that
that
you
know
it's
all
about
interoperability?
Y
All
of
my
complaints
will
be
about
interoperability
and
lack
of
interoperability,
and
that
may
well
be
that
API
is
a
missing
like
the
recording.
Api
is
missing
in
Safari
and
iOS,
but
it's
it's
also.
A
part
of
AP
is
not
being
implemented
consistently.
Between
browsers
I
can
mention
that
I'm
also
looking
forward
to
getting
more
efficient
implementation
of
video
decoding.
We
often
hit
devices,
they
can't
cope
with
the
high
CPU
requirements
of
video
decoding.
We
sometimes
hit
that
problem
as
well
at
one
of
the
biggest
things
we
need.
Y
We
need
recently
other
statistics,
so
when
we
pull
statistics
of
the
web
calls
and
we
sent
that
into
our
monitoring
application,
we
we
have
to
ap
is
fundamentally
different
api's
now
to
support
the
legacy
one
and
the
new
one
and
the
new
API
is
actually
much
more
limited
than
the
old
one.
There's
a
lot
less
information
in
there
and
there
also,
while,
while
all
the
browser's
are
not
implementing
the
new
one,
they
are
not
implementing
simple
completely.
Some
have
their.
Y
There
are
certain
bits
missing
from
from
Paris,
so
it's
really
really
hard
when
you're
trying
to
provide
a
service-
and
you
have
to
support
current
browsers
as
well
as
old,
browsers
and
all
the
complexities
of
all
the
combinations
between
all
of
them,
because
when
you're
getting
to
a
video
called,
maybe
you've
got
four
people
joining
from
four
different
browsers
and
they
all
need
to
be
working
with
each
other
person.
So
it's
an
utter
nightmare.
I.
Y
E
Y
My
profiler
also
just
mentioned
another
issue
that
is
coming
across,
so
sometimes
we
set
up
a
call
and
he
guys
were
really
telling
us
whether
the
call
is
accessible.
So
we
check
things
like
the
current
time
on
the
video
element
to
check
whether
the
video
is
flowing
so
with
whether
the
current
time
is
changing,
but
that's
inconsistent
so
most
of
the
time
it
works,
but
sometimes
we
get
media
playback
through
the
device
in
the
current
time
off
the
elements
days,
look
at
zero.
Y
E
Y
Y
Yeah
I
know
this
sounds
like
a
complaint
first,
so
I'm
really
sorry,
but
first
of
all,
I
wasn't
I,
wasn't
very
much
prepared
for
this,
so
I
apologize
but
I'm
just
trying
what
I'm
trying
to
tell
and
to
share
or
some
of
the
team
here,
a
few
goes
through
as
developers
and
as
application
providers,
which
is
quite
possibly
also
one
of
the
reasons
why
the
PB
s--
news
case
for
web
RTC.
At
the
moment
you
can
use
of
the
data
tender
for
the
libertage
video.
Y
If
you
don't
actually
have
to
deal
with
the
complexities
of
having
lots
of
prizes
and
interact
with
each
other
20.5
forms,
all
you
do
is
send
it
to
one
API.
So
it's
a
lot
easier.
So
what
he
originally
developed
without
instant
four
or
one
of
the
use
cases,
which
is
audio
and
video
conferencing,
that's
actually
really
the
hardest
and
most
difficult
use
case
that
we
have.
Y
Combinatorial
we
need
more
complex
when
you
have
to
deal
with
all
the
different
versions
of
the
standard,
and
that's
just
you
know
that's
just
the
way
of
the
world
right
now
and
I
hope
that
will
get
better
in
future.
But
you
know
that's
what
you
deal
with
right
now
and
that's
where
we
all
I
got
so
sorry
sounds
like
a
good
plan
faced,
but
I
thought
I
chair
it.
Nevertheless,
thank.
N
Z
Folks,
yes,
again,
apologies
for
not
having
put
this
down
on
the
slide
I'll
do
that
after
after
the
meeting
so
I
only
I
only
have
three
items
to
begin
with.
One
of
one
of
the
pain
points
for
us
right
now
is
this
whole
messiness
around
switching
plans
and
really
from
our
perspective.
We
have
this
that
the
situation
is
so
we
have
this
huge
change
that
needs
to
happen
that
will
not
bring
us
any
value.
Z
So
there
has
been
this
entire
discussion
about
SSR
cease
being
manipulatable
with
after
after
a
switch
to
unified
so
that
you
don't
have
to
go
and
immediately
have
our
IDM
ID
support
in
SFU,
so
that
you
can
do
changes
primarily
on
the
client
side
and
then
gradually
move
to
to
newer
to
our
IDM
ID
support
on
the
SFU
as
well.
This
this
has
been
discussed
a
lot
so
I'm
not
going
to
say
too
much
on
it,
but
I
just
wanted
to
mention
it's
one
of
the
major
things.
Z
One
thing
that's
an
implementational
concern,
that's
not
really
for
this
group,
but
I
mentioned
I'm
going
to
mention
just
the
same.
We
keep
hearing
very
very
often
about
this
problem,
about
how
not
all
windows
are
shareable
on
windows,
depending
on
specifically
those
that
use.
My
show
you
why
that's
a
problem
that
you
hit
with
chrome
and
soon
with
edge
as
well.
Z
There's
really
no
way
we
could
do
anything
to
help
audio
quality
much
like
for
you
could
say
that
for
video
we
could
add
additional
sec
on
the
SFU,
but
out
of
anna's
FEC
is
not.
Is
not
currently
supported
for
for
audio,
so
it
would
be
great
if
we
had
that
option
as
well
and
I
think
actually,
this
folds
into
the
bigger
topic
of
we'd
actually
like
to
be
able
to
add
our
own
FEC
to
to
peer
connections
and
those
those
hooks
that
have
been
discussed
in
the
past.
Z
That
is
something
that
we're
really
really
looking
forward
to
primarily
for
this
quality
issue
being
able
to
assist
it
and
adjust
it
on
the
SFU
side,
but
also
because
we'd
like
to
be
able
to
do
things
like
end
to
end
encryption.
So
there's
a
document
on
github
from
high
road,
the
describes
an
SF
use,
sorry
an
end
to
an
encryption
scheme
that
is
actually
entirely
compatible
with
our
vision
as
well.
This
is
what
we'd
like
to
be
able
to
do.
Z
The
perk
efforts
on
the
ITF
are
not
going
and
work
for
us
so
again
we're
back
in
through
these
hoops
on
the
client
side.
That
would
let
us
intercept
traffic
at
various
stages
and
inject
our
own
logic
in
there.
Whether
it's
you
know
after
encoding
before
pure
connection
story
after
peer
connection,
so
that
we
add
our
own,
does
that
make
sense.
E
N
N
AA
AA
This
is
something
that
we're
looking
forward
to
having
part
of
this
back
to
and
the
second
one
which
email
just
mentioned
is
all
the
env
topics
right.
How
do
we
really
expose
the
hooks
from
different
points
of
the
stack
and
let
developers
to
plug
in
there
that
there
are
modules
there
right,
letting
them
improve,
quality,
reliability
and
so
on
and
so
forth
right?
This
is.
This
is
something
that
we're
excited
to
see
at
some
point
becoming
part
of
the
standard
right
and
part
of
the
implementations
to.
N
AA
N
AA
There
are
some
difficulties-
yes,
specifically
around
SSRC,
Hanlin
great,
that
we
are
while
trying
to
work
around
it
as
we
speak,
but
again
right.
The
ultimate
plan
is
to
move
to
met
red
based,
multiplexing
architecture
right-
and
this
is
I-
guess
the
general
guidance
as
well,
but
again
right.
This
transition
period
needs
to
be
set
up
properly,
so
that
folks,
who
do
want
to
use
unified
plan
right
instead
of
Plan
B,
can
do
that,
while
preserving.