►
From YouTube: WebRTC Virtual Interim of August 23 2016
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
A
A
And
if
you
want
to
make
a
substantive
contribution
as
a
poster
for
the
other
spreading
their
fate
you're
supposed
to
be
a
member
so
that
you
have
have
signed
all
the
agreements
that
say,
minus
FM
is
nothing
for
fast
move
against
this.
Unless,
unless
I
tell
you
first
I
should
not
shouldn't
really
be
wrote
this
village
just
too
tempting.
A
We
have
to
kill
if
they
all
open
issues
they
don't
want
to
open.
This
is
a
new
functionality.
If
we
can
avoid
it
and
that's
it's
absolutely
necessary
in
order
to
make
good
applications
and-
and
we
don't
want
to
change
stuff-
that's
already
that
people
are
already
depending
on
unless
it's
absolutely
necessary
to
make
sensible
abs
so
that
we
have
to
make
sure
the
language
of
the
documents
say
what
we
intended
to
say.
A
D
I
am
recording
this
wizard
red
button
and
I'm,
seeing
it
for
free
tech
reworking
good.
Yes,.
F
A
Then
we
can
go
to
go
to
slide
side
file
in
the
list
of
items
for
discussion.
We
have
five
issues
for
which
there
is
a
pull
request
and
the
decision.
The
decision
have
to
take
this
weather,
whether
the
pool
reflect
because
the
effect
to
consensus
or
a
group
or
whether
we
have
to
say
something
else,
and
we
have
four
issues
on
which
their
milk
is
both
not
bullock
western.
We
will
discuss
the
right
way
forward
and
hopefully
assign
a
volunteer
or
an
editor.
It
is
easy
to
to
write
the
pull
request
you
so,
okay.
B
G
Alright,
so
the
first
issue
is
issued
685
and
PR
759,
and
it
started
out
as
just
a
question
about
a
reference
to
Jessup
around
the
following
section:
5.1
quote,
which
related
to
how
you
set
up,
create
offer
to
send
multiple
are
to
be
in
coatings,
which
is
by
setting
sending
coatings
and
also
what
a
description
does
when
called
with
a
corresponding
remote
description.
That's
able
to
receive
multiple
RGB
encodings.
G
G
The
second
question
is
whether
WebRTC
10
enables
browsers
to
receive,
as
well
as
to
send
multiple
encodings
and,
if
so,
how
that
works.
So
I
guess
we'll
probably
talk
mostly
about
the
second
one,
so
they,
as
far
as
I
know,
the
draft,
as
as
I
mentioned,
only
talks
about
the
sender
side
of
it
because
there
is
no
RTP
receiver.
G
That
set
parameters
like
there
is
with
the
sender
and
no
receiving
coding,
see,
there's
no
way
to
set
up
create,
offer
to
generate
an
offer
that
could
receive
multiple
arch,
p
encodings,
but
you
can
still
have
set
remote
description
called
on
a
remote
description.
That's
able
to
send
multiple
large,
p
encodings,
and
so
the
question
is
what
happens
when
that
occurs
or
if
that
occurs,
and
it
yeah.
G
Someone
want
to
comment
so
when
the
PRI
suggest
that
if
the
RT
receiver
can
receive
multiple
encoding,
the
parameters
retrieve
the
receiver
get
parameters
will
reflect
the
encoding
sneh
go.
She
ate
it
and
I.
Guess.
If
the
browser
can't
do
it,
then
you
don't
negotiate
multiple
and
coatings
and
there's
nothing.
G
No,
no
simulcast
reception
reflected
in
receiver
get
parameters.
So
it
is
a
little
bit
odd
in
that
you
can't
create
an
offer.
Browser
can't
create
an
offer
to
receive
multiple
coatings,
but
if,
like
conferencing
unit
or
something
offers
to
send
them,
then
it
could
accept
it
that
make
any
sense
to
people.
C
But
ours
colors
your
arms
yeah
that
make
sense
like
I
get
you're,
gonna,
I,
think
there's
a
slight
technicality
and
in
phrasing
this
here
I
think
it's
that
what
we
want
to
try
and
do
is
make
sure
that
this
fact
does
not
require
people
to
implement
nothing.
You
were
discussing
that
browsers.
Can't
do
okay,
you're
not
required
to
do
that
as
well,
that
one
dot
0.
We
certainly
don't
want
to
forbid
that
from
happening
by
some
excuse
or
extensions
back
for
anything.
G
A
question
Colin
so
right
now
the
spec
like
I
said
this:
you
can't
generate
an
offer,
but
you
could
you
could
accept
an
offer
that
can
do
it,
but
you're
not
mandated
to
do
because
I
think
I
mean.
I
don't
think
that
if
you
got
something
and
put
it
into
remote
description
that
offer
to
send
you
multiple
encodings
you're
not
going
to
generate
an
error
right
well,.
C
G
G
C
G
Creating
it
doesn't
well,
the
question
is
even
creating
an
offer.
You
can't
do.
No,
the
question
is
how
you
would
do
it
if
you
can't
manipulate
the
UK.
The
problem
is
that
you
can't
manipulate
the
receipt
on
the
receiver
side,
you're,
not
able
to
manipulate
the
encoding
parameters,
who
actually
indicate
that
you
can
receive
simulcast.
That's
the
problem,
that's
why
so,
in
the
absence
of
that
kind
of,
you
have
to
react
to
somebody
else's
offer.
That's
kind.
G
G
G
C
Sajal
I
don't
want
to
that.
There
may
be
other
ways
that
a
browser
may
decide
to
do
that.
A
browser
implementation
due
to
some
static
configuration
of
it
may
decide
to
always
offer
similar
out
every
single
offer
generate,
but
an
example
not
saying
it's
a
good
one.
Other
thing
is
I:
don't
want
to
mandate
that
anyone
has
to
implement
any
of
that.
I
just
want
to
not
forbid
that
from
happening,
because
we're
going
to
want
to
extend
it.
C
J
C
J
Think
this
yet
the
other
issues
like
do
you
simulate
how
this
actually
works
like
what
do
you
actually
get
when
multiple
coatings
are
actually
transmitted?
That's
the
other
thing.
I'm
you're
worried
about
heading
down
this
path
here,
like
my
dude,
that
really
wanted
to
leave
this
all
intentions,
but
because
I,
just
sort
of
covering
that,
as
one
small
caves
like
I,
think
can
be
kind
of
you
want
Enda
how
to
cover
a
bunch
of
different
parts
of
it
like.
How
do
you
figure
out
what
that
actually
I
put
it
on
the
track.
G
J
J
E
J
J
J
What
Jake's
death
will
do
will
be?
You
know
the
behavior
will
be
defined
in
jasa.
I'm
not
sure
what
behavior
needs
to
be
defined
here,
given
that
the
actual
processing
is
defined
in
Jacob,
like
that,
you
had
a
little
bit
unusual
like
where
this
text
is
here
like
in
the
place
where
there
are
two
key
receivers
actually
created.
G
A
A
C
J
G
G
All
right
next
one
is
issue
7,
14
and
PR
740
relating
to
auth
tokens
and
stun
turn.
So
the
question
is
basically
how
this
works,
given
the
current
RTI
server
dictionary
and
our
DC
eyes
credential
type.
Presumably
this
would
be
an
RTC
eyes,
credential
type
of
token,
and
then
the
question
is:
where
would
such
a
noel
so
can
go,
how
it
at
work
so
in
RFC,
7,
6
35.
G
C
You
know
where
I
go
forward,
I'm
confused
here,
a
little
bit.
Let's
just
go
back
one
slide
so
clearly
are
no
forward
one
slide
on
it:
okay,
I
I
think
this
is
one.
This
is
the
problem
reason
I
bring.
This
up
is
a
terminology
if
you
like.
Really
this
is
not
an
OAuth
token,
which
is
what
we
often
are
thinking
of.
C
Okay,
and
it's
not
clear
to
me
whether
we're
talking
about
the
block
of
jason
from
the
open,
with
curly
brackets
to
the
close
curly
bracket
that
has
all
the
things
in
it
or
the
base64
string.
Inside
of
that,
that's
called
access.
Token
right,
and
I
think
that
that's
the
essence
of
some
of
the
confusion
here
is
there's
three
different
wildly
different
things
that
all
we
might
all
be
thinking
with
someone
says
access
token
or
token
right.
G
C
G
J
C
C
J
That's
the
ass
up,
the
only
attribute
defined
is
an
access,
token
attributes
or
stun.
So
I.
The
only
thing
II
is
the
access
token,
and
I
think
that
the
carriage
IGAs
jam
in
the
username,
but
that's
the
reason
why
I
only
added
a
single
field,
is
that
there's
only
one
piece
of
data
that
needs
to
get
past
in
that
actually
gets
sent
over
the
wire,
and
that's
the
only
we
care
about.
J
G
G
That's
more
or
less
what
does
it
get
to
proposal?
It's
more
or
less
the
two
things
we
have
before
us.
This
is
proposal
number
one
which
is
basically
to
stick
against
ad
access.
Token
expand
expiry
as
two
attributes
of
RTI,
so
g
CI
server,
that's
the
first
one.
How
was
not
enthusiastic
about
that
and
then
the
second
one
is
basically
a
little
weird
too,
because
it
basically
as
a
typedef,
so
you
can
have
either
a
password
credential
or
token
credential,
and
then
the
token
credential
includes
the
kig.
G
I
read
and
I
was
a
little
confused
by
this
one,
because
it
kind
of
it
removes
the
username
from
our
chi
CI
server,
which
seems
like
it
could
could
be
a
backward
compatibility
problem
and
then
also
there's
no
token
type
in
the
token
credential
dictionary.
So
I
don't
know
if
I
it
can.
Somebody
comment
on
either
one
of
these
which
which
one
they
like
and
why
I.
C
The
regardless
of
how
we
pass
get
the
love
in
which
structure
the
kid
ki
alg
I
thought
that
the
browser
needed
to
know
those
to
pass
the
ice
agent,
but
the
JavaScript
didn't
need
to
know
them.
But
again,
like
you,
I
I
definitely
feel
somewhat
confused.
The
more
I
dug
into
this,
the
less
I
felt
I
understood
how
it
was
supposed
to
work.
J
C
That
leaves
you
with
a
clean
mapping
to
the
current
thing,
which
I
mean.
Maybe
we
need
to
rephrase
the
names,
maybe
passwords
a
crappy
name
but
like
we
basically
have
two
fields
very
similar
to
what
we're
passing,
but
then
the
way
that
the
term
draft
also
says.
Well,
first,
you
try
basically
a
password
based
authentication
and
then
the
path,
and
then
the
term
server
tells
you
no
go,
get
an
OAuth
token
based
authentication
made
me
wonder,
and
then,
if
that
doesn't
work,
you
fall
back
to
the
password.
J
This
is
being
offensive
again.
Why
were
you
like?
Why
would
you
even
bother
getting
the
token,
and
you
know
if
you
thought
password,
would
work
and
how
you
can
get
the
password
I
don't
come
in?
Anybody
need
to
do
that
and
it's
going
to
use
the
configuration
or
something
else
that
we,
the
the
configuration
so.
A
J
Tommy
I'm,
an
author
on
this
on
this
in
parker,
see
so
I
feel
like
at
least
I
have
some
understanding
of
how
it
should
work.
Although
I'm
memories
not
perfect,
we
get
a
zero.
If
you
want
to
do
on
a
second
opinion,
but
generally
I,
you
know
I
think
we
can
read
through
this
RC
and
there's
no
place
like
on
the
wire.
It's
very
clear,
that's
the
only
attributes
that
are
actually
content
over
the
wire
fun
example.
Here,
we're
talking
set.
J
You
know
is
the
access
token
third-party
authorization,
which
is
just
the
things
then
applying
saying,
doing
certified
authorizations
and
then
the
username
match
agree
that
contains
the
ki
d
and
so
on,
since
the
user
name
attribute
already
is
in
the
caddy.
You
know
here
in
the
verdict
in
the
API.
We
can
just
allow
the
application
the
past
in
the
caddy
as
user
name
and
then
use
the
Grampa
fight
on
that
indicates
that
the
credentials
with
that
an
access,
token
meta,
password.
A
D
Human
because
Jim
and
credential.
And
when.
A
J
G
G
Issue
720
was
father
written
by
sipo.
He
wanted
a
way
to
get
the
fingerprint
of
the
generators
certificate
from
RTC
certificates,
so
in
PR
738
there's
a
proposal
to
add
a
read-only
attribute
for
RTC
DTLS
fingerprint,
and
that
includes
the
algorithm
and
the
value
in
the
PR
es.
A
couple
of
comments.
1
is
Martin
asked.
If
we
wanted
to
hash
functions,
how
would
that
work
and
I
guess
I
would
require
a
sequence
of
fingerprints
and
cullen
us
and
comments.
I
wasn't
sure
I
understood
them.
So
maybe
you
can
explain
what
how
you,
what
you
think.
C
A
C
This
would
be
a
really
good
question
for
martin
Lineker.
I
think
in
stats.
We
should
we
the
argument
with
well.
There
was
one
that
we
were
actually
using
and
relying
on,
and
that
was
the
one
we
could
report
up
and
south
like
if
we
use
the
shot
to,
and
there
was
a
shot
one
who
character
we
use
the
shot
to,
but
I
guess
its
question
two
of
what
the
use
case
of
this
is
right.
If
it's
just
too,
it's
clearly
identified
the
certificates.
C
I
guess
we
really
pragmatic
point
of
view.
This
does
not
enter
looking
harder
to
implement
to
return
all
the
information
versus
one
of
those.
In
fact,
it
looks
even
harder
to
try
and
figure
out
which
one
returned
when
you
have
to
and
like
being
able
to
get
all
the
fingerprints
just
Sarah
strikes
me
is
a
good
thing
without
any
I
have
no
good
security
argument,
so
that
just
seems
like
a
generally,
not
bad
idea.
A
E
A
G
A
G
A
G
G
A
J
E
J
J
C
Trying
trying
to
becker,
I
think,
what
we've
got
it,
we're
talking
about
the
certificate
so
ever
to
tape
it,
and,
let's
just
imagine
you
weren't,
using
stp
or
something,
and
what
were
you
for
some
other
reason.
You
want
to
be
able
to
generate
a
fingerprint
for
the
certificate,
so
we've
got
a
proposal
to
put
a
function
in
that
would
return
you
the
fingerprint
of
a
certificate,
and
then
we
came
to
the
question
of
well.
H
J
G
So
it
can
we
move
on
we're
right
alright,
so
this
is
issue
7,
26
and
PR
757.
While
we
have
ecker
on
I
guess
you
filed
this
issue,
basically
to
add
the
UT
you
frag
to
RTC
ice
candidate
into
the
things
the
serializer
emits
and
also
modify
the
done
gathering
marker,
I,
guess:
fluffy.
You
have
a
proposal
in
PR,
757
I.
C
It's
so
Justin
and
so
I
tried
to
write
as
much
so
I
could
write
down
your
proposal.
Basically
right
up
I
mean
from
the
Arctic
sea
ice
candidate.
All
it
really
happens.
We
had
in
AU
frac,
but
then,
when
we
have
this,
no
candidate
return.
We're
we
used
to
at
the
end
of
gathering
return
the
mill
candidate
instead,
I
return.
You
know
basically
an
RTC
ice
candidate
structure
which
has
a
you,
frag
filled
into
it
and
pretty
much
everything
else
or
not.
That's
not
entirely
correct.
J
Because
this
is
additive
that
you
might
have
multiples,
you
might
have
gathering
happening
for
multiple
lungs
and
so
those
my
finish
and
have
their
own
done,
and
then
there
be
a
global
done,
which
would
be
the
null
I.
Don't
care
it.
Please
don't
have
ever
compatibility
issue
was
different.
You
soon,
as
the
things
are
most
adults
over
the
null.
The
problem
is
that
app
pages
aren't
affecting
the
gift
is
like
down.
Gazmin
for
this,
given
missing
are
going
to
fit
as
a
candidate
and
like
to
close.
J
J
J
J
C
J
D
J
Event,
you
know
and
I've
never
Eddie
talked
about
and
that
these
these
types
of
asynchronous
things
don't
really
lend
us
off
the
promises
because
they
have
lots
of
yes,
there's
multiple
source
take
change
things
going
on
and
it's
real
question
is
is
is:
is
there
some
sort
of
thing
we
want
to
do
to
make
extend
existing
I,
scattering
state
change,
things
reflective
or
so
we
just
have
by
a
new
event.
I
right
I.
C
H
J
J
H
That
but
they
put
beach
all
right,
but
that
was
going
to
you
mid-level
aversion
to
use
this.
You
can
use
him
right
for
all
the
cannon.
Call.
The
ball
go
n
lines
right,
sorry
in
percocet,
so
my
understanding
one
you
can
use
the
same.
You
cry
by
special
I,
certainly
agree.
We
should
add
your
academic
candidates,
but
how
does
that
give
you
made
like
mid-level
revolution?
Because
or
is
it
because
you're
already
on
me-
because
you
might
give
you
the
same
for
all
over
this
because
it
happening
in
cat.
A
J
That's
true:
oh
yeah,
that's
right!
You
right
here!
It's
told
me
for
the
purpose
of
determining
different
generations
of
ice
restart
and
we
wanted
to
attach
to
and
a
candidate,
because
that
way
we
could
probably
understand
what
their
inner
candidates
was
for
the
current
generation
or
from
a
previous
generation.
J
But
that
puts
us
back
to
the
same
pickle
here.
Where
would
we
can
hit
any?
We
emit
an
on-ice,
the
air
on
ice
candidates
event?
He
has
done
event
and
how
do
you
know
if
it's
for
the
current
generation
or
a
genius
generation
I
guess
there
was
my
not
be
unpinned
us
we're
going
in
the
receipt
side?
I
guess
whether
it's
ambiguous.
H
Opposed
to
it,
I'm
just
focusing
on
so
that's
what
ye
like
so.
J
J
J
So
I
felt
issue
I
I'm,
not
clear
how
we
want
to
proceed
here.
You
know
as
to
whether
or
not
the
depravity
to
emit
a
object
that
you
know
is
done.
Gathering
marker
with
the
mid
or
with
witty
you
Friday
I'm,
whether
at
Canada's
needs
to
be
able
to
consume
that
where
we
just
leave
it,
as
is
where
you
know
you
just
decorate
candidates
that
are
good
thing
with.
Are
you
grab
food
and.
J
E
J
So
we
definitely
need
is
great
on
we
can
figure
out
the
end
of
candidates
from
a
good
thing
running
candidates.
What
the
right
you
fried
should
be.
So
we
know
what
you
brag.
Wat
is
for
the
most
recently
generated
candidate,
or
at
least
from
the
ftp
going
to
get
that
on.
I
scan
is
done
event.
We
know
what
methods
are
sending
of
the
wire.
I
think
the
problem
is
on
what
happens
on
the
received
side.
They
plug
into
a
dies
candidate.
J
J
Well,
if
you're,
you
don't
have
end
of
candidates
from
the
initialized
gathering,
you
could
get
to
the
segments
and
then
this
ambiguous.
What
refers
to
that's?
Why
you
want
to
you
grabbing
I,
don't
think
we
really
need
them
in,
for
anything
is
just
like
you
know,
cause
it's
not
just
simple,
nor
any
more
its
candidate
canary.
We
thought
you
might
as
well,
but
it
is
a
separate
issue.
J
J
The
question
is
whether
we
also
need
to
mid
and
whether
it's
a
primitive
nth
order
for
those
entire
cleared
action
right,
but
again,
since
we're
going
to
generating
something
that
resisting
app
handle
like
I.
Think
we'll
need
the
same
sort
of
every
compatibility
thinking
here
as
we
would
need
for
the
bid.
C
So
the
existing
apps
that
didn't
understand
this
new
behavior
just
thought
that
they
were
passing
a
sort
of
weird
candidate
across,
but
that
the
browser's
would
know
that
that
actually
was
an
end
of
candidate
indicator.
That.
J
J
Maybe
we
maybe
we
should
do,
is
I.
Take
this
PR
for
nau
Craig
and
then
I
have
a
separate
issue
to
consider
like
what
how
to
handle
lid
and
you
cry
in
the
end
of
it,
I
think
house
ever
you
want
to
represent
that
and
you'll
have
ever
decided
only
stuff,
because
we
have
the
same
issues
with
in
either
case.
B
B
J
A
So
they
are
tcp
connection
nice
event,
as
a
candidate
attribute
on
URL
attribute
this
node
there's
no
reason
why
it
can't
have
a
you
free
atoms,
but
I
think
this
is
something
away
and
if
new
proposals-
yes
justin,
is
to
follow,
follow
ticket,
then
we'll
change
the
pr2
today
and
to
say
that
what
we
add
you
fractures
are
the
fantasy.
I
scanned
it.
That's
it
for
now.
J
G
So,
in
the
issue
discussion
I
asked
whether
we
really
need
to
set
the
negotiation
needed
slag
when
cinder
track
enters
the
end
of
state,
and
there
is
some
other
discussion
basically
saying
it
should
be
up
to
the
app
the
tracks
still
attached.
The
application
could
call
replace
track
or
just
leave.
It
Harold
asked
what
happens,
what
happens
to
receiver
track
I
guess
today
it
gets
affected
through
renegotiation,
but
if
the
negotiation
needed
flag
wasn't
set,
then
it
would
be
up
to
the
application
which
could
call
replace
track.
G
G
B
G
C
J
So
come
back
to
the
ice
connection
states
one
of
the
possible
ways
you
enter.
The
disconnected
stage
is,
if
there's
some
transitive
connection
error
that
may
resolve
itself
eventually,
so
not
necessarily
losing
consent
after
the
25
seconds
or
whatever.
But
in
this
case
that
says
why
this
checks
have
failed
and
we
don't
define
anywhere
what
we
mean
by
liveness
check.
J
If
the
lower
DC
implementation,
for
example,
knows
that
the
network
interface
went
down
for
the
local
hand
at
used
in
the
currently
selected
Canada
pair
will
be
the
implementation
Tony
to
wait
for
Liza's
chest
to
fail,
even
though
it
knows
are
going
to
fatal.
You
know
that
seems
a
little
silly
when
that
the
goal
here
is
the
left
application,
though
as
soon
as
possible.
But
something
is
wrong
so
that
it
can
I
know
form
a
nice
restart
or
you
know,
inform
the
user
or
whatever.
J
So,
if
you're
the
next
slide,
our
recommendation
is
to
leave
the
criteria
for
switching
to
disconnected
up
to
the
implementation,
but
make
it
clear
that
the
goal
is
that
in
the
implementation,
those
beliefs
that
connectivity
is
lost-
and
we
can
provide
some
examples
here,
such
as
you
know-
losing
the
network
interface
for
successful,
successively
failing
consent,
checks,
for
example,
chrome
right
now-
switches
to
disconnect
it
if
I
believe
on
five
consecutive
checks
panel
and
the
rate
for
sending
those
checks
is
higher
than
that.
The
five
seconds
required
by
the
content
check
draft.
J
E
J
J
J
J
The
media
see,
I
think,
is
a
little
different
like
I.
Think
what
you
do
want,
though,
is
some
sort
of
way
of
indicating
when
the
network
might
be
broken,
at
least
from
a
consent
check
lease
term
like
a
pro
standpoint,
which
is
different
than
packet.
Stop
driving
and,
unfortunately,
3225
is
fairly
general.
As
recommendations
of
you
know
shouldn't
being
less,
it
should
be
five
seconds,
but
you
know
basically
do
you
watch
it
every
five
seconds
so
early
as
you
probably
would
determine
something
is
busted.
It
was
15.
J
Second,
wait
too
long,
so
finds
that,
like
in
terms
of
actual
magnets
checking
or
actual
sort
of
you
know,
is
this
thing
broken.
We
probably
need
to
you
know
Lisa
not
browser
to
have
their
own
logic,
or
we
need
to
run
around
sort
of
text
of
what
we
could
learn
to
be
likeness.
I
think
that'd
be
difficult
to
consider
all
the
possible
cases
right,
especially
because,
if
I
iterate
on
this
in
our
implementations,
as
we
understand
things
that
are.
C
Right
so
I
mean
this
seems
like
okay,
so
this
is
the
likenesses.
It
seems
bound
by
the
one
second
15
second
range,
which
is
right
and
recommend
five
seconds.
I
agree
with
you,
but
does
it
must
not
be
shorter
than
one
second
or
longer
than
15
seconds
and
so
I
be
stiff
like
we
should.
We
should
maybe
points
about
there
and
then
I
mean
I
guess.
The
important
thing
here
is
is
is
what
the
event
means,
which
it
doesn't
mean
to
a
single
checks,
fail
right.
It
means
that
our
consecutive
checks
of
film.
C
G
J
C
Well,
I
think
we
need
to
specify
something
here,
that's
relatively
clear
about
how
this
all
works,
just
if
the
code
that
everybody
writes
is
as
soon
as
anything
learn
loosening
if,
as
soon
as
consent
is
not
granted,
you
instantly
start
ice
restarts
I
mean
it
looks
like
this
doesn't
work
from
a
das
point
of
view.
That
means
your
average
app
will
just
constantly
gauss,
whatever
is
trying
to
not
get
caught,
which
was
the
whole
point
of
this
time.
So
I
think.
J
That's
true,
if
you
just
tried
ice
restart
when
you
list
and
send
checks,
you
know,
you'll,
send
a
few
consent
probes
and
then
like
go
to
the
end
of
the
failed
state,
but
you
won't
be
sending
the
high
bandwidth
video
traffic.
That
you
know
was
what
the
consent
form
of
your
vote.
The
content
is
to
make
sure
that,
as
you
know,
you're
getting
your
receiving
large
amount
of
bandwidth
that
you're
presenting
your
see
that
bandwidth
in
doing
a
nice
restart,
isn't
run
making
that
worse.
C
J
Can
perform
in
on
increased
alert
and
other
times
we
give
you
any
candidates
and
promise
all
right.
Yeah
I
started
to
set
that
it's
hard
to
separate
consent
was
revoked
unless
you
get
some
error
from
your
son
check
from
like
the
network
broke,
and
we
didn't
get
responses
to
our
some
packets,
in
which
case
ice
restart,
is
in
fact
indicated.
G
J
C
A
G
I
mean
all
this
is
doing
is
basically
getting
you
to
the
disconnect
state.
A
little
sooner
I
mean,
like
you
know,
RTC
disconnect
is
only
we
only
get
into
it.
If
you
have
a
nice
done,
consent
failure,
but
here
you
get
into
it
a
little
sir
you,
the
application,
is
probably
going
to
do
the
same
thing,
which
is
a
nice
restart,
but
you
get
to
do
it
a
little
bit
sooner.
F
F
G
F
So
yeah
and
I
think
having
the
disconnected
state
fire
a
bit
before
failed
is
also
good,
because
in
some
cases
people
hang
up
or
drop
or
go
away,
and
in
that
case,
even
if
they
did
not
get
the
chance
to
do
a
nice
restart.
There's
enough
information
right
before
that
happened
like
that.
There
was
something
intrinsically
wrong
with
the
connection
at
that
point,
which
may
be
disconnected
state
and
that's
enough
logging
information
to
go
and
diagnose
that
fit.
J
E
G
Next
one
is
issue
700
an
event
for
when
the
circuit
breakers
triggered
I
think
this
originally
came
from
Varun
and
I.
Think
we
are,
some
of
the
old
slides
were
going
to
get
to
finish
them
last
interim,
so
basically,
the
res
interested
in
implementing
circuit
breakers
and
at
least
as
I
read
the
ABT
core
draft
section.
45.
What
it
means
to
cease
transmission
is
depends
on
the
application
could
mean
stopping
a
single
flow
or
multiple
bundle
flows,
and
this
is
what
it
says
in
the
text.
G
Section
8
talks
about
guidance
below
bra
so
here
are
some
questions
is
the
effect
of
triggering
the
circuit
breaker?
Should
it
be
visible
to
the
app
such
as
the
in
event
were
some
change
in
the
value
of
an
attribute,
because
the
app
somehow
needs
to
know
this?
Where
does
the
browser
can
adjust
automatically
resume
sending
when
congestion
updates,
or
maybe
you
do,
a
nice
restart
or
something?
And
you
selected
pair
changes,
in
which
case
the
actors
can
sit
there
and
not
care.
G
G
F
For
the
congestion,
control
basically
should
react
before
the
circuit.
Breaker,
basically
does
anything
so
I
think
whatever
we
have
to
date
is
supposed
to
work
within
the
envelope
of
the
circuit,
breaker
and
I
believe
they
do
tend
to
work
in
that.
So,
when
congestion,
a
base
kind
of
situation,
if
the
circuit
breaker
triggered,
then
the
congestion
controls
already
lost
something
or
it
does
not
exist
or
something.
G
Else
relation
you
know
today
we
don't
have
any
congestion
events
right,
so
the
presumption
is
the
browser.
Does
the
best
again
yeah.
I
guess
the
app
can
look
at
stats
and
decide
if
it
wants
to
shut
off.
Video
things
are
really
bad.
Is
this?
Is
this
situation
same
with
a
circuit
breaker
like
the
circuit
breaker
could
trigger
and
then
it
comes
back.
Does
the
acne
to
deal
with
it?
I
think.
J
Don't
think
that's
really
true.
I
think
that
we
should
look
at
circuit.
Breaker
is
a
basically
crude
bandwidth
estimation
and
their
proof,
congestion,
control
and
the
same
way
that
you
might
want
to
know.
If
I
more
sophisticated
condition
control
says
you
only
have
100
Kelvin,
you
know
and
I'm
going
to
stop
sending
video.
F
G
C
But
but
here's
the
problem,
I
mean
this.
This
type
agree.
This
is
crewcuts
congestion
control,
but
it
happens
in
a
very
exact
point
in
time
and
it's
thought
it's
not
congestion
control.
It
stops
everything
right.
It
stops
the
flow
completely.
It
doesn't
reduce
the
speed
of
it.
So
I
think
that
that
applications
for
diagnosis
and
reporting
and
understanding
what
happened.
I'm
not
saying
the
application
is,
do
anything
with
this
other
than
understand.
C
You
know
this
is
just
you
get
it
very
precise.
In
a
specific
time,
it's
good
to
know
what
happened.
I'm
not
saying
that
I
don't
want
to
know
about
other
types
of
congestion
and
when
they
browser
decided
to
radically
do
something,
but
that
seems
like
a
much
harder
problem
to
solve
if
they
find
windows
are
to
make
some
constant
across
browsers
all
of
those
things
where
this
one's
really
crisp
and
clear.
So
it's
easy
and
it's
pretty
traumatic
event.
It
is,
this
should
never
happen
to
your
app.
C
J
Not
sure
I
imagine
you
have
like
two
video
streams
need
each
set,
the
minimum
bandwidth
them
500
childhood
and
you
only
have.
Why
is
give
a
600.com
bit
then
one
that's
not
going
to
be
transmitted
and
you
probably
want
the
same
event
that
indicates
there
and
it's
not
a
like
you
want
to
know
if
you're
going
to
be
doing
something
transmitted.
You
know
that's
why
some
that
can
occur
in
normal
operation
in
having
HS
just
or
circuit
breaker
she's,
very
ad
hoc.
C
F
C
G
F
G
J
C
I'm
not
how
about
a
purple.
What
would
you
think
of
a
proposal
that
any
time
the
browser
stops
turns
off
a
you
know:
stop
sending
media
on
a
given
stream
that
we
get
an
event
on
that
one
which
would
include
circuit
breakers
or
other
reasons.
It
might
do
the
same
thing,
and
this
is
a
good
start
in
the
middle.
D
C
Whole
argument
that,
like
I,
mean
I,
have
regularly
said
that
the
current
API
is
not
provide
enough
information
to
really
debug
and
diagnose,
like
hi
great
systems.
Right
now,
we
need
that
in
the
one
dot
0
interface,
so
like
that's
more
community
here,
does
not
really
just
like
that.
We
need
more
work
on
errors
and
analytics
and
reporting
information
of
what
you
need
to
run.
This
seems
like
an
area
that
we
all
sort
of
a
great
aspect
still
need
some
more
for
londono.
So.
G
A
C
D
J
Sure
you
know
my
phone
would
be
like
you
know.
If
we
can
do
some
stats
and
they
I
agree
that
there's
stuff
where
you
can't
tell
right
now
it
makes
it
harder
to
body.
But
if
you
can
help
recessed,
then
adding
another
API
as
a
convenience,
then
we
should
probably
you
wouldn't
figure
out
the
radius
VI.
J
G
C
But
I
mean
I
think
it's
the
matter
requirements
things
like
whether
we
can
do
it
stats
or
how
we
do
it,
I
mean
being
able
to
tell
the
difference
between
you
know.
The
far
end
fell
off
the
planet
and
your
end
degraded
your
way
down
due
to
congestion.
Those
are
two,
you
know.
That's
that's
important
information
to.
Have
us
try
and
figure
out
what
went
wrong.
We
need
to
be
able
to
differentiate
between
those
two,
but.
J
I
think
we
can
differentiate
that
you
know
I,
think
we
have
the
things
to
tell
us
like
difference
off
the
planet.
You
know
through
the
Sun
failures
and
efforts
up
already,
and
we
can
also
and
look
at
you
know,
for
each
stream
I
think
we
have
reset
to
tell
what
current
and
coded
bitrate
is
it
as
0.
It
means
there's
nothing
sense.
J
E
C
A
E
B
A
So
in
discussion
with
Randall
and
more
people,
but
even
among
others-
hey
they
found
that
I
had
this
big
notion
that
you
can
actually
given
the
stuff
that
comes
in
and
send
a
report
to
the
cv
report,
you
can
actually
calculate
back
way
at
the
time
what
the
local
time
was.
When
the
remote
event
remote
staff
have
gathered
the
and
will
disabuse
me
gently,
let
notion
or
saying.