►
From YouTube: WebRTC Working Group Virtual Interim 9 November 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
Yeah
sure
sure
we
hope
to
be
able
to.
We
saw
some
issues
and
perhaps
making
you
at
it
was
dropped
available
after
the
meeting,
and
we
will,
of
course
this.
You
should
also
send
out
the
decisions
made,
so
everyone
not
only
scored,
can
end
look
at
the
man.
3
get
their
comments.
Then
I
want
to
tell
you
all
that
we
have
limited
editor
resources
for
a
couple
of
reasons
during
november
designers
and
early.
B
So
we
asked
everyone
who
wants
to
change
this
or
raise
issues
to
respect
to,
if
possible,
also
submit
the
pull
request
to
make
life
easier
for
the
limited
resources
we
have
for
anything.
We
have
conceded
adding
edit
resources,
but
after
some
discussions
and
content
lately
we
have
decided
that
it's
for
this
short
period.
It
would
not
make
things
much
more
efficient.
A
A
So
just
looking
at
the
issues,
we
have
a
bunch
of
issues
relating
to
ice.
We
have
a
bunch
of
issues
relating
to
the
transceiver
functionality,
methods
like
direction
and
stop
what
exactly
they
do.
We
have
some
material
on
error,
handling
and
stuntin,
a
lot
which
are
from
the
previous
meeting,
so
we're
hoping
to
make
progress
on
a
bunch
of
these
things
and
make
sure
we
know
what
the
next
steps
are
all
right.
A
A
C
Just
it
turned
out
to
be
a
lot
rather
long
list
of
things
we
have
and
we're
going
to
spend,
spend
the
time
until
the
end
of
the
meeting
and
then
we'll
stop
and
but
I
just
encourage
you
to
it.
If
we
detect
that
we're
not
getting
anywhere,
that's
just
the
table,
they
issue
and
turn
out
turn
to
the
next
issue.
C
C
A
So
just
a
reminder:
transceiver
that
stop
is
irreversible.
So
it's
not
something
you
can.
You
can
reset
transceiver
that
stopped
if
false
and
it's
the
sender
will
no
longer
stand.
Receiver
llamala
receive
a
negotiation
needed
flag
is
set.
Transceiver
that
set
direction
said,
that's
the
direction,
but
does
not
take
effect
immediately.
A
Instead,
the
future
calls
mark
the
media
description
as
described
in
Jessup,
and
the
negotiation
needed
sweat
set,
so
we're
going
to
get
into
both
of
these
in
a
few
minutes,
but
just
to
remind
people
where
we
are
also,
we
have
set
parameters
and
we
have
encoding
stacked,
off'
set
parameters.
Again
the
text
says
doesn't
call.
This
renegotiation
can
only
be
used
to
change
with
the
media
second
sending
receiving
within
the
envelope
just
to
remind
people.
There
is
only
set
parameters
for
the
sender,
not
the
receiver.
A
You
can
get
parameters
of
both
but
only
set
on
the
sender.
We
also
have
encodings
detective,
and
that
means
for
a
sender.
The
encoding
is
being
sent,
setting
it
to
false
means.
It's
no
longer
sent
for
not
XP
receiver.
There's
some
interesting
texts
there.
It
indicates
the
assistant
coding
is
being
decoded.
There's.
Additionally,
some
statements
about
setting
it
to
false
or
setting
it
to
true,
since
there's
no
receiver
that
same
parameters
method,
that's
really
not
possible.
A
So
there's
a
PR
926
that
cleans
that
up
a
bit
just
to
say
that
it
reflects
whether
the
encoding
is
being
decoded
or
not
alright.
So
issue
871
is
what
happens
when
transceiver
dot
stop
is
called.
So
we
had
a
question
about
whether
when
it
when
you
call
transceiver
that
stopped
you
have
an
up
and
in
bit
event
for
transceiver
receiver
trap
and
then
Adam
pointed
out.
We
didn't
really
have
it
off
detailed
steps
to
know
exactly
what
happened.
So
we
resolve
this
one
instead.
So
here
is
what
the
proposed
text
is
to
be
added.
A
The
transceiver
dad
stop
description,
so
we
say
that
stopping
a
transceiver
is
irreversible.
It
doesn't
say
that
explicitly,
although
I
think
most
people
seem
to
understand,
that's
what
it
does
and
we'll
the
rest
of
it
is
is
straight
from
Jessup
will
cause
future
calls
to
generate
a
zero
port
as
defined
in
Jessup
when
the
method
is
invoked,
the
user
agent
must
run
the
following
steps.
A
So
at
first,
if
it's
already
stopped,
you
abort
those
steps,
because
you
don't
need
to
stop
it
again.
Then,
if
the
connection
is
closed,
you
throw
an
invalid
state
error
exception
because
no
point
in
calling
stop
on
a
close
connection.
You
stop
sending
media
with
a
trend
with
the
sender,
stop
receiving
it
on
the
receiver.
You
set
the
ready
state
of
receiver
track
to
end.
It
set
that
stuff
to
true
and
then
mark.
The
connection
is
needing
negotiation.
B
A
So
there's
a
whole
bunch
of
examples
of
where
the
term
stopped
is
used
and
a
question
that
Taylor
brought
up
is
what
exactly
that
means,
because,
as
I
mentioned
previously,
we
have
transceiver
dot,
stop,
we
have
set
direction,
so
we
can
set
things
to
receive
only
a
send
only
and
then
we
also
have
this
inactive.
And
it's
not.
It
wasn't
exactly
clear
what
what
of
all
of
these
things,
where
we
mean
when
we
say
stopped
so
some
examples
in
enclosed
method,
we
say:
senders
and
receivers
are
now
considered
stomped
a
remove
track.
A
We
say:
if
center
stop
the
board,
be
steps,
and
then
we
say:
stop
sender
in
52,
we
say
receive
a
sender
can
be
stopped,
which
indicates
it
will
no
longer
send.
We
have
replaced
track
and
sender
stopped
returning
a
promise
with
invalid
state
error,
some
more
examples
5
for
it,
and
this
relates
to
the
transceiver
dot
stopped
attribute.
We
say
it's
true
of
either
stopped
has
been
called
or
we've
caused
the
receiver
to
be
stopped.
A
This
is
a
little
bit
puzzling
and
then
we
have
insert
DTMF
where
we
say
if
senders
been
stopped
through
an
exception
or
extenders
extenders
been
stopped
aboard
these
steps,
so
we're
trying
to
figure
out
what
exactly
these
things
correspond
to
in
terms
of
these
methods
that
we
and
attributes
that
we
have-
and
it
is
a
it-
is
a
wee
bit
tricky
but
PRS,
916
and
918
described
the
following
guesses
or
the
following
text:
substitution.
So,
instead
of
staying,
the
senders
and
receivers
are
now
considered
stopped.
A
We
say
set
transceiver
dot
stopped
to
true,
because
this
is
closed,
we're
closing
the
the
peer
connection.
It
seemed
to
be
something
that's
permanent,
so
that's
what
we
use
transceiver
dot
stopped
there
in
51
we
ever
move
track
and
I
think
a
damn.
They
have
some
questions
about
it
and
we
think
they're
that
this
was
referring.
This
could
be
referring
to.
A
We
say
centers
that
is
stopped,
then
aboard
these
steps,
and
we
say
instead
of
transceiver
dive
directions
receive
only
are
in
active
than
aboard
these
steps
and,
if
fever
that
direction
and
then
set
transceiver
that
direction
to
receive
only
there
is
an
issue
with
that
because,
as
Taylor
will
describe
just
because
direction
is
receive,
only
an
active
doesn't
mean.
The
sender
is
actually
not
sending
because
it
takes
effect.
It
doesn't
take
effect
immediately.
So
this
isn't
actually
quite
right
and
we
will
get
into
maybe
how
to
fix
that
in
a
little
bit.
A
But
the
idea
here
is
just
to
understand
in
remove
tracks
where
we,
you
know
we're
we
actually
talking
about.
We
think
it.
This
is
only
about
the
sender,
so
we
wouldn't
want
to
you
know,
said
transceiver
dot
stop
to
true
or
test.
Maybe
the
first
one
is
whether
transceiver
dot
stopped
is
true.
It's
a
little
bit
of
uncertainty
there
about
what
exactly
that
we
are
trying
to
do
there
all
right
in
remove
track
so.
C
C
F
G
A
Yeah
so
I
guess
the
question
is:
Harold
I,
whether
these,
whether
these
actions
are
awaiting
negotiation
or
whether
they
were
supposed
to
refer
to
the
immediate
state
if
they
are
awaiting,
if
they're
willing
to
wait
for
negotiation,
then
I
think
they're
correct
is
is,
but
if
you
know,
if
senders
stop
to
board
these
steps
right,
if
we're
looking
to
know
whether
the
sender
can
send
or
not,
you
can't
get
that
French
and
ce4
that
direction
and
Taylor
will
talk
more
about
it.
Dip
it.
But.
A
Can
know
if
you
can
note
those
issues
handle
that
would
be
grated.
Chris
I'll,
just
I'll
just
leave
that
where
it
is
because
I
don't
really
have
an
answer
right
now.
Okay,
so
some
more
examples.
If
I
to
ascend
there
can
be
stopped,
which
indicates
of
am
no
longer
send
media.
The
proposed
alist
is
just
to
delete
this
because
we're
just
not
sure
whether
it's
referring
to
direction
or
transceiver
dots,
copter
or
what
and
doesn't
seem
to
be
particularly
important.
We
also
had
a
question
about
replace
track.
A
There
we
say
is
its
top
returner
promise
and
there
we
rethink,
but
maybe
not,
maybe
not
exactly
that
it
that
this
is
equivalent
and
transceiver
dot
stopped
being
true,
because
they're
rejecting
the
promise
and
if
the
sender
was
just
you
know
if
it
was
just
receive
only
or
inactive,
you
shouldn't
reject
with
an
invalid
state
ever
that
make
sense
to
you.
Harold.
A
That
this
one
is
referring
to
transceiver
that
stop
okay
and
then
the
last
one
is
actually
the
definition
of
stopped
itself,
and
Taylor
pointed
out
that
this
looks
like
a
typo.
It
should
have
said
has
caused
the
transceiver
to
be
stopped
rather
than
a
receiver,
because
I
don't
know
what
the
receiver
would
have
to
do
with
it.
So
it
kind
of
makes
sense.
The
text
omit
does
make
sense
if
you
substitute
transceiver
for
receiver.
A
Ok
and
then
yet
more
examples
in
insert
DTMF.
We
have
again
a
sin:
Desmond
stop
drone
valance
pain
and
we
think
that
translates
to
return
c.
For
that,
stop
is
true,
and
then,
if
senders
stop
tube
or
three
steps,
that
one
is
a
little
tricky.
We
think
that's
stop
this
true
as
well,
but
then
the
overall,
the
problem
comes
up.
What,
if
the
sender
can't
send
right
if
it's,
if
the
actuals
or
if
the
directional
negotiated
direction
is
receive,
only
are
inactive?
A
H
A
A
B
A
B
A
B
B
B
A
G
G
So
this
is
the
access
token.
This
is
the
motor
out.
Both
client
received
these
old
parameters
from
the
authentication
servers,
and
these
are
the
information
that
the
application
receives
from
the
autistic
issue,
Sarah
and
and
from
this
information.
Not
all
information
is
used
during
the
stump
client
authentication.
On
the
access
token,
the
key
I
did
the
kid
and
the
key.
This
is
the
neck.
G
This
is
what
is
used
as
the
password,
so
it
is
used
in
the
message:
integrity,
cooperation
and
only
this
information
need
to
pass
down
to
the
ice
agent,
in
my
view,
and
and
and
all
the
other
informations
need
to
be
handled
by
the
application
by
the
old
client
part
of
the
application.
So
on
the
next
slide,
I
created
a
figure
about
this,
so
the
old
client
module
duty
is
to
renew
and
and
renewed
access
token
and
the
span
turning
the
ice
agent.
Only
getting
the
information,
these
three
information
and
these
three
information.
G
If
we
reuse
the
username
as
the
kids
at
the
mackie
as
the
credential
and
creating
a
new
access
token
in
the
interpretive
attribute,
then
we
can
pass
down
to
the
ice
agent.
This
disinformation.
I
feel
a
gap
between
the
the
actual
FC
and
the
web
RTC
api,
so
the
RC
is
only
defining
the
external
observable
behaviors
and
not
defining
the
the
behavior
and
the
interaction
between
the
stunts
on
the
ice
agent
and
the
old
client
part.
G
So
what
defining
this?
The
old
client
is
needing
some
information
from
the
stanton
client.
It
needs
the
support
either
message:
integrity,
August,
so
the
H
make
arguments
that
standard
client
soft
words,
because
it
is
needed
to
determine
the
key
light.
What
is
the
Mackay
like
the
dell?
Precision
server
will
generate,
create
the
session
key,
I'm
and-
and
this
way
I
think,
maybe
it
is
worth
to
write
them
and
something
about
about
this
I,
don't
know
your
opinion
to
write
and
clarify
more
these
things,
because
it
is
not
easy
to
understand.
G
I
think
the
actual
rescue
so
give
more
examples,
and
speaking
more
about
the
broader
context,
fold,
the
old
client
and
the
stun
client
interested
each
other
on
the
next
slide.
I
I
explained
somehow
these
functionalities,
so
the
the
old
client
part
is
to
store
the
old,
related
parameters
and
and
reused
access
program,
and
the
ice
agent
need
to
fess
up
during
the
the
piece
collection
creation.
G
The
support
is
done,
turn
the
HVAC
average
and
and
I
need
to
get
the
key
domeki
and
the
access
token
parameters,
and
when
the
old
client
renewing
this
information,
then
it
through
the
web
RTC
configuration
interface.
It
should
be
passed
down
to
the
new
new
credentials
that
you,
you
know
access
token
to
the
response
on
client.
D
G
Off
days,
actual
commercial
RF
seems,
you
say
that
that
in
the
in
the
in
the
example
part,
there
is
a
few
lines
about
the
the
client
h
make
art
will
need
to
determine
deshmukh
tilak
and
because
the
spam
piece
has
some
hash
fatality
and
accomplish
this
edge
right.
Lizzie
computing
message,
integrity
using
both
h,
medic
sh1
and
the
new
message
hundred
fifty-six
algorithms
or
somehow
it
killed
them,
set
the
key
length
that
it
is
creating.
This
is
the
only
only
only
thing
why
these
measures
listing.
G
It
needs
to
do
to
get
the
information.
What
is
the
longest
key
length
that
so
what
is
the
strongest
are
guarding
that
which
is
supporting
and
what
is
the
longest
keel
and
acting?
So
if
it
is
supporting
each
H,
Mac
SH
256,
then
it
is,
it
is
giving
that
def
k
key
and
if
it
is
needed
them
to
create
either
shorter
key
from
that.
Well,.
E
Let's
say
the
term
server
supports
new
Shaw
10,000,
okay
in
the
future
mom.
If
it
shows
the
authorization
server
in
its
relationship
with
them,
hey
I
support,
shot.
10,000,
then
the
authorization
server
could
generate
tokens
that
had
long
enough
material
to
be
usable
in
everything
it
supports,
including
shot
10,000
I
mean
I,
think
I.
I
I'm
get
I
think
that
that's
how
we
thought
this
was
going
to
work
long
ago.
I'm
not
saying
that's
how
it
should
work,
I'm
just
sort
of
like
I,
don't
understand
this
well
enough,
but
that
so.
E
G
E
E
Use
whatever
the
minimum
life
was
for
sha256
today
for
everything
you'd
be
good
to
go
and
you
can
ignore
all
of
it
and
that's
what
several
implementations
do
so,
but
the
question
is:
is
how
do
we
design
this
such
that
we
can
add
a
new
hash
algorithm
later
that
I
think
that
that
should
be
our
task
of
whether
we
got
the
standard
right.
Hey?
Can
we
add
new
crypto
later
without
breaking
everything.
G
But
the
one
decline,
though,
was
support
these
Dembele
position
server.
If
your
signal,
through
dolt
client
to
the
authorization
server
that
AI
I
need
the
h1
were
consumed,
algorithm
and
I.
Supporting
this
on
and
the
all
transition
server
will
know.
Maybe
that's
okay,
the
grades
I,
don't
know
1,000
digit
story
characters
length
key,
and
this
way
it
will.
It
will
give
back
this
kind
of
key
to
go
to
the
stanton
client
through
the
Rocky,
Mountain
I.
Guess.
E
G
E
E
I
just
suspect
that
when
we
wrote
some
of
this
stuff,
we
were
thinking
about
it
being
the
other
way
where
you
know
the
authorization
server
is
tightly
bound
to
the
term
server
in
many
ways
and
that
those
two
would
be
coordinated
to
know
the
information
they
needed
to
know
about
each
other.
To
do
the
right
thing
on
it
on
things
like
this.
G
D
G
E
I
mean
note
that
they
hit
authorization
server.
The
term
server
need
to
share
cryptographic,
information
and
configuration
information
with
each
other,
or
else
the
authorization
server
can't
create
valid
tokens
that
the
term
server
will
accept.
They
can't
I
mean
I'm,
not
saying
they
have
a
live
connection
between
each
other,
that
happens
on
every
single
call,
but
they
have
to
be
tightly
configured
together.
Others
doesn't
work.
G
G
Know
a
lot
of
the
u.s.
age,
so
the
new
age
might
have
branches
in
the
news
thumbs
up,
so
it
is
coming
hi
guess
so,
I
think
at
least
for
me
to
think
about
this.
So
this
is.
Why
told
that
the
it
is
important
to
signal
somehow
from
this
number
a
client,
this
information
back
to
the
authorization
server
according
to
RFC
example,.
G
G
Dictionary,
so
this
only
distilled
information
of
needy,
and
this
is
what
I
figured
my
proposal
and
might
be
poor
request
according
opportunities
to
somehow
extract
this
information
from
the
spam
and
pass
up
and
pass
down.
The
user
named
credential
and
excess
volcom
make
the
credential
type
is
open.
G
G
G
That
I
also
added
some
correction
that
the
that
the
in
access
token
is
also
needed
if
the
preneur
type
is
a
spoken
and
and
should
progress
as
error
and
the
board
that
if,
if
the
access
token
is
omitted
also-
and
these
are
mainly
my
my
changes
I-
I
have
also
one
question
about
the
about
to
discovering
span
term
server.
So
what
happens
if
the?
G
C
So
I
see
that
my
name
is
attached
to
the
comment
about.
Is
this
the
way
we
want
to
access
to
do
things
in
the
future
and
with
adding
one
more
token
and
will
one
more
element
dictionary?
Every
time
we
had
the
new
mechanism
and
I
found
off
the
oldest
excursion
I,
don't
care
that
much
about
it
anymore.
E
Mean
I
would
might
see
an
object
instead
of
a
another
parameter.
I
am
concerned
that
we
might
have
more
stuff
that
we
need
to
push
down
here
and,
having
like
way
a
dictionary
or
way
to
contain
that,
rather
than
gaining
it
is
more
attributes
in
the
top
level,
are
g
CI
server.
It
seems
like
it
would
be
time
well-spent.
G
Justin,
what
what
are
these
disinformation?
This
is
what
I
have
mentioned,
that
the
old
client
parameters,
not
all
informations
ladies
dreams,
authentication
them
and
over
the
fly
on
crease,
in
my
opinion,
need
to
need
to
handle,
handle
this
and
store.
This
information
understand
client
inside
the
inside.
The
browser
notarized
agent,
so
I
think
only
the
needed
parameters
are
these
three
all
right.
E
G
E
Think
that's
fine,
like
the
idea
would
be
a
credential
type.
Would
then
tell
you
what
the
value
of
credential
would
be
that
credential
type
is
o
auth?
Then
you
get
a
token
object
if
its
password,
then
you
get
a
dom
string,
which
is
the
password,
and
I
think
that
would
sort
of
set
us
up
well
for
if
we
need
to
add
more
things
for
the
oh
I
hope
you
have
different
credential
types
in
the
future.
E
You
know
we
don't
need
to
have
all
the
stuff
packed
into
the
top
level.
Dictionary
seems
like
the
color
magic
seems
like
a
good
change
to
me.
I.
You
know.
I
also
want
to
make
sure
that
the
credential
type
is
not
token
prices
like
we're
going
to
have
different
types
of
token
later,
so,
let's
call
it
an
oauth2
token
or
whatever
we
let's
make
a
name
for
the
type
of
token
it
is.
E
C
G
C
What
this
one
there's
one
interesting
property,
which
is
that,
if
you
put
in
a
union
and
side
inside
a
dictionary,
then
object
gets
more
complicated
to
have
more
variance
in
the
union
and
not
get
a
type
errors.
Then,
if
you,
then,
if
you
add
more,
add
more
names
inside
the
dictionary
so
for
expensive
ility
as
adding
more
attributes,
might
action
might
actually
be
slightly
better?
All
this
I'll
defer
to
anyone
has
feels
more
strongly
about
how
to
write
web
ideals,
but
that's
my
impression.
After
having
expelled
a
few
ideal
things.
B
E
E
Mean
if
we
do
discuss
it
with
it
back
to
me,
caes
original
point
I
think
we
need
to
write
some
text
that
clearly
explains
like
how
you
deal
with
it
and
why
you
don't
need
why
you
don't
need
this
information,
how
you
get
it
because
it's
just
going
to
cause
confusion.
Other
words
wait
so
I
don't
really
care
whether
we
do
whether
we
export
or
not,
but
either
way,
I
care
that
we
have
enough
tax,
that
people
can
figure
out
how
it
works,
which
we
we
don't
right
now,
good.
B
E
Right
well
right
now,
if
you
always
return
a
256-bit
key
things
we
find,
but
we
should.
You
know,
write
some
text
on
find
out
for
this
document,
but
there
should
be
some
way
of
understanding
what
how
to
turn
server
communicates
with
the
officer
as
colin
said,
so
that
it
can
tell
it
what
size
key
to
return.
C
E
C
G
Okay,
I
just
want
one
question
question
about
the
the
the
earth
see,
I
think
what
I
had
mentioned.
Is
it
worst
into
to
write
something
about
this
or
you
are
fine
to
skip
this?
Also
writing
down
what
happens
inside
of
our
CC
browser
context,
this
third-party
authentication
for
Stan.
So
what
are
the
differences,
because
the
actual
RC
is
not
speaking
about
about
the
stand,
client
and
old
client
interactions,
so
it
is
not
to
be
telling
this.
E
Look
I
think
it'd
be
really
good
in
this
web
RTC
dr.
have
some
is
probably
non-normative
taxes,
probably
just
some
explanatory
text.
That
explains
how
the
whole
thing
sort
of
fits
together,
because
people
trying
to
use
this.
This
is
going
to
be
the
place
where
they're
sort
of
reading
something
right
so
I
mean
hey.
Look
your
explanation
of
how
all
fit
together
today
and
those
made
it
much
clearer.
Let's,
let's
just
like
write
a
paragraph
on
that
and
put
it
in
this
document.
That
explains
how
the
whole
system
works,
because
it
didn't
look,
there's.
Never.
E
We
don't
have
a
good
document
to
put
the
system
level
information
in
this
I
mean
I,
guess
we
could
put
it
in
the
RTC
web
overview
document
or
security
document
or
something
I
don't
know.
Maybe
it
should
go
there,
but
I
think
between
Steph
and
Harold
hi
others
we
can.
We
can
figure
out
where
to
add
stuff
right.
All
right,
I
think
we
could
to
raise
write
this
down
now.
So
we
all
as
a
group,
understand
this
and
then
I
can
figure
out
exactly
what
document
it
ends
up
in
so
right.
E
C
C
B
E
Sure
so,
on
this
one
I'm
going
to
get
on
in
the
next
two
slides
to
the
actual
meat
of
it.
So
at
the
tpac
meeting
we
started,
we
agreed
in
the
broad
sense
to
add
these
more
definitive
errors
to
be
able
to
differentiate
between
different
cases
where-
and
you
know
highlight
that-
that's
you
know
where
the
actual
application
program
needs
to
be
able
to
understand
what
happened,
not
just
reporting
for
humans
for
future
debugging
processes,
or
things
like
that,
because
those
can
go
in
this
sort
of
comic
strips.
E
So
I
wrote
up
a
drop
on
that
those
sort
of
our
a
PR
and
sort
of
took
my
first
initial
stab
of
that,
and
and
thank
you
kindly
to
Dan
brunette,
who
did
a
lot
of
the
work
and
heavy
lifting
on
getting
that
structure
set
up
for
the
error
from
it.
Just
just
coffee,
the
you
know.
It's
the
same
way
we
did
and
they
get
user
me
to
drop
so
too.
If
she's
got
raised
on
this,
but
I
think
are
worth
talking
about
so
next
slide.
Please
25.
E
So
the
first
was
on
summit
on
on
stp.
If
there
was
a
parsing
error
of
the
SVP
and
it
was
invalid
stp,
I
included
reporting
back
the
line
number
where
the
parser
hadn't
come
first,
but
this
impacts
our
and
yeah
I.
Think
I,
you
know
Harold
reasonably
asked
well,
hey
is
that
is
that
needed
for
applications
to
actually
react
to,
or
is
that
just
a
convenient
for
debugging
point
of
view,
for
you
know
that?
Does
that
mean
for
Cubans?
Or
is
that
mr.
machine?
E
Basically
the
questions,
because
it's
a
good
point
I,
think
that
why
this
gets
used
to
repair
from
is
stp
doesn't
really
have
ways
to
express
mandatory
to
understand
attributes.
So
sometimes
people
put
in
proprietary
extensions
that
use
something:
that's
not
valid
syntax.
What
the
idea
it
will
cause
a
syntax
error
if
it's
received
by
something
that
doesn't
understand
it
and
the
usual
approach
to
dealing
with
that
is
to
just
yank
out
either
those
lines
or
yank
out
the
whole
in
line
with
the
whole.
E
You
know
in
line
section
that
corresponds
to
and
just
zero
out,
of
course,
that
it
is
on
and
basically
not
negotiate
and
you
see
is
happening.
Sometimes
where
there's
you
know
a
baseline,
audio
and
video
stream,
and
then
maybe
there
was
some
proprietary
things
for
multi-screen
systems
or
SVC
or
other
thing
layered
on
top
of
it,
so
it
there
is
a
very
direct
reaction.
You
can
take
to
it
when
you,
when
you
get
these
errors,
I'm
not
saying
most
apps
wouldn't
want
to
do
that.
E
Most
apps
aren't
going
to
be
dealing
with
that
type
of
essence
East,
but
should
be
very
convenient
to
know,
and
it's
something
that's
working
fresh,
it's
really
hard
to
imagine.
The
browser
doesn't
really
know
that
when
it's
when
it's
generating
the
error,
in
fact,
they're
already
sort
of
coding
it
in
the
text
string,
this
is
just
moving
it
out
to
to
arrow
straight
stuff.
So
that's
one
of
my
proposal
on
that
people
got
a
box
or
issues
on
whether
we
should
add
this
or
not
so.
C
If
this
explanation
actually
made
made
a
bit
yourself
and
they're
just
my
concerns
about
whether
with
whether
it
was
was
valuable,
it
also
gives
us
a
good
heuristic
for
what
line
numbers
return.
If
we're
not
sure
about
what
the
like,
what
line
is
witness
as
long
as
the
line
number
is
on
the
right,
&
M
line,
and
then
it's
up
to
doesn't
matter
right,
I.
E
Think
in
here'll
I
think
it
was
a
very
valid
comment
on.
All
of
these
is,
like
you
know,
hey.
Is
this
really
needed
right?
I
mean
right
now
we
should
that's
what
the
question
we
should
be
asking
about
these
errors.
Is
it
really
need
it?
So
the
next
one
for
the
next
slide
26
this
one,
depending
on
how
what,
depending
on
to
accuse
us
of
how
identity
works,
this
one's
either,
probably
not
need
it
or
probably
need
it.
It's
a
little
gray
either
way.
So
let
me
explain
what
it
is.
E
E
Well,
if
you
call
you
can
certainly
call
set
at
any
provider
to
not.
The
question
is,
is
does
that
set
to
identity
providers
and
we
try
and
insert
identity
assertions
from
both
of
the
identity
providers
from
the
calling
side
and
then
the
caller
could
use
either
one
of
those
to
verify
which
ever
wanted
to
trust
it
either
one
of
them
or
if
you
call
it
twice,
that's
going
the
second
time
just
overwrite
the
identity
provider.
E
That
was
that
the
first
time
the
drop
is
big
on
that
topic,
because
we've
always
been
a
little
bit
clear
and
we're
missing,
probably
some
of
the
texts.
I
think
that
the
way
things
are
moving.
We
certainly
want
to
keep
open
options
of
having
multiple
identity
providers
and
we
probably
want
to
support
multiple
identity
providers.
Part
of
the
changes
needed
to
support
multiple
identity
providers
was
the
fingerprint
draft
in
the
SBP.
E
That
christopher
has
been
writing
to
very
much
clear
up
and
allow
multiple
fingerprints
if
it
doesn't
make
sense
without
that,
so
we
needed
that
sorted
out.
First,
that's
getting
sorted
out.
If
we're
going
to
have
multiple
identity
providers,
then
they're
going
to
have
different
login
URLs
and
if
you
just
start
at
the
browser
you're
not
logged
into
either
of
your
identity
providers,
you
try
and
make
a
call,
and
you
get
your
going
to
get
back
to
errors.
You
know
log
in
at
any
provider
a
login
identity
provider
be.
E
We
could
probably
make
this
work
with
the
single
attribute
thing
that
we
have
today
where
it's
an
attribute
in
one
of
our
objects,
and
it
doesn't
come
back
in
the
air
and
we
can
probably
say
well,
you
know
the
last
one
wins
you
try
again
and
you
try
again
you
get
another
error
back
and
you
get
a
different
login
URL
in
the
air
condition
this
time
and
you
use
it
is
I'll
see.
E
If
we're
going
to
have
multiple
identity
providers,
I
think
there's
a
strong
argument:
that's
probably
cleaner
to
just
return
the
URL
in
the
air
or
object,
because
it's
when
you
get
the
error
object
is
exactly
the
time
you
want
the
URL
and
you
want
to
do
something
with
it.
So
I
think
if
we're
going
to
go
with
most,
if
we
view
having
multiple
gaming
fryers
in
the
future,
there's
a
very
strong
argument:
all
right
make
a
strong
argument
for
putting
it
in
the
air.
E
If
we're
only
going
to
only
have
one
identity
provider
ever
but
I
think
it's
a
toss-up
whether
it's
the
attribute
like
we
have
it
today
or
we
added
in
the
error
it
sort
of
doesn't
make
a
big
difference
to
me.
So
let
me
first
of
all
ask
did
that
explanation
makes
sense
and
then
we
can
get
into
well.
What
should
we
do?
First
of
all,
the
people
even
understand
what
I
was
talking
about.
I.
C
Otherwise,
applications
written
for
one
browser
is
mother
browser
and
that's
not
safe
at
all.
It
will
define
to
say
that
sir
there's
a
defined
editor
to
get
if
you
try
to
set
to
a
data
providers,
but
it's
not
fun
to
say
that
PS
this
works
without
the
matter,
with
the
best
different
effects
on
two
different
houses.
E
Okay,
totally
agree;
okay,
so
I
mean
I,
guess
the
question
is,
is
so
I
mean,
should
we
leave
this
sort
of
pending
open?
The
question
is:
should
I
open
a
separate
bag
of?
Do
we
allow
multiple
identity
providers
and
then,
pending
the
resolution
of
that,
come
back
to
try
and
decide
whether
we
put
this
one
attribute
in
the
in
the
error
or
not
or
is
it
would
that
be
the
right
way
to
proceed.
H
E
Okay,
we
don't
have
Martin
on
the
car.
Do
we
for
a
cur.
E
So
Arab
lemme
sorry
propose.
We
all
do
to
move
forward.
I
will
pull
this
error
string
out
of
the
pr
request
on
errors?
Ok,
because
it's
just
clearly,
we
don't
know
what
to
do
with
it
yet
so
that
we
can
move
forward
that
PR
recess.
Otherwise
I
will
open
a
new
bug
which
is
about.
Do
we
allow
multiple
identity
providers
and
I
will
open
a
new
bug
about
this
specific
error,
return
thing
and
I
will
mark
it.
E
Can
hear
crickets,
ok,
I
return,
the
calls
Legere,
ok,.
C
D
D
D
D
D
D
D
So
a
one
thing
I
wanted
to
note
before
we
move
on
is
that
I
noticed
that
as
J
subs
currently
written
is
the
only
way
transceivers
removed
from
the
peer
connection
set
of
transceivers.
So
even
if
it's
stopped
and
of
you
know
that
mines
rejected
and
it's
recycled
by
something
else,
the
transceiver
would
still
stick
around
just
in
the
stopped
unassociated
stage.
I,
don't
think
it's
really
problem
is
just
something
should
be
aware
of.
D
So
this
year
here
is
that
the
dirt
basin
needed
flag
was
section,
was
still
written
in
terms
of
tracks
and
as
opposed
to
transceivers,
and
now
you
know,
transceivers
are
we're
all
done.
Stp
negotiation
is
encapsulated
and
allspice
mountain.
There
are
some
places
on
hot
where
the
flag
was
set
are
cleared
where
it
shouldn't
have
been
or
not
set
are
cleared
where
it
shows,
then
so
I
tried
to
sort
of
just
how
clean
everything
up
and
rewrite
it
in
terms
of
transceivers.
D
So
now
the
flag
is
only
updated
mspr
when
applying
an
answer
or
modifying
the
peer
connection
in
the
stable
state.
So
if
you,
you
know,
do
something
in
a
non
stable
state
and
nothing
happens
until
you
apply
an
answer
and
then
you
can
figure
out,
you
know
is
additional
negotiation
necessary
or
not.
So,
for
example,
previously
the
flag
was
cleared
when
create
an
answer.
I,
don't
think
that
makes
much
sense,
I,
just
creating
an
answer.
Would
you
know
affect
the
state
of
the
peer
connection,
I.
D
Okay,
good
and
also
one
thing
I
removed
was
in
the
description
of
RTP
receiver.
It
mentions
that
open
attributes
are
nullified,
a
negotiation
is
triggered
and
I.
Don't
think
we
have
anything
on
the
receivers.
You
can
modify
the
trigger
negotiation,
you
need
so
I,
don't
think.
That's
not
anymore.
Yeah.
D
Alright
next
slide,
then
we
can
go
into
the
sort,
detailed
criteria.
I
wrote
for
weather
in
negotiations
be
or
not
the
first
two
rules
here
for
going
fermentation,
specific
negotiation
and
the
Athenians
data
section
are
just
carry
over
from
the
floor.
I
think
that
as
meat
of
it
is
the
rules
for
the
transceivers,
so
for
each
transceiver,
you
need
to
negotiate
this
first
of
all
the
transceivers
india
associated,
which
means
that
hasn't
appeared
in
the
description.
D
Yet
that's
pretty
clear
if
the
direction
is
has
a
send
component
and
there's
not
an
msid
line
in
the
current
local
description,
which
is
the
same
as
the
rule
for
tracks
before.
If
you
have
a
track
you
need,
and
in
this
ID
for
it
and
their
local
description
now,
just
you
know,
converted
transceivers,
essentially
all
right
on
the
next
slide.
D
We
move
on
to
more
rules
for
transceivers
I
think
this
is
where
it
gets
a
little
complicated
and
may
need
to
figure
out
what
we
mean
by
direction
before
we
really
answer
this,
but
the
goal
these
rules
is
that
negotiation
is
is
not
set
to
true.
If,
if
an
additional
negotiation,
which
you
know
not
result
in
the
desired
direction,
so,
for
example,
it
gives
you
the
example
there
below
an
a
bullet
point.
D
If
the
direction
is
set
to
send,
receive
and
I
get
a
remote
offer,
that's
receive
only
resign
in
negotiated
direction
of
spend
only
even
though
the
transceivers
direction
does
not
match
it
was
negotiated.
It
wouldn't
make
sense
to
send
me
ago
she
a
shammy.
Is
it
true,
because
you
know
the
remote
peer
is
only
offering
to
receive
and
we
we
know,
or
we
can
reasonably
expect,
that
we
would
get
the
same
negotiated
resolve.
D
D
So
any
comments
on
that,
if
you
can
try
to
take
a
little
time
to
absorb
the
rules,
but
we
have
I
split
this
into
different
rules
for
the
offering
answer
where,
if,
if
here's
the
current
description
is
an
offer,
then
the
direction
has
to
match
either
the
author
or
answer
me.
If
I
offered
you
know
x
and
you
answered
with
why
in
my
current
direction
is
x,
I'm
just
going
to
happen
except
that's
moving,
negotiated
directions.
A
E
A
Ok
so
anyway,
so
you
gave
your
example
there,
and
so
let's
say
then.
The
next
thing
that
happens
is
that
the
the
remote
offer
is
send
receive
right.
So
after
this
you've
gone
through
this
first
example,
then
again
you
get
another
offer
that
send
and
receive.
One
could
say
that
you
know,
since
the
transceiver
Direction
has
already
sent
and
received
that
nothing,
nothing
new
would
happen.
There'd
be
no
attempt
to
negotiate,
even
though
there
should
be
but
you're
saying
that
that
won't
happen,
because
the
because
it
does
because
the
direction
doesn't
match
the
answer.
D
D
A
Guess
I
was
thinking
that
right
at
sea,
it's
the
offeror,
write
the
offer
sent
the
offer
said
receive
only
the
offer,
changes
his
mind
right.
So
then
the
question
is:
how
does
he
indicate
that
and
how
does
that
cause
a?
How
does
he
change
his
mind
to
send
receive,
come
and
have
that
cause
a
renegotiation
of.
E
D
So,
in
this
case,
if
the
remote
peer
was
using
the
peer
connection
8i,
then
they
would
change
the
direction
by
calling
it
set
direction,
which
you
know
because
send
receive,
doesn't
match
what
they
offer.
That
would
trigger
negotiation
is
so
the
remote
offer
with
new
green
ago
shape
and
three
go
son
receive.
So
basically,
if
both
sides
or
clear
connections
that
you
would
always
end
up
with,
you
know
the
intersection
of
the
directions
in
a
stable
state.
A
D
D
D
D
Instead
of
trying
to
document
every
possible
state
transition,
so
if
you
look
on
the
next
slide
provides
an
example
of
how
this
looks
so
when
ice
agent
indicates
that
the
transport
state
has
changed,
which
means
it's
where
the
burden
is
on.
I
saving
you
know
a
conceptual
entity
to
you
know,
implement
a
state
change
indication
based
on
a
definition
of
the
email
and
when
this
occurs,
a
task
is
glued,
which
updates
the
transport
state
fires.
D
Yes,
so
the
previous
discussion
about
direction
and
negotiation,
you
assumed
that
when
you
false
that
direction,
you're
always
going
to
be.
You
know
when
you
read
that
review
you're
just
going
to
get
what
you
last
set
but
I
wasn't
sure
if
there's
something
we
ever
really
decided
on
and
al
argument
might
be
made
that
it
would
make
sense
that
the
direction
is
actually
was
last
ago
she
ate
or
what
was
in
the
last
current
description.
So
these
examples
sort
of
illustrate
that
point.
D
Good
idea
does
the
the
direction
of
reflect
what
was
set?
What
was
offered
or
sorry
was
answered
so
that
that's
the
question
and
we
can
move
on
to
the
next
slide,
so
I
think
the
the
tangible
benefits
of
the
direction
changing
when
you
apply
description
are
really
just
that
now
you
can
see
the
direction
those
indo
she
SD
p
and
I
guess
if
it
provides
a
nice
help
point
in
the
Navy
I
to
reference
further
up.
D
You
know
sequences
of
steps,
but
I
think
if
that's
the
real
issue,
we
could
also
just
add
a
negotiation
direction
attribute
to
fill
that
goal
and
I
actually
be
in
favor
of
that,
because
I
think
and
we
do
have
two
notions
of
direction.
It
would
be
nice
if
they
were
cleanly
separated
and
another
benefit
would
be
that
kind
of
makes
the
handling
of
negotiation
needed
for
direction.
Pretty
simple,
as
opposed
to
the
complicated
rules
on
the
previous
slide,
we
can
just
say:
negotiation
is
needed.
D
So
now
you
have
to
use
a
transceiver
a
guy,
even
if
you're
using
that
track-
or
we
add
even
more
special
rules
where
maybe
you
have
your
combination
send
receive
or
send
only
direction,
and
not
really
want
are
thinking
about
that
and
it
also
sort
confuses
what
the
directional
attribute
is,
because
it's
full
direction
that
we
want
to
negotiate
and
direction.
That
was
last
time
ago.
She
is
it's
a
sort
part
of
distinguish
between
the
two,
so
we
want
the
next
slide.
My
recommendation
is
that
we
decide
that
thing.
A
description
does
not
change.
D
The
direction
attribute
is
always
what
you
last
set
and
I
think.
If
we
do
want
to
expose
indigo
shaded
direction,
it
would
be
a
good
idea
to
have
as
a
separate
attribute,
but
it
make
clear
that
one
direction
is
the
direction
preference
and
it's
always
maybe
what
you
last
set,
and
he
has
other,
is
what
was
last
night.
I
shaded
and
what
you
know
actually
applies
when
it
comes
to
whether
I'm
sending
a
receive
media.
D
A
D
Rtp
encoding
parameters,
I
win,
think
so.
No
because
so
that's
I
would
expect
point
of
be
what
you
last
set
in
set
parameters
and
sort
orthogonal
and
not
related
to
negotiate
general.
D
B
D
D
And
if
okay,
okay,
all
right
so
I,
will
take,
the
action
does
to
make
a
PR
and
your
attribute
them.
A
All
right
so
927
is
about
what
happens
when
you
call
set
direction.
Some
of
the
confusion,
Taylor
just
talked
about,
might
be
cleared
up
by
actually
describing
the
steps.
So
this
is
the
existing
text.
We've
kind
of
gone
over,
it
doesn't
take
effect
immediately,
blah
blah
blah,
and
so
the
strawman
proposal
is
the
following
steps.
A
A
C
A
D
Is
worth
hang
yeah?
This
is
what
we
left
off.
The
feedback
mean
discussing,
so
the
catching
went
up.
What
we
want
is
to
have
the
you
frag
trickled
with
ice
candidates
and
also
with
end
of
candidates
along
with
preferably
a
bit
as
well,
because
we
need
the
youth
drag
in
order
to
differentiate
candidates
for
us
from
before
nice
restart
and
after
some
circumstances
and
the
they
should
have
sort
of
been
pulling.
D
This
up
is
that
independence
works
by
saying
the
entire
party
see
eyes,
canada,
the
null,
and
that
would
be
a
logical
place
to
put
you
fraga
now.
The
question
is:
where
does
you
for
a
go
for
end
of
candidates?
So
next
slide
a
couple
of
approaches,
so
I
think
we've
discussed
previously
our
MGR
sm57,
the
frags
added
to
our
TV,
I
sanded
and
then
in
the
candidates.
D
Edit
event,
and
did
this
would
mean
that
it
doesn't
get
a
mid,
because
you
know
the
you
frag
is
the
only
thing
in
the
event,
then,
which
is
probably
not
a
big
deal,
but
you
know
turbo
eyes
does
say
that
the
identifier
for
the
media
stream
should
be
tripled
so
next
slide.
D
This
is
what
I
think
we
sort
of
were
informally
discussing
at
the
end
of
keefak
or
I.
Think
issues
with
the
previous
proposals
would
be
that
they
they
weren't.
You
know
they
tried
to
reuse
the
ice
Canada
event,
but
they
weren't
completely
backwards
compatible.
So
I
guess
the
solution
here
is
to
just
define
a
new
event
and
new
method
to
handle
something:
that's
either
a
candidate
or
that
of
candidates,
and
it's
pretty
similar
to
TR
757,
except
for
beer.
Now
calling
it
ice
action
or
ice.
D
D
This
is
the
the
rest
of
the
proposal
we
have
with
the
union
with
two
values.
I
think
one
advantage
of
this
is
that
would
be
easy
to
extend
in
future
versions
of
the
spec,
although
maybe
put
not
a
huge
concern
at
this
point
and
then
the
the
offense
would
expose
this
dictionary-
and
I
put
here
also
expose
the
I
scanned
interface,
where
you
can
query
attributes
if
you
want
to.
D
Next
slide
August
at
this
point,
from
whatever
people's
thoughts
on
this
proposal.
Do
we
do
we
want
to
go
ahead
with
this
or
go
back
to
drawing
board
and.
D
Let's
see
I,
don't
remember,
there's
a
specific
reason.
I
think
this
that's
minor
detail
we
could
work
out
later.
I
mean
whether,
like
something
is
in
our
face
or
a
dictionary
whether
we
have
like
an.
D
C
D
Next
slide
then
well
I
just
wanted
to
send
someone
would
also
say
anything:
I
I,
like
the
general
solution
of
adding
a
different
method,
to
make
it
clear
that
this
is
different
than
what
it
was
before,
and
you
know
w
some
transition
is
apps,
move
over
the
new
think
the
old
thing
but
long-term.
This
is
much
more
robust
than
before.
F
A
new
method:
what
kind
of
web
content
is
just
protecting?
I
guess
so
I
have
quite
shallow
uses
of
these
AP.
I
swear.
I
treat
candidates
mostly
in
some
black
boxes
so
and
this
this
is
not
necessary
at
all
to
me.
So
if
someone
could
speak
up
and
defend
why
we
need
to
make
new
methods
that'll
be
great,
meaning.
D
H
Remember
the
last
time
we
talked
about
this
eka
jumped
in
with
the
problem
that
you
might
not
be
bundling
and
then
you
it
would
be
ambiguous
which
of
the
two
em
lines
you
were
talking
about.
Have
we
address
that
in
this
I'm
still
trying
to
catch
up
because
I
had
to
step
away
for
a
minute?
But
do
we
address
that
issue
or
is
it
gone
away?
Go.
A
D
All
right
so
I
mean
we
have
a
couple
of
people
in
control.
That's
and
there's
also
the
question
of
whether
PR
sm57
is
sufficient
just
using
the
existing
their
faces,
and
you
know
assuming
applications
treat
the
candidate
as
a
black
box.
I
think
the
only
issue
but
PR
sm57
would
be
if
they're,
not
community,
as
a
black
box.
Actually
trying
to
you
know,
get
the
candidates
they're,
not
gonna,
expect
it
to
be
null
so
they'll.
D
The
end
of
candidates
actually
is
an
RTC
ice
candidate
that
has
mid
and
I
you
frag.
It
just
has
no
stp
string
right,
so
I
think
we
do
have
everything
we
need.
Then
again.
The
only
issue
I
think
with
dr
77
is
that
applications
might
not
expect
an
OL
canada
string
that
can
never
be
possible
before
so.
On
that
the.