►
From YouTube: IETF102-MMUSIC-20180719-1330
Description
MMUSIC meeting session at IETF102
2018/07/19 1330
https://datatracker.ietf.org/meeting/102/proceedings/
A
A
A
So
thank
you
for
that
and
the
usual
stuff,
when
you
speak,
use
the
microphone
and
state
your
name,
and
we
have
the
beautiful
pink
square
beside
the
chairs.
Please
keep
within
that,
and
it
will
help
the
meet
echo
as
a
general
requests.
We
ask
that
you
help
out
with
reviewing
documents.
If
you
have
your
own
documents
that
you
wish
see
taken
forward
and
progress
in
the
working
group,
please
help
also
others
with
you.
Previews.
A
A
A
B
A
So,
apart
from
those
laying
in
there
are
see
attitudes
cue,
we
have
a
small
set
of
drafts
in
working
group
consensus.
We
have
the
SDP
neg,
which
saw
quite
a
lot
of
changes,
a
couple
of
recent
versions.
So
the
question,
if
is
if
people
would
like
to
see
another
working
group,
last
call
for
that
or
if
they
are
happy
with
the
changes
that
were
made
based
on
comments
as
part
of
of
the
previous
last
call,
even
though
it
has
been
quite
some
time.
F
Or
any
other
I
think
most
of
the
changes
are
editorial
nature.
It
was
like
more
like
that
eat
up
to
date,
according
to
current
procedure
about
registration,
SDPO
France
does
not
do
more
technical
changes
in
the
in
the
document,
but
and
I
think
it's
almost
ready
just
need
to
verify
that
the
latest
changes
are
address
all
the
comments,
but
if
you
want
to
do
a
hill
class
call
to
you,
so
your
decision
I
can.
A
Leave
it
up
to
the
working
group
in
present
in
the
room,
if
there
are
no
objections,
I
suggest
that
we
continue
with
regular
procedures
as
part
of
the
previous
working
group.
Last
call
addressing
comments,
as
Assad
said,
Ronnie
said:
I
believe
it
is
mainly
editorial
and
clarifications.
So
no
real
technical
changes.
B
G
First
thing:
the
update
needed
on
46
bits,
I'm
I'm
done
waiting
for
comp
and
waiting
for
countenance.
Basically,
at
this
point,
I
submitted
an
update
a
while
ago,
and
the
only
thing
I
know
there's
one
thing
wrong
with
it
now
could
lose
it
there's
a
a
normative
reference,
though
I
be
informative.
That's
the
only
thing,
I
know
about
now
I'm
holding
sending
another
update
for
that.
E
I'll
just
caution
for
clarification
times.
Flemming.
Are
you
looking
for
comments
just
to
review
what
is
done
or
were
there
any
issues
that
you
need
feedback
on
or
solution?
Okay?
So
for
as
far
as
you're
concerned,
everything
is
done
and
it's
really
up
to
people
to
comment
if
they
have
an
issue.
Okay,
thank
you.
G
Yeah
I've
asked
several
times
when
I
made
changes,
I
tweaked
the
a
B
and
F
to
clean
it
up
in
some
ways
and
in
the
process
of
doing
that,
the
the
TR
z
or
C,
whatever
it
is,
is
the
stuff
with
a
little
black
and
so
I
made
a
change
in
the
syntax
to
make
it
more
aligned.
With
this
that
has
been
documented
and
I
just
like
to
hear
anybody.
If
anybody
has
any
reason
to
believe
that.
G
H
Anyone
when
we
did
four
five,
six
six,
so
twenty
three
twenty
six,
seven
whatever
it
was
this-
will
you
renamed
some
of
the
the
a
PMF
that
the
variables
in
the
ABN
F
and
people
did
complain
about
that
in
this
case,
I
think
it's
a
sufficiently
obscure
thing
and
I
think
things
have
been
renamed
they've
just
been
something
that
was
optional.
It's
been
made
mandatory
or
whatever
it
is
so
I,
don't
think
it's
gonna
be
a
problem.
G
H
So
so
so
so
way
back
can
mean,
if
necessary,
in
about
2002
or
something
when
we
were
doing
four
five
six
six,
the
were
complaints
were
when
we
renamed
bits
of
the
ABN
F
I
think
they
came
from
Sri
GPP,
but
that
may
be
wrong
on
that.
It's
been
a
long
time
away.
I'd
have
to
search
search
battery
the
email
to
see
who
complained,
I,
suspect,
I,
suspect
it's
okay
and
you
know
possibly
we
just
want
to
last
call
it
and
send
seven.
H
E
E
H
G
G
E
A
A
Today-
and
there
are
a
few
other
documents
in
the
working
group-
we
have
the
M
music
MSRP
usage
data
channel
where
it
was
updated
recently,
but
there
were
new
reviews
or
very
little
interest
in
the
group.
So
at
this
point
in
time,
chess
proposed
to
actually
drop
the
milestone
for
this
draft,
since
there
seems
to
be
no
interest,
and
that
also
seems
to
be
okay.
With
a
new
author
of
the
documents,
we
can
revive
it.
A
A
E
He's
going
back
to
the
unknown
cashier
draft.
From
the
author's
point
of
view,
this
draft
is
done.
It
has
not
gotten
a
whole
lot
of
review.
It's
been
announced
at
TLS
and
what
you
see
WEP
as
well
I
believe
it
is
I,
have
not
followed
what
the
status
is
there,
but
could
we
get?
Maybe
you
wanted
two
people
to
volunteer
to
review
it,
preferably
people
that
have
some
security
knowledge,
no
need
to
look
down.
Jonathan
I
need
to
look
down
any
volunteers.
B
D
D
A
And
for
this
slide
we
have
used
it
in
this
working
group
for
a
while
clearly
outlining
what
dependencies
that
we
have
to
the
RTC
web,
and
this
is
starting
to
look
very
good
I
believe
most
of
them
are
in
the
RFC
editors
queue
and
the
only
document
here
that
is
not
already
there.
We
have
on
the
agenda
today
and
we
have
working
group
consensus
for
that
one.
But
let's
see
what
happens
during
that
gender
point
today,
apart
from
that,
we
are
done
with
those.
I
Okay,
this
is
a
presentation,
a
discussion
that
I
really
hope
that
we
wouldn't
have
to
do.
But
here
we
are,
it's
related
to
the
describing
offer
answer
procedures
when
we
define
a
new
attributes.
There's
no
draft
for
this
I
guess
we'll
see
what
the
outcome
of
the
discussion
today
is,
whether
it's
going
to
be
a
new
draft
incorporated
in
some
other
draft
or
we're
not
going
to
do
anything.
I
In
some
cases
you
may
even
have
then
sections
below
this,
for
example
in
bundle.
We
have
a
lot
of
sub
sections,
but
basically
this
is
the
main
sections
that
that
that
you're
you're
supposed
to
have
and
which
we
in
most
cases
have
now.
The
problem
is
that,
as
you
saw
there
on
a
previous
slide,
there
is
the
one
sub
section
which
says
generating
the
initial
STP
offer.
Now.
I
This
word
initial
has
caused
some
problems,
what
it
means,
for
example,
if
you
look
they're
basically
two
two
two
ways
that
people
interpret
it
I,
don't
think
we
need
to
go
up
here
and
say
what
people's
understanding
is.
But
the
problem
is
that
first,
some
people
say
that
this
is
the
first
offer
within
a
session,
and
this
is
basically
aligned
with,
for
example,
RFC
3261.
I
The
CSS
initial
offer
is
the
first
offer
that
you
send
within
a
session,
no
other
alternative,
which,
at
least
in
my
opinion,
is
how
how
have
drafts
SED
draft
attribute
draft
describe.
It
is
the
first
offer
within
a
session
when
this
attribute
it
actually
introduced.
I
mean
it
could
be.
The
first
session
offer
our
session,
but
it
could
also
be
later
in
the
session
unless,
of
course,
the
attribute
has
a
default
value,
in
which
case
it
will
be
in
the
first
offer
of
the
session.
I
Even
even,
if
you
don't
put
it
there,
it
will
be
the
implicit
default
value,
but
but
that's
a
special
case
and
of
course,
if
you
have
multiple
M
lines,
you
could
add
as
attribute
to
different
M
lines
in
different
offers.
You
don't
you
you
may
have
an
audio
line
at
some
point,
then
you
add
a
attribute
there
and
then
later
in
the
session,
you
may
add
a
new
M
line
for
video
and
then
you're
gonna
have
we,
you
will
add
an
attribute
to
that
that
that
meat
is
or
M
line
to.
I
Removing
it
should
be
pretty
simple,
you
just
remove
it,
but
it
may
be
some
implication
or
some
procedures
that
that
you
need
need
to
describe.
They
could,
for
example,
be
just
take
an
example.
This
STP
identity,
which
we
have
in
our
TC
web,
which
which
has
a
very
it's
related
to
the
fingerprint
attribute.
I
So,
for
example,
if
you're
removing
that
attribute
you,
you
would
also
have
to
remove
that
entity,
attributes
and
so
on
so
on,
but
but
basically
the
bottom
line
is
that
we
need
to
separate
between
this
case,
when
you
add
it
to
a
name
section
and
and
when
it's
there,
when
you
modify
it
now,
there
are
two
ways
here:
maybe
they
could
be
sick
trip
or
more,
but
this
is
the
way
first.
This
first
solution
is
to
just
clarify
what
the
initial
offer
means.
I
Whenever
you
define
a
new
estimate,
this
is
actually
what
we
did
in
bundle.
So
this
text
here
is
taken
from
from
bundle
from
the
initial
offer
section,
which
is
that's
a
procedures
in
the
in
section,
apply
to
the
first
topper
within
an
STP
session
egss
it
dialogue,
bla,
bla
bla,
in
which
the
offer
indicates
that
it
wants
to
negotiate
a
given
bundle
group.
This
could
occur
in
the
initial
offer
or
in
a
subsequent
of
rada
STP
sessions.
I
So
we
really
clarified
it
so
basically
I
mean
maybe
it
could
be
worded
better
and
so
on,
but
basically
this
would.
We
would
need
to
have
these
texts
in
whenever
we
define
a
new
STP
attribute
I
mean
we
could
mean.
It
would
be
very
I
mean
very
simple
to
modify
these
texts
and
to
to
your
favorite
attribute,
but
this
is
how
that
would
like
can
I
give
a
second
present
a
second
okay.
I
The
second
solution
is
really
to
change
the
structure,
not
talk
about
not
focus
really
on
on
any
initial
offer,
a
sub-q
cincuenta
offer,
but
look
it
from
the
feature
from
the
attribute
perspective.
So
basically
would
have
this
structure,
which
you
see
here
in
violet,
the
first,
this
section
would
be
add,
attribute
to
a
media
description
or
if
it's
a
session
level
attribute
to
read
through
STP
session,
and
then
you
have
two
subsections
send
offer.
When
you
do
it
and
send
the
answer,
we
wouldn't
really
care
whether
this
is
the
initial
offer
or
whatever.
I
This
is
the
offer
where
you
actually
add
this
attribute
to
recession,
and
then
you
would
have
it's
our
second
one
modified
attribute
or
a
modified
session.
Whatever
then
sent
how
you
send
the
offer
and
then
you
could
have
it
unchanged,
updated
removed
and
so
on
and
answer
my
personal
I
think
my
suggestion
is
alternative
to
you,
but
I
don't
really
have
any
strong
opinion,
so
I'm
not
gonna
argue
if
someone
wants.
I
If
the
community
wants
well
I'm
fine
with
that
or
if
someone
has
something
else
or
someone
thinks
we
shouldn't
do
anything
because
it
works,
what
we
have.
No,
it
works
fine,
but
but
so
I
think
maybe
a
comment
at
this
point,
because
I
think
this
is
all
all
I
had
to
say.
The
rest
is
just
the
way
forward,
but
but
in
order
to
go
there,
we
should
maybe
first
try
to
agree
on
something.
Thank.
I
H
I
I
C
C
So
the
only
question
is
whether
you
need
to
adjust
the
timing
in
some
way,
such
that
you
know
I,
guess
in
you
know
if
in
general,
if
you
couldn't
resolve
the
name
for
whatever
reason
I
feel
like
this
is
a
maybe
a
more
general
ice
issue
or
ice
or
fqdn
and
ice.
If
you,
if
you
couldn't
resolve
the
name
in
general,
you
know.
M
I
I
L
Justin
Uberti
Google
the
thing
that
distinction
is
that
the
jas
will
never
learn
the
private
IP
address.
You
know
the
the
browser
who
have
to
learn.
It
is
because
it'll
see
like
where
this
checks
are
coming
from,
but
that's
okay,
because
it's
gonna
do
the
emptiness
and
lookup
anyway.
You
don't
learn
it,
but
the
jas
won't
have
access
to
it.
N
L
C
That's
another
point:
I
was
gonna.
Make
is
I
feel
like
if
we
were
doing
this
in
the
context
of
trickle
the
the
timing
rules
get
a
lot
clearer,
because,
basically,
you
can
treat
when
the
DNS
resolution
succeeding
as
though
that
candidate
got
trickled
I'm
wondering
if
we
can
do
some
sort
of,
like
mini
mini
trickle
logic
here,
to
describe
what
you
do
when
the
DNS
succeeds,
or
until
it
succeeds
because
I've
been
clearly,
you
know
if
you
had
trickle,
that's
the
logic
you
want.
L
If
you
don't
know
how
to
handle
these
candidates,
then
you
just
have
to
skip
over
them
and,
like
I,
also
just
think
that
the
timing
like
there's
the
signaling
channel,
has
variable
timing
anyway
and
so
like
you
might
have
like
a
second
latency
through
the
signaling
channel,
like
that
seems
to
not
be
a
overall
problem.
So
if
it
takes
a
little
bit
time
to
resolve
this
name,
I,
don't
think
that's
the
end
of
the
world
either.
I
And
courser
related
to
trickle,
maybe
something
we
need
to
document
clarify
there
is
that
when
is
we
used
to
see,
pin
four
four
four
four
four
trickle
and
you
always
have
to?
If
you
understand,
if
you
understand
the
contest,
you
always
have
to
reply
with
a
200,
ok,
which
means
that
even
if
you're
not
able
to
resolve
this-
or
even
it's
an
IP
version
that,
but
you
don't
support,
you
still
have
to
send
200.
Ok,
for
maybe
that's
already
in
in
in
the
trickle
draft.
I
don't
know!
But
but
again
you
always
accepted
the
info.
J
Another
problem
Jara
lava
Strom
another
problem
I
was
curious
about
yes,
when
you
get
that
probe
from
the
other
side,
and
this
comes
in
from
yeah.
You
know
their
IP
at
the
browser
knows
the
IP
address
of
the
other
side
because
that's
in
on
the
packet,
but
he
then
doesn't
know
whether
that
matches
a
candidate
in
his
candidate
list
or
not.
J
J
I
This
is
a
hash
table.
I
said
earlier
in
in
your
internal
tables,
in
your
internal
tables,
you
need
to
use
the
IP
addresses
because
you
use
it.
For
example,
when
you
make
the
table
to
move
to
to
to
to
make
the
Foundation's
and
when
you
receive
connectivity
checks,
you
need
the
IP
address
to
see
whether
this
matches
and
something
that
you
have
and
everything
so
so
so
so
so
internally,
you
must
use
IP
addresses.
So
we
need
to
you
know,
redesign
eyes
again
so.
M
N
I
agree
with
Chris
Irsay
III.
Think
that
answers
my
question
from
earlier
to,
which
is
what
you
need
to
do,
is
think
about
this
as
an
encrypted
form
of
the
IP
address,
and
you
do
all
the
same
things
as
if
you
have
the
IP
address,
like
from
timing
point
of
view.
It's
just
like
you
had
an
IP
address.
You
can't
just
throw
these
away
and
skip
past
them.
If
you
can't
resolve
them
or
something
you
need
to
treat
them
like.
N
M
C
Only
what
is
yeah
Jonathan,
except
I
think
the
interesting
questions.
What
I
think
it's
sort
of
what
Haribo
is
getting
at
is
what
do
you
put
in
the
if
you
end
up
using
this
candidate
for
your
connectivity?
What
you
put
in
the
next
offer
answer
from
both
sides.
Do
you
still
that
does
is
the
does
the
person
who
offered
this
label
still
putting
that
in
its
candidates
in
its
MC
law
and
C
Line,
and
is
the
Peter
putting
this
in
its
C
line?
C
C
Mean
I
feel
like
the
threat
model
were
worried
about
here.
Is
people
who
are
not
on
the
local
network
find
out
about
your
hero,
local
network?
It's
you
know,
you
know,
then
you
know
this
person
you're
trying
to
work
correctly
if
you
are
on
the
same
local
network,
but
you
don't
want
to
advertise
your
internal
structure
to
drive
drive
by
JavaScript
and
so
I
feel
like
if
it
actually
did
end
up
using
this
candidate
revealing
the
IP
addresses
in
that
case
is
not
a
problem.
In
my
opinion,.
C
M
L
I
Years
old
man,
I
think
so
I
always
thought
when
I
read
this
proposal,
that
it's
only
for
the
scenario
that
the
signaling
should
not
learn
about
my
host
candidates
in
case
I,
never
get
connected.
If
you
take
that
to
the
level
of
like
there
is
a
connection
and
we
want
to
reuse
these
candidates,
I
think
then,
probably
not
for
for
a
music
but
in
general
over
create
a
lot
more
problems.
Because
then,
what
do
you
do?
Vote
get
stats?
We
have
like
ice
candidates
in
there.
L
Pure
reflexive
candidate
from
that
local
IP
and
you
can
have
like
this
timing
condition
they're
only
for
results
that
leaking
and
and
that's
something
that
has
to
be
dealt
with
and
like
pryzen
in
this
document.
But
this
probably
back
in
that
be
handling
document,
but
that's
the
only
thing
I
can
think
of
that
isn't
just
work
that
actually
requires
some
something
to
be
solved.
I.
L
I
Nobody
uses
like
fqdn
ends
today,
so,
like
I,
can
tell
you
that
in
Firefox
I'm,
pretty
sure
that
the
steps
get
generated
out
of,
like
whatever
the
ice
tech
says,
is
the
IP
address,
because
there's
simply
no
support
for
FQ,
nians,
yeah
I.
Think
what
you
have
to
put
in
the
C
line
is,
as
the
current
spec
says,
is
that
this
is
the
address
that
should
be
used.
I
I
Course
you
use
trickle
where
you
could
put
this
zero:
zero,
zero,
zero
or
whatever.
But
but
if
you
don't
in
case,
you
don't
use
trickle
I,
don't
know
if
that
needs
a
note.
They
are
saying
that
the
ADC
should
also
be
put
in
in
in
the
C
Line
I,
don't
know,
but
maybe
a
good
clarification,
because
I
think
the
case
we
want
to
avoid.
Is
that
a
bad
implementation?
Yes,
they
put
this
in
the
candidate,
but
then
they
still
put
IP
address
in
the
C
Line.
C
M
M
C
C
Really
makes
this
much
more
confusing
to
me,
because
that
basically
says
what
you,
what
you
do,
if
your
DNS
query
returned
multiple
IP
addresses
and
which
I
which
remote
IP
address
you
use
depends
on
which
candidate
you're
sending
from
and
that
kind
of
makes
my
head
explode,
especially
in
terms
of
matching
up
here.
Reflexive
addresses
so
I.
C
Yeah
I
mean
they're
complicated
rules,
but
basically
what
it's
it's.
Basically,
given
that
a
DNS
query
returned
a
set
of
IP
addresses,
which
one
you
send
to
depends
on
what
you're
sending
from.
So.
How
that
how
you
can
have
a
that
kind
of
means
that
you
have
a
candidate
whose
IP
address
is
different,
depending
on
what
it's
paired
with,
and
that
makes
my
head
completely
explode.
The.
C
This
is
not
new
for
Debian
s,
but
for
the
the
rule,
I
mean
it's
all.
It's
already
bad
room
for
this,
but
it's
up,
but
if
we're
going
to
impose
this
rule
that
aid,
the
this
address
must
always
be
that
that
candidate
must
always
represent
by
its
name.
How
I
determine
what's
that
candidate
when
there's
a
set
of
IP
addresses
it
could
be
depending
on
what
candidate
you're
talking
to
yeah.
M
It's
a
bigger
issue
with
DNS,
especially
for
endianness.
Hopefully
on
the
generous
generation
side
you
will
take
one
IP
address
and
and
one
local
name,
and
so
that
on
the
resolution,
side
of
we'll
be
simpler,
but
anyway
an
implementation
will
need
to
handle
the
case.
We
have.
There
will
be
an
fqdn
name,
no
matter
DNS
or
DNS
and
then
handle
the
case.
Wherever
will
be
civil,
most
potential
IP
others
anyway,.
L
They're,
like
I,
agree
that
this
text
about
the
rule
of
67
24
and
like
I,
tried
reading
it
the
other
day.
It's
actually
quite
confusing,
like
just
in
the
general
case,
I
sent
a
mail
to
the
ice
working
group
notice.
I
haven't
really
received
response
back.
I
think
that
just
needs
to
be
clarified
in
general,
but
I
would
like
to
try
that
sort
of
hone
in
here
is
what
do
we
want
you
in
to
do
like
what?
What
of
the
green
text
you
know?
D
Data
Roche
is
an
individual
I,
largely
agree
with
that.
There
is
one
caveat
that
ties
back
to
something
that
was
just
said,
which
is
we
if
we,
if
we
want
to
make
the
assertion
that
an
MDS
Jemma
Nina's
name
corresponds
1:1
with
an
IP
address,
we
are
going
need
to
say
that
in
here,
because,
like
normal
behavior
right
now,
when
I'm
at
home,
I
have
machines
there,
like
food,
local,
I'm,
gonna,
be
for
only
machine
and
go
to
food
out
local,
like
do
it
over
before.
D
If
I'm
gonna
be
six
cable
machine
go
to
food
out
local,
it's
gonna
hit
a
v6
address,
so
by
default.
Mdns
typically
works
on
on
both
of
those,
and
you
know
you're
matching
that
up.
If
we
don't
otherwise
say
with
you
know,
v4
candidates
and
be
six
candidates,
it'll
match
those
unless
we
constrain
it
sure.
M
On
generous
generous
inside,
we
can
say
in
the
world
if
we
were
to
see
that
you
should,
you
should
do
that
better.
The
JavaScript
may
do
whatever
we
want,
we've
generating
whatever
candidates.
They
won't
be
a
consent
that
send
that
to
the
other
side,
and
then
the
other
side
will
need
to
handle
this
case
of
multiple
potential
IP
addresses
for
one
name.
M
D
M
M
N
So
this
is
the
ICSD
PTR
when
we're
talking
about
changing
right,
so
I
think
we
do
need
to
discuss
in
favor
of
this
or
change.
This
looks
fine,
but
I
think
there's
also
needs
to
be
text
that
discusses
backwards.
Compatibility
with
this
so
outside
of
the
WebRTC
context.
No
existing
offer
answer
clients
that
do
ice
support,
MDAs
that
I'm
aware
of
outside.
M
N
N
I
think
it
should
be
in
the
you
know,
this
draft,
the
STP
okay.
This
is
this
is
okay.
This
is
generic
advice
to
people
who
are
writing
ice
SDP
about
when
they
use
this
now,
and
it
would
point
out
you
know
if
you're
a
web
browser,
you
need
to
go,
read
Justin's
documents
and
things
that
come
out
of
it
right,
but
in
the
non
WebRTC
use
cases
we
probably
need
a
little
bit
of
advice
to
say
if
you
just
start
putting
this
stuff
in,
it's
not
gonna
work.
Basically,.
C
C
N
C
Well,
I
mean
I
use.
The
IOC
stuff
is
probably
using
a
stripped
down
verse
a
little
bit
about
why
currencies
of
are
pretty
seeable.
That's
this
there
and
they
actually
update
IOT
stuff
that
we're
good
so
but
I
mean
obviously
that's
you
know,
hey
these
corner
creatures
happen
to
line
up
nicely
with
each
other.
It
does
not.
In
the
general
case,
yeah
I
mean
it's
always
a
instead
of
this
weird
cases.
This
is
definitely
a.
What
is
the
must?
What
we
know
you
won't
support
problem,
or
at
least
bust,
but
we
know
you
didn't.
B
I
You
told
me
just
to
be
clear:
I
mean
I'm
I'm
in
favor
of
like
changing
it,
to
allow
mdns
for
the
initial
advertising,
basically
as
a
as
a
protection
against
the
the
data
channel
abuse
thing
we
all
have
in
our
mind,
but
in
terms
of
like
reusing
it.
What
do
we
do
if,
if
I
receive
a
binding
requests
and
I
learned
that
as
a
peer,
reflexive
and
God
forbid,
there's,
like
the
other
end,
basically
learns
their
way
as
well,
that
his
host
candidate
is
being
netted
locally
and
then
like?
How
do
you
match
up
then?
I
Like
an
EM,
DNS
name,
I'm
I'm,
pretty
sure
we
can
come
up
with,
like
some
bizarre
scenarios
were
like
where
you
receive
a
stun
binding
request?
First,
and
then
you
get
your
trigger
candidates
later
and
then
I
have
no
idea
how
the
ice
tech
should
figure
out
how
to
match
up
these.
Like
then,
the
good
scenario,
your
mdns
will
luckily
resolve
exactly
to
the
IP
address
important
number.
You
receive
the
Sun
binding
request
form,
but
if
that
doesn't
align,
then
you're,
basically
screwed
I
mean
then
you
can
only
put
in
the
public.
L
Justin
you
ready.
This
is
I,
think
one
of
the
reasons
why
Ted
suggest
we
move
this
to
a
different
spec
rather
than
John
dumping.
This
all
into
IP
handling,
yeah
I
think
there
are
some
things
that
has
to
be
figured
out
and
like
that
P
reflexil
case
was
definitely
one
of
them,
but
to
Colin's
point
they're
like
would
you
be
okay
with
just
having
some
Texas
says?
You
know
note
that
you
know
some
clients
will
not
support
fqdn
and
you
know
they're
free
to
ignore
they
will
be
free
to
ignore
these
candidates
for
sure.
N
D
N
C
H
C
Okay,
who's
not
enough,
even
in
the
chopper
but
yeah
I,
think
this
change
I
mean
13:45
said.
Basically,
if
you
do
an
fqdn
look
up,
you
pick
one
IP
address.
You
know
v6.
If
you
have
it,
you
know
before
otherwise,
and
that's
change
to
use
six
seven.
Twenty
four
rules
and
I
am
I.
Think
I.
Remember
that
going
by
and
sort
of
nodding,
zinc,
yeah,
I
guess
that
kind
of
makes
sense
the
past,
but
now
I'm
looking
at
that
and
somewhat
alarmed
so
so
we
might
want
to
revisit
that.
I
C
I
P
I
I
send
my
candidates
to
you:
I
only
send
my
server
flexi
and
whatever
that
don't
have
any
privacy
issues
and
you're
gonna
send
your
your
candidates
back
to
me.
Then
we
create
the
pairs
and-
and
you
know
we
use
those
pairs
to
to
send
some
can
send
candidates.
But
then,
when
I
do
send,
if
there
is
known
that
when
I
start
sending
checks
to
you,
you
will
create
a
peer
reflects
a
candidate
bigger.
Q
C
Know
no,
no,
no,
that
that
will
work
if
hair
printing
is
working
because
no
bounce
off
the
your
started,
some
server
and
go
back
the
other
guy
and
that's
great.
But
the
reason
why
we're
doing
all
this
is
because
they
determined
that
they're
pinning
you
know
there.
There
are
many
there
y-you
know
many
NAT
boxes
sold
by
the
same
companies.
That's
right!
The
Browse
cell,
the
browsers
which
do
not
do
hair
fitting
properly.
L
Justin
Uberti
bag.
This
is
not
irrelevant
this
thing,
but
visual
men.
They
were
a
problem
of
okay.
Imagine
your
web
patient
wants
to
gather
addresses.
You
can
then
omit
the
emptiness
candidates
and
then,
if
hair
printing
succeeds,
you
might
actually
then
be
able
to
then
see
the
source
address
of
like
these
local
candidates,
that
we
creatives
peer,
reflexive
and
maybe
get
access
to
the
private
addresses
that
way.
So
like
this
yeah,
this
is
a
complicated
problem.
L
Well,
yeah,
a
little
soon
that
news
page
will
open
did
channels
to
themselves.
Once
we
take
away
the
front
door,
local,
try,
the
back
door
and
then
like,
if
it's
possible
to
you,
know,
remove
the
MD
nesting
and
then
like
they
that
will
cost.
Perhaps
these
IP
address
of
the
coop
you
know.
Naturally,
we
have
to
sort
of
figure
out
how
to
you
know,
map
that
to
anyway.
This
is
really
not
germane
to
this
thing,
which
is
like
a
very
focused
piece
of
text
which
I'd
still
I
custom.
Before,
with
this
column,
suggestion.
G
E
I'm
I'm
not
sure
we
have
it
quite
right
up
here.
So
let's
try
to
go
through
it.
You
know
one
by
one
in
terms
of
actually
supporting
mtns,
which
I
guess
is
the
you
know
request
that
you
guys
came
with
if
I
understand
the
sentiment
of
the
room
correctly
I
mean
people
are
in
favor
of
doing
that.
However,
there's
some
details
to
be
worked
out
as
to
exactly
what
does
that
mean?
Is
it
just
for
the
initial
you
know
offer,
or
is
it
beyond
that?
Am
I
understanding
that
correctly?
C
Jonathan
I
think
for
this
document
is
supposed
to
IP
handling.
You
know
whether
it's
fourth
I
mean
this.
This
all
I
says
is
you
may
put
it,
you
might
put
a
name
DNS
name
in
if
you
feel
like
it
and
I.
Think
it's
no.
Is
that
you
just
let
me
them
for
the
initial
offer
to
undo
it
for
subsequent
that's
much
more
in
the
domain
of
IP
handling
and
artsy
web
cuz.
That's
justified
the
policy
of
what
you
do
or
don't
do
that
so.
C
D
Adam
Roach,
individual,
yes,
I-
think
there
needs
to
be
clarity
about
whether
you're
required
to
process
these
and
I
suspect.
The
sentiment
would
be
no
so
required
to
actually
attempt
to
contact
it
right.
I
mean
certainly
have
to
parse
it,
but
do
you
need
to
act
on
it
or
would
mandating
mdns
support?
I,
don't
think.
That's
realistic!.
N
Look
all
of
this
is,
in
my
mind,
contingent
on
the
fact
that
doing
this
doesn't
can
take.
It
create
a
DDoS
vulnerability
attack
which
I'm
not
arguing
it
does,
but
I
think
that
I'm
arguing
is
one
of
the
things
we
need
to
look
at
quite
seriously
at
the
ice
level
before
we
go
forward.
All
this
not
exactly
an
SDP
issue
is
an
ice
issue,
so.
M
N
R
M
N
Dns
is
not
DNS.
What
you're
trying
to
say
is.
We
should
extend
this
to
use
DNS
and
I'm
great
with
that,
as
long
as
it
doesn't
make
a
security
vulnerability
and
I'm
working
on
the
assumption,
you
will
be
completely
correct
that
it
doesn't,
but
that
assumption
needs
to
be
checked
before
this
is
published.
D
Yeah
I
wrote
individual,
so
I
I
think
the
issue
in
you
know
applies
more
to
the
IP
handling
proposal.
That's
out
there
as
opposed
to
this,
because
this
is
just
saying
you
know,
use
mdns
available
could
be
your
food
out
local
name,
but
this
is
not
anything
to
do
with.
Uid
mechanisms
is
insane
quest
about
proactively.
D
This
is
like
hostname,
that's
fair
use
it
or
you
might
generate
some
other
mechanism
when
we
turn
around
and
start
saying,
we're
gonna
generate
these
things
on
the
fly
and
push
them
out
in
a
way
that
might
destroy
the
network.
That's
when
we
have
to
get
concerned
about
it,
I
don't
think
this
document
applies.
N
It's
not
the
IP
handling
document
is
about
what
you
pass
up
to
the
javascript
in
a
browser
that
has
like
imagine,
we
weren't
doing
anything
with
web
browsers
we're
only
going
to
do
mdns
with
sip
endpoints.
The
DDoS
thing
still
issue
is
this,
so
it's
not
the
IP
handling
document
that
needs
to
discuss
this.
It's
the
ice
usage
of
it,
which
I
agree,
is
not
this
right.
N
N
D
C
Expert
does
the
mdns
query
calls
multicast
traffickers
I
mean
as
advertised
because
well
it
gets
traffic
or
both
okay,
interesting.
The
other
thing
I
think
we
do
need
to
figure
out
for
I
mean
I,
guess
for
this.
This
is
where
you're
saying
you
can
use
F
key
again,
because
the
ice
ice
best
score
does
not
is
exactly
how
DNS
lookups
effect
timing.
Are
you
required
to
do
all
your
DNS
lookups
before
you
start
connectivity
checking
or
are
you
about
I
mean
I,
don't
be
a
bad
answer.
Yeah.
L
L
C
L
C
C
Don't
know
because
it's
the
the
weird
thing
is
it's,
this
doctorate,
that's
mentioning
FTP
ends
at
all
right.
The
I
score
on
does
not
I
would
really
rather
not
have
something
full
and
have
both
those
documents
in
major
surgery.
At
this
point,
you're
that
they're
blocking
for
2:38
yeah,
but
I
kind
of
feel
like
that
it's
fqdn
interact,
interestingly,
with
the
core
yeah.
S
We're
not
above
Microsoft
having
been
intimately
involved
with
mdns
from
the
beginning,
it's
important
for
people
to
understand
that
this
is
an
inherently
link-local
mechanism
that
is
using
an
MD
NS
name
to
represent
an
address
that
is
not
linked.
Local
scope
will
inevitably
cause
introduction
of
failures
that
would
not
otherwise
occur.
So
you
know
talking
about
private
addresses
if,
if
the
peers
are
not
on
the
same
link,
local
scope
right
this,
this
will
actually
cause
changes
to
the
to
what
happens
in
ice,
and
you
will
have
connectivity
failures
that
would
not
have
otherwise
occurred.
N
M
S
As
an
example
like
in
an
enterprise
network,
I've
seen
enterprise
networks
that
use
net10
in
a
routed
way
right.
So
if
you
would
have
to
net
ten
addresses
on
separate
networks
that
ice
connectivity
check,
would
we
pass
you
substitute
mdns
names
now
they're
linked
local
scope?
They
aren't
on
the
same
network
so
now
that
connectivity
check
will
fail.
Now,
it's
true
that
yeah
they
would
end
up
using
a
turn
server
or
you
know,
stun
or
whatever,
but
that's
that's
a
check
that
should
that
should
pass
because
they're
there
they
had
to
have
connectivity.
L
S
N
G
N
I'm
not
saying
there's
a
reason
not
to
do
it,
but
I'm
saying
it's
the
reason
we
need
to
consider
when
and
where
to
do.
It
is
large
enterprises
off
many
of
them
have
flat
IP
addresses
across
over
two
hundred
thousand
hosts.
That's
a
fairly
normal
requirement
for
switches
these
days,
and
they
are
certainly
not
broadcasting
mdns
over
a
200,000
host
domain.
I,
don't
know
how
big
they
go,
but
it's
not
that
big.
C
N
Right,
so
all
the
the
the
you
know,
the
the
MDS
proxies,
particularly
the
ones
in
the
wireless
access
points,
access
points
which
are
going
to
severely
limit
the
bandwidth
of
this.
On
most
large
enterprise
deployments,
I
mean
we're
going
to
have
places
where,
in
the
current
things
we
would
have
had
connectivity,
and
now
we
won't
now.
The
thing
that
makes
me
feel
better
about.
N
That
is
the
thing
my
friend
Justin
always
reminds
me
is
yes,
but
if
you
get
media
first,
none
of
this
matters
right,
and
so
that
makes
me
chill
out
a
lot,
but
I
think
that
that
type
of
stuff
needs
to
be
a
little
bit
discussed
somewhere
and
I've
you
up
primary
as
an
ice
issues
on
STP
issues
on
a
music
issue,
but
it's
also
not
exactly
an
IP
handling
issue.
It's
it's
an
issue
to
do
with
the
fact
that
we're
allowing
mdns
to
be
used
in
ice
I
think
we
need
to
discuss
that
somewhere.
N
C
I
mean
one
reason
why
this
kind
of
distant
curious,
the
assumption
ice
proper,
always
was
you
know,
advertised
every
address,
you've
ever
heard
of
there's
no
such
thing
as
a
privacy
issue
and
the
privacy
issues.
Things
only
really
came
in
with
WebRTC,
and
that's
where
you
start
having
these
policy
questions
of
which
candidates
do
or
don't
you
advertise,
which
this
is
a
you
know,
subset
session
related
topic
to
so
we
don't
have
anywhere
that
discusses
that
outside
of
IP
handling,
because
it
never
was
an
issue
before
now.
C
Obviously
you
know
I
mean
I,
don't
know,
maybe
it's
not
an
issue
outside
the
browser
context,
because
everywhere
else
you
know
you.
If
you
are
the
thing
using
the
I
stack,
you
can
find
out
all
the
IP
addresses
and
do
whatever
you
want
with
them.
It's
only
this
weird
trust,
separation,
B,
JavaScript
and
browser
that
this
becomes
an
issue.
L
L
E
L
L
D
Yeah
I
think
the
second.
The
second
edition
here
is:
oh
sorry,
Adam
Roche
again
individual
of
I.
Think
the
second
paragraph
here
is
pretty
uncontroversial.
The
first
one
I
think
is
probably
going
to
cause
us
to
put
probably
a
fairly
large
paragraph
in
the
security
iteration
section
that
addresses
the
issue.
C
U
V
L
N
W
R
L
L
D
Adam
wrote
so
I
I'm,
actually
a
little
more
in
favor
of
splitting
this
out,
I
think
predominantly
because
I
suspect,
once
we
get
into
the
two
issues
that
he
described,
we're
gonna
get
the
edge
to
those
in
life
and
we're
gonna
end
up
the
point
where
it's
it's
a
non-trivial
effort
that
does
delay
this
document
and
specifically
because
the
follow-on
work
Friday
handling
relies
on.
This
is
a
separate
item.
That's
gonna
take
a
little
bit
of
time
as
well.
I
think
it's
probably
doesn't
probably
harm
to
that
mechanism.
D
I'd
like
to
make
sure
we
get
this
right,
and
we
just
we
just
came
up
with
this,
and
you
know
over
the
course
of
an
hour-
expanded
parenthesized
five
words
into
you
know
probably
five
or
six
paragraphs
already
and
that
that's
a
reasonable
size
for
a
draft,
but
that's
all
it
takes
and
we're
done
good
I,
don't
think
up.
Looking
page.
S
G
T
S
P
K
N
C
N
In
this
thing
we
could
just
we
should
just
carve
a
clear
carve
out.
It's
just
like
for
this.
For
the
purposes
of
this
draft
right
here
how
this
works
with
fqdn
is
not
discussed
whatsoever
that
will
be
dealt
with
in
draft
X
and
a
pointer
to
it,
and
if
it's
a
two-page
draft
that
has
finished
two
weeks
from
now,
I
will
do
a
dance
by
Justin,
scotch
or
something
I
mean
like
I.
Don't
know
that
this
needs
to
slow
things
down
at
all
is
what
I'm
trying
to
say.
F
F
N
I
mean
III
think
the
reality
is
you
know
what
no
matter,
what
the
current
specifications
say.
They
don't
actually
tell
you
enough
to
implement
fq
d
ends
and
though
they
say
it
is
a
should
it
say
should
that
everyone
took
as
they
must
not
in
the
actual
implementations,
and
that
we
should
just
reflect
reality
on
this
one
and
just
say
like
we're
punting
this
issue
right
this.
This
document
deals
with
IP
addresses
and
we're
punting
fq
d
ends
to
the
next
document,
which
will
correctly
deal
with
both
DNS
and
Indiana's.
N
Q
E
H
E
C
I
Teresa
Jessica,
you
generally
comment
I
think
we
should
start
talking
about
our
SC
five
to
four
or
five,
because
that's
going
to
be
replaced
by
RFC,
eight,
four,
four,
five
very
soon.
Actually,
it's
part,
that's
part
of
cluster
this,
so
it's
gonna
take
a
while,
but
it's
basically
all
four
eight
and
everything
is
done
so
and
that's
ice
based
in
the
new
version
of
mice.
So
I
think
that
that's
and
that's
also
what
this
document
is.
Referencing
is
also
I
think
we
should.