►
From YouTube: IETF111-WISH-20210730-1900
Description
WISH meeting session at IETF111
2021/07/30 1900
https://datatracker.ietf.org/meeting/111/proceedings/
A
A
It's
time
to
get
going
all
right,
so
it's
friday
morning
session
one.
Hopefully
you
guys
have
seen
some
of
these
slides
before,
but
welcome
to
wish.
Here's
the
no
well
again,
hopefully
you've
seen
this.
The
basic
gist
is,
if
you
know
anything
about
ipr,
you
should
declare
it
if
you're
going
to
participate,
please
make
sure
to
follow
the
professional
guidelines
that
we
set
forth
here,
any
harassment
code
of
conduct
etc.
If
you
have
any
questions
about
these,
you
can
contact
the
chairs
or
others.
A
A
A
A
D
D
D
A
E
E
Extensibility
section.
That
was
also
requested
in
the
which
chapter
and
added
a
notion
that
I
want
to
discuss
in
this
in
this
in
this
meeting
about
the
tune,
server
configuration
and
also
just
send
for
file
responses
to
unit
union
method.
So
if
we
can
go
to
the
nexus
slides,
so
we
can
start
seeing
each
of
these
changes
in
the
tile.
E
So
basically,
what
we
have
changed
in
the
session
system
setup
is
that
now,
when
the
sd
http
processor
is
done
to
the
weapon
point
to
the
to
start
the
in
the
session,
what
we
return
is
a
a
url
that
it
is
a
use
that
it
is
the
the
url
of
the
new
web
resource
that
has
been
created
and
that
it
will
be
used
for
doing
the
the
ice,
the
ice
stuff
and
also
the
hdb.
The
light
methods
and
the
rest
of
the
protocols
is
just
the
same.
You
do
the
sdp
of
an
answer.
E
Over
there,
the
initial
sct
post,
the
dc
depos,
http
post,
is
carries
a
body
that
is
the
sdp
offer
and
in
the
sdp
answer
of
the
of
the
weapon
point
you
have
the
the
answer.
So
once
it,
this
is
done.
They
I
say
ice
is
done
between
them,
the
webrtc
producer
and
the
and
the
media
server.
The
isis
is
set
up,
and
then
the
rtp
and
rtp
flow.
This
is
carried
over
as
normal.
E
E
E
So
it's
something
that
I
want
to
add
in
the
in
the
in
the
draft
and
that's
that
and
that
just
sending
the
ice
candidate
for
one
single
m
line
is
enough,
or
we
can
just
try
to
make
a
modification
today
to
the
mime
time
to
also
support
this.
The
the
I
say,
attributes
both
the
username
and
password
and
also
the
candidate
at
a
session
level
and
get
rid
of
this
constraint.
E
This,
this
trickle
is
restarts
and
if
that's
happened,
no
further
requests.
The
http
request
has
to
be
done,
but
still
if
the
web
client
gathers
additional
candidates
after
the
sap
is
sent,
it
must
send
a
stupid
extra
request
to
show
the
so.
The
media
server
gathers
the
information
from
from
the
from
the
ice
packets
that
has
to
be
regardless,
if
the
hdb
but
is
successful
or
not.
E
C
F
And
there
I
think
about
whether
we
wanted
to
make
modifications
to
the
sdp,
frag
format,
to
accommodate
multiple
m
lines
or
whether
we're
okay
having
them
like
simply
apply
to
everyone
because
of
bundle.
I'm
happy
with
the
second
approach.
You
know
the
the
less
things
we
have
to
touch
to
make
this
work.
I
think,
probably
the
better.
E
C
B
Julius,
can
you
hear
me
okay,
so
I
have
two
comments
here.
One
is
that
I
don't
agree
so
first.
I
think
this
is
some
very
important
work.
If
we
don't
manage
to
have
something
that
is
accepted
by
both
server
authors
and
client
authors,
we're
going
to
be
stuck
with
rtmp
forever,
and
that
would
be
really
bad.
So
I'd
really
like
this
to
succeed
and
I'm
afraid
it's
going
to
go
the
weight
of
ipv6,
which
is
our
only
hope
for
the
internet,
and
it's
been
30
years
and
we're
still
at
40
deployment.
B
So
there
are
two
things
I'd
like
to
mention
on
this.
One
is
that
I
disagree
with
the
fact
that
the
server
is
allowed
to
refuse
trickle
and
to
refuse
a
restart.
I
mean
writing.
We
all
have
everyone
who
has
written
a
server
already
has
sub
support
for
ice
trickle
and
already
has
support
for
ice
restart.
B
That
is
not
a
problem
for
server
authors.
What
you
want
is
to
make
things
as
easy
as
possible
for
the
clients
and
having
to
be
able
to
deal
with
the
possibility
of
a
four
or
five
when
you
attempt
trickle
is
really
bad.
So
I
see
that
lorenzo
what
lorenzo
is
suggesting
in
the
chat
is
that
it
should
be
able
for
the
client
to
find
out
beforehand
whether
it
can
do
trickle
or
not.
E
I
I
don't
mind
to
make
it
mandatory.
It
should
be
not
I
mean,
because
I
mean
I
I
don't
implement
it
there
or
I
don't
require
negotiating
candidates
in
signaling
or
idol
superior
starts
in
my
media
server,
and
I
have
I'm
very
happy
with
that.
I
have
never
made
it
in
fact
and
because,
as
I
said,
what
I,
what
what
I
really
learned
the
candidates
from
is
from
the
ice
requests
and
not
from
the
from
the
signaling
side,
but
anyway,
I'm
happy
to
to
make
it
mandatory
to
to
support
it.
Okay,
my.
E
B
Usually
there
is
a
limited
number
of
servers,
but
what
we
are
hoping
for
is
to
make
it
easy
to
make
clients
those
are
features
that
can
be
ignored
by
the
client.
They
cannot
be
ignored
by
the
server.
So,
let's
put
the
let's
put
the
difficult
bits
in
the
server
and
let's
make
life
easy
for
the
clients.
B
The
second
point
is
that
for
us
server
authors,
it
would
be
very
helpful
if
we
could
have
a
reference
client.
So
I
asked
on
the
mailing
list
about
what
is
the
reference
client,
and
I
think
it
was
you.
I'm
not
sure
answered
that
with
a
few
pointers,
but
without
being
clear,
which
is
the
one
I
should
use.
B
So
I
would
like
to
request
from
the
people
who
are
promoting
this
document,
which
I
think
is
very
important,
that
you
give
us
a
client
that
is
easy
to
install
with
clear
instructions
about
how
to
install
it.
I'm
quite
willing
to
add
support
for
wish
to
my
server.
I
think
it's
a
great
is.
I
think
it
should
be
easy
to
do,
but
I
need
an
easy
to
compile
client
to
put
in
the
lab.
E
E
D
E
B
Okay-
and
the
third
point
I'd
like
to
make
is
that
most
of
the
server
authors
have
a
web
client
that
works,
and
that
means
that
the
incentive
to
implement
wish
is
pretty
low,
and
so
I
think
we
should
be
asking
ourselves
the
question:
what
is
the
killer
feature?
What
is
the
application,
which
is
the
killer
application
that
will
encourage
people
to
implement,
wish
tuition
their
servers?
G
A
little
concerned
that
this
specification
of
the
of
the
ice
is
a
little
underspecified
in
particular,
I
don't
think
trickle
ice
sdp
frag
can
be
used
for
ice
restarts
in
the
place
where
it's
defined,
in
particular,
there's
issues
where,
if
the
server
changes
its
ports
on
a
nice
restart,
then
you
need
to
make
sure
that
things
are
that
you
have
a
clear
transactional
model
of
offer.
Answer
for
the
ice
restarts,
and
also
you
it's
not
clear.
Are
you
not?
I
don't
think
you're.
G
G
E
E
G
E
C
Yep
sorry
waiting
for
my
microphone
to
kick
in
nils
as
an
individual
on
the
previous
slide,
you
mentioned
ice
consent
as
a
way
to
detect
whether
connection
is
dead
or
not.
I
just
wanted
to
highlight
to
the
to
the
working
group
that
the
ice
consent
by
default
according
to
the
specs
only
times
out
after
30
seconds,
so
I'm
not
sure
if
that's
for
the
scenario
is
good
enough
or
not.
C
I'm
mostly
highlighting
that,
because
the
current
implementation
in
chrome
basically
does
ice
content
a
lot
faster
than
these
30
seconds,
so
lots
of
people
seem
to
think
of
ice
consent
as
like
a
keep
alive.
I
can
quickly
verify
whether
a
connection
is
dead,
but
that's
not
according
to
the
spec.
It's
like
30
seconds,
and
I
mean
that's
obviously
to
up
to
the
working
group
to
decide
whether
30
seconds
is
fast
enough
or
too
slow.
E
Yeah,
I
think
that
that
the
30
second
is
okay
for
the
indicting
connections
and
at
ice
and
dtls
level.
Obviously
you
will
get
your
mobile
or
you
will
get
faster,
faster
indication
that
you
need
to
do
a
nice
restart,
for
example,
before
the
ic
is
timeouts,
but
I
think
that
it
is
also
good
to
to
know
that
we
have
a
specifically
for
the
for
the
server
to
know
when
the
client
has
one.
E
So
I
think
that
it
is
more
about
disconnect
the
free
resources
in
the
in
the
in
the
server
side
once
the
the
client,
for
example,
this
connector
turns
off
because
it
runs
out
of
battery
to
know
that
we
can
detect
that
with
a
30.
Second,
that
is
is
fairly.
This
is
I
mean
I
run
the
currently
are
running
westshock
and
this
connections
are
the
this,
the
10
disconnecting
by
socket
layers
at
30
seconds.
Also,
so
it's
more
or
less
what
I'm.
E
F
Respond
quickly,
what
julius
was
saying
about
motivation
for
servers
implementing
this,
and
while
I
agree
that
having
screen
sharing
mobile
devices
would
be
really
awesome,
there
is
a
pretty
strong
motivation.
I
think
at
least
you
know
speaking
as
someone
who
operates
that
kind
of
thing
to
be
able
to
make
use
of
the
rich
set
of
capabilities
that
existing
broadcast
tools
implement.
So
I
don't.
E
B
E
I
I'm
really
just
saying
that
we
hit
a
level
of
complexity
and
ice
that
makes
it
more
difficult
to
implement
this
now
and
get
it
delivered
now,
then,
then,
the
thing
to
do
is
to
stop
at
the
minimum
level
we
can
get
away
with,
and
then
a
lot
of
these
things
that
we're
trying
to
do
trying
to
talk
about
will
get
fixed
with
nicer.
In
you
know
a
subsequent
revision
or
something.
E
Yeah,
I
agree.
The
only
thing
is
that,
well,
I
think
that
we
can
revisit
before
before
publishing,
and
I
mean
I
don't
think
that
we
can
take
the
decision
right
now
to
go
to
nicer
and
remove
fights
together.
I
mean
I
think
that
nicer
should
be
the
way
to
go,
but
probably
is
too
early,
but
I
agree
that
probably
in
the
the
final
version
of
the
whis
of
a
whip
should
have
nicer
instead
of
of
a
triglycerides
restart
if
it
is,
if
it
is
ready.
G
Yeah,
I
wouldn't
want
to
gate
whip
on
nicer
because
I
feel
like
that
might
end
up
taking
longer
than
we
expect.
I
also
would
expect
that,
from
a
signaling
point
of
view,
what
nicer
needs
is
probably
not
going
to
be
that
different
than
standardized
I
mean
the
underlying
mechanism
is
going
to
be
the
same,
but
I
think
that
what
you
specify
at
this
level
is
probably
not
going
to
be
that
different.
J
And
speaking
as
hi
lagustan
speaking
as
the
the
guy,
whose
name
is
on
the
ice
nicer
draft-
yes,
nicer
is
nice,
but
it's
not
it's
not
ready
to
be
dependent
on
it's
likely
to
change
based
on
input
from
other
participants
and
and
depending
on
its
current
form.
Is
bad
luck
so
please
don't
depend
on
either.
J
E
E
So
this
same
as
as
said,
probably,
we
are
not
getting
a
an
option
to
explicitly
terminate
the
the
the
session
by
the
client
just
sending
an
http
delete
to
the
resource
url
and
and
when
the
the
sap
delete
is
requested
friendly
from
the
received
by
the
server.
They
will
also
terminate
the
ice
and
dtls
session
and
also,
as
adam
pointed
out,
a
amiga
returning
decision
must
follow
the
procedures
in
rpc
7675
to
also
regulate
the
the
consent.
E
This
is
one
this
was
specified
in
the
chat
that
we
have
to
to
to
provide
a
way
to
enable
extensions
into
for
future
or
a
way
to
extend
the
protocol
for
next
version,
and
there
was
some
discussion
about
how
to
do
eastern.
I
think
that
the
the
most
http
or
the
most
http
method
to
use
is
to
send
and
to
like,
create
a
new
url
for
this.
E
Each
of
these
extensions
that
it
will
be
handling
the
the
commands
between
the
clients
and
servers
and
the
way
to
and
how
to
the
server
can
signal
back
the
client.
With
extension,
does
it
support,
is
just
to
create
and
to
return
a
a
link,
a
header
with
the
url
of
the
of
the
of
the
extension,
and
also
adding
a
real
attribute
that
allows
in
that
start
with
idf,
but
on
sweep
server,
side,
whip
and
then
the
name
of
the
extension.
E
This
is
a
for
example.
How
a
web
push
is
is
doing
the
the
to
to
signal
the
client
that
what
is
the
url
to
to
do
all
the
all
the
commands
of
of
web
boost,
so
also
they
will
ensure
that
all
the
extension
has
to
be
in
registering
in
indiana
and,
for
example.
This
would
be
an
example
of
how
to
a
back
channel
bit
form
from
the
server
side
or
for
sending
an
event
from
the
server
to
the
client
could
be
implemented
by
using
a
server
side.
E
Event
that
I
prefer
to
do
to
http
building,
because
it
is
supported
by
natively,
supported
by
by
my
browsers,
and
you
don't
have
to
specify
what
is
the
the
web
format.
So
it
would
be
just
to
you,
send
the
link
to
a
url
to
connect
the
the
server's
id
then
object
from
the
browser,
and
you
will
start
receiving
the
event
sent
by
by
the
by
the
server
in
this
case,
and
obviously
all
the
protocolization
must
be
optional.
E
E
Is
sending
back
the
link
from
the
for
doing
the
all
the
puss
and
all
the
commas?
I
don't
know
how
web
push
works,
but
I
read
the
in
the
the
the
spec
and
I
was
using
this,
so
there
was
some
some
questions
about
it.
We
should
also
specify
how
to
configure
and
determine
student
configuration
for
the
clients.
E
I
think
that
it
is
say
something
that
it
is
interesting
and
we
should
at
least
draft
and
explain
how
this
could
be
used.
I'm
not
sure
if
making
this
normative
or
just
making
it
optional.
We
have
two
methods
of
doing
that.
First,
is
just
configuring
manually,
the
url
of
the
two
servers
on
the
not
the
eye
skin
and
the
credentials
or
this
is.
I
would
prefer
to
to
to
reduce
the
the
the
justin's
draft
phone
for
http
base,
then
basic
configuration,
because
I
think
it
fits
quite
well.
You
will
just
need
to.
E
If
you
are
a
service
provider,
you
will
just
need
to
to
provide
two
and
two
different
urls
one
for
receiving
the
tool,
server
configuration
and
another
one
for
whip,
and
you
this
will
allow
you
to
to
configure
or
and
to
configure,
to
insert
based
on
certain
thing
credentials
that
will
not.
So
you
don't
need
to
give
them
a
long-term
protection
to
the
user.
Nothing
like
that.
So
I
think
that
this
fits
quite
nicely
with
the
what
it's
with
whip,
and
I
think
that
is
the
one
that
we
should
be
using.
E
Another
thing
is
that
I
think
that
it
is
better
to
just
use
two
different
authentication
information
for
the
for
the
web
for
the
web
url
from
the
tool
server
configuration
url,
because
you
may,
for
example,
use
a
different
tool
server
provider
than
the
one.
You
are
pushing
your
your
your
your
stream
to,
for
example,
you
may
use
a
twilio
for
two
servers
because
it
is
widely
used
and
then
just
push
to
whatever
extreme
server
you
you
want
to
use.
So
I
think
that
this
we
should
keep
the
authentication
sa
separated.
B
E
The
problem
is
the
problem:
is
that
once
you
are
doing
the
std
pose,
you
already
need
to
be
to
send
this
to
the
offer
and
to
send
the
sdb
offer.
You
already
need
to
know
the
the
in
the
url,
so
we
need
to
do
we
need
a
way
or
we
need
to
configure
the
the
the
url
of
the
two
servers
on
the
credential
before
doing
the
the
initial
std
post.
So
we
need
to
show
that
you.
B
E
B
E
B
F
Is
actually
kind
of
similar
to
that,
except
for
it
it
like
does
okay
with
bcp
190,
which
is
basically
we
could
throw
it's.
We
could
have
the
rest
api
for
access
to
turn
credentials
be
advertised
on
a
dot.
Well
known,
endpoint
right,
so
you
could.
You
could
have
something
that
points
there
to
where
to
retrieve
these
credentials
from,
and
I
think
that
would
probably
get
julius
all
the
way
to
what
he
wants
worse.
You
put
one
url
in
there
and
then
they're.
F
B
B
F
Now
the
thing
that
I
wasn't
cue
to
say
is,
I
am
kind
of
hesitant
to
put
this
in
at
you
have
like
recommended
on
the
slide
here,
I'm
hesitant
to
put
it
in
at
any
normative
level,
because
I
don't
really
want
us
to
block
on
progression
of
the
the
document
you're
citing
here
I
mean
that
it's
from
2013
right,
it
hasn't.
It
hasn't
been
touched
for
more
than
a
little
bit
of
time,
and
I
mean
I
see
that
justin's
in
the
chat.
F
Maybe
he
can
speak
to
whether
he
has
plans
to
push
this
forward,
but
it's
not
it's
like.
We
couldn't
do
it
in
the
charter
of
this
working
group,
which
means
we'd
have
to
have
someone
else,
take
it
on,
and
so
I
think,
maybe
mentioning
it
as
a
potential
mechanism
in
which
would
be
a
good
thing,
but
putting
it
in
a
normative
level.
I
gives
me
a
little
bit
of
heartburn.
E
E
F
G
Yeah
so
I
mean
my
proposal,
is
I
mean
it's
similar,
but
rather
than
having
that
well
known,
have
like
some,
I
guess
you'd
almost
call
it
a
preflight
url
where
you
just
where
the
offer
where
the
exchange
is.
You
query
the
url.
It
gives
you
a
list
of
you
know
the
url
of
that
you
use
for
the
sdp
offer
answer
and
the
url
for
you
use
to
get
the
turn
to
configuration
and
maybe
url
for
any
other.
You
know
information.
G
We
might
decide
to
extend
this
with
in
the
future
and
then
that
just
gives
you
that
sort
of
the
almost
you
know
like
this
is
all
the
information
you
need
to
do
to
do
this,
and
then
you
can
query
the
other
ones.
You
know,
as
you
like,
based
on
their
properties
and
whether
that's
a
list
of
link
headers
or
json.
I
don't
care,
but
I
feel
like
that
might
be
rather
than
not
well
known,
and
that
might
be
somewhat
cleaner
authentication-wise,
but
I
don't
know.
E
E
H
Point
out
that
that
document,
the
rest
thing
got
subsumed
by
what
became
rfc
7635,
which
is
basically
an
oauth
mechanism,
for
you
know
connecting
to
turn
servers.
So
you
know
there's
a
little
more
machinery
there,
but
the
the
the
principles
are
really
the
same,
and
I
I
think
that's.
The
kind
of
thing
you
would
probably
you
know
would
be
would
be
easiest
to
lean
on.
E
J
H
Well,
it
might
be
easiest
just
to
kind
of
get
that
progressed.
You
know
and
implemented
in
webrtc,
rather
than
trying
to
pre
and
progress
another
document.
You
know
that
then,
and
have
that
as
a
dependency
here
but
yeah
I
get
the
exact
status
like
this
may
be
the
first
time.
Something
is
really
rotating
a
full
dependency
on
on
that
technique.
E
E
E
E
Well
or
something
obvious
that,
if
you
we
need
to
support
gorsan
that
http
preflight
http
options
has
to
be
supported
in
to
to
be
able
to
to
send
requests
by
specifically
web
browsers.
But
I
don't
know
if
you
think
that
we
have
to
to
explicitly
put
it
in
the
draft
that
how
to
support
it.
So
anyone
is
not
missing
that
when
implementing
a
server
or
if
it
is,
is
something
that
we
take
as
it
is
a
common
knowledge
of
a
common
http
knowledge.
E
E
F
Doing
this,
the
way
that
things
are
on
the
web
right
now
in
part,
because
it
requires
the
least
amount
of
changes,
if
we
want
to
reiterate
it
by
basically
pointing
out
hey
if
you're,
implementing
a
server
and
you
want
it
accessible
from
more
than
just
your
origin.
Then
you're
going
to
need
to
do
this
thing
over
here
and
then
point
to
the
specification
of
course,
but
I
think
doing
anything
beyond
that
is
going
to
be.
F
E
B
B
Now
what
the
thing
that
I
think
is
really
important
here
is
that
if
there
is
something
that
is
required
because
of
web
standards,
please
explain
it
to
us.
You
remember.
We
had
a
long
discussion
on
the
mailing
list
because
we
completely
misunderstood
each
other,
and
the
reason
for
that
is
that
I
understand
nothing
about
the
web.
I
understand
a
little
bit
about
networking,
but
the
web
is
completely
alien
to
me.
So
if
there's
anything
here
that
you
need
to
explain
to
people
who
are
not
familiar
with
the
web
platform,
please
do
be
explicit.
E
F
A
quick
response
to
the
assertion
that
anything
on
the
web
is
going
to
be
running
on
the
same
origin
and
basically
can
use
proprietary
protocols.
I
want
to
point
out
there
are
a
handful
of
web-based
broadcast
tools
right
now
that
are
run
by
third
parties
as
a
service
that
to
which
that
doesn't
apply.
So
you
can
look
at
things
like
melon,
app
and
stage
10
and
vdo.ninja,
all
of
which
are
like
professional
broadcast
tools
that
are
based
on
web
browsers,
and
course
handling
is
going
to
be
entirely
applicable
to
them.
F
E
E
A
Okay,
great,
so
one
of
the
things
that
we
need
to
do
is
get
this
working
group
or
get
this
draft
adopted
into
the
working
group.
I
thought
I'd
send
a
message,
but
I
just
noted
that
I
can't
see
it
on
the
mailing
list,
so
I'm
gonna
resend
that
message
to
get
the
get
the
process
started.
So
please
indicate
your
support
or,
if
you
don't
want
to
support
adoption,
tell
us
why
and
I'll
get
that
message
to
the
mailing
list.
Asap
and
I.
D
A
F
A
A
All
right
there
we
go.
Nine.
Nothing
seems
like
I'll
confirm,
we'll
confirm
this
on
the
list.
Obviously,
but
it
seems
like
this
is
probably
going
to
happen
so
sergio
don't
spin
a
new
no3
version
yet
because
the
work
group
adoption
call
go
for
o2,
but
once
we
get
that
done
then
just
submit
and
which
is
something
one
under
a
new
the
same
name
and
then
the
o1.