►
From YouTube: WebPerfWG meeting - November 5th 2019
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
B
A
A
D
A
A
Yeah
and
we're
talking
about
your
an
event:
loop,
I'm
guessing
so
a
lot
windows,
the
chair
and
event,
loop,
I'm
guessing.
E
B
E
Yeah
when
they
share
I
guess
I
wouldn't
do
similar
similar
window
origin,
I'm,
agent
and
I
think
this
is
this
is
about.
This
is
probably
a
spec
book
just
from
us
moving
from
more
event,
loop
model
to
hooking
onto
the
window,
object
and
not
realizing
that
there
could
be
multiple
window
objects.
Communicating
and
I
have
a
PR
up
which
should
address
it.
It
basically
hooks
onto
the
and
same
indoors
and.
D
F
E
I
think
that's
kind
of
what
the
PR
does
I
originally
had
it
with
event
loop
and
instead
of
and
sort
of
similar
originated
window
agents,
but
and
suggested
that
we
should
make
it
some
similar
window
agents,
which
kind
of
make
sense,
because
you
shouldn't
be
able
to
observe
the
difference
from
the
way
up
at
least,
and
it
would
give
a
bit
more
flexibility
to
the
user
agent.
Well,.
D
E
D
D
Something's
not
right
here
either.
You
have
to
say
that
list
of
idle
callback
is
associated
with
each
agent
and
then
all
similarly
similar
origin
user
in
and
share
single
list
right
like
right
now,
concept
a
list
of
Idaho
back
is
associated
with
the
concept
of
like
set
up
a
similar
origin
window
agents
know.
G
D
E
D
D
E
D
F
A
Let
me
try
to
see
if
I
got
anything
up
from
that
conversation,
so
we're
going
to
PR
HTML
with
the
concept
of
the
idle
tasks
like
the
I'll
test
you
and
then
try
to
hook
up
based
on
that
or
I.
E
A
D
E
F
G
E
B
D
D
D
D
E
Have
to
we
have
I
mean
the
point
of
the
deadline
is
to
give
a
forward-looking
prediction,
so
we
need
to.
We
need
to
decide
how
that
deadline
is
at
the
start
of
an
idle
period,
and
we
can't
change
it
right.
We
can
reset
it
back
to
zero,
where
we
can't
change
it
during
the
idle
period,
because
it's
used
as
muster
went
to
start
the
next
idle
period
like.
D
D
Guess
then
one
thing
we
have
to
take
into
town
as
water
things.
The
UA
could
reasonably
predict
right.
We
can't
say
random
stuff,
like
you,
a
have
to
guess
like
or
like
using
heuristics
to
estimate.
Why
not
like,
for
example,
in
wicked,
like
I'm,
starting
to
think
it
is
literally
impossible
to
figure
out
when
the
next
frame
will
happen.
E
So
I
mean
the
point
of
this
prediction
is
to
provide
some
knowledge
to
the
developer
as
to
what
time
they
might
have
available.
I'm
fine,
with
providing
with
providing
some
additional
kind
of
semantical
and
invariance
of
like
it
is,
is
the
the
maximum
it's
the
minimum
of
the
next
timer
and
the
next
frame,
if
that's
available
to
you
and
some
other
stuff,
but
I
think
I
mean
in
chrome.
We
are
using,
like
our
internal
tasks,
sources
to
produce
this.
E
E
E
D
E
A
D
D
A
D
A
G
E
A
E
So
I
can
talk
about
how
Chrome
data.
So
since
it's
a
forward-looking
prediction,
we're
just
looking
at
tasks
that
have
already
been
queued
with
some
delete
time
to
be
scheduled
and
like
the
the
spec
of
request.
I,
don't
call
that
means
that
any.
If
there's
any
task,
that's
cuter
on
immediately
we're
not
going
to
start
an
idle
period,
we're
the
low
priority
task
queue,
and
so
it's
only
for
every
task
queue
that
we
have.
E
We
look
at
what's
scheduled
in
the
future,
and
we
take
the
minimum
of
that
and
we
also
take
into
account
our
composite
curves
decision
as
between
it
thinks
when
it
plans
to
start
drawing
the
nice
frame
of
when
it
starts
and
scheduling
tasks,
to
draw
that
next
frame
for
requestanimationframe.
The
things
like
that.
Okay.
G
E
If
you
start
trying
to
call
a
callback-
and
you
see
they're
gonna
higher
priority
tasks,
you
can
just
cancel
that
and
then
you'll
try
again,
but
you
can
really
mean
once
the
call
back
you
started
executing
like
you're
in
user
JavaScript
land.
You
can
do
anything
really
about
that
and
the
the
deadline
on
the
object
itself
is
is
fixed,
yeah,
I,
think
I.
Think
we
do
have
the
ability
yeah.
We
do
have
the
ability
to
set
the
deadline
back
to
zero.
So
that's
the
other
way.
E
D
E
D
A
So
the
value
it's
we
didn't
really
discuss
it
more,
but
right.
A
E
G
D
E
D
I
think
is
suppose
there
was
a
pending
task,
but
most
the
pending
tasks
will
right.
He
need
any
anyway.
So
what
are
the
examples?
Water,
a
concrete
example
and
that's
a
chromium
get
called
that
is
async,
but
why
the
user-agent
to
differ.
If
we
know
like
that's
in
the
future
and
how
far
in
the
future
like
I,
just
can't
think
of
anything
and
inspect,
that
would
be
the
case
other
than
some
media
register.
Yeah.
E
D
D
A
Yeah
well
defining
I'm
guessing
we
could
like
both
from
Chrome
and
Firefox,
dig
up
those
cases
and
outline
them
right.
I
mean.
E
E
D
E
E
Think
so
I
mean
we
do
it,
we
don't
we
don't
well,
it's
so
I,
don't
know,
and
there
is
a
task
queue
which
contains
which,
which
we
know
where
the
pending
tasks
they're
coming.
It's
and
they're,
no
they're,
not
categorized
and
by
the
api's.
There
are
scheduling
these
things
because
in
some
cases,
there's
no
wave
api's,
they're
scheduling,
different
tasks
could
be
things
like.
E
A
E
D
Yeah
I
mean
so
fun
fundamental.
This
seems
poor
Manik
right
like
the
fact
that,
like
we
can't
implement
this
seems
to
indicate
that,
like
we
have
failed
specking,
this
so
I
don't
know
I,
don't
know
what
what
would
what
we
will
do
like?
Well,
maybe
we'll
just
say
just
waiting
for
the
time.
It's
good
enough.
I.
D
D
E
We
kind
of
have
a
similar
problem
like
we.
We
rely
on
the
browser
process
to
tell
us
when
to
draw
frame,
and
but
we
do
have
some
idea
of
the
frame
frequency
we
are
expecting
and
we
obviously
know
when
the
last
frame
was,
and
so
we
use
that
to
estimate
when
the
next
reamers
and
it's
obviously
not
accurate
entirely.
But
as
I
said,
the
deadlines
can
have
meant
as
a
head
of
time
prediction.
It's
not
meant
to
be
accurate
as
to
exactly
when
the
neck
timer
run.
D
E
Yeah,
it's
it's
a
prediction
and
we
can
spec
it
to
a
reasonable
degree,
but
it's
always
going
to
be
a
heuristic
in
a
prediction
and
the
other
part
of
the
spec
arm
or
the
accurate
is
the
best
it
say:
the
deadline
to
0
when
you
actually
have
a
task-
and
you
know
for
sure
that
there's
some
more
high
priority
task
knife
those
base
I-
think
we
should
spec
more
concretely
because
they
do
actually
matter.
But
this
one
I
guess
within
a
range.
But
it's
a
production.
A
So
do
we
have
agreement
that
specified
like
properly
specifying
the
time
or
deadline
is
zero
and
then
defining
the
value
of
deadlines
to
be
based
on
available
timers,
some
prediction
of
the
next
train
and
then
other
tasks
that
the
UA
may
be
aware
of.
Would
that
work
from
like
we
asked
it?
Would
that
work
from
your
perspective,
I
think.
D
So
in
a
closed
like
6:1
step
5,
it
says
the
users
and
should
choose
a
deadline
to
ensure
that
no
time
quicker
tasks
you'd
be
delayed
even
to
go
back
for
photos,
blah
blah
blah.
It
should
be
set
to
the
minimum.
Will
close
this
time
out
and
disturb
actor
timers.
A
set
vir
set
timer
set
to
internal
scheduled
time,
run
time
for
pending
animation,
Kobach's
via
requestanimationframe
right.
So.
E
D
I
mean
it's
very
what
he
was
way
of,
saying
it,
but
basically
we're
saying
that
it
should
be
minimal
of
either
they
closes
timer,
dumb,
timer,
the
fires
or
the
next
time.
Mike
request.
Animation
call
like
this
is
also
weird
like
a
next
time.
Rendering
opportunity
happens,
probably
the
better
way
to
say
it
like
I,
don't
know
why
it's
basically
goes
out
request
animation
of
a
shoe
you'll
be
really
like
when
rendering
opportunity
happens.
D
A
D
A
So
one
is
list
of
available
timers,
the
other
is
requestanimationframe
or
the
next
rendering
opportunity,
which
is
a
better
term
for
that,
and
then
the
third
would
be
a
big
blob
of
you,
a
define
words
or
you
a
define
predictions
that
can
include
everything.
That's
currently
in
the
tech
like
internal
timeouts,
audio
processing
or
other
internal
tasks,
which
are
which
the
UAE
deems
important.
So
all
that
you
be
a
note
under
the
banner
of
you,
a
defined
I.
C
B
A
Give
you
that
and
then
Stephen
is
going
to
try
the
transcripts
or
maybe
that
will
give
us
more
detailed
conversation
and
we're
at
time.
So,
let's
just
in
terms
of
action
items
Ross
can
you
take
on
the
specifying
the
bits
that
we
said
that
need
to
be
better
specified.
So
the
studying
the
deadlines
to
zero
and.
A
E
A
Long
masks
and
talent,
aggression
would
fetch,
which
are
both
exciting
topics
that
we'll
have
to
wait
till
next
time.
But
I'm
super
happy
that
we
got
to
the
bottom
of
this
particular
issue
because
it's
been
dragging
for
a
while.
So
thanks,
all
and
I
will
see
you
all
in
week
and
a
half,
if
not
before
or
in
two
weeks,
even
yeah,.