►
From YouTube: WebPerfWG call - July 2nd 2020
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).
A
B
It
sounds
good
to
me,
so
we
have
a
couple
things
today:
I
think
we're
gonna
start
out
with
Nicolas.
If
you
don't
mind
the
presentation
on
that
from
last
week,
around
page
visibility,
and
then
we
had
a
couple
things
around
tea
back
and
some
preps
for
that
coming
in
the
fall.
Obviously,
it's
going
to
be
a
virtual
conference
this
year,
but
there's
some
things
that
we
can
do
to
kind
of
get
ourselves
ready
for
that
and
hopefully
have
a
productive
tea
back
20
20.
B
C
Wrote
a
document
where
I
documented
some
of
the
problems
that
we've
been
hearing
and
one
of
them
was
the
fact
that
you
don't
really
get
a
complete
history
of
page
visibility
from
the
browser,
because
you
can
only
register
an
event
listener,
and
obviously
that
means
that
you
do
not
get
any
changes
that
occurred
before
the
listener
was
registered
and
the
other
problem
is
for
paint
I
mean
particularly.
It
seems
that
you
really
want
to
take
age
visibility
into
account,
otherwise
pane
timing
metrics
may
be
pretty
inflated
right.
If
your.
C
C
Perhaps
that
should
make
mention
of
page
visibility,
but
the
idea
is,
we
want
to
add
more
API
surface
to
address
that.
For
the
other
issue,
though,
we
do
need
an
API,
and
so
the
proposal
is
to
add
a
new
performance
entry,
and
that
way
you
can
just
use
the
performance
observer
with
a
buffered
flag
to
get
any
previous
disability
states.
C
The
only
tricky
thing
is
the
very
first
one
I
guess
you
can
deduce
the
first
one
from
the
observer
changes
plus
the
current
state
of
the
page,
but
we
could
also
just
surface
the
first
one
I'm,
not
sure
which
is
nicer,
but
in
any
case
that
entry
would
of
completeness
for
the
disability.
Basically
the
history
of
the
page,
so.
C
So
name
could
have
just
the
actual
disability
states
that
the
entry
is
referring
to
like
what
was
the
state
at
the
time
when
this
entry
was
dispatched
and
then
entry
type
could
be
something
like
visibility
state
so
that
it's
consistent
with
the
other
ApS
that
we
have
for
visible
state
and
then
startin
could
be
the
time
at
which
this
change
occurred,
which,
if
we
do
surface
and
entry
and
initial
entry,
it
would
be
zero
or
something
like
that
cause.
If
a
sweet,
drink
navigation
start
and
then
duration
I
was
proposing,
it
could
be
zero.
C
Although
Phillip
Welton
from
chrome
death
well
commented
that
maybe
it
could
contain
some
useful
information,
like
duration
of
visibility,
change
handlers
or
something
like
that.
I,
don't
know
if
that's
useful
or
confusing,
and
but
that
was
I.
Don't
know
what,
let's
think
about,
if
you,
if
it's
confusing
for
duration,
to
be
zero
or
if
there's
something
useful,
we
could
set
it
to
I.
B
B
E
C
So
we
would
there's
two
options
here:
it's
yeah
inserting
the
first
one
artificially
so
that
you
know
the
initial
state
of
the
page
or
letting
the
developer
deduce
it
from
the
actual
changes
plus
current
disability
state.
Because
of
course,
once
you
have
those,
then
you
can
deduce
the
workers
one
I,
don't
know.
How
do
people
feel
about
those
two
options?.
A
B
C
C
C
Page
visibility
is
not
technically
performance,
so
why
are
we
linking
it
to
performance
observer?
My
reasoning
is:
we
want
exactly
the
kind
of
buffering
that
performance
observer
already
provides,
so
it's
much
more
convenient
to
do
it
this
way
and
in
addition,
it
helps
tie
the
performance
data
such
as
pain,
timing
to
the
disability.
States,
so
like
another
option,
would
be
to
define
a
page
visibility
observer
that
is
separate
from
performance,
observer
and
re-implement
everything
there.
A
C
I
just
went
to
make
sure
I
run
it
through
the
crook,
since
it
was
some
someone
raised
that
and
now
for
the
favorite.
My
favorite
part
in
every
API
designed
the
name
of
of
the
new
interface
I
mean
it
could
be
something
like
this
ability
state
change,
then
the
first
one
is
not
really
a
change,
so
it's
to
just
be
visibility,
state
or
visibility,
state
and
tree.
B
C
And
it
could
even
be
without
performance
there.
Its
precedent
in
layout
shift,
so
I
mean
not
saying
it's
great.
To
have
I
mean
close
relative
is
not
really
about
performance,
headed
right
so
and
similarly,
visibility
state
could
just
drop
the
performance,
but
if
you
think
it's
better
with
performance,
we
can
also
put
it
I.
E
E
B
F
C
C
B
C
I
feel
like
for
this
one
in
particular,
we
can
make
the
buffer
slightly
smaller
than
usual,
so
with
perhaps
shouldn't
worry
too
much
about
it,
but
if
you
think
it's
better
to
gather
some
data
before
setting
the
buffer
size,
then
I
can
I
can
say.
Let
me
do
that
and
sorry
you
had
another
question
right.
Oh.
B
Just
I
don't
think
in
a
performance
observer
interface
right
now,
when
you
call
the
buffered
flag,
I,
don't
think
you
can
know
whether
the
buffer
was
reached.
B
So
you
don't
know
if
you
are,
if
you
did
hit
the
limit
other
than
you
know,
having
happening
to
know
that
particular
user
agents
buffer
limit
and
seeing
it's
that
that
that
that
count
right.
So
what
I
think
is
definitely
an
edge
case,
but
you
know
I,
don't
know
if
that
would
be
useful
for
other
things.
B
C
That's
that's
a
valid
point.
Perhaps
that's
something
we
could
add
to
the
observer
callback,
some
slack
where
it
says
that
some
buffer
was
reached
yeah
right
now,
it's
a
little
hard
for
you
to
really
know
because
you
could
get
the
number
of
entries
equal
to
the
buffer
size,
but
even
that
doesn't
technically
guarantee
that
the
buffer
is
full
well
actually.
B
B
C
A
C
G
So
Nicolas
I
just
wanted
to
add
one
more
thing:
I
think
we
already
covered
it,
but
maybe
for
just
notes
regarding
whether
or
not
to
add
an
initial
page
visibility
state
to
the
entry.
It
makes
sense
for
convenience
that
it
would
be
there
I
think
we
settled
on
that
which
is
good,
but
another
reason
perhaps
to
have
it
is.
G
If
the
first
entry
was
the
first
change,
you
would
have
to
sort
of
infer
the
first
state
by
taking
the
opposite
and
as
long
as
page
visibility
only
has
two
states
that's
possible
and
looking
at
the
current
status,
there
are
only
two
states,
but
I
was
thinking.
If
hypothetically
there
couldn't
be
more
and
it
looks
like
in
the
past,
there
was
another
one
pre-render
that
was
removed,
but,
given
that
there
used
to
be
more
I
could
foresee
potentially
there
being
more
in
the
future.
G
C
A
A
There
are
some
things
to
wrap
up
with
page
visibility.
Basically,
there
is
one
meaningful
issue
that
is
not
just
to
remove
a
single
test
that
I
should
be
doing
it
having
gone
through
it
yeah
one
meaningful
issue
which
I
think
requires
some
HTML
integration,
and
then
we
can,
you
know,
call
facials
a
believe
done
and
move
on
to
l-3
I.
A
C
A
B
B
It's
going
to
be
a
virtual
conference
this
year
and
there's
some
work
that
we
were
thinking
about,
doing
ahead
of
time,
figuring
out
what
people's
availability
might
be
for
virtually
meeting
at
some,
not
very
horrible,
but
somewhat
probably
inconvenient
times
for
everybody,
hopefully
not
too
much
for
the
people
on
this
call
and
some
other
things
that
we
wanted
to
kind
of
work
on
ahead
of
time.
You
know
I
noticed
that
you
had
created
a
forum
to
in
the
meeting
minutes.
Did
you
want
to
kind
of
leave
them.
A
Like
one
thing
we
need
to
decide
is
on
what
week
we
actually
want
the
tea
pack
for
this
working
route
to
happen,
because,
yes,
there
are
like
there's
the
official
date
for
October
26.
That
is
essentially
this
week
is
replacing
Wednesday
on
physical
defects.
So
it's
a
cross
functional
cross
working
group
day
of
sessions
where
we,
basically
people,
submit
proposals
for
sessions
ahead
of
time,
and
then
everyone
can
join
those
sessions
according
to
what's
interesting.
A
For
them,
it
is
typically
an
like
an
unconference
style
sessions
and
where
we
can
discuss
many
different
things
related
to
web
standards
the
week
for
this
working
groups.
Feedback
meetings
should
not
be
that
week
because
we
want
people
to
be
available
for
those
cross-functional
sessions.
So
we
need
to
determine
on
what
week
we
actually
want
to
meet
up,
and
so
the
forum
essentially
contains
three
questions
or
yeah.
A
So
one
is
on
what
week
we
can
have
this
working
groups
feedback
and
what
week
works
for
you
I,
don't
know
about
you,
but
my
travel
schedule
has
cleared
up
this
year.
So
it's
easier
to
schedule
things
than
it
previously
was,
but
but
still
people
may
have
previous
engagements,
so
yeah
I
want
to
gather
that
information
and
find
a
week.
That
is,
you
know,
convenient
for
most
people
or
as
many
people
as
possible,
and
then
we
were
talking
about
maybe
having
hackathon
days
before
t-back.
A
So
this
is
something
we
did
at
a
few
of
our
previous
face-to-face
meetings,
so
attached
a
hackathon
day
to
them.
Since
this
year
we
won't
have
a
physical
gathering.
We
were
thinking,
they
do,
can
have
a
few
online
sessions
of
half
days
just
working
through
issues
and
trying
to
reduce
the
number
of
open
issues.
There
are
multiple
specs
with
a
few
lingering
bits
and
pieces
that
need
to
be
nailed
down
so
yeah.
A
It
would
be
great
if
folks
would
join
that
and
basically
the
question
is:
will
you
participate
or
not,
and
then
we
can
figure
out
like
among
the
people
that
participate,
you
can
figure
out
an
ideal
date
for
that
I
think
like
from
my
perspective,
it
can
be
like
either
like
somewhere
in
August
or
September,
but
yeah.
We
can
work
through
the
ideal
date
for
that
once
we
know
the
group
of
people
that
will
actually
that
are
actually
planning
to
join
yes
to
eat
your.
You
want
to
say
something
or
oh,
no,.
H
H
We
are
going
to
also
open
next
week
for
shares
where
they're
going
to
be
able
to
put
their
target
date
for
a
face-to-face
meeting
to
help
them
avoid
overlapping
with
other
working
groups
and
also,
if
we
want
to
have
any
joint
meeting
with
all
the
working
groups.
To
also
helps
that
organization.
As
well.
A
A
A
A
A
And
then,
when
it
comes
to
the
meetings
themselves,
so
we
were
thinking
of
a
format
of
maybe
four
meetings
of
three
hours,
each
with
a
break
in
between,
because
virtual
meetings
are
harder
to
like
we
can
do
full
days
because
of
time
zones
and
because
people
will
fall
asleep
halfway
through.
So
this.
That
makes
sense
that
the
format
of
four
meetings
in
a
single
week
scare
you
or
is
like
any
any
strong
objections
to
that
format
and.
B
A
H
A
A
H
The
shop
news
before
we
move
on
to
another
topic:
we
started
the
issue
view
of
process
2020
and
the
patent
policy
2020
today,
so
that
triggers
a
six
week
review
period
for
the
Advisory
Committee
to
provide
comments
on
those
documents.
Even
if
everything
was
on
track,
those
documents
would
become
effective
as
early
as
September
15,
which
means
that
we've
been
putting
our
chattering
on
hold
to
wait
for
this
process
to
become
effective.
But
this
is
actually
on
the
way.
H
G
H
Available
it
will
become
effective,
so
you
will
be
automatically
switched
to
that
process
on
September
15.
We
are
not
going
to
be
swift,
romantically
to
the
pantora
cities
or
ever
so
so
if
we
want
it
to
adopt
a
new
Berlin
policy,
you
know
we
can
either
be
part
of
the
batch
or
we
can
actually
do
you
know
expedite
or
we
chattering
by
by
linking
to
this
new
patent
policy
sooner
yeah.
A
Okay,
yeah,
in
any
case,
the
current
Charter
takes
us
til
the
end
of
the
year
correct.
That's
me
we'll
need
to
renew
before
that,
but
that
yeah,
so
I
think
we
can
safely
say
that
we'll,
like
one
of
the
sessions
during
t-back,
will
be.
Let's
talk
about
the
new
charter
and
what
it
should
and
shouldn't
contain.
E
E
B
B
Observable
API
I
can't
remember
exactly
the
details
and
there
has
been
some
discussions
around
other
potential
data
that
could
be
included
in
the
crash
report,
but
it's
highly
privacy,
sensitive,
so
I
think
there's
a
lot
of
yeah,
not
a
lot
of
information
that
is
proposed
to
be
included.
I
can
try
to
share
the.
E
D
A
A
We
will
be
sending
link
yeah,
so
it
varies
in
some
cases.
Crashes.
Can
you
know
if
it's
a
process
crash
it
can
reveal
cross
like
like
if
the
browser
process
has
know
if
the
browser
processes
crash,
you
won't
get
a
report.
Oh
that's
easy,
but
if
it's
a
render
it
crash,
some
renders
can
contain
cross
origin
information
in
some
architectures.
So
even
if
the
site
said
it's
okay,
no
don't
worry.
Give
me
that
information.
It
may
not
be
the
site's
information
to
share
that
you
can
share
so.
A
B
E
E
E
D
E
E
Some
performance
element
timing,
entry,
and
that
is
because
I
believe
there
are
scenarios
that
most
of
the
drawing
or
the
most
of
the
expensive
drawing
is
done,
off-screen
or
placed
when
it
the
canvas,
is
hidden
or
not
even
added
to
the
Dom,
and
it
would
be
really
useful
metric
to
actually
measure.
When
was
the
canvas
visible
to
the
user
after
it
has
been
added
or
made
visible.
C
C
E
You
can
always
add
an
onload
event
handler
for
it
right
or
you
could
maybe
register
a
requestanimationframe
and
track
the
time
on
the
next
time
frame,
but
that's
not
as
accurate
right,
because
you
could
have
done
it
for
image
as
well,
and
we
already
know
that's
not
as
reliable
and
robust
as
the
performance
element,
timing,
api's
and
there's
a
reason
they
were
added.
I
guess
it
was
also
simplicity
and
accuracy
of
the
measurements.
So
it's
but
I
believe
it.
C
A
E
The
the
most
of
the
canvas
interface
is
synchronous,
so
you,
just
after
you
complete
drawing
on
the
canvas
what
happens?
It's
all
translated
to,
at
least
in
chrome,
to
GPU
instructions,
and
then
it's
get
offloaded
to
the
GPU
render
and
and
so
on,
but
at
some
point
it
gets
presented
to
the
user
after
the
paint
occurs.
E
A
A
E
A
E
E
E
B
C
C
So
as
far
as
I
see
it,
the
options
are
a
use,
something
like
now,
which
is
also
Network
reporting,
but
defines
caching,
as
the
browser
will
do
the
caching
somehow
so
pretty
hand-wavy
right
and
so
yeah
I'm
actually
not
sure
how
chrome
handles
nail,
caching,
but
certainly
not
specified
very
well
as
far
as
I
could
tell
so.
That's
one
option
and
the
other
option
is
to
try
to
get
the
current
origin
policy
implementation
and
see
if
well
my
implementation
specification
and
see
if
we
can.
C
B
C
A
Arbitrary
headers
enabling
origin
wide
things,
so
this
is
criticism
that
we
get
for
now.
It's
also
criticism
that
we
have
four
client
hints
where
there's
also
like
well-defined
caching,
but
still
caching,
that's
triggered
from
a
header
on
a
resource,
and
they
consider
origin
policy
to
be
a
more
authoritative
way
to
have.
You
know,
centralized
control
for
what
the
origin
should
and
shouldn't
be
doing.
At
the
same
time,.
A
A
A
Yeah,
otherwise,
with
one
minute
left
before
I
one
have
to
bail
for
the
next
call.
I
will
be
on
vacation
for
the
most
part
of
July.
So
two
weeks
from
now,
and
potentially
also
with
me
yeah
so
two
weeks
from
now
I'm
on
vacation
and
then
yeah
I'm
I'm
back
for
on
July
30th
I
will
be
back.
Is
there
or
a
bit
before
that?
But
that's
the
time
for
it
to
come.
After
that.