►
From YouTube: WebPerfWG call 2020 11 19
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
Hi
and
welcome
to
the
web
performance
working
group
today
we
have
an
exciting
exciting.
I
don't
know,
I'm
I'm
excited
no,
I'm
norm
rosenthal.
We
actually
have
several
neurons
norma
rosenthal
will
present
his
learnings
from
optimizing
a
preview
pop-up
at
wikipedia
and
then,
if
we'll
have
time
after
that,
we'll
talk
a
bit
about
a
few
issues
that
we
have.
That
has
been
piling
up
generally
for
the
next.
The
next
call
is
scheduled
for
december
3rd,
the
usual
time,
not
this
time.
A
Apologies
for
the
mix
up
with
scheduling
but
yeah
and
yeah.
If
there
is
a
significant
scheduling
problem
related
to
that,
please
let
me
know
and
we'll
try
to
figure
out
an
alternative
time
and
with
that
out
the
way.
No
do
you
want
to
get
started.
Yes,.
B
C
B
And
web
development,
I
I
contribute
to
a
webkit
and
chromium
and
I
also
do
actual
web
work.
So
in
this
case
the
project
was
to
take
a
very
popular
use,
a
web
item
which
is
the
preview
pop-ups.
B
B
Actually
the
mic
is
the
camera
all
right
and
yeah.
Let
me
know
if
there's
interruptions
again
so.
B
B
In
order
to
make
this
easier
in
the
future,
they
might
be
planning
more
pop-up
like
features
and
every
time
there's
a
new
pop-up,
not
just
for
wikipedia
experience.
That's
also
clicks
in
at
other
companies.
Every
time
there
are
those
slightly
off
ui
concepts
like
popovers,
there's
lots.
B
Problems,
there's
lots
of
ways
to
authorize
which
are
non-standard
and
and
it
ends
up
being
not
easy
yeah.
So
so
it
was
more
like
a
research
into
what
the
web
platform
could
provide
to
make
this
easier
in
the
future
and
to
make
it
easier
for
the
developers
to
make
fast,
ui
in
terms
of
jaggedness,
etc.
B
B
E
B
If
you
try
using
the
pop-up
for
real,
you
will
see
that
when
you
hover
a
link,
the
cursor
appears
close
to
where
you
hovered
close
to
the
mouse.
It's
not
just
centered,
with
the
link
it
actually.
B
E
B
B
B
Something
like
this
and
I
went
into
this
with
existing
code,
but
I
analyzed
it.
I
pretty
much
tried
to
rewrite
it
several
times
so
that
I
would
run
into
the
same
issues.
So
there
are
issues.
The
main
issue
is
with
the
placement
and
stacking
is
the
other
one
is
clipping
and
the
overriding
before
behavior
and
miscellaneous
I'll
get
to
all
of
them.
B
So
there's
a
big
dilemma
with
a
stacking
versus
placement.
This
is
actually
the
biggest
issue
I
found
here.
One
option
was
to
put
the
pop-up
as
a
child
of
the
link
like.
If
you
see
here,
there's
the
cutout
link
anteaters
and
then
the
the
pop-up
is
above
it
now.
The
way
I
would
think
to
do
this
is
that
ant
eaters
would
become
position
relative
and
then
the
pop-up
would
have
something
like
bottom
zero.
B
The
whole
general
text
in
wikipedia
has
a
an
isolated
stacking
context
as
a
v-index
for
practical
experts
who
could
have
had
the
isolate
yes
uses
property.
B
B
It
would
mean
that
it
would
have
to
call
something
like
get
client
wrecked
or
opposite
left.
It
doesn't
really
matter
and
on
the
link
of
anteaters
and
then
use
positioning
from
the
page
on
a
totally
separate
element
and
which
is
the
pop-up
now.
What
I
found
is
that,
well
it's
not
exactly
something
I
found
something
that's
known
in
the
internet
for
many
years
is
layout
thrashing,
but
what
I
found
is
that
layout
thrashing
is
almost
unavoidable.
B
B
So,
by
by
the
time
you
call
get
found
in
client
rex
or
whatever
it
is,
even
if
it's
very
very
early,
you
could
already
be
layout
thrashing
and
when
I
was
a
profiling,
this
particular
pop-up
preview
in
wikipedia
that's
what
I
saw
a
whole
bunch
of
layout
crashing
being
the
main
source
of
the
jagginess
and
and
even
by
batching,
all
of
the
measurements
together.
There's
still
some
of
it,
because
it's
not
entirely
avoidable.
B
And
now
I
looked
for
solutions
in
the
internet,
I
found
something
called
the
top
layer
research
which
I
started
joining
into
that
conversation.
I
think
it's
a
very
important
conversation
yeah
about
finding
a
way
to
separate
the
stacking
context
from
the
regular
dom
tree
in
certain
situations.
It's
kind
of
complex
to
do
that,
but
I
think
it's
an
important
conversation
to
push
through
with.
B
B
It
has
a
lot
of
synchronous,
get-bounding,
client-track
style
things
inside,
which
would
mean
in
many
cases
it
would
lay
out
wraps,
and
so
there
is
the
top
layer.
Research
is
another
standard
proposal
which
I
don't
like
as
much,
but
it
is
simpler,
which
is
the
relative
two
position
which
allows
positioning
an
element
relative
to
another
out
of
three
element.
B
And
now,
as
I
said,
the
slide
proper
js
is
used
by
three
percent
of
websites.
That's
because
it's
a
part
of
bootstrap,
a
positioning
popups
is
extremely
popular
and
something
that
a
lot
of
people
either
use
popper
or
tether,
or
one
of
those
libraries
or
roll
their
own
like
with
wikipedia
and
the
is
the
cause
of
a
lot
of
pitfalls
with
performance.
B
E
B
And
what
I
found
is
that
offset
x
and
offset
y
for
that
purpose
are
bored.
They
have
old,
bugs
firefox
does
completely
different
things
from
a
chromium
and
safari.
Firefox
is
actually
pretty
spec
compliant,
and
both
people
in
chromium
and
in
webkit
were
reluctant
to
move
forward
with,
say
fixing
it,
because
it's
not
clear
whether
it's
gonna,
whether
people
are
relying
on
the
bugs.
B
The
final
thing
is
that
we
can't
really
even
tell
with
use
counter
because
old
versions
of
jquery,
despite
the
use
counter
up
to
fifty
percent,
like
fifty
percent
of
websites
use
offset
ticks,
but
only
because
they
use
an
old
version
of
jquery.
B
B
So
one
big
thing:
I
learned
the
only
way
to
totally
remove
thrashing
it's
not
just
level
threshing.
It's
also
style
tracking
is
style
calculation
that
can
happen
without
paint
just
because
you
need
to
get
some
measurements.
A
B
B
B
B
B
It
doesn't
tell
you
where
an
element
is
no
yeah,
so
I
believe
this
would
might
be
a
good
intermediary
intermediate
solution
to
have
a
position
observer.
B
B
The
old
or
the
existing
wikipedia
code
uses
svg,
which
makes
a
lot
of
sense
since
I
11
still
exists
and
wikipedia
still
supports
it,
and,
and
it's
even
though
it's
2020,
that's
still
the
situation,
but
when
I
tried
to
replace
it
with
the
more
modern
clip
path,
I
found
that
it's,
it's
very
subtle,
but
this
balloon
has
a
rounded
corners
of
two
of
two
pixel
radius
and
with
clip
paths,
polygons
don't
have
rounded
corners
and
pathets
use
only
absolute
coordinates,
so
you
cannot
make
a
responsive
balloon
with
flip
path
and
now
for
the
purpose
of
this
project,
I
was
able
to
mimic
the
same
behavior
with
polygon,
just
because
it's
a
two
right.
B
B
So
what
was
happening
in
the
wikipedia
code
is
that
when
hover
happens
and
javascript
removes
the
title
attribute
and
reinstates
it
when
the
pop-up
is
hidden
now
that
code
also
appeared
in
a
profiler,
you
changed
it
on
there's
this
new
style
calculation.
Maybe
something
relies
on
the
title
archive,
even
just
for
calculation
that
don't
end
up
doing
any
layout
or
anything
it
ends
up
being
pretty
expensive.
B
I
checked
to
see
how
much
the
web
is
doing
this
in
practice,
and
I
found
that
bootstrap
tooltips.
Do
this
you
a
hover
over
a
link
to
to
show
a
pop-up,
and
it
internally
removes.
B
The
problem
is
that
prevent
default
is
not
standard,
prevent
default,
the
standard
for
prevent
default.
This
says
the
browser
should
prevent
the
default
behavior.
It
doesn't
say
what
the
default
behavior
is,
so
it's
up
to
every
browser
so
pursuing.
This
is
something
I'm
planning
to
do,
but
it's
not
standard
for
to
just
say
condensing,
webkit
folks
and
forming
folks
and
powerful
folks.
If
this
is
the
right
thing,
maybe.
C
B
A
So
is
any
browser
doing
the
right
thing,
or
what
do
you
think
is
the
right
thing
on
prevent
default.
B
A
B
Yeah
and
one
could
think
that,
because
the
websites
built
are
built
with
the
prevent
default
behavior
in
mind,
like
this
exact
line
code
that
they
put
in
link-
and
so
maybe
it
should
be
standard,
otherwise
it's
kind
of
hard.
Otherwise
there
is
code
that
just
relies
on
how
browsers
happen
to
work,
hey,
which
area.
F
E
B
G
If
you
continue,
if
you
continue
your
thread
of
that
they're
in
augmentation,
they're
pulling
from
content
that
the
link
is
ultimately
referencing,
so
it's
not
like
the
content
that
is
inside
of
them
is
not
accessible.
It's
it's
a
navigation
away
versus
a
tab.
B
Exactly
yeah
it's
so
I
think
titles
in
general.
I
think
that
there
are
opinions
to
discourage
them.
I
think
media
wiki
has
a
platform.
I
don't
speak
for
them
here,
because
I'm
I'm
just
a
analyzing.
This
allows
this
augmentation
like
when
the
preview
pop
the
previous
preview
parts
are
not
available.
Then
title
should
work.
B
This
tooltip
and
I
think
it's
okay.
As
long
as
the
content
is
accessible,
it's
okay
to
have
a
title
and
it
does
something.
But
but
the
analysis
here
is
not
opinionated
about
this.
B
B
So
the
first
thing
I
was
doing
is
perform
all
the
measurements
right
in
the
beginning.
The
first
thing
that
happens
when
the
event
comes,
I
measure
whatever
I
need
to
measure
right
away,
that
prevents
90
percent
of
layout
and
style
threshing
and
not.
It
leaves
them
to
edge
cases
which
is
nice
still.
I
need
to
use,
get
client
rates.
Get
client
tracks
does
way
too
many
things
and
it
has
to
calculate
all
the
transforms
it
has
to
find
all
the
little
links.
B
All
I
really
care
about
is
is
the
hit
rect
of
my
mouse
event.
It's
really
the
only
thing
I
care
about,
and
the
browser
already
knows
it,
but
still
I
have
to
call
get
client
tracks
which
can
take
can
take
milliseconds
for
over
a
course.
A
few
calls
can
take
milliseconds.
It's
not
a
very
scalable
function;
okay,
based
on
a
all.
B
With
semi-rounded
corners,
which
is
works
for
this
ui,
but
this
is
the
wikipedia
designer
would
want
to
do
a
rounder
rectangle.
They
wouldn't
be
able
to
use
a
solution,
and
currently,
when
default,
the
popups
extension
initializes.
It
just
removes
all
the
title
attributes
so
that
they
don't
have
to
be
removed
on
hover.
B
B
I
call
it
stacking
versus
placement
here,
there's
a
link
in
one
of
the
other
slides.
I
want
that
conversation
to
to
wake
up
about
popovers
er.
Second,
one
fixing
office
addiction
office.
Why
either
with
some
backwards
compatibility
flag
or
just
by
breaking
the
buggy
behavior
in
a
blinking
webkit.
B
B
The
fourth
one
is
a
prevent
default
for
mouse
over,
and
my
main
takeaway
to
this
group,
maybe
is
that
performance
is
not
just
about
monitoring
or
about
the
meta
and
http,
and
the
things
like
that
sometimes
performance
is
about
fixing
bugs
and
allowing
new
forms
of
expression
like
sometimes
a
performance
bottleneck
is
just
a
bug
workaround,
which
could
be
a
spec
bug.
It
could
be
an
implementation
bug
or
a
workaround
for
a
form
of
expression.
That's
missing
like
trying
to
position
a
pop-up,
because
there
is
no
standard
form
of
expression
for
pop-ups.
A
A
Thank
you.
One
final
question
about
the
offset
x
and
y
compatibility
issues,
so
you're
saying
that
currently
we
have
different
behavior
for
firefox
versus
webkit
and
chromium.
That's
correct
and
do
we
have
breakage
recorded
by
firefox
related
to
that
different
behavior,
or
is
that
just
from.
B
There
are
people
complaining
about
it.
There
are
all
the
bugs
open
about
all
this.
Okay
and
mainly
people,
just
not
using
obstetrics.
B
B
C
B
F
B
And
they
don't
consider
a
change
images,
loading
and
font
floating,
so
they.
D
My
main
comment
would
be
that
for
this
particular
type
of
feature,
it
shows
how
difficult
it
is
to
do
the
right
thing
and
how
easy
it
is
to
do
the
bad
thing
as
in
people
using
these
expensive
javascript
functions
have
no
idea
what
they're
doing,
and
we
constantly
have
to
tell
people
who
write
code
at
our
organization.
D
We
do
performance
reviews,
please
stop
using
that,
but
then
you
run
into
situations
like
this,
where,
even
if
you
say
please
stop
using
that,
they
have
no
other
choice
and
it
shows
how
things
are
in
terms
of
ergonomics,
really
the
opposite
of
what
they
should
be
like.
The
easy
functions
to
use
are
the
ones
you
want
to
avoid
and
and
doing
it
properly
is
just
like
super
complicated,
with
lots
of
very
advanced
css
and
javascript
knowledge
required
to
do
it
properly.
D
A
Or
alternatively,
like
what
I
can
imagine
us
doing
us
as
the
web
platform
at
large,
but
we
could,
assuming
we
fixed
those
things.
People
will
still
not
necessarily
adopt
the
fix,
but
then
we
could
pr
those
fixes
to
those
libraries
and
make
them
significantly
more
performant,
potentially.
B
Okay,
that
it,
it
actually
said,
don't
call
offset
x
to
avoid
layout
right
and
start
crashing.
It's
a.
I
think
people
want
to
do
the
right
thing
like
if
we
we
should
enable.
A
Yeah
cool,
so
I
guess
a
follow-up
question
would
be
like
first
of
all,
like
on
this
amazing
work,
you
filed
bugs
on
everything,
and
we
should
tell
people
like
most
or
all
of
those
things
are
not
things
that
this
working
group
can
solve,
but
we
can,
probably
you
know,
yell
at
people
until
they
would.
But
beyond
that,
like
what
other
low-hanging
fruit
is
there
out
there
like?
How
can
we
systematically
weed
out
those
pieces
that
are
causing
all
those
rendering
performance
issues
or
other
you
know,
performances?
A
B
B
Alerting
about
layer
thrashing
and
about
synchronous
measurements
could
help
er,
like
maybe
even
a
performance
entry
for
that
like
whenever
there
is
a
style
calculation
or
a
layout
that
didn't
come
from
that
didn't
come
from
painting
that
this
is
something
we.
This
should
always
be
a
yellow
flag
and
it
should
go
both
to
the
other
working
groups
and
to
the
web
developer.
A
B
Like
it's
like
a
smoke,
I
just
found
it
to
be
a
smoking
gun
in
many
situations.
So
you
just
watch
the
detective
movie,
I'm
using
like
that.
Yeah.
A
The
part
that
I'm
having
trouble
with
is
yeah
whether
like
a
different
asynchronous
api,
would
be
a
hundred
percent
replacement.
Or
will
there
still
be
cases
where.
B
And
it's
an
api
that
might
get
called
a
lot.
That's
the
thing
he
trying
to
synchronize
position
with
javascript
would
it
would
become
less
expensive
than
it
is,
but
it
wouldn't
be
a
solution.
The
solution
would
be
that
you
position
the
pop-up
that
is
placed
relatively
to
another
element
and
is
stacked
with
the
whole
page.
B
I
B
A
Yeah,
so
I
think
the
right
order
here
would
be,
you
know,
create
a
reasonable
alternative,
then
ask
them
to
move
to
it
or
pr
a
solution
and
then,
if
they
somehow
don't
or
if
people
are
using
the
older
version,
that's
not
great.
That
is
where
you
point
fingers.
H
A
Cool
so
thank
you
so
much
and
yeah
and
let's
then
we
can
talk
about
some
of
the
issues
that
we
have
on
the
agenda.
In
the
15
minutes
we
have
left
the
first
one
we
have
is
resource
or
navigation
timing
issue
number
124.
A
A
So
it
turns
out
that
currently
they
are
all
returning
zero
when
a
service
worker
is
involved,
which
is
probably
not
what
we
want
them
to
and
then
there's
a
discussion
on
the
issue
regarding
what
they
should
be
returning
and
how
we
can
specify
that,
and
I
think,
like
on
the
issue,
we
mostly
like
the
issue
coalesced
on
them,
returning
values
that
are
as
close
as
possible
to
the
actual
transferred
sizes,
because
responses,
currently,
the
response
body
has
in
in
fetch,
has
equivalents
to
the
encoded
and
decoded
body
sizes.
A
It
doesn't
have
an
equivalent
to
the
transfer
size
or
the
header
sizes,
but
we
could
probably
specify
that
and
then,
when
a
service
worker
is
putting
together,
resources
from
various
response
elements
or
from
like
typically
from
a
single
response,
object,
not
element
that
would
we
could
take
those
values
and
report
them,
and
I
guess
the
question
to
the
group
is:
does
this
make
sense
for
everyone?
Anyone
has
any
objections
or
alternative
ideas.
As
to
what
we
want
to
report
in
those.
A
A
Okay,
yeah,
so,
from
my
perspective,
this,
like
the
proposal
on
the
issue,
makes
general
sense
and
the
main
thing
that
scares
me
here
is
the
fetch
integration
pieces,
because,
right
now
we
are
not
like.
We
will
need
resource
timing
to
operate
on
a
response
object
in
order
to
make
use
of
those.
A
Response
parameters,
but
at
the
same
time
we
could
maybe
you
know,
make
that
piece
work
like
yeah.
Hopefully
it
won't
have
to
wait
for
full
fetch
integration,
but.
H
Yeah,
I
do
you
have
a
sense,
what
browser's
doing
what
processes
are
doing
right
now,
I
think,
or
are
we
just
the
only
ones
I
just
from
the
only
one
implementing
the
sizes,
because
apple
apple
isn't
right.
A
Apple,
isn't
so
in
webkit
sizes
is
not
implemented.
I
believe
that
firefox
does
have
the
transfer
size
implemented.
H
A
I
suspect
this
is
how
it's
specified
so
not,
although
I
didn't
clearly
look
but.
A
With
that
statement,
yeah,
and
it
seems
to
me
like
that,
like
properly
specifying
it
will
require
at
least
some
patch
integration,
if
not
full
fetch
integration.
A
Yeah
yeah,
but
I
don't
hear
anyone
loudly
objecting
to
transfer
size
and
encoded
size
reporting
the
results
from
the
fetch
response.
So
I
think
we
have
general
agreement
on
what
we
should
be
doing
and
now
it's
just
a
question
of
how
can
we
specify
it.
C
Can
we
fix
this
from
the
spec
point
of
view
to
align
with
current
fetch
concepts
without
fixing
the
whole
spec
to
align
with
touch.
A
E
A
Sam
whaler
from
w3c
team
pointed
out
rightfully
that
the
document
claims
to
be
normative,
but
it's
mostly
just
documentation
and
we
should
probably
fix
that
and
then
shushen
commented
that
the
document
is
outdated
and
should
be
obsoleted
and
recommending
nick's
blog
post
instead
about.
C
A
A
I
haven't
looked
at
it
lately,
I
haven't
looked
lately
either.
I
think
it
might
be
worthwhile
to
like
for
someone
to
go
over
it
and
determine
that.
I
wonder
if
anyone
feels
particularly
eager
to
take
on
that
task.
C
A
If
it's
indeed
out
of
date,
then
we
should
probably
obsolete
it,
because
we
don't
necessarily.
A
A
Venue,
rather
than
as
a
w3c
note,.
A
On
it,
I
don't
know
but
yeah
from
the
crowd
here.
Has
anyone
used
that
document
when
you
got
started
or
at
some
point
raise
your
hand
if
you
did.
A
C
C
A
A
So
yeah,
maybe
obsoleting,
is
the
right
call
here,
but
I'll
I'll
ask
around
and
see.
If
someone
wants
this
to
to
stick
around
benjamin.
I
believe
that
this
is
like
the
the
original
motivation
for
this
was
comments
from
boris
related
to
confusion.
I
may
be
mixing
things
up,
but.
A
If
you
could
also
ask
around
on
the
mozilla
side
to
see
if
someone
finds
value
in
this
document
staying
you
know
alive.
That
would
be
helpful.
I
Well,
do
I
think
I
think,
certainly
have
my
support
to
deprecate
this
okay.
C
A
So
you're
saying
maybe
2018
marcos
was
right,
possibly
yeah,
okay,
okay,
so
I
think
we
can
settle
on
obsoleting
this
unless
some
evidence
comes
up
that
this
is
required.
A
Cool
that
would
also
probably
mean
that
we
can
remove
it
from
the
charter
from
the
charter
draft.
I
don't
know
corrine.
Will
this
have
timeline
implications
or.
E
A
Okay,
cool,
thank
you
and
yeah.
We
have
one
minute
left,
so
we
can
probably
leave
the
last
one
for
next
time.
Awesome.
So
thanks
all-
and
I
will
see
you
all
in
a
couple
of
weeks.