►
From YouTube: WebPerfWG TPAC meetings 2022 09 16 - Task Attribution
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
So
for
this
last
session
of
CPAP
basically
wanted
to
talk
a
bit
about
task
attribution,
which
is
a
feature
an
underlying
like
some
infrastructure
I
built
for
chromia,
which
I
think
can
open
up
interesting
possibilities
and
I'd
love
to
so
essentially
in
the
last
I,
don't
know
how
many
years
there
has
been
a
common
theme
whenever
I
asked
people,
why
can't
we
have
things
so
for
staff,
navigation,
heuristics
and
you
know
being
able
to
keep
track
of
the
fact
that
the
user
clicked
on
something
and
that
triggered
some
other
activity
on
the
page
that
the
answer
is
like
we
can
do
that.
A
We
can't
do
long
task
attribution.
We
can't
know
what
the
resource
timing
initiator
is.
We
can
do
cross-task
prioritization,
because
we
don't
really
know
what
JavaScript
task
was
the
initiator
of
this
other
JavaScript
task,
and
if
we
wanted
to
do
that,
we'll
have
to
look
into
the
JavaScript
stack,
and
that
is
expensive,
and
we
cannot
really
do
that
in
run
time.
A
Definitely
not
for
performance
monitoring
purposes,
so
I,
essentially,
basically,
I
really
wanted
to
have
self-detication
heuristics
and
all
these
other
things
like
all
these
other
things
also
seemed
useful.
So
I
went
ahead
and
built
a
way
for
us
to
know
when
a
JavaScript
desk
initiated
another
generally.
A
The
use
cases
I
have
in
mind
for
this
theoretical
use
cases
where
this
can
be
useful,
so
self
navigation,
heuristics
I,
want
to
like
right
now,
Chrome
as
a
browser
collects
data
about
how
fast
applications
are
or
aren't,
but
we
know
nothing
about
what
happens
inside
of
sbas
and
single
page
app.
Community
has
been
claiming
for
the
last
10
years
that
sbas
are
so
much
faster
and
other
folks
have
been
claiming
that
sbas
are
so
much
slower
and
no
one
knows,
and
we
need
to
know
so.
A
I'm
like
in
the
process
of
trailing
out,
stop
Navigation
heuristics
that
are
based
on
task
attribution.
That
will
essentially
tell
us
when
a
self
navigation
has
happened
and
then
we'll
be
able
to
re-fire,
LCP
and
FCP
after
that,
heuristic
is
triggered
and
start
collecting.
Some
of
that
data,
at
least
initially
like
in
Chrome
I'm,
hope,
like
I'm,
aiming
for
An
Origin
file
that
will
also
enable
run
providers
to
collect
the
same
data,
compare
it
to
their
current
heuristics.
A
Tell
me
where
I'm,
wrong,
Prince
and
repeat
until
we
get
the
thing
right
and
then
hopefully
ship
it
as
a
web
exposed
API
that
that's
the
plan
and
that's
a
use
case.
I
care.
A
lot
about
the
other
use
case
is
long
task.
Attribution
we've
been
hearing
from
developer
like
we
shipped
chromium
shift,
long
tests
four
or
five
years
ago.
A
I
don't
know,
but
we
keep
hearing
from
developers
that
it's
great
to
get
like
to
know
that
the
main
thread
was
busy,
but
it's
very
hard
to
actually
do
something
about
it
without
field
attribution,
because
it's
very
hard
to
know
which
one
of
your
35
Third
parties
is
the
one
that
actually
ate
up
your
main
thread
and
yeah
and
task
attribution
will
also
enable
long
task
attribution
and
enable
us
to
tie,
at
the
very
least
the
URL,
of
the
main
instigating
JavaScript.
A
Then
resource
timing,
initiator
info
I
think
it
would
be
very
useful
to
help
run
providers
be
able
to
create
dependency,
trees
of
from
resource
timing
and
be
able
to
know
okay.
This
JavaScript
know
that
this
five
other
resources,
which
then
loaded
this
and
basically
be
able
to
tie
back,
attribute
device
that
were
consumed
to
this
initial
JavaScript
or
initial
resource.
A
A
All
kinds
of
things
will
be
possible:
okay,
magic
and
finally
or
I'm,
not
exactly
cross
tasks.
Crosstalk
parameterization
seem
like
also
a
useful,
primitive,
so
right
now
we
can.
We
have
priority
heads,
we
can
deprioritize
a
script,
but
then
everything
that
it
loads
can
be
high
priority
and
it
would
be
good
to
be
able
to
propagate
this
fast
priority
to
to
the
script.
Similarly,
for
post-tasks,
we
can
prioritize
a
tasks
that
we're
posting,
but
we
can't
prioritize
right
now
the
priority
of
the
test
that
it
will
execute.
A
So
if
you
can't
have
a
low
Prairie
task
that
then
cues
out
high
priority
tests-
and
it
would
be
good
to
be
able
to
propagate
this
and,
for
example,
contains
some
tasks
or
contains
some
jobs
yeah
make
sure
that
some
javascript's
always
running
low
priority
or
in
high
priority
and
and
be
able
to
have
that
kind
of
control
across
tasks
and
I
I.
Think.
A
Finally,
we
have
cross-tests,
potentially
privacy
protection,
so
we
could
say
that
certain
scripts,
for
example,
won't
be
able
to
access
high-end
per
vidis
or
we'll
be
able
to
keep
track
of
the
number
of
bits
that
a
certain
script
accumulates
throughout
its
life
with
the
various
tests
that
it
triggers,
even
if
they're
fitting
them
with
set
timeout
or
post
tasks
or
curing
micro
tasks.
We
will
be
able
to
keep
track
of
that
and
provide
that
accountability
to
the
amount
of
bits
that
each
script
collects
and
then
maybe
do
something
about
that.
A
Okay.
So
how
does
task
attribution
work
at
a
high
level?
Pitch
new
task
creates
a
passcode
if
there
are
new
tasks
that
are
created.
While
another
scope
is
active
that
scope,
it's
is
declared
as
if
tasks
parent
and
the
current
task
is
the
descendant,
and
we
have
task
IDs
that
are
passed
to
V8
and
new
promises
that
are
created
while
a
task
is
running
or
also
assigned
with
that
ancestor
ID.
So
they're
not
like
continuations
are
not
new
tasks,
but
they
know
like
they.
A
Basically
all
are
all
part
of
the
same
task.
Id.
This
is
currently
all
I
said,
is
an
implementation
detail.
A
I
would
love
for
it
to
become
eventually
a
web
platform
concept,
not
something
that
is
directly
exposed
to
web
developers.
Most
probably
I
haven't
heard
of
compelling
use
cases
to
the
directly
expose
it,
but
it
seems
like
if
we
wanted
to
build
all
those
things,
we
would
need
to
figure
out
a
way
to
say
all
the
things
I
just
said.
You
know
in
Spec
language,
inspect,
Concepts,
tied
to
the
event
Loop,
most
probably.
A
Yeah
and
so
I
basically
wanted
to
share
that
with
you
all
and
hear
your
thoughts,
and
with
that
can
we
have
recordings,
got
one
sec.
Thank
you.