►
From YouTube: Node js Tracing WG Meeting 2015-09-25
Description
notes at https://github.com/nodejs/tracing-wg/blob/master/wg-meetings/2015-09-25.md
B
Guess
I'll
go
next,
so
I'm
unbraced
barrel,
I
joined
the
working
group
when
it
first
started
by
actually
making
the
previous
meetings.
My
interest
comes
in
tracing
it.
In
my
previous
company
I
wrote
and
a
PM
tool.
They
did
tracing
for
node,
and
so
I
got
it
got
interested
and
and
so
now
I'm
also
no
source
but
still
interested
in
tracing
I.
C
Think
I'm,
you
know
she'll
idea,
no
jazz
team,
no
jazz,
platformer,
netflix
and
in
the
past,
I've
used
right
now,
it's
racing
quite
a
bit
with
djs
a
note
and
that's
what
I
like
so
we're.
Looking
for
something
else.
That's
gonna!
That's
going
to
be
able
to
give
us
that
same
functionality
and
not
so
super
excited
to
be.
D
G
So
few
interest
in
increasing
one
is
one
is
the
evening
music.
Is
that
my
talked
about
the
the
second
one
is
practical
make
time
since
we
got
google,
we
want
to
make
things
easier
for
everybody
from
a
v8
node
integration
interview
from
a
from
a
tracing
perspective,
although
I'm
not
doing
any
of
the
work
directly,
the
chrome
team
is
working
on
moving
tracer
night,
don't
need
to
behave,
and
in
that
form
it
might
become
more
useful
for
note
as
well.
So.
A
G
H
Guess
I
will
last
torsten
also
work
for
notes.
Source
I've
been
doing
some
work
with
also
dtrace
/,
also
trying
to
integrate
this
with
chrome,
dev
tools
and
other
two
links
to
make
this
easier
see
profiling
data
I
guess
my
interest
is
mostly
in
trying
to
make
it
even
easier
for
users
to
profile.
Ideally,
you
would
just
tell
note
to
profile
itself,
but
I,
don't
know
if
that's
possible,
but
we'll
see
and
I
have
also
been
in
the
very
first
meetings
but
have
been
kind
of
liking,
so
gonna
have
to
catch
up.
It's
it.
A
So
the
there
is
an
agenda
kind
of
at
the
google
doc.
That's
pointed
to
at
issue
21,
which
is
for
today's
meaning.
The
the
first
item
I
have
in
there
is
to
set
some
cadence
for
these
meetings.
So
looking
back
on
the
the
history
of
what
I
can
see
so
far
in
the
github
repo,
there
were
two
meetings,
one
on
februari
fifth
and
another
on
the
during
19th,
and
then
that
was
it.
So
it's
been
seven
months
since
we
had
one
of
these
calls
and
seems
like
we
should.
A
We
probably
want
to
talk
more
frequently
than
that
and
so
I'm
kind
of
thinking.
Maybe
we
can
try
to
do
this
once
a
month
if
things
end
up
heating
up
like
we
actually
have
some
some
stuff
people
can
kind
of
play
with
or
something
then
we
might
want
to
go
more
frequent,
but
make
every
month
as
a
target
would
probably
be
a
good
thing
to
do
any
thoughts
from
anyone.
J
K
A
A
L
A
A
A
H
G
A
A
Can
do
that
I
I
was
actually
planning
on
doing
that
before
the
suggestion
to
have
the
call
today
was
posted
in
in
the
issue
and
I
was
playing
with
doodle.
The
the
problem
of
doodle
is
I
think
it
only
lets
you
kind
of
easily
offer
up
three
days
and
and
there's
some
kind
of
yep
that
kind
of
hit
something
else
to
see
more.
A
If
you
have
more
and
then
maybe
more
than
five
or
six
got
to
be
really
unwieldy,
I'm
having
to
do
that,
because
I
have
had
success
with
doodle
in
the
past,
but
I
guess
it
was
maybe
with
a
smaller
time
range,
so
yeah,
let's
go
ahead
and
give
that
a
try,
and
was
that
Ali
that
much
or
not
yes,
okay,
so
I'm
going
to
work
with
you
to
get
that
set
up
and
will
do
that
offline.
Somehow
all
right!
A
So
so
we'll
pick
a
day
in
time
by
a
doodle
for
the
next
meeting
in
two
weeks,
good
on
the
track.
Alright!
So
looking
at
the
agenda
and
I'm,
we
probably
should
have
had
folks
add
their
initials
in
here
or
something
for
the
agenda
items.
Then
the
next
agenda
item
sells
breakdown
components,
pacing
crap,
pulling
dev
tools
out
of
blank
generic
made
of
tracing
into
smaller
parts
and
defined.
A
What
work
actually
needs
to
be
done
to
make
progress,
so
that
sounds
great
as
kind
of
a
newbie
on
these
calls
that
I've
listened
to
the
the
previous
ones.
There
seems
to
be
a
lot
of
different
areas
here,
like
a
sink
rap
like
va
très
other
things
from
even
just
the
issues
is
like
native
window
tracing
and
I.
Don't
have
a
good
feeling
for
how
all
this
stuff
fits
together.
I
do,
but
it's
it's
kind
of
my
view
and
I.
A
D
For
me
that
I
certainly
some
of
these
components
that
I
know
some
about
and
then
that's
all
that
I
don't
know
nothing
about,
and
it's
possible
that
all
of
these
components
sort
of
saw
was
the
same
issue.
So
I
guess
it
would
be
nice
if
those
that
excellent
notes
about
some
of
these
components
could
just
sort
of
wrap
out
wrap
up
what
they
are
about
and
what
the
soul,
yeah
I,
certainly
don't
know
all
of
them.
Yeah.
A
So
maybe
what
we
can
do
there
is
just
go
ahead
and
flesh
out
some
some
places
where
people
can
kind
of
talk
about
some
of
these
components.
Just
so,
we
have
some
stuff
kind
of
written
down
and
not
spread
out
across
the
minutes,
and
videos
will
not
so
why
don't
I
take
a
stab
at
that
and
create
maybe
a
directory
12
documents
to
start
with,
and
you
know
over
time,
maybe
that
moves
somewhere
else
and
then
and
then
we'll
create
a
document
sort
of
per
component
and
so
right
now
I
think
it's
pretty
clear.
A
A
A
Yeah
I
have
an
agenda
item
for
that
one
cuz
that
I
think
that
one
is
kind
of
a
no
op,
but
if,
as
we
get
to
it,
if
it,
if
it
turns
out
to
be
something
interesting,
what's
fun
of
doing
that,
generic
native
tracing
I,
guess
I
think
a
sort
of
platform
specific
tracing
like
the
windows
tracing
work
that
was
done.
It
was
mentioned
in
one
of
the
issues
as
well
as
I
heat,
race
and
LTT
engineer,
and
then
somebody
added
proof,
events,
yeah.
A
A
A
A
Sounds
good
unless
you're
on
mute
talking,
we
can't
hear
you
so
let's
go
ahead
and
plan
on
doing
that.
I
think
one
of
the
things
that
may
naturally
fall
out
of
that.
As
we
see,
some
of
these
components
will
start
to
get
a
sense
for
how
they
actually
put
together
obvious.
One
for
me
is
like
how
say
crap
fits
into
things
like
the
trace,
so
just
kind
of
you
know
solidifying
that
in
my
mind
how
that
works.
I
may
be
a
fall
out
of
this
as
we
start
to
John.
Something
has
stuck
down.
A
A
D
D
Think
you,
if
you
want
a
bit
more
history,
about
eating
bad
Trevor
in
the
beginning,
cradle
and
local
module
called
tracing
this
sort
of
provided
many
of
these
facilities
that
I
listed
as
issues
right
now
and
then
the
sort
of
it
was
decided
that,
since
we
don't
know
what
kind
of
events
is
actually
needed
from
the
user
perspective,
it's
better
just
to
have
a
very,
very
bare
bone,
a
implementation
of
it
and
then
sort
of
have
that
discussion
later
and
provide
a
more
reasonable,
epi
yeah.
But
yeah.
A
A
Pull
some
stuff
out
of
blank
to
make
it
easier
for
node
to
be
debugged
with
the
existing
chrome,
dev
tools
and
I
believe
the
response
was
looking
here
from
my
Yuri.
Was
that
says:
I
don't
think
tracing
integration
will
be
blocked
by
the
relevance
issue
that
came
up
regarding
the
yeah,
it's
new
debugging
facility,
so
I
think
this
is
sort
of
a
no
op.
Hopefully
this
will
be
a
fairly
independent
issue.
I.
A
I
That
the
reason
that
I
brought
up
was
just
like
part
of
the
dev
tools
is
like
try,
some
component
components
and
I
I.
Don't
think
it
should
be
like
blocking
anything
else.
But
but
I
would
like
to
see
us
pain
like
a
lot
more
attention
to
it
and
seeing
if
there's
anywhere,
we
can
help
out
in
making
that
all
work
and
like
that's
just
one
possible
way,
that
we
can
deliver
a
tracing,
a
tool
to
users.
A
Yeah,
like
I,
said
you
know,
I
I
would
certainly
expect
that
and
most
or
some
note,
debugging
tools
would
want
to
take
advantage
of
tracing
so
so
using
forces.
This
particular
work.
That's
going
on
as
a
testing
ground
or
for
anything
that
comes
out
of
this
working
group
sounds
great.
They
would
essentially
be
you,
customers
or
clients
of
whatever
work
comes
out
of
here,
so
in
that
in
that
regard,
absolutely,
and-
and
so
maybe
I'll
just
go
ahead
and
highlight
that
will
be
a
great
customer
for
this
work.
A
This
is
a
kind
of
side
note
and
listening
to
the
previous
hangouts
debugging
actually
came
up
in
at
least
one
of
them
and
I
suspect.
There
is
actually
a
lot
of
overlap
between
people
interested
in
debugging
and
in
this
tracing
work.
Group
I,
don't
think,
there's
a
debug
more
people
or
is
there
any
Billina
know.
A
Right
post-mortem
group,
so
it
I
mean
it
seems
to
me:
I
actually
have
a
very
strong
interest
in
debuggers
and
I
actually
contributed
to
a
WebKit
web
inspector
back
in
the
day,
it
almost
seems
to
me
like
there
should
be
kind
of
a
separate
debug
working
group
that
would
be
looking
at
things
like
to
meet
the
the
VA
debug
API,
the
debug
agent,
that
had
to
get
moved
into
node.
That
sort
of
thing
I'm
not
quite
sure
how
how
we
kind
of
resolve
that
between
the
very
strong
ties
between
tracing
and
and
post-mortem.
A
H
Think
it
also
has
to
do
with
like
what
kind
of
debugging
you're
talking
about
if
you're
talking
about
debugging
in
chrome
devtools.
Definitely
that's
much
less
related,
I!
Think
to
tracing
if
you're
talking
about
debugging
from
let's
say,
GD,
b
or
ldb,
and
trying
to
resolve
symbols
that
has
probably
more
of
a
novela
with
tracing.
A
Right
I
mean
that's
a
good
point.
You
know,
ask
people,
you
know
what
their
definition
of
debugging
is
and
you'll
get
to
answer
for
everybody,
if
not
more,
certainly
from
you
torsten,
so
yeah
there's
also
I'm
yeah.
Normally,
when
I
think
debugging
actually
took
v8
debug
API,
that's
what
I'm
kind
of
thinking
with
regard
in
regards
to
node
but
person.
You
are
obviously
have
a
strong
interest
in
much
lower
level
debugging.
That
needs
to
be
factored
in
to
that
in
small.
So
that's
an
interesting
topic,
a
totally
independent,
but
maybe
I
can.
A
All
right,
where
are
we
getting
close,
getting
close?
Alright,
the
last
agenda
I
item
I
have
in
here
on
watching
people
live
at
the
dock
right
now,
Oh
someone's
thinking
minutes.
Thank
you.
I'm
kind
of
writing
minutes
down.
I
should
have
mentioned
I'm
kind
of
writing.
Some
notes
down
plan
on
going
back
through
the
movie,
assuming
it's
actually
being
reported,
and
we
can.
We
can
make
more
elaborate
notes
from
that
as
long
all
right.
A
L
Ali
I
can
say
a
few
things
are
if
you
want
to
endemic
review,
go
ahead,
so
the
big
is
it
rajee's.
Has
it
really
been
seven
months?
This
is
how
we
get
old,
so
we've
been
focusing
on
sampling
data
and
getting
that
into
our
trace
pipeline.
So
they're
two
sets
of
probes
at
least
one
is
about
getting
v8
talking
back
and
forth
to
the
embedder
in
our
case,
chrome
telling
us
enough,
so
that
we
can
recreate
the
devtools
a
CP
profile
view
on
top
of
the
trace
infrastructure.
L
L
So
you
see
a
whole
lot
of
stream
of
that
and
then,
whenever
v8
takes
a
sample,
a
stack
sampler,
the
embedder
takes
a
stack
sample.
We
record
the
roughly
the
stack
frames
into
the
trace
in
a
kind
of
space
efficient
way,
and
so
it
you
end
up
with
is
a
file
of
samples
and
time
stamps,
and
then
you
go
and
you
look
up
that
turn,
that
into
real
function,
names
and
stuff,
and
it's
good
this
networks
and
it
works
in
our
chrome
tracing
and
it
works
in
devtools
and
I.
L
C
L
Developer
tools,
I
I,
think
so
there
there's
some
like
you'd,
have
to
read
how
the
there's
a
thing
called
the
v8
sampler
in
our
content,
folder
that
you'd
have
to
duplicate
in
the
node,
so
that
the
the
sampling
glue
is
outside
of
v8
and
in
Chrome,
but
it's
doobleh,
keable
and
probably
pretty
stable
or
hackable
right,
and
if
you
can
get
that
in
the
note
or
copy
the
way
it
works,
then
you
two
can
have
these
trace
file
like
conceptually.
Once
you
have.
This
trace
file
of
the
trace
files
and
I
can
say.
C
L
C
J
G
Yes,
so
you
know
so
all
of
this
like
getting
profiles,
programmatically
works
today,
it's
a
different
API
at
the
moment.
That's
correct,
yeah
and
what
not
is
proposing
if
you
can
be
christened
into
v8
and
then
get
know
to
start
using
it,
then
we
get
one
pipe
with
all
types
of
stuff
can
be
coming
out
of
it,
recruiting
tracing
all
the
trace
events
when
and
also
profiling,
and
what
I
will
check?
G
L
You
know
headless
or
GUI
analyses
like
show
me
the
samples
you
know
the
jas
costs
only
within
this
time
range
right
are
grouped
by
these
time.
Ranges
right,
because
you
have
one
file
with
you
know:
different
kinds
of
time
ranges
in
it,
and
then
you
have
samples
all
in
one
damn
file.
So
then
you
can
do
cross
referencing.
So
that's
why
we
merged
it
into
one
stream.
In
addition
to
just
the
simplified
plumbing.
L
H
G
G
That
is
part
of
the
that
is
placed
into
the
traces
that
v8
puts
out
today.
So
if
you
look
at
the
isolate,
we've
got
log
and
if
you're
running
with
the
CPU
profiler,
those
code
ranges
are
annotated
in
there
as
well
three.
So
so
so
this
a
few
ways
you
can
get
at
the
same
data
you
can
get
the
per
file
using
the
the
prof
events
support
in
v8
or
you
can
get
the
the
trace,
which
has
all
these
good
ranges
in
which,
what
method
or
what
function
they
correspond
to.
C
L
Its
items
we
think
it's
about
the
same
as
as
before,
right
there
was
big
it
used
to
be.
When
you
did
this,
the
same
thing
worked:
it
just
went
into
a
domain-specific
data
pipe
right,
so
we
just
made
it
now.
It's
going
into
a
common
data
pipe
so
who
the
part
of
the
reason
that
we're
being
cautious
is
we're
making
sure
that
the
what
we've
done
is
is
perfect,
neutral,
I,
look
great
I
mean
we
had
to
do
a
lot
of
tweaking
and
we
may
not
be
a
hundred
percent
under
it.
L
C
J
L
Can
do
is
just
take
a
peek.
It
like
a
pop
up
and
chrome,
take
a
peek
it.
What
happens
when
you
turn
on
Chrome's
tracing
with
the
v8
dot
CP
profile
category
on
and
then
you
you
can
look
at
the
file
right,
ignore
the
GUI
yeah
look
at
the
file
that
comes
back
and
just
start
looking
at
it
kind
of
doing
a
methyl
critique.
You
know,
and
if
you
add
me
on
threads
I
can,
I
can
add
elf
Alexei
filippov,
who
did
like
he
did
the
actual
work
here.
I'm
totally
proxying
well.
G
The
money
you
know
that
he
was
slightly
different,
the
one
you're
mentioning
so
so
that
two
types
of
things
that
you're
talking
one
is
multiple
samples
of
the
same
thing.
The
other
is
the
issue
you're
talking
about
is
the
fact
that
viet
never
reaches
a
stable
point
for,
from
a
jet
point
of
view,
there's.
C
B
C
L
Asking
the
right
questions,
one
and
one
thing
I
would
circle
us
back
to
is
one
of
the
reasons
that
one
of
the
things
that
makes
the
sampling
stuff
particularly
hard
is
that
that
won't
inside
v8
its
external
right
for
good
reasons
from
the
point
of
view
of
v8.
But
it
means
that
when
we
do
something,
there
is
work
that
the
node
community
has
to
do
to
replicate,
that
specific,
that
specific
thing
and
that
that
is
the
current
shape
of
things
right.
L
H
L
It's
not
an
API
like
again,
because
it's
in
content,
it's
it's
a
thing
that
we've
designed
it.
We've
shown
that
you
can,
you
can
just
say,
start
tracing
so
to
remember
the
way
tracing
works
is
you
say,
start
tracing,
get
me
these
probes
right
kind
of
its
like
dtrace.
In
that
sense,
it's
like.
Please
get
me
access
to
these
probes,
except
we
just
stream
all
the
results
right,
whereas
in
dtrace
you
are.
You
know
many
of
these
other
sampler
tracing
frameworks,
you're
processing,
the
crap
surprise.
L
L
L
It's
sort
of
like
you
know
if
we
did
a
block
diagram,
you
know
somehow
you'd
have
v8
down
at
the
bottom
of
the
stack
right
and
then
you'd
have
the
node
embedding,
which
would
internally
contain
a
tracing
buffering
implementation
that
would
probably
like
forked
out
to
many
different
backends
like
LT
TNG
or
whatever
sampling
subsystem,
would
be
inside
the
tracing
controller.
So
it'd
be
a
tracing
agent
inside
node
that
talks
back
and
forth
to
VA,
but
that
is
more
in
node
than
inside
v8,
but
it
gets
turned
on
and
off
by
the
basic
start.
L
L
G
So
in
a
textual
form,
where
I
wrote
something
in
a
previous
issue:
interfacing
workgroup
on
what
would
be
needed
to
pick
up
Trace
event
or
age,
and
what
are
the
components
like
how
do
busy?
We
need
another
thread
that
and
a
buffer.
That's
it.
The
sink
of
the
altar
fitted
are
coming
in.
So
it's
not
a
diagram
but
faria
it
from
it's
a
starting
point:
yeah.
C
L
K
G
L
G
So,
and
the
other
thing
I'll
add,
is
at
their
money
but
ricochets
profit
V,
it
has.
Is
they
don't
defecate
things
for
the
next
reveals
things
we
have
a
long
defecate
soon,
then
the
character?
We
can't
speak
on
behalf
of
the
vehicle,
so
so
so
API.
What
that
is
going
to
happen
to
it
in
the
future,
only
Wheatley,
indirectly,
not
so
another
type
for
that
would
would
be
the
greatest
on.
A
That
conversation
great,
what's
going
into
that
up,
then?
Yes,
specifically
what
I'm
thinking
is
I
think
it's
some
of
the
v8
debug
headers
are
things
you
know
that
specifically
I,
don't
know
what
names
are,
but
you
know
have
to
do
with
starting
and
stopping
CPU
profiles.
That's
the
level
while
I
was
thinking
at
and
I.
Think
of
that
in
terms
of
API.
A
Even
though
I
don't
know
the
exact
API
definitions
will
be
I,
understand
they
change
and
that's
fine,
but
just
trying
to
get
a
sense
for
if
it
would,
if
that
existing
API
would
still
be
there.
There
are
certainly
some
packages
that
exist
out
and
NPM,
for
instance,
to
make
use
of
those
I,
think
vh,
profiler
or
maybe
just
be
a
profiler
with
no
dash
is
one
of
those
those
packages
so
be
good
to
have
as
much
send
ahead.
A
L
You
know
seemed
like
a
good
breather
point:
ok,
there's
the
sampling
stuff
and
obviously
we
all
agree
that
that's
super
super
important
right,
the
the
other
caught
the
other
half
of
this
equation
is
structural
profiling,
so
v8
does
things
that
are
that
are
that
can
be
time-consuming
like
parsing
and
codegen.
You
know
compiling
and
GC,
and
all
these
things
and.
L
Think
when
we
get
to
that,
we'll
we'll
try
to
figure
out
slowly
how
to
encode
things
like
promised
dependencies
and
all
that
other
stuff
that
a
sink
rep
has
done
before
us,
but
do
so
in
in
sort
of
a
common
data
type
format.
But
that's
all
kind
of
everything
there
is
just
sort
of
dangling
off
of
the
basics
and
I
don't
have
like
a
really
positive
update
their
other
than
we're
like
chugging
away.
L
Great,
oh
I
would
say
one
thing
is
you
might
be
tweaking
the
data
pipe,
so
I
don't
know
if
this
happens
in
the
note,
but
in
the
browser
world
you
have
11
tab
open
that
has
many
many
different
script
contexts
associated
with
it
for
different
iframes
and
extensions,
and
things
like
that
and
for
our
profiling.
We
want
very
much
to
see
the
distinction
between
those,
so
we
want
to
be
able
to
display
our
data
string
by
domain
or
by
which
I
frame
is
taking
time
right.
Like
you
know,
oh
it's
this.
It's
this.
L
L
But
what
context
the
event
is
from
and
context
me
is
in
reference
to
do
a
unified
notion
of
frame
and
script
context,
so
that'll
probably
make
more
sense,
as
people
start
using
tracing
in
this
community,
if
rather
people
start
using
tracing
in
this
community.
But
I
wanted
to
mention
that
the
the
basic
data
pipe
is
not
completely
fixed
in
time,
and
we
do
encounter
things
where
we
have
to
tweak
it
like
this
I
just.
C
G
L
A
Again,
we
don't
have
to
go
through
this
stuff,
but
I.
You
know
I,
I,
more
or
less
understand
the
words
you
said,
but
try
trying
to
kind
of
fit
it
on
my
head,
how
it
fits
together,
I
kind
of
wonder
like
well
trace
events.
I
can
already
kind
of
provide
some
data,
so
why
is
the
structural
profiling
an
issue
and
why
is
providing
context
an
issue
because
can't
that
just
be
data
into
the
trace
of
n?
A
Let's
not
talk
about
that
now
because
we
were
actually
running
out
of
time,
but
if
you
could
keep
that
in
mind,
as
we
start
to
kind
of
build
some
docs
in
terms
of
well,
what
I'll
just
I'll
just
ask
it
if
it's
not
clear
when,
as
we
start
to
kind
of
get
a
chance,
it's
a
layering
problem,
it's
just
about
where
we
can
add
the
probes
cool
all
right.
Well,
we
have
five
minutes
left
we
work
through
all
of
the
agenda
hide.
A
A
Think
we
can
give
everybody
five
minutes
back
in
their
day.
So
thanks
a
lot
for
everyone
coming,
I'm
so
happy
that
we
kind
of
kicking
this
off.
Hopefully
we
can,
you
know,
continue
and
do
more
of
these
again
in
two
weeks.
Try
again
so
happy
everything
actually
worked.
I
was
very
scared
that
hanging
out
wouldn't
work
and
I.
Guess
we'll
we'll
see
it.
The
next
call
will
see
I'm
github
and
the
other
places
we
have
I
given
by.