►
From YouTube: 2023-02-22 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
B
A
D
All
right,
Sammy,
it
looks
like
the
first
item
is
yours.
F
See
the
exponential
histogram
PR's
are
still
open.
This
is
the
same
thing.
We've
talked
about
for
the
last
several
weeks.
F
F
D
E
D
F
B
B
Oh
yeah
I
noticed
that
some
tests
were
failing
on
Windows
kind
of
went
under
the
radar
in
the
CI,
but
I
ran
the
tests
on
in
in
CI
on
the
on
the
fork
that
we
have
and
the
tests
seem
to
be
failing
for
a
while,
so
I'm
trying
to
track
down
the
issue.
B
There
also
I
didn't
notice
because
I'm
not
not
usually
running
Windows,
so
it
kind
of
flew
under
under
my
radar
as
well
yeah,
but
there
seems
to
be
some
issue,
I
guess,
with
the
dependencies,
because
I
went
back
quite
far
in
the
history
and
it's
still
failing
so
yeah
I'm
still
investigating
that
right.
The
question
I
put
underneath
is
actually
not
really
exactly
related
to
that.
But
more.
Like
a
follow-up
question
for
the
sandbox.
F
Yeah,
so
this
is
something
Mark
and
I
were
talking
about
earlier.
The
the
out
of
memory
errors
are
likely
caused
by
a
dependency
that
updated,
and
you
know
that
wasn't
pinned
exactly
or
something
along
those
lines,
we're
using
significantly
outdated
development
dependencies.
At
this
point
for
a
from
a
testing
and
building
tooling
perspective
and
I
know
that
nav,
you
updated
the
sandbox
to
use
a
more
modern
stack
and
the
the
question
that
we
had
was
just
how
hard
would
it
be
to
apply
that
to
the
main
repo?
C
It's
possible:
the
latest
version
of
Rush
run
still
runs
on
14..
You
can
go
back
to
earlier
version
and
still
get
12
support.
If
you
want
it,
it's
not
a
small
amount
of
work.
I
guess
my
script
for
the
sandbox
does
all
most
of
the
automated
stuff,
so
not
impossible,
but
at
least
a
week's
worth
of
work,
plus
a
bunch
of
testing
Rush
does
tend
to
to
like
to
pin
things
completely,
which
is
not
how
the
main
repost
currently
work.
C
Just
purely
so
I'll
create
Sim
links
for
for
all
the
node
modules.
You
don't
end
up
downloading
the
same
thing
over
and
over
again.
F
C
C
Sandbox
today,
which
I've
got
an
uptake,
which
is
my
next
bit
like
I,
already
have
core
API
and
spanning
conventions
merged
as
of
January
16th
so
effectively.
Anything
that
was
committed
to
a
core
and
to
JS
and
contrib
is
now
also
in
sandbox.
Those
three
are
now
running:
CI
tests
complete
with
worker
tests,
which
I
don't
think
core,
has
worker
tests
for
all
of
them
and
a
new
PR
out
for
contact,
Zone
and
context
dependency,
which
I
think
one
of
those
doesn't
have
any
tests.
F
Okay,
I
guess
in
terms
of
the
original
motivation,
which
is
just
to
use
a
more
modern
to
update
the
tooling,
because
when
we've
tried
in
the
past
with
the
webpack
and
learn
a
stack,
we
weren't
able
to
do
it
because
of
old
node
version.
Sport.
Do
you
think
it
would
be
worth
switching
to
rush
and
roll
up?
For
that
reason
or
I?
Guess
I
I'm
not
entirely
clear
why
you
did
it
for
the
sandbox
so
like?
C
They
would
apply
to
the
main
repo,
so
I
mean
so
apart
from
these
are
the
two
that
I
use
all
the
time
rash
creates
like
Sim
links
for
everything.
So
you
end
up
with
a
lot
less.
C
You
know
stuff
on
your
machine
roll
up
generate
smaller
bundles
than
than
webpack,
which
is
primarily
why
I
use
that
one
yeah,
like
it's
part
of
that
API
core
and
semantic
inventions,
it's
creating
bundles
in
every
possible
type,
as
well
as
part
of
the
build
like
that's
that's
one
of
the
things
that
script
does
in
terms
of
enabling
Roll-Up.
F
D
F
F
I
guess
until
we
track
down
the
window,
it's
only
a
theory
that
the
the
old
dependencies
are
causing
this
out
of
memory
issue.
It
may
be
something
completely
different,
so
yeah.
C
I
think
it
probably
is
different.
I
know
locally
with
our
internal
projects.
Node
16
has
greater
memory
requirements
and
causes
a
bunch
of
grief
so
yeah.
It's
not
just
the
fact
that
it's
windows,
or
not
just
that
it's
the
the
runtime
framework
I
think
there
are
some
libraries
like
a
mocker
is
one
of
them
that
effectively
just
uses
more
memory,
especially
when
you're
doing
coverage.
D
F
C
Yeah
pretty
much
already
talked
about
it
so
effectively,
those
three
already
there,
as
of
January
16
sometime
between
the
previous
merge
and
then
well
since
January
16,
and
then
one
that's
well,
no
previous
one
on
January
16
there
was
a
change
incorporated
into
core
where
the
TS
config
es5
was
renamed
to
esm.
C
So
I
had
to
fix
my
scripts
to
deal
with
that,
which
is
what
the
pr
that's
that's
there
is
mainly
doing,
and
then
it's
also
adding
or
perfectly
uncommenting
the
contact,
Zone
and
resources.
The
plan
is
to
still
have
every
web
package
up
and
running
by
the
end
of
the
month.
I'm,
probably
not
going
to
have
the
examples
running
during
this
time,
like
I,
said
I
think
I
mentioned
before
there's
already
some
tests
that
are
broken,
because
the
visibility
change
is
one
that
keeps
coming
to
mind.
D
F
H
All
right
no
worries
you
can
hear
me
right,
yeah
yeah.
We
can
hear
you
all
right
perfect
because
I'm
having
some
issues
with
the
internet,
okay,
so
I've
added
this
to
the
list,
because
I
wanted
to
see
if
we
can
go
over
this
or
if
someone
has
anything.
If
anyone
is
interested
in
this
package,
I
had
some
concerns
with
this.
H
F
If
we
wanted
to
put
it
in
core
I
would
really
only
want
to
add
the
instrumentations
that
are
in
the
core
repository
to
it
right,
which
would
be
HTTP
and
grpc
only
I.
Think
at
this
point,
I
guess.
The
question
for
me
would
be
if
we
were
gonna
put
it
in
core.
F
How
do
we
decide
which
instrumentations
to
include
in
core?
Because,
with
only
with
only
HTTP
and
grpc,
it
has
limited
usefulness,
I?
Think
for
now,
contrib
is
a
reasonable
place
for
it.
We
may
want
to
add,
contribute
to
the
name,
just
so
that
it's
clear
that
it's
pulling
in
contrib
packages.
F
I
don't
know,
I
would
have
to
think
about
it
a
little
bit
more.
Anyone
else
have
an
immediate
the
auto
opinion
there
right
I
just
worry
that
this
name
may
be
something
we
want
to
use
later,
for
course,
so
we
may
want
to
to
just
add,
contribute
to
the
name
now
to
open
up
the
possibility
in
the
future
of
having
a
adding
one
in
the
core
repo.
D
F
F
So
we
probably
just
want
to
stay
consistent,
but
I
think
mostly
what
it
means
is.
It's
like
zero
configuration
and
no
code
changes.
I
F
I
F
B
We
we
may
be
able
to
to
borrow
the
auto
configure
name
that
they
have
floating
around
in
in
the
Java
repo
Auto,
configure
I
think
would
describe
very
well,
but
it's
what
it's
doing.
F
Yeah
or
Auto
instrumentation.
H
Yeah,
that
was
my
first
thought,
to
name
it
maybe
Auto
instrumentation,
because
it
suits
it
better
than
agent,
but
we
do
have
auto
transportation
node
for
in
JavaScript
that
that
includes
all
the
instrumentation,
so
I,
don't
think
that
will
work.
F
So
one
option
would
be
to
to
use
that
auto
instrumentation
package
instead
of
creating
a
new
one.
F
F
H
F
Yeah
I
agree
and
then
I.
F
Know
that
I've
seen
in
other
packages,
they
use
like
slash
register
path
for
the
for
the
Dash
R
node
flag,
just
to
make
it
clear
what
what
it's
for
and
what
we're
doing.
So.
That
might
be
a
good
idea
here
and
just
edit
this
to
show
you
what
I'm
talking
about
so
it
would
be
like
node
require.
H
F
I
Inducing
it
without
this
register
is
a
breaking
change
for
this
package,
which
is
experimental.
So
we
can
do
it,
but
it
will
break
people
because
if
they
just
import
open,
Telemetry,
Auto,
instrumentation,
node
and
now
it
will
register
an
SDK.
Then
it
will
change
the
behavior
of
the
code
which
they
need
to
be
prepared
for.
I
F
I'm
not
sure
how
the
other
sigs,
with
with
auto
instrumentation
packages
and
handled
this,
but
I,
know
that
there
is
a
specification
working
group
for
configuration
in
general.
That's
trying
to
define
a
configuration
file
format
and
that's
probably
the
long-term
answer
to
that
question
in
the
short
term.
I
think
you
would
only
be
able
to
use
this
if
you
need
either
defaults
or
things
that
can
be
configured
with
environment
variables
and
I.
I
J
At
least
in
Python,
like
the
there's,
a
command
line
wrapper
to
to
do
the
auto
instrumentation
similar
to
this
and
it's
sort
of
decoupled
from
the
installing
the
instrumentations
there's
like
optional
requirements,
sort
of
to
install
them,
but
it's
decoupled
for.
For
this
reason,.
B
I
guess
in
that
case
we
would
have
to
move
it
in
its
own
package
anyway,
not
not
move
it
into
the
auto
instrumentations
package.
F
J
So
it
does
like
a
I
believe
it
does
like
a
defensive
import
and
then
there's
also
this
other
script
open,
Telemetry
bootstrap,
which
can
you
can
just
read
the
description
right
there,
but
it
basically
helps
you
find
instrumentations
depending
on
things
that
are
already
installed
in
your
environment.
But
that's
sort
of
separate
from
the
first
thing.
I
believe
it
just
does
like
a
defensive
import.
F
F
Right,
okay,
so
we
could
do
something
similar
I
mean
in
JavaScript.
There
would
be
pretty
easy
to
have
a
similar
mechanism
where
we
just
have
a
list
of
instrumentations
that
are
tried
to
load
if
we
did
it
with
a
CLI
thing
like
this,
it's
definitely
a
different,
a
different
way
to
go.
J
Yeah,
it
looks
like
that
because
I
feel
like
it
would
be
nice
for
it
to
be
pluggable
as
well.
So
you
can,
you
know,
include
extra
ones
that
aren't
aren't,
like
part
like
checked
into
the
repository
right.
F
Yeah
I
just
don't
know
how
you
would
do
that
like.
We
would
have
to
probably
read
the
package
Json
and
then
read
the
package
Json
of
each
of
your
dependencies
and
look
for
you
know.
Maybe
we
could
Define
a
like
our
own
custom
field
in
package.json.
That's
like
open,
Telemetry,
true
or
something
like
that
or
open
Telemetry
instrumentation.
F
F
F
Like
I
I
think
what
Sami
does
provide
value
and
I
don't
want
to
block
it
based
on
you
know
something
that
we
might
do
later.
J
F
J
We
could,
you
know
like
push
off
if
we
wanted
to
add
a
separate
package
that
decoupled
it
and
then
you
know
point
this
one
to
use
the
other
one
or
something
like
that,
but
both
both
ways
make
sense.
To
me.
I
was
just
trying
to
raise
that
point.
Yeah.
C
The
comment
would
be
that
if
we
kept
this
name
I'd
like
to
have
node
in
there,
but
if
we
were
using
the
other
package
name,
then
that's
already
taken
care
of
so.
F
Yeah,
having
note
in
there
makes
sense
I,
don't
know
if
in
instrumentation
agent
like
this
yeah,
it
would
work
very
differently
in
a
browser.
It
would
have
to
be
like
a
build
time
thing
almost
like
a
webpacker
or
Lola
I.
Don't
even
know:
okay,
I'm,
not
a
browser
expert.
From
that
perspective,
so
I
don't
know,
what's
what's
even
possible
to
do.
F
Okay,
so
I
think
for
now
the
Sammy
I
would
say:
use
the
existing
Auto
instrumentation
node
package
with
like
the
slash
register
convention
and
we'll
go
from
there.
G
Yeah,
just
just
a
quick
update,
there
have
been
some
changes
to
the
pr
it
now
implements
the
API
correctly,
based
on
the
updates.
You
know
from
from
a
few
weeks
go
to
the
API.
G
There
is
still
it's
still
work
in
progress.
There's
more
tests
that
the
contributor
is
working
on,
but
I
was
gonna.
Ask
for
a
quick
review
for
anyone,
who's
interested
just
from
high
level
design
perspective
like
you,
don't
have
to
go
into.
G
You
know
the
details
of
the
implementation,
but
just
like
make
make
sure
that
the
classes
and-
and
you
know
one
thing
that
I
that
I've
been
kind
of
looking
at
is
comparing
it
to
the
trace
SDK
which
how
how
to
provider
the
the
Tracer
and
the
span
classes
are
designed.
This
is
kind
of
similar,
so
you
can
just
focus
and
take
a
look
and
just
see
if
that
makes
sense.
As
a
second
opinion
and
then
I'll
continue
up
in
the
country
contributor.
With
the
details
like
with
tests
and
stuff.
D
F
Okay,
I
will
definitely
make
sure
to
review
that
when
I
have
time
this
week,
I
was
I
was
out
all
last
week,
so
I'm
still
catching
up
on
on
things
that
changed
while
I
was
gone.
A
Yeah,
just
a
quick
update
that
this
PR
is
ready
to
be
reviewed.
There
was
just
some
back
and
forth
about
recommended
changes
that
I
did
so
it's
ready
now.
F
Okay,
yeah
I,
this
one's
been
open
for
a
really
long
time.
I
apologize
for
I
haven't
been
great
at
reviewing
this
one
either
so
I'll
make
sure
to
make
time
for
that
this
week.
Well,.
J
Yeah
this
was
mostly
just
an
FYI
I'm
planning
to
work
on
the
open
census.
Bridges,
so
there's
there's
got
to
be
I'm
gonna
make
another
bug
for
the
Trace
bridge
and
I
linked
the
spec
for
that.
If
anybody's
curious
in
the
JS
meeting
notes,
but
basically
the
idea
is
to
make
it
so
open
census
and
open
Telemetry
can
work
together,
sort
of
sharing
context
and
then
for
for
metrics.
It's
outlined
in
that
bug
already
how
it's
supposed
to
work.
F
Okay
is
that
different
than
the
it's
yo.
A
F
I
assigned
that
issue
to
you.
It
sounds
like
there
isn't
any
other
action
item
from
that.
Quite
yet,.
J
No,
not
yet
I'll
probably
hopefully
have
PR's
next
week.
D
I
Yeah
so
I
was
going
over
the
open
PLS
in
contrib.
Some
of
them
are
really
really
old,
so
this
one
is
from
a
September
and
I
talked
to
this
guy
Severin
I
asked
him
if
he's
still
interested
in
pushing
this,
and
he
said
he
that
he
wants
to
walk
on
it,
but
he's
stuck
with
the
webpack
issues
and
I
wasn't
able
to
help
him
because
I
don't
know
a
lot
about
webpack.
So
if
somebody
can
take
a
look,
never
I
know
you
have
a
lot
a
lot
of
experience
with
it.
D
F
F
It
looks
like
one
of
his
suggested
Solutions
is
to
update
to
webpack.
5.
I
did
try
to
do
that
once
in
the
past,
and
it
was
a
lot
more
work
than
I
had
thought
it
was
going
to
be,
but
mostly
ESU
was
when
I
updated
webpack.
It
required
me
to
update
other
dependencies
and,
as
I
was
updating
them.
The
old
node
versions
were
failing,
so
there
wasn't
a
way
to
to
make
them
work
together.
That
I
was
able
to
find
is
always.
I
Using
for
Dev
dependencies
right
for
running
the
tests,
yeah,
it's
not
actually
distributed,
so
there
shouldn't
be
any
effect
on
yourself
if
we
upgrade
to
web
deck
5.
F
I
C
Yeah
I
think
it
broke
the
tests
as
well
from
memory
that
was
a
while
ago
now.
F
Yeah,
it's
been
a
while
I've
tried
to
do
this
twice
now,
like
I.
Don't
want
to
just
run
super
old,
tooling
forever,
so
I
do
think
it's
something
that
we
want
to
do
eventually,
I!
Guess
it's
worth
pointing
out
that
I
tried
to
do
it
in
the
core
repo
I,
don't
know
if
it
would
be
easier
or
harder
in
the
contrib
repo.
My
expectation
is
that
it
would
be
harder
just
because
there's
a
lot
more
Dev
dependencies
and
different
versions
of
things
in
the
contribute,
though,
but
I'm
not
sure.
C
One
thing
one
thing
we'll
probably
do
is
once
the
sandbox
is
up
and
running
quite
a
branch
and
just
go
back
at
it
and
and
see
how
we
go.
That's
the
whole
point
of
the
sandbox
is
to
try
and
get
it
working,
and
once
we
get
it
working,
we
can
then
bring
it
all
back
into
Json
country
because
it'll
be
both.
At
the
same
time,
then.
C
Well,
it's
sandbox
is
already
going
to
have
rush
and
roll
up
bundles,
but
it
still
uses
webpack
to
run
the
tests.
So.
C
Yeah
I
I
want
Maine
to
always
be
in
sync,
with
contrabandjs,
so
there'll
be
no
direct
contributions
unless
it's
for
fixing
up
scripts
domain,
so
we
create
branches
and
go
hack
at
it.
C
It's
just
a
faster
thing
to
have
Contour
bingeus
in
the
one
place.
B
I
think
I
think
the
problem
was
with
Note
8
I
I
know
at
least
back
when
when
I
tried,
it
I
had
the
problems
there
and
if
I
recall
correctly,
there
was
also
some
problem
with
type
script
needing
updates
or
or
some
other
dependencies
causing
typescript
to
need
an
update,
which
then
cost
more
problems.
B
F
Yeah
I
in
any
case,
I
think
the
sandbox
is
the
right
place.
For
you
know,
that's
that's
what
it's
for,
and
it's
almost
ready
so
I
would
say.
Probably
the
smart
thing
to
do
here
is
to
wait
for
the
the
sandbox
to
be
ready
and
we
can
try
to
update
to
webpack
5
there
and
see
what
problems
we
run
into
yeah.
C
The
early
Dante
is
currently
I'm
only
dragging
over
web-based
packages,
but
we
can
change
that.
F
F
I
Hey
yeah
I
will
write
it
on
the
pl
and
ask
him
if
he
can
try
to
upgrade
to
webpack
5
and
if
it
works,
then
it's
amazing
right.
We
don't
it
shouldn't
cause
any
errors.
If
the
tests
are
all
passing.
F
All
right
and
then
I
guess,
let's
do
bug
triage
Windows
tests
failing
I'm
gonna
mark
this
as
keyboard,
since
it's
not
really
causing
any
production
issues.
Right
now
is
that
okay,
Mark
yeah.
D
B
I
have
I
have
seen
that
already
and
it
like
the
thing
that
they're
describing
may
be
related
to
to
cardinality
problem.
B
B
F
F
B
You
can
assign
it
to
me
and
I
will
do
some.
Some
investigation
just
run
it
for
a
bit
and
see
if,
if
that
causes
any
issues
and
then
follow
up
with
a
question,
if,
if
I
can't
reproduce
it,
okay.
F
Will
you
prioritize
it
then.
B
F
Actually
I
will
just
assign
a
P1
so
that
there's
no
reason
not
to
at
this
point
it
turns
out
not
to
be.
We
can
always
change
it.
B
Oh
yeah,
we
discussed
this
one
last
time.
I
think
I
asked
a
question
there
to
leave
the
container
detector
out.
I.
Think
Amia
was
coming
up
with
that
idea,
because
there's
the
eks
detector
and
then
there's
the
container
detector
and
one
might
be
overriding.
The
other
resource
attribute
with
a
different
container
ID.
B
F
B
Right,
I'm
not
too
familiar
with
that
so
I
wouldn't
know
the
answer
to
the
question.
Unfortunately,.
E
F
I'm
not
sure
if
it
does
in
any
case,
this
actually
is
a
contrib
issue.
I'm
going
to
transfer
this
to
the
contribute
Repository.
I
Yeah,
actually,
there
is
a
related.
The
pl
I
sent
it
in
the
chat.
Nowhere
in
the
documentation
do
we
specify,
when
you
give
them
multiple
resource
detectors
in
which,
although
they
will
be
merged
and
the
spec
also
doesn't
say
anything
about
it.
Actually
I
couldn't
find
any
reference
to
Resource
detector
in
the
spec
beside
places
that
just
use
it
without
defining
it
yeah.
I
Yeah
and
this
issue
is
referencing,
The,
Collector
and
The
Collector
seems
to
be
doing
the
opposite
for
what
we
are
doing.
So
if
multiple
detectors
have
the
same
resource
attribute
and
the
first
one
is
taken
in
The
Collector
and
we
take
the
last
one,
I
think
and
the
spec
doesn't
say
anything
about
how
to
do
it.
So
it
seems
like
a
bit
arbitrary,
no
and
I
think
it's.
It
should
be
clear
like
we
should
document
it
somewhere
to
say
the
order
of
the
resource
detectors
is
important.
I
I
I
D
I
I
There
is
some,
like.
You
see
the
detecting
resource
information
from
the
environment.
Here
they
use
the
term
resource
detector,
but
they
just
use
it.
It's
not
defined
anywhere
and
some
other
places
in
the
spec
also
they
use
it,
but
the
that's
it
like,
there's
no
like
head,
or
that
says
resource
detector.
This
is
what
it
means.
This
is
how
to
use
it.
At
least
I
couldn't
find
one
okay.
F
I
Yeah
but
because
the
collector
is
doing
the
opposite,
they
suspect
that
different
implementation
may
have
different
order,
which
is
unfortunate,
and
it
was
just
out
of
curiosity
to
understand
if
we
did
something
and
not
according
to
the
spec
or
if
it's
just
unspecified
it
seems
to
be
unspecified.
I
F
I
think
that's
a
different
issue
than
what
this
is,
though,
is
there
anyone
from
AWS
on
a
call
here.
F
Doesn't
look
like
it's
oh,
what's
his
name.
F
The
component
ownership
file.
F
F
J
To
the
the
op
posted
in
slack
in
the
night,
that's
how
I
found
this
because
some
Google
details
I
mean
based
on
reading
through
it
I
think
it's
working
as
intended
and
I.
Don't
think
it's
grpc
specific,
so
I
mean
I
want
to
help
them
with
their
issue,
but
I,
don't
think
there's
any
the.
J
So
they're
calling
Force
flush
not
shut
down
and
then
what
happens
is
the
I
think
what's
happening?
Is
the
container
gets
throttled
on
the
serverless
platform
and
then,
by
the
time
that
the
CPU
is
reallocated,
the
deadline
has
already
exceeded,
got.
F
It
okay
yeah
since
you're
already
looking
into
this,
can
I
assign
this
to
you
yeah
sure.
J
And
I'm
not
sure
yeah.
That's
me,
I'm
not
sure
what
tags
you
have,
but
maybe
like
question
might
be
more
appropriate
than
bug.
F
Okay,
so
I'll
remove
the
bug
label.
For
now
it
turns
out.
It
is
a
bug
you
can
always
reapply
them.
I
think
you
should
have
permission
to
do
that
right
since
you're,
an
approver.
J
I'm,
not
an
approver
in
this
repo
I,
don't
think
I
have
permission.
Oh
you're.
F
Not
okay,
if
you,
if
you
do
find
that
at
least
above
just
let
me
know
and
I'll
I'll
change.
The
labels
back
yeah
sounds.
J
F
Okay,
yeah
I
know
I
already
assigned
this
to
myself.
I
haven't
had
any
time
to
look
into
it,
but
I
will
that's
my.
F
Worker
thread,
this
is
an
old
one,
still
no
updates,
and
this
one
is
also
this
one
already
has
a
PR
which
is
still
a
draft.
Oh,
this
is
the
same.
One
I
I've
been
trying
to
help
him
look
into
this
okay
and
then
contrib,
I
guess.
F
B
Yeah
no
updates
yet
I
haven't
gotten
around
to
looking
into
it.
Yet.