►
From YouTube: 2023-02-01 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
B
B
A
Then
welcome
everybody
and
yeah.
The
first
point
on
the
agenda
that
we
have
today
is
the
bug.
Fix
versions
were
released
for
the
for
the
core
packages
and
the
experimental
packages,
so
the
issue
with
the
grpc
instrumentation
should
not
be
resolved,
which
yeah
was
blocking
the
country
release
yeah.
Anything
more
to
add,
regarding
that,
from
from
any
of
you.
A
D
Would
add
is
that
I
did
merge
the
pr
to
update
contrib,
but
I
can't
actually
do
the
release
until
I
get
an
approving
review
on
it.
So
anyone
with
permission
please
take
a
look
at
the
release
main
PR
on
contrib,
so
that
we
can
actually
get
that
release
through.
A
All
right,
thank
you,
then,
moving
on,
we
still
have
parts
of
the
exponential
histogram
PR
that
was
split
up
out
there.
The
first
part
was
merged
already
and
the
other
two
parts
are
still
awaiting
approvers.
So
if
any
of
you
are
interested,
please,
head
on
over
there
Matt
give
us
a
walk
through
in
meeting
in
in
December
and
yeah
I.
A
Well
explained,
but
it
it
definitely
needs
quite
a
few
reviews
there
to
so
that
we
can
get
that
merge.
It's
quite
a
complicated
topic,
but
yeah
explanation
was
was
quite
well
done
right
and
then
I
think
we
can
move
on
to
the
topics.
E
Yeah,
it's
it's
rebased
with
the
the.
A
Yes,
yeah
I
will
definitely
head
on
head
on
over
there
later
this
weekend.
Give
that
a
review.
So
we
can
move
that
along
all
right.
Then.
The
next
point
on
the
agenda
is
from
Martin
about
the
pr
that
separates
the
events
API
from
Deluxe
API.
E
Yeah
so
so
I
just
just
quickly.
It's
been
approved,
I
think
since
even
since
last
week,
I
think
it's
fails
on
some
unit
tests,
but
it's
not
related
into
this
PR
yeah.
A
E
There
so
I
was
just
wondering
like
there
was
one
question
about
the
name
of
the
class
event:
emitter
I,
don't
think
it's
a
big
deal
personally.
I
think
it
can.
You
know
I
would
I
would
rather
merge
this
now.
So
we
can
continue
with
the
log
SDK
and
the
other
packages,
and
then,
if
you
need
to
rename
it,
we
can
rename
it
all
of
it
later,
because
the
name
is
actually
not
defined
in
the
spec
currently.
E
But
is
there
any
concern?
If,
if
we
name
it
event,
emitter
is
like,
which
is
the
same
as
the
built-in
built-in
event
emitter
interface
from
node.
D
I
would
prefer
it
not
just
because
the
way
that
people
tend
to
import
things,
they
would
Shadow
the
event.
Emitter
name
yeah,
and
it
might
be
just
a
little
bit
annoying
right,
like
you
would
have
to
import
event,
emitter
as
Hotel
event,
emitter
or
something
along
those
lines
in
order
to
like
import
it
without
clobbering.
The
built-in
event
emitter,
so
I
would
probably
prefer
not
to
have
the
exact
same
name
as
a
built-in
global.
E
A
Yeah
but
I
think
I
agree
with
Martin
that
this
can
be
done
in
the
follow-up,
but
probably
best
to
reach
out
to
legendika's
to
to
ask
what
what
he
thinks
about
it,
but
I
I
guess.
He
will
also
see
that
point
that
that
can
be
done
in
follow-up.
E
Yeah
I
guess
I
guess
my
my
concern
is
that
discussion
about
a
name
could
take
like
another.
You
know
week
or
longer
potentially,
and
we
would
like
to
move
on
to
working
on
the
other
packages
which
this
PR
is
currently
blocking.
E
So
I
understand
the
concern
and
I'm
and
I'm
I.
Oh
I,.
E
Kind
of
have
a
discussion
about
what
that
name
should
be,
but
I
was
wondering
if
you
could
just
move
forward
with
this
particular
PR.
D
Yeah,
on
my
in
my
mind,
it
does
not
block
this
PR,
especially
since
we
just
had
a
release
like
yesterday
or
two
days
ago
or
whatever,
there's
plenty
of
time
to
update
the
name
before
we
even
have
a
release.
So
we
may
not
even
need
to
worry
about
it
being
breaking
or
anything
like
that.
Okay,
so
yeah
I
will
rerun
these
tests
and
merge
this.
E
A
All
right,
then
there
is:
are
there
any
follow-up
questions
to
that
or,
if
not
then
I
guess
we
can
move
on
to
the
next
part,
which
is
about
the
sink
resource
with
some
attributes
that
Reserve
asynchronously
are
you're
on
the
call
right,
You
can
hear
me.
Yes,
we
can
hear
you.
F
All
right
so
this
year,
I'm
having
with
this
after
the
new
release,
typescript,
is
throwing
some
issues,
throwing
some
errors
with
the
with
the
implementation
of
the
resource
class.
Since
it
has
some
private
attributes
now
and
I
have
some
compatibility
tests
that
use
the
old
resource
implementation
and
it
started
throwing
some
weird
errors
that
I
haven't
really
encountered
before.
I
spoke
to
to
Amir
about
it
and
they
helped
me
understand
a
bit,
but
we
weren't
sure
what
exactly
would
be
the
right
solution
for
this
yeah.
C
I
can
I
can
add
some
background
on
this.
So
I,
don't
know
if
you
remember
the
issue
we
had
with
the
spend
that
every
time
we
released
a
new
version,
then,
if,
if
we
use
the
two
concrete,
the
glasses
of
spin
from
two
different
versions,
then
typescript
would
not
compile,
and
if
I
remember
correctly,
we
ended
up
understanding
that
we
must
use
an
interface
and
not
the
concrete
class
right,
because
it's
very
easy
for
someone
to
create
a
setup
where
the
SDK
Trace
base
is
the
latest
version.
C
But
something
else
like
a
resource
detector
is
not
the
latest
because
it
hasn't
been
released
or
because
it's
been
the
full
version
and
not
being
updated
and
I
think
that
the
resource
class
is
like.
Currently,
it
has
only
public
properties.
So
we
didn't
see
any
issues,
but
now
we
it's
also
changed
and
also
it's
not.
It
has
a
private
properties,
so
I'm
expecting
it
to
be
a
problem
and
and
I'm
I'm
not
sure.
C
What
are
the
options
here,
like
I,
think
that
having
a
resource
interface
would
be
like
the
best
solution,
but
I'm
not
sure
if
we
can
do
it
in
a
non-breaking
change
way
and
if
it,
if
it
makes
sense
to
do
it
in
this
case,
if
it's
not
too
late,
but
anyhow,
we
have
an
issue
that
I
need
help
to
understand
how
to
to
others.
C
D
I
think
we
could
definitely
well
an
interface
is
definitely
the
right
way
to
go
and
any
function
or
method
that
receives
a
concrete
resource.
Right
now
can
just
receive
the
interface
instead
and
that's
non-breaking,
because
the
resource
concrete
class
implements
the
interface.
So
that
should
not
be
a
problem.
D
The
only
time
it
might
be.
A
problem
is
when
a
function
returns
a
concrete
resource
class,
and
we
change
that
to
be
an
interface.
Anyone
that
depended
on
internals
of
the
resource
class
would
then
be
like
the
internals
would
be
hidden
from
the
compiler,
but
that's
fine,
because
they're
internals
they're,
not
a
part
of
the
contract,
anyways,
so
I
think
that's
non-breaking
to
replace
all
of
our
usage
in
all
of
our
function.
Signatures
to
use
the
interface
instead
of
using
the
concrete
resource
class.
B
D
Yeah
I
mean
we
could
just
call
it
like
I
resource
or
something
we
haven't
done
that
in
other
places,
but
if
we
need
to,
we
can
just
to
avoid
having
the
same
name
for
both
I
think
that
that
is
a
separate
PR
from
this
one,
though.
F
So
should
I
remove
the
the
compatibility
test.
That's
breaking
the
BR
itch
test.
Is
it
I
think
it's
in
simple
span,
processor,
test.
F
F
It's
called
compatibility
if
you
want
to
search
for
it.
Yeah.
D
B
D
D
That
should
be
relatively
easy
to
fix
by
just
making
the
basic
Tracer
provider
instead
of
in
the
basic
Tracer
provider
options.
Instead
of
having
a
the
resource
key,
be
a
concrete
resource
class,
it
should
be
the
resource
interface
instead,
which
both
the
new
and
the
old
already
Implement.
So
it's
not
a
problem.
Does
that
make
sense.
C
D
D
B
A
D
F
Name
yeah,
it
makes
sense.
I
just
have
one
question,
since
the
implementation
of
the
new
interface
want
to
be
in
this
PR
should
I
should
I
like
take
this
compatibility
test
out
for
now,
and
then
add
it
after
after
the
new
update.
D
D
You
just
do
it
in
this
PR
so
that
it's
all
together
and
then
we'll
do
it
just
do
it
in
that
one
interface,
and
then
we
can
do
a
follow-up
PR
and
apply
it
everywhere
else
that
it
needs
to
be
applied.
A
All
right
sounds
good
yeah.
We
need
to
do
it
rather
rather
soon,
after
because
we
will
be
blocked
on
releases
if
we
don't.
D
Depends
so
heavily
on
each
other
that
there's
having
it
as
two
separate
PR
is
may
feel
nicer,
but
I
think
it's
not
really
providing
value.
So
I
would
just
say,
go
ahead
and
do
it
in
this
PR
all.
C
B
C
So
yes,
so
here
this
one,
so
all
of
our
detectors
were
implementing
a
detector
interface,
and
now
we
have
a
detector
sync
interface,
which
makes
sense,
but
I'm
not
sure.
If
replacing
the
existing
detectors
is
a
breaking
change
or
not
like
I
assume
it.
It
is
possible
that
someone
will
be
broken
by
this
change.
I,
don't
think
it's
very
likely,
but
I
think
it
is
possible,
and
if
it
does
then
should
we
keep
all
the
existing
detectors.
Implementing
the
old
interface
with
the
async
signature.
E
C
C
So
I
think,
if
we
want
to
be
very
strict
about
not
introducing
any
breaking
changes,
then
this
cannot
be
changed.
We
can
either
create
a
new
detector
that
implements
the
new
interface
or
just
keep
the
existing
detectors,
as
is
until
we
have
a
new
major
version.
D
Or
or
in
the
detector
sink
you
say,
detector
sync
extends
detector
and
then
you
add
a
new
method
called
like
detect
sync
or
something
like
that.
D
I
would
say:
I
probably
would
prefer
to.
C
D
I'm
going
to
create
a
milestone
for
2.0
and
start
adding
creating
issues
and
adding
them
to
the
2.0
Milestone.
Just
so
that
we
don't
forget
these
things
as
we
go,
because
I
think
when
we
get
to
the
point
that
we're
actually
releasing
the
next
major
version,
there's
already
a
lot
of
things
that
we
said,
we
want
to
do
for
cleanup
that
I
think
might
get
lost.
C
Sounds
good
okay,
and
about
from
that
this
is
really
important
and
will
improve
the
experience
of
many
users.
But
it's
really
dangerous
because
it
touches
a
lot
of
places
and
encourage
everyone
to
take
a
look
and
see
if
you
spot
anything
that
can
maybe
break
someone
or
be
incompatible.
D
D
Was
gonna
say,
I
can
manually
make
a
release
with
an
alpha
version
so
that
it
won't
get
picked
up
like
we.
We
can
release
with
like
1.10.0,
Alpha
One,
or
something
like
that,
so
that
we
can
actually
release
all
the
packages
and
try
to
update
in
a
couple
of
places.
You
know
encourage
people
to
try
it
before
we
actually
release
the
new
version.
D
Used
to
have
that
we
used
to
release
on
every
PR
merge
for
a
while.
We
haven't
done
that
in
a
while,
but
maybe
it's
worth
trying
again.
D
Yeah
I
I
have
time.
A
All
right,
thank
you,
Sami,
do
you
have
any
more
questions.
A
All
right,
thank
you,
then
does
anybody
else
have
more
topics
to
bring
up
or
something
to
talk
about?
Otherwise
we
can
move
on
to
the
back
triage
part
of
this
of
this
meeting.
D
A
Here
we
have
the
flaky
test
on
the
last
value,
aggregator
that
we
put
we
talked
about
earlier,
that's
causing
the
tests
on
the
logs
and
events,
split,
AP,
API,
split
PR
to
fail.
A
Looks
like
legendicus
already
is
having
a
look
at
the
problem.
A
A
B
A
Then
the
next
one
was
already
taking
care.
E
B
D
A
Yes,
thank
you
for
having
a
look
yeah
and
the
next
one
has
an
information
requested
label
on
it.
D
Just
waiting
for
resolution
of
a
different.
A
All
right,
and
that
is
it
for
core
I,
think
and
then
we
can
move
on
to
Country
here.
A
This
one
is
the
first
in
the
list,
the
user
interaction
instrumentation
I'm
not
too
familiar
with,
creates
too
many
traces.
D
This
is
where
I
really
wish.
We
still
had
Bart
around
yeah.
C
I
played
with
the
instrumentation
a
bit
lately
and
he
does
like
every
time
he
will
click
or
something
you
have
an
event
on
an
HTML
element.
Then
it
triggers,
but
I
think
the
solution
would
be
to
create
a
simpler
to
to
sample
out.
What's
not
interesting,
I'm,
not
sure
if
the
sampler
has
enough
data
to
make
a
decision,
but
I
think
this
is
the
way
to
Google.
B
I
think
eventually,
this
will
be
using
the
new
event
emitter
stuff,
because
this
sort
of
thing
is
really
packing
spans
to
yeah
pretend
to
be
an
event,
because
that's
all
that
was
available.
So
yeah
I.
D
D
B
We
shall
have
a
broken
discussion
about
session
ID
and
but
effectively
everything
will
have
the
same
session
ID
for
the
user.
So
that's
how
they
currently
would
get
correlated
today,
but
we
still
have
an
open
PR
from
Ted
about
effectively
having
mutable
resources,
so
that
has
to
get
resolved.
D
B
Unlikewise
a
span
would
would
be
tagged
with
the
same
session
ID
as
well.
So
you
could
say
I
want
to
look
at
everything
that
this
user
did
and
you
group
them,
rather
than
having
a
parent
span
having
the
same
crease
idea
across
everything.
That
would
never
get
closed,
which
would
be
the
other
way
to
do
it.
D
B
But
not
directly,
no
except
time
the
instrumentation
could,
in
theory,
go
okay,
I
want
to
create
a
span
for
this
and
therefore
they
would
get
correlated.
But
a
lot
of
these
events
don't
need
this
name.
D
Yeah
I'm
pretty
sure
that's
the
way
that
most
existing
roam
Solutions
work.
Anyways
is
just
time
correlation,
yeah,.
B
Time
correlation-
and
you
know
the
session
ID
or
impression
ID
I,
think,
is
what
Google
calls
it
so
yeah.
It's
not
good
questions.
Look
at
anyway,
you
get
what
I'm
talking
about
yeah.
D
D
It
says,
instead
of
only
four
does
it
say
like.
Is
this
a
result
of
a
recent
change
or
something
that
he
used
to
be
four
and
now
he's
getting
50
to
80.
A
Yeah
our
formulate
a
question
with
those
two
points
at
the
end
of
the
meeting
but
yeah
sampling
sampling
is,
is
probably
a
good
idea.
But
one
question
that
came
up
in
my
mind
is
to
is:
is
that
component
owned
by
anybody
at
the
moment
in
country
but
I,
don't
recall,
instrumentation.
D
He
does
comment
every
now
and
then,
but
he's
not
very
active
on
it.
Maybe
tag
him
in
the
comments
see
if
he
has
time
to
look
at
it
all.
D
D
On
this
example,
so
I
I
his
example.
Is
he
clicks
one
time
and
sees
four
traces?
It
looks
like
the
last
note.
Is
they
tried
it
in
a
production,
application
and
they're,
seeing
50
to
80
traces,
so
the
example
is
going
to
show
four
traces
where
he
expects
one
and
is
just
like
a
minimal
example
of
what
he's
saying
so
I'd
like
to
see
what
are
the
actual
four
traces
and
are
they
actually
spans
instead
of
traces?
So.
D
The
example
that
he
gave
us
and
see
or
if
you
want
you
can
just
ask,
can
you
please
post
a
screenshot
of
Jaeger
or
something
showing
before?
If
you
don't
have
time
to
run
the
example
or
something
like
that.
A
Yeah
I
will
I
will
ask
them
to
to
post
a
screenshot
of
them
and
then
I
guess
we
can.
We
can
follow
up
on
that
afterwards.
B
A
All
right,
then,
moving
on
yeah.
A
Wondering
if,
for
the
information
request
that
we
got
anything
back
no,
this
does
not
seem
like
we
got
an
answer
to
that.
Let's
just
same
issue
on
my
side,
so
I
guess
that
is
also
okay.
To
leave
at
that
right.
A
Have
PRS
or
are
already
assigned
this
is.
A
Stablebot
would
have
taken
care
of
it
anyway,
but
all
right
that
should
be
it
then
yeah.
D
A
You
thank
you,
everybody
for
your
time
and
see
you
next
week.