►
From YouTube: 2022-07-14 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
Yeah
so
on
this,
so
the
only
question
I
think
remaining
on
this
one
is:
do
we
want
to
merge?
A
Do
we
want
to
also
do
like
package
name,
artifact
name
in
at
the
same
time,
or
I
mean
in
the
same
release?
So
if
we
merge
that
and
do
this
in
the
same
release
or
do
we
care
and
merge
it
now.
A
And
yeah,
so
what
do
you?
Are
you
all
right
with
this
kind
of
proposal?
It's
definitely
not
ideal
right,
like
I
think
I
think
what
I
was
understanding.
Your
point
was
kind
of
similar.
We
had
similar
discussion
around
the
sdk
going
like
with
the
metrics
api.
B
A
A
A
Yeah,
I
know
about
this
much
about
java
modules,
so
I
get
easily
lost
in
that.
B
B
B
I
guess
like
multiple,
like
frameworks,
had
similar
issues,
but
I'm
not
sure
whether,
like
the
module
system.
C
A
B
Yeah
we
have
had
something
it
might
be
that
using
the
modules
is
a
prerequisite
for
using
k-link
yeah.
It
should
be
so
that,
even
if
your
jar
files
don't
have
the
module
info,
then
they
get
automatically
converted
into
modules.
B
B
Like
that,
those
jar
files
need
to
like
they
need
to
be
included
in
the
application,
because
if
they
are
inside
the
agent
then
like
there
is
no
problem.
Obviously,.
B
It
might
be
easier
to
incubate
like
incubate
the
stuff
inside
the
instrumentation
api
jar
and
just
annotate
the
experimental
stuff
as
such.
But
I
guess.
A
Also,
I
don't
recall
if
there
was
really
a
problem
with
it
as
much
as
a
preference
to
keep
a
clear
line
between
the
stable
and
unstable
modules
and
not
leak
sort
of
like,
I
think
maybe
it
was
probably
probably
based
on
experience
with
guava
on
the
and
how
many
people
depended
on
those
experimental
apis
just
because
they
were
available.
A
Which
is
part
of
what
I
was
thinking
about
this
semcon
package.
I
mean
the
thing
that
I
like
about
putting
it
under
or
calling
it
an
incubator.
A
A
And
so
a
wrap
for
the
java
9
issue.
We
just
wouldn't
include
a
java
9
module
module
info
in
this
artifact
so
that.
B
A
A
Yeah,
unless
we
go
with
the
incubator
package
name
to
make
them
different.
A
A
Oh,
we
want
to
stabilize
we're
planning
to
stabilize
the
instrumentation
api
soon,
but
we
can't
stabilize
these
classes
like,
for
example,
http
client
attributes
getter.
A
These
are
dependent
on
the
open,
telemetry
semantic
conventions
for
http
we've
kind
of
modeled
them
after
the
utes
in
the
semantic
conventions
and
those
semantic
inventions,
aren't
stable,
haven't
been
marked
stable
yet,
and
so
they
could
still
change,
and
then
we
would
be
stuck
with
like
say
they
change
and
want
to
capture
instead
of
url.
They
want
to
capture
it
in
components
like
scheme
host
path.
A
B
B
I
guess
like
we
probably
need
to
experiment
and
see
like
does
it
work
like.
If
the
only
problem
is
with
this
pair
service
attribute
extractor,
then
then
I
think
it
definitely
would
make
sense
to
keep
the
package
names
and
maybe
like
find
another
another
home
for
this
single
class.
If
it
turns
out
to
be
so.
A
And
I'm
thinking
of
like
that,
your
pr
lorry
for
the
the
context,
the
span
client
span,
mutator
oops,
we
lost
laurie.
A
C
B
A
So
maybe
in
that
case
like
in
your
this,
because
this
would
then
cause
a
problem,
but
maybe
over
maybe
for
some
things.
We
also
have
an
incubator
package
for,
but
we
could
say,
okay
for
initial.
B
I
think
the
only
problem
with
it
is
that
I
think
I
need
some.
I
need
to
instrument
something
like
this
class
needs
to
be
instrumented
and
because
of
that,
if
the
api
like,
if
the
api
and
agent
don't
match,
then
the
instrumentation
won't
work.
B
In
that
sense,
like
maybe
it
would
require
some
kind
of
internal
package
here,
like
I
don't
know
some
like
some
place
that
doesn't
get
renamed
where
to
put
the
class
that's
going
to
be
instrumented.
B
Yeah
one
of
the
problems
with
this
thing
is
that,
like
the
instrumentation
part,
is
a
bit
bit
of
magic
and
from
the
end
user's
perspective,
there
is
no
way
they
are
going
to
figure
out
why
this
thing
breaks
if
it
breaks.
A
A
So
I'm
I
don't,
I'm
currently
not
expecting
this
to
live
into
stability.
I
think
it's
a
you
know
a
great
experiment
bridge
until
we
have
that
unblocking
users,
but
I
have
a
feeling
it's
going
to
be
so
I
wouldn't
necessarily
mind
throwing
it
under
incubator.
B
A
Okay,
I
will
kind
of
summarize
some
of
the
options
here
and
we'll
get
the
materials.
A
I
can
ask
this
evening:
I
think
jack
is
still
out
on
vacation,
so
I'm
not
sure
and
honorag
is
out
with
kovid.
So
I'm
not
sure
I'll
get
much
feedback.
A
As
far
as
the
release
was
this
something
that.
A
Let's
look
at
other
things,
yeah,
so
I'll,
probably
ping
in
the
s
in
slack
ping
in
the
our
channel
and
ask
matthias
for
just
what
his
thoughts.
Oh,
I
need
to
finish
reviewing
this,
but
he's
got
several
approved
prs
that
I'm
assuming
we
should
just
merge
for
the
release
and.
A
Oh,
yes,
I
didn't
review
this,
but
thank
you
for
yeah.
This
was
a
very
mis.
This
caused
me
a
lot
of
headache,
confusion.
This
issue.
B
B
B
But
yeah
I'm
not
completely
sure
like
why
they
are
doing
like
this
thing
like,
but
that's
the
last
elastic
agent
also
seems
to
shortcut
it
and.
A
Yeah,
I
feel
pretty
solid
about
the
all
the
tests,
but
if
the
tests
pass
then
I'm
comfortable
with
something
like
that
cool,
so
I
will
get
this
merged
before.
I
think
this
would
be
a
good
one
to
add
to
the
release.
A
So
this
one
we
so
it
kind
of
depends
on.
If
I
make
the
release
well,
let's
see
what
mateish
has
to
say
in
chat
and
we
don't
have
to
release
this
week.
We
could
release
next
week
also,
if
there's
remaining
things.
B
I
think
wasn't
there
something
like
stabilizing
the
instrumentation
api.
A
A
Oh
yes,
where
did
we
do
we
put
that
right
on
the
discussion.
A
A
Let's
see,
do
I
still
have
it
pinned,
I
probably
still
have
a
pin
over
yes,
yeah
I'll,
have
a
pin
also.
A
C
Yeah,
I
have
a
couple
of
on
issues,
so
this
is
actually
a
new
well.
My
first
question
is
not
actually
new,
so
a
while
back.
C
So
I
don't
remember
exactly
which
class
it
was,
but
we
made
it
package
friendly.
You
know,
for
for
the
users
not
to
be
able
to
to
to
use
their
own
config
builder,
I
so
the
problem.
C
So
I
I
right
now
I'm
creating
kind
of
a
bridge
between
the
standard
configurations
that
we
do
on
carcass
and
the
open
telemetry
classes
and
then
there's
so
we
we
have
a
kind
of
a
namespace
for
all
configurations
that
start
with
quercus.
Even
if,
after
that,
it's
we
are
going
to
use
the
same
namespace
that
opendomg
has
so
and
because
open
telemetry
loads,
the
the
configurations
from
system
properties.
If
people
will
use
directly
the
the
system,
property
name
default
on
open
telemetry.
C
C
I
think
it
was
roberto
that
added
it
roberto
cortez,
so
it
should
be
on
under
roth
cortez
air
ad.
All
right
got
this.
A
Nope
I
had
too
many
here
we
go
yeah
yeah
this
one.
A
A
C
A
See
if
this
works
okay,
if
not
I
mean
I'm
personally
supportive,
I
just
because
I
I
like
kind
of
similarly
want
to
have
like
full
control
in
of
the
configuration
story
and
not.
A
C
Okay,
let
me
see
what
what
I
can
do
with
this,
so
there's
so
on
another.
C
A
That
doesn't
work
then
I
would
say
you
know
open
a
open,
a
new
issue
in
the
sdk
repo
okay,
to
try
to
get
attention
and
maybe.
C
C
And
on
microprofile,
there's
still
a
bit
of
a
fuzz
around
the
hotel
experimental
enable
override
enable
disable
property.
So
there's
there's
a
a
an
issue.
A
C
A
C
A
C
C
And
they
wanted
to
create
a
micro
property
for
that
which
I
am
personally
well.
C
Well,
not
me
personally,
but
here
at
reddit
we
are
kind
of
not
agreeing
on
that,
because
we
don't
want
to
to
create
properties
outside
of
the
open,
telemetry
defined
ones,
because
it
will
create
kind
of
a
problem
for
the
user
so
which
properties
live
where
and
who
manages
them.
And
but
it
causes
all
kinds
of
problems
for
them.
C
And
so
I
I
was
successful
in
trying
to
at
least
postpone
the
discussion
and
go
back
here
and
try
to
see
if
this
experimental
is
here
will
be
here
for
a
very
long
time
or
if
it's
something
that
it's
going
to
go
away
soon.
A
So
I
thought
john
mentioned:
he
thought
there
was
existing
work
going
on.
I
was
looking.
I
thought
there
was
an
issue
in
the
spec
repo
also,
but
I
spent
some
time
looking
and
I
could
not
find
it
so.
A
A
And
describe
what
that
would
do
and
that
you
know
it
could
be:
maybe
some
languages
don't
want
to
implement
it,
but
those
that
do
could
use
this
standard
property
name
and
then,
if
you
can
get
sort
of
initial
positive
reaction
on
the
issue
and
for
example,
that
could
be
just
the
you
know
getting
some
java
folks,
you
know
I'd
be
happy
to
thumbs
up
that
and
then
you
can
move
to
submitting
a
pr
into
the
spec
repo
for
the
issue.
C
A
So
I
would
say
open
the
issue
you
know
give
it
like
a
week,
maybe
if
you're
not
getting
people's
attention
ping
in
the
in
the
java
channel,
for
example,
that
since
you're
coming
from
the
java
perspective,
ask
people
to
you
know,
take
a
look
at
it
and
then
there's
on
there's
a
spec
sig
meeting
on
tuesdays
at
8
a.m.
Pacific
time,
so
that
is
kind
of
the
I
don't
want
to
say
last
resort,
but
a
lot
of
times.
A
Spec
issues
don't
get
attention,
and
so
that
is
a
place
to
go
to
physically
get
people's
attention
and
be
like
hey.
We
have
you
know
we
from
red
hat
and
the
micro
profile.
Org
really
want
this.
Can
you
know
how
do
we
move
it
forward?.
C
We
have
the
so
the
guys
at
the
microprofile
they
they
want
to,
for
some
reasons
that
I
can
tell
you
about.
They
want
to
independently
enable
and
disable
parts
of
the
spec
parts
parts
of
the
sdk,
so
they
want
to
enable
and
disable
tracing
metrics
and
logs,
and
this
is
due
to
the
particular
way
they
the
apis
are
designed,
which
are
containerized.
C
So
the
tracing
spec
is
independent
from
the
matrix
one
and
and
they
might
not
agree
to
the
same
standards
so,
for
example,
the
matrix
guys
they
they
they
are
still
discussing
and
implementing
their
own
view
on
how
to
do
metrics
and
might
not
align
exactly
with
open,
telemetry
or
micrometer
or
something
else.
But
it's
but
it's
their
problem,
but
on
the
tracing
we
are
basically
doing
transparent
use
of
open,
telemetry
tracing.
C
Yes,
yeah,
it's
it's
whatever.
We
need
to
do
it's
kind
of
wrapping
it
in
a
way
that
it's
not
trapping
it.
Even
it's
like
defining
how
micro
profile
aware
applications
should
configure
open
telemetry
so
that
so
that
the
microphone
config
can
be
used
transparently
to
configure
the
open,
telemetry
properties.
C
C
And
and
no
nobody
knows
how
logging
will
go,
but
this
leads
us
to
to
a
place
where
we
might
have
on
a
particular
application
tracing
working,
the
top
metrics
on
the
open
telemetry
side.
So
it
would
be
handy
to
have
switches
like
with
the
enable
for
the
sdk
for
each
one
of
these
and
looking
at
the
code,
it
seems
that
it's
fairly
simple,
because
the
the
enable
switch
what
it
does.
It's
basically
creates
a
an
sdk
with
default
configurations
and
no
processors
or
propagators.
A
I
see
what
you're
saying
beyond
just
the
the
auto
configure.
A
A
A
A
A
C
Okay,
let
me
check
with
them
if
they
are
okay
with
this,
if
they
are
well
no
work,
is
it.