►
From YouTube: 2022-09-23 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
B
Good
good
to
see
you
over,
hopefully
the
the
illness
there
yeah
all
right,
good
I
know
sometimes
the
other
candidating
effects
there
anyway.
B
B
B
Okay,
let
me
go
ahead
and
start
and
do
what
we
can
so
2773
was
previously
rejected
one
last
week
we
left
it
open
because
there
was
still
a
bit
of
activity
going
on
and
and
tigran
was
going
to
was
engaged
in
some
of
the
conversation
there.
B
Okay,
it
looks
like
they,
they
continue
conversation
up
to
you
know
within
the
last
week
or
so,
and
then
I
think.
The
last
comment
here
was
looks
like
about
six
or
seven
days
ago.
B
B
A
Yeah,
because
the
issue
I'm
saying
is
the
the
products
and
systems
mentally
convention
they're
getting
bigger
and
bigger,
and
we
think
have
a
low
quality
bar
like
as
long
as
as
it
makes
Generations
volatility.
But
nobody
is
doing
the
Deep
study
to
see
whether,
like
we'll
be
able
to
make
it
stable,
or
it's
just
some
like
random
thoughts
from
someone
who's
in
experienced.
So
I
I
still
probably
need
to
do
something
about
it.
A
Like
if
you
look
at
the
unit,
many
of
the
things
like
in
the
current
PR
like
start
time,
those
things
the
the
Precision
is
in
seconds-
and
this
is
the
integer
I'm,
not
sure
if
it's
helpful
or
later
we'll
realize
this
is
a
mistake,
so
I
I
feel
the
current
way
will
allow
this
thing
to
be
merged.
Is
it
like
we
just
close
one
of
our
eyes
and
assume
that
later,
we'll
polish,
all
of
them,
but
who
will
publish
that.
C
Yeah,
that's
fine!
Okay,
so
I
yeah
I
see
some
I,
don't
know
whether
it's
the
exact
thing
that
exists
for
system-
and
this
is
just
applying
the
system
once
to
metrics.
C
A
The
issue
with
like
we're
doing
a
challenge,
because
we
want
to
give
the
signal
better.
The
issue
itself
makes
sense
or
not,
and
whether
we
have
capacity
to
review
the
pr
right.
So
there
are
many
issues
that
we
reject
as
not
because
they
don't
make
sense.
It's
because
we
don't
have
capacity
or
we
don't
think
we
have
experts
here.
C
Okay,
you're,
you
were
saying
something
about
the
type
correct
that
the
type
doesn't
make
sense.
A
My
my
concern
is
given
large
storage
is
going
to
like
start
this
overall
semantic
convention
thing
just
to
lay
out
the
foundation.
I'm
not
sure
like
if
having
all
this
like
random
things
merged
in
the
current
spec
is
a
good
thing
for
us
or
not
like
later.
Someone
has
to
do
the
heavy
lifting
work
by
reviewing
every
single
thing
and
make
changes
to
make
sure
they're
correct.
C
A
A
When
I
look
at
the
current
processing
communication-
it's
horrible
it's
already,
someone
has
to
do
the
cleanup
and
I
figure
it's
something
like
like.
If
I
I'm
going
to
do
it,
I'll
probably
have
to
rewrite
the
entire
thing.
I
I
want
to
avoid
another
situation
where
we
have
the
Matrix.
Back
and
later
we
have
to
delete
everything.
I
start
from
scratch
for
this
one
I'm
not
sure
like
like
how.
A
B
I
would
also
ask
just
as
a
precursor
this
particular
individual
has
about
three
or
four
other
items
that
he's
added
similar
to
this
in
this
past
week,
where
he's
asking
about
adding
a
different
metric
and
so
forth,
yeah.
A
B
C
But
honestly,
it's
very
I
think
that
it's
totally
expected
at
some
point.
It
was
expected
at
some
point.
Somebody
will
come
to
us
and
start
adding
things.
The
problem
is
that
they're,
basically
those
things
on
things
that
are
broken,
so
somebody
should
actually
do
a
review
of
the
existing
metrics
yeah.
C
That's
a
good
one.
If
there's
already
a
group
I
didn't
know
sorry,
it
wasn't
a
word
that
you
are
meeting
with
just
suret.
Let's
say:
hey,
let's
hold
it,
let's
put
it
on
pause,
please
come
to
the
meetings
and
then
we
decide
whether
we
keep
your
PR
or
we
just
describe,
scrap
it
or
something.
You
know.
C
And
then,
in
that
case
we
could
use
reference.
It's
like
hey
I
will
like.
We
would
like
to
ask
you
to
to
pause
for
now
and
probably
come
to
this
meeting,
that
we
are
having
this
other
meeting
we
are
having,
because
we
are
actually
reviewing
the
existing
stuff
and
we're
thinking
about
changing
that.
C
Double
one:
okay,
I
only
have
I
I
only
have
seen
instrumental
stability,
not
the
plus
semantic
convention.
Working
group,
uh-huh.
A
Yeah,
so
it's
trying
to
lay
out
the
foundation,
for
example
like
like
which
one
should
we
focus
on
and
if
we
cannot
even
get
one
single
semantic
convention
stable.
Should
we
start
to
like
just
open
The
Floodgate
for
all
the
semantic
conversions,
and
should
we
do
this
in
a
way
that
we
can
cover
both
traces
or
metrics,
or
we
just
close
eyes
and
make
some
random
progress
on
cases
with
experimental
State
and
later,
when
we
do
metrics,
we
realize
everything
is
wrong.
A
A
B
A
C
Yeah
yeah
yeah,
maybe
I
think
that
yeah,
it's
really
what
you
put
there
is
fine.
Probably
we
just
need
to
explain
why
you
know
like
please
pause
for
for
now,
as
we
are
thinking
of
refactoring
the
exist,
some
of
the
existing
metrics
com
and
please
come
to
ongoing
discussions.
C
C
B
C
B
D
B
All
right
and
except
as
far
as
triage
date,
there's
we
don't
have
a
triage
state
for
this
right,
so
I
just
leave
it
I'm
gonna!
Leave
it
like
this.
A
D
This
is
a
debate.
I've
already
responded
on
I'm
watching
the
screen
now,
I'm,
not
sure
Tyler's
responded
since
his
position.
D
Just
didn't
seem
to
confused
I'm,
not
sure
this
should
be
discussed
in
this
Tuesday
meeting.
C
Okay,
so
maybe
we
can
just
tell
him:
let's
discuss
these.
The
next
spectacle.
C
B
B
B
B
A
A
C
I,
don't
remember
exactly
what
the
details,
what
I
think
I
remember
they
mentioning
The
Proposal
that
they
could
be
happy
to
adjust
what
they
have
too
much
apparel
as
long
as
it
makes
sense.
You
know
yeah
yeah,
so
so
in
theory,
we
would
have
like
a
new
hotel
Convention
that
they
could
be
jumping
from
too
yeah.
But
that's
what
I
kind
of
remember
exactly.
A
I
have
the
same
memory,
but
this
is
four
things
that
we
already
have
and
it's
something
relatively
in
shape.
My
question
is
for
things
that
we
we
don't
have
yet
hi.
Specifically,
we
want
to
invent
another
thing
given
ECS
has
been
running
for
multiple
years
and
this
seems
have
a
good
definition
already
so
like
are
we
trying
to
like
give
birth
to
a
new
child
and
then
try
to
do
something
like
Bridge
or
maybe
at
some
point
we'll
just
realize?
Oh,
that
that's
dumb.
We
should
stop.
C
C
Okay,
yeah
so
about
this
one.
This
is
an
issue
that
I
feel
based
from
the
past.
See
this
seek,
you
know,
call
I,
don't
know.
What's
the
name
of
that
guy
but
pedareth,
he
was
trying
to
you
know
to
us
that
we
actually
fail.
We
configure
tsdks
through
environment
variables,
and
some
of
them
need
like
like
we.
They
are
really
required
and
if
you
don't
sell
them
or
they
have
invalid
values
that
they
fail
and,
of
course
this
will
be
a
breaking
change
because
at
the
moment,
all
the
environment
variables.
C
If,
if
the
value
is
not
properly
set,
they
just
generate
a
warning.
So
take
your
admission
by
the
way
that
this
would
be
a
freaking
change.
You
know
if
we
allow
that,
but
one
workaround
or
kind
of
work
around
is
that
we
Define
a
devil
or
failed
fail
fast,
whatever
most
so
when
you
enable
that
any
actual
error
in
development
variables
or
some
of
those
errors
will
actually
stop
the
SDK
initially
session.
A
And
what
about
the
apis
like
in
primary
is
the
API
package
and
they
did
something
wrong
if
you
don't
really
pass
in
Parable
value
like
a
negative
value
for
explicit
bucket
histogram.
C
So
yeah,
so
at
the
moment,
this
is
only
for
environment
variables
and
environment.
Variables
are
only
used
by
the
SDK
at
this
very
moment.
So
at
this
very
moment,
if
this
would
only
include
the
sdks
I
wonder
whether
this
should
be
in
the
future
extended
to
also
API
calls,
you
know
would
fail
as
well.
I
don't
know,
but
at
least
this
is
for
for
SDK.
Only.
A
Maybe
I
haven't
spent
much
time
thinking
about
this,
but
I
I'm
a
bit
worried
about.
Is
this
going
to
add
Clarity,
or
this
is
going
to
make
it
too
complex?
Because
I
can
imagine
some
folks
might
say:
hey
I
want
to
have
a
failure.
Like
just
crash
my
applications,
I
I,
have
a
downfall
cause
cause
that
if
I
made
any
mistake
to
the
IDK,
but
for
this
OTL
PX
powder
I
don't
want
to
have
that
or
they
might
say.
Okay
for
this,
maybe.
A
But
for
the
SDK
don't
turn
that
on
so
they
can
try
to
stop
it
down
and
then
we'll
have
multiple
environment
variables
or
will
be
making
it
very
complex,
maybe
some
flag,
maybe
like
which
component
to
fill
and
also
the
definition
of
what
should
we
considered
as
a
failure
like
if
they
send
some
negative
value
to
a
histogram
or
like
positive
Infinity?
Is
that
a
filler
or
not
be
honest,
yeah
I
I
feel
it's
very
hard
to
Define
yeah.
C
A
D
A
A
C
Yeah
I
mean
so.
The
thing
is
that
for
Java,
for
if
you
use
the
Java
SDK
and
you
still
have
to
go
for
years
through
environment
variables,
then
judge
you
could
provide
something
like
go
or
talkative
or
the
agent.
That's
a
novel,
then
again,
would
it
be
bad
that
only
components
that
require
serial
code
changes
would
use
this.
A
So
for
the
agent
imagine
there
is
a
like
open
country
called
Auto
implementation.
I
I
think
what
the
call
Auto
implementation
should
do.
Is
it
take
the
the
climaxd
version
of
the
SDK
and
put
a
callback
and
they
can
do
whatever
in
the
Callback
to
fill
the
application,
so
that
SD
had
been
happy
at
all,
and
the
alternation
can
tell
the
user
how
they
want
to
enable
the
faster
filter
they
can
tell
the
user
go
and
use
this
environment
variable
or
right
configuration.
It
doesn't
have
to
be
the
go
SDK
problem.
C
C
That,
okay,
that
that's
a
good
point.
So
let's
write
down
or
actually
can
write
mentioning
the
Java,
the
Java
agent
case
and
yeah,
but
basically
yeah
and.
A
Another
minor
issue
I
I
can
see
is
the
the
terminology.
Debug
mode
is
very
convenient
because
some
language
that
sdks
they
shape
two
different
versions
of
the
package.
One
is
the
release
version,
which
is
highly
optimized.
Another
is
a
debug
version
which
you
can
always
relate
to
the
actual
source
code,
and
this
debug
mode
seems
to
be
a
feel
fast
mode
or
something
so
yeah.
A
My
model
is,
will
end
up
with
like
many
different
combinations
like
if
imagine
like,
if
we're
doing,
C
plus
parts,
we
release
the
static
library
for
production,
but
we
even
have
this
mode
enabled
in
production,
because
this
fills
us
in
production.
We
shouldn't
print
that
on
right,
so
my
worries
will
end
up
with
people
asking
hey.
You
should
have
this
High
default
turned
on.
If
you
release
a
debug
build,
if
you
have
a
release
build,
then
this
should
be
default
out
for
something.
C
Yeah
yeah,
the
main
part
was
already
mentioned
there,
because
even
in
Java,
it's
affected
us
because
they
already
have
the
debug
mode.
C
But
anyway,
really,
since
you
have
a
little
bit
more
context,
can
you
add
yourself
the
comment
there
and
saying
that
if
Java
wants,
for
example,
the
Java
engine
want
to
support
that
they
can
do
that
through
their
own
environment
variable,
but
that
could
be
technically
an
implementation
detail
and
then
say:
let's
continue
the
discussion
on
the
spec
call
or
over
here
you
know
see
what
what
what
everybody
does
and
then
I
could
say
because
of
the
current
discussion,
we're
having
here.
I
would
say
it
needs
more
info.
D
To
drop
but
I
feel,
like
we've
used
the
term
strict
mode
in
the
past.
Just
for
the
record.
Yes,
that's
another
option
like
when
you
have
instrument
conflicts,
yeah.