►
From YouTube: 2022-10-21 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
B
It's
been
better
I
woke
up
and
like
almost
immediately
tweaked,
my
back
picking
something
up,
so
those
are
not
fun.
I've
already
been
to
the
chiropractor
and
everything,
but.
A
B
A
So
I
don't
know
if
you
saw
the
projects
that
I
created
Yesterday
by
chance.
A
A
Yeah
I
think
this
is
one
so
we've
got.
This
is
the
main
project
here,
but
I've
also
created
this
metrics
API
GA
project.
That's
where
all
those
issues
came
from,
and
this
is
the
one
that
would
track
the
GA
of
metrics.
A
It's
essentially
all
just
the
verify
stuff.
It
looks
really
similar
to
what
we
did
for
the
trace.
Api
then
yeah.
Do
you
want
to
talk
about
that?
One.
B
B
Yeah
yeah
yeah
project
there
we
go.
B
Here
we
go
metrics
GA,
posterior.
A
B
Yeah,
so
we
yeah
gonna
have
to
Ping
somebody
about
that.
It's
been
open.
A
A
I,
don't
anticipate
it
getting
traction
which
is
frustrating
if
you
can't
share
it
internally
or
outside
of
open
Telemetry
users.
But
one
thing
that
was
interesting
is
like
yeah
I
can't
imagine.
It
is
right
by
default,
so
I've
changed
these
to
be
read
by
default.
So
if
you're
not
in
the
go
Sig,
essentially
you
can't
modify
this,
which
I
think
is
the
right
approach.
B
A
I
created
those
two
other
projects
or
three
of
the
projects.
I
can't
remember:
I
showed
this
ga1.
There's
the
metrics
SDK
beta,
there's
the
metrics
API
GA,
that's
the
one
I
showed
yeah.
This
is
the
one
where
I'm
kind
of
like
it
was
essentially
the
same
as
the
metrics
API.
A
Ga
this
hasn't
been
broken
apart
into
like
multiple
tasks,
but
essentially
it's
just
tracking
it
and
I
created
it,
because
this
is
I,
think
the
one
that
I
wanted
to
just
put
it
as
the
bucket
for
all
the
things
we
would
do
after
ga.
A
So
I
created
the
other
one.
It's
it's
more
of
a
placeholder.
We
can
fill
it
out
later,
but
that
was
kind
of
what
I
did
yesterday.
B
Cool
yeah,
let's
see,
can
I
rearrange
these,
of
course,
not
just,
however,
GitHub
decides
to
sort
them.
B
No,
the
the
projects,
okay,.
B
A
B
Newest
is
what
mine
was
showing
up
as
yeah,
so
I
will
share
my
screen.
Real
quick
here.
Let's
see
share,
you
should
be
seeing
the.
B
What
should
we
call
it?
The
the
Sig
dock
reminder
to
go
put
your
name
on
the
attendees
list,
just
so
that
we
know
we're
here
and
if
you
want
to
talk
about
a
specific
issue
before
we
just
work,
the
backlog
put
it
there.
B
That's
what
that
space
is
for
other
than
that.
I
was
just
going
to
work.
These
new
statuses
in
order
kind
of
just
do
a
reef
review
of,
what's
already
on
the
board
in
case,
there's
anything
that
we
think
might
be
mislabeled
misplaced
and
then
that
that's
it
that's
that's
the
game
game
plan,
any
objections.
A
No
I
think
this
I'll
just
move
this
here
so
about
the
exemplars
and
the
exponential
histograms.
Could
we
talk
about
those
sure.
B
Which
one
do
you
want
to
go
with?
First.
A
B
A
A
Supporters
on
the
left,
as
well
or
the
no
status
yeah,
which
I
don't
know,
maybe
we
should
be
actually
closing
one
in
favor
of
the
other,
but
because
I
guess
they
are
kind
of
duplicated.
At
this
point,.
B
A
Go
ahead
but
I
think
I
added
a
comments
yesterday
in
a
segmenting.
A
B
What
I
would
wonder
like
I
know
we
we
you,
you
talked
about
the
changes
to
comparability.
B
A
Then
also,
maybe
like
views
need
to
be
updated
to
they
can
enable
a
disable
based
on
a
certain
matching
criteria,
but
I
I
think
that's
not
even
a
part
of
the
specification.
So
maybe
that's
not
even
something
we
would
do.
Yeah
I
was
gonna
say:
do
we
really
want
to
be
able
to
turn
that
off
or
should
we
just
always
be
capturing
exemplarcs.
A
Our
first
draft
could
just
be
always
on
right
and
then
a
second
feature
release
could
include
a
turn
this
off.
If
users
are.
B
Yeah,
like
I,
would
wonder
if
there
are
performance
implications
because
of
the
just
stuff
and
if
somebody
might
want
to
turn
that
off,
because
they're
trying
to
run
a
very
lean.
A
A
What
that
sampling
is
not
the
right
word,
but
let's
just
go
with
sampling.
Today,
that's
sampling
rate
for
those
exemplars
would
be
like
I
think.
If
that's
something
that
you
would
probably
want
to
be
tuned
right,
yeah.
B
The
selection
criteria,
yeah
yeah,
also
a
very
good
point,
especially
for
so
like
that
I,
would
think,
would
be
another
potential
option
that
we
would
have
to
to
include.
A
A
A
B
A
A
But
yeah,
so
it
looks
like
the
option
would
take
these
like
Exemplar
filter
or
example
or
Reservoir
parameters,
and
then
these
configure
the
sampling
rate
I,
don't
know
what
the
reservoir
is.
It
sounds
like
some
sort
of
storage
mechanism,
but
I
thought
the
storage
mechanism
would
be
on
the
exported
measurements,
so
I'm
kind
of
confused
off.
B
I
believe
that's
for
performance
reasons
in
like
Java.
They
want
to
be
able
to
reserve
a
block
of
memory
to
put
the
reservoirs
so
similar
to
like
a
pool,
a
go
pool.
A
Okay,
yes,
I
mean
I.
If
that's
the
case,
then
yeah
it
sounds
like
we
would
need
to
Define
these
interfaces,
which
would
be
backwards
compatible
and
then
accept
them
as
options
to
the
meter
provider
yeah
or
as
an
option
in
the
meter
provider
yeah,
all
of
which
sounds
backwards
compatible
to
me
true
stuff.
B
B
B
A
B
That
was,
okay,
I.
Think
we've
do
we
want
do
we
want
to
make
this
a
decision
Point
here.
Do
we
just
want
to
say
we
think
where
we're
at
right
now
we
can
do
this
in
a
backwards
compatible
way.
Punt
this
this
specific
issue
and
the
Prometheus
one
into
the
post,
GA
bucket.
A
A
B
B
I
I
agree:
let's
see
you
would
take
that
one
off
and
put
on
post
GA
put
it
on
that
one,
and
then,
if
I
do
this
and
hit
refresh,
oh
it
did
man.
This
is
better
than
jira
it
automatically
refreshes
yeah.
B
A
B
A
So
this
is
one
I'm
a
little
bit
more
like
concerned
about,
because
I
know
that
we're
gonna
add
probably
a
new
type
to
like
the
metric
data
package,
but
I
know.
Josh
is
also
really
wanting
to
get
some
like
base
computation
package
into
the
project
so
that
he
can
use
it
in
The
Collector.
If
I'm
not
mistaken,
right,
Aaron.
B
I
think
we've
stalled
long
enough
that
he
put
it
into
a
different
project
for
right
now,
I'll
be
honest.
I
kinda
am
unsure
why
this
project
should
be
the
host
of
that
code.
A
I
honestly
feel
like
there
needs
to
be
a
new
repository
in
the
open,
Telemetry
organization,
but
we
seem
to
be
very
averse
to
prod
or
repositories
in.
B
I
think
right
now
it's
in
like
a
light
step
specific
project.
Okay,
I'd
have
to
I'd
have
to
go
digging
for
it.
A
Well,
I
mean,
if
that's
the
case,
then
I'm
less
motivated
to
include
it
in
the
the
Beta
release.
I
guess
because
he's
got
a
solution
going
forward:
yeah
yeah
this.
B
B
Yeah,
though
I
mean
to
be
fair,
it,
it
hasn't
changed
in
the
same
amount
of
time.
Right,
like
I,
think
there's
just
a
lack
of
willingness
to
accept
it.
Yeah.
A
Yeah
and
I
I
think
it
also
if
it
did
change
like
you'd,
see
like
what
was
that
thing
was
like
a
base
two
offset.
So
it
starts
at
zero
instead
of
like
one.
B
A
Something
like
that
yeah
so,
like
small
implementation
details,
not
necessarily
API
changes
but
and
I
I
think
they've,
already
hashed
out
all
of
those
with
Prometheus
yeah,
so
I
don't
expect.
There
will
be
any
more
of
those
changes.
A
I
don't
either,
but
I
also
I
mean
I
I,
like
I.
Don't
either
I
also
recognize
that
we
have
a
small
subset
of
like
developer
resources.
We
can
spend
on
working
on
things,
and
that
includes
reviewing
as
well
as
contributing
yeah.
So
it's
like
I
I
would
like
to
dedicate
those
to
actually
thinking
getting
things
done.
That
will
get
us
to
GA,
but
I
mean
obviously
like
when
Josh
had
a
need
to
get
this
done
like
it
motivated.
You
know
wanting
to
get
this
in.
B
I
agree
like
honestly,
I
feel
like
this
is
an
implementation
date
detail
of
the
internal
aggregation
more
than
anything
else
right.
The
one
thing
so
I
I
know
I
put
a
number
of
subtasks
here.
B
B
Could
have
you
could
have
exporters,
you
could
have
potential,
you
could
have
a
exporters
at
parse
it
and
B.
You
could
have
something
else
that
generates
that
potentially
like
through
the
what's
the
the
metric
provider.
The
me
the
the
new
thing,
the
bridge
the
bridge
can
also
speak
in
that
language
as
well.
B
All
right,
metrics
data
is
kind
of
our
P
data,
the
lingua
free,
the
lingua
Franca
of
this
open,
Telemetry
SDK
project.
So
that's
probably
the
most
impactful
beyond
having
it
all
there,
so
that
people
can
actually
use
it
all
the
way
through.
But
if
we
can't,
if
we
can't
get
all
of
it
like
if
this
is
blocked,
because
the
spec
hasn't
has
chosen
not
to
stabilize
it,
then
it's
blocked,
but
because
of
that
and
I'm.
Okay
with
that,
and
we
can
move
on
foreign.
B
A
A
All
out
Bridges,
oh
okay!
What
Bridges
currently
produce
exponential
histograms
to
those
open
census?
There's
seeing
a
head
shake
from
David?
No.
B
Open
census
wouldn't
do
it,
but
the
what
you
may
call
it
the
data,
the
go
internal
metrics,
translating
those
into
open
Telemetry
could
be
implemented
as
a
bridge
because
it
already
produces
pre-computed
histograms.
B
B
The
new
the
new
go
metrics
that
internal
go
team
produces.
B
A
B
A
B
A
B
Best
suggestion
of
this
whole
beating
just
to
have
heads
up.
We
only
have
about
four
minutes
left.
A
What
about
support
Standalone
exponential
histogram
data
structures
for
use
by
open
symmetry
components?
Three,
one:
seven
zero!
Should
this
get
moved
that
would
be
pushed
as
well
right,
yeah,
it's
Associated
right,
yeah,
okay,
okay,.
B
A
Definitely
think
it's
a
triage,
accepted
thing:
I,
don't
think,
there's
a
PR
that
we
should
be
doing
on
this
because
I
don't
think,
there's
a
consensus
yet,
but
for
just
triaging
this
I
I.
Maybe
we
need
a
new
column.
That's
like
I,
guess,
blocked
kind
of
makes
sense
like
I,
don't
know
if
that
makes
doesn't
make
sense.
A
I,
do
I
worry
about
the
proposal
to
use
Nan
because
it
can
also
have
meaning
in
other
languages
another
exporters
or
other
systems
right
well,.
B
A
But
I
mean
like
that's
what
I'm
saying
so
like.
If
you
have
a
bridge
that
records
Nan
it
comes
in
and
then
you
have
an
exporter
that
interprets
net
is
not
set,
then,
is
that
the
same
as
what
it
originally
was
used
like?
It
also
is
like
what,
if
you
record
like
some
sort
of
error
as
Nan
in
one
of
these
fields?
Well,
we.
A
A
Because
I
mean
like
bogdan's,
also
pointing
out
that
there
are
other
ways
to
try
to
approach
showing
a
zero
value
that
doesn't
use
nand.
But
it's
also
like
itarex.
B
Field
helps
well,
I
mean
the
it
would
if
this
structure
got
what
you
call
it
allocated
onto
the
stack
and
that
pointer
pointed
to
the
min
or
not
or
nil,
there
wouldn't
have
to
be
another
allocation
for
that
float.
A
But
this
is
what
I'm
saying
like
I
don't
know
if
we
necessarily
need
to
solve
this
in
the
triage
meeting,
but
I
do
think
that
we
need
to
have
some
sort
of
like
discussion
on
this
topic.
Maybe
in
the
issue
probably.
A
B
A
There's
another
thing,
that's
also
said
is,
like
you
know
and
I've
seen
in
other
places,
optional
values
have
like
booleans
associated
with
them.
So,
like
you
know,
I
don't
really
like
that.
But,
like
you
know,
you
have
Min
and
then
you
also
have
like
a
Min
set,
which
is
a
bull
value,
and
then
you
know
if
you
set
them
in
then
you
set
that
to
true
or
something
as
well.
B
Yeah,
so
how
do
we
drive
towards
consensus
on
here?
Like
just
saying
we
need
to
talk
about
it
more
in
the
the
issue
is
probably
not
not
necessarily
the
best
idea.
I
just
want
to
point
out
that
we
are
over
time.
I
could
pull
the
pack
out
of
done
and
we
can
re
restart
this
discussion
next
week.
A
B
Then
let
us
put
it
in
blocked
and
I
will
try
and
come
up
with
a
comment
saying
we
need
to
drive
to
a
consensus.
Potentially
a
PR
would
help
something
like
that.
Sorry
for
going
over
everybody.
B
Okay,
I
hope
you
enjoyed
the
the
fun
Lightning
Fast
triage
meeting.