►
From YouTube: 2022-06-16 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
D
Good
morning
sampling,
sig
bombing,
I
saw
that
mars
pr
was
reviewed
and
merged
by
peter.
Thank
you
for
that.
I
came
to
the
party
a
little
late.
D
It
already
merged
it
by
the
time
I
got
to
scrutinizing
it,
but
I
did
check
it
over
and
made
sure
I
was
following
fairly
advanced,
but
after
thinking
for
a
moment
I
decided
atmar
knows
what
he's
doing
and
the
probability
logic
seemed
seemed
good.
D
Though
that
leads
me
to
wonder
if
there
is
an
agenda
today,
because,
aside
from
reviewing
ottmar's
pr,
I
did
did
nothing
sampling
related
in
these
two.
B
D
Three
of
us
in
this
room
might
not
have
actually
have
anything
to
talk
about
I've
been.
I
have
anxiety
about
this
topic
every
week
for
this
group,
because
I
love
sampling
and
I
really
want
to
work
on
it,
but
there's
priorities
that
are
handed
to
me
each
week
and
they
don't
they
don't
actually
include
much
other
than
coming
to
screening
and
staying
tuned
and
without
atmar
anything
on
the
agenda.
I'm
not
sure
we'd
have
anything
today.
B
B
Know
your
first
yeah
go
ahead
is,
is
that
I
I
don't,
as
I
said,
I'm
a
little
underprepared
because
I
probably
will
be
working
on
this
in
the
next
couple
of
weeks,
but
the
question
of
when
to
what
to
do
with
the
2
to
the
63rd
number.
B
When
you
know
when
that's
coming
into
honeycomb,
am
I
supposed
to
drop
traces
that
have
that
or
am
I
supposed
to
set
it
to
zero
and
keep
them
or
like
and
and
like
I
said
I
don't?
Actually
I
haven't
studied
it
enough
to
even
come
up
with
a
proposed
solution.
Yet
so
it's
hard
for
me
to
have
that
conversation,
but
if
anybody
knew
it
off
the
top
of
their
head,
I
wouldn't
mind
here.
D
Yeah,
I
think
I
I
think
I
got
the
context
of
the
question
I
had
my
ears
dropped
for
a
moment,
but
so
so
I
I
mean,
I
guess,
for
the
values
other
than
263.
The
understanding
is
fairly
clear.
It's
not
true.
It's
2
to
the
negative
63
would
be
the
natural
interpretation
right.
So
we
understand
right.
B
And
honey
honeycomb
has
a
number
which
is
not
related.
You
know
tied
to
powers
of
two,
it's
an
integer
number,
which
is
one
over.
That
is
the
sampling
rate,
and-
and
so
you
know,
when
I
get
a
17,
then
I
can
turn
that
into
our
internal
sampling
rate
of
2
to
the
17th
is
the
value
that
we
store
and
that's
fine,
but
I
don't
know
whether,
when
I
get
a
63,
am
I
supposed
to
just
drop
that
or
the
way.
D
This
was
intended
to
be
used.
You
might
have
a
probability
sampler
in
effect,
that
is
assigning
all
those
numbers,
and
then
you
might
have
some
other
operator
that
is
sort
of
like
not
probability
based
who
is
saying,
oh,
and
I
want
to
see
that,
no
matter
what
like,
I
don't
care,
how
many
counts,
but
I
want
to
see
that
so
that
every.
D
D
B
D
D
So
as
long
as
you
keep
treat
the
so
there's
a
point
at
which
you
say,
I've
got
a
number.
I
need
to
invert
it
for
all
of
those
numbers.
It's
2
to
the
power
inverted.
When
you
get
to
63,
it's
just
zero
and
don't
invert
anything.
Are
you
done?
You
know
you
got
an
exception.
B
D
I
think
that's
that's,
probably
the
highest
level
overview
summary
that
I
could
imagine
there
and
I
it's
all
to
preserve
that
expected
value
that
the
samplers
intended
to
produce
okay
and
it
is
a
leap.
I
think
that
that
this
this
feels
a
little
bit
like
magic
to
people
at
first
and
but
as
far
as
we
know,
math
checks
out
and
the
the
objective
is
achieved.
D
I've
definitely
gotten
this
question
from
lots
of
people,
though,
that
are
just
sort
of
like
uninitiated
to
sampling
and
it's
hard
to
understand
how
counting
zero
can
be
right,
but
it's
balanced
by
over
counting
elsewhere.
C
So
I
wanted
to.
C
It
said
for
a
while
and
well
two
days
ago.
I
I
made
few
comments
there.
I
wonder
if
anybody
else
would
be
interested
in
reviewing
this,
I'm
I'm
honestly
a
little
bit
on
the
fence
here.
I'm
not
sure
if
this
is
really
needed.
It
looks
like
it's
complicating
thinking
instead
of
simplifying,
but
maybe
I'm
not
seeing
all
the
use
cases
so.
D
So
I
pulled
it
up
and
I'm
remembering
this
there's
been
past
debate
on
this
topic.
It's
unfortunately,
never
there's
never
been
enough
attention
or
energy
given
to
the
debate
to
like
finish
it.
It's
always
sort
of
like
a
someone
has
a
minor
reservation
and
we
stop
talking.
D
D
And
I
think
the
brief
summary
of
the
other
way
to
do
it
that
I'm
aware
of
is
to
put
the
control
of
the
sampler
like
put
all
that
stuff
that
we're
talking
about
into
the
tracer
provider
itself
and
that
actually
means
probably
redesigning
the
the
constructor
or
the
whole
setup
procedure
for
the
tracer
provider.
D
But
this
is
just
what
we
never
finished
like
it's
just
an
unfinished
project
to
say
how
you
get
configurable
sampling
on
a
per
instrumentation
library
basis,
and
I
so
without
scrutinizing
tristan's
words
in
depth
right
here.
I
can't
we
can't
necessarily
respond,
but
I
do
feel
like
there's.
D
This
doesn't
feel
necessary
to
me
just
on
the
surface,
so
maybe
I'm
responding
to
the
same
gut
feeling
that
you
had,
but
I
think
I
mean
the
the
imaginary
solution
that
I'm
I'm
thinking
of
right
now.
Is
you
pass
a
instead
of
having
a
with
sampler
option
to
your
sdk,
which
is
what
we
have
today?
D
C
Okay,
okay,
yeah!
So
if
you
could
have
a
look-
and
maybe
if
you
add
a
few
comments,
so
his
pr
would
be
helpful-
and
I
I'm
I
hope
to
see
him
at
this
meeting
but
unfortunately
he
hasn't
joined
so.
D
Right,
that's
not
the
color.
I
want
okay,
I
don't
know
if
he
knows
about
this
meeting,
but
I
will
definitely
take
an
assignment
to
review
this
in
the
coming
days
and
see
if
we
can
get
some
progress
here,
because
tristan
often
has
his
finger
on
the
pulse
and
he's
often
ahead
of
us
on
these
things
because
he's
off
working
in
a
library
with
very
few
obstacles.
D
Okay,
thank
you
peter.
I
will
look
at
that
and
I
won't
do
it
in
front
of
you
all
here.
So
let's
say,
let's
see,
is
there
anything
else
we
could
talk
about
today?
Otherwise
I
think
we
should
get
back
to
work
and
review
this.
B
I
think
that's
reasonable.
I
just
want
to
say
to
add
to
that
thing
is
like
I'm
trying
to
come
up
to
speed
on
all
this
stuff.
So
you
know,
as
we've
been
talking,
I've
been
like
browsing
through
this
stuff
and
it
seems
to
me
like
what
the
proposal
is,
is
essentially
saying:
take
the
tracer
provider
and
provide
it
an
extra
abstraction,
which
is
the
sampling
provider
to
the
tracer
provider.
C
C
C
Separate
for
for
all,
depending
on
the
instrumentation
scope,
really
with.
C
A
D
B
Up
anymore,
I
was
just
trying
to
get
additional
clarification
because
I
will
try
to
make
some
time.
I
don't
know
how
much
I'll
have
this
week,
but
to
take
a
look.
D
It's
the
way,
I'm
feeling
right
now,
just
because
we
have
meter
providers,
you
get
meters
from
and
the
user
interacts
with,
and
the
user
controls
which
global
instance
of
the
provider
they're
using
and
that's
such
we
don't
want
to
have
a
new
global
instance.
Here
the
tracer
provider
knows
it's
sampler
setup.
That's
why
I'm
using
the
word
setup,
not
provider.
I
guess
what
I
want
to
say,
but
I
think
this
is
a
very
good
conversation
to
have
in
the
slack
or
in
the
thread
for
tristan
and
others.
D
D
Great
thanks
for
the
reminder
about
this
computer
and
thanks
for
reviewing
mars
work,
and
thanks
for
coming
and
I'll,
see
you
in
two
weeks.