►
From YouTube: 2022-05-19 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
Cool
actually,
I
I
didn't
hate
that
api,
even
though
it's
internal
like
the
started.
B
A
This
started
and
yeah
yeah,
okay,
start
and
end.
I
guess
the
only
thing
that
it
doesn't
accomplish
still
is
hiding
the
end
time.
B
Yeah,
that's
one
thing
and
the
other
is
I.
I
think
that
it's
probably
a
little
bit
too
specific.
It's
probably
a
little
bit
too
oriented
to
that
one.
Certain
use
case
that
we
have
but
yeah.
A
C
I
have
I
have
one
small
thing,
so
I
open
a
pr
to
open
telemetry
java
and
it's
I
add
a
listing
of
all
the
artifacts
to
the
readme
and
I
think
that's
a
good
thing
because
it
makes
it
easier
to
find.
C
C
C
A
Do
you
have
in
here
the
list
of
your
questions.
C
B
It
did
actually
when
I
saw
your
pr,
I
actually
searched
for
the
sdk
trace
and
sdk
matrix,
because
I
was
pretty
sure
that
that
trace
was
supposed
to
be
integrated
into
the
main
sdk
module
so
yeah,
it's
weird
to
have
this
like
it's
a
separate
artifacts.
C
B
Yeah,
I
agree,
and
it's
probably
similar
with
exporters-
I
mean
if,
if
we
merged
the
all
the
sdk
components
into
one,
then
I
suppose
it
would
make
sense
to
do
the
same
for
exporters.
So
I
guess
that
whatever
decision
we
make
it's
kind
of
connected.
D
A
Well,
just
from
prior
to
feedback,
occasionally,
there's
been
some
buddy
who's
wanted,
like
individual
pieces
of
stuff
extracted
out,
but
the
general
message
we've
given
is
that
the
value
of
open
telemetry
is
that
it's
all
three
signals
combined.
A
C
Well,
yeah,
so
I
I
agree
with
that.
So
this
I
wanted
to
list
them
here
for
two
reasons.
So
one
you
know
it's
hard
to
discover
which
artifacts
are
available.
You
know
the
only
way
to
do
that
today
is
to
go
and
look
at
the
the
palm
from
the
bomb
funny
sentence
and
like
look
at
all
the
artifacts
in
maven,
and
so
that's
that's
a
pretty
roundabout
way
of
doing
that
and
like
we
don't
have
any
concise
place
where
we
say
you
know,
for
most
setups
just
depend
on
these
things.
C
I
suppose
our
docs
and
opentelemetry.io
do
that
a
little
bit,
but
maybe
we
can
do
a
better
job,
but
then
the
second
reason
is
once
you
know
that
you
want
an
artifact,
it's
hard
to
find
what
source
corresponds
to
that
artifact
like
it's,
not
obvious
that
when
you
say
like
hey,
I
have
a
dependency
on
the
jaeger
thrift
exporter
like:
where
does
where
does
that
live,
and
so
each
one
of
these
components
the
text
links
to
the
directory,
the
the
module
that
contains
the
source
as
well,
so
those
were
kind
of
the
two
purposes
and
apparently
there's
no.
C
A
C
A
B
C
A
A
So
far
seem
like
a
natural
fit
yeah.
There
will
always
I
mean.
Is
there?
Is
there
a
reasonable
approach,
or
what
could
we
tell
somebody
who's
like?
No,
I
really
don't
like
I'm
running
in
a
constrained
environment
like
an
embedded
device
or
something,
and
they
really
don't
want
a
bunch
of
extra
classes
hanging
out
they
could
repackage.
A
B
If
they
really
running
in
a
constrained
environment,
I
mean
these
jars
probably
are
not
that
heavy
right.
I
don't
think
that
the
sdks
takes
more
than
one
megabyte.
So
if
they
really
running
in
a
constrained
environment,
then
I
suppose
they
already
should
have
some
kind
of
minimization
in
their
builds.
C
A
But
I
I
think
that
this
is,
I
think
the
classes
aren't
really
in
here,
because
there's
no
way
that
it's
5k.
C
Oh,
maybe
oh
because
it
does
have
transitive
dependencies,
yeah,
the
metrics
sdk
and
yeah
trace
individually.
That's
true!
Okay!
Here
this
is
more
accurate,
so
I'm
looking
at
the
the
metrics
sdk
itself
now
so
just
the
metrics
sdk
and
that's
28k,
no,
no
285k!
It's
a
quarter
megabyte
that
sounds
better
yeah.
That
tracks.
A
But
yeah
mateusz's
point
if
I
think
it
was
mateo
shortcase
and
about
the
somebody
should
be
using
proguard
anyway,
if
they
care
about
that.
A
A
The
the
blu-ray
specification
requires
a
jdk
like
a
sorry,
a
jerry
on
every
device,
so
every
blu-ray
device
has
a
jvm
and
so
people
put
java
code
on
the
blu-ray
discs
yeah.
I
I
used
to
do
that.
Yeah.
A
C
C
A
C
Cool,
I'm
also
confused
about
the
distinction
between
extensions
and
sdk
extensions.
I
don't
know
why
we
make
a
distinction
there,
but
that's.
C
C
C
D
Just
the
first
one.
A
A
B
B
Sorry
go
ahead,
yeah,
so
the
only
problem
here
that
I
remember
was
that
it
was
hard
to
configure
the
rule-based
sampler
using
just
properties.
You
know
it
had
like
a
builder.
C
It
kind
of
reminds
me
of
this
extension
that
I
wrote
that
allows
you
to
describe
view
configuration
in
yaml
and
the
extension
parses
that
yamo
and
configures
the
views
on
your
sdk,
and
you
know
I
I
think
I
think
long
term
we
we're
going
to
have
a
a
you
know.
Some
sort
of
yaml
or
file
based
configuration
specification
and
hopefully
there'll
be
some
sort
of
accommodation
for
this
type
of
thing
that
we're
talking
about.
But
I
suppose
we
could
prototype
it
in
the
meantime.
D
A
A
Yeah,
I
think,
long
term.
The
the
ideal
solution
is
the
config
having
the
open,
telemetry
configuration
format
that
I
believe
josh
sarath
is
planning
to
work
on
and
having
a
standard
configuration
for
sort
of
rule-based
samplers.
A
A
Nikita
has
an
issue
already
for
the,
I
believe
for
the.
C
All
right
with
so
many
con
with
so
many
people
interested
in
that.
I
wonder
if
we
opened
an
issue
in
contrib,
describing
you
know
this.
A
possible
solution
of
you
know:
building
an
extension
that
incorporated
the
rule-based
sampler
with
some
sort
of
configuration
for
configuring
configuration
specification,
for
you
know
configuring
that
I
wonder
if
somebody
would
pick
it
up,
then
we
could
maybe
that's
a
more
a
less
sad
answer,
because
we
can
say:
hey,
there's
interest
in
this.
If,
if
you
go
and
build
this,
you
can
make
it
a
reality.