►
From YouTube: 2022-08-24 meeting
Description
Open Telemetry Meeting 1's Personal Meeting Room
A
A
C
Hey
hi
everyone,
I'm
audible,
hi.
C
Yeah,
let's
start,
I
see
a
few
new
faces.
Alexander
lopez
and
prismic
welcome
guys
hi
mark
good
to
see
you
back
from
vacations.
A
C
Okay,
yeah.
Let's
start,
I
do
see
one
point:
probably
we
can
start
with
that.
It's
coming
from
prismac
the
auto
instrumentation.
C
B
B
And
what
I
mean
by
that
is
a
way
to
I
don't
know,
maybe
source
code
modifying
at
compile
time
or
maybe
com
modifying
binary
at
run
time
or
something
like
that.
So
because
recently
I
I
made
a
similar
project
for
the
go
programming
language
and
I
can.
I
can
show
you
how
it
works,
and
maybe
we
can
do
something
similar
for
c
plus
plus.
B
It
was
recently
accepted
by
the
community
and
they
will
be
part
of
open
telemetry
go
instrumentation.
C
B
Yeah,
that's
that's
one
thing,
but
this
project
that
I
made
for
for
the
girl
programming
language
in
fact
is
modifying
source
code.
So
it's
it's!
It's
doing
that
at
compile
time,
so
it
injects
at
each
method,
open
telemetry
code
for
creating
spawns.
B
For
for
goal,
we
also
has
to
modify
a
function
declaration
as
it
needs
to
have
additional
context,
parameter.
C
B
B
Yes,
so
you
know
for
go
it's
it's,
I
think
much
simpler,
because
all
tools
are
available
in
a
go
standard
library.
So
we
have
a
parser.
We
have
a
asd
and
we
have
a
full
library
for
analysis
of
program
and
for
for
c
plus
plus
I
was
thinking
about
using
clunk
libraries.
B
C
A
C
A
A
B
So
do
you
see
my
screen
yeah
so
so
I
was
thinking
about
clank,
which
is
a
c
compiler
and
in
fact
it's
built
from
a
set
of
c
plus
plus
libraries,
and
it
contains
everything
that
is
needed
for
c
plus
plus
program
analysis,
and
so
we
could
use
it
just
to
inject
specific,
open,
telemetry
calls
into
functions
and
what
I
can
do
now.
I
can
show
you
how
this
works
for
go
programming,
language,
so
this
tool
that
I
created.
B
Yeah
so
so
I
will
show
you
a
very
simple
example,
so
we
have
two
source
files.
First,
is
the
mine
go
and
the
second
one
is
fib
and
it
just
calculated
fibonacci
number
and
it
has
three
or
four
functions.
One
is
a
lambda
function,
so
so
we
have
main
main
function
here
and
a
few
additional
in
fib
go
file
and
now
to
instrument
this
code.
B
We
have
to
add
some
kind
of
starting
point,
so
in
this
case
this
tool
first
search
for
such
a
call,
but
this
can
be
done
also
in
other
ways,
so
by
adding
something
into
configuration,
file
or
adding
a
switch
to
to
the
tool
about
telling
that
the
main
will
be
the
starting
point
of
instrumentation.
B
And
then
this
tool
has
a
few
passes,
so
it
searches
all
the
all
the
functions
that
are
in
the
project
and
all
calls
that
are
visible
from
from
this
place.
B
So
when
I
will
run
this
tool,
I
have
to
pass
a
few
parameters,
but
one
of
them
is
just
path
to
the
pro.
To
my
my
demo
project,
the
second
one
is
is
a
kind
of
wild
card
that
tells
that
it
should.
B
Functions
for
create
expanse,
and
also
I
am,
as
I
mentioned,
we
need
in
go
additional
context
parameters,
so
it
it
is
also
added,
and
so
the
function,
definition
and
or
declaration
is
extended
with
this
additional
context
parameter.
So,
as
you
can
see,
we
have
the
full
full
trace
in
this
case.
So
that's
how
it
works
and
using
clank
we
could
do
the
same.
B
But
of
course,
as
I
said,
for
go,
everything
is
in
the
standard
library
and
clank.
Libraries
is
a
third
party
thing,
so
that's
something
that
has
to
be
cons
taking
into
account
yeah.
So
that's
the
idea,
one
of
the
probably
ideas.
So
I
was
wondering
if
someone
was
thinking
about
such
a
tool
or
or
way
of
doing
an
instrumentation.
A
B
B
So
I
I
I'm
not
sure
I
haven't
looked
at
the
c
plus
plus
sdk.
Yet
so
I
I'm
wondering
how
contacts
is
passed
from
one
span
to
another
spawn,
because
in
go
we
have
directly,
we
are
passing
it
directly
by
a
parameter.
I
don't
know
how
it
is
done
in
c
plus
plus
do
we
have.
I
don't
know
something
like
thread
specific
storage
for
yes,.
A
C
B
So,
for
now
I
don't
see
any
a
major
challenge,
so
it
should.
From
my
perspective
it
is
feasible
yeah.
So
so
we
can
try
do
that
at
least
spock
do
that
dude
poke
and
try
it
currently
this
tool
for
go,
I'm
testing
it
with
large,
larger
applications.
I
mean
applications
that
works
in
in
not
the
toyota
examples,
but
real
real
world
applications.
C
Okay,
yeah,
it
looks
interesting
to
me.
I
think
it's
something.
Probably
it's
good,
if
I
think
you
can
try
it
out
for
c,
plus
plus
and
in
general.
This
tool
is
more
helpful
for
any
open
source
applications
or
instrumentation
libraries,
which
are
already
existing
right.
C
A
B
A
Once
the
code
is
instrumented
the
code,
the
instrumental
code
can
be
compiled
with
anything,
not
not
necessarily
c-line.
A
Sounds
interesting
just
one
comment:
if
some,
if
there
is
an
existing
library
which
is
implementing
a
class
with
a
virtual
method
and
some
application
is
a
passing,
but
I
don't
know.
A
Instrumenting,
the
the
subclass
would
be
probably
more
difficult
because
you
cannot
add
a
context
parameter
in
the
virtual
method.
Signature
so
we'll
probably
have
to
use
fabric
on
storage
to
get
the
context.
B
Yes,
so
so
for
go,
we
are
searching.
We,
we
need
to
know
exactly
what
what
methods
will
be
called
and
via
which
interfaces.
So
we
we
want.
We
want
to
be
sure
that
we
will
inject
only
the
code
into
this
method
that
are
really
used
from
from
the
starting
point.
B
Yes,
so
so
then
there
might
be
a
challenge
but
yeah.
I
I
haven't
thinking
about
that.
Yet,
but
yeah.
C
I
mean
I
just
thought
something
which
I
think
we
have
to
immediately.
Think
probably
that
can
be
over
the
course
of
time
we
can
see,
but
I
think
it
looks
good.
I
think.
Definitely,
if
you
can,
if
you
have
some
more,
I
think
if
you
want
to
really
do
some
more
study
on
this
on
c
plus
plus
and
come
up
come
back
to,
I
mean
here
to
see
how
it
works.
C
B
Okay,
so,
but
at
least
your
first
impression
is
that
it,
it
might
be
something
valuable
right
or
what
what's
your
opinion
on
that?
C
Yeah,
definitely,
I
think,
to
me,
I
think,
definitely,
some
look
something
which
is
valuable.
This
was
something
which
we
were
missing,
auto
instrumentation
in
c
plus.
We
were
not
thinking
on
this.
Probably
we
felt
that
it's
not
something
feasible
or
not,
something
which
could
be
doable.
B
C
B
C
C
A
Just
had
one
question
for
in
your
example,
for
example:
is
there
a
way
for,
for
example,
not
to
add
spam
on
function,
foo
or.
B
A
B
B
A
C
Not
really,
I
think
we
will
look
forward
to
you.
I
think
when
you
have
more
data,
or
I
mean
how
you
have
done
it
so
probably
I
think
it
would
be
good
to
really
see
at
that
time.
C
A
C
A
C
C
So
there
is
the
pr
there
is
a
pr
raised
on
for
the
specs
update,
which.
C
C
So
it
means
that
if
we
disable
and
the
sdk
the
sdk,
which
is
built
with
the
application,
it
should
be
using
the
knob
operations
or
not
all
the
knob
interfaces
this
which
are
built
for
the
api.
So
basically
it
should
be
kind
of
not
doing
any
operation.
The
span
doing
a
call
on
start
span
should
not
be
really
starting
a
span.
It
should
call
all
the
know
inherited
classes
which
we
have
for
spans.
C
A
I've
I've
not
looked
at
the
spec
itself,
but
we
probably
can
disable
it
if
we
say,
for
example,
in
the
set
trace
provider
or
check
the
environment
variable
and
do
nothing
if
it's
disabled,
so
the
code
will
still
be
linked
and
compiled
and
linked.
But
when
you
set
the
trace
provider,
we'll
just
ignore
it,
if
sometimes
environment
variable
says
to
disable
it.
A
C
C
C
C
Spans,
I
think
I
got
your
point
mark
what
you're
saying.
Probably
we
use
the
tracer
provided
provided
by
sdk,
but
it
should
check
the
environment
available
based
on
that
it
should
decide
whether
the
span
should
be
created
or
not.
A
C
A
C
A
C
C
These
are
something
which
are
set
at
the
start
of
the
fp,
just
like
any
other
environment
variables,
we
have
say
for
otlp
exporter,
all
those
are
set
at
the
start
of
the
application,
the
endpoint
of
the
otl,
p,
http
and
all
those
things.
I
think
it's
similar
to
that
this.
All
this
environmental
element
should
be
available
before
really
the
sdk
bootstraps.
A
A
C
C
Yeah,
I
don't
think
it
should
affect
us
in
some
way,
whether
so
it
says
that
as
of
now
environment
variables
are
enabled
or
disabled,
or
it
has
some
value
in
it.
But
probably
it
says
that
it
can
have
a
true
and
false.
Also.
C
C
C
C
A
C
That
means
the
otlp
protocols.
Exporter
should
emit
a
user
agent
header
to
minimally
to
him
at
a
minimum,
identify
the
exporter,
the
language
of
the
implementation
and
version
of
the
exporter.
C
C
C
C
C
I
will
say
it's
not
a
must
to
be
part
of
ga,
but
I
think
some
good
to
be
feature
complete.
I
think
it's
something
if
you
can
incorporate
it
in
the
ga.
I
think
it
would
be
good
to
have
that
will
ensure
us
that
we,
our
sdk
api
surface,
will
not
change
after
ga,
otherwise
mean
if
we
add
it
afterwards,
we
may
have
to
change
our
the
eps
surface
for
the
sdk
provide
builders
to
avoid
existing
sdk
internals.
This
is
something
mark.
C
I
think
you
probably
will
be
it's
with
you
and
I
think
at
some
time.
Maybe
you
have
to
start
we
will
get.
We
may
start
work
discussing
on
this
all
right.
C
Yes,
yeah
totally
agree.
I
think
once
we
have
a
more
stable,
stable
sdk.
I
think
probably
it's
good
to
start
on
this
add
in
my
memory
metric
exporter.
This
is
not
complete.
Probably
we
may
have
to
rewrite
it
without
using
the
otp
exporter,
so
yeah,
it's
something
something
I
think
it's
good
to
have
it
as
part
of
ga
at
support
for
pullmatic
exporter
here.
This
is
something
which
is
missing
and
that's
need
to
give
it
a
thought.
C
We
don't
support.
Pullmatrix
exporter,
we
have
periodic
push
exporter
so
which
basically
pulls
the
matrix
information
from
storage
in
regular
interval,
and
then
it
is
pushing
to
the.
C
C
Yes,
yeah,
so
basically,
we
have
to
create
a
pull
matrix
reader
which
can
be
used
so
matrix.
Reader
reader
has
encapsulates
metric
exporter,
so
we
should
have
a
push.
The
metric
leader
and
a
pool
matrix
reader,
so
pull
metric
leader
can
be
called
explicit
explicitly
at
the
application
level,
so
we
should
be
able
to
call
matrix
reader.collect
whenever
we
want
to
collect
the
matrix.
C
As
of
now,
we
don't
support
that
as
of
now,
if
we,
if
we
see
we,
we
just
add
that
the
periodic
metric
reader
in
the
sdk
and
that
is
internally
periodically
pulling
the
data
and
pushing
it
to
the
exporter,
which
is
something
probably
good
to
have,
but
I
think
it
should
be
on
demand
reading
of
the
matrix
data
also,
which
should
be
supported.
C
C
A
C
At
support
for
exponential
histogram
aggregation,
this
is
a
bigger
thing,
I'll
assign
it
to
myself
just
in
case,
unless
somebody
want
to
really
take
it,
but
I
think
this
needs.
As
I
understand
the
exponential
histogram
aggregation
is
a
more
complicated
thing
based
on
the
feedback
which
we
are
getting
from
dotnet
team,
that
this
is
something
not
easy
to
implement.
C
Probably
it
will
takes
a
substantial
effort
to
really
implement
it,
and
even
dotnet
has
not
delivered
it
as
part
of
ga
for
matrix,
so
even
they
have
they
implemented
it
after
they
went
for
the
matrix
ga.
So
I
think
if
you
can
decide,
if
it
is
something
new
will
be
needing
more
time,
you
can
postpone
it
or
after
g.
C
Remove
circular
dependencies
here
should
I
assign
it
to
myself
s
on,
I
mean
just
let
me
know
probably.
C
C
C
C
We
don't
have
that
so,
let's
give
it
p2,
but
I
think
something
we
can
see
afterwards
this.
This
is
important.
I
think
we
need
to
fix
it.
We
have
a
very
bad
circular
dependencies
between
these
and
it's
not
a
nice
way
of
doing
it.
C
C
I
think
better
to
have
a
clean
delivery,
removing
all
these
things
recording
our
interface.
This
is
more
of
a
kind
of
investigation.
If
we
can
really
do
it,
we
cannot
use
the
recordable
interface
the
way
we
did
it
for
traces
and
login
and
logger,
because
the
in
case
of
places
and
loggers
whatever
that,
whatever
trays
and
locks
the
application
was
sending,
we
have
to
send
as
same
as
it
is
to
the
exporter.
C
C
C
Do
what
do
I
have
to
add
some
more
cases
for
more
issues
related
to
benchmarking?
C
Probably
I'll,
add
it
and
then
also
need
to
add
some
issues
here
for
any
new
changes
which
have
come
into
the
specs
in
the
last
time
when
we
created
all
the
issues,
so
they
have
been
lots
of
changes,
and
so
probably
we
need
to
create
the
issues
there.
C
Yeah.
That's
all
I
think,
hopefully,
for
the
matrix.
C
Set
with
epsilon
still
required
epsilon.
I
think
this
is
probably
owen
has
given
a
nice
solution
for
this,
which
should
be
good
enough
for
that.
So
they
are
already
using
epsilon
library
in
their
application
and
they
want
to
reuse
it.
C
They
use
the
same
library.
So
basically
that's
what
he's
saying
that
they
can
set
the
cmake
find
path
to
point
to
that
library
and
then
they
should
be
able
to
build,
which
should
work
using
cma
prefix
path
which
points
to
the
live,
apps,
libraries
they
are
using
and
then
trying
to
build.
But
let's
see.
C
C
C
C
C
C
Decode
values
for
resource
attributes
yeah.
This
is
also
something
which
we
have
to
take
it.
Nothing
goes
in,
but
some
somewhere,
if
you
can
take
it
open,
telemetry
api,
singleton,
trace
provider,
do
not
play
well
in
shared
libraries,
users
and
the
heights
yeah
this.
I
think
we
already
have
one
draft
pier
from
mark,
and
probably
we
can
discuss
it
quickly
when
we
go
to
the
pr's
matrix
exemplars.
C
A
C
C
C
C
Have
been
lots
of
discussion
when
this
was
added,
and
I
think
finally
it
got
added
because
most
of
the
maintainers
and
approvals
that
time
they've
somehow
they
agreed.
Okay,
let's
keep
it
here
and
then
we
can
decide
afterwards,
but
I
did
create
an
issue
for
that
to
move
these,
both
the
two
exporters
outside
of
contemporary
pro
so
yeah,
it
needs
to
be
removed.
Sometime
from
there.
A
So
the
the
most
important
part
is
to
actually
build
a
unit
test
to
actually
checkbox
so
that
it
can
be
tested
on
our
platforms,
because
there
are
so
many
things
which
are
compiler
dependent
and
whatnot.
But
we
need
a
test
case
to
provide
the
resolution
work.
And
then
we
can
discuss
for
this
compiler
that
compiler,
which
option
should
be
used.
A
C
C
C
C
C
C
And
probably,
I
think
if
it
does
not
really
help
us
to
really
for
its
something
which
we
can
not
troubleshoot,
it's
good
to
raise
a
issue
with
grpc,
because
we
are
not
doing
anything
different.
We
are
just
trying
to
build
grpc,
given
version
of
grpc
on
mac
on
a
given
mac,
os
version,
that's
what
we
are
trying
to
do
and
if
it
is
failing.
Probably
if
it's
segmenting
fault,
I
think
definitely
we
have
to
start
looking
into
that.
C
I
have
mac
with
me,
I
can
try
to
I
did.
I
did
basically
test
this
on
window
on
on,
at
least
on
linux
and
windows.
I
didn't
have
any
problem.
All
the
tests
was
for
successful
and
it
is
successful
here
also
I
mean
the
problem
is
coming
only
for
mac
and
I
think
it's
good
to
probably
raise
the
issue.
That's
what
I'm
thinking
I
mean.
Let
me
know
what
you,
if
you
see
otherwise.
A
So
grpc
is
moving.
This
is
not
release
of
grpc
already,
so,
even
if
a
shoe
is
raised,
I
don't
think
jpc
will
fix
it,
at
least
not
in
the
same
release
of
jrpc
and-
and
this
is
assuming
it's
a
burgundy
jpc
would
be
also
a
problem
with
the
compiler
for
somehow,
but.
C
Don't
they
should
be
maintaining
the
releases,
but
they
should
not.
I
mean
because
there
would
be
there
would
be.
I
mean
projects
which
would
be
using
those
releases
and
they
cannot
just
they.
They
should
maintain
separate,
I'm
not
sure
if
they
do
a
separate
bug
fixes
for
at
the
release
level,
if
they
are
not
doing
it,
I
think
it's
good
to
even
test
it
yeah.
The
problem
is
yeah.
Please
go
ahead.
C
C
A
A
C
C
A
Please,
at
least
I
think
we
should
have
some
ci
that
test
against
maybe
latest
grpc
yeah.
C
Okay,
yeah,
I
mean
you
if
you
think
you
have
time
to
really
probably
see
this.
I
think
it's
good.
Otherwise,
probably
I'll
can
spend
some
time
when
I
get
it
this
week,
I'm
still
planning
to
raise
a
issue
on
grpc
for
this
segmentation
fault.
Irrespective
of
of
this
version,
at
least
let
them
reply
back
that
this
is
already
fixed
in
the
higher
version.
So
we
have,
we
know.