►
From YouTube: 2022-06-29 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
B
C
C
D
F
Can
sign
in
yeah
sure
you'll
still
be
on
you
you'll
still
be
on
youtube,
so
you'll
never
be
anonymous
unless
you're
training.
Here,
oh
yeah,
I
I'm
sort
of
a
fly
on
the
wall
here
today.
Yeah.
E
All
right
welcome,
matt,
I
think
robert.
We
we
should
get
started.
I
think
we
have
a
a
bunch
of
interesting,
open
pool
requests,
so,
let's
open
and
see
if
we
it's
worth
discussing
some
of
them
right
now,.
A
So
I
I
just
want
to
take
like
opinion.
There
were
two
things
like
which
we
could
do.
I
like
use
the
existing
test
app
and
introduced
a
command
line
to
separate
out
the
traces
and
metrics
at
this
point
later
left
the
place
even
for
the
logs
and
I'm
receiving
a
feedback,
it's
better
to
have
a
separate
test
app
for
all
these
three
scenarios.
If
I
understand
correctly,
I
can
no.
C
No,
no!
No!
No!
No!
This
is
not
what
I'm
at.
I
meant
that,
usually
I
prefer
to
have
separate
test
apps,
but
for
this
scenario
I
find
to
have
three
at
once
like
when
I'm
creating
instrumentation
applications.
I
prefer
to
have
it
separated,
but
for
this
like
smoke
test
manual,
instrumentation
personally,
I'm
fine
to
have
it
combined.
A
A
E
Yeah,
just
in
general
because
for
the
instrumentations
make
a
lot
of
sense
to
keep
separate
because
they
have
different
dependencies
and
we
may
want
to
test
separate,
but
in
general
just
for
for
this
case.
If
we
can
just
have
one
app
it
saved,
it's
just
made
a
kind
of
cumulative
on
the
long
run
right,
but
it
saves
a
little
bit
of
build
time
and
everything
you
know.
E
So
it's
a
it's
a
tiny
bit
for
a
local
butte,
but
remember
that
we
run
this
in
ci
and
every
one
of
us
run
this
from
damage
to
time.
So
I
I
rather
kind
of
have
just
one.
A
Yeah
the
car
like
what
we
are
discussing
is
within
a
one
app
like,
so
we
have
a
command
line
argument
which
decides
whether
we
should
do
a
tracing
or
metrics.
So
the
like.
What
we
are
trying
to
like
discussing
here
is
whether
we
should
have
that
command
line
argument
or
just
have
collectors
metrics
in
all
the
scenarios,
without
that
condition,.
E
Yeah,
I
I
think
we
we
we
should
have
both,
because
we
can
test
then
traces,
matrix,
traces
and
metrics,
just
in
the
off
case
that
somehow
they
interfere
with
each
other.
You
know.
C
I
don't
know,
but
then
so
your
apollo
suggests
to
have
this
conditionals
to
make
it
possible
to
disable
some
some
emission
right,
yeah
yeah
in
principle.
E
I
think
I
think
it
seems
reasonable
to
test
them
separate.
You
know
it's.
I
know
that
it
basically
prevents
some
kind
of
crazy
regression,
a
very
unlikely
one,
but
it
seems
reasonable
because
I
think
maybe
some
customers,
you
do
just
traces.
Some
you
just
do
metrics
and
some
most
of
them
will
do
both.
A
But
even
if
we
remove
that
condition,
it
will
not
impact,
for
example,
if
we
are
emitting
places
unless
and
until
we
have
a
listener
to
it,
we
don't
capture
any
information
out
of
it
yeah
in
in
that
way,
there
won't
be
major
impact,
but
I
I
felt
it
like:
whatever
you
are
thinking,
that's
the
same
thought
I
had
it
and
kept
it
separate,
but
no
strong
preference
here
how
to
take
it
forward.
If
need
be,
we
can
keep
it
or
can
remove
the
condition.
E
No,
I
see,
I
see
what
you
mean,
because
this
is
about
more
about
the
listening
on
our
side.
A
Then
so,
then
I
will
go
ahead
and
get
it
removed,
maybe
on
a
need
basis.
We
will
get
back
to
this
later,
so
I'll,
remove
the
condition
for
the
time
being
and
this
command
line
check.
C
C
E
A
I
completely
agree
so
then
I
read
that
I
went
there
did
further
research
also
why
it
is
happening,
because
we
calling
it
dispose
itself
should
do
a
horse
flush
from
an
open,
telemetry
sdk
perspective,
even
if
they
don't
call
them
directly,
it
should
happen,
and
then
I
realized
that
open
telemetry
that
runtime
package
does
not
have
a
dispose
implemented
it
and
that's
what
is
causing
the
issue
at
this
point,
and
I
see
whatever
the
like
the
environment
variable
that
we
are
planning
to
add
to
the
sdk
that
will
solve
this
purpose
and
right
now
I
don't
know
without
that,
is
there
an
easier
way
to
test
this
out?
A
A
Finding
what's
the
cost,
I
created
a
issue
or
showing
the
like
the
contrary
for
the
runtime
metrics
saying
that
they
need
to
like
use
the
dispose
call
so
either
we
if
we
need
to
add
as
a
workaround
like
from
which
I
like
the
option
to
we,
we
can
it's
better
to
have
that
at
some
point
in
time.
We
might
need
that.
I
believe
the
force
flash
on
metric.
C
You
both
me
and
chris
are
okay
with
it
for
sure.
Okay,
but
like
do
you
want
to
do
it
like,
I
propose
conditionally,
like
not
enabled
by
default
or
do
you
you
wanted
to
have
it
always?
No,
we
shouldn't
it
should
not
be
enabled
by
default.
Okay.
So
I'm
totally
okay
with.
A
A
So
last
one
week,
I
spent
a
lot
of
debugging
time
and
I
could
not
reach
anywhere
esther.
I
found
a
time
to
create
a
pier,
so
I
was
on
it
for
all
the
week,
so
this
one
everything
is
working
with
prometheus.
I
tried
otlp
also
similar
thing.
I
don't
know
whether
it's
for
splash
or
what
is
causing
an
issue,
especially
whenever
it
comes
to
net
framework.
A
The
metrics
are
not
getting
like
pushed
on
a
like
interval
like
what
we
expected.
We
we
had
to
wait
for
some
time.
I
thought
prometheus
is
immediate
and
we
don't
need
to
do
those
flaky
things
here
that
that's
the
reason
for
me
to
get
to
prometheus.
A
B
Yeah
so
quick
question
for
this
test.
I
wonder
what's
different
about
it
compared
to
the
http
client
tests,
because
that
test
is
dealing
with
asp.net
core
instrumentation
and
we
weren't
doing
a
force
flush
for
that
test.
I
believe
I
was
just
relying
on
the
application
exiting
and
things
getting
disposed.
A
So
I
am
not
sure,
even
if
I
I
have
we
have
like
even
my
world,
like
the
other
peer,
even
on
my
machine,
for
six
two
fails,
but
all
the
dot
net
core
passes.
So
I'm
not
sure.
What's
going
on,
I
need
to
debug,
like
the
sdk,
a
little
to
understand
why
only
the
dot
net
framework
is
fixed
to
only
that
60
seconds
and
not
the
like
door.
A
C
A
And
it's
http
client
is
only
for
net
core.
It
is
not
supported
for
dotnet
framework,
especially
if
you
are
comparing
it
is.
C
A
Again,
http
client
is
supported
only
with
dot
net
core.
It's
not
supported
for
dotnet
framework.
C
A
A
And
even
before
going,
let's
complete
the
metrics.
A
Adding
a
prometheus
is
going
to
give
a
different
challenge.
There
is
a
one
more
task
pending
on
us:
it's
adding
a
test
application
and
collector
metrics
from
the
prometheus
and
validate
it
for
that.
We
cannot
use
a
definitely
user
console
application
because
the
we
are
planning
to
use
a
prometheus
http
script.
A
The
moment
application
exists.
We
lose
that
end
point
also.
So
we
need
to
keep
an
application
live
until
like
matrix
is
validated.
So
what
we
need
to
do
is.
C
B
Yeah,
so
I
think.net
has
some
mechanisms
to
do
that.
I
know
on.net
framework,
I
want
to
say
it's
the
event
weight
handle
there
there's
a
way.
E
B
E
Yeah,
but
we
can
also
just
if
it's
very
simple
you
can
just
have
the
control
c,
as
robert
was
talking
about.
E
There
is
a
a
way
to
do
that.
A
E
C
B
Yes,
it
is
running
asp.net
core,
but
it's
spinning
that
up
in
the
background
and
it
the
the
program
makes
a
request
to
itself,
and
so,
when
that
request
completes
it
automatically
shuts
down,
whereas
our
asp.net
test
app,
it
just
runs
indefinitely
in
iis.
A
Things
to
consider
like
when
we
do
it
when
we
get
to
that
the
easier
solution
will
be
the
baby
instead
of
like
getting
on
to
the
complex
programming
stuff
like.
If
we
agree
I
think
uber
does
it
can
use
that
approach
and
next
thing
what
I
figured
out
in
while
doing
the
test,
even
though
we
call
it
a
startup
hook
test
and
we
are
renaming
it
to
smoke
test
in
those
scenarios
also,
the
profiler
is
getting
attached.
A
E
Why
why
robert
right
this
talk
about
communication
to
the
application
and
stuff
like
that
was
making
me
thinking
some
time
ago
that
perhaps
the
way
to
solve
their
because
we
still
require
admin
for
the
the
windows
container
test,
but
we
could
do
the
collection
of
the
expense
or
methods.
Whatever
we
add
to
the
test,
we
could
do
directly
on
the
container
and
then
the
test
just
extracts
that,
because
we
we
are
able
to
expose
the
endpoints
of
the
container
to
the
test.
E
E
Yeah,
I
will
put
the
hq
and
we
we
no
not
priority
now
later.
E
This
one
is
is
a
very
good
one.
I
I'm
very
excited
about
that
because
I
think
for
the
beta
stuff
and
our
keeping
kind
of
getting
interest
on
the
releases
is
to
add
instrumentation.
So
there
are
some
instrumentation,
some
libraries
that
already
added
activity
source.
E
So
if
we
just
know
the
name
of
some
good
ones
that
are
following
open,
telemetry
conventions
semantic
conventions,
then
we
should
be
adding
those.
I
think
robert
already
opened
an
issue
to
add
one
of
the
databases.
E
C
Regarding
this
issue,
erasmus,
before
maybe
you
ask
questions,
I
was
also
thinking
that
maybe
it
will
be
worth
to
add
some
contributing
guidelines,
how
to
add
a
source
instrumentation,
because
now
you
know
it
the
best
and
personal
and
also
understand
it,
but
I'm
not
sure
how
my
memory
will
work
in.
I
don't
know
a
few
months,
but
right,
I
think
it's
pretty
well
designed
right
now,
but
it's
like
one
needs
to
know
at
least
two
or
three
things
before
doing
it
in
order
not
to
lose
too
much
time.
G
E
I
think
it's
too
helpful
yeah.
I
think
I
think
you'll
be
very
helpful.
You
know
because
somebody
can
come
and
they
have.
They
know
some
library.
That
has
the
thing
and
they
can
already
do
for
us.
We
just
reveal
the
pr
if
they
have
any
instructions.
You
know
it
sounds
and
also,
as
robert
said,
I'm
sure
that
in
two
weeks
I
I
will
not
remember
what
was
the
steps
don't
know
so
having
that
dark
is
good.
E
A
Christmas-
this
is
a
very
good
like
work
on
this
one,
but
I
have
some
like
one
small
question
about
this
one.
I
did
a
review
but
did
not
go
through
the
entire
particularity,
so
we
are
using
a
the
assembly
load.
I
believe,
whenever
we
are
trying
to
load,
we
are
introducing
an
event
to
that.
A
So,
like
we
also
have
in
similar
event
in
the
instrumentation
library,
do
we
know
how
that
gets
hit?
Do
we?
I
know
it's
not
the
assembly
load
there.
We
have
something
different.
A
A
H
G
A
Yeah
maybe
like
before
we
merge,
have
a
take
a
look:
how
these
two
are
getting
invoked
when
some
assembly
is
not
there
or
we
are
trying
to
load,
there
should
not
be
any
like
condition
or
gap
between
these
two
things.
That's
why
I
might
only
worry
both
are
different,
but
can
do
the
same
job
so
that
that's
what
worries
a
little.
But
apart
from
that,
it
looks
very
good.
I'm
going
to
take
a
look
at
this
pr
in
detail
a
little
later
today,.
E
So,
but
at
the
risk
of
saying
something
very
silly,
because
I
didn't
look
at
the
pr
yet
so
this
is
for
the
case
that
we
need
to
do
some.
Some
rigid
on
the
assembly
right.
G
E
G
That's
right,
yeah.
Actually,
the
rigid
is
just
a
functionality
that
we
can
combine
with
this
loading
depends
how
the
instrumentation
needs
to
be
both
stripped.
G
And
I'm
I'm
not
sure
if
that
I'm
following
yeah
it's
getting
complicated
when
you're
looking
how
the
different
instrumentations
are
loaded.
There
are
like
lots
of
different
patterns.
So
that's
why
robert
opened
a
new
issue.
C
C
But
if
the
library
like
supports
enabling
it
like
by
default
or
by
some
static
field,
then
it's
not
needed.
E
So
so,
basically
we
we
have
three
cases
the
ideal
one.
Our
dream
is
kind
of
just
at
the
the
the
achieved
source
and
that's
done.
We
have
a
intermediary
one.
That
is,
we
can
do
how
we
manage
without
reject
that
we
add
the
source
when
something
is
load.
We
use
reflection
to
get
methods
or
something,
and
that's
enough,
we
don't
need
to
do
rigid
and
then
we
have
ridgid.
E
What
we
have
also
to
keep
in
mind
is
that
these
three
distinctions
will
bring,
and
we
are
already
aware
of
that,
but
we
have
to
think
in
ways
to
make
this
clear
is
that
the
third
case
that
requires
rigid
doesn't
work
for
startup
hooks
only
right
exactly
the
other
two,
of
course,
work
for
startup
hooks
only.
E
One
thing
for
a
long
term,
I
think,
is
that
perhaps
we
should
look
into
what
kind
of
support
we
can
give
for
the
second
case,
that
kind
of
when
the
certain
assembly
is
loaded,
we
can
do
to
help
people
right
at
that
level,
because
I
think
it
would
be
easier
for
us
to
get
more
instrumentation.
That
way.
G
Yeah,
actually,
from
the
yesterday's
total
seat
meeting,
the
general
idea
is
to
wait
when
the
instrument
not
instrumentation,
but
actually
the
client
libraries
are
going
to
implement
activity
source
directly
into
their
libraries.
So
then
we
just
need
to
subscribe
to
the
source
syntaxes.
C
But
the
erasmus,
but
in
theory
yes,
but
still
the
second
approach-
may
be
needed,
because
because
some
of
the
libraries,
for
example,
are
naming
stuff
differently
than
the
hotel
semantic
conventions.
So
there
there's
this
listener,
which
makes
the
translation
and,
for
example,
replaces
the
the
the
tags
etc
sure.
C
E
Okay,
I
think
I
at
least
have
a
better
contextual
to
review
and
kind
of
now.
I
understand
raj's
question
about
it.
G
Your
side
yeah,
actually
my
biggest
worry,
is
not
the
sex
question
about
assembly
loading.
G
I
Yeah,
so
I
I'm
thinking
that
this
could
really
only
happen
if
it
was
something
like
editor
functions
or
something
where
they,
the
user
application,
has
set
up.
Multiple
assembly
load
contacts,
because
in
this
case,
if
we're
just
detecting
assembly
loads,
that's
going
to
fire
for
each
time.
It's
loaded
into
a
context,
I
believe.
I
Actually,
I
might
need
to
double
check
that
and
I
think
the
other
concern
I
had
was
if
we
end
up
absolutely
loading
two
versions.
I
know
we
have
these
different
mechanisms
to
try
and
unify,
but
if,
for
some
reason
we
end
up
loading
a
separate
version
that
could
also
cause
a
second
load,
and
so
I
just
wanted
to
see
how
we
could
just
make
this
detection
resilient
to
it,
and
if
we
need
to
even
keep
a
count
or
not.
I
Not
really
not
the
moments.
G
I
have
just
expensive
options,
so
basically
you
know
what
are
the
default?
What
are
the
libraries
in
the
default
load
context?
So,
basically,
if
they
now
the
event
happens,
then
you
can
check,
but
you
need
to
enumerate
all
the
libraries
there
and
compare.
I
I
guess
so
this
this
method
and
this
detection
just
to
clarify,
so
this
is
used
to
make
sure
that
we
have
like
all
the
dependencies
correct
so
say
like
you
have
like
or
something
and
it
has.
It
relies
on
this
other,
like
options,
sort
of
configuration,
and
so
you
want
to
make
sure
like
each
of
those
different
assemblies
is
loaded
before
you
try
to
run
the
instrumentation.
Is
that
what
this
is
trying
to
accomplish,
or
am
I
misunderstanding,
so
can.
I
Yeah,
so
I'm
trying
to
make
sure
I
understand
what
this
piece
of
code
is
doing
and
why
we're
keeping
track
of
the
there's
an
array
of
required
assemblies.
You.
G
G
C
G
I
Like
we
saw
the,
I
think,
example
where
there's
the
mongodb
library
itself
and
then
there
is
a
hotel
adapter
as
well,
and
those
are
two
separate
packages.
G
E
Yeah,
but
I
I
see
what
what
you
mean,
but
okay
for
the
moment,
the
b
to
work
we
need
to,
but
that
is
the
case
for
reject.
So
it's
like
when
we
load
the
one
from
the
user,
then
we
load
the
one
that
adds
the
activity
source.
We
inject
that
one
right.
E
So
I
think
I
think
robert's
question
is
pertinent.
Like
do
we
have
a
concrete
case
that
we
need
more
than
one
because
you
are
dealing
with
assembly.
Load
assembly
load
usually
corresponds
to
the
module.
I
I
I
don't
remember
seeing
any
practical
thing
with
assembling
multiple
modules,
although
I
think
it's
possible.
E
Do
we
have
a
concrete
case
that
for
these
usage,
that
we
do
with
a
reflection
or
something
that
there
is
more
than
one.
I
No,
I
I
think,
even
in
the
case,
where
robert's
kind
of
navigating
around
right
now
our
detection
could
be
the
like
the
listener,
because
I
what
I'm
imagining
is
as
long
as
we
can
guarantee
that,
then
we
can
use
reflections
to
generate
the
the
like
options.
Whatever
is
needed
passed
to
the
like,
the
tracer
provider
builder.
C
C
E
Does
it
make
any
sense,
but
but
this
is
is
the
kind
of
this
is-
is
the
very
specific
case
that
we
need
to
do
regit
and
during
the
widget
we
inject
this
assembly.
We
load
this
assembly,
we
we
are
going
to
tweak
the
load
of
this
assembly.
So
that's
it.
E
I
If
yeah,
if,
if
I'm
interrupt,
I
I
think,
even
though
it
turns
out
mongodb,
we
are
doing
the
rejecting
like,
in
this
case,
they're
creating
activity
source
here.
So
this
is
where
they
would
be
emitting
those
events.
So
if
this
were
any
other
library,
these
would
be
probably
the
events
we'd
be
consuming,
and
hopefully
we
wouldn't
need
to
do
any
regions.
I
G
The
case
with
the
mongodb
is
a
bit
different
because
in
a
mongodb
you
don't
have
the
dll
available.
You
need
to
inject
it,
but
with
the
asp.net
core,
you
have
to
dll
the
waiver,
but
you
need
to
check
if
these
are
loaded.
B
So
so,
just
to
try
to
clarify
here
so
there's
a
separate
package
that
mongodb
uses
to
generate
these
activities
and
we're
responsible
for
loading
that
package.
But
we
can
only
load
that
package
once
some
other
library
or
two
of
mongodb
is
loaded
and
the
unknown
to
me
is
if
it's
only
one
library
we're
waiting
for
or
if
it's
two
of
them,
it's.
E
It's
one
one
assembly,
one
assembly
in
this
case
corresponds
to
one
module,
so
one
dll.
G
E
Yeah,
so
the
application
just
have
a
reference
to
the
the
first
library
and
we
are
just
adding
one
on
top,
but
we
do
that
vrg.
So
going
back
to
the
original
question
is
kind
of.
There
is
a
case
that
we
need
to
count
or
a
set
of
assemblies
or
for
the
reflection.
Okay,
it's
not
rigid
that
we
need
multiple
assemblies
to
trigger
some
or
not.
You
know.
E
The
discussion
is
very
good,
very
interesting,
but
in
interest
enough
time
I
think
we
need
to
move
ahead.
Hello.
K
And
us
faster
yeah,
so
basically.
E
Just
one
second,
I'm
gonna
give
you
time,
but
we
are
ready
at
45
minutes
and
we
are
still
on
the
pr's.
So
please
be
brief,
so
we
can
move
to
the
other
stuff.
G
Okay,
yeah,
so
basically
the
list
what
you
were
received
there
looking
for
the
asp.net
core
http
library
is
coming
from
the
instrumentation
here
is
the
instrumentation
operating.
So
the
instrumentation
itself
has
a
dependency,
which
is
the
agency
listener,
and
here
you
can
see
the
third
party
libraries
that
needs
to
be
available
when
we
create
it.
Otherwise
it
will
crash
if
these
libraries
are
not
available.
K
E
Okay,
I
I
would
let's
get
back
at
using
the
pr
yeah.
I
I
think
I
understand
why
you
are
getting
that,
but
I
think
I
have
other
things
to
say
about
I'll
get
back
to
that
on
the
pr
sure.
C
So
here
are
the
issues
I'm
still
starting
from
the
bottom,
so
he
was
a
feature
request.
I
don't
know
how
many
of
you
have
seen
it.
C
E
I
A
E
Yeah,
I
I
really
don't
like
start
an
application
broken
in
a
broken
monitoring
situation
and
we
have
things
to
disable
if
they
know
that
for
some
time
they
want
start
the
application
without
monitoring
without
observability
signals.
So
I
tend
to
think
okay.
If
we
have
that
escape
valve
it's
fine,
but
in
general,
if
we
have
really
broken
configuration,
it
should
not
start
the
application.
C
E
Yeah,
I
think
this
is
something
that
we
discussed
some
time
ago
and
that
that
has
been
the
pattern.
So,
let's
close
the
issue,
then
they
we
have
one
agreement
there.
With
your
comment
and.
C
So
this
one
I
created
after
I,
I
was
simplifying
working
on
the
integration
test
and
I
think,
removing
one
step
or
something
and
everything
started
to
be
unpredictable.
C
F
C
One
which
I
discovered
during
writing
integration
tests,
so
the
this
is
something
which
chris
initially
like.
I
think
that
it
was
intended
that
it
was
not
implemented
up
front
because
it
was
too
much
right.
Chris.
C
C
Because
it
like
it's,
I
would
say
that
it's
a
backing
hour,
at
least
in
our
you
know,
it's
not
specified
in
the
configuration.
B
Yeah,
it's
not
specified
implementing,
it
will
allow
for
symmetry.
I
don't.
E
E
Yeah-
but
I
I
this
is
one
of
the
things
that
I
think
we
can
live
a
long
time
without
you
know,
because
there
is
always
the
option
of
setting
the
environment
variables
per
process
or
globally
or
control
with
the
enable
disable.
So
I
kind
of
feel
that
we
can
live
a
long
time
without
this
one.
B
C
This
is
an
issue
which
I
created
yesterday
just
before
joining
the
net
submitting,
and
it's
basically
about
reviewing
the
current
instrumentation
library
guidelines
like
from
our
perspective
and
cj,
also
said
that
if
there
is
something
that
will
make
our
lives
easier,
then
we
should
probably
specify
and
look
at
it.
C
Cj
also
said
that,
unfortunately,
it
looks
like
many
people
does
not
follow
this
this
this
con
these
guidelines,
but
to
be
honest,
they're.
They
are
very
well
like
positioned
into
the
locks.
It's
not
for
sure
right
now.
It's
not
the
the
issue
that
it's
not
visible.
These
guidelines
are
like
on
the
main
rhythm,
there's
a
hyperlink
to
it,
so
I
think
we
can
just
revisit
what's
there
and
that's
it,
and
I
prefer
to
put
it
sooner
than
later,.
E
Yeah
yeah
we
we
should
definitely
do
it
later
sooner.
I
was
gonna
say
that
perhaps
we
should
divide
between
ourselves
kind
of
look
at
contribute
stuff,
and
perhaps
we
we
should.
You
should
book
it
here,
kind
of
okay,
I'm
gonna
look
at
one
or
two,
and
we
list
here
on
the
issue
of
the
one
that
we
look
at
so,
for
instance,
if
there
is
anything
that
is
just
something
we
could
just
include
already
in
our
instrumentation.
E
You
know
if
it's
something
like
the
the
the
the
sequel
one
listed
here,
we
should
be
just
adding
for
our.
C
Release
you
know
I
would
prefer
to
do
it
in
other
way.
I
would
propose
first
maybe
to
make
a
list
of
instrument,
libraries
that
we
would
like
to
instrument
and
then
just
focus
on
the
ones
that
are
like
the
one
that
we
mostly
need,
but
I'm
not
sure
how
we
can
get
this.
I
don't
know.
Maybe
in
the
neural
link
or
data
took,
you
have
some
met,
some
metrics,
which
libraries
are
the
most
frequently
used
or
something
like
that.
E
I
think
that's
like
how
long
this
issue,
so
we
keep
more
manageable.
You
know,
but
this
gives
us
a
priority
list,
which
is
good,
but
I
think
the
process.
It's
still
the
same.
You
know
whatever
one,
you
listen.
Look
you.
If
you
find
stuff
open
the
the
the
bugs
on
the
contrib
side
and,
depending
on
the
case,
we
we
can
already
incorporate
any
of
those
instrumentations
that
are
just
like
adding
just
the
active
source
thing.
B
I
know
we
had
some
issues
where
we
were
tracking,
which
libraries
use
diagnostic
source
versus
activity,
source.
H
C
So
it's
an
issue
created
today,
for
I
would
like,
if
someone
else
take
a
look,
because
I'm
not
sure
if
I
understood
correctly
the
issue
or
the
ask,
but
it
looks
for
me
like
it's
more
more
related
to
the
sdk
at
least
it's
more
the
things
that
they're
asking
for
and
also
I
think
that
they're
missing
some
context
here
or
not
everything
is
like
precisely
described.
I
did
my
best
to
answer
here
to
give
the
the
rot.
C
B
Yeah,
which
it
kind
of
without
understanding
the
use
case,
my
gut
tells
me
they're
they're,
trying
to
solve
the
the
problem
in
you
know
not
as
good
manner
and
it's
almost
as
if
they
want
that
x,
request
id
to
be
part
of
their
baggage
and
just
use
w3c
or
b3
to
to
get
normal
context.
Propagation
going.
E
Yeah
I
so
so
robert
you
you
look
at
it
seems.
Did
we
have
any
reply
after
your.
C
No,
I
think
it's
good
we'll
wait
a
few
days
and
I
will
ask
if
it's
clear
and
you
can
close
it,
make
things
sound
good
to
me
and
now
I
have
added
two
two
issues
to
add
from
my
experience
to
most
popular
sq,
like
the
relational
databases
used
right
now
in
dot
net
core
micro
services,
which
is
possible
sql,
client
and
mysql
client,
and
I
wanted
to
add
it
like,
as
the
same
for
like
for
like
for
grpc
for
the
next
milestone,
especially
that
both
of
them
should
be
easy
to
to
to
implement.
C
E
Yeah,
it's
just
synthetic
synthetic
sugar
for
people
to
kind
of
do
in
their
tracer
initialization.
So.
C
Thing
is
that
the
only
thing
in
that
package
I
haven't
found
anything
more,
and
I
think
also
that
is
what
the
guidelines
are,
are
proposing
to
create
such
package
for
convenience.
If
one
does
not
need
to
know,
you
know,
search
for
these
streams.
C
Here,
there's
no
instructions
that
anything
more
needs
to
be
done.
When
I,
when
I
quickly
quickly
look
at
the
source
code,
I
haven't
found
that
you
need
to.
I
don't
know,
enable
something
in
the
in
the
connection
object
or
something
like
that,
which
is
the
case
for
the,
which
is
the
case
for
the
mongodb.
C
E
C
E
All
right,
I
think
I
think
we
already
went
over
the
hour.
E
We
have
good
stuff
with
that
very
good
work
on
test
infrastructure.
There
is
still
more
stuff
to
go
to
go
over
the
the
improvements
in
general,
but
I
think
now
we
are
getting
to
a
moment
that
we
can
perhaps
go
a
little
bit
focus
on
trying
to
catch
the
easy
instrumentations.
E
So
we
add
more
coverage
and
make
more
interesting
for
people
to
keep
trying
the
the
distribution.
I
think
I'm
looking
forward
for
when
I'm
backing
for
a
vacation
having
the
new
release
the
new
beta
release.
E
By
the
way
I
will
be
out
for
two
weeks
so
see
you
guys
sometime
in
july,
and
one
thing
that
robert
was
already
aware-
and
perhaps
I
was
the
one
that
is
slipping
on
that,
but
I
got
a
little
bit
concerned
is
that
we
have
a
glaring
ci
gap
on
the
combination
of
windows,
specifically
what
wordsmith.net
framework
and
integration
test
that
requires
linux
docker.
E
So
as
a
step
as
a
workaround,
let's
say,
or
a
measure
for
now,
robert
added
to
the
release
that
we
should
run
locally
those
tests
to
be
sure
that
things
are
working,
but
I
would
rather
kind
of
try
to
see
if
we
can
come
up
with
some
creative
solution.
The
the
issue
is
that
github
actions
don't
support
our
runner,
because
the
runners
are
red
virtualize
it.
We
need
something
that
runs
windows,
but
the
container
visual
virtualization
should
be
linux
right.
E
So
we
we
should
look
at
ways
to
solve
that.
You
know,
okay,
that
that
was
what
I
want
to
say
before
the
meeting
ending.
Do
you
guys
want
to
bring
anything
up.
E
D
Like
yeah
thanks
very
much,
I'm
kind
of
getting
engaged
with
a
bunch
of
different
hotel
groups
and
and
doing
a
breadth
first
kind
of
hang
out
for
a
week
or
two
to
see
where
it
makes
the
most
sense.
But
thanks
all
right.
K
D
By
the
recent
profiling
work,
that's
happening
around
all
of
that,
so
I
have
some
experience
with
net
and
it's
performance
and
diagnostic
tooling
generally,
and
so
I'm
super
interested
just
as
a
nerd
in
this
context,
but
all
right.
D
E
That
that
that's
great
to
hear
eventually
hotel
gaps
do
the
profiling
here
too,
so
we
we
will
be
happy
to
have
you
working
for
us
on
that.
E
On
that
note,
for
me
see
you
guys
in
three
weeks
basically.
D
Oh,
I
I
had
one
brief
question:
if
it's
not
too
too
late
long
ago,
like
20
years
ago,
15
20
years
ago
on
in
the
microsoft.net
early.net,
but
really
preceding
it,
there
was
both
statistical
profilers
but
also
call
attributed
profiling.
You
know,
like
prologue
epilogue
hooks,
and
things
like
that
is,
is
any
of
that
covered
in
the
existing
hotel
instrumentation
for
net?
I
haven't
done
a
deep
dive
yet,
or
is
it
primarily
about
auto
instrumenting
for
trace
bands?
Only
okay.
E
At
this
time
is,
is
just
auto
instrumentation
to
add
the
activities
that
map
to
expands
and
metrics,
eventually,
as
hotel
expands
to
profiling,
then
we
move
to
that
area
we
have
been
have.
Microsoft
has
been
stand
out
on
the
collaboration
with
the
open,
telemetry
project.
E
So
when
you
get
to
profiling,
I
do
expect
a
lot
of
collaboration
with
the
folks
from
microsoft,
either
leveraging
what
they
have
for
profiling,
runtime
or
otherwise,
and
I
think
nowadays
what
microsoft
has
for
profiling
is
more
on
the
sampling
side,
because
that
the
the
one
that
added
the
probes
was
the
old.
I
forget
the
name,
but
they
were
right,
yeah.
Well,
I
I
know
I
remember
what
they
did.
They
didn't
do
sample
profile.
They
they
they
add
the
probes
to
everything,
and
you
collected
everything
right.
E
D
Yeah,
of
course,
and
and
and
we're
over
time,
but
the
profiling
efforts
as
I've
seen
so
far,
those
four
meetings
so
far,
three
or
four
meetings,
the
next
one's
tomorrow,
it's
all
sample
based
profiling.
You
know
prof
and
all
of
this
I
was
just
kind
of
curious
if
the
hotel,
auto
instrumentation
sig
in
the
past,
had
looked
at,
what's
also
possible
using
some
of
the
net
magic,
that's
a
little
tougher
to
implement
on
other
other
platforms.