►
From YouTube: 2022-11-08 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
B
B
I
moved
to,
we
were
renting
a
house
now
in
the
suburbs
we
had
our
sun,
so
I
have
a
baby,
I
think
baby.
How
old
is
Reuben
is
six
months
old
now
so.
A
C
B
A
It's
it's.
B
B
C
C
Wow
we
should,
we
should
chat.
Do.
A
You
guys
all
I'll
Shopify
right.
B
A
I'm
I'm
an
engineer
on
the
Ruby
Tracer
at
datadog
and
oh
nice,
we're
gonna
we're
gonna,
put
our
money
into
hotel
and
then
so
I
was
like
I
might
as
well.
Just
join
the
conversation,
so
I'll
see
see
what's
happening.
Yeah.
D
A
C
C
C
A
Does
that
mean
what
does
that
mean?
It
means
they're
just
sending
a
bunch
of
cash
too.
It
means
we're
putting
engineering
effort
into
they
keep
all
languages
operating
engineering
effort
into
Hotel
awesome.
B
B
Maybe
out
as
well
and
then
Francis
usually
doesn't
make
it
yeah
I'm
happy
to
I'm
happy
to
talk
around.
You
know,
do
the
usual
song
and
dance
that
we
do
here
yeah
please.
So
we
can
wait
for
a
couple
more
okay,
I,
don't
think
what
else
is
coming.
Yeah.
B
Okay,
what's
up
everyone,
my
name
is
Eric
I'm
container
of
the
contrib
repository
here
at
open
planetary
Ruby.
You
sometimes
we
do
introductions
when
there's
new
people
I
can
just
do
them.
I'll
just
do
them
for
everybody,
Sam
Handler,
Andrew
Hayworth
or
my
esteemed
co-workers.
At
Shopify,
along
with
myself,
Marco
Costa
esteemed
former
co-worker
at
datadog.
He
maintains
the
Ruby
tracing
DD,
Trace
RBA,
along
with
all
the
other
cool
things
that
datadoc
does
and
we
should
just
switch
places.
I
work
for
a
Canadian
company.
B
C
C
B
Cool,
usually
we
do
like
a
quick,
we'll
do
a
Matt
who
Matt
Ware
from
lightstep,
who
is
one
of
the
maintainers,
usually
does
like
a
spends
like
the
first
10
minutes,
just
doing
a
overview
of
the
the
previous
meeting
to
this,
which
is
for
the
was
it
the
maintainers
meeting
or
the
specification
meeting
thanks
Sig
the
spec
cig,
which
is
sexy,
often
has
points
made
that,
like
you,
know,
trickle
down
to
the
individual
languages,
sometimes
they're
just
bike
sheds,
so
we'll
just
go
over
it,
and
then
we
usually
then
yeah
after
that
we
sort
of
have
like
the
rest
of
the
meeting
open
for
any
urgent
issues
or
things
specific
to
the
Ruby
Library.
B
So
yeah
I'll
share
my
screen.
I
guess
do
I
have
anything
incriminating
the
careers
page.
One
second.
B
Can
can
you
all
see
my
yep
nice
stuff,
great
good,
great
grand,
wonderful,
no
yelling
on
the
bus,
they
canceled
the
specsig
I?
Guess
today,
yeah!
Oh
wait!
No
here
it
is
I
lied,
I,
lied,
I.
B
Oh
okay,
cool,
so,
let's
see,
looks
like
trailer.
Chase
collection
was
discussed.
What's
that
probably
is
relevant
for
us.
Let's
look
at
what
that's
all.
C
A
flavor
of
like
the
Helios
thing,
I,
don't
know,
don't
listen
to.
E
C
B
Yes,
I
mean
looks
like
there's.
Some
they're
asking
for
more
reviews
seems
like
a
relatively
contentious.
If
anyone
has.
E
E
B
Historical
context,
if
you've
tried
to
be
pretty
not
too
aggressive
with
putting
shoving
stuff
onto
a
span
such
but
people
tend
to
want,
like
the
message
body
and
stuff,
like
that,
we've
kind
of
nudged
users
to
bring
their
own
instrumentation
in
those
cases
or
their
own
middle
waves.
B
Yeah
there
you
go,
but
you
know
if
you
just
Intuit
it
it's
fine,
okay,
some
okay,
some
jvm
metrics!
B
No
one
cares
great
they've
added
some
some
runtime
jvm
metrics
I,
guess
that
are
collected
for
Java,
probably
not
relevant
for
us.
Maybe
two
I,
don't
know,
probably
not
relevant
well.
A
He
was
always
updating
it,
but
he
always
up.
He
also
updates
truffle
Ruby
right,
like
spends
a
lot
of
time
with
124
Ruby
I,
don't
know,
I
see
we
we
maintain
it
and
I,
see
I,
see
a
bunch
of
releases
and
I
see
people
using
it
every
now
and
then
I
don't
know.
Yeah.
E
Sorry
I
didn't
I
didn't
mean
to
imply
that
it
wasn't
used
more,
that,
like
I,
think
they're
stuck
on
Ruby
2,
7
compatibility,
I
I,
think
it
was
the
last
I
understood
and
I
was
wondering
like
why
they
hadn't
moved
on
and
I
I
was
kind
of
curious
about
it.
I
guess.
But
it's
also
not
on
topic,
so
we
can
be
quiet.
E
B
Me
yeah:
no,
it's
a
good
question,
especially
because
we
maintain
some
compatibility
with
JR
would
be,
although
I
don't
think
Kitty
thing
super.
E
A
E
My
question,
because
we
we
do
sort
of
limit
ourselves
to
Ruby
2
7
compatibility
primarily
for
James,
would
be
at
this
point.
So
you
know
I'd,
like
it'd,
be
nice
to
be
able
to
use
newer
language
constructs
and
stuff.
As
you
know,
Christmas
is
coming
up.
So
we're
about
to
have
a
new
Ruby
version
of
one
drops
off
support,
and
you
know
that
stuff
that
is
kind
of
actually
why
I
asked
it,
but
it's
okay,
I'll,
look
into
it
more.
My
end.
B
Yeah
I
think
I,
don't
know
offhand
I'm
sure
Charles
Nutter
is
super
responsive
on
Twitter.
Also,
if
you're
ever
quick
answer,
lazy
web
style
anyway,
I'll
just
keep
going
because
I
don't
think,
there's
anything
for
us
actually
won't
hear.
But
if
you
pull
up
jvm
metrics,
there's
some
more
in
the
specification
now
that
you're
suggested
to
grab
this
one's
kind
of
relevant
to
us,
which
is
untraced
functionality,
looks
like
Tristan
from
the
erlang.
Sig
comes
in.
B
Basically
there's
a
bunch
of
I
think
the
context
is
there
are
a
bunch
of
non
there's
like
nothing
in
the
specification
saying
like
here's,
a
helper
method
to
not
trace
this
code.
You
know
to
to
you,
know,
drop
the
context
and
the
implementations
vary.
I
think
whether
the
untraced
block
you
know
when
a
new
child
is
created,
whether
that
also
loses
the
context
whether
that
starts
a
new
Trace,
the
naming.
So
it
sounds
like
there
was
some
discussion
around
just
like
what
should
should
this
be
something
in
the
spec.
B
What
should
it
look
like?
How
does
it
work
with
the
specifications
for
sampling
and
yeah
I
mean
I?
Think
ours
is,
has
some
bugs
in.
It
is
a
little
hacky,
but
I
have
I,
don't
know
enough
about
the
others,
but
seems
like
there
might
be
something
we
need
to
look
out
for
here
in
the
future.
I
don't
know
if
there's
like
a
yeah.
A
B
E
Well,
it
would
be
nice
if
it
gets
into
the
API
spec,
because,
right
now
we
can't
really
our
the
Ruby
hack
kind
of
relies
on
sampling
in
a
way
that
doesn't
really
always
work
correctly.
And
then
you
know
other
ways
doing
it
would
be
putting
it
in
the
context,
but
like
every
Tracer
would
need
to
respect
that.
E
B
B
Something
stuff
gets
merged,
but
yeah,
probably
for
the
best
to
have
some
standards
on
this
and
super
valid
use
case.
Given
that
we
use
it
all
the
time.
Okay,
configuration
working
group,
so
I
I
think
this
is
open
by
Alex.
Boatson
could
boaten
from
split
step
response.
B
Yeah
I
think
they
just
want
to
make
a
formal
working
group
for
configuration
options:
language
agnostic
configuration
which
yeah
there's
already
in
the
specification.
There's
a
bunch
of
there
is
a
bunch
of
like
specification
level
configuration
options,
but
it
sounds
like
different
languages
have
gone
to
different
lengths
to
enable
I,
don't
know
like
per
instrumentation
configuration
and
things
like
that.
B
B
E
Yeah
that'll
actually
be
nice
I
think
they
primarily
the
only
configuration
really
specified
today
in
the
environment,
which
is
still
not
even
very
specified.
Yeah.
B
Yeah
yeah
it
it'll
be
it'll,
be
really
nice
when
it
just
is
magically
done.
We
don't
have
to
do
any
work
yeah
anyway,
I
think.
If
anyone
has
strong
opinions
about
config,
they
should
look
like
they're
looking
for
folks
to
be
involved
in
this
sig
so
definitely
drop
on
into
this
thing.
It's
unclear
when
they're
meeting
every
on
Fridays
at
in
the
afternoon.
I'm
sure
okay
sounds
that's.
Everyone
wants
to
do
that
for
a
low
Friday
meeting
on
the
calendar,
great
okay,
that
was
it
there.
You
go
that.
B
Also
for
context
these
are
recorded
and
automatically
uploaded
to
YouTube.
So
please
don't
make
any
other
Tito's
jokes
that
you
wouldn't
be
comfortable
living
on
forever
on
the
internet,
all
right,
yeah
I,
don't
you
know,
are
there
any
so
I'll
share
my
screen
again
sure.
B
Are
there
any
other
issues
or
PR's?
We
want
to
talk
about.
C
C
Yet
it'll
be
a
no-op
basic
I
mean
if
you
don't
install
this
gym,
then
you're
not
going
to
have
the
stuff,
and
it's
nothing
I
think
that's
where
the
trace
race,
oh
God,
damn
what
Trace
ID
ratio,
what
there's
like
some
sample
that,
like
some
sampler
in
there,
an
experimental
sampler
and
then
also
where
we
put
the
trace
response
propagator
in
there
and
so
we're
trying
to
get
a
version
out
of
that,
so
that
we
can
use
it
we're
going
to
try
to
add
Trace
response
as
a
optional
configuration
piece
to
the
rack,
instrumentation,
so
that'll
obviously
defaults
to
off,
but
Step
One
is
getting
this
so
I
just
I,
don't
know
PSA!
C
This
gem
exists.
If
you're
trying
to
do
something
off
spec
or
sort
of
on
the
bleeding
edge
of
the
spec,
you
might
want
to
shove
it
in
there
and
that's
that's
the
announcements.
B
Cool
yeah,
I,
guess:
okay,
so
it
sounds
like
the.
B
Okay
and
consistent
parent,
consistent
probability,
okay,
so
yeah
I
think
for
some
context.
If
I
as
I
understand
correctly,
some
of
the
probability
sampling
formats
and
structure
in
the
specification
took
a
lot
pretty
long
time
to
like
get
merged
in
getting
merged
in
after
the
1.0
release.
If
of
the
broader
like
tracing
stuff.
B
So
that's
why?
So
now
there's
some
sampling
specification
coming
out,
but
it's
not
technically
part
of
like
the
1X
release,
it's
I
guess
for
after
so
it
can't
I,
don't
know
I,
guess
we're
or
it's
still
being
discussed.
So
that's
why
it
lives
in
experimental,
but
they'll
probably
eventually
make
their.
B
The
regular
SDK
they're
pretty
standard,
just
like
here's,
how
you
receive
a
probability
from
an
upstream.
You
know
from
your
other
service
and
Pro.
Here's
how
you
propagate
it
to
the
next
Service
and
oh
I,
don't
know:
I,
don't
I'm,
making
all
this
up
as
I
go
so
I!
Think
Sam's.
E
Description,
you're,
mostly
correct
I,
don't
know
exactly
why
I
wasn't
really
as
stable,
but
one
of
the
real
reasons
it's
still
in
the
experimental
SDK
is
not
because
of
its
Upstream
status,
but
more
that
it
doesn't
have
enough.
Testing,
Francis
and
I
were
talking
about
it
when
he
was
writing
this
and
basically
there's
no
way
to
exhaustively
test
that
it's
correct.
E
They
recommend
basically
like
running
a
bunch
of
fake
traces
through
it
and
observing
the
proportion
that
was
sampled
and
not
to
determine
correctness,
and
we
did
not
build
that,
so
we
aren't
100
sure
on
it.
I
think
that's
the
main
reason
it's
in
the
experimental
directory
at
the
moment.
B
E
B
It's
like
it's
like
powers
of
two
or
something
or
base
two,
it's
sort
of
like
there's
some
base,
and
then
you.
C
B
B
So
it
uses
like
this
Trace
State.
It's
not
there's
sort
of
like
a
Trace
State
header
that
has
sort
of
lives
as
like
a
grab
bag
for
these
sort
of
things,
so
you'll
find
yeah
the
probability
within
the
Trace
State
header
and
it
gets
extracted,
and
then
there
I'm
curious
I,
actually
don't
know
the
other
one
that
I
saw
William
shipped
Upstream.
It's
like
Trace
response.
What?
What
exactly
is
that
one
doing.
C
So
great
question
Eric
the
the
editor
his
draft,
so
the
current
editor's
draft
of
the
trace
contacts
specification
supports
the
concept
of
deferred
sampling
in
the
load
balancer.
So
the
idea
here
is
that
you
have
your
load
balancer
spans
in
a
state
of
limbo.
Your
load
balancer
is
just
sort
of
holding
on
to
spans
with
an
indeterminate
sampling
decision
and
then
the
trace
response.
Header
comes
back
from
the
the
service
that
the
load
balancer.
You
know
forwards.
C
Traffic
to
confusingly
called
the
Upstream
service
in
nginx,
even
though
you
would
generally
think
of
it
as
Downstream,
but
you
receive
the
trace
header
back
in
the
load
balancer,
and
then
you
can
make
your
sampling
decision
based
on
that.
That's
the!
That
is
what
the
trace
response
propagator
is
intended
to
support
it's
kind
of
confusing,
because
you
would
never
want
to
register
it
as
a
propagator
as
their
current
designed,
because
that
just
doesn't
make
any
sense
because
you're
yeah
right
so
yeah.
That's
what
that
is.
C
There's
there's
a
there's
an
issue
open
if
anyone
cares,
there's
an
issue
open
in
the
specification
repa
where
I'm
like.
How
are
you
supposed
to
do
this
like?
What
is
the
spec
compliant
way
to
do
this
on
the
load
balancer
end?
Are
you
supposed
to?
You
know,
alter
the
sampling
decision
on
these
fans
after
they've
been
created
which
is
illegal
or
anyway,
that's
not
related
to
review,
but
yeah.
E
My
suspicion
is
that
the
hotel
spec
was
kind
of
wrapped
up
before
the
W3
spec
introduced.
This
I
think
otherwise.
C
And
there's
there's
arguing
back
and
forth
about
entry.
You
know
I've
already
I've
already
yelled
at
you
about
this,
but
now
that
I've
started
the
sentence.
The
the
people
there's
an
Otep
for
supporting
different
Hooks
and
spam
life
cycle
to
update
the
sampling
decision.
That
is
it's
like
two
years
old,
so
I
am
not
planning
on
it
being
finished.
B
So
yeah
cool,
thank
you,
okay.
That
makes
sense.
So
you
get
some
some
load.
Balancer
thing
you
most
of
those
spans
are
kind
of
junky,
but
you
don't
really
know
which
ones
are
junk
and
which
ones
are
good
stuff
until
you
actually
go
to
the
application
or
whatever
the
next
service
is
and
like
make
a
determination,
and
this
thing
helps
you
communicate
that
information
back
to
your
you
know,
nginx
or
whatever
thingy
in
front
of
it,
and
then
how
you're
actually
supposed
to
then
drop
or
modify
those
spans
is
totally
not
specified.
B
But
that
sounds
like
tomorrow's
problem,
because
now.
E
Are
there
also
yeah,
oh
sorry,
I
I'm?
Sorry,
sorry,
no.
E
I
was
gonna.
Ask
santa
if
there's
also
use
cases
where
you
might
like
make
a
request
to
a
third
party
service
and
then
get
back
a
response
with
a
trace
response
header
in
it,
and
then
you
make
some
decision
in
your
infrastructure
based
on
that.
Is
it
only
for
load.
C
Balancers
or
other
uses
you're
you're
too
smart,
yeah,
I
I
was
I.
Think
there
probably
is
some
stuff
that
you
might
want
to
do
with
third-party
services
and
Trace
response,
but
at
as
presently
constituted
the
the
spec
is
the
only
use
case
that
it
enumerates
is
the
load
balancer,
it's
kind
of
weird
to
think
about
this
happening
in
Ruby
applications
like
if
you
know
service
a
goes
to
service.
C
B
goes
to
service,
C,
I,
guess
there's
a
world
where
you
might
want
to
make
sampling
decisions
based
on
whether
a
downstream
service
is
sampled
but
I
think
yeah,
the
the
hopefully
by
having
the
propagator
there
we
can
just
like.
Let
the
you
know:
let
the
world
of
people
doing
tracing
stuff
figure
out
how
this
might
be
useful,
I,
don't
know.
Was
that
an
answer?
Your
question
Andrew
not
really
I
would
say.
I,
don't
know.
I
haven't
thought
about
that
about
that.
But
I
will
think
about
that.
Now,
that's
my
answer.
I.
B
Cool
awesome.
Well,
if
there's
any
other
I
think
that
wraps
up
discussion
of
the
experiment,
SDK
release
are
there.
Other
issues
appears
people
want
to
discuss.
Are
they
burning
questions
or
yeah
I,
don't
know?
Are
they
just
like
like
yeah
Marco?
This
is
your.
You
know
like
Are
there
specific
things
you
want
to
get.
B
A
Like
for
me,
I
was
just
trying
to
understand
like
what's.
The
I
did
see
a
few
recordings
from
from
the
past.
I
was
just
trying
to
understand
like
what's
what?
A
What
do
you
guys
do
here
like
start
to
figure
out
the
lay
of
the
land
to
see
if
we
can
help
going
forward?
Yeah
we're
gonna
start
doing
work
ourselves,
so
we
support
open
Telemetry
and
we
might,
we
might
find
like
edge
cases
or
cases
that
are
not
supported
and
yeah
just
trying
to
open
a
Communication
channel.
So
we
can
can
work
together
if
there's
anything
that
intersects.
Hopefully
there
is
stuff
going
forward.
D
Had
a
chance,
there
should
be
a
contributing
guide
and
if
there's
missing
information
in
there,
please
let
us
know.
But
essentially
you
know,
we
have
a
series
of
open
issues
that
we
need
help
with,
in
particular
I'm
speaking
on
behalf
of
the
Ruby
contrib
section
of
it
so
like
our
repositories,
are
split
into
right
now,
there's
whatever
is
going
to
be
part
of
the
official
SDK
and
that's
going
to
be
up
in
open,
Telemetry,
Ruby
right
now.
D
What's
ongoing,
there
is
metric
support
and
as
Sam,
to
show
that
PR
for
the
trace
header
for
the
trace
response,
header
and
then
everything
else
is
in
the
Ruby
contribute,
so
we're
talking
third-party
propagators
or
I
should
say
propagated
for
for
non-otel
propagation
style
instrumentations
for
for
popular
Ruby
libraries,
as
well
as
some
resource
detectors,
and
for
our
open
issues.
There
should
be
some
that
are
tagged
with
you
know,
bug
or
help
wanted
or
new
a
first
issue
that
are
all
helpful.
D
We
unfortunately
don't
have
like
a
very
comprehensive
like
instrumentation,
author's
guide
or
anything
like
that,
but
we
are
do
our
best
to
like
kind
of
help
guide
the
way.
That's
those
are
the
kinds
of
things
that
you're
interested
in.
We
certainly
could
use
help
with
documentation
with
auditing
the
things
that
will
make
us
spec
compliant.
You
know,
Andrew
does
a.
D
It's
like
we
don't
always
have
the
ability
to
catch
up
and
so
kind
of
how
we
would
divide.
Work
is
like
hey,
open
up
a
PR
and
it'll
automatically
tag
us
and
then
we'll
take
a
look
at
the
pr,
or
you
know,
comment
on
an
issue
and
say
hi
I'm
interested
in
working
on
this
and
I've.
You
know
got
a
work
in
progress
PR
over
here.
That
way.
We
we
know
that
you're
actively
there
or,
if
there's
already
an
existing
open
PR.
D
That's
not
over
the
finish
line
and
you're
interested
in
helping
that
that's
also
great,
and
if
you
are
particularly
an
expert
in
either
in
the
library
like
not
all
of
us
are,
for
example,
like
I,
don't
use
the
AWS
SDK,
but
we
have
AWS
instrumentation.
E
Aws,
specifically,
actually
I,
don't
think
many
people
on
the
call
actually
use
it.
So
I
I
was
going
to
call
out
spec
compliance
too,
though
there's
a
bunch
of
those
across
the
different
repos
and
there's
actually
projects
too.
So,
if
you
feel
so
inclined
like
there's
a
bunch
of
small
stuff
that
has
to
be
done,
released
to
release
and
yeah
we're
a
little
bit
behind
right
now,
that's
definitely
appreciated.
E
I
was
going
to
ask,
though,
like
you
know,
you
said:
datadog
was
going
to
start
kind
of
contributing
some
resources
to
it.
Did
you
have
like
any
specific
areas
in
mind
right
away
like
maybe
we
could
speak
to
something
more
specific
if
you
kind
of
had
an
area
of
Interest
or
something
that
you
were
thinking
about
already.
A
Actually
I'm
I'm
trying
to
be
like
proactive
quote-unquote
like
I'm
starting,
like
I've,
joined
this
before
I
I,
actually
have
the
dedicated
time
to
it.
So
I
haven't
started
my
work
yet
so
just
to
just
to
for
context.
Well,
what
we
want
to
do
is
we
are
probably
going
to
implement
an
alternative,
SDK
implementation
based
on
the
API.
So
what
if
I
see
happening
is
that
I
might
find
like
edge
cases
on
the
spec
or
my
final
cases
where,
like
how
come
this?
A
Is
this
this
API
implements
this
way
because
it
doesn't
seem
like
it
works
well
with
that
thing
so
and
I
might
find
bugs
that
are
like.
Oh,
this
is
a
bug
in
part
of
the
open.
Telemetry
is
like
and
I
see
when
I'm
implementing
a
custom,
SDK
I
just
want
to
just
go
upstream
and
fix
the
the
API
bug,
for
example.
So,
right
now,
like
all
the
information
I
have
from
hotel,
is
like
I
write,
all
the
spec
and
all
the
documentation
I'm
still
like
new
to
it.
A
So
I'm,
not
yeah
I,
don't
have
any
Targets
right
now,
but
I'll
probably
be
touching
the
whole
API
and
if
there's
anything,
that's
like
constructive
that
I
can
fix
and
like
help
improve
or
how
propose
change
to
the
spec
or
be
doing
that.
I'll
try
to
join
this
every
week
too,
just
to
to
see
what's
happening
here.
I
have
one
question
about
the
the
spec
that
you
said:
there
are
things
in
Ruby.
They
are
not
up
to
the
main
spec
I
assume.
There's
no
way
to
know
what
parts
are
behind
right.
E
Well,
yeah
I
mean
it's
it's
an
it's
a
manual
process
every
time,
there's
a
spec
release.
Currently
the
process
is
Andrew,
gets
a
notification
and
then
goes
and
looks
at
the
spec
and
then
determines
what
we
have
missing,
but
at
least
for
I
think
the
latest
release,
or
maybe
one
version
behind
we
do
have
issues
for
everything
that
we're
deficient
on
I,
don't
think
I
finished
the
last.
The
most
recent
audit
is
not
completely
done.
E
I
left
some
areas
unaudited,
but
like
generally
speaking,
we
do
know,
but
that's
because
we
just
go
and
exhaustively
kind
of
like
read
the
spec
changes
and
then
look
into
our
code
base.
So
if
otel
had
like
some
sort
of
automated
spec
compliance
thing,
that
would
be
fantastic,
but
they
don't
so
we
just
kind
of
have
to
go
and
investigate
every
time.
It
hasn't
actually
been
that
bad,
yet
I've
just
been
personally
too
busy
to
like
finish
the
last
audit
but
yeah.
We
just
kind
of
manually
pour
over
the
changes.
D
A
E
Oh
yeah,
so
there
is
a
project
board
on
each
of
the
Ruby
and
Ruby
contrib
repos
I
open
a
sort
of
like
top
level
spec
compliance
audit
issue
where
I,
like
you
know,
do
the
work
to
do
the
audit
and
then
I
open
a
separate
issue
for
each
one
of
the
deficient
items.
We
tag
it
with
spec
compliance
and
then
also
spec
compliance,
slash
V
whatever
and
all
of
those
things
go
on
the
board
and
then
we
track
them
in,
like
you
know,
to
do
in
progress
done.
E
E
You
yeah,
if
you
want
to
help
out
with
that
at
all,
like
that's
fantastic,
it's
it's
not
like
the
most
glamorous
work,
so
it
sometimes
Falls
by
the
wayside.
I
think
like
it's,
not
new,
instrumentation
or
new
thing,
it's
sort
of
just
like
now.
You
need
to
go
in,
like
percent
decode,
this
value
a
different
way
or
something
because
of
this
you
know
so
it
can
be
tedious,
but
if
you
have
any
capacity
or
interest,
I
would
certainly
appreciate
it
like.
It
would
be
very,
very
helpful,
I
think.
B
Another
other
things
worth
mentioning,
because
I
know
you
for
some
random
reason.
I
know
you're
pretty
good
at
this
is
our
like:
Ci
is
kind
of
a
mess
yeah
I.
B
Had
any
experience
with
that,
but
yeah,
if
you're
ever
like
looking
for
something
some
area
to
like
get
involved
in
cleaning
up
our
CI
and
making
it
run
faster,
would
be
like
a
huge,
a
huge
win
and
making
it
less.
We
have
some
flakiness
because
we're
running
it
for
like
Windows
and
we're
running
it
for
truffle,
Ruby
and-
and
we
don't
quite
have
it
as
split
out
where,
like
so
right
now,
I
think
like
when
the
Truffle
Ruby
CI
run,
Fails
Like
It
blocks
releases,
and
then
it's
like
that
or.
E
E
It
was
only
Linux
that
actually
just
runs
nightly
and
I
get
a
notification
of
the
Tails,
basically
yeah,
specifically
because
of
the
nonsense
but
yeah
anything
CI
related
like
we
run
so
many
builds
that
it's
just
really
unwieldy,
so
definitely
would
appreciate
help.
There
also
related,
like
a
gem,
releasing
actually
like.
We
have
a
lot
of
gems
to
release
and
they
are
sometimes
have
weird
version
dependencies
and
transitive
dependencies,
and
it
can
be
kind
of
overwhelming
a
little
bit.
E
B
It's
also
worth
mentioning
a
little
just
for
like
a
good
mental
model.
This
is
actually
it's
not
a
trace.
Specific
repo,
like
we
have
the
SDK
or
basically
like
open
Telemetry
Ruby,
is
supposed
to
cover
traces,
metrics
demox
right
now.
We
only
have
traces
implemented
as
like
a
signal,
but
there's
work
in
progress
on
a
metric,
swarm
and
logs
is
there's
like
a
PR
App.
That
was
pretty
rough
a
long
time
ago,
but
you
might
get
some
questions
around
that.
B
So
just
like
for
your
mental
model,
we
just
because
Ruby
doesn't
have
you
know
a
metrics
directory
in
a
logs
directory.
They
they
are
supposed
to
be
there.
We're
just
haven't
gotten
time
to
finish
our
implementations
yet,
but
metrics
is
pretty
far
along
logs.
If
there's
some
brave
soul
who
wants
to
go
jump
in
there
feel.
D
Free
but.
B
Yeah
I,
don't
know
other
than
that.
Yeah
I
don't
know,
welcome
you
you're
welcome
to
do
as
much
or
as
low
as
you'd
like,
and
you
know
we're
always
around.
If
you
have
questions
the
cncf
Slack's,
like
pretty
active,
if
you
ever
like,
have
random
questions
that
you
might
and
myself
remember
also
but
try
to
use
the
answer.
Slack
and
yeah.
A
Yeah
yeah,
no
thanks
so
much
it's
it's
nice
to
be
like
to
see
the
process
and
then
what's
happening
and
it's
nice
to
see
that,
like
there's,
just
regular
people
running
this
thing,
you
know
it's
just
people
like
us,
you
know
so
it's
all
good
yeah,
but
looking
forward
looking
forward
to
doing
the
work
going
forward
and
just
being
more
involved,
yeah.
E
No,
we're
definitely
happy
to
have
you
and
we're
definitely
just
regular
people.
Someone
once
said
we
were
the
nicest
Sig
and
and
after
that,
I've
been
like.
Oh
gosh,
I'm
gonna,
wear
that
with
a
badge
of
honor
like
we
gotta,
we
I
want
to
be
super
welcoming
in
like
anything,
you
want
to
contribute
more
than
happy
to
take
it
like
it's
there's,
no
way
that
you
know
we
can
do
this
on
our
own.
We
definitely
need
a
lot
of
outside
help,
so.
B
D
Marco
I
know
you
probably
don't
need
to
divulge,
or
you
can't
double
do
any
of
this
stuff.
So
it
sounds
like
you're
all
gonna
have
like
a
vendor
specific
implementation
of
the
SDK.
At
this
point,
you
all
have
a
sense
of
like
what
you're
gonna
do
like.
Are
you
gonna
under
your
own
implementation
altogether
or
will
be
based
on
the
existing
SDK,
because
we
have
a
bit
of
a
tight
coupling
between
our
instrumentations
and
and
the
the
main
SDK
right
or
like
our
implementation
of
the
SDK?
D
And
that's
something
also
that
should
surface
and
I
was
curious,
like
if
y'all
had
any
constraints
or
like
some
things?
That
would
make
it
difficult
for
y'all
to
use,
though,
and
contribute
to
the
SDK
that.
A
We
have
so
that's
that's
good
to
know.
That's
I
feel
like
that's.
The
kind
of
stuff
I
would
learn
and
like
midnight,
I'm
like
who
did
this?
Who
coupled
this
like
this
and
yeah?
So
all
we're
trying
to
do
is
we
want.
We
want
people
to
be
able
to
move
freely
between
data
dogs
like
jam,
the
open,
Telemetry
champ
and,
if
you're
in
a
competitor,
and
they
should
be
able
to
just
pretty
much
remove
rendered
lock-in.
A
If,
if
people
are
happy
with
that,
so
we're
just
trying
to
empower
so
that
people,
if
they
are
just
using
just
vanilla,
open,
Telemetry
they're
able
to
send
their
data
to
datadog
if
they
want
to
use
some
of
the
deadlock
specific
apis,
they
can
do
those
at
the
same
time
as
the
open,
Telemetry,
API
and
then
they're
able
to
be
able
to
use
all
data
all
open,
Telemetry,
I
mix
the
ball
I
mix
up
and
then
I'll
see
what
happens.
A
So,
like
you
said
like
it
seems
to
me
like,
because
we
want
to
mix
both
open
Telemetry
spans
and
their
dog
spans
in
the
same
Trace,
I
kind
of
have
to
do
like
a
bunch
of
heavy
lifting
to
to
do
that.
So
it
seems
to
me
like
implementing
keeping
the
API
exactly
as
is
and
implementing
it
under
that
makes
sense,
but
as
I'm
implementing
it,
it's
not
something
quick
but
like
I'll
see
how
long
like
where
the
where
the
intersection
points
are
like
where
things
are
like.
A
You
said:
if
there's
a
couple
in
between
SDK
and
API,
I'll,
probably
like
notice
that
and
I
can
raise
it
if
I
feel
like
it's
like.
This
is
a
technical
issue
at
the
open,
Telemetry
level,
but
right
now
yeah
that
what
I'm
trying
I
haven't
started.
So
it's
pretty
much
just
remove
the
vendor
lock-in.
That's
that's!
What
we're
trying
to
do!
Gotcha.
D
Gotcha,
so
if
you
were
doing
like
a
like
a
peace,
film
migration,
some
things
to
be
aware
of
is
not
like.
A
lot
of
the
code
is
based
off
of
Duty
Trace
RB,
but
it
does
not
implement
it
the
same
way
and
so
like,
if
you're,
trying
to
run
both
sdks
in
parallel,
and
you
try
to
turn
on
instrumentation
in
one
and
the
other
one
that
you're
going
to
run
into
some
problems
where
modules
are
getting
mixed
in
over
each
other.
D
So
that's
going
to
be
interesting,
I'm
interested
in
hearing
the
experience
report
from
that.
E
Yeah
I
actually
had
a
question.
I
I'm
curious
a
little
bit
more
about
the
the
vendor
lock-in
and
removing
that
because
I
I
love
that
goal.
You
know
obviously
working
in
an
open
source
project
about
it.
Right
I'm
wondering
what
like
what
kinds
of
things
make
it
really
difficult
to
use
the
SDK
as
it
is,
what
what
necessarily
like
what
types
of
Integrations
or
hooks
is.
Maybe
the
SDK
missing
that
would
need
to
like
necessitate
an
entirely
separate,
SDK
implementation,
because
it's
kind
of
a
big
surface
area.
E
You
know
I'm
wondering,
like
you,
said,
mixing
datadog
scans
and
open
planetary
spans
in
the
same
Trace.
Maybe
yeah
I
was
just
kind
of
curious
to
know
a
little
bit
more
about
that
like
what
does
that
like?
Look
like
yeah.
A
Yeah,
so
I
don't
think
we're
right
now
doing
an
IFC
draft
internal
data
dog
we
haven't
finished
yet
for
the
whole,
open
Telemetry
work
that
I'm
talking
about
right
now
across
all
languages.
So
we
actually
talked
about
using
the
is
it
called
the
spin
processor
I'm,
not
100,
sure,
there's
a
sense
of
answering
the
spin
starts
and
spin
in
that's
it.
Okay,
there's
not
a
lot,
so
I
think
some
languages
are
like.
Can
we
just
use
this
pin
processor?
A
Because
we
want
to
support
the
open
Telemetry
calls
that
clients
have
when
they
Port
their
application,
which
might
be
the
hooks
some
of
the
hooks
you're
talking
about?
It
seems
like
that
would
not
work
because
we
have
custom
Samplers
that
need
to
be
supported
and
they
kind
of
like
tied
closely
to
our
billing
cycle.
A
So,
like
we
kind
of
need
the
clients
to
support
other
data,
dog,
samplers
and
I
agree
with
you
that
the
SDK
is
a
big
surface
area,
but
much
more
than
anything
like
or
does
a
magnitude
bigger
than
the
API,
but
I
think
because
we
need
to
support
a
kind
of
bi-directional
like
the
parent
is
a
dead
dog
spin
and
the
child
is
an
open
television
standard
and
the
child
will
tell
them
understands
the
data
Dog
span,
just
because
we
want
to
make
it
completely
transparate
the
reason
I'm
saying
we
would
probably
limit
an
SDK.
A
It's
because
that's
just
our
current
assessment,
no
one
has
tackled
the
Prototype.
Yet
that
might
might
change
like
you
said.
Maybe
the
the
integration
area
is
simpler
than
implementing
the
whole
SDK,
but
that's
our
current
assessment,
we'll
start
and
then
we'll
see
where
we
where
it
goes
from
there,
but
right
now
it
is
it's
just
our
preliminary
assessment
so
we're
it's
not
set
in
stone,
exactly
everything.
A
So
what
you're,
trying
to
tell
me
like
is,
is
kind
of
quite
your
question
like
it
should
be
easier
for
a
vendor
to
to
support
the
API
than
to
have
to
implement
the
whole
SDK
like
is
that
the
goal
is
that
what
you're
trying
to
say.
E
E
The
datadog
Ruby
Tracer
as
is,
and
ideally
like
I,
would
love
us
to
sort
of
join
forces
in
a
way
and
work
on
shared
code
bases
to
a
greater
extent,
and
so,
if
there
are
like
really
odd
things
that
datadog,
Samplers
or
span
processors
or
propagator
systems
might
need
to
support,
like
I,
think
we'd
be
open
to
talking
about
like
ways.
We
might
be
able
to
support
that
without
you
having
to
boil
the
ocean
to
get
there
like
I
I,
just
from
a
duplication
of
effort.
E
Standpoint
like
there's
not
a
lot
of
people
in
the
world
working
on
tracing
implementations
for
Ruby,
like
there's,
probably
less
than
a
hundred
of
us
total
like
it's
just
a
very
small
world,
so
I'm,
basically
just
trying
to
figure
out
like
you
know
what
what
would
it
take
to
like
support
the
non-standardized
data
dog
things
in
a
way
that
makes
sense
and
like
shares
more
of
the
workload
across
everything
so
like
if
they're,
if
there
are
things
that
you
think
might
be
interesting
but,
like
don't
seem,
obviously
compatible
like
I'm
sure,
some
of
us
would
be
happy
to
talk
to
you
and
like
figure
out
or
like
work
on
a
draft
like
with
of
a
doc
or
something
like
I.
E
Don't
know
how
to
try
to
say
what
I'm
trying
to
say,
but
like
I
know,
you
can't
share
a
lot
of
plans
externally,
but
like
I'm
sure
one
of
us
would
be
happy
to
be
like
yeah,
we'll
sign
the
NBA.
So
we
can
talk
to
you
like
about
what
your
plan
is
and
then
maybe
find
like
a
less
extensive.
You
know
less
less
difficult
way
forward
or
something.
So
it's
just
something
to
keep
in
mind
and
take
back
to
people.
B
I
I
just
want
to
quickly
not
that
I'm
about
to,
like
you
know,
dump
everything.
I
did
one
of
the
data
dog,
but
do
a
quick.
There
might
be
a
difference
in
mental
models.
Sdk
in
maybe
like
the
traditional
sense
versus
an
open,
Telemetry,
sdk's,
actually
kind
of
narrow
and
open
Telemetry.
B
It's
separate
from
the
idea
of
like
so
span,
processors
and
Samplers,
and
propagators
and
exporters
are
actually
all
sort
of
like
different
Primitives
that
are
used
by
an
SDK,
but
like
are
extract
it
and
you
know
you
can
submit
you
can
you
can
have
separate
you
can
use
like
the
official,
open
television,
Ruby
SDK
with
custom,
Samplers
and
propagators
and
and
those
can
be
separate
gems,
so
I
think
what
you'll
find
is
that,
like
you
end
up
having
to
do
so
many
of
those
separate
ones
that
it
you
be
it
it
starts
to
look
like
boiling
the
ocean.
B
A
B
Yeah
I
think
you'll
find
your
your
provider.
You
have
to
implement
a
a
sampler
to
respect
the
data,
log,
specific
sampling,
headers
propagators,
to
respect
the
data,
log,
specific
propagation
headers
and
then
there's
actually
some
different.
Some.
If
there's
some
issues
in
this
concept
of
recording
versus
non-recording
spam
makes
it
a
little
funky
to
yeah
work
very
cleanly
with
some
datadog
sampling
choices,
because
non-recording
spans,
which
are
essentially
like
spans,
open
television,
decided
they
don't
want
to
sample.
They
don't
collect
attributes.
B
So
you
have
to
sort
of
modify
what
it
means
to
not
record
there.
There's
a
surprisingly
large
amount
of
like
Cuts
like
you'll,
find
there's
like
a
death
by
a
Thousand
Cuts,
but
you
might
not
need
to
for
what
it's
worth
like.
You
might
just
kind
of
need
a
wrapper
package
with
like
some
slight
modifications,
rather
than
like
rolling
the
whole
SDK
from
scratch.
So
anyway
yeah.
You
know
you
know
I'm
always
around.
If
you
know,
but.
A
I'm
glad
you're
good,
it's
good
yeah,
I
misspoke,
because
I
think
we
weren't
only
the
place
provider
and
that's
a
good
point.
I
think
I
think
this
is
already
fleshed
out
because,
like
I
said
all
languages,
there's
people
different
people,
doing
research,
you're.
A
Sense,
you
make
sense
to
what
you
said
that,
like
a
terrorist
provider,
but
it's
also
the
simpler,
but
it
seems
like
there
are
hooks
for
most
of
these.
We
understand
that
the
sampling
is
different
and
or
especially
like
I
said,
there's
the
recording
it
can
be
recording
or
not
recording.
We
know
that.
That's
the
difference
and
we
accept
that.
A
That's
incompatibility
for
now
and
that's
okay,
we'll
talk
more
in
this
coming
weeks,
we'll
see
the
stuff
but
yeah
I'm,
not
I'm,
not
trying
to
like
we're
not
we're
doing
everything
just
because
it's
fun
and
we
really
want
to
just
like
our
own
version
of
everything
right
and
to
be
honest,
like
I'm,
not
I'm,
no
expert
on
our
integration
like
the
on
otel
or
how
we're
doing
hotel
at
this
point.
But
we'll
keep
the
conversation
going.
The
next
few
weeks
or
a
few
months
and.
E
Oh
totally,
we
I
don't
think
we
assumed
you
just
wanted
to
do
it
for
fun.
I
mean
that
might
be
some
people's
definition
of
fun
and.
E
D
I
also
shared
a
little
bit
of
a
hint
of
an
older
issue
that
somebody
from
datadog
opened
up,
which
is
like
some
constraints
in
the
some
constraints,
in
how
the
SDK
is
written,
also
about
thread
using
threads
safe
variables,
try
local
variables,
and
that
is
a
implementation
constraint.
D
So
this,
like
issue
was
opened
up
to
say,
like
hey,
make
it
possible
for
the
contacts
to
be
available
in
different
threads
available
in
separate
threads
specifically
to
work
with
a
datadog,
profiler
I.
Remember
this,
and
basically
this
might
be
something
that
might
fit
into
the
custom.
Sdk
thing
you
know,
portion
just
put
on
your
radar.
A
Yeah,
no,
it
makes
sense
this
my
colleague
there
to
open
the
ticket
yeah
it's
a
profile.
The
profiler
really
likes
to
put
their
hands
everywhere
when
it's
time
to
collect
the
sampling
and
it
yeah.
But,
like
you
said
it's
it's
a
very
it's
a
very
specific
thing:
it
might
fit
the
custom
custom
implementation
better,
but
yeah,
okay,.
B
Also
yeah
I
didn't
mean
to
bombard
you
with
questions
mark
That's
good
to
see
you.
Anyone
else
have
things
that
I'll
talk
about.
We
have
seven
minutes,
but
there
is
our
design
for
my
role
as
moderator
or
whatever
resigned
in
disgrace.
Okay,.
E
E
D
D
B
D
We
got
we
got
one
approval
on
the
on
the
PG
Safeguard
against
the
host
being
no
thing.
I
asked
some
questions
on
that
one,
so
that
one
was
just
pending
and
yeah.
Andrew
is
not
here
to
talk
about
this
one,
but
like
the
auto
instrumentation
is
like
a
pain
in
the
butt
to
release
the
rails,
one
because
it's
like
the
rails,
one
references,
other
sub
gems
and
then
all
references
them
also
so
I
was
just
like
you
know
we
gotta
figure
out
how
to
remove
the
toil
there.
D
It
really
is
just
silly
because
it's
like,
for
all
intents
and
purposes
rails,
is
the
whole
gem
anyway,
so
and
to
open
up
a
separate
issue
about
it.
If
you
have
any
thoughts
jump
in
on
that
one,
where
is
it.
D
Yeah
yeah
yeah
yeah,
there's
a
a
big
old
paint,
but
basically,
like
I,
was
saying
any
bumps
that
happen
to
the
rails,
the
gems
that
rails
references
you
need
to
also
be
bumped
in
rails,
so
I'm
like.
Perhaps
we
remove
all
those
references
in
all
move
them
all
on
the
rails,
and
then
you
know
you
bump
the
child
references
bump,
rails
and
bump
ball.
That
is
not
fun
that
whole
experience
is
very,
very
unfun,
so
it
might
be
a
better
need
to
like
really
I,
don't
know,
maybe
think
about
how
it
all
works.
Still.
D
Not
so
sure
yeah.
B
I
understand
I,
guess
if
there's
a
path
to
do
it,
the
existing
config
then
just
document
that
path
foreign,
not
to
as
you
may
be
aware,
I
don't
always
see
eye
to
eye
with
other
folks
here
on,
got
the
right
way
to
do
configures
so
yeah.
B
Okay,
so
do
you
need
to
approval
on
the
PG
thing
or
I.
D
Think
it's
conversation
on
the
PG
thing
because
what's
happening
right
now,
is
it's
extracted
as
a
as
a
private
method
in
the
in
the
instrumentation
which
I.
D
To
get
away
from
using
private
methods,
yeah
that's
one
thing,
but
the
other
thing
is
it's
testing
the
private
method
using
Ruby
send,
instead
of
trying
to
reproduce
reproduce
the
use
case,
so
I
was
basically
like
hey.
Is
there
a
way
for
us
to
reproduce
the
use
case?
Otherwise
we're
just
writing
code
here
right
that
I
have
no
idea
that
we
can't
like
have
a
regression
for
really
right
yeah
or
avoid
a.
B
B
We
should
try
to
encourage
real,
but
I
guess,
let's
see
what
they
come
back
with,
if
if
they
don't
want,
if
they
truly
don't
own
the
bandwidth
to
do
a
solid
regression
test,
maybe
we
can
do
a
merge
with
an
immediate
follow-up
patch
with
the
test
ourselves,
although
I
don't
know
how
to
write
it
because
I'm
not
familiar
with
the
bug,
sounds
like
it
was
just
it's
just
like
default
rail
settings,
you
can
kind
of
get
it,
but.
B
One
sorry
yeah
there's
just
a
one
of
the
spam
tags
or
spanner
tributes
that
we
collect
for
postgres
instrumentation.
It's
like
the
potentially
nil,
which
is
for
the
host
parameter
and
I
guess
in
our
case
and
then
nil,
kinda
things
shouldn't
be
set
to
no
values.
B
It
causes
Downstream
stuff
to
choke
on
it,
I
think
when
you
try
to
format
it
into
like
the
protoba
for
otlp
format
like
things
blow
up
so
yeah,
there's
some
I
guess
that
it's
now
just
saying:
don't
include
it
if
it's
new
but
yeah
we're
kind
of
like
well,
what's
the
case
where
this
happens
so
that,
like
we
don't
just
like
you
know,
have
this
happen
again.
You
know
in
three
weeks
when
we
should
have
changed
him.
B
He
hasn't
responded
with
what
the
actual
you
know
real
world
use
cases
here,
but
I've
that
is
that's
all
I
know,
and
it's
just
from
reading
the
email
and
not
opening
the
issue
in
GitHub
and
just
reading
like
the
just
the
email,
that's
above
the,
so
you
don't
have
to
open
it
anyway.
D
Yeah
outside
of
that,
that's
all
for
the
Ruby
control
package.
It
looks
like
we're
going
to
need
another
release
of
all,
because
we're
gonna
have
race
car
and
potentially
this
bump
of
action-pack,
again
yeah
and
so
Sam.
How
you
feeling
about
doing
a
release
for
race,
car
and
all.
C
D
Yeah
and
okay,
so
with
the
old
College,
try
can
I
get
a,
can
I
get
a
review
on
this.
C
D
Which
is
the
bump,
essentially
the
gem
bump,
to
make
sure
that
rails
is
referencing,
the
appropriate
sub
module
more
of
our
sub
package,
I.
D
B
Much
it's
what
I
do
very
little
all
right,
we're
over
time,
I'm
gonna
hop!
You
guys
are
welcome
to
stay.
D
I
hope
we
have
in
schedule
away
so
come
on
back
next
time.
Yeah.
C
D
D
Weeks
we
meet
weekly,
but
is
anybody
going
to
be
around
for
the
holiday
week
in
two
weeks
y'all
be
here,
okay!
Well,
then,
then,
at.
D
Okay,
Marco
I
need
to
go
on
international
travel.
I
need
a
bodyguard.
No,
so
I
will
see
you
all
in
like
in
like
a
month
or
whatever
enjoy.