►
From YouTube: 2021-10-12 meeting
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
B
D
D
F
Nice
to
nice
to
eat
you
yeah
thanks
for
thanks
for
joining
sorry,
I'm
late.
I
was
making
my
lunch,
which
is
a
bowl
of
leftovers.
It
tastes
good.
F
Thanks
that
was
also
distracted
by
francis
damming
me
and
other
professional
chat
with
a
thousand
picks.
It's
all.
F
B
F
A
Francis
said
he's
not
gonna
join
either
no
worries
so
is
robert
going
to
attend.
A
Okie
dokie
so
then,
who
will
play
the
role
of
de
facto
maintainer
today.
F
I
can
do
it,
I
can
I'm
happy
to.
As
you
all
know,
I
like
to
talk,
I'm
happy
to
talk,
but
yeah
I
mean
if
sorry,
one
second.
F
Let
me
just
share
my
screen:
real
quickly,
do
do
you
all
want
to
do,
we
can
do
intros
because
amy
is
is
new.
I
had
I
sometimes
I'd
spam,
her
her
tweets
and
and
like
her
tweets
and
then
one
day
I
said
why
don't
you
just
attend
it's
like
we,
we
can
do
an
intro
round
real,
quick.
Well,
I'm
if
everyone
I
can
I'll
start
americ,
I'm
an
approver.
I
work
at
shopify
I
joined
recently
there
and
and
yeah
that's
kind
of.
F
I
was
a
data
dog
before
that
and
that's
cool.
C
Next,
my
name
is
andrew
hayworth.
I've
met
most
of
you
here
before
used
to
work
at
github
with
ariel
and
then
earlier
than
that
amy.
Now,
I'm
at
a
startup
doing
nothing
ruby
related
at
all,
but
I
like
talking
to
all
of
you
people,
so
I
keep
coming
here.
C
A
Hello,
I'm
mariel.
I
work
at
github,
a
former
teammate
of
andrew
on
the
observability
team.
I'm
also
lucky
enough
to
be
a
maintainer,
so
that
means
I
get
to
review
prs.
Look
at
me
and
I'm
been
pushing
the
open,
telemetry
effort
at
github,
so
lucky
them,
and
I
will
now
popcorn
over
to
tim.
H
I've
been
at
shopify
for
about
two
years
and
then
at
some
point
I
transitioned
teams
to
production
engineering,
doing
like
distributor
tracing
stuff
with
robert
and
francis.
So
I've
been
trying
to
push
some
contribution
to
open,
telemetry
and
been
attending
those
six.
B
Hey
yeah,
I'm
actually
the
maintainer
of
the
earling
and
elixir
telemetry
implementation,
and
I
recently
joined
splunk
and
joining
this
sig
in
order
to
splunk,
wants
to
get
their
ruby
distro
out
and
I'm
gonna
task
with
helping
get
that
done.
D
I
guess
I'm
last
hi,
I'm
amy
toby,
I'm
a
principal
engineer
at
equinix
or
equinix
metal,
formerly
known
as
packet,
I'm
building
an
sre
team
there
and
we're
doing
the
whole
observability
and
incidents
and
slos
approach
to
sre,
and
so
our
anytime,
you
do
that.
The
first
thing
you
got
to
do
is
roll
out
telemetry
across
the
services
you
know
and
we're
really
focused
on
tracing
and
we
have
a
large
rail
stack
and
a
couple
other
smaller
rails,
things
that
we're
instrumenting
with
with
the
os
code.
F
Cool
awesome
well
yeah,
nice,
nice
to
meet
you
all
so
the
way
we
usually
by
we
I
mean
matt
over
at
lightsaber.
Francis
usually
runs
these
meetings
is
they
they
run
down
the
the
spec.
This
specification
sig
meeting,
which
occurs
just
before
this.
We
try
to
see
if
there's
anything,
people
have
strong
opinions
on
which
it
turns
out.
People
usually
do
and
we're
just
gonna
talk
about
it,
and
then
we
kind
of
jump
into
ruby's,
specific
stuff,
but
preemptively.
F
If
people
have
things
they
want
to
bring
up
first,
that
are,
I
don't
know
specific
if
they
have
burning
questions
or
things
that
are,
they
want
to
make
sure
we
have
time
for
it.
We
can
discuss
those
first,
I
don't
have
anything
super
important.
I
can
reviews
talk
about
some
pr
reviews
with
anyone.
B
Yeah
just
that
the
so
we
were
talking
before
about
a
blog
post
for
ruby's
ga
and
the
possibility
of
combining
that
with
the
ga
for
erling
and
javascript.
E
F
Cool,
let
me
share
my
screen
also,
I
so
I
think
robert's
the
right.
I
know
and
tim
correct
me
if
I'm
oh,
no,
it
doesn't
want
me
to
sorry
tim
correct
me.
If
I'm
wrong,
I
know
robert
wanted
to
oh.
It
says
all
right.
I
have
to
quit
and
rejoin
because
zoom
permissions
one
second.
A
G
F
True,
there's
like
a
feng
shui
thing,
that's
probably
specific.
It's
like
where
you
place
your
bed
in
your
room.
First,
okay,
can
we
all
see
my
screen
cool
so
for
any
questions,
so
yeah
blog
post?
I
know
robert
was
working
on
something
internally
for
like
the
shopify
vlogs.
I
think
he
also
wanted
to
say
something
positive,
so
I
will
just
say
like
maybe
you
can.
I
can
chat
with
him
tomorrow
and
make
sure
he
pings
you,
but
I
don't
see
why
we
wouldn't
want
to
be.
F
It
feels
like
a
a
positive
thing
that
we
want
to
celebrate
so
and
yeah.
I
don't
know
what
the
timeline
spelling
not.
E
My
thing:
it's
a
in.
F
F
F
Okay,
cool
yeah,
but
I
think
that's
super
reasonable
and
hopefully,
if
he
doesn't
follow
up,
feel
free
to
ping
me
or
or
just
drop
a
note
in
the
cncf
slack
for
really
and
yeah.
I
think
it's
just
thanks.
Canadian
thanksgiving
stuff
people
are
a
little
bit
pre-occupied.
F
Yeah
keep
going
some
awesome
talks
by
the
way
people
are
interested
cool,
all
right,
anything
else
on
top
of
mine
for
others,
otherwise
I'll
kind
of
work
through
the
the
spec
sig.
I
see
this.
A
F
F
Yeah,
so
I
know
we're
interested
in
that
cool.
Oh,
maybe
a
link
to
the
issue.
Well
I'll.
Just
look
at
this
nice.
Individual
person
see
my
internal
stuff.
Whatever
excuse
me
so
so
yeah
otep152
was
merged.
The
specification
first
tristan
have
you
implemented,
schema
url
or
in
in
erlang.
F
B
There's
a
pull
request
for
it,
so
I
haven't
merged
it.
So
no.
F
Worries
so
it
does
a
few
things.
It's
it's.
Basically,
the
idea
is
you're
basically
saying
like
here's,
the
here's,
the
semantic
conventions
I'm
using
in
this
application
or
here's
the
semantic
conventions.
This
instrumentation
library
is
using.
So
there's
changes
to
the
proto,
so
we
would
need
to
so
there's
a
few
things.
One
there's
the
newest
proto
has
schema
url
as
a
field
in
instrumentation
library,
spans
and
somewhere
in
resource,
but
there's
also
changes
to
sort
of
like
the
resource
api
and
then
like
the
tracer
api,
and
essentially
it's
like
an
optional.
F
I
think
it's,
I
believe
it's
optional,
but
the
idea
is,
you
say
like
okay,
so,
let's
say
like
rails
ads,
schema
url
rails
adds
instrument
like
native
hotel
instrumentation,
and
you
know
they
want
to.
I
believe
and
correct
me
if
this
is
the
incorrect
definition.
This
is
kind
of
like
my
interpretation,
and
they
want
to
add,
like
you
know,
so
they
want
to
automatically
append
like
http
status
code
as
a
tar
as
an
attribute
they
would
use
like
the
semantic
convention
for
like
whatever
http
status
code
is
like
http
dot.
F
I
don't
even
know
what
it
is
like
dot
status
quo
and
then,
let's
say,
like
you
know
a
year
from
now,
someone
comes
along.
It's
like
oh,
like
this.
Let's
update
this
this
tag
name.
This
is
sorry.
This
attribute
name
to
be
some
other
thing.
Because
of
reasons
and-
and
you
know,
I'm
sure,
so,
that's
like
an
inevitable
inevitable
thing
that
will
happen
is
like
standards.
F
You
know
which,
and
so
they
would
update
it,
but
so
now,
like
rails,
would
have
to
either
be
rails,
would
be
in
this
weird
place
where
they'd
be
like
emitting.
They'd
have
to
either
be
like
constantly
checking
what
the
semantic
conventions
are
and
being
like.
Oh
so
they
have
this
like
thing
they're
having
to
constantly
update,
which
is
like
a
burden
on
on
library
authors.
Instead,
what
they
can
do
is
say:
okay,
I'm
just
saying
at
the
time
this
was
written.
This
is
the
schema
url.
F
What
it's
doing
is
these
schema
urls
point
to
essentially
yaml
files
that
handle
you
know
all
these
translations,
so
the
yaml
file
says
you
know
we
can
take
an
example
of
it,
so
the
yaml
file
would
look
something
like
this
and
be
like
hey
in.
In
this
version
we
rename
you
know
we
rename
these
tags
and
whatever,
and
so
it
would
sort
of
abstract
away
that
burden,
so
that
that
should
be
like
one
reason
why
we
implement
it.
F
The
other
reason
is
like
you,
may
you
know
if
you
want
to
protect
your
your
database
against,
like
you
know,
you
want
to
ensure,
like
you
know,
conventions
or
you
want
to
ensure
your
ui.
If
you
have
some
ui,
you
want
to
essentially
have
it
point
to
like
a
schema,
url
there's
some
other
things
too.
F
One
is
like
it
doesn't
have
to
be
so
these
point
to
like
open
telemetry
published
as
a
schema,
url
and
so
there's
the
instrumentation
library,
one
sorry,
I'm
kind
of
rambling
and
then
there's
the
resource
scheme,
url
resources,
what's
specific
for
the
application,
instrumentation
library
specific
for
that
specific
instrumentation.
F
So
the
thing
is
these:
don't
have
to
point
to
an
open
telemetry
published,
schema
url.
It
could
point
to
say
like
a
this
is
my
understanding.
It
could
point
to
a
github,
specific
set
of
conventions
or
translations
or
a
shopify
one
or
whatever,
and
so
like.
Maybe
you
have
some
as
a
user,
you.
E
F
You
want
to
track
your
own
thing
so
like
this
is
sort
of
like
laying
the
groundwork
for
that
there's
still
a
lot
up
in
the
air,
in
the
sense
that,
like
there's,
no
real
parrot
parentage
like
in
inheritance
relationships
with
these
schema
urls.
So
what
happens
when
there's
a
conflict?
Let's
say
your
application
is
resource,
has
like
a
schema
url
from
a
resource.
Detector
of
you
know
like
an
otlp,
schema,
url
and
then,
but
then
you're
using
some
like
internal
one.
Essentially
the
the
way
the
spec
is
written.
F
Is
it
just
says
like
we're,
gonna
drop
both
of
them
and
just
say:
there's
no
cohesive,
schema,
url,
there's
not
like
any
sort
of
sense
of
like
oh.
This
is
built
on
top
of
it.
So
like
it's
still,
I
think,
a
little
bit
of
a
vague
feature,
but
it
sets
the
groundwork
for
having,
for,
I
guess,
ensuring
you
know
like
enforcing
semantic
conventions.
So
anyway,
I
did.
F
The
pr
which
is
all
it
is
at
the
end
of
the
day
is
just
like
an
optional
string
on
a
couple
areas:
okay,
but
and
then
some
merging
behavior
on
the
resource.
Definitely
if
anyone
has
read
a
close
reading
of
the
otep
or
looked
at,
I
think
go
has
an
implementation
and
python
has
an
implementation
which
I
basically
just
stole.
You
know
like
please,
feedback
is
welcome
because
I'm
kind
of
like
I
kind
of
learned
about
it
by
implementing
but
yeah.
I
think
we're
interested
in
this
feature.
F
So
I'm
happy
to
do
the
work
here
and
does
that
help.
A
Yes,
it
helps
a
lot
because
the
one
question
that
so
snide
comment
first
xslt
in
yaml
on
to
more
helpful
comments.
The
thing
I
appreciate
is
that
we
have
the
ability
to
define
our
own
custom
schemas.
The
one
thing
that
I'm
curious
about
is
that,
is
it
it's
about?
It's
not
about
schema
enforcement
as
much
as
it
is
about
schema
mapping,
because.
E
A
F
Yeah,
there's
nothing
really.
All
it's
doing
is
saying
like
all
in
practice.
All
it
has
right
now,
like
the
yaml
file.
All
it
allows
you
to
do
is
rename
is
math,
it
doesn't
say
like
hey
these.
Are
it's
not
really
as
powerful
as
you
would
kind
of
want
saying
like
hey,
let's
these,
these
specific
values,
let's
obfuscate,
these
specific
values,
let's
drop
these,
let's
right.
F
F
B
F
The
spec
says
you
just
dropped
both
of
them.
If,
when
you,
because
when
you're
merging
resources,
you
don't
I'm
sorry.
F
You
still,
you
still
merge
the
resource
attributes
according
to
the
spec,
which
is
you
know,
pretty
traditional
merge
behavior,
but
you
know,
I
think
it's
just
like
a
naive
overwrite,
whatever
gets
merged
overwrites
anything
that
doesn't
like
if
there's
a
conflict
but
schema
uro
gets
trapped.
Essentially,
all
it's
saying
is
we
can't
guarantee
that
this
schema
that
these
this
new
you
know
like
collect.
You
know
weird
sort
of
like
grouping
of
attributes
matches
any
particular
schema.
I
think
you
I
think
it
can.
It
can
be
useful
in
a
few
ways.
F
You
could
use
it
to
adapt.
How
like
enforce.
Maybe
what
attributes
you
display
in
a
dashboard
to
say
like
okay,
previously
we
were
displaying
like
a
toggle
for
like
toggle
by
like
span.status
and
now
that's
maybe
not
a
field
anymore,
and
we
want
to
toggle
by,
like
spam,
dot
status,
underscore
code
or
something
like
that.
I
could
think.
D
Of
all
kinds
of
uses
in
my
environment,
you
know
if
some
teams
are
using
camel
case,
some
want
to
do
the
dotted
stuff,
even
even
at
that
level
the
ability
to
be
able
to
do
schema
mapping
would
be
pretty
handy
at
times.
F
Yeah,
I
think
it's
important,
especially
if
you're,
using
like
a
you
know,
especially
if
your
backend
sort
of
just
allows
you
to
doesn't
have
is
kind
of
loose.
You
sort
of
need
to
enforce
it
somewhere
up.
You
know
it's
like
practically
being
enforced,
probably
through,
like
a
bunch
of
glue
code,
so
anyway
yeah.
So
this
is
like
in
practice.
F
I
think
we'll
implement
this,
and
then
the
real
value
will
be
when
the
pr,
when
the
collector
adds
a
processor
which
should
be
pretty
trivial
once
the
go
they're
doing
some
work
in
the
go
implementation,
so
they
can
just
reuse
some
of
those
packages.
It's
still
there's
still
other
questions.
B
Yeah
I
mean
it
says
the
resource
is
undefined,
not
just
the
scheme.
Yeah
yeah,
that's
what
that's!
What
kind
of
threw
me
and
yeah
we're
going
with
just
a
merge
too
so.
D
D
F
F
D
E
F
I
think
it's
super
valuable
and
this
is
sort
of
like
the
groundwork
for
that
stuff
for
ruby.
Specifically,
a
lot
of
this
is
here,
but
in
our
actual
in
this
implementation,
because
we're
stuck
on
an
older
proto,
it's
sort
of
no
ops
on
the
exporter
right
now,
so
we
I
need
to
speak
with
francis.
I
know
he
has
a
plan
to
upgrade
to
our
the
latest
proto.
I
think
there
may
be
one
or
two
other
small
blockers,
but
this
was
the
major
one
for
upgrading.
So
that's
important.
It's
also.
We
need
that.
F
We
need
the
latest
proto
to
start
to
do
metrics,
which
is
as
yet
unimplemented
so,
okay,
cool
I'll
just
keep
moving.
One
of
other
things
is
that
did
that
help
sorry
that.
F
Thank
you
cool.
If
no
one
has
other
specific
things
we,
I
can
quickly
burn
through
the
specs,
so
you
can
sort
of
see
if
any
people
have
comments
there.
G
A
If
you
get
a
chance,
would
you
mind
just
taking
a
look
again
at
the
issue
and
then
just
chiming
in
about
what
you
know
your
thoughts
are
about
it.
You
know
what
are
the
things
I
had
considered
was
saying:
okay,
it
is
issue
for
9
47.
A
A
A
A
F
Yeah
absolutely
yeah
and
to
summarize
sometimes
with
compatibility
stuff
we're
basically
using
like
an
ariel.
You
can
probably
explain
this
yeah.
A
Sure,
if
you
look
at
the
issue,
947
you'll
see
the
details.
Essentially,
the
monolithid
github
does
not
use
bundler
or
ruby
gems
to
load
jumps
for
a
window
of
many
different,
complicated
reasons,
and
so
one
of
the
things
that
the
data
dog
library
would
do
to
resolve.
That
kind
of
issue
was
that
would
have
a
fallback
to
say.
A
I'm
gonna
try
to
see
if
I'm
compatible
with
some
version
of
some
gem
and
sometimes
actually
don't
follow
a
pattern
that
would
get
you
easily
to
the
gem
version
right,
like
the
google
protobuf
gem,
for
example,
has
it
has
it's
a
gem
version
in
a
non-standard
place,
so
datadock
had
this
utility
method
that
would
fall
back
to
a
specific
version
through
some
hardcoded
constant
value,
except
that
feels
a
little
weird
in
the
way.
A
That
kind
of
like
our
compatibility,
like
we
have
this
dsl
on
the
instrumentation
libraries
that
tell
us
hey
look.
This
is
how
you
can
check
if
something
is
compatible.
A
You
want
to
check
if
something
is
installed
and
so
on
and
so
forth,
and
so
I
had
proposed
an
idea
which
was
to
sort
of
say
I
wonder
if
we
can
support,
say
as
part
of
the
dsl,
the
ruby
gem
syntax
for
so
that
we
can
define
if
we
want
to
do
something
like
optimistic,
locking,
pessimistic,
sorry
version
control.
A
Sorry,
what
am
I
saying
pessimistic
semantic
versioning,
as
well
as
having
a
default
class
to
fall
back
on
that
we
would
use
and
do
it
in
a
declarative
way,
and
I
don't
know
what
people's
impressions
of
an
a
of
configuration
like
this
would
look
like
and
then
it
it's
it
starts
to
play
into
this
thing
about,
like
is
the
dsl
now
going
to
get
too
complicated
for
people
when
working
with
these
instrumentation
libraries?
A
I
think
because
we
also
have
this
thing,
where
we're
missing
guidance
on
how
to
write
an
instrumentation
and
what
the
expectations
are.
So
I
don't
know
I
I
want
to
fall.
I
want
to
like
not
get
stuck
in
this
declarative
whole
of,
like
a
thousand
options
that
you
can
set,
which
kind
of
worries
me
as
well.
A
So
what
we
want
to
do
a
specific
instrumentation
is
built
around
monkey
patches,
essentially
in
most
of
the
cases
yeah.
So
if
there's
a
change
to
the
api
of
a
particular
instrumentation.
E
D
H
E
D
A
I
don't
know
because
you
figure,
if
you
use
the
instrumentation
all
gem,
for
example,
that's
going
to
have
a
mix
of
a
bunch
of
different
libraries
and
then,
if
you're,
using
save
rails
5
what
you
want
to
use
faraday
10,
then
it's
kind
of
like
oh
hold,
on
whatever
these
two.
E
A
D
A
F
Yeah
a
few
things
one
is
like
the
original
issue
I
think,
is
straightforward
in
the
sense
of
like
and
sort
of
like
so
the
original
issue
sort
of
like
uncovered
this
broader
thing.
But
the
original
issue
is
like
you
know:
we
just
sort
of
need
to
do
the
work
and
go
through
these
gem,
these
instrumentations
and
find
like
which
ones
we
can
just
check
double
check.
F
The
you
know,
like
my
gem
version
instead
of
using
you
know,
instead
of
using
like
the
rubygems
helper
and-
and
you
know,
we'll
also
need
to
do
that
for
stuff
like
protobuf,
if
it
gets
added
and
a
few
other
things
and
like
that's,
not
impossible,
work
just
like
some
busy
work,
we'll
have
to
go
through
and,
like
sort
of
you
know,
make
sure
we
can
like
apply
some
reasonable
fallbacks
to
address
the
the
original
issue,
which
is
like
our
compatibility
checking,
was
just
like
a
little
lazy
and
relied
on
the
like
ruby,
gems
thing
and
that's
like
more
or
less
a
super
advanced
use
case,
but
it's
still
important
and
then
once
we
add
it
to
all
the
stuff,
we
can
also
then
enforce
it.
F
On
new
instrumentations
and
it'll
make
it
easy
because
we
can
just
say
hey,
like
just
add
this
quick
fallback
and
then
that
way
like
if
someone
comes
around
and
it's
like
hey,
I
want
to
add
the
race
car
jam
instrumentation
and
we
don't
know
anything
about
race
car.
We
won't
have
to
do
the
work
to
like
find
where
that
version
is
located.
They
can
you
know
the
contributor
can
just
add
it
so,
like
I
think,
that's
totally
reasonable
and
fine.
F
It's
it's
just
been
low
priority
for
me
to
do
outside
of,
like
you
know
this
stuff,
that's
breaking
current
users
stuff
but,
like
you
know,
so,
there's
that
and
then
there's.
The
second
thing,
which
is
like
our
compatibility
checking
is
like
essentially
just
like
a
proc
where
you
can
do
some
vague,
like
you
know
you
can
sort
of
if
I'm
understand
like
it's,
it's
sort
of
like
arbitrary
and
is
going
to
eventually
sort
of
lead
to,
and
it's
doesn't
record
any
of
these
details
so
like
we
don't.
E
F
Know
like
what
you
know
so
like
in
datadog
for
context
because
we're
a
vendor-
and
we,
like
you,
know,
customers
complaining.
We
sort
of
like
manually,
maintain
this
like
matrix
of
like
compatibility
stuff
and
it
is
a
hassle
and
it's
but
that's
fine.
The
real
hassle
is
in
the
testing
the
rigging.
F
And
sort
of
like
up
using
appraisal
effectively
and
it
leads
to
and
then
also
like
the
test
environment,
and
it
just
kind
of
leads
to
this
like
huge
test
float,
which
is
sort
of
necessary
and
we've
a
little
bit
put
off.
So
I
think
that's.
The
real
concern
for
me
is
not
necessarily
a
dsl
around
like
hey.
How
do
we
do
some
sort
of,
like
you
know,
dsl,
for
ensuring
min
and
max
versions
like
hey?
We
support
ruby
for
rails,
four
up
to
rails,
like,
for
example,
like
rail.
F
F
We
can,
and
so
there's
a
few
stuff
like
that,
where,
like
we're
not
going
to
in
practice,
we
probably
won't
be
vigilant
enough
to
be
tracking
when
every
instrumentation
you
know,
comes
out
with
a
new
major
version
and
we'll
just
accidentally
and
we'll
be
finding
out
about
these
incompatibilities
through
application
crashes,
which
is
like
a
terrible.
D
That's
unreasonable
to
to
like
make
make
the
library's
dependencies
rigid,
which
will
limit
the
test
case
explosion
for
for
you
all
right
and
then,
but
as
a
administrator
like
somebody
who
supports
a
bunch
of
teams,
writing
this
code
I'd
rather
that
it
break
in
their
test,
their
ci
process.
You
know,
or
just
they
get
a
notice,
be
like
hey
you're,
not
going
to
get
faraday,
because
you
know
we
don't
support
this
version
right
now
and
if
it's
like,
but
I
guess
what
I'm
saying
was
like.
F
Exactly
yeah,
so
I
think
I
think
we're
on
the
same
page.
It's
just
maybe
a
matter
of
like
we
need
to
clearly
define
what
the
work
is.
That
needs
to
be
done
here,
which
is
like
there's
a
few
different
parts
that
are
related
but
separate
chunks
of
work
and
sort
of
prioritize,
like
which
of
those
is
first
to
me
like
it,
feels
like
the
first
thing,
is
figuring
out
a
way
to
just
be
like
guess
what
you're
not
getting
a
waste
of
everything.
F
All
the
time
like
we
have
to
manually
say
what
the
latest
acceptable
version
is
like,
but
yeah,
I'm
not
sure
and
then
yeah.
So
I
you
know
you
expressed
a
lot
of
things
there.
I
think.
F
F
Approver,
like
you
know,
I
don't
care,
go
break,
it's
not
my
problem
but,
like
you
know,
I'm
putting
on
my
my
vendor,
hat,
which
I
still
have
in
my
closet
somewhere
like
I
need
to
you
know
like
customers
will
churn
if
they
all
of
a
sudden,
are
using
unreliable
libraries
under
the
hood.
So
you
know,
I
think
you
know.
Maybe
the
first
thing
is
like:
let's
spec
out,
let's
define
these
issues
a
little
more
clearly
and
issues,
and
then
we
can
work
on
seeing
who's
open
to
doing
that
work.
F
I
think
some
of
it's
interesting,
you
know,
like
the
testing
stuff,
is
fun
for
people
who,
like
testing
so
other
people,
but
you
know
yeah.
F
A
F
Okay
cool,
so
I
just
somehow
give
myself
more
work.
Okay,
what
else
there
is!
So
if
there's
no
other
off
the
cuff
things,
I
think
there's
a
few
things.
Maybe
we
can
talk
about.
I
can
quickly
run
through
this.
Let
me
take
five
minutes
and
run
through
the
sig
and
then
there's
one
or
two
things
for
ruby.
I
think
are
worth
mentioning
so,
okay,
metrics,
I
know
tristan,
I'm
curious
where
you
stand
on,
you
know
needs
for
the
metric
stuff.
F
I
know
robert
has
keen
interest
in
doing
the
api
implementation,
but
is
also
looking
for
help.
So
if
anyone
else
here
like
is
interested
in
the
metric
side
or
finds
a
use
case
for
it
like
I,
you
know
we
kind
of
need
to
get
started,
but
I
know
I
don't
want
to
speak
on
robert's
behalf.
I
know
he's
very
interested
in
this
work,
but
I
think
more,
the
merrier,
the
cool
stuff
cool
sampling
as
far
as
sampling.
I
think
this
is
like
close
to
something
we're
going
to
have
to
implement
soon.
F
It
seems
pretty
close
to
being
done.
If
I
recall
correctly,
it's
important
for
production
users
and
should
there's.
I
think
it's
one
of
the
blockers
is
like
the
I'm
not
super
familiar
with
the
otep.
F
I
think
some
of
the
blockers
is
like
they're,
not
quite
sure
like
what
needs
to
get
updated
in,
like
the
w3c
specification
itself,
like
whether
some
of
these
details
should
live
in
w3
c
specification
or
whether
they
can
just
like
get
added
as
a
no
tip
and
it's
fine
and
then
maybe
some
like
minor
bike
shedding
on
like
when
it's
not
bike
shedding
because
there's
math
involved
but
like
on,
like
the
right
format
of
like
how
to
represent
probability.
F
I
know
there's
some
folks
who
are
much
better
at
maths
than
me
that
are
like
have
found
some
really
thoughtful
ways
to
use
less
bytes
to
represent
probability,
but
I
think
it's
like
more
or
less
like
done.
It
feels
like
outside
of
a
few
issues
and
so
like
yeah.
If
people
are
interested
in
sampling,
this
could
be
future
work
for
ruby
soon.
F
I
don't
know
anything
about
ports.
Jesus.
If
I
want
to
talk
about
real
quickly,
I
know
things
are
getting
so
right
now,
like
it's,
not
our
exporter
doesn't
gzip.
By
default.
We
turned
it
on
internally.
I
think
we
saw
some
errors,
so
we
need
to
investigate
those.
F
So
that's
like
giving
me
some
apprehension,
I'm
not
sure
if
we
handle
like
there
might
be
some
issue
with
like
threading
and
like
the
library
the
way
like
g-zipping
works,
where,
like
I
don't
know
this
stream
things
that
I
haven't
looked
into,
but
we
saw
some
sort
of
like
scary
errors,
popping
up
like
very
infrequently
internally,
so
I'm
a
little
bit
hesitant
to
turn
gzip
on
by
default
for
ruby.
There
might
be
a
bug
in
our
ruby
implementation
of
you
know.
F
D
Some
other
lacking
issues
going
on,
but
I
haven't
tracked
them
down
yet
because
we
get
some
weird
behavior
under
load
when
the
exporter
gets
busy
that
only
that
only
manifests
under
load.
So
it's
been
really
hard
to
get
in
there
with
the
debugger
and
figure
out.
What's
going
on.
D
A
You're
welcome
to
speak
this
evident.
This
is
specifically
important
to
us
because
we
we've
only
got
as
far
as
I
know,
github
and
shopify
are
the
only
two
big
customers
that
are
using
and.
D
A
Yeah
and
like
at
least
on
the
github
side,
you
know
we're
only
using
maybe
four
or
five
of
the
auto
instrumentation
right
now
like
we.
A
F
What
you
described
amy
is
is
pretty
similar
to.
I
think
some
of
the
the
random
errors
we're
seeing
it's
like
a
very
heavy
service
that
occasionally
like
spits
out
a
you
know,
an
error
under
under
load
saying
something
around
like
I
don't
know
the
threat's
not
getting
close
properly
or
not.
I
got
to
figure
it
out,
probably
something
involving
some
low-level
stuff.
F
I
I
looked
into
the
error
randomly
and
saw
like
fluent
d,
had
a
very
similar
error
like
four
years
ago
that
they
somehow
resolved
so
I
was
like,
but
that
that
was
it.
That's
all
I
looked
at
and
I
was
like
all
right.
I
don't
think
about
it.
Anymore,
cool
yeah,.
D
It
smells
like
some
subtle,
locking
thing
that
rarely
ever
comes
up
except
under
extreme
load,
and
my
thought
was
is
if
it
keeps
coming
up
to
set
up
a
soap
test,
so
just
figure
out
like
an
independent
test,
where
we
just
soak
the
hell
out
of
it
and
just
generate
traces
that
just
go
into
the
collector
and
see
until
we
until
it
breaks
so
I'll.
Put
that
together.
If
it
fails
for
us
we're
going
to
try
again,
probably
next
week,.
A
Friends,
yeah:
do
you
have
any
log
log
information
that
you
might
be
able
to
share
specific
even
on
that
one?
Pr
like
there's
one
pr
that
the
person
has
been
impatient,
yeah
johnny
about
merging,
and
perhaps
we
can
provide
warnings
for
them.
F
Yeah,
mr
shields,
the
coolest
name
in
the
world
has
been
asking
to
get
this
out,
which
is
super
reasonable.
I
think
there's
performance
benefits
yeah.
I
think
I
need
to
maybe
sync
tim.
Maybe
we
need
to
sync
with
robert
and
francis
when
we,
but
we
could
probably,
at
the
very
least
give
you
know
like
include
some
of
the
the
eric
type
we're
seeing
and
then
some
of
the
stack
trace.
I
don't
think
there's
anything
like
proprietary
in
there,
but
it's
exactly
it's
very
similar.
F
What
you
described
amy
so
anyway,
that's
concerning
to
me
because
it
might
become
required
soon,
where's,
the
spec
cool,
so
all
right,
so
we'll
try
to
update
that
messaging.
F
Okay,
I
don't
know
they're
updating
the
conventions
if
people
have
interest
in
in
messaging
spans
or
instrumentation.
Certainly
these
working
groups
are
generally
the
right
place
to
get
involved
oftentimes
by
the
time
the
otep
hits
the
hits
github.
It's
people
kind
of
have
already
formed
consensus,
so
yeah.
I
think.
F
F
Yeah,
you
know
I've
always
felt
published,
consume
stuff
as
a
bit
of
a
like
people
are
just
of
like
covering
their
eyes
and
being
like
yeah.
Open
telemetry
works
great
for
this
stuff,
like
I
don't
think
it's
super
well
defined
and
that's
fine.
F
Are
it'll
be
fine
yeah,
it's
disconnected
from
the
that's
part
of
the
batch,
whatever
so?
Okay,
so
that's
interesting,
I
know.
There's
probably
I'm
I'm
not
super.
I
haven't
been
involved
some
okay,
some
attribute
updates
related
to
sampling.
If
anyone
has
thoughts,
I
don't
know
that's
interesting.
Oh
I
guess
so
one
one
issue
with
making
sampling
choices
is
you
need
to
have
access
to
certain
attributes
like
say,
okay,
instead
of.
F
Oh,
I
don't
know,
I
guess
the
you
have
to
know
the
name
of
where
it
came
from
or
something
so
there's
this
new
field
called
sampling
relevant
currently
in
samplers,
you
pass
in
some
information,
but
sometimes
information
like
when
it
comes
to
the
actual
implementation
and
that
information
gets
appended
to
the
span
after
the
sampling
decision
has
been
made
so
yeah.
I
haven't
looked
at
this
at
all.
E
B
This
also
brought
back
the
something
that
was
originally
or
at
some
point
in
the
spec
of
lazy
attributes
that
are
evaluated
only
if
they're
needed,
which
so
that's
brought
up
again,
it
might
get,
might
be
some
work
on
that
coming
soon.
F
Sounds
it
sounds
reasonable,
yeah
there's
it
feels
like
there's
a
few,
including
like
this
they're
they're,
making
a
change.
I
think
this
got
merged.
There's
going
to
be
a
few
of
these
pieces
of
work
that
need
to
get
done.
This
one.
Is
this
next
one's
pretty
straightforward
as
well?
It's
basically,
I
think
previously.
F
If
you
got
like
a
404,
if
your
rails
app
returned
to
404,
that
was
marked
as
an
error
on
the
server
span,
which
it's
not
you
know,
so
they
updated
that
to
to
reflect
that
so
we'll
need
to,
I
guess,
make
sure
we
have
that
handle
that
implementation.
I'm
not
sure
if
we
do.
One
thing
to
be
aware
of
is
like
your
your
your
your
error,
stats
could
suddenly
change
and
it
could
seem
like
your.
Your
services
became
a
lot
more
resilient
because
half
the
errors
aren't
getting
marked
as
errors
anymore.
F
That's
a
little
like
concerning
to
me
yeah,
it's
good
news.
This
is
once
this
is
the
sort
of
behavior
that
I
would
like
love
for
the
schema
url
and,
like
all
that
stuff
to
be
able
to
handle,
because
there
is
this
change
occurring
like
implicitly,
but
it
totally
doesn't
handle
at
all.
But
that's
something
we
need
to
this
is
work
we
need
to
do.
I
think,
and
then
there
was
some
discussion
around
metrics
okay,
so
that
was
the
spec
spec
sig
for
ruby.
F
We
have,
we
have
10
minutes.
Is
there
any
ruby
topics?
People
want
to
discuss
or
is
there
I
can
kind
of?
We
can
review
the
the
open
issues
or
pr's
it's
one
of
these
days.
Can
I
find
it
and
see
if
there's
anything,
I
have
some
pr's
up
for
that.
I
could
appreciate
if
anyone's
interested
in
reviewing,
I
tried
to
if
no
one
else's
other
things
they
want
to
discuss
I'll.
Just
ramble
about
the
stuff
I've
been
working
on
cool
yeah.
F
I
made
there's
some
config
improvements
improvements.
I
don't
know
yeah
tim
for
sure.
H
Yeah
there's
just
one
thing
that
I
wanted
to
mention.
One
main
thing
that
prevented
us
from
using
active
support
notification
a
little
bit
more
is
the
unpredictable
behavior
when
handling
exceptions.
I
left
it
in
the
in
the
sig
dock
and
it
was
just
merged.
So
if
you
guys
want
to
learn
more
about
it,
it's
all
there.
Now,
if
you
have
a
complex
subscriber,
you
can
guarantee
that
the
start
and
end
will
be
cold
before
raising
exceptions.
Awesome.
A
Oh,
this
guy
is
that
getting
rolled
out
in
a
dot
version.
I
I'm
not
sure.
F
Sorry
yeah:
these
are
good
sound
like
smart
folks
yeah.
This
is
good,
so
the
previous
behavior
of
a
lot
of
instrumentation
is
based
on
like
active
support
notifications.
But
if
there's
one
subscriber
that
throws
an
error,
there's
an
exception,
the
other
subscribers
just
don't
get
called.
So
we
were
doing
some
really
gnarly
stuff
to
kind
of
like
move
our
subscribers
to
the
front,
but
it
looks
like
rails
itself
will
handle
this
a
little
bit
more
clearly
more
cleanly
so
long
term.
F
H
Everything
yeah
I'm
still
working
on
adding
the
active
support
subscription
for
those
things.
So
hopefully,
next
week.
F
F
Oh
anyway,
I
had
a
pr
to
enable
you
can
enable
instrumentation
configuration
via
environment
variable,
and
it
also
adds
some
helpers
and
stuff
around
declaring
instrumentation
configure
option
configuration
options
deprecated
it's
a
little
bit
of
a
work
in
progress,
but
I'm
kind
of
gave
it
a
gave
it
a
shot
of
seeing
like
you
know
so
previously
like
if
you
wanted
to
you,
could
turn
some
things
off
and
on
through
environment
variables,
but
this
extends
that
I
actually
haven't
like
looked
at
the
spec
at
all,
so
some
of
the
stuff
may
be
like
way
off
spec.
F
I'm
really
not
sure.
I
just
wanted
to
kind
of
try
to
get
something
up
that
was
working
and
then
look
for
feedback.
So
if
people
have
feedback
around
this
or
have
use
cases
yeah,
your
your
feedback
is
welcome.
I
also
am
not
very
good
at
ruby,
so
I
was
having
a
little
bit
of
trouble
finding
like
the
right
pattern
to
do
this
stuff
in
so
yeah.
If
you
have,
you
know
that
sort
of
feedback
is
also
welcome,
and
ideally
what
this
also
does
is.
F
It
allows
us
to
declare
all
the
configuration
options
in
like
a
constant
like
in
one
place
for
instrumentation,
currently
right
now,
it's
sort
of
just
like
there's
a
bunch
of
invocations
of
like
option
scattered
around
that
make
it
really
hard
to
document
like
well
what
instrumentations
have
what
configuration
options,
so
I've
sort
of
added
a
syntax
for
saying,
like
okay,
if
we
just
store
all
our
instrumentations
configuration
options
like
on
this
constant
and
I've
added
some
helpers
to
them
like
take
that
constant
and
translate
those
into
like
actual
options
that
get
applied.
F
We
can,
then,
you
know,
add
a
whatever
a
rake
task
or
something
to
then
auto
generate
documentation
pretty
easily,
which
should
be
a
nice
onboarding
win
for
folks
who
are
like
you
know,
oh,
I
want
to
use
rack,
but
I'm
not
sure
how
to
obfuscate
my
query,
parameters
or
whatever
so
yeah,
take
a
look
if
you're
interested
and
again
sorry
that
the
code
itself
sucks.
D
I'm
glad
to
see
that,
because
what
we've
been
doing
with
our
rollout
is
everything
is
configured
via
environment
variables.
There's
no!
We
don't
put.
I
try
not
to
put
any
hotel
config
in
the
application
itself,
and
so
we
just
deliver
all
the
configures
environment,
variables
and
kubernetes,
and
what
that
does
is
I
default
everything
to
off
when
there's
no
environment
variables
like
if
the
exporter's
endpoint
isn't
set
everything's
in
non-recording?
F
Nice
yeah,
I
think
that's
a
good
pattern
and
some
of
this
stuff
is
going
to
get
done
via
environment
variable
like
this
way
anyway,
and
this
is
sort
of
just
an
attempt
to
like
standardize
some
of
that
stuff,
so
that
you
know
it's
yeah,
it's.
Basically
it's
easier
document
and
easier
for
onboarding,
but
yeah
you
can.
Ruby
is
a
magical
language
and
it
allows
you
to
do
lots
of
things
such
as.
Like
you
know,
you
can
just
like
easily
use
your
own
environment
variables
as
well.
Yeah,
okay,.
F
I
don't
know
that
was
kind
of
it
for
me.
F
Yeah,
I
don't
know,
I
don't
really
have
anything
else.
I
know
sorry
for
just
talking
for
an
hour
at
all
of
you.
I'm
not
used
to
doing
this.
F
Yeah
yeah
and
just
general
note,
if
there's
things
here,
that
weren't
covered
or
there's
areas
you're
interested
in
like
that,
don't
seem
prioritized
like.
I
think
the
approach
here
is
like
if
there's
work
you
want
to
do
like.
I
think
people
are
happy
to
see
any
work
done
or
any
you
know
it
doesn't
have
to
be
super
critical
stuff.
F
It
could
just
be
like
something
you're
interested
in
and
and
and
yeah,
if
you're
not
getting
reviews
on
that
stuff
like
paying
in
the
chat
and
try
to
get
reviews
like
I
know
my
colleague
made
a
pr
dad
like
an
instrumentation
that
we
need
to
review
and
stuff
like
that.
So
yeah,
I
think
no
one
is
trying
to
gatekeep
like
any
of
this
work.
F
It's
it's
just
a
matter
of
you
know
trying
to
manage
the
stack
rank
some
of
the
work
for
practical
reasons,
but
don't
feel
as
though,
if
you
want
to
do
something,
that's
lower
priority
or
applies
only
for
your
use
case
and
doesn't
have
broader
adoption
like
that's
still
super
welcome.
It's
just
you
know,
yeah,
not
I'm
not
talking
directly
to
you
amy,
but
I
didn't
want
you
to
feel
like.
I
didn't
want
you
to
feel
like.
Oh,
they
didn't.
D
D
F
D
F
That's
good
to
know
yeah
I'll
try
to
update
as
we
learn
more
internally
as
well
in
the
chat
and
yeah
also
totally
fine
to
join
these
meetings,
and
just
like
you
know,
puts
around
and
give
you
know,
comments
and
stuff.
That's
mostly
what
I
was
doing
at
datadog
to
be
honest,
so
yeah,
it's
totally
fine
as
well.
I
think
more,
the
merrier
here
and
we're
just
looking
for
input
invoices
from
the
community.
F
To
be
honest,
so,
like
it's
awesome
to
see
so
many
such
a
diverse
group
of
like
folks
from
different,
you
know
like
different
use
cases
and
and
yeah
I'll
see
you
all
next
week.
I
guess
give
you
back
30
seconds
of
your
day.