►
From YouTube: 2023-01-24 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
B
A
B
We've
We've
noticed
that
in
other
online
documents
thing
every
now
and
then
we
just
have
to
like
rename
it
actually
we'll
copy
the
content
into
a
new
one
and
then
clear
it
so
links
continue.
A
C
D
C
B
C
We
talked
about
this
a
couple
weeks
ago.
There
was
an
interesting
proposal
for
empty
or
invalid
Trace
IDs
for
the
protos,
so
you
should
generate
a
new
one,
but
that
was
very
controversial
because
it
would
look
like
you
are
potentially
missing
a
trace
that
you
would
expect
to
have
so
this
one
just
says:
tastytrace
ID.
C
As
long
as
the
satellite,
it
will
be
valid
if
it's
all
zeros
it's
invalid
or
if
it
is
less
than
16
bytes
it's
invalid,
but
it
allows
this
to
be
to
be
invalid.
If
it's
invalid,
which
I
think
is
it's.
C
You
receive
an
invalid,
so
technically,
I
think
what
should
happen
is
like
an
SDK
should
just
not
send
something.
That's
invalid,
but
if.
B
I,
personally,
don't
care
so
long
as
it's.
The
relatively
unique
identifier,
I,
think
and
I
I
haven't
read
this,
but
the
only
problem
I
see
is
if
something
if
passing
something
along,
that
is
invalid
means
it
might
not
serialized
to
otlp.
C
C
C
C
We
talked
about
this
one
last
time
as
well.
I
think
there
was
there's
already
this
pool.name.
C
Semantic
convention,
but
they
were
expanding
the
definition
in
the
case
the
pool
has
no
name
then
make
a
name
for
the
pool
as
follows:
host
Port
database
name
yeah,
so
synthesize
a
a
pool
from
the
host
Port
database
name
and
user
I
think.
C
There
was
more
of
this
general
question
about,
like
I,
think
something
came
up
in
the
python
thing
and
they
were
wondering
wondering
at
what
level
you
serve
as
the
specsig
for
specification
and
I.
Think
the
the
answer
was
kind
of
I
think
the
general
thing
that
we
usually
try
to
do
when
things
come
through.
C
This
thing
is
figure
out,
like
figure
out,
if
there's
some
existing
mechanism
to
kind
of
support
the
thing
and
if,
if
it
legitimately
looks
like
there
is
something
else
needed
then
start
to
start
with
this
back
issue,
bring
it
up
to
the
spec,
say
and
then
see
see
what
the
best
way
to
achieve
that
work
is
some
stuff
it
can
just
is.
A
C
C
Yeah
and
this
this
was
around
adding
some
like
It,
ultimately
was
spurred
by
adding
some
some
new
capabilities
to
a
play.
Metric
reader,
but
won't.
C
Think
it
was
soon
it
was
closed,
but
one
thing
that
is
specifically
here
for
metric
readers,
the
next
issue-
and
that
was
adding
a
force,
flush
operation.
C
It
exists
for
for
spandexport,
so.
C
C
B
C
In
any
case,
having
a
force,
flush,
I,
think
to
to
gracefully
shut
down,
your
metrics
pipeline
seems
seems
reasonable.
C
They
were
adding
a
recommendation.
Are
yeah
updating
this
back
to
recommend
that
by
default
database
queries
are
are
sanitized.
C
So
I
think
it's
to
avoid
sending
pii
I'm,
not
sure
what
we
I
default
to
in
in
Ruby
I
know
we
have
the
option
to
to
obfuscate,
but
all
these
things
kind
of
come
out
of
cost.
Ultimately,
so.
C
So
yeah,
all
these
things
come
out
of
costs,
and
you
know
the
debate
was:
should
this
happen
uniformly
somewhere
else
like
in
a
collector
and
is,
is
a
collector
too
distant,
for
you
know
SQL,
potentially
with
pii
to
travel,
and
if
it
is
within
your
own
trust
boundary,
it's
probably
fine
and
might
be
a
better
place
for
it.
B
D
Okay,
if
you
use,
find
Bass
ql,
there
is
no
there's
no
technology.
B
C
You
will
get
the
the
templated
statement
with
with
the
fine
params,
though
too
we
will.
You
not
I,
think
it's
only
definitely.
A
B
So,
what's
the
spec
say
the
suggest
and
a
recommendation
to
sanitize
by
default
cool,
it's
a
good
recommendation
when
possible
and
if
the
thing
that
you
were
instrumenting
does
not
give
you
a
mechanism
to
thank
you
easily,
feasibly
genericize,
query,
Sanitation
sanitization,
then
I'm
like
we
put
warnings
on
the
thing
and
say
if
you
instrument
your
your,
if
you
turn
on
database
instrumentation
and
you
use
find
by
SQL,
there's
no
way
for
us
to
sanitize.
B
That
here
are
some
options:
send
it
through
a
collector,
at
least
as
an
initial
state,
for
data
store,
query,
sanitation,
type
stuff.
So.
B
Is
it's
in
the
agenda
add
a
recommendation
to
sanitize
DB
statement
by
default.
B
C
Got
it
got
it
so
I
would
have
to
look
again,
but
I'm
pretty
sure
are
or
I
do
think
that
active
record
will
give
you
a
templated
query
with
the
buying
params
and.
C
Was
some
refactor?
They
did
in
active
record?
There
was
some.
There
was
some
big
SQL
injection
issue
in
rails
at
one
point
in
time
and
I
think
we
we
benefited
from
from
some
of
that.
But
I
don't
know
this
was
a
while
ago
and
my
memory
may
have
filled
up
with
more
important,
pressing
things
and
displaced
these
old
haunting
memories
from
the
past.
C
So
it
might
be
worth
a
little
experimentation,
especially
as
we're
moving
over
to
active
support
notification
for
everything,
but
I
think
the
thing
that
will
shoot
us
in
the
foot
is
that
we
do
have
like
native
MySQL
and
postgres
instrumentation
as
well,
and
those
definitely
are
probably
going
oh
yeah
down
at
the
client
level.
Obviously,.
D
I
think
the
gist
of
it
is
kind.
You
know
we
have.
We
we've
kind
of
just
been
cherry
picking
the
New
Relic
instrumentation,
and
so
it
might
make
sense
for
us
to
I'll
do
something
a
little
more
sophisticated
than
that.
B
D
The
the
the
redis
memcached,
all
of
them,
Dolly
I,
guess
all
of
them.
So
far
we
have
added
the
same
attribute
to.
B
B
D
D
If
you
look
at
the
J
Flex,
like
the
pr
that
was
associated
with
this
there's
a
lot
of
chatter,
around
performance
concerns
with
this,
like
we've
had
other
people
say
to
us
like
we
capped
the
limit
at
like
2
000
characters
arbitrarily
for
what
you
know,
we're
gonna,
allow
or
disallow,
because
it's
not
very
fast
right
for
for
current
implementation
and
folks
have
been
asking
to
increase
the
size
of
that
through
various
PRS.
D
D
So
systems
like
the
tests
accept
con.
You
know
context
propagation
parameters
through
comments
and
they
support
the
Jaeger
Trace
ID
format,
Trace
context
format,
so
you
can
connect
traces
generated
from
the
tests
two-year.
D
You
know
propagated
contacts
and
SQL
comments,
which
is
a
rather
curious
thing
that
we've,
you
know
that
I'm,
that's
also
in
my
head
about
our
SQL
instrumentations
in
general,
be
very
curious
about
you
know
doing
something
similar.
B
B
Let
me
make
sure
I'm
saying
that
right
in
the
open
Telemetry
organization
in
GitHub,
there
is
an
open,
Telemetry,
Dash
SQL,
commenter
project
that
has
within
it
a
SQL
comment
or
Ruby
and
rails
component
that
uses
marginalia
and
open
census
to
do
tracing
I.
Think
through
SQL
comments.
So
there's
a
few
different
paths
for
us
to
explore
there
for
that.
I
think
yeah.
D
B
I
bring
that
up,
because
there
are
some
users
of
Hotel
Ruby
who
are
probably
interested
in
this
domain
of
propagating
and
tracing
to
their
data
stores
that
support.
So
it's
something
to
look
at.
C
All
right,
channeling,
Sam,
lettuce,
yeah,
no
I,
think
it's
a
good
good
conversation
and
a
Hot
Topic,
and
one
obviously
that
we've
been
thinking
about.
Since
we
already
have
an
issue,
so
we
should
yeah.
We
should
continue
that
I
think
it's
an
interesting
CS
project.
If
nothing
else.
C
They
are
perhaps
one
of
the
oldest
working
groups,
but
we
now
have
the
new
the
new
Ted's
process,
by
which
we
hope
to
actually
finish
semantic
conventions
by
summer
2024,
the
summer
of
Tito's
to
come
and
there's.
C
It's
a
three-step
process,
working
group,
preparation
and
specification,
then
implementation
and
the
HTTP
semantic
conventions
group
is
officially
trying
to
like
kick
off
and
has
helped
wanted
and
from
our
starting
in
stage
one
right
now
stage.
One
is
happening
now
stage
two
will
begin
soon.
So,
if
anybody's
interested
this
issues
for
you
pylon
and
what
does
kind
of
be
monitoring
this
for
for
changes
and
see
see
what
it
means
for
parsing
and
all
of
our
HTTP
instrumentation.
C
C
C
C
So,
looking
at
the
agenda,
it
does
appear
that
we
have
some
some
issues
to
discuss.
Arielle.
Would
you
like
to.
D
This
was
my
attempt
at
like
bringing
some
stability
to
the
build,
so
something
that
happens.
A
lot
is
when
we
set
up
our
initial
sort
of
like
a
matrix,
build
for
the
gems.
All
of
the
services
would
start
at
once
in
one
instance
of
a
GitHub
actions
container.
D
And
within
it
you
know
it's
doing
like
Doc
reception
right,
so
you
you've
got
a
bunch
of
services
that
depend
on
it.
So
if
you're
running
Ruby
Kafka,
you
end
up
spinning
up
all
of
the
other
services
in
that
build
that
don't
relate
to
the
current
test
that
you're
running
for
that
Matrix.
C
D
D
The
other
one
Artie
Kafka,
often
seg
fault,
so
the
more
you
run
them
the
longer.
The
the
higher
likelihood
it's
going
to
set
fault,
causing
the
entire
like
Matrix
to
fill.
So
the.
C
D
Use
this
lib
Kafka
under
the
hood
so
making
a
long
story
longer.
What
I
decided
to
do
is
say:
okay,
for
a
specific
targeted
version
of
a
of
a
library
say
for
Kafka
I'm,
going
to
make
a
matrix
specifically
for
Kafka.
That
only
includes
the
gems
that
are
drivers
or
instrumentations
related
to
Kafka
and
we'll
run
those
together
in
a
group
for
MySQL
I,
have
the
MySQL,
2,
Gem
and
Trilogy
gem
in
there
for
redis
rescue,
sidekick
and
reddish,
gems,
and
so
on,
and
so
on
and
so
forth.
D
So,
by
doing
this,
we're
trying
to
increase
the
stability
of
avoiding
containers
to
crashing
containers
crashing,
bringing
the
whole
test
Suite
down
and
giving
us
faster
feedback
loop
for
those
particularly
volatile
gems,
like
Artie,
kafko
and
foreign
car,
and
that
way
you
can
kick
off
that
failed
job,
much
faster.
D
What's
the
side
effect
of
this?
What's
the
downside
of
this
is
that
we've
increased
our
footprint
again
of
the
number
of
builds
in
The,
Matrix
I.
Think
we're
up
to
with
this.
It
brings
it
up
to
44,
something
like
that
and
I
had
also
as
part
of
this
I
had
a
failed
attempt
at
trying
to
take
the
linters
out
and
isolating
the
linters
into
their
own
build,
but
I
rolled
all
that
back.
So
this
is
really
only
focused
on
the
data
store
service
related
things
to
be
isolated.
B
Not
philosophically
I
agree
with
the
direction
to
to
reduce
the
blast
radius
of
interactions,
yeah.
C
Yeah,
no,
it
looks,
looks
awesome,
looks
like
good
work
and
things
yeah
things
that
will
definitely
benefit.
B
Since
we
test
four
different
rubies
in
there
is
there
do,
are
we
prioritizing
how
many
builders
get
spun
up
to
reduce
the
number
of
builders
that
get
spun
up
or
reduce?
The
number
of
sorry.
D
D
right,
because
it
would
include
jruby,
nines
web
scale,
whoops
I
know,
but
our
organization
has
a
limited
number
of
actions,
Runners
that
it
can
use
okay,
so
we'll
be
eating
into
other
people's.
You
know:
okay,.
B
Well,
parallelizing,
on
splitting
out
the
spinning
up
of
third-party
services
that
are
themselves
instrumented
or
a
dependency
of
that's
an
instrumental
thing
is:
will
help
cool.
D
B
A
A
D
D
If
you
missed
it,
I
merged
a
dependabot
PR
that
it
checks
our
dependencies
for
us.
So
I
don't
have
to
do
this
manual,
and
here
they
go.
This
is
what
it's
gonna
look
like.
D
So
this
is
interesting,
so
A
new
cop
exists
of
you.
If
you
click
through
my
comment.
D
Somebody
added
this
new
cop,
which
was
hey,
take
all
of
your
development
dependencies
out
of
your
gem
spec
and
stick
them
in
your
gem
file.
How
do
people
feel
about
that.
B
A
B
But
I
could
see
how
it
would
maybe
make
sense
if
you
were
like
a
service.
My
my
Ruby
service,
I,
don't
distribute
as
a
gem
if
you're
going
to
work
against
it,
you're
going
to
clone
the
repo
and
so
you're
going
to
work
off
the
gem
files
declaring
the
dependencies
I
gotta
go
read.
This
cops
like
defend,
defend
yourself
cop.
Why
you
exist?
B
B
B
Also,
maybe
I
just
don't
know
if
there
are
more
options
like
appraisal
will
merge
like
I,
don't
know
what
problem
this
solves.
I.
D
Guess
look
at
the
last
comment.
This
is
great.
It's
like
a
conspiracy.
Everyone
acts
like
it's
a
well
thought
out
super
useful
feature
where
no
one
can
explain
why
yeah
so
I
found
one
discussion
about
this
before
we
use
well,
when
I
was
trying
to
understand
when
if
I
should
use
ad
development
dependencies.
B
B
A
D
D
D
B
D
You
know
everything
else
is
like,
pending
whatever
draft
mode
right
now
and
I'm
good.
So
as
people
have
time,
you
know,
if
you
see
these
little
dependabot
issues,
you
know
come
up
just
take
a
look,
make
sure
nothing's
problematic,
give
it
a
review
say
thanks
for
the
fries
and
then
merge
it.
D
You
know,
and
these
will
run
I
think
once
a
week
on
Mondays
I
did
the
same
thing
for
the
getup
actions
by
the
way.
A
C
C
C
B
New
brand
crossover,
Hotel
Ruby
distributions,
which
I
think
GitHub
and
Shopify
have
as
just
internal
company
instrumentation
rapper
libraries.
A
D
B
And
github's
hotel
I
in
my
experience
in
dealing
with,
say,
open
Telemetry
Java.
There
is
like
how
to
build
a
distro.
B
I'm
in
a
position
of
putting
cobbling
together
a
distribution
and
I'm
curious
if
we
have
any
guidance
and
if
we
don't,
if
in
my
process
of
creating
one,
we
create
that
guidance
like
start
seeding
some
instructions
on.
If
so,
you
want
to
write
a
wrapper
and
like
control
your
users
or
guide
guide
your
users
use
of
hotel
in
your
organizations
style.
B
B
I
might
what
I,
what
I
provide
to
you
to
review
is
probably
going
to
be
messy
because
you
have
more
experience.
Rapping
Hotel
Ruby
than
I
do,
but
it's
a
place
to
start.
D
B
D
D
Like
okay,
so
like,
for
example,
like
an
automatic
instrumentation
like
in
my
personal
experience,
use
causes
too
many
problems
for
me,
so
I
use
the
use
all
bucket
yeah.
The
other
thing
you
know
with
selectively
picking
specific
instrumentations
as
opposed
to
using
the
All
build
I,
don't
use.
This
bundle
require,
instead
of
using
manual
required.
D
D
So
that's
kind
of
like
Advanced
usage
I
would
not
like
you
could
probably
like
steer
people
in
that
direction,
and
the
other
thing
that
we
don't
have
is
kind
of
like
here's,
a
environment
variable
and
what
that
is
in
the
spec
and
here's
what's
not
supported
by
us
or
what
is
supported
by
us
and
what
you
would
probably
want
to
do
like.
We
don't
have,
for
example,
like
the
the
description
of
you,
can
also
enable
and
disable
instrumentations
using
these
environment
variable
settings
or
whatever
or
pass
these
through.
D
There
isn't
anything
specifically
for
in
this
documentation
related
to
how
to
configure
our
instrumentations
or
what
the
configuration
options
are.
You
have
to
go
to
the
readme
to
figure
out
each
individual
instrumentation.
D
Which
is
something
that
like
comes
up
a
lot,
any
conversations
that
I
have
with
people
and.
D
D
D
That
I
can
think
of
off
the
top
of
my
head
of
the
stuff
that
you
know.
Maybe
we
have
like
a
Sinatra
specific
section,
but.
B
As
a
as
a
vendor,
putting
together
a
distro,
a
fair
amount
of
what
you
listed
would
be
things
that
we
would
not
handle
in
our
distro
like
I
would
because
that
would
be
too
opinionated
of
how
to
configure
the
auto
instrumentation.
We
would
say,
use
all
you
list,
the
ones
that
you
want
to
modify
the
configuration
for,
and
we
would
Pat
like
teach
our
users
how
to
read
the
hotel,
Ruby
Upstream
documentation
on
this
Auto
instrumentation,
because
it's
not
honeycomb's
instrumentation,
it's
Hotel,
Ruby's
instrumentation,
so
you
learn
how
to
configure
it.
B
It's
more
of
the
the
easy
on-ramp
vendor
distributions
are
usually
about
easy
on-ramps
like
configure
for
configure
a
no
title
thing
to
send
to
a
back
end
and
then
all
the
instrumentation
is
still
like.
B
D
D
B
B
There
is
a
sampler
that
we
could
contribute.
Upstream.
D
B
C
To
think
I
will
at
least
add
to
all
this
is
yeah.
Lifestop
has
made
some
wrappers
for
some
languages
and
I
feel
like
the
configurability
varies
wildly
between
the
different
languages
and
the
amount
of
just
like
boilerplate
that
you
need
to
get
up
and
going
with
with
different
languages.
It
just
ends
up
being
like
I.
Don't
know,
I
think
us
who
are
in
a
hotel
on
a
day-to-day
basis
can
figure
this
stuff
out,
but
we
still
make
mistakes
like
yeah.
B
A
A
D
Can
tell
you
I
could
tell
you
about
GitHub
specific
implementation,
and
that
is
that
there
are
things
that
are
not
an
old
test,
Decay
right
now
that
we're
trying
to
do
all-in-one
delivery
for
right.
So
like
we
use
that
at
the
end
of
the
you
know,
for
our
metrics
otlp
for
tracing
and
then
log
fmt4
logging
in
most
cases
right,
so
we
didn't
use
the
old
testicate
for
logging
or
they'll,
tell
us
Decay
for
metrics.
D
D
There's
us
injecting
our
logger
and
adding
Trace
context
to
the
log
lines,
that's
something
that
is
done
by
our
wrapper
code.
Yeah.
We
don't
have
like
a
fully
functioning
rail
tie
as
part
of
our
instrumentation
package,
so
we've
added
a
Rel
tie
that
does
all
that
bootstrapping
of
all
those
components
and
gets
them
all
working
together,
all
all
at
once.
Right.
So
that's
what's
included
in
our
GitHub
distro,
but
it's
not
really
a
distro
or
or
like
those
we're,
not
adding
functionality
to
the
autosdk
or
the
API.
B
D
And
it's
like
it
has
all
the
default
environment
variables
and
it's
like
here
are
the
context.
Propagation
ones
that
you
need
here
are
the
the
minimum
set
of
instrumentations
that
we
have
approved
for
usage
in
production,
because
there's
a
lot
of
instrumentations
that
we
don't
use.
So
we
don't
use
all
like
this
Faraday
as
like
our
like.
If
you're
using
making
HTTP
requests
use,
Faraday,
we
use
obviously
rack
and
rails-
that's
included
included
by
default,
active
record,
the
trilogy,
SQL
and
so
on
and
so
forth.
Right
so
yeah.
B
I
guess
that's
a
that's
a
as
far
as
nomenclature
goes.
That's
probably
a
good
distinguishing
characteristic
of
an
organization's
opinionated
rapper
about
how
we
instrument
at
GitHub
versus
yeah,
a
vendor
going
like
I've
eased
the
on-ramp
to
you
using
the
hotel
with
us.
It's
like
I
still
needed
to.
Let
you
have
opinions
about
most
of
the
of
the
configurations.
Yeah
yeah.
C
So
yeah,
like
I,
had
a
lot
of
thoughts
about
this
in
the
early
days
actually
and
I
feel
like
a
lot
of
these
are
coming
back
around
right
now.
So,
like
you're,
talking
about
the
easy
on-ramp
being
one
element
and
I
think
you
know,
lifestyle
introduced
launchers
in
these
other
languages,
as
the
on-ramp
was
horrible,
it
was
actually
horrible
in
Ruby
at
one
point
in
time
also
and
Francis
and
I
would
lament
about
how
terrible
it
was
to
set
up
like
a
ruby
SDK.
C
So
then
I
don't
know,
I
hacked
together.
This
SDK
configure
and
the
configurator
and
at
least
gave
birth
to
that
whole
system,
and
it
really
like
I,
don't
know
I
feel
like
it's
not
perfect,
like
like
everything,
but.
C
And
this
is
pretty
early
in
hotel.
There
was
really
like
it's
the
wild
west
at
that
point
in
time,
and
there
wasn't
like
a
lot
of
like
how
do
you
bootstrap
your
you
know
your
system,
so
I
think
a
lot
of
so
I
think
that
kind
of
is
why
we
are
in
this
situation.
We
are
with
a
lot
of
languages
in
hotel,
like
some
people
haven't
actually
solved
this
problem
and
it
is
literally,
you
need
to
assemble
all
the
components
yourself
and
others
have
added
various
layers
of
convenience
around
this.
B
C
A
C
I
I
do
think
that
organizations
are
always
going
to
want
to
have
their
opinionated
wrapper
on
top
of
things.
You
know
in
any
any
organization
that
is
larger
than
you
know.
A
small
amount
is
probably
going
to
do
what
GitHub
has
done
or
Shopify
and
just
kind
of
have
their
their
opinionated
bundle.
B
Absolutely
that
all
of
the
stack,
Upstream
vendor
distro
need
to
accommodate
organizations
actually
being
able
to
configure
things
in
there.
According
to
their
opinions
and
not.
A
B
And
the
more
environment
variables
we
tell
some
customers
to
change
the
the
higher
the
likelihood
that
they
don't
do
it
right.
I,
don't
know
how
to
phrase
this
trying
to
keep
it
simple
and
not
have
to
like.
If
you
go
to
the
hotel
docs
and
you
see
this
wall
of
text
of
environment
variables
that
one
and
that
one
you
need
a
set
with
these
keys.
B
If
we
just
go
like
use
our
library,
our
library
looks
for
a
honeycomb,
specific
environment
variable
and
then
everything
gets
set
up,
for
you
is
convenient,
contributing,
adding
vendor
specific
things
to
contrib
is
an
option.
D
B
It
is
intended
to
make
these
things
composable.
There
could
be
like
just
a
honeycomb
exporter
that
does
the
right
thing
and
it
descends
from
the
otlp
Explorer
it
just
configures
a
little
differently.
B
B
C
It's
it's
a
bigger
harder
problem
to
make
all
of
a
hotel.
You
know
easier
to
configure
from
the
get-go,
but
I
feel
like
it's
a
better
long-term
solution.
Yeah,
you
know
the
vendor
specific
stuff.
At
least
I
think
that
we
have
at
lightstep
is
kind
of
just
like
a
stop
Gap,
it's
kind
of
just
like
all
right
today.
We
need
this
because
things
are
hard,
but.
B
The
hopes
are
also
baked
into
the
design
that,
eventually,
if
we
deprecate
the
distros
it's
at
at
some
point,
they
are
just
a
shell
around
upstream
and
go
like
okay.
You've
been
using
this,
but
know
that
it's
just
wrapping
Upstream,
because
we've
brought
you
along
while
we're
in
this
transition
period
of
figuring
out
how
to
make
it
easy.
D
Rob
so
I
I
have
to
split
but
I'm
gonna
kind
of,
like
summarize,
okay,
all
the
things
we
discussed
and
just
really
from
my
perspective,
what
my
preference
would
be
if
we
can
get
things
merged
into
contrary
people
that
would
be
best
for
us
to
the
kind
of
documentation
that
I
think
we're
missing
is
still
like
compatibility
and
hidden
features
that
are
not
available
like
you
know
how
to
configure
an
instrumentation
through
environment
variables.