►
From YouTube: 2022-10-18 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
I
walked
the
dog
early
today,
so
you
all
get
to
stare
at
my
lovely
face.
Why
I
mean
granted
I,
don't
think
anyone
else
is
coming
I,
don't
know.
Maybe
they
are
actually
I
doubt
Sam
is
Coming.
Sam
is
giving
a
talk
at
a
conference
in
New,
York,
awesome,
wow,
yeah
I
need
to
actually
watch
it.
I
I
saw
you
had
a
shirt
out,
a
preview
version.
It
was
about
like
developer
velocity
in
infrastructure
and,
like
are
Sprints
the
devil
in
infrastructure.
Are
they
not
I?
I
need
to
actually
watch
it.
A
The
whole
way
through
I
was
trying
to
cheat
and
watch
it
at
Double.
The
speed,
because
I
was
short
for
time,
but
I
need
to
actually
like
watch
it
watch
it
gotcha
yeah
how
you
been.
A
A
B
There
was
nothing
in
there
other
than
what
I
just
added
to
now.
A
C
B
You're
running
a
little
bit
overnight,
but
then
it
didn't
rain
anymore
and
you
know
for
all
you
people
watching
at
YouTube
on
YouTube
at
home.
This
Friday
is
the
Circuit
of
the
Americas
and
I'm
taking
my
son
to
go,
see
Green
Day,
they're
performing
live
there
on
Friday
night,
so.
B
Oh,
how
awesome
just
let
them
know
that
I'll
be
there
and
but
Saturday
night
is
Ed
Sheeran
and
I'm
not
going
to
that
one
and
Shaquille
O'neal
on
Sunday
I
won't
be
there
either.
B
A
B
Well,
my
Beauties,
why
don't
we
get
this
show
on
the
road
15
minute
Max
on
that
Spec's
egg?
There
man,
let.
A
B
B
I
do
not
like
it
in
a
box
or
in
the
or
the
fox.
You
know
what
I'm
saying
so
I
do
not
like
slung
specs,
no
I'm,
just
I'm
just
messing
with
you.
Man
take
your
time.
Bro,
take
your
time.
C
All
right
so
first
off
otlp
version
1.0
good
to
see
that
people
are
still
talking
about
this
and
that
more
people
are
talking
about
it.
It
did
kind
of
come
up
a
little
bit
during
the
maintainers
meeting
yesterday.
C
Who
experienced
some
breakage
due
to
otlp
in
in
the
mobile
world,
and
it
was
just
kind
of
at
least
sharing
the
pain
that
mobile
users
like
never
upgrade
their
apps.
So
even
if,
like
you
have
done,
all
the
you
know
the
stuff
to
to
deal
with
otop
changes,
users
might
be
running
old,
crusty
stuff,
that's
broken,
but
I
can't
tell
you
how
old
my
Uber
app
is
on
my
phone.
C
I
think
the
one
the
one
sticking
point
it
seemed
is
that
this
I
think
the
spec
the
span
status,
there's
something
called
description
in
the
spec
and
it's
called
message
in
the
protobufs,
which
means
it's
called
message
in
OTL
pjson.
C
So
I
think
there
is
some
sort
of
situation
where,
like
the
spec
is
technically
stable
and
that
we
want
that
message
name
to
be
stable,
so
I
think
Tigger
is
just
gonna.
Like
add
a
note
saying
message
and
description
are
the
same
thing
for
span
status
and
allow
us
to
move
on
with
our
lives.
I
think
people
were
probably
okay
with
that.
C
C
Egrin
has
a
PR
to
clarify
that
scope,
so
these
scope
attributes
are
defined
at
build
time.
They're
not
run
time
things,
so
if
you
were
tempted
to
try
to
sneak
in
like
a
process
ID
or
a
threat
ID,
this
is
saying
that
that
would
not
be
allowed.
C
I
think
people
are
okay
with
this,
but
there
were
some
commentaries
of.
We
would
like
some
bag
of
attributes
for
details
such
as
that
still,
but
I
think
people
were
willing
to
leave
that
on
the
wish
list
for
now.
C
I
do
think
it
I
think
we're
running
out
of
places
to
stick
data,
though,
if
you
want
to
rethink.
E
B
Find
phrase
but
if
I
instantiate
the
same
Tracer
and
this
in
place
different
places
in
the
code
with
different
attributes,
you
know
that's
weird
for
me
to
do,
but
there's
nothing
stopping
people
from
doing
that
from
creating
a
name
Tracer
and
then
not
specifying
attributes,
because
they're
trying
to
look
it
up.
B
For
example,
instead
of
memorizing
the
result
and
like
the
Tracer
providers
keeping
track
of
the
existing
tracers,
so
you
have
to
specify
the
name,
the
version
and
the
instrumentation
scope
attributes
at
creation
time,
but
it's
also
the
same
lookup
interface
for
and
also
same
schema
right
because,
like
it's
also,
it
also
takes
a
schema.
Url
right
for
an
instrumentation
am
I
wrong
about
this.
E
So
there's
a
comment
here
from
tigrin
that
somebody
had
said:
do
you
don't
you
mean
application
initialization
time
and
he
responded
no
I'm
actually
going
explicitly
for
build
time
since
scope
is
defined
as
a
logical
unit
of
application
code
like
what
what
are
attributes
about
this
code,
not
what
our
attributes
about
its
runtime
State,
even
even
at
an
initialization
I,
can
I
can
understand
that.
E
A
unit
of
code
is
fully
defined
at
build
time.
Application
initialization
time
would
be
different,
would
imply
something
like
process
ID
or
thread
ID,
as
a
valid
scope
attribute,
believe
it
is
not
and
want
to
prohibit
it.
A
E
A
E
Dynamic
languages
have
have
a
entered,
the
room
enter
the
chat
well.
B
A
A
E
Well,
like
you,
wouldn't
stand
up
a
different
Tracer
per
database
connection
right
yeah,
so
you
would
not
we're
agreeing
you
wouldn't
you
would
not
set
up
a
separate
Tracer
per
connection.
You'd
set
up
a
tracer
for
postgres,
and
then
you
sh.
The
instrument
did
not.
A
B
B
E
E
A
A
E
B
B
You
know
whatever
yeah
scope
of
that
Tracer
provider
or
that
context,
and
it's
like
well,
we
don't
want
people
just
throwing
stuff
everywhere,
because.
E
B
And
I
keep
getting
myself
like
I
I
need
to
change
my
vocabulary.
It's
about
the
instrumentation
and
not
the
name,
the
Tracer,
which
is
what
I
keep
calling
it
I.
C
C
C
B
B
I,
don't
think
so
either
I
think
I
think
the
context
only
keeps
track
of
the
current
span
context
and
the
context
is
threat
safe,
local,
sorry
in.
E
A
C
Maybe
it
only
makes
sense
out
of
spam.
All
right,
I
think
read
this
think
about
it,
and
if
you,
if
you
do
have
issues
you.
C
E
C
What's
next,
since
we
have
a
substitute
Sam
in
the
room,
we
maybe
should
move
on
to
the
this.
One
actually
affects
us,
wait!
No
not
yet.
This
one
has
been
re
rekindled,
support
maps
and
heterogeneous
arrays
as
attribute
values.
So.
C
Yeah
I
think
we
brought
this
up
like
a
little
bit
last
week
that
maybe
there
would
be
some
interests
in
rekindling
this
discussion
right
now.
This
is
the
state
of
the
world
in
logs,
but
the
spec
forbids
attribute
values
from
being
nested,
Maps
or
arrays
thereof-
I
at
least
kind
of
commented
on
this,
because
I
I
always
get
concerned
about
this
situation
here,
where
you
have
yeah
some
key,
pointing
at
an
array
of
maps
and
like
flattening
this
stuff
ends
up
becoming
gross.
C
If
you
need
to
for
your
back
end
because
in
order
to
keep
some
kind
of
uniqueness,
you
end
up
having
to
introduce
like
an
index
in
there
and
when
you
do
that
all
of
these
things
become
effectively
unqueryable
in
the
end
or
difficult
to
query
in
the
end,
so
I
just
wanted
to
like
really.
That
was
my
only
question.
These
other
questions
were
just
like
well,
do
we
really
need
them
and
then
I
guess
the
other
question
is:
does
anyone
actually
support
this
stuff?
But
it's
just
a
draft,
but.
C
But
yeah
those
those
are
my
questions
as
we
start
to
consider
this
stuff
moving
forward:
I'm,
not
I'm,
not
taking
a
side
yet
I'm,
just
trying
to
understand
what
this
would
mean.
C
So
next
one
I
feel
is
the
most
important
one
actually
as
it
applies
directly
to
PR
and
R
repo,
and
this
is
basically
the
author
of
this
PR
has
added
a
public
endpoint
option
to
rack
middleware,
and
the
problem
is
happening
is
that
there
is
a
request
coming
into
the
system
with
a
transparent
that
is
from
some
somebody
else's
system.
Ultimately,
so
right
now
the
rack
middleware
will
use.
C
We
use
the
parent
ID
from
transparent
and
in
the
tracing
back
end,
this
Trace
will
be
orphaned.
There
will
be
kind
of
no
parent
for
it,
because
the
traits
did
not
originate
in
this
system.
C
So
yeah.
E
E
E
A
E
I
feel
like
there's
a
there's,
a
choke
point
at
like
past
this
outside
this
boundary.
You
know
that
it's
not
you
talking
to
yourself
or
you
can
identify
you
talking
to
yourself
and
have
rules
at
the
boundary
to
strip
out
headers.
A
E
C
Yeah,
so
I
should
probably
at
least
follow
up
with
feedback
from
what
happened
at
the
spec
set
here.
My
my
thought
was
that
this
is
a
Trace
State
issue
and
I,
don't
know
once
upon
a
time.
I
did
participate
in
the
w3c
distributed
tracing
working
group,
but
it
was
free
hotel
and
it
was
really
coming
from
the
perspective
of
like
a
tracing
vendor
and
wanting
stuff
wanting
to
be
able
to
interrupt
between
vendors
and
really.
C
The
kind
of
idea
is
that
each
vendor
would
stick
a
key
in
trade
State
and
you
could
even
kind
of
have
this
notion
of
like
vendor
key
plus
tenant
ID,
and
through
all
that
information
you
could
figure
out.
You
would
basically
snapshot
some
kind
of
parent
ID
the
last
time
this
was
in
your
system
in
tray
State,
and
then
you
could
compare
these
things
and
Stitch
them
up.
C
So
that
was
kind
of
the
way
that
I
thought
this
stuff
should
probably
work
in
otel
I
did
bring
that
up
in
the
specsig
meeting
and
I
think
that
I
think
that
I
don't
think
there
was
a
consensus
that
Trey
state
that
everybody
should
be
using
trade
state
for
this
in
in
hotel
and
I
do
think
that
that
would
it
would
kind
of
make
Trace
data
a
little
bit
more
of
a
wild
west
than
I.
C
Think
even
the
distributed
tracing
working
group
had
in
mind,
but
they
did
agree
that
this
is
a
problem
for
all
hotel
sdks
and
that
we
need
a
way
to
kind
of
be
able
to
define
the
the
trust
boundaries
more
or
less,
and
that
the
at
least
the
TC
would
talk
about
this
tomorrow.
And
this
would
become
like
something
on
the
spec
Sig
agenda
to
come
up
with
a
solution
that
we
can
apply
in
all
sdks.
So
since.
E
E
According
to
the
propagation
format,
that's
been
selected
and
it
there's
been
like
I
wish
that
the
propagators
would
allow
me
to
do
outgoing
only
so
that
on
a
boundary,
endpoint
I
could
say
propagate
w3c
or
Trace,
State
and
baggage,
but
only
outgoing,
like
don't
accept
any
incoming
which
the
the
propagator
spec
and
the
carrier
and
all
that
those
layers
in
there
don't
convey
enough
information
between
them
to
make
those
decisions
or
have
any
configuration
options
on
them.
E
B
E
B
B
I
don't
know
well,
the
the
sampling
pack
is
just
true
or
false
right
now,
yeah,
whatever
I
didn't
write
that
spot.
C
And
I
think
ultimately
yeah.
That
was
another
thing
that
vendors
who
are
using
tray
state
would
kind
of
they
would
probably
manage
their
own
things
through
chase
state.
You
would
take
a
lot
of
transparent
as
as
hints,
and
he
would
propagate
it.
As
you
know,
the
rules
mandate
that
you
propagated,
so
you
would
kind
of
propagate
it
as
it
was
received
more
or
less,
but
you
would
kind
of
make
overrides
for
yourself
by
a
trade
State
be
like
oh
I
received
this
header
that
said
sample
true,
but
I.
E
I,
don't
know
that
I
I
think
I,
don't
understand
how
that's
a
vendor
concern
because
propagating
the
header
and
what
to
do
with
its
sort
of
the
like.
It's
like
an
Hotel
thing
like
the
the
instrumenting
user
who's,
not
a
vendor
like
it's,
our
customer
who's
using
otel
and
saying
I
would
like
traces
from
my
system,
but
I
don't
want
my
traces.
Looking
like
they're
part
of
my
customers
requests
coming
to
me.
C
It
was
meant
to
kind
of
interrupt
between
vendors,
I
guess
and
for
or
many
users
like
a
lot
of
companies
will
end
up
using
a
lot
of
different
vendors
tracing
Solutions
they're,
usually
not
only
lifestyle,
customers
that
might
be
lights
up
and
honeycomb
and
datadog
customers
ultimately
kind
of
through,
like
Acquisitions
and
just
whatever
else
like
everybody
has
their
own
different
observability
tools.
So
you
do
you
don't
always
want
to
just
like
discard
stuff,
because
you
might
actually
have
the
tracing
data.
E
E
I
see,
we
don't
have
a
way
to
say,
like
I,
only
want
to
add
this
header
to
outgoing
HTTP
requests,
if
it's
to
one
of
my
hosts,
but
if
I'm
making
a
client
request
to
some
third
party
I,
don't
want
to
tell
them
about
my
trace,
parentage
or
state
yeah,
which
is
what
I'm
like
propagators
need,
some
way
to
give
rules
or
allow
block
lists
of
regexes
of
URLs
or
something
so
that
you
can
say
I'm
only
going
to
propagate
to
hosts
that
are
meet.
E
My
internal,
like
regardless
of
instrumentation
propagations,
like
all
of
the
instrumentations
punt
to
a
propagator,
to
say,
like
hey,
tell
me
what
I'm
supposed
to
do.
Propagator
and
the
propagator
goes
shovels
a
bunch
of
Trace
State
stuff
on
Trace,
parent
and
Trace
State
stuff
on,
and
if
the
propagator
could
be
taught
more
about
the
the
source
of
the
stuff
that's
being
propagated
in
the
destination
of
the
stuff
being
propagated,
it
could
be
given
some
rules
about.
Are
you
really
propagating.
C
C
It
definitely
exists
within
like
Cloud
vendors
themselves,
who
are
using
hotels
like
they
don't
want
to
leak
their
the
information
about
their
internal
systems,
the
tracing
of
their
internal
systems
to
customers.
So
you
do
want
to
have
some
way
to
specify
this
so
yeah
tldr.
C
It
seems
like
this
is
a
somewhat
complex
issue,
but
it's
definitely
one
that
the
the
specsig
immediately
recognize
that
you
know
this
needs
some
TC
attention,
and
this
is
something
that
needs
to
be
solved
in
the
specs
like
they
were
talking
about
it
being
solved
in
the
instrumentation
Sig
momentarily,
but
that
was
that
feels
that.
E
Was
declared
to
be
a
bad
idea,
pretty
yeah
really
fixing
it
and
then
all
the
instrumentation
feels
bad.
When
we,
when
the
thing
this
is
about
Trace,
State
and
Trace
parentage,
that's
propagation
concern.
It's
a
maybe
we
talk
about
how
propagator
spec
could
be
given
some
sort
of
configurations
consistently.
So
the
instrumentations
just
go:
hey
propagator,
here's
what
I've
got
and
the
propagator
does
the
needful.
D
Yeah
just
I've
been
out
a
couple
of
weeks
is
the
idea
that
we're
gonna
just
keep
this
for,
for
our
Sig
we're
going
to
keep
this
PR
on
Ice
until
there's
some
Upstream
resolution
on
what
like
the
best
practices
or
are
we
taking
it
or
rejecting
it
or
is
it
I.
E
D
Yeah
I
mean
you
know:
I'm
not
going
to
not
a
hill
I'm
gonna
die
on
I
think
we're
any
help
for
that
matter.
I
don't
want
to
die.
E
Yeah
well,
it
initially
seems
friendlier
to
like
accept
this
as
a
config
option.
Yeah
it
solves
your
problem.
We
could
accept
it
as
a
config
option,
but
then,
if
six
months
from
now
we
solve
at
the
config
level
and
you're
really
supposed
to
be
turning
this
off
in
the
propagators.
We
have
a
feature
that
we
need
to
deprecate
and.
B
Yeah
I
think
in
the
I
think
for
situations
like
this
is
that
we
say
no
and
please
Fork.
D
Have
the
feature
I
don't
know
if
it's
gonna
work
but
I
think
this
user
works
at
like
a
like
autho
or
something
like
I,
don't
know
if
it's
as
simple
as
that
as.
D
I,
don't
know
what
their
actual
route
like
problem
is.
I
just
was
looking
at
their
background
anyway,
yeah
I,
don't
you
know,
look
it's
not
like.
We've
already
decided
to
just
get
some
other
sdks
doing
it
doesn't
mean
we
do
it
so.
B
I
ran
into
this
problem
myself,
where
I
deployed
to
a
cloud
provider
the
cloud
provider
was
sending
in
a
trace
parent
that
had
sampling
turned
off,
which
ended
up
shutting
my
traces
off
since
I
was
defaulting
to
parent-based
and
then
I
had
to
switch
to
always
on
and
then
I
opened.
The
bug
report
and
I
was
like
hey:
what's
going
on,
with
the
sample
flag
being
shut
off,
ruining
my
traces?
So
yes,
this
happened
to
me.
One
time
do
I
think
that
we
need
to
roll
this
into
the
riotic,
instrumentation
or
I.
B
Think
people
should
probably
fourth.
B
Because
we
could
document
the
gotcha,
but
it's
more
like
this-
is
a
problem
with
propagation
in
general,
yeah
and
the
thing
about
it
is
that
if
we're
offering
Alternatives
I,
don't
want
to
prescribe
to
people
and
be
like
because
then
we're
going
to
be
like
how
do
you?
The
next
question
is
going
to
be,
how
do
I
configure
AJ
proxy
to
do
this
or
it's
the
O2
lag
strip,
headers
or
whatever?
And
it's
like.
C
C
Hey
discussions
aren't
going
to
move
forward
on
at
the
spec
level
and
I
think
you
know.
The
public
endpoint
option
is
one
option
that
should
be
on
the
table
and
I'm
sure
it
will
get
yeah
I
think
there
are
elements
of
this
that
are
sound,
that
people
will
like,
but
I'm
sure
it
will
get.
E
It's
certainly
something
to
react
to,
because
if
we
propose
is
like
all
right,
we're
gonna
spec
specify
that
all
instrumentations
that
do
propagation
need
to
have
this
configuration
interface
so
that
you
can
fiddle
with
what
you're
propagating
I
predict
that
the
reaction
will
be
well.
Why
don't
we
do
that
in
the
propagators,
but
it's
at
least
something
to
react
to
and.
C
It's
a
starting
point,
I
think,
yeah
and
I
think
the
the
solution
will
evolve
as
they
do
and
when
you
get
enough
people
at
the
spec
level,
but
I'm
certain
that
the
problem
will
be
solved.
It
will
probably
not
look
exactly
like
this
in
the
end,
so
I
think
accepting
this
only.
B
A
D
C
Yeah
good
question
I
feel,
like
we've,
had
some
good
conversation
on
this
so
far,
do
we
have
more
to
say
for
now
I'm
sure
we
will
be
talking
about
this
for
weeks
to
come.
So
if
you,
if
you.
E
Think
about
something
this
is.
This
is
my
old
baggage
problem,
where
I
was
complaining
about
propagating
baggage
to
people
like
if
I'm
throwing
user
IDs
in
my
baggage,
so
that
I
can
trace.
It
include
a
user
ID
and
all
my
stuff
and
I
don't
have
a
way
to
turn
that
off.
An
outgoing
HTTP
calls
to
third
parties,
so
I'm,
like
anyway,
I.
E
It's
not
even
malicious.
It's
just
I'm,
if
I'm
putting
interesting
data
in
my
baggage,
so
that
in
my
world,
I
use
a
span
processor
to
duplicate
attributes
that
are
in
baggage
onto
my
span,
so
that
all
my
spans
are
wide
and
informative.
But
now
that's
a
header.
That's
going
on
all
my
HTTP
calls
to
my
systems
and
to
third-party
vendors
and.
A
E
A
C
Yeah
I
do
think
it
dovetails
with
this
one,
so
it
would
make
sense
to
try
to
piggyback
some
of
these
concerns
and
proposals
on
on
these
ideas,
as
they
start
to
evolve.
C
All
right,
it's
great
that
Sam's
at
a
conference.
Otherwise
we
would
be.
We
would
be
in
deep
trouble.
C
C
There's
really
no
way
to
convey
type
information
in
the
in
that
environment.
Variable
ultimately
is
kind
of
what
it
boils
down
to
I.
Don't
know,
there's
some
discussion
about.
Should
we
Define
distro?
C
What
is
auto
instrumentation
and
there
was
like
I,
don't
know,
talks
about
like
things
could
even
be
trademarked
like
if
they
wanted
to.
If
they
wanted
to
Define
these
things
like
a
distro
for
to
actually
be
an
open
television
treat
this
row
it
might
have
to
like
reach.
Some
kind
of
you
know
certain
level
of
of
something
due
to
trademarking.
Otherwise,
it's
I
don't
know
underground
underground
hotel,
Library.
C
C
Vote
for
the
for
the
hotel
governance
committee.
C
There
will
probably
be
some
emails
that
go
out
with
your
links
to
vote.
Here
are
your
options,
so
many
great
options
but
I
think
you
get
multiple
votes
so.
A
B
D
It
can
be
issues
in
the
past
year.
D
Certainly
have
you've
done,
I
would
bet
my
life
well,
okay,
I,
gotta,
stop.
B
A
A
B
A
B
A
lot
of
those-
and
they
can
they
can
help
kind
of
navigate
around
those
problems
or
submit
improvements
right.
B
Yeah
instrumentation
maintenance.
So
that's
my
next
thing.
So
last
release
was
like
super
tough
things
were
breaking
issues
with
bundler,
getting
upgraded
and
something
breaking
on
Windows,
which
prevented
releases
from
going
out
and
so
on,
and
so
on
and
so
forth.
You
know
these.
B
We
get
into
these
situations
where
it's
like.
Oh
there's,
the
zigzag
faults
for
like
Artie
Kafka
that
keep
happening
all
the
time.
The
Matrix
builds
were
very
con.
They're
they're
very
complex
because
it's
like
these
gems
are
only
compatible
on
this
platform
for
this
operating
system,
and
it's
just
it
just
feels
unwieldy
to
me.
B
So
one
thing
I
wanted
to
ask
about
is
how
how
much
of
how
how
dedicated
how
dedicated
or
how
committed
are
we
to
supporting
the
full
Matrix
as
it
is
so
supporting
Windows
releases,
supporting
truffle
and
supporting
jrubing
in
general
for
our
instrumentations?
B
That's
question
number
one
number
two,
as
we've
gotten
a
couple
of
new
submissions,
which
has
been
great,
but
we
haven't
done
anything
about
time,
ownership
for
long-term
maintenance
of
those
submissions.
So,
for
example,
you
know
the
race
car
submission
The,
Rake
instrumentation
we
recently
pulled
in,
and
those
are
great
contributions.
However,
you
know,
if
you
look
at
say,
like
the
go:
contribute
repository
they're
like
hey,
you
know
we
need
to
push
towards
first
party.
They
have
like
a
guide
on
there
and
they're
contributing
God.
You
really
want.
B
We,
they
really
want
to
push
for
first
party
instrumentation
whenever
possible.
Second,
is
create
your
own
Standalone
repo,
and
you
be
the
one
to
maintain
this
instrumentation,
because
we
don't
have
the
people
power
to
maintain
everything.
B
I
mean
I.
Think
the
same
thing
applies
to
us
and
I.
Think
very
uniquely.
We
have
like
this
concept
of
the
all
gem
where
it's
like
it
references
all
the
instrumentations
that
are
in
this
repository,
but
is
that
the
way
is
that
what
we
want
to
continue
to
do
and
I'm
I'm
starting
to
personally
feel
like
I,
think
we
might
want
to
start
considering
saying
no,
no
saying
no
and
keeping
these
instrumentations
like
following
the
same
policy
or
the
same
guidance
of
the
the
Goku
trip
package
is
doing
right
now.
B
A
B
B
D
I
mean
I've
one
is
that
I
think
there's
bet
yeah
what
go
is
doing,
introduces
trade-offs.
Adoption
is
harder
that
way.
Maintenance
becomes
good
I.
You
know,
I
think
you're
trading,
some
short
term
you're,
just
trading
pain
to
okay.
D
Now
the
user
is
using
some
random
thing
on
some
other
repo
and
it's
not
working
and
you're
still
having
to
you
know
you
start
getting
bugged
about
it,
and
maybe
you
can
say,
go
pick
it
up
with
them,
but
you
know
I,
don't
know,
or
you
have
to
find
some
way
to
when
you
do.
A
breaking
change
is
notify
those
repos.
So
you
know
I,
don't
think
it's
like
a
free
launch
or
anything,
although
it
does,
in
the
immediate
term,
reduce
the
maintenance
burden
of
the
instrumentation
team
yeah.
D
The
other
thing
is
just
like
I,
don't
practically
speaking
like
I
would
I
personally,
don't
have
the
I
basically
have
zero
time
to
work
on
this
right
now,
I'm
on
a
different
team
at
work
and
we're
just
really
busy.
D
B
I
think
the
challenge
with
it
is
the
domain
expertise
for
for
one
instrumentation
versus
another,
like,
for
example,
I'm
totally
fine,
like
I'll
review,
trilogy
instrumentation
commits
but
like
I,
didn't
write,
for
example
the
rack
one,
and
when
we
start
seeing
issues
come
in
about
changing
things
on
rack
I'm,
like
I'm,
deferring
to
the
person
who
originally
submitted
it
because
I'm
like
there's
some
thought
process.
Some
consequence.
D
Know
reason
why
they
made
those
choices
right,
yeah,
no,
no
for
sure
I
mean
I.
Originally,
when
I
was
pushing
for
like
the
instrumentation
can
like
the
contrib.
Repo
is
a
separate
thing.
One
of
the
things
I
mentioned
is
like
we
have.
You
know
ownership.
We
have
some
sort
of
like
finer,
grained
grain
granularity
am
I,
saying
the
same
thing
twice
of
ownership,
so
it's
like
I
mean
you
know,
there's
a
separate
file
within
like
an
ownership
file,
the
same
way
collector
contrib.
Does
it
where
it's
like
cool?
A
D
D
The
release
problem,
but
you.
E
Know
it
doesn't
solve
the
release
problem,
but
if,
if
we
could
instead
of
saying
no
to
accepting
things,
be
a
little
like
well
to
accept
this,
we
kind
of
need
some
help,
maintaining
it,
because
none
of
us
are
I'm,
not
a
Kafka
expert.
So
if
we've
got
Kafka,
instrumentation
I'm
like
I
I,
don't
I'm
not
really
qualified
to
review
this.
Who
can
we
like
in
in
submitting
this
you've,
demonstrated
that
you
have
probably
more
knowledge
than
I
do
and
we
could
see
it?
E
B
D
The
collector
has
a
process
around
this
where
they
say
like
cool,
we'll
take
it.
You
know
we'll
take
your
Alibaba
Cloud
exporter,
you
employee
from
Alibaba
like
no
one
knows
the
Alibaba
back
end
that
within
the
close,
so
it's
like
cool
you're,
you're,
the
code
owner
and
it
when
issues
come
up
or
bug.
Bug
reports
come
up.
You
have
X
amount
of
time
to
indicate
that,
like
you're,
taking
a
look
and
if
they're
and
if
you
have
some
amount
of
like
truancy
in
that
regard,
we
reserve
the
right
to
basically
deprecate
that
that.
A
D
B
I
got
all
right,
so
I
do
I.
Do
like
that.
As
a
starting
point
is
to
say
like
your,
by
submitting
this
you're
committing
to
it
right
and
if
they're
not
willing
to
commit
to
it,
then
they
wouldn't
submit
it
at
all.
So
that's
one
part
of
it.
That
sounds
good.
The
we
did
say
that
we
were
going
to
have
like
an
instrumentation
and
developers
guide
at
some
point
that
we
never
wrote,
and
that
should
be
part
of
that.
B
So
we
can
change
the
pr
template
to
say,
like
you
know,
whatever
new
instrumentation,
NPR
template
or
something
like
that
and
say
you
know
what
what
the
expectations
are
for.
Folks,
add
that
to
the
contributing
guide
for
instrumentations
and
have
that
expectation,
I
think
I'm
that
to
me
seems
reasonable
as
a
reasonable
start,
as
opposed
to
going
the
complete
opposite
direction.
Although
I
still
feel
very
strongly
that
we
should
lean
towards
first
party
instrumentation
whenever
possible,
outside
of
our
repository
when.
E
B
If
Rock
had
instrumented
itself
and
and
also
you
know
as
third
party,
we
are
depending
on
monkey
patches
in
a
lot
of
cases,
because
they
don't
have
first
class
support
for
tracing.
So
that's
what
I
mean
by
first
party,
okay,.
D
D
E
A
B
For
the
reason
that
Rick
and
the
race
car
ones,
I
will
I'll
follow
up
with
them.
I
think
that
aspect
though,
for
example,
they
used
to
participate
a
lot,
but
they
don't
have
anybody
else
kind
of
like
committing
to
the
AWS
SDK
right,
but
I'll
reach
out
to
them
as
well.
For
that
I
know,
I'm
gonna
put
Shopify.
C
E
The
leaders
and
the
pioneers
of
figuring
out
the
human
problems
of
it
yeah
yeah.
C
So,
like
I,
think
yeah
I
think
we're
reaching
this
point
where
we're
starting
to
experience
some
of
the
pain
that
they
had.
So
we
can
start
to
adopt
some
of
those
processes
and
I
I,
believe
everything
is
based
on
code
owners
and
I.
Think
yeah,
the
based
on
that
file,
like
people,
will
be
flagged
to
kind
of
review
things
over
others,
anyways
I
believe
there's
some
kind
of
hierarchy
that
can
happen
there
and
that
would
be
nice
for
for
the
owners
to
be
prioritized
over.
You
know
I.
E
Think
I
think
it'll
show
up
as
as
yes
never
mind
carry.
C
C
You
can
kind
of
like
re
readmit,
but.
E
I'd
say
if
it,
if
it
if
it
gets
into
the
unmaintained
and
starts
causing
problem
in
CI,
we're
like
well
it
we,
the
people
who
signed
up
to
maintain
that
thing
aren't
showing
up.
So
it's
coming
out
of
the
Matrix,
because.
C
E
Of
Open
Source,
that's
how
it
works
right,
that's
how
it
works
that
works
yeah,
so
we've
said
stuff
about
I.
Think
that's
a
good
plan.
We
would
like.
B
E
We
should
pull
it
in
simulcast
picture
in
picture.
A
D
D
Saltz
or
whatever
so
that'll
be
fun
complain
about
Shopify
in
person.
E
Truffle
Ruby
jruby
windows
simply.
E
I
have
no
idea.
Oh,
let
me
see
if
I
can
see
if
those
gems
are
downloaded
for
what
platforms.
D
I
have
no
idea,
I've
talked
my
head,
maybe
Robert
does
or
I
can
ping
Francis
at
all
and
see
Jr
Ruby's,
pretty
popular
there's
some
pretty
large
tracing
users
who
use
jruby.
If
I
recall
correctly
for
my
vendertips,
you
know
it
would
be
ashamed
to
drop
it.
If
there's
not
a
if
the
issue
is
just
like.
If
the
issue
is
this
thing,
is
you
know
killing
us
every
time,
then
sure?
But
if
it's
just
like
occasionally
it's
flaky,
then.
B
B
So
if
there's,
if
there's
an
alternative
that
we
can
offer
so,
for
example,
like
a
lot
of
the
instrumentations,
don't
even
aren't
supported
by
some
of
these
platforms
either
right
like
oftentimes
like
we
don't
even
have,
for
example,
like
any
of
the
tests
running
on
the
Windows
platform
for
things
that
are
that
provide
that
require,
like
native
gem,
support,
whatever
like
we
don't,
have
full
support
or
compatibility,
and
we
can
even
tell
people
what's
actually
compatible
without
looking
at
the
Matrix
and
trying
to
like
puzzle
through
what
the
Matrix
is
testing.
B
D
Suggestion
which
I've
seen
on
totally
is
we
can
Mark
I,
don't
know
what
the
support
for
these
are
marked
as
or
what
are
flexibility
this
year
is,
but
we
can
take
them
out
of
like
the
core
CI
run.
Have
them
run,
maybe
like
once
in
a
non.
You
know,
non-blocking
overnight,
run
and
mark
them
as
experimental
or
something
like
that
where
it's
like
cool
we're,
gonna
run
CI
on
it.
D
It's
if
it,
if
it
blows
up
like
you,
can
go
check
the
canary
round
and
see
if
it
blew
up
but
like
we're,
not
guaran,
you
know
we're
not
claiming
it's
GA
or
whatever.
The
term
is,
and
you
know
in
practice
that
might
be
good
enough
for
people
and
then
the
P
you
know
and
if
it
becomes
really
problematic
and
is
truly
not
like
truly
not
working.
That
ought
to
just
sort
of
gets
surfaced
to
us
by.
You
know
if
it's.
B
Specifically
I'm
talking
about
like
recent
incident
with
bundler,
where
I'm
like
spinning
my
wheels
for
bundler
on
Windows,
and
it's
like
what.
Why
is
it
stopping
me
from
releasing
stuff
right
and
why
do
I
have
to
pin
every
version
of
this
to
bundler
to
some
version
of
bundler,
so
I
can
satisfy
this
bug,
that's
specific
to
the
windows
platform
that
is
like
not.
That
was
not
a
fun
experience
for
me.
D
I
get
it
and
like
realistically,
like,
if
we're
being
honest
like
we
probably
are
treating
this
with
experimental
like
I,
don't
think
we
can
really
confidently
say
we
have
like.
You
know
full
GA
support
for
this
stuff
because,
like
as
far
as
I
know,
none
of
us
are
really
using
it,
and
so
anyway
I
don't.
That
would
be
my
recommendation
still
some
work.
It's
not
zero
work
to,
but
just
take
it.
You
know,
but
supposedly.
D
Yeah
doing.
E
Product
placement,
here's
rubygems
downloads
for
the
last
24
hours
and
we
can
scope
down
the
query
to
to
download
it
right
now.
It's
counting
how
many
gems
were
downloaded
where
the
name
contains
open,
telemetry
grouping
it
by
user
agent
and
by
geocity,
which
will
give
generally.
If
you
see
Ashburn,
show
up.
Okay,
that's
U.S
east
one,
AWS
Us
East,
one
I,
don't
know
where
is
yours
running,
but
if
I
were
to
say
this
is.
A
Eleven,
that's
like
user
agent
name
starts
with.
A
D
E
Somebody
else
using
this
Zoom
are
we
still
in
double
booked,
zoomsland
I.
B
A
I'm
here
for
the
not
vocabulary,
yeah
I
mean
I'm
working
with
tigram
and
I
was
you
know
following
PR?
You
know
it's
issue
by
tigran
from
Splunk
and
I
was
hoping
to
you
know,
but
probably
I
just
landed
in
the
wrong
seat.
Oh
okay,.
D
E
B
B
B
B
Sure
but
I
mean
like,
but
we're
looking
for
Windows
right,
yep.
E
A
A
D
All
right,
I
got
you.
A
B
A
B
B
B
B
I'm,
what
my
feeling
is
Rob
is
I'm,
seeing
John
J,
Ruby
J
Ruby's,
fine,
we
gotta
keep
Jay
Ruby
yeah.
A
B
Truffle
where's
Truffle
in
this
list.
How
does
that
identify
itself.
B
B
B
Right
all
right
sounds
good
to
me,
but
I'm
not
gonna,
do
any
of
this
stuff.
This
week
now
you
have
gotten
nerd
sniped
right,
a.
E
A
And
we.
A
A
E
E
Which
is
like
if
this
is
us
well
Point,
ID,
all
right,
so
this
is
when
I
Googled
it
it's
in
Virginia.
It's.
B
B
A
E
Do
you
know
if
GitHub
actions
fan
out
over
multiple
regions
to
like
spin
up
Runners
and
do
things
and
then
I
don't
know
the
answer
to
that.