►
From YouTube: 2023-03-14 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
B
C
My
my
wife's
boss
is
a
guy
named
Steve
Cohen,
so
sometimes
sometimes
I
get
missed
tickets.
C
B
C
Team
but
yeah,
you
know
we
bought
a
hat
once
that
counts
around
with
him.
B
Yeah,
it's
and
then
you
know,
I
even
had
signed
up
for
MLB,
because
all
the
games
are
blacked
out
here.
B
Right
yeah,
so
I
paid
for
you
know,
I
got
the
the
whole
season
and
you
know,
and
it
was
just
like
it
was
really
great
and
then
and
then
it
was
just
heartbreaking
towards
the
end
and
then
the
losing
the
first
round
of
playoffs
San
Diego's,
just
like
I
mean
just
like
my
heart's
like
and
then
it's
like.
Oh
you
know,
see
you
later
Jacob
the
ground
there
I
was
like
okay.
B
Youtube
they'll
broadcast
games
for
the
Rangers
and
the
Astros
yeah
watch.
B
C
B
It's
like
yeah,
but
it's
like
you
know,
everything's
about
hookup.
You
know
what
I'm
saying
so
yeah.
A
C
I
don't
know
I'm
like
what's
up
Robert
I've
like
checked
out
a
I
just
sit
down
at
the
time.
These
days,
I
used
to
be
like
deep
into
so
many
sports
things.
College
High,
School
I
was
like
David
to
high
school
basketball.
I
was
obsessed
with
high
school
basketball
and,
like
isn't
fairly,
that's
like
really
big.
Everyone's
a
high
school
basketball
fan.
It's
like
all
these,
like
Catholic
schools,
are
real,
big
and
stuff,
but
I
just
like
looked
up
one
day
and
it
had
been
like
three
years
and
I
haven't
really
watched
the
game.
C
A
Good
yeah.
C
B
All
at
once,
we
were
just
doing
small
talk
here:
yeah.
C
D
D
What's
that
a
pull
request
to
the
open,
Telemetry
specification,
repo,
that
is
clarification
regarding
parent
context
passed
in
the
tracing
SDK
I,
have
no
further
information,
because
I'm
just
reading
notes
that
are
I.
D
Let's
see
here
resolves
an
open,
Otep
export
span.
Context
is
remote
proposal.
The
update
otlp,
to
indicate
whether
it's
fans,
parent
is
remote.
B
The
the
Proto
boss
doesn't
have
a
notion
of
whether
or
not
the
parents
ban
was
remote
or
local.
A
D
E
B
Or
the
or
the
no
I
guess
the
consumer
spans
too
right
like
I,
guess
anything.
B
D
But
if,
if
it's
written
in
the
context
that's
get
passed,
around
I
was
like
this
context
was
remote,
so
like
I
guess,
the
carrier
is
set
like
I'm
a
carrier
I'm
setting
this
context
I'm
a
carrier
which
means
it
was
remote.
So
flip
this
to
true
and
the
next
person
that
bring
the
to
generate
a
span
from
the
current
context
would
flip
it
to
false,
and
now
it's
false
until
it
leaves
the
process
and
then
now
you
know
which
span
is
your
service
band,
because
that's
the
one
that
was
the
boundary?
B
D
I,
don't
know
where
I'm
going
with
this
is
that
auto
instrumentation
tends
to
not
know
the
layers
that,
like
I'm
thinking,
database,
comms,
client
server
database,
where
the
client
talks
to
the
server,
but
over
HTTP.
So
like
your
HTTP
spans
like
the
line
protocol
and
there's
querying
happening
at
the
database,
abstraction
layer,
but
there's
like
an
interstitial
layer
of
HTTP
happening
so
which
ones
when
the
server
gets
it
is.
Is
it
the
incoming
HTTP
span
or
the
incoming
database
handling
span
which
one's
the
server
of
spaincock.
B
But
I
think
it's
again
more
about
the
it's
about
the
appearance,
band
contact,
another
child's
band
context
so
like
if
I
were
doing
context
propagation
and
when
I
am
parsing.
The
incoming
Trace
context
at
that
point,
I
know
that
this
was
remote.
B
D
Yes,
it's
this
I
get
whatever
can
determine
that
the
the
span
context.
It
comes
from
a
remote,
that's
not
from
within
your
process.
It's
it's
been
parsed
from
a
different
process.
It
was
remote.
That
also
is
a
statement.
I
think
semantically,
about
spankind
server
versus
client
is
that
MRI
am
I
wrong
about
that
they
are
semantically
similar
or
at
least
one
might
be
able
to
be
determined
from
the
other
I
think
an
incoming
context.
That's
your
parent
was
remote
means
that
I've
got
a
Spam
kind
server.
A
A
C
B
D
D
D
C
C
C
Keys
and
values
I
feel
like
double,
will
be
on
the
details
and
there
is
a
lot
of
like
May
and
should
and
type
language
that
will
make
it
difficult
to
really
yeah
there'll
be
a
lot
of
even
if
it's
defined
as
stable.
C
There
might
be
areas
where
you
know
it's
May
or
can
type
language
that
leaves
room
for
interpretation
or
just
yeah
room
for
different
implementations
according
to
resources
and
so
on
and
yeah.
It
would
be
if
something
comes
out
of
that
it'll
be
worth
paying
attention
to,
but
I
have
a
feeling
that
I'll
be
open
for
a
while.
The
schema
thing,
if
I
could
they
admit.
There's
another
note
on
there
about
schema
Hotel
schema
that
we've
brought
up
in
the
collector
stick
last
week.
C
I
think
Splunk
has
a
bit
of
a
renewed
effort
to
think
about
whether
to
sort
of
like
push
forward
the
concept
of
the
hotel
schema,
which
kind
of
like
died
a
little
bit
that,
like
you
know
it
was
a
semi-incomplete
tool.
It
couldn't
really
do
transforms
that
super
well
or
sort
of
like
lean
into
the
elastic
common
schema.
C
It
seems
like
there's
a
need
from
Splunk
for
their
users,
so
they're
looking
into
whether
they
should
invest
into
something
in
their
hotel
or
there's.
You
know
how
much
they
would
need
to
invest
versus
pushing
people
through
this
elastic
Thing
versus
maybe
just
diver.
You
know
providing
telling
people
hey
here's
a
here's,
a
bunch
of
proprietary
processors
and
stuff
go.
You
know
to
fit
all
things
into
the
format
they
want,
but
anyway
that
was
a
context
from
The
Collector
said
last
week
it
seems
like
they
will
be
pushing
that
forward.
C
So
if
we
have
experience
there,
I
think
they're
looking
for
input.
I
had
mentioned
briefly
that,
like
we
had
sort
of
like
trialed
something
Shopify
with
regard
to
schema
and
it
didn't,
you
know
the
ground
as
much
as
we'd
like
where
it
just
has
been
deep
prioritizing,
but
I
don't
have
much
more
to
add.
Besides
it
and
I,
don't
think
it's
even
I,
don't
think
we
even
have
this
info
nanode
right
now
in
rebate.
To
be
honest,
maybe
we
do.
C
Yeah
I'm
not
I,
think
it
might
not
be
totally
implemented
like
it's
in
the
it's
in
the
produce,
but
like
I,
don't
know
no.
C
Yeah
the
second
point
I
had
seen
with
that
was
like
when
there's
a
difference,
you
just
drop
this
the
URL
instead
of
like
having
some
ability
to
like
merge
your
superset,
which
I
thought
was
like
dumb,
but
anyway,
okay,
that's
all
that
was
my
only
input
there
and
I.
Think.
Let's
move
on
to
that
who
that
wraps
up
the
overtime
portion
brought
to
you
by
also
and.
D
B
B
D
A
B
Opened
an
issue
with
the
graphql
team
about
them,
taking
over
the
tracing
support
because
I
guess
they
do
Force
like
sort
of
a
third
party,
Library
Integrations,
they
maintain
them
in
their
repo,
because
you
know
we
ran
into
a
situation
where,
like
the
builds,
were
failing
because
of
a
bug
that
came
out
in
a
point
release
as
it
was
like
sorry,
a
bug
fix
release
which
included
a
new
re-imagining
of
the
tracing
libraries
or
like
tracing
API
in
graphql.
That
is
designed
to
be
super
optimized.
B
There
were
some
behavioral
changes
as
well
I'm
going
to
try
to
capture
in
this
PR
and
in
another
related
issue,
so
that
PR
that's
there
is
to
you
know,
I
I
sort
of
like
as
an
emergency
I
went
through
and
I
said.
Let's
lock,
you
know
graphql
to
a
Max
version
of
217
and
then
I
was
I'm
skipping
to
18
and
then
moving
on
to
219.
B
2019
since
28018
is
broken.
B
But
with
it
comes
a
couple
of
changes
and
it's
you
know
this
PR
I
have
here
is
still
using.
Like
the
let's
say,
it's
like
the
Legacy
API.
B
Not
using
the
newer
Trace
packages
that
are
included
in
there,
this
is
a
new
module
and
it
kind
of
characterizes
some
of
the
behavioral
differences.
So
it
used
to
be
the
case
that,
as
part
of
the
data
that
was
getting
passed
in
you,
it
would
give
you
derived
types
but
not
anymore,
so
it's
always
going
to
render
the
interfaces
and
the
derived
types
are
missing.
B
Now
there
are
some
things
you
could
do
like
add
extra
code
to
try
to
figure
out
the
derived
types,
but
I
didn't
bother
for
right
now,
I
just
kind
of
was
just
like
well.
This
is
what
it
actually
does
with
the
code.
That's
there
now
I,
don't
know
if
anybody
would
have
any
strong
objections
or
concerns,
but
we
should
probably
get
on
to
using
the
new
API.
B
D
Of
deprecating,
the
gem
that
we
maintain
and
helping
them
use
the
tracing
that
they
have
built
into
the
graphql
library
to
emit
open
Telemetry,
and
then
it's
first
party
and
we
might
be
an
assist.
What.
B
E
Some
from
our
group
pushed
up
some
first
party
instrumentation
a
while
back
and
we
had
them
yank
it
out
because
they
didn't
quite
understand
open,
Telemetry
or
the
spec,
because
there's
some
problems
in
it.
So
like
one
of
the
things
that
kind
of
was
a
a
very
concrete
realization
of
a
concern.
Was
that,
like
first
party
instrumentation,
there's
no
guarantee
that
they're
going
to
do
things.
E
The
way
we
would
expect
them
to
do
with
open,
Telemetry
I
think
that's
like
a
reasonable
thing
that
we
should
be
able
to
approach
and
interact
with
and
like
it
shouldn't
Block
doing
it.
E
The
part
that
I
I,
obviously
care
about
that
part
a
lot,
but
the
part
that
I'm
interested
or
concerned
or
whatever
I
think
is
like
something
I'd
like
to
have
an
answer
to
and
I,
don't
know
what
it
is:
I
like
the
auto
installation,
configurability
and
all
that
stuff.
That
comes
with
us.
Maintaining
this
a
third
party,
instrumentation
Library.
E
So
right
now
a
consumer
of
open,
Telemetry,
graphql
instrumentation.
If
they
say
it's
enabled,
it
will
add
the
the
Tracer
to
graphql
through
its
like
formal
way
of
doing
it.
If
we
push
it
directly
into
graphql
does
do
we
still
have
that
mechanism
and
then
the
second
part
is,
is
how
are
the
defaults
controlled
for
what
is
Trace
like,
for
example,
we
have
some
fields
that
are
off
by
default
for
tracing
them,
because
they're
very
verbose.
E
They
generate
a
lot
of
spad
volume,
which
is
okay
for
some
people,
but
again
like
what
is
our
mechanism
for
control
and
configuration
look
like
because,
as
it
stands,
we
haven't
figured
that
out.
It's
not
figured
out
in
like
the
specification
anywhere
I'm
sure,
there's
probably
some
examples
that
exist
in
other
language
implementations,
but
it's
like
that's
something.
We
really
need
to
figure
out
and
then
there's
like
the
whole
story
of
like
how
do
we
formalize
it?
E
E
Just
it
just
seems
like
a
lot
like.
Maybe
it's
not
a
good
reason
not
to
do
it,
but
it's
just
like
there's
a
lot,
there's
a
lot
behind
what
might
sometimes
feel
like
or
what
might
feel
like
a
kind
of
like
a
small
good
first
step.
But
there's
like
this
Kirby
dragons
kind
of
thing,
I'll
pause
for
a
moment
to
let
other
people
talk.
B
The
trade-off
is
that
we
don't
have
to
be
responsible
for
constantly
maintaining
updates
like
this.
One
of
the
challenges
we
have
is
that
we
don't
have
all
of
the
domain
experts
for
every
instrumentation,
that
is
in
the
repo
we
don't
have
maintainers
for
them.
So
if
we
can
get
the
Gem
authors
to
assume
their
responsibility,
that
would
be
a
huge
win
for
us,
I.
B
Think
and
if
we
again,
it's
like
you
know
sure,
I
I'm
sympathize
with
the
fact
that,
like
the
instrumentation
in
itself
can
change
over
time
but
I
think
that's
where
we
help
like
we.
B
You
know
we've
built
in
all
the
mechanisms
to
say
like
if
you
include
this
gem,
this
graphql
gem,
then
you'll
get
Auto
loaded
or
whatever
right
by
Butler
and
as
part
of
that
auto
load
process,
because
it
uses
the
base
instrumentation.
B
Gem
dependency,
it'll
register
itself
with
the
hotel
registry
right,
so
you
know
we're
I,
don't
think
we're
going
to
lose
that
functionality,
if
we
sort
of
if
we
contribute
the
the
code
that
we
have
to
this
repo
I
think
just
longer
term
is
kind
of
like,
as
as
changes
like
this
that
happen,
that
are
big
changes
to
the
API
and
then
all
sudden
things
break,
and
we
don't
know
about
it.
B
A
E
B
E
I
just
thought:
I'd
share
that
but
like
even
just
having
it
inherent
from
instrumentation
based,
was
like
an
interesting
point
because,
like
I
have
gone
back
and
forth
on
that
that
just
that
simple
decision,
I,
don't
think
it
is
a
simple
decision
like
should
their
first
parties
like
have
their
instrumentation
coupled
to
like
our
Base
Class,
and
we
could
say
yes,
we
can
say
no
again,
I,
don't
know
what
the
right
answer
there
is
and
I
think
I
mean
I.
E
B
I
would
love
to
see
other
for
us
to
pushes
for
other
German
authors
and
stuff
like
but,
like
you
said,
I
think
the
biggest
problem
is
the
domain
expertise
in
hotel
and
if
we
can
remove
the
barrier
and
the
domain
expertise,
part
of
it
where
it's
kind
of
like
you
know.
This
is
where
you
do.
The
instrumentation
like
this
is
how
you
can
leverage
it.
I
will
address
the
API
and
not
having
to
worry
about
the
SDK
and
all
this
other
stuff.
Then
right
that'll
be
win.
I.
Think.
E
Yeah
I
think
I
definitely
think
it's
worth
pushing
a
lot
and
I
think
this
is
a
good
like
you
said
this
is
like
a
good
first
place.
To
probably
do
it
because,
like
the
owner
of
graphql,
seems
super
responsive
to
this
sort
of
things,
which
is
great.
E
We
have
people
like
inside
of
Shopify,
who
are
very
interested
in
this.
That
would
like
I,
said
some
group
kind
of
like
randomly
tried
to
tackle
it
on
their
own,
outside
of
like
our
support,
which
had
like
some
issues.
But
the
point
is,
is
like:
there's
people
who
are
have
a
vested
interest
in
like
seeing.
A
E
E
C
Ahead,
I
was
gonna
say,
like
actionable
things
here
sounds
like
we
are
on,
like
the
patch
release,
the
point
zero
point,
one
nine
patch
release
will
sort
of
like
resolve
our
immediate
issue
where
gql
instrumentation,
like
doesn't
work.
So
that's
comforting.
We
may
want
to
I
think
you
have
a
PR
to
skip
like
to
explicitly
say
we're
not
compatible
if
it
happens
to
be
that
bad
release,
which
is
like
I,
think
a
nice
thing
to
do.
C
You
know
in
practice
like
it
sounds
like
not
a
lot.
It's
only
been
out
for
like
a
day,
so
now
many
people
don't
be
using.
So
like
it's
good
to
do
and
yeah
like
it.
You
know
it's
work.
If
we
want
to
Upstream
it
just
a
matter
of
like,
besides
the
you
know
like
whether
we
even
should
it's
going
to
work
and
even
then,
like
there's
maintenance
work,
there's
you
know
like
who
handles
the
bugs
who
handles
the
the
sport
questions.
C
That's
like
not
trivial
and
stuff
to
figure
out,
but
yeah
I
think
like
for
now.
Is
there
anything
actionable
you're,
basically
saying
the
actual
thing?
Is
you
want
to
leverage
the
new
tracing
API,
which
is
lower
overhead,
and
it's
a
question
of
when
we
leverage
that
are
we
modifying
our
own
instrumentation
or
do
we
just
say
like
this?
Is
a
good
cutover
point
to
do
something?
Upstream.
D
The
Proposal
I
think
is
aerial
to
add
to
the
graphql
tracing
here
an
open,
Telemetry,
flavored
version
of
these
pairs
of
libraries,
where,
let's
say
datadog,
okay,
make
a
class
of
graphql
tracing.
That's
open,
Telemetry,
trace
and
right
use.
Graphqls
tracing
hooks
to
generate
open
Telemetry
spans
is
that
the
thought.
C
B
C
C
B
D
A
C
Okay,
good,
it's
actually
free
dpfd.
Absolutely
all
this.
B
C
C
Like
thank
you
for
doing
the
work
here,
so
you
know
catch
this
stuff
and
like
protect
us
super
appreciated
and
too
like
yeah.
I'm.
Happy
to
you,
know,
I
think
it's
good
to
do
this
first
in
our
I.
Think
it's
positive
and
we'll
realize
the
games
quickly
and
I
tend
to
agree
with
Robert,
which
is
like
there's
a
lot
of
prb
dragons,
but
maybe
it's
worth
the
virtually
notice.
C
Yeah
a
lot
of
What's
the
phrase,
a
lot
of
motivated
people
on
on
Everest
sitting
in
the
president,
so
same
with
slang
dragons
anyway,.
C
D
Your
phrase
of
the
day
of
being
actionable
I
think
we
have
two
actionable
things:
get
the
patch
release
out
to
to
fix
the
immediate
thing.
There's
a
near
term.
We
could
overhaul
the
the
instrumentation
for
graphql
and
we'll
use
the
new
tracing
interface
and
then
we,
after
all,
that
we
could
consider
the
longer
Hall
I,
don't
know
if
we
have
other
thoughts
that
you
would
want
us
to
review.
Ariel
and
other
discussion
topics.
C
A
B
B
Is
the
is
the
pr
to
review
for
the
219
support
and
then
I
think
we've
already
looked
at
it
right
and
then
adding
tests
for
rec3.
D
Yep
needs
review
is
what
I'm
hearing
yeah.
A
D
B
D
C
A
D
E
D
C
My
bad
I
thought
I
had
approved
this
last
week.
I
guess
I
didn't
I,
it
seems
straightforward.
It's
just
appraisal
stuff,
basically
yeah.
A
C
I
haven't
looked
at
at
all
foreign.
C
Api,
whatever
it
is,
the
events
thing
do
you
have
anything
on
the
agenda
for
that
Ariel
Etc,
just
another
call
to
review.
C
The
elasticsearch
pr,
we
probably
should
add,
as
an
interview
I,
haven't
reviewed
it.
It
looks
like
someone
from
elastic
made
a
PR
to
include
elastic
surgery
instrumentation.
The
implementation
is
mostly
just
a
copy
paste
from
their
whatever
elastic
is
like
an
APM,
Tracer
repo
that
looks
shockingly
similar
to
all
the
others.
C
Some
of
the
implementation
is
a
little
scary.
Looking
with
its
like
obfuscation,
defaults,.
C
C
So
there's
probably
a
lot
of
moving
Parts
here
with
like
the
specifics
of
some
of
this
attribute
names
and
stuff,
which
is
why
she
left
in
draft,
but
the
implementation
just
like
yeah
yeah.
It
was
like
the
obfuscation
code,
looked
like
pretty
gnarly
to
me
and
I
kind
of
looked
at
it
and
was
like
ooh
like
this
is
not
something
I
can
just
like
totally
say,
is
good
to
go
and
then,
and
then
I
didn't
have
time
to
look
at.
It
was.
C
C
Yeah
structure
functionality
will
change
around
because
most
of
it
seems
fine.
It
seems
like
it's
a
you
know.
It's
a
client
span.
You
pick
up
some.
You
know
you
pick
up
some
things
about
the
client
but
yeah.
Some
of
the
sanitization
obvious
stuff
looked
a
little
chunky,
a
little
scary
to
me
and
may
not
fit
super
cleanly
and
like
it's
like
exposing
its
own
DSL.
C
Just
like
backdooring
in
this,
like
you
know,
pretty
crazy
DSL,
where
it's
like
well,
do
we
want
to
just
like
suddenly
adapt
this
DSL
as
like
a
thing
we
support,
even
though
I
don't
know
what
it
is
and
yeah
and
like
no
don't
get
me
wrong,
it
seems
like
a
reasonable
I.
Don't
think,
there's
anything
that
was
like
awful.
It
was
just
like.
Oh,
like
there's
some
yeah
there's
some
moving
bits
in
here
that
just
like,
don't
feel
like
I
can
YOLO
approve
them
and.
E
C
And
you
know
like
to
be
clear:
this
is
that
wild
card
pattern
class
is
like
from
the
elastic
search.
You
know
repo,
so
there's
probably
some
confidence
in
it.
Like
I'm
sure
they've
been
you
know
it's
a
public
company.
They
appear
to
make
money
so
I'm
sure
this
works,
but
like
just.
C
You
around
yeah
there's
that
and
there's
also
like.
We
have
had
a
history
of
some
really
awful
performance
regressions
from
standardization
code,
so
like
I'm,
really
anxious
about
just
being
like
cool
and
here's,
some
more
that
we
don't
really
know
how
it
works,
but
we're
just
on
alike,
take
it
over
and
then
it's
on
stuck
one
day,
if,
like
someone
internally
adopts
this
and
then
I
get
or
we
get
yelled
at
because,
like
like,
we
haven't
even
Rob
because
it's
like.
Why
does
your
code
suck
and
it's
like?
C
Well
I,
I,
don't
know
I
just
approved
the
pr
like
sorry.
This
is
years
ago.
Anyway,
hypothetically
it
was
done
anyway,
I
rambled
kind
of
took
over
kind
of
hijacked
things,
but
I
would
say
this
is
the
it
looks
like
they
certainly
would
like
to
get
this
merged.
So
if
people
have
time,
that's
my
Roi
area
too,
not
as
high
as
cereal
stuff.
But
you
know
just
after
that.
D
B
I
just
submitted
it
merge
it
it's
done,
but
if
I
can
figure
out
how
the
hell
to
fix
this
other
thing.
A
B
I'll
get
another
release
out
for
that
too.
I
got.
It
works
on
my
machine
problem
where
the
test
pass
on
my
host,
but
not
for
graphql
in
the
code
space
for
graphql,
so
you
know,
but
actually
I,
just
synced
up
with
main
there
and
the
test
path.
B
B
B
It
doesn't
feel
good
that
if
you
look
at
run
number
one
and
it
failed.
A
D
C
B
Foreign
just
kicked
it
off
again
for
attempt
number
three
just
to
see
and
there
might
be
a
seed
value
in
there
that
I'd
be
able
to
run
the
test
with
to
see
if
it
becomes
a
problem.
But
that
sucks,
because
recently
I
noticed
that
the
rack
tests
are
failing
due
non-deterministically,
also
because
of
ordering
problems.
So
I
don't
know
just
we
need
some
stability
around
here.
Listen
that
that
was
it
for
my
agenda
on
the
list.
I
don't
know
if
y'all
have
anything
in
this,
what
I
perceive
to
be
a
somber
day.
D
B
B
E
I
guess
the
good
animal
guys
thanks
for
running
the
the
sick
today.