►
From YouTube: 2021-10-14 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).
B
A
B
I
was
really
on
the
fence
yeah.
I
decided
not
to
go
at
the
last
minute,
but
yeah,
I'm
kind
of
surprised
too.
I
don't
know
this
is
like
I
think
they're.
This
is
in
person,
slash
virtual.
I
think
yeah.
D
E
A
B
Yeah
yeah,
I
went
last
year
to
the
virtual
one
and
I'm
pretty
done
with
virtual
conferences
me
too
yeah
but
yeah.
Then
I
was
like
oh
that's,
gonna,
be
in
person,
so
I'll
go
and
then
I
was
like.
Oh
man.
This
is
like.
I
don't
know
right.
A
A
B
Goof
gone
yeah
yeah
we're
talking
about
how
not
there
this
year,
which
again,
I
think,
that's
probably
we're
gonna,
have
some
pretty
low
turnout
for
this
meeting.
So
I
don't
have
high
anticipation.
I
know
anthony's
out
for
coupons
specifically
so,
okay,
yeah
yeah.
D
I
have
just
traveled
out
out
to
california
two
weeks
ago
and
felt
it
was
a
little
too
close.
I
had
been
kind
of
itching
to
go,
having
missed
it
last
year,
but
yeah
and
with
you.
I
have
very
little
interest
in
trying
to
participate
virtually
it's.
B
So
hard
just
I
mean
I
think,
even
last
year
I
I
was
a
part
of
the
panel
for
the
hotel
days
and
like
I
just
it's
just
one,
it's
like
when
you're
just
at
the
same
place,
you
are
every
single
day
but
you're
at
a
conference.
It's
it's
just
not
the
same
yeah.
D
Much
harder
to,
even
even
if
you
could
stare
at
the
screen,
it's
very
hard
to
convince
everyone
else
that
you're
not
there
right
like
I'm
sorry
by
there
I
mean
you're,
not
in
wherever
you
normally
are
on
a
work
day,
whereas
if
you're
away
at
the
conference,
it's
much
easier
to
say
I'm
not
here,
stop
writing
to
me.
Yeah.
B
D
Yeah
I
was
there
that
was
good
yeah.
I
was
there
all
right.
In
fact,
I
think
I
went
to
there
was.
There
was
at
least
one
open,
telemetry
session
that
I
went
to
there.
I
think
it
was
like
there
was
a
sort
of
a
panel
like
I
think
liz
was
on
stage.
You
may
have
been
up
there
tyler.
I
don't.
I
was.
B
D
B
Yeah
yeah
it
was,
I
mean
in
retrospect
it
was
pretty
early
days.
It
felt
like
I
think
they
had
already
been
around
for
a
year,
so
it
felt
like
it
would
have
been
for
a
long
time
but
turns
out
organizing
the
coming
together
of
two
different
projects,
plus
a
bunch
of
new
ideas
and
plus
a
bunch
of
companies
takes
a
little
while
so
yeah
yeah,
but
yeah,
well,
cool
yeah,
we're
about
four
minutes
in
this
is
probably
what
we
can
expect
for.
The
duration
of
the
meeting.
B
Welcome
everyone,
I'm
guessing
rich
and
steve-
are
the
only
other,
two
or
well
steve
wilbury
are
not
at
kubecon
or
they
are,
and
they're
just
taking
a
break,
yeah
and
so
yeah.
I
know
anthony
specifically
told
me
he's
not
going
to
be
here,
so
I'm
definitely
excited
for
him
to
go
enjoying
that.
I'm
also
excited
to
hear
like
what
the
report
is
on
that
I've
already
seen
some
good
presentations
coming
out
of
it
from
like
the
hotel
days
as
well,
so
yeah
all
good
stuff.
B
So
yeah,
if
you
haven't
already,
please
add
your
name
to
the
attendees
list,
which
you
have
I'm
looking
at
one
for
one
and
then
I
can
start
sharing
my
screen.
Oh
that's
interesting.
B
Okay,
cool,
hopefully
you
can
see
the
agenda,
I
don't
have
too
much.
We
don't
have
a
project
boards
opened
yet
I
know
I
had
said
that
I
was
gonna
open
up
some
project
boards.
I
had
yet
to
do
that
mostly
just
been
catching
up.
I
was
out
on
vacation
last
week,
and
so
I
think
that
cleaning
things
up
is
kind
of
my
main
goal
here.
To
that
effect,
I
wanted
to
just
kind
of
like
give
a
little
bit
of
an
update.
B
B
If
you
had
your
pr
or
issue
closed
or
if
you're
listening
to
this
and
you
on,
recording
and
had
it
closed,
keep
in
mind,
closing
is
just
as
easy
as
opening
so
like.
If
I
did
that
in
error,
please
fix
it
and
I'm
happy
to
have
a
conversation
about
it.
B
That
being
said,
the
goal
is
to
try
to
move
instrumentation
into
self-hosted
repositories
or
directly
into
the
libraries
themselves,
is
kind
of
the
goal
and
I've
kind
of
talked
with
a
few
people
in
the
community
outside
of
this
sig
specifically,
and
they
had
some
concerns
about
that
and
I
think
they're
well
founded,
and
I
think
that
we
had
concerns
with
it.
B
So
I
just
wanted
to
kind
of
make
sure
that
they're
captured
and
namely
it's
just
that
there's
this
explosion
of
effort,
I
think,
is
probably
the
best
way
to
say
it
or
well.
That's
probably
not
the
best
way
to
say
it.
B
It's
a
way
to
say
it
where
you
have
multiple
groups
trying
to
maintain
essentially
the
same
instrumentation,
so
it
you
know
just
a
you
know,
call
of
spade
a
spade,
like
looking
back
at
open
tracing,
I
think
specifically,
like
datadog
had
some
instrumentation
that
they
hosted
and
then
every
other
company
just
forked
it
and
they
maintain
their
own
fork
and
they
put
their
own
fixes
in.
But
then
they
you
know
the
upstream.
It
became
really
contentious
as
to
like
how
that
was
actually
gonna
work.
B
It's
not
ideal,
and
I
don't
think
that
opencemetry
was
made.
You
know
as
a
place
to
you
know,
promote
that
kind
of
thing.
So
I
just
want
to
say
like
I
think
this
is
a
really
good
move,
specifically
with
the
developer
velocity
we
have.
Currently.
I
don't
think
that
we
can
support
an
ever-growing
contrib
repository,
so
I
think
we've
made
the
right
choice.
B
I
just
want
to
say
that,
like
we
probably
need
to
keep
working
on
this
policy
guidelines
and
saying
win
that
third
option
just
for
those
that
aren't
as
familiar
with
this
as
I
am,
if
I
can
bring
this
up
really
quick.
So
this
is
the
document
that
I'm
talking
about
where
we
recommended
the
native
instrumentation
packages,
a
dedicated
public
repository
that
self-hosted
and
then
finally,
where
the
hotel
go
contribute
repository
would
host
the
instruction.
B
So
I
think
that
I
think
this
is
a
really
good
policy
to
start
with.
I
just
think
that
we
need
to
make
sure
that
it's
clear
or
give
good
examples.
Somebody's
mic
is
on
josh
you're
unmuted.
By
the
way,
sorry
yeah
no
worries,
and
so
I
just
want
to
make
sure
that,
like
maybe
this
stock
takes
a
few
iterations
where
we
spell
out
like
you
know,
these
are
really
good
rations.
I
see
a
hand,
so
I'm
just
going
to
pause
talking
pretty
fast.
D
Either,
can
you
explain
that
exchange
that
took
place
this
morning
where
somebody
wrote
addressing
the
database
sql
package,
somebody.
C
C
D
The
policy
you
just
posted
shouldn't
this
happen,
I
think
liz
then
wrote
back
and
said:
oh
yeah,
maybe
so,
okay,
I
couldn't
really
understand
the
argument
it
sounded
like
there
was
something
there
I'm
talking.
This
is
probably
like
four
or
five
hours
ago
today,
oh
okay,
yeah,
maybe
maybe
the
third
one
down
there.
Let's
see
at
the
end.
E
D
B
Well,
so
there
definitely
is
not
a
package
with
the
database
sql
instrumentation
in
it.
Currently
we've
never
merged
one
that
doesn't
exist.
I
think
they're
con
well.
Their
point
is,
is
that
we
host
the
net
http,
so
why
aren't
we
hosting
the
database
sql,
which
I
think
is
kind
of
missing
the
point,
whereas
like
we
can't
really
keep
adding,
especially
this
is
like
one
of
the
ones
that
like
is
one
of
the
most
contentious
pieces
of
instrumentation?
We
try
to
add.
B
I
know,
there's
three
different
attempts
at
it
yeah,
so
I
I
could
probably
spend
some
time
clarifying
this.
I
hadn't
seen
this
it
does.
I
think
this
is
an
important
one,
though,
because
this
kind
of
gets
back
to
that
point.
It's
like
what
what
where's
the
the
point
where
you
you
know
tip
to
having
something
hosted
here,
and
I
think
that
that
is
something
that
we
need
to
have
a
little
bit
clearer
spelled
out.
I
definitely
don't
think
that,
like
it's,
you
know
we
have
net
hp.
D
I
wasn't
really
sure
whether
derek
was
arguing.
We
should
host
this
here
or
if
he
was
more
concerned.
Something
like
somebody
could
grab
this
this
package
path
and
then
it
would
be
really
confusing
if
they
did,
you
know
because
it
says
something
about:
can
anyone
create
a
standalone
repo
to
claim
that
import
path
is
he's
saying
I'm
not
sure
if
he's
asking
like,
would
somebody
please
do
this
or
I'm
concerned
that
somebody
might
do
this.
B
I
don't
really
understand
that
either
because
they
can't
this
package,
I
mean,
I
guess
you:
could
you
won't
ever
be
able
to
import
it
from
this
url
because
we
control
it.
We're
not
gonna
merge
anything.
So
I
don't
think
that
that's
I
mean
the
answer
to
that
question
is
no.
They
would
have
to
prefix
that
with
something
else.
I
don't
see
why
anybody
else
that
the
standard
couldn't
you
know,
write
their
own
state
of
lib,
the
instrumentation
that
seems
completely
reasonable
to
me.
B
I
do
think
the
standalone
kind
of
stands
out
a
little
bit
as
maybe
having
a
little
bit
higher
priority.
Just
from
the
fact
that
if
we
go
back
to
that
policy
we
were
talking
about
like
number
ones.
I
mean
I
don't
think
that's
ever
going
to
happen.
I
love
open
telemetry,
but
I
don't
think
the
going
authors
are
going
to
incorporate
open
telemetry
into
the
standard
library
without
open,
telemetry
itself
being
a
part
of
the
standard
library.
B
A
D
I,
if
I
remember
correctly,
they
were
around
the
same
package,
but
more
to
the
point.
I
think
there
was
discussion
about
how
should
we
do
this?
There
were
different
styles,
there
was
different
prior
art
and
there
was
this
question
of.
If
we're
going
to
work
on
the
one
that's
going
to
become
canonical,
which
style
should
we
use,
okay
and-
and
that
was
where
really
ground-
to
a
halt.
A
That's
that's
kind
of
my
interpretation
too,
like
that's
from
my
my
back
point
with
that
being
the
case
like
yes,
native
implementation,
isn't
gonna
happen,
but
it
clearly
step
two
can
happen
and
we
can
also
see
who
becomes
the
front
runner
in
that
like.
Where
does
this
get
used?
A
The
most
can't
we
before
we
decide
like
hey
this
one
versus
like
this
implementation
versus
that
implementation,
or
we
could
also
measure
based
on
like
which
one's
closer
to
the
rest
of
the
contribute,
library
or
the
rest
of
the
the
go
standard
library
or
not.
The
hotel,
open,
open,
telemetry
go
sdk
and
and
see
where
that
lies
versus,
like
you
know,
whoever
gets
whoever
submits
first.
B
Yeah-
and
I
think
that's
that's
probably
what
we
need
to
just
say
back
is
like
I,
I
mean
there's
nothing
that
says
it
has
to
be
hosted
here
like
it
isn't.
Xam
already
did
a
great
job
and
he's
hosting
his
own
and
people
are
using
it
and
if
you
know
that
becomes
the
front
runner,
because
it's
the
first
or
because
it's
the
best
like
then
maybe
that's
a
good
indicator
that
we
should
bring
that
in
here.
B
If
there
is
contention
and
the
people
that
want
to
be
forking
his
work
or
something
like
that,
but
like
I
think
that
saying
that
we
can't
have
third
party
implementations
of
standard
library
packages
is
just
not
capturing.
I
think
what
our
policy
goals
were
there.
D
Yeah,
I
I
think
that,
right,
I
remember
when
we
told
you
told
somebody
told
sam
to
scram,
the
idea
was
you
host
yours
and
joe
will,
host
his
and
mo
will
host
his
and
let's
kind
of
have
a
beauty
contest
and
we'll
see
whether
any
of
them
become
dominant,
because
there
was
some
concern
about
this.
One
has
different
performance
characteristics
because
it's
using
reflection
more
or
this
one
produces
more
spans.
D
So
I
think
the
idea
was
like
you
know:
let's
try,
multiple
and
let
people
use
them
and
that's
the
only
way
they're
going
to
be
able
to
decide
and
maybe
we'll
wind
up
with
there
is
no
winner
and
it
doesn't
matter.
E
Is
possibly
true
for
the
net
http
instrumentation,
it's
been
coming
up
in
this
discussion
about
sampling.
You
know:
must
this
the
http
implementation
declare
its
attributes
at
the
moment
that
the
span
starts
and
for
sampling
we're
starting
to
say
yes,
they
should
they
must.
They
must
should
it's
a
complicated
discussion.
E
The
point
is
that
the
reason
why
this
is
coming
up
is
that
some
instrumentation
packages
are
going
to
hyper
optimize
and
then
they
won't
know
their
attributes
at
the
beginning
of
the
span
and
some
are,
and
maybe
those
are
two
viable
http
implementations.
One
of
them
satisfies
the
sampling
requirement
and
one
of
them
doesn't
they're
two
viable
options,
though,
starting
to
look
that
way
like
there
might
be
an
official
sampling
one
and
an
official,
not
sampling,
one,
possibly.
B
Yeah,
I
think
that's
a
a
great
example
of
exactly
what
everyone's
saying
and
maybe
even
that's
something.
You
know
I
didn't
want
to
say
this
as
a
response
to
liz
but
like
maybe
that
is
also
a
good
reason
to
like
not
have
the
net
http
there,
because
there
may
be
a
better
way
and,
and
honestly,
we've
also
talked
about
in
the
past,
like
having
a
instrumentation
library,
doesn't
include
having
multiple
instrumentation
libraries.
B
You
know
it's
definitely
something
that
we
could
do
where
we
could
have
multiple
implementations
say,
like
you
know
the
database,
where
the
http
library
is
to
go
out
into
the
wild,
and
you
know,
there's
really
big
support
for
two
kind
of
like
what
joshua's
saying
like
there's,
there's
one
really
thin
and
one
kind
of
like
full
featured
and
maybe
there's
enough
community.
You
know
commitment
and
respect
like
they.
They
actually
want
the
unification
of
that
into
the
standard
into
the
contribution.
I
think
that's
totally
possible.
B
A
Okay,
so
we've
gone
around
in
circles
a
couple
times.
What
do
we
want
to
take
with
this
specific
one,
and
how
do
we
want
to
try
and
fix
it
so
that
we
don't
get?
We
don't
keep
getting
this
question
over
and
over
again,
I
think
that's
well.
B
I
don't
think
there's
actually
too
much
of
a
question
like
if
you
read
the
policy
yeah
it.
If
you
read
the
policy
like
it
answers
all
the
questions,
it's
not
based
on
upstream
instrumentation
and
the
source.
It's
not
like.
It
comes
from
the
standard
library.
We
do
this
if
it
comes
from
google
or
anywhere
else
like
we
do
something
else
like
it's
no
like
this
is
this
is
this
policy
that
applies
to
all
instrumentation,
regardless
of
the
source.
B
I
think
that
that's
that's
pretty
clear
and
I
think
that's
already
answering
some
of
those
questions.
I
think
the
only
question
that
comes
up
is
like.
Can
we
clarify
the
third
step
is
to
like
when
what's
the
barrier,
and
then
maybe
also
it
is.
The
next
question
is:
what
do
we
do
with
what's
already
existing,
because
it
is
not
released
as
stable?
It
is
still
still
experimental,
and
are
we
going
to
support
that?
The
long
term,
I
guess,
is
an
open
question.
A
Yeah,
that's
kind
of
really
more
where
I
was
getting
cut
is
like
how
do
we?
How
do
we
get
it
from
like
the
the
question
seems
to
be
like
we
want
this
in
the
standard
lib,
you
say
it's
an
option,
but
you
don't
really
say
what
we
need
to
do
to
make
it
an
option.
So
we
probably
need
to
capture
some
clarification
around
that.
B
Yeah,
I
gotcha
so
one
of
the
things
that
maybe
to
highlight
then
is
yeah.
Whoever
just
highlighted
this
as
well
need
a
sponsorship
model
yeah,
because
the
thing
that
this
policy
was
intended
to
avoid
is
where
we
were
where
we
had
a
lot
of
open
prs
that
were
stale,
we
have
instrumentation,
that
is
official,
but
it's
I
there's
people.
I
think
that
have
submitted
these
things
and
then
left
in
the
project
and
they're
no
longer
a
part
of
it.
B
So
anyone
who
wants
to
work
on
that
instrumentation
needs
to
onboard
themselves.
So
it
really
is
about
like
how
do
we
develop
a
sponsorship
model
and
until
we
can
develop
a
sponsorship
model,
I
wouldn't
really
feel
comfortable
answering
either
of
those
questions.
I
think
because
we
need
that
to
say
like
how
we're
going
to
put
that
barrier
in
place,
but
also
we
need
that
to
say
like
how
are
we
going
to
progress?
What's
already
there
for
going
forward?
B
I
can
see
a
future
where
you
have
you
know
two
different
contributors,
hopefully
from
different
companies
or
or
more
that
we
say
like
has
to
be
the
barrier
and
they're
assigned
some
sort
of
ownership
model
to
that.
B
We
also
need
a
deprecation
model
like
what
happens
when
they
stop,
you
know
contributing
and
that
number
falls
below
the
requisite
amount.
How
do
we
handle
that
situation?
So
I
think
that
we
should
develop
that.
I
don't
know
if
we
want
to
try
to
figure
that
out
today
or
if
we
want,
maybe
just
have
an
issue
track
it
but
yeah.
I
agree
like
I
think
that,
there's
something
that
needs
to
be
done
there.
B
Yeah
I,
but
I
think,
that's
pretty
much
captured
a
lot
of
the
things
I
wanted
to
talk
about.
Aaron.
Are
you
interested
in
helping
with
the
sponsorship
model.
A
Not
right
this
second,
I
have
a
little
bit
of
things
that
are
that
are
changing
on
my
side,
so
I
probably
couldn't
get
to
it
until,
like
two
weeks
from
now
or
a
week
and
a
half
from
now.
D
About
a
time
and
attention
and
responsibility.
B
I
mean
I'll
always
take
tips,
but
no
I
I
I
do
mean.
D
B
Yeah,
I
I
I
need
your
time.
That's
your
money
really
got
it.
A
D
Before
we
move
off
this
topic,
when
you
were
saying
earlier
about
how
there's
these
three
these
sort
of
three
modes,
you
know-
and
I
wonder
if
we
need
to
clarify
that
one
of
those
modes
is
really
hypothetical,
then,
like
you
know
to
just
to
dismay,
just
try
to
throw
the
right
word
to
convince
people
not
to
pine
for
it
too
much.
D
B
Yeah,
that's,
I
think
well
said,
and
it's
capturing
a
lot
of
the
sentiment,
and
I
don't
think
that
we're
alone
in
this
I've
also
noticed,
like
you
know
something
that
stands
out
to
the
prometheus
model
as
well
like
they
they've
got
some
official
support
for
certain
detectors,
actually
they're
not
called
detectors,
but
you
know
what
I
mean
I
think
and
their
instrumentation,
but
like,
like
you
said,
I
think
it's
almost.
I
like
the
word
that
you
used,
but
I've
already
forgotten
it.
B
Reckless,
I
think,
was
it
for
us
to.
Actually,
you
know
take
these
on
when
we
can't
we
can't
actually
support
them
and
yeah.
I
mean
that
it's
not
trivial,
like
we're,
I'm
kind
of
laughing
but
like
it
isn't
like
if
there's
a
security
bug-
and
we
don't
have
the
bandwidth
to
try
to
resolve
these
sort
of
things
like
that's
an
impact
to
a
large
number
of
people
that
are
using
this,
and
that's
that's
not
good,
that's
not
something.
I
think
this
community
wants
to
be
known.
B
For,
let
alone
do
you
know,
so
it
is
important.
So
what
I'm
getting
from
this
is
I've
got
a
little
bit
of
time
after
this
meeting,
I
might
open
an
issue
to
try
to
track
this
and
then
maybe
point
that
database
issue
to
it
to
say
like
what.
What's
the,
how
are
we
going
to
handle?
What
currently
exists?
B
You
know,
how
does
that
look,
and
then
you
know
how
maybe
also
another
issue
on
like
defining
that
barrier
and
defining
that
like
the
third,
the
third
stage
is,
you
know,
what's
the
barrier
there,
I
don't
know
if
that
makes
sense
to
what
you're
saying
yeah,
okay,
yeah.
D
I
think
I
think
it
does.
I
I
don't
have
much
experience
with
this,
but
for
those
of
you
who
have
touched
mpm,
you
know
the
node
package
manager.
D
I
was
a
little
confused
by
it
at
first,
but
the
way
that
if
I
understand
it
correctly,
it's
a
registry
of
mostly
blessed
names
like
shorthand
names
or
part
for
projects
that
are
hosted
elsewhere
and
that's
something
that
you
know
go
doesn't
really
offer
easily
absent,
maybe
creating
packages
of
aliases
or
something.
D
But
it
was
a
weird
thing
where
you
could
look
up
like.
Oh
this
person's
using
you
know
websocket,
which
sounds
like
something
almost
built
in
and
then
you
go
track
it
down.
It
turns
out
no,
it's
just
pointing
at
somebody's
github
repository
over
there,
but
apparently
people
agreed
that
this
shorthand
name
was
good
enough
to
say
you
know
that
name
points
there.
That
seems
almost
like
some
sort
of
middle
ground.
If
the
language
afforded
that
more
easily
or
like
we
have
vanity
urls,
we
could
get
there.
D
C
I
mean
there
is
the
open,
telemetry
plugins
page
right,
the
catalog
of
plugins,
not
right,
not
specifically
for
go,
but
for
everything
the
registry
yeah
and
yeah
npm.
I
don't
know
if
I'd
call
it
npm
blessed
packages.
I
think
anyone
who
wants
can
can
throw
something
up
there.
But
yes,.
C
B
So
I
think
that's
an
interesting
thing
and
I
did
want
to
make
sure
after
closing,
that
database,
sequel
and
exams
pr
that
you
know
this
is
a
good
case,
because
xam
has
we
have
this
listed
now
on
the
registry.
This
is
a
third
party
set
of
instrumentation
for
database
sql.
I
think
I
might
even
tag
this
wrong.
I
also
blew
up
the
website
by
merging
this,
so
ben's
watching
this
later,
my
bad
but
yeah.
B
I
think
that,
like
this
is
a
really
good
thing,
and
we
talked
about
this
before
when
we
read
the
policy
like
this
is
a
starting
point.
I
think
this
registry
needs
more
for
those
features
you
were
talking
about
steve,
like
maybe
voting
or
maybe
like
understanding
like
usage
or
something
like
that
would
be,
I
think,
helpful,
to
drive
decision
in
the
future.
B
B
I
mean
right
now:
it's
it's
completely
a
handshake
agreement,
but
it
would
be
cool
if
we
could
build,
maybe
some
tooling
or
automation,
to
validate
those
sort
of
things.
I
don't
know
how
that
would
look,
but
I
mean
it's
possible.
We
have
some
smart
people
on
the
call.
You
know.
B
Yeah,
okay,
cool
awesome
I'll
try
to
follow
up
on
this
after
this
or
tomorrow
morning,
if
I'm
being
honest,
but
I'm
gonna
try
this
after
next
thing
on
the
agenda,
I
had
added
that
I'd
like
to
get
a
release
out
next
week.
I
have
limited
time
this
week,
but
there's
these
two
bug
fixes
evan.
I
think
you're
on
the
call
as
well.
One
of
these
is
yours:
this
fixed
ipv6,
parsing
error
and
fix
http
client
ip.
B
C
A
Yep
we
got
two
thumbs
up
here
so
then
I
I
still
have
an
unanswered
question
in
my
comment.
That's
above
that
of
like.
Should
we
be
only
going
for
the
single
ip
address,
or
should
we
just
combine
those
headers
into
a
comma
separated
string
so
that
we
can
yeah
that
information
that.
B
Is
a
good
question?
Where's
the
bug,
the
actual
bug
links
to
the
submit
to
commit
the
answer
is
no
semantic
convention
say
to
use
just
the
first
one
which
is
going
to
the
client
and
to
drop
the
proxies.
I
I
thought
it
was
in
the
bug,
but
I
definitely
looked
it
up
at
the
same
time
just
to
answer
that
yeah.
A
C
Not
just
the
first
entry,
but
the
second
exported
header
also
had
comma
separated
list.
You
were
saying
which
one
should
we
take
like?
Should
you
take
the
first
one
from
the
second
one
or
the
first
one
from
the
set
combine
them
all
of
them,
but
combine
them
how
by
appending
the
second
one
to
the
first
one
yeah
comma
separation.
A
No,
no,
that
we
we
return
the
the
attribute
that
we
return
would
be
the
entire
list.
A
B
Right
yeah,
I
think
I
should
be
able
to
actually
find
that
pretty
quick.
Sorry
if.
B
That's
fine!
Well
I
let's
double
check
because
I
don't
want
to
say
the
wrong
thing
and
neither
like
definitely
want
to
be
right
so
actually
addresses
the
original
client
behind
all
proxies,
I
known
as
the
exporter
from
so
I
think
at
the
original
there.
But
this
is
not
necessarily
the
same
yeah
yeah,
I
think
it's.
The
original
is
the
key
word
there.
B
B
A
B
We
we
could
always
ask
the
in
the
spec
channel
for
slack
as
well,
but
yeah.
I
think
I
think
that's
something
that
it
was
it
was.
It
was
confusing.
I
think
that
you're
not
alone
in
that
confusion.
B
E
D
E
B
Exactly
right,
cool
evan
also
has
a
pr
out
he's
got
two
approvals.
Garth
was
the
original
author
of
the
original
issue.
I
haven't
read
through
all
of
them.
I
looked
at
the
code,
it
looked
pretty
good.
I
haven't
finished
my
approval
though
it
looks
like
so
this
is
another
one
I'd
like
to
get
in
in
the
release
that
we're
trying
to
get
out,
I'm
hoping
to
get
out
next
week.
So
if
you
have
time,
please
take
a
look.
C
Yeah,
so
this
is
relating
to
that
pr,
this
pr,
I
tried
re-running,
two
or
three
or
four
times
I
think,
and
that
test
is
always
failing
the
grpc
reconnect
and
either
the
test
race
or
the
the
other
one
like
test
race,
test
coverage
like
it's.
C
B
B
C
B
Incremental
improvements
on
it
in
the
past
and
it
went
away
and
then
things
got
a
little
bit
slower
or
a
little
bit
slower
and
then
it
just
comes
back.
I
thought
I
had
fixed
the
deadline
issue,
but
it's
just
yeah.
I
agree
it's
it's
honestly.
B
It's
doing
more
harm
than
good
and
honestly,
it
comes
from
our
connection
reset
model
in
the
grpc
exporter,
which
I'm
pretty
sure
was
questioned.
Whether
we
need
to
be
doing
that
jrpc
itself
doesn't
auto
reconnect
so
like
yeah,
and
there
was
a
bug
I
think
I
reopened
it.
I
think
it
still
exists.
We
should
probably
prioritize
that
higher
than
where
it
is
currently.
F
B
C
B
Reproduce
it
locally
and
then
for
what
it
is
like
I
I'll
post
this
here.
But
this
is
still
an
open
issue
and
I.
C
B
I
also
looks
like
this
punya's
run.
You
can
restrict
oh
yeah.
I
think
you
can
also
restrict
the
cpu
account
for
docker,
and
the
memory
count
was
also.
I
played
a
few
of
these
things
to
try
to
like
see
what
sort
of
resource
starvation
caused
it
and
it
was
just
random.
B
I
think
it
was
how
fast
the
cpu
is
more
than
it
is
the
actual
cpu
count
I
think,
had
more
to
do
with
it,
which,
on
actions,
I'm
sure
that
this
is
probably
run
somehow
through
functions,
or
something
like
that
where
maybe
it's
getting
paused
really
indeterminately.
So
I'm
not
exactly
sure
what
the
timing
issue
is
there.
B
C
B
Yeah-
and
I
might
just
recommend,
looking
at
the
holistic
issue
of
the
connection
reset
and
see
if
we
actually
need
it
if.
B
Out
that
would
probably
be
a
better
idea.
Okay,
I'll
go.
Take
a
look:
okay,
perfect,
okay,
cool,
I'm
gonna,
stop
sharing
the
screen
and
looks
like
we're
at
the
point
where,
if
anybody
has
something
else
they
want
to
talk
about
on
the
agenda,
please
speak
up.
I
also
see
somebody
left
a
chat.
B
Not
seeing
anything
else
from
anybody
josh,
I
didn't
include
in
this
the
metrics
progress
on
here,
but
maybe,
if
you
could
just
give
us
like
a
the
elevator
pitch
on
where
we're
at
on
that
one.
E
E
E
E
Where
there's
like
one
out
where
I
said
we,
I
think
we
should
streamline
the
otlp
export
path
and
there's
like
a
structural
change
there
that
will
only
affect
the
sdk
there's
tissue.
There's
tickets
open
saying
it's
time
to
think
about
whether
this
is
the
api
we
want
or
whether
it's
not
and
and
we're
there
now,
because
the
sdk
api
migration
that
I
talked
about
over
the
last
few
months
is
done,
which
means
that,
what's
in
the
api
directory,
if
you
pull
up
godoc
at
the
head
of
the
repository,
that's
it
and
anything.
E
So
any
discussion
about
hey,
let's,
simplify
or
restructure
or
put
into
new
packages
the
metrics
api
that
can
happen
now,
I'm
not
eager
to
start
that
right
away,
but
it
it
could-
and
what
I
am
eager
to
do,
though,
is
as
these
refactorings
and
I
guess,
kind
of
like
deferred
maintenances
catch
up.
There
aren't
very
many
spec
items
left
to
match
this.
The
the
draft
or
the
working
draft
of
the
metrics
sdk,
metrics
and
sdk.
E
The
big
pieces
are
exemplar
support
for
the
histogram,
especially
aggregator
and
view
configuration
and
the
view
configuration
one's
pretty
big,
but
this
is
what
that
pitch.
I
need
to
do
is
just
explain
to
you
that,
basically,
all
the
functional
pieces,
the
machinery
for
these
views
is
built,
but
the
view
configuration
is
going
to
be
this
new
piece
of
magic
glue
that
parses
some
or
read
some
configuration
struct
and
then
turns
it
into
machinery,
and
that
I
haven't.
E
I
had
a
prototype
once
so
I
kind
of
know
what's
involved
and
both
of
these
pieces.
I
have
prototyped
at
some
point,
so
I
continue
to
say
I
will
try
to
come
up
with
a
more
coherent
state
of
the
world
than
that.
You
know
it
may
be
a
document
try
for
next
week
or
the
week
after
and
and
hopefully,
and
share
my
plans
as
it
goes.
E
E
I
tried
to
do
something
internally
at
lightstep
with
the
sdk,
and
it
had
a
lot
of
sharp
edges
that
cut
me,
but
I
did
get
it
to
work
in
the
end,
so
I
have,
over
the
last
week
copied
replace
or
sorry
added
an
output
for
all
of
light
steps,
internal
services
to
use
hotels,
metrics
sdk,
and
that
includes
that
would
be
straightforward,
just
swapping
a
statsd
export
path
that
uses
an
agent
like
datadogs
that
we've
customized
with
an
otlp
export
path.
That
part
actually
did
a
few
months
ago
in
an
old
hackathon.
E
The
project
this
time
was
to
for
our
satellites
which
run
inside
of
customers.
Data
centers.
The
satellites
had
an
old
metrics
path
and
I
got
it
to
use
hotel
as
well,
and
that
meant
having
the
option
to
export
otlp
by
push
or
prometheus
by
pull,
so
that
some
of
our
customers
are
only
using
our
trace
product.
They
want
to
monitor
their
satellites
but
they're,
not
using
our
metrics
product,
so
they
want
to
use
prometheus
to
monitor
satellites.
So
I'm
doing
that
using
hotels,
sdk,
and
that
meant
configuring
sdk
that
had
both
temporalities.
E
C
So
I
was
looking
at
some
of
the
issues.
This
is
related
to
metrics
and
I
noticed
somebody
from
the
collector
thing
was
complaining
about.
Oh,
they
were
all
their
memory
was
disappearing
when
they
configured
a
histogram
exact
with
cumulative
distribution,
which
of
course
makes
sense
because
they're
just
keeping
every
single
point
since
the
beginning
of
time.
C
But
the
reason
I
ask
about
it
is
my
my
question:
is
they
they
made
a
comment
that
oh
they
couldn't
really
tell
what
was
happening
because
our
standard
out
metric
exporter
doesn't
even
try
to
print
anything
for
histograms
or
distributions.
C
E
Yeah
I
didn't.
I
would
welcome
that
this
view.
Stuff
really
is
going
to
be
a
new
module
like
I
don't
think
you
should
have
to
make
it
in
either,
but
that
would
just
kind
of
refer
to
all
the
bits
and
pieces
of
the
sdk
that
we
have
and
and
wire
them
together,
essentially
so,
printing,
the
histogram,
the
explicit
histogram.
E
That
does
make
sense,
and
on
that
note
I
have
a
pending
pr
with
support
for
this
exponential
histogram,
which
I
hubble
together
from
a
prototype
and
some
new
relic
code
from
out
of
java,
and
so
on.
That
would
that's
that's
coming
in
the
same
vein,
I
think
we
should
print
those
once
they're
there
and
so
and
we
should
support
the
collector.
I
like
that
idea.
I
did
also
see
this
issue
an
issue
about
the
exact
aggregator.
E
I'm
not
sure
if
it's
the
same
one
that
you're
referring
to,
but
it
it
it
confused,
confuses
people,
and
I
was
writing
that.
I
think
we
should
remove
the
exact
aggregator.
Very
few
people
want
that.
The
reason
it's
there
is
from
an
ancient
ancient
time
when,
supposing
you
wanted
to
forward
well,
I
you
know
you
can
compute
a
quantile
from
it.
E
E
D
E
C
B
Yeah
so
josh
just
a
heads
up
on
the
the
state
of
the
metric
stuff.
Like
no
rush
on
that,
I
understand
like
that's,
I
think,
there's
the
communities
like
needing
it
tomorrow,
kind
of
thing,
so
I
don't
want
to
pressure
you
too
much
on
that.
B
I
think
if
that's
you're
asking
the
right
questions
is
kind
of
something
I
want
to
say
because
one
of
the
things
that
I
think
that
we've
noticed
after
going
through
the
tracing
release
is
there's
a
lot
of
craft
from
you
know
many
years
ago
that
we
inherited
into
this
project,
and
the
last
thing
I
would
want
is
for
the
api
to
be
confined
into
a
way
that
doesn't
really
work
in
in
the
long
term,
and
so,
if
we
have
to
make
some
bolder
moves,
I'd
be
more
inclined
to
doing
that
than
to
try
to
bring
along
compatibility
for
early
adopters.
B
For
the
experimental
api,
I
think,
is
what
I
I
would
say,
but
yeah
I
think,
you're
asking
the
right
questions
is
what
I'm
saying.
E
Yeah,
my
previous
prior
statement
was
about
sdk
support
that
I
think
nobody
cares
about
it's
just
if
it's
there
someone's
going
to
use
it
by
mistake
for
the
api
yeah
we
gotta,
we
gotta
ask
those.
These
are
hard
questions.
I
don't
I
I'm
willing
to
upgrade
any
and
all
code,
I've
written
to
use
it,
and
I
was
supporting
a
customer.
Oh
yeah,
that's
what
it
was.
It
was
an
internal,
never
mind.
E
It
was
someone
else
trying
to
use
lightstop
and
ended
up
with
that
exact
aggregator
and
it
ended
up
badly
so
so
stuff
like
that.
There's
also
a
min
max
sum
count
that
that
one
is
on
the
edge
of
we're
debating
right
now,
whether
that
gets
folded
into
the
trivial
histogram
or
called
a
trivial
histogram
or
whether
it
maintains
the
name
min
max
some
count,
and
you
know
we
talked
about
registration
of
propagators.
I
think.
Last
week
the
same
issue
comes
up
with
all
like.
E
There
are
six
or
seven
aggregators
now
and
I
don't
ever
need
them
all
at
once.
So
should
I
link
them
all
in
that
view,
machinery
is
either
going
to
have
to
refer
to
all
of
them,
because
that's
what
the
view
spec
says,
you
can
choose
your
aggregators
or
the
view
spec
will
have
to
dynamically
load
them
based
on
which
which
ones
you
linked
in,
and
these
questions
are
all
going
to
come
up
and
I'm
I
don't
want
to
dictate
how
it
goes,
because
these
are
tricky
ones.
E
E
Yeah,
yes,
at
one
point
I
earliest
in
the
like
ancient
memory,
I
did
make
a
dynamically
loading
using
the
the
whatever
it's
called.
There's
a
runtime
package
for
loading
code
and
go.
A
A
B
If
you
haven't
seen
the
project
monkey,
I
highly
recommend
it.
It's
pretty
great
yeah
that
face
is
the
right
face,
steve
first
off
the
licensing
says,
never
use
it,
so
don't
actually
use
it,
but
I
I
think,
if
that's
sorry
josh,
I
don't
know.
If
I
cut
you
off
on
that
one.
B
Yeah
cool
is
there
anything
else
that
we
wanted
to
touch
on
in
the
metric
space
before
we
move
on
to
more
explicit
laser
stories
and
maybe
end
it
a
little
bit
early.
B
You're
doing
great,
but
yes
I
I
thank
you,
okay
cool,
then,
is
there
anybody
else
josh!
That's
interesting
that
you're
able
to
do
all
that.
I
I
think
I
kind
of
followed
what
you're
doing,
but
it's
awesome.
E
Yeah,
I
related
the
thing
from
my
hackathon
involving
getting
lightstep
satellite
to
use
both
otlp
and
prometheus,
which
is
something
I
had
built
for,
but
just
fell
into
a
mistake,
yeah
the
other.
The
other
use
case
was
our
corporate
owner,
which
is
servicenow,
has
a
team
trying
to
use
lightstaff
and
they
they
they
are
very
enthusiastic,
but
they
pull.
They
started
just
like
putting
things
together
and
they
ended
up
with
an
exact
aggregator
that
was
then
turning
the
the
measurements
back
into
gauges,
which
is
something
we've
seen
elsewhere.
E
It
ends
up
really
hurting
the
back
end
in
our
in
our
case,
and
we've
seen
this
with
the
hotel
collector's
statsd
receiver
as
well,
which
is
the
out
of
the
box
behavior
you
get
there,
which
is
to
turn
histogram
measurements
into
gauges.
E
So
that's
something
that
you
can
do
with
the
exact
aggregator
that
I
don't
want
people
doing,
and
that's
also
why
I
mentioned
the
exponential
histogram.
The
fix
for
statsd
receiver
is
to
use
the
exponential
histogram
period
and
then,
as
for
you
know,
the
situation
that
I
was
talking
with
my
corporate
overlords
about
they
they're
trying
to
use
the
hotel
api
as
a
forwarding
api
for
sensu.
Does
anyone
here
know
what
sensu
is
it's.
B
E
I
know
so
it's
an
old,
it's
an
old-ish
gometrix
framework
and-
and
I
I
use
the
phrase
untyped
metrics,
so
they
have
numbers
and
names,
but
they
don't
have
gauge
versus
counter
versus
histogram
there
and
it's
there.
It's
not
the
only
system
in
the
world
like
that,
but
you
end
up
in
this
situation
also
with
statsd,
where
you
kind
of
don't
know
what
to
do
with
a
particular
number.
You
end
up
making
a
gauge
and
then
it's
not
always
good.
E
So
this
is
how
they
ended
up
turning
histogram
measurements
into
gauge,
because
everything
is
a
gauge,
but
it
it
also
hurts
if
you
can't
aggregate
those
things
properly
so
yeah.
That
was
why
xact
came
up.
That's
my
other
user
story.
B
Cool
yeah
that
sounds
painful.
Does
anybody
else
have
any
good
user
stories
in
the
past
few
weeks,
even.
C
The
usual
suspects
and
getting
heads
not
headshakes,
I
I
have
somebody
in
our
company
who
is
trying
to
take
the
thing
I
wrote,
which
was
a
converter
from
otel
metrics
to
our
internal
metric
system
and
put
it
into
a
lambda
to
run
on
aws,
because
you
can
put
that
in
there
and
and
plug
it
in
and
export
the
metrics
still
to
our
own
internal
metric
system,
but
from
aws
I
I'm
actually
consulting
with
they
haven't
started
writing
that
code.
So
I'm
just
working
with
them.
B
C
B
Yeah
cool
well,
if
nobody
else
has
any
other
examples,
we
could
probably
end
it
a
little
bit
early
I'll
just
give
a
little
pause
here.
B
Awesome
well
thanks.
Everyone
for
joining,
like
I
said
I'll,
try
to
follow
up
on
some
of
the
things
and
evan.
If
you
have
some
time,
please
feel
free
to
take
a
look
at
that
issue,
but
otherwise
I'll
see
you
all
virtually
or
next
week,
thanks
tyler.