►
From YouTube: 2020-11-06 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
A
So,
in
order
to
use
some
prometheus
facilities,
I
don't
really,
I
don't
really
understand,
what's
going
on,
but
it
basically
includes
all
of
prometheus,
okay,
which
seems
a
little
crazy.
A
A
C
Most
like
so
there's
two
orgs
one
is
prometheus
type
of
stuff.
One
is
bad
watch
x-ray
type
of
stuff.
I
guess
there's
open
source
versus
closed
source
type
of
split,
and
so
it's
more
alolita's
org.
That's
working
on
that.
I
don't
see
it
that
much
actually.
B
Okay,
so
in
order
to
so
that,
customers
basically
can
oh
because
you're
using
the
collector
so.
B
A
I
don't
I
didn't
understand
it
sounded
like
they
needed
to
not
not
pull
the
entire
prometheus
code
base
into
their
code
base
and
rewrite
the
scraping
system
or
something
I
don't
know
it
sounded
crazy
and
I'm
glad
I
wasn't
involved,
but
I'm
glad
to
hear
honor
that
you're
a
big
fan
of
metrics,
because
I
think
we
need
to
completely
rewrite
the
java
metrics
implementation.
A
B
Yeah
I
saw
that
josh
sounded
like
that.
He
was
also
in
favor
of
what
you
were
talking
about.
B
C
A
That
the
java
implementation
is
just
like
you
know,
I'd
understand
it,
because
I
reviewed
all
of
it
when
mogden
was
writing
it,
but
it
doesn't
look
anything
at
all.
It
is
an
sdk.
I
wouldn't.
What
I
really
really
want
to
do,
though,
is
completely
re-implement
the
sdk
just
using
micrometer
under
the
hood.
C
A
So
I
don't
know
anyway,
I'm
going
to
put
an
issue
in
that
basically
says:
rewrite
the
metrics
sdk
and
we'll
see
what
happens.
I
mean
I'd
actually
be
happy
to
do
it,
but
it's
going
to
be
a
it's
a
going
to
be
a
big
big
job
and
I'm
just
trying
to
figure
out
whether
I
actually
have
time
to
do
that
and
do
all
the
pr
reviews
and
all
the
other
stuff
that
I'm
doing
as
a
major.
A
So
it
would
be
great
if
it
sounds
like
josh
has
carlos
digging
into
it,
and
if
carlos
could
go
and
do
a
rewrite,
that'd
be
fine.
I
mean
we
could
review
carlos's
carlos's
code
yeah.
A
A
If
I
were
going
to
do
it,
I
would
probably
create
a
brand
new
sdk
package
like
a
metric
metric
sdk
module,
because
we
have
a
the
sdk
has
already
broken
up
into
pieces
right,
and
so
I
would
have
a
metrics
sdk,
v2
or
something
I
don't
know,
call
it
something
else
and
just
start
re-implementing
it
piecemeal,
just
a
step
at
a
time
based
on
the
go
and
the
the
based
on
the
specs.
That
josh
has
been
writing
and
then
we
can
swap
it
out
once
it's
all
working.
A
B
Is
probably
the
main-
and
I
know
he's
done
a
lot
with
metrics
internally,
so
I
know
he's
interested
in
metrics,
so
I'm
curious
if
he
I'm
guessing,
he
probably
did
them
at
the.net
metrics
work,
but
I'm
not
sure
right.
A
A
A
C
B
C
C
A
Happen
like
they
should
just
be
publishing
it
as
a
separate
because,
all
like
all
the
languages,
they
should
be
publishing
all
the
language
bindings
and
have
a
regular
release
cycle
for
it
and
not
try
to
have
everyone
depend
on
the
latest,
like
some
arbitrary
merge.
That
happened,
which
people
have
tried
in
the
java
repo
before
and
I've
shut
them
down
like
no,
no,
no,
you
wanna
wanna
depend
on
this.
We
need
a
release.
We
need
a
version.
I
think
I
shut
you
down
on
a
rock
early
on.
A
A
Like
if
we
were
to
like,
if
we
needed
to
do
a
release
and
we
were
dependent
on
some
arbitrary
random
version,
like
you
all-
wouldn't
release
based
on
a
snapshot
right
right
yeah,
you
would
expect
a
release
when
we
released
for
sure
yeah
and
because
we
work
together
really
tightly.
We
can
coordinate
that
stuff,
but
the
photo
is
like
like
there's
no
coordination,
there,
that's
a
good
point.
A
C
A
B
So
this
morning,
so
we
had
the
the
big
zero
ten
zero
announcement.
Nice
work
both
of
you
and,
of
course
questions
about
when
we
were
gonna
release
zero,
ten
zero.
So
you
did
you
hit
the
button.
What
happened
when
you
hit
the
button
on
a
rug.
B
Weird
yeah
where's,
oh
down
here,
hero's
logging
operation
was
cancelled.
Oh.
C
C
B
Work,
cool,
oh
yeah,
and
so
I
think
nikita
and
I
both
seem
in
favor
of
staying
in
sync,
with
whatever
the
java
repo
releases
are.
B
Yes,
yes,
thank
you.
I
know
I
I
have
oh,
I
didn't
copy
paste
from
my
regular
template,
so
I
didn't.
I
lost
the
the
weekly
checking,
but
yes,
you've
been
on
that
sergey
wanted
a
couple
things
mainly.
He
wanted
the
system
metric
so
that
that
got
in
on
the
baggage.
Propagator
thing
he
said
it
was
okay,
not
to
make
it,
and
I
was
still
chatting
with
carlos
on
that
issue
a
little
bit
just
to
clarify
john.
You
know
all
about
jaeger.
A
No,
I
that
was
that
was
carlos.
I
implemented
the
jager
eager
thrift
over
http
exporter.
A
That's
something
that
splunk
specifically
needs
and
I'm
sure
no
one
else
will
use
I
mean.
Maybe
somebody
will
use
it.
I
don't
know.
I
was
actually
just
chatting
with
someone
on
gitter
who
apparently
they
have
a
jaeger
installation
in
their
kubernetes,
but
that
none
of
the
grpc
ports
are
exposed
so
they're
having
to
use
the
zipkin
exporter
to
get
data
to
it.
A
B
A
A
B
So
vendor
distros,
oh
because
I
asked
I
asked
on
jason
keller
from
new
relic
because
he
had
been
asking
on
one
of
our
issues
about
distros
and
potentially
docking
some
stuff,
and
he
said
he
hadn't
gotten
to
it
yet,
but
soon,
but
also
nikita
has
to
put
together,
is
putting
together
a
example,
a
demo
with
all
the
different
kind
of
spi
integration
hooks
so
he's
he
had
brought
this
up
on
in
this
past
monday's
maintainer
meeting.
B
As
sort
of
do
do
we
want
to
open
telemetry
wide,
have
a
repo
that
could
have
sort
of
these
example,
distribution,
things
and
I'm
kind
of
interested
to
see.
If
it's,
I
feel,
like
part,
it's
a
little
bit
more
of
an
issue
for
our
repo
than
others.
Just
because
for
for
sdk
and
other
things
for
the
distros
are
more,
I
mean
just
for
vendors
versus
for
us
we're
sort
of
in
order
to
customize
various
things.
B
A
Yeah
he's
uncovering
lots
of
interesting
configuration
glitches
in
the
sdk,
which
is
good,
like
basically
everything
you
can
figure
in
the
sdk
there's
a
different
kind,
different
shape
of
api.
To
do
it.
So
we
solve.
A
B
Oh
yeah,
so
then
this
was
the
issue
that
jason
was
chatting
on
so
nikita's
gonna
post
there
we
talked
a
little
bit
about
datadog
and
sort
of
the
future
of
tyler's.
Still
I
think
tyler
would
like
to
get
datadog
aligned
into
the
open,
telemetry
repo,
but
I
think
there's
not
a
lot
of
other
interest
there.
B
I
think
they've
gone
off
into
some
different
routes,
and
so
that's
going
to
be
challenging.
Unfortunately,
so
he's
kind
of
looking
for
small
wins
of
like
something
smaller
that
we
could
that
they
could
share
of
ours
directly.
B
He
did
mention
some
of
the
the
improvements
to
the
datadog
repo
lately
have
been
around
some
context,
propagation
improvements.
So
I
mentioned
that's,
that's
actually
on
my
agenda
starting
next
week.
I
want
I'm
gonna
start
pulling
over
those
data
dog
prs
one
by
one
where
it
makes
sense,
because
I
have.
A
B
A
Yeah,
have
you
seen
any
of
these?
I've
had
a
whole
bunch
of
pull
requests
and
master
builds
fail,
probably
five
or
six
over
the
past
two
weeks
in
github
actions
on
ubuntu
with
psych
faults
like
the
job
like
jvm,
is
segfaulting.
A
Next
time
what
happens
nikita
was
asking
if
we
could
capture
at
least
capture
something,
and
then
he
was
suggesting
we
used
gradle
build
scans
and
ci
made
to
be
able
to
look
at
that.
C
A
A
B
So
see
if
it
generates
a
one
of
those.
A
B
What
I
would
do
is
once
you
find
if
you
see
that
it's
generating
that
and
you
see
where
it
is,
then
in
the
build
script
at
the
end,
you
could,
you
know
either
cat
it
dump
it
to
console
or.
B
Yeah,
because
those
are
pretty
good
at
least
for
narrowing
down,
you
know
like
the
most
common
reason
that
I've
gotten
those
in
the
past
has
been
memory
just
out
of
physical
out
of
out
of
memory
and
no
swap
space
well,.
A
B
B
Nikita
asked
folks
if
they
care
to
review
the
after
ga
stuff
since
we're
prioritizing.
We
did
end
up
john
spinning.
We
did
a
15
minute
time
box
last
night
and
went
through
you
know
maybe
10
or
15
issues
well,.
B
Yeah
and
then
the
weekly
digest
I
hadn't
prepared,
so
I
went
through
just
the
prs
covered
all
like
half
of
the
prs.
Were
you
updating,
snapshots
and
yeah?
I
think
that.
B
Was
yeah
still
yeah,
so
still
good
participation
from
our
main
vendors?
I
am
very
I
mean.
Would
it's
would
wish
that
we
could
get
datadog
on
the
on
the
repo?
But
at
this
point
I
mean
that
they're,
I
think
they
they
just.
B
A
A
Think
bogdan
patted
his
stats
for
like
about
like
seven
last
minute,
prs
right
before
release.
B
I
haven't
given
you
grief
about
this
one,
yet,
let's
find
the
blog
blog
post,
oh
just
for
fun,
the
the
release
candidate,
the
tracing
release,
candidate
blog
post,
does.
B
It
does
and
it
it
it's
a
major
development
support,
open
telemetry
in
order
of
commits,
so
we've
got
splunk
microsoft,
google
white
set
dynatrace
new
relic
info
stellar
investor
is
even
ahead
of
amazon's
way
down
here.
So
I
I
thought
for
sure
somebody
from
amazon
would
like
would
give
you
grief
over
this.
A
You
were
you
so
does
infoseller
still
exist
as
a
company,
it
does
okay
I'll,
be
wondering
if
you
could
somehow
or
you
could
like
register
a
new
domain
domain
name
like
register
a
domain
name.
It's
like
infostellar.io
or
infoseller.I,
don't
know
something.code
and
like
start
a
company
and
show
how
well
they're
contributing
you
could
probably
raise
some.
You
could
probably
get
some
decent
funding.
Get
some
finance
just
from
that
exactly.
B
Yes,
yes,
so
john,
I
mean
you
probably
know
everything
about.
I
don't
know
at
the
yesterday's
meeting
already,
I
think
john
you
gave
us
update
on
on
slack
or
something
yeah.
A
Trying
to
remember
if
there's
anything
super
important
in
there.
Oh,
I
guess
it
sounds
like
you
know,
I
could
you
and
I
probably
need
to
make
some
sort
of
decision
about
change,
log
management
and
whether
we
want
to
try
to
propose
enforcing
some
sort
of
protocol
for
making
it
easier.
I
mean
it's
mostly
me,
I'm
the
one
who
generates
the
gosh
darn
thing,
but
it
sure
would
be
nice
if
I
could
grep
for
something
in
the
commit
history
or
in
the
merge
history
in
the
git
log.
C
A
A
A
Is
the
title
and
then
the
merge,
like
the
actual
comments
on
the
commits?
Really
not
the
description
I
didn't
know,
no,
the
description
doesn't
show
up,
which
is
a
bummer,
because
that
would
be
great
if
it
was.
If
that
would
show
up
somehow
maybe
there's
a
get
api
github
api.
We
could
use
to
pull
that,
but
the
regular
doesn't
include
it.
A
C
A
C
A
But
also
one
thing:
that's
interesting
about
so
you
made
a
comment
here:
the
change
log,
but
no
end
user
is
going
to
care
about
that
at
all.
I
don't
think
so.
One
of
the
things
is
that
I
think
is
kind
of
important.
If
it's
going
to
be
useful
is
make
sure
that
the
change
log
entries
are
actually
end
user,
relevant
and
less
maintainer
or
contributing
relevant
and
then
again
it
all
gets
tricky
right,
but.
A
C
I've
never
particularly
liked
this.
At
the
same
time,
I've
never
done
that
much
changelog
management
before
but
yeah.
I
think
so
yesterday,
when
I
brought
this
up,
nikita's
idea
was
that
we
might
do
more
automation
in
the
future
and
for
now,
as
long
as
we
have
that
thing
in
the
committed
like
it's
an
instrumentation,
we're
thinking
just
actually
being
able
to
copy
paste
from
the
pr
descriptions.
A
If
there
was
a
way
we
could
do
it
yeah
I
mean
having
it
if
it's
in
the
pr
descriptions,
that's
also
something
that
maintainers
can
edit
right
before,
like
we
can
edit
exactly
and
we
can
add,
change
log
entries
to
them
yeah.
I
wonder
I
should
look
into
github
apis,
there's,
probably
a
way
to
pull
all
the
pr's
between
something
and
you
then
maybe
grab
for
one
thing:
it's
when
it's
multi-line
like
this,
it's
going
to
be
harder
to
grip
for
harder
to
automate
so
something
to
ponder
anyway.
I
don't
know
one
line.
B
C
A
C
A
To
express
that's
how
how
I've
been
feeling
and
why
it's
not
willing
to
just
kind
of
do
the
grunt
work
yep
yeah!
It's
not
that
big
a
deal
yeah,
I'm
getting
paid
for
it.
So
you
know
at
the
end
of
the
day,
it's
not
the
end
of
the
world.
You
know
it
takes
time
same
time,
away
from
rewriting
the
metrics
sdk
yeah,
so
win-win.
B
A
B
A
For
10.00,
but
also,
unfortunately,
only
maintainers
can
see
that
which
is
a
bit
of
a
bummer
like
it's
really.
It
feels
like.
Maybe
I
should
just
put
on
my
to-do
list
like
once
a
week,
or
maybe
maybe
once
maybe
even
once
a
day
just
put
in
a
changelog
pr
with
the
stuff
that
hap
has
happened,
that's
relevant
it
just
could
like
have
the
unreleased
section
of
the
changelog
actually
be
updated
fairly
regularly.
A
B
C
A
A
Cool
hey
interesting
side.
Note
I
so
I
put
in
the
art
update
of
all
the
examples
for
oteno
found
at
least
a
couple
cases
where
I
don't
think
we
are
actually
propagating
the
context.
The
way
that
we
recommend
that
we
should
so
that
part
of
that
update
was
to
get
the
get
things
a
little
bit
more
in
line
with
how
we
should
be
doing
context
propagation.
A
Who
merged
did
I
merge
it?
Maybe.
A
Oh,
no,
I
got
it
all
done.
I
got
it
all
working,
but
there
were
yeah.
There
were
a
couple
cases,
because
there
are
some
client
server
examples
and
was
trying
to
debug
why
propagation
wasn't
working
after
I
updated
o10
and
realized
that
things
weren't
exactly
the
way
that
we
had
wanted
been
wanting
to
document
and
then
turned
out
that
the
actual
issue
was
that
the
sdk
doesn't
have
propagator
or
the
api
doesn't
install
propagators
by
default
anymore,
and
that
was
the
actual
problem.
A
A
A
C
C
C
A
C
So
interfaces
versus
classes,
I
feel
like
interfaces,
give
us
more
flexibility
and
that's
nice.
C
A
Well,
a
map
is
not
pure
data.
Right.
There's
lots
of
different
ways:
you
can
represent
a
map.
You
just
have
something.
That's
like
four
fields
like
span
context.
A
There's
quite
a
bit
of
functionality
in
attributes
and
labels.
Absolutely,
whereas,
like
span
context,
is
just
for
fields
like
it's
just
data
and
that's
where
to
me
I
mean
again.
I
don't
feel
so
strongly
about
this,
but
this
is
what
my
gut
tells
me
is
if
it's
a
pure
piece
of
data
like
it's
a
pure
piece
of
data
like
we
don't
know,
overkill.
C
A
I'm
happy
to
get
rid
of
it,
but
if
you
make
it
an
interface,
then
you
really
can't
get
rid
of
it
right
like
if
you,
but
if
you
have
it,
just
be
pure
data,
it's
easy
to
get
rid
of,
because
you
can
just
deprecate
it
and
you
people
use
it.
Then
they,
you
know,
there's
no
harm
right,
that's
the
same
for
interface,
not
if
people
have
built
all
sorts
of
crazy
functionality
into
that
interface,
then
you
can't
then
like,
if
like,
for
example,
what
what
tyler
wants
to
do
is.
A
A
B
A
A
B
A
I
actually
have
a
different
opinion
on
that
now.
My
opinion
has
changed.
I
think
that
we
should
peel
off
the
stuff,
that's
related
to
w3c,
trace
context
and
put
that
into
its
own
thing
that
we
only
put
into
the
context
and
not
into
the
span
span.
I
mean
having
this
vanity
trace
id
like
that,
belongs
in
the
span
that
the
trace
date
and
the
flags
that
belongs
in
something
else
that
shouldn't
be
in
the
span
at
all
context,
yeah
like
why
did
that
stuff?
A
A
A
B
So
what
about
for
the
agent
use
case.
A
Yeah,
the
asian
use
case
is
a
whole
other
can
of
worms,
and
this
is
where
I
think
getting
rid
of
getting
rid
of
spam
context
merging
the
pieces
that
need
should,
in
the
span
the
beat
and
taking
the
rest
of
it
out
into
its
own
hunk
of
context.
That
looks
like
baggage.
For
example,
it
looks
like
something
like
baggage.
Just
a
piece
of
data
that
can
be
propagated
would
solve.
I
think
both
those
problems
right,
you
wouldn't
have
to
have
a
little
little
spam
yeah.
I
mean,
of
course,
but.
B
Tried
the
what
we
discussed
last
night
was
that
that
I
would
take,
because
bogdan
wanted
to
see
sort
of
doc
doc
documentation
on.
Why
we
need
this.
So
I'm
going
to
take
the
the
design
dock
from
way
back
and
turn
it
into
markdown
and
publish
it
in
our
repo
and
then.
B
I'll
come
back
and
post
that
and
get
bogden's
feedback
and
because
he
seemed
he
at
least
he
seems
to
be.
You
know:
okay,
we
can
make
concessions
for
the
agent.
If
I,
if
they're
you
know,
but
let's
document
them
and
clearly
understand
that
they're
needed
and
and
yeah.
A
So
am
I
my
what
I
think
about
the
interfaces
on
those
data
is,
I
don't
feel
I'm
not
gonna
like
if
everyone
else
agrees,
I'm
not
gonna,
certainly
not
stop
it
from
happening.
Just
what
my
usual
design
gut
tells
me
is
that
if
it's
pure
data,
just
let
some
people
use
some
data,
it's
just
the
same
thing
like
when
people
are
writing
unit
tests
and
there's
a
hunk
of
data
and
they
mock
it
like
no,
don't
mock
the
data,
just
create
some
data
and
use
it
in
your
test.
B
Yeah-
and
I
I
I
would
tend
to
I
would-
I
would
agree
with
you
and
I
I
hate
to
affect
the
api
design
based
on
agent
needs,
but
it's
also,
but
you
know
it's.
A
Yes,
so
I
think
I
on,
I
honestly
will
favor
what
you
need
over
end
use
end
users
more
because
you're
the
I
mean
you're
providing
the
most
like
the
most
value,
and
I
would
also
say
I
had
something
else
I
wanted
to
follow
up
with,
which
was
crap.
It's
gonna,
be
totally
in
your
favor.
A
Oh,
I
mean
bogdan
seems
super
super
concerned
with
people
doing
crazy
stuff
with
the
interfaces
and
like
providing
breaking
implementations,
and
my
opinion
on
that
is.
If
they
do
that
and
then
in
the
future,
it
breaks
it's
kind
of.
C
A
Right,
yeah,
so
okay,
but
also
like
ogden's
example,
is
like
because
he
kept
on
mentioning
this
I'm
like.
Can
you
give
a
concrete
example
and
then
his
concrete
example
sounded
like
a
place
where
they
were
casting
to
the
implementation
yeah
their
sdk,
I'm
like
well.
Of
course,
crap
is
going
to
break
if
you've
lost
your
implementation
yeah.
I
think
that's
not
a
good.
A
A
B
No,
no,
it
didn't,
but
it
I
think.
Oh
that's
not
this
one.
B
B
A
It
feels
funny
that
that's
a
prime
example
of
someone
who's
never
maintained
an
agent.
B
A
It
would
also
be
really
good,
like
I
said
if
you,
if
you
had
like
an
example
like
if
here's,
what
that
bridging
has
to
do.
If
you
have
an
interface,
here's
how
the
bridging
has
to
work.
If
you
don't
have
an
interface,
here's
how
the
bridging
works,
if
you
have
an
abstract
class
like
that
sort
of
thing,
I
think
would
be
really
good
to
provide
some
ammunition
on
why
interfaces
would
be
preferable.
B
A
B
Yeah,
you
know
it's,
it's
there's
they're
intrigued.
I
think
they're
like
okay.
Let's
let
trask
go
off
and
see
if
he
can
make
this
successful
and
then
maybe
we'll
buy
into
it,
but
yeah.
A
It's
funny
like
now
that
I'm
in
splunk
john
bly
is
100
like
a
thousand
percent
in
the
auto
instrumentation
camp
he's
like.
Why
would
anyone
do
anything
about
this
orientation
like
he's
even
he's
way
more
gung-ho
than
I
have
anyone
seen
even
like,
even
than
you,
because
I've
always
been
in
the
camp
where
I'm
like
I'm
happy
to
manually,
install
all
my
stuff,
but
I'm
also
not
a
normal
end
user.
I'm
someone
who
has
very
narrow
requirements
about
what
I
care
about
and
yeah,
so.
B
Oh
yeah
yeah,
I
mean
I
I
mean,
and
you
know
honestly,
for,
on
the
microsoft
side,
our
customers
aren't
at
least
when
it
comes
to
monitoring.
If
they
want
something,
you
know
more
advanced
they're,
probably
looking
at
splunk
or
new
relic
or
datadog
like
they
seem
to
be
the
ones
that
I
work
with.
They
seem
to
be
pretty
excited
to
just
get
some.
You
know
basic
monitoring,
you
know
for
free,
I
mean
not
free
money
but
free
time.
Yep.
A
B
Yeah,
I
know
anurag
wanted
to
chat
real
briefly
about
builders.
A
Weekend
yeah,
I
mean,
with
with
our
overall
app
there's
only
like
one
weekend
day
right,
yeah
everything
else
is
not
not
necessarily
a
weekend.
A
Yeah,
it
looks
like
you
might
need
a
rebase
yeah,
I'm
sure
everything
needs
a
rebase,
always.
B
A
B
I
mean
we
saw
the
the
pr
list,
though,
when
you
had
70
prs
in
the
last
month,
but
that
also
is
a
part
of
why
you're
probably
rebasing
against
yourself
half
the
time.
A
Yeah
I
mean
I
like
that.
I,
like
the
top
lover
top
level
builders
personally,
but
it's.
B
You
could
like
put
a
comment
saying
planning
to
merge
tomorrow.
Any
last
last
call
last
call
for
objection.
B
But
yeah
I
would
say
with:
if
you
have
you
and
john,
I
don't
see
why
I
mean
I
think
somebody
would
have
to
come
with
a
pretty
strong.
You
know
objection
is
my
take.
A
C
Yeah,
like
the
only
like
the
biggest
change
here,
is
having
to
move
that
upper
class
to
the
internal
package.
So
it's
accessible.
That
was
a
bit
annoying,
but
I
guess
it's:
okay,.
A
Yeah
yeah,
I
did
notice
that
as
well.
I
don't
think
that's
a
big
deal,
though,
that
seems
fine
yeah
that
that
thing
I
hacked
together
apparently
is
getting
a
lot
of
use.
I
did
want
to
chat
a
little
bit
about
so
that
I
don't
know
if
you
saw
the
whole
ordered
attributes
in
order
thing.
The
spec
and
I
put
in
so
the
span
attributes
are
in
order
now
so.
B
A
A
B
A
And
so
I
was
trying
to
figure
out
if
people
get
really
antsy
about
the
fact
that
we
don't
preserve
insertion
order,
how
would
we
or
would
we
or
we
just
basically
would
we
fight
to
the
death
over
it?
But
how
would
we
alter
that
immutable
key
value
pairs
to
preserve
insertion
order?
A
I
remember
looking
at
the
the
way
that
is
it
brave
as
an
equivalent
and
they
actually
never
remove
anything
at
all
from
the
array
they
just
like
tag
it
with
like
they
have
a
separate
like
bit
array
with
things
that
are
invalid
have
been
removed,
and
so
I
wonder
if
we
could
potentially
do
something
more
like
that,
but
I
don't
know
if
you
had
any
ideas
on
oregon,
how
we
would
preserve
insertion
order
if
we.
A
A
C
A
C
B
I
would
think
for
semantic
I
mean
for
well-defined
semantic
attributes
ordering
should
be
is
totally
irrelevant.
The
only
where
place
I
can
think
of
is
like
arbitrary
key
values
I
mean
like
in
glow
root.
I
did
preserve
the
ordering
of
custom
attributes
so
that
you
could,
because
it
was
reflected
how
I
displayed
it
in
the
ui,
and
so
you
could,
you
know,
put
things
in,
but
in
our
case
I
mean
everything
is
really
supposed
to
be
these
sort
of
well-defined
semantic
attributes
that
back-ends
understand.
C
A
A
B
C
C
A
C
A
B
A
That's
the
whole
thing
like
if
you're
writing
instrumentation
like
yeah-
I
just
I
don't
understand
the
ordering
it
doesn't
make
sense
at
all
yeah
anyway
I'll
I'll
go
ahead,
I'll
put
in
at
least
an
issue
into
the
specs
to
see.
If
I
can
get
people
to
talk
about
it,
or
maybe
I
can
dig
in
and
explain
why
or
try
to
find
the
pr
where
that
got
put
in
and
maybe
there'll
be
some
sort
of
discussion,
probably
not
anyway
I'll
at
least
put
in
something.
C
B
Oh,
I
did
I
found
where
you
know
we
were
talking
about
you
all
having
admin
rights
in
the
java
repo.
A
C
C
B
Right
for
me
well,
thank
you
for
debugging,
whatever
that
bin
tray
upload
turns
out
to
be.
C
C
A
C
A
A
B
We
we
passed
a
pretty
cool,
decrement
decriminalization.
We.
A
Is
I
was
asking
her
if
she's
gonna
start
she's
like
no?
I
want
nothing
to
do
with
that.
My
practice
is
gonna.
I
will
refer
people,
but
I
that's
something
medical
marijuana.
She
will
refer
people
with
it's,
not
just
not
what
she's
interested
in
dealing
with
so
but
yeah.
This
also
I've
been
for
medical
reasons
is
super.
That's
a
fantastic
thing
that
we've
passed
on
is.
A
Yeah
that
one's
for
medical
reasons,
the
other
one
we
passed,
was
basically
decriminalizing
small
amounts
of
possession
of
most
almost
all
schedule.
One
drugs-
I
think
I
don't
know
it
was
meth.
I
don't
think
methamphetamine
was
in
the
list,
though
right.
B
The
hard
like,
basically,
your
your
addictive
hard
drugs.
A
B
A
And
she
was
like,
oh.
A
B
Well,
we,
you
know,
people
are
get
getting
used
to
now
like
after
marijuana
and
like
nothing,
horrible
happened.
So.