►
From YouTube: 2021-10-26 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).
A
Hello
different
view
of
amy:
hey,
you
have
a
white
dog.
A
B
A
B
A
B
A
Yes
and
then
I
I
I
would
be,
I
would
have
already
singed
my
eyebrows
by
that
time.
D
Buddy
not
much
trying
to
figure
out
why
my
tests
are
broken,
so
I
got.
B
You
know
just
delete
the
tests.
It'll
be
fine,
yeah,
it's
just
yeah
these
newfangled
testing
stuff
that
the
kids
are
doing
these
days.
It's
just
you
know
it
makes
everything
you
gotta
fix
them
all
the
time
and.
B
D
D
E
A
I
don't
know
doing
okay,
I
can't
I
kind
of
feel
like
my
my
week
is
incomplete
without
seeing
all
of
you,
so
I'm
so
excited
to
be
here
on
a
tuesday,
that's
exciting.
D
Yeah
and
how
are
you
yeah,
I
don't
know
it's
supposed
to
be
a
new
nor'easter.
I
don't
know
if
any
of
y'all
are
on
the
east
coast,
but
so
I
like
woke
up
early
and
I
was
like
I'll
go
to
the
supermarket
and
get
all
the
like
the
things
and
it's.
B
A
D
Marketing
in
weather,
but
yeah.
E
A
B
D
I
and
my
sister's
in
berlin
game
and
we
were
facetiming
with
her.
You
know
the
family
and
there
was
like
you
could
see
on
the
patio
like
stuff
like
flying
by.
I
guess
it
was
really
windy
like
yeah.
Maybe
I
should
like
shelter.
E
B
A
It
does
and
it's
severe
like
these
are
my
my
least
favorite
kinds
of
storms,
where
it's
like
severe
thunderstorm
warnings
overnight
when
it's
like
three
o'clock
in
the
morning
and
then
my
house
just
starts
to
rattle
like
what
is
you
know
are.
Is
my
house
gonna
get
blown
away?
I
I
just
get
so
I'm
you
know,
I'm
a
scaredy
cat
is
is
how
it
is.
You
know
what
I'm
saying.
B
A
A
The
the
last
storm
my
wife's
phone,
like
the
the
weather
service,
the
national
weather
service,
gave
her
the
alert,
tornado,
eminent
sleek
shelter.
I
stopped
and
I
said,
wait
a
minute
check
your
gps,
make
sure
it's
like
her.
Gps
was
like
targeting
her
like
30
miles
away,
and
I
was
like,
I
think,
we're
okay
for
right
now,
but
I
still
panicked
a
little
bit.
So
we
just
like
grabbed
all
the
bike
helmets
and
like
a
bunch
of
pillows,
threw
them
in
the
bathroom
for,
like
you
know,
one
time
that
happened
to
us
overnight.
B
A
And
I
I'm
gonna
keep
talking
about
the
weather.
If
you
let
me
matt
or
robert,
but
matt
so
indianapolis
polite
or
he's
not
going
to
stop
me
right.
B
F
C
It's
it's
fine!
It's
like
eight
degrees
celsius.
It's
nice!
I
always
like
describing
people
to
people
who
have
never
been
there
how
cold
it
is
because
it's
hard
to
get
a
frame
of
reference
once
it
gets
to
like
-45,
which
is
really
common
here.
You're
like
how
cold
is
that
I'm
like?
Well,
if
you
go
outside
and
you
leave
your
ears
uncovered
for
about
two
to
five
minutes
you'll,
your
skin
will
die.
B
B
C
C
No,
no
well,
it
does
like
the
sun
does
go
down
at
like
noon
or
whatever,
but
winnipeg
actually
gets.
I
think,
the
most
days
of
sunshine
next
to
calgary,
which
is
the
leading
spot
in
all
of
canada.
I
think
so.
We
actually
do
it's
freezing,
but
it's
clear
skies
and
sunshine.
So
you
get
that
nice.
Snow
blindness.
B
F
Yeah,
I
grew
up
in
michigan
and
in
middle
school.
I
I
was
far
enough.
I
was
supposed
to
ride
the
bus,
but
I
would
walk
and
I
walked
home
one
day
it
was.
It
was
a
below
zero
day,
whatever
that
is
below
zero
fahrenheit
and
ended
up
with
very
frozen
ears.
By
the
time
I
got
home
and
it
it
was
the
worst
thing
ever
like
as
they
thought
out.
F
The
pain
was
unbelievable
and
just
for,
like
I
don't
know,
probably
at
least
a
couple
of
weeks
after
the
fact
it's
like
they
were
painful
to
the
touch.
I'm
I'm
probably
lucky
that
I
still
have
ears
to
be
honest,
and
the
thing
that
I
learned
is
that
there's
just
like
absolutely
no
way
to
sleep
on
a
pillow
without
your
ears,
touching
it
in
some
way
so
bundle
up
if
you're
in
cold
weather.
D
B
D
B
F
F
Yes,
cool,
does
everybody
know
each
other
I
see
amy
is
here,
I
haven't
met
you
before,
I'm
mad.
I
think
I
think
I
have
seen
you
on
the
agenda.
Maybe
we
kind
of
missed
some
weeks
so.
B
Yeah
yeah,
I'm
I'm
fairly
new,
I'm
amy
toby,
I'm
a
principal
engineer
at
equinix
medal
and
we
just
deployed
open
tomotry
ruby
in
our
main
production
stack
last
week
and
it's
all
running
very
smoothly
now.
So
thank
you.
Everyone
awesome.
F
Cool
yeah,
if
anybody
else
does
not
know
amy,
feel
free
to
introduce
yourself.
C
G
G
A
B
F
All
right
there
we
go
so
I'm
going
to
jump
in
this.
The
first
part
of
the
agenda
actually
happened
at
the
end
of
the
meeting,
so
I'll
kind
of
cover
in
order.
But
there
was
a
quick
reminder
that
there
is
a
the
gc
election
is
going
on.
It
closes
on
the
27th.
F
So
if
you've,
if
you're
a
member
of
standing,
which
I
think
means
you've
made
at
least
like
one
contribution
over
the
last
year
or
something,
then
you
you
should
be
able
to
vote,
and
you
should
vote.
F
The
next
thing
is
somewhat
interesting,
but
basically
there
is
in
the
w3c.
There
is
a
spec
for
specs,
oh
and
it
kind
of
yeah.
It
gives
you
some
guidelines
on
how
to
write
a
spec,
and
the
main
points
about
this
is
that
specs
generally
have
like
normative
information
and
informative
information.
The
normative
stuff
are
kind
of
like
the
requirements.
These
are
the
things
that
you
absolutely
must
do,
and
the
informative
is
just
like
elaborating
on
that
stuff
to
some
degree.
F
This
guy,
diego,
is
interested
in
having
us
write
our
specs
according
to
that
w3c
spec,
and
the
reason
is
that
he
has
written
a
script
that
can
basically
parse
out
all
the
requirements
into
like
json
or
other
things,
and
you
could
conceive
of
this
being
at
a
bare
minimum.
A
great
cliff
notes
to
a
long
spec,
but
even
possibly
further
just
like
a
way
to
kind
of
like
make
sure
that
you
have
full
like
test
coverage
on
your
spec
and
in
this
pr
he
updated
the
baggage
spec
to.
F
B
So
it's
really
cool,
though,
being
able
to
generate
a
specification
that
you
can
use
to
validate
the
spec
itself
and
probably
some
downstream
language
stuff
like
being
able
to
generate
like
just
a
simple
here's.
All
the
api
calls
kind
of
thing.
I
wonder
how
hard
that
would
be.
F
Yeah-
and
I
think
that's
a
lot
of
the
motivation
behind
this-
is
all
the
stuff
that
you
could
generate,
so
I
feel
like
the
in
some
ways.
The
headings
have
some
significance.
Don't
quote
me
on
that,
but
definitely
like
these
requirement
headings
are
important,
so
I
think
the
requirements
end
up
being
the
normative
stuff
and
then
there
are
like
even
kind
of
like
these
conditional.
F
It
takes
this
kind
of
like
condition
to
conditional
requirement
kind
of
format,
and
that
can
be,
I
think,
parsed
into
some
kind
of
a
conditional.
Ultimately,
so
it's
interesting,
I
think
that.
F
B
It's
kind
of
cool,
though,
because
I
mean
the
internet's
built
on
w3c
specs,
and
so
they
figured
out
a
lot
of
the
sharp
edges
of
like
getting
people
to
conform
to
the
same
specification
and
doing
it
across
this
huge,
complex
industry.
With
a
lot
of
competing
interests,
it
seems
appropriate
a
lot
of
work
for
the
maintainer,
so.
F
Yeah,
I
think
in
general,
like
there
are,
I
don't
know
if
you
read
through
the.
If
you
read
through
the
comments,
I
think
there
are
people
who
are
a
little
resistant,
maybe
just
because
it's
a
lot
of
work
and
want
to
know
what
the
payoff
is.
But
then
I
think
there
are
people
who
are
really
excited
about
this
and
kind
of
like
see
the
the
potential
and
j
mcd.
B
B
F
F
The
same
way,
this
is
true.
Yes,
this
is
true.
You
can
kind
of
see
that
he
has
these
requirements,
at
least
so.
B
F
F
We
have
talked
about
this
a
little
bit.
I
think
we
were
talking
about
this
a
lot
when
it
was
otep,
so
when
this
was
kind
of
being
when
the
ideas
were
being
incubated
to
actually
make
it
into
the
spec,
but
this
is
actually
kind
of
the
spec
work
for
probability
sampling
and,
as
we
have
talked
about
it,
there
have
been
some
different
schemes.
This
is
a
lot
of
it
has
to
do
with
the
context.
F
Propagation
was
a
sticky
bit
so
propagating
the
sampling
probability
between
different
services,
and
this
again
dovetails
with
w3c,
mainly
through
the
w3c
trace
context
format
and
some
proposed
changes
there.
But
ultimately,
ultimately,
what
this
is
is
a
trace
state
solution.
So
I
think
the
the
dream
is
that
we
can
get
enough
of
this
information
into
the
trace
parent
header,
which
is
a
lot
easier
to
kind
of
extract
information
from
and
a
lot
more
kind
of,
uniform
between
the
different.
F
Services
that
kind
of
support
the
format,
whereas
like
trace
state
kind
of
becomes
this
highly
customizable
area
for
the
tracing
system
for
each
vendor
to
use
so
putting
data
there
yeah
it's
always
kind
of
an
escape
hatch.
If
you
need
some
sort
of
functionality
for
your
tracing
system
and
are
using
the
w3c
format,
but
it's
not
the
best
spot
for
kind
of
shared
stuff.
G
F
Awesome,
I
think
I
think
those
would
be
welcome
to
aspirations.
F
But
yeah,
so
I
I
think
it
makes
sense
to
read
through
here
when
I'm
skimming
through.
I
think
the
things
that
I
was
noticing
is
that
yeah
there's
definitely
some
work
here
to
propagate
this
p
and
r
value
in
trace
state.
And
additionally,
there
are,
as
far
as
I
can
tell
like
two
new
samplers
that
need
to
be
implemented.
The
parent,
consistent
probability
based
sampler
and
the
consistent
probability
based
sampler,
and
these
are
different
than
the
the
parent
based
or
the
probability
based.
D
F
Yeah-
and
I
guess
maybe
the
other
thing
is
r
value-
is
interesting
in
order
to
make
it
as
compact
as
possible.
It's
it's
a.
F
It's
a
inverse
power
of
two,
so
you
just
kind
of
give
the
exponent
for
it
and
it
restricts
the
it
resists
restricts
the
sampling
rates
to
like
a
constrained
set,
but
it
allows
it
to
be
quite
compact.
I
think
this
is
really
more
interesting
when
it
comes
to
the
trace
parent
header,
because
I
think
that's
something
where
like
size
is
like
a
pretty
big
concern.
G
It
was,
I
think,
I
think
it
got
negotiated
to
be.
You
know,
divisor
by
two,
so
there's
something
that
could
fit
very
small
in
the
parent
when
we
were
thinking
it
needed
to
get
pushed
up
into
trace
parent
as
a
w3c
spec,
and
then
we're
like
that's
a
big
lift
and
we
kind
of
need
sampling
now.
So
it
came
back
down
into
trace
state
as
having
decided
to
keep
it
small
so
that
at
some
point,
if
it
does
move
up
into
the
parent,
it's
already
in
a
forum,
that's
palatable.
F
A
I
don't
recall,
processing
anything
from
the
trace
date
at
all
right
now.
For
other
context,
propagators
like
ot
trace,
where,
where
did
where
should
trace
date,
go
in
those
in
like
the
jager
propagator
and.
F
So
trace
date
is
actually
part
of
the
w3c
trace
context
spec
so,
which
basically
defines
your
your
two
headers
trace,
parent
and
trace
state
think
it
might
be
somewhat
ambiguous
whether
or
not
you
can
or
should,
mix
and
match
this
with
other
propagation
formats.
A
Oh
yeah,
it's
like
it
was,
I
I
feel
like
and-
and
I
don't
remember
this,
but
I
feel
like
tree
state
was
things
that
were
vendor-specific
that
would
go,
would
go
into
the
tray
state.
Headers.
F
Yeah,
it's
definitely
so
trace.
State
is
really
for
vendor
specific
information
that
is
needed
by
the
tracing
system.
I
think
that's,
like
kind
of
the
full.
F
The
full
caveats,
because
you
also
have
baggage
stuff,
which
is
for
tracing
data,
not
for
the
tracing
system
for
for
other
purposes,
maybe
for
the
app
or
for
the
user,
but
ultimately.
F
They
never
really
solved
this
problem
so
like
this
is
one
of
the
things
that
I
think,
as
these
formats
were
starting
to
make
it
into
the
wild.
I
think
people
realized
two
things.
One
is
that
it
would
be
great
if
there
was
a
standard
because
they're
all
kind
of
propagating
the
same
essential
stuff
for
a
trace
and
two.
F
There
is
kind
of
this
need
for
there
to
be
kind
of
like
this
vendor-specific
area.
So
I
think,
rather
than
have
everybody
go
implement
these
things
again
themselves
a
lot
most,
I
would
say,
of
the
players
in
those
different
formats
kind
of
participated
to
some
degree
in
the
w3c
trace
context.
Spec.
F
I
think
the
larger
goal
is
like
if
you
need
this
stuff
migrate,
to
trace
context
in
the
long
term.
D
When
I
was
at
davidoff
there
was
an
issue
with
that.
Sorry,
if
there's
crosstalk
there
was
an
issue.
A
D
C
D
Or
whatever
that
we
is
like
started,
adding
custom
headers,
and
then
that
grew
to
include
like
sort
of
investors,
and
it
was
difficult
for
end
users
because
they
kept
having
to
white
list
like
more
and
more.
I'm
sorry
started
having
to
allow
less
more
and
more
headers,
and
it
was
just
like
this
constant
cycle.
So
yeah
it's
be
nice
if
they
had
also
participated
in
w3
suspect,
but
they
didn't
so
anyway.
D
D
This
might
be
a
good
way
to
handle
it
and,
like
you
know,
like
I
know,
like
zipkin
and
jaeger
sort
of
like
in
their
export
formats,
they
say
like
here's,
you
know
a
pen
trace
state
is
like
a
spam
tag
or
a
spam
attribute.
You
know
like
and
that's
not
convention.
That's
just
like
a
couple
guys
were
like
this
is
what
we
want
and
no
one
disagreed
so
like.
If
you
push
the
ball
forward,
I
feel
like
in
practice.
F
Cool
yeah,
as
I
think
through
this,
I
think,
trying
to
set
a
trade
state
without
a
trace
parent.
It
could
get
weird
unless
you
are
controlling
like
the
headers
of
all
of
your
systems
and
you
you
can
kind
of
control
the
outcome,
I'm
just
thinking.
If
you
send
like
a
b3
header
with
a
trace
state
from
one
service
to
one,
that's
just
all
w3c
and
it
might.
B
B
F
Yeah,
I
think
this
is
a
good
question,
though,
and
I
don't
think
that
this
is
a
definitive
answer,
but
I
feel
like
in
general,
it
would
yeah
trace
parent
and
trace.
They
definitely
are
meant
to
go
together.
Mixing
and
matching
could
definitely
get
weird
in
the
real
world.
F
But
I
will
blink
it
out.
Yeah
there
was
this
messaging
semantic
conventions
working
group.
They
have
at
least
like
come
up
with
a
a
roadmap
and
they've
at
least
published
that.
So
this
is
something
that
you're
interested
in
feel
free
to
read
through
that.
I
know
that
group
is
still
meeting
and
trying
to.
F
Yeah
determine
all
of
the
the
conventions
around
messaging
instrumentation.
F
There
is
this
state
of
hotel
road
map
that
is
also
in
progress.
I
think
it's
an
interesting
document
for
people
to
look
at
and
it
probably
should
be
updated
for
things
that
are
that
are
wrong.
I
tried
to
update
this
thing
and,
like
it
led
to
some
disaster
of
like
unapproved
comments,
and
I
was
unable
to
kind
of
add
the
the
formatting
in
there.
F
And
then
there's
a
pr
for
specifically
adding
like
a
none
option
for
hotel
propagators
like
right
now.
None
is
kind
of
like
the
absence
of
of
an
actual
defined
value,
but
it
would
be
good
to
know
if
somebody
explicitly
wanted
none
as
as
their
option.
F
And
then
yeah,
so
there
was
a
a
quick
update
from
the
metric
sig.
F
I
think
a
lot
of
this
is
just
like
moving
various
things
towards
towards
feature
freeze,
and
I
don't
know
where
we
stand
in
terms
of
having
any
or
having
begun
work
on
this.
I
know
this
is
something
that
we're
interested
in
working
on,
but
I
still
feel
like
it's
something:
that's
maybe
we
haven't
broken
ground
on
just
yet,
but
key
highlights
they're
trying
to
get
the
sdks
back
to
feature
freeze
by
the
end
of
october.
So
I
guess
that's
in
a
couple
of
days.
F
Yeah
there
was
a
question
specifically
about
the
hotel.net
prometheus
exporter.
D
That
seems
good.
I
had
thrown
one
thing
on
the
ruby
specific
agenda.
D
If
other
people
have
to
discuss
this,
isn't
super
urgent
or
anything
just
wanted
consensus.
Your
thoughts,
one
of
the
contributors
they've,
contributed
a
few
times
before
they
contributed
in
our
spec
instrumentation
vr,
so
instrumentation
for
your
r-spec,
like
tests
themselves.
So
I
left
some
feedback.
I
think
one
it's
a
little
bit
like.
Is
this
a
place?
D
Should
it
live
in
open,
telemetry,
ruby,
like
is
this
a
thing
we
want,
which
I
have
opinions
on,
but
I'm
curious
what
y'all
think
so
yeah
that
was
kind
of
what
I
was
into
like?
Is
there
any
stan?
Basically,
there's
no
standards
around
like
semantic
conventions
for
test
frameworks
that
I
could
find
there's
some
work.
D
That's
been
done
around
this
and
sort
of
the
broader
context
of
like
vendors,
supporting
the
cicd,
instrumentation
and
sort
of
as
like
more
like
productized
stuff,
but
nothing
standard
in
open
telemetry
that
I
could
find
someone
from
aws
had
written
a
library
to
do
this
and
go
that's
like
pretty
cool,
but
again
like
nothing.
D
D
I
have
some
context
because
datadog
has
an
rspec
instrumentation
package
that
you
know
I
can
use
as
reference
and
I
had
reviewed
at
the
time
but
unsure
whether
it's
something
we
want
to
support
like
whether
we
want
to
basically
take
over
maintenance
like
sign
up
for
maintenance
up,
because
there's
this
implicit
burden
of
like
eventually,
if
the
specification
comes
down
we'll
have
to
like
you
know,
update
things
to
in
to
support
that
spec
and
we'll
probably
be
way
off
spec,
because
we
don't
know
what
we're
trying
to
hit.
D
So
it's
a
matter
of
like
yeah.
Do
we
wanna
basically
do
we
wanna
support
this,
and
you
know
some
people
here
who
you
know
like?
I
don't
know
if
we
have
an
internal
use
for
it
right
now,
but
some
other
folks
may
or
you
know,
may
have
interested
as
a
vendor.
So
yeah
I'm
curious
what
people
think
I
had
said.
I
can
review
it
this
week,
but
yeah.
I
don't
want
to
review
it
if
everyone's
like
well.
D
We
don't
even
want
to
touch
this
right
now,
so
I
think
it's
cool
when
we
should.
C
Yeah
my
two
cents
on
that
is
like
again
like
this
we've
talked
about
this
much
times
it.
This
is
like
something
that
would
make
sense
to
live
in
a
contrib
repo.
If
we
had
one,
I
still
don't
really
care
for
us
setting
up
the
country
repo.
C
Why
not
just
let
it
evolve
in
this
repo,
where
we
can
actually
pull
out
some
insights
from
it
and
understand
some
of
the
things
that
people
need
and
maybe
some
limitations
at
the
api
like
we
talk
about
like
the
instrumentation
like
spec
and
semantic
conventions
like
we're
not
at
a
1.0
for
instrumentation
or
we're
not
even
close
to
it.
So
there
aren't
the
same
stability
guarantees.
C
But
if
you
see
that
as
a
good
thing
again
like,
I
think,
I'd
rather,
I
think
it'd
be
better
if
it
evolved
in
the
repo
with
us
being
able
to
see
and
watch
the
changes
rather
than
it
happened
somewhere
off
to
the
side
out
of
sight
that
we
might
not
really
keep
up
to
date
with
the
caveat
is
like
in
general.
This
is
like
more
broader
point
like
for
pros
like
this,
where
none
of
us
necessarily
have
like.
I
don't
have
a
strong
opinion.
I
don't
really
have
a
strong
frame
of
reference.
C
What
the
best
choices
are
for
this
type
of
instrumentation.
We
also
don't
want
to
like
be
like
sand
in
this
person's
boots
right
from
preventing
them
to
get
this
through,
get
this
out
and
to
work
on
it
right.
We
need
to
maybe
decide
to
be
a
little
bit
more
like
loosen
the
reins
and
just
be
like
cool.
C
This,
like
makes
sense
like
we're,
trusting
you
chris
holmes,
has
tried
to
make
a
few
contributions
in
the
past
and
it
seems
like
there's
a
lot
of
good
intent
there
so
like
we
want
to
encourage
that
and,
like
you
know
what
I
mean,
I
think
it's
just
trying
to
pull
them
in
and
yeah.
I
think
this
makes
sense.
G
I'll
speak
as
in
as
an
existing
vendor
hat
and
say
I
know
several
of
our
customers
want
instrumented
test
suites
and
so
the
more
that's
out
there.
That
is
got
more
of
a
community
around
it.
So
I
I
it
appeals
to
me
that
I
can
contribute
to
this,
but
others
can
too
and
it'll
get
better
for
everybody.
G
First
thing
that
jumps
out
at
me
is
the
user
experience
of
using
this
is
going
to
be
unlike
instrumenting,
say
a
web
service,
so
we'd
want
to
twiddle
the
readme
and
maybe
include
an
example
in
here
of
how
do
you
instrument
your
test,
suite
which,
which
is
going
to
look
different
than
most
of
the
other
gems,
but
I
mean
I'm
also
game
to
help
with
that.
So
I'd,
let's
lean
in
on
semantic
versioning,
and
maybe
this
doesn't
go
1.0
until
there
are
some
semantic
conventions
around
the
spans.
G
C
So
I'm
getting
so
I
I'm
wearing
heavily
for
like
going
through
this
and
making
sure
that,
like
at
least
in
a
sense
of
like
implementation,
it
makes
sense
and
it
seems
reasonable,
but
being
a
little
bit
more
loose
around
the
fact
that
there's
no
semantics
for
this
and
just
kind
of
like
go
with
like
I
see
this
makes
sense,
is
anyone
leaning
in
the
other
direction?
Is
anyone
being
like?
Well,
maybe
we
don't
want
to
bring
this
in?
Maybe
we
don't
want
to
make
this
part
of
the
official
repo
is
there
like?
Are.
D
C
A
Yeah,
I
think
my
take
on
this
is
you
know,
for
we
don't
want
to
include
that
in
the
auto
in
in
the
in
the
all
auto
instrumentation
by
default
should
be
separate.
A
I
think
about
the
use
case
where
people
want
to
write
unit
tests
where
they
are
checking
what
they're
tracing
and
how
that
plays
into
when
using
our
spec
to
trade,
to
to
to
collect
data
out
of
their
test
suites
and
try
to
understand
the
instrumentation
of
their
test.
Suites.
Those
things
are
going
to
be
difficult
to
disambiguate
from
one
another
so
like
in
our
case
right.
You
know
we
write
tests
for
our
instrumentation
and
we're
like
hey,
look
check
the
spans.
Did
they
write
anything?
Did
we
generate
the
right
spans?
A
B
Yeah,
that
would
be
my
concern
too,
like
I,
it
seems
like
a
really
good
idea.
That's
something
I
want
to
throw
at
my
teams,
but,
like
you
said,
we
wanted
to
be
really
clear.
What
is
instrumentation
coming
from
the
code
under
test
in
the
test
system
itself,
but
that
seems
like
we
just
don't
want
to
cross
the
streams
right
like
have
different
different
routes
for
each
that
should
work
shouldn't
it.
D
Yeah
I
mean
it's
not
to
interrupt.
It's
like
people
do
tests
where
they
want
to
look
at.
They
want
to
use
like
almost
like
snapchat
testing,
where
they
want
to
look
at
like
spans
generated
in
a
test
and-
and
you
know
for
correctness,
and
then
people
want
to
use
test
wrapping
their
instrumentation
as
a
way
to
measure
flakiness
in
their
test
suites
itself
within
some
products,
such
as
circle,
ci
or
whatever.
C
Yeah,
I'm
trying
that
feature
so
the
way
I
I
think
about
this
is
that
in
the
case
where
you're
concerned
with
like
because
I'm
working
an
application
right
now
that
they
heavily
make
use
of,
like
spans
outputted
from
their
tests
to
assert
the
correctness
of
the
code
they're
testing.
But
if
we
were
to
instrument
the
test
framework
itself
again,
that'd
be
really
confusing,
but
I
think
that's
where
it
makes
sense
that
you
have
multiple
exporters-
and
this
is
a
good
use
case
for
potentially
a
different
tracer
provider.
C
So
you
know
you
would
have
your
twister
provider
and
your
exporter
for
your
what
is
being
tested-
and
that's
probably
like
in
the
in-memory
span
exporter
the
the
console
exporter,
the
really
simple
stuff,
but
because
you
actually
want
to
collect
this
data.
Your
trace
framework,
the
tracer,
that's
tracing
the
test,
suite
it's
probably
going
to
use
a
real
exporter,
that's
actually
being
admitted
to
a
real
back
end,
and
I
think
that's
how
we
disambiguate
the
two
and
like
we
already
have
support
for
that.
It's
just
to
make
sure
that
that's
documented
and
explained
so.
C
People
know
when
they
try
to
use
this
thing,
that
they're
not
just
getting
this
jumbled
mess
right,
but
there
is
also
just
like
wondering
about
how
this
fan
context
might
garble,
but
that's
another
thing
that
I
haven't
really
thought
through
yet.
So
there
are
a
lot
of
interesting
use
cases,
but
I
think
this
is
one
of
those
instrumentations
or
one
of
these
scenarios
that
comes
with
pointy
edges
and
say,
like
this,
isn't
necessarily
going
to
be
super
straightforward
right.
Documentation
is
going
to
have
to
come
with
a
hereby
dragon
sort
of
warning.
G
I
imagine
it'll
from
I've
only
skimmed
this
so
like
standing
from
afar
and
holding
my
thumb
up,
it
kind
of
looks
like
it's
something
that
will
get
implemented
in
the
rspec
formatter.
If
the
rspec
format
will
have
a
sort
of
a
specific
instance
of
the
tracer
and
the
in
a
exporter,
configuration.
B
B
G
This
is
implemented,
I
think,
as
an
r
spec
formatter
and
then
because
r
spec
is
set,
our
spec
is
feeding
the
formatter
timing
and
then
the
formatter
is
the
thing
that
would
crank
out
spans.
So
if
we
can
isolate
that
and
like
our
spec
formatter,
you
get
your
own
tracer,
you
get
your
own
exported
pipeline
config.
That
is
separate
from
whatever
the
test
environments.
G
Hotel
wiring
is
to
test
it's
to
test
the
the
app
the
the
program
under
tests.
Own
instrumentation
has
a
yeah
there's.
Some
brackets,
like,
as
somebody
said,.
D
Too,
not
to
raise
my
hand,
but
you
can't
really
like
not
sample
a
random.
You
know,
like
the
difference
between
an
unstamped
trace,
that's
dropped
and
a
trace
that
fails
in
a
test
when
you're
testing
like
when
you're
trying
to
use
traces
to
test
flakiness
is
like
really
critical
or
or
devalues
the
point.
The
whole
point
of
the
thing.
I
would
hope
that.
D
Not
I
think
these
are
all
solvable
problems,
so
all
I'm
saying
is
it's
just
like
the
list
of,
and
I'm
in
for
like
like
robert,
I
think
like,
let's
own
it
and
whatever
like,
if
we're
a
little
behind
the
spec
at
some
point
like
it's
fine,
we're,
not
1.0,
but
it's
just
like
that.
G
Raises
a
good
point
in
in
chat
that
it
could
be
implemented
as
a
serialize
out
and
then
send
the
traces
later.
B
That
would
make
the
problem
way
simpler
to
think
about,
like
just
the
six
of
us
sitting
here.
Trying
to
talk
through
this
makes
me
think
that
probably
end
users
would
be
in
a
worse
case,
because
we
deal
with
the
stuff
every
day
and
we're
having
trouble
talking
about
it
right.
D
D
You
need
special
processors
and
exporters
and
sampling,
and-
and
so
I
just
want
to
be
clear
that,
like
owning
this,
there's
like
this
yeah
accepted
thing,
but
as
long
as
we're
all
comfortable
with
saying
like
this
read
me
will
be
like
a
little
chunky
and
have
a
bunch
of
sections.
Then
I'm
fine
with
reviewing
and
merging
it,
and
you
know
working
with
it
and
fine.
I
find
myself.
D
G
We
could
make
a
generic
we
collected,
like
hotel
collectively,
could
make
a
generic
turn
junit
xml
into
traces,
and
then
our
spec
already
does
that.
And
then
we
don't
need
a
special
gem
that
we
have
to
like
cordon
off
a
config
for
it's
just
running.
G
B
Mind
is
that
that's
critical,
it
would
be
a
good
people
to
talk
to
about.
This
are
probably
there's
some
folks
working
on
tracing
extensions
for
jenkins
and
they've
been
doing
some
really
interesting
stuff.
We
could
ask
them,
but
they,
but
I
bet
they
have
really
informed
opinions
about
how,
because
they
spent
a
lot
more
time,
gathering
these
stats
and
sending
them
around
and
stuff
and
probably
would
be
consumers
of
the
spec.
So.
D
Unit
stuff
exists
in
the
data
dog
repo
for
context,
so
there's
an
implementation
of
it.
That's
out
there
that's
apache
too,
so.
G
There's
emitting
the
traces
for
like
the
test
suite
itself
and
then
there's
the
job
runner
that's
running
the
test
suite,
which
would
be
like
a
jenkins
circle.
Github
actions,
situation
where,
like
I've,
got
a
step
in
my
ci
pipeline,
that's
running
within
it
is
an
rspec
execution
that
is
going
to
poop
out
xml
and
then,
if
you
interpret
that
xml,
you
can
get
your
spans
for
this
test
suite
itself,
I
don't
know
I
having
been
endorsing
this
gem.
I'm
now
like.
G
D
Sounds
like
there's
broad
interest
in
not
to
like
try
to
do
the
wrap
up
like
that's,
not
my
meeting,
there's
broad
interest
in
this
and
it's
worth
worst
case.
We
review
it,
accept
it
and
then
deprecate
it.
If
we
want
to
have
some,
you
know
like
broader,
more
abstracted
solution,
which
is
actually
sounds
really
nice,
but
for
the
sake
of
keeping
this
like
within
the
open,
telemetry
open,
tell
em
to
what
is
wrong
with
me
ecosystem.
D
It's
I
ca,
I'm
happy
to
review
it
and-
and
I
think
roberta
mentioned
interest
as
well,
and
we
can
try
to
get
this
merged
cool
and
yeah
and
the
rest
will
figure
out.
C
One
last,
like
suggestion,
I
thought
that
I
don't
know
if
it
really
makes
sense
if
we
should
start
making
liberal
use
of
this
is
like
we
could
flag
the
version.
The
initial
version
is
experimental,
like
as
a
warning
to
people.
I
could
say:
hey
like
this
is
experimental.
I
think
that
warns
up
a
little
bit
further
discussion,
but
it's
just
an
idea
to
throw
up
there.
I'm
also
very
comfortable
saying
this
is
like
a
0.01.
D
F
I
agree
that
it's
like
not
probably
worth
the
effort
at
this
point
in
time,
but
if,
if
this
continues
to
keep
happening,
then
maybe
that's
a
sign
that
that
I
can
trip
repo
would
be
a
good
place,
but
as
long
as
this
doesn't
get
installed
by
default
and
your
test
suite
starts
sending
spans
to
a
extremely
costly
apm
provider,
then
then
I
think
it's
fine.
C
F
Cool
yeah
is
there
anything
else
we
should
talk
about?
I
know
I
usually
hand
this
off
to
robert
to
tell
us
about
the
instrumentation
seg.
C
A
Yeah,
but
I
put
something
on
the
agenda
also
just
a
chit
chat
about
one
of
the
submissions
that
we
got
for
an
update
to
ethon
and
I
had
requested
that
the
person
create
a
you
know.
I
had
some
automated
test
coverage
and
what
that
turned
into
was
a
test
that
tries
to
call
outside
example.com
and
takes
like
30
seconds
of
timeout
or
whatever.
A
So
I
made
a
follow-up
suggestion,
which
was
hey
the
ethon
gem
test.
This
is
that
they
run
a
local
server
inside
of
that
specific
test
and
make
calls
to
it
locally.
So
it's
faster,
I
suppose-
and
so
you
know
looking
at
that
more.
It's
like
oh
there's,
a
lot
of
machinery
to
try
to
make
sure
that
this
particular.
A
Change
is
covered
because
it
looks
like
there
wasn't
any
other
test
coverage
that
was
similar
to
this.
I
guess
to
try
to
cover
this
specific
use
case,
and
so
I
wanted
to
get
folks
feelings
about
how
much
energy,
how
much,
how
sophisticated
the
test
should
be
when
we're
expecting
submissions
from
folks
who
are
trying
to
fix
bugs
or
add
features
like.
C
I
have
a
book
they
did
so
maybe
we
can
like
just
quickly
peel
the
artery
thumbnails
like.
Is
there
a
reason
why,
like
we
can't
just
like
use
like
webmock
like,
why
did
it
have
to
make
an
actual
call?
Why
did
we
have
to
spend
on
the
server.
A
A
A
Using
lip,
curl
yeah,
so
there's
an
interceptum
every
other
library
at
some
point
uses
like
netted
tv
under
the
hood.
Somehow
it
doesn't.
D
Http
rb
use
little
curl,
one,
not
blanking,
sorry
for
interrupt
nah.
I
need
to
learn
this
handwrist
feature.
Sorry.
A
Hey
no
problem,
you
you
I'm
used
to
it,
come
on,
come
at
me,
bro,
oh
no!
So
anyway,
so
what
they
were
doing
was
reaching
out
to
example.com,
and
then
they
brought
the
fact
that
they
couldn't
somehow
control
the
timeout
right.
So
I
suggested,
maybe
you
should
talk
to
the
ethon
folks
and
figure
out
how
you
can
control
timeouts
as
far
as
library,
but
when
they
asked
the
next
question,
which
was
like,
as
things
get
more
complicated
like
like
right
now.
A
C
C
Yeah,
I
agree.
I
think
this
is
obviously
going
to
come
with
like
opinions
so
like
it's
subjective
integration
tests
are
better
because
they
test
that
things
actually
work,
but
that
being
said,
they
can
be
slower.
C
If
it's
adding
30
seconds,
like
I
think,
figuring
out,
the
timeout
should
just
be
the
call
here,
but
I
don't
know
how
closely
anyone
watches
our
ci
to
get
like
windows
or
mac
tests
going
takes
like
40
minutes
and
those
are
blocking
right.
C
So,
what's
30
seconds,
if
your
upper
limit
is
40
minutes,
so
I
know
it
is
a
crappy
answer,
because
that's
how
you
get
a
slow
test
suite,
but
we
already
have
an
incredibly
slow
test
suite
because
of
windows
and
mac
ci-
and
I
don't
know-
I
don't
know,
I
think,
just
like
them
figuring
out
the
time
it
was
the
answer
here
and
you're
pushing
the
right
solution
here.
C
I
think
because,
like
if
they
don't
add
a
test
that
asserts
the
behavior
they're,
correcting
it's
just
a
regression
waiting
to
happen,
someone's
going
to
look
at
it
and
understand
why
it's
there
be
like.
Oh,
I
don't
need
to
do
the
safe
navigator
on
finish
here.
I
can
just
it's
fine
like
this.
Obviously
someone
didn't
know
what
they
were
doing
and
it's
fine
now
right,
like
I
can't
recreate
this
or
the
ci
passes
and
yeah.
Now
you
just
have
the
same
problem
again.
They'll
look
silly,
so
I
I
think
you're
pushing
in
the
right
direction.
C
C
C
It's
possible,
but
if,
if
it
really
comes
to
know,
timeouts
don't
exist
in
this
networking
library.
For
some
reason
it
sucks,
but
I
still
let
it
be
slow
for
now
and
at
some
point
someone
will
get
really
frustrated
and
fix
their
ci
and
they'll
identify
this
as
a
slow
test,
and
that
will
introduce
progression.
D
B
A
F
Yeah,
I
think,
for
the
record.
None
of
us,
like
this
testing
framework
that
we're
using,
I
think
we
all
like
mini
tests
or
our
spec,
but
not
the.
C
Abomination
right
and
and
if
nobody's
happy,
then
everybody's
happy
and
I
think
that's
what
we
decided
on
last
time.
We
talked
about
this
like
a
year
ago.
I
know
over
time
I
want
to
self-shame
myself
here.
I've
been
bad
for
keeping
up
with
reviews.
D
C
Comments
and
looking
at
issues,
I
know
that
ariel
and
eric
have
been
pretty
good.
I
think
I
just
want
to
shame
the
rest
of
us
to
be
a
little
bit
more
active.
I
think
we
need
to,
but
mostly
me
just
because
I
know
that,
like
the
rsvpr,
for
example,
it
was
so
quiet,
they
weren't
sure
that
we
they
wanted.
We
even
wanted
the
pr
and
that's
I
think,
that
sucks
and
I'm
kind
of
frustrated
with
myself
on
that
one.
C
A
Robert,
do
not
be
hard
on
you've,
contributed
so
much
to
the
project.
So
far
we
wouldn't
have
gotten
this
far
without
you,
so
please
do
not
feel
like
because
you
missed
a
tr
or
two
or
a
question
or
whatever,
and
all
of
us
are
doing
what
we
can
they'd
be
doing
this
for
free.
So,
like
don't
yeah
I
mean
don't
be
hard
on
yourself
and
thank
you
for.
A
Prank
caller,
but
no
we
got
there's
a
lot
to
cover
now
right
as
new
instrumentations
get
added,
there's
too
much
to
cover
for
a
handful
of
people
so
but
still
don't
be
hard
on
yourself.
Man.
C
A
Yeah
for
the
proof
and
run
button
also,
we
essentially
just
need
them
to
sign
the
ezcla
right
and
then
once
they
do
that
we
can
do
a
proven
run.
There's.
A
Okay,
so
I'll
make
sure
to
do
that
in
the
future
as
well
and
friends,
thanks
for
your
time.