►
From YouTube: 2020-06-11 Spec SIG
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
B
D
E
A
So
this
work
currently
in
like
a
trial
period
where
I
guess
so
at
the
beginning
of
June
they
sort
of
opened
up
so
like
restaurants,
are
open,
but
you
can
only
eat
outside,
and
people
are
pretty
good
about
the
masks,
but
the
social
distancing
is
sort
of
fallen
by
the
wayside.
I
think
people
are
just
desperate
for
human
interaction
and
so
yeah,
we'll
see
we'll
see
I
think
they
have
another
period
in
two
weeks.
They
have
an
option
to
roll
back.
If
thing,
if
cases
have
exceeded
you
know
a
number.
E
A
C
D
D
F
F
E
A
A
E
E
D
D
For
writing.
The
exporter.
I,
don't
know
if
I
already
grabbed
that
one
I'll
take
a
look.
I'll,
definitely
assign
it
to
myself.
That's
fine
for
this
issue,
in
particular
the
declaring
the
trace
part
of
protocol
the
stable.
It's
two
outstanding
concerns
right
now.
One
is
the
spec
says
that
you
can
have
array
valued
attributes,
but
that's
not
represented
in
the
protocol
at
the
moment.
There's
a
breaking
change.
D
So
that's
the
the
other
one
that
needs
some
kind
of
decision
made.
So
once
we
have
approval
or
rejection
for
this
one
that
will
help
us
move
forward
and
then
the
other
one
Tigran
has
an
example
or
has
AE
proposed
PR
and
I.
Think
the
general
consensus
is
once
these
two
issues
are
sorted
out.
We
can
move
the
traces
portion
of
the
protocol
to
stable.
The
metrics
portion
will
remain
as
alpha
because
there's
still
a
lot
of
work
being
done
on
the
metrics
side
of
the
spectrum.
E
Yeah
yeah
I'm,
aware
of
both
of
these
issues
and
recently
just
opened
a
PR
over
an
open,
telemetry
j/s
that
adds
in
instrumentation
library
expands
the
way
it
was
working
with
the
collector
exporter
over.
There
was
actually
all
wrong,
so
it
at
least
made
it
a
little
bit
more
correct,
but
yeah
I've
also
found
this
issue
that
it
give
you
at
least
two
proposals
for
for
how
to
change
this
and.
E
B
E
D
D
You
know
if
you
have
a
bunch
of
spans
coming
from
one
instrumentation
library
that
are
all
together
in
a
batch,
then
you
can
just
specify
the
name
and
the
version
once
and
then
have
a
whole
bunch
of
spans.
We've
done
to
that,
and
then
the
backends
would
just
basically
do
exactly
the
same
mapping
as
proposal
1.
But
in
theory
it's
a
little
more
efficient
in
the
protocol
and
practice
is
probably
very
little
difference
because
you're,
probably
getting
gzip.
D
D
E
Yeah
yeah
I'm
I'm
with
you
on
all
these
things,
I.
For
some
reason,
I
did
kind
of
like
proposal
to
to
some
degree,
because
I
kind
of
feel
like
I,
don't
know
exactly
if
there
are
like
huge
benefits
or
use
cases
around
this,
but
I
could
see
you
wanting
to
pass
just
a
a
set
of
common
attributes
to
your
names
tracer
when
you
get
it
for
things
that
you
want
stable
and
every
span
coming
off
of
this
tracer.
In
addition
to
instrumentation
library.
E
Information
but
like
I,
can
also
see
this
being
a
completely
imaginary
feature.
Some
things
just
sound
good
and
you're
in
your
head,
or
they
seem
like
they
would
be
useful
and
you
add
them
in
and
nobody
ever
uses
them
kind
of
thing.
So
I
could
also
see
it
falling
in
that
boat,
but
having
written
code
that
takes
a
batch
of
spans
and.
E
Gets
it
in
this
nested
structure
that
Oh
TLP
currently
wants
it?
It's
pretty
gross!
It's
like
really
like
unfortunate.
If
you
only
do
have
like
one
span
coming
from
like
a
named
tracer,
cuz
you're
like
allocating
an
array
for
for
every
one
of
them,
they
can
get
in
there.
So
yeah,
it's
like
you
end
up
with,
like
a
map
to
a
map,
to
an
array.
Ultimately-
and
it's.
E
You
could
definitely
streamline
that
process.
I
think
like
the
only
benefit
I
can
see
potentially
is
like,
as
you
said,
I
think
on
the
client
side
like
gzipping
will
get
rid
of
any
any
duplication
there.
I
don't
know
like
that's
kind
of
on
the
request
side.
I
think.
Definitely
with
how
much
processing
you're
doing
on
the
client
I
think
your
memory
usage
is
going
to
be
a
little
bit
higher
to
like
create
this
payload.
E
Maybe
on
the
receiving
side,
because
you
don't
have
to
unzip
something
into
something.
Quite
so
big,
there
might
be
some
modest,
like
memory
savings
on
the
server
side,
I
feel
like
that's
the
only
thing
that
maybe
you're
you're
saving
well
I'm,
not
even
confident
that
that's
happened.
That's
just
kind
of
my
back
with
a
napkin
reasoning.
E
Yeah,
all
this
is
probably
somewhat
relevant
to
stuff
you've
been
involved
with
Eric,
just
because
it
kind
of
has
to
do
with
with
export
pipeline.
This
is
like
specifically
with
OTO
team,
so
I
don't
know
if
anything
will
actually
change
like
on
the
readable
span.
Interface
like
I,
think
we
might
still
just
have
instrumentation
library
there
as
right
as
it
is.
A
Yeah
I
mean
I
like
in
terms
of
like
the
practical
implementation,
I
think
it's
right,
I,
don't
I
mean
I'm
I.
Think
it's
not
going
to
touch
anything
I've
done
for
a
little
I
guess
so
is
the
idea
that
there
would
just
be
on
the
instrumentation
library.
There
would
then
be
an
I
apologize
for
not
like
totally
following
along.
There
would
then
be
attributes
essentially
associated
with
that
that
we'd
have
to
you
know
it
basically
map
onto
the
particular
to
every
span.
I.
E
Mean
so
really
the
instrumentation
library
itself,
we
had
this
in
Ruby
right
now,
like
the
readable
span
or
the
span
data
has
an
instrumentation
library
struct
on
it
literally
just
holds
the
name
of
the
instrumentation
library
and
it's
version
yep.
So
this
is
probably
something
when
you
export
to
data
dog
that
you
will
end
up.
My
guess
is
you'll
just
want
to
merge
this
in
with
your
attributes
for
the
span,
but
it's
kind
of
the
way
it's
set
up
an
OT
LP.
It
went
through
kind
of
great
lengths
to
normalize
the
data.
E
A
If
there's
more
tag,
key
value
pairs,
that's
fine
too
Adam
on
it's
not
a
huge
ass,
but
that's
we
have
a
concept
called
span
types
which
mostly
ends
up
being
UI
related
how
you
display
a
DB
span
versus
of
whatever
a
webspam,
but
there's
some
obfuscation
that
occurs
at
our
our
collector,
which
is
like
a
trace
agent.
So
if
it's
like
a
DB
span
type
which
I'm
using
the
inspiration
library
name
to
determine,
there's
some
obfuscation,
sequel,
queries
for
example.
A
E
That
make
sense,
yeah
no
I
think
it
all
makes
sense.
Ultimately,
I
don't
feel
see
any
changes.
I
think
we
might
still
end
up
with
instrumentation
library
being
the
thing
on
the
spam
that
you
are
going
to
to
go
after,
but
there's
a
small
chance
I
think.
If
proposal
two
went
through,
you
would
no
longer
see
instrumentation
library.
You
would
basically
see
something
called
common
attributes,
gotcha
going
through
yeah.
E
D
Doesn't
matter
for
other
exporters,
really,
it's
really
only
relevant
to
ot
LP.
The
only
thing
that
you
need
to
be
aware
of
is
whatever
the
data
dog
exporter
from
the
open,
telemetry
collector
ends
up
doing
with
this.
Whatever
mapping
it
does
for
these
athletes,
you
should
really
do
the
same
in
the
data
dog
exporter
of
from
open
country,
routing.
A
Right
yeah,
so
at
this
point
we
don't
even
have
a
collector.
We
don't
even
have
an
exporter
in
the
collector,
it's
just
words
and
it
we're
writing
to
our
the
trade.
You
know
our
daemon,
but
I
think
the
plan
is
to
get
a
collector
in
there.
It's
just
a
more
heavyweight
there's.
A
lot
of
other
metro
generation
of
work
occurs
in
our
trace
agent.
That
was
like
thing
to
add,
so
I
think
it's
a
more.
It
wouldn't
be
me
writing
at
this.
This.
E
No,
you
know
I
think
I'm,
like
I,
think
ultimately
yeah
I,
think
for
for
you
for
Dana
dog,
like,
as
you
were
mentioning,
there's
like
a
lot
of
things
that
happen
in
your
agent
before
stuff.
Can
kind
of
leave,
leave
a
customer
system
like
obfuscation
and
scrubbing,
and
all
this
other
stuff
and
I
really
feel
like
open
telemetry.
E
It'd
be
interesting
to
see,
see
what
open
telemetry
does
for
this
I.
Think
for
like
Shopify,
you
all
don't
care
very
much,
because
you
own
your
whole
pipeline
I.
Think
for
like
a
lot
of
customers
and
when
things
kind
of
leave
when
they
leave
your
network,
like
certain
data,
cannot
leave
your
network
and
like
customer
data.
Pii
is
usually
in
that
in
that
area.
In.
D
D
E
E
E
E
E
E
Yeah
I
would
have
been
talking
this
this
discussion
a
whole
lot.
I
know
that
sampling
itself
is
an
incredibly
broad
and
challenging
topic,
and
the
needs
for
different
people
vary
wildly
and
the
actual
metadata
needed
on
your
back-end
to
understand
how
a
decision
was
made
really
depends
on,
like
the
sampling
algorithm,
that
you're
using
and
while
priority
is.
E
E
E
E
So
looks
like
there's
a
meeting
if
you're
interested
in
talking
about
this
more
there's,
also
an
errors,
discussion
group
that
is.
E
E
It's
it's
something
that
we
needed,
something
that
we
need
so
I
just
opened
this
PR
a
long
time
ago.
It
was
really
adapting
the
open
tracing
conventions
for
recording
errors
as
logs
and
open
tracing,
but
it's
kind
of
just
let's
do
the
same
thing:
I'm
events
and
open
open
telemetry
and
if
nothing
else,
just
because
or
well,
we
ended
up
adding
this
API
method
if
nothing
else,
just
to
kind
of
have
a
bookmark
for
where
we
want
to
rescue
the
errors
when
we
were
converting
over
the
instrumentation
from
DD
trace,
but
yeah.
E
E
I
do
know
that
other
languages
have
jumped
on
and
have
like
a
span
record
error
method
because,
like
they
were
in
the
same
boat
like
we
need
this,
so
we're
just
speculating
with
something,
and
they
also
have
chosen
this
one
I
believe
go-go
Lang
has
an
exact
same
method,
working
the
same
way,
I
think
Python
also
does
mm-hmm
I'm,
not
sure
if
it
exists
elsewhere,
but
at
any
rate
someday.
We
will
have
a
way
to
like
formally
do
this.
A
The
thing
I'm
kinda
cancer
about
just
being
able
to
accurately
grab
the
exception,
type
the
stack
trace
and
the
message
which
right
now,
it's
sort
of
like
I'm
guessing
at
the
right,
I
guess
they're
in
the
status
description,
they're
sort
of
like
you
know,
interpolated
in
there,
and
you
have
to
sort
assume
the
format.
So
it
seems
like
the
current
approach
using
events
with
or
this
speculative
approach.
Using
events
with
the
you
know,
key
value
pairs
is
a
little
easier
for.
A
E
So
yeah,
so
this
is
with
the
thing
I'm
talking
about
and
we
are
actually
adding
these
as
events
on
a
span.
So
these
are
all
the
things
that
you're
looking
for
I
think
the
problem
is
that
there's
also
span
status,
and
this
is
not.
This
is
not
aligned
with
span
status,
so
I
think
there
is
some
overlap.
I
could
in
interim,
see
this
method,
setting
the
status,
at
least
at
a
bare
minimum,
to
an
unknown
error.
I.
A
G
A
D
E
A
A
E
It's
it's
a
complicated
thing
and
there's
like
a
whole
industry
that
exists
around
errors
to
begin
with
and
I
think
the
thing
that's
made
it
a
little
challenging
and
open
telemetry
is
there's
always
kind
of
bad.
Like
you
know,
the
project
was
originally
supposed
to
replace
open
tracing
and
open
census.
E
So
it's
just
wasn't
big
a
merger,
so
it's
like
okay,
traces
and
metrics
will
be
like
the
things
that
were
focused
on
41.0
logging
has
always
kind
of
been
this
thing
that
we
we're
going
to
add
support
for,
but
it
was
going
to
kind
of
be
a
later
thing,
but
it's
seems
like
there's
a
lot
of
interest
in
having
logging
and
error
seems
to
be
kind
of
caught
in
this
intersection
between
logging
and
tracing
and
right.
I.
Think
that's
been
that's
been
the
problem.
E
E
E
It
was
brought
up
at
a
previous
spec,
sig
meeting
and
I
think
there's
some
ideas
for
improvement,
obviously,
but
I
think
they
wanted
to
have.
Somebody
actually
like
measure,
measure
diversity,
inclusion
first,
so
you
can
actually
tell
if
you
have
improved,
but
yeah
I
mean
clearly,
there's
always
been
like
this
open-door
policy.
E
A
E
It's
good
to
see,
if
you
will
talking
about
this,
if
you
have
more
interest
in
furthering
these
files,
it's
kind
of
checking
you
that
out
and
then
yeah
they're,
just
kind
of
talking
about
having
a
better
change
log
for
the
spec
SIG's,
so
that,
as
we
do
cut
milestones,
it's
like,
whereas
milestones
are
completed
and
releases
of
the
spec
are
cut.
We
have
a
better
idea
of
what
that
actually
even
means
at
the
sig
level,
because.
E
D
E
E
E
E
E
D
A
Thanks
for
being
responsive
guy
I
just
apologize
for
kind
of
spinning
out
the
pings,
but
thank
you
guys
for
chiming
in
it
was
helpful.
It
was
more
just
as
I'm
working
through
the
implementation
of
the
exporter,
some
questions
on
the
API
and
SDK,
but
I
think
I'm
in
a
good
place
with
most
of
it,
I
did
put
up
a
pretty
rough
PR.
I
know
this
isn't
even
the
right
repository
to
make
up
PR
against,
because
this
should
live
in
a
contribute,
though
yeah.
D
D
A
So
yeah,
basically
the
issue
is-
and
this
is
what
ended
up
happening
for
context.
The
same
thing
ended
up
happening
in
the
Python
implementation.
They
rolled
their
own
processor,
and
you
know
it's
work
that
I
guess
I
could
move
down
to
the
exporter
level,
but
because
I'm
trying
it
like
it's
basically
there
a
trace
needs
to
be
exported
as
a
complete
trace,
only
isn't
the
only
time
it
can
be
sent
to
a
processor.
So
you
can't
just
batch
a
bunch
of
spans
and
arbitrarily
say,
like
500
span,
send
them
along.
It's
got
a
wait.
A
There's
got
to
be
a
sort
of
in
memory,
you
know
record
of
the
traces
and
then
the
spans
reach
trace
and
when
that
trace
is
considered
complete,
which
the
working
definition
is
when
there
are
no
every
span
is,
has
been
finished.
Essentially,
there's
no
spans
waiting
to
be
finished
then
they
can
be
then,
for
all
intents
and
purposes,
that's
a
batch.
Basically,
so
you
know
it's
work
that
I.
Could
you
know
I
would
probably
move
it
down
to
the
exporter
and
then
make
the
exporter
have
like
a
worker
thread.
A
That's
doing
that
work,
but
yeah
right
now
it
just
lived
in
the
processor
because
it
seemed
to
be
the
closest
convention,
but
yeah
I'm.
You
know
my
goal
is
not
to
rewrite
the
book
here,
it's
more
just
to
get
something
that
works
and
so
the
same
work
than
eat.
The
bad
strand
processor
needs
to
do
needs
to
get
done.
For
the
you
know,
I
was
calling
it.
A
Yeah,
just
in
just
in
terms
of
other
stuff
there's
you
know
it's
still
pretty
rare
I
was
it's
there's
some.
You
know
it's
it's
a
little
gross
the
code,
so
I
apologize
I've
just
wanted
to
get
something
in
to
get
some
thoughts
on
the
implementation
and
there's
so
a
few
areas
that
I
have
to
address,
but
by
and
large
I
mean
yeah
I.
A
So
I
guess
let
me
know
if
what
you
guys
think
and
I
just
gifts,
I,
don't
need
an
answer
right
now,
obviously,
but
if
you
prefer
that
I
think
what
I
would
end
up
doing
is
just
use
that
if
the
preference
is
to
use
a
put
all
that
work
in
the
exporter,
I
would
just
use
the
simple
span
processor
and
then
move
all
the
you
know
asynchronous
stuff
out
to
the
exported
the
yeah.
What
do
you
guys
think
of
that
approach,
or
do
you
have
any
thoughts?
I.
E
D
It's
probably
a
naming
thing:
I
was
initially
confused
by
the
fact
that
this
seemed
to
be
kind
of
copied
and
slightly
tweaked,
but
I
looked
a
little
more
of
the
implementation
now
and
that
it's
obvious
that
you
need
some
kind
of
span
processor
because
you
need
is
like
track.
The
start
of
a
trace
really
and
and
the
end
of
a
trace.
D
So
yeah
I,
you
might
want
to
consider
well
two
things.
I
guess
one
is
the
way
we're
doing.
Batching
is
obviously
different,
and
that
may
you
may
want
to
end
up
changing
the
code
that
actually
kind
of
does
forwarding
and
like
the
timeout
values
and
so
forth.
I
just
yeah.
They
probably
completely
wrong
for
you
and
the
other
is
the
the
name
of
the
thing.
Maybe
also
you
just
want
to
change
that,
to
like
data
dog
stand,
processors
or
something
so.
A
Okay,
yeah
yeah
I
can
change
the
classes
that
I
can
I
can
make
it
a
little
more
clear.
I
was
basically
at
first
I
was
like.
Oh
maybe
I
can
just
tweak
the
best
thing
and
then
I
was
like
nope.
This
is
getting
away
from
me.
This
is
totally
different
or
not
totally,
but
this
is
different
enough
and
yeah.
A
lot
of
the
I've
left
some
values
in
there
from
the
fat
stem
processor
that
I
need
to
clean
up.
A
There's
a
lot
of
work
that
gets
essentially
laid
off
in
terms
of
like
the
timeouts
and
the
retries
to
the
data
dog
writer,
which
I'm
just
importing
in
and
sort
of
just
kind
of
is
magic.
But
it's
it's
relatively
well
tested
and
robust
code.
So
it's
made
things
a
lot
easier,
because
I
can
that's
handles
a
lot
of
the
more
difficult
work
but
yeah,
so
yeah
I'm
still
cleaning
this
up.
A
But
that's
my
goal
is
to
have
something:
that's
at
a
point
where
it
can
be
reviewed
out
on
whatever
timeline
works
and
then
I
can
start
to
work
on
getting
some
integration
adapters.
You
to
you
know,
improve
the
experience
for
for
users.
My
see
I,
think
Francis.
You
send
send
me
the
my
sequel
one,
because
that
feels
like
a
pretty
straightforward
implementation.
I
just
haven't
had
I
thought.
I
was
onto
like
knock
it
out
earlier
this
week
and
I
just
have
enough
time,
as
it
always
hits.
D
Yeah
in
many
ways,
I
feel
like
the
batch
plan.
Processor
is
most
useful
for
OT
LP
I
can't
remember
whether
I
feel
like
yeag
is
the
Yeager.
Exporter
is
also
doing
its
own
kind
of
internal
batching
as
well.
I
have
to
go
and
reread
I'm
I
wrote
it,
but
I
just
don't
remember
how
it
did
it
so
yeah,
so
I'm
not
sure.
A
H
E
A
You
notice
one
or
two
other
small
things
that
I
need
to
address.
One
is
the
Sinatra
implementation.
Has
some
brittleness
around
it
throws
it
throws
errors
on
like
404s
on
things
that
should
be
404?
Is
it
there's
some
error
handling
so
I
plan
to
make
a
small
PR
there,
and
then
I
was
just
going
to
do
a
my
tiny
update
to
the
struck
the
spend
a
destruct
to
include
whatever
it
was
tray
state,
because
there's
yeah
I
need.
E
Yeah
so
I
think
that's
that's.
The
reality
of
the
world
is
that
there
are
actually
some
people
using
open,
telemetry
movie,
even
though
we
are
I,
don't
know,
maybe
somewhat
behind
some
other
SIG's
so
but
I
think
we
don't
have
a
ton
of
users
so
like
there
are
some
rough
edges
that
haven't
been
totally
turned
out.
So
if
you
find
something
that
looks
weird,
it
probably
is
weird
and
definitely
ask
about
it
and
feel
free
to
fix
it.
If
you
find
it
yeah.
A
E
D
A
Happy
to
be
helpful,
what
does
it
mean
to
be
so
when
you
say,
like
you
know
my
so
I'm
allocating
a
certain
percentage
of
my
working
hours
are
now
to
trying
to
contribute
to
hotel
color
for
the
next
month
or
two.
What
does
it
mean
for
us
to
be
four
rwby
to
be
and
like
alpha?
Is
this
something
that
shouldn't
definitely
shouldn't
be
used
in
production?
A
If
there
is,
you
know,
should
it
be
done
and,
like
my
first
step
would
be
put
this,
we
have
some
internal
reliability
environments
that
I
was
going
to
try
to
run
this
in
when
it
gets
to
a
point
where
you
know
it's
working,
what
is
yeah
what
it
from
your
perspective
like?
What
does
it
mean?
Is
it
totally?
Do
you
guys,
like
totally,
not
want
to
put
a
stamp
of
approval
on
it?
Does
it
mean
do
it
with
caution
I'm,
just
not
sure
so
there.
D
Was
some
requirements
around
and
they
mocking
it
as
beta?
So
some
of
the
cool
languages
went
on
a
heavy
push
to
a
beta
release,
and
then
there
was
a
big
announcement
about
it.
We
chose
not
to
do
that.
We're
not
too
far
behind,
but
I
think
there's
a
few
critical
things
that
were
required
for
the
beta
tag
and
we
didn't
meet
that.
So
we
just
chose
not
to.
D
We
would
be
doing
that
right
now
if
my
team
wasn't
distracted
with
some
infrastructure
issues,
but
we're
probably
like
less
than
a
month
away
from
having
this
in
production
in
at
least
a
couple
of
smaller
acts
like
low
volume,
apps,
our
kind
of
cool
monolith
and
some
of
the
larger
body
maps.
We'll
probably
see
this
a
little
bit
later
but
like
from
my
perspective,
I
think
it's
okay
to
stop
considering
using
this
in
production,
and
certainly
other
people
have.
E
They
definitely
left
left
us
out
or
it
gave
us
like
the
option.
I
think
some
were
really
kind
of
encouraged,
or
we
essentially
required
to
do
that
and
like
one
of
the
things
that
one
of
the
big
requirements,
I
think
was
having
a
metrics
SDK,
and
this
is
a
thing
that
we
don't
yet
have.
But
I
think
it's
fine
honestly
because,
like
there
has
been
a
lot,
there's
been
a
lot
of
churn
in
metrics,
so
people
who
have
been,
who
have
had
an
SDK
have
been
kind
of
like
making
circles
with
their
SDK
I.
E
Think
that's
fine.
It's
always
easier
to
kind
of
refactor,
something
that
you
have.
What
again
is
a
lot
of
effort
has
been
put
into
that
by
a
lot
of
other
cigs
and
for
some
other
SIG's
they
have
the
staff
to
do
it
and
it's
not
like
a
big
deal,
but
we,
we
are
a
small
and
very
busy
group
and
usually
busy
with
things
other
than
open
telemetry.
So
that's
kind
of
been
our
plight,
but
our
tracing
API
has
been
in
in
good
shape
for
a
while
I.
Think.
E
For
me,
like
the
only
thing
that
I
worried
about
a
little
bit
is
that
issue
that
Francis
Wang
was
working
on
about
some
attributes
can
possibly
make
it
to
the
export
pipeline
and
oh
I've
basically
seen
it
where
like.
If
there's
a
bug
in
instrumentation
or
if
you
had
like
an
invalid,
attribute
to
a
span
and
are
using
like
the
Yeager
exporter,
your
trace
will
get
broken.
That's
the
span
will
get
dropped
or
something
so
like
that's
the
biggest
problem
which
like,
if
we
fix
that
and
then
add
some
export
options.
E
A
That's
good
to
know:
yeah
I
got
I've
said
a
thousand
times.
My
goal
is
get
something
out
the
door
with
the
exporter
and
then
try
to
just
knock
out
a
bunch
of
the
instrumentations.
Hopefully
some
rails
genius
comes
along
I.
Don't
have
to
do
that.
Yeah.
E
E
E
A
That's
that's
great
color
man
I
appreciate
it
cool.
E
E
H
B
B
J
K
Someone
signed
up
for
Netflix
with
my
email
address,
my
personal
email
address.
That's
not
me.
Does
that
mean
you
get
free
Netflix
the
free
trial,
Oh
speak
English,
and
so
my
personal
gmail
is
VHS
at
gmail
just
been
kind
of
a
curse
because
it's
not
like
I,
don't
get
spam.
I
just
get
a
bunch
of
misdirected
email
and
people
to
see.
K
K
J
C
J
J
J
J
K
J
G
H
K
L
J
J
H
G
K
K
I
wanted
to
raise
one
thing
then
I
put
on
that
one
of
the
DNI,
the
two
issues
just
to
see
what
I
mean,
what
people's
reaction
to
it:
I
like
that
you've
trying
to
put
a
little
more
teeth
on
it,
and
this
is
one
of
the
two
things
we
could
do.
We
can
say
if
you're
going
to
have,
if
you're
a
vendor,
that's
going
to
have
a
seat
on
that
GCC.
B
K
J
And
and
I
say
this
is
someone
like
weird
I,
think
a
third
person
join
our
engineering
team
and
we
would
likely
meet
those
criteria,
but
I
think
they
will
also
create
tension.
There
are
vendors
that
have
like
just
like
don't
have
engineers
other
than
like
one
like
one
person
working
on
the
project,
so
it's
gonna
create
some
tension
between
some
of
them
likely
to
be
blunt,
like
actually
Jonah
who
brought
like
they
don't
have
any
engineers
working
on
this.
Does
that
also
just
disqualify
them
like.
C
J
I
K
The
company,
or
something
like
that
I'm
just
saying
like
I
like
I,
have
to
admit
I
kind
of
like
the
idea.
I
mean
that's.
Why
I
wrote
it
down,
but
I,
don't
I
mean
we
could
have
some
criteria
around
it.
But
you
know
if
Google
or
FDA-
oh
I'm,
sorry
I,
guess
he
doesn't
work
there
anymore,
whatever.
If
a
giant
company
they
can't
make
that
you
know
it's
like
there
are
people,
but.
J
It
also
depends
on
who's
who's
involved,
like
I'll,
bring
up
Jonah
again,
isn't
because
he
raised.
The
issue
is
just
it's
a
unique
unique
example,
because,
like
he's
the
CTO,
so
he
actually
has
a
lot
of
control
over
who
they
bring
in
and
to
the
contrary,
like
in
my
case
like
I
I'm,
just
a
p.m.
at
Google
I
like
I,
have
to
fight
to
keep
funding
for
this
project
at
all,
and
and
so
it
it
would
be.
C
J
C
K
K
K
K
K
It
there
there
be
dragons
and
I
think
we
might
lose
the
GC
members
team
as
well,
which
is
yeah
what
it
is,
but
I
just
wanted
to
I
wanted
to
debate
it
a
little
bit
or
I
guess
maybe
the
better.
The
better
point
would
be
what
are
other
things
that
that
can
be
like
give
to
get
things
that
you
have
Keith
on
getting
senior
ICS
who
bring
more
diversity
or
less
white
dudes
to
the
project.
I
just
couldn't.
J
C
Press,
it's
probably
the
biggest
one,
so
blog
posts
about
you
know
like
so
what
we
used
to
do
within
kubernetes.
Is
you
reward
the
company
for
staffing,
the
things
or
doing
the
things
that
don't
directly
benefit
them,
but
you
reward
the
individual
for
doing
whatever
their
they've
assigned
for
you
know,
because
they
they
get
assigned
things.
But
if
a
company
you
know
puts
in
a
lot
of
effort,
doing
something
that
benefits
the
whole
computing
community?
C
That's
when
we
tried
to
write
blog
posts,
so
you
know
maybe
this
is
you
know
we
commit
to
highlighting
our
partners
in
this
project
or
our
corporate
partners
in
this
project
who
make
an
effort
to
stuff
with
DNI
quarterly
or
with
every
release.
We
will
make
sure
to
mention
the
companies
that
provided
that
you
know
made
an
effort
to
staff
with
not
white
dudes.
J
C
J
I
C
I
C
K
C
B
B
J
B
I
I
just
wanted
to
make
sure
that
it
was
clearly
communicated,
including
at
the
monthly
public
governance
meeting,
that
we
have
a
Governance
Committee
discussed,
adopting
the
CLA
bought
like
probably
two
months
ago,
and
it
took
like
a
months
to
actually
implement.
We
obviously
probably
should
have
been
communicating
this
more
prominently
in
github
issues
and
over
Gitter
before
actually
with
liens,
but
there
was
kind
of
this
long
drag
process
where
it
felt
like
you
know
we
we
agreed,
we
wanted
to
do
it
and-
and
we
just
were
waiting
on
the
CNCs.
I
So
when
I
did
happen,
people
are
a
little
bit
surprised.
The
major
issues
that
you
ran
into
or
that
poor
fauna
and
Splunk
didn't
have
their
CE
OS
properly
carried
over
from
the
old
system
to
the
new
system.
We
missed
them
when
we
were
doing
the
audit
of
making
sure
that
be
the
top
top
end.
Companies
were
had
all
sciency,
always
because
the
omniscient
coa
was
in
place,
and
I
was
kind
of
the.
I
I
No
we're
changing
the
automation
that
enforces
it
because
the
old
automation
was
inflexible
and
wouldn't
allow
bots
to
make
non
copyrightable
changes.
The
other
changes.
The
other
thing
that
we're
discovering
as
a
result
of
this
is
some
people
had
signed
under
a
individual
coa
who
should
have
been
corporate
cos,
so
those
people
will
be
caught
in
in
CLA
hell
and
they
do
need
to
be
for
legal.
F
F
B
I
guess
one
thing,
though,
since
we
just
found
out
that,
like
the
mailing
list,
that
things
were
communicated
on
weren't
actually
like
working,
should
we
actually
hold
off
until
we
actually
sent
in
emails.
Everyone
gets
an
update
on
what
the
CLA
is
and
then,
after
start
doing,
the
migration,
because
I'd.
I
I
M
I
I
don't
think
we
can
do
that
because
our
main
10
years
are
quite
independent
and
they
can
update
their
maintain
EEOC's
themselves
and
I,
don't
see
how
we
can
correlate
and
synchronize.
This
mailing
list
with
was
actually
some
containers
without
incurring
a
lot
of
work.
So
if
somebody
wants
to
do
this
work,
fine
I
mean
for.
I
M
I
Great,
let
me
know
what
people's
opinions
are
as
of
the
Monday
maintainer
meeting
and
I'll
decide
whether
or
not
to
enforce
on
Thursday
next
week
or
two
extended
in
another
week,
I.
B
K
I
We're
good
with
it,
except
for
two
things,
be
I,
think
renovate
bought,
they're,
open,
telemetry,
JavaScript
already
had
installed,
is
capable
of
opening
branches
and
proposing
changes
from
those
branches
and
I
believe
the
github
built
in
the
github
built
in
bot
that
updates
your
dependencies
again.
That
request
permission
to
to
update
code.
I
Those
are
the
only
two
exceptions
I
might
propose,
but
other
than
that
I
think
we
don't
read
only
on
code,
whose
is
the
correct
situation.
B
B
M
So
I
think
one
item,
maybe
I
don't
have
a
document
is
youtube
uploads,
so
we're
using
plain
I
think
that
uploads
from
zoom
to
YouTube
and
it
opposed
to
this
private
videos
and
they
all
have
the
same
name.
So
now
the
question
is:
do
we
have
any
volunteer?
So?
Can
we
like
staff
somebody
to
control
these
private
videos
with
the
correct
captions
into
public
videos
or
like
like
right
now?
There
is
a
works
team
I
asked
I'm
here
to
file
issue
is
playing
to
make
sure
that
they
can
include
timestamp
in
the
caption
yeah.
B
M
M
Yeah
so
the
issue,
and
since
we
cannot
write
automation,
we
need
to
stop
it
somehow,
so
your
human
will
parse
through,
like
so
the
real
issue.
You
need
to
take
a
video
and
you
need
to
understand
what
was
on
the
counter
understate
and
then
listen
for
the
video
enough
to
understand
which
medium
get
this
and
then
capture
it
appropriate.
J
M
M
Is
like
incorrect
timing,
so
time
may
be
enough.
I
mean
we
can
do
date,
but
we
cannot
do
missing
time
and
there
are
some
certain
problems
right.
Sometimes
our
meetings
got
occupied
by
some
random
person
just
join
a
meeting
and
sit
on
the
meeting
for
like
a
few
hours
and
then
can
be
squash
together
in
writing,
because
I.
G
J
M
J
Yeah
and
like
people
can
share
the
link
like
cuz,
like
typically
like
right
now.
They're,
not
public
and
people
have
be
reset
to
me,
like
maybe
we'll,
like
once
a
month
asking
for
meeting
recording
and
it's
typically
like
they
wanted
a
very
specific
one
and
they're
just
like.
Oh,
if
there's
any
way,
I
can
sift
through
these
that's
fine
and
so
icky,
even
if
we
just
published
them.
J
Even
if
the
titles
are
bogus
and
people
have
to
at
least
like
for
a
single
day,
click
through
them
just
to
figure
out
which
one
is
which
for
most
people's
scenarios,
that's
that's
way
better
than
what
they
have
now,
where
it's.
Let
me
find
out
that
I
even
need
to
email,
someone
who
has
access
and
then
ask
them,
and
then
they
just
give
up.
Okay,.