►
From YouTube: 2022-09-29 meeting
Description
Instrumentation: Messaging
B
B
Yeah
my
co-working
space
is
doing
a
body
this
evening
and
so
I
am
there
in
a
like
phone
phone
room,
Booth
yeah.
A
Yeah
yeah
we
went,
I
went
to
Splunk
offices
down
in
California.
They
had
a
few
of
those,
but
I
always
like
wanting
to
see
people
in,
but
then
you
get
into
them
and
it's
like
I,
don't
know
I,
maybe
I'm,
just
like
a
little
claustrophobic
I.
Just
couldn't
do
it.
I
was
just
like
this
is
too
much
but
yeah.
A
It
looks
like
David's
also
had
a
office
today.
A
A
A
Okay,
yeah
I
think:
where
are
we
at
two
and
a
half
minutes
and
I
think
we
could
probably
just
jump
in
here?
Let
me
start
sharing
my
screen
really
quick.
Then.
E
A
Okay,
welcome
everyone
thanks
for
joining.
If
you
haven't
already,
please
add
your
name
to
the
attendees
list.
We
have
agenda
items
you
want
to
talk
about.
Please
add
them
and
we
can
jump
in
here.
So
first
on
the
list
is
something
I
added
I
wanted
to
check
in
on
the
next
Milestone
that
we
have
listed
to
go
out.
Currently
there
are
three
issues
associated
with
it,
and
this
is
a
bug.
A
Foot
bug
fix
like
patch
release,
I
think
this
kind
of
ties
into
what
Aaron
was
kind
of
having
a
little
or
later
for
a
triage,
so
one
of
the
big
ones
that
I
think
is
solved
currently
is
the
default
view.
A
I
think
if
that's
something
that
should
be
applied
implicitly,
we
know
these
are
exists,
it's
something
in
the
spec
and
there's
a
PR
for
it.
There's
another
issue
here
as
well
for
duplicate
insurance,
registration
and
there's
a
related
one
for
compatible
aggregation
should
not
error
instrument
creation.
So
this
is
something
I
created
yesterday
and
it's
related
to
the
duplicate
instruments.
So
if
you
try
to
create
a
duplicate
instrument,
let's
see
yeah
there's
a
good
test
that
we
actually
do.
A
That
is
actually
wrong,
but
it's
real
easy
to
fix
it
then,
and
so
this
is
one
where
the
view
actually
will
do
a
rename.
So
in
this
case
it
renames
bar
to
Foo,
and
then,
when
you
try
to
create
this
instrument,
it's
already
been
created
directly,
so
this
food
instrument
has
already
been
directly
created.
Currently
this
errors,
the
specification
says
that
it
should
just
since
it's
compatible
instrument,
it
should
just
return
the
existing
aggregator,
and
so
then,
on
top
of
that,
there's
also
this
duplicate
instrument.
A
So
when
The
View
doesn't
do
anything
and
you
just
ask
for
the
same
thing
back
currently,
this
fails
to
return
the
same
thing
back.
So
that
was
something
that
was
solved
in
this
cash
metric
instrument
pipeline
PR.
This
was
old.
There
was
a
refactor
to
the
pipeline
that
was
done
so
I
I
was
going
to
leave
it
up
to
you.
A
If
you
wanted
to
close
this,
but
I
I'm
added,
essentially
an
equivalent
PR
here
to
actually
do
the
same
thing
to
the
pipeline
but
I
after
opening
up
this
compatible
aggregator
PR
I
realized
that
this
caching
layer,
probably
actually
isn't
at
the
right
level,
I
think
there's
like
two
levels.
We
could
probably
cache
at
like
the
provider
and
at
the
aggregation
generation
per
instrument.
A
So
I've
closed
this
and
I'm
building
out
a
redo
of
this,
but
I
think
that
there's
a
there's
a
way
to
solve
both
of
these
PRS
with
with
caching
I
was
looking
at
Java
and
python
for
compatibility,
and
so
it's
it's
there.
It's
a
work
in
progress,
so
I
I
wanted
to
kind
of
like
go
a
little
bit
of
a
deep
dive
there.
Sorry,
if
somebody
was
trying
to
say
something.
D
A
Yeah
all
right,
perfect,
so,
okay,
yeah,
so
with
that
I
think
that
these
two
three
two
two
nine
and
three
240
are
not
complete.
This
is
technically
not
complete
either.
This
still
needs
reviews,
I,
haven't
looked
to
think
Pablo
had
commented
on
this
Aaron
just
approved
this
as
well,
and
so
this
still
needs
a
review,
but
I
was
wondering
what
we
want
to
do
on
timeline
for
this
release.
A
D
My
suggestion
is,
let's
gather,
all
the
bugs
up
like
yeah
and
if
there
are
any
more
in
the
queue
that
are
labeled
as
bugs,
let's
make
sure
those
are
addressed
as
well,
if
possible,.
A
Let's,
let's
take
a
look
here.
This,
of
course,
is
not
the
best
view,
I
think
Milestone.
B
A
Yes,
that
is
definitely
something
we
still
need
to
work
on.
The
actively
metric
was
string
sliced
through
certificate
metrics.
That
is
also
something
that
needs
to
get
worked
on.
This
is
I.
Think
more
Universal,
I
think
there's
a
lot
of
context
here.
It
definitely
refers
to
our
string
slices
like
a
bigger
picture.
A
Said
they
wanted
to
pick
this
up,
I,
don't
know
if
I
assigned
this,
though
it's
a
really
complex
one,
I
I,
don't
know,
oh
it's
assigned
to
Josh.
Currently,
that's
probably
why
I
didn't
say
it?
What
do
you
think
Aaron?
Do
you
want
to
add
these
two
or
should
we
just
keep
it
at
what
it
is.
D
D
I
I
haven't
gone,
had
a
chance
to
go
and
Link
all
the
tickets
for
it,
but
I
think
I
think
that
will
address
it
because
the
refactor
allows
you
to
add
a
reader
which
allows
you
to
do
things
like
change
the
default
temp,
not
temporality,
but
default
aggregation
and
and
whatnot.
But
it
also
fixes.
D
I
think
there's
two
other
issues
that
that
talked
about
having
to
use
Prometheus
like
importing
the
original
Prometheus
with
the
Prometheus,
but
we
could
we
I
can
I
can
clean
that
up
later,
I
just
haven't
had
the
time
today
to
actually
clean
it
up.
Okay,.
A
D
D
Probably
I
think
that's
probably
a
good
idea
and
if
we
don't
have
it
done
by
the
end
of
by
this
time
next
week,
then
maybe
it's
just
say:
hey,
let's,
let's
pundit
and
then
make
a
release.
If.
A
A
A
Yeah
I,
so
I
I
think
that's
kind
of
where
we
land
on
that
Josh,
like
I
I.
Think
it's
a
great
if
we
could
find
somebody
to
do
this,
I
have
no
problem
picking
this
up
after
the
current
bug
that
I'm
working
on
but
I,
don't
know
if
that's
going
to
be
resolved
within
a
week.
A
A
Think
I'm
that's
the
wrong
button.
If
I'm
hearing
you
correctly
I
think
we
could
do
the
same
thing,
we
just
did
for
the
Prometheus
one
we'll
add
it
into
this
patch
release
and
then,
if
we
aren't
finished
with
it,
but
next
week,
we'll
move
it
into
the
the
next
patch
release.
A
Yeah:
okay,
cool,
okay,
yeah!
Let's,
let's
do
that?
Okay,
I,
don't
know!
If
there's
any
other
issues
that
we
wanted
to
include
in
this
milestone.
D
Sorry
I
just
had
a
thought
on
the
Prometheus
one
that
actually
is
a
breaking
change.
It
changes
how
the
new
function.
The
new
function,
will
return.
An
error.
I
didn't
highlight
that
yet
I
need
to
put
that
in
there,
but
is
that?
Do
we
have
appetite
for
that
right
now,
I
I
guess
it
could
be
redone
without
having
the
new
function,
but.
A
B
E
A
Remember
just
update
the
version
number
of
the
Milestone
and
I
think
that
we'll
we'll
try
it
that
way.
Okay,
there,
you
go
okay,
yeah,
perfect
cool
I.
Think
that's
it
for
well
on
Amazon,
here
yeah
so
soon
to
come.
A
I
do
think
that,
like
this
is
probably
the
right
way
to
do,
because
we
don't
really
have
anything
to
go
out
with
tomorrow.
We
should
probably
try
to
do
this.
Just
look
on
Thursdays
and
try
to
get
it
out
and
Thursday
Friday.
A
D
A
Okay,
so
we
have
yeah
the
four
slush
shutdown
they
don't
collect
pending
So.
Currently,
if
you
call
large
flush
or
shut
down
it
just
shuts
down,
even
if
you
had
already
recorded
other
measurements.
So
this
is
something
that
is
it's
nice
I,
don't
know
if
it's
worth
making
a
patch
really
stress
for
this
other
thoughts,
for
it.
E
E
A
I
also
like,
if
we
you
know
we're
a
little
Dynamic
like
if
we're
able
to
resolve
bug
issues
faster.
So
if
we've
got
like
two
or
three
to
go
out
in
a
week,
we
get
a
patch
out,
especially
at
this
early
phase
in
the
development
of
the
beta
are
working
towards
the
beta
I
know
that
people
are
using
the
alpha,
SDK
and
they're
getting
snagged
by
a
bunch
of
these
bugs
So
the
faster
we
can
I
think
free
them
up
of
the
bugs
the
faster
we
get
positive
feedback
or
any
sort
of
feedback.
A
That's
outside
of
those
bugs
I,
think
it'd
be
really
critical,
I,
don't
think
I
think
I'm
the
only
one
that
now
that
may
not
be
trying
to
think
somebody
also
plus
one
of
this
nope.
Okay,
so
I
don't
know
if
this
is
hitting
a
lot
of
people.
So
maybe
we
can
just
wait
on
this.
One
I
think
two
bugs
sounds
great.
A
Okay,
cool:
we
have
some
work
to
do
for
the
next
week.
Awesome
cool
next
up
Josh.
We
have
you
talking
about
the
restoration
of
the
exponential
mapping
functions
here,
yeah.
B
So
thanks,
it
can
be
quick.
I've
had
this
code
in
the
SDK
before
the
old
SDK
was
removed,
and
during
that
final
release
window
actually
had
a
bunch
of
change,
reviewed
and
merged
I,
don't
know
how
well
it
was
reviewed,
but
this
code's
been
running
in
production
at
white
stuff
and
we're
pretty
happy
with
it.
So
there
are
sort
of
three
parts
pending
and
I've
been
kind
of
waiting
for
a
while.
It's
not
really
urgent
for
me,
but
they're.
B
Each
each
of
the
first
two
parts
are
are
something
that
I
have
work
pending
in
The
Collector.
That
depends
on
so
I
need.
This
released
it'd
be
nice
if
we
could
get
out
so
I
just
put
one
PR
up
today,
proposing
a
path.
I
think
the
only
question
as
long
as
you
agree
that
it
doesn't
need
to
be
re-reviewed
would
be
which
package
name.
You
want
to
use
I'm,
proposing,
SDK
metric
metric
data
exponential
and
then
two
subdirectories.
B
One
of
them
would
be
the
mapping
in
this
PR
and
the
second
would
be
the
structure
which
is
just
a
pure
data
structure,
which
would
be
a
sibling
in
SDK
metric
metric
data.
Exponential
structure
and
then
I've
put
up
another
draft
I
think
a
week
ago,
or
so
with
just
a
picture
of
what
the
whole
thing
looks
like
so
really
the
question
is:
can
we
get
this
in
quickly
since
it's
been
reviewed?
B
If
you
agree
to
the
path
name,
it's
ready
to
go
otherwise,
I'm
happy
to
change
the
path
name
and
the
pr
in
front
of
you
here
is
just
the
picture
of
what
it
all
looks
like
together,
which
is
the
mapping
functions,
plus
the
data
structure
plus
a
binding
to
the
internal
of
the
SDK
and
I've
drafted
that
I
think
I
got
got
it
just
about
right
anyway.
That's
all
I
have
foreign.
B
Yeah
I
think
I
have
that
had
that
pending
against
the
new
SDK
main
right
before
you
made
the
release,
I'd
be
happy
to
rebase
that
to
show
you
the
whole
story,
I
guess,
but
there's
only
really
one
file.
You
know
it's
the
SDK
metric
internal
exponential
histogram.go.
That
shows
you,
the
the
sort
of
bridge
between
SDK
using
the
data
structure.
I've
also
got
a
statsd
receiver
in
the
collector,
using
that
data
structure
and
the
hotel
go
the
late
step.
Sorry,
the
light
step
metric
SDK
is
using
it
as
well.
A
B
I
mean
this:
this
draft
here
is
three
pieces
all
put
together
and
I'm
trying
to
say
let's
I'd
like
to
get
piece,
one
which
was
already
reviewed
back
in
the
code
base,
and
then
the
next
part
would
be
to
review
the
structure
which
is
I,
think
it's
30
22.
some
some
existing
PR
30,
something
which
I'm
not
I'm,
not
pushing
for
just
getting
the
exponential
mapping
functions.
Re-Released
is
mainly
why
I'm
here
yeah
put
that
bullet
up.
Thank
you.
A
A
This
part
metric
dat
exponential,
this
kind
of
stands
out
as
a
kind
of
a
little
bit
of
red
headering,
where
there's
a
package
that
contains
only
data
types
but
I
can't
really.
This
just
has
a
bad
smell
at
a
high
level.
I
can't
actually
say
things:
I
haven't
reviewed
it.
So
please
take
that
with
a
huge
grain
of
salt.
D
B
Yeah
I
mean
that's
what
these
these
tickets
out.
There
have
have
requested
that
maybe
we
could
I
mean
I'm,
not
I'm,
not
saying
we
have
to
do
this,
but
but
the
request
was
made
months
and
months
ago
and
we
did
discuss
having
since
otel
go,
is
sort
of
the
a
going
to
use
that
code
and
B.
You
know
a
primary
repository
for
open,
Telemetry,
Go,
stuff
I
will
say
the
idea
of
having
that
data
structure
which
is
tricky
to
implement
and
hard
to
verify
it's
800
or
so
lines
of
of
tough
code.
B
Having
that
be
hosted
by
Hotel
go
is
a
Step
Beyond,
just
providing
an
SDK
in
a
sense,
and
yet
we
in
order
to
implement
a
merge
function
or
an
aggregate
function
for
one
of
the
collector
receivers
or
one
of
the
collector
processors.
We
kind
of
need
a
way
to
implement
the
actual
mechanics
of
the
data
structure.
So
that's
what
that
that
Free
Directory
you're
you're,
pointing
at
the
structure
directory
is
which
is
just
you
know,
there's
no
locking
in
there.
It's
just
a
interface.
E
B
Well,
that's
not
in
this
PR,
because
that
hasn't
been
reviewed
yet
by
you.
Okay,
this
is
just
getting
the
mapping
function.
So
there's
another
PR,
Linked
In
the
pr
description,
which
is
not
the
one
you
were
just
looking
at,
but
the
one
you
were
just
looking
at
had
it
in
there
it's.
You
know
the
structure
subdirectory.
B
D
B
In
the
readme
that
I've
committed
to
restore
this
package,
so
if
you
okay-
but
it's
also
an
open
PR,
you
can
find
it
in
the
polls.
That's
a
draft
somewhere,
not
that
one,
the
other
one,
but
this
again
this
contains
a
copy
of
it.
Okay,
so
the
structure
the
structure
package
contains
essentially
data
structure
that
has
a
new
function.
That
returns
you
this
object.
B
It
has
an
update
function,
it
has
a
merge
function
and
it
has
accessors
to
get
you
the
basically
the
the
fields
that
you
would
put
into
an
otlp
data
structure
and
we've
found
uses
even
outside
of
the
hotel,
SDK
or
the
white
step.
Hotel
SDK,
for
example,
we
had
an
old
Legacy
protocol
internally
and
we
wanted
to
replace
that
statsd
code
path
with
Hotel
data.
We
just
raw
converted
it
into
Hotel
histogram
using
this
data
structure.
So
there's
a
couple
of
proofs
that
it
works.
B
A
Yeah,
okay
I
will
definitely
need
to
spend
some
time
looking
at
this,
but.
B
And
then,
if
you
look
at
that
final
PR,
the
work
in
progress
that
is,
the
picture
of
the
mapping
function
was
merged.
The
structure
was
accepted
and
here's
the
binding
to
the
SDK,
which
is
the
internal
stuff,
where
you
do
internal
SDK
stuff
like
decide
how
to
lock,
decide
what
mapping
you
know
what
map
function.
You
know
map
structure
to
use
as
you've
done
so
I
was
just
copying,
essentially
the
pattern
in
the
histogram
file
into
the
exponential
histogram
file.
B
A
B
If
the
most
controversial
aspect
of
that
is
that
you
know
so,
it
just
requires
another
configuration
mechanism,
so
an
option
list
that
duplicates
or
corresponds
with
exactly
the
options
that
you
would
use
to
set
the
sdks
aggregator
up.
So
if,
if
you,
if
you
decide
now,
you
don't
want
to
have
this
structure
and
this
mapping,
if
you
don't
have
the
structure
I,
don't
think
we
should
commit
the
mapping
function
again,
I'm
willing
to
host
this
in
a
light
step.
Repository
I,
just
don't
know
how
you're
going
to
get
to.
B
If
you
want
to
implement
your
own
exponential
histogram,
you
know
great,
but
this
this
was
meant
to
be.
You
know
the
hotel
go
aggregator
and
the
reference
implementation
for
actually
has
been
used
in
other
Hotel
repositories.
At
this
point.
B
What
other
Hotel
repositories
so
python
has
copied
this
using
one
of
the
step:
Engineers,
basically
Incorporated,
that
so
it's
pending.
We
think
this
implementation
is
better
than
the
one
available
in
Java
right
now.
B
A
Okay,
yeah
I
I'll
have
to
take
a
look.
I
have
some
thoughts
but
they're
really
on
informed
right
now,
so
I
don't
really
want
to
cool.
B
A
Do
I
don't
think
that
I
don't
know
if
other
people
do
but
I
don't
think
that,
like
I,
think
that
we're
going
to
need
to
provide
an
exponential
histogram
and
having
some
sort
of
implementation
for
what
you
just
described,
I
think
is
needed.
So
I
I,
don't
I
think
that
that
seems.
It
seems
fitting
unless
it's
completely
decoupled
from
that
implementation
and
then
it
could
be
hosted.
Anywhere
But,
like
I
I,
think
you're
right,
like
I
I.
Could
it
sounds
like
something
we
could
add
here.
B
B
C
B
So
metric
data
to
me
contains
the
like
abstract
data
types
for
transfer,
so
it
there's
a
struct
that
contains
Min
maximum
count,
scale,
zero
count
and
then
a
positive
and
negative
bucket
range.
That's
the
exponential
histogram
struct,
and
that
is
already
and
and
my
final
PR
work
in
progress
has
one
of
those
in
the
metric
data
top
level.
That's
when
you,
when
you
aggregate
or
whatever
the
result
is
of
the
aggregate
function,
returns
you
one
of
those
struct
objects.
B
But
this
is
the
thing
that
mechanically
maintains
that
like
update,
State
and-
and
you
know,
you
guys
don't
have
a
merge
function
being
used
in
your
in
your
SDK.
B
The
one
I
wrote,
as
you
know,
does
contain
a
merge
function
and
that's
because
it
has
an
intermediate
aggregation
essentially,
and
so
it
computes
the
intermediate
result
and
then
it
merges
them.
You
don't
need
that
merge
function,
I'd
still
asked
to
have
that
merge
function
included
in
the
data
structure
because
in
the
collector
processor
you're
going
to
want
to
merge-
and
you
know
so
and
in
other
sdks
we
want
to
merge
I,
guess
that's
what
I'm
trying
to
say!
B
A
Okay,
yeah
I
will
I'll
try
to
take
a
look
at
that.
It
is
on
my
on
my
radar
I
like
we're
trying
to
get
that
in
I.
Definitely
think
that
that's
something
I
did
do
you
know
if
the
exponential
histogram
was
released
as
stable,
I
think
there
was
discussion
about
it
last
week
in
a
spec.
B
Meeting
we
got
the
1.13
spec
release
does
contain
the
modified
boundary
equation
and
that
actually
contains
a
new
zero
field
as
well
that
I
haven't
implemented
Zero
Tolerance.
There
is
still
discussion
of
of
what
the
SDK
options
would
be
to
configure
the
zero
tolerance,
I
think
and
that's
a
good
discussion
for
the
future.
A
B
Still
Market
experimental,
but
a
lot
of
us
would
like
it
the
New
Relic
group,
as
well
as
the
white
step.
People
are
pushing
for
it.
For
you
know
best
we
can
yeah
and
and
there's
certainly
reason
not
to
call
it
stable.
Riley
came
up
with
some.
You
know,
maybe
the
name's
wrong
I,
don't
know
something
like
that.
E
Or
precede
it
at
least.
B
I'm
not
sure
how
to
answer
I'm,
not
sure
what
the
right
answer
is.
My
goal
is
to
have
the
exercise
stuff
ready
to
go
before
you
know
as
soon
as
possible.
So
that's
why
that
work
with
progress
PR
to
demonstrate
what
that
aggregator
looks
like
yeah.
A
Yeah
and
then
we
want
it
to
be
stable
before
we
release
it,
I
think
I
I,
that's
I
think
that
Anthony
timing,
wise
is
going
to
be
important
here,
because
I
think
that
having
something
in
our
stable
really
set
is
not
stable
is
not
ideal,
but
I
I
do
think
that
you
could
probably
get
the
spec
to
stabilize
the
exponential
histogram
before
that,
based
on
the
conversation
that
I
think
I
remember,
but
I
also
think
that
our
stability
of
the
state
of
metrics
SDK
is
a
few
months
out.
A
So
hopefully
that
aligns
Josh
something
you
think
that
you
can
achieve
that.
Yeah,
okay,.
E
But
if
we're
going
to
merge
it,
we
need
to
know
that
we're
accepting
that
that's
going
to
then
become
a
limitation
like
if
we
do
somehow
get
to
a
point
where
we've
got
a
stable,
metrics
SDK
that
we're
ready
to
ship
and
the
exponential
Instagram
still
isn't
stabilized,
we'll
either
need
to
rip
it
out
or
wait.
A
I
honestly
I
think
if
we
release
a
point
or
a
feature,
release
with
exponential
histogram
that
isn't
stable,
it
might
serve
some
purposes
so
I
don't
know,
I
I
could
see
ripping
it
out,
but
I
also
could
see
that
that's
such
a
far
thing
off
in
the
future
that
this
conversation
is
a
little
like.
We
still
have
to
resolve
the
whole
views
being
registered
with
a
reader
thing
and
other
spec
compliance
issues.
So.
B
B
Values
are
going
to
be
ignored
like
less
than
one
microsecond
greater
than
you
know,
100
seconds
for
latency
measurement
to
keep
your
your
plate,
your
resolution,
or
you
might
have
like
a
periodic
reset
that
can't
be
too
fast,
and
that's
that's
that's
where
we
need
probably
specification
in
cooperation.
Collaboration
with
Prometheus
like
reset
once
an
hour,
throw
away.
B
B
That's
because
that's
why
I
think
these
two,
the
mapping
and
the
structure-
that's
not
stuff,
that's
covered
by
the
SDK
specification.
That's
the
data
model
essentially,
and
at
this
point
I
only
see
this
SDK
specification
changing
or
the
name
of
the
data
model
changing
but
SDK
specifications.
A
A
B
A
No
yeah
thanks
for
bringing
it
up,
though
there's
there
have
already
been
months
of
discussion
on
this,
so
it's
a
deep
one:
okay,
cool
moving
on
to
the
spec
or
I'm
sorry
and
the
agenda
Aaron.
You
have
triaging
the
metric
Bridge
next.
D
Yeah
I
just
wanted
to
put
it
out
there.
We
have
a
large.
We
have
a
large
what
you
call
it
Milestone
that
has
the
everything
we
want
to
get
into
beta
and
probably
things
that
maybe
don't
need
to
be
in
there.
D
So
I
was
proposing
that
we
put
together
a
temporary
session,
maybe
like
once
once
a
week
for
an
hour
until
we
don't
need
it
to
just
kind
of
go
top
to
bottom,
say
here's.
What
we
need
to
do
like
does
this
need
to
be
in
there.
If
not,
can
we
bump
it
out,
or
is
it
even
relevant
anymore
and
just
make
sure
that
we
have
everything?
D
I
was
probably
going
to
put
out
a
doodle
today,
just
asking
what
would
be
a
good
time
who
wants
to
be
there?
How
many
people
we
need
there
to
to
actually
make
progress
on
it
wondering
if
that's.
A
I
think
it's
a
great
idea:
I've
been
thinking
about
it
as
well.
I,
actually
don't
think
it's
going
to
be
scoped
to
the
metrics
bridge,
but
I
think
the
metrics
beta
keeps
Cambridge
I.
Think
that,
honestly,
if
you
have
a
trio
session
that
was
ongoing
for
a
while,
we
had
a
lot
of
PRS.
That
I've
been
like
looking
at
going
like
this
easy
to
be
triage
as
well.
A
I
I
think
that
the
you
need
the
stakeholders,
so
you
need
maintainers
there
at
least
one,
because
I
think
the
doodle
is
kind
of
up
to
you
me
and
Anthony
I
I
know
the
collector
does
theirs
on
Friday
at
like
9
30
in
the
morning.
If
I'm
not
mistaken,
I
have
I'll
just
say
right
now,
you
don't
need
a
doodle
like
I
I'm
free
in
the
morning
on
Friday.
If
you
wanted
to
do
something
then
as
well
Anthony
I,
don't
know
if
you
are
as
well.
E
That
varies
week.
I
think
I've
got
every
other
week.
Some
morning,
conflicts
but
usually
9
30
I
do
have
availability,
I
used
to
have
the
Click
The
Collector
tree.
Oh,
it
looks
like
that.
Just
expired.
The
invite
that
I
had
never
mind.
I
need
to
update
that
yeah.
A
E
10
a.m:
Pacific
I
think
I
just
had
a
placeholder
in
my
calendar
that
ran
out
last
week.
Oh.
A
Yeah,
oh,
it
looks
like
the
spec
does,
has
does
one
as
well
as
eight
o'clock,
so
yeah
Aaron.
If
you
wanted
to
set
up
a
meeting,
add
it
under
the
public
open
Summit
your
calendar
for
Friday
between
8,
30
and
I.
Don't
know,
10.
just
choose
a
half
hour,
not
a
whole
hour
by
any
means.
I!
Don't
think
I
don't
do
that,
but.
A
Lot,
oh
God,
yeah
yeah.
That
sounds
like
a
good
idea
to
me
and
I
say
that
those
stakeholders
are
needed,
but
anyone
I
think
should
be
welcome,
especially.
D
Yeah
I'll
put
something
together.
A
little
agenda
of
like
I
I
want
to
focus
it
specifically
on
the
beta
at
first
and
then,
if
we
have
yep
continuing
effort
afterwards,
we
can
do
the
rest
of
the
the
backlog.
There
is
plenty
but
yeah.
Okay,.
A
Perfect
yeah:
let's
do
it?
Okay,
next
up
David
with
Prometheus
Bridge.
C
Yeah,
so
for
those
of
you
who
are
at
the
spec
meeting,
this
Tuesday
there's
still
a
significant
disagreement
about
bridges
and
how
they
should
be
done
and
I
don't
think
it's
going
to
be
I'll
do
my
best,
but
I
don't
think
it's
going
to
be
resolved
that
quickly
so
I
propose.
We
do
something
simple:
to
unblock
the
collector,
and
this
is
the
this
is
what
I've
come
up
with
as
the
least
bad
thing.
C
It
involves
the
Prometheus
exporter
having
its
own
definition
of
what
a
bridge
is,
and
then
the
open
census
Bridge
simply
implements
that
so
there's
it
doesn't
touch
the
SDK
at
all,
which
is
potentially
a
benefit.
If
we
want
to
remove
this
at
some
point,
but
yeah
feedback.
C
C
Basically,
is
the
way
that
this
bridge
currently
works.
The
only
thing
we're
missing
is
pull
and
the
only
pole,
exporter,
I'm
aware
of
that
exists
today.
Maybe
is
the
previous
one.
So
this
is
sort
of
a
it's,
not
a
long-term
solution,
but
it
will
unblock
the
collector
which
uses
the
open
census.
Prometheus
exporter.
E
Is
the
collector
still
intending
to
use
a
bridge
I
know
there's
some
discussion
the
last
time
we
we
talked
about
this
of
bypassing
a
bridge
and
just
implementing
directly
open
Telemetry
metrics
export
Beyond,
feature
gate
and
moving
that
way.
C
E
D
D
D
That's
expanding
the
the
sdk's
code
before
we
come
to
an
agreement
on
on
the
bridge
code,
but
I
think
it
would
be
just
you
know
that
that
code
would
probably
just
have
to
morph
into
whatever
ends
up
being
the
accepted
path,
and
we
have
to
let
users
of
it
know
that
it
will
change
or
or
disappear
at
some
point.
D
I
actually
do
have
like
90
there,
something
that
is
just
a
bridge
that
will
let
you
take
what
looks
like
your
Prometheus
Bridge,
but
is
just
metric
bridge
and
and
use
that
with
any
reader
I
think
that
would
probably
be
a
much
much
cleaner
approach
to
this.
If
we
go
this
route.
A
Yeah
I
I'll
look
at
this
in
a
little
while
so
I
can't
say
too
much,
but
I
do
I
do
want
to
just
unders
like
make
sure
everyone
understands
like.
There
is
a
little
bit
of
urgency
here
because,
like
the
collector
is
kind
of
blocked
on
these
sort
of
things,
but
maybe
that's
not
true,
so
I
would
love
to
better
understand
that
urgency.
So
Anthony
I
know
you're
a
lot
more
involved
in
the
collector.
A
When
is
the
next
time
that
you
all
are
meeting
and
David
I
know
you
as
well,
but
whether
there's
like
a
good
understanding
of
like
how
urgent
this
is
coming
from
The
Collector.
E
So
I
think
the
The
Collector
state
would
be
the
normal
time.
We
would
talk
about
that.
I
think,
there's
also
I
know
Jacob
from
lightstep
was
going
to
be
looking
at
contributing
more
to
the
open,
Telemetry
instrumentation
metrics
instrumentation
in
The,
Collector,
Aaron
or
Josh
I.
Don't
know
if,
if
you
know
any
more
about
what
his
plans
are
in
terms
of,
does
he
want
to
use
the
bridge
or
does
he
want
to
move
forward
the
native
instrumentation
more
quickly
so
that
the
British
might
be
skipped.
B
I'll
speak
I,
I
do
know
the
answer.
I
think
you
all
may
remember.
Gustavo
in
this
group
he's
been
working
underneath
Jacob
under
Jacob's
instruction
to
do
that
work
and
has
been
giving
David
feedback
on
the
PRS
I
think
there
is
urgency
and
I
wouldn't
make
too
much
of
it,
because
anyone
who's
getting
their
hands
dirty
and
The
Collector
code
knows
that
you
could
simply
Implement
an
open
census.
Bridge.
There
are
500
ways
to
do
this,
and
it's
not
going
to
get
solved
in
two
weeks.
B
Talking
about
in
the
hotel,
Community
or
in
the
collector
meeting
would
be
a
good
idea
and
Gustavo
is
the
person
to
be
in
touch
with
directly
I
think
he
was
running
into
problems
with
the
hotel
go
Prometheus
exporter,
not
supporting
Target
info,
which
is
great
development.
Work
for
us
to
do
just
to
move
forward.
E
A
Could
you
ping
Pablo,
Josh
or
Aaron
and
just
say
like?
Could
you
take
a
look
at
David's
PR
interview
it?
Oh
I
didn't
check
if
you
maybe
already
did
that,
but.
A
Because
I,
just
it's
been
a
week
now
and
I,
want
to
have
something.
I
know
that
the
collector
meeting
also
just
happened
yesterday
so
like
it's
gonna,
be
a
whole
week
before
that
meeting
happens
again,
and
so
if
this
solution
works
for
Pablo
and
he's
going
to
work
on
it
in
The,
Collector
I
am
motivated
to
get
it
in.
A
If
we
also
wanted
to
iterate
on
it
like
it's,
not
a
stable
API
and
the
person
that
really
matters
or
is
going
to
be
impacted
by
us,
I
think
is
going
to
be
the
collector.
So
if
you
wanted
to
change
that
in
our
next
release,
I'm
okay,
with
that
too.
B
C
Think,
but
would
you
prefer
if
the
interface
just
lived
in
the
SDK
and
nothing
else
that
would
solve.
D
The
dependence
problem,
at
least
yes
I,
definitely
would
prefer
that
I
also
have
concerns
about
the
end
of
the
life
cycle
of
this
like.
If
this
is
a
temporary
thing,
we
are
going
to
pull
it
out.
D
B
The
only
thing
I
want
to
add
on
that
is
that
I
I
feel
like
the
two
groups
could
decide
to
to
collaborate
in
a
tight
way.
We
shouldn't
be
kicking
and
screaming
at
each
other.
You
know
I'm
not
actually,
probably
either
can't
really
speak
for
either
group,
but
I
feel
like
The
Collector
group
could
decide
to
use
an
API
that
we
agree
is
unstable
and
it
just
means
that,
like
you're,
going
to
synchronize
your
releases
a
little
bit
and
I
think
that's
okay,.
C
A
D
Could
try
I
can
attempt
okay
I
can't
remember
when
it
is,
but.
A
C
A
That
Josh's
point
is
well
received,
like
we
should
try
to
collaborate
on
this
and
it
seems
like
you're
the
one
most
informed
on
it.
So
I'd
like
it.
If
you
could
join
in
something
like
this
okay,
cool
well
and
Anthony's,
going
to
be
there,
so
you
have
a
co-patriot,
okay,
David.
Does
this
kind
of
give
you
a
path
forward,
hopefully
get
Pablo
letters
you
this
see.
If
this
works
we'll
have
collaboration
with
the
collector
next
Wednesday
and
then
by
next
Thursday.
A
Awesome
with
that
I'm
gonna
progress,
this
Jamie
you're
up
next
with
the
auto
export
PR
I
can
start
sharing
my
screen
again.
F
Yeah
so
I
I,
don't
think
I
was
here
last
week,
I've
been
kind
of
all
over
the
place,
the
last
few
weeks,
so
I'm
kind
of
losing
track
of
time.
F
Here,
I
think
when
the
auto
exporter
PR
was
put
up
into
contrib,
we
may
have
accidentally
left
it
as
work
in
progress
when
the
idea
was
more
that
I
do
think
it's
in
a
spot
to
get
more
feedback
and
see
if
this
is
a
good,
a
good
thing
to
get
merged
in
with
the
idea
of
over
time,
adding
more
to
it
if
needed,
like
I,
said,
I
think
we
we
just
took
off
the
work
and
progress
label
like
yesterday
or
the
day
before,
or
something
like
that.
F
So
just
wanted
to
kind
of
point
out
that
it
is
here
and
ready
for
some
eyes
to
get
feedback
from
different
people
and,
and
the
idea
again
of
the
auto
exporter
is
that
it
works
the
same
way
as
the
auto
prop
package,
where
you
have
a
registry
of
exporters
that
you
can
use
and
that
should
simplify
the
setup
and
ultimately,
obviously,
this
had
come
up
in
conversations
when
working
on
the
hotel,
Hotel
init
slash
launcher
package,
whatever
we
haven't
fully
decided
what
we're
calling
it,
but
the
goal
would
be
that
this
could
be
used
on
its
own.
F
Just
like
the
auto
prop
package
can
be
used
on
its
own,
but
then
once
the
launcher
is
is
agreed
upon.
It
would
then
also
use
the
auto
prop
package.
It
would
use
the
auto
exporter
package
so
that
we
have
a
few
different
ways
of
initializing,
open,
Telemetry.
So
I
think
this
is.
This
is
probably
the
the
main
thing
looking
for
review
and
then,
as
a
secondary
on
that
I
guess,
I
kind
of
had
a
question
related
to
the
design
doc.
F
That
is
that
the
right
place
to
talk
about
what
that's
going
to
look
like
or
should
I
keep
it
mostly
words
and
less
a
visual
of
what
it's
going
to
look
like
so
like
we
have
the
motivation,
What,
It,
Is,
What,
It
Isn't,
that's
all
kind
of
clear
as
a
concept,
but
then
we
have
the
configure
section,
for
example,
and
it
might
be
yes,
we
absolutely
want
to
know
exactly
what
it's
going
to
look
like
here,
or
is
this
something
that
should
be
saved
for
an
actual
once
the
design
is
approved,
sort
of
a
thing?
E
Yeah
I
think
so
you're
asking
should
should
there
be
code
like
this
laying
out
what
you
yeah,
I
I,
think
that's
super
useful
to
be
able
to
reason
about
the
design
is
to
see
how
it
will
be
used.
That's
one
of
the
ways
I
find
test,
driven
development
to
actually
be,
as
well
as
more
as
design
exploration
rather
than
testing
right.
So
this
is
how
I
want
to
use
the
thing
can
I
make
that
work
yeah.
This
is
good
for
that.
F
Too,
okay,
cool
yeah,
so
yeah
I
have
like
something
local
here
that
I
want
to
push
up.
That's
going
to
change
these
a
little
bit,
obviously,
because
yeah
for
a
similar
thing
like
we
have
a
lot
that
we're
using
currently
on
a
branch
and
I,
definitely
find
it
easier
to
reason
about
code
than
words.
F
Sometimes
when
it
comes
to
you
know
how
we're
talking
about
how
this
is
going
to
work
so
I'm
just
trying
to
balance
the
two
things
together
in
this
stock,
so
that
it
makes
sense
without
being
like
here's,
the
entire
implementation
of
of
how
it's
going
to
go
so
yeah
I
guess,
go
ahead.
A
No
I
was
just
gonna,
say
yeah
that
all
makes
sense.
I,
I,
agree,
I,
think
that
having
code
also
I
just
want
to
point
out
that
I
wouldn't
necessarily
say
that
this
has
to
be
perfect
right
off
the
bat
just
giving
it
a
flavor
of
how
this
is
going
to
look
and
when
you're
actually
implemented
it.
Maybe
a
little
different
I
think
you're
you're
spot
on,
and
your
approach
to
design
doc
right
here.
F
Okay,
so
then
I
guess
right
now
again
the
the
auto
exporter,
PR
I,
think.
Is
there
ready
for
review
I'm
hoping
to
get
an
update
into
the
design
doc
today
to
point
to
those
changes
that
will
be
in
the
auto
exporter,
PR
and
I'll
put
a
note
on
there
that
says:
hey
everyone.
F
If
you
can
take
another
look
because
I
know
it's
been
a
while,
since
we've
really
gone
through
it
and
talked
about
it
in
more
depth,
so
I'll
I'll
sort
of
ping,
everyone
on
there
of
okay,
it's
ready
to
have
another
look
and
then,
if
there's
anything
else
that
comes
up
that
is
worth
looking
at
that
stops
IT
from
moving
forward.
Just
let
me
know
comments
on
the
pr
things
like
that
is
super
useful
foreign.
A
A
Okay,
I
think
we're
almost
under
the
agenda
last
thing
I
listed
kubecon
on
here
I
did
come
to
the
conclusion
that
I'm
not
going
to
be
going
this
this
week,
so
I
won't
be
there
to
see
y'all,
but
I
did
want
to
call
it
out
like
for
the
people
that
are
going
actually.
Who
is
going
on
the
call
I
know
Aaron's
gone,
David's
gone
anybody.
E
A
Anthony's
gone
yeah
I
know
that
there
I
think
it's
like
Tuesday.
They
said
there's
supposed
to
be
like
an
Hotel
get
together
or
something
like
that.
I,
don't
know
what
it
is,
but
yeah
I
just
wanted
to
make
sure
everyone
was
gonna,
go
with
sinking,
I'm
kind
of
bummed
I'm
not
gonna,
be
able
to
go,
but
just
wasn't
able
to
make
it
work.
So
next
time,
yeah.
E
Hotel
unplugged
I
think
is
what
they've
been
talking
about:
there's
a
channel
on
the
cncf
slack
regarding
that,
if
anybody's
interested.
A
A
Right
click
then,
there's
a
copy
link.
A
Well,
cool,
nothing
else
on
the
agenda.
I'll
just
pause
here.
Anybody
else
wanted
to
bring
something
up.
There's
a
lot
going
on.
A
Okay,
awesome
there
any
cool
use
cases
of
open,
telemetry.
A
I
remember:
I
had
shared
something
a
while
ago
and
I
yeah
I.
Don't
think
this
was
ready.
Yeah,
okay,
here
I
got
something.
Let
me
see
yeah
a
few
weeks
ago,
it
looks
like
now
I
created
this
thing
that
we
kind
of
always
talked
about
using
an
open,
Telemetry,
collector
Factory
to
export
without
some
trick
out.
It's
kind
of
yeah
I
know
right.
There's
it's
really
a
work
in
progress
right
now,
I
think
that
it
works,
but
that's
by
everywhere.
A
A
So
that's
where
the
the
main
meat
of
the
project
is
but
yeah
I
thought
it
was
kind
of
a
cool
idea
just
because
if
you
haven't
already
taken
a
look
at
the
collector
ecosystem,
there's
a
ton
of
exporters
there
and
it
currently
Works
where
we
can
send
stuff
from
otilgo
to
The
Collector.
And
then
you
could
use
those
exporters
but
we'd
always
kind
of
talk
about
this
as
being
just
a
translation
layer
in
our
own
ecosystem.
A
So
I
I,
guess
maybe
just
if
you
want
to
take
a
look
happy
to
collaborate.
Yeah
Josh
has
his
hand
raised.
B
So
Anthony
probably
knows
this.
Yesterday,
at
the
collector
meeting,
I
presented
a
feedback
and
a
sort
of
show
and
tell
about
this
library
of
code
that
I've
been
adapting
to
use
the
collector
P
data
package.
The
reason
I'm
doing
that
is
I
have
this
piece
of
code
in
go
that
is
to
export
from
The
Collector
into
Apache,
Arrow,
so
column,
oriented
data
structure
and
try
and
get
into
that
ecosystem
of
Apache
data
store
data,
Stores
and
but
but
the
main
goal
for
us
is
to
get
compression.
B
But
if
the
hotel
go
and
the
hotel
collector
groups
could
agree
on
a
single
protocol
definition,
then
potentially
the
hotel
go.
Sdk
could
use
it
for
export
and
then
I
would
be
able
to
swap
in
my
hotel,
Aero
adapter
and
you
would
be
getting
Hotel
Aero
export
using
the
hotel
collectors
exporter,
which
uses
my
large
Library
code
that
I'm
adapting
from
somebody
else
anyway
point
is.
A
A
Yeah
I
I
think
it's
a
common
desire.
I
I
know
that
there's
like
an
issue
because
our
span,
export
interface,
I
think
embeds
our
data
model,
so
we'd
have
to
essentially
add
to
that.
If
we
wanted
to
do
something
like.
A
E
I
think
it's
interesting.
We've
basically
come
full
circle
to
as
I
know,
some
of
the
initial
collector
exporters
were
written
as
rappers
around
go.
Exporters
honeycomb
had
done
that
I
believe
they
implemented
one
for
the
gosck,
then
adapted
it
to
to
a
collector
exporter,
eventually
got
rid
of
that
for
otlp,
but
now
we're
doing
it.
The
opposite.
A
B
Of
an
interesting
question,
how
much
would
break
or
how
breaking
is
it
for
you
to
change
the
export
signature
that
currently
uses
a
compiled
by
the
Google
official
protobuf,
compiler
and
and
to
use
the
one
that
the
collector
is
compiling?
That's
currently
in
an
internal
P
data,
internal
data
package
or
protogen,
or
whatever
it's
called?
B
If
that
was
I
mean
the
type
name
is
the
same,
but
it's
not
the
same
thing,
but
wouldn't
that
really
help
if
we
could
all
share
a
definition?
Well,
anyway,.
E
B
B
I'm,
actually,
no
I
mean
I'm
like
there's
right
now.
There's
an
Hotel
Trace
H,
there's
an
Hotel
Trace
HTTP
and
a
hotel
trace
grpc
and
those
are
the
like
grpc
clients
and
the
HTTP
clients
right,
and
they
take
these
protobufs
that
are
compiled
by
Google
protobuf.
A
E
A
E
I
think
this,
this
translation
would
happen
before
that
this
this
would
be
like.
Instead
of
using
the
otopx
corner,
you
would
use
a
collector
exporter
which
would
translate
from
the
read-only
span
to
P
data
and
then
pass
it
to
a
collector
exporter,
so
it
would
never
go
through
otlp
or
wouldn't
necessarily
need
to,
except
by
the
fact
that
P
data
is
a
wrapper
around
otlp.
A
Yeah
and
then
like,
but
to
Josh's
Point
like
if
we
could
refactor
to
use
that
I
think
that
it
would
open
up
the
export
capabilities
in
the
future.
But
I
don't
know
so.
I
I
really
don't
like
stifling
the
creativity,
but
we
are
at
time
and
I
want
to
be
respectful
of
everyone's
time.
So
hopefully
it
inspired
thoughts
and
nerds
types
you
for
the
rest
of
the
day.
But
if
not
we'll
see
you
all
next
week
or
asynchronously
bye,
all
right.