►
From YouTube: 2020-06-04 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
Hey
Harlan
nice
to
meet
you
I'm
now
I'm
working
at
Google,
I
involved
in
the
open
country
project
I
haven't
much
contributions
to
the
go
SDK
yet
mostly
done
some
stuff
on
the
collector
side,
but
in
general,
if
there
are
some
fun
stuff,
but
it
is
high
priority.
I
have
one
of
the
people
from
the
Google
who
may
be
able
to
help
me
cool
probably
have
already
met
chase
before
case
yeah.
B
Yeah
I
met
James
briefly
before
I've
seen
a
few
more
times
in
to
collect
your
meetings.
Yeah,
it's
cool!
Are
you
guys,
both
in
Australia
yeah,
yeah
yeah?
We
could
we
could
kind
of
dive
into
some
issues.
There's
I,
don't
know
what
your
appetite
for
doing
things
like
it's,
nothing,
critical,
I,
guess
a
good
way
to
put
it.
So
it's
up
for,
like.
B
C
And
that's
fun
for
me
how
many
people
you
go
with
on
the
to
go
a
stake
and
love
them
by
the
way
cuz
I
noticed
these
six
are
relatively
small.
B
B
A
B
B
C
C
B
We
definitely
the
help
would
be
useful,
like
I
think
the
reviews
are
also
kind
of
a
can
be
problematic
right
now,
for,
like
the
past
week,
it's
kind
of
slowed
down
a
little
bit,
I
think
just
with
some
general
political
issues
that
are
happening
here.
People
are
kind
of
slowing
down
their
involvement
work
in
general,
but
yeah
a
lot
of
the
times.
Some
of
the
PRS
will
sit
there
for
a
while
event,
mute
review
or
a
very
limited
number
of
use,
I
think.
B
As
for
the
specification
implementing
a
specification
where
three
issues
in
the
way
from
implementing
the
least
error
for
specification,
the
VCR
five
just
came
out
as
well,
but
I
haven't
gone
through
the
details
of
the
five,
but
I
know
that
the
major
the
both
of
the
v05
was
implementation
of
the
metrics
and
since
go
is
kind
of
the
prototype
of
english
to
be
doing
them
in.
If
we're
already
really
close
part
of
the
specification
as
well
so
I,
don't
think.
There's
too
much
there.
B
B
C
But
yes,
I,
sort
of
I
think
they
put
like
eight
in
turns
on
C++
or
something
like
that.
And
then
it's
about
five
or
six
on
each
other
language
looks
looking
go
yeah.
There's
eight
people
on
go,
but
they
kind
of
have
quite
specific
things.
I
think
sit
like
two
of
them
are
working
on
the
Google
stack
driver,
Explorer
and
then
a
couple
of
working
in
the.
C
B
C
B
A
B
B
Of
are
contributing
guidelines
as
well.
I
wanted
to
try
to
get
in
on
but
yeah
as
for
like
issue
wise
I
see
Niall
you're
Niall
box
on
github
right.
Yes,
yes,
I.
Your
latest
issue
that
you
opened
with
us,
793
I
was
like
that
was
actually
kind
of
an
interesting
one
harder
to
talk
to
talk
about
today.
So
I'm
kind
of
interested
that
you
showed
up
if
you're,
if
you're
up
for
it,
because
I
was
late,
yeah,
I'm,
sorry,
oh
yeah,.
B
B
Looks
like
Rasmus
to
be
honest.
I
was
trying
to
read
through
this
issue
that
spawned
it,
and
it
makes
a
lot
of
sense
as
to
like
the
history
on
it.
But
I
was
kind
of
wondering
what
kind
of
wondering
what
happened
between
like
here
on
March
1st,
when
we
had
it
like
trying
to
work
in
this
open,
Salome
tree
external
repo,
or
something
like
that,
and
then
like
the
idea
to
switch
to
just
going
to
having
with
the
collector
yeah.
A
To
be
honest,
I'm
I'm,
not
sure
if
that
like
decision
is
justified,
but
if
they
actually
have
some
optimizations
on
the
wire
like
on
the
generator
you've
got,
probably
probably
not
so
I,
don't
think
it
actually
why's
there,
but
I
guess
potentially
they
might
seem
like
the
implementation
inside
may
differ
from
the
generator
something
in
this
room.
Yeah.
B
B
A
Client
and
server
should
should
be
compatible
with
each
other,
so
I'm
not
sure.
Actually
what?
What
exactly
did
you
mean
there?
But
I
guess,
given
that
we
now
already
have
two
copies
of
the
generated
text
and
the
go
SDK
is
the
only
one
which
uses
the
generator
types
from
from
the
protocol
repository.
Maybe
it
makes
sense
to
actually
completely
decouple
them
to
make
it
more
explicit,
because
at
the
moment
we
have
this
protocol,
which
seems
like
a
share.
A
B
I
think
that's
it's
probably
fair.
You
had
some
really
good
points
about
how
it's
pretty
common
to
I
mean
especially
to
store
this
in
this
repository,
and
they
liked
the
idea
of
having
Ana
centralized
place
kind
of
like
what
exactly
what
you're
putting
out
so
just
to
know
drift
in
between,
like
some
sort
of
any
use
of
it,
like
it's,
a
really
well-established,
like
centralized
place
where
there's
like
a
source
of
truth,
but
I
haven't
been
able
to
work
through
this
entire
conversation
really
understand
it.
I
imagine.
A
B
Asking
us
to
do
very
similar
things
and
have
a
Sauvage
role
like
the
collector
but
yeah,
especially
with
like
the
the
proto
there's
supposed
to
be
some
a
lot
of
changes
which
we
are
talking
about
a
lot
in
the
metrics
so
much
but
yeah.
It
worries
me
a
little
bit
when
there's
gonna
be
like
these
large
changes
in
these
kinds
of
situations
where
it's
like.
A
Yeah
I
have
actually
like
contributed
to
the
proto
like
this
generation
code.
I
clean
did
not
belittle
me
again
make
sure
that
CI
now
actually
validates
the
the
gene
record
files
up-to-date.
Previously
they
were
like
completely
manual,
so
you
can
just
you
know,
push
free
or
protein
messages
without
generating
types.
Then
now
you've
dated
gogo
types
in
the
same
version
yeah
it's
a
bit
cleaned
up,
but
yeah
I
guess.
The
question
remains
whether
we
should.
A
B
A
B
B
A
A
Difference
between
them,
even
though
they
share
some
of
the
functionality
like,
for
example,
the
other
issue.
I'm,
not
sure
it's
like
it's
race
as
I,
get
a
beach
anywhere,
but
I
think
many
people
have
picture
notice
that
confusing,
but
that
SDK
exporters
have
completely
different
term
spaces
from
the
collector
exporters,
even
though
they
pretty
much
got
the
sinkhole
exploding
somewhere
and
sometimes
like,
for
example,
in
our
example,
in
Google
Cloud,
we
have
implementation
with
the
SDK
exporters
to
state
driver
api's
and
then
for
collector.
A
We
have
like
a
completely
separate
implementation
of
the
exporters
which
effectively
do
the
same,
so
they
need
two
expert,
metrics
and
traces,
but
because
of
the
different
interfaces
we
have
to
evacuate
to
implementations
or
to
say
the
same
thing.
It
would
be
nice
to
have
some
sort
of
a
breach
where
you
can
implement
only
a
SDK
exporter-
and
just
say
you
know,
use
this
like
proper
thing
to
translate
it
to
a
color
interface
yeah.
B
B
So
there's
like
a
lot
of
translation
there,
which
is
extremely
wasteful,
I,
think
on
the
back
end,
your
suggestion
that,
like
yeah
that'd,
be
like
some
sort
of
a
wrapper
like.
Maybe
you
can
take
the
SDK
exporter
types
and
then
wrap
them
to
you,
know,
use
them
also
with
the
hotel
key
export
or
the
collector
exporter.
People
also
suggested
the
opposite
as
well,
where
you
know
there'd
be
some
rapper
on
the
LP,
which
would
then
translate
it
to
work
with
the
SDKs
yeah.
B
It's
a
little
bit
of
a
it
is
definitely
based
in
history
and
there's,
not
a
really
clear
solution
here.
I
do
think
that
it's
not
a
big
enough
point
where
a
lot
of
people
have
decided
to
like
really
dig
in
to
try
to
solve
the
issue,
but
yeah
you're
you're,
noticing
it
point
that
a
lot
of
people
have
expressed
and
as
someone
who's
written
both
exporter
for
the
SDK
as
well
for
the
collector,
like
I'm
with
you
like
it's,
it's
frustrating,
it's
a
tough
cognitive
overhead.
B
B
It's
driven
by
a
different
group
of
people
in
some
ways,
there's
some
autonomous
from
the
decisions
that
are
made
in
the
more
unified
open,
Solana
tree
specification
because
they
are
in
a
somewhat
isolated
ecosystem
and
they've
inherited
a
lot
of
that
from
open
census.
You
know
they
took
a
really
iterative
approach
and
just
going
from
a
consensus
and
slowly
migrating
that
to
a
specific
thing,
with
performance
being
a
really
big
North
Star
forum.
B
Causing
some
friction
points
I
think
one
of
the
ones
that
this
is
manifesting.
It
is
in
the
OSI
LP
currently
like
a
lot
of
the
metrics
discussion
is
around
supporting
open
salami
tree
metrics
more
uniformly
in
a
thing
that
was
driven
by
the
collector
so
yeah,
it's
not
there
are
there.
Many
different
leg
facets
of
the
same.
Like
first
point.
B
And
look
if
you
have
suggestions
on
how
to
like
fix
this
or
if
you
wanted
to
really
try
to
dig
into
it.
That'd
be
awesome,
I
think
a
lot
of
people
that
would
enjoy
it,
but
I
definitely
think
that
we
should
probably
got
to
propose
a
little
together.
If
you
wanted
to
do
that
before,
you
would
actually
try
to
sum
it
up
here,
because
people
are
gonna,
have
opinions
for
sure.
B
Yeah
I
I
would
love
it
if
they
were
so
divergent
and
like
it
if
they
had
very
similar
conceptual
models.
I
know
it'd
be
really
cool.
If
we
had
some
way
to
do
some
sort
of
pass
through
where
there
wasn't
really
any
need
to
rely
on
a
translation,
maybe
it's
it
into
our
the
SDK
models
or
into
the
collector
TLT
like
I.
Think
there's
a
common
like
I
would
like
it
to
be
easier
and
I'd
like.
B
If
did
not
have
a
third
repo
that,
like
has
a
dependency,
you
know
binaries
even
bigger
kind
of
thing,
but
it's
it's
tough
to
understand
how
that
is,
gonna,
look
and
I
think
it
just
takes
a
little
bit
of
effort
in
somebody
driving
that
issue.
I
know.
Josh
is
also
kind
of
expressed
this
in
the
past
that
he
would
like
to
see
there's
some
more
overlap
and
cohesion
across
the
project,
but
yeah
on
the
other
side
like
they're
moving
fast.
They
have
a
lot
of
developers.
B
D
B
And
I
think
it's
it's
they're
doing
some
really
cool
things.
I
think
the
idea
that
there
I
don't
actually
know
where
teacher
came
from,
but
he
sounds.
He
sounds
like
a
Googler
like
you
guys,
like
he's
just
super
into
massive
optimizations
and
like
keeping
pigs
and
wire
format
to
like,
have
no
overhead
whatsoever
I'm
just
going
like
wow.
That's
that's
ambitious
and
so
I
pressed
in
some
ways,
but
it's
also
like
okay,
kid,
like
I
gotta,
use
the
thing
so
yeah
there's
been
a
lot
of
discussions
on
that,
one
for
sure.
C
B
It's
and
I
say
that
it's
pretty
impressive,
but
yeah
I,
don't
know
yeah,
so
definitely
I.
That
makes
sense
the
direction
they're
going
there.
I
don't
know
if
there's
a
way
that
you
could
also
try
the
cohesively
restructure
things
around.
There.
I've
talked
a
little
bit
about
with
them
about
this,
but
I
actually
ended
up
closing
issues
that
they
just
don't
have
the
time
to
really
drive
this
around.
B
Maybe
there's
a
way
we
can
wrap
things,
so
it
supports
both
a
more
maintainable
solution
or
more
useable
solution,
or
maybe
even
something
that
looks
analogous
to
the
tree
custody
cases,
the
translations
easier
or
even
transparent,
or
something
like
that,
but
I,
that's
again,
it'd
be
really
cool.
It
I
think
it
just
takes
somebody
really
digging
into
it
and
and.
B
B
C
Mean
if
someone
is
just
gonna
have
to
require
some
sort
of
translation
between
data
types
right,
which
kind
of
comes
back
to
that
wasteful
thing,
but
but
it
it
sort
of
it
kind
of
sort
of
feels
like
yeah.
If
someone's
running
an
exporter,
it
would
be
nice
if
they
had
the
option
if
they
want
to
arrive
to
super
ones.
They
want
it
to
be
ultra
performant
and
not
do
another
translation.
They
can.
B
B
B
Would
also
have
to
move
a
little
because
the
current
structure
over
there
is
not
ideal.
So
I
think
that
like
if
we
could
do
that,
there
could
be
some
optimization
in
front
of
us
and
I
and
that's
why
I
can't
say
like
the
future
is
a
little
bit
more
open,
because
the
everything
like
API
like
at
the
front
like
that
means
to
start
becoming
more
solidified
and
I,
think
that
that
is
becoming
more
solidified.
B
There's,
like
some
minor
cleanup
I,
think
that
needs
to
happen
before
the
general
release,
but
the
metrics
and
the
tracing
stuff
is
starting
to
look
like
I.
Think
what
we're
proud
to
release
at
least
broad
structure.
But
that
being
said
like
the
SDK
kind
of
back,
is
still
still
fluid.
I
think
the
further
back
you
go
like
the
closer
to
the
exporter
side.
It's
much
more
fluid
so
like
if
we
wanted
to
make
sure
make
that
translation,
where
eventually
the
SDK
is
just
emitting.
B
A
A
C
C
Sorry
I've
actually
started
off
just
wanted
to
implement
ruthless
diction
and
they
collector-
and
you
know
this
trying
to
figure
out
the
best
way
to
go
about
it
and
there's
lots
of
kind
of
slightly
inconsistent
implementations
that
were
open
since
this
logic,
so
I
made
a
no
tip
in
the
end
yeah.
You
have
some
time
to
look
at
that.
Coming,
though,.
B
B
A
B
B
That's
probably
not
gonna
be
okay
from
the
community
standpoint.
Like
some
I
know,
some
code
bases
are
not
as
strict
on
this
matter,
but
I
know
that
a
lot
of
other
standards
are
that,
like
any
vendor,
specific
code
needs
to
live
in
some
sort
of
other
repository
like
any
sort
of
import
of
like
the
standard,
open,
telemetry,
SDK
or
API,
is
do
not
include
vendor
specific
code,
but
I
think
that
the
rules
have
been
bent
in
the
past
and
I.
B
Think
that
like,
if
you
make
a
strong
argument,
maybe
that's
fine
I
think
it's
just
like
you.
Don't
want
to
get
into
the
situation
where
you
have
vendors
clashing
inside
of
a
code
base
that
supposed
to
be
open
source
or
you
have
any
sort
of
like
competitive
thing
or
there's
like
a
added
burden
to
the
open
source
developers
to
maintain
it's
not
doing
anything
for
the
project.
C
Yeah
kind
of
a
chance
to
me
as
well:
that's
what
I
propose!
That's,
how
I
sort
of
wrote
this
proposal
the
shooting
that
way,
but
as
like
Jay
pointed
out
most
like
I,
think
aw
asks.
Gcp
is
your
to
like
to
take
results
with
information
you
literally
either.
Just
if
it's
on,
like
GK
I,
think
you
look
up
an
environment
variable
that's
on
GCE
or
like
ec2,
you
just
call
any
and
like
a
known,
endpoint
and
get
back
a
text
responsive.
C
B
I
think
yeah
I'm
interested
to
be
cool,
I
think
there's
some
possible,
like
generalization,
probably
kind
of
like
Prometheus
and
how
it
has
like
the
surface
of
scary
stuff
where
eventually
like
they
had
like
there's
so
many
things
in
there.
They're
just
like
I,
can't
add
any
more
discovery.
Agents
to
this
code
base
like
we
need
to
just
like
unify
and
I
think
if
they
eventually
turn
it
into
like.
You
want
to
add
another
discovery
into
just
output
it
to
a
file
like
read
from
the
file
and
I.
Don't
think
it
that's
ideal.
B
B
Yeah,
absolutely
and
like
I,
said
I'm
hoping
to
get
into
better
structuring
up
some
of
the
issues,
but
super
open.
If
you,
if
you
find
any
issues,
as
you
guys
have
already
been
doing
just
you
know,
raising
him
the
awesome
and
jumping
in
definitely
feel
empowered
to
go
in
there
and
make
some
statements
like
I.
Think
that
we're
looking
for
contributions
and
I
think
that
any
sort
of
diversity
ideas
is
really
helpful
in
the
project
and.
C
Definitely
yeah
I,
don't
let
that's
something
we
can
really
make
any
commitments
on.
How
much
will
be
contributing
and
I
think's
like
two
or
three
weeks,
but
we'll
keep
coming
to
the
SIG's
a
few.
If
you
guys,
you
know,
Phil
are
you're
getting
behind.
We
can
help
contribute
more.
We
obviously
very
invested
in
making
sure
the
state
gets
completed.
C
B
Yeah
cool
that
sounds
good
yeah,
like
I,
think
there
are
definitely
other
things
that
are
worse
off
in
implementation
aspect
than
us
we're
one
of
the
few
kind
of
far
along
so
I
think
that's
probably
at
least
hopefully
that
doesn't
change
with
that
lack
of
other
people
who
was
really
good
at
getting
like
small
little
things
incrementally
like
into
the
codebase
and
just
like
getting
it
done
kind
of
thing,
so
yeah
yeah,
yeah
cool,
I'm,
open
to
hanging
out
it.
Doesn't
we
don't
have
to
go
the
whole
hour
for
sure.
A
A
A
Had
one
for
them,
maybe
you
don't
haven't
been
asked:
what's
the
right
way
to
pass
the
boolean
values
in
metrics,
so
I
found
a
different
way
in
resource
attributes.
The
boolean
type
is
supported,
so
it's
like
key
value
attributes
in
resource.
You
can
actually
pass
not
just
strings
but
also
integers
and
boolean
metrics.
You
can
only
pass
integers
and
fold
I.
A
If
there
is
any
guideline
of
like
what's
the
communica
way
of
passing
billions,
you
need
because,
like
just
for
the
context,
instead
driver
ID
is
on
the
backhand
side.
They
actually
do
support
bullets
and
in
their
scores
you
can
actually
use
that
as
a
boolean.
So
you
need
to
either
translate
the
integer
to
the
boolean
on
the
back
end
slide
or
yeah.
B
Yeah,
so
let
me
just
make
sure
I
got
it
correct,
like
there's
six
metric
types,
but
currently
those
are
augmented
with,
like
the
actual
data
values
that
you
can
actually
pass
in
currently
weeks
or
float64,
and
then
in
64
and
you're,
saying
like
what
about
billions
as
well.
B
Okay,
yeah
I
think
if
there
was
a
it's
just
been
that
those
are
the
cuz
the
defaults
and
those
are
like
the
biggest
asks
right
now.
There's
there's
always
been
on
the
background
in
the
back
burner,
or
so
my
questions
as
to
what
other
types
of
keep
we
support
and
in
that
sense
there's
Williams
have
been
pointed
out.
Enums
in
general,
it's
pretty
string
values
as
well,
which
are
those
are
the
most
interesting
one
to
me.
I
still
not
exactly
sure
how
that
it's
gonna
be
useful
order.
B
The
did
cardinality
that
you're
storing
those
would
be
just
insane
back-end,
but
maybe
people
have
some
optimizations
but
yeah.
Those
have
been
talked
about
in
the
past,
but
there
hasn't
been
a
strong
preference
for
them
and
I
think
that
the
solution
that
you
just
described
where
you
could
use
like
an
inn
and
somehow
signify
in
your
exporter
or
knowing
your
exporter,
that
you
want
to
store
that
as
a
boolean
I
think
is
really
useful.
B
But
in
saying
that
it's
it's
that's
a
hardcore
translation
right,
because
you
have
to
have
operate
knowledge
that
the
instrument
or
intended
this
to
be
shown
as
a
boolean
value
and
then
in
stackdriver
I
can
actually
show.
This
is
a
boolean
value
and.
A
Yeah
I
was
wondering
with
like
I
guess
the
most
potentially
hunky
way,
but
which
is
generic
is
to
pass
some
resource
attribute
or
some
label
where
you
would
say
that,
like
you
know,
treat
these
particular
metric
as
a
doing
something
but
yeah
like
in
the
matter
effectively
say
like
have
a
sign
or
special
kind
of
attribute.
That
is
like
conventional
thing.
If
this
attribute
is
present,
then
you
treat
integer
as
a.
B
B
B
Yeah
I
don't
have
any
immediate
reasons
as
to
why
I
would
say,
like
don't
do
this,
especially
if,
like
they're
needed
the
string
value
one's
a
little
bit
weird?
Well,
that's
not
good
I,
don't
know
if
many
people
are
gonna
be
able
I
work
for
New,
Relic
and
I.
Don't
think
we'd
be
okay
to
that,
maybe,
although
it
probably
best
some
things
up,
doing
something
like
that,
but
yeah
the
boolean
one
I
think
is
a
reasonable
ask.
Do
you
have
like
a
a
useful
example
like
a
real-world
example
where
boolean's
are
used?
A
We
do
have
an
example,
but
that's
actually
from
different
team
I
have
too
much
details.
They
were
just
saying
but
like
so.
If
they
do
I
think
they
are
using
currently
local
consensus,
which
or
something
I
can't
remember
something
that
actually
support
to
do
it
and
they
will.
They
were
considering
migrating
to
go
SDK
and
that's
the
basically
global
they
have
found.
So
the
question
was
basically
like:
either
they
need
to
only
like
beckon
side
in
this
tech
driver.
A
A
A
B
A
B
So
if
you
want
to
create
an
issues,
probably
the
specification
probably
be
a
good
place.
If
you
had
a
proposal
already,
then
putting
in
is
a
good
place.
There's
there's
like
this
bureaucratic
nightmare,
sometimes
like
attempts
was
created
to
handle
larger
issues
before
they
moved
the
specification.
So
there
should
be
consensus
there,
but
some
people.