►
From YouTube: 2021-09-16 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
C
B
C
A
All
right,
as
usual,
please
add
your
name
to
the
attendees
list.
It
says
here
we
have
nine
people
such
as
copied
from
last
week.
I'm
pretty
sure
we
don't,
but
all
good.
C
We
have
a
couple
of
hidden,
zoom
account
joined,
yeah
man.
I
don't
know
those
in
your.
A
A
Yeah,
I'm
just
a
maintainer
man.
I
don't
know
anything
about
super
user.
It's
a
new
level.
Anyways
okay
looks
like
we
have
a
couple
topics
and
pr's
on
our
agenda
yeah
as
usual.
If
you
want
to
get
your
pr
streamlined
or
issues
looked
at,
please
add
them
here.
A
Right
yeah,
so
if
you
want
to
get
your
pr
reviewed,
our
issues
addressed,
it's
the
best
place
to
do
it.
If
you
want
to
streamline
them
start
off
with
our
first
topic,
looks
like
it
was.
A
Oh,
this
is
such
a
long
read
error.
Oh
I
gotta
do
this
all
right.
Do
you
want
to
talk
about
this
for
a
bit.
D
It's
mostly
quoted
text
here,
but
yeah,
so
I
don't
know
how
familiar
everybody
is
with
the
resource
semantic
conventions.
I
know
we
populate
a
few
of
these
already,
as
you
can
see
from
the
code
I
copy
pasted
in
there.
That's
what
happens
if
you
just
use
the
resource.create
by
default
right
now,
you'll
notice
in
the
spec.
D
It
says
this
thing
that
the
triplet
of
service,
dot,
name,
space,
service,
dot,
name
service
line
since
id
must
be
globally
unique,
so
we're
not
setting
any
of
those
by
default
which
violates
that,
because,
if
you
just
set
up
two
sdks,
then
you're,
you
know
you're
not
gonna
have
a
global,
unique
resource.
So
this
isn't
a
big
deal
for
tracing
right
now,
but
with
metrics.
D
Every
resource
field
contributes
to
like
the
metric
identity
and
they
would
become
essentially
like
metric
labels
in
the
sense
that
they
would
give
you
a
separate
time
series.
So
there's
this
principle
that
if
you
have
two
sdks,
they
can't
write
to
the
same
time
series
which
we
would
be
violating
right
now.
So
this
is
something
we're
gonna
have
to
address,
and
it's
just
something.
We
have
been
consistent
with
the
spec.
D
I
think
now
like
when
I,
when
I
read
the
spec,
it's
not
clear
to
me
if
we
should
generate
a
unique
uuid
from
the
for
the
service,
that
instance
id,
but
it
does
give
that
as
an
option.
I
don't
know
if
they're
saying
hey,
the
user
should
generate
that
or
if
they're
saying
that
sdk
should
generate
it,
but
I
have
a
really
hard
time
believing
the
user
is
going
to
do
that
unless
we,
you
know,
make
it
really
easy
or
obvious
for
them.
A
Well,
I
mean
if
it's
in
the
spec
but
you're
just
you're,
just
the
point
of
contention
is
just
the
default
behavior
right.
If
it's
not
supplied.
D
Yeah
and
then
I
have
another
another
proposal,
so
you'll
notice
that
also
the
spec
also
says
something
about
the
the
service
line,
since
the
id
can
be
persistent.
D
Sorry,
what
does
it
say?
It's
a
sorry.
D
Sorry
so
it
says:
if
the
service
has
no
inherent
unique
id,
you
can
generate
a
uuid,
but
if
the
service
does
have
an
inherent
unique
id,
you
can
use
that,
in
which
case
we
wouldn't
need
to
generate
one.
So
the
question
is:
how
do
we
make
it
so
that
it's
easy
for
people
to
provide
this
if
they
have
like
an
inherent
unique
id
for
a
scaled
instance
of
their
service,
and
then
the
other
question
is
yeah?
Should
we
should
we
generate
it
if
they
don't
provide
it.
D
Or
something
right
right,
so
that's
one
thing
I
was
thinking
like
there's.
We
have
all
these
other
resource
semantic
conventions,
so,
for
instance,
if,
if
somebody's
using
like
kubernetes
and
they
have
a
container
id
or
something
like
that,
we
could
probably
safely
assume
that
that
is
a
unique
instance
id.
But
this
like
it's
not
really
specced
out
how
to
do
that.
A
E
I
think
we
should
definitely
do
this
if,
for
the
reasons
that
erin
just
mentioned,
maybe
but
yeah,
I'm
curious
how
how
we
could
get
a
unique
instance
id
these
besides
either
randomly
generated
or
forcing
them
to
specify
it.
Maybe
it'd
be
worth
it
to
ask
like
in
the
specs
or
in
the
other
cigs
how
they
did
it
to
get
something
like
meaningful
and
unique.
D
D
I
think
that's
the
same
thing
right,
because
anybody
could
pass
this
and
it
would
be
unique,
but
the
like,
if
you
don't,
have
a
default
behavior,
then
I
don't
think
any
user
is
gonna
in
their
right.
Mind
know
to
pass
service
on
instance.id
because
they
don't
read
this.
The
spec
that
shouldn't
be
like
a
user
requirement
right.
A
So
in
that
case,
it's
like
we
have
to
have
some
sort
of
you
know
easy
way
to
do
this.
I'm
I'm
thinking
this
is
what
you're
kind
of
referring
to
here
right,
either
through
environment
variable
or
I'm,
not
not
too
sure
about
this
right
here,
because
in
that
case
it's
we're
kind
of
like
forcing
them
to
pass.
It.
D
A
Right,
but
do
we
do
we
care
what
the
id
is
as
long
as
it's
unique.
D
A
D
D
Then
that
way
they
might,
you
know
like
having
those
stay.
The
same
across
service
restarts,
for
instance,
or
they
might
it
just,
might
make
more
sense
to
them
in
their
metrics
than
having
a
new
uuid,
and
the
other
thing
is,
if
you
generate
a
new
uid
every
time,
you
won't
violate
the
single
writer
thing,
but
you
will
have
a
new
time
series
every
time
the
service
restarts.
So
so
it's
gonna
increase
the
the
cardinality
of
like
your
metrics
right.
D
Yeah,
that
said,
I
don't
think,
there's
anything
wrong
with
the
with
the
unique
id
because,
like
I
don't,
I
think
at
a
minimum.
It's
not
that
big.
It's
not
a
big
that
big
of
a
deal
like
every
time.
The
servicer
starts
to
get
a
new
time
series
and
people
might
want
to
see
that,
like
hey,
I
had
a
new
rollout,
they'd,
probably
have
a
new
version
id
or
anything
or
something
like
that
already,
so
it
would
probably
create
a
new
time
series
anyway.
So
yeah.
D
Yeah
yeah,
so
so
I
think
this
is
actually
something
kind
of
separate,
but
related
is
we
don't
have
like
a
standard
set
of
resource
detectors
for
something
like
kubernetes
in
the
open
source
sense.
So,
like
google
has
a
gke
one,
and
I
don't
know
about
aws
or
other
platforms,
but
I
don't
think
anybody's
really
gonna
have
any
of
those
other
resource
semantic
inventions
populated
in
a
way
that
we
could
meaningfully
like
pull
right
container
id.
Or
what
have
you
in
so.
F
E
Tying
it
too
much
to
something,
but
we
could
maybe
use
the
the
unix
timestamp
or
the
time
that
the
instance
was
started.
I
don't
know
like
if
that
would
be
easy
to
distinguish
between
instances,
assuming
they
don't
all
start
at
the
exact
same
time
to
populate
this
id
and
then
differentiate
them.
That
way,.
D
I
mean
it's
totally
possible.
They
could
start
at
the
same
time.
Unless
you
take
like
nanosecond
precision,
then
it's
highly
unlikely,
but.
B
A
Okay,
so
it
looks
like
we're,
leaning
towards
number
one,
I'm
okay
with
number
three
as
well.
A
Okay,
anyone
yeah,
I'm
sorry
good.
A
I'm
not
so
sure
about
number
two.
I
feel
like
number
one
and
three
are
more
important
right
now,.
D
Yeah
yeah
yeah,
I
think
number
two
might
be
a
question
for
the
spec.
So
if
we
could
like
define,
define
a
way
to
like
promote
a
different
resource
label
into
the
container.instance.id,
it
would
probably
have
to
be
specked
out
pretty
thoroughly
so.
A
D
D
Yeah,
if
you're
curious
the
link
right
there,
it
says
single
via
a
single
writer
principle,
kind
of
explains
it.
So
if
anybody's
curious,
I
guess
my
next
question
then
is:
do
we
think
this
would
be
a
breaking
change
to
implement.
A
Yeah
so
like,
if
a
user
does
supply
this
like,
then
it's
always
going
to
use
that
right.
If
they
don't,
I
guess
we
will
have
a
extra
field
that
they
can
see
as
before.
They
wouldn't
have
seen
anything.
A
It
won't
break
any
behavior,
but
it
is
sorry
it
won't
break
any
code,
but
behavior
is
changing
so
but
again
like.
If
this
is
something
I
think
like
we
were
following
like
you
know,
if,
if
it's
just
something
that
we're
not
spec
complying
with
or
if
it's
like
fixing
a
bug,
I
feel
like
it's.
D
Yeah,
okay,
I
think
also
I'm
gonna-
do
a
little
more
research
and
update
this
bug
I'll
check
what
other
cities
are
doing
and
try
to
maybe
figure
out
when
this
was
at
in
the
spec.
So.
C
C
D
Yeah
yeah
got
you
yeah.
I
think
the
point
is
like
if
once
we
do
have
metrics,
this
kind
of
thing
would
be
a
breaking
change,
because
if
you
have
like
an
alert
that
that,
for
some
reason,
depends
on
the
label
not
being
there,
if
you
add
a
label,
it
could
break
your
alert
or
something
or
your
dashboard.
So.
D
I
was
sort
of
going
through
the
specs,
but
also
like
google,
so
the
google
cloud
monitoring
will
definitely
break
if
you
send
the
same
time
series
from
two
different
services
at
the
same
time.
So
right,
okay,
very
selfish,
but
we
did.
A
A
Okay,
cool
aaron.
So
are
you
going
to
be
taking
point
on
this
issue.
A
A
Cool
fix
incorrect
headers
parsing,
the
environment
variables.
I
believe
this
is
he
here
today:
hey
yeah
hi.
What's
up
man.
F
So
this
this
one
introduces
the
new
method
for
the
audiolp
exporter
for
reading
the
http
headers
from
enemy,
so
it
was
brought
up
during
like
when
we
were
adding
the
otlp
http
exporter,
so
it
specified
in
this
pack.
You
know
how
to
how
do
we
read
it,
I'm
so
introduced
in
this.
We
had
a
different
opinion
on
this,
so
yeah.
A
F
So
alex-
and
why
is
that?
Okay,
with
this
change,
he
was
requesting
questions.
I
was
wondering
you
know
what:
how
do
we
go
about
this.
A
B
Oh
yeah,
pretty
much
I
mean
what
arguing
is
the
implementation
that
pretty
much
everything
that
this
function
does
can
be
done
directly
in
the
in
the
regular
expressions.
So.
F
Yeah,
so
I
I
mean
we
could
do
that
we
could,
if
it
could
do
this,
probably
in
I'm
not
sure
like
it's
it's
going
to
be
simple
like,
since
you
know
there
is
a
range
of
characters,
that's
only
allowed
in
the
keys
and
http
names
and
values
so
yeah.
I
think
it's
okay
to
have
something
like
this.
B
B
Yeah,
because
I
mean
the
the
complicated
part
of
this
code
is
the
regular
expression
so
and
that
regular
expression
can
be.
I
mean
pretty
much
everything
that
this
code
in
this
function
does
with
d4
can
be
summarized
in
the
usage
of
the
regular
expression,
along
with
the
final,
for
example,
like
a
good
study
example.
So
right.
B
A
A
Click
on
I'll
leave
it
up
to
you
to
get
this
resolved
so.
B
F
Given
this
is
specifically,
for
you
know,
http
headers,
and
there
is
no
other,
you
know
other
places
where
we
use
it
and
who
knows
like?
Maybe
we
are
not.
Maybe
we
can
eventually
make
the
regex
as
a
part
of
the
public
api,
but
it's
at
the
end
user.
I
would
expect
some
sort
of
you
know
youtube
to
be
there.
Then
you
know
I
I
myself
using
regex
so.
F
So,
if
you're
not
already
familiar
with
this
error,
so
there
were
some
people
who
were
you
know
directly
going
to
the
sdk
spec.
They
were
not
reading
the
are
read
the
docs,
so
they
were
using
the
otp,
the
spec
recommended
values
for
the
envs
and
since
we
don't
have
like
in
our
case
it
has
to
be
deep
in
underscore
json
or
gpin
and
square
http,
but
there
is
no
jeepkin
in
the
glue
package
that
we
are
providing
them,
so
they
were
running
into
this.
F
No
entry
point
not
found
I
added
so
this
was.
I
was
adding
them
in
that.
But
then
alex
has
a
point.
You
know
if
you
want
to
make
this
has
to
otp
underscore
trace
and
we
don't.
We
can't
use
this
same
for
the
matrix
when
we
have
that
feature.
F
C
F
Yeah
that
that
made
more
sense
to
me-
and
I
can't
be
like
given
like
the
spec-
has
only
one
value
like
if
it
was
hotel
and
scored.
If
there
were
two
three
different
for
the
names
we,
it
would
have
been
easier
and
straightforward
for
us
to
use
those.
But
since
it's
the
same
one
for
these
different
signals,
I
think
it
makes
more
sense
there
to
have
it
in
the
signal.
In
the
entire
point
itself,.
A
So
so
the
options
are
one:
have
the
signal
within
the
entry
point
name
or
to
have
the
signal
within
the
actually.
I
don't
know
what
name
this
is
to
be
honest,
but
I
haven't
seen.
A
F
So
if
we
introduce
the
signal
in
the
this
identifier,
it
will
have
to
separate
communicately
from
the
like
that
deflecting
from
the
specification.
Spec
only
has
you
know
just
otlp.
If
we
make
it
plp
trace,
I
will
have
to
communicate
differently
and
there
is
a
deflection
between
the
trace
and
our
influence.
B
C
I
think
this
is
actually
specific
to
our
implementation,
but
with
these
entry
points,
so
I
I
think
like
from
a
user's
perspective,
if
I
want
to
use
otlp
to
export
data
and
I'm
using
metrics
and
logs,
for
example,
but
not
spams
or
whatever
like.
I
would
expect
that
setting
otlp,
as
my
exporter
of
choice,
will
just
work
across
all
the
signals
that
I'm
using
within
my
application.
A
B
A
I'm
I'm
not
planning
to
do
it,
but
I'm
I'm
saying
like
if
we,
if
we
do
change
the
name
of
the
the
entry
points
we
just
have
to
make
a
an
additional
like
change.
The
the
the
entry
point
name
that
we're
loading
in
like
in
in
like
site,
customize
or
something
or
wherever
we're
loading.
These.
These
entry
points
in.
A
B
I
don't
know,
maybe
you
can
discuss
other
languages.
C
I
don't
think
I
don't
think
the.
I
don't
think
anything
that
reads
the
entry
points
is
marked
as
1.0,
so
I
think
I
think
only
the
instrumentation
package
and
the
like
the
distro
maybe
uses
it,
but
those
are
not
right.1.0.
A
F
Sure
I'll
make
the
change.
If
there
are
any
questions,
please
comment
down
there.
A
B
Right
so,
except
for
a
very
few
comments
like
two
or
three
everything
is
at
rest.
Tests
have
all
been
added.
Everything
is
passing
now,
so
this.
B
A
B
Yeah
so
please
take
a
look.
I
am
currently
doing
pretty
much
the
same
thing
for
the
sdk
pr
and.
A
Yeah
I'll
take
a
look
at
this.
Also
another
note
I
remember
we
brought
this
up
a
couple
weeks
ago,
but
I
think
alex
brought
this
up
about
like
splitting
the
pr
into
different
vrs
again
like
I'm,
I'm
okay,
with
the
current
one
being
as
it
is
because
I
looked
through
the
whole
thing
anyways,
but
alex
did
point
out
that,
like
you
know
like
avoiding
work
because
like
we
already
did
a
bunch
of
stuff,
it's
like
a
quick
path
to
like
failure
or
something.
F
B
D
D
Yeah,
I
was
gonna,
say
diego,
I'm
looking
at
it
now
right.
I
think
I'm
going
to
have
some
comments,
so
I
don't
want
you
to
like
update
a
whole
bunch
of
sdk
stuff
and
then
have
to
rebase
it
on
top
of
whatever
comments
so
like
should
we
just
focus
on
getting
this
one
merged
and
we
can
like
everybody,
can
try
to
review
this
one
as
quickly
as
possible,
because
I
I
feel
I
don't
want
to
have
like
a
long
loop
to
update
everything
because
we
have
the
entire
prototype
staged.
You
know.
D
Yeah,
maybe.
D
It's
more
just
I
mean
I
can
tell
you
what
I'm
seeing
so
far
is
like.
I
think
we
can.
We
could
definitely
save
a
lot
of
it
for
later
issues,
but
the
no
it's
it's
because
I.
D
A
Yeah
sorry
aaron,
before
going
through
the
specifics,
I
think
in
general,
like
you're
right
about
like
there
will
be
comments
from
the
reviewers,
so
it'd
be
wise
not
to
update
the
sdk
every
time.
I
guess
the
prototype
every
time,
so
we
we
should
fixate
on
getting
this
in
as
soon
as
possible.
A
Should
we
set
like
an
actual
date
for
it,
because
I
think
alex
is
right.
This
has
been
out
for
about
six
years,
so.
A
Yeah,
okay,
so
if
we.
D
D
A
Yesterday,
yeah
yeah,
okay,
so
the
reviewers,
please
whoever's
you
know,
commented
on
this
before.
Please
take
a
look
at
it
asap,
I
would
say
like
by
the
end
of
this
week,
so
give
diego
like
ample
time.
Probably
I
don't
know
diego.
Do
you
have
time
to
work
on
this
next
week?
A
D
Okay,
cool
and
again,
I'm
I'm
more
than
happy
to
like
make
prs
to
your
branch.
So,
if
that,
if
that
helps
to
incorporate
stuff
faster,
we
could
just
work
on
it
collaboratively
too
sure.
C
D
B
Yeah
but
let's
also
keep
in
mind
that
we're
not
merging
this.
A
Yeah
but
like
when
we
do
merge
to
main
like
we're,
not
gonna
scrutinize
it
as
heavily
as
we
are
gonna,
be
now
so
yeah,
okay,
also
guys.
The
original
issue
is
the
metrics
api
sdk
prototype.
So
if
you
have
any
like
discussion
topics
just
put
it
here,
I
think
diego.
You
can
also
add
that
whole
test
suite.
I.
D
Sorry
one
last
quick
question:
I
think
alex
you've
done
some
project
boards
in
the
past.
For
us
should
we
add,
like
a
tag
so
that
we
can
stick
any
issues
related
to
the
metrics
api
sdk
stuff
into
a
project
board.
C
D
Is
it
just
like?
I
make
a
new
tag
on
the
issues
and
then
make
a
project
board,
and
I
can
just
target
the
tags.
I've
never
done
it
before
as
well.
So.
A
A
So
this
is
pretty
recent,
so
I
don't
think
diego
you've
gone
through
this.
Yet
but
oh
wait,
do
you
wanna
just
kind
of
like
tldr
wait?
Is
it
way.
B
A
Oh
yeah
never
mind
people
send
a
but
yeah
yeah.
We'll
we'll
take
a
look
at
this.
I
guess
diego
just
get
updated
on
what
he's
arguing
about
so
yeah.
A
Okay,
does
anybody
have
any
other
comments
on
the
eagles
pr?
A
B
All
right
so,
the
before
this
pr,
we
had
like
a
few
cli
options
for
the
open,
telemetry
instrument
command
and
those
cli
options,
pretty
much
we're
there
to
override
any
corresponding
environment
variable
that
will
be
set
before
running
that
command.
B
We
load
all
those
entry
points
and
all
those
environment
variables
and
add
them
as
options
for
the
cli
command,
so
that
when
somebody
uses
that
command,
they
can
just
type
help
right
and
all
these
options
will
be
displayed
so
that
we
automatically
have
access
to
override
every
environment
bar
a
real
till
environment
and
that's
pretty
much
it.
B
C
B
C
B
You
can
do
the
exact
same
thing,
but
this
is
I
mean
that
argument
should
have
been
made
before
when
we
added
options
for
this
command,
because
this
is
not
doing
anything
new.
It's
doing
the.
What
we
used
to
do
before
is
just
making
it
general
for
all
the
environment,
yeah.
C
Sorry,
sorry,
my
my
point
was:
if
the,
if
me
typing,
the
name
of
the
variable
like
take
the
first
one
hotel
attribute
value
length
as
a
like
flag
that
I
pass
in
is
the
same
length
as
me.
Just
setting
that
environment
variable
ahead
of
ahead
of
the
command,
there's
no
advantage
for
me
passing
it
in
as
a
argument
to
the
commander.
Maybe
there
is
tab
how
to
complete.
D
D
C
In
that
case,
I
would
expect
it
to
just
be
in
the
format
that
always
describing
in
his
in
this
last
comment
there,
but.
A
D
A
B
E
D
D
A
I
have
a
question:
is
it
and
like
I
I'm
I'm
not
even
an
advocate
for
this,
but
is
it?
Is
it
possible
that,
like
we
just
remove
this
ability
to
do
this
and
just
force
people
to
change
it
via
environment
variable?
B
F
B
B
E
That's
kind
of
what
I
liked
about
how
autumn
instrumentation
worked
right
now
it
was
like
sure
you
have
all
these
other
features
in
the
environment
variables,
but
the
important
stuff
is
exposed
in
parameters
to
this
executable,
with
helpful,
like
help
methods
or
help
descriptions,
and
then,
if
you're,
an
advanced
user,
you
want
something
more
complicated.
Then
you
start
to
go.
Look
for
the
environment
variable
that
does
the
job.
A
A
A
Yeah,
that's
also
a
good
question.
Yeah,
that's
also
a
good
point.
Yeah
well,
as
for
now
like
this,
does
anybody
feel
strongly
for
or
against
this?
We
don't
have
to
talk
about
it
here.
We
could
comment
on
the
pr,
but
yeah
diego
is
looking
for
reviewers.
So.
A
A
All
right,
I
think
that
was
the
last
pr
does
anybody
else
have
anything
they
want
to
bring
up
before
we
leave
early
for
the
meeting.