►
From YouTube: 2022-06-30 meeting
Description
Open Telemetry Meeting 1's Personal Meeting Room
A
Yeah
during
these
meetings,
many
weird
issues
all
over.
A
C
C
A
Welcome
everybody
to
another
edition
of
the
western
sick.
We're
gonna
wait
a
couple
more
minutes
to
let
our
folks
join
us.
You
should
please
add
your
names
to
the
attendees
list
and
if
you
have
any
topics,
prs
or
issues
you
want
to
discuss,
also
add
them.
There.
A
Yeah,
I
don't
see
him
online
all
right.
I
guess
we
better
start.
We
have
a
couple
of
today
share
my
screen.
A
Yeah
yeah.
Thank
you
all
right.
First
topic.
This
one
is
mine,
matrix
rc2
release
so
well,
as
you
know,
we
have
been
trying
to
get
our
rc
to
release
for
a
little
bit,
so
I
looked
into
the
into
the
issues
that
we
had
made
in
this
little
project
here.
Sorry.
A
Yeah,
so
we
do
have
this
one,
which
is
ready
to
be
merged.
The
only
thing
that
is
blocking
it
is
some
issue
that
I'm
gonna
discuss
later.
That
is
related
to
infra,
this
other
one,
which
I'm
also
going
to
discuss.
A
I
received
some
feedback
from
aaron
comments
and
finally,
this
other
one.
That
is
an
issue
someone
reported
some
time
ago,
but
I
haven't
been
trying
to
work
on
this,
haven't
been
very
successful
to
reproduce
anything
the
configuration
that
is
here.
A
I
think
it
has
some
some
issues,
so
I
was
discussing
this
with
srikant
a
while
ago
and
I
believe
that
we
could
release
rc2
without
this
one,
because
it's
a
pretty
random
issue
that
I
I
know
it's
worth
worthy
to
to
investigate
it,
but
I
don't
see
it's
justified
to
block
our
rce
because
of
this
one.
This,
on
the
other
hand,
is
much
more
important.
We,
it
has
sparked
a
bit
of
a
discussion.
A
I
have
addressed
the
comments
from
aaron
just
a
few
minutes
ago,
and
this.
A
That
we
can
merge
if
we
fix
the
infor
issue,
so
I
was
just.
I
just
wanted
to
ask
everybody
if
you
are
okay,
that
if
we
get
these
two
issues,
nourished
or
even
even
this
one
is
not
that
important.
I
guess
but
but
this
one
is
so
if
we
get
them
merged,
will
you
be
okay
with
releasing
rc2
for
metrics.
D
A
Okay,
that's
good.
B
Yeah,
I
agree
like
we
yeah
since
last
time.
Hey
later.
A
B
I
think
like
this,
we
already
had
many
fixes
and
then
changes.
I
think
this
like.
If
we
get
this
two
seven
five
five
merged,
we
can
go
ahead
and
do
rc2.
A
Okay,
perfect
later.
A
I
don't
know
if
you
heard
that
the
last
part
of
the
conversation
we
were
discussing,
if
everybody's
in
agreement
of
releasing
rc2,
if
we
get
this
one
merged.
A
Yes,
I
am,
I
heard
that
I'm
good
for
it
great
okay,
so
glad
that
we
have
an
agreement
on
that.
Okay
issues.
Oh.
E
Sorry
before
we
go
on
the
tracing
decorator,
why
is
that
needed
for
the
rc2.
A
But
we
did
mark
it
as
rc2.
There
is
a
fix
already.
This
pr
was
waiting
for
a
while,
because
the
submitter
did
not
have
the
cla
signed
a
couple
of
days
ago
got
the
cl
cla
signed.
So
this
pr
is
good
to
go.
E
A
A
It
is
completely
related
to
the
tracing
sorry,
sorry
about
that.
Okay,
I'll,
take
this
out
of
the
project:
okay,
okay,
yes,
so
next
topic,
ci
being
blocked;
okay,
bad
news.
Ci
is
right
now
being
blocked.
What
is
happening
is
that
we
added
the
instrumentation
for
sql
commander
and
that
added
a
file
that
is
just
a
few
characters
longer
than
what
windows
likes.
A
So
I
did
some
research
realized
that
this
is
a
workaround.
A
In
this
way
you
can
tell
git
that
it
can
accept
long
file,
names
and
oh
good
right
after
adding
this
ci
started
passing,
except
for
one
extremely
weird
issue
here
with
our
instrumentations
test.
So,
as
you
can
see,
it's
only
one
test
that
is
failing
and
the
test
that
is
failing
is
super.
Weird
system,
metrics
pyramid,
flask
and
django
are
failing.
B
So
this
was
this
was
an
error
that
I
encountered
when
I
you
know
when
I
was
refactoring
the
the
like
the
test
based
util
file.
This
was
fixed.
I'm
not
sure
why
you
are
seeing
it
again
now.
A
Yeah
correct:
yes,
I
looked
into
the
git
history
and
I
noticed
that
this
line
was
last
modified
by
you
and
yeah,
and
probably
that
that
was
you
fixing
this
issue.
I
also
have
no
idea
why
this
is
happening
again,
but
yeah.
This
is
happening
in
several
instrumentations
in
django
in
flask
system,
metrics
is
failing.
For
other
reasons.
B
I
have
seen
all
of
them
before
I'm
not
sure
why
they
are
here
again,
and
can
you
bump
the
country
report
to
the
latest
and
see
if
it
solves
this
issue.
A
Okay,
okay
yeah.
I
just
tried
this
a
couple
of
hours
ago
and
I
haven't
tried
with
the
with
the
show
but
I'll.
A
I
just
wanted
to
update
you
on
this
by
the
way
we
in
theory,
can
this
question
merge
now
this
issue
is
not
required,
but
I
I
didn't
want
to
do
that
because,
of
course
we
don't
want
to
do
that
with
a
failing
test
right,
but
yeah
just
to
update
you
on
how
this
this
thing
and
by
the
way,
this
is
the
ci
issue
that
is
blocking
this
pr
and
pretty
much
every
other
pr
that
we
have
right
now
in
opengl
entry
python,
because
all
the
windows
environments
are
failing:
okay,
yeah,
hopefully
I'll
keep
debugging
this
issue
and
hopefully
updating
the
show
or
something
will-
or
I
don't
know,
deleting
bumping
the
cash
cash,
because
she
I
don't
know
how
to
say
that
english
gosh,
I
will
will
work.
A
Okay
next
issue
is
also
mine.
Pr
tools,
oh
there's,
a
new
issue
here.
D
D
Yeah,
like
you
mentioned
this
one
briefly,
so
this
is
this
seems
to
be
the
desired
behavior
right
like
this
is
the
that
other
pr
that
you
added
is
testing
that
this
is
basically
what
it
does
right.
A
A
No,
no
probably
it's
intended
behavior.
I
didn't
get
that
into.
I
investigated
that
much.
I
was
just
trying
to
reproduce
and
couldn't
just
put
it
away
to
focus
on
the
rest
of
the
issues,
but
it
seems
now
that
you
mentioned
yeah.
It
could
perfectly
be
the
behavior
if
you,
if
you
can,
leave
a
comment
there
that'll
be
great.
Maybe
you
can
reach
an
agreement
and
close
this
issue.
D
In
which
case
you
know,
let
me
know,
but
yeah.
A
Yeah
I'll
also
take
a
look
at
it
and
we'll
try
to
confirm
that
this
is
the
expected
behavior.
I
I
I
was
trying
to
confirm
it
by
actually
reproducing
it,
but
I
couldn't
even
make
it
that
far
so
but
yeah
it
sounds
like
I
mean
what
do
you
say?
It
makes
sense
all
right
so
yeah,
please,
please,
okay,
what
else
we
have
here?
We
have
okay,
pr,
two,
seven,
seven
three!
A
I
think
oh
yeah,
this
one,
this
one
has,
you
know,
created
a
little
bit
of
discussion.
Thank
you
aaron
and
srikant
for
comments.
So
what
we
have
here
is
that
when
we,
when
I
made
a
change
to
the
yielding
of
values,
so
that
we
did
not
yield
non-values
aaron
asked
to
to
make
sure
that
we
keep
putting
the
same
values
for
cumulative.
A
Instruments,
even
after
consecutive
cycles
of
collection
have
happened,
which
is
of
course
expected
behavior.
Now
I
have
aaron
specifically
mentioned
about
the
permatease
metric
reader
and
in
this
pr
I
my
goal
is,
was
to
add
a
test
for
the
prometheus
metric
reader.
That
shows
that
we,
that
that
are,
we
behaved
like
that
right
that,
after
several
cycles
of
collection,
we
still
do
the
same
thing
return
the
same
result.
A
So
here
we
have
the
test
here
we
have
cycles
of
collection,
and
we
can
we
check
that
all
of
them
are.
Are
the
same
now
a
couple
of
comments
here
from
sprikant
and
from
aaron.
I.
A
That
the
idea
is
that
we
test
this
as
part
of
the
sdk,
not
only
as
part
of
prometheus.
We
want
to
make
sure
that
this
behavior
this
happens
in
the
sdk
aaron
suggested
that
we
use
in
in
memory
method
reader
to
check
this
right.
So
I
a
few
minutes
ago,
I
added
this
test.
A
It
uses
a
memory
metric
reader.
Pretty
much
does
the
same
thing.
We
create
a
counter,
we
add,
and
then
we
call
collect.
We
check
the
the
data
points
that
are
produced
and
then
we
collect
again.
We
are
the
data
points
that
are
produced
again
and
then
I
compare
the
data
points
just
a
comment
here.
A
A
D
We
are
yeah
that
looks
good.
I,
my
main,
like
confusion,
is
around
the
that,
like
the
changes,
the
prometheus
exporter
or
the
preferred.
D
A
Comment
a
few
minutes
ago:
okay,
that's
a
that's
a
very
good
point,
so
just
to
give
some
context
to
the
rest
of
the
people
here
before
I
worked
on
this
pr.
The
prometheus
metric
reader,
which
inherits
from
metric
reader,
did
not
have
the
preferred
temporality
and
preferred
aggregation
parameters
in
its
constructor.
The
metric.
B
A
Has
them
in
there,
in
its
constructor
now
aaron
points
that
the
prometheus
metric
reader
is
a
cumulative.
B
A
Reader,
which
I
I
understand,
what
what
are
once
tries
to
say-
and
I
agree
my
point
to
add
these
parameters-
is
that
it
is
not
accurate
to
say
that
a
reader
is
cumulative
in
our
data
model.
Being
cumulative
is
a
property
of
instruments,
so
we
asso
associate
the
instrument
class
with
an
aggregation
temporality,
but
the
metrograders
themselves
are
not
cumulative
or
delta.
The
other
thing
is
that
we
should
be
consistent
in
the
interfaces
with
the
interfaces.
A
D
I
don't
I
don't
know
if
I
agree
with
like
this
is
only
relevant
to
a
metric
reader.
The
instruments
don't
have
temporal
temporality
outside
of
what
the
metric
reader
says
when
it
calls
collect
right.
Yes,
regardless
this
is
kind
of
like
semantics.
Right
like
it's,
it's
not
not
terribly
important
and
and
as
far
as
the
interface
goes,
it's
totally
fine
to
have
like
a
subclass
with
a
different
constructor
right,
because
you're
not
calling
the
constructor
you're
you're
receiving
an
instance
of
a
metric
reader
you're
not
trying
to
create
a
new.
B
A
Okay
yeah,
my
my
I
do.
I
think
that
I
would
personally
prefer
that
the
interfaces
are
kept
the
same
as
much
as
possible,
but
the
the
thing
that
I
am
doing
here
is
that
we
are
raising
an
exception
here.
If
the
prometheus
metric
reader
is
is
configured
something
else
that
is
not
cumulative,
we
erase
an
exception.
A
A
That
is
what
I
think
we
are
trying
to
say
when
we
say
that
the
prometheus
metric
reader
is
cumulative.
This
is,
of
course,
a
semantics
issue.
I
agree
with
you
aaron.
A
I
think
that
this
implementation
makes
the
semantics
more
clear
to
the
user,
because
I
don't
want
us
to
tell
the
user
to
say
to
the
user
that
the
promise
used
metric
leaders
cumulative
that
that
is
just
an
inaccurate
statement,
because
metro
leaders
are
not
cumulative
or
I
mean
the
aggregation
of
morality
is
a
concept
that
is
not
applied
to
much
readers
right.
It's
a
a
little
bit
more
complicated
than
that.
You
understand,
because
of
the
preferred
aggregation
to
morality,
to
the
dictionary
and
everything
else.
D
A
A
D
Well,
okay,
I
mean
we
can't
like
protect
against
every
possible
way.
Somebody
could
accidentally
call
the
constructor
like
I.
I
don't
think
that
anybody
would
would
try
to
do
this.
They
would
probably
read
the
code
or
they'd
have
like
a
static
analysis
tool
that
would
tell
them
this
is
incorrect
or
they
would
just
try
to
run
it
or
test
their
code,
and
it
would
fail
right,
like
I
think
the
this
this
exception
is
not
like.
This
configuration
option
is
not
useful.
A
Well,
it's
also
confusing
to
have
a
child
class
that
does
not
have
the.
D
D
D
A
D
D
A
Okay,
imagine
the
in
this
situation.
Someone
is
creating
network
readers
dynamically.
Not
specifically,
someone
is
pretty
much
using
a
sequence
of
classes
at
twitter
classes
and
they're
instantiating
them,
and
they
expect
every
one
of
those
classes
to
have
this
to
accept
these
parameters,
and
if
there
is
a
class
that
doesn't
then
they
will
that
that
will
fail,
because
there
is
a
class
that
is
different
to
the
rest
of
them
and
does
not
satisfy
the
the
interface
and
then
right.
A
That
I
mean
if
the
constructor
wasn't
part
of
the
interface,
then
it
will
not
be
defined
in
the
a
in
the
astral
class.
D
D
B
A
We
just
don't
agree
in
in
this
particular
point.
That's
fine!
That's
fine!
I
mean
it's.
It's
fine
to
disagree
again.
I
guess
if
other
people
can
chime
in,
so
that
we
can
make
a
decision
also.
I
am
not
straw
super
strongly
on
this
point.
As
aaron
has
mentioned,
this
is
a
semantic
semantics
issue
it.
It
should
not
have
like
it's
it's,
not
life
or
death
for
this
project
right.
A
So
if
we
flourish
decide
that
it's
better
not
to
have
this
primary
level,
I
I
can
live
with
that,
but
I
I
would
prefer
it
to
be
like
this,
but
if,
if
voters
agree
that
we
should
go
the
other
way,
that's
fine.
So
please
leave
your
comments.
Aaron
did
you
want
to
say
something
else?
A
Okay,
okay,
but
yeah
that
that's
just
to
to
answer
the
the
question
that
you
put
there
this.
This
is
why
I
I
am
doing
this
okay.
Besides
this
this
issue,
you
are
erin
you're.
Fine,
with
this
test
prove
enough
that
we
are
checking
the
the
community.
D
Yeah,
let
me
let
me
review
it
a
little
more
closely
but
yeah
I'll
I'll
comment
back
on
I'll
reply
to
your
comment
and.
A
Take
a
look
all
right.
Thank
you.
Thank
you
appreciate
it
appreciate
it
okay,
so
this
issue
should
be
the
the
one
that
is
the
most
important
for
rc2
it's.
It
already
has
enough
approvals,
but
I
haven't
merged
merged
yet
because
of
the
discussion
that
is
still
going,
but
please,
if
people
can
comment
on
this
I'll
be
great
so
that
we
we
can
reach,
we
can
make
a
decision
on
how
to
proceed
as
soon
as
possible
so
that
we
can
have
our
hearts
too
great.
A
Okay,
so
I
I
don't
know
if
you
noticed
that
this
pr
has
been
going
on
for
a
while
and
the
submitter
has
been
trying
to
make
lint
pass
and
so
far
it
hasn't
passed.
A
We
have
tried
many
things
like
updating
the
sha,
maybe
cleaving
the
cache
or
something
anyways.
I
tried
looking
a
little
bit
into
this
and
found
something
extremely
weird,
and
I
will
laugh
if
someone
has
any
idea
how
to
proceed
here.
So
I
pretty
much
created
two
virtual
environments,
one
using
main
and
one
using
the
code
in
this
pr,
and
I
realized
that
if
we
run
the
main
code
with
the
virtual
environment
of
this
pr,
lint
also
fails.
A
So
it
seems
like
the
changes
in
this
pr
are
not
the
one
causing
lint
to
fail,
but
the
virtual
environment
itself.
So
once
that
I
realized
that
I
I
thought
to
myself:
oh
well,
probably
we
have
a
dependency
that
we
have
not
pinned
and
we
are
using
a
different
version
of
piling
or
something.
But
that's
not
the
case.
I
could
not
find
any
thing
in
the
in
the
dependencies
of
both
virtual
environments.
That
was
different.
A
And
also
the
failures
that
are
happening
are
happening
in
instrumentations
that
are
completely
unrelated
to
this
pr,
for
example,
aski
wsgi
botocore
stuff,
like
that
that
should
not
be
failing.
It's
very
weird.
I
don't
have
many
more
ideas
and.
A
Yeah,
if
someone
has
any
idea
on
how
to
proceed,
please
comment
below,
because
this
is
a
very
bad.
This
is
a
complete
instrument,
new
instrumentation
that
is
just
being
blocked
because
lent
because
of
lent.
B
A
A
This
is
sorry,
I
don't
think
this
is
a
ci
issue
because
it
fails
in
my
lab
lint
fails
in
in
my
computer,
so
yeah.
E
A
Yes,
I
can
see
the
same
failures
in
ci
and
in
my
laptop.
E
A
If
you
clone
open
telemetry
by
some
country,
then
run
lint
with
main
the
last
committee
main
if
it
passes
fine.
A
But
if
you
check
out
this
pr
in
your
laptop
using
this
and
run
lint,
it
fails,
and
if
you
keep
that
virtual
environment
and
activate
it
and
run
the
the
lend
command
that
is
python
scripts.
Each
these
length
check
only
or
something
like
that,
with
the
virtual
environment
created
with
this
pr,
but
using
the
code
in
main,
it
also
fails.
A
So
it
seems
like
it's,
the
virtual
environment
that
is
being
created
this
pair,
that
is
causing
length
to
fail,
which
makes
absolutely
no
sense,
and
it's
a
totally
crazy
idea
right,
because
the
dependencies
seem
to
be
the
same,
except,
of
course,
for
the
new
dependencies
that
were
added
in
this
pr
that
are
necessary
for
a
a.
I
o
p
cars.
A
This
kind
of
things
right,
which
should
have
no
effect
on
lint.
A
But
anyways,
I
I
try
to
write
down
in
detail
what
I
did
here.
So
if
someone
can
reproduce
or
investigate
this
further,
it
will
help
us,
because
this.
B
E
I
just
opened
up
a
new
pr.
I
didn't
put
the
cache
now.
E
Sorry,
I
I
just
I
just
ran
like
a
dummy
pr
content
trip
just
a
test
main.
Oh,
yes,
it
is
main
in
country.
A
Okay,
but
but
yeah
that
that's
pretty
much
what
I
have
here,
I
don't
know
if
anyone
has
any
other
issues
or
comments.
D
Okay,
I
was
going
to
ask
we
had
a
few
things.
I
know
I've
been
working
on
the
documentation,
I'm
doing
it
right
now,
and
then
I
have
one
like
pending
pr
to
the
open
telemetry.
I
o
doc
site
that
after
this
rc
goes
out,
I
can
publish
that
one,
because
I
don't
want
to
publish
instructions
that
don't
work
yet.
I
know
there
was
also
like
some
contrib
instrumentation.
I
think
we
did
whiskey
and
then
we
I
don't
know
that
there
was
also
some
examples:
somebody's
going
to
work.
A
Stuff
yeah,
I
I
do
have
to
work
on
on
examples.
I
haven't
pretty
much
made
any
progress
on
that.
I've
been
trying
to
focus
on
these
little
things
that
we
have
here
that
have
been
a
bit
hard
because
of
all
the
ci
issues
and
everything
else.
Once
we
get
rc2
I'll,
be
focusing
on
writing
down
the
examples
that
I
want
to
behave
also
to
work
also
as
tests,
so
so
that
we
have
this
set
of
examples.
D
Cool
yeah
that
sounds
good
was
how
about
contrib
instrumentation.
I
guess
I
guess
recon's
pr
got
merged
with
the
whiskey
stuff.
B
Yeah
there
was
also
we
merged
the
request,
matrix
instrumentation.
That
was
also
there.
Those
two
got
matched
another
couple
of.
I
created
new
issues.
For
you
know
each
instrumentation
there
were
a
couple
of
volunteers
who
wanted
to.
You
know,
contribute
that
I
think
they're
like
four
yeah.
That's
the
status
of
the
matrix
instrumentation
on
the
country.
B
D
Is
there
are
all
those
working
with
the
auto
instrumentation
agent
as
well,
and
exporters.
C
D
Awesome,
cool
yeah.
That
was
that's
what
I
put
in
the
I
updated
the
getting
started
guide,
which
is
using
the
auto
instrumentation
agent.
So
it's
gonna
once
once
the
rc
and
those
changes
go
out
that
should
all
be
working,
so
I
think
we're
in
a
pretty
good
state.
Then.
B
Yep
yeah,
I
I
tested
it
like
a
week
back
and
I'm
gonna
take
a
look
at
it
right
again
and
see
if
the
both
instrumentations
are
producing
the
metrics.
The
art
instrumentation.
B
C
C
I
just
had
a
I
had
a
quick
pr.
I
was
curious
to
get
people's
thoughts
on
it's
pretty
simple.
It's
not
it's
a
non
like
it's,
not
a
functionality,
change,
it's
just
just
like
a
refactor
of
how
we
do
the
two
json
methods
and
it
was
yeah.
I
was
curious
what
people
think
about
it,
because
it's
one
of
those
situations
where
we're
kind
of
retroactively
refactoring
it-
and
I
was
wondering
if
this
is
kind
of
correct,
correct
in
what
we're
going
for.
A
So
yeah
this
we
are
right.
B
Yeah
we
were
having
some
discussion
on
the
issue
itself,
so
latent
pointed
out
that
it's
a
breaking
change
to
introduce
the
new
format,
regardless
of
the
intention
of
the
original
component,
the
control
exporter.
B
I
initially
thought,
like
you
know,
it's
like
it's
it's
for
the
debugging
purposes.
Only
that's
intended
like
different
sdks
have
different
formats.
We
can,
we
can
go
about.
We
can
go
and
still
change
that.
But
then
again
I
thought
about
it
again.
I
think
I'm
fine
with
keeping
it.
As
is.
I
noted
down
my
you
know
what
I
think
right
now.
I
I
think
the
change
is
simple,
but
do
we
want
to
introduce
that
or
not
in
the
testing
since
it's
already
released
and
then
people
are
using.
A
Okay,
I
think
I
haven't
yet
commented
on
this
one
I'll
I'll
try
to
take
a
look
at
it
and
yeah
I'll.
Let
you
know
what
I
think.
A
Very
important,
thank
you,
aaron
for
bringing
this
vacation
so.
A
Some
people
may
be
taking
some
time
off
early
season.
I
will
be
out
july
the
10th
until
the
20th.
A
A
Yeah
we
appreciate
it
our
dedication's
all
pretty
good
aaron.
Are
you
on
vacation
right
now,
just
curious?
No!
No!
Next
next.
D
Week
next
week,
next
week,
yeah
you're
out
tomorrow-
oh
okay,
no
no
I'll
be
in
tomorrow,
and
then
god
after
that,
tomorrow's
my
last
day
until
I
leave.
E
Okay,
great,
so
we
gotta
get
those
metrics
approvals
in
before
before
next
week.
Guys.
A
E
A
We
we
should
definitely
make
make
a
decision
on
this
topic
that
aaron
and
I
have
been
discussing
before
aaron
leaves
for
vacation.
So
I
think
this
weekend
later,
we
need
your
opinion.
E
A
Well,
I
guess
that
everybody
listened
to
aaron
and
myself
exposed
points
right,
yeah,
okay,
so,
ladies,
we
can
please.
B
B
It
shouldn't
be
configurable
and
then
just
be
cumulative,
but
I
was
okay
with
having
it
configured
and
like
I
have
the
param,
and
then
there
is
an
exception
yeah.
I
I
don't
not
have
any
strong
opinion
I'm
and
with
either
way.
One
thing
that
you
brought
up
is
that
like,
if
somebody
like,
let's
say,
there's
some
entry
point
or
something
like
that
that
we
added
to
the
materials.
B
If
all
the
metric
leaders
expect
some
kind
of
similar
configuration,
let's
say
here
they
are
excited
to
communicate
on.
One
of
them
means
from.
B
I
think
that
that
would
be
helpful,
but
otherwise
I
really
don't
see
any
reason
to
have
it
here
as
a
paramount,
configurable
and
stuff.
You
know
I
mean
we
we're
accepting
something
and
then
raising
it
exception
that
it
shouldn't
be
configured
instead
of
keeping
it
static.
E
Hey
diego,
what
was
their
strong
reasoning
for
having
the
dictionary
passed
in
like?
Why
do
we
have
this
configuration.
A
Okay,
I
do
have
a
reasoning,
but
it's
not
a
strong
reasoning.
I
think
this
is
a
little
bit
of
a
matter
of
taste.
I
prefer
to
have
the
dictionary
passed
in
because
it's
more
consistent
with
the
interface
and
because,
if
it
makes
it
more
obvious
to
the
user,
what
is
happening
because
an
exception
is
being
raised
that
specifically
tells
the
user.
Okay,
you
are
using
promote,
use
the
prompt
use,
metric
reader,
wrong.
A
Again,
this
is
not.
I
mean
this
is
not
a
matter
of
life
and
death.
I
I
would
prefer
it
to
be
like
that.
But
again,
I'm
not
gonna
block
this
pr
or
anything.
If
this
does
not
go
this
way.
D
D
I
mean
we've
already
got
another
parameter
here,
that's
not
in
the
super
class
right.
We
have
prefix
and
you
can
imagine
we
might
add
other
specific
parameters
to
all
the
exporters
or
readers
like,
for
instance,
you
could.
You
could
configure
this
one
to
set
the
port
in
the
constructor
parameter
or
to
set
like
endpoint
is
something
in
otp
which
isn't
in
console.
Obviously
right
like
you're,
going
to
have
some
specialization
of
the
constructors,
and
if
you
want
to
have
like,
like.
D
Constitution
for
entry
points
or
something
like
that,
we
we
need
to
have
a
separate
like
consistent
interface
of
factories,
to
create
these
things,
but
the
actual
constructors
are
not
guaranteed
to
all
be
the
same
like
if
you
say
I
have
a
list
of
of
classes.
Metric
reader
like
classes
not
instances,
the
the
type
annotation
would
be
like
type
of
metric
reader,
which,
which
is
like
an
alias
for
just
a
function
which
takes
no
parameters
and
returns
a
metric
greater.
D
So
we
can't,
I
don't
think
we
can
rely
on
a
constructor
for
that
anyway,
we
can
add,
like
other
other
ways,
to
do
that,
like
either
separate
factory
methods
or
like
a
static
method.
E
Okay,
hey
erin:
where
was
this
the
what
you
were
referring
to
when
you
were
saying
that
constructor
is
not
part
of
the
interface?
Is
that
part
of
the
same
conversation
yeah.
B
That's
a
good
point
to
like
again
yeah.
I
agree.
It's
not.
That
was
a
weak
argument
to
say
that
you
know
to
have
them
agree.
D
We
can,
we
could
probably
accomplish
it
another
way.
I
just
think
like
in
terms
of
clarity,
I
would
say
not
having
a
parameter
not
having
a
configuration
option
for
something
you
can't
configure
is
more
clear
than.
B
E
I'm
not
I
don't.
I
don't
have
a
strong
opinion
in
terms
of
a
design
standpoint,
but
for
a
user
experience
like
standpoint,
I
kind
of
agree
with
what
aaron
said
like
like.
I,
I
don't
like
having
the
ability
to
configure
something
that
can't
be
configured
I'm
trying
to.
A
A
The
minority
opinion
is
to
not
have
this
configuration
right.
I
can
remove
this.
There
is
only
a
small
ask:
is
that
we
in
our
documentation,
let's
try
to
avoid
referring
to
the
prometheus
metrics
border
as
a
cumulative
reader.
A
A
No,
it's
it's
just
that
I
do.
I
do
think
that
that
statement
you
know
prometheus
being
a
cumulative
reader.
He
is
inaccurate
and
it's.
B
I
was
going
to
say,
like
the
regular
prometheus
user,
would
already
know
that
right
who
is
already
using
the
prometheus
client
libraries.
They
would
already
know
that
you
know
it's
all
cumulative,
so
it's
again
afraid
to
say
they
won't
expect
anything
to
be
a
temporary
to
be
configured
if
they
are
coming
to
open
telemetry
from
those
client
libraries.
B
A
All
right
cool
okay,
so
I
think
the
decision
is
to
remove
these
configuration
parameters
I
removed
them.
Then
I'll
just
need
aaron.
Please
take
a
look
at
this
test
case
to
make
sure
that
it
satisfy
the
objective
of
this
pr
that
we
verify
this
yeah.
It
looks
good
and
finally
I'll
try
to
figure
out
what
the
hell
with
the
ci
so
that
we
can
get
these
tests
to
pass,
and
hopefully
we
can
merge
these
and
have
finally
have
our
rc2.
A
Yeah,
there
is
something
very
weird
happening
here
and
it's
the
the
test
cases
are
failing
on
windows
as
you
can
see,
and
that's
because
of
the
the
thing
with
the
the
mro
and
the
things
that
we
can
have
fixed.
Maybe
this
can
be
fixed
because
with
an
update
of
shower
or
something
like
that,.
A
Fix
this
in
this
pr,
but
on
on
this
one,
this
other
one
which
one
sorry
it's
in
the
main
yeah
in
the
main
repo.
A
A
Everybody
wants
this
beer
yeah,
okay,
I'll,
do
that
and
okay
great,
thank
you
for
helping
this
decision
be
being
made
and
we
are
on
a
good
path
to
having
rc2.
D
A
If
there
are
no
more
topics,
thank
you
for
attending
and
we'll
see
you
next
week.