►
From YouTube: 2021-10-28 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).
B
B
A
A
A
A
A
A
D
A
C
C
Okay,
so
cool
looks
like
we're
four
minutes,
since
we.
A
C
Started
looks
like
we
have
one
new
attendee
today.
Is
it
june?
C
C
All
right,
if
you
don't
want
to
talk,
then
I
guess
it's
fine.
I
guess
we
can
get
started
then
cool
as
usual.
Please
add
your
name
to
the
attendees
list.
Looks
like
we
have
a
couple
topics
today
to
start
off.
A
Okay,
go
ahead.
Yes,
thank
you.
So
I
opened
up
this
draft
pr
because
aaron
asked:
how
could
we
comment
in
this
prototype?
And
yes,
so
I
am
not
intending
to
merge
this
into
money,
just
open
this
up,
so
that
other
people
can
add
comments
to
this.
So
in
the
past
few
weeks
we
have
been
discussing
the
how
to
be
compliant
with
error
handling
in
the
spec
I
put
together
this
prototype.
A
I
presented
it
yesterday
to
ted
young,
who,
I
think
is
that's
reinforced
behind
this
requirement
in
spec
and.
A
These
this
prototype
that
I
have
here
today,
I
want
to
share
it
with
you.
I
think
it
is
definitely
possible
for
us
to
implement
the
safety
mechanism.
A
I
also
think
that
it
may
require
significant
changes
into
our
design,
including
breaking
changes.
So
that's
that's
the
bad
news.
The
good
news
is
that
this
apparently
works
there
is.
There
are
several
important
design,
new
design,
ideas
that
are
incorporated
in
this
prototype,
and
I
want
to
discuss
that
with
you
all
so
this
to
give
you
just
a
quick
overview
of
this
prototype.
It
works
like
this.
A
There
are.
There
is
one
component.
Oh
sorry,
let
me
just
show
you
first,
the
files
that
I
have
here.
Okay,
so
what
I
have
done
is
that
I
created
a
mock,
open,
telemetry
api
and
I
mock
open,
203,
sdk
and
then
application
code.
A
A
Set
the
this
open
elementary
sdk
mark
that
I
have
here
as
the
sdk
that
we're
going
to
use
anticipating
that
there
can
be
multiple
sdks
installed
in
in
the
virtual
environment.
A
This
function
only
divides
this
one
by
this
one
executing
this
should
not
fail.
It
just
returns
to
the
zero
executing
this
one,
on
the
other
hand,
fails
because
this
is
a
division
by
zero.
Same
thing
happens
with
the
class
this
class.
It
has
a
method
that
also
divides
this
one
with
this
one,
and
this
one
will
fail
right
because
it's
a
division
by
zero.
As
you
can
see
in
the
application
code,
we
only
have
api
objects,
class,
zero
and
function.
Okay.
If
we
run
this
like
this,
you
get
this
result
so.
A
The
first
operation
that
we
do
here
for
divided
by
two
is
printed
here,
but
when
we
try
to
run
this
one,
it
fails.
It's
a
division
by
zero
and
the
error
mechanism
handling
on
the
error.
Handling
mechanism
works
like
this.
It
catches
the
exception
and
it
reports
a
warning,
a
python,
specific
warning
and
it
returns
a
default
value,
which
is
pretty
fine
in
this
case,
is
zero
same
thing
happens
with
four
here
pretty
much.
What
we
do
here
is
we
divide
four
by
two
and
then
we
multiply
it
by
two.
A
So
the
result
is
four:
let's
print
it
here
and
the
same
thing
happens
here.
C
revision
error
gets
raised
and
caught
by
the
open
to
limit
three
air
handling
mechanism
and
zero
is
returned.
A
The
important
thing
is
that
the
the
application
has
not
crashed
it.
It
was
able
to
run
until
the
end
like
this,
for
example
right.
If
you
run
this
application,
you
see
that
yeah.
It
runs
completely
because
every
exception
that
happens
is
being
handled
by
these
mechanisms.
Now,
the
there
is
another
way
that
we
can
run
this
using
this
standard
feature
python,
which
turns
warnings
into
actual
exceptions.
If
we
do
this,
the
application
does
not
complete.
A
The
first
warning
that
gets
raised
gets
turned
into
an
exception
division
by
zero
and
it
fails.
This
is
a
fail
fast
mode,
and
this
is
required
by
spec
well
so
far
so
good,
but
this
design
it's
based
on
several
strict
rules
that
are
necessary
for
this
work.
A
We
have
first
the
safety
mechanism
here,
which
is
private,
because
this
is
not
intended
for
anyone
to
use
directly.
It
is
a
decorator
that
has
this
pretty
much,
try
accept
it
catches
an
exception
and
erases
a
warning.
Okay,
this
decorator
must
be
implement
applied
to
every
function
or
method
in
the
api.
B
A
A
A
These
decorators
have
the
predefined
value.
That
is
to
be
returned
if
an
exception
happens
and
what
these
objects
do.
This
function
this
class
everything
what
they
do
is
that
they
act
as
proxies,
so
they
get
the
function
from
the
sdk
name
function
and
they
pass
that
the
the
arguments
that
they
received
if
this
crashes
it
gets
handled
by
the
say,
function.
So
in
this
design,
the
api
is
the
only
thing
that
the
user
has
contact
with.
F
A
Those
requirements
are
that
the
only
thing
that
the
user
has
contact
with
is
this:
are
these
api
objects
that
every
sdk
completely
implements
the
api
defined
here
exactly
like
it
is
defined?
Here
I
mean,
and
with
that
I
mean
these
same
signatures,
the
same
names
for
functions
and
classes
and
methods,
because
this
is
how
the
the
api
knows,
how
to
call
the
sdk
the
corresponding
sdk
object
by
by
its
name,
and
if
we
do
this,
then
I
think
we
can
be
compliant.
A
There
is
some
added
benefit
to
this
approach,
and
I
believe
is
that
it
finally
and
completely
defines
what
a
breaking
change
is.
I
think
we
never.
We
actually
never
got
to
answer
that
question.
I
think
that
if
we
follow
this
approach,
then
we
can
say
that
a
breaking.
D
A
A
A
Consistent
sorry
on
not
introducing
any
breaking
changes,
I
hope
that
we
can
fulfill
that.
I
am
not
sure
at
all
that
we
can
do
that,
and
I
also
have
a
few
questions
with
how
the
specific
the
specification
is
written
because,
as
the
specification
is
written,
I
think
it
may
be
self-defeating,
because.
A
The
specification,
if
I
understand
it
correctly,
has
an
api
that
is
outside
of
the
api
itself,
an
api
that
applies
to
the
sdk
for
exporters,
for
example.
So
if
that
happens,
if
there's
another
api
and
the
user
can
import
directly
from
the
other
api,
then
that
api
can
cause
exceptions
to
be
rated,
and
if
that
api
does
not
have
it
its
own
air
handling
mechanism,
something
else
can
can
fail.
A
F
Hey
diego,
I
I
looked
at
the
pr
and
just
commented
like
right
when
the
meeting
started
just
had
a
chance
to
check
it
out.
I
think
you're
like
I,
I
basically
a
lot
of
my
comments
are
the
same
things
you
just
said
about.
You
know
like
the
the
sdk.
F
A
D
D
E
Like
which
one
like
resource
or
ex
like
stuff
like
that,
I
don't
know
actually.
A
F
A
A
Well,
that
could
be,
I
mean
that
could
also
work.
I
think
the
if
the
if
every
sdk
functional
method
is
decorated
by
itself.
A
F
I
don't
think
it's
a
requirement
because,
for
instance,
like
you,
can
use
the
dependency
injection
approach,
you
don't
need
to
use
globals.
So
if
you
wanted
to
do
that,
you'd
have
there's
no
other
way
to
like
pass
in
a
sdk
implementation
of
a
meter,
for
instance
to
a
class.
I
put
a
comment
on
the
pr
with
some
code
to
try
to
show
what
I
mean.
A
Yeah,
the
sorry
later
on,
you
were
going
to
say
something.
C
Oh
no
worries
I'm
just
listening
to
aaron's,
there's
a
question:
okay,.
A
C
Oh
sorry,
diego
before
you
go
on,
let's
try
to
time
box
this
to
like
yeah,
maybe
9,
40
or
15
minutes.
A
A
Should
also
not
throw
exceptions
and
the
api
and
sdk
should
provide
safe,
defaults
yeah.
Maybe
this
does
not
require
the
error
handling
mechanism
to
be
implemented
in
the
api.
A
Maybe
it
can
be
implemented
all
over
the
place
in
the
sdk,
but
I
think
it's
more
convenient
that
the
api
handles
this
error
handling
mechanism,
because
it
can
act
as
a
central
point
to
say
shown,
but
maybe
that's
just
an
implementation
detail
relevant
to
python
anyway,
I
don't
know
if
somebody
else
has
any
other
comments
or
questions
now,
if,
if
not,
please
leave
your
comments
in
the
pr
and
I'll
stop
sharing.
If
there
are
no
more
questions.
C
Okay
yeah,
so
I
I
don't
believe
everyone
has
taken
a
chance
to
look
at
that
pr.
So
if,
if
anyone
wants
to
get
more
context,
I
believe
that
would
be
the
place
to
start
it's
okay.
If
people
don't
have
opinions
as
of
right
now,
okay,
cool
thanks
diego
for
that,
so
move
right
along,
looks
like
we
have
this
second
topic
up.
It's
the
decide
to
proceed
with
private
modules
for
experimental
features,
all
right,
this
one
yep
alex,
were
you
the
one
who
added
this,
sir?.
B
Yeah
so,
like
I
think
almost
everyone
might
be
to
go
with
the
private
module
our
alex
has
already
here.
For
that.
F
C
Aaron
so
reading
your
comment
that
that
was
pretty
much
the
reason
why
we
didn't
want
to
do
this
yeah,
so
it
would
be
like
an
easier
transition
for
users
later
on
once
it
does
become
stable,
so
they
don't
have
to
like
change
all
their
import
paths
and
stuff,
but
yeah,
because
we've
experimented
this
with
metrics
and
logs
for
quite
a
bit.
It's
just
such
a
burden
for
us
to
maintain
that,
and
we
feel
that,
like
that,
one
time
upgrade
is
like
worth
it.
You
know.
D
Think
it's
more
than
just
a
burden
for
the
maintainers.
I
think
there's
also
the
the
question
that
we
keep
getting
from
the
community
asking
like
where,
where
is
this
metrics
stop?
And
how
do
I
get
it
installed
like
it
feels
like
it's?
It's
not
an
insignificant
burden,
even
on
the
user
side,
so
yeah,
it's
a
burden
in
general,
yeah.
F
So
when
the
metrics
get
to
like
a
usable
state,
are
we
going
to
do
the
same
thing?
Hopefully.
F
D
I
mean
the
caveat
with
even
this
underscore
logs
is
that
it's
all
just
experimental
right.
So,
even
if
we
were
to
ship
like
a
a
somewhat,
not
super
functional,
metrics
implementation,
as
long
as
people
expect
that
it's
experimental,
then
I
think
that's,
I
think,
that's
okay,
but
that's
my
interpretation
of
of
shipping,
something
that's
experimental.
C
C
And
yeah,
not
the
not
the
alpha,
but
we
already
have
like
an
alpha
one,
which
is
like
the
old
metrics
implementation
right
that
people
are
getting
confused
on.
I
believe
like
so
we
haven't
like
we.
We
have
like
a
metrics
new
branch
now
right,
but
that
hasn't
been
released
yet,
which
is
like
a
com,
not
a
completely
different
api,
but
it's
fairly
different
than
the
old
one.
So
I
think
like
once
we
decide
to
release
that
like
that
one
should
already
have
like
prefixed
underscores
yeah.
C
Okay,
cool,
I
believe
we
don't.
I
don't
believe
we
have
any
objections
to
this.
If
so,
yeah
thanks
for
coming
for
bringing
this
up,
I
believe
we
will
be
moving
forward
with
this
decision.
So.
C
Yeah
no
worries
so
once
alex's
pr
is
merged,
we
can
go
ahead
and
just
like
merge
logs
into
main.
I
believe
that
was
the
idea.
C
D
Everybody
here
that
the
merge
has
to
be
like
a
rebase
merge,
not
a
squash,
merge
right.
C
Yeah,
I
made
the
mistake
of
doing
the
first
one
and
we
lost
all
of
our
commit
history
because
of
that
so
yep,
okay,
cool
nice
by
looking
at
the
agenda
today,
doesn't
look
like
there's
any
more
prs
or
issues
up.
C
F
Have
one
non-hologram
related
question
go
ahead,
I
diego
did
you
give
me
some
sort
of
special
permissions
in
the
metrics
branch
or
I
I'm
confused,
because
on
this
on
this
pr,
I
see
the
squash
emerge
button
and
I
don't
think.
A
You're
supposed
to,
I
think,
yeah
I
think
I
had
given
you
special
permissions.
We
discussed
that.
F
I
mean
yeah,
it's
fine.
I
just
thought
we
removed,
remove
them.
A
F
E
A
A
E
It's
a
draft,
I
don't
think
so.
F
It
might,
it
might
be
because
code
owners
which
I'm
going
to
check
now,
but
then
the
other
thing
was,
I
think
you
may
have
done
the
auto
merge
and
maybe,
instead
of
auto,
merging
and
showing
you
that
button.
For
some
reason
I
don't
know.
A
Well,
I
don't
know
we
can
discuss
out
of
land
if
you.
C
F
C
Oh,
I
can.
I
can
just
merge
it
because
I'm
a
maintainer,
my
bad,
never
mind
anyways
yeah
cool.
If
any.
If
nobody
else
has
any
topics
they
want
to
bring
up
yeah,
I
guess
we
can
just
we
could
just
end
early
this
week,
give
back
what
is.
D
D
C
Spooky,
okay,
cool
nice-
I
guess
that's
pretty
much!
It
then
I'll
see
you
guys
next
week
thanks
everyone
all
right,
peace
out,
guys.