►
From YouTube: 2022-11-02 meeting
Description
Open Telemetry Meeting 1's Personal Meeting Room
C
Whoops,
but
thank
you
for
all
your
hard
work
and
my
colleagues
who
are
struggling
in
Lua
also,
maybe
thank
you
as
well.
Yeah
Cheers,
Cheers.
C
C
D
D
D
Discussing
with
the
agenda
items
which
you
have
added
and
I
I
do
have
few
of
the
items
which
probably
I
want
to
talk
in
terms
of
the
issues
but
I
think.
Let's
talk
with
this,
which
you
have
added.
D
A
B
Yeah
so
well,
I,
don't
know
the
all
the
detailed
this
conversation
you
got
February,
but
the
initial
complaint
was
we
duplicated
the
method
in
viscose
chaos,
but
the
issue
is
more
very
much
more
to
it.
The
main
problem
is
that
to
work
with
the
in-memory
spawn
Explorer,
you
need
to
have
access
to
basically
all
the
spanned
data
collected
by
the
exporter
itself
and
so
far
V6
Visa
expose
all
the
basically
all
the
implementation
of
the
exporter,
so
I'm
not
sure
how
to
well.
B
B
B
D
B
D
I
mean,
though,
the
reason
why
I'm
saying
is
that
Trace
API
is
already
stable,
they're,
not
we're
not
going
to
be.
We
know
that
we're
not
going
to
change
to
have
any
change
in
the
trace,
API,
so
I
don't
see.
I
I
know
that
this
instrument
is
a
one-off
change,
which
happened,
and
probably
it
was
not
handled
in
a
very
proper
way,
both
from
specification
and
also
internally,
how
we
implement
it.
We
may
not
have
made
it
as
a
deprecated,
probably
and
kept
supporting
both
instrumentation
library
and
instrumentation
scope
without
making
this
duplicated.
D
That
would
have
been
one
of
the
solutions
and
I
mean
I
got
the
point
that
even
span
data
event
is
something.
Then
we
if
we
say
that
probably
we
have
to
ensure
that
everything
span
that
are
being
expanded,
I
mean
it's
okay,.
B
But
even
something
simple
like
today,
we
use
STD
Vector.
If.
C
D
B
So
yeah
so
for
Trace
we
have
something.
So
it's
a
question
where
we
want
to
rework
it
for
Matrix.
We
have
nothing
yet
so
maybe.
D
That's
something
which
I
feel
that
we
we
we're
not
going
to
change
this
interpretation.
Probably
we
can
say
that
this
is
something
which
can
be
used
safely
and
it
won't,
and
we
will
ensure
that
this
is
something
which
is
not
going
to
change
in
the
future.
D
Okay,
and
with
that,
it
means
that
whatever
it
returns,
spend
data
link
spend
at
a
event
or
all
those
interfaces
they
are
not
going
to
change
in
future.
D
And
for
Matrix
I
think
we
can
implement
it.
The
way
you
have
been
talking
about
I
think
that
probably
the
using
the
Vista
pattern-
okay,
so
that
yeah,
so
that
at
least
at
least
any
application
you
want
to
use
it.
They
want
to
use
the
directly
Matrix
data
they
may
not.
They
don't
want
to
use
mating
data
directly,
in
probably
very
data
content.
D
E
D
B
D
Well,
I
think,
that's
a
that's
a
fair
call,
I
mean
I
I
always
wanted
to
ensure
that
our
open
Telemetry
builds
should
we
should
be
able
to
run
our
open
dream
military
building
in
an
offline
mode,
and
then
you
know
you're
not
having
Internet
available
I
mean
if
I'm
suppose
I'm
on
a
flight
and
I
should
be
able
to
do
the
successful
build
and
at
some
until
sometime
we
were
ensuring
but
I
think
over
the
course
of
time.
Probably
we
lost
that
so
I
think
that
that's
a
fair
call.
D
D
Any
is
it
is,
it
also
required
for
for
the
low
end,
Json.
Also
I
thought.
Probably
it's
only
about
motto.
B
Ultimately,
I
would
like
to
have
these
Founders
Independence
is
so
far
I've
implemented.
This
for
Proto
went
for
endometrios.
D
Okay,
now
I
mean
do
we
need
do
we
need
any
such
any.
Such
changes
for
Json
also
like
this.
This
kind
of.
B
E
D
B
It
might,
but
it's
well
fine
package
is,
is
causing
trouble
for
us
also
because
if
there
is
a
deep
Earth
package
installed
by
default
on
the
system,
we
may
not
even
on
this
one
but
want
a
different
one.
So.
D
D
A
I'm
thinking
about
like
we
have
some
requirement
to
write
on
like
the
portal
version
like
in
protocols,
so
it's
really
like
somebody
can
change.
So
if
the
user
bring
his
own
dependency
and
which
may
be
the
incompatible,
how
can
we
tell
that
to
the
user,
whether
it's
just
in
the
compile
error.
B
A
B
B
B
B
Should
and
actually
I
phrased
an
issue
against
the
portal
repo
itself
to
ask
that
there
is
a
version
file
added
to
the
top
so
that
when
we,
when
we
get
an
open
tournament
report
or
we
can
inspect
what
the
version
is,
there.
A
A
B
You
get
issues
against
open,
Telemetry
portal,
it
would
be
there.
E
D
The
latest
version
of
open
delivery
this
place
and
let
the
build
field
yeah
they
may
have
to
come
back
to
us.
Probably
anybody
who
is
using
the
older
version-
and
we
have
to
reply
in
that
case
in
the
but
yeah
I
mean
if
they
don't
if
opportunity,
Proto
does
not
provide
a
version,
then
probably
let
the
build
fail
and
let
them
come
back,
go
through
the
documentation
and
see.
This
is
the
right
version.
To
use
I
mean
I
I.
D
B
Which
version?
No?
This
is
not
the
intent.
It's
just
to
have
the
portal
file
the
open,
Timer
file
on
the
file
system
already
somewhere
to
point
at
it.
D
B
Yeah
I
was
thinking
to
interest
labels
for
that.
E
D
Yeah
I
think
it
should
be
okay
unless
I
don't
see
any
problem.
Having
that.
B
D
B
A
D
D
D
D
We
treat
them
as
absolute
values
of
counter,
and
we
use
that
that
as
the
value
of
the
final
value
and
propagate
it
for
the
or
send
it
further
based
or
whether
it's
a
Delta
accumulative,
but
there
could
be
a
scenario
where
and
and
and
the
values
compared
based
on
what
what
attributes
are
coming
along
with
the
along
with
that
value.
But
if
the
attributes
are
filtered
out
that
creates
a
that
creates
a
confusion.
Suppose
the
attributes
are
version
is
equal
to
2
and
version
is
equal
to
one.
D
Yeah,
so
this
is
this
is
something
how
how
if
we
are
getting
some
measurement
like
measurement
value,
one
one
version
version
one
and
measurement
value
2
for
a
version
two.
So
if
the
views
are
views
are
removing
those
attributes,
then
then,
at
the
aggregation
level,
we'll
just
get
value
1
and
value
two.
So
since
it's
a
AC
counter,
what
the
agree,
the
some
aggregation
will
do,
it
will
use.
It
will
take
the
final
where
the
last
value
we
should
get.
D
It
assumes
that
both
these
measurements
are
coming
for
the
same
set
of
attribute,
because
these
are
the
null
attributes.
So
it
will
just
take
the
last
value,
and
that
would
be
the
actual
value
for
this
measurement
and
it
will
use
that
value
as
the
aggregated
Matrix,
which
is
not
the
right
viewer,
and
this
is
the
behavior
which
which
all
the
languages
have
implemented.
D
But
this
is
not
the
expected
Behavior,
because
if
we
remove
these
version
tags
I
mean
if
these
attributes
are
removed.
Actually
we
should
sum
both
these
values
and
not
just
use
the
last
value
and
that's
which,
because
because
at
least
they
are
are
we
dead?
D
I
mean
in
all
their
three
different
versions
coming,
even
though
version
values
are
different,
but
actually
these
are
all
are
related
to
Origins,
so
they
should
sum
all
these
measurements
as
they
regulated
Matrix.
So
that's
somewhat
tricky
I
I
need
to
check
how,
if
other
languages
are
going
to
do
it,
how
they
will
Implement,
because
at
once,
once
these
measurements
come
at
the
Matrix
at
the
aggregation
level,
we
don't
have
any
track
of
what
kind
of
attributes
it
was
having.
D
So
we
don't
know
whether
we're
leaving
to
submit
or
we
just
need
to
use
absolute
pointers
so
yeah,
that's
that's
just
wanted
to
call
out,
and
it's
probably
it
may
need
some
design
changes.
If
we
have
to
support
it,
but
I,
don't
know
any
idea,
I
don't
know
either
either.
We
should
maintain
these
attributes
tag
them
as
deleted
when
they
reach
the
Matrix
aggregation
level,
but
you
have
some
something
something
we
may
have
to
see
outwardly
do
this.
D
This
is
not
not
inconsistent
in
compliant
with
the
specification,
because,
as
far
as
specification,
we
should
be
doing
the
sum,
but
the
good
part
is
that
none
of
the
languages
are
compliant
as
of
now
so
yeah
I
just
wanted
to
call
out
that
we
may
have
to
fix
it.
I'm,
not
sure
how
we
will
do
it.
We
need
some
changes
in
the
way
we
have
designed
the
current
ASU
progress,
education.
D
D
We
are
having
Tom,
that's
it.
That's
a
good
good
issue
is
I.
Think
I
don't
know
how
to
solve
it.
So
the
test
cases
is
basically
trying
to
access
a
non-existing
address
and
if
we
can
have
a
local
Loop
address,
which
is
non-existing
if
we
can
use
that,
probably
that
should
not
cause
this
Firewall
Alert.
A
A
D
B
D
D
A
D
D
D
Think
something
still
in
external
letter:
it's
not
a
local
loop,
I
create
local
Loop
address.
So
it's
even
though
this
cannot
be
used
by
any
in
the
public.
Ip
address
space,
but
still
I,
think
the
HTTP
client
is,
will
try
to
create
a
connection
or
try
to
create
an
outside
connection,
so
they
still
try
to
go
outside
the
firewall
or
something
but
yeah.
Please
try
to
see
if
you,
if
I
mean
if
I
remember
when
we
changed
this.
D
D
D
A
A
D
Yeah-
and
this
is
something
I
mean
I'm-
not
expecting
using
multiple
meters,
but
I-
think
he's
going
one
step
ahead
and
I
I
say
that
I
never
tried
this
scenario,
multiple
meters
and
we
don't
have
any
tests
for
that.
Also
so
probably
I
think
to
add
some
tests
and
try
to
see
this
is
causing
any
problems.
D
We
are
in
this,
we
already
discussed
yes,
give
us
a
fall
twin
metric
provider
tries
to
lock
to
already
destructed
Global
log
Handler
at
Exit
yeah.
We
have
to
do
some
effects
for
this.
This
is
not
specific
to
metric
provider,
it
will
happen
even
for
race
provider.
It's
just
that
interest,
provided
we
during
the
shutdown
of
Trace
provider.
We
are
not
logging
anything
or
we
haven't
seen
any
errors
which
are
getting
locked.
D
So
probably
the
problem
never
came
here,
but
we
added
some
logs
internal
logs,
something
like
this
shutdown
could
be
invoked
only
once,
and
this
will
also
be
called
if,
through
the
destructor
of
Tracer
provider,
it
calls
a
shutdown
into
place,
a
provider
which
calls
the
shutdown
on
meter
contacts,
and
we
try
to
logs
this,
but
somehow
somehow
the
the
local
static
or
the
global
static
default.
Handler
is
already
already
destructed
before
this
is
called
and
it
yeah
because
of
that
I
think
it
crashes
so
yeah
either.
D
What
we
can
do
is
that,
if
we,
if
we
call
this
Colony
one
of
these
macros
blogging
macros
in
the
Constructor
of
the
meter
provider
metric
was
your
metric
provider.
I
think
that
she
will
also
solve
it
or
if
we
remove,
if
we
remove
the
shutdown,
if
we
ensure,
if
we
don't
call
the
metric
provider
shutdown
in
the
sorry
meter
provider,
shutdown
inside
its
Destructor,
which
will
ensure
that
somebody
has
to.
They
have
to
explicitly
call
this
shutdown.
That
also
solves
the.
D
D
A
B
That
mean
that
we
should
go
away
from
Global
single
times,
and
maybe
you
have
an
explicit,
initialize
and
and
shut
down
API
for
the
SDK
as
a
whole,
so
that
we
do.
D
D
At
least
for
the
logger
or
the
default
log
Handler,
we
should
definitely
yeah
remove
the
global,
remove
the
the
Singleton
implementation
for
this.
If
you
can
do
that
in
an
easy
way,
I
mean
that
should
be
a
more
proper
solution.
D
I,
don't
think
we
can
do
it
for,
for
so
so
beta
provider
is
not
Global,
Singleton
yeah,
it
is,
it
is
sorry
it
is,
but
I
don't
think
if
we
can
remove
it,
because
this
is
being
done
in
this
is
this
is
something
which
is
done
in
in
the
API
level,
which
is
all
ended.
D
D
D
D
D
Yeah
this
one
right:
yes,
yes,
which
is
right,
I,
think
it's
so
we
try
to
be
Ava
compliant
to
start
with,
but
then
we
we
never
said
that
we'll
Garden
guaranteed
yeah.
So
probably
we
can
remove
this
comment
or
we
can
change
this
in
a
to
give
give
a
better
understanding
of
that
we're
not
even
be
a
compliant.
E
B
D
B
I
think
it's
a
tricky
part,
because
if
we
start
moving
things
around
and
changing
namespace
and
whatnot
chances
are
that
we
will
have
to
do
it
again
for
the
different
reason.
So
it's
a.
E
B
Yeah
yeah,
sorry,
the
question
is,
first
of
all
how
many
different
apis
do
we
have
because
one
application
using
an
exporter
to
set
up
an
open
Telemetry,
is
using
one
kind
of
interface.
Another
application
implementing
their
own
exporter
will
probably
use
different
interfaces,
so
even
there
I
mean
one
Ado
file
can
be
public
for
one
and
and
not
for
the
overall.
So.
D
It's
It's
Tricky,
so
yeah,
so
we
have
to
document
it
according
to
that
way,
basically
for
instrumentation
libraries,
whatever
the
public
head
office
for
them
for
the
application
developer.
What
are
the
public,
whether
files
and
even
for
the
exporter
developers,
so
it
will
change
everywhere
right,
it
will
change
based
on
yes,
what's
the
role
of
the
developer,
who
is
using
the
open
database.
D
It
should
we
need
to
document
it
according
to
that,
but
yeah,
it's
a
tricky
part.
Even
we
have
to
first
figure
out
what
was.
B
For
enforcement
well
different
name
spaces.
One
way
we
can
add
a
special
Define
where,
if
you
compile
in
kid
mode,
you
only
see
inverter
files
you're
supposed
to
use
things
like
that
I
mean
there
are
different
ways
to
fix
it,
but
the
most
important
is
first
us
to
to
Define
which
other
file
is
or
can
can
or
cannot
be
used.
D
Okay,
I
think,
probably,
let's
probably
figure
out
what
are
the
public
headers
and
then
probably
start
for
each
of
the
developer
roles
and
then
probably
will
start
discussing
how
commercial
we
handle
it.
B
Yeah
and
I
guess
with
points
to
video
commutation
that
we
we
have
to
write
yeah.
D
Yeah
and
then
probably
we'll
start
talking
about
the
documentation,
I
can
pick
to
see
what
are
the
public
apis
unless
somebody
wants
to
pick
this
up
or
feel
free
to
do
it,
at
least
at
least
for
the
first
part,
or
what's
public
what's
knock
and
then
probably
we
can
talk.
Otherwise,
let
me
explain
some
spend
some
time
teaching
understand
what
what
really
is
public
and
then
probably
next
next
meeting
we
can
discuss
further
on
this
sure
and
I
can
document
it
further.
I
can
start
start
writing
in
the
comment.
D
E
D
Getting
started
doing
yeah.
D
This
is
for
foreign.
D
A
I
would
take
a
look
yeah.
A
D
E
D
D
Hey,
would
you
have
some
time
to
look
into
that
into
this
I
mean
not
to
solve
it,
but
just
to
see
if
we,
if,
what
exactly
is
they
issued
here.
D
I
think
probably
we
have
two
at
one
time
ensure
that
we
are
able
to
bump
delete.
We
are
able
to
use
the
latest
versions
in
both
seem
against
bees,
and
so
probably
you
can
discuss,
even
if
there
are
some
something
which
are
really
blocking
in
this
I
I
know
there
are
some
issues
which
we
Face
earlier
and
we
discussed
it
here
also.
D
B
D
D
D
D
D
For
suppose,
ga's
is
nothing
which
is
pressing.
As
of
now.
Support
for
exponential
is
not
very
urgent.
We
have
to
do
it.
Sometimes
just
we
already
discourage
provide
Windows.
You
know
these
are
doing.
Whenever
you
have
time
arguments,
yeah
I
can
take
it
now.
So
it's
in
memory,
Matrix
exporter,
you
have
this
something
which
we
need
to
implement
and
clean
up
so
old
Matrix.
So
this
is
yeah.
That
is.
B
Yeah
for
the
memory
metrics
as
part
of
USB
soul,
patch
by
yeah
and
I,
think,
but
we
probably
need
to
start
over,
especially
if
we
don't
want
to
expose
the
the
total
buffers.
C
B
D
Yeah
I
mean,
as
of
now,
I
think
it's
not
assigned
to
anyone
and
probably
I'm,
not
keeping
it
in
very
high
priority.
Okay,
it's
mostly
for
debugging
purpose,
but
I
think
over
the
course
of
current
ability.
We
want
to
implement
it.
B
Okay,
but
I
will
start
with
builders
in
general,
first,
okay
and
look
at
this.
Maybe
after.
D
B
D
Yeah
right
now,
I
think
the
very
most
of
them
I
mean
I,
have
seen
I
mean
I'm
kind
of
tracking
the
build
time
for
all
our
CI
actions
and
the.
E
D
Windows,
build
is
something
very
slow,
I
think,
most
of
time.
It
is
taking
lots
of
time
it's
not
using
the
cash
which
bazel
I
think
we
only
are
using
the
caching
from
the
previous
CI
bills,
but
somehow
somehow
it
looks
like
that
Windows
bazel
bit
is
not
using
that
cache.
So
probably
you
need
to
see
Health
Care
way,
that
issue
for
that,
and
that's
probably
taking
because
of
that
I
think
our
CI
builds
take
lots
in
lots
of
time.
B
D
That's
that's
a
good
point.
Yeah,
it's
possible
it's
because
of
the
VC
package.
I,
don't
know
how
the
cash
is
used
for
Visa
internally,
but
if
it
is
VC
package,
you're
saying
that
those
builds
the
VC
package
builds
are
not
used
in
subsequent
CIS
I.
D
That's
that's
possible
yeah,
but
I'll
raise
the
issue
for
that
and
probably
if
you
can
just
track
track
it
using
that
and
yeah,
but
in
general
bazel
builds
because
of
you.
They
are
using
the
cash.
They
are
much
faster,
at
least
for
non-windows
environment
as
compared
to
cmic
in
general.
Take
cloud
compared.
It
will
be
more
time
than
bazel,
so
probably,
but
still
the
basal
Windows
is
more
of
a
bottleneck
and
it
takes
most
of
the
time
out
of.
D
D
Yeah,
this
is
already
I've
taken
shared
pointers
of
facing
you
know.
Everyone
with
me
on
this
is
and
yeah.
This
is
with
me.
I'll
take
some
time
to
remove
this
dependency
of
using
the
big
particle,
getting
started,
API
documentation,
yeah,
let's
keep
it
open
till
till
we
ensure
that
we
are
have
completed
it.
We
have
all
the
documentation
meetings
and
your
rest,
I
think
we
have
already
discussed
and
for
the
PRS
we
have.
D
D
A
A
E
A
Thanks
yeah
I
also
just
submit
a
PR
to
visit
packages,
so
1
37.0
will
be
there.
A
Thanks
Tom
thanks
for
thanks,
though
yeah
the
review
was
I,
think
the
pra
is
accepted
and
will
be
merges
soon.
Yeah.
D
Thank
you
for
tracking
it.
I
mean
in
that
note.
Probably
we
can
talk
about.
Let
me
just
I
think
we
have
just
four
minutes
so
that
reminds
me
that
it's
while
I
think
Tom
you're,
we
are
tracking
the
VC
package
thing
internally.
There
was
few
people
who
also
added
open
Telemetry
in
on
Conan,
but
I.
Think.
D
A
D
Okay,
yeah
thanks
for
joining,
have
a
good
rest
of
the
week
next
week,
thanks.