►
From YouTube: 2022-01-20 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
C
Okay,
so
hello,
santos,
you
are
coming
to
the
open,
telemetry
swift
meeting.
A
Yes,
I
have
been
in.
C
B
D
D
C
C
C
C
C
Currently
most
of
the
code
is
added
by
us
in
a
more
or
less
continuous
manner,
but
from
time
to
time.
Some
other
peers
come
to
the
project
and
we
are
really
happy
to
receive
any
help
and
approve
and
help
with
anything
so
yeah
and
yeah,
and
we
and
we
keep
this
spread
as
a
as
part-time
of
our
jobs
in
our
company.
So
yeah
I
work
at
datadog
and
yeah
and
bryce
will
say
himself
so.
B
Yeah
yeah
yeah,
I'm
guess
where
to
start
yeah
I
work
at
elastic
and
I'm
the
primary
engineer
on
our
ios
sdk.
It's
it's
100
built
on
top
of
open
tiller
entry.
Swift.
So
you
know
my
work
at
elastic
is
bound
very
closely
to
the
open
telemetry
project.
So
yeah.
A
Okay,
I
can
introduce
myself
too,
so
my
name
is
santosh.
I
am
from
cisco
the
app
dynamics
group
within
cisco
and
only
recently,
I
think
about
three
months
ago.
A
I
started
participating
in
the
open,
telemetry
aspects
at
appd
app
dynamics,
and
I
am
specifically
focusing
on
the
client-side
telemetry,
not
necessarily
just
the
ios
part,
I'm
not
an
ios
developer,
but
I'm
a
kind
of
a
tech
lead
coordinating
everything
on
the
client
side,
the
end
user
side
ram
side,
so
I'm
coordinating
with
folks
working
on
ios,
android
and
javascript,
and
recently
I
think
it's
been
two
months
now.
There
was
another
sig
that
started
talking
about
client-side
specifications
specific
to
the
client-side
telemetry.
A
There
are
folks
from
new
relic
and
splunk
there.
It
would
be
great
if
you
know
some
of
you
also
join
there.
There's
a
slack
channel
called
client
side,
something
client-side
telemetry
there.
There
is
one
topic.
You
know
that
I
wanted
to
discuss
with
you
folks,
specifically
and
and
that's
why
I'm
here,
cool.
B
Yeah
yeah
I've
been
meaning
to
to
join
that
that's
sig.
More
often,
I
know
it's
it's
on
wednesdays
right
at
about
the
same
time
as
this
meeting.
C
Yeah,
I
know
that
the
splunk
are
also
writing
their
ram
a
product
in
top
of
a
telemetry
shift.
They
have
some
time.
I
don't
work
on
ram
myself
and
the
that
data.run
product
is
still
not
on
top
of
open
translation.
I
hope
someday
they
will
but
yeah.
C
I
I
work
on
ci
visibility
and
there's
visibility
on
swift,
so
I
use
open
telemetry
swifter
database
also
for
my
traces
in
my
product
for
scripts
and
yeah
yeah,
but
I
I
know
about
the
yeah
the
clean
side
and,
of
course,
is
that
one
of
the
main
focus
of
this
product
is
helping
them
having
support
for
all
the
client
side,
features
or
specs
that
they
come
up
with
because
it's
we
know
it's
a
a
first
case
use
for
for
this
library,
so
yeah.
C
Great
so,
let's
start,
let
me
share
the
document
just
to
have
something
to
look
at,
so
we
can
talk
about
so
topics,
not
much
less
meeting
actions
yeah.
We
I
I
released
a
version
of
the
library
last
week
that
supported
a
concurrency
in
swift,
the
concurrency
librarians
used
for
xcode
39.2
and
app.
C
D
C
Just
posted
the
first
version
one
hour
ago,
but
they
had
a
ask
me
anything
session
with
our
cto
in
my
company,
so
I
couldn't
check
much
of
it.
Yeah.
C
That
we
could
talk
about
sure.
So
if
we,
if
we
have
time
after
the
topic,
we
can
just
take
a
first
review.
If
you
want
or.
C
C
Okay,
okay
and
the
topic,
I
think
you
already
discussed
modeling
of
events
outside
of
a
trace,
yeah.
A
Yeah,
let
me
give
you
a
brief
overview
of
the
discussions
there
so
far
and
and
then
I
would
like
to
get
your
thoughts
because
there
has
been
no
representation
from
the
ios
side
and
that's
why
to
get
your
thoughts.
So
basically
we
are
talking
about.
You
know
events
outside
happening
outside
of
a
trace
context.
A
A
Are
you
know,
network
change,
events
and
things
like
you
know,
even
crashes.
I
know
I
saw
that
you,
you
already
support
application
crash,
so
I
wanted
to
understand
that.
So
how
do
you
you
know
mod
represent
them
which
what
data
model
you
know?
Do
you
use
to
represent
these
events
and
in
that
other
sig
we
went
back
and
forth
between
using
zero
duration,
spans
and
also
log
records.
A
I
think
initially
many
of
us
felt
that
the
this
using
a
span,
zero
duration
span-
you
know,
sounds
you
know
not
so
meaningful
and
I
think
log
records
is
is
more
appropriate
because
the
the
logs
specifically
says
that
you
know
the
logs
model
is
to
be
used
for
events.
A
A
The
events
have
a
name,
a
type
name
or
a
type
typically
and
the
logs.
Don't
the
logs
have
severity
and
severity
level,
which
you
know
events
don't
have.
A
There
are
some
minor
disparities,
but
otherwise
log
records
is,
is
a
good
fit,
but
then
again
you
know
there
was
another
argument
that
popped
up
saying
there
are
situations
where
you
can
detect
causation
like
this
event
caused
that
event,
and
that
is
best
represented
using
spans
in
a
trace,
because
you
have
the
parent-child
relationship
and
if
you
represent
two
events
as
a
parent
and
child,
you
can
say
that
hey
this
parent
event
caused
this
child
event
and
therefore
you
know
the
spans
is
a
better
fit
for
for
that
situation.
A
Of
course,
not
all
events
for
not
it's
not
possible
to
detect
the
causality
in
all
situations,
but
at
least
you
know
wherever
feasible
so,
and
that
is
where
you
know.
I
wanted
to
hear
your
thoughts.
Is
that
even
feasible
to
detect
causality?
And
if
so,
you
know
what
are
those
situations,
because
when
we
say
zero
duration,
the
context
propagation
also
is
tricky.
You
know
so
so
they
will
be
situations
I
mean
for
these
situations.
The
duration
is
not
a
you
know:
zero
duration.
A
They
will
have
a
certain
time
span.
So
do
you
have
use
cases
scenarios
in
the
ios
where
this
is
applicable,
or
this
is
feasible.
B
Yeah
that's.
This
is
something
that
I've
often
like
thought
about
is
just
like.
You
know
it
would
be
really
nice
if
there
was
a
way
to
record
an
event,
that's
not
associated
with
a
span,
because
you
know,
unlike
the
server
side
kind
of
agents,
there's
really
not
always
a
span
running
on
ios.
You
know
it's
like
a
very
more
like
ad
hoc
kind
of
thing,
but
whether
it
should
be
represented
as
a
span
or
as
a
log,
it
seems
yeah.
B
It
seems
like
it's
really
just
trying
to
work
with
what
we
got
rather
than
what
we
should
have
in
terms
of
like,
like
an
event
like
it
makes
sense
as
a
log
to
me,
but
yeah
you
bring
up
a
good
point
where
logs
are
logs,
you
know
like
they're
their
specific
thing
and
adding
an
event
to
it
is
kind
of
a
weird
you
know,
usage
of
that.
B
I
can't
think
of
a
real
world
instance
where
you
would
have
a
like
a
event
that
is
tied
to
another
event,
a
result
of
another
event.
Do
you
have
an
example
of
that.
A
B
Yeah,
that's
that's
something
that
I've
tried
to
model
in
in
spans
as
well,
and
I've
always
just
used
the
duration
of
the
the
click
event
button
or,
like
the
click.
You
know
method
as
like
the
span,
duration
and
then
trying
to
capture
you
know
through
propagation
like
what
actually
happens
after
that.
So
yeah.
It's
interesting,
I
still
think
yeah.
Maybe
it
makes
more
sense
to
treat
events
as
spans
and
yeah
and
yeah,
but
yeah
it's
it's.
It's
a
tough
one.
It's
a
tough
one.
A
Yeah,
so
maybe
you
know,
you
know,
take
your
time
think
about
it
and
then
maybe
we
could
revisit
this
next
week
or
later.
If
you
get
any
ideas.
I
just
wanted
to
put
this
in
your
minds
because
I
I
know
it's
not,
there
were
not
many.
You
know
use
cases
there
too,
but
it
was
more
like
hey.
A
If
we
include
this
as
one
of
the
options
in
the
apis,
then
you
know
whenever
in
the
future,
if
we
need
to
represent
causal
events,
then
this
is
the
api
to
use.
C
Yeah,
I
personally
think
that
zero
duration
span
is
a
trick.
I
mean,
is
a
sort
come
to
a
problem
that
you
don't
know
how
to
solve.
So
you
use
this
trick.
So
that's
not
fair.
I
mean,
if
you
don't
have
any
other
thing,
then
yeah
it
works,
but
it's
not
what
you
expect.
C
So
I
I
was
thinking
if
we
cannot
solve
that
with
the
things
we
have.
We
might
need
another
instructor
that
talks
about
sessions
that
can
have
events
in
it
and
where
you
represent,
that
I
mean
you
don't
have
to
send.
I
mean
one
of
the
things
that
happens
with
expanse
is
that
you
have
to
end
the
span
in
order
for
that
to
be
an
entity
and
to
transmit
everything
that
happens,
that
something
cannot
work
for
some
for
for
some
features.
C
For
example,
when
you
are
running
an
application
you
couldn't,
you
cannot
represent
an
application
by
a
spam,
because
the
application
can
end
abruptly
with
a
grass,
and
you
cannot
then
done
this,
that's
fun.
C
C
So
maybe
there
should
be
a
way
that
cannot
be
solved
only
on
the
client
side,
but
the
back
end
must
know
about
the
context
of
things
that
are
happening,
so
they
will
sense
that
you
send
here
start
my
app.
The
backend
knows
that,
after
that
it
will
create
something
that
will
last
as
much
as
the
app
and
that
thing
so
I
don't
know
if
that's
something
we
can
solve
only
from
the
client
side,
or
it
should
be
something
more
in
the
spec
that
should
also
share
with
the
with
the
process.
C
That
is
how
to
say,
reading
that
instrumentation
and
and
getting
sense
from
the.
A
I
don't
know,
maybe
yeah
yeah.
I
think
sessions
specifically
can
be
detected
on
the
back
end
as
well,
but
then
they
require
you
to.
A
You
know
judge
based
on
whatever
is
received
so
far,
so
we
need
some
good
heuristics,
our
good.
You
know
signals
from
the
client
which
are
kind
of
deterministic
to
say.
Okay
now
the
session
has
started
now.
The
session
has
timed
out,
or
the
session
has
explicitly.
You
know,
ended
in
cases.
C
A
C
Need
a
new
construction
that
it's
just
signals
for
example,
and
and
create
something
around
that
signal
thing.
That
is
something
that
may
could
could
have
yeah
an
origin
or
a
parent
or
or
a
reason
that
is
apparent,
can
be
a
root
in
a
similar
way
to
to
expanse
itself,
but
with
differences
that
yeah
that
are
really
different.
So
I
would
say
we
need
something
more.
B
C
Fix
that
behavior
tricking
the
the
constructs
we
already
have
yeah
and
support
yeah
yeah.
That's.
B
I
think
that
in
all
my
experience
I
see
this
problem
in
in
apm
a
lot
where
the
initial
offering
is
focused
on
like
the
server
side
kind
of
thing,
and
then
you
know,
client
side
comes
along
and
then
we
try
to
fit
all
of
the
all
of
the
the
the
constructs
of
client
side
into
the
server
side
model,
and
that
ends
up
not
working,
and
I
kind
of
feel
like
we're,
reproducing
that
with
the
open
telemetry
spec,
where
we
have
these
three
signals
that
open
telemetry
uses-
and
they
were
kind
of
you
know-
envisioned
from
the
server
side
model
and
now
we're
we're
trying
to
fit
client
side
into
it.
B
So
I
I
kind
of
feel
like.
Maybe
what
we
should
look
at
is
the
span.
Spec
does
have
the
idea
of
an
event,
and
the
idea
is
that
event
is
associated
with
a
span,
but
maybe
we
need
to
elevate
that
event
so
that
it
can
be
a
standalone
kind
of
object,
but
still
provide
the
ability
to
associate
with
a
span
if
necessary,
right.
Maybe
that's
where
we
should
kind
of
kind
of
go
at
this
from.
C
Yeah,
I
mean
it's
a
behavior
with
a
with
an
event,
but
the
problem
is
that
you
don't
have
a
span
to
contain
it,
so
maybe
events
could
be,
as
you
say,
yeah
that
could
work
perfectly.
I
mean
having
an
event
that
can
have
or
can't
have,
a
parent
that
is
a
span,
a
related
span
or
can
be
as
a
root
event
with
some
information
that
just
is
transmitted
yeah.
That's.
F
C
B
Knock
on
my
door,
yeah,
you
know
what
might
be
so
this.
This
is
an
interesting
idea,
like
maybe
the
the
binding
of
the
event
with
a
span
needs
to
be
a
little
bit
more
generic,
and
maybe
you
can
have
a
root
event.
That's
that
kicks
off
spans
right.
B
So
this
I'm
thinking
like
an
application
launches
and
that's
an
event
and
then
spans
might
be
associated
with
that
launch
event
after
the
fact,
or
you
might
have
a
span
that
generates
events,
and
then
you
can,
you
know,
kind
of
associate
them
back
and
forth,
but
yeah.
Maybe
we
can
play
around
with
that
idea,
a
little
bit
just
a
thought.
A
Sure
sure,
so
what
I'm
hearing
is
that
a
standalone
event
is
is
what
is
preferred,
and
I
want
to
say
that
the
in
the
logs
sig,
I
have
attended
log,
sig
and
you're
right
that
most
of
the
people
come
with
the
the
back
end.
Telemetry
experience
more
than
the
the
end
user
side,
the
client
side-
and
I
don't
understand
the
hesitation,
but
they
are
very
reluctant
to
add
another
top
level
signal.
A
A
So
then
I
think
we
could.
You
know
standardize
on
that,
because
I
think
that
covers
most
of
the
scenarios,
because
the
see
the
use
case
of
identifying
causal
relationship
between
events,
it's
more
of
a
you
know,
a
possible
thought.
I
mean
a
thought
for
a
possibility,
but
people
don't
have
clear
examples
on
you
know
what
those
are
at
this
point:
yeah.
B
Yeah
well
yeah,
it's
it's
it's
just
odd
because
we
do
have
the
idea
of
an
event
in
the
spans,
but
the
problem
is:
is
that
spam
like
they're
a
little
bit
the
the
relationship
between
them
is
too
narrow,
like
spans,
only
have
child
events,
but
I
kind
of
view
it
as
events
might
have
child
spans.
I
don't
know,
I
think
I
think
it
could
go
either
way.
B
I
see
at
least
on
client
side,
because
we
don't
always
have
a
span
running.
You
know
it.
Doesn't
it
doesn't
make
sense
like
you
do
like
yeah,
because
you
have
yeah.
This
is
just
like
the
the
client
side
versus
server
model.
You
know
servers
have
an
entry
point
that
have
a
terminus,
whereas
client-side
is
just
nebulous
like
application
is
on
and
you
don't
necessarily
want
to
have
a
span.
That
starts
every
time
the
application
comes
into
the
foreground
or
is
yeah.
C
C
A
That
doesn't
fit
yeah,
I
think
sessions.
Clearly,
you
know,
should
not
be
modeled
using
spans.
You
know.
Instead,
you
could,
you
should
look
at
you,
know
session
start
session
timeout
or
a
session
and
app
coming
to
the
foreground
app
going
to
the
background.
So
we
should,
you
know,
identify
yeah.
C
As
logs
is
like,
loss
are
things
that
come
to
the
output.
Somehow
that
you
know
logs
are
output,
you
saw
to
the
user.
These
are
not
thought
to
be
for
users
consuming.
So
I
I
I
don't
like
that
approach
I
mean
yeah.
You
can
use
that
the
same
way
you
can
use
a
sponge
with
zero
duration.
I
mean,
when
you
only
have
a
hammer,
you
see
nails
everywhere,
it's
like
tricking
the
the
system
or
the
language
or
or
the
construction
to
fitting
your
needs
yeah.
It
doesn't
feel
right
for
me.
B
A
It's
interesting
yeah
a
session
can
contain
anything
and
everything
yeah.
You
know
that
happens.
You
know
in
that
duration
or
in
that
session
span,
and
then
we
could
have
a
different.
You
know
strategies
on
what
defines
a
session.
You
know,
you
know
I
think
different
people
might
have
different
ideas
about
it,
yeah
yeah.
A
So
I
have
a
question.
I
see
that
you
support
reporting
crash
of
your
ios
app.
I
saw
somewhere.
How
do
you
represent
that
today?.
C
You
mean
bryce.
A
C
I
mean
in
data
we
we
support
project,
we
have
a
crash
reporter
for
ram.
I
mean
we
have
a
crash
report
there
for
applications
running
yeah.
I
mean
yeah
go
ahead.
I
also
I
also
support
crashing
inside
tests,
so
I
represent
them
as
a
spam.
With
that
with
the
duration,
I
mean
what
they
do
for
if
you're
asking
about
what
I
do
for
tests,
for
example
in
my
project,
that's
not
the
ram
the
crash
reporter
in
datalog,
but
I
captured
the
process
you.
C
I
am
using
a
crash
reporter
and
I
am
setting
custom
data
in
the
crash
report
that,
if
a
crash
happens,
it
saves
the
that
data
in
the
class
information
so
each
time
I
span
in
my
case
each
time
a
test
is
created.
That's
my
root
span.
Let's
say
that
I
want
to
save.
I
save
that
information
inside
the
cross
report
and
whenever
the
application
crosses
the
next
time
is
launched.
C
C
I
recreate
the
span
with
the
start
time
that
I
had
the
crash
time
that
I
recovered
from
the
crash
information.
I
might
not
get
that
time
of
the
crash,
but
then
I
I
set
some
fixed
time.
I
said
yes,
one
millisecond
because
of
I
don't
want
to
set
zero
duration
response
and
that's
the
way
I
do.
C
I
mean
you
could
save
any
number
of
expanse
there
in
the
crash
reporter,
and
you
said
that
when
you
create,
but
that
can
also
take
some
time
from
you
and
I
think
that
they
are
using
the
same
functionality
in
the
in
our
run
product
because
yeah
they.
I
implemented
that
in
the
crash
reporter
the
setting
custom
data
function,
the
crash
reporter
we
use
because
we
need
it
for
the
test
thing,
but
they
are
also
using
the
same.
C
A
Okay,
so
when
the
application
starts
again,
when
you
report
it,
you
report
that
crash
through
as
events
in
a
span.
C
I
I
mean
I
do
that
for
the
tests
for
sure
I
recreate
the
span,
because
it's
for
me
in
my
product
is
its
test
is
a
span.
It
can
have
child
spans.
There
are
network
requests
and
things
like
that.
But
if
the
test
crosses
inside
the
test,
then
I
report
a
span
with
a
status
of
failure,
but
yeah
it's
a
span
with
the
duration.
I
get
from
the
crash
time
a
minute.
The
start
time
of
the
span.
A
Okay-
and
I
know.
C
For
the
run
thing,
they
also
said
other
things
I
mean
they
are
saving
the
ram
session,
that
it's
not
a
spam
and
they
are
saving
other
views
on
things
that
are
not
strictly
expands
in
the
for
our
own
product.
They
they
don't
not.
Everything
is
modeled
with
response
and
trade,
so
they
are
recovering
other
things.
A
Also,
okay,
and
before
the
crash,
you
start
the
span
because
there
was
a
network
request
going
or
even
otherwise
you
start
a
spam.
C
In
my
case,
it's
because
I
create
a
I,
for
I
create
a
spam
for
each
test
and
everything
rises
inside
the
test.
C
Once
the
test
has
started
so
yeah,
I
have
a
spam
alive
and
whenever
I
create
a
test
span,
I
set
that,
as
the
I
said,
that
in
the
cross
reporter
custom
data,
so
I
can
recover
that
okay,
okay.
C
A
Okay,
all
right,
I
think
that's
good
information.
A
So
I'll
take
this
back
to
the
other
team
on
on
wherever
you
know
what
your
opinion
is
and
if
you
want
to
join,
you
know
you
could
join
that
those
meetings.
It
will
be
super
helpful
to
have.
Somebody
represent
the
ios
set
of
things
yeah
I'll,
try
to.
C
Okay,
okay,
I
can
I
mean,
if
there's
a
topic
that
can
be
interesting
about
the
spec
and
things
like
that
I
can
join.
I
mean
I
don't
usually
join
because
I.
B
C
C
Yeah,
so
just
because
I
am
not
about
program
or
client
client
topics
currently
so
just
was
missing
those
because
I
knew
bryce
went
usually
or
from
time
to
time
and
yeah,
and
they
told
that-
and
I
expected
them
to
tell
us
what
kind
of
specs
they
want
to
have,
or
tax
or
whatever,
and
we
would
support
that.
But
if
they
want
to
add
something
more
structural
or
something
like
that
yeah
I
yeah.
I
will
be
interested
in
discussion.
So
if
it's
gonna
be
next
week,
I
will
probably
try
to.
A
C
Okay,
so
let's
see
I
will
check
when
is
it
next
week.
B
It's
it's
on
wednesdays,
half
an
hour
before
this
meeting
starts.
C
On
wednesdays,
okay,
let
me
check
yeah
for
me,
it's
a
bit
late,
so
yeah.
That
was
the
main
reason
I
was
not
attending,
because
it
was
very
run,
focus
yeah.
B
B
I've
been
meaning
to
go
to
it,
but
it's
just
been
a
busy
like
it
kind
of
started
during
the
holidays,
and
then
I.
F
C
So,
okay
about
I
have
one
thing
I
I
saw
here
in
the
pr
that
I
have
just
checked
it
very
much
in
yeah.
Is
this
one.
B
Yeah
this
is
this
is
just
leakiness.
I
don't
know
why
that
changed,
but
yeah.
Maybe
I
should
probably
move
that
back
to
5.2.
C
B
C
The
problem
is
that
it
needs
a
really
new
xcode
version
for
that,
so
we
can't
have
users
yeah.
Yes,
yes,
that
was
the
first.
I
saw
yeah.
F
C
F
B
Yeah,
so
if
you
look
at
the
stable
gosh,
what
is
it
stable
meter
provider
there's
some
interesting
stuff
in
and
and
I'm
still
kind
of
like
sorting
it
out?
Maybe
if
you
go
to
it's,
I
think
it's
down
at
the
bottom
stable
meter
provider
and
like
this
stable
metric
reader.
B
C
B
It
replaces
the
metric
processor
that
we
had
okay,
and
so
the
idea
is
that
it's
yeah,
it's
very
similar.
I've
opted
to
use
a
an
array
of
metric
readers
rather
than
a
like
and
so
like
in
the
metric
processor.
B
The
way
that
you
had
it
set
up
is
you
had
one
metric
processor
that
could
be
a
metric
processor
of
processors,
but
instead
I've
just
decided
to
opt
for
an
array
of
metric
metric
readers,
okay,
but
yeah-
I
don't
know,
maybe
maybe
it's
not
the
best
way
to
do
it,
because
then
you're.
C
B
Going
to
have
like
this
no
op
metric
reader
in
there,
and
so
it's
yes,
I'm
open
to
refactoring
that
to
to
the
way
it
should
be
or
the
way
it
was.
But
the
idea
is
so.
B
You
have
readers
which
will,
which
will
read
all
of
the
meters
that
you
might
have
and
there
is
a
possibility
of
multiple
readers,
but
each
reader
only
has
one
exporter,
and
so,
if
you
have
multiple
readers,
they
will
read
every
single
meter
that
you
have
each
one
will
and
I
think
that
might
be
a
little
bit
different
than
the
original.
The
original
design-
maybe
maybe
maybe
not.
B
I
think
that
maybe
I
think,
maybe
in
the
original
design,
the
metric
processors
all
went
to
the
same
exporter,
whereas.
C
Now
yeah,
I
will
see
if
the
exporter
just
went
to
its
metric
and
and
report
reported
that
it
saves
that
value
yeah,
yeah.
C
See
that
it's
its
metric
has
its
own
reader
writer
to
output.
B
No,
actually
so
so
each
so
there
are
many
meters
to
per
reader
or
well
hold
on
hold
on
hold
on.
Let
me
stop
this,
so
you
have
many
meters,
but
each
reader
will
so
from
yeah.
So
so
you
have
a
you,
have
the
meter
provider
and.
D
B
A
meter
provider
can
have
multiple
meters
associated
with
this
yeah.
F
B
B
However,
inside
there
each
the
the
now
a
metric
exporter
is
associated
with
a
metric
reader,
so
they're
they're
a
couple
and
they
they're
there
can
be
multiple
pairs
of
those
and
so
for
every
meter
that
you
have
in
a
meter
provider.
B
B
And
so
they
have
similar,
they
have
they're
very
standard
like
collect
and
you
know
shut
down,
but
the
the
other
interesting
thing
is.
B
I
think
that
the
the
shared
state
is
is
is
more
interest
or
a
little
bit
different
now
too,
because
if
you
add
a
new
meter
that
needs
to
be-
or
rather,
if
you
add,
a
new
reader,
all
of
the
meters
need
to
also
now
send
to
that
reader.
All.
C
Okay,
so
it's
like
a
like,
like
a
bit
like
the
observer
pattern,
where
you
register
your
your
listener
and
you
get
notified.
B
But
then
yeah
then
there's
also
like
push
yeah
it's
it's,
I'm
still
trying
to
sort
it
out,
but
there's
like
push
and
pull
readers
possibly
and
then
there's
another
new
construct,
which
is
quite
interesting,
called
the
gosh.
What
is
it
called?
I
was
just
looking
at
it.
I'm
going
to
go
a
view,
so
a
view
is
allows.
B
B
C
B
C
B
Requires
like
the
way
that
you
set
up
a
view,
is
you
will
tell
the
the
meter
provider
like
hey,
I'm
looking
for
this
type
of
instrument
that
might
have
this
name
and
then
it'll
have
to
search
through
all
the
the
instruments
and
then
pull
out
that
you
know
create
a
linkage
to
that.
That'll
that'll
start
watching
that
that
instrument.
C
Yeah,
I
I
will
I
will
I
will
I
will.
I
will
take
a
look
yeah
to
your
pr.
Do
you
have
the
the
aspect
is
in
the
open.
B
Some
of
it's
pretty
vague,
so
yeah
yeah
yeah.
C
Yeah
I
I
I
see
that
there
are
lots
of
changes
in
many
places
yeah,
but
thanks
for
these
kind
of
things
that
it
will
pay
off
later
for
sure
for
the
users
yeah.
C
B
Yeah
right
now,
it's
just
it's
it's
mostly
just
stubs,
but
I'm
going
to
keep
working
on
it
and
try
to
flush
out
the
the
whole
hierarchy
of
meter
provider
to
reader
to
exporter.
B
I'm
going
to
try
to
sort
that
all
out
and
then
also
implement
all
of
the
all
of
the
measurements
that
we
should
have
like.
The
gauge
asynchronous
gauge,
histogram
up
down,
counts
and
counts,
and
I
think
there's
one
more
that
I
can't
remember,
but
that
that's
kind
of
where
my
focus
is
and
then
we
can
start
trying
to
sort
out
how
how
we
we
add
in,
like
the.
C
C
B
And
I'll
try
to
I'll
try
to
make
sure
that
everything
is
is
extensible
for
that
view
stuff,
if
I
can
but
yeah
the
the
interesting
one
of
the
main
changes
to
all
the
all
of
the
measurements
is
that
we
don't
just
have
a
name
now,
they're
also
the
the
metric
or
the
meter
provider
also
has
a
optional
unit
of
measurement
for
that
meter,
as
well
as
a
description
and
those
are
all
those
are
all
facets
that
can
be
used
for
the
view
to
kind
of
identify,
yeah.
C
B
They're
just
interesting
yeah,
so
so
the
hard
part
is
is
like
like
right
now.
The
way
that
we
have
it
set
up
is
the
the
meter
provider
has
key
value
pairs,
identifying
the
the
meters
and
that's
how
you
like
you
know
you
have
the
name
of
the
meter
and
then
that
that
links
to
the
the
meter
object
in
the
arrays.
B
But
it's
like
it's
like.
How
are
we
gonna
to
to
kind
of
like
you
know
we
might
might
have
weird
bindings
for
or
we
might
need
to
add,
a
new
kind
of
way
of
storing
that
so
that
we
can
get
at
the
name
and
the
unit
and
the
and
the
description
as
well
when
looking
up
those
meters
so
that'll
be
an
interesting
problem
to
solve.
C
C
I
remember
there
was
a
there
was
something
like
that
in
some
framework,
but
I
don't
know
if
that
really
is
used
everywhere,
anywhere
or
yeah
and
yeah
also
yeah.
We
can
also
check
how
some
of
those
things
are
implemented
in
on
apple
switch,
metrics,
maybe
how
they,
I
don't
know
if
they
are
managing
also
units
or
types
of
family
that
maybe
some
ideas
can
come
from
from
there.
B
Okay,
yeah
yeah,
if
you,
if
you
can
find
that
and
share
it
in
the
oh,
the
swift
channel,
that
would
be
cool.
I'd
like
to.
B
C
B
Yeah,
that
would
be
cool
yeah.
I
think
just
one
last
one
last
really
cool
thing
that
I
that
that's
in
the
spec
is
the
the
new
like
reader
exporter
binding,
provides
ways
to
allow
for
different
rates
of
of
collection
of
stuff.
So
that
was
something
that
that
was
really
irking
me
with
having
one
exporter
is
like
it's
only
every
60
60
seconds,
and
I
have
this
cpu
measurement
and
it's
really
hard
to
get
a
good
idea
of
what
the
cpu
usage
is.
B
If
it's
only
every
minute,
because
it's
like
you
know
I'll,
be
clicking
stuff
and
the
cpu
will
be
like
15
percent,
and
then
it
goes
back
down
to
zero
when
I'm
not
doing
anything,
and
then
it
does
the
gauge
at
the
one
minute
mark
and
it's
like.
Oh
cpus
is
just
zero
and
it's
like
well.
If
it
was
going
like
every
ten
seconds,
then
we
might
get
a
better
a
better
average
of
what's
going
on.
But
you
know
I
don't
want
to
change
all
of
the
metric
collections
for
that.
Just
that
one.
C
C
Cool
all
right,
okay,
yeah.
I
will
try
to
find
some
time
tomorrow
to
review
this
and
not
everything
I
can
think
about.
B
Yeah,
hopefully,.
C
C
Okay,
great
all
right,
cool
all
right
have
a
good.
F
C
Yeah
nice
weekend,
for
you
too
see
you
next
week,
yeah.
C
E
And
I
I'm
also
getting
some
symptoms
like
I.
C
C
What's
that,
I
I
mean
that
I
hope
you
you
are
vaccinated,
so
it
will
yeah.
E
Yeah
yeah,
I
also
got
a
booster,
so
yeah,
so
I
have
yeah
great.
Then
it
should
be
really
really
easy
to
to
ride
it,
yeah
yeah,
so
yeah.
So
I
I
see
that
there
was
another
member
joined
today
from
from
yeah.
E
E
C
C
From
the
client
side,
part
asking
about
our
opinions
about
how
to
how
to
model
some
things
in
the
like,
the
events
that
happen
in
the
app
where
you
don't
have
a
spam
and
that
can't
have
relations
of
parent's
heart.
He
wanted
to
know
our
opinion
because
we
didn't
go
to
the
client-side
meeting.
E
C
I
said
that
maybe
bryce
said
he's
gonna
probably
go
next
week.
I
will
try
also
and
check
it
yeah.
My
opinion
is
that
we
need
another
structure,
different
that
tricking,
zero
duration
expands
for
that
is
overkill
and
that
using
logs
for
that
is
under
kill.
So
what
what?
What
is
that
again
say
that
again
yeah
yeah?
They
they?
They
said
that
they
were
planning
using
zero
duration
response
for
representing
that.
That
is
not
what.
C
That
serialized!
No,
I
didn't
question
for
my
english:
zero
duration,
expanse
serial
durations,
zero,
zero,
zero
expanse,
no
expanse
without
duration
with
duration
zero,
but
that
that
that's
a
trick
without.
C
C
E
C
E
This
is,
I
think,
bryce
bryce
had
this
problem
before
right.
Remember.
He
was
talking
about
this
difference
in
the
you
know,
ntp
issues
and
you
know
because
the
client.
C
C
E
C
F
C
Bit
artificial,
at
least
for
my
taste,
it's
a
bit
too
artificial
using
that
for
things
that
are
not
really.
C
Yeah,
that's
right
that
that's
what
they
wanted
to
do
to
use
expanse
and
setting
duration
to
zero,
but
that's
not
a
span!
That's
another
thing
that,
yes,
sir.
C
E
Like
what's
on
the
header,
that
is
going
big
part
of
the
context,
but
I
mean.
F
F
C
E
E
E
C
E
Yeah,
hopefully
I'll
get
more
time
to
work
on
open
telemetry
after
okay.
C
E
C
Sorry,
that's
what
I
I
I
know
there
is
a
complaint
with
that
name
and
I
thought.
E
The
affiliation,
I
think
they
have
like
a
they-
have
a
repo
with
that
affiliation.
So,
but
you
know
hopefully
I
am
going
to
be
able
to
spend
more
time
on
open
telemetry
sources.
E
C
E
Platform,
yeah,
you
know,
platform
is
snowflake
and
then
you
have
all
the
you
know,
observability
stuff.
On
top
of
that,
so
we
have
a
open,
telemetry,
endpoint,
okay,
and
I
think
we
know
we
are
still
you
know
like
kubernetes.
We
are
more
the
kubernetes
and
you
know
the
cloud
and
the
container
side
right.
So
I
think
this
will
be
kind
of
a
you
know:
green
field
right
for
open
telemetry.
So
I
think
this
will
yeah.
Hopefully
I
get
to
work
more
on
open,
telemetry
and
great.
C
E
O'clock
call
so
I
you
know,
if
I
finish
early,
then
I'll
come
and
join
you
guys
11
to
11
30.
yeah.
So
today
it
went
down
for
a
while,
so
yeah,
but
I
I've
been
reading.
You
know
the
the
comments
on
the
on
the
repo
and
in
the
dark
and
so
once
yeah,
so
all
the
one
yeah.
So
I
think
one
of
the
things
I
I
noticed
was
the
open,
telemetry
documentation
has
gotten
much
better
and
we
need
to
add
stuff
on
the
swift.
E
C
Okay,
okay,
yeah,
probably
I
have
that
task
assigned
to
me
a
long
time
ago,
but
you
know
I
am
a
coder.
So
documenting
is
something
that
you
always
do.
The
last
yeah.
C
E
E
Be
a
good
way
to
you
know.
So,
let's,
let's
pick
up
the
thread,
because
I
can
do
this
offline
too
right.
Whenever
I
get
a
chance,
so,
okay,
we
can
yeah.
We
can
yeah,
because
I
think
I
looked
at
the.
I
looked
at
all
the
languages
and
you
know
they
have
some
very
detailed
instructions
for
java
and.
C
E
C
E
E
E
C
Hope
it
just
takes
one
or
two
days,
medium
symptoms
and
just
go
away.
Yeah,
because.
E
E
C
That's
right,
it's
crazy.
I
have
seen
many
many
people
yeah
my
daughter
at
the
school.
Several
of
these
fellows
are
ill
or
have
been
ill.
We
are,
you
know,
I
think,
saving
from
that,
or
at
least
if
the
test
or
the
didn't
say
positive,
any
of
us
so
yeah
we
are,
we
are
escaping,
but
yeah,
probably
some
some
of
these
days.
We
will
probably
pick
it
up
also
because
it's
crazy,
the
numbers
are
totally
out
of
the
graph
and.
E
C
E
C
Yeah
yeah:
we
cannot.
We
can
some
days
yeah,
okay,
so
I
I
really
hope
it's
did
you
pass
it
really
soft
and
that
yeah
everything
just
pass
away
early
and
yeah
and
you
will
have
better
defenses
for
sure
all
right,
good.