►
From YouTube: 2021-08-12 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
It's
becoming
that
way,
unfortunately,
but
yeah
my
house
is
a
little
older,
so
get
some
portable
acs.
So
I'm
not.
A
B
Dying
because
I
probably
would,
if
I
didn't,
have
them
but
yeah,
it's
it's
yeah.
New
construction
is
becoming
more
common
to
see
acs.
A
A
How's
the
east
coast,
going
on
these
days.
It's
warm
it's
it's
like
92,
where
I
am
right
now,
not
that
bad.
B
Yeah,
I
remember
my
days
in
new
england.
92
is
like
it's
okay,
it's
hot,
but
it's
the
humidity.
That
would
always
kill
me
yeah,
and
that
was
just
I
couldn't
stand.
It.
A
Yeah,
it's
like
a
light
freeze
helps,
but
I
was
out
walking
for
about
20
minutes
earlier
and
came
back
and
pretty
much
take
my
shirt
off.
B
David,
are
you
in,
I
want
to
say
you're
in
you're
in
like
tahoe
at
some
point,
but
I
feel
like
maybe
you're,
a
bay
area
person.
C
B
B
A
I
was,
I
was
reading
code
today
and
I
discovered
your
name
all
over
it.
I
was
reading
the
open
census
code
for
trace
propagation.
A
Like
the
bridge
there's
like
a
grpc,
you
know
something
like
an
oc
grpc
package
that
extracts
trace
headers
out
of
grpc
metadata,
and
I
saw
your
name
in
there.
There
is.
A
I
have
no
recollection
of
this,
it's
possible.
I
did
something
we
we
have
a
program.
We
didn't
write
that
is
instrumented
with
open
census,
we're
trying
to
figure
out
how
to
get
traces
to
propagate
into
it
and
determine
that
it's
not
going
to
be
easy,
if
at
all
possible,
but
I
could
tell
you:
we've
been
there.
B
Proof
yeah
I've
heard
dave
was
pretty
prolific.
Maybe
it's
just
his
alter
ego.
There's
actually
three
of
me
yeah.
Sometimes
I
feel
that's
true
for
bogdan,
for
how
often
he's
committing
it's
just
incredible
cool.
Well,
it
looks
like
we
are
definitely
well
into
the
hour.
So
let's
get
started
here.
If
you
haven't
already,
please
add
your
name
to
the
attendees
list
and
if
you
have
any
agenda
items
that
you
want
to
actually
talk
about.
B
Please
add
them
here
as
well
and
we'll
jump
in
as
usual
to
looking
to
our
active
project
board
that
we're
going
on
right
now
in
the
past
week.
There's
definitely
that's
not
right!
31
yeah!
That's
not
how
you
do
math,
so
that
definitely
got
quite
a
lot
done.
I
think
in
the
past
week
I'm
still
making
some
some
good
progress
for
it,
so
excited
about
that
definitely
excited
to
get
a
stable
release
actually
released.
So
that's
always
making
me
happy.
B
Let's
jump
in,
I
think,
to
what
is
actively
open
and
being
worked
on
in
these
three
categories.
There
is
a
new
pr
related
to
this
issue
that
we
can
kind
of
jump
into.
I
opened
up
yesterday.
This
is
a
follow-on
to
the
removal
of
the
arrays
from
the
attributes,
the
array
type
being
replacing
those
with
the
slices
themselves
which
just
merged.
B
Today
I
was
looking
at
the
code
while
I
was
writing
this
and
there's
this
any
function,
which
I
think
we
added
in
kind
of
the
same
vein
where
you
can
just
provide
a
very
generic
duct
type
interface
type
and
we'll
figure
it
out
we'll
kind
of
like
handle
it
for
you.
B
The
problem
with
that
is
very
similar
to
what
we
just
got
rid
of
with
the
array,
and
I
think
it
actually
suffers
from
the
same
problem
as
eventually
as
if
you
pass
an
array
of
the
wrong
types,
then
this
could
actually
result
in
missing
data.
But
on
top
of
that
we
also
do
conversions
in
this
type
that
are
one-off
and
not
universal,
to
open
telemetry.
B
So
it
is
definitely
not
going
to
be
cross-compatible
across
languages
and
there's
definitely
some
one-off
choices
that
could
essentially
bite
people
when
open
telemetry
changes
or
even
users,
implementations
change.
B
One
thing
that
I
you
know
I
was
kind
of
thinking
like
if
you
pass
this
and
the
implementation
is
looking
to
see
if
it's
marshalling
json,
if
it
implements
the
stringer
method,
if
the
underlying
type
or
the
concrete
type
changes
when
you're
using
this
s
print,
all
of
those
things
are
going
to
change
your
value,
which
are
like
really
unexpected
from
the
user's
perspective.
So
the
goal
here
is
to
remove
this
and
defecate
it.
So
there's
an
open
pr
for
that.
B
If
you're
interested
it's
in
in
the
issue
itself,
please
take
a
look
or
comment.
I'd
love
to
have
some
feedback.
If
you
have
other
ideas
on
where
to
go
with
this
one,
that
being
said,
jumping
back
into
the
list
here
we
have
this
right
here.
I
think
there
was
some
movement
on
this
sorry.
This
is
the
actual
issue,
and
then
this
is
the
pr
associated
with
it.
I
think
robert
was
jumping
in
on
this
one
and
he's
doing
a
really
good
job,
providing
some
feedback.
B
I
think
it
might
need
some
more
reviews,
which
is
what
I
might
have
said
last
week,
but
yeah.
I
think
some
more
eyes
on
this
most.
I
think
me
and
robert
try
to
provide
some
sort
of
like
conclusion
as
to
what
our
assumption
is
but
yeah.
I
think
that's
just
waiting
on
reviews
and
aside
from
that,
the
only
thing
else
is
in
active
work.
Is
this
deprecation
of
the
hotel
test
package
I'm
still
working
on
this?
It
is
essentially
could
be
deprecated
in
the
main
repo.
B
Currently,
however,
it's
used
very
extensively
in
a
very
copy-pasted
pattern
throughout
the
contrib
repo.
So
when
we
go
and
upgrade
and
we
actually
do
another
release,
this
is
going
to
be
a
hassle.
I
think
I'll
say
if
we
tried
to
release
this
and
have
somebody
do
all
of
these
changes
in
one
pr.
So
I'm
trying
to
break
this
apart.
B
I've
got
the
pr
for
the
hotel
grpc
right
now,
I'm
working
on
one
for
the
hotel
http,
the
rest
of
like
these
are
all
pretty
copy
pasted,
so
I
don't
imagine
them
being
too
complex,
I'm
starting
with
the
more
complex
ones,
the
grpc
and
the
http,
but
making
some
progress
and
making
some
headway
on
this.
It's
also
really
positive.
It's
giving
us
some
really
good,
I
think
understanding
of
what
we
need
to
test
and
it's
padding
out
the
trace
test
package
with
those
constructs.
So
I
think
it's
an
effective
exercise.
B
Yeah,
I
would
love
some
help.
I
was
actually
just
thinking
about
that
this
morning.
I
could
definitely
break
this
list
apart
into
an
actionable
list
with
checklist
into
this
issue.
I
just
didn't
know
anybody
else
was
interested,
but
yeah
evan.
If
you're
putting
your
hat
into
the
pile,
I
will.
D
B
Yeah,
if
you,
if
you
see
some
particular
ones
in
here,
that
you
want
to
grab,
please
just
comment
in
this
issue.
Otherwise
I
will
try,
after
this
meeting,
to
get
like
a
checklist,
and
maybe
we
can.
I
don't
want
to
like
go
full
on
just
make
issues
for
every
single
one
of
these,
because
that.
B
Of
bloat,
but
I'm
happy
to
coordinate
and
share
this
issue
with
you.
If
you
want
to
jump
in
here,
yep.
B
Cool
awesome,
yeah
and
anybody
else
on
the
call
I
would
love
to
help.
So
it's
much
much
appreciated
and
then
I
think
that's
it
for
all
of
the
active
open
issues
right
now
that
we
have
going-
and
I
think
the
only
other
things
are
things
that
you've
actually
already
talked
about.
I
know
I
would
like
to
maybe
talk
a
little
bit
about
this,
but
also
I
wanted
to
ask
anthony
that
you're
back
from
vacation,
if
there's
any
movement
on
the
next
round
of
amazon
interns.
B
E
Yeah,
so
I
think
that
the
it
wouldn't
be
the
next
round
of
interns.
It
was
hope
that
the
interns
who
are
wrapping
up
this
weekend
next
would
have
some
bandwidth.
I
think
that
that
bandwidth
ended
up
being
directed
at
the
collector.
E
B
Awesome
thanks
for
the
heads
up
yeah,
that
sounds
fine.
I
I
kind
of
expected
that
as
well
there's
a
lot
of
stuff
to
do
in
opencl
in
general,
so
no
worries
we'll
just
add
into
the
the
burn
down
list.
I
think
we're
making
some
great
progress
so,
let's
just
get
into
them
cool
the
only
other
one
was
this.
B
I
think
I
it's
kind
of
been
sitting
in
the
back
of
my
mind
there
just
to
remind
people
is
it's
a
rename
of
a
package
rename
and
potentially
this
could
impact
a
few
other
packages,
but
the
idea
here
is
that
the
sdk
trace
package
would
be
renamed
to
sdk
trace,
and
I
think
this
has
been
submitted
as
an
idea
a
few
times.
I
think
we
just
need
a
vote
on
this
and
I
don't
know
the
best
form.
I
was
thinking,
maybe
something
in
slack.
We
could
get
like
a.
B
I
don't,
I
think,
there's
like
a
polling
mechanism
there.
If
you
haven't
figured
out,
I'm
not
so
great
at
technology
sometimes
but
yeah
like
I
feel
like
we
just
need
a
vote
on
this
is
if
we
want
to
do
this
or
not-
and
I
would
probably
just
go
to
the
default
like
it's
pretty
late
in
the
game
to
be
renaming,
it
kind
of
makes
sense.
It
helps
you
know,
there's
some
real
use
case.
That's
given
here,
so
maybe
we
could
have
some
people
read
over
it.
I
don't
know.
B
B
It
has
the
potential
to
have
some
some
weird
effects,
but
yeah
it's
a
really
small
change,
but
the
impact,
I
think,
is
going
to
be
more
impactful
than
than
just
the
level
of
effort
so
yeah.
I
would
love
to
to
see
people's
opinions
on
this
one.
Maybe
if
I
don't
see
any
movement
in
the
next
week,
I
could
try
to
post
something
in
slack
but
yeah.
I
don't.
I
think
that
sounds
like
something
we
should
do
here.
B
Cool
with
that,
I
think
that's
the
enough
of
a
review
of
the
project
board.
It's
looking
pretty
healthy.
We're
definitely,
like
I
said,
seeing
some
movement
from
left
to
right,
so
everybody
else
is
something
else.
I
wanted
to
talk
about
on
here.
I'll
just
pause
for
a
second.
E
I
don't
think
it's
on
here,
but
related
to
the
interns
wrapping
up
their
their
terms.
Garrett
has
an
outstanding
pr
in
contrib,
actually,
two
of
them
for
lambda
instrumentation.
E
That
could
use
some
additional
reviews,
so
the
first
one
needs
to
land
before
the
second
one
can
really
be
effectively
reviewed
because
they
include
the
same
code
and
github
doesn't
provide
a
great
ui
for
saying
this.
One
is
built
on
top
of
the
other
one,
but
if
we
could
get
some
more
reviews
on
this
one,
I
think
several
people
from
aws
have
reviewed
it,
but
we
need
some
additional
reviews.
B
Okay,
yeah:
I
will
copy
this
from
last
time.
B
E
I
think
that's
all
I've
got
for
now.
Okay,.
B
Awesome,
the
agenda's
pretty
light
on
a
meta
issue.
Here
I
was
hearing
from
other
cigs
that
the
82
pages
that
we
have
in
our
agenda
docs
can
sometimes
cause.
I
guess
82
is
not
that
big,
but
some
other
sigs
have
more
notes
and
they're
causing
some.
You
know
google
doc
lag
when
they
try
to
open
it
and
what
I've
seen
and
specifically
in
java
was
they
just
went
back.
B
I
think
three
months
or
something
like
that
or
two
months
and
anything
older
than
that
they
just
put
in
an
archive
dock
and
linked
it
in
their.
You
know
top
header
or
something
right
here,
which
I
don't
know
if
anyone
cares
or
has
opposition
to
that.
But
I
was
thinking
of
doing
something
similar
just
to
kind
of
reduce
the
the
load
on
on
this.
I've
seen
a
thumbs
up
from
evan
cool
yeah,
pretty
simple
one.
B
E
Do
it
perfect
hasn't
been
an
issue
for
me,
but
I'm
fine
with
that.
I
think
there's
probably
not
a
whole
lot
of
value
in
keeping
really
old
stuff
around,
at
least
not
in
the
same
dock.
As
long
as
we've
got
a
link
to
it
and
people
can
still
access
it
when
they
need
it.
B
Yeah,
that's
kind
of
my
thing.
The
specification
one
is
the
one
that's
killing
me
right
now,
but
yeah.
I
agree
like
having
things
from
two
years
ago
is
not
really
that
important.
I
will
have
to
find
out
how
to
maybe
I
have
to
own
a
dock.
Okay,
I'll
figure
this
one
out,
but
that'll
be
interesting
cool
other
than
that
I
think
somebody's
typing.
I
don't
know
if
this
is
garrett,
but.
F
Oh
yeah,
that's
me
my
stuff
in
the
next
like
day
or
so
is
going
to
be
migrating
over
to
blocktake.
Somehow
we
haven't
figured
out
exactly
how
that's
going
to
be
structured,
but
yeah.
B
Okay,
cool
I've
seen
josh
type
as
well,
so
I'll
just
pause
and
we
can
hand
it
over
to
each
other.
G
Thank
you
for
merging
one
of
the
pr's.
I
had
up
I'll
keep
trying
to
review
prs
to
help
out
and
in
the
main
repository,
and
I
have
one
that
I
need
to
merge,
do
merge
conflicts
for
which
I
think
from
my
opinion,
would
be
next
and
all
that
stuff.
We
talked
about
last
week
and
I
will
start
producing
more
of
them
to
finish
this
movement
of
the
sdk
api
and
the
the
next
one
that
I'm
aware
of
the
way.
I
will
do
it
until
I
learn
otherwise.
G
Is
that
the
metric
descriptor
object
currently
has
the
instrumentation
library
name
and
version
in
it,
and
I
want
to
pull
that
out
to
separate
the
descriptor
from
the
instrumentation
library
stuff,
because
that's
that
entanglement,
that's
causing
other
trouble
and
so
instrument
type
instrument
type
had
to
go
first,
which
is
why
the
pr
that
you
just
merged
went
in
and
next
will
be
to
move
the
instrument.
G
B
It's
pretty
great,
it
can
be
our
mascot.
I
like
it.
My
daughter,
I
think
that's
even
cuter,
now
cool.
Is
there
anybody
else
that
has
some
pr's?
I
wanted
to
talk
about
before
we
jump
into
the
next
phase.
B
Awesome
cool
I'll,
stop,
sharing
I'd
love
if
anybody
has
any
user
stories,
we're
definitely
pretty
early
on
in
the
meeting,
but
I
also
don't
want
to
keep
people,
but
this
is
one
of
my
favorite
topics
is:
if
anybody
has
some
successes,
pain
points
for
using
open
source,
both
the
contrib
and
the
main
repo.
B
E
Not
so
much
a
user
story,
but
there
was
a
twitter
thread
that
I
was
reading
the
other
day
where
someone
was
talking
about
having
tried
several
open,
telemetry
sdks
and
they
found
that
go
was
the
best
of
them,
even
though
our
documentation
was
slightly
lacking,
which
I'll
take.
We
know
our
documentation
can
be
improved.
But
it's
some
good
feedback
to
hear.
B
Yeah,
that's
great.
I
love
all
that
kind
of
feedback
because
I
agree
we
should
improve
our
docs
yeah
cool
yeah.
If
you
have
a
link
to
that
twitter
thread,
I'd
love
to
read
it,
but
yeah.
G
Much
like
last
week,
I
have
a
story
that
I
can
keep
short.
If
you
like
it's
another
prototype
that
I
used
to
go
sdk
for
last
week
I
showed
I
had
prototyped
the
probability
sampling
that
was
in
pr
form
number
2177,
which
is
in
a
draft
now,
just
because
I
pointed
out
at
it
enough
times,
and
I
put
up
one
yesterday
named
2180,
which
is
a
rate
limited
sampler
prototype.
G
This
was
in
response
to
a
spec
issue,
asking
for
us
to
spec
out
a
standard,
leaky
bucket
type
rate
limiter,
and
I
we've
been
light.
Step
wants
to
make
probability
sampling
happen,
so
I
demonstrated
how
to
do
probability
sampling.
It
turned
out
really
nice.
I
don't
think
I
should
show
it
to
you
the
way
I
did
last
week,
but
basically
it
boils
down
to
I'm
going
to
say
50
lines
of
code.
It
does
some
tricks
to
put
to
pull
apart
an
ieee
floating
point
number,
which
is
like
the
first
time.
G
I've
ever
needed
math
float64
from
bits
in
my
life,
which
is
like
the
first
real
cool
usage
for
that
for
me,
and
then
it
basically
has
to
do
a
little
linear,
linear
interpolation,
between
two
powers
of
two
and
then
it
flips
a
coin
to
decide
whether
it's
gonna
use
the
low
probability
or
the
high
probability,
and
then
it
calls
into
the
trace
id
ratio
based
sampler
for
corresponding
bracketed.
G
You
find
which
powers
of
two
you're
between
and
then
you
just
keep
adjusting
the
rate
to
target
your
outcome,
and
if
the
distribution
of
rate
or
the
rate
of
arrival
is
stable,
you'll
eventually
get
it.
And
if
it's
not
it'll
you'll
track
it
pretty
well,
and
it's
just
a
proof
of
concept,
you
can
improve
the
logic
to
estimate
using
all
kinds
of
tricks,
exponential,
smoothing
you
can
model
it.
G
You
can
do
all
kinds
of
stuff,
but
but
the
basic
idea
is
you
can
keep
adjusting
your
probability
and
you
can
do
that
accomplish
it
accomplish
that
by
a
probability,
choice
and
a
and
two
powers
of
two.
So
anyway,
that's
in
2180,
if
you're
interested-
and
we
talked
about
it
this
morning
in
the
sampling,
sig.
B
That's
really
cool,
so
that
did
you
do
a
little
demo
in
the
sampling.
Stick.
G
Okay,
I'll
show
you
what
I
when
I
share.
I
have
you'll
see
what
I'm
in
the
middle
of
here
I
am
merging
to
emerge.
I
need
to
check
out.
I've
got
a
million
branches,
it's
this
one
and
then
let's
just
stash
that
oh
jeez,
I'm
I'm
merging.
I
can't
do
this.
G
G
G
Actually,
don't
think
I
recorded
it,
but
I
I'm
not
gonna,
I'm
not
gonna
challenge
my
get
state
right
now,
anyway.
There's
a
program.
Let
me
just
open
the
file.
G
I
like
well,
I'm
going
to
have
to
show
you
and
I'll
have
to
unshare
this
window,
because
I
can't
show
it
to
you
in
emacs,
without
fixing
my
git
state.
Okay,
let
me
show
you
the
actual
pr
and
I'll
just
quickly
show
it
to
you.
G
G
Yuri
is
asking
for
me
to
spell
it
out
in
words,
which
of
course
I
should
have
done,
but
I
was
late
last
night,
here's
this
pulling
apart
floating
point
number
to
get
its
powers
of
two
and
then
linear
interpolation
between
the
others,
and
this
is
constructing
the
two
probability
samplers
above
and
below
the
target
probability,
and
then
here's
my
test
that
I
want
to
show
you.
G
G
The
assumption
at
the
start
is
that
you're
going
to
sample
one
to
one,
so
you
start
out
without
any
any
probability,
sampling
and
so
the
first
period
you
always
overdo
it
in
this
experiment,
but
the
the
whole
point
of
that
probability,
sampling
work
is
just
is
to
say
that
there's
something
we
can
add
up
at
the
end
for
the
spans
that
we
received
to
estimate
the
total
number
of
them,
and
that's
this
so
I
walk
through
the
spans
of
my
test
exporter.
G
I
set
a
default
count
of
one
since
that's
the
natural
count
and
then
I
look
for
this
sample
adjusted
count.
This
is
what
the
otep
says.
If,
if
there's
a
number
there,
I
expect
it
to
be
an
integer,
because
that's
the
way
the
sample
was
written
and
then
at
the
end
I
I
asked.
I
expect
that
my
total
of
adjusted
accounts
is
close
to
my
actual
number
of
spans
that
were
in
the
population,
and
it
works
out
with
this
particular
test.
G
With
my
arbitrary
numbers
that
I
put
into
it
that
you
get
a
six
percent
error,
so
I
wrote
the
test
there.
You
also
can
expect
a
reasonably
close
rate
to
your
target
and
if,
if
you're,
if
the
rate,
the
actual
rate
constantly
overestimate
or
is
higher
than
your
estimate
and
you're
going
to
get
slightly
over
sampled
and
if
you're
constantly
falling
below
the
expectation
you'll
slightly
undersample,
but
over
time,
it
averages
out
anyway.
That's
the
current
state
of
this.
G
It's
basically
to
show
that
we
can
do
approximately
rate
limited
sampling
in
a
sampler,
and
there
was
some
discussion
in
the
in
the
meeting
that
that
you
could
go
check
out.
That
was
a
lot
more
in-depth
about
how,
if
you
want
to
hard,
if
you
do
want
to
do
a
hard
rate
limit,
you
need
to
do
something
in
the
exporter
as
well
as
in
the
sampler
and
and
possibly
you
want
to
share
state
between
the
two,
and
maybe
you
don't
need
to
so.
There's
there's
some
stuff
in
the
meeting
about
that.
G
B
Yeah
definitely
appreciate
it
definitely
cool.
I
think
anybody
else
have
something
they
wanted
to
show
for
the
user
stuff.
Otherwise
you
can
probably
just
end
this
pretty
early.
Surprisingly
cool,
all
right,
I'm
seeing
nothing
from
steve.
I
think
everyone
will
see
all
virtually
or
next
week,
bye.