►
From YouTube: 2021-06-18 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
A
Hello,
did
you
all
get
emails
from
hbo
max.
A
Did
our
info
leak
did
our
private
info
leak?
No,
they
sent
an
integration
test,
email
to
all
customers.
B
No,
I
honorag,
and
I
had
this
discussion
previously.
They
don't
get
hbo
max
in
japan
for
some
reason.
A
B
B
B
B
Last
week
was
two
we're
on
a
we're
on
a
two-week
roll.
Like
last
week
we
had
the
the
app
dynamics-
folks
yeah
interesting-
they
didn't
show
up
this
week,
but
yes,.
B
That
I
don't
remember
their
screen
the
screen
names,
but
yes,
they
did
not
sign
in
at
least.
A
A
C
A
So
let
me
let
me
I
guess
I
could
share
my
screen
and
show
you
what
I
mean
give
me
just
a
moment.
B
So
you
know
what
we
have,
how
we
set
that
up
in
the
instrumentation
repo
is
we
have
this
one
accept
pr
job
that
depends
on
all
the
previous
ones,
and
then
we
just
put
the
the
branch
protection.
We
use
that
one
except
pr.
B
A
A
A
There's
no
tls
issues
here.
This
is
just
plain,
jane
http,
so
I
don't
know
if
that
helps,
but
it's
definitely
not
a
and
I
tried
switching
between
localhost
and
127.001
that
didn't
make
any
difference.
I'm
not
sure
what
else
I
mean
there.
Maybe
there's
a
way
we
can
get
access
to
the
actual
hostname
hostname,
and
that
would
work.
A
This
particular
test
has
localhost
hard
coded
in
it
at
the
moment.
But
that's
the
only
thing
I
can
imagine
is
that
something
has
changed
with
the
networking
and
what
is
considered
localhost
somehow.
But
I
don't
really.
I
looked
through
the
the
changelog
on
the
ubuntu
on
the
github
ubuntu
image,
and
I
couldn't
do
anything
that
looked
interesting
or
useful.
B
But
everything
good
and
now
this
everything's
passing
again
yeah
everything
works
on
1804.
I
don't
know
it
doesn't
make
any
sense.
A
To
me,
I
was
hoping
that
github
would
let
us
run
against
the
previous
image
of
ubuntu
2004,
but
there's
they
don't.
I
even
tried
just
hacking
some
stuff
in,
but
there's
they
don't
have
any
support
for
running
on
anything
of
whatever
their
default
is
whatever
ends
up
in
your
in
your
run
time
right
bring
your
own
container.
A
B
Let's
see
so,
we
had
the
outreachy
intern,
joined
and
shared
on
her
project
and
we're
recommended
to
open
the
discussion
item
and
that
folks
would
review
it.
It's
basically
jmh
tests
of
specific
instrumentation
library,
instrumentations
and
open
tracing,
so
example
servlet
or
here's
the
jdbc
one
that
we're
just
also
chatting
about,
and
so
it's
just
it
jam
makes
benchmarks
of
these
instrumentations.
And
so
the
basic
idea
of
the
project
is
to
create
those
also
for
open,
telemetry
and
then
compare.
A
We
have,
we
have
separate
jmh
modules
inside
each
or
not
modules
source
paths
inside
each
module
that
has
jmh
tests.
You
have
an
issue
to
combine.
A
If
there's
a
reason
to
do
that,
so
the
that
issue
in
particular
was,
I
think,
to
make
it
so
we
don't
have
to
copy-paste
the
batch
spam
processor
benchmark
into
every
single
implementation
of
the
batch
of
that
idea.
So
it
wasn't
necessarily
to
combine
every
single
jmh.
B
A
Have
like
at
least
one
common
set
that
could
run
against
a
bunch
of
different
versions
of
the
same.
B
C
B
Yeah
yeah,
I
think
this
project,
though,
is
or
at
least
the
open
tracing
ones,
are
all
they're,
not
java
agent.
C
A
B
A
B
Was
the
they
didn't
have
the
java
agent
until
later.
C
A
Yeah,
that's
that's
true.
Definitely
worth
commenting
on
that
discussion
item,
though.
B
Yeah,
what
will
be
interesting
is
if
we
I'm
not
sure
whether,
like
how
much
is
in
some
of
those
cases,
whether
it's
reusable
inside
of
the
java
agent
is
a
different
question,
but
certainly
adding
those
library
instrumentations
would
be
good.
B
Api
and
then
lots
about
metrics
that
I'm
gonna
john,
do
you
want
to
pause,
eating
and
talk
about,
or
should
I
stumble
badly.
A
Well
so
josh
earth
is
taking
on.
You
know
this
task.
We
talked
about
this
that
he's
going
to
try
to
do
the
new
version
of
metrics,
and
so
the
big
question
is,
as
we've
talked
about
before:
do
you
take
the
existing
thing
and
break
everything,
or
do
you
create
a
new
package,
and
I
think
that
I
think
what
came
out
of
the
discussion
today
at
least
my
opinion
was
we
want
to.
A
A
You
know
it's
contrary
to
what
we
had
talked
about
before
on
rug,
that,
like
just
do
it
in
place
and
break
everything,
but
I
think,
given
that
I
really
would
like,
like
we
had
some
quite
good
good
discussion
today
about.
Can
we
just
make
counters
be
longs
and
not
have
a
double
counter
like
nobody
cares?
Nobody's
counting
doubles
and
can
we
make
everything
else
just
be
doubles
because
nobody
wants
to.
I
mean
it
just
works
for
everything
else
like
micrometer
does.
A
So
if
we
want
to
do
something
like
that,
that's
a
fairly
significant
change
to
the
api
structure
and
josh
also
would
like
to
add
optional
parameters
to
pass
the
context
in
explicitly
for
every
metric
reporting
for
for
exemplars
and
such
like,
and
that's
also
a
fairly
significant
change.
So
it
kind
of
feels
like
it
might.
We
should
really.
I
would
prefer
that
we
take
this
time
to
really
rethink
and
build
an
api
that
isn't
just
open
census
times.
A
It
is
going
to
be
there
there's
the
big
pain,
though
do
we,
what
do
we
name
the
module
and
what
do
we?
How
do
we
package
the?
How
do
we
package
the
code?
Josh
was
saying
I
don't
know
what
ide
ide
he
uses,
but
it
was
very,
very
unhappy
with
having
two
modules
with
the
same
packages
that
it
got
very
confused.
A
Yeah,
well,
it's
a
bad
idea
for
a
bunch
of
reasons
yeah.
So
then
what
do
we
do?
Do
we
do
we
create
a
new
package
and
then,
at
the
end,
rename
the
old.
The
new
package
back
to
the
old
package,
which
is
the
old
package
name,
is
the
right
package
name.
So
do
we
do
having
a
moment
where
we
delete
the
old
one
and
rename
all
the
packages?
I
don't
know,
and
the
module
name
like?
What
do
we
publish
the
module
name
as
like
there's
a
whole
bunch
of
weird
questions?
There.
C
A
A
A
Yeah
yeah
anyway
yeah.
I
guess
the
other
thing
is
we
don't
have
to
actually
publish
the
new
one
at
all
yep
until
it's
ready
like
we
just
know,
we
just
don't,
publish
yeah,
leave
the
publishing
out
until
it's
done
yep
or
at
least
until
it's
ready
to
appoint,
to
replace
the
old
one
that
maybe
that's
what
we
do
is
we
just
say:
let's
hold
off,
let's
not
publish
anything
even
at
dash
alpha
and
then
when
we're.
We
think
what
we
have
is
ready.
A
A
I
I
agree
with
also
and
also
maven
central
is
such
a
mess
with
all
of
the
different.
You
know,
historical
module
names
that
we've
gone
through
and
and
the
absolutely
terrible
search
function
on
any
of
the
maven
central
interfaces.
It
doesn't
even
repository,
I
think,
works.
Okay,
do
you
use.
C
A
C
A
The
question
is
like
what
I
really
want
to
see
is:
I
want
to
see
everything
sorted
by
recency.
Does
this
do
that
sorting
by.
A
A
C
I
think
I
tried
using
search
maven
recently
because
I
didn't
even
know
about
it
like
I
thought
this
was
official
for
the
longest
time.
I
thought
this
is
maven
central
search
and
sounds
like.
Oh,
this
is
just
some
third-party
thing,
so
I
tried
the
official
one,
but
now
I
understand
why
there's
a
third
party
thing
yeah
this
one
takes.
A
D
B
B
C
A
Well,
it
could
be,
but
I
mean
I
think,
that
josh
said
there
were
legitimate
usages
that
he
knew
of
for
long
for
64-bit
counters.
C
A
A
B
A
C
Yes,
it
was
the
proto,
and
so
now,
instead
of
getting
as
I
get
as
a
double
or
get
as
long
depending
on
what
it
is
and
it's
hard
to
even
know
which
one
it
is
and
that
was
actually
insanely
annoying.
So
it
would
be
nice
if
each
type
only
like,
I
don't
think
combining
the
types
improved
it.
I
think
it
made
it
worse.
A
C
C
A
A
A
A
This
whole
question
that
whether
buckets
histogram
buckets
should
be
definable
at
the
api
layer,
and
this
was
about
this
was
most
actually
ended
up
being
most
of
the
time
that
we
spent
in
the
the
metrics
sig
today
was
discussing
this
so
no
easy
answer,
so
the
I
think
that
the
crux
of
the
matter
was
josh
was
feeling
that
instrumentation
authors
have
the
best
knowledge
about
kind
of
what
relevant,
histogram
buckets
should
be,
and
my
argument
was
what
like
no
like.
A
I
think
all
that,
like
if
you're
instrumenting,
http
client,
all
you
know
is
that
it
has
a
lower
bound
of
zero
as
its
time
like.
Never.
Besides
from
that,
it
completely
depends
on
what's
being
called
and
how
it's
used
and
yeah.
So
anyway
went
back
and
forth
it
was.
It
was
probably
45
minutes
of
the
metric
sig
today.
C
D
A
I
think
that
most
of
us
have
been
trying
to
push.
This
is
an
sdk
configuration
concern,
but
the
but
the
real.
The
real
issue
is:
what
do
we?
What
do
you
use
as
the
default
histogram
configuration
like
if
someone
doesn't
specify
it?
What
do
you
do
because
there's
not
really
a
good
default
prometheus
has
decided
that
the
default
is
something
that
kind
of
makes
sense
for
http
requests
on
the
server
side,
which
I
think
josh
was.
Finally
con.
Josh
earth
was
finally
convinced
this
might
be
a
decent,
a
reasonable
default.
If
it's
not
configured.
C
A
A
Is
this
is
kind
of
the
other
side
of
this
whole
argument
is,
if
you
have
a
dynamic,
if
you
have
a
dynamic
histogram,
if
you
have
one
that
like
is
exponential,
buckets
or
some
sort
of
or
logarithmic
buckets-
or
I
don't
know-
I'm
not
an
expert
here,
but
then
it
doesn't
matter,
but
at
the
moment
we
don't
really
have
that
as
a
standard
thing
in
open,
telemetry
yeah.
So
that's
definitely
kind
of
the
better
way
to
do.
A
It
is
just
have
something
that
dynamically
adjusts
itself
to
whatever
comes
in,
and
I
think
josh,
the
other
josh
lightstep
josh
was
saying
that
there's
something
called
open,
histogram,
which
is
exists
now,
which
is
actually
a
donation
from
some
company
metrics
company.
I
don't
remember,
I
don't
remember
who
it
was
yeah,
this
log
linear
histogram
that
they're
donating,
but
there's
some
issue
with
getting
it
accepted
into
the
cncf.
A
I
don't
know
some
licensing
or
something
weirdness
that
has
to
be
worked
through
before
it
can
be
officially
in
the
cncf,
but
I
think
the
idea
was
if
we
could
get
this
thing
into
yeah
circonus
is
the
company
that
developed
it
that
this
would
be
a
really
great
option.
That's
super
flexible
and
just
works
really
really
well
so.
A
A
So,
anyway,
I
think
that
everyone,
most
people,
agree
that
if
we
could
just
get
something
like
this
in
there,
then
we
wouldn't
have
to
worry
about
bucketing
anyway,
except
if
your
backend
doesn't
support
log,
linear,
histograms
or
have
a
way
to
map
them,
and
I
think
otlp
doesn't
currently
have
a
way
to
map
these
histograms
into
the
data
model.
So
that's
also
a
blocker
at
the
moment
for
making
that
happen.
A
Oh
yeah
yeah
yeah,
there's
tests
that
rely
on
data
ordering
my
recommendation
was.
We
should
instead
build
some
fluent
test
stuff,
like
we
have
for
tracking
yeah.
A
C
A
A
A
D
B
Some
concurrency
not
really
but
but
I
did
mention
that
you
have
reviewed
some
of
our
concurrency
stuff
before
what
is
this
exemplar?
Oh
yeah,
and
that's
how
we
got
on
the
try
tree
discussion.
B
D
C
B
B
A
Yeah
yeah,
the
allocations
are
good,
it
was
I
was.
There
was
some
discussion
about
why
the
garbage
collection
didn't
change,
and
my
suspicion
is
that
g1
is
probably
just
super
good
about
getting
rid
of
that
very
short-lived
garbage
that
doesn't
make
that
much
difference
that
if
we
ran
this
with
cms,
we
would
see
a
very
different
answer.
B
A
But
if
you,
if
you
had
a
zone
in
your
in
your
heap,
that
was
like
just
the
like-
I
mean
there
needs
to
be
some
yeah
there's
some
escape
analysis,
but
I
think
it
could
be
very
simple
because
if
you
know
that
it's
literally
gone
in
and
out
of
scope
without
any
other
references
to
it
within
the
method,
in
that
particular
case
at
run
time,
not
at
not
at
compile
time,
then
you
would
know
you
would
never
need
to.
You
need
to
promote
it
to
the
next
gen.
But
I
I'm
not
a
gc
expert.
A
B
Yes,
yeah,
it's
actually
the.
I
wonder
how,
because
of
that
particular
allocation
issue
that
got
fixed.
It
did
feel
like
something
that
could
potentially
have
been
jitted
away
by
itself,
although
clearly
it
wasn't-
or
there
was
some.
A
B
B
All
right
I
mean
I
meant
escaping
out
like
allocated
on
the
stack,
but
do
any
vms.
C
C
Character
string,
it's
supposed
to
detect
that
as
a
character
like
because
somewhere
in
that
there
should
be
a
constant
32.
It's
a
trace,
id
dot,
hex
size
or
whatever
variable
we
have,
and
if
it
collapses
all
that
together,
it
should
be
able
to
detect
that
that's
a
fixed
size
string.
But
in
a
particular
case
it
was
a
string
concat.
A
C
B
And
having
the
overhead,
the
allocation
overhead,
essentially
with
and
without
the
agent.
B
Nikita
asked
who
was
screening
up
in
the
release
notes
he
was
looking
for
the
k,
http
stuff
and
felt
like
this
should
be
under
a
breaking
changes.
Banner
like
it
should
be
more
highlighted,
and
then
we
talked
about
that
with.
B
C
B
B
C
C
B
Okay,
so
I
was
gonna
remove
the
three.
I
think
I
went
a
little
overboard.
D
A
B
Yeah-
and
I
know
that
josh
is,
they
are.
B
B
B
B
Oh
man,
well,
I
learned
a
little
bit
more
about
cradle
still,
not
I
I'm
still
like.
I
don't
understand
it
enough
to
to
now.
Can
you
are
you
able
to?
How
do
you
debug,
like
I
noticed
with
the
kotlin
like
it,
worked
pretty
nicely
as
far
as
I
was
able
to
drill
into
like
the
shadow
plug-in
code
actually
showed
up
in
intellij
but
like
if
you
debug
or
how?
How
do
you
understand,
follow
the
gradle.
A
Yeah,
if
you
can
get
the
source
into
idea,
which
is
the
actually
the
hardest
part
because
it
doesn't
like,
let
you
navigate
to
you-
can
navigate
it
from
calls.
If
it
knows
where
what
you're
trying
to
call
like
it's
harder
in
groovy
than
it
is
in
kotlin
yeah,
and
if
you
could,
but
if
you
can
get
the
source
to
show
up,
then
you
can
put
breakpoints
in
and
it
works
just
fine.
B
So
what
do
you
do
you?
You
run
your
just
run
your
regular
task
in
intellij
and
it
will
pause.
A
Yeah,
you
could
yeah
definitely
like
if
you
run,
I
think,
even
recently,
there's
actually
little
gutter
icons
to
run
gradle
tasks
or
something
I
don't
remember
exactly
what
it
is,
but
yeah
you
can
run
it.
You
can
run
it
in
debug,
like
from
the
gradle
like
in
the
right
hand,
thing
where
there's
the
gradle
menu
you
can
run
with
debug
in
there
and
it
works.
C
B
C
B
B
I
think
I
was
getting
distracted
by
another
problem
which
actually
existed
already.
If
you
look
at-
let's
say:
okay
http.
B
So
we
are.
We
are
publishing
the
alls
for.
C
And
by
the
way,
sorry,
I
remember
you
can
finish-
I
guess
maybe
for
performance
reasons
or
something
that
looks
like
now
on
tasks.
Intellij
has
a
checkbox
to
enable
script
debug,
it's
not
enabled
by
default.
B
You
can
okay,
okay,
cool,
so
the
and
this
this
predated
your
changes,
because
I
noticed
this
was
also
for.
B
Yeah,
so
I
was
trying
to
get
that
so
I
was
trying
to
fix
like
three
different
things
at
once
that
I
just
kept
going
like
breaking
one
thing.
Every
time
I
broke
one
thing
I
fixed
another,
fixed
plumping
problem,
yeah
yeah,.
B
C
A
A
A
A
A
By
the
way
I
saw
I
need
to
hunt
this
down.
I
saw
an
interesting
thing
on
twitter,
where
someone
had
with
a
simple
script:
kind
of
reproduce:
the
really
nice
bin
tray,
publish
to
maven
central
feature
without
having
to
put
your
secrets,
splatter
your
secrets
everywhere.
I
need
to.
I
need
to
dig
it
up.
I
don't
really
quite
understand
it,
but
it
seemed
kind
of
interesting
splatter,
your
secrets
everywhere.
A
Well,
you
know
I
mean
if
you
want
to
publish
like
we
have
to
publish
david
central
now
we
have
to
put
our
gpg
key
into
ghea
and
we
have
to
put
our
username
and
password,
whereas
we
we
used
to
be
able
to
do
that
just
centrally
in
bin
tray
right
and
not
have
to
put
it
into
our
ci
yeah,
like
you
could
log
into
bin
tray
and
just
manage
it.
There.
C
A
Quite
understand
it,
but
it
might
involve
like
running
something
at
the
command
line,
with
your
secrets
locally
to
as
one
part
of
your
publish
step
or
something
I
anyway,
I
don't
know,
I
thought
it
was
just
something
that
I
could
look
into
yeah
I
need
to.
I
of
course,
lost
lost
it
as
soon
as
I
navigated
out
of
twitter,
so
I
may
never.
I
may
never
find
it
again.
It
might
have
been
a
fake
news
right.
It's
on
twitter,
I
mean
it
was.
There
was
the
repo
that
you
could
go.
A
Yes,
oh
on
rock,
did
you
see
the
feature
request
in
no-tell
java
for
an
aws
yep?
That's
very
dynamic,
sampler.
C
Yeah
we've
been
wanting
to
implement
all
that
we've
been
planning
to
implement
after
you
soon
now
that
there's
an
issue
that
makes
me
even
happier
to
implement
it
yeah
cool,
because
it's
basically
the
only
feature
that
we're
missing
in
hotel
right
now
can
protect
your
sdks
once.
That's
in
then
no
point
at
all
for
x-ray
stickers.
So
yeah.
C
B
Now
is
that
do
you
have
to
be
using
x-ray
back-end
to
use
that.
C
C
C
A
C
D
C
B
Yep,
okay,
you
june,
there's
a
word
for
it
in
japanese
in
in
california.
In
where
june
june
gloom
I'm
trying
to
think
where,
where
I
lived
that
june
glue,
I
guess
maybe
the
bay
area
has
june
gluten,
who
has
june
gloom
somewhere
in
california,
where
I've
lived
once.
C
B
Yeah,
no,
that
makes
sense
because
you're
you're
you're
about
you're
fairly
similar.
I
think
what
are
you
took
your
parallel.
B
B
C
B
Yeah
learned
where
pittsburgh
was
in
this
morning's
call.
B
Because
josh
lives
there
and
I
had
no
idea
that
it
was
so
far
west.
B
Oh
look
at
this.
I
mean
you
likely
know,
but
look
how.
A
B
He
said
people
in
pittsburgh
consider
themselves
to
be
east
coast,
but
the
rest
of
the
east
coast
doesn't
consider
them.
C
B
Not
you're,
not
real
east
coast,
that's
funny,
yeah
yeah,
I
mean
look,
how
close
they
are
to
columbus,
ohio.
C
C
B
A
B
My
my
spec
pr
isn't
going
anywhere
or
my
spec
issue
for
this
isn't
going
anywhere,
but
that's
okay,
we'll
just
we'll
just
won't,
do
anything
on
either
side,
we'll.
C
B
D
C
B
Yeah,
what's
he
doing
now.
C
C
B
C
For
example,
like
passing
in
the
url
and
getting
the
path
out
of
it
or
something,
but
I
pointed
him
at
the
way
our
extractors
have
that
base,
like
the
most
of
the
logic,
is
in
the
base
class.
That's
not
overwritten.
The
overrides
are
just
to
get
those
attributes
and
then
he
said,
oh,
that
sort
of
makes
sense.
C
C
C
B
So
what
does
what
does
the
wrapper
full
rapper
get
us
like
a
recruit
being
a
request?
What
are
we
talking
about
rappers
so.
C
B
C
B
C
B
Yeah,
could
you
have
his
his
phone
number
his
his
contact.
C
A
A
A
C
B
Yeah
yeah,
we
have
yeah,
I
interested
that's
some
feedback.
I
need
to
give
my
manager.
We
have
a
somebody
on
our
team,
a
more
junior
person
also
who
does
two
languages,
and
I
think
that
it
is
really
tough.
C
B
C
B
B
They
have
a
shim
kind
of
like
a
couch
db
kind
of
how
the
we
have.
They
have
their
own
tracing
api,
and
then
they
have
the
open,
telemetry
shim
to
send
open.
B
And
what
I've
done
for
our
distro
and
I'll,
probably
upstream
it
it's
or
I
will
upstream
it
at
some
point-
is
auto-injecting
that
shim
into
the
class
path
so
that
it
gets
picked
up
and
used
and
the
advantage
one
thing:
that's
nice
there
is
oh
well,
I
guess
the
client,
the
duplicate
client
span
thing
is
still
a
problem
because
they're
not
attaching
it
into
our
they're,
not
it's
still
just
a
normal
client
span.
B
It's
not
going
to
our
special
client
span
key,
so
it
actually
wouldn't
suppress
the
nested
nettie,
because
they're
primarily
nettie
under
the
hood,
although
you
can
also
plug
in
okay,
http.
C
B
Oh,
she
opened
a
a
spec
issue.
That's
probably
what
you're,
okay.
C
B
I
think
let
me
get
let's
talk
about
this
next
week.
Yeah.
B
Cause
I
actually,
I
haven't
chatted
with
her
since
she
submitted
this
there's
some
clarification.
I
need
also
yeah.
C
Yeah,
so
okay
no
worries,
so
this
is
actually
the
one
I
had
in
my
mind,
not
the
other
one,
and
it
came
up
like
perfectly
timed
with
ken
bringing
up
the
topic
and
the
slack.
So
that
was
interesting,
like
I
think
ken
brought
that
up
a
couple
of
days
ago,
right
right,
the
yeah,
yeah
yeah.
It
says
everyone's
talking
about
this
now.
B
Yeah
yeah,
so
actually
it's
really
cool
because
ludmila
was
in.
She
had
left
microsoft,
maybe
like
a
year
ago
and
she
came
back
and
she
joined
just
recently
and
joined
the
azure
sdk
team.
Doing
training
because
she
used
to
be
in
the
racing
team
responsible.
A
B
Like
the
distributed
tracing
for
azure
monitor
for
a
long
time
for
a
while
yeah,
so
yeah
we'll
get
a
lot
of
good,
effective.