►
From YouTube: 2021-04-23 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
B
C
B
B
I
mean
the
other.
The
other
thing
is,
I
think,
traffic
brought
this
up.
Do
we
need
to
use
thread
local
random?
Does
it
have
any
advantages
over
just
having
a
thread
local
with
a
random
inside
it?
That
works
just
fine,
it's
a
few
couple
fewer
lines
of
code,
but
I
mean:
does
it
really
make
that
much
difference?
Can
we
just
I
mean
yeah?
I
don't
know,
I
think,
that's
a
valid
question.
C
C
C
D
That,
like
certainly
it
could
you'll
get
a
shared
copy
with
everybody
else,
who's
using
thread
local
random,
so
that
that
you'd
get
benefit
there.
B
C
C
D
C
A
heart
they
have
a
different
model-
I
don't
know
probably
yeah,
so
it's
it's
very
deeply
implemented
in
the
jdk
I
was.
I
did
not
expect
that
so
yeah.
That
must
be
why
personally,
this
is
like
they
also
have
add
to
jdk.internal
that
vm
annotation
contended
on
this
field,
whatever
that
triggers
hotspot
to
do,
I
don't
know.
B
E
C
B
C
D
C
B
C
B
B
B
D
Tried
to
get
a
little
cat
cat
nap
in
before
before
this
meeting.
B
So
other
stuff
we
talked
about
today,
jack
berg
showed
up,
which
is
nice,
and
you
wanted
to
know
about
view
registry
auto
configuration,
and
I
said
that
thing
is
so
prototypey
at
this
point.
I
don't
think
we
should
be
building
auto.
It's
auto
configuration
for
it,
yet
until
we
have
spec
for
that.
B
That
I
mean
that's
one
thing:
that
is
one
option.
You
can
also
do
it
just
via
the
type
the
instrument
type.
B
Right
but
my
I
mean
it
sounds
like
from
the
I've
been
listening
into
the
metrics
api
group,
because
it
sounds
like
they're
going
to
put
something
in
that
they're
planning
on
putting
something
in
the
api
called
a
hint
api
which
is
going
to
be
kind
of
like
views,
but
for
instrumentation
people
to
like
suggest
what
sort
of
aggregation
you
might
want
to
use.
B
D
B
Make
sense
so
the
people
who
do
that
I
have.
I
have
done
that
before
mostly
when
I
was
mostly
when
kotlin
was
very
young
and
I
was
trying
to
get
it
to
work
with
jackson,
data
bindings
and
needed
parameters
on
things
for
jackson
data
bindings
to
pick
things
up,
but
that's
I
have
that's
the
only
time
I've
used.
It
is
when
trying
to
like
take
kotlin
code
or
it's
a
kotlin
code.
I
think
it's.
I
think
it
was
kotlin
code
or
maybe
it
would
have
been.
Also
lombok
like
make
lombok
worked
with
jackson,
yeah.
B
C
No,
I
mean
we
definitely
wouldn't
require.
We'd
still
have
the
name
field.
So
if
you
compile
with
parameters,
you
don't
need
it.
If
you
don't
you
need
it,
but
unfortunately,
that's
not
a
compile
time
check.
That's
just
a
hand-wavy
check,
so
it
is
a
bit
questionable,
but
at
the
same
time,
if
you
compile
primaries,
you
get
around
some
of
the
noisiness
that
you
don't
like
about
this
whole.
C
D
D
Because
I
mean
if
the
that
was
my
my
main
barometer
was,
if
other
annotations
are
doing
it,
then
I'm
like
okay,
otherwise,
I
would
say
no.
I.
E
C
C
C
C
D
B
D
Yeah
yeah,
I
I
mean,
I
guess
I'm
just
because
it's
not
so
it's
not
like
common,
and
it
would
be
so
easy
to
break
something.
E
C
E
D
E
D
C
C
D
C
C
D
Yep
yeah:
that's
why
I
was
hoping
to
find
more
more
better,
clearer
dock
on
that
screen,
but
this
seems
to
be
a
pretty
good
indication
that
they
are
using
that.
D
Is
that
how
you
found
this
was
through
that
they
are
using
this
for
annotation
based
stuff.
C
C
C
C
E
B
C
A
C
D
C
C
C
John
doesn't
know
what
he
is
saying:
you'll
figure
it
out.
C
C
D
D
Did
we
discuss
yes,
sleuth
material
gave
us
sort
of
an
update
on
his
sleuth
sleuthing.
D
D
Interrupt
working
better,
you
mentioned
that
you
thought
that
we
could
do
a
a
bridge
or
you
know,
have
you
seen
the
sleuth
context
stuff
or
just
the
brave
content,
because
I
know
you
know
the
brain.
C
Good
surprise
right,
I
know
grpc
has
some
packs
related
to
the
reflection
service.
Maybe
some
bad
behavior.
D
C
D
D
That
yeah
not
a
whole
lot
else.
We
just
we
went
through
a
prs
from
the
last
week,
which
there
were
just
a
lot
of.
D
A
lot
of
build
small,
build.
D
Yeah,
you
know
all
this.
I
was
just
thinking
if
any
discussions
interesting
discussions
came.
D
D
D
Things
in
flight,
this
one
did
you,
are
you
done
with
it?
No
okay.
D
Oh
right
right,
I
remember
nikita.
D
D
C
C
C
D
Yeah
we've
got
and
and
yeah.
This
is
not
a
problem
to
we.
We
should
instrument
them,
should
decorate.
E
D
Yeah,
that
was
the
one
thing
we
had
discussed
with.
Nikita
was
not
in
partic
not
to
this
case,
but
that
as
long
as
we're
not
adding
overhead
to,
if
somebody's,
not
using
a
callback
that
seems
reasonable
to
instrument
callbacks
where
they,
where
they
do
exist,.
D
I
did
notice,
he
also
did
this
one.
I
remember,
I
think
who
laurie
mentioned
lori
had
taken
a
look
at
it.
Also
this
book-
and
I
had
looked
at
this
also
at
the
time
and
though
it
felt
simple
at
first
but
then
I
was
like
oh
adding
the
listener.
Then
you
have
to
track
the
one
that
we
wrap
and
add.
Yeah
then
remove
it.
So
I
was
like
no
but
looks
like
matish
went
for
it.
So
that's.
D
A
D
Create
a
new
context
basically,
and
they
could
even
set
their
own
trace
id
span
id,
I
guess,
and
then
we
would
well.
We
would
parent
that
it'd
be
a
little
weird,
I'm
not
even
sure
what
that
use
case
really
would
entail.
A
D
D
C
D
Little
bit
of
pushback
from
a
user
who
thought
that
was
that
the
agent
should
respect
the
trace
parent
header,
but
then,
when
I
pushed
back
on
that
they
said,
oh,
I
don't
really
need
that
behavior
it.
I
actually.
What
I
want
is
it
to
for
the
agent
to
not
break
the
trace
so
like.
Why.
C
C
C
C
C
D
Yeah
not
like
a
redirect
but
like
doing
that
servlet
dispatcher
forward.
Oh.
D
Yeah
or
the
internal
include
right,
and
so
you
know
I
mean
we
don't
have
those
anymore,
but
I
not
and
it
I
feel
like
if
somebody
if
somebody
wants
to
make
a,
I
would
be
okay.
If
somebody,
you
know,
really
wanted
them.
We
could.
I
don't
even
realize
this
question
yeah,
I
just
saw
the
removement.
I
was
yes.
C
C
A
C
D
Case
yeah
because
I
was
gonna
also,
I
was
gonna
look
at
those
the
other
spans
that
we
create
that
have
a
little
bit
more
value
than
this.
I
think,
but
are
still
kind
of
noisy
and
our
and
our
tests
show
that
is
those
send
error,
send
error
and
send
redirect
spans
so
anytime,
we
instrument
response,
dot,
sender
and
response
dot,
send
redirect
and
again
those
do
those
we
already
capture
that
info.
If
we
return
to
redirect
or
return
the
stat,
basically,
the
status
code.
A
D
B
B
I
mean
partly,
I
just
I
feel
like
we
need
somebody
like.
We
need
somebody
involved
in
this
project
and
knows
it,
and
I
have
time
at
the
moment
because,
because
of
like
the
ftk
is
kind
of
quiet,
so
just
seem
like
a
good
time
to
dig
in.
I
can
learn
something
new.
D
Yeah
I
mean
android
is
a
huge
java
ecosystem,
part
of
the
java
ecosystem.
Yep,
do
you
all
have
have
you
had
customers
for
android
stuff,
yet.
B
B
Yep
one
thing
that's
interesting:
I
have
not
even
attempted,
and
I
think
it
might
even
be
kind
of
impossible
to
track
http
requests
in
the
context
of
activities
and
fragments,
just
because
of
the
the
extreme
async
nature
of
the
way
that
works
like
the
fragment,
like
the
fragment.
Will
life
cycle
be
completely
done
and
then,
via
callbacks
and
bindings,
the
data
from
background
events
will
show
up
so
getting
any
kind
of
context.
Linkage
between
them
is
fairly
dicey.
B
Oh
okay,
but
you
like
load
up
the
screen
and
then
a
whole
like
stuff
just
happens
behind
like
via
binding
internal
binding
framework
and
all
sorts
of
things
like
that,
to
get
it
all
wired
together,
so
figuring
out
how
to
actually
hook
in
hook.
The
context
together
is
relatively
tricky,
but
also
maybe
it
doesn't
matter.
D
Yeah,
so
not
so
trying
to
do
more
than
just
your
basic
http
client,
starting
the
trace,
starting
the
distributed
trace
from
the
device
yeah,
but
also
trying
to
capture
some
some.
B
B
Yeah,
it
would
be
really
cool
to
be
able
to
do
that
and
I'm
not.
I
don't
know
how
possible
it
is.
Okay,
I
mean
it's
probably
about,
like
literally
I've
gone
through
the
android
tutorials
at
this
point,
so
it
may
be
possible,
but
there's
also
some
interesting.
I
mean
I'm
basically
just
using
all
of
the
things
that
I'm
trying
to
track
right
now
are
using
the
built-in
android
apis
for
activity,
lifecycle,
callbacks
and
fragment
lifecycle,
callbacks
and
there's
all
sorts
of
like
interesting
things
like.
B
If
you
want
to
parent
the
fragment
creation
under
the
activity
creation,
you
have
to
keep
scopes
around
in
ways
that
make
me
very
uncomfortable,
like
you
can't
you
can't
put
them
in
a
nice
tri-finally
block,
for
example,
because
because
it
spans
the
only
the
method
that
we
have
access
to
so
anyway,
are
you
using
our
library
instrumentation,
I'm
using
the
library
instrumentation
for
okay,
hdpe
yeah,
and
it
works
great.
It's
fantastic
super
easy.
C
No,
I
mean
I
don't
because
the
reactive
pattern,
why
is
it
reactive?
It's
because
when
the
user
hits
a
button,
you
want
to
react
to
it
and,
like
that,
can
happen
all
over
the
place
and
you
might
need
to
debounce
and
all
these
things
so
responding
to
ui
reactively
makes
sense
and
you
have
a
request
response
pair.
It's
just
one-to-one
like
I
don't
find
the
reactive
model
to
help
with
that.
In
my
experience,
like
the
dagger
producers
work
well,
even
just
mapping.
C
C
B
Extension
also
attribute
annotation
yeah,
but
again
I
think
there's
there's
a
trickiness
in
like
what
is
the.
What
is
the
context
like
they're,
like
I'm,
not
even
sure
what
you
want
to
propagate.
In
that
case,
that's
where
things
get
tricky,
because
those
often
the
co
routines
are
being
called
not
in
any
sort
of
normal,
rendering
lighting
might
be
called
they're
called
when
button
pushes
like
do
you
want
to
start
capturing
spans
around
every
button?
Push
I
I
don't
know
like.
These
are
the
things
where
I
think
you're
you
could
get
super.
B
It
could
get
super
super
noisy,
especially
when
you're
working
on
a
mobile
device-
and
you
don't
want
to
be
like
pushing
enormous
amounts
of
number
of
spans
around
so
like
figuring
out
what
the
appropriate
thing
to
instrument
and
whether
it
even
makes
sense
to
have
like
the
responsive
button,
push
show
up
in
some
sort
of
context
of
something
else.
I
just
I
don't
know.
C
D
B
Other
thing
is:
there's
no
automated
way
to
instrument
button
pushes
right.
The
user
is
going
to
have
to
do
that
themselves
if
they
want
to
add
some
sort
of
instrumentation
around
button
pushes
so
that's
that's
the
other
like
there.
There
just
are
no
hooks
in
android
for
for,
like
third-party
instrumentation
to
come
in
and
grab
access
to
all
that
all
that
stuff.
B
B
B
C
I'm
not
doing
it.
I
remember
even
the
google
maps
the
search
button
that
required
about
three
rpcs
to
service
and
if
that
button
had
a
span
around
all
of
those
that
would
be
pretty
cool
but
like
it
requires
me,
it
requires
me
to
know
that
that
search
button
has
all
these
bands
that
should
deserve
a
spam.
I
wouldn't
want
that
on
every
button,
but
yeah,
I
guess
that's
sort
of
the
pattern
that
we'd
end
up
with,
but
definitely
would
want
to
spin
the
button.
B
Yeah
yeah,
I
think
that
it's
interesting
that
looking
at
the
splunk
web
based
rum,
implementation
and
just
the
way
the
ui
works
is
they
they
don't
worry
about
tracking
kind
of
that
they
don't.
They
worry.
They
kind
of
track.
Events
like
what
happened
during
this
user
session.
What
sort
of
stuff
happened
and
they
tie
it
all
to
a
session
id
and
then
there's
just
things
that
happen
during
the
session.
It
doesn't
really
matter
if
there's
parenting,
proper
parenting
or
proper
context,
propagation
or
anything
like
that.
B
B
Well,
you
still
get.
I
mean
if
you
report
timings
on
things
that
happen
like
for
http
for
external
calls
and
those
things
still
need
to
be
distributed
like
distributed
tracing
for
external
calls
very
important,
but
like
the
internal,
like
the
internal
machinations
of
what
thing
led
to
the
next
thing,
they
just
do
it
via
time.
They
don't
worry
about,
because
it's
because
there's
one
user
in
a
session
right.
B
It's
not
like
it's
a
giant
multi-threaded
app
with
you
know
thousands
of
requests
coming
in
per
second
that
you
want
to
crack
the
details
of
it's
just
a
user
doing
stuff,
and
from
that
perspective
it
really
just
matters.
What's
the
timeline
within
the
session,
what
happens
yeah
so
tracking
that
kind
of
causal
relationship
between
activities
doesn't
matter
so
much
when
what
you're
really
just
doing
is
tagging
things
in
time.
B
B
Don't
think
no,
the
user
has
to
add
it
manually.
We
don't
like
I
know,
like
new
relic,
has
a
thing
where
yeah
it
will
automatically
add
it
to
the
application.
I
really
only
tell
you
that
that
thing
was
a
super
nightmare
to
maintain
it,
because
we
actually
like
that
code
came
down
from
the
new
relic
servers.
It
wasn't
because
it
had
to
be
kept
up
to
date
and
the
people
who
built
the
built,
the
javascript
for
not
the
same
team
to
build
the
java
agent.
B
D
Yeah
you
instrument,
I
mean
what
I've
seen.
People
do
is
instrument
the
the
servlet
output
stream,
detect.
D
D
D
But
user
yeah-
I
know
I
I
I'm.
I
know
they
want
to
do
that.
Our
our
product
managers
want
us
to
do
that.
So
that's
why
I'm
tracking
what
other
people
are
doing.
B
No,
I
I
do
not
believe.
As
far
as
I
know,
the
splunk
java
agent
is
not
doing
anything
like
that.
We
do.
The
splunk
java
agent
does
respond
with
an
additional
header,
like
it
adds
a
header
about
trace,
ids
or
something
that
are
generated
on
the
server.
I
don't
remember
why
or
how
that's
used,
but
it's
used
by
the
browser
in
some
way.
Oh.
D
C
D
The
good
news
there
is,
if
you
want,
I
have
I've,
already
implemented,
that
in
the
java
instrumentation
in
our.
A
B
D
The
new,
even
the
current
one,
we
have
to
send
back
a
response
header,
this
weird
handshake
of
whether
it's
it
really
has
to
end
up,
as
in
our
back
end
indexing
which
there
so
they're
getting
rid
of
it.
We
won't
have
to
do
it
and
they
keep
telling
me
you
know
soon.
A
C
D
Well,
well,
I
might
have,
I
might
not
have
gotten
the
generic.
You
know,
text
map
setter
route.
Just
you
know,
we
have
one
specific
need
and
one
specific
header.
D
B
D
Hey
now
that
I
know
you
have
it,
I
I
may
yeah,
we
should
just
put
it
in
the
hotel.
Instrumentation
we're
not
limited
in
our
repo
to
what's
w3c.
D
Deleted,
no
it's!
So
what
it
is
is
the
like
institut
our.
What
do
you
call
instrumentation
key
like
your
cust,
like
your?
What
what
back
end
bucket?
Are
you
sending
it
to
and
if
all
of
your
telemetry
goes
to
one
single
back-end
bucket,
then
that's
fine?
We
don't
need
that
correlation.
A
D
C
D
Somewhere
right:
well,
it's
just
like
a
indexing
because
on
the
back
end,
if
they
know
the
trace
id-
and
they
don't
want
to
sir
in
the
past,
what
they
didn't
want
to
do
is
search
all.
C
C
That's
something
like
with
defense,
because
I
know
the
response
header
stuff
would
always
just
makes
my
head
hurt
is,
I
think,
usually
it's
in
the
context
of
delaying
the
sampling
decision.
Also,
so,
instead
of
zero
or
one
you
have
like,
I
think
also
brave
supports
it,
and
amazon
also
supports
this
question.
Mark
type
of
thing
where
you.
D
D
D
C
D
D
C
D
C
I
forget
what
they
said.
Probably
shouldn't
say
it:
anyways
yeah
yeah,
that's
a.
E
C
D
Now,
what
about
tail
based
sampling.
A
C
C
D
D
Yeah,
okay,
yeah
you're,
not
using
the
collector,
not
yet
it's
it's
it's
it's
not
our
preferred
language
implemented
in
our
preferences
right,
so
yeah,
there's
some
and
and
there's
already
existing
sort
of
agents
that
live
everywhere,
and
I
don't
know,
there's
it's
interesting
and.
A
D
Also
interesting
what
I
know:
there's
people
are
very
kind
of
interested
in
what
aws
are
doing
and
and
but
mostly
just
the
collector,
like.
I
keep
forwarding
on
the
so
tigran
made
some
response
to
whether
you
were
nikita
about
oh,
the
schema,
the
telemetry
schema
stuff,
whether
we
would
you
know
implement
that
in
the
the
mappings
in
every
sdk,
or
only
in
the.
D
In
like
there
would
be
version
one
of
these
semantic
attributes
and
then
and
the
sdk
or
the
instrumentation
would
tag
it
saying.
This
is
the
version
that
I'm
using
and
the
collector
would
have
a
mapping
layer
to
support,
renaming
and
different
things
like
that,
so
that
we
wouldn't
be
totally
locked
in
to
you
know
v1
forever.
D
Which
seems
good,
I,
like
I
like
the
comment
about
reducing
the
like
the
oh,
not
the
overhead,
but
making
it
easier
to
add
semantic
conventions,
attributes
because
you're
not
so
worried
that
these
are
gonna
have
to
live
forever.
So
it
doesn't
have
to
be.
D
C
Well
and
then
your
collector
seems
to
make
sense,
because
I
mean
not
sdks,
because
it's
the
type
of
complexity,
you
don't
want
in
your
clients.
I
think
it's
either
in
the
back
end,
or
I
mean
it's
in
a
backend
of
some
sort,
I
think
usually
because,
for
example,
you
wouldn't
want
to
implement
that
logic
in
android
apps,
ios
apps,
like
totally,
you
generally
keep
the
logic
out
of
your
clients
so
and
then
collectors
a
nice
way
so
that
all
the
bactends
don't
have
to
do
this,
but
batons
could
also
implement
if
they
want
to.
D
D
They're
due
to
accepting
otl
right
exactly
yeah
yeah
yeah,
which
is
cert
one
of
the
things
that
you
know,
which
is
a
good
option
for
us.
If
we
don't
go
the
collector.
A
D
C
C
Yeah
I
finally
like
I
was
having
some
trouble
and
then
I
finally
realized.
I
mean
it's
something.
I'd
already
found
a
couple
months
ago
and
then
forgot
that
they
still
hadn't
fixed.
This
simple
spam
processor
doesn't
force
flush,
we
fixed
it
in
the
java
sdk
a
couple
months
ago,
but
and
I
was
using
the
simplest
band
processor
just
because
I
thought
first,
I
should
keep
things
simple
before
I
make
things
more
complicated
or
better
or
whatever,
but.
D
D
The
otlp
exporters
are
async
or
they
sync.
D
C
D
Is
somebody
else
doing
like
python
yep.
D
C
D
Auto
is
anybody
they're
working
on
the
the
auto
instrument
open
telemetry.net
automatically.
E
A
C
Also,
I
mean
I
also
feel
as
if
like
there's
so
many
spec
issues,
I
want
to
be
more
active
on,
but
I
just
don't
have
time
right
now,
so
really
looking
forward
to
this
pedaling
cool.
E
Well,
good
luck.
Thanks,
see.