►
From YouTube: 2021-04-28 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
B
Yeah
yeah,
but
so
today
you're
not
there.
Sorry
could
you
be
what
that
means
right
now
at
7
a.m.
For
you,
yeah
thanks
for
joining
swirly.
A
Yeah
he
said
something
about
being
in
a
bit
different
time
zone.
So
it's.
A
A
B
B
B
C
Oh
yes,
mattash
welcome
to
two
to
wed
twe
twen's
day
wednesday.
Yes,
really
trying
to
invent
trying
to
make
new
words
for
different
cover
art
time
zones.
C
Everything
else
is
it's
about
time.
Things
are
not
always
us
centric.
B
C
Cool
yeah,
you
all
are
also
welcome
to
schedule.
You
know
something
that's
in
the
the
middle
of
my
night,
but
you
know
it
is
nice
to
overlap
here
and
there.
C
Yeah
sorry,
I'm
trying.
C
B
C
It's
a
little
bit
more
like
a
laptop,
oh
okay,
it
is,
it
is,
you
know,
touch
and
you
can
take
off
the
you
can
take
off
the
screen
from
the
keyboard,
and
so
it
is
just
like
a
yeah,
and
it
works
just
like
that.
Also.
B
C
Nope,
my
laptop
is
not
happy
about
being
moved.
This
is
probably
the
first
like
the
first
time.
I've
unplugged
it
in
like
a
year
from
the
death
from
my
desktop.
I'm
not
even
sure
why
you
know
I
have
a
well.
I
guess
I
have
a
laptop
because
before
I
before
the
pandemic,
I
actually
the
plan
was
to
go
to
microsoft
campus
every
few
months
to
visit.
So
it's
nice
to
have
the
portable.
C
So
how
about
you
mathias,
you're
you're,
a
long
way
from
splunk
headquarters?
How
did
you
get
pulled
into
splunk.
A
Yeah,
well,
I
am
away
from
splunk
headquarters,
but
we
have
an
office
in
krakow,
so
and
hank
has
been
actively
hiring
some
people
in
craft
and
that's
how
I
got
the
job.
It
was
kind
of
funny
because
I
actually
started
the
recruitment
process
right
before
the
pandemic
or
just
as
it
hit,
and
it
stretched
from.
I
don't
know
two
weeks
to
three
months,
so
I
haven't
even
seen
the
krakow's
back
office
at
all.
I
don't
know
where
it
is.
I
just
got
my
laptop
in
the
mail
and
I
started
working.
A
A
C
How
is
it
yeah
we
have
done?
Statuses
like
nikita
was
asking
status
on
the
virus,
the
vaccine
roll
out.
Oh
my
god,
my
monitor
worked
in
different
countries.
Primarily,
I
think
nikita
was
commenting
on
japan's.
The
the
the
state
of
the
the
poor
honorary
is
not
gonna
get
vaccinated
until
like
2023
or
something.
B
B
C
C
I
pull
up,
or
do
you
want
to
go
through
and
I'll
take
notes
want.
C
28,
okay,
so
I
see
yeah
you,
you
decided
it
is
wednesday.
It
says
4,
28,
pacific
time
in
the
notes.
A
A
Yeah
I
had
the
chance
to
read
it.
I
just
wasn't:
I
hadn't
responded
yet
so
actually
what
you
described
here
with
this
another
response
was
my
first
idea
about
it,
but
then
I
thought
that
okay,
we
don't
really
have
label
response
at
any
other
extractor,
so
it
might
be
kind
of
strange
for
this
one
to
be
a
little
bit
different,
but
now,
on
the
other
hand,
you
are
absolutely
right
about
this.
B
A
B
A
B
A
B
B
On
the
same
page,
nice,
so
the
reset
super
conflict
wanted
to
make
sure
you
get
a
chance
to
review
the
stress,
because
I
think
you
originally
changed
to
resetting.
Did
you
get
it?
Did
you
get
to
read
this
yet?
No
like
seven
six,
but
it
is
making
sure
to
only
reset
if
there
was
a
week
rather.
B
B
Because
jrpc
does
use
the
contact
before
we
do,
and
so
if
we
reset
that
gets
deleted
yeah
and
before
we
didn't
have
the
bridge
that
still
was
okay
but
even
ignoring
bridges
and
whatnot
I
mean
grpc
could
be
using
hotel
context.
It's
a
it's
a
usable
context,
implementation,
in
which
case
then,
regardless
of
bridge
it
would
be
the
same
breakage.
So
in
principle,
we're
not
really
supposed
to
reset
the
context
in
case.
C
Yeah
yeah,
so
only
if
there's.
B
B
B
C
B
C
B
So
metrics
the
first
version
of
trying
to
do
any
metrics
inside
our
code
base
or
I
guess
not.
We
have
a
system
metric,
but
these
are
request.
Metrics
I
mean
we
don't
have
to
go
into
too
much
detail
right
now.
You
can
read
some
review
comments
later,
but
it's
just
taking
the
form
of
something
that
gets
called
on
start
and
end
with
attributes
and
then
within
it.
It
has
to
manage
its
own
state.
B
Like
I
didn't
know
what
I
was
getting
into
when
I
started
writing
and
I
was
like
oh,
I
need
to
have
the
state
variable
in
the
context,
or
else
there's
no
way
to
end
their
because
we
don't
know
what
labels
to
use
at
the
end
of
the
request
or
we
don't
know
what
the
start
time
was
at
the
end
of
their
question.
So
and
then
we
end
up
needing
this
sort
of
state
in
the
context
which
was
a
bit.
It
wasn't
what
I
was
anticipating,
but
I
guess
it's
sort
of
normal.
C
B
C
C
End
yeah
active
requests
right:
it's
actually
recording
a
metric
at
the
beginning
and
recording
a
metric
at
the
end.
Yeah.
B
B
Yeah,
and
so
the
other
aspect
was
in
our
semantic
conventions.
Currently
we
define
different
labels
for
different
metrics,
like
these
things
were
defined
for
active
request
in
the
spec,
and
then
these
things
were
defined
for
duration
in
the
spec.
They
were
different,
so
I
have
a
different
set
of
labels
for
each
metric.
I
hope
that's
not
the
final
direction,
because
this
is
pretty
annoying.
C
Okay,
okay,
yeah-
and
these
are
just
two-
so
these
are
http
server
and
is
that
state
that
you're
passing
across?
Is
that
just
part
of
the
http
server
instrumentation.
C
B
C
You
have
to
to
create
a
factory
I
see
for
the.
C
B
C
B
B
B
B
B
B
I
think
it's
fine
and
I
think
it's
sort
of
required
in
some
of
these
cases.
We
can
try
to
not
go
overboard.
Some
libraries
go
overboard
on
it
and
it
gets
makes
a
very
confusing,
but
for
the
especially
when
we're
working
on
alpha
signals,
I
think
it's
for
the
instrumentation
api.
It's
really
important,
because
there's
no
way
we
can
experiment
without
having
these
knobs.
I
think.
C
Yeah
yeah,
I'm
I
mean
I'm,
I'm
totally
fine
with
that.
Having
a
marking
it
unstable.
B
C
Want
that
that
to
hold
up
being
able
to
make
the
instrumentation
api
release
it
stable
before
the
metrics
api.
C
B
B
C
Yeah
I
like
it,
do
you
want,
I
mean,
I
think
this
is
awesome.
I'd
love
to
get
you
know,
merge
our
first
metrics
something
in
there.
B
Just
so
that
we'll
be
cleaning
this
up,
and
then
I
mean,
if
matthias
has
changed,
it
gets
merged.
First,
then
I'll
update
mine,
move
it
to
the
http
package
or
whatnot
or
otherwise.
Maybe
either
of
us
could
probably
do
that.
That's
the
only
thing
but
yeah,
so
I
just
upgrade
some
javadoc
and
unit
tests.
B
I
will
finish
my
brs
today,
but
I
think
yeah.
I
think
mateus
will
finish
first,
because
I
have
a
few
days
of
vacation
starting
tomorrow
we
have
golden
week,
oh
yeah,
what's
golden
it's
a
week
that
has
like
japanese
holidays
in
succession,
so
national
holidays,
so
thursday,
monday
tuesday,
wednesday
holidays.
C
B
A
Yeah
and
like
80
or
90
of
it
is
just
imports
so
I'll
make
this
reviewable
today.
But
I
can
talk
to
you
and
explain
what
what
I
tried
to
do
here
anyway.
A
So
I
wanted
to
create
a
module
that
will
have
all
those
classes
that
are
supposed
to
be
used
by
java
agent,
instrumentation
developer
or
extension
developer.
A
If
I
can
borrow
the
terminology
that
nikita
uses
in
its
extension
class
builder-
and
I
set
out
to
untangle
all
those
masks
that
we
have
in
java
agent
link
and
I
was
able
to
extract
all
those
classes
that
you
hear
and
the
image
imagine
the
appear
here
to
a
separate
module,
and
there
are
two
things
that
are
pretty
new
to
this
to
our
instrumentation
link,
and
these
are
the
agent
extension
and
agent
extension
tooling
classes
if
you
could
maybe
find
them
in
the
pr
yeah.
I'm
not
sure
if
this
is
an
easy.
A
Yeah
I
can
share
my
screen
if
we
or
yeah-
let's
maybe
talk-
I
probably
don't
need
to
do
that
since
it's
here
already,
so
I
extracted
the
sort
of
a
parent-based
interface
of
all
extensions
modules
included,
and
this
is
this
interface
actually
has
the
same
capabilities
as
our
own
byte
body,
something
customizer.
So
I
duplicated
that
one
and
it
basically
has
one
one
method:
that's
worth
talking
about,
and
it's
the
extent
method
which
receives
an
agent
and
some
object
called
to
link
and
retains
an
agent.
A
So
you
can
do
pretty
much
whatever
you
want
in
it.
Instrumentation
module
now
extends
this
implements
this
interface.
So
all
our
instrumentation
logic
is
also
an
extension
is
some
kind
of
an
extension
and
if
you
switch
to
the
tooling
class,
maybe
oh
yeah,
not
the
internet.
A
A
A
C
On
this
api,
nice
on
this
class
here,
like
the
create
type
pool,
was
that
I'm
not
remembering
what
instrumentation
that
was
used
for
it's
used
in
muzzle
in
the
muzzle
reference
matter,
which.
C
Okay
and
the
other
things
here
I
see
so
right,
I
see
so
the
the
instrumentation
modules
don't
have
a
base
class
anymore.
They
just
have
to
implement
the
interface.
A
Oh
they
do,
they
do
have
the
instrumentation
module-based
class,
which
has
all
the
logic
of
applying
advices
and
stuff.
It's
just
the
the
instrumentation
module
class
itself
implements
this
interface,
so
we
have
like
a
base
interface
for
every
possible
extension.
A
But
yeah
it
definitely
should
so.
I
changed
the
package
name
of
several
classes
and
I'll
try
to
split
this
into
two
comments,
one
with
all
those
actual
changes
and
the
other
one
with
moving
stuff
around.
So
the
second
one.
A
A
A
A
B
A
B
B
A
A
B
A
So
this
is
the
fix
for
that,
and
this
is
actually
pretty
similar
to
the
unintermittent
abstract
method,
one
that
I
I've
done
something
a
few
months
ago,
but
it's
kind
of
reverse.
So
we
have
now
helper
classes
that
might
extend
some
library
based
classes,
and
this
was
the
case
in
rx
java
at
the
back
and
what
we
do
now.
A
We
collect
field
references
in
helper
classes
and
if
they
are
defined
in
or
declared
in
the
hyper
classes,
we
mark
them
as
declared
if
they
are
not-
and
this
is
just
usage
of
any
field-
they
are
marked
as
undeclared,
and
the
new
thing
is
here
is
that
we
get
all
those
undeclared
fields
that
were
used
in
a
helper
class
and
check
if
they
are
declared
somewhere
else
in
the
type
here.
A
So
now,
if
we
have
a
field,
that's
called
then,
if
we
use
a
field
somewhere
in
what
some
of
the
methods
on
our
hypercast
the
field,
that's
called
qs
and
we
don't
find
that
field
somewhere
in
the
base
class
or
the
superclass
of
the
base
class.
Then
it's
then
it's
an
error
and
muzzle
stops.
The
instrumentation.
A
But
it
was
like
we
were
actually
collecting
like
field
usages.
So
whenever
we
referred
to
a
field
in
a
content
of
a
method
in
a
body
of
methods,.
A
C
B
B
C
B
C
B
Of
a
extension
to
your
function,
you
can
register
and
then
it
just
adds
the
files.
So
if
you
add
our
lambda
layer
here,
function
will
just
automatically
get
traced
instead
of
x3,
and
we
want
to
have
a
similar
layer.
That's
published
for
upstream
as
well.
There's
been
open,
telemetry
that
publishes
in
otp,
but
we
don't
have
any
adobes
that
can't
be
published
here.
B
B
C
Oh,
I
understand
you
need
a
an
open,
telemetry
aws
account
owned.
B
C
C
B
B
C
Neat,
and
so
that's
with
the
do-
you
have
one
with
and
one
without
the
java
agent.
B
B
We
have
one
with
the
java
wrapper,
so
we
have
two
java
versions.
Okay,
the
java
wrapper
one
doesn't
have
too
much
start
time.
So
it's
okay.
Obviously
you
have
to
instrument
your
clients
yourself
in
that
case
right
job
agent,
it's
slow,
but
once
it's
worn
then
I
guess
it's
good
so,
depending
on
your
function,
pattern
that
might
work
okay
for
you
yeah
always
the
wrapper
is
going
away
either
way.
In
my
opinion,
they
would
suck,
but
the
cold
start
time
is
so
long,
even
with
the
wrapper.
It's
around
15
seconds
really.
B
Yeah,
I
think
just
I
mean
class
loading.
We
know
class
loading
is
slow
in
java
right.
So
even
if
we're
not
doing
byte
code
instrumentation,
just
loading
neti
and
all
these
things
it
takes
a
while
wow
someday.
I
want
to
experiment
with
replacing
so
in
the
rap
ratchet
I
picked
jrpc
and
eddie.
Just
to
make
sure
I
don't
run
into
problems,
only
is
what
we
normally
use,
but
I'm
going
to
look
at
grpcc
just
to
see
if
that
behaves
differently
and
may
be
better
for
lambda.
Your
raw
throughput
isn't
as
important
as
binary
size.
C
What's
the
yeah,
it
keeps
seeing
more
about
like
growl,
stuff,
yeah.
B
C
I
thought
there
was
some
some
fun
bug
with
the
scala
random
that
if
you
had
an
app
a
gra
app,
it
would
statically
compile
the
random
function.
The
seed.