►
From YouTube: 2021-10-12 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).
A
B
Oh
yeah,
so
I
have
a
pr
that
removes
almost
all
those
virtual
field:
computers
now
usages
and
the
only
one
that's
left
this
one
in
the
we
have
a
few
executor,
instrumentation
helper
classes
in
the
java
engine,
instrumentation
api
module
and
there's
one
particular
usage
in
executor-
advice
youtube.
Maybe
I
think
oh
no
helper
advocate
is
the
keto
advice
helper.
B
Get
that
one.
Probably
that
kind
of
seems
to
oh
yeah
line
40
something
here
it
is,
it
kind
of
seems
to
make
sense.
I
mean
to
be
to
be
atomic
because
you
probably
want
to
have
some
sort
of.
B
I
mean
any
instance
of
propagated
context
attached
to
the
task
at
all
times,
because
the
propagated
context,
class
itself
has
also
a
virtual
field
with
the
actual
open,
telemetry
context,
and
it
has
some
sophisticated
set
context,
clear
context,
methods
that
actually
check.
If
something
overwritten
the
context
has
overwritten
the
context
or
not.
B
B
A
B
You
click
on
the
set
context,
method
of
the
propagated
context
and
just
move
to
its
definition,
which
is
yeah.
It
seems
to
be
this
one,
so
the
only
thing
we
lose
is
that,
like
the
blog
statement,
because
you're
true
in
saying
that
one
of
oh,
I
that's
going
to
win,
but
we
are
not
going
to
notice
that.
B
B
A
Else
has
other
thoughts.
B
B
A
Honorable
does
this
need
to
happen
before?
Can
we
change
annotations
after
declaring
stability.
A
D
C
A
C
C
C
B
A
So
that
that's
my
only
my
only
thought
on
this
concern
is:
do
we
is
this?
Is
there
what's
the
purpose,
but
I
don't
really
have
any
serious
objection
to
if
you
know
like
almost
like,
I
see
it
as
okay,
maybe
like
a
debug
flag,
it
feels
like
that
we
would
you
know,
but
that's.
B
A
And
even
if
we
do
expose
something
well,
we'll
have
the
ability.
C
C
C
C
C
The
question
may
be:
is
how
actually
to
achieve
that,
because
I
have,
I
have
the
related
question
about
that
unstable
api
or
what
type
of
annotation,
how
exactly
that
annotation
is
supposed
to
work.
C
D
Stable
api
is
really
only
for
end
users.
It,
like
libraries,
aren't
allowed
to
use
unstable
api,
which
then
defeats
the
purpose
of
the
instrumentary
api
from
the
one
stable
api.
So
this
is
actually
a
pretty
big
problem.
I
guess
how
can
anyone
write
instrumentation
as
long
as
our
extractors
are
unstable,
but
how
can
we
make
the
extractor
stable,
with
this
meant?
Cementite
conventions
are
unstable
where
it
gets
stuck
here.
I
think.
D
D
D
C
So
I
see
at
least
two
profits
of
publishing
stable
instrument,
their
api,
even
without
semcons
extractors.
One
is
that
things
anybody
can
still
write.
I
don't
know
either
internal
spam
instrumentations
two.
When
open
telemetry,
instrumentation
finally
publishes
some
stable,
like
http
convention,
then
we
release
http
extractors
and
everything
else
is
already
like
stable.
It's
it's
lesser
steps,
smaller
steps.
B
If
they
implement
a
custom
attributes
extractors
themselves,
it
will
be
stable,
so
they
won't
have
to
change
and
the
actual
end
users
won't
have
to
change
any
custom
attributes
extractors.
They
they
add,
even
though
the
ok
http
library
instrumentations
might
need
to
change.
You
know
the
http
extractor,
but
that
will
be
completely
invisible
to
the
end
user.
D
I
guess
we're
thinking.
If
an
external
instrumentation
uses
our
instrumentation
api,
they
still
need
to
use
the
open
home
gpi
to
add
the
attributes
until
they
make
any
attributes
extractors
table
that
they
could
at
least
use
the
other
functionality,
and
we
hope
that
they
follow
this
rather
than
using
our
offline
structures.
A
C
You're
right
right
hotel.
I
believe
that
the
official
denomination
was
native
instrumentation.
C
B
I
think
there
was
also
a
mention
of
ordering
the
spam
something
rubber,
so
we
have
this
order
interface
in
our
extension
api.
Maybe
it
would
be
a
good
idea
to
copy
or
move
that
to
the
sdk
airport.
D
C
D
C
C
B
There
is
also
this
issue
of
cooperation
between
java
and
library,
instrumentations,
which
I
didn't
dare
to
mention
here,
because
nobody
has
any
idea
how
to
implement
it.
C
C
C
C
D
B
B
Yeah,
I
think
that
it's
like
mostly
uncharted
territories
and
there's
a
lot
of
changes
that
we
need
to
do
before.
It's
this
whole
module
or
whatever
it
turns
out
to
be
before
before
it's
actually
usable
by
somebody
other
than
us.
D
I
don't
know
how
much
it
simplifies
it,
but
it's
too
claimed
test
and
the
extensions
like
for
junit5
then,
ideally
also
an
http
server
test.
Migrated
javascreen
pattern,
like
those,
I
think,
are
enough
to
provide
a
lot
of
benefits.
So
if
the
work
for
that
is
small,
then
that
would
be
the
one
to
separate
to
publish
a
stable
in
the
beginning
without
the
rest
of
the
stuff.
C
First,
I
totally
agree
that,
yes,
those
are
http
test
or
http
server.
That's
indeed
the
first
very,
very
usable
article
that
we
have
to
publish,
but
I
cannot
imagine
their
conversion
to
java.
D
D
C
A
Why
do
we,
like
stability
of
testing
comments,
seems
much
lower
priority
to
me
because
breaking
like
there's
not
the
problem
of
breaking
changes
is
not
as
severe
as
foreign.
D
C
Yeah
so
tests,
yes,.
D
C
B
A
No,
but
that
wasn't
what
he
was
proposing,
but
I
would
be
okay
with
that.
D
A
A
We
have
at
least
we
have
controller
spam.
Most
of
them
are
suppressable
now,
via
the
controller
span,
config
that
I
added,
because
I
I
I
have
it-
I
disabled
it
in
our
distro,
mainly
because
we
have
no
way
to
go
over
the
controller
span
to
get
the
server
span,
which
is,
and
that's
like
one
of
the
top
use
like
very
common
user
needs,
is
to
add
attributes
onto
the
request.
Spam.
A
Because
I'm
not
telling
our
users
about
unstable,
instrumentation,
api,
okay,
fair.
C
C
B
A
C
A
That's
what
that's,
what
the
config
option
we
have
right
now
does.
Is
it
suppresses
controller
spans
but
keeps
the
it
still
captures
the
route.
A
Yeah
I
mean
the
feedback
that
I've
gotten
in
the
past.
Is
some
people
really
like
them
like
from
your
standard
spring
web
and
vc
kind
of
thing?
It
shows
you,
basically
your
annotated
methods
and
it
some
people
like
them.
A
I
wanted
to
have
them
but
kind
of
the
downside
kind
of
outweigh
that
for
now,
but
yeah.
I
would
definitely
check
nikita
before
in
your
in
your
distro,
because
competitive
I
like
I've.
Even
I've
heard
that
as
a
competitive
disadvantage
even
before
we
when
we
didn't
have-
and
now
we
don't
have
them
again,
but
people
said
that,
oh,
but
you
know
xyz
product
has
these.
I
can
see
these
extra
spans
right.
D
D
A
Does
with
does
brave,
have
internal
spans.
D
D
D
So
I
was
just
thinking
of
making
a
group
called
eye
open,
telemetry,
internal
or
something
so
that
it's
at
least
not
as
obvious
for
consum
like
for
directly
referencing,
but
it's
published
so
that
our
users
can
run
our
build
yeah.
So
and
so
then
I
think
I'm
just
kind
of
a
manual
workflow
in
our
repo,
but
we
just
anytime
that
sdk
releases.
Then
we
can
run
it
so
it
shades
and
publishes.
E
D
D
D
D
D
A
A
A
It
used
to
ignore
all
of
this,
so
initially
I
tried
to
remove
this
and
because
we
don't
have
access
to
the
class
loader
here,
so
I
can't
check
this
and
bootstrap
class
loader
and
is
there
a
better
commit
that
doesn't
have
all
the
other
junk?
Yes,
probably
this
first
one.
A
And
so
I
added
here
basically
where
it
is
this
and
it's
the
bootstrap
class
loader
and
I
removed
this
and
then
I
also
added
there's
one
other
place.
We
have
to
check
over
here.
A
A
A
It
actually
wasn't
technically
errors.
It
was
that
thing
that
you
know
when
the
test
runs
and
if
there
were
any
transform
errors,
it
will
fail
the
test
at
the
end
it
used
to
plague
us
a
lot
more.
I
haven't
actually
run
into
that
recently.
A
C
A
Because
this
one,
the
one
over
in
the
agent
installer,
this
only
applies.
So
it's
not
the
dot
shaded,
it's
everything
under
java
agent
and
it's
only.
C
B
B
Because
it
uses
that
try
to
efficiently
check
if
the
class
should
be
ignored
or
not,
and
it
contains
all
other
ignores.
C
C
B
I'm
wondering
what
what
packages
do
we
have
in
the
bootstrap
quest
holder?
It's
probably
I
o
and
telemetry
java
agent
bootstrap,
and
what
else.
A
There
were
four,
I
looked
yeah,
I
had
the
same
question
slf
for
j,
okay,
bootstrap
shaded,
and
there
was
one
more
I
have.
I
have
it
handy
here.
B
A
D
B
That's
true,
but
on
the
other
hand,
if
we
drift,
then
the
fact
that
we
don't
have
the
potential
new
package
ignored
here
means
that
we
might
be
able
to
detect
that
drift
if
it
for
some
reason,
starts
to
cause
instrumentation
errors
in
the
tests.
D
A
But,
given
how
yes,
given
how
that
was
a
good
test
nikita
you
that
this
is
confusing
enough
to
not
be
worth
or
to
be
worth
that
risk
of
drift.
A
Just
that
your
feedback
on
how
confusing
this
is,
I
think,
is
a
good
reason
to
go
with
the
simpler
solution
that
mateish
is
proposing.
Oh
you.
A
A
That's
on
this,
so
it
adds
some
complexity
to
the
startup.
D
A
I'd
say
I'm
moderately,
I'm
weakly
in
favor,
just
as
a
I
I
but
yeah.
I
would
really
like
the
smoke
test.
Although
I
mean
we
could
say
it
was
something
that
we
support
as
like
a
best
effort.
Only.
D
I
see
a
point
in
that
like
that's
more
confusing
right,
but
it
is
not
support.
C
A
They
shouldn't,
but
so
the
one
the
customer
who
reported
this,
I
told
them.
You
know
that
once
we
figured
out
what
the
problem
was,
I
told
them.
You
know
that
they'd
have
to
upgrade
and
they
came
back
today
and
said
you
know.
Oh,
does
your
older
release
still
work
with
8u25.
D
Interest,
oh,
you
can
have
my
vote,
I'm
okay
with
it,
but
with
the
bare
minimum
of
this
smoke
test
like
I
don't
this
best
effort
thing
that
sounds
too
weird
to
me
like.
If
we're
gonna
support,
we
should
actually
support
it.
I
mean
we
have
the
test.
If
we
don't
have
the
test,
then
we
don't
support
it.
That's
my
line.
I
think.
A
All
right:
well,
let's
see
what
yeah
laurie
comes
back
with
and
nikita.
If
you
want
to.
A
A
C
A
Thank
you
mateish.
This
was
awesome,
I'm
very
excited
for
nettie
request,
metrics.
C
A
Yeah
it's
so
it's
this
similar,
but
the
strict
context
check
is
even
harder
is
like
a
more
subtle
like
in
this
case.
It's
just
a
matter
of
like
the
controller,
we're
ending
the
response
during
the
controller
span,
but
in
but
yeah
in
that
strict
context
check
case.
A
A
The
context
open
so
like
stay
say
like
the
start,
async
for
async
servlets.
That
context
is
open
and
you
complete,
you
do
the
complete,
but
your
context
is
still
open,
but
you've
sent
your
you've
finalized.
Your
response.
A
So
in
take
this
example
here,
let's
look
at
this
example.
We
do
context
start.
So
this
is
the
serlet
async.
So
we
open
a
an
hotel
context
here
and
then
we
do
context
complete.
This
is
the
servlet
complete
the
response?
So
now
the
response
is
getting
is
closed
back
to
the
client,
but
our
hotel
context
doesn't
end
until
this
line
here
because
we
opened
it.
A
D
A
D
D
D
A
D
D
A
A
Basically
says:
okay,
when
the
current
thread
context
finishes,
go
ahead
and
finish
everything
out,
because
we
actually
in
the
the
one
of
the
apps,
did
unfortunately
some
some
of
that
logic
after
and
I
wanted
to
capture
that
in,
but
I'm
not
sure
if
that
was
still
the
wise
to
capture
that
as
part
of
the
response
time,
because
people
look
at
that
server
metric
as
response
time.
D
A
D
D
A
Yeah
yeah,
the
strict
context
check
topic
is
here
to
stay.
D
D
B
D
F
I
looked
at
the
new
relic
as
a
reference
implementation,
and
it
does
seem
to
kind
of
do
this
circular
array,
so
you
don't
have
to
copy
array
every
time
you
want
to
every
time
you
want
to
insert
a
because
it
gives
you
stuff
like
negative
indexing
and
things
like
that.
But
if
there
is
like,
if
we
just
want
to
go
with
something
slightly
similar
and
not
have
to
include
any
new
relic
code,
then
I'm
I'm
happy
to
to
give
that
a
shot.
And
then
we
can
yeah.
D
F
Make
it
easily
easily
exchangeable
in
like
an
interface
and
just
try
different
implementations.
F
Yeah,
no,
that's
a
good
point.
Yeah
it
is,
it
is
just
a
prototype,
so
I
just
certainly
not
locked
in
but
yeah
we
can.
We
can
just
do
a
simple
implementation
and
then
go
from
there
like
it.
D
F
D
F
It
does
do
a
bit
of
a
bit
of
the
lifting.
It
did
save
me
a
bit
of
time
putting
it
in
because
it
can
do
things
like
you
know
like
the
negative
indexing
and,
and
it
kind
of
does
it
handles
a
bit
of
the
scaling
stuff
yeah.
F
Don't
think
it'd
be
too
hard.
I've
mainly.
D
E
F
I'll
have
a
look
at
it
tomorrow
and
let
you
know
how
I
go
and
in
terms
of
benchmarking
I
haven't
looked
into
the
existing
benchmarking.
Sdk
is
that
is
there
a
kind
of
like
a
set
way
of
doing
that.
D
B
A
Think
so
I
did
look
at
this
and
my
my
best
understanding
is
yeah,
probably
because,
even
though
the
field
that
we're
storing
this
in
is
volatile,
it
didn't
look
like
we
were
always
reading
or
writing
to
that
field
in
order
to
inherit
basically
visibility
for
these,
but.
A
D
A
D
D
D
Yeah,
like
one
thread,
is
just
if
there's,
if
you're
only
doing
I
o,
then
that
thread
is
blocked
waiting
for
any
event.
That
event
could
be
really
an
http.
It
could
be
many
requests
or
it
could
be
like
one
thread
will
be
handling
many
connections,
I
think,
but
the
connections
always
like
once
it
gets
assigned
to
a
thread
and
that
threads,
the
one
that's
listening
for
the
I
o
events
for
that
connection
and
handles
the
cortex.
A
All
right
so
problem
solved
there
we
go.
A
All
right
and
what's
this:
why
is
this
slot.
C
So
I
actually
think
that
until
we
have
second
use
case
for
that,
let
it
be
specific.
D
A
C
C
D
A
All
right,
any
other
old
and
by
old
I
mean
like
more
than
three
days
old,
that
we
need
to
unblock.
A
I
don't
think
so:
yeah
we've
got
a
lot
of
draft
stuff
and
this
one's
yeah
this
one's
gonna
emerge
soon.
I
think
windmill
is
just
finishing
up
some
stuff.
It
was
great
justin
looked
at
it
today.
Thank
you,
justin.