►
From YouTube: 2021-04-14 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
C
B
A
C
B
Honor
exiting
zugler
the
ex-googler.
Maybe
he
knows
the
google
docs
magic
he's
an
anonymous.
D
B
C
B
D
A
C
C
A
A
B
C
A
B
Okay,
yeah
and
this
part
will
only
be
in
the
release
notes
not
in
the
change
log
yeah,
so
I
mean
that's.
Certainly
another
option
is
to
remove
this
all
from
the
changelog.
D
D
A
B
C
We
could
do
that
but
or
we
don't
have
to-
I
don't
know
if
it's
definitely
better,
I
wouldn't
say
so:
instrumentation
users
or
just
users,
users.
B
C
C
A
B
Yep,
I
will
cool.
I
will
split
that
up
tomorrow
by
those
two
groups
and.
C
B
B
B
That's
true,
and
just
as.
C
D
A
B
This
was
the
only
thought
that
I
had
did,
but
then
I
was
looking
and
was
caffeine
was
was
caffeine
already
in
one
zero.
D
B
D
C
B
Right
because
we
swapped
out
the
whole
guava,
the
you,
it
was
the
guava
cash
week,
swapped
out,
okay,
so
that
would
be
my
only
concern.
I
know
that,
like
in
my
in
our
distro,
I
will
definitely
patch
this
in.
If
it's
not
in
one
one
zero.
B
B
I'm
okay
with
it
with
releasing,
as
is
just
I
I
did
want
to
give
you,
though
my
personal
like
for
our
customers.
I
will
be.
A
B
C
Are
you,
is
this
clustered
anything
just
unsolvable?
There's.
D
B
Not
really
you,
you
just
have
to
be
super
careful
on
okay.
So
what.
B
At
so,
the
problem
is
that
so
a
library
can
make
makes
assumptions
when
it's
holding
a
lock
right
that
it's
only
doing
safe
things
under
that
lock
and
the
problem
is,
if
it
triggers
loading
of
a
class
under
that
lock,
then
now
you've
expanded
out
the
number
of
things
that
are
happening
right
instead
of
having
this
very
narrow,
lock
of
only
doing
these
three
things.
B
C
C
C
C
C
B
But
that's
caffeine
is
introducing
one
one
lock
right.
The
question
is
what
other
lock
it
could
be.
B
B
Yeah,
that
weighting
is
basically
another
you're
right,
that's
not
technically
another
lock.
It
could
be
like
some
other
weight
type.
C
A
B
Yeah
I'll
I'll
ask
in
the
slack
yeah,
but
I
think
kind
of
with
is
like
that
class
loader
lock
right.
If
the
caffeine
grabs
the
light
hits
has
its
lock
and
then
triggers
loading,
and
then
it
has.
But
the
other
thread
already
has
that
class
loader
lock
and
then
in
the
transform.
Now
we
hit
that.
B
Okay,
well
one
one
one
class
loader.
I
think,
because
the
one
class
loader
has
the
lock.
B
A
So
we
have
a
customer
who
uses
react,
1080
and
they
want
to
add
some
extra
attributes
to
that
director
native
span
currently
we're
doing
that
inside
different
reactor
like
http,
client
callbacks,
and
it
seems
that
there
is
no
current
span.
That
is
not
the
correct
current
stand
inside
those
codecs,
so
we
have
one
like
tactical
issue.
We
just
have
to
update
our
afternoon
instrumentation
on
for
reactor
instrumentation,
but
all
callbacks
are
called
inside
the
correct
span
context,
but
maybe
there
is
a
bigger
and
more
generic
problem.
Is
that.
A
A
B
Yeah,
so,
okay,
that
makes
sense
what
you're
asking
I
kind
of
guess
that
so
my
thought
there
is,
if
that,
if
we
don't
put
it
into
the
thread,
local
right
into.
B
Span
current
context:
by
doing
the
tactical
thing
everywhere,
our
downstream
I
mean
we
can
give
some
api
to
get
that
from
like
a
netty
connection,
a
you
know,
a
request
object.
You
know
whatever,
wherever
we're
holding
that,
but
it's
not
going
to,
and
we
still
have
to
do
the
tactical
stuff
everywhere
or
we
won't
automatically
correlate
downstream.
B
C
A
A
B
A
B
Yeah
yeah,
it
uses
micro
profile
context.
Propagation.
I
don't
know
how
do
I
share
here?
Oh,
I
know
just
in
the
dark.
A
Yeah
thanks,
okay,
so.
C
The
afternoon
specifically,
I
wonder
if
there's
a
way
to
I
mean
I
don't
know
what
the
implementation
is,
but
since
we
instrument
reactor
like
maybe
it's
something
yeah
like
in
their
internals,
maybe
they're
not
calling
these
callbacks
from
a
reactor
callback,
but
they
could
be
or
something
like
that,
because
I
feel
as
if
instrumenting
reactor,
not
solving.
The
problem
feels
weird
to
me.
A
A
A
C
Do
we
consider
our
job
to
instrument
every
single
callback
possible
or
we
make
calls?
Because
I
mean,
of
course
you
have
your
customer
use
case.
There
must
be
a
use
case,
but
I
on
the
flip
side,
if
there's
some
api,
that's
really
just
like
a
monitoring
api
like
their
callbacks
just
to
they're
not
intended
to
trigger
downstream
calls
because
they're
just
monitoring,
then
maybe
those
aren't,
even
in
our
scope
of
what
we
instrument
because
they're
not
they're
more
of
a
maybe
a
platform,
because
these
do
unresolved
appraisals.
A
C
Reason
they
want
that
so
to
take
a
very
narrow
look.
What
we
could
say
in
this
case
is,
we
should
just
have
our
instrumentation
at
these
events,
anyways,
because
it
seems
useful
let's,
but
that
doesn't
mean
we
have
to
propagate
the
context
into
these
callbacks.
If
it
seems
like
such
a
small
use
case.
A
B
A
C
B
C
A
B
B
In
this
case,
nikita
do
you
know
like
because
if
we're
only
instrumenting,
you
know
if
they
provide
this,
then
I
think
that's,
okay.
You
know
like
if
there's
no
resolve
handler
on
resolve
handler,
we
wouldn't
instrument
anything
and
so.
A
I
I'm
not
sure
about
instrumenting,
but
we
certainly
will
not
like
make
contacts.
A
D
A
A
A
Yeah,
okay,
did
we
try
to
to
ask
raphael
about
type
initializer
thingy?
I
understand
it's
like
the
problem
is
that
currently
red
body
doesn't
work
with
type
initializer
but
yeah.
I
think
it
like
looks
like
it
should.
B
Yeah
am
I
wrong,
I
think
you
know.
I
think
we
need
to
spend
a
little
bit
more
time,
trying
before
asking.
D
A
D
C
B
B
A
B
Yeah,
I
think
I
agree
with
what
matters
is
saying
that,
like
I
mean
the
one,
that
the
name
that
we
sort
of
advertise,
as
that
we
recommend
people
say,
should
be
the
one
without
the
version.
A
B
This
is
slightly
different.
This
one
is
the
name
that
you
use
to
suppress
instrumentation
with,
whereas
the
okay,
the
instrumentation
name,
that
gets
reported
as
the
library
info.
Okay,.
A
A
In
order
in
order
in
order
to
suppress,
we
shouldn't
need
that
version
specific
key,
because
the
idea
is
that
in
the
in
the
one
application
in
the
every
point
in
time
there
should
be
only
one
version.
So
I
want
to
suppress
that.
That's
it
yeah
in
library
name,
we
need
version
yet
that
may
be
useful,
but
in
order
to
suppress
no
just
just
name
it.
That's
a
very
good
point.
A
A
A
Probably
we
we
even
we
can't
even
have
it
that
way,
so
we
we
we
are
going
to
have
anyway,
some
documentation
which
says
this
module
has
this
id.
If
you
want
to
disable,
and
that
should
mention
name
without
version-
that's
okay.
At
the
same
time,
in
the
troubleshooting
guide,
we
can
say
that,
in
order
to
disable
this
specific
span,
you
can
take
a
look
at
the
instrumentation
library
name
and
use
that
as.
A
Well,
so
our
documentation
still
talk
about
only
ideas
which
is
like
couch
base
and
that's
it,
but
the
viewer
goes
to
troubleshooting
guy
guy.
He
they.
They
can't
themselves
find
the
correct
version,
specific
idea
as
well.
We
don't
we
don't
list
it
anywhere,
but
they
know
how
to
find
it
in
order
to
manually
disable.
It.
B
Me
cool
I,
if
nobody
else
does,
I
will
post
to
the
issue
tomorrow
with
these
notes.
A
Okay,
we
no,
we
don't
want.
No,
we
don't
want
to
merge
this
one
because
we
don't
want
to
list
additional
names.
I
believe.
B
B
B
But
I
did
have
you
know
kind
of
my
I'm
not
quite
sure
like
the
for
me
that
I
wouldn't
mind
going
to
the
other
repo
to
do
doc
stuff,
but
I'm
also
used
to
at
microsoft.
We
have
a
different
docs
repo
and
I'm
always
going
there.
So
that's
kind
of
just
like
a
work
flow.
B
C
B
C
B
B
Yes,
and
this
one
is,
this-
is
the
box
and
for
docs
we
decided
yes
to
the
same
process
as
fdk
pros
include
updating
the
same
pr's
as.
C
I
would
also
suggest
I
might
drive
the
talk
about
the
instrumentation
api
in
the
context
of
the
dock,
because
it's
not
just
java
that
uses
instrumentation
api,
so
I'll
probably
do
something
at
some
point.
Okay,
if
I
could
finish
this
damn
lambda
stuff
but
yeah,
but
not
dislike,
so
I'm
happy
to
leave
all
the
distribution
stuff
to
someone
else
and
then
I
think
one.
B
I
answer
your
question
here
nikita
yes,
I
did.
I
didn't
realize
nine
days
I
looked
at
it
like
nine
days
ago
today.
So
yes,
I
do
plan
to
do
this.
Okay,
few
more
days.
A
B
A
Then
maybe
if
we
get,
if
we
have
a
couple
of
minutes,
can
we
talk
a
little
bit
about
that
extension?
Loading.
A
One.
Do
we,
so
if
user
provides
directory
of
jars,
the
folder
of
jars,
do
we
expect
that
every
jar
is
the
self-contained
thing
which
contains
all
transitive
dependency
as
well.
B
B
I
mean
I
like
the
idea
of
isolating
them,
but
that
also
is
not
as
simple
of
a
deployment.
B
C
A
If
we
start
supporting
putting
things
into
user
class
part
build
style,
then
certainly,
then
slim
jars
are
better,
but
if
this
is
deploy
time
dependency,
essentially
not
not
build
time
so
end
user
will
download
them
manually,
not
via
gradle,
then
in
most
cases,
so
fat
jars
are
easy
flat.
Jars
are
more
complicated
for
developers
of
those
flat
jars,
but
just
side
jars
without
shape
without
shading
just
using
shadow
plugin.
It's
it's.
It's
easy.
C
Yeah
and
safer
so
like
it
might
be
a
bit
harder
for
those
authors
to
write
the
thing
by
shadowing,
but
then
they
won't
have
the
support
burden
of
having
to
deal
with
weird
conflicts
within
our
extension
class
loader.
So
I
think,
even
for
the
author,
it's
a
better
experience.
If
you
don't
have
to
worry
about
this.
C
A
A
Wait:
instrumentation
api.
We
should
yeah
instrumentation
api
extension
extension
jar.
We
will
certainly
contain
extension
api.
B
Yeah,
so
if.
C
B
A
B
So
if
we
didn't,
if
we
weren't
trying
to
isolate
the
extensions
from
each
other,
we
could
do
something
like
this,
which
would
be
pretty
simple
where
this
these
are:
parents,
the
this
parents,
the
bootstrap
extension
parents
this.
So
it
only
has
access
to
these
things
and
then
our
agent
class
loader
parents,
the
extension
class
loader.
So
it
has
everything
what
what
are
they
doing.
B
Right
so
what
I'm
saying
here,
but
if
you
make
a
if
you
split
agent
class
loader
with
the
sdk
up
here
and
then
agent
class
loader
down
here
this
way,
but
yeah
this
way.
This
way
extension
class
loader
only
has
access
to
stuff
up
here,
so
extension
class
loader,
can't
access,
say
neti
jrpc,
but
that
doesn't
help
with
isolation,
because
if
this,
if
this
brings
nettie
and
grpc
here,
we're
hosed
anyway,
we
can't
there
can't
be
a
parent
child
relationship
between
these.
B
D
A
A
A
B
For
this
to
be
isolated,
like
this
has
to
be
off
to
the
side,
and
but
the
we're
gonna
have
to
register
the
the
spis
right,
all
the
basically
the
so
this
would
be
like
off
to
the
side.
B
Because
they
are
going
to
in
general,
they
won't
find
that
on
their
own.
We
would
have
to
do
something
special
for
that.
Okay
and
then
one
one
other
thing
to
let
you
know
about.
If
you
haven't
seen
the
if
you,
if
we
do
want
to
split
our
class
loaders
sorted
into
two
right,
we
have
that
one
inst.