►
From YouTube: 2021-03-03 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
A
B
B
B
B
C
C
A
B
B
B
B
Okay,
so
two
things
I
had
wanted
to
talk
about:
one
is
library
instrumentation
releasing
what
does
it
mean
to
release
or
how
can
we
get
to
the
point
of
releasing
library
instrumentation
as
non-alpha.
B
So
we
brought
it,
it
came
up
in
the
or
john
brought
it
up
in
the
maintainers
meeting,
and
so
one
of
the
things
was
they
said
it
was
the
kind
of
we're
like
okay,
fine,
you
can
release
the
java
agent,
but
they
we're
not
thrilled
about.
I
guess
I
thought
there
had
been
a
decision,
but
clearly
I
was
wrong
about
telemetry
stability
being
required
or
or
not
required
for
one
zero.
C
B
So
I
think
so
tigran
seemed
like
he
was
going
to
push
forward
with
this,
but
it
sounded
like
potentially
a
couple
of
you
know
it
could
take
some
time.
Obviously,
so
I
don't
know,
if
are
you
okay
with
that
anrag?
I
know
you
want
to
put
out.
C
A
C
B
B
B
A
A
So
I
mean
we,
I
mean.
I
also
think
that
there
are
two
users
or
like
two
uses
of
every
library,
of
the
information
that
we
release.
One
is
actual
users
who
touch
their
application
code
and
another
is
well
telemetry,
backend
and,
as
was
said
during
maintainers
meetings,
nobody
never
before
ever
ever
guaranteed
stability
of
the
telemetry
data.
A
Actually
no
single
project
said
that,
yes,
we
will
not
make
incompatible
changes
with
like
telemetry
data,
so
maybe
it's
okay,
that
we
continue
to
do
so
for
for
some
time,
while
still
saying
users,
you
can
instrument
your
connection,
pools
with
this
library
or
your
awsa
clients
with
this
library,
and
you
will
not
run
into
dependency
dependency,
hell
or
breaking
changes
due
during
upgrades,
because
that's
what
means
guarantee
for
the
end?
Users.
A
B
Well,
let's,
let's
push
it
next
week
in
the
maintainers
meeting
again
say
that
you
know
that's,
you
know
we
would
we.
That
would
be
our
our
proposal.
A
B
I
have
a
question:
does
the
hotel
java
agent
extension
or
the
hotel,
essentially
the
hotel
java
agent
initializer
stuff,
gets
you
what
you
want
without
having
to.
A
A
B
You
can
I
mean,
depending
on
what
you
want
to
do
right,
you
can
do
a
lot
with.
What's
in
the
the
auto
configure.
B
B
It
is
yeah
you're
right,
but
it's
not
in
the
java,
it's
not
in
the
java
agent
just
and
it's
yeah.
Therefore,
it's
probably
more
stable
than
whatever
we
have.
C
C
A
C
B
Yeah
I
feel
like
this
is
a
really
great
combination
for
a
lot
of
stuff,
so
I'd
like
to
kind
of
go
down
that
road
and
turn
this
into
like
do
what
I
think
it
josh
streth's
idea
was
great
about
the
class
loader
isolation
for
loading,
the
initializer
jar,
so
that
it
wouldn't
have
any
conflicts
with
anything
else
that
we
have
in
the
agent
class
loader.
B
We
would
only
share
the
open,
telemetry
api
and
sdk
through
to
that
class
loader,
so
that
would
give
us
a
pretty
tight
contract
around
interop.
B
A
C
B
C
C
B
That's
a
good
point:
the
initialize
the
exporter
is,
but
the
initializer
is
very
similar
to
this.
So
we
could.
This
could
be
an
alias.
A
A
B
B
And
then
the
other
thing
that's
different,
the
exporter
jar,
I'm
just
looking
at
the
code
now
because
I
forgot
we
are
using
the
java
agent
spis
instead
of
the
the
auto
configure.
B
A
Yeah
that
I
can
understand
yeah,
but
my
logic
is
still
we
we
want
to
provide
the
functionality
that
you
can
load
a
separate
jar
into
agent,
isolate
it
and
install
any
component
that
you
find
inside
that
jar.
A
B
No
because
the
old
exporter
jars
used
a
special
java
agent,
spis.
B
B
A
B
A
A
So
answering
your
question,
my
my
thought
is
that
when
we
release
1.0
of
java
agent,
we
have
agreed
that
we
should
very
carefully
and
explicitly
list
all
the
things
that
we
declare
stable,
which
means
that
we
can
not
declare
stable
exporter,
jar
and
initializer
jar.
So
we
can
just
throw
them
away
in
some
point
in
time.
B
Yep,
I
I
I
like
this
yeah
I'll
I'll.
Do
that
and
then
yeah,
hopefully
I'm
I
I'll
work
on
this
maybe
next
week.
I
don't
think
it
should
be
a
big
deal
and
it
would
be
cool.
B
I
just
didn't
want
to
yeah.
I
just
wanted
to
make
it
really
clear
beyond
just
our
list,
also
like
cool
okay,
so
that
I'm
good
with
that
so
I'll
make
that
I'll.
Do
I
can
do
that
rename
tomorrow
I
mean
unless
yeah
I
mean,
and
we
can
still-
we
should
still
be
able
to
release
this
week.
A
A
B
Oh
yeah,
so
what
I
would
I
propose
would
propose
is
I'll
submit
a
pr
with
what
onrag
in
the
java
repo
was
there
a
convention
for
how
you
were
documenting?
There
was
a
release.
C
B
Okay,
so
I
will,
I
will
submit
this
tomorrow
and
then
we
can
review
on
thursday
slash
underogs
friday,
and
then
we
can
release
sound
like
a
plan.
Yep.
B
B
Currently
they
depend
on
right,
the
instrumentation
api
artifact
and
since
that's
alpha,
my
thought
was
and
andrea.
Let
me
know
if
this
is
my
understanding
from
what
the
how
the
sdk
repo
was
doing.
This
is
not
having
like
you
can't
have
non-alpha
artifacts
depend
on
alpha
artifacts,
so.
C
C
C
C
C
C
B
C
B
Okay,
okay,
so
the
idea
is
that
as
long
as
it's
impul
implementation,
not
api
users
would
have
to.
They
won't
get.
C
B
Okay,
yeah,
I
I'm
happy
to
take
your
recommendation.
B
Nikita.
Do
you
have
any
preference.
C
B
Think
that
the
idea
is
that
people
had
like
the
way
I've
seen
the
sdk
repo
navigate.
This
is
tell
people
you
have
to
use
the
bomb
and
then,
as
long
as
you
use
the
bomb,
then
you're
guaranteed
to
since.
A
C
A
C
A
C
A
Okay,
in
that
case,
it's
a
little
bit
less
of
all
of
a
problem.
Okay,.
A
C
B
C
C
B
C
C
About
this
caching
stuff
for
some
reason
yesterday,
I
was
playing
around
with
it
before
bringing
in
that
concurrent
link,
hash
map,
or
something
like
that.
I
thought
I'd
just
see
what
happens
if
I
depend
on
caffeine
and
use
shadow
jars,
minimize
thing
to
only
get
the
classes
that
we
use
and
when
I
tried
that
instead
of
the
one
meg,
which
is
the
normal
jar,
it
was
cut
down
to
around
350k
without.
B
C
B
C
A
B
C
B
C
B
Yeah,
definitely
I
mean
definitely
with
that
size.
If
that's
I
mean
I've
had
kind
of
trouble
in
the
past,
with
or
at
least
it's
been
a
little
bit
of
a
nuisance
in
the
past,
using
those
minimized
things,
but
that's
more
been
like
for
a
big,
bigger,
yep.
C
B
Yeah
and
also
it's
a
probably
not
a
lot
of
reflection
and
weird
stuff
going
on
in
there.
C
It
automatically
generates
like
100
different
types
of
implementations
for
different
types
of
caches
and
then
reflectively
loads,
the
one
that
was
configured
by
the
user,
so
it
actually
does
have
that
but,
like
I
found
it
very
easily
by
just
writing
a
unit
test,
I'm
pretty
confident
I
would
find
those.
So
I'm
not
too
worried.
C
B
Yep
yep,
I
I
think
we
probably
have
some.
If
we
really
wanted
to
save
a
few
hundred
k
here
and
there
I'm
sure,
there's
places
we
could
attack.
C
Have
you
ever
thought
about
the
idea
of,
if
not
even
like,
we
could
have
a
slim
jar
or
not
but
like
using
grpc,
okay
gtp
instead
of
an
idi
like,
I
think
that
would
cut
down
on
a
lot
of
startup
overhead
and
jar
size
and
might
reduce
throughput
expands
or
might
not?
I
think
the
performance
of
location
versus
an
idea
is
not
a
well-defined
thing,
there's
an
assumption
that
it
has
better
performance,
but
in
practice
I
think
there'd
be
similar.
C
At
some
point,
we
probably
need
better
benchmarks.
First.
B
Yeah,
what
in
the
sdk?
C
C
It
was
really
just
a
random
comment,
like
I
was
really
happy
with
this
rocket
mqpr.
I
could
just
convert
it
into
library,
instrumentation
right
in
front
of
my
eyes
like
I
was
like
well,
this
actually
works,
like
you.
B
Yeah
yeah
yep
and
yes,
and
also
think
thankful
for
the
number
of
different
reviewers
that
we
have,
but
not.
B
B
B
Cool,
I
will
just
leave
it
like
this
and
let
you
all
once
matesh
is
online
just
pick
something
to
merge.
B
C
C
I
wonder
if
they
tweaked
something
it
was
just
so
interesting
to
see
like
it
never
happened,
and
then
all
of
a
sudden
since
yesterday,
like
the
test
order,
changed
for
whatever
reason.
Maybe
the
test,
maybe
java,
just
based
the
test
on
the
month
of
the
year
or
something
I
don't
know
so
if
it
flipped
to
march
and
now
the
test
order
is
different,
I
don't
know,
but
that
started
failing,
because
that
was
really
surprising.
B
Well,
you
all
are
here:
what
do
you
think
about
this?
So
for
cassandra
for
db
sequel,
table
we're
stripping
off
the
key
space.
B
And
frank
had
done
that
originally,
because
key
space
is
already
captured
as
a
separate
tag,
the
which
I
don't
have
any
issue
with
that.
Just
that
I
mean
it's
not
what
the
spec
says,
though,
for
db
sequel
table.
A
B
No,
no,
the
dot,
db.sequel.table,
and
so
actually
maybe
this
is.
The
answer
is
that
cassandra,
maybe
shouldn't
even
be
capturing.
C
B
B
B
Okay,
I
I
can
check
on
that.
Maybe
maybe
it
is
capturing
it
as
tv
cassandra,
but
I
can
figure
that
out
and
reply
there.
B
B
All
right
anything
else.
B
Cool
have
a
good,
have
a
good
wednesday
and
see
you
soon.
Good
night
see
you.