►
From YouTube: 2021-05-28 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
B
A
A
B
C
We
had
another
quiet
morning
meeting
just
me
john
and
tyler.
B
B
Have
a
lot
of
direct
reports?
Well,
so
none
of
them
are
official
direct
reports
he's
just
kind
of
organizing
steve's
behalf.
B
C
Hey,
who
knows,
what's
what
started
the.
C
Seen
yet
some
don't
watch
how.
C
How
to
resolve
disputes
maintainer
disputes.
B
A
C
Yeah
I
thought
about
I
well,
I
knew
it
wasn't
the
the
java
repo
because
it
it's
all
about
if
there's
ties,
if
the,
if.
B
C
C
Yet
yeah
yeah.
B
So,
speaking
of
not
that
really
at
all,
but
maybe
adjacent,
we
need
people
to
work
on
metrics.
B
Well,
yeah,
because
so
there
they
have
it.
There's
the
new
api,
which
mostly
just
looks
like
some
renames
of
the
existing
api,
which
is
not
a
big
deal,
but
the
sdk
is
actually
there's
a
whole
bunch
of
new
features
that
are
going
to
be
required
of
the
sdk
that
we're
going
to
have
to
somebody
needs
to
figure
out
how
to
make
them
work.
B
So,
for
example,
I'm
going
to
need
to
be
able
to
have
the
ability
to
have
both
a
push.
Pull
export
for
metrics
in
the
same
sdk
running
at
the
same
time
also
need
to
be
able
to
have
two
push
exporters
running
at
different
intervals,
with
different
metrics
assigned
to
different
exporters.
B
So
anyway,
the
there
was
a
question
during
the
api
sdk
insignia
this
afternoon,
like
what's
the
status
of
java
how's
the
prototype
going
like
we
haven't,
started
and
there's
no
plan.
So
I
said
I
would
bring
it
up
at
the
same
time
and
talk
at
least
about
what
we
should
plan.
A
B
Was
originally
why
he
had
to
pull
out-
and
I
remember
yeah,
so
my
guess
is
that
once
he's
out
of
the
collector,
he
will
want
to
come
back
and
work
on
metrics.
So
I
haven't
been
very
enthusiastically
starting
to
work
on
metrics,
because
I
know
he
will
come
in
and
come
through
the
project
like
a
freight
train
when
he's
as
he
always
does
so
anyway.
I
don't
know
what
the
plan
should
be,
but
I
said
I
would
at
least
bring
it
up
with
you
and
we
can.
B
I
also
bring
it
up.
So
I
don't
know
if
it's
something
you
have
interested
in
interested.
I
also
are
trying
to
recruit
jonathan
to
come
and
help
us
out
and
see
if
he
knows
metrics
and
knows
java.
A
A
B
B
A
B
The
the
different
exporters
for
different
metrics,
though,
is
a
little
bit
different,
because
then
you
just
not
just
sending
stuff
to
both.
That's
you
have
to
choose
where
it
should
go
depending
on
criteria,
and
I
don't.
I
don't
know
that
even
what
that
api
looks
like
or
how
that
works.
Probably
later,
did
you
do
the
view
api.
B
So
I
think
they're
they're
planning
on
doing
a
feature
freeze
so
they're
trying
to.
I
think
I
think
riley
is
trying
to
cut
a
experimental
not
quite
frozen,
but
like
a
release
candidate,
essentially
for
the
api
by
the
end
of
june
and
then
end
of
july.
I
think
the
plan
is
to
he
has
the
schedule,
I'm
just
trying
to
remember
that
off
top
ahead,
but
then,
by
the
end
of
july
I
have
a
have
a
real
feature.
Freeze
on
the
api.
B
Well,
I
think
riley
is
trying
to
be
realistic,
given
how
long
it
takes
anything
to
get
approved
in
the
spec
he's
trying
not
to
over
promise.
At
this
point,
given
you
know,
just
even
getting
people
to
review
the
prs
is
is
difficult
actually
well,
let
me
go
and
see
if
I
can
find
his.
B
He
posted
his
schedule
of
the
notes.
Today,
let
me
go
look.
B
A
B
A
B
Yeah,
the
accounting
is
a
little
funny
there,
but
yeah
where
I
thought
his
schedule
was.
He
did
not
posted
here.
Let
me
see
if
I
can
find
it
here.
It
is
a
google
doc.
I
will
send
put
the
link
in
the
chat
or
maybe
I'll
put
it
in
our
notes
since
we're
here
nope.
That's.
A
What
do
you
mean
by
language
specific,
in
the
sense
that
if
one
language
like
go
is
already
prototyping
java
could
still
come
in
when
it's
in
a
more.
B
B
So
end
of
may
experimental,
metrics
api
spec,
that's
not
going
to
happen.
I
mean
we're
not
going
to
get
there
in
the
next
week.
I
don't
think
that's
possible.
B
A
Yeah,
I'm
so
wonder
if
this
my
manager
might
have
seen
the
stock
somehow,
because
I,
when
you
mention
the
metrics,
they
pay
they're,
trying
to
make
stables.
I
assume
that
went
like
a
month,
and
so
I
said
it's
going
to
be
stable
soon
and
he's
like.
Isn't
it
supposed
to
be
september
or
something
then
it
looks
like.
B
That
is
the
case
yeah.
This
is
the
this
basically
is
moving
the
metrics
api
stability
up
a
month
than
it
was
before
I
kind
of
was
like.
Can
we
stabilize
api
before
the
sdks
yeah?
So
this
is
that
this
is
that
plan
okay
anyway
stable
at
the
end
of
august
of
the
api
and
feature
freeze,
sdk,
but
then.
B
Otherwise,
yeah,
I
think
what
we
have
right
now
is
just
essentially
what's
there
with
some
renames
and
maybe
a
couple
deletes,
because
I
think
the
batch
api
is
being
removed
from
the
first
cut,
so
we'll
be
able
to
remove
the
batching
one.
I
think
what
we
really
need,
though,
is
probably
a
plan
of
how
we
want
to
build
the
new
one
like
do.
We
want
to
try
to
morph
what
we
have
right
now
in
place.
B
We
call
it
n
metrics
for
new
or
non-blocking
matrix
like
what
do
we
call
our
metrics
right
now.
It's
called
api
metrics
right,
so
we
need
now
like
api
and
metrics
or
new.
C
B
A
B
B
Right,
that's
what
it's
just,
I
think,
there's
a
there's
a
skeleton
of
the
sdk
specification
at
the
moment,
right,
yeah,
yeah,
that
I
think
riley
is
hoping
by
the
end
of
june
to
have
that
flushed
out,
but
there
are
what
so
what
there
is
in
the
api
or
actually
there's
an
otep.
I
don't
remember
the
number
oh
tab,
146
or
something.
B
Where
is
the
number
here?
148
is
o
type
146
hey,
I
remember,
and
otep
146
yeah
has
basically
a
whole
bunch
of.
B
B
A
We
have
a
great
okay,
I
see.
A
I
don't
think
I'm
interested
in
developing
something
from
scratch
matrix
if
there's
no
space,
that's
kind
of
the
way
I
mean
because
especially
because,
like
I
mean
I
have
so
little
visibility
into
the
spec
discussion
themselves,
yeah,
it's
just
hard.
So
I
definitely
like
I.
I
didn't
agree
to
prototype,
so
I
said
josh
agreed
to
prototype
and
you
can
delete
java
from.
B
Yeah,
I
don't
know
anyway.
I
think
that,
because
josh,
the
other
josh
was
very
busy
working
on
the
data
model,
he
said
he
was
not
going
to
be
available
to
do
that.
B
A
B
And
so,
if
there's
a,
if
there's
a
vendor,
who
feels
it's
important,
they
should
step
forward
and
give
us
some
resources.
It
sounded
like
well,
it
sounded
like
riley
was
gonna
ask
around
github,
but
that
was
not.
There
was
no
java.
That
was
gonna,
be
ruby
and
go
not
java,
and
somebody
was
going
to
ask
somewhere
else,
but
I
don't
remember
where
it
was
oh
well.
I
and
I
asked
jonathan
that
he
was
interested
in
helping.
So
let's
see
what
happens
but
I'll
just
bring
it
up
in
the
maintainers
meeting.
B
A
B
Yeah,
I
think
it's
I
probably
can
work
on
that
like
do
with
somebody
named
some
renaming
and
some
deprecation
of
the
old
names.
I
think
so
it's
going
to
be
a
little.
If
we
just
I
mean
the
easiest
thing
would
just
be:
do
some
hard
renaming
and
some
deleting
and
stuff.
If
we
do
want
to
go
through
the
deprecation
cycle,
it's
going
to
be
a
lot
more
painful
because
I'll
have
to
have
duplicate
entries
for
everything
and
all
of
the
registries
and
et
cetera,
et
cetera,
so
have.
A
B
So
all
right!
Well,
I
will,
I
will
I'll,
take
the
action
and
try
to
do
the
messaging.
I'm
also
going
to
reach
out
to
bogdan
and
just
be
like
hey.
Do
you
have
time
to
do
this
work?
Because
I
guess
the
answer
is
no
yep,
so
but
I'll
I'll
do
my
due
diligence
and
ask
at
least.
B
Yeah,
I
don't
know
what
the
status
is
on
that
at
all.
I
mean
I
thought
I
mean
this
was
many
many
months
ago,
but
there
was
a
whole
bunch
of
work
that
was
going
to
have
to
be
done
to
make
prometheus
like
the
prometheus
scraper
in
the
collector
work
right
or
I
don't
know
whether
there's
a
prometheus
scraper
or
exposing
the
prometheus
data
to
a
scraper
or
both
might
have
been
both
anyway.
I
don't
know,
I
don't
know
if
that
is
something
that's
also
being
worked
on.
B
A
B
Yeah
any
any
more
thoughts
on
that
the
allocation
cost
of
recording
an
exception.
A
Yeah
I
mean
again
like
I'm
always
of
the.
If
we
can
make
it
even
a
little
bit
better.
We
should
because
it's
a
core
component
and
people
don't
want
instrumentation
to
do
bad
stuff,
so
I
mean
yeah.
I
don't
know
if
you
remember
my
original
optimized
thing
that
was
using
the
it
was
using
the
red
local
buffer
and
it
does
that
and
the
allocation
cost
is
are
reduced.
So
why
won't
that?
Just
like
save
one
allocation,
though.
A
A
B
Yeah
nikita
miscalculated
his
nanoseconds
twice
today,
yeah
so
yeah,
I'm
sure
he
was
I'm
sure
he
was
overreacting
right
I
mean
12
microseconds
seems
fairly
reasonable
to
me
for
doing
the
copy
of
a
stack
trace,
although
in
his
in
his
benchmark
it's
all
like
a
very
short
stack
trace,
but
it
might
be
interesting
to
exactly
generate
bring
generated
spring
stack,
trace,
yep
and
see
how
long
that
takes
to
actually
copy
over.
A
Yeah
yeah,
so
if
there's
something
we
can
do,
that's
probably
I
mean
that's.
These
sort
of
sdk
is
the
ones
where
go
all
out
and
micro
optimizations
I
mean
every
time
I
mean
I
think
we
buzzed
on
hector
news.
I
don't
know
if
that
quotes
coincides
with
the
timing
of
nikita
or
not,
but
any
time
he
wasn't
hacked
because
then
people
like.
B
A
A
C
A
B
A
A
B
So
we
you
have
one
that
says:
instrumentation
you've
got
one
that
says
auto
instrumentation
agent
or
something
like
that.
Like.
A
A
A
B
That
you
could
yep
do
these.
A
Yeah,
the
zipkin
has
its
rule
that
they
don't
want
to
do
supporting
issues.
Issues
should
just
be
about
development,
because
that's
why
we
did
this.
We
don't
have
that
rule,
but
we
can
still
have
links
in
our
thing
anyways
just
like.
If
you
want
to
do
something
more
casual,
like
casual
chat
or
something
we
can
have.
A
C
And
we
did
google
this
morning
and
there
is
some
the
github
roadmap,
like
that.
We
found
an
issue
about
adding
templates
to
discussions
and
github
people
said
it's
on
their
roadmap.
B
A
B
B
Completely
unrelated
to
this
discussion,
just
while
it's
on
top
of
my
head
there's
at
least
one
more
open
telemetry
person
in
portland-
that's
victor
victor
liu
from
I
think
he's
from
microsoft.
C
B
A
C
C
Anurag,
I
just
wanted
to
thank
you
for
the
who
this
this.
Yes,
I
was
gonna.
C
C
A
C
Yes,
so,
let's,
if
it's
not
merged
by
morning,
I
will
merge
it
yeah.
A
C
I
know
we
have
some
weird
still.
There
was
a
another
one
of
those
where
gradle
didn't
wasn't
running
some
test,
that
laurie.
A
A
C
A
And
then
maybe
it's
better,
like
I'm
thinking
of
migrating,
the
instrumentation
gradle
files
to
kaufman,
because
there's
lots
of
them
and
they're
very
small
and
you
can
sort
of
see
what
that
does
to
performance
of
the
build.
I
mean,
theoretically,
it's
not
supposed
to
rebuild
them
that
much
so
it
must
be
okay,
but
while
trying
it
out,
we
might
find
it
has
to
recompile
more
often
than
we
like
a
cotton
recompiling.
The
script
is
much
slower
than
groovy,
but
it.
A
C
As
like,
as
a
scripting
language,.
A
A
A
C
A
B
A
B
A
I
don't
know
if
that,
like,
I
know
it
resolves
dependencies
during
the
build,
not
during
configuration,
so
I
don't
know
if
the
number
of
dependencies
affects
configuration
it
might,
and
that
could
be
one
of
the
reasons
why
it's
so
slow,
but
I
don't
think
like
it's
supposed
to
only
resolve
the
builds
dependencies,
not
the
project's
dependencies
during
configuration,
yeah
anyways.
So
that's
ever
since
laurie's
sent
that
thing
about
configuration
on
demand.
I've
been
trying
to
get
her
like
it's
what
it
was
eight
seconds
at
that
time.
A
A
Maybe
no
because
that's
griddle
just
running
maybe.
C
B
A
C
C
Yeah
I
was,
I
was
in
favor
of
it.
I
mean
I
still
am
in
favor
of
it,
for
the
reasons
primarily
for
the
reason
that
it,
I
was
convinced
it's
impossible
to
instrument.
A
C
So
at
that
point
you
know
I'd
rather
throw
it.
You
know
at
least
turn
it
off
by
default.
A
Yeah
at
some
point,
someone
probably
just
has
to
make
that
a
big
project
yeah,
it
means
re-instrumenting.
Some
of
the
libraries
I
mean
rat
package,
I'm
interested
in.
I
want
to
introduce
library
instrumentation,
because
I
feel
as
if
broader
instrumentation
is
much
more
invasive
than
it
needs
to
be
right
does
have
some
looks
like
even
for
context
propagation,
I
think,
like
one
of
the
developers
for
rat
pack
was
also
sort
of
a
zip
code
member.
A
A
C
A
Performance
we're
getting
a
lot
of
these
issues
about
the
web,
flux,
performance
being
bad.
A
C
I
feel
like
we
need.
We
need
a
repro
or
yeah.
A
It's
hard,
at
least
that
would
allow
some
examination
I
mean
half
the
performance
is
pretty
unexpected
only
like
it's
either
a
bug.
So
that's
what
like,
if
we
knew,
which
library,
maybe
we
can
examine
to
try
to
find
some
bug
if
it's
not
a
bug
like,
for
example,
a
bug
might
be
blocking
so
that
it
blocks
the
vent
loop
or
something
that
I
don't
know
that's
possible.
C
Yeah,
I
would
I
mean
I
like.
A
It
good
suggestions,
if,
if
that
x,
that
we
find
is
reactor,
that
would
definitely
be
a
promising
information
to
know
about,
because
then
we
do
know
that
the
performance
of
that
is
pretty
bad.
So
we
could
point
that
to
the
reactor
team
also
and
then
make
some
progress.
I
think
sure
in
the
meantime,.
C
Yeah,
if
we
just
rule
out
the
exporter
and
even
the
you
know,
sampling
at
zero
percent.
A
C
Because,
like
my
first
thought
is
memory
like
like,
if
you
drive
like,
if
you're
trying
to
do
those
throughput
benchmarks
and
you're
just
saturating
your
app?
If
we
have
you
know
bad
effect
on
allocations
that
can
really
drive
that
kind
of
a
big
difference.
A
C
But
I
would
expect
you
know
a
lot
of
the
allocations
to
be
done
in
the
exporter
in
the
two-span
data.
Part
of
the
exporter,
like.
B
B
Yep,
I
mean
I
mean
unless
you
turn
keyboard
on
and
debug
will
turn
on
the
logging
one
but
yeah
it's
not
supposed.
A
B
C
A
B
B
I'm
sure
it's
been
a
long
time,
but
the
two
span
data
is
actually
should
be
a
lightweight
thing
and
mostly
it's
just
a
wrapper
around
the
the
existing
area.
B
B
Yeah,
I'm
sure
there's
a
lot
of
allocations
in
the
conversion
protocol.
This
is
why
I
mean
I
have
had
in
the
back
of
my
head.
I
think
I've
mentioned
this
before
the
idea
is
an
otlp
native
sdk,
oh
yeah,
I
try.
I
actually
tried
to
do
that
and
the
otlp
does
not
like
it.
You
can't
do
it
otlp,
like
you,
it's
just
the
the
data
structures.
Do
not
work
with
the
read
write
span
like
you,
just
would
have
to
bridge
would
be
very
awkward
and
very
strange
like
it.
Just
does
not
look
like
the
like.
B
I
think
well,
I
made
a
I
made
a
a
spend
45
minutes
trying
it.
I
don't
know
six
months
ago
and
quickly
said
this
is
going
to
be
way
more
painful.
A
B
Weird
to
hold
on
to
some
the
the
over
the
over
writing
stuff,
so
I
anyway,
I
was
like
yeah.
I'm
not
gonna,
spend
time
on
this,
but
yeah
there
probably
would
be
a
clever
way
to
do
it.
Well,
especially,
there
might
be
a
clever
way
to
do
it
well
did
bogdan.
What
did
we
do
with
his
like
write
directly
to
the
like,
avoid
the
pro
objects
right.
A
A
My
idea
I
had
was
adding
a
experimental
underscore
with
tlp
exporter,
so
that
at
least
would
be
such
too
and
to
get
some
live
data
out
there.
If
we're
not
confident,
that's
one
approach,
but.
B
B
B
C
Blue
green
test,
it
like
have
it
just
a
randomizer,
and
you
know
five
percent
of
the
time
somebody
that
launches
their
app
with
the
new
experimental
one.
C
Yeah
check
the
ip
block
is
in
azure.
C
Not
yet,
but
that
that's
the
plan
for
extending
beyond
our
core
languages.
B
Yep
yeah,
I
guess
the
well
we've,
had
plenty
of
performance
complaints
that
aren't
using
hotel.
So
I
don't
think
it's
well.
I
think,
there's
a
general
issue
like
people
who
don't
understand
how
to
do
performance
tuning
who
are
like,
oh,
let's
swap
out
our
old
agent
for
a
new
one
and
then
just
expect
everything
to
work
perfectly
they're,
probably
in
for
a
surprise.
B
So
we've
definitely
come
across
this
with
some
customers
who
really
don't
understand
how
to
do
performance
tuning
like
we've
looked
at
their
jvm
settings
and
they're
like
they're
nonsense
like
they
literally
make
no
sense
in
combination
with
all
with
what
they
have
set
up
or
they're.
You
know
running
with
the
really
old
garbage
collector
with
settings
that
haven't
been
supported
since
java
6,
and
things
like
that.
So
sometimes
it's
hard
to
know
is
the
performance
problem
real
or
is
it
just
that
they
don't
know
how
to
tune
their
jvm?
C
C
B
C
C
Yeah
we
send
with
we
do
sampling,
I
mean
you
can
turn
on
sampling
and
we
send
like
a
item,
count
that
then
the
back
end
is
a
it's
a
multiplier
right
to
generate
metrics,
but
the
metrics
aren't
precise.
Then
so
I
know
that's
why
we're
looking
forward
to
having
metrics
generated.
A
A
C
Oh,
how
to
even
decide
upstream,
like
more
than
just
the
because
the
problem
that
I
have
with
ours
is
that,
like
you,
if
you
do
it
upstream,
your
downstream
nodes
all
need
to
be
configured
with
that
same
sampling
rate.
So
you
send
that
same
item
count,
even
though
the
the
decision
was
made
upstream
at
the
if
it's
head
based
sampling.
C
So
that's
why
I
would
I
like
that
otep,
I'm
hoping
that
that
hotep
goes
through
because
then
we'll
propagate
the
the
percentage
in
the
trace
state
but
yeah.
I
don't
know
how
like
well
like
new
relic,
does
stuff
like
that.
Also
with
like
operation
names
like
getting
kind
of
it'll
sample.
Some
of
your
outliers
also.
A
A
A
A
A
Which
is
annoying
like
I
feel
as
if
the
wrapping
mechanism
of
open
is
not
very
good.
A
Brave,
had
a
different
way
where
it
would
like
samplers
could
either
make
a
decision
or
not,
and
if
it
didn't
make
a
decision.
It'd
go
on
to
the
next
sampler,
but
an
open,
telemetry,
there's
actually
a
direct.
A
C
A
C
Maybe
we
can
drop
the.
C
C
Yeah,
I
actually
like
the
if
you
like
originally
with
you
know,
checks
filtering
them
out,
but
I
kind
of
realized
that
you
know
it's.
Basically,
I
may
as
well
give
it
first
time
yeah,
that's
what
we.
A
X-Rays
to
support
this
that's
one
gap.
We
have
returned
history
right
now,
because
we
haven't
implemented
an
experience
in
korea
and
we've
been
sort
of
one
x-ray
sampler
that
oh,
like
x-ray,
has
remote
sampling,
oh
cool
yeah,
and
that
the
config
itself
looks
very
similar
to
this,
like
it
has
percentage
and
path-based
percentages
and
whatnot,
and
then
you
can
update
that
on
the
server.
C
C
A
A
Yeah,
at
the
same
time,
so
seeing
how
you
have
a
downstream
solution
to
this,
that
does
give
me
a
bit
more.
That
does
make
me
feel
a
bit
more
comfortable
with
us,
starting
with
something
downstream.
Also,
we
might
do
that.
I
mean
I've
actually
been
in
the
past
couple
of
weeks.
I've
been
trying
to
push
through
that
also
like
until
that
happens,
I
mean
it's
not
a
big
deal
for
us
to
have
our
x-ray
sampler.
If
someone
wants
to
switch
upstream,
they
just
change
the
config
setting.
A
A
A
C
Cool
well
thanks
again
for
paying
attention
to
the
and
knowing
how
to
even
do
something
about
the
intellij
gradle
slowness.