►
From YouTube: 2021-05-19 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).
C
C
C
A
A
A
A
A
A
A
C
A
Anyway,
yeah
the
only
thing
that
I
commented
in
the
pr
that
I
have
not
done,
and
I
actually
don't
know
how
to
do,
which
is
this
whole
thing
where
you're
not
allowed
the
otlp
specifies
semantically
you're
not
allowed
to
have
instagram
buckets
or
or
sorry
histogram,
but
no
histogram
count
or
some.
Maybe
it's
some,
maybe
a
histogram
sum
or
the
summary
sum
that
has
negative
values
included
with
it
needs
to
be
purely
all.
A
Values
recorded
must
be
positive
in
order
otherwise
you're
supposed
to
leave
the
sum
out
of
the
proto,
and
I
think
it's
for
something
to
do
with
with
the
prometheus
compatibility
that
I
don't
really
understand
and.
A
A
B
A
A
A
A
C
D
C
C
A
A
What
was
there
was
a
short.
It
was
a
pretty
short
meeting.
Oh
yeah,
this
whole
stuff
about
the
schema.
You
know.
Have
you
taken
a
look
at
all
any
of
that
stuff
with
the
adding
a
schema
to
the
tracer,
the
tracer
instrumentation
library?
So,
in
addition
to
a
name
and
a
version,
you
get
a
schema,
you
want
to
put
a
schema
and
it'll
be
a
published
schema.
A
C
A
Discussion
about
my
proposal
to
stabilize
the
api
just
I
think
that
there's
questions
about
how
complete
it
has
to
be
before
we
stabilize
it
like
do.
We
need
to
have
the
hint
api
into
it
which
hasn't
inspected.
All
I
think
do
we
need
oh
batch.
You
know
the
other
batch
recordings
I
think
are
going
to
be
probably
left
out
of
the
initial
api
version,
which
is
also
good
because
I
think
they're
more
controversial
than
people
think
so
that's
probably
not
going
to
be
in
the
initial
metrics
api
1.0
either.
A
Basically,
let's
bundle
up
a
whole
bunch
of
recordings
all
at
once
and
shift
them
as
a
batch,
but
there's
some
disagreement
about
what
it
means.
Yeah
christian
wants
to
make
sure
that
all
of
the
recordings
at
the
same
time
stamp
on
them
yeah
remember
that
yeah
bogdan
doesn't
want
that,
doesn't
think
that's
possible
or
useful.
A
So
all
right,
I'm
happy
to
have
them
left
out.
I'm
happy
to
have
be
a
simpler
api
to
start
with.
A
So
I
think
that
riley
was
going
to
basically
push
out
the
sdk
stabilization
for
a
month
and
pull
in
the
api
stabilization
stabilization
by
a
month
into
the
plan.
So
from
a
timeline
perspective,
focus
on
getting
the
api
really
nailed
down
hey,
there
was
a
conversation.
It
was
discussion
with
your
name
on
it.
I
don't
know
if
you
saw
the
you
got
called
called
out
about
aws
in
it
was
an
auto
hotel.
Java
discussion
is
that
true.
C
And
so
so
I
mean
it's
not
on
every
spin,
like
aws,
like
he's
so
he's
using
x-ray,
it's
x-ray
also
it's
only
on
the
top
the
root
spin,
it's
not
on
every
spin
similar
to
a
tailpiece,
that's
good,
but
even
then
I
think
he's
saying
this.
Especially
the
process
line
is
generally
ridiculous
for
java
programs.
A
I
I
think
that
process
the
process,
the
command
line
is,
is
something
that
we
should
make
opt-in.
That
thing
is,
I
mean
it
really
like
it.
Actually,
we
have
to
specifically,
we
had
to
have
a
specific
fix
in
our
splunk
metrics
collector
exporter,
to
throw
out
attributes
that
were
too
big
because
they
basically
were
being
rejected
by
the
back
end,
because
it
was
more
than
some
number
of
characters.
A
A
I
mean,
obviously,
if
you're
doing
it
manually,
it's
easy
to
just
pick
off
the
pieces
we
want.
But
but
even
then
I
couldn't
get
the
process
without
getting
that
command
line
and.
A
C
C
C
A
I
don't
have
any
good
ideas
yet,
but
maybe
I
mean
maybe
do
our
resources
have
names
at
the
moment
like
the
like
the
exporters
and
samplers
and
stuff?
Maybe
we
need
our?
Maybe
we
need
our
resource
providers
to
be
named,
so
you
can
easily
bring
them
in
and
bring
them
out
just
as
command
line,
comma
separating
command
line.
C
A
Yeah
yeah
yeah,
so
we
did
that
and
then
they
could.
We
could
have
people
either
not
put
anything
and
just
get
out,
get
all
of
them
or
put
the
incl
put
in
include
or
enabled,
and
then
the
comma
separated
list
of
what
you
want,
enabled
or
other
of
the
names
or
disabled.
And
specifically
when
you
want
to
disable
and
then
you
probably
have
as
much
flexibility
as
you
need.
A
B
A
A
A
C
He's
been
pretty
responsive
in
case
he
was
first
trying
out
x-ray,
sdk
and
he's
like
asking
for
some
features.
He's
like
don't
do
x-ray
sticky
to
hotel
please
and
he
jumped
into
that
and
he's
been
pretty
active,
so
I
think
he
could
help
implement
stuff
too,
like
he
also
like
you've
seen
several
issues
right.
He
also
filed
issue
about
the
sampler
and
whatnot,
but
he
got
his
custom.
Sampler
working
and
things
are
going.
Okay,
I
think.
A
C
A
Yeah
that'll
be
super
cool.
New
relic
has
a
feature
like
that
with
our
agent,
which
is
very
very
handy,
especially,
we
would
often
give
customers
like
just
a
custom
jar
that
they
could
drop
into
their
extensions
directory,
and
agent
would
pick
it
up
and
use
it
like.
C
A
Field
the
field
instrumentation
people
would
often
do
that.
They're,
like
consultants
going
in
they're
like
hey,
here's,
a
little
custom
jar
and
they'll
drop
it
in
and
you'll
get
your
custom
instrumentation.
A
A
A
I
one
thing
I'd
be
nice,
like
I
have
a
really
hard
time,
knowing
how
much
of
that
is
just
like
intrinsic
overhead
of
the
sdk
and
how
much
of
it
is
the
agent
on
top
of
the
sdk
next.
If
we
could
figure
out
how
to
separate
that,
like
the
performance
overhead
into
those
pieces,
I
don't
have
a
good
handle
like
what
is
the
agent?
Does
the
agent
do
extra
processing
on
the
hot
path?
C
C
C
A
E
C
A
Yeah,
it
would
be
really
handy
to
plug
that
in
to
be
able
to
plug
that
in
as
an
agent
option
and
then
be
able
to
test
pure
agent
overhead,
because
then
you
do
all
the
instrumentation,
like
all
the
bytecode
instrumentation
would
all
be
there
and
I'll
be
in
place.
It
would
just
be
doing
nothing
exactly
yeah.