►
From YouTube: 2021-08-03 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
B
B
He
mentioned
to
me
earlier
today
that
he's
still
looking
for
for
reviews,
and
this
is
stock.
The
first
one
about
adding
optional
data
point
flags
to
all
metric
setup
data
points
in
order
to
support
stillness
markers.
It
has
already
three
approvals,
but
we
need
more
definitely
especially
from
the
metrics
groups.
B
B
The
second
one
is
for
allowing
non-monotonic
and
empty
songs
on
instagrams,
and
it's
a
similar
situation,
because
this
one
was
opened
by
josh
suret
with
some
holidays
this
week
at
least
there's
some
discussion
going
on
there,
but
in
general
we
need
more
eyes
as
well
and
the
last
one
that
he
added
was
base
two
exponential
histogram
protocol
sport.
It
was
opened
a
pair
of
weeks
ago.
B
It
has
no
approvals,
but
it
needs
more
reviews,
and
there
was
of
course
some.
There
have
been
some
comments
already,
but
we
need
actually
a
little
bit
more
to
move
forward,
probably
the
last
one
on
the
base.
Two
exponential
histogram
is
the
least
important
at
this
point,
but
please
take
a
look.
B
Okay,
let's
move
on
then
probability
sampling
has
two
attempts
to
review.
J
macdig
was
also
working
with
those
ones
he's
at
this
moment,
building
a
prototype
for
this
long
story
short.
We
want
to
add
additional
information
when
samplers
are
used
to
basically
add
information
regarding
that
this
part
called
adjusting
out
adjusted
count
where,
basically,
when
you're
sampling,
you
are
adding
information
of
how
many
spans
are
something
similar
to
the
sponge
you're
sending.
B
So
that's
important
and
there's
two.
As
I
said,
the
168
specify
how
to
propagate
head
sampling
probability.
This
is
the
main
one
and
the
one
that
is
which
has
which
has
a
lower
number
one
and
48
wheels
on
top
of
that
very
related.
B
And
finally
nikita
you
want
to
talk
about
composite
samplers.
A
A
So
in
my
head,
the
logical
way
to
solve
that
is
to
use
some
endpoint
sampler.
If
this
pan
originates
from
this
endpoint
sampler
will
say
drop
it,
but
at
the
same
time,
all
other
traces
or
spans
should
use
some
sampler
like
trace,
trace
id
based
or
parent-based.
What
not
so
it's
it
comes
out
like
there
is
a
need
for
arbitrary
composition
of
samplers.
A
A
Yeah,
so
there
is
one
one
specific
example:
parent-based
sample
which,
by
by
definition,
is
composite,
but
we
but
spec
doesn't
say
anything
in
general
about
samples
composition.
If
we
don't
provide
any
guidance
for
like
there
is
a
environment
very
variable
based
configuration
of
samples,
for
example,
should
we
support
some
sort
of
sample
composition
over
there
as
well
like
I
want
this
emperor
and
that.
C
B
I
know
that
jim
id
has
some
ideas
about
this.
I
don't
know
what
are
the
exact
details,
but
I
will
even
if
you
cannot
communicate
that
to
the
next
sampling
see
call
which
is
on
tuesday
this
thursday.
I
will
add
it
there.
So
at
least
I
can
we
can
get
an
idea
about
that.
I
think
yeah.
I
think
he
was
a
little
bit
afraid
of
of
of
one
specific
bad
side
about
that,
but
I
don't
know
what
what's
the
detail
there.
B
C
If
you
could
search
for
some
existing
issue,
I
guess
when
people
introduce
the
parent
base.
I
remember
saying
someone
mentioned
it's.
It
might
be
good
idea
to
formalize
that,
and
also
I
have
a
question
regarding
the
scenario:
do
you
think
in
that
way
the
sampler
might
be
specific
to
some
instrumentation
library,
for
example,
you
have
you
have
application
running
you
take
some.
C
You
take
some
in
incoming
requests
and
you're
doing
some
other
stuff,
and
the
only
some
like
the
sampler
you
want
to
apply
is
only
going
to
work
for
a
specific
instrument
that
that
the
library
is
taking
the
request
anything
else.
You
don't
specify
the
sample
at
all,
because
it's
just
a
waste
of
the
time.
You
know
that
you're
not
going
to
get
any
any
like
crawler
or
some
like
synthetic
input.
In
this
way,
I
think
that
sampler
might
not
be
a
global
thing,
so
you're.
C
To
the
instrumented
library,
because
I
I've
seen
that
in
open,
telemetry.net
folks
were
talking
about
performance
and
they're,
saying:
okay,
if
I
have
an
application
that
can
take
some
like
message,
queue
items
and
I
can
also
take
incoming
http
requests.
I
want
the
asp.net
core
layer
to
just
filter
things
out.
Instead
of
calling
the
sampler,
because
sampler
is
relatively
slow,.
A
C
Yeah,
and
if
you
need
some
example,
you
can
keep
me
offline,
I
can
see
some
open
like
open
time
trade
download
folks
added
some
filter.
I
I
think
they've
measured
the
performance
and
decided.
I
think
a
filter
can
achieve
much
better
performance,
although
it
seems
like
not
well
aligned,
so
you
can
filter
out
the
data
either
as
a
sampler
as
a
spam,
processor
or
even
like
earlier
as
a
filter,
only
instrumentation
library,
but
it
seems
like
there's
a
big
performance
with.
A
B
You
perfect
and
talking
about
samplers.
I
just
want
to
bring
attention
to
this
small
pr
that
anurag
has
for
adding
aws
x-ray,
stamper
environment
variable.
Do
the
specification,
I
think
it's
looking
great
it's
fine.
The
only
detail
is
that
we
may
consider
as
a
follow-up,
probably
to
put
the
vendor-specific
stuff
in
a
separate
file.
B
Yeah.
We
don't
have
to
discuss
it
here,
because
it's
a
matter
of
organization.
I
think
so.
Please
discuss
that
on
the
issue
itself
or
the
pr
itself.
Actually,
please,
and
finally,
just
for
your
information.
I
mentioned
this
yesterday
at
the
maintainers
meeting,
but
in
case
you
didn't
make
it
there.
We
merged
pr
to
have
otlp
http
use
a
different
port.
You
know
like
the
default
port
will
change.
We
don't
want
to
have
both
otlp
grpco
and
http
transports
to
use
the
same
port.
This
was
accepted.
B
B
Other
than
that
we
don't
have
more
issues
on
the
agenda,
so
I
don't
know
if
somebody
wants
to
discuss
something
else.
If
not,
maybe
we
can
just
get
a
small
summary
of
what's
happening
in
the
metrics
world,
as
we
have
been
doing
in
the
previous.
B
C
B
C
All
for
that,
thank
you
and
the
sdk
part.
I
I
think
we
made
a
good
progress
last
week,
so
we
try
to
scope
down
the
view
pr
by
removing
something
that
is
either
not
super
clear
or
people
are
not
all
happy
with
that
and
try
to
focus
on
the
core.
So
the
ick
tries
to
target
the
minimum
viable
solution
right
now,
and
that
makes
us
being
able
to
make
progress.