►
From YouTube: 2021-06-15 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).
B
B
B
Okay,
I
guess
we
can
start.
Thank
you
for
joining
us,
usually,
first
of
all,
as
you
guys
may
remember,
we
were
trying
to
triage
new
issues
coming
as
part
of
this
call.
B
C
C
C
For
example,
I
have
this
question
like
which
I'm
working
on
the
metrics
apn
sdk
spec,
and
then
there
are
a
lot
of
folks
trying
to
ping
me
offline
either
like
slack
or
somewhere
else,
and
they
mentioned
hey
riley,
there's
a
problem
in
the
metric
semantic
convention.
And
will
you
answer
my
question
or
will
you
fix
the
problem
when
I
look
at
the
semantic
combination,
I
I
think
so
far
it's
far
from
ready,
so
I
I'm
not
sure
if
we
should
make
progress
there
or
we
should
just
set
the
right
expectation.
C
And
how
should
we
communicate
that
so,
for
example,
currently
I
have
to
tell
people
that
the
metrics
api
sdk
is
not
done
and
all
the
terminology
might
change,
so
it
probably
doesn't
make
sense
for
me
to
go
and
change
the
semantic
combination
right
now
and
also
there's
a
limited
bandwidth
on
my
side.
So
it'll
be
great.
If
we
can
clarify
that,
I
I've
seen
the
other
semantic
convention
in
the
similar
situation,
for
example,
like
the
database
part,
we
know
that
there
is
an
issue,
but
do
we
have
a
project
to
track
them?
C
E
So
riley
I
mean,
as
josh
said,
we
have,
we
met,
met
and
discussed
the
instrumentation
effort
and
again
maybe
we
can
create
a
group
or
you
know
we
can
just
select
one
of
the
folks
from
the
semantic
conventions
team
group.
That's
sorry,
the
instrumentation
group
that
can
help
there.
C
B
C
Okay,
so
this
is
the
only
issue
that
we
should
change.
The
other
part
I,
I
think
they're
already
like
labeled
correctly
so
I'll
I'll,
follow
up
with
ted
and
create
a
project.
E
Had
taken
riley
on
the
instrumentation
side,
we
had
taken
some
action
items,
so
I've
been
I'm
supposed
to
reach
out
to
michael
lee,
from
from
your
team
and
and
then
sync
up
with
others
like
josh.
C
Yeah
yeah,
it
looks
to
me
like
there
are
multiple
steps,
for
example,
like
we
probably
cannot
make
great
progress
on
the
matrix
semantic
combination,
simply
because
the
metrics
api
sdk
they're
still
under
construction,
so
setting
that
expectation.
So
we
can
communicate
in
a
consistent
way
to
people
who
have
those
questions.
Then.
E
B
All
right,
thank
you.
Thank
you
so
much
for
that.
Let's
go
to
the
next
one,
which
is
you,
oh
okay.
Actually,
that
was
part
of
that
sorry.
I
thought
that
it
was
a
additional
comment.
You
put
there.
John
watson
reminder
please.
F
B
Okay,
next
item.
G
Bryce
please
hi
there
yeah,
so
I'm
I'm
trying
to
get
this
pr
working
for
the
additions
for
mobile
agent
carrier
information
and
it
seems
like
there's
some
sort
of
issue
with
the
table
generation,
and
it
looks
like
that
sorry.
G
I
was
poking
around
trying
to
collect
all
my
all
the
documents
that
I
wanted
to
look
at,
but
I
think
armin
was
was
commenting
on
it,
saying
that
there's
something
wrong
with
the
with
the
the
keyword
that
I'm
using
versus
the
allow
custom
values-
and
I
just
wanted
to
get
that
clarified-
should
be
I
shouldn't
be
using,
may
should
it
be
using
optional
when
I
use
custom
values,
I'm
really
hoping
that
that's
the
problem,
because
yesterday
I
was,
I
was
running
these
tests
and
one
of
them
would
the
the
fix
the
make
fix,
would
add
a
space
in
and
it
would
fix
the
table
error
that
I
was
getting
the
table
check
error.
H
H
Generate
method:
well,
how
did
you
create
the
microphone
in
first
place.
G
Oh,
the
the
markdown,
I
just
wrote
it
initially
is
it
should
I
remove
it
all
and
then
run
the.
H
Generate
and
then
that
that
explains
a
lot
I
will
add
a
another
comment
pointing
you
into
the
right
direction,
because
there
would
be
a
generator
tool
for
that.
That
would
take
the
yaml
and
then
make
markdown
out
of
it
and
as
we
can
see,
it's
not
that
well
discoverable
at
least
not
always
yeah
I'll.
Add
another
comment.
Okay,
sorry
thanks!
Thank
you.
B
Yeah,
I
think
that,
that's
honestly,
a
minor
detail.
What
I
would
like
to
have
is
more
people
pay
attention
to
this.
I
was
taking
a
look
myself
at
the
changes
themselves.
They
look
fine,
but
we
need
more
people.
I
think
that
the
errors
we
can
iron
them
out
offline,
but
please
for
more
people.
Please
review
this.
As
I
said,
this
looks
correct
from
my
perspective,
but
we
need
more
eyes.
B
All
right
next
one
I
put
that
music
exporter,
so
there's
an
issue
regarding
having
environment
variable
supporting
more
than
one
exporter,
and
this
should
seem
to
be
initially
trivial.
But
then
great
great
feedback
came
from
john
and
always
saying
that.
Well,
there
are
some
details
about
that.
I
don't
know
if
there's
if
there
are
any
opinions
about
that.
B
F
C
Yeah
there
are
scenarios
where
you
have.
You
have
two
explorers.
One
is
the
pool
model,
sending
data
to
premiere
phase,
another
one
is
push
sending
to
otlp,
or
there
could
be
scenario
where
you
have
two
weak
spotters.
One
is
sending
something
very
critical
like
the
heartbeat
information
every
second
and
then
you
have
something
that
has
lower
frequency,
like
the
temperature
would
only
send
every
one
minute
or
even
lower
frequency.
B
B
C
Support
now
yeah,
so
it
has
been
a
an
issue
on
the
previous
sdk
spec,
and
I
see
multiple
folks
supporting
that
and-
and
we
discussed
this
in
the
sixth
meeting
so
so
far,
there
seems
to
be
a
a
good
amount
of
use
cases,
and-
and
this
is
something
fundamental-
we
only
allow
one
exposure
later.
We
want
to
add
more,
we
might
change
the
design,
so
I
figure
at
this
moment.
It's
like
something
we,
we
should
figure
out.
I
May
I
ask
if
this
issue
is
more
about
multiple
exporters
or
more
about
using
environment
variables
for
configuration?
I
feel
like
it's
actually
about
configuration
using
environment
variables,
because
we've
talked
about
how
for
the
sdks
both
in
tracing
and
metrics,
like
you
really
might
want
to
have
something
much
more
sophisticated
than
choosing
your
exporter
name.
You
might
have
lots
of
variables,
including
export
frequency
and
interfacing.
It's
going
to
be
sampling
configuration-
and
these
are
all
like
really
nice-
to
have
like
magic,
configurators
and
environment
variables
are
severely
limiting
enviro.
I
D
Of
are
we
trying
to
go
down
the
path
of
encoding,
our
entire
environment,
language,
like
our
configuration
language
into
environment
variables,
or
do
we
want
to
give
people
a
secondary
option,
that's
kind
of
like
an
offband
file
or
some
other
way
of
passing
in
config?
That's
not
just
so
like
an
environmental.
I
I
D
I
I
kind
of
think
that,
maybe,
as
as
part
of
the
instrumentation
effort,
we
should
just
kick
off
standard
configuration
for
sdk
processor,
pipeline
flows
across
the
sdks
and
see
what
see
what
happens.
A
I
And
I
think
that
the
work
we
have
today
is
to
say
that
there's
a
protocol
object
somewhere
in
our
proto
tree,
that
that
tells
you
what
a
configuration
struct
looks
like
and
then
at
least
you
don't
have
to
talk
about
specifying
yemo
or
json,
or
something
like
that.
But
we've
already
got
a
struct
for
it
and
potentially
then
you
know
the
user.
The
programmer
can
configure
it
with
a
hard-coded
call.
B
Work.
Okay,
so
let's
have
a
comment
there
that
we
want
to
work
on
these,
where
we
are
going
to
kind
of
halt
it
like
this
pr
specifically,
and
that
this
will
become
part
of
the
instrumentation
team.
Does
that
sound
about
right.
F
It
seems
different
than
instrumentation
to
me
this
sounds
more
like
it's
a
convenient.
If
I
think,
if
we
want
to
do
this,
we
should
not
mix
it
up
with
instrumentation,
because
instrumentation
group
is
going
to
have
an
enormous
amount
of
work
to
do,
and
it
is
not
involved
with
configuration
of
an
sdk.
F
B
D
In
the
we
actually
talked
about
this
in
the
instrumentation
kickoff,
how
there's
like
a
set
of
horizontal
things
that
need
to
happen
and
then
a
set
of
vertical
instrumentation
components,
one
of
the
horizontal
things
that
people
want
to
know
is:
how
do
I
configure
my
instrumentation?
How
do
I
turn
on
and
off
specific
metrics?
How
do
I
like
configure
what
http
error
codes
are
reported
or
what
failure
means
for
that
that
you
know
like
if
we're
going
to
provide
out-of-the-box
instrumentation,
that's
supposed
to
be
ubiquitous
across
all
these
different
things.
D
How
do
users
configure
it
in
a
consistent
way
is
a
horizontal
task
that
is
like
and
and
and
coming
back
to
like,
how
do
I
export
the
stuff
in
the
sdk?
It's
it
it's!
It's
that
same
horizontal
task,
so
you're
right
that
we
shouldn't
call
it
instrumentation,
but
it's
like
a
necessary
component
for
instrumentation
to
succeed.
So
that's
why
I'm
calling
it
out
like
we
talked
about
the
instrumentation
kickoff.
I
think
it.
D
I
think
it
deserves
a
lot
of
time
and
attention
and
yeah,
I
think,
partly
we
have
to
figure
out
how
we're
architecting
metrics
to
get
the
metrics
component
right.
I
was
looking
at
java
and
the
single
exporter
limitation
right
now
and
trying
to
think
through
ideas
for
how
to
get
around
that,
but
I
think
fundamentally
it
doesn't
matter
whether
or
not
we
have
a
single
exporter
in
java.
D
But
I
think
the
reason
we're
seeing
the
pressure
on
the
environment
variables
is
because
we
haven't
given
users
anything
but
environment
variables.
So
that's
the
obvious
place
to
go.
Yeah.
John
sorry,
there
was
a
there
was
an
instrumentation
kickoff
that
ted
organized
with
a
bunch
of
people
to
try
to
not
pull
in
specifically
not
pull
in
sdk
api
maintainers.
To
try
to
like
not
overload
sdk
api.
F
D
Okay,
yeah
that
that's
I
will.
I
will
mention
that
there
is.
Are
you
on
the
instrumentation
slack
channel
the
hotel.
F
E
Yeah
josh.
I
think
that
was
the
action
item,
that
there
would
be
a
formal
say
for
that
or
a
formal
group,
at
least
for
metrics.
B
Okay,
let's,
let's
add
a
comment
there
and
once
riley
you
create
that
group
for
instrumentation.
We
can
just
mention
them
yeah.
Thank
you.
Let's
do
that.
Okay!
B
Next
item
is
about
adding
environment
variables
for
the
attribute
limits
for
links
and
events.
We
only
have
them
for
expanse.
At
this
moment.
It's
very
straightforward.
I
had
this
probably
bad
initial
idea
of
having
a
single
environment
variable,
but
I
think
it
does
make
sense
to
have
three
of
them
because
you
may
want
to
have
different
values
for
the
limits
of
spans
span,
attributes,
link,
attributes
and
event
attributes,
so
pretty
straightforward,
just
go
and
guys
and
just
go
check
that
I
don't
think
it's
it's
super
polemical.
B
And
the
final
one
is
one
pr
that
has
been
standing
there
for
a
long
time
now
it's
about
splitting
out
otlp
grpc
versus
http
endpoint
configuration
essentially
it's
bringing
back
the
insecure
part
of
the
environment,
variable
just
making
it
optional,
and
it
has
two
reviews.
B
But
my
impression
is
that
not
enough
people
have
been
reviewing
that
and
that's
going
to
affect
all
the
all
the
sigs
if
we
decide
to
change
yes,
so
please
really
really
really
asking
you
for
your
feedback.
I
think
two
approvals
for
a
pr
like
these,
even
though
it
seems
small,
it's
not
enough.
We
really
need
at
least
a
pair
of
more
reviews.
C
A
relatively
stable
version
of
the
matrix
semantic
convention,
so
what
we
have
now
is
an
experimental
version
of
the
api.
It
should
be
like
80
there.
We
expect
some
minor
changes
before
the
feature
freeze
and
there's
an
ongoing
discussion,
whether
we
should
add
another
instrumentation
like
the
instrument
to
measure
the
duration.
C
C
Meanwhile,
if
you
are
blocked
on
the
semantic
convention,
I
I
think
so
far.
There
is
no
good
answer,
so
what
probably
you
can
do
is
like
the
the
tracing
part,
you,
you
release
the
library
with
the
the
instrumentation
and
you
mark
that
as
experimental
based
on
your
best
knowledge
from
the
semantic
convention,
and
you
join
the
instrumentation
and
semantic
convention
group,
but
I
I
hope
that
you
know
it
should
be
there
and
and
try
to
accelerate
that
part.
So
that
that's
the
answer
I
can
give.
D
Can,
if
I
can
add
to
that
specifically
with
some
job
advice
from
what
I'm
seeing
I'm
trying
to
prototype
the
new
api,
so
a
few,
a
few
simple
things:
one
is:
if
you're
touching
counters
up
down
counters
and
gauges
right,
you
should
be
totally
fine
if
you're
leveraging
summaries
coming
out
of
java,
that's
gonna
switch
to
histogram
and
that's
gonna
be
a
bit
different
right.
If
you're
touching
anything
in
the
api,
like
views
or
aggregators
or
any
of
the
like
internal
hooks
to
process
that
may
change,
the
way
exporting
happens.
D
Might
change
given
like
the
current
tentative
design
of
the
sdk
and
this
whole
multi-exporter
discussion
so
depending
on
which
pieces
you're
touching
the
api
different
pieces
might
change,
but
the
like
the
shape
of
the
data
package
in
the
sdk
should
be
exactly
the
same.
That's
based
on
the
metrics
data
model
and
it's
it's
up
to
date.
Effectively.
It's
missing
exemplars,
but
you
know
not
a
big
deal.