►
From YouTube: 2022-11-01 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
B
A
A
You,
how
was
your
Halloween,
did
you
celebrate
or
did
your
neighborhood.
A
That
cool
cool
yeah
we
were
we're,
really
excited.
It
was
our
first
year
in
like
a
house,
and
we
thought
it
would
be
a
big
growing
up.
It
was
a
big
Halloween.
You
know
in
my
neighborhood
was
Halloween
like
so
I
got
all
excited,
I
decorated,
only
like
three
people
showed
up,
and
one
of
them
was
like
an
infant
in
a
who
cried
and
then
the
other
people
were
like
awkward
like
must
have
been
16
or
so
yeah
too
old.
For
this
like
go,
get
a
job.
B
A
B
A
A
C
A
There
was
some
discussion
in
the
chat,
but
the
implications
seem
to
be.
A
lot
of
people
are
kind
of
busy.
A
Of
conversation
in
the
chat
today,
yeah
software
right
who,
who
is
the
time
for
that
I'm
here
for
the
thought
leadership,
yeah
I-
can
sure
Rob
had
wanted.
I
mean
oh,
like
what's
in
the
chat?
Was
you
can
probably
review
the
chats
better
Robin
wanted
to
have
an
option
to
like
kind
of
shove
stuff?
That's
normally
events
onto
attributes
for
a
span.
A
What
would
otherwise
because
yeah
span
events
are
kind
of
like
a
red-headed
stepchild
within
the
pipelines
of
The,
Collector
and
back
ends
of
various
folks
and
so
I
think
there
were
some
you
know
rather
than
deal
with.
Attempting
to
you
know
boil
the
air
shouldn't
fix
all
that
you
just
kind
of
wanted
to
shove
it
onto
an
attribute,
so
you
can
do
some
processing
and
I
think
there's
some
back
and
forth.
A
There
I
think
we
ultimately
I,
think
the
consensus
was
sort
of
like
use
a
span
processor
in
app
or
there
were
some
PRS
Upstream
to
give
the
transform
processor
first
class
support
for
span
events,
but
that
is
a
larger
project.
That
I
think
some
of
the
collector
folks
are
working
on
I
kind
of
advocated
for
just
just
you
know
my
usual
stance
of
who
cares
about
maintaining
stuff,
just
shove
stuff,
on
as
a
config
option
that
you
can.
A
You
know
toggle
and
that
got
the
usual
pushback
from
usual
folks,
that's
kind
of
where
that
stood
yeah
I
mean
I
I
can
review
the
the
maintainer
spec
on
my
own
I.
Don't
want
you
to
have
to
waste
your
time.
I
also
happy
to
wait
around
for
a
few
minutes
for
other
folks
or
if
there's
anything,
you
want
to
discuss
I'm
happy
to
be
a
resource.
Whoever
I
can
I.
Don't
I
haven't
done
much
on
this
repo,
so
I
don't
think.
There's
much
I
can
add.
B
A
B
That
I
guess
but
I
don't
know
there
was
this
question.
Should
we
consider
metric
Point
kind,
a
breaking
change,
which
I
think
is
somewhat
controversial?
Oh
those
looks
like
we
have
a
Rob
welcome,
Rob.
A
B
B
So
like
the
protocol.
Kind
of
like
technically
allows
metric
points
for
a
Time
series
to
be
either
floater
doubles,
but
I
think
in
practice
like
they
should
at
a
bare
minimum,
be
the
same
for
the
same
type.
For
like
a
a
run
like
a
lifetime
of
that
time.
Series
she
kind
of
restarted
the
time
series.
Maybe
it
could
change
values,
but
I
think
I,
think
back
ends
vary
and
how
well
they
support
this.
B
If
at
all,
but
I
think
the
question
was
like
so
yeah,
it
seems
like
the
protocol
technically
allows
it.
So
the
question
was:
should
it
be
like
a
breaking
change
for
like
the
semantic
conventions?
I
guess,
if
you.
B
A
metric
from
from
being
like
an
integer
to
a
float,
or
vice
versa,
or
just
allow
flexibility
in
how
you
actually
want
to
report
it
there.
So.
A
B
A
The
idea,
being
you
receive
a
bunch
of
Time
series,
you
know
during
some
apps
life
cycle
and
you
evaluate
the
first
one.
You
determine
whether
it's
a
number
or
a
point
or
maybe
there's
an
indicator,
whether
it's
a
number
or
a
floating
point,
and
then
you
don't
have
to
keep
checking
for
every
single.
You
know.
Entry
right,
you
would
just
check
once
is
that
the
idea,
but
technically
you
do
have
to
check
every
single
time
right
now.
B
I
think
your
backend
needs
to
have
some
way
to
reconcile
this
and
I
think
it
gets
just
more
complicated
in
that,
depending
depending
on
how
all
your
labeling
works
out.
I
think
you
could
be
merging
time
series
from
multiple
sources
and
each
one
of
them
could
potentially
be
sending
them.
You
know
they
they
could
be
consistently
sending.
You
know
you
could
have
one
process
that
is
consistently
sending
ants.
You
could
have
one
that's
consistently
sending
doubles,
and
then
it's
like.
What
do
you
do
now
right.
B
So
be
curious,
facetious.
B
C
Feel
like
every
time
we
say
vendor,
it
means
back
end
because
people
are
going
to
choose
to
do
their
own
on-prem,
Telemetry
storage,
even
if
they're
in
for
pain,
doing
so
yeah.
So
it's
not
just
like
we're
pumped
into
the
vendors.
They
can
figure
it
out.
They've
got
all
that
VC
money.
It's
like
anybody,
who's
going
to
store
this
data.
C
Has
this
problem
for
what
it's
worth.
If,
if
we
were,
if
speaking
on,
we
have
a
honeycomb.
If
you
were
to
send
us
this
with
a
field
coming
some
attribute
on
a
metric,
the
value
coming
in
with
different
types.
A
A
B
A
I
can
ping
the
folks
internally
here
and
let
them
know
that,
because
I
know
Robert
probably
has
more
to
say
here
than
I.
Would
he's
been
more
involved
in
the
metric
stuff
yeah,
it
seems,
like
seems
like
a
complex
question
and.
A
Yeah
I,
don't
know
it's
about.
You
know
it's
a
good
point
it
it
feels
like
it
ought
to
be
a.
It
would
be
nice
if,
in
the
semantic
conventions,
you
could
I
guess
in
some
in
some
comp
right
now
you
sh
you
do
denote
what
if
it's
a
float
or
a
integer
or
whatever,
if
I
call
correctly,
it
would
be
reasonable
to
me
that
that
changing
that
would
be
a
breaking
change.
B
A
B
I
think
we
are
mostly
crossing
our
fingers
that
this
doesn't
happen
in
that
things
consistently
come
in
as
one
type
or
or
another
just
out
of
the
natural
Order
of
Things,
but
I
I
think
this
change
would
kind
of
open
the
floodgates
to
has
to
deal
with
that
much
more
regularly.
A
Yeah
I'd
be
curious.
How
like
other
open
source
metrics,
like
does
Prometheus
offer
any
sort
of,
like
you
know,
convention
here
that
they
I,
don't
really
think
they
yeah
I,
don't
know
like
yeah
I,
guess:
I'm
curious
how
this
is
handled
in
other
places.
They
mentioned
okay,
open
metrics.
They
say
they
allow
it
to
switch.
B
I
think
that's
how
we
ended
up
getting
this
structure
and
the
Proto
if
I'm,
remembering
the
discussion
properly
so.
A
Cool
yeah
Rob
before
this
I
tried
to
kind
of
summarize
the
chat
stuff
just
to
for
Matt's
sanity.
So
I
didn't
have
to
reread
everything,
but.
C
C
Like
I
could
see,
maybe
a
config
value
to
say
like
turn
off
those
span,
events
or
not,
and
not
what
format
do
you
want
to
take
them
in
yeah,
but
also
with
there?
There
is,
as
you
would
noted,
in
the
chat
movement
of
foot,
to
try
to
do
things
in
the
collector
and
it's
a
matter
of
whether
people
using
the
instrumentation
want
to
miss
out
like
running
the
collectors,
not
insignificant,
yeah.
A
Yeah
you.
C
A
A
Yeah
I
mean
it's
a
recorded
call,
so
you
deserve
my
my
therapy
session
for
another
point
in
time.
A
A
C
A
C
A
C
Challenge
is
that
that
we've
had
is
that
in
sdks
the
processors
are
not
pipelines,
but
in
the
collector
they
are
right.
A
A
Yeah
not
exactly
Canadian
or
like
Plug
and
Play
ish.
You
can't
just
have
code
Snippets
or
you
say
like
add
this
thing
in
yeah.
Ariel's
brought
this
up
a
bunch
which
is
like
there's
sort
of
this
drift
in
what
just
like
terminology
around
like
what
processor
means
and
that's
like
confusing
to
people
and
then
like.
Is
that
intense?
It's
if
it's
intentional
like
why
and
I
don't
know,
I
wish
you
all
way
more
than
luck.
A
Yeah
I,
don't
have
anything
else.
You
know,
I
have
been
working
on
this
too.
Actually
so
I'm
just
here
to
listen
and
keep
playing
I
guess,
but
I
have
nothing,
no,
nothing
to
bring
up
or
no
pressing
questions
or
PR's
or
anything
I
think
Robert
had
merged
some
changes
to
the
my
sequel
instrumentation
that
maybe
ought
to
get
added
to
Trilogy
and
postgres
I'm,
not
sure
if
he's
done
that
or
if
someone
else
has
or
whatever.
A
Basically
we
have
to
like
some
stuff's,
not
necessarily
like
was
it
utf-8,
encoded
yeah?
If
you
go
down
to
it's
like
seven
or
eight
down,
handle
encoding
errors
in
my
SQL
obfuscation,.
A
A
So
that's
probably
a
PR
a
low
hang
for
PR
that
ought
to
get
done.
I
think
someone
had
mention
they'll
get
around
to
it,
but
okay,
but
yeah.
So
we
should
probably
that's
I
know
that's
an
open
thing
that
probably
ought
to
get
done.
A
It's
merged,
but
in
the
comments
Ariel
points
out
like
it's,
this
is
for
my
sequel,
but
the
problem
exists
probably
for
the
PG
instrumentation
Trilogy,
which
is
just
another
like
my
sequel,
Library,
so
I
think
we're
maybe
testing
this
internally
still
I'm,
not
totally
sure
I'm
involved
in
some
different
stuff,
okay,
yeah
I,
guess
Robert
was
saying
he
wants
to
verify
that
the
fix
you
put
up
like
addresses
some
of
the
issues
we
were
seeing
before
you
know
patching
the
other
stuff.
A
I
can
I
guess:
I
should
probably
follow
up
and
see
like
I.
Don't
think,
we've
shipped
this
thing
yet
I
think.
Maybe
next
week
we're
supposed
to
that's
all
I
got
I
haven't
been
paying
attention
to
anything
else:
cool
wow.
A
Researched
it
director
that
merge
or
is
that
open,
yeah,
that's
interesting,
the
Azure
resource
detector
looks
interesting,
cool
I,
think
I
think
we
kind
of
have
some
not
like
our
I'm,
not
sure
what
our
right
now,
like
our
resource
detector,
if
I
remember
correctly,
is
sort
of
just
like
just
gcp
or
AWS.
A
A
Gcp,
okay,
yeah.
B
A
This,
like
the
auto
detector,
maybe
just
as
like
hard-coded
to
only
check
gcp
yeah,
so
we'd
have
to
update
anyway
I
it's
nice
to
see
Azure
getting
some
love.
We
should
probably
I,
should
probably
review
that-
and
it's
probably
a
good
nudge
to
say,
like
we
should
probably
add
AWS
as
well,
but
we're
not
AWS
users,
so
I
wouldn't
have
trouble
doing
that
anyway.
B
Yeah,
that's
fine,
I
have
have
you
covered.
Everything
then.
Is
there
anything
else
that
we
did
want
to
talk
about.
A
One
day:
oh
no
cool
thanks,
everybody
yeah
good,
seeing
y'all
as
well
cheers.