►
From YouTube: 2022-05-26 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
Go
pretty
good
yeah
busy,
as
always
yeah
spend
a
lot
of
time.
Reading
the
metric
specification
and
a
bunch
of
other
projects.
B
C
C
A
Time,
no,
you
know
what
it
is.
I
totally
forgot.
We
went
back
to
the
old
system
where
there's
only
like
three
or
four
meeting
links
and
I
changed
the
time
and
it
overlaps
with
you.
I
did
not
think
about
that.
Oh
shoot.
Okay,
okay,.
A
D
Am
I
right
so
I
don't
think
we
want
to
use
this
one
long
term,
but
the
one
I
just
put
into
chat
was
from
the
java
instrumentation
sig,
usually
yeah,
they're
right
before
us
so
long
term.
We
probably
want
to
find
a
different
one,
but
I
expect
they're
probably
wrapped
up
now,
and
we
can
hop
over
to
that
one.
A
Yeah,
let's,
let's
do
that,
I
will
let's
see
copy.
I
don't
know
how
to
copy
this.
There
we
go.
I'm
gonna
update
the
zoom
meeting
in
the
agenda.
D
A
So
I
see
there's
one
for
the
instrumentation
messaging,
which
is
definitely
not
going
to
overlap
with
this.
I
was
thinking
of
just
using
that
one.
It's
974,
whatever
okay.
A
Yeah,
I
do,
let
me
grab
a
link,
I'm
gonna
post
it
here
in
the
chat,
anthony
and
then
I'll
post
it
into
the
doc.
A
And
I'll
update
the
meeting
event:
okay,
yeah,
sorry,
tristan,
we'll
see
all
the
gopro
over
there
no
problem,
but.
C
C
Yeah
we'll
get
started
in
a
second,
I
guess
yeah,
it's
already
five
after
because
of
all
that,
so
probably
trying
to
there's
anything
else,
I
meant
to
put
in
the
agenda
besides
gustavo's
point
a
board
of
agenda
item
on
baggage
spam
processor.
That's
probably
the
I
mean.
That's
definitely
the
most
important
one
to
discuss
right
now.
C
Anything
else
that
pressing,
except
for
the
demo
app
which
crap
I
forgot
to
push
my
branch
I'll,
put
a
link
to
that
in
the
slack
channel
today.
C
If
anybody
wants
to
take
a
look
at
the
the
phoenix
code
for
it's
just
for
crud
stuff
for
managing
a
feature
flag
database,
so
pretty
pretty
simple,
I
was
I
don't
know
if
there
is.
If
there
is,
I
might
recreate
it,
but
there's
not
like
there's
a
lot
of
stuff
that
phoenix
generates
into
the
application,
and
it's
probably
not
it's
a.
C
I
mean
not
enough
of
an
issue
that
I'm
gonna
try
to
remove
it,
but
if
there's
like
a
way
to
generate
one
that
doesn't
have
so
much
because
I
mean
maybe
most
people
looking
at
it
are
gonna
be
experienced
with
phoenix,
so
they
won't
be
overwhelmed.
Looking
through
the
code
and
being
like
is
any
of
this
stuff
needed
or
where's
the
the
stuff.
That's
actually
open
telemetry,
so
it
might
not
be
a
big
deal
but
yeah.
I
was
just
surprised
by
that.
C
E
There
is
three
flags
I
believe
you
can
pass,
it's.
No,
it's
meow,
no
ecto
and
no
live
view,
and
then
it
would
be
quite
bare
bones.
There.
C
I
am
using
the
html
and
the
ecto
stuff
so
but
maybe
remove
the
live
view
stuff
or
was
it
the.
C
Yeah,
I
guess
it's
mainly
because
it
has
live
view
and
live
component,
and
that
makes
it
larger
the
underscore
web
module
has
a
bunch
of
stuff.
Endpoint
router
all
are
full
of
stuff
that
aren't
needed,
and
I
really
want
to
break
it
by
trying
to
piece
these
things
out,
but
yeah,
something
somebody
might
want
to
take
a
look
at
when
if
they,
if
they're
able,
we
can
move
on
to
oh
yeah,
I
guess
go.
People
might
also
keep
showing
up
good
one
gustavo
thanks,
so
the
baggage
span
processor,
that's
well.
C
First,
I
mean
thanks
for
adding
that
that's
been
needed.
It's
going
to
be
very
useful,
but
we're
I
discussed
the
other
day,
some
other
people
in
who
work
on
either
go
or
the
spec
itself
about
my
understanding
of
span
processors,
and
I
guess
the
my
understanding
was
a
little
wrong
and
the
spec
needs
some
clarification.
C
So
it's
not
quite
as
rigid
as
I
read
it
to
be,
and
so
we
don't
necessarily
need
to
do
the
decorator
thing,
because
what
I,
what
I
understood
from
the
spec,
was
that
any
span
processor
had
to
be
able
to
have
an
exporter
attached
to
it,
which
just
it
doesn't
make
sense.
C
But
what
it
means
is
well
what
it
says
is
the
built
it's
only
referring
to
the
built-in
span
processor,
so
the
one,
the
the
one,
the
simple
and
the
batch
it
doesn't
mean
that
every
spam
processor
has
to
implement
it.
So
it's
just
some
poor
wording,
and
so
that
makes
it
a
lot
less
awkward
and
the
it's
each
spam
processor
is
supposed
to
implement
a
force
flush
which
I
also
took
to
be
related
to
the
exporter.
But
it's
just
in
general.
C
If
there's
any
state
with
in
a
spam
processor,
you
would
need
to
flush
it,
so
it
doesn't
necessarily
apply
to
an
exporter.
C
So
the
problem
is
more
the
the
way
I
changed
the
configuration
to
be
what
I
thought
was
simpler,
which
is
just
spam
processor
and
you
can
give
it
batch
or
simple.
C
C
Do
we
have
any
other
ideas,
or
should
we
do
some
sort
of
decorator
thing?
So
you
can
say
that
on
the
batch
processor
you
want
this
other
thing
to
run,
so
it
adds
baggage.
And
then
you
don't
have
to
care
about
the
ordering,
because
it
would
be
a
separate
thing.
You
wouldn't
have
to
make
sure
it
runs
before
the
batch
processor.
C
I
don't
know
I'm
opening
the
ideas.
I
think
I've
stu.
I
was
originally
leaning
into
either
a
new
configuration
or
the
decorator,
but
now
I
think
that
might
be
leaning
into
just
the
the
way
it
is
now
but
and
going
back
to
the
old,
not
going
back.
So
the
the
new
way
of
saying
span
processor
with
a
single
item
will
still
work,
but
in
our
examples
and
everything
we'll
go
back
to
and
have
the
support
for
that
it
can
just
be
an
atom.
E
E
The
semantic
between
spain
process
processor
that
doesn't
export
and
one
that
does
export,
also
something
that's
bugging
me
a
little,
because
if
it
experts,
we
want
it
to
be
at
the
end
of
the
spam
process
of
this
hours,
because
if
there's
anything
before
that
modifies
this
pen,
it
has
to
be
before
exporting.
Otherwise
it
doesn't
make
sense.
It
almost
does
make
sense
I
mean.
Maybe
it
could
be
a
small
processor
that
just
I
don't
know,
vlogs
show
us
the
doubts,
but
that
could
be
exported
too,
but
yeah.
E
I
thought
maybe
there
could
be
like
a
main
processor
thing
like
this
is
guaranteed
to
be
the
last
and
then
there
is
secondary
and
I'm
not
sure
it
could
be
awkward,
should
configure
a
main
processor
and
then
configure
secondary
processors,
and
if
you
only
have
one
you
just
do
processor
dots
or
if
you
only
want
to
do
a
list
with
one.
You
could
also
do
at
least
with
one
it's
a
little
strange.
E
C
F
What's
it
it
wasn't
simple
processor,
then
it
just
ended
up
being
batch
processor
with
a
set
of
export
of
one.
No
because
the
I
thought
that
was
somebody
suggested
from
some
other
language.
C
I
suggested
that,
but
then,
when
we
discussed
simple
processor
with
some
the
person
with
the
trigger
guy
right,
yeah
yeah,
the
purpose
of
the
simple
one
was
to
be
synchronous,
so
it
would
block
until
it
was
exported
and
we
wouldn't
have
that
functionality.
If
we
did
it
with
the
batch.
If
we
did
it
just
like,
because
it
would
still
be
pass
it
off
and
then
move
on.
E
At
the
end
of
the
day
regarding
the
configuration
issue,
I'm
fine
with
either
way
like
creating
a
main
main
process
of
this
export,
that
processor
list
and
then
decorator
is
processor
list
or
just
have
a
processor
that
lists
that
sort
of
dependence.
And
finally,
if
e3
yeah
yeah.
C
I
think
we'll
go
with
just
the
processor
list
to
not
differ
from
so
like
your
current
configuration
will
be
fine,
we
can
merge
it,
as
is
there
might
be
some
updates
because
or
oh
wait,
I'm
looking
at
your
configuration
again.
C
E
E
C
C
Technically
should
be
spam
processors,
so
yeah
I
get
the
only
so
yeah.
I
think
that's
pretty
much
subtle
unless
anybody
has
anything
else
to
say
on
it
and
we
can
get
this
merged.
But
that
then
brings
to
the
issue
of
publishing,
which
found
out
is
not
as
simple
as
I
thought
it
would
be,
because
hex
doesn't
allow
mixed
ownership
of
an
organization
and
a
user.
C
So
I
don't
know
if
we
want
to
just
start
adding
people
to
the
open,
telemetry
hex
org
that
we
trust
I
mean
as
long
as
it's
not
someone
just
completely
handing
off
a
if
it's
someone
who's
just
completely
handing
it
off
and
what's
nothing
involved
with
it.
Obviously
we
won't
add
them
or
if
we
don't
know
them
at
all,
but
just
to
keep
it
simpler.
So
so,
like
you,
gustavo
can
keep
publishing
it
to
hex.
C
I
guess
that
would
be
an
option.
I've
always
shied
away
from
ci
just
because
I
feel
like
it's
gonna
start
publishing
something
bad
if
nobody's
looking
at
it,
but
it
could
be
an
option
for
oh
good.
How
would
we
do
that?
C
F
Because,
if
you,
even
if
you
yeah,
because
your
tags
would
be
super
complicated
because
of
the
amount
of
repos
and
then
we
also
would
not
be
able
to,
it
would
be
fairly
complicated
to
try
to
quickly
republish
something.
C
C
With
yeah
with
x,
you
can,
you
can
just
say
they
should
I'm
just
saying
like
that.
Hex
should
add
this
functionality.
Yeah
the
org
yeah,
like
I
said
ping
in
the
channel
again
cause
nobody
ever
responded
to
the
fact
that
you
came
through.
C
F
C
F
F
C
All
right,
so
you
added
this
propagator
process.
Propagator
item
to
the
agenda.
F
Was
this
yeah?
I
asked
andrew
to
look
at
it
because
it
was
related
to
that
change.
That's
holding
up.
F
I
wouldn't
it
would
be
nice
to
get
even
more
feedback,
because
essentially
what
I
I
mean
like
one
thing
is
probably
fine
and
I
don't
think
it's
like
controversial
at
all,
which
was
just
returning
changing
the
return
type,
if
no
context
is
found
to
just
be
an
empty
map,
because
that's
how
hotel
context
does
it?
If
it
doesn't
find
the
context,
then
it
just
returns
a
blank
map
yeah,
which
I
think
is
then
safe
to
say,
attach
and
it's
just
empty.
F
So
that
part
is
not
really
controversial.
I
think
but
then
I
I
did
it
add
a
new
function
of
get
context.
F
Well,
it's
just
kind
of
like
people,
so
that
was
what
was
ending
up
in
the
ecto
one
was
you
had
to
like
get
the
current
see
if
there
was
something
if
there
was
a
trace
running
at
all
locally
and
then,
if
not
go,
do
it?
I
don't
know
that
the
I
guess
one
thing
was
like
a
question
was
with
this
method.
I
don't
think
that
there's
any
issue
of
needing
to
attach
and
detach
some
new
context
right,
like.
F
F
F
F
C
F
That,
if
you,
if
you
manually,
call
those
things
like
you
need
to
like
be
very
cautious
about
calling
detach
yeah
good
point,
but
I
don't
know
that
it
belongs
in
this
library.
I
think
I
mean
like
it
probably
needs
some
long
and
hotel
context
yeah.
It
should
be
in
the
main
in
the.
C
F
F
C
I
think
the
only
confusing
part
is
calling
it
parent
because
it
means
parent
process.
Doesn't
it.
C
Because
you're
not
getting
the
well,
I
guess
it
is
the
parent
context,
because
you
want
to
make
children
out
of
it
like
you,
don't
want
to
use
it
as
the
current
context,
but
it's
also.
F
F
F
F
C
C
F
C
C
I
don't
remember
exactly
what
that
does,
but
that
might
be
something
to
consider.
Yeah.
F
E
D
E
E
D
E
You'd
have
a
predict
propagator
that
you
can
inject
and
stretch.
E
C
Oh
yeah,
so
there's
a
thing
with
the
head
would
work
the
profit
the
propagators
use
from
remote
span
to
get
to
create
the
span
context
from
what's
propagated,
which
sets
it
to
is
remote.
True
and
that's
how
it
knows
the
stuff
a
bit
like.
Not
it's
not
a
local
spam
that
can
be
updated.
F
C
E
Exactly
if
you
pass
an
empty
header,
it's
no
lock,
no
operation
is
done,
so
the
same
thing
could
be
true
for
predicts.
C
Yeah,
that's
the
tough
one;
it
probably
it
wouldn't
so
it
wouldn't
use
that
because
we're
not
going
to
be
injecting
it.
If
there
was
an
injection,
it
would
be
the
case.
But
since
it's
not
reading
a
text
map
from
the
other
process,
it
would
need
to
be
like
something
new.
I
mean
I
might
not
even
be
able
to
reuse
much
of
the
existing
propagator
stuff.
It
would
just
it
want
to.
C
C
E
C
Yeah
yeah
we
just.
We
only
have
the
text
map
right
now,
so
something
could
be
added,
but
but
yeah
I
mean
I
guess
technically
you
could.
F
C
C
You
would
just
call
it
would
basically
just
be
renaming,
get
context
into
extract.
That's
what
I
think
it
would
end
up
being
yeah,
that's
what
I'm
saying:
yeah
yeah
and
using
create
remote
span
when
doing
this,
so
that
it's
a
clearly
a
remote
span,
context
that
was
decoded
or
fetched
so
that
it's
not
used
locally.
C
E
C
C
F
F
F
C
C
Well,
no,
you
could,
you
could
still
extract,
have
the
extract
return,
the
current
processes
current
context,
but
it
would
say.