►
From YouTube: 2022-09-12 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
B
A
I
think
normally
morgan
drives
this,
but
he's
not
here.
So
I
guess,
let's,
let's
do
the
check-ins
for
the
specifications
seek
where
the
release
is
blocked.
On
an
issue,
I
added
the
link
to
the
maintenance.
Talk
it's
about
the.
A
A
So
the
issue
is
there:
if
you
have
any
thoughts,
please
comment
on
it
on
the
logs
we're
we're
working
on
adjusting
this
sdk
wording,
primarily
after
the
recent
changes
to
the
logging
api
specification,
so
basically
refining
the
api
and
sdk
logging
spec
and
after
that,
and
I
guess
the
work
has
already
started
on
the
java
and
at
least
to
make
it
compliant
with
this.
With
these
recent
changes,
that's
on
logs-
and
I
can
also
talk
about
the
product-
I
don't
think
there's
anything
new
in
the
products
or
no
updates
there.
A
Okay
php
anyone
from
php
wants
to
say
anything.
B
C
Yep,
just
what
it
says
there
we
have
two
issues
left
for
a
metrics.
Ga
one
of
them
is
just
documenting
it
on
the
website,
which
is
obviously
important
for
ga,
the
other
one's
not
really
blocking.
We've
been
waiting
on
spec
clarification
for
a
while.
It's
about
removing
unused
label
values,
but
we're
shooting
for
a
ga
release
in
the
next.
I
don't
know
a
week
or
two
we'd
probably
like
to
get
a
tc
spec
compliance
review.
C
I
see
somebody
answered
my
question
there,
it's
optional,
but
the
sig
may
request
it.
Does
anyone
know
what
the
process
is
to
request
a
review?
Should
I
just
create
a
community
issue
or
something.
A
D
Yeah,
I
can,
I
can
jump
in
here
so
we're
still
working
towards
a
metrics
sdk
alpha
release,
94
complete,
it
is
slated
for
I
think
october.
6
was
our
timeline
when
we
expected
to
go
out,
but
we're
really
trying
to
have
that
happen
sooner
than
that
date
so
also
related
to
the
project
we
have
1.110
coming
out
this
week.
This
is
just
the
release
for
all
of
our
sable
apis
and
sdk.
D
One
of
the
things
that
we
didn't
add
is
the
scope
attributes
in
this
release.
We
specifically
took
it
out,
so
we,
I
think
tigard
you
mentioned.
This-
is
something
that's
blocking
the
spec
release
as
well,
but
just
kind
of
a
heads
up
that
like
this
is
also
something
that's
blocking
our
progress
as
well.
Well,
I
guess
not
blocking.
Is
there
something
we're
aware
of
and
not
including
feature-wise,
so
yep
just
wanna
call
it
out
the
go
auto
instrumentation
seg
also
had
a
kickoff
meeting.
D
Last
week,
we're
gonna
be
on
a
bi-weekly
meeting
cadence.
If
people
wanted
to
drop
in
that
interest
in
the
project
next
week,
we're
going
to
do
some
demos
of
the
proposals.
So
if
you're
interested
stop
buying.
E
Yeah,
I
can
talk
about
that,
so
we
worked
on.
I
mean
to
portray
this.
I
think
we
spent
some
time
on
supporting
the
shared
libraries
for
linux
and
mac
os.
I
mean
this
basically
for
gcc
and
c
lang.
That's
for
the
scenarios
with
the
sim,
where
the
application
are
built
with
with
such
a
way
that
the
visibility
of
symbols
are
hidden
and
that
support
was
not
there
and
that's
that's
something
substantial.
We
have
added
last
week,
support
for
linux
and
mac
for
windows.
E
We
still
don't
have
any
solution,
how
to
add
supports
for
dll,
so
that
is
still
something
which
is
not
supported
yet
matrix
sdk
for
ga.
Our
work
is
happening,
it's
bit
slow
as
of
now,
but
we
are
hoping
to
pick
it
up
in
next
couple
of
weeks,
so
no
no,
no
date
as
of
now
for
the
g8
but
yeah
it's
ongoing.
As
of
now.
A
Ruby
swift,
collector
roster
line
community
demo,
no
updates
anyone
from
comms.
Yes,.
F
Hello,
so
quick
update,
austin
parker,
who
was
the
maintainer
and
lead
of
the
comsig,
has
stepped
down
he's
gonna
be
doing
community
roles,
so
myself
and
severn
newman
are
now
the
two
co-maintainers
and
co-leads
of
the
website,
repo
and
docs
and
vlog.
So
yeah
hi.
We're
doing
that.
F
The
major
issue
that
I
see
happening,
as
I'm
sure
you've
read,
is
that
there's
no
metric
stocks
at
all
and
in
fact
we
have
pulled
some
metric
stocks
from
some
of
the
language
sigs
as
well
or
some
of
the
language
sig
subsections
because
they
were
like
wildly
out
of
date
using
incorrect
apis
and
like
people
were
filing
issues.
Saying
like
this
doesn't
even
type
check
we're
like
okay,
clearly
a
problem.
What
like
there
might
be
a
good
like
pattern
that
we
could
apply
like.
F
I
suggested
that
one
of
the
maintainers
of
a
language
sig
does
like
a
code
dump
of
like
okay.
This
is
like
the
right
way
to
initialize
stuff,
and
then
we
can
basically
take
it
and
run
with
it
from
there.
But
then
across
you
know,
seven
or
eight
languages
that
that's
difficult
to
like
initiate
on
our
end,
and
so
anyone
who
is
working
on
metrics,
where
you
think,
like
all
of
your
apis,
are
relatively
stable.
F
If
you
have
an
ability
to
just
do
like
a
doc
dump
or
like
just
unstructured,
doesn't
doesn't
have
to
be
good
or
anything
like
that.
We
can
start
fleshing
that
out,
but
it
is
a
worry
that
you
know
by
you
know
the
metrics
sdk
releases.
There
may
not
be
any
docks
available
and
we
want
to
not
have
that
be
a
thing.
C
Yeah
we
already
talked
about
this,
I
think
last
week,
but
I'm
working
on
one
for
js,
a
doc
stump
similar
to
what
you
just
described.
I
agree
like
we're.
Holding
our
ga
release,
at
least
until
the
documentation
is
usable,
doesn't
have
to
be
amazing,
but
I
I
don't
want
to
release
a
wandeto
with
no
docs,
so
we're
holding
until
at
least
that
point.
D
I
think
I've
asked
this
in
the
past,
but
I've
never
really
got
a
clear
understanding
of
where
the
partition
of
what
the
docs
on
the
open
telemetry.io
website
should
contain
versus
like
the
language
specific
docs,
because
I
think
the
the
problem
that
you're
mentioning
is
is
systematic
to
the
structure
that
we
currently
have.
D
The
docs
on
the
website
are
a
copy
of
the
language
docs,
and
I
think
that
that's
something
that
is
not
serving
us
so
like
a
lot
of
the
languages
themselves
have
their
own
hosting
sites
for
docs
like
read
the
docs
or
and
go
there's
just
package.dev,
or
something
like
that
right
and
I
I'm
not
familiar
enough
with
other
languages
to
really
speak
on,
but
I
know
that
they
exist
when
you're
copying
docs
from
the
existing
languages
to
the
open,
telemetry.io
site,
like
all
you're
doing,
is
taking
a
snapshot
in
time
and
you're
almost
guaranteed
to
have
this
exact
issue
in
the
future,
where
things
aren't
going
to
compile,
because
the
packages
are
going
to
change
or
something
like
that.
D
I
know
that
we've
tried
to
version
the
docs,
so
that,
like,
if
you
you
know,
are
trying
to
use
the
telemetry.io
docs
you'll
always
have
some
sort
of
like
stable
version
of
the
package
you're
using,
and
I
think
that
helps,
but
it
also
duplicates
effort
right
like
not
only
do
you
have
to
publish
docs
to
the
language
specific
site,
but
you
also
have
to
publish
it
to
teletrade.io.
D
I
I
also
wonder
also
like
users
like
do.
We
have
a
good
understanding
for
the
users
of
these
docs
like
who
uses
these
stocks
like
when
they
come
to
the
telegram
io
site
like?
Are
they
looking
for
just
an
intro?
Are
they
looking
for
examples?
Are
they
looking
for
complete,
like
functional
listings
and
like
they're,
wanting
to
search
them
the
same
way
they
do
the
other
sites
like?
I
think
that
a
better
understanding
of
that
would
help
direct
effort,
because
I
know
we
have
a
limited
amount
of
bandwidth
here.
F
Yeah,
oh
just
quick
time
is
this
the
sort
of
thing
that
we
can
have
a
discussion
about
here,
or
should
we
or
are
there
other
things
that
we
should
get
to
now?
I
think
it's
fine.
We
have
time,
okay,
cool
yeah.
F
I
didn't
want
to
absorb
a
whole
bunch
of
time,
so
some
anecdotal
evidence
that
I've
had
just
in
my
job
working
with
end
users
of
open
telemetry
is
they
go
to
the
website
and
they
expect
to
have
all
docs
for
how
to
do
stuff
to
be
there
and
then
they
get
confused
when
like
they
have
to
go
to
different
repos
or
like
it
so
well,
I
should
clarify
conceptual
docs
that
are
language
agnostic.
They
100
expect
to
be
there
conceptual
docs
that
are
specific
to
a
language
but
are
not
like
an
api
reference.
F
They
also
still
expect
to
be
there
with
a
copy,
pastel
code
sample.
That's
going
to
work,
and
you
know
like
if
there's
a
tiny
thing
they
have
to
fix
up
or
something
it's
not
that
bad.
So
long
as
like
all
the
major
concepts
actually
exist
and
there's
a
code
sample
api
reference
like
they
like
java
developers,
expect
like
javadoc.
For
that
I
would
imagine,
go
developers.
I
haven't
talked
to
them
specifically
about
this.
Expect,
like
the
package
docs
for
that
which
do
exist
and
are
are
quite
extensive.
F
Some
of
the
things
that
are
missing
on
those
language-specific
docs
like
on
the
go
one,
for
example,
it's
a
great
api
reference,
but
there's
no
like
okay.
How
do
I
like
pull
in
these
three
apis
to
then
initialize?
This
thing,
the
way
that
I
want
to
or
like.
What's
a
good
example
of
I'll
use
like
a
span
link
as
an
example?
F
How
do
I
go
create
a
spam
link
now,
like
you,
could
go
figure
that
out
by
kind
of
walking
through
each
of
the
different
api,
calls
and
use
your
ide
with
intellisense
to
kind
of
get
you
what
you
need,
but
copy-pastable
code
sample,
there's
better.
I
think
with
tracing
the
fact
that
it's
like
not
part
of
like
a
built
artifact
from
like
an
sdk,
has
been
better
just
because
most
of
the
tracing
sdks
are
stable,
and
so
there
haven't
been
a
lot
of
turn
but
metrics.
F
There
has
been
churn
for
sure,
because
you
know
it's
I'm
sure
you're
all
aware
the
apis
have
changed
quite
a
bit
over
the
past
year,
and
so
we
haven't
like
established
any
principles
around
how
to
best
incorporate
these.
F
I
know
that
in
the
past
stuff
used
to
be
entirely
on
the
doc
site,
and
then
it
was
all
moved
into
sub
modules
for
each
of
the
different
language
repos,
with
the
intent
being
that
that
would
be
things
would
be
better
kept
up
to
date,
but
then
that
didn't
actually
happen,
and
so
that
doesn't
mean
that
that
couldn't
that
that
would
be
the
same
outcome.
If
that
were
to
happen
again,
but
then
we
started
pulling
stuff
back
into
the
website
repo
and
just
updated
the
docs
more.
So
I
don't
know.
F
I
think
that
was
probably
just
more
of
a
bandwidth
issue
like
either
on
the
dockside
and
language
specific
cigs
like
there
were
just
not
enough
people
with
enough
time
to
actually
document
stuff
and
keep
it
up
to
date.
We
have
more
of
that
now,
but
I
think
with
metrics
specifically-
and
I
imagine
this
will
probably
happen
for
logs.
It's
just
a
matter
of
knowing
like
okay
are
the
apis
stable
such
that
they
can
be
documented
or
are?
F
Is
there
realistically
going
to
be
churn,
and
thus
you
know,
should
we
just
tell
people
to
like
hold
off
on
using
it
for
now?
I
think
we
ran
into
that
problem
with
metrics,
which
is
why
we
ultimately
pulled
stuff,
but
yeah,
I
don't
know,
I
think
we
should
probably
have.
D
So,
based
on
what
you
just
said,
like
it'd,
be
really
helpful.
Speaking
of
the
gosig,
but
I
imagine
this
is
others
is,
if
you
can
present
to
the
sig
themselves,
some
list
of
topics
they
need
to
document
not
necessarily
like
have
to,
but
like
a
good.
A
good
guideline
would
be
ideal.
I
know
that
our
as
we're
kind
of
saying,
like
our
metrics
offering,
is
becoming
stable
and
go,
and
I
could
see
in
the
next
milestone
us
really
strategizing
and
focusing
on
docs,
but
it'd
be
really
a
wasted
effort.
E
D
Kind
of
like
what
you're
saying
with
like
spam
links
like
how
do
you
create
a
spam
link
to
have
the
same
thing
of
like
all
those
listed
topics
of
what
you
need
for
the
metrics
for
our
documents?
If
you
could
just
I
mean
it
really
could
just
be
like
an
issue
or
something
like
that
like
or
or
something
in
a.
I
don't
know
a
piece
of
paper.
A
F
Well,
I
I
would
also
say
that,
like
so
between
me
and
severan
like
we
can,
we
can
usually
figure
out
how
to
do
everything.
It's
just
a
matter
of
like
okay,
like
is
this
like
it's
usually,
it's
usually
initialization
like
the
the
right
way
using
quotes
to
initialize
things
in
a
given
language
is
usually
the
hardest
thing
to
go
and
figure
out,
and
then,
after
that,
like
it's
fairly
straightforward
to
know
how
to
document
it,
but
I
think
one
thing
that
we
could
probably
do.
Maybe
this
would
probably
help
is
on
the
com.
Stick.
F
We
could
go
back,
make
this
an
agenda
item
to
go
and
figure
out.
Okay,
assume
a
given
language
is
ready
to
get
documented
for
metrics.
What
generally,
should
the
structure
of
like
how
to
use
metrics?
What
should
that
look
like?
Should
that
be
like
an
individual
page?
Should
there
be
sub
sections
on
these
specific
pages?
If
so,
what
this
should?
Those
generally
look
like
and
then
come
back
to
be
like
hey.
This
is
how
like
we
think
it
could
be
generally
be
presented
pretty
well.
F
F
D
B
D
You
could
just
if
we
get
an
issue
where
you're
saying
like
literally,
if
you're
asking
like,
can
you
show
us
how
to
initialize
this?
Like
that's
great,
if
you're
asking
us,
can
you
write
a
paragraph
or
two
around
this,
like
that's
great
as
well,
but
having
some
sort
of
ask,
because
I
I
see
like
our
next
documentation,
milestone
being
like,
let's
come
up
with
something
we
think
is
good,
which
I
don't
want
it
to
be,
just
not
what
you
guys
need
as
well.
So.
D
Might
you
know,
but
like
yeah
I'd
rather
not
be
churned.
F
Make
sense,
okay,
I'll,
make
sure
that's
an
agenda
item
for
our
next
meeting
and
then
we
should
have
something
we
can
follow
up
with
different
folks
on.
A
All
right
thanks,
phillip
tyler,
you
have
a
couple
issues
there.
D
D
D
I
guess
if
you
use
a
ratio
sampler
to
our
parent-based
ratio
sampler
afterwards,
so
that's
not
ideal.
I'm
wondering,
I
guess
I'll
bring
it
up
on
this,
because
I'm
wondering
if
other
maintainers
have
seen
this
or
they're
tracing
ratio-based
sampling
algorithm
does
this
and
if
that's
the
case,
if
they
have
a
solution
or
if
also
then
we
should
define
it
at
the
specification
level.
If
that's
something
we're
at
a
point,
we
could
do.
D
I
can
also-
I
guess
I
don't
really
know
where
to
create
an
issue.
I
guess
is
the
question
because
I
don't
know
if
this
is
a
specification
issue,
we
can
obviously
try
to
solve
this
and
go,
but
I
also
think
that
this
is
something
that's
more
than
just
one
language
implementation
here
yeah,
I.
G
Could
suggest
creating
an
issue
in
the
specification
and
cc
anthony
mirabella?
Okay,
I
would
say
that's
the
thing.
The
other
thing
is
that
geometry
is
not
here,
but
I
think
that
the
idea
is
that
the
probability
you
know
like
these,
the
current
city,
radio
samplers,
will
be
like
deprecated
eventually.
So
there's
a
few,
you
know
things
there
I
mean
in
theory.
G
We
would
like
to
start
using
in
the
future
at
some
point,
the
probability
samplers,
the
new
ones,
which
there's
only
limitations
in
gel
and
gold
and
they're,
not
part
of
the
core.
Yet
so
right.
B
Yeah,
I
think
the
consistent
probability
sampler
is
the
the
specified
way
out
of
this.
If
we
wanted
to
add
to
the
spec,
I
think
we
would
need
to
like
specify
take
from
the
middle
of
the
trace
id
instead
of
the
front
of
the
trace
id
something
like
that,
but
that's
still
going
to
require
multiple
languages
to
change
their
implementations.
D
Right
yeah,
it's
a
it's
a
kind
of
I
don't
know
the
right
solution
and
like
there
may
be
more
technical
computer
science
ways
to
solve
this
as
well,
but
I
just
want
to
raise
the
issue.
Okay,
I'll,
take
it
as
an
action
to
create
an
issue
at
the
spec
level
for
it.
D
The
next
thing
that
we
also
found
was
in
google.
We
were
truncating
attributes,
namely
strings
and
their
representation
in
as
utf-8.
I
think
it
might
actually
be
utf-16,
but
we're
truncating
at
the
byte
level,
not
the
character
level.
D
So
we
had
a
bug
there,
and
I
think
that
josh
mcdonald
also
saw
this
in
a
few
other
languages,
where
the
specification
defines
truncation
at
a
character,
it
doesn't
really
define
what
a
character
is
at
that
point,
but
I
think
it's
just
kind
of
assumed
that
it's
like
a
uniform
unicode
code
point
or
an
ascii
code
depending
on
the
the
language.
D
One
of
the
things
that
was
kind
of
interesting
was
like
in
the
truncation
algorithm.
We
also
realized
that
there
is
a
possibility
for
invalid
characters
to
exist
in
the
string
itself.
So,
like
there's
a
lot
of,
I
think
the
truncation
algorithm
that's
underspecified,
and
I
think
this
is
also
an
area
where
I'm
wondering
if
other
sigs
have
spent
some
time
trying
to
get
this
right.
And
if
we
should
create
an
issue
to
try
to
get
this
a
better
specification
around
this.
A
So
the
product
of
specifications
says
that
the
strings
must
be
valid
called
unicode,
I
think,
or
maybe
udf8
doesn't
really
make
a
difference
anyway.
What
it
means
that
is
that
all
the
senders,
all
the
exporters
have
to
supply
valid
unicode
strings
to
the
marshalling
functions.
A
The
only
place
where
that
that
enforcement
can
happen
is
the
sdk
right.
So
we
need
to
enforce
that.
Somehow
I
I
guess
we
need
to
refine
the
the
the
limiters
that
we
have
the
specification
for
the
limiters
to
define
more
precisely
what
it
means
right
to
to
to
essentially
try
and
truncate
it
in
a
way
that
is
valid,
for
the
experts
to
use
probably
should
be
filed
as
an
issue
against
the
spec
repo.
D
Okay,
all
right,
I
can
do
that
as
well.
Yeah,
one
of
the
things
tristan
that
we're
interested
teagan
that
really
came
up
was
just
like
what
happens
if
there's
invalid,
are
you
supposed
to
replace
it
with
a
valve
or
like
a
like
a
standard
like
this
is
under
invalid
code?
Or
do
you
just
drop
it?
So
I
think
that
this
is
again
like
a
good
area
for
yeah.