►
From YouTube: 2022-11-15 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
B
A
C
B
C
D
E
C
F
D
A
A
So
I've
been
wanting
to
get
involved
with
contributing
code,
docs
kind
of
everything
to
otel
for
a
while
now
and
I,
just
wasn't
sure
where
to
start
and
someone
on
Mastodon
said:
Ruby
had
the
nicest
folks
and
then
I
saw
your
name
on
the
list.
Oh
it's
Rob
and
so
I'm
like
all
right.
Let
me
start
with
Ruby
I.
Could
dust
off
my
Ruby
skills
I
could
do
it.
E
D
B
E
An
agenda
let
up
and
drop
it
in
the
in
the
chat
we
usually
start
off
as
by
doing
a
specsig
recap,
it's
kind
of
the
meeting
before
this.
It's
a
long,
boring
meeting
that
most
people
don't
want
to
sit
through
that
I
usually
do
so
I
I.
Try
to
summarize
that
in
you
know,
10
or
15
minutes,
depending
on
on
how
generous
Sam
is
being,
and
then
we
kind
of
move
on
and
talk
about
things
related
to
the
Ruby
repository.
G
E
Cool
there's
a
fairly
long
and
contentious
topic
to
start
things
off
and
I'm
still
trying
to
understand
all
of
what
the
problem
is
here,
but
basically
I.
Think
golang
has
had
some
issues
with
the
generated
protos
and
things
that
change
in
the
protos
that
end
up
somehow
breaking
the
the
go
SDK
and
my
understanding
is,
you
know
so
I
guess
this
issue
is
saying
generally:
do
not
expose
the
generated
protovall
structs
in
in
the
API
anywhere
and
I'm,
not
sure
how
that's
actually
happening
like
I.
E
Think
I
think
it's
okay
to
use
these
in
your
exporters,
but
not
okay,
to
use
them
elsewhere,
and
maybe
this
is
what
was
happening
in
golang,
but
maybe
I'm
still
trying
to
piece
everything
together.
I
think
they
are
depending
on
like
a
external
package
and
I
think
they
can
get
in
a
situation
where,
like
the
external
package,
changes
and
breaks
the
exporter,
even
though
yeah,
even
though
it
necessarily
hasn't
been
like
a
breaking
change,
it's
like
a
new
method
was
added
to
an
interface
or
something.
B
E
E
E
I
think
this
is
to
kind
of
bridge
bridge
the
gap
between
push
and
pull
systems
and
knowing,
if
yeah
it,
if
a
system
that
is
up
should
be
up
and
I
think
in
the
Prometheus
world.
It's
you
kind
of
know.
If
you
have
your
scrape
configs
and
you
try
to
scrape
something-
that's
not
there,
but
in
like
a
pull-based
system,
you
may
or
may
not
know.
If
a
if
a
system
is
up
so
I
think
they
would
like
to
specify
something
on
how
this
would
work.
There's
interest,
I
think
Heim.
E
A
So
that's
something
I'm
learning
the
Prometheus
like
pole,
based
model
I'm
used
to
the
kind
of
agent
system
I,
would
expect
that
up.
Metrics
come
from
an
external
Pinger
Source.
Do
we
I
mean?
Are
people
really
seeing
this
up
metric
as
something
that
otel
needs
to
be
responsible
for
I,
almost
I,
almost
I'm
like
oh,
my
gosh,
get
you
know
ping
dumb
or
something
or
make
your
own
Pinger.
C
E
D
E
E
D
E
E
There
was
some
discussion
about
being
able
to
add
attributes
that
aren't
part
of
the
semantic
conventions,
but
prefixing
them
with
an
X
and
then
I
think
that
discussion
ended
up
whenever
you
do
that.
It's
like
well
what!
If
what
what
happens
when
it
transitions
from
being
experimental
to
being
non-experimental
and
the
end
result?
E
Is
you
end
up
supporting
all
the
names
forever,
even
though
you
know
there's
only
supposed
to
think
a
grace
period,
it's
hard
to
it's
hard
to
make
those
changes,
because,
as
soon
as
you
remove
support
for
the
X
attribute,
like
you
will
have
a
lot
of
people
coming
to
complain
that
things
broke,
which
is
harder
usually
to
deal
with
than
just
leaving
support
for
all
of
it.
So
I
know.
E
E
E
I
guess:
that's
there's
some
bike
shedding
about
whether
these
metrics
should
have
like
a
posix
prefix
or
not.
If
they
are
kind
of
more
widely
appliable
or
if
they
are
really
kind
of
like
encoding,
some
posix
conventions
that
wouldn't
apply
to
other
systems
and
next
week
There's
Thanksgiving,
it
seems
like
they
will
skip
the
meeting.
So
there
you
go.
Thank
you
Tito's
for
the
Tito's
alarm,
Bell.
E
F
Yeah,
actually,
the
release
that
is
truly
busted
is
instrumentation
all
because
of
the
transitive
dependency
conflict
and
we
have
to
get
articopca
out
to
get
the
new
instrumentation
all
out,
and
obviously
this
is
just
a
complete
nightmare
and
I.
Don't
really
think
I
I,
don't
really
want
to
talk
about
what
the
solution
is,
but
I
do
the
solution
to
the
overall
problem
unless
others
others
do
but
like
there's
an
open,
GitHub
issue
about
like
how
we
could
have
this
happen.
F
F
H
F
F
E
F
G
Set
out
an
opinion
that
people
can
react
to
We
should
enable
them,
because
we
want
to
discuss
instrumentation
and
contribute
things
in
contrib
and
discuss
core
things
in
core.
It
does
give
us
another
Avenue
to
tell
people.
You
didn't
ask
this
in
the
right
place.
So
we'll
need
to
be
nice
about
that,
but.
D
H
Pretty
reasonable,
I
think
in
the
past
we
were
maybe
a
little
bit
more
hesitant
to
do
so
because
it
was
going
to
just
be
the
same,
like
smaller
group
of
people
looking
at
both
repos,
and
we
were
like
hey,
why
don't
we
put
it
in
one
place
to
kind
of
reduce
the
overhead,
but
the
contrib
repo
has
a
group
of
people
who
are
lovingly
taking
care
of
it
now
so
I
think
it
makes
sense
to
give
them
a
place
to
discuss
I
think
that's
my
two
cents
on
it.
I'm
happy
to
be
around,
though.
D
F
G
C
G
G
C
F
C
C
G
We're
sorting
that
out,
I
think
my
question
I
think
my
question
is:
is
I.
Expect
it
to
be
simple,
was:
were
these
two
active
model
serializers
an
active
job,
instrumentations
emitted
from
the
rails,
roll-up
gem
intentionally
or
is
it
just
an
oversight
of
when
they
were
made?
They
didn't
get
added
to
the
rails.
Gem.
G
H
G
H
G
H
One
of
the
things
you
have
to
watch
out
for
active
job,
though,
is
like
I,
believe
I
haven't
actually
tested
it
or
in
a
long
time
like
if
someone
has
like
psychic
instrumentation
and
a
rescue
instrumentation,
you
add
the
active
job
instrumentation,
they
don't
just
like
be
friends
with
one
another.
You
get
this
like
clashing
duplication,.
G
We
could
we
could,
at
least
if
we
decide
not.
If
we
decide
to
admit
it,
we
could
at
least
document
it
and
say
so.
You
want
to
use
the
rails
gem.
Bringing
all
these
things
in.
Are
you
using
active
job,
bring
that
in
because
it
doesn't
play
well
with
others
when,
like
turns
out
monkey
patching
commonly
used
classes,
you
don't
want
two
things.
You
don't
want
two
monkeys.
H
Yeah,
it's
kind
of
like
the
same
story
as
like,
and
well
it's
not
the
same
story.
It's
a
roundabout
thing
of
like
when
someone
wants
to
instrument
the
elastic
search
stuff.
You
basically
were
getting
duplicate
spans
for
if
you
had
like
a
net
HTTP
instrumentation
you'd
get
your
same
Appliance
past
the
same
like
it's
very,
very
repetitive
and
like
it
would
create
pretty
noisy
low
quality
traces.
So
again,
it's
just
you
know.
If
we
do
that
people
will
learn
the
hard
way
or
well.
D
C
Hi
cool
that
all
makes
sense
to
me,
not
super
urgent,
so
don't
we
just
said:
looked
at
a
random
PR
this
morning,
someone's
we
have
a
bunch
of
like
just
random
strings,
sprinkled
in
our
code
base.
That
mostly
are
just
like
specification.
You
know,
attributes
or
cut
you
know
constants
or
things
like
that,
so
he
those
PR's
just
mostly
doing
some
cleanup
work.
C
A
couple
of
them
are
on
I,
don't
know
experimental,
stuff
or
stuff
specific
to
Ruby,
so
this
PR
adds
them
right
now
to
the
like,
semcom
gym
in
the
common
module
and
yeah.
My
feedback
was
like
I,
just
I'm,
not
sure
whether
I
think
it's
only
like
two
or
three
should
live
in
some,
especially
now
that
we
have
experimental
as
a
thing.
It's
like
a
a
true
citizen
or
whatever
of
our
of
our
repo,
whether
we
should
you
know
be
organizing
the
some
of
these
they're.
C
F
The
only
argument
for
not
throwing
these
in
experimental
is
to
first
of
all
make
sure
that
Matt
doesn't
get
a
chance
to
talk,
even
though
he
unmuted
himself
and
secondarily
so
that
people
don't
have
to
pull
in
like
another
gem
just
to
just
to
have
like
the
like,
because
then
we'll
be
shipping.
The
experimental
SDK
with
everything
like
it'll
just
it'll,
be
bundled
alongside
because,
like
everyone's
going
to
be
using
these
constants,
whether
they
like
it
or
not.
So
that's
sort
of
why
I
thought
maybe.
F
G
No
Matt
doesn't
get
done
I'm
going
to
add
to
that,
like
the
only
feedback
I've
gotten
on
instrumenting
Mastodon
is
God,
that's
a
lot
of
dependencies,
so
another
adding
another
gem
to
that
list.
I,
don't
think
that's
gonna
yeah
I,
like
the
idea
of
like
maybe
open,
Telemetry
semantic
conventions.
Maybe
we
have
an
unsemantic
conventions
namespace
within
this
Gem
and
go
like
look.
We
needed
names
for
things
in
magical,
constants
or
not.
G
E
Yeah,
these
later
comments
are
kind
of
trending
towards
at
least
my
two
cents.
Was
it
going
to
be
that
the
semantic
conventions
constants
should
be
things
that
are
actually
in
the
spec
semantic
conventions?
And
these
other
constants
that
we
have
littered
around
the
SDK
should
have
possibly
a
another
home.
C
C
It,
oh
it's
just
a
good
Otep,
I,
think
or
you
know
something
along
those
lines.
I
think
it
has
high
value
in
the
end
user
group
that
I
attended
once
or
twice
before,
I
get,
you
know,
add,
kicked
in
I
stopped
attending
people
had
a
common
issue
with
that
across
languages
and
Ruby
was
the
only
sick
that
had
like
a
good
solution
or
it's
a
pro.
You
know
it's
a
better
solution.
What's
up.
C
D
E
G
There's
a
PR
through
Mastodon,
which
is
having
a
moment
it's
sort
of
a
talking
about
Chicken
and
the
Egg
problems.
It's
the
does.
Mastodon
accept
this
PR
without
anybody
actually
using
it
to
see
if
it's
producing
useful
Telemetry,
and
is
it
too
much
so
people
who
know
things
both
about
Mastodon
and
about
open
Telemetry
might
be
worth
taking
a
look
at
that
PR
and
seeing
what's
up
interestingly
mastodon's
using
a
gem
that
has
been
updated,
I
think
since
2018.
G
Adder
encrypted
Adder
encrypted
patches,
active
record
to
encrypt
and
and
or
redact
fields
on
it.
It
monkey
patches
active
record
the
old
way
by
aliasing
Method
at
the
class
method
level
and
our
active
record
instrumentation.
Does
it
the
new
way
with
prepend
and
thusly
when
their
interaction
is
they
call
each
other
infinitely.
G
So
that
PR
now
has
nerfed
the
direct
active
record
instrumentation,
which
I
will
go
back,
Robert
sold
me
on
it.
You
said
you
get
instrumentation
on
callbacks
and
I'm
like
I'm
in
I.
Don't
want
the
active
support.
Notifications
anymore
I
want
my
callbacks,
but
it
doesn't
work
with
Mastodon,
so
I,
don't
foresee
Adder
encrypted
changing
its
ways
since
it's
old,
so
I
don't
know
what
we
can
do
about
our
patching
in
situations
like
that.
G
E
This
this
is
the
thing
that
started
happening.
Look
like
module
prepaid
did
not
used
to
be
a
thing
back
in
the
early
days
of
monkey
badging.
So
there's
kind
of
just
been
like
this
Evolution
and
transition
that
the
Ruby
Community
has
been
going
through
and
I,
don't
know,
I
feel
like.
We
must
be
on
the
module
pre-pened
side
of
that
Evolution
things
were
kind
of
like
mixed
at
one
point,
so
I
think
we
really
need
to
probably
just
Trend
towards
towards
using
using
module,
prepend
and.
G
Oh
I'm
not
suggesting
that
we
that
we
go
back
I'm,
just
it's
a
thing
to
be
aware
of
and
I
don't
know.
If
it's,
if
there's
anything,
we
can
do
about
it
except
I.
Don't
know,
there's
some
dark
arts
to
figure
out
if
I
don't
even
know,
if
you
could
detect
that
a
module
that
something's
been
aliased
and
not
do
the
pre-pen
or
do
something
different
or
is
there
a
load
ordering
so
that
one
doesn't
grab
the
other.
E
The
way
I
don't
know
a
previous
employer
I
worked
at
the
way
that
we
solved.
This
was
four
things
that
this
was
happening.
A
lot
on
there
was
like
an
option
to
install
wide,
v-pan
or
install
by
Alias
method
chain,
but
the
instrumentations
were
subtly
different
in
order
to
make
that
work,
and
it's
kind
of
a
can
of
worms
that
you
probably
would
like
to
avoid
is
is
Adder
encrypted.
Is
it
it's
its
own
kind
of
gem.
E
E
G
D
C
You're
doing
I
am
like
I
did
I
I
actually
am
confused.
I
always
felt
like
you're
supposed
to
put
like
it's
supposed
to
be
like
the
API
lives
in
third
party,
like
you
know,
repos
and
then
like
users
can
bring
their
own
SDK
to
I
I,
never
understood
how
it
actually
works
like
if
we're
gonna
include
open,
Telemetry
and,
like
the
actual,
like
I,
don't
know,
git,
lab
or
or
Mastodon
or.
D
C
E
C
Think
what
you've
done
is
the
best
like
most
practical
way
to
like
get
something
out
the
door,
because
realistically
I
don't
think
we
can
take
a
patch
I.
Don't
think
we
want
to
add
all
the
config
options
and
the
neither
side
is
going
to
take
a
patch
for
this.
So
like
this
is
the
best
you
can
do.
I
think
it
was.
A
common
problem
between
users
would
be
using
two
vendors,
instrumentation
libraries
and
one
vendor
would
have
used
instrumentation
using
the
Alias
method
and
the
other
with.
G
C
E
E
G
B
G
G
E
Yeah,
so
maybe
on
that
note
yeah,
it's
awesome
to
see
this
I
was
going
to
come
back
and
and
see
yeah
see
again.
We
have
Paige
here
and
I.
Think
Paige.
You
were
asking
about
documentation
and
other
ways
to
kind
of
like
participate
and
help,
and
maybe
maybe
the
Ruby
Sig
would
be
a
good
place
to
start
so
I'm
wondering
if
we,
what
things
are
you
really
trying
to
get
into
and
we
can
see
what
things
we
we
have
on
Deck,
okay,
great.
A
Question
a
few
things
now
that
I
am
not
getting
paid
to
code.
I
feel
a
lot
of
ways
about
being
so
far
away
from
the
projects
that
I'm
talking
about
all
the
time.
So
I
do
want
to
be
exposed
to
the
like
realities
of
where
we
are
at
with
otel
and
like
I'm,
just
kind
of
pick
a
language
so
I'd
like
to
contribute
I
can
do
build
stuff,
I'm
I'm,
very
curious,
I
kind
of
want
to
write
up
like
what's
the
life
of
a
release.
What
does
that
take
now?
A
E
E
E
A
E
A
G
G
G
So
it
would
be
an
interesting
like
little
integration
test,
but
the
rest
of
open
Telemetry
demo
is,
is
a
not
trivial
to
yeah
deal
with
the
running
of
it's
an
interesting
idea,
though,
to
have
just
like
a
very
simple
thing
out
there,
that's
instrumented
that
maybe
could
get
hooked
in
as
an
integration
test,
but
our
our
test
Suites
pretty
wicked
right
now.
Oh.
A
G
I
think
as
it
exists
now,
the
Ruby
component
of
the
open,
Telemetry
demo
wouldn't
do
much
more
of
an
integration
test
than
what
like
the
Sinatra
test.
Suite
does
today
the
Sonata
instrumentation
test
Suite
does
today,
but
that's
a
like
I,
like
the
smell
of
what
you're
cooking
it'd,
be
it'd,
be
good
to.
A
A
C
A
C
G
H
Every
day
we
tattoo
stuff
typically
gets
tested
like
it's
nice,
because
most
of
shopify's
applications
don't
run
on
like
say
like
rails,
six
or
seven,
they
just
point
a
head
right
like
so
we're
always
on
the
latest
rails,
wow
and
as
a
result
like
a
lot
of
our
dependencies,
stay
very
very
up
to
date.
So
if
there's
like
a
conflict,
we
find
out
pretty
quickly
when.
D
A
H
Conflict
I
mean
like
it
went
boom
we'll
find
out
quickly.
The
less
fortunate
ones
are
when
the
instrumentation
just
stops
working
and
nobody
knows
right.
So
those
are
less
fun
because
like
unless
we're
actively
just
like
spot
checking-
or
you
know,
looking
at
the
quality
of
the
traces
coming
out
of
certain
applications,
we
may
never
notice
like
I'm
right
on
that
note.
A
A
A
F
F
C
Practically
speaking,
what
Tony
asked
is
probably
relevant,
which
is
the
hotel
Ruby
Channel
within
the
cncf.
Slack
is
generally
where
this
stuff
will
get
surfaced
from
other
parties
or
internally.
If
we
see
something
at
all,
we'll
try
to
mention
in
the
slack
and
anyway,
that
I
think
that's
like
his
official
ish
of
a
channel
as
you'd
want
to
find.
Besides
people
opening
issues
yeah,
and
we
have
no
real
support
guarantees
as
far
as
they
know,.
G
F
C
E
E
G
G
E
G
H
D
G
E
E
E
H
H
It
is
like
kind
of
a
I'm
gonna
just
be
a
dick
one
of
those
stupid
helpers,
or
you
can
just
add
that
one
line
says
like
trace
my
method
for
me,
so
you
don't
actually
have
to
think
about
your
instrumentation.
That
demand
doesn't
seem
to
be
going
anywhere.
Everybody
who
wants
that
is
wrong
and
that's
fine,
but
maybe
we
should
give
them
a
proper
way
of
doing
the
wrong
thing.
I,
don't
know
I'm
a
little
bit
conflicted.
We
have
a
prototype
internally
that
we're
using
to
slow
down
people
being
like
Oh.
H
H
Of
it
right
sorry,
yes,
yes,
the
spec
is
actually
seeming
like
it's
in
a
very
dense,
stable
State,
but
the
implementation
in
Ruby
is
it's
like
half
done
synchronous
instruments.
Are
there
there's
configuration?
We
have
it
separated
in
such
a
way
that
you
could
technically
bring
it
into
the
main
SDK
main
API,
a
non-invasive
way.
H
We
haven't
released
any
gems
of
it,
but
the
export
path
and
asynchronous
instruments
are
completely
undone
right.
Now,
Andrew
Hayworth
was
taking
a
shot
at
asynchronous
stuff,
yet
you
can
find
some
issues.
The
other
like
unfortunate
reality
is
like
my
internal
Focus.
That
work
has
moved
away
from
this
a
lot.
So
it's
hard
to
devote
time
to
this.
Unless
I'm
like
burning
the
midnight
weekend,
oil
which
I've
been
trying
to.
H
H
If
we
can
get
even
synchronous
instruments
all
done
and
then
the
export
path
down
with
otlp
I
would
be
in
a
position
where
I
could
start
testing
it
internally
on
a
poor,
unsuspecting
team,
and
then
we
could
actually
get
this
to
a
point
where
we
could
encourage
people
to
use
it
and
give
feedback.
But
right
now
it's
not
even
in
a
usable
state
for
that.
So
the
hard
part
is
getting
that
export
pipeline
done
and
things
like
views
and
exemplars
again
entirely
untouched.
So
there
is
a
bit
of
work
here
still.
A
H
Companies
I've
talked
to
because
I've
like
through
this,
the
ncf
slack
I've
gotten
some
messages
from
other
companies
and
it
seems
to
be
pretty
commonly
used
statute.
There
is
the
discourse
Prometheus,
something
gem,
it's
kind
of
tricky
like
scrape
style
metrics
for
Ruby
applications
because
they
do
the
whole
pre-forking
process
and
then
so,
if
you
scrape
it,
you
might
scrape
a
particular
like
process
that
doesn't
have
any
of
the
information
from
any
of
the
other
requests
and
then
the
next
time
you
scrape
you
get
it.
H
So
you
get
this
like
round
robin
massive
incorrect
metrics,
so
you
effectively
have
to
set
up
some
sort
of
like
sidecar
or
whatever
you
want
to
call
it
something
where
they're
you're
pushing
metrics
to
this
other
thing
that
then
gets
screened
and
kind
of
circling
back
to
the
spec
conversation
of
like
scrape
targets
being
an
indicating
indicator
of
uptime.
It's
like
well
the
thing
you're
scraping
this
up,
but
that
doesn't
mean
the
application
is
so
maybe
that's
not
like
whatever
I'm
not
gonna.