►
From YouTube: 2022-12-06 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
C
A
B
C
F
Not
so
bad
I
was
about
to
say,
I'm,
sorry,
I'm,
monitoring
a
deploy
on
the
other
half
of
the
screen.
So
forgive
me
if
I'm
a
little
distracted.
C
B
So
yeah
it
is,
it
is
the
usual
time
the
crowd's
a
little
bit
light,
but
probably
not
unusual.
For
this
time
of
the
year,
welcome
Andrew
I'll
go
ahead
and
share
my
screen
go
to
the
specsic
recap,
which
was
also
pretty
light,
so
we
will
have
plenty
of
time
to
talk
about
whatever.
B
All
right
very
short
meeting
there
was,
there
are
PR's
to
add
a
couple
of
of
attributes,
which
is
adding
an
open
shift
as
a
cloud
provider.
B
Santosh
had
a
big
question,
but
the
meeting
ended
well
before
8
30.,
so
we
do
not
talk
about
it.
I
think
we'll
we'll
talk
about
it
next
week.
Maybe
the
most
interesting
thing,
and
it's
not
even
all
that
interesting-
is
that
there
is
a
PR
for
adding
attribute
limit
configuration
for
log
records.
B
We
had
this
for
spans
already
there's
kind
of
been
like
a
moratorium
on
adding
new
environment
variables,
because
there's
too
many,
but
since
this
one
is
basically
just
the
log
version
of
what
we
already
have
for
spans
I
think
people
were
cool
with
it
and
that
was
it.
B
Return
forward
and
uninteresting
so
here
we
are
on
our
pretty
empty
agenda.
Wait
but
not
completely
of
the
agenda.
D
D
So
there
was
I
I'm,
probably
not
the
best
person
actually
to
present
this
context,
I'll
give
a
very
short
summary
and
then
maybe
Sam
can
jump
in,
but
essentially
we
merged
changes
into
the
instrumentation
that
started
running
a
code
path
by
default
that
was
previously
not
exercised
by
default,
and
this
code
path
turned
out
to
tough
problems
in
that,
in
some
instances
it
could
actually
mess
with
rails,
routing
and
wrap
requests
to
different
endpoints
than
what
you
expected
and,
generally
speaking,
that
is
probably
a
bad
thing
anyways.
D
What
I
wanted
to
talk
about
today
was
basically
should
we
stop
monkey
patching
rails
and
now
that
rail
7
has
a
lot
better
support
for
guaranteeing
that
active
support.
Notifications
can
get
delivered
and,
like
you
know,
one
subscriber
can't
necessarily
crash
all
of
them
wondering
if
this
isn't
the
time
to
start
actually
pushing
towards
notification
based
instrumentation
for
rails
rather
than
monkey
patching
the
theory
being.
D
D
F
Yeah
I
guess
like
so
this
is
Robert
is
just
listening.
If
that's
okay
for
our
viewers
on
the
Tito's
YouTube
cast
Robert
has
made
a
comment
that
says
that
it's
he's
just
listening
Yeah.
So
basically,
this
is.
This
code
was
actually
super
old.
It
just
wasn't
on
by
default
this
this
method
and
yeah
I,
guess
the
the
main
thing
is
like
we're
mucking
about,
as
Robert
might
say,
we're
mucking
about
in
in
gem
internals
using
undocumented
apis,
and
we
were
not
attentive
to
side
effects.
F
So
we
did.
We
talked
about
this
a
bit
internally
and
we
were
like
oh
it'd,
be
really
cool
if
we
could
run
Hotel
Ruby
tests
against
the
rails
test,
all
the
rails,
gems,
test
Suites
using
you
know,
Edge
or
whatever
the
latest
on
Maine.
A
F
So
that's
like
one
thing
we
thought
about,
but
anyway
that's
maybe
that's.
My
take
is
like
oh
yeah.
There
were
side
effects,
there
might
be
other
things
that
have
side
effects
that
we
don't
know
about,
might
be
cool
to
run,
specs
against
rails
main
test
Suite.
B
Yeah,
so
as
to
the
original
kind
of
suggestion
that
you're
making
Andrew
about
trying
to
not
monkey
patch
rails,
if
we
can
get
away
with
it
and
using
access
support,
notifications,
I
think
that
is
definitely
a
a
good
idea
moving
forward.
B
So
the
issue
with
access
support
notifications
is
that
I'm
kind
of
stating
something
and
hoping
to
be
corrected
on
this
one
is
that
if
you,
if
you
have
an
unhandled
exception
somewhere
in
there,
you
might
not
get
like
the
end
event
for
something
in
progress,
and
you
end
up
with
just
a
unended
span,
is
that
is
that.
D
D
What
we
did
with
our
active
support
notifications
in
the
past,
because
some
of
the
instrumentation
does
use
it.
We
basically
would
rip
out
the
entire
span.
D
Sorry,
what
am
I
trying
to
say
there's
basically
like
a
notification
fan
out
class
inside
of
rails,
and
we
would
actually
replace
that
with
our
own
in
a
way
that
made
sure
it
was
less
likely
to
crash.
So
it
gave
us
a
little
bit
more
safety,
so
we
might
still
be
able
to
rely
on
that
for
older
versions
of
rails,
but
we
could
also
gatekeep
this
to
only
like.
You
know,
rail,
7
and
up
you
use
the
new
safer
instrumentation
rails.
You
know
six
and
Below.
D
You
have
some
monkey
patching
and
you
know
maybe
that
instrumentation
just
kind
of
gets
Frozen
in
Time,
but
yeah
I
I
believe
on
rail
7.
It
should
be
safe
to
do
really.
The
hardest
part
is
making
sure
that
we
are
like
first
in
the
chain
of
notifications,
to
make
sure
that,
like
our
span,
would
capture
as
much
of
the
timing
as
possible.
D
And,
of
course,
everything
needs
to
emit
a
notification
in
order
for
it
to
work
and
I.
Think
that's
largely
the
case,
but
I
would.
We
would
need
to
audit
that
as
well
to
make
sure
that
we're
not
going
to
like
Miss
stuff
or.
D
B
B
I
know
it's
like
a
pain,
and
you
know
it's
is
it
paying
the
pr
to
rails
when
we
need
to
change
something
there,
but
I
think
you
know
if,
if
we
did
it
in
a
way
where
we
were
being
constructive,
I
think
it
would
be
welcomed
ultimately,
and
the
other
pain
is
just
that,
then
you
can't
rely
on
it
until,
like
you
know,
three
or
four
years
have
passed
and
people
are
on
a
version
with
with
these
changes,
but
at
the
same
time,
but
we
should
still
start
doing
it,
because
I
think
this
will
be
ultimately
if
rails
ever
does
have
native
support
for
open,
Telemetry
I
would
see
it
coming
through
actually
support
notifications,
more
than
anything
and
probably
be
a
layer
on
top
of
that,
so
helping
to
set
up
rails,
for
this
makes
a
lot
of
sense.
B
I
think
it
does
probably
require
some
research
to
make
sure
that
things
are
actually
fixed.
Today
and
if
not
the
we
should
try
to
help
fix
them
or
actually
support
notifications,
and
the
thought
that
I
had
is
that,
depending
on
what
changes
were
made
to
actually
support
notifications
in
order
to
fix
this,
this
issue
with
exceptions
and
the
flow
it
might
be
possible
to
just
monkey
patch
active
support
notifications
for
older
rails
to
have
the
same
behavior
and
then
just
use.
D
I
guess,
like
you
know,
we
want
to
do
less
monkey
patching,
but
one
one
place
to
Monkey
patch
is
perhaps
better
than
multiple,
but
that
is
that's
kind
of
similar
to
what
we
were
doing
for
the
like:
the
active
dispatch
or
whatever
the
router
one
is
I
forget
or
not
not
router.
The
action
view
action
view
that
one
was
notification
based
from
the
beginning
and
we
and
we
were
essentially
monkey
patching
the
active
support
notification
stuff
to
make
it
a
little
safer.
So
we
already
had
precedence
for
that.
D
I
think
it'd
be
fine
to
keep
doing
that.
I
will
open
an
issue
copying
or
capturing
what
we
just
talked
about
and
then
maybe
we
can
start
looking
at
it,
maybe
breaking
off
some
chunks
of
it.
Crush
some
code
Crush
some
code,
maybe
someday
in
this
parallel
hypothetical
world,
where
I
have
more
time.
B
You
could
say
Tito's
Christmas
dinner.
It
will
never
get
boring
just
throwing
that
in
there
yeah.
D
Paige,
do
you
know
what
this
is
also
hi
I,
don't
know
if
there
were
introductions.
First,
it's
nice
Avenue
face,
but
also
I
realize.
Maybe
you
don't
know
the
in-joke
about
Dido's
vodka.
A
No
I'm
familiar
with
Tito's
vodka
from
being
a
college
student
when
Tito's
came
on
the
scene,
but
tell
me
what
is
its
significance
to
this
group.
B
F
B
I've
been
waiting
for
for
them
to
live
up
to
their
end
of
the
sponsorship,
but.
D
F
It's
sort
of
like,
if
you
remember,
when
influencers
the
in
a
bid
to
seem
more
legit,
they
would
do
fake
sponsored
posts,
be
like
I,
Love,
My,
Adidas
sneakers.
Thank
you
so
much
to
Adidas
for
giving
these
to
me
and
that
was
fraud
and
I
guess
yeah,
wait
to
clarify
this
is
a
joke,
we're
not
sponsored
by
tdos.
So.
A
D
A
D
D
Fine,
that's
fine!
Since
I'm
since
I'm,
currently
hogging
the
mic,
I'm
Andrew
I
work
at
Shopify.
Previously
it
was
at
GitHub
got
involved
with
open,
Telemetry
and
then
somehow
now
my
day,
job
is
actually
not
focused
on
tracing,
but
I
am
still
trying
to
to
do
as
much
yeah
yeah
I'm
sad,
actually
very
sad,
but
I'm
still
trying
to
be
as
involved
as
possible,
and
that's
me
a
lot
of
my
co-workers
on
the
call.
F
Yes,
I
I
think
maybe
we've
met
on
a
previous
one
of
these
calls,
but
I'm
saying
I
work
at
Shopify
I
try
to
review
BRS
for
the
open,
Telemetry,
Ruby,
contrib
repo,
as
best
I
can
and
that
you
know
do
my
I.
Try
I
try
my
best
Robert
also
works
at
Shopify,
but
he's
just
chilling,
so
he
doesn't
have
to
say
anything.
A
I
can
go
next.
I
am
Paige
Paige
Cruz
I
work
over
at
chronosphere,
where
all
of
the
observability
companies,
like
you
know
what
is
the
pitch?
A
Basically
we're
a
very
nice
UI
and
back
end
for
open
source
instrumentation,
so
Prometheus
metrics
send
them
to
US
Open,
Telemetry,
traces
and
one
day
logs,
send
them
to
us,
and
we
do
some
stuff
I've
been
following
open
Telemetry
since
the
baby
days
when
we
had
open
tracing
versus
open
metrics
when
I
worked
at
New
Relic
on
the
tracing
team
and
now
five
years
later,
I'm
ready
to
finally
put
on
my
big
boy
pants
and
come
contribute
and
be
an
open.
You
know
really
get
into
this
world
of
Open
Source.
A
Instead
of
just
being
an
end
user
I
want
to
take
a
very
more
active
role,
because
I
think
this
is
the
best
place
to
find
a
long-term
Tech
home
with
as
many
people
that
we've
got
collaborating
in
companies.
Oh,
it
tells
where
it's
at
and
it's
not
going
anywhere
and
it's
only
getting
a
bigger
footprint.
So
why
not
be
a
part
of
that
and
Matt?
We've
worked
together
at
light
step
and
New
Relic,
but
give
me
give
me
your
intro
anyway
for
the
call
for
posterity.
A
B
I
made
some
contributions
to
open
Telemetry
Ruby
in
the
early
days
now,
I
normally
just
ramble
at
these
meetings,
but
I
I
would
like
to
play
more
of
a
role
at
some
point
in
time.
It's
just
yeah
work
is
crazy
and
your
roles
and
responsibilities
are
always
changing
and
that's
kind
of
kind
of
the
situation,
but.
F
You
contribute
to
other
Hotel
projects,
or
are
you
like
crushing
code
internally.
B
I
was
writing
code
internally.
I
am
writing
I'm.
Sadly,
writing
a
little
bit
of
JS
right
now
and
I
have
done
some
little
little
bits
of
work
on
the
collector
here
and
there,
but
I
don't
know
that
that's
much
needed
any
longer,
so
so
yeah
I
feel
like
figuring
out
exactly
what
I
do
is
also
a
goal
of
mine,
yeah.
F
C
C
F
A
B
B
But
you
could
you
could
write
some
code
for
if
you
want
at
least
at
the
time,
I
think
now,
all
their
stuff
is
legitimately
open
sourced,
but
I
did
start
working
on
open
tracing
Ruby
when
I
was
at
New
Relic
and
that
was
that
was
actually
yeah.
B
I
I
definitely
saw
the
need
for
kind
of
retracing
ecosystem
to
be
non-proprietary,
and
you
just
have
like
a
community
set
of
tools
for
this
stuff
and
I
was
I,
was
around
when
open,
tracing
and
open
census
merged
to
the
form
open
Telemetry.
So
that's
like
the
longer
history
and
then
yeah
I
think
Francis
was
around
the
beginning
of
this
project
and
then
I
was
around
I'm.
Sure
Francis
was
disappointed
to
have
me
holding
him
back
in
the
early
days,
but
that's
that's
how
it
started.
D
And
he
he
still
drops
in
from
time
to
time.
We
we
talk
to
him
a
lot
about
hotel
stuff,
but
he's
he's
a
very
busy
man
too.
So
he's
sort
of
into
meetings,
not
in
others,
but
I
wanted
to
ask
patrons
who
said
you
want
to
like
start,
you
know
finding
an
open
source
home
and
working
in
your
presence
at
this
meeting
means
you're,
probably
like
Ruby
I
would
assume.
D
Are
there
any
areas
you
wanted
to
work
on
like
because
we
definitely
could
use
help
like
we've
got
a
million
and
a
half
things
that
we
want
to
do
and
a
small
handful
of
maintainers
who
don't
have
time
to
do
much
of
it?
So
is
there
anything
you
were
thinking
about.
A
Yeah,
so
it's
very
selfishly
my
personal
love,
like
personal
goals,
that
I
see
overlapping
with
open
source
contributions,
is
the
biggest
Gap
that
I
see
for
observability.
A
Right
now
is
I
can
get
in
a
group
of
a
room
like
this,
and
we
can
talk
shop
and-
and
we
understand
spans
traces,
exemplars
cardinality
when
I
go
to
talk
to
my
friends
that
are
not
deep
in
this
world
of
observability
haven't
even
heard
of
you,
know,
open
tracing,
open
census,
I
see
this
really
big
gap
and,
and
part
of
me
thinks
it's
not
that
there's
not
101
level
content,
but
that
content
is
not
reaching
people.
The
way
that
it
needs
to.
A
So
one
of
the
things
I'm
going
to
do
next
year
is
have
an
intro
to
instrumentation
Workshop
I,
take
to
like
kubernetes
Community
Days
shop
around
just
really
publish
it
open
source.
Anyone
can
use
it
where
I
I
want
to
contrast
what
you
get
with
auto
instrumentation
gets
you
80
of
the
way
there
super
cool.
Oh
my
God
add
some
manual
instrumentation.
Let
me
show
you
how
to
put
an
attribute
on
a
span
or
baggage
or
whatever
we
call
it.
A
These
days
and
see
how
much
more
powerful
it
is
and
I
think
giving
people
that
direct
comparison
and
showing
it
is
not
that
hard
to
manually
instrument
for
the
benefits
you
get.
That
is
something
I'm
looking
to
do,
probably
with
the
hotel
demo.
App
I
want
to
use
as
much
of
the
stuff.
That's
already
out
there
as
possible,
so
I
can
contribute
back.
But
as
far
as
this
goes
I,
the
skills
I
can
bring
to
the
table
that
wouldn't
have
me
learning
on
the
job
would
be
bill.
D
A
I
can
do
that
stuff
and
I.
It's
been
a
few
years
since
I've
professionally
written
Ruby,
so
also
throw
me
the
I'm
happy
to
start
taking
the
easy
medium
stuff
things
that
would
be
a
waste
of
your
time
and
work
up
to
some
bigger,
bigger
stuff.
When
I'm
a
little
transparently
nervous
about
is
I,
merge,
something
and
then
Shopify
you
know
deploys
it
like.
Oh
my
God
I,
don't
I
want
to
be
able
to
buy
my
roller
skates
right,
I,
don't
want
to
take
Shopify
down.
C
D
D
D
I
use
it
internally.
A
lot
when
I'm,
just
like
trying
to
test
something
and
I'm
like
I,
need
a
crap
ton
of
spans
that
all
look
different
I
know
I'll
fire
up
the
web
store.
So
as
far
as
like
kubernetes
and
yaml
Engineering,
we
don't
actually
have
a
ton
of
that
in
this
sig
yeah,
the
most
yaml
engineering
I
think
we
have
is
in
our
GitHub
actions,
workflow,
which
by
all
means,
if
you
have
any
desire
like
it,
is
a
place.
We
tend
not
to
go
unless
we
absolutely
have
to.
A
Because
I've
been
seeing
that
stuff
about
Trace
tests
being
open
source
instrumenting,
your
CIA
CI
CD
stuff,
which
could
be
yeah
yeah.
D
That's
okay:
my
brain
was
going
a
million
different
ways.
Anyways
de-risking
releases
is
something
like
we
were
talking
about
internally
at
Shopify
this
morning,
because
yeah
something
did
get
merged
and
it
did
cause
a
problem
like
we.
We
got
it
but
yeah
anything
that
adds
safety
to.
It
is
certainly
welcome.
I
I,
don't
know
it's,
you
know
open-ended,
so
whatever
it
looks
like,
but
don't
feel
bad
about,
like
you
know
not
having
written
Ruby
for
a
while,
like
you
know,
we'll
we'll
help
get
PR
as
to
where
they
need
to
be
so
like.
D
A
D
Any
issues
you
see
that
seem
interesting,
my
my
perennial
plug
is
the
spec
compliance
issues.
There
are
usually
a
lot
of
like
little
things
to
change
in
the
API
of
the
SDK
that
are
like
sort
of
bite-sized,
so
any
of
those
you
want
to
grab
always
welcome.
Oh.
C
D
D
Also
oh
I
was
gonna,
say
documentation
too,
like
if
you're,
if
you're
interested
like
because
you
mentioned
the
web
store
and
the
web
store
the
Ruby
component,
there
is
in
Sinatra,
because
I
could
bang
that
out
in
like
an
evening
while
watching
TV,
rather
than
anything
more
interesting,
if
you're
interested
at
all
in
like
making
the
Ruby
component
of
that
look
better
yeah
great
by
all
means
like
super
happy
to
do
that
too,
and,
like
our
own
internal
examples,
are
not
great.
D
A
Nice
and
they're
so
I
mean
gosh
rails
is
I,
mean
I
GitHub
I,
like
I,
mean
yeah
all
the
places
y'all
have
worked
like
New
Relic
has
Ruby
rails,
running
okay,
okay,
I
got
a
nice
list
here
and
we
meet
every
Tuesday
or
every
other
Tuesday.
Every.
A
D
F
Yeah,
there's
I,
there's
a
cup
there's
a
couple
good.
First
issues
that
I've
made
somewhat
recently
that
that
Docker
C
group
detector
might
be
good.
Actually
I
didn't
make
that
one.
But
if
you're
in
the
obsolete,
SRE
type
mindset
that
could
be
I
mean
that
that
benefit
a
lot
of
people.
So.
C
B
Today,
maybe
something
interesting
will
show
up
and
then
I
think,
suggestions,
I,
think
the
Sig
is
always
open
to
suggestions
if
there's
something
like
I
would
really
like
to
do.
X
and
I
think
this
would
improve
the
project.
B
Yeah
comp
suggests
and
yeah
most
likely.
If
it's
a
good
idea,
we
will
support
it
and
we
can,
you
know,
make
make
an
issue
to
track
it.
Etc.
B
Yeah
and
then
there's
the
the
core
repo.
This
is
more
kind
of
like
the
the
API
and
an
SDK
machinery.
A
And
speaking
of
the
the
Truffle
Ruby
stuff
is
there,
I
was
looking
at
Argo
CDs
project
and
I.
Think
last
year,
cncf
paid
for
a
a
pen
test
and
like
reading
the
report
was
really
really
fascinating.
I'm
curious,
if
we
do
anything
formal
like
that
or
it's
kind
of
like
security,
is
everybody's
job.
Let's
be
aware,.
A
Was
another
talk
I
wanted
to
do?
Is
this?
You
know
we're
pushing
all
this
open
source
stuff
on
you.
How
secure
is
it
to
run
some?
You
know
we're
running
this
stuff
right
alongside
yel's
code
and
I
want
to
make
sure
people
feel
confident,
not
just
in
us
being
on
guard,
but
we're
still
a
baby
project
right
or
that's
something
that
could
come
later.
Eh
we'd.
A
C
D
Trying
to
think
what
oh,
the
name
of
this
is
eluding
me.
It
was
an
acquisition.
Github
did
and
it
was
like
some
sort
of
structural
analysis
of
the
code
to
try
and
point
out
problems,
including
security
vulnerabilities.
Oh
now,
I
can't
remember.
If
the
life
of
me
what
this
product
was
called,
we
integrated
it
here
at
one
point
and
then
code,
ql
wait.
Why
does
that
sound?
Familiar?
Okay,
never
mind
I'll,
think
about
what
it
was.
B
D
Yes,
we
did
a
little
bit
where
we
were
like
this
is
cool
GitHub
provides
it
we'll,
try
it
sure,
and
that
was
about
as
far
as
we
ever
got
like
okay
I'm
sure
we
would
be
open
to
something
a
little
bit
more
formal
or
even
even
testing.
That's
trying
to
do
like
okay,
so,
like
speaking
of
testing
like
fuzzing,
might
be
an
interesting
thing
that
we
would
be
kind
of
interesting
static
analysis
like
I'm
sure
none
of
us
would
really
say.
No
to
that.
D
Also,
though,
like
more
broadly
like
performance
testing
is
something
we've
been
talking
about
too.
There
was
a
change
recently
that
made
the
memcache
the
dolly
instrumentation,
significantly
faster,
at
least
in
corner
cases,
where
you
were
inserting
very,
very
large
values,
and
it
was
by
significantly
I
mean,
like
eye-boggling
amounts
of
speed,
increased
most
people
didn't
notice,
but
like
we
do
have
Performance
Corner
cases
like
that,
I
think
we're.
D
You
know
that
type
of
infrastructure
and
setting
that
up
enrollment
like
I,
don't
know
how
to
do
it,
and
so
yeah
we're
really
open
to
just
about
anything
that
might
be
cool
or
might
help.
One
of
the
main
concerns
we
have
is
just
that:
there's
just
not
a
lot
of
us
maintaining
it.
So
you
know
that's
the
the
barrier
to
a
lot
of
this
is
like
yeah
I
would
do
it,
but
I,
don't
know
how
and
I
don't
have
time.
B
Yeah
so
I
think
in
addition
to
all
this,
like
there
are
some
projects,
kind
of
kicking
off
or
ongoing
in
the
larger
Hotel,
Community
I,
think
I
think
maybe
I
mentioned
last
week,
but
the
last
week
yeah
so
semantic
conventions
for
instrumentation
is
like
a
big
deal.
B
No
instrumentation
has
is
currently
GA,
it's
all
kind
of
like
experimental
and
it
kind
of
has
a
status,
because
because
we
expect
things
to
change
a
little
bit
and
I
think
open,
Telemetry
is
actually
being
quite
ambitious
in
this
area
and
I
think
it's
it's
good,
I
I
think
the
the
reasoning
behind
it
is
very
good
I
think
in
practice
it
makes
things
quite
hard
and
I
guess
the
things.
B
The
thing
that
is
good
and
also
hard
is
trying
to
unify
kind
of
the
output
of
Telemetry
from
you
know:
instrumentation
between
languages.
So,
like
you
know,
HTTP
client
instrumentations
between
languages
should
kind
of
create
generally
the
same
spans
with
generally
the
same
attributes
on
it.
Under
the
same
circumstances,
your
HTTP
server
instrumentation,
you
should
do
the
same
and
generally
traces
should
kind
of
look
and
feel
the
same
between
different
languages.
B
There
shouldn't
be
like
every
language
is
off
on
its
own
path,
doing
its
own
things,
which
has
generally
kind
of
been
it's.
Definitely
my
experience
in
Vendor
land,
how
things
were
kind
of
working.
It
was
hard
to
kind
of
because
it's
so
hard
to
get
everybody
to
agree
on
things.
People
just
go
off
and
do
it
on
their
own,
and
you
end
up
with
a
bunch
of
different
things,
so
semantic
conventions
for
instrumentation
I
think
you
know
right
now
we
are
kind
of
in.
We
are
in
this
area.
B
Where
we've
all
the
cigs
have
done
a
lot
of
work
to
create
instrumentation.
We
all
kind
of
have
followed
What
semantic
conventions.
Are
there
but
I
think
we've
in
addition
added
some
specialness
I
think
between
the
different
languages
and
with
the
semantic
conventions
project.
The
idea
is
to
is
to
make
sure
that
everything
is
uniform
and
then,
ultimately,
there
is
kind
of
this.
This
plan,
this
timeline,
I,
don't
know
when
Ted
presented
this.
B
B
So
he
came
up
with
this
schedule
and
a
group
to
kind
of
like
focus
on
it
to
try
to
at
least
get
things
done
by
July,
1st,
2024
and
I.
Don't
know
he
has
through
Q3
2023
different
domains
to
kind
of
work
through
and
I.
Don't
know,
I,
don't
know
what
the
additional
time
is
is
probably
just
a
buffer
zone
which,
which
is
good
to
build
in
but
anyways
as
so
each
quarter.
B
I
think
some
domains
should
have
like
semantic
conventions
that
are
firming
up,
if
not
trying
to
reach
GA
and
what
that
will
mean
as
well
to
kind
of
take
a
pass
through
our
instrumentation
make
sure
that
we're
recording
sounds
where
we're
supposed
to
make
sure
that
we
have
the
right
attributes
there.
If
there
is
like
some
additional
stuff,
we
need
to
kind
of
figure
like
stuff
that
we
have
that
we're
recording.
That's
not
specify.
B
These
are
things
that
we
need
to
figure
out
to
do
with
so
so
I
said
this
is
a
whole
stream
of
work
that
is
definitely
accessible.
I
feel
like
it
would
be
fun
for
a
little
while
and
then
maybe
start
to
drive.
Somebody
crazy,
but
at
any
point
in
time,
I
think
if,
if
all
of
us
can
pitch
in
and
do
a
little
bit
of
it
like
it,
it
is
definitely
something
that
that
will
need
to
be
done
so.
B
E
D
Yeah
we
could
do
that
dance,
I
didn't
want
to
do
it
because
it's
annoying,
but
that
doesn't
mean
it
shouldn't
get
done.
I
can
open
those
release.
Prs
yeah,
no
I,
know
I,
know
all
right.
Okay,
where
are
our
release?
Tooling,
really
does
need
some
help.
That's
another
big
Paige,
if
you're
familiar
with
really
stooling-
and
you
have
thoughts,
there's
an
open
issue
in
the
contribute,
though
about.
D
Why
is
this
painful?
What
can
we
do
better?
It's!
It's
we've
got
some
automation
that
works,
but,
like
it's
there's
just
a
lot
of
manual
steps
and
a
lot
of
like
dependencies,
and
things
like
you
have
to
do
things
at
a
very
specific
order
for
it
to
work
right,
most
of
which
is
sort
of
just
knowledge.
We
pass
around
by
word
of
mouth
as
we
and
we
frequently
forget
and
yeah
I
I
do
certainly
but
yeah
I
real
I
can
go
open.
D
I
can
go
start
the
process
and
start
kicking
the
PRS
off
and
stuff
and
get
that
going.
We
we
should
all
right.
E
D
Figured
if
it's
unsafe
for
one
spot,
it's
probably
unsafe
for
all
of
them,
you're
welcome
many.
E
D
A
D
Don't
know
if
we
have
anything
else
really
super
to
talk
about
Ariel.
One
thing
we
were
we
did
discuss,
which
you
might
have
opinions
on
and
I
was
about
to
open.
I
was
typing
out
an
issue
for
is
it's
time,
maybe
to
bite
the
bullet
and
move
our
rails
instrumentation
to
be
notification
based
beyond
the
theory
that
it
is
probably
always
going
to
be
safer
than
monkey.
Patching
large
code
bases
that
we
don't
fully
understand
and
the
general
consensus
was
yeah.
D
D
But
I
don't
know
if
you
had
any
thoughts
on
that
before
we're.
You
know
settled
like
we
had
much
else
to.
E
D
D
So
the
implementation
wise
it
would
work
essentially
the
same
way
that
the
action
view
stuff
works
today,
which
is
it's,
got
a
custom
listener,
because
you
don't
have
to
just
provide
like
a
callback.
You
can
actually
provide
a
class
that
listens
to
the
start
and
the
Finish
hooks.
Okay,
and
we
do.
We
start
the
spin
and
we
store
the
context
key
and
we
do
some
magic
and
then
we
finish
it
in
the
finish.
So.
E
E
Is
it,
are
you
going
to
be
our
project
manager
on
that
one
Hayworth,
and
can
we
split
it
up
and
jump
on
in,
like
yes,.
E
Okay,
thank
you
because
a
very
curious
thing,
I
noticed
this
morning,
is
that
we
had
a
huge
spike
in
find
by
sqls
that
had
no
child
database
spans
attributed
to
them
and
I.
Think
it's
because
it's
like
going
through
find
by
SQL
but
hitting
the
cache.
So
it
doesn't
end
up
hitting
the
driver
and
causing
the
driver
to
trigger
a
child
span.
So
I'm
very
curious
about
switching
to
as
notifications
for
this
interesting.
D
Yeah
I
don't
know
if
I'm
you
know,
assuming
making
several
assumptions
and
leaping
to
places
that
may
not
be
warranted.
I,
don't
know
if
you
would
have
seen
a
span
underneath
it
in
the
notification.
Only
world
or
not
I
mean
if
it
never
hit
the
driver.
I
mean
it
actually
makes
me
wonder
like.
Would
you
have
ever
seen
any
expand
at
all
for
something
that,
like
I,
don't
know
when
active
support
fires,
its
notification,
for
that,
like
I.
D
B
Yeah,
the
one
thing
that
I
remember
with
active
record
and
the
notifications
that
were
around
is
that
there
was
still
some
monkey
patching
that
needed
to
be
done
in
order
to
bully
get
visibility
into
all
the
operations
that
you
want
in
this.
So
I
think
we
may
find
that
as
we
move
towards
access
support
notifications
that
we
might
need
to
see
if
there
might
be
some
there
might
be
some
notifications
missing.
B
So
we
might
need
to
go
through
the
process
of
trying
to
introduce
those
into
rails
itself,
but
but
again,
I
think
that's
fine,
I
think,
ultimately,
whatever
we
can
do
to
improve
active
support
notifications
to
make
it
entirely
usable
for
for
otel
will
be
better
for
rails,
better
for
active
support,
notifications
and
ultimately
kind
of
pave
the
way.
I
think
ports
having
a
seamless
way
to
kind
of
integrate,
otel
into
rails,
possibly
natively
and
and
yeah
I.
B
Think
one
of
the
other
thoughts
and
suggestions
was
that
I
think
yeah
and
Andrew
was
kind
of
mentioning
that
there
was
this
flaw
with
X
support
notifications.
Where,
sometimes
you
will
get
your
start
event,
and
if
there
was
an
exception
that
happened,
your
end
event
will
never
come.
So
you
end
up
with
this,
like
lingering
started
span
with
no
end,
and
this
is
the
thing
that
we
need
to
verify.
It
was
fixed
in
rail
7.,
but
if
it
was
fixed,
yeah.
B
D
That's
a
good
question:
I,
don't
know!
That's.
B
Guess
we'd
actually
see
if
rails
will
back
Port
those
changes
to
maintain
versions,
anyways
and.
E
So
what
I
would
say
from
that
point
is
we
can
make
a
decision
and
say
that
for
our
rewrite
of
these,
your
pause,
like
a
specific
version
of
this
six,
whatever
support
stops
at
like
version
38
or
whatever
of
the
Rails
instrumentation
and
39
going
forward,
is
going
to
be
seven
on
right
and
that's
that.
B
E
B
E
Hey
if
it's
okay,
I've
got
one
more
follow-up
before
I
have
to
vanish
right,
and
that
is
because
it's
gonna
be
related,
also
because
you
were
talking
about
the
semcom
project
that
Ted
proposed
and
I'm
talking
about
some
time
and
some
internally
for
the
instrumentations
and
if
we're
revisiting
active
support,
I
would
love
for
us
to
come
up
with
our
own
internal
conventions
for
at
least
rails
or
MVC,
or
something
along
those
lines,
and
someone
had
opened
the
pr
requesting
that
we
add
metrics
values
as
tags
for
span,
attributes
and
I
kind
of
said.
E
Hey.
These
are
metric
values.
We
probably
it
is
I
would
do
not
agree
that
we
should
add
them
as
span
tags.
I
never
responded
to
the
person
because
I
won
on
vacation
and
my
family
got
sick
and
so
on
and
so
forth.
But
I
was
wondering
if
folks
wanted
to
chime
in
on
this
and
tell
me
you
know
whether
I
should
why
I
should
change
my
mind.
So
that's
what
number
204
there.
D
I
have
thoughts,
I.
Think
I,
responded
to
your
comment.
I
do
actually
disagree
and
think
you
should
change
your
mind,
but
specifically
that's
because
we
don't
have
a
metrics
SDK
that
we
can
ship.
Yet
it
is.
You
are
correct
ultimately,
but
in
the
interim
those
specific
attributes
seemed
really
useful
to
me
and
I
could
see
why
someone
might
want
those
ones.
I
would
be
comfortable
if
we
said
like
this
is
an
exception.
We're
not
going
to
do
this.
D
If
somebody
comes
up
and
adds
more
we're
going
to
like
look
very
very
carefully
at
this
and
see
like
do
you
have?
Is
it
a
very
good
reason
to
do
so,
but
I'm
also
I'm
also
going
to
say
I
only
hold
that
opinion
maybe
moderately
strongly
like
it's
not
it's
not
a
super
super
strong
feeling,
it's
sort
of
like
those
ones
might
matter
a
lot
specifically
but
I,
don't
know.
D
E
So
yeah
my
take
on
this
is
that
if
the
vendor
doesn't
support
a
way
that
is
like
customized
to
say,
like
I,
want
to
see
a
particular
span
and
all
of
his
children
and
then
do
some
sort
of
a
calculation
or
roll
up
of
the
entire
time
spent
on
all
the
child
spans
right,
because
we're
kind
of
doing
that
computation.
This
is
kind
the
person
said.
Hey
this
feature
is
missing
Downstream
in
the
vendor,
I'm
using
so
basically,
what
we're
trying?
E
What
they're
trying
to
do
is
build
a
feature
into
the
instrumentation,
that's
missing
in
their
vendor
implementation
and
so
I
think
to
myself
like
okay,
what
you!
What
you're
trying
to
do
is
capture
that
roll
up.
That
rails
thinks
it
is,
and
then
comparing
those
so
that
you,
you
have
a
quick
shorthand
into
seeing
all
of
your
child
span
timings
that
are
summed
up
right.
E
So
it's
like
every
the
sum
of
all
the
action
view:
child
spans
that
came
from
this
action
controller
Index
right,
which
is
the
the
type
span
server
the
spanking
server
and
then
all
of
the
internal
spans
whose
names
are
action,
view
blah
blah
blah
or
whatever
or
coming
from
active
action
view
instrumentation.
E
Those
are
features
that
are
missing,
I,
think
in
the
vendor,
and
not
something
that
necessarily
should
be
exported
by
us
now,
if
we
exported
them
as
metrics,
because
their
next
thing
was
hey.
If
we
export
these
as
metrics
I
want
really
high
cardinality
attributes
was
the
thing
that
they
were
also
saying
like
they're
like
they
disagree,
it
should
be
exported
as
metrics
because
of
that
because
they
want
to
be
able
to
narrow
down
what
view
is
really
slowing
down
associated
with
all
these
other
attributes.
E
Allegedly,
the
exemplars
are
gonna
fix
that,
but
that
feels
again
more
like
a
missing
vendor
feature
of
like
hey
the
vendor
doesn't
know
how
to
interpret
a
roll-up
of
all
the
things
that
happen
in
the
child
like
right,
like
a
vendor
can
say
you
know
you
can
say
the
same
thing
for
SQL
time.
How
much
SQL
time
did
I
spend
in
this
action?
View?
E
Well,
the
vendor
can
say
from
type
span
server.
You
could
look
at
all
of
the
child
spans
within
the
same
service
boundary
and
then
do
a
sum
of
all
of
the
span
timings
that
where
the
DB
dot
system
is
MySQL
or
postgres,
and
they
can
build
a
visualization
around
that
like
how
would
they,
how
would
their
system
or
vendor
support
this
notion
of
hey,
look
I've
got
an
attribute,
that's
named
DB
and
it's
3650.
E
What
semantic
meaning
does
that
provide
for
the
vendor
to
do
something
like
are
people
gonna,
do
the
vendor
say?
Oh
I
can
do
a
query
on
this
attribute
where
the
value
is
between.
You
know
three
thousand
and
five
thousand
now
they're
building
histograms
themselves
through
right
right
and
this
it's
kind
of
like
you.
C
E
I
I'm
not
keen
on
adding
these
these
metric
type
attributes
on
the
instrumentation.
That's
my
that's!
My
take
I
didn't
write
all
of
that
up
because
God
knows
I
need
my
fingers
past
this
Soliloquy
or
whatever,
but.
D
Well,
that's
a
fair
point:
I
mean
I.
Think
the
trade-off
in
my
mind,
was
like
yeah
it'd,
be
cool.
If
the
vendors
can
do
this,
I
don't
know.
If
anyone
really
can
today
and
like
if
we
tell
this
person,
no,
we
say
no,
your
vendor
should
do
this
or
your
Telemetry
backend
should
do
this.
D
Are
they
going
to
go
and
push
their
vendor
or
Telemetry
back
ends
to
do
that,
or
are
they
just
going
to
work
around
it
and
that's
kind
of
what
I
was
wondering
I
do
agree,
it
should
be
exported
as
a
metric,
even
a
high
cardinality
metric,
and
if
people
don't
like
the
high
card,
Melody,
the
metrics
API
defines
or
SDK
defines
views
that
can
reduce
that
cardinality
before
export.
So
I
do
think
that's
where
it
should
go
eventually.
D
E
E
C
E
D
Ahead
of
time,
why
did
you?
Why
did
you
let
these
ones
in
and
yeah?
That's
a
fair
point.
I
I
think
I
said
what
I
said,
because
those
ones
specifically
I
know
are
very
useful
when
you're
debugging
rails,
but
I
mean
I
I,
hear
what
you're
saying
about
not
setting
a
precedent,
so
I
would
be
comfortable.
Taking
that
stance
as
I
would
I
would
adopt
that
as
my
own
and
say
yeah
we're
we're
not
going
to
do
it
because
you
know
we
don't
want
to
have
to
defend
our
decision
forever.
D
As
you
know,
when
we
deny
the
next
ones
which
I'm
having
big
Recollections
of
HTTP
hooks
right
now
and
how
maybe
we've
learned
our
lesson
here
so,
okay
yeah,
do
you
want
to
update
that
issue?
The
the
pr
with
the
reasoning
or
I.
E
Will
eventually
what
I'll
do
is
I'll
if
I
can
find
this
video
eventually
on
YouTube,
since
none
of
them
are
ever
tagged,
I'll,
try
to
tag
this
conversation
and
put
it
in
there
and
then
provide
the
transcript
of
what
we
discussed
here,
but
I
Andrew.
This
commentation
move
mostly
between
you
and
I
I'm,
hoping
other
people
can
chime
in
and
provide.
B
Yeah
go
ahead.
I
will
at
least
chime
in
to
say
that
this
code
definitely
yeah.
This
is
a
this
code,
could
be
a
standalone,
could
be
something
Standalone
that
this
user
could
it
could
use
in
any
way,
it's
just
adding
a
an
additional
subscriber
that
takes
the
current
span
and
adds
some
attributes
to
it.
So
ultimately
they
have
a
Way
Forward
on
this
I
think
I'm
not
sure.
B
What's
going
to
happen
with
this
semantic
conventions,
project
but
I
feel
like
the
goal
here
is
to
try
to
like
unify
stuff
as
much
as
possible
between
the
languages
and
I
think
the
more
custom
things
that
we're
adding
the
more
things
we're
gonna
have
to
figure
out
during
during
that
project,
because
I
think
there's
going
to
be
some
I,
don't
know,
I
feel
like
there's
going
to
be
some
discussions
where
this
semantic
conventions
group
is
probably
going
to
say.
B
These
are
the
conventions,
no
more,
no
less
and
then
I
think
different
cigs
are
going
to
come
in
and
say
hey,
but
we
have
all
these
other
custom
things
that
we've
added
that
we're
not
willing
to
part
with.
How
do
we
kind
of
deal
with
those
and
I
think
there's
going
to
have
to
be
some
conversations
back
and
forth
to
ultimately
reach
like
the
middle
ground
there,
but
I
don't
know,
that's
just
my
my
senses
of
how
that
project
is
going
to
go,
but.
B
But
yeah
so
it's
hard
to
say
how
we
should
handle
these
things
without
knowing
what
the
expectations
are
coming
from
the
semantic
conventions
group,
because
I
imagine
like
right
now
we
don't
have
a
metrics
SDK,
so
we
are
heavily
span
based,
so
kind
of
the
line
between
one
should
go
on
a
span.
What
should
be
a
metric?
We
don't
have
that
line
because
all
we
have
are
spans,
but
I
think
the
the
semantic
conventions
working
group
will
hopefully
Define
also
metrics.
B
That
should
be
recorded
for
different
instrumentation,
which
I
think
will
be
helpful
for
us,
because
I
think
in
a
lot
of
other
language
cigs
once
they
end
up
having
a
ga,
metrics
SDK,
it's
like
the
pr
start
rolling
in
to
add
all
these
different
metrics
and
the
sigs
have
to
kind
of
figure
out
like
do
we
need
these.
Can
we
add
these
and
it
just
ends
up
being
one
of
these
things?
B
Where
there's
a
lot
of
a
lot
of
discussions
like
this,
to
figure
it
out,
I,
don't
I,
don't
think
anybody
has
the
answers,
because
we
don't
have
enough
Direction,
I
think
from
the
project
as
to
what
should
be
here
and
what
should
not
sorry.
That
was
a
long
follow-up.
F
E
E
All
right,
I
know
we're
already
at
time,
but
I
haven't,
heard
Robert's
voice
and
just
like
a
unmute
and
a
hello
would
make
me
feel
great.
E
Hey
sending
internet
love
over
the
over
across
countries.
E
I
heard
my
man
well,
my
friends,
I'm
sorry
that
I
missed
the
first
half
of
the
meeting,
but
thanks
for
letting
me
barge
in
and
Chit
Chat.