►
From YouTube: 2023-01-31 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
A
I'm
good
I'm
trying
to
get
my
headphones
working,
I
guess
one
second
I,
don't
know
how
to
use
a
computer.
Okay,
test,
microphone,
Zone,
wired.
A
B
What
is
up
Matt
we're
two
days
in
a
row:
dude
how's.
A
Bootstrapping
what
it
match
I
was
actually
just
in
the
hotel.
Javascript
SDK
was
driving
myself
yesterday.
There
you
go,
it's
almost
like
we
work
on
the
same
stuff.
B
Yeah
no
kidding
and
I
need
a
class
I
need
to
just
go
back
a
little
bit.
Did
you
use
the
word
Mensch,
as
in
a
really
nice
guy,
yeah.
A
Kidding
despite
being
from
Colorado
and
wearing
Patagonia
vest,
I
am
Jewish
and
was
born
in
the
Northeast.
So
there
you
go.
A
B
Exactly
yeah
because
I
am
I'm,
Roman,
Catholic
and
I'm
Latino,
so
yeah
this
is,
we
didn't
have
too
many
years
words
flying
around
my
family.
You
know.
B
A
Yeah,
that's
right.
Oh
Mel,.
B
Did
you
see
that
it's
gonna
be
a
history
of
the
World
Part
Two?
No,
it's
on
Hulu!
It's
gonna
be
coming
out
this
spring
I
think.
B
Well,
we
got
an
ice
day
here
in
Austin,
so
everything
in
my
backyard
is
Frozen.
A
B
Schools
are
closed
and
I'm,
not
in
my
office,
because
my
son
is
doing
remote.
Schooling
from
there
got.
A
B
Hence
my
beautiful
living
room
background
is
that.
D
D
B
D
A
D
No,
it's!
It's
all
good
call
me
when
you
can
and
it's
it's
good
to
see
you
so
wow
thanks.
D
So
yeah
I
guess
we
can
get
into
the
specsig
recap,
even
with
the
small
crowd
theater
specific
recap:
this
one
was
actually
kind
of
long.
So,
let's
see
if
I
can
keep
this
focused,
but
the
HTTP
semantic
convention
group
has
kind
of
Spun
up
and
they're
asking
for
I
think
like
15
minutes
of
spec
cigs
to
kind
of
go
over
issues,
so
they
they
came
with
some
issues.
They
definitely
burned
up
the
15
minutes
and
did
not
get
through
all
of
them,
but
I'm
trying
to
recap
the
important
ones.
D
This
was
the
most
important
one
kind
of
a
new
suggestion
for
https
fan,
naming
currently
the
I
believe
this
fact.
It
says
that
it
should
be
the
route
or
the
HTTP
verb,
but
I
think
they
would
really
like
to
have
the
verb
plus
route
if
available
and
yeah
I
think
people
were
thumbs
up
on
that
idea.
D
Although
there's
schema
translations
that
can
happen
and
if
you're
like.
Maybe
we
can
just
upgrade
this
via,
like
a
schema,
translation,
Rule
and
apparently
that
is
not
possible
with
the
current
capabilities
of
schema
translations
but
I
think
all
instrumentation
is
kind
of
experimental
anyways.
So
breakage
is
kind
of
I
think
pop,
possibly
possible
I.
Think.
D
So
yeah
there's
some
more
discussions
like
should
the
HTTP
type
go
in
the
name
like
it's?
If,
if
you
just
get
get
as
the
span
name,
how
do
you
know
it's
not
like
a
redis
fan
or
something
those
bike
sheds
were
had
I
think
ultimately
having
the
hype
in
the
name
was
not
super
popular,
but.
B
B
D
B
Yeah,
just
that
the
the
span
names
have
two
like
for
really
low
granularity
there
right
for
the
for
for
low
cardinality,
like
that's
great
they're,
just
hard
to
reason
about
when
it's
just
like
I,
just
see
a
bunch
of
guests,
because
this
is
when
we're
streaming
through
them,
but
because,
like
even
if
you
look
at,
for
example,
like
other
instrumentation
spans
like
for
databases,
they're
optional,
they
grow
optionally,
depending
on
what
the
driver
provides
for
you.
So
it's
like
the
operation
name,
the
tables
that
are
involved,
blah
blah
otherwise
fall
back
to
this.
C
B
A
But
then
you
get
subscribed
and
then
you
get
notifications
for
like
a
year
and
then
you
feel
guilty
when
you
unwatch
the
issue.
Yeah.
D
A
rough
life
so
I
think
we
ended
up
skipping
this
next
one
but
talked
about
the
forwarded
four
business.
Basically,
some
servers
will
terminate
TLS
and
then
they
will
record
telemetry
and
I.
Think
the
difference
of
whether
or
not
you
see
HTTP
or
https
can
cause
problems
can
cause
worry
and
panic
for
folks,
but
I
think.
Ultimately,
the
yeah
scheme
scheme
should
be
the
actual
scheme.
D
I
think
is
what
people
were
saying
and
that
there
should
be
another
attribute
to
let
you
know
if
it,
if
the
income,
if
the
404
at
a
different
scheme
so
I,
think
the
yeah.
The
original
proposal
was
to
lie
about
scheme
and
say
what
what
the
incoming
or
what
what
deported
for
would
be.
What
the
port
forwarded
for
in
the
scheme
rather
than
the
actual
scheme,
but
I
think
the
tldr
was.
It
should
be
a
separate
attribute.
D
There
is
a
right
now.
There's
an
HTTP
flavor
attribute
and
lumila
was
saying
we
can
ditch
that
attribute
and
use
existing
attributes
net
at
protocol
name
and
NetApp
protocol
version.
Instead,.
D
And
I
don't
know
there
was
one
thumbs
down
from
from
Riley
on
this,
so
he
was
talking
on
his
thumbs
down.
He
was
actually
a
thumbs
up
on
the
idea
but
thumbs
down,
because
there
are
more
attributes.
We
can
do
this
with
I
guess
like
there's
like
http
or
like
you,
aren't
all
that
scheme
or
something
http.s
game,
and
he
wants
it
to
be
urida
scheme
and
just
kind
of
keep
going
more
General
on
everything.
D
B
B
B
Positional
arguments
in
the
protos
then
it'd
be
like
oh
okay,
yeah,
fine,
but
I'm
gonna
be
like
processing
terabytes
of
attribute
names.
You
know.
D
D
I,
don't
think
anything
else
was
discussed
from
the
HTTP
semantic
conventions
because
we
ran
out
we
already
kind
of
ran
over.
At
this
point.
D
D
Basically,
this
person
would
like
to
be
able
to
generate
metrics
from
all
spans
and
came
up
with
two
ways
that
you
could
do.
This
one
is
deferred
sampling.
So,
regardless
of
what
your
SDK
says
still
send
these
things
off
to
a
collector
and
then
the
collector
can
record
the
metrics
and
then
actually
do
drops
fan
if
it
was
not
sampled
I
think
there
are
a
lot
of
problems
with
that
idea,
but
it's
it
is
nit.
D
The
other
way
was
basically
to
have
processors
within
each
SDK
that
will
create
metrics
for
all
spans
I
think
there
actually
is.
Probably
it
sounded
like
there
are
some
SDK
mechanics
that
make
this
impossible,
also
with
way
processors
work,
but
the.
C
D
Idea
is
create
all
of
your
metrics
in
process
and
then
export
those
I
think
yeah.
Those
those
were
the
ideas
presented.
Josh
McDonald
was
saying
that
you
can
have
the
best
of
both
worlds,
actually
with
probability
sampling.
D
As
long
as
you
have
the
probability
on
the
span,
you
know
how
how
much
of
a
population
that
span
represents?
You
can
reconstruct
the
metrics
later
on
without
having
to
send
all
of
that
way.
You
can
sample
your
spans
in
the
process
and
then
you
can
send
them
to
a
collector
and
a
collector
could
do
your
span
to
metrics
conversion.
You
know
with
with
a
good
degree
of
accuracy
but
I,
think
I
think
the
debate
continues.
D
Moving
on
there's
APR
to
default
to
the
database
statement
being
sanitized,
so
I
think
we
talked
about
a
little
bit
last
week.
It
looks
like
there
might
be
a
second
PR
that
talks
a
little
bit
more
about
how
it
should
be
sanitized.
D
D
Another
interesting
thing
that
was
coming
up
is
that
this
user
would
like
to
know
basically
like
entry
and
exit
spans
to
a
service,
mainly
entries
fans,
to
a
service.
I,
guess
and
there's
not
really
a
great
way
to
detect
this
I
guess.
If
you're,
just
kind
of
looking
at
the
spans.
D
Span
because
it
has
you
know,
parent
ID,
but
the
other
service
entry
spans.
Look.
He
really
wanted
to
know
like
if
there
was
a
remote
span
context,
so
people
were
on
board
that
there
should
be
a
way
to
do
this
easily,
but
again,
maybe
not
necessarily
with
the
proposal
that
is
presented
here.
D
It
does
seem
like
you
should
be
able
to
follow
the
client
to
server
span
situation
and
figure
this
out,
but
I
guess
this
will
be
a
future
conversation
that
they
were
asking
him
to
show
up,
maybe
at
the
messaging
sigs
I
think
they
also
have
this
General
need
as
well.
D
So
we
are
at
time
and
we
did
start
running
out
of
time
at
the
specs
Sig
as
well.
I
think
the
last
thing
that
we
actually
got
to
was
this
issue
and
I
will
summarize
in
one
minute
I
think
this
person
was
asking
to
remove
attributes
from
an
instrument
API
but
I
think
what
he
was
actually
asking
was
how
to
forget
a
Time
series
that
is
no
longer
kind
of
active
in
a
process
when
you
have
cumulative
export
but
anyways.
D
Let
us
let
us
end
the
specsic
recap
because
we're
at
time-
and
that
was
generally
as
far
as
we
made
it
during
the
meeting,
so
there
are
some
issues
that
we'll
probably
carry
over
to
to
next
time.
Anyhow,
any
questions
comments
concerns
on
any
of
these
things.
D
Of
course,
let's
take
a
look
at
our
agenda,
see
if
we
have
any
any
issues
burning
questions.
B
The
unhappy
report
from
front
it's
HTTP,
Target
sanitization,
the
rack
instrumentation
doesn't
sanitize
as
another.
Does
the
Sumatra
one
I
don't
think
action-packed
does
use
like
the
filtered
path
from
the
active
support,
request
or
action
view
request
or
whatever
sorry
action
pack
request.
Maybe
I
don't
know.
I
can't
remember
so
many
of
these
packages
now.
B
So
that
caught
some
people
by
surprise,
I
caught
me
by
surprise
I,
guess
also
that
we
were
recording
the
query
string,
the
unfiltered
query,
string
in
the
rack,
middleware
and
I
suppose
HTTP
clients
as
well
I,
had
some
chat
with
Robert
about
this
in
the
in
the
channel,
and
they
were
you
know.
He
had
mentioned
that
there
was
some
discussion
about
this
before
and
it's
kind
of
like
well.
B
We
just
I
think
that
the
the
resolution,
if
I
recall
at
the
time
of
that
discussion,
was
action-pack's
gonna
do
what
it
does
because
rails
provides
that
feature
for
us,
but
Rec
doesn't
provide
that
feature.
So
that's
why
we're
not
sanitizing
it,
but
it
might
be
possible.
You
know
it
might
be
okay
to
like
bring
an
option
in
to
sanitize.
It.
B
And
I
realized
you
know
we're
going
to
get
back
into
that
discussion
about
whether
or
not
instrumentation
should
be
busy
doing
these
things
or
if
we
should
be
deferring
this
kind
of
stuff.
B
So
what
I
think
I'm
going
to
do
is
instead
of
kind
of
now
that
I'm
saying
this
all
out
loud,
and
we
just
saw
that
DB
statement,
sanitization
discussion,
I
think
I
might
just
do.
Take
this
conversation
to
specification
instead,.
D
So
it
sounds
like.
Ultimately,
the
issue
is
anytime.
You
have
a
URL
you
or
maybe
it's
a
URI
who
knows
where
this
Falls
the
taxonomy,
but
anything
that
has
possibly
query
params.
They
should
be
excreted
treated
consistently.
So,
whether
it's
you
know
as
an
http.url
attribute
or
an
http.target
attribute
or
or
anything
else
or.
B
Just
just
anything
that
could
be
construed
as
sensitive
right.
It's
like
we
don't
Auto
record
request,
headers,
for
example,
in
in
BRAC.
You
know
you
we
configure
it.
So
you
could
say,
like
you
know
you
you
have
to
you,
have
to
opt
into
this.
B
Because
you
know
the
risk
of
of
collecting
this
data,
but
just
like
any
attributes
in
general
again
the
database
statement
I
can't
think
of
any
other
instrumentations
right
now
that
would
off
the
top
of
my
head.
That
would
contain
sensitive
data
in
a.
D
Yeah
I
think
that
makes
sense.
I
would
have
to
consult
us
back
to
see
what
it
says
about
collecting
these
attributes
in
the
first
place.
But
if,
if
it
says
they
should
be
collect,
if
HTTP,
Target
and
acp.url
are
things
that
like
monster
should
be
detected,
they
probably
should
be
detected
by
default
without
the
query
params
and
you
could
add,
like
an
option
to
collect
everything
opt-in
to
everything
but
but
yeah
making
sure
this
spec
is
explicit
about
this
and
that
we
consistently
handle.
It
sounds
like
a
good
idea.
B
Yeah
because
resource
attributes
like
what
I
mean
by
resource
actually
I
mean
like
not
now
resource
actually
I
mean
URI.
Resources
can
also
contain
sensitive
information
in
it
or
potentially
like
personally
identifiable
information
right,
because
people
use
restful
routes,
restful
URLs,
so
it's
kind
of
like
oh
I'm,
going
to
include
a
dereference.
You
know
UPI
ID
or
something
that
could
be
construed
for
like
a
gdpr
concerns
right,
a
username
or
something
like
that.
D
However,
it
seems
like
people
always
put
sensitive
stuff
from
the
URI
and
want
want
to
at
least
be
able
to
control
this
in
the
for
for
Telemetry
back
ends,
so
I
always
I
always
have
to
give
in
and
say
all
right,
don't
do
it,
but
since
you're
doing
it,
we
will
give
you
away
too
to
ignore
this
problem.
D
Arielle,
you
will
take
this
to
the
specsig,
possibly
and
regret
this
for
life,
but
also
make
a
change
for
the
better
for
hotel.
A
I
was
when
I
last
stopped
watching
the
episodes.
A
couple
of
the
characters
were
working
on
moving
a
lot
of
rails
instrumentation
to
use
active
support.
Notifications
I
was
just
curious
if
we're.
If
that
is
any
progress,
I'm
just
curious
I,
don't
have
any
oh
actually
mailer
instrumentation,
okay,.
B
On
Friday,
Andrew
and
I
were
going
to
pair,
however,
we
were
unable
to
sync
up,
so
we
didn't
start
on
it.
Okay,.
A
B
But
I
think
like
I,
wanted
to
kind
of
like
just
revisit
like
the
revisit
it
holistically
right,
because
there's
all
these
other
instrumentations
that
we're
potentially
missing
and
I
kind
of
wanted
to
like
revisit
all
the
different
implementations
that
were
out
there
in
the
wild,
which
is
going
to
be
like
the
B-Line,
the
DD
Trace
RB,
looking
at
a
New
Relic
kind
of
coming
up
with
a
plan
kind
of
like
rethink
how
all
these
are
put
together
and
registered
to
try
to
give
us
some
sort
of
you
know
ability
to
to
leverage
it
more
and
perhaps
some
changes
to
the
notifications
instrumentation
itself,
because
right
now
you
know
like
you,
try
to
use
it
for
a
race
car.
B
But
there
needed
to
be
some
heavy
customization
that
occurred
right.
So
Chris
went
and
kind
of
redid
it
and
created
his
own
span,
subscribers
and
stuff
so
that,
because
he
needed
to
customize
the
spans
a
little
bit
more
and
so
I'm
wondering
like.
Is
there
some
changes
that
need
to
happen
in
active
support,
instrumentation
itself.
C
B
Order
to
support
this
and
what
I
was
doing
was
I
also
wanted
to
drop
that
support
for
rails
five,
since
it
was
end
of
life.
I
got
a
thumbs
up
from
people,
but
I
wanted
to
wait
until
this
was
done
so
until
the
migration
was
completed,
to
merge
it
and
what
I'm
and
I'm
thinking
that
I
want
to
try
to
avoid
running
into
problems
with
compatibility
issues.
Yeah.
A
B
B
You
oh,
like
Andrew,
also
pointed
out
that
we
missed
we
end
up
losing
some
granularity
of
things
like
he
said,
we'll
lose
the
granularity
of
like
when
a
user
calls
a
specific,
active
record,
API
method
like
create
or
destroy.
You
see
that
in
that
comment,
because
the
SQL
active
record
event
doesn't
have
that
granularity
in.
B
I
work
on
the
observability
team.
If
you
have
a
problem.
C
D
B
If
you're,
not
using
your
mobile
device
to
respond
to
issues
you're,
not
living.
A
D
B
You
all
want
to
take
a
look
at
that
030,
because
I'm
trying
to
get
I'm
trying
to
get
this
bumped
out
the
pr
that
it
came
from
yeah.
The
pr
that
it
came
from
doesn't
include
the
instrumentations
there's,
not
a
link
in
our
read
me
to
the
pr
that
included
them.
B
Instrumentation's
feet:
Bang
there
303
Bang.
B
B
Patch
version
for
all
we
only
bump
the
minor
versions
or
the
major
versions
in
all,
and
then
hopefully,
dependabot
will
take
care
of
the
rest,
makes.
D
Sense
yeah,
so
it's
AWS
one
was
not
actually
a
change,
graphql
was
ACP,
client
was
rails,
Kaka
was
not,
Sinatra
was
not.
B
A
And
was
that
just
miss
like
a
missing
dependency
or.
D
A
D
Yeah
I
will
I
will
come
back
and
check
this
out
for
updates
and
we'll
be
happy
to
yeah.
A
A
At
me,
in
slack
when
but
updated
changelog
that'll
I'll
give
it
the
old
stamp
yeah
two.
B
Stamps
promise,
if
you
do
a
refresh
hey.
A
B
Oops
sorry,
I
didn't
do
it
as
a
suggestion.
Like
a
fool,
let
me
do
a
suggestion.