►
From YouTube: 2023-02-10 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
D
E
B
Okay,
so
Josh
do
you
want
to
so
we
can
talk
project
board
stuff,
but
we
have
sort
of
the
overhanging
big
item
of
ECS.
B
So,
okay,
so
I
broke
out
the
final
two
ECS
alignment,
mappings
renames,
that
we
would
need
for
http
the
net
attributes
which,
from
Ludmila
and
I,
met
with
Alex
to
work,
those
out
and
then
the
exception
ones,
because
I
really
don't
think
we
can
I.
Think
it's
I,
don't
think
it's
meaningful
to
make
HTTP
semantic
conventions
stable.
Well,.
C
So
that
one
I
want
to
call
out,
because
I
I
I
have
a
pull
request,
that
I
was
actually
filling
out.
The
description
for
to
Mark
service
dot,
star
that's
stable
for
resource.
So
at
least
you
have
something
to
attach
to
right.
C
B
C
Yeah
we
have
to
be
a
little
more
careful
with
that
related.
We
did
this
thing
where
we
said
we're
not
specifying
what
an
exception
looks
like
and
we're
requiring
all
the
language
sigs
to
specify
what
an
exception
looks
like
so
I'm
actually
planning
to
come
back
to
language.
Six
and
say:
hey
did
we
do
this
because
it's
not
listed,
we,
we
never
wrote
it
down.
C
So
I!
Guess,
let's
look
at
that.
First,
because
that's
what
I'm
most
nervous
about
so
we
can
spend
the
most
time
on
it.
The
rest
of
them
don't
seem
terrible.
Well,
I
should
say
they
don't
seem
quite
as
scary.
B
So
yeah,
so
all
of
these
things
make
since
to
me
I
think
you
know
going
to
the
languages
is
reasonable
documenting.
It
is
reasonable,
I'm,
not
sure
what
to
do
about
that.
B
But
what
you
were
saying
as
far
as
renaming
like
aligning
do
you
think
aligning.
This
is
a
no-go
because
of
because
the
tracing
SDK
is
already
stable.
C
That's
kind
of
my
fear
a
little
bit
because
we'd
be
changing
how
the
SDK
operate
so
there's
the
reason
I'm
worried
is
there's
a
method
you
can
call
on
the
SDK
and
when
you
call
that
method
it
does
XYZ
right,
and
so
it's
a
it's
a
helper
method,
but
it's
documented
to
basically
say
if
you
call
record
exception,
we
will
automatically
fill
out
these
fields
right
and
that's
in
our
stable
version
of
our
tracing
SDK.
So
that's
why?
C
Yes,
the
stable
version
of
the
SDK
points
to
an
unstable
part
of
the
SDK
or
an
unstable
semantic
conventions,
and
that's
a
little
freaking
weird,
but
that
that's
my
concern
here
is
like
I
I
I'm
wondering
if
that
particular
like
how
much
push
we
can
have,
on
the
exception
part
from
their
end,
because
what
we
could
do
is
we
could
keep
tracing
with
its
current
semantic
conventions
and
we
could
allow
error
kind
of
a
thing.
You
know
it
just.
It
feels
weird
to
me,
though,.
C
The
other
thing
I
wanted
to
to
look
at.
If
you
look
at
their
error,
do
they
have
a
notion
of
whether
or
not
the
error
is
escaping
the
current
scope.
D
C
Right
because
it's
log
based
right,
so
that's
the
the
other
thing
yeah,
no
corresponding
the
other.
The
other
thing
we
could
think
about
here
is
effectively.
Ecs
only
has
errors
that
escape
the
scope
and
we
could
Define
a
conversion
where
you
know.
If
you
have
an
exception
that
escapes
the
scope,
you
can
turn
it
into
an
ECS
error.
C
I,
don't
think
yeah
anyway.
That's
another
thing:
I
was
thinking
about
if
I
think
about
like
an
error
versus
an
exception,
but.
B
I'm
trying
to
think
if
I
feel
that
strongly
about
that
the
a
rename
would
break
SDK
stability.
B
C
I
mean
that's
that's
fair,
so
you're
saying
like
in
for
a
penny
in
for
a
pound,
we
would
need
to
at
a
minimum,
get
that
scope
I.
Don't
know
why
the
scope
bit
lives
in
semantic
conventions
by
the
way,
I'm,
not
sure
who
propose
that
it's
kind
of
an
oddity
in
the
more
I
looked
into
it,
to
try
to
figure
out
the
justification
of
it.
But
we
would
need
to
get
that
kind
of
added
right.
Sorry,
which
bit
exception
and
error.
B
C
Can
you
can
you
go
back
to
the?
Do?
You
have
the
ECS,
what
does
their
stack
Trace?
How
is
that
specified
yeah?
That's
my
other
concern.
C
F
C
D
F
F
C
Because
it's
a
wild
card
and
it's
extended,
which
means
I,
think
you
can
actually
put
like
an
array
of
the
stack
trace
and
stuff.
That's
that's
my
fear,
like
there's,
not
enough
information
in
here
in
that
spec,
for
unless
we
have
somebody
who
actually
knows
ECS
well,.
C
Yeah
I
guess
the
two
things
that
I
worry
about
right
is:
if
we
churn
here
it's
it's,
the
name
is
not.
C
The
name
might
not
be
as
bad
as
the
format
right,
because,
right
now
we
have
we
made
a
decision
where
all
otel
stack
traces
are
just
strings
and
then
languages
are
supposed
to
document
what
those
strings
mean
and
what
they
use.
So
a
tool
provider
would
have
to
look
up
how
to
parse
every
stack
Trace
by
language.
We
actually
probably
should
well
I.
Think
we
do
require
that
you
have
a
language
version
in
your
resource
attributes.
C
So
between
the
resource
attribute
and
the
stack
Trace,
you
should
be
able
to
figure
out
how
to
parse.
You
know
every
single
exception
as
a
tooling
provider.
However,
we
don't
know
what
ECS
did
there.
Did
they
take
the
same
approach
or
did
they
do
something
more
clever
if
they
did
something
more
clever?
That
does
seem
kind
of
like
a
breaking
change
or
a
big
churn
for
the
SDK
to
go
like.
Oh
now,
you
have
to
parse
exceptions
right.
That's
that's!.
C
B
I
think
there's
two
good
reasons
to
do
it.
One
is
the
that
we
already
have
two
examples
in
our
repo
of
wanting
to
share
that,
and
the
other
is
Nev's
comment
here
that
that
the
user
agent
is
getting
Frozen
anyway
and
there's
going
to
be
kind
of
subfields
under
there
in
the
future
anyway.
So
it's
nice
to
have
that
grouping.
B
This
one
right
that
that's
the
one
misalignment
is
splitting
it
into
path
and
query
string.
These
are
all
straight
renames.
B
These
are
I,
think
straight
renames
stack,
Trace
dot.
This
should
be
stack,
tricks.txt.
B
This
one
there's
lots
of
weird
lots
of
misalignment
where
the
main
beam
the
net
peer.
We
use
net
Peer
Net
peer
to
mean
the
remote
one,
whereas
they
just
always
label
that
server
and
client
so
on,
as
opposed
to
myself
and
my
peer,
so
it
kind
of
depends
will
depend
on
spam
kind,
I
guess
if
we
could
have
a
schema
transformation
that
renames
for
a
specifics.
Mankind.
D
B
A
E
Yeah
so
on
HTTP
clients,
you
would
usually
for
specific,
HTTP
client.
You
have
a
different
apis,
saying:
okay,
this
is
my
proxy
and
this
is
my
real
destination.
E
A
Okay,
I
mean
there's
still
a
lot
of
edge
cases
like
I,
I'm
I,
think
we
I
mean
in
Java
instrumentation.
We
have
HTTP
clients
that
do
not
get
the
logical
address
easily,
at
least
so.
It
still
would
be
kind
of
a
question
what
you
should
actually
should
you
use
like
a
server,
IP
or
server.ip,
if
you're
not
sure,
if
it's
a
proxy
or
if
it's
the
actual
server
that
you're
sending
a
requested.
A
But
like
I'm
asking
this
question
because,
like
this
seems
kind
of
complex
to
me,
I
I
liked
the
previous
I
mean
the
current
somatic
conversion,
because
it's
it's
conceptually
simple
I
mean
there's
like
a
logical
address
and
that
thing
that
you
see
in
the
URL
that
you're
calling
and
the
actual
socket
address,
and
it's
simpler
to
understand.
At
least
from
the
instrumentation
older
point
of
view.
E
E
Yeah
because
it's
like
dawn
like
it's,
already
been
like
that
for
adult
semantic
conventions,.
E
E
It
sounds
like
not
in
general,
is,
is
very
specific
and
not
probably
straightforward
for
users
to
understand
I,
I
I'm,
not
sure
myself,
what
exactly
it
means
there
in
ECS
and
it
doesn't
like
if
it's
this
significant
and
maybe
it
is
I
wonder,
do
we
see
a
future
where
we
introduce
a
clear
one
in
two
open,
Telemetry
and
ECS
like
if
they
converge?
Maybe
it's
something
that
we
will
add
there.
B
Yeah
we
could
propose
server
sock
address,
which
I
think
you
had
proposed
at
one
point
already,
with
Noah
service
socket
address.
B
Okay,
let
I'll
update
the
issue
with
these
comments.
B
We
don't
have
a
lot
of
time
left
Josh.
What
do
you
think
and
Riley?
What
do
you
think
is
the
most
important
thing
to
discuss.
B
B
Do
you
think
are
Next
Step
from
making
this
this
decision.
G
Yeah,
so
so
there
are,
there
are
two
things:
I
I
feel
we
should
do.
Number
one
is
I.
I
should
follow
up
with
highlights
I
like
I
I
told
him
about
it.
He's
talking
to
the
elastic
folks
and
and
I
I
feel
like.
We
should
give
him
a
few
days
and
and
I'll
check
with
him
before
next
Monday
and
see
whether
he
will
hear
back
and
he's
not
hearing
back
at
least
I
won't
know.
G
We
should
merge
the
old
type
as
well.
If
that
couldn't
work,
then
we
just
close
the
old
tab.
We're
saying
like
easy
as
an
open
packages.
Unlikely
to
merge
we're
still
happy
to
collaborate,
doing
something,
translation
or
something
and
the
HTT
semantic
convention.
Given
if
we're
going
to
make
the
change,
it's
not
going
to
align
well
with
ECS
anyways,
let's
just
stick
with
what
we
have
to
minimize
the
breaking
chain
for
existing
customers.
G
In
that
way,
we
still
achieve
the
stability.
It
has
been
running
for
a
long
time.
I
haven't
seen
like
fundamental
Gap
and
and
meanwhile
I
I
think
trust
trust.
We
we
should
reach
more
voice.
I
know
you're
already
doing
that
in
the
spec
room
in
the
maintenance
report.
Just
to
tell
people
there's
a
certain
timeline.
If
they
have
any
issue
that
might
block
the
HTTP
semantic
convention
stabilization,
they
should
ask
clear
that
before
that
time,
otherwise,
for
example,
we're
saying
like
starting
from
March,
we
might
have
some
Polish
work
to
do.
G
G
B
So
the
the
current
plan
with
the
following
the
new
Ted's
process,
we
have
four
more
weeks
for
this
working
group
to
present
sort
of
final
recommendations
to
the
community
about
in
four
weeks,
for
the
community
to
review
and
then
two
weeks
to
make
stable
yeah
about
timeline.
Yeah
I
will
I
will
do
this,
I
think
that's
really
good
and
then
I'll
continue
using
the
spec
yeah.
B
That's
probably
a
good
idea:
I
can
use
some
maintainer
because
there's
that's
a
different
set
of
folks
that
show
up
to
that
meeting
to
raise
awareness,
but
then
in
the
spec
meeting
I'll
continue
and
we
can
talk
on
Monday,
but
since
we
won't
have
Josh
on
Monday
Josh
thoughts
on
what
I
should
use
next
Tuesday
spec
meeting.
C
Or
are
we
ready
to
have
the
exception
discussion.
C
We
need
to
I
know
we
need
to
I
okay,
so
just
for
reference
I'm
planning
to
propose
stability
for
service,
dot,
star
and
telemetry.sdk.star
resource
attributes.
C
What
I
don't
want
to
do
is
cause
that
to
not
have
any
time
to
talk
about
HTTP,
cements
convention,
things
that
we
need,
but
I
also
think
that
that's
a
blocker
for
you
is
having
any
kind
of
stable
resource
right.
I.
Think,
in
my
mind,
exception
is
the
thing
my
next
top
priority
of
unblocking
you
but
I
feel
like
this
Net
stock,
pure
name
thing
is
probably
a
more
useful
discussion
for
us
to
have,
because
that's
really
the
heart
and
soul
of
when
we
think
about
ECS
compatibility.
C
You
have
enough
details
in
there
and
enough
kind
of
what
do
I
want
to
call
it
like,
like
impedance
mismatch
for
us
to
walk
into
issues
like
I
I.
Have
that
as
well
with
the
other
first,
two.
So
for
the
exception
discussion
we
have
some
impedance
in
this
match.
I
think
between
that
what
that
stack,
Trace
field
really
is
and
what
we
did,
but
I
don't
know,
because
we
have
to
find
out
for
sure
with
service.star.
C
We
have
one
conflict.
Everything
else
is
exactly
the
same
that
conflict
I'm
planning
not
to
resolve
for
reasons
that
I'm
specifying
in
the
pr,
because
I
think
we
can
actually
just
adopt
their
other
name
as
well,
but
anyway,
that's
that's
a
detail.
What
I
think
is
more
interesting
is
when
you
look
when
we
walk
through
that
Net
versus
server
and
client
right,
there's
a
lot
of
there's
a
lot
of.
Basically.
C
C
So
I
feel
like
maybe
that
would
be
the
best
topic
of
discussion,
because
we
might
as
well
get
that
discussion
out
of
the
way
so
I
would
plan
for
that
discussion.
I'd
prep
for
it.
However,
you
want
I'd
make
sure
that
the
way
you
present
it
is
the
way
you
want
people
to
be
thinking
about
it
right,
but
that's
in
terms
of
the
most
difficult
conversations
to
have.
That's
probably
your
number
one.
B
Back
to
this,
so
I,
don't
like
I,
understand
your
point
that
that
it
makes
the
HTTP
stuff
way
more
valuable
but
like
from
a
Java
instrumentation
perspective,
this
wouldn't
block
us
from
releasing
stable,
HTTP,
instrumentation
I,
don't
think,
although
I
know
you
have
thoughts
about
around
the
permit
because
of
the
Prometheus
mapping.
That's
what
your
concern
is.
C
Yeah
so
I
guess
the
the
problem
I
have.
Is
you
couldn't
Mark
the
whole
thing
as
stable
unless
all
the
stuff
that
you
produce
is
stable?
If
you
think
about
it
right
so,
if
I
produce
a
span
and
those
resource
attributes
aren't
stable
but
I
call
them
stable.
Now,
I
have
an
issue
where
I
can't
change
them
right.
G
Wait
for
the
focus
on
getting
the
accusing
stable
forgot
about
anything
else
for
now
and-
and
we
should
clearly
understand
like
even
if
we
can
mark
the
HTTP
cemented
Convention
as
stable
in
practice,
a
lot
of
things
are
still
stuck.
So
it's
like
P0
in
P0.
We
have
two
items,
but
we
shouldn't
bundle
them
together.
Otherwise
it
it
was
low
as
well.
So
we
we
just
got
something
stable,
knowing
that
only
having
stable
might
not
solve
any
problem,
and
that's
okay,
because
at
least
we
got
one
thing
stable.
C
C
F
C
G
C
I,
don't
think
the
HTTP
semantic
convention
group
needs
to
take
care
of
it.
I
guess
is
the
way
I'd
phrase
it.
C
Yeah,
actually
I
was
just
going
through.
If
you
look
at
how
we
Define
what
Telemetry
is
allowed
and
attributes
and
things,
it
is
totally
fine
to
do
exceptions
later.
C
B
Okay,
so
awesome
all
right.
This
gives
us
a
lot
to
chew
on
do
we
have
I
think
we're
I
forget
if
we're
meeting
next
Monday,
our
general
is
this:
no
Monday
is
not
our
general
working
group,
so
I
will
post
in
the
chat
and
if
Josh,
if
you
have
a
chance
to
look
at
I,
don't
know
if
you
looked
at
look
Noah's.