►
From YouTube: 2023-01-31 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
C
A
C
C
A
Awesome
I'm
doing
pretty
good.
What,
where
are
you
in
Australia
I'm.
A
A
C
What's
the
progress
like
on
the
stability
of
the
semantic
convention
at
the
moment.
A
We're
there
seems
to
be
good
momentum
now
in
terms
of
people,
recognizing
that
we
need
to
get
this
done.
We
need
to
get.
C
A
Stuff
stable,
so
I
know,
we've
we've
tried
a
couple
times
in
the
past
and
I
know.
You've
been
part
of
that
effort
and
been
run
into
various
roadblocks,
and
so
Ted
had
put
together
a
proposal
on
the
process
of
kind
of
a
to
reboot
the
process
and
how
trying
to
address
the
some
of
the
block,
blockades
that
had
come
up
before
and
so
we're
trying
again.
A
Is
that
still
yeah,
so
we
did,
we
decided
to
just
go
for
it
and
put
a
date
on
it
and
try
and
see
if
we
can
see
if
that
also
helps
to
like
I,
think
part
of
getting
people
like
yourself
or
other
people
back
interested
is
was
like
they
were
kind
of
burnt
out
and
so
hoping
that
yeah.
A
Yeah,
so
we're
hoping
with
the
community
support
and
and
a
date
and
visibility,
because
this
is
this
issue
has
gotten
visibility
through
both
the
technical
committee
and
the
governance
Committee
of
semantics
convention
stability.
A
B
D
A
I
should
have
yeah,
also
just
pinged
Dennis.
A
A
What's
your
recommendation
for
like
how
to
make
progress
just
continue
pushing
in
the
the
bi-weekly
meetings.
B
Okay,
having
to
really
prepared
but
I'm.
B
Have
a
heart
attack
here
so
for
for
this
one
I
I
have
a
gut
feeling
that
okay
it'll
take
five
or
four
people
to
have
fight
here
and
given
is
already
milliseconds.
Maybe
we
we
can
like.
Maybe
we
can
put
a
stick
on
the
ground
and
saying
for
now
we'll
use
many
seconds
and
we'll
wait
for
the
overall
semantic
convention
to
evolve
and
if
it
turned
out
people
aligned
only
in
seconds
by
default,
then
we'll
just
introduce
another
HTTP
metrics
and
both
of
them
would
exist
as
the
user.
B
So
so
my
my
bottom
line
is,
we
should
agree
on.
A
semantic
convention
will
never
be
perfect
if
the
goal
is
to
have
a
perfect
semantic
convention
at
the
initial
stable
release,
I
I
would
call
it
a
pipe
dream.
B
It
can
take
forever,
like
maybe
a
year.
3000
will
have
the
perfect
semantic
convention
that
everyone
is
happy,
so
I
I'm
pretty
sure
that
the
first
version
we
released,
maybe
two
years
later,
when
we
look
back
it'll,
be
stupid
to
make
mistake,
but
it's
better
than
still
having
things
fly
in
the
air
and
not
having
anything
stable
in
two
years.
A
Yeah,
so
is
the
unit
part
of
Isis
the
unit
part
of
the
identity.
A
B
Okay,
three,
yes,
but
the
the
challenge
here
is
like
currently,
the
isdk
does
not
understand
the
unit.
It's
just
part
of
the
identity.
So
if
you
try
to
have
both
they'll
just
be
treated
as
two
different
things
and
I
I
think
that's
perfectly
fine
for
any
system
that
doesn't
recognize
units.
They
simply
like
just
take
the
name
and
try
to
do
re-aggregation
like
you
receive
things
from
multiple
places
and
they
have
the
same
name.
Then
during
a
dangerous
law
and
and
I
I
guess
that's
fine.
A
Okay,
but
that
helps
me
just
to
frame
frame
that
a
little
bit
and
we
can
yeah
and
here's.
B
Some
additional
input
like
there
are
scenario
where
people
don't
like
any
floating
number
at
all.
They
want
to
say,
like
every
operation
here
is
a
very
precise
integer
in
nanoseconds
and
I
want
this
procedure.
B
I
like
I
I,
just
don't
want
anything
to
be
float
number
and
cause
that
drift
so
in
in
that
spirit,
I,
like
or
I
can
describe
it
in
another
way
like
why
the
heck
do
we
have
so
many
different
units
for
time
like
we
have
seconds
we
have
year
month
and
micro
seconds,
that's
because
in
different,
like
that's
coming
from
physics,
we
need
all
of
them.
If
we
only
need
one,
I
was
like
2000
years
ago,
maybe,
like
some
ancient
civilization,
already
solve
that
problem,
but
the
fact
they
didn't
solve.
B
That
problem
is
telling
me
that
we're
not
smarter
than
them,
and
they
it
took
2000
years
for
not
not
having
this
problem
fixed.
It
might
take
us
another
2000
years,
not
fixing
this
problem
on
the
other
side.
I
totally
understand
the
intention
from
from
this,
and
the
open
Telemetry
project
has
a
very
high
commitment.
We
don't
want
to
give
something
that
is
against
premises
with
this
approach.
They've
literally
agreed
on
the
general
thing,
like
everything
should
have
the
seconds
by
default
thing,
or
maybe
we
just
decide.
B
B
Yeah,
so
for
for
this,
one
I
personally,
don't
have
strong
opinion
here,
but,
like
I
I
receive
a
lot
of
questions
from
people
who
use
open
television
and
and
I
like
I'm,
collecting
feedback
from
them
and
every
time
I
hurt
someone
like
they're
they're,
telling
me
I,
don't
understand
the
by
I.
B
Think
bias
is
very
common
and
I
have
to
point
them
to
some
I'm
saying
this
is
a
international
standard
and
they
told
me
like
I,
don't
even
know,
there's
an
international
standard,
and
if
you
show
this
like
International
standard
to
anyone
that
hadn't
read
this
before,
they
would
think
like
they
have
no
idea
what
that
is.
So
my
question:
it
seems
counter-intuitive
and
I
I,
don't
have
the
empathy
I
want
to
give
people
something
that
they
would
understand,
knowing
that,
maybe
it's
not
super
accurate.
B
But
my
question
is
like
like
is
that
International
standard
just
one
of
the
many
like
100
International
standards
that
nobody
knows
or
that
something
established
so
I'm
I'm,
not
expert
in
this
domain?
But
the
fact
I
keep
getting
those
feedback
makes
me
wonder.
At
least
we
should
be
very
clear.
I
want
to
avoid
a
situation
where
we
take
some
quote-unquote
International
standard
and
later
realize.
This
is
one
of
the
100
things
in
the
nebula
and
nobody
ever
know.
A
A
But
I
wanted
to
Let's
I
think
we
really
need
the
larger
group
I.
Just
kind
of
wanted
to
that.
Helped
me
to
frame
I.
Wasn't
I
hadn't
thought
through
like
what
some
of
the
potential
mitigations
were.
If
we
can't
get
agreement
here.
B
Yeah
so
there's
a
philosophy
part
like
like:
do
we
optimize
for
the
alignment
with
premises
and
open
metrics,
or
do
we
optimize
for
something
that
we
believe
right
or
do
we
optimize
based
on
the
user
feedback
or
we
optimize
based
on
some
standard?
So
my
my
gut
feeling
is
be
practical.
Probably
won't
have
more
weight
than
being
like
ideal.
A
So,
let's
go
through
the
board
real,
briefly
and
sort
of
kind
of
get
a
sense
of
like
some
of
these,
for
example,
this
one
and
this
one
where
I
think
is.
A
We
basically
agree
that
you
know
not
all
the
metrics
have
to
be
implemented.
There's
certain.
You
know
we
have
to
decide
on
which
ones
and
this
one
of
you
know,
there's
certain
attributes
that
aren't
required
that
we
need
to
just
put
in
and
I
I
can
try
to
put
in
some
PR
or
language
around
that,
although
I
guess
this
one
for
the
required
stuff.
Maybe
that
ties
to
the
yaml
pieces
has
Lynn
Miller
or
Dennis.
Have
you
all
been
following
the
yaml?
The
latest
yaml
metrics
changes.
A
D
C
Over
a
year
to
get
merged,
but
it
finally
got
merged
the
other
day.
So
yeah.
E
Cool
by
the
way,
there
is
another
change
in
the
build
codes
that
we
now
have
specific
group
type,
which
is
called
attribute
group.
So
we
can
Define
attributes
in
agnostic
way
without
referring
the
traces
or
metrics,
and
we
can
refer
to
those
attributes
from
different
specs.
C
Thank
you,
yeah
yeah,
that
was
I,
think
I
ran
into
the
same
problem
as
you
as
I
wanted
to
Define
an
attribute,
but
I
could
only
really
Define
it
attached
to
a
metric
or
a
span
so
led
me
to
create
something
similar
in
the
that
metrics
PR,
which
is
like
a
metric
group
where
you
define
attributes
and
then
the
metrics
underneath
them
will
use
them,
but
I
think
I
think
thank
you.
Those
two
things
are
kind
of
doing
similar
things,
but
I
think
they
can
work
together.
C
I'm,
not
sure
if
we
should
consolidate
on
your
approach,
I'm
just
using
only
attribute
groups
or
should
we
keep
the
metric
groups
I.
E
Think
that
they're
related
right,
so
we
can
Define
attributes
in
one
place,
but
I
think
we
still
need
metric
group
metric
definition.
It's
just.
We
can
refer
from
metric
definition
in
the
same
way
as
we
can
refer
from
spans
to
attributes
different
some
rules.
C
Yeah
yeah,
that's
both
yeah.
Absolutely
the
the
weird
thing
about
this
page,
specifically
the
HTTP
metrics
page.
That's
up
here,
is
that
it's
kind
of
hard
to
read
Because.
If
you
look
at,
for
instance,
they've
got
all
the
attributes
listed
in
one
place
and
then
they've
got
it's
rather
a
client
or
it's
a
client
and
the
server
or
it's,
but
then
there's
it's
more
complicated
than
that,
because
sometimes
it's
on
the
server
and
it's
like
depending
on
XYZ
and
then
sometimes
it's
on
the
client
conditionally
required.
F
C
Think
the
structure
that
I
came
up
with
in
the
build
tools
satisfies
it,
but
it's
get
this
as
a
result.
This
page
needs
to
be
rearranged
because
you
can't
really
generate
this
kind
of.
C
E
Should
we
actually
get
going
and
on
board
the
old
HTTP
specs
onto
the
changes
that
we
have
and
build
tools.
C
Yeah
I
think
there's
a
couple
of
things
that
need
to
be
done
before
we
can
actually
move
metrics
stemcom
in
the
spec
to
use
that
one
of
the
items
that
was
open
from
my
PR
was
there
was
a
couple
of
things
about.
C
You
know
the
table
generation
and
how
the
table
looks
like
that
kind
of
thing,
which
is
which
is
fairly
straightforward,
like
the
amble
structure
is
all
there
and
we
all
agree
on
the
yaml
structure,
but
yeah
the
the
tables
like
how
it
you
know
the
columns
and
which
tables
to
generate,
and
that
kind
of
thing
was
the
basically
the
markdown
rendering
was
the
other
thing
that
we
need
to
do
in
the
build
tools
it
there
is
a
basic
markdown
render
already
for
metrics,
but
there
needs
to
be
I
think
it
needs
to
be
a
bit.
C
There
was
a
few
improvements
that
need
to
be
called
out
for
it.
So
that's
that's
the
one
thing
I
think
needs
to
be
done
in
the
build
tools
before
we
can
move
the
metrics
some
problems
to
be
generated.
C
I
could
work
on
it.
I
don't
have
heaps
of
time
at
the
moment,
I
I,
so
I
could
get
it
to
at
some
stage.
It
wouldn't
be
my
first
thing,
but
if
no
one
gets
around
to
it,
I
can
probably
get
around
to
it
some
stage,
but
if
someone's
really
wants
to
do
it
now,
then
then
then
feel
free,
yeah
I'm
happy
to
provide
any.
A
C
A
Able
to
open
issue
or
issues
for
them.
C
C
Oh,
you
know
what
there's
already
an
issue
for
it:
nice
I'll
link,
I'll
put
it
in
the
the
doc
foreign.
D
A
A
For
some
of
these,
that
are
probably
requires
some
closer
look.
Should
we
just
list
and
make
a
point
to
review
certain
ones
before
next
time.
B
To
feel
like,
like
changing
the
assignee
is
a
good
idea.
Currently
it's
all
assigned
to
some
TC
members
who
who's
not
working
on
this
many
convention.
D
C
C
B
A
E
Yeah
I
can
add,
I
can
change
a
sickness
on
this
area,
so
I
think
it's
just
a
question
of
getting
you
trust
into
instrumentation,
DG
and
WG,
and
they
think
this
this.
We
should
do
anyway.
E
B
A
A
Yeah,
so
does
it
so
you're
if
we
assign,
should
we
just
go
eeny,
meeny
miny
mode
to
assign
different
ones?
Does
anybody
want.
B
So
I'll
be
clear:
I'll
I'll,
review
the
proquest,
but
I
probably
don't
have
benefits
to
look
at
individuals,
especially
I
open.
Some
of
the
the
issues.
A
Yeah
I
mean
I,
don't
mind,
being
the
assignee
for
all
of
them
and
or
whatever
that's
you
know,
means
as
far
as
driving
it.
E
Well,
I
mean
can't
mean
I
can't
take
a
look
at
present
them
by
next
time
and
be
able
to
explain
what
what's
needed
there.
C
A
B
One
yeah:
some
of
these
issues
are
just
like
when
I
go
through
the
the
spec
I
I
figure,
there's
a
guide
and
just
want
to
make
sure
we're
super
clear,
so
avoid
the
situation
where
we
published
this
semantic
convention
has
stable.
Then
people
have
two
different
interpretation
that
would
give
you
a
completely
different
result.
A
A
Yes
and
Riley,
we
very
much
appreciate
your
time
of
joining
these
meetings.
So,
yes,
that
is
not
a
problem
to
just
as
long
as
you
can
be
around
to
review
and
help
hit
the
merge
button
eventually.
B
And,
and
who
else
do
we
need
in
order
to
get
a
PR
merged.
A
B
A
B
Less
worried
about,
like,
like
the
community
members,
not
waiting
to
review
this
because
normally
I've
seen
like
a
lot
of
people
they're
reading
the
contribute,
I'm
more
worried
about
what,
if,
like
we
got
10
approvals
and
they're
they're,
very
good
discussion
in
the
end,
someone
has
to
make
a
call.
There's
nobody
willing
to
make
the
call
and
I'm
here,
but
I'm
I
cannot
be
the
only
one.
We
need
another
one.
So.
A
A
C
A
lot
of
those
calls
get
made
when
there's
just
a
discussion
in
the
spec
and
there's
obviously
two
ways
you
could
go.
Is
there
like
a
vote
in
the
TC
or
something
like
that.
B
Many
different
ways,
I
I,
feel
like
90
of
the
issues
are
resolved
by
social
pressure
So.
Eventually,
people
will
see
where
the
direction
is
going
and
they'll
just
pile
up
and
then
there's
a
clear
winner.
There
are
some
hard
time
where
a
particular
one
has
a
very
strong
opinion
and
they
block
the
pr
in
this
way,
normally
it'll
be
escalated
to
the
TC.
If
it's
important
and
sometimes
the
teacher
will
also
have
this
challenge
of
people
not
agree
on
the
same
thing
and
then
either
will
vote
Vote
or
it'll
just
land.
B
A
F
A
A
B
B
Like
here
we
can
see
like
this
can
be
added
as
an
optional
Dimension
that
people
can
opt
in
so
by
default
it's
not
turned
on
and-
and
you
can
have
it,
but
you
have
to
know
if
you
turn
it
on
it'll,
be
a
different
metric
stream.
No
one
can
defer
this
even
after
the
stable
release,
but
if
we
decided
okay,
we
cannot
do
that.
Then
it'll
it'll
be
a
question
for
now.
B
A
A
And
that
yeah,
that
was
my
sort
of
original
proposal
in
the
tracking
issue,
is
only
duration.
A
We'll
do
that
next
time,
HTTP
route,
let
me
know
I
saw
you
were
following
this:
oh
I,
keep
hack
was
going
to
join
I,
think
I'll
think,
but
I
had
not
followed,
although
I
kind
of
understand
the
the
issue
so.
B
And
the
the
concern
here
is
that
if
we
don't
document
that
people,
just
like,
add
their
own
dimension
on
top
of
our
existing
stuff
and
later
we
decided
to
add
the
the
same
thing,
but
we
decided
to
call
it
a
different
name
than
people
who
already
gave
some
name.
They
would
be
crazy.
It's
more
like
the
namespacing
things
like
do.
We
reserve
some
namespace
for
people
to
invent
their
own
thing
and
and
I
feel
that's
a
much
bigger
topic
than
this
group.
A
Yeah
I
mean
for
this
one
actually
did
remember
looking
but
Miller.
What
would
be
the
I
mean
the
issue
with
like
I
mean
I
could
see
it.
I
know
I
understand
your
point
of
that.
Instrumentation
generally
can't
figure
this
out
on
its
own
on
the
client
side,
but
if
somebody
was
doing
manual
instrumentation,
potentially
they
could,
if
we
just
had
it
as
an
optional
on
the
client
side,.
E
Can
you
hear
me
now
yeah,
yeah,
I'm,
sorry
something
happened
yeah,
so
it's
a
great
question
of
what
users
should
do
when
they
want
to
enrich
default
telemetry,
and
there
is
a
separate
question:
what
value
does
it
bring
the
standardize
attributes
that
users
can
only
add
manually
right
so
like
I'm,
not
against
making
it
optional,
but
I
don't
understand
how
does
it
help
it?
It's
introduces
a
new
configuration
option
for
all
the
Java
instrumentations
right,
but
it
could
hurt
more
than
help.
A
I
see,
okay
can
I
assign
this
one
to
you.
E
B
B
F
D
B
I
know
this
is
open,
Telemetry
and.
D
B
A
custom
dot
prefix
you
put
whatever
under
that.
You
don't
worry
about
people
taking
your
your
name.
E
But
we
have
attributes
that
it's
not
a
closed,
Set
HTTP
dot
right.
We
have,
for
example,
HTTP
headers
I
mean
if
we
just
say:
okay,
everything
that
starts
with
HTTP
is
reserved
it's
fine.
But
do
we
want
to
put
this
limitation
on
open
set
of
things.
E
E
A
I
think
the
headers
I
mean
it's
reasonably
closed
in
terms
of
at
least
what
they
describe,
what
you
can
put
in
them.
What
those
names
would
be.
Okay,.
A
And
I
know
we
we're
over
time
and
I,
but
thank
you
all.
If
you
have
time
we'll
just
the
last
two
and
then
what
I
think
we'll
do
for
next
time
is
we'll
go
through
these
and
if
you
can
go
through
all
of
them,
regardless
of
if
you're,
the
assignee
or
not.
If
you
have
time
that
would
be
great.
A
F
F
F
So
maybe
yeah
I
can
take
a
look
at
this.
Just
understand
what
how
we
can
proceed
with
this.
If
you
might.
A
All
right,
awesome,
yeah,
so,
let's,
let's
try
and
I,
wanted
to
try
to
keep
I
think
having
the
three
times
a
week
but
shorter
and
then
trying
to
decide
what
we're
going
to
do
between
meetings
get
some.
Some
work
done.
Async
also
will
be
ideal.
B
I
I
have
one
last
question:
do
we
know
specifically
in
the
general
semantic
convention
working
group?
What
are
the
the
issues
that
will
block
this
work
stream.
A
A
B
Yeah,
this
is
what
I
I'm
going
to
be
super
worried
about,
because
in
Matrix,
like
we
initially
started
with
some
scoping
document,
it's
an
old
half
and
we're
saying
anything
that
is
not
in
this
old
have
will
be
out
of
scope
for
the
initial
Matrix
effort
and
if
you
do
think
you
want
to
catch
the
chain,
you
should
just
go
and
block
the
old
type
after
those
have
got
merged.
The
door
is
closed.
F
Yeah
from
the
previous
specific,
from
the
various
experience,
it
was
the
case
so
EG
where
I've
been
adding
and
it
was
exactly
the
the
same
experience
but
I
think
this.
This
time
we
might
have
some
like
I
Define,
some
specific
time
time
range
when
people
can
participate
and
just
bring
all
their
ideas.
But
after
some
point
I
believe
we
just
need
to
kind
of
draw
a
line
saying
that
no
other
changes
will
be
taken.
A
A
Would
we
consider
marking
HTTP
tracing
stable
like
say
we
end
up,
you
know
we're
stuck
on
some
of
these
metrics
related
things.
Would
we
consider
releasing
HTTP
tracing
as
stable
without
metrics.
C
I
think
that's
the
idea.
I
think
it
was
in
my
mind,
that
was
the
original
idea
like
the
metric
stuff
was
not
going
to
be
marked
stable
at
the
that
was
going
to
be
later
on,
but
so
I'd
be
happy
for
that.
So
yeah
from
that
tricks
to
be
later
makes
sense.
If
there
is
extra,
if
there's
extra
challenges
for
metrics
being
stable,
then
I
think
that
makes
sense.
E
I
actually
would
rather
not
even
though
I
really
want
them
to
get
stable,
because
if
we
already
knew
what
it
means
for
a
semantic
conventions
to
be
stable,
we
could
have
more
confidence
saying:
okay,
probably
metrics
will
not
introduce
something
breaking,
but
here
well,
we
just
never
gone
through
it
and
we
can
say
it's
frozen
or
we
can
make
tracing
stable
once
metrics
are
frozen
or
I
would
leave
us
some
door
to
back
off
the
Discover,
something
unexpected
to
metric
site.
A
E
This
is
the
same
story
for
traces,
so
we
I
don't
see
an
issue
here
where
we
Define
stability.
But
even
though
we
have
some
language
in
the
specification
saying
what
semantic
convention
stability
is
it's
it's
supervac
and
like
a
traditionally
required
attribute
for
traces
is
probably
breaking
too
and
I.
I
made
an
attempt
to
Define,
what's
not
breaking
and
then
I
got
stuck
and
got
blocked
so
that
I
couldn't
make
any
progress.
E
So
if
there
is
an
issue
in
a
general
working
group,
a
general
semantic
conventions
for
a
group
I
can
post
a
link
to
the
chat,
but
unless
we
Define
what
changes
are
breaking
for
traces,
we
cannot
declare
anything
stable.
A
Okay,
yeah!
Let
me
let
me
review
this
before
next
time,
because
I
thought
that
yeah
I,
obviously
a
over
optimistic
about
what
it
means
on
the
tracing
side
to
be
stable,
was
was
more
clearly
defined.
B
I
have
one
question
regarding
the
the
reveal
participation
from
the
communities
to
folks
think
here
like
if
we
give
a
recommendation,
maybe
in
the
spec
report
or
the
maintenance
report
by
saying
for
the
next
few
months.
If
someone
send
a
pull
request
in
the
spec
repo
about
some
others,
many
convention,
we
we
should
defer
that
after
we
ship
the
HTTP
just
because
we
want
hyper
Focus,
because
I've
seen
a
lot
of
things,
they're
they're
dragging
attention
or
or
do
you
think,
it's
okay
for
them
to
compete.
The
resource.
E
There
will
be
mostly
extensions
and
then
well
if
we
had
some
other
place
to
send
people
to
that,
doesn't
get
scrutinized
right.
People
could
party
on
forever.
Thank
you
defining
other
names.
A
One
thing
that
might
work
we
could
ask
to
have
you
know
a
certain
amount
of
time
every
week
in
the
spec
meeting
to
say,
Hey,
you
know
to
prioritize
to
work
through
issues
and
to
sort
of
demand
attention.
B
A
Yeah
yeah:
well,
let's
I'll,
add
it
to.
E
Yeah
and
I
guess
there
is
it's
an
interesting
question.
So
there
is
a
spec
like,
let's
see
messaging
credit
base,
and
there
are
some
common
databases
like,
for
example,
open
source
things
so
like
having
rabbit
and
Kafka
makes
sense,
but
having
service
bus
or
event.
How
I
I
don't
understand
why
right
I
have
I
can
host
it.
B
E
And
then
like
it
would
be
nice
if
there
was
a
place
for
it.
There
is
a
tooling
problems
where,
if
I
hosted
somewhere
else,
I
cannot
benefit
from
schema,
Transformations
or
any
tooling,
from
open,
Telemetry
right,
but
it's
separate
it
doesn't
matter,
but
basically,
if
we
can
put
a
line
somewhere,
if
you
have
a
framework
and
say
okay,
the
technology,
the
general
semantic
convention
plus
plus
popular
open
source,
Stacks
or
popular
Stacks,
leave
here
and
everything
else
goes
to
a
specific
place.
E
B
E
Yes,
that's
it's.
E
By
the
way,
there
is
a
odd
up,
I'm,
not
sure
if
you
saw
I'm
posting
link
here,
Daniel
dealer
is
defining
the
procedure
for
experimental
semantic
conventions,
I'm,
not
sure.
If,
if
it's
baked
enough,
it's
only
about
attributes,
but
it's
somehow
related
and
sounds
like
people
are
working
on
semantics
convention,
stability
from
different
places.
D
E
Well,
it
solves
a
difference,
the
migration
story
for
attributes,
but
it
sounds
like
people
think
about
semantic
conventions
as
a
set
of
attributes
and
I.
Don't
agree
with
this
point:
I,
don't
think
we
can
Define
stability
of
attribute.
A
Yeah
and
I'm
kind
of
I'm
not
unsure
why
the
approach
that
Java
instrumentation
has
been
taking,
isn't
good
enough.
We've
just
tied
everything:
that's
not
in
the
semantic
conventions
behind
an
experimental
flag.
E
B
E
It's
it's
good
for
attributes,
but
it's
it's
not
the
way.
It's
how
you
introduce
new
ones,
but
if
you
need
the
whole
Behavior
right.
So
if
this
attribute
is
set,
it
means
something
and
you
should
expect
I,
don't
know,
links
or
events
of
certain
kinds
be
available
and
you
can
Define
how
the
attributes
are
edit,
but
like
they're,
the
behavior
I.
B
Understand
yeah
I
feel
semantic
convention
is
very
ambitious.
If
you
look
at
the
elastic
common
schema,
it's
just
talking
about
the
the
attribute
name
and
the
value
and
the
the
meaning.
It
doesn't
talk
about
how
different
Telemetry
can
be
connected,
like
with
a
trace
like
how
parent
child
should
be
modeled,
and
how
do
you
handle
retry
it
has.
The
common
schema
has
nothing
about
that.
A
Retries
that
I
I
know
this
was
on
oh
yeah,
but
Miller
I
will
I
send
this
one
to
myself,
but
I'm
definitely
going
to
want
your
thoughts
on
this.
We
chatted
about
this
in
the
Java
meeting
last.
D
A
Yeah
matesh
has
started
to
try
to
implement
the
retry
spec.
So
that's
where
it
came
out
of.