►
From YouTube: 2023-02-15 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
B
Do
you
know
Riley
what
happened
and
spec
seek
yesterday?
Was
there
any
discussion
on
ECS.
C
Okay,
so
there's
no
discussion
on
the
stack
seek
there's
a
because
ecis
is
more
about
the
governance.
It's
not
about
specification,
there's
a
separate
discussion
right
after
the
specific
meeting
and
we
might
Alex
from
ECS
due
to
the
short
notice
he
he
didn't
manage
to
invite
other
folks
from
elastic.
So
we
literally
followed
up.
C
We
put
folks
in
the
same
slack
Channel
with
the
open
climate
governance
and
the
ECI
stocks,
who
only
believer
in
charge
of
the
ownership
of
UCS
that
can
make
decision
whether
they
made
a
donation
or
not,
and
the
discussion
is
continuing
and
reporting
a
deadline
saying
by
end
of
this
month.
We
need
to
make
a
call.
C
So
if
we
don't
see
any
hope
of
getting
that
alignment,
then
we
know
that
we
won't
be
able
to
work
with
UCS
to
close
that
then
we'll
just
close
the
old
Tab
and
we'll
take
our
existing
stuff
mark
them
as
table.
So
that's
the
message
we
delivered
to
ECS,
folks
and
and
I
I
feel
like
so
far.
Folks
are
willing
to
collaborate.
There
are
some
additional
thing
ecis
folks
will
have
to
dig
into
before
they
give
us.
The
final
answer.
C
The
timeline
is
managed
by
Trask
I
I.
Don't
want
to
give
you
some
random
thing,
but
but
my
understanding
is
by
end
of
this
month
we're
going
to
give
a
proposal
to
open
Telemetry
and
that
proposal
could
either
be
based
on
the
ECS
schema
or
it
could
be
based
on
the
current
schema.
C
Without
a
lot
of
breaking
changes,
we
might
still
leverage
some
ECS
game
if
that
makes
sense,
and
that
doesn't
include
a
lot
of
surprising
breaking
changes,
but
we'll
have
that
proposal
and
then,
after
that,
we'll
probably
take
a
month
to
big
head
before
we
announce
table.
But
you
should
check
the
actual
timeline
proposed
by
task.
I.
Think
that's
the
source
of
Truth.
B
No
problem
Riley
on
board
like
caught
us
up
on
the
ECS
discussion.
E
E
A
E
The
status
here
I
mean
a
separate
status
for
ECS
stability,
so
that
we
can
just
focus
on
the
non-ecs.
E
E
B
No
but
I
should
I
plan
to
create
a
PR.
Today
draft
version
and
we'll
see.
E
Cool
yeah,
no
worries,
that's
great
I.
Think
I
liked
the
direction
that
you
were
proposing
here.
E
Since
this
is
a
I
would
like
to
get
more
oh
I'm,
having
trouble
the
schema
is
about.
Failing
schema
check
is
failing.
B
B
E
Was
I
think
it
was
passing,
but
then
Carlos
mentioned
that
I
needed
to
put
correctly
mentioned
I
needed
to
put
this
into
the
resources
section.
Instead
of
the
span
section.
A
E
There's
no
other
I
checked,
there's
no
other
resource.
Oh
yeah,
okay,
I
will
but
I'll
figure
that
out
yeah
Riley.
Let's
wait
to
merge
this
one
for
yeah.
C
I
understand
so
there's
some
minor
things
not
going
to
break
a
lot
of
things.
I
think
ECS
is
doing
something
smarter.
We
should
learn
from
it.
That's
okay,
but
they
won't.
They
won't
have
to
merge
that
in
a
hurry
or
no
yeah,
so
just
wait
for
it.
Yeah
I
feel
like
we
can
wait
for
the
UCS
decision
and
if
the
decision
is
yes
most
of
the
PRS,
if
the
decision
is
no,
then
we
merge
these
minor
things
and
for
like
HTTP
URL,
then
it's
just
too
big
to
change.
E
This
one,
so
this
makes
sense
the
only
my
only
hesitation
is
I,
don't
love
the
I
feel
like
it's
like
complicated
I
feel
like
the
dot
app
part
here
is
confusing
I
understand
it's
like
Network
app,
it's
a
seven
layer.
Networking
thing
right.
B
Yes,
I
think
so
do
you
if,
if
you
think
we
should
change
it,
and
it
depends
on
this
yes
too,
but
if
let's
forget
about
this,
yes,
would
you
want
it
to
look
like
as
a
net
protocol
name
or
what
would
you
that.
E
Would
be
a
good
name
preferences,
just
net
protocol
name
and
net
protocol
version
and
unless
we
really
think
we
need
this
extra
namespace
but
I,
don't
I,
guess
I
I,
don't
know
enough
about
networking
to
make
that.
B
This
was
originally
a
an
idea
of
10
from
Dana,
Trace
and
I.
Guess.
If
we
make
a
proposal,
we
should
suggest
how
we
will
be
able
to
express
both
and
I
think
we
already
have
some
attribute
to
put
I,
don't
know
IP
there
right,
not
transport
or
something.
B
B
D
Yeah,
so
we
were
talking
there,
we
were
talking
about
HTTP
flavor,
and
this
is
kind
of
application.
Level
right,
so
it's
gonna
different
type,
so
yeah
as
as
they
put
here
like
I,
am
phdp
and
mqtt.
So
it
is
not
a
another.
It's
on
top
of
TCP
and
UDP
right.
D
E
What
would
be
a
protocol
Seward
net.transport.protocol.
D
Transport
is
still
like
a
transport
layer
right,
so
we
have
this
like
a
seven
layers
of
the.
What
is
it
the
OSI
model
transport
level
is
one
of
them
like
an
L4
I
believe
an
application.
Level
is
L7
right,
so
I
believe
that
that's
the
reference
to
it.
E
Okay,
I'll,
look
over
and
I
see
that
transport
yeah.
E
So
I
guess
didn't
know,
wouldn't
we
then
have
like
net.transport.protocol.name
and
net.transport.protocol.version.
E
Yeah
I'll
I'll
open
a
issue
to
discuss
just
what
this
means
and
how
we
see
the
like
future
things,
so
we
can
get
start
getting
feedback
from
Christian
in
particular,.
B
E
E
And
then,
let's
hop
over
to
the
general
stuff,
we've
got
the
seconds
versus
milliseconds
discussion.
Let
me
remember
what
we
discussed
in
the
spec.
E
Oh
yes,
we
discussed
the
buckets
is
a
problem.
The
current
explicit
bucket,
explicit
histogram
buckets,
are
aligned
to
milliseconds.
Currently
so
Jack
sent
a
PR
I
think
I
dropped
at
this
right.
F
E
In
the
Java
repo,
you
sent
a
draft
to
explore
a
histogram
advice,
bucket
advice,
Riley
brought
up
in
that
issue.
Whether
this
is
whether
the
histogram
is
blocking
or
not.
C
C
I
mean
there
are
three
outcomes:
number
one
we
decide
a
and
people
who
prefer
B
would
be
unhappy.
Number
two
would
decide
to
be
people
who
prefer
a
would
be
unhappy
number
three.
We
don't
make
decision
and
everyone
will
be
unhappy.
So
we
made
a
decision
and
someone
unhappy
and
totally
fine
with
it
it's
much
better
than
we
don't
make
decision.
E
So
my
question,
for
you
is
well
let's,
let's
chat
whether
so
there
this
PR
I,
my
thought
here:
I
agree
that
it's
not
technically
blocking
but
I
feel
like
it
will
be
hard
to
get
support
approvals.
I
agree
without
the
hint
API
being
merged.
First.
C
I'm
less
worried
yeah,
but
if
you
want
to
use
this
to
push
Jack
to
make
progress
on
the
hint
API
I
think
it's
fine.
Okay,.
E
E
To
specify
buckets
buckets.
C
No,
so
the
unit
is
not
something
we're
concern
here.
The
like
the
semantic
convention
for
metrics
are
basically
things
for
HTTP
duration.
It
has
milliseconds
and
then
grafana
permisses
folks.
They
prefer
second,
because
that's
what
open,
metrics
and
and
premises
have
been
yielding
so
open
time
should
make
a
shift,
but
then
there's
a
challenge,
because
in
open
Telemetry,
if
you
define
any
histogram
instrumentation,
you
will
get
the
default
buckets
and
the
default
buckets
is
like
something
500
a
thousand.
Those
are
biased
towards
milliseconds.
C
If
we
change
to
seconds
those
default
buckets
make
no
sense
yep.
The
question
is
like
hey:
if
we
switch
two
seconds,
then
the
the
default
buckets
will
will
make
no
sense,
and
this
will
be
a
breaking
change
if
we
change
the
default
bucket.
So
the
suggestion
is,
we
don't
change
the
default
buckets,
but
for
HTTP
duration,
we
use
the
hint
API
to
give
a
proper
set
of
buckets
and
JK
is
going
to
solve
that
problem.
I
believe
it's
a
very
simple
thing:
straightforward,
because
it's
very
scoped.
B
Budget
yeah
yeah
I
kind
of
like
it
because
in
in
Azure
sdks,
we
we
have
some
histograms
that
are
sometimes
better
expressed
in
milliseconds
and.
C
B
E
Breaking
changes,
I'm
kind
of
avoiding
these
to
maybe
I'll
I'll
ask
Josh.
If
we
can
maybe
focus
on
these
two
in
next
weeks,
General
oh
I
set
up
a
weekly
General,
so
actually
we're
meeting
tomorrow.
E
No
I,
don't
remember
oh
yeah
here
it
is
oh
we're
meeting
tomorrow
at
8
A.M
with
okay.
It.
B
But
that
that's
fine,
if
you're
going
to
join
I
will
go
to
messaging
and
we
don't
need
to
all
be
in
the
same
place
here.
E
Cool
yeah,
mainly
I,
want
to
use
that
to
and
I'll
let
Josh
know
today.
So
he
can.
You
know
ahead
of
time
that
I
want
to
for
tomorrow.
E
Yes
and
please
go
to
the
messaging
Liz
Miller.
The
messaging
group
needs
everybody's
thoughts
and
prayers.
E
Oh,
tomorrow's:
tomorrow's
general
semcom
meeting
conflicts
with
the
messaging
Sig
semantic
convention.
Sig
is.
B
I
think
it's
the
same
timeline
that
Pat
suggested
messaging
as
part
of
it,
but
they
don't
remember
when
exactly
they
supposed
to
happen.
It's
not
like
in
the
focus
spring
queer
doing
updating
the
conventions
for
the
past
year
and
a
half,
and
there
are
some
changes,
fundamental
changes
that
we're
discussing
there,
but
I
think
from
the
attributes
perspective.
If
ECS
won't
happen,
it
will.
We
are
done
with
attributes.
A
C
B
E
Metrics
requirement
level
requirements
level,
this
kind
of
ties
to
Riley
what
you
were
thinking
of
around
like
a
log
level,
Maybe.
C
Oh
yeah
yeah,
so
so
they're.
Let
me
explain
this.
There
there's
three
different
ways
of
thinking
about
this
problem.
One
extreme
side
is
people
coming
from
the
logging
space
they're
saying
if
I
use
logging,
API
I
can
describe
the
log
level,
whether
it's
verbors
or
not,
then
there's
a
default
expectation
like
maybe
like
information
level
will
just
get
dropped
on
the
floor.
Anything
error
and
critical
will
be
sent.
So
can
I
have
that
thing
in
the
API,
then
it's
a
matrix,
API
design
problem.
C
The
second
thing
is:
oh,
not
necessarily,
but
we
have
some
convention,
so
maybe
in
a
semantic
convention
we
know
what
are
the
things
that
are
required.
What
are
the
things
that
are
like
implemented
by
the
instrumentation
library
but
not
enabled
by
default?
The
customer
can
enable
it
and
the
the
On
The
Other
Extreme
people
are
saying:
let's
don't
do
it
so
we'll
just
make
all
the
metrics
available
and
some
of
them
are
required.
C
Some
of
them
are
optional
from
the
instrumentation
authors
perspective,
whether
they
are
like
Sunday
by
default
or
they're
turned
off
by
default.
It
sounds
in
The,
Collector
or
the
agent
should
control,
still
message
and
I.
Think
for
our
conversation,
we
should
be
very
focusing
on
like
in
the
spec,
which
one
do
we
require
and
for
HTTP
I
think
it's
very
simple.
This
is
just
HTTP
duration
period.
C
E
B
Yeah
and
I
it
made
me
or
something
so
there
is
a
bunch
of
the
documents
and
links
were
referred
to
from
HTTP
semantic
conventions,
including
this
document.
So
if
we
use
requirement
levels
essentially,
we
also
need
to
move
this
document
to
to
stable.
So
let
me
do
this
exercise.
Let
me
do
a
draft
PR
to
change
status,
to
stable
and
find
all
the
links
and
see
what
else
are
we
missing,
and
maybe
we
should
do
my
work
than
we
expected.
B
This
is,
this
actually
affects
General,
spend
conventions
right,
and
we
probably
want
to
separate
stable
part
from
unstable
Parts
just
to
understand
the
scope.
E
That's
a
really
good
idea:
yeah
I'll
put
your
name
there,
but
please
feel
free
to
ping
me
and
I'm.
Happy
to
you
know.
This
is
something
I
can
I
can
do
also.