►
From YouTube: 2023-04-05 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
B
B
A
A
We
get
a
lot
of
questions
about
it
about
it.
In
the
a
lot
of
the
Java
Hotel
Java
instrumentation
users
are
using
Prometheus.
A
C
A
A
B
And-
and
we
had
a
TC
meeting
this
morning,
so
the
idea
is
ideally,
we
would
want
to
be
data
driven
by
understanding
how
many
folks
are
relying
on
it
and
how
much
pain
they
would
get
but
I.
Unfortunately,
as
the
people
working
on
observability
and
normally
we
don't
have
any
solution
usage,
so
maybe
open
source.
We
just
shouldn't
monitor
things.
So
so
on
The
Collector
side
we
don't
have
any
usage
Telemetry,
but
on
the
individual
instrumentation
site,
we
have
some
information.
B
For
example,
I
can
tell
that
and
we
have
17
million
downloads
of
the
esp.net
core
instrumentation,
which
is
the
HTTP
storage
mutation,
and
we
also
have
something
about
the
Legacy
SMS
and
the
client
second
guys
a
rough
number
of
hey.
What
are
the
total
download
numbers
of
this
package
and
also,
what's
the
what's?
The
percent
of
the
US,
for
example,
I
I
briefly,
look
at
the
donut
space.
I
can
tell
one
third
of
the
customers.
Even
open.
Telemetry
Donuts
are
using
some
sort
of
HTTP
client
and
server
side.
Instrumentation.
No,
no,
it's
critical!
B
Then
it's
very
easy
to
convince
the
donut
maintenance,
not
because
they
there
is
many
of
them
are
in.
My
DM
is
because
we
have
one
third
of
the
users
relying
on
these
critical
packages.
It
wouldn't
make
sense
just
to
trim
this
thing
over
sudden
and
break
them.
So
in
this
way
the
maintenance
would
be
reasonable
and
be
willing
to
support,
maybe
for
a
year
right,
but
if
it
comes
to
like
maybe
python
or
something
they
don't
have
a
loud
house,
HTTP
users,
then
fine
with
me
like.
B
Why
do
we
want
to
order
policy
and
require
everything
like
maybe
like
the
rust
Community
is
just
saying.
We
just
got
this
HTTP
instrumentation
yesterday
and
you
want
me
to
either
slide.
Let's
grow
up
right,
so
so
my
position
is
we
we
try
to
be
reasonable
and
and
I
think
charge.
Storage
is
on
the
same
on
the
same
direction.
It's
basically
saying.
Ideally,
we
want
to
run
the
data,
but
unfortunately
we
don't
have
all
the
data
so
we'll
try
to
catch.
Some
numbers
try
to
be
reasonable
But.
C
A
So
my
reading
of
the
tigran's
concerns
the
especially
around
the
build
tools:
code
generation.
These.
B
C
A
As
a
vendor,
it
doesn't
want
people
emitting
these
new
semantic
conventions
for
the
next
six
months.
He
wants
basically
a
six-month
Gap
where.
B
In
that
way,
I
wouldn't
say
we
introduced
all
these
versions
or
something
we
should
just
either
big
disclaimer
in
the
HTTP
cement
convention.
By
saying
anyone
who's
trying
to
implement
a
new
thing,
they
should
follow
the
thing
like
like
url.4
or
something,
and
you
already
have
something
you
want
to
upgrade-
don't
just
make
the
like
the
blind
update
for
now
make
sure
you
keep
supporting
this
thing
until,
like
this
post
update,
number
done
right.
A
I
see
so
what
you're
saying
is
put
a
date
say
October.
You
know
first
yeah.
A
And
for
say,
Java,
we
would
not
add
we
would.
By
default,
we
would.
Our
default
state
would
need
to
continue
emitting
the
today's
yeah
spec.
B
Version
we're
saying:
hey,
you
shouldn't
just
go
and
switch
the
new
version
you
have
to
support
both
and
by
default,
usually
use
the
previous
version
and
put
the
exact
version,
and
then,
after
this
certain
date,
you
have
the
freedom
to
remove
the
old
support,
and
let's
just
focus
on
that
very
problem.
Instead
of
trying
to
invent
this
mechanism
because
that
magnetism,
like
any
mechanism
I,
would
say:
if
it's
not
solving
three
problems,
then
we
shouldn't
introduce
that
like
I
hope
the
issue,
one
is
the
only
one
that
we
have
to
solve.
A
So
to
the
tour
generation
point
the
like
a
lot
of
the
instrument,
so
we
would
release
So
today
we're
at
spec
version
what
one
dot
19
1.20,
something
like
that,
whatever
we're
at,
we
would
release
say
1.21
with
the
new
conventions,
but
we
would
just
have
a
disclaimer
saying
that.
Okay,
so
we
released
the
new.
So
then
the
build
tools,
project
Auto,
generates
all
these
constants
for
different
languages
from
the
Hat.
A
A
But
then
they're
not
that's,
they're
not
going
to
be
able
to
get
other
changes
like
that
were
made
to
database
or
elasticsearch
conventions
or
any
of
those
other
changes.
If
they
can't
update
the
constant
that
the
build
tools
I
mean
they
can't
use
the
constants
for
those
they
would
have
to
manually.
B
A
A
I
do
understand
the
I
think
the
the
concern
is
around
these
the
constants
and
and
there's
that
yeah.
B
So
so
I
I
can
give
the
offer,
for
example,
we're
going
to
say
the
work,
we're
saying
python.
We
need
this,
and
python
doesn't
have
people
working
on
the
thing.
Riley
will
just
go
and
copy
the
constant
and
try
to
merge
them
into
a
single
file.
I
mean
this
is
like
one
hour
actual
shitty
work
for
me
I'm
having
to
do
it
it's
much
better
than
us
information
effective
until
without
even
knowing
whether
that
maximum
is
generic
or
not.
A
B
Yeah
and
and
I
and
ultimately
I
don't
even
see
a
reason
why
we
need
to
run
this
create
a
hybrid
mode.
We're
saying
this
is
a
stable
version,
but
not
the
Implement
that
right
now
wait
for
Signal
like
on
October
or
like
whatever
like
time
you
feel
comfortable.
This
is
the
time
you
should
start
to
implement
that
and
before
that
we
give
six
months
notice,
all
the
vendors
they
can
try
to
change
their
stuff
just
to
be
prepared
for
it,
and
this
will
be
a
one-time
change
and
after
that,
we're
done.
B
And
another
thing,
I'm
thinking
I
know,
there's
a
third
party
Tool,
both
at
some
point
they're
trying
to
contribute
to
open
times.
So
it
is
a
command
line
tool
which
will
generate
open
climate
payload,
and
maybe
we
can.
B
You
have
six
months
of
time
to
make
all
the
changes.
It
seems
pretty
generous
and
also
it
gives
us
a
way
and
not
just
HTTP.
If
anything.
Previously
we
struggle
with
like
we
worry
about
making
changes.
Now
we
have
this
window.
We
can
make
a
lot
of
changes
just
to
align
the
long-term
thing
we
should
embrace
it.
B
This
yeah
and
the
one
one
thing
I
guess
a
lot.
How
vendors
might
be
concerned
is
here
we
may
say:
instrumentation
libraries
are
talking
about
existing
one
or
the
new
ones,
I
I
guess
they
would
want
all
of
them
because
they
don't
want
all
of
a
sudden,
a
like
rust
to
start
Implement
a
new
one,
that's
in
the
new
data
and
they
haven't
prepared
for
it.
But
maybe
we
can
clarify
I.
B
B
Are
we
okay
with
them
just
implementing
the
new
one
or
they
have
to
do
the
old
one
or
side
by
side,
because
this
seems
like
we're
asking
them
to
do
a
throwaway
work,
which
is
very
clear
so
that
that
part?
Maybe
we
can,
we
can
just
get
more
feedback.
The
worst
case
is
they'll.
Just
like
do
some
shitty
work
right.
B
Yeah
what
they
do
exactly,
for
example,
if
Reserve
like
open,
Telemetry
rust,
they
try
to
implement
things
that
are
the
same
before
October.
Now
we
have
to
do
both
I.
Don't
want
to
do
that.
Then
the
suggestion
might
be
okay,
you're
lucky
you
haven't
started
this
thing,
so
you
can't
just
wait
for
the
thing
to
be
stable
as
after
October,
then
you
have
a
clean
start
right.
D
I
guess
one
potential
problem
with
that
would
be
if
we
make
that
assertion
and
we
have
the
new
conventions
living
in
the
build
tools,
but
there's
also
other
parallel
work
going
on.
How
do
they
go
about
consuming
that
package,
like
in
the
case
of
node?
It
generates
this
npm
package,
which
is
generated
at
build
time
to
populate
it
all,
and
while
it's
technically
possible
to
have
different
instrumentations
use
different
versions,
it
gets
messy,
increases
the
size
of
the
code,
but
it
really
is
I
I
think
the.
D
If,
if
we
add
it
to
1.21,
then
some
other
ones
come
into
1.22.
D
A
B
D
A
Yeah,
so
the
proposal
here
is
that
node
and
other
languages,
Java,
is
in
the
exact
same
boat.
Would.
D
A
Then
you
make
a
internal
class,
at
least
in
the
Java
world.
We
would
just
have
an
internal
class
that
has
con
the
we
would
manually
create
those
constants
for
the
new
things.
Yeah.
D
B
B
D
D
A
I
can
confirm
on
the
Java
side
that
you
know
we
would
just
we've
done
this
in
the
past
already,
where
we've
gotten
ahead
of
the
convention,
so
we
couldn't
copy.
We
couldn't
use
the
officially
generated
code,
so
we
just
had
our
own
temporary
constants
and
still
until
the
we
were
able
to
update
so
yeah
I
can
confirm
on
the
Java
side.
It's
you
know
an
hour
of
work.
A
Yeah
I
think
probably
I
mean
yeah
I'll
I'll
put
that
out
there
I
do
expect
some
pushback,
because
people
some
people
really
like
these
auto-generated
constants,
even
though
all
they
are
is
literally
just
constants
from
the
spec.
A
I
had
a
couple
other
things:
oh
yeah
little
nevinite
chatted
about
this
one,
and
so
the
one
kind
of
open
question,
I
think
is:
should
we
combine
since
this
is
breaking
change
on
HTTP
client
IP
can
which
is
reasonably
common?
Should
we
or
can
we
bundle
that,
in
with
the
ECS
transition
period,.
E
A
D
Yeah,
that's
what
I'm
saying
today
that
is
100
generated,
so
there
is
no
manual
aspect
of
that,
so
I'm
just
calling
that
out
until
yeah.
A
E
A
So
the
concern
about
those
four:
let
me
pull
up
the
right.
A
Items
here
the
concern
with
these
four
is
that
there
there's
no
we're
not
saying
that
one
is
there's
any
better
than
the
other
and
is
just
purely
subjective,
and
it
sounded
like
at
least
a
couple
folks
thought
in
the
Otep
that
if
there
was
in
that
case
of
like
a
tie,
that
we
would
pick
the
hotel
ones.
E
A
And
I'm
not
sure
that's
the
case
or
not,
but
I
think
we
will.
It
will
help
us
if
we
can
just
take
these
four
out
we're
already
Gonna
Change,
most
of
the
net
ones
to
client
or
server,
and
we
do
think
there's
good
reason.
Technical
reason
for
doing
that.
E
A
So
I'll
create
a
new
issue
for
that
and
then
I
was
going
to
go
through
this
I'll
go
through
this
issue.
This
was
sort
of
the
initial
brainstorming
issue
and
I'm
gonna
try
to
close
this
out.
I
think
most
of
we
have
sub
topics
for
most
of
sub
issues.
For
most
of
these
already.
A
E
Do
you
think
I
should
explain
this
pull
request
into
individual
things
like
URL
or
HTTP,
or
server
or
client
all
right,
since
this
PR
already
got
some
attention?
Maybe
you
should
keep
going
here?
How
do
you
think
you
should
do
it.
E
A
B
So
the
trust
I
I
have
to
like
both
our
step
impact
questions.
Number
one
is
probably
answer
that
before,
but
I'm
skypeful.
So
how
important
do
you
think
this
night
related
things
are
tied
to
the
HD
key
stabilization?
If
it's
not,
then
my
suggestion
is
maybe
we
can
just
separate
the
thing
and
focus
on
HTTP
one
and
if
it's
not
like
you
think
we
must
solve
this
as
a
single
transaction.
B
My
gut
feeling
is
I
haven't
seen
a
lot
of
feedback
from
elastic
folks
and
another
has
been
a
lot
of
feedback
for
HTTP,
but
for
next
part
I'm
a
little
bit
funny.
So
do
we
know
what's
their
position,
maybe
maybe
they
have
strong
reason
why
elastic
common
schema
is
better
or
maybe
they're
even
unhappy
with
knights.
They
want
to
change
to
something
else.
So
I
haven't
seen
a
clear
position
from
them.
A
So
on
these
ones,
ladmill
and
I
talked
through
this
on
Monday
and
there's
some
nice
advantages
to
splitting
these
into
client
and
server
sites.
A
E
C
A
Yes
with
it
wasn't
we,
we
took
some
of
the
good
things
from
open,
Telemetry
and
applied
that
to
ECS
and
made
this
serversocket
dot,
client
and
client
socket,
instead
of
they
have
some
funky
Nat,
client.net
and
server.net.
That
doesn't
really
feel
good.
A
And
we
did
talk
extensively
with
Alex
about
the
NAT
stuff
and
yeah
he
was.
He
seemed
to
be
a
lot
like
accepting
of
us
that
he
didn't
seem
to
love
what
they
had.
B
The
one
yeah
so
one
one,
quick
update
from
the
announcement
side
so
suppose
ECS
and
often
times
said,
will
prepare
the
draft
blog
post
both
down
the
initial
round
of
internal
review,
because
we
don't
want
to
make
it
more
public.
Before
we
made
sure
we
don't
put
anything,
that's
very
misleading
or
like
sensitive
there,
I
I
feel
by
Thursday
this
week.
B
We
should
be
ready
to
turn
this
into
a
more
probably,
not
not
making
a
final
post
but
I'm
going
to
change
the
Open
Times
rewinding
to
a
PR
on
open
permission,
open,
100.io
repo.
So
you
guys
can
see
and
comment
on
that
and
elasticity
will
probably
put
it
somewhere
like
in
their
stating
environment
and
then
we'll
we'll
do
the
final
runoff
review
and-
and
they
have
some
things
like
the
marketing
team
I
think
like
after
they
did
the
announcements
on
April
the
18th,
which
is
the
EO
cubecon,
now
all
just
published.
B
A
Yeah,
thanks
for
I
see
you
keep
pushing
on
that.
A
A
A
Okay,
cool-
and
do
you
remember
I-
was
thinking
of
I
may
have
some
time
to
attack
to
do
that.
Smallish
PR,
to
remove
schema
translations
from
stability.
A
The
way
we
left
it
on
Monday
with
Josh
was
I
think
he
was
going
to
do
that
next
week.
Was
that.