►
From YouTube: 2023-04-03 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
B
There's
we're
trying
to
figure
out
what
to
do
about
recordings
in
open
Telemetry,
because
some
folks
are
concerned
about
their
likeness
being
copied
and
used
for
I.
Don't
know
if
there
sounds
like
there's
some
I'm,
not
really
sure
why
some
banks
are
using
voice
to
authenticate.
C
A
Okay,
so
you
know
we
will
need
to
have
avatars
for
our
video
have
modified
voice,
all
the
the
information
about
ourselves.
That's
what.
B
Yeah,
you
know
how
on
the
iPhones
there's
like
you
can
do
that
like
for
not
for
the
voice,
but
for
like
have
a
talking
Avatar
and
you
could
like
ruin
it'll
like
mimic.
What
you're
doing
I
haven't
seen
that
I'm
sure
that's
possible
on
Zoom
I've.
Only
seen
like
you
know,
like
the
dog
years
or.
C
A
B
I
haven't
been
on,
I
mean
Facebook
would
be
the
I
remember
on
Facebook
that
was
like
the
like.
Every
that
was
the
only
time
I
would
ever
get
messaged
was
you
know,
because
everybody
would
get
a
notification.
C
B
All
right,
let
us
reboot.
B
I.
Think
I'll
try
to
refresh
that
in
my
mind
before
Friday
and
I'll
ping
mitesh.
If
he
can
join-
and
we
can
talk
about
that.
A
A
A
A
B
C
A
A
B
B
Okay,
that's
good
context.
I
think
I
can
chat
with
Nev
and
see
if
we
can
resolve
his
concerns.
A
No,
if
enough
is
fine
moving
forward
with
that
I
think
that
that's
all
that's
done.
B
A
C
C
B
A
Yeah
so
I
found
it.
I
was
sick.
I
I
was
trying
to
implement
the
conversion
in
Python
to
accept
the
follicle
for
the
older
version
and
they
were
I
realized
I,
don't
really
understand
all
the
mappings,
so
I
decided
to
the
full
exercise
to
switch
specs
and
to
generate
the
new
semantic
conventions
in
Python
and
then
I'll
sunboard
their
cement,
their
instrumentation
libraries.
A
So
the
first
thing
is
that
yeah
most
of
the
mappings
are
straightforward,
but
I
actually
I'm
really
not
happy
about
the
nut
thing
and
the
problem
is
not
is
the
name
and
the
idea.
So,
let's,
let's
get
back
for
a
second,
so
they're
not
is
transparent
to
both
sides
right
the
quantum
server
until
unless
you
use
a
specific
protocol
to
know
that
not
exists,
you
wouldn't
even
know
so.
B
I
thought
we
were
I
agree
about
not
liking
the
net
I
thought
we
were
proposing,
potentially
to
use
socket
here
to
do
server.socket.address.
A
We
had
a
combination
of
things
so
I
think
we
we're
supposed
to
use.
We
wanted
to
introduce
server
socket
address,
but
the
IP
and
Port
remained
from
that.
So
if
we
introduce
socket
that
thinks
get
better.
A
A
Awesome:
okay,
so
if
we
do
sockets,
then
then
I
still
think
that
The
Proposal
we
had
is
not
perfect.
So
if,
if
I
switched
to
my
screen
for
a
second
yeah
yeah,
because.
A
I've
been
creating
some
charts
to
wrap
my
head
around
it,
so
I'm
going
to
share
the
screen.
Let
me
see
if
I
can
find
pictures.
C
A
Okay,
let
me
see
if
I
can
explain
this
so
on
the
client
side,
we
know
two
combination
of
things.
One
is
what
we
are
calling
this
guy,
that
describe
a
server
and
a
direct
peer
right
so
which
can
be
proxy.
It
can
be
the
server.
So
we
need
to
have
a
combination
of
server,
address,
server,
port
and
role
RT
about
February
2
for
a
second,
so
about
the
direct
peer.
Let's
say
it's
a
server
socket
address.
C
A
A
B
A
Okay,
then,
then,
if
you
look
into
server
side,
so
the
server
on
server
is
what
server
knows
about
itself,
but
they're
right
again,
two
combinations,
one
is
what
came
into
the
host
header,
where
client
called.
A
A
So
these
two
are
logical
things,
and
this
are
the
local
information
about
the
server.
So,
for
example,
with
the
trance
in
Docker,
then
it
would
put
potentially
things
here,
as
it
looks
internally
for
itself.
C
A
So
if
we
introduce
in
this
one
here,
then
it
could
be
anything
that
server.
A
C
C
C
B
Yeah
I
think
it's
capturing
it
now
locally
for
me,
so
I
can
save
it
off
later.
Okay,
we'll
see
I
gotta
try
this
on
a
couple
of
Hotel
meetings
this
week,
okay,.
A
A
B
A
A
Right
up
but
forget
the
names
for
now:
there's
fully
qualified
exists
per
server
only,
but
not
for
the
client
right,
so
client
does
not
need
one
of
them.
But
since
it's
a
socket
thing
it
has
an
address.
A
A
Yeah
but
then
the
problem
happens
for
server
essentially
for
Server.
We
need
two
different
things:
three
different
things:
the
DNS,
the
proxy
DNS
and
the
IP
address,
whatever
it
is,
the
there
we
get
really
into
the
conflict
with
ECS.
A
A
B
If
the
other
case
here
would
be,
if
you
were
just
connecting
to
an
IP
address,.
A
Uh-Huh
then
this
guy
would
be
S1
and
there
will
be
no
damage.
B
C
B
Mean
I
I
think
it's
okay
to
not
just
to
to
not
duplicate
things
when
we're
like
to
stay
with.
As
long
as
I
mean
I
see
the
the
alignment
has
at
the
field
level
and
less
at
the
logic
level
of
like
when
to
apply
it
when
to
fill
these
fields.
B
A
Okay,
okay,
then
it
makes
sense.
Okay,
so
then,
thus,
oh
in
our
case,
what
I'm
suggesting
that
IP.
A
And
on
the
circuit
level,
still
since
IP
is
an
address,
actually,
we
should
not
use
IP,
and
maybe
we
should
use
a
breast,
and
maybe
this
one
should
be
dummy.
If
we
stay
aligned
with
the
ECS
terminology,
we
used
to
call
it
name
right
now.
We
can
call
it
domain.
What.
A
A
Okay,
so
let's
see,
let's,
let's
start
with
a
simple
case-
we
are
connecting
directly
to
this
thing
and
we
don't
Define
server
domain
for
our
conventions
that
they're
connecting
directly
to
this
AP
address
than
the
socket
address
would
be
this
one
in
this
example,
just
really
kill
this
socket
address
so
for
the
simple
case,
direct
connection,
it
doesn't
even
matter,
but
for
the
proxy
the
stock,
the
server
address
would
still
be
the
example.com
and
or
the.
A
A
B
So
what
about
I
mean
is
proxy
just
a
can
we
not
is
proxy,
maybe
like
a
special
case
and
not
does
it
help
to
generalize
that
I
mean
essentially
it
proxy
was
and
I
don't
really
know
what
a
Nat
is,
but
I
was
assuming
and
that
is
kind
of
a
proxy
or,
and
that
is.
A
A
It
will
have
maybe
public
properties
for
for
instrumentation
to
see
that
the
connection
is
established
as
proxy,
and
then
you
know
this
guy
will
do
the
DNS
for
me
and
I:
don't
need
to
do
it
myself,
right
or
I
will
do
it
against
proxy
that
comments,
that
of
example.com
right
in
case
of
not
it's
the
thing
between
me
and
the
internet
that
hides
my
IP
address
from
the
internet
right.
If
it's.
A
A
Yes,
there,
so
there
are
websockets
proxies
on
the.
What
does
this
level
four
of
Network
stack
right,
so
there
could
be,
but
mostly
like
the
DNS
is
privileged
to
TP
on
this
specific.
So
we
should
expect
to
have
this
thing
mostly.
C
A
A
B
So
ignoring
proxy,
is
there
a
problem.
A
There
there
is
misalignment,
no,
there
is
no
problem,
so
this
is
the
whole
new
thing
that
we
never
had
before.
It's
not
enthusiastic,
it's
our
thing
and
ECS
looks
at
it
differently.
Probably.
A
A
A
B
B
A
Mm-Hmm,
okay,
so
then
it
sounds
like
with
the
turkey
disk
similar,
look
and
feel,
but
no
assumption
that
something
existing
should
work.
B
A
Okay,
yeah,
so
then
I
don't
think
there
are
problems,
but
what,
in
fact
suggesting
is
that
I
I
like
that
they
have
the
server
and
client
and
that
that
there
is
a
symmetry
we
used
to
be
very
symmetrical.
We
had
peer
and
post,
but
if
we
switch
to
server,
then
we
have
less
attributes
because
they
are
not
symmetric.
In
fact,.
A
A
Nice
awesome
yeah,
so
then,
if
we
switch
the
socket,
the
only
slide.
Questionable
thing
is
that
we
need
to
have
them
made
an
address
and
we
might
not
use
it
perfectly.
But
then
we
can
get
feedback
from
Alex
to
see
how
well
he
likes
it
and
then
yeah
no
problems
on
this
front.
A
So
if
we
change
the
prefix
for
this
thing,
then
we
should
probably
change
the
name
of
the
event
it
used
to
be
exception,
and
now
it
should
probably
be
ever.
A
B
Yeah
I,
it's
not
clear
to
me
if
that
is
part
of
those
things
that
are
part
of
stable
I
mean
I,
don't
it
some
people
have
said
that
they
thought
that
exception.
We
couldn't
change
exception,
because
it
is
part
of
the
SDK,
but
at
least
I
know
in
Java
I
mean
we
spell
out.
It's
binary.
Compatibility
is
what
we
guarantee
and
so
that
I
don't
think
that
limits
us
from
changing
it.
B
A
A
Yeah
and
the
last
thing
I
discovered
that
okay,
two
two
more
things
first,
is
that
the
sampling
I
mentioned
in
the
morning
that
this
attribute
Transformations
the
polyfills.
They
cannot
work
as
a
processor
and
as
the
case
they
have
to
either
happen.
A
B
Yeah
I
think
that's
a
really
good
point
the
Samplers
and
makes
me
wonder
about
yeah.
B
B
A
So
the
upper
grades
and
those
breaking
changes
probably
is
very
difficult
for
all
languages
and
they
stay
very
far
behind.
So
essentially,
it's
not
clear
what
to
stay
compatible
with
it's,
not
the
version
before
stability
like
before.
You
see
us
plus
one.
It's
a
whole
set
of
breaking
changes
before
that
they
didn't
onboard
yet,
plus
that
one.
A
Yeah,
so
it's
more
like
the
tracking.
How
do
we
reach
stability
and
how
do
we
resource
all
the
updates
across
the
board
to
make
sure
people
actually
went
there.
A
So
I
think
we
should
be
very,
very
strict
on
this,
so
I
think
the
prerequisite
is
that
we
are
not
maintaining
anything
we're
not
making
older
versions
work
like
1.11.
We
should
first
go
and
update
python
to
the
version
that
is
prior
to
ECS,
so
the
current
then
from
there
we
can
guarantee
the
com.
We
can
add
the
flag,
it
might
be
harder
for
python
because
they
don't
have
clear
extractors
and
it's
a
lot
of
duplication
but
I
think
it's
manageable.
A
This
third
place
where
it
can
be
hard,
is
Javascript
I
guess
because
they
have
tons
of
libraries,
but
that
their
languages
should
be
easier
because
they
at
one
or
two
HTTP
clients
where
it
should
be
manageable.
Anyway,.
B
B
Yeah,
and
so
we
could
I
mean
I
I,
think
that
somebody
had
thrown
out
like
we
should
have
a
six-month
transition
period
or
you
know,
whatever
the
trend,
you
know
a
year
transition
period,
where
we
support
that
flag.
A
Yeah,
okay,
it
makes
sense
yeah.
So
then
the
last
thing:
if
we
need
to
do
this,
we
should
follow
the
process
we
discussed
today
that
we
don't
remove
attributes
when
we
transition
to
ECS,
we
mark
them
as
deprecated.
We
have
the
flag
and
the
build
tools
support.
Then
one
that's
generated
they
are
still
generated,
but
as
deprecated.
B
A
B
A
Cool
okay,
so
then
I
will
update
the
draft.
Pr
I
will
split
it
into
different
things,
and
I
will
also
work
on
the
python
prototypes.
A
Implementing
instrumentations
this
way
will
also
be
much
nicer,
because
we
then
don't
need
to
support.
We
don't
need
to
create
policy
for
different
signals
right.
B
I
can
do
the
same
for
Java.
A
B
Cool
I
think
I
will
probably
open
on
a
spec
issue
with
this
proposal,
because
that
has
been
coming
up
a
few
times
and
I
think
it
would
be
good
to
get
some
more
clarity
on
that
yeah.
B
A
B
Yeah,
oh
yeah
I
saw
your
you
started
thinking
about
messaging
ECS
under
messaging.