►
From YouTube: 2022-07-08 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
A
B
B
A
third
person
just
entered
the
document,
so
I
assume
somebody's
going
to
be
joining
sometime
soon.
The
only
thing
I
had
on
the
agenda
was
carlos
wanted
to
get
the
two
open
tracing
shim
ers
merged
before
the
release.
D
B
Sometimes
the
food
arrives,
usually
thursday
night
is
take
out
night
and
sometimes
it
arrives
during
the
meeting,
and
sometimes
it
gets
here
before
the
meeting.
B
B
And
you're
still
at
microsoft,
are
you
going
to
transfer
into
the
blizzard
division
when
the
blizzard
acquisition
happens.
E
D
Right
so
you
y'all
are
gonna
release
this
week
still.
B
D
Yeah
you're,
probably
not
too
interested
in
the
gory
details
of
the
well,
the
http
semantic
convention.
Stability,
work
got
sort
of
derailed
by
diving
into
what
net
pure
name
and
that
pure
ip
really
mean
and,
and
it
ended
up
splitting
out
into
net
net,
like
kind
of
logical
like
net
pure
name
and
socket
level
like
so
now
we
have,
or
the
proposal
is
to
have
like
net
sock,
pure
name
and
or
net
sock,
pure
address
and
net
sock,
pure
port
and
yeah.
So
I
was
trying
to
help
move
that
along.
D
Since
there's
not
a
lot
of
eyes,
not
doesn't
seem
to
be
a
lot
of
interest,
but
it's
blocking
http
semantic
stability.
So.
D
If
there's
no
well,
I
guess
this
is
I
mean
it
only
affects
the
no
op
mode.
D
D
But
I
guess
the
other
thing
is
in
the
splunk,
distro
and
they're
using
this
in
the
splunk
distro
is
we
have
a
customized
hook
where
we
take
in
existing
config
and
we
can
like
conditionally,
set
other
things
based
on
what
the
user
configured,
and
so
we
don't
really
have
a
way
to
do
that
in
the
auto
configure.
D
I
mean
we
could
add
a
hook
in
auto
configure
to
do
that.
That
takes
the
config
properties
and
returns
a
config
properties
or
matthias
was
wondering
about.
You
know
what,
if
the
agent
you
know
is
responsible
for
constructing
the
config
object
and
passing
it
to
the
sdk,
which
sort
of
ring
a
bell
of
what
the
carcass
folks
wanted
at
one
point,
sort
of
to
be
able
to
completely
basically
replace
the
config
properties
and
only
allow
people
to
configure
it
through
their
config
source.
B
B
Like
all
of
the
I
mean
all
of
the
the
end
like
do
you
have
to
have
your
own
copy
of
all
the
n
bars
and
all
of
the
system
properties
and
that
stuff?
In
order
to
do
that,.
B
D
That
takes
and
the
config
properties.
D
Map
we
don't
necessarily
want
to
create
a
whole
new
config
properties,
object
because
the
it's
not
it
doesn't
have
setters
on
there.
B
B
Mutable,
oh
sure,
but
if
you
grab
it
you
could
always
wrap
it.
I
mean
if
you
can
get
a
map,
you
could
always
wrap
it
in
a
beautiful
hatch
map
and
be
ready
to
go,
but
we
don't
have
that
so
never
mind
I
mean
we.
Could
I
mean
that
that
if
that
would
be
something
helpful
like
expose
like
do
we
have
a
way
to
get
all
the
properties
on
this
thing
right
now
or
do
you
have
to
just
know
your
properties
ahead
of
time.
D
Either
we
pass
in,
we
get
a
config
properties
and
we
create
a
we
return,
a
new
one,
which
I
guess
is
kind
of
what
you're
saying
of
wrapping,
but
it
might
be
more,
it
might
be
easier
just
to
take
into
config
properties
and
pass
back
a
new
map
that
should
get
overlaid
on
top
of
that
config
properties,
essentially
like
a
new
properties
supplier,
but
it's
a
buy
supplier.
I
don't
know,
is
that
a
thing.
B
A
buy
supplier,
no
because
you
don't
end
up
with
you,
can't
have
two
return
values
so
you'd
have
to
like
have
a
supplier
of
a
pair
or
something
like
that.
Right,
no.
D
No
sorry
yeah,
I
I
got
it
mixed
up
in
my
brain.
I
guess
just
a
function.
It
just
takes
in
a
config
properties
and
returns.
Yes,
yeah!
That's
a
function.
B
D
We're
trying
to
solve
well
there's
two
different
problems
here
and
matesh
kind
of
asked,
and
he
said
these
two
are
essentially
the
same
issue.
This
is
the
no
op
problem.
D
I
think
it's
just
for
debugging,
but
this
one
is.
I
know
that
splunk
uses
this
where
and
so
they
need
a
hook
somewhere.
So
we
had
a
hook
that
took
our
final
config
object
and
allowed
you
to
produce
a
new
config
object
from
that.
B
D
B
I
mean
that's
why
I'm
trying
to
just
ask
if
that's
what
the
problem
is
trying
to
be
solved?
Is
you
want
to
be
able
to
utilize
the
auto
configure
config
that
gets
produced,
but
then
you
want
to
let
agent
well,
you
want
to
let
the
agent
to
do
some
customization,
because
there's
things
that
aren't
from
the
sdk
and
you
want
to
let
extension
agent
extensions
also
over
add
their
own
properties
to
it.
Is
that
basically
the
and
then
be
able
to
hand
that
back
to
the
sdk
to
be
able
to
work,
I
guess.
D
Oh,
I
see
what
you're
saying,
because
the
if
we
want
to
modify
that
makes
sense
if
we're.
E
A
D
Yeah,
I
agree
try
to
think
of
matisse's
concern
here.
B
B
A
B
D
D
Reading
but
I
don't
see
a
reason
why
in
that
case,
when
it's
basically,
instead
of
using
the
empty
config
properties,
we
can
you
know
duplicate
that
for
the
no
op
case.
C
D
Okay,
so
let
me
I
will
respond
with
our.
Let
me
write
down
our
points
just
so
I
don't
get
them
wrong.
C
D
D
You
call
the
method
because
the
way
that
this
guy
works
is
you
implement
the
provider.
D
B
D
Those
would
be
passed
in
because
it
would
be
a
new
method
here.
Oh.
B
B
With
a
weight,
this
is
what
I'm
confused
about.
What
is
like,
what
does
the
sdk
do
with
the
resulting
map.
D
It
merges
it
back
into
it,
creates
a
new.
The
sdk
would
create
a
new
config
properties,
merging.
B
B
Got
I
got
it
got
it
got
it:
okay,
oh
yeah!
That
makes
sense.
Sorry,
there's
a
lot
of
there's
like
who's
doing
what,
where
yeah
yeah
I
got
it.
Okay,
that
makes
sense,
and
the
only
reason
we
do
do.
Map
string
string
is
because
we
don't
really
have
a
concrete
implementation
of
config
properties
that
would
be
exposed.
B
B
D
B
C
B
B
B
B
D
Hey
just
when
I
I
just
saw
jason
right
here.
D
Okay,
but
then
this
one
is
more
powerful
if
you
want
to
make
decisions,
if
you
want
to
based
on
so,
I
guess
here
based
on
oh
yeah,
and
so
this
kind
of,
I
think,
is
what
like
the
exporter
example-ish,
that
john
was
guessing
about
based
on
some
other
property.
You
want
to
set
something
at
if
absent,.
B
The
delete
case
is
an
interesting
one
though
like
would
you
want
to
be
able
to
remove
a
property
and
would
would
putting
a
null
value
in
the
map?
Have
that
I
don't
know?
I
have
no
idea.
Does
this,
I
guess.
Does
this
solution
proposal
solve
the
delete,
a
property
use
case,
and
is
that
a
real
use
case?
I
don't
know.
B
B
B
D
B
B
B
D
B
B
D
A
E
D
D
Who
is
java
was
java,
so
I
I
kind
of
explained
that
you
know
if
we
wanted
to
add
it
like
built-in
support
for
oauth,
that
that
would
require
spec
issue,
but
that
I
some
kind
of
a
hook
in
the
exporters
that
allowed
sort
of
a
generic
authentication
type
of
a
hook
that
authentication
providers
would
need
to
be
able
to
would
allow
them
to
do
what
they
needed
to.
D
And
that
might
it
might
be
just
that
simple
of
you
know
lambda
that
gets
called
each
time
to
populate
the
headers.
D
I
thought
it
was
him
mentioned
there
back
in,
but
I
don't
know
who
they
are.
Okay,.
E
A
A
End
up
with
a
continuing
stream
of
methods
for
different
authentication
mechanisms.
If
we
go
with
the
simple
one
right
now
because
like
for
example,
they
don't
need
the
request,
parameters
or
headers
or
anything
for
oauth.
But
then,
if
they
go
to
a
different
like
aws,
auth
requires
them
to
sign
it
or
something.
And
so
and
then
we
need
a
yet
another
method
for
registering
that.
So
I
prefer
to
go
straight
to
something:
that's
more
generic
than
just
a
lambda
that
returns
a
string.
E
C
D
B
A
I
think
it's
relatively
safe
to
have
a
function
that
accepts
headers
and
payload
and
let
it
return
something
so
maybe
that's
the
hook
pad.
We
wouldn't
have
the
actual
helpers.
Unless
someone
wants
to
contribute
them,
maybe
in
contrib
or
something,
but
at
least
that
hook
itself
sounds
relatively
futuristic.
D
A
E
So
oauth,
as
a
mechanism
has
like
a
setup
right,
there's
like
a
little
bit
of
a
handshake
dance
that
happens
at
the
beginning.
Is
that
right,
I'm
rusty?
I
guess
the
implication
is
like
if
you,
if
we
went
with
option
one,
then
it's
kind
of
assumed
that
on
first
request
the
setup
would
happen
and
whatever
this
component
is,
would
be
stateful
enough
to
hold
on
to
that.
A
B
I
am
studiously
and
intentionally
ignorant
on
all
of
that
stuff,
so
I.
D
A
F
D
So
he
brought
up
where
these
should
live
potentially
or
about
brought
up
potentially
moving
them
to
contrib
and
for
two,
I
think
two
reasons
well
one
they
definitely
don't
really
fit
under
the
instrumentation
folder
yeah.
We
don't
shade
them.
Do
we
no
right.
D
E
E
D
E
B
We
use
spring
boots,
but
they
they.
It
was
all
manual
open,
tracing
instrumentation,
so
right
something
just.
A
D
D
As
to
emphasize
that
it's
an
alternative
to
it's,
maybe
not
like
what
we're
promoting
necessarily
over
sleuth,
that's
an
alternative
yeah,
I'm
still
very
wary
of
the
whole
bridging
I
I'm
kind
of
I
kind
of
a
fan
of
the
having
hotel
pure.
E
D
D
Yeah,
so
it's
another
kind
of
similarity.
Another
maybe
analogous
piece
is
the
static
instrumenter
static,
instrumentation
as
a
like
a
way
to.
D
D
D
Yeah
I
thought
that
made
sense
long
term.
You
know
once
as
treating
it
as
like
a
first
class
side
by
side
with
the
java
agent.
E
D
Koopawatch
is
continuing
to
work
on
it,
okay,
but
I
don't
think
it's
reached
stability.
I
mean
at
a
point
where
it's
at
least
we
haven't
mentioned
it
in
release,
notes,
yet
thinking
that
it
was
very
much
still
a
work
in
progress,
sure
sure.
D
I'm
looking
forward
to
it,
though
I
know
that
for
sure
we
have
I
I
can
drum
up
users
for
that.
Oh.
D
D
Yeah
yeah
we're
just
starting
to
push
out
runtime
attached
to
users
who
were
we
had
a
subset
of
users
old
on
our
old
product
that
we
had
a
spring
boot
starter
for
it,
and
people
loved
that
spring
boot
starter
and
there's
been
a
certain
class
of
those
that
even
given
all
of
the
improvements
and
new
stuff
in
the
java
agent,
they
have
resisted
moving
off
of
their
spring
boot
starter.
D
Which,
because
I
I
asked
I
kind
of
pushed
on
you
know
some
of
them
like
asking
why
why
they
like
it
so
much
what
what
we
would
need
to
do
to
get
them
into
the
auto
instrumentation
world,
and
it
was
very
much
about
the
they
like
it
all
to
be
deployed
and
bundled
up
into
their
single
spring
boot
app
so
that
you
know
they
can
shove
it
out
everywhere
and
not
need
to
change
all
their
deployment
scripts
and
push
around
a
java
agent
and
push
around
a
config
file.
D
D
Wow,
this
is
interesting.
If
you
it's
an
interesting
jmx
issue,
I
was
just
trying
to
get
more
eyes
on
it,
because
it's
a
proposal
from
the
app
dynamics,
folks
for
jmx
configuration
kind
of
generic
exporting
jmx
metrics
from
the
java
agent
and.
D
Well,
that
materials,
a
nice
idea
that
this
could
potentially
be
used
to
replace
the
jmx
component
by
the
library,
because
the
library
instrumentation
could
be
pointed
at
a
remote
jm,
a
remote
ambient
server
yeah.
D
Format
based
and
went
with
sort
of
jack's
idea
of
you
know
just
go
ahead
and
bring
in
a
yaml
config
for
it
and
have
a
property
file
property
that
points
at
that
file
and
then
eventually,
when
we
will
merge
all
the
yaml
together
into
the
master
config.
D
So
yeah,
if
that
interests,
you
take
a
look,
I
was
just
trying
to
get
feedback
to
them
since
that's
a
decent
before
they
go
too
far.
E
D
Duplication
for
sure
I
don't
know.
E
Yes,
that's
super
interesting.
That's
I
mean
the
runtime
metrics
are
the
same
thing
as
the
jvm
metrics
that
I'm
saying
right:
okay,
yeah,
yeah
yeah.
I
think
it's
super
interesting,
yeah
and
sort
of
unifying
it
might
be
nice,
especially
if
these
can
be
overlaid
or
blended.
You
know
if
there's
like
one
of
these
configs
that
ships
with
the
agent
and
then
a
user
supplied
one
if
needed
or
a
distro
supplied
one,
then
they're,
blended
or
overlaid
that
that
seems
like
it
could
work
nicely.
A
E
D
D
E
A
Just
for
random
share.
Okay
then
make
a
poc
of
a
wasm
webassembly
transformation
processor
that
can
be
useful
for
these
distros.
I
think
because
then,
customers
don't
rebuild
their
distro
or
sort
of
the
support.
If
splunk
is
interested,
maybe
they
can
help
some
to
sponsor
this.
If
not
it's
not
a
high
party
for
me
or
anything.
Yes,
I
want
to
make
sure
people
know
about
it
as
an
option.
E
Interesting,
that's
super
cool
and
you
never
sleep.
D
C
D
A
A
E
I
did
want
to
just
maybe
spend
two
minutes
talking
about
jager.
My
skin
made
your
my
stream
sure
I
I
I
joined
because
I
missed
this
morning.
I
had
some
other
things
going
on,
but
also
you
know
that
jaeger
thing
to
me
is
is
interesting,
and
I
I
mean
you
probably
saw
my
comment,
but
it
does
feel
like
a
liability
that
we
continue
supporting
it,
and
I
I
respect
your
point
like
it's
not
like
it's
ancient,
like
it
literally
just
got
deprecated,
but
the
longer.
E
C
B
Is
that
is
jaeger,
exporter
support
for
all
languages?
Has
that
all
been
deprecated
and
repo's
closed
off
right?
Because
if
so,
this
sounds
like
something
we
should
be.
It
should
be
treated
at
the
spec
level,
it's
like,
because
we
have
requirements
in
the
spec
right
now
to
have
some
get
your
sport
knots
ripped
in
particular.
E
D
E
E
Which,
which
doesn't
surprise
me
that
makes
sense
and
also
the
thrift
ones,
it's
unfortunate,
because
you
know
yet
again
the
mono
repo
fails
us.
The
jaeger
java
client
has
the
thrift
bindings
in
it
and
if
that
were
its
own
repo,
then
that
would
not
have
impacted
us.
We
could
just
keep
sourcing
those
thrift
definitions,
but
because
they're
in
the
mono
repo,
that's
deprecated,
they
kind
of
were
collateral
damage.
E
B
E
B
A
E
Yeah,
okay,
but
they
might
they
might.
It
might
be
backfiring
if
they're
creating
more
work
like
if
they
have
a
long-term
supported
agent
that
still
supports
thrift
and
they
don't
have
java
bindings
that
are
updated
anymore.
For
that
they're
gonna,
create
problems,
crates,
work.
A
E
E
A
We're
looking
at
go,
they
do
also
generate
the
thrift
and
vendor
it
in
so.
E
There
is
oh
there's
a
there's,
one
more
class
that
we
depend
on
it's
the
exception.
It's
like
it's
all,
the
thrift,
binding
classes
and
then
there's,
like
some
exception
class,
that
we
special
case
in
the
exporter.