►
From YouTube: 2023-03-28 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
B
B
I
guess
we
can
start
it's
a
two
minutes
after
start
time,
yeah,
let's
start
I,
don't
see
yeah
holiday
base.
Never
mind
also,
don't
forget
to
add
your
name
to
the
attend
these
list.
Please,
okay!
So
let
me
share
my
screen.
The
first
item
is
by
Robert
from
the.net
after
instrumentation
group.
He
mentioned
that
he's
not
coming
by
the
way
there
we
are
so,
but
he
wanted
people
you
know
to
get
people's
attention.
B
B
B
Jack
until
you
already
reviewed
that
so
that
that's
good
I
think
we
should
get
maintainers
opinion
on
this
one,
just
in
case
or
actual
instrumentation,
probably
instrumentation
teams
in
staff
Trask
is
still
well
not
around.
What
I
would
love
to
hear
he's
I'm
here,
I.
C
Hey
hey,
this
is
not
for
http,
but
we
can
I
can
ask
the
HTTP
group
about
this.
B
B
Okay,
in
that
case
yeah,
please
consider
it
consider
reviewing
that
one,
the
next
one
tigram
yeah,
do
you
have
the
link
for
out
it's
here.
Sorry
I.
A
Do
yeah
it's
it's
a
pull
request
to
clarify
how,
in
the
future
objective
changes
to
the
product
can
be
made.
It
was
it's
a
subset
of
another
PR
which
goes
I,
guess
farther
than
that
and
and
introduces
other
guarantees
about
how
the
future
of
the
Proto
can
evolve.
So
I
split
it
just
to
make
sure
that
we're
able
to
to
make
progress
right.
So
this
one
I
think
is
non-controversial.
A
So
please
take
a
look
if,
if
you
agree
please
approve
and
for
the
rest,
I
will
just
fix
the
other
PR
to
remove
these
bits,
and
then
we
can
discuss
what
remains
there
anyway.
This
is
just
part
of
the
effort
to
to
do
its
layering.
The
Proto
1.0.
A
There's
a
there's,
a
tool
that
we
use-
I
don't
know
if
it's
it
can
be
configured
to
do
exactly
what
we
want,
but
we
can
look
into
that
thanks.
B
Thank
you
so
much.
The
next
one
is
one
that
is
not
actually
mine,
but
this
has
been
lying
around
and
it's
about
whether
we
should
collect
or
not
collect
by
default.
B
You
know
the
database
statement
because
of
you
know,
privacy
reasons,
and
there
has
been
back
and
forth
opinions
on
this.
One
and
I
was
trying
to
summarize
things
and
I.
Don't
know
whether
it
makes
sense,
but
I
wrote
down
there,
but
essentially
I
think
that
what
people
want
is
to
enable
these,
in
whatever
form
it
exists
like
in
Java,
there's
something
in
place,
and
you
collect
that
by
default.
B
If
there's
some
kind
of
Sanity
session
mechanism,
if
there's
none,
then
there's
a
big
suggestion
of
leaving
it
out
and
in
all
cases
you
provide
an
option.
So
the
user
can,
in
all
cases,
control
whether
you
know
distribute
by
default
or
not
I,
don't
know
whether
that
could
make
sense
for
everybody.
B
See
I
guess
I
I
could
be
curious
about
number
three,
like
you
think
it's
a
lot
of
work
to
support
the
option
to
allow
users
to
live
without
adjusting.
C
E
F
F
Opinion
of
number
two
I'm
Number
three
being
preferable
just
because
sanitization
is
some
more
work
in
my
opinion,
but
either
one
probably
works
and
because
queries
should
shouldn't
have
sensitive
data
in
them.
Anyways
should
be
in
the
args.
But
obviously
we
can't
rely
on
that.
But
if
it's
an
option,
then
the
user
can,
if
the
users
are
going
to
put
sensitive
data,
they
can
disable,
including
in
in
the
spans,
seems
reasonable.
B
B
Connection,
sorry,
you
were
saying
something
trust
and
then
I
lost
connection.
Oh.
C
Yeah,
what
I
was
going
to
say
is
that
with
number
two
I
just
didn't,
if
you're,
not
if
you're
not
doing
sanitization,
then
you
wouldn't,
then
it
would
be
opt-in
for
users.
F
E
F
See
that
makes
sense.
B
Yeah,
we
could
start
with
that,
and
you
just
hit
the
rate
based
on
the
feedback.
I
am
yeah
as
far
as
mentioned.
If
there's
no
feedback
like
negative
feedback
about
what
Java
does,
we
can
start
with
that
and
if
people
really
want
to
actually
have
these
included,
even
if
it's
like
exposing
all
these
things
with
all
the
initiality
station,
then
we
can
discuss
it
by
then,
but
not
before.
B
Okay,
okay,
so
if
that
makes
sense,
because
I
don't
hear
any
other
opinion
on
this,
mother
I
will
mention
in
the
in
the
issue
in
the
pr
that
we
want
to
go
forward
to
yeah,
hopefully
that
no
I'm
just
leaving
the
door
open
for
further.
You
know
tuning
for
that.
Okay,
thank
you
perfect.
So
I
will
mention
that
moving
on
the
next
one
is
about
adding
whether
we
should
be
allowed
or
not
to
specify
invalid
links.
B
Sorry
response
sorry
valid
response
context
or
leaves,
even
though
it
doesn't
make
sense,
there's
a
potentially
scenario
there
and
which
is
that
you
are
recording
a
batch
for
example,
and
then
you
find
out
that
some
of
the
messages
wouldn't
include
the
actual
parent,
but
there's
still
a
lot
of
very
valuable,
valuable
information
about
them
like
retries
or
message
number
message
size
whatever
and
in
that
in
those
cases,
probably
you
want
to
keep
that
information.
B
Even
if
there's
no
parents-
and
there
were
there-
were
arguments
against
that
and
anyway
and
I-
guess
that
this
is
probably
the
most
interesting
comment
there
and
probably
one
thing
that
I
would
like
to
leave
out
for
now.
Is
this
because
that's
a
different
discussion
we
will
be
talking
about
later,
which
is
representing
links
as
events,
but
just
thinking
about
these
just
what
about
what
people
think
about
this?
H
This
would
be
helpful
for
an
occasional
issue
that
comes
up
for
some
AWS
users,
where
the
application
load
balancer
sends
a
trace
context,
header
that
has
just
a
trace
ID,
but
no
span
ID,
and
thus,
once
that's
received
by
a
receiving
service,
it
gets
treated
as
an
invalid
context
and
drop
in
a
new
new
Trace
ID
and
span.
Id
are
created
being
able
to
link
that
in
that
case
would
be
helpful.
D
So
so
my
suggestion
would
be.
We
can
include
this
data
if
the
user
provided
such
data,
at
least
as
you
can
see
the
data
and
understand,
what's
happening
if
it's
time
to
drop
it
on
the
floor.
This
will
be
very
hard
for
the
user
to
understand
unless
they
turn
on
some
internal
SDK
troubleshooting
tool.
G
G
D
D
It's
not
not
something
I
feel
we
we
could
control
and
maybe
just
give
the
user
availability
an
individual
instrumentation
libraries.
They
can
decide
how
to
handle
this.
Maybe
they
just
record
this
or
if
they
have
specific
reason.
As
long
as
it's
clear,
the
user
can
understand.
Actually
that's,
probably
okay,.
D
G
This
I
think
explains
if
there
was
really
invalid,
like
not
properly
generated
or
partial
Trace
context
and
I
would
agree
there.
What
about
if
there
is
no
Trace
context.
G
H
G
I
cannot
speak
about
all
implementations,
but
I
know
that
Java
even
ux
Rock
the
context.
You
only
know
that
there
is
no
context,
because
it's
invalid
it
gives
you
an
instance
of
context,
an
invalid
one,
essentially
as
an
instrument
like
instrument
instrumentation
using
propagator.
You
don't
even
know
if
there
was
a
context
and
it
was
invalid
or
there
were
no
context
at
all
yeah.
F
B
Put
a
comment
in
the
chat:
I
guess
that
yeah
back
then
you
had
raised
concerns
about
this
in
the
past.
I
think
I
am
not
correct
me,
but
anyway,
what
you're
saying
now
is
that
you
will.
You
would
like
to
have
a
config
option
at
the
at
the
SDK
level,
so
basically
touch
links
with
an
invalid
span
contacts
our
job,
because
in
case
you
don't
want
to
pollute
your
back
end.
G
B
I
B
B
Aaron,
you
are,
you
were
mentioning
in
a
comment
like
in
a
backend
whether
there's
significant
significant
difference
between
a
link
with
those
zeros
or
a
one,
with
a
valid
that
was
sampled
away.
Yeah.
That's
an
interesting
question.
I
Have
done
my
my
two
cents
here?
It
may
not
be
a
difference,
but
but
there
is
a
chance.
So
if
we
know
that
we
are
sampled
out,
maybe
arguably
maybe
we
should
drop,
we
should
offer
early
an
option
to
drop.
That
link
as
well
again
may
not
be
useful
for
everyone,
but
some
back-ends
may
act
weirdly
if,
if
that's
not
varied,
but
but
what
I'm
trying
to
say
is
I
think
we
should
offer
users
the
capability
of
filtering
out
these
links,
somehow
not
necessarily
on
I,
mean
an
option
in
SDK.
I
Another
option
is:
maybe
we
do
spam
processor
or
something
that
does
this
or
whatever
other
component.
That
allows
me
to
filter
this
out
somewhere
in
the
in
the
in
the
collector,
is
also
a
possibility,
but
I
think
it
would
be
nice
to
have
it
in
the
SDK
somewhere.
G
Sorry,
essentially,
we
need
a
way
to
remove
a
link.
We
need
to
make
sure
it
exists,
correct.
G
G
B
B
B
Basically,
he
mentions
that
you
know
it's
actually
a
long
post.
We
don't
have
to
read
here
holy,
but
basically
he
mentioned
why
he
thinks
that
they
good
links
will
work
better
as
events
and
one
of
the
reasons
that
he
mentions
about
this
is
that
well
we
we
at
least
the
way
we
are
them
these
days,
because
we
don't
have
an
actual
model,
but
if
we
were
to
model
them,
probably
we
should.
We
are
going
to
start
modeling
that
we
should
probably
consider
this
I
am
mostly
curious
about
work,
dance
opinion
on
this.
B
B
So
no
answer,
I,
guess
that
in
that
case,
as
I
said
before,
this
is
a
rather
long
issue,
which
is
great,
but
of
course
he
needs
some
time
to
you
know:
okay,.
I
G
I
read
this
issue
and
I
think
what
Josh
is
proposing
here
is
that
we
have
the
same
data
structure.
That
link
can
be
it's
still
a
link
it
can
be
attached
to
it
can
be
an
event,
then
it
has
Trace
ads
1080..
B
H
So
I
think
what
Josh
was
recommending
here
is
that
the
add
event
method
would
take
an
additional
option
with
link
context
and
when
it
receives
that
option
instead
of
creating
an
event,
it
would
create
a
link.
So
it
wouldn't
be
changing
anything
about
how
links
are
modeled
or
they're
the
the
fact
that
they
are
indeed
special.
A
H
You
but
I
think
he's
he's
recommending
adding
those
even
to
things
that
are
added
at
the
start
via
links,
so
links
added
later
by
add
event
would
be
no
different
than
Link's
edit
at
the
start,
except.
J
A
So
I
think
this
is
really
two
different
proposals.
One
is
about
extending
what
links
can
represent.
The
other
is
how
you
record
the
links
through
the
API,
so
that,
instead
of
using
a
special
ad
link
API,
you
use
a
more
generic
ad
event
API
with
some
special
parameters
to
indicate
that
it
needs
to
become
a
link
in
the
data.
B
Think
I
think
it's
like
records,
like
exception
more
like
that,
but
anyway,
whatever
the
name
would
yeah,
it
would
be
like
you
are
actually
we're.
I
Yes
sure
we
again,
everything
is
possible,
but
I
would
keep
the
concepts
differently.
The
Proposal,
of
extending
the
span
links
data
model
I'm,
all
about
like
we.
We
know
it's
not
enough.
We
know
we
have
to
add
more
things,
but
that
doesn't
necessarily
mean
we
need
to
overload
the
the
event
API
for,
for
these
agree,.
A
A
Anything
that
you
do
with
a
span.
Technically,
you
could
say
it's
it's
it's
some
sort
of
an
event
right,
that's
logically,
not
wrong,
but
there
is
a
reason.
You
model
the
apis
with
you,
give
them
different
symbolic
names
right
to
make
it
easier
to
understand.
B
B
Okay,
thank
you
so
much
for
that.
Oh
this,
the
next
one
from
gym
idea
as
well
I
am
far
less
involved
in
this
one.
By
the
way,
so
I
will
let
metrics
experts
talk
about
this
one.
B
I
see
approval
by
the
way
by
Jack
and
Riley,
so
yeah
I
don't
know
what
what
else
do
we
still
have
pending
comments
there
by
the
way,
like.
J
I
haven't
I,
haven't
reviewed.
Josh's
last
commits
that
he's
pushed
and
his
in
his
comments.
You
know
there
is
a.
There
was
a
round
of
reviews
that
was
last
week
that
you're
you're,
showing
here
where
there's
a
couple
of
questions
about,
is,
is
phrasing,
and
you
know
I
I
tried
to
reply
to
that
series
of
comments
as
best
as
I
could
and
and
Josh
has
just
pushed
to
commit,
which
I
think
addresses
any
outstanding
issues
there.
J
So
you
know
I,
guess
the
the
folks
that
have
reviewed
it
and
approved
it
like
Riley
and
myself
need
to.
You
know,
re-review
it
and
and
see
if
anything's
changed
and
the
folks
that
have
commented
on
it
and
and
haven't
approved
it
should
take
another
look
and
and
see
if
their
comments
have
been
addressed,
and
so
you
know
I'm
not
sure
exactly
who
those
are,
but
probably
the
metrics
approvers
minimally.
B
J
Yeah,
so
if
you've
commented
this
on
this
in
the
past,
please
take
another
look.
I
think
this
is
a
really
important
thing
to
get
consistent
across
the
SDK
implementations,
and
you
know
I'm
excited
to
get
this
merged.
B
Now
to
the
Curiosity
thesis,
we
are
planning
to
do
the
usual
monthly
release
like
next
week
for
April.
Is
that
something
that
you
would
like
to
see
these
included
in
the
entire
release,
or
should
we
even
if
we
get
enough
approvals
till
we
wait?
What's
your
feeling
on
that
Riley
and
Jack.