►
From YouTube: 2023-02-13 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
B
Okay,
I
think
Tom
is
not
online.
I
cannot
see
online,
so
you
can
just
send
him
a
message
and
probably
we
can
start.
B
B
Regarding
the
specs
so
from
last
week,
I
think
we
discussed
about
allowing
the
adding
links
after
span
creation.
This
looks
like
something
probably
this
is:
has
sufficient
Quorum
I
think
this
is
something
which
looks
like
most
of
the
people
have
already
approved,
and
hopefully
we
have
the
first
breaking
change
in
the
traces.
After
a
long
time,
I
mean
some
something
which
is
going
to
break
at
least
our
API
compatibility
and
also
I
think
go
is
also
going
to
suffer
with
this.
These.
B
This
is
something
which
is
going
to
be
their
API
also,
but
I
think
this
is
something
which
has
which
is
going
to
be
added,
and
that
would
be
I
think
would
be
chance
for
us
to
consolidate
any
other
Avi
breaking
changes
which
we
feel
probably
which
we
want
to
add.
It
I
think
this
would
be
a
good
time
to
consolidate
everything
and
probably
have
one
release
with
the
which
breaks
our
API
will
increment
our
API
version
right
now.
B
It
is
one
I
think
we
can
increment
it
to
two,
but
that's
just
probably
probably
so,
I
think
we
have
to
plan
it
accordingly.
Once
once
this
gets
submerged.
B
Yeah,
so
it's
visible
right.
B
I
think
there
is
only
one
blocking,
and
this
blocker
has
been
basically
from
the
golang,
because
this
is
going
to
break
their
API
API,
but
I
think
it's
it
will
eventually
get
merged,
so
this
would
be
API.
Making
change
for
us,
and
probably
I
was
I
was
just
thinking
that
I
could
get
a
few
more
changes
which
we
have
been
planning,
but
I
think
we
cannot
read
just
because
there's
no
changes
for
API
making
a
really
good
Planet.
B
B
D
B
I
think
it
is
something
probably
we
all
can
do
it
I
mean
I,
don't
think
anyone
single
person
should
be
doing.
We
all
have
been
so
you're
applying
to
the
issues.
To
probably
anybody
remembers
any
of
the
issues
which,
which
was
something
I
think
it
was
a
valid
concern,
a
valid
change,
but
we
just
couldn't
do
it
just
because
it
was
easier
between
I
think
this
is
the
time
we
can
add
that
price
I'll
be
anyway
going
through
I
had
a
few
of
the
issues.
B
I
remember
them,
so
probably
I'll
tagged
them
with
the
ABA
before,
but
really
quickly.
I!
Think.
Anyone,
if
you
remember
any
infused
foreign.
B
And
move
towards
log
apis,
because
I
think
this
this
would
be
also
eventually
get
stabilized
to
both
log
API
and
SDK
would
be
Market
stable,
not
happening
yet,
but
log
API
is
being
changed.
The
name
of
login
is
being
changed
to
log
Bridge,
API
or
log
SBI,
so
any
of
any
one
of
these
PR
would
be
merged
once
this
is
merged.
I
presume
that
log,
API
and
log
sbk
would
be
marked
as
stable.
B
So
yeah,
and
apart
from
that
I
think
another
issue:
does
log
API
in
SDK
need
a
way
to
enable
log
11
inquiry
for
it?
I?
Don't
think
this
is
being
added
as
of
now
you
all
gave
to
here,
because
this
is
not
the
front
facing
log
API
which
which,
if
the
product
specification
is
providing
it
basically
a
backend
API
which
is
supposed
to
be
used
by
the
instrumentation
framework
login
framework.
So
it
will
not
have
a
code
for
log
level.
B
I
have
been
talking
about
adding
the
support,
at
least
at
the
SDK
level,
because
it
makes
sense
if
we
can.
It
does
make
sense
that,
as
if
we
can,
if
we
can
configure
in
SDK
to
enable
a
log
level
and
the
API
can
check
the
clock
level
and
accordingly,
it
can
do
that.
Do
the
logging
but
I
don't
think
even
that
would
be
added
in
the
current
a
current
SDK
specs.
B
It
may
be
added
afterwards
but
yeah,
as
of
so
I
think
the
log
level
will
not
be
added
as
part
of
blog
SDK.
B
Eventually
it
may
be
added,
it
would
be
definitely
not
added
in
log
back
in
deployment.
So.
B
D
B
The
end
users,
just
because
C
plus
plus,
is
the
only
language.
This
does
not
have
any
language
specific
login
framework,
so
there's
no
standard
login
framework
in
C
plus
Place,
unlike
other
languages,
so
we
may
have
to
come
up
with
some
convenience
or
some
extent.
Some
some
I
won't
say,
send
them
at
some
convenient
API.
B
So
yeah,
let's,
let's
solve
for
those
respects
there
was
one
I
think
issue
raised
in
see
open,
Telemetry,
C,
plus
plus
Also,
regarding
these
Response
Code
for
otlp
http
different
languages
are
handling
it
in
a
different
way
for
I
mean
by
things.
B
Which
one
is
this?
Okay,
this
is
for
collector
I.
Think
collector
looks
like
it
reads:
200
to
300
success,
Java
also
teach
200
to
300
is
Success
200
to
299,
probably
not
300.,
python
is
200
and
202
go
is
also
200
and
C,
plus
plus
is
also
200
and
200.
B
B
But
I
think
most
of
the
resource
semantic
convention,
which
we
have
been
using
would
be
stable,
so
service
name
would
be
something
which
would
be
stable
and
I
think
we
can
probably
see
here.
D
B
C
B
Yeah
yeah
the
coordination
generation
would
be
same,
but
this
will
mean
that
there
won't
be
any
further
changes
in
the
gamma
point.
Once.
B
So
give
that
Clarity
that
there
won't
be
any
further
changes,
yeah
yeah,
so
it
basically
means
that
the
service
name
and
the
Telemetry
SDK
group
here
I
think
these
two
would
be
stable.
B
Yeah
Telemetry
SDK
is
marked
as
stable
yeah.
This
means
a
lot
not
to
complete
results
on
this
multi-convention,
but
parts
of
that.
Ultimately,
as
we
said
and
component
unit,
these
are
not
more
than
stable,
but
Telemetry
SDK
would
be
stable
and
the
service
attributes
would
be
stable
and
and
these
these
both
we
are,
we
are
setting
as
the
default
values
we
are
setting
at
the
default.
Values
for
both
of
them
are
in
service.
B
B
So
I
mean
I
have
been
following
up
this
because
for
our
hotel,
CPP
I
have
been
writing
process
Matrix,
instrumentation
Library,
so
this
was
something
I
have
been
closely
following
it
right
now.
We
have
process
securitization,
which
is
more
of
a
difference
in
the
process,
if
you
time
since
the
last
measurement,
but
this
is
not
the
proper
way
to
get
these
Matrix,
because
this
needs
previous
state
has
to
maintain
in
the
instrumentation
library
to
get
this
value.
B
So
I
think
there
are
some
issues
in
in
using
this
Matrix
process,
CPU
utilization,
so
yeah.
That's
what
I
was
saying
that
if
you
can
modify
this
to
represent
the
real
time
percentage,
if
you
previously
I,
haven't,
got
a
given
process,
something
which
we
get
it
from
the
CPU
column
in
the
top
command
as
an
abdominal
counter
and
for
process.
B
If
you
count
the
problem
is
that
at
least
for
C,
plus
plus
application,
which
we,
if
on
Linux
there
is
no
proper
way
to
get
the
the
list
of
all
the
processors
which
are
current
or
all
the
cores
in
the
machine
which
are
being
used
by
the
current
process,
because
those
scores
can
keep
on
changing.
It
makes
sense
with
for
the
dotnet
runtime
and
also
for
jvm,
because
I
think
in
for
for
both
of
them.
B
We
can
configure
this
set,
of
course,
which
would
be
used
by
by
the
runtime
and
the
virtual
machine,
but
at
least
for
for
all
the
languages
compiled
languages.
There
is
no
way
to
get
that
so
every
time
the
CPU
account
would
be
total
number
of
cores
in
the
machine
and
the
same
value
would
be
imported
every
time.
B
B
Okay,
anything
else,
I
mean
I,
have
gone
through.
This
text
didn't
find
anything
which
probably
is
going.
C
Of
the
pr
number,
but
there
is
something
in
the
specs
which
is
a
PR
right
now,
which
is
moving
some
of
the
wording
from
the
API
to
SDK
related
to
naming
naming
traces,
naming
attributes
things
like
that,
and
when
that
is
merged,
we
will
have
to
move
some
code
as
well
from
the
API
to
the
SDK.
Okay.
Typically,
the
code
that
use
regular
expressions
to
make
sure
that
the
attributes
are
named
named
are
given
way,
for
example,.
B
Is
it
for
matrix?
B
C
It's
for
everything.
Let
me
okay.
C
Oh
right
in
the
middle
20
3171
this
one
so.
D
C
B
Talks
about
the
actual
logic
inside
that
I
mean
so
he
it
just
talks
about
validating
those
names.
Right
I
mean
validating
the
instrument,
name
and
versions,
and.
B
B
C
B
We
do
it
but
I
that's
in
the
SDK.
It
was
not
possible
to
use
it
in
the
API.
B
D
C
That
happened,
but
what
is
possibly
it
will
affect
some
code
and
also
it
may
affect.
It
may
be
a
breaking
change.
I,
don't
know
yet.
B
B
B
B
I
think
this
is
a
good
I
think
we
should
add
I,
think
only
only
the
header
files
which
are
part
of
the
API
basically
should
be.
It
means
by
default
installed
so
yeah
using
something
so.
C
C
By
the
way,
with
pattern,
metrics
I
think
it's
it's
a
bug,
which
is
a
leftover
from
the
time
we
had
the
underscore
metrics
and
metrics.
B
B
B
I
think
yeah.
B
Anyone
who
I
really
want
to
be
interested
to
take
it
over
or
we
can
leave
it
as
Help
Wanted.
Just
in
case
I
mean
it
is
something
I
think
it's
anyone
can
do
it
I
mean
it's
somebody
who
wants
to
contribute
externally.
Everything.
B
B
B
We
have
to
see
further,
it
looks
like
this
is
happening
while
shut
down,
but
probably
to
develop
it
further,
whether
it's
related
or
not,
I
can
I'll
check
it
or
what?
If
we
want
to
pick
it
up
otherwise,
I'll
pick
it
up.
D
B
C
B
B
B
First,
you
always
show
that
the
terror
is
there
and
it's
related
to
the
grpc
library,
but
in
the
the
same
error
is
not
coming
from
Ubuntu,
probably
something
to
do
that
these
liabilities
can
find
are
installed
from
the
package
and
if
that
is
something
making
some
difference,
but
you
have
to
check
this
further,
so
so
I
know
Ubuntu.
If
I
was
I
just
had
quick
check
in
Ubuntu
When
I
See,
This
Target,
all
the
grpcs
libraries
was
already
part
of
defined
in
this
target.
B
But
when
I
see
this
Target
and
Alpine,
the
grpc
libraries
were
symbols
for
missing.
Sorry
all
the
drpg
symbols
were
missing
from
this
Target
in
L
fine,
but
they
were
present
in
Ubuntu,
so
that
was
something
I
found
and
then,
after
that,
I
didn't
have
time
to
further
investigate.
B
B
B
C
B
B
So
so,
if
we
use
with
otlp
HTTP
with
cmake
that
will
not,
and
that
alone
will
not
work
right
now,
I
mean
that's
correct
right,
Mark
I
mean
we
have
to
use
this.
Also,
yes,.
D
C
Both
are
needed
eventually,
I
think
we
can
clean
that
up
and
and
just
choose
one.
Yes,.
B
B
But
I
think
the
term.
You
were
saying
that
it
was
working
in
okay,
not
this
sorry
yeah
in
VC
package.
We
had
okay,
it's
in
the
discussion
earlier.
You
can
quickly
see
that
this
one.
B
B
So
probably
it
is
working
in
Windows
I.
Think
it's
not
working
in
line
at
least
I,
don't
know
whether
it's
that's
a
difference
in
at
least
in
four
six
or
non-windows.
We
have
to
provide
both
these
two
or
in
fact,
all
the
three
grpc
or
grpc
libraries
also
installed.
Then
we
have
to
give
all
the
three
and
I
said.
No,
so
yeah.
D
D
B
B
B
B
B
B
So
yeah
I
think
probably
it's
it's
not
a
mandatory
attribute,
it's
still
as
a
recommended
attribute
in
the
semantic
convention.
So
we
will
see
what
specs
says.
As
of
now
I
mean
I,
don't
know
as
it
as
long
as
it
is
recommended
in
specs
is
saying
that
SDK
should
not
be
generated.
B
B
C
D
B
But
I
think
nothing,
nothing,
something
which
we
have
to
do
it
urgently.
I
think
not
every
language
is.
B
B
B
D
A
I
mean
the
crash
is
just
the
port
is
used,
it's
an
exception
that
is
not
handled
so.
C
B
A
B
Yeah
and
if
it
is,
if
the
exception
is
reaching
till
the
application,
the
SDK
I
think
we
can
I,
don't
know.
Probably
we
have
we,
we
should
do
or
not,
but
I
think
as
of
now
it's
okay,
if
we
even
if
we
document
it
that
they
should
use
it
correctly
so,
but.
B
I'm
not
sure
if
we,
if
we
can
catch
this
exception
at
the
application
Level,
we
can
do
that.
Otherwise,
I
think
it's
okay!
If
we
just
documented.
B
A
B
B
Okay,
I'll
probably
follow
up
and
probably
will
close
it.
We
discuss.
It
should
be
closed.
Add
example
yeah.
This
is
already
logs
because
complete
this
is
still
pointed
if,
if
they
are
using
a
domain
liability
which
is
installed
already
installed,
is
made
by
appointment.
Yeah
I
think
this
is
also
who
has
this
suppose
to
me.
Yeah
I'll
accept
it
add
example.
C
Coming
back
to
the
previous
one,
with
shared
pointers,
but
if
we
change
back
and
forth
between
STD
and
ostd,
that
could
be
a
breaking
change.
So
if
we
are,
we
we
may
take
the
opportunity.
C
B
B
D
B
C
B
C
B
B
B
B
And
we
are
not,
we
didn't
announce
it
yet.
C
B
B
C
I
think
we
need
to
see
how
exactly
STD
and
ostd
works
and
possibly
split
the
compiling
option
between
11
14,
17
and
20,
as
opposed.
C
D
B
B
B
B
Okay,
this
was
basically
providing
the
pool,
as
part
of
you
know,
that
span,
processor
or
whether
I
know
in
town.
You
know
spinal
position
options
here,
but
yeah
somebody's
want
to
implement
Authority
is
there,
but
I
mean
if
you
want
to
make
it
a
help,
wanted
or
somebody.
So
that
would
be
also
a
good
thing,
because
this
yeah
I
think
lots
of
people
would
want
to
use
their
own
threadful
customer
platforms
to
create
threads
or
at
least
for
the
advantages
so.
B
We
agree,
I
mean
it
differ.
Something
I
think
it
was
something
I
propose
like
if
you
can
add
that
threat
pool
as
part
of
batch
pad
processor
options,
and
then
this
is
something
which
we
can
add
this
support.
And
then,
if
somebody
is
by
default,
it
could
use
a
default
rate
pool
which
is
uses
our
default
language
sets
and
what
can
provide
a
custom
ones
if
you
want
to
use
it,
but
yeah
P3
I,
think
if
you
want
that
somebody
can
take
it
that.
B
B
B
Update
to
build
tutorial
yoga
yeah,
this
I
think
we
should
not
do
it
benchmark
test
yeah.
We
already
discussed
yeah.
This
I
think
I'll
leave
it
to
Mark.
Whenever
you
have
time
to
see
this
sure.
C
B
I
think
this
is
dropped.
I'll
close
this
one
open
tracing,
shim
I,
think
this
guy
is
very
patient
if
they
probably
try
to
close
the
terminal
just
because
I'll
review
it.
This
is
for
sure,
and.
D
B
B
Fixing
she
make
this
is
I.
Think
top
you
had
some
can
you?
Can
you
just
see
if
you
still
have
some
concerns
or
is
okay.
B
Okay,
I
think
that's
all
we
have
for
the
peers
and
yeah
I.
Think
that's
that's
anything
else.
We
want
to
discuss.