►
From YouTube: 2023-02-22 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
B
Yeah
sorry
for
Missing
the
the
time.
Thank
you
for
for
okay,.
A
Let's
see
I
really
can't
make.
B
A
D
A
A
D
A
D
A
For
the
most
part
use
status
code,
except
for
like
connection
failures
right
where
we
get,
you
know
status
code,
although
maybe
we
don't
get
an
HTTP
span
either
or
we'll
have
to
come
back
to
this
connect.
C
C
A
C
So
I
think
it
it
is.
It
is
useful
so,
but.
C
C
C
C
Yes,
and
for
the
backward
compatibility
reasons,
if
we
can't
introduce
new
thing
later,
we
can
remove
and
use
200
positive
numbers
before
and
maxint
minus
status
codes
to
represent
anything
that
we
want
in
future.
C
A
Is
it
I
mean
we
could
use
like
zero
here
the
mean
something
but
I?
Don't
love.
A
A
C
B
C
B
D
A
Potentially
Josh
tomorrow
also
for
and
Armin
since
he
I
think
they'll
both
be
in
the
meeting
in
the
morning.
He
had
originally
opened
this
or
commented
on
it
real
quickly.
Do
we
think?
Do
we
think
that
this
is
something
we
should
consider
before
stability.
C
B
D
A
Already
have
status
code,
so
we're
already
I
guess
so
I
guess
maybe
this
would
be
an
alternative
for
somebody
who
is
too
sensitive
to
cardinality
to
capture
status
code
itself.
B
Right
so
it
looks
like
people
don't
want
to
use
status
code
for
metrics,
because
it
can
have
a
lot
of
values
there
so
high
cardinality,
but
instead
they
want
something
more
generic
but
like.
If
we
reuse
the
same
attribute
but
change
the
type.
Let's
say
from
end
to
string.
B
A
C
A
D
A
B
So
basically,
in
mados
proposed
proposal,
we
have
like
a
two
things.
The
first
one
is
the
point.
When
instrumentation
Library
cannot
kind
of
capture
the
retry
capture
retrace,
then
it
should
be
just
one
one
span
yeah.
Basically
it
is
written
here
and
if
it's
possible
then
yeah
all
the
stuff
should
be
just
expressed
with
rice,
but
one
thing
which
actually
I'm
not
really
sure.
B
B
Should
be,
should
we
yeah
basically
like
it,
should
we
think
about
this
as
like
a
two
layers,
kind
of
TCP
and
HTTP
connection
established,
then,
will
it
just
work.
A
So
I
remember
what
the
main
concern
was
in
the
spec
media
was
related
to
The
Connect
span,
but
it
was
I
think
a
bigger
question
whether
the
connect
span.
A
Should
be
captured
at
all,
so
Bogdan
was
making
the
case
that
that
you
know
that,
with
a
hdb
connection
with
the
connection
pool,
you
grab
your
connection
and
you
use
it
and
where
does
the
connect
span
even
fit
in
there.
C
Can
you
hear
me
now:
okay,
yeah
I'm
I'm,
going
to
reply
to
mataj,
but
actually
I
want
to
propose
that
so
we
we
use
the
second
point
in
time.
We
believe
that
individual
tries
don't
cover
something
important.
So
basically,
instrumentations
should
create
HTTP
version
attempt
on
the
main
grade
and
encompassing
Span.
In
addition
to,
if
it's
impossible,
they
should
always
create
an
encompassing
span
and
then
it
kind
of
fits
nicely
into
the
there
is
no
connect
span.
C
There
is
no
concerns
if
something
happens
in
between
or
for
example,
if,
if
like
it's,
not
just
connect
right,
so
I
can
connect.
If
it's
the
first
time,
then
I
can
receive
headers
and
then
I
can
read.
The
Stream
and
problem
can
happen
when
I'm
reading
the
stream
right
and
it's
valid.
If
you
can
capture
this
stream
reading
to
capture
it
all
right
and
then
it's
basically,
we
have
encompassing
spell
and
exceptions
for
events
can
tell
us
what
happens
underneath
if
we
care.
A
A
C
Yeah,
so
if
you
can
capture
the
Gnostic
ones,
capture
them,
if
you
can't
but
can
capture
the
outer
thing
capture
it,
if
you
can
capture
both,
which
is
actually
not
super
common,
it
listed.net
it's
not
possible,
but
if
you
can
capture
both,
then
you
may
capture
out
there
the
logical
thing-
and
it's
probably
not
the
same
so
so
there
is
a
logical
thing
right
if
there
are
nested
ones
and
for
option
two
here
when
there
are
non-nested
there
is
a
physical
thing.
C
C
A
A
If
we
do
get
a
connect
failure,
we
don't
have
anywhere
to
record
it
like
we
could
create
and
we
don't
have
the
HTTP
details
needed
at
that
point.
We
don't
know
if
it's
a
get
or
a
post
or
anything
in
order
to
create
a
fake
HTTP
span.
B
Should
be
just
maybe
just
siblings,
meaning
that
we
have
just
like
a
sequence
of
spans
like
a
connect
will
be
the
first
one
or
like
a
TCP
level.
Spam
will
be
the
first
one
and
then,
if
it
works
fine,
then
we
just
go
to
http,
and
then
we
have
all
this
logic
with
rice
with
stop
level
like
a
logical
spam
or
something
I
mean
for
me.
A
B
Is
there
any
space
when
we
actually
established
kind
of
HTTP
connection
and
then
stand
get
requests,
and
this
part
is
not
really
clear
from
okay?
The
way
how
I
understand
it
is
basically
we're
just
establishing
statistic
connection.
It
might
be
also
TLS
layer
on
top
of
that,
and
then
we
just
send
all
this
get
posted
whatever.
It
is
right
like
we
operate
after
that
we
operate
on
HTTP
level.