►
From YouTube: 2023-03-07 meeting
Description
cncf-opentelemetry@cncf.io's Personal Meeting Room
A
Not
too
bad
hoping
this
is
our
last
week
of
HTTP.
This
is
our
last
week
on
the
schedule
of
do.
B
You
need
anything
to
help
close
it
across
a
lot,
so
I've
been
very
busy
last
couple
months
and
I
wasn't
on
this
HTTP
specific
subject,
but
I
just
got
some
agreement
from
my.
B
D
B
C
E
A
And,
let's
see
it
through
some
things,
yeah
Briley,
so
we
were
chatting
a
little
bit
about
this
right
when
you
left
this
morning,
the
exception
stuff
and
I
shared
with
Ted
kind
of
the
earlier
discussion
we
had
about
trying
to
you
know
separating
HTTP
or
that
HTTP
doesn't
necessarily
rely
on
exceptions
and
that
we
could
stabilize
that
after
and
he
seemed
to
okay
with
that.
That's
cool
I
was
thinking
a
little
bit
more
though,
and
I
had
a
one.
Can
one
additional
concern?
A
So
if
we
don't
make
a
recommendation
on
that,
you
know
instrument,
server,
instrumentation
or
client
in
certain
client
spans
should
capture
exceptions,
wouldn't
that
lead
I
mean
that
could
lead
to
different
behavior
in
different
languages.
Some
people
thinking,
oh,
these
exceptions-
are
too
noisy,
I
shouldn't
catch
them.
Some
thinking
hey.
This
is
really
valuable.
Info
I
should
catch
them.
C
Uh-Huh
yeah
about
a
question
I
feel
the
same
question
would
exist
for
another
million
scenario
still
say:
hey
I
have
some
certificate
exchange
and
how
do
I
model
that
I
have
some
underlying
gcp
connection
research?
How
do
I
do
that?
What
about
DNS
resolution
so
I
would
say
like
by
except
all
these
are
nice
to
have
problems,
but
not
now.
A
C
Can't
wait,
you
can
shift
the
0.9
version
and
wait,
for
the
exception,
part
to
be
stable,
yeah.
My
point
is
always
like:
if
you
have,
if
you
want
to
go
to
the
fitness
there,
there
are
like
10
stars,
then
you
have
to
make
the
first
step
and
I
I
just
feel
we
won't
be
moving.
If
we
don't
know
number
one
where
we're
going
number
two
is
we're
saying
we
don't
know
where
we're
going.
So
what
just
goes
on
a
random
Direction,
and
hopefully
after
that,
we'll
be
able
to
figure
out
where
we're
going.
C
So
my
assertion
is
no:
if
not
do
that,
number
three:
is
we
don't
know
how
to
make
three
stars,
then
we're
talking
about
how
we
make
five
steps,
so
so
I
just
want
to
all
the
people
accountable
by
saying.
First,
we
need
to
be
very
clear
on
what's
the
direction
and
where
we're
going
number
two,
we
need
to
First
be
able
to
make
a
baby
step
and
we
can
out.
Then
we
should
stop
doing
this.
A
Okay,
so
if
I
can
rephrase
just
to
make
sure
that
I
got
it,
you're
saying
that
we
would
come
back
to
this
question
likely
when
we
stabilize
exception
semantic
convention.
C
C
And
maybe
a
fair,
fair
thing
we
can
put
here
is:
we
think
the
exception
is
a
very
common
thing.
So
by
just
working
on
HTTP,
we
don't
feel
we
got
good
understanding,
so
we
want
H3
to
be
stable
first
and
then
maybe
we'll
just
pick
messaging,
and
if
we
have
both
No
One's
Gonna
figure
out
how
to
make
exception
work.
So
that
won't
be
the
second
thing.
Maybe
that's
the
third
thing
that
we
work
on.
B
Just
to
be
clear,
so
what
we're
saying
are
exceptions
for
the
conventions
we
mean
any
kind
of
errorful
message:
that's
not
in
the
happy
path
right
so
for
the
go-long,
equivalent
being
like
an
error,
return
value
as
opposed
to
it's.
A
This
time,
okay,
we're
talking
about
in
particular
this
semantic
convention,
which
my
understanding
does
cover
the
it's
not
just
Java
exceptions.
It's
whatever
your
language,
yeah.
B
A
B
A
Yeah,
so,
basically,
once
we
finish
stabilizing
HTTP
semantic
conventions,
we
will
We
some
some
group
will
tackle
stabilizing
exception
semantic
conventions.
B
D
A
This
is
oh
yeah
just
wanted
to
get
a
second
opinion
on
that
I
can
un
remove
the
HTTP
semcon
locker
from
this,
because
I
feel
like
this
clarifies
now
that
we
can
evolve
the
schema
or
can
we
on
the
metric
side?
We
still
don't
have
an
answer
on
the
metric
side.
A
A
F
Breaking
dropping
is
fine,
oh
sorry,
popped
opt
in
with
breaking
is
fine.
A
Yes,
I
was
still
stuck
in
the
consumer.
Telemetry
world
guarantees
are
on
consumer
telemetry.
A
Sense,
okay,
cool
and
then
I
wanted
to
check
in
vanilla
with
you
on.
C
Let
me
go
back
to
the
previous
one
when
you
said
cool,
what
does
that
mean?
Your
code
is
removing
the
bloggers
so.
A
F
Yes,
I
think
the
zero
in
HTTP
status,
the
only
place
where
it's
important,
at
least
for
as
I
I
am
concerned,
is
that
we
have
a
we
need
to
decide
if
we
need
to
represent
HTTP
failures
to
differentiate
the
HTTP
failures
and
if
we
don't
do
it
now,
HTTP
failures
that
are
result
in
absence
of
response
right.
So
if
we
don't
do
it
now,
then
later
on,
we
won't
be
able
to
use
this
attribute
and
we'll
need
to
extend
the
HTTP
status
quad.
Instead,.
A
F
D
F
A
Stability;
okay,
all
right
that
and
drop
from
those
issues.
A
Before
we
get
to
this
one,
this
was
all
I
had.
Is
there
anything
that
else
that
we
should
use
today's
time
for
I.
F
Created
a
new
issue
as
a
result
of
morning
discussion
regarding
schema
transformation,
my
concern
yeah
I'm,
sorry
I-
can
find
the
link,
but
you
probably.
F
Yeah
so
basically
I
think
just
Dennis
you
weren't
in
in
the
morning
Telemetry
semantic
convention
stability
meeting,
but
the
concern
brought
up
there
that
with
ECS
alignment,
for
example,
we
are
not
ready
to
rename
attributes
after
stability,
but
it's
something
that's
currently
explicitly
allows
renaming
attributes
and
considers
it
a
non-breaking
change.
F
So
my
it
feels
like
a
contradiction
there
is.
There
is
not
no
confidence
in
the
community
that
we
can
rename
at
least
their
massive
change,
but
to
explicitly
allow
it,
and
with
this
issue,
I'm
suggesting
to
make
schema
transformations
to
keep
it
experimental,
but
the
enforcement
part
Josh
is
working
on,
would
rather
be
stable.
So
I
would
prefer
us
to
ship
stable
version
of
HTTP
semantic
convention
without
a
way
to
rename
attributes
after.
But
it
was
okay,
it
was
experimental
way
to
rename
attributes
after.
E
Yeah,
it
makes
sense
so
just
just
to
clarify
you're
saying
that
we
want
to
go
with
the
with
the
current
or
I
mean
if
there
is
no
ACS
in
place,
we're
going.
We
want
to
go
with
the
current
attribute
names
and
we
won't
be
kind
of
supporting
renaming
attributes.
F
No
I
don't
know
about
easy
as
well
like
the
community
will
decide,
but
what
I'm
saying
there
we
are
not
ready
to
rely
on
schema
transformation,
yet
at
least
okay
stability.
A
Blocking
renamed
until
there's,
everybody
is
convinced
that
we
can
do
the
schema
Transformations
yep.
B
So,
on
that
note
actually
Riley
have
you
had
any
contact
with
tigrid
in
the
TC
regarding
the
schema
versions
and
whatnot.
B
Sorry
about
that,
so
you
said
my
mic
is
noisy
yeah,
okay,
my
apologies.
So
have
you
had
any
contact
with
tigran
regarding
the
schema
versions
and
processor.
B
We
never
caught
them
Okay.
The
reason
I
bring
that
up
is
because
we've
been
mentioning
this
ECS
schema
right
and
an
internally
to
Splunk
right.
We
have
the
old
signal
FX
way
of
doing
things,
and
it's
kind
of
you
can
kind
of
think
of
it
as
a
similar
problem.
It
seems
very
related
to
me,
at
least
to
this
elastic
containers
came
up
in
the
sense
that
we
have
an
existing
set
of
metrics
with
you
know.
There
are
semantic
conventions,
you
know
internal
to
our
company
and
we're
trying
to
standardize
an
Hotel.
B
C
Oh
yeah,
so
I
added
him
to
the
ECS
related
conversation,
but
it's
unclear
when
you
mentioned
like
versioning.
What
does
that
mean
to
me
and
I
I?
Guess
it's
just
similar
like
how
we,
how
we
can
use
different
natural
languages
and
you'll
be
able
to
translate
right
so
I
have
less
hope
that
the
Universe
would
just
speak.
One
single
language,
so
I
apply
the
same
methodology
to
romantic
conventions.
B
C
B
The
selfish
reason
I
brought
that
up
is
because,
if
there
is
an
ongoing
discussion
about
this
General
type
of
use
case
about
hey,
we
have
this
external
schema
that
we
would
like
to
have
supported
to
some
degree
or
aligned
with
open
Telemetry.
You
know
if
there's
any
prior
art
and
more
minds
are
bashing
us.
This
problem
I'd
rather
not
be
alone
yeah.
Thank
you.
I'll
follow
up
later
yeah.
C
And
if
you
look
at
the
the
name
of
ECS
is
already
saying
it's
elastic
common
schema
right.
The
fact
it
has
a
common
schema
is
essentially
telling
everyone.
There
are
many
uncommon
schemas
and
these
maps
to
the
pharmacy.
So
in
Microsoft
they
also
have
a
Microsoft
common
schema
in
the
same
Spirits,
because
we
know
like
office
Windows
either
always
keep
inventing
their
own
schema.
Having
something
common
is
good
enough.
A
So
Ludmila
on
this
we're
just
trying
to
think
of
next
steps
and
who
yeah
yeah.
F
I
was
thinking
that
maybe
I
can
bring
it
up
on
tomorrow's
fact
meeting
and
raise
awareness.
I
just
want
to
see
what
people
think
about
schema
transformation
and
if
that
would
shame
in
in
the
same
way
he
came
in
today.
What
others
would
think
about
it.
A
D
A
C
This
is
the
old
inflammatory,
collector
has
offered
some
schema
translation
feature,
folder,
also
folks,
requesting
that
they
have
scenario
where
inflammatory
constant
directly
to
some
back
times
without
going
through
and
I
tried
towards
agent,
and
they
cannot
run
the
factor
anywhere.
So
they
expect
some
schema,
translation
and
currently
I
think
we
have
some
limitations,
there's
certain
things
that
the
schema
Administration
would
be
able
to
do,
and
there
are
also
things
that
it's
not
the
I
expect
over
time.
That
might
be
more
like
sophisticated.
Some
people
can
do.
C
C
And
also
with
open
times,
you're
getting
popular
one
thing:
I'm
I'm
guessing
is
there
are
many
other
existing
things
like
you
can
imagine,
there's
a
patch
log.
There
is
log
Windows
low
levels
here,
so
the
login
part
is
tricky
and
I
I
guess
the
last
day.
Folks,
we
have
seen
this
in
the
past
two
decades
so
like
with
the
like
elastics
and
open
Telemetry
working
together,
ideally
like
with
some
version
of
effort
and
I
guess
we'll
bring
in
a
lot
of
expertise,
foreign.
A
Okay,
yeah,
so
just
very
briefly
and
I
I
just
took
a
look
at
it
right
before
the
meeting.
It's
pretty
much
the
same.
F
A
F
A
I
see
yeah.
A
E
But
it's
also
a
little
bit
contradicted
right,
because
if
you
don't,
if
it
should
not
or
must
not
exhibit
HTTP
managues,
then
which
semantically
should
examine
like
what
it
is.
E
Well,
for
this
logical
span,
I
believe
they
still
might
want.
To
put
some
I
mean
it's
really
I
mean
it's.
It's
really
probably
really
hard
restriction,
if
you
want
to,
if
we
want
to
configure
it
in
this
particular
way,
I
think
it
should
be
some
something
simpler,
meaning
that
if
you,
if
you
want,
if
you
want
to
capture
all
this
with
and
if
it's
possible
to
capture
all
this
retries
and
redirects,
then
oh
sure
sure
go
for
it.
E
If
you
want
to
have,
if
you
want
to
like
put
even
more
spans
like
a
those
logical
spans,
okay
go
for
it
if
it
and-
and
basically
this
is
the
first
point-
and
the
second
point,
if
in
case
it
is
not
possible
to
capture
all
the
read
rice,
okay,
okay,
so
just
just
create
a
this
encompassing
one
and
you're
done.
F
D
F
D
F
It
shouldn't
because,
if
I
want
to
write
a
query,
give
me
all
the
HTTP
failures
right.
I
want
to
compare
apples
to
apples
if
I
compare
logical
operation,
failure
rate
with
our
retrace
versus
all
HTTP
request
failure
rate,
they
are
not
comparable
right,
so
we
we
need
a
way
to
prohibit
people
from
having
both
in
the
same
application
that
we
can
cannot
distinguish.