►
From YouTube: 2020-08-11 Spec SIG
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
B
B
D
E
E
F
F
E
Okay,
let's
go.
We
have
so
many
items.
So
thanks
for
joining
sorry
for
the
background
noise,
I
was
supposed
to
not
be
working
today,
but
you
know
some
changes.
Okay,
so,
first
of
all
daniel
dayla
tracy
state
api,
underspecified.
G
Yeah,
so
I
put
this
on
the
agenda
because
I
opened
this
issue
about
a
week
ago
and,
let's
see
a
couple
of
days
ago,
andrew
added
the
release
after
ga
label,
and
I
just
wanted
to
contest
that
label.
Essentially,
in
my
opinion,
this
cannot
be
done
after
ga,
as
it
will
require
breaking
changes.
G
So
at
the
very
least,
I
think
that's
something
in
the
spec.
You
know
we
we
could
remove
the
public
api
entirely,
which
would
free
us
to
define
it
later,
but
a
bunch
of
the
sigs
have
already
implemented
trace
state
differently
from
any
other.
You
know
on
their
own
because
there's
been
no
api
specified
for
it.
G
D
Makes
makes
sense
to
me,
I
don't
know
if
if
anyone
or
you
have
time
to
pick
it
up
and
and
come
up
with,
a
proposal
will
be
good,
otherwise
removing
is
not
necessarily
the
worst
option.
In
my
opinion,
I
know
I
know
there
are
some
some
parties
that
need
that
so
yeah
both
are
valid.
H
For
example,
I'm
wondering
if
we
there
is
a
pr
to
add
the
jager
propagator
to
the
required
propagators.
I'm
wondering
if
this
is
implementable
without
having
to
trace
it.
I
don't
know,
maybe
it's
no
problem,
but
maybe
it
is.
D
H
E
I
E
Let's
start,
let's
remove,
let's
put
it,
you
know
with:
let's
use
the
correct
label
for
this
issue.
D
E
H
Yeah,
the
next
question
is
about
the
after
ga
versus
before
ga
labor.
I
think
we
have
some
issues.
I
know
one
which
are
linked
which
are
low
priority,
so
we
wouldn't
really
care
if
we
get
them
in
before
release,
but
on
the
other
hand,
we
can't
really
get
them
after
release,
because
it
would
be
a
breaking
change.
So
I'm
wondering
what
we
should
do
with
such
issues,
maybe
like
just
close
them
put
before
release
but
priority
three
or
four
at
them.
D
I
think
if,
if
there
are
priority,
two
or
four
most
likely
in
my
opinion,
are
things
that
are
on
the
fence,
but
usually
should
not.
We
should
not
have
anything
that
is
breaking
change.
Maybe
I
will
take
another
look
today
at
all
the
the
things
that
are
marked
for
ga
and
see
if,
if
that's
the
case
and
raise
the
priority,
if,
if,
if
that's
a
breaking
change,.
D
Yeah
so
also
the
example.
The
five
to
five
that
you
gave
for
me
is
not
a
breaking
change,
because
most
likely,
the
solution
in
my
mind,
is
that
they
truncate
the
run
and
they
just
use
the
random
part
in
in
the
data
case.
H
I
didn't
mean
it's
possible
you're
wrong.
It's
it's
possible
that
it's
do
ever
without
breaking
change.
D
Oh,
no,
no,
no,
no,
no
yeah,
but
in
general
like
if
we
have
made
any
mistake
in
in
prioritizing
or
in
labeling
these
things,
it's
good
to
say
that
and
one
option
is
mark
them.
If
you
are
on
the
fence
and
stuff
mark
them
on
the
release,
change
the
label
to
be
released
required
for
the
ga
and
then
comment.
Why
and-
and
we
can
make
a
decision
that
hey
we're
gonna
solve
this
issue
in
this
way.
So
no
need
to
do
this
thing.
E
You
great
okay,
next
item
andrew
probably
we
should
have
done
the
inbox
for
the
issues
before
the
rest
of
the
items.
Sorry,
sorry
for
that.
A
A
A
Okay,
8
10
right
now
triage
nuclear
newly
opened
issues
in
pr
since
last
six
meeting.
So
I
have
a
scan
of
this.
It
looks
like
since
seven
days
ago,
one
two
three
four
five
they've
been
triaged
properly.
This
one
came
in
three
hours
ago.
A
All
the
other
ones
have
a
required
for
ga
or
not
or
after
ga
label.
So
this
one
just
needs
a
required
for
ga
or
after
ga
label.
I
A
Okay,
john,
would
you
be
able
to?
Would
you
be
the
appropriate
person
to
answer
the
question.
I
D
I
would
like
to
to
hear
other
opinions.
I
mean
I'm
not
against
this.
I
would
just
want
a
big
consistency
and
I
want
to
to
make
a
consistent
decision
across
all
the
sikhs
that
this
is
a
thing
that
we
want
to
support.
So
essentially
the
person
was
was
asking
us
hey.
You
should
split
at
least
the
sdk.
I
think
you
guys
convinced
that
api
should
not
be
split,
but
also
the
person
also
asked
for
that
between
different
signals.
So
I
would
like
to
hear
others
opinion
as
well
here.
D
I
D
I
Get
the
the
motivation
for
this
was
ken,
and
some
folks
who
have
been
working
on
micrometer
are
very
interested
in
open,
telemetry,
open,
telemetry's
tracing
implementation,
but
would
like
to
have
an
alternative
for
metrics
to
use.
Micrometer.
D
C
I
In
the
binary
yeah
they're,
that's
very
specific,
because
they're
they're
implementing
in
quarkus,
which
is
a
pre-compiled,
pre
nade,
it
generates
native
code,
and
they
want
to
make
it
as
lean
as
possible,
with
as
few
classes
as
possible.
That
aren't
needed.
B
G
Case
yeah
in
in
js
we
have
it
split
as
multiple
artifacts
package
size
is
our
major
limitation
there
as
we're
deploying
to
the
browser
anything
we
can
cut
out.
As
a
win,
I
see.
D
B
Should
do
it
will
probably
be
hard
for
different
languages
like
for
c
plus
plus?
There
is
no
such
issue
at
all
like
at
least
time.
The
link
will
do
the
optimization
and
for
download
that's
not
a
concern
and
for
javascript.
I
think
you
always
want
to
minimize
and
only
pay
for
what
you
need
for
the
web
scenario
right.
You
always
try
to
split
to
as
fine-grained
as
possible,
and
I
I
I
think,
probably
the
specs
should
describe
the
desire
and
each
language
will
have
their
own
flexibility.
D
Correct
correct,
I
mean
in
c
plus
class,
as
you
said,
you
you
made
different
buildable
units.
So
even
if
I
depend
on
this,
I
can
depend
on
only
the
buildable
units
that
I
need.
So
it's
still
a
step
towards
this
direction,
which
is
not
available
in
java.
But
I'm
asking
if,
if
this
is
an
agreement,
let's
have
a
statement
that
the
language
should
offer
in
the
language
specific
way
a
possibility
to
depend
only
on
the
the
parts
that
you
need.
That's
most
likely
what
we
need
correct.
B
Is
that
required
thing,
or
we
think
respect
shouldn't
mention
that
at
all,
so
the
spike
asked
for
something
that
means
it's
required.
Respect
never
mentioned
about
something.
That
means
it's
not
required.
Each
language
have
their
freedom
and
I
think
it's
impossible
for
the
spec
to
have
a
full
list
of
what
is
not
required.
D
But
the
problem
is,
the
problem
is:
if
the
supports
and
and
java
does
not
support,
people
will
come
and
say,
hey.
Why?
Why
js
supports
and
java
does
not
support,
and
I
don't
want
to
be
the
answer
that
the
maintainers
are
are
the
bad
guys
and
they
they
did
not
want
to
do
this.
I
want
to
be.
The
answer
is
okay.
If
the
here
is
what
the
spec
says
like
does
it
make
sense.
C
I
agree
with
really
I
don't
think
spec.
It's
a
spec
concern.
I
think
it's.
It
needs
to
be
a
decision
made
by
each
of
the
language
maintainers
based
on
the
requirements
of
that
particular
language.
It's
critically
important
for
javascript,
because
it's
bundled
and
it's
downloaded,
it's
a
browser
site
thing.
It
doesn't
really
matter
for
c
plus
plus
because
it
gets
optimized
out
by
linker.
C
A
I'm
running
up
against
my
time
box,
seven
minutes
so.
D
Andrew
it's
your
time,
let's,
let's
put
a
label
for
require
g
to
make
a
decision
of
of
what
do
we
want
to
do,
and
then
we
can
mark
it
as
after
ga.
If
there
is
nothing
that
we
do
or
something.
F
A
A
A
Five
is
still
closed
same
number
as
last
week
and
the
other
point
the
trace
issues
so
p1
trace
issues,
our
very
specific
concentration,
which
I'd
also
like
to
go
over
in
the
topic
following
this
is
the
follow-up
items
of
what
we
wanted
to
move
the
needle
on
on
the
p1
trace
issues.
A
A
D
D
E
D
D
I
would
I
would
review
reread
again
and
if
it's
trivial,
because
this
one
is
just
a
change
log,
I
would
add
an
entry
that
we
didn't
wait
for
48
hours
because
because
and
I
will
press
the
merge
but
again
I
I
think
I
don't
know
what
your
what's
your
motivation
with
this
going
through
these
things
right
now,
I
think
there
are
a
couple
of
them
very
old
that
we
need
to
discuss.
Maybe.
A
D
A
Okay,
sure,
then
we
wait
for
the
time
limit.
I'm
just
trying
to
make
sure
we
know
where
what
our
expectations
are.
So
we
can
push
right
up
against
expectation
in
order
to
make
sure
our
velocity
is
at
it's
optimal
to
keep
progress
forward,
but
I
can
really
wish
the
time
to
discuss
other
prs
as
well.
A
D
A
Okay,
all
right,
then,
I
think
that.
D
J
C
Okay
sure
there's
a
way
to
have
multi
repository
projects
in
the
product.
A
E
D
D
There
is
a
pr
open
that
there
are.
There
is
a
discussion
there.
I
think
you
you
can
look
at
the
seven
two.
One
is
the
pr.
I
think.
Yes,
that's
the
pr
that
we
we
have
the
discussion
so.
D
A
D
H
K
Yeah
I
mean
bogdan
again,
this
is
lita.
We
have
been
going
through
the
spec,
you
know
from
the
aws
team
and
you
know
again
kind
of
creating
a
gap
analysis,
but
we
were
going
to
report
that
in
the
sampling
sake,
but
it
would
be
nice
to
have
be
more
specific
for
sure.
D
E
Oh
yeah,
there's
a
pr
for
that,
and
actually
I
mentioned
it
in
the
agenda.
There's
no
review
so
far,
but
yeah
we
have
a
this
is
initial
uninitial,
so
it
will.
It
should
have
this
issue
fixed.
There
are
two
already
existing
issues.
So
if
you
are
into
open
tracing,
please
or
you
are
familiar
with
review.
D
E
D
E
F
A
Okay,
so
that
that
concludes
this
item.
That's
it.
D
Perfect
next
item
is.
H
Left
yeah:
this
is
about
the
pr
if
someone
could
share
the
screen
again
or
maybe
like
me
to
share
the
screen
yeah.
That
would
be
okay.
Helpful.
Thank
you
because
I'm
not
if
you
could
go
on
the
pr
yeah.
This
pr
suggests
two
things.
H
One
thing
is
is
inherited
from
an
older
pr
to
specify
that
we
should
record
an
exception
if
it
leaves
the
scope
of
a
span,
so
that
doesn't
say
that
we
should
not
record
exceptions
in
other
cases,
but
in
that
case
we
should
record
exceptions,
and
the
other
thing
it
does,
which
is
a
bit
controversial,
it
seems,
is
to
add
boolean
semantic
conventions
for
the
for
the
exception.
H
Events
that
says,
is
the
exception
escaped
or
not,
and
I
was
wondering
what
you
think
how
useful
such
an
attribute
is,
or
if
you
think
this
is
unneeded
yeah,
because
I
think
it
would
be
rather
helpful,
especially
if
you
have
multiple
exception.
Events
on
a
span
which,
which
you
can
have
then
usually
one
of
them
or
none
of
them-
will
be
an
exception
that
escapes
the
scope
of
the
span,
and
with
that
attribute,
you
would
be
able
to
pick
that
one
out
and
maybe
show
it
more
prominently.
D
H
Yeah
I
added
the
after
ga
label.
Actually
because
it's
I
don't
think
it's
that
critical
yeah,
but
it's
still
useful
and
I
think
it
would
be
easier,
a
bit
easier
to
implement
it.
If
we
do
it
with
the
whole
exception,
specifications
still
being
very
new
and
I
think
I'm
not
sure
about
the
status
of
other
six.
But
I
don't
think
many
six
have
implemented
record
exception.
D
D
So,
okay,
what
I'm
trying
to
say
is
please
everyone
say
that
I
am
wrong
that
we
really
need
this.
If,
if
that's
what
you
believe
and
then
we
will
spend
time
thinking
about
this,
if
people
are
like
me
that
believe
we
can
add
this
later
and
is
not
required.
Just
support
my
comment
and
then
we
can.
We
can
evaluate
again,
I'm
not
saying
too
close
or
to
to
to
shut
up
your
your,
your
pr
or
anything.
F
Say
so
my
opinion
would
be
if
we
want
to
edit,
then
now
would
be
the
best
time,
because
six
haven't
implemented
the
the
convenience
api
for
serializing,
the
stack
traces
and
so
on
yet,
and
so
it
would
be,
it
would
be
done
in
one
term
and
also
instrumentations
implementing.
It
would
be
correct
right
from
the
start,
because
it
will
be
less
likely
that
they
will
add
this
should
but
not
must
requirement
when
it
is
introduced
later
on.
So
if
we
have
it
right
from
the
start,
it
is,
it
is
way
more
reliable
and
yeah.
K
L
I
tend
to
agree
with
that
sentiment.
I
think
the
big
question
is
like
do
we
need
it
or
not,
and
I
know
we
have
been
having
in
errors
working
group
on
thursday
mornings
this.
It
might
be
a
good
place
to
continue
this
discussion.
I
would
I
would
strongly
encourage
and
come
up
with
with
the.
D
H
To
the
specs
so
yeah
and
I
think
it's
more
like
an
edge
case
where
you
can
market
and
in
the
common
case
you
you
don't
know
in
the
general
case,
you
don't
know
it,
but
that
edge
case
of
knowing
it
is
common
enough
that
it
makes
sense.
I
think
in
most
there
will
be
many
cases
where
this
attribute
is
absent,
because
you
just
don't
know
it,
but.
H
D
Yeah,
if,
if
that's
the
case,
I
don't
know,
I
don't
know
if
we
should
have
it
or
not,
but
maybe
the
working
group
can
come
up
with
a
good
good
reasoning.
Why?
Even
though
we
don't
have
it
everywhere,
it's
still
a
super
awesome
thing
to
have.
K
H
D
So
you
are,
you
are
saying
if,
if
I
have
a
span
that
covers
let's
say,
http
call
and
the
http
client
library
throw
an
exception
in
java,
for
example
yeah
I
don't
handle
that
exception.
I
I
catch
it,
but
I
throw
it
back.
I
continue
to
throw
do
I
record
or
not
the
exception.
That's
your
question.
H
Yeah
I
mean
my
question
is:
should
should
we
specify
that
this
should
be
recorded?
Yes,
I
think
it
should
be
recorded
yeah.
I
also
think.
D
H
F
H
Fine
example
in
python,
every
iteration
is
ended
by
a
stop
iteration
exception,
so
that
would
be
a
lot
of
noise.
If
we
call
we
recorded
all
of
them
exactly.
D
So
I
think
I
think
then,
then
I
don't
know
exactly
what
the
current
state
is,
but
I
think,
having
a
clear
focus
on
having
a
clear
documentation
of
when
we
should
record
exception
should
be
our
priority
before
we
add
extra
informations,
about
the
exception
at
least,
determine
when
to
record
the
exception.
That's
critical
for
ga,
then
then
we
based
on
these
whole
cases
we
can
determine
if
we
need
to
to
make
a
different
call
between
let's
say
having
a
bull
and
to
mark
some
of
them
or
not.
A
E
Yes,
andrew
no,
it's
carlos
the
next
level
yeah,
but
they
already
mentioned
this.
Oh
okay,.
D
J
A
Is
there's
okay
I'll
bring
it
up
again,
yeah?
Are
there
things.
D
I
I
did
that
I
I
put
labels
for
for
the
ones
that
I
triage
and
everything
for
require
4g
release
after
and
stuff.
I
I
did
the
specs,
the
ones
that
I
needed.
I
didn't
add
the
priorities
and
areas
for
the
moment,
because
areas
don't
apply
too
much.
D
K
D
G
D
And
the
even
the
the
166
is
not
marked
with
release.
Yes,
okay,
so
this
one.
If
I
read
it
correctly,
people
are
asking
for
supporting
just
bytes
as
one
of
the
type
like
battery.
D
D
D
Secondly,
I
want
to
say:
should
we
discuss
this
in
in
the
spec,
because
we
also
need
to
to
have
an
api
to
support
to
add
these
values,
not
only
the
way
to
transport
them.
C
D
D
I'm
going
to
move
it,
I'm
going
to
transfer
it
to
effects
and
change
the
labels
and
stuff.
Yes,.
A
B
C
So
this
was
not.
This
was
not
meant
to
be
a
benchmark
that
we
publish
for
external
people
to
have
a
look
and
see
how
the
protocol
performs.
This
was
more
about
having
a
tool
for
people
who
work
on
the
design
of
the
protocol
to
make
decisions.
C
For
example,
josh
was
recently
working
on
the
metrics
and
he
used
the
benchmarking,
my
personal
repository
to
do
the
benchmarking.
I
think
others
who
would
need
to
do
a
similar
work
with
benefit
hunters.
This
was
primarily
for
that
for
people
who
are
working
on
the
protocol
itself.
C
D
Okay
different:
how
hard
would
be
to
move
from
your
code
to
have
something
introduced
for
that.
C
D
As
you
said,
there
is
a
desire
we,
we
also
have
another
issue
to
related
to
performance
which
is
mark4ga.
So
if
you
have
time-
and
you
can
help
with
that-
that
would
be
great.
C
Okay,
I'll
have
a
look,
I'm
not
sure
about
the
prime
and
the
partly.
I
guess
you
just
can't
fork
in
and
use
the
fourth
version.
If
you
want
to
do
something
and
josh
did
that
when
he
needed.
Let
me
have
a
look
basically,
if
it's
easily
doable,
I
I
will
do
that.
Otherwise,
maybe
I
just
provide
some
instructions
for
others
to
use.
B
D
Oh
yeah,
the
the
current
spam
proto
definition
does
not
have
the
trace
flags
and
from
the
span
context
and
kevin
is
actually
asking
for
that.
I'm
fine
with
this
change
is
not
necessarily
required
for
ga
in
terms
of
proto
being
backwards
compatible.
We
can
add
it
later,
but
I
would
like
to
make
a
call
do
we
want
to
have
this
before
g
or
not.
C
K
K
D
K
D
So,
in
the
meantime,
I
would
put
require
for
g
just
to
make
sure
we
we
don't.
We
don't
make
the
release
without
at
least
making
a
decision.
Is
this
important
or
not?
And
if
the
decision
is
we
don't
care
about
this
for
ga
we're
just
gonna
flip
the
the
label,
but
to
make
sure
we
don't
forget
about
it.
I
put
a
really
required
for
ga
label.
K
K
D
Yes,
the
189
is
a
p1.
K
D
A
D
A
A
A
D
Okay,
I
added
the.