►
From YouTube: 2021-12-14 meeting
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).
A
A
Okay,
let's
start,
thank
you
everybody
for
joining.
Let
me
share
my
screen:
okay,
yeah.
A
Okay,
let's
go
to
the
agenda.
The
first
item
is
this
one:
it's
a
pr
that
tiran
posted
a
few
days
ago
and
it's
basically
about
defining
a
workflow
every
time
we
get
a
new
issue.
A
Please
define
some
timeline,
you
know
like
we
have
to
actually
answer
to
the
issues
and
the
actual
maintainers
will
have
to
assign
somebody
that
can
follow
up
at
the
issues.
Basically,
it
has
enough
approvals
and
we
will
be
merging
that
today.
A
Okay,
that's
from
my
side.
Nikita
semantic
convention
schema
extensions.
B
Yes,
that
that
question
was
born
from
recently
grant's
pull
request
about
semantic
language
version
stability
and
said
that,
like
pre-requirements,
for
declaring
a
semantic
version,
stable
or
kind
of
requirement
and
that
emitted
telemetry
have
to
have
schema
schema
assigned
to
do
that.
So
my
my
question
is
what
if
we
have,
for
example,
database
client,
it
emits
some
telemetry
part
of
that
telemetry
is
defined
by
the
database.
B
A
What
could
be
the
problem
about
actually
having
death
in
the
specification
in
the
main
specification,
if
we
define
a
language
specific
directory,
let's
say:
could
that
be
a
problem.
A
B
C
C
C
Yeah,
I
I
don't.
I
don't
know
about
that.
I
mean,
I
think
it's
just
I'm
thinking
about
it
from
the
customer's
perspective.
You
know
I
haven't
thought
through
this
nearly
as
much
as
you
have
it
sounds
like
nikita,
but
just
the
idea
that,
like
you
know,
you
have
a
user
and
they're
looking
for
instrumentation,
and
maybe
they
have
one
choice
that
works
better
than
what
we
provide
in
the
open
symmetry.
C
B
Yeah,
I
think,
the
pro
so
that
question
is
probably
easy
to
to
answer.
We.
If
you
have
a
telemetry
and
telemetry
has
assigned
a
schema,
then
you
have
to
like
comply
with
that
schema,
and
you
know
what
what?
What?
What
can
you
expect.
B
D
Yeah,
I
think
I
actually
think
tigran
had
this
as
an
extension
or
it's
kind
of
a
future
possibility
in
his
original
tap
like
this
concept
of
a
parent,
a
parent
schema
that
basically
inherits
you
can
inherit
from
this
scheme
and
add
other
attributes.
I
think
it's
possible
to
do
that
extension,
but
it's
not
there
yet.
So
I
think
that
sounds
okay.
B
D
E
Okay,
next
week
last
week,
we
discussed
this
idea
that
client-side
telemetry
needs
different
conventions
and
it
seemed
like
this
is
a
solution
for
that
sort
of
problem
as
well
like
right
now
we
have
one
semantic
convention
and
it
doesn't
fit
everything,
and
if
we
had
the
ability
to
compose
them
and
and
link
them
together,
maybe
a
solution
to
that
problem
as
well.
E
E
He
would
say
that
he
built
the
current
schema
support
to
do
exactly
and
narrowly
what
he
needed
it
to
do
and
that
he
always
intended
to
leave
room
to
do
this
sort
of
thing.
But
it's
going
to
require
research
and
investigation
to
make
a
good
multi-schema
composition
system.
I
guess
is
probably
what
he
was
thinking.
B
A
E
Probability,
sampling
hi.
I
just
wanted
to
say
quickly
what,
where
we
are
with
this,
as
you
may
know,
it's
a
fairly
complicated
and
large
pr.
I
was
looking
at
the
size
of
this
file.
I've
produced
and
it's
larger
than
the
existing
sdk
spec
by
50.
So
there's
a
question.
A
few
questions
have
come
up
like
it
is
very
detailed.
It
is
perhaps
too
detailed
and
hard
to
understand
so
one
one
piece
of
feedback:
I'm
getting
is
that
that
we
could
use
a
higher
level
introduction
to
the
whole
concept
of
sampling.
E
That
is
not
really
a
specification,
but
more
of
a
user's
introduction
to
the
idea.
So
that's
one
sort
of
feedback
I'm
getting
and
the
other
sort
of
feedback
I'm
getting
is
more
implicit,
but
it's
it's
sort
of
like
here.
We
are
with
one
proposal.
E
It's
been
standing
for
about
a
month,
or
so
too,
and
and
I'm
not
producing
a
different
proposal,
so
maybe
we've
forgotten
what
the
alternatives
are,
and
I
think
that
that
is
a
risk
here
is
that
we
have,
you
know,
done
a
lot
of
research
and
figured
out
how
to
do
this
particular
approach
to
sampling
and
we
excluded
a
few
other
approaches
to
sampling
along
the
way,
which
maybe
we've
forgotten.
E
Now
how
that
that
would
look,
so
I'm
afraid
that
those
of
us
who
are
technical
and
and
involved
in
the
very
fine
detail
of
this
specification
might
want
to
see
what
the
proposed,
what
the
alternates
look
like
at
this
point
as
a
reminder
to
to
help
us
decide
whether
this
is
what
we
want
or
not,
but
then
I
still
end
up
back
at
that
first
point:
this
is
a
specification,
not
a
user
guide
or
user
introduction,
and
if
you're
expecting
to
come
to
this
specification
and
understand.
E
Oh,
this
is
how
I'm
going
to
change
my
organization's
approach
to
sampling
in
the
coming
year.
This
is
not
a
solution
that
you've
been
waiting
for,
and
we
are
still
it's
still
ahead
of
us,
so
I
think
the
solution
that
most
of
us
are
waiting
for
is.
I
can
configure
my
sdk.
I
can
tell
it
which
spans
to
view
and
which
bands
not
to
view
which
spans
are
sampled
at
fifty
percent
and
which
spans
are
sampled
at
zero
percent,
which
bands
are
like.
Is
there
a
rate
limit?
Is
there
a
remote
source
for
this
configuration?
E
That's
what
people
are
asking
for
and
that's
what
people
are
going
to
get
when
we
actually
work
on
that
problem.
This
is
just
a
solution
for
counting
spans
that
have
been
sampled.
So
in
some
sense
it's
no
more
meant
for
you
to
understand
it
than
the
existing
samplers
which
are
built
in
and
have
much
less
said
about
them
and
specifications.
E
So
I'm
afraid
that
we're
changing
the
bar
for
specification
quality
on
the
fly
here
in
a
way
that's
slowing
us
down-
and
I
will
I
will
conclude
my
remarks
here
just
by
saying
I
think
we
could.
We
could
merge
this,
but
we
could
also
slow
down
and
remember
what
all
the
alternatives
are
and
write
user
guides
as
well.