►
From YouTube: 2021-12-09 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
C
D
B
A
C
B
C
C
So
just
two
things
on
the
agenda:
I'm
guessing
the
there
might
be
a
couple
more
but
we'll
get
started
with
josh's.
I
replied
just
before
this
in
the
the
slack
channel
that
there
is
no
retry
right
now
and
as
far
as
I
recall,
grpc,
it's
an
extension
like
similar
to
load,
balancing
and
a
few
others,
but
it's
not
part
of
the
the
spec
spec.
C
So
the
client
doesn't
do
it
automatically
the
client's.
Actually,
my
library
as
well.
So
I
can
add
that
if
the
extension
is
is
worth
it
in
the,
but
I
did
plan
to
support
the
retry
based
on
the
return
of
the
the
failure
from
the
otlp
protocol
and
but
I've
been
putting
that
off
as
post
ga,
because
I
didn't
think
it
was
a
blocker
for
1.0.
B
Right
so
there's
only
two
things
I
want
to
call
out.
First
of
all,
the
spirit
of
the
spec
was
that
if
you
just
use
the
grpc
retry,
that's
a
bit
that
should
be
available,
and
if
it's
not
that's
what
I
was
kind
of
curious
about
that,
you
should
just
be
able
to
use
that
and
you're
done
like
like
it's,
not
it's
not.
B
It
wasn't
meant
to
be
onerous
for
the
grpc
aspect
for
http
you
might
you
know
you
might
have
to
implement
something
whatever,
if
you
can
do
it
in
a
non-breaking
way,
you're
fine
for
1.0.
But
if
you
think
that
it
could
break
your
interface
in
some
way
or
users,
then
it
would
be
blocking.
B
But,
like
that's
that's
up
to
you
like,
I
don't
I
don't
know
if
it's
breaking
or
not,
if
it's
not
gravy
same
with,
actually
the
attribute
limit
count,
I
think,
if
it's
not
breaking
your
1.0,
you
could
release
a
1.0
without
it,
but
I
would
recommend
against
it
depending
on
how
much
instrumentation
you
expect,
just
because
that's
like
the
only
mechanism
that
we
offer
in
open
telemetry
to
limit
things
in
configuration
like
that's
it.
B
So
yeah
that
one
has
been
more
nervous,
but
the
the
grpc
one
I
was
just
kind
of
like
this-
should
have
been
free
for
you
and
if
it
wasn't,
I
kind
of
want
to
understand
why
you
know
like.
Is
it
hard
to
get
that
extension?
Was
it
like
a
pain
in
the
butt
that
sort
of
thing.
C
Right,
the
second,
the
first
one
I
find,
is
only
connection
back
off,
not
retry,
but
I
swear.
There's
a
retry
spec
that
I've
just
been
putting
off
implementing.
B
Yeah
for
reference,
there
was
a
discussion
in
the
java
seg
around
retry
as
well,
and
the
clarification
yeah
the
clarification
that,
because
someone
was
implementing
retry
on
top
of
grpc
and
the.
B
From
tigran
was
basically
like,
you
should
just
use
what
your
pc
provides
out
of
the
box
and
if
it
doesn't
have
something
that's
on
jrpc
for
the
grpc
endpoint
for
hdp,
it's
it
you
know,
then
then
it's
a
little
different.
C
C
B
C
C
Yeah,
okay,
yeah,
there's
nothing
there,
but
yeah
so
well.
That
was
good
to
know
because
that
means
I
will
implement
it
in
grpc
box,
our
our
grpc
library,
instead
of
directly
in
the
directly
in
an
open,
telemetry
exporter
and
that'll,
be
good
for
everyone,
because
other
people
use
that
library
as
well-
and
I
have
a
nice
document
proposal
to
work
off
of
for
that.
C
I
guess
I
should
make
sure
this
was
actually
the
accepted
proposal,
but
yeah
cool
and
for
http.
It's
yeah
it'll
go
in
the
exporter,
and
it's
just
it
shouldn't
be
that
I
mean
I
didn't
think
it
would
be
too
much
trouble
to
do
for
either
http
the
same.
B
Yeah,
I
totally
understand,
and
the
the
other
thing
is
like
the
in
terms
of
the
exporter
I
couldn't
find
documentation
outside
of
just
reading
it
like
what's
supported,
what's
not
supported.
I
don't
know
if
you're
planning
to
to
write
that
as
part
of
ga.
It's
not
like
a
it's
like
like
one
of
my
goals
here
with
doing
all
the
eval
is
just
to
make
sure
documentation
lines
up
with
expectation
from
other
languages
and
then
expectation
of
like
what's
actually
in
the
code.
B
So
I
think
there's
I
don't
know
if
there's
missing
documentation
or
can't
find
it,
but
that
would
be
my
recommendation
too.
When
it
comes
to
that
exporter.
For,
like
the
spec,
you
only
have
to
support
one
of
the
three
things
you're
doing
as
long
as
it
goes
to
a
local
collector
and
has
a
retry
you're
gravy
right
from
that
standpoint.
B
But
you
know
it
you,
you
went
beyond
great,
but
when
it
comes
to
the
spec
matrix,
we
we
don't
actually
have
a
way
to
clarify
which
one
you
chose
to
be
like
your
primary
use
case.
And
so,
if
you
document
it
and
say
you
know
what
this
grpc
one's
our
primary
use
case,
you
can
go
mark
your
spec
compliance,
matrix
plus
plus
for
all
the
exporter
stuff.
As
long
as
it's
clear
to
users,
what's
what's
supported,
what's
stable
and
like
what
features
are
there?
Okay,.
C
Yeah
right
now,
it's
just
in
the
exporter
read
me
that
the
default
is
http.
I
planned
on
updating
the
website
docs,
because
right
now
it
just
uses
the
exporter.
That's
to
standard
out,
and
people
have
been
having
trouble
and
asking
questions
in
the
our.
B
C
So
yeah,
that's
the
question
is
so
the
primary
user
documentation,
I
guess,
will
be
the
so
hex
pm.
Let
me
I'll
show
you
and
I
should
I'll
link
to
I'll
need
to
there
yeah
there's
badges.
I
should
add.
I
guess
to
the
top
level,
read
me
because
so
these
read
me
go
into
the
package
manager.
B
C
Those
will
ultimately
be,
I
guess,
the
what
would
be
considered
the
the
main
documentation
because
it'll
be
what's
in
the
package
manager
the
website
and
then
the
website.
I
guess
I'm
not
exactly
sure
what
do
you
know
what
other
sigs
are
doing
mainly
for
that,
because
I
was
going
with
kind
of
some
broad
examples
on
the
the
main
website
and
to
help
people
get
started
and
then
link
them
out
to
the
name,
documentation.
B
It's
it's
a
bit
of
a
mass
chaos
situation
there
so,
like
here's.
Fundamentally,
your
language
ecosystem
has
expected
documentation
and,
like
expected
ways
that
people
look
for
documentation.
B
The
website
should
link
to
there
as
fast
as
it
possibly
can,
and
if
there
is
no
expectation,
then
use
the
website.
That's
right
so,
like
I
think,
there's
a
lot
of
documentation
like
for
c
plus
plus.
I
know
that
java's
moving
a
lot
of
their
documentation
into
it,
but
they
also
link
out
for
javadocs
python.
It's
it's
that
it's
not
read
to
me.
What
is
it
like,
or
whatever.
B
Interesting
that
that
is
what
they
link
to
and
that's
what
they
maintain
as
their
set
of
docs.
It's
do
you
remember
what
the
name
of
that
site
is.
I
always
forget.
I
was
call
it
read
to
me,
but
that's
wrong.
It's
like.
B
Oh
right,
right
c,
plus
plus,
did
move
to
that
okay
right
because
and
they
did
get
oxygen
working
but
like
whatever
your
ecosystem
leverages.
That's
that's
kind
of
where
you
want
to
push
to
it.
Yeah.
B
And
the
other
thing
that
your
your
read
me
in
at
your
top
level,
repo
just
linking
into
docs
from
there
as
well
too.
You
know
seo
is
a
thing
that
is
actually
super
critical
for
open
source,
but
I
would
arg
when
I
say
seo,
I
mean
stack
overflow,
optimization,
but
that's
different
anyway.
A
I
searched
your
name
and
I
was
like
he's
done
that.
B
Yeah
yeah
yeah,
I
used
to
be
a
scala.
What
evangelist,
I
guess,
would
be
the
appropriate
term.
There
yeah.
B
Yeah
before
it
was
light
bend,
I
actually
left
around
the
time
it
renamed.
So
I
was
yeah
anyway
yeah.
I
did
a
lot
of
the
tooling
stuff
and
we
were
working
on
observability
around
the
time
I
left.
So
it
comes
full
circle.
You
know.
A
A
C
I
was
hoping
he
would
be
here
so
yeah,
okay,
that
sounds
good.
The.
C
Okay,
so
is
it
mainly
for
so
we're
going
to
get
the
spam
limits
done
because
yeah
one?
I
agree,
probably
that
the
the
need
to
be
able
to
limit
somehow
so
that
they'll
just
blow
up
the
system
with
some
ottoman,
auto
instrumentation
and
just
we've
had
the
time
and
it's
nearly
there
so
spam
limits
I
want
to
get
in
and
then
the
is
it
pretty
much
sounds
like
documentation,
clarification
on
the
exporter
stuff
and
we're
pretty
much
good.
B
Yeah
yeah,
the
the
yeah
everything
everything
I
found
was
basically
little
knits.
So
it's
I
mean
it's.
It's
in
good
shape
the
the
documentation
bit,
though,
just
make
sure
that
everything
is
super
clear,
what's
stable
and
what's
not
that's
like
our
big,
our
big
effort,
everybody
wants
to
adopt
this
and
they're
asking
when's.
It
done,
when's,
it
ready
and
so
being
very.
B
We
still
need
room
to
experiment
because
there's
still
thing
like,
especially
in
the
instrumentation
space
right
like
knowing
knowing
which
spans
will
remain
and
which
ones
we're
planning
to
change,
as
we
experiment
with
how
to
instrument
using
this
stuff,
making
sure
that
that's
clear
to
users
is
like
the
biggest
part
of
this.
Just
you
know
setting
up
clear
user
expectations.
C
Yeah,
that's
a
actually
yeah.
That
brings
up
a
good
point
with
the
the
instrumentation
libraries
the
way
of
the
open,
telemetry
airline
contrib
repo.
Where
we're
keeping
those
is
there,
I
mean
I'd
assume
it
would
be
a
pretty
large
undertaking,
because
I
mean
I
know
java.
Probably
it's
like
hundreds,
the
the
for
going
for
marking
any
of
those
is
1.0.
Is
there
a
process
or
do
we
just
use
our
best
judgment
to
say
we're
not
going
to
be
changing
these
and
it
follows
the
semantic?
B
Yeah,
there's
a
there's,
an
active
pr
right
now
around
semantic
convention
stability
and
how
we
define
it
and
how
we
define.
So
we
don't
want
to
mark
things
as
stable
until
we
can
change
them.
That's
ironic
right,
but
that's
that's
that
we
want
to
be
able
to
change
it
safely,
and
so
we
don't
want
to
just
lock
in
and
never
change
and
basically
be
the
java
api
from
actual
java
itself.
You
know
for
like
the
next
20
years.
B
Ever
changes,
but
that
said
like
we
do
need
to
provide
stability,
so
so
there's
a
there's
a
pr
out
right
now.
I
highly
recommend
reading
it
from
tigran,
but
in
terms
of
marking
any
of
those
libraries
as
stable.
What
the
there's
different
there's
different
ways
to
do
this,
so
the
collector,
for
example,
right
marks
itself
is
stable
and
it
has
a
bunch
of
things
that
receive
and
interact
with
telemetry,
and
when
you
build
a
distribution
of
the
collector,
you
have
components
which
are
not
stable.
B
So
what
the
hell
does
this
mean?
This
means
the
configuration
of
that
giant
component
is
stable
and
the
pieces
of
it
that
are
stable
users
can
rely
on,
but
they
need
to
go
do
investigation
of
which
pieces
are
unstable.
So
if
you
provide
like
an
instrumentation
distribution
there
there's
a
lot
like
in
java.
Their
agent
is
marked
to
stable,
but
the
specific
instrumentation
is
clearly
called
out.
B
What's
stable
and
what's
not
effectively
all
of
it's
unstable
right
now,
but
the
the
configuration
for
the
agent
itself
will
not
break
kind
of
a
thing
right.
It's
like
this
weird
dichotomy
that
we're
doing
that's.
That's!
That's
really
meant
to
be
an
interim
thing.
I
I
I'm
expecting
http
stability
to
you
should
be
able,
in
q1
to
start
marking
things
as
stable
around
http
instrumentation.
B
If
you
need
to
make
a
distribution
that
includes
both
stable
pieces
and
unstable
pieces,
because
your
users
want
that
it's
just
the
the
thing
that
we
want
to
provide
is
clarity
of
hey.
These
traces
will
never
change
and
are
stable.
These
chase
traces
we're
experimenting
with,
and
please
give
us
feedback,
but
we
might
move
them
around
to
make
sure
they're
more
useful
for
you
right,
and
we
just
want
to
make
sure
that's
super
clear
to
people
that
there's
a
anyway
there's
a
long
discussion
on
that
right
now,
so
you
can
join
in
it's.
B
I
can
send
you
the
pr.
Let
me
find
it.
C
Yeah
I
found
it,
I
got
it
open,
I
can
link
to
it
in
the
chat
and
the
our
document.
Yeah
yeah.
I
know
that
that's
going
on
at
splunk
as
well
cause
so
many
yeah
pieces
are
some
is
1.0
something's,
not
I'm
working
on
the
ruby
distro
for
splunk
and
they're
1.0
with
the
tracing,
but
a
bunch
of
other
stuff
isn't
1.0
like
the
exporter.
So
what
are
we
yeah
we're
building
a
distro?
We
can't
call
the
distroness.
B
A
C
Brian's,
not
here
I
was
hoping
he
would
be
here
for
this
next
item,
because
I
just
wanted
I'm
opening
a
pr,
hopefully
soon.
I
thought
I
would
have
it
by
now,
but
a
bunch
of
stuff
happened
all
morning
for
kind
of
changing
up
some
stuff
with
how
our
internally
code
for
configuration,
so
it's
nothing,
external,
so
user
or
anything
so
changed
there.
But
after
doing
the
span
limits
set
up,
I
needed
to
make
some
changes
there,
opening
the
pr
so
hoping
to
get
eyes
on
it
early.
C
So
I
wanted
to
bring
it
up
here,
but
brian's
not
here
so
we
can
move
on
to
is
there
isaac?
Is
there
anything?
Oh
wait
assuming
josh
that
was
it
right
there
isaac?
Is
there
anything
you
wanted
to
discuss
questions
you
had
about.
A
Mostly
just
come
to
listen,
but
I
I
will
say
tracing
you
know
as
a
as
a
feel-good
story.
Racing
continues
to
be
making
moves
for
us.
So,
like
you
know,
I
was
talking
to
someone
in
charge
like
in
charge
of
a
lot
of
group
groups
inside
of
our
firm
and
they
seem
to
be
wanting
to
get
tracing
onto
so
evangelizing.
In
my
own
small
way,.
A
I
think
there's
a
bunch.
Our
team
is
like
maybe
there's
two
teams
that
use
the
one
I'm
here
for
the
sig
of,
but
the
rest
of
it's
just
whoever,
like
whatever
it's.
I
can't
divulge
too
much.
That's
what
I'm
saying.
A
B
C
A
C
A
B
So
so
the
reality
of
open
source
is
the
the
the
higher
you
are
up
in
the
project
in
like
a
thing
that
like
describes
people
who
do
it
the
lower
down,
you
are
in
the
totem
pole
of
how
much
work
and
grunt
work.
You
have
to
do
just
that.
C
Nothing
else
and
nothing
got
any
messages
from
gregor
brian's.
I
guess
greg's,
not
maybe
you
lost
internet
call
it
a
day
and
save
30
get
30
minutes
back
sound.
B
C
B
Right
yeah,
let
me
know
if
you
need
anything
else
for
me
for
your
1.0
and
it
was
a
pleasure
reading
some
airline
code.
It
was
it
was
a
nice
refresher,
like
you
know
what
I
mean
walk.