►
From YouTube: 2021-01-12 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
B
D
D
E
E
B
How
do
you
take
a
star
and
pin
it
down?
I
don't
remember
the
lyrics
actually.
A
Shall
it's
six
minutes
past
the
hour?
Shall
I
begin
all
right
so
I'll
start
with
going
over
quick
status
of
the
ga
spec
burn
down,
which
is
just
a
summary
of
this
project
here
that
I'm
filtering
on
p1
issues,
we're
tracking
22
in
the
to
do
column,
six
have
associated
prs
and
we've
resolved
57
just
a
little
bit
of
movement.
From
last
week,
a
couple
of
prs
bunch
of
pr's
three
from
p1
priority
have
arrived
in
progress.
A
If
you
want
me
to
dive
into
anyone
any
of
them,
let
me
know
else
I'll
go
on
to
the
next
issue,
which
is
time
boxed
of
triaging
newly
opened
spec
otep's
protect
proto
issues
from
last
week.
I
was
looking
at
this
earlier
and
there's
not
too
many
like
five
or
six.
So
we'll
start
here.
C
I
think
it's
related
to
the
yeah
exactly
this
one
great.
Thank
you.
C
I
guess
I
gotta
read
this,
but
by
the
way
I
I
think
I
had
already
asked
you
this
question
in
private
that
whether
you
think
this
is
important
enough
to
be
defined
before
ga.
D
D
We
can't
declare
ourselves
stable
and,
to
a
certain
degree,
can't
really
declare
ga
until
we've
figured
this
out,
so
I
do
think
we
need
to
take
a
stab
at
it
at
least,
but
I
also
feel
like
this
relates
to
some
other
issues
in
general,
we're
seeing
around
metrics
that
I
don't
know
if
I
want
to
get
into
right
here,
but
there's
a
question
around
what
kind
of
telemetry
do
you
want
to
see
out
of
your
system
and
with
metrics
there's
a
question
of
like?
D
Do
you
auto-generate
that
from
traces
to
some
degree-
and
I
actually
think
it's
a
similar
question
potentially
to
how
do
we?
How
do
we
have
the
end
user
define
what
invariants
they
need
to
see
in
their
telemetry
make
sure
we
don't
violate
that
anyways.
That's
some
vague
stuff,
but
I
do
think
we're
going
to
hit
some
related
problems
as
we
dig
into
metrics
further
around.
How
do
you
define
this
kind
of
telemetry
schema?
G
I
think
I
don't
want
to
take
too
much
time
here.
I
believe
we
can
address
this
after
the
ga
in,
in
my
opinion,
what
we
need
to
do
now
is
make
a
decision
that
it's
okay
to
postpone
this
after
ge
and
then
the
details
we
can
discuss
later.
G
G
I
was
thinking
that
we
should
have
a
way
to
for
for
the
emitters,
to
declare
what
they
are
meeting
and
then
the
receivers
to
understand
all
the
emitters
emitted,
which
would
make
it
possible
for
the
emitters
to
change,
to
make
changes
which
would
no
longer
say
they
are
breaking.
They
are
not
breaking,
they
are
changes
which
can
be
handled
right.
So
in
a
way,
I
have
this
feeling
that
this
is
sufficiently
different
from
the
concept
of
an
aki.
G
That
means
to
approach
it
differently,
where
we
should
no
longer
think
of
it.
In
terms
of
this
is
an
api.
We
cannot
change
it
because
it
will
break
stuff.
Rather
it's
it's,
it's
data
that
has
schema
and
data
tends
to
evolve,
schemas
tend
to
change.
This
is
the
reality
and
particularly
for
telemetry.
It's
a
similar
reality
right
like
for
many
other
things
which
are
data.
D
I
I'm
seeing
the
need
for
something
that's
effectively
a
control
plane.
I
kind
of
predict
something
like
that
will
start
to
emerge
and
I
kind
of
see
maybe
all
of
this
telemetry
stability
work
is
maybe
related
to
that
yeah,
where
we
need
some
amount
of
metadata
and
some
kind
of
feedback
loop
occurring
here.
G
D
B
I
think
this
is
just
an
editorial
change.
I
think
in
the
in
the
attributes,
yep
exactly.
B
D
B
So
this
is
a
this
is
an
interesting.
I
know
the
background
on
this,
because
I've
been
working
with
honorable.
H
B
B
That's
required,
and
I
think
that's
really
the
core
of
the
issue
here,
and
I
don't
know
that
we
need
to
address
it
immediately,
but
I
think
it
is
something
that's
a
little
loosey-goosey
in
the
specs
and
it
would
be
good
to
have
a
better
understanding,
especially
for
maintainers,
about
how
to
make
these
decisions.
D
D
B
H
Yeah,
ted
ted.
I
think
it's
it's
exactly
what
john
said:
it's
we,
we
define
in
the
specs
that
there
has
to
be
a
process
or
a
spam
processor
that
implements
the
batching
logic.
Okay,.
E
H
D
H
D
I
mean
it
makes
sense
to
me,
especially
now
that
we're
looking
at
the
sdk
surface
area,
like
with
a
closer
lens
and
thinking
about
you,
know,
maintainability.
D
Definitely
anything
that
could
be
moved
internally
without
that
could
reduce
surface
area
without
reducing
functionality,
sounds
like
a
wind.
I
I
mean
in
this
case
there
is
this.
I
think
rather
awake,
paragraph
respect
that
says
you
should
be
able
to
filter
spends
and
each
spam
processor
should
be
the
start
of
a
pipeline.
There
is
this
wording,
I
don't
exactly
know
what
is
meant
for
it,
but
you
could
interpret
it
that
you
should
be
able
to
like
make
your
own
spam
processor.
That
then
calls
the
batch
spam
process
or
something
like
that,
and
this
possibility
would
be
removed
if
you
can't
instantiate
or
access
the
batch
spam.
I
D
H
D
H
Correct,
but
right
now
I
think
I
think
this
is
related
to
the
story
of
initialization,
for
everyone
to
understand
is
like
right
now
we
ask
users
to
think
about
the
fact
that
there
is
a
processor
you
need
to
initialize
one,
but
but
if
we
make
this
only
a
configuration,
the
the
whole
idea
is
okay.
H
Whenever
you
configure
the
entire
thing
you
see
there.
Okay,
I'm
doing
some
batching
here
are
the
properties
for
the
batching,
so
is
more
like
it
makes
it
makes
initialization
easier
to
think
for
user
and
not
have
to
understand
the
entire
stack.
B
Yeah-
and
this
was
that
was
exactly
the
impetus
that
onorag
had
when
you
put
in
a
pr
in
java
to
remove
it
and
I'm
like
wait,
hold
hold
the
phone.
We
need
to
talk
about
this
so
yeah.
How
much
does
the
user
need
to
know
about
the
internals
of
the
sdk
and
kind
of
how
these
pipelines
are
built?
If
all
they
really
want?
Is
they
want
to
hook
up
an
exporter
with
some
batching
in
it?
Yeah.
I
Yeah,
if
we
want
to
remove
the,
if
you
want
to
allow
removing
the
type,
I
think
it
would
need
to
be
done
before
j,
because
if
we
release
type,
then
we
can't
easily
remove
it.
D
I
think
it's
something
we
should
figure
out
sooner
rather
than
later.
It
sounds
like
maybe
there's
a
just
related
question
around
just
initialization
in
general,
and
how
can
we
simplify
that
for
people
but
we're
going
to
remove
stuff?
Let's
do
it
now.
D
So,
do
you
want
to
make
it
required
for
ga.
B
D
We
we
just
decided
that,
if
we're
going
to
remove
this,
we
need
to
do
it
sooner
rather
than
later.
We
need
to
verify
that
we
won't
regret
removing
this.
D
C
Okay,
I
have
a
question:
are
they
so?
I
remember.
I
remember
that
this
in
java
we
have
a
very
exotic
processor
wrote
that
one
disruptor,
I
think
it's
called.
Are
we
really
thinking
about
leaving
those
ones
out,
or
I
don't
know
it's
something
to
consider.
H
H
H
I
think
I
think
it's
it's
reasonable
to
say
that
before
we
need
a
decision
or
we
need
to
put
more
thoughts
into
this
before
g8,
so
it's
required
for
ga,
but
the
action
item
would
be
to
make
a
decision
if
this
is
good
or
bad,
and
if
this
is
good,
apply
the
changes
to
the
specs
otherwise
leave
it
as
it
is.
C
A
J
H
K
I
bet
if
you
decided
to
onorak,
he
would
write.
He
would
write
something
up
tonight,
but
but
we
need
to
explain
to
him.
H
What
exactly
we
expect
the
decision
to
be
made
and
what
are
the
pros,
cons
and
stuff.
L
H
A
B
Yeah
so
service
name
and
required
resource
attributes
bringing
it
up
yet
another
week
we
really
haven't
gotten
much
further
on
this.
We've
got
three
approvals
since
we
removed
all
the
approvals
and
had
people
re-read
and
reapprove.
B
I
don't
know
how
many
approvals
we
need,
but
I
think
this
is
a
big
deal
and
I
think
at
least
for
java.
I
feel
like
this
is
kind
of
the
last
significant
blocker
for
getting
a
1.0
release
candidate.
B
H
John
you,
you
talked
about
this
a
lot
if,
if
we
were
to
to
do
the
the
minimal
change
right
now,
that
will
allow
us
in
the
future
to
either
make
it
non-optional
or
make
it
optional.
What
would
that
be?
H
B
So,
to
reiterate
the
three
options
so
we
can
make,
we
can
make
sure
that
I
remember,
and
we
all
remember
what
we're
talking
about.
One
option
is
that
what
is
currently
written
in
the
spec,
which
is
to
say
every
resource,
must
have
a
service
name
and
that
there
must
be
a
fallback
defined
by
the
sdk
for
that
resource.
So
that
means,
if
I
create
the
default
resource
or
just
an
empty
resource,
even
it
will
come
along
with
the
default
service
name
when
attributes
are
pulled
from
it.
B
That's
option
one
which
is
currently
what's
written
in
the
spec
option.
Two
is,
I
think,
what
is
presented
in
this
pr,
which
is
to
say
that
the
there
that
the
sdk
will
provide
a
resource
that
is
that
has
the
default
set
on
it,
that
exporters
and
whatever
needs
resources
can
access
if
they
feel
like
they
need
it
and
option
three
is
to
make
it
so
that
the
sdk
will
assign
those
default
attributes
when
the
resource
is
assigned
to
one
of
the
signal
providers
like
the
tracer
provider
or
the
media
provider.
B
C
C
So
I
don't
have
a
preference
and
I
could
approve
anything
that
you
john,
as
whatever
you
approve,
because
I
know
that
you
spent
like
enough
time,
but
the
last
time
I
checked
there
were
some
unresolved
comments
and
I
think
that
the
problem
based
on
you
know,
given
the
fact
that
we
track
one
more
week,
is
that
people
are
just
not
sure
we
need
some
solution
like
you
know,
one
of
them
and
just
go
with
it.
D
I
mean
there's
a
lot
of
systems
that
need
this,
and
you
can
even
ask
a
question
of
what
does
it
even
mean
to
deploy
a
service
that
can't
be
grouped
or
identified.
D
C
C
C
I
think
that's
the
simplest.
So
in
theory
that
should
be
the
one
we
could
go
for.
B
B
Yes,
just
just
thinking
this
through-
is
this
something
we
need
to
figure
out
how
to
encode
in
this
in
the
aml,
because
we're
trying
to
generate
trying
to
generate
the
code
for
semantic
conventions
and
resource
attributes
or
how
like
how
much?
How
much
of
this
do
we
have
to
think
through
and
or
if
we
decide
to
make
that
choice.
At
this
point.
D
We
would
want
to
add
it
to
the
yaml.
I
would
think,
if
we're
adding
more
constants,
this
is
just
one
more
constant
right
that
it
has
to
get
auto-generated
yeah.
I
would
say
right
and
so
that's
step.
One
is
make
a
default
available
to
people
and
then
the
next
step
is
do
we
want
to
auto
insert
this
default
into
resources,
and
then
the
third
is:
do
we
wanna
halt
if
the
end
user
fails
to
override
the
default?
C
Yeah
on
the
robotics
proposal,
we
just
would
do
the
first
one
and
then
just
you
know,
allow
exporters
to
say
like
if
resource
doesn't
have
service.
Name
just
use
this
one.
So
we
don't
do
two
and
three,
which
is
like
we
go
for
that
and
specify
that
in
a
clear
fashion
in
specification.
That
should
be
enough.
H
Yeah
and
that's
my
goal
and
the
and
also
it
seems
that
we
can,
if
we
tell
people
that
hey
if
there
is
no
element
in
the
resource,
object
use
this
default.
Then,
even
if
we
add
that
default
to
the
resource,
the
behavior
will
be
the
same.
The
result
will
be
the
same,
even
though
user
do
a
an
extra
if
check,
because
you
will
always
get
the
default
value,
so
they
don't
get
a
null
or
not
present
in
the
object.
H
So
we
we
have
a
path
forward
even
to
add
it
to
the
resource
and
make
it
as
a
result
of
get
from
the
resource.
D
Josh
mentions
that
instance
id
we
have
service
id,
it's
the
equivalent.
The
prometheus
requirement
is
that
correct,
josh.
J
I
was
just
throwing
out
an
observation
in
the
chat
there
that
in
the
prometheus
environment,
you
have
two
mandatory
labels.
One
is
job,
one
is
instance,
and
you
have
to
have
labels
to
separate
all
your
data
streams
in
that
system,
so
there
must
be
one
that
is
unique.
D
D
I
think
we
would
do.
I
think
it
would
be
a
positive
change,
the
industry,
if
we
made
people
do
that
frankly,
but
to
me
it's
just
a
question
of:
do
we
complain
at
them
or
do
we
do
we
halt
if
they
don't
and
if
we
halt?
D
N
I
feel
like
this
is
complicated
somewhat
in
that
resource
resource
resolution
can
be
async
and
take
time.
You're,
like
you,
don't
know
the
resource
at
startup
like
a
number
of
these
values
that
you
end
up
that
end
up
in
your
resource,
come
from
an
http
request,
something
that
can
fail.
N
So
I
don't
know
of
all
the
options
that
are
on
the
table.
I
really
like
kind
of
the
default
resource.
You
just
start
off
with
something,
and
then
you
just
keep
merging
stuff
on
over
the
top
of
it,
and
you
will
always
end
up
with
a
value
you
it
might
get
progressively
better
but
yeah.
That's.
D
J
N
J
In
the
go
sdk,
we
have
a
like
a
get
resources
with
timeout
operation
that
needs
to
run
first,
and
I
don't
think
we've
truly
correctly
addressed
what
happens
in
case
of
a
hang
there,
but
the
the
the
structure
of
the
sdk
is
that
you
wait
for
that
information
before
you
keep
going
yeah.
H
At
least
also
also
the
instance
id
mat
is
not
the
the
vm
is
id
the
instance
id
that
george
refers
to
it's
it's
a
bit
different
is,
I
think,
it's
an
instance.
J
E
E
H
You're,
correct
you,
you
cannot
use
that.
So
that's
why
I
wanna.
I
know
prometheus
uses
in
society,
which
is
an
overloaded
term,
and
I
think
we
we
most
likely
what
we
call
service
instance
id
is
what
they
want
for
prometheus,
which
is
something
that
we
say
it's
globally:
random
in
the
scope
of
a
service
name.
D
I
think
we
think
I'm
required
personally,
that's
what
my
gut
is
telling
me
that,
how
how
do
you
make
them
required?
I
think
it
would
mean
halting
which
breaks
a
breaks,
a
core
principle
yeah.
D
H
D
Time
it's
just
a
question
of
like:
is
it
a
whole
bunch
of
things
showing
up
is
like
default
service
name
in
your
back
end
like?
Is
that
fine
and
then
you'll
just
go
like
oh,
I
need
to
change
this
and
likewise
with
instance,
id
we
just
auto
generate
it
like
that.
Maybe
that's
the
most
reasonable
middle
ground
here
that
at
least
ensures
the
data.
D
D
So
I
can't
remember
which
one
of
that
proposals
this
is,
but
that
would
be
the
proposal
where
we,
the
default
object,
gets
stuff
written
down
into
it
like
we
write
down
a
random
id
and
we
write
down
default
service
name
and
then
you
can
overwrite
those
things.
So
that's
actually
what's
currently
written
into
the
spec.
D
B
J
B
D
H
B
Present,
that
was,
that
is
option
number
two.
I
remember
which
option
that
is
anyway.
I
think
this
is
this
has
been
one
of
the
proposals
and
I
think
it's
also
fine.
Chris
christian,
I
think
I
don't
remember
what
the
I
think
that
was
one
of
the
phases
that
this
pr
went
through
and
the
one
that
I
approved
originally.
I
Yeah,
that
was
the
original
phase
of
this
pi.
F
Yep
I
mean.
I
You
mean
you,
you
would
need
to
have
some
additional
shards
between
the
metrics
resource
and
tracer
resumes
yeah,
but
it
should
be
doable.
I
guess.
I
But
anyway,
id
is
goes
beyond
that.
We
are
talking
just
about
so
his
name
is
great.
D
H
D
Yeah,
like
I
think,
that's
what
matt
was
talking
about
if
it
just
the
default,
the
first
resource
object
you
get
is
already
filled
in
and
then
you
can
merge
it
with
other
things
to
make
make
more
defaults
or
sorry
make
more
resources.
Then
it's,
then
it's
a
good
thing
and
it
seems
simple
right.
I
just
I'm
a
little
vague
on
whether
or
not
we
have
a
centralized
place
right
now
to
create
that
initial
default,
that
all
the
providers
can
grab.
H
D
J
I
J
H
D
H
Yeah,
there
is
only
one
exception,
I
think
which
says
if,
if
the
value
that
supposedly
wins
is
empty,
does
it
override
or
not
and
the
the
the
the
reason?
The
reason
why
and
I
think
that
that's
the
most
important
part,
because
the
reason
why
we
wanted
that
is,
if
you
have
two
ways
to
get
the
vm
essence
id
and
one
succeeds
and
the
other
one
fails.
You
want
to
doesn't
matter
which
one
is
the
order
you
want
to
to
end
up
with
the
one
that
succeeded.
D
I
don't
know,
isn't
that
how
merge
works
anyways
if
you
haven't
included
about,
I
guess
you're
saying
you've
included
the
key,
but
the
value
is
nil
or
something
equivalent.
N
Yeah,
I
think
you
can
avoid
that
special
case
just
by
not
adding
the
thing
to
the
resource,
if
if
it
did
fail,
if
it
was
empty
and
then
you
always
end
up
preserving
the
non-empty
value.
D
D
Yes,
okay!
Well,
I'm
feeling
good
about
this.
I'm
happy
to
review
it
in
writing
one
more
time
once
you
make
the
update
christian,
but
this
sounds
like
sounds
like
we're:
finding
a
solution
to
what's
been
a
pretty
big
blocker,
so
I'm
excited.
B
Just
as
a
as
a
corollary
to
all
this
discussion,
does
anyone
have
concerns
of
the
fact
that
we
don't
have
a
spec
for
the
jager
exporter
and
we're
trying
to
release
yes.
B
M
D
B
N
I'm
not
a
taker,
and
I
don't
know
jaeger
very
well,
but
the
one
thing
that
I
see
coming
up
is
that
jaeger
has
a
number
of
different
export
formats,
so
depending
on
which
version
of
jaeger
expert
you're
familiar
with
the
spec,
may
end
up
skewing
one
way
or
another.
But
I
know
there's
like
there's:
there's
a
udp
variation.
There's
there's
there's
one
that
uses
it
in
http
transport.
N
There
might
be
more
and
we're
just
kind
of
running
into
this.
With
some
of
the
there
are
some
environment
variable
specs
for
for
the
jager
exporter,
and
they
are
definitely
for
one
of
the
implementations
and
not
the
other
like
the
default
port.
It's
like
a
compact
thrift
over
udp
or
something
for
port.
D
H
That
that's
that's
another
good
suggestion
ted.
We
should
just
say
this
is
the
official
one
that
we
support.
We
just
don't
say
we
support
jagger,
because
jager
has
how
many
like
10
versions
or
whatever.
H
J
Would
presume
that
jager
has
given
a
correct
translation
between
its
ten
formats
so
that
we
should
just
choose
the
latest
and
greatest
of
their
formats
and
and
stay
there.
I
think
that's
the
proto
version,
but
I
don't
know,
I
believe,
that's
true
as
well.
It's
it's
very
close
to
the
open
tracing
data
model
in
theory,
if
they.
D
N
C
C
M
L
It
could
be
wrong,
but
I
think
that
in
js
we
just
use
the
official
client
library,
like
jaeger,
has
a
client
library
and
we
just
use
that-
is
that
not
acceptable.
B
D
H
F
A
We
still
have
a
few
other
items
on
the
agenda.
One
is
actually
just
one
last
item:
jonathan
ivanov
brought
up
something
related
in
comment
to
the
pr
we
were
just
discussing
about
service
name.
P
Yeah,
hey
I'm
jonathan,
I'm
quite
new
here
and
I
work
in
the
spring
team
at
vmware
and
I
had
just
a
few
questions.
This
is
one
of
them,
so
the
name
of
the
application
you
integrate
open,
telemetry
with
is
not
necessarily
a
service.
It
can
be
something
as,
for
example,
a
desktop
application.
P
D
J
Jonathan
to
your
other
question
about
metrics,
we,
I
have
a
draft
agenda
and
I've
shared
it
with
a
few
people
and
we're
still
developing
that
agenda
for
friday,
but
you
should
join
and
I
think
this
is
the
right
audience
to
start
sending
feedback.
There's
also
a
github
channel
for
metric
specs
and
a
thursday
meeting
dedicated
to
metrics,
where
we'd
love
to
talk
about
it.
P
B
J
A
We've
run
up
against
the
end
of
this
meeting,
but
I
can
carry
this
over
to
the
next
week,
where
we
have
an
outstanding
item
to
talk
about
open,
telemetry,
gta,
high
level
items.
This
is
a
tracking
project
to
track
high-level,
ga
items,
so
we
make
sure
we
have
progress
outside
of
the
specification
repo.
We
just
happen
to
be
tracking
in
the
specification
repo,
but
this
spans
across
all
the
repos
in
open
telemetry,
that's
required
for
ga.