►
From YouTube: 2021-10-21 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
C
B
C
B
Some
people
show
up
five
ten
minutes
late,
trying
to
check
if
anybody's
gonna
be
making
it
today.
But
if
not,
then
we
can
just
discuss
the
ga
stuff.
D
C
Awesome,
I'm
going
to
give
you
a
caveat
that
I
read
joe
armstrong's
phd
and
I
may
have
written
a
little
bit
of
erlang
for
fun.
Nice.
C
And
that's
about
it,
but
I
I
used
to
work
with.
I
was
with
like
bend
or
type
safe,
did
a
bunch
of
akka
only
with
the
model
yeah.
B
Do
you
have
a
volume
on
your
mic
because
you're
pretty
quiet,
I
don't
know
if
it's
on
my
end
or
not?
Oh,
you
know
what
I'll
keep
raising
the
volume
and
it
doesn't
help.
C
B
Yeah
I
mean,
I
guess
we
couldn't.
I
haven't
heard
anything
back
in
any
of
the
channels
yet
so
people
might
not
have
anything
to
discuss
today.
So
we
can
get
started
with
ga
discussion.
C
Yeah,
so
basically
the
way
the
way
I
did
this
for
c,
plus
plus
and
what
I'm
gonna
do
is,
is
go
through
and
go
through
the
spec
compliance
matrix,
go
through
your
code
and
just
see
if
things
line
up
when
it
comes
down
to
like,
what's
the
best
way
to
express
something
in
erlang,
I'm
not
going
to
ask
questions
there,
mostly.
What
I'm
going
to
be
asking
is,
if
I'm
someone
new
and
I
come
in,
do
I
know
what's
stable?
C
C
So
if
like,
for
example,
you
you
don't
have
the
environment
variables
set
up
right
as
long
as
it's
clear
to
users
that
that
doesn't
work
or
isn't
needed
and
it's
clear
how
they
use
otlp
beautiful
anywhere,
where
it
like
you
know,
changes
the
name
from
the
spec.
If
it's
matching
the
convention,
not
gonna
care,
if
it's
like,
I
don't
know
how
to
consume
the
stable
pieces
of
the
language,
then
I'll
make
comments
right,
but
that's
the
the
bit.
C
The
only
thing
I'm
I'm
really
looking
at
here
is:
if
I'm
a
user
coming
in
and
I
use
erlang.
Does
it
work
the
way
I
expect
from
other
languages
right
and
does
it
like?
Is
it
clear
to
me
what
I
can
depend
on
what
I
shouldn't
depend
on
or
what's
like
in
progress?
That's
all
right,
yeah,
so
for
c
plus
plus,
I
remember
they
had
kind
of
metrics
and
traces
kind
of
entangled
a
little
bit
initially.
C
So
we
just
we
didn't
actually
change
any
of
the
code.
We
just
made
it
clear
in
like
header
files
and
directory
structure
like
what
was
actually
public
and
what
wasn't?
Oh,
okay,
that
sort
of
thing:
yeah
yeah,
that's
that's
about
it!
I'm
not
I'm
not
really
worried
about
any
any
implementation
details
unless
you
have
concerns
like.
Is
there
anything
you're
worried
about
with
going
1.0
or
you
wanted
extra
someone
to
take
a
look
at
you
know.
B
I
think
it's
mostly
been
hammered
out.
The
only
thing
that
I've
kind
of
been
kicking
around
recently
is
the
we
don't
actually
have
a
simple
span
processor,
because
I.
B
I
didn't
see
a
way
of
it
actually
working
and
instead
use
the
batch
processor
configured
to
basically
right
away,
send
stuff,
but
the
simple
span,
processor,
and
how
processors
are
set
up.
I
don't
know,
was
just
a
a
general
concern
in
there.
I've
had,
I
guess,
does
it
actually
make
sense.
B
Oh,
does
it
say
something
like
that
I
was
reading.
I
just
remember
it
said
that
it
had
to
send
it
immediately
or
something,
but
I
mean
it.
C
A
processor
that
which
passes
finished
spans
and
pass
export
friendly
and
span
data
representation
to
the
configured
spanx
board
as
soon
as
they're
finished
yeah.
Okay,
if
you,
if
you
have
a
batching
span
processor,
that
just
doesn't
batch
and
feeds
on
like
that's
totally
fine
too.
C
If
I
recall
correctly,
there
were
two
use
cases
for
simple
right.
One
is
like,
if
you
run
in
a
like
a
lambda
environment
and
you
work
fine,
you
know
then
cool.
That's
one
of
the
reasons
that
we
have
simple.
B
C
I
think
I
I
know
too
many.
I
don't
remember
what
details
I
can
talk
about
for
google,
but
I
can
tell
you
that
it's
a
very
real
thing.
We've
talked
about
a
lot
yeah.
I
think
I
think
as
long
as
you're
sending
it
immediately.
That's
that's
all
you
really
need.
So
if
you
like,
it's
simple
actually
required.
Let
me
take
a
look.
It's
it's
supposed
to
be
built
in.
Is
it
that
you're
not
able
to
just
name
it
simple?
Even
it's
just
a
bash
processor
that
has
a
cue
size
of
like
zero.
B
Right
yeah,
so
I
was
considering
just
making
a
configuration
option
named
simple
that
configures
the
batch
processor
for
the
user
in
a
specific
way.
That's
what
what
I'll
probably
do
just
to
have
it
there,
and
it's
kind
of
I
mean
it's
useful
for
tests
right
now
in
the
testing.
We
set
up
the
patch
processor
this
way
each
time
manually,
so
it
might
be
simpler
to
just
call
it
a
simple
processor,
but
it
yeah
it
just
made.
C
B
Lot
of
extra
work
that
seemed
like
it
was
going
to
cause
like
if
anybody
accidentally
used
it.
It
was
going
to
cause
a
lot
of
problems
unless
they
were
in
this
specific
environment,
which
now.
C
The
other
use
case
that
I
forgot
about
was,
if
you're,
using
the
collector
as
your
batcher.
B
C
Yeah
you're
running
it
as
an
agent
right
next
to
you
and
then
you're
doing
say
like
tail
sampling
in
the
collector,
where
the
collector
is
connected
to
some
control
plane.
Now
again,
this
is
in
the
future
right.
Not
now,
because
right
now,
you
can
do
head
sampling
in
the
collector.
I
think
there's
a
tail
sampling
processor,
but
it's
not
quite
the
same,
but
that's
the
other
use
case
that
I
forgot
about.
C
There's
like
the
cloud
run,
use
case
and
the
the
function
use
case
right,
aws
lambda,
then
there's
the
local
collector
agent,
where
you
just
want
to
feed
immediately,
get
it
get
it
out
of
process
right
away
and
that
way,
if
you
crash,
you
know
you're
not
holding
on
to
anything,
although
erlang's,
better
with
crashing
than
most
languages,
but
that's
a
different
story.
B
Yeah,
the
but
the
so
there's
there's
if
there's
still
an
exporter,
that's
sending
to
something,
even
if
it's
a
local
but
less
of
a
problem,
if
it's
local,
but
it's
still
essentially
queuing
these
up
to
send
and
it's
sort
of
moving
the
batchness
the
exporter
in
our
case
process,
instead
of
the
batching
being
set
up
by
the
batch
processor,
so
it
kind
of
makes
for
worse
batching.
In
our
case,
I
guess
because
you're
still
cueing
these
to
send
you're
just
sending
them
one
at
a
time,
essentially.
C
Actually
yeah
yeah
wow.
B
Also,
partially,
why
I
was
concerned
about
it
because
then,
if
we
block
while
sending,
then
you
have
to
worry
about
timeouts
because
now
and
span
could
time
out
and
before
when
you're
using
the
batch
processor.
It
doesn't
time
out,
and
that
creates
confusing,
because
the
user
can
now
be
crashing
with
a
timeout
when
they
end
the
span.
C
C
Okay,
there's
there's
a
lot
of
fun
that
can
happen
there.
Yeah
yeah,
the
the
the
cool
thing
is
context,
then,
is
valid
in
the
exporter.
The
not
cool
thing
is
context
is
valid
in
the
exporter.
Okay,
so
I
I
I
understand
like
to
be
spec
compliant.
I
I
would
recommend
implementing
the
simple
processor,
the
the
thing,
the
thing
that
I
like,
let's,
let's
pie
in
the
sky,
a
magical
world
where
we
can
do
whatever
the
hell
we
want.
C
Where
I'm
just
writing
things
out
as
fast
as
they
come
in
right,
that's
that's
what
I'd
love
to
see
from
this
in
practice?
The
way
it's
spec'd
is
just
to
allow
that
use
case
and
no
one's
ever
to
implement
it.
Probably
so,
okay,
yeah
yeah,
I
don't
know
if
that
helps
or
hurts,
but
yeah
the
the.
If
you
just
make
your
batch
processor,
send
immediately
everything's
gravy
for
that
yeah,
okay,
cool.
B
Oh
except
yeah
I
mean
we
wouldn't
be
blocking
still
in
that
case
on
the
end,
so
the
spec.
B
B
Look
at
that
because
yeah
it
makes
more
sense
to
to
block
there
because
then
you
don't
accidentally
suspend
or
shut
down
before
the
request
is
finished,
I
mean,
after
the
request
is
finished,
but
you
haven't
seen
hey
brian,
so
we're
discussing
some
ga
code
review
stuff.
He
was
asking
sort
of
any
concerns.
I
have
about
the
how
we
cover
the
spec
and
one
on
top
of
my
mind,
was
the
simple
processor.
So
I
brought
that
up
because
I've
been
looking
at
that
sorry.
A
B
C
A
B
B
Not
going
to
it's
not
going
to
hurt
the
use
cases,
except
for
maybe
that
future
one
of
you
know
memp
file
where
you're
actually
doing
this
for
high
performance,
but
in
the
use
case
of
like
a
lambda
environment
where
you
might
be
suspended,
but
you
want
to
get
this
band
sent
when
you're
recording
before
your
request
ends
being
handled.
B
B
No
yeah,
that's
basically,
I
think
what
would
what
the
simple
processor
will
be.
B
Well,
no,
I
mean
it,
won't
it
it'll
only
be
quick.
If
there's
I
mean
it
would
only
be
swapping
them
a
lot
if
there's
continually
spans
being
ended,
which
is
only
a
potential
future
in
which
we
can
later
optimize
the
simple
span
processor,
if
there
ever
becomes
the
day
where
it's
not
just
used
for
cases
like
lambda.
C
You
force
flush
right.
C
B
Yeah
yeah,
I
mean
yeah,
we'll
get
we'll
have
this
covered.
It's
just
it'll
be
a
question
later,
potentially
of
making
it
more
efficient.
C
Yeah
yeah
and
I
gotcha
I
got
you
the
the
other
thing
I
want
to
say
about
the
lambda
use
case
is,
I
do
think
the
force
flush
is
the
better
way
to
do
that
in
general
of
use
a
batch
processor
and
then
force
flush.
It's
only
simple
span.
Processor
is
only
good
if
you
have
one
span.
C
Cool
any
other
concerns
you
have
with
the
spec
yeah.
A
I
haven't
gone
through
like
the
fine-tooth
comb
on
anything
else,
but
it
nothing's
like
jumping
out.
I
think
we
do
need
to
do
that.
That
step,
though,
probably
because
I
know
that
we
have
things
checked
off
in
the
compliance,
but
just
to
do
one
pass
of
verification
through
this
back
to
what
we
have.
A
B
No,
so
we
had
an
option
to
start
span
to
pass
a
sampler,
and
it
turns
out
that
was
from
the
0.1.0.
B
Hotel
spec
and
must
have
been
who
knows
how
long
ago
I
was
out
for
like
four
months
and
was
I
wasn't
even
going
to
expect
sig
meetings
and
stuff,
so
there's
stuff
like
that
that
got
dropped
that
now
finding
needs
to
be
removed.
Yeah.
I
found
that
out
yesterday
when
I
was
right.
I'm
writing
an
otep
for
per
tracer
configuration
and
I
included
a
line
about
that
and
everybody
was
like.
You
can't
do
that.
B
A
C
Yeah,
I
said
that's
my
own
library,
though,
do
you
support
instrumentation
library
yeah?
By
that
I
mean
schema
urls.
Sorry,
oh.
B
The
schemer
ue
url,
no,
we
have
a
pull
request
for
that
and
I
think
I
just
have
to
take
it
over
because
the
person
hasn't
been
responding
so
for
months.
C
C
If
you
wanted
to
target
pre
schema
url,
I
I'm
fine
with
that
as
well.
Here's
why
I
think
schema
urls
will
be
super
valuable,
but
there's
missing
specification
around
them.
Yeah.
B
C
B
C
C
Do
you
have
any
this
is
straight
up
api?
Do
you
have
existing
instrumentation
in
the
components
or
not
just
curious.
C
C
Yeah,
zip
and
jager
both
need
to
be
implemented.
That
was
one
thing
c,
plus
plus
added
kind
of
last
minute,
the
spec
on
it's
pretty
nice
and
easy,
but
if
it's
not
if
it
so
again,
if
there's
technologies
that
you
think
one
of
them
uses
thrift,
if
you
don't
have
that
available,
then
we
can
call
that
out
as
well.
B
Right,
that's
why
we're
not?
We
don't
have
jaeger
technically,
there's
thrift,
libraries
for
erlang,
but
nothing
we
I
want
to
use
and
jaeger
accepts
zipkin
so
like
when
I've
worked
places
that
we've
had
to
export
to
jaeger
from
erlang.
I've
just
used
the
zip
code,
zipkin
exporter,.
B
C
You
can
so
so
the
one
thing
I'm
gonna
call
out
just
as
a
word
of
warning,
your
exporter
for
otlp
right
now,
just
does
traces
right,
so
eventually,
metrics
and
logs
will
be
in
there
too.
C
Do
you
want
those
to
be
separate
modules
like?
How
do
you
want
to
deal
with
stability
and
experimentation
and
that
kind
of
stuff?
Just
it
I'm
not.
I
don't
I'm
not
trying
to
to
to
judge
what's
there
now.
I
just
I'm
asking
you
right
and
then
the
second
thing
is,
if
you
wanted
to,
have
it
be
a
separate
app
thing,
that's
what
a
lot
of
other
languages
have
done.
Is
they
actually
have
a
separate
kind
of
module
for
otlp,
a
separate
one
for
jaeger,
separate
one
for
zipkin
and
then
with
the
metric
spec?
C
They
have
a
separate
one
for
prometheus.
You
know
a
separate
one,
for
whatever
else
will
be
there
and
in
java
there's
there's
like
five
otlp
exporters,
there's
like
a
specific
one
for
http
json,
a
specific
one
for
http,
grpc
or
proto,
and
it's
it's
kind
of
nuts,
so
yeah
yeah,
whatever
you
do,
is
fine.
A
I
mean,
I
think
we
might
end
up
that
way.
I
mean,
like
my
concern,
really
is
we
have
just
because
we
have.
I
don't
want
to
like
make
anything.
I
don't
want
to
do
anything
that
will
like
make
the
iot
crowd
be
like.
We
can't
use
this
because
it's
adding
grpc
box
and
like
all
these
other
things
that
I
don't
want
to
run
on
my
iot
device,
yeah.
B
Well
then,
yeah,
that's
why
I
was
splitting
them
out
originally,
but
we
now
have
optional
application
support,
so
this
might
be
less
of
an
issue
in
otp
24,
so
consider
it
both
ways,
not
sure
which
direction
for
sure
we'll
go
yet,
but
something
we'll
have
to
discuss,
because
I
also
don't
know,
I
think
elixir
people
are
oh
wait.
B
Yeah
we
hey.
D
I
know
there
were
some
changes
to
the
meaning.
C
A
C
D
Lung
are
you
the
yeah,
yeah,
okay,
I'll
I'll
figure
it
out
with
jonah?
I
apologize
with
sorry
with
morgan.
Apologies
for
that.
Why
don't
you
continue
and
I'll
be
quiet.
B
Yeah,
I
kind
of
wanted
to
go
to
the
eb
pf1
yeah
anyway.
If
this,
if
we
ended
early
nice,
what
are
we
talking
about
yeah,
so
yeah?
It's
something
we
need
to
decide
on,
especially
with
metrics
to
consider
but
yeah.
A
I
think
that'd
be
the
answer.
Then
josh
is
like,
I
think,
we'll
keep
him
separate
for
now.
Okay,
because
we
can
always
build
up
one
of
those
larger
libraries
that
includes
like
everything.
C
B
C
The
other
thing
I'm
going
to
ask
for
help
with,
because
I
don't
know
is
compatibility
restrictions
of
erlang-
I
will
just
be
asking
you
hey
is
this?
Is
this
a
compatible
thing
to
do
like?
Is
this
going
to
break
in
the
future?
That's
that's
the
other
concern,
but
yeah.
A
Okay,
so
right
now
we
do
have
a
hard
cut
line
of
otp
21
plus,
because
we
make
a
lot
of
that.
One
had
a
lot
of
features
that
we
use,
but.
B
We
are
kind
of
taking
the
the
oldest
one.
That
ericsson
supports
is
the
way
I'd
like
to
to
move
forward
so
yeah.
I
think
that's
like
three
versions,
two
or
three
versions
and
major
versions.
B
D
I
pasted
the
zoom
link
in
in
the
chat.
Oh
so
again,
sorry
for
crashing
your
party.
C
I
might
have
to
go
to
that
one
then
too
yeah,
so
real,
quick,
then
I'll
say
around
the
support
bit.
You
pick
whatever
you
think
like
like
it's
it's
up
to
like
when
it
comes
to
knowing
the
erlang
space.
You
know
better
than
anyone
else.
What
is
the
most
popular
languages
are,
and
you
just
tell
us
all.
I'm
gonna
ask
is
that
it's
crystal
clear,
like
if
a
user
comes
to
the
erlang
documentation
right.
The
other
thing
is
that
I
want
to
understand.
C
I'm
going
to
do
a
quick
google
search
for
open,
telemetry,
erlang
and
see
what
the
top
hit
is
and
if
it's
not
your
docs,
then
I'll.
Ask
you
some
questions
and
see
if
we
can
write
blogs
or
something
to
fix
that
right,
because
we
might
need
to
do
seo,
I
hate
doing
seo,
but
we
might
have
to
do
some
of
that
too.
E
C
I
look
forward
to
this.
I
am
way
the
hell
overloaded
this
week,
but
I'll
try
to
get
you
a
so
what
I
did
before.
If
you
want
to
see
a
bug,
I
can
send
you
what
I
did
for
c,
plus
plus
I
just
copy
and
pasted
the
compliance
matrix
and
like
what
I'm
reviewing
with
like
an
x
mark,
and
then
I
open
bugs
with
things
that
I
find
in
there.
If
you
want
to
do
a
precursory
review
of
everything
before
I
go
to
town,
go
for
it.
C
If
you
want
to
do
this
together,
I'm
fine
with
that
as
well.
It
would
be
a
long
long
meeting,
but
I'm
happy
to
do
that
and
yeah.
Hopefully
we
can
get
it
done
quickly.
B
Cool
yeah,
I'm
gonna
do
a
precursor
and
brian.
Maybe
we
can
slack
through
it
a
bit
together
and
get
that
done
before
next
week.
C
Awesome
awesome
look
forward
to
it
man
this
is
gonna,
be
awesome.
I'm
I'm
really
excited
to
see
all
these
1.00.
This
is.
This
is
super
fun.