►
From YouTube: 2021-10-14 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).
C
B
A
B
Yeah
we
were
starting
to
get
everything
set
up
with
cowboy
and
phoenix
and
whatnot
and
just
wanted
to
know
if
there
was
already
a
plan
for
how
to
do
the
context
propagation
there
between
the
two
or,
if
we
should,
you
know,
go
ahead
and
propose
something
or
if
he
explicitly
didn't
want
it
to
be
there.
A
D
A
Oh
yeah
brian
says
he's
gonna
be
late,
but
he's
gonna
be
here
in
within
10
minutes.
A
But
yeah
I
could
just
he
already
knows
the
other
agenda
items
are
just
announcements
of
we
have
a
our
ce3
is
out,
and
technical
committee
is
working
on
figuring
out
how
to
review
our
are
going
ga
he
has
to
go
through
the
technical
committee
review,
which
usually
works
by
someone
in
the
committee
who's
familiar
with
the
language
reviewing
it,
but
since
that,
doesn't
that
person
doesn't
exist
in
this
case,
they've
been
trying
to
figure
out
how
to
move
forward
and
they're.
A
C
B
Do
you
know
how
the
technical
committee
works
like?
Could
you
be
on
the
technical
committee
or
are
you
I
guess
you're
not
already.
A
Yeah,
I'm
actually
not
sure
I
meant
to
look
into
that,
because
the
government
governance
committee
elections
are
coming
up
like
in
a
couple
weeks.
I'm
not
sure
I
can't
remember.
I.
B
Was
just
thinking
because
you've
been
super
involved
in
like
the
back
and
forth
on
all
the
spec
stuff
and
you
know
like
helping
guide
decisions
there.
So
I
don't
know
what
constitutes
being
technical
committee.
I
only
ran
up
about
like
the
membership
and
contributors
and
stuff
like
that.
But
now
there's
so
many
terms,
yeah,
I'm
not
sure
if
they
get.
A
Put
on
there
by
the
governance
committee
or
if
they're
also
elected,
I
don't
think,
they're
elected.
So
let's
go
through
the
government's
commit
committee
which
I'm
not
going
to
be
running
for,
but
was
considering
it
but
splunk
you
can
only
yeah.
It's
just
a
number
of
people
at
splunk
are
running,
so
I'm
probably
hopefully
gonna
run
next
time
and
then.
A
And
then
yeah
once
we
have
ga
I'm
working
with
a
couple
other
people,
one
from
ruby
and
javascript,
to
write
a
blog
post
announcement
about
our
gas,
because
they're
also
going
ga
right
now
so
that'll
be
nice,
something
on
open,
telemetry's,
blog
combined
with
them
big
announcement.
That's
pretty
much!
It
yeah
a
lot
of
works
going
on
especially
a
lot
of
stuff
coming
in
through
contrib
and
being
updated,
and
I
think
we're
pretty
close
to
pre.
A
E
B
Well,
I
guess
yeah
as
we
get
things
integrated
at
work,
we
can
give
any
feedback
we
have
about.
Does
it
seem
ready,
or
is
there
a
gap
missing
and
contrib
or
something
we're
kind
of
just
planning
to
drop
in
all
the
easy
contrib
stuff
at
first
and
just
get
started,
and
you
know
hook
it
up
to
a
few
vendors
and
see
yeah,
see.
D
B
A
B
Let's
see
anything,
something
that
passed
in
mind
for
me
is
just
that,
like
the
docs
and
getting
started,
guide
probably
still
need
more
work,
and
I
have
been
meaning
for
months
to
help
out
with
that
notice
having
it.
So
maybe
this
will
be
a
good
excuse
to
write
up
some
stuff
as
we're
learning
it.
D
B
D
I
can
probably
write
a
few
things
up
too
after
I
get
it
done.
One
thing
I
did
notice
is
that
everything
like
I
see
in
several
examples
where
it
talks
about
doing
open,
telemetry
dot.
I
don't
remember
what
it
was
like
register
application,
tracer
yeah,
something
like
that
and
that
just
okay
yeah
I
was
like
I
was
looking
around.
I
don't
see
this
anymore-
is
that
I
think
that's
still
on
the
website.
Docs
yeah,
it's
not
on
the
actual
readme,
but
it's
in
some
of
the
example
code
I
mean
I
can.
D
A
B
So
there
used
to
be
a
per
application
tracer
that
you
would
have
to
register
yourself
and
now
there's
still
a
per
application
tracer,
but
it's
automatic.
A
C
B
A
Can
be
will
be
fixed
but
yeah
for
now
it's
just
such
an
edge
case.
Yep.
B
Cool
well
brian
seth
and
I
came
bearing
gifts
of
questions.
B
C
So
I'm
gonna
do
it
it's
just
we
were.
I
was
waiting
to
release
cowboy
until
after
we
had
one
three
done,
because
I
just
it
wasn't
out
yet,
and
I
didn't
really
want
to
mess
with
having
to
upgrade
that
one
as
well,
because
it
had
propagation
in
it
so
yeah.
It
literally
got
published
last
night
rc3
got
published
two
days
ago,
yeah.
D
I
was
using
more
than
48
hours
ago,
so
I
was
using
your
fork
until
this
morning
so
yeah
last
night
I
saw
it.
I
was
like
oh
look
four
hours
ago.
He
got
this
working
with
the
rc3,
so
the
the
the
phoenix
open
telemetry
phoenix,
but
that
doesn't
have
the
propagation.
C
Yeah,
so
the
next
step
will
be
that
the
phoenix
library
will
just
check
to
see
if
there
is
a
span
context
already
and
if
it
does
on
start,
then
it
will
just
not
pull
the
propagation
and
it
will
just
do
the
update,
attribute
step.
B
So
the
problem,
though,
is
that
the
cut
the
cowboy
one
runs
in
a
different
process
than
the
phoenix
one
right.
I
don't
know
if
you're
already
aware
of
that,
so
you
have
to
like
what
I've
done
in
the
past.
Is
you
magically
you
like
put
something
in
a
nets
table
that
like
has
the
cowboy
process
id
as
a
key
and
then
in
the
phoenix
one
you
can
say
like?
What's
my
ancestor
and
like
you
can
find
or
like
call
her,
I
think
or
something
I
forget
which
one
it
is.
C
C
Yeah
so
there's
another
issue
that
in
ecto
that
I
have,
we
have
to
figure
out
a
solution,
for
there
is
a
pr
open
for
the
ecto
one
that
still
needs.
If
you
want
to
do
a
review
on
it,
the
main
issue
that
I'm
looking
for-
and
I
don't
know
if
the
contributor.
B
Parallel
preloads
and
also
there's
an
async:
it's
not
async
stream.
There's
some
similar
thing
to
async
stream.
That
hector
does.
C
Yeah,
so
I
gave
him
direction
on
it.
I
haven't
he.
He
was
paying
about
it
again
yesterday,
but
if
we
solved
that
one
it
would
solve
the
cowboy
one
as
well.
B
Right,
that's
why
I
would
propose
the
the
x
thing
so
that
the
parent
can
like
put
a
breadcrumb
with
its
own
pid,
and
then
the
child
can
look
up
that
breadcrumb
using
the
ancestor
pid
like
basically,
the
table
would
be
a
key
value
pair
of
parent
pid
and
trace
context
or
whatever
we
want
to
put
in
there.
Yeah.
B
C
It
was
actually
good
that
it
waited
because
the
one
three
or
the
yeah,
the
rc3
actually
like
raised
an
issue
with
the
cowboy
implementation
because
of
the
long-running
separate
process.
C
So
I
had
to
yeah,
I
had
to
change
it
so
that
basically
like
if
you
got
a
parent,
if
you
got
a
header
for
a
trace
context
and
you
pulled
it,
it
got
set
into
the
cowboy
part
which
is
fine.
That's
it
should
be
on
that
span.
But
then,
when
it
closed
out
like
there
was
still
a
current
context,
so
you
would
inherit,
like
whatever
the
previous
thing
was.
If
it
didn't
have
new
stuff.
B
C
B
C
Yeah
and
there's
like
insanely
limited
information,
that's
on
it
yeah.
I
remember
yeah.
That
was
a
struggle
at
one
point
which
is
fine
like
we.
The
whole
idea
is
just
like
with
phoenix
as
like.
You
know,
you
start
a
span
and
then
you
keep
adding
to
it
and
modifying
it
until
for
with
whatever
you
can
like.
B
C
Yeah
precisely
so
like
right
now,
I
think
you
would
essentially
end
up
with
just
nested
if
you
were
to
like
run
it
today.
The.
C
B
C
C
But
I
want
the
same
solution
whatever
it
is.
I
want
the
same
solution
for
both
because
boiled
down.
This
is
a
propagation
issue
and
it's
a
part
of
the
propagation
that,
like
we
don't
have
any
answer
for
it,
and
people.
B
C
B
A
C
A
B
B
B
D
B
B
C
The
thing
that
I
would
like
to
find,
though,
like
my
goal
of
a
solution,
is
that
we
in
the
contrib
repo
that
we
have
a
like
ancestor,
propagator
or
something
that
does
a
that,
does
a
one
look
back
it
only.
It
looks
back
like
one
spot
or
it
does
come
with
a
gen
server
or
something,
and
then
that's
what
the
other
libraries
use.
B
Yeah,
I
I
get
where
that's.
B
Yeah,
yeah,
and,
and
also
the
other
nice
thing,
is
that
in
the
contrib
repo
we
can
like
since
they're,
both
the
same
repo,
we
could
implement
it
in
the
cowboy
and
the
phoenix
and
like
expect
that
they'll
work
well
together.
For
some
release
of
I
mean
we
don't
release
the
whole
contribute
anyway,
but
at
least
like
inside
that
same
mono
repo.
You
have
consistent
behavior.
C
B
But
you
definitely
need
that
common
library,
because
the
behavior
is
only
going
to
work,
if,
like
you,
have
it
implemented
the
same
way.
It's
just
a
magic
x,
key
yeah,
pretty
much.
But
I
was
going
to
ask
like
I
get
where
you're
going
with
only
looking
back
one
ancestor.
But
what
if
we
implemented
the
x
lookup
to
take
like
you
know,
do
a
select
on
ads
and
match
any
of
these
ancestors
and
like
just
give.
If,
if
there's
anything,
tell
me
about
it,
but
I
think
that'd
be
efficient
enough
or.
C
B
Yeah
yeah,
I
mean,
I
guess
at
scale
like
anything,
can
be
a
problem,
but
I'm
just
thinking
like.
Maybe
there
wouldn't
be
that
many
things
in
the
table,
because
it's
like
only
in
flight
requests
right
now,
so
because
you're
going
to
clean
them
up
as
soon
as
you're
done
like
there
might
only
be
15
things
in
the
table.
C
A
So
this
this
whole
thing
in
the
I'm
leaving
out
the
ecto
side,
the
cowboy
phoenix
side.
This
is
only
if
you
want-
or
you
want
a
span
that
starts
at
cowboy
and
span
at
cowboy
and
users
don't
have
to
do
that.
We
should
probably
recommend
that
they
don't
unless
they
really
don't
want
to
miss
that
in
that,
like.
C
C
It
doesn't
take
a
lot
of
time.
But
it's.
B
Yeah,
I
mean
that's
good
to
be
aware
of.
We
should
definitely
like
have
something
in
the
docs
about
that.
At
the
point
where
we
have
talks
about,
you
know:
here's
how
you
should
set
this
up.
So
were
you
saying
you
were
going
to
recommend
that
people?
Just
don't
do
the
cowboy
thing,
because
I
think
my
issue
there
is
that
the
telemetry
that
phoenix
gives
you
is
actually
not
that
good,
and
I
was
counting
on
cowboy
to
give
me
better.
B
Telemetry,
like
phoenix,
doesn't
make
harley
any
guarantees
that
it's
ever
gonna
tell
you
about
something
if
it
crashes
in
the
middle,
whereas
cowboy
is
pretty
reliable,
that
it
either
gives
you
an
early
error
or
start
at
least
you
know
something
happened
where
phoenix
is
not
as
good
about
that.
So
you
you,
you
don't
get
a
span
at
all
in
some
cases.
B
If,
if
something
crashes
at
cowboys
level,
then
phoenix
may
just
never
tell
you
anything,
and
you
don't
know
that
you
had
that
failed
request.
It'll
just
return
a
500
back
to
the
user.
It's
like
specifically
an
early
error
would
be
that
right,
like
it
never
got
to
phoenix,
because
there
was
an
early
error.
Well,
you're
gonna
have
because,
like
some
tote,
some
like
cookie
was
too
big
or
something-
and
you
just
never
know
about
that.
C
B
B
B
Thing
where
cowboy
starts
the
connection,
you
start
the
phoenix
thing,
but
then,
like
the
client
closes
the
connection
and
cowboy
ends
the
process
and
phoenix
never
sends
an
exception
or
an
end.
So,
like
you,
just
never
get
that
trace
completed
on
the
phoenix
side,
whereas
cowboy
does
send
an
end
event.
So
you
like
at
least
know
that
request
happened
for.
However
many
minutes
until
the
client
decided
to
hang
up
right.
You
know
that
that
long
request
happened
and
like
potentially
what
work
happened
in
between
so
like.
C
In
those
cases
too,
like
like,
I
get
it,
but
the
thing
I
was
gonna
say
on.
That
point
is
we
do
have
I
mean
it
hasn't
been
touched
in
a
year
and
a
half
but
tristan
did
start
the
process,
monitor
option
and
in
that
case
like
that
would
also
be
a
solution
where,
if
you
don't
want
to
do
at
the
cowboy
level-
and
you
just
want
to
track
phoenix
and
your
request,
volume
is
low.
Whatever
then
using
the
monitor
solution
is
something
that
we
had
already.
C
B
A
I
was
gonna,
I
was
gonna
say,
though,
with
the
someone.
Obviously,
some
of
that
can
just
be
metrics
and
not
traces,
so
you
don't
have
to
do
the
look
up
the
process,
dictionary
and
or
use
the
nets
table.
That's
the
stuff
that
creeps
me
out
and
but
the
would
you
be
able
to
somehow
get
that
those
metrics
into
the
span
that
phoenix
creates.
B
I
wonder,
like
probably,
I
think,
it'd
be
the
same
propagation
problem
right.
If
you
wanted
to
attach
the
metrics
to
the
trace,
you
still
have
to
propagate
it
like
if
you're
talking
about
just
statsd
sure
you
can
fire
them
off
like
stateless.
B
C
Okay,
they
just
run
yeah.
I
guess
that
that
is
one
question
tier
is
so
like
my
intention
with
the
phoenix
integration
to
cowboy
is
like
it's
not
gonna
run
a
separate.
C
C
It
does
have
the
full
url,
but
I
don't
know.
If
that's
I
mean.
C
A
tooling
thing,
because
most
tooling
is
going
to
show
like
at
the
top
level
of
those
requests.
It
would
just
say
cowboy
and
then
then
it
goes
into
like
a
monitoring
thing.
So
if
you
wanted
to
monitor
at
the
at
the
total
request
level,
the
problem
then
is
like
no
tooling
supports,
like
put
alert
like
track
this
stream
of
information
and
alert
on
it.
C
C
I
don't
know
how
much
like
you
would
have
to
do
math.
I
think
we
could
record
events
as
to
like
when
phoenix
starts
and
ends
like
the
different
parts
of
the
call
stack.
C
C
B
Right-
and
I
guess
that's
another
point-
slash
reason
that
I
do
want
to
have
a
cowboy
span.
Is
that
like,
if
you
don't,
if
you
only
have
a
phoenix
fan,
and
you
try
to
do
that
kind
of
monitoring,
you're
actually
just
lying
to
yourself,
because
you
aren't
tracking
anything
before
phoenix,
starts
telling
you
that
it's
doing
something?
No.
B
If
you
only
had
phoenix,
you
wouldn't
know
about
anything
before
phoenix
got
its
hands
on
it,
and
that
depends
on
literally
your
application
code
having
a
phoenix
telemetry
plug
in
it
right
I
mean
you
can
add
more
telemetry
plugs
in
there.
If
you
want,
I
mean
you
could
I'm
just
saying
like
the
way
that
a
normal
app
gets
generated
right,
it's
like,
after
all,
the
static
stuff,
it's
after
all,
the
like
jason,
pre-processor,
stuff
sure.
C
B
B
B
Yeah,
I
think
what
I
think,
what
what
I
would
want
is
to
have
the
cowboys
fan
be
there,
but
but
update
it
with
all
this
all
the
metadata
from
phoenix,
but
then
also
have
like
a
phoenix
controller
action
span
right
like
once,
you
get
into
your
actual
application
code
where
it's
like
this
is
the
act,
the
actual
controller
action
that
I'm
running.
B
C
If
you
were
interested
in
that,
there's
not
really
any
useful
info
on
it
that
I
found
but
yeah
so
like
the
whole
thing
is
like
once
you
get
plus
past
plug
telemetry,
you
get
a
router
event
and
then
that's
pretty
much
it
at
that
point.
It's
gonna
go
through
the
controller,
and
if
you,
basically
you
can
put
another
span
in
your
controller,
it
just
can't
include
the
actual
return.
B
C
B
Well,
I
guess
all
of
that's
kind
of
deep
in
the
weeds
for
us
at
this
point,
but
we'll
we'll
report
back
what
we
find.
B
I
guess
our
next
action
then
as
far
as
seth
and
I
go
is
that
we
can
look
at
trying
to
come
up
with
a
generic
context,
propagator
thing
that
we
can
use
for
ecto
and
for
cowboy,
because
I
think
I
mean
I
guess
the
other
option
that
we
have
personally
is
that
we
can
just
not
use
cowboy
for
now
or
we
could
have
it,
but
it
isn't
connected.
That's
also
a
thing
we
could
do
so.
C
C
B
C
Then
it
goes
in
like
yeah,
then
it
goes
in
the
reverse,
because
then
you
have
to
say
like
from
the
cowboy
process
say
what
is
the
child
process?
I
called
find
the
current
context
of
that.
One
pull
that
back
up
so
you'd
have
to
like
do
the
whole
mapping
thing
and
then
you'd
still
have
to
ask
for
the
context
and
pull
the
context
out
which
at
that
point
by
the
time
it
gets
up
to
there
it's
going
to
be
an
s.
C
B
I
I
implemented
it
was
gross,
but
I
implemented
this
at
bleacher
report
now
I
don't
have
access
to
the
code,
but
like
we,
we
had
a
bi-directional
communication
between
cowboy
and
and
phoenix
all
working
and
doing
what
we're
talking
about,
but
it
was
gross
so
like
it
was
all
you
know
like.
If
this
is
here,
I
guess
use
it.
I
don't
know
where
it
came
from,
but
you
know
that
kind
of
thing
like
in
the
in
the
cowboy
telemetry
thing
like
when
it
closes.
B
C
B
B
Like
like,
we
all
know,
this
won't
happen,
but
I
think
a
better
long-term
solution
is
just
to
have
cowboy
have
a
hook
that
would
like
you
know,
like
from
the
other
process,
somehow
call
its
parent
and
like
pass
something
through
right,
like
some
user
configurable,
whatever
you
want
to
have
passed
through
from
your
cowboy
connection
to
your
like
when
it
starts
your
phoenix
endpoint,
for
example,
it
could
pass
a
parameter
through
for
you.
That
would
be
a
better
solution,
but
I
don't
see
how
that
would
ever
happen
in
real
reality.
B
B
So
one
of
the
issues,
something
that
you
would
configure
in
there
like
here's,
how
I
want
my
web
server
to
talk
to
my
phoenix
like
somewhere
in
there,
you
would
say,
like
you
know,
pass
whatever's
in
this
predict
key
as
a
parameter
to
this
thing.
You
know
you
would
just
tell
it
that
declaratively
and
then,
whenever
it
starts
up
phoenix,
it
would
say,
like
oh
cool,
I
have
a
context
in
my
pdx
I'll
pass
it
in
and
then,
when
the
end
point
starts
to
run,
it
would
just
have
that
yeah.
If.
C
Yeah
I
mean
like,
if
you
want
to
put
together
like
an
s
thing.
I
think
one
of
the.
C
No
thanks,
it's
good
feedback.
I
I
probably
I
might
work
on
it
a
little
bit
this
weekend,
we'll
see
I'm
on
call
so
well.
We.