►
From YouTube: 2020-07-30 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
A
So,
with
the
context
thing,
my
issue
is
with
your
most
recent
suggestion.
It's
been.
I
can't
remember
the
reason
that
I
didn't
like
I.
I
know
there
was
an
issue
for
why
I
didn't
think
we
could
just
put
the
whole
context
back
and
I
can't
remember
what
it
is.
I've
been
trying
for
days,
so
maybe
there
isn't
an
issue.
C
C
Well-
and
that's
that's
a
I'm
glad
you
mentioned
that
I
mean
like
said,
because
that's
not
my
goal
right,
my
goal
is,
and
they
and
the
the
spec
kind
of
exactly
says
what
I'm
trying
to
do
here
right,
it's
it
when
it
goes
into-
and
this
is
part
of
the
body
of
the
new
context,
detach
and
detach
stuff,
which
is
a
slightly
separate
issue.
C
C
And
what
I'm
saying
is
I
want
this
so
that
I
can
write
open,
telemetry,
instrumentation,
libraries,
yeah,
sure,
okay,
right
and
I
am
100
you're
totally
right.
It
like,
if
there's
any
possible
way
that
somebody
can
use
with
spam
instead
or
the
elixir
equivalents.
They
should
totally
do
that
because
it's
much
more
convenient
it's
easier.
B
C
And
then,
and
they
can't
get
in
trouble,
they
cannot
possibly
get
in
trouble
right,
whereas
the
even
current
ctx
right,
exposing
that
information,
that's
not
for
our
users.
A
C
That's
for
us
right
and,
and
what
I'm
proposing
here
is
is
stuff.
That,
again,
is
it's
not
for
our
users,
it's
for
us
so
that,
even
if
our
users
are
stuffing
up,
even
if
our
users
are
using
the
rest
of
those
functions
in
a
way
that
rsc2119
says
they
should
not
yeah
per
the
spec
that
the
the
the
moment
they
hit
a
any
of
our
code,
whether
it's
with
spam
3,
putting
the
context
back
so
that
it
guarantees
that
at
least
none
of
the
spam
expanded
its
callers
spans
won't
get
stuffed
up.
C
That's
that's
awesome.
Self-Defense.
Thank
you
with
span
and
I
want
one
of
our
instrumentation
libraries.
It's
forced
to
hook
up
via
telemetry
or
a
telemetry
like
api,
which
means
it's
not
able
to
intercept
the
the
call
to
the
next
thing
on.
I
want
our
instrumentation
library
to
be
able
to
do
the
same
thing
as
with
span
3
does
in
the
in
the
middle
of
the
sdk,
so
that
it
can
go
well.
Any
of
your
code
that
I
called
through
might
have
stuffed
up,
but
at
least
I
can
put
everything
back
just
like
with
spam.
C
Three
does
so
that
anyone
upstream
of
me
doesn't
get
stuffed
up:
okay,
yeah,
just
a
self-defense
thing
right,
because
there's
nothing
right
because,
like
I'm,
a
big
fan
in
general
of
giving
people
really
clear
signals
and
if,
if
we're
setting
up
a
situation
in
which
exceptions
can
cause
spans
to
get
unwrapped
in
the
wrong
at
the
wrong
times
in
the
wrong
order
and
stuff,
then
that's
it's
going
to
be
diabolically
hard
to
debug
that
it's
going
to
leave
people
with
a
bad
taste
in
their
mouth
that
they
might
talk
about
to
their
mates
with
the
word
open,
telemetry
attached
to
it.
C
Yeah
well,
and
and
and
that's
the
other-
that's
why
I
said
sort
of
telemetry
like
situations
as
well,
because
you
can
imagine
the
same
kind
of
thing
happening
if
you're
in
a
gen
server
and
there's
any
you
know
and
you're
relying
on
continue
and
going
off
and
do
something
else
and
the
result
comes
back
from
somewhere
else
and
then
then
you're
able
to
pop
the
context
right.
All
that
kind
of
manual
span
management
I
and-
and
you
see
that
in
real
life
in
even
in
things
like
phoenix,
live
view.
C
And
it's
also
it's
coupled
to
another
thing
which
I
I
don't
know
it
might
be
peculiar
to
the
elixir
world,
which
is
that
we're
combining
the
erlang
communities
crash
early
crash,
often
right,
if
there's
anything,
just
crash
the
whole
process
out
right
and-
and
so
the
idea
in
the
erlang
world
is
that
exceptions
are
an
exceptional
thing.
They
don't
happen
that
often
and
when
they
do,
they
have
very
small
splash
damage,
because
you've
got
a
little
tiny
ginseng
with
only
one
thing
to
do,
and
it's
totally
appropriate
to
to
just
take
that.
C
One
thing
out
takes
out
a
little
part
of
the
request
and
the
whole
rest
of
the
system
keeps
running
right
great
in
in
the
elixir
world.
We're
often
dealing
with
data
that
comes
from
systems
that
are
done
or
not
under
our
control
and
and
exceptions
happen
a
lot
more
often
like
we're
not
getting
data
from
our
own
code,
we're
getting
data
from
somebody
else's
code
and
that's
combined
with
a
a
performance,
optimization
thing
where
we
we
we
got
very
big
early
on.
It
seems
on
avoiding
serialization
to
other
processes.
C
So
we
will
do
almost
anything
to
do
all
of
the
work
in
one
process
and
it's
taken
to
its
extreme
in
something
like
absinthe,
where
you'll
have
a
whole
graphql
operation
chart
right
with
many
many
many
operations
rolled
up
in
it
and
if
there's
any
one
exception,
that
kind
of
inherited
erlang
treatment
of
I'll
just
take
out
the
entire
process
combines
with,
and
I'm
trying
to
do
as
much
work
in
one
process
as
possible
to
end
up
with
a
kind
of
a
ridiculous
outcome
right
anyway,
a
bit
of
a
canter
ant
there,
but
yeah
like
in
the
elixir
world
exceptions
are
happening
all
the
damn
time,
and
so,
whereas
from
like
an
erlang
program,
a
perspective,
you
might
be
going
like
your
length
should
take
out
the
entire
tree.
C
You're
not
going
to
get
you
know,
all
the
spans
will
just
get
cleaned
up
by
the
sweeper,
not
in
not
in
alexia
land,
where
quite
often
having
to
put
in
a
whole
bunch
of
rescues,
so
that
our
graphql
operation
doesn't
get
completely
unwound
we're
trying
to
catch.
Just
one
thing
turn
that
into
an
error
and
when
we
do
so,
if
our
spans
get
screwed
up,
we'll
have
no
idea.
What's
going
on.
A
Yeah,
it's
yeah,
it's
difficult
to
yeah.
Contexting
is
easy
to
do
if
you
just
use
a
variable
like
go
or
if
you
just.
C
C
Yeah,
it's
a
whole
sort
of
thread:
local
storage,
continuation,
local
storage,
lucky
people-
actually
node
can
have
some
problems
with
that
too
I've
I've
worked
on
trying
to
get
the
that
stuff
going.
You
have
to
deal
with
this
concept
called
async
contexts,
and
it's
diabolical.
C
C
Yeah,
it's
it's
just
that
thing
where,
because
the
you
know,
if
you're
trying
to
do
like
propagation
having
you
need
a
copy
of
the
spam
context,
so
you
can
then
turn
it
into
a
set
of
trace,
headers
or
whatever,
and
and
so,
if
I'm
dealing,
if
I've
got
a
function
and
I'm
dealing
with
spams,
then
I
I
I
expect
to
get
a
span
context
as
a
return
value
either
the
one
that
I
just
got
or
the
previous
one
and
I'm
still
a
bit
yeah.
C
It's
first
thing
in
the
morning
for
me:
I'm
a
bit
fuzzy
on
which
one
I
prefer
on
when
starting,
but
then
the
and
and
it's
it
again
it's
it's
kind
of
our
problem
that
we've
got
this
separation
between
the
the
span
context
and
the
tracer
context
right,
and
that
con
that
tracer
contract
value
is
is
required
because
the
span
context
doesn't
have
the
parent
in
it
right
right,
yeah.
C
So
so
that
forces
us
having
both
of
those
concepts
so
yeah,
I'm
kind
of
okay
with
having
the
you
have
both
of
those
available
like
in
in
my
gen.
C
C
C
You
know
a
little
libraries
like
if,
in
the
particular
case
of
telemetry,
like
I
want
to
give
somebody
another
wrapper
function
that
just
says:
okay,
look,
you're
gonna,
hopefully
you're
dealing
with
something
that
gave
you
an
integration
point
via
there
with
spam,
telemetry
with
spam,
3
and
so
you're
going
to
get
a
start
and
then
either
a
stop
or
an
exception.
C
A
Want
to
give
people
it
could
be
possible
if,
if
we
have
one
piece
in
the
process
dictionary,
that's
the
span
context
of
the
current
span,
then
a
key
that
is
the
span
context
itself.
That
looks
up
the
trace
context.
So
when
you
close
a
particular
span
and
maybe
you're
passing
in
the
span
context
like
in
your
case,
it
looks
up
what
the
trace
context
is
from
there
in
the
process
dictionary
and
it
gets
the
right
one
for
when
it
was
created.
C
That
would
make
the
whole
problem
go
away.
Yeah
sure
I
also
like
that,
because
I
mean
we've
got
to
also
keep
in
mind
that
the
value
of
n
is
going
to
be
nice
and
low.
C
I
mean
most
of
the
time
that
I'm
dealing
with
this
like,
if
I'm,
how
to
put
this,
if
I'm
hooking
up
to
something
like
this
via
telemetry,
it's
wrapped
around
an
operation
that
takes
a
millisecond
at
least
all
right
and
there's
going
to
be
a
number
of
network
hops,
but
like
the
the
worst
trace
graph,
I
ever
looked
at
had
a
thousand
entries
in
it
in
in
the
rare
case,
it
gets
into
the
tens
of
thousands.
C
C
Yeah
that
that
could
that
could
tie
things
up.
I
don't
know
I
mean
the
good
news
is
if
you
oh
yeah,
yeah
like
if
you
made
that
the
sport
that
the.
B
C
To
put
it
that
could
then
all
be
internal
in
ot
tracer
default
and
you
could
change
the
implementation
and
it
wouldn't
affect
anybody
at
all
right,
in
which
case
you
could,
if
you
had
current
ctx
and
set
ctx
to
bridges
in
the
time
being,
while
you're
playing
with
that,
then
you
could,
once
you
made
the
you
know,
made
the
problem
go
away.
You
could
then
deprecate
those.
C
Right,
well,
it's
it's
just
I
I
again
it's
a
super
ambitious
project
like
standardizing
all
of
this
stuff
across
all
the
languages,
but
just
I
mean
the
the
difficulty
just
you
guys,
understanding
where
the
hell
I'm
coming
from
as
an
elixir
programmer.
C
Right,
good
yeah,
thanks
yeah,
good
good.
Good
luck!
You
know
bridging
across
python
go
erlang
and
lua.
A
C
Anyway,
at
least
the
ruby
programmers
can
just
like
monkey
patch
everything
yeah
like
I
I'd
so
love.
C
Anyway,
so
brian
you
were
going
to
split
out
the
phoenix
it's
already
done.
C
B
No,
so
I
I
split
out
and
created
an
entire
phoenix
bridge,
because
we
don't
wanna,
we
don't
wanna
conflate
phoenix
into
plug
good
call,
especially
because
it's
also
gonna,
like
I
got
the
pr
merged
this
morning
for
the
exposing
context
headers
to
the
sockets.
Now
so.
B
Start
I
mean
like
we'll,
probably
start
a
span
whatever
like
if
you,
if
you
want
to
do
that,
but
then
then
treating
the
messages
on
more
of
an
async
level
of
producer
consumer
versus.
B
So
we
can
let
those
spans
flow
without
like
doing
it
like
as
events,
especially
because
events
don't
seem
to
be
supported
by
anything
right
now,
exporter
wise,
so
it's
kind
of
useless,
plus
the
events
won't
ship
to
the
collector.
I
think
until
this
fan
closes
out.
C
Oh
interesting
yeah,
I've
yeah,
I
implemented
a
kind
of
a
crude
hack
where
I'm
I'm
doing
linking
with
honeycomb
just
by.
C
C
I
I
like
using,
I
like
using
events
for
some
of
those
because,
like
especially
when
you
get
one
of
those
handle
events,
it's
it's
nice
and
small.
When
we
get
to
phoenix
live
view,
I
mean
I've
got
an
integration
method
that
works
with
the
the
views
themselves,
like
catching
their
handle
event,
handle
info
mount
render
and
and
other
callbacks.
C
The
fun
thing
in
that
one
which
bears
on
your
work
with
the
controllers
and
channels
is
you've,
got
the
first
request
for
the
html
that
gets
passed
through
to
the
phoenix
live
view
using
the
process
of
the
con
right
right
and,
and
we
want
to
kick
off
a
span
there,
and
then
we
we
have
to
smuggle
in
phoenix
live
view
at
the
moment.
The
span
context
out
by
having
the
the
templates
rendered.
C
Put
it
put
it
in
a
meta
tag,
yeah
and,
and
then
the
the
javascript
at
the
other
end,
can
scrape
it
out
of
the
you
know,
use
it
uses
the
dom
to
find
the
meta
tag,
pull
the
value
out
and
then
pass
it
back
down
when
it
initializes
the
socket.
C
I
believe
so
we're
forced
to
include
it.
Maybe
that's
just
for
the
live
reload,
but.
B
Would
be
nice
because,
that's
the
other
part
that's
going
to
have
to
get
like
at
some
point.
Kind
of
added
is
sticking
in
the
phoenix
client
like
generation
of
the
generation
of
the
stuff
in
the
context
headers
in
yeah
right
automatically
like
not
having
to
like
do
it
manually
and
then
like
the
linking
is
what
I
was
thinking
of
doing
was
like
you,
get
the
original,
like
parent
trace
id
yeah.
C
B
C
Yeah
and
what
I'm
I
mean,
what
I'm
doing
is
I'm
using
I'm
linking
the
like.
I'm
glad
you're
mentioning
using
trace
like
the
linking
capability,
because
you
know
that
that's.
C
Right
and
and
the
and
the
system
at
the
other
end,
if
you
leave
the
spans
open
and
say,
for
example,
as
long
as
the
live
view
process
survives
right,
it's
it's
terrible,
so
tristan,
that's
where
we
end
up
using
that
capability,
you
you
built
in
where
we
can
start
a
span,
get
its
context,
end
it
and
leave
the
spans
context
as
the
current
context
of
that
process.
C
B
We're
not
handling
producer
consumer
differently.
Are
we
like
you'd,
essentially
just
like
I
guess
you'd
have
to
like,
like?
Are
you
supposed
to
like
start
and
close
a
span
really
quickly?
That
has
like
I'm
a
producer?
I
here's?
How
I
think
it
would
happen
is
just
like
I'm
going
to
send
a
message.
So
I
start
a
spam,
I'm
producer,
I
send
it
at
the
end
of
the
send
I
close
out
and
on
the
other
end
it's
because
it's
got
to
operate.
I
guess
somewhere
to
an
http
client.
C
Oh
okay,
I
think
it.
If
you
take
a
look,
I
think
I
mean
drawing
on
a
fair
amount
of
the
the
history
here
I
mean
like
if,
if
you're
waiting
for
a
response
right,
I.
C
Yeah,
I
think
I
think
that
would
be
where
we
would
you'd
want
to
use
something
like
linking
right,
because
you
know
normally
what's
happening.
There
is
the
the
the
producer
is
in
a
very
short
amount
of
time
like
producing
an
event.
That's
going
to
then
cause
you
know
havoc
elsewhere
right,
in
which
case
you
know
starting
a
span.
While
you
get
that
ready
and
ending
it
once
it's
ready
right,
it's
fine
right
and
then
you
take
the
context
of
that
and
you
send
it
off
with
your
trace
propagation.
C
B
C
B
C
B
B
Tackle
that
piece
for
a
little
while,
like
my
main
concern
right
now,
is
like
I'm
trying
to
get
this
stuff
stood
up
on
the
product.
I'm
working
on
right
now,
because
we
have
a
lot
to
observe
there
and
uncover
and
there's
nothing
so.
B
B
So
like
like,
we
should
probably
talk
about
the
monitoring
thing
in
a
second
here,
but
so
writing
the
the
phoenix
one
one
of
the
things
that
came
up
and
I'm
I
feel
like
this.
This
definitely
needs
to
move
into
something.
B
If
we're
gonna
combine
the
repos
at
some
point,
then
I
guess
it
just
makes
sense
to
just
go
ahead
and
put
it
in
the
api
as
a
helper.
But
one
thing
that's
not
specified
anywhere
and
the
specification
whatever
is
actually
having
you
know
like
helper
things
to
especially
for
like
different
clients
to
get
the
spam
statuses
correct
so
basically
like
every.
B
A
B
C
B
Spit
out
like
what
the
proper
thing
is,
yeah
that'd
be
good
in
the
I
think.
That's
gonna
have
to
be
because,
like
I
had
it
here
and
then
I
like
started,
writing
stuff
to
wrap
around
like
finch
or
like
just
let's
just
say,
like
wrapping
around
an
http
client
period,
and
I'm
like.
B
And
it's
all
the
same,
I
mean
like
the
only
difference
is
like
do
you
want
to
like
potentially
change
what
the
message
is,
so
I
I
was
just
curious,
tristan
like
where
should
this
live
like?
Should
I
just
put
this
in
ot
spam
in
the
api.
C
Yeah,
I'm
the
only
I'm
kind
of
fine
with
the
first
two
segments
there
that
I'm
not
I'm,
not
a
big
fan
of
utils
I
and
while
elixir
itself
is,
is
full
of
like
4000
line
modules.
I
I
think
that
you
could
totally
instead
have
a
module
that
just
does
this
one
thing
that
just
deals
with
status
to
http.
B
Yeah
and
that's
like
there's
other
things
in
here
that
we
haven't
done
like
the
status
creation
like
this
is
kind
of
being
backwards
done
in
the
exporter
right
now
by
just
like
the
proto
like
stuff,
but
it
seems
like
it
does
say,
must.
C
Yeah,
oh
yeah
yeah.
Thank
you
for
again
yeah,
I'm
a
big
2119
fan,
yeah,
so
plugs
similar
module
is
called
plug.con.status.
So
I
think
you
could
just
totally
call
it
opentelemetry.http.status.
A
So
the
create
a
new
status
just
means
that
we
do
in
the
open,
telemetry
module.
You
can
create
a
new
status
so
yeah.
That's
all.
That
means.
B
B
But
like
the
flip
side
of
this
is
like
this
is
all
great,
but
it
like.
Doesn't
it
doesn't
key
on
the
thing
that
I
was
like?
Oh,
what
I
really
need
is
a
helper
that
takes
care
of
like
this
mapping.
C
You
know
so
back
to
that
I
mean
my
reading
of
that
particular
part
of
the
must
is
like
they
talk
about.
These
are
the
http
codes
and
then
they
talk
about,
and
these
other
things
are
the
the
canonical
codes
that
we
associate
with
those
http
codes.
C
Yeah,
I
I
mean
if
you
call
get
canonical
code
with
say,
conflict.
B
C
C
B
C
Do
you
mind
me
zooming
out
just
for
a
moment
tristan
I
keep
going.
I
keep
going
back
to
the
spec
to
try
and
find
stuff
and
when
I
do
searches
in
that
repository
instead,
what
I
find
is
oteps-
and
it's
striking
me
that
there
might
be
a
there-
might
be
a
situation
where
we've
got
like
the
written
law,
which
is
the
spec
and
then
you've
got
all
the
case
law,
which
is
the
oh
chips.
B
Oh,
it's
even
worse
than
that,
though,
guys
because,
like
the
protos,
don't
line
up
necessarily
what
the
spec
says
either
like
I
opened
a
issue
the
other
day
of
like
unknown
like
in
this
in
here.
Right
like
it
says
like
this,
is
unknown
yeah
and
then
they
say
like
unknown
in
other
places
as
status
kinds.
But
then,
like
you,
go
in
the
proto
and
it
says
unknown
error
so,
like
there's,
conflicts,
definitely
like
in
the
spec
against,
what's
like
in
the
proto,
even
yeah
the
yeah,
because
I
go
here
for
like
whatever
they
is
legit.
A
B
B
Here
and
then
just
like
picking
like
whatever
the
last
tag
was,
if
you
want
to
say
like
okay,
this
is
how
it's
working
now
and
then
I've
just
started.
I
know
tristan's
doing
this
for
sure,
but,
like
I
just
watched
this
guy
to
see
like
what
changes
they're
making
because
there
was
the
exporter
broke
because
they're
they
they
outright
changed
the
underlying
protos,
because
they're
not
honoring
at
all.
Right
now,
like
oh
yeah,
it's
a
proto
buff,
so
I
increment
things
like
they're,
just
like
nope,
don't
like
that
delete
it
now.
B
C
My
god,
okay,
so
brian,
like
totally
ping
me
on
the
pull
request
for
this
I'd
love
to
play
code.
Golf
with
you
on.
C
On
the
status
code
thing
yeah
like
that
yeah
get
that
tidied
up,
because
I
I
we're
gonna
be
using
like
at
the
moment
we've
got
everybody's.
C
Yeah
right,
it's
good.
It's
going
to
be
across
everything
and
there's
just
something
about
the
way
with
the
code
that
we're
writing
at
work.
At
the
moment
where
we
keep
like
every
time,
we
grab
an
external
dependency
to
deal
with
some.
You
know
cloud
providers
stuff.
C
We
we
have
to
write
a
different
shim
around
their
strange
use
of
either
hackney
or
http
poison
or
whatever
the
hell.
It
is
right,
they
all
use
different
clients,
but
they
provide
a
behavior,
thankfully
to
let
us
get
between
them
and
it,
and
then
we
re-shim
it
into
a
tesla
client
which
which
then
always
uses
the
same
way
to
get
outbound.
C
B
Yeah,
like
I'm,
waiting
on
chris
to
finish
up
some
telemetry's,
there's
telemetry
revisions
on
finch,
but
in
like
the
meantime,
it's
just
like
I'm
doing
the
same
thing
of
shimming,
where
I
just
like.
I
I'm
wrapping
the
request
with
a
span
I
set
like
base
attributes
and
then
after
it
creates
the
request,
object.
C
B
C
Right
and
then,
and
then,
when
you
get
one
layer
in
you,
hit
the
phoenix
router,
you
can
update
the
span
name
again
with
the
route,
which
is
much
more
useful.
I
I'm
so
cross.
I
can't
do
that
in
open
senses-
oh
one
other
note,
brian
just
for
when
you're
hacking
all
of
this
stuff
in
I
don't.
I
think
finch
is
more
in
the
you.
You
get
to
name
your
own
gen
server,
kind
of
style
of
he's.
B
C
C
B
C
So
so
I
just
I
just
wanted
to
add
one
thing,
and
then
I
need
to
find
out
what
tristan
said
on
the
agenda,
because
we
are
not
sticking
to
it
when
when
you've
got
another
trick,
I
mean
that
the
trick
I'm
using
for
the
fenix
live
view
is
very
handy
for
any
of
these
nasty
integrations
where
they,
where
you
use
their
macro,
and
so
it's
your
module
and
you
use
their
macro
and
then
it
defines
a
whole
bunch
of
things
that
that
the
the
code
path
thing
goes
through.
B
C
You
can
take
any
of
those
methods
that
they
dropped
in
and
then
you
could
def
overrideable
and
then
call
super.
You
know
call
the
one
that
they
wrote
from
yours
and
and
then
you
can
put
that
in
a
a
with
span.
C
It's
it's
like
this
magic
integration
method
that
gets
you
in
the
call
path.
Whenever
somebody
used
a
macro
to
shove
code
into
a
module
of
yours,
and
I'm
just
mentioning
it
because
again,
a
lot
of
people
are
getting
their
telemetry
wrong,
but
when
the
way
that
they
work
is
involves
that
pattern
you
can
go,
I
don't
need.
B
C
Which
is
what
I
did
with
the
phoenix
live
view,
but
I'm
just
mentioning
it
because
I
haven't
written
it
up
yet
and
it's
totally
something
I
I
can
never
run.
B
C
You
write
it
up.
Yeah.
B
B
Like
just
seeing
like
the
f
response
from
like
when
I
made
the
pr
that
had
like
the
base
implementation
of
just
adding
the
phoenix
and
ecto
exporters
or
bridge
libraries,
they
were
like
wow.
That's
all
that's
in
this
pr
and
it's
like
yeah
and
then
like
once,
you
have
it
running
whatever
and
you
have
the
the
collector
configuration
is
so
dead,
simple
to
do
like
locally.
I
was
doing
like
zipkin
and
jager
at
the
same
time
exporting
and
then
added
light
step.
B
It
took
longer
to
sign
up
the
demo
account
for
lightstep
than
it
did
to
like.
Have
it
exporting.
I
mean
it's
like
four
four
lines
of
config
and
it's
like
cool
like
done
so
I
think
like
people
like
once
they
get
into
this
they're
gonna
we've
got
it
running
like
they're
gonna
be
like
holy
crap
like
I
can
just.
I
can
just
switch
and
trial
other
people
like
like
at
work.
I
was
just
like
you
know,
like
it's
cheap
enough,
guys
like
we
can
actually
just
like
run.
C
C
They're
they're
the
only
two
that
can
actually
deal
with
the
high
cardinality,
so
yeah
at
the
moment.
So
great
you
know
get
into
those
and
then,
if
you
know-
and
I
like
the
you
know,
if
someone
else
is
doing
a
better
job
of
say,
metrics
great
use
them
for
the
metrics
I
mean
metrics
have
their.
I
mean
if
you're
dealing
with
usage
saturation.
You
know
those
kinds
of
concepts,
then
metrics
are
wonderful.
Go
do
that
too.
A
A
A
A
B
I
mean
the
only
other
commonality
is
like
there's
a
lot
of
attributes
that
you
end
up
setting.
However
they're
all
kind
of
I
mean
like
there's
the
con.
The
plug
one
is
like
generally
the
same
between
like
phoenix
and
blood,
but
other
people
are
gonna.
Have
it
stuck
in
different
places
as
to
like
how
the
hell
you
get
that
information.
C
Yeah,
good
good
good
luck,
populating
net
dot
star
like
I
I
I
can
get
the
I
can
get
the
incoming
net
star,
but
I
cannot
get
the
local
stuff
I
was.
I
was
gonna,
try
and
get
a
pr
and
plug
cowboy
for
that,
but
I
got
distracted.