►
From YouTube: 2021-10-20 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
No
no
go
ahead.
A
Yeah
in
camas,
which
is
across
the
river.
C
C
Did
you
know
so
you
know
constance
right.
Obviously,
of
course,
her
mom
worked
at
expo
86
and
she
still
has
her
badge
like
constance.
Has.
D
Her
mom's
badge,
I,
my
older
sister,
worked
at
the
mcbarge,
the
floating
mcdonald's
barge
at
x86,
and
I
think
she
also
still
has
her
badge
nice,
but,
yes,
a
very
different
city.
D
D
Heritage
home
that
used
to
be
here
that
was
moved
or
something
I
don't
know.
This
place
was
built
no
six
anyways.
Let's
give
people
a
couple
more
minutes
to
join
so
starting
next
week.
The
meeting
will
likely
shift
to
the
time
that
we
wanted,
because
I
will
have
fixed
all
of
our
zoom
issues
and
created
different
zoom
issues,
but
I'll
have
fixed
the
current
ones
that
people
are
complaining
about.
So
a
quick
question,
while
people
start
to
join
was
our
preference
to
do
this
at
8,
30
a.m,
pacific
or
8
o'clock
a.m.
D
B
Once
in
a
while
I
might
be
late,
I
need
to
drop
my
kid
off
at
school,
okay,
but
yeah,
I'm
doing
so.
A
B
D
Fun
all
right.
That
was
my
only
update.
I
was
not
here
last
week,
so
I
think
we
were,
I
think,
john
or
someone
or
evo
mentioned
he
caught.
He
caught
me
up
and
said
we
reviewed
the
plan
doc
and
I
think
you
see
in
the
agenda
we're
going
over
the
trace
response,
header
and
the
missing
functionality
section
this
week,
yeah.
D
Yeah,
it's
a
small
group.
I
know
that
evo
and
sea
route
from
splunk
will
not
be
joining,
or
at
least
I
know
the
evo
won't
because
I
think
he's
flying.
He
was
in
california
earlier
this
week
and
I
think,
he's
on
a
plane
flying
home
right
now.
D
And
yeah
for
us,
it's
our
big
annual
conference
this
week,
so
I'm
guessing,
we
will
not
see
them,
can't
speak
for
other
people.
D
It's
five
past,
so
we
might
as
well
kick
this
off
so
yeah.
I
gave
the
update
on
the
calendar.
Missing
functionality
section
appears
to
be
the
next
topic.
B
Today,
we'll
we
can
go
through
a
few
things
that
I
wrote
under
the
missing
functionality
section,
and
this
is
just
based
on
my
you
know,
a
brief
understanding
of
you
know
open
telemetry
concepts.
I
am
relatively
new
compared
to
you
folks,
so
let
me
know
as
we
go
so
so
I
think
the
first
and
the
most
important
one
was
like
what
is
the
signal
we
should
be
using
for
indicating
events
coming
from
the
client
side
right.
B
I
think
I
think,
given
that
the
events
you
know,
data
model
exists
only
within
a
span
today,
like
there
was
a
span
event
object
within
a
span,
it's
not
available
at
a
high
level.
I
guess
you
know
that's
why
events
couldn't
be
used,
and
I
guess
that's
why
the
the
new
folks
at
splunk
ended
up
using
the
zero
duration
span,
but
now
that
there
is
a
log
data
model,
you
know
it
is
probably
closer
to
an
event.
B
There
are
some
fields
which
are
not
relevant
to
an
event
like
severity,
but
but
then
they
are
optional
anyway.
So
if
we
were
to
you
know,
standardize
on
using
the
log
data
model,
then
I
I
guess
we
will
need
to
work
with
the
log.
Stick
to
make
sure
there
is
enough
distinction.
You
know
between
the
two
for
two
different
use
cases,
but
what
those
would
be.
I
have
no
idea
at
this
point.
B
There
was
also
another
thought
that
came
up
in
the
last
meeting,
where
we
said
that
if
it
is
possible
to
detect
one
event
led
to
another
on
the
client
side,
then
you
can
model
that
as
a
trace,
so
that
you
can
link
those
events
as
pants
under
address
right,
but
it
was
just
a
thought
there
weren't
any
specific
examples
we
could
come
up
with
where
it
is
indeed
possible,
like
for
example,
you
you
click
a
button
and
then
it
led
to
you
know,
making
a
a
network
call
right.
B
So
my
understanding
is
it's
it's
hard
to
identify
that,
but
if
we
can,
that
will
be.
You
know
really,
you
know
useful.
B
So
I
think
this
should
probably
be
discussed
with
the
the
individual
like
the
js
and
the
android
and
the
ios
teams,
and
only
if
that's
possible,
then
we
those
situations
could
be.
You
know
modeled
as
a
trace
are
included
in
the
trace
and
it
doesn't
have
to
be
a
network
request
as
such.
It
could
be.
Let's
say
you
clicked
something
and
in
response,
another
event
happened
right
like
you
rendered
another
screen,
so
you
want
to
see
how
long
that
takes
so
the
trace
can
be
entirely
on
the
client
side.
C
So
it's
going
to
be
tricky.
I
can
imagine
it
would
be
very
tricky
for
us
to
figure
out
how
to
advise
people
writing
instrumentation
on
when
to
use
a
log
and
when
to
use
a
span,
given
that
they
they
look
sort
of
the
same,
except
for
this
causality
piece
where
logs
don't
have
any
causality
baked
into
them
at
all,
whereas
fans
clearly
have
causality
baked
into
them
with
their
parent
right
so
that
they
get
trying
to
figure
out
when
how
to
make
that
advice
and
how
to
make
those
kind
of
lay
down.
A
There
was,
there
was
also
the
proposal
or
idea
of
having
like
a
streaming
streaming
sdk
and
where
you
could,
you
could
capture
like
the
start
of
a
span
independently
of
the
end
of
the
span
and
then
like
on
the
back
end
tie,
you
know
actually
create
generate
the
span
from
those
two
events.
B
I
think
for
for
these
situations,
I
guess
we
need
concrete
use
cases
that
can
be
really
modeled
and.
D
Like
specific
examples,
right,
like
like
user
clicks
on
a
button-
and
it
sends
h
like
a
request
out
here
right.
B
B
I
I'm
not
aware
of
any
examples,
as
in
I'm
not
aware
of
anything
being
possible
to
detect.
B
We
we
know
that
you
know
you
like
every
page
navigation,
you
know
after
the
initial
page
load
is,
is
a
result
of
some
user
action.
Generally,
yes,
generally
yeah
yeah.
A
Yeah
I
mean
so
like
the
the
one
thing
that
I
know
that
that,
like
the
current
js
instrumentation
does
for
there's
that
user
user
interaction
plugin
and
so
it
tracks
like
clicks
or
certain
types
of
interactions
on
the
page
on
the
in
the
browser,
and
then
it
will.
A
A
So
it
treats
the
interaction
itself
as
a
span.
But
you
know
it's
in
its
interaction
like
once.
The
actual
ui
is
already
loaded.
A
So
the
initial
there's
also
that
initial
user
interface
or
page
load,
it's
also
in
an
interaction
that
that
I
think
is
represented
as
a
different
span,
but
anyway
yeah
so
like
there's
that
in
existing
instrumentation
does
treat
user
interactions
as
as
spans.
A
So
I
it's,
but
it's
it's
a
little
bit.
You
know
there
was
certain.
Obviously
there
was
a
decision
made
made
to
track
the
duration
of
the
span,
a
certain
way,
that's
very
specific
to
browsers.
So
I
don't
know
how
it
translates
to
mobile
devices
or
other
client
devices.
A
B
Yeah
duration
is,
I
think,
is,
is
less
important
than.
B
Are
you
saying
that
in
in
the
javascript
world,
in
the
browser
world,
it
is
possible
to
link
a
user
interaction
to
let's
say
a
network
activity.
A
It
is
yeah,
and
it's
by
it's
it's,
because
that
interaction
is
represented
as
a
span.
B
A
B
Okay,
if,
if,
if
that's
possible,
then
that
you
know
those
can
be
included
in
the
trace,
there
is
already,
but
if
it
is,
if
the
effect
is
a
network
call,
there
will
be
a
trace
anyway,
so
you
would
have
more
internal
spans
at
the
top.
B
Then
there
won't
be
an
existing.
There
won't
be
a
tray,
so
you
will
have
to
have
a
new
trace
for
that
purpose.
Right,
yeah
yeah!
You
would
okay
yeah
okay
number
one,
so
the
next
topic
was
a
session,
whether
we
need
a
data
model
for
a
session,
and
I
think
the
original
proposal
from
aws
you
know
suggested
having
a
session
object.
B
As
a
let's
say,
a
container
under
which
you
know
you
will
have
the
events,
the
different
events
on
on
the
client
side,
but
then
it
will
be
a
significant
change
to
the.
B
B
So
I
was
wondering:
do
we
really
need
another
object
for
the
session
itself
to
to
include
all
this,
the
properties
of
that
session,
but
again
in
a
trace?
We
only
include
a
trace
id
in
each
of
the
spans.
There
is
no
separate
trace,
object
right,
it's
kind
of
implicit,
so
I
I
don't
know
if,
for
a
session
you
you
would
want
to
include
anything
as
session
properties
or
are
just
derive
it
on
your
own.
From
from
the
different
events,
any
thoughts.
C
C
Whichever
so,
I'm
not
sure
that
there's
that
we
need
a
data
model,
but
we
do
or
should
at
least
explain
the
concept
of
a
session
and
what
the
id
represents
like
whether
like,
for
example
in
browser,
deserves
in
a
single
page
load
or
does
it
represent
a
more
time
oriented
interaction
from
the
same
user
or
whatever?
I
don't.
C
C
Okay,
okay,
I
think
I
think
that's
true,
although
there
is
also
there,
I
guess
there
is
also
that
maybe
this
again
isn't
data
model,
but
maybe
this
is
more
of
conventions
yes,
but
like,
for
example,
at
splunk
we
do
generate
a
new,
as
I
mentioned,
I
think,
last
week
or
two
weeks
ago
we
generated
a
new
span
at
the
beginning
of
a
session
and
when
the
session
id
changes
we
generated
a
span
for
that
as
well
so
and
that
actually
probably
would
be
a
better
model
as
an
event
on
it,
obviously
or
as
a
log,
a
log
entry.
C
So
that
is
something
that
I
think
could
be
useful
and
again
it's
not
really
a
data
model
for
a
session.
It's
more
just.
You
know
what
the
session
represents
and
what
instrumentation
should
do
about
sessions.
Okay,.
B
So
the
same
thing
you
know
said
in
a
different
way:
would
it
be
correct
to
say
that
if
let's
say
we
have
different
named
events,
there
are,
let's
say
semantic
conventions
for
naming
different
standard
events.
Then
we
could
have
one
name
for
a
session
as
an
event,
let's
say
to
indicate
the
start
of
a
session
or
an
end
of
the
session.
B
Although
there
is
a
risk
of
I
mean
at
this
point,
probably
there
aren't
any
like
I'm
wondering:
will
sampling
have
any
effect
of
these?
You
know
one-off
objects.
B
If
we
lose,
let's
say
a
session
start
or
a
session,
and
then
you
know
that
especially
the
sessions
start.
You
know
that
can
be
a
loss,
so
this
sampling
should
take
into
consideration
to
to
not
drop
certain
types.
C
D
C
The
error,
I
only
want
to
see
more
or
whatever
got
it,
so
that
is
going
to
mean
that
could
definitely
if
we
are
using
log,
the
logs,
open,
telemetry
logs
to
represent
rom
events,
we'll
need
to
think
about
what
sampling
means
or
what
it
should
mean
or
what
it
could
be.
So
I
yes
absolutely
I
agree
with
you
100.
Is
that
what
I'm
going
to
answer?
Okay,
good,
okay,.
C
That's
actually
probably
something
to
capture
up
in
the
in
item.
One
is
what
what
to
do
about
what's
what's
up
with
sampling
and
events
and
logs.
A
C
C
Like
we,
I
don't
think
we're
going
to
be
able
to
make
just
like
unilaterally
say.
This
is
what
you're
like
what
your
timeout
should
be,
or
whatever
there
could
be
some
somebody
who
wants
a
session
to
last
forever
as
long
as
the
browser
is
open
or
as
long
as
the
mobile
app
is
open,
but
I
think
at
least
writing
down
the
idea
that
this
is
something
that
is
likely
to
occur
and
then
modeling.
What
like
what
would
happen
at
the
end
of
the
session
or
what
would
happen
if
there,
when
a
session
id
rolls
over.
A
C
I
don't
know,
I
know
that
I
mean
just
speaking
for
splunk.
We
use
it
essentially
to
track
for
billing
purposes.
How
many
sessions
did
a
given
living
company
you
use
in
a
given
amount
of
time,
so
I
don't
we
are.
We
have
not
gotten
to
the
point
yet,
where
we're
talking
about
using
sessions
to
model
behavior
or
to
to
I
don't
even
know
to
rap
behavior.
But
it's
a
good
point.
It's
interesting.
If
that's
something
that
you've
seen
done.
C
Like
how
many
sessions
converted,
for
example,
like
how
many
sessions
did
people
actually
check
out
in
their
with
their
their
shopping,
cart
and
pay
money?
That
kind
of
idea.
C
Just
just
for
interest,
do
you
do
session
like
if
somebody
let's
say,
for
example,
you
have
a
web
page
where
you're
selling
stuff
and
they
do
check
out
and
they
go
through
the
payment
process?
Do
you
then
recommend
that
people
create
a
new
session
at
that
point,
or
could
they
have,
for
example,
four
or
five
checkout
processes
per
session
and
all
of
them
convert?
If
they're,
you
know
buying
things
here
and
buying
things
one
and
one
at
a
time.
A
Yeah,
so
I'm
not
sure
like
we
don't
actually
do
that
like
neurallic
doesn't
do
that.
But
you
know
this
is
more
kind
of
a
like
google
analytics
kind
of
thing,
but
you
know
I
think
the
topic
will
come
up.
A
I
think
people
want
to
tie
thai
like
the
performance
of
the
of
the
application
to
like
how
how
user
interact
with
it,
but
I
think
in
general,
like
I
mean
I
don't
know
if
I
had,
if
I
had
to
guess
like
you,
would
have
just
one
session
as
long
as
the
user,
like
is
continuing
spending
time
in
the
in
the
client
application,
but
I'm
not
sure.
A
I
guess
it
depends
like
what
the
what
the
user
of
the
sdk
or
of
the
instrumentation,
what
kind
of
questions
they
would
like
to
answer.
C
B
Okay,
so
the
the
next
one
is
about
the
difficulty
with
sending
with
the
browsers
sending
the
trace
parent
header
for
some
of
the
requests,
and
you
know
I
also
want
to
understand
you
know
the
purpose
of
the
trace
response.
Header
the.
B
So,
if
I
understand
correctly,
you
know
I'm
a
little
confused
here,
because
martin,
you
use
the
term
initial
page
load
right,
but
my
understanding
is
that
let's
say
you
go
to
you
type,
a
url,
page
loads
and
let's
say
the
page,
had
you
know
no
javascript,
just
anchor
tags
right
and
then
you
click
on
another
link.
Even
that
second
page
also,
you
know
the
browser,
won't
you,
you
won't
have
the
ability
to
send
the
transparent
header.
You
know
even
for
such
requests
right.
B
B
A
A
That
the
open
thermometer,
js
sdk,
has
a
plugin
for
that
they
called
document
load.
B
Later
right
in
response,
so
basically
they
get
the
so
that
is
that
is
where
the
idea
of
the
trace
response
comes
in.
Where
you
require
the
back
end,
so
basically
the
the
client,
the
browser
does
not
start
to
trace.
You
know
the
browser
makes
a
request.
Now
the
back
end
starts
a
trace
and
then
in
response
it
sends
the
trace
id
back
to
the
browser
indicating
this
is
the
span
id.
You
should
use
for
your
spam
because
I
have
said
this
as
my
parent
right.
B
Initially,
it
had
a
component
called
a
proposed
parent
id,
where
the
back
end
proposes
to
the
client
that
this
is
the
span
id
you
should
use
for
for
for
your
spam,
but
but
I
think
it
was
later
dropped,
and
so
I
I
wanted
to
understand
what
was
the
intention
of
introducing
this
response
header,
because
in
all
the
tickets
and
all
whatever
documentation
I
could
find,
I
know
there
was
no
indication
of
how
that
response
is
being
read
by
the
javascript
in
the
browser,
because
you
cannot
read
response
headers
right
in
javascript,
you
it
from.
B
On
those
pages
you
can
read
only
on
the
adjacent
fetch
calls.
If,
if
that
is
true,
then
how
is
this
trace
response
even
being
used.
A
Yeah,
I
think
I
don't
think
it
was
intended
just
for
browser.
I
think
it
was
intended
for
other
use
cases
where,
where,
like
you,
you
need
to
provide,
maybe
like
maybe
like
I'm
trying
to
remember
the
use
case
that
they
have
there.
B
Was
there
was
one
use.
D
B
It
can
indicate
that
to
its
parent,
you
know
to
its
downstream
service.
That
is
one
benefit
of
out
of
distress
response,
but
is
that
the
motivation
for
introducing
this
header,
I
think
that
morgan
do
you
know?
I
think
I
think
you
are
part
of.
D
I
was
I'm
not
as
active
in
the
I'm,
not
really
active
in
the
w3c
group
anymore,
but
the
the
there
were
two
purposes
for
the
trace
response,
header
that
were
very
similar
and
so
the
the
first
was
for
client-side
instrumentation,
okay
and
the
other
was
for
really
specific
to
cloud
providers,
and
it
was
so
that
if
they
generated
logs
or
something
in
in
their
actual
like
right
on
their
their
network
edge
for
their
load,
balancers,
for
example,
that
they
could
use
the
trace
response
and
then
append
the
right
trace
id
of
those
logs.
B
D
If
a
request
of
no
trace
id
came
in
like
this
was
these
were
conversations
between
google,
microsoft
and
amazon.
Actually,
just
really.
Google,
microsoft
and
the
whole
idea
was
like
on
google's
edge
or
microsoft's
edge.
Request
comes
in
to
some
api.
It
was
no
trace
request.
They
generate
some
logs
inside
of
their
load
balancers
inside
of
the
edge
there's
no
trace
id
to
correlate
those
logs
with
they
don't
want
to
generate
the
trace
id
from
from
scratch
on
the
edge,
because
I
don't
know,
maybe
there's
some
logic
and
how
it
needs
to
get
generated.
D
Maybe
they
need
the
sampling
decision,
for
example,
and
so
they
could
use
the
response
to
determine
if
they're
going
to
correlate
those
logs
of
a
trace
and
set
other
metadata.
C
D
It's
on
the
first
load
of
a
page
in
the
browser,
because,
theoretically
I
guess
the
server
would
have
to
embed
it,
and
they
don't
want
to
do
that.
So
this
is
really
for
the
js
that
loads
in
the
browser
it
needs
to.
I'm
trying
to
remember
the
exact
scenario,
something
like
daniel
dyla
would
know
better
than
me,
or
some
of
this.
B
It's
not
a
standard
header,
so
the
browser
is
not
going
to
initiate
generate
a
trace
parent
header
in
that
request.
Yeah.
B
Yeah
but
but
that's
what
I
think
I
clarified
with
martin,
that
it
is
not
only
for
the
initial
page,
but
there
are
different
types
of
requests.
The
browser
makes
right,
one
is
the
the
page
or
or
even
quality
document,
and
then
there
are
these
ajax
calls
right
so
for
ajax,
apparently
you
know
you
do
have
control
the
browser,
javascript
s,
control
to
to
insert
this
header
yeah.
So
it's
only
for
these
base
pages.
You
know
that
you
don't
have
controls
yeah.
D
So
I
think
there
may
have
been
a
third
scenario
for
trace
response
as
well,
and
that
was,
if
you're
sending
a
trace
across
some
boundary
to
an
organization,
that's
going
to
start
its
own
trace,
where
they
don't
want
to
trust
your
sampling
decision
or
your
id.
They
would
respond
if
the
id
they
had
so
that
in
some
group,
debugging
situation,
if
you've
ever
had
to
you,
could
link
both
of
those
up
together
but
again.
C
C
I
actually,
I
actually
think
that
is
very
relevant
for
this
group.
Okay,
like
for
an
android
app,
it
may
be
calling
six
apis
there
you
go
yeah.
Do
we?
Do
we
actually
want
that
android
app
to
be
like
sending
trace
parent
headers
to
all
those
apis
and,
like
you
said
well,
maybe
we
do.
I
don't
know,
but
then
should
we
also
then
be
looking
for
that
header
and
yeah.
I
don't
know,
do
something
with
that.
Probably.
C
If
but
if
this
kind
of
cross-organization
api
story,
which
is
going
to
be
very,
very
common
in
mobile,
yes,
like
do,
I
wonder
whether
we
want
to
for
mobile
apps
or
mobile
http
requests.
Just
say:
don't
don't
do
that
like
like
if
you
get
a
if
you
get
a
trace
like
this
trace
response,
header
back
use
that
to
stitch
things
together
or
in
some
way
it.
D
Presumably,
presumably,
if
I'm
a
developer,
I
make
let's
say
I'm
a
developer.
Who
makes
a
game,
I'm
I'm
rovio.
I
make
angry
birds,
even
though
I
know
that's
way
out
of
date.
These
days
and
let's
say,
I
add,
a
voice
over
ip
thing
so
that
any
I
don't
think
these
exist,
but
pretend
there's
a
multiplayer
angry
birds
mode.
D
I
think
it's
single
player
only
but
pretend
there's
multiplayer
and
I'm
playing
with
my
friends
such
that
we
can
voice
chat
each
other
and
that
voice
chat
goes
through
some
sas
service,
maybe
twilio,
even
though
I
think
twilio
usually
goes
up
the
phone
network
but
whatever,
and
so
the
calls
that
I
have
to
my
angry
birds
back
end.
I
want
to
use
the
same
trace
ids
throughout
my
whole
stack
right,
my
client
side,
traces
that
the
id
goes
all
the
way
down
through
all
my
back-end
surfaces,
all
the
way.
D
My
databases-
and
I
get
this
amazing
incredible
trace
of
everything
that
happens
in
a
user
interaction
and
life
is
good
for
my
calls
to
twilio
or
whatever
sas
providers
I'm
using.
They
have
their
own
tracing
system,
they're,
not
going
to
trust
my
ids,
because
maybe
I'm
going
to
use
the
same
id
for
every
request
right
unless
I'm
part
of
their
organization
they're
not
going
to
trust
the
ids
that
I
create,
because
I
might
goof
it
up
right,
we're
across
a
trust,
boundary
and
so
they're
going
to
use
their
own.
D
Well,
I
don't
have,
I
generally
wouldn't
have
their
internal
trace
ids,
but
now
I
do-
and
I
can
you
know,
on
the
phone
of
support
when
I'm
yelling
at
them.
I
can
say
well,
it's
trace
id
foo
and,
and
and
that's
that
was
stashed
inside
of
my
own
traces
somewhere,
I
mean
in
theory,
maybe
for
using
the
same
trace,
analytics
software.
They
could
stitch
those
together,
but
realistically
just
having
the
id
there
is
important.
So
then
I
can
just
say
at
least
like
yeah.
D
D
C
I
think-
and
this
may
be
a
common
term,
but
I
think
sir,
it
call
calls
this
calls
the
ones
that
you
trust.
Oh,
she
has
a
name
for
it.
It's
like
you're,
I
don't
know
where
it
was
anyway,
but
there's
there's
this
group
of
urls
that
you're,
probably
you
do,
want
to
send
trace
transparents
to,
but
then
there's
all
the
other
urls
that
you
probably
don't
want
to
be
sending
transparents
to
or
that
they're
not
gonna.
It's
not
gonna
be
helpful
and
might
in
fact
end
up
being
harmful.
D
D
I
mean,
which
is
why
I
think
of
the
w3c
spec.
It
says
if
you're
receiving
things
across
the
trust
boundary
do
not
like
log
the
trace
id
right,
but
do
not
do
not
propagate
it,
because,
yes,
like
you,
could,
for
example,
if
they're
just
blindly
trusting
them,
you
could
just
send
in
you
could
basically
dos
their
observability
system
by
sending,
in
a
whole
bunch
of
requests
the
same
trace
id
yeah,
yeah,
okay,.
B
Okay,
so
I'm
thinking
martin,
maybe
if
you're
interested
you
know,
we
could
join
one
of
the
js
meetings
and
get.
D
A
Yeah
I
want
to,
I
want
to
say
a
couple
things.
I've
just
been
listening,
but
the
for
the
you
said
santosh,
you
said
in
browser,
there's
the
initial
document
load
and
then
there's
the
ajax
calls
there's
also
other
resources,
like
images
and
css
and
scripts
themselves
that
those
you
don't
have
access
to
those
headers
either
correct,
correct
yeah.
I.
A
A
And
the
other
thing
that
I
wanted
to
say
is
that
you
know
sounds
like
there
are
lots
of
different
use
cases
actually,
but
are
like
the
immediate
like
use
case
or
need
that
we
had
at
least
a
neurologic
for
our
instrumentation
was
sampling
because,
like
you
can't
you
can't
make
a
sampling
decision
or
a
trace.
A
A
So
this
you
know
like
if
you
on
the
back
end
like,
if
you
could
did
your
you,
could
make
your
sampling
decision
on
the
back
end
and
just
throw
the
client.
Yes
collect
the
span
or
not
collect
the
span.
C
Yeah,
this
actually
ties
a
little
bit
back
to
the
session
question,
because
I
know
there
has
been
discussion
about
potentially
sampling
obsessions
and
would
would
we
want
to
introduce
the
idea
of
that
when
the
session
is
created
like
having
a
mobile,
app
or
whatever
be
able
to
decide
whether
to
say
this?
Is
you
know
in
my
one
percent
or
10
of
whatever
sessions?
I
actually
want
to
start
reporting
data
for
so
the
session
sampling
session
level
sampling
is
something
we
should
probably
at
least
put
on
the
list
of
stuff.
We
haven't
figured
out.
B
Yeah,
okay,
I
think.
C
A
A
B
Yeah,
it
actually
makes
sense
because
you
know
currently
you
know
for
individual
traces.
You
know
you.
Let's
say
you
have
a
sampling
flag,
but
the
problem
with
that
is
in
a
in
a
given
session.
You
know
it's
useful
to
either
have
all
the
different
traces
are
none
of
them
right.
You
know
if
you
have
some,
but
not
the
other.
D
B
You
know
so
if,
if
that
sampling
decision
on
the
trace
is
is
flowing
from
the
sampling
decision
of
the
session,
I
think
that
would
be
useful.
B
I
could
be
wrong,
but
it
could
at
least
be
an
optional,
could
be
an
option
if
somebody
wants
to
use
it.
C
B
B
Okay,
so
the
next
one
is
the
semantic
conventions.
I
think
there
are
a
lot
of
things
in
in
in
the
challenges
section
for
which
I
believe
the
answer
to
many
of
them
is
to
define
the
semantic
conventions
we
could.
B
And
then
you
know
reveal.
B
Yeah,
okay,
the
next
one
is,
I
have
put
the
like
john.
You
mentioned
that
the
ape
the,
although
the
proto
definition
you
know,
supports
the
binary
data
for
an
attribute
value,
but
not
the
api.
C
Will
be,
I
think
it
will
be
very
challenging
to
add
binary
data
to
the
to
api
span,
attributes
not
impossible,
but
you're
going
to
have
to
make
a
very,
very
strong,
a
very
a
very
strong
case
for
it.
But
I
will
say
the
log
model
I
believe,
has
binary
baked
in
right
now,
even
though
the
span
attributes
do
not
so
that
may
be,
it
may
be
an
option.
I
could
be
wrong
in
that,
but
I
believe
I
believe
log
attributes
are
more
open.
B
C
That
could
also
be
a
case.
Okay,.
C
A
C
C
Morgan,
I
thought
that
our
synthetics
product
did
something
with
browser.
Screenshots
didn't
it.
Oh.
D
Synthetic
synthetics
synthetics
is
the
it's
it's
the
browser
right
like
it's
taking
screenshots
of
itself
of
itself,
yeah,
not
not
of
this.
What
we're
talking
about
here
is
more
like
there's
a
really
popular
product.
D
I
think
we
usually
call
like
storybook
tell
a
story,
some
rum
like
product
that
gives
you
screenshots
and
basically
like
if
you
run
an
ecommerce
site,
for
example,
you
can
use
it
and
it'll
show
the
entire
flow
of
one
user
over
the
last
hour
that
led
up
to
the
purchasing
decision,
so
you
can
see
even
like
where
their
mouse
was
hovering
and
all
that
don't
get.
D
Is
the
ecommerce
site's
useful,
but
certainly
if
we
implement
this,
we
should
be
aware
that
people
will
have
legitimate
concerns
about
it
and
those
concerns
will
be
valid,
so
yeah,
that's
something
worth
debating.
C
B
Yeah
yeah
actually
browser
also.
It
is
probably
possible
because
I
know
I
I
was
in
a
support,
call
with
turbo
tax.
You
know
they
they
make
you
just
you
know
press
shift
twice
or
something
and
then
they
get
permission
and
then
they
can
see
your
screen
there.
You
go
yeah.
D
B
D
D
I'm
gonna
drop
jump
off,
but
please
don't
end
on
my
account.
Yeah.
B
Yeah
yeah
yeah.
No,
I
think,
there's
only
one
left
even
I
need
to
drop
five
minutes
before
10.
yeah,
so
the
last
one
john.
I
I
wanted
to
get
your
feedback
on
this
so
today
the
authentication,
if
I
understand
correctly,
is
available
only
in
the
collector,
like
only
the
collector,
has
some
extension
that
can,
you
know,
do
authentication
with
backex.
You
know,
but
the
agent,
the
the
clients
themselves,
of
whichever
language
you
know
they
probably
do
not
support
authentication
in
in
their
tlp.
B
Even
when
talking
to
any
backend,
including
collector,
is
that
is
that
understanding,
correct.
C
So
it's
totally
up
to
the
who
owns
the
back
end
right.
I
know
that
actually,
martin
probably
knows
new
relic
has
otlp
ingest
direct
direct
ingest
and
it
must
be
authenticated
in
some
way.
C
We
at
splunk
right
now
for
mobile,
we're
using
and
browser
actually
we're
using
zipkin,
and
we
do
authenticate
like
we
require
you
to
send
a
token
in
a
header
or
not
it's
actually
not
in
the
header,
whatever
form.
B
C
So
the
java
agent
allow
like
at
least
for
ota
for
open
telemetry,
allows
you
to
set
arbitrary
headers.
So
you
can,
you
can
add
headers
at
as
a
configuration
to
your
otlp
calls
so
that
it
could
support
authentication
header
based.
C
B
C
B
Okay,
with
with
java
agent,
it
is
possible
to
write
additional
extensions
right.
You
you
can
so
there
is
no
interface
for
an
auth
plugin.
C
A
B
So
for
server
side
there
is
an
there
is
at
least
an
option
available
where
you
can
put
the
collector
in
your
network.
So
your
java
agent
talks
to
the
collector
in
without
authentication.
B
The
collector
has
an
authentication
plugin
to
to
to
make
sure
you
know
it
authenticates
with
your
back
end
in
the
cloud
before
it
can
send.
You
know
the
the
the
data,
whereas
for
the
client
side
there
is,
there
is
no
option,
you
know
where
do
you
put
the
collector,
so
the
clients
themselves,
you
know,
must
support
authentication.
C
B
Then
it's
also,
you
know
who
owns
it
right,
who's
responsible
in
in
the
case
of
java
agent.
Let's
say
the
customer:
will
you
know,
deploy
the
collector.
You
know
in
the
same
network
as
your
backend
app
is
running
on.
C
B
D
B
D
B
D
B
Again,
you
know,
given
that
you
know
different
companies.
You
know
we
might
be
using
different
ways
of
authentication.
So
therefore,
the
way
the
collector
has
a
plugin
for
authentication,
you
know
we
should
have
a
plugin
for
an
extension
plugin
whatever
for
for
authentication
in
in
all
the.
C
B
So
either
we
standardize
on
on
what
that
token
is,
even
if
it's
it's
public
we
and-
and
I
think
it's
going
to
be
unlikely-
that
we
are
going
to
use
the
same
mechanism.
B
In
that
case,
you
know
you
just
leave
it
as
a
as
a
plug-in
and
then
leave
it
to
the
individual
implementations,
but
either
ways
you
know
you
need
some
at
least
a
fictitious.
You
know,
authentication
or
header.
B
A
Yeah,
so
maybe
I
will,
I
will
ping
them
okay,
once
it's
changed
with
the
calendar,
I
will.
I
will
thank
them
and
and
daniel
and
like
whoever
else
is
so
I've
attended
that
sick
few
times,
so
I
can
invite
them
to
the
to
the
half
hour
and
okay
yeah.
I
also
feel
like
like
there's.
Probably
it
probably
sounds
to
me
like
we
want
to
define
use
cases
like
more
specific
use,
cases
of
kind
of
describing
like
what
this
is
going
to
look
like
and
then
yeah.
A
I
you
know,
there's
still
like
there's
a
lot
of
questions
in
my
mind.
So
do
we
do
we
have
like
different
people,
propose
different
things
and
then
just
like
discuss
the
different
ideas
or
how
would
we
move
forward
with
making
decisions.
B
I
don't
know
one
thing
I
can
think
of
is
to
be
more.
You
know
specific
in
what
our
proposal
is,
because
there
are,
some
of
these
could
be
just
ideas,
yeah
and
then
vague.
So
if
you
could
translate
into
you
know,
okay
here
is
the
change
to
let's
say
an
api
or
a
spec,
or
you
know
what
what
change?
What
pr
you
know
you
you
want
to
raise.
C
Probably
the
first,
the
the
first
thing
we
want
to
do
before
we
get
to
the
spec
is
to
write
up
a
fairly
comprehensive
otep.
That
kind
of
has
the
plan
in
it
and
all
of
the
things
that
we're
intending
to
write
and
write
as
a
part
of
the
specification
so
like
have
like
what
is
what
is.
Why
are
we
doing
this?
What
are
the
use
cases?
What
are
what
are
the
you
know?
What
is
the
plan?
B
Okay,
but
do
you
think
we
should
wait
for
at
least
some
discussion
from
the
js
team
and
android
diverse
teams,
and
once
we
have
at
least
some
basic
agreement,
then
we
could
filter
out
the
necessary
ones
and
then
put
that
in
the
totem.
C
A
Yes,
maybe
next
time
we
should,
we
should
go
through
all
these
things
and
maybe
prioritize
like
which
things
you
want
to
tackle.
First,
like
maybe
the
next
few
months,
and
then
you
know
some
things
that
can
be
done
later.
Okay,
all
right.