►
From YouTube: 2022-02-16 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).
B
C
A
First
of
all,
I
wanted
to
just
touch
base
follow
up
on
our
conversation
from
last
week
when
we
talked
about
the
proposed
data
model,
and
I
wanted
to
see
if
folks
have
a
chance
to
think
about
it,
then
if
there
were
any
more
thoughts
or
comments.
B
I
think
I
added
some
notes
to
it,
but
I
don't
think
it's
necessary
to
go
over
here.
Maybe.
A
A
Okay
and
then
I
also
I
know
that
stefan
had
a
question
about
other
types
of
events
in
mobile
like
so
I'm
not
familiar
with
mobile,
so
I'm
kind
of
relying
on
on
folks
who
are
experts
here,
but
like
view
activity,
life
cycle
events
and
app
lifecycle,
events,
I'm
assuming
they
would
also
be
better
represented
as
logs
than
spans.
D
I'm
pretty
sure
there
are
some
aspects
of
the
activity
life
cycle
that
make
a
little
bit
more
senses
certain
traces
so
like,
for
example,
app
start
or
even
view
rendering
tends
to
have
sort
of
multiple
things
that
happen,
which
are
all
related
to
that
app
start
or
to
that
view,
render
right
so
part
of
the
activity
life
cycle.
D
If
I'm
not
mistaken,
there
is,
you
know,
there's
a
view
created,
there's
another
view
destroyed,
that's
happening
simultaneously,
but
they're
all
related
to
that
end,
user
experience
that
is
sort
of
rendering
that
view
and
they're
all
they're
kind
of
happening
in
parallel,
but
they
do
result
in
that
same
broader
time
for
the
view
or
the
app
start
to
render
fully.
Those
are
the
ones
that
I
I
think
is
definitely
some
that
there's
a
causal
plus
like
a
time
element
for
them,
which
is,
I
don't
think,
can
be
fully
represented
in
a
log.
B
Yeah,
I
think
the
the
main
issue
that
I
have
with
putting
those
sorts
of
activities
under
a
span
is
that
it's
really
like
they're
kind
of
you
know
like
you
could
put
a
span
start
and
like
if
you
did
load
or
spam
start
and
in
view
did
appear
and
end
it
in
that,
like
you
know,
narrow
function,
but
the
problem
that
I
always
kind
of
run
into
is
how
do
you
like
to
find
the
broader
kind
of
like
review
load
period
and
like
how
do
you
define
like
when
to
stop
it
and
like
what
you
know
like
if
you
don't
have
customer
triggered
like
behaviors
such
as
like
network
traffic
and
stuff,
like
that,
then
it
kind
of
misses
the
misses.
B
The
broader,
like
activities
of
loading,
a
view
like
you
need
to
go,
make
that
network
request,
but
unless,
like
you,
can
somehow
tie
that
to
the
load
activity,
then
it
kind
of
gets
disassociated
with
the
span.
I
don't
I
don't
know
like
it's.
It
could
go
either
way,
but
it
just
seems
like
there's
a
lot
of
oh
there's.
A
lot
left
like
in
terms
of
actual
getting
stuff
captured
in
just
a
span.
C
At
least
on
android,
it's
very
easy
to
capture
the
timing
of
every
step
of
the
the
activity
and
fragment
creation
and
destruction
process.
So
it
feels
like
it
would
be
a
bummer
to
lose
that
timing
information,
because
it
is
very
easy
to
capture
and
can
be
very
important
if,
like
if
a
user,
if
a
developer
has
put
in
some
very
expensive
operation
on
them,
that's
running
on
the
main
thread
during
part
of
their
application.
Like
the
activity
loading
life
cycle,
it'll
really
easily
show
up
as
a
slow
span
in
there.
F
A
E
A
E
C
E
D
I'm
pretty
sure
since
android
8,
so
basically
I
think
that's
api
level,
26
27,
I'm
sorry,
I
I
forget
the
number
of
mappings,
but
I
think
there
were
like
since
that
time
we've
there's
a
lot
of
interesting
information.
You
can
get.
You
may
not
be
able
to
get
it
all
the
way
back
to
let's
say
android,
four
or
five.
If
I'm
not
mistaken,.
G
B
So
I
just
I
haven't
been
able
to
look
at
the
log
api
all
that
closely.
Is
there
not
any
way
to
add
like
an
attribute
to
a
log
at
all,
or
is
it
just
a
pure
like
string
of
text?
That's
all
you.
C
B
B
If
the
span
does
not
complete
in
the
cases
of
like
you
know,
app
crashing,
or
you
know,
application
background
that
sort
of
stuff,
I
I
kind
of
feel
like
it
would
be
safer
to
use
a
log
for
these
sorts
of,
like
you
know,
view
did
load
kind
of
event
and
then
create
like
a
or
you
know
create
a
log
for
the
log
for
this.
This
sort
of
action
that
causes
other
things
to
happen
add
an
id
to
it
and
then
attribute
that
id
to
like
other
stuff
that
happens
nearby.
A
So
the
log
the
log
data
model
has
has
like,
like
a
trace
id
on
it
or
and
parent
id,
so
you
can
link
it
to
a
spam.
So
I
you
know,
like
the
the
idea.
Is
that,
like
you,
could
you
could
always
capture
logs?
A
And
you
know
if,
if
like
the,
if
you're
able
to
to
generate
a
span
as
well,
that
like
kind
of
represents
the
activity
like
a
duration
of
the
activity
and
links
to
other
work,
type
of
work
that
you
want
to
capture,
you
can
do
that.
You
know,
but
but
it's
optional
like
the
it
sounds
like.
Maybe
some
some
sdks
might
be
able
to
do
that.
Some
sdks
may
not.
A
But
for
like
a
session
replay
kind
of
thing
like
you
would
always
always
get
the
logs,
so
you
could
you
would
at
least
get
the
this,
the
sequence
of
events
that
happened
in
it.
You
know
at
the
very
minimum.
A
Okay
tonight,
the
next
thing
that
I
want
to
talk
just
quickly
talk
about
is
we.
When
we
talked
to
ted
last
week,
I
think
we
talked
about
starting
a
poc.
I
I
have
talked
to
martin
about
maybe
collaborating
on
the
poc
for
browser.
If
anyone.
E
A
Is
interested
in
in
that
like,
let
me
know
we
can
we
can
pull
you
in
into
that
effort.
I
also
would
like
to
ask
like
if
anyone
would
be
interested
in
or
able
to
work
on
a
poc
for
mobile,
so
I
think
we
need
to
have
that
as
well.
We
should
have
that
as
well.
B
I
don't
think
I'm
gonna
be
able
to
have
the
time
coming
up
here.
Pretty
soon
I've
got
a
my
wife's
due
date
is
like
next
monday.
A
And
then
this
is
probably
a
question
more
question
for
ted
because
he
suggested
the
poc,
maybe
maybe
morgan,
you
might
know
if
we
open
an
otap.
How?
What
should
the
poc,
like
show
like
in
this
particular
particular
case,
like
we're
talking
about
just
generating
different
types
of
data
so
like
we
can
imagine
that
this
data
can
be
generated,
but
like
should
we?
Should
the
poc
be
more
focused
on
the
api
or.
F
Yeah,
john,
you
might
john
watson,
probably
actually
has
more
expense
experience
pushing
an
oteps
than
I
do
so.
John
even
opinion.
F
Okay,
never
mind,
I
mean
I
mean
it's,
it's
really.
It's
gonna
be
a
few
things.
It's
gonna
be
videotep,
it's
gonna,
be
the.
I
mean
it's
gonna
be
sort
of
a
summary
of
our
learnings,
but
I
suspect,
like
the
biggest
questions,
that
this
will
answer,
the
ones
that
we'll
put
in
the
hotep
will
be
the
ones
that
we're
talking
about
right
now,
right
like
when
you
use
logs
when
you
spans,
what's
semantic
expansions
to
use
things
like
that,
but
we
might
want
like.
I
am.
F
I
work
on
specs
and
things,
I'm
not
usually
the
one
who
writes
houston
hoteps,
so
we
may
just
want
ted
or
someone
else
to
try
to.
Unfortunately,
I
don't
think
he's
on
the
call.
F
A
Okay-
and
the
last
thing
I
had
on
the
agenda
was
I
wanted
to
circle
back
to
the
discussion
we
had
about
identifying
client-side
apps.
A
It
sounds
like
the
app
name.
Space
is
very
contentious
like
there's
and
it
might
be
difficult
to
get
an
agreement
on
that.
So
at
least
for
now
I
would
have
done
is
separated
adding
the
browser
attributes
into
its
own
pr.
A
So
that
can
proceed
so
take
a
look
at
this
pr
and
then
I
also
wanted
to
just
gotta
get
a
sanitize
check
from
everyone.
A
For
do,
we
do
like
different
vendors
need,
have
the
need
to
actually
identify
like
browser
and
mobile
separately
for
different
processing
and
ui
purposes.
E
A
Yeah,
what
I
meant
more
specifically
is
like
when
we
talked
about.
We
talked
about
client-side
telemetry,
but
I
think
that
there
are
really
two
different
types
on
the
on.
The
back
ends
like
there
is
we're.
Gonna,
have
a
different
processing
and
different
analysis
for
browser
different
and
different
from
mobile
applications,
at
least
that's
what
neuralic
has,
and
I
just
wanted
to
so
when
we
talk
about
identifying
type
I
just
wanted
to
just
I
pretty
much
know
what
the
answer
is,
but
I
just
wanted
to
get
everybody's
like
yes
like
this.
C
Yeah,
I
think
it's
splunk.
We
do
that
via.
I
think
it's
remember,
there's
a
plat,
an
attribute
that
describes
the
platform,
basically
android,
ios
or
or
browser.
I
don't
remember
exactly
how
it
works,
but
we
definitely
need
something
to
distinguish
between
those
things,
because
the
analysis
pipeline
needs
to
do
different,
different,
generate
different
kinds
of
metrics
or
different
kinds
of
information,
depending
on
where
the
data
is
coming
from
for
sure.
A
Yeah,
okay,
so
I
think
for
browser
like
just
the
presence
of
the
browser
attributes
would
probably
be
enough
for
for
mobile.
I
wasn't
sure
because
we
talked
about
device,
but
device
could
potentially
you
know
be
other
things,
not
just
mobile
devices,
so
you
know
I
I
think
at
that
point.
You'd
have
to
not
just
look
for
the
presence
of
device
attributes,
but
also
look
at
the
value
of
of
those
attributes.
Yeah.
C
That
I
do
not
know,
but
they
certainly
send
a
browser,
an
attribute
about
what
browser
is
running
yeah,
so
there's,
but
there
is
actually.
I
think
I
don't
remember
who
somebody
just
brought
up
a
really
good,
interesting
and
relatively
complex
case
that
we
can't
ignore,
and
that
is
embedded
web
views
in
mobile
apps.
C
It's
actually
there's
there's
interesting,
it's
an
interesting
case
because
you
kind
of
have
a
an
application.
Now
that
is
both
a
mobile
and
a
web
application,
and
it's
really
nice
to
be
able
to
stitch
those
things
together
under
a
common
session
id.
C
And
only
if
you
have
ways
to
communicate
spans
down
into
or
context
down,
I
don't
know
that
at
least
on
android.
I
don't
think
there's
any
way
to
to
communicate
something
as
complex
as
context
between
the
the
native
and
the
webview.
You
basically
get
some
strengths.
B
Yeah
I
know
that
like
in
the
react
native
platforms
like
there
is
a
bridge
between
the
two
that
you
can
communicate,
but
maybe
in
things
like
cordova,
where
it's
just
a
pure
webview,
it
might
not
necessarily
matter
because
there's
not
really
a
lot
of
os.
C
Even
in
the
android
webview,
you
don't
get
a
lot
of
like
the
standard,
android
webview.
You
don't
get
a
lot
of
communication
channels
to
get
that
information
like
you
can
get
some
data,
but
there's
no
function,
and
I
don't
know
if
yeah.
E
C
C
H
C
I
mean
that's
what
we
that's,
what
we
do
it's
in
our
instrumentation.
Is
we
inject
some?
We
inject
some
data
that
can
be
picked
up,
but
I
think
the
way
it's
done
in
ios
is
completely
different
than
the
way
it's
done
on
android,
because
the
communication
protocols
are
between
native
and
web
are
very
different
in
those
spaces
like
what
you
can
do
like,
I
think,
on
ios
you
only
get
function
and
on
android
you
only
get
data
so
figuring
out.
A
common
api
is
a
little
squirrely.
E
G
So
there
seems
to
be
two
problems
right.
I
think
one
is
to
identify
that
we
are
in
that
situation
of
embedded
view
inside
an
app
second
is
to
be
able
to
share
this
session,
so
the
session
sharing
might
be
a
complex
topic,
but
what
about
the
identification
that
you
know
it
is
an
embedded
situation.
G
How
do
we
want
to
represent
them
through
the
attributes
and-
and
I
also
want
to
ask
you
know-
should
we
define
in
the
spec
that,
if
just
the
presence
of
any
browser
dot
something
means
it's
a
browser
and
any
device
dot?
Something
means
it's
a
it's
a
mobile.
I
mean
it
could
be
a
mobile
app,
but
what,
if
should
the
spec
also
say
that
you
know
these
should
not
be
present
simultaneously.
G
Yeah
and
in
the
document
that
ted
shared
at
the
very
bottom
in
one
of
his
screenshots,
there
was
an
attribute
called
client.platform.
It's
new
to
me,
but
client,
client,
yeah,
client.platform,.
G
So
the
client
name
space-
I
don't
know
I
have
never
seen
before.
G
If,
in
case,
you
know
that
is
any
relevant
here,
which
which
documents
are
you
referring
to
the
document
where
he
talks
about
why
we
should
just
stick
to
service
dot,
name,
okay,.
G
G
Yeah,
I
I
think,
let's
maybe
think,
over
the
embedded
case
and
once
there
is
some
minimum
agreement.
You
know
clarity
on
that.
Maybe
we
could
just
define
you
know
some
rules
and
and
then
see
what
reactions
you
get.
Let's
say
we
can
say
that
if
it's
a
browser,
you
know
you
only
can
have
browser
name
space,
but
no
device
name
space.
If
there
is
then
it's
an
error
right.
G
If,
if
it's
a
mobile
app,
it
has
to
have
mobile
sorry
device
name
space,
but
no
browser
name
space,
but
if
it
has
both
and
it's
an
embedded
case.
G
A
I
mean
that
probably
we
can
define
as
a
as
a
new
type
of
attribute.
Okay,
we
still
need
to
talk
about
like
client-side
specific
attributes,
like
the
bundle
name
or
we
haven't
resolved
that
yet
so.
A
I
wanted
to
really
quick
like
circle
back
to
the
os
name.
Attribute
is,
is
like
these.
Do
you
is
there?
Do
you
think,
there's
a
need
to
identify
like
mobile
phones
and
tablets
from
other
and
like
from
other
devices
such
as,
like
you
know,
I
don't
know
like
embedded
devices
or
you
know
like
android
is
used
in
many
different
places.
So
like
is
there
a
need
to
identify
that
it's
like
a
mobile
mobile
phone.
D
No
os
name
is
just
it's
just
android
or
ios.
So
but.
D
World
it
might
be
ipad
os
and
ios,
but
as
far
as
I
know
the
android
side,
it
would
be
the
same
today.
Sorry,
that's!
What
we've
explored
versions
might
be
different,
flavors
or
android
might
be
there,
but
I'm
not
sure
if
there
is
a
way
to
pull
that
out
like
if
you
want
to
do
oxygen,
os
versus
android
os,
or
something
like
that.
D
D
Like
your
os
can
be,
it
could
be
on
a
tablet.
It
could
be
on
a
phone,
it
could
be
on.
Something
else,
too,
is
that
motor
device
type
attribute,
or
something
like
that.
B
A
So
I
mean,
I
guess
like
what
I'm
trying
to
maybe
like
also
figure
out
is:
is
there
a
need
to
define
from
from
like
the
and
you
from
the
the
application
owner
perspective
like
define
like
that?
This
is
a
like
a
client-side
mobile
application.
A
You
know
like,
would
there
be
a
different?
Would
there
be
a
need
on
the
back
end,
I
guess
to
like
distinguish
like
a
mobile
application
for
client-side
application
from
maybe
I
don't
know
like
if
there's
anything
else
that
can
run
like
any
other
type
of
application
that
can
run
on
mobile
devices
that
you
would
might
want
to
capture
telemetry
from.
B
A
Okay,
yeah,
I'm
just
you
know,
I'm
maybe
like
kind
of
my
questions,
but
I'm
just
trying
to
like
suss
out
like
if,
if
the
os
attribute
and
device
is
good
enough,
essentially.
E
H
G
But
it
could
be
a
standard
way
of
doing
this.
You
know
across
multiple
platforms.
Maybe
it
will
be
good
to
define
that,
like.
Let's
assume
that
you
know
the
back
end
is
the
authority
for
the
time
and
and
then
how
do
the
clients
synchronize
with
that?
So
only
only
thing
is
you
know
it
requires
at
least
one
communication
with
the
back
end
to
to
identify
what
is
the
difference.
B
As
you
can
see
on
on
the
in
swift
or
in
ios,
we
call
this
the
candy
crush
issue
where
people
will
like
change.
Their
phones
dates
to
get
like
things
unlocked
in
candy
crush.
But
we
solved
this
by
just
creating
like
an
asynchronous
clock
that
that
talks
to
some
external
authority,
because
the
yeah
the
devices
clock
can't
be
trusted.
C
C
So,
on
java,
for
example,
there's
a
system.nanotime
which
is
guaranteed
to
always
increase
with
the
correct
nanos.
Even
if
the
wall
clock
drifts
back
and
forth,
it's
a
it
is
monochronically
accurate.
You
can't
take
that
nano
time
as
like
being
meaningful
in
and
of
itself,
but
it
should
be
exactly
meaningful
for
measuring
differences
between
two
calls
to
the
nanotime.
C
Well,
that
is
a
requirement,
but
if
you're
using
the
wall
clock
to
be
measuring,
start
time
and
end
time,
you're
going
to
get
who
knows
what
you're
going
to
get,
especially
if
someone
you
know
messes
with
their
messes
with
their
clock
like
price
was
saying,
but
if
you,
if
you
just
basically
at
startup
time,
at
least
this
is
what
we
do
on
android
for
splunk
is
at
startup
time.
You
just
record
a
a
one
wall
clock
time,
and
then
everything
should
be
nanos
from
that
point
on
to
do
durations.
A
B
Yeah
that
works
really
well,
if
you're
only
looking
at
the
things
that
happen
on
the
device,
but
once
you
start
bringing
in
distributed
tracing,
it
gets
pretty
pretty
ugly
because
there's
a
lot
of
you
know
the
clock
or
the
clock
on
the
device
can
be
pretty
off
compared
to
like.
What's
on
the
server
so.
A
Do
you
have
do
you
have
a
link
by
chance,
martin
to
that
issue
or
pr
that
you
mentioned
yeah,
I'm
trying.
G
Yeah,
I
think
this
will
be
good
to
be
part
of
the
specs
so
that
all
the
different
services
in
a
distributed
trace.
They
use
the
same
mechanism
for
handling
the
time
drift.
A
E
G
Yeah
yeah
not
yet,
but
at
least
for
our
requirement
it.
It
is
sufficient
that
there
is
an
ability
to
set
a
header,
and
our
plan
is
to
just
set
the
the
the
authorization
header
with
the
better
token.
G
But
I
I
see
there
is
a
another
github
issue
on
the
spec
about
having
for
the
expert
authentication.
I
I
don't
know
what
they
have
in
mind.
I
think
they
they're
talking
about
aligning
the
interfaces
with
the
same
that's
available
in
the
collector.
G
G
In
in
the
splunk's
case,
I
think
last
time
we
saw
you
know
you
just
pass
an
auth
as
the
url
parameter
right,
so
is
that.
G
G
A
Okay
sounds
good,
then
sounds
like
we're
done.
We
can
take
back
15
minutes.
I
linked
the
issues
also
in
this
link.