►
From YouTube: 2022-11-03 meeting
Description
Open cncf-opentelemetry-meeting-3@cncf.io's Personal Meeting Room
A
B
B
B
I,
don't
know
if
you
got
introduced
before
hi,
hey
hello,.
C
C
Yeah
I'm
Stephen
from
app
Dynamics
as
well.
It's
been
a
while.
E
Great
yeah
yeah
with
the
the
yeah
he
he
contacted
like.
E
Around
that,
because
they
were
interested
in
in
working
on
print,
the
limited
with
Swift
and
yeah
I
I
hope
you
I.
A
E
A
bit
late
he's
in
the
channel
That,
probably
half
past
five
I
mean
in
half
an
hour
I
mean
because
he
had
a
another
meeting.
Oh.
D
E
Okay,
so
okay.
D
Okay,
so
I
want
to
okay,
you
folks
can
complete
your
agenda
first
and
then
I
I
will
talk
about
the
events.
E
A
E
Sorry,
do
you
know
the
document
if
you
have
any
topic
apart
from
this.
D
E
E
D
Yeah
so
yeah,
okay,
I
see
okay.
A
C
E
D
Okay,
all
right,
let
me
share
my
screen.
I
want
to
share
something.
D
Yeah
yeah
no
I,
it
will
do
automatically
yeah,
okay,
okay,
so
this
is
rum.
Events
schema
for
browsers
that
we
in
an
exercise
we
did
with
the
browser
team.
D
D
And
and
then
the
First
Column
we
identified,
we
basically
discussed
like
there
are
some
attributes
that
some
are
some
vendors.
Don't
have
some
attributes
that
all
of
us
have
so
then
we
identified
okay.
What
what
are
the
in
attributes
that
we
are
finally
interested
in
that
that
we
expect
to
have
you
know-
and
this
is
what
we
identified
in
the
first
column
and
then
eventually,
we
there's
another
document.
D
It's
still
loading
yeah,
so
so
so
these
are
the
you
know,
final
set
of
attributes
for
each
events
we
have
identified-
and
this
is
what
we
are
planning
to
you
know,
publish
to
ourselves.
Saying:
okay,
you
know
run
browser.
Rum
is
going
to
emit
different
events.
D
D
So
that
is
the
exercise
you
know
we
have
been
doing
since
last
couple
of
months
and
I'm
I'm
wondering
if
we
could
do
something
similar
for
the
mobile.
Also,
although
in
Mobile
there
is
one
more
Dimension,
it's
not
just
across
multiple
vendors.
D
Even
you
know,
we
want
to
standardize
across
IOS
and
Android
if
possible
as
much
as
possible
right
so
I
could
start
a
document
with
this
and
I
think.
To
begin
with,
we
could
simply
start
with
the
most
basic
events
that
we're
interested
in
and
I
I
can
think
of
a
couple
which
is
the
error
event.
Like
you
know,
your
application,
you
know,
got
an
arrayed,
it
didn't
crash,
it's
just
an
error
and
that
error
is
not
necessarily.
You
know
during
a
network
request
right.
D
E
Yeah
I
think
that
those
are
scenarios
that
are
currently
solved
by
using
locks,
probably
yeah
living
an
error
that
happens
in
the
app
nothing
like
that.
Yeah.
D
Yeah,
so
that
is
this,
you
know
event
we're
calling
it
exception,
because
open
Telemetry
spans,
you
know
they
already
call
it
exception.
All
the
error
would
have
been
more
appropriate.
D
The
second
event
is
that,
and
third
is
the
Crush
right
so
for
a
so
network
change
event
is
specific
to
mobiles.
You
know
the
browsers,
you
know,
cannot
detect
Network
change
unfortunate.
A
D
D
And
the
third
one
is
the
cache
crashes
and
then
for
the
crash,
so
Stephen
good
that
you're
here
I
think
last
week,
nachos
said
that
the
like
they
also
do
crash
analytics.
But
in
their
case
the
size
of
the
crash
dump
is,
is
not
big.
It's
in
a
few
kilobytes
or
a
few
hundred
kilobytes,
but
not
in
megabytes.
What
so
we
use
the
pl
crash
reporter
right
from
Microsoft
and
nacho.
Do
you
guys
use
any
third
party.
E
Yeah
we
we
currently
use
PL
class
reporter
also,
okay
and
I
have
also
used
in
the
past
KS
crash,
that's
very
similar
to
PL
Crush
reporter
and.
D
E
Okay,
KS
crash
K
is
KS
case
crash
yeah.
That's
fine,
yeah
I
used
that
in
the
past,
before
joining
a
data.
Okay
in
my
previous
company
that
was
acquired
by
data,
we
used
KS
crash
because
it
was
also
a
good
crash
reporter
doing
very
similar
things
to
PL
cross
reporters.
So,
okay,
but
there
is
no
one,
no
no
company
behind
it.
So
it
was
more
like
open
source
thing.
E
E
The
advantage
of
this
versus
real
reporter
I
would
say
is
that
it
gives
a
structured
error
that
you
can
consume
in
a
structured
manner,
while
peel
cross
reporter
is
more
oriented
to
giving
a
upper
class
like
reports
and
it
and
it's
less
configurable
but
yeah
both
are
I,
think
those
are
the
more
the
stuff
for
crash
report
and
I
also
think
that
Sentry
uses
a
version
of
this.
D
Yeah,
what
we
could
do
is
we
could
design
our
crash
event
in
such
a
way
that
the
users,
the
customers,
you
know-
can
configure
what
they
want.
E
Yeah
well,
one
of
the
things
I
am
concerned
about
harassing.
Is
that
specifically
on
iOS
I,
don't
know
in
Android?
Probably
it's
very
similar
I
am
not
an
expert
in
that
okay,
yeah.
E
This
discussion,
okay,
the
thing
is
about
the
size
of
the
I
mean
usually
what
you're
getting
across
in
a
production
app
is
an
unsymbolicated
cross.
Okay,
so
you
only
get
addresses
of
the
functions
and
addresses
of
the
class
points.
You
don't
get
even
the
name
of
the
module
or
things
like
that.
You.
E
And
that's
not
very
useful
and
that's
what
you
get
until
you
symbolicate
that
so
for
the
crust
to
be
useful,
I
mean
the
same
way.
I
say
that
it
works
just
a
small
kilobyte
or
just
a
100.
E
String,
it's
also
not
very
useful,
just
with
the
string
and
for
symbolicit
in
that
unit,
quite
a
lot
of
quite
a
lot
of
infrastructure.
Okay,
so
you
need
that
this,
the
symbol
files,
for
example,
that
somehow
the
developer
must
provide
to
your
platform.
So
you
can
symbolicate
that,
and
things
like
that,
so
I
I
have.
A
E
Have
doubts
about
how
would
what
will
mean
supporting
process
in
in
you
know,
pen
Telemetry,
to
be
something
you
really
useful,
I,
don't
know
what?
What
do
you
think
is
different.
C
Well,
I
I
I
agree.
Obviously,
symbolication
is
required
or
it's
useless.
We
do
that
now
in
our
existing
non-uh
open
Telemetry
product,
in
that
we
capture
the
the
symbol
piles
and
we
do
a
bunch
of
work
on
the
with
the
data
from
PL
crash
reporter
on
the
device
before
sending
it
up
to
our
back
end
and
then
the
back
end
has
a
large
set
of
processes
that
do
the
symbolication
and
make
something
useful
out
of
the
crash
report.
C
C
Is
a
is
a
complex,
complex
issue
and
it
gets
you
in
some
ways:
it's
more
complex
than
Android
as
well,
given
the
profusion
of
deoxification
tools
that
they
use,
you've
got
to
deal
with
those
as
well,
that's
less
common
than
iOS,
but
but
you
certainly
deal
with.
C
For
the
most
part,
for
the
most
part,
it's
just
those
two
but
again
there's
something
it's
something
else.
One
needs
to
deal
with.
One
needs
to
get
the
the
the
files.
I
can't
remember
the
name
of
them.
C
I
can't
remember
the
file,
but
but
yes,
you've
got
to
deal
with
that
as
well.
So
so
there
is
a
complication
there
and
it
is
not
a
client-side
only
issue.
E
Yeah
I
mean,
but
you
can
send
it
as
it
says,
modify,
but
that
file
is
not
it's
almost
not
useful.
If
you
don't
have
the
simple
files
that
are
usually
megabyte
size,
100.
E
Size
depending
on
the
size
of
the
app
so
yeah
I,
don't
maybe
we
can
send
still
the
grass
as
unsymbolicated
with
this
open
Telemetry,
but
that
won't
be
a
useful
signal
without
preprocessing.
First
yeah.
C
There's
a
there'd
be
a
back-end
process
that
would
have
to
exist
between
the
developer
and
the
vendor
in
order
to
get
the
symbol
files
to
where
they
need
to
be
in
order
to
process
the
symbolication
at
the
end,
so
that
it's
clearly
not
a
solution
that
can
be
solely
in
this
code.
You
know
inside
yeah.
E
E
Because
I
am
doing
in
process
because
I'm
not
running
on
device,
but
that
that's
that's
an
exception.
Usually
you
don't
you
don't
have
any
any
value
on
the
on
the
crosses
without
without
the
symbols.
So
so
that
was
my
concern
about
this
signal
of
the
crisis
I
mean
having
across
is
important.
Maybe
it
can
be
as
a
signal,
but
it
won't
be
useful
until
a
back-end
process
happens.
D
No
I
think
all
the
vendors
who
are
in
this
business
I
think
they
understand
right,
I,
think
what
you're
talking
like
I
I
do
not
I'm
new
to
iOS,
but
Stephen
has
been
with
that
for
a
long
time
and
he
fully
understands
so
I'm.
Assuming
the
same
with
you
folks,
people
who
are
working
on
the
highways.
You
know
you
understand
that
I
think
in
general.
That's
true
too,
that
a
lot
of
times
the
code
is
obfuscated
and
until
you
you
know
symbolicate,
you
know
you
know
it's
useless
right.
D
So
it's
kind
of
natural
and
I
think
even
in
JavaScript
world,
we
we
want
to
convert
stock
traces
with
you
know
all
the
obfuscated
symbols
to
the
original
names
on
the
server
side.
It
doesn't
happen
on
the
agent
in
the
app
because
that
symbol
file
is
is
not
available
there,
so
I
think
it's
I
think
we
can
go
with
the
assumption
that
it's
a
Well
understood
process.
We
can
document
it
as
well.
Okay,.
B
Yeah
on
the
on
the
browser
side,
do
you
use
the
the
hard
file
so.
B
Right
yeah,
but
it's
useful
when
you
have
a
crash
right,
which
you
cannot,
you
know
easily
determine
it.
D
It
it
helps
for
sure
it
gives
you
more
information,
because
at
a
crash
you
know,
if
you
knew
you
know
what
went
back
and
forth
as
a
session.
You
know
it's
helpful,
you
know
you
could
follow
the
same
steps
and
and
then
you
know,
try
to
reproduce
the
the
the
error
but
but
in
the
absence
of
it
you
know
hard
file
capture
is
something
that
you
will
have
to
it's
a
separate
feature.
You
have
to
enable
it
and
it
it
is
more.
D
You
know
bulky,
you
know,
there's
a
lot
of
content,
so
in
the
absence
of
it,
the
least
you
you
need
is
still
you
you
got
in
the
stack
Trace,
its
symbols
are
all
you
know,
ABCD,
you
know
weird,
so
you
you
would
still
want
to
convert
it
into
the
original
symbols
and
that
that
you
know,
irrespective
of
the
hard
file
capture
you
know
you
still
have
to
do
it.
B
Right
yeah!
No,
but
do
you
have
some
may
of.
D
Support
for
heart
file,
capture,
yeah,
okay,
I'm,
not
sure
I,
haven't
seen
I
have
only
seen
the
requests
asking
for
it,
not
in
open
Telemetry.
So
far,
yet
yeah
I
think
one
reason
I
can
think
of
is
there
are
like
even
compare
even
take
the
example
of
a
crashed,
dub
right
or
a
cold
dump.
Transfer
of
that
is
it
considered
a
Telemetry
I
think
the
the
span,
events
and
the
metrics
I
think
they
do
not
support
capturing
large
content.
D
They
are
typically
meant
and
the
all
the
otlp
con
constraints
you
know
are
designed
around.
You
know
small
content
like
they
have
each
otlp
message
could
be
64
KB
or
something
I
think
there
were
some
similar
restrictions.
D
They
are
put
on
the
protocol,
I
think
they're,
all
designed
for
smaller
payloads
and
heart
file
could
get
large
because
you
are
not
only
capturing
I
think
there
are
options,
but
at
the
largest
level
you
can
capture
the
HTML
content
as
well,
not
just
your
headers,
so
it
can
get
large
so
so
transporting
that
to
a
back
end,
a
Telemetry
backend
will
have
to
be
done
offline,
not
through
otlp,
and
then
you
pass
the
metadatan
or
TLP
right
like,
for
example,
like
the
crash
report,
I
was
thinking
yeah.
D
Of
course
you
know
it's
spending,
discussion
and
I'm,
not
very
familiar
how
it's
done
today,
but
one
approach
I
was
thinking
is
the
crash
dump
itself
you,
you
enable
a
mechanism
to
upload
to
some
S3
or
some
location
and
then
pass
that
location
and
the
timestamp
and
some
metadata
in
an
event
in
the
TLP.
B
Yeah
I
think
that's
where
my
question
was.
You
know,
because
this
is
also
something
similar
in
terms
of
you
know,
for
a
debug
session
between
a
browser
and
a
backend
right,
and
it
is
the
only
way
to
actually
do
it.
If
you
do
not,
you
know
something
crashed
and
you
have
no
way.
D
Actually
it
it
is
available
in
some
ways
right,
because
every
Network
request
you
are
sending
your
capturing
through
span.
So
it
is
not
that
it's
not
available
it's
available
in
a
in
a
different
form,
not
in
the
heart
form.
B
Yeah
yeah
no
I
think
similar
to
I
was
thinking
in
terms
of
you
know
when
you're
talking
about
you
have
a
crash
on
Android
and
iOS
right,
so
it
could
be
emitted
because
you're
making
a
network
request,
but
it
crashed.
Now.
You
know
the
sponge
will
not
tell
you
that
right.
What
exactly
why
it
crashed,
it
could
be.
You
know
no.
D
So
what
we
need
to
do
is
you
you
need
to
have
a
concept
of
a
session,
and
sessions
is
also
something
that
you
know
we
are
talking
in
the
browser
world.
We
need
to
bring
it
here,
so
you
need
to
have
a
session
ID.
You
know
that
brings
together
all
the
network
requests
that
happened
in
a
session,
so
so
the
moment
there
was
a
crash
so
that
when
you
send
a
crash
event
also
you
know
you
you
need
to
attach
that
session
ID.
D
So
that
way,
if
a
crash
is
part
of
a
session,
you
can
see
what
were
the
network
requests
are
what
were
the
like
in
the
browser
we
are
even
designing,
for.
Let
me
show
you:
there
is
a
user
interaction.
This
user
action
is
like
button
clicks.
Where
did
the
user
click?
The
coordinates?
You
know
the
there
are
the
coordinates
and
the
element
which
element
which
button
you
know
the
user
clicked
right,
so
you
can
to
a
large
extent.
D
You
know
you
can
replay
conceptually
in
the
mind,
but
you
know
you
can
replay
what
the
user
did
right
right.
B
Which
is
exactly
you
know
what,
so
that
is
what
you
get
from
an
half
file
and
that's
where
I
was
thinking.
You
know,
because
the
crash
also
gives
you
that
replay
sort
of
you
know
in
some
sense
you
it
tells
you
the
place
where,
where
the
crash
happened
right
because
yeah,
but
but
to
make
sense
of
that,
you
need
all
this
decim
files.
That
will
tell
you
where
you
know
the
the
chain
of
the
the
path.
That
of
the
call.
D
So
so
that
could
be
another
feature
where
what
you
could
do
is
you
could
always
capture
the
entire
request
response
and
including
the
body
headers
and
then
keep
it
with
you
in
in
your
local
buffer.
You
know
on
the
Discord
somewhere
and
only
when
there's
a
crash
along
with
the
crash.
You
take
that
and
you
know
report
because.
B
B
So
so,
when
you
so
the
way
to
send
it
is
on
the
next
instantiation
of
the
app
yeah.
C
B
So
so
that
is
exactly
what
happens
when
you,
you
know,
have
a
browser
crash.
Where
bro
you
know
you
cannot
determine
from
the
so.
The
support
will
ask
you
know:
can
you
run
a
create
a
hard
file
and
then
send
it
to
us,
so
they
can
re.
You
know
walk
through
right
what
happened
and
when,
when
that,
so
that's
that
is
where
I
was.
You
know.
D
So
I
think
a
couple
of
things
need
to
happen
here.
I
think
this
will
be.
This
will
have
to
be
like
a
designed
properly
to
you
know,
be
a
standard
across
browsers
and
the
you
know:
Mobile
Android
iOS,
you
know
different
scenarios.
There
are
two
things:
I
can
think
of.
One
is
the
heart
file
format
the
heart
file
capture,
like
even
in
the
mobile
apps.
Do
you
want
to
capture
in
the
same
format?
I
think
heartfelt
is
already
a
standard,
so
you
could
capture
things
in
that
format.
D
The
second
part,
is
you
have
one
component
in
your
agent
in
your
app
that
is
capturing
this,
but
not
sending
it.
You
know
it
is
capturing
and
it
has
let's
say
a
buffer
size
of.
Let's
say
you
know,
500
events,
so
so
the
last
500
events
it
captures
anything.
You
know
older
than
that
you
discard,
so
you
always
keep
the
last
500
events
and
then,
along
with
the
crash,
when
you
report
you
report
that,
so
you
know
something
like
that,
something.
D
Of
course,
the
the
technology
used
to
capture
will
be,
of
course,
the
platform
dependent,
but
the
the
like,
you
know
it's,
it's
the
same
as
the
how
the
instrumentation
and
Telemetry
apis
are
different
today.
So
there
are
Telemetry
libraries,
you
know
the
three
signals
right:
traces
spans
and
traces
metrics
and
events
logs,
so
so
that
you
know
there
could
be
a
standard
on
how
to
capture
and
represent
the
you
know,
Network
requests,
and
so
that
is
the
our
file
format
yeah,
so
the
Telemetry
part
is
common.
D
A
B
So
so
my
my
question
is
in
the
JavaScript
on
the
web:
JavaScript.
Is
there
any
concept
of
you
know
the
similar.
D
No
okay,
I
am
not
familiar
I
think
it
may
not
be
easy
as
well,
because
in
the
in
the
browser
world
it
is
the
browser
that
gives
you
the
hard
file.
I
I'm
not
familiar
I
could
be
wrong.
I'm,
not
familiar
how
your
Javascript
app
can
extract
that.
D
A
D
Access
to
some
of
the
things
like
Ajax
requests,
the
the
once
the
page
has
loaded
the
the
JavaScript
can
make
API
calls
so
that
it
can.
It
has
access
to
the
request
and
response,
but
the
base
page
itself,
HTML
I,
don't
think
it
has
access
to
its
Network
aspects.
B
No
I
think
from
what
I'm
I
recall,
so
this
would
be
more
of
a
support.
You
know,
after
the
fact,
it's
a
user
initiated
activity
right
you
have
to.
You
have
to
go
to
the
you
know,
whichever
browsers
Network
Tab,
and
you
know
you
have
to
say,
yeah.
B
Yeah
yeah
yeah
so
in
this
case
a
crash
with
automatic,
and
you
know
you
can
analyze
the
crash,
whereas
there
I
think
there
is
some
user
initiation,
it's
not
automatically
created
for
you,
but
it's
there
for
you.
You
know
that
you
had
export
and
typically
it
is
asked
by
the
support
when
they
have
so
I'm.
Recalling
all
this
from.
You
know,
New
Relic
and
you
know
how
the
installers
ask
for
a
harder
fire.
When
you
had
this
kind
of
issues
with
the
browser,
so.
D
So
we're
not
maybe
I
think
we
should
create
a
document
and
and
then
write
this
down
and
then
we
could,
you
know,
distribute
it
with
the
Android
folks
and
browser
folks
and
then
arrive
at
you
know:
Common
design,
yeah.
E
I
I
think
I
think
so.
I
think
our
document
can
be
can
be
a
good
point
to
to
really
talk
about
things
and-
and
collabority
I
mean
probably
talking
about
that
also
in
some
meetings,
but
probably
something
that
we
can
move
with
our
respective
teams
mobile
teams.
So
they
can
also.
D
Yeah
and
I
would
still.
That
is
something
after
the
events
like
you
know,
first,
we
did
the
spans
and
now,
let's
work
on
the
events.
D
Right
and
the
three
events
that
we
talked
about,
there
are
a
crashed
report
and
the
network
change
and,
and
at
least
the
three
to
begin
with.
Maybe
there
could
be
more
and.
E
And
yeah,
also
probably
I,
don't
know,
probably
can
be
application
lunchtime.
Maybe
it's
gonna
be
something
similar
to
the
first
place
view
or
something
like
that
in
the
web.
But
the
lunch
time
is
a
bit
applications.
Time
is
a
very
valuable
metric
different
to
the
loading
of
views,
but
yeah,
maybe
I,
don't
know
if
that
could
be
also
unified,
somehow
with
web
events
or
or
yeah
that
that's
something
that's
a
value
that
probably
all
all
platforms
or
Ram
platforms
are
getting.
Is
the
time
to
launch
the
application.
D
So
what
I'll
do
is
I'll,
create
a
similar
Excel
sheet
and
share
it
with
you,
folks,
where
you
could
then
fill
in
you
know
the
attributes
that
yeah
you
are
interested
in
are
your
existing
products.
You
know
they
they
send
yeah,
and
then
we
could
identify
that
in
open,
Telemetry
events.
You
know
what
are
the
attributes
that
we
want
to
capture.
B
Yep
and
if
you
can
actually
I
don't
know,
maybe
address
links
to
the
document,
I
think
that'll
help
us.
A
B
I
think
this,
this
kind
of
stuff
there
is
a
lot
of
backend
support.
That
has
to
be
there
right
for
you
to
make
sense.
You
can
I
can
only
add
this
stuff
as
long
as
you
have
a
decent
file,
you
have
I
know
the
mapping
file,
and
you
know
all
this,
so
you
have
to
be
able
to
map
the
the
build
ID
with
a
crash
file
right,
that's
kind
of
the
for
that
there
is
nothing
from
the
open
Telemetry
you
can
do.
B
D
Okay,
okay,
actually
Splunk.
We
had
a
conversation
with
some
folks
in
Splunk
who
are
working
on
the
Android,
and
today
there
is
no
Android
open,
telemetry
Android
reports.
D
But
on
top
of
java
you
know
they
have
instrumentation
for
Android
and
only
so
Splunk
has
built
one
and
they
are
willing
to
contribute
it
to
open
telemetry.
D
So
we
had
met
some
of
the
folks
and
they
agreed
to
you
know,
move
it
over,
but
it
hasn't
happened
so
I'll
follow
up.
But
do
you
know
them
by
any
chance.
B
No,
no
I,
don't
I'm
in
a
different
Arc
yeah
I
can
I
can
check.
I
can
check,
though,
if
you
have
the
right
contact
and
so
on.
Yeah.
A
E
E
But
yeah
they
they
currently
have
not
contributed
much
to
the
project
except
improving
the
idea.
Exporter
I
think
that
was
a
bit
updated,
but
yeah
yes,
has
a
background.
D
All
right,
so
that's
all
on
this
topic:
I
will
create
a
document,
and
you
know
share
it
before
next
week,
if
possible,
and
that
way
now
we
could
come
prepared.
E
Yeah,
that's
that
that
would
be
great,
so
we
can
focus
a
bit
more
and
also
talk
with
I
mean
I
can
pass
a
document
to
my
to
the
ram
team
at
datadoc
that
it's
not
using
open
Telemetry
now,
but
probably
events
and
things
like
that
are
the
same
okay.
They
are
building
on
top
of
penetration
now,
but
I
see
planning
to
move
to
open
Telemetry
in
the
near
future.
I
don't
know
exactly
about
their
plans,
but
okay,
but
yeah
I.
D
And
one
other
quick
question:
has
anybody
started
working
on
the
implementation
of
the
events
API,
the
logs
API.
E
Now
that's
a
topic
that
we
we
have
I
mean
we
talked
in
the
past,
as
I
told
you
last
week.
Still
I
think
there's
still
no
work
on
this,
but
I.
That
was
one
of
the
topics
I
wanted
to
talk
today.
What
was
the
progress
there
on
yeah
I?
Don't
know
if
Stephen
could
be
interested
in
helping
with
some
area
of
the
Tracer
sure.
C
E
Yeah
I
know
Lux
is,
is
support
for
locks,
it's
basic
for
the
we
were
talking
in
the
past
about
that
be
not
to
us.
E
Yeah
also
also
around
that
he
yeah,
so
it
might
start
some
development
in
the
around
that,
but
still
no
no
clear
dates
right.
A
C
E
D
All
right,
thank
you,
then,.
D
So
that's
all
on
this
topic:
I
will
drop
off
if
folks
can
continue
with
yours.
Okay,
thank
you.
E
Thanks
Santos,
thank
you
very
much,
okay,
so
for
the
next
of
the
topics
just
to
follow
up
with
last
meeting
actions,
so
we
the
documentation
of
open
Telemetry.
Your
docs
are
ready.
I,
don't
know
if
you
check
those
stfm,
but
there
is
really
interesting
documentation.
Now
in
in
the
yeah
here
in
the
open
Telemetry
I
o
Docs
was.
E
Writing
and
Publishing
recently
it
has
yeah
lots
of
support
for
the
information
done,
how
things
work
yeah.
This
is
quite
new,
so.
E
Okay,
you
know
you
have
that
also
the
channel
and
it's
like
that.
We
try
to
provide
super
quickly
yeah.
So
here
we
there
are
some
things
about
money,
instrumentation
about
creating
expense
and
really
interesting,
the
the
active
span
and
starting
spans
as
active.
This
is
really
really
useful.
You
should
really
just
always
said
active
when
you
create
a
sponge,
so
your
side
response
can
get
the
context
automatically,
and
things
like
that,
so
this
documentation
is
done
is
they're,
approved,
I.
E
Think
Bryce
is
working
on
updating
some
of
this
to
be
more
complete,
but
still
on
several
peers
for
the
rest,
yeah.
We
have
to
change
the
readme.
Oh
yeah
forgot
about
it.
Sorry
I
will
say
that
read
me
and
add
the
price
document
to
the
readme,
so
people
can
get
proper
information.
E
The
status
of
the
Tracer
yeah
that
one
I
don't
know
if
it's
correctly
updated.
No,
we
have
to
Bryce
was
going
to
update
it.
I
mean
traces
are
definitely
in
a
production
ready
stuff.
Many
companies
are
already
using
that
in
production,
already
Magics
could
be
alpha
or
beta,
because
it's
a
bit
updated
outdated,
but
this
is
definitely
ready
for
production
and
for
the
next
more
things.
E
These
are
the
new
dogs
that
he
was
doing
and
the
events
Now
new
issues
in
the
project
are
new
peers.
A
E
This
there
were
no
news
about
the
pr
created
about
around
an
issue
creating
the
Tracer
provider,
the.
E
E
And
that
is
library
was
leaked
with
an
old,
open,
Telemetry
yeah
that
had
that
tissue
with
the
treasure
provider
with
multiple
threads,
so
I
think
that
was
their
problem,
so
yeah
these
things
must
be
done.
Contributor.
E
So
I
will
add
you
to
the
to
the
committer
to
the
progress
in
the
project,
so
you
can
approve
commits
also
in
the
project
now
I,
don't
know
exactly
what's
the
name
for
the.
E
So
yeah
that
I
will
I
will
create
that
PR
yesterday,
if
you
are
okay
with
that
and
yeah
about
the
logs
thing,
be
not,
did
you
see
it
visible?
Do
you.
B
Think
yeah
I
I
started
at
the
you
know.
Looking
at
the
the
Java
one
right,
I'm
I
wanted
to
actually
I
mean
we
can
take
it
offline
or
something
no
on
slack
or
something.
But
yeah
I
wanted
to
see
how
I
can,
because
they
have
things
separated
out
in
API
and
SDK,
and
so
on
and
I
wanted
to
kind
of
follow
the
same
model.
B
E
We
had
a
very
similar
model.
Product
teams
in
the
industry
model
leads
for
traces
is
very
similar
to
that.
So
and
metrics
also
started
very
similar
yeah
right,
so
we
can.
It
can
be
a
good
model
for
that.
E
C
I
actually
I
I
just
finally
got
everything,
sort
of
integrated
and
running,
and
we
have
built
a
little
test
thing
for
processing,
Network
requests
and
the
darn
thing
works
perfectly.
I
have
no
complaints
at
all.
At
this
point
great,
although
I
haven't
made,
it
do
anything
difficult
yet
so
it
was
just
like
send
a
network
request
and
darn.
If
traces
and
spans
don't
show
up
it,
it
worked
like
a
champ
and
so
far
no
nothing
I
need
change.
C
I
haven't
had
to
make
a
change
in
in
your
source
code
to
make
my
thing
work,
yet
I
haven't
I
still
I
haven't
touched
any
of
the
open
source
stuff
at
all
in
order
to
make
everything
work.
So
that's
that's
really,
terrific!
That's
the
first
time
that's
ever
happened
trying
to
integrated
open
source
project,
and
you
guys
should
be
proud
of
how
well
it
all
works,
and
but
now,
of
course,
comes
the
hard
part
of
adding
you
know
now.
C
I
have
to
add
more
data,
our
specific
data
items
into
into
our
into
the
stream
and
and
and
then
there's
lots
of
discussion
at
my
company
about
various
events,
some
of
which
are
probably
more
abstract
than
light
crashes
that
are
pretty
esoteric.
Some
of
the
things
that
we
do
are
kind
of
bar
removed
from
something
you'd
call
Telemetry,
I,
I
I'd,
prefer
to
you,
know
save
crashes
and
stuff
for
for
later,
given
the
complexity
of
it
all.
E
C
And
concentrate
just
on
simple
things
just
being
able
to
report,
you
know
basic
metrics,
you
know,
I,
think
the
network
transition
event
is
an
important
one.
I
think
our
customers
like
to
know
simple
things
like
you
know,
battery
level,
you
know
memory
available
things
like
very
basic
basic
metrics
about
the
device.
C
We
also
have
a
large
section
that
Santosh
alluded
to
that
that
tracks
the
interactivity
with
the
user
interface
where,
where
touches
occur,
and
we
have
an
elaborate
system
in
our
current
product-
that
grabs
screenshots
from
the
app
periodically
and
then
allows
one
to
basically
almost
make
a
video
out
of
replaying
the
session
sort
of
thing
as
it
occurred
on
the
on
the
client
device,
and
that
gets
really
quite
elaborate
because
you
have
to
hide
certain
things
and
it
gets
really
quite
in
the
bowels
and
it's
very
specific
to
the
device
where
the
we
we
do
the
same
in
Android,
but
there's
not
a
line
of
code
anywhere
the
same
between
the
two
of
them.
C
E
It
kind
of
yeah,
so
some
of
the
work
that
I
I
remember,
Bryce
was
doing
for,
for
the
Apple
metrics,
for
example,
create
updated
our
metrics
to
support
pretty
pre-cooked
Mercury
metrics
like
the
ones
that
Apple
sends
daily
or
weekly.
So.
A
E
Can
ingest
metrics
with
a
pre-cooked
data
yeah,
so
that
can
be
useful
for
you
and
also
about
those
events.
I
think
it's
basically
the
problem
that
Santos
was
having
also
to
to
have
a
something
standardized
between
all
the
all
the
languages
and
all
the
web,
because
yeah,
it's
and
the
product
for
us
is
that
the
that
client-side
spec
is
based
on
logs
on.
E
C
A
whole
lot
without
log
support
in
order
sort
of
the
underlying
mechanism.
That's
going
to
allow
the
rest
of
this
to
be
layered.
On
top
of
is
least
the
way
I
read
it.
Yeah
yeah.
C
I
mean
basically
I've
gotta
I'm
working
on
my
management
to
say,
look
I'm
going
to
need
to
they
jumped
into
the
Hotel
stuff
and
I
guess
on
the
browser
side,
on
the
Java
side,
I
guess
things
were
a
little
further
along,
so
they
were
able
to
do
some
things
and
they
said
well.
Can
you
turn
that
on
in
iOS
and
it's
like
well
not
quite
yet.
We
need
log
support.
C
We
need
some
things
so
I'm
trying
to
get
them
to
understand
that
we
need
to
to
allocate
a
bunch
of
time
to
to
helping
the
effort
and
getting
these
things
together.
So,
hopefully,
I'll
have
success.
There.
Santos
is
excited
about
it
now,
which
is
good
he's
more
of
them,
he's
on
the
he's
on
the
architecture,
side
and
he's
more
excited
about
getting
these
things.
What
I
had
proposed
for
crashes
was
a
little
different
than
I.
C
Think
what
San
Jose
well
I
was
thinking
that
we
we
I'd
like
to
see
an
event,
an
Otay,
an
open,
Telemetry
type
event,
noting
that
a
crash
occurred,
but
not
wrapping
all
the
crash
data
into
that,
because
that's
big
and
and
a
bit
proprietary
depending
on
error,
not
proprietary
but
but
different
depending
on
the
platform,
and
things
like
that.
C
Yeah
yeah,
so
it
was
more
like
sure,
so
they
crossed
yeah
like
I'd,
like
to
log
in
open
Telemetry,
that
a
crash
occurred
at
such
and
such
a
time
and
date
and
and
if
a
particular
vendor
wants
to
actually
capture
the
crash
data
and
send
it
somewhere
and
then
disembowlocate.
It
great
feel
free.
B
A
C
E
It
gets
serialized
together
with
the
cross,
so
you
can
save
there
whatever
you
need,
I
mean
your
trace.
Your
span,
ID,
whatever
you
need
yeah.
A
E
Refer
to
the
crash
so
that
they
also
had
way
I
don't
know
they
had
in
the
past.
They
also
have
a
way
to
write
another
file,
together
with
the
grass
that
you
can
read
after,
but
I
I
added
a
comment,
so
you
could
just
pre-cook
everything
when
you
have
it
running
and
whenever
it
crashes
it,
it
gets
saved
directly
with
without
you
needing
to
spend
time.
After
that,.
C
Yeah,
well,
we
yeah
yeah,
we
just
dump
it
to
disk
and
and
leave
and
then
the
next
time
we
start
up.
We
see
if
there's
any
records.
C
We
have
an
existing
mechanism
with
local
artifacts
that
made
that
work
easily
for
us
in
the
app
Dynamics
Regular
product.
But
again
I
was
thinking
that
I'd
I'd
I.
Imagine
there's
other
types
of
events
like
I
was
talking
about
our
screenshots
that
we
do
I
I'm,
that's
a
rare
case,
not
everybody's,
going
to
want
to
do
that.
It
takes
a
lot
of
data
and
you've
got
to
really
want
to
use
it
yeah.
C
It's
the
sort
of
thing
that
that
the
event
data
would
be
interesting,
but
the
actual
the
content,
the
amount
of
data
you
need
to
store
it
you're
going
to
be
able
to
replay
and
stire
Screen
sessions
like
a
movie
is
is,
is.
E
Expensive
and
yeah,
and
also
you
need
the
price
from
the
European.
So
you
have
to
watch
probably
right
to
record
your
tickets.
Yeah.
C
Know
in
the
future,
yeah
yeah
and
a
great
deal
of
it
is,
is
only
useful
to
you
know:
Enterprise,
apps
and
stuff.
Some
of
our
you
know
some
of
our
large
customers
use
it
for
development
to
see
you
know
before
they
release
anything,
it's
sort
of
outside
the
scope
of
of
straight
Telemetry.
You
know,
I
think
it's
beyond
the
scope.
B
I
brought
up
that
hard
file
thing,
because
you
know
heart
file
to
me
is
the
closest
to
a
crash
file
on
the
browser
you
know,
but
I
don't
think
open.
Telemetry
is
even
thought
about
it
right
on
the
hotel
web,
so
I
think
that's
kind
of
the
similar.
E
I
have
a
one
more
question
Stephen
just
before
in
the
in
the
meeting,
because
it's
it's
about
your
stack,
what
you
are
inducing
of
of
the
library.
If
you
plan
to
use
I
mean,
are
you
using
the
otlp
exported
basic
and
just
adding
the
open,
Telemetry
SDK,
or
are
you
planning
to
write
exporters
or
are.
C
My
the
product
managers
and
I
have
way
too
many
of
them
all
have
many
many
ideas
of
all
the
features
they
want
to
support.
So
it's
going
to
end
up
probably
needing
a
custom,
X
Border,
there's
only
so
much
stuff.
C
I
can
add
on
here
right
now,
ltp
exporter
to
our
implementation
and
I
and
and
of
course,
the
standard
out
one
just
for
debugging
is
very
handy
but
and
and
I
was
I
really
liked
the
the
data
dog
one
as
an
example
of
of
of
how
elaborate
one
one
of
these
can
can.
A
C
The
company
wants
to
build
an
open,
Telemetry
based
back
end
as
well
separate
from
our
current
existing
back
end
and
they're,
going
to
run
it
as
sort
of
two
parallel
products
as
far
as
I
know,
but
again
I'm
kind
of
out
of
the
loop
on
it
on
you
know
what
the
Market
goal
the
desires
are
for
a
product
base,
I'm
just
making
prototypes
for
them
to
assess
and
and
judging
but
right
now,
I'm
using
the
oltp
exporter
speaking
to
a
standard
implementation
that
one
of
the
guys
on
our
server
side
put
together
and
sending
spans
and
traces
out
that
way.
C
They've
got
a
implementation
for
the
browser
as
well.
That's
doing
similar
things
and
right
now,
they're
mostly
make
seeing
if
making
sure
that
they're
all
similar
and
that
the
data
is
something
they
can
consume
with
a
common
back
end,
and
so
that's
sort
of
where
the
experiment
is
right.
Now,
once
they
get
to
the
point
of
telling
us
what
features
we're
going
to
support
like
crashed
or
whatever
that's
when
a
a
custom
exporter,
May
may
may
need.
C
Right
yeah,
if
we're
sending
data
to
some
like
if
we
split
the
crashes
like
if
we
get
a
crash,
we
want
to
send
an
event,
but
we
also
want
to
send
you
know
the
crash
log
file
off
to
S3
or
something
then
that's
a
that's
a
different
thing.
E
Working
well,
you
shouldn't
have
problems
there
either.
Probably
the
only
thing
is
about
the
use
on
using
the
bats
exported
or
the
simple
exporter.
B
C
Yeah,
it
seems
it
seems
great
at
this
point.
I've
I've,
been
I've,
been
I've,
been
very,
very
pleased
how
quickly
it
all
came
together,
and
you
know
you
set
up
a
sample
and
it
actually
worked,
and
data
came
out
of
it,
and
I
was
yeah.
E
It's
really
high
quality,
I
would
say
yes.
A
E
Okay,
anything
more
be
not
I.
Just
could
you
a
bit
sorry,
no.
B
No,
no,
that
that
that's
fine,
we
can
talk,
you
know
offline,
but
you
know
I'm
I
need
to
go
through
the
logging
and
and
then
I'll
ask
you.
We
can
talk
on
slack
or
something
okay,
great.