►
From YouTube: 2022-10-05 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
A
A
A
B
So
looks
like
Ram
is
not
able
to
join
today.
So
if
others
don't
have
any
other
topics,
then
we
can
continue.
Our
schema
discussion.
B
So
we
yesterday
we
discussed
we
completed
discussing
the
Ajax
events
and
today
we
are
going
to
discuss
I,
think
user
action
and
error,
and-
and
these
are
the
messages
for
which
we
will
be
using
the
the
event
data
model,
or
rather
the
event
signal.
B
So
so
for
us
we
we
just
stored
the
stack
trace
and
and
the
error
message.
The
error
object,
but
I
I
see
Microsoft
you
never.
You
have
severity
level
line,
number
column
number
and.
A
May
we
break
it
down
a
lot
more
so
and
the
exceptions
is
actually
an
array
of
exceptions
and
it's
not
just
the
single
stack
right,
so
exceptions.
This
isn't
quite
the
stack,
It's
actually
an
object,
but
it's
close
enough.
B
I
see
in,
in
which
cases
do
you
have
more
than
one
stock
Trace.
A
In
JavaScript,
we
don't
it's
really
a
case
of
the
the
model
that
we
have
for.
The
SDK
is
the
same
across
all
languages.
So
for
it's
historical
I
see,
okay,
but
yeah
generally,
we
only
have
the
one.
Sometimes
you
can
can
get
a
Cascade
of
exceptions,
which
is
why
it's
an
array
so
that
you
know
that
exception.
One
was
caused
by
exception
two,
and
sometimes
it's
more
useful,
but
in
JavaScript
we
only
populate
the
first
one.
A
I
can
talk
to
some
of
these,
like
type
name,
is
effectively.
We
pull
out
the
type
of
error
so
for
JavaScript.
If
it's
a
custom
error,
then
that'll
be
whatever
the
the
error
name
is
otherwise.
It's
like
your
type
error
or
just
error,
or
whatever
the
the
native
object
is
that
got
thrown
and
the
reason
we
pull
it
apart
is
in
fact
the
stack
looks
different
across
the
board
for
different
browsers.
A
A
A
B
A
That's
effective
the
source
of
those
two
fields
and
yeah
they're,
not
there
for
all
browsers,
but
for
the
browsers
that
do
have
them.
We
take
them
and
generally
they're
encoded
in
the
message,
which
is
why
the
past
the
message
a
little
bit.
A
Yeah,
these
are
really
probably
your
name.
Well,
you
can't
get
them
in.net,
I,
don't
think
now
for
their
languages,
we're
having
the
discussion
internally,
where
they're
not
looking
at
pulling
it
apart,
they're
just
saying
message:
stack,
trace
and
then
making
human
readable.
So,
whatever
I'm
going
to
code.
B
So
severity
level
can
you
talk
about
it.
A
A
B
And
and
we
need
ID
and
properties
and
error
source.
B
A
I,
don't
know
we
actually
have
an
exception
ID.
Where
do
we
get
that
from?
There
is
an
ID
I
think
it's
past
as
but.
A
For
the
question,
I
think
all
right,
yeah,
it's
not
populated
for
auto
exceptions.
So
let's
ignore
that
one.
B
It's
it's
used
to
represent
nesting.
A
A
Properties
is
an
object
that
so
properties
and
measurements
identical
measurements
as
mustard
there,
it's
just
a
liquid
yeah,
their
own
class
of
property.
So
just
an
embedded
object.
Okay,
so
you
we.
We
have
the
default
on
automatic
on
error
stuff,
but
people
can
throw
their
own
exceptions.
So,
if
they're
further
on
exceptions,
they
can
give
us
properties.
A
Again,
this
is
like
a
user
supplied
value.
So
if
you
look
online,
you
do
a
residence
403
of
this
file.
You'll
see,
there's
a
function
there
to
say
create
auto
exception,
which
takes
it
as
the
last
argument
to
accept
the
this
is
the
code
that
calls
that
you
can
see
like
the
last
thing
it
sends
is
the
event
so
yeah
the
the
stack
of
the
aerosols
aren't
populated
for
that
portion,
but
the
stack
is
passed
out.
B
B
Okay,
Martin
is
not
here,
so
we'll
look
at
Splunk
foreign.
B
Do
we
have
do
we
do
we
record
errors
today?
Is
there
a
open,
Telemetry
package.
B
Okay,
what
is
start
and
end?
What
do
they
represent?.
A
B
A
Yeah,
didn't
we
find
the
name
for
that
yesterday?
What
was
it.
A
A
It
was
the
instrumentation
fetch
that
was
the
introvertation
name.
We
eventually
tracked
it
down
to.
A
A
Name
that
was
it
yeah,
sorry
for
t2's
component.
That's
the
scripting.
B
But
here
I
think
we
are
going
to
have
an
event
yeah.
A
So
yeah
so
then
type
name.
That's
really
we're
already
saying
that
because
I'm
saying
it's
the
event
or
the
error
thing
but
I
think
there's
already
like
a
bunch
of
exception.
Semantic
conventions,
Define.
B
A
Yeah,
which
is
what
we
have
like,
we
have
a
name
field.
It's
not
here,
which
we
call
I,
don't
remember,
just
disappeared
out
of
my
head.
B
A
B
Yeah
we
recently
talked
about
it
in
the
log
Sig
and
what
was
introduced
in
the
in
the
semantic
convention
for
logs
was
how
to
record
exceptions
in
in
other
log
messages
in
other
log
records
or
events.
So
it
need
not
be
an
exception
event
or
an
error
event.
B
But
if
you
want
to
additionally
attach
exception
information
how
to
use,
how
are
you
supposed
to
record
that
yeah
so.
B
Yeah
yeah
but
yeah,
but
for
the
sake
of
consistency
we
can
use
the
same
term
exception.
Yeah
foreign.
A
A
A
C
That's
the
same
as
type
name
with
Microsoft
so
which
op
error
object;
type.
It
is
from
yeah.
A
They
have
very
two
very
similar
Concepts
to
what
I
did
in
identity,
so
an
identity
which
we
don't
do
in
Azure
monitor
yet
yeah,
apparently
I
created
a
hash
for
a
collection
of
properties
so
that
once
it
got
to
the
back
end,
we
could
then
collect
it
all
together
and
say
anything
but
the
same
hash.
A
We
could
say:
okay,
we've
just
got
like
thousands
of
of
this
exception,
rather
than
having
to
have
a
query
that
looked
at
the
the
type
name
and
the
stack
traits
and
kind
of
thing
we
quickly
just
generated
that
I'm
not
running,
and
the
first
in
session
I
actually
had
a
count.
So
I
could
say
this
is
the
you
know
the
the
10th
exception
within
and
I
recent
I've
placed
on
a
time
frame.
So
it
was
the
Cascade
of
errors
that
we
talked
about
earlier,
so
yeah.
A
B
Last
one,
the
stack
hash:
how?
How
do
you
constructed
no,
as
in
different
browsers,
give
the
same
error
in
different
forms
right,
correct.
A
Which
is
what
the
hash
is
trying
to
to
resolve
so
effectively
it
processed.
A
What
would
I
do
it
was
the
message
and
the
type
so
I
think
I
didn't
I
didn't
take
the
stack
as
it
was
because
it's
different
across
browsers,
but
I
tried
to
match
it
up
on
it.
I,
don't
remember
the
exactly
I
can't
go
the
stack
on
the
back
end.
I
didn't
do
it
actually
on
the
browser,
because
we're
head
or
an
error
well
I
think
he
still
has
its
own
Aero
handling
I'm
getting
set
back
there.
So
I'm
still
running
my
code
for
that,
but
yeah.
A
B
Yeah
I
know
useful
for
sure,
but
is
it
better
than
on
the
back
end.
A
The
reason
I
did
it
on
the
back
end
is
I
pulled
the
exception
apart,
so
I
could
get
the
individual
components
from
a
JavaScript
perspective
in
the
browser.
It
is
a
bit
of
nasty
bit
of
code.
You've
got
to
write
and
even
just
entering
the
hash
takes
a
bit
of
CPU
time
as
well,
so
yeah
it
would
be
better
on
the
back
end,
but
we
would
need
to
pass
down
enough
components
to
be
able
to
do
that.
A
B
A
A
Net
C,
plus
plus
Java
node
I
can
definitely
see
that
the
advantage
there,
the
first
error
in
session
yeah
that
is
extremely
useful
for
identifying
the
Cascade
of
Errors,
because
you
you
get
a
you
get
one
error,
you
would
you
definitely
get
multiple
errors
after
the
fact
generally,
and
really
you
only
want
to
look
at
the
first
in
the
session,
so.
B
A
No,
they
are
they're,
not
required.
So
the
way
I
did
the
the
error
account
is
I.
Would
whenever
I
get
an
on
error,
I
would
effect
a
report.
Okay,
this
is
error,
number
zero
and
then
increment
it.
So
if
I
get
another
error,
Within
I
think
it
was
a
couple
of
seconds.
It
would
increment
the
error
cap,
but
then
I
think
a
kick
his
timer
off
and
that
would
reset
the
error
count
back
to
zero
so
that
you
can
actually
see
the
the
Cascade
of
errors
as
they
occurred.
A
So
it
then
just
concentrate
on
the
on
the
first
one
that
was
causing
the
problem,
but.
A
I
calculate
on
the
back
end
yeah,
but
but
the
error
cap
came
from
the
front
end
if
you
actually
go
to
a
like
an
identity
page,
whether
it's
like
Office,
365
or
the
MSA
ones,
and
cause
an
exception,
you'll
actually
see
the
exception
going
out
the
door.
It's
just
adjacent
problem.
B
A
Okay,
no,
this
is
not
the
same
error.
This
is
a
case
of
I.
I
have
an
error.
This
is
the
the
this
is
the
this
is
error
number
X
and
you
just
keep
incrementing
what
that
error
number
is
for
that
session.
I
also
reset
it
after
a
certain
period
of
time.
So
it's
it's
not
a
case
of
this
is
the
same
error
for
this
particular
event.
Oh.
A
Because
if
you
get
the
first
error,
sometimes
that
can
cause
a
Cascade
of
Errors.
So
if
something
fails
to
initialize,
you
might
get
extra
exceptions
down
the
track,
because
something's
now
trying
to
use
the
thing
that
didn't
get
initialized.
A
I
found
it
definitely
useful
for
dashboards
that
I
created
so
I
I
did
it
on
the
back
end.
So.
A
Calculating
on
the
front-
and
we
could
do
it
by
by
saying
let's
take
the
type
it'd-
probably
be
a
combination
of
the
the
message,
the
type
and
maybe
the
line
and
column
number
so
I
had
different
Fields
related
to
the
actual
page,
so
that
I
knew
I
could
narrow
it
down
to
the
page
that
was
causing
there.
B
B
Okay,
that's
this
added
object.
B
A
It's
not
even
here
yet
either
so
effectively.
We
have
two
versions
of
this
when
we
find
that
they're,
essentially
the
same
with
a
slightly
different
set
of
properties,
foreign.
A
Yeah,
so
actually
you're
already
here.
So
if
you
go
back
to
the
root
node,
so
application,
insights,
Js
and
then
extensions
and
there's
a
click
analytics
in
there,
so
yeah
this
one
here.
So
this
isn't
loaded
by
default.
People
have
to
choose
to
grab
this,
but
this
is
the
thing
that
grabs
all
of
it.
So
if
you
look
in
events,
I
didn't
write
this
one.
One
of
the
questions
previously
worked
good,
so
yeah.
These
are
the
sorts
of
things
that
we
grab
for
that's
the
configure.
A
A
B
B
And
and
then
what
is
additional
here.
B
A
Yeah,
so
this
is
the
behavior
file
that
we
showed.
This
is
where
it
gets
used
yeah
if
you
go
back
and
look
at
look
at
the
event
and
look
at
events,
so
these
are
the
individual
server
yeah
page
action
on
where
we're
in
so
those
are
the
things
that
will
actually
generate
stuff
indoor
yeah.
So
that's
our
lovely
embedded
event
that
data,
which
is
a
nested
object.
A
Yeah
sort
of
a
coordinates
where
it
happened-
page
name,
parent,
ID,
populated
data
at
present,
that's
already
where
the
magic
happens.
A
I'm
not
that
familiar
with
this
one,
but
out
of
the
two
versions,
the
internal
version
existed
before
I
joined
the
team,
and
this
was
it
effectively
a
copy
of
that
it
was
created
for
another
team
who
reuses
this
for
their
own
bit.
It's
in
the
readme.
It
says
time
taking
a
milliseconds
to
use
it
to
click
the
element.
Since
the
initial
paint
line.
A
No
because
the
time
stamp
is
the
time
at
which
the
the
event
was
so
so
line.
61
was
called.
A
Well,
it
could
be
most
of
the
time
they're
probably
the
same,
but
we
do
have
effectively
lazy,
initialization
aspect
as
well,
so
we
have
the
ability
to
effectively
record
these
events
which
effectively
just
become
function
calls
while
the
SDK
is
still
loading
once
the
SDK
is
loaded,
it
replays
them.
So,
therefore,
in
those
cases
the
timestamp
is
different
to
this
time,
Direction.
A
I
think
that
is
yeah.
So
if
you
look
on
look
at
the
readme,
so
if
you
go
back
to
the
click
analytics
route,
there's
a
nice
summary
in
that
back,
one
more
yeah,
so
the
readme
yeah.
So
if
you
scroll
down
it
has
actually
has
a
table
lamp
which
explains
there.
You
go
so
yeah,
so
the
data
tags
so
affected
that
you
mock
up
the
element
with
you
know:
data
Dash
attributes.
A
A
B
B
Okay
I'll
add
some
of
these.
Let's
see.
B
B
A
lot
of
type-
and
this
could
be-
where
did
you
read
the
different
types.
A
Source,
you
know
there,
we
go.
B
B
A
Yeah
I
think
there's
a
bunch
of
stuff
a
bit
like
herpe
head
where
they
just
collect
a
bunch
of
standard
stuff.
Like
your
page
name,
where
it
happens.
The
parent
IDs.
B
Those
will
be
their
common,
I
guess,
I
think
from
the.
A
A
A
A
Until
I'm
creating
the
source
space
type-
okay,
this
can
be
passed
in
but
yeah.
We
we
use
this
on
the
back
end
to
the
the
to
our
validation.
A
A
Yes,
so
in
the
click
analytics,
it's
it,
the
user
actions
always
click
event.
The
web
action
is
derived
from
another
field
and
I'm
just
trying
to
find
those,
but
so
this
is
a
common
thing
that
we
have
across
all
of
them.
So
every
one
of
our
events
has
a
base
type,
which
one
is
this
so
I'm
now
in
well,
click
analytics
Okay.
So
that's
not
the
one.
A
So
like
for
exceptions,
it's
exception
data,
it's
I,
guess
what
we
use
for
identifying
the
type.
So
we
could
almost
not
have
this.
B
Okay,
so
let's
get
that
so
time
to
action?
Also
I'm
thinking
we'll
we'll
skip,
because
it's
at
the
most
a
few
milliseconds
off
right
compared
to
the.
B
Okay,
then
we
can
just
keep
these.
C
Currently
it's
the
element
type
like
a
or
button
or
it
will
work.
Okay,
yeah.
The
targets
are
a
bit
inconsistent
right
now
because
it
takes
the
event
Target
which
like,
if
you
have
a
pattern
inside
of
it,
it's
done,
there's
an
icon
that
does
text
and,
depending
on
whether
it's
a
user,
it's
on
the
text,
Story
icon,
is
different,
so
it
would
need
a
little
bit
of
standardizing.
B
Right
and
do
we
also
need
the
the
let's
say:
if
it's
a
link
or
a
user
clicks
on
a
link,
do
we
are
a
button?
Do
you
want
to
do
you
want
to
capture
the
text
of
that
button.
C
Probably
not
party
first
so
having
a
way
to
define.
B
And
the
reason
I
I
haven't
looked
at
the
you
know,
this
click
analytic
products,
but
if
you
were
to
show
what
people
are
clicking
more
on
a
given
page,
you
know
you
for
for
customers
it
will
be
easier
to
follow
if
we
tell
them
hey,
they
are
clicking
on
this
button
named
so
and
so
text
versus
the
other
button
name
so,
and
so
some
other
text.
C
B
And
is
this
event
type
equivalent
to
something
we
already
have.
B
32
this
event
type
here.
C
B
Okay,
so
good,
then.
B
Okay,
so
I'll
I
I
need
to
populate
this
table
I'll.
Do
it
thank
you
and
then
maybe
next
time
we
meet
I'll
I
would
have
completed
this
and
I
will
put
some
of
this
information.
Also
in
the
other
document.
B
A
Yeah
I,
just
it
was
going
for
the
code.
This
is
just
a
bullying
flag,
so
it's
obviously
someone
requested
it
at
some
point
in
time.
A
B
In
in
our
case,
hopefully
there
will
be
a
session
ID
that
will
tell
us
if
the
user
is
logged
in
or
not.
A
Yeah,
okay:
this
is
like
it
would
be
nice
for
the
pretend
to
join
again
today,
so
we
could
try
and
push
the
thermal
resources
but
yeah
the
session
ID
I
think
is
if
they
give
it,
we
need
that
promise
now.
I
feel
like
some
of
the
other
ones.
Here,
like
the
page,
like
line
six
right.
A
A
Attributes
so
if
you
click
on
a
sub
element,
we'll
walk
up
the
Dom
to
a
certain
point,
yeah
grabbing
all
the
data
attributes
for
that
user
action.
A
There
can
be
a
lot
of
information
depending
on
how
the
page
is
constructed,
but
I
feel
like
the
the
shopping.
The
is
the
we
have
a
couple
of
big
customers
of
this
for
the
internal
version
and
yeah
they.
If
you
look
at
the
source
of
the
page,
you'll
see
data
tags
all
over
the
place,
but
they
that's
because
they
effectively
have
components
like
I
can
see
it
in
my
head.
A
I
can't
think
of
the
name
affected
where
you
can
like
buy
products
affect
each
image,
is
its
own
little
I
think
they
use
a
react
component.
A
A
Yeah
yeah:
it's
considering
we're
the
only
ones
that
have
it
here
and
it
might
be
a
case.
The
other
guys
haven't
got
to
populating
this
year,
which
is.
B
You
know
yeah
what
I
mean
is
like
you
can,
keep
it
but
keep
it
turned
off
by
default,
and
then
customers
that
need
it
that
are
going
to
depend
on
it.
You
know
they
can
turn
it
on.
A
Yeah
deal
with
it,
you
don't
mark
up
the
page
and
then
there's
no
data
to
grab
the
same.
So
as
I
say
the
click
analytics
extension
is
they
have
to
explicitly
go
and
load
that
it's
not
like
the
default.
Cdn
SKU
is
doesn't
have
it.
B
So
this
is
again
an
object.
It's
it's
a
dictionary
key
value
here:
okay,
what
is
Page
type
error,
page
form
page.
A
That's
that
behavior
file
that
we
showed
a
while
ago,
which
is
like
huge
yeah.
It's
again,
it's
I
think
we
also
get
them
from
yeah.
It's
not
in
oh
yeah.
There
you
go
so
yeah.
You
can
mark
up
a
page
and
say
this
is
the
behavior
that
they're,
seeing
and
I
think
it's
a
combination
of
these
two
and
say
we
Supply
it.
We
don't
actually
directly
set
these.
These
are
all
derived
from
the
markup.
A
A
B
This
will
require
detailed
instrumentation
right,
like
you
would
you
need
to
identify
which
one
applies.
A
C
So
I
said
an
idea
on
the
last
two
minutes
of
the
sake.
What,
if
just
freaking
out
that,
if
you
put
the
element
on
attributes,
data,
Dash,
Auto
Dash,
something
on
the
element,
then
it
will
be
put
as
an
attribute
on
the
correct
generated
and.
B
Sorry
I
I'm
not
following
can
can
one
of
you
put
that
here
somewhere
if
we
need
it.
A
Well,
it's
just
the
case
of
the
instrumentation
that
does
the
work
rather
than
grabbing
every
attribute
or
every
data
Dash
attribute.
It
only
grabs
a
subset
of
them.
So
the
proposal
was
it
has
an
Hotel
Dash
prefix.
A
A
Yes,
sorry,
the
URI
is
the
page
UI
the
target.
Uri
is
it's
a
click
on
an
anchor
example
that
says:
okay,
so
where
it's
going
to.
B
So
so
that
that's
what
I'm
wondering
that
you
can
click
on
a
button?
You
click
on
an
image
so
for
each
type
of
element
that
you
click.
A
Yeah,
which
is
like
yeah,
that's
so
it
depends
on
the
thing
you're
clicking,
that's
where
the
content,
the
the
target
URI
and
all
the
data
attributes,
so
the
page
tags.
So
that's
what
they're
all
about
and
yeah
and
by
default
we
have
like
AI
Dash
is
the
prefix.
So
it's
data
Dash
AI
Dash,
but
you
can
configure
it
with
your
own
prefix,
so
it
can
be
like
data
Dash
and
then
whatever
else.
B
A
So,
as
I
said,
the
content
would
be
the
text.
The
target
URI
would
be
the
anchor
of
the
image
wherever
else,
but
then
we
also
have
the
behavior
and
the
data
tag
for
the
markup
and
let's
say
for
the
dated
tags:
we
do
walk
up
the
Dom,
not
necessarily
the
thing
you
clicked
on,
but
it
could
be
the
data
tags
from
the
parents
or
the
parents,
parent.
B
A
B
All
right
talk
to
you
next
time,
then
yep.