►
From YouTube: 2022-06-14 meeting
Description
Instrumentation: Messaging
A
B
C
E
Hello,
martin,
did
you
have
anything
on
the
agenda.
A
E
G
Maybe
just
upfront
some
information.
I
will
invite
some
of
our
the
agent
mobile
agent
developers
next
week
to
the
tuesday
and
wednesday
meeting.
So
if
you
have
any
questions,
let's
talk
about
or
let's
focus
on
mobile
next
week,
please.
E
G
I
can
tell
you
your
otep
already
draw
some
rules
and
draw
some
attention
to
other
agent
developers,
so
they
are
already
discussing
it.
So
it's
in
a
very
positive
way.
G
So
it's
fine
from
open
telemetry
experienced
people
in
our
company,
so
for
them
they
don't
see
any
issues.
E
Okay,
and
would
that
be
android
or
ios
or
both
server
sides,
the
the
folks
that
you
mentioned.
G
The
folks
that
I
mentioned,
I
have
ios
android
and
also
hybrid,
so
flutter
and
react
native
okay.
E
Okay,
if
it
is
android,
I
think
you
know
there
was
one.
There
is
one
help
that's
needed,
which
is.
E
There
is
a
splunk
agent
for
android,
and
I
think
you
know
long
ago
john
watson
mentioned
that
they
are
open
to
contributing
that
to
open
telemetry,
but
they
don't
have
resources.
The
stupidest.
G
We
were
experimenting
a
bit
with
the
ios
agent,
also
from
splunk,
so
we
have
some
experience
there.
Android.
I
would
need
to
check.
E
Yeah
yeah,
of
course,
of
course-
and
I
I
don't
see
anyone
from
splunk
here,
I
guess
so
I
I'm
I'm
wondering
you
know
why
they
have
a
lot
of
functionality
in
their
public
reports,
but
you
know
like,
for
example,
there
was
one
topic
which
is
a
server
timing,
header.
You
know
that
that
you
know
that
is
used
to
send
the
trace
parent
back
to
the
client
from
the
server.
E
There
is
a
bunch
of
logic
that
they
added
to
their
repos,
but
it's
generally
helpful
for
for
everyone.
So
I'm
wondering
you
know
what
is
their
criteria
for
what
they
would?
You
know
added
to
their
reports
versus
what
they
would
add
it
to.
E
You
know
the
the
common
repo,
the
open
telemetry
reports.
Are
there
any
general
guidelines.
G
E
And
and
okay
before
we
go
discuss
deeper
on
martin's
topic
ted
one
quick
question:
like
did
you
get
it?
Did
you
get
time
to
look
at
the
the
ephemeral
resources
proposal
to.
D
Make
one
no
sorry,
I
had
kind
of
a
a
rough
week
last
week,
yeah
I'm
getting
the
covet
vaccine,
but
I'll
try
to
get
it
to
you
guys,
hopefully
in
the
next
tomorrow,
the
day
after
that,
if
I
don't
have
it
ready
for
tomorrow's
meeting,
then
certainly
before
the
next
one,
okay.
E
Okay,
yeah,
so
I
guess
it's,
it
will
be:
okay,.
F
E
We
can
discuss
you
know
once
you
have
some
initial
writer,
but
I'm
guessing
that
there
will
be.
You
know,
naturally
a
lot
of
resistance
from
the
other
folks
right.
So
do
you
do
you
want
to
like
discuss
with
them?
First
before
you
create
a
proposal
or
you,
you
would
use
that
proposal
as
a
means
to
discuss.
D
D
You
know
this
is
I
just
the
the
one
thing
I'm
I'm
not
fine
with
is
just
stapling
the
session
id
everywhere,
because
we
don't
have
a
good
place.
To
put
it.
That's
so,
I
would
say:
that's
like
the
one
non-solution
that
seems
easy,
but
is
it's
not
good?
So
people
don't
like
this
solution?
That's
fine!
You
know
but
make
a
counter
proposal
in
that
case,
but
we
have
to
do
something
here
so
so
I
think
I
think
it
will
be.
D
It
will
be
fine,
as
as
part
of
that
proposal,
I'm
going
to
include
an
explanation
for
how
we
can
keep
existing
resources,
as
you
know,
non
immutable
right,
so,
okay,
hopefully
hopefully
that
will
mollify
them.
Okay,.
C
I
know
daniel
was
going
to
try
and
follow
up
with
a
community
group
that
doesn't
seem
to
have
any
traction.
Do
you
have
anyone
we
can
try
and.
D
Yeah,
so
so
the
issue
is
just
now:
you
created
the
problem,
is
you
created
an
issue
and
no
one's
responding
to
it?
Yep.
C
D
Okay,
thanks
I'll
bring
I'll
bring
that
up
with
the
gc
and
tc.
A
Okay,
so
let
me
share
my
screen.
A
Okay,
so
as
as
we're
working
on
working
on
the
the
events
api-
and
I
I
imagine
hopefully
soon
that-
will
get
get
some
agreement
and
I
think
I
would
like
to
start
thinking
about
specific
semantic
conventions,
and
you
know
I
think
I
mentioned
that
last
week,
that
the
two
that
kind
of
come
to
mind
is
is
kind
of
or
three
top
top
importance
for
me
is
timing,
events,
interactions
and
exceptions.
A
So
so
the
one
thing,
the
one
that
I
started
working
on
in
more
detail
is
timing,
and
I
wanted
to
just
get
some
feedback
from
what
folk
you
know
what
folks
think
like
this.
This
could
look
like
so
timing.
A
Events
like
so
I
listed
kind
of
a
number
of
different
different
events
that
that
could
happen
in
browser
that
have
that
are
important
from
like
analysis
of
what's
happening
on
the
page,
and
I
call
them
timing
because
they
they
have
like
a
time
stamp
like
a
relative
time
stamp
to
navigation
start.
A
So
you
know
when
you're
looking
at
performance
like
it
tells
you
something
like
it
tells
you,
for
example,
like
how
quick
the
the
page
was
loaded.
You
know
how
you
know
when
it
first
thing
was
painted
when
the
you,
when
first
input
happened,
meaning
like
when
the
first
time
the
user
interacted
with
the
page
there's
also
things
like
element:
timing,
user
timing,
like
custom
timing,
there
is,
there
is
a
navigation
timing.
Obviously
there
is
a
resource.
I've
got
I've
done
resource
all
of
these.
A
So
most
of
these,
like,
if
you
look
at
the
mdn
mbn,
well
w3c
specs,
then
these
are
apis
that
all
of
these
have
you
know
if
you
look
at
these
in
detail
like
so
so
there
is
like
the
the
base
class
for
this
is
performance.
Entry
performance
entry
has
like
the
type
of
the
the
event
or
the
timing,
and
then
it
has
a
start
time
and
duration,
which
the
start
time
is
pretty
much
always
there.
I
think,
which
is
the
relative
timestamp.
A
A
There
is
a
you
know:
there's
such
a
thing
as
the
performance
entry,
which
is,
which
I
think
is
pretty
much
identical
in
my
mind,
to
like
timing,
timing
events
that
I'm
thinking,
if
you
look
at
then
the
details
like
like
then
like
the
individual
types
of
entries
so
like
the
enter
set,
name,
entry
type
start
time,
duration,
but
then
the
individual
ones
have
different
attributes
depending
on,
like
you
know,
for
that
type
that
are
specific
for
that
type.
A
So,
and
so
I
don't
know
like,
for
example,
element
timing
has
an
identifier
which
is
unique
to
it
largest
contentful
fans
will
have
things
like
render
time
the
url
size
resource
resource
timing
obviously
has
a
bunch
of
other
stuff
that
has
that
is
related
to
making
the
connection
and
downloading
that
resource
and
navigation
timing
has
actually
extends
resource
timing,
and
it
has
additional
attributes
for
events
that
it's
like
sub
events
that
happen
during
window
window.
Loading
loading,
the
page
such
as
non-interactive
down
complete,
obviously
load
events.
A
So
so
those
are.
Those
are
so
like
a
lot
of
these
events
that
I
could
think
of
like
come
from
this
all
common
api.
There
are
two
that
that
I
don't
see
that
I
think
we
might
want
to
map
to
to
capture
which
are
window
in
unload
and
page
like
hiding
page,
maybe
that's
a
visibility.
This
is
maybe
a
different
type
of
event,
but
window
unload
for
sure
like
would
be
useful
to
capture
and
that
that's
not.
E
One
fundamental
question:
are
you
you're
you're
listing
this
under
events,
so
my
my
understanding
is
that
anything
with
respect
to
like,
let's
say
the
timing,
events
right,
the
timing,
events
with
respect
to
jack's
calls
api
calls
are
better
represented
as
events
within
their
corresponding
spans
right.
The
you
know
like
yesterday,
you
know
I
was
discussing
internally,
whether
like
in
our
company,
whether
it
makes
sense
to
pull
them
out
as
independent
events
with
with
you
know,
with
the
link
to
the
corresponding
span.
E
The
the
problem
is,
you
know
for
a
lot
of
the
metric
calculations.
You
know
you
know
you
would
have
to
you
know
subtract.
You
know
one
marker
from
the
other,
so
you
would
need
to
actually
group
the
events
of
that
particular
span.
You
know
together,
which
you
know.
F
E
Be
in
a
significant,
you
know
complex,
so
for
that
reason
you
know
we
felt
that
it.
You
know
they
are
better.
You
know
inside
the
spans,
as
they
are
already
today
right
in
the
the
current
instrumentations
we
have
for
the
xhr
and
even
the
document
load
they
they
are
already
inside
the
there
are.
There
are
events
inside
the
span,
so
are
you
thinking
of
moving
them
out
too.
A
Yeah,
so
that's
that's
part
of
the
question
like
I
like.
I
know
that
the
existing
instrumentation
that
we
have
now
uses
spans
like
we
use
spans
for
window
load,
we
use
spans
for
resources
and
then
I
think
the
load
event
has
has,
like
you
said,
span
events
linked
to
the
span
which,
which
I
think
are
I
don't
know
exactly
from
top
of
my
head,
but
I
probably
like
the
dumb
content
loaded,
dumb,
interactive,
like
events
on
on
the
span
right,
I
guess.
A
So
so,
as
I'm
thinking
about
this,
like
you
know
like
if
you
look
at
like,
for
example,
like
the
resource
resource
timing,
like
from
the
api
perspective,
the
whole
like
the
whole
thing
is
one
event
with
all
these
attributes.
A
Right
so
there's
there
is
one
there's
one
event
for
like
a
one
resource
and-
and
then
you
you
get,
you
know
all
these
attributes
on
that
and
the
same
for
navigation.
You
could
potentially
have
just
one
event
that
has
all
these
attributes
on
it,
as
opposed
to,
like
you
know,
like
extracting
all
of
these
values
and
creating
event
for
each
of
those
right.
So
so
so
like
like
a
couple
of
the
like,
so
I
was
I
started.
A
I
didn't
finish
this,
but
I
started
writing
down
options
of
what
I
think
these
things
could
look
like
like
in
this
in
the
spec.
So
so
we
could
have.
We
could
have
you
know,
one
combination
namespace
for
all
events
that
represent
timing,
so
that
could
be.
There
could
be
like
a
one
option
so
in
in
that
case,
like
the
event
name
could
be
either
timing
or
it
could
be,
maybe
even
like
more
generic,
like
a
performance
entry
like
if
you
wanted
to
match
the
the
w3c
space,
I
think.
A
So
so
you
would
have
yeah.
So
in
that
case
like
if
you
you
know
like
like
going
back
to
this
example
here,
like
they're
like
the
base
classes,
performance
entry,
then,
like
you
know,
there's
resource
timing
is
extends
that
and
there's
like
navigation
timing,
extends
that,
and
so
like
yeah
is
this.
There
are
some
base
attributes
on
the
performance.
Entry
like
resource
timing
adds
some
additional
attributes
and
navigation
timing
ads.
A
It
is
some
additional
attributes,
so
you
could
have
if
you
did,
if
you
did,
did
it
this
way,
then
I'm
wondering
like
do
we
want
to
what
one
question
is
like
do
we
want
to
enumerate
all
these
all
these
attributes,
or
do
we
want
to
just?
I
think
we
mentioned
this.
We
want
to
just
have
a
single
single
attribute,
something
like
timing,
value
or
you
know,
entry
value
or
something
like
that.
That
has
a
stringify
json
of
all
these
things.
All
these
attributes
that
come
from
from
the
from
the
api.
C
Yeah,
so
this
gets
back
to
how
I
think
the
the
general
event
should
look
so
effective.
We've
got
we're
gonna
have
the
event
domain,
the
event
name.
I
think
we
should
just
have
an
event
data
blob,
whether
we
call
it
value
whatever.
I
don't
do
not
want
to
call
the
timing.value
and
it
would
be
the
nested
attributes
so,
depending
on
the
event
type
that
will
define
what
event
data
looks
like.
So
in
these
particular
cases,
we've
got
all
the
w3c
standards.
That
say
a
performance
entry
looks
like
this
so
effectively.
C
This
data
object
is
really
a
case
of
that.
We
asked
the
browser
to
give
us
that
entry
and
it's
it
just
put.
We
just
put
it
there
in
the
case
of
santosh's
thing
about.
You
know
spans
versus
log
events.
I
I
want
to
define
one
set
of
events
and
that
event
can
exist
either
here
using
the
the
log
event
api
or
it
can
appear
in
a
in
the
span
event.
C
That's
part
of
the
motivation
for
my
separate
thing
of
rationalizing
the
the
attributes.
If
nothing
else,
if
we
can
get
the
nested
attribute
defined
on
a
span
event,
that
will
be
good
enough
for
this.
The
only
issue
with
span
events
is
there
is
a
limit
by
default.
128
events
per
span.
C
Some
of
these
perk
timing
ones
make
sense
to
be
on
the
span,
some
of
them,
don't
in
the
case
of
ajax,
that
that's
100
make
the
perfect
sense
to
be
associated
with
spain.
E
So
can
I
share
for
just
a
minute
martin,
I
have
a
sample
span
open
on
my
screen.
I
don't
know
how
to
share
it
and
I
yeah.
E
Okay,
so
now
I
think,
if
I
understand
you
correctly,
we
have
these.
Currently
the
current
javascript
instrumentation,
the
document
load,
instrumentation
library.
You
know
it
creates.
E
You
know
each
of
these
parameters
as
a
as
an
individual
event,
yeah
right.
So
instead,
what
you
are
suggesting
is
that
you
know
this
should
be
something
like.
C
C
F
So
one
thing
I'd
like
to
caution
there
with
the
data
blob
is,
I
don't
think
we
want
to
basically
take
the
performance
timing,
object
and
stick
it
into
this,
as
is
we
should
still
define
which
parts
of
the
performance
navigation
object?
Are
we
going
to
read
it
to
this
top
level,
because
I
know
that
could
be
a
huge.
F
Data
blob,
if
you
simply
just
ask
for
browser
performance
navigation
object,
that
is
huge,
the
payload
is
going
to
be
insanely.
Big
we've
run
into
issues
with
you
know
that
hitting
the
limits
and
things
so
we've
got
to
be
careful
about
which
parts
of
the
navigation
object.
We
take
the
tree
test
top
level
I
mean
or
tier
one.
If
you
will,
we
can't
simply
just
say
we're
just
going
to
take
all
the
performance,
navigation
objects
and
stick.
E
F
I
like
that
idea,
because
somebody
else
defined
it
w3c
where
the
spec
defined
it
will
just
say
we'll
use
it,
but
that's
just
huge.
We
need
to
take
a
subset
of
it.
That's
my
recommendation.
F
We
can
look
across
we,
you
know
we
can
look
across
companies,
see
which
ones
we've
been
commonly
using
and
you
know
come
up
with
the
distal
list.
I
think
I
see
so.
F
A
So
I
think
I
I
yeah.
I
agree
with
that.
I
think
I
I
do
want
to
just
point
out
that
essentially
we're
just
copying
right
the
what
is
what
is
already
defined
defined
in
the
w3c
spec
and
if
they,
if
they
add
something
like
we
might
need
to
like
like
make
sure
that
our
spec
is
synchronized
with
their
spec.
C
Or
we
just
say
we're
following
the
w3g
spec:
these
are
the
the
known
fields.
If
there's
something
we
want
to
specifically
exclude
from
what
ram
was
saying,
then
we
then
we
say
we're
excluding
these
and
we
don't
ingest
them.
Okay,.
F
C
Yeah,
because,
depending
on
the
browser,
you
don't
get
all
them
anyway
and
depending
on
the
navigation
like
for
the
navigation
ones,
that
if
there
was
no
dns
lookup,
you
don't
get
the
dns
lookup
entries,
for
example,.
A
Okay
back.
C
I
I
think
we
should
enumerate
them
out
and
and
see.
What's
there
I
know
internally
for
office.
I
just
grabbed
the
navigation
event,
put
it
into
a
jason,
apply
it
because
they
wanted
us
yeah,
but
that's
optionally.
It's
turned
off
by
default
and
they
can
optionally
turn
on.
They
want
to
include
it
with
the
ajax
requests.
E
Yeah
that
that's
what
I
I
would
do
too,
but
I
think
from
what
ram
is
saying
that
it's
it's
like.
Have
you
seen
cases
where
the
browser
exposes
a
a
lot
of
parameters
and
yeah?
You
know
there.
F
I
think
let
me
correct
me
if
I'm
wrong
he's
much
more
experienced
in
that
area.
It's
I
believe,
react
pages.
You
know
the
individual
components
will
have
their
own
load,
timings
and
things.
So
the
performance
planning
object
is
a
nested
object,
so
each
one
would
have
if
you
have
a
thousand.
Sometimes
there
have
been
pages.
Where
thousand
components
are
there?
Each
component
reports
its
own
timing.
C
Yeah,
I
think
we
explicitly
exclude
that
we
just
say
it's
the
page,
navigation
timing
and
the
resource
navigation
timing.
I
I
I
don't
think
we
want
to
say
the
navigation
is
everything
that
was
required
to
load
this
page
and
include
all
the
resources
yeah.
I
definitely
don't
want
to
that's
not
where
we
want
to
go
yeah.
F
Okay,
so
so
we
we
have
to
be
specific
about,
you
know
which,
which
parts
of
that
you
know
performance,
standing
object.
We
take
if
it's
at
the
top
level,
then
then
it's
fine,
but
it's
got
to
be
a
list
inclusive
list
or
something
or
an
excluded
list
or
something.
But
we
need
to
do
that.
Otherwise,
yeah
we
blanket
take
it.
We
will
run
it
dishes.
C
Yeah,
I
think
the
w3c
don't
include
that
they
are
enumerated
out
separately
they're.
Not
this
is
the
the
page
and
all
the
resources
that
were
included
there
are
actually
separate
ones,
so
I
think
we're
right
for
that
by
following
the
w3c
spec
but
yeah
for
react
and
angular
and
stuff
like
that
that
create
their
own
timings
yeah.
That
is
a
different
story.
C
And
I
guess
who
are
on
the
subject
of
timing.
This
is
also
why
I
think
we
probably
should
be
attending
the
profile
meeting,
because
this
is
the
browser
form
of
profiling.
C
So
I
I
definitely
like
to
see
the
profiling
sig
use
the
events
api
for
logs
and
looking
at
their
doc
this
morning.
They
have
that
listed
as
an
option
in
their
kickoff
meeting.
C
I
think
we
want
to
definitely
push
them
down
our
path,
so
they
use
the
the
logs
event
api
and
they
go
in
to
find
their
own
events
with
their
in
their
own
domain.
For
that
stuff,
and
I
definitely
have
nested
objects.
E
A
F
C
Yeah,
I've
had
no
more
comments
from
from,
in
the
last
two
spec
meetings,
yeah,
so
yeah.
I
was
definitely
going
to
go
to
do
that
if
nothing
else,
if
we
just
define
it
for
logs,
because
it's
not
defined
in
there
today
and
I
I
want
to
try
and
expand
it
out,
I'm
pretty
sure
I'm
going
to
get
pushback
to
say
attributes
everywhere
can
be
nested,
which
is
what
I
want
the
comments
for,
but
I'll
start
the
pr
this
week
and
then
we'll
get
the
real
comments.
I
think.
E
Yeah,
it's
a
slight
digression
to
martin's
topic,
but
quickly
on
on
on
the
nested
attributes.
I
think
it
from
from
reading
the
comments
it
feels
like.
The
major
opposition
is
really
that
you
know
there
isn't
a
good
or
easy
mapping
to
non
otlp
destinations
like
jager,
and
you
know
other
folks.
You
know
they
how
they
handle
attributes.
You
know
we
are.
We
are
limited
by
that.
C
E
E
C
A
Just
I
have
a
question
about
this,
like
I'm
still
not
100
clear
on
when
we
wouldn't
use
nested
attributes,
because
we
we
talked
about
you
know
for
for
events,
different
events
like
having
the
value
actually
stringified
in
a
single
field.
So
when
would
be,
when
would
it
be
useful
for
us
to
have
nested
like
or.
C
Always
so
I
I,
I
think
effectively
shape
and
event
should
be
domain
name,
maybe
type
but
probably
definitely
type,
because
I
think
we're
going
to
have
some
sub
like
you've
got
here
with
timing
and
name.
So
if
you've
got
subtypes,
I
would
have
preferred.
I
would
would
have
preferred
to
see
a
single
name
field
which
included
all
of
that.
So
the
domain
was
the
first
part,
but
we've
already
already
got
an
agreement
where
that's
separate
so
be,
I
think,
domain
name
type
and
then
value
and
the
value
is
the
nested
object.
E
C
E
If
you're
going
that
route
nav,
I
I
sometimes
wonder
you
know
if
we
should
go
back
and
make
a
strong
argument
that
you
know
we
shouldn't
be
used,
lock
records
at
all,
because
you,
I
think
the
definition
that
you
are
giving
for
the
structure
of
an
event.
Is
you
know
it?
It
is
an
event
within
an
event
right,
there
is
a.
There
is
a
log
record
which
has
exact
same
structure
that
you
just
described.
C
Oh
yeah
we're
using
log
as
a
transport-
that's
really
all
we're
doing
so,
which
is
also
why
I
think
profiles
should
be
using
logs
of
transport
rather
than
defining
their
own
signal,
which
was
brought
up
in
the
the
spec
meeting
this
morning.
It's
like
well,
if
profile
is
going
to
be
its
own
signal,
then
eventually
its
own
signal,
and
if
we
just
get
too
far
down
that
path,.
D
E
E
You
know
people
might
stop
using
the
the
the
semantic
conventions
right.
I
think
you
know
the
the
current
convention
is
to
use
the
namespace
style,
the
dotted
notation.
C
Which
is
just
the
both
and,
like
you,
create
way
more
events
well,
depending
on
your
application,
you're
going
to
create
way
more
events
than
you
are
spans
like.
If
you
look
at
the
timing,
one
if
you've
got
a
a
span
with
an
ajax
request
in
that
how
many
events
we're
going
to
tack
into
that.
So
the
whole
idea
is
to
keep
it
as
small
as
possible.
E
Okay,
in
that
case,
should
we
extend
my
myotip
to
include
this
third,
the
event,
data
or
event
value
attribute
to
the
semantic
conventions.
C
No
because
I
I
think
the
it
depends
on
the
name
of
the
event
so
really
having
the
event
saying.
Domain
and
name
is
fine,
but
but
what
we?
What
we
have
to
define
here
is
for
this
timing
event.
So
for
argument's
sake,
we
call
that
the
domain
is
browser.
The
type
is
timing.
The
name
is
timing,
then
we're
defining
what
that,
what
the
shape
that
looks
like
so
people
know
how
to
pull
it
apart
when
it
gets
to
the
back
end.
C
So
some
event
names
might
also
need
some
other
higher
level
ones.
I
type
that
I've
just
been
talking
about,
but
I
think
we
just
define
you
know
what
are
the
standard
sets
of
attributes
for
that
event,
but
as
far
as
the
log
entry
is
concerned,
it's
it
always
has.
A
domain
always
has
a
has
a
name
and
then,
depending
on
the
name,
we're
defining
from
there,
which
is
the
document
that
martin's
working
on
which
ram
we
need
to
pull
apart
our
internal
stuff.
So
we
can
contribute
to
this
and
find
some
time.
E
C
C
I
agree
yes,
every
event
should
use
the
same
thing,
but
depending
on
the
name
of
the
event,
it
may
not
make
sense
that
it
has
data.
C
E
Okay,
the
is
there
any
expectation
on
these
keys
inside
the
data.
F
So
do
we
have
any
semantics
for
that
defined
yeah.
E
Because
I
already
see
you
know
there
is
a
you
know:
component
attribute,
you
know
today,
you
know
which
is
not
following.
You
know,
which
is
which
doesn't
have
a
namespace
so
is,
is,
are
we
allowed
to
have
attribute
keys
without
the
namespace
today.
C
I
believe
you
are,
you
can
send
anything
you
like
and
in
terms
of
third-party
names.
You
could
do
that,
like
people
can
send
their
own
events
today
right,
they
can
say,
expand
event
frame
and
put
whatever
they
like
in
it,
and
then
the
vendor
decodes
that,
based
on
their
own
requirements,.
C
And
I
think
what
we're
defining
here
is
like
the
existing
instrumentations
have
really
been
a
hack
to
have
zero
length
spans
to
descend
events
down,
span
events
or
just
attributes
for
a
zero
length
span.
I
think
what
we're
now
defining
as
part
of
this
group
is
okay,
we're
actually
going
to
set
in
stone
what
the
events
look
like,
and
if
that
means,
we
have
to
go
and
change
the
existing
instrumentation
so
effectively.
C
C
So
it's
part
of,
for
example,
as
part
of
the
document
load,
which
is
just
listening
to
the
page
load
event.
It
is
just
a,
although
I
think
the
existing
interpretation
tries
to
reach
into
the
resources
to
get
the
to
get
the
time.
We're
just
saying
it's.
It's
browser
timing,
you
know
type,
maybe
is
page
load
or
as
part
of
the
data.
Maybe
we
define
the
name
in
there
and
then
we
have
values,
and
I
don't
know
this
is
this
is
what
we're
trying
to
do
is
define
the
shape
of
the
event.
C
Okay,
but
yeah.
I
agree
that
you're
having
event
data.
Definitely
I
you
know
that
that
would
be
the
semantic
convention,
and
I
would
have
that
as
an
optional
field,
because
you,
you
will
have
events
which
are
just
domain
a
name.
It's
just
a
hey.
This
thing
happened.
E
E
You
know
which
are
which
contains
any
data,
and-
and
what
is
data
right
is?
Are
these
attributes
you
know
constitute?
Do
they
constitute
data.
C
E
You
know
what
what
I'm
trying
to
say
is:
let's
say
your
exception
right
exception
today.
Already
has
this
structure.
C
E
C
We
could
have
these,
this
is
an
exception
and
we
say
the
event
name
is
whatever
the
so.
The
event
domain
is
whatever
the
event,
name
is
exception
and,
as
part
of
a
an
event,
an
exception
event
we're
following
the
existing
conventions,
but
there
are
additional
exception
fields.
I
think
we
want
to
include.
C
A
So
I
mean
the
alternative
to
this
would
be
to
not
have
a
generic
event
event.
Data
attribute,
but
instead,
like
have
you
know
for
timing
like
you
would
have
a
new
namespace
timing
and
which
would
have
only
like
timing,
dot
data,
and
then
you
would
have
the
you
would
have
the
option
like
in
the
semantic
conventions,
if
you
want
to
like,
like
put
it
in
a
single
field
or
if
you
want
to
flatten
it
out
into
multiple
fields.
C
So
if
you
said,
instead
of
instead
of
being
event
dot
data,
it
was
like
event
name
dot
data
which
would
be
another
way
to
do
it
too.
But
I
I
I
don't
know
I
I
know
ted's
talked
about
in
the
past,
of
having
composition,
so
the
you
could
actually
merge
things
together.
E
Actually
that
that
might
be
relevant
right
here,
where
you
know
for
a
for
a
given
page
load
for
a
given
url
that
was
fetched,
there
could
be
different
types
of
you
know,
timing,
events
and
the
data
for
each
of
them.
You
know
you
would
want
to
put
them
in
in
different
objects.
C
E
Okay,
I
think
what
might
help
is
if
we
list
all
the
performance
timing
attributes
in
in
one
place,
then
I
think
it
will
give
us
a
good
idea.
C
E
Okay,
okay,
maybe
martin,
maybe
you
know
you
and
I
you
know
we
can.
You
know,
let's
work
on
this
offline.
A
Sure
yeah,
like
let
me
just
tell
them,
I'm
going
to
share
my
screen
one
more
time
and
just
go
through
like
some
of
the
things
that
I
have
been
on
my
mind
before
I
do
that.
A
Actually,
I
want
to
say
one
more
thing
about
the
event
data
attribute
so
like
the
first
step
that
I
was
imagining
that
you
would
do
is
is
in
the
spec
define
in
the
semantic
conventions
in
the
spec
define
the
event
dot,
attributes
yeah
so
right
so
so
it
would
be
even
what
when
is
event.name
domain
name
used
and
how
it
is
used.
A
C
Well,
there's
two
level
of
data
which
I've
been
trying
to
avoid
complicating
answer.
Your
first
question
we
would
say
for
this
domain
name,
then
it's
its
content
is
in
event
data.
C
There
is
effectively
a
fourth
field,
which
is
custom
data,
which
is
effectively
random
properties
that
the
application
or
the
developer
or
the
the
page
wants
to
include
with
the
event
which
don't
fall
into
the
resource
category,
because
they're
unique
to
that
event,
we
could
define
that
as
event
dot
properties
or
we
could
say
as
part
of
the
data
element.
It
has
a
field
properties.
C
C
They
come
from
the
application,
so
we
don't
care
about
what
they
are,
but
the
application
does
when
it
hits
the
back
end.
They
use
that
for
qualifying
whatever
they're
playing
with.
A
C
C
Yeah,
but
I
think,
as
part
of
the
event,
we
want
to
define
what
its
name
is.
Okay,
so
so
we
have
a.
We
have
a
root
name
that
says,
put
put
your
own
attributes
in
here,
and
it
is
it's
up
to
you
to
define
what
your
semantic
convention
for
that
is.
But
any
back-end
vendor
should
just
take
this
data
and
push
it
in
into
the
back-end
storage
for
the
application
to
do
with
what
they
want.
A
Yeah,
so
so
I
just
want
to
make
sure
that
that,
when
we're
defining
this
that
we're
clear
like
what
these
attributes
would
be
used
for
for
different
domains,
I
guess
like
because,
like
we're,
not
this
like
the
event,
namespace
would
be
used
for
not
just
browser
but
for
all
different
kinds
of
domains
like
going
forward
right.
So
so
I
think,
like
we're
right
now
we're
when
we
talk
about
putting
things
into
into
event.
Data
like
we're
thinking
about
up
can
optimizing
for
browser.
C
No,
I'm
I'm
thinking
about
optimizing
for
everything,
not
just
the
browser.
It's
like
you
know.
I've
worked
in
the
the
back
end
as
well,
so
you
know
I've
been
doing
stuff
for
quite
a
while,
and
it
is
just
a
case
of
from
a
memory
and
a
performance
perspective
with
your
serializing
and
holding
it
whatever.
C
Not
repeating
a
bunch
of
stuff
is
just
always
efficient,
which
is
why
I'm
in
two
minds
about
whether
we
say
the
data
includes
one
of
the
fields
of
data
as
a
as
a
semantic
invention
is
the
custom
properties
or
whether
we
say
okay,
let's
just
let
have
it
move
it
to
the
top
level
and
say
it's
event.custom
properties,
I'm
in
two
minds
for
that.
C
I
think
the
data
is
going
to
be
optional
depending
on
the
event
depending
on
the
domain.
So
I
yeah
I'm
in
two
minds.
A
A
A
C
Yeah,
probably
so
so,
in
fact,
I
I
think
it's
eventually
enumerating
every
hotel
defined
event
and
as
part
of
that
hotel
defined
event
it.
That
is
a
combination
of
domain
and
name.
So
do
we
list
that
based
on
domains?
Yes,
do
we
just
place
on
yeah,
but
yeah?
We
do
need
to
go
around
and
enumerate
them
so
grouping
them
by
domain
would
make
sense.
C
I
do
see,
however,
they're,
like
with
the
timing
event
case
here
with
w3c,
we're
very
browser
specific,
but
I
do
see
that
we're
going
to
have
common
events
that
are
shared
across
domains
so
defining
the
domain
is
browser,
while
it's
good
to
start
with
that,
I
I
think
we're
eventually
going
to
collapse
it
down
and
say
this
is
just
an
hotel
timing
event
or
it's
an
hotel,
something
event,
and
then
we
might
have
specific
ones
which
are
hotel
browser
as
the
domain,
for
example,.
A
Yeah,
so
I'm
I'm
thinking,
I'm
just
thinking
how
to
organize
organize
the
semantic
dimensions
once
we
start
defining
them
right.
So
maybe
maybe
we
can.
I
imagine
it's
gonna,
be
it's
gonna
be
different
from
what
it's
been
so
far.
So
far,
it's
essentially
just
like
there's
like
a
single
document
for
the
namespace
right
and
so
we're
gonna
have
something
like
either
we're
gonna.
A
Have
it
more
structured
like
domain
and
then
like
under
the
domain
folder,
you
would
have
individual
like
namespace,
and
then
you
would
also
have
a
common
common
namespace
or
common
set
of
attributes
or
or
we
do
something
like
we
list
the
wheels
like
the
there's,
a
timing,
namespace
and
but
but
when,
when
you
use
the
timing
namespace,
you
have
to
set
event
dot
domain
to
the
browser
you
know
so.
C
Else
so
we
we
say
this
is
what
an
event
looks
like
and
then
we
start
defining.
Okay,
and
these
are
the
sorts
of
events
you
get
based
on
the
domain.
So
these
are
shared
common
events
and
then
specific
events
yeah.
But
I
think
that
that
that
first
level
is
saying
this
is
what
an
event
looks
like
in
all
cases.
C
I
think
it's
a
start
and
we
will
probably
once
we
go
and
define
more
events
we'll.
We
may
come
back
and
alter
that,
but
I
think
if
we
just
say
an
event
looks
like
this:
okay
and
I
think
that
would
fit
nicely
in
in
the
standard
spec
but
yeah
when
it
comes
to.
How
do
we
enumerate
all
the
events
and
how
do
we
document
those
events?
That's
where
it's
going
to
get
interesting
because,
as
I
said,
we're
going
to
have
a
bunch
of
domain,
specific
ones
for
browser
and
common
ones.
C
So
we
want
to
say
well
for
a
browser.
Runtime
you've
got
the
set
of
common
ones
and
you've
got
the
browser
ones
for
mobile.
You
get
the
set
of
common
ones
and
the
mobile
ones,
rather
than
just
defining
them,
or
maybe
we
just
say
I
don't
know,
I'm
thinking
out
aloud.
C
There
is
a
button
prominently
and
like
likewise
the
profiling
one.
I
can
see
the
profiling
assuming
we
can
get
them
to
use
this
mechanism.
We
say
the
the
profiling
domain
will
be
the
profiling,
whatever
the
name
is
and
they'll
go
and
define
their
own
set
of
things,
which
is
probably
going
to
be
language
specific
because,
like
java,
which
seems
to
be
the
what
the
documentation
is
talking
about,
the
moment
is
very
heavy
on
its
set,
but
there's
also
a
common,
like
you
know,
cpu
and
memory
and
stuff
like
that.
C
A
Okay,
we've
got
to
open
it.
Five
minutes
left.
A
Let
me
just
just
say
one
more
thing
here,
like
I
started
going
down
the
list
of
options
just
for
the
completeness
of
the
thought
like
here,
so
I
was
thinking.
One
option
would
be
have
like
a
generic
event,
name
like
the
timing
or
performance
entry
and
then
all
the
details
having
in
the
value.
A
I
was
actually
originally
thinking
that,
like
that,
we
would
have
a
namespace
that
corresponds
with
the
event
name
but
sounds
like
we
prefer
to
have
event
value.
So
in
that
case,
and
like
then
list
the
the
attributes
you
want
to
pull
out
for
that
timing,
yeah
so
part
of
that
definition
then
would
be
some
kind
of
name
if
name
event,
which
would
be
like
it's
a
timing
and
the
name
would
be.
C
A
So
the
other,
the
other
option
that
I
just
wanted
to
list
for
completeness,
was
instead
of
having
a
generic
timing
name.
It
could
be
indeed
like
the
actual
individual
event
type,
which
is
from
from
the
w3c
spec
would
be
like
navigation
paint.
A
You
know
largest
contentful
paint
event
resource,
and
then
you
would
just
you
would
skip
the
type
and
you
would
just
have
the
the
fields
for
that
from
that
type.
Right.
C
C
Actually,
now
that'd
be
an
option
one
as
well.
It
would
be
a
timing
event
like
the
last
con,
the
largest
contentful
paint,
which
is
like
fcp
paint
and
the
event
like
the
marks
are
particularly
they
just
marks
so
that
that's
where
that's
really
a
case
of
there
is
no
data
for
a
mark.
Apart
from
its
time.
C
C
And
resources
is
sort
of
like
navigation.
This
is
just
the
timing
thing
so
yeah
I
I
would
prefer
option
one
because,
like
yeah,
if
you
start
listing
out
all
of
option,
two
we're
we've,
then
gonna
go
and
define
every
one
of
those
where,
if
we
just
say
we
have
naming,
and
then
we
have
type
and
then
type
follows
a
w3c
standard
of
these
number
of
fields,
then.
A
So
so,
then,
then,
going
from
that
like
if
we
have
a
single,
if
you
have
a
single
type
event
name
for
all
these,
all
these
things
then
like
in
that
spec,
like
for
timing
like
we'll,
have
to
list
out
all
of
these
individual
ones
with
the
attributes
we
want
to
include
for
each
of
them.
C
A
Yeah,
so
I'm
just
I'm
just
following
to
what
the
realm
ram
said.
Is
that,
like
we,
probably
that
we
don't
want
to
just
stringify
the
whole
performance
entry,
we
want
to
pick
out
the
values
that
we
that
we
care
about.
C
A
C
And
we
just
harvested
it
from
w3c
for
now
and
then
we
can
put
some.
We
see
wishy-washy
words
about,
if
they
add
new
ones.
We'll
do
this
to
follow
rams
thing.
We
just
say
we.
We,
we
don't
support
nesting
of
resource
navigation,
so
the
navigation
event
does
not
include
the
timings
for
all
the
resources,
for
example,
which
I
don't
believe
the
navigation
event
has
for
browser,
but
for
mobile
devices.
I
can
see
once
you
start
getting
those,
especially
once
they
come
next
week
when
stefan
gets
them
here.
A
Okay,
I
had
some
additional
questions,
but
let's
just
let's
just
continue
that
tomorrow,
because
I
think
we're
out
of
time.