►
From YouTube: 2023-02-15 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
A
B
Will
wait
a
minute
or
two
for
nav
and
Martin
to
join
and
then
we'll
start.
C
Yeah
Fred
asked
me
to
attend
this
meeting
and
we
need
a
boxer
from
the
PC
for
this
sig
I'm.
Not
yet
sure
I
will
be
that
person,
or
maybe
someone
else,
but
for
now
I'll
be
here
and
I
can
stay
for
half
an
hour
this
time.
Unfortunately,
I
have
another
thing
to
attend,
but
yeah
sure.
B
Okay,
I
think
everyone
is
here.
We
can
start
so
to
quickly.
You
know,
go
over,
you
know
the
things
that
we
have
made
progress
on
so
far.
We
have
we
needed
an
events
API
which
we
have
now
it's
it's
still
evolving,
but
mostly
I
think
you
know
we
are
good.
B
We
needed
a
way
to
further.
You
know,
put
our
our
data
into
it,
and
I've
suggested
that
we
put
the
payload
into
a
separate.
You
know
section,
you
know
we
started
calling
it
event.data
and
that
required
support
for
map
as
attribute
value
and
that
we
are
we're
close
to
getting
that
approved
now.
B
Telemetry,
you
know
for
the
client-side
instrumentation
for
the
browser
instrumentation
at
least
we
have
identified
a
bunch
of
events
that
we
have
documented
and
I'll
I'll
share
that
link
with
you,
I
think
the
most
important
item
that's
pending,
you
know,
for
which
you
know
we'll
need
your
help,
but
again
is
is
basically
there
are
some
parameters
we
think
are
applicable
at
the
resource
level,
but
they
they
also
change.
B
Instead
of
you
know
repeating
them.
In
every
you
know,
span
every
signal
spanner
log
record
can
I
be.
Ideally,
you
know
would
like
to
keep
them
at
one
place
and
also
some
additional.
You
know
background.
You
know,
prior
to
open
Telemetry
the
the
Legacy
systems
that
we
all
had.
They
didn't
have
the
concept
of
a
resource.
So
you
know
it
is
true
that
you
know
these
attributes
that
we're
talking
about
that
that
we
want
to
put
in
on
the
resource.
B
You
know
they
have
been
part
of
the
individual.
You
know
events
already,
but
given
that
there
is
an
opportunity
to
you
know
centralize
them,
you
know
we.
We
want
to
explore
that
because
I
think
you
know
given
the
overall.
You
know
other
reasons
why
you
know
the
the
size
is
getting
larger.
B
We
believe
this
will
be
a
good
benefit
to
reducing
the
size
of
the
payload,
and
there
are
several
terms
we
have
used
and
at
this
point
I
would
like
to
start.
You
know
maybe
using
the
term
state
of
the
resource.
So
you
know,
given
that
typically
State
changes
for
a
resource,
so
I
I
wanted
to.
You
know
go
back
to
that
discussion,
so
how
anybody
else
want
to
say
anything
on
to
add
to
this.
C
I
remember
we
had
the
discussion
about
the
session
ID
going
into
the
resource,
but
I'm
forgetting
the
details.
Now
I
will
need
to
go
back
and
read
the
the
proposal
there
and
and
then
maybe
we'll
see
what
we
can
do
right.
I
I
remember
one
of
the
things
that
prompted
this
was
avoiding
the
repetition
of
the
session
ID,
but
I
think
there
was
more
to
that
right.
It
wasn't
just
the
optimization
of
the
payload
side
so
but
I'm
forgetting
now.
What
was
the
other
reason
argument
in
favor
of
doing
this?
C
D
It's
a
case
of
for
rum
systems
rather
than
using
the
trace.
Id
people
tend
to,
in
fact
use
the
session
ID
to
link
multiple
page
loads
together
so
yeah.
If
you've
got
the
session
ID
across
everything,
it'll
help
group
them
together.
So
in
the
the
notes,
I
I
dropped
in
Ted
the
thermal
resource
Otep,
which
was
his
first
attempt
at
trying
to
get
that
through
I,
don't
think
it's
gone
anywhere,
though
so
I
think
you've
got
details
in
there
from
memory.
I
haven't
loaded
in
a
while.
C
Okay,
I'll
read
that
I'll
read
that
we'll
try
to
I
guess
go
back
and
remember
what
I
was
thinking
then
and
then
maybe
maybe
I
will
try
to
comment
on
the
old
app
if
I
have
anything
new
to
offer
there
right.
B
Yeah
in
that
context,
I,
you
know
there
was
a
discussion
around
adding
another
set
of
attributes
on
the
resource
to
identify
the
metric
I
think.
A
C
Yes,
there
was
one,
but
it's
it's
a
can
of
worms
that
we
do
not
want
to
open
prematurely
right.
It
has
consequences
everywhere
and
I'm,
not
sure
we're
ready
to
to
figure
out
what
that
means
right.
So
Josh
and
I
worked
only
to
beat,
and
we
saw
that
it's
just
it's
just
too
complicated
for
now
and
we're
there's
so
many
things
that
are
unfinished
there,
that
adding
one
more
new
complex
thing
to
open
Telemetry
didn't
seem
like
a
wise
idea.
C
So
maybe
I
guess
this
is
another
thing
that
can
be
driving
that
effort.
So
maybe
it's
time
to
go
back
to
that
right,
so
we
paused
it
intentionally.
So
let's
let
me
do
that
right.
Let
me
reread
the
old
app
about
the
session
IDs
and
see
I
guess
maybe
it's.
It
means
that
it's
time
for
us
to
also
rethink
that
the
resource
ID
needs
to
have
a
mutable
portion
right,
yeah
and
then,
if
that's
true,
then
then
it
solves
other
problems
as
well
problems
as
well.
So
maybe
we
should
do
that.
C
I'm,
not
sure
and
I
want
to
make
sure
we're
not
we're
not
unnecessarily
creating
more
complexity.
Here.
Maybe
there
is
a
solution
just
for
session
ID
that
that
doesn't
require
more.
Those
changes
will
be
very
expensive
with
the
resource.
The
smartest
immutable
in
many
places
and
making
this
change
in
a
way
that
doesn't
break
a
lot
of
stuff
is
quite
hard.
I
think
we
there
is
a
possibility
to
do
that,
but
it's
definitely
not
trivial.
So
I
prefer
not
to
do
that
right
if
there
was
another
way.
I.
C
Definitely
don't
want
to
to
to
to
do
that
right,
so
I'll
do
that
right.
Let's
after
this
meeting,
I
will
I
will
go
and
read
the
Otep
and
I
will
see
what
is
it
that
there
is
other
possibilities
and
what
can
I
offer
here.
B
A
B
C
Yeah,
so
that
that
that
one,
the
the
issue
about
resource
contributing
to
metric
identity,
it's
it's
a
very
small
slice
of
of
the
actual
bigger
problem
there.
Really
we
have
a
large
document,
Josh
and
I
wrote
together,
but
we
intentionally
did
it
didn't
present
it
to
open
Telemetry,
because
we
saw
it's
just
getting
unmanageably
large.
C
So
maybe
there
is
a
way
to
to
solve
this
as
a
kind
of
a
subset
of
that
problem.
I,
don't
know
right,
I'll,
think
about.
B
C
Yeah,
by
the
way
you
mentioned
one
more
thing
that
the
event
data
requires
that
we
allow
domestic
maps
in
the
attributes.
We
already
allow
that
for
log
records,
oh
yeah,
we
already
love
it.
So
that's
not
blocking
at
all
correct.
D
Correct
yeah
yeah:
that's
why
we
chose
logs
as
the
transport,
although
with
your
change,
which
it
got
like
loads
of
sign
up
last
week,
has
it
been
merged
yet
I
think
you
know
yeah.
D
C
We
had
a
few
people
who
had
strong
objections
and
I
don't
want
to
rush
and
merge
without
without
those
folks
having
an
opportunity
to
say
to
speak
up.
Yeah
yeah.
B
Okay,
any
other
topics
from
others.
C
B
The
session
ID
is
one
of
them.
Let
me.
B
So
we
have
a
bunch
of
attributes
that
we
have
identified
that
fit
into
that
category
and
session.
Id
is
one
of
them.
C
Sorry
they
change
I
mean
yeah,
they
change,
but
they
are.
They
need
to
be
the
same
across
a
batch
of
events.
Coming
from
from
that
resource.
B
B
Yeah
browser
the
client-side
instrumentation
is,
is
kind
of
very
different
from
the
life
cycle
of
the
back
end,
telemetry
I
think
the
key
difference
is
I.
Think
every
the
scope
of
each
object
is,
is,
let's
say,
a
Trace,
whereas
on
the
client
side,
a
sequence
of
traces,
you
know
you
want
to
link
them
together,
saying:
okay,
the
all
these
traces
were
emitted
as
part
of
this
user
session.
B
So
so
that's
why
there
is
a
different
dimension.
That's
unique
to
the
client
instrumentations.
C
And
the
window
size
obviously
changes
if
you
resize
the
the
window,
I
guess
so,
okay,
why
is
it
not
an
event
why
that
doesn't
need
to
be
a
resource
attribute,
I'm,
curious.
B
It
can
be
an
event,
I
think
it's
a
probably
you
know
kept
here,
because
if
it's
an
event,
you
know
what,
if
that
event
is
somehow
you
know,
lost.
D
Yeah
it's
a
case
of
yes,
you
could
have
a
resize
of
it,
which
would
change
the
size,
but
if
you've
got
a
page
actually,
where
someone's
clicking
on
a
page,
you
may
need
to
want
to
know
what
the
size
of
the
page
is
and
yeah
it
displays.
Different
different
teams
use
this
information
differently.
B
And
I
think
the
the
fourth
one
is
also
interesting.
These
are
for
the
single
page.
Applications
you
are
on
the
same
page
are
not
on
the
same
page.
You
know
you
could
be
on
different
pages
like,
for
example,
you
know
like
I
can
I
can
jump
to
any
hash
fragment
so
like
I
can
jump
to
to
any
any
place
within
the
page?
And
then,
while
we
are
in
that
state,
you
make.
You
know,
API
calls
from
the
browser.
C
So
the
the
page
is
not
reloaded,
but
the
URL
changes.
You
are
the
same,
the
same
page,
the
same
SDK
state.
So
we
use
the
SDK
is
not
reloaded,
but
the
URL
is
different
yeah
because
you
may
navigate
and
in
spas
you.
The
navigations
may
be
quite
significant.
It's
not
just
the
that
you're
jumping
to
a
different
anchor
right,
correct,
yeah,.
F
B
And
the
end
user
ID,
the
the
user
ID
also,
you
know
you
could
log
out
and
log
in
as
a
different
user
while
being
on
the
same
instance
of
the
page.
D
C
So
I
guess
what
can't
change
in
that
case
right?
Is
it
just
the
browsers
how
to
coded
things
that
cannot
change
yeah.
B
So
so
we
categorized,
we
have
two
tables
resource
and
resource.
In
a
very
you
know,
this
is
just
we
used
some
temporary
names.
What
cannot
change?
Is
the
user
agent
right
yeah?
Not
many,
though
yeah
I
think.
B
B
E
B
E
B
No,
that's
what
I'm
saying
so
I
think
for
me
for
our
you
know,
considerations
in
our
product.
You
know,
that's
not.
You
know
a
priority,
so
if
it
is
for
you,
maybe
then
yeah
I
think
you
should.
You
know
add
it
here.
C
So,
just
to
just
to
understand,
you
are
you're,
seeing
it
as
the
same
resource
until
the
the
window.
Unload
event
right
is
that
correct
and
until
that
event
the
resource
is
the
same
resource,
but
it
may
have
some
mutating
attributes,
obviously
correct,
but
then,
when
the
window
is
downloaded,
whether
it's
due
to
a
page
refresh
or
maybe
it's
a
pub
is
closed,
the
new
one
is
opened
or
whatever
that
can
be
considered
a
different
resource.
In
that
case,
right
until
then,
it's
the
same
resource.
That's
that's
how
you
see
it.
Is
it
correct?
Yes,.
A
C
F
C
So
I'm
not
sure
I
understand
that
the
resource
there's
a
there's,
a
there's,
a
web
page
or
an
SBA
which
is
loaded
in
a
in
a
browser
window
that
that
has
an
instance
of
open,
Telemetry
SDK
running
in
it.
That
opt-in,
SDK
today
requires
that
you
have
one
resource
associated
with
it
right
as
a
resource
that
will
be
also
included
in
on
all
emitted
telemetry
going
outgoing
from
that
SDK
instance.
F
I
just
mean
that
over
time,
like
I'm
gonna
I
wish
you
a
draw
and
sent
to
the
screen.
So
at
the
start
you
may
have
one
one
resource,
but
over
time,
for
example,
also
logs
in
here,
you
can
have
secondary
source
which
takes
over
so
the
proposal
for
inferior
resources
was
based
on
that,
instead
of
resource
being
given
to
the
SDK,
you
gave
a
resource
provider
and
dynastic
evidence
is
a
resource.
It
does
the
resource
provider
hey.
What's
the
current
resource
at
this
moment
of
time.
C
Yeah
I
I
get
it
so
I
you,
you
want
to
I
guess
to
model
the
the
the
attribute
that
can
mutate
over
time.
By
saying
it's
not
the
same
result.
The
resource
change,
I,
guess
yeah.
You
can
say
that
as
well.
The
other
way
to
think
about
that
is
it's
the
same
resource,
but
the
resource
has
two
portions.
One
is
immutable
and
the
other
is
mutable
portion,
but
it's
still
the
same
resource.
Well,
whatever,
whatever
way
you
describe
it,
I
guess
is
more
kind
of
semantics
right.
It's
conceptually!
C
No
different,
okay,
I
I,
understand
now
I
see
so
it's
it's
not
multiple
resources
simultaneously.
It's
it's
different
resources
over
time.
Right,
there's
one
resource
that
may
change
to
become
a
different
resource.
That's
one
way
to
describe
things.
The
other
way
is
to
say
it's
the
same
resource,
but
some
part
of
that
resource
has
changed
and
it's
still
the
same
resource.
The
next
result
is
probably
indistinguishable
in
the
telemetry.
A
B
Even
even
for
the
so
I
was
wondering
if
this
is
a
requirement,
only
I
think
you're
you're,
not
questioning
the
requirement
that
you're
just
saying
that
the
solution
is
is
hard.
You
know
without
having
side
effects
at
other
places,
correct.
C
I'm
not
questioning
the
requirement
I
understand
that
you
want
to
record
it,
but
at
the
same
time
I
guess
the
solutions
may
be
very
different
right,
so
we'll
need
to
figure
out.
How
do
we
solve
this?
Well
yeah
I
mean
what
what
they're
describing
is
obvious
right.
There's
a
page,
the
page
changes
state.
There
are
a
few
things
that
don't
change.
Other
things
do
change.
You
will
need
to
include
the
mutating
changing
portions
with
some
of
the
events,
so
yeah
I
mean
that's,
that's
understandable,
yeah,
no
I,
don't
need
to
question
those
I.
Think.
C
C
B
I
think
so
what
we
did
is
we
we
listed
I
mean
there
are
existing
implementations
of
the
different
instrumentations,
but
with
the
new
event
model
that
we
are
coming
up
with,
we
we
Revisited
and
we
have
identified.
You
know
the
structure
of
each
of
those
events
and
and
then
put
that
in
in
this
page.
This
is
still
in
my
Fork.
We're
planning
to
you
know,
raise
a
PR
against
the
spec
repo,
so
I
I.
Can.
Let
me
put
this
in
the.
C
C
C
B
Except
the
mobile
all
the
browser,
folks
yeah,
we
work
together.
Okay,.
G
I'm
here
I
am
curious
from
the
mobile,
with
my
mobile
hat
on
too
tight.
What
will
happen
when,
when
the
mobile
conventions
get
dialed
in
and
how
much
overlap
there
will
be,
and
whether
or
not
you
know
it
warrants
or
will
warrants
pulling
those
out
into
a
common
set
of
conventions
like
clients,
conventions
or
something
that
slightly
covers?
You
know
more
generic
and
covers
both
browser
and
mobile,
but
I
haven't
had
time
yet
to
go
over
the
list
and
get
a
strong
sense
of
how
much
overlap
there
is
yet.
C
C
G
B
B
Currently,
you
know
you,
you
emit
event,
so
if
there
is
any
way
to
represent
that
that
screen
position,
the
the
which
screen
you
are
on,
maybe
then
that
that
is
the
equivalent
for
this,
the
user,
ID
and
the
impression
ID
I,
don't
know
much
about
the
impression
ID,
but
the
user
ID
applies
that
you
are
on
the
same
mobile
app
instance,
and
then
you
you
can
log
out
and
log
in
as
a
different
user.
C
Yeah,
so
even
the
screen
screen
size
can
change.
I
think
you
can
do
the
split
string
thing
all
right.
You
can
put
two
apps
on
one
screen,
so
I
think
that
that
changes
in
the
app
as
well
so
I
think
there
is
a
lot
of
things
that
that
are
similar,
I'm
more
curious
about
things
that
are
different
there.
I
guess
right.
G
Or
some
things
that
might
might
have
a
a
similar
setting
in
Mobile,
but
just
aren't
that
important
to
the
use
case.
I
guess
you
know
like
if
if
there
is
a
user
ID,
but
mobile
people
are
more
sensitive
about
it
or
aren't
concerned
with
it
I'm.
Just
picking
on
that
as
an
example,
but
there's
bound
to
be
others
where
you
can
sort
of
find
this
similar
convention
in
Mobile.
But
it's
just
not
as
important
to
someone
doing
rum.
C
B
H
And
the
attribute
name
is
screen
reads
and
then
the
description
says
window
width.
B
D
Okay,
either
one
ma'am:
do
you
have
any
reference
I.
I
Was
I
was
going
to
refer
back
to
the
Excel
spreadsheet,
where
we
arrived
at
the
final
land
and
use
them
and
the
descriptions
from
there
I
don't
know
if
they
got
translated
really
well
so
I'm
gonna
go
look
at
that,
but
I
think
we
call
it
viewport
again.
We
went
through
all
these
discussions
right,
yeah.
I
The
spreadsheet
we
should
just
use
it
okay,
and
do
you.
H
B
Yeah
I
mean
there
are
other
places
too,
where
I
had
a
browser.page.url,
and
you
know
and
I
couldn't
answer.
Why
that
you
know
multiple
name:
space,
multiple
levels
of
instances
are
required.
B
Possibilities
of
you
know
additional
fields
in
that
namespace.
We
can
do
that,
but
I
think
the
general
tendency
is
to
avoid
in
adding
a
lot
of
hierarchy.
H
H
H
Space
then
it's
obvious
to
like
Apostles
that
are
related
right
because
they
belong
to
the
same
namespace.
But
otherwise,
if
they're
just
like
text
and
can't
infer
it
automatically.
B
What
did
you
share?
Martin.
E
Just
started
going
to
documentation
on
the
properties
are
on
the
window
of
the
browser.
I
mean
not.
Obviously
it's
a
lot
of
things,
but
yeah.
I
So
once
again,
we
did
yeah
I'm
gonna
you're,
welcome
to
look
at
that
spreadsheet
and
you
know
see
if
you're
already
capturing
anything
other
than
screen
within
height
from
the
viewport
and
bring
you
the
discussion
to
add
them.
I
But
we
did
go
through
across
all
of
the
vendors
that
participated
in
the
in
that
exercise
and
collected
all
of
the
schemas
that
they're,
using
and
and
figured
out
that
viewport
width
and
height
are
the
only
two
things
that
we
generally
cared
about
and
that's
reason
why
they
only
showed
up
and
nothing
else.
But
if
you
have
an
example
and
a
use
case,
I
think
we
should
discuss
about
that.
Yeah.
B
You
can
also
add
your
comments
in
this
PR.
Maybe
I
should
you
know,
raise
a
PR
against
the
the
the
public
open
Telemetry
report
now,
so
that
we
can
get
more
feedback.
Sooner
than
later,
we
are.
H
H
B
I,
don't
know
somebody
else,
maybe
T2
who
do
you
know
who
called
it
out
in
the
jsig.
E
So
I'm
not
sure,
if
that's
what
I
said,
but
I
would
I
would
say
now
that
the
interaction
spans
are
doesn't
mean
that
they
they
will
go
away.
I
mean
that
instrumentation
can
still
exist,
but
we're
proposing
that
events,
the
interactions
are
captured
as
events,
because
it's
more
it's
more
reliable
and
as
like
the
primary
method
of
collecting
interactions
with
the
events.
If
you
want
the
correlation,
you
can
still
have
instrumentation
that
creates
a
span
or
like
a
context
across
events.
H
E
It
might
be
possible
like
this.
That's
what
this
instrumentation
does
right,
yeah,
so
so
I
guess
this
doesn't
replace
the
the
interaction
span.
E
So
our
recommendation
would
be
to
use
the
event
because
it's
reliable
and
it
captures
what
what
actually
happened
reliably
the
the
span.
If
you
need
the
correlation
that
you
can
have
that
instrumentation
on
top
of
it
and
that
so
the
span
and
the
span,
would
you
know
these
all
these
events
that
happen
during
the
span
would
be
linked
to
the
sorry
all
these
events.
That
happened,
like
the
interaction
event,
that
happened,
would
be
linked
to
the
span,
interaction
event.
A
D
Yeah,
so
give
an
example:
maybe
so
if
you
have
the
instrumentation
that
in
fact
we
just
monitoring
clicks,
then
effective,
the
page
is
loaded.
There
is
no
span
that
spans
the
entire
length.
You
know
a
duration
of
the
page
because
it
could
be
days
so
when
they
click
on
a
leak
which
might
just
do
something
simple,
then
it
would
have
an
event.
However,
if
that
link
is
doing
a
navigation,
then
there's
part
of
that
navigation,
it.
It
will
very
likely
create
a
span,
and
then
anything
that
happens
during
that
duration
of
this
band.
D
E
So
we
have
a.
We
also
have
a
a
few
historical
documents
linked
at
the
top
of
the
this
meetings
notes
document.
E
A
B
Like
I
have
a
table
at
the
bottom
of
this
page,
where
I
I
said
you
know
the
current
like
what
is
the
proposed?
You
know
the
type,
whether
it's
a
span
event
or
a
span
or
an
event,
and
for
user
action.
I
I
kept
as
event.
H
B
So
maybe
yeah
this
this
one
is
is
still
up
for
debate
or
you
know,
discussion.
D
I
I
disagree,
so
I
I
think
the
the
payload
is.
The
user.
Action
will
be
the
event
data
whether
there
is
or
there
is,
the
span
will
be
up
to
the
application
to
come
to
hell
wants
to
represent
it
in
terms
of
semantic
conventions
which
ones
particularly
you're.
Talking
about
I
mean.
H
So
I
looked
at
the
current
instrumentation
for
user
interaction
and
there
are
some
attributes
there
that
doesn't
make
sense
and
I
looked
at
the
specification.
I
tried
to
find
like
semantic
conventions
on
what
should
be
recorded
on
such
spin
and
I,
couldn't
find
it
and
I
was
wondering
if
you
plan
to
to
work
on
it
or
not,
I'm,
not
sure
how.
D
It
goes
yeah,
so
so
as
an
event,
so
so
we're
we're
defining
that
an
event
will
effectively
be
a
combination
of
event,
name,
event,
domain
and
then
the
event
data.
So
rather
than
going
through
the
heavyweight
semantic
convention
process
for
everything
we're
going
going
to
end
up
saying
in
a
nice,
concise
document,
but
this
domain
name
combination.
This
is
the
data,
so
that
will
actually
so
we'll
have
like
four.
D
We
don't
have
the
event,
so
it's
okay,
so
the
domain
is
browser
and
the
name
is
user
action,
we'll
say
for
the
event
data.
It
will
contain
these
fields.
So
for
the
user
interaction
event,
we
will
Define
the
convention
of
what
it
looks
like
exactly
where
that
lives.
We're
still
open
for
debate,
we're
proposing
in
something
like
this
document
or
this
location.
But
we
have
had
pushback
from
tigrin
and
the
logs
team.
They
wanted
to
put
it
into
the
JS
repo.
D
We
several
of
us,
don't
believe
the
JS
repo
is
the
right
location,
because
this
is
going
to
be
hopefully
more
generic,
but
so
yeah
in
the
shorter
answer
is
yes,
there
will
be
some
conventions
for
declining
what
the
data
component
of
the
Avengers
that
we
will
Define.
In
fact,
the
spreadsheet
may
already
contain
the
fields
for
that
particular
one.
B
So
I
think
we
we
should
start
discussing
and
finalizing
those
aspects
as
well
now,
because
you
know
now
that
the
event
API
is,
you
know
we
we
have,
you
know
very
soon.
We'll
start
you
know,
prototyping.
All
of
these
user
action
is
something
that
we
might
take
up
with
the
you
know
towards
the
end,
but
at
least
the
the
initial
ones,
the
page
and
the
the
HTTP.
You
know
some
of
these
I
think
we
should
be
clear.
D
Yeah
I'm
actually
working
on
fixing
that
typo
in
my
to
find
the
shape
with
event
data.
So
we
can
try
and
get
that
in
because
I
think
that's
the
the
next
thing
we
want
to
get
in.
So
we
can
then
start
defining
where
we
put
these
combinations.
H
B
No
I
think
that's
what
yeah
there
is
there
is.
There
are
other
cases
too.
Where
there
is,
there
is
a
you
know,
things
are
mixed,
so
we're
not
the
first
ones
yeah.
D
Yeah,
the
the
spreadsheet
has
the
user
actions
already,
where
we
have
like
time
stamp.
Page
name
element
XPath
type,
yeah
we've
already
talked
about
it,
so
the
the.
If
you
look
in
the
spreadsheet
that
Ram
posted
there's
a
tab
in
there,
which
is
the
user
action
so
they're
the
proposed
names
that
we've
identified
so
effectively
it's
the
column.
A
is
what
we
said
will
be
in
the
event.
Columns
C
onwards
were
where
we
collected
all
the
event
names
used
by
existing
companies
to
try
and
derive
the
the
common
set.
B
G
Are
you
asking
specifically
like
about
mobile,
for
example,
yeah?
We
talked
about
this
a
little
earlier
in
the
call
back
when
tigran
was
on.
We
think
there
probably
is
some
overlap,
but
it
hasn't
really.
No
one.
No
one's
done
an
analysis
yet
of
like
what
which
conventions
probably
map
over
and
should
exist
in
both
so
that
Still
Remains
work
to
be
done.
H
D
Yeah,
the
original
goal
of
the
spreadsheet
was
to
try
and
collect
everything
from
browser,
mobile
and
all
the
rest
to
do
exactly
that
to
try
and
find
the
common
name,
but
when
we
first
created
them
that
we
only
had
us
browser
guys
involved.
So
that's
why
the
browser
namespace
exists,
but
we
haven't
set
them
in
stowed.
Yet
we
could
easily
remove
the
browser
name
prefix,
if
necessary.
G
B
All
right,
if
there
are
no
further
questions,
we
can
stop
here.