►
From YouTube: 2022-08-24 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
B
B
Okay,
I
guess
we
can
start.
I
don't
see
nav
on
the
call.
He
has
started
one
document
to
collect
the
events
from
different
vendors
and
and
with
the
intent
to
see
how
we
can
standardize
on
on
what
we
want
to
collect
for
each
individual
event,
type.
B
Please
fill
in
the
data
in
this
document
that's
mentioned
here.
I
think
I
say
nav
on
the
call.
Now
I
think
there
is
in
in
your
events
document
there
is
some
information
already
is.
Is
there
any
analysis
you
have
from
that?
So
far,.
C
B
Okay,
one
goal
I'm
trying
to
have
is
assuming
that
the
the
current
instrumentations
are
mostly
sufficient.
You
know
there
might
be
a
few
changes
we
could
make.
I
I
for
the
document
load,
fetch
and
xhr.
I
think
those
are
the
core
events
you
know
coming
as
spans.
B
B
The
it
is
from
from
the
spec
it
looks
like
the
only
way
to
do
that
is
to
mark
the
instrumentation
as
table,
but
it's
a
little
confusing
because
the
instrumentation
library
or
the
software
itself
may
not
be
stable.
We
only
want
to
have
the
telemetry,
that's
emitted,
you
know
to
this
table.
So
do
you
have
any
thoughts
on
how
to
distinguish
between
the
two
when
you
mark
something
as
stable.
C
I
I
think
it
only
gets
marked
as
stable
once
all
the
conventions
are
there.
So
I
I
think,
was
it
last
week
when
you
asked
daniel
the
question:
yeah
you're
gonna
get
the
same,
wishy-washy
question
answer
from
me:
it's
just
like
the
the
specs
need
to
be
stable.
First,
one
thought
I
did
have
is
what
one
thing
we
should
be
looking
at
doing
is
as
part
of
defining
our
rum.
C
Events
is
also
providing
a
definition
of
how
a
log
event
is
mapped
as
part
of
the
span
event,
especially
now
considering
it's
looking
like
the
span
event
or
sorry.
The
nested
attributes
is
going
to
be
potentially
never
happen
or
extremely
long
pole.
Before
that
happens,.
B
Yeah,
I
I
think
I
saw
one
of
your
tickets
where
you
you
suggested
that
we
move
the
performance
timing
data
into
one
one
single
event,
which.
D
C
Correct
yeah
that
way,
it
better
represents
what
what
martin's
already
defined
as
part
of
the
performance
metrics
but
yeah
having
individual
events
for
each
value
is
just
wrong.
I
know
t2
wouldn't
do
the
thing
in
the
zip
code.
Exporter
doesn't
currently
do
it,
but
it
just
means
we
need
to
pick
the
zip
code
exporter
as
well.
So.
C
Because
yeah,
it
really
is
it's
a
signal
event
that
happens
to
have
multiple
attributes.
So
I,
while
I
don't
like
it,
I'm
happy
to
say,
let's
just
change
it
to
attributes,
which
is
where
I'm
was
going
about,
help
defining
how
we
map
our
log
event
into
a
span
event,
because
at
the
end
of
the
day,
I'd
like
to
see
redefine
an
event
once
and
they
could
be
under
either.
One.
B
Yeah,
I
think
he
also
mentioned
that
even
with
zipkin,
the
spec
does
include
a
suggestion
to
you
know,
convert
the
the
span
events.
I
think
they
go
as
annotations.
C
C
Okay,
so
so
for
instrumentation,
there's
like
an
instrumentation
base,
so
if
that
isn't
stable,
then
anything
we
build
on
top
of
it,
it's
going
to
be
problematic
yeah
in
terms
of
like
what
I'm
going
to
interpret
like
I,
I
haven't
really
digested
all
this
discussion
because
I
said
I've
been
away.
Okay,
interpretation
for
not
yet
like
for
no
don't
care
from
about
browse
perspective
yeah.
C
If
the
spec
and
the
hispanic
infections
are
stable
was
like
it
was
just,
there
was
no
standardizing
instrumentation,
it
went
stale.
C
Yeah,
I'm
not
really
following
the
fourth
point
but
define
how
which
is
different
yeah.
I
don't
know
it's
a
I
I've,
really
you
know
being
on
the
the
west
coast.
You
know
I
I've
been
back
at
work
for
10
minutes
now
so
but
I'll
go
through
these
and
have
a
look.
I
think
you
had
a
link
in
here
as
well
for
something
that
I
saw
so
I
had
a
quick
look
at
this
before
jumping
on
the
call.
A
I
don't,
I
do
just
quick
comments
so,
like
you,
you
talked
about
going
through
the
existing
instrumentations
like
document
load
and
fetch,
and
that
we
have
in
contrib
yeah.
B
Because
I
felt
those
are
the
core,
the
most
immediately
relevant
ones
as
the
first
ones,
to
start
with.
A
So
as
we
define
as
we
define
these
semantic
conventions,
and
also
should,
I
think,
should
we
maybe
take
each
of
those
instrumentations
and
talk
about
like
whether
it's
still
like
what
changes
would
need
to
be
made
need
to
be
made
to
that
instrumentation
also,
if
it
still
makes,
makes
change
sense,
because
you
know
right
right
now,
like
you
said,
like
they're,
creating
spans,
but
some
of
the
some
of
the
attributes,
what
the
data
we
want
to
capture
as
events.
A
So
are,
we
are
we
going
to
be
updating
these
these
packages,
or
are
we
going
to
be
deprecating
them?
Are
we
going
to
be
supplementing
them
with
other
packages.
B
Yeah,
I
have
listed
a
few
in
items
here
so
so
we
we
need
to
rework
the
events.
The
the
performance
timing
info.
B
There
is
a
ticket.
There
is
an
issue
for
that
and
you
know
couple
other
minor.
You
know
attributes
we
need
to
remove.
B
C
B
Yeah
other
than
this,
I'm
not
very
familiar,
you
know
what
else
you
know
to
be
considered.
I
do
have
you
know
the
spans
in
in
json
form,
I
can
put
them
somewhere
and
give
the
links.
You
know.
If
you
want
to
review
them,
you
could.
A
C
So
it's
probably
to
more
directly
answer
your
question:
martin,
rather
than
rework
the
existing
ones,
it
might
be
safer
just
to
say
we're
going
to
introduce
a
different
set
that
conforms
to
the
the
definition
we
define,
because
that
no,
no
doubt
there
are
people
out
there
who
are
currently
using
these
the
way
they
are
so
they're
just
going
to
change
it,
we're
going
to
break
them
so,
but
if
we
have
a
different
set
that
can
conforms
to
the
I'm
going
to
use.
The
word
correct,
but
I
know
people
complain
about
that.
C
The
correct
way
of
doing
things
or
the
the
specified
way
of
doing
things.
It
would
just
be
easier.
C
Yeah
just
sort
of
mark
them,
as
deprecated
saying
you
know
with
the
message
saying:
please
go
use
this
other
one,
but
you
know
it's
gonna
be
probably
years
before
people
stop
using
it.
Yeah.
B
Exactly
I
think,
I've
like
I.
I
feel
that
the
changes
are
not
too
many.
I
think
that's.
The
first
thing
we
need
to
you
know
identify
like
what
are
the
changes.
If
it's
only
these
changes
that
I
have
listed,
then
then
I
think
it's
good
to
just
change,
but
we
give
a
path
right.
We
we
don't
change
it
right
away.
We
give
a
flag,
you
know
to
you,
know.
B
So
so
those
on
the
call
you
know
who
are
already
using
these
instrumentations
do
they
have
any
feedback
on
whether
it's
working
well
for
them
or
do
they
want
to
see
any
changes.
E
I
think
I
I
feel
like
this
is
maybe
off
topic,
but
I
think
more
than
the
content
like
as
a
user,
I'm
more
interested
in
seeing
like
the
minification
and
like
browser
friendly
stuff,
because
it's
just
like
some
of
these
packages
are
it's
just
really
big,
to
include
open
telemetry
in
in
a
browser
app.
So
that's
the,
I
think,
that's
more
the
feedback
than
actually
like
any
specifics
of
the
package.
No.
B
That's
that's
an
orthogonal
topic
right.
I
think
we
have
had
several
discussions
on
on
this
topic
and
and
yeah.
If
there
are
any
suggestions,
please
let
us
know,
but
the
suggestions
were
to
use
protobuf
and
and
then
compress
it-
and
I
know
compression
you
know
is,
is
has
a
lot
of
you
know
issues
that
never
has
pointed
out,
but
there
is
a
compression
api
that
the
browsers
are
introducing.
So
you
could
certainly
use
the
compression
api
yeah
to
compress
the
the
payload.
A
So
for
video,
do
you
mean
the
the
size
of
the
payloads
or
like
the
size
of
the
binder.
B
Yeah
the
size
of
the
js
bundle
that
I
think
nev
is
working
on
it
right.
I
think
he
yeah
never
do
you
want
to
talk
about
that.
C
Yeah,
that's
that's
the
first
goal
of
the
the
sandbox
I
just
hadn't
jumped
over
had
a
look
at
the
the
pr
that's
out
there.
It's
still
blocked
because
of
like.
When
was
this.
It
was
a
really
old
commit
51b
958e
that
was
signed
with
the
the
github
actions
bot
so
effectively
where
the
sandbox
is
at.
Is
it's
taking
a
complete
copy
or
merging
the
js
and
the
api
repos
over
into
this
new
sandbox
repo?
C
So
it's
got
the
complete
history
and-
and
that's
where
it's
blocked
at
the
moment,
because
there,
because
of
this
one,
commit
they've
now
locked
it
down,
so
that
everyone
has
to
sign
the
cla
except
the
github
actions.
Bot
can't
sign
the
cla.
B
But
what
is
the
approach
you're
taking
nav
to
introduce
the
libraries.
C
So
I
I
have
lots
of
different
techniques.
I
was
gonna
do
in
a
different
branch.
Let
me
there's
like
1000
positive.
Actually,
let
me
see
I
can
find
the
readme
reference
where
I've
actually
highlighted
what
I
plan
to
do
here,
but
yeah
there's.
Oh,
I
just
did
that.
Okay,
okay,
I'm
going
to
go,
find
it
and
I'll
post
it
in
the
chat
later
but
effectively.
C
C
The
only
way
to
stop
that
is
to
effectively
unname
space
them
so
provide
helpful
functions
and
helpful
implementations
and
that's
a
big
radical
change
from
where
I
tell
this
today,
which
is
why
I
proposed
creating
the
sandbox
and
why
the
sandbox
got
created
so
that
I
can
just
go
in
there
and
beat
the
crap
out
of
it.
It's
part
of
my
original
pr,
where
I
just
copied
the
files
over
as
part
of
the
build
in
the
sandbox.
It's
also
creating
effective
web
ready
bundles.
The
idea
behind
those
is
we.
C
I
can
then
create
tests
on
those
to
say.
Okay,
this
prebuilt
bundle
should
be
no
bigger
than
x.
E
C
E
No,
that
makes
sense.
Thank
you.
I
didn't
mean
to
derail
the
discussion
about
the
api
itself.
I
was
just
giving
my
feedback
that
I
think
mostly
using
the
packages
has
been
fine
for
for
us,
but
it's
it's
really
more.
The
bundle
size.
B
Yeah
sure
I
think
there
are
two
independent
aspects
right
I
think
for
for
for
the
team.
That's
working
on
the
the
agents
themselves.
I
think
yeah,
the
the
bundle
size
is
important.
How
it
performs
on
the
wire
is
important,
but
there
are,
there
is
a
team
on
the
back
end.
You
know
that
consumes
the
data
that's
being
sent,
and
you
know
they're
asking
us
to
to
freeze
you
know
the
the
structure
of
the
data
that's
being
sent,
and
it's
for
that
purpose.
B
In
fact,
there
is
a
there
is
a
specification
specifically
talking
about
telemetry
stability.
It's
a
little
interesting.
The
title
says
telemetry
stability,
but
it
is
the
instrumentations
that
are
marked.
You
know
stable
or
unstable.
B
A
So
I'm
guessing,
I'm
I'm
wondering
if
we
could
use
this
the
concept
of
schema
for
for
this,
so
that
we,
you
know
if
he,
if
he
starts
sending
a
specific
schema
like
if
it
if
it
changes
at
some
point
like
that
it
does
an
indication
to
the
backhands
that
you
know.
B
A
Yeah,
I
don't
know
if
it's
marking
the
schemas
table,
but
it
will
be,
would
be
versioned.
So
if,
if
you
updated
the
instrumentation
to
a
different
version
of
the
schema,
then
there
would
be
a
you
know
way
to
for
for
the
consumers
to
know
how
to
translate
the
new
data
to
the
previous
one.
A
B
Yeah,
I
think
they
only
talk
about
attributes
or
I
haven't
really
reviewed.
The
current
schema.
A
I
would
think
so
because
it's
ub
like
if
you
remove
or
add
attributes,
then
the
schema
should
handle
that.
B
Yeah,
but
these
this
is
not
attributes.
This
is
one
event
versus
multiple
events,
so
I
think
the
the
ask
in
this
is
you
know
today.
We
we
send
this
performance
timing
entries
as
individual
events
yeah,
but
instead
they
ask
is
to
have
just
one
event
and
and
have
this
as
attributes
on
that
event.
B
Yeah
sure
yeah,
so
it
means
so
that
is
a
big
structural
change.
C
Yeah
that
that's
a
breaking
change
so,
which
is
why
I
put
in
there
sort
of
marking
it
as
deprecated,
because
I
don't
think
we
can
remove
it,
because
I'm
almost
positive
someone
is
using
it
today.
B
A
lot
of
people
use
it
for
sure.
I'm
just
saying
that,
given
that
these
are
not
stable
anyways,
you
know
if
we
work
towards
the
goal
of
marking
them
stable,
then
you
know
it
it's
okay
to
make
this
change
and
give
a
path
for
folks
that
are
relying.
B
C
There
yeah,
but
it
does
raise
probably
a
side
question
rather
than
using
the
schema
if
we,
if
we
like
today,
we've
defined
events
as
having
a
name
and
a
domain,
but
if
we
also
potentially
have
an
optional
version,
so
we
say
our
first
version
has
no
version
and
then,
if
we
have
to
introduce
something,
that's
not
an
optional
extra
attribute
or
value.
C
We
introduce
a
version
field,
so
we
can
say:
okay
well
version
x,
sorry,
event
x,
we're
going
to
like
rev
that
past
then
event
y,
because,
like
some
of
these
events
are
just
going
to
like,
become
stable
and
stay
that
way,
but
we're
probably
going
to
have
new
events
in
the
future
that
you
know
we
want
to
keep
working
on
and
rather
than
having
that
as
a
schema
level
change,
which
would
be
the
entire
event
definition.
C
Yeah
the
the
concept,
the
only
it
exists
today,
based
on
what
joshua
was
talking
about
at
the
schema
level
that
you
were
talking
about,
martin
so
yeah,
but
I
think
we
want
to
play
a
little
bit
more
fine-grained.
So
and
I'd
say
for
the
first
one
we
can
say,
though
there
is
no
version
so
yeah
effectively.
If
the
version
attribute
is
missing,
then
it
is
version
one
and
we
only
introduce
it
if
we
have
to
later,
because
we've
got
something
that,
and
that
would
be
a
case
of
where
you've
got
a.
C
B
C
Yeah,
but
we're
talking
about
how
the
events
are
effectively
included
in
spain,
which
is
where
I
spoke
earlier.
We
should
have
a.
We
should
come
up
with
a
standard
way
and
we
probably
should
have
a
spec
for
it.
This
is
redefine
event
x
and
generally
events
are
mapped
from
a
log
event
to
a
span
event
like
this,
which
is
probably
going
to
be
instead
of
being
event,
data
affecting
the
event
data
becomes
individual
attributes,
and
then
we
get
into
a
discussion.
Do
we
worry
about
event
data
because
of
that?
C
Because
of
the
nested
battery
issues
I
I
spoke
to
our
back
end
guys
before
I
I
went
on
leave
and
apparently
the
hlp
otlp
receiver.
They
have
does
not
handle
nested
attributes
at
all,
so
they
can't
even
handle
general
logs.
At
this
point
they
were
even
unaware
that
that
existed
so.
C
Yeah,
but
at
the
moment,
where
they're,
primarily
focusing
on
the
server-side
stuff,
so
they're,
not
even
thinking
about
like
I,
I
don't
have
a
way
to
send
it
any
even
these
existing
client
events
to
them
so
because
they
have
no
mapping.
C
C
You
know
given
a
a
effective
formula
for
like
a
an
algorithm
so
that
we
don't
say:
okay,
we
have
document
load
events
which
you
can
send
in
a
span
as
a
span
event
like
this
or
you
can
send
it
as
a
log
event
like
this.
B
But
but
that
is
again
an
incremental
feature
right,
so
so
on
the
testing
api,
you
know
we
could
introduce
a
tracer
option
to
to
basically
emit
the
span.
Events
as
independent
events
use
the
logger
api
and
emit
them
as
independent
events.
A
C
We
never
know
different
people
will
want
to
do
different
things.
C
Depends
I'm
suggesting
hotel
should
be
able
to
define
it
and
treat
them
it's
either
way
in
terms
of
specific
vendors.
No
for
a
vendor,
they
may
say:
okay,
we're
only
going
to
have
span
events.
C
So
therefore
you
use
this
set
of
instrumentations
or
you
set
this
config
flag
on
as
part
of
the
distro
and
if
they
want
to
move
to
it
to
a
different
vendor,
then
the
other
vendor
says:
okay,
well
we're
using
that
that
instrumentation,
but
you
need
to
change
this
config
flag
or
you
need
to
give
a
different
instrumentation,
and
that
would
just
be
part
of
the
whatever
the
definition
of
a
distribution
is
that
the
vendor
has
is
what
would
be
there,
but
under
the
covers,
it's
all
just
hotel
code.
A
Okay,
but
it's
essentially
we're
kind
of
admitting
that
that
we're
going
to
be
supporting
both
logs
as
events
and
span
events,
and
they
should
be
basically
the
same.
C
Oh
yeah
yeah
they
should
have
the
same
set
of
data
yeah
unless
we
do
have
an
event
like
logs.
That
is
nested
that
we
can't
represent
in
a
span
event,
but
even
then
the
attribute
definition
today
talks
about
if
it's,
if
it's
like
a
nested
structure.
So
if
it's
not,
you
know,
string,
ins
etc.
It
should
be
represented
as
json,
so
there's
already
sort
of
a
mapping
for
how
a
nested
attribute
should
be
represented
today
on
a
span.
B
Okay,
so
so,
if,
if
you
think
we
should
work
on
defining
a
schema
for
each
of
the
events,
yeah,
which.
C
Is
the
document
that
random
started
he's
actually,
which
is
probably
why
he's
not
here,
yeah.
B
Have
you
folks
looked
at
json
schema,
it
allows
you
to
define
a
schema
for
for
your
for
your
json.
We
could,
we
could
define
the
schema
with
this.
It
seems
to
be
part
of
an
ietf
standard,
are
getting
there.
B
No,
I
think
these
two
somehow
have
to
be
combined
in
the
semantic
conventions.
Semantic
conventions
are
only
talking
about
the
individual
attributes
right,
but
but
if
an
event
has
to
have
these
four
attributes
versus
an
event
p
having
these
three
attributes
and
what
is
the
data
type
for
each?
You
know
if
you
know
we
could
get
to
that
level
of
specifying
the
event
structure.
B
C
B
Because
I
I
think,
if,
if
we
were
to
just
stick
to
the
current
semantic
conventions,
then
then
then
we
already
have
it
right.
I
think
we
could
define
the
semantic
convention
for
for
each
of
the
attributes
and
and
and
then
you
know
be
done
with
it,
but
is
that
all?
B
C
That
that
I
think,
is
really
defining
it.
So
if
we
stopped
it,
if
we
stop
sending
a
required
attribute,
that
would
be
a
problem
so
from
a
validation
perspective
that
become
invalid.
C
But
if
we
stop
sending
an
optional
attribute,
then
I
think
it'd
be
fine.
B
Missing
today,
is
you
know
if
it
doesn't
define
the
context
in
which
the
attribute
is
missing,
attribute
is
required.
A
Yeah
yeah
it
is,
I
mean
it
does
say
that
it's
optional
required
or.
A
Right
right,
so
so
so
yeah
we
have,
we
still
have.
We
still
have
to
get
to
figure
out
how
to
define
semantic
dimensions
for
individual
events.
So,
like
you
know,
the
the
current
semantic
conventions
are
kind
of
tied
to
not
you
know
not
types
of
spans.
A
I
mean
they're
tied
to
like
namespaces
so,
but
I
think
at
least
my
understanding
was
that
we
were
thinking
about
going
to
in
the
direction
of
you
know,
supplementing
the
current
current
or
how
organized
the
semantic
dimensions
are
with
with
advanced
event
types,
and
I
had
that.
I
had
that
pr
still
open
in
my
in
my
repo
for
like
the
navigation
and
the
resource
timing,
events
where
I
proposed.
You
know
that
what
day
that
would
look
like,
I
can
send
it
to
you
again.
A
C
So,
probably
to
answer
your
question
santosh:
if
we
can
do
it
with
the
ammo
file,
then
we'll
do
it
with
the
ammo
file.
If
we
can't
do
it
with
the
ammo
file,
then
yeah,
maybe
we
go
to
find
something
like
the
json
schema
or
maybe
we
can
try
and
create
json
schema
from
the
ammo
file,
but
yeah
like
what
martin
said,
we
we
really
want
to
define
it
and
say
this
event.
C
A
So
sometimes
take
a
look
at
I
I
added
a
link
to
that
pr.
You
know
you
don't
have
to
look
at
it.
You
know
go
through
that
today,
but
it's
maybe
look,
take
a
look
at
it
and
see
if
that,
if
you
have
any
questions,
if
it
makes
sense
yeah
that
one.
C
C
Well,
this
is
the
navigation
event,
so
so
navigation
browser
is
the
name
domain
and
then
the
accurate
table
below
is
the
is
the
values
that
would
be
on
this.
Okay.
A
Yeah,
the
idea
was
that,
under
semantic
conventions
like
you
would
introduce
folders
for
domains
and
under
those
folders
that
you
would
have
you
know,
files
for
individual
events
with
their
semantic
conventions.
B
Okay,
I
have
gone
through
this
once,
but
let
me
go
through
it
again.
It
looks
like
I
haven't
fully
absorbed
everything
here:
okay,.
C
I
haven't
looked
at
the
doc
to
see
where
it's
updated
to,
but
it
really
is
a
case
of
ram
is
going
to
be
out
for
the
next
two
weeks,
so
it's
unfortunate
that
he
could
make
it
today
going
through
and
either
creating
separate
working
groups
to
say,
okay,
this
set
of
people
who
were
who
nominated
themselves
in
the
doc
we'll
go
and
effectively
define
each
individual
event
rather
than
having
you
know,
all
of
us
doing
it.
C
I
think
santosh
you
penguin,
wanted
me
to
get
the
the
ajax
one
which
I
can
go
harvest.
It's
called
a
dependency
from
a
microsoft
perspective.
B
C
Yeah
yeah,
I
should
be
able
to
put
that
into
this
stock
and
yeah,
and
once
we
go
through
and
define
it,
which
looks
like
we've
already
got
an
open,
telemetry
web
and
when
someone's
put
in,
we
then
take
their
stock
and
start
creating
specs
from
it
unless
it
needs
more
work,
because
realistically
this
should
be
straightforward.
We
just
need
to
come
to
an
agreement.
Okay,
these
are
the
set
of
values
that
we
want
to
have
and
if
vendor
a
calls,
it
fred
and
ben
toby
calls
it
bob.
A
C
C
C
And
just
get
defined,
and
then
we
can
say:
okay
well,
probably
a
bit
like
in
the
javascript
sig.
We
now
have
that
15
minutes
at
the
end.
Maybe
we
then
start
saying:
okay.
Well,
let's
pick
on
events
and
start
going
through
them
once
we've
got
them
defined.
C
Okay,
yeah
and
I
think
that
way,
that'll
help
you
santosh
in
terms
of
your
back
end,
guys
saying:
okay,
this
is
what
it's
going
to
look
like
and
you
can
work
with
them
to
to
see
if,
if
it
looked
like
that,
whether
it
worked
for
them
or
how
much
effort
that's
going
to
be
for
them,.
B
Okay,
any
other
topics.
A
The
only
the
only
other
thing
that
I
wanted
to
I
just
mentioned
is
I
have
I
have
that
pr
open
for
logs
api
in
javascript
sdk.
I
think
it's
getting
close
to
being
being
looking
good.
So
if
you,
if
you
haven't,
looked
at
it
yet
and
you
care,
please
take
a
look,
I
think
the
most
important
thing
there
is
just
the
shape
of
the
api,
which
is
in
the
the
types
like
the
the
logger,
which
methods
the
logger
has
how
it's
created.
A
I
think
once
once
that's
merged,
then
then
we
can
start
working
on
the
log
sdk
exporter,
yeah.
A
A
E
Yeah,
I
I
have
a
branch
going
and
it
ended
up
being
a
little
bit
more
complicated
than
I
thought
so.
I'm
adding
browser
support
for
basically
to
make
it
happen.
You
have
to
add
browser
support
for
the
sdk
trace
base,
sorry
exporter,
proto,
exporter
base
and
there's
another
package
as
well,
and
we
also
have
to
make
the
send
function
able
to
send
buffers.
So
I'm
I'm
slowly
working
through
that.
I
have
time
this
week.
So
if,
if
I'm
not
done,
then
I'll-
probably
just
what
do
you
think
is
the
best
way
like?
E
A
Yeah
yeah
totally,
we
can
help
out
some
if
you,
if
you
need.
C
Probably
the
other
thing
that
I've
been
noticing
a
lot
of
stuff
in
terms
of
the
actual
browser
matrix
that
otol
supports.
It's
really
only
going
to
be
the
latest.
I
think
I
vaguely
recall
seeing
symbols
in
the
existing
code,
which
won't
work
for
any
version
of
ie.
C
I
think
some,
some
safari
or
opera
browsers,
don't
support
symbols.
So,
unless
you
happen
to
use
like
some
sort
of
trans
transpile,
the
matrix
is
like
the
latest
and
that's
it-
I'm
uncertain
how
to
treat
this
with
the
the
sandbox.
I
think
it's
too
late
to
try
and
go
and
add
backward
supports.
I
think
it's
just
going
to
be
case
of
hotel
will
be
this
min
bar
and
that's
it
so.
Anyone
who
needs
to
target
all
the
browsers
is
going
to
have
a
problem.
C
It's
but
not
gonna,
be
fun
for
me
either,
because,
while
our
next
version
we're
dropping
es3
support
like
a8,
we're
still
supporting
ie9
to
11.
D
Sorry
I
mean,
if
you
want
to,
if
you
want
to
talk
about
transplanting.
To
be
honest,
all
of
the
class
usage
is
a
lot
of
wasted
space.
Likewise
yeah,
including
instrumentation
based.
I
would
rather
just
completely
nuke
that
and
just
make
it
a
function,
just
run
it
once
and
it
you
just
default
apis
but
yeah.
Whatever.
C
Yeah,
I
agree,
I
would
prefer
to
see
everything
as
interfaces
and
then
we
just
have
functions
that
return
objects
that
conform
to
those
interfaces.
So
we
can
do
exactly
that
because
that's
a
big
problem,
which
is
actually
one
of
the
things
I
was
going
to
play
with.
D
C
I've
been
playing
around
with
that
so
because
yeah,
the
ultimate
goal,
is
we
want
someone
to
go
and
create
an
instrumentation
using
the
existing
api
definition,
and
if
we
have
to
create
a
completely
separate
web
api
that
should
be
able
to
load
and
run
that
on
top,
it
should
be
compatible.
C
And
until
this
pr's
get
merged
no,
so
the
the
current
pr,
I
said,
is
blocked
because
of
this
cla
bot,
which
supposedly
got
raised
in
the
dc
meeting
last
thursday,
but
I
have,
let's
have
a
chance
to
catch
up
with
daniel
since
then
there's
one
more
pr
that
has
to
happen
after
that
one.
So
this
first
pr
is
really
just
bringing
over
the
history
into
this.
Like
auto,
merge
thing
so
effectively,
it's
just
a
script
that
will
run
will
run
regularly
to
bring
over
any
any
new
code.
C
The
second
script
will
then
massage
that
merge
data
into
the
form
that's
going
to
be
compilable,
it'll,
be
it'll,
use,
rush
and
create
the
the
bundles
then
that'll.
That
will
be
our
main.
So
our
main
branch
so
effect
would
be
two
automated
scripts
to
get
from
js
api
into
into
that
definition,
and
then
we
can
go
start
creating.
You
know
separate
branches
to
work
on
these
different
things.
C
Yeah
yeah,
like
in
terms
of
the
second
step,
my
very
first
pr,
which
I
it's
now
closed,
but
it's
sitting
on
the
sandbox
as
pr
one
brought
all
that
over,
but
it
was
like,
like
1400
files
or
something-
and
you
know
daniel
said
it
was
un
reviewable,
so
he
wanted
the
history.
So
that's
why
we're
going
down
this
path?
C
Okay,
that
does
mean
my
mundane.
Script
has
to
be
a
little
bit
more
intelligent,
because
one
of
the
things
has
to
do
is
effectively
rename
all
the
packages
to
put
sandbox
in
them.
So
it's
like
open,
telemetry,
sandbox
dash
and
whatever
the
existing
name
is
which
my
script
did
have
in
it,
but
I'll
have
to
make
it
probably
slightly
smarter,
because
I
want
to
make
it
completely
automated.
A
A
Okay,
if
not,
then
thank
you
and
talk
to
you
next
week.