►
From YouTube: 2022-03-02 meeting
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
B
A
A
Both
added
xhr
and
fetch,
at
the
same
time
you
added,
usually
I
did
funny.
C
A
A
I
wonder:
where
does
this
log
get
duration?
Does
it
have
some.
B
D
D
No,
it's
too
warm
for
snow,
okay,
you're
in
portland.
We
got
a
little
dusting
of
snow
with
like
last
week
in
february,
but
not
very
much.
D
B
So,
john,
just
to
get
you
up
to
speed
if
you're
interested
we've
been
working
on
the
prototype.
B
And
made
some
progress,
we
added
an
api
kind
of
a
draft
api
for
capturing
events
in
the
javascript
sdk,
and
then
we
updated
a
few
instrumentations
to
capture
events
and
also
have
been
working
on
visualizing.
D
Well,
so
is
that
using
no
logs,
just
all
as
spams.
D
B
D
E
D
D
A
B
Yeah,
I
mean
so
what
would
you
like?
There
certainly
are
some
some
kind
of
most
important
measurements
that
you
want
to
get
from
all
sessions
like
the
count
of
views
like
page
views,
you
want
to
get
like
a
some
like
for
throughput.
You
want
to
get
some
some
measurement
latency,
whether
it's
like
load
event
or
lcp
or
something.
B
So
it
hasn't
been
clear
to
me
like
if
those
make
sense
as
as
metrics
or
if,
if
you
know
should
they
should
just
be
a
log
and
then
metrics
generated
on
the
back
end.
A
Because
if
I
want
something
to
be
a
metric-
and
I
want
it
to
be
on
a
timeline,
then
I'm
gonna
just
send
one
one
data
type,
not
two
and
the
back
end
has
to
like
whatever
generated
I'm,
not
gonna,
send
like
a
metric
for
largest
content
for
paint
and
then
a
something
else
for
it
to
show
in.
Maybe
we
can
just
well,
I
guess
the
background
can
just
put
the
metrics
on
a
trace
also
or
the
system
view.
I
mean.
B
F
B
Martin,
do
you
just
wanna?
Do
you
wanna
show
what
you
have
or
do
you
want
me
to
just
go
from
the
latest?
That's
one.
A
A
A
And
what
data
do
we
have.
F
Okay,
okay,
can
you
go
back
to
that
browser
like
how
come
we
can
see
logs
in
a
trace
like?
Is
this
one
place?
Oh,
it's
a
special
list.
This
is
sessions.
B
A
F
So
for
these
fans,
the
corresponding.
F
A
F
A
F
F
Which
one
which
event
you
want
to
see?
Let's
say:
let's
look
at
the
click,
only
click
event.
F
Basically,
the
attributes
of
the
click
itself
should
be
not
a
span
attribute,
but
should
be
either
should
be
an
event
attribute.
C
G
B
Yeah,
so
we
have
just
worked
on
like
this
is
kind
of
where
we
are
right.
Now
we
were
able
to
generate
events.
We
have
some
basic
api,
but
we
haven't
yet
like
looked
at
the
attributes,
the
semantic
conventions
or
attributes
so.
F
B
It's
it's
it's
kind
of
yeah
so
like
for
for
user
interactions,
we
talked
about
generating
always
always
logs.
All
those
events.
F
So,
every
time
there
is
a
spam
there
will
also
be
a
log,
but
every
time
there
is
a
log
there
need
not
be
a
spam
right
like
for
okay,.
F
Okay,
yeah
semantic
conventions-
you
know,
need
not
be
part
of
the
poc,
because
that
you
know
can
happen
later
too,
because
here
the
purpose
was
to
demonstrate
the
appropriate
data
model
to
use
for
different
situations.
F
Once
we
agree
on
that
proposal,
then
I
think
the
semantic
conventions,
maybe
on
a
few
important
ones
like
the
the
service,
dot
name
or
any
other
top
level,
lane
spaces
that
we
agree
on
that
might
be
sufficient.
Just
one
or
two.
B
That
sounds
fine,
no
ted
mentioned
that,
like
maybe
in
that
old
proposal,
we
should
have
some
something
about
like
what
something
about
semantic
conventions,
but
maybe
that's
not
that
secondary.
F
Yeah
sure
I
mean
we
can
add
a
few,
I
mean,
let's
add
a
few,
but
it
doesn't
have
to
be
complete
completely.
A
A
A
B
Okay,
yeah,
so
john
and
santosh
have
put
links
to
to
the
repo
and
the
demo
page,
so
you
can
actually
go
to
it
and
test
it
yourself.
If
you
want.
B
And
yeah.
C
A
So,
like
start
writing
it
there
and
then
yeah
use
the
poc
to
like
give
examples
or
go
case
by
case.
So
I
did
start
the
old
tab.
Yeah
I
mean
the
initial
cooldown.
Yes,
okay,.
B
B
Okay,
so
I
think
we
have
the
next
steps,
then
our,
like
finish
finish
the
prototype
just
with
some
additional
attributes
finish,
the
visualization.
B
D
D
B
And
I
would
I
would
basically
the
recommendation
would
be
to
to
generate
the
events
and,
and
if,
if
you
know,
and
if
the
sdk
is
capable
of
of
you,
know
generating
causality
or
durations,
it
would
also
generate
spans
and
link
those
logs
to
the
spans.
Those
events,
events
to
the
expanse.
A
F
Yeah,
maybe
we
can
add
an
explicit
note
saying
that
the
the
events
you
know,
although
there
is
an
option
to
use
inside
the
span,
we
are
choosing
to
keep
them
separate
and
link
them
to
the
span
as
we
are
going
as
there
is
some
talk
about
deprecating
the
spam
events
anyway,
in
the
long
term.
B
There
is
yeah
there
is
that
one
use
case
that
I
think
I'm
kind
of
trying
to
figure
out.
If
that
makes
sense,
which
is,
if
you
look
at
hold
on,
I
should
share
my
screen
real,
quick.
B
These
these
are
the
spans.
B
For
individual
resources-
and
they
have-
they
all-
have
this
list
of
these
events.
These
span
events
that
have
to
do
with
like
the
stages
of
the
network
requests.
B
So
I
do
my
look
up
and
you
know
secure,
like
ssl,
shake
shake
handshake
and
request,
request
response
like
that,
so
these
are
basically
duplicated.
Every
every
network
club
will
have
these
events,
and
so
that's
the
only
exception
where
I
kind
of
feel
like
they
only
make
sense
like
within
the
context
of
the
span.
D
B
F
So
I
was
going
to
say
that
these
events
are
today
inside
the
span,
because
there
was
no
option
right
when
this
was
implemented.
But
if
this
were
to
be
implemented
or
with
the
new
model
that
we're
talking
about,
how
would
you
do
it?
You
wouldn't
put
them
inside
the
span
right.
There
will
still
be
separate
events,
but
with
a
with
the
pointer
to
the
corresponding
span.
B
A
F
F
Have
agreed
on
a
suitable
data
model,
and,
and
eventually
you
know
when
these
things
get
stable,
we
want
to
move
to
a
uniform
representation
like
everywhere.
Things
are
present
in
the
same
way,
so
we
could
like
document
and
identify
a
plan
to
even
move
these
things
out.
Initially,
they
can
be.
You
know,
inside
the
span,
the
back
ends
will
have
to
be
modeled
to
you
know,
pull
these
events
from
inside
the
spans.
F
For
now,
until
you
know
they
are
moved
out,
yeah,
maybe
maybe
for
a
certain
time
now
we
could
send
them
twice.
I
don't
know
what,
if
it
has
an
implication.
B
Yeah,
I
think
also
like,
if,
if
you
had,
if
in
the
future
like
we
like
all
of
the
events,
were
separate
from
spans,
I
think
any
like
visualization
or
like
any
ui
that
that
displays
like
would
have
to
look
at
like
is
this.
Is
this
event
linked
to
a
span?
B
A
I
wonder
if
we
save
some
data
on
the
wire,
if
we
have
them
as
event
spans
spam
events
like
otherwise,
you
would
have
to
to
every
event
you
would
have
to
add
spam.
Id
may
possibly
trace
id.
Maybe
otlp
somehow
makes
this
more
concise
but
yeah,
because
that's
quite
a
lot
of
more
data.
A
F
So
that's
another
thing
that
should
come
out
of
this
poc.
Is
that
how
many
events
we're
talking
about
for
which
there
are
you
know
spans?
I
think
my
understanding
was
that
there
are
a
lot
more
events.
You
know
which
don't
have
causality.
You
know,
and
so
the
number
of
events
that
we're
talking
of,
for
which
there
is
the
spam
is
relatively
you
know,
maybe
like
five
or
ten
percent.
I
I
could
be
wrong,
but
some
smaller
percentage.
F
F
A
I
agree
also
the
bundle
size
well,
this
thing
doesn't
change
the
bundle
size,
but
the
bundle
size
is
also
very
people
start
complaining.
If
it's
big.
F
I
think
that
yeah,
maybe
we
should
take
it
up
with
the
respect
group
where
you
know,
are
there,
you
know
options
available
to
to
optimize
the
size
of
the
payload.
You
know
still
retaining
the
you
know
the
original
data
model.
F
B
G
Oh,
yes,
sorry,
I
missed
the
first
half
of
the
meeting,
but
but
yeah
I'd
love
to
have
a
look
at
the
prototype
and
I
I
think
I
just
need
to
catch
up
this
week.
I'm
sorry.
B
No
problem,
there's
a
there's,
a
link
to
the
prototype
in
in
the
meeting
notes.
A
D
G
I
think
where
we
found
it
to
be
very
useful,
I
mean
the
places
where
we
don't
have
zero
duration
spans,
of
course,
in
networking
like
so
anything
around
http
events
or
requests
and
responses.
Anything
around
app
starts
screen
render
times
is
another
area
that
we're
actually
investing
in
right
now.
So
those
are
things
that
make
a
lot
more
sense
as
a
span,
because
they
actually
have
a
time,
but
I
mean
yeah.
We
can
always
find
different
ways
of
visually
representing
it.
G
I
think
the
causality
piece
is
something
we're
still
trying
to
wrap
our
heads
around.
As
with
the
logs
data
model,.
F
G
B
B
There's
there's
actually
one
more
topic,
martin,
that
I
thought
of
in
the
poc.
Do
you
think
we
should?
We
should
add
events
for
page
loads
and
route
changes
hard
and
soft
navigations
yeah?
The
sba
support.
A
A
Yeah
you
can
yeah,
I
guess
you
can
lock
them
more,
I'm
not
sure
how
much
causality
you
can
have
there
exactly
yeah.
You
cannot
like.
I
don't
think
you
can
connect
route
change
event
to
spend
the
whatever
fetches
the
page
does.
B
So
the
browser
like
the
the
brow,
the
neural
browser
agent,
tries
to
do
that
like
where,
like
you.
B
A
A
B
Okay,
all
right:
well,
does
anyone
have
any
other
topic?
They
want
to
talk
about.
G
Okay,
I
had
one
quick
question:
have
we
discussed
locations
semantics
at
all?
I
may
have
missed
this
in
prior
conversations.
F
What
about
it?
I
I
I
always
used
to
think
that
it
is
something
that
the
back
end
could
derive,
because
the
ip
address
will
be
available
to
the
back
end.
G
Yeah,
I
think
it's
more
just
like
what
is
the
convention
that
we
all
want
to
agree
upon.
I
completely
agree
with
you
santosh,
like
a
lot
of
it,
has
to
be
derived
at
the
back
end.
Also,
the
thing
is
like
do
we
want
to
limit
it
to
ip
addresses,
or
do
we
want
to
also
want
to
encompass
things
like
latitude
and
longitude
which
can
be
picked
up?
You
know
from
your
device,
especially
when
it's
mobile.
F
There
was
another
you
know
discussion.
You
know
a
few
days
ago
that
there
might
be
some
customers
that
want
to
route
all
the
traffic
via
a
collector
going
through
their
organization,
but
I
guess
there
will
be
only
a
few
of
such
you
know
cases
overall,
but
in
that
case
the
ip
address
information
you
know
could
be
lost.
F
Remainder
right,
yeah
yeah,
unless
it
is
explicitly
you
know
handled
unless
it
is
explicitly
added
somewhere
in,
in
which
case
you
know,
it
would
be
good
to
define
some
70
conventions.
B
Or
would
it
be
possible
to
have
like
a
web
specific
receiver
processor.
F
Typically,
in
an
http
proxy
any
anytime,
there
is
an
intermediate
proxy
for
http.
They
are
supposed
to
add
a
header
called
x,
dash
forwarded,
dash
for
x,
forwarded
for
and
then
every
time
there
is
another
proxy.
You
know
they
are
just
they're
supposed
to
open
their
ip.
You
know
to
that
list,
comma
separated,
so
we
could,
you
know,
do
something
along
the
same
lines,
but
I
think
the
question
is:
do
you?
Do
you
want
to
support
this
use
case
or
defer
it?
You
know,
is
it
important?
F
G
To
be
honest,
however,
if
we
think
about
custom
and
like
putting
on
my
hotel
hat,
not
necessarily
my
splunk
hat,
I
as
a
customer
like
if,
if
I
want
to
maintain
my
own,
I
don't
know
in-house
stack
of
either
logs
or
spans
or
something-
and
I
don't
have
to
then
set
up
like
a
much
more
complex
ingest,
end
point:
I
can
kind
of
use
the
collector
to
forward
that
if
I
want
to,
because
the
collector
is
supposed
to
be
able
to
handle
a
lot
more
scale-
and
it's
actually
optimized
for
that,
so
I
won't
have
to
necessarily
rewrite
a
new
ingest
point
right.
F
Yeah
so
yeah,
I
think
we
need
to
understand.
You
know
what
what
are
the
capabilities?
You
know
they
want
in
that
architecture,
where
the
data
goes
through
their
systems.
F
If
it
is
about
masking
of
you,
know
personal
data
there,
then
maybe
you
know
we
could
enable
that
as
plugins
on
the
on
the
source
itself
like
when
the
agents
generate
the
data,
you
know
right
there,
you
know
you
enable
them
to
define
some
plugins
to
to
mask
the
data.
If
that's
acceptable,.
G
Actually,
we've
done
something
similar
kind
of
like
redaction
a
little
bit
in
our
libraries
as
well.
If
I'm
not
mistaken,
but
you
need
to
like
it's,
I
don't
think
it's
a
very
complex
thing
that
we've
enabled,
if
I'm
not
mistaken,
so
like,
I
think
we
look
for
specific
tags
or
we
look
for
specific
matches.
G
I
don't
know
how
how
configurable
someone
really
wants
to
make
it
right,
because
when
it
comes
to
pii
data
or
pci
data,
also
for
that
matter
or
phi
data,
there's
a
variety
of
like
different
kind
of
things
you
want
to
match
against,
and
every
country
is
different
too,
so
it
can
actually
lead
to
a
very
complex
rule
set,
which
is
actually
not
very
easy
to
build
like
out
of
the
box,
but
yeah,
I
think
redaction
rule
sets
are
very
interesting.
It
comes
up
in
conversations
beyond
client
instrumentation.
A
Ideally,
there
would
be
a
way
for
the
agent
to
get
the
configuration
from
the
back
from
the
server,
so
customers
could
easily
change
it.
Otherwise
it
would
take
a
long
time
for
some
companies
to
upgrade
their
agents,
and
this
is
for
backend
agents.
It's
okay,
but
for
client-side
agents
to
pull
the
configuration
from
somewhere
else
is
kinda
tricky,
especially
for
browser,
because
you
would
want
to
get
the
configuration
before
you
start.
Creating
expands.
F
Yeah,
I
think
the
remote
configuration
is
going
to
be
another
complex
topic.
I
think
we
should
talk
about
it
after
this
proposal
is
done.
F
G
I
mean
to
martin's
point:
should
we
at
least
agree
upon
a
like
semicon
on
around
this,
or
do
we
feel
like
that's
not
relevant
for
the
base
use
case.
G
I
mean
it's
an
incoming
ip
address,
it's
not
technically
being
sent
by
the
client
of
lemon
tree
itself
when
it
comes
to
the
ip
address,
but
I
don't
know
something
to
think
about.
I
think
lat
long
is
one
of
those
things
on
mobile
that
I
think
deserves
some
some
convention
because
it's
set
at
the
client
device.
You
can't
derive
that
from
an
incoming
connection.
F
You
know
it
has
to
come
from
the
source.
It
will
be
more
precise
than
you
know
what
can
be
derived
so
yeah.
Maybe
then
martin,
maybe
I
don't
know
how
you
want
to
track
this.
So
for
some
of
these
things,
maybe
we
should
start
creating
tickets
to
track
our
issues
right.
G
John,
you
keep
pointing
us
to
it.
Sorry
I
need
to
have
my
second
cup
of
coffee.
Is
it
called
resource
semantic
conventions?
I
think.
G
F
So
in
the
same
repo
you
could
yeah
to
submit
an
issue
and
then
assign
it
to
an
officer.
B
Sounds
good
anything
else.