►
From YouTube: 2022-02-02 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).
C
A
A
A
A
But
we've
got
a
couple
of
pr's
first
thing
I
just
quickly
raised.
I
said
I'd
do
this
and
I
finally
got
around
to
it.
Raising
this
pr
for
error
statuses
regarding
the.
A
This
error
status,
so
it
was,
it
was
mentioned
by
dennis
that
the
aerostat,
just
because
a
client
receives
a
4xx,
doesn't
necessarily
mean
that
it's
a
it's
an
error
on
the
client,
so
it
seems
reasonable
to
leave
it
on
set
there.
So
I'm
interested
to
get
more
people's
thoughts
on
it
there,
because
I've
actually
raised
the
pr
for
it
here
and
yeah.
It's
a
it's
a
short
one
in
terms
of
your
examples
dennis
regarding
that
kind
of
thing
like
what?
What
do
you
think?
The
top
examples
are
for
that.
C
Yeah,
so
we
have
some
some
cases
when,
for
example
like
something
like
404
or
409
or
4,
2
9
cannot
be
considered
errors
on
the
client
side,
just
because
it's
that's
the
way
how
communication
works
so
and
basically
it
might
be
really.
You
know
really
tough
to
say
that
all
the
4xx
represents
error
errors
on
the
client
side.
So
there
are.
There
are
some
examples
or
age
cases
so
actually
to
to
cover
this.
C
It
might
be
beneficial
to
have
like
to
keep
this
and
set
basically
mirroring
the
same
behavior
that
we
have
for
for
a
server
right
now.
A
Yeah,
I
think
it
makes
sense,
so
that's
just
their
early
stages
just
want
to
let
everyone
know
it's
there
dennis
you
had
a
pr
you've
put
in
the
agenda.
C
Yeah,
so
this
is
a
really
old
one.
Actually,
you
see
it's
from
october.
So
recently
I
got
a
lot
of
feedback
from
different
groups
of
people
and
thanks
thanks
all
for
providing
this
feedback,
so
the
most
one
of
them,
the
most
interesting
there
was
from
probably
python
instrumentation
group,
so
they
they
were
saying
that
it's
not
really
clear
for
them
how
to
actually
structure
all
this
stuff.
So
I
did
a
lot
of
rewardings
so
yeah.
C
If
you
can
open
the
file
itself
like
empty
file
itself,
maybe
we
can
just
briefly
see
what
we
have
there.
C
A
I'll,
just
let
you
I'll
just
let
you
drive
there,
you
go
yeah.
C
It
will
be
so
let
me
choose
the
desktop,
probably
this
one,
okay,
so
oops.
So
what
I
was
thinking
about,
so
what
it
is
a
lot
of
windows.
Maybe
I
can
just
open
this
one
here.
C
Just
what
it
is,
so
that
should
be
it
right
all
right,
so
we
do
have
this
section
for
retries
and
redirects,
and
I
tried
to
clarify
this
like
as
much
as
possible
and
just
wanted
to
to
have
your
feedback
here.
Basically,
we
are
saying
here
that
we
are
we're
having
more
than
one
physical
request
in
in
for
advice
in
the
redirects
cases,
and
here
we're
stating
that
a
client
spam
should
be
created
for
each
of
these
physical
spams.
C
So
this
is
the
first
thing
when
we
use
a
word
should
previously,
it
was
must,
but
it
was
confirmed
by
aws
sdk
folks
by
on
rack
and
by
nav
from
javascript's,
instrumentation
or
javascript
sdk
that
it's
not
really
possible
for
them
to
to
create
these
pens
for
each
physical
request.
So
that's
why
we
have
this
shoot
here
and
basically
we
are
also
saying
here
that
there
is
no
spam
created
for
logical
or
this
encompassing
request.
C
So
we
are
not
having
this
top
level
spam
for
for
each
logical
request
and
then,
when
it
comes
to
a
relationship
between
these
newly
created
spans,
I'm
saying
here
that
link
should
be
used
and
it's
basically
should
be
used
to
link
each
subsequent
spam
with
the
immediate
previous
physical
request.
So
we
have
just
this
linked
list,
representation
kind
of-
and
also
I
put
here
this
clarification-
that
these
stance
should
be
created
as
symbolics,
not
as
like
a
children
of
of
the
previous
previous
try
or
like
a
previous
redirect.
C
Also,
I
put
this
this
statement
here
that
in
case
there
is
no
top
there,
we
have
no
trace
started
up
front.
Basically,
we
have
no
top
level
span
created
in
this
case.
This
spans,
for
it
right
and
redirects
will
correspond
to
different
completely
different
traces,
and
that's
exactly
why
we
use
link
here.
Basically,
we
are
expecting
users
to
interpret
this
spans
with
links
to
actually
reconstruct
this
logical
request.
B
B
It's
it's
very
hard
for
me
to
parse
this
information
fyi.
So
I
do
have
examples
here,
so
I
can
show
it
later
yeah
so
that
may
be
yeah
go
ahead
james
and
then
I
will
go
to
the
examples,
maybe
that
I
think
that
explains
my
problem.
A
I
was
just
like
in
the
case
that
there
is
no
trace
started
before
retrieser
redirects.
Wouldn't
there
always
be
at
least
another
span
started
if
there
was,
if,
if
you
were
retrying,
a
request
like
I'd,
expect
there
already
to
be
like
a
span
open
for
that
initial
request.
B
That
should
be
the
case
correct.
I
I
expect
that
to
be.
C
The
case
well,
actually,
if
I
just
correctly
it
will
not
be
this
top
level
span.
So,
okay,
let
me
put
it
this
way
once
again,
so
there
is
no
kind
of
encompassing
spam
created
for
each
request
like
a
retry
or
redirect
chain.
Just
because
we
don't
know
up
front
will
be
any
retrace
or
redirects
right
when
we
are
starting
communicating
with
the
server.
C
So
like
we
have
several
options
here
we
can,
we
can,
you
know
artificially,
create
this
in
comparison,
encompassing
requests
or
encompassing
spam,
but
this
basically
means
in
case
there
are
no
no
retries
on
redirects.
We
will
be
just
duplicating
data.
Can
can
you
show
the
example?
Maybe
you
have
there
some
of
these
things?
C
Definitely
so
basically
examples
are
the
following.
So
we
have.
I
have
these
two
examples
like
in
in
in
case
we
do
have
span,
trace,
started
and
no
space
no
trace
started
so
for
trace
started
like
we
just
assumed.
We
have
this
inbound
request
and
then
we
have
some
outbound
requests
and
basically
this
one
should
be
read
right,
so
we
have,
but.
C
Well
because,
basically,
that's
the
only
kind
of
generic
possibility
for
us
to
reconstruct
this.
The
overall
kind
of
retry
and
redirects
span
or
like
a
structure
of
these
pants
so
like
an
alternative
here,
is
to
have
this
spans
created
as
as,
like
a
part
having
parallel
par
and
child
or
relationship
right.
But
in
this
case
it
will
be
this
gets
server,
then
it
will
be
again
get
server
and
the
it
will
be
really
highly
nested
kind
of
spans
created
here.
So
I
think
I
think
we
haven't
defined
yet.
B
To
to
to
use
the
links
here,
I
think
you
need
another
enum
that
says
a
bit
of
relationship,
some
some
kind
of
relationship
there
that
you
want
to
put
between
them
because
links
usually
at
least
so
far.
We
were
seeing
them
as
parent
child,
mostly
in
case
of
batching
and
stuff.
Like
that,
so
probably
we
may,
we
may
have
to
extend
that.
B
C
C
Well,
you
know
there
is
no
explicit
attributes
that
it
indicates
this,
but
you
know
there
are
some
some
another
like
a
pieces
of
information.
We
can
easily
get
to
understand
this.
The
first
thing
is
this:
you
know
host
and
the
path
and
like
a
basically
spam
name
sure,
but
it's
time.
Okay,.
B
But
same
thing
you
can
get
without
having
that
link
same
resolutions
you
can
use
without
that
link.
If
you
just
remove
that,
you
can
use
all
the
things
like
the
path,
the
300,
the
fact
that
there
is
an
http
request,
retry
count
to
infer
that
this
is
actually
following
by
the
other
one.
So
I
I
think
I
think,
unless
maybe
I'm
wrong,
but
I
don't
think
it
gives
you
any
information
that
lee.
Unless
we
make
it
we,
we
do
add
some
extra
information
there
to
help
you
identify
that
this
is.
C
What
you
need
correct,
so
I
see
a
point
and
like
up
for
for
this
case.
Links
doesn't
like
they
don't
don't
bring
much
information
here
in
addition
to
what
we
have
just
because
we
have.
We
already
have
this
three
right.
Yeah.
C
Yeah
in
case
there
is
no
trace
started.
We
don't
have
these
three
anymore
right.
We
have
totally
separated
isolated
traces,
and
here
the
only
like
a
possibility
for
us
to
kind
of
reconstruct
the
overall
behavior
that
happens
on
the
on
the
client
side
is
to
have
these
links,
and
this
is.
This
is
really
like
a
you
know,
really
possible
case.
For
example,
if
you
have
a
client
application
like
a
mobile
application,
there
is
no
top
levels
panel.
C
There
is
nothing
that
we
can
think
about
as
like
a
trace
has
been
started
right,
so
it
will
be
just
like
an
initial
point
in
this
case
it-
and
it
can
be-
it-
can
like
bring
this
additional
retries
and
redirects
yeah.
B
B
Now
now,
for
this
case,
we
may
consider
this,
and
I
I
I
can
see
where
this
will
help.
I
was
mentioning
first.
In
the
first
case,
I
don't
think
we
should
add
the
links
in
the
second
case,
that's
a
very
interesting
point,
and
there
is
also
a
notion
of
session
which
we
should
consider
to
to
map,
even
even
though
the
request
may
not
be
necessary,
retries
within
a
session.
For
example,
there
is
this
notion
of
session,
especially
in
web
browsing
stuff
like
how
do
we?
C
I
don't
have
clear
answer
to
this,
but
yeah
definitely
there's
something:
that's
we
can
use
to
represent
this
session
so
basically
yeah
from
from
this
perspective,
it's
like
a
logical
kind
of
transaction
or
session,
and
it
makes
sense
you
know
to
to
have
some
information
to
to
be
able
to
reconstruct
it.
I
believe
the
same
can
be
done
for
like
http
2
scenario.
When
you
have
just
channel
open-
and
you
know,
requests
risk
like
a
messages
basically
flowing
over
this
open
channel.
It
can
be,
it
can
be
done
the
same
way
as
well.
C
Concept
of
session,
as
of
now
in
open,
telemetry
right,
so
there
is
no
kind
of
way
how
we
can
represent
it.
I
believe-
and
actually
that's
this
came
out
of
similar
discussion
several
months
ago
in
the
same
group.
Basically
we're
just
trying
to
you
know
model
it
differently
from
different
perspectives,
and
we
clearly
saw
that
it
was
possible.
You
know
from
all
the
instruments
that
are
already
available
links
give
us
a
lot
of
flexibility,
and
it
allows
us
to
you
know
to
to
represent
overall
overall
behavior
yeah,
but
but.
B
You
may
not
derive
the
right
thing
on
the
backhand
side.
So
that's
what
I'm
trying
to
say
it's
like
it
may.
We
may
require
to
have
some
kind
of
relationship
or
something
there
to
describe
this
kind
of
relationships
because,
as
you
pointed
there
are
way
too
many
relationships
that
we
can
model
here
and
I
don't
think
we
can
derive
all
of
them
from
arguments
from
from
things.
That's
that's
my
worry
that.
C
Yeah,
that's
a
really
good
point,
and
but
it's
really
valid
points.
The
thing
is
that
we
don't
have
any
at
the
moment
right.
We
don't
have
any
kind
of
indication
that
we
can
put
next
to
link
saying
that
this
is
like
a
you
know,
session,
or
this
is
rich
ryan
or
this
is
redirect
or
something
else
right,
so
definitely
something
that
we
probably
need
to
continue
to
work
on
just
to
clarify
yeah.
C
That's
correct,
yeah,
I
don't
really,
you
know,
have
an
opinion.
What
those
attributes
can
be
represented
right,
how
how
they
can
be
represented
and
which
exact
attributes
we
can
put
there.
C
Maybe
it's
something
we
can
also
think
about
going
forward,
so
just
just
to
clarify
type
of
relationship
right,
but
you
know
we
can
definitely
add
it
here,
but
it
should
be
some
high
level
concept,
as
you
said,
some
kind
of
ename
that
we
can
rely
on
putting
all
these
values
for
these
attributes.
Right
so
looks
like
it
should
be
some
kind
of
super
set
of
different
types
of
relationship
which
we
can
use
here.
C
B
C
Well,
we
don't
we
don't
in
this
case,
we
don't
have
like,
or
it's
not
really
possible.
You
know
to
say
for
sure,
do
we
have
like
at
this
top
level
span
or
not,
or
do
we
have
this
traces
started
or
not
in
case
we
do
reit
rights
or
redirects
from
the
very
beginning
right
so
having
these,
you
know,
conditions
stating
that
okay,
if
we
have,
if
we
have
this
top
level
span,
we
need
to
do
this
and
if
we
don't
have
any,
it
should
be
different.
C
Just
brings
more
kind
of
you
know,
complexity
to
overall
to
overall
representation.
So
from
my
perspective,
it
will
be
good
to
have
it
consistent
right,
but
at
the
same
time
it
allows
us
to
cover
all
the
cases.
B
B
I'm
not
I'm
not
writing
much
instrumentation
these
days,
but
you
may
also
have
troubles
passing
around
the
previous
span,
because
the
previous
span
is
done
at
the
moment
when
you
need
to
do
it.
The
retry
correct,
like
you,
you
have
a
span
which
was
done
because
then,
probably
you
have
some
back
off
there
like
retry
with
the
back
off,
so
you
may
retry
in
20
seconds.
B
That
span
is
already
done.
I
don't
know
to
hold
that
memory
around
for
the
span
context
to
to
record
the
link.
That
may
also
be
a
problem
like
in
the
same
way
as
it's
a
problem
to
have
the
logic
of
of
checking.
If
you
have
a
parent
or
not,
it
may
be
a
problem
to
to
to
pass
around
the
like
memory
between
between
the
two
requests.
C
Yeah,
that's
all
also
really
valid
points,
and
you
know
I.
It
can
be
also
even
generalized
more
when
we're
talking
about
this
contract
propagation
right.
So
let's
put
it
from
some
slightly
different
angle.
So,
for
example,
we
have
this
notion
of
session
or
transaction,
and
we
want
to
you
know
propagate
some
kind
of
context
from
one
spawn
to
another
to
represent
that
all
these
you
know
somehow
currently
correspond
to
the
same
session.
C
So
this
kind
of
you
know
problem
of
context
propagation.
We
also
have
this
in
scope
of
figuring
out
for
v1,
for
these
mining
conventions
specification,
stable
version,
first
table
version,
but
again
this
is
kind
of
really
somehow
slightly
different
problem,
which
also
applicable
here
for
sure.
C
I
fully
agree
with
that,
but
it's
it's
like
a
you
know,
it's
probably
another
topic:
how
how
we
want
to
propagate
this
stuff,
and
definitely
we
also
have
this
several
options,
so
we
can
have
some
context
to
be
flown
within
the
requests
within
this
session,
so
it
should
be
kind
of
like
some
ambient
context
and
whatnot
yeah.
We
can
think
about
using
baggage
here
or
something
like
this
yeah.
So
definitely
I
I
fully
agree
with
you
that
this
context
should
be
saved
or
stored
somewhere
to
be
able
to
so
so.
B
C
The
entire
transaction-
just
because
I
don't
know
it
will
be
transaction
or
not
right
in
case
of
retries,
I
cannot
say
for
sure,
will
it
be
any
kind
of
retry
and
to
be
honest,
for
majority
of
cases
will
be
no
retries
right,
maybe
for
like
80
or
even
even
90
percent
of
cases
will
be
adjusted.
C
It
will
be,
it
will
be
basically
this
one,
so
in
case
it
will
be
trace
started.
It
will
be
this
case
if
it
will
be
no
case
started.
It
will
be
kind
of
this
case
so
yeah.
If
you're
talking
about
just
like
a
200.
It
will
be
just
this
case,
but
what.
B
I'm
trying
to
say
is
the
majority
of
the
cases.
You'll
have
a
trade
started
and
we
already
agree
that
links
there
may
not
necessarily
help,
so
we
don't
need
links
then,
on
the
second
one.
On
the
second
example,
I
would
say
that
we
may
take
the
risk
and
build
that
span.
The
transaction
span
all
the
time.
C
Well,
yeah,
I
actually
not
really
follow
why
we
should,
where
we're
thinking
about
this.
This,
like
this
trace,
is
already
started,
so
it
should
be
a
lot
of
cases
when
it's
not.
I
I.
B
C
Right,
but
it's
not
like
a
100
percent
cases
right
so
definitely
the
for
the
majority.
Probably
that's
the
case
like
when
we're
talking
about
back-end
apps,
and
we
do
have
this
instrumentation
enabled
for
inbound
requests.
It
definitely
will
be
there
but
in,
but
if
not,
it
will
not
be
there.
So
I
don't
think
that
we
can
rely
on
this.
C
Stating
that
spam
will
be
or
like
a
trace
will
be
always
started
so
yeah.
That's
why
you
know
it
make
it
harder,
because
we
cannot
cover
all
the
all
the
cases,
but
this
approach
looks
like
we
just
just
you
know
re
using
links
as
instruments
to
represent
different
types
of
relationship
not
only
par
and
spam,
but
any
kind
of
relationships.
C
This
instrument
allows
us
to
model
to
model
all
this
stuff
yeah.
I
agree
that
going
forward.
It
might
be
also
beneficial
to
clarify
which
type
of
relationship
we
are
representing,
with
links,
for
example,
with
attributes,
but
yeah,
that's
probably
can
be
done
as
a
second
step
here.
B
Okay,
I'm
I'm
very
worried
about
this
proposal.
If
I,
as
I
said
it's,
it
adds.
B
So
especially
the
first
the
first
example,
it
adds
kind
of
double
propagation,
so
you
are
kind
of
doing
double
propagation.
You
do
a
parent
child
propagation
from
the
request,
server
context
and
you
do
from
siblings
as
well,
so
so
now
that
we
only
we
have
to
do
one
context,
propagation
kind
of
now.
We
have
to
do
double
propagation
here,
one
that
goes
from
from
from
the
the
server
span
and
the
other
one
that
goes
from
siblings.
B
C
Well,
yeah
again
like
this:
is
this
valid
point
but
as
as
you
said
this,
this
propagation,
when
we
have
trace
started,
is
definitely
something
that
we
already
have.
So
we
already
have
this
propagation,
so
we
we
know.
If
so
we
are.
C
One
more,
I
think,
one
more
on
top
of
the
one
that
we
already
have
to
do
like
correct
correct,
but
we
cannot
say,
like
I
mean
for
some
cases
we
don't
have
any
trace
started
and
that's
the
problem
right.
So
this
this
this
like,
if
we
don't
have,
maybe
maybe.
B
B
I
have
to
carry
everything
that
I
inherit
plus
extra
so
like,
and
we
don't
have
a
space
where
to
carry
that
extra
information,
not
even
in
the
context.
We
don't
have
that
space.
We
we
need
to
create
the
second
entry
into
the
context,
dial,
that
is,
the
parent
and
the
sibling,
which
is
absurd.
In
my
opinion-
and
I
think
it's
it's
a
huge
change
that
you
are
going
to
ask
us
to
do
well
more
than
you
can
imagine
versus
trying
better
to
identify
the
cases
and
and
forced
to
have
a
transaction
span.
C
For
sure
for
sure,
so
we
can
we
can
do
this
like
we
can
start
a
discussion
again.
So
we
can,
we
can,
you
know,
engage
more
folks
just
to
see
what
they
were
their
points.
So
basically
what
you
are
saying
just
to
summarize
we
can
in
this
case
we
can
reuse
the
fact
that
trace
is
started,
so
we
can
avoid
using
creating
links
right,
but
in
case
there
is
no
trace
started.
Links
are
still
needed
or
or.
C
Yeah,
so
if
we
like,
as
I
said
like
it,
looks
like
there
is
no
way
for
us
to
indicate
a
front,
will
it
be
any
kind
of
rate
rise
and
redirects,
because
it's
not
like
a
user
transaction
right
and
what
what
would
be
the
downside?
We
will
create
a
fake,
not.
C
C
And
that's,
it
is
correct
so,
and
the
problem
here
is
that
for
majority
of
cases
it
will
be
just
no
retries
or
redirects,
and
instead
of
one
we
will
just
say
we
will
be
just
sending
two
spams
every
time
right.
So
basically,
that's
that's
why
we
just
double
the
the
traffic
from
from
any
kind
of
instrumentation,
and
that's
that's
the
also
the
problem.
C
So
the
thing
here
is
like
you
know,
it's
always
a
trade-off,
and
probably
the
question
is
how
how
we
want
to
make
it
more
efficient
from
from
from
from
both
sides
right,
so
how
we
want
to
avoid
duplicating
data
and
increasing
the
drive,
the
amount
of
traffic,
but
in
the
same
time,
how
we
can,
which,
which
technique
we
can
use
to
to
reconstruct
this
case,
for
example,
on
the
on
the
platform
side,
when
we.
B
C
So
yeah
in
this
case,
basically
it
will
go
against
probably
current
practices
right,
because
we
still
like
from
this
case.
We
already
have
this
kind
of
request.
We
do
have
a
response,
we
do
have
our
results,
but
we
cannot
use
this
result
and
it
will
be
only
identified
like
a
going
forward
right,
so
it
will
be
some.
It
should
be
all
like
this.
C
B
C
Yeah,
so
that's
that's,
basically
what
I
was
saying
about
this
context,
propagation
thing,
so
it
probably
should
be
better
clarified
in
the
specification
itself
right
how
how
to
how
to
behave
in
these
cases.
So
it's
just
like.
I
know
you
know
some
some
just
representation
of
this
case,
but
the
the
overall
question
is
wider
and
what
like
the
the
technique
that
I
was
using?
Actually
I
do
have
some
some
prototype
or
not
even
prototype,
but
the
implementation
of
this
4.net
and
we
do
have
it
for
python
as
well.
C
So
I'm
working
with
folks
who
are
helping
to
do
this
with
python
and
basically
for
nets.
The
you
know
the
overall
trick
was
to
use
the
same
exact
instance
of
request
object.
So
basically
we
can
say
that
okay,
we
can
reuse.
The
same
request
object
to
do
retries
or
redirects,
and
within
this,
within
this
instance
of
this
object,
we
can
store
some
some
additional
information,
for
example,
this
link
or
can
expand
context
of
the
previous
spam.
C
Yeah,
so
just
just
a
couple
of
words
about
the
redirect
case,
it's
really
similar.
The
only
difference
here
is
that
we
have
this.
Like
a
you
know,
302
code,
and
we
do
have
another
span
like
this,
and
there
is
no
additional
attributes,
so
there
is
no
retry
or
redirect
or
sorry
there
is
no
retry
account
but
yeah.
Essentially
they
are
pretty
much
the
same
from
the
structure
perspective.
They
are
just
exactly
the
same
yeah.
Basically,
that
that's
was.
A
A
Like
you
don't
to
me,
it
seems
like
bogdan
met.
You
are
suggesting
that
it
should
always
well.
This
should
be
basically
a
parent
child
spans
of
each
other,
if
I'm
engaging
that
correctly,
but
it
can't
be
that
case.
You
can't
like
do
that
in
the
second
case,
but
I'm
not
sure
why.
Why
can't
you
just
there's
like
no
parent
spam
available,
you
can't
see
it
or
you
mentioned
mobile
applications
or
something.
B
In
mobile
application
there
is
a
case
or
or
the
initial
page
load,
for
example,
for
a
web.
I
heard
that
multiple
times
so
essentially,
when
you,
when
you
go
in
chrome
and
the
type
the
url,
the
initial
page
load,
you
you
don't
have
a
chance
to
start
anything
because
the
chrome
does
not
chrome
does
not
allow
you
to
execute
any
code
until
you
have
to
load
the
library
and
so
on
and
so
forth.
So
there
are
a
bunch
of
things
that
can
happen
there.
B
B
So
if,
if
we
do
have
a
way
to
to
to
model
the
sessions
which
will
be
between
multiple
requests
between
retries,
whatever
all
the
things
notion
of
a
session,
then
I
don't
think
you
are
left
with
any
cases
in
this
second
thing
or
very,
very
very
rarely
that
I
may
not
care
that
you
double
the
span
problem
that
you
mentioned
so
so
I
think
I
think
I
think
if,
if
you
are,
you
know
any
distributed
system
like
if
you
are
not
an
edge
node,
you
most
likely
receive
something
if
you
did
not
receive,
we
created
something
when
the
request
came
so,
first
of
all,
if
you
have
everything
instrumented
correctly,
okay,
so
every
time
when
you
receive
a
request,
an
incoming
request
from
somebody
you
are
in
a
middle
note.
B
So
this
request
has
to
come
from
nowhere,
and
this
comes
from
nowhere
or
doesn't
have
a
parent
logical
is
where,
where
in
on
mobile,
apps
and
web
apps,
where
we
said:
okay,
that's
where
we
have
to
model
sessions
and
then
all
of
these
things
will
become
child
children
of
of
a
session
span,
and
the
second
case
would
be
we.
If
you
have
background
operations,
background
operations
here,
meaning
I
know
you
have
some
some
triggers.
For
example,
you
you
have
a
crown
job.
Okay,
you
have
a
chrome
job.
That
does
something.
That's
that's
another
example.
B
Okay,
now,
okay,
sure
you
don't
have
you,
you
are
not
you
are
not.
Maybe
you
don't
model
it
correctly.
If
you
don't
have
a
a
span
when,
when
you
from
a
chrome
job
operation,
you
start
doing
an
http
request,
you
may
not
do
it
correctly,
but
I
think
the
solution
should
not.
I
don't
think
the
solution
is
asked
to
start
relying
on
building
these
links
and
helping
the
user.
Maybe
the
answer
is:
hey
just
create
a
span
when,
when
you
start
your
chrome
app
like
you
have
that
option
to
to
to
do
it
correctly.
C
Well,
yeah,
I
I
I
fully
agree
with
some
some
points
that
you
brought
and
definitely
can
simplify
the
overall
landscape
right,
but
the
thing
is
that,
for
example,
for
cron
jobs,
which
is
really
well
at
example.
You
basically
need
to
have,
for
example,
you
have
a
cron
job
and
you
have
a
script
in
python
and
in
this
case
you
also
need
to
bring
a
lot
of
different
dependencies
just
to
instrument
this
chrome
job.
C
I
assume
many
different
for
many
different
cases
now,
we'll
be
no
one,
be
just
doing
this
like
on
purpose
right,
so
yeah,
sorry,
muted.
But
the
thing
is
that's
like
now:
we
trying
to
do
our
best
effort
and
like
if
you
look
at
this
from
the
different
angle.
I
fully
agree
that
if
you
have
some
representation
of
session,
it
shouldn't
be
kind
of
just
top
level
spam
representing
this
passion
session.
It
can
be
something
some
another
way
of
doing
the
same
stuff
right,
but
if
it
habits
this
basically
falls
in.
C
B
Be
it
could
be
the
starting
point
of
the
discussion,
but
on
the
other
side,
I
don't
know
if
you
saw
the
recording
from
this
morning.
We
made
a
mistake,
and
now
I'm
fighting
with
everyone
about
that
mistake,
so
because,
because
this
project
is
on
the
stability
road
doing,
some
experiments
like
this
may
lead
to
to
not
necessarily
the
best
design
and
it's
almost
impossible
to
change
after
so
that's.
Why
that's?
C
After
for
sure,
for
sure
and
the
goal
exactly
the
same
to
to
bring
some
mining
conventions
hp,
so
many
conventions
to
this
to
the
stable
state,
that's
exactly
the
goal,
and
this
is
the
one
of
the
item
that
we
have
in
scope.
One
of
the
four
items
that
we
have
in
scope
for
this.
So
yeah
I'm
open
to
two
suggestions,
and
we
can.
We
can
decide
like
how
we
can
improve
this
okay,
so
yeah.
My
personal
suggestion.
B
Personally,
ignore
the
case
where
you
don't
have
a
parent
and
treat
that
you
have
a
parent
and
leave
that
broken
for
for
the
first
table.
It's
not
gonna
work,
it's
broken,
we
know
it's
broken
and
we
can
fix
it,
but
I
think
you
will
cover
80
90
percent
of
the
cases
with
with
that
and
then
and
then,
once
you
have
that
covered
everywhere,
you
can
start
discussing
about
these
edge
cases,
and-
and
so
you
can
get
to
stability
by
just
having
that
without
over
engineering.
B
C
Right
thing
for
sure,
for
sure:
that's
exactly
why
we
have
this
shoot
instead
of
any
kind
of
must
so,
basically
now
we
are
saying
that
we
we
can.
You
know
we
are
suggestion,
suggesting
you
to
model
this
with
links,
but
if
you
don't
want,
you
don't
have
to
right
or
you
have
some
really
good
reason
not
to
do
that.
So
that's
the
thing
so
should,
if
I
should
is.
B
You
sh,
you
should
do
it
unless
you
have
a
good
reason
to
not
do
it
right
if
you
so
so
you
are
not
suggesting
you
are
strongly
encouraging.
So
should
is
more
or
less
strongly
encouraging.
Make
is
a
situation
and
must
is
a
requirement.
So
so
you
are
strongly
encouraging
people
to
do
this,
which
means
I
have
to
have
an
exception
to
not
do
it.
It's
not
it's
not
that
I'm
lazy
or
I
don't
want
to
do.
It
is
because
I
have
a
clear
exception
that
I
don't
want.
I
cannot
do
it.
B
C
Okay,
okay,
got
it
yeah.
I
see
a
point
like.
Can
you
also
advise
like
what
exactly
it
can
be
the
past
going
forward?
The
thing
is
that
we
already
have
this
request
approved.
C
We
try
a
couple
of
folks,
so
I
just
just
wanted
to
figure
out
what
can
be
the.
I
don't
personally,
don't
know
how
it
works
in
open,
telemetry
as
a
whole
like
what
will
be
the
yeah.
C
We
do
have
this
from
from
joshua
and
from
ted
yeah
yeah
yep.
B
So,
first
of
all,
did
you
see
the
other
pr
you're
gonna
be
added,
as
the
owner
thinks
he
here
so
two
zero
two.
So
you
saw
this
one.
B
B
C
C
You
more
work
yeah,
so
just
just
trying
to
do
my
best.
You
know
to
make
it
to
make
some
progress
here,
but
in
the
same
time
to
you
know,
make
a
best
effort.
Definitely
I
I
see.
I
see
all
these
points
that
you
brought,
and
you
know
we've
discussed
this
several
times
looks
like
we've
like
previously.
It
was
consensus
that
we
found.
Is
this
like
a
solution
with
links,
but
now
we
definitely
can
reconsider
it
and
see
like
I
look
for
some
better
solution.
B
You
see,
the
tigran
is
assignee
on
this.
Okay,
you
should,
if,
if
you
are
interested
just
ping
him
on
slack
or
on
somewhere,
like
just
being
him
actively
tigran,
come
and
resolve
this,
you
are
an
assigned
to
this.
C
Okay,
all
right
so
okay,
so
I
will
discuss
it
with
you
guys.
Yeah,
I
mean.
C
Well,
yeah,
the
the
the
way
I
saw
it
before
this
meeting.
Actually
it
looks
like
that
we
have
this.
You
know
consensus
and
I
was
just
doing
some
improvements
to
resolve
some.
You
know
unclarity,
so
the
same
exact
proposal
is
here
starting
from
october,
so
the
last
change
is
mostly
about
rewarding
all
this
stuff
and
make
it
more
clear
in
adding
more
examples.
So
there
is
no
change
like
in
essence.
C
So
I
was
thinking
about
you
know,
making
some
steps
completing
it
this
week,
but
looks
like
we
need
to
now.
We
need
to
do
more
efforts
here,
so
just
trying
to
understand
what
will
be
the
next
step
in
in
terms
of
finding
new
solution
or
like
I
just
you
know,
removing
something
or
make
it
broken
from
the
very
beginning,
as
you
suggested,
and
to
cover
just
some
subset
of
cases
so
yeah
all
these
topics
are
here.
A
C
I
believe,
like
it's
not
not
really
bring
more.
You
know,
clarity
when
it
comes
to
parent
child
or
links,
because
basically,
links
are
just
extension
of
par
and
child
right
from
this
perspective,
so
you
can
model
to
parent
child
with
links,
but
not
what
was
not
vice
versa,
and
if
you
decide
to
go
with
parent
child,
you
need
to.
You
still
need
to
store
this
context
between,
tries
to
be
able
to
represent
this
pile
trying
to
basically
indicate
this
was
my
parent
right.
C
So,
from
this
perspective,
it's
pretty
much
the
same
links
just
give
you
more
flexibility
in
terms
of
the
overall
tree
like
when
we
have
this
client
server.
Client
server
will
not
will
not
be
really
like
a
kind
of
nested
right,
so
the
structure
of
this
three
will
be
basically
the
same
having
them
siblings
rather
than
children.
A
B
I
I
commented
my
thoughts
there
feel
free
to
ping
me
any
time
you
want
on
slack,
I'm
very
available
great
great.
Thank
you.
A
Oh
we've
got
anything
else,
we'd
like
to
discuss
or
five
minutes
to
the
hour,
so.
A
Okay,
I
look
forward
to
it
all
right.
It
seems
like
we
got
nothing
else
to
discuss,
but
I
think
we
had
a
productive
discussion,
so
that
was
good
thanks,
guys
probably
ended
it.
Okay.