►
From YouTube: 2023-02-02 meeting
Description
Instrumentation: Messaging
A
B
A
D
F
C
E
Please
add
your
name
to
the
list
of
attendees
in
the.
E
E
Went
over
that
I
think
three
weeks
ago
and
sorted
out
some
issues,
some
open
issues
here
to
look
into
and
I
remember,
I
took
overlooking
into
those
first
two
here
and
I.
Didn't
get
get
enough
time
to
look
into
this
one.
But
I
looked
into
this
capture,
Kafka
Caster,
ID
and
added
some
comments
there
and
I
actually
saw
that
it's
the
purpose
or
the
main
purpose
of
this
Caster
again
cuff
case
that
people
say
yeah,
often
hostname,
IP
and
port,
or
any
combination
of
those.
E
E
A
document
here
that
describes
this
and
I
think
with
this
understanding.
In
my
opinion,
it's
definitely
worth
including
this
into
the
Kafka
specific
semantic
conventions,
because
I
also
think
Kafka
people
put
it
as
dimensional
metrics,
that's
mentioned
somewhere
in
this
document
and
I
think
the
use
case
is
a
valid
one.
So
if
nobody
has
any
different
opinion,
I
will
trust.
Yeah,
probably
ask
Jack
to
open
the
PRN
at
the
Kafka
cluster
ID
to
the
Kafka
specific
conventions.
E
E
F
Yeah,
so
I
can
quickly
update
you
on
this
or,
if
you
want
to,
we
can
switch
to
this
sorry,
adapt.
C
F
So
for
the
first
one:
where
can
you
open
this
year.
F
I,
oh
sorry,
I
didn't
realize.
I
need
to
leave
a
comment.
Yes,
I
I
really
will
comment
the
short
summary:
let's
make
it
consistent,
let's
remove
compressed
because
compress
this
can't
believe
I'm
not
available.
Let's
remove
bites
right.
F
The
only
thing
I
think
that
is
interest
could
be
an
interesting
discussion
is,
do
you
want
batch
payload
size
or
it's
more
like?
Do
we
want
to
introduce
it,
or
can
we
do
it
later?
The
video,
if
it's
super
important
to
have
it.
E
F
Then
I
can
send
the
pr
and
I
can
update
the
attribute
names
to
be
consistent.
They
can
I
can
leave
a
comment
on
this
issue
or
we
can
review
it
if
you
are
first.
F
C
F
Yeah
and
on
the
other
one
so
that
cure
versus
topic
So,
based
on
that,
so
the
original
comment
was
referring
to
amqp
spark
so
basically
address
it's
not
an
entity.
It's
not
a
synonym
of
pure
topic.
It's
like
the
rather
endpoint.
Thank
you.
So
mqp
does
not
define
terms
like
your
topic,
but
it
defines
address
as
the
destination
was
this
in
mind?
F
G
F
Destination.Address
right
right
address
can
be
ambiguous
right
and
it
can
be
way
too
specific.
We
have
socket
address
if
we
need
to
in
general
attributes,
so
my
opinion
here,
it's
not
applicable
I
can
I,
don't
know
document
it
somehow,
but
I
don't
think
that
it's
the
right
abstraction
for
cure.
F
G
Yeah
and
I
think
the
main
suggestion
here,
though,
is
just
to
allow
it
allow
this
to
be
extendable
right
and
and
allow
like
if
q
and
topic
isn't
meaningful
in
the
messaging
system,
you
need
using
there's
some
other
word
there.
Then
we
just
allow
that
other
word.
F
Yeah
and
you're
also
had
a
great
point
on
this,
that
it
can
be
useful
in
socket
that
your
instrumentation,
even
though
it's
not
strictly
messaging
but
it's
just
allowing
extending
the
list,
Bianchi
and
topic,
would
be
a.
G
A
E
F
A
E
A
G
Well,
I
guess
the
point
being:
it's
kind
of
the
maintenance
is
a
little
bit
distributed,
which
makes
this
spec
easier
to
maintain
the
messaging
spec
and
then
there's
you
know
the
individual,
amp
Kafka
or,
whatever
you
know,
specific
definitions
of
how
they
extend
this.
You
know
is
sort
of
the
distributed
nature
of
the
definition
of
this
value.
They
all
have
to
list
any
new
values.
They
would
include
there.
G
E
Know
I'm
in
one
option
would
also
be
that
maybe
have
this
kind
of
as
part
of
the
general
convention.
Centimeter
is
the
messaging
system
specific
conventions
already
here
in
this
document
and
I
mean
one
thing
we
could
say
that
you
see
if
we
use
an
other
value
here,
then
that
should
be
in
the
system,
specific
conventions
that
should
be
mentioned,
but
then
it's
still
open
if
there
is
like
a
system
that
doesn't
have
any
open,
Telemetry
semantic
specific
conventions
and
I.
Think
Daniel
point
to
end
is
that
we
say
okay.
E
That
should
somehow
correspond
to
to
some
like
external
external
conventional
specification
and
should
not
be
just
any
kind
of
random
value
that
people
put
there.
G
Yeah,
although
my
understanding
is,
there's
gonna
be
like
a
core
messaging
spec
and
that's
what
we're
working
on
and
there'd
be
like
separate
specifications
for
sqs
Kafka,
ampp
Etc,
and
it
would
have
to
at
least
go
on
one
of
those
system
specific
specifications
before
it
could
go
on
any
instrumentation.
That
would
be
the
idea.
E
Okay,
so
basically
we
just
say:
okay:
here
we
have
keyword
topic
and
then
you
want
to
add
other
value
that
needs
to
be
defined
in
the
system.
Specific
conventions
also
here
in
this,
in
the
scope
of
this
open,
Telemetry
semantic
conventions.
E
G
H
The
only
only
problem
is
that
if
it's
in
separate
documents,
storage
ability
or
or
people
finding
them
might
be
harder,
but
if
they
are
looking
already
for
the
emergency
system,
then
they
see
that
there's
the
file.
So
that
could
also
also
be
another
way
you're
looking
at.
But
if
you
just
look
at
try
to
find
it,
maybe
they
already
give
up
because
it's
not
in
the
main
document.
But
yes,
all
right,
I
think
it's
not
a
big
deal.
D
H
Yeah
exactly
I
think
there's
something
not
not
similar,
but
the
database
specs
there's
a
list
of
all
the
databases
and
things
like
that.
But
it's
a
bit
different
from
here,
but
yeah
I
think
both
ways
there
are
pros
and
cons
and
it
it
would
be:
okay,
I
guess.
If
it's
separate.
A
A
H
Yeah
yeah,
so
so
the
the
code
generation
for
the
conventions.
It
works,
different
I,
guess
in
in
different
languages,
but
but
usually
it
just
goes
to
the
yaml,
all
the
ammo
files
and
then
just
and
gets
all
of
them.
So
maybe
maybe
what
you're
saying
it's
also
yeah?
Maybe
it's
possible
to
just
aggregate
all
them
and
then
put
under
the
same
category.
But
what
I
see
is
happening
is
that
there
is,
for
example,
when
the
code
is
generated.
H
Usually
people
put
the,
for
example,
the
AWS
attributes
inside
a
class
for
AWS,
and
then
all
the
things
are
inside
that,
but
again
like
the
instrumentation
Alters.
Writing
this.
They
will,
for
example,
install
the
semantic
conventions.
Library
I,
don't
know,
for
example
the
Java
one
they
will
install
it
and
then
they
will
look
for
their
attributes
and
they
will
find
it.
They
don't
need
to
care
about
the
other
attributes.
So
I
think
it's
it's
fine
that
is
separated.
H
A
H
Other
way,
yeah
the
way
most
to
do
today
is
that
there
is
a
so
there's
the
the
semantic
connections
tools
and
you
feed
it.
A
ginger
template
this
python
template
template,
so
you
feed
it
a
template
to
it
with
your
class
and
so
on,
and
then
it
will
just
read
about
us
from
the
mfi
on
and
then
replace
them
in
the
template.
H
The
way
you
tell
it
to
do
so,
if
you
want
to
create
an
email,
maybe
we'll
create
an
enum
and
so
on,
but
you
go
but
the
scripted
the
the
tools
work
by
by
systematic
versions
per
semantically,
so
you
will,
for
example,
iterate
to
all
the
yellow
files
and
then
one
by
one
it
will
apply
the
yellow
file
to
your
template.
H
So
you
ended
up
with
with
either
a
class
with
all
the
things
inside
separate
by
subclasses,
for
example,
or
different
files.
The
way
I
guess
it
it
doesn't,
it
can
change
whatever
you
want
to
do
against
this.
E
To
agree
that
we
kind
of
can
split
this
up.
We
basically
open
up
this
definition.
Here
we
say:
okay,
it's
not
just
restricted
to
keyword
topic
and
like
specific
messaging
system.
Conventions
can
kind
of
extend
this
and
when
yeah
I'm
not
sure
how
to
kind
of
best
model
this
in
the
in
the
Yammer
in
the
Yammer
files,
but
I
think
that's,
then
something
that
we
would
need
to
look
into.
F
H
F
E
Okay,
then,
it
seems
to
be
agreed
to
kind
of
to
kind
of.
E
E
Structure
my
domain
question
I
want
to
answer
whether
we
need
this
ability
to
add
links
after
span
creation,
but
I
also
saw
Alex
put
some
very
concrete
questions
here.
So
maybe
we
start
with
your
point
LX,
because
I
think
when
we
start
here
we
will
take.
We
will
take
longer
so
okay.
I
I
Okay,
hi
everyone
so
I
linked
a
pull
request
for
an
Otep
which
is
quite
old
and
got
stuck
in
some
discussions.
So
I
want
to
revive
this
and
and
brought
this
actually
to
to
the
specifications
like
this
week,
and
they
propose
to
discuss
this
in
this
group
as
well.
It's
the
other
one
actually.
I
Because
yeah
this
182
yeah,
so
the
context
is
so
some
of
the
systems
try
to
differentiate
whether
a
span
is
an
entry
point
or
a
service
or
not,
and
a
problem
we've
based
today
with
some
of
the
messaging
or
actually
one
concrete
example,
is
with
the
consumer
spans
is
that
we,
when
we
receive
a
span,
it's
not
possible
to
differentiate
whether
it's
a
span
that
has
a
remote
parent,
so
basically
a
continuation
of
a
of
a
trace
that
came
from
a
producer
or
whether
it's
a
span
that
is
just
doing
some
active
polling
on
the
messaging
queue.
I
So
in
that
case
it
would
be
a
local
parent,
and
the
proposal
here
is
to
actually
expose
the.
This
is
remote
attribute
from
the
parent
span
context
on
a
span
as
an
additional
otlp
yeah
attribute
on
the
span,
so
that
systems
can
basically
do
this.
Differentiation.
I
I
A
A
If
you
have
a
messaging
span
and
you
want
to
know
like
like
a
consumer
messaging
span-
and
you
want
to
know
the
producer
to
get
to
the
producer,
so
it
can
either
be
a
link
or
it
can
be
a
parent
of
pen
right.
But
if
you
have
a
parent
you
can
tell
if
it's
the
ambient
context
that
was
just
there
by
accident
or
if
the
parent
is
actually
linking
to
the
consumer.
So
you
can
like.
A
E
I
also
don't
see
any
like
issues
from
messaging
point
with
this
proposal
Alex
what
was
some
pushback
where
we're
getting
some
pushback
on
this
proposal?
No.
I
That
there
was
no
pushback,
but
I
think
we
are
still
missing
approvals
for
this
to
be
implemented.
So
yeah
just
want
to
raise
this,
so
there
was
some
discussion,
but
no
really
pushback
I
think
only
what
would
be
required
is
that
we
add
an
additional
method
to
the
API
in
the
specification
to
have
this.
Has
remote
parent
call,
but
it
wouldn't
be
a
breaking
change
right.
It
would
be
just
something
that
would
extend
the
current
functionality.
I
E
Yes,
I
think
from
my
side,
I
can
have
a
look
at
the
at
the
pr
and
maybe
add
comments
and
I
can
also
add
the
comments
that
we
that
we
looked
over
this
in
the
messaging
work
group
and
that,
from
our
point
of
view,
there
is
no
concerns
and
it
actually
seems
to
be
a
useful.
I
E
Okay,
thanks
Alex
for
bringing
this
here
and
then
I
would
say:
let's
switch
over
to
the
to
the
discussion
on
the
on
the
old
tab,
a
diverse
some
discussions
in
the
last
two
weeks-
I've
not
been
a
I've
missed
I've
missed
out
on
this.
E
Unfortunately,
but
some
one
thing
I
wanted
to
discuss,
and
maybe
that
ties
into
some
of
the
discussions
from
the
last
week
is
the
need
for
headings
band
links
after
spam
creation,
because
I
think
that's
something
we
need
to
figure
out,
because
if
we,
if
there
is
a
requirement
that
something
we
need
to
to
push
for
to
get
this
into
the
spec
and
one
of
the
main
use
cases
we
need.
This
are
the
are
the
pull
use
cases
let's
see
if.
E
If
I
see
one
yes,
yes,
I
see
this
here.
This
is
a
full
based
consumer.
The
process
is
bands,
and
here
we
have
this-
this
receive
span
here
that
basically
models
a
kind
of
a
call
to
receive
messages,
single
message-
or
in
this
case
is
it
receives
like
a
batch
of
two
messages
and
what
we
or
what
is
required
here
in
this
with
this
proposal,
is
that
this
receive
span
is
linked
to
the
publish
or
create
Span.
E
In
this
case,
there's
no
credits
by
digital,
publish
spends
to
see
span
has
links
to
the
published
spans
of
the
two
messages
that
I
received.
However,
in
many
cases
when
this
receive
operation
is
started
or
when
the
user
triggers
this
receive
operations,
it's
not
known
yet
which
messages
are
received.
The
message-
our
own
messages-
are
only
received
during
this
receive
operation
here,
which
means
that
we
cannot
add
links
to
those
public
span
for
those
received
messages
at
the
beginning
of
the
receive
operation.
B
So
one
one
question
I
have
regarding
this
and
just
to
be
clear:
I'm
not
opposed
to
the
idea,
but
a
potential
workaround
would
be
to
capture
a
timestamp
of
some
sorts
at
the
beginning
and
then,
when,
after
the
messages
actually
received,
then
kind
of
post
date
or
sorry
predate
like
the
the
start
time
to
what
it
was
beforehand.
F
And
there
is
a
problem
with
this
because
imagine
I'm
doing
the
pool
operation
and,
let's
say
I'm,
pulling
over
http
and
everything
that
happens
before
I
started
my
span
later
on,
like
all
the
HTTP
course
in
your
logs,
they
are
not
correlated
to
the
receive
operation.
F
C
F
E
E
A
I
think
there
are
two
concerns
regarding
this
change.
One
is
I
think
the
specs
would
be
explicit
that,
even
though
links
can
be
added
after
speculation
time,
they
should
only
link
to
a
context
that
started
before
the
span
like
so
there's.
Nowhere
loops,
like
someone
can
steal,
generate
a
scenario
when
the
links
create
a
loop,
but
at
least
saying
that
we
don't
want
it
to
happen.
A
This
is
one
thing
and
the
other
thing
is
regarding
sampling,
because
currently
sampling
receives
the
links
array
and
it
assumes
that
all
the
links
are
available
at
sampling
time,
and
this
will
change
this
assumption
so
like
it
has
to
be
clear
that
anyone
with
a
sampler
needs
to
be
prepared
to
handle
this
situation.
E
The
regarding
the
sampling-
we
had
some
discussion
about
this
with
Josh
from
the
sampling
stick
and
I
mean
his
his
take
on
this
is
that,
from
the
sampling
point
of
view,
adding
links
after
span
creation
should
not
be
a
problem
for
several
reasons,
also
that
currently
there
is
no
sample
that
takes
links
into
accounts
and
the,
and
even
if
there
will
be,
he
said
that
adding
links
after
Spain
creation
does
not
significantly
hinder
any
sampling,
sampling,
kind
of
assumptions
or
goals
that
we
that
we,
that
we
can
that
we
have
and
can
meet
so
from
from
from
his
from
his
from
the
sampling
point
of
view.
E
At
least
Patrol
said
he
said
he
will.
He
will
be
fine
with
the
change
and
from
the
other.
The
other
point,
the
circular
links
that's
an
interesting
point
of
view,
but
I'm
not
that's
an
interesting
thought,
but
I'm
not
sure
how
that
can
be
avoided,
because
when
they
receive
a
context
here
that
we
link
to
the
context
doesn't
tell
us
when
it
was
started.
C
A
It
doesn't
tell
us,
but
in
this
specific
case
we
know
there
cannot
build
a
Loops
right
like,
but
if
it
goes
into
the
spec
like
without
the
context
of
messaging,
like
saying
that
the
link
can
be
added
after
spend
creation,
then
we
should
add,
like
a
requirement
or
a
warning,
to
not
record
links
in
ways
that
create
loops
for
the
general
case.
Not
for
this
case.
A
Yes,
that
makes
sense,
and
regarding
sampling
there
is
one
use
case
which
for
which
sampling
is
important
like
sometimes
when
you
create
a
span
for
a
batch
of
messages,
and
you
want
to
decide
if
this
span
is
sampled
or
not.
You
can
enumerate
all
the
all
the
links
and
see
if
one
link
is
sampled,
then
sample
a
current
spend
right.
So
I
hope
it
was
clear,
but
I
think
the
benefits
of
adding
links
after
spend
creation
is
greater
than
the
sampling
issues,
so
I
totally
support
it.
E
Yes,
I
agree:
it
definitely
takes
away
some
guarantees
that
are
there
now,
when
you
say,
okay,
you're
guaranteed
to
have
order,
links
in
Spain
creation,
but
I
think
Josh's
point
was
that
this
guarantee
is
not
that
useful
for
for
sampling
anyway,
I
mean
there's
this
case,
I'm
ready
to
mention
work
and
say:
okay,
I
sample
this
receive,
spend
when
or
or
maybe
even
one
just
one
of
those
published
spans
are
Stanford,
but
that
leads
to
that
actually
leads
to
cases
where
then,
the
sampling
rate
gets
really
gets
out
of
control,
because
you
cannot
control
the
sampling
rate,
especially
when
you
think
about
having
large
batteries
you
may
be
receiving
like
100
messages.
E
A
Yeah-
and
there
is
one
additional
use
case,
which
could
be
helpful
to
have
a
link
and
adapters
pen
creation
and
so
I
encounter
the
situation
where
you,
when
you
set
a
timer,
then
when
the
timer
expires,
you
you
need
the
connection
to
the
context
that
register
the
time.
Also,
if
I'm
saying
do
do
something
in
30
seconds
and
I
have
a
new
Trace
when
the
30
seconds
expires,
I
want
to
know
where
it
came
from,
like
connect
the
two
with
the
link
for
this,
for
this
link
to
be
added.
A
What
I'm
currently
doing
is
I'm
using
a
span
processor.
The
span
processor
won't
start
a
hook,
but
this
processor
is
only
invoked
after
the
span
has
already
been
created,
and
you
cannot
add
the
link
at
this
point
of
time.
So
having
this
possibility
will
really
make
it
easier
for
this
use
case
to
be
implemented.
E
I
mean
regarding
a
special
dictionary:
I
have
a
question.
I
just
read
like
a
summary
on
slack
another
quick
chat
with
the
middle
about
discussions
you
had
in
the
last
week,
but
I
think
it
kind
of
touched
on
this
scenario
here
and
actually
like
involving
the
process
span
and
US
kind
of
like
also
spanking
out
process
bands,
especially
for
those
I,
think
a
poll
based
receive
scenarios
and
the
wonder
if
this
would
would
change
anything
on
this
picture
here
and
on
our
requirements
regarding
spend
links.
G
D
C
E
F
C
F
F
E
E
E
D
A
Formal
definition
describes
the
kind
based
on
the
pill
and
child
relationship.
So
if
it's,
if
the
parent
is
remote,
then
it
should
be
consumer
and
if
it's
local,
then
it
should
be
internal.
F
I
think
spark
oh
I'm
going
to
check,
but
I
guess.
G
G
A
I
I
agree
that
it
makes
sense
that
it
will
be
a
consumer
or
a
client,
but
the
I
think
it
depends.
If
it
has.
If
one
of
its
children
is
remote,
then
it
should
be
a
client.
This
is
the
current
wording.
I
don't
agree
with
them,
but
this
is
how
it's
currently
written.
F
The
client
I'm
I
have
the
spec
company.
Now
it
says
that
client
indicates
the
spend
the
scripture
requests
to
some
remote
service.
The
spot
is
usually
the
parent
of
a
remote
server
stand
and
does
not
turn
until
the
responses
was
received.
It's
not
requirement
to
be
a
parent
of
a
server's
plan,
okay,
and
it
makes
sense
if
I'm
writing,
HTTP
Clan
dot
dot.
Yet
right,
it's
a
client's
pen
here,
if
I'm
writing
service
bus
that
receive
it
makes
sense
that
it's
a
client's
plan.
It's
the
operation
I
initiate.
B
So
I
actually
have
one
other
case,
also
where
I
think
having
a
adding
a
link
after
spend
creation
could
be
useful
and
that's
with
the
the
the
processing
span.
A
Yeah,
so
if
you
go
to
the
second
paragraph,
the
one
that
starts
with
the
first
probability
a
bit
above
yeah,
so
it
said
it
says
biological
we
mean
that
the
span
is
logically
a
remote
child
or
parent
from
the
point
of
view
of
the
library,
that's
being
instrumented
spins
with
remote
parent,
but
so
saying
that
the
logical
means
that
it
doesn't
have
to
be
a
real
parent.
F
It's
saying
that,
since
the
receive
call
is
a
parent
of
broker
corresponding
server,
but
which
does
not
exist
right.
So
it's
not
that
there
have
to
be
a
child.
That's
never
guaranteed
to
have
a
child,
because
the
other
system
doesn't
can
it's
not
necessarily
instrumented
or
ever
will
be
right.
It's
the
operation.
I
initiate
that's.
My
client
call.
F
F
A
A
Yeah
yeah
I
agree
with
you
like
I.
Don't
think
the
text
is
useful
as
it
is
now,
but
I
need
to
read
it
into
the
details
and
I.
E
Think
it's
a
bit
misleading,
especially
this
this
for
this
part
here
where
it
says
that
the
consumer
span
indicates
that
the
span
describes
a
child
of
an
asynchronous
producer
request.
E
I
mean
that
makes
sense
in
the
term
of
like
The
Logical
child,
but
not
when
you're
interparent
right
relationship,
because
when
you
think
about
batching
scenarios
that
we
have
basically
a
consumer
span
can
have
several
logically
parents.
D
F
Described
what,
if
I,
if
I'm,
practicing
a
batch
of
messages?
I
cannot
use
consumer
kind
at
all.
A
E
Yeah
yes,
I
mean
I,
will
write.
I
will
write
an
issue
for
two
kind
of
for
revising
this,
this
text
here
and
there.
We
can
then
technically
this
and
take
this
up
as
we
make
progress
in
the
discussions
here.
H
Yes,
I
just
wanted
to
before
we
move
to
the
link
discussion.
I
just
wanted
to
to
post
a
link,
because
you
guys
had
some
examples
about
need
for
for
adding
links
after
this
pen
creation
and
there's
a
I'm,
not
sure
if
everybody
is
aware,
but
there's
an
issue
already
where
there's
a
lot
of
discussions
about
it.
And
if
you
have
your
examples,
then
it
would
be
great
if
you
can
post
there
in
the
issue,
there
was
a
PR
I.
H
E
Yes,
there's
lots
of
discussion
and
also
examples
connected
here
and
it's
interesting
because
yeah
again
not
again,
people
stumble
across
distance
at
add
their
use
cases
and
examples
here
so
yeah.
If
you
have
one,
please
add
it
to
because
if
we
decide
that
yeah
we
need
links
after
spring
creation,
then
we
will
do
it
in
in
the
in
connection
with
this.
Existing
issue
and
properly
are
reopen.
J
But
by
the
way,
I
have
a
small
question:
are
we
confident
that
this
will
be
a
blocker,
because
if
this
that's
the
case,
I
can
take
a
stop
myself.
I
think
that
one
of
the
problems
with
this
issue
is
that
there's
a
lot
of
discussion
happening
there,
but
somebody
needs
to
go.
Read
the
entire
thing
again,
summarize
things
and
get
it.
You
know
back
to
life.
I
could
take
that
task.
Myself
and
I
can
present
that
in
the
specification
next
Tuesday.
E
Yes,
that's
that
that's
that's
the
main!
That's
the
main
point.
I
I
want
to
clarify
today
whether
this
is
kind
of
a
requirement
for
us
whether
we
need
to
add
these
links
up
to
spend
creation.
I
think
we
do
and
I
just
wanted
to
get
like
alignment.
Okay,.
J
Perfect,
so,
okay,
so
if
you
have
like,
we
only
have
like
four
minutes
or
five.
So
in
the
meantime,
I
will
go.
Read
the
entire
thing
again
summarize
things
and
I
will
be
poking
you,
so
you
can
in
case
you
want
to
provide
additional
feedback
and,
of
course,
if
you
have
already,
some
extra
scenarios
in
mind
feel
free
to
pull
those
things
as
well
like
from
now.
E
Awesome
so
just
to
clarify
here
then
in
the
last
five
minutes
that
we
have
left,
regardless
of
whether
like
this
receive
Spain,
is
now
of
type
client
or
consumer,
and
we
also
need
to
clean
up
this
Bank
kind.
E
Definitions,
I,
think
in
terms
of
the
link
situation
here,
I
think
that's
how
we
want
to
model
this
also
because
maybe
there
might
be
additional
links
from
process
bands,
but
I
think
we
definitely
want
to
have
this
receive
operation,
and
we
want
to
have
this
receive
operation,
link
to
the
publish
context
of
the
messages
that
are
received
and
I.
E
Think
if
we
want
to
have
this
modeling,
then
we
need
to
add
spam
links
after
span
creation
and
Trust
yeah
wanted
them
to
check
here
whether
there
is
any
doubts
about
this
way
of
modeling.
This.
This
scenario
over
the
whether
there
is
kind
of
anybody
thinks
there's
alternative
solutions
that
would
not
require
adding
spending
software
span.
Creation.
A
So
I
think
it
is
needed
for
this
case
at
least,
but
there
is
also
another
case
where
it
is
needed.
Maybe
it's
not
a
receive?
It's,
not
a
bullying
request,
but
it's
a
like
a
push
request,
but
messages
are
live.
Arriving
is
a
stream,
so
you
can
tell
at
the
beginning
of
the
operation.
You
can't
tell
all
the
messages
in
advance.
You
get
them
one
by
one
like
iterate
over
them
and
in
this
case
we
will
also
need
the
links
to
be
created
after
spend
start.
A
Well,
this
came
up
like
I
think
or
two
ago
in
the
JS
contributor,
someone
tried
to
write
an
instrumentation
for
a
library.
I
can't
remember
the
name,
but
I
can
look
it
up
and
they
got
stuck
at
this
situation
where
it's
like
iteration
API,
and
it
can
tell
all
the
messages
in
advance.
So.
J
J
That
would
be
fantastic
and,
of
course,
you
don't
have
to
go
and
check
the
actual
details
like
as
long
as
you
point
me
to
that
specific
framework
or
broker
I
can
just
dig
myself,
but
yeah
could
be
fantastic.
E
That
that
big
that'd
be
awesome,
so
some
things
Carlos
for
taking
this
over
I
mean
I
will
also
join
then
on
Tuesday
and
then
maybe,
if
there's
any
like
questions
or
doubts,
then
we
can
yeah
have
a
discussion
on
Tuesday
in
the
overall
spec
meeting.
E
This
point
clarified
I
think
we
still
still
have
some
open
questions
about
the
those
link
parent
trial
relationships
also
about
span
kinds
here,
especially
in
that
context,
but
yeah
as
there's
only
one
minute
left
I
think
we
will
Reserve
that
for
next
week
and
also
amir's
document
here
and
also
titles,
question
I
will
move
that
over
to
to
next
week's
agenda
and
then,
let's
say
continue.
E
Continue
continue
there
because
yeah
I
think
the
processing
span.
That
is
a
great
open
question.
The
link
semantics
that
is
anyway,
something
we
need
to
clarify
to
and
I
will
also
add
a
discussion
point
about
the
span
kinds
in
especially
this
scenario
here
and
then,
let's
continue
discussing
next
week,.
B
Would
it
be
helpful
to
have
another
meeting
to
move
things
along
faster
before
then.
C
E
Think
discussion
time
is
always
always
valuable.
I
think
the
challenge
is
just
to
find
like
a
meeting
time
where,
where
a
significant,
where
we
get
a
significant
quote
of
people.
C
J
Yeah,
actually
we
we
are
on
time,
but
I
wanted
to
say
that
think
about
these
as
a
repossibility.
You
know,
like
the
HTTP
instrumentation
group,
I
think
they
are
meeting
two
times
a
week,
trying
to
speed
up
progress
so
that
that's
always
a
possibility
so
think
about
that.
E
I
mean
I
will
add
like
a
question
to
slack
Channel
about
the
about
the
days
and
times
I.
C
E
We
would
need
to
have
it
in
the
morning
because
we
have
some
significant
course
of
APAC
or
contributors
from
Europe,
so
it
would
need
to
be
like
probably
an
eight
a.m.
Meeting
and
yeah
I
will
I
will
post
on
slack
about
the
about
to
kind
of
see
what
availability
people
have
for
for
an
additional
8
Dam
meeting
another
weekdays
right
and
then
we
can
go
from
there.