►
From YouTube: 2022-08-23 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
A
A
A
A
So
it's
kind
of
the
reverse
of
what
we
already
offer.
So
cedro
has
a
comment
on
here,
something
about
like.
Should
we
even
have
this
in
the
repo?
Does
it
make
sense
for
us
to
provide
this?
I
think
so,
but
I
just
wanted
to
get
some
other
opinions
and
maybe
discuss
it
a
little
bit
so
like
what
what's
our
like
general
plan
for
the
cellular
stuff,
it's
not
long
time
plan
to
be
in
the
main
trip
right.
It
would
be
like
moved
away
to
the
country,
people.
A
So
this
this
should
be
a
contrib.
Yes,
it
can't
today,
because
the
log
emitter
stuff
is
all
internal.
So
right
now
the
siri
log,
extensions
and
the
event
source
extensions
are
in
the
main
repo
once
that
stuff
goes
public,
we
should
move
it
over
to
contrib,
so
the
the
shims
can
be
moved
to
the
contract.
One
way
once
we
expose
publicly
the
thing
which
they
need
like
log
emitter
sdk
stuff.
A
A
I
feel
like
it's
nice
to
have
it
all
in
one
package
yeah,
I
mean
the
pro
program
which
I
see
here
is
like.
There
is
nothing
open
elementary
specific
here,
which
makes
me
wonder
like,
should
we
like
even
increase
that,
because
it's
probably
like
fine
to
have
it
like
once,
but
then
we
basically
open
the
gate
for
more
things
in
the
future.
A
So
if
we
do
not
intend
to
have
that,
I
mean
it's
not
an
open
element
respecting.
So
if
you
do
not
intend
to
have
it
in
the
main
point
maintain
it,
then
we
should
not
do
it
like
one
offs.
We
can
do
it
in
the
control
grip
as
a
separate
component,
so
we
obviously
make
it
clear
that
we
don't
want
it
in
the
same
package.
So
if
anyone
wants
to
add
a
new
enricher
for
something
else
like.
B
A
Want
to
enrich
with
baggage-
or
they
want
to
do
something
more-
they
can
create
their
own
component
in
the
country
just
to
keep
the
bar
high.
Otherwise
we
will
be
like
managing
a
lot
more
stuff
than
we
then
do,
because
this
is
pure
serious
meaning.
If
I
look
at
the
code
correctly,
there
is
no,
I
mean
it's
related
to
upon
elementary,
but
it's
really
just
some
interesting
for
siri.
A
B
B
B
Yeah,
I
don't
exactly
I
mean
you've
looked
at
it,
the
most
but
yeah
somehow
somehow
pulling
those
out.
If
that's
even
possible
is
my
question.
I
guess.
A
Right
now,
how
you
get
a
log
emitter?
Is
you
get
it
off
the
logger
provider?
So
it's
kind
of
tied
to
the
sdk.
A
A
That
is
essentially
that
I
feel
like
we
had
a
need
for
just
yet
another
thing,
but
I
could
try
it
I
mean.
Can
we
use
the
internals
visible
to
a
project
in
contract?
Like
one
house
like
the
steady
log
integration
which
we
have
sorry,
not
the
interest
or
the
overall
cellular
and
even
source
thing?
Can
we
keep
it
in
the
corner
ripple
and
still
have
the
internals
visible
from
the
main
report
to
that
individual
packet
project?
Is
that
like
technically
feasible?
A
A
A
A
A
So
that
would
be
like
slightly
like
different,
but
conceptually
similar
like
we
just,
would
play
the
current
baggage
and
attach
it
to
dogs
like
similarly
for,
like
even
the
existing
prometheus
exporter,
not
the
open,
elementary
premises,
exporter,
the
official
or
unofficial
prometheus,
and
they
could
also
use
something
like
this.
They
want
to
attach
the
baggage
as
an
additional
label,
so
it
could
be
a
prometheus
specific
component,
even
though
it's
related.
A
That's
why
I
was
saying
that
we
should
not
do
it
in
the
first
place.
We
should
see
if
we
can
do
that
in
the
contract
before,
like
all
the
serial
related
stuff
single
package
would
be
better
unless
anyone
I
mean
I
I
like
I
mean
I
share
the
concerns
like
we
don't
want
to
have
like
too
many
packages
for
cereal,
even
in
the
country
report,
so
we
could
potentially
bundle
that
into
all
same
package.
A
B
A
60
percent
of
what
you're
saying
sorry,
I
lost
a
little
bit.
Could
you
repeat.
A
Yeah,
your
voice
is
a
little
muscle
in
c2,
okay,
sorry
yeah,
I'm
in
a
different
laptop
in
a
different
environment.
So
sorry
about
that,
that's
much
better
yeah.
So
I
was
just
saying
like
first,
we
should
see
what
is
the
general
practice
of
study
log
in
richards
like
do
they
do?
Sirilogue
has
an
official
report
where
people
can
contribute.
If,
yes,
then
we
can
simply
contribute
it
there,
so
it
will
be
officially
maintained
by
the
city
law
community
instead
of
us,
because
this
does
not
require
any
internal
access.
A
Okay,
do
I
have
some
trouble
hearing
as
well,
so
maybe
I'll
just
reconnect
and
come
back
in
a
couple
of
minutes,
but
meanwhile
you
can
continue
this
discussion.
Let
me
try
to
reach
out
yeah.
I
think
he
said
like
to
see
what's
the
general
practice
for
contributing
to
like
steady
dog
and
riches
and
if
there's
some
other
place
that
we
could
just
contribute
to
instead
of
hosting
it
on
our
depo
or
contra
people
on
the
main
people
counter
people,
then
we
would
probably
want
to
do
that.
B
B
Yeah,
I
think
it's
worth
investigating,
but
I'm
just
doing
a
quick
search
right
now,
just
nougat.org
serial
log-
and
I
don't
know
I
don't-
have
a
solid
sense
of
things
yet,
but
it
seems
like
there
are.
B
There
is
a
collection
of
like
official
stereolog
packages,
but
all
of
them
seem
relatively
generic
or
like
not
like
associated
with
any
specific
vendor
concern.
It's
not
like
stereolog
is
hosting
a
contrib
repository
of
sorts,
at
least
that's
what
it
looks
like,
but
there
are
vendor-specific
serial
log
packages,
but
they're
usually
produced
by
those
vendors
or
hosted
by
those
vendors
from
what
I
can
tell.
B
There's
a
cereal
log,
syncs
application
insights
package
yeah.
I
think
I
would
probably
agree
with
that
assessment
as
well.
When
I've
looked
at
their
stuff,
they
don't
really
make
a
distinction
between
you
know:
surlock,
proper
and
stereolog
community
that
I
remember
looking
through.
I
mean
it's
obvious
when
you
start
looking
at
their
documentation,
which
ones
they
kind
of
highlight,
but
then,
like
these
other
ones
like,
I
saw
microsoft,
eye
logger
in.
A
B
B
A
B
A
It's
all
work
with
the
contributor
on
like
the
readme
updates.
We
can
merge
it
in.
We
haven't
released
it
yet
so
we
can
decide
if
we're
going
to
keep
it
here,
move
it
to
contrib
toss
it
out
completely.
A
I
mean
my
goal
with
creating
this
sirilog
package
with
the
log
emitter
was
so
that
we
could
put
out
an
alpha
or
a
beta
and
get
feedback
on
the
log
emitter
api.
Does
this
work?
Does
it
fit
in
appropriately
stuff
like
that?
So
I'd
like
to
get
it
out
there
at
some
point,
but
there's
no
no
timeline
or
urgency
behind
it
yeah.
A
Since
we
just
finished
like
two
alphas,
maybe
next
release,
we
can
start
releasing
the
siri
log
stuff
and
if
you
ever
do
that
lift
and
shift
from
the
main
ripple,
it
would
be
like
the
users
wouldn't
notice,
because
you're
not
changing
the
package
name,
just
the
repo,
where
we
host
things
and
at
that
time
it
would
probably
be
okay
to
have
the
integers
there.
A
Yeah
because
there
is
a
good
chance
that
one
of
the
maintainers
from
the
main
report
would
be
owning
the
serilock
like
adapter
or
ship,
whatever
we
decide
to
call
it,
but
then
there
is
like
add-ons
on
top
of
it.
So
we
could
decide
at
that
time
should
we
bundle
it
into
the
same
package
as
planche
was
suggesting
or
should
be
like
movie
two
separate
packages
like
all
the
entrances,
because
this.
B
A
One
literature
for
activity,
then
we'll
have
package
and
like
what
else
there
could
be
more
things
coming,
so
we
can
decide
it
like.
Once
we
reach
that
stage,
I'm
generally
concerned,
the
core
report
should
not
be
like
having
too
many
things
not
really
expected
from
the
open
elementary
spec
in
the
country
report.
We
have
like
more
flexibility,
we
can
host
it.
Then
the
only
question
is,
should
be
split
it
into
its
own
package
or
bundled
into
the
overall
city
log
thing
itself.
A
I,
if
there's
no
other
agenda,
there's
one
issue
that
and
also
I
see
vishwash
is
here,
so
we
can
talk
about
that.
Okay,
can
you
update,
like
all
the
releases
like
all
the
releases
were
done
last
week
right?
Oh
yeah,
like
the
alpha
2
release
was
done.
Last
week,
yeah
I
saw
like
someone
had
opened
an
issue
yeah.
I
tagged
you
there.
I
don't
know
whether
it's
related
to
the
newly
released
package
or.
A
A
Yeah,
this
is
from
9.4
to
9.5.
This
is
something
that
I
remember
telling
you
also
about
it
and,
like
I
was
looking
at
that
prometheus
exporter.
Demo
right
will.
B
A
Yep,
oh,
it
has
a
reaper
also,
so
we
should
be
able
to
quickly
check
that
yeah
I
mean.
Basically,
we
want
to
make
sure
all
the
most
recent
releases
they
all
work
with
each
other
very
well
like
if
something
doesn't
work
either,
because
we
did
not
release
some
component
or
we
broke
something.
We
want
to
fix
it,
because
we
generally
ask
people
first
thing:
okay,
go
and
update
you
the
most
recent
and
it
should
be
in
a
working
state.
A
Yeah,
and
also
like
with
the
latest
one-
this
is
not
a
package
anymore.
I
mean
the
package.
Name
has
changed
yeah,
so
maybe
like
that.
That
would
be
the
reason
why
they
are
missing
the
method
access
exception,
because
the
error
is
thrown
from
app
dot
use
prometheus
endpoint.
So
they
are
missing
that
package
like
open,
elementary.prometheus.asp.net
core,
maybe
that's
the
fix
for
this
issue,
but
that
was
only
introduced
last
week
right,
so
that
was
okay
yeah.
This.
A
A
A
I
want
to
quickly
ask
like:
was
there
any
discussion
about
the
sirilogue
versioning?
Do
we
want
to
question
it
as
the
known
core
component,
or
do
we
want
to
give
it
a
individual
question
number
any
anyone
discuss
that
or
do
we
want
to
discuss
that
now?
We're
in
we
could
discuss
that
now
because
at
least
the
next
release
onwards,
we
should
start
doing
it,
because
that's
the
only
way
we
can
validate
things
are
moving
in
the
right
place.
So.
A
B
A
Yeah
I
mean,
or
like
my
blinds,
like
any
like
thoughts
on
that,
you
don't
really
want
to
tie
it
with
the
core
sdk
anyway.
So
then,
the
only
other
option,
right
now
with
the
current
setup,
is
non-core
which
would
make
it
like
rc
9
dot,
something
or
we
could
do
like
individual
version
like
just
just.
Let's
just
call
it
like,
like
some
version,
followed
by
alpha,
because
it's
like.
Oh
it's
not
like
a
release
candidate.
A
B
A
Yeah,
okay,
okay,
so
when
we
do
the
next
release
like
like,
maybe
we
just
think
and
make
sure
like
we,
we
don't
have
like
the
the
script
which
we
use
for
releasing
the
manual
instructions.
It
does
not
take
into
account
any
third
time.
We
only
have
two
tags,
the
core
one
and
then
on
core
one.
Now
we
have
potential
third
or
fourth
one.
So
we
should
probably
need
some
updates
to
the
actual
release
process
to
make
sure
like
we
don't
produce
a
wrong
version
in
any
case.
A
So
when
we
okay,
let's
agree
to
like
do
the
release
whenever
we
do
the
next
one,
we
could
probably
do
it
by
like
in
the
next
couple
of
weeks,
because
we
have
a
lot
of
ti
related
improvements
coming.
So
maybe
we
can
club
that
into
the
next
one
and
start
the
releasing
of
blogging
adapters
as
well.
A
Okay,
yes,
there
is
no
conflict.
We
will
mark
that,
like
we
will
release
it
independently,
as
in
the
version
being
independent
from
core
and
on
core.
B
B
This
issue
is
being
addressed
by
fish
wish
in
a
different
way,
but
I
think
that
again,
in
the
spirit
of
of
landing
folks
works,
I
would
like
to
open
my
I'm
suggesting
that
I
would
open
this
back
up
and
I
think
that
there
was
just
some
conflicts
requests
that
he
resolved
those
and
then
I
think
it's
a
good
improvement,
but
yeah
it's
gonna,
be
it's
gonna,
be
kind
of
much
of
it's
going
to
be
overwritten
by
what
I
think
you're
doing
right.
Please
rush.
A
Yeah,
that's
right,
but
I
think
it's
still
good
to
watch
this
one,
because
this
basically
fixes
the
reflection
part
that
we
are
doing
so.
I
think
like,
irrespective
of
the
changes
that
I
am
doing.
I
think
this
is
still
helpful.
So
can
can
you
reopen
like
ln
and
yeah
the
like?
Is
there
conflicts
you
study?
There
are
conflict.
B
A
A
Okay
sounds
good
thanks
for
bringing
it
up.
Maybe
we
should
occasionally
look
at
all
the
stale
one
and
see
if
there
is
anything
which
we
should
practically
reactivate,
maybe
like
once
a
month
or
so.
A
And
yeah
there
is
that
other
pr
which
I
think
I
thought
I
think
it's
not
gone
still,
but
the
histogram
saying
so.
I
think
we
can
have
that
in
our
next
alpha
release
or
next,
whatever
at
least
we
do
so,
maybe
like
I,
I
left
it
unmerged
to
give
blanche
a
chance
to
look
at
it.
A
I
think
I
upload
it
and
yeah
I'm
just
there,
but
I
don't
think
he
explicitly
approved
so
maybe
randy.
If
you
can
take
a
look
see
if
it
looks
good.
A
A
Is
we
should
be
able
to
modify
the
stress
test
to
do
like
validation
as
well?
So
in
the
like
early
days
of
the
sick,
like
we
had
its
trust
us,
which
riley
was
using
to
test
the
circular
buffer?
So
it
had
this
violation
like
you
produce
like
like
10
million
or
100
million
whatever
number
of
events,
and
then
you
check
whether
all
of
them
were
trained
out
correctly
from
the
circular
buffer.
So
it
kind
of
knows
that
at
the
end
of
this
test
the
test
was
not
only
testing
the
stress.
A
That
way,
we
have
like
very
high
confidence
that
we
can
let
it
run
for,
like
days
in
different
architectures
in
different
machines
like
32-bit
64-bit,
and
confirm
that
okay,
we
are
not
like
having
any
thread
issue,
because
right
now
we
can
run
trust
us
for
like
days.
We
do
it
all
the
time,
but
there
is
nothing
which
validates
that
the
correctness
yeah,
if
you
can
create
an
issue
or
maybe
there
is
an
issue
already.
If
not,
can
you
create
one
so
we
can
work
on
it
later.
A
Should
I
even
search
for
it?
I
don't
know
if
there
is
anything
yeah
yeah
I
mean
you
can
clear
them,
because
the
fact
that
we
did
not
have
a
I
mean
the
this
pr.
Without
this
pr
the
code
was
broken,
but
we
did
not
catch
it
with
any
unity
so
that
that's.
That
means
we
have
a
testing
gap
for
multi-threaded
scientists,
so
we
can
create
an
issue
just
for
that.
I
mean
whether
we
solve
it
with
stress
test
or
something
else
that's
to
be
discussed
in
the
offline.
A
So
he
said
like
there
was
an
issue
to
be
discussed
later
on,
so
it's
this
one
and
I
think
cj.
You
also
commented
on
it
today,
but
like
yeah,
I
tell
you
that
this
one
it
does
have
this
there
is.
This
is
a
legit
issue
and.
A
A
But
when
I
used
to
see
cli
commands
dot
net
run,
then
I
could
see
that
the
issue
was
actually
there.
I
mean
the
the
parent
id
is
not
set
correctly
and
there's
a
reason
for
that,
but
one
why
it
like
it
worked
on
visual
studio
run,
but
not
on
dot
and
cli,
and
so
the
original
issue
is
this.
There
are
two
apps
two
ways
core
apps
and
when
he
sets
asp.net
core
instrumentation
and
http
client
instrumentation,
when
he
enables
both
of
that
on
both
the
apps,
then
all
the
propagation
is
correct.
A
A
But
if
the
caller
app
has
the
http
client
instrumentation
disabled.
A
A
B
Is
this
feels
kind
of
similar
to
the
the
grpc
instrumentation,
where
you
know
grpc
built
on
top
of
http
will,
if
you
have
both
enabled
it
creates
two
spans?
B
Did
something
to
support
that
to,
and
I
forget
what
I
think
I
did
it,
but
it
was
a
while
ago
you,
the
grpc
span,
would
be
the
remote
parent,
and
maybe
we
need
to
consider
doing
something
like
that
here
with
asp.net
core,
but
it
might
be
trickier.
A
So
vishwash
do
you
wanna?
Like
talk
more
about
this,
I
think
you
investigated
a
lot
more.
B
A
So
apparently
it
creates
its
own
activity
if
there
is
a
activity,
dot
current,
which
is
not
null
so
basically,
in
this
case
you
have
an
asp.net
core
instrumentation
enabled
it
creates
the
activity
and
then
the
user
is
making
http
client
call,
but
the
http
client
instrumentation
is
not
enabled.
However,
the
asp.net
core
activity
is
active,
so
the
http
client
still
chooses
to
create
its
own
activity
and
uses
that
activity
as
a
trace
parent.
A
So
when
we
get
the
transparent
header
on
the
node
b,
their
asp.net
code
just
basically
uses
that
transparent
to
attach,
like
initialize,
the
the
parent
ids
there
for
the
respirator
activity.
So
in
that
screenshot,
where
you
like,
saw
that
invalid
id
that
invalid
span
id
is
coming
from
http
client,
but
because
we,
the
user,
is
not
instrumenting
it.
We
are
not
exporting
that
activity.
So
it's
just
basically
like
it's.
A
It's
like
it
shows
in
the
the
back
end,
it's
like
an
invalid
id,
but
so
that
that
is
the
reason
that
this
user
is
seeing
this
issue
one
second,
but
there
is
no
bug
here
right.
There
is
no
incorrect
spin
id
it's
just
that
the
user
did
not
export
the
history,
so
so
the
bug
yeah,
so
I
mean
like
there's
like
it
is
by
design,
but
from
user's
perspective.
It's
it's
a
bug
because,
ideally
like
they
are
not
instrumenting
http
clients.
A
So
then
what
should
happen
is
the
asp
net
core
activity
should
be
the
parent
of
the
asp
net
core
activity
of
project
b,
so
project
a
has
incoming
request
coming
in
asp.net
core
creates
the
activity
that
activity
should
be
the
parent
of
the
activity
created
in
project
b
of
asp.net
core,
not
sure
whether
we
can
do
that,
because
this
it's
like
imagine
like
other
cases
where
http
client
is
indeed
creating
activity.
We
have
enabled
instrumentation
right,
so
there
could
be
a
sampler
which
decides
to
drop
like
that
particular
span.
A
Only
then
also
we'll
get
the
same
thing,
so
it's
more
about
like
incomplete
trees,
because
some
spans
were
not
exported.
It
could
happen
because
of
like
sampling
or
like
just
telemetry,
being
simply
dropped
by
the
back
end
or
lost
in
transport.
In
this
case
it's
not
even
exported.
So
there
are
like
several
cases
where
this
can
really
occur
right,
so
I
I
don't.
I
mean
I'm
still
trying
to
see
whether
it's
something
which
we
should
do
anything
about.
B
I
mean
the
tricky
thing
about
this
is
that
it
is.
I
believe
that
this
is
a
different
experience
than
what
users
would
get
from
open,
telemetry
in
other
languages,
given
that,
like
in
this
case,
if
they
did
not
enable
http,
client
instrumentation
like
in
java
or
something
right
or
whatever
their
equivalent
is
then
open.
Telemetry
never
actually
creates
that
spam,
and
so
there
is
no,
there
is
no
span
id
that
is
missing.
Okay,.
B
I
just
had
to
refresh
my
memory,
but
basically
it's
it's
easy
here,
because
you
know
grpc
is
right
at
the
same
point
of
like
http
client,
it's
like
the
the
call
is
going
outbound.
So
we
have
this
dialogue
here,
saying
hey
if
we're
suppressing,
downstream
instrumentation
or
whatever
it's
just
not
invoked
for
some
reason,
then
we
make
sure
that
the
grpc
span
is
the
remote
parent.
B
A
I
have
added
one
link,
also
in
the
docs
as
well.
There
is
a
related
similar
issue
because
we
currently
have
the
behavior,
where
we
do
not
create
the
activity
itself.
If
the
sampling
decision
is
unfavorable,
but
just
text
says
otherwise,
we
should
be
creating
a
new
span
with
the
new
spidey
new
spaniard,
even
though
we
are
not
going
to
export
it.
A
So
in
this
case
also,
the
issue
of
like
missing
span
would
occur,
because
if
you
have
like
abc
and
b
is
the
middle
guy
and
we
did
create
a
new
span,
so
c
thinks
that
its
parent
is
b
but
b
is
not
exported
because
of
sampling,
so
they'll
the
exp,
the
end
user
experience
would
be
like
very
similar,
but
it
manifests
due
to
a
different
reason
than
this
one.
So
it's
very
much
related
to
this
issue.
A
Ultimately,
what
you
see
in
like
zikken
or
like
your
back
end
is
okay.
You
have
a
trace,
you
have
spans
from
the
same,
but
there
are
like
broken.
They
are
broken
in
the
middle
because
you
cannot
find
the
parent
spine
id.
You
cannot
find
the
span
with
which
has
been
used
as
a
parent
spine
id
in
some
places,
so
both
looks
same
and
yeah.
I
mean
in
this
case
for
http
client.
We
probably
want
to
do
something
to
mitigate
or
even
like,
if
it's
not
possible,
then
educate
users
to
do
something.
B
A
A
Like
one
other
thing
is
like
I
don't
know
whether
is
this
an
issue
even
in
dot
net
seven,
where
http
client
uses
the
new
way
of
creating
yeah.
A
B
A
See
no,
I
mean
this.
This
has
like
it's
not
related
to
like
the
activity
source
or
anything.
It's
it's
just
like
the
http
client
decides
to
create
the
activity.
If
there
is
a
activity
dot
current,
not
null
so
yeah,
I
I
actually
meant
to
ask
it
slightly
differently,
so
in
dot
net
seven
before
dot
number
seven
http
line
will
always
create
the
activity
by
using
the
new
activity
constructor.
So
there
is
a
guarantee
that
it
will
always
create
activity.
A
Now,
in
the
second
case,
where
they
do
the
activity,
source,
dot,
start
activity
and
if
you
do
not
enable
it,
there
is
no
one
subscribing
to
that
particular
activity
source.
In
that
case,
it
will
not
create
the
activity
at
all,
but
we
know
that
it
falls
back
to
the
old
way
and
still
creates
activity.
But
without
that,
like
backup
fault
fallback
option,
it
would
have
been
just
like
java
or
any
other
language,
because
there
is
a
library
called
http
client.
It
calls
like
start
span
or
stat
activity.
A
It
detects
that
okay,
no
one,
is
interested
in
my
expand,
so
I
won't
even
create
it,
but
when
it
propagates
it
will
propagate
what
is
available
in
the
context.
Would
that
would
happen
to
be
the
asp.net
core
one,
so
in
that
case
we
would
be
able
to
see
the
same
expected.
The
user
expected
behavior
in
this
case
the
receiving
side,
would
parent
to
the
request,
request
the
asp.net
core
request,
not
the
http
client
one.
A
So
that's
why
I
was
wondering
like
in
dot
net
seven
is
that
fallback
mechanism
to
always
create
activity
was
not
there.
We
would
not
be
facing
this
issue
starting.net
seven.
Oh
yes,
yeah-
that
that
is
right.
Yeah,
however,
like
this,
this
issue,
surfaces
and
dotnet,
seven
as
well
yeah.
A
Now
it
for
sure,
because
if
stat
activity
returns
null
then
immediately
http
client
releases,
we
need
to
fall
back
and
create
the
activity
using
the
legacy
way.
Yeah
yeah
just
copy
paste
it
where
it
comes
from
in
the
http
client.
Like.
Oh
sorry,
I
copy
paste
it
in
the
chat.
A
Yeah,
that's
exactly
so.
It
tries
to
create
using
the
new
way.
If
it
is
now,
then
it
uses
the
fallback
mechanism
which
is
to
do
the
constructor,
so
that
is
done,
probably
not
probably,
that
is
done
to
maintain
backward
combat.
I
think
yes,
yeah
yeah,
that's
right!
Yeah!
It's
for
backward
compatibility.
B
Actually,
question
about
this,
so
so
61
right
here
it
is:
where
is
the
new
one
where
it's
creating
it
from
the
activity
source?
B
And
if
that
happens
and
say
they
don't
have
the
http
client
or
I
guess
we're
looking
at
yeah,
the
http,
client
and
instrumentation
library
enabled
do
we
export
a
span
that
has
not
been
enriched
with
like
open
telemetry's
conventions.
A
A
In
61
it
tries
to
create
the
activity
using
the
open,
limited
api,
the
activity
source
api.
If
http
client
instrumentation
is
there,
then
we
have
that
add
source
for
this
particular
source,
so
we'll
create
activity.
But
if
it's
not
there,
then
the
activity
won't
even
be
created,
so
in
61
they'll
get
a
null.
So
at
that
point
it's
very
similar
to
like
java.
A
Like
you,
you
don't
it's
as
good
as
saying
instrumentation
is
not
enabled
for
that
particular
library,
but
http
client
wants
to
maintain
backward
compatibility,
so
it
tries
to
create
like
activity
by
hand
by
using
the
new
activity
constructor
like
bypassing
everything
in
the
directory
yeah,
and
in
that
case
we
do
get
that
like
active
in
v.
In
the
I
mean
your
question
was
like:
do
we
export
it?
In
that
case,
right.
B
A
Yes,
so
61
would
never
be
null
if
open
elementary
http
client
instrumentation
is
enabled
oh
okay.
It
can
still
be
null
due
to
sampling
so
and
if
you
choose
to
have
a
sampler
which
happens
to
drop
this,
then
yes,
but
otherwise
we
would
force
the
activity
to
be
created
in
61
itself,
and
then
we
enrich
it
with
all
the
additional
tags
using
the
like
payload.
A
Not
to
my
knowledge,
so
if
you
have
a,
if
you
have
a
like
tracer.startspan.
A
B
A
If
I
ever
think,
like
it's
the
fallback
mechanism
which
might
be
at
play
here
like
64
to
689,
we
have
the
fallback
like.
If
the
activity
is
not,
I
mean,
if
activity
is
null,
it
means
okay,
open
elementary
says,
don't
create
the
span,
but
here
it's
not
respecting
it
and
still
creating
it,
just
to
keep
its
own
backward
compatibility,
and
this
would
not
be
the
case
with
any
other
languages.
A
So
if
you
do
not
have
that
block
of
code
there
like
64
to
69,
then
all
that
happens
is
http
client
attempts
to
create
a
new
span.
Then
it
releases
a
set,
not
needed.
So
all
it
does
is
propagates
activity,
dot
current,
which
would
be
the
asp.net
core
one,
and
in
that
case
we
wouldn't
have
the
missing
span
or
missing
connection
issue.
A
A
So
like
we,
when
we
run
when
we
run
this
on
visual
studio
through
the
visual
studio
run,
it
works
right
because
then
we
have
a
listener,
enabled
yeah.
So
on
visual
studio
like
in
the
debug
mode.
I
think
there
is
another
diagnostic
listener
present
which
gets
attached
via
enter
interest,
and
they
have
like
a
very
odd
looking
condition
where
it
says.
A
If
the
diagnostic
listener
is
enabled,
then
don't
create
the
activity,
so
it
doesn't
create
the
activity
and
it
just
works
so
like
it
will
it
behaves
differently
in
debug
mode
as
compared
to
like
so
you're
saying
in
debug
mode
line.
68
is
not
executed,
it
doesn't
create
yeah
yeah,
exactly
that's
same
as
like
not
having
this
fallback
mechanism
at
all,
right
yeah,
but
yeah,
that's
in
only
in
debug
mode
like
and
just
start.
If
you
do
dot
net
run,
which
or
release
mode
like
even.
A
A
It
would
be
like
a
lot
of
hacks
involved,
but
yeah,
let's
see
like
what's
the
best
way
like
we
need
to
investigate
how
to
best
fix
it
like
should
we
work
with
the
http
client
owners
to
probably
provide
a
flag
to
disable
this
fallback
creation,
or
it
could
be
like
something
else.
I
think
we
need
to
spend
some
time
to
understand.
What's
a
best
way
to
solve
this.
A
I
think
the
the
linked
issue,
which
I
mentioned,
where
even
if
the
sampler
returns
null
it,
has
the
similar
thing,
even
though
it
manifests
in
a
different
way.
This
is
also
like
it.
This
will
also
have
the
equal
implication
like
when
you
export.
You
would
see
that
span
says
hey.
My
parents
fan
id
something,
but
no
matter
where
research,
you
will
not
find
that
parent
span
anywhere,
because
it's
simply
dropped
by
the
sampler.
A
A
I
mean
no
but
like
here.
We
can
just
put
it
on
the
user
right
that
their
sampling
is
descriptive,
so
they
missed
out
on
that
actual
stand,
whose
parent,
who,
whose
id
got
exported
as
a
parent
span
of
something
else,
yeah
yeah.
You
can
say
that
like.
If,
if
you
have
a
sampler
which
does
not
return
consistent
sampling
decision,
then
yes,
you
can
have
broken.
A
Traces
would
be
a
very
easy
explanation,
as
opposed
to
the
current
case,
where,
like
users
simply
did
not
want
enable
http
client
and
we
produced
an
incomplete
trace,
so
so,
for
now,
the
only
work
around
we
can
tell
user
is
to
just
enable
both
of
them
if
they
want
to
see
into
entries
or
like
overwrite,
the
trace
parent
in
the
request
outgoing
requesting.
A
A
I
think
the
workaround
is
here.
They
need
to
create
adhd
instrumentation
so,
but
then
there
might
be
a
reason
why
they
did
not
want
to
enable
it.
I
haven't
read
the
actual
issue
fully,
so
I'll
need
to
see
like
what
would
be
the
reason
why
they
haven't
hold
their
reason
for
like
not
enabling
http
client
instrumentation.
Okay,
so
just
like
they
accidentally
noticed
this
or
they
have
a
strong
reason
to
not.
A
We
have
a
similar
issue.
I
think
island
opened
few
months
back
like
if
you
enable
http
the
vice
versa,
like
you've,
enabled
http
client,
but
you
do
not
enable
sp
net
core.
We
do
have
like
some
issue
with
the
default
sampler,
because
the
default
sampler
is
parent-based
and
since
the
asp.net
core
instrumentation
is
not
enabled,
you
will
have
an
activity
which
has
sampling
like
faults
and
naturally
the
http
client.
When
it
tries
to
create
an
activity,
it
will
also
be
ensemble
because
the
default
tumblr
is
apparent
based.
A
B
A
Yeah,
I
actually
wonder
like
whether
like
like,
if
we
just
ignore
the
fact
that
http
client
and
espn
core
are
two
special
libraries
who
creates
activities
on
their
own
rules.
What,
if
like
this,
was
general
like
any
other
random
foo
and
bar
libraries
in
that
case,
are
we
in
compliance
with
the
specs?
I
can
see
that
asp.net
core
and
sdp,
and
they
don't
really
follow
or
portal
spec.
They
just
do
their
own
thing,
because
they
are
part
of
the
dotnet
runtime
ripple.
A
So
we
can
say
that
okay,
it's
like
they
have
every
right
to
be
like
deviating
if
it
makes
sense
for
the
dot
net
ecosystem,
but
for
like
any
other
foo
bar
libraries.
If
we
I
mean,
if
this
is
still
an
issue
for
any
other
libraries,
then
we
need
to
see
whether
we
better.
How
can
we
fix
that?
A
Okay,
just
imagine
like
abc
like
just
a
being
one
library
b
being
another
c
being
another
and
all
of
them
creates
activity,
and
there
is
a
in
the
process
communication
between
like
say,
b
and
c.
What
would
be
the
parent
for
the
span
created
in
c,
and
then
that
would
be
like
a
good
official
way
of
confirming
okay,
we
are
not
violating
any
spec,
the
http
client
thing
being
a
very,
very
special
thing,
and
then
we
can
say:
okay.
This
is
a
one-off
thing
and
we
can
do
like
some
hacks
to
help
the
customer.
A
But
if
it's
a
general
thing,
then
we
should
try
to
solve
it
in
a
generic
way.
Not
do
any
hacks,
so
like
visual
before
investigator.
Can
you
like
add,
like
more
notes
to
the
issue
which
was
open
like
explaining?
What
is
the
reason
why
this
behavior
and
we
can
continue
to
have
that
discussion
in
the
github
issue
itself
like
to
see
if
there
is
any
better
way
to
solve
this
or
like
we
may
need
to
work
with
http
client?
Do
not
have
that
fallback,
et
cetera.
B
Yeah
I
just
like
the
the
issue
you're,
referring
to
in
case
that's
at
all
useful.
A
All
right
thing:
let's
move
to
any
other
topic
if
we
definitely
need
to
like
follow
on
this
one,
because
this
might
be
like
very
common,
like
you
can
put
like
a
warning
or
something
in
the
not
if
you're
in
sp
net
core
world
make
sure
you
enable
both
instrumentations
ways.
Trouble
awaits
you
and
linked
to
these
issues.
While
we
solve
this,
so
it
can
be
like
a
more
like
a
document
issue
for
now
until
we
have
the
perfect
solution.
A
I
also
think
like
this
might
be
like
something
which
other
libraries
face
when
they
do
the
migration
from
diagnostic
source
to
activity
source,
because
http
client
did
that
migration
and
I
was
hoping
that
that
migration
would
fix
it.
But
since
they
have
that
fallback
plan,
it
didn't
really
fix
it.
But
there
are
other
libraries
which
I
think
must
translate
and
even
as
usd
case
they
all
did
the
migration
to
activity
source.
A
I
to
the
best
of
my
knowledge.
They
don't
have
this
fallback,
so
there
is.
B
A
B
A
No,
it
would
still
do
the
propagation
like
it's
just
that
they
won't
create
the
activity
using
the
new
activity.
Api,
yeah,
okay,
okay,
got
it
so
the
only
thing
which
they
will
miss
is
they
try
to
do
activities
source
dot
start.
They
realize
that
it's
null,
so
they
won't
have
that
fallback
to
do
activity
equal
to
new
activity.
They
will
still
do
the
transparent
from
activity
dot,
current
dot
id.
That
would
at
that.
B
A
B
A
Be
activity
dot
current
and
if
we
did
create
activity
then
it
would
also
be
correctly
reflected
in
activity
or
current.
So
I.
B
A
We
have,
if
you
scroll
up
somewhere.
A
It's
not
creation
alone;
we
do
start
it,
so
it
will
become
the
current
yeah.
We
have,
I
think,
alan
you
had
that
documented
in
the
the
micro
service
example,
where
we
try
to
create
activity
and
if
we
did
not
end
up
creating
one,
we
should
still
retrieve
the
activity
dot
current
and
propagate
it,
so
that
logic
should
be
like
documented
in
the
at
least
in
the
micro
service
example,
which
is
only
example.
We
have
about
manually
propagating
context.
A
Like
if
you
like,
can
you
open
the
repo
and
look
for
the
micro
service
example?
You
should
have
like
some
comment
or
like
some
nothing
so
go
to
examples.
Microservice
utils,.
A
A
If
you
can
always
do
like
activity
dot
current
that
would
yeah
I
mean
I
was.
I
was
referring
to
this
like
the
http
client
like
here
when
they're
injecting
headers,
they
inject
it
from
whatever
this
method
yeah.
So
they
need
to
have
that
fallback
thing
similar
to
what
our
micro
service
example
is
doing.
A
I
mean,
of
course,
we
need
to
like
make
sure
like
it's
the
right
thing,
but
logically
that
seems
the
right
thing
to
do,
because
I
believe
the
micro
service
example
was
also
missing.
It
like
we
were
not
propagating
it
correctly,
but
then
we
like
later
fixed
it
like
that
71
line
71
in
this
class
was
added
like
way
later.
A
So
we
did
add
it
at
that
time,
because
someone
reported
like
an
issue
in
some
other
case,
and
then
we
realized,
even
our
own
example,
was
missing
this
thing,
so
maybe
in
the
case
of
http
client,
this
would
be
the
right
thing
I
mean,
of
course
I
need
to
like
create
a
little
bit
more,
I'm
on
my
phone
right
now
so
yeah
at
least
like
we
showed
like
just
document
this
add
links
to
the
like
the
related
issue,
which
I
posted
in
the
meeting
notes
as
well,
because
when
we
look
at
it
we
want
to
make
sure
all
of
them
are
considered
in
and
then
make
a
decision
which
does
not
break
anything
else.
A
Okay,
so
when
we
like,
there
are
a
few
pr's
I
was
on
vacations,
I
haven't
reviewed
the
dependency
injection
improvement,
so
let's
try
to
get
that
included
in
the
next
release,
along
with
the
login
collected,
still
yeah.
Okay.
Thank
you.