►
From YouTube: 2019-10-04 Java SIG
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).
B
D
A
C
A
A
C
But
as
soon
as
we
change
the
like
I,
don't
know
how
you
would
do
the
export
is
one
by
one,
because
you
would
be
breaking
the
other
exporters
as
soon
as
the
processors
are
sending
span
data.
Oh
I,
see
yes,
yeah.
That's
the
real
problem
is
that
the
API
this
the
export
API
soon
as
you
change
that
you
break
no.
A
C
C
So
I
think
it
feels
like
to
get
the
broadest
brought
us
approach
across
the
community
having
a
programmatic
approach
for
as
the
primary
one
within
some
sort
of
way
to
make
it
super
easy
to
choose
what
your
sort
or
automatically
figure
out,
which
option
of
configuration
is
being
used,
would
be
great,
but
I
haven't
thought
of
a
good
way
to
do
that.
Yet
so
so.
A
I
think
it
would
be
good
by
the
way
if
we
can
have
a
sample
application
that
we
can
play
with
yeah
I
agree.
We
have
a
good
place
to
put
sample
applications.
That's
one
question
that
I
don't
think
we
can
create
a
directory
called
the
examples
or
something
like
that,
and
we
can
have
two
all
of
these
things
there.
So.
D
D
C
A
Yeah
the
other
thing
that
I'm
not
clear,
so
there
are
a
lot
of
people
mentioning
that
they
want
to
have
multiple
tracer
in
the
application
that
that's
kind
of
tricky.
In
my
opinion,
because
you,
you,
then
have
to
pass
the
tracer
somehow
all
the
time
and
I
don't
know
if
that's
reasonable
or
not
I
mean
in
Java
applications.
I
know
that
there
are
multiple
class
loaders.
You
most
likely
have
multiple
tracers,
because
you
have
multiple
class
loaders,
but
multiple
SDKs
are
multiple
tracers.
Multiple
multiple
I
mean
multiple
instances
different
instances.
A
A
D
A
A
D
D
A
C
A
C
A
Okay,
because
the
problem
that
I
have,
if
you
have
two
different
factories
and
then
one
span
for
one
request
that
passes
these
two
factories,
how
do
we
share
spans
and
stuff
like
that?
It
will
be
very
hard
to
achieve
correctness.
Yeah
I
agree
with
you,
so
so
the
main
idea
is
probably.
We
should
go
with
the
requirement
that
we
have
only
one
factory
and
then,
but
unless,
unless
somebody
convinced
us
that
we
we
do
have
use
case
for
for
that
to
not
be
the
case.
I
agree.
A
There
are
some
issues
where
we
marked
as
need
help.
If
anyone
wants
to
do
work,
I
mean
John.
You
know
you
do
a
lot
of
things,
but
if
anyone
else
wants
to
pick
up
some
work,
look
at
the
need,
help
issues
that
we
marked
and
just
say
that
you
want
to
take
that
and
I'm
more
than
happy
to
to
get
help
from
that.
C
B
Hi,
my
name
is
Stefan
I,
just
a
question.
One
thing:
I
was
thinking
about
we're
using
open,
tracing
and
started
looking
into
bit
open
telemetry,
and
one
thing
I
was
thinking
of
that
might
be
useful
feature
is
to
have
a
we
had
the
global
tags
that
you
had
to
do,
tracer
that
ends
up
on
every
span
and
would
be
interesting
to
support
having
a
supplier
function
that
you
can
register
there.
B
A
B
B
B
A
Yes,
but,
but
do
you
think,
do
you
think?
Do
you
think
this
is
part
of
the
API
or
part
of
the
SDK
that
implements
the
API
the
question?
What
I'm
asking
is
who
which
person
or
what
is
the
role
of
the
developer?
Who
is
gonna
decide
on
these
things?
So
we
try
to
say
that
the
API
is
is
gonna,
be
used
by
by
the,
for
example,
third-party
library
developer
or
a
application.
A
final
application
developer
correct.
A
B
B
A
B
A
D
The
EPA
is
API
Society,
it
makes
sense,
yeah
ACK
I'm,
not
that
sure
you
know
in
the
end
it's
and
implementation
and
I
think
we
should
reduce
as
many
dependency
as
possible
what
you
know,
especially
the
tricky
ones,
but
otherwise
I
don't
see
any
like
big
problem
in
this
case,
depending
on
a
lot
of
the
stuff.
Yes,.
A
A
D
A
Yeah,
so
okay
I
will
I
will
push
hard
on
the
by
the
way
I
talked
with
the
GRDC
folks.
They
are
open
to
do
that.
Change
that
we
discussed
to
move
into
a
different
package,
but
I
I
do
not
see
any
reason
if
that
gets
moved
to
any
other
package.
I
do
not
see
any
reason
to
not
use
that,
but
from
from
what
I
understand
it's
like
yeah,
maybe
maybe
that's
the
answer
correct.
Maybe
you
are
correct
that
the
answer
is.
He
believes
that
we
bring
the
whole
G
RPC
dependency,
yeah.
D
A
A
D
There
there's
one,
but
it's
not
related
to
the
Java,
specific
implementation
or
API,
and
it's
about
something
we
talked
about
last
week
about
adding
multiple
attributes
to
span
in
a
single
call.
So
I
open
an
issue
on
the
specification.
I
think
I'm
going
to
be
having
just
like
an
optional
operation
for
doing
these,
and
then
Sergei
mentioned
that
we
could
probably
achieve
that
using
label
set.
A
Yes,
I
need
I,
think
I
think
we
should
clarify
on
the
label,
set
I,
think
what
you
should
should
do.
If
you
ask
me
what
you
should
do,
you
should
just
open
a
PR
on
under
spec
and
just
say
on
the
set
attribute
section
in
the
API.
You
should
say
that
in
Clemente,
API
should
allow
people
to
set
multiple
attributes
in
one
call,
for
the
reason
that
you
mentioned
they're
like
avoid
locking
and
stuff
like
that.
So
just
do
that
and
ignore.
D
A
And
then
I
will
try
to
address
the
label
set
problem,
because
because
the
idea
of
label
said
that
George
has
is
way
different,
because
what
he
wants
is
actually
to
to
serialize
them
in
a
way
that
then
those
can
be
shipped
directly
to
the
wire.
Compare
with
what
what
the
intent
for
this
one
is.
Okay,
make
sense:
okay,.
C
A
What
I
will
do
is
the
following:
I
will
create
a
milestone.
I
will
close
all
the
milestone
that
we
have,
and
we
are
not
related.
I
will
create
a
milestone.
That
match
is
the
14th
and
then
the
next
month
release
and
put
all
the
issues
under
those
milestones.
As
you
make
sense,
perfect,
yes,
missions
and
then
we
we
can
have
a
easy
way
to
see
which
one
are
in,
which
release
perfect.
A
A
C
A
Fine
but
I
anyway,
my
my
I
hope
I
hope,
I'm,
not
I'm,
not
the
bad
cop
here,
but
for
me
for
me,
as
I
said
as
a
developer
I
like
to
to
know
and
to
predict
some
of
these,
these
performance
improvements
or
not
and
then
having
a
way
to
measure
them.
I
think
Jim.
Jm
h
does
a
good
job
of
allowing
people
to
benchmark
things.
Yep.
C
A
Finally,
having
I
mean
all
my
thing
was
like
I'm
I'm
here
with
a
change
I'm,
just
suggesting
that
hey
these
need
some
profiling,
give
me
some
more
thoughts,
yep,
yep
I,
don't
think
it
didn't.
That
needs
to
be
done
before
alpha,
but
yes,
I,
as
I
said,
I
did
not
specify
to
doing
it.
I
just
said
that
hey,
we
need
to
put
some
more
thoughts
here.
Yeah
I
agree,
it's
fine!
It's
good!
Ok,.
A
D
I
did
some
changes
to
that
which
I
was
hoping
to
try
out
to
push
after
the
call.
Basically,
I
am
now
specifying
also
just
probably
a
good
question
for
you
now
yeah
I
guess
we
cannot
remove
the
time
stamp
creation
from
my
list,
because
I
guess
you
need
that
for
something
else.
So
I
am
specifying
now,
instead
of
get
receiving
and
using
a
time
stamp,
I
am
using
two
components,
which
is
just
the
seconds
and
the
nanos.
So
people
are
forced
to
you
know
to
provide
those
two
components.
A
Let's
not
do
that.
There
is
another
question
that
I
didn't
put
it
anywhere,
should
we
steal
I
think
currently,
because
we
have
to
be
Java
7
compatible.
We
don't
use
instant
so
and
even
with
instant
I
I.
Have
this
question
in
my
mind:
should
we
just
move
to
just
a
simple
long,
nano
time
nanoseconds
or
in
nanoseconds
from
a
pork
yeah.
D
A
By
the
way,
one
thing
that
I
observed
and
I
think
you
had
a
pod
and
that's
why
I
pushed
back
on
on
pushing
the
implementation
right
now
is.
There
is
a
difference
between
system
nano
time
and
none
of
time
since
support
system
nano
time
is
a
monotonic
time.
That
is
not.
It
does
not
represent
the
eport.
A
A
C
A
That's
exactly
my
problem:
I
I,
so
I
have
two
problems
with
this.
One
is
yes,
I
want
to
make
sure
that
people
do
not
confuse
system
nano
time
with
this,
because
there
are
two
different
things
and
second
thing:
I
want
to
discourage
people
to
set
a
time
when
they
don't
need
a
time
like
I
want
to
make
sure
that
the
set
a
time
only
when
Sun,
when
they
have
a
previous
time,
stem
not
for
the
current
time,
I.
A
Happy
interesting,
the
reason.
The
reason
is
we
we
want
to
do
some
optimizations
in
the
backend
on
the
on
the
implementation,
for
example,
to
calculate
all
these
using
nano
time.
As
a
from
from
a
previous
moment
when
we
read
their
pork,
so
so
the
whole
trick
with
the
timestamp
converter.
Is
we
read
the
import
once
in
the
system
nano
time
in
the
same
moment
and
then
every
event
we
calculate
as
a
delta.
A
Compare
with
that
moment
that
way
we
guarantee
ordering
of
the
events,
so
we
guarantee
that
a
star
time
of
a
span
if
we
have
a
happens
before
relationship
between
two
different
moments
will
reflect
that
into
the
trace,
but
but
if
we,
if
they
provide
us,
always
the
epoch,
there
will
be.
There
will
be
no
guarantees
that
we
can
offer
like
this
sure,
but.
A
C
A
A
Do
you
know,
do
you
know
if
there
is
a
need
to
have
the
kind
computing
from
somewhere
else
or
is
just
the
need
that
the
moment
when
the
start
should
have
been
done?
It
just
happens
to
be
earlier
and
we
can
call
clock
dot
now
then
I
think
I
think
is
the
second
case.
Can
you
confirm,
with
the
I
think
Brandon
I
think.
A
Confirm
and
maybe
clarify
the
specs
that
we
we
can
have
this
model
with
the
clock
and
now
and
that
will
satisfy
all
the
requirements
because
because
that
will
be
better
John,
yeah
I
agree
with
you,
but
I
don't
know
if
the
need
is
to
have
a
way
to
support
a
way
to
specify
the
time
from
external
source
or
from
from
the
current
process
but
happen
before.
Are
you
thinking?
Are
you
thinking
like
an
incoming
span?
No
I
think
I'm
thinking.
A
If
there
is
a
library
that
lets
say
if
there
is
a
third-party
library
and
you
only
integrate
with
them
and
they
and
then
let's
say
for
every
request,
they
give
you
some
metadata
of
when
I
started
the
request
and
they
give
you
already
a
timestamp,
and
you
cannot
do
a
clock
now.
So
there
is
no
auction
called
club.
Now.
A
Yeah,
that's,
that's!
Fine!
That's
fine!
Stefan
perfectly
agree
with
you,
but
my
my
question
is
if,
if
the
use
case
that
we
try
to
cover
with
this
start
time
is
the
fact
that,
let's
assume
some
third
party
library,
let's
call
it
the
foo-
gives-
gives
me
a
start
time.
That
is
not
complete.
With
my
clock.