►
From YouTube: 2020-07-24 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).
B
A
Very
great,
it's
a
very
good
day
today
here
in
prague,
so
yeah,
you
know
like
it
feels
warmer
because
I
am
very
close
to
renek
to
the
city
center.
So
the
heat
stays
here,
yeah.
A
Yeah,
it's
very
nice.
I
have
never
been
hearing
winter.
I
always
came
to
mexico
in
winter.
So
so
you
like
hot
places.
I
guess
yes
today,
it's
like
unexpected!
Well,
I
mean
I
really
like
the
weather,
but
there's
no
ac.
A
A
Part
of
the
fun
yes
by
the
way.
Thank
you
so
much
for
doing
all
the
reviews
on
a
lot
of
stuff.
I
have
been
super
busy
with
other
stuff,
so
having
worries
but
yeah
like
checking
out,
I
will
probably
have
you
review
a
pr
of
mine
for
resources,
probably
by
the
way
we
want
to.
Basically
you
know
we
have
this
embar
resources
part,
but
we
want.
I
think
that
the
plan
is
to
also
allow
users
to
specify
that
to
system
properties.
A
Yeah
definitely
so
that's
a
good
thing
exactly
so
yeah,
so
that's
gonna
be
very,
very
interesting
to
implement
and
of
course
I
would
like
you
know,
since
we
are
only
one
week
away
from
yeah
one
week
away
from
doing
the
next
release
yeah,
I
would
like
to
have
something
done
yeah
this
week,
so
I
will
try
to
spend
cycles
doing
that.
A
Yeah
andrew
from
like
from
like,
except
he
was
supposed
to
join
he's
not
here
for
some
reason,
bogdan,
sometimes
he's
too
busy,
as
you
can
imagine,
or
yeah,.
B
A
C
A
Yes,
we
are
waiting
for
for
the
rest
of
the
contributors.
John
is
off
this
week
and
I
don't
know,
let
me
book
box
and
just
in
case
for
probably
he's
busy,
probably
he's
sleeping,
but
probably
he's
only
distracted.
So
let
me
write
him.
C
All
the
way,
through
next
wednesday,
I
did
a
sync
up
with
him
on.
Oh,
we
had
this
a
cup
right
with
triage
for
java
issues.
Earlier
this
week,.
E
E
Kind
of
slow
morning
here
yeah,
how
come
no,
I'm
saying:
where
is
everyone.
A
Yeah
yeah
yeah.
I
think
it
feels
a
little
bit
emptier
because
john
he's
on
holidays
and
bolton
has
been
late
lately
so
and
usually
we
get
some
other
contributors
randomly
like
trust
he's
not
today
with
us,
so
I
just
dropped
him
a
message
well
a
few
minutes
ago
to
box
them
because
in
the
agenda
there's
only
basic
number
of
issues
and
for
both
of
them
I
want
foreign
feedback
so
also
any
key
that
usually
well
not
always
quite
often
he
joins
this
call.
But
I
guess
he's.
A
A
B
B
I
opened
an
issue
long
time
ago
about
creating
a
setup
method
for
allowing
to
configure
the
meter
and
trace
provider
in
the
context.
Also,
christian
provides
also
a
pr
that
implements
that,
and
this
nicely
it's
typed
together
with
now
your
issue,
carlos
about
resource,
so
find
a
way
to
allow
users
to
manually,
define
their
own
resources
and
not
just
read
them
from
environmental
variables.
B
A
Decided
to
restart
an
update,
so
whatever
the
problem
yeah
I
was
I
was.
I
was
a
little
bit
worried
that
you
would
be
busy.
It
was
like,
hopefully,
he's
only
late,
so
good,
yeah,
okay,
yeah
yeah.
Sorry
giovanni
was
talking
about
resources.
I
don't
know
you
catch
part
of
it
or
not.
If
not
germany,
you
could
probably.
F
So
so
let
me
give
you
my
my
understanding
of
this.
We
in
the
specs
we
say
that
resource
should
be
set
only
once
at
the
beginning
of
the
app
and
then
it
should
never
be
changed.
F
B
F
We
do
have
we
do
have
this,
let's
call
it
problem
and
also
we
do
have
to
set
up
exporters
and
so
on
and
a
bunch
of
other
things.
Okay,
and
there
is
a
good
question:
what
should
we
do
between
the
star?
The
between
the
the
the
so
there
is
a
there,
is
a
problem
of
initialization
order.
F
Okay,
if,
if
we
allow
people
to
have
to
write
a
private
static
tracer
in
in
their
classes,
we
do
have
a
static
initialization
ordering
problem
so
service
loader
solves
that
problem
by
making
sure
that
at
least
we
give
them
the
right
instance
in
the
in
the
in
that
statement,
the
problem
is
still
there
is.
There
are
other
configs
that
we
want
to
to
install
one
option.
Even
if
we
have
a
setup
method,
I
would
be
fine
to
have
one
of
that.
If
we,
the
setup
method
is,
is
not
create.
F
F
Yes,
that's
that's
a
strong
requirement
because
some
of
the
fields,
some
of
the
things
may,
as
the
spec
says,
cannot
be
changed
during
the
lifetime
of
the
process
and,
secondly,
is
to
define
the
the
behavior
of
the
the
sdk.
If
setup
is
not
called.
F
A
F
Saying
to
create
more
than
one,
so
I
mean
global
instance,
so
so
global
has
to
be
only
one.
Correction
has
to
be
only
one.
If
it's,
if
you
pass
it
around,
you
can
pass
whatever
instance.
We
already
have
a
builder
for
for
sure.
We
have
a
builder
for
trace
provider
which
have
we
should
support
to
set
all
these
fields.
So
there
is
no
question
if
you,
if
you
construct
your
own
thing
with
the
builder,
we
should
support,
and
I
think
we
have
that.
F
Okay,
so
so
that's
that's
the
the
problem
that
we
have,
the
one
that
is
created
by
the
spi.
F
A
Let
me
double
check
a
bit
sure
take
your
time.
This
is
a
very
important
thing,
so
we
can.
We
can
spend
as
many
seconds
as
we
need.
F
F
E
Did
somebody
have
some
code
that
they
could
pull
up
to
kind
of
help
me
visualize
what
what
and
follow
along
what
we're
talking
about.
F
So,
first
of
all,
this
is
the
the
tracer
sdk
class,
the
provider.
Okay,
the
implementation
we
have.
We
have
a
builder
that
allows
you
to
create
a
new
instance,
okay,
which
allows
you
to
set
the
clock.
If
you
want
to
set
the
ids
generator
ids
to
set
a
resource
and
maybe
should
allow
you
most
likely
should
allow
you
to
to
add
the
spam
processors
as
well.
I
mean
to
be
a
fully
builder
that
you
can
construct
the
entire
thing
and
that's
that's
what
we
have,
but
there
is
a
problem
here.
F
Tyler
and
the
problem
that
I'm
mentioning
is
not
here
is
in
the
api
global
thing.
So
in
the
api
we
have
the
the
spi
loading
pod
and
mostly
because.
C
B
F
F
E
F
E
F
Yeah
it's
in
the
load
spi,
so
they
are
looking
for
for
specific
implementations
and
stuff.
One
option
that
we
can
do
is
our
sdk
to
initially
pass
an
instance
uninitialized
instance,
or
something
like
that:
okay
and
then
that
uninitialized
instance
is
able
to
to
have
a
method
initialize
which
accepts
everything
that
the
builder
accepts
and
then
becomes
initialized,
and
then
nothing
you
can
change
on
that.
Does
it
make
sense?
What
I'm
trying
to
say.
E
E
E
So
then
you
could
potentially
set
the
the
instance
that
you
initialized
in
that
class,
which
would
then
be
loaded
here
on
initialization.
It
feels
like
a
hack,
I
don't
like
it,
but
it's
not
gonna
work,
correct.
F
Because
because
again,
I
I
do
understand,
you
have
this
problem
in
mind
that
that
you
cannot
initialize,
because
you'll
use
this
class,
which
will
auto
initialize,
which
is
one
of
the
problem.
But
the
other
problem
is
if
there
is
another
class
that
you
load
before
that
code,
that
you
run
to
initialize
you'll
still
trigger
the
initialization.
E
No,
not
necessarily
so
I'm
saying
so
you
have
this
like
separate,
like,
let's
call
it
tracer
override
a
separate
class
called
tracer
override
and
you
load
that
before
you
load
this
open,
telemetry
class.
Okay,
you
set
that
you
set
a
variable
in
that
field.
E
Sure
before
this
class
gets
loaded,
and
then
you
load
this
class,
which
references
that
other
class
that
overload
class.
F
E
F
Correct
and
that
this
may
be
referenced
by
your
main
thing:
your
main
function
in
your
main
function.
You
may
have
a
you
need
something
else
that
that
that
class
has
a
private
static,
open,
telemetry
that
get
three
sir
in
it,
and
hence
nothing
will
work
so
so
by
a
side
effect
by
a
completely
different
class.
That
is
instrument.
It
is
open
telemetry,
because
you
want
to
look
to
initialize
that
before
open
telemetry.
E
E
Another
option:
we
could
provide
a
system
property
to
prevent
the
eager
initialization
of
that.
F
E
E
F
A
A
F
Should
they
why
why
should
they
be
registered
versus,
should
they
not
be
set
only
once
if
we
have
this
only
once
initialization,
why
do
you
want
the
spam
processor
to
be
updated?
That's
the
way
it's
in
specification.
E
So
you
either
pass
in
a
specific
instance
or
you
pass
in
or
you
you
do
say,
init
with
service
loader
and
don't
pass
anything
in.
F
E
F
So
the
problem
you,
if
it
was
only
one
object,
would
have
been
working
great,
but
because
of
this
problem
there
may
be
a
lot
of
get
tracer
with
different
library,
instrumentation,
libraries
and
stuff,
like
that.
So
I
have
to.
If,
if
I
let
user
give
me
a
complete
new
trace
provider,
I
have
to
replay
all
the
operations
that
happened
before
so
remember.
The
problem
with
they're
going
to
be
a
private
static,
open,
telemetry,
get
tracer
full
that
happens
before
dc
need.
F
I
have
to
replay
that
with
a
new
provider,
my
idea
was
my
idea
was
like
this
part
of
of
the
provider
having
a
map
to
all
the
tracers
and
stuff,
like
that,
it's
it
will
be
loaded
initially,
because
it's
no
no
brainer
on
that,
and
the
only
thing
that
that
I'm
you
are
allowed
to
to
swap
is
just
the
tracer
implementation,
just
the
the
the
real
tracer
implementation,
but
this
part
of
of
let's
say,
of
the
the
the
map
between
the
all
this
instrumentation
library
part
will
not
be
able
to
be
swappable.
E
So
you're
trying
to
support
having
this
init,
called
it
anytime
and
potentially
having
like
a
get
tracer
called
before
it,
and
going
and
updating
those
references.
E
Yes,
I
understand
I'm
just
I'm
not
arguing
with
you,
I'm
just
making
sure
I
understand
what
you're
suggesting,
because
that
does
impose
a
significant
increase
in
complexity.
F
And
that's,
that's!
That's
what
I'm
trying
to
achieve,
because
otherwise
there
will
be
surprises
of
initialization
ordering
and
people
may
have
spent
for
this
class
or
may
not
have
spans
for
this
class
at
one
point,
because
they,
if
they
happen
to
get
the
tracer
before
or
in
need
or
after
they
will
have
or
not
have
spent.
A
F
F
C
E
Makes
sense,
I
think
it's
a
good
direction,
I
think
requiring
service
loader
for
everything
is,
is
going
to
potentially
cause
problems
down
the
road.
So
if
we
can
figure
out
how
to
have
additional
options,
that's
going
to,
I
think,
end
up
being
a
better
in
the
long
run.
F
But
how
how
how
can
we
have
additional
options?
So
we
can
drop
completely
the
service
loader
we
can
drop
completely.
The
global
instance
is
that
reasonable
for
auto
instrumentation
is
nice.
You
can
inject
the
code,
whatever
you
want
and
wherever
you
want,
but
for
manual
instrumentation,
I
think,
would
be
a
pain
to
to
pass
the
actually.
E
It's
the
other
way
around,
with
with,
in
my
opinion,
with
automatic
instrumentation,
it's
much
easier,
having
a
static
reference.
F
Yeah,
but
what
I'm
trying
to
say
is
for
for
auto
instrumentation
you
can.
You
can
get
this
global
thing
on
your
side
and
you
can
every
time
when
you
buy
code,
manipulate
things
you
just
you
just
you
you,
essentially
you
inject
bytecode
in
every
class
to
put
a
private
static
whatever
and
you
do
whatever
you
want.
There
correct.
E
F
But
is
it
is
it
that
okay
do
we
want
to
go
that
path
and
then
every
everywhere
we
ask
people
to
move
the
pass
around
trace
provider,
metrics
provider
or
maybe
an
open
telemetry
at
all,
with
all
these
providers
and
all
the
things
object.
C
F
E
F
E
I
mean
most
likely,
someone
will
just
store
it
in
a
static
variable
somewhere
themselves.
A
Okay,
anyway,
I
really
feel
that
we
need
to
talk
with
code,
so
yeah,
let's
continue
directing
on
that.
I
just
would
like
to,
by
the
way,
raise
a
pair
of
issues,
sorry
for
tyler,
for
interrupting
that.
Please
don't
forget
about
bringing
that
back
when
we
trade
over
this
draft,
the
other
two
things.
Well,
we
couldn't
go
through
andrew's
issue
about.
We
know
we
triaging
putting
priorities,
so
the
idea
is,
of
course,
to
select
better
what
what
to
work
on,
but
the
two
other
things
I
wanted
to
to
get
booked
on
substantial.
A
Dear
hunter,
probably
we
can
discuss
in
the
issue,
but
one
thing
that
I
I
have
been
wondering
for
a
long
time
is
that
there's
there
was
a
pr
specification
that
was
merged
about
doing
auto,
auto
resources,
detection
and
it
was
merged.
It
was
approved
everything
and
in
java
we
we
do
that
for
environment
variables,
but
given
the
fact
that
system
properties
are
semantically
similar
or
equivalent,
would
it
make
sense
for
us
automatic
support
for
them
as
well
or
not
yeah
or
for
resources?
A
F
A
Probably
so
I
will
so,
I
will
probably
prepare
a
pr
myself
or
I
will
put
sergey
with
a
developer
here
like
step
to
work
on
that.
We
are
interested
in
playing
with
that.
Okay.
F
The
other
things
is
the
auto
detection,
which
is
not
system
properties
and
things
like.
How
can
we
detect
that
we
are
in
gcp
and
or
in
azure,
azure
or.
E
F
F
Service
resource
loader
or
whatever
resource
loading,
allow
people
to
write
this
class
and
then
once
during
the
initialization
we
get
sorry,
we
get
a
list
of
these
possible
loaders,
we
try
to
load
all
of
them
and
then
we
merge
all
the
all
the
results
in
the
final
resource.
So
something
like
that,
we
don't
have.
A
But
we
should
have
yeah.
We
need
to
disclose
that
underreacts.
He
proposed
an
idea
of
sp.
You
know
spi
loaded
packages,
which
I'm
not
sure
it's
the
right
idea,
but
we
need
to
prototype
a
few,
a
few
possibilities
on
the
side
which,
which
you
want
to
go
for.
Okay,.