►
From YouTube: 2019-10-03 Python 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).
C
D
D
D
D
E
E
E
E
E
B
B
B
B
B
B
E
E
E
E
E
E
E
E
E
E
F
So
after
the
alpha
release,
I
start
to
test
this
twist,
contacts,
implementation
and
I
noticed
in
our
behavior
is
very
different.
So
this
is
a
simple
example:
I
have
and
I
run
this
w3c
validation
service,
which
is
also
written
by
me
and
used
previously
by
open
sensors
across
several
different
languages
and
I
noticed
we
got
a
lot
of
fears.
F
It
looks
like
imitation
is
whenever
we
receive
some
invalid
incoming
twist
contacts,
header
Lotus,
draw
arrow,
which
is
weird
so
I,
think
the
the
goal
of
this
like
API
and
SDK
implementation
is,
if,
even
if
someone
send
us
the
illegal
twist
contacts,
we
should
be
able
to
recover
from
that.
Instead
of
working
the
application
or
generating
any
arrows
yeah.
F
D
F
C
F
I
think
this
also
brings
a
question
like
we.
We
probably
should
have
some
like
p1
p2
p3
g0
attacks
after
alpha
I
expect
people
will
start
to
give
us
some
feedback,
so
will
receive
some
random
issue
in
github
and
having
that
priority
will
help
us
to
yeah
coordinate
so
sounds
useful,
yeah,
okay
and
the
second
one.
F
So
we've
been
asked
by
the
spec.
So
this
is
a
cross
language
effort.
Basically
we're
saying
any
vendor
specific,
like
exporter
logic,
should
move
away
from
open,
telemetry
ya
know.
So
country
is
not
very
clear
to
me
whether
we
will
like
under
open
telemetry.
It
all
be
like
a
separate
repo
where
people
can
contribute,
so
that
gives
people
much
better
experience
when
they
try
to
discover
what
we
have
or
we
want
all
the
exposures
to
move
to
whatever
real
hot.
As
long
as
they
don't
belong
to
open
telemetry,
yeah.
E
F
E
E
E
B
F
I
would
propose
that
we
get
an
idea
like
like
what
what
does
p0
p1
p2
mean
and
do
we
need
p3,
so
I
think
I
have
a
pretty
good
definition.
You
might
had
T
theorem
is
a
showstopper,
so
we
should
start
doing
anything.
Just
fix
that
so
in
the
emergency
and
p1
means
something.
That's
really
important
that
we
want
to
get
hit
done.
People.
E
F
B
F
F
E
F
H
E
B
F
F
E
F
F
E
D
E
F
E
A
B
E
E
F
F
F
E
G
H
E
I
F
I
Jointly,
we
we're
gonna,
like
add
logging
or
log
correlation,
we
gotta
add
it
in
the
spec
first
and
and
I,
don't
know
if
you've
already
talked
about
this
at
all.
Yeah
actually,
like
you,
have
this
on
the
agenda
for
like
like
for
q1,
but
this
there's
no,
but
yes,
this
is
like
no
plan
for
this
to
be
before
the
1.0
release,
but
it
would
have
to
be
on
the
spec
for
us,
this
is
like
no
anything
we
can
sneak
in.
F
And
besides
the
Tris
contacts
like
compliances,
you
know
what
her
is
a
major
like
how
four
zero
one
zero
one
released
that
in
your
mind,
I,
probably
need
to
tag
them
and
make
sure
we
address
them
in
alpha.
There
are
point
to
release
so
contract
from
what
I've
seen
the
like.
The
only
thing,
I
I,
think
we're
missing
is
a
the
compared
with
trace
contacts.
B
E
F
J
F
G
F
D
F
E
G
E
E
G
Based
on
the
cliff
or
to
me
specification
the
be
in
the
truth,
proto
source
files,
so
I
already
did
that
so
currently,
when
you
install,
when
you
run
it
installed
or
up
install
the
files
are
generated
automatically.
So
there
is
a
Serie
of
discussion,
the
region
going
not
only
in
this
repo
bar
in
the
other
episode.
Where
is
that,
currently
the
trip
protocol
genes
to
read
the
old
one
and
they
are
going
to
use
that
year?
Is
he
as
the
new
one?
G
So
the
current
implementation
that
we
have
enough
consensus
on
in
this
period
is
using
truth.
So
this
is
not
totally
clear
for
me
if
we
want
to
introduce
your
PC
well
now
or
if
we
want
to
delete
I
look
around
in
the
different
limitation
and
we
have
different
things.
Gold
is
using
trip.
Java
is
using
your
PC,
so
there
is
not
like
common
implementation.
G
E
G
E
G
So
yes,
I
should
have
seen
that
when
we
we
are
going
to
introduce
your
PC,
we
are
not
going
to
remove
trade.
Probably
we
will
have
to
keep
compatibility
but
the
same
time
I'm
not
sure,
but
maybe
that's
the
best
way,
because
as
far
as
I
know,
the
agent
of
Yeager
doesn't
support
your
PC
right
now.
So
the
only
thing
in
Jigar
support
injured.
This
is
just
the
collector.
So
we
singing
to
your
street
to
talk
to
the
agent.
G
By
the
way,
and
just
to
finish,
there
is
also
some
discussion
of
ow
if
this
Porter
should
send
the
traces
to
the
agent
and
to
the
collector,
both
of
them
just
one
of
them.
So
this
is
the
same.
I
look
around
on
the
different
plantations
in
open
science,
or
something
like
the
agent
is
mandatory.
The
collector
is
optional.
Ingo
is
something
like
if
the
collector
is
configure,
the
traces
are
not
sent
to
the
agent.
So,
basically,
to
summarize
also
here,
there
is
not
a
common
approach
to
the
agent
collector
reporting
of
the
traces.
G
J
To
meet
that
sounds
good,
but
used
to
be
also
in
a
better
position.
I
would
like
to
specify
these
details
in
the
in
the
readme
of
the
exporter
itself.
You
know
like
hey.
This
is
what
we
support
at
this
moment
and
whatever
you
know
so
at
these
people
don't
have
to
go
and
check
the
code
to
figure
out
what
we
are
supporting
and
we
are
what
we
are
sporting.
G
E
G
E
G
F
Multiple
lines
and
crispy
and
I
remember:
he
pointed
out
that
it'll
be
idea.
I
do
to
have
a
like
one
line,
configuration
and
also
we
talked
about
this
like
long
time
ago,
that,
like
we
don't
in
general
and
a
flask
or
pyramid
people
would
love
to
use
configuration
so
it'll
be
good
to
like
have
you
know
her
configuration
story?
The
second
thing
is
today,
like
we
require
extra
places,
shutdown
which,
in
organ
senses,
we
have
this
nice
grace
period
and
I
exit
hook,
so
we
automatically
shut
down
those
things.
I
won't
know
like.
E
K
I
mean
I
think
one
thing
that's
worth
considering
is
yeah
I,
think
that
is
desired
and
make
it
magical,
but
at
the
same
time
I
think
it's
what
the
configuration
is
actually
a
good
thing.
What
if
we
need
to
switch
the
way
that
work,
I
would
say
the
one
thing
about
you
know
the
consideration
that
could
probably
consider
it
under
dicing
the
way
it
works,
or
at
least
having
like
a
normal
API
right.
K
So
like
yeah,
a
little
I
mean
that
kind
of
make
sense
to
me,
but
also
the
way
that
you
said
so.
The
preferred
creature
is
different
than
you
said,
like
the
sound
processor
stuff,
and
also
this
doesn't
include
the
example
where
you
actually
choosing
the
propagator
either
to
be
good.
If
the
propagator
and
the
tracer
follow
the
same
sort
of
like
CI
design,
looks
very
go
by
line
yeah.
I
If
you
mentioned
having
static
config
like
like
Django's
config,
and
that
was
really
nice
and
open
census,
so
I
think
there's
like
good
reason
to
prioritize
this.
We're
gonna
work
on
you
know
say
getting
a
one
line
like
runtime
config
solution.
I
think
like
having
a
big
static.
Config
solution
would
be
great
too,
especially
since
we're
gonna
need
this
soon
for
metrics
yeah.
I
E
F
K
Very
rarely,
mr.
first
are
the
conversation
I'm
happy
I
am
I
was
thinking
we
could
work
together
on
this,
but
did
you
ever
see
to
litter
approach
in
mind?
I
was
thinking.
Maybe
I
could
just
take
the
test
cases
that
you've
already
ran
against
them
and
either
copy
them
into
open,
telemetry
or
I.
Guess
include
something
that
was
import,
that
that
package,
or
something
and
tried
to
run
those
tests.
Yeah.
F
Yeah
I
think
having
that
external
dependency
and
run
as
part
of
the
CI
is
important
because
I
know
still
some
amazing
cases
in
the
you
know:
phase
contacts
test
so,
for
example,
like
their
scenario
where
we
allowed
manipulate
the
tri-state
and
I
plan
to
add
more
test
cases
to
the
Senate
report.
Okay,.
F
I
Yeah,
if
you
look
at
the
spec,
says
we're
supposed
to
have
a
Prometheus
exporter
already,
but
you
know,
of
course,
the
matrix
exporter
interfaces,
inspect
I.
Think
we
need
to
do
this
before
GA
I.
Don't
think
we
need
to
have
this
in
the
next
two
weeks,
also
before
better,
probably
well,
I.
Think
that
depends
what
what
I
mean
by
data
everything
yeah.
It.
K
Metrics
I'm
also
curious
what
the
actual
sort
of
full
flight
story
is
like
because
I've
seen
the
examples
in
the
TR
understands,
we
have
to
actually
attach
your
exporter
to
every
metric
that
you
care
about
and
so
I
feel
like
I
use
a
Goldy
standpoint,
it's
a
little
bit
harder.
Then
you
know,
if
you
look
at
how
creamy
this
kind
of
Brooks
right
that
client
library
is
just
a
single
Tim,
that
this
registers
a
bunch
of
metrics
and
then
it's
kind
of
magically
binding.
K
I
K
B
K
B
B
I
B
F
For
the
unit
has
ten
coverage
there
so
already
a
PR.
Try
to
add
the
high
coverage.
I
think
that
approach
looks
good
in
general,
although
there
are
some
like
feedbacks
and
there
was
no
response
on
that.
So
I
wonder
like
hacks,
where
you
could
pick
this
up
and
move
forward,
sure
thing
and
so
Chris
anything
want
to
add
I.
Think
I
probably
can
help
here.
I
So
that's
great
yeah
I
mean
I.
Think
it's
okay
to
like
merge
this
in
once
coverage
is
working,
even
though
it
will
say
that
all
of
our
you
know
subsequent
PRS
will
will
fail
because
we
don't
have
full
coverage
and
then
we
can
just
go
through
and
the
people
that
wrote
the
code
that
isn't
covered
by
unit
tests.
Gonna
unit
tests.
F
D
I
E
E
E
J
I
E
J
I
I
So
the
other
two
things
that
I
added
here
to
check
that
are
getting
examples
written
for
the
features
that
we
have
already
and
your
skin
and
Riley
I've
already
done.
This
I
think
it'd
be
great.
To
have
more
of
these
as
we
decide
which
features
are
going
into
which
at
least
like
which
alpha
release
and
then.
I
Yeah
we
talked
about
the
fact
that
propagators
right
now
are
kind
of
like
a
draft
implementation
of
propagators
because
they
weren't
finished
in
the
spec
and
I
I.
Don't
I,
don't
know
what
the
plan
for
the
spec
is,
but
I
I
think
we
ought
to
have
like
a
solid
spec
for
propagators
before
the
beta.
Whatever
we're
calling
the
data.
I
K
I
B
D
F
I
E
Maybe
a
bit
relate
related
to
the
coat-collar
edge
I,
just
added
and
work-in-progress
PR
to
test
more
code
with
my
PI
I
know
that
we
don't
want
to
add
type
annotations
to
the
SDK
everywhere,
but
I
think
we
should
at
least
check
that
we
don't
like
call
any
type
functions
with
wrong
parameter
types.
Unless
the
chaos
does
that
make
sense,
yeah
exists.
This.
E
I
E
Know
if
you're
right
x
equals
span
call
parent
turn,
then
it
my
pleasure
would
be
able
to
detect
it.
Xe
suspend
sync
at
close
range.
It
should
it
should
detect
some
issues,
but
we
will
see.
F
Okay,
now
go
to
the
exam
host,
so
I
have
a
name
PR,
which
I
called
the
simple
exam
house.
So
there's
a
feedback
from
use.
Okay,
so
I
don't
know.
What's
your
feedback
here
and
I
I
would
thinking
that,
for
example,
people
would
probably
would
prefer
to
have
a
very
simple
thing
so,
like
like
don't
be
like
put
a
lot
of
like
prerequisites.
Just
give
me
the
most
simple
way
to
get
started
so
similar
like
the
hello
world
application.
You
see
so
give
me
a
last
lines
of
code
as
possible.
F
I
want
to
get
start
and
also
the
example
probably
will
cover
like
if
you
just
use
the
API.
What
would
be
the
behavior
I
think?
Currently
we
don't
have
a
place
to
explain
that
and
I
tried
some
folks
inside
Microsoft
asked
them
to
use
this
open.
Telemetry
and
I
figured
they'd
always
come
to
me
and
ask
about
this
concept,
so
I
think
in
the
readme
file
we'll
have
in
the
very
beginning.
This
is
the
most
simple
way
how
you
can
use
the
API
plus
the
ice
DK
to
send
a
hollow
word
and
later.
F
If
you
have
metrics
logs
for
follow
the
same
way
and
then
go
explain
if
you're
the
component
developer,
you
can
use
the
API
package
only,
and
this
is
how
it
should
work
without
the
ice
DK
you
got,
nothing
performance
is
super
gold,
and
the
only
thing
you
have
is
a
propagation.
So
with
that
high
I
think
having
a
like
a
package
like
what
you
see
has,
but
it's
something
different,
so
the
simple
start,
I
think
is
more
focusing
on
giving
people
the
concept
without
having
them
to
run
a
full-fledged
application.
Well
for
the
useless
application.
F
I
think
it's
more
like
an
integration
thing,
so
you
want
to
know
how,
in
real
world
those
things
should
work,
and
if
you
look
at
the
simple
application
it
trying
to
give
people
example
from
different
angle,
with
the
ice
DK
without
die,
Steve:
hey!
How
do
you
come
fake,
a
propagator?
How
do
you
use
everything
by
default,
just
having
a
hard
work?
It's
hard
to
fit
into
one
example
app.
K
I
mean
if
it's
your
point
and
you
agree
that
we'll
probably
have
a
new
decree,
multiple
different
examples
for
different
use
cases.
I
think
the
one
thing
I
would
notice
right
now.
You
know
I
see
your
point
about
that,
but
their
server
implementations
probably
two
lines
different
than
what
the
example
is
today
and
so
I,
don't
necessarily
see
a
strong
need
to
kind
of
like
bifurcate
those
I
mean,
maybe
there's
a
long
term
contention
there
and
the
other
other
thought
I
do
have.
K
Is
you
know
we
do
or
have
way,
and
their
unit
test
actually
verifies
the
behaviors
as
expected
and
I
think
you
know,
I
mean
the
example
is
great,
but
there's
nothing
in
their
talks
or
I
mean
these
examples
to
actually
validate
the
behaviors
as
expected,
and
so
so
now
I
worry
about
the
maintenance
cost
of
these
things,
because
there's
no
direct
feedback
to
someone
who's
making
changes
that
there's
change
is
actually
broken
on
some
behavior
that
we're
expecting
to
kind
of
showcase.
So.
B
K
Are
kind
of
my
my
two
main
concerns
you
know,
I
would
say
if
the
examples,
and
also
like
to
your
point,
I
think
it
would
be
great
if
it
was
one
that
was
literally
just
the
open,
telemetry
API.
So
your
examples
also
directly
not
only
import
from
the
API,
but
also
import
from
SDKs
and
extensions
as
well.
So
I,
don't
necessarily
see
the
arguments
with
the
example
posted
that
you
know
when
you
stuff
like
this
doesn't
seem
to
illustrate
that
this
is
just
me
a
pure
API
example.
This
is
an
example.
F
K
Yeah
I
mean
I
would
do
this,
I
mean
we
already
have
a
server
example.
That's
pretty
similar
to
what
you
have
the
only
differences
that
the
example
that
you
posted
includes
the
stand,
processing
which
could
easily
be
added
to
the
existing
example,
and
then
the
client
could
also
be
edited.
I
mean
I,
see
the
value
for
a
separate
client.
One
I
would
love
to
see
that
added,
but
maybe
also
itself
has
some
sort
of
testing
around
that,
although
they
should
do
that.
Yeah.
K
I
Mean
I
I
think
even
if
it
duplicates
what's
in
there
today,
like
I,
think
when
we
have
all
of
the
tests,
we
intend
to
have
it'll
make
sense
to
have
separate
simple
and
like
full-fledged
installable
examples
like
this
one,
because
it's
it
really
is
just
like
two
files,
I
think
it's.
It
serves
the
purpose
of
the
example
better
for
it
not
to
be
like
buried
in
another
package
right.
It's
like
pretty
obvious
what
you're
supposed
to
do,
if
there's
just
a
server
hi
and
a
client
that
PI.
I
Like
I
think
that
the
big
benefit
of
the
example
app
is
going
to
be
like
when
we
have
like
say
examples
that
use
the
loader
to
load
different
implementations
or
that
show
that
you
have
like
different
behavior,
depending
on
link
your
dependencies
like
right,
whether
you
can
load
the
SDK
and
so
right
now,
if
people
are
just
running
the
examples,
in
the
same
say,
virtual
and
where
they're
developing
I
mean
that
the
the
behavior
is
pretty
straightforward
right.
But
if
you
want
to
like
show
how
the
app
works
you
know,
depending
on
your
environment.
I
So
if
you're,
a
library
that
only
depends
on
the
API
or
if
you're,
a
vendor
that
like
depends
on
the
API
but
includes
their
own
exporter.
Like
then
I
think,
there's
like
a
really
good
reason
to
have
this
packaged
as
a
as
an
example
app.
But
right
now,
for
the
simple
example.
I
think
the
app
is
just
like
some
extra
mental
overhead
for
people
using
it
for
the
first
time,
I.
I
I
K
I
K
F
G
G
J
Carlos
yeah,
so
basically
I
created
a
small
PR
for
updating
start
start
span.
So
yeah
you
guys
could
take
a
look
after
the
call
or
the
next
days.
Chris
already
gave
some
feedback,
which
is
very
nice
and
but
yeah
I'm
waiting
for
for
you
guys.
The
other
thing
is
I
will
be
removing
tracer
a
great
span,
or
at
least
having
a
PR
for
that.
So
for.
E
Create
span
I'd
like
to
not
remove
it,
but
like
make
it
an
internal
and
internal
only
API,
because
I
found
it
very
useful
when
implementing
a
new
SDK
on
top
of
the
official
SDK
that
you
can
just
overwrite
create
span
in
the
tracer
and
return
your
own
span
implementation.
So
maybe
we
should
just
remove
it
from
the
API
but
keep
an
underscore
create
span
in
the
SDK.
E
F
However,
that
so
it's
like
something
far
away
so
yeah.
I
I
This
means
that
we
could
still
release
a
later
version
as
OH
dot,
one
dot
o
because,
like
a
B
and
RC
all
get
version
before
the
ones
without
the
like,
like
the
alpha
beta
RC
signifier,
but
we
have
to
figure
out
what
to
do
for
Oda
to
dot
o
like
whether
we
also
include
a
and
tell
people
that
this
is
an
alpha.
The
problem
with
doing
it
this
way
is
the
like.
I
The
version
system
and
the
milestone
stack,
calls
them
like
Oh
dot,
1
alpha
2,
alpha
3
alpha,
which
kind
of
implies
that
there
could
be
like
an
o1,
o2
and
o3
that
are
not
alphas
Ravana's
afterwards.
So,
basically,
just
that
the
release
schema
in
specs
doesn't
match
like
python
conventions,
so
I
think.
I
I
Okay,
and
so
the
last
thing
I
got,
is
we
have
this
air
handling
dock
now
in
the
spec
repo
that
says,
basically,
we
should
eat
all
errors
all
the
time.
We
are
not
really
doing
this
yet
because
we
don't
have
typed
text
everywhere,
like
one
easy
way
to
do.
This
might
be
to
like
at
some
like
the
entry
points
in
the
API.
Just
have
a
try.
I
You
know
catch
finally
or
to
like
put
a
decorator
on
some
like
API
methods
that
are
like
in
the
SDK
put
like
on
popular
like
high
level
methods
that
just
eat
all
exceptions
and
like
one
interesting
thing
about
this,
is
that
we
probably
want
to
eat
all
the
exceptions
in
production,
but
then
in
testing
we
want
to
actually
like
let
them
escape
and
get
like
meaningful
errors
and
tests.
So
I
think
this
thing
that
we
should
start
thinking
about
soon,
but
I
don't
have
any
specific
plan
for.