►
From YouTube: 2022-03-25 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).
A
A
B
B
Were
you
talking
about
like
in
the
future
once
after,
like
if
you
want
to
remove
it
in
the
future,.
A
Yeah
say
like
say:
we
change
periodic
metric
readers
so
that
it
doesn't
extend
this
internal
class
anymore.
Would
that
be
considered
a
change
that
is
not
binary,
compatible.
B
C
For
example,
the
java
dock-
that's
messed
up
because
it's
there,
but
without
a
link,
is
just
sort
of
a
mysterious
sub
super
class
that
shows
up
there
so
generally,
like
we
have
a
implementation
that
is
private
and
the
user
is
using
an
interface
and
then
that
implementation
can
extend
from
internal
classes.
But
we
usually
don't
have
a
public
class
in
extending
an
internal
one.
A
So
the
way
that
the
aggregation
class
handles
this
is
you
know,
aggregation
is
an
empty
interface,
but
you
can't
implement
it.
The
the
we
we
assert
that
your
aggregation
also
is
an
instance
of
aggregator
factory,
and
I
suppose
I
could
do
something
similar.
They.
C
A
Yeah
I'll
I'll
see,
if
I
can
do
something
about
that,
the
that
kind
of
got
you
there
is
that
the
so
we
I
assert
that
we
we
don't
want
outside
implementations
of
metric
reader,
yet
because
the
way
that
a
an
sdk
meter
provider
registers
its
collection,
callback
with
the
reader
is
kind
of
unspecified
and
there's
potential
for
logic
to
be
put
there
and
specified
someday.
So
you
know
we
want.
A
We
want
to
make
that
not
implementable
today
for
practical
purposes,
but
the
the
implementations
of
metric
reader
are
not
all
in
one
package
like
they're,
distributed
across
a
variety
of
packages,
and
so
that's
kind
of
the
trick
is
like
how
do
we
have
things
like
the
open
census?
Shim
be
able
to
implement
a
metric
reader.
A
Yeah
well,
just
like
you
know,
out
of
a
curiosity,
do
you
know
if
that
would
be
binary
compatible
if
we
remove
that
extends
like
if
we
extended
from
if
periodic
metric
raiders
stopped,
extending
abstract
metric
greater
in
the
future,
would
that
be
binary
compatible
or
no.
A
A
So
like
in
the
future-
let's
just
you
know
hypothetically
if
we
were
to
keep
this
in
the
future,
if
we
were
to
break
that
and
no
longer
extend
abstract
metric
reader
we'd
have
a
change
in
the
api
diff,
but
you
know
there
wouldn't
be
any
like
breaking
changes
as
far
as
binary
compatibility.
That's
that's
interesting.
C
C
A
Right:
okay,
well
yeah
I'll
I'll,
explore
solutions
to
that.
B
Yeah,
I
do
think
that
this
being
a
the
super
class
of
this
makes
it
like
almost
too
easily
discoverable-
and
you
see
oh
hey,
there's
this-
I
can
do
this
like
I,
I
pulled
it
up
in
the
in
intellij
when
I
was
looking
at
that
and
of
course
I
missed
that
it
was
internal.
C
A
B
A
Yes,
I
opened
a
draft
pr,
you
know
proposing
which
artifacts
get
marked
as
stable,
and
so
I
think
you
know
sdk
metrics
and
that
one's
obvious,
but
then
the
the
questionable
ones
are
more
questionable
or
whether
we
want
to
mark
the
the
exporters
as
stable
right
away.
I
don't,
I
don't
see
any
reason
not
to
because
the
metric
exporter
contract
is
isn't
going
to
change
at
this
point.
So.
A
There
was
a
couple
that
I
so
the
ones
that
I
wasn't
broaching
yet
or
I
you
know
I
don't
think
are
worth
you
know
stabilizing
are
the
metrics
testing
classes
that
those
just
haven't?
I
think
gotten
enough.
C
A
C
And
we
do
that
as
an
exercise
of
adding
those
classes
into
our
testing
artifact,
not
the
like
we're,
never
going
to
stabilize
the
metrics
testing.
As
far
as
I
know,
we
would
integrate
those
classes
into
the
writing
artifact.
We
have
right
now,
so
we
can
work
on
the
while
we
do
that
and
then.
Finally,
the
extension,
the
jupyter
extension
can
manage.
Metrics
and
that'll
be
really
nice.
C
A
C
A
This
is
actually
the
first
project.
I've
worked
on
with
the
cerche,
so
yeah,
I'm
probably
not
the
best
person
to
dictate
patterns
with
that.
A
A
A
Yeah
that
actually
brings
up
a
good
point,
so
the
prometheus
collector.
So
I
think
the
prometheus
specification
is
still
unstable
as
well
about
how
to
map.
A
C
C
C
A
Well,
the
pr
got
merged
to
sdk
as
mixed,
so
the
sdk
document
is
now
mixed
and
the
only
thing
that's
not
stable
is
exemplars
that
happened
on
you
know
a
couple
of
days
ago.
Only
oh.
C
A
And
so
the
the
other
change
that,
I
guess
is
kind
of
controversial
is
whether
or
not
to
enable
metrics
by
default
now
from
auto
configure,
it's
we're.
Gonna
have
to
do
it
at
some
point.
I
think
just
the
it's
just
a
question
of.
I
guess
maybe
how
to
communicate
this
change
because
it
could
catch
some
people
by
surprise.
A
What
I
imagine
happening
is
folks
using
auto
configure
taking
the
default,
which
is
the
metrics,
are
disabled
and
you
know
maybe
they're
just
exporting
traces
to
a
back
end
like
zipkin
or
jager,
and
all
of
a
sudden
they
start
to.
They
don't
have
prometheus
set
up
or
anything
like
that,
so
they
just
are.
C
A
A
Yes,
but
that
doesn't
actually
mean
so
we
still
have
a
compile-only
dependency
on
that
otlp
all
dependency,
and
so
you
still
have
to
explicitly
bring
that
in
in
order
to
for
it
to
work.
A
Yeah,
that
is
weird
now
now
that
you've
changed
the
serialization
and
we
have
the
you
know
the
okay
http
version,
perhaps
a
lot
of
the
reason
to
keep
them
keep
it.
A
C
B
Exactly
well
I
mean
we
could
make.
We
could
conditionally
like
if
they're
using
jager
or
zipkin
exporters.
We
could.
B
But
I
guess
you
know
it
is:
it
does
send
a
good
message:
kind
of
to
honor
x.
Your
earlier
point
about
micro
profile
of
that
open
telemetry
is
both
metrics
and
tracing.
C
C
B
C
C
A
Yes,
okay,
so
that
you
know
the
pr
that
that
implements
from
the
specification
also
says
that
metric
readers
should
have
the
ability
to
influence
the
default
aggregation
by
instrument
type,
and
so
you
know
you
can
imagine.
I
can
imagine
like
a
scenario
where
that
would
be
useful.
So
let's
say
so.
Aggregation
is
definable
as
part
of
views
right,
so
you
can
say.
A
A
All
of
a
sudden,
your
catch-all
rule
to
say
you
know,
change
all
histograms
to
exponential,
is
going
to
create
two
metric
streams
for
the
one
instrument
that
tries
to
add
a
description,
because
you
know
views
don't
layer
well
together,
like
if
there's
two
views
that
match
an
instrument.
You
get
two
metric
streams
and
so
a
good
alternative
to
that.
To
doing
that,
at
the
view,
level
is
to
configure
a
metric
reader
to
switch
the
default
aggregation,
and
so
you
know
I
don't
know
the
use
cases
I
I
don't
know
any
likes.
A
A
There's
two
five-way
functions:
one
is
for
temporality,
so
you
can
say
the
reader
has
the
ability
to
influence
temporality.
A
C
A
Yeah,
so
every
instrument
has
a
default
aggregation
that
applies
to
it.
So
counters
use
the
sum
aggregation
by
default.
You
know
you
could
do
something
nonsense,
though,
and
switch
them
to
use
the
histogram
aggregation,
and
then
you
could
see
the
distribution
of
like
your
your
sums
that.
A
Okay,
okay,
you
know
up
down
counters,
also
use
some
as
the
aggregation
and
here's
one.
That's
really
confusing
gauge
instruments
use
an
aggregation,
that's
called
last
value
and
then
last
value
aggregates
into
a
gauge
again
on
the
wire
format.
So
it
goes
gage
last
value
gauge,
which
is.
B
A
Think
so
histograms
will
be
common
because
you
might
want
to
change
your
bucket
boundaries
and
I
think
that
would
probably
be
the
most
popular
use
case.
A
You
could
you
know
you
could
come
up
with
some
kind
of
contrived
use
case
where,
like
let's
say,
you
have
a
back
end
that
doesn't
support
histograms,
but
you
want
to
get
like
the
count
of
your
number
of
http
requests,
and
so
maybe
you
change
http
server
duration
to
from
a
histogram
which
is
default,
aggregation
to
a
sum,
and
so
you
lose.
B
A
A
A
B
Oh
yeah,
is
he
back
on
monday
or
the
following.
A
C
B
Yeah,
so
the
javadoc
now
do
we
publish
to
or
do
does
like
javadoc
io
automatically
pick
up
the
open,
telemetry
stuff,
okay,
and
so
would
oh
and
it
probably
just
pulls
down
the
javadoc
jars
from
maven
central.
So
we.
C
A
C
A
A
C
C
Because
we'd
still
be
using
the
tool
chain
feature
so
it'll
automatically
download
if
they're
java's
yeah
same
with,
if
they're
using
java
11
right
now,
it
automatically
downloads
java
17.
So
instead
of
download
18.
C
A
Javadocs
that
get
published
would
would
have
this
all
you
know
neatly
generated
yeah,
not.
C
B
C
There
was
just
I
do
remember
like
there's
a
bullet
plane
about
automatic,
like
it
doesn't
even
do
some
tax
highlighting
really-
and
I
was
like
yeah,
because
java
features
evolve
so
fast
or
something
we
wouldn't
be
able
to
keep
the
highlighting
in
in
sync
or
something
I
think
that
was
in
the
jeff.
So
there
is,
I
mean
I
might
still
give
it
a
there's,
a
doclet
that
tries
to
implement
that
functionality.
That
would
be
usable
in
older
java.
C
C
C
C
C
B
C
C
B
C
Java
18,
but
there
is
the
big
change
that
now
it's
all
utf-8
by
default,
so
that
they
did
instead
of
the
platform
default
or
whatever
the
card,
all
the
default.
Oh,
like
so
far,
the
standard
best
practices
have
been
always
passing.
The
standard,
carsets.utf8
yeah
not
use
the
platform
default,
but
with
java
18
they
finally
switched
the
default
to
utf-8.
B
C
C
A
End
up
being
like
significantly
more
than
like
eight
bets,
they're
eight
bytes
for
your
eight
bits
per
character.
Then.
B
B
A
C
A
Yeah
dude,
just
do
japanese
characters
tend
to
contain
more.
Like
information
per
you
know,
per
character
than
a.
B
A
C
The
relative
yeah,
maybe
yeah,
two
three
like
english
word,
that
is
six
characters,
would
just
be
one
or
two
characters
in
japanese,
oh
wow,.
A
C
Contextual
and
the
annoying
thing
is
that
they
have
two
scripts,
and
so
you
tend
to
mix
the
letters
that
are
based
on
chinese
letters.
Those
tend
to
be
nouns,
you
know
the
grammar
tends
to
be
a
japanese,
specific,
simpler
script,
and
so
you
mix
these
into
your
sentence,
and
so
that
tends
to
be
the
trigger,
and
so
for
me
it's
really
for
japanese
people
since
they're
used
to
it
it's
still
sort
of
okay.
C
For
me,
it's
really
hard
to
read
japanese
for
like
ninten
old
8-bit
games,
because
they
didn't
have
enough
space
for
the
japanese
text,
so
they
would
be
using
that
phonetic,
one
for
everything,
including
the
nouns
and
then
without
the
spaces.
That's
actually,
I
think.
Actually
they
did
include
spaces
for
those,
luckily,
because
otherwise,
it'd
be
way
too
hard
video
and
then
it's
pretty
hard
to
read
when
they
use
the
phonetic
script
for
everything.
A
I
wonder
how
man
there's
so
many
like.
I
wonder
how
like
text,
processing
works
when
there's
no
obvious
delimiter
between
words,
much
harder
yeah,
so
cjk.
C
Processing,
especially
japanese,
I
guess
chinese
also,
they
don't
use
spaces.
Maybe
I'm
not
sure
I
think
so
so
cjk
tends
to
be
that's
why
every
like
elasticsearch
has
a
cjk
module
like
all
the
text.
Processing
things
have
a
cjk
extension
or
a
module,
or
something
like
that
as
well,
because
it's
way
different
and
makes
it
much
harder.
B
I'm
pretty
sure
my
son
is
going
to
college
in
tokyo,
no
wow
yeah,
that's
where
his
wants
to
go
to
temple
university
in.
C
B
C
C
C
I
mean
I've
experienced
both
it's,
I
think
the
u.s
university.
Still
you
do
have
to
do
a
bit
more
work.
I
mean
you
might
drop
out,
but
in
japan
you
can
graduate
even
without
really
doing
any
work,
because
they
just
tend
to
focus
on
graduating
people,
not
teaching.
B
C
A
My
gosh,
they
don't
want
that
negative
statistic:
yeah
bringing.
C
A
Their
yeah
exactly
applications
yeah.
C
A
They're
so
good
at
manufacturing
and
in.
C
A
I
bet
those
those
kind
of
old
institutions,
then
are
are
super
important
to
you
know
continuing
to
educate
the
population
overall.
C
It's
true
yeah,
like
the
training
of
these
bigger
companies,
like
they
say
like
if
you
don't
take
that
training,
so
companies
like
american
companies,
like
google
or
something
their
training,
wasn't
very
good.
Now.
I
think
it's
okay,
but
back
in
like
if
a
new
grad
joined
a
company
like
google,
they
wouldn't
really
be
able
to
join
a
japanese
company
because
they
missed
out
on
the
training
that
everyone
gets,
which
teaches
a
lot
of
the
fundamentals
for
japanese
business
practice
and
stuff
as
well.
A
Is
it
kind
of
like,
like
an
apprenticeship
ship
in
you,
know
united
states
trades
terms
or.
C
C
C
C
C
Everyone
joins
their
company
april
first,
like
as
a
new
grad
you
graduate,
and
then
you
join
a
company,
but
first
that's
what
it
is.
So
it
also
helps
them
set
up
the
training
program.
I
guess
because
all
companies
are
getting
their
new
beds
at
the
same
time,
yeah
there's
a
batch
and
yeah.
Some
of
it,
I
think,
is
just
fun
or
whatever,
like
you,
join
ntt
to
work
on
ai
or
something
but
they'll
still
have
you
do
training
for
a
few
months
in
a
cell
phone
shop.
C
C
C
B
C
B
B
Yeah,
assuming
he
does
this,
I
will
I'm
gonna
go
over
I'll.
C
B
I
think
it's
around
beginning
of
august,
okay,
school
start.
B
B
A
C
C
Eventually
you
just
get
used
to.
I
mean
there's
people
that
just
never
learn
japanese
and
still
live
in
japan,
and
I
don't
know
how
they
can
be
fun.
Be
happy
here
like
that,
but
like
for
me,
one
of
the
big
things
was
like
when
I'm
writing
the
train.
There's
all
this
japanese,
and
so
I
would
just
try
to
then
look
up
some
words.
B
Yeah,
I
think
that
those
the
phone
translate
stuff
tends
to
be
pretty
good
now
also
yep,.
C
B
A
C
C
C
C
B
Oh
yeah,
no,
he
wants
to.
I
think
that
they
have
to
take
japanese
there
anyway,
and
I
think
it's
about
the
school
is
a
about
50,
japanese
and
about
50
foreigners.
C
Yeah
when
I
came
like
in
the
beginning,
I
was
doing
the
standard
thing
like
hanging
out
with
all
the
other
overseas
students
as
well,
but
most
of
them
were
pretty
boring
and
I
just
preferred
the
japanese
students
in
my
lab,
so
I
just
focused
friendships
on
that.
I
think
that
helped
me
focus
on
japanese
language
as
well.
Oh
yeah,
yeah,
yeah
yeah,
because
otherwise,
most
especially
if
it's
just
a
one-year
program-
and
it's
just
the
overseas
students
talking
to
each
other
in
english
entire
time,
so
most
of
them
don't
actually
get
better
at
japanese.
C
B
Did
I
not
stamp
it?
I
didn't
know
if
anybody
else
was
I
mean
I
I
didn't
scroll
through
it.
I
don't
know
if
anybody
else
is
actually.
B
B
B
Help
me
understand
reactor
better.
Actually,
I
really
because
I
I
was
glad
you
did
these
ones,
because
the
I
think
they
look
a
lot
better
or
they
make
they
make
more
sense
to
me
in
java.
C
Yeah,
my
my
standard
rule
of
thumb
is
not
having
weird
if
statements
inside
the
assertions
like.
Instead
of
that,
I
would
rather
copy
paste
the
assertions
it
does
make
it
a
lot
easier
for
me.
So
I
tend
to
do
that
in
the
old
test
they
had
like
if
mono
cast
a
mono
like
this.
This
sort
of
logic
does
not
belong
in.