►
From YouTube: 2021-10-15 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
C
So
we
don't
get
the
excuse
of
not
doing
any
work
because
we're
at
coupon
we
cannot
use
the
excuse
of
other
people
are
at
google.
A
Yeah,
what
do
you?
I
only
have
only
my
only
references
I
went
to
spring
one
a
couple
years
ago.
It
sounds
like
it
would
also
be
very
large.
B
A
Yeah
this
morning
I
just
talked
through
a
lot
of
what
we
talked
about
on
tuesday
about
instrumentation
api
stability
and
moving
the
semantic
attribute
extractors
to
internal
package.
B
B
A
I
thought
it
made
a
lot
of
sense
from
the
from
the
versioning
perspective
that
the
release
the
sdk
releases
don't
really
correspond
to
semantic
attribute
releases.
If
anything.
B
But
also
this
bigger,
I
mean
the
bigger
issue
of
like
if
you
like,
two
different
instrumentations
might
use
two
different
versions
of
semantic
attributes
and
if
we're,
if
the
spec
is
relying
on
schemas
to
move
between
those
versions,
it
doesn't
make
sense
to
try
to
like
there's
like
I
don't.
How
do
we
do?
How
do
we
do
that
when
their
scheme
is
involved
and
where
we
might
actually
literally
be
changing
some
of
those
constants
as
a
part
of
that
schema
evolution
anyway?
C
B
C
B
Interesting
yeah,
we
still,
I
think,
there's
still
some
thinking
that
needs
to
be
to
go
on.
I'm
gonna,
I
already
put
another
or
put
an
issue
in
the
containers,
notes
to
talk
about
them
at
least.
Is
anybody
else?
The
real
question
is:
is
anybody
else
actually
generating
that
stuff
or
is
job?
The
only
thing
that
is
so.
C
B
C
A
Maybe
yeah,
and
so
we
kind
of
got
these
two
different,
but
similar
issues,
semantic
attributes
and
then
kind
of
somewhat.
Similarly,
what
we
do.
C
C
C
C
C
B
Well,
none
of
the
semantic
conventions
are
1.0
yet
so
they
can
completely
decide
to
change.
Yeah.
C
B
C
A
This
and
that
kind
of
ties
to
one
of
the
things
ludmilla
was
trying
to
push
of
not
having
both
http
and
like
database
semantic
conventions
on
the
same
span,
but
that
doesn't
really
help
with
like
net
and
hd,
which
do
live
together.
C
C
B
B
B
C
The
performance
should
be
better
for
static
finals,
but
there's
also
the
option
just
not
like
only
accepting
we're
not
even
going
to
change
the
so
we
can't
get
rid
of
activity
like
if
we
only
accepted
strings
and
then
memoized
in
our
implementation,
then
we
don't
even
have
to
publish
anything
those,
but
it's
probably
too
late
for
anything
like
that,
and
the
performance
would
probably
be
worth
sometimes
well.
We
still
have
to
figure
out
how
to.
C
B
C
Yeah
we're,
but
if
it's
published,
then
it's
published
something
like
not
like
really
just
string
this
only
way
and
then
whether
it
gets
memoized
into
a
singleton
attribute
activating
the
implementation
detail
and
just
let
the
strings
be
the
api,
it's
the
safest,
but
I
think
it's
our
api
would
not
work
well
with
that
anyways
yeah,
yeah.
B
B
A
Yeah,
what's
honorable.
B
C
C
Yeah,
it
can
be
bdfl.
Probably
I
think,
even
if
there's
lots
of
people
with
merge
rights,
it
is
good
to
have
clear
like
who's
active
who's.
The
most
active
in
that
people
like,
for
example,
especially
like
contributes,
obviously
I'm
probably
the
only
active
person
there.
Other
people
do
sort
of
review,
but,
like
I
at
least
I
know
I'm
committed
to
following
up
on
issues
or
prs.
C
A
Yeah,
I
mean
it's
a
lot.
I
was
thinking
about
that
today,
a
little
bit
and
I
mean
especially
the
it
feels
like
the
sdk
repos
really
picked
up
activity
lately
and
so
like
I,
I
can
barely
keep
up.
Even
just
you
know,
skimming
those
pr
everything
going
on
in
the
java
repo.
B
C
Yep,
so
I
don't
know
I
mean
I
think,
the
current
steam
which
separate
like
either
we
have
a
common
group
and
then
these
implicit
primary
or
we
just
go
with
the
current
scheme.
They
both
seem
fine,
like
I
don't
see
a
big
reason
to
change
either.
I
guess
because
I
would
expect
some
of
the
other
maintainers
to
still
just
not
really
do
the
repo
anyways.
B
B
B
B
Since
constance
who's-
last
name,
I
I'm
not
gonna
attempt
to
pronounce
but
she's
one
of
the
gc
members
she's
on
that
team.
So
she
definitely.
She
definitely
is
very
involved
in
making
sure
that
we
have
the
right
observability
for
there,
so
she
would
definitely
yeah.
I
noticed.
A
So
one
seven,
I
I
told
josh,
who
was
just
curious
about
the
because
we
had
discussed
exemplars
last
week,
whether
to
include
those,
and
so
I
gave
updated
him
that
we
had
talked
last
week
and
that
you
were
comfortable
since
metrics
is
alpha
and
it's
not.
It
only
powers,
metric
exporter,
yeah.
B
A
B
We
actually
instrument
micrometer
and
use
that
for
all
of
our
metrics
and
in
this
form
distro
at
the
moment.
A
C
A
Oops,
I
was
also
a
dash
topes
because
the
there
was
a
contrib
draft
pr
floating
around.
That
was
a
micrometer
hotel
bridge,
but
it's
going
in
the
wrong
direction,
not
the
direction
that
most
of
us
know.
B
A
So
I'll
probably
try
to
press
the
button
tomorrow
afternoon
and
as
usual,
if
that,
if
I
get
delayed,
it
will
be
your
monday
slash
my
monday.
C
C
A
You
put
in
the
the
one
seven
update
yeah.
We
just
needed
somebody
to
write
release,
notes.
C
C
I
think
there's
one
mindset-
hopefully
it's
gotten
a
bit
better
but
like
instrumentary-
tries
to
handle
the
standard
cases
and
semantic
attributes,
but
it's
not
meant
to
be
shoehorned
into
like
where
some
helper,
it
doesn't
work
you're
supposed
to
just
do
it
using
the
api
directly.
That's
the
general
idea.
A
Yeah
earlier
at
this
point
now
that
we're
like
83
out
of
99
yeah.
C
A
C
B
B
And
I
don't
know
if
we
even
need
a
log
sync
provider,
and
I
I
just.
I
think
we
should
just
sit
down
and
really
rethink
how
the
whole
log
sync
api
is
currently
put
together,
because
I
don't
think
it
was.
It
was
put
together
independently
and
not
really
following
in
the
patterns
that
we
have
in
the
sdk.
A
A
C
A
A
C
C
Yeah,
I
guess
so
so
you
just
have
to
bridge
the
whole
login
sdk's
public
api
anyways.
A
But
I'm
not
sure
that
helps
that
doesn't
really
help
us
for
internal.
B
This
is
interesting
because
this
is
the
opposite
direction
of
all
the
extension
points
for
for
for
tracing
right,
because
this
is
now
that's
like
where
the
user
is
going
to
bring
the
front
end
rather
than
the
back
end.
For
like
exporters
and
stuff
I
mean.
C
C
C
C
C
So
what
about
when
the
classes
are
the
same?
Do
we
get
played
up
or
we'd
still
bridge?
Somehow
I
guess
we're
bridging.
So
it's
okay
like,
for
example,
a
resource,
would
be
exposed
to
this
api,
I'm
sure,
and
that's
up
till
now
that
was
in
our
agent
class
letter
and
had
no
nothing
was
not
involved
in
bridging
or
anything
once
that
gets
involved.
Is
it
okay.
A
It
just
means
more
surface
area
that
we
would
have
to
bridge
for
the
logging
sdk.
I
mean
oh
wait.
Yeah.
C
I
guess
if
there's
something
like
set
resource
view,
just
know
that
or
something
like
we
wouldn't
let
the
user
configure
like
we.
Don't
we,
the
user
generally
can't
configure
the
sdk,
and
so
now
we
would
probably
need
to
disable
the
configuration
apis
for
the
login
sdk
end
agent.
Is
that
what
we're
thinking.
C
A
B
A
Interface
and
if
we
can
keep
this
interface
really,
you
know
reasonably
clean
and
say
that
you
know
if
you're
building
an
appender,
you
should
only
use
yeah.
C
B
B
I
mean
honestly,
this
was
something
that
ken
asked
for
around
tracing
also
to
have
like
a
separate
package
that
was
all
of
the
apis
that
are
actually
relevant
for
configuring
and
customizing.
Your
sdk
like
to
put
the
spam
processor
api
and
the
order,
and
basically
and
the
spam
data
and
that
kind
of
stuff
into
one
package
that
would
be
separate.
B
B
Just
saying,
though,
and
I
think
he
wanted
this
for
oh
size
for
native.
A
B
B
Yeah,
I
don't
think
the
spec
needs
to
care
yeah,
although
I
think
actually
ted
has
also
requested
something
like
this
just
in
general,
to
define
very
carefully
define
all
of
the
extension
points
and
have
that
be
it
separate
like
a
separate
part
of
the
spec.
A
C
A
A
This
was
just
a
follow-up
from
an
issue:
matisse
had
that
and
and
name
is
fabrizio
from
fabrizio.
A
There
had
a
proposal
for
shortening
our
the
monstrosity
of
our
environment,
variable
names,
getting
rid
of
common,
introducing
kind
of
a
categorization
which
then
removes
the
need
for
this
one
there.
So
you
know
it's
not
short,
but
nice,
nice
progress.
A
So
we
would
do
similar
for
the
other
ones,
so
I
think
we
give
thumbs
up,
and
so
matish
is
going
to
send
something
after
one,
seven.
B
A
That's
awesome,
there
was,
you
have.
A
Effects
you
have
some
documentation.
I
saw
recently
added
about
how
to
add
jvm
args
for
different
servers.
B
A
Yeah
I
was
going
to
at
some
point
propose
that
us,
if
we
wanted
to
collaborate
on
doing
that
in
upstream,
because
I
have
my
own
horrific
version
of
that.
You
know
the
different
app
servers
and
people
still
complain
about
this
stock.
A
A
No,
no,
we
do
what
is
yeah,
we
do.
We
do
gzip
our
ingestion
data.
B
B
C
B
A
B
A
B
A
Like
it's
so
perfect,
with
both
of
our
distros
support,
the
now
have
that
extension
mechanism.
So
that's.
B
C
B
Well,
we
could
also
like
throw
in
the
light
step
one
while
we're
at
it
right
all
three
distros
at
once.
I'm
sure
it'll
it'll.
A
A
And
we've
got
a
google
distro
now.
B
C
C
C
B
B
B
C
C
A
A
A
C
A
C
B
If
you,
if
you
go,
if
the
stamping
happens
in
the
instrumentation
and
then
right
now,
I
think
I
mean
I
haven't
actually
looked
in
detail
at
jack's
pr
draft
pr,
but
I
assume
it
pulls
data
out
of
the
context
to
get
that
stuff
into
the
sdk
log
records
correlate
yeah.
Do
we?
Are
we
duplicating
effort
there
like?
Should
we?
B
A
C
C
A
A
C
It
does
seem
to
make
sense.
Instrumentation
report
is
still
sort
of
library.
Instrumentation
of
a
logging
framework
like
similar
to
how
we
instrument
any
http
library
to
create
spams
the
instrument
logger
to
create
log
records
generate
through
an
offender
framework
stuff
conceptually
definitely
makes
sense
in
instrumentation.
C
It's
fair
enough
yeah,
and
so
even
that's
just
libra
instrumentation
though
I
think,
because
especially
once
we
have
the
split
we're
going
to
have
that
same
split.
We
discussed
where
we
have
this
log
of
vendor
api,
and
so
then
that
also
follows
the
model
like
everything
in
the
instrumentation
required
only
depends
on
that
api,
not
the
sdk,
yeah,
and
so
somehow
it
all
just
seems
to
fit
together.
B
C
B
C
B
B
B
A
Well,
and
for
us
I
mean
I've
had
to
hack
in
logging
instrumentation
into
our
agent
distro.
C
C
C
C
So,
what's
special
about
java
agent,
it
could
be.
A
B
C
C
C
A
Yeah
add
the
export
like
as
a
exporter
as
a
helper.
A
But
I
will
be
very
happy
to
delete
my
my
logging
instrumentation,
auto
instrumentation
that
I
carried
over
after
I
deleted
it
from
upstream
yeah.
A
Yeah
and
there's
a
log
as
a
spin-
I
I
initially
we
initially
did
it.
I
initially
implemented
it
as
events
span
events,
but
those
are
limited
and
they
stick
in
memory,
so
spans
felt
better.
C
It
was
only
with
the
government's
committee
stuff.
I
noticed
that
yuri
had
moved
to
facebook,
I'd
always
thought
of
him
as
the
uber
guy,
and
then
he
even
mentioned
in
his
in
his
spiel
or
whatever
about
it's
interesting
that
I
get
to
work
both
on
the
spam
side
and
facebook
like
they
have
their
model.
That's
only
events.
They
don't
have
spams
on
facebook.
I
don't
know
if
you've.
C
It
was
interesting
that
instead
of
modeling
time
like
the
durations
and
stuff
in
your
app
like,
if
you
just
send
the
vents
up,
your
back
end
can
do
it
anyways
and
that
can
be
easier
like
that
can
be
more
flexible
for
different
workloads
rather
than
just
request
response
person.
That's
sort
of
the
theory
behind
it.
A
B
A
B
A
B
B
B
C
C
C
C
A
C
C
A
Is
it
we
already
shipped?
Well,
I
guess
we
can
reuse
those
spi's
right.
I
mean
that
because
the
spi
artifact
is
stable
right.