►
From YouTube: 2021-04-02 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).
C
C
B
Okay,
I
my
pr
that
I
was
hacking
around
on.
I
underverted
it,
but
I
don't
think
we
necessarily
want
to
merge
that.
C
A
A
We
were
just
chatting
this
morning
this
morning,
speaking
about
hc
ssl
tests
for
the
instrumentation
yeah.
C
A
Yeah
tyler,
because
I
had
seen
in
the
datadog
repo,
they
have
added
ssl
tests
for
all
of
the
http
clients,
and
he
was
saying
that
was.
It
was
a
pain.
A
Well,
but
if
we
want
to
test
all
the
different
client
instrumentations.
A
With
ssl,
doesn't
it
I
mean
we
have
to
deal
with
each
client?
We
can't
how
does
that
not
be
a
pain
code
wise,
but
the
url
is
just
http
versus
https
right,
oh
right
right,
but
configuring
them
to
use
the
certs
the.
I
guess
that's
true,
that's
something!
You
know
any
time
that
I
get
yeah.
C
B
C
A
So
oh
yeah
john
had
brought
up
and
I
guess
neither
of
these
affect
you,
but
because
it's
9
am
pacific
time
meetings,
but
I
think
we're
going
to
merge
the
java
and
the
job
instrumentation
9
em
meetings
into
the
thursday
9
a.m,
meetings
for
a
while,
at
least
instead
of
four
meetings.
Three
meetings,
yeah
yeah.
B
A
A
C
C
A
So
yeah
I
mean
it
was
primarily
around
vendors.
A
So
how
would
vendors,
because
that
was
brought
up
snapshot?
Snapshots
idea
was
brought
up,
but
the
concern
there
is
the
that
they
go
away.
What
would
I
haven't?
Looked
at
the
sona
type
snapshots
to
see
how
much
in
the
exact
version.
C
C
Like
I've
had
our
aws
build
running
against
these
snapshots
for
the
past
month
or
so
now
and
it's
never
broken.
Actually,
when
you
make
a
release,
though,
then
we
I
mean,
I
don't
use
the
snapchats
for
release,
but
if
a
vendor
wanted
to,
I
don't
know
if
that
has
to
force
us
to
release
sooner
than
we
want
to.
C
Mostly
thinking
about
the
end
user,
like
when
they're
seeing
these
versions,
what
do
they
think?
Because
I'm
not
as
concerned
about
the
vendors,
I
think,
and
so
if
they
see
lots
of
releases
they're
getting
instrumentation
faster,
they're
happy,
that's
a
good
thing,
of
course,
if
they're
seeing
the
version
number
get
out
of
sync
with
sdk
and
get
confused,
that's
a
bad
thing.
So
it's
sort
of
the
trade-off
against
those.
C
A
Think
yeah,
I
would
say
our
like
our
users
direct
users
of
the
repo.
I
mean
we
can
just
point
them
them
to
snapshots.
A
A
Probably
the
primary
desire
for
that
yeah.
I.
C
A
Yeah
I
I
agree
that
that's
the
I
mean
that
that's
the
simplest
solution.
C
A
What
is
your
thought
on
consistency
versus
like
on
demand
like
yes,
there's?
Is
it
better
to
stick
to
a
schedule,
or
you
know
if
there's
no
big
changes,
nobody
wants
one
necessarily
to
push
it.
C
C
And
I
mean
there's
urgent,
like
actually
the
door
dash.
I
guess
what
is
it
the
bat
spam
processor
optimizing
person
ping
me
like
when?
Is
it
going
to
get
released
and
I
said
that
should
get
released
early
next
month?
We
do
a
monthly
release
like
okay
cool
and,
like
just
saying
that
it's
very
easy
to
tell
that
to
users
and
generally,
I
think
they
get
it
like.
If
that's
the
schedule,
that's
the
schedule
once
we
start
doing
one
offs,
then
we're
like.
Oh
can
I
do
my
one
off.
B
C
A
I
I
made
up
this
based
on
yeah.
C
A
A
Weird
yeah
so.
B
D
A
Item,
oh,
I
had
that
from
the
we
were
talking
about
stuff
john
encouraged
us
all
to
monitor
the
hotel
java
instrumentation
channel,
which
I
still.
A
A
Yeah
and
then
went
through
the
the
prs
again
and
which
I
think
is
good.
It
brings
up
some
also
discussion.
A
And
trying
to
think
like?
Can
the
library
instrumentation
call
something
to
tell
the
java
agent
like?
Can
we
have
an
api?
It
calls
to
tell
that
java
agent
to
back
off,
but
your
grpc
example
was
really
good,
where,
like
even
if
you're
using
it,
you
may
hook
it
into
one
intercept.
You
may
hook
it
into
one
grpc
service,
but
not
another.
C
A
A
I
mentioned
that
you
know
maybe
that
actually,
some
of
it's
probably
already
the
duplicate
nesting
is
probably
already
sick,
working
because
of
client
and
server
just
decline
in
servers
fans,
but
materia
correctly
brought
up
that
that
we're
not
bridging
those
yet
so
we
are
still
getting
duplicates
there.
A
But
that
that's
solvable,
but
still
then
it's.
I
guess
the.
A
C
A
I
think
just
that,
there's
not
a
consistent
yeah
like
it's
all
one-off,
but
I
I
think
that's
I
mean,
if
that's
what
it
is
and
at
work
when
we
kind
of
build
up
some
patterns
and
stuff
around
that.
B
A
A
Tyler
was
curious.
Why
we
had
switched
on
to
not
having
the
agent
back
off
as
opposed
to
the
agent
taking
precedence
so,
which
was
a
good
reason
why?
I
documented
our
decision.
A
C
A
A
That
out
so
I
did
mention
that
I
mean
we're
open
to
changing
that.
If,
for
some
reason
it
turned
out
to
as
mataj
was
thinking
about
it
or
anybody
was
thinking
about
it,
it
turned
out
to
be
easier
one
way
or
another,
but
I
think
for
now
yeah,
but
this
was
sort
of
tyler's
proposed
reference
with,
like
anything
built
into
the
library,
would
take
precedence.
D
A
D
Okay,
yeah
yeah,
I
mean
I,
I
actually
I'm
trying
to
join
more
yeah.
So
I
I
have
been.
D
Okay
yeah,
so
I
was.
I
was
saying
that
yeah
I
I
I
used
to
join
early
on.
I
I
work
at
new
relic,
I'm
a
field
instrumentation
engineer.
I
used
to
work
with
john
watson,
so
yeah
yeah,
so
yeah
so
yeah.
You
know
I've
been
following
some
of
these
conversations
but
yeah.
I
need
to
get
more
involved
before
I
contribute.
D
I
mean
I,
I
the
yeah,
the
the
order
of
presidents
actually
does.
It
sounds
interesting
to
me,
especially
you
know,
given
that
you
know
if
you
so
I
work
with
a
lot
with
customers.
So
in
a
you
know,
customer
wants
to
add
instrumentation
a
certain
way
right.
D
So
what
you
guys
are
saying
is:
if
you
has
a
framework
that
has
instrumentation
use
that
first,
then
you
have
the
the
contrib
or
the
auto
instrumentation,
and
then
you
have.
Third,
is
the
you
know,
user
preference
right,
that's
kind
of.
A
So
yeah,
so
when
you're,
what
do
you
think
the
customers
sort
of
default.
D
Yeah
the
default
is,
is
fine
with
you
know,
because
the
library
authors,
obviously
you
know
they
know
the
the
they
are
the
best
in
the
best
position
to
write
the
instrumentation.
So
even
from
a
user
perspective,
they
would
trust
that
and
then
you
have
the
auto
instrumentation
and
then
you
have,
but
the
the
I
think,
the
my
perspective
is.
D
You
know
you
should
give
the
user
at
least
a
choice,
to
add
what
they're
really
you
know,
because
you
cannot
override
what
the
user
preferences
right
so
that
in
in
that
sense
you
know
this
should
be
way
of.
I
mean
if
there
is
a
flag
or
something
to
give
the
user
a
choice
to
add.
Let's
say
they
want
to,
you
know,
add
some
instrumentation
or
attributes
or
some
something
that
they
want
to
collect.
That
is
not
in
the
framework
or
in
the
auto
contrib
projects.
D
A
Yeah
no,
no,
but
the
you
know
having
you
know
that
connection
to
the
customer
is
a
is
really
valuable
for
us.
It's
a
experience
that
you
have
with
you
know
what
customers
would
expect,
because,
ideally
out
of
the
box,
we'd
like
to
have
the
least
surprising
behavior
to
customers
and
then
for
sure
giving
them
knobs
to
yes.
Yes,.
D
A
A
A
Search
result
25
just
see
if
anything
jumps.
A
Out
we
talked
about
strict
context.
A
I
made
a
pitch
for
people
who
are
using
the
sdk
to
use
that
and
that
we
are
still
trying
to
use
it.
Tyler
was
mentioning
was
asking
why
we
cared
about
context
leaking
because
in
their
repo
they
have
a
very
specific
need
where
they
don't
send
a
trace
and
they
don't
send
any
of
the
spans
in
the
trace
until
it
the
whole
trace
has
ended.
A
A
So
I
was
just
explaining
that
you
know
we
could
leak
spans
in
then
we
could
be
parenting
to
the
wrong
span,
so
yeah,
potentially
there's
some
stuff
in
the
well.
You
had
been
looking
at
some
of
the
data
dog
stuff
around
tightening
up.
I
think
that's
why
they
were
tightening
up
around
the
the
runnable.
C
C
C
A
Did
talk
about,
I
didn't
mention
that
to
folks
the
plan
that
we
had
discussed
done
earlier
this
week,
just
as
a
heads
up,
jason,
plum,
put
in
one
more
pitch
for
multi,
multiple
repos,
not
or
not,
mono
repos,
but
nikita
brought
up
a
good
point
that
the
contrib
repo
will
not
be
a
mono
project.
A
Each.
My
instrumentation
will
kind
of
be
a
separate
independent
project,
so
we
won't
have
quite
that
same
mono,
repo
mono
project
feel
so
it
seems
like
a
pretty
good
compromise.
The
way
that.
C
You
should
I
mean
we'll
have
to
see:
I'm
actually
not
convinced,
that's
going
to
work
that
well,
because
we
do
of
course,
have
some
tooling
that
we're
going
to
need
to
apply,
and
I
think
we
want
to
release
everything
at
the
same
time
right
so
generally,
then
it'd
be
one
build,
so
we.
C
And
I
don't
want
to
have
100
gradle
wrappers
checked
in
either.
That
just
seems
weird.
So
that
being
said,
I
I've
never
tried,
but
if
there's
anything
wrong
with
it,
I
feel
we
can
fix
it
like
intellij.
We
should
be
able
to
actually
import
the
modules
we
care
about,
and
that
would.
C
A
Yeah-
and
that
makes
sense
I
mean
I
have
less
confidence-
we
can
fix
it
in
the
main
repo,
where
we
have
all
the
tooling
and
everything
intertwined,
but
in
the
contrib.
A
A
C
A
Also,
that
was
a
great
great
idea.
Yeah.
A
A
C
A
C
A
This
was
a
very
timely
fix.
You
saw
that
halo
4
ran
into
this
problem
with
his.
C
A
A
A
Yes,
it's
good
to
see
everybody
everybody's
working
on
nettie.
I,
like
this
everybody's,
got
netty
on
the
brain
right
now,
like
back
to
back
from
three
different
people,
nettie.
A
Pr's
yeah,
and
then
I
caught
is
that
where
did
it's
this
yeah?
I
called
out
this
our
external
contributor
contribution
that,
with
span,
I
think,
is
pretty
cool.
Yep
yeah.
A
C
Going
to
give
up
on
this
packaging
pr,
it
doesn't
seem
like
it's
going
to
be
feasible
because
it
did
move
too
much
into
each
sub
project,
which
means
I
don't
have
a
good
way
to
work
around
the
fact
that
a
lot
of
our
classes
are
on
the
bootstrap
class
loader
during
tests
right
now,
including
iog,
our
pc
context,
storage
override.
So
I
might
just
punt
on
that
issue
for
no.
C
A
C
A
C
C
C
A
A
D
C
A
C
C
A
C
A
C
A
C
Which
is
a
good
start
if
it
works?
Actually,
that's
a
good
point,
though,
like
if
I
shade
the
usages
get
shaded
all
I
wanted,
so
maybe
okay,
I
think
I
know
enough
to
try
to
explore
various
hacks.
C
C
Oh
and
then
instrumenter
api
did
this
come
up
at
the
meeting
nikita
didn't
approve.
I
wonder
if
he
doesn't
like
my.
A
A
On
one
of
these
things
somewhere,
I
don't
know
there's
so
many.
I
commented
just
now
when
I
was
oh.
I
liked
the
forcing
people
to
have
to
implement
those
methods.
They
returned
null
yeah
because,
presumably
also
they
should
document.
Why
not
I
mean-
or
it
would
be
nice
because
yeah
otherwise
people
will
be
like.
Oh.
A
Yep
yeah
cause,
I
say
I
say,
merge
it.
I
mean
I'm
happy
to
click
the
merge
button,
move
move
to
the
next
phase.
A
A
A
I
remember
when
I
ported
this
stuff
over
the.
I
remember
this
feature
they
have.
They
have
multiple
prefixes,
not
just
the
inst,
because
they
have
multiple
class
loaders
for
some
other.
C
C
C
C
B
C
D
B
A
C
C
Work
on
it
or
something,
and
that
was
happening
for
the
past,
like
I
think
they
upgraded
a
week
ago,
though
so
I'm
guessing
yesterday,
it's
like
a
minor
update
on
top
of
that,
but
they
switched
from
1804
to
2004
a
week
ago.
I
think-
or
maybe
it
took
some
time
to
propagate
and
just
happened
yesterday-
that's
possible
yeah.
C
B
A
How
do
you,
because
I
mean
that's
a
good
thought
of
like
and
we
have
lots
of
versions
of
dependencies
and
things
I
mean
if
we,
how
do
you
on
the
sdk
side?
Are
you
updating
dependency
versions?
We.
B
A
A
B
C
B
I've
got
a
question
for
you:
there
was
a
so
in
spring
cloud:
sleuth
hotel.
There
was
a
user.
B
I
don't
know
it
was
a
bug
or
an
issue
or
something
logged
that
we
don't
include
a
version
of
grpc
nete
shaded
as
a
dependency
like
we
have
that
as
a
the
user
has
to
provide
their
own
transport
right.
We've
been.
I
think,
we've
been
kind
of
stuck
with
that
line
for
a
long
time
that,
if
you're
going
to
use
a
grpc
thing,
you
have
to
provide
your
own
transport.
We're
not
going
to
give
one
to
you.
C
B
And
so
the
complaint
and
the
the
the
complaint
was
that
spring
cloud,
sleep,
hotel,
also
didn't,
and
so
the
users
didn't
know
it
was.
It
was
one
level
removed
about
how
to
actually
get
a
functional,
grpc
exporter.
B
One
level
removed,
that's
interesting,
so
my
recommendation
to
springfield
people
was
like.
Maybe
I
mean
this
is
kind
of
we've
kind
of
left
this
up
to
the
distributions
and
they're
kind
of
a
distribution
of
open
telemetry-
and
you
know
it's
not
an
agent,
but
it's
kind
of
a
distribution
that
uses
it
so
if
they
want
to
recommend
using
a
specific
version
of
nete
than
any
implementation
here,
we
see
net
implementation,
that's
something
to
be
done
on
their
level,
not
something
that's
done
on
the
sdk
level.
C
Definitely
something
to
be
done
on
their
level
because
they're
yeah.
I
agree
that
they're
the
end
user
experience
like
batteries
loaded
things,
so
they
should
decide
all
the
batteries
that
doesn't
mean
the
sdk
shouldn't.
Do
it
I'm
in
the
end,
it's
whether
you
explicitly
decide
your
transport
or,
if
you
have
a
problem
with
the
default,
you
exclude
using
palm
exclusions
or
something
and
tweak
it
around
which
isn't
that
hard.
C
C
C
B
C
And
so
that
that's
the
same
argument
with
salute
and
if
we
started
seeing
issues
on
our
sdk
repo
itself,
like
I
wouldn't
be
opposed
to
it.
I
don't
think
it's
like
conceptually
a
bad
thing,
because
it's
just
a
trade-off
of
which
user
you
prioritize,
like
the
one
that
has
to
exclude
or
the
one
that
has
to
opt
in.
C
B
C
C
On
grpc,
1.2
or
something,
but
they
don't
add
the
bomb
because
they
thought
they
just
had
to
add
that
to
get
their
transport
working,
but
their
version
doesn't
align
with
what
we
used.
Then
it
would
crash
and
so
that,
after
I
said
this,
I'm
actually
leaning
a
bit
more
towards
defining
adding
the
transport
by
default,
because
it's
pretty
easy
to
run
into
strange
behavior
if
they
accidentally
mess
up
the
version
alignment.
B
B
C
C
B
C
C
B
I
think
they
they
are
trying
to
use
the
the
open
tonight
defaults.
If.
B
The
open
telemetry
as
your
your
smooth
back
end,
I
see,
which
I
think
makes
sense.
I
mean
they're
like
if
you're
going
to
use
open
telemetry
you
want
to
use
all
the
telemetry.
If
you
want
to
use.
If
you
want
to
use
zip,
can
you
probably
would
just
want
to
use
regular
old?
You
know
very
closely,
so
I
see
braves
loose.
C
B
And
I
think
I
am
absolutely
positive-
it
can
be
done
better.
What
is
there
is
seems
to
be
working
in
all
the
cases
that
I
have
done.
I
mean
I
put
in
a
bunch
of
bug
fixes
to
make
sure
that
the
things
actually
got
the
contacts
got
propagated
properly,
but
it
wasn't
using
the
it
wasn't
using
the
like,
the
actual
storage
that
you
built
like
the
storage
bridge,
it's
using
it's
still
kind
of
hacked
together,
and
I
don't
think
it's
good,
but
it's
working,
so
I
would
want
to
mess
with
it.
Okay,.
B
D
A
Yeah,
I
I
do
think
you
know
bundling,
I
would
just
bundle
it
all
together.