►
From YouTube: 2020-10-23 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
C
B
B
B
B
B
Yeah
we
used
thrifted
new
relic,
a
ton
for
most
of
our
internal,
like
all
of
our
our
kafka
messaging
was
all
use.
Thrift
as
the
binary.
C
A
B
A
B
A
Yeah
people
have
done
freaking,
amazing
things
with
my
sequel.
It's
pretty
incredible.
How
much
like
people
have
pushed
the
limits
on
that
to
scale.
B
A
B
Well,
I
should
probably
find
I
don't
even
remember
what
we
talked
about.
What
what
day
is
it?
Is
it
thursday?
It
is
that
let
me
go
look
at
the
notes
and
see
what
we
talked
about
because
you
weren't
there
were
you
trask
all
right.
Let
me
open
them
up.
B
I
can
probably
share
my
screen
and
we
can
recall
we
didn't
talk
about
much,
but
of
course
it
took
a
long
time
because
it
always
does
just
have
one
share
yeah
here
we
are
yeah,
so
I
went
through
the
my
usual,
the
usual
ga
burned
down
stuff
we
were
down
to,
I
think,
we're
maybe
even
down
to
eight
issues.
Now
it's
getting
small
yeah,
I
just
kind
of
forced
the
issue
on
your
pi
1805.
B
Oh
we're,
probably
going
to
move
we're
going
to
try
to
move
the
meeting.
No
one
is
objected,
so
we're
going
to
think
we're
going
to
move
the
meeting
to
be
9
30.
A
More
likely
to
make
it
yes.
B
Yeah
the
tc
is
gonna
is
moving
their
meeting
to
be
like
8
30
to
9
30
on
wednesdays,
so
yeah,
oh.
So
this
was
a
super
interesting
one
1830,
which
was
your
issue.
Is
that
right?
Honorable,
you
remember
all
my
number
right,
the
repackaging
yeah.
We
had
a
good
discussion
on
this
and
I
put
a
big
comment.
B
Bogdan
recalled,
I
don't
know
how
he
recalled
that
you
can't
have
two
jpms
modules
that
export
the
same
package
yep
in
a
single
application,
and
this
is
actually
a
super,
strong
technical
reason
why
we
should
break
split
up.
I.
C
B
C
B
Reading
jls
diffs,
which
was
it's
always
exciting,.
B
Certainly
not
written
to
be
consumed
by
normal
humans.
I
don't
think
anyway.
I
think
I
thought
I
think
I
found
the
exact
paragraph,
though
that
forces
it
to
be,
and
it
was
in
the
class
loading
section
which
I
thought
was
interesting
so
yeah
I
was,
I
was
trying
to
grok
the
class
loading
specification
for
jpms
and
I'm
like
no
no
brain
doesn't
want
to
process
this
right
now
so
yeah.
So
we
talked
about
that.
That
was
good.
A
A
B
B
C
A
C
A
C
A
A
C
B
Yeah
yeah,
so
the
other
rest
of
the
time
was
taken
up
talking
about
all
of
the
context,
interaction
and
the
implicit
context,
keyed
and
all
that
kind
of
stuff.
Since
logan
hadn't
ever
looked
at
it
so
kind
of
walked
him
through.
He
started
raising
all
sorts
of
objections
and
then
we're
like
it's
all
implemented
by
tracing
context
utilities
under
the
hood.
We
can
do
whatever
we
want
with
it
he's
anyway,
so
he
was
really
worried,
so
his
main
worry
just
to
relay.
B
Exposing
the
key
you
know
or
exposing
in
quotes
the
key
in
a
way
that
wouldn't
let
us
customize
the
mapping
of
putting
pulling
data
out,
and
I
don't
think
that's
true
at
all-
and
everyone
at
the
end
was
like
I
mean
everyone,
I
think,
was
very
scratching
our
heads
and
trying
to
understand
where
it
was
coming
from
and
he
finally
relented.
My
guess
is
he
had
something
in
his
head
that
he
ran
to
ran
into
a
grpc
context
that
isn't
an
actual
issue.
A
Yeah
he's
explained
this
thing
before
about
why
it's
why
he
wants
us
to
keep
the
keys
private,
which
was
the
going
from
open
census
context
to
open
telemetry
context
or
open
census
to
it
was
something
so
yeah.
I
guess
that
they
had
a
bad
experience
in
open
census
or
when
they
were
trying
to
figure
out
how
that
would
map
forwards,
but
yeah.
The
key
is
private.
As
long
as.
C
A
B
C
C
B
C
B
B
B
It's
all
good
yeah.
I
think
this
is
one
of
the
things
you
gotta
working
with
him.
His
like
he
has
all
of
this
history
and
baggage
of
problems,
and
he
tries
to
put
that
pattern
on
everything
that
he
sees
that's
different
than
what
he's
used
to
and
sometimes
just
seems
to
be
talked
down
so
anyway.
That
was
that
was
basically
it
for
the
meeting
yeah-
and
I
am
I
mean,
I'm
super
stoked-
I
think
we're
actually
getting
an
api.
That's
a
joy
to
use.
C
B
C
A
A
Were
saying
on
tuesday
makes
a
lot
of
sense
as
far
as
helping
you
know,
like
just
merging
you
know,
purging
faster,
you
know,
and
if
they
want
to
review
it
after
it's
been
merged.
You.
A
Comments
and
make
suggestions,
that's
fine,
but
I
think
that's,
a
big
part
of
the
pain
is
just
you
know,
waiting.
C
B
Yep
yeah,
I
had
a
talk
with
steve
flanders
about
it
yesterday
and
he
agreed
that
you
know
it's
a
democracy.
If
you
want
to
have
an
opinion,
you
need
to
get
in
there
and
have
an
opinion
quickly
and
not
just
slow
things
down,
and
he
was
like
yeah
I
mean
make
things
happen.
Get
it
done,
don't
worry
so
much
about
he
mean
he
could
just
concur.
That's
that's.
We
should
proceed
this
way
like
awesome.
B
B
B
East,
I
have
a
view
of
the
mountains.
I
have
a
view
of
my
neighbors
where
there's
been
two
masons,
repairing
their
chimneys
sitting
on
top
of
their
very,
very
steep
house
for
the
past
two
days
in
the
cold
and
old
wind.
B
A
Cool,
so
on
the
instrumentation
side,
we
talked
a
while
about
the
muzzle
stuff
that
paul
was
trying
to
run
muzzle
outside
of
the
repo
in
his
repo.
So
he
had
some
issues
that
couldn't
really
understand
on
the
fly
and
so
suggested
he's
gonna
make
his
repo
available
so
that
we
can
kind
of,
I
think,
mateo
sure
I
probably
be
interested
in
poking
around
on
there.
C
So
I
guess
they
were
planning
on
running
their
build
with
java
8.
Did
you
discuss
that
at
all
or
no
like
we
had
that
issue.
A
A
C
A
Yeah
these
were
some
more
like
yeah
other
muzzle
yeah
yeah
a
little
tricky
that
it's
kind
of
baked
in
to
a
lot
of
our
our
stuff
but
yeah.
Well,
I'm
glad
that
you
know
he's
interested
in
trying
to
make
that
work
outside
the
repo.
It
would
be
a
cool
thing
to
we
really
would
it
would
be
great
for
people
writing
custom
instrumentation
to
be
able
to
use
it.
A
A
Oh
yeah,
I've
never
heard
that
until
I
started
at
microsoft,
but
that's
like
a
big
term.
People
here
use
a
lot
like
the
north
star
is
like
your
guiding
light.
You're.
C
C
A
Well,
yeah,
he
said
so.
He
has
a
somebody,
he
knows
at
jfrog
and
so
the
group
ids.
I
guess,
because
we
changed
group
ids
like
when
we
and
did
the
initial
sync
to
jcenter.
A
They
like
picked
up
our
group
ids
at
that
point,
and
then
it
doesn't
pick
up
our
new
group
ids
after
that,
so
we
have
to
like
unhook
it
and
re-hook
it
up
to
get
it
to
resync,
but
I
think
nikita
wasn't
totally
sure,
but
I'm
pretty
sure
that
it's
only
about
group
ids
and
not
artifact
ids,
because
I
mean
everybody
changes,
artifact
ids
all
the
time,
but
the
group
and
the
group
ids
are,
I
think,
finalized.
Now.
I
think
we're
we're
all
set
as
just
there's
some.
A
B
A
A
It
feels
good
there's
once
we
pull
that
out,
they'll
be
like
I'm,
not
quite
sure
exactly
what
we'll
do
with
like
these
two
things,
but
not
too
worried
none,
none
of
all
the
important
things
I
think
are
good
there,
and
then
we
talked
through
the
byte.
C
Code,
yes,
that
there
was
no
way
I
was
gonna
figure
that
out.
That
was
insane.
A
A
A
And
so
that
gets
actually
loaded
by
the
jvm
and
then
when
the
transform
finishes
and
it
tries
to
load
your
transformed
span.
It
throws
a
duplicate
class
file
class
error,
but
we
could
that's
not
necessarily
required
for
the
particular
span
instrumentation
to.
B
A
That
so
that's
one
thought
this
thought
manually
calling
tree
transform
on
the
span.
A
I
don't
think
this
is
I'm
gonna
comment
that
one
out
I
think
that
was
not
a.
I
think
we
decided
that
wasn't
an
option
instrument
cloud:
oh
yeah,
so
then
the
question
is
like:
where
do
you?
How
do
you
reliably
preload
spam?
A
Because
it's
on
the
user's
class
path,
and
so
we
could
instrument
class
loader
check?
You
know?
Oh
here's!
A
new
class
loader
see
if
it
has
a
resource
name
for
that
span
and
do
the
helper
class
injection
outside
of
the
transform
then,
which
would
solve
that
problem,
which
I
kind
I
mean
it
could
be
a
nice
general
solution.
A
C
Yeah,
so
my
one
random
hack
attempt
that
I
thought
might
be
this
up
like
I
went
to
the
cluster
instrumentation
and
at
the
very
start
of
it,
I
added
for
the
load
class
instrumentation.
I
just
added
class
dot,
four
name
iota
open,
telemetry.spam,
hoping
that
that
would
trigger
it
to
get
instrumented
that
didn't
work,
and
I
didn't
explore
anything.
C
A
A
A
C
A
Yeah,
because
I
put
I
what
I
did
is
I,
when
I
put
the
class
for
name
span
and
now,
did
you
did
the
application,
dot,
io,
dot,
telemetry
dot.
A
A
Yeah
so
yeah
and
I
think
thought
this
was
a
good
sort
of
general
explanation
of
the
problem
or
the
pitfall
avoid
instrumenting
default,
static
methods
on
interfaces,
parent
classes
that
your
helper
classes
extend
or.
C
A
C
C
A
A
Oh
oh
into
the
sdk
I.
A
C
A
C
B
I
was
trying
to
figure
out
how
to
get
long
back
out
of
our
string
trace
ids
anyway,
something
got
merged.
I
emerged
lots
of
things.
B
Oh,
did
you
see
the
the
pr
that
was
put
in
by
ken
to
change
powell's
association.
A
B
C
A
C
A
B
B
B
Yeah
which
holy
moly,
I
don't
know
if
you've
looked
at
that
code,
I
went
running
away
screaming
when
I
saw
mongodb
in
the
middle
of
it
is.
B
A
So
I
just
the
reason
I'm
confused
is
why
adrian
is
not
is,
is
trying
so
hard
not
to
make
friends
with
in
the
open,
telemetry
community
repo.
B
C
B
B
B
I
mean
literally
the
day
that
acquisition
happened.
New
relics
stopped
caring
about
app
dynamics.
They're
like
yeah
they're
in
a
big
giant
corporation
they're
gonna
get
eaten
up
and
nothing
to
worry
about
there
yeah
they
can
you
can
you
can
monitor
your
cisco
routers
with
apti,
but
you
know
nobody
else
cares.
B
A
A
Honorable,
a
question
came
up
today
about
you
and
the
issue
about
jcenter
and
maven
central,
where
nikita
switched
it
to
maven.
Central
first
was
your
comment:
people
did
it
or
there
was
a
divide
of
people
who
thought
your
comment
was
that
you've
seen
people
switch
to
maven
central
first
and
not
go
back
or
or
that
you
were
surprised
that
we
reverted
this
and
which
one
sorry
I'll
find
it
stop
being
vague,
so
nikita's
pr.
A
A
B
C
A
No,
that's
the
the
snapshot,
repo
yeah.
B
A
But
that
that's
for
publishing
not
for
I
think
this.
This
rotation
here
is
for
download
just
the
downloads.
B
C
C
B
B
C
A
B
B
We
should
volunteer
to
instrument
sauna
sauna
type
code
with
open
telemetry.
C
B
A
A
C
Yeah,
actually
so
that's
a
good
so
yeah
I
may
as
well
discuss
it
here.
I
one
idea,
I
also
have
is
to
sync
up
the
dependency
management
like
at
the
very
least
not
have
libraries
dot
and
one
repo
and
depth
start
in
another
repo.
That's
always
driving
me
crazy,
but
generally
the
strategy
generally.
The
strategy
I
take
for
dependency
management
is
to
use
a
fake
bomb
project.
C
That's
used
for
internal
dependency
management
and
just
use
the
normal
dependency
specifiers
in
the
gradle
files
without
versions.
That's
been
a
pattern.
That's
worked
pretty
well
for
me
because
then
you
like
there's
already
an
idea
for
a
dependency
like
come
together,
guava
or
whatever,
and
having
yet
another
id
depth
guava
or
something
that
doesn't
feel
beneficial
compared
to
just
using
the
normal
specifier.
C
C
A
Now
do
you
do
you
put
everything
in
there
or
just
things
that
are
shared
across
more
than
one.
C
A
A
A
Sounds
good
to
me,
I
would
suggest
pick
whichever
repo
is
easiest.
C
A
C
C
B
C
I
use
zulu
also,
usually
just
at
least
back
into
the
adopt
open,
like
their
download
in
japan,
was
really
slow,
while
zulu
was
fast.
That's
the
only
reason
I
picked
zulu.
I
think
that
up
seems
to
have
improved
their
cdn
since
then,
but
last
year
it
was
taking
like
half
an
hour
to
download
jdk
from
adopt
in
japan.
B
B
C
C
B
Well,
anyway,
I'm
happy
to
I
mean
I'm
happy
to
get
that
thing
merged
it
built
just
fine
for
me
on
java,
not
build
on
java
8.
That
is
definitely
true.
We
will
be,
it
will
be.
No
turning
back
at
that
point
is
that
because
they're
they're,
the
google
java
format,
gradle
plug-in,
was
built
with
java
11
by
code,
so
yeah
exactly.
B
C
B
C
A
B
Could
probably
change
we
can?
If
we,
if
we
force
java
11,
that
means
we
can,
we
can
start
using.
We
can
get
rid
of
the
animal
sniffer
and
use
the
release.
Yeah.
C
C
B
Hey
I've
got
a
question
for
you,
so,
as
I
said,
I
was
distracted,
trying
to
figure
out
how
to
extract
how
to
turn
trace
ids
and
in
spanish
into
long
back
into
longs,
because
thrift
wants
them
as
longs
again
and
we
removed
all
those
apis
to
get
with
which
exposed
things
as
longs.
So
I'm
trying
to
figure
out
I
mean
I
could
expose
some
more
methods
on
the
trace,
id
and
spanidy
class
helper
classes
that
just
call
to
call
back
to
that
basic
base,
16
encoding
class
or
whatever
big,
ending
encoding
class.
B
B
It
doesn't
generate
padded,
but
it
parses
padded,
okay,
which
I
also
I
knew
you
couldn't
use
it
to
gently
unless
you
manually
padded,
which
is
a
little
annoying,
but
it
appears
to
parse
padded
just
fine.
So
I
wonder
if
I
could
I
mean
if
I
can
get
away
with
just
doing.
I
should
write
a
little
test
that
basically
just
runs
over
all.
B
Well,
all
possible
alongs
will
take
a
while,
but
randomly
chooses
longs
like
does
a
random
long
run
it
as
long
as
I
can
and
compare
the
output
of
those
two
and
make
sure
if
it
ever
like.
I'm
worried
about
corner
cases
like
around
like
negative
negative
numbers
and
things
like
that
and
how
that
might
fall
out.
C
B
A
B
B
C
B
C
C
A
Yeah
and
now
we're
on
java,
15
and
tomorrow,
we'll
be
on
java
17.
B
B
I
would
I
would
be
willing
to
build
with
14,
but
I
don't
know
15
I
mean
in
order
even
to
order
to
use
15
you're
supposed
to
sign
some
thing
that
says,
I'm
willing
to
use
this,
even
though
it's
got
experimental
string,
long
string
features
and
all
sorts
of
strange
stuff
in
it
like
when
I
tried
to
use
it
an
idea
it
made
me
like
check
a
box
that
says
I
understand,
I'm
using
experimental
features,
which
I
thought
was
very
strange.
So.
A
B
B
14
is
not
lts,
but
I
think
14
doesn't
have
any
experimental
features
in
it.
They've
they've
made
all
the
features
that
were
experimental
into
official
or
remove
them,
whereas
15.,
I
think
they're
kind
of
doing
this
on
it.
There's
still
three
there's
cycles
they
go
through
and
15
has
the
new
experimental
feature
of
multi-line
strings
and
I
don't
think
14
had
any
experimental
features
in
it
at
all.
Anyway.
C
I
don't
have
any
concrete
examples,
maybe
it's
good
to
try
to
find
some.
It's
almost
always
as
javadoc
renders
a
bit
better,
but
it
might
not.
There
might
not
even
be
any
that's
the
thing
that
I
tend
to
see,
though
every
java
does
improve
the
javadoc
a
little
bit.
B
A
C
A
Oh,
I
had
a
question
about
the
the
gradle
six
seven
that
will
that
download
the
java
version
that
you
run
the
build
with
or
just
the
for
running
tests.
C
C
B
It
actually
makes
me
really
happy.
I
have.
I
have
gigabit
ethernet
and
gigabit
gigabit
internet
over
fiber
when.
C
C
C
B
When
I
worked
at
nike,
you
know
this
was
15
years
ago
I
worked
in
their
business
to
business
like
I
worked
on
their
business
to
business
website,
the
back
end
of
that,
and
they
had
to
like
the
html
that
they
rendered
had
to
be
as
tiny
as
possible,
because
a
lot
of
their
like
their
people,
who
were
buying
directly
from
nike
the
business
to
business
portal,
were
like
tiny,
shoe
stores
or
convenience
stores
in
southeast
asia,
where
they
were
using
their
visa
machine
as
their
modem.
B
A
Project
yeah
for
honorable
for
running
I
mean
the
only
concern
I
had
was.
If
people
had
to
download
you
know
java
15
themselves
to
run
it,
but
if.
C
A
A
C
A
C
8,
because,
like
java
8
is
just
going
to
be
there
forever.
A
B
C
A
A
A
You
have
to
support
these
old.
You
know
old,
apps,
yeah,
so
yeah
I
like
java
8
and
it's
I.
Finally,
you
know
I
finally
had
I'm
learning
streams
and
all
that
you
know
all
that
stuff
via
intellij.
I'm
like.
B
A
C
B
Yes,
there's
a
great
excellent
article.
I
can
point
you
at
I'll
have
to
dig
it
up.
There's
a
fantastic
argument,
article
that
deconstructs
about
dynamic
and
why
it's
there
and
how
it
works.
C
B
C
Yes,
so
I
saw
that
comment
on
like
lambda
instrumentation
strategy
or
something
that
which
we
haven't
enabled
because
of
a
very
old
issue
that
seems
to
be
fixed.
I
wonder
if
we
can
instrument
those
now.
A
Yeah
so
I
mean
in
general
the
jv
the
the
class
file
transformer
the
at
the
jvm
level.
It
doesn't
pass.
C
B
C
B
C
B
B
A
Well,
this
is
closed,
but
I
thought
the
last
thing
I
had
seen
on
here
was
that
it
wasn't
like
recommended
for
production
or
something
using
this
yeah
are
using
this
strategy
in
production
yeah.
I
I'm
gonna
say
that's
a
no.
It's
2018.,
okay,.
A
Yeah
here
I'm
going
to
add
that
to
next
week
I
will
ask
phone
a
friend
again:
we've.
B
Apparently
got
some
some
of
the
best
by
code
instrumentation
experts
in
estonia
now
working
on
the
project,
so.
B
Also
true,
but
all
of
the
peter's
buddies
are
like
it's
funny.
You
know
it's.
It's
very
strange
in
all
the
splunk
marketing
around
the
plumber
acquisition
they're
talking
about
how
like
experts
in
bci
like
this,
like
they're,
throwing
around
the
bci
acronym,
I'm
like
no
one
ever
calls
it
that,
like
in
the
real
world
that
I've
ever
run
into,
but
I
don't
know
apparently
apparently
that
that's
the.
A
B
Know
I
don't
know
you
heard
that
right
we
did
the
new
relic,
not
really
the
splunk
acquired
properly.
C
B
All
right,
well,
I'm
gonna
take
off
evening
and
afternoon.
A
B
A
Over
it
yeah
yeah
yeah,
I
got
to
go
and
yeah
empty
out
the
the
the
hose
so
that
it
won't
free,
freeze
or
something.