►
From YouTube: 2020-11-12 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
D
A
A
A
A
A
We're
working
on
currently
so
I
think
good
progress
there,
nikita's
keeping
us
keeping
us
honest
there,
making
sure
we're
driving
those
down.
Thank
you.
Mikita.
A
Let's
talk
about
start
with
spi's
because
I
know
there's
a
there's
been
a
bunch
of
discussion
on
issues
and
pr's
this
week
about
them,
and
I
saw
just
barely
saw
nikita
put
in
something
also.
E
So,
for
instance,
if
you
want
to
exclude
some
classes
in
a
different
way,
how
auto
does
it
or
let's
say
you
want
to
have
a
different?
I
think
it's
called
listener.
A
What's
a
buddy
listener
so,
for
example,
for
logging,
some
things.
E
To
bite
body
agent
builder,
because
if
there
is
something
needs
immediate
fix,
we
cannot
wait,
you
know
for
hotel
to
submit
the
fix
and
get
it
merged
and
wait
for
our
next
release
version.
A
So
you're
thinking
like
bite,
buddy,
adding
your
own
like
if
you
need
access
to
the
low
level
bite
buddy
or
for
general
instrumentation.
That
would
go
through
the
instrumentation
api.
E
A
So
pavel
I
heard
on
the
use
the
use
case
of
excluding
classes.
The
only
thing
I
mean
about
that
is.
That
seems
like
a
something
that
we
are
trying
to
support.
We
have
a
different
hook
for
it
that
other
hook
has
just
broken.
E
And
it's
again
like
it
will
be
in
our
full
control
and
for
responsibility
to
to
you
know,
provide
the
right
configuration
for
by
so
if
we
break
it,
it's
it's
our
mistake.
A
Yeah,
I
I
don't
have
any
issue,
I
mean
it
seems
like
a
reasonable
thing
to
expose
it's
it's
nice
to
have
a
concrete
use
case
that
it's
really
needed
for
before
we
add
it
is
my
only
thought.
E
So
yeah,
for
example,
to
exclude
those
classes
or
again
like
there
is
functionality
which
is
broken,
but
it
demonstrates
like
what
kind
of
issues
we
run
into
and
we
couldn't
proceed
because
something
was
broken
and
we
couldn't
have.
We
didn't
have
a
way
to
fix
it
ourselves
before,
although
it
does
and
the
release
of.
E
You
know
it
takes
time
to
to
get
something
in
sometimes
you
know,
maintainers
may
have
some
objections
and
again
it
will
take
considerably
longer.
A
A
I
don't
know
what
do
you
think
of
I
mean
I
think
we're
open
to
adding
it.
With
I
mean,
could
we
fix
the
exclude
classes
for
now
since
that's
broken,
and
then
wait
for
you
know
for
a
use
case
to
emerge
that
we
don't
have
a
hook
for
already
and
then
add
that.
E
A
E
D
Thinking
that
well
e,
if
we
so
I
can't
I
kinda,
understand
paul's
point
of
view,
because
if,
if
you
are
developing
an
java,
auto
instrumentation
agent
on
top
of
open
telemetry.
D
D
Especially
right
now
hearing
what
what
paul
has
to
say
about
that
when
I.
But
when
I
take
the
look
at
the
next
pull
request
about
agent,
installer
customized
wrote
what
was
it
that
that
that
little
bit
like.
A
A
On
this
one
yeah,
so
how
about
would
it
if
we
named
this
like
bite,
buddy
customizer,
to
make
it
like,
because
I
think
one
of
nikita's
thoughts
and
one
of
the
issues
was
making
it
like
really
clear
what
these
spis
are
for.
A
Obviously,
just
by
buddy
customizer,
it's
basically
exposing.
If
we
had
other
potential
hooks
that
we
needed
to
expose
for
bite
buddy,
we
could
add
them
also.
D
D
A
So,
let's
go
to
this
one:
you
want
to
give
us
a
same
kind
of
overview,
pavel.
E
E
It's
basically
an
api
where
you
can
submit
request,
headers
or
bodies
do
some
magic
on
it
and
then
block
requests
it's
for
a
security
product,
something
like
android
proxy,
but
in
a
process,
and
so
I
want
to
put
the
implementation
or
which
is,
you
know,
comparably
to
the
open,
telemetry
sdk
into
agent
class
loader
and
have
the
api
inside
bootstrap.
A
E
E
E
E
E
Maybe
we
could
have
like
component
installer
or
rename
the
tracer
installer
to
component
installer,
the
tracer
installer.
The
component
installer
would
install
the
tracer,
but
also
another.
You
know,
implementations
like
this
that
are
located
inside.
E
D
I
and
you
want
to
load,
say
it
again:
you
want
to
load
that
in
agent
class
loader,
because.
A
I
think
kavos,
what
you're
saying
is
you
basically
want
to
do?
No,
not
that
oh
yeah,
this
kind
of
thing
right.
E
A
So
nikita
in
general,
for
these
hooks,
do
you
have
thoughts
to
have
thoughts
on
going
more
to
like
more
general,
like
less
but
more
generalized,
hooks
or
more,
but
more
specific
hooks.
D
My
first,
my
first
thought
always
during
this
pull
request
during
this
discussion
is:
can
we
make
it
simpler
to
make
all
these
inventor?
Well,
you
have
vendor
distribution,
you
all!
You
have
your
like
three
main
class,
you
you!
You
have
your
well
your
entry
point
into
all
this
process.
Can
we
make
it
somehow
easier
to
make
it
totally
like
separate
you?
You
have
the
agent.
D
Can
you
do
it
somehow
in
there,
and
the
problem
I
see
is
that
well,
we
like,
we
still
have
a
lot
of
a
lot
of
like
functionality
that
you,
you
cannot
reuse
in
your
agent.
E
Maybe
if
you
share
this
trust,
I
will
send
in
the
message
there's
this
the
diagram.
Would
I
draw
it's
actually
very
simple:
the
bootstrap
sequences,
it's
basically
just
these
scriptings
yeah,
yeah,
okay,.
D
E
What
I
want
to
do
is
to
what
I
would
do,
maybe
is
call
the
tracer
installer
before
the
agent
installer
and
from
the
agent
installer,
remove
all
the
open,
telemetry
bits.
What
we
see
there.
What
we
just
saw
right
so
tracer
installer
is
called
first
then
the
agent
installer
and
tracer
installer.
I
would
rename
it
as
a
something
component
installer
and
have
an
api
there
that
loads
anything
that
loads.
Something
and
you
are
in
the
agent
class
loader
and
you
can
do
whatever
you
want.
So
let's
say
you
want
to
install
open,
telemetry
exporter.
E
E
The
only
the
only
thing
what
you
have
to
be
careful
about
is,
for
example,
the
exporters
setting
the
auto
exporters
environment
variable
the
next.
The
implementation
of
this
sdi
would
have
to
check
whether
should
install
something
or
shouldn't
instead
of
we
do
it.
Now,
in
the
agent
right,
we
take
that
responsibility
from
the
spi
debate.
A
So
one
one
potential
issue
with
flipping
the
order
here
is
that
we
want
to
install
the
bite
buddy
the
instrumentation
as
early
as
possible,
because
so
that
our,
if
we
do
more
stuff
before
we
install
the
bite
buddy,
that
increases
the
number
of
classes
that
have
loaded
and
increases
the
number
of
classes
that
we
have
to
retransform
and
re-transforming.
A
Yeah,
but
I
I
mean
I
like
that
idea
of,
like
maybe
for
the
component
installer
you
have
a
before
and
and
after
I
don't
know
something
that,
because
that
that
comes
back
to
this
white.
A
Body,
but
we
do
we
run
this
little
snippet
here
nikita
before
bite
buddy,
for
this
painful
reason,
yeah,
that's
it.
E
A
E
A
Yeah
I
mean
so
we
we
don't
necessarily
need
to
implement.
You
know
our
tracer
installer
using
that,
although
that's
kind
of
cool
and
could
help
prove
that
out,
keep
that
in.
But
I
do
kind
of
like
that
idea
of
component
installer
and
having
a
before
and
after.
E
A
Oh
right,
okay,
I
mean
we
could
pull
out.
You
know
something
that
a
rapper
that
then
delegates
to
those.
B
D
Well,
my
general
answer
is
yes,
I'm
I'm
thinking
about
like
more
details,
I
think
and
and
I'm
trying
to
to
acce
accessing
by
body
that
simple
and
I
can
well.
I
can
understand
that
easily
enough
yeah,
you
can
do
you
you,
you
can't
want
to
do
anything
with
bite
body,
because
that's
your
vendor
agent
yeah,
so
you
should
have
access
to
byte
by
day.
I
can.
I
can
understand
that
here
I'm
still
a
little
bit
confused
about
the
specific
use
case
and
but
I'm
trying
to.
D
And
if
we,
if
we
continue
adding
new
like
hooks
here
it
will,
it
will
become
even
more
complicated
and
at
some
point
in
time
we
have
to
start
thinking
about.
D
C
D
Yeah
absolutely
well,
every
code
that
currently
exists
was
done
for
reason
and
taking
the
into
account
trade-offs
that
were
effective
and
at
the
time
that
code
was
written.
Absolutely
I'm
trying
to
understand
is
our
current
situation
and
our
current
needs
justify
some
hard
looks
into
that
code
and
restructuring
that.
A
D
E
So
I
think
this
code
should
be
should
be
maybe
extensible
to
some
degree,
but
also
should
be
very
optimized,
so
I
wouldn't
be
creating
something
super.
E
A
A
D
That
by
the
way,
so
it's
still
that's
that's
a
good
question.
A
It's
a
lot
of
stuff
to
think
about,
and
maybe
paul
do
you
want
to
give
that
sort
of
component
installer
a
try
and
then
the
other
thing
to
nikita's
point
with
nikita.
Just
added
the
examples.
If
you
look
in
the
repo
now
for
everybody,
this
is
really
helpful.
Did
it
merge
nikita?
No,
it's
not
merged.
A
Okay,
so
this
will
be
merged
shortly.
A
But
this
goes
through
and
has
examples
of
each
of
the
different
spi's
and
pretty
good.
Like
I
mean
not,
you
know
real
examples,
but
reasonably
real,
like
how
somebody
would
want
to
use
them,
and
also
it's
got
a
let's
see.
If
my
comment
is
here:
oh
I
skipped
over
it
a
really
nice
smoke
test,
verifying
stuff
that
was
customized
that
it
all
worked.
A
D
C
We
have
a
new
resident,
muzzle
expert.
I
would
totally
suggest
for
for
him
to
move
it
out
to
a
separate
project
and
publish
it
as
a
separate
gradle
plug-in.
D
F
D
C
So
I
think
the
the
the
limitation
that
we're
trying
to
work
around
here
is
for
the
gradle,
the
the
the
maven
comparison
part
correct.
G
E
D
F
F
C
Somehow
have
to
publish
two
plugins
or
well
so
maybe
the
way
that
I
would
recommend
doing
that
is
refactoring
the
the
gradle
plugin
to
be
a
little
bit
more
generic
and
maybe
provide
an
interface
that
says:
hey,
I'm
going
to
be
responsible
for
downloading
a
range
of
artifacts
from
maven
and
applying
a
function
to
it
and
then
having
that
function
be
a
separate
thing
that
is,
is
not
tied
to
that
plug-in.
E
E
A
All
right,
let's
that
was
a
much
needed
conversation.
I
would
don't
have
a
ton
of
time
left.
So,
let's,
let's
skip
to
sergeys,
make
sure
we
get
to
sergey's
questions.
B
A
What
do
people
think
about
d
by
default?
Looking
for
the
properties
file
in
the
directory,
where
the
agent.
C
Is
so
I
think
that
that's
a
little
so
when
I
was
at
new
relic,
I
found
that
having
a
default
location
tended
to
result
in
more
confusion
than
having
it
be
explicit.
E
E
A
And
will
you
still
will
it
still
will
users
if
they
use
the
open
telemetry
agent
settings?
Will
those
override
the
settings
in
your
config
file.
E
E
C
Outcomes
go
ahead.
I
was
going
to
say.
I
think
that
the
config
file
should
have
precedence
just
above
defaults,
but
system
properties
and
environment
variables
should
override
the
config
file.
That's
my
opinion.
A
Pawel
because
I'm
doing
something
similar-
and
I
actually
I
there's
not
a
hook
currently,
but
I
would
prefer
to
suppress
the
open
telemetry
default.
I
mean
configuration
story
and
only
have
one
configuration
story.
A
D
A
Because
some
vendors
have
different
configuration
approaches
that
also
need
to
configure
other
things
too
in
the
agent
and
exposing,
and
also
vendors,
may
not
want
to
expose
all
of
those
configuration
options.
They
may
want
a
simpler.
You
know
less
configurable
agent.
D
E
Want
to
even
mismatch,
you
know
this
match
our
configuration
with
total
configuration.
I
don't
want.
Oh.
E
E
A
F
A
Exporter
spam
factory
right
right,
so
we
do.
We
do
pass
that
in
so
paul.
If
you
put
your
properties
in
the
property
source,
we
we
pass
those
into
when
we
configure
the
spam.
The
exporters.
E
A
Sorry,
I
don't
follow
the
so
in.
I
think
what
we're
asked
you're
saying
to
only
have
the
vendor
ones.
We
would
like
if
there
was
a
way
to
not
load
the
default
ones
and
only
take
the
property
source
provider,
and
maybe
that
could
even
be
a
setting
on
the
property
pretty
side
that
says
you
know
don't
use
defaults
or
something
like.
A
The
one
update
I
have
for
you
is
honorag
has
mentioned.
Somebody
from
the
aws
team
has
been
benchmarking
the
java
agent,
and
so
hopefully
they
will
be
contributing
that
to
this
repo
those
benchmarks
and
then
we'll
have.
You
know
we'll
be
able
to
use
those,
at
least
for
some
guidance.
A
It's
I'm
not
sure
if
we
want
to
go
that
far
to
publish
like
official
like
you
know,
require
minimum
requirements,
because
it's
kind
of
it
depends
so
much
on
the
app
and
it's
sort
of
controversial.
And
if
you
look
at
other
apm
like
apm
vendors,
don't
I
don't
think
any
of
them
publish
that
kind
of
information?
A
But
we
do
want
to
have
some
benchmarks
that
we
can.
You
know,
run
continuously
and
you
know
we
can
point
people
to
those.
G
A
D
D
E
A
Yeah,
there's
definitely
something
there's
some
investigation.
That
would
be
good,
there's
an
issue
here
and
not
just
heap
memory,
but
it
seems
like
native
memory,
which
could
be
related
to
class
classes
getting
loaded
or
so.
This
is
the
issue
that
honorag
mentioned.
That
has
pointed
somebody
from
the
aws
team
to
so
hopefully
we'll
get
some
some
benchmarks,
some
at
least
to
start
with,
but
yeah,
I
think
profiling.
It
will
be
fun
for.
A
A
A
A
So
you'll
see
there
were
a
bunch
of
these
across.
I
like
the
alphabetical
breakout
because
they
were,
there
were
huge
prs
if
you
haven't
seen
it,
take,
take
a
look
at
the
new,
the
new
stuff,
it's
nice
and,
in
particular,
oh
addresses
an
issue
that
resolves
an
issue
that
pavel
had
opened
up
a
long
time
a
long
time.
This
feels
so
long
ago.
A
Where
did
we
go
here?
Oh
we
had
a
an
author
of
a
library
author
of
this
library
wanted
to
added
their
library
to
our
list
of
instrumentation
and
in
particular
they
we
don't
have
auto
instrumentation,
but
they
have
added
that
instrumentation
themselves
in
their
http
4k.
A
So
what
we
did
for
now.
We
put
kind
of
put
this
asterisk,
that
the
support
is
provided
by
the
library
which
is
cool
in
this
version.
And
then,
if
you
link
in
here,
it
shows
how
to
add
that
support.
A
Oh,
a
bug,
not
a
bug
fix,
but
a
bug
repro,
which
is
equally
helpful,
if
not
more
helpful.
Sometimes
we
had
we
had
some
release.
Publishing
problems
with
zero.
Ten
part
of
the
problem
is
the
the
bin
tray
publishing
jfrog
publishing
keeps
timing
out
so
we're
looking
at
solutions
for
that.
E
A
D
A
It's
a
it
makes
sense,
it's
not
something.
It
was
kind
of
almost
just
a
workaround
that
we
were
thinking
to
help
with
the
publishing,
but
we
need
to
fix
that
publishing
anyway.
So
yes,.
A
Yeah
pablo,
maybe
once
nikita
and
once
nikita
merges
that
pr,
if
you
can
add
an
example
like
that
to
that,
that
would
also
help
to
make
sure
that
we
don't.
You
know
we're
aware
of
that
kind
of
extension.
E
A
Oh,
we
haven't
done
no,
no
they're.
All
there
we
haven't
made
that
change.
We
discussed
the
change,
but
we
won't.
We
won't
make
that
change.
If
you
need
it.
A
Yeah,
sorry
folks
we're
running
overtime
and
there
was.
There
was
one
thing
that
I
remember
wanting
to
chat
about
related
to
that,
but
we
will
chat
next
week,
nothing's,
nothing
urgent,
oh
yeah.
If
you
want
so
we
meet
on
tuesday
nights,
which
I
know
is
an
inconvenient
time
for
a
lot
of
people,
but
happens
to
be
the
only
time
that
myself
and
nikita
and
honorag
are
all
awake
at
the
same
time.
A
But
we
do
have
notes
from
those
meetings.
So
if
you
have
see
anything
there
that
you
want
to
chat
about,
that's
always
something
you
know
who
can
ping
us.
Also
there.
A
Yeah
yeah,
they
are
they're
on
the
they're
on
the
google
calendar
and
they
do
get
recorded
also
in
they
should
be
on
youtube.
A
A
Yes,
yes,
your
wednesday
morning,
yes,
yes,
there
are
public,
they
are
recorded
on
youtube.
They
are.
We
try
to
take
trying
to
take
better
meeting
notes
during
them,
but
every
everyone
is
welcome.
They're,
totally
public.