►
From YouTube: 2021-03-31 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
C
Yeah
yeah,
it
was
spring
break
where,
where
I'm
at
so.
C
C
Yeah
so
I
can
go,
I
don't
really
have
have
anything
particular
on
the
agenda.
I
just
figured.
We
can
quickly
go
through
the
prs
that
that
are
out
there
from
what
I
can
see.
It
doesn't
look
like
there's
any
big
blockers.
It
looks
like
zack.
You've
had
a
chance
to
take
a
look
at
at
all
of
them,
and
so.
D
C
A
I
have
two
topics:
okay,
first
one
I
was
watching
just
before
this
meeting
the
the
last
meeting,
so
someone
is
looking
at
it.
So
basically
I
remember
that
zac
were
asking
about
plans
and
things
like
that.
I
don't
know
if,
if
paulo
had
any
chance
to
grab
this
list
of
what
we
want
to
do,
but
even
if
he,
if
he
did
or
didn't
he
got
the
time.
A
D
Yeah
I
got
he
mentioned
that
there
is
a
couple
things
like
links
and
baggage
and
whatnot.
It
was
just
like
a
private
message.
Let
me
see
if
there's
anything
else,
but
it
seemed
like
most
of
the
other
things
that
are
gonna
be
added,
are
going
to
be
just
additive
and
not
really
overlapping.
What
we
have
so,
let's
see
yeah,
maybe
changing
a
couple
of
tabs
kind
of
like
the
http
convention
that
we
just
did
and
then
mostly
events,
links
and
baggage,
and
so
that's
what
paolo
told
me.
A
Yeah,
I'm
asking
also
because
I
know
that
paulo
didn't
have
any
time
to
do
it.
Personally,
I
didn't
have
time
at
all
also
this
week,
but
I
don't
know
if
we
should,
I
don't
know,
start
to
plan
or
using
github
to
make
things
more
transparent,
even
as
headers
without
details,
even
at
you
know
such
high
level
point
of
view.
A
D
D
E
A
E
E
E
No
totally
yes,
yes,
sorry!
I
was
actually
out
not
following
things,
because
I
was
playing
with
profilers
for
the
last
couple
of
weeks,
so
I'm
not
completely
up
to
date,
but
if
there's
anything
I
can
do
with
that
kind
of
stuff,
I
can
spend
half
an
hour
or
an
hour
whatever
time
it
is
required
to
kind
of
help,
set
it
up
or
show
it
around
to
like
just
do
it.
I
just
don't
know
what
what
is
going
on
right
now.
So
I'm
kind
of
I
need
some
jump
start.
E
Oh
no,
no!
No!
I
was
just
playing
around
like
with
making
things
work
and
like
there
is
not
much
to
present
I
can
like.
I
can
have
lots
of
interesting
conversations
about
that,
but
I
don't
want
to
distract
from
the
mainstream
work.
So
what.
A
I
see
so
that
was
one
topic,
and
I
also
maybe
wanted
to
pro
propose
that.
Maybe
if
you
have
time
today,
we
can
even
go
through
the
list
and
titles,
maybe
together
just
to
have
an
alignment,
and
second
thing
is
something
that
basically
erasmus
started
working
on.
Is
that
basically
in
splunk,
we
want
to
start
to
create
something
like
like
a
branded
version
of
this
auto
instrumentation
for
hotel,
and
I
think
that
our
like
stuff
will
be
something
like
initially
a
kind
of
poc.
A
A
We
have
some
idea
that
maybe
we'll
not
need
to
create
a
real
fork,
maybe
we'll
be
able
to
just
have
some
sub
modules
and
use
some
weak
wicks
set
properties-
I
don't
know-
maybe
some
dedicated
just
you
know
just
have
an
ability
to
replace
some
stuff
so,
for
example,
to
create
our
own
installer,
maybe
to
change
some
stuff
like
to
set
the
default
and
variable
variables
which
have
different
values
and
stuff
like
that.
A
Meaning,
meaning
personally,
I
think
that
if
all
of
us
will
follow
with
forking
the
repos,
I
think
it
will
be
a
nightmare
to
maintain
for
all
of
us
like
having
some
kind
of
compound
componentization.
Even
some
github
modules,
I
think,
will
be
better
for
us.
If
we
just
you
know,
work
on
one
repository
like
the
core
stuff,
like
this
auto
instrumentation,
stuff,
exporters,
etc,
and
having
some
way
to
to
customize
the
thing
which
is
specific.
E
C
Yeah
so
greg.
So
I
believe
what
robert's
talking
about
is
being
able
to
create
the
vendor-specific
versions
of
this
software
and
what
that
process
is
going
to
look
like
for
each
vendor
and
so
splunk's
experimenting
with
not
using
a
hard
fork
of
the
repo
and
instead.
E
Basically,
basically,
essentially,
rather
than
compile
time
configuration
some
sort
of
conflict
based
configuration.
A
E
It
makes
sense
I
mean,
but
but
in
terms
of
what
we
can
do,
it
depends
on
two
things.
First
of
all,
in
the
details
of
of
what
what
you
come
up
with
and
second
is
on
this.
C
E
Several
key
team
members
working
on
it,
so
it
would
be
hard
to
promise
to
commit
to
things
without
them
being
here.
A
E
A
E
But
we
can
commit
to
looking
at
it
and
considering
it
seriously.
A
A
At
this
model,
not
that
you
just
wanted
to
have
you
know
initial
feedback
if
it's
a
good
plan
even
to
work
on
such
stuff,
because
if
you
told
from
the
beginning
that
you
know
that
anything
else
in
a
fork
will
not
work
or
there
are
some,
you
know
things
that
are.
You
know,
for
example,
already
that
we'll
need
to
make
too
many
stuff
that
it's
not
realistic
for
us,
then
that's.
Why
we're
asking
right
now.
E
I
don't
see
a
problem
as
such
in
the
principle.
I
think
the
principle
is
actually
even
better
the
only
problem
so
so
there
is,
there
is
actually
two
two
things.
One
is
like
the
whole
idea
of
not
having
a
hard
fork,
I
think,
is
a
good
one.
E
The
the
from
our
perspective,
the
main
concern
is
more
just
like
how
much
regular
integration
work
it
needs
to
happen
to
like
have
prs
flow
back
and
forth.
I
think
the
leadership
philosophy
is
essentially
that
like,
if,
essentially,
if,
if
we're
spending
some
significant
amount
of
time
for
for
just.
E
Like
outside
of
data
dog
reports,
then
they
really
start
asking
hard
questions,
how
it
benefits
the
business,
and
if
we
can
answer
them,
then
we
can.
If
we
cannot,
then
they
go
like
spend
less
time
right.
E
So
that's
that's
the
kind
of
high
level
product
leaderships
approach
to
this
and
we
need
to
operate
within
those
call
chains
and
then,
in
terms
of
implementation,
as
in
sub
modules,
up
to
you
to
experiment,
I
just
like
we,
we
recently
looked
at
at
having
a
product
built
from
several
repos
and
we
as
soon
as
somebody
mentioned,
the
the
term
sub
modules.
E
Everybody
who
was
involved
in
this
conversation
said:
oh,
my
god,
oh
no,
no,
no,
no,
no,
and
that
was
even
before
we
had
the
chance
to
actually
discuss
details
and
in
the
past
I
also
made
like
experience
that
sub
modules
are
like
hard
to
work
with
you
get
like,
but
I
see
no
principal
problem
with
that.
It's
just
usually
it
ends
up
and
if
you
like,
google,
it
and
read
about
it,
you'll
find
examples
in
our
case
where
we
needed
to
work
with
several
repositories.
E
We
have
we
haven't
finished,
but
we
probably
will
just
create
a
convention
because
essentially
we
have
code
that
depends
on
on
each
eye
on
each
other,
like
there
is
like
one
piece
of
product
and
another
piece
of
product
on
the
share
code.
And
how
do
we
do
this?
E
So
we
have
created
a
multi-put
dependency,
at
least
at
this
point,
where
we
have
like
some
shared
code
in
one
repo,
and
then
everybody
else
depends
on
this,
and
that
is
with
an
idea
that
the
shared
code
changes
much
less
frequently
than
other
things
so
and
then,
if,
if
there
is
a
change
that
requires
co-changing,
both
repos
we're
actually
still
working
through
how
we're
going
to
do
this,
I'm
not
sure
it
will
work
in
this
particular
case.
E
Maybe
it
will,
but
it's
just
as
a
consideration
that
sub-modules
can
like
can
be
really
annoying
to
work
with
when
you
have
like
a
logic,
but
generally
I
want.
I
think
it's
a
good
idea
like
what
you
want
is
actually
what
I
would
want
to.
I
just
don't
know
how
realistic
it
is
to
implement
it.
G
I
think
it's
fine.
Currently,
we
still
need
to
think
it
through
also.
A
So
maybe
I'll
just
quickly
say
that
I
was
also
thinking
of
other
other
approaches,
but
I
felt
that
they'll
be
simply
unrealistic,
meaning
like
people
hate
sub
modules,
because
that's
where
packaging
like
there
are
nougats
and
stuff
like
that.
But
I
think
that
making
nougat
packages
for
vix
projects
and
stuff
like
that
for
also
we
have
cpp
and
other
stuff.
I
think
that
it
will
be
simply
too
big
step
to
make
packages
all
this
stuff
and
it
will
be
safer
to
first
start
with
step
modules
and
think
of
a
compile
time
like
customization.
E
There
is
also
one
other
possibility
that
you
can
consider,
and
I
really
don't
know
whether
it's
it's
worth
it,
but
just
the
consideration
right.
E
So,
theoretically,
speaking
as
an
open
telemetry
project,
you
want
a
lot
of
vendors
and
unlimited,
like
you
know,
blah
blah
blah
practically
speaking
at
this
stage
there
is
neuralic
splunk
us
and
microsoft
kind
of
hanging
out,
so
so
in
the
world
where
in
reality,
we
have
like,
on
the
order
of
five
vendors,
actually
contributing
and
actually
discussing
things
and
the
rest
might
just
clone
a
repo
and
silently
do
something
but
they're
not
contributing
right.
E
Maybe
we
can
like
not
go
as
generic
as
the
sdk
guys
went
where
they
have
a
structure
here
is
like
generic
thing
goes
here
and
when
the
things
go
there,
but
not,
maybe
we
can
just
have
like
one
monolithic
thing
without
even
sub
modules,
just
structured
folders,
where
you
can,
if
you,
if
you
have,
because
I'm
guessing
and
correct
me
if
wrong,
that.
E
Because
for
now,
for
us,
everything
is
open
source
even
stuff
that
we
don't
want
to
be
part
of
of
open
telemetry
if
it
runs
in
customer
machines,
it's
open
source,
so.
E
I
don't
know
how
it
is
for
you,
but
if
it
is
a
case
two
maybe
we
can
just
have
like
structured,
folders
and
stuff
that
is
like
just
for
your
branding
goes
into
some
folder,
where
it
is
kind
of
isolated
from
the
rest,
so
that
you
can
even
like
choose
not
to
merge
it
or
choose
to
like
merge
it,
but
it's
very
easy
for
everybody.
Congratulations.
A
A
A
The
only
thing
which
is
done,
at
least
for
us
in
this
splunk
like
this
or
this
distro
stuff,
is
just
making
something
like
by
default,
enabling
the
the
stuff
for
splunk.
So
it's
just
a
matter,
for
example,
as
setting
the
environmental
variables
in
the
in
as
default
and
stuff
like
that,
and
there's
no
almost
no
code
more
code
like
that.
E
Yeah,
so
so
exactly
like,
if
it's
like
imagine
that
we
display
a
branded
logo
at
startup
which
we're
not
but
like,
as
I
think,
maybe
everything
all
the
logos
can
be
in
the
repo
just
in
different
folders.
Everything
is
hardcoded
and
then
another
or.
E
Yeah
yeah
like
exactly
but
the
thing
is
all
of
these
like
it
may
be.
It
may
be
architecturally
better.
That
autel
is
just
like
another
report,
but
it's
just.
That
is
the
one
that's
selected
by
default,
unless
you
like
make
some
other
setting.
E
E
I
think
that
might
be
like
that
is
what
I
was
pushing
for
before
you
guys
started
working
very
actively
on
it
and
when
there
was
essentially
mainly
just
I
was
contribution
which
is
changing
now,
but
that
is
what
I
was
kind
of
pushing
towards
before
when
when
code
was
flowing
much
less
and
I
was
like
why
any
divergence
so
just
as
a
consideration,
I
guess.
C
Yeah
and
the
the
other
thing
that
I
want
to
call
out
there
is
that
not
all
vendors
will
want
to
have
their
own
distribution
of
this.
I
for,
for
example,
for
new
relics,
side.
We
won't
need
our
own
distribution
of
this.
We
can
just
go
with
the
vanilla,
open,
telemetry
version
of
it,
there's
no
branding
requirements
or
anything
there
so,
but.
E
You
may
you
may
actually
you're
not
sure,
because
maybe
so
in
terms
of
branding
yes
like,
as
in
I
don't
know,
the
metaphorical.
C
About
x,
but
exporting
also
as
long
as
the
vanilla,
open,
telemetry
distribution
supports
the
standard,
open,
telemetry
exporters
and
by
standard
I
mean
things
like
zipkin
jaeger,
otlp
new
relic
will
be
able
to
consume
it.
E
And
you
might
also
have
a
different
set
of
default
settings
depending
on
you
know,
maybe
performance.
A
E
Yeah
and
perhaps
neuralic
wants
a
to
export
via
the
collector
or
directly
depending
on
what
they
want,
and
they
may
mainly
like.
If
they
choose
to
export
directly,
then
they
might
have
their
like
by
default,
their
own
thing
plugged
in
so
it's
just
like
a
different
level
of
customization.
So
branding
branding
is
already
a
kind
of
customization,
but
I
meant
like
genuinely
a
vendor-specific
customization.
C
Yeah
and
in
practice
most
of
our
customers
are
running
their
own
collector
instance
to
begin
with,
because
they
like
to
be
able
to
filter
out
any
sensitive
attributes
on
the
data
before
sending
it
out.
So.
D
So
a
question
I
was
going
to
ask
a
little
bit
earlier,
which
I
think
helps
sort
of
guide
how
we
want
to
approach.
This
is
how
do
we
what's
our
plan
for
setting
like
the
different
configuration
knobs,
because,
like
one
way,
we
have
right
now
is
just
via
environment,
variable
and
so
like.
D
If
everything
is
done
that
way,
then
we
can
basically
we
that
would
call
for
building
everything
all
the
different
options
in
the
hotel,
repo
and
then
just
having
like
your
distribution
channel
like
the
msi
or
whatever
you
know,
set
the
proper
environment
variable
to
go
through
that
convention
or
if
we
don't
want
to
combine
everything
and
see
this
one
repo,
then
that's
an
area
where
okay,
so
then
each
repo
is
going
to
have
to
provide
its
own
exporter.
D
A
A
D
D
No,
that
one's
just
we
tell
people
a
series
of
steps
so
yeah.
If
our
goal
is
to
basically
just
have
the
different
like
we
already
have
the
zipkin
and
the
jaeger
exporter
and
just
keep
that
and
not
really
have
additional
vendor
logic
inside
this
repo,
then,
and
just
configuration
based
on
environment
variables.
Then
it
sounds
like
yeah.
D
The
approach
for
just
doing
a
sub
module
could
work
because
you
just
might
need
to
change
something
about
some.
You
know
configuration
or
deployment
where
you
don't
actually
need
to
vendor
in
code,
but
then,
if,
if
there's
some
other
code
you
wanted
to
vendor
in
then
I
don't
know
how
difficult
situation
would
be
to
have
a
sub
module.
E
Well,
with
only
an
environment,
basically
runtime
configuration,
I
think
some
things
can
be
covered
by
it,
but
we
won't
completely
get
away
with
it
simply
because
branding
might
also
affect
how
you
configure
at
runtime
right,
because
right
now
at
least
a
lot
of
variables
you
set
are
the
data
that
we
did
branded
so
and
for
us
it
will
be
nearly
impossible
to
move
away
from
it
simply
because
it
would
be
breaking
the
customers
right,
like
you
already
have
things
setups
with
those
particular
environment
variables
names
in
production.
E
So
if
a
vendor
wants
to
start
configuring
things
without
dd
branding
and
we
need
some
kind
of
prefix
right,
we
will
need
to
have
some
sort
of
structure
in
variables
that
we
configure
like
prefix,
followed
by
some.
You
know
not
just
random
stuff,
so
we
may
or
may
not
be
able
to
use
the
current
conventions
there,
depending
on
what
the
other
vendors
want.
A
E
I
doubt
it
is
for
two
reasons:
one
is
breaking
change
that
would
break
it,
change
that
that
would
mean
we
can
support
two
sets
of
variables,
but
that
you
know
imagine
the
customers
running
stuff
in
production.
We
are
saying
you
you
can
no
longer
upgrade
and
unless
you
change
all
your
deployment
scripts
that
set
environment
variables
and
second
is
also
consistency
with
with
other
vendor
products.
E
So
if
you're
configuring
say
say
you
have
whatever
splunk
and
you
you
have
a
tracer
for
for
java
and
you
just
happen
to
configure
java
tracer
underscore.
E
I
don't
know
and
enable
this
enable
that
and
then
you
have
the
same
for
go
and
then
you
have
go.
Tracer
underscore
enables
this
enable
that
and
then
yeah
and
so
on,
and
then
you
you
just
go
for
the
net
or
for
the
net
you
go
hotel.
E
Dot
net
tracing
enables
enable
that
and
then
for
the
for
customers.
It's
it's
confusing,
so
vendors,
at
least
for
us.
It's
it's
not
the
most
important
priority,
but
it's
not
completely
out
of
the
picture
to
have
configuration
be
consistent
across
different
products,
so
we
won't
be
able
to
break
with
it.
Just
for
the
sake
of.net
so
again,.
C
Related
to
that,
I
I
I
want
to
understand
a
little
bit
more
about
how
datadog's
handling
hotel
java,
because
datadog
also
donated
their
their
tracer
to
the
java
sig,
and
I'm
wondering
how
data
dogs
dealing
with
any
divergences
there.
E
That
is
a
good
question.
I
don't
know
if
the
details
about
that
very.
D
Elementary
understanding,
which
is
at
this
point,
they're
not
really
doing
a
lot
of
there's
a
lot
of
friction
between
porting
stuff,
and
so
it's
not
like
a
constant
upstreaming
one
way
or
the
other,
and
so
I
think
the
integration
points
are
pretty
small
now,
where
they,
you
know,
take
only
fixes
that
are,
like
you
know,
pretty
large
or
impactful.
I
don't
think
there's
a
lot
of
code
flow
going
in
the
way
at
this
point,
as.
A
G
A
E
So
I'm
not
familiar
with
this
on
that
level,
but
on
a
high
level,
I
I
described
correctly.
There
is
essentially
the
history
of
it
that
we
donated
the
code
and
initially
there
was
a
a
lot
of
code
flow
in
both
directions,
but
this
happened
way
longer
ago
than
the
net
and
over
time
it
has
subsided
significantly
and
I'm
not
sure
so
that
I
recently
actually
spoke
to
the
guy
who
was
driving
all
that
work.
E
Exactly
with
this
question,
I
wanted
to
know
how
it
went
and
how
they
organize
the
repos
and
everything,
and
he
said
that
early
on,
which
is
by
now
over
a
year
ago.
E
He
was
actively
creating
a
lot
of
aprs
to
have
data
the
code
flow
into
hotel
after
the
after
the
donation
as
well,
and
it
was
being
reviewed
and
merged
and
then
over
time
it
the
product,
diverged
and
now
he's
doing
it
very
rarely,
and
maybe
auto
people
are
like
older
people
can
do
it
themselves,
but
they
also
do
it
relatively
rarely.
E
Taking
targeted,
integrations
and
then
also
datadock
takes
some
bug
fixes
back
from
from
open
telemetry,
but
also
in
a
targeted
way,
rather
than
almost
everything,
so
they
they
have
given
up
on
the
on
this
idea
of,
like,
let's
be,
have
very
little
divergence
to
make
integrations
easy,
and
one
of
the
reasons
for
it
just
for
context
is
microsoft
has
resourced
a
bunch
of
work,
so
they
have
like
had
a
bunch
of
people
working
full
time
on
on
this,
so
they
were
working
with
datadog
they
when
they
donated
the
tracer
and
then
the
microsoft
guys
were
making
a
lot
of
contributions
and
they
made
at
some
point
their
architectural
decision
to
diverge
because
they
had
different
priorities
so
and
because
they
were
investing
a
lot
they
like,
where
whoever
contributes.
E
C
Yeah,
so
at
a
high
level,
because
there
is
that
divergence,
I'm
curious
how
datadog
is
approaching
those
differences.
So
is
there
a
separate
data
dog
version
of
the
hotel
java
tracer,
or
is
it
still
a
separate
theta
dog
tracer
and
a
separate
hotel
tracer,
or
are
they
one?
In
the
same?
I
guess
that
at
a
high
level,
I'm
trying
to
understand
that,
because
that
also
affects
how
you
do
things
like
handling
the
different
environment
variables
so
that
you
can
remain
consistent
across
all
of
your
different
languages
and
so
on.
E
It's
actually
a
good
question,
because
I
think
part
of
the
question
is
more
a
way
of
about
how
people
think
about
it
rather
than
technical,
because
technically
it's
just
called
data.
I
think
in
terms
of
how
people
think
about
it
at
data
doc
in
the
java
team.
It's
a
separate
choice.
D
E
Like
when
I
talked
when
I
talked,
I
wanted
to
know
some
details
about
how
they
organize
the
reports,
and
then
they
say
things
like
you
will
wear
working
closer
together.
Now
they
diverged
from
us
this.
This
is
like
a
formulation
that
more
or
less
like
this
they're
phrasing
it
so
this
this
is.
This
tells
you
something
about
how
they
think
about
it.
C
E
Like
yeah,
so
in
terms
of
naming
of
around
of
like
customer
facing
configuration,
I
don't
see
a
problem
with
supporting
multiple
things
as
long
as
they
like
can
be
made,
not
contradictionary,
but
I
don't
see
that
we
can
kind
of
move
away
from
current.
Maybe
if
we
ever
strip
some
completely
new
like
big
breaking
change
version
of
the
of
the
tracer,
then
we
can
talk
about
it,
but
even
then,
this
consistency
across
products
is
is
important.
E
I
think
from
the
product
perspective
people-
and
I
I
would
I
would
if
I
was
a
product
person-
I
would
agree
with
that-
they
will
say
like
look.
How
do
you
explain
the
customer
that
everything
else
you
configure
with
data
dockers?
Did
you
underscore
and
suddenly
you
do
auto
underscore
like
why
they
go
like
as
a
customer
there.
They
sometimes
want
everything
they
go
like.
Oh,
this
is
so
great
that
datadog
is
doing
open
telemetry,
but
why
is
it
different?
I'm
annoyed
so.
D
A
It's
worth
bringing
up
because
it's
important
so
the
ch.
I
think
that
what
greg
told
right
now
is
very
important,
because
it
means
that
for
data
doc,
for
example,
it
should
be
possible
that
different
names
of
environmental
variables
should
work
at
least
to
make,
for
example,
customers
life
happy
like
like
with
like
happier
right
correct.
A
E
I
am
guessing
so
yes
and
it
may
or
may
not
be
similar
to
for
other
vendors.
If
say
you
say
hey,
I
have.
E
We
would
like
to
have
some
kind
of
splunk
splunk-like
configuration
either,
because
we
have
other
products
that
are
already
like
this
or
whatever,
so
it
or
compatibility
or
whatnot
so
and
then
there
there
will.
There
is
al.
Also,
you
may
not
want
to
support
some
options.
This
is
another
thing
so
like
personally,
I
I
am
very
keen
on
having
as
little
options
as
possible,
because
every
every
optional
thing
that
the
customer
can
do
means
more.
E
More
support,
cost
and
other
vendors
might
take
make
individual
decisions
slightly
differently,
but
they
might
say,
hey,
there's
a
particular
configuration
option
that
we
want
to
support
or
not
yeah.
A
E
A
E
Not
necessarily
so
here's
an
example.
This
is
just
this.
This
is
artificial
example,
so
just
bear
with
the
spirit
of
another
details.
So
say
we
decide.
Okay,
we
want
to
run.
We
want
to
be
able
to
configure
the
export
at
runtime.
The
customer
should
should
should
be
able
to
just
change
the
environment
variable
and
say
I
usually
export
into.
E
I
don't
know
zipkin
and
now
I
want
to
export
into
jager
right,
and
we
might
say
you
know
what
and
then,
if
we
allow
this,
you
first
make
this
this
setting
and
then
the
setting
cannot
possibly
be
just
one
setting,
because
usually
there
is
some
sort
of
different
data
model,
something
else
and
something
in
addition
needs
to
be
configured
and
then
you
do
it
incorrectly,
and
you
call
up
our
support
saying
why
things
are
broken.
We
don't
want
to
support
this.
We
never
want
this
to
happen.
E
This
is
not
a
scenario
that
our
business
invests
in.
We
might
support
jager
and
zipkin,
but
it
should
be
something
that
you
take
our
configuration
distribution,
you
install
it
and
it's
never
configurable,
because
we
pre-configured
that
to
make
it
work.
So
we
don't
want
to
expose
the
setting,
because
we
don't
want
to
have
customers
who
first
changed
it
and
then
call
us
up
and
say
I
paid
you
money
and
now
help
me
configuring
it.
E
So
we
might
want
to
say
setting
not
supported,
even
if
open
telemetry
made
the
decision.
Yes,
we
do
want
to
support
it,
because
it's
a
business
decision
for
us
to
have
less
configurable
stuff
as
an
example
right
and
and
so.
E
So
I
would
I
would.
I
would-
and
I
realized
that
this
is
a
very
approximate
answer,
so
I
would.
E
I
would
first
go
with
compile
time
if,
if
in
doubt,
I
know
that
we
as
developers
and
I've,
seen
it
like
ways,
developers
really
like
generalizing
and
making
things
configurable
and
creating
abstractions,
but
I
think
experience
and
across
different
products
and
teams
shows
that
really
abstractions
for
the
sake
of
professions
is
bad
and
when,
unless
we
have
a
customer
who
actually
wants
to
configure
things
and
it's
a
clear
business
case
and
they
want
they
go
like
unless
you
make
it
configurable,
we
don't
pay
you
money
and
if
you
do
make
it
configurable,
we
do
pay
your
money.
E
And
if
you
want
to
make
them
different
across
vendors,
then
it
should
be
a
super
easy
compile
time
configuration.
So
if
we,
if
we
have
some
sort
of
one
place
where
we
have
like
a
source
file
or
a
build
file
or
whatever
it
is
where
we
like,
set
these
options.
So
it
is
configurable,
but
it's
configurable
before
you
build
the
product
not
for
for
the
customer
and
only
for
us.
So.
A
E
For
example,
if,
if
like,
if
if
something
happened,
where
you
say
there
is
a
hotel
standard
that
requires
this
and
this
to
happen,
but
it
it
costs
you
at
runtime,
it's
you
know
there
is
like
some
measurable
slowdown
in
terms
of
how
we
affect
the
customer
application.
It
is
quite
likely
that
we
would
say
well
when
a
customer
can.
E
Yes,
when
the
customer
comes
and
says
we
want
it
and
we
pay
money
if
it's
supported
and
we
don't
pay
money.
If
it's
not,
then
we
will
would
say
okay.
Well,
then
we
support
it.
But
if
something
is
like
well,
it's
open
telemetry
standard,
but
everything
gets
slower
and
no
customer
asked
for
it.
Then
it
would
probably
say
no.
E
C
D
Is
this
discussion
a
little
premature,
because
once
we,
if
we
finish
building
out
what
the
hotel
compliant
thing
is,
then
we'll
have
known
like
what
all
the
configurability
knobs
are?
So
I'm
wondering
if
we
should
try
and
make
target
first
the
hotel
compliance
before
doing
the
vendor,
or
is
this
just
like
a
thought
exercise?
A
D
D
A
C
A
D
Update
from
my
end
so
there's
it's
kind
of
the
same
update
as
last
week.
For
me,
I'm
trying
to
port
a
couple
of
changes
back
from
hotel
into
datadog,
but
there's
a
lot
of
code
going
in
right
now,
so
it's
kind
of
slow
getting
the
changes
back
in,
but
trying
to
make
my
best
efforts
to
do
that.
A
Have
you
seen
the
markdown
that
is
in
hotel
regarding
regarding
cherry
picking
and
stuff.
A
A
A
A
The
first
step,
which
I
was
like
resolving
right
now
today,
meaning
is
also
the
names
of
the
dll,
so
this
prefix
from
data
dog
tracer
for
you
for
us
like
for
the
hotel,
it's
named
open,
telemetry
dot,
auto
instrumentation.
So
these
are
the
three
most
frequent
changes
that
you
need
to
remember:
okay,.
D
Yeah
at
this
point,
it's
just
going
over
the
changes
getting
just
reviewing
them,
and
I
mean
we
actually.
We
have
so
many
changes
that
we're
trying
to
take
in.
So
it's
actually
usually
a
backlog.
I
think
we
now
have
like
almost
20
pr's
open,
so
just
gotta
go
through
them
as
get
yeah
just
get
through
them.
E
I
have
a
question.
I
have
an
off
topic,
actually
chris,
if
you,
if
you're
like
I,
it's
an
off
topic
question.
So
if
you're
gonna
go,
then
let
me
know
it's
a
pro
profiler
related.
I
was
just
looking
again
at
what
you
guys
have
and
playing
around,
and
do
I
see
it
correctly?
You
have
wall
clock
profiles
right
so
in
in
your
thing,
you
you
collect
how
much
real
time
elapsed.
C
E
So
my
question
is
about
customers
how
customers
are
using
it
because
a
lot
of
the
time
you
just
see
everything
just
waiting.
C
So
you've
got
a
good
question.
Unfortunately,
I
do
not
have
any
good
answers
about
how
customers
are
using
the
the
profiling
functionality.
C
I
know
one
common
use
case
for
the
the
profiler
functionality
that
we
have
is
you
got
somebody?
They
don't
actually
have
the
source
code
available
for
their
application
and
they
want
to
just
kind
of
get
an
idea
of
what's
running
in
their
system,
and
so
they
run
the
profiler
to
kind
of
get
this
overview
of.
What's
running.
D
C
Once
they
have
that
overview,
then
it
kind
of
gives
them
an
idea
of
whether
or
not
there's
any
gaps
in
the
information
that's
already
being
presented
via
the
auto
instrumentation,
and
at
that
point
they
can
then
identify
places
to
apply
custom
instrumentation
where
they
want
some
additional
insight.
C
E
I
understand
makes
sense:
no,
if
they
don't
have
circuit,
they
could
just
decompile
it.
I
guess,
but
that
way
that
that's
meter.
F
C
H
E
Yeah
I
was,
I
was
looking
at
all
this
and
in
linux
you
can
get
some
information
from
from
this
from
the
ois
about
cpu
timings,
but
in
windows
I
either
you're
running
as
admin,
in
which
case
you
can
listen
to
etw
for
context,
changes
which
will
tell
you
about
cpu
timings.
But
since
we
are
running
in
all
these
applications
that
we're
often
not
admin
etw
is
not
a
thing
and
in
windows
you
can.
E
That
is
somehow
is
actually
real
time,
because
you
know,
if
I,
if
I,
if
I
walk
stacks,
if
I
do
a
sample
based
profiling,
I
know
how
much
real
times
passed
between
you
know
my
samples
for
a
given
thread,
so
I
can
do
walk
look.
I
can
also
ask
the
operating
system
how
many,
how
many
cycles
the
thread
used
up
between,
like
you
know
between
then
and
here.
So
I
can
get
a
rough
idea
about
how
much
work
actually
happened.
E
Versus
waiting,
but
clocks
are
dynamic
and
and
clocks
can
be
even
different
between
cpu
cores
so
yeah.
So
I
have
no
idea
how
to
like
as
a
rough
indication.
You
know
this
thread
made
a
lot
of
progress
versus
this
didn't
makes
them
a
lot
of
progress.
Yes,
that
that's
good,
but
to
actually
make
it
comparable.
You
know
across
time
and
thread
and
everything
I
I
don't
know
so.
C
C
So
it's
similar
to
the
cpu
time
model
that
you
mentioned,
but
it's
possible
that
your
span
was
sitting
there
waiting,
but.
E
C
Really
yeah,
so
you
can.
You
could
have
async
code,
so
yeah
I'd
have
to
do
some
pseudo
code,
but
the
idea
is
you've
got
a
method,
and
in
that
method
you
can
you're
gonna
wait
for
some
other
method.
That's
gonna
take
five
seconds.
C
Necessarily
now
you
don't
know
without
looking
at
the
code.
If
your
outer
method
is
going
to
be
doing
work
during
those
five
seconds
or.
C
E
Yeah
makes
sense:
okay,
well
I'll,
keep
looking
and
I'll.
Let
you
guys
know
what
I
find
out
yeah
so
far,
yeah
wall
clock
is
good,
but
you
know
I
look
at
it
myself
and
you
know.
If
you
write
this
profile,
then
one
thing
that
the
profile
is
doing
is
waiting
for
it
for
itself
to
have
for
itself
to
have
something
to
to
do
and
then
doing
it
and
most
of
the
time
it's
just
waiting.
E
So
if
you
it's
very
funny,
if
you
do
just
walk
up
profiling
of
your
of
an
application,
then
it
falsely
says
that
the
most
expensive
thing
in
the
application
is
the
profile
itself,
because
because
it
has,
it
has
like
a
method
that
is
doing
something
once
per
once
per
blue
moon
and
most
of
the
time
it's
just
waiting.
But
it's
always
there.
C
Yeah,
it's
I
I
don't
know
if
you've
had
re
requests
to
exclude
the
profiler
code
from
your
profiles.
H
But
so
far
we
have
no
product
we're
just
like
okay,.
E
You
know
looking
at
things
if
it
was
again,
if
it
wasn't
customer
machines,
you'd
be
seeing
it
as
a
as
an
open
source
repo.
It's
just
like
you
know,
investigations
and
playing
around
so
hopefully
at
some
someday
in
the
future.
It
will
be,
but,
like
you
know
so
far,
it's
just.
D
C
All
of
a
sudden
you'll
see
that
oh
there's
instrumentation
running
in
the
applications,
and
now
people
are
seeing
that
code.
Oh,
how
much
time
is
your
instrumentation
taking
wait?
What
did
you
do
to
my
code?
That's
not
there
and
then,
if
you
have
background
threads
all
of
a
sudden,
those
are
showing
that
it's
taking
a
lot
of
time
right,
but
is
that
affecting
the
cust
the
performance
of
the
customer's
application?
D
We
already
have
confusion
from
our
call
site
instrumentation
because,
unlike
yours,
where
it's
like
well,
we
have
call
target
now,
but
it's
it's
still
not
privatized,
so
we
always
show
up
on
the
call
stack
and
customers
see
that
in
their
exception,
like
hey,
did
you
do
this?
We're
like?
No?
No.
We
didn't
do
that.
D
D
D
We
had
some
microsoft
people
jump
in.
I
don't
know
if
you
noticed
them.
I
think
they're
actually
waiting
for
the
that,
like
the
event
pipe
presentation,
so
they'll
probably
be
back.