►
From YouTube: 2023-03-01 meeting
Description
cncf-opentelemetry meeting-2's Personal Meeting Room
A
I,
don't
know
who
else
is
going
to
show
up,
but
you
and
I
might
be
it.
The
Microsoft
folks
like
see
Joe
and
Michael
and
utkarsh.
They
are
all
like
some
half
day
meeting
type
of
thing,
so
they
are
unable
to
make
it
this
afternoon.
B
B
B
In
in
the
dock,
yeah.
A
A
B
A
A
B
Yeah
I
I
actually
asked
this
in
the
channel,
but
I
think
I
asked
today.
So
since
they
are
gone,
maybe
they
just
didn't
see.
It
I
noticed
that
the
release
notes
that
we
release
at
least
I
guess
I'll
just
probably
can
zoom
here
they
seem
to
only
generate
the
release.
Notes
from
the
previous,
like
non-stable
release,
yeah
wow
I'm
curious.
If
there's
a
way:
hey,
hey
Martin,.
B
No
worries
I'm,
just
curious
if,
like
we've
discussed
or
if
there's
a
mechanism
for
like
viewing
complete
release,
notes,
I,
guess
between
stable
releases,
because
right
now
the
only
way
I
could
see
to
do.
It
was
just
like
navigating
through
all
the
individual
Alpha
and
beta
and
RC
releases
and
kind
of
like
aggregating
them
myself.
A
Yeah
100,
we
were
actually
just
talking
about
this
myself
and
some
of
the
other
maintainers
I
to
just
clarify
I,
think
I.
Think
I
was
making
the
same
kind
of
comment
as
you
to
siegia
the
other
day
basically
like,
but
you
can
see
my
screen
now
right.
A
A
Yeah
yeah,
yeah,
yeah
and
so
yeah,
like
one
thing,
and
obviously
we
can
just
update
this
and
just
make
this
make
this
nicer.
It's
it
just
hasn't
raised
up
far
enough
on
my
to-do
list
yet,
but
I
I,
agree,
I,
think
coming
in
here
and
just
like
basically
giving
a
summary
of
like
all
the
things
that
happen
so
that
you
don't
have
to
go
through
all
of
these
releases
to
see
the
change
would
be
nice
for
sure.
A
C
I
am
in
the
process
of
doing
the
1.4
docs
at
the
moment.
I
don't
know
whether
anybody's
seen
that,
but
my
plan
was
to
write
a
blog
post
about
1.4
about
Lessons
Learned
and
that
kind
of
stuff
as
I
kind
of
round
up
other
stuff.
That
was
good
stuff.
That
was
done,
but
I
don't
know
whether
you
want
to
do
that
for
every
release,
but
I
think
it
does
reserve
some
narrative
each
time,
rather
than
just
a
release,
notes.
A
C
Anything
similar
there's
a
couple
that
have
done
something,
but
I
mean
it
doesn't
even
have
to
be
on
open
Telemetry.
I
o,
if
they
don't,
want
it
but
pulling
it
somewhere,
even
if
we
stick
it
in
the
repo.
C
A
B
I
just
think
from,
like
the
perspective
of
Sam
somebody
who
only
updates
between
stable
release
to
stable
release
like
when
I
see
that
new,
stable
releases
has
been
released.
I
would
love
to
just
go
and
say:
okay,
here
are
all
the
chain.
You
know
hear
all
the
changes
that
all
right
like
apply
to
my
my
upgrade
from
one
three
to
one,
four
or
whatever,
totally.
A
B
A
Yeah
yeah
I
think
it's
it's
definitely
a
little
bit
of
both
I
mean
we've
definitely
gotten
feedback
that
this
looks.
This
is
difficult
to
read.
I
know
that
there
are
other
projects
that
do
a
lot
better
than
us
and,
like
I,
think
one
that
was
cited
was
maybe
the
collector
yeah.
C
A
Do
a
pretty
good
job
I,
don't
know
how
much
of
this
is
automated.
We
actually
have
it
on
kind
of
our
to-do
to
to
to
see
if
they
have
any
automation
going
to
to
drive
this
or
whether
you
know
they
just
hand,
roll
I
mean.
C
C
If
you
think
about
what
release
notes
are
trying
to
do,
I
see
them
as
two
things
one
to
get
people
to
update
and
the
other
one
is
for
people
who
want
to
know
why
they
shouldn't
yeah
so
they're
looking
for.
C
C
Achieve
both
of
those
two
use
cases
so
yeah,
some
of
them
might
need
it.
I
mean
specifically
this
one
I
think
this
one
absolutely
needs
a
a
bit
of
a
release,
notes
digestion
from
all
the
problems
that
we
had
all
the
things
that
we've
learned,
that
we're
going
to
do
differently,
but
also
the
big
new
features
around
dependency
injection
and
the
new
interfaces
and
the
dev
exchanges
and
all
of
that
kind
of
stuff.
C
B
Want
everybody
and
they
they
do-
have
some
type
of
automation,
Alan
to
use
them
I,
don't
know
it's
like
dot,
ch
login
thing,
I
was.
If
you
go
back,
it's
like
up
one
more
directory.
It's
in
the
root,
oh.
B
Nothing
yeah
gotcha,
so
I,
don't
know
like
as
part
of
the
pr
they
like
put
in
these
files
and
then
they
get
somehow
like
rolled
up
I,
don't
really
know
how
it
works,
but
I.
A
Either
I
haven't
heard
of
it
interesting.
A
Yeah
well:
okay,
yeah
thanks
for
thanks
for
pointing
that
out,
I
we'll
have
to
look
at
this
or
I
guess
we
will
have
to
look
at
this
see
if
there's
something
we
can
borrow.
A
C
C
I
yeah
I've.
Don't
think
that's
for
me
because
I
have
a
habit
of
creating
nasty
issues,
so
I
don't
think
people
will
take
that
one
as
sarcasm
from
me.
Yeah.
A
Anyways
noted
our
release,
notes
are
in
need
of
some
TLC
and
we're
talking
about
it.
So
you
know,
maybe
maybe
what
I'll
do
in
the
meantime.
Just
for
this
last
release,
I
think
it's
great
Martin
that
you're
you're
thinking
about
a
blog
post,
but
at
least
for,
like
the
release,
notes,
you
know
short
of
short
of
any
automation
that
we.
A
C
A
A
Unless
there's
more
to
say
about
that
Dan,
do
you
want
to
talk
about
this
refreshing
provider
configuration.
B
Oh
yeah,
so
I
don't
know
if
there's
so,
okay
like
we,
we
instantiate
or
configure
whatever
you
want
to
call
it
like
the
Tracer
provider.
B
Maybe
we
specify
a
header
that
includes
something
like
an
API
key
and
we
get
a
lot
of
feedback
like
what
if
the
API
key
needs
to
roll,
what's
the
Cadence,
the
role
needs
to
roll.
What,
if
my
service
doesn't
restart,
etc,
etc.
As
far
as
I
know,
like
the
configuration
that's
passed
right
provider
is,
unlike
not
changeable,
like
after
the
provider's
been
instantiated.
A
Yeah
not
really
today,
it's
been
discussed,
but
just
to
clarify,
like
you
know
you
this
API
key
example
that
you
used
like
you're,
imagining
like
if
you
misconfigure
the
API
key
SDK,
comes
up
like
maybe
you're
using
the
otlp
exporter.
It
data
doesn't
get
transmitted
because
you
get
403s
or
whatever.
A
C
In
the
real
world
example,
is
you
stole
everything
in
key
Vault
in
Azure
and
somebody
rotates
the
Keen
key
Vault,
but
if
you
only
ever
took
it
at
startup,
you
have
to
restart
your
service
to
get
it
right
right
and
that
I
can
see
is
100
scenario.
That
I
know
has
happened.
Yeah.
B
A
Yeah
well,
there's
been
some
some
discussion,
but
nothing
of
like
much
substance,
yet
I
think
there'd
be
a
fair
amount
of
work
there.
A
lot
of
the
work
that
you
know,
Michael
did
with
the
di
options
and
configuration
kind
of
support,
I,
I
think
I
think
some
of
that
actually
set
a
good
stage
for
us
to
better
entertain
that
type
of
thing
in
the
future.
He
took
advantage
of
I
still
feel
like
I'm
a
little
bit
clumsy
with
my
knowledge
of
this
stuff,
but
he.
C
A
Of
like
the
eye
options
monitor
in
a
lot
of
places,
so
I
could
really
were
to
look
at
the
code.
Like
say,
the
I
think
the
otlp
exporter
should
be.
A
A
A
A
So,
to
my
knowledge
like
this,
this
I
options,
monitor
class
enables
I,
don't
know.
If
what
it
does,
it
must
set
up
some
sort
of
a
watcher
like
when
configuration
changes
like
you'll
get
notified,
but
the
thing
is
we're
not
actually
doing
anything
we're
not
taking
advantage
of
that
today,
like
from
this
context
of
like
the
otlp
exporter,
that
would
mean
us
to,
like
you
know,
establishing
a
new
channel.
A
C
Also
needs
to
be
passed
down
further
as
well,
so
where
it
is
there
at
the
moment,
we're
passing
down
the
value
of
it.
You'd
need
to
actually
pass
down
the
eye
options,
monitor
even
built
even
further
below
for
things
like
apis
to
use
to
be
picked
up,
you
could
have
two
different
classes
of
it,
but
the
where
we
have
it
at
the
moment.
You
can
see
there
where
it
does
current
value.
Current
value,
isn't
the
monitor
value
it's
what
the
value
was
when
he
got
it.
C
C
A
Also,
it's
it's
in
its
I
mean
this
is
kind
of
an
interesting
note
like
I,
haven't
yet
studied
this
Otep
much
myself,
but
open
telemetry
I'm,
just
gonna,
that's
not
where
I
want
to
go.
A
Yes,
yes,
that's
that's!
What
I
was
going
to
pull
up
Jack
open
that
up
it's
fairly
recently,
it's
got
to
be
one
up
some
issue.
It's
pull
request
configuration
proposal
that
was
actually
Alex
opened
it
up
at
Jack
and
him
are
I've,
been
working
for
a
while
now
on
this
yeah.
So
this
idea
of
a
file
based
configuration
might
also
be
a
a
factor
that
helps
enable
some
of
this
stuff.
A
You
know
the
ability
to
some
settings,
I
I,
don't
know
I,
don't
know
if
other
vendors
have
done
this,
but
I
know
like
the
New
Relic
agents
work
in
such
a
way
that
all
of
them
have
like
a
file
based
configuration,
and
you
can
update
a
lot
of
the
values
within
the
file
and
it'll
pick
get
picked
up
at
runtime
and
and.
A
I,
don't
I,
don't
know
how
you
comment
on
the
issue.
If
you
want
wow,
there's
already
106
comments
on
it,
but.
A
A
C
C
Was
that
was
the
first
time
when
that
I
realized
the.net
Community
is
what
it
is.
It
was
wow
they
can
come
out
with
pitchforks.
It's.
B
A
I
think
that
I
think
that's,
maybe
part
of
the
inspiration.
Also
yaml
I
mean
at
the
end
of
the
day.
Yaml
does
have
some
nice.
Nice
features
that
Json.
Doesn't
you
know
you
can
comment
things
which
is
important
to
some
people?
It's
you
know,
arguably,
maybe
a
little
bit
easier
to
read
by
as
a
human,
maybe
I
don't
know,
that's
definitely
arguable.
You.
A
Xml,
you
know
has
I,
think
fewer
people
excited
about
it,
but
yeah
XML
is
definitely
you
can
comment
and
you
can
easily
validate.
You
know
with
an
xsd.
Do.
C
But
yeah
I
think
that
that
particular
one
there
probably
doesn't
have
as
much
of
a
bearing
on
the.
How
do
we
reload
stuff.
A
Right,
no
I,
don't
I,
don't
think
this
necessarily
touches
on,
like
you
know,
Dan's
issue
of
refreshing
I.
Think,
though
it
would
it's
a
natural
extension,
though,
if
we
had
file
basic
configuration,
that
being
able
to
edit
that
file
and
and
have
the
SDK
dynamically,
you
know
adapt
to
changes
thus
far.
You
know
the
open.
A
Telemetry
specification
has
been
very
heavy
with
like
configuration
via
environment
variable
which
technically
aren't
immutable,
but
you
know
in
many
cases
are,
are,
for
all
intents
and
purposes,
so
there's
that
and
then
also
there's
the
there's,
the
op
amp
protocol
and
I,
don't
know
quite
where
that
project
is
at,
but
that's
basically
the
idea
of
of
like
a
server-side
configuration
so
like
an
SDK
reaching
out
and
asking
for
its
configuration.
C
A
B
It
came
up
in
the
and
user
discussion
last
week.
B
Can't
remember,
I,
know
you're
talking
about.
A
Anyways
not
very
satisfying
answers
for
you
on
this
today.
I
think
you
know,
if
you
specifically
for
this,
this
API
rotation
or
API
key
rotation.
You
know
how.
How
would
you
want
that
to
work
today?
If
we
were
to
like
consider
like
a
one-off
type
of
thing,.
B
Yeah
I
don't
know,
I
didn't
give
it
that
much
thought.
Maybe
maybe
I
should
just
create
an
issue
for
the
specific
use
case.
A
C
C
C
A
C
I
mean
this
comes
down
to
what
me
and
Bunch
were
talking
about
when
we
wanted
the
Di
and
options
as
part
as
a
core
part
of
all
of
this
was
that's
a
pattern.
That.Net
developers
know
it's
ingrained
into
the
way
that
configuration
Works
in.net.
Therefore,
if
you
pass
down
on
options,
if
you
register
in
options
and
you
register
in
options
based
on
some
other
options,
you
expect
all
of
the
things
to
respect
that
at
the
moment,
it
doesn't.
A
On
open
Telemetry,
I
o
yes,
I
have
not
reviewed
them.
Yeah.
C
So
this
is
a
larger
refactor
they've
merged
somebody
else's
update,
which
was
a
really
light
update
to
use
with
tracing.
C
C
So
not
just
here
you
go
install
these
packages.
They're
pre-released,
they're
pre-release,
because
things
aren't
released,
was
quite
literally
the
text
that
was
it's
pre-release
because
things
aren't
released
yet
like
okay,
the
getting
started
is
the
one:
that's
had
the
biggest
change
yeah.
This
is
the
one
that's
been.
That's
had
quite
a
big
refactor.
C
So
what
we've
been?
What
I've
been?
Basically
doing?
If
you
look
at
the
lines
21
there,
what
we
were
doing
previously
was
install
all
of
the
packages
that
are
pre-released
all
the
instrumentation
packages.
C
Now,
what
I've
done
is
I've
split
it
into
two
things:
install
the
hosting
package,
which
is
not
pre-release,
then
add
the
instrumentation
passage,
the
automatic
instrumentation
package
for
asp.net
core,
which
is
pre-release,
and
the
reason
why
it's
pre-release,
which
is
37
through
43.,
so
that
people
can
see
the
specific
differences
and
the
reasons
why
so
I've
done
that
throughout
pretty
much
the
entire
file
so
split
out
the
stuff,
that's
pre-released
from
not
so
that
it's
two
explicit
commands
that
we
can
explain.
C
I've
made
the
examples,
a
lot
more
succinct,
I'm,
going
to
move
up,
that
configuration
resource,
configure
resource
on
63
up
to
the
main,
open,
Telemetry,
one
that
is
in
the
repo,
but
not
in
this
I-
have
decided
not
to
take
the
feedback
of
use
dependency
injection
for
everything
when
it
gets
down
to
the
manual
instrumentation
section.
C
So
what
that's
actually
doing
at
the
moment
is
using
a
static
for
everything
right
now
here:
yeah
diagnostic
config,
there's
a
diagnostic
config
class.
That's
static
in
the
program
CS
that
you
can
call
I've,
not
done
it
as
one
big
wall
of
text
for
single
for
minimal
apis
to
try
and
make
it
a
little
bit
more
readable
for
people,
because
what
we
were
ending
up
with
was
basically
big
walls
of
text
that
were
single
file.
C
Applications
which
is
great
for
I,
don't
think
it
helps
the
people
who
need
help
at
this
stage,
because
the
people
who
are
working
on
minimal
apis
really
understand
minimal
apis,
and
if
they
do,
then
yeah
they're,
not
they're,
going
to
need
less
onboarding.
C
There
was
a
comment
from
cgo
about
removing
the
manual
instrumentation
from
this
whole
thing
me
and
Philip
Carter
disagree
in
that
a
it's
quite
small
in
its
implementation
and
B.
That
is
the
key
part
of
it
that
people
can
do
that
manual
instrumentation.
C
C
So
the
example
that
you
see
here
the
ones
that
you
see
in
the
examples
folder
was
it
you
Danny
who
went
through
the
yeah.
B
And
this
isn't
that
comment:
I
left
on
your
that
when
you
posted
recent
Slack
but
yeah,
that's
exactly
yeah,
I
I,
agree,
I,
don't
know
if
it's
a
problem
per
se,
but
it
it
is
different.
B
Maybe
it's
good,
though,
to
have
both
examples
also
and
be
like
you
can
use
Statics
if
you
want,
or
you
can
follow
this
di
pattern.
If
you
want
yeah.
C
C
B
B
I
mean
it
wasn't
a
problem,
it
was
just
you
know,
I
think,
there's
just
a
few
ways
to
do
this.
Obviously
so
it's
yeah,
maybe
it's
good,
to
have
examples
of
you,
know
different
ways
and
then
people
can
choose
the
the
methodology
that
you
know
they
agree
with
the
most.
C
Yeah
well
I
wanted
to
make
the
the
docs
the
simplest
possible
thing
that
they
could
be
and
I.
Think
bringing
independency
injection
in
that
way
complicates
things
because
now
I've
got
to
have
a
class
with
a
Constructor
for
my
controllers
that
injects
the
thing
pulls
them
out,
which
I,
don't
think
is
part
of
getting
started.
C
So
yeah
that
this
one
is
a
a
larger
refactor
but
I
think
it
gives
people
just
the
things
they
need
at
the
time
when
they
need
them,
rather
than
big
walls
of
text
with
VAR
service
name
in
them.
Foreign.
C
Would
ever
build
these
things
yeah,
but
you
can
see
where
we
do
tracing
there
now,
what
we're
doing
is
bringing
Trace
in
here's
the
packages
that
are
not
pre-release
now
bring
in
asp.net
core,
which
is
pre-release,
and
here's
a
no
on.
Why,
specifically,
that
is
pre-release.
A
Yeah
no
I,
like
it
I
think
this
is
a
great
Improvement.
Also
it's
just
exciting
to
see
all
this
one
for
work
highlighted
and
nicely
documented.
C
I've
just
got
to
answer
those
couple
of
comments,
but
I
just
need
you
guys
to
sign
off
on
it
and
that
you're
happy
with
it
and
then
we'll
get
that
one
out,
there's
more
stuff
to
be
done
on
the
docs
talking
to
Philip
about
some
of
the
things
that
we
can
bring
in
and
put
more
documentation.
Essentially,
what
we
need
to
do
is
take
the
documentation
out
of
the
repo
and
into
the
doc
site,
because
that's
where
people
are
looking
they're,
not
looking
at
the
GitHub
for
troubleshooting,
so
things
like.
C
Why
doesn't?
Why
do
my
dependencies
not
work
together?
We
need
a
troubleshooting
page
on
open
Telemetry.
I
o
not
on
the
GitHub
repo,
because
the
sort
of
people
who
really
need
that
information
are
not
going
to
go
to
the
GitHub
repo
and
find
it
okay.
So
I
need
to
put
some
of
that
in
there
for
all
of
that
comes
secondary
to
just
getting
a
decent
guide.
C
And
I'm
volunteering
to
look
after
the
docks,
because
I
act
at
the
same
level
as
Blanche
can
in
terms
of
writing
the
code.
But
I
can
write
a
decent
piece
of
documentation.
A
C
And
this
the
next
one
was
me
as
well
Okay,
so
we
have
a
package
which
is
the
otlp
exporter
for
logs,
yes,
because
logs
is
not
signed
off
yet
yes,
however,
we've
also
put
other
things
in
like
the
console
exporter
we'll
take
logs,
so
the
console
export.
So
if
you
go
into
the
console
exporter.
C
Console
exporter.logs
right,
it's
like
console!
If
you
look
the
tag,
Transformer,
there's
a
log
record
export
there
you
go
that
exists
inside
the
console
exporter.
A
A
C
A
We
we
I
actually
I'm,
actually
the
one
that
did
the
original
cut
of
the
TLP
Lodge
at
quarter
some
some
number
of
years
ago
or
like
it
feels
like
two
years
ago,
but
no,
we
did
it
because
at
that
point
in
time,
right
like
it
was
kind
of
a
weird
point
in
time.net
was
really
the
first
language
and
I
actually
is
still
the
only
language
that
declares
some
sort
of
like
stable
logging
support,
even
though
that
was
all
done
at
a
time
before
any
of
the
of
the
spec
for
logs,
but
stable,
the
data
model
wasn't
even
stable
when
when.net
did
this.
A
A
If
you
go
to
like
any
of
the
other
languages,
none
of
their
logging
components,
API
SDK
or
you
know
much
less,
like
the
exporters
like
like
the
otlp
logging
exporter,
and
none
of
them
are
stable
across
any
of
the
other
languages
that
I'm
aware
of,
but
This
was
done.
Primarily
at
a
time
when
you
know,
maybe
maybe
we
can
reconsider
it,
but
the
Proto
right
is
what
governs
the
open
telemetry
protocol.
A
A
We
can
discuss
I'd,
be
curious
to
see
what
other
people's
opinions
are
separate
from
that,
though
I
don't
know,
I
think
maybe
you're
aware
Martin
that
for
a
while
now,
Blanche
has
had
this
logs
Branch
out
there,
okay
tracking,
along
with
the
stability
of
the
log
API
SDK
at
the
spec
level,
which
is
hopefully
imminent,
just
to
kind
of
like
bring
the
spec
issue
up.
C
A
Well,
that
was
November
number
of
months
have
gone
by
and
I
think
it's
gotten
pretty
close
so
like
I
have
not
really
been
leading
this
effort,
but
I
did
start
this
issue
and
I
started
the
punch
list
of
of
things
and
and
I
actually
did
along
with
my
colleague
Jack
we
we
went
through
a
couple
of
just
the
the
punch
list
items
and
on
the
spec
and
really
you
know
we
have
prototypes
now
in
at
least
three
different
languages:
java.net
python
of
of
log
appenders,
so
like.net,
you
know,
Blanche
did
the
one
for
Syria
log
did
the
one
for
Event
Source.
A
We
already
had
ilogger
support,
but
anyways,
that's
necessary
prototypes
and
languages
for
stabilizing
or
declaring
spec
stable,
so
we're
tracking
well
there
and
then
really
it's
it's
Jack
and
and
tigran
who
have
been
kind
of
the
the
leaders
of
the
logsig.
It's
on
them
to
basically
present
the
the
specification
to
the
larger
Community,
both
at
the
maintainers
and
the
specification
meeting.
I.
Think
that's
going
to
come
in
like
the
next
week
or
so.
C
A
Hope,
and
with
that,
if
things
progress
well
and
the
the
broader
Community
signs
off
on
this
and
the
spec
is
stabilized,
then
some
of
this
kind
of
becomes
may
become
a
moot
point
in
in
the
nearest
term,
because
blanch
will
log
will
will
merge
in
his
main
logs
branch,
which
is
probably
wildly
out
of
date
with
Maine
at
this
point
and
we'll
hopefully
quickly
begin
to
stabilize
the
the
logs
work,
thereby,
you
know
rolling
in
like
the
logs
otlp
exporter
into
the
main
package,
which
has
always
been
the
the
plan.
It's
just
I.
C
Being
the
trigger
point
for
merging
those
two
for
all
the
other
stuff,
yes,
but
for
merging
those
two
I'm,
not
convinced
that
there
is
a
a
reason
to
stop
it.
For
two
reasons:
a
data
model
is
fine.
Now,
so
the
protos
are,
you
know
pretty
solid,
but
the
other
side
of
it
is
well.
We've
already
got
precedent
because
we're
doing
it
in
the
console
export.
C
So
if
we're
already
doing
it,
Why
does
adding
one
thing
that
works
differently
than
the
others,
so
where
this
is
coming
from,
I
have
a
I'm
doing
a
talk
next
week
for
a.net
User
Group
called
practical,
open,
telemetry
for.net,
which
is
all
of
the
things
in.net,
and
how
do
you
do
them
and
also
what
are
they
useful
for?
So
what
would
you
use
span?
C
Events
for
what
would
you
use
span,
links
for
what
types
of
metrics
are
available
and
what
would
they
be
useful
for
and
how
would
you
use
them,
but
obviously
also
includes
logs,
because
that's
the
main
question
I
get
asked
these
types
of
things
so
when
I
went
to
try
and
do
it
and
I'm
like
oh
I-
need
to
install
this
other
pre-release
package
about
logs
for
the
logs
exporter,
but
everything
else
appears
to
be
stable
to
everyone
else,
and
that
was
just
a
little
bit
of
an
irk.
C
If
you
like,
when
I
was
trying
to
do
it
and
trying
to
write
the
slides
to
present
it
so
yeah
it
just
if
I,
if
it's
like
that
for
me,
I
can
only
imagine
what
it's
like
for
other
people
that
aren't
as
close
to
it
so
merging
that
one
in
would
be
a
nice
first
step
and
I.
Don't
think.
That's
based
on
all
the
other
things.
C
Yeah
and
I
will
drop
a
link
in
here
for
further
reading
for
people
something
I
put
together
today.
C
That
was
on
my
blog,
because
it's
come
up
quite
a
lot.
I
mean
you
understand
that
you've
heard
of
the
12
Factor
app
methodology.
C
That
kubernetes
Can
Read
from
logs.
C
C
In
series
you
know,
you're
gonna
have
to
do
this
in
a
particular
way.
It's
not
really
going
to
work
so
don't
use
a
console
export,
so
I
wrote
that
thing
about
logs
and
how
how
and
why
12
Factor
was
a
thing.
C
C
C
C
The
zoom
but
yeah
I
did
that
just
about
the
log
stuff,
because
I
was
all
on
the
logs
type
thing
and
console
locked,
but
I
am
also
looking
at
a
console
exporter,
formatter
mm-hmm,
because
I
think
our
console
export
format
right
now
isn't
amazing.
A
There
has
been
some
I,
don't
know
if
we
have
an
issue
already,
if
we
don't
absolutely
open
it,
but
we
have
talked
about
something
like
that.
In
the
past,
I
mean
I,
don't
even
know
what
search
for
a
console
format,
something.
A
No,
but
one
thing
that
I
I
think
that
maybe
some
of
the
other
languages
that
have
done
is
to
have
a
configuration
setting
for
logging
out
in
the
otlp
Json
format,
which
I
think
would
be
a
natural
choice
for
for
additional
format
for
the
console
exporter,
in
fact,
maybe
the
default
Choice
rather
than
you
know,
I,
don't
disagree
with
you
Martin
that
our
format
is
not
phenomenal.
C
C
Because
that's
kind
of
exposing
an
implementation
detail
but
yeah
being
able
to
say
if
you
look
at
how
serolog
works
and
that's
you
know,
elephant
in
the
room,
that's
what
we're
competing
against
for
adoption,
we're
competing
in
the
log
space
for
sorry
logs
market
share
and
dot
net
developers
are
sticking
their
ways.
C
So
we've
got
an
uphill
battle
already
and
if
we
don't
allow
at
least
a
decent
colorized
output
for
logs,
which
already
have
the
message
format
then
we're
on
the
back
foot.
What
we
do
for
the
others
could
be
very
interesting.
C
So
what
we
do
for
metrics
and
how
we
log
metrics
out
same
with
traces,
I,
don't
know,
I
have
built
a
trace
view,
a
local
Trace,
View
and
back
to
console,
which
is
pretty
cool
but
I.
Don't
think
that
would
ever
be
part
of
this,
so
so
yeah
I'm
going
to
create
an
issue
for
it,
because
I
think
we
need
to
have
a
think
about
it.
I
don't
mind
doing
the
work
on
it,
because
it's
quite
interesting
and
I
think
I'm
not
going
to
break
anything
by
doing
it.
C
Whereas
you
know
working
in
the
powers
of
the
tracing
provider,
I
could
probably
break
a
lot
of
things,
but
yeah
I,
just
I
thought
I'd
mention
that
I'm
gonna
do
something
like
that
cool.
C
But
this
is
all
coming
about.
How
do
we
do
onboarding
and
I
hope?
You
know
I've
been
consistent
in
the
the
ways
I've
been
talking
about
this.
This
is
all
about
how
what's
the
on-ramp
for
people.
How
do
we
get
people
on
board
quickly
and
that's
what
I'm
trying
to
push
with
everybody
to
understand
that
we
need
to
make
that
easy
for
people
and
at
the
moment,
It
Ain't
Easy.
A
Yeah
I
agree
in
fact
just
a
little
bit
of
tangent.
You
know:
I
think
that
that
was
actually
it
was
that
General
theme,
I,
think
which
was
guiding
a
lot
of
Noah
and
David's
comments
on
your
PR
Dan,
which
is
that
that
is
definitely
on
their
mind
as
well.
A
I
know,
David
has
given
a
number
of
demonstrations
using
open
Telemetry,
so
he's
had
his
hands
on
it
and
found
it
to
be.
You
know
a
little
rough
around
the
edges,
usability,
wise
onboarding
type
of
thing,
and
so
he's
he's
beginning
to
come
in
with
his
own
opinions.
So
I
think
I
think
this
is
a
a
ripe
space
for
us
to
continue.
A
C
I
mean
the
dogs,
the
dogs
PR
that
I've
got
in
there.
That's
you
know,
step
one,
let's
document
what
we
have
in
a
better
way,
but
then
step
two
is
definitely
about
document
all
the
rest
of
the
things
that
we
haven't
documented.
Yet
the
advanced
use
cases
the
interesting
use
cases,
the
opinionated
use
cases
get
all
those
done,
and
then
how
do
we
make
this
on
ramp
a
little
bit
easier?
And
how
do
we
bring
that
into
everything
that
we're
doing
going
forward?
C
So
when
we
talk
about
exemplars,
when
we
talk
about
histograms,
how
do
we
make
those
easy
for
people
to
do
and
understand,
and
how
do
we
make
sure
there's
docs
lined
up
for
it
as
well,
but
that's
the
thing
I
wanted
to
mention
when
there's
quite
a
few
more
of
us
on
here
that
I
think
we
need
APR
ready
to
go
for
the
next
big
feature
we
had,
it
can't
be
a
it
can't
be
a
secondary
thing.
C
The
main
doc
site
needs
to
be
ready
to
go
on.
Our
1.4
release
went
out
which,
luckily
it
was
because
somebody
had
already
done
a
PR,
because
the
hotel
demo
was
broken
so
anyway,
yeah
just
about
out.
A
Of
time
yeah
well,
this
is
this
is
good.
This
is
good
work
and
thanks
for
sharing
the
the
blog
post.
Oh
look
at
that.
One
too
you've.
A
A
The
the
stress
tests
that
we
have,
where
are
they
on
this
bill,
I,
think
we
have
these
stress
test.
Suites,
okay,
I'd,
be
curious
to
see
like
how
many
you.
C
A
The
benchmarks
folder
I
I
I
get
is
what
inspired
which
you've
written
up
so
far.
The
stress
tests
I
know
that
one
thing
that
sijo
was
really
focused
on
when
we
were
originally
doing.
The
metrics
work
was
like
operations
per
second
with
and
without
open
Telemetry,
and
he
was
using
this
this
stress
test
for
for
metrics
and
gauging
gauging
things
with
that.
Again.
As
I
said
myself,
I
probably
should
you'd
want
to.
C
Look
at
yeah
I
can
I
can
have
a
look
at
those,
and
you
know
the
the
whole
point
of
that
post
was
pretty
much
the
same
as
what
that
seems
like.
It's,
therefore,
which
is
to
give
people
an
idea,
a
very
granular
level,
what
it's
going
to
cost
them
for
implementing
open
Telemetry
in
their
application.
C
Is
it
more
memory?
Is
it
going
to
slow
down
my
requests,
because
literally
every
single
conference
that
I've
been
at
talking
about
this
stuff
that
comes
up
yeah?
Doesn't
it
make
my
system
slower
to
add
open,
telemetry
and
I'm
like
well?
No,
he
doesn't
I
mean
technically
it
does,
which
is
what
the
post
says.
It's
like.
Yes,
it
does,
but
if
you
optimize
it
at
that
level,
probably
don't
use.net.
B
C
You're
worried
about
40
nanoseconds
I'm,
not
sure
done.
That's
for
you.
A
Yeah
I
think
no
I
I've
gotten
this
question
too.
A
lot
and
I
I
really
appreciate
that
you're
you're
putting
out
some
I
like
this
blog
post
for
I.
Guess
it's
something
we'll
be
able
to
point
folks
too.
It's
a
concern.
Another
big
aspect
of
it
is
that
it's
and
I
think
you
address
this
in
your
in
your
dock.
Is
it's
totally
dependent
on
the
work
workload?
A
You
know
if
you're
I
think
you
I
think
you
address
this
like
if
you're,
if
you're
creating
a
lot
of
activities
per
request
right
then
maybe
you're
shooting
yourself
in
the
foot
or
you're
you're,
trying
to
trace
things
at
a
finer
grain
level
than
you
really
need
to,
but
or
maybe
you
have
processors
that
are
doing
some
really
heavy
weight,
inefficient
work.
You
know
custom
processors
like
it's
an
SDK,
so
you
can.
You
can
do
Parts
things.
C
Yeah
I
mean
we
had
somebody
recently
that
was
doing
some
stuff
in
a
processor
to
do
with
the
user
context
and
the
user
context
was
being
blocked,
and
that
meant
the
processor
was
running
for
about
half
a
second
every
time
yeah.
Until
the
other
thing
was,
the
the
lock
was
dropped
and
people
were
complaining
that
they
got
us
pause
on
their
website.
C
It's
like
well
you're,
doing
bad
things,
stop
doing
bad
things
and
copy
the
object
out,
or
do
that
in
a
different
context
like
that.
That's
not
use
a
volatile
read
whatever
the
hell.
That
means,
but
yeah
use
all
of
the
things,
but
if
you've
got
something
that's
running
on
the
end
of
every
dispose
of
every
activity
and
it
takes
five
milliseconds
to
do
it-
that's
going
to
add
to
your
response
times
so,
but
that's
the
things
that
people
don't
understand.
C
C
There's
the
New
Sentry
one
which
has
its
own
processor,
which
calls
up
stack
frames
as
it.
It's
a
processor
four
stack
frames
so
that
they
can
get
their
exception
stuff,
which
is
fine
until
you
understand
what
the
apps
they
know.
C
That's
a
big
problem
for
performance,
inline,
request
performance
right,
so
I'm
trying
to
educate
people
but
I'm
most
trying
to
give
people
a
or
like
people
like
us,
a
place
to
point
to,
which
is
why
I'm
doing
it
non-vendored
it's
not
going
on
the
honeycomb
blog
it's
to
go
on
cncf,
so
we
can
all
point
to
it,
which
is
why
I
want
you
and
cjo
and
people
like
that
to
weigh
in
on
it.
A
Yeah
I
I,
like
I,
like
it
I,
think
I,
know
you're
really
focused
on
like
the
in-line
performance,
but
you
know:
there's,
let's
see
Blanche
and
I
have
actually
been
talking
or
we
have
talked
in
the
past
about
tuning
the
otlp
expert.
He
spent
a
lot
of
time
a
while
back
tuning
the
memory,
consumption
of
of
the
Zipkin
and
Jaeger
exporters.
A
There's
work
that
we
can
do
on
the
on
the
hotel
or
the
otlp
exporter
to
further
reduce
its
memory
Footprints.
Even
in
like
batch
batch
kind
of
situations,
you
know
we
can,
we
can
reduce.
There
are
things
we
can
do
to
reduce
the
number
of
allocations,
which,
ultimately,
you
know
has
has
an
impact
on
a
workload.
You
know
in
terms
of
GC
and
ultimate
like
CPU,
which.
C
A
A
A
Yeah,
technically,
that's
that's,
you
know
any
of
the
any
of
the
signals.
Metrics
and
traces
can
also
have
objects
as
tag
values
and
we
do
hold
on
to
them
and
we
to
string
them.
C
A
Let's
see
yeah
bro,
that's
good,
but
it's
the
exporter
that
does
the
actual
transform.
Now
it's
the
exporter,
so
so
the
they'll.
A
The
I
have
this
like
tag
transform
thing,
but
all
the
exporters
have
their
own
transform
because
they
have
to
transform
to
like
their
own
their
own
native
format.
So.
C
Yeah
I
mean
we
can
do
a
series
of
these
things
and
then
we
can.
My
plan
was
we'll:
do
the
series
of
them
there'll
be
blog
posts
and
then
ultimately
we
can
then
take
those
and
make
them
in
once
we've
got
a
doc
section
for
them.
We
can
have
a
performance
dock
that
we
can
point
to,
but
at
the
moment
let's
just
get
some
blog
posts
out,
because
blog
posts
are
more
ephemeral
and
because
they're
more
ephemeral,
we
can
be
a
little
bit
looser
with
them
with
the
doc
stuff.
C
So
but
yeah
there's
a
whole
host
of
whether
it's
memory
stuff
all
that
kind
of
stuff,
but
people
really
what
people
really
care
about
when
they
talk
to
me
about
it
is.
Is
it
going
to
slow
down
my
requests
because
they're
running
in
kubernetes
they've
got
a
ton
of
extra
memory
that
they
can
wait?
They
can
be
working
with
what
they
really
care
about
is
CPU
cycles
and
the
request
performance.
A
C
Think
that's
then
we
can
start
adding
more.
You
know
this
is
a
you
know.
The
the
memory
consumption
stuff
is
a
really
interesting
one
and
I
think
that's
probably
something
we
should
tackle.
C
I've
only
got
so
much
time.
I
can
spend
on
these
things,
but
let's
do
another
one
on
memory,
but
if
you're
happy
with
that
post
I'll
pass
it
on
to
see
Joe
next
and
I've
been
doing
it
in
stages.
Yeah.
C
Awesome
anyway,
I'll,
let
you
go
from
my
side,
though,
if
you
can
have
a
look
at
the
docs
PR
same
with
you
Dan.
If
you
have
another
look
at
it
now,
it's
you
know
pretty
ready
now,
with
all
the
other
sections
value
your
feedback
as
soon
as
you
went
through
the
ringer,
with
the
with
David
and
Damien.
B
C
Awesome
and
hopefully
we'll
get
that
out
in
the
next
week
or
so.