►
From YouTube: CNCF Open Telemetry Community Meeting 2020-01-22
Description
CNCF Open Telemetry Community Meeting 2020-01-22
B
B
C
C
C
C
One
thing
I
noticed
by
kind
of
looking
through
there
is
that
these
were
called
like
tracer
Factory
and
meter
factory
and
they're,
calling
them
traits
of
provider
and
meter
provider.
It's
more
of
a
language
agnostic
way
of
referring
to
these
things.
So
my
guess
is
at
some
point
we
might
want
to
follow
suit.
I've
I've
at
least
heard
some
some
disgruntlement
about
having
the
name
factory
in
our
code
base,
so
I.
C
D
C
C
C
The
project
timeline
and
milestones.
Basically,
there
is
an
issue
that
specifies
what
what
the
milestones
are,
what
should
be
in
each
one
and
they're
like
woefully
out
of
date,
like
so
many
other
things,
so
there
was
a
request
that
hey.
Maybe
we
should
update
this
and
make
it
look
a
little
bit
more
like
reality,
just
for
four
people,
sanity
I
think
there
are
a
lot
of
people
who
are
like
working
on
this
stuff.
C
C
C
C
C
C
And
the
whole
purpose
of
those
things
is
they're,
just
like
endless
issues.
I
think
there
are
at
least
like
four
pages
of
issues
on
the
spec,
depo
and
I.
Think
we
made
it
through
like
one
page
today,
which
is
good
and
we're
trying
to
burn
those
things
out
and
get
this
V
0,
3
kind
of
nailed
down
and
then
kind
of
know
what
we
need
to
do
for
these
0-4.
C
Think
if
you
asked
about
this
in
your
rack,
PR
is
like
what
I
knew
about
errors
and
I
mentioned:
hey,
there's
some
semantic
conventions
for
those.
Let
me
just
pull
those
up
and
then,
as
I
went
dick
mnemonic.
Where
are
the
semantic
conventions
for
errors,
so
they
disappeared
and
I
know
they
were
there
because
we
originally
copied
over
like
the
open
tracing
ones.
The
open
tracing
ones
had
semantic
conventions
for
what
they
call
logs,
which
are
events
in
open
telemetry.
So
it's
standard
names
and
then
standard
attribute
names
for
them,
but
yeah
it
turns
out.
C
There's
there's
problems:
and/or,
there's
bigger
problems
that
led
to
that
stuff
getting
removed.
There
is
this
guy
Kevin
Brock
Hoff,
who
has
an
odd
tab,
opened
about
some
of
these
things
and
he's
specify
he
wants
kind
of
typed
event,
a
special
event,
type
for
errors
and
I.
Think
maybe
that's
gonna
be
necessary.
Just
because
of
well,
probably
mainly
like
stack
trace,
stack
trace,
is
a
thing
that
doesn't
neatly
fit
into.
C
D
The
array
valued
thing
was
added:
well
the
example
that
I
used
to
motivate
that
was
memcache
multi
gets
where
you
want
to
include
all
of
the
keys
as
an
attribute,
and
otherwise
you
need
some
kind
of
semantics
conventions
around
like.
How
do
you
group
these
things?
Do
you
you
know
comma
separate
them,
or
what
do
you
do
so
array?
Valued
attributes
was
a
fix
to
that
yeah
I
think
I
could
be
leveraged
for
four
stack
traces
as
well
stack,
trace
support
is
being
removed
from
the
open
telemetry
protocol.
C
Yeah
yeah,
so
that
just
occurred
to
me
so
I
might
might
propose
that
in
some
way,
or
at
least
try
to
get
in
touch
with
this
guy
Kevin,
who
has
been
kind
of
leading
the
charge
on
this
stuff
and
just
kind
of
saying.
Maybe
this
is
like
an
easier
way
around.
It.
I
think
yeah,
I,
forget
I,
know:
attributes
have
like
a
length
limit
of
like
256
characters
or
something
like
that.
255
one
of
those
numbers.
C
But
yeah
I
think
the
whole
point
of
saying
that
is
I've
at
least
realized
that
errors
are
a
thing
that
are
missing
and
I
feel
like
we
really
can't
have
a
1.0
without
them.
At
least
that's
that's.
My
take
like
I
almost
feel
like
you
know,
users
of
telemetry.
If,
if
you
gave
him
a
choice,
it's
like
you
know
we
can
give
you
like
your
traces
of
where
we
can
give
you
your
errors.
C
D
C
C
There
is
that,
and
there's
kind
of
this
hole
discussion
about
how
that
doesn't
fit,
doesn't
fit
all
use
cases,
but
I
think
the
status
can
help
indicate
that
something
went
wrong,
which
is
great,
but
we
need
to
find
ways
to,
like
you
know,
document
exactly
what
happened,
preferably
like
in
the
air
class,
the
error
message
and
a
stack
trace
of
the
thing
that
went
wrong,
but
yeah
stand
up
status
being
that
enumeration
of
T
RPC
codes.
I
think
is
another
swamp
of
sorts.
C
So
yeah
I
guess
for
now
for
instrumentation
I,
don't
know
exactly
what
to
recommend
other
than
we
should
totally
have
to
do.
X'
about
this.
You
should
definitely
anything
that
is
like
anything
you
see
in
the
data
dog
code
or
whether
it's
like
a
rescue
clause,
and
there
is
you,
know,
documenting
the
error
on
the
span.
I
would
say,
keep
the
clause
and
just
added
to
do.
We
need
to
document
this
error
when
semantic
conventions
come
and
that
way
we
don't
miss.
Something
and
it'll
be
easy
to
kind
of
fill
in
the
blanks.
C
C
C
D
C
G
E
C
Cool
so
yesterday
I
merged
this
PR
for
the
instrumentation
installer.
Thanks
for
all
the
reviews
on
that
and
then
I
opened
another
PR
for
preliminary
SDK
config
I
kind
of
had
this
other
draft
SDK
config
and
it
was
like
configure
all
these
things,
some
of
which
don't
yet
exist,
and
then
I
kind
of
pared
it
down
to
these
things
that
actually
exist
because
I
feel
like
that's.
This
is
kind
of
like
the
the
loose
end
from
the
auto
instrumentation
stuff.
So
this
is
how
you
kind
of
can
interact
with
that
and
install
instrumentation.
C
So
I
think
Francis
have
already
left
a
handful
of
comments
here
and
they
were
pretty
good
ones
like.
Should
we
be
able
to
set
the
tracer
factory
and
as
I
thought
about
that,
I
was
like.
You
know
that
doesn't
make
a
lot
of
sense,
actually
I
think
like
if
you're
using
the
default,
SDK
you're
gonna
use
the
tracer
factory
from
that
SDK.
C
C
C
F
C
Right
we're
some
eyes
at
some
point
and
then
there's
a
draft
for
rack,
instrumentation
I.
Think
a
few
of
us
have
left
some
comments
there
in
general.
I
think
I
had
a
lot
of
comments
about
trying
to
like
not
repeatedly
create
strings
and
and
other
things
so
I
think
yeah
I
think
it's
worth
just
kind
of
like
mentioning
this
and
maybe
even
documenting
this.
The
porting
guide
that.
C
Instrumentation
should
strive
to
minimize
allocations
as
much
as
possible
and
yeah
like
I,
guess
it.
It
is
like
a
different
mindset
like
writing,
instrumentation
and
I'm
thinking
back
from
when
I
transitioned,
like
you
know,
into
this
monitoring
and
observability
field
from
just
kind
of
being
like
a
a
a
ruby
developer,
and
it
did
it
didn't.
Take
it
kind
of
like
this
mental
shift
and
I
felt
like
a
child
like
in
many
ways
like
the
Ruby
that
I
ended
up
having
to
write
in
the
monitoring
world.
C
It
wasn't
as
beautiful
and
wasn't
as
nice
and
concise
like
it's,
sometimes
ugly
and
crappy,
but
with
with
good
reason
so
I
think
yeah
I
think.
Ultimately,
I
became
like
a
little
bit
more
disappointed
in
Ruby
itself,
which
I
think
is
a
true
a
true
sign
of
actually
just
getting
more
on
under
the
hood
with
a
lot
of
stuff.
It's
like
I,
don't
know.
C
D
F
C
H
F
C
If
you,
if
you
are
regularly
creating
a
lot
of
garbage,
you
end
up
with
ultimately
a
larger
kind
of
standing
heap
size
and,
like
users,
tend
to
notice
this,
you
have
like
a
lot
of
people.
You
know
they're
really
trying
to
maximize
their
instances.
So
you
know
people
who
are
using
like
cloud
providers
or
whatever
like.
If
you
they
run
their
app
without
your
tool
and
then
they
add
your
tool
like
sometimes
it
bumps
bumps
up
the
memory
usage
like
on
Heroku
or
something
they're
instances
just
start
getting
like
killed
and.
B
Right
if
I've
had
some
fragmentation
issues-
and
you
know
a
couple
of
years
ago
and
so
I
revisited
those
I
found
a
really
nice
blog.
So
basically,
the
the
the
Ruby
runtime
just
isn't
releasing
memory
back
to
the
OS
is
one
of
the
main
problems
and
there's
like
a
really
easy
fix.
Apparently
that
needs
to
be
broadcast.
Everybody
like
hey
this
joyful
bike,
shedding.
B
F
F
C
It
is
this
one
line
missing
that
it
has
maybe
made
it
into
more
recent
movies.
If
not
I,
don't
know,
somebody
needs
to
to
open
a
pr2.
Ruby.
Add
that
in
what
yeah
like
I
know,
there
are
other
issues
such
as
just
like
sheep
fragmentation,
and
you
know
if,
if
pages
are
not
completely
empty,
if
there's
a
single
object
on
them,
they
will
never
ever
get
freed.
Unless
you
know
the
object
is,
is
removed
from
the
page.
So
that's
right.
B
E
C
C
C
C
That
was
applicable,
and
that
is
right
now
we're
kind
of
we're
using
we're,
passing
config
to
adapters
as
hashes
and
I
think.
This
is
like
a
really
pragmatic
way
to
do
stuff
right
now,
I
think
longer
term,
having
a
slightly
more
sophisticated
config
object
might
not
be
a
bad
idea
and
I
think
it
should
be
able
to
be
initialized
from
a
hash,
but
I
think
data
dog
has
a
pretty
interesting
config
class,
where
you
can
kind
of
specify.
C
You
know
the
the
name
of
a
thing
like
a
type
for
it,
and
maybe
it
would
be
nice
if
you
can
even
like
specify
like
a
coercion
for
a
thing
so
like
if
you
expect
something
to
be
an
array,
but
somebody
might
want
to
give
you
a
single
value
or
array.
It
might
be
nice
to
be
able
to
kind
of
document
that
this
thing
is
expected,
or
it
should
be
coarse
to
an
array,
no
matter
what
you
get
so
like.
C
C
F
F
Okay,
yeah
I
think
there's
a
in
terms
of
inspiration.
Then
I
don't
know
if
you've
looked
at
any
of
the
dry,
RB
gems
or
the
way
those
folks
write
stuff,
but
they
have
a
gem
called
dry
configurable,
which
is
really
nice
and
hat
and
probably
has
some
of
the
internal
guts
that
are
battle
tested.
So
even
if
we
wanted
to
you
know
internalize
some
of
that
work,
we
could
do
that,
but
that
might
be
you
know
when
we're
looking
at
it.
F
Sometimes
I
worry.
You
know
when
we're
doing
that
we
ended
up
with
kind
of
our
half-baked
version
of
a
thing,
not
us,
but
anytime
I
say
you
know
what
I
can't
bring
in
that
gym,
because
it's
got
the
stuff
that
I
tried
to
do
it
myself,
like
yeah.
So
anyway,
it's
just
if
we're
looking
for
inspiration,
it
is
open
source.
It
is
something
that
that's
been
around
for
a
while,
so
it
might
be
worth
taking
a
little
inspiration
from
yeah.
C
C
F
Yeah
I'm
glad
you
looked
at
him
I,
just
it's
something
I,
always
think
about
when
I
find
myself
going,
you
know
I
want
to
maybe
I'll
just
you
know
like
you,
I've
been
following
someone
good
stuff
for
years
and
years
and
watching
on
the
romm
stuff.
Everything
so
I
think
pretty
highly
of
this
stuff,
but
I
totally
understand.
C
C
C
I
also
hope
so.
I
also
understand
that
this
thing
is
ridiculous
and
we
definitely
need
to
yeah
I,
think
whatever
I
can
do
to
help
out
like
getting
this
stuff
to
do
and
definitely
I
appreciate
the
comments
that
have
come
in
so
far
and
I
understand
it's
just
kind
of
like
this
iterative
process
of
like
chipping
into
the
problem
and
understanding
like
what
exactly
has
happened
so.
C
C
F
B
Okay
and
then
there
were
I
thought
there
were
a
couple
more
to
do's.
Besides
the
error
handling
off
to
take
a
second
look
and
see
if
there
was
another
big
piece,
we
were
missing,
I,
don't
know
those
off
the
top
of
my
head.
I
have
to
take
a
look,
but
I
think
the
error
handling
was
the
big
one.
That
I
had
a
question
on,
and
then
we
can
bring
that
out
of
out
of
draft
I.
B
D
B
B
I
B
Mine,
just
for
is
quantization
and
too
happy
of
the
adapter,
2
or
3
of
the
adapters
was
only
in
a
couple
of
places,
but
this
was
one
of
them
trying
to
strip
off
the
parameter
sleeve.
The
query
parameter
on
soon
I
didn't
know
if
that
would
be
an
operation
of
an
exporter
or
some
other
piece
in
the
pipeline
or
whether
the
adapters
responsibility
to
be
the
quantization
span.
Yeah.
F
F
They
make
some
decisions,
you
know
because
they
care
about.
What's
going
to
their
back
in
so
they're,
pretty
opinionated
about
what
they're
gonna
send
and
man
Francis.
My
opinion
was
just
we
can't.
We
can't
do
that,
necessarily
because
we're
kind
of
agnostic
in
terms
of
that,
but
that
might
be
that
rich
configuration,
your
timeout
mat
where
it's
like.
F
We
could
make
some
decisions
or
we,
you
could
say
you
know,
give
us
a
give
us
a
method
or
something
to
filter
the
headers
or
you
know,
give
us
the
rules
by
which
you
want
us
to
do
that
or
you
know,
but
it
almost
feels
like
send
every.
If
we're
gonna
do
the
header
stuff
send
everything
for
now
and
then
work
on
reducing
that
later
or
provide
configurability.
F
D
I
think
we
either
need
to
adopt
the
semantic
conventions
and
support
that
in
some
way
or
we
need
to
leave
it
configurable.
My
experience
I've
used
a
quite
a
few
different
backends
at
scale,
and
every
time
I've
tried
a
new
one.
I'd
had
to
go
and
adjust
instrumentation
or
do
some
kind
of
mapping
in
the
collection
pipeline,
because
every
like
every
back-end
has
different
expectations,
slightly
different
expectations
about
what
you're,
what
you're
sending
what
they
caught
anomaly
limits
are
things
like
that,
whereas.
F
C
D
D
D
Yeah,
the
suggestion
is
HDTV
clients.
Fans
should
be
using
conservative
low
cardinality
names
form
from
available
parameters
of
an
HTTP
request,
such
as
the
HTTP
method
name.
So
it's
kind
of
like
the
default
is
ultra
conservative
and
then
provide
configuration
hooks
to
to
use
slightly
less
conservative
names.
If
you
want
to.
C
C
C
So
so
there
you
go,
yeah
confirmed
the
the
community
will
be
dropping
in
in
four
minutes.
I
guess
there
are
two
things
on
my
mind
that
I
will
throw
out
there,
and
one
is
I,
am
curious,
like
what
people
would
think
about
possibly
moving
this
meeting
to
absolutely
any
other
time
spot
than
this.
One
like
Tuesday
or
Thursday
at
this
time
would
work.
C
Maybe
people
can
just
like
think
about
this,
and
maybe
next
week
I
can
make
a
doodle
of
what
times
people
can
make
it,
and
if
everybody,
if
we
can
all
agree
on
another
time
to
move
this-
and
there
is
a
freeze,
the
link
during
that
time-
maybe
we
can
do
that
and
the
whole
reason
behind
this
is
kind
of
interested
in
participating
in
the
javascript
sing
a
little
bit
but
they're
kind
of
at
the
same
time.
So
it's
cramping
my
style.
There.
F
C
Cool,
so
maybe
I
will
make
a
doodle
and
either
I
could
yeah
I
won't.
We
can
do
it
and
we
can
talk
about
this
at
the
next
meeting.
It
will
be
at
the
same
time
and
then
we'll
see
if
there's
a
something
else,
convenient
what
I
really
hope
that,
like
reify
something
convenient
I
wouldn't
want
to
move
it.
If
it
would
mean
that
people
that
have
been
attending
can
attend
anymore,
so
yeah.
C
So
I
won't
make
that
deal
or
I
guess
we
only
have
two
minutes
left.
Is
there's
really
not
enough
time
to
go
into
this
context
prop
stuff,
but
I
guess
we
can
think
about
over
the
next
week,
like
I'll,
keep
addressing
feedback
from
from
that
as
I
have
time
and
we
can
solve
stuff
over
or
the
PR,
but
if
it's
looking
like,
maybe
we
need
to
dedicate
some
time
to
it.
C
It's
like,
we
could
think
about
trying
to
chop
that
PR
up
into
chunks,
I
feel
like
it'll,
get
nasty
a
little
bit
just
because
I
would
kind
of
like
you
know
this
chain
of
PR
that
are
kind
of
like
you
know,
based
off
of
each
other,
and
but
if
it
must
be
done,
it
can
be
done.
It
might
not
be
super
super
easy
and
with
that
the
community
seems
to
be
appearing
so
I
guess
we
can
probably
close
out
close
out
this
meeting
and
I'll
see
you
all
next
week,
yep.
C
C
K
K
Okay,
so
welcome
everybody
to
the
first
open,
telemetry
community
meeting
of
2020.
It
looks
like
we
have
a
big
crowd
today,
which
is
excellent.
I'm.
Also
glad
about
that,
because
we
move
the
meeting
times
starting
I.
Think
this
meeting
some
glad
everybody
founds
the
new
calendar
entry
on
the
shared
calendar-
that's
fantastic,
please
add
any
items
to
the
agenda.
If
you
want
us
to
discuss
them
and
I
think
there's
already
a
couple
topics
there
and
then
we
can
get
started.
K
So
the
first
topic
on
the
agenda
right
now
is
an
update
on
project
status
and
a
schedule
for
beta
bicubic
on
I
added
this
I
suppose
I
should
be
the
one
to
talk
about
it,
but
I
think
this
will
be
a
broader
discussion
with
few
people
chiming
in
so
I
wanted
to
talk
with
all
the
attendees
in
this
forum
about
whether
we
think
we
can
make
what
sort
of
release
will
be
at
by
the
time
we
had
to
keep
con
conference.
I
know
for
a
number
languages.
K
G
K
G
K
G
K
G
K
L
G
But
in
order
to
be
called
better,
I
think
the
other
thing
that
we
need
morgan
is
sigma,
thinner,
put
down
a
document
of
things
that
needs
to
be
done
in
order
to
call
bet
on
the
things
that
can
be
tracked,
so
so
I
think.
That's,
that's
probably
a
good
start.
So
who's
gonna
lead
this
effort.
Maybe
you
can
offer
yourself
yeah.
G
H
L
H
G
L
So
maybe
you're
related
to
that
is
we
wanted
to.
There
is
a
docker
hub
account
named
open
telemetry,
which
we
wanted
to
use
to
publish
the
docker
images,
and
it
now
goes
to
unleash
an
account.
We
don't
know
all
registered.
The
account
currently
docker
has
a
procedure
for
transferring
the
ownership.
If
you
have
a
registered
trademark,
which
we
do
for
open
telemetry,
so
I
created
a
ticket
with
docker
hub
to
do
this,
but
got
no
replies.
If
anybody
has
a
better
channels
to
escalate
this
with
docker,
we
should
do
that.
We.
K
L
L
K
G
M
K
So
the
next
item
on
the
agenda
is
the
actual
cig
updates,
which
we
usually
leave
for
last.
So
all
this
ass
again
does
anybody
else
have
any
topics
they
want
to
discuss
today,
either:
focus
on
cube
können
and
the
beta
release,
or
just
generally
other
topics
they
want
to
talk
about
before
we
move
to
sake,
updates,
I.
I
Think
my
peaceful
to
talk
about
the
prototypes
that
we're
working
on
in
each
language
has
particular
metrics
and
context
propagation,
because
we're
doing
this
thing
now,
where
we
have
a
hotel
and
then
we
have
a
couple
prototypes
for
each
of
the
other
steps
and
then
really
hopefully,
by
implementing
the
prototypes.
We
learn
what
kind
of
questions
to
raise
in
the
adepts
before
when
we
write
the
spec
and
then
we
use
the
spec
to
come
back
and
write
the
actual
implementation.
I
G
Experience
that
I
have
in
Java
almost
put
together
a
prototype
for
for
context
propagation
and
what
I
asked
him
is
what
what
I
would
do.
All
the
time.
I
asked
him
to
start
splitting,
that
into
small
pieces
that
are
testable
and
and
easy
to
review
and
carefully
review
them.
So
I
would
suggest
that
you
as
a
signal
tenor.
You
push
hard
on
the
on
the
author
of
that
pr2
to
split
it
into
more
reviewable
and
testable
changes.
I
G
Yeah,
that's
that's
probably
something
that
the
lies
that
people
told
us
that
they
are
pushing
very
hard
this
week
to
get
out
the
specs
for
the
convex
prop
Josh
mentioned
that
yesterday
in
the
in
the
specs
meeting.
So
I
wish
that's
gonna
happen
and
I
think
he
is
correct.
You
are
right
that
we
should
have
push
more
harder
on
the
on
respects,
but
in
wait
for
that
I
don't
know,
but
I
think
I.
I
G
A
K
L
L
Okay,
so,
basically,
since
the
last
community
meeting
that
changes
primarily
making
it
even
I-
guess
right
here
for
production
usage,
we
have
better
coverage
by
tests
by
end-to-end
tests.
We
have.
We
have
built
from
the
country
version
which
previously
had
the
the
minimal
version
or
the
core
version.
Only
now
we
have
to
tube
official
builds.
Both
are
also
available
in
docker
hub
feature
wise.
We
have
two
new
components:
the
memory
limiter
and
the
kubernetes
tagging
and
I
think
that's,
that's
primarily
the
in
terms
of
what
we
have
new
since
the
last
time.
G
G
G
So
we
we
started
to
split
and
merge
some
of
the
context
propagations
the
the
non-controversial
changes
like
renaming
the
disability
context
who,
to
correlation
context
and
couple
of
other
things
like
that
for
for
metrics,
we
did
I
already
added
that
by
the
way
is
the
next
item.
But
for
metrics
we
did
a
good
progress.
We
have
now
all
the
instruments
in
implemented,
but
we
do
not
have
the
aggregations
that.
E
G
A
A
E
A
K
I
Is
that
you
yep
okay,
so
like
some
of
the
episodes
are
big
blogging
issues
are
context,
propagation
and
metrics
aggregation.
We
have
two
big
PRS
open
for
this
now,
besides
that
we
just
have
a
bunch
of
cleanup
work
to
do,
but
we
are
largely
on
track
to
get
this
right
about
much
excellent.
Thank
you.
Go.
E
Rahul
so
histogram
aggregation
was
recently
added
and
this
week
and
I
think
most
of
the
aggregation
up
there
for
colanow,
so
metrics
is
pretty
much
complete,
except
for
the
view,
specification
and
I
think
that
we
need
the
spec
for
that
as
well.
So,
with
respect
to
context
propagation,
the
PR
is
there
I
think
it's
still
in
progress.
Hopefully
we
should
have
it
ready
for
you
get
this
week,
yeah
and
waiting
for
spec
to
identify
all
the
better
requirements
so
that
we
can
appropriately
open
issues
and
fix
the
missing
gaps.
Basically
perfect.
K
C
C
C
Sdk
configuration
is
a
headache,
but
there's
some
PRS
actually
make
that
pretty
easy
reduces
the
number
of
lines
of
boilerplate.
You
have
to
oh
right
to
set
up
an
SDK
and
our
our
readme
is
woefully
out
of
date
and
I
think
this
is
true
of
it's
not
only
Ruby
but
some
other
things.
So
we
have
a
PR
that
brings
that
up
to
date
and
kind
of
throw
out
some
estimates,
so
we're
kind
of
just
hashing
out
how
we
feel
what
those
estimates,
but
soon
our
readme
will
be
a
little
bit
more
reasonable.
K
M
Yep
well
so
that
Matt
is
doing
okay.
We
there
is
a
big
topic
for
now
or
not
for
us
it's
how
the
converse
is
done.
A
team-best
dotnet
has
some
abstractions
open.
Tillymann
has
some
abstractions
and
we
really
want
to
marry
them
together.
So
we
are
discussing
water
but
like
since
we
still,
we
still
have
some
opportunity
to
change
open
to
the
API
a
little
bit
specifically
for
dotnet
to
a
better
transition
to
future
versatile
and
hopefully
doesn't
fight
we'll
bring
even
more
goodness
and
in
terms
of
built-in
functionality.
M
Besides,
that,
like
we
have
some
matrix
works
as
we
started,
and
we
implemented
some
api's,
but
they're
not
completed
yet
and
aggregations
are
also
not
ready.
We
have
only
few
of
them.
We
have
a
some
interest
to
this
matrix
course,
especially
with
like,
even
though
like
what
is
like
to
tow
like
this
is
futile
yeah,
so
they
actually
looking
for
open
today,
minions
a
waiting
for
us
to
anywhere
that
I
hope
will
have.
Some
customers
are,
and
yeah
decidedly
just
keep
pushing
towards
0
for
3
and
beta
okay.
H
Yeah
today,
I
finally
got
merged
in
trace,
tracer
provider,
stuff
and
registration
and
documented
it.
We
were
also
had
a
very
old
readme,
but
now
with
documents
named
tracers
and
how
to
use
them
best
practices
things
like
that
nice
we
have
the
metrics
API
ready
to
merge,
but
the
SDK
still
has
the
good
ways
to
go,
but
I
mean
it's
got
basic
stuff
in
there
counters
and
gauges
and
whatnot
Blake's,
our
API.
So
technically
it
works
with
a
lick,
sir,
but
to
have
an
idiomatic
elixir.
H
K
Perfect,
normally
we
do
a
website
update
I,
don't
think
we've
done
much
work
on
the
website.
Recently
we
were
doing
a
lot
before
cubic
on
us
and
I'm
guessing
it'll
pick
up
before
Q
con
you
all
right.
So
that's
it
for
the
sick,
check-ins
and
I
can
see
that
Sergey
right
now
is
adding
additional
topics.
M
Okay,
so
I
just
wanted
to
thank
everybody
for
their
submissions.
We
were
asking
people.
What
can
we
do
for
cubic
or
maintain
your
truck
and
interesting
submissions
and
as
a
result
of
the
submission,
it
says
that
the
Java
auto
instrumentation
and
the
collector
work
yes
actually
combine
to
represent
a
means
very
well.
So
we.