►
From YouTube: 2023-01-25 meeting
Description
Open Telemetry Meeting 1's Personal Meeting Room
C
B
E
D
D
D
D
D
C
D
E
D
Okay,
I
think
I,
it
should
be
fine,
I
mean
I.
Think
we
agreed
that
King
Lock
would
be
incremental.
It
would
be
difficult
for
anybody
to
really
see
what
all
is
duplicated
over
the
course
of
time
and
probably
having
one
single
document
which
outlines
which
describes
I
mean
all
the
features
which
have
been
the.
D
E
D
C
E
D
D
D
Okay,
yeah
discussed
CPP
and
just
go
to
the
next
one.
I
mean
correct
me.
If
you
want
to
further
discussion,
discuss
CPP
and
CPP
contact
built,
yeah,
say
thanks
for
Mark
I.
Think
I
saw
your
comments
on
this.
I
just
wanted
to
have
another.
Go
try
to
try
to
see
how
it
looks
like
if
we
do
the
changes
I
mean
how
it
will
look
like,
and
you
know
whether
it's
a
very
minimal
change
and
then
just
wanted
to
get
the
more
wider
view.
D
D
So
so
probably
I
just
felt
that
let
me
have
another
go
on
this.
I
mean
try
to
see
that
the
more
of
a
wider
view
of
how
everybody
feels
on
this
change.
Yeah
I
mean
this
I
I
this
this
was
I
mean
adding
is
with
contrary.
Option
was
something
which
was
more
futuristic,
I,
don't
I,
know
I
mean
with
the
content.
Repo
is
not
something
which
is
as
of
now
well
maintained.
D
We
know
how
what
what's
the
dependencies
on
on
the
main
trap
of
each
of
those
components
have
so
I,
don't
think
with
something
which
is
available
immediately
but
yeah.
This
was
something
I
think
I
just
thought.
Probably
in
future
we
can
do
that.
I
have
gone
through
your
comment.
Mark
I
mean
I.
Think
for
the
first
respect
for
the
legal
I
felt
that
this
should
not
have
any
legal
issues
just
correct
me,
because
we
are
not
checking
in
any
any
file
as
such,
which
has
a
different
license
or
a
different
copyright.
D
It's
just
that.
It's
an
optional
kind
of
optional
feature
anybody
who
want
to
use
it.
They
want
to
build
a
contract
or
they
want
to
build
any
external
component.
They
can
use
that
feature
which
we,
if
if
if
we
add
that,
but
just
correct
me,
I
mean
if
you
really
feel
that
this
this
has
a
legal
binding
I
mean.
C
A
E
C
D
Yeah,
if
we
use
it,
that's
something
which
is
awesome.
We
don't
be
even
we
don't
give
the
name
of
the
repo
anywhere
yeah.
Even
if
it's
it's
been
good,
so
superbal
I
mean
still
it's.
How
applications
want
to
how?
How
somebody
who
want
to
utilize
this
open
Telemetry?
They
want
to
use
it.
They
then
off
which,
if
they
want
to
use
this
feature,
they
know
they
know
which
which
Rapport,
which
directory
they
are
including
it
with
the
build.
So
they
should
have.
D
D
But
yeah
that
that's
one
of
the
issues
which
probably
you
pointed
out,
the
technical,
disc,
I
agree
components
in
Upper
Telemetry,
contrary,
but
not
well
maintained
all
right.
This
is
something
which
we
cannot
use
as
of
now,
and
it's
not
definitely
upgraded
version
of
packages.
It
depends
upon
now
any
security
bugs
we
don't
know.
Possibly
it
would
be
there.
C
D
D
Still
correct
me,
I
mean
even
if
somebody
wants
will
create
a
super
build,
they
still
want
to
have
to
add
I
mean,
as
you
mentioned,
both
the
dependencies
they
have
to
build.
Both
both
I
mean
the
main
repo
and
the
component
sure,
and
they
have
to
ensure
so
the
I
mean
this
was
I
mean
talking
about
the
second
point
which
I
probably
this
was
more
I
mean
this.
This
option
is
more
to
provide
a
convenience
to
somebody
who
really
want
to
utilize
open
Telemetry.
D
E
D
To
definitely
build
everything
or
I
mean
all
the
components
so
just
to
provide
an
the
I
mean.
The
motive
was
just
to
provide
a
convenience
for
somebody
who
want
to
integrate,
create
a
single
build.
I
mean
to
do
on
to
just
want
to
do
only
single,
build
and
create
a
package
with
the
external
component,
along
with
the
main
routing.
It
was
something
I
think
have
seen
opencv
also
using
it.
They
also
do
the
same
thing.
D
D
Okay,
in
other
places,
also
basically,
basically
for
the
for
the
sdks
which,
which
brings
with
which
can
which
can
have
extensions
like
we
have
the
components
which
can
be
integrated
with
with
that
SDK
and
they
can
be
utilized
by
the
customers
of
the
SDK
or
of
the
components
so
just
to
provide
a
layer
of
con
of
convenience
on
top
of
the
sgk
I
know.
I
know
this
brings
I
mean
it
just
has
a
single
ad
subdirectory
or
something,
and
anybody
can
try
to
add
anything.
D
D
E
D
C
D
D
Sure
sure
no
I
think
that
should
work
I,
don't
think
that
should
have
a
problem,
but
let
me
just
test
it
out
and
let
me
see
how
it
it
goes
so
I
mean
right
now:
I
can
probably
I
think
I
already
viewed
that
link.
How
can
I
know
what
the
approach
I
am
using
is
and
still
that
is
causing
a
problem
to
my
customers.
D
D
E
C
C
D
C
It
does
a
lot
of
things
and
my
concern
is
to
try
to
break
it
into
smaller
pieces
so
that
smaller
pieces
can
be
merged,
and
the
reason
for
that
is
the
first
part
is
that
there
is
refactoring
on
environment,
variable
and
I.
Think
that
is
ready
to
go.
It's
it's
very
simple,
but
it
does
a
lot
of
cleanup
which
is
needed.
The
second
part
is
implementing
just
SSL
alone
in
the
HTTP
exporter,
so
that
as
a
spec
already,
it's
just
that
we
don't
use
the
client
certificate,
the
client
key
and
the
certificate
today.
C
So
we
we
can
do
that
and
we
will
be
compliant
with
respect,
because
this
spec
is
already
defined
for
this.
So
this
is
the
second
part
that
can
go
in
which
is
causing
more
trouble.
Is
the
implementation
for
TLS
versions
to
set
the
minimum
and
maximum
clearance
version?
It's
a
new
feature
which
is
not
even
in
the
spec,
shows
or
raised
an
issue
a
long
time
ago,
but
it
got
no
attention.
C
D
E
C
I
can
look
at
that
so
in
in
general,
so
where's
the
question
of
how
to
slice
the
patch
in
in
smaller
pieces,
but
there's
also
a
bigger
question,
which
is
how
to
test
it
in
CI
and
I
have
no
ID
because
so
first
of
all
is
through
test
SSL.
So
we
cannot
just
have
the
me
HTTP
server
read
or
read.
One
that
implements
is
a
cell
and
I.
C
D
E
D
C
E
D
C
Everything
so
when
doing
tests
locally
I,
just
look
at
Vlogs
and
I,
see
if,
if
it's
connecting
or
not
and
if
some
span
is
given
to
the
collector,
what
I
don't
know
is
how
to
do
that
automatically
to
test
that
the
connection
is
fine,
and
the
problem
here
is
that
if
we
have
a
client
using
the
HTTP
exporter,
the
exporter
can
work
or
can
fail.
But
there
is
no
return
code
written
to
the
client
there
is.
There
is
no
external
sign
that
we
can
test
to
make
sure.
C
D
D
B
D
C
B
E
C
And
there
is
also
testing
of
negative
conditions
like
what,
if
you
set
up
a
serious
version
that
does
not
exist.
What
do
you
visit
a
Visa,
but
I
mean
in
that
case
the
exporter
will
fail,
but
we
need
to
at
least
test
that
it
doesn't
crash
and,
if
possible,
to
test
whether
it
returns
a
proper
error
or
explaining.
C
D
D
E
D
B
D
E
E
D
Yeah
so
so
I
think
I
mean
Mark.
If
I
understand
you're
going
to
split
it,
the
first
the
first
period
would
be
for
the
environment
variable
path
and
then
probably
so:
okay,
okay,
fine
and
then
the
two
final
will
go.
I
won't
be
reviewing
this,
as
of
now
probably
I'll.
Be
reviewing
this
all
of
here
that
you
will
be
raising
I
hope
that
should
be
fine.
D
D
D
E
D
D
D
It
provides
the
flexibility
that
this
I
mean
in
the
code.
It
provides
the
flexibility
that,
like
that,
the
user
should
need
not
be
using
the
same
order
for
the
argument
they
can
pass
this
arguments
in
any
order
and
the
implementation
will
take
will
be
a
positional
type
of
the
arguments.
All
the
types
are
different
So
based
on
the
type
of
the
arguments
they
can
read.
They
should
be
if
they
I
mean
the
implementation
will
process
it.
D
So
the
flexibility
is
that
the
order
of
user
need
not
give
all
the
correct
order,
but
just
just
the
problem
I
see
is
that
I
mean
it's
by
seeing
this
the
Prototype
of
this.
They
don't
know
what
all
things
that
should
be
passed,
but
if
I
mean
I
am
okay
with
that,
with
the
documentation
which
details.
What
are
the
accepted
parameters
in
the
of
this,
but
I
mean
if.
D
C
D
So
this
slides,
yeah,
okay,
this
lines
basically
means
that
it
can
accept
any
number
of
arguments
and
based
on
the
type
of
the
arguments,
the
actual
prototype
of
this
function
will
get
created.
So
this
is
more
of
our
body.
The
variable
arguments
which
we
haven't
seen
if
this
is
a
template
form
of
that
variable
arguments,
so
I
mean
so
I
mean
my
only
concern
was
that
this
is
this.
D
I
have
to
go
through
the
documentation
and
then
I
have
to
realize.
Okay.
This
is
how
I
have
to
pass.
So
it's
not
self-explanatory.
As
most
of
the
I
mean
it's
not
very
much
military,
as
as
the
added
prototypes
we
have
in
the
EPS
office
like
earlier,
we
were
having
something
called
a
severity
like
if
I
see
here
earlier
how
we
used
to
have.
D
So,
probably
just
to
compare
it
with
the
earlier
approach.
We
were
using
something
like
this.
If
you
see
here
so
so
anybody
who
is
using
this,
they
know
that
okay,
the
first
argument
is
the
severity.
Second,
is
the
body?
Third?
Is
attributes?
Trace
ID
stand
ID
this
this
now
we
change
it
change
it
to
just
having
one.
B
D
E
C
C
D
E
D
Feels
that
is
this
is
not
a
nice
approach.
That's
the
only
thing
not
not
in
terms
of
supportability
I,
don't
I
I,
don't
feel
that
that
could
be
a
concern.
It's
just
that
this
is
something
which
looks
better,
or
this
is
something
which
looks
better.
Providing
multiple
overhead
functions
in
this
well.
E
C
D
Think
it's
yeah,
so
it's
around
seven
to
eight
different
overloaded
functions,
which
gets
replaced
with
someone's
a
single
one
and
the
the
benefit
is
that
now
we
don't
need
to
worry
about
the
order
of
these,
so
we
don't
I,
don't
need
to
remember.
I
I
should
just
know
what
all
arguments
I
have
to
pass
that
some
somehow
documentation
may
give,
but
I
don't
need
to
worry
about
what
order
I
have
to
pass
whatever
order
I
will
pass.
It
will
automatically.
D
C
If
it
builds,
why
not
just
put
things
in
perspective
to
today?
Well,
someone
complained
about
the
bill
break
in
some
template
code
in
no
STD
span
and
looking
at
the
code,
it
was
a
C
plus
plus
template
using
meta
programming
and
and
sdg
neighborhood
and
whatnot,
and
the
Lambda
and
I
mean
it's
the
kind
of.
D
B
D
C
D
D
Also,
we
can
directly
provide
the
key
value
attributes
initial
as
an
initialization
list,
which
we
won't
be
able
to
provide
so
I
think
we
have
to
add
another
function,
so
if
they
want
to
directly
provide
the
attributes
they
have
to
use
this
function,
make
attributes
and
then
pass
that
and
then
then
that's
that's
a
minor
thing
which
probably
would
should
be
documented
and
then
they
can
use
it,
but
yeah.
That's
another
minor
change:
okay,
yeah
yeah!
That's
that's
something!
I
think
these.
This
was
it
another
stream,
but
other
than
that
I
think
it's
provides.
D
The
flexibility
difficult
to
maintain
I
agree
we'll
need
somebody
with
somebody
like
who
went
to
really
who
knows
better
programming
that
will
to
really
keep
maintaining
otherwise
we'll
take
some
time
for
us
to
understand.
That's
yeah,
you
know
done
done
very
nicely,
I
mean
it
uses
the
typewrites
and
to
basically
ensure
that
the
arguments
can
be
passed
in
any
order,
and
then
that
basically
uses
these
Setter
methods.
D
D
C
D
Yeah
any
any
problem
which
comes
up
I
think
once
once
we
make
it,
make
it
stable.
If
we
find
out
any
problem
in
this
template,
but
I
could
probably
quote,
it
would
be
difficult
to
really
handle
change.
The
API
well
yeah
that
that
some
concerns
I
also
had
so
yeah.
It
took
me
lots
of
time
to
really
understand
how
he's
doing
it.
What
exactly
is
happening
in
this
code.
E
D
E
A
A
Think
the
critical
thing
is
here:
the
memory
thing
right
because
building
a
DL
for
all
memory
allocation
and
free
and
we
need
to
make
sure
it
happens
inside
the
DL
for
for
the
objects
and
we
women
tend
yeah
I,
think
yeah
most
time
are
there
known
issues
and,
in
my
mind,
were
already
included
in
the
VR,
so
feel
free
to
take
a
look
from
my
side,
I
think
I.
The
only
thing
missing
in
API
is
I
I'd
like
to
add
more
PR
on
some
some
tricky
stuff
for
this.
For
this
change.
D
A
D
D
A
D
A
I
think
all
the
classes
which
can
show
show
up
in
user's
code
we
need
to,
we
may
need
to
expose
them
like
if
the
user
wants
to
call
call
their
function
in
in
our
DL.
We
need
such
export,
but
if
for
some
utility
or
some
function
which
doesn't
provide
the
memory
allocation,
so
the
inline
the
method
in
the
user's
code?
That's
fine!
So
then
the
API
tag
is
not
necessary.
A
D
A
D
A
C
C
Yes,
I
I,
don't
understand
exactly
the
problem
with
memory
education
due.
Is
it
the
case
that
the
place
doing
allocating
memory
must
be
the
same
Library
as
the
place
destroying
memory?
And
if
this
is
the
case
under
the
network,
we've
shared
the
shared
pointers
that
we
have,
because
today
we
can.
The
application,
for
example,
can
create
an
exporter
that
will
create
a
SharePoint
or
sharep
pointer
with
a
reference,
give
it
to
open
Telemetry
and
then,
when
open,
to
inventory
shuts
down.
C
A
C
A
A
A
D
A
little
bit,
okay,
I
will
probably
let
him
be
back.
We
can
talk
about
the
other
things.
Meanwhile,
just
interesting
time.
We
just
have
seven
minutes
so
I
just
had
one
like
very
quick
I
think
probably
discussion
on
using
Hotel
CPP
Trace
in
the
multi-threader
racing
applications.
They
have
been
multiple
questions
coming
on
that
how
users
should
be
using
it.
It's
a
scenario
is
like
creating
a
span
in
one
function
or
one
thread
and
exporting
it
from
another
parenting.
The
span
Creation
in
different
thread.
D
We
don't
have
enough
documentation
whether
we
should
document
it
or
whether
we
should
do
any
change
to
make
it
work.
How
the
other
I
don't
have
how
whether
other
six
support
the
scenarios
where
the
span
can
be
created.
It's
just
I
mean
the
reason
why
it's
different
from
the
current
existing
supported
scenario
is,
in
the
current
supported
scenario,
our
the
in
case
of
implicit
context
management.
D
The
context
of
this
parent
is
told
in
the
thread
local
storage,
so
the
problem
is
like
if,
if
I
want
to
parent
a
given
span
from
a
different
set,
it's
not
possible
because
the
context
storage
is
maintained
in
the
in
the
previous
slide
on
the
earlier
thread
and
I
cannot
really
or
we
cannot
use
this
context.
Management
in
multi-threaded,
environment
or
even
even
in
the
scenarios
where
I
want
to
create
a
span
in
one
of
the
function
and
then
try
to
try
to
clean
it
up
in
the
other
functions.
C
I
happen
to
have
a
multi-faded
application
and
what
I'm
doing
is
I'm
first
of
all,
I'm
not
I'm,
not
using
runtime
context
at
all
I'm,
not
using
Federal
Storage
I'm,
explicitly
keeping
track
of
the
balance
plan.
All
the
time
yeah
so
that
when
creating
expander
can
I
can
attach
it
to
the
parents,
because
I
know
what
the
point
is
and
things
like
that.
So
fight
has
been
working.
Okay
in
small
testing,
with
I
think
on
this
part,
but
sure
this
this
kind
of
code
I
mean
it's
a
drawing
time.
D
Probably
I
think
that
I
agree
that
the
current
that's
the
current
view
of
that's
the
current
best
with
how
or
that
in
the
current
implementation.
That's
how
it
has
to
be
done.
The
parents
pan
need
to
be
keep
track
of
the
parents
plan
and
when,
when
we
are
I
mean
I
think
the
scope
where
we
are
creating
a
child's
parent.
The
parents
plan
should
be
available
at
that
time.
So
that's
how
it
has
to
be
done.
Yeah.
C
D
Yes,
that
was
a
wrong
approach.
Definitely
that
that's
I
think
he
was
using
scope
in
in
the
multitude,
probably
the
family,
I
think.
In
that
case
he
was
not
aware
that
the
scope
is
specific
to
a
threat,
local
storage
and
he
was
somehow
trying
to
use
it
in
some
other,
so
it
which
is
not.
It
won't
work
in
that
case
yeah,
but
but
yeah,
that's
I,
think.
D
Probably
it's
good
to
document
I
just
felt
that
I,
probably
I,
can
create
a
task
for
documenting
it,
because
we
didn't
document
that
scenario
in
a
better
way,
not
to
use
it
in
multinated
applications
or
in
general.
How
to
use,
without
runtime
storage,
to
specify
the
parent
test
argument.
So
yeah
probably
I'll
create
a
task
and
I.
E
D
C
So
we've
named
typically
when
you
open
a
socket
and
say
that
it
goes
to
to
a
given
hostname,
if
you,
if
you
have
only
one
network,
it's
a
host
name
for
that
network,
but
we
can
have
named
the
network
on
top
of
that,
so
that
you
can
have
a
blue
Network,
a
black
Network,
a
green
Network
and
so
forth,
and
there
is
some
API
to
call
to
basically
tell
Fred
that
it
belongs
to
a
given
Network
and
I.
Don't
know
today,
or
this
works
with
open,
Telemetry
and
I'm.
Assuming
it.
We
need
some
changes.
D
E
D
Okay,
I
well,
just
the
last
one
I
think
we
see
package
to
support
with
the
TLP
https
cmake
option.
I
think
this
was
something
slack
you
know
hit
storm
I
mean
if
you
just
still
around,
probably
whenever
you
are
creating
a
package
next
time
for
VC
package
changes.
Probably
just
have
you
know,
just
just
keep
keep
keep
coming.
Probably
you
can
add
this
as
a
feature
also:
okay,.
D
E
C
D
D
B
C
B
D
D
A
Open
and
I'm,
showing
C
plus
plus,
especially
for
the
library
which,
in
just
our
library,
which
usually
I
think
requires
high
performance.
So
if
they
for
logging
currently
giving
the
there's
no
way
to
start
I
think
for
to
lose
a
way
to
stop.
Logging
like
I,
think
in
linear,
should
like
to
to
remove
the
Tracer
provider.
I
think
there's
nothing
use
the
API
like
to
set
off
lag
to
this
to
see
I
want
for
this
local
session.
A
I
want
to
disable
it
temporarily
or
enable
it
yeah,
I
I
was
thinking
about
me
either,
adding
this
to
spec
or
in
C,
plus
plus.
We
have
the
flexibility
to
like
to
add
a
wrapper
API,
which
is
not
a
standard,
but
we
have
it.
We
had
like
a
like
logger.enabled
or
is
enabled
such
such
as
a
quick
flag
check
in
there
and
if
it
passed,
then
we
can
call
the
logger.log,
because
the
log
Accord
we
have
many
wrappers
I,
think
if
we
build
that
into
DL.
A
That's
I
think
there
are
some
some
significant
cost,
because
emailing
can't
happen
there
so
yeah.
If
that
the
user
want
to
disable
lock
or
in
the
log.
Doesn't
it's
not
exported
with
the
only
core
internet
Decay
for
many
core
levels?
So
that's
a
significant
cost
I
think
for
for
like
for
how
to
Loop.
So
if
we
have
a
like
a
flag
which.
E
D
D
D
A
D
D
C
C
A
A
E
D
But
that
does
not
mean
that
it
will
not
call
the
apis
the
the
normal
line
method
that
will
still
call
those
methods
and
the
new
logo
provided
method.
And
then
even
the
logger,
which
would
be
created
by
node,
will
again
they
have
to
create
a
logger
from
the
get
a
new
logger
from
the
logger
provider
and
then
start
using
that
it's
a
complete
initialization.
D
A
E
C
C
D
C
C
There
is
another
point
also
which
is
especially
to
avoid
that
so
that
to
give
everyone
know
and
then
we'll
have
a
need
to
disable
things.
So,
instead
of
everyone
implementing
their
own
flag
to
disabled
things,
there
is
an
item:
invest
pack,
which
is
an
alignment
viable
which
is
open,
Telemetry,
SDK,
disabled
I,
think
and
that
one.
E
A
C
Disable
all
the
free
providers.
So
it's
not
it's
not
per
signal,
but
at
least
it's
it's
a
way
to
completely
disable
the
SDK,
and
so
there's
an
issue
for
that.
This
is
an
example
where
the
fix
depends
on
a
cleanup
in
in
the
environment
library
in
the
first
place,
but
there
is
a
way
to
to
basically
check
for
that
environment
variable
when
setting
a
tracer
and
if
your
SDK.
D
A
A
D
A
Lot
level
is
like
below
the
level,
then
just
just
an
internet
check
and
then
do
a
really
totally
ignore
the
lock.
That's
a
successful
acceptable,
but
we
don't
want
to
make
it
a
function,
call
and
then
check
the
nested
are
and
do
that
because
in
our
you
know,
situation
we
like
we
pass
the
parameter.
We
may
need
to
pass
many
strings,
so
in
many
parameters
inside
do
some
conversion
right,
I
think
that's,
that's
not
a
lightweight
operation,
so
I
mean
from
from
what
I
get
from
from
Market.
A
C
Dynamically
is
more
complicated,
I
think
the
autolistically
disabled
is
just
globally
to
start
an
application
without
the
SDK
and
without
open
telemetry.
The
all
the
trace
providers
were
single
terms.
So
if
someone
sets
that
Singleton,
all
the
files
of
the
application
will
be
impacted
as
instantly
so
there's
no
way
to
isolate
one
thread
of
execution,
for
example,.
A
A
E
C
A
A
D
C
A
D
B
B
C
Like
how
many
signals
do
we
see
of
a
of
a
recorded
of
a
dropped
things
like
that,
so
we
can
definitely
say
if
a
provider
is
a
new
provider
or
not
as
well
what
sounds
feasible,
but
the
thing
is:
if
you
start
doing
that,
then
your
applications
start
to
be
aware
of
fdk
as
ddk.
So
you
start
asking
for
SDK
is
tracing
enabled
so
that
if
it's
not,
then
I
don't
know
calling
tracing
and
it
makes
multiple
code
buffing
application.
A
C
B
A
C
A
I
see
yeah
yeah
I
can
I
can
try
that,
because
actually
in
the
concern
comes
from
the
idea
right
like
for,
for
none
deal
for
starting
a
library
or
in
The
Logical,
maybe
in
line
with
aggressive
optimization
for
better
for
currently
for
our
with
DL
build.
Actually
the
API
will
always
cross.
There
will
make
a
I
think
indirect
the
call
into
another
Library
so
that
cost
Maybe
significant.