►
From YouTube: 2023-02-08 meeting
Description
Open Telemetry Meeting 1's Personal Meeting Room
A
B
D
So
probably
we
can.
Let
me
see
if
Tom
is
going
to
join.
D
D
Yeah,
if
I
mean
well
I,
think
I
just
have
one,
probably
whenever
you
have
time
for
this
PR
one
PR
on
matrix,
it's
just
basically
adds
The
Benchmark
tests
for
some
aggregation.
D
I
also
have
put
the
results
of
that
benchmark.
I
mean
right
now
we
don't
have
a
very
full-fledged
performance
metrics,
which
we
are
getting
for:
the
our
Matrix
internal
Matrix
implementation,
so
I'm
just
trying
to
capture
some
of
the
benchmarks,
which
would
be
important
for
us
to
really
show
if
anybody
is
interested
to
see.
What's
what's
with
how
good
is
the
performance
of
the
The
Matrix
SDK?
That
will
just
get
some
figures
so
I'm
just
trying
to
add
some
some
tests
for
that
benchmark.
D
Okay,
I
think
we
can
start.
We
have
a
quorum
now
so
probably
with
the
agenda.
I
think
there
were
a
couple
of
specs
item
which
I
just
wanted
to
call
out.
So
what
was
allow
allow
adding
links
after
spam
creation?
This
is
I
think
this
has.
This
ask
has
been
coming
for
a
few
times
in
C,
plus,
plus
Community,
also
and
I.
Think
even
the
other
six
also
are.
The
similar
task
has
been
there
that
it
should
be
possible
to
add
links
after
spread
creation.
D
There
is
a
PR
for
this,
which
is
already
approved.
If
I
see
it's
already,
at
least
by
two-
please
even
listen.
Hopefully
it
should
go
through.
That
means
that
we
may
have
to
modify
our
API
Trace
API.
That
would
be
some
that
would
be
Avi
breaking
change,
but
I
think
this
is
something
important,
as
they
have
been
multiple
ask
for
that
to
to
once
it
gets
merged.
Probably
we
also
have
to
start
looking
into
we'll
add
the
support
for
this.
D
D
So
I
think
warrant
is
already
implementing
the
log
API
in
SDK.
Changes
as
per
the
current
specs,
so
would
be
good
for
good,
should
not
be
affecting
us
so
much.
The
only
problem
here
is
that
we
see
it
as
a
log
API,
but
this
is
actually
not
a
log
API.
It
is
the
log
backend
API.
So
that's
the
discussion
going
on
and
I
think.
We
also
had
some
comments
on
this.
We
want
it.
D
If
we
wanted
that,
if
it
is
called
as
log
API,
then
we
should
have
log
levels
also
part
of
the
API
to
at
least
to
where
to
check
the
level.
The
current
level,
the
log
level
should
be
something
which
should
be
able
to.
We
should
be
able
to
configure
at
the
SDK
level
and
at
the
API
we
should
be
able
to
check
the
log
level
and
then
based
on
that
the
actual
log
we
should
trigger
the
actual
login.
D
So
this
is
not
as
part
of
the
spec
size
of
law
So.
The
plan
is
to
change
it
from
log
API
to
log
Spa
kind
of
service
provider
interface,
not
an
application
programmer
interface,
because
the
current
API
is
not
something
which
application
programmer
can
directly
use
it.
It
is
more
as
a
backend
API,
and
then
we
I
mean
it
seems
that
we
as
a
C
plus
plus
Community,
does
not
have
any
log
API
or
any
any
free
log,
any
log
framework
which
provided
directly
provided
by
the
language
package.
D
So
that's
something
I
think
probably
we
have
to
do
as
a
separate
step,
but
yeah,
let's
see
first,
what
where
this
goes
discussion
goes
as
of
now.
It
looks
like
that
we
will
have
a
log
SPI
and
log
SDK
and
yeah,
so
there's
already
a
PR
for
that,
and
then
there
won't
be
any
log
API
which
would
be
provided
by
the
specification.
D
B
D
G
D
Yeah
I
mean
I,
had
some
minor
comments,
I
mean
it
looks
good
to
me.
I
know.
I
know,
Mark
has
some
valid
concerns
about
that
which,
which
I
think
I
agree
on
that,
but.
D
This
would
be
something
which
we
can
Market
as
an
experimental
once
it
goes
through
and
then
probably
have
a
more
thorough
discussion.
How
should
we
support
dlls
with
multiple
targets
separate
deal
with
multiple
targets,
and
so
that's
something
probably
we
will
continue
discussing
on
that,
but
I
think
Mark
I
think
you
had
done
more
thorough
review
on
that
I'll.
Let
you
to
decide
and
to
do
to
probably
talk
about
this.
If
you
want.
H
So
the
most
of
well
at
some
percent
earlier,
but
because
of
the
way
we
had
the
STD
editor
and
whatnot,
but
this
is
all
resolved.
Now
as
a
side
note,
I
wanted
to
remove
the
the
flag
in
the
in
GitHub
itself
to
to
remove
the
requesting
for
change
flag,
but
I
could
not
do
it.
I
don't
know
if
if
there
is
a
way.
H
Because
I
tried
to
enter
a
new
review
only
as
a
comment
and
it's
that
change
requested
flag
just
takes
it.
It
will
not
remove.
D
H
Just
to
say
that
I
am
no
longer
asking
for
specific
changes.
It's
it's.
B
F
Yeah
but
I
agree
in
the
current
traffic
proposal
can
stay
in
preview
or
for
a
while
for
a
few
release,
and
then
we
decided
later
like
we
still.
We
can
still
make
change
like
to
build
it
into
multiple
deal.
Let's
currently
I
added
in
the
dark
web
to
say
to
imply
this
is
for
for
preview.
Maybe
we
need
a
flag
in
a
like.
They
make
to
see
enable
preview
to
enable
this
feature
to
implicate
imply
that
that
explicitly.
D
Or
probably
just
mark
this
as
underscore
preview
or
something
openly
build
import,
we
already
have
one
cmake
option
right
to
enable
this.
If
we
just
suffix
it
with
preview,
will
that
work
or
not,
not
probably
probably
add
a
flag
would
be
better.
Let's
not
change
this
here.
D
D
D
It
I
just
wanted
to
discuss
it
today
before
before
approving
it.
For
me,
the
changes
looks
very
good.
Very
nicely
done,
I
mean
with
some
after
those
comments
regarding
the
default
delete
and
those
ones
those
are
fixed.
Certain
the
changes
looks
very
clean
to
me,
but
I
just
wanted
over
and
along
with
Mark,
which
already
can
spend
some
time
on
that.
If
he
has
this
week,
probably
I
think
that
you
can
wait
for
that.
A
H
D
Okay,
apart
from
that,
regarding
probably
we
can
talk
about
this
PR
with
I
see
miscellaneous
copyright
is
being
dis
added
as
one
of
the
items.
B
B
H
A
bit
strange
to
me
to
see
different
different
rules.
That
being
said,
I
just
looked
at
Java
go
over
repository
and
we
all
use
the
same,
the
same
sentence
which
is
without
Europe,
so.
B
H
D
Yeah
I
mean
that
that
that's
probably
more
coming
from
the
consistency
across
across
to
other
six
I
remember
when
I
initially
created
a
first
PR
for
this
copyrights.
The
year
was
there
in
that
change,
but
then
I
got
a
request
from
other
members
to
remove
it,
just
because
other
things
are
not
using
it
so
that
time,
I
removed
that
year
from
that
copyright
notice.
Okay,
that
is
what
something
which
we
are
following.
So
probably,
let's
stick
to
that
when.
H
D
Yes,
the
only
thing
from
I
was
the
what
the
trace
ID
should
be.
It
should
be
ignore
that,
or
should
we
remove
or
hear
from
both
of
them
right?
Okay,
so.
B
C
D
H
H
Yes,
well,
we
we
don't
have
to
typically
I
saw
that
some
checks
or
NCI
are
required
and
some
we
didn't
say
I
are
required
and
some
are
not.
But
in
any
case
we
we
tend
to
to
make
sure
that
everything
is
clear
when
we
merge
anyway.
So
it's.
D
C
B
D
H
D
Yeah,
it
won't
be
somewhere
in
the
code
if
I
understand,
to
make
it
mandatory
right,
yeah,
I.
Think,
probably
if
it's
not
required
in
this
PR
I
mean
as
correct
me.
It
is
not.
No
change
is
required
for
this
in
the
pr
that
should
be
something
we
have
to
separately
do
once.
It
is
merged
in.
B
B
B
H
B
D
D
D
I
mean
they
asked
about
that
thought
before.
Really
we
make
it,
we
move
the
company
DTW.
So
if
we
move
just
this,
like
with
this
header
file
as
a
separate
sub
module
repo,
so
that
we
don't
have
the
third
party
license
code
in
the
main
report,
I
mean
that
something
which
we
can
do
it
before,
really
moving
the
complete
DBW
but
yeah.
How
can
we.
D
H
D
Yes,
exactly
this
is
something
I
think
I
think
would
make
that
repo
as
a
sub
module
and
I
think
it
would
be
cleaner
approach.
As
of
now
till
we
till
we
move
the
complete
DW
in
contrib.
B
D
Enforce
computation
is
done,
update,
bezel
support
for
zygard.
This
I
think
we
don't
need,
as
we
are
deprecating
again
so
yeah
I
think
we
should
be
good
to
close
it.
We
left
the
actual
owner
of
the
pr
to
close,
otherwise,
we'll
make
a
note,
add
benchmark
test.
Yeah
I
think
I
already
discussed
this.
This
Sr,
yes,
the
arrival
building,
Hotel
CPP
extension
I
think
this
is
something
Mark
is
on
your
plate.
Wherever
you
want
yeah.
D
Yeah
exporter,
sanitizes
invalid
characters,
yeah
I
think
it
is
with
you
yeah
yeah
I
mean
just
talking
about
Primitives
I.
Think
I
did
some
design
changes
on
that
I.
Think
probably
you
would
have
gone
through
that.
You
already
approved
that
PR
yeah.
D
Yeah
that
will
help
us
at
least
the
duplicate
to
event,
which
we
were
seeing.
The
Matrix.
D
But
the
only
only
side
effect
of
that
change
would
be
that
now,
every
time
a
fetchik
or
pull
request
is
coming
from
Prometheus
agent
or
the
server
that
time
there
would
be
some
delay
because
we
are
going
to
at
that
time.
We
are
going
to
do
actual
collection
of
The
Matrix
earlier.
D
D
If
we
can
use
try,
I
was
trying
to
use
to
use
the
ghs
here
or
IO
for
some
caching
thing,
but
I
think
probably
I'll
close
this
PR
open
tracing,
shim,
yeah
I
think
I
did
some
review,
but
I
don't
I
mean
I
think
it
was
not
a
more
power
review
would
like
to
spend
some
time
on
this.
This
guy
has
been
very
patient
or
on
this
PR,
so
probably
I
think
I'll
spend
some
time
to
speak
to
review
it
more
thoroughly.
D
This
also
I
think
fixing
C
make
to
build.
G-Test
I
have
approved
it,
but
I
see
Tom
is
doing
some
discussions
with
the
order
of
the
pr,
so
we'll
wait
once
Tom
approves
it.
I
think
we
should
be
good.
D
You
yeah
fix
open
Telemetry
CPP
dependency
in
HCI,
yeah,
I
thought
I.
Think
I
had
a
thought
about.
It,
I
mean
probably
I
think
we
I'll
say
that
let's,
let's
keep
this
open
and
let's
see
if
our,
if
our
existing
CI
breaks
again
and
then
probably
we
can
look,
look
think
about
moving
it
to
the
docker.
D
Then
the
change
will
not
help
us
to
really
to
speed
up
the
complete
throughout
your
tests.
So
I
mean
that
that's
my
opinion,
probably
that
it's
a
good
change,
definitely
moving
to
the
docker
and
then
building
everything
in
a
more
controlled
way.
But.
B
D
Yeah
I
welcome
that
this
button,
implement.tlp
yes
to
TLP,
HTTP,
SSL,
yeah
I,
see
the
marketing
lots
of
changes,
but
I
mean
just
waiting
for
the
the
split
Piers.
H
So
a
couple
of
things
so
I
still
need
to
to
split
all
the
all
the
environment
code
in
a
different
PR
to
to
review
it
and
also
I,
have
an
update
on
testing.
I
did
manage
to
do
functional
tests
talking
to
a
real
Hotel
open
telemetric
collector,
both
in.
B
H
And
there
is
a
in
the
maintenance
tests,
visual
section
to
generate
research
for
testing
and
another
one
trying
the
functional
test.
H
H
And
as
part
of
that,
I
found
a
bug
in
the
async
implementation
for
verexplorer,
which
is
in
the
in
the
issues.
B
H
I
debugged,
it
was
just
when,
when
doing
ending
with
span,
which
caused
an
export
and
for
some
reason,
your
export
is,
is
failing
right
away.
D
No
I
think
it's
okay,
probably
let's,
let's
first
see
this.
Yes,
so
I
think
this.
This
looks
I
mean
if
it
is,
this
is
working
I,
think
probably
it's
they're
able
to
test
it.
I
think
that's
a
good
task,
and
okay,
probably
I,
want
to
definitely
have
a
look
into
this
PR,
the
whatever
just
want
to
see
once
well,
once
it
ready
I'll
review
it.
H
The
adverb
you
will
see
the
test
results
to
see
how
much
is
tested.
B
H
B
B
B
F
H
H
If
you
scroll
totally
at
the
end,
you
will
see
the
result
of
all
tests.
D
H
So
I
will
split
the
all
the
environment
variable
in
a
separate
PR
to
get
that
Trunk
by
trunk
because
of
the
size.
D
D
D
D
H
D
D
Based
on
that
assumption,
only
I
made
a
comment
that
in
that
case
we
should
have
an
explicit
Force
for
Force
flush,
which
can
be
called
by
the
application
before
doing
a
shutdown
but
yeah,
probably
I,
think
I
would
have.
You
can
do
that
once
more.
H
I
have
a
question
also
on
the
async
mode.
Today,
it's
a
it's
a
feature
which
is
enabled
at
compile
time,
so
we
build
either
in
sync
mode
or
in
async
mode.
A
H
In
the
code,
I've
seen
that
the
HTTP
client
is
capable
of
doing
both
so
I
was
wondering:
should
it
be
a
compile
time
switch,
or
should
it
be
that
the
client
can
decide
whether
to
send
synchronously
or
async
asynchronously.
D
The
the
only
reason
why
we
have
kept
it
as
a
compile
timer.
That
is
something
which
is
a
preview.
B
D
We
kept
it
as
a
preview
is
not
that
it
is
not
stable,
but
it
was
kept
as
a
preview,
because
the
the
export
the
specification
for
the
exporter
export
is
that
it
should
return
back
only
after
doing
all
the
exports
if
I
understand.
That
is
not
what
we
are
doing
in
the
async
mode.
So
if,
when
the
processor
called
in
async
mode,
when
a
processor
called
the
export
method,
when
it
returns
back,
it
is
still
doing
the
export
exports
are
still
happening.
D
So
it's
not
specific,
it's
not
compliant
to
the
actual
specification,
and
that
was
the
reason
why
we
kept
it
as
a
as
a
as
a
compiled
and
flag
just
to
be
compliant
with
the
specification
yeah,
so
specification
more
talks
about
the
synchronous
upload
so
whenever,
as
per
the
specification,
if,
if
the
export,
if
the
exporter
export
method
returns,
that
means
that
all
the
all
these
is
all
these
pans
should
have
been
exported
all
or
should
have
been
either
failed
or
exported,
and
then
it
should
return.
So
the
processor
will
not.
D
Processor
does
not
have
any
pump,
I
mean,
and
that
was
next
export
should
processor
to
do
should
call
other
the
subsequent
export
only
when
first
returns
successfully
or
unsuccessfully.
So
that's
that's.
Something
was
the
reason
why
we
kept
it
as
a
flag.
Okay,.
H
B
H
Once
this
preview
flag
is
removed,
will
someone,
as
still
as
have
a
capability
to
send
in
synchronous
mode
or
will
we
have
to
use
async.
B
A
D
Okay,
I
think
probably
that's
your
comment.
Your
question
is
valid
now
earlier
this
was
not
asynchronous
earlier.
It
was
mentioned
that
the
export
should
return
successfully
for
a
given
set
of
events,
and
then
the
subsequent
export
should
start,
but
now
I
think
they
have
made
it
explicit
that
it
can
be
asynchronous
also
so
yeah
in
that
case
I.
Think,
probably
you
will
it's
a
very
correct
I
mean
in
that
case
we
should.
It
should
be
configured
configuration
driven,
not
the
compile
time
driven
once
once
we
remove
this
flag.
B
D
D
B
We
can
have
of.
B
D
So
when
then,
this
async,
when
the
HTTP
client,
the
async
functionality,
was
added
in
HTTP
client
at
that
time
the
specification
was
not
ready
for
that
asynchronous
mode.
Specification
still
said
that
it
should
be
synchronous,
but
I
think
over
the
course
of
time.
The
specification
changed
when
we
didn't
modify
it
out
so
yeah.
D
Out
of
boundary
exception,
yeah
I
think
this
is
a
valid
issue,
probably
I
think
so
we
can
pick
it
up.
Really.
You
have
been
but
I
think
it's
just
the
test
case,
which
is
varying
so
nothing
very
urgent.
D
D
So
probably
I
think
if
everybody
is
fine,
it
should
be
good
to
close
it
as
something
which
we
cannot
support
and
the
replication
should
ensure
that
duplicate
values
are
not
getting
added.
Don't
forget
the
values.
The
same
key
are
not
getting
added
again
so
and
I'll
close
it.
Unless
somebody
has
any
concern,
please
let
me
know
now.
D
D
Do
you
think
it's
related
to
any
change
which
you
have
done
recently
or
something.
B
B
D
Yes,
no
it's
not
six
months
but
I
think
I
said.
D
Yeah
except
carry
into
injector
and
next
after
interfaces,
I've
tagged
it
as
a
help
wanted
in
case
somebody
wanted
to
change
it,
or
we
can
also
no
I,
don't
think
we
should
changing
it
into
a
into
a
API
breaking
way.
So
it
should
be
something
if
we
can
do
it.
H
So
I
looked
at
a
bit
at
it,
so
it's
it's
easy
to
add
a
new
interface
for
the
injector
load
and
extractor
alone,
that
which
is
fine.
But
then
at
some
point
there
is
a
single
turn
to
get
the
yeah
and
it's
unclear
whether
we
can
support
support
both
a
Singleton
for
the
text,
map
range
or
or.
D
D
H
D
B
H
End
up
with
code
unimplemented
in
your
own
implementation,
which
is
sort
of
weird,
but
it's
it's
working,
yeah.
D
Yeah
I
mean
I,
understand
it's
it's
a
bit
weird
having
having
these
methods
for
for
the
interface
that
it
is
not
really
required.
D
D
Okay,
I
think
I
did
comment
it
last
week
yesterday,
so
he
was
saying
that
he
was
not
seeing
that
his
metric
results
so
with
the
recent
change
in
The
Primitives,
if
he's
just.
D
If,
if
he
is
able
to
see
the
results
and
also
I,
think
eventually,
we
can
just
fix
the
documentation
for
the
port
numbers
and
I
I'll
raise
a
chain
for
the
documentation,
because
right
now
the
documentation
is
also
pointing
to
the
earlier
implementation.
But
it
should
anyway
change
for
the
new
implementation.
So
as
part
of
that,
I'll
also
fix
the
port
number.
C
B
D
D
D
B
D
Okay,
I
think
this:
is
that
honest
with
you.
B
D
D
D
Working
on
that
and
I'll
add
another
based
on
the
log
CPI
and
logs
SPI
and
sdgation
stabilization
create
another
issue
to
create
a
log
API
which
probably
we
have
to.
We
could
do
something
specific
to
our
sync,
some
kind
of
convenient
Library,
it's
more
disc
I,
have
put
the
discussion
here.
If
somebody
want
to
look
more
into
that,
what
exactly
we
are
talking
about,
having
our
own
implementation
of
logs
API.
D
D
B
D
Looking
into
these
issues,
I
think
just
wanted
to
call
out
next
couple
of
weeks.
I
have
some
internal
Microsoft
commitments,
I'm
I'll
be
I'll,
be
there,
but
I
may
not
be
doing
I
mean
I
won't
be
I'll,
be
reviewing
the
pr
that
probably
looking
into
the
issues,
but
I
would
not
be
very
active
for
next
couple
of
weeks,
but
I
don't
think
it
will.
It
will
be
blocking
in
terms
of
any
of
the
pr
should
somebody
raising
for
the
issues.
So
they
just
wanted
to
talk
to
call
out
that.