►
From YouTube: 2020-09-14 C/C++ SIG
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
B
Hello,
can
you
hear
me
yeah?
I
can
I'm
good,
I'm
figuring
out
my
video
set
up
just
a
moment,
good.
A
C
Portland
there
is
no
city
in
the
world
where
it's
verse
currently
so
foreign
actually
is
the
worst
city
in
the
world.
Currently,
when
it
comes
to
air
pollution,.
A
A
I
I
read
the
newspaper
like
in
in,
like
indian,
some
city
has
like
pm
2.5,
like
2
000
points.
I've
never
seen
that
in
other
places.
A
I
I
know
in
in
beijing
once
it
reached
800
and
in
shanghai.
I
I've
seen
in
person
like
I,
I
went
to
microsoft
shanghai
office
and
that
that
day
it
was
600
pm
2.5,
oh
wow
yeah.
So
I
think
in
portland
the
current
is
500
and
I
think
it's
almost
toxic
and
and
you
should
like
stay
safe,
never
go
outside.
I
remember,
probably,
five
years
ago
I
went
to
microsoft
shanghai
office
in
the
morning,
and
after
that
I
read,
the
pm.
2.5
is
500.
C
A
Yeah,
we
need
to
wear
a
mask,
feel
free
to
do
that.
I
think
it's
important
to
get
healthy.
A
A
So
I
only
have
one
topic:
that's
coming
from
me,
so
I
want
to
explore
the
upcoming
work.
I
I
know
like
jordy
johannes
you
mentioned,
like
I
plan
to
do
some
like
help
on
the
project
management,
I'm
totally
open
to
that,
and
actually
we
look
for
some
like
triager
role.
I
think
we
can
explore
that
I'm
open,
so
I
can
sponsor
that
people
on
top
and
for
my
topic,
so
I
want
to
explore.
I
I
think
the
open
climatry
spec
is
almost
locked
down
for
the
tracing
part.
A
You
can
easily
map
that
so
now,
the
the
so
that
has
been
driving
that
discussion
and
this
morning,
what
I
heard
is
we
have
concluded.
So
you
can
look
at
the
pr
there's,
a
old
type
of
pr
talking
about
the
status
and
the
goal
is
to
reduce
the
complexity.
So
I
believe
what
we'll
end
up.
Having
is
just
three
status.
One
is
onset,
it
means
it's
undefined.
A
We
don't
know
another
one
is
set,
it's
basically
a
set
of
status
to
be
not
okay
and
the
instrumentation
library
should
do
that,
and
then
it
is
a
value
okay,
which
can
be
overwritten
by
the
application,
developer
or
the
operator.
So
basically
people
who
consume
those
they
can
claim.
Oh,
the
library
is
saying
it's
not
okay,
but
from
my
perspective
I
think
it's
okay,
because
it's
underlying
detail,
for
example,
I'm
doing
http
retry
and
I
know
it's
not
fun.
A
It
doesn't
hurt
my
business
so
I'll.
Just
ignore
that
and
don't
report
that
as
error
and
give
me
a
page
or
something
so
that
seems
to
be
very
reasonable.
Based
on
my
observation,
like
on
the
other
sdk
work
streams,
so
based
on
this,
I
think
the
tracing
spike
is
almost
ready.
We
can
rely
on
that.
A
So
to
me,
I
think
it's
a
the
right
time
for
us
to
stabilize
the
tracing
part
any
like
missing
thing:
we're
going
to
look
at
the
matrix
to
get
all
the
facial
parity
and
we'll
need
to
take
a
serious
test
on
the
performance
and
the
memory
so
for
open
time
redundant.
We
have
done
that.
We
know
exactly
how
many
bytes
we
allocate
for
each
critical
operation
and
for
c
plus
class.
I
think
we'll
just
do
the
same,
and
here
we
have
tom.
A
So
probably
the
first
time
you
joined
this
meeting
and
tom
has
been
joining
the
open,
telemetry
spec
meeting
and
he's
taking
a
lead
on
the
performance
and
concurrency
requirement
for
the
spec,
so
it
will
be
used
across
the
languages.
I
think
class
class
will
be
a
good
place
for
us
to
to
do
some
experiment
and
use
that
to
drive
this
back
so
tom.
You
want
to
say
hi
to
everyone,
hello,
everyone,
nice
to
meet.
You.
A
So
tom
yeah,
so
so
tom
has
been
in
microsoft
for
many
years
and
he
has
experience
working
on
the
low-level
stuff
like
the
operating
system.
Some
of
the
kernel
mode
stuff
and
like
cpu
also
is
a
is
an
expert
on
doing
all
the
instruction
level
optimization,
and
I
hope
it
will
be
a
great
addition
to
this
project.
A
And
thanks
for
the
sponsorship,
johannes,
oh
no
problem,
it's
awesome
to
have
you
on
board
tom
thanks
yeah.
So
so
what
we'll
be
working
on
the
tracing
stabilization
and
I
I
think
the
goal-
probably
I
would
say
like
by
end
of
this
year-
we'll
have
the
the
release,
like
the
initial
release
of
the
open
time,
3c
plus
pass
with
a
pretty
confident
tracing
api
and
sdk
implementation,
and
also
the
the
building
exporters.
A
I'm
not
sure
if
we're
going
to
have
all
the
exporters
as
required
by
the
spec,
but
at
least
I
would
expect
with
support.
We
have
a
good
support
of
one
of
the
most
famous
tracing
speller,
besides
oklp
foreign
and
and
for
the
other
scenario
like
currently.
If
you
look
at
the
spec,
the
exporter
is
not
re-entrance
safe,
which
means
no
matter
is
the
simple
export
processor
or
the
sim.
A
Like
the
batch
processor,
the
exporter
can
be
only
called
at
once,
even
if
you
have
multiple
threads
and
in
some
of
the
scenarios
like
linux,
there's
lttng
and
on
windows,
there's
etw,
so
those
high
performance
scenarios,
people
don't
use
any
batching
and
they
don't
like
all
they
do.
Is
they
take
the
data
and
immediately
they
call
the
operating
system?
So
it's
a
cisco
and
after
you
return
from
the
cisco,
you
know
the
icdk
is
no
longer
owning
the
data.
A
The
data
is
owned
by
the
underlying
operating
system
and
we
have
done
the
the
work
in
the
open
time
tree
download,
sdk
to
test
the
performance
and
I've
seen
like
millions
of
spends
per
second.
So
for
c,
plus
plus.
I
I
think
for
the
high
performance
scenario.
We
probably
need
to
have
a
similar
thing.
It's
not
going
to
be
against
the
spec,
but
more
like
an
addition
that
supports
concurrency.
So
I
talked
with
alolita
from
aws,
and
it
seems
like.
Aws
is
also
interested
in
that
scenario.
A
B
A
B
I
do
have
working
implementation
done
in
a
very
crude
way
right
now,
I'm
just
implementing
the
api
instead
of
doing
the
proper
exporter,
I'm
going
through
straight
through
api,
and
it
does
work
for
the
standard,
etw
events
when
I
mean
standard
idws
like
pretty
much
key
value
pairs,
that's
how
etw
works,
but
there's
also
a
way
for
us
to
incorporate
a
custom
protocol
that
allows
marshalling
like
tree
objects
like
pretty
much
message
pack
or
json
style
objects,
and
that
itself
means
that
event
carry
is
a
like
binary
payload,
that
an
agent
that
monitors
for
incoming
events
should
take
and
decode
into
a
structured
nested
option
right.
B
So
there
are
two
alternatives
for
the
first
one.
Whatever
I
have.
Unfortunately,
right
now
depends
on
a
proprietary
header
that
we
are
trying
to
release
an
open
source,
and
our
pm
on
our
end
is
trying
to
help
me
with
this.
I
do
not
anticipate
issues
with
releasing
it
it
it's
not.
No,
it's
it's
not
the
rocket
science,
not
the
top
secret
source.
Similar
stuff
is
already
done
in
c
sharp,
for
example,
it's
just.
B
We
are
also
share
a
few
microsoft
internal
apis
for
that,
but
at
the
same
time,
like
the
second
part
that
they
mentioned
structure
to
object,
it
doesn't
really
depend
on
any
proprietary
source.
Even
now
we
can
go
and
implement
some
sort
of,
like
I
don't
know,
otlp
to
message,
part
message
pack
to
tlp
transform
and
even
if
interns
start
right
now,
they
do
not
have
dependency
on
any
proprietary
software.
So
I
think
there
are
a
few
options
how
we
can
make
it
happen
within
the
next
few
weeks.
A
Yeah,
so
the
update
from
my
site
is
for
for
this
one.
It
seems
to
be
attracting
some
of
the
interest
from
the
community,
so
their
cncf
internship
project,
we
might
be
able
to
get
someone
from
the
community
just
join
us
and
help
on
that,
and
I
think,
having
this
like
concurrent
or
engine
safe
exporter,
will
help
us
to
drive
the
extreme
performance
and
also
how
has
to
drive
the
memory
usage
and
all
the
concurrency
things
so
hope.
Hopefully
that's
an
addition
and
for
the
etw
exporter.
A
B
Just
wanted
to
add
that
my
understanding
is,
we
do
have
internal
customers,
not
just
azure
services
customers,
but
like
wider
microsoft,
work,
customers
that
would
be
willing
to
onboard
280w
export.
D
B
Is
a
separate
story
remember
when
we
were
originally
discussing
c
api,
we
said
that
perhaps
c
api
deserves
a
separate
community
yeah.
B
Exactly
yeah,
we
get
questions
like
from
driver
folks,
they
say
well
absolutely
zero
c,
plus
plus
just
just
see.
What's
in
it
for
me
and
well,
we
don't
really
have
an
answer
right
now.
Yeah.
A
C
I
have
a
question
to
the
concurrent
spam,
processor,
yeah
and
also
regarding
like
this.
How
does
it
look
in
this
back
yeah,
but
I
mean
I
think,
even
with
the
currently
out
I
mean
it
will
be
possible
for
vendors
just
to
provide
their
own
processor
that
basically
are
not
just
providing
the
exporter
yeah,
just
providing
processor
and
exporter
and
basically
having
this
a
small
chunk
and
putting
it
in
which
would
fulfill.
C
I'm
curious
what
what
comes
out
there
for
this
pack,
because
it
seems
to
be
a
use
case,
more
common.
A
Things
yeah,
so
the
goal
is
to
try,
try
it
in
a
few
like
languages.
So
start
it
from
the
c
sharp,
and
I
think,
given
c
plus
plus,
is
a
high
performance
scenario.
We
probably
will
try
it
here
and
then
we
can
see
if
gold
will
be
interested
and
if
we
have
enough
learning,
we
can
try
to
put
that
in
the
spec.
The
reason
I
I
think,
currently
it's
not
high
hard
putting
this
back
is
because,
as
I
mentioned
earlier,
so
the
spec
is
trying
to
get
the
tracing
part.
A
C
B
Sounds
great
because
I
think
we
should
always
keep
that
option.
Even
if
we
provide
a
full-fledged
sdk,
we
should
still
keep
that
door
open
for
somebody
to
implement,
potentially
a
header,
only
implementation
of
the
sdk
if
they
really
need
to.
If
you
I've
been
poking
around
with
a
few
customers
there,
some
of
them
actually
love
the
header
only
id,
and
it's
like
due
to
simplicity,
no
binary
artifacts,
you
just
point
them
to
whatever
source
headers
yeah.
C
And
I
think
I
would
love
some
reviews
from
those
two
pr's,
because
I
think
when
we
have
those
merged,
I
would
feel
in
a
comfortable
point
actually
telling
people
you
can
start
experimenting
with
open
the
diametry
cpp,
because
we
have
orders
banned
parenting
in
place.
Then,
and
the
api
I
mean,
like
the
scope
thing,
with
the
managing
the
active
span.
C
That's
still
like
a
bit
of
an
open
question
for
me,
so
it
becomes
your
input
there,
but
giving
that
that
is
a
solution
there's
merged.
I
think
we
have
the
api
pretty
much
in
place
and
I'd
be
curious.
Just
seeing
people
starting
to
experiment
with
that
and
see
what
input
we
got
because
functionality
devices
would
be
there
like,
we
would
be
able
to
export
to
some
vendor
back
end
and
see
correctly
parented
span
there,
including
attributes
yeah,
and
I
think
that's
a.
C
A
A
So,
okay,
so
coming
back
to
my
next
topic,
so
currently
we
don't
have
a
good
internal
troubleshooting
log.
If
we
face
some
problem,
will
will
either
throw
exception
if
the
user
allow
us
to
throw
or
we'll
just
terminate
the
process
and
for
some
non-critical
issue
of
example,
the
exporter
is
not
configured
correctly,
there's
no
general
guidance.
How
people
are
going
to
report
arrow
and
the
open
telemetry
in
general.
A
We
have.
This
idea
of
the
sdk
should
be
able
to
do
internal
logging,
and
that
brings
two
questions.
One
is
the
internal
logging
itself.
The
second
one
is:
do
we
have
some
idea
about?
How
do
we
build
the
logging
api?
So
things
I
can
share
here
is
for
net
we're,
taking
different
approach
so
for
customers
who
want
to
write
logs,
the
answer
is
just
use
the
dotnet
standard
logger,
which
is
the
I
logger
and
for
the
sdk
internal
troubleshooting.
A
They
will
use
the
the
internal
like
class
provided
by
the
open,
telemetry
api,
which
gives
people
ability
to
write
internal
log
and
that's
also
based
on
a
subscription
model.
So
people
can
either
tell
the
internal
system
hey.
Please
don't
owe
the
logs
to
etw
on
windows
or
a
plan
file
on
linux,
or
they
can
even
turn
on
turn
off
the
log
at
any
certain
moment.
A
So,
in
order
for
this
project
to
scale
better,
I
think
at
some
point
we
will
need
to
have
this
internal
logging
for
sure,
and
another
thing
is
like
we
need
to
figure
out.
Do
we
want
to
invent
a
logging
api
or
just
borrow
something,
and
do
we
want
this
logging
api
to
be
the
same
customers
when
they
try
to
use
logging
or
we'll
just
have
two
different
set
of
apis?
One
is
reserved
for
the
open,
telemetry
project
itself.
A
Another
one
is
the
value
for
customers,
and
should
we
just
borrow
some
existing
stuff
or
invent
something
from
scratch.
So
those
are
the
things
to
start
think
about,
and
we
have
some
at
least
I
have
some
learning
from
the
open
time
tree
down
that
so
I
probably
can
can
discuss
with
tom
internally
and
then
like
like.
I
want
to
see
like
johannes.
A
Oh,
oh,
I
could
definitely
be
interested
yeah,
yeah
and,
for
example,
like
how
neuralic
expect
like
do
you
expect
if
the
customer,
misconfigured,
ick
or
the
sdk
is
not
functioning
correctly,
you
would
want
them
to
go
to
the
local
machine
and
collect
the
log
or
there's
a
scenario
where
the
sdk
will
send
a
constant
stream
of
internal
log
through
the
agent,
and
you
can
turn
on
turn
off
or
change
the
velocity
level.
What
are
the
scenarios
required?
So
probably
we
can
start
from
the
scenario.
What
do
we
expect
and
then
drive
some
initial
design.
A
See
that
information
yeah
and
we
we
need
that
freedom
home
at
minimum.
We
need
some
heartbeat
to
know
if
the
sdk
is
not
working
at
all
or
it
is
working.
We
also
get
some
usage
data,
for
example,
which
version
of
the
sdk
the
user
using
and
are
the
latest
version
or
not,
and
then
for
the
sdk.
Some
of
the
critical
configuration
like
what's
the
sample
rate?
Is
there
any
data
job
like
any
like
data
loss
happened?
Is
there
like
any
crash
or
something?
So
this
is
also
critical.
A
C
A
A
I
think
it's
all,
depending
on
how
we
implement
that,
if,
like
a
proper
implementation,
could
avoid
this,
but
in
like
in
donet
it
is
just
hard
because
we
use
I
logger
analog
itself
is
very
pluggable,
so
people
plug
in
something
and
then
they
feed
back
the
data
into
the
same
stream.
Then
we
need
to
like
the
only
way
we
can
detect.
A
If
there's
a
loop
is
to
tell
from
the
call
stack
and
it's
expensive
right,
you
don't
want
to
travel
the
call
stack
for
every
single
log
so
or
like
put
something:
that's
right:
local
storage,
it's
just
too
expensive
for
us.
In
this
way,
we
decided
okay,
we're
going
to
have
our
smaller
version
of
the
logging
api.
That
doesn't
give
you
a
lot
of
flexibility,
and
it
will
give
you
the
minimum
feature
but
required
for
this
internal
log
and
still
you're
able
to
subscribe
in
a
slightly
different
way.
But
the
concept
is
the
same.
A
In
this
way
we
can.
We
can
bring
the
problem
into
a
smaller
problem
instead
of
making
a
problem
like
making
a
solution
that
seems
to
be
working
by
introducing
yet
another
problem.
This
is
the
the
current
thinking.
C
Is
probably
in
the
start
using
a
third-party
logging
framework
or
logging
tool
just
also
because
the
logging
api
just
still
seems
pretty
far
out
the
thing
I
think,
unless
we
release
like
open
telemetry,
c,
plus
plus
restraining
support,
we
need
some
kind
of
logging
facility
there,
but
I
see
there's
an
option
that
maybe
at
some
point
in
the
future.
We
can
just
switch
from
this
like
digital
third-party
library
and
use
our
blogging
tools.
A
Yeah,
I
think,
as
long
as
we
have
that
level
of
abstraction,
either
a
micro
or
some
like
function
abstraction,
we
can
easily
replace
that
yeah,
okay
and
the
any
any
topic
here
or
I
can
move
to
my
last
one,
okay.
So
the
last
small
thing
I
I
think
alolita
asked
me
so
just
for
contacts
already-
is
the
evangelist
and
has
been
driving
the
effort
from
aws
side,
so
she's
working
with
mark
carter
from
aws
on
the
cloud
watch
and
integration.
A
So
I
think
it's
not
ready
to
harden
that
part,
and
I
understand
that
a
lot
of
like
extra
memory,
allocation
and
contention,
and
I
think
we
probably
will
live
with
that
and
by
making
it
clear
in
our
documentation
we're
going
to
have
the
first
release
we're
going
to
tell
people.
This
mattress
part
is
just
to
measure
the
end-to-end
workflow,
but
it's
not
supposed
to
be
a
high
like
a
fish,
highly
efficient
production
thing
that
people
should
use.
So
I
want
to
get
feedback
from
from
you
guys.
A
What's
your
take,
do
you
think
you
want
to
see
more
progress
on
the
matrix
part
or
you
think
the
the
current,
like
my
thinking,
putting
logging
like
before
matrix,
would
more
align
with
your
understanding.
C
One
thing
about
other
bodies:
when
you
say
yeah,
you
expect
drastic
changes
in
the
matrix
part.
Those
are
changes
also
in
the
api
like
in
the
metrics
api.
Yes,
I
think
when
we
expect
dramatic
changes
in
the
api,
it
doesn't
make
sense
to
do
much
hardening.
There.
A
Yeah,
it's
so
smart
yeah,
so
for
the
for
the
matrix
api,
the
spec
part
I've
seen
there
are
a
lot
of
like
high
priority
issues
and-
and
I
quickly
scan
through
some
of
the
issues
I
think
clothing
them
would
require
significant
changes
and
even
those
are
breaking
changes.
I
think
hardening
them
will
cause
a
lot
of
throwaway
work
for
us.
E
A
E
A
C
And
basically,
I
set
it
up
based
on
the
matrix
that,
like
in
the
spec
filled
out,
and
it
just
created
issue
for
all
the
parts
that
were
missing,
and
I
also
like
moved
other
issues
in
here.
That
kind
of
belonged
there
and
in
progress
you
see
here.
Those
are
the
two
issues
I
have
props
prs
up
for
the
first
one
is
replacing
the
static,
gilbert
context
handler
yeah.
C
The
second
one
is
adding
functionality
to
marco
spain
is
active
and
not
do
it
by
default.
On
spain.
A
C
We
should
do
there's
nothing
special
to
talk
about
it's
just
that
this
is
that's
not
yet
an
exhaustive
list
that
just
created
this
from
this
yeah
matrix,
basically,
that's
just
a
functionality,
and
that
doesn't,
I
think
this
project
doesn't
include
any
performance
work
or
that
will
definitely
be
a
separate
separate.
That
should
be
definitely
grouped
in
a
separate
project,
because
otherwise
this
one
gets
blown
up.
E
All
right,
so
so
the
outcome
of
this
is
that
we'll
be
in
conformance
with
the
spec
and
then
maybe
we
need
another
project
where
we
are
at
a
point
where
we
feel
like.
We
actually
have
a
an
alpha
or
beta
release
like
some
some
version
and
then
we're
adding
in.
We
actually
need
to
add
in
the
logging
capability.
C
A
Yeah,
I
think,
for
alpha
release.
This
is
already
a
good
enough
list
and
I
probably
will
work
on
some
of
the
the
ci
infrastructure
work.
For
example,
like
I
noticed
we
don't
have
a
a
lot
of
like
like
consistency
in
the
code
based
people
for
different
line
endings.
They
have
like
typos
those
things
previously.
I
had
some
pr
for
the
spec
and
for
other
sdks
we'll
just
go
and
like
borrow
those
like
ci
jobs
and
and
checking
here.
C
And
also
regarding
this
board,
I
just
said
anybody
who
has
us
an
issue
or
something
that
they
want
to
have
in
the
alpha
release.
Just
put
it
on
them:
open
forum,
okay,
yeah.
A
Yeah,
so
so,
probably
like
johannes,
I
I
think
I
I'm
happy
to
sponsor
like
started
to
have
the
triager
permission
and.
A
A
Permission
from
me,
so
the
the
product
first
like
a
proposal
from
from
the
maintainer
to
grant
to
grant
extra
information.
So
our
awesome
appearance,
just
wonderful,
I'm
not
putting
you
as
a
as
a
trader
without
you,
knowing
because
I've
been
putting
a
treasure
without
me.
Agreeing-
and
I
appreciate.