►
From YouTube: 2020-06-22 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
A
A
A
B
C
I
think
we've
an
update
on
the
sampling
API,
so
the
first
PR
that's
in
was
the
header
definition
and
there's
been
a
lot
of
comments
on
that.
I
think
that
one
is
blocked
from
merging
by
another
PR
for
attributes,
and
there
are
two
work-in-progress
PRS
after
that
for
the
always
on
and
always
off
samplers.
B
D
C
D
E
B
B
F
B
Open
issues
when
I
look
at
that
I
think
this
is
good
to
merge
this
block,
many
others
and
I
won't
see
any
real
authentication.
Although
we
have
some
common,
like
the
base
around
like
the
ownership
and
around
the
know,
standard
versus
standard,
but
I
think
this
is
something
we
can
address
in
subsequent
er's.
We
want
to
merge
that
now.
I.
F
G
F
G
As
long
as
we
keep
the
shoe
open
because
reasons
why
I'm
worried
about
this
is
like
I
have
an
example,
reference
implementation
that
just
takes
the
API
surface
and
implements
it
CW
export.
So
for
that
flow,
and
if
we
eventually
provide
an
80
double
exported
using
exporter
interface,
there
might
be
quite
a
few
optimizations
to
avoid
the
unnecessary
copies
for
the
keys
and
values.
G
B
I
think
this
is
related
to
the
comments
from
all
in
and
regarding
the
the
stream
bill.
I
think,
probably
stream
vo
is
only
intended
for
the
like
API
parameters.
As
long
as
you
know
that
API
is
not
going
to
do
anything
later
on,
so
it
will
take
the
stream
bill.
Do
the
work
and
return
control
and
after
that,
nobody
would
access
that
value
anymore.
G
Yes,
in
part,
there
is
a
change
that
I
posted
for
you
just
now
about
that
in
comparator,
to
string
you,
because
in
the
standard
sto
I
breathe,
is
a
comparator
operator
on
it
and
that
allows
to
use
no
establishing
view
as
a
key
in
the
map
and
then
basically,
somebody
who
passes
set
of
values
either
to
add
event
or,
as
attributes
may
potentially
just
avoid
the
unnecessary
name
like
the
container
that
represents
the
values.
Is
then
assuming
it's?
G
B
A
good
question,
so
here
you
go,
they
have
more
than
she
passed
past
experience.
I
am
probably
the
old
school
sleep
has
class.
So
my
question
is
people
use
stream
view
I
expect
them
to
have
a
struct
instant
access
after
you
gave
that
as
a
parameter
and
after
that
the
control
is
returned
back,
so
they
don't
claim
any
memory
ownership.
My
question
is:
if
we're
going
to
put
stream
wheel
inside
a
like
part
of
a
class
like
a
member
or
a
structure,
is
there
some
convention
that
we
can
tell
people?
B
F
B
Better
okay,
in
this
case,
if
we're
saying
it
shouldn't
be
stored,
are
we
still
okay
to
have
stream
view
as
a
member
of
a
class
or
not
Emily,
we
are
saying
in
that
class
is
also
like
it's
likely
cancer.
So
if
this
is
a
view,
that
class
is
also
called
like
spam
deal
or
some
deal,
but
you
got
you
got
to
take
this
consume
that
immediately
and
then
return
control
and
assume,
like
you,
can
never
access
the
data
again
then
I
think
that's
an
established
pattern.
B
C
F
F
G
I'll
open
the
separate
issue,
and
perhaps
we
can
either
have
a
template
method
or
like
I'm,
trying
to
generalize.
How
do
we
make
that
spawn
data
class
usable
in
both
cases?
I
understand
the
case
where
we
need
to
copy
the
value
objects
and
there's
another
case
where
we
may
have
with
and
that's
what
I
would
to
address,
because
in
our
practice
there
is
a
lot
of
other
services
that
are
going
to
be
dealing
with
practically
synchronous.
Exporter
I,
like
super
hyper,
efficient
synchronous
exporter
and
on
that
part,
I'd
like
to
avoid
them.
B
So
I
wasn't
expecting
like
a
long,
PR
or
marketing
document
I've
emerging
into
this
repo,
this
repo,
not
at
least
where
we
host
those
like
design
dogs,
probably
Asia,
or
something
as
long
as
we
can
discuss
and
comment
and
I
wonder
if,
like
we
can,
we
can
split
this
up
into
a
few
PRS.
The
first
one
is
about
the
matrix
API.
B
The
second
one
is
a
gem.
How
like?
How
do
you,
the
matrix,
API
itself,
can
be
splitted
up
into
the
instrumentation
API
and
the
consumption
API
and
the
next
PR
can
be
the
matrix
as
DK
or
implementation,
and
it
doesn't
have
to
be
the
entire
like
one
single
PR,
everything
can
be
just
starting
from
a
simple
example
and
then
try
to
cover
the
other
scenarios
and
currently
I've
seen
some
performance
issue
in
the
API
design.
I
want
to
explore
that
this
is
the
current
status.
I,
don't
know
like
your
miss.
B
F
We
should
I
would
prefer
and
I
think
about
is
in
there
just
start
to
work
like
sketching
out.
The
API
I
think
we
should
like
separate
API
and
SDK
and
start
to
sketch
out
the
API,
because
certain
things
that
are
very
important
for
this
decay
are
not
of
that
much
importance
for
the
API,
yeah
and
already
parts
of
the
API
that
are
pretty
stablish,
that
you
can
take
over
from
trace.
So
like
the
tracer
provider.
F
B
Just
to
make
sure
we're
on
the
same
page,
so
what
I
would
prefer
is
to
follow
what
we
have
like.
We
have
been
doing
with
the
contacts
hard
to
send
a
small
PR
on
API
design
and
then
having
separate
PRS
about
the
implementation,
how
to
connect
with
PR
I--'s
Dalton.
In
that
way,
we
can
have
a
very
like
clear
scope
for
each
PR
make
that
small
and
people
can
come
and
make
a
move
forward.
Yeah.
H
Early
I
mean
thank
you
for
that.
You
know
guidance
because
again
you
harness
and
your
comments
are
very
useful
and
and
will
structure
the
PRS
and
break
them
up
into
focused
areas.
So
I
agree
there
should
be
a
series
of
peers
and
you're
already
starting
to
you,
know,
work
on
the
code
as
well
as
have
been
working
on
the
test
tests.
First,
so
again
we
will
take
a
divide-and-conquer
approach.
B
Ok,
cool
and
for
the
contacts
API
I
got
a
ping
during
the
weekend
and
there's
some
clarification
needed
from
me.
So
I
wonder
if
this
is
something
we
can,
we
can
discuss
in
the
github
er
or
issue
if
it's
hard
I'm,
like
I'm
open
to
spend
time,
but
in
general
I
figure.
This
pattern,
like
figure
out
you
know
when
they
come
to
this
project,
who
prefer
to
schedule
a
separate
meeting
or
the
prefer
to
talk
like
share
video
or
something.
My
worry
is
like
this
project
we
had
people
come
and
people
leave
all
the
time.
B
It's
always
more
efficient
to
catch
someone
in
person
to
clarify
things.
My
worry
is
that
contacts
is
not
captured
in
the
project
and
later
for
them.
If
I
decided
to
move
to
a
different
project,
I'll
take
this
private
knowledge
away
and
that's
becoming
a
job
security.
So
I
would
rather
ask
to
slow
down
a
little
bit
and
clarify
things
in
issue
or
PR
to
make
it
easier
for
the
others
in
the
future
to
come
here
and
understand.
Why
are
we
making
this
decision
and
it
was
the
like
underlying
consideration,
as
I
said,
I'm
open?
B
No
I
mean
in
general,
I
I
got
several
like
like
work
streams
like
pinging
me,
I
know
like
I'm,
trying
to
help
people,
but
sometimes
my
various.
We
have
this
like
site
communication
and
it's
not
captured
and
later
we
come
here,
we'll
have
others
reveal
and
they
don't
have
the
contacts.
So
this
is
a
potential
communication
gap
and
I
wonder
if
we
can
like
try
to
attach
more
to
github
issue,
NPR's
sure
sure
we.
I
B
Yeah,
so
I
replied
on
the
dinner
channel
privately
to
you,
and
my
take
is
if
this
is
a
general
requirement,
I'm
seeing,
for
example,
the
matrix
part
contacts
part.
We
all
need
some
like
data
structure
in
the
api
for
ABI
compatibility
and
if
we
see
the
common
need,
I
would
say
like
eventually
we
need
to
add
that,
and
probably
is
the
right
time
to
explore
that
now,
regarding
like
like,
what's
the
bar
I,
think
we
need
to
have
a
high
bar,
otherwise
we'll
be
replicating
the
entire
istea
into
the
API
package.
B
So
we
need
to
decide
what's
the
right
thing
and
I
would
suggest
that
we
do
decision
based
on
the
following
principles.
So,
first
we
list
all
the
requirements
we
have
because
people
will
come
different
but
like
background
different
work
stream,
they
might
have
different
thing.
I've
seen
like
Jani
is
trying
to
add
something
like
iterable,
that's
trying
to
hide
the
detail,
we
add
vector
or
something
basically
we're
telling
people.
This
is
a
linear
structure.
B
Inside
memory
you
can
access
randomly
and
the
size
is
always
cached
and
some
people
prefer
a
linked
list,
because
the
maintenance
cost
is
lower.
But
the
downside
is,
you
cannot
do
the
random
access
by
index
and
also
it's
hard
to
know
account.
You
can
always
maintain
a
separate
thing.
So
these
are
the
things
I
think
we
should
discuss,
and
probably
we
need
to
like
have
a
place.
I
can
treat
one
issue
to
track
that
the
the
general
feedback
have
for
this
community's
country.
I
think
the
way
we
we
do.
B
Okay,
good
yeah,
so
I
probably
was
started
github
issue
and
out
consolidate
what
so
far
they
asked.
I
I've
got
regarding
the
api
compatibility
and
will
say
like
what's
the
right
balance
and
I
in
there
other
topics
related,
but
not
strictly
like
premier.
Well,
the
first
thing
is
like
how
many
data
structures
need
to
replicate,
and
although
we
designed
not
didn't
structure,
this
is
admin
arrays.
As
max
pointed
out,
we
have
more
and
more
data
structures.
It's
both.
B
Then
we
start
to
create
this
guide
for
people
to
adopt
the
API,
because
the
like,
if
they
use
something
that
the
familiar
is
they
got
version
or
they
need
to
do
the
mental
like
mapping
and
a
lot
of
people.
This
seems
to
be
a
big
like
overhead,
so
you
could
always
like
to
think
about
like
how
much
you
pay
for
a
particular
feature
versus
the
return
you're
going
to
get
from
that.
So
max
were
thinking
like
do.
We
need
to
have
two
different
modes
and
I
understand.
B
G
G
B
Envoy
that
so
we
can
make
something
simple
and
we
Hanks
or
whether
we
need
to
build
additional
layer
to
convert
like
whatever
iced
tea
out
for
our.
This
is
something
high
school
SEO
is
it's
just
too
much.
The
product
will
eventually
become
even
bigger
than
urban
climb
trees.
It
has,
as
they
cells,
probably
not
or
for
the
exposure
for
the
pages.
A
Yeah
not
in
case
of
protein,
the
initial
draft
of
their
designs,
a
chef
is
working
on
the
data
aggregator
and
Internet
in
the
span
processor
we're
trying
to
cover
first
the
scope
and
the
requirements
from
the
and
then
like
go
for
it
for
this
solution.
I
think
this
week,
oh
sure,
the
designs
of
the
specific
component
and
meanwhile
they're,
like
kind
of
prototyping
starting
writing
code,
to
see
what's
missing
our
current
gap.
A
That
max
also
raised
is
that
the
spanish
reboots
that
some
of
them
are
missing
so
we're
trying
to
map
what
are
the
attributes
that
we
need
participate
for
the
tracy
and
then
what's
missing
the
json
file
that
max
provided
is
elements,
but
still
so
it's
important
so
no
matter
the
gap
and
what
should
be
added
in
order
to
have
the
complete
integration
and
also
we
had
question
about
the
work.
Should
we
add
the
two
pages
directory
like
we
had
some
discussions
with
max
as
well,
but
that's
something
that
once
we
start
adding
PRS.
A
B
F
G
Your
harness
I
think
in
some
sense,
gee
pages,
may
benefit
from
ability,
for
example,
go
unconditional.
We
say
that
everything
is
getting
recorded,
so
it's
like
in
some
way
when
we
build
it
on
top
of
the
rest
that
be
the
right
approach,
the
monitor
all
activities
happening,
irrespective
of
whether
we
load
external
plug-in
or
it's
another
exporter,
would
still
want
to
monitor.
G
Everything
in
that
sense
is
like
a
separate
higher
level
pass-through
and
how
do
I
merge
that
I
mean
I
was
thinking
that
some
other
customers
would
also
prefer
to
build,
will
see
pages
without
G
pages
right.
So
in
that
sense,
it's
like
what
would
be
the
best
place.
I
I
see
what
you're
saying
as
if
it's
one
of
the
exporters,
but
if
it's
the
monitoring
facility
for
all
possible
exporters
more
privileged
than
the
rest,
what
could
be
the
right
place
so.
B
G
G
G
B
Probably
more
aligned
with
there
with
this
with
this
bag
and
Donna
has
a
little
bit
different,
so
Donna
her
in
Java.
They
have
this
like
in
Java.
You
know
the
hierarchy
has
to
follow
the
package
structure,
so
it's
a
little
bit
different
I
believe
conceptually
like
we're
aligned.
So
this
is
the
ice
DK
layout.
This
is
the
I.
G
Just
looked
at
how
it's
done
in
Java,
they
have
control,
so
that's
pretty
much.
What
I've
been
suggesting
should
we
have
something
like
modules
or
country?
Ok
can
share
the
link
in
a
chat.
So
it's
like
open
sis
is
Java,
3
master,
slash
country,
and
then
there
are
modules
under
country.
So
should
we
add
country
been
C++
repo,
no.
B
G
B
J
Special
will
be
working
on
DCP,
exporter
and
Nadia
will
be
working
on
exporter
and
we
started
working
this
week
and
I
have
a
few
questions.
Like
one
question
is
about
the
repo
as
well.
It
looks
like
the
vendor.
Specific
exporters
need
to
be
in
their
own
very
poor,
so
we
are
looking
to
create
a
separate
areas
for
GCP
and
which
will
be
explored.
J
K
Sorry,
yeah
I'm,
Sneha
and
I'm,
currently
working
on
the
GSP
exporter,
and
so
one
question
I
had
about
moving
TCP
to
its
own
repo.
Is
that
how
would
we
like
get
reviews
for
it
like
if
I
wanted
to
add,
like
reviewers
for
my
code,
it's
easier
to
do
when
you
do
put
pull
requests
and
you
can
automatically
add
reviewers
to
your
own
code.
B
F
J
G
G
We
use
only
standard,
C++,
11
features
right
now.
Overall,
I
had
the
proposal
actually
with
respect
to.
This
is
related
to
my
work,
STL
vs.
nice
to
be
that
if
we
could
map
and
alternate
implementation,
maybe
we
have
to
provide
some
options
to
remap
from
noise,
TD
classes
to
a
DSL
or
whatever,
but
I
see
value
in
when
your
static
were
linking
against
certain
standard.
G
From
your
perspective
library,
you
may
want
to
remap
from
no
SD
classes
to
your
classes
to
avoid
the
via
necessary
translations
and
transforms
later
on,
but
in
the
mainline
project
we
deal
with
our
own
OCD.
Our
tests
run
against
an
OST
and
to
remain
compatible
with
other
vendor
precompiled
binaries
I.
Guess
we
have
to
follow
the
same
I
see
so.
J
G
F
Have
a
bit
of
more
like
a
clarification
here,
because
I
think
for
when
the
first
bits
about
using
a
it
is
anything
40
Shebib
and
you
and
your
own
repository.
You
can
basically
you're
free
to
do
what
you
want.
There's
no
Commons
OTL
being
in
the
open
the
remedy,
cpp
repository,
I
I,
don't
prefer
to
see
like
that.
The
additional
dependencies
are
limited
as
far
as
possible
because
we'll
just
make
the
distribution
of
telemetry
cpp
easier.
So
if
you
can't
go
without
it
is
a
dependency
there.
That
would
be
preferred.
G
I'd
like
to
follow
up
with
you
offline
like
if
you
can
educate
me
a
little
bit
more
on
a
BSL
implementation
in
detail,
I
can
share
the
prototype
work.
I've
done
to
compare
no
STD
versus
STL
versus
GSL,
which
is
Microsoft
guidelines,
support
library
which
implements,
for
example,
geocell,
span
class
and
I,
see
opportunity
like
how
we
can
set
up
the
generic
mechanism
for
us
working
things.
If
you
guys
have
a
strong
need
to
swap
certain
classes
for
your
own,
more
efficient
implementation,
I.
G
And
the
fifth
necessarily,
maybe
even
at
like
templating
and
now
we're
single
again,
these
breaks
certain
features
such
as
ABI
compart,
but
I
gauge,
like
number
of
scenarios
I
have
and
in
some
cases,
I'd
prefer
curve
over
compatibility
with
other
vendors
right,
and
that
I
believe
personally,
is
a
strong
motivation
for
some
of
our
flows
rely
on
more
efficient
library,
implementation
and
I.
I
want
to
keep
that
available
to
others
like
the
compile
time,
at
least
as
possible
bill
flag
or
something.
G
H
I
Riley,
this
is
a
Lolita
again,
we
are
actually
looking
at
building
out
the
exporter
for
Prometheus.
You
know,
which
is
an
issue
which
is
an
open
issue
right
now
on
the
repo
and
and
we've
already
started
working
on.
The
technical
design
so
again
would
like
to
go
through
the
same
process
as
we
did
with
the
metrics
API
SDK
design,
and
then,
of
course
build
it
out.