►
From YouTube: 2019-10-14 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).
B
A
D
B
B
Okay
may
I
start
so
I
I'm
trying
to
kick
off
this
I've
heard
so.
First,
we
need
to
get
people
and
I
I'm,
proposing
that
we
have
a
max
and
me
joining
as
a
maintainer
and
implore
for
this
and
I
plan
to
send
the
initial
PR
I
think
I
can
send
to
PRS
in
this
month.
The
first
one
is:
we
have
this
initial,
like
folder
hierarchy
for
the
project,
Abdullah
team,
Python
and
I
understood
any
of
how
open
today
IDK
should
be
organized.
B
I
hope
that
we
can
get
people
help
to
review
I'm,
not
very
familiar
with
the
modern
build
systems
haven't
being
a
C++
engineer
for
the
past
few
years,
I
started
as
a
compiler
developer,
but
really
I
move
forward
to
different
languages,
so
I'm
willing
to
learn
I
might
be
like
all
the
Styles
C
developers.
So
I
look
forward
for
your
feedback.
I
mean
wow
I
want
to
know
like
who
have
a
additional
time
and
can
help
out
the
PR
reveals
I
want
to
see.
We
got
enough
population
or
I
need
to
look
for
more
people.
B
E
E
B
Great
then
makes
me
much
happier
it's
Connor
time
boring,
because
I
came
from
the
Python
community
and
I
host
a
meeting
there
together
with
Chris,
and
we
used
to
see
like
10
or
like
20
people
there
and
now,
like
oh,
like
we
got,
started
and
see
less
faces
and
me
makes
me
a
little
bit
worried
being
it's
got
to
happen.
No
I
at
least
like
to
folks
trying
to
make
this
was
the
amount
of
power
we
can
get.
D
Riley
I
noticed
that
the
you
sent
a
PR
number
three
initial
lead
me
and
contributing
documents,
and
the
reviewers
listed
on
that
PR
are
three
people
and
I'm
confused
about
that
already.
If
I'm
listed
as
a
as
an
approver
I
should
see
this
I
didn't
know.
This
was
a
PR
until
just
now,
for
example,
is
there
so
I
really
understand
how
github
gets
its
list
of
maintainer
and
approvers
I.
B
B
B
So
my
general
understanding
is
if
there
is
a
cross
language
design
or
something
big,
we
should
go
to
the
spec
and
drive
for
the
cross
language
thing
and
instead
of
trying
to
spend
energy
here,
what
we
do
here
is
we
read
the
spec
and
follow
up
and
implement
that
unites
secret
and
if
we
notice
some
specific
challenging
in
C++
freedom
hope
you
look
for
like
lot.
Three
algorithm,
they
look
for
super
performance,
then
we
realized.
Oh,
this
is
something
want
to
bring
back
to
a
spec.
B
D
Yeah
on
that
note,
I
just
updated
an
attempt.
That's
been
standing
for
a
couple
of
months
on
on
global
initialization
I.
Think
anyone
here
should
consider
reading
that
I'll
post
a
link
in
the
chat,
but
if
you're
in
the
road
test
repo
it
so
except
number
five
and
there's
a
work-in-progress
PR
just
a
moment,
I'll
find
a
link
yeah.
C
Yeah
I
read
that
we
reason
why
I
mentioned
the
global
singleton
is,
of
course
it's
usually
super
easy
to
invert.
That
and
people
do
that,
and
then
we
encountered
quite
interesting
combinations
of
libraries
end
up
in
the
same
process
and
with
certain
initialization
order
guarantees.
It
presents
a
challenge
to
rework
the
code
of
every
library.
So
it's
like
anyways.
Let's
discuss
this
outside
of
the
hospice.
A
B
E
I
agree
but
I'm
just
for
the
C
support,
like
the
people
that
have
asked
for
it
for
open
tracing.
A
few
examples
are
like
the
Gloucester
file
system.
They're
all
based
in
seeing
post
fresco
live
s
for
something
like
that.
It's
come
up
a
few
times.
I,
don't
know!
If
would
plan
on
developing
it
concurrently
or
not
or
like
finish
the
C++
API
and
then
serve
Ariana
and
would.
E
B
A
E
It
varies
so
like
the
person
I
was
working
on
Postgres
Cole,
who
was
able
to
work
with,
is
interested
in
tracing
with
Jaeger.
He
was
able
to
put
something
together
as
like
a
module
that
loads
into
post
grad
school
at
runtime
and
uses
the
underlying
C++
tracer.
There
are
some.
There
was
some
interest
and
like
a
complete
C
stack
and
someone
from
Jaeger
actually
started
developing
a
CT
tracer
and
the
C
API.
It's
just
a
lot
of
work
and
I.
Don't
know
if
anyone.
C
Questioned
here
so,
if
say,
we'll
have
respectful
capi
and
we
have
two
implementations.
One
is
binding:
the
C++
stacker
and
the
other
one
for
those
who
actually
need
like
embedded
C
or
something
they
have
a
complete
implementation
and
B
or
C.
It's
just
kind
of
maintaining
one
had
that
as
a
single
source
of
truth,
how
the
the
I
should
look
like
then.
C
For
our
use,
cases
for
Microsoft
internal
is
the
key.
We
do
have
C
consumers,
but
to
them
it
wasn't
an
issue
to
link
the
C++
code
within
it's
just
a
one
at
AC
API,
but
they
could
still
either
dynamically
load.
The
ship
was
blast
library
or
they
could
still
figure
out
how
to
link
the
C++
library
into
their
process
statically.
So
there
was
no
need
for
a
pure
C
implementation
said
the.
C
C
A
A
That
also
means,
like
no
exceptions
as
part
of
the
API
right
yeah.
C
That's
what
we
follow
as
well
like
there's
a
good
partner,
Jason
HPV
library,
where
they
have
that
macro,
where
you
can
have
your
own
thrall
and
you're,
always
like,
depending
on
the
fear
building,
if
no
accept
or
if
you
are
building
with
exceptions,
you
can
customize
that
behavior
in
a
way
like,
but
ideally
no
exceptions
on
it.
You.
C
Of
course,
do
in
C++
I
think
we
need
to
discuss.
I
need
to
learn
a
little
bit
more
because
I
heard
about
the
headers
for
HDL
like
if
we
can
rely
on
something
that
would
be
actually
like
this
place,
like
header,
only
dependency
on
something
that
can
be
expressed.
Header
only
I
think
that
is
appropriate.
C
Generally
with
the
steel,
we
had
some
competition
it
just
maybe
it's
unique
to
Microsoft
and
the
unique
to
the
way
how
we
build
our
ratios.
Our
different
organizations
prefer
building
STL's
differently,
rendering
at
binary
incompatible,
ABI
incompatible,
one
with
another.
That's
where
we
ended
up
having
it'll
build
from
source
code
for.
A
C
B
Now
a
lot
of
cases
with
symbol
directly
to
the
ingestion,
so
is
we
have
in
the
backend
and
I
think
having
that
HTTP
client
library
that
works
across
platform
and
X
ability
to
replace.
That
would
be
something
we
look
forward
to
to
contribute
and
work
with
your
guys.
It's
a
really
cool
piece
and
it's
very
helpful,
and
we
don't
want
to
just
have
one
developer
like
max
Matheny,
insightful
of
all
libraries
and
platforms
under
wine.
C
So
just
some
feedback
from
over,
and
so
our
current
telemetry
SDK
runs
on
Windows
Linux
Mac
everywhere.
So
we
had
a
roster
of
clients,
supported
lip
curl
on
Linux,
with
open,
SSL
and
optionally,
no
tos
and
dirty
HTTP
client,
as
well
as
wininet,
and
the
CPP
restes
declare
that
Microsoft
has,
but
we
also
ended
up
having
to
bind
to
Apple's
HTTP
client
that
comes
in
their
framework.
C
So
if
we
start
with
the
same
structure
that
identifies
like
the
core
interface
for
an
HTTP
client,
irrespective
of
where
it
is
bound
to
what
can
create
implementation,
it
is
bound
to
and
then
basically
describe
the
contract.
And
then
we
can
like
throw
in
like
three
for
reference,
implementations
and
suggest
that
if
you
guys
need
another
custom,
HTTP,
client
library,
then
you'd
have
to
code
it
to
that
interface.
So.
B
B
Hope
you
take
this
sahasrara
SDK
and
they
run
on
our
environment.
Where
initially
we
can
see,
we
use
lead
core
and
people
say
no.
We
have
this
native
HTTP
client,
which
is
recommended
by
apple.
So
you
guys
should
just
remove
that
customer
implementation
he's
the
standard
library
and
also
you
can
imagine
when
we
run
this
on
Windows
with
no
spoke
about
how
us
no
I,
don't
want
to
you.
B
Sleep,
chrome,
I,
want
you
guys
to
use
the
Windows
HTTP
client,
because
that
one
is
using
the
windows
like
global
proxy
said
he
and
since
that,
even
worse,
if
you
run
run
the
SDK
in
Android,
meaning
some
like
big
head,
then
people
are
saying
they
want
your
eyes
decline.
Sleep
has
passed
callback
to
the
reverse
tree
and
I
came
to
the
Android
Java
HDH
e-library,
so
that
you
want
to
use
the
specific
proxy
they
want
to
use
some
special
activity,
the
validation
this
will
be
respected
by
the
operating
system.
A
B
C
Has
its
own
HTTP
client,
so
we'd
rather
bind
to
that
right.
So
as
long
as
we
have
a
same
definition
of
common
interface
for
I,
do
have
a
bit
of
a
concern,
though,
like
I've
been
reading
about
web
sockets,
which
is
a
great
idea,
I
mean
for
the
streaming
proposal
like
man,
it's
interesting,
it's
becoming
interesting.
How
do
we
abstract
away
from
the
just
HTTP
posts
into
something
universal
enough?
So
internally
we
don't
do
things
like
as
sophisticated
as
WebSockets
yet
turn
away.
B
B
E
C
Understanding,
let's
say
if
you
do,
post,
okay,
post,
okay,
post,
okay,
right
and
you
would
do
this-
is
a
handshake
once
so,
the
overhead
of
whatever
headers,
that
you
post
isn't
really
that
big.
So
it's
like
to
the
client
code,
it
seems
like
back
to
back
HTTP
posts.
Let's
say
in
some
operating
systems,
you
shoot
my
then
just
two
simultaneous
connections
to
individual
hosts
and
basically
on
to
connections
you'd,
be
firing
back-to-back
and
for
anything
that
hasn't
gotten
200.
D
I
see
I
feel
like
this
we're
talking
about
how
to
implement
HTTP
transport,
and
it
just
feels
out
of
place
for
a
baton,
matriz,
meaning
I,
like
I
know,
when
the
other
languages,
we
provide
this
propagator
API
and
we
have
injecting
we
have
extract
and
you
can
extract
from
one
context
and
then
there
needs
to
be
an
example
given
for
every
of
one
of
these
libraries
that
it's
out
there
up
like
how
do
I
extract
and
put
it
in
this
specific
libraries,
HTTP
libraries,
headers
and
then
you're
done.
You
don't
have
to
do
anything
more.
C
Mostly,
yes,
that's
why,
when
I
was
speaking
about
a
sane
abstraction
interface
for
the
HTTP
client
like
we
can
actually
prescribe
a
recipe
for
how
that
HTTP
client
bubbles
up
the
response
code
for
a
post.
So
then
you
can
react
and
have
some
custom
logic
on
your
exporter.
If
I
haven't
gotten
200.
Okay,
what
do
I
do?
Do
I
need
to
mark
that
much
for
subsequent
that
he
said
or
bind
some
more
complex,
exponential,
back-off
retry
logic
like
if
I
tried,
10
times
and
I
fail
like?
C
A
D
Over
climate
change
sounds
like
like
I
mean,
for
example,
I
know,
Ryan
has
worked
on
a
transport
using
live
event.
That
is
a
way
we
send
our
data
to
the
light
step
system
and
I
mean
I.
Guess
that's
a
reusable
component
but
I
originally,
when
we
start
this
conversation
thought
they're
talking
about
how
users
will
instrument
their
code
and
I
realize
that
that's
not
what
we're
talking
about
yeah.
D
Is
this
proposed
protocol
called
OT,
LP
I?
Think
from
Tigran
and
I
I?
Don't
have
a
strong
opinion,
but
it's
been
proposed
and
it's
basically
it's
got
a
lot
of
reliability
built
into
it
and
we
looked
at
that's
Ryan
and
me
and
sort
of
weren't
sure
that
it
was
possible
to
implement
using
a
very
low
dependency
step
such
as
live
event,
and
it
worries
us
that,
like
in
order
to
implement
Oh
TLP,
you
got
to
have
G
RPC,
for
example.
D
Right
now,
which
has
been
a
tremendously
difficult
thing
to
keep
like
dependency,
is
sorted
for
and
it
just
raises
the
size
that
dependencies
which
I
know
C++
programmers,
often
are
trying
to
eliminate
dependencies.
So
we
could
have
more
than
one
option
here:
I'm,
not
sure
what
if
you
haven't
Oh
Sophie,
you
might
take
a
look
because
it
does
sort
of
depend
on
GRDC.
C
C
D
I'll
try
and
find
a
link
to
this.
This
protocol
proposal,
but
Ryan
had
been
trying
to
Ryan
and,
unlike
Steph,
had
been
basically
lobbying
for
an
alternate,
basically
something
that's
much
simpler
and
we're
not
sure
how
we're
not
sure
where
that's
going
I
think
at
the
moment,
everything's
at
Ryan
not.
C
And
I
believe
we
can
share
some
thoughts,
not
necessarily
like
you
know
the
final
implementation,
but
some
thoughts
that
do
not
depend
on
concrete
protocol
like
what
exact
binary
serialization
protocol
we
use.
So
you
know
client
is
a
separate
abstraction
layer
and
then
there
is
a
concrete
protocol
that
you
would
post
using
that
quiet,
Thanks.