►
From YouTube: 2021-02-02 meeting
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
B
A
B
A
B
B
A
B
B
A
B
C,
plus,
plus
11
is
fully
supported
by
open,
telemetry
c
plus
hardware.
I
think
that's
that's
declared
or
are.
A
B
Don't
want
to
yeah.
I
think
this
is
mostly
about
compatibility
right
because
we
are,
we
are
going
to
mostly
like
act
as
an
sdk
or
library.
So
if,
if
some
other
library
you
can
build
only
people
who
require
c
plus
plus
ever,
I
think
we
can
do
that
so
mostly
for
compatibility,
so
we
don't
want
to
require
our
consumer
to
upgrade
it
to
the
platform
because
17,
I
think-
and
this
oh
yeah.
B
I
think
we
talked
about
this
before
and
there
are
a
few
reasons
like
some
libraries
we
rely
on
like
absolute,
which
also
supports
a
platform
eleven.
So
that's
one
of
the
reason
we
support
it,
but
this
may
change.
I
think,
in
the
future.
B
Yeah,
actually
I
I
haven't.
I
have
a
topic
today
actually
to
follow
for
the
sa
c,
plus
plus
standard
support
like
in,
I
think
I
saw
the
app
ceo,
safe,
plus,
plus
library,
upgraded
the
requirement
to
to
save
plus
plus
14.
I
think
so.
There's
one
pending
pr
in
in
open,
telemetry
plus
plus,
which
tries
to
upgrade
the
grpc
to
latest
version.
B
I
think
grpc
relies
on
relies
on
app
seo,
so
that
causes
problems
right,
because
we
still
support
c
plus
plus
11,
but
appso
requires
free,
supporting
break
there
right,
wow,
okay,
I
just
pinged
our
two
other
contributors
who
seems
to
both
have
conflicts
today,
so
they
may
not
join
in
today's
meeting
with
me.
A
A
A
Yes
and
obviously,
I've
been
now
working
on
some
common
stuff
for
propagators,
which
I
promised
on
the
one
of
the
when
you
when
you
approve
the
b3
propagator,
yes,
okay,
I
declared
I
will
be
working
on
this
common
stuff
for
so
that's
on.
I'm
working
on
this
right
now.
B
C
I'm
also
new
to
the
joining
meeting,
I'm
from
a
company
called
aqua
security,
and
we
we
contacted
connect
on
github
today
to
see
the
state,
the
state
of
the
project,
we're
considering
using
it
at
some
stage
in
production
and
we're
trying
to
try
to
understand
if
we
can
contribute
to
make
it
fast.
C
Telemetry
we
are
starting.
We
want
to
start
to
use
the
metrics
in
our
our
our
code
and
we
didn't
decide
on
a
library
yet
so
open.
C
C
Thing
to
do
so,
I'm.
C
Yes,
we're
here,
I'm
here
to
hear:
what's
things
that
are
going
on
to
understand,
if
you
can
help
in
and
use
it
at
some
stage,
some
stage.
B
Yeah
sure
sure
help
is
welcome
and
I
think
compared
comparing
to
other
languages
like
donate
or
go.
I
think
we,
our
progress,
is
still
slower
than
others,
but
I
think
I
don't
have
a
fixed
timeline
for
for
how
we
need
to
reach
like
the
1.0
stage.
So
we
may
just
discuss
that.
You
may
get
issues
yeah,
but
I
think
in
microsoft
we
we
have
two
and
or
two
or
three
people
fully
working
on
this
yeah
with
the
working.
D
C
You
know
thomas
you
and
splanco
or
microsoft,
places
that
are
trying
to
use
it
at
some
level
of
of
production
state
stage,
even
if
it's
an
alpha
or
or
you're,
just
like
purely
contributing
to
towards
the
stage.
Even
though
you
don't,
you
know,
you're,
not
instrumenting
libraries,
yet.
B
I
think
we
for
now
we
don't
recommend
to
to
integrate
into
into
production
like
some
some
functionalities,
as
you're
meeting
like
you
can
instrument
ad
instrument
course
to
your
c
plus
library,
that's
fine,
but
we,
I
think
we
don't
have
proper
gators
like
jrpc
and
he's
me
yeah,
probably
http,
I'm
not
sure.
I
think
it's
also
not
completed
right
so,
but
if
you
you
are
willing
to
contribute
and
try
the
integration,
I
think
the
support
is
yeah.
A
Well,
yeah,
I
well
well
at
splunk
what
the
the
part
which
I'm
well
involved
in,
and
I
can
reveal-
is
that
we
plan
to
release
this
quarter:
open
telemetry
module
for
apache
for
http
server.
So,
okay.
A
B
E
I
guess
hey,
I
just
heard
about
apache
and
cpa.
That
sounds
like
a
great
idea.
Would
this
be
something
that
we
should
add
to
the
country,
repo.
A
C
E
That's
great,
and
I
think
we
discussed
last
week
that
if
we
have
some
proposal
how
to
do
it,
we
can
actually
discuss
if
we
eventually
pull
it
in
the
main
report
and
claim
that
we
will
also
support
c
api
as
part
of
this
community
like
again
depending
on
how
it
goes.
A
B
B
A
A
B
E
Guys
we
can
either
take
it
offline
or
I
can
quickly
share,
maybe
a
link
in
the
chat
window
and
in
another
telemetry
related
sdk.
We
recently
open
source
that
the
main
idea
why
we
open
source
it
is,
we
decided
to
borrow
some
ideas
for
the
open,
telemetry
sdk
from
there.
E
E
It
only
needs
to
look
up
a
single
c
function
and
the
rest
is
all
kind
of
martial
through
that
ipc
style
interface,
where
you
specify
I
want
this
up
call
these
parameters-
and
this
is
the
cisco
id
and
then
it
maps
to
the
c
plus
plus
class,
like
in
this
case,
where
we
have
the
full
feature:
open,
telemetry,
plus
plus
sdk
api.
It's
like
switch
case
if
log
event
this.
If
log
trace
this,
if
stat
span
and
spam
this,
you
see
what
I'm
saying
so
it's
like
I'll
share
the
link.
E
B
Okay,
yeah,
so
please
feel
free
to
touch
build
back
to
our
contributor.
I
think
max
just
submitted
the
initial
layout
of
our
contributor
report.
I
think
we
will
we
will
build
the
other
the
build
scripts
to
make
it
work,
then
add
more
examples
and
then,
like
you're
gonna,
meet
your
apache
instrumentation
to
that
library.
B
Welcome
and
yeah-
I
have
one
question
I
asked
mentioned
that
before
you
join
me
max.
I
think
this
is
say
plus
plus
14
support.
I
saw
I
think
I
saw
the
app
so
which
requires
3,
plus
plus
14
mile
right
and
open
telemetry.
C
B
Plus
class
like
for
when
built
with
ot
otrp,
we
require
grpc
and
the
grpc
requires
appsource,
a
plus
plus
library,
then,
which
means
we
should
we
should
upgrade
our
language
standard
to
say,
plus
plus
14.
E
Okay,
tom,
maybe
just
fresh
in
my
mind,
because
I
had
the
meeting
about
it
about
an
hour
ago.
I
have
a
prospective
customer
and
it
seems
like
they
are
still
on
visual
studio
2015,
which
has
a
good
support
for
11.
and
so
so
so
support
for
c,
plus
plus
14..
E
Hopefully
we
can
code
it
around
so
that
we
use
just
the
features
that
work
there,
but
I
mean
right
now.
Last
time
I
tried
we
were
okay,
mostly
except
that
abseil
like
actually
our
there's.
Another
reason
why
you
mentioned
abseil
right,
yeah
m
park,
variant
that
we
use
for
for
no
std
variant
does
not
compile
in
visual
studio
15.
E
Then
again,
how
far
are
we
taking
dependency
on
upsell?
In
general?
You
mentioned
that
for
jrpc
it
will
be
required
right,
okay,
so
from
what
I
know
is
that
specifically
upsell
variant?
Just
that-
and
I
didn't
try
anything
else-
worked
fine
for
my
use
case
with
visual
studio,
15
and
with
c
plus
plus
11
support.
So
I'm
sorry
for
kind
of
holding
it
back,
but
I
think
there
was
another
reason
for
the
gcc
4.8,
the
min
we
support.
E
I
I
I
have
doubts.
We
need
to
check
how
far
it
is,
if
any,
if
it
supports
c
plus
plus
14,
but
it
was
needed
for
the
linux
for
the
envoy
proxy
and
for
a
few
other
scenarios
like
we
had
customers,
not
microsoft
customers,
but
other
company
customers
needing
c
plus
possible,
because
they
are
enterprise
servers
and
I
think
in
another
project.
We
discussed
that
red
hat
enterprise
linux,
for
example,
with
the
base
tools
that
it
supports.
E
It's
like
out
there
officially
supported
at
least
until
2022,
which
means
that
it's
kind
of
hard
to
make
that
decision
and
say
hey,
and
now
we
do
just
c
plus
plus
14
and
above
because
it's
it's
not
that
I
don't
want
it.
I
do
want
it
this
to
happen,
but
we
actually
have
customers
which
may
hold
us
back
on
11
for
for
a
bit
longer.
B
E
Like
I,
I
don't
know
how
we
are
going
to
solve
this
grpc
and
c
plus
plus
14
dependency
issue.
Is
it
just
going
to
be
like
hard
stop
and
we
say,
and
now
the
otlp
protocol
requires
grpc,
which
means
that
the
whole
open,
telemetry
solution
requires
min
c
plus
plus
14.
B
I
think
the
issue
I've
seen
is,
I
think,
there's
a
pending
pr
right,
which
tries
to
upgrade
the
grpc
to
the
latest
one
1.38
or
something
but
yeah-
that's
the
latest,
but
that
one
that
that
vr
is
the
build
is
broken
because
I
think
our
booking
on
gcc
4.8
yeah,
because
that's
that's
what
that's!
What
that.
E
Was
added
for
the
reason-
and
that
was
added
for
the
reason
like
like
I'm
advocating
now
for
not
for
my
customer
but
explaining
you
guys
why
the
historical
reason
that
I
think
was
either
lightstep
or
android
proxy
guys
that
needed
it.
That
is
why
it
was
added
as
part
of
ci
in
the
first
place
and
now,
if
we're-
and
that
was
done
like
a
year
ago
now,
if
we
are
to
break
it,
then
I
guess
we
need
to
have
a
wider
community
representation
and
the
bigger
discussion
about
it.
C
I
I
just
checked
in
the
middle
of
the
meeting
I
think
only
just
gcc
5.0
there's
a
full
support
for
cpp
14..
I
can
tell
you
that
in
our
company,
shame
on
us
4.6
is
still
their
requirement.
E
Before
for
windows
and
for
a
few
flaws,
I
know
how
to
kind
of
hack
it
you
see,
we
have
this
with
otlp
build
parameter,
and
only
if
we
build
with
or
tlp
we
require
a
grpc
upsell
and
all
these
dependencies
right
now
I
can
tell
if
I
run
on
windows
and
only
with
the
etw
sync,
and
only
with
the
atw
agent
listener.
I
can
build
without
otop
and
I
can
say:
oh
I'm
fine.
E
I
can
even
build
with
visual
studio
15,
but
that
way
you
would
never
get
grpc
and
otlp
protocol
supported
if
you
are
still
on
15..
So
it's
like
half
break
half
featured
open
telemetry
sdk,
but
that
will
cover
my
customer
base.
Now
I
don't
know
if
the
others
are
going
to
be
as
happy.
If
we
say
hey,
48
is
no
longer
supported,
that's
probably
going
to
be
frustrating
to
a
few
linux.
Folks.
B
E
It's
like
you
know
that
a
fairy
tale
when
you
get
either
the
berries
or
or
or
you
know
like
I'll,
tell
you
this
later.
It's
like
you
have
to
choose
one
right.
C
One
of
your
problems
of
not
upgrading
jrpcs,
I
saw
that
open
sensors
had
a
code
written
in
grpc
and
if,
for
some
reason
you
will
you
will
have
that
sort
of
of
a
contribution
of
you
and
jrpc,
you
will
have
to
upgrade
it
you.
You
won't
want
to
fork
it
from
an
old
version
and
put
it
inside.
So
it
looks
like
at
some
stage.
E
Let's
see
what
we
are
losing,
maybe
if
we
are
upgrading,
let's
describe
what
other
configurations
are
still
supported
and
like
we
were
thinking
about
having
one
binary
build
for
linux.
Specifically,
that
includes
all
of
the
standard
exporters,
including
what
elp
and
I
think
it's
a
requirement
to
have
hlp
tom.
Is
it
required
to
have
both
otlp
with
protobuf,
or
is
it
okay
to
have
just
a
tlp
with
json,
or
do
we
need
still
to
have
that
dependency?
Even
if
we
export
which
lp
json.
B
E
E
So
does
it
mean,
then
we
should
then
make
that
call
and
say:
hey
c,
plus,
plus
14
and
gcc.
Five
is
now
the
min
bar
because,
like
anyways
I'd
rather
see
other
maintainers
chime
in.
B
B
D
E
E
Yes,
so
provide
subset
features
for
bs,
2015
c,
plus
plus
11.
B
Okay,
I
think
that's
fine
yeah,
I
think
yeah.
This
looks
so
good
to
change.
I
think
I
believe
if
we
change
it
into
a
break
compatibility
like
some
library
may
not
be
supported
here.
We
must
be
very
careful
to
agree.
E
B
I
think
I'm
good
on
this
any
other
issues
we're
going
to
discuss.
I
think
thomas
has
issue
about
or
the
the
question
is
from
jakarta
about.
The
planning
of
open,
telemetry,
c,
plus
class
or
max
may
have
more
information.
A
E
Are
you
asking
me
or
who
sorry
I
didn't
get
foo
thomashi?
I
think
tom
is
asking
about
some
questions
that
I
missed,
because
I
joined
play.
C
Yes,
I
I
was,
I
was
asking
about
attack,
timelines
and
and
state
of
production
we're
trying
to
to
see
if
we
can
use,
even
if
even
a
pre,
a
a
pre
pre,
better
better
stage
of
this
product
in
in
our
software
and
understand
and
also
to
to
contribute.
So
we
so
it
will
get
a
stated
that
that
we
can
use
it.
We
started
we
st
we.
We
would
want
it
also
for
tracing
and
also
for
metrics.
We
started
to
check
on
metrics.
C
C
It
was
missing
just
one
line
of
code
to
make
it
working
and
and
we're
trying
to
understand
where
it
where
this
project
is
standing
at
and
if,
if,
if
a
contribution
from
our
company
can
bring
it
a
stage
that
we
can
use
it,
at
least
at
least
without
the
risk
of
of
harming
our
system,
even
even
if
we
don't
get
the
full
features
set
out
of
it.
E
So,
first
of
all,
we
are
trying
to
make
sure
that
traces
and
the
logs
are
stable
and
these
are
going
to
be
part
of
the
first
version,
one
release
and
at
the
same
time,
as
you
mentioned,
I
also
have
internal
customers
like
at
microsoft
that
I
would
love
to
onboard
to
what
we
have
right
now
like
try
whatever
right
now.
I
see
that
we
will
definitely
need
to
release
a
few
new
tags
like
right.
E
Now
we
have
that
git
tag
version
0,
zero,
one
or
something
we
need
to
have
periodic
like
every
two
weeks,
new
git
tags,
where
we
say
version
zero,
two
and
zero
three,
and
I
be
needing
a
new
tag
with
some
of
my
painting
prs
on
review
like
for
windows,
480w
export,
and
I
think
you
should
send
in
your
fixes
for
the
metrics.
E
But
let
me
elaborate
on
metrics
a
little.
There
was
a
meeting
with
prometheus
authors
I
think
last
week
and
there
are
going
to
be
changes
to
api
surfaces
related
to
metrics.
E
So
whatever
we
have
for
metrics
right
now,
it's
not
planned
to
be
part
of
version,
one
ga
it
is
subject
to
change,
and
I
think
you
have
an
opportunity
to
influence
how
it
should
change
so
making
it
work
is
a
good
step
and
the
core
part
like
the
engine
is
probably
gonna
remain,
but
there
will
be
differences,
changes
on
api
surface
of
it,
and
there
will
be,
I
don't
know,
like
version
1.1
or
something
that
will
be
done
later
in
the
pipeline
after
we
released
the
first
stage
with
traces
and
log
support.
E
Tom
does
it
align
with
your
understanding.
C
Yes,
so
it's
it's
like
the
the
main
risk.
C
E
Understanding
is,
they
are
actively
engaging
prometheus
author
and
we
do
rely
right
now
on
some
prometheus
supply,
plus
client
implementation
for
the
metrics
that
will
probably
stay
there.
I
don't
know
if
we're
going
to
rewrite
that
from
scratch,
I
mean
the
core
eventing
channel
part
the
current
implementation.
I
think
we
need
to
reshape
it,
but
not
rewrite
it.
E
I
hope
it's
not
like
going
to
be
a
full
rewrite,
reshape
it
to
fit
within
what
community,
like
larger
cross
language
community
decides
for
the
metrics
api
to
any
customer
early
customer.
E
I'm
I'm
thinking
like
you'd,
probably
have
some
logic
like
that
on
boards
to
open
elementary
api
searches,
like
I
don't
know,
a
single
function
that
emit
metric
or
log
metric
or
log
metric
with
parameters,
and
that
way,
as
long
as
you
provide
your
own
kind
of
shim
helper
shrimp
and
all
of
the
code
uses
your
shin,
it's
going
to
be
easier
for
you
to
patch
it.
When
open,
telemetry
api
changes,
you
just
fix
it
in
one
spot
right.
E
Yes,
that
will
probably
be
the
best
way
to
try
experimental
stuff.
E
It
won't
change
as
much.
Hopefully.
Yes,
yes,
that's
our
understanding.
So
there
is
one
issue
which
I
am
trying
to
kind
of
push
through
in
a
few
pr's.
I
want
byte
buffer
support.
E
I
need
byte
buffer
support,
so
I
need
to
emit
a
trace
with
a
8-bit
byte
buffer,
which
is
not
currently
supported
by
spec,
because
my
underlying
channel
supports
it
like,
and
if
I
provide
my
custom
open,
telemetry
exporter,
I
need
to
be
able
to
express
a
byte
buffer
on
api
surface,
so
that
part
is
missing
the
rest.
Personally,
in
our
scenarios
it
seems
like
it
satisfies
most
of
our
current
needs
and
hopefully
the
tracing
api
is
not
going
to
change
significantly
if,
if
at
all,.
E
64-Bit
integers
are
accepted
eight
bit
bytes
a
span
is
not
part
of
spec.
Okay
string
is
part
of
spec,
but
I
want
to
differentiate
utf-8
string
from
a
byte
buffer,
because
these
are
different
types
in
my
telemetric
flow.
E
So
I
actually
added
some
code
like
half
byte
support,
half
byte,
arrays
or
something,
but
we
also
discussed
with
google
forks
that
the
way
how
variant
works
from
the
abi
stability
perspective.
We
cannot
just
randomly
add
new
types
or
we
cannot
change
the
layout,
which
means
that
if
we
release
version
one,
we
need
to
make
sure
that
this
api
surface
covers
the
full
set
of
types
that
we
intend
to
support
in
version.
One
and
my
personal
opinion
is
that
bytes
by
buffers
I'm
missing
right
now
we
need
it
like.
A
A
E
Okay,
I'm
I'm
a
bit
lost,
so
it's
automatic,
like
we
do
have.
A
E
We
need
to
check
with
valley.
Lalit
is
in
a
different
time
zone,
so
he's
gonna
be
online.
Next
meeting
we
have
alternating
time
zones
like
alternating
meetings.
The
other
week
is
the
morning
time.
Yeah.
E
B
A
B
Okay
out
of
the
future
here,
any
any
other
questions.
E
Can
I
ask
you
guys
about
like
a
release
and
consumption
experience?
So
while
we
were
planning
to
have
a
vc
package
support
which
we
will
have,
I
also
have
customers
on
windows
who
that
used
to
consume
source
code
from
nougat
packages.
E
It's
funny.
It's
like
in
internal
build
environment.
E
E
For
example,
version
0.2
and
I
was
wondering
like
I
can
contribute
a
few
scripts
that
allowed
to
package
a
snapshot
of
the
source
code
in
the
nougat
package.
Maybe
two
packages
because
for
the
other
is
the
case,
there's
api
package,
which
allows
just
the
abstract
interfaces,
the
structure
and
the
sdk
package,
which
gives
the
implementation
of
the
api.
Maybe
two
nougat
packages
like
from
the
maintenance
perspective.
E
A
So
well,
I
can
share
that
for
the
apache
stuff
we
will.
We
are
using
basel
build
and
we
are
trying
to
get
it
together
from
basel.
So
I
I
cannot
tell
you
on
the
experience
of
using
any
other
build
system,
unfortunately,
but
this
is
what
we
are
trying
to
make
it
work
and-
and
it's
it's
it's
working
now,
I
would
say
for
for
as
a
module
for
basel.
E
Yes,
if
we
know
that
it's
like
right
now,
we
originally
committed
to
support
a
cmake
in
bazel,
and
I
think
google
guys
are
heavily
contributing
to
bazel
as
well.
I
think
splunk
and
google
right
I
use
bazel.
E
We
use
some
projects
at
microsoft,
uc
mac,
the
others
use
ms
build
and
we
don't
officially
support
ms
build
and
we
never
committed
to
support
ms
build
cmake
can
generate
ms
build,
but
teams
that
want
nuget
packages
that
don't
even
need
or
want
to
use
cmake
they'd
rather
keep
using
their
custom.
E
Ms
build
setup
that
does
nougat
package
install
and
saying
what
the
package
name
is,
and
then
it
extracts
from
that
pretty
much
zip
archive
the
source
code
of
the
sdk
for
them,
and
I
wanted
to
contribute
a
few
scripts
that
just
take
the
the
snapshot
of
the
source
tree
and
package
it
into
a
nugget
package.
That's
that's
it
under
the
tools,
repo
right
under
the
tools
directory,
not
gonna,
break
anybody.
E
Yes,
so
it's
like
for
that
flow
on
windows,
we're
trained
to
have
an
option
where
the
whole
thing
is
header
only
right,
at
least
for
one
of
the
flows.
So
from
that
perspective
I
don't
really
need
a
library
file.
I
can
just
package
these
source
headers
into
nuget
yeah.
E
E
To
to
write
any
build
rules,
you
just
do
include
open,
telemetry
whatever,
and
then
you
link
the
needed
system
like
windows,
sdk
libraries
and
there's
no
need
to
provide
any
lib
file
or
a
file
for
that
full.
That's
great,
less
artifacts.
To
maintain.
B
E
E
Much
nuget
source
code
nugget
we
created
like
in
our
own
build
flow
and
we
clone
it
for
our
internal
build
system
that
way.
Customers
don't
have
to
clone
from
open
from
like
from
github.
That
makes
them
less
vulnerable
because
we
are
only
a
snapshotting,
stable,
known
good
versions
and
the
not
directly
cloning
from
open
source.
B
E
I
can
contribute
a
readme
that
shows
this
is
how
you
can
create
a
nuget
package
and
maybe
a
generic
instructions
like
how
to
consume
nugget
packages,
but
I'd
assume
that
it's
public
knowledge
outside
of
our
scope
and
there's
no
no
ci.
I
don't
think
I
want
to
add
the
ci
to
verify
the
nougat
package
creation
script.
E
D
A
A
Question
so
one
more
one,
more
ques,
I
send
it
on
zoom
again
related
to
the.
A
That
this
is
a
header
only
thing
which
I
already
contributed
with
b3
propagator
and
expanded
a
little
bit
string
view
to
be
able
to
search
for
one
single
character,
which
is
an
equal
sign
in
my
case,
yes,
and
I've
added
some
test
case.
However,
I
was
struggling
to
use
string
view
well,
no
std
string
view
and
pause
or
no
or
position
with
which
is
equivalent
of
not
found,
basically
a
character
which
is
minus
one,
and
this
is
where
I
left
the
comment
with
a
6b.
A
Which
I
posted
to
zoom,
because
whenever
I
tried
to
use
this
string
view
and
pause,
I
had
the
linker
problem
related
to
like
this
symbol
is:
never
it's
not
defined
anywhere,
but
maybe
just
so
I
was
wondering.
Have
you
found
any
any
any
workarounds
or
any
best
practices
to
how
to
overcome?
Because
now
I
left
to
fix
me.
But
this
is
not
nice,
so
I
was
wondering,
have
you
do
you
have?
I
recommend.
E
It
like
usually
what
we
do
is,
let's
say,
there's
some
standard
library
implementation,
whatever
a
friendly
licensed
implementation,
stl
or
gcc
clang,
whatever
we
take
a
look
at
how
they
do
it
and
then
then
do
clean
room
porting,
and
it's
like
I'm
also
trying
to
make
sure
that
whatever
we
do
for
the
new
std
classes
is
also
compatible.
E
If
you
just
alias
pretty
much
to
the
standard
library,
because
in
a
few
flows
when
I
build,
for
example,
for
windows,
customers,
I'd
rather
use
standard
library,
implementation
of
everything,
including
string
view,
including,
like
a
span
on
latest
c
plus
plus
20,
so
that
I
don't
take
dependency
on
our
non-standard
backward
and
for
security
reasons.
Customers
may
trust
the
battle
tested
library
that
comes
with
their
compiler
of
choice,
rather
than
our
hacky
attempt
to
backport
latest
things
to
evolve
like
11
and
14..
E
So
anything
that
you
see
that
there's
some
conflict
between
how
standard
library
behaves
and
non-standard
library
behaves
feel
free
to
fix
it.
If
you
see
that
some
definitions,
missing
and
pause
is
missing
on
that
noise
td
string
view
added,
if
there
is
a
problem
like
this,
I
have
hit
it
in
a
few
tests
and
if
you
search
there's
some
check
if
they
have
a
standard
library
than
this
else,
if
non-standard
library,
then
this
expectation,
so
we
do
have
a
few
ugly
places
like
this.
E
Ideally,
we
need
to
try
to
minimize
this
by
bug,
porting
and
fixing
our
hacking
implementation
reasons
why
we
even
have
it
why?
Why
why
we
are
not
using
upsell
for
example,
or
why
we
are
not
using
standard
library?
I
think
we
were
planning
to
build
a
binary
like
for
linux
that
is
again
compatible
with
gcc
408,
which
we
mentioned
for
another
reason
and
gcc7,
so
that
you
can
actually
dynamically
load
and
never
use
any
of
the
standard
library
classes
and
service
surface
by
porting
our
own.
E
I
mean
the
whole
thing
like
if
we
absolute
gcc4
for
the
reasons
of
upgrading
grpc,
then
do
we
still
even
have
that
binary
incompatibility
at
the
steel
level?
Can
we
just
like
anyways
if
you
can
fix
it
fix
it.
B
E
See
to
me
when
you
add
something,
and
it
is
known
to
work
that
good
way
with
standard
library,
for
my
header
only
build
of
sdk,
I
would
even
build
with
the
standard
library
and
I'm
not
gonna,
be
affected
by
this
weird
operation
of
non-standard.
E
So
I
think
we
need
to
catch
those
spots
and
can
contribute
back
to
community,
because
the
best
where
it's
going
to
be
used
is
on
linux,
most
likely
that's
where
originally
this
non-standard
non-std
string
view
requirement
came
from.
A
A
E
The
theory
is
that
standard
library,
implementation
of
string,
view
or
string
even
on
recent
compilers
may
not
necessarily
be
compatible
memory
layout
wise
between
different
compiler
versions,
and
that's
why
we
came
up
with
this
interesting
concept
of
re-implementing
all
foundation
classes
ourselves.
E
A
Backboard
sorry,
but
was
this
case
really
found
somewhere
in
the
wild,
that
this
is
a
problem
like
to
do
that.
This
is
a
pro
that,
because
I
have
actually
this,
this
is
the
first
library
in
my
I
don't
have
much
experience,
but
this
is
like
the
first
time
in
my
life
when
I
see
that
this
kind
of
a
problem
is
solved
this
way,
but
maybe
there
are
other
libraries
which
I'm
not
aware
of,
but
I
have
not
seen
anything
like
this:
isn't.
E
Upsell
sort
of
solves
some
of
this-
I
know
upsells
sort
of
sort
solves
backward
aspect
of
it.
E
Abseil
gives
similar
concept
of
this
yeah,
but
in
addition
to
this,
we
were
trying
to
make
sure
that
the
memory
layout
of
recompiled,
open,
telemetry
library,
instrumentation
library,
is
compatible
with
any
host
manix
z,
whether
it
is
built
with
gcc
four
gcc,
five,
six
seven
doesn't
matter,
you
can
still
load
the
shared
library.
E
I
can
tell
you
that
this
is
very
artificial
requirement
from
windows
standpoint,
because,
starting
from
visual
studio
2015,
there
is
actually
a
api
compat
guarantee
for
15
17
and
19.
so
for
windows.
For
me,
I
don't
even
need
that,
like
I'm
looking
at
this,
but
originally
we
made
that
decision
that
we
respect
that
that
we
build
that
mini
stl
inside
our
library
that
we
keep
maintaining
it
and
on
api
surface
we
only
expose
types
using
that
mini
seo,
known
std
stl.
E
We
can
bring
that
topic
for
discussion
again,
like
maybe
when
there's
a
bigger
community
representation
on
this
call
to
ask
why,
like
sometimes,
I
also
have
the
same
question:
okay,
thanks.
C
Max
mikey
said
you
said
it
that
on
windows,
you
have
a
guarantee
about
abi
compatibility.
C
C
E
E
So
then
I
can
say
that,
at
least
for
now
for
the
three
current
compilers
that
we
have
today,
15
17
and
19,
I'm
mostly
gonna,
be
fine.
Even
if
I
build
with
the
standard
library
see
it's
like,
I
care
a
bit
less
about
it,
but
I
think
on
linux
the
guys
said
that
if
you
do
these
tricks
with
std
string,
for
example,
if
you
pack
instead
this
train
and
try
to
compile
across
gcc4
and
main
exe
process,
that
main
binary
that
is
built
with
like
gcc,
eight
or
nine.
C
It's
much
worse
than
that,
even
even
compiler
flash
simple
compiler
flights
can
can
break
the
api.
From
my
experience.
E
E
That's
like,
I
think,
some
of
the
folks
who
originally
were
on
this
project
and
who
are
strongly
advocated
for
that
they
are
no
longer
joining
community
meeting
anymore.
I'm
just
recapturing
recapping
you,
the
story
of
where
it
originally
came
from.
E
Variant
was
borrowed
from
an
park
variant
and
I
don't
like
it
because
it
doesn't
compile
with
visual
studio
2015
string
view.
I
I
did
the
clean
room,
our
implementation
of
a
few
methods
peaking
into
three
different
implementations.
I'm
not
going
to
hide
that.
I
looked
at
how
it's
done
in
clang
in
visual
studio,
surplus
plus
stl,
so
I
didn't
just
steal.
E
I
looked,
I
adjusted
and
I
hand
written
the
logic
that
works
in
the
same
way
as
it's
required
by
the
c
plus
plus
standard.
What
else
string
view
span
for
the
span?
I
don't
remember
where
it
came
from
us,
but
I
also
give
an
option
to
substitute
with
microsoft.
Gsl
spam,
because
gsl
spam
implements
precisely
the
the
standard
like
it
was
an
early
reference
implementation.
So
I
allowed
to
build
with
that.
E
B
B
E
Though
it
is
a
not
a
patch,
it's
like
a
mit
license,
but
you
can
actually
replace
our
span
with
gsl
span
and
all
tests
pass
or
you
can
build
with
the
c
plus
plus
20
span
std
span
and
all
tests
work.
E
Sorry
guys,
I
have
a
hard
stop
at
four.
I
need
to
join.