►
From YouTube: 2020-06-08 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
A
A
Can
first
with
a
quick
update
here
to
give
people
example
like
how
how
I'm
thinking
we
can
we
can
run
this
so
for
the
contacts
API.
Actually
we
have.
This,
like
small
group,
engage
her,
so
we
have
like
for
folks
here.
Kirk
has
the
the
Google
like
host
and
also
have
like
two
other,
like
interns
here
and
I've
seen.
The
first
appear
so
seems
like
we're:
making
good
progress.
A
A
C
A
C
A
A
C
D
What
hi?
My
name
is:
Oliver
I
work
with
men,
Odin
and
Nick
on
this
simple
API.
So
right
now
we
we
have
like
a
starter
project.
We
can
work
at
which
we
implement
open
census,
API
like
tracing
API,
but
right
now
we
have
this
started
implementing
the
sampling
API
like
open,
telemetry
sampling
API,
since
it
seems
like
there's.
No
implementation
of
this
sub
cific
feature
yeah.
So
basically
we
tried
open
census,
one
yeah,
but
not
really
working
on
the
open,
telemetry
API
I
see
yep.
C
E
Again,
what
we
have
been
doing
is
starting
to
work
on
a
technical
design
that
we
will
submit,
for
you
know
the
review
by
by
the
open
telemetry
team,
and
we
have
also
done
a
initial
analysis
of
the
implementations
in
Python
and
go
as
well
as
Java
and
again.
You
know
a
lot
of
the
insight
that
you
had
shared
in
terms
of
Java
being
slightly
differently
implemented
compared
to
Python.
Well,
we
can
validate.
That
is
true.
E
What
what
we
plan
to
do
is.
Actually
we
have
a
detailed
plan
that
we're
building
in
terms
of
the
metrics
and
the
the
different
parts
of
the
API
and
the
instruments
that
are
in
you
know
we
plan
to
build
first,
and
so
we
will
submit
the
design
first,
get
your
you
know,
get
reviews
and
then
start
hopefully
start
implementing
by
next
week.
Okay,.
E
A
E
A
B
So
we
have
initial
design
that
defines
the
scope
of
the
project
and
we
shared
it
with
max
max
keep
some
useful
feedback,
and
we
had
some
discussions
for
the
last
week
tomorrow.
We'll
have
another
sink
about
three
pages
together
with
Max
and
Sergei.
Sergei
is
leaving
the
three
pages
effort
across
all
open
Kalama
tree.
B
F
A
G
So
I'm
Elia
from
Google
and
I'm
host
vigil
as
co-host
and
my
dad,
jibun
and
CEO
Chopra
are
in
terms
last
time.
We
are
not
sure
what
it
exactly
which
exported
exactly
to
work
once
all
right
now
we
we
started
to
urban
New
York
on
TCP
exporter
and
again
potentially
oil
to
be
exported
and
haven't
yet
determined
to
finish.
G
A
G
A
A
Okay
for
the
I
think
this
one
Jonas
I
approve,
so
if
you're
you're
ready,
we
can
merge.
This
I
think
max
has
a
question
not
sure
if
we
put
in
the
PR
comment.
So
the
question
is:
when
we
understand
we
expect
just
a
timestamp
or
we
need
a
class
like
and
spend
options,
but
I
from
my
perspective,
I
think
it's
fine
with
your
current
approach.
F
H
F
Just
gonna
be
easier
than
also
if
we
ever
come
up
with
a
projection
to
other
languages,
it's
easier
when
we
keep
it
simple
request
course
is
easier
to
write
a
maintaining
projection
and
I.
Think
the
overhead
shouldn't
be
that
much
if
there's
a
limited
set
of
fields
that
we
expose
and
perhaps
certain
can
be
fully
reusable
for
that's
one
option:
sundance
plan
options,
just
maybe
we
should
make
it
an
option.
I.
H
Mean
I
can
tell
you
by
event
with
the
two
separate
classes.
I
starts
with
options
has
several
things
in
it:
a
selector,
steady,
time-stamped,
the
absolute
x
time
is
like
a
span,
kind,
might
have
other
things
like
parenting,
information
and
finish
time
span
right.
The
finish
time
spends
options,
just
as
steady
and
timestamp
so
try
to
keep
like
those
two
sets
exclusive.
H
If
unified
it,
then
you
know
we
might
come
into
this
situation
where,
for
example,
for
this
span
kind
did
we
need
to
specify
there.
There
are
different
values
in
the
start,
it
has
to
start
end
to
end,
and
then
it's
kind
of
a
question
what
in
that
case,
and
they
try
to
make
it
clearer,
the
basically
opposite
so
exclusive,
and
there
are
none
of
those
so
do.
A
H
F
Comment
is,
since
we
have
this
end
span
options
as
optional
parameter
right
and
when
we
start
this
Bank
won't
it
be
reasonable
that
we
already
supplied
like
a
complete
set
of
options
and
start
and
is
optional.
We
don't
play
any
where,
as
if
it
were
the
same
colors
genetic
functions,
we
could
have
modified
the
ones
that
we
passed
initially
with
supplying
some
extra.
A
A
F
A
I
A
A
Okay,
cool
thanks,
yeah
I
think
we're
all
cover,
and
this
PR
is
a
fairly
straightforward.
Well,
so
this
is
the
the
CI
part
used
to
buy
the
spec,
and
some
other
like
open
time
truly
opposed
to
do
symbolic,
spell
check
people
made
type
or
something
CR
should
capture
that
as
long
as
it's
in
the
markdown
files,
we
can
extend
that
to
other
horsepower
later.
A
A
A
A
H
A
F
So
as
a
compile-time
motion,
so
my
intent
is
to
put
support
both
methods
for
those
who
do
to
whatever
performance
reasons
or
other
static.
Linking
reasons
prefer
to
use
standard
library,
they
can
use
the
standard
library
as
an
implementation
of
anything
we
have
under
the
university
and
those
who
need
dynamic
library
and
covered
compatibility
with
pre-built
exporters.
Those
who
would
use
the
the
university
implementation
for
we
discussed
this
with
Ryan
brief.
We
span
like
the
standards
but
not
open
to
a
mattress
pad
span
class
is
not
right
now,
part
of
official
standard.
F
Yet,
but
it's
implemented
in
in
a
few
different
reference
implementations.
So
I
used
Microsoft
guidelines,
library
for
that,
as
an
example,
it's
a
fairly
mature
library
which
has
quite
prominent-
you
know,
use
user
Bay,
but
you
know
four
thousand
stars
on
github
I'm
going
to
avoid
you
instead
mod
your
journals
just
for
them.
My
for
work,
I
preferred
using
sub
modules
just
easier
for
me
to
do.
F
F
So
I'm
not
offering
this
as
the
only
it's
one
of
the
options
so
that
we
can
use
for
the
span
implementation
so,
rather
than
providing
our
own
open
telemetry
span
for
statically
linked
projects,
it
might
lead
this
and
so
far
alike.
I
tested
it
works.
Fine
for
me,
so
I
was
gonna,
add
a
see,
Mike
build
option
and
that
familiar
with
basil
I
need
to
look
into
how
to
do
that
to
decide
whether
the
libraries
built
with
STL
or
if
it's
built
with
the
open,
telemetry
nor
city
and
another
thing.
This
is
more
of.
F
We
need
to
discuss
this
about
how
respect
the
attribute
value
class
for
the
very
end.
I
have
questions
here.
Is
this
something
that
open
telemetry
describes
prescribes,
or
is
this
our
own
invention,
and
it
is
a
implementation
dependent
because
right
now,
I
see
that
it
might
be
missing.
Certain
constructs
such
as,
for
example,
in
terms
of
expressing
types
if
we
use
CCC
in
types
such
as
aimed
in
6040,
we
might
as
well
use
you
into
the
32
underscore
t
as
well.
F
Instead
of
one
signed
in
I,
don't
remember
particular
detail
why
I
had
to
do
it
elaborate
I'll
check
and
elaborate,
but
I
think
it's
gonna
be
a
bit
more
consistent
if
we
stick
to
one
typing
system,
the
other
part
is
how
do
we
express
Boyd
box
because
it
seems
like
for
this
type.
We
did
not
really
like
maybe
I'm
missing
something
but
I
actually
needed
to
have
spent
for
the
byte
buffer
and
the
costs
are
heavily
despised
and
I.
F
Guess
something
that
we
should
not
be
using
I'm
thinking
about
older
code,
where
folks
have
not
necessarily
been
using
string
view
string
view
requires
the
wrapper
class
and
some
code
might
have
been
already
using
Mike.
You
know
cause
char
star
and
typecasting
from
the
byte,
arrays
least
string
materials
with
no
terminator.
F
So
I
was
thinking
that,
instead
of
boxing
it
into
a
string
you
every
time
should
we
have
possibly
an
option
in
that
variant
which
suggests
that
this
is
in
fact
not
as
a
span,
not
a
string
view,
but
a
raw
C
style
string.
So
that's
open,
like
I,
like
to
discuss
this.
If
we
need
to
divide
it
into
the
implementation,
but
I
do
have
a
strong
opinion
that
we
probably
should
add
a
you
ain't
ATT.
F
So
that's
so
far
like
sharing
those
thoughts,
it
was
quite
a
bit
helpful
to
have
a
map
with
an
enum
which
describes
the
index
of
a
variant,
because
when
we
refer
to
element
we
do
STD
get
whatever
alternative
and
we
specify
the
index
of
that.
So
to
my
personal
preference,
it's
nicer
when
we
alias
this
in
the
same
order,
and
then
the
code
looks
a
bit
cleaner
because,
instead
of
get
seven,
we
would
actually
explain
what
exactly
we're
getting
like
what
type
it
that's
related
to
this.
F
You
know
no
custody
PR.
The
rest
I
collapsed,
whatever
unrelated
changes
and
I'm
going
to
split
it
into
a
smaller
PR.
So
the
rest
looks
pretty
much.
You
know
there's
one
more
thing
both
when
I
was
solving
a
specific
issue
like
I
also
implemented
an
exporter
for
Jason
for
console
for
etw,
for
like
pretty
much
statically
linked
implementation
over
as
decay
without
exporter,
but
like
as
a
reference
point
I
had
the
use
case
where
I
needed
a
null
key
value.
Tempted
iterable,
like
I,
was
looking
at
your
feedback.
F
If
it's
something
that
we
might
want
to
consider
having
as
part
of
the
standard
others
such
as
an
empty
collection
like
null
object
collection,
which
has
zero
values,
but
it's
still
iterative.
So
I
had
this
matter
where
I
needed
and
it's
pretty
much
a
fairly
simple
implementation.
I
was
thinking.
Maybe
we
should
that
it
side
by
side
to
where
we
have
the
other
key
value,
terrible
implementation,
that's
about
it,
the
rest,
I'm
close!
Yes,
the
other
thing
is
close
method.
F
This
is
my
feedback
in
a
similar
system.
We
never
had
to
have
a
precise
specifier
with
close
with
microseconds
and
explicitly
requesting
the
timeout
parameter,
because
in
a
way
we
had
this
indirectly
configurable
through
options.
So
I
was
wondering
if
we
should
have
something
like
a
tracer
options.
Instead,
then,
if
we
have
trace
eruptions,
we
may
invoke
close
method
and
obtain
the
actual
timeout
value
from
those
trace
routes.
F
F
F
H
H
F
Right:
here's
my
here's.
My
scenario
and
I'd
like
to
elaborate
on
this
before
we
make
a
change.
First
of
all,
I'd
like
to
commit
to
support
the
noise,
the
implementation
in
all
of
our
tests
and
it
cannot
be
broke.
Like
you
know,
rule
number
one
never
break
the
you
know
yesterday
extended
line
self-sufficient
implementation.
That's
my
fundamental
assumption.
F
Rule
number
two:
let's
say:
office
application
like
office
desktop
Windows,
Microsoft
onedrive.
These
guys
may
not
even
want
to
add
any
shared
library
at
all.
They
prefer
to
avoid
loading
anything
dynamically
due
to
security
considerations,
they'd
like
to
build
it
all
statically
and
just
be
done
with
it.
F
They
still
want
to
benefit
from
the
standard
API
surface
and
the
standard
offering
that
we
provide
to
them
and
when
we
build
with
a
specific
STL
and
specific
library,
say
Microsoft
guy
and
guidelines
support
library,
we'd
like
to
avoid
all
these
transforms
and
just
use
it,
as
is
as
if
the
customer
code
would
have
implemented
it
and
if
they
use
STD
a
string
view.
These
guys
are
not
gonna
instrument
all
of
their
code
with
no
CD
single.
They
would
prefer
to
use
it
as
such.
F
So
that's
why
I'd
like
to
have
that
aliasing
option
available,
because
we
are
in
essence,
still
cloning
that
code
and
we
are
in
our
newest-
did
the
implementation
trying
to
keep
it
as
close
to
standard
as
possible.
So
that's
why
I'd
like
to
offer
the
adoption
like?
If
we
are
cloning
it,
why
not
offer
an
option
we're
just
unifying
the
two
into
one
so
that
there's
no
transform
necessary
and
another
argument
is
somebody
might
say:
hey,
I
trust,
my
steel
I
trust
Wes,
whatever
you
guys
came
up
with.
F
What's
in
it
like
trust
issues,
you
know
and
then
I
can
tell
hey.
You
rely
on
your
STL
and
if
there's
a
bug
in
STL,
you're
gonna
go
and
chase
the
STL
developers
for
that
visual
C++.
Str2
fix
it
rather
than
chasing
us
to
fix
a
security
bug
in
our
newest
TV
implementation,
but
we're
gonna
do
risk
ourselves
as
well,
because
we
are
no
longer
gonna,
be
saying.
F
F
Let's
say
let's
say:
I
would
like
to
prefer
the
standard
open,
telemetry
and
custom
exporters
to
some
other
systems
right
and
then
for
a
jewelry
say
he
you
guys
go
with
the
noise
to
the
implementation,
because
that
is
what's
giving
you
all.
These
fancy
exporters
that
you
need
so
I'm
just
kind
of
giving
people
more
freedom.
I
realize
that
yeah.
F
A
I
can
see
some
like
usage
scenario
here,
I
think
like,
although
one
can
debate,
the
scenario
can
be
fulfilled
by
having
another
layer
of
rap
hurt
and
turning
the
STL
another
one
into
announced.
Another
one.
This
something
we
can
discuss
in
OPR
never
benefit.
I
can
tell
the
top
thing
from
my
my
mind
is
in
this
way
we
can
compare
the
versus
our
now
standard
version
side-by-side.
So
if
there's
future
changes,
we
might
be
able
to
catch
that
in
the
CI
and
be
able
to
port
the
like
a
current
fishery.
Our
announced
an
implementation.
F
Right,
so
this
is
the
core
part
right,
so
we
have
this
pass
remover
and
utility
right,
and
these
the
private
implementations
that
we
got.
This
is
a
a
fork
into
the
standard
library.
Implementation
except
span
is
still
not
standardized
of
C++
20
and
I
had
to
end
up
using
another,
yet
another
alternate
implementation
as
an
option,
so
this
is
pretty
much
it.