►
From YouTube: 2021-06-11 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
A
A
So,
on
my
side,
I
have
two
topics.
One
is
from
jonathan
about
the
timing
thing.
The
second
one
is
about
the
hint
pr
and
if
you
have
any
other
topics,
please
open
that
to
the
agenda.
A
Okay,
thanks
jonathan,
so
for
the
timing
that
issue
has
been
around
for
a
while,
and
I
I
think
so
far
it
seems
from
the
java
award.
People
have
bigger
interest
and
the
motivation
is
if
we
just
give
a
very
generic
instrument,
we
ask
individual
developer
to
measure
the
time
themselves.
A
So
the
suggestion
here
is
to
provide
a
dedicated
instrument
for
measure
the
time
and
and
by
default,
we'll
give
user
the
best
time
we
have.
My
little
concern
is
I'm
not
sure
if
in
some
languages
you
will
have
a
hard
time,
for
example,
plus
plus.
That's
why
I
look
for
more
feedback
and
looking
at
the
the
latest
feedback.
I
I
think
at
least
from
from
judge
it
seems
like
c
plus
class.
We
believe
can
we
can
use
the
monotonic
time
and
for
python
it
seems
folks
are
interested,
although
they
don't
see
this
super
urgent.
A
But
it's
it's
a
it's
very
nice
to
have
thing,
and
also
I
I
I
think
from
what
I
read
here.
There
are
bad
practices
and
if
we
can
have
this
instrument,
it
might
solve
the
problem.
So
my
question
here
is:
if
we're
saying
we
want
to
put
this
before
the
feature
freeze.
A
Would
people
think
this
is
something
too
big
like
like
who
can
work
with
me?
I
I
can't
probably
help
on
the
back
part.
I
I
need
support
from
you
guys
on
the
prototype.
A
And
given
I
have
to
work
on
the
sdk,
so
I'll
probably
try
to
scope
down
we'll
just
have
the
default
timer
instead
of
allow
people
to
do
all
the
crazy
stuff
like
they
can
specify
their
own
timer
source
or
something.
So
I
can
do
that.
My
question
is:
who
can
work
with
me
on
the
on
the
spike
and
the
prototype.
A
B
I
can
help
you
out
by
like
reviewing
things
and
and
and
such
things,
and
also
like
showing
you
examples,
though,
in
my
next,
like
one
two
weeks,
will
be
very
hectic.
A
I
mean
if
I
just
send
a
pr
and
it's
revealed
by
some
folks
here
we
got
a
reasonable
confidence,
but
we
don't
have
a
good
prototype
that
that
makes
me
worried,
because
when
I
do
the
previous
one
I
at
least
I'm
running
some
python
prototype
and
I'm
working
with
vector
on
the
c
sharp
prototype
without
a
prototype,
I'm
not
confident
to
send
the.
Even
if
I
got
approval
here,
I
personally
won't
feel
confident.
A
At
least
I
I
want
to
see
we
can,
we
can
explain
in
the
spec.
Here
goes
the
api
example
from
java.
When
you
look
at
the
spike,
you
can
see
the
python
c-sharp
example
I
put
there.
I
want
to
see
some
concrete
example
and
we
can
mention
by
default.
This
one
will
use
sun
timer
and
we
explain
what's
the
requirement
and
if
I
want
to
explain
that
I'll
need
some
concrete
input.
A
B
Don't
don't
we
already
have
it
like?
You
wrote
one
for
for
python,
josh
wrote
one
for
java
and
victor
wrote
one
for
c
sharp.
B
As
far
as
I
remember,
we
do
have
like
all
these
three
measures,
time
yeah.
So
as
far
as
I
remember.
B
A
A
B
Oh
so,
for
for
that
I
can,
I
can
show
a
couple
of
them
for
for
java.
You
can
you
can
check
out
like
how
micrometer
does
this
or
how
the?
How
does
the
api
does
it
look
like
and
also
you
can
check
out,
drop
wizard
metrics.
A
Yeah,
so
can
you
can
you
put
a
pr
or
something
in
the
java
repo
just
to
show
that,
and
so
people
can
comment,
and
we
can,
we
can
use
the
discussion
there
to
see
what
are
the
things
we
need
to
capture
in
the
spec,
for
example,
I
have
the
question:
if
we
use
a
timer
by
default,
are
we
going
to
allow
people
to
specify
different
units?
This
is
my
question
and
I
have
no
idea.
How
should
we
do
that?
A
A
D
E
And
I
can
put
together
a
straw
man
and
go
it's
not
something
that
we
would
create
a
pr
against
the
the
sdk
for,
but
I
can
give
it
just
that
that
has
an
implementation.
B
A
A
Okay,
so
the
next
topic
I
have
is
about
the
hint
pr
so
just
for
folks
who
who
were
not
in
this
meeting
before
so
so
we
have
three
things
that
are
entangled,
and
I
briefly
explained
it
here.
A
But
there
are
scenarios
we
might
want
the
library
owner
to
give
the
hint,
for
example,
they're
saying
I
have
a
histogram
we
want
them
to
be
able
to
describe
these
are
the
default
buckets
that
I
want
my
customers
to
use.
This
is
why
we
have
the
hint
hint
is
part
of
the
api,
so
library
developers
can
use
that,
and
we
also
have
a
concept
called
the
view
which
is
in
the
sdk
spec.
We're
saying
with
the
view.
The
application
developer
has
the
ultimate
power
to
configure
how
they
want
to
con
consume.
A
The
metrics,
for
example,
which
dimension?
Are
they
going
to
take?
Which
dimensions
are
going
to
drop
and
what
are
the
histogram
buckets?
So
initially
I
thought
that
we
want
to
do
the
view,
because
view
is
the
ultimate
like
setting
from
the
developer
and
we
always
respect
the
view
at
the
highest
priority.
A
But
it
seems
when
the
view
api,
spec
pr
is
out
there
a
lot
of
questions
and
some
of
them
are
blocking
questions.
So
people
have
the
question
hey.
If
I
have
a
view-
and
I
have
the
instrument,
what's
the
what's
the
order
when
you
determine
what
we
should
take,
so
I
figure
without
the
the
hint
pr
it's
very
hard
to
explain
the
full
picture
to
people.
That's
why?
In
the
last
meeting
I
decided
okay,
I'm
going
back
to
provide
the
hint
api.
A
So
here
goes
the
hint
api
and
just
to
give
you
an
idea
in
this
pr,
I'm
trying
to
solve
the
problem.
How
how
are
we
trying
to
allow
epic
application
developers
to
do
less
and
achieve
the
default
behavior
from
the
library?
So
you
can
see
here-
and
I
received
some
question,
but
most
of
the
question
I
believe,
is
more
related
to
how
the
isdk
is
going
to
handle
the
hint.
So
so,
please
take
a
look
and
see
if
this
pr
is
clear.
It
helps
you
to
understand.
A
A
So
when
you
review
the
pr,
I
want
you
to
review
both
of
them
and
and
start
to
like
correlate
how
things
are
working,
I'm
going
to
stop
here
and
see
if
anyone
has
questions
like
any
anything,
you
feel
unclear
about
the
hint
view
and
where
we're
going
or
it's
clear
to
you,
because
I
figure
based
on
the
the
pr
comment
and
questions
I
I
got.
I
think
a
lot
of
people
are
still
not
having
a
good
understanding
about
the
hint
and
view
I'm
trying
to
understand.
A
If
this
is
something
too
big
and
we
should
scope
down
or
it's
it's
just
because
people
join
late
without
the
prior
knowledge,
and
I
want
to
get
feedback
to
make
sure
the
hint
and
view
are
something
people
are
reasonable.
Instead
of
we're
pushing
something
that
that
is
not
going
to
be
used
by
by
many
folks,
so
I'll
stop
here
and
collect
feedback.
B
I
I
haven't
checked
those
those
pr's,
but
is
I
I
believe
I
have
like
some
high
level
understanding
about
the
view
api
and-
and
I,
as
far
as
I
know,
it
is
like,
after
you
record
your
measurements
with
one
of
the
instruments,
you
can
basically
convert
those
instruments
into
another
one,
for
example,
if
you,
if
you
use
the
counter,
then
you
instrumented
your
code
with
the
view
api,
you
can
turn
it
into
a
histogram.
Is
that
right.
A
That's
something
you
can
possibly
do,
but
I
wouldn't
say
it's
the
the
number
one
scenario
I
think
number
one
scenario:
is
you
you're
an
application
developer
and
you
take
a
library
that
is
already
instrumented
and
the
library
is
giving
you,
for
example,
a
counter,
and
that
counter
has
a
crazy
number
of
dimensions
like
20
dimensions.
It
has
all
the
created
detail
that
you
don't
need,
and
all
you
want
is
hey.
A
I
want
to
get
the
the
number
of
http
incoming
requests
and
I
only
care
about
three
dimensions:
the
http
method,
okay,
the
http
verb
and
the
status
code,
and
I
want
the
sdk
to
aggregate
all
the
other
dimensions.
So
I
I
don't
get
like
like
one
million
time
series,
so
with
the
view
api
they
give
customers
ability
to
configure
on
the
icd
side.
So
this
the
view
part
is
only
available
for
application.
A
Developer
library
developers
are
not
supposed
to
have
access
to
the
view
it
gives
the
ultimate
saying
for
the
application
developer,
what
they
want,
and
with
that
you
can
imagine
they
can
specify
which
dimension
I
want
and
we
can
potentially
adjust
the
aggregation
if
the
default
aggregation
is
not
what
we
want.
You
can
also
see.
This
is
my
histogram
bucket.
You
can
do
a
lot
of
things
or
like
change.
The
name
change
the
description
if
they
want.
B
This
also
means
that
you
will
be
able
to
basically
like
filter
out
like
accept
and
deny
complete
set
of
metrics.
Yes,.
D
I
have
a
question
riley,
I
read
your,
I
read
your
pr
and
you
have
defined
the
key
attributes
and
the
bucket
are
those
going
to
be
defined
in
the
spec
or
is
that
more
of
a
semantic
convention
thing
or
or
can
people
just
add
what
they
want
in
the
hint
or
so
I'm
trying
to
understand
is
that
while
specifying
the
spec
or
is
that
open-ended.
D
Would
you
clarify
request
yeah,
so
in
the
hint
api
you
you,
you
provide
the
hint
and
in
your
spec
you,
you
specify
two,
the
the
key,
the
attribute
key,
which
is
a
list
of
string
and
then
a
bucket,
which
is
a
list
of
doubles.
So
the
question
is
generically.
You
said
you
marked
them
as
optional.
Are
those
those
two
things
like
right
there
like
online
1000,
you
know
and
1001
are
those
hard
specified.
The
spec
are
those
driven
by
semantic
convention
or
can
people
add.
D
A
A
D
D
A
D
D
A
G
I
I
haven't
read
through
your
pr
yet
riley,
but
I
I
mean
the
principle
behind
the
view:
api
and
the
hint
api
seems
well
understood
to
me.
Okay,
thanks
and
I
will
I
will
read
through
your
pr
shortly
and
I'll
provide
more
feedback
based
on
what
I
find.
E
So
I
think
the
the
question
that
victor
is
asking
is
an
important
one
that
we
should
consider
at
this
point
in
terms
of
how
do
we
assure
that
we
have
specified
it
in
any
way
that
allows
that
future
flexibility
are?
We
are
we
going
to
say
that
the
hint
is
something
that
is
a
list
of
key
value
pairs,
and
then
we
provide
a
hint
semantic
convention
that
describes.
E
I
think
we
probably
could
right
so
have
a
set
of
conventions,
for
these
hints
that
say
attribute
keys
is
a
list
of
strings
that
are
keys
from
the
semantic
conventions
book
is,
is
a
list
of
floats
that
describe
the
the
boundaries
for
buckets
in
a
histogram,
and
then
then
we
could
add
to
that
simply
by
adding
to
the
conventions
for
hints
and
give
us
a
way
to
describe
them
and
potentially
programmatically
implement
them.
The
same
way,
we've
done
the
resource
and
trace
semantic
conventions.
A
E
E
E
G
G
A
A
A
E
Maybe
I
I
think,
to
the
extent
that
we're
treating
these
as
strictly
typed,
you
know
if
we're
to
say
that
hint
is
an
object
that
has
attribute
keys,
which
is
a
list
of
strings
and
buckets,
which
is
a
list
of
floats
extending
that
with
new
values
in
the
future
may
be
problematic.
For
some
languages
like
in
go,
adding
a
new
field
to
a
structs
is
technically
not
a
backwards
compatible
change,
because
there
are
shorthand
initialization
methods
for
structs
that
don't
specify
names
that
would
break
if
you
add
a
field
to
it.
E
So
if
we
want
to
go
that
way
route
we
may
have
to
say
this
should
be
a
map
of
string
identifiers
to
some
type.
You
know
whether
that's
it
go,
be
an
empty
interface
or
something
that
can
be
duct
typed,
which
would
be
much
more
like
your
python
example
here.
I
think,
but
I
think
in
python
that
that's
that's
more
natural,
but
for
more
strongly
typed
languages
that
can
become
complicated,
and
then
that
leads
us
into
the
the
question
of
okay.
E
A
Okay,
I
I
hear
you
so
I
think
what
I
I'm
I'm
missing
here.
It's
not
asking
people
to
do
that.
This
is
just
examples
that
people
might
consider
and
I'll
add
something
here
like
whether
people
want
to
use
strong
type
or
they
want
to
use
a
dictionary.
It's
a
language
decision
and
we
give
that
flexibility.
The
spec
is
not
asking
people
that
they
must
use
stronghold,
because
not
every
language
has
that
philosophy
and
number
two
I'm
going
to
add
here
is
whatever
decision
they
they
made
here.
They
should
also
consider
the
future
extensibility.
D
So
so,
for
based
on
your
structure,
here
I
mean
it
seems
like
if
we
just
loosely
define
this
to
be
it's
a
you
know.
Let's
say
it's
a
map
of
a
key
that
is
currently
defining
attribute
key
and
bucket
to
a
list
of
objects.
Is
that
enough
flexibility
going
forward?
I
mean
without
having
to
fully
you
know,
lock
down
or
define
or
open
up
the
schema,
just
a
basic
scheme
of
being
a
you
know,
a
tuple.
E
Well,
I
think
that
would
be
very
flexible,
but
it
would
then
take
away
the
choice
from
languages
that
did
want
to
be
strongly
typed
and
did
have
good
extension
mechanisms
for
structures
that
are
strongly
typed
if
they
wanted
to
do
that
route.
So
I
don't
know
that
I
would
necessarily
recommend
it,
I
think,
being
just
just
ensuring
that
we're
not
foreclosing
either
path
would
be
good.
A
Okay,
so
I'll
follow
up
and
add
the
clarification
here
and
please
take
a
look
after
I
update
the
pr
thanks:
okay,
jonathan,
it's
your
turn
now.
B
B
B
Yeah
so
when
I
instrument
my
own
code
and
let's
say
I
am
using
open
telemetry,
but
only
the
matrix
part
of
open
telemetry.
F
B
I
be
able
to
add
the
an
example
to
that.
So,
for
example,
I
am
using
a
third
party
in
another
open
telemetry
tracing
library,
should
I
be
able
to
add
like
them
to
to
my
instruments
or
any
arbitrary
tag.
I
want,
like
example,
semi.
B
B
A
A
B
I
am
not
sure
I
have
anything
to
do
with
the
sdk
at
that
point
like
what
I
want
to
do
is
basically
like
one,
then
I
will
add
that
that
example
to
a
recording
that
will
be
like
that
will
not
be
on
the
level
of
attributes
that
will
be
on
the
level
of
the
name
of
the
metric
or
the
description
of
the
metric
or
the
or
the
set
of
attributes
like
next
to
it,
and
when
it
will
be
like
reported,
it
will
not
show
up
as
a
new
attribute.
B
A
Right,
maybe
we
have
different
understanding,
so
I'll
explain
my
understanding,
I
think,
in
the
views.
Example,
in
the
back
end,
you
might
get
something
saying
hey.
This
is
the
total
number
of
http
requests
on
the
server
side,
where
you
end
up
with
the
exception
and
by
the
way.
This
is
one
particular
example.
Here
goes
the
entire
http
requests
payload,
and
this
is
the
body
this
is
the
header.
This
is
what
example
means
to
me.
B
Yeah,
we
are
talking
about
the
same
thing,
but
what
I
am
saying
is
that,
like
will
the
user
be
able
to
do
this
and
add
a
example
to
the
2d
recording.
H
Yes,
so
the
way
well,
if
I
can,
if
I
can
jump
in
please
so
the
way
that
the
things
have
been
designed
up
to
this
point
is
that
the
the
api
user
will
not
be
the
ones
who
say
there's
an
exemplar.
H
What
will
happen
is
the
sdk
which
has
access
to
the
context
during
the
recording
of
a
particular
metric
will
then
pick
out
whatever
span
is
in
the
context
at
that
moment,
when
the
recording
happens
and
put
it
into
like,
if
you're
doing,
instagram
record
it
as
an
exemplar
in
the
appropriate
bucket
of
the
histogram
that
it
for
that
recording
that
was
just
made
so.
H
H
E
You
want
to
record
the
trace
context,
though
right
it
sounds
like
jonathan
is
asking
about
exactly
external
information
and
riley
had
the
example
of
something,
like
an
exception,
a
stacked
race
that
wouldn't
be
in
the
trace
context.
But
you
have
some
information
that
you
want
to
attach
as
an
example
of
non-trace
context,
information
that
should
be
recorded
with
this
metric.
H
A
Yeah,
I
agree
with
strong.
I
I
think
people
are
saying
we
only
use
matrix
api
for
now
and
now
we
need
exemplar.
What
should
I
do?
I
I
think
my
answer
would
be
here
goes
the
actual
api
from
open
telemetry
that
are
not
part
of
the
matrix
api
and
you
should
use
that
to
associate
the
data,
so
we
can
leverage
example,
and
I
I
think,
the
the
in
the
spike
part
examples
should
be
described
in
the
sdk
spec.
A
G
Thank
you,
and
not
at
all,
trying
to
sort
of
design
an
api
on
the
spot,
but
just
to
sort
of
throw
out
like
I
could
envision
possibilities
that
even
could
fit
within
the
current
api
design.
Like
you
could
say,
hey
I'm
going
to
make
the
exemplar
data
I'm
going
to
provide
it
through
as
a
tag,
but
then
I'm
going
to
use
the
hints
api
to
say
that
that
tag
is
not
a
good
tag
to
use
as
a
dimension.
So
you
know
don't
do
that.
G
I'm
not
saying
that's
a
good
design
necessarily,
but
you
could
imagine
you
know
that's
one
possibility
of
how
you
might
fit
it
in,
but
I
I
do
agree
with
john,
like
I
think
otech
would
be
great.
I
think
you
know
design
work
and
api
work
to
sort
of
figure
out
how
the
pieces
fit
together
is.
Is
the
right
next
step,
yeah.
H
Because
I
think
there
actually
will
be
data
modeling.
That
needs
to
happen
as
well,
because
I
believe
I
haven't
looked
at
the
data
model
in
detail
recently,
but
I
believe
in
the
data
model,
the
only
thing
that
is
allowed
for
an
exemplar
is
a
trace
id
span
id,
like,
I
don't
think,
there's
other
exemplar
data
in
the
data
model.
D
H
I'm
going
to
go
look
at
the
data
protocols.
B
Absolutely
absolutely
so,
just
one
quick
follow-up
question
with
that
so
by
the
definition
of
examples
will
be
only
an
sdk
feature.
It
should
go
to
the
same
timeline
as
we
have
right
now,
so
it
should
go
to
feature
freeze
by
end
of
july
right.
A
B
Is
is
that
defined
anywhere
right
now,
or
it
is
something
that
that
should
be
defined
later.
A
So
the
ick
spike
is
still
early
stage
in
the
experimental
version.
I
tried
to
capture
the
essential
things,
such
as
the
explorer
processor
and
the
view,
and
and
also
the
isdk
meter
provider
part,
because
people
can
configure
that
I
think
exam
exemplar
can
be
a
strategical,
so
any
example
or
covering
all
the
sdk
record
exporters
like
otlp,
which
formats
those
things.
I
I
think
it's
a
nice
to
have
it's
not
a
blocker
for
the
experimental
release.
H
So
I
did
just
look
at
the
data
model,
so
I
mean
victor
you.
You
are
correct,
it's
basically
the
recording.
So
the
number
that's
the
number
that's
there
and
a
set
of
attributes
with
optional
transcendence
vanity.
H
I
mean
I
assume
that
attributes
are
also
optional,
but
the
the
at
least
as
I
understood
the
goal,
is
that
the
attributes
are
some
filtered
set
from
the
span
that's
being
recorded,
but
it
doesn't
actually
make
it
clear
in
the
data
model
spec
that
that
has
to
be
the
case.
H
So,
anyway,
I
think
I
still
think
it's
a
good
idea.
We
there
should
be
an
otep
to
describe
what
this
api
might
look
like,
since
it
would
be
in
addition
to
the
api
that
I
don't
think,
riley
or
anyone
has
really
been
considering
by
the
way.
This
is
a
typo
or.
G
Okay,
yeah,
I
honestly,
I
don't
recall,
hearing
it
until
probably
doing
this
work.
So
even
though
I
I
don't
think
it's
a
jargon
word,
I
think
it's
a
legit.
You
know
regular
english
word.
H
A
Okay
and
anything
else
from
you
jonathan,
regarding
this
topic,
so
thank
you.
Okay,
so
I'll
probably
need
to
capture
this.
If
there's
no
github
issue
I'll
I'll
check
and
make
sure
we
capture
that
in
our
project
board
and
yeah,
we
still
have
15
minutes
left.
I
want
to
see
if
anyone
still
has
the
question
about
the
intense
view.
I'm
happy
to
answer
here,
because
I
figure
like
there's
still
a
lot
of
questions
and
many
folks
don't
seem
to
understand,
but
I
believe
there's
no
other
discussion
topics,
so
we
can.
I
Have
sorry
go
ahead?
Yes,
sorry,
I
do
have
questions,
they
are
not
related
to
hint
or
any
other
topic
that
has
been
displayed.
Okay,.
A
F
All
right,
yeah,
thank
you,
so
yeah,
just
a
quick
question.
Today
the
python
sega
was
about
pretty
much
what
happened
to
the
bound
instruments
that
we
had
before.
A
Okay,
so
the
bond
instrument
is
not
in
the
scope
for
the
initial
stable
release.
We
we
try
to
scope
down
and
make
sure
we
can
release
a
stable
version
of
the
spec
and
later.
If
we
need
anything
that
might
require
us
to
have
high
performance,
we
can
add
those
things
on
top
of
these
existing
stuff
without
breaking
things.
So
it's
not
it's
not
in
the
current
scope,
which
means
when
we
release
the
matrix
api
at
stable.
I
don't
expect
the
initial
release
will
cover
the
bomb
instrument.
A
Nice.
Okay,
thank
you,
and
if
you
see
a
big
need
from
the
python
stick,
I
would
suggest
to
follow
up
your
github
issue
and
we
can
track
there
and-
and
I
probably
can
make
that
in
the
in
the
upcoming,
for
example.
Then
the
next
release
of
the
matrix
after.
I
F
I
F
So
just
the
other
question
pretty
much
the
same,
but
with
the
views
is
view,
something
that
are
we
planning
to
implement,
or
should
we
try
to
implement
in
these
new
prototypes.
A
Yeah,
so
view
is
something
that
I
I
I
think
most
of
the
folks
would
agree
that
vo
is
required
to
have
a
reasonable
experience.
So
without
view
I
I
think,
will
not
be
confident
to
release
a
stable
version
of
the
ic
case.
Back
with
that,
I
I
think
view
and
hint
there
are
important
hints,
probably
without
hint,
I
think
we're
we're
able
to
survive,
but
without
the
view
we're
basically
asking
application
developer
to
take
all
the
dimensions
from
the
library
which
is
not
going
to
work.
F
All
right,
so
I
just
checked
the
sdk
document
and
it
pretty
much
just
says
views
is
in
to
face
so
I
guess
should
we
should
we
expect
that
news
will
be
defined
in
the
sdk
document
yeah
for
purposes
of
develop
developing
this.
I
A
A
Pr
this
is
the
view
part
pr
right.
I
can
put
the
link
here
so
to
take
a
look,
and
it
will
give
you
some
idea
and
even
we're
still
iterating
on
this.
I
wouldn't
suggest
you
spend
too
much
time
writing
the
prototype,
because
what
what
you
did
there
might
change
very
quickly.
So
what
my
suggestion
would
be,
is
you
look
at
the
pr
and,
and
you
put
comments
contributing
ideas
and
and
once
we
agreed
that
pr
got
merged?
I
I
think
it's
a
good
sign
that
we
can
start
to
do
more
prototype.