►
From YouTube: GitHub Quick Reviews
Description
Powered by Restream https://restream.io/
A
A
So
hello,
friends
good
morning,
good
afternoon,
good
evening,
today's
backlog
review.
Hopefully
we
can
make
some
dents
into
our
remaining
backlog.
Let's
start
from
the
top
down,
maybe
and
then
just
see
how
far
we
get
all
yet.
A
Of
course,
I'm
intentionally
not
sharing
my
screen,
because
I
thought
we
raised
this
difficulty
level
to
level
three
where
you
have
to
do
blind
reviews.
A
C
B
Way:
marshall.,
whatever
string
from
pitter
and
then
span.
C
Yeah
yeah:
this
would
be
a
no
allocation
way
to
do
it
really
really.
I
think
this
should
actually
be
a
pair
of
apis
one,
the
actual
workhorse,
which
is
like
just
wcs
len,
like
here's,
a
char
star,
tell
me
how
many
characters
it
is
before
you
hit
the
null
terminator
and
then
a
helper
api.
On
top
of
that,
which
is
okay.
Now
that
you
have
a
char
star
and
a
new
end
make
a
span
from
it.
E
C
Because
index
of
is
free
to
assume
that
the
entire
span
that
you
pass
into,
it
is
validly
dereferencable
now
that
it
could
start
at
the
end
and
work
its
way
backwards
or
something
if
it
wants
to
yeah.
I
mean
it
would
be
a
silly
implementation,
but
it
would
fit
the
contract
with
the
api
right.
D
C
Example,
if
you
pass
in
an
infinite
length
span
like
we
might
just
say:
oh,
let's
just
read
the
unaligned
first
vector
from
it
and
then
look
for
the
null
terminator.
Well,
that
unaligned,
read,
could
actually
seg
fault.
C
A
E
B
Yeah
I
mean
this-
this
wouldn't
really
make
sense
for
t
star
because
we're
we're
using
the
char
star
semantic
of
being
terminated
by
a
zero
value
like.
B
Only
if
you're,
asserting
that
the
bytes
are
are
ascii.
D
C
I
I
don't
think
that's
necessary,
because
again
only
only
wide
and
narrow
strings
are
the
things
that
people
care
about
here,
so
that
would
be
char,
star
and
bite
star.
The
yeah.
B
B
E
C
A
B
A
B
C
A
C
A
Yeah
I
mean
yeah,
I'm
not
advocating
for
generic.
I
was
just
curious,
but
like
yeah
I've,
I
completely
missed
the
point
in
the
beginning
where
this
was
sniffing,
the
length
yeah
yeah
and
then
it
would
return
span
of
char
right
and
span
off
bite,
probably
right.
B
A
B
B
Okay
and
after
all,
it's
s
z,
because
it's
string,
zero,
terminated.
E
C
F
C
C
Length
you
can't
because
sterling
presumably
would
return
a
new
end
and
span
create
span.
If
the
string
length
exceeds
intel
max
value.
Will.
E
B
A
Citation
needed
yeah,
I
mean
I
was
in
an
impression
show-
was
signed
as
well,
because
I
think
it
has
an
interesting
conversion
to
end
as
well.
So,
like
that's
seems
odd.
If
that
would
be.
B
A
Fair
enough,
I
guess
that's
true
fair
enough.
I
mean
I
mean
my
care
level
is
low.
I
mean
like
if
I
don't
know
how
popular
utf-32
is
these
days,
but
like
yeah
I
mean
I
have
zero
problem
with
the
api.
So
if
I
mean
whether
we
do
indoor
you
win
to
me
is
also
largely
irrelevant,
because
it's
a
pointer
anyway,
so
you
don't
really
care.
A
A
But
how
often
do
you
create
buffers
over
that?
I
don't
know.
A
You'll,
probably
from
our
team,
the
closest
person
who
can
do
it
isn't
from
china
yeah,
I
don't
know
I
mean
if
people
feel
strongly.
I
I
don't
mind
the
end.
One.
B
E
A
D
D
D
C
C
D
E
C
B
But
those
are
dealing
with
code
points
which
are
always
defined
to
be
a
less
than
24
bit
value
as
opposed
to
ucs,
32
or
so
utf-32,
which
is
dealing
with
a
I
mean
we
can
look
at
the
unicode
spec
and
see
what
they
call
it.
C
But
jeremy's
right,
though
those
apis,
the
apis
that
take
end,
often
work
with
code
points,
whereas
utf-32
is
a
sequence
of
code
units
which
are
ostensibly
scalar,
scalar
values
not
code
points.
The
the
distinction
is
it's
valid
to
pass
an
unpaired
surrogate
character
into
many
of
the
apis
that
take
end,
but
it
would
never
be
valid
to
have
those
inside
of
utf-32
string.
C
Yes,
just
because
it's
an
it's
an
older
api,
but
the
apis
in
the
globalization
library
do
tend
to
understand
unpaired,
surrogates
and
leave
them
alone
like
when
you
try
to
do
a
case
conversion,
for
instance,
okay,.
A
B
B
I
think
that
putting
it
as
you
went
star
makes
it
like
it
is
a
little
bit
better
signaling
and
the
fact
that
you
have
to
write
the
extra
conversion
means
that
you're
less
likely
to
do
it
on.
You
know
numerical
data
that
that
it
is
a
stop
and
think
about
this
and
make
sure
it's
a
string.
So
I
think
it's
both
more
honest
and
the
the
slight
hurdle
of
please.
B
C
A
Yeah
yeah,
but
that's
what
I'm
saying
like
I
mean.
Usually
when
we
add
the
method
we
already
have
prior
art
and
then
we
can
make
a
decision.
Let's
hold
back
utf
yeah
32
until
we
need
it.
B
A
C
C
Literally
marshall.sterlingmarshall.wcs
lin
like
sterling,
would
be
lowercase.
Strlen,
wcs
lin
would
be
lowercase,
wcs
lem.
A
B
C
B
C
A
A
It
has
block
copy
byte
length,
get
buyer
and
set
byte
and
they
all
deal
with
like
array,
the
generic
array
type.
You
know
uppercase
array
yeah,
I
think
memory
copy
is
the
only
one
that
takes
pointers,
and
but
I
mean
it
seems
like
it's
more
like
that.
Like
martial
is
such
a
shirt
today
I
would
not
put
an
api
there
if
you
actually
care
of
using
it,
because
martial
is
really
more
about
interrupt
right
versus
this
one
is.
I
have
some
memory
and
I
would
like
to
find
the
zero
in
it
sure.
A
A
C
F
B
C
Yeah
and
the
behavior
of
the
api
is,
if
you
pass
in
a
null
pointer,
it
throws.
If
you
pass
in
a
nominal
pointer,
it
either
returns
your
avs.
B
Yeah
just
like
sterling
yeah,
so
I'm
going
to
ask
the
sdl
question
of
do
we
do
we
need
them
to
take
a
do?
We
need
the
stir
inland
variant
of
where
you
tell
it
the
stop.
Looking
after
this
point.
C
E
Well,
so
the
purpose
of
the
length
is
for
security,
so,
for
example,
in
c
it
is
fairly
common
to
define
a
fixed
width
buffer
as
part
of
struct.
For
example,
you
might
say
the
maximum
length
of
this
name
is
128
characters,
so
you
never
want
to
read
past
128
characters,
but
the
null
terminated
character
could
be
at
digit.
You
know
14,
because
the
name
of
this
is
5700
xt.
B
Yeah
so
yeah,
I
I
think
that
we
would
need
it,
possibly
as
a
defaulted
to
negative
one.
I
mean
the
question
is:
what
type
are
we
taking?
Presumably
we're
taking
a
another
size
t
as
the
input,
and
then
that
means
that
we
would
need
proper
overloading
where
one
of
them
is
sterling
and
one
of
them
is
stir
inland.
E
B
A
A
I
mean
the
one
yeah
I
think
that
seems
more
reasonable,
like
I
don't
like
the
index
of
suggestion
for
span,
it's
not
that
it's
wrong,
it's
just
it's
a
completely
different
way
of
doing
it,
and
it's
not
very
obvious
that
that's
what
you're
supposed
to
do
when
you
have
a
length
right.
So
I
actually
like
the
idea
of
that.
Just
having
two
overloads
that
remind
you
that
there
is
a
length
you
may
want
to
pass
in.
If
you
have
one
or
maximum,
I
guess
is
the
way
to
put
it
right.
D
C
So
index
of
again
is
it's
assumed
that
you're
going
to
return
negative
one
if,
if
the
null
terminator
isn't
found
before
the
maximum
length
is
hit,
whereas
with
a
proper
sterling
api,
you
would
want
to
return
the
maximum
length
instead
of
negative
one.
E
A
B
Correct
return
value
would
be
an
unsigned
native
integer.
B
A
E
B
A
B
I,
like,
I
think
that
z
like
null,
is
the
name
of
the
character
whose
value
is
zero,
so
I
could
be
okay
with
null
it's
just
a
little
weird
because
we're
not
using
byte
star
as
bite,
star
we're
using
bite
star
as
actually
char
star
and
and
there's
a
little
bit
of
a
signal
disparity
there.
So
I
what
I
like
for
zero
is
it
it
seems
being
a
little
clearer
that
we
mean
a
zero
value
in
a
byte,
not
we're
using
this
byte
star,
as
just
an
arbitrary
memory
address
that
we're
interpreting.
C
Yeah
considering
this
is
primarily
meant
for
interrupt
scenarios.
I
I
do
think
noel
is
the
better
name,
because
when
you're
dealing
with
c,
it
is
called
null.
A
Nice
all
right,
then
you
bike
shed
this
one
to
death.
I
think
alternate
you
can
bike
about
random,
okay,.
E
So,
basically,
today
there
isn't
a
way
to
get
a
64-bit
random
value
without
having
to
construct
it
yourself.
This
is
mainly
a
convenience
api
for,
in
fact,
getting
a
64-bit
integral
in
the
right
length.
A
E
B
E
Notably,
the
things
that
are
missing
from
system
random
today
are
long.
You
long
and
I
believe,
single
from
the
primitive
types
the
others
are
fairly.
Things
like
byte
short,
are
fairly
easy
to
do.
Using
the
existing
int
apis
yeah.
E
A
A
C
B
A
A
E
A
I
mean
it
seems
like
weird
that
I
mean
normally,
I
think,
of
these
apis
as
being
constrained
by
some
sort
of
business
size
right,
and
it
seems
weird
that
we
have
like
some
weird
slice
through
this
whole
thing.
I
think,
given
that
we
normally
do
these
things,
they
will
probably
add
next
float
as
well,
yeah
or
sorry.
Next,
single,
this
seems
reasonable
to
have.
A
C
C
So
presumably
the
correct
behavior
of
that
call
site
would
actually
be
if
you're
on
a
64-bit
platform
call
rand.nexlong
instead
of
rand.next.
A
E
E
So
if,
for
example,
if
you
said,
I
want
a
double
in
2
to
the
power
of
53
and
2
to
the
power
of
54,
you've
got
basically
two
values
to
pick.
A
Yeah,
that's
what
I
heard
don't
be
fancy
all
right,
so
that
means
we
have
next
in
64
next
in
64,
with
long
max
radio
1964
with
long
min
value
and
long
max
value,
and
we
have
next
single
with
no
arguments.
B
A
They're,
the
non-natural
ones.
C
I
I
think,
that's
what
we
already
do
for
unsigned
32-bit.
D
A
A
Anyway
of
three
years,
so
here's
an
issue
from
actually
from
this
year
never
mind.
C
There
are,
there
are
overloads
largely
of
existing
apis
or,
I
shouldn't
say,
overloads
analogs
to
existing
apis.
C
C
Yeah
so
for,
for
example,
append
all
lines
async
the
current
overloads.
Take
I
enumerable
of
string.
These
proposed
overloads.
Take
I
enumerable
of
read-only
memory
of
char,
I'm
not
sh,
I
I
don't
know
if
we
have
existing
apis
that
take
like
I.e,
numerable
of
memory.
A
No
and
I
have
to
say
like
I,
unless
somebody
gives
me
a
scenario
for
this,
I
have
a
hard
time
buying
that
scenario,
because
it
seems
like
it's
the
it's
it's
this.
On
the
one
hand
side
you
have
low
level
information
like
with
only
memory
and
the
other
side.
You
have
this
super
high
level
thing,
which
is
I
enumerable
of
something,
and
then
you
want
to
write
it
to
an
existing
file,
which
seems
now
you're
in
the
0.003
percentile.
B
Does
read
only
sequence,
it
only
gives
back
spans
right,
not.
C
Memories-
that
is
a
good
question
I
feel
like.
I
should
know
this.
B
Because
if
it
gives
back
memories,
then
you
could
theoretically
have
a
read-only
sequence
of
char
and
write
a
yield
return
enumerable
that
just
returns
all
of
its
subsequent
span
pages.
It
gives
back
pages.
So
that
would
be
the
scenario
for
that
now,
maybe
we
say
it's
not
common
and
we
don't
care
yet,
but
that
would
be.
F
B
So
if
yeah,
if
you
had
read
only
sequence
of
char,
then
I
have
to.
C
A
I
think
he
was
just
trying
to
answer
the
question
of
like
when
would
you
ever
have
a
combination
of
read-only
memory
and
I
innumerable,
and
he
was
saying
that
if
you
traverse
a
read-only
sequence,
you
would,
and
you
would
have
an
iterator
that
basically
would
give
you
the
memories
to
the
individual
items.
In
that
thing
that
you
found.
B
But
the
ones
the
simple
ones,
the
append
all
text
of
read-
only
span
chart
contents.
Again,
that's
a
that's
a
quick
way
of
writing
a
substring
to
a
file
without
needing
to
make
a
copy
yeah.
So
all
the
ones
that
take
read-only
memory
of
char
instead
of
string
that
aren't
themselves
part
of
another
generic
composition
seem
easy,
but
back
to
what
you'd
said
levi
of
not
all
of
these
are
overloads.
Some
of
them
are
new
concepts.
B
B
Okay,
because
like
if
we,
for
example,
didn't
have
writal
text
that
took
path
and
contents
without
the
encoding,
then
we
should
add
the
string
overload
if
we
were
adding
it
here.
But
if
these
are
all
just
overloads
of
drop
in
replacement,
rosh
char
instead
of
string,
then
then
they're,
fine
or
yeah.
You
know
I.
A
Mean
I
would
go,
I
mean
I
would
say
like
the
only
thing
I
see
in
this
issue
here
is
this:
is
he's
basically
pointing
out
the
first
one
would
be
you
just
called
to
bite
array
which
would
allocate,
which
I
don't
know
like
the
only
reason
why
I
use
these
file.
Convenience
helpers
is
usually
because
the
things
I'm
processing
is
small,
or
I
just
you
know,
cook
up
a
quick
and
dirty
command
line,
app,
which
I
don't
care,
what
performance
it
does.
A
And
yes,
if
I
test
something
that
is
low
level,
I
can
see
myself
wanting
to
persist,
something
that
is
read
only
memory,
because
that's
you
know
the
algorithm
that
I'm
actually
testing,
but
then
I
also
don't
care
about
perfs.
I
wouldn't
mind
doing
two
array
in
those
cases
right,
but
like
the
second
one
is
a
bit
more
nuanced.
I
think
he's
saying
that
you
know
creating
a
file
stream
instance
with
an
appropriate
file.
Options
is
verbose.
A
That
is
the
thing
I
think
we
may
want
to
address.
It
seems
reasonable
for
me.
We
already
have
these.
You
know
open
and
create
ones
on
file.
They
give
you
back
a
file
stream.
Maybe
we
should
have
one
that
would
deal
with
creating
something
that
is
truly
asynchronous
right,
because
that's
today
also
not
great
because
file
streams.
Last
argument
is
async.
You
know
and
that's
verbose
right,
but
that's
the
only
thing
that
I
see
myself
doing
for
file
to
say
like
yeah.
B
That
I
mean
I
feel
that
in
the
design
guidelines
that
the
file
class
itself
actually
comes
up
as
that
it
was
added,
because
there
was
a
collective
feeling
that
I
have
this
data
already
built.
Why
do
I
need
to
say
you
know
using
a
file
stream?
B
Go
write
this
like
that,
especially
the
text
ones,
because
you
have
to
open
up
the
file
stream.
You
have
to
open
the
the
string
writer
version
of
it
or
the
text
writer
or
whatever,
and
and
that
it's
the
like
wow.
It
feels
like
saving
this
text
to
a
file
or
saving
these
bytes
to
a
file
when
I've
already
have
them,
and
I
don't
need
to
do
it.
There
are
no
pieces
that
I
can
do
because
that's
the
pin
doll
lines.
Async
is
totally.
You
know
what
just
build
the
stream
right.
B
Sorry,
the
pinned
all
lines
of
innumerable
is
like
yep.
Until
that
we
feel
that's
a
common
scenario.
You
can
do
it
with
opening
the
file
stream,
but
the
appendal
text
of
read-only
span
chart
contents
or
the
appendal
text.
Async
of
read-only
span
chart
contents
to
me.
That
seems
like
we're
just
making
an
accelerator
for
a
thing.
People
would
do
commonly
and
it
totally
meets
the
the
purpose
of
the
file
class
of
you
have
your
data
already.
You
just
want
to
save
it
to
a
file
you're,
avoiding
the
dance.
A
So
I
agree
with
that:
it's
just
that.
I
think
we
need
to
be
careful
what
data
we
are
supporting
right
because
it
seems
weird
to
me
to
say
I
have
all
my
data
already
in
memory,
but
I'm
I'm
highly
geared
for
performance.
So
I
do
everything
with
only
memory
chunks,
but
then
I'm
kidding,
I
couldn't
be
bothered
with
writing
two
lines
of
code
that
actually
invited
to
a
file.
That
seems
like
a
weird
combo.
B
If
we
had
span
and
memory
back
in
the
beginning,
I
think
that
there's
a
good
chance
that
we
would
have
said
string.substring
returns
a
slice
instead
of
a
new
string,
and
then
this
would
have
been.
The
obvious
thing
to
take
is
take
an
arbitrary
slice
of
a
string.
B
A
But
the
restrictions
are
the
same
right
I
mean,
if
I
give
you
read-only
span
of
char,
that
is
by
this,
is
nowhere
near
as
convenient
as
a
string
for
the
very
simple
reason
you
can't
assign
it
to
a
field.
For
example,
right,
like
you,
have
all
the
red
struct
semantics
that
that
come
with
that
which,
in
mainland
scenarios
you
don't
care
like.
Why
would
you.
B
A
A
I
have
a
hard
time
seeing
that,
but
yeah
it's
a,
I
mean
it's
a
fair
point
I
mean,
like
I
mean
to
me:
it's
more
like
file
is
really
meant
as
a
as
a
as
a
as
a
fast
way
to
write
a
file
right,
because
we
have
string
reader
stream,
reader
and
and
and
stream
to
basically
make
arbitrary
processing
of
files,
and
I
think
to
me
the
way
I
read
the
book
was
the
the
incident
in
the
2-0
time
frame
was
people
who
came
from
a
java
background
got
it.
A
People
who
came
from
any
other
background
didn't
get
it
because
they
were
not
as
they
were.
Not
so
you
know
exposed
to
this.
You
know
combinating
things
in
the
pipeline.
Basically,
but
then
we
are
saying
here
we
have
these
building
blocks
that
people
use.
In
context
of
these,
you
know
things
but
they're,
not
aware
of
stream
or
or
or
you
know,
text
writers,
and
that
seems
somewhat
backwards
to
me.
A
Like
I
mean,
I
think,
if
you
already
have
data
in
memory,
you
most
likely
have
it
in
a
convenient
fashion
right,
which
is
a
by
the
way
or
a
string
like.
I
really
have
a
hard
time
seeing
somebody
I
already
have
a.
I
knew
a
read-only
memory
of
char
and
I
just
want
to
have
it
in
one
line
to
a
file.
Then
you
know
the
alternative
would
have
been
three
lines
of
code
right,
we're
not
talking
about.
You
know
50
lines
of
code
right,
so
it's
not
like
it's
not
like
it's.
A
B
A
D
B
That
you
probably
don't
already
have
the
innumerable
as
a
convenient
thing
and
once
you
had
to
write
the
yield
return,
you
could
have
just
written
the
span
rate
yourself,
but
the
the
write
all
text
of
read-only
span
of
char
contents.
Again,
I
think
that
that's
a
you
did
some
trimming
or
some
some
manual
splitting
and
it's
the
you.
You
had
some
some
data
and
you're
writing
it
into
three
different
files.
You
have
it
in
one
string.
It
just
saves
on
the
copies.
B
A
That's
the
thing
to
me:
convenience
kind
of
like
it's
somewhat,
it's
somewhat
backward
if
the
thing
you're
doing
is
already
so
convoluted
that,
but
now
the
persistence
is
not
the
thing
that
breaks
it.
It
seems
a
bit
weird.
You
know
what
I
mean
like
I
mean
you
already
did
extraordinary
gymnastics
to
get
these
only
memories.
To
begin
with
or
read,
you
know
spans
right,
you
usually
don't
have
them
that
handy
right,
that
they're
not
floating
around
that.
Much
that
that
that
is
a
common
thing.
A
I
mean
you
have
strings
all
the
time
I
get
it
even
by
the
ways
are
already
somewhat
rare
right,
but
then,
when
we
actually
talk
about
spans
and
memories,
now
you're
really
talking
about
the
long,
the
you
know
the
really
long
end
of
the
spectrum,
and
these
are
mainline
apis.
So
do
we
really
want
people
when
they
write
their
first
sharp
program?
They
want
to
just
write
a
text
file
and
they
call
write
all
text
and
they
open
intellisense.
There's
like
five
overloads
or
eight
overloads
with
you
know
unspeakable
names
like.
C
We've
had
this
question
posed
in
the
past,
where,
where
is
it?
Is
it
reasonable
for
us
to
keep
using
like?
What's
the
intellisense
experience
like
as
a
crutch
for
adding
or
not
adding
an
api?
But
I
definitely
do
agree
with
the
overall
point
you're
trying
to
make,
though,
like
the
string
is
clearly
an
exchange
type
used
commonly
within
the
ecosystem
span
and
memory
tend
not
to
be
frequently
used
exchange
types.
They
tend
to
be
things
that
you
pass
down
to
leaf
apis,
but
not
necessarily
like
between
different
modules.
A
Yeah
I
mean
to
me:
it's
like
you
know.
If
you
have
a
read-only
span
of
char.
How
fast
can
I
get
a
string
out
of
that
right
so
that
I
can't
call
write
all
text,
for
example
like,
and
if
this
is
literally
like
you
know,
two
string
or
two
array,
then
I
don't
care
it's
like
yeah
yeah,
it's
not
performant
but
like.
If
you
really
wanted
to
perform
it,
then
I
think
it's
okay
for
you
to
write
the
you
know,
file.open,
you
know
and
then
stream.
A
B
That's
fair,
so
I
mean
I'm
willing
to
believe
the
or
be
convinced
by
the
suggestion
of
that
file
is
a
type
that
we
want
to
not
like
that.
It's
too
early
in
the
learning.net
pipeline
to
be
suggesting
memory
and
span
like
that
is
a.
I
don't
know
that
I
100
believe
it,
but
I
that's
an
argument
that
makes
sense
to
me
and
again
definitely
I'm
with
you
on
the
you.
You
don't
have
an
I
enumerable
of
read-only
memory.
B
So
this
is
okay
to
me,
but
if
we
want
to
say
no,
that's
also
okay
to
me
for
the
ones
without
the
iron
rules.
B
B
Next
to
the
string
thing
that
I
understand,
I
guess
what
seems
weird
to
me
is
that
it
took
string
path
instead
of
rosh
jar
path
that
like,
if
once
it's
going
to
take
a
a
rosh
jar
on
one
of
them.
Why
didn't
it
take
it
in
all
of
them,
but
yeah?
So
maybe
maybe
this
is
fine
for
no,
because
the
workaround's
not
hard,
and
if
you
want
the
accelerator,
you
can
write
it
yourself.
Three
lines.
C
B
B
C
A
Yeah
yeah,
could
you
post
a
comment
and
take
carlos
on
it,
because
I
think
he
owns
that
sure
I
think
yeah.
The
question
to
me
would
be
like:
should
we
make
at
least
creating
an
asynchronous
file
stream
easier
right?
Because
today,
that's
like
you
have
to
call
the
file
sim
overload
that
takes
like
12
arguments
or
something
right?
A
A
B
B
A
C
I
think
what
jeremy's
saying
is,
we
should
add
and
overload
the
takes
os
platform.
A
I
think
you
lost
me,
I
think.
C
A
A
The
funniest
thing
is
when
we
approve
a
completely
different
api.
It's
not
like
we've
written
it's
just
instead
of
doing
this.
How
about
we
give
you
the
building
block
over
there?
What
you
asked
for?
Well,
that's
what
you
get.
Is
there
any
of
context
on
this
one
here.
B
I
was
gonna
say,
didn't
we
already
talk
about
it,
but
I
realized
I'm
looking
at
teams
instead
of.
C
Yeah,
what
are
they
wanting
to
add?
Ssh,
ftps
and
websockets.
B
So
ws
and
wss
totally
makes
sense
because
we
added
websocket
support
ftp,
I
mean
these
are
just
strings,
so
they're
fine.
A
B
A
C
Yeah
well,
both
corey
and
mija,
commented
on
the
issue
saying
that
they
they
think
this
is
a
worthwhile
change,
and
I
would
imagine
that,
had
this
involved
a
lot
of
co-churn,
they
would
have
spoken
up.
A
Yeah
I
mean
carol
seems
to
be
more
in
the
like
what
waited
as
a
dad
thing,
but
it
seems
he
was
convinced
at
the
end
that
it's
just
a
shorthand
for
well-known
schemes.
I
mean
it
seems
fine,
I
mean
like
it
was
literally
just
a
well-known
string.
I
think
it's
totally
fine
to
have
that.
D
B
That
is
very
vague
right
and
the
only
thing
that
I
really
see
in
it
is
yes,
we
have
in
the
uri
parser,
we
have
a
table,
we
build
the
thing
once
and
the
all
the
table
has
is
like
if
we
understand
the
scheme,
what's
the
default
port
and
what
are
some
syntax
flags
that
are
associated
with
it,
which
I
don't
know
what
our
breaking
change
model
would
be
at
this
point
like
would
we
start
rejecting
ssh
colon
values
that
we
used
to
accept,
but
that
would
be
the
that
would
be
the
non-api
follow-up
is.
B
B
B
C
Ui
parsing
support
for
it
like
looking
through
the
uri
parser.
It
actually
looks
like
it's
not
all
that
difficult
to
add
knowledge
of
new
protocols,
because
it
looks
like
they
just
maintain
a
set
of
internal
of
internal
requirements
that
a
particular
structure
has,
and
it's
just
a
bit
wise
combination
of
that.
So
if
they
wanted,
if
you
can
describe
ssh
using
their
existing
flags,
then
it's
just
choose
the
right,
bitwise
combination
and
go.
D
B
Uri
doesn't
throw
if
you
give
it
the
string,
you
know
ssh
colon
sure
whatever
and
then
now
they
add
the
the
uri
parser
logic
for
it
and
it
fails
then,
but
I
don't
know
what
their
failure
modes
are.
So
that's
things
for
them
to
consider,
but.
C
B
Maybe
but
yeah
so,
but
for
websocket
and
wss.
Those
are
both
already
supported
by
uri,
parser,
yeah,
sftp
and
ftps
are
not
ssh
is
not
intelnet
is
so.
Some
of
these
are
we've
done.
The
we've
already
done
the
parser
work,
exposing
the
cons.
It
has
no
real
ramification.
The
other
half
are
the
question
of.
Do
we
also
add
special
casing
to
them
in
uri
parser
and
then
what
breaking
change
does
that
entail
yeah?
C
That
yeah
and
it's
it
would
be
up
to
up
to
ncl
if
they
want
to
support
like
sftp
and
ftps,
for
instance.
I
I
honestly
didn't
realize
ftps
was
a
thing
like
I
I
thought
you
know.
Sftp
was
kind
of
the
universal
standard,
but
whatever.
B
Yeah,
it's
a
little
more
complicated
than
like
smtp
with
start
tls
versus
smtp
over
a
tls
channel
like
because
I
think
sftp
came
out
of
ssh
and
ftps
is
do
ftp
with
tls
and
that
they're
actually
very
similar
letters
for
very,
very,
very
different
protocols
like
one
of
them
ie
supports
and
the
other
one.
It
doesn't.
D
A
Are
you
open
there
yeah?
So
I
don't
think
this
one
is
ready
really
like
it.
There
was
a
lot
of
feedback
on
the
api.
I
should
probably
mark
this
back
as
needs
work.
A
So
basically,
what
this
was
for
right.
So
this
is
the
we
have
this
thing
in
wpf
and
winforms,
where
sorry
not
there,
we
have
it
in
the
visual
basic
dll
that
you
can
basically
say.
I
want
a
single
instance
app
and
when
you
do
that
they
do
the
heavy
lifting
of
creating
a
mutex
for
your
own
startup
and
then,
if
the
app
is
already
running,
they
basically
forward
the
comment
on
arguments
to
the
existing
instance.
A
So
that's
actually
quite
nice
because
it
means
it's
very
easy
for
you
to
write
a
multi-instance
app
like
in
a
document
editor.
Basically,
and
then,
if
somebody
opens
another
instance,
they
just
forward
you
the
path
to
the
existing
instance
and
you
just
open
the
file
and
the
existing
one,
which
is
actually
quite
nice.
A
The
problem
is
that
we
want
some
building
block
that
is
lower
level
than
vb
so
that
we
can
support
it
directly
in
forms
without
requiring
all
of
the
vbb.dll
to
be
used
by
someone,
and
you
can
also
make
some
aesthetic
arguments
that
c-sharp
people
are
offended
by
referencing
vb.dll
in
order
to
do
signal
instance
apps.
A
But
the
question
is
what
it?
What
that
would
mean
and
the
when
I
talked
with
carefully
and
there
was
an
idea
about
providing
a
building
block
like
this?
We
don't
really
expect
this
to
be
used
by
most
people.
A
If
you
know
we
expect
people
to
use
the
the
the
apple
specific
ones
and
I
think
that's
fair
feedback,
and
I
think
there
was
not
much
traction
on
the
vb
side
on
the
of
this
or
the
winform
side
of
it.
So
I'm
inclined
to
just
say,
need
to
work
and
then
come
back.
B
B
If
all
it's
doing
is
creating
the
the
pipes
for
you
and
it's
just
doing
it
in
a
standardized
way,
then
that
would
make
sense
as
an
accelerator
in
io,
but
it
would
need
a
different
name.
Presumably
this
is
all
about
the
figuring
out
what
we
want
app
model
components
to
look
like.
I
was
trying
to
look
for
a
namespace
that
felt
related
to
that,
and
the
closest
I
got
was
you
know,
sta
thread
and
mta
thread
which
are
both
in
system,
which
I
don't
think
is
a
good
fit
for
this.
C
I
was
looking
through
some
of
the
discussion
just
now
and
I
I
can
understand
this
for
ui
based
applications
when
forms
wpf
and
so
on.
I
I
can't
say
I
can
think
offhand
of
any
examples
of
console
applications
that
have
a
single
instance
model
and
I
can't
really
think
of
services
that
have
that
model
either
and
the
reason.
B
C
C
C-Sharp
compiler
is
not
a
single
instance
app.
Is
it
like,
I
type
csc
at
the
command
line.
It
actually
launches
an
instance
of
csc.
B
They
definitely
have
the
weird
thing
with
the
server
mode,
which
is
why
you
have
to
when
you're
moving
to
a
new
checkpoint.net
runtime.
You
have
to
kill
all.net
processes
because
they've
done
the
they
spin
up
a
host,
and
then
they
essentially
just
marshal
the
arguments
into
that
host
and
then
it
that
way
it
didn't
need
to
spin
up
all
the
dlls
again.
A
D
B
So
it's
the
it's
the
notion
of
if
you
would
have
any
sort
of
console
process
that
is
like.
Oh,
if
I'm
the
first
runner,
then
I'll
also
be
the
work
executor.
Otherwise,
I
just
enqueue
things
to
the
executor
and
terminate
granted
it's
way
less
common
in
console
applications.
I'm
just
saying
it's
not
nonsense.
Yeah.
C
B
B
Yeah
I
mean
they
may
want
the
named
pipe
for
doing
something
like
you
know:
the
ntp
command
talking
to
ntpd,
but
yeah,
but
they're
not
really
they're,
probably
not
the
same
process.
They
might
be
the
same
process.
Unix
tools
really
like
switching
on
arg0,
but
the.
B
B
A
Yeah,
I
think
this
one
needs
more
love,
I
think
so
to
me.
There's
two
sides
to
it:
there's
the
yeah,
I'm
not
gung-ho
about
the
namespace
choice.
I
don't
think
it
belongs
in
system.windows
and
we
don't
really
have
a
an
app
model
namespace
today
there
was
another
idea
that
glenn
was
pushing
fairly
hard
at
some
point,
which
was
the
reuse.
The
host
builder
pattern
from
asp.net
across
all
other
app
models.
I
think
olia
was
recently
looking
into
this
for
wpf
and
wind
phone,
so
that
wouldn't
make
sense.
A
So
presumably
there
might
be
some
sort
of
app
model
namespace
that
we
open
up.
For
that.
I
highly
doubt
that
we
would
share
things
like
application,
because
you
know
they're
just
too
big
to
do
anything
with
them
and
they
will
never
go
away
in
wpf
and
winforms,
but
like
these
common
services,
like
you
know
single
instance
mode
or
you
know,
even
some
windows
registration,
you
know-
maybe
the
taskbar
tray
thing
that
we
have
they're
not
really
related
to
ui
choice,
right
they're,
just.
D
A
A
Just
based
on
what
what
chuck
basically
sketched
up
as
a
as
a
building
block
out
of
the
vb.dll
and
so
yep,
that
was
it.
A
A
Yeah
almost
all
would
be
very
heavy
apps,
probably
anyway.
So
let's
close
this
one
for
now
is
eric's
on
the
call.
F
D
F
D
F
B
Yeah
yeah,
I
see
the
difference
here
of
yes,
it's
this
is
you
and
pitter,
instead
of
independer,
so
right.
C
A
Okay,
so
I
have
another
question:
okay
for
primitive
types
that
have
compiler
support
for
operators.
What
happens
if
we
define
a
custom
operator?
E
I
believe
it
depends
on
the
exact
scenario,
and
neil
is
probably
the
person
to
talk
to
about
this.
Like.
A
C
B
A
A
Okay,
never
mind
so
because
my
suggestion
was
to
make
them
custom
operators
on
the
thing,
but
that
wouldn't
work
then
so
then
I
guess
you
have
to
rename
them
it's
the
only
other
option.
B
Yeah
yeah,
because
because
nuand
is
really-
and
this
is
what
we
just
did
with
the
string
length-
that's
really
size
t
or
rather,
if
you
have
size
t,
it's
really
do
it,
and
and
adding
offsets
in
c
makes
sense
with
the
size
of
so.
F
B
Yeah,
so
these
are
the
four
things:
you're
adding
all
exist
as
reft
source
and
signed
into
pitter,
byte
offset
and
element
offset,
and
now
we're
just
adding
the
unsigned
version,
but
using
the
the
new
type
alias
now
that
we
have
it.
A
B
A
C
That
makes
sense
to
me:
there's
we
we
honestly
should
be
directing
people
who
are
using
these
apis
toward
using
endnt
and
newent
wherever
practical.
So
I
I
don't
think
we
should
be
adding
in
the
future
into
our
uint
paste,
overloads
and
uint
flows.
E
E
E
E
E
B
A
B
B
Right,
but
I
mean,
like
part
of
the
question,
would
be
like:
what's
what's
the
kind
of
thing
that
you're
writing
when
you're
doing
it
and
how
much
other
work
like
that,
are
you
already
doing
so?
What's
the
is
it
really
buying
anything
or
is
it
just
one
step
in
the
process
and
if
it's
just
one
step
in
the
process,
then
maybe
it
needs
a
more
holistic
proposal.
B
C
So
we've
we've
actually
had
some
some
prototypes
internally
of
working
ref
types
with
overloaded
operators,
but
legit
kind
of
falls
over
when
it
sees
them,
which
is
unfortunate.
Well,.
E
And
I've
got
a
proposal-
that's
half
baked
right
now
on
exposing
a
new
type
called
pointer
of
t
in
system
that
operates
essentially
like
null
of
t
in
that
it
would
have
special
language
and
runtime
support,
and
so
you
could
then
use
pointer
types
in
generics.
C
A
A
A
One:
it's
three,
seven,
five,
three,
nine
three,
seven,
five,
three:
nine
and
ability
to
perform
zero.
But
it
reads
in
stream:
pipe
reader
will
wait
for.
B
C
A
C
Yeah
so
you're
right
there
is
a
read
byte
api
on
stream
that
returns
negative
one
when
the
end
of
file
is
reached.
B
Okay,
so
this
would
have
to
do
all
sorts
of
crazy
things
to
like
using
read
byte
instead
of
using
read
to.
B
C
B
Like
it
seems
like
it,
I
just
don't
even
know
how
this
would
work
is
really
the
problem
you.
I
don't
know
that
we
can
add
the
api
and,
if
the,
if
we're
adding
an
api
that
is
changing,
visible
behavior,
which
it
obviously
is,
then,
presumably
we
need
to
add
it
in
the
in
or
then
we
need
to
make
the
decision.
Are
we
adding
it
as
compatibility,
or
we
think
that
the
new
feature
is
what
people
want
by
default
and
we
rarely
make
that
decision.
B
Though
so,
if
you
read
stream.read,
what
we
put
on
msdn
total
number
of
bytes
read
into
the
buffer
can
be
less
than
the
number
of
bytes
allocated.
If
that
many
bytes
are
not
currently
available
or
zero.
If
the
end
of
the
stream
has
been
reached
so
in.net
one,
we
should
have
said
negative
one
if
the
end
of
the
stream
has
been
reached.
But
we
didn't,
we
said
zero,
and
that
means
you
can't
ever
return
zero
in
the
middle.
A
B
I
mean
if
you're
using
a
memory,
mapped
file
for
communication.
You
have
to
invent
your
protocol
of
how
you
know
where
to
look
so.
Presumably
that's
you
write
a
length
and
set
up
your
I'm
ready
for
you
to
read
bit
somewhere.
B
B
Essentially,
what
you're
supposed
to
write
for
stream
is
a
while
loop
and
you
say,
while
current
red
equal
read
where
you're
writing,
comma,
the
offset
that
we've
already
read
so
far
is
not
equal
to
zero,
because
once
it
once
the
local
read,
returns
zero.
That's
telling
you
there's
no
more
data
and
you
move
on
because
we
don't
expose
a
is
eof
or
or
anything
like
that,
and
we
we
a
long
time
ago,
we
buried
it
all
in
read
returning
zero
and
so
again
that
that's
just
the
I
don't
understand
what
this
would
do.
C
At
the
base,
so
by
the
way
like
I'm
following
through
the
various
links
in
this
issue,
sslstream.read
does
actually
return
zero.
Today
to
mean
I
sorry
ssl
stream.read,
you
can
pass
in
a
zero
length
buffer
and
it
will.
It
will
return
an
incomplete
task
until
data
is
received
and
then
it
returns
zero
from
read.
But
I
believe,
that's
only
if
you
pass
in
a
zero
length
buffer,
which
makes
sense,
I
guess-
and
maybe.
B
C
A
C
C
So
the
the
method,
the
method
blocks
or
returns,
an
incomplete
task.
If
data
is
not
available
as
soon
as
data
becomes
available
or
the
end
of
the
file
is
reached,
it
either
returns
xero
had
you
called
the
synchronous
api
or
completes
the
task?
Had
you
called
the
asynchronous
api.
B
C
Yeah
this,
so
one
interesting
thing
here
is
that
this
is
actually
kind
of
an
ambiguous
scenario
in
stream.read,
because
I
don't
think
anyone
ever
here
we
go
to
do
the
implement.
So,
according
to
msdn,
the
implementation
will
block
until
at
least
one
byte
of
data
can
be
read
in
the
event
that
no
data
is
available.
Read
return
zero
only
when
there
is
no
more
data
in
the
stream
and
no
more
is
expected,
such
as
a
closed
socket
or
an
end
of
file.
C
So
it
looks
like
from
from
reading
msdn
here
it
sounds
like
passing.
A
zero
length
destination
buffer
was
supposed
to
be
considered
an
error
because
otherwise
there's
no
way
to
fulfill
the
contract
of
the
api
correctly,
but
it
sounds
like
certain
types
have
already
violated
that
contract.
So
now
we're
just
asking
for
more
types
to
do
so.
C
Yeah
but
again
like
we're,
reading
msdn
very
strictly,
like
msdn
very
clearly,
says:
zero
means,
there's
no
more
data
in
the
stream
and
you've
reached
the
end
of
the
file
period,
which
the
implication
of
that,
even
though
it's
not
explicitly
said
is.
It
is
invalid
to
pass
a
zero
length
array
as
a
destination
into
the
read
apis.
B
B
It's
also
unclear
what
happens
if
you
I
mean,
I
mean
it.
Is
it's
clear
that
you're
you're
violating
the
intent
of
read
if
you
pass
zero
as
the
count
that
it's
allowed
to
read.
F
C
B
B
Presumably,
it's
then
the
feature
work
of
all
the
follow-up
of
say
that
this
is
the
new
behavior
of
that
somehow
we're
getting
this
feature
into
stream,
and
I
don't
even
know
what
it
means,
because
you
can't
you
can't
really
tell
the
difference
between
it,
returned
0
for
end
of
stream
and
you
return
0,
for
I
don't
have
data
yet
followed
by.
Oh,
I
have
data
like
you.
B
Can
you
can
accidentally
depend
on
the
task,
not
completing
and
then
eventually
completing
to
mean
there's
either
data
or
someone
close,
the
stream
now
yeah
that
it
has
gone
into
a
terminal
state
and
you
got
zero
because
you
told
it
zero
is
the
biggest
number
it's
allowed
to
write,
but
they
could
have
easily
checked
that
in
a
precondition
and
returned
early,
and
that
totally
seems
like
a
thing
that
could
easily
happen,
and
it's
stream
really
isn't
set
up
for
this,
and
I
think
we
need
to
do
something
different
on
stream
in
order
to
make
it
work.
B
Detail
that
violates
what
the
abstraction
was
documented
as
having
as
a
behavior.
You
know
I
mean
again
the
the
joy
of
virtual
everything
is.
You
can
write
down
what
it's
supposed
to
do,
but
you
can't
write
down
what
it
actually
does,
which
is
my
template
method
pattern
is
better
you
can
you
can
limit
your
derived
types
into
what
they
can
do.
C
I
think
I
think
there
are
workarounds
that
asp.net
can
do
at
the
moment
in
order
to
make
this
work
better.
A
C
C
Makes
sense,
like
the
scenario
makes
sense
what,
basically,
what
they're
doing
is
they?
They
have
a
bunch
of
idle
connections
and
the
only
way
to
pull
those.
The
only
way
to
ask
do
these
connections
have
any
data
that
I
should
be
reading
is
to
call
their
read
apis,
but
read
requires
you
to
pass
a
buffer
in
so
at
minimum.
What
they
would
have
to
do
is
they
would
have
to
pass
a
one
byte
buffer
into
every
single
one
of
these
read
apis,
while
they're
just
waiting
for
all
the
tasks.
C
To
finish
so
now,
if
they're
waiting
for
twenty
thousand
concurrent
reads,
most
of
which
are
just
idle,
they
have
twenty
thousand
one
element,
byte
arrays,
just
kind
of
sitting
around
doing
nothing
and
what
they
want
is.
Can
we
eliminate
allocating
those
one
byte
arrays
and
just
pass
in
a
shared
singleton
empty
buffer
into
all
of
these.
C
A
C
Yes,
but
you
had
intended
to
read
you,
you
had
intended
to
read
the
data.
Anyway,
you
were
just
waiting
for
data
to
arrive.
B
I
mean
I
I'm
guessing.
The
scenario
here
is
that
the
stream
type
reader
shouldn't
create
its
4k
buffer
if
the
stream
that
it's
wrapping
is
already
closed,
like,
I
think,
that's
sort
of
the
question
it's
trying
to
answer
and
that
if
you
or
if
you
never
ended
up
calling
read
after
data
was
available,
then
why
did
we
need
the
buffer
but
yeah?
If
this
is
depending
on
an
implementation,
behavior
of
stream?
We
need
to
do
that
first
and
then,
why
would
you
ever
set
this
to
the
other
behavior
like?
C
Well,
normally,
the
way
that
you
do
this
if
you
were
following
like
proper
handle
handling,
is
you
would
gather
the
handles
of
all
the
streams
that
you
care
about,
and
then
you
would
pass
every
single
one
of
those
handles
into
a
call
to
wait
for
any
and
then
wait
for
any
would
tell
you
what
stream
actually
pinged
and
has
data
for
you.
I
mean
it's
basically,
a
poll
routine,
just
asynchronous.
A
A
B
A
C
B
A
Well,
either
that
or
you
do
the
trick
divided
in
other
places
right,
because
realistically
these
things
only
have
to
be
supported
at
the
bottom.
Everybody
in
the
middle
basically
gets
to
just
dispatch
the
underlying
stream-
and
this
goes
back
to
this
whole
like
we
need
a
delegating
stream
realistically.
So
we
can
do
these
kind
of
forwards
right,
because
realistically
yeah
the
thing
in
the
middle
doesn't
care.
The
thing
in
the
middle
just
says:
well,
does
the
thing
beneath
me
have
any
data
right,
and
so
we
wouldn't
have
to
actually
plumb
it
all
the
way
through.
B
It
has
data
every
wrapper
stream
on
another
stream
has
data
if
it
has
locally
buffered
something
or
its
underlying
stream
has
data,
so
there's
actually
probably
work
at
every
single
level,
because
I
very
few
stream
implementations
don't
buffer
their
data
locally
yeah.
A
B
Right
but
yeah,
I
think
here
it
sounds
like
what
we're
really
trying
to
do
is
change
stream.
So
so
this
needs
a
thing
other
than
the
triage
review.
B
Do
you
promise
that
the
stream
you're
having
me
wrap
is
going
to
do
that
thing?
If
so
I'll
depend
on
it?
Otherwise,
the
exceptions
on
you
for
passing
me
the
wrong
value,
which
effectively
means
that
this
is
a
throw
on
failure,
api,
which
we
have
guidance.
That
says,
don't
expose
a
boolean
for
throw
on
failure,
yeah,
so
yeah.
This
just
needs
more.
A
Somebody
in
chat
is
asking:
why
don't
we
invite
david?
I
mean
like
we.
We
usually
do
this.
If
the
issue
up
front
is
big
enough,
we
usually
don't
invite
people
up
front,
because
we
can't
tell
them
when
their
issue
pops
up
in
a
two-hour
review
slot,
and
so
it's
just
wasting
people's
time,
but
this
one
seems
like
we
should
have
a
dedicated
meeting
on
for
an
hour
and
talk
with
him
about
it.
B
A
Yeah,
I
mean
I
mean
we
own
stream
right,
so
it
doesn't
seem
unreasonable
to
actually
have
a
design
discussion
of
whether
this
should
actually
be
part
of
stream
right.
It
would
not
be
it
wouldn't
be
reviewed
more
like
alchemy
design,
meaning
yeah
and
then
the
other
issue
that
we
still
have
is
this
one
here,
which
is
also
pretty
largish.
I
just
read
the
comments.
A
Basically,
what
it's
proposing
is
making
some
things
public,
that's
not
the
bad
part.
The
more
problematic
part
is
this
enum
here,
so
it
basically
gives
you
great
detail
about
what
happened
in
the
parser,
and
that
seems
reasonable
at
the
surface,
but
it's
dubious
whether
that
actually
helps
and
so
there's
kind
of
different
cam.
A
So
you
know
cyrus
was
looped
in
because
he
actually
did
the
parsing
for
regex
in
roslin
to
basically
allow
you
to
colorize
regexes
and
also
give
a
bit
more
of
an
experience
in
visual
studio
with
code
completion
when
it
comes
to
actual
authoring
regexes,
and
he
basically
said
this
data-
he
wouldn't
be
sufficient
to
do
that.
A
So
the
question
is
really:
who
would
be
consuming
these
error
codes
to
do
anything
meaningful
with
them
because
it
wasn't
supposed
it
wasn't
sufficient
to
do
colorization
or
parsing
or
anything
useful
in
an
ide
which
these
things
you
know
would
only
be
useful
for
right.
Normally,
you
don't
care
about
why
the
regex
wasn't
developed
unless
you
provide
the
tooling
for
that.
A
A
A
A
B
A
I
mean
for
parcels,
it's
I
mean
for
lectures.
Yes,
you
get
away
with
them,
but
for
parsers
it's
hard
and
regex
are
logically
trees
right
because
you
have
like
you
know,
parentheses
and
groupings
and
stuff
so
like
giving
you.
The
actual
position,
isn't
necessarily
useful
right.
If
you
have
a
missing
closing
paren,
it's
always
the
length
of
the
string
right,
because
that's
when
you
find
out
that
you
have
a
missing
closing
plan
so
like
that's
not
like,
I'm
I'm
skeptical
that
this
was
all
that
would
help
as
well.
A
It
probably
would
help
in
some
cases,
but
what
I
like
about
this
proposal
is
it's
just
giving
you
an
int
right
versus
this.
One
gives
you,
like,
I
don't
know
30
euro
members,
which
seems
a
little
bit
over
the
top,
so
yeah,
like
I
think,
who's
on
the
hook
for
this
prashant,
I
think,
is
on
the
hook
for
this
yeah.
So
like
yeah.
We
will
invite
him
next
week
and
talk
about
that
as
well.
A
B
C
B
B
I
was
trying
to
to
change
the
habanzig.
B
Imperative
action
and
realize
they
don't
know
enough
german
grammar,
so
you
should
watch
the
producers
because
haven't
seen.
Dostocho
band
is
awesome.
B
In
the
movie,
the
producers
and-
and
I
think,
you'll
like
the
newer
one
with-
will
ferrell
and
matthew
broderick,
the
they're
holding
an
audition
and
someone
auditions
with
the
song.
That's
made
up
for
the
the
play
and
the
movie
and
whatever
habenziga
hort
dusto
joband.
It's
actually
a
really
cute
song.
I
think
it
plays
bavarian
music
very
well.
He
likes
the
oompa
and
it's
good
dancing
and
will
ferrell's
fun
so.
B
C
Paycheck
back
in
is,
he
played.
B
I
mean
yeah,
you
don't
want.
I
mean
the
67
one's
good
too,
but
the
new,
the
remake
is
it's
version
of
haban
zika
horta,
japan,
it's
either
new
for
that
one
or
or
that's
the
better
one.
Oh.
D
B
A
couple
of
the
songs
were
written
when
he
made
it
a
stage
play
from
the
version,
and
I
think
that
this
is
one
of
the
new
ones,
so
they
are
different.
If
you
see
the
broadway,
if
anybody
does
the
broadway
showing,
then
the
2000,
whatever
movie
is
almost
it's
pretty.
D
B
Stage
setting
because
I've
also
seen
it
on
stage-
and
it
was
just
that
movie
but
without
being
able
to
change
scenery
as
much.
B
Yeah,
you
should
you,
should
totally
watch
it
and
I
think,
as
a
german
you'll
get
an
extra
kick
out
of
it
sure
it
makes
fun
of
the
most
famous
german
in
very
fun
and
and
and
laughable
ways.
B
C
B
C
He
and
his
son
max
brooks
actually
did
a
psa
at
the
start
of
the
pandemic,
and
max
works
was
standing
outside
of
his
father's
house.
Basically
saying
like
hey
stay
home,
you
know
help
keep
help.
Keep
our
comic
talent,
like
my
dad
alive.