►
From YouTube: .NET Design Reviews: GitHub Triage
Description
00:00:00 - Approved: Rename MemoryHandle.PinnedPointer to Pointer https://github.com/dotnet/corefx/issues/24426#issuecomment-337305000
00:18:08 - Needs Work: Disposables in CoreFx: SerialDisposable, CompositeDisposable, etc https://github.com/dotnet/corefx/issues/14891
00:40:23 - Approved: Add DateTime.UnixEpoch and DateTimeOffset.UnixEpoch fields https://github.com/dotnet/corefx/issues/24449#issuecomment-337315763
00:51:47 - Approved: [ExcludeFromCodeCoverageAttribute] should be applicable to assemblies https://github.com/dotnet/corefx/issues/24694
01:49:06 - Approved: Add Base64 conversion APIs that support UTF-8 for Span(T) https://github.com/dotnet/corefx/issues/24568#issuecomment-337334707
C
C
E
C
A
D
D
C
K
A
We
can't
I
mean,
like
the
only
reason
we
should
probably
bring
in
awesome,
because
I
think
he
only
comes
at
11:00.
So,
okay
thinking
earlier
that
we
can
okay,
this
one
I
did
I
just
didn't
do
all
right.
So
then,
so
we
just
go
with
this
one
then,
before
we
I
don't
have
strong
opinions,
because,
as
already
on
the
screen
that
I.
L
H
E
M
G
H
It
just
unlisted:
do
you
have
any.
D
In
PR
there
will
be
net
new
types
right:
it's
it
still
not
closed,
whether
they
will
map
to
this
type
or
not.
We
haven't
made
that
decision,
but
there
are
compatibility
problems
if
we
want
to.
If
we
would
want
to
map
the
new
language
teacher
stood
to
this
type,
but
regardless
whether
you're
posting
it,
whether
the
language
teacher
will
be
mapped
to
this
new
type
or
not,
it
would
be
I
think
good
to
expand.
What
in
pointer
has
in
terms
of
API
surface
plus.
D
D
There
were
people
who
wanted
to
keep
in
pointer,
you
invite
them
to
be
super
super
simple,
and
then
there
were
people
who
wanted
to
have
more
and
more
operations,
and
we
kind
of
ended
up
over
years
in
the
middle
and
I.
Think
the
Battle
of
having
in
pointer
being
super
super
super
simple,
is
already
lost,
so
I
think
it
just
makes
sense
to
you
know.
Fill
in
the
API
is
that
people
want
to
use
is
the.
D
This
is
the
this
is
the
native
in
discussion
that
we
have.
If
we,
if
we
decide
to
map,
then
language
features
to
this
type.
Then
yes,
but
it's
there
are
so
many
issues,
for
example
like
language
native
int,
the
types
you
would
like
them
to
other
checked
unchecked
blocks
in
in
the
language.
We
cannot
do
it
for
in
pointer
because
it
would
be
a
breaking
change.
For
example,
there's.
P
D
Are
some
ideas
how
we
can
I
call
it
hack
it
such
that
it
kind
of
works,
but
they
all
kind
of
solutions?
So,
as
I
said
to
summarize,
it's
not
yet
clear
whether
the
language
teachers
will
map
to
this
type
or
not,
but
I
think
it
just
makes
sense
to
add
API.
Smack
people
have
been
asking
for
these
api's
for
when.
D
So
many
people
said
that
if
we
don't
map
the
language
features
to
these
types,
they
will
keep
using
these
types
anyway,
because
they
have
too
much
code
invested,
and
you
know
using
in
pointers,
hearing
pointers.
So
they
said,
even
if
you
don't
provide
the
language
mapping
to
these
types,
we
still
want.
Additional
API
is
because
we
we
need
them
and
we're
not
gonna
move
away
from
each
pointer.
D
D
B
P
B
M
O
B
P
E
E
B
D
A
D
A
See
the
question
is,
like
you
know,
what
do
we
do
of?
This
means
that,
even
if
you
were
to
add
operator,
then
if
the
language
starts
to
actually
map
to
these
types
in
the
corner
ourselves
again
into
weird
position
right
now,
the
compiler
has
to
decide
to
it
called
operators
or
do
I
will
treat
them
as
intrinsic
sign,
and
we
don't
want
to
do
that,
which
kind.
Q
B
D
H
D
Q
A
C
N
D
U
R
D
D
B
T
B
Q
C
E
U
T
Yeah,
so
we
added,
we
already
agreed
to
add
two
new
methods
to
sock
it
or
extension
methods
on
socket
for
reading
and
writing
really
sync
right,
a
sync
with
memory
and
we
added
them
to
the
socket
task.
Extensions
class,
because
that's
where
the
existing
task
based
methods
for
socket
are
I,
think
you
know
when
this
was
originally
added.
T
T
You
know
if
someone
was
passing
an
array
segment
and
socket
flags,
whatever
the
compiler
will
say.
Oh
I
I
can't
I
can't
use
that.
With
this
new
overload
and
it'll
fail
to
compile,
which
means
we
have
two
options:
either.
We,
we
add
them
to
socket
ass
extensions
as
we
originally
decided,
or
we
add
them
to
the
socket
class,
as
instance
methods
and
we
had
instance
based
methods
of
the
existing
extension
methods
to
the
socket
class
as
well.
X
T
No
just
because
we
haven't,
we
haven't
decided
to
add
meth
memory
based
over
those
four
that
we
could.
We
just
haven't
yet
and
the
same
percent
to
a
sink.
So
there's
a
question.
You
know
we
could.
We
could
just
define
the
new
methods,
as
they're
currently
checked
in
there
on
socket,
ask
extensions.
We
could
just
leave
it
like
that
or
we
could
choose
to
move
those
to
socket,
as
instance,
methods
and
also
add
the
ones
that
could
result
in
conflicts
to
socket,
as
instance,
methods
at
which
point
the
socket
task.
T
C
T
P
T
B
B
K
T
C
A
A
T
A
C
A
A
Into
special
I,
don't
think,
and
we
can
try
but
I'm
pretty
sure
the
compiler
results
over
those
between
the
only
problem
is.
You
have
so
many
issues
where
you
know
when
you
only
consider
instance
method,
it
finds
a
unique
match.
But
if
you
find,
if
you
consider
overloads
from
extension
methods-
and
you
would
no
longer
find
a
unique
one-
then
they
discarded
it
volatile
it.
A
So
the
only
considerate
of
a
name
and
they
physically
cannot
resolve
it.
So,
in
other
words,
let's
say
I
have
an
overload
that
takes
an
int
on
the
instance
and
then
you
also
add
extension
methods
and
you
say
I
add
one
that
takes
a
long.
Let's
say
sorry,
I'm
sorry
the
instance
one
has
it
long,
I
think
sense,
but
it
takes
an
Internet
passing
in
it.
A
It's
the
extension
method
will
be
the
better
fit,
but
that
don't
consider
the
extension
methods,
because
this
one
already
applied
I
think
that's
true
I
think
they
only
consider
overloads
on
the
extension
time
if
they
cannot
find
a
match
on
the
instance
method
at
all.
But
it's
not
just
by
name
it's
given
the
types
so.
P
G
A
D
D
A
E
D
D
A
D
D
D
B
B
B
A
E
C
G
C
A
U
A
C
E
C
You
created,
or
we
created
the
guidance
really
yeah
I.
Remember
that
it
was
like
at
some
point
like
we
said
like:
oh,
we
can
add
new
extension
methods
and
naught
and
encore
types
he
said
like.
Oh,
we
don't
want
to
just
blow
to
the
number
of
number
of
methods
on
that
and
the
type
of
stuff
doesn't
apply
here,
but
Sony.
D
You
might
ask
like
what
will
people
see
as
a
benefit
of
moving
them?
I
give
this.
You
know
perf
improvement,
or
maybe
they've
heard
from
customers
that
there's
some
confusion
about
why
these
methods
are
extensions.
People
cannot
find
them
whatever.
There
is
on
this
and
we
should
probably
move
if
there's
some
reason
to.
Z
E
T
H
H
B
B
L
A
X
D
B
It's
not
that
it
does
different
thing
like
well,
it's
it's
on.
We
have
to
file
time,
which
is
the
windows
value.
We
have
that
on
date
time,
but
the
to
you
next
time
is
only
update
time
offset
date.
Time
is
terrible
because
it
has
the
I,
don't
know
if
I'm
local
time
or
UTC
time,
and
then
it
always
assumes.
Whichever
thing
you
don't
want
it
to
do,
and
big
time
offset
removes
that
ambiguity
by
always
knowing
what
time
zone,
but.
D
B
D
A
I
think
it's
fine
to
expose
yeah
I
have
to
say,
like
this
whole
idea
of
like
not
giving
it
on
day
time,
because
when
people
to
push
to
date
and
offset
I'm
thinking,
that's
not
yes,
fair
is
a
question.
I
think
the
better
way
to
do
it
is
expose
the
API
and
then,
for
example,
have
an
analyzer
that
detects
that,
because
you're
using
day
time,
don't
you
know,
UNIX
epoch
probably
want
to
look
at
data
officer,
because,
whatever
you
know,
you're
doing
you're,
probably
assuming
the
wrong
things
anyway.
A
D
Because
I
understand,
when
you
called
a
time
that
UNIX
epoch,
it
will
give
you
a
UTC
instance,
basically
date
and
kind
I
don't
see
but
little
it
drops.
But
it's
probably
UTC
it's
it's
a
perfect.
It's
actually
what
the
epoch!
It
is.
It's
even
better
than
representation.
Then
they
time
offset
frankly,
because
the
offset
doesn't
make
too
much
sense
for
UTC
plate
times.
Well,.
B
B
It
doesn't
make
sense,
it's
just
a
wasted
space.
The
work
I
mean
the
problem
with
date.
Time
is,
if
you
take
a
day
time
and
you
subtract
UNIX
epoch
if
you're
a
local
if
the
date
time
that
is
the
left
hand,
side
of
the
expression,
you
is
the
unspecified
day
time
for
your
answer,
makes
no
sense,
but
we.
D
H
A
A
O
O
B
V
C
Api
prove,
and
so
let
me
let
me
ask
one
more
question,
so
what
bothers
me
a
little
bit
on
the
data
offset,
we
have
the
two
UNIX
thangai,
which
seems
to
be
at
least
the
one
of
the
major
like
you
know,
reasons
for
that.
Would
it
make
sense
to
add
instead
of
adding
the
UNIX
epoch
and
exposing
that
and
to
UNIX
matter
like?
Would
it
satisfy
all
of
the
requests
or
is
there
another
scenario
where
the
UNIX
epoch
is
really
interesting?
B
O
D
O
A
Again,
I'm
someone
hesitant
to
like
try
to
guide
people
by
omission
of
API
and
the
reason
it
doesn't
work
is
that
if
you
can
find
the
API
I
think
you're,
not
necessarily
likely
to
read
up
on
hold
the
whole
thing
works
right.
You
just
do
some
other
crazy,
hackneyed
new
daytime
offset
your
thing
there
and
then
you
save
time.
A
B
A
Work
is
fine,
but
it's
very
easy
to
detect
a
second
Alice's
that
you
they
typed
up
null
daytime,
look
UNIX
epoch
and
let
me
say
you
know
what
you
should
call
the
method
of
passing
arguments
regards
to
me
and
it
seems
like
a
better
fix
into
all
of
our
way.
To
not
give
people.
Api
is
just
because
their
name
is
used
because
then
like.
If
there
are
valid
ways
you
can
use
the
API,
then
the
question
is:
what
way
do
we
add
by
not
giving
you
the
API?
It
seems
backwards.
C
R
B
F
D
B
D
D
B
W
AA
T
Partially
wrong
or
I
was
wrong
about
my
explanation.
The
issue
is
exactly
that.
The
compiler
does
find
the
conversion
from
array
segment
to
memory,
so
it
binds
more
strongly
to
the
new
instance
method,
then,
to
the
old
extension
method,
which
would
be
fine,
except
that
they
have
different
return
types
and
so
any
code
that
was
performing
operations
on
the
return
of
the
operation
rather
than
just
awaiting
it
will
break.
A
Is
like
you
should
avoid
a
situation
where
we
have
to
do
this
on
an
ongoing
basis
to
basically
demand
that
the
compiler?
That's
all
the
solution
between
new
and
old
and
then
just
have
them
either
consistently
on
the
extension
class
or
consistently
on
this
since
type,
so
that
it's
easier
for
us
to
reason
about
overload
to
me
at
them.
T
T
A
A
AC
Look
at
this
before
right,
so
what
has
changed
since
the
last
time
we
looked
at
this
I'm
a
couple
of
things.
First,
we
don't
have
them
as
encapsulated
into
decoder,
encoder
classes,
nextrip
types
they
are
just
flat.
Api's
second,
is
we
added
renamed
with
some
of
the
arguments
based
on
some
feedback
and
method
names
and
added
a
s,
file,
block
optional
argument
or.
D
D
Transformer
right
now,
so
there
was
basically
base64
outer
tie
and
then
two
nested,
subtypes
or
utf-16
and
utf-32
scan
last
week.
This
whole
thing
with
nested
types
really
doesn't
work
super
well
in
our
ecosystem.
That
is
not
ready
for
this,
like
we
didn't,
have
many
nested
types
in
the
framework,
so
our
intelligence
doesn't
work.
Refactoring
like
I
now
concluded
we
shouldn't
be
doing
nested
types
or
shall
we
fix
it
Tony
or
maybe
fix
the
Tony,
but
fixing
tuning
is
like
not
a
super
practical
solution,
so
we
are
nested
and
that's
why
they're?
D
O
O
D
D
O
AB
D
O
D
A
Don't
know
I
think,
logically,
in
generally,
when
you
say
encode
and
decode,
you
expected
to
be
encode
ins
into
whatever.
That
format
is,
which
means,
if
I
say
encode
for
basics
before
expect
to
get
a
basic
support
string
back,
because
that's
the
total
building
code
and
that
context
means.
But
if
I
say
m
for
encoding,
I
expect
to
get
an
impact
back
right,
I
don't
get
expensive
normal
I,
see
you
falling
back,
I
mean
that's
I,
think
jelly
ho
ho
and
Cody
cook
and
I
mean
different
name.
A
O
O
D
D
W
N
M
H
O
O
O
O
AB
D
D
D
D
D
A
I
have
to
say
I
kind
of
prefer
that,
because
it
it
names
what
the
thing
does
the
reverse
today,
this
one
is
not
too
bad
because
ECF,
encode
and
decode
as
the
primary
operations,
and
it
gives
you
a
vocabulary,
but
the
other
one
was.
Actually.
You
have
a
bunch
of
methods,
and
there
is
nothing
you
can
say
across
the
salt
type
that
will
allow
you
to
like
four
things
or
giving
you
a
sort
of
like
keyboard
based
vocabulary.
But
now
we
have
a
district
here
is
encode
and
decode.
Is
we
actually
make
this?
D
O
O
A
D
A
I
think
that
is
something
that
we
just
have
to
think
about
in
general,
because
it's
kind
of
a
can
have
two
concepts.
Coding
coding
with
very
different
API
shapes
I
mean
that
something
ages,
I,
don't
know
how
you
would
call
this
thing
like.
If
you
don't
call
it,
we
said
to
call
it
something.
Just
because
you
don't
name
the
API
I
mean
there's
no
name
for
the
console.
D
M
M
D
D
A
A
D
Also,
we
have
a
hierarchy
of
the
encoding
like
there's
a
base,
encoding
type
atom,
base64,
encoding
kind
of
implies
that
it
inherits
from
the
coding
base
line.
It's
just
a
64th
transforms
like
yeah.
We
had
some
suffix
or
no,
you
know
like
it
doesn't
add
value.
This
is
actually
why
I
called
the
basics
before
without
any
suffix,
because
everything
else
after.
A
A
D
Q
P
A
G
A
D
O
A
E
A
D
AB
A
A
X
A
AC
AC
A
Of
this
realm
and
someone
you
want
to
tell
them
how
to
use
the
api's.
So
what
kind
of
things
do
you
give
them
so
they're
successful?
So,
for
example,
if
you
point
them
to
an
MSDN
article
that
says,
here's
how
to
do
buff
operations
right,
if
you
give
this
whole
thing
a
name,
and
you
have
types
that
reflect
that
name,
people
can
find
stuff.
They
can
learn
a
technology
like
that.
A
AC
O
M
H
M
O
D
Q
A
O
Levi
about
the
operation
cells,
so
I
wanted
a
trustee
in
place
in
u.s..
First,
because
I,
don't
think
operations
has
made.
Senses
returned
enum,
promos
api's,
because
if
the
in
place
really
open
can
be
done
in
one
shot,
it
doesn't
really
make
sense
to
do
it
any
other
way
and
as
a
result
and
the
reason
that
I
say
that
is
because
you're
generally
going
to
be
expanding
or
contracting
the
buffer,
which
means
that,
if
you,
if
you
right
now
with
data,
basically
there
are
some
lights
and
the
other
number
that
you
have
an
encoded.
O
Well,
you've
now
overwritten
up
with
the
output,
so
that
data
is
now
perfectly
lost
in
power.
He
can
never
hold
nothing
in
so
needs.
More
data
is
kind
of
a
medium.
A
status
as
far
as
it
plays
encoding
knows
now.
Also
consider
needs
more
room
well
again,
if,
if
you're
doing
an
operation
within
a
single
buffer-
but
you
really
do
need
to
check
ahead
of
time
to
the
buffer-
is
large
enough
to
contain
the
entire
result,
because
again,
there's
no
way
for
the
color
to
restart
any
heavenly
procedure.
O
O
That
you're
calling
this
in
a
loop
and
you're
reading
data
from
from
a
network
of
ur,
and
you
get
a
55
bytes
of
data
from
your
from
your
incoming
network
stream,
like
that's,
not
evenly
divisible
by
3
in
stock.
So
how
do
you
like?
Are
you
supposed
to
add
padding
to
that?
Or
are
you
supposed
to
return
back
to
the
colony
for
data?
No,
so.
O
P
AC
D
O
You
can
call
another
people.
The
idea
is,
if
you
have,
if
you
have
a
generic
transform,
API
said
with
this
word
interface,
you
would
probably
want
to
call
it
this
final
part
just
because
you
don't
know
what
it's
going
to
do
with
that
information
may
be
able
at
having
nav
opposed,
but
it's
another
person.
Importantly.
V
D
AB
O
R
O
A
A
Am
I
only
concerned
that
I
guess
I
have
and
I
don't
know
how
to
solve
it?
It's
like
this
whole
idea
of.
Like
you
know,
if
you
happen
to
be
in
cou
boundaries,
you
can
return
the
bytearray
to
the
caller
and
everything
is
fine.
You
only
get
the
case
when
you
end
up
in
the
you
know.
I
still
need
one
stuff,
but
I
guess.
If.
E
P
D
Why
did
we
return?
Operation
starts
from
the
in
place
because
I'm
pretty
sure
that
they
used
to
be
like
not
return.
You.
Q
W
Q
D
G
G
D
D
Need
to
be
really
fascinating.
Well,
let
me
so.
Basically,
we
showed
to
the
to
this
API
review
committee,
the
loop
that
people
need
to
write
and
all
the
checks
and
everybody
in
the
room
said
what
the
heck
is
this.
This
is
too
complicated.
Can
we
just
return
an
enum
that
tells
you
exactly
like
what
is
the
state.
O
P
H
A
Yeah
grew
press.
The
last
time
look
at
the
at
the
generic
it
was
like
doing.
The
check
seems
like
if
you
didn't
write
a
comment
above
the
check.
It's
not
easy
to
tell
by
the
check.
D
It
was
exactly
because
it
it's
an
enum,
because
now
it
returns
information
about
the
things
that
otherwise
you
would
have
to
compute
from
the
returned
ends
like
this.
If
we
ate
at
boo
right
Monaco,
try
yeah,
it
was
try.
So
this
is
one
condition,
and
then
you
had
to
like
look
at
the
ends
and
see
whether
this
input
buffer
length
is
the
same
as
the
you
know,
number
of
consumes
bytes.
Then
it
means
that
it
consumed
everything.
D
O
O
AC
D
D
O
The
enum
value,
but
I,
don't
think
it's
required,
because
what
you
could
have
done
is
before
would
call
a
transformation
without
execute,
like
query:
how
much
output
and
query
the
size
of
the
output
buffer
that
we
need,
because
we
have
api's
to
get
the
size
and
then
just
make
sure
that
your
output
has
that
available
and
I
call
transform,
execute.
And
if
it's
not
enough,
just
go
to
properly.
But.
D
O
O
AC
We
can
be
again
pinning
this
discussion
because
it
is
medicine
pool
or
you
know
that
could
be
separate,
but
when
we
push
on
the
other
questions
ahead
well,.
A
I
mean
cam
and
it
seemed
fun.
Take
decisions
I'm,
not
really
nervous,
because,
let
me
just
end
up
with,
like
you
know,
we
do
the
same
loop
you
just
described.
He
goes
to
the
back
to
the
top
of
the
list
and
then
three
weeks
ago,
they're
in
the
same
discussion
again
I
need
to
close
the
lease
in
a
few
things
or
at
least
give
you
a
homework
that
is
years
secretive.
A
Stick
it
there
we
go
so.
The
other
thing
is
that
we
said
there
is
the
the
general
shape
of
the
EP
is,
like
you
know,
in
culver
city
code
seemed
like
we
settled
on
that.
Then
there
was
the
discussion
on.
Do
we
even
at
all
for
the
in
place,
things
should
be
in
place
differently.
There's
the
third
one,
which
is
even
if
it's
not
in
place
like
what
does
EBI
shrimp
look
like,
and
it
seems
like
there
is
this
opinion
in
the
room,
maybe
that
we
don't
need
to
in
them
at
all.
A
We
can
just
simplify
the
patterns
with
the
pointing
people,
know
what
to
do,
which
seems
fine,
but
then
I
think
what
I
would
like
to
see
is
you
know
the
extracted
code
for
the
general
pattern,
not
just
a
specific
example
where
you
know
that
the
outfit
sizes
change
like
that,
but
it's
actually.
This
is
the.
This
is
like
the
copy
and
paste
review,
except
you
have
everybody
to
ride,
and
well
the
only
people.
D
P
D
M
D
O
Say
if
you
did
want
the
concept
of
operation
says
you
can't
always
get
it
before
the
dull
features
from
the
others
good.
So
the
caller
could
see
if
the
conversion
was
incomplete
by
looking
at
my
sister
device
written,
but
he
doesn't
know
conversion
using
leaks
because
he
ran
out
yet
the
data
or
happened
yeah.
This.
O
O
O
O
In
the
outer
loop
has
to
end
your
loop,
who
end
of
you
run
out
of
days
because
you're
reading
input
data
from
some
read-only
buddies
from
interest
rate
from
something
when
you're
out
of
data,
that
if
you
terminate
your
loop,
because
you
have
nothing
left
to
do
and
then
you
follow
into
another
loop,
Richards
I
might
have
a
buffer
of
remaining
bytes.
Let
me
pop
that,
through
the
trans
lock,
so
that
you
transform
at
that
point,
never
return
true
without
so.
Z
A
K
A
And
then
the
option
may
have
is
like
maybe
overthinking,
and
maybe
we
can
just
say
instead
of
having
you
know,
to
return
arguments
like
fights
written
bytes
met,
we
can
formalize
this
into
a
single
struck.
That
has
two,
ladies,
who
just
can't
Curie
properties
of
this
guy.
That
says,
am
I
done
yet
because
it
really
only
the
only
thing
you
have
to
do
is
compare
these
relative
to
each
other,
usually
compared
with
you
input
science
to
know
what.
A
Let's
see
if
we
actually
look
at
codons,
okay,
with
the
last
time
I
saw
it
looks
like
you
know,
this
is
less
than
or
equal
to
this
one
and
this
is
bigger
or
that
that
one
is
that
it
seems
like
these
are
things
that
you
can
easily
get
wrong
like?
Oh,
it's
repeat,
that's
them,
but
it's
like
that's,
not
equal
to
or
something
and
then
I
think
that
was
the
conclusion
as
well.
Instead,
given
that
stuff
operation,
you
kind
of
want
to
make
it
not.
D
Very
error-prone,
so
I
propose
that
basically
asan
goes
offline
and
kind
of
analyzes,
both
of
the
code,
samples
and
sense.
Something
saying
you
know
either
there's
a
bug
in
Levi's
code
or
maybe
doesn't
give
you
enough
information
for
some
scenarios
that
he
didn't.
You
know
consider
and
we
should
also
either
I've.
O
AC
Now
start
going
through
the
other
yeah
so
well,
the
questions
was
that
our
import
length
only
takes
an
int
which
gives
you
somewhat
of
an
upper
bound
by
a
few
bytes.
Potentially,
if
we
have
a
specific
implementation
that
looks
at
the
inputs,
man
as
well,
we
could
get
the
exact
precise
why,
with
phase
64
and
pausing,
so
it
yeah.
If
you
always
take
a
look
at
the
example,
pretty
that's
encoded
that
has
there.
If
you
give
the
input,
which
is
necessary,
noises
compute
from
base64
encoding,
no,
no!
No!
AC
This
is
bytes
to
you,
T
of
it.
This
isn't
good.
The
example
code
shows
encode,
okay,
so
so
it's
seven
characters.
Seven
values
right:
I
put
ABC
to
square
one
in
return
in
return,
eight,
if
you
are
file
box
false,
otherwise
it
will
return.
Twelve
and
I
kind
of
limitation
returns
twelve
regard
because
it
doesn't
have
that
pool
as
far
as
block.
Why
would
it
do
eight?
What
happens
with
the
G?
Since
there
are
file
block?
You
expect
more
to
come
so
you're,
not
including
the
padding
in
decides
that
would
have
been
excluded.
D
Okay,
so
there's
also
there's
something
that
we
need
to
talk
about
like
I.
Imagine
that
the
method
to
compute
the
length
you
would
not
call
it
with
the
next
block
that
should
be
counting.
You
would
call
it
with
like
when
you
have
all
the
data,
and
you
call
it
without
the
look
like
honestly
when
you
call
it
in
the
rope.
Like
you
know,
feeding
the
next
block
I
would
never
call
this
man.
You.
AB
O
D
O
Opposition,
what
will
it
optimize
it?
You
could
allocate.
It
say
that
your
output
buffer
is
currently
1k,
but
this
method
is
telling
you
like
I'm,
really
going
to
give
you
a
key.
If
you
give
me
this
much
input
data,
you
could
ahead
of
time
make
sure
that
you
have
an
aka
out.
The
number
that
way
transforms
us
a
little
bit
more
efficient.
You
can
get
away
the
number
times
the
loop.
D
AB
A
A
AC
D
Review
I
suggested
it
I
think
it
makes
sense
it
get
max
counter
from
here,
because
I
was
afraid
that
like
if
we,
if
we
in
some
of
these
IP,
is
we
kind
of
get
lucky
a
link
and
ethics
are,
you
know,
give
people
exact
numbers
in
some
other
AP
is,
we
won't
be
able
to
give
them
exact
numbers
and
then,
if
there
will
be
confusion,
so
this
should
be.
This
is
the
upper
bound?
Yes,.
A
D
AB
AC
D
AC
A
M
O
E
T
I'm
sorry
I
missed
it,
but
did
we
revisit
the
decision
of
whether
this
type,
whatever
we
kind
of
lost
you?
Can,
you
repeat
sorry
I
was
wondering
if
I
may
have
missed
it,
but
to
be
revisit
the
decision
of
whether
this
class
would
ever
potentially
have
anything
besides
utf-8
on
it,
because
if
it
would
that
max
decoded
bytes
thing
could
be
ambiguous
right.
D
O
T
O
Y
O
O
M
O
F
D
Also
like,
if
we
can
come
up
with
a
nice
name
that
includes
utf-8,
it
would
just
be
safer.
I
think
we
had
problems
that
you
know
it
doesn't
roll
off
the
tongue.
Quite
if
you
had
utf-8
to
this
name,
but
maybe
if
we
can
come
up
with
something
is
just
we
make
it.
You
know
explicit
and
edits,
future-proof
all.