►
From YouTube: .NET Design Reviews GitHub Triage
Description
00:00:00 - Approved: Consider adding MemoryMarshal.GetOwner(ROM) https://github.com/dotnet/corefx/issues/27237#issuecomment-367072826
00:25:14 - Approved: Add DecompressionMethods.Brotli? https://github.com/dotnet/corefx/issues/26995
00:38:11 - Approved: Add file enumeration extensibility points https://github.com/dotnet/corefx/issues/25873#issuecomment-367082482
B
B
If
he
has
it,
maybe
we
won't
be
able
to
review
it
today,
because
basically,
he
wants
ability
to
get
some
internals
from
I
think
from
sequence
from
sequence
position
and
if
you
want
to
do
it
like
March
on
like
ap
is
but
I,
don't
think
the
issue
is
yet
filed
and
I
told
him.
I
worry
a
bit
like
we
keep
adding
more
and
more
of
these
Marshall
back
doors.
So
that's
worry
number
one.
Secondly,
we
need
to
decide
how
they
go
to
the
same
class
or
they
on
some
different
I.
B
Think
the
position
was
just
giving
everything
anyway,
no
yeah,
but
it
gives
you
wrong
information
because,
basically
inside
there's
an
intent
and
object,
but
they
are
like
masked.
So
if
you
access
it
on
on
memory
sequence
or
read
on
the
sequence
you
getting
some
like
negative
int,
which
people
complain
that
you
know
they
thought
it's
an
index.
But
it's
a
negative
int.
What
do
I
do
a
bit
while
you
have
to
apply
a
mask
so
now
he
wants
an
API
to
apply
the
mask
for
you
and
I
kind
of
would
like
to
discuss
it.
B
A
A
B
Kind
of
you
shouldn't
be
doing
it.
That's
why
he
wants
to
pull
it
off
memory
marshal
and
I.
Think
it's
consistent.
This
one
is
consistent
with
some
other
ideas
that
we
added
to
memory
Marshall.
The
only
thing
I
don't
understand
is:
why
is
he
proposing
that
the
return
type
this
object
and
not
on
memory.
A
B
B
C
B
B
B
B
A
A
F
C
B
G
B
B
A
B
C
B
B
A
C
D
B
B
Should
not
be
a
extension
method
either
I
should
probably
just
we
only
have
any
well,
no
I
think
the
reason
that
on
memory
of
D,
you
have
try
gate
array
is
that
it's
kind
of
nuts?
It's
not
a
loophole
for
the
read
on
leanness
of
the
type
I
think
Ben
wants
specifically
read-only
memory,
which
then
gives
you
something.
That's
basically
read/write
and
all
these
are
on
memory
power.
Yes,
interacted,
like
only
March.
We
also
have
try
gate
array
at
exit.
B
I
G
I
E
B
B
B
A
Is
so
I
think
there's
two
different
sides
here?
One
of
them
is:
do
we
want
to
hide
the
fact
that,
in
the
same
field
of
restore
their
own
memory,
we
also
store
the
string
and
array?
Because
if
the
answer
is
yes-
and
you
will
always
have
to
do
a
cos
anyway,
because
you
need
to
basic
several
doses
that
string
and
neck
as
they
return
now.
A
B
A
A
A
B
B
K
B
B
B
B
A
B
B
C
J
B
L
I
A
K
A
C
B
J
A
J
B
A
B
A
B
A
B
C
A
B
B
A
G
I
O
I
M
Know
think
in
our
minds
socket
has
long.
Does
it
minute
cancellation
token,
where
we
register
with
it
and
dispose
of
the
underlying
connection.
My
screen,
I'm.
Sorry,
no
for
the
connection
base
socket
has
a
static
method
that
allows
you
to
cancel
the
operation.
It's
like
cancel
connect
a
sink
or
something
like
that,
and
so
we
register
a
cancellation
token
and
we
fire.
We
can
we
invoke
that
operation
when
cancellation
is
requested,
so
I
can't
find.
F
O
C
B
C
C
M
C
M
So
another
timeout,
when
you
enable
expect
100,
continue
for
an
HTTP
request.
Basically,
what
you're
doing
is
you're,
rather
than
sending
all
of
your
post
content
you're,
sending
a
header
that
says,
expect
100
continue
and
you
delay
sending
your
content
until
you
hear
back
from
the
server
that
says:
okay
go
ahead,
but
sometimes
you
never
hear
back
from
the
server
and
you
just
want
to
send
the
content
anyway.
M
So
most
implementations
will
just
send
it
after
some
period
of
time,
regardless
of
what
that
you're,
not
from
the
server
and
we're
currently
hard-coded
to
one
second,
but
we'd
like
to
be
able
to
make
that
configurable.
So
this
is
the
API
that
would
let
you
do
that
if
you
scroll
down
you'll,
see
that
this
is
exposed
under
different
names
and
other
places.
A
A
B
B
A
A
M
M
G
A
A
B
C
C
B
B
B
C
B
B
A
E
O
C
B
M
E
A
G
Most
part,
of
course,
is
still
alive.
There
are
a
couple.
We
should
go
one
last
time
over
counter
with
the
bits
that
are
potentially
we
consider
options
on
one
of
the
being
the
properties
of
each
bit.
We
expose
on
the
file
system
tree
right
so,
and
there
was
a
question
specifically
for
you
on
that,
like
ghost
even
brought
up
the
whether
or
not
they
should
be,
you
know
an
entire
version
stream.
Well,
it's
I
like
it
down
further
than
being
the
file
system.
G
Entry
right
there,
yeah
so
was
exposing
string
for
directory
or
sorry
for
the
root
directory
in
original
director.
Those
are
the
so
original
to
root
directories.
What
you
pass
in
to
do
your
search
there,
conceivably
and
probably
will
be
an
opportunity
to
pass
in
a
veto
expanding
for
your
your
source,.
C
G
B
G
B
G
You
know
I,
just
the
only
reason.
I
didn't
add
it
right
off.
That
is
because
look,
it
wasn't
super
surprise
there
as
far
as
a
constructor
for
the
numerator,
because
it
wasn't
super
high
priority,
we're
trying
to
like
shape
the
tree
as
hard
as
I
can
to
make
sure
the
other
stuff
actually
just
working
in
a
way.
That's
describable.
G
There
are
some
definite
subtleties
about
how
this
stuff
works.
I've
been
iterating
on,
like
the
differences
between
units
and
windows
and
stuff,
find
rational
ways
to
expose
things,
because
one
of
the
things
that
came
off
that
is
also
you
know
it,
which
we
knew
already
like
Windows,
gives
you
back
all
the
data
in
one
chunk
right,
UNIX
does
not
that
introduces
a
bit
of
a
timing
question
there.
You
know
if
you
don't
actually
go
and
act
access,
the
things
that
require
doing
doing
the
Van,
Eps
tab
or
saturate.
G
You
don't
get
the
data
as
it
was,
and
you've
got
the
original
data
back.
So
it
ends
up
that
like
doing
that
makes
the
cost
of
an
emerging
on
UNIX
like
three
times
as
much
as
it
is
to
not
get
it.
So
that's
what
that
one
I'm,
just
kind
of
lean
I
decided.
The
best
thing
to
do
is
just
describe
when
you're.
Actually
looking
at
the
data
being
passed
to
you,
it's
imperative
to
you
can
actually
grab
the
day.
Didn't
care
about
it
not
like
sit
on
it
wait.
G
So
the
behavior
of
the
API
is
that
it
won't
fill
things
out
like
that.
It's
actually
gonna
call
for
it,
but
it
won't
change
after
that
back
down
there.
So
whatever
we
get,
it
will
stick
and
I
constrained
it
to
the
one
thing
that
we
can't
get
so
like
I
get
the
get
the
existing
flags
for
units
right
before
whether
it's
it's
the
point
where
I
can't
get,
which
is
read-only
real
ones,
the
only
mode
we
have
to
make
other
calls
to
go
and
see,
but
that's
the
only
flag,
I
switch
from
that
point.
G
G
What
can
you
don't
get?
The
I
could
add
these
hidden
as
well,
because
it's
another
common
is
on
both
platforms.
So
I
already
have
it
specific
the
convention
out
of
crept,
but
we
have
that
convention
is
what
we
map
to
okay,
all
right.
Something
about
where
are
you
yeah
sit?
The
caveat
to
that
is
that
you
can
check,
you
can
see.
Well
am
I
on
UNIX
does
my
filing
it
begin
with
a
period,
but
there
is
a
tax
with
that
too,
because
then
we
have
to
convert
from
utf-8
right
to
utf-16.
So
is
it.
G
I,
don't
think
name
stuff
outside
of
the
ASCII
space
and
do
you
so
that's.
Why
like,
if
you
wait,
whether
it's
asking
or
now,
no
no
tell
me
well
cuz
they
match
right.
I
mean
the
first
starter.
27
code
points
make
some
things
really
easy
internally
apply
it
like
so
like
hidden
is
easy
because
you
just
look
for
space,
then
export
you
keep
it
on.
We.
B
G
Operation
so,
but
like
though,
there
are
certainly
optimizations
because
of
the
code,
the
the
the
optimizations
for
hidden,
you
don't
have
to
convert
right.
You
can
look
at
the
utf-8
buffer,
which
we
actually
have
in
this
case
and
I
can
say.
Look
good.
I
certainly
do
that
for
the
special
directories
I
look
for
period
in
period
actually.
M
G
G
G
C
C
G
H
G
H
M
G
You
know
there's
a
tax
on
that,
but
that's
that's
an
implementation.
Well,
the
only
thing
that
like
really
impacts
your
code
in
any
significant
ways,
just
the
the
atomic
the
audacity.
Actually
it's
getting
some
of
the
rockers,
but
that's
again
if
you're
writing
custom
thanks
for
that,
and
it's
not
it's
really
kind
of
hard
to
fall
over
in
those
cases
too,
but
it
is
possible.
It's
just
the.
A
G
G
G
That
you
get
out
is
not
an
atomic
there's,
no
way
to
make
it
atomic
communiques
right.
So
that's
just
right.
We
make
it
as
atomic
as
we
can
be,
and
we
can
help
push
that
a
little
further.
If
we
had
it's
hidden
to
it
as
well,
because
then,
then
you
can
get
all
the
attributes
are
very
few.
The
attributes
are
actually
get
back
on
there
to
try
and
map
them,
but
then
you
get
all
of
its
pieces.
Can.
P
P
P
B
A
G
C
G
G
G
Actually
you
know,
as
I
went
through
and
I
changed
it
to
use
a
ray
pool
to
throw
like
all
the
directory
games
and
that
one
faster
and
something
slower
so
like
I
I'm,
not
jumping
forward
on
that
when
I
had
to
spend
more
time
with
that,
but
like
there's,
clearly
lots
of
opportunity
to
continue
to
make
changes
here
that
that
could
that
will
in
some
cases,
will
definitely
get
better,
but
when
week
12
I
mean
internal
stuff
without
changing
the
api's.
So
I'm,
fine
with
that,
are
you,
okay,
with
adding
his
hidden
to
it?
I
G
I
A
Match
casing
mesh
type
remember
this
is
the
all
entities
to
my
all
and
then
those
ones
just
take
that
guy,
it
seems
reasonable,
I
think
I'm,
not
Tally's,
shown
the
existing
API,
but
I
think
these
ones
are
just
all
those
on
the
existing
ones.
Right,
yes,
like
past
heart
pattern
already
exists,
and
then
there
is
a
search.