►
From YouTube: .NET Design Reviews: GitHub Triage
Description
-12:-04:-25 - Approved: Support casting ReadOnlyMemory(T) to Memory(T) https://github.com/dotnet/corefx/issues/23879#issuecomment-346117097
00:08:18 - Approved: Move (ReadOnly)Span.DangerousGetPinnableReference to MemoryMarshal https://github.com/dotnet/corefx/issues/25412#issuecomment-346119552
00:16:46 - Approved: [ExcludeFromCodeCoverageAttribute] should be applicable to assemblies https://github.com/dotnet/corefx/issues/24694#issuecomment-346119984
00:19:25 - Approved: String.Contains(char) https://github.com/dotnet/corefx/issues/25094
00:20:49 - Approved: Expose UnixDomainSocket https://github.com/dotnet/corefx/issues/10981
00:25:43 - Approved: TryFormat with format strings as ReadOnlySpan(char) or string https://github.com/dotnet/corefx/issues/25337#issuecomment-346126703
00:43:37 - Rejected: "ItemRef" - Ref element accessor for types that already have an indexer. https://github.com/dotnet/corefx/issues/25189#issuecomment-346139011
A
B
A
A
B
A
C
D
D
D
G
G
D
A
Like
we
didn't
add
the
so
basically
I
think
originally
we
had
the
discussion
of
adding
an
unsafe
that
the
basis
does
it
for
span
of
T
to
read
only
so
far
from
we'd
only
span
of
T
to
span
of
T.
Is
they
do
people
get
rid
of
that
in
favor
of
basically
getting
the
raft
T
and
then
using
the
unsafe
function?
To
get
sorry,
we
the
only
ref
of.
H
D
E
Yeah
I
mean
I
think
that
there
is
sort
of
the
uber
point
that
emerged
as
part
of
this
or
prior
to
this,
that
we
should
have
this
memory
Marshall
class,
and
we
agreed
to
that,
and
we
already
have
it
and
it's
checked
in,
and
some
of
these
members
I
believe
are
even
already
checked
in
alive
to
go
check
and
then
it's
a
question
of
what
gets
moved
there
from
api's.
We
already
have
so
do
we
move
the
dangerous
get
pinnable
references
like
take
them
off
the
span
types
and
put
the
monta
memory.
E
Marshall
know
what
does
their
name
become?
Do
we
remove
it
dangerous,
try,
get
array
from
read-only
memory
and
put
it
onto
memory
Marshall
and
what
is
its
name
become?
We
should
probably
take
stock
of
where
we
currently
are,
what
we've
already
approved
and
probably
have
a
new
issue
that
that
covers.
That
sounds
good
to
me.
D
E
I
I
C
E
We
already
we're
already
in
the
process
of
doing
that
for
read-only,
spans
indexer,
so
I
would
assume
we
would
do
the
same
thing
here.
I,
don't
know
if
there's
a
good
reason
not
to
I
believe
the
we'd
have
to
check
with
Vladimir
I
believe,
there's
already
support
now
in
c-sharp
for
getting
the
address
of
a
refereed.
Only
so
the
the
pinning
argument,
kind
of
or
the
difficulty
with
pinning
goes
away
so
I
would
I
would
assume
would
be
refereed
only,
but
we
should
double-check
with
Jana
and
bladder
so
host.
E
E
J
D
E
D
F
I
D
E
E
C
C
D
E
D
E
K
D
A
K
M
M
A
E
E
A
The
sealing
I
think
this
is
something
where
I
somewhat
disagree
with
Christophe,
but
I
can
see
crystals
argument
from
the
point
of
view
that
the
guide
I
may
have
used
for
the
last
15
years,
and
it's
not
necessarily
a
reason
to
change
it,
but
pretty
much
in
my
entire
code
bases.
If
you're
going
get
up,
you
will
see
that
almost
all
my
tabs
will
always
seal,
because
my
gem
will
stances
that
types
have
to
be
designed
to
be
extensible
and
I
find
it
always
super.
A
Weird
people
derive
from
my
sides
just
to
add
a
few
properties
which
is
really
what
it's
for
I
mean,
like
extension,
methods,
get
you
pretty
far
without
having
you
get
you're
having
to
extend
the
times,
but
it's
really
about
the
state
you
want
to
attach
to
that,
but
giving
it
in
half
the
time
you
don't
own.
You
know
the
creation
path,
you
don't
get
to
choose
to
type,
so
it
doesn't.
It
doesn't
work
anyway
and
the
other
habit
just
find
it
weird
that
I
have
to
deal
with
customers.
A
C
A
E
E
There's
I'm,
not
aware
of
any
scenario
for
accessing.
In
that
said,
I
don't
see
a
downside
to
it
other
than
it's
an
extra
member
that
you
know.
Maybe
it
would
cause
a
problem
in
the
future.
I
have
no
idea,
but
I'm
always
I
mean
maybe
I'm
too
conservative
in
this
regard,
but
I
always
that
on
the
side
of
not
exposing
something
unless
we
find
a
use
for
it,
I
think.
A
Generally
speaking,
I
said
before,
like
I
I'm,
always
a
person
that
likes
symmetry
like
if
I
can
pass
into
value,
it
seems
somewhat
logical
to
be
able
to
get
it
out
of
it
as
well,
because
constructor
arguments
are
usually
meant
to
be.
You
know,
stored
somewhere,
they're,
usually
not
like
used
and
then
forgotten
mmediately.
A
A
The
I
mean
it.
Is
it
you
care
to
say
that
the
scenario?
Basically?
Is
you
instantiate
this
type
very
locally,
and
then
you
immediately
forget
it
anyways,
because
you're
you
only
basically
tell
it
to
use
the
you
know,
a
path
for
a
socket
right,
which
you
can
only
do
on
UNIX
right,
so
yeah
I
mean
again
I
could
go
either
way.
It
just
seems
odd.
E
I,
don't
have
any
I
mean
I
if
it
was
me,
I
wouldn't
at
it,
but
I
have
no
problems
like
I.
Don't
think
it'll
cause
any
problems.
So
if
we're
doing
it,
you
know
if
the
symmetry
is
important,
like
that's
fine
or
if
you
know
somehow,
I
don't
know
looking
at
it
in
the
bunker
or
the
debugger
would
see
the
private
field.
A
H
A
B
B
E
E
Ammeter,
so
you
can
pass
it,
you
know
D
or
G,
63
or
whatever
right
now.
That
format
parameter
is
a
string,
and
we
did
that
simply
because
we,
you
know
who
would
generate
a
custom
format,
parameter
dynamically
and
passed
that
in
apparently
we
do
it
in
like
string
dot
format,
for
example,
or
string
builders
out
a
pen
format
we
parse
out
of
the
format
string.
That's
the
you
know
the
substring
that
represents
the
format,
parameter
on
a
hole,
and
then
we
pass
that
to
the
format.
E
So,
if
we
could
today
to
do
that,
you
have
to
do
a
substring.
If
we
allowed
these
try
format
methods
to
instead
take
a
read,
only
span
of
char.
You
could
just
do
a
slice
without
allocating
and
passed
that
in
there's
another
case
I've
seen
where
you
dynamically,
like
big
integer.
Does
this
it
dynamically
generates
a
format
based
on
the
number
of
digits
like
a
digits
argument
and
so
it'll
append
D
with
that,
you
know
format
at
digit
number.
E
E
Part
of
the
issue
is
today
I,
don't
know
if
we
have
a
good
reason
for
this
or
not
there's
no
implicit
conversion
from
string
to
span
R
from
string
to
read
only
span
and
because
of
that,
if
we
made,
if
you
just
change
the
existing
overloads
to
take
a
read,
only
span
to
call
it
with
a
string.
You'd
have
to
tack
on
like
dot
as
read
only
span,
which
makes
a
little
bit
more
difficult
to
use.
H
E
L
E
L
E
L
An
archetype
to
me
based
off
having
used
the
stuff
like
I'm
it
like
you,
you
can
pass
an
array
into
a
read-only
span
and
they're
a
byte
I,
guess
a
of
T
and
it's
very
natural
and
lets
you
drop
in
these
better
replacements
without
needing
to
make
the
overload
so
I
think
that
we
should
probably
whether
or
not
it's,
for
this
add
the
implicit
conversion
from
string
to
read-only
span
of
char.
Unless
you
know,
yon
wants
to
give
a
reason.
L
D
A
The
only
thing
I
could
imagine
is
that
it's
me
is
the
bird
to
do
something
similar
to
what
job
is
done
with
their
encoding
for
Strings
I
could
imagine
a
wall.
You
said
there
are
certain
strings
that
are
not
utf-8
but
like
whatever
I
coding,
they're
using
I
forgot
about
the
name
is
some
some
western
Latin
thingy,
which
is
effectively
you
know
single
byte
for
most
of
the
term
then
read
only
like
we'd
only
start
off
the
charm
might
be
inconvenient.
We
may
have
another
or
below
the
text.
A
I
A
D
A
L
Did
in
in
corefoundation
their
string
it
it
can?
No
it's
an
ASCII
string
or
utf-8
string
or
utf-16
string,
and
the
only
time
that
it's
ever
relevant
is,
if
you
ask
it,
can
I
have,
can
I
see
your
bytes
and
then
you
have
to
tell
it
I
believe
your
encoding
is
this
encoding?
If
you
get
the
if
you're
right,
then
it
will
give
you
the
bytes
and
if
you're
wrong,
it
will
not,
and
so
it's
the
sort
of
test.
L
L
Do
even
TF
32
strings
because
in
some
cultures,
they're
more
efficient,
maybe
that
so,
if
we
were
going
to
make
string,
have
multi
encoding
than
then
the
implicit
Torito
way
span
of
chart
could
be
expensive,
but
to
char.
It
would
also
be
more
expensive
than
a
current
is
I.
Don't
know,
I,
don't
know
how
much
we
would
be
willing
to
change
the
intrinsic
wet
string
beads.
L
L
But
either
way,
I
agree
with
Carol's
assessment.
That
number
two
is
probably
better
than
number
one.
So
just
on
the
things
that
are
outing
to
a
span,
there's
things
that
are
writing
to
a
span.
We
take
the
format
as
a
read-only
span
of
char
and
then,
if
we
get
feedback
that
this
is
too
much
of
a
pain
in
the
butt
to
use,
we
can
add
the
overloads
later.
E
L
D
D
L
D
L
And
now
shrink
I'm
like
when
you
switch
from
string
to
read-only
span
of
charge,
you
lose
dust,
you
lose
the
ability
to
distinguish
between
Nullah
empty
meaning
that
you,
if
before
empty
meant
something
if
no
limit
do
the
default
thing
and
empty
met.
Do
something
else
that
make
it
knowledgeable
and
you
don't
want
to
I'm
just
saying
that
would
be
the
only
reason
to
not.
A
A
L
A
E
Don't
think
that's
a
valid
argument
for
these.
Maybe
it's
fine
for
the
utf-8
ones
for
those
scenarios,
but
I
mean
these
try.
Format
methods
are
primarily
about
avoiding
a
string
allocation,
the
cost
of
doing
the
formatting.
If
it's
the
same
between
the
two
string
and
the
try
format,
it's
really
just
is
there
the
string,
allocation
or
not
I,
don't
know
why
custom
format
impacts
that
I
mean.
L
I
E
L
M
D
A
A
The
opposite,
like
I,
mean
you
base,
is
saying
that
the
add
new
try,
format,
api's
and
those
would
never
take
a
string
base
for
us
and
it
would
always
take
we'd
only
spend
off
char
string
and
then,
if
later
on,
we
would
add
the
implicit
conversion
from
strings
to
be
the
only
span
of
char.
Then
you
just
you
know,
call
this.
Just
don't
have
to
use
the
as
you
don't
we
spend
anymore.
They
can
use
posit
indirectly
right.
E
C
E
J
L
D
I
M
L
D
L
D
L
I
I
A
N
N
In
particular,
we
want
that
in
immutable
immutable
array,
at
least
right
builder,
in
particular,
because
I
you
can
see
so
much
annoying
code
where
you
have
a
stack
of
T
in
Rosa
and
they
want
to
change
what
you
have
on
the
top.
So
you
have
to
push
it
and
change
it.
Chocolate
change
it
and
push
it
back.
If
you
can
get
a
reference
to
the
top,
you
could
just
change
it
in
place
right,
but
but
it's
not
possible
because
it's
already
builder
underneath
and
that
has
by
Wow
indexing.
N
C
N
N
Don't
leave
FTC
BB
doesn't
support
at
all.
If
you
be
supports,
fifty
doesn't
support
read-only
of
key
for
now
I
see
and
it
doesn't
support
them
specifically
because
I'm
attracted
it
doesn't
know
about
read-only
at
all,
but
we're
also
poison
the
signature.
So
it's
we
intentionally
make
it
not
consumable
from
VB,
because
you
should
better
smart
and
know
how
to
use
it
or
you
should
not
be
able
to
use.
We
only
make.
A
N
A
N
N
A
A
A
N
D
D
N
O
Just
the
name
right,
yeah
yeah:
this
is
not
a
discussion
on
whether
or
not
we
should
add
this
any
given
tight
its
discussion
on
when
we
decide
that
to
have
the
Throwdown
over
list
of
tea
or
anything
else,
Throwdown,
we
can
take
bike
shutting
out
of
equation,
it's
like
if
we
do
it.
This
is
how
we're
going
to
do
it
so
what
we're
doing
with
no
yet
now
we
are
like
shedding.
H
A
I
I
A
N
K
N
K
D
N
N
A
Talked
to
him
briefly
offline,
like
I
mean
he
didn't
even
hear
what
the
proposal,
apparently
so
I
explained
a
bullet
or
him.
He
was
insisting
that
arrays
do
not
return.
We
have
T
that
I
showed
him
in
the
debugger
that
actually
does
return.
50
do
then
the
say
I
think
he
buys
into
the
scenario.
We
understand
the
media
and
the
message
similar
to
me
didn't
seem
super
excited
about
the
name,
but
it
also
don't
think
you
had
like
a
massive
problem
with
it
either
like.
D
F
N
D
D
D
D
A
D
A
O
D
N
N
K
N
N
E
N
Convoluted
there
is
a
special
API
which
we
call,
and
that
is
specifically
because
we
cannot
kill
to
the
element
as
a
reference
way
we
can
for
array,
but
we
cannot
forest
thing.
So
that's
why
this
strange,
API
X's
were
like
offset
from
the
beginning,
something
and
the
moment
you
want
to
have
it
any
changes
to
the
string
but
separated
into
objects,
or,
let's
do
you
jf8
or
compression,
or
something
you
run
into
this
hardware
requirement
that
the
first
character
must
start
after
the
beginning
of
the
object.
N
N
D
N
Will
definitely
have
a
redundant
reference
index
ale
but
then,
like
you're
returning
it
character,
char,
pirate,
so
I
four
bytes
to
version
two,
so
I
think
string
is
a
no
one.
Extremists
too
simple
the
dictionaries
problem,
another
extreme,
that's
too
complicated
I
could
see
it
by
real
indexer
on
the
dictionary.
It
will
have
a
take
the
key
and
give
you
it
certainly
can
be
implemented
and
probably
can
even
be
used,
but
feels
it's
a
bit.
That's
actually
per
day
was
for
I
guess.
N
N
K
N
A
So
yeah
I
think
there's
nothing
wrong
dictionary
turning
you
Rafi,
because
I
think
it
is
again
I've
seen
cases
where
it's
someone
who's
strong
to
avoid
dictionary,
lookup
sent
me.
The
reason
I
think
is
user
is
because
we
added
other
methods
to
avoid
the
double
yes
and
I.
Think
this
is
getting
rid
of
that
problem.
We
can
just
say
we
don't.
We
only
have
one
method
now.
N
A
F
N
N
N
N
A
N
N
D
You
should
have
probably
warning
at
the
item
earth
like
the
same
stuff.
We
have
in
iterators.
Like
don't
change,
you
know
the
underlying
data
structure,
basically
the
same
hopes
for
the
item,
wrath
yeah.
Whenever
is
you
know
because
you
drawing
on
the
implementation
that
the
thing
is
not
going
to
go
away
or
you
might
get
unlucky?
If
you
insert
things
and
or
you
know
it's
resized
through
stuff
yeah,
so
something.
O
To
say
we
had
that
whole
like
what
happens
like
when
we
did
this
in
the
dory
we
had
the
whole,
like.
Oh
my
god
like
if
we
make
list
of
t,
have
a
ref
indexer,
because
we
could
change
things.
We
do
were
bound
they're
like,
but
then
someone's
gonna
take
a
rest
to
the
item.
They're
going
to
index,
they're
gonna
set
it
and
it's
going
to
break
expectations.
B
O
D
D
O
Grieve
you,
but
as
there
will,
you
put
it
on
concurrent
college
that
one
is
trickier
because
of
the
inherent
because
the
that
one
is
trickier,
I
think
because
list
is
a
type
where,
if
you're
using
it
in
a
multi-threaded
fashion,
you
you're
already
broken.
Unless
you
have
taken
the
steps
to
protect
it,
where
it's
like
the
expectation
is
your
list
of
T
is
being
used
in
a
single-threaded
fashion,
except
if
you
have
explicitly
had
a
case
where
you
are
doing
the
proper
walking
and
you
have
controls
are
to.
O
A
That's
where
we
have
the
limits.
How
far
we
can
go
was
allowed
to
say,
like
I
mean
really.
Is
it
earlier
that
dictionaries
one
common
examples
you
wanna
you
want
to
avoid
the
double
look
up
to
do
like
an
indigent
increment,
for
example,
right
which
for
concurrent
dictionaries,
you
wouldn't
be
able
to
do,
because
you
have
to
call
it
try,
get
or
update,
because
you
once
the
opted
to
be
atomic
right
right.
N
A
N
N
A
Actually,
no,
we
have
like
I,
don't
know,
should
we
did
it
on
hash
set
was
sorted
said,
but
on
on
the
immutable
ones
like
Andrew
edit,
the
like
big
vector,
like
the
pollen
was
they
had
somewhere
in
the
project
system.
They
had
a
set
of
file
paths
and
they
wanted
to
get
the
actual
casing
out
of
it
to
be
added
an
API
on
said,
but
you
could
basically
like
given
given
some
representation
that
would
match
it
would
give
you
the
actual
instance
back
then.
O
A
O
Graph
yeah,
you
yeah,
that's
you!
You
want
the
ref
to
the
World
Tour
Preferred,
when
I
think
I
think
any
argument
made
for
sack
you
can
you
can
make
a
DQ's
for
cues
they
litter?
Oh
now,
you'd
want
this.
The
taught
you
the
next
thing.
That's
coming
out
like
how
many
lists
as
well
they're,
just
as
official
yeah.
D
A
N
C
N
D
A
A
A
So
things
like
that
that
leads
to
me,
like
I,
mean
I,
don't
think
we
have
a
single
name
would
have
a
pattern.
It
says,
take
whatever
thing
you
have
like
an
index
or
the
names
item
usually
for
peak
it's,
you
know
peak
and
you
suffix
it
with
rapid.
That
will
be
the
pattern
name,
and
that
seems
reasonable.
It's
reasonable
and
short.
N
N
A
H
N
O
The
the
benefit
of
breath
only
like
basically
its
breath
is
useful
in
a
couple
different
places
particular
like
holding
on
to
a
slot
to
add,
later
or
also
potential
or
the
essentially,
the
other
one
is
rather
than
holding
on
a
list
and
and
index.
You
can
hold
on
to
a
breath
and
use
that
as
a
place
to
insert
the
other
one
is
on
immutable
types
on
immutable
types.
The
value
that
ref
provides
is
essentially
avoiding
copies
to
the
stack.
Essentially
when
you,
when
I,
want
to
basically
I
have
an
immutable
array
of
giant
freaking
struck.
O
O
Can
say
like
get
instead
of
in
a
case
today
we
are
like,
if
I
want
to
do
an
insertion,
and
you
had
to
you
today
healing
what
you
can
do.
It
is
either
pass
a
delegate
or
you
pass
that
collection
in
an
index
and
say
insert
bill
and
breath
and
indexers.
You
just
pass
a
ref
to
the
location
and
say
just
sign
here,
so
we
will
have
to
add
one
more
method
on
list
item.
Ref
returns,
the
rest,
it's
a
ref
returning
method.
You
can
take
a
rest
of
that
value.
O
D
D
D
H
N
D
N
N
I
H
H
N
D
N
A
N
A
D
A
N
A
D
D
O
O
O
That
the
fact
that
the
immutable
rays
weren't
two
of
them,
the
first
two
were
grouped
and
so
every
time
I
look
them
like.
We
only
did
those
two
musical
things.
Why,
like
my
mind,
can't
read
further
down
the
list
but
having
the
builders
and
the
arrays
have
item
reps
would
significantly
like
there
are
number
places
were
like
yeah.
We
could
significantly
change
upper
future
types
could
just
have
indexes
which
are
right,
I,
think,
but
again.
D
N
O
The
only
downside,
the
the
only
so
when
you
talk
about
like,
should
I
have
a
buyout
indexer
versus
a
virus
indexer
by
ref
indexers,
except
for
one
case
they
allow
for
maurices
than
five
out
once,
for
instance,
you
can
pass
them
by
breath.
You
can
take
an
index,
turn
pass
it
by
ref.
Also,
you
can
dot
through
on
mutable
structs
and
you
can
assign
through
two
new
trucks.
You
can
only
do
that
by
ref
index.
You
can't
do
it,
the
bivalve
indexer,
the
only
downside
is,
you
can't
call
a
reflection.
A
N
N
A
N
N
I
N
Right,
yeah
yeah:
they
fully
support
them.
It's
just.
You
cannot
declare
your
own
so
like.
If
you
have
abstract
class
with
refrigerant
thing,
then
good
luck
using
that
and
maybe
not
ever
right,
but
you
could
use
it.
You
could
call
it
an
aside
through
it.
It
understand,
but
if
variable
is
not
new
to
be
be,
it
knows
about
by
ref
parameters.