►
From YouTube: .NET Design Reviews: GitHub Quick Reviews
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
D
Right
so
you've
resolved
the
name
through
DNS.
Each
address
family
you
resolve,
needs
a
separate
packet
and
in
certain
instances
you
only
want
a
single
address
family,
so
you
can
optimize
things
by
sending
packets.
This
essentially
just
enables
that
just
adding
address
family
parameters
to
all
of
these
methods,
yeah
and
also
adding
cancellation
tools,
training
value
tasks.
A
A
B
D
G
Mean
in
general,
with
those
api's
because
they're
callable
to
do
both
either
overlapped
or
not.
You
know
the
it
can
give
you
back
an
answer
for
the
sink
case,
but
if
you
call
it
with
an
overlapped
offer
complete
asynchronously,
even
if
it
had
the
data
already
available.
Some
api's
allow
you
to
say
to
complete
synchronously,
even
in
that
case,
but
maybe
worth
check
right.
A
G
A
What
if
I
say
that
the
behavior
of
the
operating
system
today
is
that
if
you
pass
in
an
overlapped,
it
completes
on
a
completion
for
thread?
Maybe
some
future
API
allows
you
to
pass
a
flag
that
says
no
I
really
wanted
the
synchronous,
if
possible,
yeah
like
would
we
be
cheating
ourselves
in
the
foot
by
foreclosing
the
ability
to
return
a
value
task,
I
mean
yes,
I!
Think
we're.
F
G
G
G
If
it
to
me,
it
really
comes
down
to
what
do
we
expect
the
usage
to
be
if
we
expect
that
people
are
only
gonna
ever
just
await
this
thing,
then
you
know,
value
task
is
fine.
My
concern
was
something
like
this,
as
someone
might
end
up
doing
their
own
caching
like
they
might
call
get
hosts
entry,
a
sink
store,
the
return
tasks
into
a
dictionary.
G
That
seems
like
a
perfectly
reasonable
thing
to
do,
and
then
every
subsequent
use
just
goes
to
the
dictionary
and
says,
give
me
the
task
of
IP
host
entry
for
Microsoft,
calm
and
all
await
that
and
again
did
that
with
the
existing
API
you'd
be
fine.
If
you
did
that
with
the
new
one
that
returned
value
tasks,
you
be
hosed.
A
D
G
If
we
can
get
Windows
to
do
the
synchronous
thing,
then
it's
worth
considering
value
tasks,
otherwise
I
would
prefer
tasks
because
I'm
concerned
that
people
that
that
that,
for
this
API
in
particular,
there
is
a
usage
pattern
that
might
lead
people
to
a
bit
of
failure.
Now,
if
you
can
prove
to
me
that
no
one
in
the
history
of
you
know
you
say
yet,
does
that
then
are
any
way
too
strong?
G
If,
if
you
can't
find
any
examples
where
people
are
doing
something
like
that,
then
maybe
I'm
wrong
and
it's
fine,
but
you
know
in
lots
of
an
API,
is
like
streamed,
read
async
I've,
never
seen
anyone
store
the
result
into
a
dictionary
for
the
lifetime
of
the
process,
but
for
something
like
get
me,
the
IP
address
for
this
domain.
Name
like
that
seems
like
a
somewhat
reasonable
things.
Do
sure
that
sounds
sense
thing.
A
F
Pretty
sure
yeah
I
think
the
value
test
versus
task
is
issue
and
I
kind
of
I
was
especially
interested
because
of
the
interrupts
like
I.
Don't
we
we
discussed
the
guidelines
a
month
ago
about
when
to
use?
Are
you
tasks
when
to
use
tasks,
and
we
didn't
actually
hit
the
case
of
what
happens
in
interrupts
with
completion?
Part
API,
like
you
know,
what
are
the
rules
in
such
cases?
Firm.
A
H
D
A
F
D
D
A
F
A
D
G
C
Those
eyes
now
and
I
think
Linux
has
some
but
I
don't
know.
If
we
expose
anything,
we
don't
have
that
anywhere,
because
that
might
be
a
big
thing.
It's
like,
if
you're
on
a
tethered
connection
on
a
laptop.
You
don't
want
to
do
parallel
connections,
but
if
you're
on
a
unlimited
bandwidth
desktop
connection,
you
would.
A
D
B
F
K
F
A
What
I
believe
it's
not
a
breaking
change
to
take
a
small
enon
like
this,
maybe
one
that
contains
only
one
value
and
make
it
flatter
the
two-string
behavior
gene
is
doesn't
it
it?
Wouldn't
because,
if
you're
going
from
one
value
to
several
like
to
string,
never
would
have
returned
anything
without
a
comment.
C
M
A
D
We're
not
sure
exactly
which
one
we
want
to
use
right
now,
there's
happy
eyeballs
we've
talked
with
the
windows,
tcp
team
and
they
have
their
own
algorithm
that
they
call
happier
eyeballs.
So
we
still
have
to
talk
it
with
them
and
see.
If
that's
what
we
want
to
do
here,
plus
I,
think
Arabia
is
often
try
to
express
intent
rather
than
a
specific
standard.
No.
A
Do
not
always
because,
if
there's
especially
when
we're
talking
about
protocols
like
generally
speaking
when
we're
calling
an
API
that
implement
a
protocol,
we
want
them
to
follow
a
very
particular
protocol
or
a
difference.
Maybe
v2
of
that
protocol,
just
something
to
consider
I
guess
like
if
this
is
in
the
end
going
to
be
8305
and
if
some
future
are
F.
A
A
F
Okay,
that's
Ted,
I
would
move
this
parameter
to
be
last,
since
it's
a
new
addition
and
then
I
would
not
ship
it
as
a
flags.
Enum
I
will
rename
it
to
something
like
you
know,
connection
strategy
or
algorithm,
and
then
it's
just
a
list
of
options.
If
we
need
to
add
another
one
that
you
can
mix
in
I
think
it
would
need
to
be
a
separate
parameter.
Otherwise
you
know
it.
A
G
A
A
F
F
D
F
C
L
C
L
G
A
So
what
we
are
so
what
do
is
we
could?
We
could
say
this
Imam
is
very
particular
to
like
a
multicast
or
any
caste
type
thing
I.
Think
someone
had
mentioned
strategy
earlier.
As
a
name,
yeah
I
mean
strategy.
I've
only
ever
seen
the
term
strategy
used
in
a
type
name
when
dealing
with
retry
strategy.
To
be
honest,
things
like
back
off
limits,
we
could
say.
D
F
Yeah,
if
we
think
that
we
will
have
more
options
related
to
this,
you
know
the
connections
priority.
Then
it
shouldn't
be
Amina.
The
pattern
is
to
have
like
a
small
stroke
that
has
many
different
fields.
If
you
are
in
this
situation,
the
the
thing
of
like
claiming
an
enum
where
some
fields
can
be,
you
know
combined
and
some
don't
we
already
know.
This
is
a
complete.
D
F
M
C
L
H
H
N
A
A
A
C
The
I
Triple
E
spec
defines
some
min
and
Max
functions.
The
one
that
we
are
currently
exposing
does
not
curve
does
propagate,
not
numbers.
We
should
also
expose
one
that
does
not
propagate,
not
numbers.
Not
only
is
it
more
efficient,
but
it's
useful
in
various
graphics
applications
and
other
scenarios.
C
C
Something
in
infinity
is
infinity
right
and
if
you
have
max
of
1
and
not
a
number,
the
current
API
returns,
it
depends
on.
It
will
return
out
a
number,
and
if
and
this
one
will
always
return
one,
because
it
is
a
number
not
number
of
man
and
man,
its
then
or
is
throw
its
nan
and
it
depends
which
which
nan
you
get
depends
on
the
algorithm.
Typically,
it's
the
first
man.
F
C
Is
very
a
few
people
in
various
performance
oriented
scenarios
that
have
requested
this?
We
were
originally
going
so
in
the
2008
spec
for
a
triple-a.
They
did
not
define
the
not
number
behave.
Behavior
and
various
API
is
like
the
C++
API
do
not
propagate
not
number,
so
we
were
going
to
fix
our
API
to
be
consistent
with
that
that
broke
WPF.
So
we
had
to
revert
it.
C
F
C
F
M
C
Would
be
non
efficient
like
Cochin
would
be
terrible.
Are
there
any
other
if
you
have
a
boolean
and
you
have
to
branch
to
decide
which
algorithm
you're
doing
in
a
hot
path
yeah,
but
I
would
assume
that
it
would
just
be
an
intrinsic
the
trip.
Would
there
no
intrinsic
for
max.
You
have
to
do
special,
it's
multiple
instructions
to
do
the
coke
Jen.
Oh
he's
saying
that
that
would
I.
A
A
C
Have
this
like
it's
their
ad
number
subtract
number,
no,
the
the
behavior
for
ad
and
others
is
clearly
defined
for
not
number,
which
is
it
returns,
not
a
number,
whereas
max
the
behavior.
You
want
on
whether
you
want
a
number
or
not
number
varies,
based
on
the
application
and
algorithm
you're
writing.
C
A
L
C
A
C
C
Let
see
the
problem
is,
is
that
both
are
I
Triple
E.
This
is
just
a
particular
variant
of
the
actively
spec
called
max,
not
sure
so
so
that
might
be
confusing
too,
because
it
might
imply
that
the
other
version
is
not
I
Triple
E
compliant,
which
is
what
I
Triple
E
remain.
Remainder,
implies
about
the
float
modulus
operator.
It
is
not
I,
Triple,
E,
compliant.
C
M
Like
you
say,
someone
needs
to
consider
not
a
number,
but
if
you
aren't
somebody
who
is
well
versed
in
I,
Triple,
E
754
and
you
think
of
doubles
just
as
integers
with
decimals,
then
max
number
and
max
doesn't
mean
anything
to
your
like
there's
two
options
here,
one
of
them
shorter
I'm
going
to
go
with
the
shorter,
which
is
the
behavior
that
everyone
else
has
already.
But
it's
you're,
not
you're,
not
considering
nan
at
all.
You're
I.
Don't
know
why
there
are
two
of
these
max
is
the
thing
I
want.
Yeah.
C
M
C
It
depends
on
the
algorithm
and
whether
or
not
nan
is
important
for
WPF
they
use
nan
to
represent
no
value,
and
so
it's
important
for
them
that
they
maintain
the
existing
nan
propagation
behavior
for
other
algorithms
like
where
you're
doing
games,
and
you
want
to
ignore
nan
and
perform
an
action,
preferring
actual
value.
You
want
to
not
propagate
the
name.
C
F
C
F
D
F
C
What
I
was
saying
is
I
think
max
is
one
of
those
api's
where
it's
like
two
string.
Most
people
aren't
going
to
type
dot
two
string
and
then
be
like
Oh.
What
are
all
the
overloads
here
and
explore?
They're
gonna
type
dot,
two
string
open
friend,
maybe
dot
TOS
tab,
open
friend
and
they're
not
going
to
care
because
they
know
what
they're
doing
yeah
or
think
they
know
where
they're
doing.
F
This
yes
I,
but
I,
don't
I
think
some
some
will
be
basically
will
do
like
will
notice
that
and
like
oh,
what
the
heck
is
another
one,
not
returning
a
number,
and
we
also
have.
F
C
F
C
F
D
C
D
C
Most
software
code-
that's
written
today,
tends
to
ignore
Nan's
in
favor
of
writing
mathematical
white
code,
which
is
why
the
defaults
on
most
platforms
isn't
strict
I,
Triple
E,
it's
generally
like
FP
fast.
When
you
do
oh
three
compilation,
because
people
want
to
be
able
to
write
code
that
looks
like
math
and
they're.
Okay
with
minor
rounding
errors,
and
so
they
don't
want
a
net
propagation.
They
don't
want
it
exceptions
to
be
thrown
when
an
an
is
encountered,
etc.
They
just
want
to
kind
of
pretend
it
doesn't
exist,
but
then
you
make.
M
What
people
should
probably
eat
it
to
me?
I
think
the
other
one
makes
more
sense
if
I
asked
for
max
of
Bandhan
something
it
means
garbage
in
garbage
out,
man
went
in
man
comes
out
yeah.
This
question
made
no
sense
and
that's
clearly
what
we
just
think
the
safer
and
so
so
I
don't
know
that
it's
clear
that
any
plate.
M
M
A
F
M
M
C
Then
you're
going
to
have
a
lot
of
code
where
you
see
math,
dot,
max
XY,
true
or
XY
false,
because
people
don't
like
using
names
before
optional
parameters
all
the
time
and
so
and
that's
going
to
be
probably
more
confusing
than
someone
saying
bath
max
and
math
max
number
and
then
hitting
f12
to
go,
see
the
docs.
You
said:
what
do
you
pool.
A
M
A
People
are
clinical
unless
they
go
out
of
their
way
to
discover
that
they
have
to
call
me
because
they're
just
going
to
keep
using
max
min
and
so
on.
That's
my
expertise
is
that
acceptable,
even
though
we're
also
saying
people
should
be
using
these
over
just
about
yes,
I
think
that
they
should
baby
yeah.
J
C
M
C
M
C
M
F
C
It
likely
be
a
we
basically
deprecated,
math
math
has
move
all
the
types
to
the
two
there
effective,
including
the
like
integer
math
methods
and
moving
forward.
We
do
that
may
be
a
good
design,
considering
we're
adding
new,
primitive
halves,
who
may
add
more
in
the
future,
and
this
is
going
to
keep
coming
up,
and
maybe
we
can
schedule
a
slot
for
that.
One
mo
gets
back
yeah.
M
A
M
A
C
M
N
F
C
F
C
I
could
finish
flushing
out,
I
could
finish
going
through
the
new
hydro
police
back
in
old,
look
together
the
full
list,
but
there's
also
like
of
all
the
float
in
double
eight
that
exists
today.
These
same
need
to
be
added
for
half
and
the
same
will
eventually
be
it
needed
the
functionality,
but
we
won't
need
a.
C
A
A
C
Think
it
was,
these
api's
are
fine
to
be
added.
We
just
need
to
decide.
Are
we
going
to
moving
forward,
be
adding
things
to
the
the
primitive
types
directly
in
which
case
they
don't
get
added
to
math?
They
get
added
to
those,
otherwise
we're
adding
things
to
there,
and
we
should
only
add
them
there.
So.
A
H
H
A
Of
wish
that
that
we
could
just
like
look
at
a
module
level
attribute
or
something
like
that,
and
just
say
this
is
the
version
of
math
back
by
the.
F
C
A
C
A
A
A
A
A
M
M
C
A
A
That
was
the
whole
point.
Oh
yeah.
There
is
some
discussion
in
here
as
to
why
we
don't
want
to
do
that
and
there's
no
way
for
the
jet
to
that.
You
can,
in
certain
cases
alight
it.
But
if
you
have
an
inside
of
the
loop,
the
trip
will
not
be
able
to
the
light
and
inside
of
it
you
have
to
access
an
outside
the
loop
first,
in
order
for
it
to
be
alighted
inside
the
loop.
Okay.
A
So
yeah,
so
that
was
ultimately
the
problem,
so
the
suggestion
was
okay.
Instead
of
calling
it
get
reference
calling
carihuela
the
this
name
is
intentionally
chosen.
The
ref
in
the
name
implies
that
it's
returning
a
wrath
and
not
escape
like
you
should
you
should
read
the
value
as
a
raft
here
and
not
as
a
team.
A
Just
like
how
get
reference
you
should
not
try
to
dereference
unless
you
know
for
a
fact
of
safety.
So
couldn't
you
just
have
a
name
that
describes
the
semantics
a
bit
better.
So
the
the
semantics
are
get
me.
The
reference
to
where
the
zeroeth
element
would
have
been
inside
of
the
arrival.
I
could
call
it
that
or
throw.
M
F
Surviving
overloaded
basic
you
can
color
collide,
have
used
the
existing
name
that
we
wanted
to
use
them.
Just
add
you
know
throwing
or
something
like
this.
We
have
some
ID
is
that
the
suffix
we're
throwing
between
see,
basically
it
can
throw
on.
Now
you
have
to
read
dogs.
This
is
on
muscle
memory,
Marshall
ID,
now,
I'm,
not
sure
it
with
reading
dogs
and
I
mean
it's
an
option.
Frankly,
I
don't
care
too
much,
because
it's
an
API
or
memory
Marshall
for
you
such
a
fun.
You
know
yeah,
it's
already
and.
A
M
A
A
A
Sorry
would
we
ever
add
an
overload
restraint.
We
already
have
string
a
string
dawn,
get
pinnable
reference
and
the
reason
there's
some
discussion
in
here
as
to
that
to
string
dog
and
pinnable
reference
is
always
the
safe
method
to
call,
even
if
the
string
is
empty,
because
then
we
return
a
reference
to
the
null
terminator.
So
if
you
dereference
it
this,
if
you
pass
in
an
empty
array
and
dereference
it,
potentially
all
hell
could
break
loose
because.
C
A
With
the
reason
that
you
need,
the
rest
of
instead
of
the
teat
star
might
be,
might
be
advantage
type
well,.
C
A
K
K
A
J
A
D
C
Not
empty
yeah
yeah,
the
only
question
is
no
ability.
What
what
is?
What
does
suppress
message
do
for
the
default
reason,
if
anything
actually.
F
A
C
J
H
A
A
A
C
F
C
M
M
A
M
D
M
F
C
C
D
F
C
M
Gotten
the
answer
from
try
get
value
and
then
now
you
have
to
tell
the
dictionary
you
want
to
add
it
back.
So
it's
gonna
go
yeah
for
trying
it
out
if
it
was
an
extension,
it's
a
real
method.
You
can
do
it
better,
but
these
are
extensions,
so
I
mean
you
can
like.
What
do
you
want?
Presumably
what
he
wants
us
to
avoid
invoking
be
Hong
Kong
if
the
values
present?
Yes,
it
is
one
line
as
as
an
overload
to
the
existing
value.
M
F
M
A
M
D
A
C
D
M
L
A
A
M
M
Next
version
of
the
book,
I
wrote
a
thing
on
some
breakage
changes
and
one
of
the
things
that
I've
said
that
when
we
talk
about
things
like
this
in
the
past
is,
if
you
already
would
have
thrown
for
null.
We
don't
care
that
you
would
break
over
a
low-resolution
with
an
old
rule,
because
if
it
was
non-working
code,
but
if
null
weren't,
you
can't
make
a
little
over
and
it
all
works
here.
So
that
means
that
our
general
stance
on
this
would
be
that
that
is
an
unacceptable
over
by.
F
D
M
F
F
Do
is
implied
for
by
you
know
that
the
only
problem
that
we
have
with
this
API
is
that
it's
breaking
the
person
does
a
lot
of
work,
reads
and
yet
proposal,
and
then
so.
We
should
also
maybe
provide
more
informations
like
other
persons
that
we
have
is.
Is
it
even
worth
it?
It's
just.
A
neat
statement
could
be
like
write.
Some
justification.
Ye
yeah.
D
A
M
A
M
C
C
F
F
M
M
M
A
F
N
A
C
F
M
Reason
that
I
know
that
something
of
like
collection
or
somebody
having
a
weak
King
versus
a
strong
thing
came
up
is
because
it
was
on
a
static
or
was
hit
its
static
to
friends
safe
and
there
was
a
static,
add
and
a
static
read-only
collection,
and
it
was
not
thread
safe
and
you
can
kill
the
process.
Yeah.
H
M
M
Like
okay,
so
there
are
multiple
definitions
of
him,
one
of
which
is
RFC
74-68,
which
removes
all
of
the
complicated
parts
from
them
and
only
leaves
the
parts
that
people
think
and
has
so
pim
supports.
The
thing
him
supports
attributes
which
no
one
understands
which
is
at
the
end
of
the
begin
line
you
can
have
a
new
one
in
and
you
can
have
what
look
like
HTTP,
headers
and
part
of
why,
when
I
added
crypto,
key
I'll
support,
I
didn't
want
to
add
PIM
is
because,
frankly,
it
was
kind
of
a
I.
A
K
M
This
supports
RFC's,
seven,
you
version
which
is
killing
without
attributes,
and
so
then
the
the
API
essentially
is
so
the
try
find.
Is
you
give
it
a
rush
chart?
It
finds
the
first
valid
begin
in
pair
in
it
outs
the
portion
of
the
data
that
contained
when
a
64
data
and
produces
what
the
label
one
so,
whether
it's
a
certificate
or
private
key
or
whatever.
So
it
pulls
that
back
as
data
elements
so
yeah.
M
So
you
get
the
base64
data
the
label
and
then
because
it
already
had
to
do
all
of
the
work
of
seeing
how
many
base
64
bytes
were
there
and
where
the
whitespace
was
and
all
that
stuff.
While
looking
for
the
in
header,
it's
the
like,
it
can
also
tell
you
like
this
is
how
big
the
base64
decode
is
going
to
be
so.
F
F
F
M
F
N
M
H
F
A
You
get
the
range
and
then
to
you,
you
have
your
original
input
and
you
do
location,
dot
and
dot
inside
brackets.
That's
how
you
slice
the
younger
so.
A
Not
really
publicly
exposed
so
say
that
we
have
that
right.
So
what
you
would
end
up
doing
is,
you
would
say,
read:
read
only
span
char
input
equals
something.
Try
find
input,
blah
blah
he'll
range
range.
If
we'd
only
spend
remainder
input
range
and
there
are
you
food
yep
that
works
just
fine,
so
I
mean
that's.
Why
I'm
is
inclusive
exclusive.
D
M
M
M
M
K
M
K
F
E
M
J
M
M
F
N
P
A
F
A
F
A
M
Already
a
binary
encoding
if
you're,
if
you're
reading
from
the
wire
just
under
his
mind,
like
don't
ya,
don't
bother
with
basics
before
okay.
Yes,
we
wanted
to
say
that
we
cared
about
chart
eight
for
in
the
future.
Then
I
think
one
of
the
earlier
proposals,
I
had
was
try
find
instead
of
returning
instead
of
having
the
struct
adjust
out
in
all
the
stuff.
M
M
F
F
F
F
A
M
A
A
A
So
it
doesn't
simplify
things
to
get
rid
of
come
feels
entirely
and
just
say.
Like
here's,
that's
was
mentioned
earlier:
here's
where
the
label
is
here's
worthy
base64-encoded
Eidos
and
leave
it
to
someone
else
like
the
convert
class.
To
do
all
this
formula
basically
make
this
as
dumb
as
possible.
Just
think.
It's
literally
just
a
parcel
like
the
data
begins
here,
ends
here
so.
M
Out
three
ranges
in
didn't:
just
three
ranges:
well,
actually,
one
spike,
because
they,
it
was
how
big,
because
it
already
had
to
go
through
all
the
like.
What
was
what
stuff
in
here
was
new
lines
versus
state
transitions
and,
like
it
already
knows
how
many
semantic
characters
there
were
in
the
base.
64
so
know
that
the
pan
64
DS.
E
F
A
M
Yeah
my
my
original
was
like
try.
Fine
or
my
original
counterproposal
was
try.
Fine
next
field
data
comes
in
and
label
contents
contents.
The
cloak
decode
link
that
chars
read
comes
out
a
charge.
Bread
was
actually
ambiguous
because
well
I
told
you
how
far
you
need
to
slice
for
the
next
one.
I
didn't
tell
you
where
the
data
actually
started,
because
you
it'll
this
will
skip
any
non.
It
doesn't
require
that
it's
starting
at
the
first
time
ever
so
I
did.
F
M
Developers
twenty
humans
in
the
home
so
they're,
like
the
reason
that
I
now
have
the
link.
Do
we
want
to
start
with
so
for
one
I
guess
the
structure
we
could
make
the
structure
support
char
eight
in
the
future,
by
making
it
not
being
spans
and
making
it
arrangements
like
a
good
return
ranges
that
expands
and
then
we
could
use
it
for
then
we
could
use
a
guitar
and
I
have
less
it'll
grow
and
then
it
doesn't
need
to
be
a
restaurant.
A
M
M
M
N
A
M
So
I
took
your
suggestion,
the
thing
out
and
you
you
had
the
suggestion
to
counter
the
fact
that
we
need
four
of
everything
because
of
memory
and
span
and
redolent
states
and-
and
I
christophe
hasn't
finished
doing
that
chapter
yet,
but
I
actually
I
wrote
that
into
the
next
set
of
guidelines.
You
know
and-
and
he
could
have
also
argued
on
email
and
hidden,
be
so
like
we
haven't.
A
Yeah
Oh
other
piece
of
feedback,
I
would
probably
from
find
I
would
probably
return
the
range
and
help
the
fields,
because
normally
the
final
L
parameter
is
the
thing
that
gets
returned
from
an
oven.
A
M
But
that's
that
we
could
like,
if
you
look
at
which
data
is
someone
wants
when
they
call
this
in
which
data
is
the
extra
data
yeah,
the
location
is
the
extra
data,
that's
what
they
don't
care
about
it.
So
it
makes
more
sense
to
to
me
to
have
the
data
be
once
and
the
location
be
that
what
you
disagree.
Why
are
we
returning.
D
A
M
A
K
A
G
G
So
basically,
we've
got
these
two
overloads
on
crank
click
create
link,
token
source,
but
it's
pretty
common.
Actually,
in
fact,
maybe
even
the
most
common
usage
is.
You
want
to
create
a
new
cancellation,
token
source
that
you
want
to
be
able
to
cancel
yourself
and
have
cancelled
when
some
other
token
so
like
if
you're,
if
you're
implementing
an
API-
and
you
accept
the
cancellation
token
but
then
also
internally,
you
want
to
be
able
to
force
cancellation
to
occur.
G
You
want
to
basically
call
crate
link
token
source
with
a
single
token,
but
we
don't
have
an
overload
that
takes
a
single
token.
So
you
end
up
kind
of
implicitly
calling
the
params
overload
interest
up
and
you.
So
if
people
know
this,
they
end
up
calling
the
first
overload
with
the
actual
token
and
then
either
defaults
or
cancellation
token
dot,
none
I'm,
suggesting
we
just
add
one
that
takes
the
single
token.
You
know
we
already
have
it.
We
just
need
to
flip
it
public.
A
M
M
M
G
M
G
No,
that's
you.
You
often
all
really
like
the
whole
purpose
of
this
new
overload
is
because
you
generally,
you
often
are
going
to
call
cancel
on.
Are
you
because
I
think
every
time
I
wanted
that
it
I
just
the
only
and
the
only
reason
to
use
this
new
overload
is
if
you're
gonna
call
cancel
on
the
single
yeah
a.
M
G
Token
right
it's
the
case
of
you're
implementing
some
API.
Let's
say:
let's
say
you
want
to
implement
an
API
that
accepts
the
cancellation
token
and
you
want
to
have
some
internal
timeout,
so
you
accept
a
cancellation
token
from
elsewhere,
and
then
you
know
internally.
You
call
cancel
on
your
thing
when
there's
a
timeout
or
whatever,
so
you
you
create
a
cancellation
token
source
that
you
can
yourself
cancel,
but
you
also
want
to
be
cancelled
when
the
external
token
is
cancelled
right.
M
I
guess
what
I've
always
done
for
that
for
the
timeout,
but
again
we
have
a
discussion.
Christoph
wants
to
have
of
should
timeouts
use
cancellation,
token
timeout,
or
should
we
use
timeout
exception,
which
is
those
are
two
of
the
two
ways.
We've
told
people
to
do,
timeouts
that
if
I
wanted
to
do
the
like
link
this
weather,
it's
okay
and
I'm
like
cancel
yourself
after
30
seconds,
then
I
don't
care
to
call
cancel
I'm,
letting
it
handle
it.
For
me,
I
just
I've
always
just
wanted
to
talk
things,
but
this
is
the
way
it
is.
A
To
expect
so
I
have
a
doubles.
Advocate
question
Steve
yeah,
remember
how
the
language
team
was
considering
having
like
params
read
over
spend
of
T
yeah.
That's
probably
gonna
happen
for
well.
Hopefully
that
happens
perceived
nine
but
yeah.
If
we
had
that
word,
we
just
rely
on
that.
Instead
of
adding
a
single
parameter
overload-
or
would
we
add
this
anyway,
just
for
convenience,
I.
G
Well,
no
okay,
I
think
I.
Think
the
logic
was
the
spans
ones
were
gonna
take
precedence,
and
if
you
were
adding
a
span
overload,
then
it's
your
accepting
the
fact
that
people
might
find
the
new
one
instead,
even
if
you
intended
for
it
to
be
more
efficient,
I
guess
you
intended
for
you
more
efficient
I
mean.
If,
if
that
happened,
you
know
protential
ii,
we
would
have
a
params
read-only
span.
Cancellation
token.
G
G
G
Thing
I'm
not
entirely
sure
about
in
the
fullness
of
time.
Levi.
We
probably
would
what
I'm
not
sure
about
is
exactly
what
code
the
compiler
is
going
to
generate
for
the
call
sites.
The
most
recent
proposal,
I
heard,
was
that
initially
it
would
just
allocate
an
array,
but
it
would
do
so
in
a
way
that
it
might
be
able
to
avoid,
for
example
like
if
you
made
the
call
inside
a
loop,
it
could
hoist
the
array
allocation
to
outside
of
the
loop
I.
J
G
Buying
things,
so
it
doesn't
really
matter
yeah
and
Mike.
I'm
sort
of
assuming,
like
I've
I,
can't
remember
the
last
time,
I've
needed
more
than
two
arguments
to
create
link
into
open
source
like
the
by
far
the
most
common
cases
are
one
or
two
so
I,
don't
even
know
if
we'll
want
to
add
the
params
span.
But
what
we
can
see
when
you
know
when
that
happens,.
M
M
M
G
G
M
Right
look
is
the
one
that's
built
in
based
on
the
timer.
Would
not
what
we
can
just
automatically
do,
because
it's
just
things
where
I've
seen
people
where
they're
doing
like
Oh
HTP
client
I,
like
I,
got
a
cancel,
which,
why
did
it
cancel?
Is
it
canceled,
because
someone
said
cancel
it
didn't
cancel
because
we
gonna
type
out
right.
G
Basically
be
timeout,
exception
derives
from
operation
cancel
of
exception
and,
if
a
cancellation
token
source
is
cancelled
because
of
a
cancel
after
or
having
a
time
up
as
to
its
constructor,
then
we
would
sort
of
do
the
right
thing
and
throw
the
timeout
exception.
Instead,
technically
it
could
break.
You
know,
order
of
cache
blocks
and
things
like
that
yeah,
but
that's
sort
of
how
I
would
ideally
merge
the
two
thing
to.