►
From YouTube: WG API Expression Bi-Weekly Meeting for 20200414
Description
WG API Expression Bi-Weekly Meeting for 20200414
A
A
B
A
C
D
A
A
D
E
A
F
D
F
A
A
Can
share
with
you
can
share
the
folder
with
it
with
a
group.
That's
everybody
in
so
if
you
are
create
a
document
in
the
folder,
it's
still
owned
by
me
and
you
can
transfer
ownership
that
easily.
So
that's
where
the
shared
drives
come
in,
because
the
shared
drive
owns
the
objects
inside
the
people,
so
they
send
Paris
and
a
message
because
I
read
something
about
a
she's
moved
for
Jacobo
news
team.
C
F
A
B
A
C
C
B
A
G
Because
you
had
some
parallel,
yeah
I
mean
there's
very
few
time
slots
in
the
week
that
are
your
friend
and
so
there's
necessarily
going
to
be
conflict.
This
just
happens
to
come
exactly
with
another
meeting
that
I
actually
generally
have
to
go
to.
Ok,
so
I'll
just
have
to
do
my
best
to
juggle
and
split
time.
Yeah.
A
G
G
It
becomes
the
scheduling
problem
and
we
should
write
a
scheduler.
F
G
C
H
H
A
C
C
B
Can
you
attach
the
meeting
notes
links
well,
so
I
can
access
it.
Maybe
I.
D
B
I
C
All
right
so,
since
we
have
an
adjusted
document
now
and
people
with
limited
time
value
effort
to
be
here,
do
Tim
and/or
Mike
have
particular
things
that
you
think
we
should
discuss
sooner
rather
than
later,
and
the
goal,
or
at
least
one
of
my
goals
for
this
group,
is
that
we
should
actually
do
things
and
not
just
talk
about
them.
So.
C
C
G
That's
fair,
I
also
think
part
of
the
responsibility
of
these
working
groups
is
to
be
aspirational
and,
if
we're
only
defining
things
that
we
have
time
to
work
on
we're
going
to
have
very
lackluster
results,
whereas
you
know
having
some
I,
don't
want
to
say
intellectual,
but
some
sort
of
abstract
conversations
about
shoulds
will
give
us
a
better
course
forward.
Not
that
it
should
be
exclusively
that.
But
if
it's
exclusively
not
that
then
I
feel
like
we
lose
track
of
where
we
actually
want
to
go.
We're
only
looking
as
far
as
our
feet.
G
G
I
believe
so,
except
we
need
to
well.
Let
me
I'll
speak
for
myself
right,
like
I'm,
very
interested
in
API
expression
and
API
machinery,
and
and
as
a
consumer
and
and
frequent
reviewer
of
API
as
I
find
myself
saying,
gosh
I
wish
we
had
this
thing
right.
I
have
very
little
technical
time
to
contribute
right
now
and
for
the
foreseeable
future.
If
that
means
that
I
should
not
attend,
I,
guess
I'm!
Okay
with
that,
but
I've,
that's
not
what
I
was
hoping
to
create
so
I
I.
C
A
A
C
G
G
All
right,
some,
a
few
weeks
ago,
I
guess
a
month
ago,
before
all
of
this
started
happening
I
had
suggested
to
Jordan
that
we
start
with
an
outline
and
actually
go
like
I
opened
up
the
changing
API
doc
and
a
that
conventions
Docs
and
realize.
There's
a
ton
of
overlap
and
there's
no
coordination
between
them
and
actually
the
changing
api's
doc
is
like
two
docs
jammed
into
one
doc,
like
literally
just
cat,
mated,
together
and
the
way
I
would
have
attack.
G
C
H
H
G
Actually,
we
have
two
documents
that
are
both
trying
to
be
both
sure.
Specifically,
we
have
the
API
conventions,
doc,
which
is
20
some
pages,
which
is
updated
fairly
regularly.
That
I,
don't
know
I,
don't
know
when
the
last
time
I
read
it
was
and
not
everybody
agrees
with.
Frankly,
there
are
things
in
there
like
all
optional
fields
should
be
pointers,
and
there
are
other
API,
reviewers,
really
yeah,
I,
only
partially
enforced
that
and
then
there's
the.
G
Maybe
not
the
flowchart,
but
a
more
complete
outline
of
what
do
we
think
all
of
the
interesting
use
cases
are
for
people
who
are
making
API
either
building
new
API
s
or
making
changes
to
the
existing
API,
and
if
we
didn't
capture
sort
of
all
of
the
things
that
we
identify
as
interesting,
weird
corner
cases,
for
example,
pluralizing,
a
field
or
adding
a
new
field
or
how
do
detect.
You
know
immutable
fields.
Those
sorts
of
things
are
all
interesting
use.
G
You
might
be
surprised,
I
mean
there's
a
lot
of
people
who
just
want
to
add
one
field
to
the
name
pod
or
something
like
that,
and
they
don't
see
themselves
as
creating
novel
new
CRD
and
schema
lifecycle.
They
just
want
to
add
that
one
new
field
and
I
don't
think
we
haven't
done
a
really
good
job
of
explaining
to
them
exactly
why
they
have
to
jump
through
some
of
the
hoops
they
have
to
jump.
Yeah.
C
C
Api
is
in
general,
which,
like
not
what
they're
trying
to
do
so
that's
right
and
like
often
it's
some
sort
of
feel
some
sort
of
feature
that
we
really
do
want,
like
the
like
ephemeral
containers
or
pods,
for
example,
and
like
the
amount
of
hoops
these
people
have
to
jump
through
to
get
an
API
changed
in
I.
Think.
G
Ephemeral
containers
is
an
apple
because
it's
so
huge,
but
you
know
take
examples
like
we're,
adding
it
literally
an
informative
field
to
service
call
that
protocol
there's
really
no
implementation
of
it,
and
it's
not
clear
what
the
person
who's
adding
this
field
really
has
to
do
or
why,
except
that
they
happen
to
work
at
Google.
So
they
can
talk
to
me
or
Jordan
or
you.
C
C
G
Hence
my
suggestion,
and
if
we
can
find
an
outline
like
if
we
can
come
together
on
an
outline
I
think
actually,
if
first
draft
of
these
docs
would
be
binding
and
pasting
stuff
out
of
the
original
Doc's
into
the
new
skeleton
and
then
massaging
the
transitions,
and
actually
that
would
could
result
in
a
pretty
good
doc
with
obvious
holes.
Whereas
today
we
have
a
terrible
dock
with
non-obvious
holes.
Yeah.
If
you
look
at
our
like.
C
Website,
documentation,
they've
got
the
they've
got
quite
a
system
of
like
I,
don't
even
know
what
they're
called
but
there's
like
how-to
guides
and
the
concept
guides
and
and
like
tried
to
change
documentation.
You'll
often
get
comments
on
them
from
them
like.
Oh,
this
belongs
in
a
in
a
concept
section,
not
a
hockey
section
or
something
like
that.
I
think
we
need
a
similar
theory
or
or
our
API
evolution
documentation,
so
that
so
that
we
can
take
them
exchanges
at
their
time
without
it.
Turning
into
a
giant
mess.
G
Although
I
think
that
maybe
the
needs
here
are
a
little
bit
less
rigorous,
but
it
also
goes
I.
Think
to
my
point
of
like
a
flow
of
chart,
forces
principles.
I
think
we
have,
we
have
to
figure
out
both
here's
the
principle,
here's
why
you're
being
asked
to
do
something,
and
if
you
don't
care
about
that,
here's
the
steps
you
need
to
follow
to
do
it.
Yeah.
C
G
Ya,
I'm
very
interested
in
it
a
sudden
backlog
last
week.
I
would
have
had
time,
except
for
a
customer
issue
that
ate
up
most
of
the
week
and
so
this
week,
I'm
paying
for
last
week.
But
I
have
my
to-do
list
next
to
me,
and
three
or
four
of
the
items
on
my
to-do
list
are
around
API
documentation
backlog
that
I
owe
for
cigar
ch.
So
if
I
can
get
a
co
victim
here,
a
volunteer
voluntold,
then
I
would
be
willing
to
sit
and
run
through
the
existing
box.
H
So
I'm
interested
too
and
Carly
don't
really
have
time,
but
I'm
probably
gonna
offer
some
I'm
still
trying
to
figure
out
how
Tim
remark
about
and
now
he's
talking
about,
two
things
actually
outline
and
use
cases,
and
so
like
I,
understand
how
to
put
together
an
outline
of
principles.
For
example.
Use
cases
are
another
matter.
G
Yes-
and
maybe
we
really
do
want
to
Doc's,
but
they
did
cross,
did
you
cross-reference
each
other
and
deploy
how
you're
coming
at
the
problem
would
be
which
doc
you
want
to
start
with?
I
think
we
need
a
comprehensive
principles
and
conventions
doc
like
this
is
what
we're
asking
you
to
do
and
why
and
then
the
I
just
want
to
add
a
field.
Tell
me
how
I
can
do
these
common
things
and
cross
linked
to
the
this
is
why
you're
being
asked
to
do
this?
H
I'm
thinking
the
flowcharts
really
err
would
be
oriented
around
use
cases
right,
I
mean
I.
Could
imagine
freedom
in
this
case
degrade
a
flowchart?
Yes,
so
I
think
so
maybe
we're
still
talking
about
the
same
to
basically
doc
kind
of
documents.
I.
G
Let's
I
have
to
go
look
at
the
calendar
and
figure
out
when,
but
let's
see,
we
can
find
time,
maybe
later
this
week
to
just
jump
on
a
call
and
walk
through
the
existing
docks
and
see
if
we
can
extract
the
existing
outlines
like
the
table
of
contents
exists,
and
then
we
can
start
arguing
about
how
they
maybe
should
look.
Okay,.
C
G
C
G
H
G
Okay,
so
there's
a
lot
of
people
now
now
volunteering,
if
we
I
mean
we
could
do
it
in
this
time
slot
it
doesn't
seem
like
it's
gonna.
Be
that
effective.
If
10
people
are
doing
it
all
in
tandem,
but
maybe
we
can
lay
out
some
tasks
and
have
everybody
go
off
and
do
sort
of
an
independent
version
and
then
merge
them.
F
H
C
G
C
C
C
C
H
I
also
agree
with
Tim's
thing:
we
should
you
know
kind
of
talk
about
aspirations,
even
if
we
don't
schedule
work
on
them
immediately.
I
find
the
questions.
You
know
this
first
topic
here.
You
know
a
desired
state
for
the
scheme
of
features.
I
think
that
is
an
interesting
thing
to
discuss.
I
think
this.
The
second
one
kind
of
is
about
the
well.
A
H
That's
I
I
have
less
to
say
about
that,
but
it
sounds
like
an
interesting
topic.
I,
certainly
like
the
idea
of
tools
that
can
help
with
this
stuff,
so
I
find
a
third
interesting.
A
third
and
fourth
really
I,
also
agree
finishing
up.
Ssa
I
mean
if
there
are
problems
with
SSA.
You
know
we
definitely
take
care
of
that.
H
G
C
C
H
The
last
bullet
here
I,
don't
know
if
what
exactly
what's
meant
by
that
it
sounds
like
it's:
either
overlaps
with
the
first
bullet
or
the
second
boat.
Well,
we
just
talked
about
the
docs
or
kind
of
stuff
I've
been
working
on,
which
is
the
you
know.
The
developer
needs
to
to
read
and
understand
yeah.
A
H
See,
okay,
so
we've
been
talking
about
the
issues
of
what
a
API
developer,
who
started
talking
about
the
problems
that
an
API
developer
has
both
creating
and
updating
an
API
there's
another
collection
of
issues
which
is
a
developer
of
a
client.
He
just
wants
to
consume
an
API.
The
listing
documentation
for
a
developer
is
actually
pretty
bad.
If
you
try
to
write
a
controller,
all
the
things
you
have
to
understand,
you
go
to
look
at
the
documentation
for
the
pieces.
You
use
to
write
a
controller.
H
C
H
F
C
C
And
like
I
feel
like
API,
you
know
what
sort
of
use
cases
Sita,
what
sorts
of
use
cases
does
our
API
Tila
intend
to
support
is
useful
to
both
API
authors
and
API
users,
but
it
doesn't
like
that
was
enough
information
to
go
off
and
write
a
controller
I
think
that
would
be
a
separate
right
like
there's
there's
a
lot
of
details
that
wouldn't
necessarily
be
in
that
kind
of
document.
Yeah.
B
B
A
Looking
at
the
the
point
about,
do
we
want
to
keep
types
code?
How
do
we
publish
api's?
That's
something
where
you
have
to
take
in
mind
how
clients
use
it,
because
it's
it's
useless
to
expose
something
it's
unusable.
Like
types
go,
I
we
had
people
who
didn't
want
to
use
our
types
go
but
build
custom
ones,
because
it's
a
big
group
of
repository,
so
I
think
it's
it's
at
least
at
some
point,
part
of
let's
go
yeah.
C
C
H
I
agree:
we
have
to
draw
a
boundary
somewhere,
I
guess
I'm.
Okay,
with
you
know,
saying,
the
problems
of
clients
are
mainly
outside,
but
you
know,
as
it
was
also
said.
You
know.
The
whole
point
of
this
is
to
support
the
clients.
You
know
both
of
the
users
and
the
controller
authors,
but
you
know
we
have
to
bear
their
problems
in
mind.
I
think.
C
Like
understanding
use
cases
or
the
other
point,
the
sets
of
operations
that
this
API
style
makes
easy
and
the
said
that
it
makes
hard
like
understanding
that
and
therefore
which
clients
it's
appropriate
for
that
seems
mandatory
I,
don't
see
how
we
can
succeed.
If
we
don't
do
that
step,
but
hand-holding
clients
through
using
it
successfully,
is
maybe
not
the
goal,
at
least
at
this
group.
Well,.
H
For
for
AP
acts,
we
can
focus
on
the
API
expression
issues
without
having
to
worry
about
all
the
details
of
documenting
stuff
for
clients,
I
think
we're
green
I.
You
know
we
have
to
keep
the
problems
of
clients
in
mind.
That's
the
whole
reason
this
stuff
exists.
It's
the
only
reason
this
stuff
exists
is
because
their
clients.
C
A
H
C
C
H
A
A
So
we
said
that
we
collect
our
goals
for
119
and
that's
mostly
making
chip
CTO
suicide
apply
default
and
yeah.
The
one
part
is
this:
concise
a
suicide
apart
transition
I
think
you
had
the
discussion
about
if
we
can
do
it
because
of
the
breaking
changes,
I'm,
not
sure
what
the
resolving
this
was,
because
I
found
that
we
have
several
issues
regarding
this
yeah.
H
C
H
H
H
B
It
is
sitting
to
control
right
now,
but
that's
really
confusing
because
you
may
have
one
of
the
problem
we
have
is
that
guy
inside
of
my
conflicts,
we
saw
side
apply
if
you're
setting
an
annotation,
whiskey
barrel.
It's
always
give
girl.
So
Julian
is
working
on
changing
that,
so
that
each
cup
girl
come
in
has
its
own
manager.
C
H
C
C
The
issue
with
making
service
I
define
default
is
to
make
sure
is
use
default
that
that
there's,
many
of
the
current
expectation
doesn't
can
yeah
I'm,
actually,
okay,
with
changing
the
expectations
somewhat,
because
we
are
adding
worth
more
features
like
I
know.
In
some
cases
people
are
going
to
see
a
conflict
when
they
switch
over
to
server
side
apply,
but
I
think
that's
actually
useful
for
conflict,
because
as
long
as
we
do
it
correctly
right,
like
you,
shouldn't,
get
a
conflict.
C
If
you
didn't
change
anything
or
only
change
the
things
that
you
were
managing
the
client
side
applied,
we
should
try
and
avoid
giving
contact
conflicts
in
that
situation.
But
if
you're
getting
a
conflict
because,
like
HP,
a
changed,
the
replicas
feel
or
you
R
and
some
other
cute
control
command-
that,
like
change,
your
image
or
something
and
when
you
switch
over
to
server
side
apply
for
the
first
time
you
get
a
conflict
on
that
and
I
think
that's
fine
and
actually
useful,
because
it's
telling
you
something
about
your
system
that
you
may
not
have
known.
C
B
J
C
F
C
F
B
Yeah
absolutely
I
I
talked
to
Julian
like
twice
this
week
or
last
week,
so
hopefully
he's
not
not
doing
anything.
B
B
C
E
C
C
F
J
E
C
C
F
C
C
B
Okay,
okay,
I
think
I.
Remember
why
I
was
not
super
leering
with
that.
Yes,
it
should
be
in
the
types
that
we
should
do
exactly
what
CL
is
out
to
me,
but
what
I
didn't
want
was
to
have
the
field
in
the
types
go
and,
and
then
not
defaulting
with
these,
by
reading
these
and
actually
defaulting
by
reading
these,
because
I
suspect
someone
is
going
to
come
up
and
add
these
fields,
someone
expect
the
right
thing
to
happen
and
I.
C
C
Yeah
I
think
there's
two
positive
two
possible
things:
we
could
you
automatically
omitting
defaulting
functions
is
definitely
one
of
them.
The
other
one
is
emitting
a
test
that
verifies
that
the
existing
default
defaulting
function
actually
fulfills
the
contract
specified
in
the
API.
Now
we
might
want
that
either
way
like.
If
somebody
puts
in
a
major
with
defaulting
functions,
we
should
force
that
it
actually
just
what
the
API
claims
it
does.
I.
B
C
Not
necessarily
because
our
existing,
like
especially
the
conversion
functions
we
generate
default,
but
then
you
can
implement
them
manually.
However,
you
want
to
and
that
suppresses
the
generation
of
the
default
function
or
the
conversion
option.
So
if
this
works
similarly
like
we
generated
defaulting
function,
but
then,
if
you
write
your
own
hand
built
one,
it
doesn't
necessarily
VOC
that
one.
B
F
B
H
B
C
C
We
don't
have
that
problem,
because
this
is
specifically
about
key
fields.
Defaults
for
value
fields.
Server-Side
apply.
Has
it
doesn't
really
care
much
about,
but
for
key
fields?
It
needs
those
to
be
set
to
even
know
which
item
in
the
list
you're
talking
about.
So
those
do.
You
have
to
be
sent
statically
and
I.
Think
that's
a
reasonable
requirement
that
key
field
T
field
default
should
be
expressible.
Statically.
H
C
D
C
Think
it
is
a
good
constraint
and
I
do
think
we
want
to
like
anything
that
is
expressed
with
people
statically
any
default,
it
is
accessible,
statically
I
think
should
be
expressible
statically
in
the
schema,
which
means
that
we
can
automatically
put
it
in
API,
documentation
and
stuff
like
that.
So
I
think
that
that's
just
a
good
thing.
H
B
C
B
A
B
A
And
I
try
to
look
more
into
the
the
problem
I
had
with
the
CRTs
and
haven't
made
much
progress
on
it,
because
it's
it's
weird
I,
don't
know,
what's
happening,
I,
actually
try
more
I
still
and
thinking
a
lot
about
different
ways
to
solve,
based
in
general,
it's
comparable
to
to
what
we
discussed
right
now,
because
I
would
expect
like
that
to
be
represented
in
a
schema
that
a
field
doesn't
get
persisted.
If
you
change
it,
but
it's
hard
to
express.
A
But
what
I
thought
about
would
be
an
option
to
solve
in
the
short
term
would
be,
instead
of
having
only
the
strategy
to
define
a
custom,
json
schema
that
could
generate
the
resetting
and
the
field
manager
stuff
and
put
it
there
and
run
a
cogeneration
rate
which
could
make
it
easier
and
testable
because
it's
both
generated
both
from
the
same
source.
But
that's
just
a
fall
I
had
over
the
weekend,
so
I'm
not
sure.
B
B
In
the
least
because
the
apply
status
first
in
the
list,
because
it's
it's
really
changing
the
behavior
of
our
people
to
like
they
are
not
going
to
see
the
same
yes,
sir,
before
and
I'm
sure,
a
lot
of
people
are
going
to
freak
out
if
they're,
if
they
do,
is
that
the
it
seems
more
easy
to
do.
Yeah
yeah.
A
So
it
might
even
seem
like
a
I,
could
first
starter
once
the
cap
is
done
it.
It
might
be
something
quite
easy
to
do.
For
somebody
says
yeah
desert,
you
have
cool
already
anyway,
no
I
didn't
we
discussed
how
to
do
it
and
I'm,
not
sure
if
there
is
code
in
the
cab,
but
I
didn't
think
so
we
discussed
only
with
reduce
you
mentioned
doing
with
with
generation,
and
we
have
the
old
and
new
objects
at
some
point.
So
we
could.
B
G
A
I
B
E
E
Doesn't
really
make
a
statement
about
that
this
is.
This
would
be
necessary,
no
matter
how
we
do
the
defaulting
and
it
doesn't,
it
doesn't
give
people
that
chose
to
accidentally
think
their
default
thing,
since
we
haven't
added
the
type
psycho
thing
or
the
open,
API
thing:
it's
just
the
SSA
schema
change.
E
B
B
B
The
one
one
we
compare
to
make
sure
it's
the
same
yeah,
it's
it's
not
a
good
experience
right.
You
change
the
field
and
all
of
the
sudden
it
tells
you
hey.
This
doesn't
work.
You
have
to
do
something
else,
I
hate
it.
When
the
tools
tell
the
tools,
hey,
you
actually
thing
else,
you're
like.
Why
don't
you
do
it
for
me,
and
in
this
case
we
could
just
say
if
you
have
the
slash
slash
default
on
top
of
a
field,
this
is
going
to
take
precedence
of
the
code
yeah
or
you
could.
A
Deliver
slash,
slash
default
stuff,
but
on
occasions
where
it's
implemented
in
it
with
custom
code,
you
don't
do
it
and
you
fail
for
those
and
people
should
not
I
say
it
shouldn't
be
common
to
you
to
override
default
in
like
custom,
stuff,
I,
don't
know
if
it
if
that's
wrong
assumption.
But,
yes,
we
have
both
worlds
and
it
would
be
working
and.