►
From YouTube: GraphQL.js Working Group - 2022-06-29
Description
GraphQL is a query language for APIs and a runtime for fulfilling those queries with your existing data. GraphQL provides a complete and understandable description of the data in your API, gives clients the power to ask for exactly what they need and nothing more, makes it easier to evolve APIs over time, and enables powerful developer tools. Get Started Here: https://graphql.org/
B
I
have
a
new
working
setup
and
it's
new,
the
railroad,
but
it's
too
hot
at
my
my
place
to
close
the
window
so
yeah
we
train
noise.
I
will
call
the
window
temporarily
so
sorry
about
that
in
advance.
B
Yeah,
it's
quite
hard,
then
yeah.
It's
like
ukraine.
You
have
very
interesting
climate.
We
have
temperature
from
minus
20
to
plus,
like
40
minus,
not
every
winter,
but
one
yeah
sometime,
usually
it's
like
at
least
minus
10,
minus
15,
every
winter
and
like
past
14
this
summer,
so
a
bunch
of
different
calls
from
winter
to
like
summer.
B
There
is
no
like
yeah.
I
think
we
can
start
with
official
part
and
yaakov
will
join,
but
we
will
save
more
time
for
like
discussion,
yeah
yeah
carl
said
he
will
join
in
like
10-15.
B
Okay,
because
actually
wait.
B
I
did
some
some
thinking
about
his
second
topic
item.
I
will
send
you
in
in
the
chat
my
proposal,
it's
like
so
my
issue
and
also
so
like
troubles
that
yaco
have
with
refactoring
visitor
and
just
finish
like
that
comment
and
since
we'll
probably
discuss
it
as
one
as
possible
solution
would
be
great
here.
B
Agenda
and
yet
today
is
29th
19th
first
one
yeah,
so
let's
send
the
link
in
a
chat
just
in
case
somebody
else
will
join.
So,
let's
start
with
agenda
and
yeah
as
usual.
You
need
to
sign
documents
if
you're
watching
on
youtube.
Obviously
it's
too
fast
and
we
both
signed
it,
but
if
you're
watching
on
youtube
and
want
to
join,
please
feel
free.
It's
very
easy.
You
can
add
yourself
to
agenda
items
agenda.
B
B
It's
not,
you
know,
don't
need
to
involve
your
company
or
anything
so
in
introduction
of
attendees
hi.
My
name
is
one
jakob
will
join
in
10
minutes
and
we
use
a
hatch.
B
I
think
it
last
time
we
actually
have
like
almost
full
screen
or
like
one
before
a
previous
one.
So
we
have
it
attendance.
I
think,
like
it's
some
so
people
almost
most
of
the
people
have
yeah.
B
So
bus
adjustments
like
the
cc
and
stuff
yeah.
B
Yeah
yeah,
but
it's
only
a
way
to
to
specify
where
you
want
to
like,
not
know
ability
assertion
outside
the
right,
alex
initially
wanted
to
do
it
without
without
without
wet
syntax,
and
just
like
say
you
cannot
put
on
no
no
assertion
inside
the
race
only
on
array
as
a
whole,
but
because
deal
breaker
for
for
some
people,
yeah,
and
I
have
a
train
coming
up.
Oh
yeah,
so
that's
why
alex
said
that
I
don't
think
anyone
particularly
like
that
syntax,
but
that's
the
only
mechanism.
B
We
have
passes
future
proof
in
future
if
we
create
something
else
beyond
exclamation
mark
same
with,
like
question
mark,
you
need
to
be
able
to
specify
where
you
want
a
reaction
on
non-knowability
inside
the
rail
outside
the
array,
so
yeah
so
with
syntax
goten,
but
it's
only
syntax
is
not
at
the
experimental
facts,
so
we
can
break
it
right.
Next
release
between
releases,
it's
experimental,
you
you
get
with
ice
g
notes
only
only
if
you
enable
like
experimental
files.
So
it's
right.
There
is
a
warning
cover
everywhere,
saying
it's
it's
experimental.
B
So
a
merger
can
change
it.
Yeah,
it's
still
stage
one!
So
technically
we
must
at
this
experiment
it's
not
like
even
stage
two
meaning
quite
its
shape
up.
B
It's
like
I
alex
wanted
to
experiment
with
it,
so
I
edited
it
basically
like
maybe
no
background,
and
we
have
time
to
discuss
it
in
your
way
before
yak
will
join
alex
main
reason
why
alex
needed
in
the
core,
because
yop
use,
apo,
ios
and
apollo
ios
used
graphql.js
for
code
generation
and
he
convinced
maintainer
of
oppo
ios
to
actually
use
this
feature
even
as
experimental,
so
he
needed
to
craft
kjs
to
like,
and
he
also
wanted
to
contribute
something
to
guild
graphql
kojon.
B
B
I
think
it
it
will
come
so
like
I
had
the
proposal
of,
I
don't
like
receptionality,
because
previous
version
people
just
said
like,
if
you
specify
stream
in
default
server,
can
ignore
it
and
it's
weird:
it's
a
directive.
You
specify
it
why
it's
ignored,
so
I
propose
this
adding
increment
incremental
to
power
field,
and
I
think
people
agreed
on
that,
but
we
didn't
discuss
it
on
my
working
group.
Yet
I
I
like,
I
came
up
with
idea
during
working
group,
but
I
don't
think
like
people,
we
had
like
a
real
discussion
for
it.
B
It
looked
like
everybody
agreed
and
we
ironed
in
final
finalizing
like
stuff.
Finally,
and.
A
B
I
I
like
okay,
I
missed
what
I'm
36
59.
B
Oh,
I
opened
him
asking
yeah.
He
calls
it
himself
because
he
have
like
commit
taxes
go
what
what's,
probably
because
he
had
a
bunch
of
stuck
there
and
two
previous
pairs
matched
so
now
here
down
to
only
one
pr.
So
I
think
he
he
just
like
did
the
fresh
start
on
that
one
yeah.
But
technically,
I
think
it's
two
stage
one
so,
okay.
B
One
can
one
thing
why
we
previously
didn't
want
like
two
things
that
we
previously
didn't
want
too
much.
First,
is
that
after
much
execute
will
start
returning,
I
think
it
readable
and
it's
like
breaking
again
yeah
and
second
thing:
it's
like
we
discussed
that.
Not
all
all
servers
are
ready
to
serve
in
incremental
delivery,
but
with
like
with
topo
proposal
of
incremental
field,
it's
actually
solved
so
this
issue,
I
think,
like
we
can
figure
out
how
to
do
that.
B
Right
now
me
and
jakob
kayakov,
by
the
way
like
we're
discussing
with
sahaj,
we
didn't
want
to
start
agenda
discussion
without
you,
so
we
waited
and
in
meantime
we
discussed
rope
pr
about
stream.
Rope
swear
about
streaming
g4
and,
if
it's
possible
to
merge
it
on
the
experimental
flag
same
as
coin
control,
know
ability-
and
I
was
saying
quite
it's
possible-
the
only
thing-
the
big
like
thing
about
it.
It's
like
change
of
return,
type.
B
So
it's
like
separate
discussion
and
we
can
either
discuss
it
at
the
end
of
this
call
or
like
asynchronously,
just
wanted
to
explain
what
we
discussing
before
we
joined.
B
C
Sorry
and
sorry
I
was
late.
B
Yeah
so
let's
start
with
agenda
items
and
I
actually
watched
agenda.
B
I
will
open
it
so
yeah.
The
first
agenda
item
is
instance
of,
and
you
edited
yakov
so
like
microphone
to
you.
C
Sorry
about
that,
so
basically,
I
think
there
were
a
couple
of
reasons.
The
title
is
a
little
misleading.
The
title
is
like
the
future
of
instance
of
or
something
like
that
or
direction
of
it,
but
the
question,
but
that
sort
of
gets
it
backwards
like
what's
wrong
with
instance
of
so
I
think,
offline
or
asynchronously,
I'm
not
sure
how
much
was
published.
We
we
but
ivan.
We
were
discussing
it.
C
It
seems
like
the
most
important
issue
that
instanceof
is
connected
to
is
is
the
is
the
is
the
currently
the
future
in
terms
of
whether
we're
going
to
be
supporting
esm
and
and
and
com.js
formats
going
forward
and
how
it
would
be
possible
to
have
like
interoperab
interoperability
between
the
libraries.
I
think
that's
what
we
correct
me
if
I'm
wrong,
I
think
that's
what
we
sort
of
focused
on
that
and
then.
C
Blocker
because
well
because
there's
then
there
would
be
like
dual
packages.
I
mean
there'd,
be
two
versions
potentially
of
graphql,
not
necessarily,
I
should
say
two
two
instances,
not
necessarily
different
versions,
but
two
instances,
one
as
is
in
a
c
js
and
the
one
and
esm
and
and
and
the
instance
of
checks
would
fail.
C
C
I
think
like
the
topic,
so
the
topic
could
maybe
be
renamed,
because
we
only
need
to
decide
whether
to
move
ahead
with
symbols.
If
we
have
a
a
thought
as
to
what's
going
on
with
esm
and
cjs,
at
least
that's
my
understanding,
so
I'm
not
sure
whether
we
want
to
start
the
discussion
with
that.
C
In
terms
of
of
how
things
are
looking,
I
think
if
I
recalled
our
last
meeting
or
maybe
was
that
one
two
times
ago
we
were,
we
were
thinking
of
of
going
ahead
with
esm,
only
support,
and
basically,
if,
if,
if
that's
the
case-
and
we
got
some
support,
I
think
from
an
interested
commenter,
the
author
of
graphql
upload
jaden,
I'm
sorry.
C
If
I'm
pronouncing
that
correctly
yeah,
you
know
it
was
very
pro
going
esm
only
so
if
we're
going
usm
only
there's
really
not
much
to
worry
about.
I
I
think,
and
that's
where
I
am,
but
if
but
if
we're
trying
to
support
both,
then
maybe
we
would
want
to
introduce
symbols.
That's.
A
C
How
I
understand
the
discussion
that
was
held
asynchronously
up
to
this
point.
B
Yeah,
so
to
explain
my
point:
it's
like
a
long
grinding
issue,
and
I
like
I
was
in
the
moment
where
the
that
I
was
just
starting
to
contribute
where,
when
we
added
instance
of
check
at
a
project
like
our
custom
instance
of
check
and
our
custom
instance
of
check
is
problematic,
even
after
even
after
or
an
added
like
a
global
thing,
yeah,
it's
still
problematic
considerable
that
I
expense
a
hatch.
B
I
live
near
the
trend,
trucks
and
it's
super
hot
here,
so
I'm
like
causing
a
window
when
a
train
comes
and
open
afterwards,
so
issue
that
was
solving
here
is
that,
with
instance
of
check,
is
that
our
api
wasn't
designed
to
be
to
be.
B
I
would
say
like
there
is
two
separate
issues
like
three
separate
issues
supporting
like
non-graphql
js
version
of
graphiel.js
interfaces,
so
custom
implementation,
which
is
not
like
inherited
bases
like
need
a
use
case
and
except,
like
executor,
I
didn't
saw
any
other
project
like
creating
stuff
and
I
believe,
like
it
should
be
use
case
based.
So
if
somebody
came
up
with
use
case
for
like
third
party
implementation,
we
need
to
either
address
it
in
graphql,
js
or
say
like
yeah,
it's
valid
use
case.
B
I
don't
think
so.
It's
like
it's
like
you
see
this
visitor
like
right
now,
the
entire
immigration
thing
it's
like
junk,
jungling
juggling
of
stuff
and
with
supporting
between
multiple
version.
We
replace
both
with
knives
and
start
juggling
with
nice,
because
it's
like
it's
so
easy
to
make
mistake,
it's
so
hard
to
think
about
like
compatibility
between
different
version
yeah.
So
for
me
we
can
discuss
solution,
but
I'm
like
strongly
against
supporting
not
the
same
code
base.
So
if
something
is
created
by
the
same.
B
B
It
solves
some
technical
issues,
I'm
like
against,
or
on
opening
foot
gate
and
saying
we
support
like
everything
so
I'm
like,
if
we
can
and
like
do
you
know
thing
is
separate
thing:
we
slowly
move
in
that
direction
and
like
we
need
to
pass
it
over
finish
one,
but
basically
we
cannot
tie
to
version
because
version
is
like
we
release
and
pam
branch
for
people
to
experiment
with
release
like
canaries
countries
have
like
separate
versions,
but
we
release
in
pembranch
and
in
general
ecosystem
people
just
use
whatever
is
main
or
in
like
certain
branches.
B
So
if
solution
basically
say
like
this
schema
created
with
this
hash
and
it's
passed
to
execute,
execute
created
with
with
commit
the
compatible,
if
it's
different
comments,
they're
not
compatible,
we
can,
since
we
do
build
step
anyway,
our
source
code
is
typed
creep,
but
we
need
the
build
step
for
like
number
of
reason.
We
have
like
couple
transformation
and
we
generate
version
js
and
version
tests
and
other
stuff.
So
we
need
the
build
step
anyway
and
during
build
step,
we
can
inject
like
it
hash
in
strings
in
symbols
anywhere.
B
So,
like
my
my
like,
if
it
I
with
compatibility
to
to
commit
hash,
I'm
like
I'm,
okay
with
experimenting
and
comparing
pros
and
cons
if
we
go
beyond
version
of
third
party
implementation,
I'm
like
it's
too
much.
B
To
so
like
use
case
that
not
many
people
reported
like,
I
don't
think
we
actually
have
any
issues
asking
for
third
party
implementation
of
something
or
like
support
between
different
versions.
I
think
people
were
pretty
much
pretty
much.
Okay
with
like
updating,
graphql
expert
defendership
crossroads
to
the
whole
thing.
C
I
think
I
think
that
you
know
all
that
is
is
definitely
fair.
I'm
I'm
just
wondering
meaning,
so
we
basically
whittled
it
down
to
you
know.
Our
use
case
right
now
is
right.
Is
the
esm,
cjs
compatibility.
A
C
So
are
we
are
we
I
mean
I
I'm
definitely
no
expert
in
in
you
know,
making
sure
that
those
are
cross-compatible
and
and
even
in
well
in
anything
really,
but
I'm
just
noting
that
I
think
we
had
decided
tentatively
to
go
esm
only
so
and
then
maybe
we
don't
have
a
use
case
at
all,
but
or
or
did
we
not
really
for
sure
decide
that
you
know
we
may
be
backing
away
from
that?
So
I
mean
that's
part
of
I
think
what
lauren's
pr
is
about.
C
You
know
he
doesn't
want
us
to.
You
know
back
away
from
that.
I
think
and
and
he
may
represent
a
a
contingent-
I
don't
know
and
then
there's
the
esm
only
contingent.
So
I
guess
I
mean
I
think
it's
somewhat
somewhat
dependent
on
that
right.
B
Yeah,
my
in
a
sense
idea
like
when
we
discuss
it.
I
use
forward
as
an
example
for
js.
B
So
for
one
time
we
discussed
like
what
to
do
with
flow
typings
when
we
switch
to
typescript,
because
we
like
break
people
consuming
graphqs
in
flow
and
it
it
was
adding
like
complexity
to
the
entire
immigration
process,
even
like
somebody,
someone
from
guild
offered
like
to
maintain
them.
But
in
reality,
what
we
did
we
created
a
issue
asking:
should
we
remove
it?
B
There
was
no
nothing
strong
for
we
went
away
and
remove
it
and
listen
if
anybody
will
scream
in
pain,
basically
like
in
a
sense
like
if
we
break
somebody
or
not
and
for
for
flow.
It
was
a
great
like
great
step
because,
like
nobody
like
it
look
like
nobody
cared.
I
spoke
like
with
facebook
guys
and
they
agreed
that
it's
necessary
step
and
they're
okay
to
bite
the
bullet
right.
C
B
B
B
B
We
probably
need
to
edit,
especially
if
more
people
will
come
or
if,
like
we
discuss,
so
I'm
like
right
now
and
it's
probably
located
back
because
there
is
at
least
somebody
say,
says
they
needed
in
case
of
fall.
Nobody
said
so.
B
We
can
do
even
more
or
in
a
sense,
I
can
probably
like
pin
with
his
issue
and
see
like
if
more
people
will
react
and.
B
Probably
have
some
discussion
like
how
breaking
it
is,
but
I'm
like
feeling,
probably
will
it
now
from
my
attempt
to
switch
with
repo
to
sm.
I
just
have
like
bunch
of
projects.
It's
not.
Yes,
I'm
ready
like.
I
cannot
switch
our
repo
to
yes,
I'm
only
here,
because
docosaurus
is
not
supporting
esm
fully.
So
it's
all.
Quite
it's
look,
like
my
assumption
was
I
assume
it
early
that
everybody
okay.
This
is
some
what
I
see
it's
like
it's
like
50
type
script
and
got
normally
some
support
only
in
latest
version.
B
B
You
need
to
maintain
and
test
both
both
like
things,
so
I'm
like
treated
like
a
temperature,
even
if
it
suggests
we
probably
had
suggest,
but
even
but
only
for
couple
releases,
maybe
for
a
year
or
two.
It's
not
like
long-term
thing,
so
I
don't
think
we
should
design
api
around
it.
That
said,
our
current
instance
of
mechanism
is
bad
and
it's
it's
bad,
I'm
like
totally
agrees,
but
so
I'm
using
warrant.
They
are
not
as
like,
cjs
mgs,
specifically,
but
possibly
as
possible,
replacement
of
instance,
of
check
right
now,.
C
Okay,
that's
pretty
fair,
so
so
I
think
I
think,
then
the
only
thing
to
talk
about
specifically
in
terms
of
lawrence
pr
and
using
symbols
is,
is
the
relationship
between
versions
which
you
partially
covered
with
your
suggestion
that
we
use
git
commits,
I
think,
what.
A
C
Was
pointing
out
is
that
we
we
have
another
option,
which
is
that
we
can
say
that
we
only
support
the
same
versions,
but
we
can
stop
checking
for
them,
meaning
meaning
package
managers.
You
know
usually
take
you
know,
take
the
center
stage
and
making
sure
that
all
the
you
know
all
the
packages
are
managed,
and
so
we
can
say
that
any
errors
are
not
really
our
responsibility
and
that
you
can
use
the
same.
You
know
you
know
even
for
example,
esm
and
cjs.
C
You
have
to
use
the
same
version
and
it's
on
you
to
make
sure
that
you're
doing
that.
So
I
think
that's
what
I
think
when
lee
added
the
this
ins.
C
Actually
I'm
not
sure
about
that,
but
you
can
maybe
tell
me,
but
when,
when
this
instance
of
check
was
originally
introduced
to
check
to
help
also
check
versions,
I
think
that
was
around
the
time
that
library
switched
to
being
a
peer
dependency
and
it
was
very
important
to
make
sure
you
know
in
terms
of
the
overall
ecosystem,
that
people
were
using
it
that
way.
But
now
I
you
know
it's
been
some
time
since
then.
So
maybe
you
know
just
to
keep
our
lives
simpler.
C
B
Yeah,
I
can
provide
little
bit
context
about
that,
so
back
when
lee
edited
he
had
did
not
like
just
preventive
measure
for
some
future
problem.
I
think,
like
we
use
per
dependency
for
very
long,
even
before
what
we
and
pm
like
all
and
then
have
weirder
behavior
around
peer
dependency.
Sorry.
B
Okay,
so
yeah,
I
will
continue
and
maybe
jakub
can
watch
it
in
youtube.
So
oh
might
be
listening
at
the
same
time,
so
lee,
basically
that
it,
because
we
had
bunch
of
issues
with
errors,
saying
and
like
solution
for
them,
was
remove,
not
models
and
do
npm
install
so.
B
A
I
think
there
are
a
couple
issues
with
instant
soft
checks.
One
is
like
well,
first
question
is:
why
are
we
checking
and
sometimes
like
a
whole
debate?
We
can
have
right
and
then
keeping
esm
and
cgs
support
aside,
which
is
like
potentially
in
getting
rid
of
instances
of
that
you
can
get
it
to
work.
The
other
problem
is
like.
A
Let's
say
I
am
using
graphql
js
in
my
server,
but
then
I'm
also
using
code
gen
and
then
what
could
happen
potentially
is
like
you
have
like
lower
version
of
graphical
dns
in
your
server
but
graphql
code.
Gen
is
not
like
the
latest
version
and
then
now
you're
in
this
problem
of
like,
like
it's
conflicting
versions
in
your
dependency
trees,
and
then
you
this
year
there
was
like
you
can't
use.
You
need
a
graphical
js
to
be
one
version
which
is
like.
I
see
that
happen,
like
almost
every
like
a
few
weeks
later.
A
I
would
see
this
one
of
these
things
pop
up
and
then
the
only
way
to
fix
this
is
like
in
the
yarn.
I
I
just
forced
on
resolution
to
point
it
towards
version,
but
that's
not
like
a
good
fix,
because
they're
like
npm
doesn't
support
it,
but
on
pnpm
I
can
do
it.
So
it's
like
like
keeping
like
you
know,
bundling
on
the
side
and
like
why
runtime
jack
it's
more
like.
B
It's
it's
like
actually,
a
very
good
point.
So
what
happened
before
lee
added
instance
of
check
it
was
the
same.
Mass
package
was
described
as
per
dependency,
but
people
were
saying
like.
B
Yeah,
this
is
exactly
right.
Issues
was
issue
was
always
the
same
issue
was
like.
B
If
you
never
pass
values
between
graphql
js
instances,
nothing
bad
will
happen
if
graphql
code
just
use
graphqs
as
internal
dependency
and
output
just
generated
got
nothing.
Quick
theoretically
should
happen.
If
your
not
model's
folder
is
not
screwed
and
we
all
know,
npm
is
not
ideal
and
yard
not
ideal.
So.
B
It's
because
in
some
sort
of
works
like
that,
even
if
we
figure
out
even
if
we're
much
like
warren
pr
currently
like
is
it
implemented.
Currently
it
will
solve
some
of
the
problem,
but
it
doesn't
show
others,
for
example,.
B
B
B
Theoretically,
we
can
say
we
support,
like
all
previous
graphics
version.
So
if
you
create
schema
with
like
really
old
graphql
js
version-
and
you
pass
it
to
execute,
it
will
work,
but
it
it
really
means
we
can
cannot
use
anything.
They
bad
into
graphql
schema
and
I
use
graphql
schema
as
an
example.
We
have
a
bunch
of
types
errors,
for
example
like
if
we
change.
B
B
A
B
React
previously,
you
could
not
use
mixed
like
different
react
version.
React
is
also
per
dependency
and
previously,
before
version
18,
you
could
not
mix
react
versions.
I
think
version
18,
like
last
version
or
one
before
yeah.
A
You
can't
mix
versions
with
the
react
but,
like
let's
say
I
have
like
a
package
like
you
know,
react
16.5
into
my
three
and
then
I'm
installing,
like
some
dependency,
which
is
like
16
dot,
something
like
it
does
not
really
matter
which
16
I
am
on,
they
would
work
fine.
I
don't
get
any
errors.
I've
even
seen
things
work.
B
Yeah,
so
a
react
in
a
sense
like
two
things
like
react
before
like
it
was
a
pretty
big
feature
for
them
to
support
different
direct
versions.
I
never
tried
different
subversion
so
like
not
a
major
version
but
subversion.
I
never
tried
to
mix
them,
but
react
always
was
a
bird
dependency
in
all
libraries
I
used.
B
A
C
A
Let's
take
graphical
as
an
example:
graphical
is
an
old
react
project
if
graphical
was
running
on
react
12,
but
then
they
wanted
a
new
part
of
the
editor
just
to
use
react
18
because
of
suspense
features,
and
no
one
actually
wanted
to
migrate
legacy.
Graphical,
like
I
guess
in
that
case
for
react.
It
makes
sense
to
invest
time,
but
for
us,
is
it
really
worth
the
effort
like
yeah?
Does
it
really
make
sense,
like
you
wouldn't
be
like
that,
does
not
have
a
spark.
A
We
do
have
a
spec
to
follow
so,
like
you
can't
be
saying,
like
I'm,
gonna
run
on
react
with
kafka
15
and
then,
as
graphql
17
comes
out,
you're
basically
spec
non-compliant,
because
in
17
let's
say
client
control.
Nullability
is
a
stack
feature
right
like
basically
it's
non-spec
compliant
server.
So
we
don't.
We
don't
even
want
to
encourage
people
using
old
versions.
You
want
people
like
easier
paths
to
migrate
to
new
versions,
but
then
for
us
the
problem
is
like
I
could
be
on
16.5
and
in
some
dependency
3.
B
Imagine
somebody
adds
a
new
method
to
to
to
schema
course
and
it
get
backported
to
16x6
it's
obvious.
B
If
person
needs
something
he
had,
he
had
it
and
asked
he
cannot
backboard
and
they
probably
say
yes,
because
there's
no
functionality
so
usually
when,
like,
for
example,
like
schema,
coordinate
with
good
example,
schema
coordinates,
add
schema
to
every
type
and
field
and
argument
and
something
it's
a
schema,
coordinate
and
we
have
like
chain
of
pr's
to
edit
and
they
had
like
schema,
coordinate
to
every
object,
but
they
also
change
validation
code
and
some
other
code
to
actually
use
with
schema
coordinates
in
the
error
messages.
B
B
So
it's
get
like
and
somebody
asked
to
buy
ported
currently
or
we
release
it
after
17
version,
so
it
gets
backported
and
releases
16.8,
for
example.
So
what
happened
and
that
after
that?
So
if
you
use
schema
from
16.7
before
this
change,
nothing
have
like
schema
coordinator
touch
it
arguments,
fields,
types,
nothing
have
scheme
according,
but
the
code
inside
validation
expect
things
to
have
like
scheme.
According.
B
So,
let's
see
so
problem
here,
yeah
we
we
can,
if
we,
if
we
do
a
little
bit
of
tension
around
the
tree
for
back
parting
things,
so
we
can
say
if
person
want
to
backpost
him.
Accordingly,
we
can
say
like
it's,
okay
to
add
fields:
it's
not
okay
to
backport,
like
validation
code
that
use
them.
We
can
say
that,
but
it's
created
a
bunch
of
issues
right
now,.
A
A
Let's
just
cut
two
things
like
schema
coordinates
is
something
I
can
add
like
15
16
would
be
fine,
no
one
will
actually
complain
about,
but
then
the
forehand
stream
is
also
a
new
spec
feature.
That's
it
that
gets
approved,
but
then,
even
if
you're
trying
to
backport
it
you're,
basically
non-compliant
graphql
server
right
because
you're
not
supporting
the
foreign
stream.
At
that
stage,
like.
B
One
thing:
spec:
don't
have
breaking
versions,
it's
all
incremental,
it's
incremental,
so
streaming
g4
added
as
server
may
support
stream
and
default.
So
streaming
deferred
is
optional.
B
B
Like
we,
we
have
certain
guarantees
you
can
treat.
Everything
is
breaking
change,
but
for
aspect
is
what
we
mean
by
non-breaking
thing,
for
a
spec
is
that
if,
if
from
client,
we
can
change
implementation,
we
can
change
like
other
things,
but
if
it
worked
previously
for
clients,
it
will
continue
to
work
for
them
and
the
remains
back
combined.
So
all
the
servers
need
to
remain
spec
compliant,
so
stream
defer,
say
it's:
if
server
server
might
support
streaming
differ,
not
not
sudo
required,
so
okay,
yeah
yeah,
but
it's
not
a
good
example
here.
B
Okay
and
spec
will
never
add
streaming
before
breaking
for
another
reason,
execute
right
now,
execute
return.
You
is
a
result
or
promise
of
result.
Stream
d4
is
breaking
for
graphql
js,
it's
not
breaking
face
pack,
but
is
breaking
for
graphql.js,
because
after
we
implemented
execute
will
start
returning.
B
No,
no,
like
result
promise
of
result-
and
I
think,
like
I'm,
I'm
not
sure,
probably
yeah-
promise
of
a
synthetic
so
at
least
like
free,
free,
new,
new
bike
return
type
and
it's
breaking
change.
If
some
something
previously
returned
like
number
or
string.
If
now
it
will
also
return
array,
it's
breaking
change
right
because.
C
B
B
By
the
way,
like
one
of
another
example
of
problematic
thing,
one-off,
add
like
a
new
flag
to
every
input,
type
being
like
true
or
false,
and
when
we
work
with
schema,
we
expect
this
property
to
be
present.
B
B
There's
like
producer
site
and
consumer
side
or
graphql
js,
like
not
every
library
has
some
way,
but
it
just
provides
a
single
function
and
you
pass
input
and
receive
output.
Nothing
is
bad,
but
on
graphql
you
can
create
a
schema
and
pass
it
to
some
function.
So
it's
mean
there
is
a
loop
outside
of
graphql.js
itself.
Open
third
party
code,
so
like
codjen,
for
example,
create
schema
and
pass
it
into
validate.
B
It
was
situation
when
graphical
tools
was
on
a
new
version
of
graphql
supported,
newer
versions
and
typed
graphql
was
behind
supporting
one
version
back
only
so
without
peer
dependency
and
without,
like
twist
check,
we
in
situations
that
type
graphql
create
a
schema
with
one
version
of
graphql
js.
B
B
Graffia
because
we
have
the
swoops
yeah
yeah,
it's
it's
like
our
ipa
is
huge,
like.
I
think
our
api
is
bigger
than
react
in
a
sense.
Last
time
I
checked,
we
have
like
400
functions,
public
one
courses
and
functions.
We
have
like
a
couple
of
dozen
classes
and
hundreds
of
functions,
so
it's
it,
creates
a
mess,
and
especially
since
there
is
like
a
really
strong
ecosystem
of
things
built
around
like
bunch
of
wrappers
like
you
can
create
schema
with
like
kraken
next
type,
so
many
yeah.
A
A
Like
instead
of
saying,
16
or
because
because
essentially
in
in
the
world,
let's
say
becomes
schema
coordinates
and
then
we
have
17
with
the
defer
and
stream,
which
also
has
schema,
coordinates
in
theory.
The
schemas
I
generated
with
both
of
them
would
be
the
same,
but
the
executed
is
different
right
or
the
validation
could.
Validation
is
also
same
because
I
mean
okay.
Well,
maybe
not
the
different
stream
parts
of
it.
But
then,
if
I'm
not
including
the
fur
and
stream,
everything
would
be
the
same
on
two
different
breaking
versions
of
graphically.
Yes,.
B
Okay,
let's
right,
there
is
like
a
couple
of
things
stuff,
it's
like.
First,
we
need
to
experiment
before
spec
release,
which
is
like
idea
of
fact
process
much
stuff
people
use
it
in
production
before
we
add
stuff
to
his
pack,
it's
like
because
we
can
break
rocky
js.
We
cannot
break
spec.
B
We
could
still
have
like
a
I
mean,
but
it's
technical,
let's
focus
on
other
points,
it's
just
like
in
reference.
It's
like
it's
deserved
with
some
separate
discussion,
but
it's
even
outside
of
this
group
is
how
we
decide
to
develop
spark
itself
back
to
what
what
in
scope
of
this
group,
there
is
internal
things,
for
example,
schema
coordinates.
B
B
It's
it's
not!
It's
not
related
like,
for
example,
it's
harder
to
find,
but
I
can
send
you
a
link
afterwards.
B
B
So
I
added
two
schema,
so
every
schema
created
by
this
version
of
graphql
js
and
all
future
version
will
include
this
method
right
and
execute.
I
I
in
execute.
I
use
that
method,
so
it's
also
compatible.
But
if
I
go
one
version
before
I
edit
into
a
schema
and
generate
schema
there,
that
instance
of
a
schema,
don't
have
this
method.
B
A
A
I
don't
care
if
there's
type
info
in
v6,
v17
or
not
in
that
case,
it's
more
like
internals
of
like
graphical
code
and
yeah,
wouldn't
matter
but
like
let's
say
graphical
code
is
on
v17
and
nexus
is
on
v16,
okay,
but
the
schema
generated
from
nexus
is
like
spec
compliant
like
nothing
is
wrong
and
then
again,
ideally
you
shouldn't
be
past.
It
like
to
my
understanding,
you
really
shouldn't
be
just
surpassing
the
schema
object.
You
should
be
parsing
your
own
schema
in
that
case
and
building
a
new
schema
object.
B
B
B
B
B
Yes,
I
do
see
your
screen
so
in
this
change
it
happened
like
last
month
I
had
a
new
method,
get
filled
to
a
schema
object,
it's
not
required
by
spec.
It's
not
just
implementation
detail
and
I
use
this
method
in
execute
and
subscribe
and
basically
like
to
remove
code
duplication.
We
had
like
this
function,
the
same
function
code
here
and
inside
execute,
and
I
just
like
move
it
to
a
schema.
So
here
in
line
480
now
we
have
like
a
schema.getfield
right.
B
B
A
B
B
B
It's
like
it's.
This
function,
save
us.
B
We
have
like
other
otherwise
similar
like
here,
get
implementations,
for
example,
yeah,
it's
wrapper,
so
in
a
sense,
some
of
the
functions
will
do
cache
and
call
speed
up
things.
B
B
B
It's
it's
not
even
functions
like,
for
example,
yeah.
I
can
try
to
find
them
one
off
by
bench
to
explain
you
and
one
offer
is
obviously
spark
feature
but
yeah.
I
need
to
check,
but
I
expected
to
have
yeah
definitions
and
yeah.
Here
you
see
it's
like
every
type
you
have
boolean
and
it's
not
optional.
It's
is
a
true
or
false,
so
records
that
use
it
expect
true
or
false.
It's
not
expected.
Like
truth,
also
undefined
code
is
packed
with
property
to
be
there.
B
C
B
This
is
right
problem
of
supporting
multiply
version.
We
can
do
that,
but
it's
like
a
big
sacrifice.
We
freeze
iowa
api
and
the
white
migration
process.
Deprecation
process
is
like
basically
impossible.
We
cannot
remove
stuff,
it
becomes
like
very
complicated,
I
would
say
so:
okay,
yeah
it's
complexation
and
before
we
it's
one
hour
and
yakov
is
not
connected
back.
B
So
we
have
this
problem,
always
with
instance
of
things
that
lee
added
it's
just
a
way
to
make
user-friendly
error
so,
without
instance,
of
check
user
create
schema
by
one
version
of
graphic
js,
pass
it
to
another
version
of
graphql.js
to
do
execute
inside
execute.
We
get.
We
called
schema.getfield
definition
and,
without
instance,
of
check
your
user
will
just
have
a
crash.
Saying
object,
don't
have
get
field
definition
method.
B
Which
is
bad
user
experience
and
user
will
open
issue
on
our
repo,
and
it
will
be
hard
to
investigate
and
wait
it's
a
homeless.
What
lee
added
is
explicit
check.
It's
mean
before
execute
start
executing
before
we
try
to
call
with
method.
First
thing
we
we
do
with
the
schema
is
to
check
if
it's
schema
and
if
it's
schema
but
created
by
another
version,
we
get
user-friendly
errors
saying
your
schema
is
not
compatible
created
by
different
version
and
it's
not
compatible
with
this
version,
so
instance
of
custom
instance
of
that,
please,
lea
added
it.
B
B
Okay,
yeah
anything
else
before
we
post
and
about
yeah
discussion
topic
with
yakov,
I
think
we
can
discuss
it,
asynchronously
yeah.
We
can
discuss
next
working
group
or
something
yeah
and
they're,
pretty
both
pretty
technical.
I
think,
like
I
explained
my
position
about
warrants
they
are
and
what
what's
deal
breaker
and
what's
not,
and
about
visitors,
basic
technical
issue,
I
I
wrote
a
comment
and
I
wait
until
like
yaakov
will
answer
it
yeah
anything
else
to
discuss
before
we
like
finish
this
one.
A
B
Yeah
thanks
and
bye
thanks
for
listening,
it
was
yeah
yeah
good
that
we
now
all
on
one
level,
and
we
can
like
discuss
like
how
to
deal
with
this
issue
because
it's
multi-year
one
issue
so
yeah.