►
From YouTube: SES Meeting: Get Intrinsics
Description
Jordan Harband, champion of the Get Intrinsics Proposal, joins us to iron out the SES strategy group’s position in advance of the plenary meeting in the coming week. We concluded that the proposal was good as-written because it satisfies the criteria: hidden intrinsics are distinguishable by parsing the first component of its name and checking whether a property exists on the global, and enumeration order places language intrinsics, the global object, and host intrinsics in order.
So, Repair Intrinsics for Hardened JavaScript can rely on the heuristic that all new undeniable intrinsics will be both before the global object and have no corresponding global, hardening such objects as they are revealed by language evolution.
A
B
B
B
B
B
A
A
B
B
C
B
Jordan
harband:,
like
for
the
262
ones.,
I
might
remember
if
I
made
it
deterministic
for
the
host.
hook,
but
yeah,
in
that
case,
you'd
have
to
like.
guess
which
the
last
one
was
which
isn't
super
ideal.
well,
no,.
You
don't
have
to
guess.
it's
just
you
run.
no,.
You
run
until
you
backtrack
in
order.
B
D
B
B
C
B
B
B
A
A
C
A
A
A
A
A
A
B
B
A
Kris
“cowbert”
kowal:
things
that
should
go
above
global
and
things
that
should
go
below,
and
then
would
have
to
communicate
somehow
delimiter
was
to
say,.
This
is
where
262
stops
and
the
web
begins..
It
wasn't
to
say,
just
go
to
scrollable
and
stop
with
your
hardening..
You'd
still
want
to
go
through
everything..
It
would
just
be
categorizing
them..
We
don't
go
through
everything.
E
Kris
“cowbert”
kowal:.
That
would
be
that
that
aren't
part
of
ecmascript.
but
still
well,,
it's
it's
actually
even
more
sinister
than
that..
There
are
hidden..
There
are
intrinsics.,
never
mind
hidden
intrinsics
into
6,
2
that
we
don't
care
to
harden
right..
We
don't
care,
but
like,,
because
we're
not
sharing
them
with
a
guest
compartment,
for
example,.
Don't
have
any
need
to
harden
the
intel
api,
for
example.,
but
that's
always
gonna
be
an
enumeration.
In.
A
A
B
B
A
A
A
B
A
A
B
B
B
chip
morningstar:
you
know,
like,,
as
you
said.,
you
only
provide
host
like
language
objects
to
compartments,.
So
if
it's
above
the
line,,
then
you
provide
it
compartments,
and
if
it's
below
the
line,
you
still
want
to
lock
it
down,,
but
you
still
may
not
want
to
inject
it
or
anything
like
that.,
yeah,
ii.
Think
that's
right.
cause
the
the
whole
purpose
of
this
proposal.
D
Chip
morningstar:
is
to
chip
morningstar:,
give
you
a
means
to
defend
yourself
against
future
changes
to
the
language
which
were
anticipated.
the
time
you
wrote
your
shim,
or
whatever,,
and
and
there's
sort
of
2
different
categories.,
there's
one,
or
what
are
all
the
things
we
wanna.
We
might
possibly
want
to
lock
down.
D
A
A
A
Kris
“cowbert”
kowal:
after
scuttling,
they
get
to
the
global
object.
and
there's
nothing
there.
That's
of
any
use.
and
the
the
reason
this
is
possible
is
because
we
specifically
do
not
harden
certain
objects.
specifically,
the
global
object
in
the
global
context,
so
that
they
can
be
erased.
Essentially.
A
A
A
E
E
E
B
D
D
A
A
A
A
E
A
Kris
“cowbert”
kowal:
and
having
common
global
object
isn't
even
sufficient
to
that
end
because
we
don't
enumerate
all
properties
on
the
global
object..
We
do
enumerate
all
of
the
properties
on
shared
intrinsics,
like
the
array,
constructor,,
etc.
so
all.
It
tends
to
be
the
case
that
all
interesting
language
features
happen
to
be
properties
of
existing
shared
intrinsics.,
but
it
is
possible
for
some
things.
Like.
A
E
E
E
A
E
B
B
B
B
B
E
A
E
E
E
E
A
A
Richard
gibson:,
the
no
the
known
shared
intrinsics.,
I
might
be
wrong
on
that,
though.
no,.
I
think
I
think
we
do
start
at
global,
but
if
not,
that
would
be
like,,
that's
a
fix
for
us
to
make..
It's
basically
the
same
logic
that
already
exists
with
a
special
case.
Saying,,
the
global
itself
isn't
going
to
get
hardened.,
but
it
is
going
to
be
a
route
for
traversal.
E
E
E
E
Kris
“cowbert”
kowal:,
either
through
analyzing,
the
string,
or
through
looking
for
an
object
property,
or
whatever
it
is
right..
So
we
have
2..
We
have
2
predicates
that
we
think
are
reasonable
to
ask
of
2,
6,
2.
to
maintain
as
an
invariant
to
the
language,,
one
of
which
is
is
this
provided
by
the
language
or
provided
by
the
host.,
and
we
use
the
global?.
This
is
a
delimiter.
A
B
B
B
Jordan,
harband:,
based
on
jordan,
harband:
like
you
like
with.,
if,
if
we
include
it
all
to
all
intrinsics,,
you
don't
need
to
do
any
object.
Traversal.
you
just
go
through
the
list
and
do
everything.,
and
but
you
you
get
to
determine
like,
is
you
parse
out
of
the
string.
and
then,
you
say,
is
that
on
the
global?,
if
so.
B
B
B
B
B
B
Jordan,
harband:
way
to
kind
of
sneak
that
information
in
there
as
a
delimiter
for
all.,
but
we
don't
even
have
to
do
that..
We
could
use
null
as
a
delimiter
in
both
cases
or
something,
right?
it
just
just
anything.
That's
not
a
string
could
be
a
delimiter
in
both
cases.
If
you
wanted
to
be
really
explicit
about
it.
but,
like.
B
E
E
A
Kris
“cowbert”
kowal:
yeah.
provided,
provided
that
there
is
a
language
invariant,
that's
preserved,,
that
language
invariant,
preserved
to
make
it
so
those
questions
continue
to
be
reliably.
Answerable.
yeah,
ii
am
highly
invested
in
providing
as
few
opportunities
for
monkeying
with
things,
with
host
hooks.
As
I
can.
so
yes,.
There
will
be
all
sorts
of.