►
From YouTube: IETF-JSONPATH-20210615-0900
Description
JSONPATH interim meeting
2021/06/15 0900
C
I
don't
know
I
can
this
is
glenn
here.
I
can
chat
to
greg
on
slack
if
necessary,.
C
C
B
C
Yeah
he
says
that
he
has
tried
using
his
email
address
as
his
username.
It's
not
working,
he
says,
he's
got
two
saved
passwords,
neither
of
them
work.
E
C
Yeah
he's
just
put
an
image
up
of
password
reset
and
it
says
didn't
find
a
matching
account.
If
you
don't
have
an
account,
you
can
find
one.
So
it's
looking
a
bit
difficult.
C
C
Yes,
okay,
he
says
how
other
telephone
numbers
somewhere
in
the
emails.
F
C
C
B
G
Okay,
okay,
so
let's
get
started
so
just
just
I'm
just
going
to
crack
through
it.
This
is
the
note
well
this
isn't.
This
is
an
ietf
meeting.
You
should
all
know
well,
but
just
in
case.
G
No
okay
I'll
do
I'll
I'll.
Do
it
with
the
board
once
it's
available,
but
we
should
probably
tim
between
us.
We
should
probably
keep
notes.
G
Do
basic,
thank
you,
tim
blue
sheets,
I'm
fairly
certain
that,
because
this
is
a
meat,
echo
call
that's
automatically
generated.
That's
what
usually
happens
is
that
the
case.
A
G
Correct
okay,
so
I
don't
have
to
deal
with
that
javascribe
I'll
watch
java.
So
I
don't
think
we
have
anybody
additional,
that's
exclusively
in
java,
so
that
shouldn't
be
a
problem.
Is
there
any
gender
bashing
or
anything
in
particular
that
people
want
to
make
sure
that
we
talk
about.
B
B
Yeah,
marco,
that's
right,
we
marco,
has
has
resigned
his
editorship
now,
as
we
know.
In
fact,
all
the
editorial
work
has
been
done
by
carsten
and
glenn,
and
stefan
we
thank
them.
For
that.
That's
great.
We
really
appreciate
the
support
james
and
I
haven't
had
a
chance
to
talk
this
over
yet
we'll
figure
it
out,
but
but
marco
will
no
longer
be
with
us
in
that
capacity.
B
As
for
the
agenda,
should
we
karsten
at
this
point?
Should
we
ask
you
to?
I
gather
you
have
been
pulling
together
issues.
A
H
B
I
I
think
we
I
think
we
have
consensus
that
in
future,
the
pull
requests
should
be
smaller
and
more
surgical.
Please
thank.
B
A
I
C
Just
while
we're
waiting,
a
suggestion
is
if
someone
needs
to
put
up
a
large
pull
request
and
it
may
be,
that's
no
longer
needed.
But
if
someone
needs
to,
they
can
also
put
a
draft
pull
request
for
comments
and
then
pull
it.
Apart
into
surgical,
small
pull,
requests
for
merging.
A
E
Trying
to
avoid
that
in
future
is
okay,
but
the
my
pull
requests.
E
Break
it
down
to
no,
no,
no.
E
A
Finish
the
current
full
request,
because
that
already
has
pushed
us
aside,
everything
else.
So
we
can
finish
that
now
and
and
I'm
dreaming
that
we
can
merge
it
today,
so
we
can
then
start
creating
little
pull
requests
on
specific
issues,
because
I'm
already
dying
to
to
make
a
few
of
those.
But
I'm
not
going
to
muddy
the
water
further
by
by
doing
this
right
now.
So.
J
A
Is
we
can
merge
102
today.
J
Real
quick
on
what
we
were
just
talking
about
number
102,
I
did
put
in
a
few
comments.
J
A
I
think
I
I
have
covered
everything
that
came
in
until
eight
time,
but
yeah.
There
are
lots
of
comments
in
there
that
really
need
to
lead
to
further
pull
requests.
So
we
cannot
simply
take
these
off.
All
three
98
99
and
102
have
comments
that
we
still
have
to
act
on,
but
that's
just
not
productive
to
try
to
tweak
the
the
big
per
request
we
we
have
to
act
on
these
things
in
small
pull
requests.
A
Yeah,
that's
good,
okay,
yeah!
So
let's
talk
about
the
the
one
thing
that's
remaining
of
97
and
the
that's
the
the
definition
of
of
the
term
string,
and
I
think
the
the
underlying
problem
is
that
we
are
basing
our
terminology
on
that
of
82
59,
which
really
has
been
written
to
define
what
adjacent
text
is
so
it's.
A
It
was
not
possible
for
political
reasons,
to
actually
fix
this
document
to
talk
about
the
data,
so
it
talks
about
the
text
and
so
the
the
it
doesn't
even
have
a
definition
section,
but
the
the
definitions
that
are
in
the
text
essentially
define
how
the
json
text
looks
for
something
and
not
how
the
json
values
look
like.
A
A
So
it's
important
to
keep
in
our
mind
whether
we
are
talking
about
a
literal
or
we
are
talking
about
the
the
number.
That
is
the
data
behind
that
literal
and
that
becomes
important
now,
because
we
we
are
defining
string
literals
that
actually
are
different
from
those
of
json
all
other
literals.
It
seems
I
don't
know
if
that's
actually
true,
but
to
me
it
seems
all
other
literals.
J
Good,
the
the
discussion
that
we
had
last
time
was
that
we
wanted
to
use
values
and
that's
what
this
pr
does.
This
part
pr
uses
the
term
values
we
never
considered
using
the
word
literals
any
time
with
until
you
brought
it
up
like
a
few
days
ago,
and
no
discussion
has
been
around
the
word
literal
until
you
brought
it
up,
and
so
my
question
is:
can
we
take
the
pr
as
it
is
for
what
we
discussed
and
then
use
another
pr
to
change
it
for
these
literals.
B
A
But
the
the
new
thing,
of
course,
is
that
the
literal
that
we
are
accepting
for
strings
is
more
permissive
than
the
literal
provided
by
jason
and
what
the
the
remaining
part
of
pr97
actually
does
is
redefine
string,
and
that
would
be
read
as
redefining
string
values
as
the
literal
and
that
doesn't
make
sense.
A
So,
that's
why
I
think
we
really
have
to
to
separate
these
two
aspects.
Clearly,
I
think
so
far
it
didn't
really
matter,
but
as
soon
as
we
actually
have
specific
literal
syntax
in
jsonpath
that
that
differs
from
the
one
we
import
from
h259.
We
have
to
be
careful
here
and
I'm
not
not
wedded
to
the
term
literal.
That
has
just
been
the
the
term
that
has
been
used
for
for
the
last
45
years
or
so,
but
we
can
also
say
notation
or
something
else.
A
J
I'm
fine
with
the
term
literal.
I
understand
it.
I
guess
my
question
is:
I'm
still
fuzzy
on
the
distinction
you
have
between
a
literal
and
the
value
is
the
value
the.
For
instance,
if
you
have
the
word
foo
in
quotes,
they're.
J
A
J
J
All
right,
well,
my
pr
doesn't
necessarily
do
that,
though
my
pr
doesn't
look
at
comparisons
stefan's.
Pr
does
later,
when
he's
talking
about
expressions.
J
B
J
So
it
sounds
like
for
the
purposes
of
expressions
we
want
to
make
a
distinction
between
how
it's
represented,
which
is
the
literal
and
the
value
behind
it,
which
is
what's
actually
going
to
be
used
in
the
expressions
and
maybe
that
can
be
extended
to
indexes,
for
example,
where
we
use
the
square
brackets
with
a
quotation
or
with
a
string.
Literal
quotations
as
well.
A
A
So
this
is
essentially
the
the
remainder
of
the
pr
everything
else
already
has
been
mud,
and
this
defines
string
after
saying
we,
we
import
number
object
array
and
so
on
from
from
8059.
A
A
J
A
J
I'm
happy
taking
stefan's
his
breakdown
of
of
what
the
string
is
from
his
pr
and
reusing
and
using
that
for
a
string.
Literal.
A
So
we
can
import
that,
like
we
import,
number
and
array
and
object
and
then
of
course,
at
some
point
in
the
text,
there's
a
b
and
f
for
how
we
actually
notate
strings,
and
I
I
also
have
a
suggestion
how
to
change
that.
But
that's
not
the
subject
right
now,
but
I
think
we
can
simply
lose
this
particular
pull
request.
B
B
That
that
we
adopt
a
definition
in
from
from
glenn's.
E
C
A
Yeah
again,
let's
merge
one
or
two
first,
but
I
think
one
of
the
next
projects
will
be
separating
separating
out
the
expression
language
from
the
way
it
is
used
in
in
the
various
selectors
and
part
of
that,
of
course,
will
be
defining
the
literal
aspect
of
the
expression,
language,
okay,
so
there's
lots
of
other
syntax.
That
needs
to
be
defined,
and
we
can
define
this
the
literal
syntax
with
that,
so
that
the
selectors
can
really
focus
on
on
just
saying
how
you
line
up
the
selectors
and
and
what
the
individual
kinds
of
selectors
actually
mean.
A
But
again,
this
is
about
the
definition
section:
it's
not
about
the
the
sections
that
actually
specify
things.
It's
just
the
definition
and
again
my
my
proposal
is
to
import
string
with
the
other
8259
terms
and
do
not
do
what
this
pull
request.
But
the
remaining
parts
of
this
pull
request
are
proposing.
I
mean
tons
of
stuff
has
been
merged,
so
we're
mostly
done
with
this
pull
request,
but
that's
the
one
thing
that
couldn't
be.
J
J
Are
people
comfortable
with
that
approach?
We
need
to
add
in
at
some
point
the
string
literal
from
the
expressions
or
a
variant
thereof,
but
that
needs
to
go
up
into.
A
Somewhere
well,
the
abn
f
is
right
now
smeared
over
the
text.
So
whenever
we
introduce
the
new
concept,
we
have
a
little
bit
of
abn
f
going
with
that
and
at
some
point
we
probably
want
to
consolidate
that
into
an
appendix
as
well.
But
right
now
it's
it's
easier
to
just
keep
the
smeared
abner.
J
I
I
understand
that,
but
for
right
now
it's
specifically
targeting.
J
J
I
come
on
89,
I
believe.
A
J
A
J
J
A
Yeah
expert
two
six
two
then
calls
it
string
letter.
A
B
C
A
I
haven't
checked
everything
else,
so
I
think
we
at
some
point.
We
need
to
actually
look
at
implementations,
for
instance,
if
there
is
a
backslash
eggs
in
in
a
string
literal,
what
two
implementations
do
there,
so
the
the
the
ecmascript
string
literals
are
even
more
permissive,
so
they
would
actually
have
a
definition
of
what
the
backslash
x
means.
Why
jason
simply
rejects
the
the
backslash
x
because
it's
not
defined,
and
we
have
to
find
our
position
on
that
continuum.
B
A
But
that's
actually
interesting
because,
for
instance,
if
you
have
a
backslash
n
somewhere
in
in
your
expression,
if
the
json
parser
already
does
one
level
of
unescaping,
then
what
the
the
string
literal
parser
for
json
path
will
see
is
a
new
line
and
right
now
it
will
choke
about.
J
B
C
Move
along
yes,
do
we
need
to
just
agree
that
the
I
forget
which
the
pr
number
was
I
get
confused
by
all
those
pr's,
but
the
one
that
greg
was
working
on.
Can
that
now
be
closed,
if
that
was
the
only
outstanding
bit
and
we've
merged
all
the
rest.
A
Well,
that's
your
discretion
is
the
editors,
but
that
would
be
my
recommendation.
J
Yeah
there
was
niggly
bit
that
I
had
on
line
430,
where
I
had
changed
a
self
self
referential
json
path
to
the
word
query,
but
I
can
let
that
go.
C
A
A
To
have
this
this
discussion
here,
but
I
think
we
need
to
actually
attack
it
from
different
angles.
So
with
that,
let's
talk
about
98,
which
is
stefan's
giant
pr
and
that
actually
comes
in
three
flavors
98
99,
which
was
a
fix
so
that
it
actually
builds
and
then
102,
which
is
a
rebase
on
the
the
101
which
came
out
of
97..
A
And
there
are
a
number
of
issues
actually
uncovered
by
by
this
pr,
which
we
don't
necessarily
all
have
to
answer
before
we
merge
one
or
two,
but
we
need
to
be
aware
off,
and
so
one
thing
that
that
came
up
in
various
places
is:
what
is
the
amount
of
power
that
we
actually
want
to
to
give
to
the
expression
language?
A
The
second
question
was:
should
we
have
function
calls
so
the
the
pr
currently
has
parts
of
the
syntax
for
for
a
function
call,
but
I'm
not
even
sure,
is
that
actually
you
do.
We
define
a
function
yet
in
the
pr.
A
So
that's
something
I'm
raising
the
issues
I
have
slides
for
for
the
individual
issues
later
and
the
the
third
one
was.
Can
we
actually
juggle
with
structured
values
so
that
the
term
that
8259
uses
for
containers
is
structured
values
and
the
term
it
uses
for
the
atomic
values
is
primitive
values?
And
again
I
would
propose
to
using
the
8259
terms,
even
if
I'm
not,
I
wouldn't
have
chosen
them,
but
that's
what
they
have.
What
we
have
structured
values.
E
E
Why
not
add
all
members
of
an
array
and
why
not
divide
the
result
by
the
by
the
count
of
the
numbers
in
this
array
resulting
in
median,
and
so
you
approximating
the
the
median
arithmetic?
I
I'm
afraid
about
the
fact
that
it
gets
quite
complex
very
fast,
so
I
propose.
B
A
A
We
have
to
be
very
specific.
What
what
we
mean
by
this
so,
for
instance,
comparison.
If
anybody
here
thinks
that
comparison
is
simple,
think
about
unicode
normalization.
Think
about
floating
point
comparison.
There's
lots
of
booby
traps
there
already
so
arithmetic
kind
of
seems
like
a
slam
dunk
compared
to
that,
but
I
think
we
can
keep
that
open.
Maybe
that's
not
something
we
have
to
define
right
now,
but
of
course,
where
we
have
examples
that
already
use,
arithmetic
or
or
structured
values,
and
so
on.
A
We
have
to
decide
whether
we
keep
them
or
whether
we
mark
them
with
not
really
supported.
Yet,
let's
see
whether
we
will
or
something
like
that.
C
My
question
about
some
of
these
features
is:
do
they
pull
their
ways?
You
know
I
can
see
that
they
have
use
cases
and
we
can
invent
use
cases
if
we
want
to,
but
are
they
really
generally
useful
compared
to
the
amount
of
complexity
that
they
introduce
both
for
the
implementer
and
for
the
user?
You
know
the
reader
of
the
the
spec.
B
B
Sorry
normalization,
please
you
know,
let's
not
go
there
and-
and
you
know,
as
regards
numbers,
you
know
8259
says
what
you
can
expect
is
you
know.
Ieee
754
doubles
and
that's
all.
C
B
Expect
and
and
given
you
know,
those
base
assumptions,
I
just
don't
really
believe
that
that
you
know
comparison
is
is
is,
is,
is
that
big
a
deal
so
so?
I
think
that
saying
arithmetic
is
a
slam
dunk
compared
to
comparison.
I
know
unconvinced.
A
It
has
taken
a
while
before
people
started
to
understand
that
so
so
originally
people
were
slapping
nfc
on
everything
whenever
they
had
a
data
value
in
the
hand
they
just
normalized
that
as
a
matter
of
course,
and
that
has
turned
out
to
be
the
wrong
way
of
doing
it.
So
the
only
place
where
you
actually
want
to
have
normalization
around
is
in
comparison.
E
Okay,
it
was
not
my
intention
to
say
that
comparison
is
easy
and
then
arithmetic
is
is
complex,
but
jsonpath
is
for
selecting
things
out
of
the
json
text,
and
and
for
this
you
need
comparison
for
the
first
place.
You
ask
for
a
special
path,
and
maybe
you
can
compare
a
value
with
a
given
value
for
selecting,
and
this
is
what
we
need
for
filters
just
comparison
on
the
first
place
in
the
first.
J
J
H
E
A
The
idea
behind
an
extension
point
is
that
when
you
pass
a
json
path
expression,
you
can
find
out
whether
an
extension
point
has
been
used
and
look
up
with
whether
your
implementation
supports
that
extension
point
and
then
then
apply
the
implementation.
So
it's
a
framework
in
which
extensions
can
be
put
in
without
destroying
the
the
rest
and
arithmetic,
of
course,
can
can
could
be
added
in
a
functional
notation.
A
Then
it
would
be
easy
to
just
make
the
whole
functional,
notation
and
extension
point
that
that's
actually
what
we
did
in
ctdl,
where
we
have
the
the
control
operator
concept
and-
and
so
we
made
it
a
functional
extension,
and
that
would
be
one
way
of
doing
it.
I'm
not
saying
that.
I
prefer
that
or
I
dislike
that,
but
that's
certainly
one
approach.
J
Important
for
the
purposes
of
this
pr
leaving
functions
out
is
good,
but
we
should
definitely
revisit
it.
J
J
Seems
like
having
that
defined
by
the
spec
seems
something
like
we
should
be
able
to
have
some
kind
of
operator
that
that
works
between
values
that
way
and
that
maybe
isn't
a
good
extension
point
but
function
calls
is
a
great
extension
point,
because
extensions
can
define
new
functions,
etc
and
all
those
functions
would
have
could
have
a
similar
syntax.
J
B
E
B
Does
anybody
I
I
smell
a
consensus
here
so
should
do?
I
generally
perceive
consensus
that
we
should
be
minimal
in
this
respect
for
this
pr
head.
I
think
so.
Okay,
not
hearing
any
objections,
good
ooh
carsten.
We
hadn't
finished
that
slide.
Yet.
A
Well,
I
I
was
just
trying
to
find
out
what
we
just
decided,
so
this
is
what
we
will
have
to
change
to
actually
do
what
we
just
said,
and
I
just
would
like
to
hear
what
people
think
we
should
be
doing
with
this.
E
E
E
It's
a
comparison
because
you
can
always
take
every
element
of
an
array
and
compare
it
with
it
with
with
with
the
value
in
question.
E
So
it's
a
shortcut
for
or
syntactical
yeah
shortcut,
okay,
the
question
is
of
if
we
are
allowed
to
apply
the
inoperator.
E
B
In
this
respect,
c
bergmer's
comparison
is
interesting.
You
know
he
looks
at
filter
expression
with
greater
than
greater
than
equals
in
is
actually
explicitly
called
out
there.
E
No,
no,
the
the
the
most
important
logical
expression
is
the
exist.
The
existence,
if
we.
E
At
dot
key
has
the
value
false
the
boolean
value
false.
You
must
explicitly
have
a
way
to
test
it
for
false.
You
can
do
it
by
equal
equal
false,
but
you
can
also
use
the
negation
operator
and
there
is
a
table
that
complies
mostly
to
javascript
how
the
not
operator
works
with
strings.
Empty
string
is
false.
C
A
point
about
the
in-op:
it
seems
to
drag
in
the
decision
about
whether
we
allow
structured
literals.
B
B
B
A
B
So
so
I
so,
as
speaking
under
my
chair
hat,
I'm
always
smelling.
Do
we
have
consensus
here?
Can
we
settle
this
thing
and
I'm
in
favor
of
taking
things
out.
C
I'm
a
bit
uncomfortable
still
within,
because
I
think
it
seems
to
sneak
in
comparison
with
structured
values,
which
I
think
we
need
to
discuss
separately.
A
A
J
Yeah-
and
this
particular
operator
is
actually
quite
unique
from
say,
the
equality
operator
in
that
the
the
left
side
takes
a
single
value
and
the
right
side
contains
it.
It
must
be
a
container
or
an
array
yeah.
It
doesn't
make
sense
to
have
it
the
other
way
around,
whereas
equality,
a
equals
b,
then
b
equals
a.
J
G
B
We
have
two
paths
forward:
one
is
to
adopt
basically
the
whole
screen
full
of
stuff,
simply
because
that
allows
the
the
you
know
the
emerge
to
proceed,
and
we
get
a
you
know,
a
single
document
that
we
can
work
on.
The
other
is
to
drop
a
few
of
these
things.
Now
I
am
not
particularly
religious
on
either
path
forward.
E
A
Question
is
that
okay
with
you
yeah,
but
it
just
needs
to
be
done
because
saying
that
let's
do
that
is
not
doesn't
mean
that
it's
actually
done.
So
let
me
just
try
doing
this,
so
we
we
just
took
out
the
calc.
Well,
we
took
out
funk.
E
In
op
is
going
away;
no,
no,
I
don't
want
to
go.
Have
it
go
go
away
just
to
discuss
in
a
later
pr.
It
needs
more
evaluation.
I
think.
J
A
Different
procedure
in
mind:
okay,
so
I
don't
think
we
always
have
to
work
below
the
water
line,
so
we
we
might
have
things
in
our
document
that,
where
we
don't
quite
know
how
it
works,
but
we
still
say
how
it
works
together
with
everything
else,
one
once
we
have
defined
how
it
works
and
if,
if
we
always
cut
away
everything
that
is
not
defined,
then
we
have
an
essentially
empty
document
most
of
the
time,
because
there's
always
one
detail
that
isn't
defined
yet.
Okay,.
C
G
E
And
do
some
investigation
about
that.
G
B
A
E
A
Language
and
yeah,
so
the
I
think
the
the
the
boundary
can
be
discussed
where
that
exactly
is,
but
I
I
would
consider
a
boolean
expression
to
be
part
of
the
expression
language,
but
on
the
other
hand,
there
are
some
things
here
like
like
the
dot
selector
and
index
selector,
where,
where
the
expression
language,
then
then
grabs
back
into
the
selector
syntax.
E
C
Just
looking
at
the
logical
operators,
we
don't
seem
to
have
any
they've
got
syntax
wrong,
but
I
can't
see
a
way
of
grouping
together,
disjuncts
with
inside
brackets
or
maybe
defining.
Maybe
maybe
we
need
to
define
the
precedence
of
abandon
or
but
ignoring
that
for
the
moment,
if
you
want
to
have
explicit
brackets,
I
don't
think
there's
a
syntax
for
that.
C
C
Yeah,
so
I'm
wondering
if,
if
there's
a
argument
here
for
emitting
logical
expressions
from
this
pr
and
bringing
them
back
with
the
full
power
in
a
later
pr,
I'm
in
favor
of
them
by
the
way.
But
I
think
they
need
to
be
done
right
as
it
were.
A
Yeah,
I
think
we
we
can
use
the
the
syntax
out
of
ecma
262
and
simply
compress
this
down
to
the
parts
that
we
actually
need.
So
I
think
that
should
be
an
easy
exercise,
and
I
can
do
this
after
this
meeting.
A
Yeah,
but
that
doesn't
make
sense:
okay,
okay,
I
I!
I
don't
think
that
the
the
whole
gamut
of
of
potential
expansions
of
this
actually
make
sense.
So
I
think
the
more
traditional
way
of
nesting
things
what
would
be
easier
to
to
get
right.
C
B
Okay,
I
think
we've
arrived
at
a
point
where
we
have
an
action
item
which
carson
has
volunteered
for.
Thank
you,
and
I
think
we
have
some
expectations
of
what
we're
going
to
get
from
it.
And
why
don't
we
let
him
loose
to
do
that
and
see
what
we
get.
A
Okay,
okay,
so
I
think
we
covered
about
two-thirds
of
my
remaining
slides,
already
yeah,
so
comparing
with
structured
values,
I
think
that's
on
the
to-do
list.
We
will
do
that
later.
Don't
do
it
now
right.
B
Whenever
I
see
these
things,
I
want
to
immediately
flip
over
to
see
bergmur's
thing
to
see.
J
Yeah,
okay,
I'll
reiterate
my
my
implementation
does
support
any
json
literal.
A
B
A
Okay,
so
I
think
we
yeah
there's
one
one
thing
that
I
already
mentioned,
but
I
want
to
mention
again,
so
we
have
had
this
terminology
problem
that
we
need
to
distinguish
atomic
from
complex
or
simple,
from
structured
or
whatever,
and
given
that
8259
uses
primitive
and
structured.
J
A
Okay,
then
we
had
the
discussion
about
selecting
a
value
versus
selecting
a
node.
That's
one,
one
more
terminology
problem
that
we
haven't
fully
understood
yet,
and
the
the
text
currently
mostly
talks
about
selecting
a
value,
but
I
think
it
actually
should
be
saying
we
are
selecting
nodes
because
we
are
not
selecting
not
just
selecting
the
value.
We
are
also
selecting
the
position
in
the.
E
A
Argument
yeah,
but
whenever
you
need
a
pointer
into
the
argument
as
a
return
value
of
the
json
path,
query
and
updating
is
one
case,
but
there
may
be
other
cases
where
it's
useful
to
have
that
pointer.
A
I
E
A
B
J
J
Yeah
because
I
initially
added
node
and
node
list
based
on
our
previous
discussion
and
then
it
looks
like
that
was
rolled
into
what
karsten
did
for.
B
E
No,
I'm
fine
with
we
stiff,
define
this
concept
and
and
the
concept
of
node.
C
E
If
you
do
not
have
your
egg,
of
course,
you
can
apply
the
the
query
again,
but
you
cannot
tell
if
you
have
the
number
five
from
where
exactly
in
the
document
it
is.
But
if
you
have
the
the
path,
the
normalized
path,
you
can
do
it.
This
is
only
possible
using
nodes,
and
this
is
what
we
need
to
explain
carefully.
Yes,.
J
One
my
initial
reaction
to
this
is
not
to
allow
it.
I
think
it
should
start
with
a
letter.
A
B
A
J
A
B
So
so
this
is
so
if
we,
if
we,
if
we
were
to
assert
that
a
dot
member
name
cannot
start
with
a
digit,
you
could
you're
stepping
on
a
slippery
slope.
You
could
say:
okay
well,
it
has
to
match
the
ecmascript
name,
production
which,
which
we
actually
did
in
in
an
implementation
where
we
were
sanity,
checking
json
scripts
jsonpass
before
we
applied
them
and
we
rejected
those
that
did
not
match
the
ecmascript
name,
production
which
actually
worked
pretty
well.
B
C
Yeah
I
mean
what
my
reaction
to
that
question:
what
does
dot
one
apply
to
an
array
mean
was
well
we're
applying
a
dot
child
member
member
child
to
a
an
array,
so
it
doesn't
select
any
nodes.
E
D
A
E
E
Might
be
discussed
controversially.
J
Now,
there's
something
that
I
brought
up
in
the
pr
the
while
the
the
wild
card
as
it's
defined
in
the
the
union,
as
it's
defined
in
the
pr
which
I
think
you
have
here,
does
not
include
wild
cards.
J
E
If
we,
if
we,
if
we
disallow
the
descendant
selector,
we
restrict
the
union
selector
on
one
level
of
object
or
array
level.
Is
this
correct,
so
the
wildcard
selects
all
so
the
wildcard
doesn't
need
to
be
inside
of
the
brackets
and
it
doesn't
need
to
be
combined
with
other
selectors
logically,
so
the
next
question
would
be.
Should
we
disallow
it
explicitly
for
this
logical
reason
or
keep
it
inside
for
stupid
users?
I
don't
know
from
yeah.
J
C
I
think
I
agree
with
what's
being
said.
I
think
the
only
reason
for
keeping
wildcard
in
there
is
for
uniformity
so
that
you
can
say
you
can
make
general
statements
about
unions
being
like
single
values.
If
you,
if
you
allow
square
brackets
with
a
wild
card
inside
then,
is
that
a
degenerative
degenerate
form
of
a
union
without
a
comma
yeah.
So
it's
a
uniformity
thing
for
me
yeah.
I
don't
feel
strong,
particularly.
A
A
syntactical
ambiguity
thing,
because
the
way
it's
currently
written,
it's
not
quite
clear
whether
you
enter
into
a
union
selector
or
into
a
what's
the
other
thing
called
selector.
C
C
E
C
E
Argument
greg
also
mentioned
the
fact
that
this
is
the
only
place
where
the
white
space
plays
a
role
in
in
in
the
abn
f,
and
this
is
a
good
question:
do
we
really
need
this
white
space
in
yeah.
E
E
B
So
I
think
I'm
smelling
agreement
that,
even
though
you
really
probably
shouldn't
need
the
star
inside
the
the
selector,
then
inside
the
union,
the
there's
no
harm
in
having
it
there
for
syntactic
uniformity.
J
Because
each
one
of
those
diff
each
one
of
those
makes
a
distinct,
not
distinct,
but
they
make
a
defined
set
on
that
current
level
of
selection,
wildcard
selects
them
all
which
makes
the
others
redundant
and
descent
selectors.
We
don't
have
a
syntax
for
it
and
it
would
go
beyond
the
current
level.
I
I
A
J
Okay,
we
are
not
adding
the
wild
cards.
C
I
think
I'm
not
sure
we've
got
consensus
on
that
because
I
mentioned
it.
The
synthetic
uniformity
thing
just
so
we'd
discussed
it,
but
I
don't
think
it's
a
particularly
good
idea
if
we
can
admit
it.
Let's
admit
it.
E
J
E
J
J
Explicitly
disallowing
it
also
kind
of
steers
authors
path,
authors
into
building
good.
J
I
E
Them
I
wrote
a
comment
one
hour
or
one
and
a
half
hour
before
our
session
and
it's
there.
E
E
A
H
A
J
I'm
putting
something
into
there's
something
into
the
chat,
which
is
the
atfu
equals
string
without
the
parentheses,
and
it
looks
it
looks
like
the
question
marks
part
of
the
atfu
to
me.
It
doesn't
look
like
it's
denoting.
Here's
here
comes
an.
C
A
C
The
other
argument
I
think
has
been
mentioned
is
the
comparison
project,
and
you
know
how
many
implementations
already
support
this
abbreviated.
Syntax
suspect
it's
none,
so
we're
kind
of
departing
in
philosophy,
from
what
we've
done
so
far
in
a
way,
yeah.
J
I
feel
this
may
be
maybe
not
a
release
two,
but
like
an
enhancement
on
jason
path.
After
we
release,
it
could
be
something
like
that
kind
of
like
we
have
versions
of
c
sharp
or
versions
of
c
that
come
out
like
carson.
You
were
saying
that
c
eventually
stopped
requiring
those
in
a
later
version
of
the
language.
J
J
I
put
another
example
in
the
chat
of
using
one
of
these
expressions
in
a
union
syntax
as
well.
E
Good,
if
it
is
a
if
it
is
a
good
programming
or
notation
style
to
use
parenthesis,
it
may
be
also
a
good
notation
style
to
always
leave
a
blank
after
the
question
mark.
E
C
Another
concern
is
interoperation.
I
know
that
if,
if
we
allow
this
compact
syntax,
then
implementations
that
implement
the
standard
will
interoperate.
C
E
Of
at
least
one
implementation
that
allows
to
leave
the
parenthesis
out.
B
E
A
Actually,
I'm
not
sure
about
that.
I
I
haven't
sat
down
and
and
pushed
the
abn
f
into
a
puzzle
generator.
So
there
might
even
be
some
technical
benefits
and-
and
I
think
I
will
do
that
when,
when
things
have
stabilized
enough
for
for
that
to
be
a
worthwhile
exercise,
so
kirsten
do
you
mean
a
syntactical
benefit
of
of.
I
E
If
you
find
syntactically
a
bracket
open
bracket
and
a
question
mark,
you
can
pass
until
you
find
the
closing
bracket
or
the
next
comma.
It's
it's!
I
I
don't
see
a
syntactically
benefit
anywhere.
E
J
A
Okay,
we
have
13
minutes
and
three
more
slides.
So
let's
try
to
get
through
those
the
white
space
issue,
the
the
abn
f,
needs
to
tell
you
where
white
space
is
allowed
and
that's
yeah,
not
fun,
but
that's
a
feature
of
abn
f
abnf
is
not
a
two
level:
parser
scanner
combination.
It's
a
single
level
specification.
A
So,
wherever
you
you
want
white
space
to
be
allowed,
you
have
to
put
it
in
and
that
of
course
requires
yeah
a
lot
of
repetition.
So
this
is
just
just
to
show
a
random
piece
of
abnf
that
I
have
recently.
A
Generated
yeah,
so
I
actually
use
the
capital
s
as
the
name
of
white
space
to
reduce
to
minimize
the
amount
of
noise
that
this
generates.
But
it's
still
some
noise.
J
I
think
it
makes
sense
inside
operations,
flight
or
inside
operators
of
a
union
like
around
the
brackets
and
commas
and
stuff
like
that.
Just
to
kind
of
increase
readability
inside
expressions
around
operators
would
be
fine.
I
would
not
necess
well,
you
could
do
it
between
selectors,
possibly
one
person.
J
Right,
one
person
did
comment
to
have
multi-line
queries
for
complex,
for,
if
you
wanted
to
do
a
multi-line
query,
you
could
do
that
to
make
it
more
readable,
which
would
mean
that
you
need
wide
space
between
selectors
as
well.
E
B
Is
the
general
feeling
that
that
they
should
be
allowed
where
it
doesn't
screw
up
the
grammar,
or
that
we
should
be
generous
with
white
space
or
stingy
with
white
space?
I
mean
generous.
There's
there's
no
doubt
that
once
things
get
complicated,
white
space
can
dramatically
improve
readability,
but
is
that
going
to
screw
up
any
other
aspects
of
expressions.
B
A
C
J
A
Yeah,
let's
see
how
we
do
this,
but
I
think
the
the
observation
has
been.
We
want
to
sprinkle
it
over
the
thing
and
except
where,
where
it
actually
would
hurt-
and
that
has
been
to
be
done
in
a
future
pr.
J
J
A
A
H
E
Yes,
predictability
is
a
point
and
testability
also,
but
this
is
not
for
for
the
end
user
disability,
so
I
my
my
opinion
is
to
leave
it
to
the
implementations.
E
The
order
is
always
the
same.
Whatever
implementation
you
are
using,
but
is
it
really
worth
the
effort?
A
Well,
I
find
it
nice
if
I
can
write
test
cases
without
having
to
do
special
gymnastics
for
for
sorting
things
before
I
compare
them
so
right
now,
I'm
I'm
one
example
where
I'm
fighting
with
is:
when
python
puts
out
xml
it
randomly
puts
out
the
attributes
in
any
order.
That
depends
on
a
seed
that
that
is
fed
into
a
random
generator.
When
python
starts,
so
you
can
be
sure
that
you
run
it
twice,
it's
very
likely
that
they
will
be
different
and
that
really.
B
Hurts
we
had
an
early
earlier
discussion
about
determinism
and
I
think
the
consensus
was
against
requiring
determinism.
Is
this
a
separate
issue?
No,
that's
exactly
the
same
issue.
J
I
think
there
are
actually
some
frameworks
that,
like
like
python,
for
example,
that
you
absolutely
cannot
get
the
get
determinism
from
it.
A
Yeah
so
for
sibo
we
decided
it's
not
a
basic
requirement
to
be
deterministic,
but
specific
application
can
ask
for
deterministic
encoding
and
then
it
is
a
requirement
for
implementations
that
want
to
support
these
applications.
E
Extension
point
and
something
we
we
tell
them
what
would
be
yeah.
J
I
think
we
can
have
probably
define
a
deterministic
order
and
then
label
it
with
the
should.
I
think
that
would
be
fine
that
way.
It's
not
a
requirement.
J
B
J
C
J
A
B
Yeah
but
then
then,
if
he
changes
the
the
underlying
server
all
his
tests
break,
if
he
hasn't
been
careful
about
doing
equals,
testing,
and
so
on
I
mean
I,
I
think
it
I
think
the
te
I
think
the
spec
needs
to
be
clear,
that
this
is
not
a
thing
that
is
part
of
the
just
the
jason
path
landscape.
You
know
it's
a
good
idea,
recommend
it
put
a
shirt
in,
but.
C
I
guess
another
comment
is
that
when
you're
actually
coding
the
implementation
of
the
descendant
selector,
you
have
to
make
a
decision,
it's
hard
to
see
how
you
would.
I
personally
don't
see
how
you
would
just
end
up
with
a
non-deterministic
implementation.
So
you
have
to
make
a
decision
and
it's
fairly
arbitrary.
So
you
can.
H
J
B
Yeah,
okay,
so
we've
got
an
action
here
to
do
a
pr
on
this,
but
you
know
nobody
is
in
favor
of
ironclad
requirement
of
determinism.
We
are
essentially
out
of
time.
Carson
show
us
your
last
slide.
I
think
there's
one
left.
H
A
We
do
have
the
reaction
on
the
implicit
conversions,
but
they
actually
came
up
again
with
this
concept
of
truthy
and
I
think
we
need
to
understand
how
we
can
avoid
pulling
in
implicit
conversions
through
the
back
door.
A
E
Now
there
there
is
no
implicit
conversion
to
the
boolean
except
the
existence.
The
existence
aspect,
if
a
member
or.