►
From YouTube: IETF-JSONPATH-20220118-1000
Description
JSONPATH meeting session at IETF
2022/01/18 1000
https://datatracker.ietf.org/meeting//proceedings/
A
B
B
D
Before
I
do,
for
those
of
you
just
joined,
I'm
terribly
sorry
for
the
food
following
up
by
the
original
meeting
booking
I
put
in
the
wrong
day
right
time
wrong
day,
which
is
why
you
got
that
terrible
error
message.
D
So
sorry
about
that.
I
got
to
do
my
usual
boilerplate,
so
let
me
just
crack
on
with
it.
This
is
an
official
ietf
meeting.
So
please
know
well
the
usual
times
all
right.
You,
you
guys
are
familiar
with
this,
but
I
have
to
say
it
just
the
australia.
Are
we
gonna
do
what
we
usually
do,
which
is
try
to
take
links
as
best
as
we
can
and
I'll
edit
them
afterwards
and
send
them.
D
Okay,
I
can
do
that-
and
I
can
make
some
time
later
in
the
week
to
edit
them
up
once
available,
because
this
is
such
a
last
minute
booking
I'll,
try
to
make
sure
that
the
body
is
available
if
there,
if
it
isn't
I'll
chase
me,
don't
go
about
that
blue
sheets.
We
don't
have
to
do
because
this
is
a
meat
echo
thing
jabba
scribe.
D
We
can
obviously
just
keep
an
eye
out,
there's
only
a
handful
of
us,
so
it
shouldn't
be
too
too
bad
and
is
there
any
agenda
bashing
that
we
want
to
do
before
we
get
started.
A
I
thought
kirsten
slides,
captured
the
essentials
pretty
well
cool.
D
D
One
thing
I
would
like
to
talk
about
towards
the
end
is
perhaps
have
a
bit
of
a
discussion
around
what
our,
what
the
working
group's
thoughts
are
around
ietf,
113,
future
meetings
and
so
forth,
particularly
given
that
the
iatf
has
announced
that
it
will
be
doing
in-person
hybrid
meetings
this
year
with
two
european
cities,
so
vienna,
vienna,
philadelphia,
london
have
been
announced.
D
E
A
Because
we
we
approved
them.
D
Yes,
but
because
this
is
because
I've
created
a
new
meeting,
I'm
gonna
have
to
mess
around
with
downloading
the
slides
first
and
then
re-uploading
them,
I
guess
is
what
I
have
to
do
is
that
right,
casting
well.
E
Use
the
the
button
in
the
top
right
corner.
D
E
So
yeah,
I
think
you
know
which
day
it
is.
I
mean
quickly
summarize
the
status.
Yesterday
morning
we
had
37
open
issues.
Yesterday
evening
we
had
12
open
issues,
and
this
is
because
I
closed
25
of
them,
so
some
of
them
had
long
been
marked.
As
can
we
close
this
now
or
something
like
that
and
some
well,
I
tried
to
summarize
them
for,
for
this
slide
set
and
it
turned
out.
We
had
kind
of
done
them
or
they
were
about
a
part
of
jsonpath.
E
So
I
marked
those
with
with
another
tag
which
was
like
revisit
afterbase
or
something
like
that.
But
anyway
I
closed
25
because
they
are
not
relevant
to
the
current
state
of
the
document.
So
there
are
now
12
open
issues
and
some
of
them
are
marked
tasks,
and
these
just
need
to
be
done.
I
mean
somebody
has
to
check
all
the
the
abnf
and
and
all
that,
so
we
probably
don't
have
to
discuss
it.
E
It's
just
work
that
needs
to
be
done
and
probably
when
it's
done
then
then
quickly
checked
by
people
in
in
this
group.
But
there's
not
really
anything
we
can
discuss
today,
so
I
have
like
seven
or
eight,
for
which
I
have
slides
so
just
to
remind
was
there
a
question.
E
B
E
E
It
means
that
we,
we
have
a
much
more
manageable
set
of
items
before
us
and,
of
course
it
doesn't
mean
that
we
won't
come
up
with
new
issues
or
that
that
we
maybe
even
will
find
things
in
the
issues.
I
close
that
that
do
need
to
be
done,
but
I
think
in
general
the
cash
is
a
little
bit
more
manageable.
G
Question
my
personal
feelings
on
closing
the
issues
is
that
we
shouldn't
if
they
are
not
done.
I
recognize
that
these
are
these
issues.
Many
of
these
issues
are
not
done
because
we
are
explicitly
not
including
them
in
what
you're
calling
the
base
document
or
the
the
I
guess,
the
first
draft
of
or
whatever
this
is,
and
I
I
just
feel
that
the
the
issues
should
remain
open,
put
the
label
on
them
sure,
but
if
they're
not
done,
I
believe
the
issue
should
stay
open.
E
G
But
how
do
I
know
later,
which
ones
of
those
are
done
or
which
ones
aren't
or
which
ones
we
decided
just
not
to
include
that's
what
the
label
is
for.
G
E
E
C
In
general,
I
think
that
the
things
that
are
left
open
are,
let's
just
say
at
this
point,
carsten's
best
guess
at
what
we
need
to
do
before
we
do
go
to
work
in
group
last
call,
and
I
think
that
there
were
a
lot
of
decisions
made
there,
and
it's
perfectly
unsurprising,
if
one
or
two
of
us
disagree
with
with
one
or
two
of
those
decisions
and
if
so,
please
bring
it
bring
it
alive.
G
G
G
E
I
Can
I
make
a
suggestion
it
seems
like
greg
is
asking
for
the
issues
which
are
currently
labeled
revisit
after
based
on
to
be
reopened?
Yes,.
I
Okay
hold
on,
but
why
don't
we
just
do
that
after
the
base
is
done.
Yes,
then,
we're
all
happy.
E
G
G
G
G
C
Okay,
since
this
has
already
been
done,
the
the
discussion
is
perhaps
not
that
useful
at
this
point
in
time,
particularly
given
that
we've
agreed
that
once
we
have
a
a
last
call
working
group
last
call,
we
will
go
and
reopen
those
issues
that
are
that
are
so
marked,
and
that
should
get
us
back
to
a
place
where,
where
greg
would
like
to
see.
E
So
absent
equals
absent.
That
was
one
of
the
the
items
where
we
ended
the
last
discussion
with
discuss
this
on
the
list.
Unfortunately,
not
much
discussion
has
happened
on
the
list,
so
the
question
is:
when
we
compare
empty
node
lists
in
comparison.
Are
these
equal?
Are
these
not
equal?
So
do
we
do
something
that
we
can
imagine
as
converting
them
to
undefined,
or
do
we
something
that
we
can
imagine
as
converting
them
to
not
a
and
yeah?
E
That's,
that's
just
the
device
in
our
minds
to
think
about
this,
so
this
device
assumes
that
in
in
the
processing
of
the
expression
language,
we
are
not
limited
to
json
values,
which
we
apparently
not
because
we
have.
C
Paths
in
there,
so
I
think
this
I
think
this
issue
is
reasonably
well
understood.
We
understand
what
we're
disagreeing
about
here
and
I
might
you
know
last
time
we
discussed
this
some
of
us
some
of
us
thought.
Well,
obviously,
it's
like
not
a
number,
and
some
of
us
thought.
Obviously
it's
like
absent,
and
so
the
idea
was
we
get
some
use
cases
and
help
us
help
and
see.
If
that
changed
anybody's
point
of
view,.
E
If
we
decide
that
we
go
for
the
number
answer,
then
somebody
who
wants
to
to
empty
notice
to
be
equal
has
to
add
two
existence
tests
and
all
this
with
the
result
of
the
comparison,
and
somebody
who
wants
know
for
the
answer,
of
course,
can
can
use
the
comparison
directly.
E
C
So
I'm
confused.
I
think
that
so
just
to
be
clear,
I
I
like
I,
I
would
not
want
them
to
be
equal
if
they're,
both
you
know
just
not
there
so,
and
this
is
the
case
where
dot
a
and
dot
b,
neither
of
them
are
there.
C
E
Yeah,
but
you
have
to
check
both
because
if
you
only
check
one,
the
other
one
might
exist.
G
B
Well,
unfortunately,
jason
power
jason
doesn't
defines
undefined.
This
is
the
issue
here.
I
think
I
would
like
to
prefer
set
preference
on
on
on
the
first
converts
to
undefined
it's.
G
It
would
evaluate
those
values
it
would
return
null
for
both
of
them,
because
it's
dot-net
and
it
doesn't
understand
that
null
is
an
entity,
but
then
it
would
compare
null
to
null
and
you'd
get
it
true.
E
G
B
We
we
choose
the
not
a
number
style
because
of
javascript,
where
the
comparison
not
a
number
equals.
Not
a
number
is
false.
This
is.
B
G
C
Sorry
in
ieee
arithmetic,
not
a
number
does
not
equal,
not
a
number.
I
think.
E
Yeah
and
the
the
reason
they
did
this
essentially
was
that
they
changed
floating
point
processing
from
actually
producing
a
value
from
actually
producing
an
exception
in
cases
like
this
to
producing
a
value
so
to
make
existing
code
work
right,
they
had
to
make
this
value
anti-social,
so
it
doesn't
compare
equal
even
with
itself,
but
that's
because
that
that's
essentially
that
the
nan
a
value
actually
mirrors
or
reflects
what
would
have
been
an
exception
so
that
that's
a
kind
of
retrofit
mechanism
and
it
sure,
hurts
when
you
actually
work
on
floating
point
numbers
that
you
cannot
simply
compare
to
a
not
a
number.
E
So
you
have
to
have
these.
These
is
nan
things,
so
you
cannot
have
a
table
that
that
says,
for
instance,
what
do
you
expect?
E
Because
if
the
table
has
has
another
number,
then
you
have
to
check
that
explicitly
so
so
a
lot
of
code
actually
looks
like
the
the
second
to
last
color,
second,
to
last
line
on
this
slide,
and
that's
why
I
really
would
like
to
go.
Go
back,
it's
not
exceptional,
not
to
find
something.
C
C
Say
two
things.
First
of
all,
just
to
me
it
seems
crazy
if
I
write
a
test
saying
you
know
if
a
equals
b
and
there
isn't
an
a
and
there
isn't
a
b
and
it
comes
back
saying
equal.
You
know
that
that
just
seems
deeply
wrong
to
me.
Having
said
that,
I've
I've
never
actually
done
anything
like
this
or
felt
the
need
to
do
anything
like
this
or
encountered
a
situation.
So
I
shouldn't
put
too
much
weight
on
my
own
opinion.
C
That's
why
I
think
we
were
looking
for
use
cases
and-
and
I
I
still
haven't
seen
any
if
other
people
are
strongly
convinced
that
this
should
come
out
to
be
equal.
B
What
about
arrays
with
with
holes
in
it,
sparse
arrays?
If
you
address
a
well
non-existing
an
element,
error
element
with
non-existing
index
or
an
index
with
non-existing
value?
What
is
resulting
also
an
exception.
C
G
B
E
So
most
use
cases
that
really
can
use
the
the
first
line
of
one
are
cases
in
which
properties
of
an
object
default,
and
so
you
essentially
don't
have
it
if
it
has
its
default
value.
So
if
you
compare
to
two
objects
with
respect
to
this
value,
both
will
be
undefined
and
they
will
compare
the
same,
which
is
exactly
what
what's
right,
but
again,
there
are
tons
of
other
cases,
but
the
the
absence
means
default
case
is
actually
covered
nicely
by
one.
G
Yeah
I
agree
since
json
is
used
for
serialization
or
data
representation
and
very
often
defaults
aren't
stored
or
aren't
present
in
that
serialization.
Then
that's
a
great
use
case
for
number
one.
C
G
I
E
D
C
C
C
C
Okay,
so
I
I
suggest
that
we
take
it
to
the
list
that
we're
going
to
treat
these
as
undefined
and
consider
them
to
be
equal.
Both
sides
are
absent.
B
We
also
need
to
put
examples
like
this
into
the
to
be
worked
out.
Example
section.
I
I
consider
this
very
important
example
section
marked
as
a
task
and.
E
Okay,
I
think
the
the
I
think
this
is
exactly
the
same
slide
I
had
in
one
drive,
so
there
seem
to
be
about
three
ways
to
handle
regular
expressions,
selecting
one
existing
regular
expression,
flavor,
providing
a
way
to
plug
in
regular
expressions
or
keeping
the
whole
subject
out
of
the
base
document.
E
E
E
The
the
first
question
that
came
up
was:
do
we
want
to
have
the
parsing
flavor
of
regular
expressions
that
that
originally
came
from
the
ed
editor,
where
regular
expressions
were
used
for
searching
on
a
line
or
do
we
want
to
use
matching
regular
expressions?
And
it
seems
to
me
that
the
prominent
example
for
parsing
regular
expression
is
ecmascript
and
a
prominent
example
for
a
matching.
Regular
expression
is
w3c
xsd,
regular
expression.
E
E
E
So
my
proposal
here
was
to
do
something
that
that
apparently,
jason
schema
or
people
also
have
advocated
building
a
modest
subset
and,
of
course,
in
a
way
that
this
actually
fits
with
both
camps.
So
I
actually.
G
All
right,
I'm
gonna
go
pull
it
up,
because
I
am
very
deeply
into
that
involved
in
that
spec.
G
E
G
G
B
C
The
best
way
to
go
at
this
moment
I
I've,
actually
changed
my
opinion
I
was.
I
was
of
the
opinion
that
this
was
just
too
hard
and
that
I
was
I
was
in
favor
of
leaving
it
out,
but
having
after
having
had
a
look
at
the
irex
document
in
its
recent
form,
I
think
I'm
also
with
with
stefan
here.
C
I
think
that
the
value
to
users
of
having
something
there
for
regular
expressions,
even
even
basic
regular
expressions,
has
has
a
lot
of
value,
but
I
think
that
there's
a
big
80
20
point
in
there
and
you
know
you
get
a
lot
of
value
from
basic,
regular
expressions
and
the
advanced
features
you
know
are
much
less
interesting
and
I
think
I
believe
very
strongly
that
our
chief
priority
in
in
writing.
C
This
document
is
that
if
somebody
builds
a
json
path
according
strictly
to
this
document,
they
have
a
maximum
hope
of
interoperability
and
and
for
all
of
those
reasons.
My
path
forward
at
this
point
preferred
would
be
to
to
to
develop
the
irex
document.
We
should
clearly
do
it
separately
because
other
people
will
probably
use
it
and
and
when
we
are
discussing
things
in
iregas,
we
should
have
a
prejudice
to
throw
it
out
if,
if
it's
difficult
or
complex
or
is
apt
to
be
a
source
of
interoperability
problems.
C
So
my
suspicion
is,
we
end
up
with
you
know:
irex
minus
minus.
We
still
probably
end
up
losing
a
few
things
that
are
that
are
in
there
now,
and
I
think
that
would
be
a
pretty
good
outcome
now
for
that
to
work.
G
Right,
so
what
I've
I've
pasted
a
link
to
the
the
schema
core
section
6.4
in
the
the
chat
to
this
meeting,
and
it
gives
some
guidance
for
schema
authors,
not
implementers.
G
E
E
D
E
B
Is
included
which
is
defined
by
jsonpath,
but
not
by
jsonpath,
but
by
regex,
which
is-
and
this
is
a
charming
aspect
from
I
regex-
that
it's
also
usable
by
other
drafts.
G
B
G
Over
trying
to
bend
over
backwards
to
implement
their
own
thing.
C
C
That's
fair,
you
know,
I'm
I've
been
in
this
seat.
I
have
a
strong
so
back
about
three
or
four
years
ago
now
we
were
building
a
product
called
step
functions
over
at
aws,
which
has
become
quite
popular
and
familiar.
We
didn't
think
about
this.
You
know
we
just
said:
oh
we'll
use,
jsonpath
right
and,
and
then
it
turns
out
that
it
just
depends.
C
C
So
maybe
I'm
being
overly
hurried
here,
but
are
people
comfortable
with
that
path
forward?
The
last
bullet
point
here
in
the
slide:
is
there
anybody
comfortable.
C
Okay,
so
next
good,
thank
you
greg.
That
leads
us
to
the
next
question.
Are
we
comfortable
with
having
a
separate
document,
and
I
regast
internet
draft
and
eventually
probably
rfc
as
opposed
to
wiring
it
straight
into
json?.
I
I
I
like
keeping
it
separate,
but
my
main
concern
is:
will
we
get
the
right
constituency
of
people
interested?
You
know
if
we,
as
jason
path
working
group,
start
this
red
xp
thing?
Will
it
emit
certain
other
interesting
parties?
You
know
how
do
we,
how
do
we
socialize
it?
How
do
we
draw
people
in
who
would
have
a
good
input
to
it.
E
So
the
advantage
advantage
of
having
it
in
a
separate
document
is
that
we
can
shop
this
around
with
all
the
people
who
already
have
regular
expression
scars
on
on
their
face,
and
I
think
that
that's
something
that
we
could
get
very
good
feedback
for.
D
Another
consideration
that
we
should
make
as
well
as,
if
we're
going
to
do
it
this
way
the
interdependency
we're
effectively
saying
that
we
can't
move.
We
can't
move
to
the
point
of
having
an
rfc
until
we've
got
the
the
regex
document
as
rfc
or
something
like
that.
Right.
A
G
Yeah
you
have
to
reference
the
rfc
in
in
the
jsonpath
dock.
I
would
probably
be
willing
to
say
that
the
the
jason
schema
folks
would
be
willing
to
also
or
would
be
happy
to
also
include
such
a
document
into
jason's
schema.
We've
been
having
some
discussions
on
what
specific
versions
of
of
regex.
We
should
be
supporting
with
the
upcoming
drafts,
because
what
we're
using
is
quite
old.
B
Build
the
example
section:
we
will
also
include
examples
with
regular
expressions
and
if
we
do
so,
we
we
need
or
we
we
can
reference
to
that
subset,
which
is
called
I
regex
and
and
then
tell
that
it
conforms
to
exactly
this.
C
D
C
E
Yeah,
I
think,
procedure
wise.
We
should
make
a
decision
now
confirm
it
on
the
mailing
list,
that's
different
from
taking
it
to
the
list,
which
means
we
are
not
ready
to
make
a
decision.
So
I
would
like
to
make
sure
this
is
captured.
E
Then,
if,
if
we
actually
go
for
a
separate
document,
then
right
now
we
have
a
hole
and
we
could
fill
this
hole
by
writing
a
new
document
or
we
could
fill
this
whole
by
adopting
an
existing
document
as
a
working
group
document.
Obviously
I
have
a
slight
preference
here,
but
that's
procedurally,
that's
the
decision
we
have
to
make
at
some
point
then.
E
So
I
personally,
I
think
the
readiness
of
iraqis
was
even
larger
than
the
readiness
of
of
jsp
the
json
path
document,
but
yeah,
that's
my
personal
view.
C
Yeah,
my
proposal
would
be
that
we
would
adopt
irex
0
0
as
a
working
group
document:
zero:
zero,
not
zero.
Two
are
you
up
to
zero?
Two?
Now?
Yes,
okay.
The
latest.
I
I
I
C
Of
course,
of
course,
this
is
an
interim
meeting
right,
we're
not
we're
not
decreeing
consensus,
but
we're
going
to
take
this
for
confirmation
to
the
wg.
E
C
E
C
E
That
is
not
really
an
issue
yeah.
I
think
that
the
main
remaining
issue
is
what
about
backslash
p
and
do
do
we
keep
this
in
and
if
yes
do,
we
make
any
further
restrictions
on
that,
because
that's
for
somebody
who
comes
from
from
a
non-ascii
version
part
of
the
world.
C
I
absolutely
agree,
I
don't
think
I
think
it's
totally
unacceptable
for
the
iatf
to
do
anything
that
is
ascii
limited.
I
would.
I
would
hope
that
the
iasg
would
would
kick,
which
is
such
a
draft,
I'm
a
little
bit
out
of
touch
with
what
the
implementations
do
do
they
have
good
general
coverage
of
like
the
unicode
character
classes
and
so
on.
E
C
Yeah,
I
I
I
agree,
but
is
the
is
the
universe
of
things
you
can
say
on
backslash
p
well
agreed
on.
E
Well,
the
the
only
two
the
current
draft
is
very
explicit
for
with
respect
to
what
you
can
put
in
a
back
hp,
and
I
think
that,
what's
in
there
is,
is
universally
supported
by
backslash
p
implementations
because
as
an
implementer,
essentially
at
some
point,
you
have
to
decide.
Oh,
oh,
my
god,
I
have
to
pull
in
the
unicode
character
table
and
once
you
have
that
character
table
it's
kind
of
trivial
to
do
all
the
things.
C
C
Okay,
now
I
think
this
is
in
a
good
place,
so
so
I
think
we
have
a
mental
work
item
figuring
out
what
needs
to
be
done
to
iregas
to.
C
E
There's
there's
one
issue
that
we
haven't
really
discussed
is
what
is
actually
the
syntax
we
put
around
those
regular
expressions,
so
there's
probably
something
like
equals
tilde
and
then
there
is
the
the
question:
do
we
actually
use
slashes
for
the
regex,
but
do
we
just
use
standard
string
syntax?
E
I
would
prefer
to
use
standard
string
syntax,
so
we
don't
have
one
more
problem
to
work
with
and
also
it's
it's
a
little
bit
of
an
alert
that
this
is
not
necessarily
compatible
with
existing
implementations
that
use
slashes.
C
Right
you're
correct,
that's
an
issue
and
I
think
it's
not
controversial
to
like
the
string,
but
the
compatibility
argument
is
strong.
If,
by
using
slashes,
we
would
get
regex
just
work
with
a
bunch
of
existing
implementations
that
would
be
very
attractive.
E
I'm
a
little
bit
more
on
the
side
of
trying
to
prevent
false
intervality.
C
Okay,
so
I'm
just
gonna.
Let's
just
take
a
note
that
we
need
to
work
on
the
embedding
sequence.
E
Okay,
so
this
has
been
said
a
few
times,
we
need
those
examples
and
there
is
an
open
issue
for
that
labeled
as
task-
and
this
is
just
a
reminder
that,
yes,
we
should
make
that
appendix
and
it
would
really
be
great
if
people
were
contributing
examples
we
can
put
in
there.
We
still
need
to
find
a
structure
for
that
appendix
so
I'm
not
necessarily
suggesting
that
you
put
them
in
as
a
pull
request,
but
sending
a
simple
message
to
the
mailing
list.
E
C
Are
there,
is
there
anything
we
can
steal,
I
mean,
are
there?
Are
there?
Are
there
good
example,
suites
out
there
in
open
source
places
or
other
rfcs
where
we
can
go
and
just
import
them.
B
G
G
C
And
see
if
anybody
else
has
provided
a
a
set
of
examples,
so
we
don't
have
to
create
them.
E
C
Yeah,
I
think
I'd
like
to
just
emphasize
this.
That
would
really
appeal
to
working
group
members
to
put
a
little
bit
of
work
into
this.
Go
find
some
things
that,
because,
for
two
reasons,
first
of
all,
I
believe
strongly
that
the
value
of
a
spec
is
is
greatly
increased
by
having
good
examples
in
it,
and,
secondly,
doing
this
often
reveals
disagreements
where
we
thought
we
had
agreement.
G
Yeah,
let's
paste
it
in
the
the
json
pointer
spec,
it's
got
some
measly
examples
down
at
the
bottom.
G
It's
rfc
6901,
yes,.
G
G
And
to
be
fair,
stefan's
original
blog
post
had
examples
as
well.
C
C
On
this,
because
you
know
the
per
itf
principles,
the
mailing
list
is
a
first-class
citizen,
but
I
would
agree
that
in
this
particular
case
it
would
be
nice
to
have
them
on
github.
E
Along
yeah,
speaking
of
json
pointer,
we
still
have
to
do
in
the
document
right
there
that
we
need
to
explain
what
the
difference
from
jsonpointer
is.
This
again
is
something
that
just
needs
to
be
done,
and
this
is
just
a
reminder
that
that
we
need
to
do
that.
I'm
not
aware
of
anything
controversial
about
that,
but
maybe,
after
writing
the
text
I
will
be.
G
I
C
E
Okay,
good,
so
if
there
is
nothing
controversial
there,
the
next
item
on
the
list
is
number
55,
which
I
think
is
essentially
covered,
but
this
this
was
input
that
that
is
a
year
old
and
I
think
we
we
actually
have
to
go
back
to
the
web
of
things,
people
and
ask
whether
they
are
now
happy
with
what
what
we
have.
E
So
we
have
to
find
the
right
point
in
time
when
we
actually
do
that.
We
can
do
this
at
working
with
last
call
time,
or
we
can
do
this
a
little
bit
earlier.
E
That
was
essentially
a
statement
of
interest
by
the
w3
web
of
things
project
and
and
a
couple
of
requirements
that
they
came
up
with.
G
C
E
Well,
it's
taken
us
at
face
value,
they
want
to
use
jsonpath
and
they
want
us
to
be
aware
of
that,
and
we
probably
need
to
to
talk
to
actually
find
out
whether
we
we
are
done.
C
E
Okay,
there
was
a
review
10
months
ago
and
we
maybe
want
to
ask
for
a
refresh
of
that
review
at
some
point
and
implicitly.
E
C
C
G
D
6901
seems
to
bring
reference
to
null
character,
unico
characters
and
that
implementer
should
handle
those
and
trying
to
summarize
the
other,
the
top
one
in
691,
basically
saying
that
dead,
end
pointers
and
the
implementers
should
anticipate
those.
But
I
think
that
we've
largely
covered
both
of
those
points
already
correct.
D
I'll
just
read
it
out
verbatim
in
jsonpointers
applications
using
jsonpoint.
You
should
anticipate
situation
where
a
reference
to
an
actual
json
value
is
not
guaranteed.
C
G
C
Well,
I
noticed
that
in
in
json
itself
in
8259-
and
this
has
always
been
true
way
back
to
the
very
first
specification.
It's
always
said
that
implementations
may
impose
limits
on
sizes
and
number
ranges,
and-
and
things
like
that-
and
you
know,
as
greg
said,
you
know,
deeply
nested
queries
or
something
like
that.
It
was
perfectly
reasonable
for
implementations
to
defend
themselves.
C
You
know
somebody
sends
a
json
path.
That's
three
megabytes
long.
G
One
to
look
at
would
be
graphql.
They
have
very
because
you
can
very
easily
do
loops
referential
loops
in
graphql
and
make
absurdly
long
queries
that
are
perfectly
valid.
So
what
they've
done
is
in
the
spec
they've
actually
said
that
implementations
need
to
support
like
a
max
depth
and
they
have
various
calculations
and
stuff
for
that.
But
their
security
considerations
are
fairly
robust
and
might
be
some
reference
material.
C
Yeah
I
mean
things
that
would
be
reasonable
in
you
know
a
simple
event:
filtering
thing
where
you
might
be
handling
very
high
volumes.
It
would
be
very
reasonable
to
have
quite
modest.
You
know
limits
on
that,
whereas
for
somebody
who's
doing
large-scale
document
transform
work.
D
I
So
someone
could
sort
of
turn
all
the
knobs
to
zero
and
miss
out
points
of
the
spec
and
still
claim
an
interoperable
implementation.
C
C
In
fact,
in
in
in
in
json,
it
actually
does
give
you
some
guidance.
Specific
guidance
such
as
you
know
assume
that
numbers
are
basically
javascript
numbers,
which
means
you
know,
754
float64
and
don't
assume
you
can
do
any
more
range
or
precision
than
that
and
that's
useful
guidance.
C
So
if
somebody
could
think
of
specific
things
to
watch
out
for
those
would
be
welcome.
C
I
F
C
Once
you've
accepted
the
jsonpath
as
being
syntactically,
okay,
that's
a
really
good
point
glenn,
because
we
should,
in
fact,
if
we
are
going
to
allow
implementations,
to
express
these
kind
of
limits,
we
could
say
that
they
can
only
express
limits
that
can
be
detected
at
compile
time.
Basically,
so.
C
E
We
cannot
do
anything
about
that,
but
silently
failing,
for
instance,
is
probably
not
a
good
thing,
because
if
you
rely
on
the
absence
that
the
jason
path
expression
told
you
that
that
can
lead
to
grave
mistakes.
C
But
nonetheless,
I've
taken
note
of
glen's
point
here
and
I
think
it's
correct
is
that
if
we're
going
to
allow
limits,
there
should
be
limits
that
can
be
detected
at.
I
don't
know
if
we
have
a
term
for
this,
but
you
know
at
parsing
time
right.
E
C
G
My
issue
with
it
from
an
implementation
from
an
implementer
point
of
view,
is
that
if
I
have
I
I
understand,
let
me
start
off
by
saying
I
understand
the
there
being
a
distinction
between
values
and
nodes,
but
even
if
I'm
processing
a
specific
value
and
that
value
occurs,
let's
say
that
value
is
an
object
with
the
same
keys
and
the
same
values
for
those
keys.
G
If
I'm
processing
that
object
once
I've
processed
that
value,
I
shouldn't
have
to
process
that
value
again.
I
can
handle
that
internally
in
my
implementation,
by
saying:
oh,
here's
a
here's,
the
hash
for
that
object,
and
I
don't
have
to
I
don't-
have
to
process
that
object
again.
I
just
go
fetch
its
results
and
insert
it
in
in
here.
If
we
want
duplication
or
if
we
want
to
filter
out
the
duplicates,
then
I
can
know
to
do
that,
but.
C
Right,
but
from
this,
but
from
the
viewpoint
of
somebody,
who's,
writing
the
json
path
and
then
processing
the
results
that
come
back
from
you
know
executing
it
at
what
point
is
their
life
can
be
made
better.
H
G
I
mean
it
depends,
do
they
want,
you
know
what
do
they
want?
Do
they
just
want
the
values
in
which
case,
or
do
they
just
want
the
resulting
values,
in
which
case
duplicates
don't
really
help
them?
Do
they
want
the
duplic?
Do
they
want
nodes,
in
which
case
duplicate
nodes,
don't
help
them,
but
duplicate
values
could
or
do
they
want,
like
the
raw
selection
set
of
of
the
query
which
do
showing
the
duplicates
would
give.
E
The
the
context
is
not
using
the
term
node
set,
except
where
we
talk
about
xpath,
so
it
has
a
node
list
and
that
to
me
implies
that
there
is
no
duplicate
detection
going
on,
and
so
we
we
could
save
this
explicitly,
which
probably
would
help
with
not
getting
that
question
too
often.
E
B
B
B
C
C
B
The
proposal
is,
don't
don't
remove,
duplicates
in
this
slide.
Is
this
correct
exactly
yes,.
I
G
Can
we
put
it
in
the
spec
that
it's
that
implementers
or
implementers
should
or
are
required
to
have
an
option
like
a
switchable
option,
and
then
we
specify
a
do?
We.
C
Have
any
other
switchable
options
because
start
right.
A
G
I
I
think
the
optimization
you
suggested
greg
of
having
a
kind
of
memento
pattern
and
and
saying
I've
already
computed
the
answer
to
this
I'll
reuse.
The
result
it's
perfect
because
it's
it
conforms
to
the
the
spec
without
complicating
the
spec,
and
it
gives
the
kind
of
performance
that
maybe
your
users
are
looking
for.
It's
internal.
C
G
C
D
C
The
json
path,
by
the
reasons
that
tell
whether
to
expect,
duplicates
or
not.
B
In
fact,
removing
nothing
is
the
mud,
the
most
intuitive
thing,
and
so
jsonpath
does
what
what
it
is
supposed
to
do
and
a
lot
more
makes
sense.
Yes,.
B
The
most
minimal
effort
to
do
and
the
outcome
corresponds
to
exactly
this
philosophy.
C
C
C
So
so
greg,
I
think,
you're
you're
kind
of
the
outlier
here,
but
I
think
what
what
I
think
would
help
is
a
proposal
about
see,
because
I
think
what
a
lot
of
what
we're
worried
about
is
how
complex
it
would
be
to
specify
what
happens
and
what
to
expect.
C
G
Okay,
I
think
the
most
complex
bit
will
be
the
the
final
output
and
tying
that
into
the
the
the
ordering
that
we
had
had
talked
about
previously,
like
node
ordering
in
the
output.
G
B
G
Well,
I
mean
because
we
had
previously
talked
about
the
the
sequencing
of
the
output,
because
we're
do
you
do
the
the
node
and
then
it's
children
and
the
children
or
the
children.
C
Okay,
so
my
my
feeling
is
that
at
the
moment
we
are
probably
you
know
absent
some
new
thinking
on
this
we're
going
to
go
forward.
I've
also
heard
that
we
need
to
make
sure
that
we
have
examples
that
make
it
clear
that
duplicates
can
come
back.
E
B
G
G
If
we
have
defined
union,
if
we
have
since.
D
B
Okay,
if
you
are
confused
because
of
it,
may
mean
that
we
we
indeed
need
more
description,
but
it
may
also
be
a
historical
confusion,
so
yeah
users.
G
B
So
do
you
suggest
the
the
naming
is
that
could
be
a
better
naming
renaming.
C
C
B
C
C
And
right
here
it
explicitly
says
note
that
any
node
selected
more
than
one
of
the
union
selectors
is
kept
as
many
times
in
the
node
list.
That's
not
all
that
yeah!
That's
not
that
so.
We've
actually
been
quite
clear
here
about
duplicates.
In
this
case.
I
think.
E
C
B
G
E
C
It's
it's
not
perfect,
but
I
prefer
it
to
union.
The
word
union.
G
G
D
G
G
E
G
G
C
Yeah
so
because
that's
true,
we
explicitly
wrote
a
rule
saying
you
can't
do
that
so
so
I
suggest
this.
I
suggest
that
we
go
ahead
and
look
at
the
actual,
you
know
make
a
pr
and
see
how
it
reads
once
it's
actually
in
front
of
us.
E
G
I
would
be
happy
with
slice
selector
I
I
proposed
using
index
because
that's
that's
what
the
square
brackets
mean
to
me.
It's
an
indexer.
E
Slice
selector
has
enclosing
square
brackets
and
this
is
meant
to
not
have
those
included
enclosing
spare
square
brackets,
so
we
actually
have
to
take
apart
the
production
for
slice
selector
into
the
brackets
and
the
slice
index
that
is
inside
the
slide
selector,
and
then
we
can
use
the
slice
index
inside
the
list
selector
as
well.
F
E
Yeah,
the
the
disadvantage
of
of
the
current
selected
terminology
is
that
the
thing
that
is
in
front
of
the
term
selector
sometimes
describes
the
result
and
sometimes
describes
the
form
of
the
selector.
So
a
filter
selector
doesn't
give
you
a
filter
at
the
end,
it
gives
you
it
employs
a
filter
and
similarly
a
list
selector
doesn't
give
you
a
list.
Well,
it
actually
does,
but
this
is
not
the
one
that
the
selector
is
about.
It
employs
a
list
to
to
create
the
selector.
E
C
B
E
Okay,
so
time
is
running
short
a
bit,
so
let's
go
to
the
next
one.
E
Yeah,
there
is
one
of
the
issues
that
I
didn't
dare
to
close,
but
really
it's
a
won't
fix.
We
currently
don't
have
an
automatic
commit
check
for
the
abn
f
and
we
we
we
don't
need
it.
E
Yes,
I
think
the
summary
is
already
in
there,
so
the
next
steps
that
I
would
propose
is
actually
processing
these
issues
here,
of
course,
doing
doing
another
sanity
check
after
we
have
done
the
the
processing
and
then
do
the
working
blast
call,
because
that's
the
only
way
to
get
people
to
actually
review
review
things
in
detail,
and
we
also
should
send
this
working
plus
call
to
a
number
of
places.
C
So,
are
you
suggesting
we're
not.
C
E
E
So
my
timeline-
I
I
didn't-
I
usually
use
a
rose
colored
slide
for
timelines
because
I
need
to
wear
my
rose
colored
glasses
to
to
do
timelines.
My
timeline
would
be
that
we
get
this
written
up
within
about
three
weeks
and
then
we
have
a
document
where
we
do
the
sanity
checks.
Maybe
we
need
to
re-spin
this
once
more,
but
we
should
do
the
wrecking
blocks
call
within
february
and
and
preferably
early
february,.
C
C
I
C
E
D
Yeah,
I
think
we
already
know
that
that
complicates
it
with
other
working
groups.
Referencing
other
documents
in
their
own
working
groups
that
are
incomplete
is
fine,
but
there's
a
timing,
dependency
thing.
C
D
D
E
D
So
ietf
113
is
in
march
in
vienna.
It
sounds
like
we
we've
got
about
what
six
eight
weeks
until
then,
I'm
unsure,
if
it
makes
sense
for
us
to
do
another
interim
before
then,
so
we
have
a
couple
of
options.
D
D
I
know
this
has
caused
issues
in
the
past
with
some
people.
I
guess
what
I'm
liking
fielding
from
the
working
group
is.
Should
we
have
a
meet
at
113
it
just
just
to
be
clear.
It's
unlikely
that
I'll
be
able
to
attend
in
person
because
of
corona
and
scheduling.
F
And
such
so
whatever
it
is,
it
will
be
a
largely
online
affair,
I
presume,
but
just
just
filling
thoughts
from
the
working
group.
What
do
you
all
think.
C
You
know
the
ihf
is
our
hosting
organization
and
when
you
sign
up
to
do
a
working
group,
there's
sort
of
an
implicit
commitment
to
give
people
who
are
attending
iitf
a
chance
to
wander
in
and
hear
what's
going
on
so
so
now
we
had
the
unfortunate
experience
of
the
where
we
tried
to
do
one
and
it
was
on
north
american
time
and
basically
nobody
showed
up.
But
this
is
going
to
be
on
european
time.
E
Yeah
there
are
working
groups
that
never
go
to
itf
meetings
and
that
only
do
interims
like
the
seller
working
group.
So
this
is
this
is
not
unheard
of.
The
advantage
of
going
to
an
itf
meeting
is
that
it
might
be
easier
to
get
other
people
pulled
in,
but
also
there
might
be
a
conflict
in
this
slot,
but
which
makes
it
much
harder
to
pull
other
people
in
and
the
the
advantage
of
going
outside.
C
E
C
The
working
group
is
not
wildly
enthusiastic
about
having
a
meeting;
I'm
certainly
not
going
to
insist
james.
What
do
you
think.
D
Or
book
another
interim
meeting
around
the
time
of
around
about
the
same
time,
either
before
or
after.
E
D
Okay,
the
only
thing
I'm
working
then
is
draft
submission.
If
that's
going
to
interfere.
When
is
that
going
to
be.
D
Yep
so
that
itf
113
is
19
sorry
week,
starting
the
21st
of
march,
so
the
latest
we
can
do
it
is
weeks
starting
monday,
the
7th
of
march.
F
C
Week,
starting
either
the
21st
of
february
or
the
28th
of
february.
E
No,
no
not
february,
we,
we
are
still
waiting
for
the
working
rubles
call
to
complete
at
that
time.