►
From YouTube: IETF-JSONPATH-20230110-0800
Description
JSONPATH meeting session at IETF
2023/01/10 0800
https://datatracker.ietf.org/meeting//proceedings/
A
C
B
B
A
F
C
It
away
okay
for
the
recording
for
later
this
is
the
Json
cloth
working
group
meeting
for
January
2023
I'm,
your
host
change
with
Tim
and
my
co-chair,
since
this
is
an
official
ITF
meeting.
The
note
well
applies
here.
These
went
well
note,
taking
will
usually
happen
as
it
usually
does.
I
will
get
done
later.
Blue
sheets
are
automatic,
there's
only
a
handful
of
us,
so
we
don't
need
to
worry
about
the
chat
too
much.
C
C
F
C
C
D
D
B
Well,
the
the
difference
in
the
Json
path.
Excuse
me
in
the
JSO
usage
is
that
JSO
already
uses
anchors,
so
they
they
would
have
to
Define
how
how
anchors
are
used
and
since
they
essentially
are
tied
to
the
ecmascript
definition
of
regular
Expressions.
The
I
think
the
the
usefulness
of
iRig
X
for
Json
schema
org
is
that
it
provides
a
boundary
of
for
for
the
gamut
of
functionality
that
you
can
expect
to
be
interoperable.
So
you
you
won't
use
it
literally.
G
C
B
B
C
D
B
B
Foreign,
yes,
so
my
summary
of
this
issue
would
be
we.
We
are
copying
the
Json
Syntax
for
Unicode
escapes,
and
we
note
that
JavaScript
actually
has
added
backslash,
you
brace
chord
Point
Breeze
as
an
alternative
syntax.
So
you
don't
have
to
do
the
arithmetic
for
splitting
up
a
non-bmp
character
into
two
surrogates
and
what
remains
from
this
issue.
F
B
B
D
B
B
B
Yeah,
so
the
issue
was
that
the
current
grammar
is
is
very
liberal.
You
can
essentially
put
in
a
space
everywhere,
except
inside
tokens
like
like
strings
or
numbers,
and
the
question
was:
do
we
want
to
take
away
some
of
these
opportunities
to
in
certain
spaces,
because
I
mean
every
single
one
of
those
is
a
potential
interability
issue.
B
You
have
to
test
all
of
those
in
your
implementation,
but
on
the
other
hand,
it
turns
out
that
generally,
when,
when
you
try
to
remove
one
of
them,
it
turns
out
that
that
you
really
need
lack
of
imagination
to
be
sure
that
you
actually
can
remove
that.
So,
for
instance,
if
we
look
at
the
the
change
that
that
is
on
the
page
on
on
child
longhand,
of
course
you
wouldn't
put
spaces
within
the
the
brackets.
C
I
I
would
not
wearing
my
chair.
I
would
I
would
agree
that
we
shouldn't
get
rid
of
what
optional
white
space
simply
for
the
usability
issue
of
of
people
from
fumbling
around
trying
to
figure
out
how
to
write
their
queries.
And
if
it's
big
and
complex
enough
they'll
end
up
needing
they'll
want
to
put
in
White
space
in
there
to
figure
out
how
it
works
and
by
not
having
support
for
the
spaces.
Then
that
means
implementations
will
have
to
do
some
sort.
D
G
B
F
F
D
G
D
B
D
B
F
G
F
B
B
D
G
D
G
D
A
D
Okay,
how
about
this
slightly
strange
scenario?
I
suppose
there
was
a
function,
implementation
which
just
crashed.
You
know
memory
whatever
occasionally,
and
someone
wanted
to
code
a
guard
on
a
function
on
a
predicate
in
a
filter
expression
to
avoid
that
happening,
and
you
know,
and
because
we
reordered
the
the
evaluation,
because
an
implementation
reordered
the
evaluation
it
crashed
anyway
didn't
check
the
guard.
First.
B
D
F
D
B
D
B
D
Well,
it's
not
specification,
but
it
would.
It
would
help
it
would
get
around
the
the
problem
with
that
particular
implementation.
So
you
know
they're
up
against
it.
They've
got
an
issue
in
production.
This
thing
is
crept
out
of
the
woodwork
or
they
can't
upgrade
the
jsonpath
implementation
because
the
fix
isn't
available
and
they
put
a
guard
in
and
it
works
some
of
the
time
and
it
works.
You
know
not
all
the
time.
D
B
D
B
D
D
G
G
D
D
F
B
D
D
Yeah,
so
this
was
behavior
that
we
added
recently
because
the
match
and
search
functions
needed
a
comparison
to
true
or
similar
to
be
usable.
It
wasn't
possible
to
just
say,
question
mark
match
in
the
filter
expression,
so
we
introduced
the
optional
Boolean
to
to
deal
with
that.
Thank
you,
Karsten,
but
the
existence
test
text
still
talks
about.
D
B
D
Yeah
I'm
just
concerned
concerned
about
referential
transparency.
If,
if
you
have
enough,
if
you
have
a
match
that
returns
a
true
value,
how
is
that
different
from
any
other
true
value,
for
instance,
dollar
dot
Foo
with
a
Boolean
value,
true
in
an
existence
test
or
dollar
dot
Foo
with
with
value
false?
More
importantly,
in
the
existence
test.
D
B
Well,
yeah:
we
have
special
casing
in
the
sense
that
a
function
that
has
a
Boolean
return,
signature
actually
is
interpreted
in
a
different
way
than
a
value
that
we
cannot
get,
because
we
only
look
at
the
existence
of
the
way
we're
not
the
value
of
the
value.
So
yes,
it's
a
special
case,
but
it's
a
well
protected
special
case.
You
cannot
really
confuse
that,
so
you
could
put
in
an
identity
function
just
for
for
for
fun.
D
D
F
B
D
B
C
D
D
G
B
Yeah
there
are
two
elements
to
this
discussion:
one
is
that
the
the
grammar
implicitly
allows
nested
expressions
and
we
probably
should
be
explicit
about
that
and
the
other
one
is.
Do
we
get
to
reference
the
the
equivalent
of
the
ad
design
in
the
higher
enclosing
expressions
and
to
me
it
seems
the
discussion
indicated.
This
probably
would
be
a
nice
feature
to
have,
but
it's
also
not
that
easy
to
come
up
with
an
example
that
actually
needs
that.
G
G
G
G
To
let
focus
on
this
problem,
but
it's
quite
late
to
the
party
we
we
should
defer
this
for
the
future.
So
another
point
is
what
shall
we
do
with
with
empty
filter
Expressions?
So
when
you
have
a
bracket,
a
question
mark
and
and
the
closing
bracket
immediately
don't
be
allowed.
Is
this
it?
It
might
return
false
as
a
it's
in
a
syntax
error,
so
syntax
error?
B
G
B
G
D
F
B
B
G
D
B
G
F
B
F
D
D
B
G
B
I
I
was
thinking
of
maybe
using
the
the
tooling
that
I
have
for
ABN
F
to
write
a
various
simple
tool
that
allows
us
to
dump
a
set
of
examples,
both
positive
and
negative
examples
into
the
ABF
and
see
if
it's
properly
accepted
or
rejected,
so
that
that
will
be
a
few
lines
of
code
and
I.
I
could
write
that
word
or
we
could
figure
out
how
to
use
the
existing
avnf
tools
to
do
that.
D
B
D
G
I
was
stumbling
yesterday
about
a
child's
longhand
child
segment
and
child's
longhand
and
dot
white
card
shorthand
and
the
same
at
the
descendant
segment.
This
is
a
descendant
segment
and
descendant
child.
This
is
more
logical,
more
more
more
clear
and
I
was
thinking
about
how
to
rename
the
child
segment
similar,
but
I
had
not
good
solution.
I
would
try
to
think
about
this
a
little
longer.
This
is
what
I
mean
about
going
through
the
ABN
f
one
more
and
to
support.
B
Yeah,
so
we
did
the
usual
thing:
we
we
made
the
changes
that
are
needed,
but
we
need
to
reflector
step,
but
let's
do
that.
Reflector,
step
and
I
think
it's
probably
best
if
we
do
this
in
small
steps.
So
so
each
of
us
will
propose
some
improvement
instead
of
having
a
monster
PR
that
changes
everything
and
then
needs
to
merge
with
another
monster.
Pr
that
also
changes
everything,
but
slightly
differently.