►
From YouTube: IETF110-JSONPATH-20210310-1200
Description
JSONPATH meeting session at IETF110
2021/03/10 1200
https://datatracker.ietf.org/meeting/110/proceedings/
A
A
We
may
have
to
just
start
without
him,
and
I
hope
that
he
joins
us
before
we
make
any
sort
of
real
progression
just
so
that
we
can
stick
to
time.
A
A
A
Okay,
I've
tried
to
call
tim
and
I
couldn't
get
through
so
let's
just
let's
just
get
started,
because
this
shouldn't
be
an
overly
complicated
affair
so
good
day,
people.
This
is
the
jason
path
working
group
for
ietf
110.
A
just
before
we
get
started
the
usual
you've
probably
seen
this
before,
but
but
I
have
to
do
the
the
reminder.
This
is
the
note.
Well,
it
applies
to
this
meeting
as
it
does
to
all
other
ietf
meetings.
A
A
Can
I,
before
we
get
into,
I
guess
the
meat
of
the
discussion?
Do
we
have
somebody
that
can
be
a
note
taker?
Please.
A
If
you
send
it
out
afterwards,
I'm
sure
we
can.
We
can
help
out
with
anything
that
you
miss.
Thank
you.
As
for
a
javascribe
I'll,
given
that
we're
we're
quite
a
small
group,
I'm
more
than
happy
to
do
that.
Is
there
any
agenda
bashing?
Is
there
anything
that
anybody
has
a
particular
subject,
they'd
like
to
bring
up
before
we
get
started.
A
Nope,
okay,
the
agenda's
fairly
minimal
today,
usually
I
would
expect
him
to
be
driving
this
he's,
unfortunately
not
here,
but
I
guess
maybe
is
a
good
starting
point.
If,
if
our
editors
and
those
who've
been
working
hard
on,
the
draft
could
perhaps
provide
some
sort
of
informal
update
about
where
things
are
when
you
you
recently
have
had
submitted
the
renaming
of
the
draft.
B
Okay
not
used
to
the
protocol.
I
can
just
summarize
where
we're
at
I
mean
a
while
back.
We
we
formed
the
working
group
and
got
a
emerged
draft
together
of
the
draft
that
I'd
been
editing
and
the
one
that
stefan
gerstner
and
carson
have
been
working
on,
and
so
we
got
that
merged
draft
and
there's
been
a
little
bit
of
editorial
work
on
it
and
that's
now.
B
The
current
base
zero
zero
version
in
the
documents
area
on
data
tracker,
but
more
recently,
there's
been
a
flurry
of
activity
on
the
mailing
list
and
in
issues
discussing
various
aspects
gradually
coming
to
conclusions
in
various
places,
but
a
lot
of
discussion
going
off
in
various
tangents.
So
I'm
not
sure
whether
we're
converging
or
diverging
at
the
moment,
but
there's
certainly
a
lot
of
discussion
going
on,
but
no
sign
of
any
pull
requests.
So
there's
no
writing
going
on.
B
A
Okay,
okay,
it
probably
doesn't
help
us
that
neither
stefan
nor
marco
are
here
that
makes
this.
That
makes
this
meeting
a
little
bit
difficult
to
sort
of
make
any
sort
of
progress.
A
I
understand
this
there's
currently
a
handful
of
issues
that
are
that
remain
open,
some
of
which
are
more
contentious
than
others.
A
D
D
Editorially
with
with
the
document,
but
also,
I
think
we
we
really
need
to
get
a
few
issues
out
of
the
way
that
that
really
need
to
be
decided
before
we
can
make
such
a
substantial
further
progress
and
for
me
right
now,
the
most
important
issue
is
what
is
exactly
going
to
be
the
processing
model
that
we
assume
and
we
have
had
some
discussions
on
on
the
list.
D
But
I
think
we
we
haven't,
really
managed
to
structure
these
discussions
yet
in
a
way
that
that
leads
to
progress-
and
I
think
that
that's
really
the
next
big
item.
But
I'm
I'm
also
aware
that
there
are
lots
of
little
issues
that
that
are
that
do
merit
working
on
on
them
on
their
own.
A
Okay,
that,
that's
that's
very
that's
a
very
good
update.
Thank
you
for
that
carsten
in
terms
of
the
processing
model,
as
you
described
it
is
that,
to
me,
sort
of
sounds
like
an
effort
for
perhaps
is
that
something
that
we
could
discuss
in
in
this
time
slot
now
and
make
some
sort
of
meaningful
discussion
about,
or
is
that
something
you
think
we
should
take
to
the
mailing
list.
D
Well,
I
think
it's
always
better
if
it's
carefully
prepared
if
a
discussion
is
carefully
prepared,
but
given
that
we
don't
have
that,
may
we
could
do
a
little
bit
of
impromptu
discussion
and
and
see
where
we
are
with
respect
to
that.
A
Sure
I'm
gonna
take
your
lead
on
that,
because
I
think
that
you're
far
more
informed
on
that
than
I
am.
D
Thank
you
for
throwing
back
the
ball
yeah.
So
do
you
want
to
do
this
now?
Are
there
other
things
you
want
to
do?
First,
I
don't.
A
D
Okay,
so
there
are
23
issues
on
on
the
github
repository
that
are
marked
as
open,
and
I
think
we
haven't
really
managed
to
to
label
them
appropriately
and
we
haven't
really
managed
to
to
boil
them
down
to
issues
that
are
narrow
enough,
that
you
actually
can
resolve
those.
D
But
there
are
issues
like
sequence
of
results,
duplicates
and
selector
output,
and-
and
maybe
the
whole
discussion
about
the
the
term
union
that
that
really
point
to
to
some
some
uncertainty
about
what
the
processing
model
is
now,
given
that
that
jsonpath
has
thank
you,
given
that
that
jsonpath
has
learned
from
from
xpath
in
its
youth,
I
think
it's
interesting
to
look
at
the
the
xpath
processing
model,
which
is
really
about
node
sets,
as
as
the
the
most
important
data
structure
that
that
is
being
used
at
least
an
abstract
fashion.
D
And
then
you
have
steps
that
go
through.
These
note
sets.
So
the
whole
thing
has
a
really
long
story.
Much
of
this
was
developed
in
in
the
late
80s
early
90s
in
the
context
of
the
diesel
project
document,
style,
semantics
and
specification
language
in
in
iso
iesc.
D
But
basically,
the
idea
is
that
you
have
ways
to
to
step
from
the
nodes
that
that
you
are
currently
sitting
on
and
that
you
have
ways
to
to
select
subsets
and
and
compute
values,
and
at
the
time
we
had
an
expression
language
that
was
based
on
this.
That
is
what
that
was
doing,
all
these
things,
that
that
got
turned
into
a
special
thing
for
xpath
and
and
xslt.
D
But
I
think
basically
the
idea
is
the
same,
but
the
the
important
part
to
make
this
whole
thing
work
is
that
you
need
to
know
where
you
are
and
and
what
actually,
the
the
how
the
sequence
of
steps
actually
are
connected
together.
So
what
are
the
intermediate
results
and
for
xpath?
That's
a
pretty
obvious
structure.
D
That's
the
the
note
set,
so
you
essentially
identify
a
subset
of
the
set
of
positions
that
are
defined
in
the
document,
so
the
node
set
has
a
few
interesting
properties,
and,
and
one
of
them
is
that
it
actually
does
not
record
duplicates.
D
So
you,
you
only
can
mark
a
note
once
in
a
note
set,
so
that
automatically
gives
you
one
property
which
is
interesting
and
also
the
the
set
is
not
by
itself
ordered,
so
that
there
is
no
aura
information
left
by
by
the
various
operations.
I
hope
somebody
is
taking
good
notes.
A
Jeffrey
is
he's
also
asked
a
question
asking
if,
if
we've
got
a
specific
issue
in
the
issues,
I
presume
jeffrey
you're,
referring
to
the
processing
model,
is
that
correct.
D
Okay
yeah,
so
just
judging
from
the
title,
I
think
27
is
one
of
them.
23
is
one
of
them.
21
is
one
of
them.
I
think
even
15
is
14
and
15
are
one,
so
these
are
just
different
ways
of
of
approaching
the
elephant
and
and
finding
trunks
and
legs
and
so
on
off
of
the
overall
question.
What's
the
processing
model.
A
Okay,
with
with
those
issues
you've
identified,
we
can
just
quickly
create
a
label
and
attach
it
to
the
ones
that
that
you
think
fall
under
that
and
then
so
that
we
can
collect
them
together.
If
that's
useful
sounds
good.
D
D
The
the
first
one
that
that
really
brought
up
the
problem
to
to
the
service
yep.
A
A
D
No,
I
cannot
say
that,
but
I
can
try
to
find
them
again,
so
I
think
15
at
least
touches
on
this
21
23
and
27
touch
on
these.
A
D
Yeah,
I
must
admit
that
I
added
some
labels
in
hoping
that
I
would
spawn
a
process
where
other
people
add
labels,
but
that
hasn't
happened
yet,
maybe
because
people
are
unsure
about
what
what
they
are
allowed
to
do
here
and
and
what
not
and
I'm
kind
of
on
the
wikipedia
be
bold
side,
you
will
be
called
back
if
you
do
something:
that's
not
appropriate.
A
Given
the
amount
of
participation
that
we
have
is
fairly
minimal,
the
and
thank
you
for
the
link
in
the
chat
christina,
given
the
amount
of
we've
got
fairly
a
small
community
here
in
this
working
group,
I
would
say
just
go
ahead
and
do
stuff,
obviously
in
a
considerate
manner.
So
if
people
need
access,
I
need
to
twiddle
things
to
help
you
get
this
work
done.
I
will
do
it.
That's
that's
that's
my
role.
D
Well,
definitely
because
how
this
thing
that
replaces
the
note
set,
which
I
have
called
collection,
just
as
a
placeholder
word,
but
how?
How
that
is
structured,
actually
tells
you
how
what
the
fate
will
be
of
duplicates.
A
Yup
no,
this
looks
like
it.
We've
just
got
some
ongoing
discussion
in
there,
so
21
23
27.
I
think
I've
covered
all
of
them
except
14.
A
Include
okay,
so
it
sounds
like
one
of
the
actions
we've
now
got
is
to.
I
guess:
try
to
triage
these
a
little
bit
further.
Okay,
I'm
trying
to
think
about
how
we
should
keep
going
with
this.
D
B
Okay,
yeah
I'll
buy
it,
so
it
depends
on
the
type
of
reference
that
the
node
set
makes
into
the
into
the
json
document.
That's
input!
How
did
you
see
that
going?
Would
you
would
that
be
a?
Would
it
point
to
a
node?
Would
it
be
a
reference
to
a
node
in
some
sense
in
in
the
original
document
or
a
copy
or
what
or
a
value?
No,
it's
a
reference
reference.
Okay,.
B
Yeah
I
mean
certainly
in
the
in
my
input
draft.
The
processing
model
I
had
in
mind
was
that
we
built
up
a
an
ordered
list
of
elements.
As
we
went
through
the
jsonpath
selector.
B
You
know
the
components
of
the
selector
and
and
didn't
remove
duplicates,
and
I
just
had
in
mind
efficiency
really
and
sensitivity
to
the
current
implementations.
I
don't
think
there
are
many
if
any
implementations
that
remove
duplicates
so
that
that
was
my
goal,
but
certainly
having
a
maybe
not
a
node
set,
but
a
node
list.
You
know
an
ordered
collection
of
nodes
would
satisfy
that.
B
It
wouldn't
have
to
be
values
because
they're
immutable,
so
reference
into
the
original
structure
would
be
helpful
and
also
it
would
tie
down
the
fact
that
the
intermediate
values
passing
through
were
all
sub-nodes
sub-trees
of
the
original
document.
B
Yeah-
and
you
know
admittedly,
there's
a
trade-off
there,
because
from
the
user's
perspective,
they
may
prefer
not
to
have
duplicates,
but
it
depends
how
quickly
they
want
the
result
to
be.
A
Okay,
is
there
any
other,
I'm
trying
to
think
if
there's
any
sort
of
value
of
us
going
through
the
rest
of
the
issues
that
are
here,
make
most
of
the
time
that
we've
got
to
discuss?
All
of
this.
D
The
next
step
would
be
to
actually
go
through
these
issues
with
the
assumption
we
are
using
a
node
list
and
see
whether
the
examples
that
have
been
brought
up
in
these
issues
actually
bring
up
problems
with
that
or
advantages
of
of
doing
that.
That
may
be
a
bit
tedious
to
do
in
the
meeting.
So
I'm
wondering
how
to
do
this.
So
if
I
was
doing
this
in
a
classroom,
I
would
just
divide
the
students
into
small
groups.
D
A
Yep,
that
would
be
the
only
challenge
that
we
have
is
that
our
editors
aren't
about
so
which
I
met.
I
kind
of
ran
on
the
assumption
that
they
would
be
in
the
best
position
to
do
this,
but
this
is
certainly
something
that
that
I
can.
I
can
do,
and
I
guess
if
we
can
get
some
help
from
other
members
of
the
working
group
who
have
prepared
to
look
at
a
couple
of
issues.
D
Yeah,
I'm
afraid
mideco
doesn't
have
breakout
rooms
which
I
would
use
for
that,
so
it's
hard
to
to
duplicate.
We
have
done
this
in
this
new
thing
to
think
research
group,
occasionally
that
we
just
we
all,
went
into
one
of
the
four
corners
of
the
room
and
and
try
to
get
some
some
work
to
get
done
and
then
came
back
together
and
reported
on
on
what
the
four
corners
found.
But
you
cannot
do
this
in
mid
echo.
A
Yeah
I
mean
we've
got
this,
there's
23
issues
here
and
I
don't
think
all
of
them
apply
to
what
what
we've
agreed
about
with
regards
to
the
node
set.
A
A
Yes
jeffrey
if,
if,
if
folks
would
like
to
help
out
with
regards
to
sort
of
triaging
these
priority
stickers
as
carson
described,
that
would
be
that's.
That's
sort
of
the
ask
here.
A
A
If
there's,
if
there's
anybody
who'd
like
to
help
out
with
that,
please
speak
up
now.
B
Are
we
talking
about?
Are
we
talking
about
doing
in
real
time,
taking
a
break.
A
I
think
that
would
be
a
bit
too
finicky
and
about
as
amusing
as
paint
drying.
I
think
this
is
one
of
those
things
where
we
go
away
offline
and
have
a
look
at
this,
because
it
may
stir
ongoing
discussion
that
may
need
to
come
back
to
the
list
or
so
on
and
so
forth.
A
But
it
sounds
like
that.
We've
got,
we've
got
a
bit
of
an
action
there,
and
this
is
something
that
I'll
I'll
talk
to
tim
go
ahead.
Jeffrey.
C
If
we
don't
have
other
things
to
talk
about
in
particular,
then
it
might
actually
be
a
good
use
of
time
to
like
assign
assign
an
issue
to
to
each
person
who
volunteers,
okay,
stop
talking
for
five
minutes
and
then
come
back
and
report
back
on
the
issues.
C
A
D
Well,
I
think
I
talked
over
you
I'll
take
issue
14.
E
A
That's
now
assigned
to
you
jeffrey:
do
you
have
a
if
you
don't
have
a
choice?
I
will
give
you
15.
C
A
B
Can
I
just
interrupt,
I
mean
I
think
signing
and
one
of
these
issues
to
someone
at
this
point
is
probably
not
the
best
ploy,
because
we're
just
talking
about
doing
some
analysis
coming
back
and
discussing
it,
whereas
we
don't.
I
personally
don't
want
to
be
quite
lumbered
with
one
of
these
issues
after
the
meeting.
A
E
F
E
E
E
E
E
A
For
time
it
might
be
an
idea
that
we
sort
of
make
moves
in
about
say
five,
ten
more
minutes
unless
people
feel
they
need
a
bit
more.
A
B
C
C
C
C
A
Okay,
shall
we
continue.
A
Please
well,
I,
I
guess,
does
anybody
want
to
talk
first,
I
guess
there's
more
of
a
question.
A
C
Sure
could
someone
else
take
notes
for
any
discussion
that
that
arises
for
this
section.
A
Sure,
if
nobody
else
will
step
up
to
it,
I'll
put
myself
on
mute
and
take
what
I
can.
C
Sure
so
I
looked
at
15,
which
is
selector
definitions.
Basically,
an
attempt
to
start
a
processing
model
and
karsten
has
chimed
in
there
with
the
collections
idea
and
the
the
impact
of
of
defining
this
as
a
list
seems
to
be
that
that
nobody
is
doing
that
by
default.
All
of
the
discussion
is
talking
about
which
set
of
objects.
C
Particular
things
are
returning
and
not
what
order
they
are
and
in
javascript
objects,
kind,
only
kind
of
have
an
order.
I
looked
up
the
json
rfc
and
it
says
sometimes
they
they
make
the
ordering
of
object.
Members
visible,
so
we're
going
to
have
if,
if
we
keep
an
order
here,
we're
going
to
have
to
say
that
it's
we're
going
to
have
to
say
what
order
things
things
show
up
in
which
could
be
could
be
ambiguous
for
objects.
C
Otherwise
everything
seemed
consistent
and
actually
to
be
assuming
definitely
like
that
that
duplicates
well.
It
doesn't
mention
whether
duplicates
happen,
but
it
doesn't
describe
deduplicating.
C
D
Yeah,
so
the
the
whatever
8259
says,
the
practical
matter
is
that
that
jsonpath
evaluator
will
sometimes
have
to
work
on
json
data
and
not
on
a
json
text,
and
the
json
data
may
not
be
presented
in
a
way
that
that
provides
any
ordering
to
the
map
entry.
So,
for
instance,
javascript
tends
to
to
put
numeric
map
keys
first
and
and
then
the
the
other
map
keys
and
so
on.
So
there
is
some
reordering
going
on
in
implementations
in
python.
D
That
reordering
is
actually
different
between
each
invocation,
which
can
be
a
lot
of
fun
with
it
when
you're
trying
to
debug
things
so
yeah
json
objects
do
not
have
order
and
any
selector
that
actually
relies
on
an
order
in
a
map
is
early.
Fine,
of
course,
arrays
do
have
order,
and
I
think
that
that
generally,
if
you
filter
an
array,
the
result
of
that
data
should
have
the
same
order
that
the
original
array
had.
B
Yeah
I
was
going
to
make
a
comment
about
the
ordering
of
objects.
I
think,
if
I
remember
correctly,
the
only
selector
component
selected
that
actually
would
that
would
impact
is
something
like
an
asterisk.
You
know
square
brackets
asterisk
to
pick
out
all
the
values
of
an
object,
and
in
that
case
yes,
the
output
would
be
the
order
will
be
non-deterministic,
but
I
think,
aside
from
that,
I
think
ordering
would
work.
B
B
But
suppose
someone
had
introduced
square
bracket
star
twice
over
somehow,
not
sure
exactly
how
that
would
work,
then
I
think
the
removal
of
the
duplicates
there
afterwards
still
work
we'd
end
up
with
the
original
non-deterministic
order.
Potentially,
so
I
think
think
it
could
be
made
to
work
together.
Don't
think
we
need
to
throw
out
ordering
just
because
of
the
behavior
of.
D
So
if
I
think
that
that's
a
question
that
that
really
should
be
an
issue
on
its
own,
do
we
expect
json
path
to
be
deterministic,
and
so,
if
you
apply
the
same
creator,
the
same
json
data
item.
Do
you
get
the
same
result?
A
A
Sorry
is
there
a
better
summary
of
what
you've
just
said
and
I'll
make
this
into
a
new
issue
right
now,.
B
Let
me
just
express
an
opinion
while
we're
here.
I
don't
think
it
should
be
deterministic
in
this
case,
because
I
think
that
would
be
too
constraining
on
the
implementations
and
I
don't
think
it
actually
matters,
because
I
think
applying
star
to
a
an
object
is,
is
probably
a
pretty
rare
use
case.
I
don't
think
it
really
matters,
and
if
you
exclude
that
case,
then
it
becomes
deterministic
everywhere
else.
B
I
think
you
know,
I
think
recursive
descent
is
a
possible
area
of
where
we
need
to
tighten
up
the
order,
but
I
think
you
could
you
can
have
a
well-defined
order
there.
I
think
everywhere
else,
sequences
and
so
on
are
well
defined.
B
A
Okay,
carsten.
D
Yeah,
I'm
in
the
middle
of
typing
that
issue,
but
anyway,
so
the
the
processing
model
discussion,
our
algorithm
discussion
under
number
14
was
actually
pretty
far
along
when
it
kind
of
derailed
into
it's.
It's
funny
to
to
see
people
fixing
the
mark
down
there
when
it
derailed
into
a
discussion
about
what
we
can
do
about
recursive
or
nested
filters,
and
so
on.
I
think
that
that's
a
completely
different
issue,
so
it
doesn't
really
have
a
bearing
on
on
this.
D
D
D
The
result
is
an
of
the
applying
the
selector
as
a
node
list
that
is
appended
to
the
result
list,
and
I
think
that
that's
where
the
first
little
question
marks
are
coming
up,
what
exactly
a
pending
means
here
and
then
the
the
proposal,
exactly
that's
the
one
I'm
reciting
here
the
the
proposal
says
the
the
results
are
put
through
an
operator
that
is
called
this
thing
that
I
have
no
idea
what
it
does,
but
some
probably
something
about
duplicates
and
that
then
becomes
the
current
node
list
again
and
when
the
the
path
selector
is
that
excuse
me,
the
the
path
expression
is
completed.
D
We
just
return
the
current
note
list.
That
makes
a
lot
of
sense
to
me,
except
again
that
I
don't
know
what
append
and
distinct
are
meaning
exactly
so
there
are
several
options
not
taken
here.
So,
for
instance,
the
the
selector
does
not
get
the
whole
node
list.
The
selector
gets
each
entry
of
the
node
list
each
element
of
the
node
list
in
turn,
so
you
cannot
write
selectors.
That
order,
note
lists
or
do
something
interesting
like
build
an
average.
You
only
ever
get
selectors
that
work
on
on
a
single.
D
Note
and
yeah,
I
said
it's
not
clear
what
this
thing
means
and
I
think
we
just
already
discussed
about
selectors
that
that
take
objects
exclusively
take
items
out
of
an
unordered
map
or
json
object.
I
think
we
covered
that
already,
so
I
I
would
love
to
hear
what
people
think
about
this
append
and
distinct
operations.
B
For
me,
the
obvious
meaning
for
append
is
really
concatenate.
I
would
think
the
selector
brackets
item
would
return
a
node
list
and
append
would
concatenate
the
two,
but
I
admit
it's,
it's
ambiguous,
isn't
it
it
could
be.
You
know
something
else
distinct,
I
would
hope,
runs
through
the
node
list
in
linear
order
and
when
it
encounters
a
duplicate,
removes
the
later
occurrence
of
the
duplicate
and
then
keeps
going.
So
essentially,
if
you
get
the
same
node
multiple
times
over,
only
the
first
occurrence
survives.
B
Say
again,
what
is
it
okay,
yeah
appointed
to
this
reference
to
the
same
node.
A
Okay,
I
take
it.
Please
are
we
going
to
be
adding
some
comments
onto
these.
F
A
That
was
a
question
is:
are
we
going
to
put
some
comments
onto
these
issues
clarifying
what
we've
just
discussed.
D
So,
unless
there
is
more
discussion
that
calls
in
the
question
whether
these
decisions
are
right,
we
can
make
a
pull
request
out
of
that,
and
the
pull
request
should
should
be
specific
and
detailed
and
unambiguous
enough
that
we
really
can
discuss
it,
in
particular,
against
the
the
various
implementation
survey
pieces.
We
have
that
we
can
run
against
this.
B
Okay,
I
mean-
maybe
it's
appropriate
at
this
point
to
then
discuss
the
way
the
working
group
operates,
because
currently
it's
very
much
in
mailing
list
and
issue
mode,
and
I
don't
see
the
editors
rushing
to
produce
pull
requests
at
the
moment
by
editors,
I'm
referring
to
stefan
and
marco,
who
were
the
editors
that
we
were
kind
of
expecting
to
to
do
most
of
the
writing,
and
I
I
too
favor
poor
requests,
but
it's
you
know
I
could
plow
into
this,
but
I'm
trying
to
hold
off
and
focus
on
the
reference,
implementation
and
the
compliance
test
suite
so
carsten,
were
you
thinking
you?
A
Okay,
that
sounds
like
we've
made
some
good
progress
there.
That
pull
request
can
okay
sounds
like
we
can
put
a
summary
on
to
some
of
these
issues,
then
I'll
do
that
afterwards
after
I
speak
to
tim
and
make
sure
that
he's
he's
abreast
as
to
what
we're
doing
it
sounds
like
we've
made
a
fair
work
of
progress
today.
A
B
So
number
27
actually
was
a
one.
It's
just
a
single
comment
issue
and
it
referred
mainly
back
to
number
23
duplicates
in
selector
output,
and
there
was
a
bit
of
discussion
back
and
forth.
Wasn't
really
a
consensus
there,
but
the
the
last
proposal
seemed
to
be.
It
wasn't
immediately
counted,
so
it
was
tending
towards
removing
duplicate
nodes,
but
not
duplicate
values,
which
is
consistent.
I
think,
with
the
node
list
idea
we've
been
discussing
today.
B
It
does
make
me
wonder
about
certain
languages.
I
just
have
the
back
of
my
mind.
You
know
things
like
I
don't
know.
Haskell
or
some
other
languages
that
pass
by
value
might
be
difficult
to
distinguish
a
node
from
a
value
reference
from
a
value,
a
duplicate
node
rather
from
a
duplicate
value.
A
I
wouldn't
be
worried
so
much
about
the
languages
unless
there
is
already
implementations
in
those
languages.
Yeah.
B
There
isn't,
as
far
as
I
know,
so
that's
that's
fine,
so
yeah.
There
was
interesting
discussion
in
number
23
about
whether
removal
of
duplicates
would
affect
ordering
and
consequently
make
the
spec
non-deterministic,
but
I
don't
think
it
would.
I
think
if
we
defined
it
carefully
and
said
that
later
nodes
were
discarded
in
preference
to
earlier
nodes.
Sorry
in
favor
of
earlier
knows,
then
you'd
get
deterministic
ordering.
B
I
think
that
could
even
be
done
during
the
processing
of
the
selector
stages.
So
as
soon
as
we
end
a
selector
stage
and
we
going
back
to
the
previous
piece
of
pseudo
code
as
soon
as
soon
as
we
removed
the
duplicates,
then
we
could
continue.
B
A
A
Okay
sounds
like
we've
made
some
impact
on
that,
just
just
to
recap.
For
my
own
own
sake,
it
sounds
like
in
terms
of
sort
of
wrapping
up
this.
A
It
sounds
like
that
we
need
to
well
I
need
to
as
chair
after
I
talk
to
tim,
discuss
to
leave
a
summary
over
what
over
sort
of
what
we've
concluded
with
in
the
meeting,
but
that
on
these
on
these
various
issues
I
would
say
we
should
probably
leave
the
issues
open
for
the
time
being,
because
if
there's
any
further
discussion
about
it,
we
can
keep
it
going
and
then
cast
them
when
you
have
the
time.
A
I'd
be
appreciative
for
your
pull
request,
and
I
know
you
don't
you
don't
have
time
in
the
next
couple
of
weeks,
but
at
some
point
that
would
be
kind
of
key.
Is
there
anything
I've
missed
there.
A
A
Sure
we
can
go
through
that
is
there
anything
in
particular
you'd
like
to
go
through
next.
D
Yeah,
I'm
not,
I
didn't
really
prepare
for
this.
So
I'm
sorry,
I
think
the
other
issue
that
got
a
lot
of
discussion
is
the
the
whole
expression
language
issue.
D
Well,
one
other
thing
we
could
do
is
go
through
issue.
53.
E
G
A
A
B
I
mean
it
seemed
to
me
that
the
next
step
here
was
for
someone
to
simply
make
a
choice
in
each
line
of
the
table
and
then
put
in
a
pull
request,
and
I
was
kind
of
holding
my
breath
opening
up
hoping
that
stefan
would
do
that,
but
he
hasn't
decided
to
so
far.
So
it
just
seemed
like
a
coin.
Tossing
exercise
really
to
pick
on
some
terms.
D
D
D
A
Don't
sound
like
it,
but
if
there
is
anybody's,
please
speak
up
now.
The
other
option
that
we
can
do
is
reiterate
this
on
the
mailing.
A
D
B
Just
checking
you
mentioned
minus
characters
in
names.
I
didn't
see
that
in
the
issue
53
I
was
looking
at.
A
There's
also
the
script
expressions
point,
but
I
don't
want
to
open
a
pan
of
worms
on
that.
One.
A
D
A
D
A
Yeah
there's
this
is
big.
Is
it
worth
us
offline
splitting
this
up.
G
D
D
So
node
is
really
talking
about
the
specific
position
in
in
a
tree,
and
item
is
just
talking
about
what's
at
this
position,
but
in
in
the
end
they
have
the
same
meaning
within
this
document.
B
Just
in
passing,
I
mentioned
a
mathematical
alternative
for
the
name,
value
pair
or
key
value
pair
in
an
object
will
be
maplet
would
be
what
we
want.
Maplet
m-a-p-l-e-t,
oh
you
know.
If
we
wanted
to
something
that
was
completely
neutral
from
a
computing
point
of
view
and
a
you
know,
unique
term,
it's
got
a
well.
You
know
mathematical
history
behind
that
term.
If
we
wanted
to
use
it.
A
D
A
D
A
D
So
we
move
on
to
the
next
point
yeah,
so
the
differentiation
thing
we
have
a
sample
plus
one
and
of
course,
I'm
happy
with
the
plus.
D
D
E
A
I
mean
if
it
could,
maybe
maybe
it
could
be
editorialized
and
and
perhaps
summarized
a
little
bit
further.
Instead.
B
We
could
yes,
yeah
that'd,
be
good,
I
think
I
think
bringing
it
in
earlier
in
the
document
risks
kind
of
leaning
on
xpath
for
semantics
and
being
a
bit
unclear
about
the
semantics
in
our
document,
which
I
don't
like.
So
I
think
it
appendix
would
be
a
good
idea.
Good
point.
G
D
Yeah
this
seems
to
focus
on
the
term,
but
I
think
the
interesting
question
is
about
duplicate
removal.
I
think
that
that
stefan
has
has
a
view
that
that
is
not
so
much
emphasizing
duplicate
removal,
so
that
would
be
different
from
what
we
just
said
in
in
the
processing
model
discussion.
F
B
Yeah,
I
guess
if
we,
if
we
did
go
the
way
that
we
discussed,
then
there
could
still
be
duplicate
values
in
the
output.
Yes,
so
I
I
think
we
could
get
away
with
the
term
union
there
on
the
understanding
that
it
was
a
union
of
node
references.
B
You
know,
but
it's
it's
might
be
a
bit
misleading.
When
you
look
at
the.
If
you
look
at
something
like
the
compliance
test
suite,
you
would
see
union
producing,
apparently,
duplicated
output,
duplication
values.
D
Well,
the
other
problem
is
that
it's
really
a
set
operation,
so
it
doesn't
result
in
an
order,
but
we
we
just
discussed
that
we
do
want
to
create
a
particular
order,
yeah
very
good
yeah.
E
D
D
I
think
we
can
just
push
this
to
the
processing
model
issue.
G
D
A
D
A
B
There
may
be
an
obscure
implementation
that
happens
to
be
dependent
upon
by
you
know
thousands
of
applications.
So
it's
hard
to
tell.
A
Okay,
is
it
worse
than
is
this
to
be
split
off
into
a
separate
issue,
so
that
we
can.
A
Okay,
respect
implementations:
this
shouldn't
be
fairly
controversial.
D
Yeah,
I
think
we
should
have
a
little
bit
of
introductory
text
that
that
explains,
I
mean
we
should
not
overdo
this,
but
we
should
really
explain
at
the
beginning
what
the
the
consideration
considerations
were
that
went
into
this
process,
and
so
I
think
this
this
is
good
material,
but
it's
mostly
just
editorial
writing
this
up.
I
don't
think
there
is
a
big
disagreement
on
the
principle
that,
of
course,
there
always
will
be
disagreements
on
what
the
principle
actually
means
in
a
specific
case,
so
so,
for
instance,
in
the
filter.
D
Expressions
point
my
view
would
be
that
if
an
implementation
does
this,
this
is
fine,
but
we
don't
have
to
respect
that
aspect
but
yeah.
So
it's
a
good
motherhood
and
apple
pie
point
to
make
in
the
introduction.
A
Okay,
that
sounds
like
an
issue
so
that
we
can
get
the
editorial
work
for
it:
reference,
yeah,
okay
and,
lastly,
we've
got
error,
handling.
D
A
D
No,
no,
you
can
still
close
an
issue
even
if
it's
already
happening
so
my
objective
would
be
to
actually
split
this
issue
54
up,
so
we
can
actually
close
issue
54
because
we
have
the
other
issues
that
completely
replace
it.
A
Yep
going
going
through
my
list
and
I
think
that
the
sorry
brain
fart
it
sounds
like
we've
got
a
few
a
few
issues
here:
we've
got
one
two:
we've
got
with
the
exception
of
union
and
duplicates
everything
else
will
be
turned
into
a
new
issue.
A
This
is
something
that
I
can
do
with
tim
in
terms
of
creating
the
new
issues
and
if
people
in
the
working
group
want
to
check
my
homework
to
make
sure
that
I
haven't
misrepresented
the
issues
or
or
said
anything
untoward.
That
would
be
very
much
appreciative
and
then
once
we've
got
those
issues
we
can
refer
back
to
54
and
then
close
them.
D
A
Yeah
and
and
boring
mechanical
things
are
what
I'm
best
at.
A
The
first
step
is
admitting
you
have
a
problem,
so
christina's
just
asked
a
question:
apologies
for
only
just
noticing
it
christine.
Would
you
like
to
ask
it
to
on
the
mic
or
do
you
want
me
to
relay.
A
It
so
christina
asks.
Is
there
any
way
that
we
could
quickly
resolve
not
to
change
the
most
basic
object,
dot
first
dot,
second
and
other
pathing
syntax,
so
that
implementers,
who
are
using
those
things
today,
can
feel
safer
that
they
are
not
going
to
be
get
hit
with
a
breaking.
D
D
So
dollar.a.b
certainly
would
be
one
such
example
where
we
should
be
very
clear
on
what
that
means,
but
there
are
little
issues
like
what
is
it
if,
if
one
of
the
items
in
in
that
sequence
actually
is
an
array
that
we
haven't
fully
defined
yet.
A
Okay,
creating,
I
think,
I
think,
that's
probably
the
best
idea
for
now.
B
Can
I
just
ask
a
question
of
christina:
is
she
happy
with
the
current
draft
behavior
for
the
for
the
pathing
syntax.
B
A
Okay,
cool,
you
know
it
sounds
like
we
made
some
progress.
We
still
have
about
30
minutes
left
on
the
clock
is,
do
we
want
to
keep
going
through
issues
or
have
we
got
enough
actions
to
to
move
ahead
with.
A
G
Yes,
hi.
I
just
wanted
to
suggest
that
you,
the
chairs,
organize
an
interim
meeting.
G
Sometimes,
between
this
itf
and
the
next
one,
and
it
would
be
good-
I
was
looking
at
some
of
the
people
who
are
active
in
the
in
the
github
issues
and
some
of
them
seem
to
be
in
a
different
time
zone.
So
it
might
be
good
to
have
like
a
doodle
poll
and
make
sure
that
editors-
and
at
least
these
people
can
be
present.
A
Yep
that
works
for
me.
I
would
thank
you
for
bringing
that
up.
I
was
going
to
do
it
eventually
would
just
just
a
straw
poll
of
people
would
would
something
say
the
approximately
the
first
week
or
two
of
may
be
appropriate,
given
that
the
next
atf
meeting
is
not
until
mid
july.
If
I
remember
correctly,.
A
I'll
be
honest
and
say
it
probably
favor
the
times
and
dates
for
a
lot
of
the
editors
that
are
editors
as
well
as
my
co-chair,
but
I'll
put
it
I'll
I'll,
create
it
as
an
action
to
take
it
to
the
mailing
list
and
come
up
with
some
times
and
dates,
and
we
can.
We
can
set
set
up
something
in
in
may
or
even
april.
D
Which
has
been
very
very
just
briefly
like
to
point
to
issue
55,
which
is
interesting,
because
it's
not
really
an
issue
that
that
we
can
act
on,
but
it's
a
statement
of
interest
from
from
standards
activity
that
that
is
making
use
of
this
or
plans
to
make
use
of
this.
D
D
But
I
think
it
would
be
good
to
to
read
this
issue
and
maybe
further
issues
we
we
get
from
from
other
people
who
who
want
to
use
this
work
and
think
about
what
what
our
response
should
be.
So
if
they
have
un
unfulfillable
expectations,
we
probably
should
say
that
or
if
there
are
activities
that
needs
need
to
be
split
off
and
and
are
not
currently
in
in
the
purview
of
this
group,
we
shall
try
to
to
detect
that.
A
A
Now
that
sounds
completely
reasonable.
Looks
like
there's
been
some
discussion
on
this.
Might
it
be
of
use,
I
don't
recall,
seeing
this
posted
to
the
mailing
list.
Would
it
be
an
idea
to
do
so.
D
A
B
A
Sure,
however,
I
think
the
the
position,
the
position
that
myself
and
tim
have
made
on
that
hasn't
really
changed.
But
if
it's
not
working
for
the
working
group,
if
people
want
to
chime
in
with
any
any
views
that
they
have
I'd
like
to
hear
them.
B
I
was
just
going
to
say
I
wonder
if
we
should
try
to
distill
discussion
in
the
mailing
list
into
issues,
because
I
feel
that
we're
kind
of
in
two
camps
at
the
moment
some
people
that
prefer
the
main
list.
Some
people
prefer
issues
and
there's
a
slight
disorganization
about
it.
D
D
So
it's
just
a
question
where
we
need
to
collect
information
and
then
make
a
decision.
These
should
always
be
converted
into
issues.
So
people
can
start
writing
pull
requests
because
pull
requests
really
should
be
closing
issues,
so
I
mean
we
may
occasionally
have
pull
requests
that
don't
have
an
issue,
but
the
normal
situation
would
be.
B
A
Okay,
so,
okay,
if
there's
no
more
issues
for
us
to
go
through,
I
think
we
can
probably
wrap
things
up
for
today.
A
Through
okay
I'll
I'll,
take
that
silence
as
we're
good
to
go.
I'd
like
to
thank
all
of
you
for
your
time
today,
including
newcast
and
glenn,
and
jeffrey
for
helping
me
out
here
being
without
a
co-chair
I'll,
be
putting
out
a
poll
to
the
mailing
list
later
today
or
in
the
next
coming
days.
I'll.
I
better
make
a
promise.
I
can
actually
keep
and
we'll
see
you
on
the
mailing
list
and,
if
not
we'll
see
you
at
the
next
atf
in
july,.