►
From YouTube: OpenActive W3C Community Call / 2020-09-09
Description
The Data Set Site Specification cont..
- Discussed in the W3C calls of 15th July and 26th August
- Resolved (?) issues
* Different feed types
* Data catalog specification
* Spoof-proofing identifiers
- Open issues
* Handling of base URI
* Publisher/creator/maintainer
* Feature assertion
* Status codes
A
Okay,
welcome
to
the
w3c
community
group
meeting
for
the
9th
of
september
2020.
topic
for
the
call
today
is
going
to
be
further
amendments
to
the
data
site
data
set
site
specifications
standard.
I
will
begin
sharing
my
screen.
I
guess
for
context.
This
is
the
third
call
in
a
row
we've
had
dedicated
to
the
data
set
site
specification.
A
A
Okay,
are
you
seeing
the
presentation
there
yep?
That's
very
good.
I
do
okay,
so
nick,
essentially
I
just
went
through
the
issues
list
attached
to
the
github
repository
and
started
pulling
out
stuff
that
looked
like
it
needed
more
work.
A
The
there
were
some
that
seemed
to
be
probably
unproblematic
in
that
we
could
probably
just
resolve
them
as
as
discussed
on
the
thread
so
for
differentiating
feed
types.
That
is,
I
believe
the
proposal
was
just
to
use
the
kind
value
as
the
value
of
the
of
the
type,
which
seems,
I
guess
it
seems
eminently
sensible
as
a
solution.
What
is
your
sense
of
the
practice
of
people
who
are
publishing.
B
Well,
I
mean
right
now
that
there's
a
there's
three
places
in
fact
that
this
information
is
stored.
So
there's
that
there's
the
type
sorry
there's
the
identifier
there's
a
kind
and
then
there's
also
this
alternative
type.
I
think
it's
called
I'm
just
going
to
try
and
bring
it
up.
So
actually,
if
you
could
bring
up
the
issue,
that
would
be
helpful.
B
Yeah
yeah
right
so
is
there
more
on
this
issue.
Is
that
that's
it
that's
about
it
right,
yeah,
so
yeah.
So
the
challenge
is
that
if
you
click
on
that
meta.json
link,
actually
that's
in
there
yeah.
So
at
the
moment,
there's
this
in
the
feeds
themselves
they're
using
kind
of
a
combination
of
these
things.
B
Interestingly,
in
the
harvester
that
I
don't
know
how
they
are
the
harvester
that
for
the
states,
page
works,
but
the
harvester
that's
built
into
this.
The
test
suite
didn't
actually
need
to
do
anything
clever
other
than
know
whether
that
it's
a
parent
or
child
type,
because
everything
else
is
actually
embedded
in
the
in
the
data.
So
what
I
thought
we'd
need
to
do
is
have
a
kind
of
a
switch
on.
B
What's
there
the
same
as
field,
but
I'm
not
sure
we
actually
need
to
do
that
now,
but
but
then
I
mean
it's
probably
still
useful
anyway,
for
you
know
having
some
consistency
to
know
what
types
are
in
there.
So
I
think
it's
a
nice
politeness
for
the
data
consumer
yeah
yeah.
B
So
I
think
the
challenge
we've
got
is
really
this,
as
you
can
see
in
green
on
this
chart
here
course,
instance,
sub
event
and
headline
event,
sub
event
and
slot
an
individual
slot
kind
of
demonstrate
that
we've
got
because
of
the
way
the
model
is
constructed.
B
We've
got
types
that
are
being
used
in
different
contexts,
which
is
slightly
annoying
and
we
haven't
really
formally
defined
these
anywhere,
and
so,
when
you
can,
you
can
get
a
feed
of
events,
but
you
can
also
get
a
feed
in
theory,
of
course,
instant
sub
events
and
so
within
the
net
implementation.
B
Within
the
test
suite,
we
use
these
strings
which
are
in
this
diagram
to
differentiate
between
those,
because
obviously
they
need
to
be
treated
differently,
because
they
are
kind
of
fundamentally
different
things,
even
though
they're
modeled
in
exactly
the
same
way,
apologies
for
the
siren.
So
I
think
it
would
be
good
to
formally
define
these.
I
think
if
we
could,
it
probably
means
that
we
need
to
to
define
them
in
the
open,
active
name
space.
B
So
that
would
be,
I
guess
so
connective
slash,
but
but
then
the
tricky
thing
there
is
that
if
we
define
them
in
the
objective
name
space
that
that
some
of
these
overlap
with
the
types
themselves
so
facility
use
is
obviously
the
same
as
the
facility
use
type.
B
So
whether
we
need
an
enumeration
which
is
like
type
of
feed
which
is
kind
of
where
you'd
maybe
have
facility
use
feed.
Maybe
we
just
prefix
everything
with
feed
or
another,
maybe
feeds
the
wrong
word
actually,
because
it's
not
even
feed,
because
this
is
this
is
the
type
of
the
object.
B
So
it's
it's
a
really
a
gnarly
thing
actually
from
the
modeling
perspective,
because
you've
kind
of
got
this
situation
where,
like
that,
you've
got
this
perfect
model,
but
it's
not
quite
enough
information
to
for
things
to
stand
alone
when
you're
when
you're
describing
them.
You
know,
if
you've
just
got,
you
just
got
a
slot
on
its
own.
You
don't
know
whether
that
slot
is
an
individual
slot
or
a
slot.
B
Unless
you
look
at
the
facility
used
to
determine
if
it's
an
individual
facility
use
or
a
facility
use-
and
the
same
goes
for
every
one
of
these
there's
basically
and
there's
a
nice
little
function
in
the
tests,
we
have
in
the
dotnet
sdk,
which
takes
the
event
and
sub
event
and
figures
out
which
of
these.
It
is
on
the
basis
of
what
are
what
are
you
and
what's
your
parent
and
therefore
what
what
is
the
the
relevant
string
that
identifies
you
and
converts
forwards
and
backwards
between
those
two
things?
B
Okay,
so
so
you
said
that's
in
the
test
suite
yeah,
so
I
can
maybe
share
share
the.
B
Because
we've
got
testing
oh
yeah
right
right,
right,
yeah
sure
we
can
well,
we
can
test
for
that.
Tesco
doesn't
test
for
it.
It
infers
it
currently
right
on
each
object,
but
but
yeah
I
mean
we
could
we
could
we
could
test
for
it
to
confirm
it
works,
but
the
I
mean
the
yeah,
so
the
yeah.
The
issue
is
more,
this
kind
of
modeling
weirdness
between
the
two,
where
you've
got
yeah
where
that
logic
exists,
so
the
two,
the
other
one
is
the
test
interface.
So
actually
you
can.
B
I
can
show
you
this
one.
If
you
go
to
the
test
interface,
that
might
be
useful,
so
open
active.
I
o
slash
test
hyphen
interface.
B
So-
and
this
is
a
really
good
example
where,
if
you
scroll
down
to
the
example
second
example,
there
you
go,
that's
that's
just
got
some
json
in
it.
This
is
it
yeah,
so
you
can
see
here
that
the
way
the
test
interface
is
built
in
order
to
specify
the
type
you
actually
need
to
do.
You
have
to
specify
two
things
to
specify
the
type
and
the
super
event
type
and
the
organizer.
In
fact,
in
the
super
event,
so
that's
the
structure-
that's
currently
laid
out
there.
B
You
don't
want
to
have
to
give
it
like
this,
and
that,
like
this
and
the
parent,
therefore,
that
and
that's
why
there's
this
logic
at
the
moment
in
all
these
places,
which
switch,
which
converts
between
those
two
things
to
detect
this
type,
almost
so
so
yeah.
I
guess
I
guess
my
suggestion
would
be
that
we
would
probably
need
to
create
a
enumeration
which
has
which,
where
these
don't
clash.
A
B
We
can
validate
against
it
and
clarify
what
yeah
that
this
is
a
an
event
series
something.
So
I
don't.
I
haven't
really
looked
into
schema
to
see
where
this
is
done
elsewhere,
because
I
imagine
there's
going
to
be
situations
in
schema
where,
for
example,
something
can
be
a
hotel,
but
it
can
also
be
a
type
of
hotel
stuff
like
that.
So
does
bound
to
be
someone
someone's
probably
done.
Has
this
situation
somewhere.
A
Yeah,
I
think
I
think
the
enum
suggestion
is
good
and
I
think
obviously
looking
at
schema.org
for
patterns
is
good.
I
guess
my
my
concern
is:
how
do
we
know
that
publishers
are
going
to
get
this
right.
A
Given
that
the
logic
is
actually
a
little
bit
complex-
and
I
can
imagine
if
we
don't
communicate
that
extremely
clearly,
it's
going
to
get
incorrect
assertions
are
going
to
be
made
yeah.
Unless
I
mean,
unless
we
do
something
whizzy
where,
by
creating
the
data
set
site,
there's
some
sort
of
dynamic
querying
of
that.
But
that
just
seems
to
be
making
crazy.
B
Well,
this
is
this
is
the
real
challenge
right,
so
I
mean
it's
just
to
what
extent
do
we
need
to
describe
what
these
feeds
are
in
the
dataset
site,
and
so
it's
obviously
this
is
what
I'm
saying
with
the
implementation
in
the
test
suite,
at
least
it
doesn't.
All
you
need
to
describe
is
parent
or
child.
You
can
actually
do
that
with
a
boolean.
B
Is
this
a
parent
or
a
child?
If
it's
a
if
it's
a
parent,
I'm
expecting
to
find
children
associated
with
it?
If
it's
a
child,
I'm
expecting
to
find
parents?
If
it's,
neither,
then
that
that
doesn't
matter,
which
is
the
really
the
only
relationship?
B
I
think
you
need
to
process
them
because
everything
lives
in
some
event
or
in
facility
use
in
terms
of
the
the
parent-child
relationship,
so
you
can
just
look
for
the
id
that
corresponds
and
everything
else
is
identical
in
terms
of
the
data
structure,
so
I
mean
the
other
route
here
we
go
down
here.
Is
we
just
don't
we?
B
We
keep
this
enum
as
a
kind
of
internally
num
for
tooling
and
for
sdks
and
things
like
that,
where
that's
declared
separately
and
then
we
don't
formalize
it,
and
then
the
data
set
site,
we
just
say
parent
child
rather
than
trying
to.
A
A
To
make
more
sense
to
me,
because
I
feel
like
the
the
labor
of
getting
that
right
is
going
to
fall
on
either
the
publisher
or
the
consumer.
B
Yeah,
it's
just
another
thing
to
validate,
I
mean
yeah,
I
think
that's
the
thing
it's
it's.
I
guess
the
one
of
the
things
is
that
you
won't
be
able
to
just
look
at
a
feed
and
look
at
a
data
site
and
know
programmatically
what
type
of
stuff
it
has
which
is
kind
of
annoying
facilities
or
sessions.
Things
like
that.
B
So
I
mean
for
that,
there's
actually
an
alternative,
which
is
that
we
have-
and
this
is
something
that's
already
in
the
in
the
metadata
as
well,
but
we
could
have
an
enumeration
which
is
actually
a
slightly
more
abstract
enumeration,
which
is
kind
of
facilities,
sessions,
courses,
events,
you
know,
and
so
we
can
meet
the
needs
of
the
consumer
trying
to
determine
that's
so
so,
okay,
so
a
way
of
a
way
of
saying
that
would
be
that
you've
got
you've
you
you
could
say
this
is
a
type
of
sessions
feed
and
it's
it's
a
parent
sessions
feed
and
it's
a
child
sessions.
B
Feed
and
you'd
know
what
that
meant,
and
so
we
basically.
A
Why
not
do
it
as
tightly
as
possible,
like
is
splitting
the
difference
really
that
satisfactory
solution?
What's
that
saying?
Well,
well
I
mean
so:
we've
we've
got
sort
of
a
choice
between
not
specifying
and
having
walls
or
specifying
at
the
level
of
parent
child
specifying
at
the
level
of
quite
specific.
You
know
every
single
individual
type
that's
listed
in
that
diagram
and
then
there's
a
kind
of
in
between
the
difference
of
specifying
the
more
abstract
category,
but
not
all
of
the
subclasses.
B
Yeah,
I
mean
the
thing
the
thing
is
without
specifying
well,
I
think
we
probably
do
need
to
have
a
level
of
specific
specificity
in
there
because
we
need
to
be
able
to,
like,
I
said,
programmatically
know
what
types
are
in
a
in
a
feed.
That
seems
like
a
reasonable
thing
to
to
ask
right
like
how
many
data
set
sites
support
facilities
like
we
probably
need
to
support
that
query.
B
A
I
guess
if
it's
well,
if
we're
assuming
that
validation,
ensures
that
publishers
are
doing
this
correctly
and
we
don't
actually
have
a
validator
right
now.
Yeah.
A
I
guess
jason,
that's
an
interesting
question.
What
kind
of
validation
do
we
need
for
the
data
set
site?
I
feel
like
jason
schema,
would
probably
do
the
trick
in
this
case.
B
Well,
so
the
the
validation
we
have
already
does
to
an
extent
do
this
because
it
has
to
for
the
test
suite
right.
So
if
you
go
into
the
validator
right
now,
you
can
look
at
a
data
set
site
in
there
and
it
will
it
will
validate
it
and
all
the
validation
for
that.
The
validator
just
extracts
the
json
ld
from
the
page
and
then
validates
json
ld.
It
obviously
doesn't
validate
any
of
the
other
html
markup.
A
B
Right,
but
we
have
that
problem
anyway,
right
so,
and
the
thing
that,
with
the
test
suite
potentially,
is
that
we
can
extend
it.
I
mean
it's
not
in
the
current
scope
of
work
for
the
stuff
that's
going
on
now,
but
it
could
be
extended
to
to
check,
for
example,
feeds
against
each
other.
In
fact,
sorry,
that's
something
that
it
already
does.
It
just
doesn't
have
a
test
for
it,
so,
for
example,
to
detect
orphans,
that's
a
thing
that
it
does
right
now.
B
So
if
you've
got
parent-child
relationship,
obviously
there's
a
situation
where
the
ideas
don't
match
so
you're
kind
of
having
to
get
more
than
just
surface
level.
Validation
you're
actually
checking
the
integrity
of
those
two
things.
So
that's
what
you'd
need
to
do
for
the
data
site
as
well
is
just
check
that
integrity
is
the
feed
I'm
accessing.
B
This
is
actually
the
same
for
urls
as
well.
As
that
you
know,
a
lot
of
the
urls
in
the
feed
probably
need
to
be
checked
for
integrity.
So
is
the
url
that
something's
referencing
gonna
return
a
404,
or
is
it
an
actual
real
url.
A
A
I
think
I'm
I'm
sort
of
tempted
to
say:
should
we
put
it
in
right
now,
just
as
a
parent-child
level,
on
the
grounds
that
this
is
something
that
we've
more
or
less
got
tooling,
that
we
can
test
against
and
put
it
on
some
kind
of
road
map
for
for
tightening
that
validation
in
the
future
and
tightening
that
kind
of
specificity
in
the
future.
B
B
And
also
bear
in
mind
that
we
currently
do
do
this
validation
anyway,
as
part
of
every
new
data
set
site
that
gets
published,
because
they
already
include
the
identifier
which
has
one
of
these
strings
in
it.
It's
just
it's
not
formalized,
so
I
I
don't
know.
If
we
I
mean,
maybe
I
guess
I'm
suggesting
we
don't
worry
about
validation,
so
much
as
a
concern
in
terms
of
the
solution
we
choose,
because
I
mean
yeah
data
set
sites
already
have
well.
The
whole
thing
already
has
a
level
of
manual
validation
anyway,
but
okay.
A
When
you
say,
validation
mean
manual
checking
right
well,
so
I
mean
this
is
my.
This
is
my
concern
is
that
my
concern
is
entirely
that
publishers
enter
the
wrong
value
into
their
declaration,
and
then
you
know
we're
actually
actively
misleading
data
consumers.
Basically.
B
Yeah,
but
I
guess
I
guess
my
yeah,
what
I'm
thinking
there
is
that
that's
that's
the
issue
we
have
with
the
whole
thing
right
like
if
they,
if
they
put
something
wrong
in
their
feed,
and
we
haven't,
checked
it
overall
until
we
have
a
validation
that
covers
it,
then
that
there's
misleading
happening,
whether
it's
the
price
in
the
wrong
place
or
there's
all
sorts
of
issues
with
some
of
these
fees
that
we've
kind
of
had
to
manually
come
and
kind
of
fix.
So
I
guess
this
is
it's
more
like
is
this?
B
You
know
it's
more
about
supporting
features,
but
maybe
this
is
I
mean.
Maybe
it's
a
question
for,
because
obviously
we're
just
this
is
the
scope
of
this
call.
We
probably
can't
it
sounds
like
it's
about
requirements.
Maybe
we
should
talk
to
the
data
consumers
and
find
out.
You
know.
I
know
we've
got
people
on
the
call
who
usually
attend
are
those
people,
so
maybe
maybe
the
thing
is
to
decide
whether
it's
a
requirement,
in
fact
to
to
figure
out
which
feeds
which,
from
a
kind
of
wanting
to
know
that
information
when
you're
listening.
B
Just
figure
out
what's
needed:
okay,
yeah,
probably
probably
I
mean
yeah.
I
suggest
that's
weird.
A
B
Yeah
I
mean-
and
we
can
also
take
an
action
over
to
look
at
schema
and
see
what
they've,
if
they've
sold
it
in
the
clever
way.
You
never
know
there
might
be
an
inspiration
in
there
that
yeah
they've
really
good
practice,
because
this
must
be
a
problem
like
there's
a
lot
in
that
model
that
can
get
used
in
different
contexts
yeah.
A
B
Really,
I
think
so
I
mean
this
one
is
actually
lifted
straight
from
schema,
there's
not
much.
We
were
doing
in
there
other
than
using
the
has
part,
but
that's
also
in
the
google
documentation
in
a
in
roughly
the
same
context.
So.
A
Okay,
so
I
think
there's
nothing
to
do
there,
except
really
to
close
the
issue
and
acknowledge
that
it
needs
to
be
documented,
so
pagination.
So
this
is
about
providing
the
total
number
of
items
in
a
feed
which
we
discussed
previously
right.
That's.
B
That
interesting,
because,
with
the
test
suite,
I
I
think
I
think
I
agree
with
you
actually
I'm
looking
at
that,
because
I
know
that
came
out
your
kind
of
harvesting
experience
somewhat
and
I
did
in
the
test
week.
We
did
the
same
thing.
It
was
a
massive
feed
of
some
provider.
I
can't
remember-
and
it
was
I
was
looking
at
this
thing
thinking
how
percentage
bar
would
be
useful
right
now.
B
B
B
Total
items
or
something
yeah,
we
probably
should,
I
think,
there's
something
stuff
in
schema
that
does
this
already
in
maybe
for
lists
or
something
and
there's
also
hydra.
Hydra
has
a
total
items,
type
thing
as
well,
so
maybe
between
hydra
and
schema,
we
can
be
inspired
with
that.
One
sure,
okay,.
A
B
Yeah,
so
I
mean
the
good
thing:
is
the
data
site
libraries
currently
are
designed
for
dynamic
rendering?
So
it's
it's
designed
for
the
idea
that
you
might,
for
example,
have
different
brands
if
you're,
if
you're
legend
and
gll
is
one
of
your
customers
and
you
might
want
to
produce
a
site
per
brand
and
that's
why
it
doesn't
do
it.
B
Doesn't
just
statically
produce
html
that
you
then
copy
and
paste
into
your
site,
it's
a
little
bit
more
there,
so
so
yeah
I
mean
that
would
be
really
easy
for
anyone
to
do,
and
also
the
caching
recommendations
around
it
as
well,
cache
it
for
15
minutes
or
whatever.
That
would
work.
So
I
yeah
and
obviously
we
I
think
we
talked
about
this
last
time.
We
didn't
need
that
number
to
be
a
hundred
percent
accurate
because
it
will
change
as
paging
continues.
A
Okay,
so
yeah,
I
can
see,
I
guess
the
coding
there
has
to
be
a
bit
more
defensive
than
it
would
be
for
some
of
the
other
variables
that
are
dealt
with
and
dynamic
page
generation,
but
nothing.
Nothing
too
involved.
B
Well,
I
guess
there's
a
good
question
about
whether
we're
going
to
make
that
required
or
optional
for
the
data
set
site
spec,
which
is
interesting.
I
mean
it's
gonna,
be
a
lot
of
work
for
people
to
update
to
well
not
a
lot,
but
there
might
be
situations
where
that's
more
work,
but
then
it
might
be.
It
depends
how
valuable
we
think
that
percentage
bar
is.
I
suppose.
A
A
Okay,
spoof
proofing
identifiers,
so
this
was
about
defining
a
url
pattern
for
identifiers
that
would
constrain
options.
Basically,.
A
B
A
Just
wanted
to
erase
it
here
to
make
sure
there
weren't
any
sort
of
nuances.
I
was
missing
there,
okay,
so
those
are
the
ones.
Those
are
those
that
classified
as
potentially
resolved
and
we've
we're
now
half
an
hour
through,
but
great.
B
A
B
Yeah,
what's
that
issue
I'm
trying
for
which
one?
This
is
now.
B
Well,
so,
okay,
so
right,
that's
because
the
cut
there's
currently
an
inconsistency
in
this
in
the
open
booking
spec
there's.
Currently
the-
and
this
is
this-
is
primary
primarily
for
caching
reasons,
depending
on
you've
how
you've
configured
your
cdn.
B
So
the
idea
here
is
that
the
orders
feed
may
be
at
a
different
url
to
the
rest
of
the
booking
endpoints,
so
the
booking
endpoints
will
all
have
the
same
base
url.
That
was
a
discussion
happening
in
schema.
B
It
seems
everyone's
aligned
on
that
and
I
think
originally,
the
version
0.3
of
the
of
the
booking
spec
had
like
a
very
hyper
media
vibe
to
it,
so
you
could
just
find
urls
and
use
them
from
this
yeah,
and
it
was
it
was
kind
of
every
response
would
give
you
a
url
that
you'd
go
to
for
the
next
response,
but
I
think
we
kind
of
realized
that
most
things
now
don't
really
work
like
that.
B
I
think,
apart
from,
if
you're
going
full
hydra,
it's
generally
the
case
that
you
can
get
a
swagger
spec
that
all
the
endpoints
are
where
they
are
and
that's
very
well
defined,
and
they
don't
move
around
depending
on
what
you
click
on
and
when
you,
when
you
build
your
your
implementation,
you're,
not
thinking,
I'm
just
going
to
grab
a
url
from
here
and,
let's
post
to
it
a
bunch
of
personal
data
right
because
that's
just
all
kinds
of
madness,
so
you
kind
of
want
to
know
consistently
where
you're
posting
stuff
and
that
doesn't
come
dynamically
out
of
the
responses,
so
so
that
that's.
B
Why
that
that
moved
away
from
hype
media
towards
a
base
url.
I
think
that
seems
like
it
fits
very
well
with
all
of
the
the
schema
stuff
that's
going
on.
However,
there's
this
one
remaining
question,
which
is
that
well
there's
a
a
feed
which
all
the
other
feeds
can
be
anywhere
right.
B
There's
no,
we
haven't
got
any
constraints
about
the
open
data
feeds
they
can
live
wherever
you
want,
and
that's
for
caching
reasons
again
because
you
might
want
to
put
them
somewhere,
where,
like
gladstone,
for
example,
has
a
special
cache
edge,
cache
cdn,
which
specifically
handles
rpd
and
exists
in
a
completely
different
domain
to
all
of
its
implementations,
all
of
the
local
installations
that
exist,
and
so
the
data
set
sites
on
the
local
installation.
B
But
then
the
actual
fees
themselves
are
in
a
completely
separate
setup
and
that
works
really
well
and
there's
no
need
to
constrain
that
stuff.
And
so
I
suppose,
in
a
similar
way
I
mean.
Let's
take
that
example.
Maybe
gladstone
would
want
to
use
that
same
edge
cache
for
the
orders,
feed
there's
some
nuances
around
authentication
there,
which
make
that
a
little
bit
more
tricky.
B
B
If
it's
defined
in
the
data
sets
back
at
the
dataset
site,
so
it's
almost
it's
it's
to
purposely
deferred
it
to
the
dataset
site
to
a
feature
that
doesn't
exist
right
because
we
haven't
put
it
in
yet
so
so
it's
a
question
of
whether
we
want
to
have
yeah
how
we
want
to
kind
of
handle
that
I
mean
do
we
want
to
actually
try
to
get
a
get
that
feature
in.
A
I
guess
what
this
sits
in
tension
with
is
the
discussion
about
endpoint
url,
with
the
idea
that
that
you
know
you've
got
endpoint
url,
which
just
points
at
the
the
initial
endpoint
for
the
interaction
and
then
you've
got
a
documentation
url
which
points
to
some
kind
of
swagger
or
endpoint
description
or
whatever
the
is,
and
so
I
suppose
in
that
scheme
of
things,
supposing
that
everybody
had
an
open
api,
swiper
documentation
available
that
orders,
feed
url,
would
presumably
properly
sit
in
there.
Exactly
that's
right.
C
A
B
Tally
with
your
understanding,
yeah,
that's
exactly
right!
That's
exactly
right,
and
if
we
put
that
in
that
property
would
then
override
whatever
swagger
would
say
about
where
that
that
endpoint
was
supposed
to
live.
It's
kind
of
annoying
from
swagger
point
of
view,
although
I
think
you
might
be
able
to
do
something
in
the
swagger
dock
to
add
a
new
base
url
or
something
which
you
can
write.
Yeah.
A
They've
got
a
actually,
I
actually
looked
at
it.
It's
quite
simple,
which
is
you
know
if
given
a
if
given
base
uri
but
a
path
declared
underneath
that
is
not
relative,
it's
interpreted
absolutely
so
that's
you
know.
B
So
they
could
just
customize
the
swagger
yeah,
so
yeah
I
mean
I
guess
it's
it's
funny,
isn't
it
it's
almost
there's
a
bit
of
me.
That's
like
well,
there's
no
point
constraining
us.
I
mean
we're
just
literally
doing
it
for
some
for
modeling
reasons.
Oh
we
can't
if
we
can't,
if
we
don't
define
a
property,
it's
just
well.
Why
don't?
We
just
keep
it
simple,
so
I
guess
why
yeah
so
maybe
there's
no
reason
not
to
yeah.
A
I
think
I
mean
I
guess
I
guess
the
risk
is.
If
we've
got
the
risk,
I
guess
is
a
divergence
between
your
other
api
documentation,
which
we've
kind
of
implicitly
recommended
elsewhere
and
an
explicit
declaration
in
the
dataset
site.
I
guess
that's
a
question
of
you're
sort
of
maintaining
two
different
declarations
about
where
the
order
speed
lives.
B
Yeah,
that
is
true.
Well,
I
suppose
the
other
way
of
looking
at
it
is
that,
like
this
is
this
is
a
completely
speculative
feature.
We
have
absolutely
no
idea
if
it's
going
to
be
something
that
people
are
going
to
want
or
not.
So
maybe
we,
you
know
we
wait
and
see
how
the
implementations
go
or
you
know
because
we
could
yeah.
B
I
mean
I'm
just
I'm
really
thinking
you
know
after
gladstone
and
legend
we'll
probably
have
more
of
an
idea,
but
because
the
trouble
with
this
one
is
that
it
is,
there
is
a
there
is
authentication
over
the
top
of
it.
So
it's
right
not
super
easy
to
edge
cache.
Unless
you
take
the
api
keys
into
account,
which
most
caches
don't
most
cdns,
don't
do
that
so
you're
probably
gonna
have
to
have
some
internal
caching
mechanism
anyway,
if
you're
gonna
do
that,
then
you
know
it's,
it
seems
yeah,
so
I
mean
it.
B
So
I
guess
the
other
option
is
that
we
make
a
beta
property
that
is
like
whatever
this
you
know
put
this
somewhere
as
a
bit
like
we've
done
with
the
barcode
stuff
in
the
booking
spec
itself,
we've
kind
of
acknowledged
that
we've
we've
put
we've
built
the
spec
with
a
known
limitation
and
said
we
don't
know
enough
about
barcodes
at
this
stage
to
know
how
this
is
going
to
be
used.
B
So
we're
going
to
purposely
introduce
to
be
just
beat
a
property
that
will
have
to
be
used
with
the
booking
spec
in
order
for
this
to
work,
but
we
want
to
learn
through
implementation
experience
and
actually
that's
a
very
minor
part
of
the
spec.
So
I
mean
yeah
one
way
we
could
do.
It
is
just
put
a
piece
of
property
in
for
now
and
then
in
the
future.
B
If
it
becomes
obvious
that
this
is
something
that
people
want
to
do,
we
can,
then
you
know
we
can
adapt
the
tooling
to
read
that
beta
property
and
yeah
and
and
test
for
it,
and
things
like
that.
A
Okay
sounds
good.
I
think
I
think
beta
property
makes
sense
there
yeah,
I
thought
cluttering
is
too
much.
A
A
Yeah
and
I
think
the
sort
of
conclusion
you
arrived
at
and
I
believe
I
plus
one
bit
there,
which
was
just
to
keep
booking
surface
basically
yeah,.
B
A
It's
quite
a
bespoke
kind
of
requirement.
Isn't
it.
B
Yeah-
and
it's
really-
I
mean
it's
going
to
just
complicate
it
for
everyone
that
we
in
the
open,
active
ecosystem
that
wants
to
use
this
property
for
something
definitive.
It's
going
to
just
make
it
a
lot
more
messy,
so
yeah.
A
Okay,
let's
maybe
just
retain
the
original
or
well
the
current
implementation.
I
suppose
current
approach,
yeah,
okay,
return.
A
Okay
feature
assertion:
okay,
this
one's
a
bit
more
meaty.
I
think
so
this
is
about
indicating
oh
hold
on
bad
link
there.
This
is
about
indicating.
A
Oh,
that's
annoying,
which
features
a
particular
data
set
of
or
feed
assemblage
I
suppose
implements
now.
How
did
you
is
it?
The
conformance
certificates
just
lived
in?
Well,
I
think
conformance
certificates
is
related
right.
That's
that's
kind
of
one
approach
would
be
yeah,
so
the
first
approach
is
your
option.
One
is
profiles
defined
via
certification
and
then
the
second
one
is
more
just
a
kind
of
use
conforms
to
and
then
an
enum
basically
to
describe
what
is
available.
A
That,
given
the
conformance
certification,
isn't
a
reality
yet
and
it
won't
be
for
a
while.
We
have
an
enum,
but
we
also
have
an
option
of
url
pointing
to
further
description.
B
Well,
this
is
an
interesting
one
right
because
the
the
conforms
to
it.
Well,
I
mean
now:
it's
been
defined
in
schema
a
little
bit
more
closely
to
to
d
cap,
which
is,
I
guess,
where
they're
getting
inspiration
from
now.
It
does
look
like
yeah.
It
does
look
like
the
it's
almost
it's
it's
a
there's,
maybe
a
level
of
interaction
implied,
so
you
probably
want
to
say,
conforms
to
modeling,
open
book
inspect
one
and
then
some
other
property
which
the
confact
is
conforming
to.
That
would
mean
that
you've
got
that
property.
B
You
said
I
mean
story
knight.
A
B
So
so
the
the
feature
spec
is
actually
part
of
so
so
conforms
to
itself.
Right
doesn't
give
you
anything
about
enums,
it's
just
this
is
the
spec.
It
conforms
to.
A
B
Semantics
have
conforms
to
this
is
kind
of
overloading
it
with
some
stuff
that
it
really
doesn't
do
right.
Okay.
So
what
you're
really
saying
is
that
this
web
api
conforms
to
open
booking
api
there
and
the
open
booking
api
has
a
feature
enumeration
facility
within
it,
which
that
spec
defines
or
links
to
another
device
so
you're.
B
That's
probably
that
I
think
that's,
probably
the
more
semantically
correct
way
of
of
looking
at
it,
because
then
you're
well,
you're
you've
got
a
really
well-defined
feature
profile
thing
in
the
booking
spec,
which
is
what
this
is
saying,
we're
conforming
to,
because
yeah,
because
conforms
to
doesn't
do
any
of
that
like
what
does
it
mean,
for
example,
to
have
these
hashes
after
that
like?
Does
it
mean
full
conformance
with
each
one?
Does
it
mean,
like
I
don't
know
like
there's
only
we'd
have
to
make
it
up.
A
Yeah
or
add
some
documentation,
yeah
we'd
have
to
make
separate
urls
for.
B
Yeah
and
the
other
thing
is
that,
if
you're
looking
at,
if
you're
google
crawling,
this
conforms
to
is
going
to
be
recognized
and
obviously
they
are
expecting
like
one
or
two
standards,
so
they
can
link
to,
and
you
know
reference
so.
A
Okay,
I
think
I
feel
like
that
leaves
us
with
a
conundrum.
Unfortunately,
like
I,
I
take
your
point,
but
the
difficulty
is
I
mean:
when
are
we
getting
conformant
certificates?
Definitely
a
good
idea,
but
you
know
they're
quite
speculative.
B
At
the
moment,
well,
so
the
performance
certificates
are
already
generated
by
the
the
test
suite
that
that
that
that
works,
whether
we
want
to
allow,
I
mean
that
they're
self-certificates
self-certifications
at
the
moment,
so
they're
not
there's
not
a
paid
for
option
all
the
rest
of
it
there's
other
questions
about
whether
we
offer
those
in
the
future
or
what
that
looks
like,
but
but
yeah
I
mean
right
now
we
do
have
the
future
of
conformance
certifications
that
we
can.
We
can
use.
Well,
I
suppose,
maybe
the
the
question
before.
B
That
is
what
extent.
Well,
I
guess,
if
a
conformance
certification
is
an
assertion
of
specific
conformance
within
a
spec
right
as
a
concept,
then,
actually
you
know
that
assertion
could
be
anything
right.
It
could
just
be
jason.
It
doesn't
like
sorry.
What
I
mean
by
that
is
conformance
certificates
is
like
a
big
thing
that
could
be
lots
of
things,
but
for
what
we
need
it
to
be.
B
A
B
Right
but
it
could
literally
be
you
know.
First
party
I
mean
it
could
just
be
I've
confirmed.
I
assert
these
things
and
the
conformance
certification
I
mean
if
yeah
and
then
how
that
that
json
gets
generated,
is
then
a
question,
maybe
there's
a
some
kind
of
digital
signing
of
it
that
allows
it
to
be.
You
know
that
then,
then,
almost
the
conformance
certificate
becomes
the
digital
signing
of
the
assertion,
but
the
assertion
itself
is
the
like.
We
can
build
like
that.
A
Okay,
so
what
we
need,
what
we
need
is
a
property
name
for
that
yeah
the
enums
to
be
defined,
and
then
I
guess
some
thought
about
what
a
third-party
confirmation
of
that
would
look.
B
Like
well
so
this
might
be
a
good
segue
into
if
you
just
go.
If
you
jump
to
the
conformance
certification,
one
there's
some
jason
in
there,
that
does
do
a
little
bit
of
that
it
might
be.
I
think
you
have
to
go.
Oh
yeah,
perfect.
B
Yeah,
so
maybe
yeah,
if
you
just
scroll
down
to
the
json
in
that
blob
right.
B
B
Feature
assertion
or
something
yeah,
and
so
therefore
the
conformance
certificate
is
actually
just
certifying
feature
assertion,
if
you
like,
which
may
be
a
sub
property
of
that
so
and
then
it's
about
this
feature
implemented,
and
maybe
we
have,
I
mean
at
the
moment
I've
just
used
scores
because
we've
used
scores
everywhere
in
this
example,
but
you
can
imagine
that
we,
we
define
a
like
a
standard
control
vocabulary
like
we've
done
elsewhere,
that
can
be
evolved
and
and
contain
whatever
it
needs
to
contain,
and
that
would
I
mean
this
just
yeah
just
so
the
main
feature
implemented,
uses
scars
and
then
there's
also
like
an
opportunity
type
implemented,
which
is
which
which
actually
that's
interesting,
so
opportunity
type
implemented,
is
kind
of
what
we
were
talking
about
earlier
with
the
problem
of
feeds
right.
B
Actually,
if
you
scroll
down,
I
can't
remember
what
we've
done
there.
What
is
what
does
it
say?
Oh
yeah,
okay,
so
it's
just
using
strings,
it's
not
very
useful
but
yeah.
So
that's
that's
kind
of
referencing
the
earlier
conversation
about
that
whole
problem
and
like
serving
kind
of
a
similar
need
here
and
then
and
then
the
associated
media
in
this
example.
So
the
way
the
test
suite
generates
conformance
certificates
is
that
is
the
digital
signature
is
the
associated
media.
Okay.
B
So
that's
the
thing
that
basically
means
that
everything
else
is
valid.
So
I
yeah
you
could
almost
you
could
almost
move
associated
media
awarded
to
and
recognized
by
into
conformance
certificate
thing,
and
then
it
could
be
referenced
that
there
could
be
an
assertion
there,
that's
references
or
vice
versa,
something
like
that.
A
B
Yeah,
do
we
want
to
use
scores
for
the
enum
or
use
or
use
the
model
enums
that
we
have
in
other
places
like
gender?
We
could
use
that
and
this
control
vocabulary
versus
I
mean
yeah
yeah.
A
A
Very
annoying,
how
can
I
move
my
zoom
bar
there?
We
go
can't
get
to
tabs
because
of
zoombar.
That
was
everything
that
I
had
down
that
needed
resolution.
Did
we
cover
the
status
codes
point
at
the
end?
Oh,
I'm
sorry,
good
point
status,
codes
right,
sorry,
so
this
is
what
do
we
do
with
feeds
that
have
been
removed
or
unpublished?
A
I
kind
of
thought
so
I
kind
of
wondered
how
much
this
was
something
to
do
with
the
spec
per
se,
how
much
it
was
like
return,
a
410
code.
You
know
we're
telling
you
return
the
appropriate
http
status
code
if
you've
unpublished,
something,
and
then
we
just
issue
guidance
for
how
you
how
you
deal
with
that
from
a
gdpr
perspective,
etc.
B
Yeah,
I
guess
I
guess
probably
we
need
well.
I
imagine
if
the
data
set
site
specification
is
defining
the
data
site
endpoint,
it
probably
just
needs
a
little
status
code
section
in
there
right
that
says,
this
end
point
should
return
to
200
unless
x,
in
which
case
yeah
like
410
great
I
mean,
if
that's
that's,
the
appropriate
one
here
gone,
makes
makes
more
sense
than
found
not
found.
I
guess
yeah,
like
not
fans
currently
implemented
in
in
obviously
one
place,
but
that's
not
a
reason
to
use.
B
It
necessarily,
and
I
guess
a
good
that's
a
good
example
where
having
it
in
the
spec
is
going
to
be
really
helpful,
because
then
we
can
point
to
that
and
say
right.
We
need
to
move
that
to
a
410,
so
they
have
a
404.
Now,
because
that's
what
the
spec
expects
and
that's
what
data
consumers
will
react
to
yeah.
A
B
Yeah,
that's
true,
but
I
guess
there's
it's
interesting
because
with
http
status
codes
they
tend
to
mean
different
things
in
different
contexts
and
often
they're
overloaded,
so
it
might
be
worth
and
also
for
an
implementation
perspective
and
a
testing
perspective,
making
it
really
clear
what
the
expectation
of
those
semantics
are,
so
that
you
know
when
you've
conformed.
I
mean,
let's
say
that
we,
you
know
we
got
the
test.
B
Ideally,
you
have
a
test
interface
thing
that
says
turn
off
your
data
set
site
and
you
turn
it
off
and
then
you
test
to
see
if
it's
returning
a
410
and
then
you
turn
it
on
again
and
then
that's
your
that's
your
future
to
test,
which
I
guess
it's
hard
to
kind
of
code
and
make
explicit
in
there.
If
we
don't
have
to
associate
it.
A
B
Maybe
we'll
talk
about
different
things.
I
was
just
saying:
make
the
status
codes
explicit
in
the
spec:
oh
okay,
right
right
right,
rather
than
communicate
them
declaratively.
B
B
I
mean
exactly
in
this
case:
it's
just
the
thing.
The
thing
is
that
the
the
this
mind
of
the
404
might
persist
because
they've
literally
taken
the
software
out
of
their
system,
so
it
might
just
be.
I
mean
mandating
co
mandating
jason
to
describe
that
might
be,
or
almost
more
burdensome
than
just
saying.
You
know
just
you
configure
your
ios
server
to
return
the
status
code
for
this.
This
end
point
ongoing
and
then
you
can.
B
You
can
rip
out
whatever
software
you
had
in
there
and
just
make
sure
you're
still
returning
that
status
code,
which
is
why
I
actually
wonder
whether
we
need
to
say
maybe
410
or
404,
to
be
kind
of
to
allow
whatever
infrastructure
or
frameworks
that
are
being
used.
Kind
of
full
flexibility.
A
I
guess
the
I
guess
the
tricky
bit
is
404
implies,
to
my
mind,
maybe
something's
broken
and
you
you
might.
This
might
be
populated
again.
The
problem
is
just
on
our
end,
whereas
410
is,
this
has
been
deliberately
removed,
wipe
your
data.
B
C
A
Yeah
yeah
yeah.
Well,
okay,
let's
continue
that
discussion
then,
on
the
on
the
thread
I
mean,
I
think,
we're
close
to
resolution,
but
we've
got
we've
got
three
minutes
here
so
I'll
just
say:
aob
with
three
minutes
on
the
call.
B
Yeah,
I
think,
from
my
perspective,
that's
that's
good.
Obviously,
the
the
spec
itself
I
haven't
had
a
chance
to
look
at
yet
in
terms
of
once
they
fix
the
respec
issue.
You
can
read
it
in
a
friendlier
way.
B
Sure
no
worries-
and
I
guess
the
other
point
on
that
was
obviously
the
last
call.
I
think
we
talked
about
a
deadline
to
the
community
or
at
least
moving
this
all
forward,
so
I
haven't
seen
that
maybe
that's
being
pushed
back
because
of
the
schema.org
stuff,
which
now
is
in
some
state.
A
A
B
Well,
yeah,
I
guess
like
maybe
we
just
need
to
we
still
haven't
got
the
spec
in
we've
got
a
few
issues:
we've
integrated
stuff.
So
maybe
it's
a
case
of
just
doing
the
work.
We
need
to
do
to
get
this
back
into
a
place
where
it
does
everything
and
then
maybe
next
call
the
call
after
assessing
whether
yeah,
okay,
that
makes
sense
yeah
need
to
have
a
clear
well
basically
clear.
I
mean
whenever
you
want
for
me
personally,
to
review
I'm
happy
to
whenever
that
makes
sense.
Sure
yeah
well.
A
Like,
as
I
said,
I'll
fix
the
respect,
issues,
update,
update
the
github
issues
and.
B
A
I
think
some
of
the
issues
are
more
or
less
resolved,
in
which
case
I'll
put
them
in
the
spec
and
then
the
ones
that
need
discussion
will
be
discussed
yeah
but
I'll
ping.
You,
when
I've
got
this
fact,
readable
and
available
online
cool.
Okay,
okay,
talk
to
you
later,
nick,
all
right,
thanks,
jen,
all
right.