►
From YouTube: CORE WG Interim Meeting, 2021-05-12
Description
CORE WG Interim Meeting, 2021-05-12
A
Okay,
so
welcome
everybody
to
this
virtual
interim
meeting
of
the
co-working
group,
I'm
marco
de
galco,
my
co-chair
is
I'm
ximenez
and
I
should
show
you
this
usual
slide.
It's
an
official
idf
meeting,
so
the
not
well
applied.
We
are
recording
if
you're
not
familiar
with
this
take
time
to
become
familiar
with
it.
It's
not
just
about
ipr,
also
about
code
of
conduct,
so
be
nice
with
each
other,
and
the
agenda
for
today
is
about
three
items.
Mainly
first,
we
go
through
the
href
document.
A
We
had
the
resubmission
last
week,
so
it's
about
checking
the
status
and
the
remaining
open
points
before
it
can
proceed
to
working
group.
Call
second
point
is
an
idea
from
christian
on
a
possible
new
co-op
option
that
started
out
from
a
discussion
on
edoc
actually
and
third
point.
We
can
check
the
status
current
issues
and
way
forward
for
the
problem,
details
document.
A
No
okay,
then
we
can
right
move
on
to
the
first
item:
href
that
should
be
carsten,
and
you
want
me
to
share
anything
if
you
uploaded
anything
right
now,
kirsten
or
you're
going
to
share
slides.
B
B
We
can
look
at
the
cdl
briefly,
yes
and
see
how
how
the
internals
have
changed
a
bit.
B
B
More
okay,
so
what
you
are
seeing
there
is
the
cddl
for
a
cri
reference,
which
is
the
the
thing
that
allows
you
to
write
both
absolute
cris
and
the
things
that
get
combined
with
the
base
cia
to
do
relative
resolution
and
what
we
essentially
did
is
we
made
this
much
closer?
B
Oh,
that's
the
previous
version
yeah,
but
but
it's
close
enough
yeah.
I
didn't
remember
that
that
we
fixed
one
thing
in
the
repository,
so
essentially
the
the
3986
struct.
B
B
Which
contains
the
host
and
port
and
that's
the
part,
that's
wrong
in
this
city,
it
subsumes
the
the
scheme
under
the
authority
that
that's,
of
course
not
right.
Then
we
have
the
path,
the
query
and
the
fragment,
and
there
are
a
few
changes
here.
First
of
all,
as
in
cohab,
we
are
dividing
the
query
into
multiple
query:
parameters
separated
by
the
enzymes.
B
B
So,
instead
of
giving
a
a
scheme
and
authority,
you
can
give
a
discard
value,
and
this
is
essentially
a
consolidated
dot,
dot,
slash,
dot,
dot,
slash
dot,
dot,
slash
from
from
the
reference,
but
of
course
a
relative
reference
doesn't
have
to
have
a
dot
dot.
B
So
the
the
discard
value
you
get
if
you
just
give
a
path,
is
one
because
you
throw
away
the
relative
path,
throw
away
the
last
component
of
the
base
path
and
if
you
don't
have
a
path
at
all
and
and
give
only
give
query
and
end
or
fragment,
then
the
discard
value
is
zero,
because
you're
not
changing
the
base,
the
path
from
the
base.
B
So
that's
pretty
much.
It's
the
same
high-level
structure
it
was
before,
but
before
we
had
numbers
on
on
each
of
the
components,
and
now
we
have
a
syntax
that
you
essentially
have
to
step
through
linearly
to
find
out
which
parts
of
the
reference
go
into
which
component
or
we
call
them
section,
because
this
card
is
its
own
section.
B
And
if
you
go
down
to
a
table
that
that
should
follow
yeah
exactly
that
table,
that's
essentially
the
six
sections.
We
have
the
five
components
scheme
this.
This
is
really
authority
scheme
authority,
path,
three
and
fragment
are
work,
the
same
as
as
in
39.86,
and
this
card
is
that
that
number
that
compresses,
the
dot,
slash
stuff
and
tells
you
how
to
to
operate
on
on
the
base.
Part
and
the
e
value
essentially
tells
you
how
much
of
the
base
path
you
actually
throw
away.
So
as
soon
as
you
actually
give
a
path.
C
Understood
what
I'm
reading
correctly,
because
I
think
it's
the
first
time
I've
read
this
if,
if,
if
it
starts
with
a
false
or
-1,
it's
a
co-app
or
could
be
a
string,
but
if
it's
an
integer
0
to
127,
then
it's
a
discard.
Yes,
okay,.
B
So
actually
there
is,
I
don't
know
if
it's
in
in
this
version,
maybe
you
really
should
go
to
the
repository,
I'm
sorry.
We
have
submitted
that
when
we
had
about
eighty
percent
done
and
then
we
worked
on
it
some
more
and
now
we
have
about
ninety
percent
done
and
that's
in
the
repository.
B
Well,
that's
a
that's
a
great
question
because
essentially
we
have
to
decide
what
we
actually
want
to
do
here.
So
is
this
going
to
replace
your
eyes
and
I
think
it
won't.
It
really
is
a
representation
of
the
the
same
concept
that
your
eyes
use
in
the
context
of
constrained
resources.
B
B
Yeah,
I'm
sure
we
can
change
things
even
when
we
have
everything
in
a
row.
D
D
F
Usually,
if
you're
in
the
branch
and
then
click
on
the,
if
you
go
to
the
branch,
that
the
pull
request
is
based
on
and
then
click
on
the
on
the
editor's
copy,
then
you
redirect
it
into
new.
B
F
Syntax
but
don't
go
into
the
pull
request,
go
into
the
go
into
the
code.
F
Haven't
set
up
this
branch
that
is
not
master
but
okay,
new
syntax
and
then,
if
you
click
on
okay
yeah,
it
just
copy
isn't
built,
because
this
is
a
bit.
C
C
A
B
No,
no,
the
the
new
syntax,
the
the
news,
the
tip
of
the
new
syntax
per
request
is
not
submitted
yet
because
we
have
to
finish
this
one
thing
about
this
card
and
the
e-value
of
this
card.
B
B
Yeah
we
should
be
using
martin
thompson's
thing
and
and
then
it
would
be
easy,
but
we
haven't
done
that
yet
and
true
or
bytes
means
it's
an
authority
and
a
non-negative
integer
is
discard
and
the
text
is
path
and
array
is
query
and
null
value.
If
you
see
a
null
value
as
the
first
element,
then
you
have
a
fragment
only.
So
this
is
kind
of
the
the
prefix
table
that
anybody
who
knows
about
parser
generators,
probably.
D
Recognizes
maybe,
as
a
quick
background,
why
we
made
this
change
a
long
time
ago
we
had
an
interim
meeting
where
we
discussed
this
very
same
section
and
the
cdl,
and
we
notice
that,
of
course,
when
we
have
our
euro
eyes
and
cris,
we
need
to
store
the
values
of
these
components
and
we
also
need
to
have
delimiters.
So
we
know
when,
where
the
values
start
and
and
and
what
they
mean,
and
at
that
point
in
time
we
had
in
cddl
where
the
delimiters
were
a
bit
verbose.
D
D
The
downside-
maybe
of
of
this,
is
that
you
have
to
spend
a
bit
more
time
on
thinking
on
how
to
implement
this,
and
the
code
for
for
decoding
cri
also
becomes
slightly
larger,
because
you
know
you
have
to
basically
implement
a
small
parser
that
that
you
uses
a
one,
look
ahead
to
determine
what
what
is
the
lecture
structure
of
a
cri
instead
of
having
attacked
key
value
pairs
that
tell
you
immediately
what
it
means.
D
So
I
I
think,
I'm
very
happy
personally
with
the
compactness
of
the
new
format.
I
think
I
need
still
need
to
do
sleep,
one
or
two
nights
over
this.
If
I'm
happy
with
the
slightly
more
complicated
parser
that
everybody
would
have
to
implement
the
problem
a
bit
overall
with
your
eyes,
is
that
nothing
of
this
is
intuitive.
Most
of
your
eyes
are
defined,
algorithmically
and
trying
to
understand
what
discard
and
and
so
on
means
often
leads
to
frustration.
D
You
have
to
implement
it,
and
then
your
code
tells
you
what
happens
when
you
do
different
things.
It's
a
bit
unintuitive
to
try
to
understand
what
exactly
is
going
on.
That's
with
the
frustrating
part
here.
B
Yeah,
so
essentially
what
micro
does?
Is
it
just
steps
through
the
curry
and
finds
those
six
sections
that
are
shown
here
and
essentially
puts
pointers
for
these
six
sections
somewhere
into
a
variable?
So
you
know
where
they
are
in
in
the
structure
and
then
essentially
you
have
a
fully
decoded
ci
in
front
of
you.
F
Those
don't
even
necessarily
need
to
be
saved
in
between
so
I've
done
a
few
experiments
with
with
the
rust
based
cri
parser,
and
there
are
optimizations
that
basically,
if
done
right,
the
the
compiler
should
be
able
to
optimize
out
much
of
this
and
become
and
and
then
relative
resolution
becomes
a
becomes
a
topic
of
parsing
over
those
two
two
streams
and
logs
again.
B
B
No,
I
think
we
just
have
to
to
finish
those
last
ten
percent,
and
so
essentially
what
we
did.
B
B
A
bucket
of
algorithms
that
can
do
something
on
text
strings
and
we
abstracted
the
information
model,
and
then
we
we
implemented
the
information
model
in
a
different
representation,
but
still
the
same
information
model
with
a
few
simplifications-
and
I
already
mentioned
percent
encoding,
which
allows
a
lot
of
complexity
in
uis
and
and
we
got
approximately
eighty
percent
of
the
value
of
of
percent
encoding
by
by
doing
thing
rights
in
the
first
place
and
by
by
actually
parsing
the
the
query
component
into
an
array.
B
So
there
is,
you
would
have
to
have
a
very
weird
application
that
still
needs
percent
encoding
here,
so
we
can
yeah
thomas.
My
problem
is
my
internet
connection
is,
is
from
deutsche
telekom
and
it's
not
working
very
well
at
the
moment.
B
Okay,
so
we
actually
had
a
had
a
class
explored
an
hour
ago,
because
telecom
completely
lost
the
routing
and
sorry.
D
Yeah,
what
what
I
think
is
left
to
do
is
finalizing
the
the
thoughts
on
the
discard.
D
Then,
as
I
said,
I
think
we
have
a
few
bucks
in
in
the
document
and
we
we
need
to
flush
those
out
and
the
best
way
to
do.
That
would
be
if
we
could
get
a
bunch
of
independent
implementations
that
literally
just
look
at
the
draft
and
and
try
to
create
an
introverbal
implementation
from
that
and
and
then
we
we
might
have
some
questions
on
disagreements
on
and
how
things
are
to
be.
So
those
implementations
we
have
so
far.
The
authors
are
a
bit
spoiled
by
by
talking
to
each
other.
D
C
C
So
can
you
say
again
so
this
work
is
going
to
be
referenced
in
corral,
I
think
and
then
who
is
there
other
users
of
it
that
you
anticipate.
E
D
I
can
imagine
that,
basically
everywhere,
where
you
want
to
express
your
eyes
in
seabor,
it
would
be
a
candidate
to
use
sierra
eyes
instead.
So,
for
example,
carsten
has
his
authorization
iaef
format,
which
is
using
a
lot
of
your
eyes,
and
that
might
be
also
a
candidate
for
cris.
C
So
I'm
just
thinking
so
I
would
recognize
that
it's
a
this
thing
from
a
tag
or
from
the
fact
that
it
starts
with
a
an
array
rather
than
a
a
byte
array.
If
I
had
something
that
was
already
using
your
eyes,
could
I
how
could
I
put
this
in
unambiguously.
C
F
C
A
A
A
Shot.
Okay,
thank
you.
We
can
move
to
the
second
topic,
the
new
option.
From
christian
and
kristen.
You
want
to
try
sharing
your
screen,
you're
doing.
F
F
F
F
You
you
send
18
bytes
of
well
known,
slash
adhoc,
which
is
a
significant
portion
of
the
of
the
length
of
a
of
a
message,
one
and
yeah.
So
the
the
question
that
came
up
basically
during
during
during
during
the
hackathon
before
the
last
itf
was:
can
we
do
better,
and
my
impression
is
that
we
can
quite
easily
do
better
by
simply
defining
an
option
that
shortens
that
uri.
F
It
would
be
kind
of
an
alia
alias
to
to
having
the
uri
path
option,
and
that
would
be
it,
and
if
someone
were
motivated
to
say
that
here
this
could
be
your
eye
value,
then
it
could
be
used
for
well-known
core
two
one
of
those
would
get
the
three
bytes
out
quite
optional.
One
of
those
would
get
a
four
byte
option,
but
that
would
be
relatively
simple.
F
The
question
to
me
so
the
question
here
is:
can
we
do
better
and
I
think
we
can
even
without
doing
it
differently
on
the
wire,
because.
F
Well,
the
basic,
basically
every
in
every
situation,
where
you
would
serialize
those
two
options:
11
dot
well
known
and
ad
hoc.
Instead,
you
serialize
a
bit
later
in
the
in
the
options
stream
and
option
501
and
the
your
eye
parser
at
the
at
the
other.
End
of
the
thing
needs
to,
of
course,
needs
to
recognize
that
option
and
says
yeah.
I
haven't
seen
any
uri
option
or
proxy
scheme,
but
I
see
this.
F
I
know
this
and
instead
of
kind
of
going
through
my
tree
of
of
resources,
I
right
away
slot
this
into
well
known
at
home.
F
C
And
your
point
three:
is
it,
are
you
giving
it
a
specific
thing
for
core,
or
are
you
giving
it
a
a
a
ayanna
table
for
future
well-knowns.
F
So
if,
if
I'm
going
back
one
slide,
I'm
not
doing
anything
at
all
so
far,
this
is
just
introducing
the
idea
and
all
those
are
of
course
options.
So
we
could,
we
could
have
just
hard-coded
these
two
things.
We
could
have
a
nayana
table.
Okay,
this
is
pretty
much
open
open.
So.
B
Far,
we
would
map
play
your
zero
to
ad
hoc,
and
then
you
want
to
call
and.
E
B
F
And
I
mean
it's,
it
would
be
you
and
options,
so
it
would
be
virtually
unlimited
anyway
yeah.
I
I
understand
so
this
is
something
we
could.
I
mean
if,
if
it
turns
out
that
ad
hoc
runs
into
a
limitation
there,
this
could
be
done
there
locally,
defining
that
option,
leaving
and
basically
only
needing
needing
that.
One
point
clarified
about
whether
we
assign
core
manually
whether
we
have
ayanna
policy
those
details.
F
The
other
thing,
of
course,
is
that
this
creates
an
interoperability
problem,
because
parties
may
or
may
not
know
what
that
is
short
for
and
for
ad
hoc
it
would
be
pretty
clear-cut,
because
if
this
were
to
be
created
with
ad
hoc,
then
this
should
be
mandated
for
fighter
implementations
and
problem
solved.
F
Retrofitting
it
into
well-known
core,
of
course,
is
not
that
easy
and
would
usually
create
a
for
or
something
bad
request
case,
which
would
then
need
to
be
resolved.
So
typically,
this
would
be
used
in
deployments
where
everyone
is
aware
that
this
is
being
used
here,
because
we
are
so
also
constrained,
but
it
might
be
nice,
and
this
is
where
the
generalization
topic
starts.
F
It
might
be
nice
to
allow
intermediaries
to
recognize
that
and
possibly
even
help,
recognizing
it
or
allow
libraries
to
consume
those
shortcut
options
in
a
more
in
an
easier
way
than
having
to
manually
implement
additional
code
for
an
additional
option,
and
I
think
this
is
where
shiki
could
come
in.
F
So
everything
I'm
describing
from
here
on
is
purely
a
matter
of
describing
the
same
procedure
differently.
It
would
on
the
wire
and
in
some
implementations,
look
just
as
I
described
it
before,
but
implementations
could
choose
to
use
an
understanding
that
is
based
on
chick
and
then,
for
example,
be
updated
will
easily,
because
I'm
thinking
of
the
ao
co-op
implementation
that
I
maintain,
I
would
not
implement
a
new
option,
but
just
add
an
entry
to
a
table
of
of
known
options
that
have
this
sick,
compressing,
deep
compression
style
route.
F
So
once
there
is
a
format
for
for
compression
and
decompression
rules,
and
I
think
there
is
not
yet
this
can
be
become
machine,
readable
and
used
for
different
options.
So,
for
example,
we've
talked
about
the
the
binary
host
names
that
that
corey
cory's
offer
at
some
point.
Someone
could
be
annoyed
by
having,
by
the
only
up
place
in
their
own
in
their
embedded
code
that
still
has
hex
encoding
and
there
being
the
being
or
decimal
encoding
being
the
uri
host
option,
because
it
has
a
byte
string.
F
That
is
a
an
ip
literal
and
describe
maybe
even
for
their
own,
even
only
for
their
own
application,
a
uri
host
literal
option
that
transforms
the
byte
string,
that
is
4
bytes
for
ipv4
and
16
bytes
for
ipv6
back
and
forth
into
a
uri
host.
That
is
st
or
hex
encoded,
and
I
think
there
are
there.
It's
not
not
in
the
not
in
the
currently
registered
options,
but
in
in
in
the
options
that
are
around
in
in
current
internet
drafts.
I
think
there
is
some
potential
for
describing
some
options
in
terms
of
such
compression.
F
That
doesn't
mean
that
they
have
to
be
normatively
described
that
way,
and
it
doesn't
mean
that
they
have
to
be
implemented.
That
way,
but
I
think
it's
it
would
be
a
convenient
check
to
see
that
those
option
semantics
are
easy
to
understand
because
they
can
be
understood
in
terms
of
another
option.
That
is
just
a
bit
more
verbose
or
that
would
need
that
was
actually
sent
in
the
request,
not
in
the
response.
F
F
Now,
there's
only
a
limited
set
of
responses
that
can
be
combined
back
there,
but
then
again,
there's
only
two,
but
by
the
very
nature
of
how
the
how
oscar
attack
works,
there
is
only
a
limited
set
of
outcomes
as
well,
so
I
think
it
would
be
expressible
that
way.
The
questions
that
come
from
this
are
then
hey.
Do
we
want
this
especially?
Do
we
want
this
urgently,
because
so
how?
How
pressing
is
the
need
on
the
network
side
to
to
save
those
15
bytes?
B
F
I
don't
even
think
that
in
I
think
that
in
many
cases
this
would
not
be
adding.
To
that
note,
it
would
be
more
like
adding
shake
to
the
code
generator
that
creates
your
co-op
library,
possibly
or
not
at
all,
or
just
to
use
it
as
an
understanding.
So
if,
if
some
component
already
does
use
chick,
then
they
will
probably
implement
it
that
way,
because
they
have
that
all
already
available.
If
it
does
not,
there
is
still
I
mean
you
can
implement
the
you
can
implement
everything.
That's
in
it
all
squared
hope.
F
Without
this
no
question
you
it's
just
an
alternative
way,
that's
opened
by
describing
it
as
and
by
the
way.
It
also
works.
If
you
view
it
through
that
lens.
C
B
So
there
could
actually
be
an
appendix
somewhere
that
shows
how
to
use
shake
properly
to
to
make
this
great
for
ad
hoc
and
make
all
the
numbers
look
really
great.
Yeah.
F
It
could
it.
It
could
also
be
just
a
bit
a
piece
of
of
guidance
on
how
to
on
on
kind
of
a
check
that
you
can
do.
If
you
define
a
new
call
option
that
says,
if
you're,
defining
a
code,
option
and
you're
unsure
about
whether
you're
doing
it
right,
try
to
phrase
it
in
terms
of
existing
options
and
just
make
it
a
conversion,
and
then
you'll
see
whether
it
works
out
or.
D
D
Yeah,
I
think
this
is
interesting
christian.
I
need
to
digest
it
a
little
bit
more,
but
I
mean
the
general
approach
I
think
is
right.
I
mean
from
from
from
the
set
things
where,
where
we
have
these
overhead
issues,
then
at
least
addressing
the
the
well-known
uri,
I
think,
is
a
step
in
the
right
direction
and
then
maybe
even
go
further,
but
I
I
need
to
think.
A
More
just
a
comment
here
on
the
parallel
between
this
and
the
addoc
option
for
the
oscar
dock
document.
Maybe
the
parallel
with
well
not
edok.
Exactly
for
that
combined
request
is
a
particular
case
and
not
a
general
case,
because
from
another
point
of
view,
that
combined
request
is
trying
to
target
the
exact
same
medical
resource
that
was
targeted
by
adolf
message
one.
A
F
Yeah,
but
I
wouldn't
rely
on
an
ad
hoc
option
hitting
anything
other
than
the
well.
Not
I
have
because
if
you
have
multiple
ad
hoc
resources,
those
might
those
might
be
completely
independent
processors
because
they
are
for
completely
independent
applications
and
those
might
even
use
the
same.
The
same
the
same
number
space
in
different
ways,
because
they
rely
on
the
request
coming
in
over
that
channel
again.
F
So,
if
also
the
else
correct
hook,
option
were
to
be
used
with
any
other
ad
hoc
resource.
I'd
be
a
bit
careful
here.
A
D
Propose
I
mean,
are
the
different
options
here,
or
is
it
one
of
one
proposal
with
more
potential
granularity,
so
that
I
mean
other
op?
Is
one
option,
one
alternative
that
we
define
a
co-op
option
which
handles
your
eyes
or
sorry
well
known
somehow
that
compresses
those.
F
A
thing,
a
single
option
that
deals
with
well
not
with
well-known
uris,
is
what
I
think
is
the
the
core
of
every
way
this
could
go
the
the
and
whether
we
do
that
depends
on
how
how
important
it
is
for
the
network
applications
to
to
have
this
and
anything
else.
That
is
then
basically
just
is
background
is
additions
on
top
of
that,
but
it
wouldn't
change
how
that
single
option
say
501
would.
D
Look:
okay
thanks
because
I
think
that
that
is
definitely
worth
doing,
but
not
I
mean
not
restricted
to
endo,
but
for
for
a
general
case.
I
think
so.
I'm
in
favor
of
that.
F
More
or
less
going,
I
was
going
for
the
for
the
two
byte
space
taking
one
of
the
lower
ones.
That
would
not
usually
be
hit
by
any
combination
of
earlier
options,
plus
earlier
options,
plus
a
small
data.
F
Primarily
because
the
the
the
suggestion
is
to
what
I'm
trying
to
get
across
is
that
we
could
do
this
without
even
without
standards
action.
If,
of
course,
we
want,
we
get
the
experts
to
agree
that
this
is
something
useful.
Then
we
could
lower
it
into
the
into
the
one
byte
into
the
oneplus
one
byte
branch,
and
it
would
work
there
just
as
well.
E
D
I
have
a
question
on
a
talk
that
may
be
stupid,
but
I
haven't
read
the
draft.
If
I
have
a
uri,
how
do
I
know
if
I'm
supposed
to
use
dtls
or
attack.
D
B
F
This
general
assumption
that
anything,
that's
signed
by
the
by
the
by
any
of
those
all
known
to
be
good
cas
is
valid,
but
can't
go
up
as
you
could
be
in
either
situation
and
whatever
tells
you
whether
you
you're
supposed
to
trust
the
the
large
list
of
cas
or
whether
you're
supposed
to
have
a
key
for
that
for
the
particular
application,
as
it's
used
with
with
lightweight
m2m,
will
also
tell
you
whether
you
are
supposed
to
use
anything.
D
So
essentially,
every
server
could
have
its
own
endpoint
for
setting
up
ad
hoc.
It
doesn't
have
to
be
well
known
if
you
can
discover
that
server
specific
endpoint
somehow
and
while
we're
at
it.
We
also
have
the
situation
and
coral
that
you
want
to
agree
on
the
vocabulary
or
some
dictionary
or
vocabularies,
and
we
sticking
that
into
the
same
context.
What
simplify
things
a
lot.
D
But
yeah
I
mean
the
use
of
well-known
ad
hoc
is
is
an
example
in
in
in
in
the
ad
hoc
craft.
So
so
you
don't
need
to
use
it,
but
it's
sort
of
one
one
way
to
get
this
uniform
and
an
easy
way
to
to
identify
ad
hoc
messages,
but
as
you
as
you
propose
on,
as
as
I
said
before
here,
this
is
really
up
to
the
application.
How
you
want
to
identify
that
you
receive
other
messages,
so
you
can
assume
things
that
are
available
to
both
sides.
D
D
F
I
think
there's
a
bit
of
a
there.
There
are
different
application
spaces
and
some
have
this
context
and
some
do
not
and
class,
as
you
mentioned,
if
you
get
tossed
the
uri
and
I'll
told
to
start
with
that,
then
you
like
that
context
and
and
then
well
known
course,
pro
well
known
at
hope.
It's
probably
a
good
good
way
to
try
starting
atop
and
see
what
comes
with
it.
D
So
but
unknown
coral
next,
can
you
stick?
Our
bouquet
varies
into
a
well-known,
maybe
maybe
that's
some
some
follow-up
difficult
discussion,
but
we
keep
discussing
these
context.
Information,
for
I
think
quite
some
time,
and
one
area
is
definitely
these
security
parameters
and
another
area
now
shows
up
in
coral
with
negotiating
dictionaries
and
vocabularies.
B
Yeah,
I
think
what
what
most
people
don't
don't
think
about
is
the
situation
where
you
get
trust
the
uri
is,
is
I
mean
that's
not
exactly
how
it
works?
B
There
must
be
a
reason
to
actually
dereference
that
ui,
and
that
reason
is,
is
the
first
element
of
that
context
and
if,
if
the
the
that
part
of
the
system
that
supplies
the
reason
can
also
define
other
things,
then
we
have
more.
In
that
context,.
D
G
G
Which
is
maybe
a
good
idea
or
having
you
know
the
id
format,
backed
by
something
else,
which
is
not.
B
A
G
G
Okay,
a
quick
recap
of
the
timeline,
so
we
adopted
the
draft
in
may
2020
and
on
the
premise
that
this
would
be
a
coral
embodiment
rather
than
the
the
royal
seaboard
that
I
that
that
was
in
the
individual
draft
and
and
the
idea
was
that
we'd
used
it
to
put
to
the
test
some
of
the
coral
machinery
and-
and
this
was
zero
zero.
Then
we
published
zero
one
just
before
the
cutoff
preceding
itf-108.
G
If
I
remember
correctly
and
and
this
included
some
restructuring
of
the
of
the
content
and
and
possibly
also
a
bunch
of
small
fixes.
G
And
then
the
discussion
revolved
around
a
bunch
of
pretty
thorny
design
points.
Basically,
the
the
extensibility
of
the
semantics
of
pts
and
and
in
parallel,
how
one
discovers
any
per
api
extended
semantics.
G
G
We
have
restarted
the
activity
on
on
pt,
together
with
the
rest
of
coral,
okay.
So
next,
please
so
open
issues.
If
so,
so,
if
you
look
at
the
open
issues
on
github
you'll,
see
roughly
three
categories.
G
First
is
the
purely
editorial
staff
that
just
sits
there
waiting
for
one
of
us
to
to
you
know
to
shake
his
laziness
and
go
through
the
10
15
minutes
rigmarole
to
drag
the
issues
done,
and
the
second
category
comprises
a
bunch
of
issues
that
are
related
to
the
completion
of
this
so-called
common
data
model.
G
The
common
data
model
is
split
roughly
in
two
regions.
First,
the
set
of
basic
attributes
that
we
know
have
general
applicability,
even
if
they
won't
be
there
in
any
case
they
can
be,
they
can
be
optional,
but
we
know
they
exist
for
all
cases
and
then
there's
a
second
set
of
optional
attributes
that
we
call
per
feature.
Clouds
came
up
with
this
momentum
and-
and
these
are
targeted
to
use
cases
that
we
think
are
common
enough
to
be
worth
codifying.
You
know,
but
but
they
may
not
make
sense
in
the
general
case,
for
example.
G
If
so,
if
you
need
to
forward
the
api
errors
to
a
to
your
log
analytics
engine,
you
may
need
to
enrich
the
the
basic
error
report
with
some
specific
declaration,
for
example
the
cooperative
code
that
would
otherwise
get
lost
at
the
transaction
layer
right,
and
this
is
the
kind
of
attributes
that
we're
looking
at
here
per
use
case,
and-
and
this
is
work
in
progress.
We
know
what
to
do.
We
have
identified
the
various
bunch
of
categories
that
we
want
to
codify.
G
We
haven't
done
it
fully
yet
there's
an
open
pr,
I
think,
pending
since
a
while
and-
and
this
may
be
an
order,
magnitude
or
more
for
compared
to
to
what
is
needed
to
the
editorial
stuff
that
I
was
mentioning
before,
but
but
it's
still
bounded
effort
and
then
and
then
there's
a
third
category
and
and
this
this
bit
is
concerned
with
the
extensible
part
of
the
data
model,
and
this
is
where
we
spent
most,
if
not
all,
of
our
meeting
time
until
december
last
year,
going
through
the
problem
space,
you
know
analyzing
the
existing
solutions
and
so
on,
and
we
don't
have
a
complete
answer.
G
Yet
we
came
up
with
a
tentative
proposal
that
currently
sits
in
in
a
draft
that
is
unpublished,
which
is
likely
to
need
some
polishing
before
we
can
present
now.
The
goal
that
we
set
for
for
these
mechanisms
are
quite
ambitious.
We
want
it
to
be
easy
to
register
for
a
developer
with
ideally
zero
friction
and
be
stable
as
well.
We
still
want
to
allow
a
computer,
a
representation
of
of
of
the
data.
G
We
have
the
complementary
problem
of
discoverability
of
the
api
characteristics
and
pd
sits.
There
is
one
instance
of
the
bigger
problem
right.
G
We
also
have
a
composability
problem,
and
by
that
I
mean
how
we
mix
and
match
different
api
spaces
in
a
non-clashing
way,
and
then,
on
top
of
that,
we
want
to
avoid
segmentation
between
private
and
public
spaces.
In
fact,
my
the
initial
proposal,
we
did
suffer
from
the
what
carson
called
the
x-dash
problem,
but
which
means
the
fact
that
private
code
points
can
trickle
and
persist
into
the
public
sphere.
G
G
Okay,
so
what's
the
plan,
this
is
a
tentative
thing,
so
I
would
like
to
discuss
this,
but
before
some
idea
is
that
before
the
cutoff
next
meeting,
we
we
should
address
the
all
the
issues
in
category
a
and
b
from
from
the
previous
slide,
it
looks,
looks
like
an
achievable
plan
to
me.
I
would
like
to
hear
klaus
and
heimer.
Jamie
is
not
here
right,
but
klaus
is
here.
What
do
you
think.
D
I
have
to
admit
that
I
haven't
read
the
latest
draft,
so
I
I
think
that
there's
the
opportunity
to
make
some
progress
until
the
the
idf
deadline,
I'm
not
sure
how
far
we
can
get,
but
we
should
definitely
bring
bring
the
latest
ideas
into
the
draft
and
and
see
how
much
we
can
do
so.
I
I'm
into
it
for
the
next
couple
of
weeks.
D
G
Okay,
but
you
know
it
it's
looking.
I've
looked
recently
at
the
issue
list
and
scanned
through
the
various
things
that
are
there,
and
there
are
a
few
of
them
that
that
seems
very
reasonable
and
achievable.
So
I
think
I
would
do
that.
D
Problem
details
draft
is
probably
mostly
an
exercise
and
defining
the
vocabularies
for
for
the
different
error.
G
Types,
that's
exactly
exactly
what
I
wanted
to
say.
So
these
two
categories,
a
and
b,
can
be
solved.
You
know
because
they
pertain
the
data
model
and
they
are
abstract.
But
then
we
have
the
the
accessibility
part
that
we
need
the
wider
choral
discussion
regarding
the
dictionaries
and
discovery
to
converge
into
some
usable
mechanism
and,
as
you
say
once
once
that
is,
that
is
done.
It's
just
a
matter
of
reusing
the
relevant
bits
of
whatever
we
have
designed
there
in
in
pd,
but
but
until
that
is
done
it.
This
draft
needs
to
wait.
D
G
D
So
you
can
already
use
vocabulary
from
different
places
in
the
same
corridor
document
to
just
the
mechanics
on
how
you
do
it
exactly
on
the
wire
is
not
finalized.
Yet.
G
Okay,
so,
but
then,
of
course,
right
yeah
well,
at
the
same
time,
if
you
want
to
look
at
it
positively,
this
could
become
an
incentive
to
make
progress
on
the
other
side
right.
So
since
I
tend
to
be
an
optimistic
fellow.
G
You
know
yeah,
I
think
I
think
this
is
where
the
presentation
ends,
and
this
is
sort
of
the
plan
for
the
short
term
and
mid
term.
A
B
Before
we
go
into
the
details,
I
have
two
two
more
high-level
questions.
One
is
the
the
json
problem
details
document
has
been
out
for
a
while.
Is
there
maybe
anything
we
can
learn
from
that.
G
Experiment-
I
don't
know
they
are
doing
the
piece
right
at
this
point
in
time,
so
maybe
they
have
gone
through
the
process
of
absorbing
lessons
from
from
existing
deployments.
G
Eight
or
seven
b's
in
yeah
http
api.
You
would
find
okay.
B
B
So
essentially,
even
even
though
it
really
isn't
that
important
from
a
technical
point
of
view,
from
a
point
of
view
of
of
getting
the
the
processes
in
place
and
and
the
inner
stuff
and
so
on,
I
think
the
it's
really
important
to
think
about
how
you
can
avoid
doing
your
first
second
class
citizen
thing
here.
B
The
the
one
area
where
I
think
we
managed
to
hit
that
is
in
the
the
core
sids,
and
maybe
we
should
think
about
how
because
sits-
or
maybe
an
approach
like
horse
hits,
but
but
we
we.
I
hope
we
soon
have
because
it's
available
to
us
how
that
can
actually
solve
the
the
problem
of
having
to
give
different
efficiency
classes
to
different
people,
because
I
mean
in
reality,
sids
will
will
be
four
bytes
or
less.
So
it's.
G
D
Okay
yeah,
so
in
our
discussions
we
came
up
with
a
bunch
of
criteria
like
we
really
want
to
have
a
low
barrier
to
entry.
Ideally,
if
you
just
want
to
prototype
your
application,
you
shouldn't
have
to
first
interact
with
an
ina
registry,
or
so
before.
You
even
get
can
put
some
dummy
values
in
into
your
proof
of
concept
and
then,
of
course,
we
want
to
have
compact
identifiers.
D
We
want
to
be
able
to
mix
and
match
error
types
and
and
error
details
from
from
different
domains
and
and
so
on,
and
we
we
have
looked
at,
sits
and
and
sits,
are
good
for
the
the
the
the
the
world
where
it's
applied
to
so
you
you
have
these
consecutively
increasing,
ids
very
frequently,
and
you
have
already
some
mechanism
to
formally
define
the
the
available
properties
and
and
so
on.
So
this
is
a
really
good
approach
to
that.
D
But
I
guess
coral
and
problem
details
in
the
end
might
be
even
less
formally
organized
than
than
the
comai
world,
so
you
have
really
developers
who
are
building
some
apis
and
they
need
an
error
code
and
they
just
need
to
put
a
value
somewhere
in
their
code
and
and
not
go
to
a
website
first
and
register
an
account
and
and
so
on,
and
one
proposal
that
we
are
having
now
is
these
namespace
tables,
as
we
have
called
it.
D
It
has
a
github
issue
which
allows
you
at
least
to
use
items
on
multiple
vocabularies,
and
it
doesn't
matter
if
those
vocabularies
are
then
some
private
ones
or
public
ones.
You
can
just
pull
from
from
any
vocabulary
that
exists
and
it
doesn't
even
have
to
be
defined.
For
for
this
purpose,
it
just
has
to
have
a
certain
number
scheme.
D
So
a
lot
of
ayana
registries,
for
example,
can
be
already
pulled
in
using
these
name
service
tables,
but
there
are
still
some
unresolved
issues,
even
with
those
namespace
tables,
and
in
particular,
how
do
you
know
which
namespace
table
to
use-
and
that
brings
us
to
the
earlier
point
on.
If
we
have
more
context
for
uri,
but
we
stick
our
security
information
and
our
vocabulary
name
space
tables
into
that
as
well.
So
there's
still
a
bit
of
work.
D
A
We
have
some
minutes
more,
just
a
double
check
with
you,
thomas.
If
you
wanted
to
go
through
some
issue,
in
particular,.
A
A
B
B
I
just
mentioned
something
completely
different,
but
that
may
be
useful
in
a
world
that
uses
sibo
and
json
and
all
these
things
you
know
about
the
json
path
activity
and
on
tuesday
we
had
a
long
discussion
about
whether
we
could
be
doing
regular
expressions
there
and,
of
course
the
main
problem
is
that
there
are
so
many
to
choose
from
and
so
yeah
this.
B
B
So
if
you
care,
you
might
want
to
have
a
look
at
the
iregx,
a
proposal
that
I
submitted
today
and
I
hope
we
will
be
able
to
integrate
this
into
jsonpath,
but
it
might
also
be
useful
in
other
query,
languages
like
in
the
discovery
context
for
for
core
and
so
on.
So
yeah
just
just
have
a
look.
If
you
care
about
regular
expressions,
if
you
don't
don't
waste
your
time.
A
Of
course,
we'll
meet
again
in
two
weeks
same
time
same
venue.
The
main
topic
for
that
meeting
is
coral
and
the
main
coral
document,
and
also
fully
a
comparison
or
discussion
around
its
relation
with
sdf.
C
C
C
I
I
have
implemented-
and
we
are
depending
upon
this
code,
this
stuff
to
proceed
in
the
anima
constrained,
voucher,
work,
okay,
so
yeah.
C
B
C
B
Also,
merge
them
and
publish
a
new
draft
and
and
everything
yeah
well,
that
would
be
step
two
yeah.
B
Yeah,
I
think
the
the
chairs
at
some
point
need
to
decide
who
the
authors
of
these,
who
the
editors
of
these
these
documents
are
so
yeah,
but
I'm
also
interested
in
knowing
are
there
technical
reasons
why
things
are
as
weird
as
they
are
in
in
these
areas,
and
that's
why
I
really
would
like
to
talk
with
the
existing
authors
to
get
this
done.
So
I
wouldn't
do
this
lightly.
D
B
So
I
think
the
the
interesting
thing
to
look
at
is
the
the
web
of
things.
People,
the
w3c
people,
have
thing
descriptions,
and
now
they
are
emulating
sdf
with
the
thing
description
templates
and
maybe
that's
something
we
want
to
look
at
in
the
choral
space
as
well,
so
we
we
have
instances
and
classes,
and
now
I
have
no
idea
how
this
really.
B
Yeah,
mostly
about
the
the
incense
as
a
star
thing,.
D
B
B
G
An
ignorant
question
here,
so
how?
How
do
you
serialize
it
a
thing,
description
thing.
F
I've
looked
briefly
at
the
draft
and
it
looks
like
it's
primarily
really
taking
taking
a
json
ld
ensuring
there
are
no
inline
contexts
around
anymore
and
then
translating
it
pretty
much
won't
want
to
to
see
more
full
compactness
with
some
optimizations
but
model
wise.
It's
it's
a
translation.
E
G
A
Okay,
guys,
if
there
are
no
other
points,
we
can't
hear
the
meeting
here
thanks
all
for
a
good
discussion.
Talk
to
you
later
in
two
weeks.