►
From YouTube: IETF-CORE-20221012-1400
Description
CORE meeting session at IETF
2022/10/12 1400
https://datatracker.ietf.org/meeting//proceedings/
A
A
A
A
A
A
So
it's
three
past
I
think
we
can
start
welcome
everyone
to
this
internal
meeting
of
core.
This
is
Marco.
My
coaches
are
I'm
Jimenez
and
Carson
Borman
and,
as
usual,
the
not
well
applies.
Please
be
familiar
with
that.
It's
not
just
about
IPR.
It's
also
in
the
special
about
our
product
conduct,
so
be
nice
with
each
other,
and
the
agenda
for
today
is
most
about
two
points.
A
Some
proposals
to
discuss
some
possible
additions
to
the
href
documents
that
I
discussed
with
some
other
guys
a
couple
of
weeks
ago
and
and
then
a
topic
related
to
core
and
an
anima
that
will
also
continue
during
the
iotops
interim
meeting
following
or
in
fact,
we
plan
to
close
this
meeting
latest
at
15
UTC
and
to
switch
to
iot
Ops.
A
Where
again,
especially
this
second
topic
from
the
agenda
will
be
discussed,
but
we
can
do
some
early
discussion
and
preparation
here
already
I'll
try
to
take
notes,
but
any
help
is
appreciated
other
than
that
any
more
items
for
the
agenda
or
any
bashing.
A
I've
just
had
some
proposal
to
share
related
to
the
URI
scheme,
encoded
in
the
in
the
CRI
in
particular.
This
is
just
a
snapshot
from
from
the
current
draft
from
the
cddl
definition
of
the
CRI,
where
the
URI
scheme
can
be
expressed
for
sure
as
a
simple
text
string,
but
possibly
also
as
a
negative
integer,
and
currently
there
is
a
list
of
well-known
most
important
schemes
with
their
Associated
negative
integer,
so
I
started
to
think
how
we
can
ensure
interoperability
also
in
an
extensible
way.
A
The
moment
additional
schemes
may
be
considered
and
if
we
want
to
use
in
a
compact
way,
these
negative
integers.
Basically,
since
there
is,
there
is
no
registry
for
those
integers
at
the
moment.
I
think
this
already
holds
from
a
general
point
of
view,
but
otherwise
there
have
been
a
few
cases
in
point
about
this
foreign
yeah,
and
this
is
the
main
one
that
actually
triggered
this
thinking.
A
We
have
the
multicast
notifications
document,
for
which
there
is
a
pull
request
about
using
Cris
on
it,
in
fact,
so
this
is
an
extract
from
from
the
pull
request.
So
what
we
have
been
proposing
there,
where
essentially
information
specified
by
the
server,
indicate
as
a
CRI,
the
addressing
information
of
the
server
and
this
parameter
TPI
SRV
and
then
additional
information
to
put
all
the
clients
on
the
same
page
also
expressed
as
a
CRI.
A
We
also
used
to
have
an
Ayana
registry
around
basically
to
to
give
indication
of
what
TPI
details
should
say,
depending
on
the
particular
scheme
used
in
TPI
SRV,
and
that
registry
will
also
become
it
becomes
easier.
The
moment
we
switch
to
to
CRI
so
here
on
the
on
the
right.
You
can
see
how
the
the
registry
would
look
like
with
its
new
structure
and
if
you
check
also
the
proposal
and
the
pr
in
fact
that's
already
thinking
of
negative
integers
as
registered
somewhere.
A
But
again
we
don't
have
a
registry
at
the
moment
and
related
to
this
well,
a
subset
of
TPI
SRV.
Basically,
with
all
this
in
mind
again
and
would
be
used
also
in
another
document,
where
a
co-op
option
would
indicate
addressing
information
of
the
server,
although
limited
only
to
to
the
scheme
and
Authority
without
local
part.
But
still
the
scheme
is
involved
again.
A
So
putting
all
this
together,
it
seemed
like
there
were
enough
reasons
to
start
thinking
of
introducing
an
actual
registry
to
track
those
negative
integers
I've
had
a
discussion
about
this
with
Thomas
hosati
and
hang
on
a
few
days
ago,
but
I
wanted
to
bring
it
here
to
The.
Proposal
is
actually
it's
actually
about
two
points.
One
is
pretty
easy.
A
We
just
noticed
that
the
list
of
well
known
schemes
already
associated
with
the
negative
integer
could
actually
be
Extended
up
front
to
include
also
the
the
co-op
plus
schemes
already
existing
I.
Don't
know
that
possibly
warrants
also
some
reordering
of
the
actual
negative
integers
value
assigned,
but
that's
a
minor
thing.
So
this
would
be
the
the
first
point
and
it
should
be
relatively
easy
and
uncontroversial.
A
The
second
point
is
that
about
the
the
actual
registry,
as
I
mentioned
before
and
in
turn
it
includes
a
number
of
some
some
points
to
consider
and
to
discuss
as
a
tentative
structure.
Yeah
I
was
thinking
of
something
like
that.
It
should
be
relatively
easy.
Just
the
scheming
question,
with
the
same
name:
it
takes
from
the
main
URL
schemes
registry.
A
Of
course,
the
negative
integer
value
that
we
want
to
associate
with
it
trying
to
be
very
conservative
about
the
the
negative
integers
that
are
encoded
in
one
or
two
bytes
in
cboard,
and
those
should
really
be
protected
and
used
in
the
interest
of
iot
use
cases,
while
we
can
be
more
flexible
for,
for
other
ranges
to
be
used
for
whatever
use
case
and
then,
of
course,
a
reference
to
the
document
that
is
registering
the
scheme
and
or
the
negative
integer,
because,
of
course,
those
already
defined
in
href
would
point
to
the
href
document,
which
defines
only
the
integers
and
and
not
any
new
scheme.
A
But
then
this
would
continue
with
additional
details
on
the
registry.
Like
the
policy,
hopefully
specification
required
to
be
good
enough
and
then,
of
course,
what
should
be
initially
registered
and
how
to
define
policies
for
future
registration.
A
Well,
it's
easy
to
imagine
that
the
negative
integers
already
defined
in
the
ehref
document
would
be
a
part
of
the
initial
population
of
the
registry,
but
other
than
that,
and
here
is
where
I
started,
to
get
additional
input.
Also
from
Carson
the
other
day,
we
may
decide
to
go
through
the
already
existing
URI
schemes
and
select
only
a
few
of
those
that
we
believe
are
expected
to
be
used
in
Cris
and
pre-register
the
negative
integers
for
those
schemes
too.
A
But
then,
after
that,
meaning
following
a
dhrf
document,
what
to
admit
to
be
registered
and
if
I
understood
Carson
suggesting
correctly
yeah.
It
would
be
good
to
give
only
a
single
opportunity
for
a
future
new
scheme
to
be
sorry
to
continue
with
the
registration
of
a
negative
integer
for
the
CRI,
meaning
in
that
exact
document
where
that
new
scheme
is
proposed
and
registered
itself,
and
no
more
after
that.
A
B
Yeah,
so
the
the
reason
we
originally
didn't
see
this
as
an
extensible
list,
was
that
if
you
make
this
list
extensible
and
apply,
it
apply
the
the
ability
to
encode
negative
integers,
two
existing
schemes
during
the
the
lifetime
of
Cris.
B
B
So
from
from
an
interoperability
point
of
view,
you
you
create
a
problem,
and
also
since
senders
will
typically
be
aware
of
this
problem,
you
essentially
force
them
to
continue
to
use
the
text
form
because
that
that
is
by
definition,
interoperable,
even
if
there
is
a
an
integer
form
because
yeah
you
don't
know
whether
you
are
a
consumer.
B
The
consumer
of
this
here
I
is
going
to
know
about
the
registration
and,
and
actually
the
whole
point
about
the
UI
is
that
you
can
formulate
it
without
actually
knowing
who
the
recipient
of
that
thing
is.
So
it's
supposed
to
be
Universal.
That's
why
it's
called
a
UI,
so
that
that
is
a
big
problem.
B
So
if
we
really
want
to
add
the
ability
to
register
new
integers
for
schemes,
this
only
can
be
made
interoperable
by
essentially
registering
the
negative
integer
at
the
buff
of
the
new
URI
scheme,
and
that
is
of
course
difficult
to
to
actually
make
sure
it
gets
done,
because
people
register
new
UI
schemes
all
the
time
and
that
really
would
be
necessary
to
add
the
the
integer
registration.
B
At
the
time,
the
UI
scheme
itself
is
registered
I'm,
not
even
sure
that
this
works
100,
because
the
UI
scheme
may
be
in
use
for
a
while
before.
B
Finally,
it
is
being
registered
because
it's
much
easier
to
avoid
collisions
by
by
using
text
strings.
So,
if
I
have
a
new,
a
full
scheme
that
I
want
to
use
the
the
draft
for
that
may
live
for
two
or
three
years
before
the
foo
actually
is
registered
as
being
registered.
This
is
the
way
things
go
and
you
would
have
to
essentially
do
an
as
early
as
possible
early
allocation
to
make
sure
that
the
negative
integer
is
available
with
that
for
a
CRI
usage.
B
What
we're
really
doing
here
is
we
are
essentially
adding
a
column
in
the
UI
scheme
registry,
but
I'm
not
sure
we
we
can
phrase
it
this
way.
So
saying
this
is
a
separate
registry
may
be
easier
to
do,
but
we
probably
should
be
discussing
this
with
the
UI
scheme.
People
over
at
the
URL
scheme
list
as
well.
B
Yes,
yes,
so
yeah
we
don't
need
a
registry.
B
If
we
don't
want
to
allow
for
future
registrations,
because
then
then
it's
just
the
task
now
we
have
to
do
to
make
sure
we
have
negative
integers
defined
for
all
UI
schemes
where
it
makes
sense.
But
if
we
want
to
allow
future
registration,
then
we
have
to
think
about
this.
This
effect
of
having
implementations
that
know
about
the
the
UI
scheme,
but
don't
know
about
the
negative
integer
for
that.
A
C
Yeah
hi
I
had
actually
two
two
questions,
so
one
is
just
about
the
negative
integers.
So
also
is
there
a
reason
not
to
use
positive
integers
here.
A
C
C
If
you
register
the
the
scheme
at
the
birth,
let's
say
because
yeah
it
could
be
a
the
newer
device,
that's
sending
a
URI
to
someone
who
is
a
bit
older,
so
the
recipient
and
still
cannot
convert
CRI
to
a
URI
in
that
case,
because
it's
using
a
new
new
created
scheme.
Basically,
so
you
still
have
that
same
problem,
but
maybe
it's
not
as
as
bad
in
that
case,
because
the
recipient
any
case
doesn't
know
what
to
do
with
the
scheme.
But
I
think
it's.
C
D
On
Kristen
here
on
the
topic
of
convertibility,
that's
to
some
extent
a
bit
of
a
you
could
call
the
second
class
citizen
in
that
sense
that
everything
about
Cris
is
optimized
for
processing
them
as
they
are
and
Eve,
and
that
comes
at
the
cost
of
conversions
being
being
a
bit
harder
and,
in
this
particular
case,
maybe
not
being
possible
as
well
as
they
as
using.
It
would
be
possible,
so
the
the
and
so
I
from
that
I
think
that
it
would.
A
B
There's
also
the
question
of
comparing
see
our
eyes.
You
can't
really
compare
them
if
you
have.
B
B
So
this
is
something
we
we
also
may
want
to
look
into.
So
there
is
some
some
canonicalness
to
that,
so
that
that's
also
maybe
a
problem.
We
should
be
looking
at.
D
I,
don't
think
the
comparing
problem
are
the
canonic.
The
comparing
problem
is
that
hard,
because
if,
if
you
want
to
determine
identity
of
resources,
you
should
know
anyway,
whether
those
are
normalized
under
the
under
the
scheme,
specific
normalizations,
which
in
general,
which
a
general
tool
cannot
know.
So
it
will
only
be
limited
to
to
syntactic
normalization,
and
if
it
can,
if
it
knows
the
protocol
or
if
it
doesn't
scheme,
it
will
also
know
the
number.
D
Yes,
this
does
introduce
the
small,
a
small
syntax
based
normalization
issue
that
is
usually
not
there,
but
I
think
the
vast
majority
of
identity
comparisons
would
be
by
parties
knowing
also
the
also
the
normalization
rules
of
the
of
the
scheme.
So
what
what
good
is
identity?
Comparison?
If
you,
if
you
can't
be
sure
whether
it
works
anyway.
B
A
B
B
But,
on
the
other
hand,
not
being
able
to
extend
it
also
is,
is
somewhat
unsected
sacrifactory,
because
that
creates
this.
The
second
class
system
for
for
newer
uis
I
mean
it's
not
that
expensive
to
to
usually
use
scheme
names
that
that
are
kind
of
reasonable
and
and
not
very
long,
but
it's
still
a
second
class.
A
Okay-
and
this
is
on
the
second
biggest
part
of
this
whole
thing-
otherwise,
the
The
Proposal
one
trying
to
get
back
to
the
right
slide.
Sorry,
yeah.
This
point
is
said
to
be.
This
should
be
less
complicated
and
controversial.
Do
you
think
this
is
also
fine
to
add.
B
But
certainly
the
one
on
the
slide,
but
we
actually
should
should
do
that
task
of
looking
at
the
scheme
registrations
and
come
up
with
likely
UI
schemes
and
and
put
them
in
here.
A
B
Well,
I
think
I
know
enough,
would
do
text
on
proposal
number
two
proposal.
Number
one
really
requires
this
this
going
through
the
hundreds
of
UI
schemes
that
that
we
have
right
out
there
and
yeah
it's
easy
for
a
single
person
to
miss
something
so
I
think
we
should
maybe
set
up
a
little
Wiki
where
we
collect
these
things
before
we
actually
do
appear.
C
So
I
just
had
an
idea,
so
there
are
some
schemes
that
are
like
a
very
useful
to
be
used
here
and
yeah.
Maybe
other
schemes
that
exist
but
but
are
not
probably
not
useful,
but
you
could
actually
just
take
the
entire
list
of
schemes
and
then
yeah
assign
to
everyone
a
number.
So
let's
say
not
the
one
byte
or
two
byte
number,
but
maybe
a
bigger
one.
C
A
C
D
C
Yeah
I,
don't
think
enough.
Converter
in
a
typical
kind
of
typical
use
case
would
have
to
have
the
whole
list
on
either
the
ones
to
encounter.
But
if
you
want
the
yeah
like
a
generic
converter,
indeed,
it
needs
yeah.
This
database,
basically
yeah.
C
Right
so
I
was
thinking
of
if
you
have
something
like
URI
to
CRI,
a
web
page
service
or
whatever.
That
could
just
carry
the
whole
list.
That's
quite
easy,
because
it's
not
that
much
schemes
that
we
have.
C
B
Ute
button
again
so
I
I
just
put
three
categories
here:
one
for
for
to
ice
cubes
that
where
we
definitely
want
to
use
short
identifier
for
so
that
would
be
the
one
plus
zero
ones.
So
so
the
numbers
minus
one,
two
minus
24
have
have
a
short
representation.
Then
the
next
ones
would
be
the
the
24
to
minus
24,
to
minus
256,
actually
minus
25
to
to
minus
256.,
which
are
still
reasonably
short.
B
And
then
we
have
the
OnePlus
twos,
which
we
have
enough
of
so
we
probably
don't
have
to
dive
into
with
the
OnePlus
Falls.
That
was
just
for
ordering
the
the
importance
or
likelihood
of
coming
up
with
specific
UI
in
in
a
CI
UI
scheme
in
the
CI.
C
Okay,
yeah
got
it
so
it's
the
kind
of
the
category
based
on
the
length
encoding,
yeah.
D
Just
one
note
on
the
side:
the
Bluetooth
stick
has
already
done
this
exercise
once,
although
not
with
limiting
with
not
with
the
distinction
between
the
low
and
the
high
numbers,
but
they
do
have
a
list
of
I
think
every
scheme
that
was
ever
registered
mapped
to
16
bit
something
number.
So,
oh,
we
may
or
may
not
want
to
want
to
start
from
there.
B
The
other
that,
of
course,
will
be
hard
to
enforce
our
registration
requirements
on,
but
if
somebody
has
done
the
work
that
that
is,
of
course,
useful
information,
can
you
dig
up
a
pointer
to
that
document?
We'll.
D
A
Okay
thanks
a
lot,
then
we
can
move
on
to
the
next
item.
We
should
be
well
prep
work
for
the
ism
wider
discussion
later
into
the
iotops
interim
and
it
would
be
about
the
resource
type
Target
attribute
or
possibly
another
Target
attribute
for
the
anima
document.
B
Yeah,
so
we
we
did
this
gymnastics
to
accommodate
Michael,
who
I
think
is
in
a
different
meeting
right
now.
B
C
So
I
was
just
curious
what
what
the
current
status
is
and
thinking
about
using
a
resource
type
definition
for
that.
C
I
I
think
I
read
from
the
males.
There
were
some
doubts
about
whether
that's
the
right
yeah,
basically
the
right
way
to
do
it.
B
Yeah
correct
the
the
doubt
is
whether
what
we
are
registering
there
actually
is
a
resource
type.
But
let's
have
that
discussion
when,
when
Michael
is
actually
able
to
follow,
which
will
be
in
half
an
hour
over
in
the
iot
Ops
meeting.