►
From YouTube: IETF-CORE-20210929-1400
Description
CORE meeting session at IETF
2021/09/29 1400
https://datatracker.ietf.org/meeting//proceedings/
A
A
A
A
A
And
this
is
an
official
ipf
meeting,
so
we
are
recording
now
the
not
well
applies.
Please
get
familiar
with
it
if
you're
not
already,
it's
not
just
about
ipr,
it's
also
about
conduct,
so
please
be
nice
with
each
other,
and
the
agenda
for
today
is
mostly
on
status
and
progress.
Checking
on
a
number
of
working
group
documents
we'll
go
through
cinema
data
ct
now
under
isg
review,
with
a
few
comments
to
address,
and
we
have
a
presentation
from
carson
about
that.
A
A
C
It's
basic
mostly
mostly
responding
to
ben's
points
that
needs
to
be
done
by
me.
Euron
has
already
taken
a
few
steps
there
and
I've
just
come
out
of
a
situation
of
too
much
work,
so
I'll
think
I
can
respond
to.
Those
soon
should
only
need
just
another
update.
There
is
nothing
major
in
there
anymore.
A
Thank
you,
christian.
Okay.
The
next
point
is
about
the
two
core
course:
documents:
young
siberian
seed
and
they're,
also
under
sg
review,
with
some
comments
to
address
it's
mostly
about
a
few
announcements
on
them.
So
michael
richardson
opened
a
number
of
issues
in
the
working
group
repository
to
track,
especially
those
commands
to
address
and
make
it
a
bit
more
systematic
at
the
work
on
addressing
them,
and
then
we
also
took
a
step
forward
to
accelerate
the
work.
A
More
and
after
some
discussion,
we
agreed
to
set
up
a
design
team
to
work
on
those
issues
and
those
comments,
and
we
are
going
to
have
a
first
meeting
for
that.
So
it's
mostly
the
authors
and
michael
for
for
the
first
meeting
and
and
it's
going
to
be
most
likely
recovering
and
the
precise
schedule
would
will
be
probably
nailed
down
tomorrow
during
the
first
meeting.
So
this
should
hopefully
help
advance
in
the
documents
faster.
A
D
Well,
I
still
have
to
actually
read
through
the
all
this
stuff,
so
this
is
kind
of
top
of
my
agenda
now.
So
I
think
tomorrow
I
will
have
in
tomorrow
in
the
design
team
meeting.
I
will
have
an
overview
of
what
we
need
to
do
and,
of
course
we
can
also
in
parallel,
do
simple
stuff
like
like
fix
editorial
comments
and
so
on.
D
D
So
we
have
to
examine
what
the
comments
were,
that
we
got
and
and
whether
we
can
solve
them
by
just
recurring
to
rfc
7951,
which
was
not
did
not
exist
at
the
time
when
we,
when
we
did
this
work
or
whether
we
are
getting
into
seriously
difficult
territory.
And
in
that
case
I
would
be
prepared
to
discuss
this.
Some
more.
E
D
D
D
A
F
D
It's
too
easy
to
switch
off
your
microphone
while
selecting
slides
okay.
So
so
I
made
a
few
slides
based
on
the
isg
comments.
We
got
on
the
cinema
data
ct
document,
so
there
are
lots
of
comments
that
just
can
be
done
by
the
authors.
So
I'm
not
going
to
discuss
these,
but
there
are
a
few
that
maybe
are
not
not
worthy.
D
D
Second
item
ben
pointed
out
that
content
coding's
nest.
I
completely
forgot
about
that,
probably
because
we
have
never
had
a
use
case
where
we
would
actually
use
that
so
the
current
syntax
for
a
content
format
spec
only
allows
one
content
coding
and
worse
the
registry
for
content
format.
Numbers
only
allows
one
content
coding.
So
that's
maybe
a
separate
item.
D
We
can
keep
in
the
backs
of
our
minds
that
content
format
numbers
cannot
cover
that
case,
but
fortunately,
due
due
to
the
previous
slide,
we
we
still
can
talk
about
these
cases
and
yeah.
So
where
we
currently
say
something
like
application,
slash,
jason's
at
deflate
to
provide
a
single
content
coding,
we
could
provide
the
multiple
ones
using
some
syntax.
We
could
come
up
with.
So
we
could
say
something
like
add:
deflate,
comma
as128,
gcm
or
we
could
use
the
add
sign
again.
D
D
The
the
example
I
used
here
is
one
where
we
have
one
content
coding
that
actually
does
compression
the
deflate
content
coding,
which
is
widely
used
and
another
content
coding
which
actually
does
encryption.
Now.
This
is
a
weird
thing.
We
are
very
unlikely
to
ever
see
this
in
a
cinema
file,
but
it's
the
only
reasonable
example
for
more
than
one
content
coding.
I
could
come
up
with
based
on
the
content
codings
that
I
currently
registered.
So
I
expect
to
use
that
example
in
in
the
draft
as
well.
D
So
that's
one
way
of
doing
it
and
unless
people
think
the
the
common
stuff
is
better
than
the
double
add
sign,
I
think
we
should
go
there.
The
the
other
approach,
of
course,
would
be
to
say
we
don't
do
nested
content
codings
in
co-op,
obviously,
because
there's
only
one
item
for
for
content
coding
in
the
registry,
so
we
don't
need
to
do
it
in
cinema
either
and
we
we
won't
handle
nested
chronic
coatings
at
all.
G
D
D
D
C
Other
than
us
not
having
any
use
cases
or
possibly
nobody
having
any
is
there
anything
that
would
stop
us
from
redefining
the
co-op
content
format
registry
to
say
that
that
can
be
a
list
and
everyone
currently
so
far
only
input,
nothing
or
a
single
item
in
there.
So
it's
a
single
item
list
for
those.
D
Okay,
so
with
that,
let
me
move
to
the
next
item
so
ben.
Can
I
caught
us
trying
to
to
define
abnf
rules
in
a
way
that
actually
makes
sense
and
and
is
not
sucking
up
all
the
legacy
of
the
last
30
years?
And
of
course
the
question
is:
can
we
do
that?
So?
Is
there
maybe
something
magic
about
abn,
f
rule
names
that
makes
people
assume
that
using
the
same
rule,
name
in
different
documents
still
has
the
same
semantics,
and
I
think
mostly,
I
do.
D
Nobody
expects
them
to
be
the
same.
So
if
you
have
a
production
token
in
your
abn
f,
you
expect
that
to
be
different
from
any
other
token,
you
have
in
any
other
rfc,
and
actually
I
also
noticed
that
that
when
he
pointed
out
there
is
a
new
version
of
http
coming
out
a
new
revision.
Excuse
me
not
a
new
version.
D
I
have
no
idea
why
it's
mentioned
in
the
changes,
so
it
obviously
was
was
a
deliberate
change,
but
I
would
expect
that's
another
one
of
those
legacy
things
because
there's
no
point
in
doing
that,
so
yeah,
the
the
the
http
people,
people
are
doing
it
themselves,
so
why?
Why
wouldn't
we
be
doing
it
and
the?
The
third
observation
is
that
http
uses
an
extended
form
of
abnf
with
the
the
pound,
sign
or
hash
sign,
meaning
something
different,
and
we
wouldn't
use
that
syntax
in
in
any
case,
because
we
would
use
basic
abnf.
D
D
That
you
see
in
the
copy
from
70
to
30,
the
o
is
obsolete
because
it's
the
obsolete
form
of
white
space
and
we
don't
need
obsolete
white
space.
So
in
our
syntax
there
is
a
star
sp
there
yeah.
So
I
would
propose
to
to
handle
this
as
an
editorial
request
to
explain
what
we
have
been
doing,
but
not
to
actually
change
anything
in
how
we're
doing
it.
G
So
I
think,
like
that,
one
challenge
here
that
cinnamon
can
be
used
with
a
variety
of
different
transfer
protocols.
It
could
be
hp
could
become
ppmqtt,
it
could
be
a
file,
so
we
don't
have.
Of
course
you
cannot
say
this.
Is
the
error
code
to
use,
and
maybe
this
document
would
not
be
the
right
place
to
go
on
on
this
level,
but
I
don't
think
there
might
be
use
for
another
document
that
gives
in
in
general
the
explanation
on
this
kind
of
errors.
This
appropriate
error
code
for
co-op
or
http.
D
G
G
True
true,
I
was
wondering
if
this
is
the
right
document
or
if
there
would
be
a
different
document,
I
mean
overall,
covering
all
sorts
of
current
and
future
sentiment
extensions
and
giving
the
guidance
guidance
there,
which
might
be
more.
You
could
actually
take
all
my
more
details
around
this
that
might
be
possible
way
forward,
but
yeah
no
strong
opinion
here.
D
Yeah,
so
to
me
this
is
a
reasonable
question
to
raise
in
a
review.
So
what?
What?
What's
the
plan
here
and
maybe
just
adding
a
sentence
or
two
that
nothing
in
this
document
changes?
The
overall
error
handling
model
in
centimeters
might
be
a
good
thing,
but
I
don't
think
we
have
any
any
idea
about
specific
error
handling.
I
mean
we
are
not
telling
an
implementer
of
units
that
they
have
to
pull
down
the
registry
every
time
they
encounter
unsupported
unit
or
something
like
that.
It's
just
something
that
that
is
kind
of
implied.
D
D
So,
finally,
as
I
said,
I'm
not
going
to
bother
you
with
the
editorial
stuff.
There
will
be
some
some
more
editorial
stuff
in
in
the
repository
and
I
plan
to
to
have
another
interaction
with
ben
kadak,
because
his
discuss
probably
is
the
most
complicated
one.
And
if
that
goes
well,
I
plan
to
submit
a
new
version.
So.
C
Yeah
I'd
kind
of
give
say
a
few
lines
on
about
where
we
are
and
then
I'll
start
screen
sharing,
because
I
think
this
will.
This
will
help
a
bit
more
sure.
So,
in
general,
in
the
in
the
design
team
meetings
for
coral,
we
set
up
to
focus
on
a
few
on
a
few
kind
of
focus
points,
because
otherwise
there
is
too
many
variables
to
to
keep
in
mind.
C
C
So
this
is
nodes
representing
your
eyes
and
edges,
representing
statements
that
are
made
about
the
the
thing
on
the
start
of
the
statement
thing
of
the
end
of
the
statement
pretty
much
like
web
links,
but
with
the
difference
that
statements
are
not
really
depending
on
each
other.
That
is,
if
you
have
a
statement
that
is
slot
one
can
receive
an
order.
C
By
doing
this
thing,
then
this
is
true
independent
of
whether
you
come
to
slot
one
from
these
coffee
machines
has
a
coffee
slot
or
from
some
other
interaction,
now
other
interactions
for
other
interactions.
That
thing
might
not
be
that
useful,
but
it's
still
true,
so
the
model
is
described
as
having
statements
with
subjects
with
predicates
or
objects
or
in
in
link
contact
link
terminology.
That
would
be
a
context
and
linked
target
under
relation.
C
One
is
to
have
say,
immutable
atomic
literals
that
just
have
a
few
properties
that
are
probably
serialized
as
child
nodes,
but
we
don't
talk
about
how
that
happens
in
the
information
model.
That's
this
approach
here.
Kind
of
everything
is
in
the
literature
and
if
two
literals
have
the
same
complex
value,
that
is,
they
have
the
same
literal
value
and
have
the
same
language
tag
and
they
have
the
same
left
to
right
indication
and
they
have
the
same
xsd
type,
yada
yada.
C
C
That
is
not
necessarily
identical
to
any
other
occurrence
of
kaffir
bashir,
but
it
can
have
be
a
subject
of
further
statements,
so
there
can
be
then
a
statement
that
says
this
term
that
we
are
talking
about
here.
This
particular
cafe,
bachelor
height
home,
has
the
language
german
and
has
writing
direction
left
to
right,
other
properties,
etc.
C
I
think
this
is
mainly
a
matter
of
phrasing
it
because
for
practical
use
there
will
need
to
be
restrictions
in
place
anyway,
to
avoid
that
there
we
are
too
much.
There
is
too
much
fan
out
down
below
this,
so
you
can
still
should
talk
about
the
german
language
from
here,
but
things
are
getting
pointless.
C
So
if
there
are
any
strong
opinions,
I'm
I'm
happy
to
hear
those
if
there
are
not
so
strong
opinions
just
experience
anything
else,
two,
but
either
of
those
it
will
probably
be
one
more
thing.
I'd
like
to
point
out
before
we
go
to
the
structured
information
model
is
that
there
are
still
forms
in
there,
but
on
the
for
the.
As
far
as
the
information
model
is
concerned,
they
are
just
statements,
so
there
can
be
a
few
optimizations
in
how
they
are
serialized.
C
There
can
be
guidance
on
how
they
are
used,
but
for
the
information
model
is
just
this.
Slot
has
a
thing
like
the
like
the
literals.
Above
that's
a
thing.
That
is
only
this
particular
thing.
It
doesn't
necessarily
need
to
have
a
uri
of
its
own,
but
it
has
a
submit
uri.
You
can
submit
there
using
that
method
post
with
this
particular
content
format,
and
these
and
those
are
those
are
the
fields
you
need
to
fill
in.
C
C
But
I
don't
think
that's
really
so
surprising,
because
whenever
someone
receives
information
about
a
thing
that
that
they
could
interact
with,
they
will
follow
that
information
and
make
sure
they
have.
There
has
to
be
a
trust
model,
whether
they
really
should
do
it
that
way.
But
in
the
end,
if,
if
the
permission
is
checked
out,
they
will
try
to
make
coffee
with
you.
C
C
A
new
addition
here
is
now
the
structured
serial
structured
information
model.
This
is
basically
a
tour
and
I
still
don't
have
a
precise
term
that
describes
the
swelling
graph
theoretical
terms,
a
tour
through
the
nodes
that
are
supposed
to
be
expressed
in
a
particular
order.
So
you
can
say
that
this
is
kind
of
the
first
slot
in
the
second
slot,
and
it
also
contains
information
about
where
you
go
down
next.
C
So,
for
example,
with
this
in
this
in
this
possible
serialization,
you
go
down
the
you
explain
what
an
app,
what
an
app
slot
is,
the
first
time
you
encounter
it.
So
you
say
here
we
have
a
coffee
slot
slot
one
you
can
do
orders
like
that,
and
this
is
of
type
coffee
slot,
which
we
call
cafe
heidelbergdo,
and
then
there
is
this
other
coffee
slot,
which
is
of
type
app
slot
and
hopefully,
by
the
way.
You've.
C
Now
remembered
it
and
we
call
it,
we
call
this
particular
slot
that,
but
there
can
be
different
civilizations,
so
you
can
express
the
same
information.
The
second
time
you
pass
over
it
or
you
can
express
it
both
times
and
different
applications
or
different
constrained
applications
might
have
requirements
on
how
those
are
processed,
how
where
those
are
present.
So,
for
example,
some
application
might
not
be
able
to
use
that
information
when
it
is
just
processing
its
data
about
this
slot.
C
C
I
have
a
hunch
that
much
of
this
can
go
away
anyway,
if
we
have
profiling
nailed
down
precisely,
but
at
least
the
part
about
the
ordering
will
not,
and
so
at
least
some
of
this
should
probably
stay
in
there,
and
this
summarizing
is
what
I'd
like
to
or
what
we.
What
we
like
to
evaluate
other
models
say
what
asdf
needs,
what
we
have
in
the
existing
applications?
What
web
links?
Would
you
look
like
what
rdf
would
look
like
when
using
it
with
this
with
coral?
A
A
C
C
Notation
that
I
should
have
mentioned
before
on
kind
of
the
different
fronts
that
we're
exploring
this
together
with
using
packed
hebrew,
the
the
extensions
to
diagnostic
notation,
that
custom
presented
last
time
in
in
sieber
what
we
would
use
for
the
time
being
and
probably
going
forward.
But
this
this.
This
should.
This
should
be
in
the
in
the
next
examples
that
we
produce.
A
A
C
H
H
So
latest
development
we
have
been
active
in
in
spurts
in
bursts,
and
so
we
have
now
managed
to
at
least
clarify
quite
a
few
things,
but
still
there
is
always
scope
for
new
feedback
and
and
review
comments.
So
please
please
go
ahead
and
do
that
and
in
our
github
pages
we
have
three
issues
remaining:
two
technical
ones
and
one
editorial
and
they're
all
under
active
discussion.
So
let's
I
think
that
it's
more
important
to
discuss
the
technical
issues.
H
The
editorial
issues
can
be
done
quite
easily,
but
we
just
wanted
to
discuss
a
couple
of
things.
So.
Firstly,
this
is
a
technical
issue
which,
which
came
up
when
we
were
discussing
the
band
attribute
now
just
a
background
band
modifies
the
behavior
of
gt,
which
is
greater
than
or
lt,
which
is
l
less
than
which
indicate
the
limits
where
the
the
notifications
should
arrive
so
lt
and
gt.
They
indicate
the
limits
where
the
value
of
the
resource
should
cross
before
triggering
a
notification,
and
this
is
a
notification
one
and
only
one
notification.
H
Right
so
and
and
the
behavior
you
can
see
below
in
the
slide,
so,
for
example,
if
the
server
has
a
resource
and
the
state
changes
between
the
stages
from
zero
to
one
two,
three,
four,
five,
six
and
so
on-
and
this
is
the
base
state
and
that
the
client
sends
a
conditional
observe
with
gt
equals
six
as
the
parameter.
So
it's
a
query
parameter
so
the
moment
the
resource
value
switches
from
four
to
ten.
H
H
Now,
if
you
want
to
have
a
band
in
which
you
want
to
get
notifications
for
a
specific
group
of
resource
states,
you
can
set
a
ban,
for
example
like
this,
so
if
it
is
greater
than
2
and
less
than
12
and
your
band
is
enabled
so
then,
if
you
actually
get,
if
you
put
the
the
query
in
then
you
get
notifications
within
the
band.
So
you
get.
You
get
notifications
for
three.
H
Four
and
ten
and
eleven,
but
of
course,
if
it
is
two
and
twelve,
then
you
don't
get
notifications
and
that
proceeds
until
it
drops
and
if
it
drops
below
this
this
limit,
you
don't
get
on
the
questions
either.
So
then
again,
you
know
if,
if
the
resource
states
start
changing
between
this
part,
you
you
get
notifications
again
and
that's
basically
how
band
works
now.
H
H
It's
qualifies
as
a
as
a
bank
command,
so
there
have
been
comments
on
github,
so,
firstly,
the
first
command
is
is
whether
the
ban
attribute
is
defined
as
a
boolean
type,
but
its
usage
is
determined
only
by
its
presence
and
its
value.
So
we
should
probably
explicitly
explain
that
which
basically
means
that
you
accept
both
band
equals.
One
and
band
equals
zero
setting
the
notification
band.
H
C
H
H
The
alternative
here
is
not
having
band
at
all
and
then
in.
In
other
words,
you
just
allow
gt
and
lt
trigger
all
notifications,
so
in
other
words,
we
modify
the
text
in
such
a
way
or
the
behavior
of
gdnode
saturated
each
time.
The
resource
states
change
after
crossing
the
threshold.
You
keep
sending
notifications
instead
of
just
one
now.
H
F
F
F
Maybe
there's
a
problem
with
that,
but
I'm
just
going
to
stick
to
the
one
comment
and
and
if
we
get
to
the
other
subject,
I'll
I'll
join
the
queue
again.
But
my
comment
was
yeah
that
if
there
was
some
way
to
have
it
be
able
to
be
expressed
as
a
boolean
in
a
different
sort
of
model
other
than
query
parameter.
That
would
be
a
big
that
would
be
good.
F
D
D
F
F
F
I
don't
know
if
we
really
had
a
way
to
do
that,
provide
a
way
to
turn
it
on
and
off
and
go
from
one
behavior
to
the
other.
I
don't
recall
that
even
being
really
part
of
the
use
case
that
it
seems
like
once
I
set
up
notifications,
I'm
going
to
either
have
it
be
one
mode
or
the
other,
but
not
sure
about
going
back
and
forth
and
then
with
multiple
bands.
That's
an
interesting
question.
Theoretically,
you
could
have
multiple
bands,
but
you
would
need
labels
other
than
gt
and
lt.
F
F
Notifiers,
okay,
so
if
you
have
a
resource
that
has
two
notifiers
and
you
have
a
a
band
and
one
and
a
different
band
in
the
other,
then
they
could
each
notify
when
their
conditions
are
met.
But
I
don't
think
that's
exactly
what
we're
describing,
but
that
that
may
be.
Maybe
what
we
were
thinking
also.
H
Yeah
yeah,
so
so
there
were
that
it
was
yeah
to
notify
us,
but
but
the
way
it
was
written
was
that
there
are
multiple
bands
set.
So
it's
it's
kind
of
confusing
how
how
we
can
actually
do
that
by
not
giving
any
like
further
guidance
on
on
whether
it's
even
possible.
There
are
multiple
bands
and
something
that
we
really
need
to
consider
carefully
before
we.
We
include
that
text
in
there.
F
H
Yeah,
I
I
it's
actually
already
in
the
in
the
github
issue.
So
so
please
just
have
a
look.
It's
probably
the
editorial
github,
if
you
actually
so
that's
it's
there,
but
let's,
let's
take
that
offline
and
decide
whether
we
keep
it
inside
or
not
christian
you're.
On
the
cue
yes,
could
you
go
back?
One
slide,
please
yeah
sure.
C
So
I
think
I
can
phrase
this
now
more
precisely
than
I
did
in
chat
before
what
I'm
seeing
here,
that
the
band
does
not
trigger
when
there
is
a
jump
through
the
through
the
values,
is
a
bit
inconsistent
with
the
explanation
that
we
had,
at
least
in
a
few
discussions
back.
I've
not
been
following
too
closely
of
a
view
of
that
resource.
So
if
we
have
this
bad
scope,
so
this
this,
this
band
attribute
was
greater
than
less
than
looks.
C
To
me
like
this
is
a
view
of
the
resource
that
says:
take
the
value
whatever
it
currently
is
and
clip
it
low
and
high
to
to
the
to
say,
2
and
12..
So
whenever
you
receive
whenever
it
goes
higher,
don't
tell
me,
because
I
don't
care
anyway
and
whenever
it
goes
lower,
don't
either,
but
ignoring.
This
jump
in
the
middle
here
violates
this.
This
model,
because
now
now
kind
of
having
or
not
having
a
value
in
between
makes
things
look
different.
C
So
what
I
would
have
expected
here
on
leave,
leaving
out
any
less
than
or
less
an
equal
thing
is
that
as
soon
as
it
goes
to
the
lower
threshold,
it
would
at
least
report
either
the
lower
threshold
or
the
lower
threshold
two
or
the
value
one.
But
at
least
tell
me
because
what
I
would
implement
with
this
is
say
some
kind
of
regulation.
C
H
It's
so
so
band
is
used
in
two
ways:
it's
it's
used
for
what
we
call
in-band
in-band
notifications,
which
is
this
basic
behavior
and
you
can.
You
can
swap
the
values
around
so
you
can
have.
I
also
have
band
signaling
out
of
band
behavior,
so
that's
basically,
when
less
than,
for
example,
less
than
less
than
two
and
greater
than.
B
H
So
so
that's
that's
the
way
it
has
been
specified
now
I
I
don't
have
an
opinion
on
on
whether
we
should
actually
start
to
give
notifications
when
the
threshold
is
crossed.
The
other
way
when,
when
it
switches
from
example,
from
from
sorry
from
14
to
1.,
I
do
believe,
there's
a
there's,
a
use
case
for
this,
whether
it's
interesting
for
the
the
client
to
actually
know
that
the
threshold
has
been
crossed
the
other
way.
I
don't
really,
I
can't
really
say.
C
C
That's
not
like,
and
it's
not
like.
They
can't
kind
of
you
as
long
as
you're
in
there
you'll
receive
you.
The
server
may
send
notifications,
so
you
send
a
notification
that
something
is
crossing
through
and
the
client
can
still
ignore
that
notification,
but
not
getting.
It
is
harder
harder
to
ignore.
F
Michael
yeah,
I
think
I
see
what
you're
saying
and
that,
even
if
we
wanted
to
say
we're
using
band
to
define
high
and
low
error
limits
and
I'm
getting
notifications
when
I'm
high
and
when
I'm
low.
If
I
stop
getting
notifications,
I
don't
know
if
that's
because
of
some
setting
of
you
know,
pmin,
pmax
or
or
that
or
whether
I've
you
know
returned
to
the
normal
band
or
not.
So
we
want
to
make
we
want.
F
We
probably
do
want
to
get
a
notification
when
we
cross
back
into
the
normal
band
or
when
we
go
back
into
that
band.
So
it's
more
like
you
could
think
of
it
as
limit
crossing
as
in
the
other
mode,
and
you
can
also
think
of
it
as
state
change
when
you're
going
from
an
out
of
band
to
in-band
or
in-band
to
out-of-band,
and
I
need
to
look
at
the
equations
again
to
see
if
that
happens,
I
suspect
that
maybe
it
doesn't,
and
we
need
to
think
about
that.
H
F
H
So
this
is
a
the
second
issue
and
it's
about
an
ongoing
discussion.
We've
always
had
about
the
presence
of
proxies
and
middle
boxes.
That
could
be
the
path
between
the
client
and
the
server
and
in
the
discussions
that
we
have
had.
If
you
look
at
the
the
lower
part
of
the
slide,
it
kind
of
captures.
What
is
what
is
the
problem,
and-
and
this
doesn't
happen
to
every
conditional
attribute-
it
happens,
for
example,
with
pmax
now
what's
happening.
H
The
proxy
may
choose
not
to
return
the
the
value
of
the
resource
to
the
client.
If
the
value
does
not
change,
so
we
have
been
trying
to
figure
out
ways
in
which
we
can
handle
that
and-
and
it
I
think,
from
the
perspective
of
conditional
attributes,
it's
it's
possibly
kind
of
ambitious
to
try
to
find
a
workaround
behavior.
But
I
think
what
we
can
do
is
that
we
have
agreed
that
we'll
put
some
some
text
in
the
implementation
consideration.
H
So
I
I
kind
of
propose
a
text
to
say
that
this
specification
defines
conditional
attributes
it
can
be
used
with
call
observe.
It's
recognized
that
the
presence
of
one
or
more
proxies
between
the
client,
the
server
can
interfere
with
clients
receiving
resource
updates
and
therefore
the
server
may
use.
For
example,
the
max
h
option
to
mitigate
this
by
setting
max
h
to
be
less
than
or
equal
to
p
max.
H
F
It's
that
another
way
to
think
about
this
is
that
t-max
is
only
guaranteed
to
work
on
a
sort
of
a
link
basis,
not
so
much
link
in
terms
of
a
802
154
mesh
but
link
in
terms
of
endpoints
of
the
protocol.
So
if
you
wanted
the
behavior
end
to
end,
you
would
have
to
implement
pmax
in
the
proxies.
I
guess
is
the
side
effect
of
that,
but
I
think
yeah.
H
A
H
B
C
D
E
A
A
H
A
You
can
find
all
the
links
in
the
minutes
of
this
meeting
too,
and
the
only
official
feedback
I
could
track
in
the
last
few
weeks
was
a
mail
to
the
mailing
list
from
esco,
basically
happy
enough
with
the
current
text
as
a
good,
updated
overview
of
the
activities
of
the
working
group,
just
pointing
out
that
if
the
text
becomes
too
detailed,
there's
of
course
a
risk
to
increase
the
chance
to
update
the
text
again
in
the
future.
So
of
course
it's
something
to
keep
in
mind.
A
D
A
A
Thanks
for
bringing
this
up-
and
I
know
you
wanted
to
do
a
a
comparison
of
all
the
new
texts,
so
we
are
looking
forward
to
it
and
in
general
we
solicit
more
comment
and
feedback
on
the
list
or
directly
in
the
coding
d.
I
think
francesca
will
also
wanted
to
check
with
other
ads
for
possible
suggestions
on
how
to
proceed
in
general.
Even
anything
from
that
side
already.
B
There
was
not
much
reaction;
this
is
it.
As
I
said
last
time,
this
would
be
a
work
to
to
to
simplify
core
getting
the
work
done
and
not
getting
stuck
in
asg
evaluation
over
and
over
again
for
the
same
reason,
which
is:
is
this
really
in
charter
every
document
coming?
Is
this
really
in
charge
and
having
that
discussion
all
over
again,
but
of
course
we
don't
want
to
update
the
charter
if
that
doesn't
help
getting
things
done
and
and
makes
things
harder
instead
of
simpler.
B
D
D
We
will
do
maintenance
on
eggs
instead
of
defining
what
exactly
that
maintenance
will
be
now,
of
course,
if,
if
there
is
somebody
like
stewart
who
wants
to
have
the
word
dnssd
in
the
charter,
that
may
not
work
for
that
part,
but
maybe
we
should
do
do
a
critical
read
where
maybe
the
the
work
that
we
have
already
have
done
is
the
best
description
of
of
the
kind
of
work
we
were
doing.
Its
maintenance.
D
I
mean
the
charter
could
say
this
working
group
does.
What
mark
thinks
is
the
right
thing
to
do
and
it
would
probably
still
pass
the
isg,
but
this
is
actually
missing
one
other
point
of
a
charter,
which
is
that
it
can
be
used
by
the
working
group
chairs
to
beat
up
the
working
group
members
when
when
they
try
to
do
something
weird
that
is
outside
the
scope
of
the
working
group,
so
that
that
is
actually
a
function
that
is
needed.
B
I,
of
course
there
needs
to
be
consensus
on
the
rechartering
text
and
and
the
fact
that
we
even
need
rechartering
and
thank
you,
marco
and
whoever
else.
I
don't
know
if
I
may
have
worked
on
this,
but
thank
you
for
starting
this
work.
I
think
if
we
can
get
it
done
and
get
consensus,
and
you
know
it
would
be
helpful,
but
if
we
don't,
then
we
should
not
try
to
yeah.
We
should
do
what's
best,
so
any.
B
D
Which
is
completely
changing
the
way
in
which
I
read
the
same
words
right
right
right,
that's
a
good
thing,
yeah,
so
yeah.
I
think
one
has
to
be
really
careful.
I
completely
agree
that
we
need
to
do
something
about
collab
over
bluetooth,
but
yeah.
The
the
the
way
we
phrase
this
here,
I
think
is,
is
really
important
and
I
think
we
have
a
much
better
understanding
how
to
handle
the
security
issues
than
we
had
10
years
ago.