►
From YouTube: IETF-TAPS-20220202-1600
Description
TAPS meeting session at IETF
2022/02/02 1600
https://datatracker.ietf.org/meeting//proceedings/
A
A
A
D
A
A
A
A
A
B
F
One
I
had
opened
brian
approved
and
michael
you
had
a
comment.
Shall
we
discuss
this
one?
Essentially,
this
was,
I
believe,
what
we
had
agreed
at
a
previous
meeting
to
say
you
know
we're
not
going
to
mention
framers
before
we've,
even
defined
them
as
part
of
message
context
in
this
vague
scope
thing
instead,
we're
just
going
to
essentially
have
that
be
in
a
lighted,
parameter
that
there's
a
variant
later.
That
can
take
a
framer.
D
F
E
F
Know
some
languages
do
allow
essentially
saying
that
if
you
have
a
different
number
of
parameters,
it
is
a
different
function
and
that
they
can
both
be
called
add
and
both
be
called
get,
even
if
they
take
different
numbers
of
things
and
some
don't
allow
that
yeah.
So
I
I
don't
know,
what's
the
rule
for
our
made-up
example,
language
here.
C
C
I
think
for
this
one
we
really
have
to
make
our
mind
about
what's
better,
for
that
either
have
so
we
have
a
scope
for
param
for
the
whole
pokey
system
already,
and
we
have
defined
this
scope
to
be
able
to
accommodate
protocol
specific
properties
and
implementation,
specific
properties
and
now
for
the
framer.
It's
the
question
whether
we
we
really
add
the
scope
parameter
here
or
we
say.
C
Oh
yes,
it's
a
special
case
of
of
the
scope
we
already
have
up
and
up
in
the
definition
of
properties
and
and
it's
it's
an
implementation,
specific
properties.
That's
just
prefixed
with
the
name
of
the
framer,
so
you
you
could
take
it
that
way
or
you
could
could
take
it.
The
original
way
with
the
scope
parameter
here
and
explicitly
name
the
framers.
So
I
have
no,
I
guess
what's
better,
but
what's
more
consistent.
F
C
F
F
F
C
So
the
question
is:
do
you
really
have
to
to
solve
the
name
clash
problem
here?
So
if
you
give
give
your
so
first,
it
could
be
that
your
implement.
So
let's
assume
you
have
a
framer
that
implements
a
protocol
and
more
or
less
ex
exposes
a
protocol
specific
property.
That
way,
so
you
implement
a
framer
yeah.
C
C
A
F
A
F
And
not
append
yeah,
so
I'm
going
to
just
tag
you
here
to
say:
can
you
propose
text
for
how
we
can
prepend
the
name
of
the
framer
as
a
namespace
for
the
properties
we
had
yet,
and
I
think
this
upper
part
could
actually
remain
the
same
and
it's
really
just
in
the
framework
part.
We
say:
hey,
there's
a
way
to
name
space
properties
for
framers.
E
F
F
Okay,
we
had
this
pr
from
sean
and
I
I've
merged
a
bunch
of
sean's
pr's,
which
is
great
this
one
I
was
not
as
sure
on
this
is
just
a
typo,
but
let's,
let's
take
a
look
at
the
change
here,
so
it
used
to
say
at
least
one
local
endpoint
must
be
specified
for
listen
now
it
says,
should,
and
at
least
one
remote
must
be
specified
for
initiated
outbound
connections
and
now
says.
Should
I
think
I
just
disagree
with
that.
Joan
do
you
want
to
help
us
discuss.
H
A
F
I
mean
for
listening
yeah
because
you
could
say
I
don't
care
about
my
address.
I
don't
care
about
my
port
and
just
assign
one
to
me,
so
I
think
you
could.
That
could
become
a
should,
but
I
think
we
definitely
need
us.
We
still
need
to
keep
a
must
for
remote,
like
you
need
to
say
what
you're
talking
to.
C
I
was
stressed
here
I
was
so
for
the
first
one.
I
definitely
should
for
the
second
one.
If
you
have
something
strange,
unlike
ospf,
where
you
know
you
can
only
talk
to
a
specific
multicast
address,
you
might
be
able
to
skip
that
and
just
derive
this
from
from
the
other
parameters
on
the
local
endpoint,
but
that
that
sounds
very.
J
D
Settles
it
and
it's
just
a
way
of
phrasing,
this
in
text.
So
I
think
we
all
agree
that
well
about
this
condition
about
if
listener
must
have
local
may
have
remote
and
vice
versa,
and
based
on
what
sean
just
said,
I
think
we
should
leave
the
sentence
as
it
is
because
nobody
actually
proposed
a
semantic
change
to
the
mustang
and
shoot
okay.
So
just.
L
I
I
I
F
F
Michael,
you
commented
here
right,
so
this
is
in
the
implementation
draft
talking
about
framers.
We
had
an
issue
just
to
mention
that
you
must.
The
implementation
must
make
sure
that
calls
on
a
framework
are
synchronized,
so
that
operations
like
adding
a
framer
or
passing
through
data
is
synchronized
with
actually
sending
and
parsing
data
and
that
those
can't
overlap
in
an
inconsistent
manner.
F
F
J
J
That
was
added,
I
mean
that
seemed
to
kind
of
say
what
brian
was
complaining
about,
that
it
was
the
same
object
that
changed
from
that
is
expressed
both
as
mutable
and
immutable.
Actually,
I'm
not
sure
what
the
problem
was
from
the
beginning,
because
the
text
actually
talks
about
two
different
objects.
I
think,
but
I
think
the
new
text
is
clearer
so,
except
for
this
last
part
that
I
didn't
understand.
J
B
D
F
Essentially,
explaining,
I
think
this
is
in
response
to
one
of
martin's
issues
that
the
way
that
an
application
expresses
preference
or
one
of
the
ways
to
express
a
preference
for
what
address
to
choose
when
raising
different
addresses
is
via
selection
properties,
and
it
gives
the
example
of
using
temporary
privacy
addresses,
and
if
you
prefer
that,
then
you
end
up
preferring
v6.
If
you
require
it,
you
require
v6,
etc
and
just
kind
of
explaining
how
that
can
influence
things.
E
F
H
H
M
H
D
I
E
F
H
F
B
F
F
L
L
F
L
L
I
L
B
H
F
F
B
F
H
Looking
at
the
security
parameters
and
the
things
that
are
set-
and
I
was
like
yeah
most
of
it's
there,
because
you're
setting
identities
and
private
keys
and
they're
flowing
back
and
forth,
but
what
I
didn't
notice
was
if
you
were
going
to
use
pre-shared
keys-
and
I
don't
know
if
you
need
to
specifically
call
that
out
here
or
if
that's
already
supported,
and
so
that's
really
was
like.
Should
we
be
pointing
there
or
not.
H
A
C
M
H
D
F
D
Yeah,
so
the
issue
is
that
is
that
you
can
create
a
message,
context
and
adjust
stuff
in
it
at
any
point,
and
it
doesn't
have
anything
to
do
with
connection
or
selection
properties
and
then
you're
expecting
that
when
you
then
later
use
it,
it
should
inherit
like
if
you
don't
do
anything
about
it.
Actually,
if
you
just
create
it,
but
you
don't
change
it,
then
it
should
maybe
get
a
value
from
a
connection
property
inherited
from
that
right.
So,
for
instance,
reliability.
D
If
you
don't
like,
you,
create
the
message
context,
but
the
context
would
have
an
object
for
deciding
about
messenger
reliability
and
the
default
would
be
reliable
when
you
use
it
for
a
connection.
But
the
issue
is
that
the
object
itself
is
not
connected
to
a
connection
somehow,
so
I
can
create
it
at
any
time.
Long
before
I
even
make
a
pre-connection
right,
and
it
should
somehow
be
made
clear
that
it
inherits
the
property
of
the
connection
that
it's
used
on
when
a
message
is
sent
or
you're
using
this
message
context.
D
J
C
C
This
is
how
I
think,
when
writing
this,
we
anticipated
it
and
bento
sorted.
No,
no,
no
a
message
context
should
have
all
the
states
of
all
the
properties
in
the
way.
The
cent
will
do
it
and
if
you
want
it
that
way,
you
would
really
need
to
con
to
create
the
message
context
from
the
connection
and
then
make
sure
that
you
can
use
it
for
introspection
and
other
stuff
to
make
sure
what
would
happen
to
that
sand.
D
F
D
Now
the
next
part
of
this
is
that
my
proposal
on
how
to
do
this
here
was
to
say
that
we
can
just
say
that
they
all
have
a
default
and
the
default
is
ignore
and
ignore
means
that
it's
going
to
be
overwritten
by
the
connection
behavior
when
you
use
it
for
the
connection.
So
unless
you
set
it
to
something
specifically,
it's
going
to
have
that
behavior.
C
D
B
D
A
D
I
mean
that
is
kind
of
natural.
I
mean
you
choose
something
you
say
I
would
like
it
or
not,
and
if
you
get
it,
then
it
becomes
a
boolean
situation
right.
But
the
proposal
here
for
using
a
selection
property
is
kind
of
just
to
misuse.
That
type,
because
that
type
exists
and
say
we
basically
redefine
the
values
to
mean
true
and
false,
which
I
think
is
pretty
ugly,
because
that
means
semantically.
C
D
It
becomes
awkward,
doesn't
it
because
you,
you,
then
have
message
context
like
in
principle.
Somebody
can
create
so
many
message
contexts
and
put
the
put
a
few
preferences
in
there
and
then
also
create
connections
and
say
I
have
this
in
this
preference
for
this
connection,
and
I
want
to
use
this
message
context
with
these
preferences.
On
I
mean
it
becomes
just
a
strange
mix
of
things.
F
F
If
it's
possible,
then
it
overrides
it.
I
I
think,
maybe
to
some
degree
the
only
thing
we're
missing,
and
rather
than
trying
to
have
some
default
value
in
a
message,
context
already
being
filled
out.
Can't
we
just
add
a
paragraph
that
says
if
the
application
does
not
add
a
specific
message,
property
to
a
message
that
all
properties
from
the
connection
are
inherited
and
just
say
like
there
is
no
default
value.
There
is
no
inherit
value,
you
just
didn't
add
it.
If
you
didn't
add
it,
then
it's
not.
There.
D
B
D
G
L
D
A
F
Feel
free
to
make
a
pr,
I
think,
adding
something
to
all
of
the
properties
or
whatever
makes
sense.
It
may
be
nice
to
just
have
some
text
up
above
because
that's
the
area
that
the
person
who
filed
the
issue
got
confused
in
yeah
and
so
just
explicitly
saying
by
the
way
when
you
look
below,
it
explains
how
they
inherit
and
what
their
defaults
are.
E
F
A
A
F
F
F
F
L
F
N
F
A
A
N
B
M
A
E
A
N
I
mean
like
you
can,
but
if,
if,
if
we
are
all
planning
on
going
to
vienna,
which
I'm
not
sure
we
are
all
planning
on
going
to
vienna,
but
let's
you
know,
let's
alternate
universe
this
and
say
everyone's
going
to
vienna.
I
would
propose
that
the
chairs
say
at
the
beginning,
or
you
know,
whenever
we
do,
the
the
updates
like
and
working
group
last
call
begins
now
right,
because
I
don't
think
it's,
I
don't
think
it's
feasible
to
target
ending
the
working
group
last
call
before
vienna.
A
A
N
I
I
strongly
suspect
that
I'm
gonna
well,
I'm
gonna
do
something
that'll,
probably
a
lot
of
other
people
are
doing
is
check
the
agenda
or
check
the
the
agendas
and
the
attendee
list
about
two
weeks
before
and
then,
like
you
know,
rush
book
it
two
weeks
before
right
because,
like
worst
case,
I
can
bicycle
home
from
vienna.
It'd
take
a
while
it'd
be
fun,
but
worst
case.
N
I
can
do
that
so
like
I
also
I'm
not
sure
what
the
travel
restriction
situation
looks
will
look
like
at
google
at
that
point
in
time.
Like
you
know,
if,
if
I
have
to
get
warren
kumari
to
sign
off
on
my
expenses,
then
that's
a
maybe
if
I
have
to
get
like
ben
traynor
to
sign
off
on
my
expenses,
that's
a
no!
So.
A
Cool
yeah,
so
I
guess
the
orthogonal
question
to
that,
because
we
are
going
to
have
a
hybrid
meeting
and
remote
attendance
should
be
sort
of
more
well
supported
than
at
previous
in-person
meetings.
We
can
always
have
a
tabs
session
and
I
was
wondering
if
ietf
113
will
be
a
good
time
for
us
to
have
a
tabs
session
again
to
sort
of
maybe
start
the
working
group
last
call.
A
N
A
A
A
F
F
K
Well,
I
think
brian's
tsv
art
talk
is
has
not
been,
has
not
been
overcome
by
any
of
the
changes.
The
document
we've
made.
Since
that's
true,
that's
true.
That's
like
a
that's
like
a
good!
Well,
I
mean
it's
like
20
minute
talk,
plus
40
minutes
of
arguing.
But
like
that's
that's
a
I
mean,
that's
that's
a
relatively
large
time
commitment,
but
if
people
really
want
to
know
what
it's
all
about,
like
that's
about
a
good
entry
point
as
you'll
get
so
that's
easy
to
find
in
the
youtube
archives.
I
M
K
What
the
heck
is
this
about
thing,
because
I
just
can't
keep
that
state
and
like
it,
and
it
turned
out
the
useful
thing
to
send
to
the
isg
like
you
can
skim
this.
You
know
a
thousand
words
about
what
the
heck
this
is
about
in
a
readable
way
and
if
and
if
the,
if
you're,
looking
for
an
entry
point
that
is
less
than
an
hour
of
time,
investment.
That
would
be
a
good
thing
to
do
just
really
accessible
text
about
what
the
heck
this
is
about
and
like
some
example
code
or
something.
K
K
N
N
K
K
N
K
A
N
M
Can
I
say
something
before
you
go
like
finalizing
the
wikipedia
thing,
because
I
I
mean.
Let
me
just
understand
the
thing:
the
wikipedia
anybody
can
edit
right.
So
are
we
saying,
like
a
taps,
working
group
taking
this
wikipedia
page
as
a
like
as
a
information
center
kind
of
thing,
and
I
take
the
ownership
of
that
one
and
what?
If,
if
somebody
just
screw
around
that
wikipedia
and
does
something.
A
N
E
K
Well,
I
mean
they're
they're
references
though
I
mean
so
that
the
way
that
the
way
that
manifests
itself
is
that,
like
you,
need
references.
So
if
I
write
an
article
about
myself
and
there's
no
media
about
me,
then
that
doesn't
get
anywhere
but,
like
you
can
point
to
internet
drafts
and
rfcs
and
that's
like
that's
the
documentation
that
they
need
right
and
like.
N
So
so
I
think
the
idea
and
what
what
alto
did
with
this
right?
The
idea
is
to
make
sure
that
people
who
are
not
necessarily
like
paying
attention
to
the
working
group
can
fall
into
having
information
about.
What's
going
on
right,
like
so
the
advantage
of,
if
we
put
something
on
a
working
group
wiki,
and
then
we
put
something
on
wikipedia,
I
I'm
pretty
sure
where
the
natural
seo
is
going
to
rank
those
in
the
search
rankings.
In
you
know,
your
favorite
search
engine.
K
It
is
true
that,
like
track
or
whatever
the
new
tool,
I
mean
the
one
issue
is
they're
like
migrating
from
track
to
something
else,
and
I
don't
know
what
something
else
is
yet
but
like
both
are
equally
fine,
I
don't
think
it'd
be
the
worst
thing
in
the
world
on
publicly
accessible
information
on
something
like
wikipedia
I
mean
we
don't
have
to.
Like
quote:
unquote,
maintain
it.
K
C
So
we
would
just
should
make
sure
that
everyone
agrees
that
putting
it
on
wikipedia
means
more
or
less
leaving
it
out
there
for
anyone
to
yeah,
add
things
to
it
and
make
sure
that
we
don't
call
that
we
own
the
text
and
it's
so
in
in
a
year
it
can't
it
can
happen
that
it's
not
brian
text
anymore,
and
if
it's
right,
we
should
be
fine
with
that
yeah.
That's
actually
what
we.
N
M
C
M
N
N
A
C
K
N
Yeah
and
like
the
the
apnic
blog
post
was
beginning
of
2021
right
like
so
it's
newer
than
some
of
the
bugs
that
we're
talking
about
today.
So
like
it's,
it's
accurate
to.
You
know
the
level
of
accuracy
that
we
were
talking
about
here
right,
I
don't
actually
remember,
but
the
tsp
talk
I
gave
that
was
in
person
right.
That
was,
you
know
that
was
in
the
into
the
luvian
past.
So.
M
M
Blog
post,
we
do
we
can
anyway,
somebody
else
in
this
world
can
take
in
taps
wikipedia
page
and
like
refer
to
all
the
rfcs
and
block
posters
that
we
are
doing
so
that
should
should
not
be
a
big
deal,
but
as
a
main
communication
method,
I
I
don't
want
to
take
the
taps
working
group
to
take
the
ownership
or
keep
it
a
content
in
any
sense.
So
that
can
be
my
only
concern,
but
I
really
like
what
you
guys
are
doing
here
discussing
so
please
continue
that.
Thank
you.
I
I
guess
right,
I
guess
we
could
have
a
a
sort
of
deadline
for
trying
to
get
all
the
prs
in
and
see.
If
we
can
agree
on
them.
We
don't
need
a
real
meeting
if
they're
pretty
close,
then
maybe
we
could
just
sign
them
off
and
then
we
have
a
more
clean
position.
So
maybe
just
an
email
around
reminding
us
when
that
comes
up
would
be
enough
rather
than
having
an
intro
a
week.