►
From YouTube: IETF-TAPS-20220615-1500
Description
TAPS meeting session at IETF
2022/06/15 1500
https://datatracker.ietf.org/meeting//proceedings/
A
C
A
C
A
A
A
Okay
cool,
so
we
are
currently
working
on
three
different
drafts
that
you
are
potentially
going
to
hear
about
today
and
yeah.
If
you
had
any
thoughts
on
those
drafts.
You're
also
welcome
to
send
an
email
to
the
list
and
share
your
feedback
on
the
draft,
and
we
also
have
a
github
where
we
are
collecting
issues.
We're
gonna
be
probably
looking
at.
C
A
C
D
Yeah
jian,
I
think
I've
asked
you
to
write
your
queries
in
a
mail
and
send
it
to
the
chairs
and
me
I'm
the
real
director,
and
then
we
can
sort
it
out
your
questions
from
there,
because
I'm
kind
of
getting
I'm
not
sure
like.
I
understand
what
exactly
you
want.
So
please
write
an
email
to
the
chairs,
tap
shares
and
taps
eddy.
D
E
B
A
A
F
F
G
B
B
E
E
E
I
E
I
E
I
E
E
I
I
I
actually,
I
don't
think
so.
I
mean
it's
it's
kind
of
the
opposite
of
being
having
low
latency
right,
so
in
case
of
low
latency,
you
may
want
to
lose
only
one
path
and
that's
the
one
with
the
lowest
round
trip
time,
whereas
in
case
of
scavenger
you
may
be
able
to
handle
more
delay
that
comes
from
interleaving,
multiple
paths,
so
you
have
to
reorder
things
on
the
receiver
side
and
now
stuff
may
take
longer,
but
that's
that's
kind
of
the
trade-off.
I
J
J
A
I
A
I
I
B
I
Well,
it's
very
easy.
I
mean
you
may
have
three
paths
and
you
have
three
connections.
Let's
say
three
connections
and
two
connections.
They
they
use
two
parts,
so
you
would
use.
I
mean
you
would
put
it
on
the
third,
where
it
gets
into
gets
out
of
the
way
of
all
the
other
traffic.
So
at
least
your
own
traffic
isn't
being
harmed
by
it
and.
I
G
B
K
K
Isn't
it
we
got
the
whole
time
we
discussed?
This
was
past
preferences
in
different
ways.
We
said
an
implementation
can
take
this
things
like
capacity
profile
into
account
when
doing
a
past
choice,
and
I
think
for
for
scavenger.
It's
the
same.
I
would
expect
for
for
today's
implementation
rather
to
use
the
default
person
if
available,
enable
that
if
an
implementation
is
smarter
and
we
we
have,
for
example,
some
kind
of
satellite
low
cost
high
bandwidth
link
available
in
five
years.
K
E
G
E
J
G
E
E
G
A
J
G
K
If
you
have
the
more
examples
you
have
the
the
more
open
this
will
be,
otherwise
people
will
say:
oh
we
don't
see
scavenger
here
we
can't
implement
that
or
on
the
other
hand,
we
end
up
with
a
lot
of
choices
that
will
be
much
difficult
different
from
implementation
to
implementation.
So
I
think
it's
diffic,
that's
my
fear.
If
we
remove
more
and
more
of
that
stuff.
K
Yeah,
so
so
I'm
I'm
completely
fine
with
was
making
the
was
making
the
capacity
profile.
This
general
description
clear
that
it's
just
a
list
of
examples
I
still
would
prefer
to
have
at
least
the
scavenger
type
in
the
list,
because
it's
something
that's
at
least
implementable,
and
we
have
an
idea
how
to
implement
it.
A
E
J
Yeah,
I
think
we're
actually
doing
the
right
thing.
I
think
we're
accentuating
the
idea
of
scavenger
being
a
congestion
control
type
thing
where
you
can
make
relative
priority
decisions
in
taps
and
we're
not
saying
that
we
can
choose
between
network
class
and
the
week.
One
actually
has
low
cost
or
a
high
available
bandwidth
and
whether
we
think
that's
acceptable
to
use
which
I
think
actually
can
be
tricky
trade-offs.
J
E
You
know
very
early
on,
say
you
know.
Look
at
this
section
of
the
api
document
for
property
preferences
be
more
explicit,
that
these
are
selection,
property
preferences,
and
I
took
this
suggestion
to
say:
okay,
you
know.
First,
your
talk
about
prohibit
properties
and
require
properties
and
just
kind
of
add
the
references
turbine.
Is
that
fine
with
you?
Since
you
opened
the
issue
yeah.
E
I
Well,
I
mean,
as
the
description
says,
right
that
I
I
what
I
did
is
I
I
read
this.
I
read
this
email
and
it
has
a
good
number
of
suggestions
that
are
just
small
wording,
fixes
you
see
a
timer
timer.
I
don't
know
tiny
stuff
like
that.
So
this
is
my
effort
to
put
all
the
tiny
stuff
together
and
in
doing
so,
I
also
noticed
that
there
are
these
references
to
property
names
that
use
well,
that's
that
has
to
be
the
old
style
that
we
once
had
before.
We
invented
property
right.
B
H
E
I
I
I
I
H
D
D
I
L
L
A
I
B
E
I
Yes,
we
had
that
issue
that
the
checks
failed
if
there
was
a
space
at
the
end
of
lines.
Yes,
yes
and
I
I
wrote
a
tiny
piece
of
python
called
for
myself
to
just
find
lines
to
have
a
space
in
the
end
and
remove
them
with
it
with
a
push
to
the
master.
And
after
doing
this,
the
pull
requests
after
that
didn't
have
didn't,
show
a
mistake
so
because
they're
probably
written
against
the
master
branch.
That
has
it
if
that
has
fixed
right.
E
E
I
I
E
A
H
E
E
L
A
L
L
I
would
say
a
framework
can
encapsulate
or
encode
outbound
messages,
decapitate
or
decode
inbound
data
into
messages
and
provide
other,
oh,
how
to
say
framing
without
saying
framing
yeah
or
right.
It's
like
the
language
is,
there
makes
it
sound
like
it's.
A
one-to-one
mix
between
the
framer
does
a
thing
and
a
message
pops
out,
which
is
what
we're
saying
is
not
the
race
right,
like
the
framework,
can
be
a
completely
stateful
thing
that
you
know
is
running
an
entirely.
E
L
A
A
E
E
E
J
E
E
A
F
H
I
I
H
I
F
E
I
E
J
E
F
F
F
F
J
E
E
F
G
F
K
K
E
E
I
E
I
E
E
F
L
F
J
F
E
F
F
F
E
B
A
E
E
A
H
A
A
A
A
A
So
I
guess,
once
we
have
gone
through
all
of
our
implementation
draft
documents,
we
will
most
likely
need
maybe
another
sort
of
path
to
see.
If
there's
anything
we
missed
when
aligning
our
three
documents,
because
we
were
gonna
hold
architecture
and
interface
until
implementation
is
done
yeah,
but
then
I
think
we
should
get
shepherds
right.
Abs,
I
think,
is
the
next
part.
J
A
A
So
we
had
basically,
we
had
assigned
three
different
people,
each
of
which
is
not
an
offer
for
that
particular
draft
that
they
would
be
shepherding
and
we
have
okay,
I'm
seeing
implementation
with
brian
interface,
anna
and
architecture.
Michael,
is
what
I'm
seeing
from
my
notes
from
an
interim
we
had
last
year.