►
Description
Evolvability, Deployability, & Maintainability (EDM) Program Meeting, 2021-04-22
A
All
right
stop
the
side.
Chatter
now,
so
what
we
would
like
to
do
is
you
know,
figure
out
how
to
actually
and
just
ship
this
dock
and
get
that
moved
on,
because
we
think
it's
useful
to
do
and
so
to
that
end,
having
a
smaller
group
actually
just
talk
about
practically
what
we
should
do.
I
think
it's
totally
fine.
A
A
On
this
editorial
pr,
I
think
the
only
thing
we
can
bike
shed
on
is
the
name
of
this
section.
A
I
I
don't
particularly
care
what
it
is,
but
I
don't
know
the
existing
title
felt
not
totally
representative
of
the
whole
sections
like
when
we're
going
over
in
the
abstract,
like
in
the
introduction.
We're
like
section
two
highlights
historical
issues
with
difficulties
and
transitions,
new
political
features
and
then.
A
B
B
Well,
maybe
it
is
so
one
of
the
things
that
we've
observed
is
that,
for
instance,
in
in
tls,
the
fundamentally
the
design
is
good,
but
that
didn't
prevent
it
from
being
impossible
to
to
evolve
using
that
design.
B
C
C
C
Correct
it
was
the
only
point
of
slight
disagreement
on
the
pr
the
so
like
tommy
rephrased
it
and
me,
like.
I
was
kind
of
not
too
happy
with
the
rephrasing.
C
I
see
I
see
what
you're
saying
yeah
I
I
mean
I,
I
don't
feel
too
strongly
here.
I
the
only
thing
that
we're
missing
with
yours
is
the
fact
that
it
implies
that
all
implementations
are
imperfect,
which
I
fundamentally
agree
with.
So
it's
not
something
as
a
wise
man
would
say.
I'm
not
gonna
lie.
A
Bugs
bugs
and
bugs
in
implementations,
limit
protocol
solution,
yeah
all
right,
we'll
bike,
another
time
going
into
issues,
that's
that
productively
discuss.
B
So
we're
gonna
start
with
34..
You
could
always
run
the
web
client
that
probably
won't
generate
echo.
A
A
Cool
so
the
first
thing
that
stood
out
to
me
in
my
last
pass
of
this
was
we
have
this
section:
independency
is
better,
I'm
saying
you
know
active
use,
dependency
is
better.
Caution
is
advised
to
assume
to
avoid,
assuming
that
building
dependency
on
extension
is
sufficient
to
ensure
that
it's
going
to
be
available
long
term.
So
it
seems
like
we
want
to
be
able
to
point
out
an
example
of
like
I
guess
you
know
if
this
were
to
happen
in
http.
A
Can
correct
me
if
I'm
wrong,
martin
it'd
be
like
hey,
we
have
these
great
headers,
but
for
some
reason
in
2005
people
stopped
creating
new
headers
and
then
at
after
that,
everyone
ossified
on
a
certain
level,
a
certain
set
of
them
and
said
anything
new
out
of
this
is
probably
going
to
be
attack.
Let's
reject
it
all
blah.
B
You're
talking
about
the
http
scenario,
where
you've
got
all
these
web
application
security,
things
that
sit
in
the
front
and
say
oh,
this
is
what
the
web
looks
like
in
april.
22Nd
2021
and
that's
it.
A
B
Yeah,
I'm
not
sure
that
that's
exactly
the
the
point
that
I
I
picked
up
with
my
comment
there.
It's
probably
more
along
the
lines
of
all
those
cases
where
the
extensions
exist,
but
then
it
was
a
finite
set
and
it
never
went
anywhere.
The
only
thing
that
I'm
personally
aware
of
is
taylor.
Signature
algorithms
had
a
a
nice
problem.
You've
got
a
couple
of
implementations.
B
A
Although
I
mean
in
this
section
we're
trying
to
say
that
essentially,
dependency
is
not
enough
to
eliminate
the
the
dangers
of
disuse
that
it
is
not
not
sufficient
to
solve
the
problem
and,
arguably
arguably
the
example
of
s.
I,
like
you,
didn't,
have
to
speak
the
s
I
extension
in
order
to
be
able
to
speak
tls
at
all.
It
wasn't
a
core
dependency
of
the
protocol.
Although.
C
Let's
just
to
clarify
here,
tommy,
the
the
fact
that
s-
and
I
was
used
is
not
really
the
relevant
point.
It's
the
fact
that
anything,
other
types
of
names
inside
the
sni
were
never
used.
That's
the
that's
the
point
here.
The
extensibility
joint
of
the
tls
extensions
was
used
and
that's
why
that
one
hasn't
rusted
shut,
it's
the
other
one
that
did.
A
Yeah
understood
so
I
mean
going
to
the
what
you're
trying
to
point
to
for
signature
algorithms.
Like
clearly
I
mean
you
have
to
speak
signature
algorithms
to
work
at
all,
but
it
was
enough
of
a
finite
set
that
you
were
like
eh.
These
are
the
ones
I
support
and.
B
Yeah
and
then-
and
it
meant
that
there
was
no
way
to
add
to
this
set,
and
so
the
cool
thing
that
we're
talking
about
here
is
what
what
makes
it
possible
to
add
to
the
protocol
and
the
fact
that
we
had
something
didn't
necessarily.
G
B
That
there
was
any
any
chance
that
things
could
be
added.
I
think
that's
probably
the
caveat
that
we're
looking
for
rather
than
an
example,
though,
where
is
it
now.
A
I
wonder
if
yeah
so
in
the.
A
A
A
A
A
A
Rat
hole
comment,
but
it
just
the
the
current
text.
You
know
just
ending
this
pair,
essentially
ending
the
section
on
active
use
of
like
we
have
no
advice
for
you.
A
B
Mean
yeah,
it's
just
it's
an
interesting
effect,
so
the
theory
there
that
I
have
at
least
is
that
as
long
as
the
the
cycles
of
addition
are
faster
than
the
the
product
cycles
of
the
people
who
are
going
to
ossify,
then
we're
probably
okay,
but
anything
slower
than
that
is,
is
going
to
be
a
problem
potentially.
B
A
B
Well,
it's
going
to
depend
right
on
on
a
whole
lot
of
factors
right.
So
if,
if
you've
got
a
product
cycle,
that's
a
month
long
because
you're
actively
responding
to
things,
then
you
might
forget
by
the
time
that
you
put
out
the
fifth
version
of
the
protocol,
that
things
might
change
and
you
might
accidentally
also
find
it.
But.
E
There's
also
like
counter
examples
to
that,
where
you
could
you
could
ship
stuff
after
I
guess,
the
middle
boxes
have
updated.
For
example,
like
say
they
had
some
like
wonky,
like
machine
learning.
That
was
like
I'm
gonna,
try
and
learn
what
the
shape
of
tls
looks
like
and
then
eventually
like
they
learn.
What
tls
looks
like
and
then
you
try
to
change
something
and
they
fall
over
or
whatever,
like
you've
done
things
in
the
right
order,
but
like
it
just
took
time
for
the
ossification
to
occur,
for
whatever
reason.
B
B
A
B
Right,
I'm
not
even
sure
that
it's
future
work
that
would
be
generically
applicable
either.
So
if
you
look
at
something
like
http
that
there's
nothing,
we
can
do
and
there's
probably
nothing.
We
need
to
do
because
there
are
so
many
people
out
there
already
putting
new
stuff
on
on
top
of
the
protocol
that
anyone
anyone
trying
to
also
find
this
one
would
would
find
it
very
challenging.
But.
C
Well
so
one
thing
there
was
the
discussion
that
we
had
with
like
mdot
some
number
of
months
ago
about
like
doing
some
coordination
and
having
all
the
browsers
like
start
greasing
at
the
same
time,
to
kind
of
power.
Through
this
you
know
you
know,
power
will
wash
the
rust
away
kind
of
deal
and
that's
something
we
could
potentially
talk
about
as
a
technique.
B
A
B
A
Right
yeah,
so
what
makes
sense
for
one
protocol.
G
B
A
B
A
Very
good,
since
we're
doing
pretty
good
on
time
one.
A
B
B
Yeah,
I
think,
that's
probably
probably
a
shuffling
thing
that
we
can
do
here
so
that
the
classy
address
space
probably
belongs
back
into
two
all
right
and
there's
some
other
text
there
that
might
need
to
move
so
you're
right.
There's
just
some
shuffling
might
be
the
right
way
to
do
this,
and
maybe
maybe
once
that's
done,
we
can
find
some
things
to
cup
right
yeah.
I
was
kind.
A
Of
wondering
like
do
we
do
we
need
kind
of
a
third
point
in
this
in
this
section
in
section
three,
are
we
happy
with
it
being
like.
B
Active
use,
it's
good
dependency
is
good
yeah.
The
title
would
be
wrong.
I
think
in
that
case,
if
you,
if
this
was
to
be
retained,
I
think
it's
it's
about
you
know.
Maybe
it
can
go.
Maybe
you
can
go.
I
think
we
could
probably
fold
the
positive
bits
into
the
previous
section
and
then
the
negative
bits
can
go
up
into
two
two
and
okay
yeah.
Let's
work
through.
G
Mention
that
I
I've
gotten
this
very
late,
so
I
used
the
time
to
go
through
the
document.
Read
it
and
at
least
in
my
hired
state
it
felt
pretty
good.
G
G
So
I
think
I'm
gonna
fold
from
this
call
and
go
to
bed.
B
B
A
A
I'm
happy
with
that.
So
for
some
of
these
editorial
things,
do
you
want
me
to
propose
things?
Did
you
want
to
try
to
reshuffle
things
either
way.
E
B
A
C
Oh
just
well,
so
you
got
this
one
covered,
but
I'm
happy
to
spend
some
time
and
provide
text
as
well.
A
Great
yeah
did
you,
let's
switch
over
to
what
I
was
going
to
have
you
chat
about
for
greasing
and
thoughts
about?
Do
we
mention
quick
in
here
sounds
good.
C
You
want
to
take
over
the
minutes.
I
put
the
link
to
the
doc
in
the
in
the
chat
in
the
chat,
yeah
all
right
and
remind
me
which
issue
I
was
going
to
discuss
again.
You
told
me
yesterday
and
then
I've
said
it.
A
C
D
D
C
D
All
right,
can
you
see
it?
Yes,
cool.
D
E
C
Brilliant,
the
oh
there
we
go
yeah,
the
chrome
sharing
thing
goes
over
all
the
other
controls,
it's
like.
Obviously,
the
bottom
of
the
screen
in
the
middle
is
prime
real
estate
yeah.
C
So
the
the
document
does
a
really
good
job
about
talking
like
or
what
we
can
do
when
we're
creating
a
new
protocol,
and
so,
like
you
know,
quick,
and
I
think
I
mean
in
a
way,
you
can
really
tell
that
this
was
written
by
someone
involved
in
the
early
design
of
quick,
because,
like
quick,
is
the
poster
child
for
all
the
good
ideas
in
here,
but
it
doesn't
talk
too
much
about
how
to
like
retrofit
them,
because,
like
that's
what
we
did
with
for
tls,
but
we
now
don't
necessarily
tell
that
story,
or
you
know
it
doesn't
say
like.
C
C
Do
we
have
thoughts
there?
Should
I
just
write
a
pr
see
if
there
are
any
ideas
in
there.
B
Yeah,
so
I
was
going
to
come
at
this.
I
thought
you
might
be
coming
at
this
from
a
different
angle,
which
is
good.
Yes,
it
does
sort
of
not
really
talk
about
the
difference
between
new
and
retrofit
there's
different
opinions
on
this
piece.
B
C
Really
I
mean
I
still
remember.
I
was
telling
tommy
this
yesterday
that
I
tried
to
interrupt
with
apple
in
london
and
they
crashed
their
implementation
and
you
went
what
because
it
was
because
I
had
a
transfer
parameter
that
they
didn't
know,
and
you
went
oh
and
you
wrote
the
pr
like
right
away
to
add
the
greasing
like
that.
That
was
kind
of
like
in
the
moon
moment,
and
it
was
the
right
thing
to
do
too
and.
A
H
H
But
I
think
that's
like
that's
why
I
would
distinguish
here
because
that's
always
the
better
option.
Only
if
you
got
yourself
into
a
debt
and
then
you
have
to
start
doing
this
ugly
hack,
which
is
called
greasing.
A
Well,
so
I
guess
to
unpack
that
bit
is
what
you're,
referring
to
mainly
the
fact
that
you
have
to
reserve
given
code
points
as
grease.
Only
code
points.
C
It
will
reserve
them
and
also
it
encourages
implementation
to
use
those
greasy
code
points
and,
and
that
actually
helped
it
wasn't
just
apple
like
we,
for
example,
google
didn't
handle
grease
http
frames
at
the
beginning,
because
we
did
we
needed
to
write
a
special
parser
for
the
like
dump
the
data
on
the
ground
as
it
comes
in,
because
someone
could
send
you
four
gigabytes
of
grease
frame
and
you
don't
want
to
put
those
in
memory
before
you
drop
them.
H
Yeah,
it
helps
you
to
detect
implementation
limitations
very
efficiently
right,
but
it
also
has
a
cost
and
if
you
could
have
the
same
effect
by
just
using
any
kind
of
extendability,
that
would
always
be
better.
I
mean
yes.
B
Yeah,
I
suspect
that
this
is
really
just
a
matter
whether
it
be
a
new
protocol
or
an
old
one
of
providing
filler.
So
there's
going
to
be
points
in
time
where
you
add
new
stuff
and
those
events
will
be
important
right
and
ultimately,
you
want
that
cadence
to
be
fast
enough,
that
you
never
have
any
any
period
where
no
one
encounters
anything
new
such
that
they
can
deploy,
something
that
occupies
on
whatever
the
current
state
is,
and
so
greece
allows
you
to
artificially
increase
that
cadence,
which
is
part
of
the
reason
why.
B
A
Yeah,
but
I
mean,
but
anyway
it's
interesting
but
like
when
the
some
of
the
sexy
headers
went
down,
I
mean
they
had
bugs
with
some
crazy,
like
backhand
apple
servers
that
were
being
dumped.
So
right
that
clearly
would
have
benefited.
A
You
know,
potentially
from
those
types
of
things
being
greased
more.
So
I
guess
the
question.
A
As
you
were
arriving
and
I
think
what
we
discussed
there
was
to
say,
let's
you
know,
I
personally
think
that's
you
know
that's
kind
of
like
a
future
research
thing
or
topic
we
can
discuss
more
but
like
we
can
just
more
concretely
say
here.
You
know
we
don't
know
what
it
is.
More
experimentation
needs
to
be
done,
it's
going
to
be
per
protocol
and
per
deployment,
and
then
I
think
we
should
have
that
discussion
in
a
program
meeting
in
the
future.
A
Going
back
to
this,
like
the
concern
you
bring
up
maria,
though
I
I
wonder
if
it's
worth
like
saying
in
the
grease
section,
very
explicitly
how
like
the
the
mechanism
that
tls
uses
is
one
algorithm
that's
suited
to
its
code
points
on
an
existing
protocol,
but
maybe
you
know
expand
the
definition
to
say
like
you
can.
Do
it
many
many
different
ways,
because
you
know
for
quick,
quick
said
we
have
this
enormous
number
space
and
so
we're
just
going
to
reserve.
A
You
know
parts
of
it
and
it's
very,
very
cheap,
so
it
doesn't
have
that
cost
problem
are
they
referring
to,
but
you
could
also
have
a
grease
mechanism
that
says
you
know
like
these
numbers
are
not
going
to
be
forever
allocated.
It's
not
like
they're,
going
to
be
blacklisted
forever
off
of
the
analyst.
B
Everyone
agrees
that
there's
going
to
be
the
one
change
to
the
protocol
right,
so
we
turn
we,
we
reverse
the
order
of
the
client
random
or
something
like
that
right
and
the
problem
there
is
that
it
just
comes
down
to
coordination
problem
and
we're
being
lazy
in
in
quick,
we're
just
saying
that
these
versions
are
reserved
forever
and
that's
easy,
because
we
do
it.
We
do
that
once
we
don't
have
to
worry
about
coordinating
the
changes,
but
you
could,
you
could
say:
well,
you
know
for
2021
these.
A
C
I
mean
so
so
we
actually
discussed
this
a
little
bit
yesterday
and
quick
the
idea
being
that
if
we
have
compatible
version
negotiation,
we
can
start
mucking
with
this
stuff,
and
that
could
be
fun.
You
don't
actually
completely
need
compatible
version
negotiation
because
you
know,
if
you
just
say
well
use
the
one
you
have,
but
it's
still
like
it
allows
us
to
do
more
things
like
this.
Where
maybe
chrome
starts
minting,
you
know
unilaterally.
C
Min
and
google
start,
you
know
minting
a
new
version
every
six
months
and
we
tell
people
hey.
We
have
this
one.
If
you
want
to
play
along,
you
can
we'll
still
support
the
main
one.
But
if
you
know
there's
enough,
you
know
maybe
a
second
browser,
or
you
know
a
second
cdn
that
plays
long
it'll,
help
break
some
some
boxes
that
don't
that
can't
handle
this.
B
Yeah
yeah
and
the
question
is
whether
the
the
effort
involved
in
coordinating
that
is
justified
by
the
benefits
that
that
come
from
it,
and
this
is,
I
think,
one
of
the
key
core
debates
about
greasing
as
well
is
is
the
effort
that
we're
putting
in
you
know,
reserving
all
the
code
points
sending
all
the
extra
stuff
and
writing
all
the
code.
Does
that
justify?
B
It's?
Is
that
effort
justified
relative
to
the
the
benefits
that
we're
seeing
and
certainly
we
saw
benefits
from
greasing
tls
early
on.
That's
because
tls
is,
and-
and
by
that
I
mean
the
deployment
ecosystem
is-
is
just
a
complete
disaster.
B
H
H
That's
the
important
part
like
how
you
use
it
is
a
different
question
and
like
I
would
be
really
careful
to
just
see,
look
at
tls
and
see
that
greasing
work
very
well,
and
just
take
this
as
an
example
or
like
an
evidence
that
greasing
works
well
everywhere,
while
there
are
other
ways
to
just
use
these
extended
bulletins.
A
So
here's
a
question
for
kind
of
us
all
because
I
think
maybe
you're
right
that
it's
important
to
have
active
use.
I
think
it
is
important
also
to
have
like
artificial
use
like
to
have
the
ability
to
do
artificial
use
of
the
extension
point
when
you
don't
actually
have
a
new
protocol
thing
to
do
like
we
can't
rely
on.
If
we
need
a
monthly
cadence,
we
can't
rely
on
someone
coming
with
a
new
standard
or
a
new
experiment.
A
E
A
E
Well,
it
serves
more
it's
more
than
that,
but
that
is
one
of
his
purposes.
B
I
I
think
that
just
reading
through
the
section
here-
there's
probably
probably
the
real
point
that
miriam
might
be
looking
for
here
is,
is
what
david
wrote
down
in
the
comment,
which
is
the
the
artificial
stuff,
is
no
real
substitute
for
the
real
stuff
when
it
comes
to
to
making
sure
that
you
can
use
those
extension
cut
points,
because
if,
if
people
are
hard
coding
their
code
to
throw
away
everything-
oh
that's
true,
then
you
may
end
up
in
a
situation
where
you
just
can't
deploy
it,
even
though
the
greasing
stuff
gets
through,
and
so
the
the
the
idea
that
greece
is
is
filler
for
the
real
stuff
is
probably
something
that
we
could
say
more
directly.
H
H
B
That's
actually
not
a
bad
suggestion,
even
though
I
don't
like
adding
extra
levels
to
the
tree,
we
could
probably
change
the
name
of
the
section
to
to
falsifying
active
use.
Yes,.
E
B
And
recognizing
the
fact
that
that's
all
it
really
is,
and
with
that
title
and
maybe
with
a
couple
of
tweaks
to
the
language
there
that
conveys
the
the
message
that
we,
what
when
you
say,
you're
falsifying
activities,
it's
very
clear,
then
that
you're,
not
it's
not
the
real
thing,
that's
not
as
good
as
the
real
thing,
and
maybe
that's
that's
better
than
than
what
we
have
all
right.
I'm
gonna
make.
A
A
Greece,
as
like
the
title
for
the
strategy,
so
we
can
describe
it
in
these
more
abstract
terms.
Yeah.
H
Where
else
you
need
to
grease
right,
it's
not
like
really
part
of
the
solution.
It's
more
a
definition.
A
Although
it
is,
I
mean
I
do
find
that
it
is
a
defensive
design.
I
I
agree
that,
like
active
use,
virgin
negotiation,
cryptography
grease
or
whatever
we're
going
to
call
it
like,
you
know,
artificial,
active
use
or
something
seem
to
be
of
one
ilk
like
they
are
things
you
do.
A
H
Message
would
be
right,
so
the
invariants
are
there
to
ossify,
and
you
know
that
and
you
write
it
down
and
everything
else.
You
need
to
make
sure
it.
It
stays
extend
or
it's
just
changeable
in
that
case,
and
so
you
have
to
act
on.
A
A
C
On
the
topic
of
invariance,
just
a
quick
slight
aside
that
I
thought
was,
it
might
be
interesting.
I
had
a
conversation
yesterday
with
ryan
hamilton,
where
it
was
like
right
after
the
quick
intro
we
were
talking
about.
You
know
using
versions
versus
using
extensions
for
new
features,
and
I
mentioned
to
him
like.
I
was
surprised
that
google,
quick,
like
realized,
relied
so
heavily
on
on
versions
even
like,
let's
say
when
they
added
the
datagram
frame.
C
It
was
a
new
version
and
he
made
the
really
good
point
that
when
they
first
started
this
they
had
no
idea
what
the
variance
would
be.
They
didn't
know
what
they
were
doing
where
they
were
going,
and
so
it's
really
easy
to
look
back
and
say.
Well,
yes,
the
client,
hello
and
google
quick
had
an
invariant
format,
the
version
field
being
there
had
invariant
format.
C
These
bits
were
variant,
those
turned
out
not
to
be,
but
when
you're
at
the
beginning,
it
is
a
lot
harder,
and
I
thought
that
was
an
interesting
point
in
you
know
the
design
of
all
these
things.
I'm
not
sure
if
this
is
something
that
we
could.
That
would
be
useful
to
add
to
the
document,
but
I
thought
it
was
interesting
to
noodle
about.
H
C
I
mean
how
it
should
be.
You
know
like
we
were
also
designing
standards
at
the
atf
right.
We're
not
only
doing
things
that
other
companies
have
done
before.
H
C
A
Yeah,
anyway,
back
to
the
yes
here,
so
I
guess
you
know
we
bring
up
the
section
on
variance
kind
of
its
position.
Also,
I
mean
I
would
also
kind
of
ask
the
same
thing
for
the
effective
feedback.
H
B
B
The
second
one
might
actually
be
almost
as
important,
if
not
depending
on
your
view,
like
ryan,
seems
to
take
the
view
that
the
version
negotiation
stuff
is
more
important
than
being
able
to
do
the
extensibility
thing,
and
I
at
some
level
I
agree,
and
yes,
we
like
crypto
and
then
we
get
into
the
weaker
stuff
and
the
ordering
is
kind
of
deliberate
in
that
way.
But
it's
only
approximate
in
terms
of
what
we
think
are
the
the
best
options.
B
Changing
the
way
the
text
gives,
the
impression
of
the
value
of
things
is
the
right
way
to
to
level
set
on
these
yeah.
H
But
I
don't
think
these
are
kind
of
like
comparable
options
right.
The
version
negotiation
is
a
little
bit
like
something
that
you
could
also
call
active
use.
Probably
it's
just
like
giving
you
more
detail
about
how
to
do
things
specifically
about
versions
and
and
greasing
is
like
how
you
get
active
use.
H
B
H
Yeah,
so
crypto
just
helps
you
to
protect
against
in
network
funcification
yeah.
C
Or
about
reducing
the
number
of
parties,
which
is
you
know
something
that's
mentioned
earlier
in
the
yes
in
the
draft?
I
thought
that
was
a
really
good
way.
To
put
it.
E
Does
it
also
run
the
risk
of
like
triggering
potential
misbehavior,
though,
like
I
just
keep
falling
back
on
ech
like
the
fact
that
we're
using
ech
is
probably
going
to
make
some
people
upset
and
the
networks
might
behave
differently
because
they
have
ossified
on
the
fact
that
they
already
have
the
clear
text
and
not
the
cipher
text.
So,
like
I
don't
know,
it
cuts
both
ways,
though.
B
Of
course,
I
think
I
think
to
the
point
that
ach
is
a
new
thing
is
part
of
the
problem
there
right.
So
if
we,
if
we
started
with
everything
encrypted
from
the
outset,
then
no
problem,
and
once
you
manage
to
equip
things
then
then
you're
in
a
better
state
than
you
were
before,
but.
C
To
nest
and
be
it
makes
things
harder
to
read.
I
think
some
of
that
well.
D
B
Reminds
me
of
the
the
old
oh,
this
is
a
proposed
standard.
All
proposed
standards
are
equal
right,
that's
it's!
It's
not
really
the
case,
and
I
don't
think
anyone
reading
through
should
be
getting
that
impression
and
if
they
do,
then
that's
a
problem
with
the
text
and
not
a
problem
with
the
structure.
B
H
B
H
I
wonder
if
this
really
has
to
be
a
list
like
this
or
if
it
would
be
stronger
if
you
have
like
two
or
three
main
sections
here,
one
on
active
use
and
then
a
separate
one
on
crypto
and
then
maybe
a
separate
one
on
invariants
or
put
parents
somewhere
else.
I
don't
know
yeah.
A
A
There
crypto
and
invariance
I
feel
like
are
things
that
are
less
about
using
the
deployment
they're
a
little
bit
more
like
these
decisions,
you
make
of
saying:
hey,
I'm
going
to
choose
these
as
in
variance
I'm
going
to
choose
to
use
crypto
and
then
it's
kind
of
orthogonal
from
using
it
or
deployments
just
like.
I
did
it
and
I
hope
it
worked.
I
hope
it
helped
me
and
then
effective
feedback
is
almost
like
the
other
end
of
saying,
like
it's
not
even
really
about
the
protocol
necessarily
itself.
A
It's
just
like
the
super
active
like
keep
paying
attention
to
how
your
protocol
works,
but
there's
like
different
levels
of
activity
and
engagement
and
yeah.
I
think
what
this
is
all
pointing
to
and
maybe
is
like
the
thing
we
need
to-
maybe
we'll
never
figure
it
out,
but
like
we
could
talk
about
in
the
future.
Is
that
thing
of
like
you
know?
What
is
this
cadence
and
it's
like
realizing
that
protocols
are
not
static,
things
that
there
are
cadences
that
they
have
to
their
evolution
of
their
lifetime
and
that
determines
both?
H
Yeah,
so
I
I
think,
like
the
one
thing
I'm
really
unhappy
about
this
list
is
that
it
has
like
active
use
on
the
same
level
as
everything
else,
because
I
think
active
use
is
really
the
main
point
here.
So
it
should
be,
should
be
you
know
its
own
main
section
or
whatever.
It
should
be
more
in
focus.
It's.
H
B
I'm
struggling
with
structure
things.
It
may
be
that.
B
The
way
to
structure
this
is
this:
there's
a
section
four
which
has
active
use
and
as
the
title
and
we
then
sort
of
carve
off
the
the
the
secondary
things
like
the
invariants
and
the
feedback,
maybe
not
the
grease
and
maybe
the
crypto.
I
don't
know
into
a
and
other
other
techniques
for
managing
sort
of
that.
One.
A
A
You've
already
done
so,
oh
lovely,
I
don't
think
we
probably
have
time
to
review
merchant
right
now,
but
let's
merge
that
in
and
then
I
I
can
just
write
up
an
issue
to.
I
can
try
to
take
a
stab
at
rearranging
the
things
and
you
can
just
see
if
we
hate
it
or
not.
B
Yeah,
I
I
think
it's
probably
going
to
be
reasonable
to
say
two
two
lists
of
three
items
would
work
and
what's
interesting
about
that,
is
that
the
4.1,
which
is
basically
the
what
is
active,
use
kind
of
becomes
the
top
level
for
a
new
section.
A
B
B
A
A
Okay,
we
are
pretty
much
at
the
hour,
but
I
think
we're
you
know
getting
eyes
on
this
and
making
progress
do.
E
You
think
there's
an
issue
on
the
the
threat
model.
I
I
don't
know
if
we
want
to
address
it
at
all,
you
can
feel
free
to
close
it,
but
like
and
we're
at
the
top
of
the
hour
anyway.
So
we
we
can
just
kind
of
move
on.
A
B
I
just
hadn't
seen
that
yet
that's
fine,
some
thinking
on
that
might
be
necessary.
I
think
you
probably
got
some
ideas
here
that
would
be
useful
to
to
work
through
whether
or
not
it
changes
anything
in
the
document,
and
I
don't
know.
C
I
think
having
a
pr
would
help,
because
I
I
like
the
idea
in
general,
but
I
wouldn't
know
how
to
write
it.
But
if
you
have
a
pr,
then
we
would
totally
use
to
say
if
it
makes
sense
or
not.
Yeah.
C
It's
not
quite
that
it's
just
like
there
are
cases
where
it's
like.
Okay,
I
don't.
I
can't
be
bothered.
This
is
a
okay.
I
think
there's
something
here,
but
I
don't
know
how
to
write
it
myself.
E
H
So,
what's
the
what's
the
sorry,
what's
the
what's
the
timeline
here
I
mean
this
document
has
been
around
for
a
while
and
I
think
it's
a
pretty
good
shape
and
sure
we
have
some
more
things
we
want
to
discuss,
but
I
think
we
should
try
to
get
it
out
the
door
in
a
reasonable
amount
of
time.
H
A
C
Yeah
aiming
to
get
it
fully
published
by
the
next
itf
would
be
good,
if
not
that,
at
least
in
the
hands
of
the
rfc
editor
by
then
would
be
nice.
H
Then
we
should,
we
should
try
to
figure
out
how
we
get
to
a
version
that
we
can
run
an
adoption
call
for
in
the
ib,
because
we
now
also
have
this
feedback
around
for
adoption.
Calls
right.
C
F
H
A
C
And
I
think
we
have
a
few
issues,
but
there
are,
they
all
have
owners.
So
if
we
like
in
the
coming
weeks
like
next,
you
know
try
to
have
prs
for
all
those
we
should
be
able
to
land
them,
and
then
you
know
by
some
point
in
may:
bring
it
before
the
iab
can.
E
B
B
Can
we
meet
again?
Can
we
say
let's
try
to
get
the
things
we've
been
assigned
to
do
done
in
about
two
to
a
bit
weeks
and
then
meet
again
shortly
after
that,
and
then
just
go
through
the
issues
and
say:
yes,
we've
done
it?
No,
we
don't
plan
to
do
it
and
have
a
final
sanity
check,
because
there's
a
couple
there
that
I,
I
suspect,
it's
just
the
case
that
they
were
opened
and
we
don't
have
any
fixed,
exactly
yeah.
It's
too.
A
So
let's
say
the
week
of
may
10th:
let's
see.
H
And
can
I
can
I
ask
one
more
thing
so.
B
H
Like
you
know,
for
this,
this
meeting
was
focused
on
this
document
and
like
having
a
small
group.
Talking
about
is
actually
very
helpful,
but
I
think
what
we're
seeing
here
is
that
otherwise
people
are
not
feeling
committed
to
be
part
of
the
program
right
they
might
or
might
not
show
up,
and
we
discussed
about
having
memberships
to
actually
kind
of
motivate
people
to
actually,
you
know,
feel
being
part
of
the
program
and
so
on.
A
Yeah,
I
think
it's
largely
schedules
and
stuff,
but
we
we've
had
some
fairly
consistent
folks
and
I
I
think
on
my
end,
you
know
I
once
I
would
hope
after
this
next
meeting,
we
kind
of
move
out
of
talking
about
just
this
document
and
talk
talk
about
some
of
these
other
ideas
and
specifically
around
like
these
deployability
cadences
and
how
that
affects
things,
and
I
think
that's
a
good
time
to
also
kind
of
solicit
people
to
say:
hey
pay
attention
come
back
in,
but
yeah
we
can
chat
more.
H
Yeah
I
mean
I'm,
I'm
not
I'm
not
sure
about
this
whole
membership
idea.
If
that's
a
good
thing,
a
bad
thing,
and
if
we
would
have
in-person
meetings,
maybe
more
people
would
show
up
anyway,
because
then
it's
a
different
situation,
but
but,
like
this
whole
point
about
membership
for
me,
was
exactly
that
people
feel
committed
and
that
people
actually
think
they
are
mandatory
to
show
up
right
kind
of
instead
of
just
like
you
know
it
doesn't
matter
if
I'm
there
or
whatever
yeah.
A
H
It
I
mean,
as
I
said,
this
was
really
focused
on
this
document
and
in
that
case
it
it's
actually
nice
to
have
a
small
group
who
are
actually
those
people
who
are
actually
engaged
about
program
about
the
draft
but
like
in
general.
If
you
ever
want
to
have
a
more
wider
discussion,
it
actually
is
helpful
to
have
a
few
more
people
as
well.
I
think.