►
From YouTube: IETF-TAPS-20210922-1500
Description
TAPS meeting session at IETF
2021/09/22 1500
https://datatracker.ietf.org/meeting//proceedings/
A
Text
into
the
into
the
notes,
it
asks
me
to
sign
in
by
the
data
tracker,
so
I
think
that's.
Oh.
A
B
B
B
B
C
A
Screw
that
I
mean,
is
there
a
cost?
I
mean
it's
not.
I
can't
see
any
incentive
in
doing
that.
Well,.
C
D
A
Guess
now
I
can
see
you.
Yes,
I
can
see
you
have
a
guitar
on
your
wall.
I
think
that's
like
standard
issue
for
for
anybody
in
networking
or.
A
F
G
G
A
That
in
your
your
kit,
set
up
in
your
office,
michael
I'm
loving
my
office,
not
in
your
office.
A
So
there's
no
separate
blue
sheet
feature
right.
We
still
want
to
keep
track.
K
B
A
I
I
must
have
missed
that
feature.
A
The
miracles
of
well
now
there's
an
echo
miracle.
B
Oh
yeah,
and
just
in
case
people
haven't
seen
it
there
is
a
gallery
gallery
view
on
the
top
right.
Second,
second
button,
which
is
gallery
view,
and
you
actually
get
these.
D
B
Yeah,
but
I
don't
think
I
have
actually
ever
used
the
gallery
view
in
my
deco.
A
A
Of
a
sort
of
a
linear
agenda
presentations
at
the
meeting-
and
so
the
kind
of
the
way
we've
been
using
webex
for
the
interims
is
a
little
bit
different.
It's
cool,
although
it
doesn't.
Oh,
I
see
the
names
you
have
to
hover
to
see
the
names
for
the
pictures.
That's
I
think
the
video
quality
is
really
good.
I'm
impressed
now,
of
course,
we're
going
to
push
to
do
this
all
and
gather
right.
So
we
got
to
make
our
avatars
and
you
know
everybody
get
their
exit
skills
going.
L
B
So
our
first
agenda
item
is
actually
martin
demoing,
his
code,
brian,
I'm
not
sure
if
you
were
going
to
drive
the
github,
but
that
would
be
yeah.
C
I
am
not
on
my
corp
machine
right
now,
which
means
I
am
you
know,
allowed
by
security
to
drive
to
github,
but
but
I
wanted
to
see
if
there
was
a
way
that
I
could
point
it
at
a
window
as
opposed
to
the
full
screen.
But
well
we'll
have.
B
Martin
go
for
it,
let's
see,
okay,
but
now,
I'm
not
sure
if
it
switches
you
to
a
presentation
you
automatically,
but
it
does.
Martin
is
just
starting
a
screen
share.
Okay,.
H
M
K
K
This
effort
has
two
purposes
two
main
purposes:
one
is
to
help
build
a
linux
ecosystem
for
taps
by
you
know,
I
mean
it's
nowhere
near
ready
to
do
that,
but
eventually
to
to
build
this
up
to
a
thing
where
people
might
actually
use
it
and
put
it
on
their
on
their
boxes
and
their
linux
servers
and
whatnot
and
and
start
using
it,
and
then
the
second
thing
was
to
exercise
some
of
the
ideas
in
in
transport
discovery
in
that
draft.
K
So,
let's
see
probably
the
best
way
to
show
you
is
just
to
this
is
a
this
is
a
sample
program.
I
wrote
this
is
just
an
an
echo
server,
so
you
you
know
you
telling
it
into
it
and
you
you
type
stuff,
and
it
sends
it
back
to
you.
It's
a
very
common
toy
program.
Of
course,
just
give
you
a
sense
of
what
a
program
would
look
like
using
this
interface.
K
So
that's
what
you
see
here,
so
here's
the
one
when
you
get
a
new
connection
from
a
listener.
It
then
you
know,
initiates
a
receive
and
then
the
other
one
interesting
is
see.
When
we
receive
some
data
it
will,
then
you
can
see
there
in
line
119.
It
sends
that
and
then
it
starts
receiving
it
again.
So
you
guys
obviously
remember
the
interface,
that's
what
it
looks
like
and
then
the
main
program
is
here:
it
initializes
an
endpoint.
K
It
sets
some
transport
properties
and
then,
let's
see,
then
it
frees
all
the
preconnection
stuff.
It
doesn't
need
it
anymore
and
then
just
sits
around
and
waits
until
you
terminate
the
program
like
most
listeners,
so
I've
already
compiled
that,
and
so,
if
you
just
run
the
program,
there's
a
ton
of
debugging
here,
which
you
guys
might
find
somewhat
interesting.
K
And
so
now,
if
I
do
stuff,
it
echoes
it
and
I'm
putting
in
a
ton
of
so
you
can
see
sort
of
the
bugging
stack
here,
see
what
you
can
see
and
this
kind
of
ties
into
transport
discovery
stuff,
there's
a
this
is
sort
of
the
protocol
level
and
then
there's
a
tax
level,
and
I
don't
actually
print
out
function,
traces
for
the
oh
there.
It
is
no.
I
do
actually
my
mistake
so
then,
of
course,
there's
the
the
application
level.
K
K
The
last
thing
I'll
show
you
about
the
taps:
interface,
the
taps,
application
interface,
because
you
guys,
you
know
this
is
not
necessarily
all
that
new
revolutionary
is.
This
is
kind
of
what
the
the
include
looks
like
again.
K
This
is
all
written
in
vanilla
c,
so
there's
a
bunch
of
callback
function,
pointers
and
then
because
this
is
pure,
vanilla,
c,
there's
sort
of
a
a
manual
object-oriented
interface,
where,
like
you,
you
have
a
handle
that
you
pass
to
taps
to
for
whatever
object,
you're
you're
looking
at
so
there's
the
endpoint
stuff,
there's
transport,
property,
stuff
and
you'll
see
that
really
there's
very
few
things
that
work
right
now
completely,
but
the
listener
works,
pre-connection,
stuff,
listener,
stuff,
message,
stuff
connection,
stuff,
all
right,
so
that's
that
by
the
way,
here's
the
github
just
so
you
can
see
that
it
exists.
K
K
Okay,
so
now
the
on
the
transport
discovery
side,
what
I've
done
is
essentially
written
a
wrapper
for
tcp
sockets,
and
you
can
think
of
that
as
a
prototype
for
how
anywhere
these
wrappers
would
work.
You
could
also
have
a
native
protocol
implementation.
It
did
the
tap
stuff,
but
in
the
near
term,
we'll
probably
be
doing
these
wrappers
and
part
of
what
this
thing
does.
Is
it's
got,
and
then,
if
there
was
an
installer
would
create
a
yaml
file
that
that
looks
exactly
like
this.
K
The
udp
code
doesn't
actually
exist
yet
so
so,
if
you
ask
those
properties,
things
will
go
very
badly,
but
essentially
it's
describing.
What
is
the
protocol?
What
property
does
it
provide,
and
where
can
I
find
the
dynamic
library
for
this,
this
library
for
that
protocol
and
so
taps
will
then
read
that
in
and
then.
K
Certain
template
in
terms
of
callbacks
that
it
accepts
and
and
some
of
the
the
you
know,
the
basic
function
calls
so
you
see,
listen,
stop,
send
and
receive,
which
is
all
a
minute
at
this
time
and
then
of
course,
there's
a.
K
And
then
this
is
the
actual
implement.
This
is
the
actual
wrapper
of
tcp
sockets
that
that
handles
all
that.
So
you
know,
I
would
say
the
only
thing
I'd
say
about
this
implementation.
It
doesn't
make
an
effort
to
be
portable
and
scalable
like
it
does
use
lib
event,
which
is
I
understand
it
if
I'm
not
an
expert
on
a
live
event,
but
it's
a
relatively
portable
means
of
eventing.
K
You
don't
have
to
use
lib
event
in
your
application
or
in
your
protocol,
but
taps
it
selfies
that
it
assumes
that
and
it
has
got
it's
going
to
have
some
some
ways:
it'll
be
more
efficient
if
you
use
that
or
there's
certainly
some
things
that
there's
certainly
the
flexibility
if
you're,
if
your
protocol
uses
something
else
to
use
that
instead
in
the
model
and
then
there's-
and
you
know,
there's
things
like
it
takes-
it
takes
iovex.
K
If
you
want
to
use
them,
for
you
know
zero
copy,
and
that
kind
of
thing,
so
it
there's
a
lot
of
stubs.
There's
a
lot
there's
a
lot.
It
hasn't
been
done
for
sure
on
both
ends,
in
particular
the
one
piece,
the
architecture
that
is
really
missing
is
there
should
be
a
taps
d
that
is
reading
these
yaml
files
and
loading
all
the
protocol,
information
into
shared
memory,
and
but
right
now,
I'm
doing
it
on
the
pre-connection,
which
is
probably
not
the
best
thing
to
do.
K
Computationally,
that's
probably
the
biggest
architectural
piece
that
is
missing,
but
other
than
that.
All
of
all
the
pieces
are
here,
really
needs
to
be
enriched.
So
certainly
I
would
welcome
really
anything
at
this
point
from
like
you
know,
your
your
syntax
is
all
wrong
and
you
should
use
a
different
different
thing
there
to
to
like
code
contributions
or
or
feature
ads
or
bug
fixes,
etc.
So
that's
that
I'm
happy
to
take
any
questions
about
this.
A
So
martin
did
you
I
I
maybe
I
faded,
and
you
mentioned
this:
did
you
implement
any
kind
of
end-to-end
verification
that
that
the
selective
protocol
works?
You
know
the
somewhere
between
probing
and
racing.
K
K
So
I
I
I
stubbed
that
I
kind
of
commented
out
a
lot
of
that
just
built
the
listener
just
so
there
would
be
something
so
right
now.
It
is
actually
it's
going
through
the
list
of
protocols
which
is
pretty
small,
but
all
the
stuff
exists
to
like
look
at
the
properties
that
are
requested,
look
to
the
properties
that
are
presented
and
then
kind
of
take
a
score
for
that,
and
then
it's
ignoring
that
score
and
just
picking
the
first
one
right
now,
but
there's
some
stuff
there.
K
I
think,
ultimately,
this
this
taps
d
thing.
One
reason
it
should
be
a
damon
is
that's
where
all
the
security
stuff
would
live
to
verify
that
these
things
work,
you
could
even
sort
of
test
the
library
output
at
a
maximum
and
like
verify
the
claims
of
the
protocol.
That
would
be
a
useful
thing
that
I've
not
even
started
to
think
about
coding,
but
it
would
be
a
neat
thing
to
do.
Does
that
answer
your
question.
A
It's
just
a
high
level
question
of:
did
you
implement
it,
but
I'm
not
sure
that
you
fully
understood
the
question
that
you
know
when
we,
when
we
chartered
caps
the
sort
of
the
big
sort
of
researchy
area,
in
my
mind,
was
around
you
know:
how
do
you,
how
do
you
select
if
you've
got
multiple
protocols
that
can
that
can
work,
and
so
there's
a
the
probing
and
racing
piece
and
that
there
was
going
to
be
some
logic,
and
I
expected
that
there
was
going
to
be
some
experimentation
needed
to
sort
of
figure
out
on
what
kind
of
heuristics
do
you
use
to
verify
that
something
is
working?
K
So
there
are
some
hooks
for
racing.
It
does
not
currently
race.
The
thing
that
is
most
developed
in
this
code
is
an
effort
to
sort
of
score.
The
various
implementations
against
the
requested
properties.
A
lot
of
that
machinery
is
there
it's
currently
being
ignored,
but
that
is
there
and
yeah
I
mean
the
intent
is
eventually
implement
racing,
but
it's
it's
not.
That
doesn't
really
exist
right
now.
N
K
G
G
Really
helpful
implementation,
if
you
had
just
something
slightly
more
than
a
test
workload
to
work
with
so
you're
sure
you're
hitting
reality
more
often.
K
Yeah,
that
would
be
nice,
so
my
my
like
my
corporate
justification
for
working
on
this
is
that
I
I
think
we
believe
that
it
would
be
good
for
for
for
the
ecosystem,
for
our
applications
to
have
these
plot
to
make
things
more
platform
agnostic.
K
So
it
is
not
so
much
that
we
are
pushing
a
particular
application
as
just
the
the
idea
that
this
would
be
present
that
if
this
was
attractive
enough
to
exist
on
most
linux
boxes,
that
would
be
good
for
us.
I
could
look
at
something
like
engine
x,
but
that
is
very
complicated
and
I
don't
I
don't
think
we're
ready
at
this
point
to
to
to
go
there,
but
that
that
actually
good
advice,
and
once
this
is
in
a
little
bit
more
a
little
bit
less
a
rickety.
C
G
Because
quickly,
just
like
hands
out
the
socket,
so
if
you
can
deal
with
like
the
tap
side
of
things,
so
you
could.
You
could
probably
look
at
some
of
the
quick
implementations
and
see
I
mean
there's
like
five
or
six.
You
could
probably
do
a
quick
survey
and
figure
out
which
one
is
the
thinnest,
but
you
need
to
implement
the
udp
stuff.
First.
L
A
You
know
if
we
ever
see
things
like
udp
being
blocked
as
a
security
measure
of
having
something
that
detects
that
and
fails
over
the
tpp
so
now
you're
using
like
really
widely
available
protocols
and
just
looking
for
does
it
work.
End-To-End.
K
K
The
transport
discovery
stuff
and
number
two
again
build
something
that
maybe
is
the
root
of
something
that
would
be
production
quality
at
some
point
so,
like
you
know,
there's
a
lot
of
work
to
do
just
to
get
this
up
to
a
minimum
level,
functionality
which
we
are
not
really
at
yet,
and
that
has
to
be
the
priority,
but-
and
I
also
would
really
like
to
wrap
open
ssl
in
particular
for
various
reasons,
but
obviously,
contributions
are
welcome
and
and
kind
of
suggested.
K
Steering
directions
are
welcome
and
I
think,
there's
a
longer
term
term
goal.
Tom's
suggestion
of
having
a
target
application
real
application
is,
is
quite
attractive.
Regarding
quick
implementations,
I'm
not
really
100
sure
if
people
are
suggesting
that
we
run
the
quick
application
on
top
of
this
or
we
put
it
underneath
we
built
a
wrapper
for
it
to
be
a
a
like.
An
available
protocol.
K
K
I
would
hope
that
this
you
know
as
this
picked
up
like
if
it
did
really
take
off
a
little
bit,
then
people
would
be
motivated
to
wrap
various
things
and
I've
tried
to
make
it
easy
like
this
has
obviously
been
written
with
an
eye
towards
making
it
clean
for
a
protocol
developer
to
plug
into
this
architecture
without
having
to
do
major
surgery
to
their
software
and,
like
I
said,
people
are
interested
to
look
at
the
markdown
docs.
K
That
kind
of
see
what
that
looks
like
and
look
at
the
tcp
implementation
or
wrapper,
such
as
it
is
to
get
a
sense
of
how
that
might
what
that
might
look
like
and
I'm
I
think,
I'm
over
time.
So
I'm
happy
to
take
other
questions,
and
certainly
you
know
email
if
people
want
to
get
involved
in
this.
A
Before
we
move
on
completely
martin,
how
did
you
find
the
usefulness
of
the
the
implementation
document?
Was
it
helpful
for
you.
K
Yeah,
so
actually
you
raised
that
ua
is
kind
of
like
the
third
thing
that
was
out
of
this
was
kind
of
an
effort
to
implement
the
docs
as
they
are
from
from
scratch.
I
mostly
use
the
api
draft,
the
the
interface
draft,
that
was
by
far
the
most
useful
to
me.
The
implementation
draft
was
most
useful
for
trying
to
figure
out
all
the
pre-connection
and
racing
stuff,
at
least
that's
where
I
went
to
try
to
figure
out
how
to
do
it.
K
I
I
mean,
as
I
kind
of
I
could
I
put
that
work
on
hold,
so
you
know
there's
that
I
found
it
kind
of
confusing
and
a
little
little
heavyweight.
Frankly,
I
think
I
found
some
comments
in
this
area.
K
You
know
there
was
already
issue
to
this
effect,
but
I
was
kind
of
the
thing
that
I
kind
of
kind
of
surprised
me
when
I
on
this
implementation
effort.
I
didn't
appreciate
it
until
I
started
writing
code
was
how
little
the
preconnection
does.
N
K
I
didn't
appreciate
that
until
I
read
the
thing
really
carefully,
it
kind
of
worked
out
the
problem
code
so
that
yeah
there
were
a
couple
of
issues
filed
in
that
in
that
space,
that
was
kind
of
the
main
thing
and
I've
been
I've
been
still.
I've
actually
still
wrestled
a
little
bit
with
like
the
notion
of
adapting
tcp
to
this
notion
of
the
like
the
message
construct
in
in
taps
and
there's
not
like.
K
K
Although
I
haven't
completely
implemented
a
lot
of
eom
stuff
and
all
that,
because
it's
not
really
necessary
for
what
I
have,
but
it's
not
like
100
clear
how
those
things
map
and-
and
I
think
I
understand
now
where
like
there's
a
message-
context
which
is
constant
but
that
the
like
the
actual
buffer,
the
message
data
changes,
each
receive
partial,
that's
still
very
kind
of
not
clean
and-
and
I'm
not
100
sure
how
I'm
going
to
do
that
properly.
Once
once,
the
machinery
exists.
C
C
We
think
that
networking
is
held
back
by
the
fact
that
the
tcp
stack
is
basically
an
interface
designed
to
move
a
tape
across
a
room
and
see
right
and
so
like
we
tried
to
go
as
far
as
we
could
in
the
other
direction,
just
to
make
sure
we
weren't
stuck
in
the,
and
it
is
interesting
to
see
that
you
were-
I
mean,
of
course
you
can
implement
it
and
see.
C
Obviously,
but
it's
it's
interesting
to
see
how
you
implement
it
and
seeing
that
it's
actually,
it
doesn't
look
that
much
dirtier
than
any
I'm
in
c,
and
I
need
object,
oriented
concepts,
sort
of
sort
of
thing
right,
so
yeah.
K
Yeah
so
like
there
are
many
reasons
for
doing
it.
In
c
I
you
know,
our
data
plane
is
in
c,
and
so
I'm
used
to
trying
to
pound
object,
oriented
pegs
into
into
into
non-objective
holes,
but
I
mean
yeah
so
like
obviously
it's
kind
of
like
the
the
base
thing
to
have
it's
sort
of
like
the
most
fundamental
thing
to
do.
It
allows
you
know,
there's
some
performance
advantages.
Of
course
you
can
always
build
stuff,
on
top
of
it,
that's
a
little
less,
a
little
more
user-friendly
and
then.
K
Lastly,
I
have
a
like
a
not
well
developed
idea
that,
like
in
the
very
very
long
run,
this
might
become
a
thing
in
the
kernel,
in
which
case
doing
it
in
c
is
valuable
again,
not
a
very
well
developed
thought,
and
maybe
there's
some
huge
problems
that
I'm
not
seeing
here,
but
anyway,
those
are
the
and
and
fourth
and
and
not
not
unimportantly,
on
probably
most
fluent
in
vanilla
c
today,
because
that's
where
I've
been
doing
work
for
other
purposes
last
few
years,
so
that's
yeah!
A
Well,
thanks,
martin,
what's
the
what's
the
next,
what's
the
next
milestone,
we
should
check
in
again
on
this.
You
do
you
have
plans
like
specifically.
K
Yeah
well,
like
I
mean
there's
some
issues,
there's
some
obvious
things
to
do
like
there's
some.
I
would
like
to
have
well
right
now
it
like
you
can
actually
break
it
pretty
easily
by
like
it'll
pick
udp
and
he's
not
there
and
stuff
like
that.
So
I
want
to
make
it
like
not
break
like
eliminate
those
sorts
of
rough
edges.
Fill
in
like
allow
you
a
lot.
You
know
let
people
initiate
connections
that
kind
of
basic
thing,
and
you
know,
there's
there's
a
lot
of
directions.
K
I
want
to
go
with
this.
Obviously,
if
there
are
people
who
want
to
contribute,
that
would
be
if
there's
interest
in
a
particular
thing
that
would
that
would
certainly
you
know,
people
who
are
going
to
put
in
some
code
or
get
a
lot
of
votes
and
how
this
is
going
to
go.
K
It's
target
rich
environment
for
sure
I
would
I
would
like
to,
but
beside
after
that,
like
basic
fragility
is
addressed,
I
would
probably
like
to
pursue
the
transport
discovery
stuff,
maybe
put
in,
like
I
mentioned
to
tapsd,
because
that's
it
seems
like
that's
where
this
is
unique
relative
to
the
other
coding
efforts
in
particular,
but
yeah
like
I
said,
lots
of
people
get
a
vote
if
they're
willing
to
help.
K
A
So
I
guess
the
next
next
agenda
item
is,
you
know
going
to
the
github.
I
first
amaya
culpa
this
week
kind
of
exploded
on
me
in
terms
of
busyness,
and
so
I'm
I
am
behind.
On
my
email.
I
saw
that
we
got
a
couple
of
gen
art
reviews
from
robert
sparks.
A
Have
those
been
ported
into
the
issue?
Tracker?
Okay?
So
that's
what
that
flurry
of
issues
were
that
that
I
saw
come
out
yeah.
Do
you
teresa?
Do
you
know?
Has
there
been
any
activity
in
the
sector
or
what
was
the
other
one?
Was
it
the
art
yeah.
B
At
least
I
didn't
see
it,
let
me
check
back
what
it
says
now
yeah.
It
says
secure
review,
but
it
also
says
august,
which
I
mean
art
art
actually
pushed
their
due
date
by
a
month
and
then
they
actually
did
it
section,
not
sure,
but
I
think
they
had.
B
I
think
they
had
an
assigned
reviewer.
Actually
oh
yeah,
who
has
also
accepted
that
assignment,
but
has
not
completed
the
review.
Okay,.
L
Yeah,
the
review
of
from
robert
was
really
detailed.
I
was
really
impressed
by
that.
C
Yeah
I
was
I
was
I
looked
at
the
review
and
there's
very
little
of
that-
that
I
don't
think
we
have
to
address
with
basically
actually
either
texting
the
document
to
clarify
things
or
going
back
and
thinking
about
stuff.
So
it
will
be
like
it's
kind
of
annoying,
because
it's
a
lot
of
work.
We're
about
to
hit
issue
number
one
thousand
on
these
on
these
documents.
C
C
So
I'm
not
sure
what
the
the
most
productive
way
to
do.
This
is
because
I
have
also
like
beyond
looking
at
the
thing
and
going
that's
a
long
email,
and
then
you
know
just
skimming
it
to
see
what
you
know
if
there
was
actually
going
to
be
work
out
of
this,
I
have
not
had
a
chance
to
really
do
anything
since
the
last
interim
you'll
see.
I
still
have
a
few
things
that
are
assigned
to
me
that
have
that
I
haven't
done
yet.
Oh
wait.
C
I
do
have
at
least
well
back
in
february.
I
do
have
at
least
one
pr,
so
I
would
say,
let's
probably
go
to
the
pr's
first
get
those
merged
and
then
spend
a
little
bit
of
time.
Signing
tommy
up
to
do
all
of
the
changes
that
robert
asked
for,
or
at
least
figure
out
who's
the
right
person
to
do
some
of
these
things
right.
J
Small
editorial
ones,
so
there
are
quite
many
that
I
picked
as
I
just
saw
it's
a
small
paragraph.
One
sentence
I
didn't
want
to
miss
any,
but
many
of
these
issues
are
just
remove
a
sentence
here
fix
this
sentence.
There
yeah
fix
that
paragraph.
So
I
mean
this
this
I
can.
J
L
C
Think
that's
good
cool
by
the
way.
This
is
awesome.
You
can
actually
tell
meat
echo
just
to
share
a
window,
so
this
is
like
yet
another
vote
for
continuing
in
meet
echo.
I
think
kyle's
855
were
probably
going.
You
know
increasing
consistency
of
formatting
spelling.
I
think
that
one
is
going
to
go
into
merge
conflict
hell
soon.
We
should
possibly
restart
that
one.
Let's
look
multicast
example
code
corey
looked
at
it,
tommy
fixed
it.
C
I
would
like
you
to
take
a
look
at
this
one,
having
recently
done
a
tsv
art
review
that
was
very
multicast
heavy.
I'm
like
I
have
learned
how
much
I
don't
know
about
multicast.
So
look.
D
C
Gory
is
also
not
here
correct.
I
think
we
used
all
of
our
gory
time
with
the
73
issues
that
he.
L
L
J
J
That
makes
a
difference,
but
I
can't
take
the
api
editorials.
I
wonder
if
I
can
take
them
all,
but.
M
If
there
is
some
you
think
I
should
handle,
I
can
also
assign
them
to
me.
C
There
are
a
few
non-editorial
issues
on
architecture,
and
it
seems
like
at
least
one
of
these
or
the
separation
of
architecture
and
interface
documents.
One
is,
it
seems
like
we
should
have
the
opinion
as
a
working
group,
so
I
would
say
we
start
with
the
art
art
review
architecture
like
everything
up
to
this
already
has
someone
assigned
to
it.
At
least
I
would
maybe
start
with
8.79
the
ones
that
are
not
marked
editorial
and
start
either
discussion
or
triage
out
to
people.
Does
that
sound
like
a
good
way
to
go.
N
O
C
Right
like
so,
I
I
think,
that's
that's
the
ask
is
go
through
here
and
say:
okay,
the
normative
thing
is
the
api
and
architecture.
Just
basically
gives
you
an
introduction
to
the
concepts
to
understand
how
the
api
works,
which
is
not
what
we
set
out
to
do,
but
I
think
that
one
of
the
reasons
that
it
wasn't
what
we
set
up,
I
think
there
are
two
reasons
that
surprise
and
whatever
are
our
weapons.
C
I
think
there
are
three
reasons
that
we
started
to
do
it
that
way
reason
number
one
is
just
straight
up
from
the
way
the
charter
is
is
put
together.
C
The
architecture
might
have
independent
in
a
world
where
we
didn't
get
all
the
way
to
finishing
the
api
dock,
which
is
not
the
world
we
live
in.
The
architecture
might
have
had
independent
utility
right
in
terms
of
like
influencing
people
to
shape
their
apis.
Like
this,
because
I
mean
the
charter
basically
said
hey,
we
can
do
some
api
work
in
order
to
to
experiment
with
these
concepts,
but
not
necessarily,
you
know
to
define
the
future
interface
to
the
transport
layer.
I
think
we
got
farther
than
that
right.
C
The
second
reason
to
do
that
is
that
we
wanted
to
finish
the
architecture
document
first
and
say
these
are
the
things
that
we
agree
with
and
basically,
if
we
end
up
making
changes
to
the
api
dock
that
don't
fit
into
that
architecture.
Well,
we
shouldn't
do
that
and
that's
not
actually
how
we've
ended
up
working.
The
way
that
we've
ended
up
working
is
we've
now
gotten
to
the
point
where
most
of
the
concepts
are
being
built
in
the
api
dock
first
and
then
back
ported
to
architecture.
C
If
architecture
is
disagrees
with
it
right,
it's
like
it's,
okay!
Well,
if
the
api
says
this,
the
api
is
what
we
want.
Therefore,
the
architecture
should
bend
to
the
api.
We
haven't
done
much
of
this,
but
like
a
little
bit
like
the
last
one,
was
connection
grouping,
we
had
to
go
back
and
rephrase
connection
grouping
in
the
architecture,
so
I
can
absolutely
see
why
you
know
robert
coming
to
this.
C
With
you
know
a
background
and
a
whole
lot
of
you
know
the
internet,
but
coming
to
this
document
for
the
first
time
might
think
that
this
looks
weird
right
because,
like
this
separation
between
architecture
and
api,
is
I'm
going
to
say
obsolete,
but
it's
certainly
historical.
It
certainly
reflects
the
state
of
these
documents
a
long
time
ago.
C
My
counter
argument,
my
argument
to
not
doing
this
would
be.
I
don't
wanna,
it's
a
bunch
of
work.
I
mean
it,
but
it's,
but
I'm
not
like.
L
So
of.
C
L
L
A
I
was
going
to
ask
the
same
question
that
I
mean
paraphrasing
what
I
think
brian
said
was.
The
separation
was
helpful
for
our
process,
but
may
not
be
as
important
to
the
the
reader
coming
in
after
the
fact.
A
And
so
it
does
raise
the
question
about
whether
you
know
is
there
value
in
having
two
documents?
Could
you
could
a
new
reader
skip
the
architecture
document
and
pick
up
interface
and
make
use
of
it.
J
J
C
C
I
mean.
Maybe
the
subsections
of
section
4
right,
like
so
4,
4,
1
and
4
2
could
move
into
the
interface
document
or
401
could
move
an
interface
and
42
could
move
into
implementation
right.
C
M
Okay,
then,
I
turn
up
the
level
a
little
bit
better.
Yes,
that
helps
so,
I
think
for
two:
we
can't
move
into
the
implementation,
because
it's
really
the
background
on
how
we
think
about
racing
and
how
we
look
at
in
into
the
general
concept,
how
everything
falls
together
so
about
for
one
moving
to
api.
This
could
be
done,
but
then
I
think.
J
F
M
M
But
the
still,
I
think
this
is
a
discussion
we
maybe
want
to
put
up
on
the
mailing
list
and
get
more
opinions
on
what
does
it
mean
for
people
reading
it,
and
is
it
really
helpful
to
have
a
shorter
architecture
with
much
longer
api.
C
Well,
I
think,
from
a
from
a
a
document
user
standpoint
right
like-
and
this
is
the
other
thing
is
this-
is
the
second
paragraph
here
right
like
consider
reformulating
this
as
an
overview
right
like
so
we
start
with
the
principles
like
so
that.
C
He's
right
that
this
document
is
basically
is
the
way
that
it
is
because
of
the
process
that
we
use
to
get
to
it.
But
if
we
take
a
pretty
hard
edit
pass
at
arch
and
say,
okay,
we're
going
to
make
this
an
overview
introduction
in
principles
right
like
so,
and
you
can
even
keep
the
you
know
the
short
name
of
the
document
and
call
it
sorry.
F
C
C
Belong
in
the
implementation
document,
correct
or
in
the
api
document,
if
they,
if
like
non-implementation
of
them,
is.
C
Going
to
lead
to
interoperability
issues,
so
I
think
the
first
thing
we
could
do
on
this
is
take
all
the
normative
language
out
right
and
then
the
second
thing
we
could
do
that
would
be
easy,
would
be
consider
renaming
the
doc.
J
J
J
D
D
C
F
D
Now
I
mean,
I
think
I
I'm
thinking
like
this,
I
mean.
Obviously
we
are
pretty
expert
on
doing
things
and
taps
right
now,
but
I
mean
these
documents
should
also
tell
us
more
about.
Like
I
mean
it
does
tell
us
about
the
how
taps
works
and
how
things
are
expected
and
things.
Obviously
it
puts
some
requirements
on
the
rest
of
the
things
that
we
do,
but
it
is
a
good
intro
so
to
say
not
yet
aim
for
it,
but
it
does
keep
some.
K
So
when
I
went
through
this
implementation
exercise,
I
did
not
really
use
this
document.
I
did
have
the
benefit
of
being
in
this
working
group
and
just
picking
up
the
architecture
through
osmosis.
K
So-
and
I
will
say,
like
the
one
thing
that's
been
very
useful
for
the
architecture
document
is
figure,
one
in
figure,
two,
which
explains
the
the
it's
a
diagram
of
the
architecture,
which
has
been
very
useful
for
slidewear
about
what
how
I'm
doing
in
taps.
Yes,.
C
Well-
and
this
is
like
also
figure-
three
and
figure
four
have
been
insanely
useful
for
slidewear
as
well
right,
it's
like
the
in
terms
of
like
what
is
a
connection
object
and
how
is
this
slightly
different
having
this
lifetime
thing
here
like
I,
I
think
that
I
think
that
robert's
onto
something,
I
think
that
this
document
is
almost
a
really
good
introduction
to
the
high-level
concepts,
but
it
says
in
its
abstract
that
it's
an
architecture,
definition
and
the
point
that
he's
making
with
like
the
first
sentence
of
this
review
and
the
first
sentence
of
the
interface
review,
that
michael,
that
you
added
here,
is
the
normative
architecture.
C
Lean
into
this
document
being
published
as
the
hey,
you
would
like
to
understand
enough
about
taps
to
write
slider
about
it
right,
which
is
actually
an
important
thing,
for
you
know
early
adoption
of
an
effort
to
implement
something
like
this.
We
turn
that
into
this
document,
and
then
it's
basically,
you
know
we
rename
it
to
an
introduction
to
the
transport
services
architecture.
C
And
I
think
that
like
having
having
having
a
non-mutable
sort
of
like
thing
that
says,
here's
the
goal,
here's
what
we
set
out
to
do,
I
think
that's
important
as
well.
So
I
think
that
basically,
this
is
a
there's
a
not
very
much
to
do.
C
There's
a
this
is
less
work
way
to
to
address
this
right,
which
fixes
my
I
don't
want
a
problem
where
we
basically
then
do
a
pass
over
this
document
and
say:
okay,
we're
gonna,
we're
gonna,
make
sure
this
document
is
useful
as
an
introduction
to
the
architecture.
Does
that
seem
reason
a
reasonable
thing
to
put
time
into.
O
D
A
Yeah
that
that,
well,
that
the
that
it
really
isn't
there
isn't
a
hard
dependency
on
any
information.
That's
in
the
architecture,
documents.
D
M
M
B
B
C
C
Entities
within
the
architecture
capitalized
do
we
want
to
keep
that
as
a
as
a
as
a
convention,
because
if
you
we,
I
mean
it's
same
as
672
right
like
there's
the
post
yeah,
it
just
needs
to
be
consistent.
C
B
C
So
we
know
we
need
to
do
that,
pass,
we
haven't
done
it
yet.
Then
we
have
check
using
normative
text,
which
is
something
that
we're
going
to
do
in
architecture
as
part
of
this
separation
architecture
interface
document.
So
those
will
get.
A
A
I
C
C
Yeah
it's
supposed
to
be
down
here,
I'm
I'm,
I'm
not
sure
what
happened
here.
Let's
see
hold
on
taps
arch.
Let's
go
to
the
code.
C
I'm
not
sure
how
this
got
added
here:
let's
get
it
to
the
wrong
place.
O
A
C
So
that
is
all
of
the
non
editorial
commentary
on.
C
J
J
L
C
So
one
way
to
fix
this
in
the
api
would
be
to
to
get
rid
of
all
of
our
types
right,
which
I
think
yeah,
which
would
cause
implementers
to
have
exactly
the
look
that
martin
has
on
his
face
right
now.
So
we
probably
shouldn't
do
that.
I
don't
know
I
I
don't
know
if
you're
following
him,
a
look
came
across
your
face
was
like
get
rid
of
types
you're
like,
so
we
probably
shouldn't
do
that.
The.
I
J
I
mean
there
were
a
couple
of
cases
where
I
probably
had
something
like
minus
one.
There's
an
error.
A
card
rightly
complained
that
this
is
very
c
yeah,
and
so
we
said:
okay,
there
are
these
special
defined
error
values
because
still
it's
an
integer
otherwise,
so
that
is
ugly,
I
mean
well
special
yeah,
I
mean
I.
I
J
C
C
All
it
is
yeah
yeah.
M
Model
number
making
it
consistent
there
and
full
coverage
equals
infinity.
A
C
J
B
J
J
Has
the
same
timeout
for
keeper
lives?
Has
it
pounds
on
send
and
receive
rate?
Has
it
numeric,
especially
when
you
unlimited.
L
Can
we
just
copy
the
rust
approach
and
define
a
type
which
is
a
result
which
is
either
an
error
code
or
or
a
whatever,
and
then
just
say
this?
This
returns
a
result
which
is.
C
J
L
We
can
do
it
like
that
or
we
can
do
the
the
way
you
do
it
in
a
typed
language
and
say
actually
it's
it's
an
enumeration
of
this
type
or
this
type
and
put
a
note
in
the
introduction
that
says
we
use
this
type
and
if,
if
you're
implementing
it
in
c,
it
probably
turns
into
a
magic
error
code,
but
you
wanna
name.
It.
C
I
would
use
the
I
would
use
a
type
enumeration
instead,
it's
ugly
but
like
there
would
then
be
maybe
guidance
in
implementation
that
hey,
if
you're
doing
this
in
a
language
that
doesn't
use
type
enumerations,
then
do
something
like
from
a
type
theoretic
standpoint.
This
is
actually
a
type
of
numeration.
C
J
M
J
I
I
The
same
complaint
about
yeah
separation
of
documents.
C
M
C
C
All
these
blue
cycles
yeah,
but
it's
going
to
be
like
one
pr,
that's
going
to
close
yeah
five
issues,
so
it's
gonna,
be
it's
gonna,
be
super
satisfying
to
land
that
one
that
is
wrong.
It's
going
to
remove
many
small
things,
transport
property
names
in
4.1.
I
think
you
need
to
be
more
specific
than
alphanumeric.
I
don't.
J
C
So
there's
like
this
was
this
actually
goes
back
to
this
discussion,
and
I
think
I
was
in
the
minority
here
and
thinking
that
this
was
not
a
super
necessary
thing
to
do,
if
only
because
you're
going
to
end
up
with
a
bunch
of
names
that
are
not
going
to
end
up
being
partial
as
identifiers
in
any
in
every
language
that
you
might
want
to
to
implement
right
like
so,
you
know
if,
if
you're
trying
to
get
rid
of
delimiters,
then
you
can't
use
dashes
and
you
possibly
can't
use
underscores.
C
N
M
Double
check
with
the
underscore
what
we
did
there.
I
think
we
had
the
special
case
that
we
wanted
the
other
underscore
to
automate
it
automatically
mutate
to
whatever
concept
you
had
as
word
separators
in
your
language,.
C
I
mean
we
do
specify
already
in
the
api
dock
that,
like
hey,
you
should
make
this
look
like
you
know.
You
know
you
should
make
this
look
like
your
language,
because
otherwise
people
are
going
to
be
like.
Oh,
this
is
some
weird
java
stuff.
I
don't
want
that
or
this
is
some
weird
python
stuff.
I
don't
want
that.
L
C
Right
this
was,
we
decided,
we
didn't
want
to
do
an
iana
registry,
because
it's
a
lot
of
work
and
we
come
back
and
do
that
later.
If
it
turned
out,
we
needed
it.
L
M
So
we
wanted-
and
this
was
something
we
discussed
with
gory
quite
intensely-
had
to
have
something
that
can
automatically
translate
across
implementations.
L
J
So
I
I
I
like
having
the
same
names
for
things
actually,
so
I'm
not
such
a
fan
of
honda
spy
on
the
specifying,
because
then
you
know
I
imagine
to
two
implementers
talking
to
each
other
and
they
can
talk
about
things
like
the
capacity
profile
and
the
parameter
called
something
and
property
called
something
it's
the
same
name,
so
I
think
I
mean
without
having
to
go
into
capitalization
or
underscore
or
whatever,
but
at
least
having
the
same
names
for
things
I
mean
that
right.
You
know,
that
is
that
that
has
value.
C
Right
all
of
these,
and
each
of
these
translation
things
even
in
a
world
where
people
are
not
being
100
consistent.
These
are
these
are
regular
expressions
that
will
parse
down
into
a
finite
state
automaton
that
can
fit
into
a
tiny
amount
of
silicon
right
like
so.
It's
like
right,
using
reg
x's
to
bash
from
one
one
context
into
another,
is
like
maybe
not
super
pretty,
but
it's
I'd
hazard,
a
guess
that
it's
been
at
least
15
of
everyone's
job
in
this
room.
At
some
point,
it's
computers.
C
What
we
do
I'm
a
fan
of
underspecifying
here,
but
I
understand
that,
like
that,
like
underspecifying,
makes
certain
people
twitchy,
including
probably
robert,
actually
our
art,
art
reviewer.
But
I
say
I
would
leave
this
up
to
michael
as
a
as
a
how
much
time
he
wants
to
spend
on
it.
Executive
decision.
C
Do
we
need
host
or
service?
Actually
we
mean
service.
C
C
F
C
It
could
be,
there
might
be
some
application
level
semantics
that
you
know
when
you're
setting
up
all
of
these
these
remote
endpoints.
You
know
that
they're
the
same
there
might
be
multiple
front
ends
like
transport
front
ends
that
will
go
to
the
same
application
back
end
right
right.
The
point
we
want
to
make
here
is
like
you,
shouldn't
have
two
remote
endpoints,
one
of
which
is
a
web
server
and
one
of
which
is
a
mail
server,
because
then
you're
going
to
have
a
bad
day,
yeah.
O
C
M
I
think
we
ended
up
and
having
multiple
end
points,
you
can
specify
our
set
of
endpoints.
We
can
specify
just
to
make
sure
that
you're
able
to
manually
specify
candidate
lists.
L
Does
it
work
for
these?
Well,
I
suppose
v4,
but
for
v6,
where
you
can
have
multiple
ipv6
addresses
for
the
same
right.
L
I
J
B
O
C
Well,
I
mean
the
point
is
the
point
that
robert
is
making
here:
is
that
it's
inconsistent
right
like
if
you
can
have
a
name
and
an
ip
address?
Well,
what
do
you
do
with
that?
If
the
name
resolves
to
something
other
than
the
ip
address
right,
and
I
think
this
is
one
of
those
things
where
underspecified
ambiguity
is
probably
okay
and
like.
Given
that
there's
not
much
clarity
on
this
sentence,
we
can
just
delete
it.
I
think
that's
the
right
thing
to
do
so.
C
L
C
Why
does
the
api
have
both
the
ability
to
say
local
specifier.with
interface
and
to
set
interface
requirements
on
transport
properties?
Why
not
just
keep
the
ladder
and
remove.
J
M
So
if
you're
implementing
stuff
like
link
local
or
multicast,
you
really
have
to
specify
the
interface
as
especially
firefighter
with
system
local.
You
need
that
yeah!
Well,
you
can
do
that
as
a
property
on
the
pre-connection
right.
It's
just
a
syntactical
thing,
but
that's
weird
having
having
a
linked
local
address
on
the
on
the
end
point
and
then
specify
the
interface
on
a
pre-connection
is
weird.
H
C
So
I
really
see
the
pre-connection
thing
is,
is
like
kind
of
a.
It
would
be
something
that
you
could
have
like
a
policy
thing
right
like
saying
hey,
I
want
to
bind
this
application
only
to.
C
The
wireless
interface,
the
wired
interface
or
whatever,
whereas
the
local
specifier,
is
really
something
that
you
might
get
from
some
other
place
in
the
application.
K
C
C
You
might
want
to
have
in
your
application
a
different
user
interface
for
the
endpoint
specification,
as
opposed
to
the
user
interface
for
the
for
the
transport
property
specification.
You
might
not
even
bubble
up
the
transport
property
specification
to
the
user,
whereas
the
the
endpoint
you
would
bubble
up
to
the
user
right.
C
K
C
C
K
C
M
Yes,
we
you
have
to
re,
you
have
the
preference
in
the
in
the
properties.
You
have
the
preferences,
interface
and
sensor
type.
C
M
C
Well,
so
there's
there's
a
there's
like
martin
raises
a
really
good
issue.
Let's
say
that
your
transport
properties,
which
have
not
bubbled
up
the
ui,
have
even
just
a
non-uh
enumerated
list
of
interface
identifiers.
They
just
have
like
so
some
interface
properties
and
then
you
hand
a
local
specifier
to
the
same
free
connection
that
has
a
set
of
interfaces
that
do
not
meet
those
properties
boom.
You
can't
connect
yeah
right.
C
K
Yes,
this
is
like
I
mean
it's
more
code
and
it's
another
foot
gun
to
create
weird
failure,
cases
that
are
hard
to
debug
and,
like
I
would
like
what
gets
bubbled
up
the
user,
I
think
is
just
it's
neither
here
nor
there,
like
the
application,
is
gonna
need
a
knob
to
declare
interface
preferences
and
whether
that's
in
the
endpoint
of
the
transport
parameters.
K
I
mean
like
when
you
do
this
you're
going
to
have
you're
going
to
create
an
endpoint
you're,
going
to
create
transport,
transport
properties
and
you're
going
to
build
all
that
stuff.
I
mean
you
could
almost
have
it
be
one
giant
blob
in
principle,
although
yeah
I
mean
I,
I
really.
I
really
feel
strongly.
There
should
be
one,
and
I
don't
have
a
strong
opinion
about
the
like,
which
one
it
is
did
the
only
real
like
practical
difference
is
with
the
pre-connection.
K
M
Turns
up
to
the
following
conflict:
if
you
really
want
a
hard
specification,
it's
much
better
in
the
endpoint
specification.
M
If
you
want
really
express
a
preference
like
please
avoid
cellular,
then
it's
much
much
better
in
the
preference,
because
we
don't
have
the
preference
cons
in
the
ends
per
point
specifier-
and
this
is
the
conflict
we
have
to
tackle
here.
C
C
M
Yes,
but
that's
that's
not
up
to
the
flexibility
the
api
has,
so
what
you
could
do
is
you
could
specify
a
list
of
12
interfaces
as
explicit
identifiers,
yes
in
the
in
the
endpoint
and
then
have
preference
in
as
a
property
that
says
which
one
you
most
like.
C
In
the
remainder
of
this
call,
because
I
think
we
have
two
other
things
we
want
to
talk
about,
how
about
I
assign
you
and
martin
to
this
one
and
you
go
figure
it
out.
N
E
Yeah
I
mean
we
should
we
can
do
one
more
one.
Discussion
on
the
call
today.
C
J
L
D
C
L
C
L
C
I
did
that
for
a
while
at
the
beginning
of
the
pandemic,
and
then
I
realized
that,
like
so,
my
camera
is
a
fuji
xt2
which
is
like
you,
notoriously
it's
an
arm-based
computer,
strapped
to
a
heat
sink
in
the
shape
of
a
camera,
and
it
has
various
thermal
issues
and,
like
my
normal,
my
normal
day
of
like
four
to
six
hours
of
meetings,
it
couldn't
do
like.
L
Yeah
panasonic
have
one
which
will
run
all
day
every
day
without
heat
issues,
because
it's
designed
for
the
for
video,
like
this.
C
I
J
C
B
We
got
two
excellent
and
long
reviews,
I'm
not
sure
if
anybody
has
actually
gotten
back
to
robert
saying.
Thank
you.
These
were
excellent
reviews.
If
nobody
wants
to
do
it,
I'm
happy
to
just
send
you.
B
C
D
D
I
I
wanted
to
disgrace
this
question
because
we
have.
We
have
some
issues
with
this
github
review
process
and
I
was
one
of
the
thing
like
when
I
was
really
reviewing
giving
comments
to
http
documents.
They
actually
discussed
it,
they
closed
this
one
and
then
I
was
in
a
position
not
to
get
engaged.
So
the
question
here
is
basically
I'm
asking:
if
you're
taking
these
issues
or
from
the
review
command
is
creating
issues,
will
you
be
are?
Are
we
going
to
involve
the
reviewer
in
that
process?
D
Saying
like
saying
like
you
can
come
and
chat
with
us
at
github
many
author
reviewer
might
not
do
that
or
might
do
that,
but
at
least
we
need
to
update
them.
With
this
information
like
here,
I
have
created
issues
and
here
you
can
engage
and
if
that
doesn't
engage
then
basically
you
will
resolve
these
things
and
basically
send
another
note
like
here
what
we
have
done
so.
J
A
A
A
L
J
J
A
If
teresa
is
going
to
send
a
if
teresa
is
going
to
send
a
message
to
him
thanking
him,
we
can
just
add
a
note
onto
the
message
that
says
you
know
we
welcome
your
involvement
in
the
resolution
of
these
they're
all
logged
in
our
issue.
Tracker,
and
you
know
you
can
be
as
involved
as
you
want
to
be
in
subsequent
discussion,
and
you
know
I
think
that
that's
you
know
we,
I
I
it's
a
good
point.
A
We
want
to
open
the
door
and
make
it
clear
that,
if
he's
interested
in
seeing
if
any
particular
aspects
that
he
wants
to
see
how
they're
disposed
of
this
is
a
way
for
him
to
find
them
and
follow
them
express
opinions
on
them,
but
I
don't
think
we
want
to
block
on
it.
I
mean
I
think.
D
A
D
Don't
think
so
it's
just
information
to
so
that
I
mean
the
reviewer
feels
comfortable
about
it
and
anyway
you
will.
When
you
resolve
these
things,
then
basically
you
need
to
reply
to
his
comments
right
and
this
will.
E
D
A
lot
of
email
conversation
going
on
because
there
are
lots
of
issues
now
and
we
have
some
resolved
things,
and
maybe
this
is
unlike
pre
preview
of
like
this
is
what
we
are
going
to
do
and
you
can
get
engaged
or
you
can
wait
until
we
resolved
everything.
B
Yeah,
so
I
I
have
done
a
couple
of
reviews
on
documents
where
people
have
added
me
as
a
reviewer
to
the
pr
that
is
supposed
to
resolve
that
issue,
and
then
I
can
give
feedback
on
the
github
or
I
can
just
reply
through
the
email
saying
yeah.
It
looks
good
and
I
really
I
thought
that
was
really
useful,
especially
for
a
long
and
sort
of
documents
with
a
lot
of.
B
C
Is
rj
sparks
on
github,
so
I
mean
like
we
could
even
say
hey.
If
you'd
like
we
can,
we
can
add
you
to
the
to
the
pr's
that
will
that
will
fix
these
issues.
D
That
would
be
great,
I
mean
also
like
going
forward
when
you
will
have
this
sort
of
like
itf
review
or
icdb.
We
can
follow
the
same
procedure.
Basically,
this
problem
here
is
like
some
time
we
do.
We
have
some
reviews
and
I
don't
I
don't
use
github
and
then
basically
there's
a
different
thing,
but
if
you
have
we
have
a
reviewer
on
the
github,
I
mean
that's
a
very
good
way
of
engaging
them.
A
B
K
There's
an
iesg
retreat
at
that
time.
In
that
week
I
don't
have
details
on
when
exactly
yeah.
K
I'm
sorry
it'd
be
in
like
afternoon
utc
because
it's
a
mix
of
americans
and
europeans
so,
like
anything
else,
yeah
ietf.
It's
like
this
time
slot
something
like
that.
A
A
K
D
K
A
Is
it
an
ib?
Is
it
joint
receipt.
D
L
A
K
K
And
so
that's
it
right
so
that
week
is
the
workshop,
but
it
is,
it
starts
at
what
is
that
six
or
eighteen
hundred
utc?
So
if
we
did
it
at
the
normal
time,
it
wouldn't
conflict.
K
I'm
sorry,
seven
eighteen
hundred,
etc.
K
N
K
H
A
If
you
think
the
we
don't
have
all
the
istar
folks
who
would
be
invited
to
this
thing
on
the
call,
but
if
those
who
are
are
willing
to
to
have
a
long
day,
I
mean
if
it's
a
if
it's
a
you
know
a
full
day's
workshop
and
we're
adding
a
two-hour
working
group
meeting
on
the
front.
That's
it's
a.
C
K
L
A
So
I
I
think
the
20th
is
at
least
feasible
and
maybe
we
add
in
open
bar
just
to
soften
the
blow.
A
It's
I
I
think
it's
either
that
or
we
cancel
until
december,
because
I
don't
want
to
try
to
do
it
in
november
with
the
ietf,
unless
people
feel
strongly
otherwise
and
so
we're
gonna,
then
we
so
we're
and
then.
B
Monday
was
kind
of
bad
and
doodle
actually.
C
Have
the
one
technical
thing
they
still?
Let
me
do
at
google
is
mondays
from
5
7,
so.
A
All
right
mondays
from
friday:
seven,
it
is
tuesday.
We
want
to
help
you
with
your
transition,
brian,
the
full
pointy
head,
my
transition.
E
H
B
A
Friday
then
we
lose
philip
and
brian.
If
we
do
wednesday,
I
think
we
lose
phillip.
Is
there
anybody
else,
who's
blocked
on
the
22nd?
In
this
slide.
H
A
Let's
do
that,
it's
never
perfect!
I
I
think
that
maybe
after
the
holidays
in
january,
we
revisit
scheduling
again
make
sure
that
this
works
for
folk.
If
there's
a
lot
of
istar
things
that
fall
on
wednesdays,
maybe
wednesdays,
isn't
a
good
sustaining.
L
C
C
And
like
it'll,
be
150
entertaining
I'm
not
sure
how
productive
it'll
be.