►
From YouTube: IETF100-TRILL-20171113-1740
Description
TRILL meeting session at IETF100
2017/11/13 1740
https://datatracker.ietf.org/meeting/100/proceedings/
A
C
D
E
D
C
So
we
should
probably
we'll
start
with
the
administrivia.
Is
he
gonna
be
here
in
a
couple
minutes
we'll
go
ahead,
then
so
welcome
to
trill
hi
John
wave
to
John
on
the
on
the
monitor,
John
Hudson's
working
on
the
monitor,
hi
John
anyway.
Welcome
to
trill
we'll
go
ahead.
John
Donald
is
secretary
and
doing
lots
of
our
reviews
for
us.
I'm
sue
Harris
in
case
you're,
listening
remotely
on
this
recording
later
go
ahead.
This
is
a
note.
Well
can't
read
it.
We've
been
this
working
groups
been
really
good
about
putting
inside
PR
together.
C
So
thank
you
go
ahead.
Next
doctor
we
were
in
our
last
run
to
the
finish
line.
So
thank
you
all
for
reading
the
docs
Thank
You
Fenway
for
working
hard.
Please
help
with
joint
review.
We've
got
how
many
documents
Donald
will
come
to
the
list
so
guys
we've
got
a
lot
to
do
to
get
to
the
finish
line
everything's
about
there.
Much
of
this
has
gone
through
two
or
three
reps,
but
we've
got
to
push
through
to
get
through
a
whole
bunch
in
eight
weeks.
So
take
a
deep
breath
now.
C
C
C
C
Boring
we've
had
this
is
this:
is
our
last
sitting
around
the
table?
Well
in
it
sit
up
in
front,
don't
be
shy.
We're
gonna,
we're
gonna
have
a
bit
of
fun.
We
have
to
RFC's
published
the
MTU,
yes
and
the
ultimate
forum
for
multi-level
and
our
optimization
went
today.
Woohoo,
that's
it
that's
a
really
good
one.
I'll
go
ahead:
we've
got
two
in
publications
requested
state,
so
Ali's
got
two
new
ones,
and
McDonald's
and
I've
got
the
four
of
these
drafts.
C
B
C
I
H
C
Please
confirm
that
you
know
send
a
message
to
him
saying
please
I
made
the
changes
confirm
the
two
like
these,
because
last
time
he
confirmed
and
said
no
I
didn't
you
know
absolutely
they
weren't.
There
is
what
he
said.
Yes
did
I
miss
anything
else
on
the
BFD
okay,
this
is
a
tag-team,
don't
truck
in
working
through
plastic,
all
those
directory,
assisted
in
caps,
multi,
ecology
and
transport
over
MPLS.
So
please
send
a
comment
in
I.
C
C
These
are
the
drafts
not
listed
above
that
we're
going
to
go
through
in
the
next
eight
weeks.
So
we've
got
the
group
king,
the
single,
nicknamed
the
unique,
nickname
Donald's,
got
a
respect,
a
trill
over
IP,
oh
yeah.
We
were
going
to
take
that
out
of
that.
That's
a
lot
of
the
headache
there
and
we're
not
going
to
get
done
unless
we
D
function.
Are
we
ok
with
that
that
the
the
O
who
gave
us
the
excellent.
C
B
B
C
B
G
C
But
the
trill
data
encoding
and
the
vendor
Channel.
We
talked
about
and
split
out
from
other
drafts,
so
those
are
to
you.
We
were
thinking
because
Donald
had
taken
them
out
of
the
security
to
make
it
cleaner.
Those
who
are
gonna
try
to
adopt
this
next
week
and
pushed
it.
You
want
to
talk
yeah,
let's.
B
Talk
about
them
for
a
sec
are
they're,
pretty
mature.
The
vendor
channel.
One
is
really
just
a
simple
way
so
that
these
are
our
bridge
channel
messages
between
our
Burridge's
as
the
way
as
a
way
where
vendors
can
have
their
own
to
find
their
own
by
putting
a
vendor
o
UI
in
there.
So
I
figure
you
know
in
if
you,
if
the
working
group,
it's
not
continued,
then
there's
no,
it's
probably
harder
to
get
standardized
things
through
and
it
may
be
more
beneficial
to
provide
this
escape
valve
for
vendors.
It's
very
simple
draft.
B
B
B
It
dates
from
quite
a
while
ago
and
at
the
time
it
was
the
chip.
Vendors
were
saying
that
they
didn't
want
to
do
it
yet
because
they
wanted
to
see
how
things
were
going
with
drill
and
stuff.
So
as
far
as
I
know,
it
hasn't
been
implemented
in
hardware,
so
you
ve
to
do
usefully.
Do
it
for
most
applications.
That
drill
would
involve
fast
path,
implementation,
it's
simple,
but
so
I
don't
know.
C
B
D
C
C
C
We
need
to
get
this
stuff
through
so
from
now
to
the
end
of
December
is
at
six
weeks.
Hey
six
or
eight
in
six
weeks
is
the
best
way
to
get
all
of
this
done
so
that
we
can
go
ahead
and
then
spend
January
and
February
getting
it
through
this
ice
Jean.
So
we
really
need
to
have.
You
know,
go
home.
Tell
the
folks.
We
really
need
to
have
enough
when
we
do
the
calls.
It's
not
like
we've
done
it
in
the
past.
C
It'll
be
two
weeks
and
if
we
don't
get
enough,
we'll
have
to
go
on
without
it
I
think.
There's
a
Lea's
Dettol.
If,
if
you
go
back
to
what
I
have
one
more
no
other
way,
I'm
sorry,
my
bad
all
the
ones
that
are
in
working
who'd
last
call
and
that
at
least
got
seven
of
these-
that
she
could
already
have
going
forward
and
then
we're
going
to
add
another
six.
That's
a
real
lot
for
her
to
get
through.
So
we've
got
to
push
these
working
group
calls
so.
D
C
That's
small,
so
folks,
I'm
gonna
thousand
I'm,
probably
gonna,
send
you
lots
of
notes.
Like
please
review
an
answer.
We
got
our
optimization
I'm.
Sorry,
that's
a
cheer
for
all
the
hard
work
bit
you
put
in
so
folks.
We
really
have
six
weeks
and
a
lot
of
these.
These
lit
this
last
call
for
directory
and
tab
assistant
and
monthly
through
quality,
will
probably
close
this
week.
Transport
is
two
weeks
and
then
we
have
six
weeks
or
go
ahead,
but
we
have
six
weeks
for
all
these
other
documents.
C
C
F
E
So
under
no
that
linen
table,
especially
in
the
Acura,
operates
and
the
secondary
is
increasingly
refreshed
and
another
table,
and
then
you
know
also
suppose
a
fine
current
neighbor
a
while
under
station.
So
this
is
a
motivation
for
this
matter
and
the
note
and
history
when
you
have
presenter
so
smart
I
know
the
drafter
unit
right
have
home
okay,
so
my
god
I'd
have
a
meet
him
from
818,
618,
8
and
19
1
and
19
for
95
and
98
Michi.
E
E
Yes,
I
asked
portico
others
entry
AHA,
no,
because
the
Trahan
oh
and
has
an
image
that
camiseta
and
the
canola,
a
generator
general
gen
generally
be
fragmented
and
trail,
yes,
is,
as
is
during
the
use,
the
Pitons
and
the
nose
and
the
average.
So
we
moved
the
content
of
the
trahana
ronita
and
we
can
verify
that
the
smarter
handle
is
a
paper
trail.
C
C
E
C
E
B
E
Second,
deck
notification,
so
the
reason
is
that
the
smart
and
another
station
cannot
support
the
Assadi
and
the
Kannada
Center
is
at
the
SP.
So
if
we
used
the
pusher
key
rectory,
there
is
some
security
issue
and
I'm
really
in
teacher
for
the
I.
Don't
know
that
to
deal
with
it's
not
a
message,
so
we
single
smart
unknown,
know
the
only
support
her
poor
dear
emissary,
excellent.
C
Right,
I
think
we're
ready.
Let
me
ask
you
three
questions.
Okay,
have
you
gone
through
all
the
reviewers
comments
and
responded
to
them
and.
E
E
C
B
C
E
B
E
C
B
C
C
C
B
So
talk
about
the
goop
King
draft,
which
is
work,
move
draft
and
what
needs
to
be
done
to
finish
things
up
there.
So
just
go
over
a
little
bit
what
it
does
so
still
Trent
troll
standardizes
communication
protocols
that
need
security
services,
sometimes
encryption
and
authentication,
and
for
those
to
work
you
need
to
get
keys,
distributed
to
the
places
so
and
particularly
to
meet
modern
security
standards.
The
keys
need
to
be
periodically
we
negotiated
refresh.
You
know
you
need
some
form
of
dynamic
keying
scheme,
I
mean
you
can
also
support
fixed
keys,
but
it's
not.
B
If
that's
all
your
support,
you
don't
really
meet
modern
criterion,
so
most
of
the
existing
trills
all
existing
troll
security
is
unicast,
so
it's
got
stuff.
So
you
can
secure
things.
Point-To-Point
the
trill
over
IP
draft
uses
IPSec,
bichri
and
there's
also.
These
are
Birds
channel
messages
which
are
control.
Messages
that
are
not
is
is
their
eyes
eyes
has
its
own
security
features,
which
are
also,
of
course
available
to
troll,
but
garbage
control.
Channel
messages
are
a
different
kind
of
thing.
B
I
B
B
That
supports
native
multicast
and
you're
doing
trill
over
IP
in
that
network.
You'd
want
to
use
that.
So
that
means
some
way
to
kind
of
secure
these
messages.
You
know
a
multicast
group,
so
these
are
just
a
list
of
the
ways
you
can
do
this
sort
of
thing.
You
can
only
see
really
a
unicast.
So
if
you
have
point-to-point
security,
you
can
just
use
a
bunch
of
point-to-point
links
and
send.
If
there's
a
hundred
things
in
the
group,
you
can
send
100
copies.
Well,
that's
not
really
necessarily
two
grades
from
efficiency.
B
Quite
a
few,
you
can
distribute
a
shared
secret
to
the
group,
so
that's
pretty
efficient,
because
now
you
can
just
send
it
once
you,
the
underlying
multicast
data
distribution,
but
using
this
one,
its
group
key
to
secure
it.
There
is
the
problem
that
other
members
of
the
group
can
generally
forge
the
source
address
that
doesn't
include
originator,
authentication.
B
If
you
use
a
shared
group
key,
you
can
use
public
key
cryptography
which
will
lets
you
solve
that
problem,
but
it's
very
inefficient,
like
orders
of
magnitude,
more
inefficient
than
the
shared
group
key
and
there's
even
more
exotic
things.
You
might
be
able
to
do
but
are
pretty
complicated,
so
the
idea
is
that
in
this
draft
is
to
use
the
approach
to
a
shared
secret
key.
So
somehow
you
have
to
get
the
shared
secret
key
distributed
to
the
members
of
the
group,
so
they
all
have
it
and.
B
If
there's
a
case
where
there
are
some
messages,
where
the
fact
that
any
member
of
the
group
could
Forge
remember
a
message
showing
a
source
address
of
any
other
member,
so
you
don't
have
like
originator
authentication
in
those
those
are
fairly
rare
cases,
I
think
and
in
those
cases
you
can
still
always
fall
back
to
doing
serial
unicast
do
serve
a
unicast
at
the
point-to-point
security.
You
don't
have
any
question,
because
if
you
receive
something
securely
over
a
point-to-point
Channel,
it
must
have
come
from
the
other
end.
So
you
know
where
the
origin
was
so.
B
The
existing
graph
displays
a
generic
way
to
distribute
keys
to
a
group,
and
it
assumes
that
it'll
be
profiled
for
any
particular
use
that
profiling
mostly
includes
how
to
determine,
which
is
the
designated
member
of
the
group.
That's
sending
out
the
keys
and
controlling
the
keying
for
the
group
at
any
particular
time
and
also
the
envelope
so
the
the
group
de
janeiro
group
keying
protocol
in
the
air
that
does
not
really
specify
any
envelope.
So
you
need
to
like
a
PFD
message.
B
Should
you
need
to
specify
how
it's
a
envelope
to
address
for
a
particular
use?
So
there
are
two
profiles
in
the
current
draft:
one
for
the
Arbor
channel
message
case
and
one
for
the
trill
over
IP
case.
So
it
uses
the
pairwise
keying
it's
in
place
to
do
secure
distribution
of
the
group
key
soon,
as
you
have
do,
have
pairwise
security
in
place
and
yeah.
That's
pretty
much
what
it
does.
So
you
can
distribute
a
group
key
and
it
can
be
in
use.
B
Then
you
can
distribute
the
next
generation
of
group
key
and
then
after
you've
gotten
it
distributed
to
all
the
stations.
You
can
then
send
out
a
message
telling
them
to
switch
to
using
the
newer
key,
so
you
don't
have
to
you-
can
kind
of
minimize
the
problem
of
a
period
of
time
when
you
don't
have
a
valid
group
key
out
there.
It
also
in
the
case
where
you
don't
have
a
valid
key
for
any
reason.
You
can
always
say
always
fall
back
to
serve
a
unicast,
even
though
it's
not
for
a
efficient.
B
So
this
is
so
this
glue.
King
thing
is
really
just
about
keying.
The
actual
data
for
map
is
is
a
different
question,
so
for
the
Phil
over
IP
it
uses
the
IPSec
data
format
enveloping
and
stuff,
and
for
the
the
our
channel
case
it
uses
the
DTLS
data
format.
So
this
is
an
example.
There's
a
designated
node
distributing
kita
the
group
members.
This
is
what
the
mesenteric
messengers
look
like.
There's
some
sort
of
native
point-to-point
security
envelope
around
the
whole
thing.
B
There's
a
minimum
amount
of
headers
outside
then
there's
a
user's
key
wrap
and
has
the
actual
keying
material
and
stuff
inside
this
inner
key
wrap
so
proposed
changes
when
as
to
split
it
into
two
graphs.
So
the
generic
group
keying
mechanism
will
be
in
a
separate
graph
that
disk
covers
that
and
the
trill
specific
profiling
would
be
in
a
separate
draft.
So
there's
two
graphs
and
the
only
change
to
the
generic
group
keying
mechanism
is
supposed
is
to
add
agility
for
the
key
wrap
algorithm
currently
uses
AES
key
wrap.
B
There
are,
may
be
other
group
key
wraps
algorithms
in
the
future.
It
has
algorithm
agility
for
other
things,
but
not
for
the
actual
key
wrap
algorithm.
Currently
so
I
think
that
would
be
making
more
acceptable
security
people
and
the
profiling
is
complete
for
the
Arbor
Channel
case.
The
trill
of
our
IPA
is
a
little
bit
more
work.
So
let's
finish
that
up.
So
those
are
the
proposed
changes
to
the
draft.
B
B
B
C
J
Yes,
I
just
wanted
to
clarify
because
I'm
happy
to
push
these
work.
The
drafts
through
I
don't
want
you
to
think
that
I
am
saying.
I
haven't
asked
other
ideas,
280
sponsor
documents,
they're
left,
because
I
don't
think
that
the
work
is
worth
a
while.
It's
the
reason
that
I'm
doing
this
push
is
because
I
feel
that
the
working
group
does
not
have
the
energy
that
a
lot
of
these
documents
are
there.
Everything
we've
been
pushing
a
number
of
documents
through
in
the
last
couple
of
years,
but
we
need
to
just
get
these
done.
J
I,
don't
that
having
more
time
would
cause
them
to
be
better
or
to
have
more
energy
for
them
to
be
completed.
I
feel
that
what's
really
needed
is
the
deadline.
It's
not
an
estimation
of
the
quality
or
the
amount
of
work
that
you
all
have
put
in.
I
really
do
appreciate
that
so
I
just
want
to
be
clear.
This
isn't
unhappiness.
The
quality
of
the
documents
that
have
been
coming
through
is
excellent
and
I
do
usually
have
very
few
specificity
comments.
J
J
C
C
Ten
or
so
drafts
after
we
added
two
so
that
we're
trying
to
get
all
these
through
to
RFC
back
out
count
ten
with
the
group
splitting
ninety
nine
okay,
I
miscounted
one
anyway,
that's
it
for
today.
Thank
you
very
much
for
all
the
hard
work.
Thank
you
for
the
all.
Smart
end,
nodes
work
and
the
ARP
optimization
work.
It's
really
good.