►
From YouTube: IETF101-TRILL-20180319-1740
Description
TRILL meeting session at IETF101
2018/03/19 1740
https://datatracker.ietf.org/meeting/101/proceedings/
A
B
C
C
Guess
we'll
get
in
their
way
I'm
Donna
least
like
with
hallway
and
I
guess
I'm
acting
this
chair
for
this
meeting
in
the
absence
of
the
chair
since
I'm
working
group
secretary,
so
people
should
be
aware
of
the
note
wall
some
people
have
seen
before,
but
there
are
rules
concerning
IPR.
If
you
contribute
to
the
IETF
in
any
way.
C
C
We
do
have
the
one
RFC
it's
been
published
since
the
last
meeting.
We
have
a
whole
bunch
of
documents
that
are
on
the
RFC
editors
queue,
including
one
more
that's
not
shown
here
of
these
centralized
replication
was
added
to
the
our
theaters
queue
a
while
ago,
and
it
just
really
is
in
auth
48.
It
needs
a
little
bit
of
work.
C
But
we
have
we
address
flush.
Let's
set
our
bridge.
Tell
another
our
bridge
to
forget
some
learn
to
back
addresses,
centralized
replication
is
a
way
to
slightly
different
way
to
handle
the
distribution
of
multi
destination
frames.
You
know
active
active
case.
The
directory,
assisted
encapsulation
uses
a
directory
feature
to
control
encapsulation
at
end
nodes,
but
doesn't
require
any
change
in
the
standards
for
our
bridges.
Let's
then
do
the
D
capsulation
ecn
support
adds
ecn
to
the
trill
header.
C
It's
fill
header
extension
had
some
spare
bits
in
a
multi
topology
should
be
pretty
straightforward,
as
multi
topology
routing
multi-level
also,
this
is
a
multi-level
which
has
unique
nickname.
So
all
the
nicknames
in
the
troll
campus,
including
across
all
the
level
one
areas,
have
unique
nicknames.
C
C
C
C
And
we
have
some
other
grafts,
so
there
are
two
draught
ITF
trilha,
multi-level
single
nickname,
which
is
a
nickname.
Aggregated
version
provides
for
even
larger
scaling,
because
different
level,
one
areas
and
the
multi-level
routing
can
have
duplicated
nicknames
between
them,
because
they're
aggregated
to
a
single
nickname
that
represents
the
area
when
they're
in
level
2
or
in
other
level
ones.
There's
a
trail.
Parent
selection
is
an
informational
graph
that
in
quite
get
through
or
group
consensus,
so
I.
C
It
depends
on
the
new
area
director
who
may
want
to
say
some
words
at
at
some
point,
which
will
hopefully
be
ad
sponsored,
there's
also
two
drafts
concerning
group
keying
and
group
key
profiles.
These
are
ways
to
do:
group
keying
for
for
over
IP
multi
destination,
packets
and
group
keying
for
multi
destination.
Our
bridge
channel
messages
actually
add
the
route.
Keying
draft
is
a
very
general
group,
keying
algorithm
and
the
one
in
profiles,
profiles
of
those
two
particular
uses.
C
There
also
are
some
yang
grafts,
which
sue
hairs
has
been
primarily
handling,
who
is
under
the
weather's
about
as
a
health
problem.
So
she
is
not
here
and
I'm,
not
sure
what
to
report
on
those
I
is
a
bubbly.
She
does
plan
to
complete.
Those
is
a
basic
Trillian
graft
and
one
on
om
and
one
on
performance
management.
I
mean
there
are
our
FCS
on
how
to
to
do
am
and
performance
measurement
in
drill,
but
needs
to
be
the
yang
documents
for
those.
F
E
C
F
C
F
F
Going
on
is
we
had
a
lot
of
work
to
finish
up
in
drill
to
get
things
done
and
in
fact,
after
this
meeting
I
haven't
expected,
I
am
planning
on
closing
12
after
the
meeting
we've
gotten
all
of
the
documents
there
that
we
interested
they're
getting
through
this
working
group
for
the
pair
obstruction
documents.
If
you
have
good
enthusiasm
for
pursuing
that,
you
have
a
couple
of
routes.
F
F
Okay,
good,
so
the
only
piece
I
wanted
to
say
this
is
it
is,
you
know,
we've
gotten
an
amazing
amount
of
work
through
with
trail,
and
there
are
deployments
and
as
an
if
there
are
needs
to
do
more
work.
You
know,
as
I
said,
the
routing
area
working
group
will
be
available.
I
am
expecting
to
close
in
about
now
and
has
the
mailing
list
remain
open
for
conversation?
F
C
So
I
guess
I
have
this
presentation
I
sort
of
took
what
presentation
I
gave
on
trill
a
while
ago
and
I
extended
the
just
on
time.
Perhaps
the
we
have
a
a
cake
and
back,
unfortunately,
people
who
are
listening
remotely
will
not
be
able
to
partake
of
the
cake
that
was
just
rolled
into
the
room,
but
I
think.
Maybe
we
could
take
like
a
real
brief
recess,
because
I
want
to
take
a
picture
of
the
cake
for
any
pieces
are
cut
out
of
it,
and
other
people
are
welcome
to
take
some.
A
C
As
I
have
this
presentation,
hey
I
took
a
presentation
I've
made
on
trill
some
years
ago
few
years
ago
and
I
tweaked
it
around
several
has
way
too
many
slides,
but
I'll
zip
through
the
ones
that
are
people
should
already
know.
But
it's
a
few
of
the
ones
are
kind
of
interesting
to
think
about
again.
So.
C
C
C
There
we
go
okay,
so
so,
basically,
troll
is
very
simple
idea
in
a
calculate
the
famous,
with
a
hop
count,
route
from
via
eius
eius
and
d
capsulate
than
when
they're
on
it
rates
all
these
great
features.
So
the
reason
that
this
wasn't
done
before
was
that
when
bridging
was
first
designed,
they
were
all
these
constraints.
That
people
hardly
remember
these
days.
I
mean
you
couldn't
first
of
all,
were
the
Ethernet.
Couldn't
you
couldn't
change
the
frame?
C
There
was
no
bit,
you
could
use
and
early
equipment
really
rigidly
enforced,
the
Ethernet
length
limit
of
fifteen
hundred
and
twelve
White's.
If
you
good
the
address
or
1500
data,
so
you
couldn't
encapsulate
it
wasn't
it
it
would
not
work
and
he
was
able
not
to
change
any
of
the
end
node.
So
bridging
had
these
enormous
constraints.
Suspending
for
you
is
a
great
solution,
but
it's
really
pretty
brittle.
You
know
one
of
the
differences
between
driving
bridging
is
the
routing
only
send
stuff
out
if
it
gets
enabling
messages
on
a
port.
C
So
it's
sort
of
fails
in
the
safe
direction
if
the
camp
process
or
else
to
process
control
messages,
whereas
rigging,
send
stuff
out
all
reports
and
bless,
it
gets
messages
that
indicate
it's
unsafe,
so
it
fails
in
the
unsafe
direction,
tends
to
melt
things
down
with
beta
loops.
The
drill
is
designed
to
combine
the
features
of
both
routers
and
bridges
has
lots
of
other
features.
C
So,
which
are
now
possessed
by
other
protocols,
but
there's
tables
grow
only
with
the
size
of
the
number
of
of
throw
routers
to
our
bridges,
not
with
the
number
of
n
stations
it's
compatible
with
our
IP
routers.
So
there's
the
troll
canvases
transparent,
two
routers,
like
bridges,
are
sports,
the
VLANs
and
so
forth,
as
MTU
support,
active
active
connection
at
the
edge
padded
directory
features
and
molding
topology
and,
of
course,
like
spanning
tree
I
mean,
unlike
every
other
protocol.
C
Besides
been
in
Crete,
it
has
a
poem
which
is
algo
I'm
version
2,
which
actually
written
by
Ray
Pro
Maria's
son.
The
radio
wrote
the
original
algorithm
for
spanning
tree,
so
I
won't
burden
you
with
reading
this,
but
it's
here
also
I
put
algo
I'm
the
problem
for
spanning
tree
in
the
Supplemental
slides
at
the
end.
C
So
it's
an
interesting
because
there's
sort
of
this
hierarchy
of
things,
so
bridges
don't
see
hugs
or
repeaters,
and
do
you
have
hubs
and
repeaters
and
stuff
it
just
sort
of
looks
like
a
bigger
cable
like
multi-drop,
Ethernet
or
something
but
the
fact.
The
hubs
and
pre
Peters
repeat
everything.
That's
that's
that
mode
of
operation
is
terminated
by
a
bridge
which
looks
of
the
stuff
and
might
decide
not
to
forward
it
or
or
whatever.
C
Then
full
switches
are
above
that
to
them.
Bridged
lands
appear
to
be
transparent,
but
they
terminate
the
spanning
tree
algorithm.
So
a
spanning
tree
is
never
sent
through
it
through
a
troll
campus.
You
can
take
it
make
a
bridge
land.
You
can
replace
the
bridges
one
at
a
time
by
troll
switches
and
as
you
do
so,
it
gets
increasingly
efficient
and
the
spanning
tree
domains
get
chopped
and
come
smaller
and
smaller,
which
is
a
good
thing.
A
spanning
tree
doesn't
work
very
well
when
you
get
large
domains.
C
People
control
switches
run
the
troll
protocol
and
in
routers,
where
three
at
the
top
level,
everything
below
that
troll
canvas
looks
transparent
to
IP
routers,
but
IP
routers
terminate
the
the
tro
protocol
and
don't
forward
there.
They
drill
things
through
through
them,
and
this
is
mostly
determined
by
how
they
treat
various
multicast
addresses.
C
So
a
spanning
tree
would
work
in
the
meantime
for
like
days
or
all
the
paperwork
in
this
major
hospital
was
being
done
by
hand
like
on
paper
and
pencil,
which
other
I
mean
even
the
new
favor
article
specifically
says
spanning
tree.
You
don't
find
newspaper
articles
mentioning
spanning
tree
very
often
so
radio
decided
to
do
something
better
about
this.
So
this
is
the
main
history.
Part
of
this,
so
I
went
back
a
little
ways.
C
So
she
presented
to
the
IETF
and
a
working
group
was
chartered
so
initially
that
just
had
to
share
a
Burtonwood
mark
and
margaret
then
Wasserman
coal
Kong
was
the
internet
ad
when
it
was
chartered
fairly
soon
thereafter,
the
internet
ad
was
changed
to
mark
Townsley,
Margaret
didn't
run
again
and
fairly
soon,
a
little
bit
longer
boycott
e.
The
year
later,
I
was
added
as
co-chair,
so
it
looked
like
in
a
2008
that
it
was
like
done.
You
know
was
about
to
come
out
when
a
semi
fatal,
MTU
problem
was
found
with
the
protocol.
C
So
a
whole
bunch
of
MTU
features
were
added
to
it,
including
I
Sai,
as
probe
in
act
PDUs
and
various
things
that
can
AMT
you
feature
to
to.
So.
If
the
feature
means
you
can
be
safe
with
we're
somewhat
reduced
empty
use,
it
also
makes
it
possible
for
it
to
sort
of
know
what
the
MTU
is
and
be
able
to
take
advantage.
Campus-Wide,
if
you
happen
to
have
all
your
links,
support
jumbo
frames
so
icky
you
can
use
jumbo
link,
state
protocol
messages
and
so
on
and
so
forth
with
it.
C
The
only
things
you
can't
do
jumbo
are
things
like
hellos,
because
you
want
to
do
hello
is
to
discover
a
new
or
bridge
to
find
out,
maybe
that
our
bridge
has
to
lower
empty
you,
because,
if
you
use
using
anyway
so
there
it
looks
like
for
some
reason,
trill
didn't
meet
in
the
July
2008
or
2009
meetings.
Every
other
meeting
in
this
entire
interval
that
fill
working
group
met
at
least
there's
no
minutes
arrey
in
the
Proceedings.
So
there's
this
problem,
ineffability
statement
finally
came
out,
which
I
think
is
slightly
out
of
date.
C
Then
there's
another
change
in
internet
area
directory
to
ralph
groms.
That
was
convenient
because
of
the
time
ralph
groms
sat
about
a
hundred
feet
from
me,
so
right
needed
ad
attention.
I
could
get
up,
I
was
consulting
to
Cisco
at
the
time
and
it
was
very
convenient
yes
for
me,
yeah
I
I.
Don't
think
he
minded
so
much
I
mean
you
know.
I
feet
me
to
go
away
or
come
back
later.
I
certainly
would
do
that
anyway.
So
why
so?
C
This
could
be
Solis
MTU,
probably
cause
something
like
a
two-year
delay
was
somewhat
less
less,
but
it
moved
to
2008
to
2010
so
find
the
a
base
protocol
was
a
publication
was
requested,
and
then
it
was
I
took
a
while
for
her
to
get
approved
and
on
March
15th.
The
base
protocol
is
approved
and
at
that
time
the
ether
types
have
been
allocated,
the
block
of
multicast
addresses
was
allocated
III.
C
C
Multicast
number,
it's
kind
of
adjacent
to
the
ones
are
actually
used
by
bridging
and
I've
seen
unhappy
expressions
on
the
face
of
some
people
when
reading
this
and
looking
at
but
anyway
so
we
had
an
OP
in
2010
very
early,
just
control
plane,
no
data
was
passed,
but
you
know
exchanging
hellos
and
establishing
adjacency
and
stuff
like
that
and
in
lieu
took
because
the
approval
it
took
a
while
to
get
through
the
arcs,
the
others
queue
is
in
2011.
When
the
base
protocol
spec
6325,
surely
a
foundation
of
stuff
got
published,
it
was
held
up.
C
I
guess
by
the
use
of
is,
is
which
went
through
the
eius
eius
working
group
and
at
the
time
there
was
some
friction
and
it's
decreased
with
time.
So
now
I,
don't
think
there's
any
real
problem
or
anything.
If
drill
wanted
something
from
the
ice
ice
working
group
now,
I,
don't
think
Joe
would
have
any
more
or
less
difficulty
than
anybody
else
would
getting
anything
out
of
the
ice
ice
working
group
and
very
early
there
was
a
very
earlier-
was
a
RFC.
It
fell
over
PPP
on
a
different
link.
C
Protocol
12,
a
ppb
doesn't
have
any
ether
types
uses,
PP
beep
code
points
and
everything
like
that
sort
of
established
the
architecture
for
trill
link
usage
and
we
got
recharger.
We
got
reach
charted
periodically
in
2012.
There
was
a
second
control
plane
Interop.
At
that
point,
John
Hudson
was
added
his
working
group
secretary.
We
did
but
a
burst
of
documents,
they're
agreeing
amid
OAM
requirements,
directly
assistance,
high-level
design.
C
C
Eric
would
mark
left.
As
a
working
group
co-chair
and
John
Hudson,
who
had
been
secretary,
became
the
working
group
co-chair
and
sue
Harris
became
working
group
secretary.
There
was
also
this
sort
of
massive
burst
of
documents,
and
if
you
look
at
this,
you'll
notice
the
whole
first
bunch
like
all,
but
the
bottom
two
are
consecutive
RFC
numbers
and
that's
because
they
were
all
held
up
on
the
7176.
Trill
you
serve
is
is
changes
as
I
recall.
Anyway,
there
was
a
normative
dependency
which
so
these
things
are
all
piled
up
and
can
plop.
C
They
all
came
out
at
once
a
whole
bunch
of
stuff
in
here
very
formative,
improvements
and
won't
go
through,
but
PFD
support.
There's
clarifications,
Corrections
and
updates
city
from
experience.
Yeah
at
these
interrupts
knew
just
to
say
there
were
especially
the
early
ones.
There
were
some
questions.
People
had
about
what
the
spec
mentor,
where
it
might
be
ambiguous
or
something
and
there's
only
was
there.
C
And
the
sorry
statement
problems
have
a
back
to
act
around
this.
You
first
of
all
of
C's
came
out
turns
out.
Most
of
these
are
standards
for
the
standards
and
I
was
on
an
author
on
many
of
them,
which
was
made
some
people
unhappy,
so
I
would
be
I
swapped
with
Sue
hares,
and
she
became
a
co-chair
and
I
became
secretary.
C
2015
is
the
the
basically
that
Trillo
am
graft,
welcome.
It
was
published
2016,
we
sort
of
have
more
documents
coming
through
where's
way,
see
the
implementation
of
active,
active,
active
work
and
started
very
early.
We
took
a
while
to
get
through
on
our
way
so
on
and
so
forth,
various
extensions
and
sittings,
and
in
2017
we
had
some
of
these
RCS.
It
was
obviously
to
earlier
trail
RFC's
when
we'd
found
additions
or
improvements
or
clarifications
to
make,
and
so
far
on
2018.
C
Only
one
RFC's
come
out,
which
is
the
ARP
and
they
were
discovery
optimization.
There
will
be
more
in
2018
what
would
suspect
from
the
huge
list
in
the
artist
at
risk
you
so
is
it
if
the
I
think
this
is
complete,
I'm
sure
it
cumulative
lists
of
the
companies
that
participated
in
the
trill
and
reveal
the
events
at
the
mercy
to
Hampshire
in
her
on
bility
laboratory.
C
A
couple
references
one
is
an
article
in
the
Internet
Protocol
journal
on
Trello
relatively
early,
but
still
and
then
also
the
reference
to
the
original
paper
presented
by
Radia
at
infocomm
2004,
actually
before
the
word
trills
sort
of
been
invented.
Just
referring
to
them
as
are
ridges
or
routing
bridges.
C
So
that's
my
trill
history,
presentation,
okay,
say:
I,
threw
on
a
couple
of
the
algorithm
for
spanning
tree
and
a
bunch
of
slides,
explaining
how
troll
works
and
how
multi
destination
works.
Why
did
phony
claims?
Why
is
is
was
selected
as
opposed
to
is
a
OSPF
etc.
How
Mac
learning
Weirich
societies
that
are
are
on
the
end.
C
C
E
C
C
G
C
Is
a
YouTube
video
of
presentation,
I
made
yeah
presentation
to
nanog
was
quite
a
while
ago.
I
think
I
did
that.
But
there
was
a
survey.
It
was
cold.
The
debate,
debate,
I,
don't
know
the
thing
I
didn't
like
so
much
about
it.
I
was
there
I
presented
trill
and
several
people
presented
shortest
path.
Bridging
and
you
know,
I
felt
kind
of
outnumbered
and
stuff
biased,
so
I'm
not
sure
I
did
a
particularly
good
presentation,
but
anyway,
I
think
you
can
find
all
that
on
YouTube
or
at
least
a
in
the
nanog
files.