►
From YouTube: IETF98-TRILL-20170327-1520
Description
TRILL meeting session at IETF98
2017/03/27 1520
A
B
C
C
D
D
E
A
F
B
G
B
One
other
thing:
this
is
going
to
be
more
of
a
discussion
and
friendly
talk,
so
it's
always
more
interesting
if
you're
up
toward
the
front-
and
we
all
can
talk,
I
realize
you
might
feel
comfortable
in
the
back,
but
we
always
like
it.
If
you
come
to
the
front,
if
you
don't
there's
no
forcing
you,
but
I
certainly
would
love
to
have
many
of
you
closer
her.
B
Okay
with
that
will
start
we
and
besides,
if
you,
if
you're
using
later
laptop,
the
tables,
are
much
more
effective.
Okay.
This
is
our
note.
Well
how
many
people
have
never
seen
a
note?
Well
good.
This
note
well
looks
like
all
the
rest
of
them.
Please
read
it.
It
talks
about
submissions
and
I
pr's
and
what
your
contribution
is.
You
should
read
and
understand
your
friendly
note.
B
B
We've
refined
it,
but
it's
always
tough
to
look
at
the
document
one
more
time,
but
I'm
going
to
ask
several
people
to
just
please
read
these
documents
were
we're
making
the
end
of
the
active
activist
passed
through
we're,
making
the
end
of
some
of
the
directory
assistance
pipe
drafts
and
we're
going
on
to
the
multi-level
and
low
pay
apology,
please,
I
think
we've
got
it
as
I
said
by
George.
I
think
we've
got
it,
but
please
read
okay.
This
is
our
agenda.
For
today.
B
So
that's
our
topics
today,
we'll
have
five
minute
wrap
up.
Okay,
so
we've
gotten
Sutton
two
drafts
in
the
RC
Ataturk
you
we've
got
two
in
the
publication
request
and
we've
got
two
graphs
that
got
returned
at
this
point.
The
ARB
optimization
is
gone
through
one
revision.
Thank
you
very
much
use
out
and
we
are
back
to
just
one
more
check
on
it
and
then
it
will
go
back
to
Elias
kind.
B
It'll
go
through
working
group
last
fire
at
call
and
then
aleeah's
review
again,
please
when
we
start
these
review
processes,
I
appreciate
your
active
looking.
There
are
two
past
working
group
last
call
with
drill
over
IP
and
trill
resilient
trees,
I've
sent
in
some
comments
to
the
author's
we
need
to
quickly
revise
these
after
ITF.
B
So
I
will
actually
hang
on
several
of
the
authors
like
Margaret
over
here
in
and
try
to
get
some
of
these
things
and
I
just
really
appreciate
your
help,
because
getting
these
through,
as
you
will
know,
Margaret
from
being
chair
and
other
things
getting
it
through,
the
last
toughest
fun
is
getting
it
through.
The
last
process
is
difficult.
We
have
a
draft
and
working
group
last
call
and
I've.
Let
it
sit
through
this
time
period.
We
haven't
had
many
comments
on
it,
but
I.
B
I
think
it's
there's
been
some
interest,
so
please
comment:
we
have
graphs
that
have
revision
and
I
think
the
smart
end
nodes
in
a
dress
flush.
The
address
flush
will
go
back
through
working
group.
Last
call
because
it
didn't
finish
the
smart
end,
nodes
I,
think
you
know
what
needs
to
happen
there
and
we
will
again
go
through
working
through
class
call.
Now.
What
does
all
this
mean?
What
we're
doing
is
we've
done
tremendous
good
work
and
we're
pressing
it
all
toward
RC,
so
help
your
fellow
authors
by
pressing
it
toward
RFC.
B
B
Issues
we've
gone
through
the
active,
active
we've
gone
through
the
directory
assistance,
multi-level
multi,
Nick
thing.
Thank
you
is
there.
Is
it
here?
Oh
there
it
is
I
got
it
nope
that
didn't
work.
Well,
let
me
go
back
okay,
one
more
so,
for
example,
we
have
the
trill
data
center
interface
and
we
have
the
trill.
B
We
have
some
true
trill
transport
over
mpls
I
will
try
I.
Think
most
of
those
graphs
are
fairly
well
cooked.
I
will
ask
you
to
please
review
the
yang
documents
will
cover
next.
What
you're
hearing
me
say
is
I
think
we're
sort
of
to
the
end
of
several
large
pieces
of
work.
But
the
only
thing
hanging
out
is
is
security
that
we've
got
some
individual
drafts
and
the
any
urgent
thing
we
hear
from
you
we're
at
a
place
where
most
working
groups
go
into
a
hiatus
or
sort
of
go
into
a
death
rebirth
thing.
B
This
isn't
because
technology
isn't
working
it's
because
in
reality,
you've
done
a
really
really
good
job,
with
we've
gotten
through
a
whole
bunch
of
stuff
I.
Think
we've
gotten
to
all
the
product
lines
what
they
need
for
those
who
are
shipping
trill
product
it
would
be
or
trill
sort
of
product.
Tell
us
if
we've
gotten
everything,
but
we've
tried
to
focus
on
what
we've
got.
If
there's
something
we
needed
we're
heading
toward
that
G
we've
gotten
through
the
first
piece,
we've
defined
everything
will
let
it
set.
B
If
there's
some
urgent
problem,
I'm
sure
the
80s
will
pick
up
and
move
it
out,
but
we're
at
that
you
know
coming
to
one
end.
Why
is
that
important?
I
really
want
to
get
all
the
rfcs
through
the
process.
Can
I
probably
send
it
ten
times
now
hope
read
the
dress.
Okay,
and
these
are
the
documents
we
have
done.
You
see
this
status
and
these
are
milestones
and
looking
with
these
milestones,
we're
going
to
readapt
the
milestones
or
out
of
faith,
but
we
will
we've
completed
a
whole
bunch.
This
is
a
tremendous
working
group.
B
Its
technical
capabilities
have
been
a
joy
to
work
with,
but
we
have
a
few
overdue,
mostly
because
we've
recycled
a
couple
of
these
in
review,
I'd
like
to
catch
up
our
interviews
and
we
were
going
to
recharter
or
shut
down
the
working
group
I'm,
not
sure
we're
going
to
make
July.
We
may
be
november,
but
I
think
we're
heading
toward
that
hiatus.
That's
your
chair!
Trying
to
so.
If
you
got
some
challenge
on
that,
that's
where
I
think
we're
heading
again,
because
I
think
we
finished.
B
Let
me
give
you
an
example
of
another
working
group
that
went
through
this.
Bgp
went
through
a
whole
long
cycle
where
we
had
a
whole
lot
of
work,
and
then
we
went
into
hiatus
for
a
while.
We
sort
of
say
that,
because
we
really
didn't
have
in
something
interesting
to
say.
But
then,
when
we
came
back,
we
had
a
lot
of
interesting
things
to
say
and
we're
still
finding
things
that
are
changing
in
the
BGP
environment
is
just
for
a
while.
B
Okay,
how
many
people
here
have
worked
with
yang
I
think
he's
out
okay
good
now.
This
is
a
little
bit
of
talk
about
a
challenge
when
we
had
our
yang
models
are
trilling,
which
is
the
basic
configuration
the
p.m.
and
the
OEM,
the
PM,
which
actually
works
out
the
ohm.
We
did
a
lot
of
work
ahead
of
everyone
else
and
then
tisa
thought
it
would
be
a
really
good
idea
to
have
lime
and
to
make
the
true
model
go
under
lime.
B
Lime
is
finishing
its
work
and
I.
Don't
know
that
it
got
to
everything
we
thought
it
would
so
I'm,
faced
with
three
options
and
I
just
for
those
who
are
working
with
gang
and
drill
option.
One
is
the
working
group
last
call
and
send
through
the
process
the
modules,
because
it's
my
understanding
from
deepak
and
maybe
other
people
who
have
implementations
that
you
have
implementations
of
the
existing
drill
gang
models.
They
will
not
be
exactly
up
to
latest
snuff
off
for
some
of
the
requirements
and
format.
B
So
some
stuff
is
easy
for
format
to
change
and
some
stuff
is
fairly
hard.
My
first
option
is
we're
not
going
to
really
address
the
hard
changes,
we're
going
to
say
these
are
adopted
and
used
in
that
case,
I
will
collect
from
deepak
and
any
work
view
that
have
drill
and
figuration
yang
models,
the
information
and
send
it
to
the
isg.
B
The
second
option
is
to
change
the
trill
p.m.
and
OEM
to
be
dependent
on
line.
That
means
you
have
the
line
base
and
then
we
have
to
do
new
models.
For
that
d.
Fox
original
comment
to
me:
an
email
was
no
I'd
rather
stay
with
the
original
gang
model.
Okay,
we
can
do
both,
but
that
takes
a
bit
more
energy
I'm.
Well
to
do
that.
But
I'd
like
to
start
now,
because
yang
models
have
taken
us
a
while
any
feedback
you
sound,
ok,
I
would
like
to
see.
B
D
B
F
B
B
If
anyone
wants
both,
you
must
tell
me
within
this
week
because
we'll
go
on
and
I
will
tell
the
lime
chair.
That's
nice!
If
you
get
someone
to
volunteer
go
forth,
but
we're
going
to
stay
focused
on
the
first
option.
Ok,
and
if
you
feel
shy
about
the
mic,
come
and
talk
to
me
afterwards,
but
I
will
make
a
fairly
call.
B
Let
me
just
give
you
a
little
information
about
some
of
the
challenges
now
that
I've
gone
through
the
major
question,
which
is
one
of
the
challenges
that
is
that
in
the
future,
in
the
sort
of
lime
time
frame,
we
would
have
to
work
with
the
revised
datastore
and
revised
data
stores
changed
the
OP
state
to
be
separate
than
the
rest.
It
also
has
taken
dynamic
data
stores
such
as
trill
r2s
might
be
into
that.
B
H
Making
FGL
Tantalus
against
is
contagious,
as
in
the
present
is
eliminating
32
digital
space,
invention
for
forums
for
bringing
FJ
equal
to
negative
evidence
emerging
ever
so.
What
exactly
all
and
again
inhale
was
in
the
inner
head
under
the
original
date.
Our
Garrett
is
so
what
we
have
thought
out,
LG
refrigerator,
and
this
document
proposes
changes
in
the
actual
package
structure
that
is
describing
the
subsequent
section
of
this
document
in
the
good
next
leg.
H
H
H
H
H
This
is
how
it
looks
like
so:
trill
header
is
whatever
we
are
using
as
of
now
so
FGL
header
it
is
having
that
is
in
the
town.
So
you
can
see
here.
We
have
some
information
regarding
a
chill
headed
I
beat
our
with
key
priority
field,
and
we
are
bringing
entropy
labor
also,
and
the
entropy
label
is
because
what
we
now
have
is
that
we
have
to
go
all
to
the
deep
packet
inspection
to
find
out
for
doing
a
load,
balancing
on
course,
I'd
means.
H
H
H
H
H
Ok,
he
field
is
a
two-bit
field
whose
value
provides
information
on
the
National.
That
is
priority
field,
so
the
priority
field,
incredible
value,
0
of
people-
indicates
that
reality
field
is
offered
from
the
inner
wheel
and
header,
which
is
a
forwarding,
including
DC.
I
means
DC,
the
eligibility
drop
eligibility
indicator
and
so
for
most
significant
bit
will
be
in
use
or
priority.
For
least
significant
bits
should
be
set
to
0
and
ignored
by
doing
packet,
forwarding
position,
and
so
can
they
go
for
next
level.
H
Okay,
value1
non-people
indicates
that
clarity,
Figgis
clarity
field
is
offered
from
inner
mpls
headers
exp
bits,
which
is
a
3-bit
field,
and
three
most
significant
bit
will
be
used
for
priority.
Highly
significant
bit
should
be
set
to
zero
and
ignored
while
doing
pack
according
decision
valued
20
people
indicates
reality.
Field
is
copied
from
inner
IP
for
headers,
tosser,
dhcp
killed
and
hence
all
eight
which
will
be
used
for
taking
forwarding
decision
and
the
value
three
of
a
bit
field
indicates
that
purity
field
is
copied
from
inner
eye,
be
a
success.
H
H
Okay,
this
entropy
field,
entropy
label,
is
a
16-bit
field
which
is
calculated
based
on
has
generated
by
in
that
package.
Layer
to
address
means,
MAC
addresses,
source
or
destination
Methodist
and
layer.
3
addresses
IP
for
source
of
destination,
addresses
are
ipv6,
losa
destination
addresses,
are
l
for
addresses,
means
tcp
or
UDP
port
numbers,
a
combination
of
all
Diego
and
then
you
to
use
to
to
load
balancing
distribution
are
low
distribution
of
Dayton
ecmp
of
core
our
bridges
without
doing
deep
packet
inspection.
Now.
H
H
So
by
bringing
in
trophy
level
what
we
are
trying
to
do
is
again.
We
are
giving
a
free
hand
to
means
unless
later
or
someone
who
want
like
whether
he
want
to
use
an
inner
ear.
Two
hitters
are
in
the
l3
headers
and
you
don't
do
a
deep
packet
inspection
and
without
that
doing,
entropy
level
will
help
us
to
do
a
load
balancing
and
it's
like.
H
Okay,
so
now
FGL
will
come
here
after
this
entropy
level.
This
fine
grind
level
is
a
24-bit
level,
which
is
same
as
whatever
it
was
earlier.
So
what
we
have
married
this
as
a
contiguous
later
cuz
this
contributes
so
this
indicates
the
frame
belongs
to
it,
which
particular
broadcast
segments.
So
on
top
of
that,
we
have
reserved
for
future
purpose.
I
think
you
can
go
for
next
line.
H
So
what
exactly
it
will
provide
us?
What
are
the
uses
that
we
are
getting
from
here?
Is
that
with
entropy
label
load
distribution
in
the
easily
turn
on
me,
cmp
for
our
bridges
and
achieving
of
qos
until
can
be
done
based
on
customers
requirement
and
top
of
that?
What
we
can
do
here
is
that
there
are
some
models
in
mpls
for
achieving
us
like
white
model
short
fight
model,
all
those
things
we
can
bring
into
the
trail
also
alert.
If
really
and
some
customer
wants.
H
It
could
be
like
that,
but
for
achieving
us
until
what
I
can
see
here
is
that
right
now
we
don't
have
that
much
of
free
hand
provided
to
the
administrator
if
he
wants
to
stay
whatever
is
there
in
the
trail?
Header
is
indicating
to
him
or
if
you
want
to
change,
so
we
can
give
him
a
free
hand.
If
you
want
to
take
the
decision
based
on
the
inner
headers
which,
in
the
hidden
there
is
a
to
head
up
our
l3
hidden,
if
l3
header
means
again
how
he
want
to
do
it.
H
So,
by
bringing
this
priority
field
into
FGL
header
I
think
we
are
giving
a
free
hand
to
a
customer
r2r
administrator
how
he
wanted
to
take
a
decision,
and
another
thing
that
is
most
important
is
that
a
change
of
different
mode
of
services
easily,
like
we
land
based
service
instance,
we
learn
bundle
service
instance.
We
learn
our
bundles
resistance,
etc.
So
it
will
become
very
easy.
When
we
bring
a
team
outside
the
inner
header,
we
can
go
for
next
line.
B
So
well,
Donald's
going
to
the
mic.
Let
me
ask
an
operational
question.
This
proposal
hints
that
you've
got
customers
who
are
leading
these
changes.
Can
you
provide
some
more
detail
on
why
the
original
stuff
original
fine-grained
label
wasn't
able
to
accomplish
some
of
these
pieces
and
what
your
customer
scenario
don't
give
enough?
So
we
can
tell
the
customer,
but
if
you
would
just
give
us
a
little
background
on
what
customers
you're
working
with,
why
it's
important
to
them?
That
would
really
help
me
in
my
evaluation.
So.
H
All
right,
no,
we
don't
have
any
customers
as
of
now,
so
what
we
can
see
the
different
technologies
for
we
are
like
what
we
have
right
now
is
tell
always
means
want
to
have
something
like
FGL
to
be
outside
the
inner
packet
headers.
So
what
we
are
doing
here,
we
are
putting
FGL
inside
a
inner
packet
header,
so
they
always
have
means
want
it
will
be
outside
of
it
and
top
of
it.
H
As
I
told
that
entropy
level,
what
we
here
we
are
proposing
will
have,
it
will
give
a
very
good
chance
for
load
distribution
there
as
the
condo
in
deep
packet
inspection,
that
that
will
bring
a
lot
of
overhead
to
the
for
our
bridges.
So
it
has
to
go
all
the
way
in
split
in
order
to
freeze
energy
fields
so
and
I
also
chillin
of
us.
Until
what
will
we
right
now
we
are
doing
is
we
are
using?
Sometimes
we
land
header
outer
wheel
and
header.
H
H
We
need
to
make
this
FGL
means
to
bring
the
sexy
and
outside
the
inner
header
and
to
put
in
the
trail
header
I
did
it
will
become
easier
so
because
sometimes
it
is
confusing
when
we
see
the
ethertype
well,
if
two
times
it
has
been
repeated,
so
so
we
are
wasting
you
around.
Some
two
plus
two
means
four
byes
where
to
tell
that
this
is
the
a
third
type.
H
B
E
H
B
H
B
Okay,
thank
you
just
a
minute
down
over
in
drama
theme,
I'm
asking
about
use
cases
I.
G
G
G
H
E
I
I
And
that's
a
little
this
little
point
these
lights
here,
I
was
thinking
they
proposed
changing
the
version
number,
which
is
kind
of
a
big
deal
in
the
trouble.
Petrolhead
er
there's
other
ways
of
doing
it,
which
would
be.
Are
there?
Are
these
there's
some
flag
fields
in
the
trail
header
which
are
critical
flags
over?
There
said
they
would
cause
any
switch
that
didn't
understand
it
to
reject
it,
which
don't
require
changing
the
version
at
all.
F
G
J
Margaret
Colin
from
pain,
land
security,
the
first
question
I
have
I
guess
is
I,
did
we'll
go
over
the
drafts
for
this
meeting,
but
I
didn't
find
this
one.
So
what
is
the
draft
that
this
is
documented
in?
J
J
It
sounded
like
we're
going
to
have
a
new
header
version
that
says
that
the
header
that
follows
is
an
FTL
header
and
then
we're
going
to
have
a
bit
in
the
FGL
header.
That
says
it's
a
valid
FGL
header.
Now,
what
is
it
if
that
it
isn't
set
an
invalid
FGL
header
I
mean
I,
don't
I,
don't
understand
what
the
is
that
space
always
going
to
be
reserved,
but
if
that
fits
one
we
should
do
something
with
it
and
if
not,
it
should
be
all
zeros.
I,
don't
understand
what
this
needs.
That.
B
J
H
H
Sometimes
the
htl
values
might
be
all
zeros.
So
then
we
can
tell
that
there
is
a
invalid
FGL
value,
which
is
there
in
this
header,
and
you
can
directly
drop
it.
So
that
is
the
purpose
why
we
are
linking
the
I
bit.
So
if
this
is
not
the
way
to
commence
like
to
handle,
it
I
think
we
still
have
time
and
we
can
think
about
it,
but
right
now
the
main
intention
is:
if
something
which
is
all
the
value
is
zero,
then
implicate
the
art
bridges
to
directly
drop.
That
frame.
J
B
J
One
thing
I
should
think
you
should
at
least
consider
here
is
that
deciding
what
to
do
by
looking
at
a
value,
and
then
a
field
is
computationally
more
expensive
right
than
just
deciding
what
to
do,
based
on
a
field
and
I
wonder
if
it's
possible
and
I'd
have
to
look
and
think
no
to
have
a
way
that
we
encode
one
field.
J
That
could
be
as
much
as
10
bits
long
because
you're,
one
two
and
three
are
going
to
take
two
bits
plus
you're
a
bit
field
right.
If
we
have
a
way
to
encode
one
field
that
we
could
map
all
these
values
to,
so
that
there's
a
consistent
set
of
fields
instead
of
having
three
different
types
is
something
I
would
wonder
about
in
terms
of,
if
there's
a
meaningful
way
to
do
it,
I
mean,
if
they're,
completely
disparate
information,
that
they're
trying.
J
H
The
reason
behind
why
this
these
are
dope
fields,
because
one
it
indicates
that
convey
we
have
copied
is
miss
Perry
liquid.
Then,
when
we
want
to
copy
it
back
means
when
it
is
going
out
of
30
campus.
So
if
some
customer
want
in
such
a
way
that
if
you
want
to
copy
the
priority
bits
back
to
its
original
hidden,
so
it
will
help
him
to
know
which
very
he
has
to
copy
it
back.
So
whether
it
is
n2
header
there
is
LT
hidden.
So
where
exactly
you
want
to
copy
that?
That's
why
we
have
divided.
H
H
Correct,
yes,
that's
why
it
means
we
have
I
mean
different
type
of
models
in
qos.
As
what
I
was
discussing
like
we
have
pipe
model
me,
hair
and
universal
model.
We
have
short
pipe
model.
These
models
means
means.
Each
model
is
different
from
other
in
the
case
where
sometimes
we
don't
want
to
copy
it
back.
Sometimes
we
want
to
copy
it
back
to
the
virginal
header,
so
the
reason
is
just
to
do
the
edge
carrbridge.
It
will
indicate
that
where
you
want
to
copy
it
back
means
whether
to
the
end
to
head
over
a
three-headed.
J
Yeah
I'm
pretty
sure
it
is
it
safe
to
modify
all
of
these
are
empaths
so
but
yeah
I
Donald,
wants
to
say
something
about
this
I.
I
I
That
there's
always
a
for
bit
handling
priority
for
the
drill
data
packet
and
basically
the
grotto
eligibility
and
the
three
bits
of
priority
and
the
trill
switch.
Ports
are
Bridgeport's
and
they
could
have
these
8
q's.
Having
additional
information
around
is
fine,
but
it's
sort
of
would
be
good
if
there
was
a
spirit,
priority
and
a
wonder.
Crop
eligibility
field
available
and
having
additional
information,
which
is
ok,
but
the
exit.
The
trill
architectures
would
have
assumes,
there's
that
for
bit
summary
of
the
priority
for
handling
things.
J
C
J
B
B
G
G
Greatest,
hopefully,
a
way
of
encouraging
people
to
read
into
this.
Assuming
that
this
all
be
written
up
is
it
there
are
a
lot
of
things
we
wanted
to
do
when
thinking
about
Joe
that
we
didn't
do,
because
there
was
so
much
resistance
to
any
sort
of
hair
change
of
any
sort,
and
the
fear
was
that
it
would
cause
all
the
ripple
effects
in
the
market.
I've
Donna
pointed
out,
there
are
no
silicon
chips,
supporting.
E
G
B
B
B
B
B
B
I
C
I
I
The
basic
the
VLAN
on
the
native
traffic-
if
you
configured
it
this
way,
gets
mapped
into
an
FGL
at
the
ingress,
our
bridge
and
then
the
egress.
It
gets
mapped
back,
I
guess
the
mappings
can
be
different,
and
it's
also
prides
away
where
you
can
have
the
purdy
us
set
for
the
traffic
which
is
different
from
that
externally
and
can
be
restored.
I
mean
don't
I
push
them
spoken:
yeah,
bushnell,
okay,
yeah,
priority
of
yellow
okay.
I
So
this
is
what
it
looks
like
it's
very
similar
to
the
previous
diagram,
and
you
might
wonder
how
this
strange
double
ethertype
thing
came
out.
It's
it
sort
of
comes
from
the
ethernet
from
the
tag
stacking
stuff
I
mean
if
you
go
back
far
enough.
A
certain
large
commercial
vendor
started
putting
to
customer
VLAN
tags
in
a
row
to
indicate
essentially
a
24-bit
label
and
which
that
was
never
approved
as
the
standard
by
any
standards
body,
and
in
fact,
what
I
Triple
E
did
is
day
later
said.
I
You
need
to
use
a
service
VLAN,
followed
by
a
customer
VLAN.
If
you
want
to
effectively
get
that
because
they
thought
putting
two
copies
of
the
same
tag
in
the
row
was
evil
or
pertaining
to
pick,
it
certainly
was.
It
was
never
standardized
that
it
never
approved.
So
not
certainly
wasn't
kosher
so
anyway,
but
so
we
sort
of
wanted
to
do
something
like
that.
But
I
don't
remember
if
there
was
a
liaison
or
anything,
but
basically
we
were
basically
told
anything
we
wanted
to
do
with
that
customer
or
service
VLAN
tags.
I
I
So
one
of
the
things
is:
let's
assume
it
is
a
draft
ins.
The
standardization
of
the
proposal
that
Goomer
has
presented.
There
will
be
a
problem.
You
might
have
a
link
with
different
ports
on
it
from
different,
our
bridges
that
use
different
FGL
encodings
and
that's
not
going
to
work
very
well.
So
does
the
one
of
the
details
that
we
would
need
to
clean
up
to
do.
I
This
is
to
have
a
bit
in
the
capabilities
of
the
either
in
the
port
game
abilities
or
the
switch
capabilities
indicating
which
variation
they
were
using
and
if
you
guys
are
all
the
same
on
the
link,
everything
is
wonderful,
you
might,
it
might
be
better,
be
purport
if
there
could
ever
be
something
you
could
convert
from
one
to
the
other
or
you
could
have
it.
/
switched.
I
B
I
So
yeah
I,
whoever
has
chuckling
before
and
Mar,
both
our
names
on
the
Venus
draft,
so
Buddhist
slides.
So
the
next
step
would
be
to
produce
a
draft,
so
I
think
it
should.
Obviously
a
first
draft
doesn't
have
to
cover
everything,
but
it'll
be
good
for
it
to
at
least
mention
all
the
issues
and
have
a
sketch
of
a
solution
to
any
of
them.
B
Thank
You
Donald
we're
going
to
go
on
to
the
next
thing
on
our
agenda,
which
is
the
parents
election.
Are
you
is
route
from
a
Kumar
good,
you're
here,
wonderful
in
person,
let
me
see
if
I
can
get
these
slides
up.
Please
take
the
once.
I
get
your
slides.
It
will
be
just
easily
and
thank
you
for
redoing
the
slides
for
us.
B
F
F
F
So
what
is
the
problem
that
the
draft
a
scientist
out?
It's
basically
that
the
parent
default
pattern
selection
rules
standardized
in
the
treadle
reconstruction
process
can
lead
to
situations
where
parents
for
a
particular
node
in
the
tree
shift
unnecessarily
and
I'm.
Aware
of
this?
Creating
problems
for
customers
in
the
field,
and
the
draft
here
basically
tries
to
provides
actually
two
different
solutions
which
can
be
used
for
this,
and
I
will
explain
what
they
are
and
which
one
is
preferred
as
I
go
along.
So
the
crux
of
it
comes
down
to
this.
F
Basically,
the
second
pilot,
the
lower
paragraph
on
the
page.
Ok,
so
if
and
I'll
skip
one
slide
mode,
so
it
comes
down
to
this.
So
basically,
if
you
are
building
3
number
J
and
you
have
a
situation
where
a
node
can
be
pulled
by
three
different
parents,
then
the
virtual
chooses
the
parent
for
that
particular
tree.
Is
it
use
it
orders?
All
the
parents,
in
ascending
order
of
seven
octave,
is,
is
ID
and
then
in
dec
sets
them
up
so
that
the
first
one
has
an
index
of
0.
F
The
last
one
has
an
index
of
and
not
P
minus
1
or
whatever
it
is,
and
then
the
parent
choice
is
then
made
as
J
minus
one
more
p.
Where
j
is
the
tree.
Number
and
p
is
the
number
of
parents.
So
what
does
what
problem
does
this
cause?
So,
let's
take
a
let's
take
the
example
of
this
spine,
lift
network.
That's
shown
on
the
right
side.
Here
are
basically
let's
say
that
this
we
are
talking
about
tree
number
two,
and
that
is
the
root
node
for
that
tree.
F
You
start
off
the
SPF
at
eight
and
then
it
tries
to
go
ahead.
It
tries
to
pull
the
nodes
that
are
optimally,
reachable
from
it
so
on
the
very
first
shot
is
going
to
pull
in
one
two
and
three,
because
those
are
the
nearest
nodes
that
are
attached
to
a
so
and
four
one.
Two
and
three
there
is
no
choice
of
parent,
because
it
is
the
tree
root
is
a
and
only
a
can
be
the
parent.
F
Now,
as
the
second
step,
none
will
start
looking
at
other
nodes
which
are
reachable
through
one
two
and
three,
and
now
you
will
find
that
you
can
now
reach
B
and
C.
Now,
when
you
try
to
pull
in
B
and
C
into
the
tree,
you
have
a
choice.
They
can
be
pulling
through
one
through
two
or
three.
So
that's
the
rule
that
the
rule
that
was
on
the
previous
slide
is
basically
acting
as
a
tiebreaker
for
this
particular
situation.
F
So
in
this
case,
what
you
will
do
is
you
so,
let's
say
they
in
ascending
order
of
I,
sighs
ID
there,
the
parents
are
one
two
and
three
okay,
so
one
is
considered
to
be
our
index.
02
is
considered
to
get
index.
One
and
three
is
considered
to
be
at
index
2.
So
because
this
is
32
and
you
have
three
parents
potential
parents,
you
will
compute
the
index
the
preferred
index
as
2-1
mod
3,
which
is
equal
to
1
mod
3,
which
is
equal
to
1.
So
you
will
go
with
so
the
node
at
index.
F
1
will
basically
too.
If
you
see
the
list
there
right
so
too
will
become
the
parent
for
gay
and
see
so
the
tree
looks
as
it
does
on
the
left
side
arm.
So
this
is
fine,
but
what
happens
when
let
node
number
one
went
down
so
again:
you're
forced
to
recalculate
the
three
at
that
point.
At
that
point,
what
happens
is
a
pulls
in
two
and
three
directly.
F
Now
again,
when
you
come
when
you're
pulling
in
B
and
C,
you
have
a
choice:
they
can
be
pulling
through
two
or
they
can
be
pulling
through
three
so
again
order
the
parents
in
ascending
order,
so
they
are
2,
is
at
index
0
threes,
our
index
1.
You
run
through
the
calculation,
which
is
basically
three
number
minus
1
mod
number
of
parents,
which
is
in
this
case
2
minus
1
mod
two
parents
now
so
is
equal
to
1
mod
2,
which
is
one.
F
F
I
will
talk
about
the
bogie
approaches
and
why
I
prefer
one
of
them,
so
the
simplest
easiest
way
to
solve
this
is
using
the
affinity
sub.
Here.
It's
a
sub
theory
of
the
router
capability
tlv,
which
is
already
used
in
many
other
scenario.
Culture
and
it's
published
by
the
so
in
terms
of
semantics,
is
published
by
the
planet
identifying
which
children
it
wants
to
have
a
parent
relationship
with
and
the
specific
tree
on
which
this
is
to
be
done.
It's
a
it's
quite
powerful
and
needs
to
be
used
with
certain
caveat.
F
So
the
way
the
approach
one
works
is
that
the
operated
configures
parent
stickiness
on
a
particular
preferred
mode.
So
in
the
previous
example,
if
you
go
back
to
this
slide,
it's
on
unknown
number
two
he
can.
He
can
configure
this
parents
tequila
saying
that
my
preferred
parent
in
all
three
situations
is
to
for
that
particular
level
in
the
tree.
F
So
once
once
he
configures
this,
what
happens
is
that
the
the
two
then
runs
through
a
tree
calculation
again
when
it
runs
through
the
tree
calculation.
It
does
not
use
the
default
reviews
and
at
the
same
time,
at
that
point
it
has
not
published
any
affinity
but
ignores
its
own.
Any
affinity
tells
me
that
it
might
have
generated,
but
it
because
it
has
now
a
preferred
parent
skaters.
F
It
acquires
the
so
when
you're
in
the
SPF
construction,
you
see
that
you
are
in
contention
to
be
a
parent
for
those
children
and
you
it
asserts
that
parent
relationship,
basically
so
immediately
bill
pools
DNC
in
the
previous
case
through
through
itself
and
and
it
becomes
the
parent
of
BNC
now
once
it
has
run
through
one
calculation
and
and
it
through.
It
has
waited
for
some
measure
of
stability.
Let's
say
a
few
check
a
certain
time
interval
or
something
like
that.
Then
it
goes
ahead
and
publishes
and
affinities
of
jail
me
saying
that
hey.
F
I
am
the
preferred
method
for
children,
B
and
C
on
three
number
two.
So
once
this
goes
out
in
the
network,
everybody
knows
that
too
wants
to
assert
that
current
relationship
and
they
all
bind
it
to
two
as
the
parent.
Okay.
So
subsequently,
what
happens?
So
if
you
take
the
same
situation
in
the
previous,
the
did
this
approach
one
scheme
implemented.
What
happens
is
that
when
one
goes
down
you
are
even
before
one
went
down.
F
Presumably,
when
two
was
configured
with
the
parent
stickiness,
it
went
ahead
and
published
an
affinity
sub
tlv,
which
went
to
every
node
in
the
network.
So
even
when
one
went
down,
nobody
is
using
the
default
trail.
Calculation
anymore.
They
already
know
that
two
wants
to
be
the
parent,
because
it
has
sent
its
affinity
Kelby
and
it
which
is
valid
right
now.
So
everybody
will
pick
you
as
a
preferred
parent,
so
the
parent
will
not
shift
23
when
one
goes
down,
basically,
which
is
what
you
wanted
to
achieve.
F
So
there
is
a
second
way
in
which
you
can
do
this.
This
approach
is
a
little
bit
harder
to
implement
arm,
and
so
the
way
its
proposed
is
that
so
the
draft
presents
are
modified
form
on
the
Dijkstra
algorithm.
Where,
basically
you
are,
you
are.
If
multiple
parents
can
pull
in
a
particular
child
at
the
same
optimal
cost,
then
you
can
insert
a
policy
filter
in
the
free
construction
thing.
What
which
will
drive
your
parents
election?
Basically,
the
policy
filter
is
abstract.
F
You
can
put
whatever
you
can
hold
it,
however,
you
want,
but
the
what
the
draft
cops
about
is
that
you
make
it
match
your
previous,
stable,
pre-construction.
Okay,
so
you
can
begin
with.
Let's
say
you
started
off
with
till
the
default
rule
rules
and
your
tree
look
like
like
it
did
on
the
left
side
of
the
slide
gear.
So
once
you
know
that
too
was
the
parent
in
subsequent
recalculations.
You
will
continue
to
be
maintained
to
as
the
parent,
because
you
are
poly,
because
your
policy
is
always
referencing.
The
previous
table
draft
reconstruction.
F
So
so,
by
doing
this,
you
basically
preserve
whoever
was
the
parent
when
the
first
drill
parents
election
was
applied,
so
this
will
also
solve
the
problem.
But
basically
here
the
difficulty
with
this
approach
is
that
you
have
to
you
have
to
kind
you're
kind
of
doing
the
doing
a
kind
of
time
synchronization
in
a
distributed
Network.
Basically,
so
this
is
harder
to
do
and
with
some
caveats
it
might
work,
but
it's
probably
best
suited
for
like
small
to
mid-size
networks,
but-
and
this
slide
indicates
how
that
works.
F
So
when
one
went
down
because
too
was
the
preferred
parent,
you
continue
to
latch
22
in
your
subsequent
recalculations.
You
policy
guides
that,
but
still
between
the
two
approaches,
approach,
one
is
substantially
better
preferred
arm
and
because
it's
guaranteed
to
work
everything
that
you
need
to
implement
approach,
a
is
already
there
in
trill
and
no
change
to
the
protocol
is
needed,
but
I
still
put
approach,
be
there
as
a
backup.
F
However,
there
are
in
the
published
version,
there
are
a
few
mistakes
and
logical
inconsistency
that
I
need
to
fix.
I
have
a
private
draft
which
I
will
probably
upload
after
the
after
I.
The
current
idea
on
one
of
the
questions
I
had
was
you
know:
should
we
carry
approach,
be
all
the
way
till
the
end
or
just
remove
it
and
knew
that
approach
a
arm?
F
The
other
thing
is,
the
policy
driven
up
mechanism
listed
in
approach
be,
can
be
used
in
some
other
situations,
particularly
if
you
want
some,
there
can
be
alternative
default
parents
election
groups
it
so
you
can
probably
come
up
with
other
formulas
which
will
try
to
do
a
better
job
of
distributing
the
links
on
between
the
parents,
the
children
between
the
parents.
Basically,
if
there
is
any
approaching
interest
in
pursuing
that,
maybe
we
could
probably
use
approach.
F
B
E
B
Or
is
that
the
original
affinity
Fieldy
that
we've
had
a
mother
okay?
So
the
change
is
really
which
is
this
an
application
or
since
we're
not
changing
bites
or
is
it
just
a
mechanism
for
changing
parents
election,
in
other
words,
we're
using
the
same
affinity
tale
be
or
type
a
we're
changing
the
paragraphs
in
RFC
7780?
Is
that
what
I
understand
correctly?
No,
it's
just
a
blinding
the.
F
F
B
So
it's
it
sounds
like
an
informational
I'm.
Sorry,
if
it
you
know,
you're,
not
changing
anything.
It's
an
informational,
good
idea
on
how
to
prevent
your
customers
from
flapping,
but
if
you're
not
changing
anything
in
the
in
the
dress
sounds
like
if
you've
got
customers
that
have
problems.
This
is
a
real
important
thing
to
look
at,
and
you
know
I
would
state
those
things
to
the
the
list
and
indicate
the
problem.
You
know
underline
the
problems
you're
having
with
customers
so
that
we
can
consider
that
in
the
process
of
adoption
it
sounds
like
the
custom.
B
F
I
should
like
there
was
basically
a
presentation
in
my
previous
job.
F
B
B
I
F
I
B
So
so.
I
Changes
from
the
last
version
it
ends,
I
see
it
now.
It
only
supports
full
directory
information
because
how
to
do
push
was
unclear
and
the
this
is
sade
stuff
in
there,
which
was
pulled
visually.
These
are
all
results
of
my
comments
on
the
on
the
previous
version
and
exactly
what
macular
as
it
uses
for
where's
multicast
frames.
I
It's
been
changed
to
the
all
our
bridges
multicast
at
us.
Let's
see
this
is
this
is
a
more
changes
in
this
version.
If
you
get
a
native
multi-destination
frame
from
a
non-smart
n-no,
then
you
have
to
send
the
krill
data
package
to
the
campus
and
also
out
of
port,
to
the
look
that
back
to
that
same
length,
so
that
the
smart
end
note
on
the
link
will
pay
attention
to
it
so
easily.
I
B
I
Were
the
previous
fundus
didn't
really
cover
all
the
cases
correctly,
so
there
is
another
graph
which
is
the
directory
assisted
encapsulation,
which
might
be
working
with
last
call.
At
the
same
time,
the
topics
are
similar
enough
that
it
might
be
convenient
to
be
able
to
review
them
both
at
the
same
time
and
I.
B
Of
our
our
directory
assistance
or
well,
we've
come
to
the
end
of
the
session
and
thank
you
for
your
attention.
We
will
call
it
into
the
clothes
Donald
if
you
will
post
that
that
any
questions
to
the
draft
and
then
I
will
post
later
in
the
week
of
working
group,
last
call
blue
sheets
please
and
who
are
if
you
will
post
a
comments
about
your
your
you
know,
general
some
general
idea
of
customer
needs.