►
From YouTube: IETF94-OSPF-20151104-1300.webm
Description
OSPF meeting session at IETF94
2015/11/04 1300
A
We're
gonna
go
ahead,
get
started
droop,
you
wanna
says
you
sent
me
an
email
that
you
were,
that
there's
gonna
be
some
remote
attendees,
okay,
I'm
to
the
mic.
Yeah
come
for
the
night,
but
one
thing
everybody
have
you
see
the
note
wells
and
any
IP?
Are
you
know
about
your
and
you're
participating?
This
discussion.
D
I
am
true
from
Huawei,
so
in
association
with
I
sock
back
door,
who
always
hosting
a
remote
hub-
and
we
did
it
yesterday
for
pc
working
group
session
today
we
are
have
40
SPF
and
the
plenary
afterwards.
So
our
aim
was
to
get
the
bunch
of
vendors
who
are
anyway
there
in
bangalore,
but
who
had
no
opportunity
to
come
to
the
ITF
who
have
worked
on
a
sphere
for
interested
in
those
protocols
to
get
in
the
same
room,
go
over
the
agenda
beforehand.
D
Look
at
how
the
working
group
progresses
and
later
on
once
the
working
group
is
over,
discuss
what
they
have
landed.
They
have
any
questions:
how
do
they
post
the
questions?
What
are
the
process
in
which
the
idea
of
works?
So
apart
from
the
windows,
we
also
had
some
university
students
who
are
interested
in
this
process.
So
it's
going
to
be
a
good
experiment
and
we
will
get
back
to
the
working
group
based
on
the
feedback
that
we
have
we
are
getting,
and
this
is
something
that
anyway,
I
TF
is
interested
in
tomorrow
morning.
D
There
is
a
barb
off
on
how
to
conduct
remote
hubs,
and
how
can
we
make
this?
This
process
go
smoother,
so
we
will
be
sharing
each
other's
experiences
of
various
places
like
in
Latin
America
and
in
India,
where
this
experiment
is
already
started.
Thank
you.
So
please
say
your
name
on
the
mic
more
clearly
the
same.
Thank
you
is.
A
C
E
E
No
dad
min
tag
had
a
lot
of
Disqus
from
iesg.
The
good
news
is
all
of
this
is
cleared,
so
it
should
pretty
much
get
pushed
out
of
is
g
soon.
Spfd,
we
finished
protein
rope
last
call
AC
had
an
open
question
end
of
last
month.
We
need
to
nudge
the
authoress
to
answer
on
that
and
after
that
we
can
request
for
publication.
E
Yeah,
so
these
are
the
two
which
are
very
close
to
the
last
call
of
tardiness
on
our
part.
There
was
a
lot
of
other
things
we
were
trying
to
push,
but
pretty
much
gonna
push
this
first.
One
is
for
for
use
cases
where
the
devices
don't
have
ipv6
capability
at
all.
So
it's
a
very
simple
draft.
The
two-part
metric
is
also
pretty
simple,
has
had
a
lot
of
discussion.
I
think
I
saw
Jeffery
in
the
room.
A
E
The
active
documents-
actually
you
will
see
presentation
on
most
of
this
today-
was
pure
v3,
extended-release,
say
AC
respondus,
with
the
new
concept
of
sparse
mode,
which
is
basically
to
simplify
the
migration.
You
don't
have
to
worry
about
the
old
versus
new
and
you
can
only
again
I'll.
Let
AC
do
more
justice
to
that
segment.
Routing
drafts
also
have
updates.
Today.
The
v2
one
has
some
conflict
resolution
updates
and
we
also
have
update
on
the
yang
model
direct.
E
E
So
we
would
love
to
have
some
more
discussion
on
the
working
group
and
then
we
can
last
call
that
document
next
flow
spec
was
presented.
I.
Think
a
couple
of
five
years
ago
there
were
two
competing
versions
to
merge
them
into
one.
We
have
some
prototypes
coming
along,
so
we
can
probably
share
some
news
on
that
in
the
mailing
list.
Entropy
level
is
also
it's
waiting
on
the
MPLS
working
group
to
proceed
further
and
will
progress.
That
document.
E
A
bunch
of
new
documents
first,
one
is
linked,
overload
a
lot
of
lot
of
passionate
discussion
on
the
mailing
list.
If
you
have
been
following
in
the
last
month,
each
bit
is
the
hospital.
Was
it
offers
POV
3
router
bit,
so
that
also
was
not
accepted.
Encapsulation
capabilities
is
all
sorts
of
tunneling
which
we
are
trying
to
do.
It
captures
in
our
ILS
SE
now
what
what
encapsulations
use
support.
So
these
are
all
new
working
group
documents.
E
There
are
some
other
documents
progressing
in
other
working
groups.
We
actually
had
a
discussion
on
the
beer,
for
example,
where
we
agreed
that
we
will
progress
it
in
the
beer
working
group,
because
most
PF
changes
are
small.
Beer
is
still
going
through
some
major
architectural
stuff,
so
that
will
progress
there
and
there
are
some
documents
sitting
in
sea
camp
again
with
our
agreement
from
that
I
think
that's
pretty
much
it.
The
next
up
is
a
sea.
Yes,
you
do
the
major
recharter
updates.
E
C
Charter
is
written
in
such
a
way
that
it
is
self-perpetuating
and
I,
don't
see
any
need
to
make
it
too
specific.
So
what
I
actually
did
is
I
actually
took
all
the
work
we
finished
and
marked
it
done
and
I
and
including
stuff
that
was
started
and
finished
before
it
even
hit
a
milestone.
Ciety
I
got
us
back
to
the
future
in
that
and
added
the
new
stuff
we're
doing.
Basically,
we
have
four
tracks
a
little
bit
more.
C
We
have
well
four
tracks,
and
I
think
I
was
I
put
the
these
are
things
I
put
into
the
charter,
the
newt.
The
things
that
aren't
done
are
all
part
part
of
these
for
try.
Let
me
go
through
what
they
are.
Basically
there's
protocol
extensions
using
TL
fees,
we're
pretty
close
to
that
with
ospf,
be
too
that's
on,
that's
I
mean
we're
actually
gonna
fit.
It
sits
in
the
RFC
queue
and
we
probably
won't
need
to
do
any
more
of
that.
Ospfv3
I've
got
a
president.
C
I've
got
a
presentation
today,
where
I've
simplified
it
more
or
less,
given
an
option
that
doesn't
have
any
backward
compatibility
implications,
other
the
implications
for
the
functions
that
use
that
now.
The
second
track,
after
that,
after
the
extensions,
are
applications
that
make
use
of
the
new
TLDs
or
the
RFC
4970,
the
router
ospf
router
information
LSAs,
you
know
those
are
for
both
and
these
these
are
generally
most
all
these
applications
are
applicable
across
both
ospf,
be
to
in
ospf,
be
three.
C
This
would
be
the
place
where,
if
we
did
ever
have
an
overlapping
part
of
part
of
our
meeting
with
I
esaias
the
application,
this
applications
area
tends
to
be
pretty
standard
between,
I
should
say
between
the
three
protocols
we
have
OS
52,
ospfv3
and
I
sighs
and
finally,
we
have
Matt
protocol
enhancements.
These
are
you
know
little.
You
know
big
or
little
or
big
optimizations
changes
like
the
experimental,
ttz
or
small
ones
like
what
we
did
with
the
monnet
enhancement
to
the
database
exchange.
Things
like
that.
C
You
know
things
that
actually
affect
the
protocol,
mechanics
of
ospf
the
draft
you're
going
to
see
from
Padma
the
stub
neighbor.
This
would
fit
kind
of
in
him,
I
would
say
in
a
medium
size,
it's
not
a
small
enhancement
but
not
as
cataclysmic
as
t
TZ.
Maybe
that
was
the
wrong
adjective,
but
not
is
not
his.
That
is
big
as
t
TZ,
ok,
you
know,
and
then
then
we
have
security.
C
We're
done
for
now,
but
I'm
sure
there'll
be
additional
requirements.
Additional
things.
For
now
we
have
completed
all
the
drafts
to
support
the
scarp
requirements
for
manual
key,
so
I
I
think
we're
I
think
we're
pretty
much
done.
You
know
there's
a
lot
of
other
considerations.
The
focus
IV
focus
on
private
privacy.
Maybe
we'll
have
to
have
encryption,
we'll
have
to
figure
that
out
for
multicast
I,
don't
know
multi,
you
know
key.
You
know
you
know
in
which
would
include
something
to
do
multicast
key
distribution-
I
guess
it
wouldn't
have
to.
C
But
that's
that's
another
one.
You
know
that
there
was
always
this
discussion
about
whether
now
that
we
have
manual
key
distribution,
would
we
do
something
for
automated
key
distribution
as
well
and
and
that
then
the
final
track
is
manageability
and
I?
Think
we,
you
know,
we
have
the
yang
model,
I!
Think
that's
about
it.
For
now
there
might
be
I
mean
I'm
sure
they'll
be
augmentations
to
that
as
well.
C
Okay,
here's
where
we
are,
the
draft
has
been
pretty
stable.
We
got
some
good
reviews
out
of
people
like
David
I
saw
he
was
it
walked
in
and-
and
it
seems
like
it's
really
ready-
I
mean
it
looked
seemed
like
it
was
pretty
well
baked.
Even
even
we,
a
couple
of
us
have
even
looked
at
what
it
would
take
to
implement
it
and
there's
still
no
implementations,
which
is
unfortunate.
C
So
even
though
they're
required
these
extensions
are
required
for
segment
routing
and
when
I
looked
at
it
and
we
looked
at
what
it
would
take
to
do
it,
the
big
barrier
in
terms
of
it
seems
to
be
the
migration,
because
the
migration
is
quite
complicated,
with
the
three
step
process.
So
what
I
did
in
this
version
is
something
people
want
it.
Well,
here's
the
previous
migration
paths.
Of
course
you
could
do
a
Greenfield
network.
C
Then
you
can
take
off
the
configuration
of
the
of
the
instance
that
you're
trying
to
migrate
from
so
and
similarly,
the
draft
includes
a
single
instance
approach
to
this,
where
you
first,
you
advertise
both
the
extended
lsas
and
the
existing
L
essays,
but
you're
still
using
the
SPF
I
mean
your
Studios
in
the
existing
LSA
is
the
legacy.
Let's
call
them
legacy.
L
essays,
that's
right,
call
them
in
this
for
the
SPF,
then
you
switch
over
to
use
the
extended
L
essays
for
the
SPF,
make
sure
everything
works.
And,
finally,
you
purge
the
legacy.
C
C
So
what
I
added
is
a
new
mode,
it's
the
sparse
mode
and
basically,
what
is
what
it
is.
Is
you
you're
always
using
the
legacy?
L
essays
and
what
you
do
is
you
use
the
extended
essays
only
for
the
new
functions
and
backward
compatibility
is
the
responsibility
of
whatever
function.
It
is
you
know
just
like
in
segment
routing.
We
have
to
cover
back
backward
compatibility
or
whatever
everyone's,
so
you
only
need
to
include
the
TL
these
that
are
necessary
for
these
new
functions.
C
C
Compliance
you
could
implement
just
the
sparse
mode.
This
is
to
try
and
get
make
sure
that
this
migration
is
not
a
implementation
barrier,
adoption
barrier
or
you
could
implement
the
full
support
and
the
sparse
mode
or
you
could
do
the
whole
ball
of
wax.
You
could
do
the
full
mode,
the
all
three
steps
in
the
migration
and
the
minimum.
You
always
have
to
do
the
sparse
most
for
new
new
functions.
C
Now.
What
I'm
I'd
like
to
do
is
I'd
like
that,
since
I,
don't
I,
don't
see
what
I'd
like
to
do
is
just
take
this
complicated
mode
where
you
do
it
with
one
instance,
I
I'd
almost
like
to
deprecate
it.
It's
not
really
very
I,
don't
know
if
you
need
to
call
it
deprecated
cousin
always
implemented
it,
but
I'd
like
to
take
it
out
of
the
draft.
C
Why
do
I
keep
doing
that
here?
I
already
went
through
this,
the
complexity
and
because
we
don't
have
implemented
after
two
years
it
it's
a
it's
represented
a
barrier
and
the
other
thing
is
the
multiple
instance
route.
You
know
the
other
mount
route
to
have
an
instance
generating
the
old
LSAs
and
when
generating
the
new
lsas.
That's
kind
of
well
understood,
because
that's
what
you
do
anytime
you're
migrating
between,
so
you
want
to
migrate.
Your
network
from
OspA
from
is
is
to
ospf.
C
You
would
do
the
same
thing,
so
it's
it's
pretty
well
understood
and
really
by
doing
it
in
a
single
instance,
you're,
really
not
saving
yourself.
That
much!
You
know
you
know
implementing
this
complex
because
I
as
ospf
is
I/o
bound,
so
I
and
most
of
it
you
know
it.
Flooding
is
responsible
for
most
of
the
the
scaling
issues,
so
an
integrated
migration.
In
my
opinion,
it
doesn't
doesn't
doesn't
say
that
much
over
the
two
instances.
C
There's
another
reason
to
it:
actually.
Actually,
if
you
do
separate
instances,
you
actually
have
better
isolation
between
the
old.
Oh,
the
process
are
the
ospf
instance
running
the
originating
the
legacy
essays
and
though
SPF
instance
originating
the
extended
lsas,
because
both
are
running
espia
f's
independently
and
you
can
and
they
both
have
their
own
local
rib.
So
you
can
really
verify
that
they're
getting
the
same
things
before
you
and
look
at
look
at
them,
whereas
in
the
case
where
you
just
switch
over,
you
don't
know,
you're
doing
a
leap
of
faith
that
it's
gonna.
C
You
know
it's
going
to
going
to
have
the
same
results
now
in
some
platforms.
If
you
do
separate
instances,
you
even
get
them
as
separate
processes,
so
you're
actually
getting
better
scaling,
because
you're
you're
spreading
the
load
over
more
processes
and
just
because
of
scheduling
and
the
reality
of
how
things
are
implemented.
C
You're
going
to
it's
going
to
be
possibly
even
less
than
the
less
of
a
load
with
multiple
instances
and
I
can
say,
there's
less
chance
of
bugs,
because
you
know
this.
This
code
is
only
going
to
be
implemented,
I
mean
implemented.
It's
not
going
to
be
executed
many
times
a
person
would
only
do
this
migration
once
unless
they
weren't
successful.
C
C
A
C
C
C
Now
one
thing
that's
happened.
Is
we've
had
a
lot
of
discussion
on
what
to
do?
If
you
get
complexing
information,
especially
with
respect
to
a
mapping
server
because
of
a
mapping
survey,
you
can
have
multiple
guys
advertising
the
same
prefix
and
how
to
resolve
conflicts
that
has
risen.
There's
a
long
discussion
on
the
is
is
late
list
and
we're
kind
of
waiting
for
that.
Cole,
less
and
less
Ginsburg
was
good
enough.
C
To
put
it
all,
put
some
ideas
in
the
draft
and
we've
started
a
discussion
spring
on
how
we're
going
to
get
it,
get
everyone
to
degree
how
we
resolve
conflicts
when
they're,
especially
when
different
routers,
are
advertising
the
same
prefix
with
different
segments.
I
IDs,
you
know
the
segments,
the
SID.
If
you
call
from
segments
routing
nodes,
SIDS
and
prefix
it's
our
global
are
are
globally
global.
It
should
be
globally
consistent
within
a
routing
domain
and
finally
Oh.
Finally,
one
thing:
one
thing
for
the
prefix
SI
prefix
length,
attributes
draft.
C
We
had
some
early
allocations
for
the
exists.
I
mean
for
this
draft,
the
segment
routing
we
had.
We
had
early
allocations
for
ospf
router
information
LSA,
but
we
didn't
have
them
for
the
prefix
link
attributes,
because
that
was
a
draft
itself.
So
those
registries
had
been
been
created,
but
now
they've
been
created
due
to
the
draft
advancing.
So
we
got
early
allocations
of
these.
There
are
some
their
applications
based
on
prefix
length,
attributes
that
are
also
going
to
request
early
applique
early
allocation.
I.
Think
we'll
do
that.
So
we
can
just
keep
those
things.
C
This
is
nobody's
implemented
this
that
I
know
about
I'm,
hoping
now
that
I
have
the
sparse
mode.
It
will
do
some
act.
It'll
motivate
people
to
implement
those
pfv
three
segments
routing
draft.
There
was
one
open
source
implementation
in.
I
think
it
was
in
budapest
I,
don't
remember,
then
the
name
of
it
right
off
the
top
of
my
head,
but
they
had
we've
exchanged
some
emails,
so
I'm
gonna,
try
and
see.
C
Yes,
okay!
Yes,
yes,
for
those
52
will
do
that.
We're
sorta,
although
it's
not
I'll,
say
one
more
thing
about
that,
although
it's
not
necessary
because
we've
we've
taken
that
conflict
resolution
out
into
a
separate
draft,
I
think
it
would
be
not
necessarily
have
that
draft
have
have
all
the
people
that
discussion
over
and
make
sure
that
you
know.
We've
incorporated
that
into
the
into
our
implementations
and
that
would
that
would
demonstrate
that
it
was
sound.
C
F
Testing
their
again
from
cisco
systems,
I'm
here
to
present
update
for
the
03
y,
a
model
for
ospf
on
behalf
of
other
co-authors
of
the
job
and
also
other
color
on
the
slave
on
on
the
model
itself,
excellent
piece
so
agenda.
I
will
tell
you
that
we'll
talk
about
change,
we
make
from
different
from
0
1
for
the
base
model
and
also
different
for
the
SL
model
from
01,
and
we
discount
that
a
new
OS
k,
pft
module
and
we'll
talk
about
some
next
steps.
F
Now
we
add
Helen
as
the
co-author
of
the
base
model,
thanks
for
her
continued
support
and
come
in
and
everything,
and
so
we
don't
have
a
lot
of
change
this
time.
It's
basically,
what
we
have
here
are
the
first
change
you
make
is
that
before
we
have
extra
three
identity
we
have
ospf
and
from
our
spear
we
divide
4-spd
23.
F
We
think
it's
nice,
because
now
we
can
use
some
trick
to
to
Reverend
Bob
OS
people
between
v3,
so
yeah,
but
it
turned
out
that
when
people
implement
it
because
they
have
no
idea
that
issues
so
you're
supposed
to
instantiate,
just
OS,
pvt
or
or
sp3
type,
but
not
worse,
gift
I,
so
they
do
have
a
constant
confusion.
So,
after
some
discussion
we
decide
to
just
we
were
we
removed
in
the
media,
one
just
half
hours
baby
to
agree
three.
Instead,
I'm
also
an
arching
cream
make
is
that
the
area
ID
type?
F
We
have
a
union
before
and
it's
a
pop
OH,
the
third
to
be
numbers
and
bonaparte,
and
after
some
discussion
we
think
that
is
so.
First,
not
all
malaysia
for
both
and
and
the
mips
only
support
golf
or
so
after
some
discussion
we
decided
to
just
keep
the
report
also
and
so
now
left
to
the
American
station.
They
can
get
work.
What
they
want
to
change
it
to
32-bit,
for
representation
is
done.
F
Many
men
stationed
right
in
the
tagging
plantation
anymore
and
do
some
correction
about
the
array
ID
type,
because
before
we
also
have
some
incorrect,
some
mystic
there
sometime,
we
use
express
a
new
power
cord
or
explicit
UN
32.
Now
we
change
all
to
the
standard
error,
ID
type
which
map
to
it
on
the
quad
and
because
we
have
our
things
that
we
have
a
new
PF.
F
We
have
a
BF
dmm
owners
submitted
in
last
ITF
so
because
of
that
is,
if
I
some
grouping
so
welcoming
that
we
move
our
BF
d
supporting
in
the
base
model
into
its
new
module.
This,
in
order
to
avoid
anyone,
does
just
use
itäôs
been
module
which,
if
it
is,
if
it's
forced
to
import
vfg
module,
then
they
all
need.
Do
it
all
the
time,
even
if
the
platform
say
are
low
and
poppin
does
it's
appropriate
p.
So,
by
moving
to
an
arm
of
your,
then
we
solve
the
problem.
F
C
Got
one
thing
to
say
about
the
the
area
type
most
implementations
support
either
a
number
or
a
dotted
quad
in
the
CLI.
But
since
this
is
a
programmatic
and
information
interface,
we
thought
well.
We
might
as
well
just
pick
one
and
makes
it
a
lot
a
lot
simpler
and
and
doesn't
require
us
to
keep
which
one
the
person
you
know
which
one
it
was
configured
in
a
matter
of
fact.
I
know
that
most
implementations
support
multiple.
F
F
So
now
all
notification
basically
contained
some
stand
apart,
so
we
have
always
have
our
notification
header
instead
I'm
going
to
tell
you
which
instant
ospf
that
generate
the
trap
or
the
duplication,
and
then
we
have
an
outcropping
to
say
that
which
interface
is
generating
the
trap,
and
then
we
have
an
outcropping
for
which
neighbors
doing
that
and
then,
of
course,
each
notification
have
some
new
king
spesific
field
in
the
next
one,
for
example.
So
here
I'll
just
show
your
example
that
we
have
interface,
state
change,
notification
and
devastation
notification.
F
You
see
the
blue
part,
we
they
all
share
the
routing
instant
header
to
indicate
which
instant
is
coming
from,
and
then
we
have
the
red
part
that
tell
you
what
cup
interface
we
rolling
normal
interface
shambling
and
for
the
interface
state
change.
That's
it
our!
Then.
You
also
need
to
stay
for
the
interface
layer
also,
but
for
a
neighbor
State
changed
notification.
We
also
have
that
the
neighbor
State
grouping
that
you
can
eat
help
the
neighbor
information
and
then
fall
with
its
own
state
information.
F
So
all
the
art
is
just
to
expect
example,
but
all
the
other
notification
now
phone
is
similar
thing,
similar
white.
So
it's
more
structured.
We
also
cleaned
up
some
of
the
other
before
I'll
say,
for
example,
in
some
links
election
cut
before
we
can
technically
you
can
have
both
you
can
genuine
location
with
same
interface.
Virtual
laying
and
sharing
together
right
now
make
the
choice.
Today
is
more
correct.
F
Next
one,
please:
okay,
now
ospf
sr
so
in
last
ITF
for
the
spring,
meaning
that's
discussion
going
on
24
the
need
of
her
Pollock
OS,
r,
GP
Brock,
so
in
the
hands
afterwards,
they
make
a
change
so
in
the
a
young
model
to
support
purple
srgb.
So
we
make
the
change
accordingly.
In
a
nice
map
as
our
model,
the
vessel
at
new
bra
container
current
protocol
srgb
under
the
instance
containers
and
is
a
feature.
F
F
Ok,
next
one
so
OS
with
bfg
are
so
it
is
similar
to.
I
saw
that
it
is
a
new
module,
but
it's
in
the
same
ospf
for
yang
draft
document.
We
create
both
a
Korean
state
and
convict
state
for
the
VFP.
The
major
thing
that
now
we
need
to
use
the
grouping
from
vfds
that
they
also
taken
before
we
can
convict
EFE
on
off
on
an
interface
in
the
PFE
draft
yang
draft.
They
also
allow
per
protocol
configuration,
have
some
proofing.
F
F
C
This
was
a
cylinder.
This
was
just
sort
of
a
discussion
with
the
BFD
module
to
whether
or
not
you'd
configure
these
things
within
under
BFD
for
the
interfaces
or
whether
some
a
lot
of
implementations
had
them
in
the
protocol.
So
this
was
a
compromise
you
and
if
you
have
this
feature,
you
can
do
it
either
way.
Yeah.
F
So
so
right
now
we
create
a
hope.
We
grow
on
continues,
BFD
our
grouping.
But
if
you
look
at
be
a
Fijian
model,
they
actually
have
an
olive
grouping,
but
we
cannot
use
it
because
we
use
it
that
way.
We
cannot
control
where
the
BF
is
and
if
you
are
not
myself,
so
if
they
make
a
change,
maybe
we
will
just
make
use
of
your
groupies,
but
right
now,
that's
what
we
have.
Ok
next
one,
please
so
next
steps,
so
why
what
we
are
looking
at
is
more
mission
effort.
F
So
if
anyone
know
of
any
vendor
is
doing
vision,
it's
good
to
know
so
that
we
can
talk
about
that
and
try
to
see
if
we
can
test
it
out,
see
how
it
works
and
for
the
final
model,
part
wheel
is
no
for
loss
from
Las
idea
of
this
pre
stable.
We
look
at,
though,
do
we
need
more
state
and
we
look
at
past
actually
and
it
turned
out
the
policy
trough,
there's
something
from
a
poet.
F
That's
an
issue
there
that
it
doesn't
specify
the
you
can
reduce
away
from
OSP
poker,
but
not
you
don't
know,
which
instance
so
they'll
fix
it
first,
so
we're
going
to
talk
to
them
and
when
they
are
ready,
then
we
could
make
use
of
that
and
at
the
distribution
and
consecrated
things
in
ospf.
Another
thing
we
didn't
get
to
do
is
now:
we
won't
use
at
the
inheritance
container
and
the
content
or
complete
right
now.
We
just
have
a
container.
We
want
to
sterilize
on
as
a
features
when
you
want
inheritance.
F
How
does
it
work
and
we
also
get
some
new-
requires
about
adding
things
to
the
OSP
model
just
before
ITF
I
think
people
asking
whether
we
should
at
ntr
no
I'm,
not
a
Mountie
and
also
and
also
worker,
which
should
at
the
dopamine
tech,
so
we
locate
it
after
we
come
back,
go
back
up
after
this
idea,
either
as
a
features
or
module.
We
know
what
we'll
figure
it
out
and
so
what
we?
F
What
what
the
other
thing
that
could
affect
our
malo
is
opera
stage
payment
today,
mark
group
there's
some
talk
about
what
this
should
be
was
to
data
store
or
use
the
OSI
model.
So
you
check
to
see
whether
we'd
have
a
consensus
and
see
if
it's
close
or
weeks
till
we
do
something
and
of
course,
we
also
looking
for
big
feedbacks.
Actually,
we
saw
some
feedback
from
island
in
the
very
least
we're
going
to
prior.
To
that
any
question:
that's.
C
Okay,
thank
you.
I
just
want
to
take
a
seal
in
them
again.
I'd
want
to
encourage
people
to
review
and
comment
on
the
policy
model
in
the
routing
working
group,
because
we're
kind
of
looking
at
I've
been
in
discussions
with
my
comments
and
see
if
some
of
you
come
up
with
the
same
comments
independently.
G
G
G
G
Secondly,
usually
the
spokes
have
limited
capacity,
so
they
are
actually
storing
a
lot
of
information
which
is
actually
unnecessary
for
them
from
their
cubs
perspective,
handling
and
flooding
of
large
database
to
each
one
of
the
spokes,
actually
limits
the
amount
of
spokes
you
can
actually
handle,
and
today
you
know,
SPF.
There
is
no
mechanism
to
aggregate
entry
area
route
or
even
filter
them
x.
G
G
The
hug
routers
will
define
a
new
type
of
neighbor
relationship
with
their
spoke
and
actually
send
them
and
modified
router
lysate
with
limited
information
just
enough
for
what
he
needs
to
actually
forward
their
traffic
and
we'll
see
a
little
bit
more
about
what
this
modified
form
of
the
router
is
a
is
for
now
the
router
I
say
the
local
writer
like
there's
a
coil.
It
is
going
to
have
a
default
route,
which
is
an
able
to
hug
Oh
disposed
to
translate
this
graphic.
A
G
G
G
G
So
the
link
the
local
router
laces
are
going
to
be
flooded
on
the
stub
link
zones
to
the
Stokes
and
a
hug
receiving
its
own
local
router
I,
say
not
on
a
stepping-stone,
actually
can
detect
this
configuration
and
act
upon
it
that
could
be
by
policy
either.
We
can
actually
revert
back
to
the
normal
ospf
process
or
login
error,
a
hug
receiving
and
now
that
hubs
local
route
real
estate
can
also
act
upon
it
if
it
receives
it
on
a
normal
link,
which
means
that
we
miss
configuration.
G
However,
if
it
receives
it
on
a
normal
stub
link
I'll,
then
you
should
just
acknowledge
it
to
make
the
recency
come
out,
but
you
should
not
use
it
for
its
normal
is
a
a
hub
should
not
actually
receive
the
local
router
LLC
on
any
other
of
its
normal
length.
So
I'm
going
to
illustrate
that
in
the
next
slide.
G
So
if
you
look
at
this
example,
imagine
hub
to
using
the
green
arrows
is
actually
transmitting
the
normal
router
all
I
see,
as
we
know
it.
Only
stub
links
to
the
spokes
he's
actually
selling
the
verified
privacy,
which
is
the
local
one,
with
the
default
prefix
the
spokes,
on
receiving
it
I'm
going
to
reflect
just
as
more
or
less
p
process.
However,
if
you
look
at
hub
3,
which
was
also
upgraded,
it
will
actually
acknowledge
it,
but
it
will
not
forward
flooded
any
further.
G
However,
if
we
look
at
the
other
case
of
may
need
a
hug
one
which
is
misconfigured,
you
will
actually
receive
the
local
router
LSA
and
try
to
reflect
it
back
on
these
other
links.
Now,
at
this
point,
if
the
local
router
oil
is
coming
from
hub
tube,
you
can
actually
act
upon
it
because
its
own
router
listen.
You
can
detect
that
he
got
it
a
very
normal
link
and
that's
not
that
signals
in
this
configuration
and
if
this
route
girl
I
see
worth
actually
coming
from
HUD
three
helped.
G
C
C
E
A
C
C
G
C
G
G
C
I
I
G
I
C
C
Many
people
think
this
is
a
problem
we
should
solve
here,
assuming
I,
think
I
think
it's
good
one
Yeah
Yeah
right
right,
okay,
okay,
so
we'll
take
its
list.
There's
one
thing:
I
wanted
to
say
and
aaliyah
kind
of
hit
on
it
too
and
I'm
gonna
take
an
action.
I
think
I
have
IPR
on.
This
was
a
different
company.
C
My
prior
job,
so
send
I
mean
I,
think
it
may
be
IPR
I'm
not
going
to
make
the
judgment.
It's
close,
it'sit's
again
solving
the
same
problem.
I,
don't
know
if
the
mechanisms
aren't
close
enough
will
let
well
let
them
decide
right.
So
I'm
gonna
take
that
action.
I
think
there
are
other
people
that
may
have
something
close
enough,
that
you
should
also
inform
you.
You
know
see
if
your
companies
need
to
want
our
intent
to
disclose
something
on
this.
So
that's.
E
C
E
A
H
I'm
going
talks
about
the
ospf
cheating
attributes,
two
years
and
years
case,
what
we
are
trying
to
achieve
here
and,
on
behalf
concert
next
letters.
So
over
the
years
we've
defined
many
many
many
link
attribute
advertised
as
some
key
attributes.
We
always
see
as
the
sub
teal
with
stealing
TV
double
time.
Traffic
engineering
information-
that's
like,
so
we
had
many
discussions
on
the
list.
What
exactly
t
topology
is
and
I
think
RFC
3630,
it's
very
clear.
If
link
is
described
in
T,
let
us
say
it
is
part
of
T
topology,
so
it
is
there.
H
J
H
J
H
Juniper,
listen
yeah
next
slide,
so
thanks
to
you
think,
is
coming
up.
We've
got
new
unified
container
to
advertise
new
link
attributes
and
it's
progressing
quite
well.
Cc
shared
with
us.
It's
getting
to
be
published
Watson,
so
the
important
point
you
don't
have
to
make
I
don't
make
it
part
of
tea
or
HP
technology.
It's
really
up
to
receiver
to
decide
next,
like
listen.
H
So
there
are
many
different
solutions.
Are
things
we
could
use
this
information?
The
LFA
is
one
come
to
mine
and
we
start
the
discussion
about
three
years
ago.
How
can
we
use
information
available
today
in
T
database
for
things
like
ellen
page,
which
doesn't
use
to
eat
apology
at
all
so
finally,
and
what
I'm
presenting
to
you
its
ability
to
copy
the
attributes?
H
You
are
interested
in
the
first
comment:
amount
of
SSR
lg's
and
in
virtual
delay,
even
directional
jitters
assignment
parameters
and
be
able
to
use
them
to
do
some
computation,
which
is
not
xpf
in
this
case
not
advising
T,
and
this
problem
exists
in
many
implementations,
Anthony
Lewis.
So
next
slide,
please.
H
So
the
two
solution
to
that
one
is
to
use
t
ll,
say
another:
one
is
two
years
extended:
wrinkle
essays
and
copy
this
information
to
them
and
then,
through
the
separate,
so
here
we
are
assessing
they
use
it
both
next
slide,
please!
So
if
we
use
t
allah
says
the
problem,
even
when
t
is
not
enabled
and
in
most
implementations,
explicit
configuration,
you
need
to
configure
it
to
work.
H
If
we
start
using
TLS
say
we
explicitly
make
link
part
of
key
topology
well,
it
might
not
be
the
intention
number
two
there's
fruit
in
no
way
for
him
for
ospf
to
recognize
whether
advertisement
has
to
be
made
part
of
d
topology
or
a
GP
next
slide,
please.
So
the
option
two
is
to
use
defendant
in
LSA
and
caught
me.
Kate,
by
case
information
from
are
not
glued
this
information
in
aunty
permit.
H
So
this
information
could
be
flows
across
the
network
and
consumed
by
the
routers
without
making
them
part
of
the
apology
next
slide,
please
so
the
advantages
are
and
you
don't
need
to
create
te
topology.
If
you
don't
need
to,
we
don't
touch,
t
allah
says
so
there
are
backwards,
compatible,
3630
and
clear
distinguish.
If
you
look
from
operational
perspective,
it's
always
good
to
have
clean
distinguish
between
t
and
AGP
databases
ability
to
look
into
both
without
mixing.
H
H
H
K
C
A
J
Chris
bowers
with
juniper,
so
I've
made
some
pretty
detailed
comments
on
the
list
about
this
that
the
two
main
things
are
in
terms
of
the
draft.
Actually
I
I
still
don't
get
a
clearer
understanding
of
exactly
what
a
te
application
is
or
not
I've
suggested
is
that
does
that
mean
RSVP
and
I
got
a
very
vague
response
back
so
trying
I
think
it
would
be
very
useful
to
actually
try
to
define
in
the
document.
J
What
do
you
mean
by
a
te
enabled
application,
and
that
may
lead
to
some
understanding
about
how
difficult
the
interpretation
of
these
two
will
be?
The
second
is
I've
talked
about
backwards,
compatibility
issues
and
the
general
response
seems
to
be
like
well.
If
you
did
this
in
the
past,
you
were
wrong,
and
so
there
are
no
backwards
compatibility
issues,
and
you
know
if
you,
if
you
were
to
rewind,
say
two
years
ago
before
this,
these
extensions
existed.
J
I
think
you
would
have
seen
no
problem
at
all
with
using
the
SR
LG
information
for
remote,
LFA,
so
I,
I
think
the
backwards
compatibility
compatibility
issues.
I
should
be
taken
very
seriously
I.
It
may
be
a
problem
to
solve,
but
simply
saying
that
backwards
compatibility
is
not
an
issue
is
not
going
to
be
productive.
H
B
Orange
oh
I
kind
of
agree
with
the
previous
command
hospice
are
I
know
when
we
need
a
third
informations
in
our
I
GP.
We
enable
T
if
it's
the
way
it's
implemented
in
our
deployed
implementations,
some
I
don't
see
the
need
to
tweak
the
original
specification
to
do
it.
Otherwise,
if
t
means
that
the
thing
you
need
to
enable
to
have
a
surge
isn't
even
able
T,
and
maybe
it's
because
it's
here
probe,
maybe
just
a
guess
right
on
AC
comments,
because
it's
tears
more
enabling
t4
ospf
people
than
it
is
for
I
size.
H
Again,
going
back
to
back,
our
sensibility
is
not
going
to
break
your
network.
You
can
still
use
since
these
enable
you
can
still
use
that
in
order
to
look
into
a
surges,
our
for
people
who
are
looking
to
do
so,
there's
going
to
be
a
way
to
do
it
cleanly
without
enabling
TV.
She
has
nothing
to
do
with
little
stormy.
I
Hi,
oh
yeah,
the
snow
has
I'm
curious.
How
many
people
see
this
as
a
problem?
I
know
that
when
we
did
LFA
it
talks
all
about
srl
geez,
it
doesn't
have
a
reference
to
the
extensions,
but
clearly
that
was
the
only
place
you
could
get
us
or
LG
stuff.
Now,
I
information
I,
don't
have
any
strong
reason
to
believe
that
very
many
people
have
implemented
the
SLG
aspects
but
I
think
I.
I
Certainly
at
that
time
there
I
others
anyone
who
was
talking
about
it
assumed
that
you
would
just
get
the
SLG
information
from
the
teapot
apology.
I
I
do
share
the
concerns
about
migrations
or
changes
or
I
have
a
very
hard
time
hearing.
You
come
in
and
say
well,
this
is
a
protocol
problem,
so
we
can
ignore
the
operate,
but
we'll
just
fix
it
and
there's
not
an
operational
aspect.
If
it
were
to
happen,
I'd
like
to
have
our
focus
be
the
other
direction.
You
know
solving
the
operational
this.
H
I
H
Be
great
to
hear
from
operators,
people
we
are
doing
it
for
whether
you
think
it's
usable
solves
problem
and
again
since
I
p
past
your
other,
not
just
like
a
lot
a
remote,
they
become
more
and
more
popular.
That
might
be
more
and
more
people
willing
to
make
use
of
it
and
they'll
have
a
village
to
do
is
cleanly
without
never
in
traffic
engineering,
energy,
peas,
us
especially.
A
B
H
So
I
think
drop
would
greatly
benefit
from
exact
definition
of
T.
Application
will
work
on,
but
they
live
in
a
side
definition.
It
gives
you
ability
to
distribute
information
which
previously
could
be
only
distributed
through
T
extensions.
Now
you
get
ability,
distribute
information
without
having
configuring.
In
fact,
engineering.
C
Me
just
say
so,
and
what
I'm
I'm
a
co-author
as
well
I
didn't
write
it,
but
I
will
write
participate
in
this
discussion
of
the
backward
compatibility.
I
think
the
thing
is
that,
with
all
the
stuff
being
putting
into
prefix
length,
attributes
you're
going
to
have
those
anyway,
so
this
would
allow
you
to
do
it
in
two
LS
a's,
since
we
already
got
one
disjoint
piece
of
information
rather
than
having
it
in
three
places
now
so
that
was.
J
Chris
bowers
juniper
not
sure
if
I
exactly
understood
that,
but
I
mean
one
way
forward
is
if
the
motivation
is,
we
are
adding
lots
of
new
stuff.
You
draw
a
line
in
the
sand
and
say:
okay,
anything
moving
forward
goes
here
with
this
new
definition
but
stuff
that
has
been
defined
for
many
years
and
has
been
assumed
to
be
in
that
place.
Leave
it
where
it
is
no.
H
C
C
I
mean
I
mean
it.
There
is
a
there
is
a
difference,
I
think
between
te
and
90,
because
typically
the
TE
stuff
goes
to
a
different,
can
go
to
a
different
entity
to
compute
the
path
or,
as
the
other
stuff
doesn't
necessarily
I.
J
C
H
J
H
So
again,
I
think
we
presented
two
options.
One
is
to
continue
doing
what
we've
been
doing
for
last
15
years,
another
one
to
make
it
better.
The
help
of
this
bias
to
second
opinion,
because
you
propose
to
use
it.
However,
again
we
are
asking
you
working
group
to
your
opinion
and
what
witches,
but
should
we
do?
Thank.
C
This
prefix
link
tags.
It
seems
like
an
obvious
thing
to
do
and
I
want
and-
and
we
have
had
this
draft
for
a
while,
but
the
agendas
have
been
sort
of
full
at
the
last
couple.
I
etfs,
so
I
didn't
I
didn't
have
any
real
burning
need
to
present
it,
but
this
one
it
was
I,
said:
okay.
Now
I'm
going
to
present
this
one,
and
I
did
a
clean
up
on
the
draft
and
and
put
together
the
presentation
so
right
now
you
know
this
this.
C
What
this
does
it
just
allows
you
to
advertise
one
or
more
Ministry
of
tags
and
associated
with
a
prefix
or
length
either
in
ospfv3
or
v2.
It
makes
use
of,
like
I
said
in
the
ekstrom,
are
going.
It
makes
use
of
the
extended
the
TLB
extensions
in
the
two
protocols.
A
lot
of
it
I
looked
very
closely
at
RFC
5150
diocese
raft,
so
hopefully
even
the
ious
guys
guys
would
like
this
one.
C
Ho
caught
on
it,
but
the
one
thing
I
didn't
do
is
I
is
I
had
the
64-bit
tags
in
as
well,
but
I
looked
and
nobody
had
implemented
a
matter
of
fact
when
we
say
multiple
tanks,
the
multiple
text
people
had
implemented
were
a
very
small
number
like
two
or
three.
It
was
nobody.
You
know
nobody
dimp
lamented
like
20
tags
or
anything
like
that.
So
I
just
moved
this
to
an
appendix
I.
Think
the
original.
C
C
Which
we
can
hopefully
know
these
are
too
controversial,
I'm
going
to
talk
to
whether
or
not
I
use
case
falls
into
the
use
of
these
tags.
This
is
one
of
the
things
that,
for
the
node
tag
draft,
the
use
cases
were
actually
as
controversial
in
the
final
discussions
as
the
extension
itself,
but
I
think
most
of
these
are
pretty
pretty
much
pretty
straightforward.
We
could
do
it,
use
all
these,
for
instance,
for
controlling
redistribution
right
now.
C
We
can
only
do
it
for
external
routes,
because
that's
the
only
place
you
can
put
a
tag
or
selecting
prefixes
either
for
protection
or
for
prioritization
and,
like
I,
said,
there's
always
a
fine
line.
If
it's
something
it
may
be
something
now
that
we
have
till
these,
we,
it
may
be
something
you
don't
want
to
use
a
tag
for
if
it's
something
really
that
needs
to
be
standardized,
you
could
have
a
specific
tlv.
Just
for
that
purpose,.
C
For
the
ospf
be
to
encoding,
we
love
it
both
for
the
links
and
prefixes.
Just
you
in
the
LPL
essays
and
it's
simply
a
list
of
one
or
more
32-bit
event.
Miss
admin
tags,
ospfv3
I
took
advantage
of
the
extended
l
essays
and
you
can
see
for
all
the
ones
that
where
you
have
prefixes,
you
can
advertise
a
prefix
LSA
and
all
the
ones
that
are
and
the
and
the
router
the
only
place
you
have
a
link
again
is
the
router
LSA
in
the
router
linked
he'll,
be
in
extended
essays,
it's
the
same
tags.
C
Okay,
this
is
the
key.
He
won.
The
abrs
and
NS
a
translator's
must
propagate
tags
across
area
boundaries.
So,
if
you're
doing,
if
you're
generating
summaries,
you
should
propagate
the
tags,
the
number
of
tags
depends
on
how
many
tech
support
I'm
not
going
to
you
know
you
don't
have
to
support
more
tags
than
the
router
support.
So
if
it
only
supports
three
tags-
and
you
end
up
with
five-
you
only
have
to
propagate
the
number
you
support.
C
Again,
ospfv3
must
support
one
prefix
or
link
tag
and
for
ecmp
prefixes
you
can
propagate
you
see
you
should
propagate
the
tags
from
the
contributing
paths
if
you
have
different
prefixes,
if
you
support
multiple
tags
now
these
when
we
discuss
this,
these
are
all
subject
to
negotiation
or
working
group
consensus
and
I
thought
I
had
one
more
slide,
oh
well,
the
other
thing
that
I
was
going
to
say:
I
must
not
up.
I
must
not
have
uploaded
I
thought
I
did
recently
the
one
other
operational
consideration
is
for
the
NSA
and
external
LSAs.
C
The
first
tag:
you
must
use
the
existing
the
existing
mechanism,
so
you
know
we
have
a
tag
field
in
NS,
al
essays
and
external.
So
for
your
first
tag,
you
must
use
that
just
for
backward
compatibility
anyway,
next
step
more
two
more
discussion
on
the
list
of
this
area.
Has
anybody
read
this
Jeff?
Okay,
okay,
so
one
Peter
or
myself
will
I,
don't
know
will
probably
be
discussing
the
other
one
so
much.
This
may
may
take
a
while
before
we
get
to
this
one.