►
From YouTube: IETF112-LSR-20211111-1200
Description
LSR meeting session at IETF112
2021/11/11 1200
https://datatracker.ietf.org/meeting/112/proceedings/
A
I
have
a
m1
mac
mini
super
happy
with
it
it's.
It
goes
faster
than
my
octa-core
intel
laptop,
but
I'm
shocked
at
how
fast
it
goes.
Yeah
anyway,
I
upgraded.
B
C
Column
in
there
there
was
a
column
in
the
new
york
times
yesterday
about
how
how
great
that
processor
is
like
really.
The
new
york
times
is
like
having
fanboy
articles
about
cpus,
okay,.
A
B
B
Ying's
ends
online
jen's
online
too
she's
been
for
a
while.
So
we
can
just
start.
A
Okay,
yeah,
okay,
yep.
B
Okay,
good
morning,
everyone,
buenos
dias
for
those
of
you
who
are
pining
for
madrid
and
happy
veterans
day.
B
This
is
the
note,
well,
I
know
you've
seen
it
a
lot
in
the
other
ones.
The
crux
of
it
is:
you
need
to
comply
with
all
these
ietf
rules,
including
declaration
of
any
ipr.
You
know
about
related
to
work
in
the
ietf,
and
your
attendance
of
the
meeting
constitutes
participation.
B
The
other
one
people
have
asked
to
call
out
is
the
kona
code
of
conduct.
I
don't
think
we've
had
too
much
problems
with
that
out
of
the
out
of
our
working
group
we
haven't
had,
but
that's
also,
where
we're
go,
we're
obliged
to
treat
everybody
with
respect.
B
B
B
Oh,
I
would
hope
in
the
next
couple
months,
even
if
we
wait
for
the
bfd
yang
biz,
because
it's
it's
a
very
short,
very
small
change
to
the.
A
I
don't
know
ac,
I'm
not
sure
if
we
have
anybody,
that's
vision
impaired,
but
that's
what
you
were
just
talking
about
are
young
data
models
for
ospf
and
the
young
data
model
for
isis.
A
B
Models
the
base,
the
base
models,
the
100
plus
page
documents
on
which
all
the
other
gaming
documents
are,
are
augment.
B
I'm
not
going
to
say
anything
about
the
only
one
I'm
going
to
say
something
about
is
the
flex
algorithm.
We
had
some
changes
to
that.
We
had
an
errata
for
the
extended
attributes,
isis
and
ospf
drafts,
that's
89,
20
and
89
19,
those
rfcs
and
now
that
it's
finally
through
the
second
working
group
last
call
it's
waiting
on
john
to
review
chris.
The
reverse
metric.
It's
in,
I
is
it
still
in
ietf
last
call
that
kind
of
completed
without
any.
A
C
I
sent
you
review
comments,
you
you
took
care
of
them,
then
I
went
to
ietf
last
call.
C
The
ietf
last
call
is
probably
finished
by
now,
but
I
have
that
on
my
list
of
things
to
check
up
on,
but
it's
presumably
if
the
last
call
hadn't
finished
already
it'll
be
finished
by
the
end
of
this
week,
and
we
can
kick
it
on
to
the
next
step.
Great.
B
We
got
these
two
sr
yang
models,
ospf
sr
and
isis.
These
are
augmentations
to
the
base.
I
don't
put,
I'm
not.
I
wouldn't
put
too
high
a
priority
on
these
until
we
get
the
base
done,
I
mean
we
can
like.
We
can
start
definitely
start
the
process,
though
the
distinction
of
these
drafts,
the
last
drafts
that
don't
have
lsr
last
working
group
graphs
that
don't
have
lsr
in
the
title
from
when
we
merged
the
working
groups
about
five
years
ago,
I'm
not
going
to
say
anything.
A
A
It
was
five
years
ago,
that's
wow,
something
like
anyway
yeah.
B
We're
gonna
cover
flood
reduction.
Today
I
won't
say
anything:
bfd,
strict
mode,
that's
been
implemented,
it's
pretty
simple
draft.
I
think
we
could.
We
could
go
through
this
one
actually
pretty
quickly,
and
this
is
one
we
hope
to
get
done,
because
we
have
the
isi
sitting
waiting
waiting
for
miss
refs
refs,
so
we'd
really
like
to
get
the
ospf
v3
extensions
for
srv6
done.
B
B
We
have
a
only
one,
I'm
going
to
say
anything
about
is
the
flexible
bandwidth,
delay
and
constraints.
There
was
a.
There
was
a
discussion
on
the
asla
that
never
completed,
and
today
I
mean
that
the
generic
metric.
B
We
didn't
really
get
through
that
discussion
on
the
list,
but
there's
a
now
there's
a
new
proposal
for
asla
bit.
Hopefully
we
can
discuss
whether
that
can
resolve
the
controversy
with
the
generic
metric.
B
B
E
F
A
Yeah,
so
let's
flip
back,
can
you
flip
back
to
the
existing
working
group
documents?
Certainly
page
one?
I
guess
one
of
three
yeah.
So
I'm
curious,
you
know
we
have
all
these
experimental
ones
do
any
of
the
authors
of
any
of
these
have
to
can
they
report
like?
Has
it
been
deployed?
A
Is
anyone
using
them?
Are
they
being
experimented
with?
I
guess
the
question
would
go
out
to
wemo
tony.
What
was
the
other
one?
I
guess
the
other
one
was
flood
reflector,
which
presented.
B
A
No
okay!
Well
we're
not
gonna,
hear
anything
about
those,
then
okay
yeah.
So
I
guess
then
the
I
wonder
if
there's
anything
else,
we
could
say
no
to
see
if
we
could
push
some
of
these.
Through
I
mean,
are
they
are
any
of
these
blocked
on
us?
A
Does
anybody
know
if
any
of
these
are
blocked
on
us?
These
working
group
documents,
if
you
want
to
say,
hey,
look
pay
attention
to
mine.
You
guys
need
to
do
this.
B
Let
me
look
at.
Let
me
look
at
anybody
else.
Ask
for
the
reverse
metric
they've
asked,
for
I
mean
they've,
said
it's
ready
for
working
group.
Last
call
you
know
it's
just
the
oas,
ospf
corollary
to
the
existing
ice.
D
B
A
A
B
Not
none
of
the
other
ones
that
really
they
really
asked
for
working
for
a
glass
call,
and
there
hasn't
been
much
discussion
on
on
on
the
list
on
as
we
get
further
on.
B
A
A
B
A
I'll
show
you
what
it
looks
like
just
in
case.
Let's
see
so
it's
right
here,
this
guy
is
that
the
one
you
clicked
on?
Yes,
oh
man,
okay!
Well,
oh
it
says
all
the
slots
immediate,
oh
okay,
so
maybe
it
pops
something
up
on
your
screen,
but
you
can't
see
it.
A
D
A
Yeah,
my
audio
I'm
getting
is
fuzzy.
Is
anybody
else
getting
that
I
mean
I
might
re
reset
my
audio.
G
That
go
ahead,
les
all
right,
okay,
so
this
is
an
update
on
fast
flooding.
G
G
Arrow
keys,
oh
I
see
I
see,
I
see
I
have
the
numbers
okay,
so
in
september
the
authors
of
these
two
drafts,
which
we've
been
talking
about
for
some
time
now,
agreed
to
produce
a
combined
draft.
That
was,
I
believe,
just
before
the
interim
meeting
that
we
had
on
this
topic.
G
G
There
is
also
a
flow
control
aspect
depending
upon
what
approach
you
take
and
for
flow
control.
We
do
need
some
tlvs
for
that
have
to
be
supported
for
interoperability,
and
we've
been
debating
back
and
forth
for
some
time
about
which
approach
is
better
and
we've
basically
decided.
We
don't
have
to
make
that
choice
at
this
point.
What
we
want
to
do
is
to
support
experimental
work
going
forward
in
terms
of
how
the
algorithms
are
working
and
that's
what
we've
been
talking
about.
G
So
this
is
not
a
technical
presentation.
This
is
basically
just
a
a
quick
overview
of
the
combined
draft
content,
and
I've
done
this
basically
by
picking
the
key
sections
in
the
document
and
describing
the
content.
That's
in
these
sections,
so
we
have
a
section
an
introductory
section
and
the
discussion
of
historical
behavior.
G
G
The
first
two
already
were
present
in
draft
ukraine,
although
we've
renamed
them
a
bit
and
we
introduced
four
new
ones
which
were
related
to
things
that
the
you
know,
one
or
or
the
other
algorithm
was
making
use
of.
G
Section
6
contains
the
description
of
the
two
algorithms
that
have
been
discussed
separately
in
the
in
the
different
documents
in
the
past.
So
section
6.2
is
the
algorithm
from
draft
ukraine.
The
description
has
been
updated.
6.3
is
the
algorithm
from
draft
ginsburg.
That
description
has
also
been
updated,
but
they're
they're.
Fundamentally,
you
know
the
the
same
two
algorithms
that
were
previously
in
separate
documents
and
we
added
an
appendix
to
track
changes
and
another
appendix
to
track
open
issues.
At
the
present
time
there
are
no
open
issues.
G
We
managed
to
resolve
all
of
the
open
issues
that
we
had
during
the
discussion
of
putting
together
the
combined
draft,
but
we'll
use
that
to
track
issues
in
the
future
and
now
we'd
like
to
request
working
group
adoption.
This
has
been
discussed
probably
for
a
couple
years
now
and
we'd
like
to
to
move
forward
with
with
the
work,
and
you
know
we
will.
You
know
the
respective
teams
are
still
doing
implementation
and
testing
and
there'll
be
further
reports
and
update
to
the
draft
in
the
future.
G
B
B
Okay,
okay
yeah,
I
think
we're
gonna
we're
definitely
gonna
do
given
the
interest
and
the
you
know
the
the
number
of
meetings
we've
had
on
this.
This
has
been
probably
the
most
discussed
topic
on
working
group
list
for
a
long
time.
We're
going
to
do
working
group
adoption
immediately.
The
one
question
is
whether
we
do
standard
tracker
experimental.
I
think
we'll
we'll
we'll
we'll
include
that
in
adoption,
we're
kind
of
running
short
in
time.
We
didn't.
B
B
But
anyway,
good
work,
I
mean
to
everybody
on
the
design
on
on
both
of
the
drafts
and
the
discussion,
and
I
I
I
read:
it
looks
ready
for
adoption.
I
I'll
probably
have
some
more
comments,
but
purely
editorial.
B
J
I
Yeah
look
at
this.
Something
works.
Yes,
so
pdf,
you
don't
really
see
the
title
doesn't
matter
anyway.
It's
the
update
on
what's
going
on
with
the
flat
reflection
draft
version,
four.
I
Okay,
so
yeah
four
came
out
that
was
based
on
some
discussions
and
implementation
deployment
experience
things
especially
to
ac,
who
really
find
competitive
stuff.
I
don't
think
how
much
was
on
the
list.
How
much
was
actually
authors
they
see.
I
So
you
know
multiple
sections
have
been
changed,
clarified
added,
so
substantially
nothing
changed,
but
a
lot
of
clarification
has
been
put
into
the
document,
more
detailed
changes,
so
glossary
always
a
nice
thing
and
then
we
broke
out
and
clarified
the
section
about
the
two
deployment
modes
which
you
know
we
call
no
tunnel
deployment,
so
basically
no
shortcuts
and
l1
tunnel
deployment.
So
basically
you
shortcut
the
thing
with
full
or
partial
mesh
of
l1
shortcuts
and
you
use
basically
the
tunnels
as
next
tops
what
we
also
throw
in
based
on
already.
I
You
know,
discussion
about
some
future
stuff
and
implementation
experience.
We
threw
in
the
tunnel
type
discovery
because
so
far
there
was
only
discovery
of
whether
you
know
where
the
end
points
are
and
who
is
doing
what
role.
But
there
was
no
description.
What
kind
of
tunnel
can
you
set
up
if
you
want
to
set
up
automatic
tunnel?
And
thankfully,
now
we
have
the
rfc
9012,
which
deals
in
extent
with
extensive
inexpensive
detail
with
the
subject,
so
the
acceptable
tunnel
end
points,
including
the
caps
and
all
the
other.
I
You
know
just
that
you
need
to
get
it
done,
are
being
announced
and
a
flag
basically
shows
up
where
the
the
end
point
is
flood,
reflector,
adjacency
or
an
l1
shortcut
end
point,
and
of
course
you
know
you
can
run
the
stuff
just
fine
with
statically
configured
tunnels
or
any
you
know,
tunnel
solution
that
that
you
prefer
to
you
know
this
auto
discovered
stuff,
all
right,
even
more
detail.
That's
mostly
implementation
and
deployment
feedback.
I
The
master
of
multiplicities
of
tlvs
has
been
moved
mostly
to
shoot,
and
the
violation
should
be
locked
and
clarified,
and
only
the
first,
you
know,
instance
should
be
used.
I
think
it
was
based,
maybe
even
on
aces
comments.
I
The
other
interesting
thing
is
that
we
added
the
sub
tlbs
for
discovery
of
the
different
roles
and
the
tunneling
points.
We
moved
that
not
only
in
l1
but
also
into
l2
as
adva
advertisement
as
the
scope
of
advertisement.
I
That's
simply
based
on
the
observation
that
some
people
desire
to
run
it
with
underneath
not
having
isis
l1,
but
possibly
something
else.
So
the
advertisement
cannot
fly
in
l1.
They
should
fly
in
l2.
I
Then
isis
metric
on
all
flood
reflector
adjacency
should
be
uniform,
that's
basically
from
deployment
they
don't
have
to
be,
but
it
will
save
you
potential
a
lot
of
trouble.
If
you
keep
you
know
the
flood
reflector
adjacency
is
uniform
in
terms
of
the
metric
distances,
then
again
deployment
and
implementation.
I
It
wasn't
entirely
clear
that,
within
an
area
isis
area,
the
cluster
id
should
be
uniform.
That
does
not
prevent
you
technically
to
run
multiple
cluster
in
an
area,
but
it
still
prevents.
I
mean
it
was
pretty
clear
in
the
draft,
but
on
the
implementation
experience
you
know,
we
pointed
that
stuff.
We
made
it
even
more
clear
that
a
single
client
has
to
use
the
same
cluster
id.
I
So
the
only
way
he
can
interact
with
another
cluster
would
be
for
a
normal
l2
adjacency
with
another
client
in
another
cluster,
and
we
clarified
also
that
the
cluster
id
is
used
to
actually
check
adjacency
formation.
If
you
decide
to
be,
you
know
a
client
or
a
reflector
right,
so
old
ants.
I
What
we
see
is
that
the
numbers
in
terms
of
additional
scalability
you
can
get
with
it
are
pretty
much
what
we
expected
based
on
kind
of
simulation,
theoretical
work
thinking
through
the
stuff
before
implementation,
and
we
also
see
that
it
does
not
require
forklift
of
the
network,
but
you
can
actually
move
over
without
losing
capacity.
I
So
basically
one
node
at
a
time,
if
you
do
it
correctly,
technically
speaking
or
in
terms
of
substance,
there
were
no
significant
changes
right,
except
all
this
clarification
work
that
I
pointed
out,
and
I
think
again
that
this
is
ready
for
last
call.
I
do
not
think
we'll
have
any
substantial
additions
or
changes
and
currently
ongoing
work,
which
is
already
spinning
out
of
it
is
the
bgpls
which
is,
of
course,
the
next
thing
that
is
needed
that
has
been
introduced
in
idr.
I
This
meeting
jordan
is
doing
most
of
this
work
there
and
we're
also
starting
on
auto
flood
reflection,
which
is
the
same
thing,
is
auto
evpn
in
rift,
which
basically
means
that
you
can
bootstrap
routers
without
any
configuration
whatsoever,
and
you
end
up
if
you
build
something,
also
variation
of
the
claw
network,
you
end
up
with
a
fully
configured
claw,
plus
the
flood
reflection
already
automatically
being
configured
on
top
of
it.
That's
the
last
one
that
I
have.
I
I
think
yes,
so
from
my
side
or
from
the
other
side,
the
outstanding
request
is
to
go
with
this
thing
for
a
last
call
and
see
whether
more
things
pop
up.
If
anybody
has
implementation,
wants
to
interrupt,
and
so
one
and
so
on
as
usual,
pretty
welcome
implementation
is
relatively
simple
from
our
experience,
so
no
highly
encouraged
thanks.
I'm
done.
A
So,
with
a
contributor
hat
on
at
the
chair
hat,
I
reread
the
draft
yesterday
and
had
a
couple
of
thoughts.
Have
you
considered?
You
know
one
of
the
obvious
ways
to
deploy.
It
is
with
the
with
the
tunnels
between
the
l1
l2s
for
forwarding.
A
Have
you
considered
like
just
they
automatically
working?
You
know
like
they're,
auto
setup,
not
even
signaled
at
all,
but
you
know
just
because
of
the
fact
that
they're
once
they
join
up
with
the
reflector,
then
they
just
you
know
they
can
program
themselves
to
accept
encapsulated
traffic
from
the
other
clients.
A
I
I
A
I
A
So
that's
what
I
was
getting
at
is
that
that
up
front
rather
than
signaling
tunnels,
just
saying
what
type
of
tunnel
and
then
everyone
just
does
it
anyway.
It's
just
that
this
was
a
thought
about
it.
Well,.
I
So
see
the
tunnel
saying
you
know:
if
you
talk
to
four
customers
what
tunnel
they
prefer,
they
got
fi.
You
get
five
to
six
answers
right.
So
that's
that
that
was
exactly
the
discussion
we
were
having
and
we
tried
to
steer
them
towards.
You
know
gre,
and
you
know
every
tunnel
has
it
worked
and
silicon
support
problems
and,
frankly,
the
most
advanced
customers.
They
they
end
up,
preferring
the
no
tunnel
option
and
that's
why
we
worked
on
that
stuff
most
extensively.
I
A
So
I
have
just
a
follow-up
question
on
the
tunnels.
You
know
you
in
the
draft.
I
think
you
might
have
added
the
ttz
comparison.
It
did
get
me
thinking.
You
know
the
ttc
gives
you
a
full
mesh
of
adjacencies
and
and
you're
cutting
down.
You
know
that's
the
main
advantage
here
is
the
adjacency.
The
reflector
keeps
your
adjacency
hub
and
smooth
yeah.
D
A
I
Hidden,
that's
that's
the
difference
right.
This
draft
basically
hides
the
full
nash
by
no
showing
that
whatever
a
double
star
and
that's
architecturally-
really,
you
know
the
only
difference
in
the
rest
is
kind
of.
You
know
mechanics
and
little
smart.
You
know
curriculus
and
so
on.
What
is
of
course,
drastically
different.
Is
the
no
tunnel
option
right,
but
this
is
far
more
sophisticated
to
deploy.
I
mean
this
is
a
much
sharper
knife
I'll
bite,
no.
I
How
should
I
put
it?
It's
it's
very
appealing
right
because
you
get
rid
of
all
the
time
stuff
and
you
get
very
good.
You
know
properties
in
terms
of
scalability,
but
you
have
to
know
what
you're
doing
when
you
start
to
flood
all
the
l2
into
l1
and
l1.
You
know
out
in
terms
of.
A
I
A
Yes,
let
me
wemos
he's
not
in
the
queue
but
he's
put
in
the
chat.
Ttz
does
not
use
tunnels.
A
K
Can
I
can
I
speak
it
for
400
seconds
yeah
for
ospf
tdz
for
the
tdz
ages,
fully
connection?
They
they
don't
use
tunnels,
they
use
the
shortest
path,
because
that's
a
forwarding
from
one
age
to
another
age,
to
this
age.
A
I
Yeah
yeah,
I
mean
I
remember
the
ttc
differently,
but
it's
a
while,
since
I
read
it,
you
know
when
there
were
initial
discussion
eight
years
ago,
when
I
was
in
new
mexico
about
you
know
why
the
whole
thing
would
loop
and
then
the
full
mesh
came
up
and
so
on
yeah.
But
you
know
this.
This
thing
was
specifically
built
to
scale
up
l2
right,
so
we
couldn't
yeah,
no,
not
all
the
stuff
into
l2,
and
then
the
l1
does
underneath
what
l1
needs
to
do?
I
Yes,
all
right,
so
I
hope
it
was
enlightening
to
a
point.
I
think
the
drop
is
now
very,
very
detailed
if
somebody
were
to
attempt
an
implementation.
B
Yeah
yeah
us
this
is
ac
speaking
as
a
working
group
chair,
I
I
we
went
through
some
iterations
and
we
handled
all
the
things
that
myself
and
other
people
had
said
needed
to
be
clarified,
and
I
think
it's
ready
too
I'd
like
some
of
the
longtime
isis
developers
like
stefano
strata,
less
ahmed
and
some
of
the
others.
I'd
like
would
be
good
if
you
would
review
it
as
well.
Oh
bruno
and
bruno.
B
No,
no
nokia,
I
think.
E
B
A
D
B
L
Okay,
so
I'll
be
presenting
the
application
specific
link
attributes
any
application
bit
on
behalf
of
my
co-authors.
L
So
we
look
at
what
problem
we
are
trying
to
solve.
With
this.
We
will
look
into
some
of
the
details
of
protocol
extensions
and
how
is
the
backward
compatibility
handled.
L
So
the
problem
statement
is
that
network
operators
may
want
certain
link
attributes
to
be
used
by
all
current
and
future
applications.
L
So-
and
we
have
many
examples
of
you-
know
such
attributes
and
how
the
network
evolved
over
many
years
so
and
we
we
have
made
asla
mandatory
for
some
applications
such
as
flexalgo,
so
asla
allows
for
an
attribute
advertisement
where
link
attributes
applicable
to
one
applications
or
some
applications
or
all
the
applications
currently
defined.
L
L
So
the
reason
why
we
call
it
limited
support
is
because
rfc
8919
and
8920
they
do
not
allow
application
to
use
attributes
from
zero
length.
So
there
is
a
zero
length
sabm.
L
If
you
advertise
attributes
under
that
zero
length,
sapm,
then
any
other,
then
any
application
can
use
it
unless
any
the
application
has
any
other
attribute
advertised
with
an
application
bits
it.
So,
during
a
discussion
in
the
mailing
list,
there
were
opinions
that
more
granular
control
over
attribute
advertisement
for
any
application
versus
application.
L
Specific
is
useful,
like
some
micro
attributes,
operators
want
to
design
it
in
an
application
independent
way,
so
that
those
attributes
are
always
available
for
any
application
that
are
currently
defined
or
in
defined
in
future,
and
some
specific
applications
may
be
designed
to
be
used
in
an
application
specific
way.
So
this
we
are
trying
to
solve
this
problem.
Where
you
know
this
can
be.
This
control
can
be
provided
in
asla.
L
So
the
details
of
protocol
extension,
so
the
the
so
any
any
bit
is
a
standard
bitmask
and
it's
a
bit
number
four.
So
that
implies
that
any
application,
so
any
attribute
advertised
under
this
can
be
used
by
any
application.
L
So
obviously,
a
node
that
does
not
understand
a
bit.
It
cannot
process
the
attributes
under
under
the
advertisement,
which
has
an
abe
asla
tlb,
which
has
an
a
bit
set.
So
it
ignores
that
and
processing
of
the
other
bits
that
remains,
as
per
eight,
nine
one,
nine
and
eight
nine
two
zero
and
this
draft
mandates
that,
if
same
attribute,
is
advertised
under
asla
with
a
bit
and.
L
There
is
an
advertisement
under
asla
with
specific
application
bit
set,
the
one
which
which
is
application
must
use
the
one
which
which
has
the
application
bit
set.
I
know
that
takes
the
precedence,
so
this
is
in
order
to
avoid
so,
let's
say
some
nodes
are
upgraded.
They
suddenly
start
advertising
attributes
with
a
bit
set
and
the
older
nodes.
They
will
pick
up
attributes
under
the
application
specific.
L
L
So
until
all
nodes
are
upgraded
to
support
a
bit.
Obviously,
the
nodes
won't
be
able
to
use
the
attributes
which
are
advertised
under
8-bit
and
hence
the
upgraded
nodes
also
would
have
to
advertise
in
in
the
attributes
under
a
bit
as
well
as
under
application.
Specific
bit
set
and
the
processing
of
the
scbm
with
zero
length
continues
to
follow
eight
nine
one,
nine
and
eight
nine
two
zero.
No,
no,
nothing
is
this
trap
doesn't
modify
any
of
those
procedures.
L
Yeah
request
further
review
and
comments
on
the
draft
and
one
and
also
request
working
group
adoption
and
one
more
thing.
I
want
to
clarify,
based
on
the
on
ac's
comment
in
the
beginning
of
the
meeting
today,
so
this
is
quite
independent
from
generic
metrics,
so
this
is
just
trying
to.
L
Provide
for
for
network
operators
who
want
to
design
certain
attributes
in
an
application,
independent
manner
and
advertise
them.
This
is
just
giving
that
that
of
possibility
so
that
those
attributes
can
be
advertised
in
an
application
independent
way
inside
asla
yeah.
That's
the
clarification,
any
questions,
comments.
B
Okay
yeah,
I
I
realized
that
this
didn't
get
us
off
the
hook
for
the
for
the
discussion
of
the
continued
discussion
of
the
generic
metric.
The
one
question
I
have,
if
you
have
the
a
bit
set
and
other
bits,
you're
you're,
going
to
use
the
whatever
is
advertised
in
that
attribute
in
in
either
case.
B
So,
what's
the
advantage
over
the
a
bit
over
the
zero
length
bit
mask,
I
mean
you
know
the
application,
I
forget
the
acronyms
for
them.
What's
the
advantage
that
I
couldn't,
I
couldn't
see
that.
L
Yeah,
so
the
difference
is
that
so,
let's
say,
if
you
the
you,
if
you
have
designed
one
attribute-
let's
say
admin
group
in
an
application
independent
way
so
in
in
your
network
right
so
red
color
means
one
gig
link,
and
that
means
this
red
color
is
any
application
can
use
this
red
color.
This
admin
group,
so
admin
group
is
an
attribute
which
is
designed
in
the
network
to
to
be
used
by
any
application.
L
So
now
you
can
advertise
it
under
a
bit
and
there
are
some
other
applications,
some
other
attributes
that
you
want
to
design
it
in
an
application
specific
way.
For
example,
you
want
there
is
some.
Let's
say
some
bandwidth
related
attribute
that
that
you
want
to
design
it
and
advertise
it
in
in
an
application
specific
way,
so
so
different
for
srt
different
for
flex
and
go
and
so
on.
So
you
can.
You
can
do
that.
L
I
mean
you
can
advertise
that
one
under
the
bandwidth
one
under
the
application
specific
bit
set
and
the
the
admin
group
under
a
bit
set.
L
So,
whereas
if
you
advertise
the
admin
group
under
sabm
with
all
sabm
with
length,
zero
and
then
for
let's
say
for
flex,
algo
and
srt,
you
have
advertised
bandwidth
now
you
cannot
use
the
admin
group
for.
A
L
A
So
one
attribute
uses
zero
length,
everything's
golden
you,
do
another
attribute
with
a
specific
and
then
all
of
a
sudden.
You
have
to
stop
listening
to
that
zero
length.
To
me,
that
seems
like
a
fault
in
the
original
design
like
why
don't
we
just
fix
that
like?
Why
do?
Why?
Is
it
even
that
way,
and
maybe
less
can
answer
that
question.
G
Yes,
yes,
okay,
so
we've
had
extensive
discussion
about
the
technical
issues
on
the
list.
You
know,
including
the
the
points
that
you
know
ac
and
chris
are
are
asking
about.
I
don't
want
to
go
into
that
here.
I
don't
think
we
have
the
time
to
go
back
and
forth.
G
G
The
authors
of
this
document
have
decided
that
a
certain
portion
of
you
know
the
way
to
support
all
applications
or
any
application
if
you
will
is
not
to
their
liking,
and
so
they've
proposed
an
alternate
form
of
advertising.
This
the
two
definitions,
the
definition,
the
zero
length
definition
that
exists
in
the
asla
rfcs
and
the
a
bit
that's
defined
here-
are
not
compatible.
G
This
now
puts
us
in
the
position
of
instead
of
having
a
standard
which
everybody
can
use
to
support
interoperability.
We
now
have
multiple
published
solutions.
You
know,
some
of
which
are
favored
by
one
set
of
people
and
some
of
which
are
favored
by
another
set
of
people.
This
introduces
a
lot
of
interoperability
issues.
It
requires
vendors
to
do
multiple
implementations
of
the
same
functionality,
not
because
one
of
the
solutions
is
deficient,
but
simply
because
some
folks
have
decided.
G
We
prefer
to
do
it
this
way,
and-
and
this
to
me
does
not
serve
the
industry.
D
A
A
D
A
A
So
that's
a
reason
not
to
go
with
it
in
your
opinion,
correct
right,
okay,
great
so
we're
out
of
time
on
this.
The
question
that
I
was
asking,
I
guess,
you're
saying
go
read
the
list
on
me
is
why
the
zero
length
can't
be
fixed.
It
sounds
a
little
broken,
but
I'll
we
can
do
that
on
the
list
since
we're
out
of
time
and.
L
Yeah
sure,
and
with
this
just
just
one
comment
with
respect
to
leicester's
observation,
so
I
I
think
it's
not
right
that
you
know
this.
The
two
solutions
are
not
interoperable.
How
they
can
interrupt
has
been
discussed
just
now
in
the
previous
slide,
I
mean
it's
not
clear.
We
can
discuss
again,
but
I
do
think
that
this
very
much
interrupts
with
whatever
has
been
deployed
in
89,
19
and
89..
You
know
it
just.
A
Gives
another
option:
let's
do
it
on
the
list
because
I
mean
showing
something
doesn't
interop
is
pretty
easy
right
I
mean,
and
we
can
all
technically
look
at
that
and
look
at
the
example
I
mean
you
usually
can
show
an
example
right.
So
if
less
is
right,
you'll
throw
an
example
out
and
we'll
go
yep
les
is
right
and
if
les
can't
come
up
with
an
example,
then
you're
right,
that's
easy
enough
right,
yeah
yeah,
the.
B
Yeah
we'll
take
its
list,
it
stuck
out
in
my
mind.
It
said
when
everybody
has
upgraded
and
there
was
no
nothing
in
the
protocol
to
do
that.
It
seemed
like
it
was
an
operational
interoperability
and
not
a
protocol
interoperability,
but
anyway.
A
Okay,
who's
up
next,
all
right.
A
A
A
Can
you
talk.
A
A
M
M
So
I
think
we're
all
familiar
with
the
scenario,
so
I
just
described
the
solution
directly.
The
pm
solution
there
are
the
the
main
the
main
step
is,
is
open,
receiving
the
node
link,
failure,
information,
which
is
perfectly
either
in
the
range
of
the
androidized
summary
of
this,
the
abr
or
element
or.
M
M
M
So
then
we
are
only
some
of
the
avr
can't
reach
the
failure
prefix,
the
abr
that
can
reach
the
prefix
will
under
advertise
the
specific
load
to
this
qm
prefix.
So
this
procedure
is
the
same
as
the
river.
M
And
avr
will
only
send
out
the
such
message
about
all
the
internal
routers
for
the
search
mechanism
and
avr
can
also
control
want
to
send
out
such
message
based
on
sl-like
configuration
to
say
it
is
one
some
link
or
some
link
failure,
but
it
is
not
within
the
interesting
prefix.
So
the
apr
should
not
send
out
some
message,
so
it
keeps
people
the
the
the
original
performance
of
the
idp.
M
M
M
We
have
discussed
this
solution
online
or
offline,
and
here
just
our
conclusion
for
this:
even
the
notification
solution
proper
proper
one
general,
even
a
notification
container
varying
eyes
to
deliver
the
positive
or
negative
pulse
event
based
on
the
flooding
scope,
lsp
also
1750.
M
Currently,
it
is
on
focus
only
on
the
same
use
case
as
a
pm
chapter
and
based
on
the
discussion
we
think
for
4
pm
and
for
the
current
use
case.
Both
solution
can
achieve
the
same
result
and
about
40
am
the
existing
led,
pdf
format
and
procedure
can
be
utilized
and
the
solution
is
the
same
same
procedure
for
ospf
and
iss,
and
we
think
it
is
easy
to
implement,
deploy
and
debug
in
the
protagon
network
and
there
we
also
can
consider
some
misbehavior
for
the
unspotted
node.
M
By
the
way
we
have
simply
the
such
behavior
can
be
controlled.
You
know
kind
of
considering
in
in
department
and
for
even
the
notification
solution.
We
think
it
is
almost
the
same
procedure
as
qm.
The
difference
lies
mainly
how
to
encode
such
a
such
message
and
for
we
think
for
general
solution
such
general,
some
even
notifications.
M
It
should
be
compared
with
other
existing
solutions
for
the
positional
requirement,
because
it
is
not
able
to
solve
only
the
principling
reachable
announcement
and
I
may
think,
because
the
currently
there
is
only
one
use
case
for
such
a
certain
mechanism.
We
think
it
is
a
mature
to
introduce
regional
parts
mechanical
within
idp
and
currently
and
for
spf.
M
M
M
G
A
Yeah
so
chair
hat
on
they've
been
asking
for
adoption
for
a
while.
The
event
thing
is
new.
I
agree
with
what
you
just
said
less
in
a
perfect
world.
That
would
be
the
case,
but
then
again
asking
for
adoption
is
one
way
to
answer
that
question.
A
It
may
be
not
the
perfect
way
to
answer
that
question,
but
it
is
one
way-
and
I
mean
I
I
agree
without
my
chair
hat
on-
I'm
not
sure
we
need
this
right,
but
it's
not
for
me
to
say
by
fiat.
A
We
ac
did
put
something
out
on
the
list
to
try
to
engage
people
again,
and
I
don't
think
a
lot
got
said.
Yeah.
E
A
No,
that
that
was
all
he's
gonna
say
it
is
eijon
has
another
presentation.
He
only
has
got
three
minutes
left.
I.
A
B
I
didn't
I,
I
didn't
see
much
support
other
than
the
offers
and
I
saw
one
non-offer
supporting
the
other,
the
event
notification
so.
A
You
know,
I
think
it's
fair,
you
know
everyone
has
a
right
to
ask
for
an
adoption
and
right.
You
know
I
I
don't
know
so.
If,
and
everyone
has
a
right
to
say
we
shouldn't
adopt
this,
and
these
are
the
reasons
we've
sort
of,
let
it
been
letting
it
percolate
to
try
to
get
people
to
express
opinions
and
without
seeing
a
lot
of
negative
opinions,
it's
it's
hard
not
to
just
grant
the
adoption
call.
I
Yeah,
I
think
so
I
think
this
is
all
trash
can
in
the
igp
I
mean
nothing.
New
has
been
done
forever
because,
hey,
that's,
not
everybody
can
get
everywhere.
One
possible
solution
is
to
ban
or
encourage,
maybe
everyone
with
this
kind
of
suggestions
to
go
towards
the
service,
instant
stuff
or
whatever.
We
call
it,
which
I
think
is
a
good
idea
just
run
apart
on
igp
at
much
lower
priority
and
don't
trash
the
main
thing
that
holds
the
whole
thing
together.
A
That
sounds
like
a
great
comment
for
the
adoption
call
or
or
for
have
coming
out
before,
but
yeah
yeah
and
as
a
as
a
working
good
member.
I
I
agree
with
you.
C
Maybe
the
local
mute
button
in
the
lower
right
is
clicked.
A
I
don't
know
yeah
well,
why
don't
we?
Why
don't
we
move
on
to
the
next
presentation
and
if
we
have
time
at
the
end,
we'll
come
back
aigen
you
you
you've
got
to
figure
out
how
to
get
yourself.
Unmuted
you've
got
it.
A
Okay
now
you're
there,
but
let's
let's
come
back
to
your
presentation:
if
I'll
try
to
bring
people
in
tighter
and
see
if
we
can
come
back
to
it:
okay,
okay,
no
problem
all
right,
thanks,
sorry,
okay!
So
next
up
is
linda.
H
What
do
you
mean
you
want
me
to
present
or
what.
H
Arrow
right:
okay,
okay,
good
good!
Thank
you.
So
this
is
about
lsr
isis,
ospf
extension
for
serving
the
5g
edge
computing
services.
H
H
Okay.
So
a
little
bit
background.
So
this
is
for
the
edge
computing
environment,
where
the
servers
are
very
close
to
the
proximity
of
the
5g
size
like
upf
and
the
psa
package
session
anchors
and
for
the
mission
sensitive
applications.
It
is
very
common
to
have
multiple
servers
for
the
same
application
all
close
together.
H
There
are
many
use
cases
being
discussed
in
3gpp
for
this
kind
of
applications
and
this
kind
of
deployment,
and
so
for
the
itp
domain.
You
may
have
a
small
set
of
routers
and
then
the
e-waste
router
will
have
the
server's
edge
server
edge
computing
server
directly
attached
to
the
routers.
So
this
is
the
basic
background
so
very
frequently
and
it
has
been
used.
H
The
many
benefits
of
using
anycast
basically
is
to
leverage
the
network
condition
to
distribute
the
packets
so
that,
if
there's
changes
in
the
ue
behavior
like,
for
example,
lots
of
users
move
to
one
location
for
like
concert
or
conference
call
or
gathering
and
the
servers
nearby.
The
cell
tower
may
be
overbooked
over
utilized
and
at
this
point,
that,
like
the
network,
routing
distance
to
the
others
may
be
very,
very
small
difference.
H
So
compare
with
the
utilization
of
the
server
so
that
the
network
itself,
if
any
cases
used,
can
automatically
migrate.
The
traffic
to
the
location
where
there
are
less
utilized
servers
available,
compute
available
and
also
it
is
make
it
possible
to
eliminate
the
single
point
of
failure
like
a
dns
resolver
or
the
load
balancer
like
typically
today
we
have
their
net
application
layer
load
balancer,
which
has
many
servers
attached
behind,
but
to
a
network.
H
There's
only
one
address
where
we
see
the
network
load
balancer,
but
if
the
condition
network
condition
to
the
load,
balancer
has
some
issues
then
impact
the
the
performance
so
using
any
task
can
without
can
eliminate
that
issues.
Also,
many
clients
may
use
stealth
ip
addresses.
They
may
only
retrieve
the
ip
address
once
a
while,
so
when
they
move
from
one
cell
tower
to
another
cell
tower,
they
may
still
have
the
old
ip
address,
and
this
is
become
more
acute
in
a
5g
environment,
because
in
4g
you
don't
have
that
much
issue
4g.
H
Maybe
the
entire
city
only
has
one
packet
gateway,
so
everybody
traffic
doesn't
matter
where
you
move
to.
Unless
you
drive
very
far
away,
your
traffic
all
entered
into
the
same
ingress,
router
to
the
iq
network
or
the
same
packet
gateway,
but
in
5g
environment
there
will
be
many
many.
H
Those
user
plan
functions
and
packet
anchor
points
so
that
when
user
move
the
the
traffic
move
and
their
associate
the
local
network
local
network
routers,
they
were
going
through
the
different
ingress
routers.
So
so
that's
basically
the
benefit
using
any
cast
and-
and
then
this
right.
This
particular
contribution
of
the
draft
is
to
bring
up
another
layer
of
matrix
in
compute
the
past
shortest
path.
H
Since
those
those
servers
they
are
attached
to
the
the
egress
routers
and
like,
for
example,
the
site
can
have
different
value
or
preference
and
they
may
have
different
capacity,
and
so
we
want
to
bring
this
information
into
the
igp's
constraint,
past
compute,
okay,
so
the
overview
of
the
solutions,
pretty
simple.
H
So
one
thing
is
the
the
egress
router
need
to
advertise
the
site
cost
of
the
in
the
id
prefix
reachability
tlb,
to
indicate
the
the
value,
the
cost
value,
to
reach
specific
prefix
and
and
bear
in
mind
that
not
everybody,
not
every
service
in
this
ldl
required
is
preferred
or
specialized
compass
compute.
Only
a
small
set
of
prefix
like,
for
example,
if
you
use
this
aws
wavelength
zoom
right.
If
you
want
a
low
latency
service
as
a
as
a
client,
they
have
to
register
their
service
with
the
provider.
H
So
so
this
this
is
and
then
we're
using
using
the
flat
algorithm,
there's
a
new
flag
set
to
indicate
that,
for
this
specific
prefix
that
you
need
to
consider
the
site
cost
to
balance
the
the
actual
cost
to
the
specific
prefix.
H
So
that's
the
basic
overview
of
this
and
then
for
the
flex,
algorithm
definition
of
flags
of
tlv.
We
introduce
a
new
flag
to
indicate
that
this
side
cost
matrix
has
to
be
need
to
be
included
to
deliver
to
derive
the
constrained
path
to
the
prefix.
So
that's
just
a
new
flag
need
to
be
introduced
once
this
new
flags
to
be
introduced
and
then
the
as
the
egress
node
it
can
advertise
the
the
cost.
H
The
aggregate
costs
which
include
the
like
a
low
back
a
load
index
may
the
egress
node
may
measure
the
the
package
to
the
node
or
from
the
node,
from
the
prefix
to
the
prefix,
and
also
including
the
site,
preference
and
other
other
parameters.
They
compute
aggregate
cost
and
this
cost
can
be
advertised.
H
A
L
D
L
To
load
balance
the
traffic
onto
these
application
servers.
So
to
me
it
looks
like
it's
a
problem
of
load
balancing
onto
the
servers
and
probably
what
I
feel
is
metric
probably
is
not
the
right
way,
because
the
moment
you
change
the
metric
based
on
load
or
preference
or
whatever
your
traffic
gets
skewed
towards
that
metric.
And
then
you
have
an
oscillation
right
and
you
have
more
ues
connecting
to
this
particular
servers,
because
this
metric
looks
better
so
because
you're
describing
this
as
a
load
balancing
problem.
L
What
I
think
you
really
want
to
do
is
to
use
the
factors
that
you
have
described
like
the
load,
the
preference
and
the
site
cost,
etc
to
compute
the
balance
and
low
and
use
it
as
a
load
balancing
mechanism,
basically
unequal
load,
balancing
on
ecmp
any
cost.
Prefix.
A
A
Do
it
with
the
pizza
you're
right
I
mean
the
hysteresis
here
are
very
important
right
because
if
you
add
one
client
to
the
thing
and
then
the
metric
changes
and
then
you're
bouncing
back
and
forth
between
the
two
sites
right
for
each
client
that
gets
added.
So
you
have
to
put
some
kind
of
hysteresis
on
this,
so
that
that
doesn't
happen.
But
eventually
that
is
what
you
want
to
happen
right,
that's
load
balancing!
You
know
going
back
and
forth
between
the
two
you
just
maybe
don't
want
to
happen
on
a
per
client.
L
Exactly
so
to
me,
it
appears
that
it's
a
load,
balancing
problem
and
you
know
changing
metric-
doesn't
really
isn't
really
going
to
solve
the
problem.
So
I
think
there
are.
There
are
other
ways
of
doing
it
and
even
igp
right.
The
load
balancing
is
generally
done
on
next
top
equal
next
stops,
though,
you
can
do
load
balancing,
you
can
do
unequal
load
balancing
as
well,
but
on
an
anycast
prefix
that.
A
E
B
Link
metric
and
the
one
on
the
stub
stub
link
that
gives
a
nod
to
the
stub
link.
E
H
Yes,
yes,
a
few
sections
because
just
last
few
days,
discussion
with
peter
and
I
shrink
the
slides,
but
I
haven't
got
a
chance
to
okay.
B
Update
the
draft
and
the
other
one
is,
if
you're
not
using
the
raw
metrics,
if
you're
just
using
a
an
aggregated
metric,
you
can
do
this
today.
With
I
mean,
irrespective
of
all
the
problems
with
load,
balancing
using
routing
that
right
and
chris
were
discussing,
you
could
do
it
just
with
base
ospf
just
make
the
mix
the
anycast
address
an
external
route
and
use
the
type
2
metric.
That's
your
for
the
external
metric!
That's
your
type
2
metric
there
and
it
all
works
anyway.
D
A
On
a
slop
at
this
point,
so
next
presentation,
thank
you
linda
and.
A
O
We
know
that
I
start
leveraged
source,
routine
mechanism
and
the
process
and
prs
labels
are
as
a
value
side
in
the
package
header.
However,
the
current
mechanism
in
isa
do
not
have
the
specific
path
contractor
signaling,
so
that
the
past
mto
information
cannot
be
abandoned
in
advance.
Therefore,
it
cannot
be
ensure
that
the
packet
size
is
less
than
the
past
mto.
O
For
the
sis
attention
we
defined,
we
defined
a
new
sub
tree
called
link
mtu
subtree
to
carry
the
mto
of
the
interface
associated
the
link.
The
link
mto
is
advertised
as
an
optional
subtitles
of
the
six
rider.
The
ties,
the
less
of
the
types
failed,
the
less
fair,
the
last
failed
and
the
value
failed
are
one
better
one
byte
and
two
bets
respectively.
O
The
link
mto
is
advertised
as
an
optional
subtly
of
the
ospf
way
to
attain
the
link
to
awa
in
the
ospf
way
to
attain
the
link
opaquer
lsa
for
ospf
raceway.
The
link
mto
is
advertised
as
an
optional,
subtle
way
of
the
link
to
in
the
ospf
wave
3
e-rotor
rsa,
the
last
of
the
type
field,
the
the
last
failed
and
the
value
failed
are
all
two
batters.
O
O
A
O
A
A
Question
I
it
looks
like
there's
a
couple
people
in
the
line.
I
would
just
throw
out
the
comment
that,
after
reading
it,
I
think
your
tie
break
should
just
be
lowest
time
to
you
and
forget
all
that
stuff
about
encoding,
ids
and
all
that
go
ahead.
Jeff.
A
J
So
my
criticism
is
not
specifically
targeting
you
know:
how
do
we
carry
this
inside
of
the
protocol
mechanism?
My
comment
to
you
is
that
the
general
issue
that
we
seem
to
run
into
in
the
real
world
is
that
we
can't
believe
what
the
protocols
are
telling
us
in
a
lot
of
cases.
So
I
know
that
you
want
to
carry
this
stuff
in
the
idp
to
signal.
You
know
to
some
front
end.
Here's
what
my
path
should
actually
be,
whether
you
believe
it
or
not.
J
J
The
general
problem
that
you're
going
to
have
is
you're
told
the
path
mtu
is
x,
but
some
component
link
that
may
be
part
of
that
path.
Part
of
a
backup
link,
part
of
a
fast
reroute
scenario,
pick
any
number
of
scenarios.
The
path
m2
is
not
going
to
be.
You
know
what
you
actually
are
told
that
it
is,
and
it's
important
to
not
trust
that
that's
my
comment.
Thank
you.
G
Yeah,
so
I
share
jeff's
concerns.
I
have
a
limited
enthusiasm
for
this
draft,
but
in
rig
my
comment
is
in
regards
to
isis.
If
the
working
group
decides
to
move
forward
with
this
draft,
there
are
already
two
code
points
that
potentially
can
be
used
to
advertise
this
one
of
them,
particularly
the
per
link
advertisement,
was
defined
by
trill.
G
A
Thanks
les
easy.
B
Yeah
yeah
we've
got
my
comment.
Speaking
of
working
member
is
we've
got
a
a
long
without
this
for
decades
it
seems
like
just
a
protocol.
You
know
bringing
this
in
because
it's
it
can
be
done.
I
don't
know.
I
looked
at
the
use
case
of
sr
and
what
I
would
typically
do.
Is
you?
Definitely
you
don't
you
know
if
you
want
to
generate
mtu's,
I
would
just
take
the
lowest
one
in
the
in
the
whole
domain
and
use
that
as
my
sending
them
to
you,
so
I
could
make
sure
that
any
opposed
labels.
B
B
A
Yeah,
the
other
okay.
Well,
these
are
good.
These
are
good
list
comments,
I'll,
throw
I'll
throw
out
there
that
you
know
we
we're
tunneling
like
crazy,
but
in
the
last
you
know
five
years
or
so
so
we
that
that's
getting
worse.
A
And
worse,
but
yeah
anyway,
these
are
good.
Let's
have
the
discussion
on
the
list.
It's
a
new
draft
thanks,
zing,
all
right,
we
might
be
back
on
time
almost.
N
Okay,
hello,
everyone,
it's
xiaomi
speaking.
This
presentation
is
on
signaling
flow
id
label
capabilities
and
the
flow
id
readable
label
depth
using
igp
and
the
bgpls.
N
N
We
already
have
obviously
a
1988
that
defines
the
mechanism
to
signal
entropy
label
capability
and
entropy
readable
label
depth
using
iss
and
vgpos,
and
we
also
already
have
obviously
a
at
9080
89
that
defines
mechanism
to
signal
entropy
level
capability
and
entropy
readable
label
depths
using
ospf
and
bgpls.
N
N
N
N
N
The
borrowed
flc
flag
is
bit
number
four,
which
is
next
to
the
elc
flag,
defined
in
section
three
of
c
1988.
A
Because
you
only
have
three
minutes
left,
I
don't
think
you
have
to
necessarily
go
through
every
encapsulation
here.
Okay,
maybe
just
stick
to
the
main
points.
A
N
Yeah,
so
this
job
to
just
model
after
rc,
1988
and
1989
to
advertise
the
flow
id
label
capability
and
variety
rather
readable
level,
level
depth
using
rgb
and
pgpos.
So
that's
the
intention.
That's
the
point.
A
Yeah,
okay,
so
is.
N
A
Is
this
is
pretty
new?
My
al,
my
comments
are
quick.
You
you,
I
don't
know
why
you
did
prefix.
You
say
it's
useful,
but
whatever
we
can
talk
about
that
on
the
list,
but
you
you
also
say
that
every
interface
has
to
support
it.
So
it's
almost
like
when
you
propose
to
like
somebody,
I'm
gonna
sell
you
a
car
for
a
billion
dollars
and
then
they
say
that's
ridiculous.
I'll
give
you
500..
A
A
A
But
you
know
just
like
ifit
came
into
the
working
group
and
they
wanted
to
do
things
that
didn't
have
to
do
with
forwarding.
They
wanted
to
signal
capabilities
that
didn't
have
to
do
with
forwarding.
I
think
that
violates
what
an
ig
igp
should
be
advertising
right.
That's
my
opinion.
Working
group
chair
not
as
a
working
choice.
N
I
understand
yeah,
I
understand
your
point.
Okay,
falc
is
used
for
performance
performance,
it's
not
for
direct
forwarding
all
right
right.
B
And
state
your
first
comment:
more
simply:
rfc
9088
and
908.9
advertise
this
mpls
capab
these
mpos
capabilities
on
a
link
and
node
level.
Yet
you
put
it
at
the
prefix
level.
I
don't
know
why
you
put
it
at
the
prefix
level.
It
seems
like
this
should
be:
have
a
similar
form
and
structure
to
those
two
rfcs,
as
opposed
to
advertising
things
on
a
prefix.
B
G
Less
so,
first
yeah
so
first
off,
I
agree
with
chris's
comment
about
this,
not
necessarily
being
appropriate
for
the
igp.
G
I
think
the
mystery
about
why
you've
chosen
prefix
reachability
is
you've
been
misled
by
the
entropy
drafts,
because
we
had
a
special
case
there,
where
we
wanted
to
advertise
entropy
capability
out
of
the
area,
and
so
we
needed
to
use
a
prefix
reachability
advertisement
to
do
it,
but
it's
really
a
node
property
and
what
you're
talking
about
here
is
also
a
known
property.
So
if
we
were
to
advertise
it,
it
should
go
into
router
capabilities,
but.
A
All
right,
thanks,
y'all,
we
gotta
move
on
so
up.
Next
is
the
anacost
property
run.
P
Okay,
hi
everybody
I'm
from
dt
today.
I
would
like
to
talk
about
updates
to
unicast
property
advertisement
for
spf
on
behalf
of
co-author
next.
Oh
sorry,.
P
Oh
yeah,
yeah,
okay,
I
found
the
bot
okay.
Yes,
this
is
the
moto
business,
both
both
smps
prefix
seed
and
ipv4,
ip
and
ipv6
prefix
may
be
configured
as
any
cost
and,
as
such,
the
same
value
can
be
otherwise
advertised
by
multiple
daughters.
P
Fc
7684
defines
ospf
v2
opic
ios
a
based
on
tr
way
that
can
be
used
to
associate
additional
attributes
with
prefix
or
links.
P
And,
however,
three
bits
have
been
defined
for
fc.
8362
extends
the
ios
a
format
by
encoding,
the
existing
ospf
v3,
I
always
say
information
in
tla
and
for
allah
and
allowing
advertisement
of
additional
information
with
additional
tr
way.
Its
prefix
is
advertised
along
with
an
eight
eight
bit
field
of
capability
by
using
the
prefix
options,
but
the
definition
of
any
cost
of
like
to
identify
the
prefix
as
any
cost
has
not
yet
been
defined.
P
P
That
can
be
reused
in
multiple
of
different
rsa
as
as
sub-tree,
and
we
also
introduce
a
new
flag,
a
unicast
flag
in
the
prefix
attribute
sub
tree
to
advertise
the
unicast
property,
and
this
flat
is
for
the
wearable
lens.
Prefix
attribute
subtle
way.
P
So
in
so
in
the
case
of
ospf,
we
need
to
the
prefix
attribute
sub
tov
is
a
subtlety
of
the
ospf
v2
extended
prefix
tray,
as
defined
in
rfc
7684
and
in
the
case
of
ospf,
which
three
the
prefix
attribute.
Subtree
is
a
subtlety
of
the
intra
air
prefix
tray
inter
error,
prefix
cova
and
external
prefix
tailway
for
the
next
slide
is,
is
fold
attribute
the
the
definition
of
attribute
tr
way
for
osp
for
the
ospf
way
to
the
following
the
following
flag.
P
Please
say
that
the
first
figure
has
been
defined
and
the
first
eight
bit
as
reserved
for
the
previously
defined
uspf
v2,
extended
prefix
tr
way
for
the
an
for
the
a
under
un
beat.
Please
refer
to
abscess,
seven,
eight,
seven,
six,
eight,
four
and
ebay.
Please
refer
to
fc
nine
zero,
eight
nine,
and
in
this
document
we
defined
a
new
flag
ac
flag,
as
the
new
flight
is
used
to
advertise
unicast
property
in
the
case
of
ospi
v3.
P
Please
see
the
figure
2
and
also
the
first
8-bit,
as
result
for
the
pro
previously
defined
uspf
v3
prefix
options
and
the
ac
like
is
used
to
advertise
undercast
properties.
So
so
we
have
defined
the
following
processes:
use
processing
rules
for
easy
flights.
The
first
is
when
the
prefix
is
configured
as
uncast.
The
ac
flag
should
be
cited.
P
Otherwise
this
flag
must
be
clear,
and
the
second
is
if
both
in
flight
and
ac
flight
assigned
the
receiving
daughters,
I
must
ignore
the
end
flag
and
the
second.
The
third
is
ac.
Flag
must
be
preserved
when
the
prefix
is
properly
propagated
between
areas
and
the
last
is
the
same.
Prefix
can
be
advertised
by
multiple
daughters
and
that,
if
any
at
least
one
of
them
sets
the
easy
flag
in
its
advertisement,
the
prefix
should
be
considered
as
any
cost
in
the
last
last
slide.
P
The
first
is:
if
there
is
a
device
in
this
network
that
don't
support
extensive
of
the
prefix
triple
sub
troa,
then
the
device
that
supports
the
prefix
triple
sub
toa
should
advertise
the
field
of
the
capability
of
the
prefix
by
using
prefix
obstacles
or
previous
flags
and
the
prefix
attribute
sub
tree.
P
Otherwise
only
use
the
prefix
attribute
subtle
way
to
advertise
the
field
of
capital
capability
of
the
prefix
and
the
the
second
is
pre
prefixes
and
our
test,
along
with
the
field
of
capability
by
using
the
prefix
or
tributary
array,
then
the
field
of
capability
of
the
prefix
in
the
prefix
or
triple
sub
two.
We
sell
preview
and
as
well
as
the
prefix
sweeping
sub
troa,
is
used
to
advertise
the
field
of
capability
and
the
device
supports
the
prefix
attributes
of
troa.
P
Then
the
field
of
capability
in
the
prefix
or
treatment
supply
should
be
prevailed,
and
the
last,
if,
if
is
if
prefixes,
is
advertised
along
with
the
field
of
capability
by
using
only
the
prefix
substance
or
prefix
flags,
then
the
field
of
capability
in
the
prefix
offices
of
prefix
apply
flags
sharper
will
approve.
Yes,
this
is
our
priorities.
That's
all
thank
you
and
request
request,
review
and
feedback
from
a
working
group.
A
B
Guess
I'm
at
the
top
of
the
queue
ac
linda,
I'm
speaking
as
member.
The
first
question
is:
what
is
the
use
case
for
any
castle?
I
thought
the
whole
idea
was.
The
is
is
that
it
was
that
it
was
transparent.
B
B
That's
not
the
case
because
you
still
have
to
you
still
should
set
it
for
the
existing
flags
for
backward
compatibility,
because
the
it's
part
of
the
prefix
in
both
cases
and
both
for
both
7684
and
the
ospf
b31
8362
or
whatever
it
is
the
the
the
the
prefix
options
in
the
case
of
ospf
b3
and
the
and
the
flags
field
in
the
case
of
ospf
v2,
extended
prefix
is
part
of
the
same
field
that
has
the
address,
so
you
really
have
to
see.
So
you
can't
you
can't
not
use
it.
Yeah.
A
So
my
comment
is
actually
the
same.
I
think
this
maybe
we
need
to
go
over
this
on
the
list,
but
when
I
I,
when
I
read
the
you
know
the
sort
of
transition
that
that
last
slide
talked
about
to
me,
it
sounded
wrong.
A
At
least
when
I
read
the
document
it
was
like
it
preferring
like
the
new
advertisement
over
the
old
one
is,
is
almost
always
wrong
right,
because
the
the
non-implementers
or
people
who
don't
understand
it
are
picking
something
different,
but
I
could
have
misread
it
either
case.
I
think
that
needs
some
review
on
the
list.
B
B
F
F
As
shown
in
in
this
diagram,
the
number
first
path
is
the
normal
igp
path
which
has
the
shortest
metric,
and
we
may
notify
the
link
bit.
Error
ratio
of
every
link
by
the
router
to
the
controller
using
bprs
and
the
controller
will
calculate
a
policy
path
which
has
the
lower
link
beta
ratio,
as
shown
by
number
two
parts,
so
that
different
service
may
use
different
parts.
F
The
format
is
shown
here.
It
is
similar
to
u-directional
link
losslab
theory.
The
type
is
to
be
defined.
The
length
is
a
constant
value
four
and
we
define
one
byte
flags
and
only
use
one
bit
here.
We
call
a
normal
speed
and
a
bit
the
8-bit
is
set
when
the
measured
value
exceeds
this
configured
max
maximum
threshold
and
will
clear
when
the
value
falls
below
is
configured
reuse.
Threshold.
F
A
A
A
My
experience,
the
operator
I've
worked
for
was
deutsche
telekom
right,
so
it
was
way
up
high
in
the
pecking
order,
and
we
worked
on
a
new
new
network
and
for
us
any
post,
fec
bit
error
rate
meant
you
took
the
thing
out
of
service.
It
was
no
percentage
any
prefect
bit
error
rate
unless
it
was
immediately
spiking
just
meant
the
thing
was
working
right.
A
Now
there
on
all
the
routers
there's
there's
you
know,
alarms
and
different
things
that
can
watch
this
stuff.
I'm
just
not
sure
how
useful
this
is.
I'm
interested
in
hearing
from
other
people
and
other
specifically,
you
know
either
other
operators
at
the
you
know,
sort
of
all
free,
but
also
you
know,
maybe
does
this
have
some
kind
of
application
in
lossier
networks-
I
I
don't
know,
but
I
I'm
worried
that
this
is
too
too
fine
of
a
grain
to
be
looking
at
on
the
routing
level.
B
Yeah
I'll
just
I'll
just
come
in
there
speaking
as
working
group
member,
I
see
the
encoding
looks
fine.
I
mean
it's
similar
to
the
other,
extended
te
metrics
that
we
spent
a
lot
of
time
on
years
ago,
but
I'm
wondering
what
bit
error
rate
offers
you
over
and
above
packet
loss.
B
They
seem
like
they're,
you
know,
basically
both
at
a
high
level.
Metrics
of
you
know
the
link
quality
and
really
it
seems
like
that.
What's
delivered,
the
packet
loss
is
a
more
important
metric
than
the
the
lower
level
at
the
you
know
at
layer,
one,
the
bit
error
rate
the
physical
layer.
So
I
I
I
don't
really
see
the
need.
A
Yeah
randy
agreed
with
me
in
the
chat
that
you
know
if
you,
if
you
see
post
fact
bit
errors,
you
just
take
it
out
of
service
yeah.
So
just
this
is
all
just
echoing
the
same
thing
right
then
ac
is
saying
the
same
thing
we'll
take
it
to
the
list,
but
this
may
not
be
that
useful.
B
So
the
question
now
is:
do
we
want
to
try
and
load
itunes,
stub,
link
or
advertisement
presentation
quickly
or
not.
M
Okay,
yeah,
so
let's
so,
this
is
the
advertise
of
the
stop
link
attribute.
M
So
so
there
are
many
players
and
many
situations
and
the
sub
interface
interface
are
used
commonly
in
network,
but
currently
there
is
no
suitable
place
to
advertise
the
staff
interface
and
their
associated
attributes,
and
we,
we
also
investigated,
agreed
possible
solution
and
there
is
no
country.
There
is
no
ceiling
place,
so
we
think
it
is
necessary
to
extend
the
oil
pfo
and
the
ice
protocol
to
transfer
the
stabling
and
release
the
attributes
and
the
updated
content
for
the
current
contract
is
based
on
the
discuss
on
the
main
list.
M
We
found
that
the
interest
in
ospf
and
the
interest
digitality
kelvin
isis
is
a
suitable
place
to
contain
such
information,
so
we
define
the
sublingually
or
or
under
sub
the
subtle
way
to
carry
the
associated
attitude
within
osp
and
I
iss,
and
we
also
define
the
link
type
for
the
sub
sub
link,
because
there
are
various
types
of
stub
link.
So
we
can.
We
need
to
further
distinguish
them
and
the
other
existing
subtlety
for
the
link
attribute
can
be
reused.
M
And
we
also
con
discuss
the
other
theory
that
is
the
there
is
a
legacy
already
defined
in
rc,
36,
30
and
30,
and
they
we
also
investigate
the
american.
M
This
theory
can
be,
can
be
used
or
not,
and
we
think
the
such
the
length
of
the
arena
should
be
mainly
used
for
the
link
within
the
itv
protocol,
but
the
sub
link
is
not
in
the
id
protocol
and
and
the
if
we
use
such
a
sub
here,
the
limit
type
regions
within
such
a
circular
way,
so
the
should
also
philosophy
should
be
extended,
and
there
is
no
sub
theory
to
describe
the
prefix
that
associated
with
the
studying
so
based
on
this
consideration,
we
think
it
is
fine
to
define
my
new
neutral,
similar
theory
to
container
information
for
the
sub
link.
M
Okay,
this
is
the
the
update,
the
content
and
the
update
of
results.
I
think
it's
also
discussed
on
the
mainly
intensely,
and
mainly,
we
also
ask
for
the
working
group
to
adopt
okay,
so
yeah,
linda.
H
This
strat
seems
like
we
can
use
that
for
distribute
the
aggregate
costs
as
well.
So
I'm
curious:
what's
the
stop
link
how's
the
difference
from
like
a
prefix
lsa
like,
if
is
the
sampling,
is
always
for
prefix.
M
No,
no,
I
I
think
the
stop
link
account
can
be
used
to
connect
many
many
prefix.
You
know,
for
example,
in
your
in
your
list
for
the
5g
edge
computer.
I
think
the
sublingualism
is
linked.
That's
the
located
on
the
boundary
of
the
s
so.
A
M
A
That's
the
interesting
case
we're
out
of
time.
I
I'm
always.
I
don't
know
whether
it's
appropriate
or
not
to
run
over.
I
think
virtual.
We
can
run
over
a
little
bit,
but
it's
a
little
bit
better
than
when
you're
no.
M
A
Yeah,
I
I
mean
I
I
would
encourage
people
to
talk
about
this
on
the
list.
This
is
another
one
where
I
don't
think
we're
gonna
do
an
adoption
call
right
now,
because
you
know,
I
think,
there's
still
some
questions
as
to
whether
it's
needed.
I
know
as
a
contributor.
A
I
I
don't
see
why
it's
needed,
but
you
know
that's
just
me
as
a
contributor,
but
I'd
like
to
see
some
more
discussion
on
the
list
about
this,
particularly
if
people
have
ideas
about
how
to
how
to
identify
stub
links
without
a
tlb
right,
because
I
think
that
could
be
done
so
then
we
wouldn't
need
this,
but
anyway
more
on
the
list,
and
I
think
that's
it-
for
presentations.
B
A
B
A
A
Okay,
great
thanks,
everybody
any
other
comments.
We
do
have
this
virtual
time.
So
then
we're
gonna
throw
some
tomatoes
at
us
or.
B
B
A
A
I
think
I
got
another
plan.
It's
recorded
here,
so
yeah,
it's
not
recorded
at
the
end
of
the
session
when
you're
just
gathered
around
the
chair
table
john
says
running
over
is
not
ideal.
I
agree,
that's
always
been.
My
thought
is
that
people
need
to
plan
and
it
disadvantages
people
that
have
other
things
to
do
all.