►
From YouTube: IETF114-RIFT-20220728-2000
Description
RIFT meeting session at IETF114
2022/07/28 2000
https://datatracker.ietf.org/meeting/114/proceedings/
A
Okay,
so
it's
four
o'clock
now
here
in
the
room:
we
have
three
people,
I,
don't
know
how
many
online.
Let
me
actually
let
me
check.
B
We've
got:
let's
see
how
many
about
15
people.
A
I
see
our
main
presenter
is
here:
Jordan
head
I
think
we
can
get
started.
I
see
that
Alvaro
is
online,
but
our
are
you
there.
A
A
A
B
From
here
so
agenda
for
today
we
are
going
to
provide
quick
update
also
with
regards
to
rechartering
and
work.
You
would
like
to
progress
and
then
Jordan
is
going
to
talk
about
updates
to
the
base
spec
and
Delta
VPN
KB
registry
I.
Think
we
reached
good
consensus
with
Ayanna
Fox
here
and
the
assistant
recharging
discussion
website.
B
So
Alvar
thanks
for
your
review,
it
came
back
so
we
are
working
on
it.
Applicability
statement
is
still
pending.
It
is
review
young
models
yesterday
submitted
to
isg
KV
registry.
As
we
said,
British
consensus
with
Ayana,
and
now
we
need
Shepherd
right
up
and
Rift
segment
routing
has
been
adopted
as
a
new
working
group
document.
A
So
then
the
next
presentation
is
from
Jordan
head.
Let
me
just
download
the
material.
C
C
C
So
with
alvaro's
feedback
that
came
back,
here's
you
know
some
of
the
stuff
that's
going
to
come
up
as
the
major
feedback
would
be
some
of
the
ASCII
diagrams,
didn't
exactly
they
weren't
parallel
with
their
SVG
counterparts.
Now
that
there
aren't
requirements
for
things
to
be
exact,
but
they
need
to
be
at
least
comparable,
so
that
both
are
both
formats
of
the
document.
Have
you
know
roughly
the
same
experience
things
like
textual
references
and
diagrams
and
so
forth.
C
Other
clarifying
changes
to
yeah
next
slide,
just
a
little
bit
of
a
visualization
for
those
that
are
interested.
So
what
I
mean
by
not
parallel
so
the
SVG
on
the
left.
C
You
know
pretty
cool
and
so
forth,
but
the
the
two
ASCII
versions
on
the
right,
the
the
top
is
the
before
and
the
bottom
one
is
is
after
it,
you
know
more
clearly
reflects
the
point
of
you
know
the
in
this
case,
intertography
inner
top
of
fabric
rings
on
the
Fabrics
again,
the
svgs
are
kind
of
more
of
a
should
than
a
should.
C
But
again
it's
about
it's
an
accessibility
thing
right.
The
need
to
be
able
to
read
relatively
similarly
next
slide.
C
D
Yeah
one
thing
that
came
to
mind
just
now
is
that
the
svgs
are
of
course
nicer
and
and
easier
to
understand.
If
I
was
reading
the
document
I
don't
know.
If
we
can
put
a
note,
maybe
especially
at
the
beginning
of
that
first
section
say
you
know:
go
look
at
the
PDF
type
thing,
because
most
people
are
going
to
go,
look
at
the
HTML
version
or
something
and
they
might
not
get
the
the
benefit
of
looking
at
and
all
the
and
all
the
nice
pictures.
C
Yeah
I
know
this
was
when
we
were
first
doing
the
svgs.
It
came
into
a
whole.
It
was
a
whole
thing,
figuring
out
what
was
allowed,
what
wasn't
you
know,
grayscale
versus
color
and
so
forth-
we're
not
actually
allowed
well.
C
We
were
advised
not
to
reference
specific
formats
in
the
text
at
all,
at
least
that's
what
I
saw
it
ties
back
to
Geneva,
Convention,
sort
of
things.
It's
pretty
interesting
as
far
as
accessibility
and
readability
in
certain
parts
of
the
world
may
not
have
access
to
something
that
could
render
HTML
properly
things
like
that,
but
yeah
so
wanted
to.
But
okay
next
slide
just
kind
of
other
stuff
there
I
know
you
mentioned
the.
C
We
should
request
early
allocation
for
the
multicast
addresses
and
the
UDP
ports,
we'll
I'll
make
a
note
to
the
chairs
or
I'll
send
email
to
the
chairs.
When
we
push
version
16.,
you
know
be
on
that
various
editorial
things:
readability,
Clarity
aspects,
some
reference
fixes
and
I
still
eat
responses,
so
there'll
be
a
couple
of
things
that
will
have
a
little
bit
of
a
discussion
about
the
vast
majority
of
feedback.
C
Is
super
straightforward,
next
slide
and
there's
some
things
unrelated
e-review
that
I
kind
of
noticed
so
well,
first
off
I'll
be
added
as
the
sixco
author,
thanks
for
helping
me
weed
through
that
it'll
require
a
minor
Shepherd
update,
but
ads
mentioned
that
that's
a
pretty
straightforward
thing
did
some
fixes
for
minor
fixes
for
the
thrift
schema
with
regard
to,
what's
already
reflected
in
the
KV
draft,
they
just
a
little
bit
out
of
date.
C
There
other
SVG
fixes,
just
you
know
size
scale
things
some
of
the
in
the
previous
versions.
They
were,
they
scaled
really
really
poorly
and
it
was
it
was
and
other
kind
of
SVG,
XML,
optimization
and
that
had
to
do
with
there's
actually
a
three
Meg
limit
on
the
draft
upload
tool
that
cause
all
sorts
of
fun
problems.
C
We
had
to
render
the
pdf
version
and
the
pdf
version
doesn't
render
the
svgs
very
well
but
anyway,
so
that
now
everything
should
kind
of
fall
into
and
they'll
get
the
full
benefit
like
the
HTML
version
I'm
going
to
well.
We
we
have
a
get
issue
open
with
them,
but
hopefully
there
won't
be
pushback,
so
I
might
bug
you
depending
upon
things,
but
okay
next
slide
I
think
that's
yeah!
That's
it
other
questions.
Before
move
on.
C
Okay,
just
gonna
provide
a
brief
update
on
auto
flood
reflection
next
slide.
So
just
a
quick
refresher
for
those
aren't
familiar.
It's
it's
an
experimental
draft
out
of
LSR,
but
the
short
version
of
it
is
so
you
know
there
are
lots
of
reasons
to
use.
You
know
flat
single
area
igps,
but
they
come
with
drawbacks.
C
You
know:
flooding
State
convergence
right,
every
node's
got
I,
know
hold
that
state
and
then
compute
that
state
so
there's
a
link
to
the
actual
flood
reflection
draft,
but
we'll
kind
of
circle
and
with
Rift
here
in
a
second
next
slide.
C
So
flood
reflection,
like
Rift,
is
very
well
suited
for
cloth.
Apologies,
so
Rift
will
take
care
of
the
underlay
and
then
the
auto
FR
extensions
will
use.
You
know:
Rift
CTP
and
the
LIE
exchanges
to
build
a
flood
reflection
topology
on
top
of
that
so
next
slide.
C
So
where
does
it
actually
fit
in
so
flood?
Reflection
requires
loopback
addresses
system
IDs
for
Isis
Network
entity,
titles,
which
just
makes
the
single
address,
and
then
cluster
ID
to
actually
segment
the
flood,
reflection,
domains
and
so
Auto
FR
has
to
actually
build
and
derive
those
variables
using
the
distributed
State
we
get
in
the
rift
so
next
slide.
C
So
what
that
really
kind
of
looks
like
is
so
Z
rip
CTP.
Will
you
know
it'll
drive
each
level
it'll
build
the
Riff
fabric
next
slide,
the
the
FSM
will
help
derive
all
of
the
flood
reflection
roles
and
variables,
and
you
know
basically
take
the
what
would
be
configuration
data
and
then
next
slide
and
then
finally,
you
end
up
with
Isis
flood
reflection
on
top
of
everything,
so
the
adjacences
has
come
up.
You
have
the
flood,
Reflections
apology
and
there
you
go
you
basically
plug
in
everything.
You
have
a
whole
fabric.
C
Okay
next
slide.
So
as
far
as
the
actual
work
in
the
draft,
so
there
were
some
concerns
with
how
you
know
the
iso
address.
Derivation
wasn't
very
clear
where
it
came
from.
Unless
you
went
in
and
looked
at
the
thrift
model
it's
there,
but
it
should
be
kind
of
aligned
in
the
text
as
well.
C
The
kind
of
did
a
new
section
refactoring
of
sections
for
all
of
the
extensions
required
to
Rift.
So
it's
like
Auto
FR
works.
So
that
would
be
you
know
the
finite
State
machine,
no
tie,
advertisements
and
kind
of
any
applicable
schema
updates.
Basically,
prior
to
kind
of
refactoring,
it
felt
a
little
disjointed.
You
have
to
bounce
around
the
draft
to
figure
out
what
was
what
was
doing,
what
so
more
readable.
Now
next
slide.
C
Okay,
we
added
a
flood
reflector
preference
value,
which
means
we
need
to
update
the
election
procedures
to
factor
that
in
and
then
the
key
value
ties
are
more
of
a
editorial
stuff.
So
a
little
bit
of
normative
language
and
kind
of
we've
since
reworked
the
game
graph
just
a
little
bit.
So
we've
got
the
request:
criteria:
aligned,
okay,
so
next
slide.
C
Okay,
just
a
little
bit
of
a
you
know,
visual
fluff,
if
you
will
as
to
how
what's
clarified
so
we
those
are
the
functions
that
are
tied
to
what
are
actually
in
the
draft
in
the
appendix
that
kind
of
show.
You
know
here's
how
we
derive
it,
and
these
are
the
you
know
the
individual
functions
that
build
the
parts
of
the
address
that
we
need.
So
it's
that
image
isn't
there,
but
it's
described
in
text
now
versus
not
next
slide.
C
Right
so,
like
I
said
what
what
it
was
before
was,
with
things
being
so
disjointed,
put
everything
in
a
single
section
and
said:
you
know:
here's
the
flood,
reflective
version,
the
cluster
ID
and
the
preference
values,
and
there
were
a
little
there's
a
little
bit
of
a
functional
change
with
that
that
wasn't
described
before,
to
kind
of
factor
in
for
interoperability
with
nodes
that
do
run
it
nodes
that
don't
run
it
nodes
that
do
support
it,
but
don't
want
to
run
it
so
next
slide.
C
So
quick,
look
at
the
actual
reflector
election.
Previously
it
was
just
High,
Sid,
low
acid,
second
highest,
second
lowest,
that's
basically
to
prevent
you
know
any
single
node
coming
in
from
flipping
the
entire
Fabric
and
killing
all
the
adjacencies.
C
C
Like
I
said
for
key
value
ties
lots
of
normative
stuff.
Basically
we
made
it
say
that
you
know
you
don't
have
to
carry
everything,
but
you
probably
should
and
then
the
like
I
said
the
actual
request
component
of
it
is
aligned
with
what
the
key
value
draft
says
next
slide.
C
So
that's
that's
really
it.
We
are
requesting
working
group
adoption
for
that
one,
but
we'll
add
in
some
multi-plane
considerations.
I
know
that's
kind
of
on
a
back
burner
for
that's
really
just
any
questions
about
autofr.
A
All
right
so
and
the
option
it
would
need
to
to
probably
need
the
the
charter
update
to
adopt
it.
Yeah.
C
Okay,
Auto
evpn,
similar
to
Auto
FR,
but
you
know
build
a
European
topologies
on
that,
but
just
a
quick
update
with
this
one
next
slide
so
kind
of
similar
to
the
auto
F.
On
these
things
were
you
know
we
use
this
the
temple
product
FR
and
then
we
notice
it
problems.
So
the
similar
structural
changes
made
things
a
lot
more
clear
and
readable.
C
So
again
it
was
just
feeling
disjointed
previously
and
you
know,
jumping
around
the
documents,
never
good,
similar
stuff
to
the
auto
F4,
where
the
key
value
ties
as
well,
some
Norm
to
language
things,
and
you
know
Ayana
things
were
tightened.
There
was
a
little
bit
of
a
model
change
in
that
certain
piece.
Well,
we'll
we'll
come
to
it
and
then
you
know
editorial
and
knits,
and
things
like
that.
So
next
slide.
C
Okay,
so
again
just
kind
of
kind
of
aligns
with
what
we
did
with
autofr,
so
the
the
actual
extensions
just
kind
of
more
contained
where
they
should
be,
as
well
as
the
actual
Behavior.
This
was
never
really
spelled
out
before
and
that
they
version
fabric
ID.
The
auto
evpn
model
itself
actually
has
to
be
advertised.
In
a
no
tie
at
least
once
in
each
Direction,
so
we
can
again
it
has
to
do
with
interop,
and
then
there
were
kind
of
relevance.
Complementary
scheme
updates
for
that
yep
next
slide.
C
Okay,
so
with
the
kit
with
the
key
value
ties
the
auto
evpn,
the
the
Telemetry
model
we
use
to
push
Upstate
Upstream
had
a
couple
of
things
in
the
thrift
model
that
have
now
since
moved
with
common
model.
Overall,
so
you
know
be
used
outside
of
it.
So
it's
kind
of
just
a
it's
more
of
a
cleanup
thing,
but
it's
an
important
one.
C
Be
the
structure
of
the
that
bit
in
common
Thrift
also
changes
just
a
little
bit,
but
again
that
just
aligns
with
the
values
that
we're
going
to
request
from
Ayanna
for
well-known
oui
and
experimental
key
types
again
normative
language
stuff
right,
don't
don't
have
to
do
these
things,
but
should
and.
C
Okay,
same
thing:
here
we
are
still
kind
of
work
on
some
of
the
data
center
interconnect
stuff,
so
it'll
basically
be
a
different
Auto
VPN
model
to
you
know,
get
track
more
multi-plane
considerations
and
we've
got
a
lot
of
this
stuff
built
in
already,
but
there's
some
other
implications.
We
need
to
work
on
and
then
some
actual
operational
deployment
considerations
next
slide
and
I
think
that's
it.
C
The
only
slide,
nothing
terribly
exciting,
but
it's
kind
of
worthy
of
an
update
going
through
the
shepherd
View
for
the
Yang
model
made
me
realize
that
this
would
not
do
well,
so
I
refactored
it
to
take
into
consideration
all
of
the
Ayanna
stuff.
You
know
expert
review
requirements,
guidance
for
them.
You
know
registry
structure
as
well,
so,
like
Jeff,
said
the
shepherd
write-up's
already
in
progress
there,
but
next
slide.
C
So
this
may
be
a
little
bit
of
a
question
for
ad
depends
on
what
they
want
to
do.
I
I
sent
an
email,
initially
saying
you
know,
or
the
Ayanna
had
reached
out
to
me.
They
did
an
early
review
actually
before
this
and
kind
of
said.
Well,
you
know
you're
you're,
requesting
a
registry
from
something
that
isn't
there
yet
there's
no
top
level
Rift
category
yet
because
it's
requested
in
the
base
spec
I
was
thinking
about
already
just
you
know,
we'll
move
that
request
into
this
draft,
but
I
changed
my
mind.
I,
don't
know!
C
D
Right
sorry,
I
saw
your
email,
but
I
was
a
little
bit
confused
because,
yes,
the
register
doesn't
exist,
they
they
have
been
doing.
Ayanna
has
been
doing
recently,
I
think
they
started.
Maybe
this
ITF
reviewing
documents
that
they
see
on
agendas,
so
they
probably
have
never
reviewed
the
base
spec
right,
yeah
right,
so
they
don't
know
that
that
is
coming
so
yeah.
You
know
they
so
no
I.
You
should
just
assume
that
that
exists,
because
in
any
case
we're
going
to
publish
this
after
the
base-
spec
yeah
yeah.
C
Yeah
and
I
I
talk
to
you
think.
Oh
maybe
maybe
this
move
along
faster
video
split
in
hairs
right
so
leave
it,
as
is
I'll
correct
the
email
the
but
there
is
I,
actually
have
to
specify
the
rift
registry
that
I'm
requesting
out
of
so
very
brief
update
to
version
two
but
literally
sends
yep
to
me.
E
Tony
P
Juniper,
so
that's
more
question
to
Alvaro.
So
do
you
prefer
us
to
stick
the
hole
into
one
huge
registry,
king
of
Rift,
or
should
we
break
off
the
rift
base,
spec
registry
and
then
the
key
values
registries.
C
A
Okay,
I
think,
thank
you.
That's
all
the
updates.
We
have
we
just
and
zip
through
the
presentations.
Now
it's
coming
to
the
before
we
go
to
the
next
agenda
and
do
you
have
any
comments,
questions
from
anyone.
A
E
Tony
P
again
so
just
as
update
the
community,
so
Jordan's
pushing
the
newest
version
of
the
draft
up,
because
you
know
we're
basically
breaking
tools.
All
the
time
size,
wise
reviews
going
on
and
I
am
updating,
Bruno's
open
source
implementation
to
the
newest
spec,
which
I
already
did
but
there's
one
thing
that
Bruno
was
supposed
supposed
to
do,
but
he's
in
communicanto
for
a
moment,
so
that
will
be
done
within
the
next
week
or
two
and
test
it
against.
B
So
currently
Charter
has
a
base
spec
applicability,
setting
young
model.
All
documents
are
in
advanced
State
and
progressing
really
well
right.
So
we
have
adopted
another
number
of
items
that
kind
of
happen:
the
charter
half
not.
So
we
would
like
to
update
Charter
to
explicitly
include
all
the
work
on
key
value
registry
and
all
the
work
where
Rift
is
used
really
for
bootstrapping
stuff,
so
otvpn
flood
reflection,
so
segment
routing
is
adopted,
but
again
not
exactly
in
the
chapter.
So
I
would
like
to
include
it.
B
Multicast
work
has
been
stranded
for
quite
some
time,
but
we
plan
to
restart
working
working
group
potentially
after
next
ATF,
and
you
would
like
to
include
the
new
Charter
as
well.
So
it's
there
when
we
start
so
we
don't
have
any
off
Charter
work
for
the
couple
of
accumulative
Tony.
E
Yeah,
so
the
especially
the
multicast
I
mean
the
whole
multicast
was
stuck
on
this
last
issue
of
the
tree
joint
at
the
top
right,
and
that
was
before
the
whole
Auto
VPN
autoflot
reflection
would
actually
delivers
a
completely
distributed
stable
election
at
the
top,
for
you
know,
whatever
one
two
three
roots
that
possibly
can
be
used
to
resolve
this
multicast
dilemma
in
which
the
multicast
at
the
end
was
stuck
and
I.
Think
it's
worth
a
push.
E
If
it
doesn't,
you
know
it's
impossible
to
solve
well
I
mean
no
other
igp
managed
to
solve
it
reasonably.
If
we
manage
to
actually
get
these
multicast
working
and
merge
it
at
the
top,
this
will
be
a
pretty
big
pop
in
terms
of
protocol
design.
So
I
just
encourage.
You
know
more
than
happy
to
have
a
look
and
help
with
the
stuff
as
well.
To
give
you
know
the
one
last
look,
whether
we
can
merge
all
these
trees
that
we
already
figured
out
at
the
very
top
and
get
the
whole
solution
working.
B
D
Yeah,
it
would
be
really
nice
if
others
in
the
working
group
had
opinions,
because
you
know
it's
important
to
to
you
know,
get
reviews
and
people
to
look
at
the
drafts
and
everything
else
other
than
that
yeah.
This
should
looks
good
to
me.