►
From YouTube: IETF111-SPRING-20210726-1900
Description
SPRING meeting session at IETF111
2021/07/26 1900
https://datatracker.ietf.org/meeting/111/proceedings/
A
B
Good
morning,
good
afternoon,
good
evening,
good
night
welcome
to
spring
walk.
B
Please
not
the
not
well
by
contributing
to
the
itf.
You
agree
to
follow
itf
processes
on
policies
which
are
described
here.
So
please
with
a
look.
B
B
So
please,
please
feel
free
to
to
to
contribute,
because
this
is
a
topic
for
the
work
group
now.
B
So
today's
meeting
you
you
can
update
the
minutes
by
directly
checking
and
correcting
your
your
name
of
comments
and
contributing
to
the
ether
pad
in
term
of
queue
management.
You
can
enter
the
queue
by
pressing
the
res
and
icon,
but
you
need
to
separately.
Send
audio
to
speaker,
so
you
need
to
send
audio
chairs,
cannot
send
your
audio.
B
B
B
I
would
like
to
thank
the
design
team
for
the
hard
work
really
a
lot
of
works
and
lots
of
dedication,
and
now
it's
it's
up
to
the
working
group
to
review
the
commands
and
to
contribute
and
to
own
the
work.
B
There
is
a
related
document
in
ibp
and
workflow
for
the
team
to
typo,
which
has
been
accepted
by
ippm
before
we,
we
took
it
on
spring.
B
B
There
was
a
revised
id
needed
it
was
published
today.
So
please
check
whether
your
comments
are
a
part
of
it,
and
there
is
another
one
occurring
now
right
now,
which
is
a
draft
idea
screen
mps
pass
segment.
B
B
And
finally,
we
have
two
two
documents
in
rc
editor.
B
F
F
Me
when
to
advance
yeah,
thank
you
charles.
This
is
attention
I
will
present
to
the
slides
on
behalf
of
the
design
team,
any
comments
or
question.
Please
interrupt
me
directly
next
slide.
E
F
F
Okay,
here
we
list
the
the
two
main
tasks
running
in
the
design
team.
One
is
a
requirement
to
draft
and
the
the
other
is
analysis.
Draft.
F
F
F
So
here
is
the
latest
status
of
a
requirement
draft
compiled
to
revision.
Five.
In
this
version
we
just
remove
the
section
5.2
and
change
a
few
words
next
page.
F
According
to
the
comments,
the
members
of
the
design
team
agreed
that
this
requirement
is
a
normal
idf
rule
that
should
be
satisfied
by
default.
So
we
remove
that
next
page
and
then
we
directed
to
the
status
of
analysis
draft
so
compiled
to
the
first
revision
of
the
analysis
draft
we've
presented
in
last
atf
meeting.
F
F
F
F
F
Vc
draft
defines
the
site
of
see
the
behaviors
to
access
smaller
seats
within
the
sr
basics.
Hider,
it's
a
very
interesting
solution.
F
It
introduced
the
uet
flags
to
unify
the
traditional
i76
and
unify
the
seed
forwarding
behaviors.
When
the
uet
is
zero
zero,
it
will
represent
the
classical
ipv6
address
when
it's
zero
one.
It
will
be
a
32
bit.
F
F
F
This
figure
is
a
typical
reference
network
for
the
analysis
and
isa
domain
consists
of
three
subdomains.
F
F
F
F
The
compression
proposal
should
minimize
the
number
of
the
additional
forwarding
state
store.
At
a
note,
our
analysis
result
is
that
cc
we
said,
and
the
ui
dsr
minimize
forwarding
state
store
at
the
node
cih
moves
per
segment
state
from
the
packed
to
fib,
okay
next
page,
and
let's
continue
to
see
the
srv6
based
solution.
F
G
F
Oh,
I
I
don't
know,
maybe
there
are
some
noise,
but
I
I
don't
know
why
so.
Okay,
yeah.
F
Is
almost
finished,
so
I
think
that
yeah.
E
I
I
Just
you,
we
can
continue
probably
further
down,
and
you
know
when
we
get
to
the
q
and
a
section
but
yeah,
no,
no
comments.
F
F
Summarization
cih
doesn't
for
the
operational
requirements.
All
the
proposal
can
provide
lossless
compression,
non-rotating
information,
flexible
ipv6
planning
and
the
scalability
requirements.
All
the
proposal
meet
scalability
requirements,
all
the
proposals,
support
16-bit
and
32-bit
seeder,
and
all
the
proposals
can
be
deployed
with
the
isabel
6
best
solution
at
the
same
time
and
for
the
security
requirements.
F
All
the
proposal
can
support
the
security
requirement.
So
next
page-
and
this
page
gives
the
overall
result
channel
shows
the
relative
degree
based
on
conclusion
we
just
mentioned
since
it
meet
all
the
requirement
items
visit
and
the
uids
are
allies.
Satisfied
in
forwarding
efficiency
and
srv6
based
crh
cannot
support
based
and
seed
summarization
requirements
and
let's
satisfy
the
on
forwarding,
efficiency
and
state
efficiency.
F
F
I
think
in
order
to
achieve
this
point,
the
design
team
took
more
one
years
and
we
had
the
more
than
100
meetings
and
exchanged
over
1000
emails.
E
So
folks,
guyan
is
up
first
in
the
queue
anybody
else
who
wants
to
speak
put
yourself
in
the
queue
andrew
second
guyan
go
ahead.
I
Yeah,
I
just
had
a
brief
question:
how
long?
What
is
the
timeline?
I
guess
to
finally
pick.
I
guess
pick
one
of
the
four
solutions.
Also
I
guess.
As
far
as
the
final,
would
it
be
really
picking
one
or
would
would
there
be
a
combination
of
multiple
solutions
as
the
final,
I
guess,
the
end
result
of
the
whole
design
team
exercise.
I
guess
to
come
up
with
that.
I
E
What
the
working
group
chooses
to
do
is
up
to
the
working
group.
If
a
working
group
really
wants
to
choose
one,
they
can
if
they
want
to
choose
all
those
which
for
which
there
are
sufficient
people
willing
to
work
they
can
if
they
want
to
make
some
other
choice.
Well,
as
long
as
the
chairs
can
understand,.
I
A
You
guys
firstly
good
evening
or
good
morning,
depending
on
where
you
are
so.
I
just
wanted
to
make
a
comment
that
you
keep
this.
A
G
H
J
We're
not
interested
in
the
srv6
control
plane
or
for
that
matter,
any
other
control,
plane,
they're,
driving
the
whole
thing
from
a
controller.
G
E
K
Hi,
can
you
hear
me
yes
just
wanted
to
make
a
real,
quick
comment
that
looking
at
the
presentation
and
the
materials
and
the
draft
and
and
and
and
watching
the
comparisons
in
background
it
seem
seems
like
ccd,
is
almost
a
superset
looks
like
a
preferred
solution
going
ahead,
at
least
from
an
individual
point
here?
Are
you
going
to
issue
an
adoption
call
anytime
soon.
F
Go
ahead,
witching,
yeah,
so
I
think
the
current
assisted
draft,
which
have
combined
several
different
solutions,
such
as
microsieve
gc
and
even
the
basic
conception
of
a
visit.
F
F
We
hope
the
working
group
can
consider
some
adoption
and
how
to
do
the
adoption.
I
think
it
depends
on
the
working
group
decision.
Thank.
G
G
Did
that
answer
your
question?
I
know
you
directed
the
chairs,
but
I
don't
want
to
just
brush
over
it
so
an
act.
It
would
be
good.
K
Yeah
at
some
point,
I'd
like
to
understand
what
chairs
are
thinking
and
if
and
and
what
the
timelines
are
with
the
issuance
of
the
call
looks
like
good
analysis
has
been
done,
at
least
from
my
perspective
and
looks
like
a
a
a
comparative
analysis
is
looking
solid.
That's
one
of
the
reasons
I
asked
thank.
E
You,
the
one
thing
I
will
say
about
what
the
chairs
are
thinking
is
we
want
to
know
what
the
working
group
is
thinking.
We
would
really
like
to
see
discussion
of
these
two
drafts
on
the
working
group
by
the
working
group
on
the
email
list,
so
we
can
tell
because
what
we
want
to
do.
Well,
we
have
opinions
I'll
grant
you
that's
not
what
matters
we
want
to
know
what
the
working
group
wants.
K
I
think
I
will
reply
to
that
joel
with
all
due
respect.
The
working
group
is
going
to
be
a
a
a
set
of
opinions
coming
from
different
individual
members.
I
think
at
some
point.
B
G
Daniel
you're
next,
in
the
queue.
L
Yeah,
do
you
guys
hear
me,
I
think?
Yes,
yes,
we
can
good
joe.
Can
you
go
back
there?
The
slide
just
before
next
step,
where
we
had
the
summary
in
the
conclusion
yeah
that
one.
So
the
last
comment
at
the
bottom.
L
All
requirements
are
not
equally
important
and
the
questioning
of
the
fuss
about
it's
more
is
obviously
so
I'm
I'm
very
surprised
to
see
that
coming
in
that
slide,
because
we
had
a
requirement
document
before
doing
the
analysis
that
the
working
group
kind
of
varied
or
had
like
a
thumbs
up
or
thumbs
down
and
to
me,
like
the
analysis
document,
is
based
on
the
agreement
of
the
working
group.
L
So
if
they're
not
equally
important
anymore
after
we've
done
the
conclusion
it's
kind
of
weird,
is
that
how
I
should
interpret
that
that
we
had
an
agreement
on
the
requirement
before
doing
this
analysis?
I'm
I'm
talking
to
the
the
the
design
team.
If,
if
that
was
discussed
amongst
your
amongst
the
team
before.
C
F
Yeah,
so
I
think
it
in
the
design
team
there
is
no
importance
assigned,
I
think
maybe
four
some
requirements
must
and
for
us
should
and
we
provide
all
the
possible
requirements
from
the
design
team
side,
and
here
we
list
this
sentence,
because
there
are
some
members
in
the
design
team
who
have
some
concerns.
F
G
Waiting
this
is
jim,
I
have
to
say
it
caused
me
confusion
as
well,
because
my
reading
of
this
slide
is
either
either
the
solution
supports
the
requirement
or
it
does
not-
and
this
is
saying
the
you
know-
here's
what
here
are
the
requirements
here
are
the
solutions.
These
ones
support
that
requirement
these
ones
don't
now.
Obviously,
when
you
choose
a
solution,
it
may
be
that
some
of
the
requirements
you're
not
particularly
concerned
about
and
therefore
it
it
doesn't
matter
to
you.
G
But
the
fundamental
conclusion
here
surely
is
that
the
solution
either
meets
the
requirement
or
it
doesn't,
and
it's
not
really
a
case
of
whether
that
requirement
is
significant
to
a
particular
operator
or
not.
Is
that
correct.
F
Yeah
yeah,
so,
okay,
I
understanding
your
comment.
So
what
do
you
do?
You
have
some
suggestion
to
design
team.
L
Know
I
was
just
trying
to
understand
it,
but
go
ahead
daniel,
sorry!
L
L
With
with
this
we're,
not
we're
we're
not
applying
a
fair
trade
treatment
for
all
four
solutions.
If
we
only
waiting
at
the
end
of
the
analysis
to
find
out
that,
oh,
but
this
one
is
a
should,
and
this
one
is
at
last,
this
is
what
I'm
trying
to
say.
L
N
N
G
I
I
think,
it's
also
fair,
to
say
joel
that
as
part
of
that
discussion,
if
people
want
to
make
specific
statements,
you
know
such
as
you
just
made
wind,
then
that's
perfectly
fair
to
do
so
on
the
mailing
list
and
that's.
G
A
decision
as
to
you
know
what
we
do
as
a
working
group
moving
forward,
but
to
be
fair
at
the
moment,
there's
virtually
no
discussion
of
either
requirements
or
the
analysis
document
on
the
working
on
the
mailing
list.
So
it's
been,
you
know.
Obviously
the
chairs
have
got
their
own
opinions
and
they
may
not
actually
agree
with
each
other
from
you
know.
Particular
solutions
perspective,
but
it's
not
you
know
we're
trying
not
to
influence
this.
We
want
to
hear
from
the
working
group
as
to
you
know.
G
What
do
we
want
to
do
with
these
documents?
Do
you
know
how
have
these
documents
satisfied,
what
the
design
team
was
asked
to
do,
and
how
does
the
working
group
want
to
move
forward?
So
mailing
list
discussion
is
really
being
encouraged
by
the
chairs.
At
this
point,.
O
N
P
All
I
can
really
say
is
just
to
pile
on
to
what
was
previously
said
it
would.
It
would
be
great
if
the
if
the
working
group
can
can
come
out
and
and
pick
one
of
these.
I
think
that,
as
a
design
team
member,
we
spent
a
lot
of
time
and
a
lot
of
effort
a
lot
of
hours
and
a
lot
of
deep
analysis.
P
That
was
done
very
well
and
very
technically
true,
and
I
think
it'd
be
be
great
to
get
the
comments
on
that
and
validation
of
of
that
analysis
and
requirements
that
was
produced
by
the
working
group.
Q
A
very
tough
walk
at
the
past
year,
so
please
go
ahead,
take
a
look
of
the
text
and
they
will
help
to
this
for
the
discussion
and
since
we
have
multiple
vendors
have
made
them
make
their
choice,
of
which
solution
would
be
the
better
one.
So
we
we
may
accelerate
the
the
peer
the
the
progress
of
the
standard
progress.
I
think,
because
our
customers
want
the
solutions
yep.
Thank
you.
O
O
O
O
O
O
O
Let's
compare
two
two
drafts,
so
one
draft
is
disrupt.
Another
draft
is
an
informational
draft
so
that
information,
another
graph
recorded
draft
a
and
a
disk
draft.
We
call
the
dropper
t
so
these
two
drafts,
I
have
some
difference-
draft
a
talks
about
using
any
cast
c
to
protect
a
failed
node
and
also
for
protect
e-p-e-c.
O
O
The
overlapping
part
between
two
drafts,
which
is
the
both
drafts,
try
to
protect,
know
the
seat
on
the
failed
note,
however,
even
for
this
at
this
common
part,
they
use
different
approaches
in
draft
a
draft
day.
Talks
about
configuring
route
hold
timer
on
each
on
each
node
to
hold
the
routes
after
another
failure
for
some
given
time.
O
R
If
I
understand
correctly
so
you
suggest
that,
because
not
all
nodes
may
support
their
timer,
then
the
convergence
in
the
domain
will
be
problematic,
but
it
appears
that
the
same
is
applicable
to
this
proposal,
so
it
has
to
be
heterogeneous
in
a
domain.
So
all
nodes
has
to
support
this.
R
So
why
you
think
that
all
nodes
will
support
this
proposal
and
rather
than
supporting
the
timer?
Thank
you.
O
Yeah,
that's
a
good
question,
so
the
thing
is
that
in,
for
example,
for
configuring,
every
node,
so
for
some
of
the
nodes
just
support
that
one
in
in
a
in
this
case.
The
production
may
not
look.
For
example,
one
example
is
that
if
the
shortest
path
is
now
the
on
the
nose
which
supported
this
one,
then
the
traffic
were
dropped
in
the
nodes
which
now
support
this
feature.
However,
for
drafter
t,
because
we
for
each
approximate
capability,
it
will
identify
the
capabilities.
O
So
then
we
can
select
a
rows
which
supports
on
the
node
which
supports
that
capabilities.
So
the
properties
solution
may
be
a
little
bit
better.
That's
the
my
point
because
they
have
more
information
in
the
network.
They
consider
that
cover
giving
nodes,
and
then
they
send
the
traffic
along
the
nose
with
the
capable
to
that.
On
view
to
the
adjacent
note
of
the
feather
note,.
S
Can
you
hear
me
yes
yeah,
so
I
think
my
first
comment
was
same.
What
greg
made
that
if,
if
a
receiving
router
doesn't
support
draft
t,
then
it's
the
same,
I
mean
it
won't
be
able
to
support
the
procedure.
So
I
think
the
what
you're
trying
to
say
is
the
advantage
of
t
is
that
if
certain
plr
doesn't
support.
S
Local
failure
mechanisms
for
protecting
a
node,
then
drafty
has
a
better
approach
where
it
can
advertise
that
information,
and
then
traffic
can
come
to
that
node
which
supports
being
proxy
and
then
provide
protection.
So
one
thing
that
I
would
like
to
note
is
so
far:
whatever
protection
mechanisms
have
been
built
around
segment
routing,
be
it
ti
lfa
or
micro,
loop
avoidance,
and
all
these
are
following.
S
I
mean
they
all
follow
the
post
convergence
path
and
never
divert
from
post
convergence
path,
whereas
this
solution,
if
if
if
there
is
a
certain
if,
if
all
nodes
are
upgraded,
probably
drafty
also
follows
that.
But
if
certain
plrs
are
not
upgraded,
then
it
it
attracts
the
traffic
towards
the
nodes
which
are
supporting
this
proxy
mechanism,
which
means
that
you
are
diverting
from
the
post
convergence
path.
S
So
I
think
what
we
need
to
discuss
is
draft
t
seems
to
assume
that
you
know
protection
is,
in
certain
cases
more
important
than
being
comple.
I
mean
being
compliant
to
the
post
convergence
path,
and
how
much
important
that
use
case
is
is
something
that
that
we
need
to
discuss.
O
Yeah,
I
think,
as
I
answered
same
as
a
break
as
india,
so
you
for
some
of
them
not
support
and
then
maybe
we
can
even
provide
some
sort
of
protections,
but
for
the
for
the
better
one,
yes
configuration
and
that
may
not
work.
That's
my
point.
T
Can
you
hear
me
now
yeah,
we
can?
Oh
sorry.
A
T
Was
needed?
Okay,
I
just
want
to
make
a
small
note
on
what
chad
I
mentioned.
You
know,
regardless
of
you
know.
These
drafts
in
particular,
is
that
sometimes
the
trfa
pass
may
diverge
from
the
post
convergence
pass,
because
if
you
are
doing
node
protection
or
srlg
protection,
then
the
the
tlfa
paths
may
not
follow
the
post
conversion
path
and
it
may
differ
so
it
wouldn't
be
the
first
case,
but
it's
not
that
I
want
to
make
a
particular
comment
on
the
draft,
but
that
situation
can
happen
even
with
the
regulatory
lfa.
E
U
I
think
the
next
presentation
is
srv6
midpoint
protection
unit
agenda.
If
I.
U
Because,
based
on
the
agenda,
the
the
presentation
is
srv6
midpoint
protection
and
the
slides
presenting
now
is
the
next.
U
U
So
why
we
need
srv6
midpoint
protection
when
there
is
some
failure
in
the
transit
node,
we
could
use
fr
mechanisms
such
as
tifa,
but
when
there
is
a
field
node
in
the
endpoint
of
the
srv6
path,
the
existing
fr
mechanisms
cannot
be
used
to
restore
the
reachability.
U
It
could
provide
endpoint
protection
for
srm
mps
case,
and
the
goal
of
this
document
is
to
provide
an
srv6
midpoint
protection
in
the
srv6
network.
When
an
srv6,
endpoint
fails
and
srv6
proxy
forwarding
node
can
replace
the
failed
endpoint
to
perform
my
srv6
endpoint
function
and
forward
packet
to
the
right
destination,
and
next
page.
Please.
U
Actually,
the
mechanisms
is
quite
straightforward:
yeah
srt,
srv60,
passive
endpoint
is
failed.
U
The
midpoint
protection
could
be
divided
into
two
periods
in
the
first
period
before
the
igp
has
convert
converged
when
the
repair
node
is
adjacent
to
the
field
node
and
finds
that
the
field
node
is
an
endpoint,
it
will
perform
proxy
forwarding
and
send
packet
to
the
next
or
some
reasonable
endpoint
in
the
second
list
and
the
second
period
after
igp
convergence
has
happened
any
upstream
node
that
has
been
converged
and
deleted.
The
fib
to
the
field.
U
U
In
the
document
we
specified
the
srv6
midpoint
protection
behavior,
basically,
the
when
the
repair
node
is
a
transient
node
or
an
endpoint
node.
The
behavior
is
the
same.
It
is
divided
into
two
cases
as
as
I
have
explained
in
the
previous
slide.
U
The
first
case
is
when
the
node
is
adjacent
to
the
fifth
node,
and
the
second
case
is
the
node
is
a
remote
node
and
the
behavior
is
similar
same
segment
left
this
creases
updated
the
ipv6
destination
and
the
fibo
cup
on
the
updated
destination
address
and
forwarding
the
packet
according
to
the
matched
entry,
and
when
the
repair
node
is
an
endpoint.x
node.
U
Please
so
actually,
this
is
the
some
simple
plan
for
this
document.
This
document
has
already
been
presented
in
rtgwg
for
several
times
and
we
collect
some
feedbacks
and
updates
document.
U
U
R
Thank
you
for
the
presentation.
I
have
a
question
so
the
note
that
provides
the
protection
is
expected
to
have
a
state
and
notion
about
their
edges
and
their
capabilities
so
that
it
can
select
if
detected
the
node.
R
Actually,
my
question
is
like
twofold:
first
is
how
this
node,
that
switches
to
another
egress
knows
that
the
alternative
is,
is
still
capable
and
can
be
used,
and
it's
a
more
another
question
is
more
like
generic
and
philosophical,
probably
why
you
propose
to
use
local
protection
for
this
rather
than
to
have
end-to-end
protection,
so
not
to
create
the
state
in
a
transit
node
and
let
their
ingress
note
to
do
the
switchover
if
something
fails.
U
Thank
you
greg
for
your
question
for
the
first
question.
If
I
understand
this
right,
it
is
about
when
we
find
an
endpoint
has
failed,
we
switch
to
another
endpoint
to
forward
the
packet.
How
do
the
node,
how
does
the
node
and
know
that
the
other
endpoint
is
still
work?
If
I
understand
this
right?
Is
that
your
question.
R
Yes,
because
if
you're
switching
to
and
to
node,
which
is
not
working
or
cannot
be
reached,
then
so
there
is
no
benefit.
U
Actually-
and
this
is
another
case
when
the
the
existing
node
find
its
neighbor
has
failed
and
the
neighbor
is
the
endpoint,
it
will
switch
the
destination
address
to
the
next
endpoint.
If
the
neighbor
of
the
next
endpoint
finds
that
the
the
end
point
is
also
failed,
it
will
switch
to
the
next
endpoint.
U
So
that
is
that
is
another
behavior
after
this,
because
the
first
thing
you
will
find
that
my
adjacency
has
failed
and
it's
an
end
point,
and
if
the
some
other
nodes
find
that
the
switched
endpoint
is
also
failed,
it
could
do
the
similar
thing
to
to
forward
the
packet
unless
this
is
the
last
endpoint.
U
Yeah
the
the
next
next
question
is
why
we
won't
use
end-to-end
protection.
A
yes
end-to-end
protection
will
work,
but
we
think
and
end-to-end
protection
maybe
have
their
limitations,
such
as
it
depends
on
bfd
or
some
other
detection
mechanisms,
and
also
local
repair
will
be
more
fast
and
in
some
cases
maybe
vendors
will
like
to
see
the
the
packet
won't
last
for
a
long
time.
So
we
think
local
repair
is
also
requested.
K
R
That
there
needs
to
be
some
mechanism
to
detect
that
the
neighbor.
R
It
depends
yes,
but
again
there
could
be
scenarios,
but
I,
I
think
of
scenario
where
bfd
must
be
used,
so
you
cannot
really
rely
only
on
link
layer,
state.
R
I
don't
think
that's
the
case
because,
regardless
whether
it's
a
single
hop
or
multi-hop
bfd
can
be
used
aggressively
and
detect.
R
Def
failure
at
the
same
time,
so
I
don't
see
much
difference
between
single
hop
or
multi-hop
and
that's
been
used
in
the
networking
for
a
long
long
time.
U
Actually,
it's
just
like
tfa.
If
we
use
tfa
for
the
transient
node,
we
can
to
treat
it
as
a
local
repair
mechanism.
So
we
could
use
this
endpoint
protection
for
the
endpoint
as
local.
If
we
do
not.
G
Bruno
you're
next
and
then
then
robin
and
then
we'll
close
the
queue
there,
because
we
in
the
interest
of
time.
B
Brother
speaking
as
an
individual
contributor,
could
you
move
to
slide
four?
Please.
B
Thank
you.
So
the
slide
is
not
very
readable,
but
my
understanding
is
that
the
transit
node
is
a
regular
ipv6,
loader
doing
a
ipv6
routing
and
you
are
proposing
that
the
transit
ipv6
node
is
changing
the
ipv6
header,
both
destination
address
and
srh,
and
this
could
be
seen
as
a
violation
of
src
8200.
U
Actually,
in
my
understanding
when
the
tlfa
works-
and
it
will
just
like
to
have
a
new
sr6
encapsulation
for
the
for
the
packet
to
build
up
a
new
new
tunnel
to
the
next
destination,
I
think
it
is
similar
we
have
to
make
the
the
repair.
Node
has
some
extra
capability
to
complete
the
protection.
U
B
U
Yes,
you
are
right,
the
the
transit
note
may
be
requested
to
to
change
the
srsx
header,
but
we
think
it
is
in
the
limited
domain
and
the
the
transient
node
is
also
trusted,
so
we
we
think
this
may
be
reasonable
to
give
it
some
capability
to
do
this,
and
the
mechanisms
can
be
complicated
if
the
case,
the
scenario
is
reasonable.
B
Maybe
this
would
need
to
be
discussed
in
six
months
because
it's
it's
a
change
of.
U
8200
sure
we
can
also
bring
it
to
the
list,
and
we
can
also
forward
it
to
six
men
to
to
guarantee
that
this
will
work.
B
T
M
Okay,
I
got
other
explanation
to
the
greg's
comments.
I
think
the
bmd
is
used
for
the
link
local
for
the
local
protection
and
the
this
is
just
the
one
link
is
just
one
bfd
session,
but
if
the
pfd
is
used
for
the
under
to
under
failure
detection,
that
will
be
first
pass.
That
means
for
every
sr
path.
There
will
be
a
pfd,
so
that
will
increase
the
deployment
burden.
So
I
think
that's
the
the
challenge
of
the
under
to
under
detection.
The
second
one
I
think,
the
the
blue
for
the
local
detection.
M
I
think
that
the
detection
period
can
be
shorter,
but
if
the
bfd
is
used
for
the
end
to
end
that
the
timer
needed
to
need
to
be
set
to
a
long
value.
So
that's
also
is
this
case
in
the
practice.
G
Hey
thanks
for
I
actually
closed
the
queue
greg
because
we
need
to
move
on
to
the
next
presentation.
Perhaps
we
can
take
that
comment
or
that
discussion
to
the
mailing.
G
E
V
V
An
sr
policy
allows
for
allows
for
multiple
candidate
parts,
and
but
only
one
active
candidate
passed
can
be
provisioned
in
the
forwarding
plan
and
used
for
traffic
steering
at
the
same
time,
so
another
candidate
pass
may
be
used
as
a
backup
in
order
to
specify
the
protection
relationship
between
canada,
parts
confederation
or
local
policy
may
be
needed
on
the
head,
and
the
backup
candidate
house
provides
protection.
Only
when
all
the
segment
lists
in
this
active
can
be
passed.
Are
invalid
scenarios
like
load
balancing,
may
need
a
segment
list
for
protection.
V
V
V
V
And
this
draft
also
proposes
extensions
of
bgp
os.
The
t
distribution
draft
describes
a
mechanism
to
collect
the
sr
policy
information
that
is
locally
available
in
a
node
and
advertise
it
into
bgpos
updates.
It
includes
a
kennedy
parts
pass
theov
and
the
segment
list,
tov,
s-flag
and
b-flag
are
originally
defined.
In
kennedy,
past
state
tv
and
two
new
flags
of
similar
meanings
in
segment
list
theory
are
defined
in
this
draft.
The
s
flag
indicates
that
the
segment
segment
list
is
in
administrative
short
state
if
set
the
b
flag
indicates
in
the
segment
list.
V
N
I
I
will
be
segment
routing
an
extension
to
second
routing
to
provide
redundancy
protection
next
slide.
I
So
this
draft
provides
a
an
extension
to
segment
routing
to
provide
redundancy
protection.
It
utilizes
the
deadnet
framework
architecture
of
to
provide
ultra
high
lossless
transmission
of
data.
You
know
so
it
provides.
You
know
a
variety
of
use
cases
where,
where
high
high
redundance
redundancy
protection
is
required,
where
there's
ultra
low
loss
of
traffic
and
you
in
utilizing
the
dead
net
framework
that
architecture
it,
we
introduced
two
new
sid
type
services.
I
So
there's
a
a
new
replication
sid
which
which
performs
the
the
prp
rf
function,
the
replication
function
in
the
definite
framework,
and
then
you
merge
it,
which
is
at
the
egress
node
that
provides
the
pef
elimination
function
for
the
overall
pre
pre-off
framework,
providing
that
function.
I
So
with
with
that,
the
goal
of
the
draft
is
to
be
able
to
provide
the
ultra
high
ultra
high
transmission
of
traffic
and
flows
by
duplicating
the
packets
and
then
eliminating
the
duplicates
at
the
merge
point
to
provide
a
lossless
data
transmission.
I
So
in
this
diagram
here
we're
we're
showing
just
an
example
with
srv6
redundancy
protection
and
how
that
would
work
so
with
srm
pls.
It
is
very
similar
that
you
would
have
a
redundancy
node
where
you
have
the,
where
you
have
a
a
replication
sid,
where
we,
where
the
packets
are
replicated
as
a
redundancy
node,
where
two
or
more
copies
are
created.
I
Each
replica
is
transmitted
via
via
different,
destroying
forwarding
paths
that
may
occur
along
the
path
at
the
top
and
then
the
path
at
the
bottom.
And
then
at
the
merge
point,
we
have
a
a
service,
a
second
service
in
emerge
sid,
which
actually
does
the
elimination
provides
a
pef,
pef
function,
dead,
net
framework,
pdf
function
to
eliminate
copies
of
the
of
duplicates
and
then
provides
a
lossless
transmission
from
ingress
to
egress.
I
See
some
history
so
at
ietf
104
105.
This
draft
was
presented
in
dead
network
group
as
the
deadnet
srv6
data
plane
encapsulation
at
ietf
106.
This
was
presented
again
at
the
deadnet
work
group
suggests
and
was
suggested
to
take.
I
Take
the
take
the
presentation
take
the
draft
to
spring
and
then,
after
that,
in
ietf
109
was
this
draft
is
presented
again
but
presented
now
in
spring,
and
the
proposal
was
redefined
and
as
a
general
protection
mechanism
and
basically
like
a
live,
live
protection
so
really
the
the
architecture
it
was
based
in
initially
on
the
net
framework.
However,
the
the
single
use
cases
was
was
what
the
framework
was
based
on,
but
now
this
is
it.
I
The
the
draft
is
really
more
of
a
ubiquitous
where
in
any
use
case,
where
ultra
high
low
ultra
high
traffic
flows,
where
ultra
low
loss
flows
are
required,
such
as
5g
or
network
slicing,
where
low
loss
traffic
is
required.
High
qos
and
availability,
video
voice,
video
or
whatnot,
where,
where
there
are
cases
where
that
that
might
be
required,
this
solution,
could
it
could
provide
provide
the
framework
for
that
and
it
could
be
a
a
good.
A
solid
solution
for
it
for
that
next
slide.
I
So
the
draft
structure
shows
scenario
of
redundancy
protection
in
sr,
so
we,
as
I
mentioned,
we
have
two
two
new
segments
that
were
introduced
for
redundancy
protection.
So
we
have
the
redundancy
product.
We
have
the
redundancy
segment,
that's
at
the
ingress
node
and
then
a
merge
segment
of
the
egress
node.
So
the
redundancy
segment
within
the
net
framework
is
providing
the
the
packet
dub
replication
prf
function
and
at
the
merge
segment
providing
the
packet
elimination
function
for
the
for
pre-off
and
metadata
to
support.
I
Redundancy
protection
is
supported
as
well
segment
routing
policy
to
support
redundancy
protection
as
well,
so
that
is
that
is
supported
as
well
to
provide
provide
that
with
srte
next
slide.
I
I
In
the
case
of
srv6,
we
we
define
a
new
n
n
r
n
function
is
defined
for
srv6,
for
the
merge
point
merge
segment
to
perform
the
packet
elimination
function
on
the
merging
point
is
this:
is
the
goal
of
the
merge
segment
in
the
case
of
srv6,
we
have
a
new
behavior
defined
at
n.
I
I
So
on
the
left,
the
programming
messer
is
the
programming
for
the
redundancy
segment
for
srv6.
So
when
so,
when
srv6,
when
the,
when
the
sra
job
is
processed
for
with
srv6
programming
first,
so
the
standard
processing
happens
what's
included.
I
In
addition,
is
the
allow
in
blue
at
line
so
six
and
so
seven
to
allow
flow
identification
and
the
sequence
number,
if
indicated,
and
then
do
duplicate
duplication
of
the
packets
as
the
number
of
active
sids
listed
and
b.
I
So
so,
at
the
beginning
of
the
floor
at
the
at
the
during
the
processing
once
once,
the
flow
flow
identification
happens
and
the
packets
are
duplicated,
then
the
standard
processing
happens
similar
to
any
other
topological
segment.
On
the
merge
side,
when
the
merge
segment
the
same
thing
here,
the
standard
srv6
processing
occurs
with
the
of
the
initial
when
the
srh
is
processed.
What's
added
in
there
is
is
in
blue,
acquiring
the
secret
sequence,
number
the
receive
packet
and
look
up
in
the
table
and
it's
a
sequence
number
doesn't
exist.
I
I
Basically,
what
we're
doing
here
is
it's
similar
to
the
dead
net
architecture,
where
the
where
the
sub
layers,
the
service
sub
layer
function,
is
really
what's
happening,
with
the
service
segment
with
the
redundancy
and
merge
segment
where
we
have
sequence
numbers
so
we're
we,
we
have
a
sequence
number
that
we're
matching
on
on
the
traffic
flows
and
that
sequence
number
is
actually
matched
actually
at
the
merge
segment
to
determine
which
which
flows
in
which
packets
are
duplicates
that
need
to
be
removed
at
the
bird
segment.
So
that's,
that's
really
the
key
there!
I
It's
really
in
this.
In
this
framework,
it's
providing
the
that
service
sublayer
function
in
in
tracking
each
flow,
with
a
sequence
number
and
determining
which
packets
to
be
removed
at
the
merge
segment.
Next
slide.
I
So
a
packet
packet
walk
with
srv6
redundancy,
so
how
how
the
flow
would
happen
so
going
from
left
to
right.
So
we
have
at
r1,
we
have
our
ingress,
node
and
then
node
r,
so
and
at
the
source
node.
The
srv6
srh
header
is
added
at
the
source
node
and
then
it's
forwarded.
The
headers
cop,
the
srh
processing,
happens
in
the
packages
forwarded
along
the
steered
path.
The
packet
hits
the
redundancy
node
so
at
the
redundancy
node
the
service,
the
replication
sid
is
as
processed
so
at
that.
I
At
that
point,
the
sequence
number
is
added
into
the
into
the
srh
header
and
the
and
the
tlv
is
set
for
the
for
the
f
label
and
that's
labeled
for
for
the
dead
net
architecture,
our
ceo
1925
and
then
and
then
the
package
process
received
at
the
merge
nodes
that
the
merge
node,
the
sequence
number
is
matched
and
there
when,
when,
when
the
service
said
the
emerge,
node,
the
merge
segment
is
processed,
the
sequence
numbers
matched
and
then
duplicates
within
the
within
that
sequence.
I
Number
are
eliminated
so
performing
basically
the
pdf
function
within
pre-off
and
then
the
packet
is
forwarded
to
the
egress
next
slide.
I
So
addressing
received
comments
thus
far
on
the
mailing
list,
so
so
the
first
one.
What
is
what
is
the
location
to
add
the
flow
id
and
sequence
number,
so
we've
had
an
agreement
that
there
is
an
advantage
to
add
the
fi
and
sn
at
the
ingress
node,
so
that
has
been
updated
in
revision.
Four.
G
Go
this
is
jim,
I'm
sorry
to
interrupt.
You've
got
maybe
one
more
minute.
I
I
Well,
dude
all
right.
The
next
one
is
word:
steel,
replication,
sit
and
redundancy
protection,
so
clarifications
on
the
modification
of
the
replication.
Sid
number:
three
clarification:
the
redundancy
segment
with
multiple
pads,
so
clarification
there
and
then
flexibility
with
the
redundancy
segment
to
be
topological
or
or
services.
I
So
relationship
to
deadnet,
so
there
there
has
been
some
conversation
back
and
forth
with
on
the
deadnet
architecture
and
whether
are
the
implementation
of
the
redundancy
sit
redundancy
protection
and
whether
it
violates
or
fall
or
if
it
follows
a
deadnet
framework.
So
we've
had
some
feedback
back
and
forth
related
to
that
so
reduction.
Redundancy
protection
is
it's
a
generalized
mechanism,
so
we're
using
the
framework,
but
it
can
be
used
there.
There's
any
any
use
case
where
you
need
ultra
low
low
loss.
I
I
Oneplus
n
protection
ultra
high
reliability
for
for
segment,
rounded
networks,
ultra
light
ultra
reliable
protection
in
many
use
cases,
so
any
use
case
where
you
have
voice
video
or
where
you
need
ultra
high
reliability.
Ultra
high
reliable
data
transmission
across
the
segment
routed
network
where
low
loss.
I
So
drafts
revision-
since
you
know,
since
by
atf-110
so
so
use
case,
clarifications,
said:
targets
to
provide
ultra
reliable
transmission,
add
more
use
cases.
We
did
some
reorganization
with
the
srv6
scenario,
an
inflexible
place
to
add
the
flow,
identification
and
sequence
numbers
at
the
ingress.
So
we've
decided
you
know
it
updates
in
section
three
and
four
for
the
ingress,
node
and
redundancy
node
node
and
then
change
changes
to
the
segments,
the
pseudo
code.
I
Accordingly,
in
section
four
and
regarding
segment
usage
in
sr
over
mpls,
you
know
referencing
rfc
8964
in
section
four
and
lastly,
a
separate
redundancy
policy
specification
so
spinning
off
into
a
different,
possibly
different
draft
section.
Six
next
slide.
I
So
we'll
we'll
take
some
more
discussions
onto
the
mailing
list
and
we
want
to
get
feedback
from
the
work
group
through
the
changes
like
an
update
since
since
the
last
itf
110,
and
now
whether
you
know
how
to
how
the
worker
feels
about
the
draft
and
and
the
use
cases,
we've
expanded
the
use
cases.
So
it's
not
geared
just
for
detonate,
even
though
the
framework
is
based
on
dead
net
based
on
the
pre-op,
but
we're
leveraging
dead
net
the
architecture
that
could
be
used
for
any
type.
I
Any
really
any
use
case
where
you
have
where
you
have
high
high,
where
you,
where
you
require
high
through
high
transmission
and
low
and
really
low
or
zero
loss
of
traffic,
zero
packet
drop.
So.
G
Thank
you
thanks
again,
so
so
greg.
If
it's
quick,
I'm
gonna
allow
one
question,
because,
in
fairness
of
other
presenters
I
want
to
make
sure
they've
got
time
to
give
their
presentations.
So
if
you
can
go
ahead
greg,
but
I'm
going
to
have
to
cut
the
cue
after
you.
R
Analyze
why
this
method
is
more
beneficial
than
just
one,
plus
one
protection
when
you
select
the
working
source
and
the
protection
source
and
do
the
switch
over
not
per
packet
but
by
source.
Thank
you.
G
Okay,
sorry,
everyone
else
on
the
queue
I'm
going
to
have
to
cut
it
there.
If
you
can
take
your
questions
to
the
mailing
list,
we
need
we
need
to
get
on.
We've
got
three
more
presentations.
I
want
to
make
sure
they
get
time.
So
so,
joel
next
presentation.
G
S
W
E
W
W
W
In
fc,
1874
54
defies
srv6
hmac
theory
for
ipv6
source
address
and
segment
routing
header
integrity
protection.
This
method
may
bring
a
program
because
hmac
verification
is
based
on
symmetric
king.
That
means
all
network
nodes
need
to
be
verified
have
to
share
the
same
king.
So
when
a
single
points
king
was
lacking,
then
all
the
trusted
doming
was
compromised.
W
Next
slide,
the
first
part
we
define
a
new
type
for
tle,
which
called
us
theory.
The
two
is
used
for
signature
protection
based
on
a
cemetery
sacred
kings
and
used
to
protect
source
edges
and
header.
Ascii
id
is
a
for
object
or
pack
number
which
identifies
the
hash
signature,
algorithm
and
certificate
arial
number
used
for
signature,
authentication
and
us
will
be
used
to
store
the
signature
content
us
is
to
protect
segment
routinely
6
header,
but
also
protect
gsru
b6
header
next
slide.
W
W
W
In
part,
two,
how
to
verify
signature
always
seek
nature.
Verification
is
required
at
king
nut,
blog
notes.
It
also
divided
into
three
steps.
First,
enable
enable
signature.
Verification
at
king
knows
order
notes,
will
do
the
verification
step.
2
request
the
public
king
certificate
from
the
controller.
W
Only
the
kingdoms
who
want
to
verify
the
signature
well
recurs
the
public
incentivic
certificate
step,
3
calculate
the
harsh
value
according
to
the
header
and
use
the
public
key
to
decrypt.
The
signature
finally
make
these
two
variables
compared
if
we
equal
them
forward
and
otherwise
this
card
next
slide.
Thank
you.
W
We
have
a
t4
optim,
optimization,
how
to
see
in
a
word,
use
a
local
hash
table
to
record
the
history
that
so,
when
some
hash
value
is
filed
in
the
table,
then
no
need
to
do
our
more
decryption.
W
G
Thank
you
ron
go.
G
J
J
W
Thank
you,
rome,
it
is.
Is
it
sure
that
a
segment
routine
hazard
a
can
contain
many
different
variables
in
our
we
just
covered
several
king
king
variable
kim
parameter
in
the
segment
routine
header
such
like
source
address.
W
And
the
second
and
the
second
question
is
I
unders
I
from
my
point
answer:
you
are
care
about
their
king
point,
because
most
of
the
points
are
muted
is
the
right.
W
W
Sorry,
I
I
don't
get
the
keen
point
of
your
second
question.
Can
you
discuss
this
this
question
on
my
on
the
mail
list.
H
H
H
Yeah
so,
as
I
said,
the
the
internet
marketing
data
field
that
we
are
going
to
extend
here
for
the
essential,
the
srh
tlb.
It's
basically
the
same
that
we
have
defined
for
rpvc
for
the
pv6
options
adder.
So
we
have
the
2-bit,
that
is,
the
loss
and
delay
bit
also
called
marking
fields
and
from
monitoring
identification.
H
H
So
the
use
of
this
dlv
is
quite
simple,
so
the
ingress
node
is
the
only
one
that
can
write
these
tlb
and
can
may
add
the
alter
mark
dlv
in
the
srh.
If
this
functionality
is
supported,
of
course,
regarding
the
intermediate
and
egress
node,
they
can
only
read
this
dlv
and
if
an
intermediate
or
rigorous
node
are
not
capable
of
processing
this
dlv,
they
can
simply
ignore
it.
H
It
is
worth
highlighting
that
if
there
are
nodes
that
are
not
capable
of
processing
this
dlv,
this
is
not.
This
is
not
an
issue,
because
the
measurement
can
be
done
only
for
the
nodes.
H
According
to
a
discussion
we
had
in
six
months,
we
had
the
new
section
on
controller
domain,
because
8799
introduced
the
concept
of
specific
limited
domain
solutions
and
alternate
marking
is
considered
as
an
let's
say,
an
innovative
approach,
an
innovative
extension
to
a
pv6
that,
for
some
reason,
such
as
policies
option
supported
security
requirement.
It
is
strongly
suggested
to
limit
this
kind
of
application
to
limit
to
limited
domains,
and
so
we
added
this
requirement
also
in
the
context
of
srv6.
H
H
J
Thank
you
for
this
work.
The
ietf
has
defined
a
small
handful
of
routing
extension
headers.
If
you
want
the
alt
mark
to
be
available
only
to
the
srh,
you
might
consider
putting
it
in
an
srh
tlv.
J
H
Yeah
yeah.
This
is
a
good
question
because,
as
I
said
yeah
in
six
months,
we
we
already
defined
the
the
same,
the
same
option
same
tlv
that
can
be
encoded
as
a
destination
option.
So
in
this
case
it
is
applicable
to
all
the
routing
gather,
but
with
this
draft
we
only
apply
as
a
htlv
because,
of
course,
for
the
case
of
srv6,
it
may
be
preferred
the
use
of
sir
htlv,
because
you
know
in
some
cases
the
destination
option
are
not
well
support
and
maybe
in
the
future,
if
srv6
will
be
used.
J
H
H
G
Okay,
no
more
questions
so
so
last
presentation
chang.
I
hope
I
pronounced
that
correctly.
X
Okay,
I'm
changi
from
china
mobile.
I
will
present
the
draft
on
behalf
of
the
co-authors
next
slide.
Please.
X
Symmetrically,
if
we
want
the
spfd
packet
go
because
symmetrically,
we
need
the
especially
in
srt
srv60
scenario.
The
srh
header
will
be
that
srh
header
will
be
double
less
long.
X
X
X
X
If
the
destination
address
is
itself
and
if
the
sr
header
flag
bit
file
is
sent,
if
the
condition
is
is
satisfied
that
the
reflector
node
will
do
the
swept
operation
besides
the
normal
spfd
procedure.
Next
next
slide,
please.
X
This
is
more
detail
about
the
procedure
in
the
picture.
There
are
four
node
and
switch
a
is
the
init
node
and
it
will
push
the
sid
list
abcd
in
as
a
header
and
set
the
as
I
had
a
flag
bit
file
to
one
one.
It
runs
as
spfd
packet
chair
travel
to
the
switch
d,
that
is
the
reflector
switch
d,
will
do
normal
spft
operation
and
it
it
will
will
swap
the
seed
list
to
this
ecba.
G
See
chang:
this
is
jim,
we're
gonna
run
out
of
time,
so
if
you
can
quickly
wrap
up
and
greg,
if,
if
you
don't
get
cut
off,
then
go
ahead
with
your
question.
Y
X
This
is
just
a
conclusion
in
the
draft
we
just
reused
the,
as
I
had
a
flag
bit
file
to
indicate
if
we,
if
the
reflector
node
need
to
do
the
swap
operation
and
the
reflection
just
needed
to
support
the
swap
operation,
this
is
clean
and
tidy
next
slide.
Please.
X
X
R
I
noticed
that
the
draft
is
tagged
as
informational,
but
it
looks
like
you
are,
requiring
allocation
of
one
of
the
fields
in
the
flags
to
do
the
reversal
of
the
seed
list,
and
another
is
probably
more
generic
is
that
I
I
think
that
similar
can
be
achieved
if
you
concerned
with
their
length
of
their
seedless
with
a
binding
sid,
and
then
it
would
be
applicable
to
both
the
service,
6
and
srm
pls.
R
Also,
the
concern
is
that
sbfd
is
not
defined
for
multi-point
networks,
so
that
I
wonder
if
you
plan
address,
only
point-to-point,
sr
policies
or
point
to
multi-pointers
are
policies
as
well.
Thank
you.
X
Okay,
thank
you.
This
is
an
interesting
question
for
the
first
first
draft.
We
just
consider
the
point
two
point
scenario:
maybe
we
we
will
reverse
draft
and
in
the
next
next
chapter,
we'll
consider
more
scenario
like
the
p2
mp
scenario,
and
we
can
talk
in
the
mail
list.