►
From YouTube: IETF103-RTGAREA-20181108-1610
Description
RTGAREA meeting session at IETF103
2018/11/08 1610
https://datatracker.ietf.org/meeting/103/proceedings/
A
A
A
First
of
all,
because
I've
seen
many
working
groups
show
the
all-russian
of
the
note.
Well,
this
is
the
actual
right
version.
Now
this
version
talks
about
intellectual
property.
Yes,
just
like
the
old
one
also
has
pointers
to
things
like
the
code
of
conduct.
How
many
of
you
were
the
ITF
plenary
last
night?
A
One
is
to
many
honestly
who
came
to
the
microphone
to
talk
about
not
the
best
conduct
in
the
ITF
people
being
attacked
directly
people
not
being
given
the
chance
to
express
their
opinion
and
sort
to
explain
their
ideas,
and
so
it's
important
that
we
understand
that.
That
is
obviously
not
the
way
we
should
behave
in
the
ATF
I
have
no
specifics
or
no
knowledge
of
anything
of
this
happening
in
the
routing
area.
A
But
you
know
it's
important
for
us
to
to
be
on
the
lookout
for
that,
especially
for
the
chairs,
as
we
conduct
our
meetings
to
make
sure
that
we
give
everyone
chance
to
to
speak
everyone
a
chance
to
present
their
opinion,
defend
their
proposals.
You
know
whatever
they're
doing
and
that's
the
other
reason.
Why
again,
this
is
the
note.
Well
it's
important
because
it
talks
about
intellectual
property
and
some
of
the
links
at
the
bottom
are
about,
for
example,
the
code
of
conduct
and
the
anti
anti
harassment
procedures
in
the
IETF.
A
There's
also
some
links
that
we
have
here
for
other
materials
in
the
routing
area.
As
you
can
go,
take
a
look
at
absolute
meeting.
There's
blue
sheets
are
going
around
somewhere.
Is
anyone
on
the
jabber
room
by
any
chance?
Can
someone
please
get
on
the
jabber
room
just
in
case
we
have
someone.
Thank
you
drew
well
there.
We
have
several
people
remote.
If
the
remote
people
are
listening,
which
I
hope
they
are.
A
Yes,
you
can
use
if
you
have
a
room.
If
you
want
to
say
something,
you
can
also
come
to
the
mic
and
we'll
see
you
from
here
and
we
can
let
you
speak
as
well.
So
here's
the
agenda
we're
going
to
talk
about
the
area
of
course.
First
then
we're
going
to
have
the
routing
directorate
report
and
then
we're
going
to
have
an
update
one
year
later,
of
where
we
are
with
DC
routing
work
and
then
hopefully,
at
the
end,
we'll
have
some
time
for
for
any
other
things
that
you
guys
want
to
talk
about.
A
So
as
always,
we
want
to
request
that
you
review
documents
we
s3
and
the
rest
of
the
is.
You
have
to
read
everything,
and
sometimes
it
looks
like
the
documents
were
not
reviewed
properly
in
the
working
groups.
You
have
the
routing
area
Directorate,
which
does
a
lot
of
reviews,
and
it
helps
a
lot,
but
please
make
sure
that
the
other
are
people
reviewing
the
documents
through
adoption
and
that
last
call
and
discussion.
If
you
need
to
grab
volunteers
or
volunteer
someone,
please
do
that
to
make
sure
that
the
quality
of
the
documents
goes
up.
C
A
You
thanks
really
good
comment.
Separates
are
not
only
supposed
to
review
documents
as
well
as
the
chairs
they're
supposed
to
also
Shepherd.
That's
why
it's
called
Shepherd
the
document
through
the
rest
of
the
process.
So
if
you're
a
shepherd,
remember
that
your
job
doesn't
end
when
you
fill
out
the
Shepherd
right
up,
that's
just
the
beginning
right!
You
still
have
to
go
through
the
whole
process.
In
case
you
know,
there's
some
stick
up
with
AG
or
you
know
some
review
later
to
you
know
help
the
authors
go
through
that.
A
A
If
you
don't,
then
it
means
whatever
the
dictionary
says
that
that
word
is
right.
It
also
says
something
else
that
is
important
in
that.
Not
every
specification
to
be
normative
has
to
use
29,
18
21
19
keywords,
but
you
can
still
make
a
specification
normative
and
clear,
without
necessarily
using
those
words.
So
just
the
action
here
is:
please
make
sure
that
the
documents
don't
have
the
alternate
119
boilerplate,
that
we
use
the
one
at
the
bottom
here.
That
lists
the
key
words.
A
We're
always
looking
for
feedback.
Please
talk
to
us
about
ourselves,
talk
to
us
a
lot,
each
other.
You
know
anyway,
you
want
to
provide
us
with
feedback
of
things
that
we're
doing
well,
or
especially
things
of
me
to
improve.
Please
let
us
know
this
Center
of
the
year
is
very
important
because,
as
you
know,
the
NomCom
or
so
I
hope
you.
A
A
It
said
no
one
else
and
I'll
come
here.
I
don't
see
you,
but
in
case
you
want
to
give
feedback.
You
can
talk
to
them.
You
can
go
to
the
webpage
data
tracker,
slash
mount
come
slash,
28:18
and
provide
feedback
there.
The
feedback
can
be
anonymous.
If
you
want
to
I'm
sure
that
you
can
tell
them
and
give
them
an
honest
feedback
and
that
they
can
anonymize
it
for
you
as
well.
A
Usually
they
use
that
not
only
to
make
the
selection
for
routing
ad
or
for
any
ad
in
this
case,
but
they
have
been
doing
for
the
last
couple
of
years.
They
have
given
us
feedback
and
what
the
community
says
so
that
we
can
improve.
So
if
you
don't
feel
like
talking
to
us
directly,
especially
bad,
go
talk
to
the
NomCom
and
they
can't
take
care
of
taking
down
into
consideration
for
selection
and
to
give
us
feedback
later.
Also,
this
is
not
just
about
about
any
of
you
once
we
get
feedback
about
some
other.
A
A
A
The
other
thing
they
will
say
here
is
that
we
are
always
on
the
lookout
to
grow
the
leadership
talent
inside
the
area.
It
was
important
because
we,
the
three
of
us,
are
probably
not
going
to
do
the
ad
work
forever.
The
chairs
change
once
in
a
while.
So
we're
always
looking
to
to
find
people
who
can
do
those
jobs
are
willing
to
do
it.
A
That
means
couple
things
it
begins
the
head.
There
was
a
good
pool
of
people
that
were
nominated.
You
know,
ten
nominees
were
outing.
It
he
was
was
a
good
poll,
but
a
tips
on
the
client,
which
means
that
only
to
accept
it
and
there's
two
slots
so
yeah
you
can
do
the
math.
You
can
add
now
that
doesn't
mean
that
it's
an
automatic
selection,
the
NomCom,
is
going
through
the
process.
A
Anyways,
that's
why
I
invite
you
to
go,
give
them
feedback,
because
if
we
are
not
the
right
people
or
especially,
if
we
need
to
improve
right,
you
know,
please
let
them
know
and
the
two
people
that
accept
it.
Of
course,
never
and
I,
so
we're
going
to
a
third
term
I,
don't
know
many
a
these
who
go
past
a
third
term.
So
we
need
to
keep
that
in
mind
right
for
the
next
time,
and
hopefully
we
have
a
lot
more
people
who
accept
nominations
the
next
time
around.
So.
A
A
A
You
have
always
loved
this
meeting
because
we
used
to
give
updates
of
the
working
groups.
We
used
to
lighten
up
all
the
chairs
that
everyone
will
give
a
two
minute
update
of
what
they're
working
was
doing
this
year
at
the
start
of
this
year.
We
changed
that
a
little
bit
where
we
have
been
using
some
time
to.
You
know
socialize
a
little
bit
more,
which
was
at
the
first
ITF
of
the
year
when
we
had
the
routing
ad
change,
when
idea
was
replaced
by
Martin
and
we've
been
putting
the
status
updates
from
the
wiki.
A
How
many
of
you
know
where
the
wiki
is?
Okay?
How
many
of
you
who
are
not
a
chair
know
where
the
way
he
is
okay,
so
we
figured
that
you
know
the
the
wiki
has
been
useful,
but
probably
not
as
useful
as
we
would
want.
So
in
one
of
the
last
meetings
that
we
have
the
chairs,
we
talked
about
changing
the
the
audience
that
we're
making
the
the
updates
for
so
what
we
hope
you
start
seeing
now
is
a
status
update
on
the
page
of
the
data
tracker
working
group.
A
Yes,
so
you
should
be
able
to
go.
Look
at
that
and
get
an
update,
so
you
know
or
more
importantly,
so-
that
external
people,
maybe
people
who
come
to
the
ATF
too
much-
maybe
people
who
are
not
participating
in
specific
working
group
but
do
come
to
the
ATF,
get
a
sense
of
what
the
workgroup
is
doing.
What
are
they
interesting
things
that
are
going
on,
so
we
should
start
seeing
that
that
now
this
time
we're
running
this
is
an
experiment.
A
A
A
This
came
from
a
question
that
was
brought
up
at
the
plenary
in
ITF
102,
where
it
was
asked
about
implementation
requirements
in
the
ATF
as
a
whole.
Thanks
or
the
IAG
had
discussed
that,
and
we
decided
it
was
really
hard
to
have
one
policy
for
everyone.
So
we
decided
that
this
would
be
better
served
working
by
working
group.
There
are
working
groups
that
already
have
policies.
A
So
we
hope
that
this
soon
will
come
up
in
a
list
near
you
for
that
discussion.
This
doesn't
mean
that
every
working
group
should
require
two
implementations
for
interoperable
implementations,
or
anything
like
that.
This
means
that
we
hope
there
is
a
policy
for
the
work.
So
the
work
group
may
decide
that
the
policy
is
five.
Interoperable
implementations
may
decide
to
may
decide
one
and
may
decide
that
what
we
want
is
a
RFC
79,
42
type
section
and
the
RCS
or
in
the
drafts.
A
It
may
decide
that
there
is
no
requirement,
there's
many
reasons,
but
we
want
this
for
the
chairs
and
hopefully
in
connection
with
their
working,
you
think
about
this
and
considerate,
so
that
you
know
we
can
actually
articulate
what
that
policy
is
I,
hope,
the
working
participants
or
by
you
anyone
else
who
is
or
by
you
know,
a
sir.
It's
ever
so
Stewart.
You.
E
Surely
the
name
Stewart
III
didn't
realize
he'd
put
it
there,
but
I
was
really
worried
that
one
size
doesn't
fit
all
in
any
working
group.
So
whatever
your
policy
is,
you
need
the
chair
exception,
or
the
ADA
exception,
or
something
or
other,
because
you
know
you
know
I'm.
There
are
always
cases
where
it's
an
information
about
something
or
it's
a
correction
that
everyone
knows
needs
to
be
had
or
whatever
there
always
needs
to
be
an
exception.
E
A
A
Yeah,
so
there's
been
several
suggestions
of
where
we
can
put
this.
It
could
be
in
the
status
update.
It
could
be
somewhere
else
on
the
on
the
air
tracker
itself.
Some
worker
groups
have
a
wiki
page
when
we
discuss
this
with
the
chairs
someone's
just
so
that
maybe
we
could
even
make
it
part
of
the
Charter,
so
one
size
doesn't
fit
all,
and
the
ten
here
is
for
everyone
to
have
a
policy.
A
You
know
we
are
the
IETF.
We
will
need
an
order
code,
so
you'll
be
good
to
at
least
substan
what
the
policy
is.
There
are
many
reasons
why
we
may
or
may
not
require
from
the
protocol
that
we're
that
we're
developing
the
market
that
we're
serving
the
type
of
industry.
You
know
many
many
things
why
you
would
want
or
not
to
have
requirement
implementations
in
your
working
group.
D
Just
to
let
folks
know,
I'm
John
has
told
us
the
very
bad
news
and
he's
stepping
down
as
pce
chair
and
also
for
for
this,
so
actually
we're
on
the
lookout
to
help
somebody
to
help
Amy
okay.
So
if
anybody
would
like
to
to
step
up
to
this
to
help
out,
they
can
tell
you
what
it
involves
and
we'd
really
appreciate
it,
and
we
thank
John
very,
very
much
for
everything
he's
done
for
us.
Thanks.
G
D
G
Last
time,
I
counted
49,
routing
area
experts
who
are
awesome,
people
who
freely
give
their
time
to
help
the
ADEs
in
processing.
All
the
many
drafts
that
we
have
to
go
through
in
the
routing
area,
they're
appointed
by
the
ADEs
they're.
The
purpose
of
having
a
Directorate
is
basically
to
ensure
quality
of
the
drafts.
So
we
review
most
of
the
routing
area
drafts
as
they
pass
through.
Itf
last
call.
G
We
will
review
from
time
to
time
other
routes
and
related
drafts
at
IETF
last
call
as
well.
If
you're
a
working
group
chair,
you
always
have
the
opportunity
to
ask
the
Directorate
to
perform
an
early
review
of
any
draft
that
you
have
in
your
working
group.
It's
useful,
particularly
if
you
feel
that
a
draft
needs
more
review
and
you're,
not
getting
that
review
inside
your
working
groups
to
get
some
eyes
on
that
from
the
Directorate
and
from
time
to
time,
but
the
rector
will
assist
the
ADEs
in
making
judgment
calls
about
particular
issues.
G
G
The
reviews
that
the
80s
sent
to
the
Directorate
and
the
gray
shows
the
reviews
of
the
chairs
sent
to
the
Directorate
and,
and
you
can
see,
but
at
the
start
of
year
we
had
a
lot
of
reviews
because
for
chair
of
the
the
80s
of
a
keeping
is
very
busy
with
with
40
reviews
of
the
ITF
101
period,
that's
tape
it
off
from
the
ADEs
in
the
next
two
periods,
but
the
the
chairs
have
increased
for
load
that
they've
been
putting
on
us.
So
it's
nice
to
see
where
we'd
be
getting
somewhat
early.
G
Reviews
requests
coming
in
from
the
chairs
I'd
like
to
see
more
about,
because
I
think
it's
actually
useful.
But
when
Dawkins
comes
ITF
last
call
of
the
Directorate
have
already
seen
them.
Then
we
already
have
someone
ramps
up
on
the
draft
and
the
draft
is
presumably
already
had
the
benefit
of
them
looking
at
it.
D
G
Of
reviews
come
from
outside
there
at
the
razzing
area
altogether,
and
certainly
no
one
working
group
loads
as
much
as
the
rest
of
the
ITF
so
few
more
stats.
What
do
we
find
in
our
reviews?
The
graph
on
the
Left
shows
what
do
we
find
when
we
actually
review
documents
from
inside
the
area?
Well,
half
of
those
documents
we
find
issues
and.
G
Occasionally,
we
will
send
the
document
back
through
the
working
group
and
say
look.
This
is
not
ready
where
there
are
major
issues
and
we
don't
think
it's
ready
to
be
published.
Obviously,
that's
the
chairs
in
the
ad
to
really
decide
what
happens
to
the
document,
but
it
does
sometimes
happen.
It
must
the
three-percent
slicing
the
pie
charter.
G
Sometimes
we
just
think
that
reviews
are
perfect
and
the
rest
of
the
time
we'll
find
some
needs,
but
nothing
to
block
the
draft
part
of
the
right
is
the
same
thing
but
for
drafts
coming
from
outside
area.
So
again,
but
pretty
similar
I
think
that
generally,
we
find
slightly
fewer
issues
in
those
sorts
of
drafts,
probably
because
of
the
background
required
to
review
them
in
depth.
But
but
we
still
do
a
pretty
good
job
of
haunting
out
issues,
and
one
thing
I
think
is
interesting-
is
of
these
types
of
graphs
are
quite
stable
across
time.
E
G
G
A
Thank
you
so
now
we're
going
to
go
into
an
update
on
that
they
are
Center
routing
work.
So
if
you
remember
a
year
ago,
we'd
met
in
Singapore
and
in
Singapore
we
had
a
a
bar
about
DC
routing.
The
intent
of
that
bar
was
to
talk
about
the
requirements
and
potential
interest
of
the
community
in
working
on
new
road
and
proposals
to
be
used
in
a
data
center
that
both
we
asked.
You
know
how
many
people
are
interested
in
a
couple
of
the
represented
there,
and
the
other
thing
to
be
interest
in
that.
A
So
now,
when
you're
later
would
want
to
do
is
provide
an
update
or
where
we
are,
and
also
at
the
same
time,
to
try
to
renew
the
interest
in
these
proposals.
And
you
know
people
participate
in
the
work
groups
as
well,
so
we're
gonna
have
an
update
from
Ella's
VR
and
from
rift
on
where
they
are
and
what
the
thing
was
G
is
about.
So
Gunther
I
think
this
one
percent
this
one.
H
Hello,
everybody,
so
what
I'm
doing
here
besides
freezing
is
you
know
coming
to
the
realization
that
I've
been
blessed
to
entertain
you
all
with
my
you
know,
least,
eight
victor
routing.
You
know
technology,
so
indeed
we
started
this
journey
about,
like
you
know
about
a
year
and
a
half
ago,
when
we
start
to
explore
this
space
and
about
a
year
ago
we
started
with,
like
you
know,
think
about
charging
this.
H
This
particular
rocking
group,
so
I'm
gonna
be
covering,
like
you
know
more,
like
you
know
what
it
is
about
and
some
of
the
topics
which
you
know
keep
us
like
an
old.
You
know
kind
of
warm
and
where
we
are
right
now
with
the
working
group.
So
initially
you
know
when
the
working
group
actually
got
started.
H
I
thought
I
gave
is
a
very
clear
message
actually
said,
like
okay
guys,
you
have
like
one
year
to
actually,
you
know,
deliver
like
a
set
of
documents
and
after
the
year
I
will
actually
assess
how
well
you
know,
everything
actually
has
progressed
and
if
to
also
understand,
if
you
like
any
real
major,
you
know
gate
stops
or
something
like
that
and
then
we're
gonna
see.
You
know
what
will
be
happening
and
afterwards
with
a
particular
work.
So
I
think
at
this
point
in
time.
H
H
We
said,
okay,
you
know
in
between
IDs
we
have
to
do
some
interim
meetings,
so
that
is
what
you
know
we
are
doing
also,
and-
and
so
you
know
from
what
I'm
seeing
it
really
helps
well
in
keeping
continuous
evolution
actually
in
your
deliverables.
So
if
you're
not
doing
it
actually,
I
would
highly
recommend
you
know
to
really
start,
not
thinking
about
it.
It
really
brings.
You
know,
progress
in
the
work
you're
delivering
so
initially
when
we
started,
you
know
our
journey
in
the
data
center.
You
know
we
were
just
trying
to
understand.
H
You
know
why
is
data
center
routing
different
from
any
other
kind
of
environments
now
about
a
year
and
a
half
year
ago
we
actually
did
some.
You
know
some
further
analysis
about
it
now,
I
think
about.
Like
a
week
week
and
a
half
two
weeks
ago
there
was
an
anode
meeting,
and
actually
it
was
a
very
good.
You
know
overview
by
by
Tony
Lee
and
by
Chris
Martin
on
white
data
center.
You
know,
routing
environment
actually
has
some
slightly
different
properties
than
routing
in
the
traditional.
H
H
Just
you
know,
as
a
background
set
of
information
so
going
to
my
agenda,
I
have
about
like
120
slides,
have
about
30
minutes,
so
I'm
trying
to
keep
up
with
everybody
here
so
also,
you
know
just
be
aware,
you
know,
make
sure
you
know
I
am
the
chair,
but
I'm,
not
the
smartest
person
in
the
room.
It's
like
it
would
be
half
if
there
would
be
any
person
here
with
a
very
difficult
question.
Then
we
have
like
smart
people
that
mean
room
actually
wrote
those
particular
specifications,
so
they
were.
H
Actually,
you
know,
give
you
the
answer,
and
the
one
thought
actually
I
will
give
you
is
that
you
know
the
one
thing
that
is
more
scary
than
a
question
is
an
answer
so
think
about
that
one.
So
there
is
something
I've
learned.
You
know
my
journey
here
at
IETF,
some
very
smart
people
here,
so
my
agenda
is
first
to
actually
go
over,
like
you
know
where
we
are
with
you
know,
least,
I
depicted
routing.
So
what
our
milestones
and
an
embedded
we
are
I
know
what
are
the
deliverables.
H
Then
I'm
gonna
give
like
a
very
high-level
overview
about.
Like
you
know
what
is
the
technology
about
and
then
at
the
end,
something
you
know,
I
only
have
like
one
slide,
because
I
didn't
really
know.
You
know
what
was
going
to
happen
exactly
my
working
group
today,
because
you
know
every
day
is
a
surprise
and
everybody.
It's
always
like
a
miraculous
kind
of
thing.
What
happens?
Unite
your
working
groups
so
now
I
know
I
just
lost
it
so
and
that's
still
alive.
So,
let's
go
on
so
this
is
where
we
are.
H
We
started
in
Singapore
a
year
ago.
We
are
almost
at
the
end
of
our
journey
at
the
next
ideas.
We're
gonna,
not
really
you
know,
try
to
assess,
know
how
stable
is
our
specification.
Where
are
we
and
you
know,
and
also
try
to
understand,
like
you
know,
where
is
the
you
know,
growing
interest
in
you
know
from
the
people
you
know
applying
these
kind
of
technologies
so
going
forward.
We
have
like
one
in
two
incoming
and
then
at
the
end
you
know
ITF
104
in
Prague.
H
You
know
we'll
see
you
know
what's
going
to
be
happening,
yes,
so
so
from
a
milestone
perspective,
we
only
have
a
few
because
we
only
had
a
few
months
work
to
actually
you
know
given
to
us
to
work
on
these
things.
So
the
main
thing
is,
we
have
an
applicability
statement
like
you
know,
we're
you
know,
does
it
make
the
most
sense
to
use
links
that
vector
routing?
How
can
you
apply
it
because
there
are
like
different
kind
of
you
know,
application
models?
H
You
know
how
you
can't
really,
you
know,
use
this
thing
and
then
we
have
like
two
other
deliverables.
You
know
it's
like
speaking
about,
like
you
know
how
to
use
links
as
Victor
out
of
extender
to
Dykstra
and
how
to
use
link
States
Victor,
you
know
distribution
using
BGP
transport,
so
the
last
two
we
put
into
one
single
document,
just
you
know,
from
a
simplicity
perspective
and
from
a
readability
perspective,
and
also
because
you
know
we
already
had
like
some
work.
H
You
know
you
know
previous
alternate
space,
so
it
seemed
as
a
logical
thing
to
do.
You
know
to
progress
the
work
forward
because
we
only
had
like
limited
amount
of
time.
We
also
were
given
the
task
to
actually
look
into
yang
and
so
we're
waiting
a
few
dependencies
there
on
some
other
working
groups
to
to
progress
also,
once
that
is
done
or
more
stable,
we're
gonna
keep
on
progressing
at
work
also.
H
Now
the
good
thing
is
that
there
is
some
sort
of
an
ambition
to
keep
this
thing
living
because
March
2019
was
kind
of
like
our
deadline,
but
this
deadline
for
this
yang
thing
is
like
in
July,
so
it
sneaky
done
at
the
door
a
little
bit,
so
we
still
have
hopes
so
looking
into
the
LS
of
your
components.
So
the
way
we
see
link
state,
vector
routing,
it's
sort
of
you
know,
exists
sort
of
like
four
different.
You
know
areas
here
so
the
first
one.
H
If
you
look
at
the
ground,
you
know
at
the
bottom,
one
is
like
each
router,
yeah
and
I
need
to
describe
itself
me
to
describe
these
neighbors
and
B
to
describe
the
cost
of
artists
neighbors-
and
it's
actually
I
saw
you
know,
assess
also
some
other
attributes
to
it.
But
the
minimum
set
is
like
Who
am
I.
Who
is
my
neighbor
and
I'll?
Do
you
know
what
is
the
cost
of
arts
my
neighbor?
So
you
need
to
harvest
that
in
some
sort
of
a
way.
H
So
that
is
something
you
know
which
can
be
done
via
and
our
whole
set
of
different
protocols.
We
have
been
you
know,
discussing
the
solution
space
for
that
the
different
protocols.
What
actually
are
out
there,
the
two
most
you
know
obvious
ones
which
actually
provide
us.
The
right
hooks
to
that
information
would
be
like
you
know,
link
state
over
Ethernet
or
lldp,
and
right
now
you
know
in
the
in
our
working
group.
H
You
know
we
actually
are
going
to
start
a
working
group
adoption
con
for
the
link
set
over
Ethernet
elements
so
from
the
moment
actually
of
harvest
that
set
of
information.
You
need
to
encode
it
into
something
so
and
the
way
we
encode
it.
You
know
we
actually
encoded
into
what
we
call
like
link
state
vectors.
So
this
actually
is
just
you
know,
an
encoding.
You
know
mechanism
to
describe.
You
know
the
router
itself
with
the
neighbors
and
wanted
to
post
well,
if
the
neighbors
as
a
minimum
set.
H
It's
proven
everybody
knows
how
it
works
and
from
my
experience
we
can
grow
it
as
big.
As
you
know,
whatever
it's
like,
you
know
a
big
massive
thing.
We
can
do
it
whatever
we
desire
right
now,
our
desire
is
just
put
like
only
a
few
things
actually
into
it.
That
is
just
the
encoding.
Now,
from
the
moment,
you
describe
your
router,
you
link
state
vector,
you
need
to
distribute
those
things.
I
know
those
elements
of
the
router
between
all
the
and
routers
in
the
environment,
so
that
is
actually
our
floating
mechanism.
H
So
the
way
we
are
doing
that
and
so
each
link
state
you
know
victim
can
be
reflected
as
as
an
NL
or
I,
and
so
it
can
actually
be
distributed
by
BGP
and
that's
what
we're
using.
So
in
our
case,
we
are
just
using
BGP
as
a
distribution
mechanism.
Now
again
from
a
link,
steady
vector,
routing
perspective.
There
is
only
one
of
the
choices.
Why
do
we
use
BGP?
H
I
will
explain
later
on,
but
right
now
you
know
it
is
seen,
as
you
know,
something
which
is
out
there
and
it
seems
to
make
sense
at
this
point
in
time
from
a
time
to
market.
You
know
perspective,
and
it
seems
the
correct
thing
to
do
at
this
point
in
time
now
at
the
end.
So
the
only
goal
is
that
each
router
will
actually
have
the
links
that
the
vectors
from
all
of
the
other
relatives
in
the
network
environment.
H
So
that
actually
means
each
router
will
have
like
a
full
view
of
topology,
just
the
same
as
if
the
router
would
be
like
an
ICS
route
that
having
all
the
other
space
you
know
the
topology
looks
like,
and
for
that
moment
onwards
you
can
actually
start
doing
dextra
and
if
you
do
the
dextra
the
end
result
will
be
your
routing
table.
So
that
is
where
we
are
so
that
is,
you
know
what
is
happening
and
for
what
you
can
see
to
top
things
here.
You
know
in
the
red
square.
H
I
should
not
really
use
red
because
red
means
bad
in
this
case.
It's
a
good
thing
is
actually
what
we
call
like
links
at
victor
out
using
bgp
SPF.
So
it's
a
very
high
level.
You
know
overview
it's
reasonably
simple.
So
why
are
we
using
bgp?
Because
bgp
is
probably
one
of
the
most
cool
protocols
ever
invented.
I
H
Is
because
it
actually
pays
for
a
lot
of
people,
you
know
their,
you
know
their
salaries
and
things
like
that.
It's
like
a
never-ending
stream
of
like
wonderful
things.
You
can
do
with
it.
So
that's
why
we're
using
you
know
BGP
in
our
technology,
because
it
is
just
ultimately
cold
now,
at
the
same
time,
you
know
it
is
you
know
it's
robust
and
simple.
Now
that
is
debatable.
Of
course
you
know
some
people
come
up
with.
H
You
know
things
to
do
with
it,
which
is
not
really
that
robust
and
need
of
simple
at
all,
but
what
we
are
doing
with
it
is.
Actually
you
know
we
are
using
it
in
a
very
simplistic
kind
of
fashion,
and
if
you
actually
read
the
specification,
you
know
you
actually
will
understand
also
that
what
we
are
doing.
H
It's
not
really
rocket
science,
it's
actually
just
using
BGP
in
a
more
simplistic
camera
version
and
if
you're
looking
to
you
know
even
spinning
SPF
into
it,
you
know
you
will
see
that
a
lot
of
the
forwarding
and
a
lot
of
the
the
ways
of
you
know
distributing
all
these
pieces
of
information
has
been
simplified.
Also
so
I
will
be
covering
some
of
those
elements.
So
BGP
provides
is
a
few
things
it
provides.
Reliable
transport
is
guaranteed
an
order
delivery.
It's
incremental
updates.
H
You
know
there
is
like
so,
unlike
in
link
state
routing,
we
don't
have
a
flooding
problem
here,
okay,
so
there
is
no
25
different
proposals
for
floating.
There
is
one
and
it
works
really
really
well
and
it's
proven
to
work.
Well,
it's
loop
free
and
it's
fantastic.
It's
awesome
so
and
at
the
same
time
we
also
have
like
different
mechanisms
to
actually
you
know
bring
this
into
the
network,
so
you
can
actually
use
like
router
flex.
You
can
use
like
controllers
to
actually
program,
no
more
information
downstream,
and
then
you
know
now.
H
The
most
important
thing
probably
from
this
slide
is
that
the
way
the
the
routing
actually
works?
Everything
is
tisha
view.
You
know
it's
computed
in
a
distributed
fashion,
so
there
is
no
one
piece
in
the
network
which
computes
the
topology
for
everything
everything
is
calculated.
You
know
from
a
routing
perspective,
distributed
and
each
single
node
itself,
so
that
is
an
important
artifact
and
at
the
same
time
you
know
a
lot
of
people.
You
know
running
data
centers
already
using
you
know,
you
know
BGP,
so
you
have
some
massive
scalable
data.
H
Centers
running
like
I,
don't
know
like
ten
thousand
one
hundred
thousand
boxes,
they
run
BGP.
You
have
some
other
set
P
of
C.
You
know
some
other
set
of
people
are,
unlike
massively
scalable
data.
Centers
existing
I
look
like
50
routers.
They
also
run
BGP
because
it's
like
really
really
cool.
Now
we
actually
offer
them.
You
know
like
a
new
set
of
technology
to
actually
know
run
with
this.
H
Okay,
that's
a
surprise,
so
the
high
level,
you
know
perspective
from
the
insight
vector
out
there.
So
the
target
use
case
for
reset
vector
routing
is
actually
you
know
the
the
massive
scalable
data
center.
So
the
plan
is
to
actually
you
know,
bring
your
political
awareness
and
to
actually
do
dextra
in
networks.
You
know
of
gigantic
size
to
actually
not
speak
about
networks
of
like
$50
or
a
thousand
routers,
but
like
10,000,
100,000
and
more
because
that
is
the
size
of
what
BGP
can
scale
to
so
that
is
really
really
nice.
H
Now,
the
advantage
you
actually
will
have
by
having
the
topological
in
our
environment
is,
you
can
do
potentially
smarter
things
with
your
environment,
because
if
you
know
the
topology
you
can
do,
you
can
actually
start
doing
things
like
loop,
free
alternates.
You
know
you
can
actually
avoid
risk.
You
know
reach
between
groups
and
so
on
words.
You
can
do
some
traffic
engineering
know
whatever
crazy
thing,
you
can
dream
off.
You
know
if
you
know
the
topology,
you
can
actually
do
this.
You
know
with
this
by
extending
the
technology
itself.
H
The
target
also
is
to
actually
use
this
as
an
underlay
routing
mechanism,
and
so
you
actually
will
reduce
the
underlayer
outing,
and
on
top
of
that
you
will
have
flow
layer,
2,
VPN,
BGP
and
whatever
you
know,
you
know
all
the
VPN
things
happening.
Layer,
3
VPN
happening
on
top
of
that
this
is
just
underlay.
H
From
an
encoding
perspective,
so
we
could
have
used
or
reused
the
BGP.
You
know
Safi
from
bt
pls
now
from
a
backward
compatibility
perspective.
We
said:
ok,
let's
not
do
that,
even
though
it's
actually
perfectly
possible,
we
actually
select
a
real
selector.
We
get
or
we
get
allocated
like
a
new
stuffy
for
this
just
from
our
backwards
compatibility
perspective.
It
also
means
we
need
to
actually
have
like
a
new
capability
for
this.
H
Now
here
you
see
my
so
the
reason
actually
I
was
laughing
with
this,
so
I
have
like
his
new
iPad
and
it's
actually,
you
know
iPad
pro
and
a
pen.
You
can
actually
make
like
very
nice
graphs
with
that.
You
know
it
make
pictures,
but
it's
not
very
useful
being
a
very
pro
in
this.
If
you
make
a
PDF
out
of
your
slider,
it
looks
like
this,
so
my
graph
actually
is
gone.
So
that's
a
bit
sad,
so
actually
what
you
would
be
seeing
here
just
visualize
this
close
your
eyes.
H
So
close
your
eyes,
you
know
dream
you're
in
this
room.
You
know
fantasize
apart
my
picture.
Okay
and
what
you
see
here
is
like
another
smalls.
You
know
the
different
routers
on
the
left-hand
side,
router
one
two
three
four
and
five,
and
they
look
like
routing
logos
and
you
have
like
a
nice
blue
lines
between
them
and
next
to
the
routing
locals
visualize.
You
know
close
your
eyes.
Take
your
time.
H
Okay,
you're,
like
in
Bangkok,
okay,
visualizer,
but
not
out,
I
thought
about
anything
else
and
then
mix
it
around
visualize
yourself,
a
puzzle,
piece
yeah
and
that
puzzle
piece
is
actually
your
link
state
vector,
and
that
is
what
I'm
drawing
here
so
I
see
everybody's
smiling
you're.
Probably
thinking
about
my
puzzles
I
think
so,
and
the
important
thing
is
that
a
puzzle
actually
is
encoded.
H
As
you
know,
in
a
way
of,
like
you
know
like
with
the
BGP
LS
and
low-riding
coding
itself,
you
know
with
the
new
stuff,
which
is
you
know,
plan
to
be
proposed,
so
that
is
my
magic.
Invisible
slide
here,
which
I
didn't
really
check,
I.
Think
of
how
to
block
my
slide
is
clicking
on
the
button
thing.
So
I
blame
him
anyway.
So
what
sort
of
peering
models
do
we
have?
So
traditionally,
you
know
if
you
look
into
how
massive
scalable
you
know,
datacenters
are
configured
by
using
BGP.
H
If
you
actually
have
like
all
routers
are
not
directly
connected
peer-to-peer
and
your
prefixes
ago,
you
know
through
them,
so
there
is
something
which
is
possible.
Also,
here
with
the
link
States,
you
know
vector
routing,
you
know
mechanism,
but
it
doesn't
have
to
be
because
if
you
know
the
topology,
you
know,
of
course,
lots
of
massive
scale.
That
is
know
the
topology.
You
can
actually
download
it.
H
It
is
actually
very
well
possible
to
actually
not
put
like
you
know
like
a
route
reflector
in
place
or
a
controller
in
place
and
P
with
that
one
directly,
so
that
you
actually
reduce
the
amount
of
feeding
configuration
in
your
network
environment,
so
those
details
and
those
specifications
and
some
of
the
requirements
and
what
can
be
done
is
actually
you
know
this.
You
know
documented
very
well
in
the
applicability
draft,
so
if
anybody
would
have
a
you
know,
a
nice
question
about
that,
be
you
know
feel
free
to
come
to
the
mic.
H
You
know
again,
we
have
people
here
to
actually
answer
your
difficult
question,
so
that
is
where
we
are
so.
The
other
thing
to
keep
in
to
mind
is
that
what
we
are
doing
here
is
not
a
new
technological
principle.
So
what
is
being
done
here
is
sort
of
like
I,
wouldn't
call
it
reinventing,
but
more
standardizing,
an
approach
which
has
been
used
by
you
know,
which
has
been
disclosed
already
by
Google
in
one
of
the
sitcom
papers
like
you
know,
which
is
know
like
Jupiter
rising
has
they
could
have
cost
a
pop.
H
H
This
picture
actually
turned
out
better,
which
is
fantastic
because
I
didn't
make
it
with
my
ipod,
Pro
and
my
pen.
So
that's
like
a
thousand
dollars
thrown
away
so
anyway,
I'll
give
it
to
my
kids.
So
so
a
few
things
actually
changed
in
BGP.
So
what
you
see
it
on
the
left
hand,
side
is
your
classic
BGP.
You
know
decision
process
what
we
have
done
with
link
stage
Victor
out
them,
but
Victor
routing
we
simplified
that
so
I
know
everybody
knows
BGP
inside
out.
You
know
in
the
audience
here
yeah
so
face
one.
H
Normally,
you
know
what
you
will
do
is
that
the
calculation
of
degree
of
preference,
so
all
of
the
rights
you
get
for
one
of
your
neighbors,
you
sort
of
like
give
it
a
certain
preference.
You
know,
so
you
can
compare
it
later
on
in
phase
two.
You
actually
strano
compare
off
all
the
previous.
You
have
with
the
preference
thing
and
you
figure
out
the
best
one
and
you
put
it
into
your
local
rib
and
then
at
the
phase.
Three.
No,
you
actually
start
distributing
goes
round
again
to
watch
your
neighbors.
H
So
there
is
a
traditional
process
and
phase
one
and
phase
two
is
what
most
people
commonly
know
or
actually
what
is
known,
as
you
know,
the
part
effective
algorithm,
and
that
is
what
we
are
changing
here
so
phase
one
and
two
in
this
case
is
being
replaced
by
this
new
thing
by
the
SPF
algorithm.
So
that
actually
means
you
know
we're
gonna,
be
doing
the
BGP
and
a
right
selection
and
at
the
same
time
you
know
we're
gonna
do
nearly
in
parallel.
H
H
So
in
traditional
classic
BGP
phase,
1
phase
2
and
phase
3,
they
are
consecutive
from
each
other
in
the
Lynx
and
Victor
routing
algorithm.
It
doesn't
really
have
to
be
the
case
because
we
can
actually
do
those
things
in
parallel,
which
actually
makes
it
that
if
you
have
a
piece
of
information
that
needs
to
be
transported
through
the
network,
it
can
go
really
really
really
much
faster,
very
similar,
the
same
speed
as
a
traditional
AGP
like
a
SAS
or
OSPF.
H
So
some
of
the
advantages
is
already
you
know
spoken
about
is
a
little
bit
all
the
notes.
We'll
have
a
complete
view
of
topology,
so
you
can
do
like
really
crazy
stuff
or
you
can
do
smart
stuff
and
so
up
to
you
and
what
you
want
to
do
with
this
it
another
benefit
of
that
is
that,
if
something
actually
breaks
in
the
network
environment,
just
like
with
the
traditional
link
state
protocol,
you
only
communicate
the
failure
of
what
you
have,
because
SPF
will
pick
up.
I
will
understand
and
recompute
the
impact
of
that
particular
failure.
H
You
know
truly
topology
with
classic
BGP.
If
something
actually
happens,
you
have
to
really
knows
all
the
routes
were
sitting
behind
that
failure
again,
which
makes
everything
much
more
heavy
slower
and
more
complicated
than
you
know
a
little
bit
harder.
You
know
from
that
perspective
at
the
same
time,
if
you
at
the
same
time,
yes,
if
you
know
the
topology,
then
you
can
actually
make
smarter
or
more
optimal
routing
decisions.
You
know
for
the
traffic
through
your
network,
you
don't
always
have
to
go
to
watch.
H
You
know
the
shortest
path
towards
your
next
hop,
because
you
know
the
topology,
you
can
make
smarter
decisions,
it
is
up
to
the
implementer.
You
know
to
actually
you
know
to
create
a
code
or
actually
you
know,
do
those
smarter
things.
So
in
essence,
you
know
that
reality
is
there
so
almost
to
the
last
slide.
So
what
is
hot
in
link
state
vector
routing?
Well,
right
now,
you
know
I,
think
I've
the
impression
that
the
draft
section
you
know
or
have
it's
me
as
chair
of
course,
trying
to
you
know,
market
my
products
here,
yeah.
J
H
The
draft
selection
of
incredibly
good
quality
easy
to
read.
It
will
make
your
day
happier
even
if
you're
in
Bangkok,
it's
fantastic
yeah.
So
it's
super,
it's
wonderful,
so
the
BGP,
you
know
SPF
draft
itself,
so
we're
gonna
be
doing
a
working
group
last
column
that
you
know
we're
gonna,
be
releasing
that
very
soon.
I
invite
everybody
in
a
room
here
to
actually
have
a
very
good
read
on
that.
So
what
we
actually
need
to
further
improve
these
documents,
because
you
know
they
actually
went
through
the
writing.
Directorate
I
went
to
the
operational
Directorate.
H
All
of
the
commands
actually
have
been
addressed,
but
at
the
same
time
you
know
it
does
really
need
to
have
like
more
inputs
from
you
know,
a
lot
more
people,
the
more
people
who
can
actually
give
like
you,
know
constructive
feedback
or
give
information
about
what
you
know
can
potentially
be
changed.
What
is
good
and
what
is
bad?
You
know
please
let
us
know.
So
we
can
really
really.
You
know
further
improve
the
document
and
make
sure
that
you
know
what
we're
doing
here
is
developing
the
right.
You
know
the
right
material.
H
So
that's
point.
One
two
and
three
and
then
the
other
thing
you
know
what
actually
has
been
hot.
Very
you
know
heavily
discussed,
is
you
know,
as
a
remember,
you
know,
as
you
can
see,
actually
could
not
see
on
that
picture
with
that.
You
know
with
your
eyes
closed
the
little
puzzle
pieces
so
that
puzzle
piece
needs
to
be
created.
Now
you
can
create
this.
You
know
either
by
by
some
sort
of
a
divine
resource.
H
Probably
you
configuring
all
of
the
link
States
and
you
know
identifying
that,
but
it's
not
going
to
make
like
you
know
somebody
operates.
Router
is
very
incredibly
happy
because
you
don't
want
to
do
that.
So
you
want
those
puzzle
piece,
so
you
want
to
link
state
vectors
to
be
created
by
themselves
and
automatically.
So
that
means
you
need
to
have
a
technology
out
there
that
can
harvest
the
right
sets
and
the
right
hooks
of
information
for
you
to
create
that
link
state
vector.
So
we
have
been
looking.
H
You
know
around,
you
know
all
the
different
kind
of
technologies
we
have
been,
you
know
being
pulling
the
ball
left
and
right
and
all
the
different
things
of
what
is
available.
We
see
the
same
kind
of
movement,
you
know
I,
can
you
know
in
the
idea
all
working
group
and
we
started
to
compare
them
now.
One
of
the
key
elements
here
is
that
the
requirements
you
know
of
what
we
have
from
al
instead
vector
routing
perspective,
are
actually
different
from
the
requirements
which
there
are
from
an
idea
perspective,
so
link
cell
vector
out
technology.
H
Is
you
know
a
little
bit?
You
know
little
bit,
you
know
it's
not
really
deployed
that
much
because
it
doesn't
really
exist,
so
we're
not
really
restrained
by
the
fact
that
whatever
we
do
has
to
be
a
technology
which
proven
you
know
track
record
yet
so
we
can
do
a
bit
more.
You
know
we
can
be
a
bit
more
creative.
So
what
we
have
decided
now
you
know
in
in
our
in
the
lsv.
Our
working
group
is
to
actually
know
start
a
working
group.
H
H
So
the
way
we
are
doing
this
is
by
making
it
extendable
by
using
the
magic
of
tlvs.
So
these
are
like
wonderful
everything
is
with
deal
of
each
nowadays,
so
we
use
still
these
because
it
is
almost
as
cool
as
PGP.
So
so
we
going
down
that
road,
so
that
is
where
we
are,
and
that
is
what
we
have
right
now
and
again:
I
love
the
way.
H
It's
you
know
it's,
it's
almost
over
trivial
simplicity
and
there
is
no
IPR
test,
which
is
something
that
Randy
Bush
seems
to
love
a
lot
really
a
lot
and
me
too,
so,
that's
fantastic.
So
that
is
where
we
are
right
now.
So
that
is
a
quick
update
about
you
know.
Alice
VR,
I
hope
that
you
know
I
was
able
to
entertain
you
a
little
bit.
You
know
with
my
blessings
about
Ella's
VR,
so
any
questions
please
come
to
the
mic,
but
don't
expect
me
to
answer
them
because
we
have
smarter
people
in
the
room.
A
K
K
K
So
the
first
I
want
to
talk
about
the
issues
with
existing
solutions.
Why
did
we
move
to
ebgp
the
first
place?
The
two
main
drawbacks
with
a
GP
in
mstc
is
that
the
first
one
is
the
failure
impact
scope.
Some
people
call
it
plus
radius,
a
small
change
or
link
up
and
down
anywhere
in
a
small
corner
of
the
data
center
is
flooded
everywhere
and
that
triggers
SPF
recalculation
on
every
note,
and
that
applies
to
lsv
r2
as
well.
K
Another
issue
is
the
rich
connections
in
the
data
center
network
makes
flooding
unnecessarily
redundant
and
inefficient.
So
those
are
the
two
main
problems
with
a
GP.
That's
then
people
move
to
the
ebgp
solution,
but
there
are
also
two
issues
with
that.
The
first
issue
is
that,
with
a
well-defined
technology
class
network
in
data
center,
ideally.
K
The
routing
tables
could
be
very
simple,
for
example,
a
leaf
node.
All
your
needs
is
a
default
routes
on
the
next
level.
You
only
need
a
default
route
and
more
specific
subset
routes
to
destinations
destinations
south
of
it
that's
under
normal
situations,
but
if
you
just
if
you
do
that,
only
and
then
in
some
situations
you
may
end
up
a
black
hole.
K
The
next
issue
is
a
BGP.
Is
that
some
because
we
have
massive
EMP
paths
in
the
network,
for
example,
and
node
could
have
32
64
100
128
years
and
for
every
prefix
you
will
learn
the
same
periphery
from
all
those
peers
and
you
have
to
keep
all
those.
So
that's
in
the
control
plane,
it's
a
lot
of
resources
wasted
and
lots
of
stay
talking
to
keep.
K
K
Ok,
so
if
you
look
at
this
picture,
we
have
three
layers:
the
purple,
purple
layer,
green
layer
and
then
the
pink
layer.
So
we
do
linked
States
and
distance
vector
in
two
directions
on
the
right
side,
where
we
are
showing
that
we're
flooding
the
linked
States
all
the
way
from
on
the
northbound
all
the
way
to
the
top.
Well,
actually,
on
the
on
the
left
side,
we
do
the
same
thing
as
well.
K
K
The
link
state
from
this
green
note
is
flat
itself
by
one
hub,
owning
and
also
reflected
north,
so
that
the
other
green
node
at
the
same
level
will
also
learn
it.
Why
do
we
want
to
do
that?
It
would
become
clear
later
now
anyway,
so
because
we
fly
the
link
States
all
the
way
to
the
north,
then
every
node
will
know
the
topology
full
topology
south
of
it
and
under
and
then
on
the
southbound.
With
we
use
link
States.
We
we
do
a
distance
vector
routing
to
to
propagate
default,
owning
by
one
hub.
K
K
K
Let's
say
this
leaf:
node
gets
the
packet,
it
has
two
choices.
You
are
following
the
default
route:
they
can
choose
either
a
screen
note
or
the
screen
notes
as
fine,
and
let's
say
if
it
shows
this
note-
and
it
again
has
two
nodes-
a
two
choices
to
go
this
one
or
this
one.
But
if
this
link
is
down
and
this
one
still
choose
to
go
this
way
and
the
traffic
will
reach
there
and
has
nowhere
else
to
go
so
that's
the
blackening
issue.
If
you
don't
do
anything
else
now,
how
do
I
solve
this
problem?
K
K
Spf
calculation
based
on
the
topology
flooded
northbound,
then
it
will
realize
that
he
had
a
route
towards
p1
using
this
link
here.
He
also
will
realize
that
his
body
here
does
not
have
a
pink
to
this
note
here,
so
his
body
will
not
be
able
to
reach
p1
once
he
gets
that
conclusion.
It
will
send
a
specific
route
for
p1,
using
the
link
using
the
distance
vector
part
of
this
protocol
and
notice.
I
also
also
note
that
this
p1
routes
will
stop
at
this.
Green
node
is
not
propagated
further
down.
K
It's
just
one
half
it
stops
here
now.
If
you
have
a
packet
arriving
on
this
nook
trying
to
reach
p1-
and
let's
say
this:
node
picks
its
node
as
a
lookup,
you
find
the
default
routes
and
then
the
packet
ends
up
here
now.
This
green
node
here
now
has
two
routes.
One
is
the
default
routes.
The
other
one
is
the
most
specific
p1
routes
that
this
kink
node
has
automatically
inject
it.
Now
the
packet
will
follow
the
p1
routes
reaching
this
pink
note
down
here,
instead
of
going
to
the
top.
K
So
that's
how
we
solve
the
Brioni
issue.
So
with
this
automatically
aggregation,
we
can
use
default
routes
and
plasti
were
specific
routes
through
the
destinations
south
of
south
of
the
nodes,
and
if
link
fader
happens
in
somewhere,
if
there
is
a
needs
to
do
the
auto
this
D
aggregation,
it
will
happen.
K
Component
of
this
rift
protocol
is
zero-touch
provisioning.
If
we
go
back
to
this
picture,
the
owning
these
two
top
nodes,
we
called
top
of
the
fabric
needs
configuration
very
simple
convictions,
indicating
that
they
are
at
the
top
other
than
that
no
configuration
is
needed
on
the
notes.
All
you
need
to
do
is
to
power
up
connect
the
cables
and
that
they
were.
K
The
nose
will
discover
each
other
decide
to
mean
set
up
the
routing
adjacencies
and
if
you
happen
to
have
me,
miss
keep
learning
and
the
protocol
will
make
sure
that
the
logical
topology
that
we
obtain
from
from
the
routing
adjacency
that
we
decide
to
do
that's
the
topology
topology
will
be
still
conformed
to
a
strict
north
north,
bound
South
Mountain,
a
class
type
of
topology.
That
way,
you
don't
run
into
routing
notes,
and
that
kind
of
thing
so.
K
There
are
a
lot
of
other
features,
either
peyote
or
ed,
ed,
on
of
the
rifts
of
automatically
optimal
reduction,
flooding,
flooding
reduction
and
not
balanced.
We
mentioned
earlier
that
either
is
a
issue
with
a
valina
ICP
routing
the
flooding
are
unnecessary,
unnecessary
flooding
and
that's
is
reduced
here
and
it
has
built
in
mobility,
support,
key-value
store
and
fabric
bandwidth
balancing.
K
K
K
K
Yeah
I
guess
that's
the
last
slide,
so
I
I
thought
I
was
only
going
to
give
overview
of
the
protocol,
so
I
didn't
put
a
steady-state
there,
but
I
can
talk
about
it,
so
the
working
group
was
chartered
immersed
in
London.
We
have
we
had
one
intern
meeting
in
May
and
we
have
been
that
the
core
team
have
been
working
having
having
weekly,
sometimes
twice
weekly
design
meetings
that
those
meetings
are
most
of
them
are
recorded
and
announced.
K
We
welcome
open
party
participation
so,
but
typically
it's
just
the
core
members
of
the
team,
but
it's
open
to
everyone.
Our
original
all
the
open
source
implementation
has
been
going
on
brutal
righteous
men.
It's
doing
been
doing
a
fantastic
job
with
his
helpers
and
that
overall
protocol
we
right
now
we
are
at
zero
three
revision,
that
zero
for
revision
is
evolving
and
our
original
plan
was
to
submit
to
iesg
for
the
pace
based
protocol
in
February.
K
L
M
K
Another
thing
I
forgot
to
put
there
and
I
know
that
some
Lara
wanted
to
hear
about
is
the
the
thrifter
model
we
actually
last
time.
We
actually
talked
about
this
extensively
in
in
the
working
group
meeting,
so
the
in
O'day's
routing
protocols
are
specified
with
packet
formats
and
despite
and
that
beats
reader
for
what,
for
whatever
and
with
rift
the
pack,
we
don't
have
packet
format,
the
old-style
packet
format,
definition,
any
more.
Everything
is
RIF
model-based
that
makes
development
a
lot
easier.
K
N
C
Alright,
so
Enterprise
I
mean
lots
of
stuff
has
been
suppressed,
so
rift
is
loop
free,
which
means
you
can
take
any
path
to
anywhere.
So
if
you
have
a
rift
is
doing
through
any
cast.
So
not
this
any
cast
like
IP,
which
is
bound
by
a
CMP,
basically
but
rift.
When
you
see
the
same
address
on
the
fabric,
doesn't
matter
what
the
distance
is,
and
you
can
split
the
traffic
also-
and
you
know,
based
on
the
bandwidth
distance,
whatever
you
want
more
than
that.
C
O
A
A
P
So
no
Anderson
I
have
a
real
name.
We
got
the
mail
or
the
working
group
shares
got
the
mail
from
the
ADEs.
How
we
in
the
future
need
to
update
the
working
group
status
I.
Think
it's
a
brilliant
idea.
It's
the
right
place
to
have
it,
though,
when
you
look
at
the
working
group
page,
it's
extremely
hard
to
actually
find
the
working
group
status
where
you
need
to
click
the.
Q
A
I
A
P
P
A
So
we
could
put
this
on
my
said:
I.
We
had
never
used
this,
so
I
originally
thought
that
what
would
happen
is
that
there
would
be
a
little
window
like
thing
right
here.
That
would
show
the
actual
summary.
So
that's
why,
when
I
sent
this
was
chairs
now
we
said
you
know,
let's
do
this
a
couple
paragraphs
or
something
but
I,
think
a
tab.
Is
this
a
great
idea
and
and
since
yeah
they
do
it
for
us
anyway
said
it
would
be.
P
E
It
so
if
you're
gonna
find
it
where
you
expect
to
find
it
in
some
ways,
it
really
belongs
on
the
bed
on
the
bottom
of
the
Charter,
because
the
Charter
has
got
the
current
status
of
the
working
group.
It's
got
what
you
what
it's
work
our
package
is,
it's
got
what
it
has
done
and
it's
got
what
it
is
expected
to
do
and
isn't
that
exactly
you
know,
status
information
and
where
you'd
expect,
if
I
may
be
an
amplification
of
the
status
information.
A
A
A
Just
do
it
that
so
that
we
can
see
the
actual
you
know,
status
of
when
you're
going
to
deliver
things
etcetera.
So
the
intent
of
the
status
update
this
we're
talking
about
is
that
here's
a
short
summary
because,
yes,
okay,
you
go
read
the
the
Charter.
You
know
what
the
worker
who
was
supposed
to
be
doing.
A
We
want
to
know,
especially
for,
as
we
said,
external
audiences,
what
the
worker
was
doing
now.
You
know
if
I'm
interested
in
in
in
any
working
group,
let's
say
LSR,
you
know
what
what
is
it
discussing
now.
You
know
said
work
for
me
to
go
to
the
meeting.
What
are
the
drafts
that
maybe
I
should
go?
Do
it
now,
because
otherwise
you
look
at
the
document
agent
there's
a
lot
of
documents
and
it's
not
necessarily
easy
for
anyone,
not
following
the
work
group
to
tell
exactly
what's.
A
A
A
Yes,
there
are
many
things
or
there
are
some
things:
I
guess
we
can
do
about
them.
The
first
order
is,
of
course,
the
conflict
list
that
is
put
into
the
session
requests,
and
please
revise
that.
We
have
tried
unsuccessfully
before
to
try
to
prioritize
that
list,
so
that
would
be
easier
for
the
Secretariat
and
for
us
to
understand.
You
know
conflicts
that
we
have.
We
usually
try
as
much
as
we
can
to
avoid
any
significant
conflicts.
However,
you
know,
there's
always
gonna
be
something.
A
Maybe
we
can
avoid
more
than
we
have
avoided
now,
but
but
yeah,
that's
those
initial
step.
The
other
thing
that
would
be
really
useful
for
us
is
as
the
preliminary
agent
that
comes
out.
If
you
see
anything
that
we
should
be
avoiding
those
no
and
put
it
on
the
conflict
list
for
next
time
as
well,
so
that
we
try
not
to
keep
stepping
on
that.
But
but
yes,
I
mean
we
try
to
do
the
best
we
can
this
week
because
we
have,
on
the
last
day
or
half
of
us
they
may
be.
R
R
Think
it's
a
waste
of
the
day,
there's
only
like
four
small
groups
that
are
meeting
tomorrow
with
that,
based
on
what
I
saw
on
on
the
room,
reservations,
I,
think
it's
is
it's
a
waste
of
the
resources
that
were
paying
for
in
the
hotel
and
I.
Think
a
much
better
use
of
the
day
is
what
we've
done
in
the
past.
A
J
S
Support
to
previous
speakers-
Gregory
egt.
Yes,
please
even
their
time
we
used
to
have
on
Friday
for
group
meetings,
helped
to
relieve
conflict
and
another
plea
to
work
in
group
Cheers.
Please
be
very
considerate
when
you
request
working
group
sessions
listing
possible
conflicts
with
other
groups
in
the
area,
at
least
in
the
area,
because
otherwise
running
from
one
group
to
another
to
have
presentations
here
and
then
presentation
there.
It's
a
lot
of
stress.
D
Say
something,
then
you
can
answer
my
question
in
this
income,
so
one
of
the
ideas
actually
kicked
out,
which
was
might
be
even
more
revolutionary,
was
that
to
just
have
one
hour
sessions
for
all
the
working
groups
said
they
more
effectively
made
no
use
of
their
meeting
time
and
they
could
use
interim
meetings.
You
know,
for
other,
you
know
discussions
but
to
go
just
and
to
one
hour
and
are
grouting
our
reactions,
one
of
the
only
ones
that
really
requested.
You
know
their
two-hour
sessions
to
two
and
a
half
hour
sessions
whatever.
D
So
do
you
really
feel
right
that,
because
I
would
reduce
the
complex
to
it
right
if
everybody
just
went
to
one
hours,
but
do
you
really
see
the
utility
right
to
have
the
discussions
in
person
maybe
I'm,
leading
in
but
right
that
we
have
in
person
we
have
we're
here
and
that
we
make
good
use
of
that?
Time
might
have
the
two
hour
of
sessions
david.
B
E
Not
kind
of
support
what
David
said:
I
belong
are
going
to
take
a
different
point,
which
is
that
we
do
the
we
use
the
eighteenth-century
method
of
figuring
out
conflicts
right.
People
who
understand
what's
going
on
sort
of
figure
out
from
you
know
what
they
know
what
the
conflict
should
be.
We
live
in
the
21st
century
with
big
data
in
all
those
sorts
of
things.
So,
though,
it
should
be
possible
to
generate
a
conflict
terms
hint
using.
You
know
what
we
know
about
who
was
in
the
meetings,
and
you
know
who
is
on
the
draft.
E
Well,
even
if
you
didn't
figure
out
from
the
blue
sheets,
which
you
probably
could
do
only
online
I'm,
not
sure
with
near
modern
handwriting
techniques,
if
you
didn't
do
that,
you
could
at
least
figure
out
who
was
active
in
which
area
and
try
and
minimize
the
the
overlaps.
You
could
do
that
from
the
mailing
lists,
which
are
public
information.
You
could
do
this
from
the
author
lists,
etc.
A
Yeah
I
just
want
to
point
out
not
one
of
more
excuse,
but
one
more
circumstance
this
week
this
week
for
the
first
time,
we
added
a
conflict
with
every
working
group
for
the
routing
area
working
group.
So
it
used
to
be
that
in
the
past
we
had
other
routing
groups
meeting
while
deliver
our
twg
was
meeting.
So
this
time
we
conflicted
that
which,
of
course,
you
know
that's
how
things
go
ahead
and
then
I
want
to
ask
another
question
similar
to
what
they
were
asked
before.
I
want.
A
Food,
but
will
take
all
the
credit
for
that
you
know.
Take
the
blame
for
the
rest
of
the
AG
will
take
the
credit
for
the
food.
So
we
talked
about.
You
know
the
fact
that
mostly
the
routing
area
of
groups,
some
of
them,
are
the
ones
who
select
some
of
the
bigger
slots.
One
thing
that
we
have
to
do
this
week
was
have
maybe
two
sessions,
a
two-hour
session
and
the
one-hour
session,
maybe
later
and
by
couple
groups,
met
twice
but
at
different
times
in
the
week.