►
From YouTube: IETF114-HOTRFC-20220724-2200
Description
HOTRFC meeting session at IETF114
2022/07/24 2200
https://datatracker.ietf.org/meeting/114/proceedings/
A
A
We
are
still
getting
slightly
organized.
We
are
most
of
the
way
there,
but
we
are
not
all
the
way
there.
So
if
anybody
would
like
to
come
up-
and
it's
going
to
be
speaking,
come
up
and
make
sure
that
everything
is
okay
for
them,
that
will
be
great,
and
I
look
forward
to
knowing
what
the
answer
is
very
soon.
A
A
A
B
C
A
D
A
A
A
A
D
D
F
A
F
D
E
F
B
A
A
F
A
A
H
D
I
J
A
A
A
A
A
A
A
A
A
So
when
you're
speaking
and
you,
if
you,
if
you
happen
to
hear
people
applauding
that-
is
your
opportunity
to
recognize
what
a
great
job
you've
done
and
passed
back
the
micro
microphone
to
the
next
speaker,
we
don't
do
questions
after
each
talk
here
that
basically,
the
thing
with
hot
rfc
is
that
this
is
where
you
come,
and
you
provide
follow-up
so
basically
saying
what
you
want
people
to
do
and
how
you
want
them
to
find
you
to
do.
It
is
part
of
each
talk.
D
And
I'm
removing
the
mask,
because
my
accent
is
worse
enough
already.
I
hope.
That's
fine
still
louder.
Okay,
all
right
people
on
the
back
falling
over
okay
yeah,
I'm.
I
hope
it's
okay
to
remove
the
mask,
while
I'm
here
on
the
microphone,
because
my
accent
is
already
bad
enough
and
I
don't
need
a
microphone
to
make
it
worse
and
maybe
if
we
can
get
the
levels
up,
that
would
help
people
to
really
not
eat
the
mic,
not
sure
if
the
technician
is
still
here,
while
spencer
is
still
working
and
going
going
going.
A
D
All
right
welcome
everybody,
so
we
we
started
to
think
about
what
might
be
interesting
to
do
about
energy,
green
networking,
sustainability
and
so
on,
and
then
you
step
back
and
say.
Well,
you
know
what
is
already
been
done
in
the
ietf
and
when
we
started
to
write
this
together,
it
actually
went
through
a
lot
of
different
areas
which
reminded
me
of
really
the
monty
python,
the
life
of
brian
right.
So
what
have
the
romans
ever
done
for
us?
And
you
know
you
start
with
oh
yeah,
the
itf.
D
That's
just
lln
low
power
networks,
iot,
that's
all
it's
energy
related
and
when
you
then
go
through
what
diet
have,
and
I
think
it's
a
lot
more.
So
the
purpose
of
the
document
is
to
become
maybe
an
individual
submission
document
for
education
of
the
community
and
to
help
you
know,
steer
interest
in
new
work,
but
also
to
broaden
the
adoption
of
the
specific
technologies
mentioned
now
I'll
I'll.
Just
you
know,
show
you
here
the
the
content
of
the
document
a
lot.
D
What
I
think
really
helps
preserve
energy
from
the
iatf
work
is
really
incidental,
meaning
all
the
technologies.
We've
done
to
some
extent
do
save
energy
compared
to
their
prior
non-digitized
solutions
and
then
the
other
part
that
really
made
the
itf
technologies
become.
Very
energy
saving
is
the
the
saving
through
scale
right
building
the
internet
converging
all
these
different,
separate
network
applications
into
one
network.
So
those
things
are
big
areas
in
the
document
and
then
looking
into
specific
applications
like
telecollaboration
computing
with
sustainable
energy.
So
there
are
really
specific
areas
I
think
of
interest.
D
Obviously
I
hope
I'm
missing
a
big
area.
Some
people
will
jump
up
and
say
you
know
we
can
contribute
other
stuff.
That,
incidentally,
relates
to
energy
saving
as
well,
then
the
obvious
one
low
power,
lossy
constraint,
networks-
we've
got,
you
know
almost
an
area
worth
of
working
groups
that
are
working
on
various
technologies,
and
then
there
are
some.
D
You
know
sample
technologies
where
you
know
I
just
had
some
background
and
interest
in,
but
hopefully
there
will
others
be
added
to
it
and
then,
of
course,
the
specific
networks
being
built
to
actually
support
the
generation
of
power.
Smart
grid
synchro
phaser
networks,
that's
in
the
totally
different
approach
and
then
energy
management.
We
had
an
e-man
working
group
and
later
on,
there
were
ideas
to
do
power,
awareness
in
networks,
but
they
didn't
pan
out
so
far
to
form
working
groups.
But
there
are
some
good
initial
drafts
next
slide.
D
D
G
F
F
It
show
okay,
it
doesn't
show
here:
okay,
well,
okay,
so
yeah
so
welcome
everyone.
So,
basically,
this
presentation
is
briefly
to
build
on
what
charles
was
just
presented,
presenting
on
challenges
and
opportunities
in
green
networking,
and
there
are
two
drafts
actually
which
this
presentation
relates
to.
Please
next
slide.
A
A
A
A
F
F
Okay,
so
basically
yeah.
This
builds
a
little
bit
also
on
what
charles
was
presenting
earlier.
So
the
question
is
why
green
networking,
or
why
sustainability
actually
and
obviously
this
is
one
of
mankind's
grand
challenges
and
the
question
is
basically
well
while
network
applications
themselves
are
key
enabler,
of
course,
the
sun
sustainability.
What
about
the
networks
themselves
and
clearly
the
net
zero
mandates,
etc
will
apply
to
the
network
providers
as
well.
Sorry,
we
lost
connectivity
again.
A
A
F
F
What
are
ways
in
which
the
itf
and
the
network,
at
the
network
level
how
we
can
contribute
to
making
networks
more
sustainable-
and
this
may
be
perfectly
a
smaller
factor
than
general
hardware,
advances
and
deployment
factors
and
antenna
technology,
but
still
of
course,
there
are
certain
things
which
are
within
our
control
next,
one
next
slide.
Please.
F
Okay,
and
so
there
are
quite
a
few
opportunities
and,
of
course,
challenges
associated
with
them.
We
divided
them
into
four
different
yeah
into
four
different
levels,
basically
one
for
one,
basically
going
bottom
up
at
the
device
or
at
the
equipment
level.
It
starts
with
providing
proper
instrumentation
and
visibility
into
the
right
types
of
metrics
which,
because,
unless
you
have
visibility
to
this
all
the
other,
well,
then,
basically,
it's
kind
of
like
future.
What
else
you're
attempting
to
to
do?
F
F
If
you
will
to
enabling
other
mechanisms,
such
as
for
ins,
fast
discovery
for
state
reconvergence,
if
you,
for
instance,
dynamically
take
resources
in
and
out
of
service,
for
instance,
network
addressing
to
have
smaller
tables,
to
maintain
et
cetera,
et
cetera
at
the
network
level,
there
is
a
question
of:
can
we
do
more
energy
where
routing
path
configuration
takes
energy
and
sustainability
into
account?
What
would
be
some
of
the
control
protocol
extension
that
would
make
sense
there
and
then
find
basically
also
at
the
architecture
level.
One
can
look
at
some
of
those
things
as
well.
F
Basically,
when
it
comes
to
where
do
we
place
content?
Where
do
we
place
computation?
Of
course,
people
are
already
looking
into
this,
but
not
so
much
with
the
energy
or
sustainability
in
the
foreground,
but
certainly
an
aspect
that
that
plays
a
role
there
as
well,
and
so
there
are
many
pieces
and
support
that
can
contribute
to
this.
Not
one
brand
solution
is
required.
F
Next
next
slide,
please,
and
so
basically,
we
want
to
generate
basically
more
discussion
on
this,
and
so
we
have
posted
two
drafts
on
this
topic
and
would
appreciate
feedback
comments
on
this.
There's
a
problem
statement
and
then
basically
metrics
is
the
first
specific
work
item
and
we
are
also
still
looking
for
collaborators
and
actually
the
proper
landing
spot
here
here
in
the
idf
between
the
authors,
we
will
have
an
informal
site
meeting
on
tuesday
from
1
to
2
in
salon
9..
If
you
are
interested
in
joining
and
contributing.
F
K
C
C
C
This
process
have
actually
reached
its
very
first
milestone
actually
this
month
and
they
decided
to
actually
announce
the
selected
algorithms
that
are
going
to
be
standardized
for
the
key
exchange
process,
meaning
that
that's
the
process
in
which
you
generate
a
key,
and
you
also
share
that
team,
with
a
participant
to
arrive,
to
arrive
to
confidentiality
and
also
this,
they
decided
to
standardize
the
algorithms
for
authentication,
mainly
digital
signatures.
This
is
just
the
first
milestone.
C
C
So
why
is
this
important
for
the
itf,
because
one
of
the
things
that
one
might
think
is
to
just
take
the
pos
quantum
algorithm,
the
signature
algorithm,
for
example,
and
swap
it
in
the
place
where
a
classic
algorithm
might
have
been.
This
is
not
so
simple
because
all
of
the
post,
quantum
algorithms
in
in
specific,
the
signature,
algorithms
have
bigger
parameter
sizes
when
compared
with
the
classical
counterparts.
C
Don't
have
all
of
the
difficult
properties
that
we
love
and
also
certain
algorithms
are
now
currently
being
standardized
by
the
iita.
For
example,
in
the
ppm
group,
don't
really
have
a
postquantum
counterpart
that
can
be
attested
for
a
high
security,
just
like,
for
example,
syrian
united
states
threshold
signatures
to
have
past
quantum
alternatives,
but
they
have
not
been
properly
assessed
for
their
security,
excess
life.
Please.
C
If
you're
ever
interested
interested
we're
running
a
workshop
on
this
specific
challenge
of
putting
quantum
cryptography
network
and
protocols,
it
is
going
to
be
potentially
collocated
with
the
niss
workshop
on
november.
So
please
reach
out
if
you're
interested
in
participating
to
it
and
there's
also
currently
a
new
itf,
mainly
spqc,
that
you
can
join
and
continue.
The
discussion
next
is
like
and
with
that.
Thank
you
very
much.
K
K
There
have
been
a
reincarnate
rules
for
a
packet
up
to
256
bytes,
the
programming
environment,
like
java
card
6
billion
device
deploy
every
day,
6
billion
java
cards,
and
there
are
secure
software
manager
but
something's
called
global
platform
in
order
to
list
the
letter
upload.
The
software
next
slide,
please
so,
while
connecting
secure
element
to
internet,
it's
very
simple:
we
want
online
cryptographic
resource
for
internet
users
and
we
want
to
identify
this
resources
by
a
uniform
resource
identifier.
K
The
issue
there
is
no
tpip
stack
in
secure
elements,
so
we
need
additional
processors
to
manage
the
networks.
We
need
to
support
global
platform
for
on-demand
protocols.
We
need
the
defined
protocol
to
access
secure
element,
resources,
we
need,
we
need
to
define
ticker
element
naming
and
we
need
to
define
attestation
procedures
for
on-demand
application
for
next
slide.
Please.
K
So
that
mine
is
a
free,
atf
draft.
In
order
to
to
support
this,
one
is
called
rox,
it's
an
old
draft
and
this
graph
is
enabled
to
support
a
global
platform
over
tlx.
The
second
one
is
called.
The
second
draft
is
called
tlssc
for
tls
for
secure
element.
It
supports
a
tls
ps,
key
in
a
secure
element,
and
this
gives
a
tls
interface
to
secure
element.
K
K
So
at
this
moment
there
are
open
software.
One
is
tls
sc
for
java
card.
It
works
with
java
card
3.04,
it's
available
in
github
and
the
second
open
software
version
5,
which
works
with
windows,
ubuntu
and
raspberry
pi,
and
it
is
an
implementation
of
racks
and
tls
from
server,
and
it
supports
a
multiple
communication
interface
like
pcic,
i2,
sim
and
sim
array,
and
that's
it
I'm
done.
G
G
It
actually
typically
establishes
a
secure
communication
channel
with
the
cloud-based
service
and
so
next
slide.
It
was
kind
of
obvious
to
combine
the
two
steps
of
attestation
with
dls
and
that's
what
we
have
been
doing
in
in
a
recently
published
document,
and
also
software
we've
been
working
on.
The
solution
is
simple:
to
augment
the
dls
exchange
with
the
attestation
information
and
to
provide
a
proof
of
procession.
G
The
software
we
are
working
on,
which
will
hopefully
will
be
able
to
release
shortly,
uses
a
combination
of
the
embed
dls
with
dls
1.3,
a
platform
agnostic
library
to
access
security,
hardware
called
basic,
and
on
the
cloud
side
there
is
a
verification
services
service.
We
need
to
use
which
colleagues
of
mine
have
been
working
on
called
verizon
and
we
basically
munch
all
these
three
components
together
to
accomplish
a
workable
system
next
slide,
and
here
here's
the
document
and
pointer
to
the
projects
we're
using
and
soon.
A
J
J
Why
are
your
slices
so
special,
because
it's
a
moment
it's
moving
so
fast?
Unfortunately,
this
is
originally
a
ppt.
Has
some
video
clip
if
you're
interested,
I
can
give
you
you
can
shoot
me?
Email
first,
is
that
50
satellites
moving
to
different
directions?
Another
50
percent
this
this
will
cause
the
satellite
networking
becoming
our
interleaved
mesh
network
and
moving
at
a
very
high
speed,
also
including
the
self-rotation
of
earth,
makes
the
movement
is
even
more
complicated.
J
J
So
itf
has
done
a
lot
of
work
for
many
years
for
satellites,
but
unfortunately,
most
of
our
work
are
focusing
other
layers.
For
example,
we
have
done
a
dtn,
it's
not
fitting
to
air
you
and
we
have
done
air
for
works,
which
is
about
transport
quotation
layer,
and
we
have
done
the
coding
for
the
satellites.
J
The
first
question
we
have
to
answer
is
that
is
the
ip
networkings
needed
for
air?
U,
because
many
people
challenge
me
say:
starling
doesn't
use
it,
that's
true,
but
we
have
many
reasons
to
to
to
predict
that
the
l3
technology
is
still
best
for
the
aerial
satellite.
Firstly,
the
scalability
second
is
the
interworking
with
internet
and
the
third
one
is
that
3gpp
has
already
expected
that
the
satellite
network
is
ip
network.
J
J
H
That
said
next
slide,
please,
if
you
look
at
the
state
of
webrtc
today,
this
is
what
users
are
seeing
all
that
open
standards,
based
goodness,
is
hidden
behind
a
proprietary
walled
garden
next,
what
next
likely?
What
users
need
is
the
ability
to
communicate
with
any
other
user,
no
matter
which
service
provider
is
providing
their
webrtc
contact
and
they
should
not
just
be
able
to
choose
a
messaging
provider
and
talk
to
everybody.
They
should
be
able
to
change
their
messaging
provider
at
any
time
without
switching
costs.
H
H
The
problem
I
see
in
the
current
end-to-end
messaging
model
is
that
all
we
get
is
end
to
end
security.
We
think
about
securing
the
packets
as
they're,
going
from
the
sender
to
the
receiver,
but
that
doesn't
close
down
all
the
points
of
vulnerability
where
a
hostile
government
can
do
something
hostile
next
slide.
H
H
Well,
the
first
thing
we
need
to
do
is
to
complete
the
messaging
stack
and
I've
been
looking
at
webrtc,
and
you
know,
I've
got
this
threshold
key
infrastructure
that
manages
public
and
private
keys
people
across
devices
and
makes
it
really
easy
for
the
end
user
is
transparent
security.
So
what
I
want
to
do
is
to
take
my
security
stuff,
my
tki
and
bolt
it
on
top
of
webrtc.
H
H
Okay,
so
why
I'm
going
to
be
here
till
friday,
if
you're
interested
in
this
type
of
project,
please
talk
to
me,
you
know
tell
me,
I'm
not
a
webrtc
person.
I
can
see
how
to
do
it.
I've
kind
of
sort
of
got
my
prototype
running
and
I'll
be
hopefully
running
that
prototype
by
london,
but
I
would
really
appreciate
help
on
choosing
the
best
pass
path
through
webrtc
and
the
best
of
breed
things
to
go
to
market
with.
So
thank
you
for
your
time
and
mathmesh.com
has
the
further
details.
A
M
M
As
you
know,
automation
grows
and
becomes
more
dynamic.
The
factory
floors
will
require
more
sophisticated
controllers
with
more
processing
time,
more
processing,
capabilities,
more
flexibility
and
more
interoperability
as
support,
and
the
controllers
sitting
in
the
cloud
such
as
virtualized
prcs,
would
become
more
and
more
popular
as
another
use
case.
In
remote
driving
a
remote
human
driver
actually
controls
or
operates
a
vehicle
from
a
distance
through
communication
networks
and
remote
driving
is
extremely
useful
when
autonomous
driving
becomes
you
know
short
or
for
short,
and
in
some
hazardous
environments
where
human
cannot
access.
M
M
But
if
those
end
devices
are
networked,
there
would
some
would
have
some
issues.
So
those
common
elements
or
abstract
elements
by
using
ocn
actually
enable
us
to
have
discussions
on
those
requirements
and
the
issues
next
slide.
Please
so
we
define
our
thing
as
operation
like
it's
a
short
name
for
operation
control
network,
it's
a
interconnection
of
a
few
devices
and
their
controllers
for
the
exchange
of
data
to
cause
and
monitor
change
to
the
end
equipment.
Next
slide.
Please.
M
So
what
are
the
issues
we
want
to
consider
in
ocn,
so
with
the
controllers
being
sitting
in
the
cloud
and
there
are
more
and
more
field
devices
being
deployed?
Maybe
it
is
time
to
explore
the
opportunities
and
methods
to
use
iep
to
interconnect
the
controllers
and
field
devices
and
to
build
ocn
and
high
precision.
M
Communication
are
needed
because
you
want
to
guarantee
the
key
performances
and
to
provide
a
more
granular
qs
to
build
ocn,
and
you
know
there
are
multiple
types
of
you
know
like
a
media
such
as
you
know,
5g
radio
internet,
you
know
all
those
kind
of
different
media
and
as
well
as
there
are
multiple
types
of
field,
bus
communication
protocols.
We
want
to
use
o7
to
convert
them
in
a
large
scale,
with
some
reference
model
and
then
addressing
needs
to
be
supported.
M
Like
you
know,
auto
configuration
and
also
header
region
is
addressing,
and
we
also
need
to
build
message
types
for
example.
What
are
common
characteristics
between
the
controllers
and
field
devices
and
the
last
but
not
least,
is
the
security.
You
know
we
want
to
ensure
the
security
of
communication
between
the
field
devices
and
the
controllers
and
next
slide.
Please
all
right,
so
we
are
going
to
host
a
list
of
meetings
always
saying
tomorrow
lunchtime
in
room.
Actually,
this
is
wrong
slide.
M
A
B
B
So
if
we
remove
the
access
to
the
sni,
ech
is
pushing
security
to
the
end
point
now,
unfortunately,
in
the
paper
we
show
we
can
not
trust
the
device
and
not
trust
the
brother.
We
found
enough
evidence
of
money
in
the
browser
attacks
and
worse.
If
you
really
want
to
protect
the
people,
you
not
only
need
to
access
the
smi,
but
you
need
to
read
the
entire,
the
entire
page.
B
B
B
So
it's
pretty
simple:
we
need
first
clarification
on
the
client-facing
server,
so
the
back-end
server.
Is
it
really
left
to
just
implementation
or
do
we
need
the
protocol
based
for
the
client-facing
server
to
the
back-end
server,
and
in
this
case
we
are
back-to-middle
box
problem
if
ech
constitutes
cannot
be
judging
parties?
How
do
we
integrate
a
third-party
security
component
and
is
there
a
form
of
recognition
like
management
of
of
enterprise
or
personal
security
problems?
B
E
E
So
websites
typically
use
cdns
to
deliver
their
content
efficiently
to
the
end
users,
and
they
have
this
cluster
of
servers
spread
across
across
the
globe,
typically
called
cd
in
edge
and
at
the
edge
we
have
protocols
like
tcp
and
http
that
dictate
the
data
transmission
rules.
So
there's
multiple
options
available.
E
E
So,
in
fact,
when
we
look
at
the
client
here
or
the
user
base,
we
see
that
there's
a
diversity
there.
Users
come
from
different
regions,
have
different
last-minute
connections
like
2d,
3g
4g,
and
it's
a
problem
because
the
performance
of
the
protocols
is
sensitive
to
the
type
of
network
and
devices.
E
So
assuming
that
us
an
edge,
has
three
users,
one
with
low
latency,
one
with
high
bandwidth
and
one
with
high
loss
network
parts.
The
choice
of
optimal
configuration
can
be
very
different
based
on
the
type
of
network
and
in
fact
we
did
some
measurements
at
scale.
That
actually
showed
that
there's
a
tremendous
opportunity
here,
there's
an
opportunity
to
improve
web
performance,
and
the
figure
here
shows
the
improvement
in
ph
low
time
for
websites
and
there's
up
to
70
improvement
at
tl.
E
If
we
are
using
some
sort
of
an
article
to
configure
the
right
configuration
on
the
per
connection
basis.
So
next
please
so
in
this
work
we
propose
a
system
that
optimizes
web
performance
by
systematically
reconfiguring
the
networking
stack
so
basically
taking
into
features
such
as
users
network,
the
device
and
website.
E
E
E
L
Go
ahead
and
read
it,
the
basic
premise
here
is
that
unicast
is
a
real
problem.
I've
been
working
on
this
for
a
while.
I've
been
talking
to
a
lot
of
isps,
got
some
ideas
about
how
to
get
them
to
transport
it.
But
then
the
next
step
is:
what
are
we
going
to
transport
next
slide?
Please,
and
the
basic
idea
is
ip.
Multicast
is
what
we
want
to
use.
So
we
proposed
a
way
to
trans
to
to
let
quick
use
that
next
slide.
Please.
L
L
So
server
knows
what
channels
it
got,
what
what
data
it's
getting
from
these
channels
and
and
the
multicast
is
going
to
be
used
for
server
to
client
only
so
this
matches
up
with
the
security
consideration
stack
that
we
presented
in
sec
dispatch
in
112.
We
think-
and
we
are
looking
to
get
a
prototype
running
next
slide,
please
so
I
just
wanted
to
let
people
know
about
it.
L
I'll,
be
talking
about
this
a
little
bit
more
in
the
quick
working
group
on
thursday
we've
got,
you
know,
we've
got
the
draft
we'd
love
to
get
feedback
on
it.
Anybody
who
wants
a
review
we've
been
working
on
an
implementation
with
the
w3c
multicast
community
group.
If
you're
interested
you
can
join
us
there
and
contribute
to
that,
my
email
is
there,
you
can
ping
me
and
please
come
to
quick
and
or
find
me
anytime
this
week
and
give
me
comments
and
I'd
love
to
have
them
or
my
co-author
max.
Here
raise
your
hand.
L
A
A
I
I
I
We're
all
very
used
to
thinking
about
performance
in
terms
of
throughput.
Everybody
knows
how
many
megabits,
their
home
internet
connection,
is,
and
that's
good
for
big
downloads,
but
for
everything
else
that
we
do
for
maps
driving
directions,
weather
forecasts
every
time,
you're
sitting
watching
an
app
with
some
little
spinny
wheel,
waiting
waiting
waiting.
It's
not
because
you
don't
have
enough
bandwidth!
I
It's
because
you
have
too
much
delay.
This
affects
lots
of
things
that
are
very
obvious,
like
online
gaming,
video
conferencing,
it
affects
many
other
things,
even
if
you're
watching
a
two-hour
streaming
video.
If
you
want
to
skip
to
another
chapter,
then
suddenly
it's
the
latency
that
affects
the
performance
of
that
okay.
I
We
present
this
in
round
trips
per
minute,
because
most
people
like
metrics,
where
more
is
better
and
milliseconds,
is
a
fairly
abstract
concept
to
a
lot
of
people.
Even
for
us
in
this
room,
how
much
is
a
millisecond?
Well,
it's
too
small
to
notice.
Does
it
really
matter
so
we
wanted
a
number
that
is
more
people
relate
to
more
easily
next,
please.
I
Since
last
year,
waveform
has
put
out
a
really
good
buffer
bloat
test.
Professor
william
hawkins
from
university
of
cincinnati
has
got
an
open
source
implementation
in
go
and
more
recently,
ookla
makers
are
probably
the
most
popular
speed
test.
App
and
website
on
the
internet
have
added
a
measure
of
working
latency.
I
I
I
So
we
know
there's
a
problem:
what
are
we
going
to
do
about
it?
The
answer,
in
my
opinion,
is
l4s.
It's
a
new
method
of
doing
congestion,
control
in
partnership
between
the
end
points
and
the
network.
It's
called
low,
latency,
low
loss,
scalable
throughput,
and
I
can't
explain
all
of
this
in
one
slide,
but
the
summary
is
it
works
to
keep
q
short
by
keeping
q
short.
We
keep
round
trip
delay
short
as
soon
as
a
queue
starts
to
build
up
at
the
bottleneck
link.
I
It
uses
ecn
marks
to
tell
the
sender
to
slow
down
next
one.
Please
we
had
a
lot
of
representation
at
the
hackathon.
Let's
move
through
these
pictures
quickly
go
ahead.
Next,
please,
we
had
two
cmts
racks
set
up.
Next
we
had,
we
took
over
a
big
chunk,
go
on
next
one,
please.
We
took
over
a
big
chunk
of
the
hackathon
room
next
couple
of
slides.
Please,
we
had
a
whole
bunch
of
equipment,
piled
up
on
desks
a
lot
of
energy,
so
to
learn
more
there's
a
great
buy
tag
report.
I
If
people
ask
you
about,
latency
you'd
only
spend
an
hour
explaining
it
yourself
pointing
to
this
report.
If
you
prefer
video,
there's
an
apple
developer
conference
presentation
about
15
minutes
long
by
my
colleague,
vidi
goyle
who's
here
today
and
last
slide.
Please
come
to,
can
you
go
to
the
last
live?
Please
can.