►
From YouTube: libp2p in 2018 - Juan Benet
Description
Originally recorded during the Berlin Developers Meetings from July 9-13, 2018.
A
Okay,
great
hey
everyone,
so
thank
you
to
the
VP
for
that
awesome,
introduction
to
the
poppy
in
general.
What
I
wanted
to
talk
about
was
focus
on
where
were
apps
as
it
projects
in
great
part,
in
how
we
develop
some
of
this
stuff
and
think
through
the
priorities
that
we
have
ahead,
so
that
we
can
set
up
ourselves
to
pretty
much
scale
it
a
lot
of
development
and
get
to
solve
some
of
those
really
really
hard
hard
issues
right
so
we're
talking
about
in
in
the
long
term.
A
We
hope
to
bring
about
drastic
improvements
to
how
people
use
networking
in
their
applications
and
platforms
and
systems,
and
it's
totally
achievable,
but
it
will
require
a
lot
of
work
on
very
specific
areas
that
we're
getting
ourselves
set
up
to
do,
but
we're
not
fully
there
yet
so
I
want
to
talk
a
bit
about
how
we
can
we
can
get
there.
I
wouldn't
touch
a
bit
on
kind
of
vision
and
mission.
I
think
they've
already
covered
this
excessively,
so
I
can't
don't
have
to
spend
a
lot
of
time
on
this
at
all.
A
This
shared
a
lot
with
ipfs,
just
simply
because
this
is
where
a
lot
of
the
project
came
out
of,
but
over
time
it's
gonna
become
kind
of
its
own
identity,
but
I
think
this
slide
still
captures
that
the
idea
is
really
well.
We
want
a
networking
stack,
the
supports
the
end,
use
the
end
style
distributed
environments
where
you're
not
really,
depending
on
any
kind
of
intermediaries
at
all,
so
the
internet
works
that
way
the
web
doesn't
and
then
a
lot
of
other
platforms.
A
If
right
now,
there
was
a
major
disaster
in
a
city,
a
lot
of
the
normal
communication
applications
that
people
use
to
talk
to
each
other
would
suddenly
stop
working.
A
lot
of
the
normal,
centralized
messaging
systems
depend
highly
on
a
set
of
conditions
working
and
if
any
one
of
those
links
in
the
chain
breaks,
the
entire
system
comes
down.
So
that's
not
how
the
internet
was
designed.
A
That's
how
modern
web
applications
and
mobile
systems
work,
and
so
we
need
to
go
back
back
to
the
old
designs
and
you
know
kind
of
like
unwind
and
go
down
a
different
path,
and
so
a
lot
of
that
is
thinking
through
the
hard
problems
in
building
those
large-scale
platforms
to
work
like
the
internet,
which
is
much
harder
and
then
think
through.
How
can
we
build
software
libraries
to
enable
those
applications
and
systems
to
do
it
easily?
A
It's
very
easy
today
to
build
and
depend
on
cloud
systems
and
to
not
do
those
normal
architectures,
but
it
wasn't
easy
at
the
beginning,
people
put
in
a
lot
of
work
and
effort
to
make
all
of
those
libraries
so
that
it
is
easy.
Today,
we
just
have
to
do
that
before
the
peer-to-peer
type
of
things
worth
just
keep
this
in
mind.
I
think
people
were
around
the
last
few
days
already
kind
of
sell
this,
but
just
keep
this
in
mind,
as
you
think
about
the
PvP
in
general.
A
Think
about
the
probably
vast
majority
of
the
internet
doesn't
have
the
level
of
connectivity
that
the
main
cities
in
the
world
have
it's
highly
mobile
and
that's
something
that
we
need
to
make
more
of
a
focus
in
our
projects.
And
then
we
don't
really
describe
these
the
this
set
of
problems
and
so
on
or
recover.
All
of
this.
These
are
the
all
the
web
problems
from
a
profess.
A
lot
of
them
are
just
about
the
networking
stack
and
how
data
is
structured.
A
On
top
of
that
platform,
same
thing,
they're-
probably
a
lot
more
projects
here,
but
lipid
apiece,
gaining
in
in
kind
of,
is
growing
as
a
system
that
people
depend
on,
and
so
we
hope
to
you
know
a
year
have
that
grid
expand
quite
a
bit
there's
already
a
lot
of
that
are
not
represented
there,
but
more
will
grow.
We
wanted
to
mention
that
there
was
kind
of
like
an
inversion,
interestingly,
where,
where
I
peeled
II
suddenly
might
actually
become
useful
tool
if
it
appears
as
a
library.
A
This
is
you
know
an
attempt
at
a
mission,
not
quite
not
quite
there
yet
but
we'll
think
through.
We
keep
workshopping
this
I
hope,
but
really
think
of
making
a
networking
stack.
A
modular
networking
stack
to
make
robust
peer-to-peer
systems
easy
to
build.
It
is
not
easy
to
build
robust
peer-to-peer
systems
today
at
all,
so
things
like
process
addressing
has
to
be
covered.
I.
A
So
when
people
think
about
implementing
something
like
a
choral
DHD
or
you
know
some
of
the
newer
DHD
that
have
been
designed
in
the
last
five
years,
they
can
just
focus
on
that
as
a
problem
and
use
whatever
toolkit.
There
is
in
lipid
EP
but
then
provide
whatever
interface
they
need
to,
so
that
other
systems
can
start
depending
on
that
and
right
now.
Development
in
lipid
V
is
highly
integrated
across
a
lot
of
things,
and
so
it's
kind
of
difficult
for
for
people
to
do
that.
But
it's
getting
a
lot
better.
A
So
I
think
this
year
is
way
better
than
last
year
and
I.
Think
next
year,
it'll
be
even
easier.
I
think
I'd
be
interested
in
talking
to
people
around
this
in
the
next
couple
of
days
about
if
you
implemented
a
specific
version
of
this,
for
example
the
quick
transport.
How
much
did
you
have
to
worry
about
the
rest
of
the
lipid
EP
interfaces?
A
So
I
wanted
to
show
some
highlights.
We
split
out
planning
from
from
ipfs,
so
we've
been
on
this
like
fun,
going
effort
to
like
split
out
the
library
and
and
so
on,
and
so
it
required
effort
to
just
make
sure
that,
like
really
they're,
distinct
projects
and
planning
is
happening
separately
and,
though,
there's
a
lot
of
integration
and
sharing
of
people.
We
really
have
two
two
different
threads
and
this
will
help,
especially
all
the
other
people
that
are
coming
to
depend
on
p2p,
to
participate
in
the
p2p
ecosystem.
A
Without
having
to
worry
about
understanding
or
thinking
about
how
a
BFS
organizes
itself
implementations,
it's
exciting
to
have
rust
added
to
the
sets,
this
was
over
the
last
year
and
I
think
also
over
the
last
year,
Jas
got
to
a
level
of
connectivity
and
support
that
the
remarkable
improvements,
especially
in
the
browser
and
so
on,
and
go
got
massively
more
performant
and
also
got
quick.
So
that's
a
you'll
hear
about
more
more
of
that
later
on.
A
A
No
I'll
fix
it
later,
one
waste
your
time,
so
you
know
we
have
these
kind
of
working
groups
budding.
There
may
be
more
later.
I'll
talk
about
some
of
some
of
them.
So
with
that
I
wanted
to
talk
about
what
we
should
think
about
focusing
on
in
the
next
couple
of
quarters,
and
my
camp
has
already
done
a
lot
of
thinking
and
and
structuring
of
this
and
he'll
talk
about
later.
A
And
you
know
each
of
these
dimensions
like
features
break
out
into
ton
more,
and
you
know
you
need
you
just
growing
the
the
volume
of
cover
over
all
of
these
different
dimensions.
When
you
are
building
a
software,
it
just
consumes
a
lot
of
time
and
so
getting
to
a
really
good
level
of
completion
with
on
any
of
these
dimensions.
It
takes
a
while,
and
so
if
you
want
to
ship
something
you
have
to
like,
take
a
lot
of
shortcuts,
and
so
we
did
that.
A
Then
you
know
we
got
things
usable
and
and
dependable
to
some
degree.
But
now
we
need
to
keep
rolling
that
up
and
keep
diving
deep
into
each
of
these
dimensions
and
improving
things
so
leveling
up
or
docks,
leveling
up
or
future
sets
leveling
up
testing
and
QA
leveling
up
performance
level,
specs
and
so
on,
and
this
isn't
a
you
know.
We
work
on
this
for
a
while
and-
and
you
know
we
get
it
to
the
you
will
max
it
out
all
the
way
to
the
end
and
like
we're
set.
A
A
One
other
thing
is
that,
because
lipid
apiece
highly
modular,
we
could
each
of
the
different
components
whether
it's
different
implementations
or
different
sub
pieces
can
proceed
on
their
own
at
their
own
pace
and
their
own
rate
to
some
degree,
except
for
the
things
that
are
dependent
on
as
a
group.
So,
for
example,
things
like
interfaces
things
like
testing
things
like
documentation
and
examples,
and
communications
and
websites,
and
and
so
on.
A
We've
gotta
get
getting
that
also
useful
to
think
about
the
Colonel
shell
and
commands
kind
of
structure
from
Unix.
This
is
pretty
similar
to
how
the
PDP
works
right,
so
you
can
think
of
the
the
kernel
of
Lapita
P
being
the
addressing
and
the
switching
both
right
now
circuit,
searching
in
the
future
we
might
bring.
A
But
but
again
this
the
coupling
makes
it
easy
for
us
to
proceed
in
development
in
one
area
without
having
to
worry
about
the
rest
of
the
project.
I
think
right
now
we're
not
in
a
good
we're
in--okay
place,
but
we
need
to
get
better
so
levelling
up
with
a
lot
of
things
and
scaling
our
development
so
yeah.
You
know
one
objective
that
we
have
for
2018
is
making
a
lipid
be
a
first
class
project.
That
means
you
know
some
working
group
structure
or
a
core
working
group
to
kind
of
coordinate.
A
All
of
it
clearly
articulated
requirements
for
different
implementations,
a
great
roadmap,
that's
something
we're
we're
working
on
and
also
increasing
the
the
full
the
amount
of
full-time
dedicated
people.
Then
this
around
quality
dependability.
This
I
want
to
talk
more
with
people
about
this,
but
just
leveling
up
our
our
work
on
these
kinds
of
things.
You
know
getting
correctness,
higher
performance,
completeness,
scalability
and
so
on
and
I
want
to
touch
on
a
couple
of
things
here.
One
is
visualizing.
Protocols
can
become
extremely
useful
to
understand
how
to
how
to
work
with
them.
A
So
one
project,
I'm
I'm
really
excited
to
think
about
during
this
year
and
either
doing
this
year
or
next
year.
It's
building
a
structure
for
people
to
be
able
to
visualize
things
like
a
DHT
or
things
like
the
stream
multiplexers
and
so
on.
Here's
an
example
of
a
visualization
of
a
protocol.
This
is
spot
point
and
this
is
a
visualization.
That's
generated
straight
from
the
log
output
of
a
point
notes.
A
So
those
are
six
back
windows,
they're
all
running
a
little
bit
of
heat,
and
we
take
the
log
output
of
those
and
consumed
create
those
logs
and
then
just
visualize
logs,
and
so
we
get
a
trace
of
what
it
looks
like
to
to
have
a
protocol
run,
and
it's
built
in
such
a
way
that
this
visualization
is
built
in
such
a
way
that
we
can
just
write
all
of
those
log
outputs
to
a
file
and
then
just
play
them
back
in
the
future.
Now
you
can
imagine
things
like
controls
here.
A
Being
yours
like
free,
is
somewhere
and
be
like.
Oh,
what's
happening
here
like
what
is
this
weird
message
that
are
like
I'd,
never
expected
and
there's
some
amount
of
like
custom
visualization
they
have
to
do
per
protocol
like
this
is
a
visualization
of
falcone,
where
you
have
a
chain,
and
you
have
like
the
market,
the
storage
market
structure
and
so
on.
Something
like
a
DHT
would
look
pretty
different,
but
some
of
the
components
might
be
reusable
and
at
least
the
the
interest.
The
architecture
of
the
you
know.
A
A
The
other
thing
around
testing
is
that
we
need
to
start
doing
really
large-scale
tests
like
we
need
to
get
to
a
point
where
we
are
testing
millions
of
nodes
and
get
ourselves
all
set
up,
so
we
can
do
testing
with
billions
of
notes.
This
is
planetlab
I,
don't
know
if
you're
familiar
with
it.
It's
a
large
scale,
testing
environment
that
I
kima
built
a
couple
decades
ago,
and
more
most
academic
peer
to
peer
systems
were
tested
on
this
things
like
coral
things
like
academic,
probably
and
so
on,
and
we
also
have
another
version
of
this.
A
So
this
is
another
project
that
we
might
consider
doing,
maybe
this
year,
potentially
next
year
around
getting
our
own
testing
infrastructure
to
a
point
where
it
is
easy
for
somebody,
who's,
writing
a
new
DHT
to
say:
hey,
you
know
what
I
want
to
test
out
how
this
is
new.
We
think
this
thing
works
really
well
in
simulation.
A
You
know
this
pipeline
of
going
from
ideas
through
specs,
encoding
and
so
on,
and
a
lot
of
lip
yupi
work
is
pushing
the
research
boundaries
and
is
pushing
the
frontiers
of
what
we
know
about
peer-to-peer
systems
and
already
today
we
do
a
lot
of
stuff
that
we
need
to
kind
of
inform
the
world
about.
So
this
might
mean
writing
papers.
A
This
might
mean
collaborations,
with
with
other
groups,
there's
an
enormous
amount
of
interest
in
peer
to
peer
and
decentralized
systems.
That
goes
all
the
way
back
to
you
know,
late
90s
and
early
2000s
and
in
some
cases
much
further
back,
but
there's
a
lot.
There
was
a
lot
of
important
activity
happening
across
a
bunch
of
universities
in
that
time
period
and
that
a
lot
of
that
just
fell
flat.
It
was
used
in
the
early
period,
appear
applications
of
the
unifor.
A
You
know
the
first
peer-to-peer
wave,
but
with
the
kind
of
there's
a
whole
bunch
of
reasons
why?
But
there
was
effectively
a
peer-to-peer
winter
and
a
lot
of
those
those
groups
stopped
working
on
this
stuff,
but
many
of
their
their
implementations
and
solutions
were
extremely
good.
They
just
didn't
get
built.
They
died
somewhere
here
right.
A
So
a
lot
of
the
really
good
peer
to
peer
ideas
and
things
that
are
going
to
make
peer
to
peer
system
scale
are
just
you
know,
stuck
in
the
ideas
or
or
paper
domain
where
there
were
great
results
achieved,
and
then
those
groups
moved
on
to
other
things.
So
it's
time
for
us
to
go
and
and
figure
out
what
those
might
be
and
bring
them
in
like
actually
get
get
them
to
a
point
where
you
can
use
them
and
another
cent
in
other
places.
A
It's
connecting
with
the
people
that
we're
doing
that
work
and
saying
hey.
We
now
have
a
good
structure
to
build
and
test
all
of
these
things,
and
you
know
yes,
the
Entenmann's
principle
got
violated
when
we
like
made
the
TCP
and
UDP
protocols,
be
like
the
only
things
and
like
you
know,
you
couldn't
really
make
new
transfer
protocols,
but
we're
working
on
like
fixing
that
and
like
now
it
with
low
p2p.
You
can
totally
deploy
a
new
transfer
protocol
and
you
don't
have
to
worry
about.
A
You
know
you
just
overlay
it
over
UDP
in
it
and
you
can
build
applications
on
top
of
it.
So
all
of
that
kind
of
stuff
can
revive
an
enormous
amount
of
activity
in
all
over
those
groups,
because
if
we
give
them
an
easy
way
to
test
out
implementations
and
then
ship
them
into
real
live
applications,
that's
gonna
generate
a
ton
of
activity
in
this
world,
and
so
this
is
something
we
should
think
about
this.
A
So
part
of
that's
part
of
the
goal
of
this
event
to
understand
what,
as
a
user
live
to
be,
what
do
you
need
out
of
it,
and
you
know,
build
a
registered
registry
to
track
those
use
cases
and
means,
and
so
on,
so
that
when
we
tackle
those
things
in
a
quarterly
basis,
we
figure
out
what
to
what
to
prioritize
on
what
to
do,
and
another
thing
I
want
to
talk
about
is
is
scaling
up
development,
so
we've
already
started
to
form
working
groups.
We
want
to
double
down
on
that.
A
We
want
to
sort
of
measuring
what
we're
doing
right
and
wrong
there,
better
planning
and
scheduling,
and
all
that
kind
of
stuff
increasing
the
resources
we
are
doing.
Things
like
the
journal,
Club
principle
I
think
we
covered
coral.
We
covered
a
few
other
things
that
are
relevant
to
peer
to
peer.
If
you
want
to
get
involved
in
that,
please
do
and
I
wanted
to
touch
on
how
to
speed
up
our
development,
so
I
think
different
from
other
systems
like
ipfs
and
others
were
a
lot
of
the
system.
A
Is
it
needs
to
be
highly
integrated
other
than
the
p2p
core
or
kernel?
You
don't
really
have
to
think
very
much
about
syncing
with
other
groups,
while
you're
billing
building
an
independent
implementation
of
a
subset
of
these
these
protocols.
So
you
know,
each
of
these
things
can
proceed
in
parallel.
I
kind
of
follow
its
own
own
timeline
and
often
doesn't
require
the
context
of
understanding
of
getting
everything
else.
A
So,
instead
of
thinking
of
little
to
be
advancing
as
a
you
know,
unit
with
you
know
a
group
of
people
having
to
have
high
contexts
and
everything,
it's
really
a
bunch
of
little
modules,
advancing
at
their
own
rate,
and
once
you
know
some
of
them
reach
a
certain
level
of
completion,
then
they
get
pulled
into
like
the
implement
the
working
implementations.
So
this
suggests
that
we
can,
you
know,
organize
our
work
around
these
things
much
better.
In
that
we
could.
We
could
describe
things
like
hey.
A
We
need
a
Bluetooth
transport
and
here's
what
you
two
look
like:
here's,
the
interface
it
needs
to
match
and
then
kind
of
put
that
out
as
a
as
a
project,
description
that
that
needs
to
happen,
and
then
people
that
want
that
thing
or
that
might
be
good
at
building.
That
thing
can
just
focus
on
that
and
not
have
to
worry
about
all
the
rest
of
the
stuff.
That's
mostly
documentation
and
project
management,
stuff
that
that
we
haven't
gone
there
with,
but
I
think.
A
If
we
set
ourselves
up
in
this
cord
to
do
this,
then
we're
gonna
be
able
to
scale
the
development
pretty
dramatically,
and
you
think
of
this
is
becoming
also
RFP
worthy
stuff.
Where
we
can
say
you
know
what
we
really
need:
a
visual
transports,
because
I
want
to
like
aim:
two
phones
up
each
other
and
like
visual,
like
QR
codes.
It's
like
the
only
thing
that,
like
I,
can
use
in
this
crazy
setting,
but
I
don't
really
have
the
time
to
go.
A
Do
this
so
I'm
gonna,
like
yeah,
hire
some
other
group
that
can
do
this.
That
doesn't
know
that
much
about
loop.
It
would
be
necessarily
but
have
a
well
scope
than
a
project
that
they
can
dive
into
the
code
and
build
that
one
thing
and
then
and
then
go
off
and
I
think
this.
If
we,
if
we
do
this
well,
then
we
can
scale
the
development
of
all
the
stuff
very
effectively
in
the
next
in
the
coming
year.
A
So
you
know
here
are
some
ideas
of
working
groups.
We
might,
we
might
form,
there's
more,
but
things
like
testing
and
visualization
testing
here
being
like
large-scale,
now
we're
testing
and
then
starting
to
think
about
use
cases.
So
let
me
to
keep
lipid
b4
blockchains
lipid,
b4,
browsers,
OS,
filesystems
mobile
there's
a
lot
of
things
there.
Those
were
five
that
I
think
are
high
in
my
mind
and
we
are
we're
gonna
be
trying
out
a
thing
around
projects
and
I
think
Michael
touched
on
this
more
later,
but
or
on
earth.
A
A
A
That
kind
of
like
project
management
structure
will
will
again
help
us
parent
license
scale
the
development
process
where,
because
all
of
the
lipid
B
modules
are
so
many,
you
don't
need
a
lot
of
people
working
on
some
specific
things
won't
need
all
that
context,
and
so
this
is
why
we're
doing
this
is
what
we're
experimenting
with
this
kind
of
structure
and
the
last
thing
I'll
mention
is
hey.
We
should
have
a
lot
of
fun.
A
Damn
it
a
lot
of
us
hack
on
these
things,
because
it's
pretty
awesome
to
do
some
of
the
stuff
like
being
able
to
build
like
it,
creates
a
network
in
the
browser
that
could,
you
know,
have
millions
of
browsers
connected
to
each
other
and
have
you're
playing
a
game
or
something
that
kind
of
awesome.
Hacks
I
think
we
need
to
do
more
of.
A
Let
p2p
is
now
at
a
point
to
support
that
kind
of
stuff,
I
love,
seeing
like
random
things
like
this,
where
you
know
somebody
will
post
a
link
and
they'll,
be
like
this
crazy.
You
know
to
test
out
of
the
HD
implementation.
You'd
leave
your
browser
open,
and
you
know
it
starts
doing
a
bunch
of
requests
and
it
could
maybe
you
can
visualize
a
network,
and
then
this
is
a
way
that
we
can
get
also
a
lot
more
people
engaged
and
and
thinking
about
the
the
cool
things
that
can
happen
with
these.
A
These
systems
so
I
think
we
should
just
put
a
bit
more
time
and
effort
into
into
letting
you
know
doing
this
kind
of
stuff
like
it's
wow.
It
sounds
like
I'm,
saying:
hey
work
hard
on
having
fun,
but
but
really
just
about
celebrating
more
what
what
we're
doing
and
yeah
telling
the
world
about
it
all
right,
so
I
think
that
and
and
I
wanted
to
touch
on
kind
of
what
protocol
so
we're
thinking
through.
How
do
we
interface
like
organist,
there's
like
the
project
and
this
organizations
that
are
doing
work
with
the
project?
A
B
B
Hye-Won
I
have
a
question
about
I,
guess
just
a
pipian
ipfs,
like
philosophy,
you
know,
w
talked
about
the
PTP
being
like
a
location.
Oh
sorry,
no
content
addressed
assistant
means
have
a
location
address
system.
Do
you
guys
see
a
way
to
kind
of
still
attach
location
addressing
somehow
into
the
system,
or
is
this
kind
of
just
like
think?
B
A
I,
guess
is
the
question:
are
we
open
to
exploring
new
kinds
of
ways
of
doing
networking
where
you're
doing
the
addressing,
by
other
things
like
yeah
like
geolocation,
or
something
like
that?
Don't
sound,
awesome
and
interesting
and
we
should
explore
them.
No
question:
I.
Think
what
where
we're
at
now
is
the
the
core
lowest
level
of
the
p2p
like.
Why
is
it
a
ways
that
a
circuit
switch
thing
and
not
a
packet
switch
thing?
A
A
That
should
be
a
circuit
switched
Network
to
begin
with,
but
packet
switching
is
like
is
coming
right
and
so
similarly
other
kinds
of
networking
ideas,
like
perhaps
some
of
the
content,
centric
networking
stuff
or
come
to
an
address,
networking
stuff
back
from
ndn
and
and
other
you
know
x,
IA
I,
don't
know
if
you've
heard
about
these
systems
but
they're
all
remanded
imaginings
of.
If
you
were
to
build
the
Internet
today
from
scratch,
like
the.
A
If
you
were
to
redesign
the
network
layer
of
the
Internet
like
the
IP
layer
of
the
Internet,
would
you
do
it
by
just
doing
these
processes
address
or
these
host
addresses?
Or
would
you
do
it
with
something
completely
different,
and
we
should
absolutely
explore
that
kind
of
stuff?
I
think
it
would
be
that's
kind
of
like
a
track,
and
that
would
be
kind
of
a
working
group
idea.
A
If
there's
enough
people
interested,
we
can
form
those
kinds
of
working
groups,
but
yeah
it'll
be
demand,
driven,
as
opposed
to
you
know
explicitly
forcing
forcing
yourself
to
go
in
that
direction
a
priori.
A
big
priority
for
us
is
to
you
know
what,
if
is
already
used
by
a
lot
of
groups,
and
so
we
want
to
get.
We
want
to
make
sure
that
the
the
a
large
amount
of
our
bandwidth
goes
into
supporting
those
those
users
but
yeah.
We
should
explore
that.