►
From YouTube: Building a distributed video distribution platform
Description
In this presentation from Day 2 of the #SwarmOrangeSummit, Jelle Gerbrandy from Paratii presents their decentralised video platform in the making that lets users share video content without the middleman and get rewarded for it.
A
Okay,
guys
hello,
I'm
yeah,
like
a
blondie,
I
work
for
a
parity
and
we're
making
a
decentralized
video
platform
just
want
to
agree
with
Victor.
Please
interrupt
me
and
it's
much
nicer
to
discuss
them
to
just
listen
to
me.
Okay
and
so
I'm
gonna
do
two
things
I'm
going
to
explain
to
you
very
quickly
what
our
project
is
about
and
then
I
have
some
like.
There
is
remarks
that
I
think
may
be
relevant
for
this
warm
community
for
from
what
we've
encountered.
A
Yes,
I
think
I
can
okay,
so
the
project
is
parity,
we're
doing
it.
We
have
a
team
of
about
eight
to
ten
people,
now,
half
of
them
in
Brazil,
half
of
them
in
Europe,
there's
one
guy
in
Egypt
and
so
from
say
the
PR
kind
of
side.
It
looks
like
this,
so
what
we
are
going
to
do
it's
in
Portuguese,
but
go
get
it
so
decentralized.
Video
platform
means
the
monetization
is
in
your
hands.
It
means
there's
no
intermediaries
and
there's
even
it's
the
content
that
you
want
to
see.
A
So,
basically,
so
the
project
is,
we
would
like
to
make
to
build
a
decentralized
YouTube
using
the
various
tools
and
various
projects
that
are
in
the
crypto
and
specifically
in
the
ethereum
ecosystem,
all
right
and
so
in
practice.
What
we
need
to
build
is
something
like
this.
So,
first
of
all,
we
have
to
build
like
the
video
pipeline,
which
basically
means
that
we
have
to
make
a
building
is
system.
What
you
do
and
YouTube
you
upload
a
video.
A
The
video
will
need
to
be
transcoded.
Eric
just
explained
perfectly
what
that
means.
It
needs
to
be
persistent
right.
We
need
to
store
it
somewhere.
This
is
why
I'm
here
we
need
to
make
a
discoverable.
So
we
need
to
have
this.
We
have
this
a
bunch
of
videos
and
we
need
our
users
to
be
able
to
find
them,
so
they
need
to
be
indexed
somehow,
and
then
we
need
to
deliver
them
to
the
users
in
a
fairly
fast
way.
A
So
don't
have
to
weigh
too
much
and
then
we
need
a
bunch
of
applications
to
actually
view
the
video
and
video
player
portal
stuff.
Okay,
so
that
is
like
the
basic
infrastructure.
We
need
to
pay
for
this
infrastructure
right
I
mean
the
whole
point
of
one
of
the
motors
behind
platforms
like
YouTube
is
that
people
can
actually
make
money
on
that
work
with
in
different
way.
So
once
we
have
this,
we
want
to.
We
are
experimenting
with
ideas
about
how
to
monetize
this
stuff.
A
We
want
to
make
an
open
platform
where
users
can
choose
their
way
of
monetization
and
a
third
point
which
is
particularly
relevant
not
only
now
in
this
Facebook
disaster
time
that
it
does
it
because
we're
all
building
this
on
open
a
theorem
system,
the
privacy
anonyme
anonymity
issue
becomes
like
a
really
difficult
thing,
because
basically
we
have
all
our
date
in
the
open
and
we
need
to
hide
them
somehow
and
okay.
So
we've
been
working
about
on
this
for
about
a
year
now,
and
so
we,
of
course
we
don't
have
this
right.
A
I
mean
it's
enormous,
this
project,
but
what
we
do
have
is
like
a
first
version
of
this
video
pipeline.
I
will
tell
you
what
it
we
are
specifically
later
and
we
have
made
written
this
like
generic
library,
which
basically
gives
us
a
uniform
interface
to
to
interact
with
which
our
applications
can
be
directly
two
different
pieces
of
the
pipeline,
and
we
do
have
a
web
player
and
we
have
a
portal
and
radiator
pilots
with
20
uploaders.
We
have
250
videos,
and
this
week
we
should
open
it
up
a
bit
more.
A
A
A
So
this
is
what,
but
it
looks
like
now.
So
these
are
the
basic
steps
of
what
we
need
to
do
with
videos.
So
first
they
need
to
be
transcoded.
Then
we
need
to
store
them
somewhere.
We
need
to
index
them.
They
need
to
be
curated
in
a
sense
that
we
need
to
organize
the
videos
in
playlists.
We
need
to
prop
beef.
A
If
we
put
them
on
our
branded
portal,
we
need
to
censor
them,
get
rid
of
the
child
porn
and
we
have
to
deliver
them
to
the
users,
and
so
this
is
what
we
have
now,
but
we've
built,
because
the
idea
was
we're
gonna
build
some
stuff
with
whatever
tools
we
can
found
and
then
try
to
decentralize
all
the
pieces
step
by
step,
and
so
for
the
transcoder.
We
was
just
our
own
friends
collar,
which
is
simple,
but
we
are
looking
very
much
at
life
beer.
A
A
There's
some
differences
in
requirements
because
live
beer
is
doing
live
streaming,
which
is
you
can
ask
if
you're
interested
okay
for
the
persistence
layer,
so
these
are
basically
two
parts,
so
we
need
to
store
the
videos
themselves
and
we
need
to
store
data
about
the
videos
right
yeah,
whose
is
it
what's
the
title
if
you
want
to
get
paid
for
it
stuff
like
that,
and
so
what
we're
using
is
ipfs
and
I
will
explain
to
you
why
we're
not
using
swamp
in
the
next
slide,
so
the
videos
are
stored
in
ipfs
in
the
data
it's
partly
start
in.
A
Basically,
we
store
video
records
on
ethereum
with
a
reference
to
a
hessian
ipfs,
where
all
the
real
data
are.
This
is
flexible.
It
worked
well
and
we
would
have
liked
to
use
the
elements
one
I'll
get
to
that
and
for
the
indexing.
So
I
think
this
is
a
real
interesting
point
here,
and
so
you
can
store
date
on
a
blockchain
right.
It's
basically
a
database,
but
you
cannot
really
search
very
well
on
the
blockchain,
and
so
what
you
need
to
do
is
index
the
data
right.
A
You
need
to
make
a
blockchain,
Explorer
type
of
thing,
and
so
I,
don't
know
of
any
I
mean
I.
Don't
know
how
to
do
this.
Of
course
we
can
just
put
the
database
index
stuffing
and
do
this,
but
I
don't
know
of
any
decentralized
solutions
for
that
any
mature
projects-
and
this
is
something
if
you
guys
know
of
something
I-
would
be
very
happy
to
learn
about
that,
and
the
curation
parts
are
now
basically
not
doing
anything.
A
A
For
the
delivery,
so
for
actually
getting
the
video
studio,
sirs
we're
using
IP
FS
and
we
were
hoping
to
only
use
that.
But
we
need
a
gateway
because
it's
yeah,
it's
just
not
fast
enough
and
also
there,
where
you
would
like
to
use
warm.
So
why
didn't
we
use
Swan,
okay,
but
the
basic
reason
is
they
so
one
of
our
first
requirements
or
the
first
ideas
was
that.
A
A
Most
of
our
new
nodes
will
be
video
players,
people
watching
stuff
in
the
browser,
okay
and
so
forth.
That
idea
to
work
we
would
have.
We
have
to
instantiate
a
node
in
the
browser
right
and
so
ipfs
has
a
JavaScript
implementation
and
has
implemented
transport
web
RTC
transport
layer.
So
with
some
tweaking
I
mean
it
was
not
immediate.
A
We
actually
have
now
an
IPF
as
note
which
runs
in
a
browser.
So
this
means
that
why
why
this
screenshot
is
so
interesting,
because
here,
what
we're
actually
doing
is
if
you
tweet
one
of
our
videos,
you
know
player
and
you
open
your
Twitter
feed
in
your
browser.
This
will
instantiate
an
IP
FS
node,
which
will
both
now
download
and
surf
content,
and
I
mean
this
is
really
cool.
A
A
B
B
A
A
Okay,
so
so
this
diagram,
I'm
gonna,
go
over
it
it
very
quickly.
It's
it's
because
it's
more
meant
to
illustrate
how
messy
our
our
stuff
is
going
to
be
and
how
much
better
it
could
be,
and
so
basically,
what
happens
in
this
stuff-
and
this
in
this
video
pipeline
is
this.
A
This
network,
where
which
is
messy
and
most
on
the
mostly
on
terms
of
I,
mean
all
these
steps
have
to
be
done.
That
kind
of
messy
we
cannot
avoid,
but
we
are
using
different,
like
communication
protocols
and
systems
all
over
the
place.
So,
for
example,
here
within
the
ipfs
nodes,
what
we're
basically
using
is
an
adaptive
version
of
the
bit
swap
protocol
because
bad
these
nodes-
these
are
all
ipfs
nodes.
They
are
communicating
anyway,
so.
B
A
Added
a
little
layer,
some
messaging
on
top
for
yeah,
say
therefore
for
saying
I
would
like
to
transfer
all
this
job
or
I've
done
that
kind
of
thing.
Well,
of
course,
the
communication
with
etherium
not
goes
over
HTTP
and
over
WebSockets
the
index.
This
is
just
HTTP
stuff,
and
so
what
we
find
ourselves
is
getting
this
system,
which
is
getting
more
complex
and
complex,
and
I
wrote
it
a
little
bit
small,
but
I
mean
the
vision
of
the
of
the
Wed
3e
Tyrian
stack,
which
combines.
A
A
Communication
between
browsers
of
a
wet
RTC
is
very
heavily
throttled.
If
you
start
sending
more
data
than
I
mean
we
didn't
do
any
systematic
experiments,
but
it's
like
like,
if
you
start
considering
sending
like
more
than
5
mega
bytes
of
data,
which
is
very
little
to
the
saint-pierre.
It
will
just
stop
stopping
it.
I
mean
wouldn't
work
around
that
and.
A
So
am
I
yes,
I
have
to
explain
also
so
the
person
who
would
have
should
have
given
this
talk
is
I
will
go
through
the
last
slide.
So
you
can
see
him
this
guy
yeah
and
he's
got
Muhammad
ayah,
but
he
cannot
come
here
because
he's
living
in
Egypt.
These
are
like
this
is
during
the
low-level
peer-to-peer
work
and
he
would
have
answered
your
question
perfectly,
but
we
live
in
a
limited
world,
I
mean
so
we
ran
into
this
day.
A
Well,
actually,
Japan
they're,
dropping
like
crazy,
and
so
I
mean
the
point
I
want
to
make
is
that
projects
like
like
ours,
which
is
very
specific
about
video
of
it's
warm
in
general,
and
it
kind
of
has
to
face
this
at
a
certain
point
right.
If
we
get
browser
makers
or
ISPs
who
somehow
don't
like
all
this
peer-to-peer
stuff,
we
can
build
all
the
queue
technology
we
want,
but
we
will
be
killed.
B
A
Mean
I
think
I
mean
yeah
I
mean
if
you
had
AB.
If
you
have
a
like
a
free
functioning
market,
you
would
say:
okay,
if
consumers
want
beer
to
beer
stuff,
they
will
get
it
from
there.
Nice
piece
right,
but
we
don't
have
that
kind
of
market
in
a
big
part
of
the
world.
Like
sight
again,
my
friend
Mohammed
III
is
in
Egypt.
There
is
no
question
that
consumers
can
influence
the
behavior
of
ISPs
no,
but
also
in
the
States
I
mean
they're
very
they're.
Also
very
many
other
interests.
B
A
Are
you
going
to
implement
swarm
in
the
browser
and
they
said
no,
because
they're,
very
shitty
peers
and
so
I
can
confirm.
They're
really
shitty
peers
I
mean
this
all
and
people
watching
a
video
the
world
all
right.
They
will
download
stuff
and
they
will
finish
the
video
and
go
away
right,
and
we
have
that.
The
fact
that
there,
in
the
browser
I
mean
it's
just
like
lots
of
limitations
for
programming
in
the
browser
and
they're,
also
all
different
right
and
I
just
want
to
here
to
give
some
like
a
scientific
field.
A
To
that
thing,
this
is
a
from
an
article
by
Flavio
Ribera,
one
of
our
advisors,
and
so
he
actually
it's
also.
He
did
an
experiment
where
this
was
with
with
web
torrent,
where
an
episode
I
mean
the
top.
So
was
an
experiment
with
a
torrent
where
people
would
see
videos
and
and
there
and
then
try
to
download
as
much
as
possible
from
their
peers
instead
of
hitting
a
central
server,
and
so
the
top
graph
here
is
the
100%
is
if
their
hits
do
their
to
the
centralized
server.
A
If
not
if,
in
the
normal
case,
where
you
don't
have
peer-to-peer
connections-
and
this
is
what
reduction
in
hits
to
the
central
server,
when
you
add
mattered
web
torrent,
and
so
in
the
best
of
cases
they
got
like
a
reduction
of
like
30
40
percent.
This
is
it's
not
so
bad.
Of
course
I
mean
it's.
It's
really
very
good.
If
you
are
costs,
if
your
job
is
paying
for
a
CDN,
but
it's
very
long
way
off
from
our
idea
of
using
IP
FS
or
something
like
swarm
as
a
single
content,
distribution
network.