►
Description
This talk took place at the Honk Kong 2015 Blockchain Workshops organised by COALA http://coala.global/
http://hk.blockchainworkshops.org/
A
You
ready
okay,
so
we
need
actually
consensus
making
it
on
the
agenda.
We
had
to
a
fork
of
the
agenda.
We
had
one
of
the
website
and
one
which
is
handed
out
so
I'm
gonna
reconcile
the
fork
with
my
for
consensus
rules
and
why
you
have
the
floor
on
the
distributed
word
for
the
internet
and
I
PF
s.
Take
it
away.
B
Thank
you
thanks
for
having
me
so
quick
show
of
hands
how
many
people
know
what
ipfs
is
already
cool
substantial
amount.
So
we
like
to
call
the
IP
fest
a
distributed
web,
because
it's
an
effort
to
make
web
sites
and
web
applications
operate
in
a
distributed
context.
Another
name
for
ipfs
is
also
the
permanent
web
or
the
merkel.
Web
and
ipfs
is
a
protocol
to
upgrade
how
the
web
works,
but
I
profess
loves
blog
chains
in
general.
B
The
structure
of
this
talk
is
that
I'll
talk
a
bit
about
the
web
and
why
there
are
some
problems
there
that
we're
trying
to
address.
Some
of
the
things
that
have
been
discussed
in
this
conference
will
feature
there
to
some
extent,
I'll
describe
what
IP
FS
is
and
how
its
structured
I'll
talk
about,
how
it
relates
to
block
chains
and
I'll
conclude.
B
If
there's
time,
we
tried
out
with
our
discussion
on
how
the
on
the
project
itself,
because
I
think
it
has
a
lot
to
say
about
how
the
internet
gets
patched
today
and
how
major
developments
happen
in
in
general,
all
right,
so
I
like
starting
with
this
image,
because
it
is
a
very
clear
construction
of
the
different
kinds
of
networks-
and
this
is
this-
comes
from
Paul
Baran,
who
was
one
of
the
inventor
inventors
of
packet?
Switching
way
back.
B
This
was
when
AT&T
was
switching
or
like
go
tell
coach
we're
switching
from
switch
lines
to
packet,
switching
way
before
the
Internet,
and
he
characterized
networks
as
being
either
centralized
decentralized
or
distributed.
And
the
major
point
here
is
that
in
a
centralized
network,
it's
easy
to
think
about
it,
because
there's
just
one
thing
that
does
all
the
work
and
a
bunch
of
clients,
decentralized
networks,
kind
of
shard
that
and
gain
some
replication
or
some
resiliency.
B
In
hardware
that
you
get
get
this
kind
of
fabric
that
you
can
break
apart
in
any
any
kind
of
way
and
the
whole
thing
will
still
kind
of
work,
and
so,
when
you
think
about
the
internet
and
the
protocols
that
make
it
up,
there's
of
course,
agreement
that
helps
understand
how
the
that
help
made
and
formulate
this
amazing
machine
that
we
that
we
have-
and
it's
all
really
about
specs
code
and
computers
right.
We
have
a
whole
bunch
of
ideas
that
we
synthesize
into
agreements
into
protocols.
B
We
turn
those
into
code,
which
means
just
taking
the
ideas
in
the
specs
and
massaging
them
into
a
program,
and
then
we
run
the
programs.
This
means
the
Internet
is
extremely
malleable.
It
means
that
anybody
can
actually
come
in
and
change
the
internet.
All
they
have
to
do
is
come
up
with
a
set
of
ideas
that
are
good
enough,
implement
it
and
ship
it.
If
it's
successful,
people
will
adopt
it
and
people
will
use
it,
and
people
will
eventually
make
it
part
of
the
core
of
the
Internet.
This
is
what
happened
to
Bitcoin.
B
In
fact,
we're
doing
it
already.
We
have
lots
of
over
between
fifty
and
a
hundred
thousand
websites
already
run
on
ipfs.
So
that's
an
interesting
all
right,
so
the
web
and
the
internet
not
the
same
thing
right.
The
Internet
is
the
wires.
The
web
is
the
applications
on
top
of
those
wires
in
those
computers
and
when
you
think
about
the
applications
that
we
run
today
on
the
web,
these
applications
more
and
more
rule
all
of
our
lives.
I.
Think
about
how
much
data
that
you
generate
goes
into
some
of
these
web
applications.
B
Is
that
the
web,
though
it
started
in
a
distributed
sense,
and
this
kind
of
very
peer-to-peer
notion
that
everyone
was
going
to
run
both
an
HTTP
server
and
a
client
and
be
able
to
share
documents
with
each
other?
It's
completely
centralized.
Now
we
run
browsers
and
we
consume
content
from
web
servers,
and
we
talk
to
the
web
servers
through
like
these
little
carefully
carefully
constructed
ways,
but
you,
as
a
web
browser,
don't
really
get
to
publish
data
into
the
network.
B
It's
everything
is
mediated
by
a
set
of
servers,
which
also
means
that
if
you
want
to
download
data,
you
have
to
go
and
bring
it
down
from
those
servers
and
it
becomes
an
extremely
inefficient
mess
right
like
if
all
of
you
right
now
started
downloading
Gangnam
style.
We
could
even
suppose
that
there
were
seven
or
eight
links
in
between.
We
would
end
up
wasting
tons
of
bandwidth.
B
We
should
calculated
based
on
the
number
of
views
when
I
made
the
slide
it's
about
almost
500
petabytes
of
data
coming
off
of
Google
servers,
let
alone
times
8
for
all
those
links.
You
know,
depending
times
whatever
the
diameter
of
the
network
is
gets
worse
when
you
think
about
offline
use
cases
right.
So
if
all
of
us
were
collaborating
on
a
Google
document
or
some
sort
of
application
through
the
web
and
the
internet
fell
apart
in
our
connection
to
the
internet
fell
apart,
it
would
just
cease
operating
the
web.
B
Web
apps
are
not
designed
today
to
continue
to
operate
in
the
offline
case.
They
are
not
offline
first
and
at
the
time
that
I
have
been
decided
as
well.
There's
all
these
other
very
critical
pieces
of
humanity.
Now
that
just
cease
to
work,
if
your
latency
or
bandwidth
are
above
or
below
certain
limits
and
I
think
this
is
unacceptable.
I
think
that
we
as
engineers
need
to
pick
up
the
game
here
and
fix
this
major
problem,
because
it
actually
is
really
critical.
B
When
you
think
about
how
people
are
using
these
things,
we
tend
to
design
web
applications
as
this,
you
know
thing
that
kind
of
is
supposed
to
work
in
the
best
case
and
we
hand
them
out
to
people
and
people
fall
in
love
with
them
and
they
use
them
all
the
time.
And
then
we
don't
stop
to
think
about
what
happens
when
the
model
of
execution
that
we
thought
about.
Isn't
the
one
that
applies
in
their
daily
life?
B
We
don't
stop
to
think
about
when
their
connectivity
breaks
and
that
dependence
that
they
developed
on
the
software
we
created
ends
up
hurting
them
quite
a
bit,
and
you
know
think
about,
like
all
the
devices
that
people
are
getting
nowadays
and
you
know
think
about
Internet
of
Things
and
so
on.
These
things
aren't
capable
of
sharing
data
through
the
web
at
all.
B
They
share
it
through
usually
native
protocols
that
they
implement
the
whole
idea
of
this
amazing
idealized
web
of
documents
that
we
were
all
going
to
share
and
collaborate
through,
and
these
webs
of
applications
don't
extend
into
mobile
and
don't
extend
into
the
Internet
of
Things.
Yet,
and
so
this
is
another
problem
like
the
web
is
getting
kicked
out
of
these
devices
and
the
value
that
were
brought
by
integrating
everything
together
is
being
stopped
from
entering
here,
and
so
this
is
another
issue
that
we
need
to
fix.
B
There's
of
course,
information
silos
when
you
think
about
all
of
the
databases
out
there
of
the
big
social
networks,
it's
really
their
data
right,
like
it's.
It's
sort
of
your
data,
but
not
really
it's
like
you
sort
of
have
rights
on
it,
but
they
control
it,
and
certainly
you
can't
link
it
to
each
other.
You
can't
link
it
to
other
pieces
of
data
on
the
web
in
such
a
way
that
will
remain.
B
Should
that
website
shut
down
or
kick
you
out
or
whatever,
so
that
data
is
the
the
whole
point
of
the
web
was
to
create
pieces
of
data
that
link
to
each
other,
and
if
that,
all
of
that
is
mediated
through
specific
entities,
then
the
utility
and
value
of
your
data
depends
on
those
entities.
I
think
the
people
that
know
this
best
are
the
people
of
Egypt.
B
When
suddenly,
they
woke
up
one
morning
to
the
fact
that
their
internet
had
been
completely
shut
down
and
their
communications
were
gone,
people
had
been
communicating
with
each
other
through
social
networks
and
so
on,
and
suddenly
nothing
worked,
and
so
this
is
again
another
problem
when
you
design
a
communications
system
or
a
communications
application,
and
you
don't
think
about
what
happens
when
a
government
decides
to
shut
down
Internet
access
for
people.
That's
a
big
problem.
Of
course.
B
Thankfully,
people
deployed
mesh
networks
and
so
on
and
and
they
were
managed
to
get
internet
access
back,
but
that's
not
a
given
and
so
applications
need
to
be
built
to
deal
with
these
kinds
of
use
cases.
Applications
need
to
run
in
local
area
networks.
Think
about
okay,
so
that
was
a
man-made
disaster.
But
what
about
natural
disasters?
What
happens
to
the
web,
then?
What
happens
to
all
these
communication
infrastructure
is
when
there's
earthquakes,
floods?
You
know
super
volcanoes
exploding
right,
I
mean
what
what
do
we
do
then?
B
What
happens
if
major
disasters
happen
to
data
centers
and
suddenly
we
don't
have
our
data
anymore,
a
company
will
say
sorry
we
lost
it.
What
are
we
I
mean
natural
disaster?
We
didn't
account
for
that
and,
of
course,
a
few
companies
do
and
actually
are
big
enough,
that
they
do
think
about
the
sort
of
thing
and
they
do
replicate
your
data
across
a
few
data
centers,
but
it's
not
perhaps
as
replicated
or
as
safe,
as
you
might
believe,
and
it's
not
as
safe
as
you
may
want
it
to
be.
B
There's
also
the
problem
with
book-burning
right
I
mean
we
have
been
criticizing.
Book-Burning
as
this
horrible
thing
that
happens
when
a
society
kind
of
goes
crazy
and
we
think
see
book-burning
as
this,
the
ultimate
sin
against
humanity.
We
see
humanity
as
the
product
of
language
and
technology
and
knowledge.
B
Those
are
the
things
that
distinguishes,
and
yet
some
societies
burn
books-
and
we
see
this
as
like
a
trait
of
the
things
going
really
really
badly.
And
yet
today
we
burn
books
all
the
time
we
burn
books
daily.
We
burn
books
every
day
whenever
you
move
and
occupant
on
the
web
and
and
a
URL
no
longer
points
to
where
it
used
to
whoever
had
a
link
to
it
and
now
cannot
see
it
for
them.
It's
a
book
burned
for
them.
They
may
not
be
able
to
find
the
document
anymore.
B
They
may
not
be
able
to
access
it.
They
may
not
even
have
a
search
that
works
fully.
So
the
critical
point
of
the
web,
which
was
to
create
this
idealized
notion
of
documents
linked
to
each
other,
has
a
problem.
These
documents
are
documents
on
computers
and
you
can
burn
those
links
and
you
can
burn
those
computers
too.
So,
each
of
some
of
the
critical
critical
problems
I
hope
that
I've
inspired
in
you
a
sense
of
the
urgency
of
these
matters
and
why
it
is
important
to
upgrade
the
web.
B
So
ipfs
is
this
project
to
make
the
web
work
in
the
distributer
case.
Work
offline
first,
be
more
permanent,
be
safer
on
the
to
the
user,
move
the
contour
around
in
a
smart
way
and,
of
course
be
faster,
because
if
you
don't
make
it
faster,
no
one's
gonna
use
it.
Actually,
this
all
started
by
trying
to
make
it
faster
and
then,
after
a
while,
all
of
the
other
properties
kind
of
fell
out,
which
is
cool,
I
profess,
is
a
hyper
meteor
transport
protocol,
the
same
thing
as
HTTP
and
the
goal
is
to
match
the
interface.
B
Exactly
things
shouldn't
have
to
change
what
applications
should
not
be
different
at
all.
You
should
still
be
able
to
run
anything
that
you
run
in
HTTP
over
IP
FS
with
minimal
to
no
modification.
Some
things
will
be
harder,
of
course,
like
the
more
complicated
web
applications
will
be
more
difficult
to
to
translate,
but
you'll
gain
some
very
interesting
properties.
B
I
profess
is
the
product
of
looking
back
through
the
last
kind
of
25
years
of
developments
since
the
web
has
been
created
and
think
about
what
the
web
would
have
looked
like
today,
if
we,
if
those
ideas
have
been
around
when
Tim
berners-lee
invented
the
web,
so
we've
come
up
with
a
whole
bunch
of
good
ideas
right
since
the
way
invention
of
the
web.
What
would
it
look
like
if
those
kind
of
made
it
in
you
might
think
because
we're
at
the
blockchain
worship?
Why
isn't
that
going
there?
B
Why
isn't
the
blockchain
sort
of
like
part
of
this
well
Bitcoin
and
the
blockchain
didn't
come
up
with
Merkel
links
that
was
kind
of
an
older
idea,
so
it
would
have
made
it,
but
it
was
kind
of
already
done,
and
so
these
protocols
kind
of
separate
out
into
a
stack.
The
web
is,
of
course,
the
great
application
platform
that
we
know
s.
Fs
was
a
fantastic
protocol
that
made
sure
that
we
have
a
secure
way
to
do.
Naming
get,
gives
us
the
whole
notion
of
version
versioning
of
data
and
Merkel
links.
B
That's
where
I
I
learned
about
them,
they're
even
older,
but
I
learned
about
them
and
GUID
and
of
course,
BitTorrent.
Has
this
amazing
way
of
moving
around
content
very
officially
through
networks
and
the
HDS
allows
us
to
find
contents
and
the
whole
thing
is
designed
so
that
it
works
over
any
network.
So
this
is
not
true,
HTTP
HTTP,
you
can
make
it
work
over
any
transport,
but
in
general,
most
web
servers.
Just
don't
do
that.
They
only
work
over
TCP.
B
So
if
you
wanted
to
run
the
web
over
Bluetooth,
if
you
wanted
to
run
the
web
over
audio
or
something
like
that,
you
have
to
work
really
really
hard
to
do
that,
which
means
that
nobody
does
it,
and
so
the
the
stack
kind
of
breaks
down
into
these
sections,
something
we
were
calling
lit
p2p,
which
is
a
set
of
protocols
that
a
whole
bunch
of
other
projects,
could
use
and
really
the
core
part
which
we're
nicknaming
either
the
miracle
dog
or
the
miracle
web
or
IP
LD.
And
that
is
the
core
part
of
the
protocol.
B
The
idea
is
to
come
up
with
one
core
format
that
makes
sense
for
all
of
these
distributed
data
structures
for
all
these
distributed
systems
that
want
to
interact
through
the
web
through
the
network
and
address
content
with
each
other
and
moving
around,
and
so
the
the
stack
here
there's
a
set
of
protocols
in
this
stack
that
you
can
see
that
are
ipfs
specific.
But
the
whole
point
is
that
you
could
swap
them
out
you
could
you
could
actually
use
HTTP
in
exchange
in
ipfs?
B
B
Ip
was
such
a
good
idea
because
it
created
this
thin
waist
around
the
IP
protocol,
we're
doing
the
same
thing,
but
for
data
itself,
so
we're
coming
up
with,
like
the
the
thin
waist
of
what
it
means
to
define
these
data
structures
that
are
distributed,
and
it's
not
just
data.
It's
just
data
structures.
B
It's
you
need
a
way
to
express
data
structures,
the
same
way
that
other
people
express
them
today
and
if
you
have
ever
used
git,
you
know
that
it's
a
Merkel
tree
or
something
like
it
and
I
won't
explain
fully
what
a
Merkel
tree
is
if
you've
heard
about
a
blockchain.
The
whole
idea
is
that
you
have
some
block
and
you
have
the
hash
pointing
to
a
previous
block,
and
that
creates
this
chain
of
links.
That's
all
that
amiracle
link
really
is,
and
that's
the
property
that
we
care
about.
B
But
the
point
is
that
get
and
a
whole
bunch
of
other
protocols
have
these
different
trees.
There's
all
these
different,
separate
repositories
of
information
and
even
bitcoin
and
blockchain
czar.
This
massive
Merkel
tree
like
this
huge
chain,
all
along
the
way
you
have
this,
these
Merkel
links
I
professor
Merkel
forests.
The
whole
point
is
to
bridge
together
the
distinct
different
systems
by
coming
up
with
the
same
way
of
linking
between
them,
so
you
can
think
of
Bitcoin
and
gets
and
BitTorrent,
and
that
and
so
on,
as
pieces
of
the
same
system
as
a
whole.
B
That's
how
you
achieve
progress
on
the
network
if
you
come
up
with
a
protocol
and
its
really
great,
but
like
your
implementation,
isn't
helpful
or
friendly
to
developers
thanks
for
try
again,
it's
just
not
gonna.
It's
not
gonna
move
the
needle.
It's
gonna
have
to
wait
until
somebody
else
manages
to
bring
that
down.
B
That
is
a
hash
right,
so
the
block
on
the
Left,
when
you
hash
that
you
get
the
value
that
goes
into
that
it's
included
in
the
block
on
the
right,
and
this
is
a
merkel
link.
You
get
the
ability
to
check
integrity
of
this
huge
chain
as
it's
forming
through
that
property.
Of
course,
blotches
have
this
additional
thing,
or
at
least
in
particular
the
Bitcoin
blockchain.
B
B
When
you
look
deeper
at
a
blockchain,
it
is
pointing
to
a
set
of
transactions
right.
This
is,
these
are
all
still
Merkel
links
and
the
transactions
that
aren't
yet
included
are
sort
of
in
that
pool
somewhere,
and
the
process
of
extending
the
chain
is
taking
the
transactions
that
are
valid
and
putting
them
into
blocks
and
extending
the
chain.
B
What
happens
when
you,
when
you
want
to
include
you,
know,
data
on
the
chain
that
doesn't
fit?
We've
talked
this
entire
conference
about
including
things
like
contracts
and
Records
and
important
documents,
and
so
on
in
the
blockchain,
but
you
can't
actually
put
it
in
there
through
a
transaction,
because
you
don't
want
every
single
node
ever
to
have
to
to
to
store
it.
B
So
you
do
the
same
thing
that
the
blockchain
does
and
you
put
a
hash
to
it
as
a
link
right,
but
these
these
this
content,
which,
by
the
way,
is
starting
to
be
all
sorts
of
stuff.
There's
like
legal
records
contracts
code,
you
I've,
seen
email
over
bitcoin,
pretty
much
anything
right.
What,
but
what
happens
with
those
links?
You
can't
click
on
a
link
like
that
there's
just
some
hash
right.
B
B
You
have
a
whole
bunch
of
servers
and
if
you
want
a
specific
file,
you
have
to
talk
to
a
specific
server
to
get
it
back
or
a
specific
set
of
servers,
even
if
a
whole
bunch
of
other
computers
in
the
network
have
it
doesn't
matter.
You
have
to
go
to
talk
to
a
specific
one
and
there's
there's
a
reason
for
this.
It
made
sense
at
the
time,
but
maybe
not
anymore.
B
So,
of
course,
if
you
wanna
put
in
the
blockchain,
you
don't
do
that.
You
put
the
hash
of
the
content.
So
why
don't
we
build
a
system
to
just
address
everything
by
hash?
And
yes,
there
are
some
already.
But
the
point
is:
why
don't
we
make
the
web
itself
work
this
way,
and
this
is
what
I
profess
is
about
making
the
web
itself
work
with
a
hash
based
file
system,
and
so,
instead
of
this
picture,
you
get
a
picture
where
any
note
that
has
the
content
can
distribute
the
content
to
you.
B
Because
again,
you
need
no
trust.
It's
the
same
thing
as
you
get
and
baturin
the
same
thing:
you're
getting
get
and
the
same
thing
you
get
in
Bitcoin
block
chains
and
so
on.
Well,
the
benefit
is
that
it's
a
file
system
right
like
we're
talking
about
it's
good,
to
store
rock
records
and
documents
and
directories
and
so
on,
and
the
resolution
works
exactly
as
you
would
expect
it
to
work
in
a
file
system.
The
same
way
that
you
would
expect
it
to
work
in
the
web.
B
You
have,
you
can
have
directories
that
point
to
other
objects
and
so
on.
This
is
exactly
what
get
data
and
why
it
became
so
so
successful.
These
are,
of
course,
miracle
links
and,
if
people
kind
of
think
you
know
what
it
did
and
why
it
was
successful,
the
whole
idea
was
that
there
was
a
previous
version,
control
called
sen
and
before
that
they
were
CVS
and
so
on,
and
the
model
was
that
it
was
centralized
right.
B
You
had
one
server,
maintaining
your
versions,
and
everyone
would
talk
to
that
one
server
and
so
to
make
any
update.
You
would
have
to
ship
that
server
that
updated
to
that
server.
This
was
not
very
robust
or
resilient.
If
you
cut
any
one
of
the
links,
the
hope
you
you
couldn't
talk
to
it,
so
you
couldn't
work,
you
couldn't
make
any
update
if
the
central
server
went
down.
Of
course,
everything
fell
apart.
Its
improvement
was
to
make
the
entire
thing
distributed
to
make
it
offline
first.
B
This
is
the
same
thing
that
Bitcoin
did
in
a
sense.
Any
node
in
this
network
is
capable
of
maintaining
its
own
record
of
the
versions
and
talk
to
each
other.
So
if
part
of
the
network
goes
down,
doesn't
matter,
anybody
can
still
work
and
you
will
sync
back
when
you
get
together.
If
the
server's
go
down,
doesn't
matter
at
all,
you
can
still
work
and
talk
to
each
other.
B
This
is
what
I
profess
is
doing
to
the
entire
web,
we're
moving
the
web,
apps
and
websites
and
documents
and
archives,
and
everything
that
you
can
touch
on
the
web
to
work.
This
way,
it's
what
I
call
hyper
speed,
because
it's
really
fast,
if
you
download
something
and
you
have
it
locally
and
you
have
the
hash-
you
never
have
to
download
it
again.
B
B
What
everyone
here
wants
out
of
the
block
chain,
or
you
want
some
of
the
things
that
people
want
out
of
a
blockchain
right
as
I've
listened
to
people
talk
throughout
the
blockchain
workshop,
many
people
that
are
interested
in
block
chains,
not
from
a
you
know,
smart
contra
sense,
but
from
a
as
a
way
to
store
data.
It's
the
same
thing:
it's
if
you
get
the
same
kind
of
properties
here,
and
so
this
is.
This
is
a
a
different
way
of
thinking
about
how
to
store
information.
You
can
you
the
important
property
of
the
blockchain.
B
Is
that
it's
timestamps
right,
and
so
you
can
think
of
putting
data
on
ipfs
and
taking
the
the
route
hashes
from
ipfs
and
then
time,
stamping
them
into
the
blockchain
and
vice-versa.
You
could
take
the
blockchain
itself
and
put
all
of
it
on
ipfs,
in
fact,
there's
people
that
are
starting
to
build
block
chains
that
are
designed
to
be
entirely
on
IP
of
us,
because
it
makes
people's
life
easier,
like
you,
don't
have
to
work,
do
all
this
peer-to-peer
stuff.
That
is
difficult.
B
So,
when
you
think
about
websites
today,
they
work
kind
of
like
this
right,
there's
a
whole
bunch
of
servers
that
talk
to
each
other
and,
of
course,
there's
clients
too,
and
everybody
has
a
big
database
and
the
data
is
kind
of
like
within
the
databases
and
that's
it
and
you
talk
to
each
other
over
these
wires.
But
the
problem
is
that
all
of
that
data
is
really
interlinked,
but
those
links
again
refer
to
servers
in
the
a
confessed
model.
B
It's
about
flipping
that
on
its
head
and
saying
let
the
data
connect
to
each
other
and
so
on,
and,
of
course,
if
anybody
remembers
linked
data
and
the
Semantic
Web.
This
is
the
same
idea
right,
like
let
data
connect
to
each
other.
The
one
issue,
though,
is
that
the
link
data
was
still
dependent
on
those
links
is
still
dependent
on
on
those
servers.
B
It's
still
addressed
by
mostly
not
all
link
data
is
address
this
way,
but
the
large
majority
of
link
data
is
addressed
by
by
those
locations,
and
so
it
makes
it
extremely
difficult
to
work
with,
because
those
links
can
go
down.
You
don't
want
to
query
like
20
servers
at
once
and
so
on.
What
you
really
want
is
something
like
it
or
the
blockchain,
where
you
can
take
entire
portions
of
data,
move
them
to
one
location,
even
completely
go
offline
and
still
have
everything
in
operation
and
so
yeah.
B
Creating
this
this
distributed
network
where
data
can
link
to
each
other
and
this
with
Merkel
links,
and
you
can
move
it
around
pretty
easily.
It
turns
out,
though,
to
be
a
lot
of
pretty
much
like
the
barrier
to
this
is
making
new
ways
of
like
basically
matching
all
the
interfaces
that
people
are
expecting
out
of
how
you
make
websites
and
web
applications,
but
make
it
a
whole
work
with
this
kind
of
distributed
effort
cool
all
right.
So
hopefully
that
explains
what
I
profess
is
about.
B
So
this
started
because
I
wanted
to
build
a
a
virgin
data
set
package
manager,
so
I
wanted
to
make
it
easier
for
scientists
to
move
around
data
and
simultaneously
I
also
wanted
to
make
a
like
some
sort
of
storage
layer
for
some
sort
of
like
distributed
agents.
I
could
talk
to
each
other,
but
that
was
kind
of
like
a
secondary
goal
and
as
I
got
into
into
merging,
get
and
BitTorrent
I
realized.
B
So
you
could
actually
just
sink
in
and
restructure
how
the
the
data
structures
worked,
make
a
better
protocol
and
use
that
instead
and
as
I
started
thinking
more
and
more
about
it.
It
became
clear
that
this
had
vast
implications
for
the
web
itself
and
it's
only
sort
of
after
the
fact
that
this
became
clear.
B
B
And
if
we
follow
that
one
thing,
then
everything
can
can
talk
to
each
other
again
right
like
let's
not
break
the
web,
and
so
this
this
is,
of
course,
all
open
source,
as
all
important
protocols
should
be-
and
it's
you
know,
there's
a
huge
community
now
behind
this
I
sort
of
started
it.
But
even
the
ideas
came
before
me
like
this
is
kind
of
the
product
of
decades
of
and
thinking
and
engineering,
from
a
lot
of
really
good
people.
B
My
task
has
been
mostly
integrating
ideas
into
producing
a
good
interface
that
matches
all
of
the
pieces
and
there's
a
again
a
large
community
on
it
already.
There's
tons
of
users,
lots
hundreds
of
contributors
and
also
quite
a
bit
of
traction
already.
So
we
we
describe
our
project
as
an
alpha,
and
yet
people
are
using
in
production
all
over
the
place.
There's
the
like
I
said
50
to
100
thousand
websites
on
it,
there's
tons
of
different
block
chain
or
Bitcoin
or
aetherium
related
companies
that
are
using
it.
B
It
recently
was
shipped
in
a
very
large
provider
of
like
network
that
attached
storage
devices
with
a
massive
install
base,
so
that'll
be
interesting.
When
that
comes
online,
it's
it's
gaining
a
lot
of
traction
and
moving
really
quickly,
which
is
great,
but
at
the
same
time
it's
it's
kind
of
important
too,
so
to
keep
to
kind
of
like
step
back
and
think
about
how
this
project
is
moving
forward
and
sale
a
little
bit
a
couple
of
sameness
about
how
these?
B
People
don't
do
this,
and
it's
and
I
started
a
company
to
build
this
out,
because
I
wanted
to
have
like
an
independent
group
and
I
think
this
is
another.
Another
thing
that's
important
about
how
these
protocols
emerged.
When
you
think
about
Bitcoin,
it
was
a
person
anonymous,
put
it
out
there,
what
you
think
about
get
it
was
built
by
the
kernel
hackers,
not
a
company
when
you
think
about
BitTorrent.
It
was
built
a
one
individual
and
so
on.
B
B
But
what
I
want
to
build
is
a
a
lab
where
it's
completely
open-source
and
people
can
come
together
and
work
on
things,
no
matter
what
organization
they're
part
of
and
sort
of
like
a
like
a
strong
arm
for
the
IETF,
a
group
that
can
take
the
moment
to
look
at
the
whole
internet
stack
think
about
what's
missing
what
would
be
really
important
to
add
and
then
just
do
it.
So
that's!
What's
what
the
effort
behind
ipfs
is
becoming
it's
about,
creating
a
lab
like
this
a
we
call
it.
B
The
protocol
lab
it's
kind
of
like
Tesla
meeting
in
the
the
protocol
stack.
But
if
you
go
back
to
this
slide
about
specs
code
on
computers,
the
problem
that
we
have
today
is
that
so
much
research
exists
and
it's
great
I
mean
it's
amazing
that
there
is
so
much
research
there's
so
many
good
ideas.
But
the
issue
that
I
see
often
is
that
I
can
even
Mia
it's
like
25
years
I
had
it's
light
years
ahead.
B
We
it's
like
alien
technology
to
us
to
like
us,
mere
mortals,
and
the
issue
is
that
so
much
what
we're
talking
about
today
has
been
solved
already
by
academics
decades
ago,
and
why
are
we
not
seeing
those
results
today
and
it's
because
this
funnel
is
broken?
It's
because
when
we
go
from
research
to
development,
there's
a
huge
gap.
Most
academics
come
up
with
really
great
ideas
and
move
on
to
the
next
paper,
but
very
few
people
actually
take
the
time
to
implement
those
pieces
of
ideas
into
a
system.
B
C
A
C
Haven't
looked
I
have
like
only
played
around
a
little
bit
with
ipfs
is
a
very
interesting
project.
My
question
is:
it
looks
very
good
for
like
static
content.
What
about
dynamic
conferences
if
I
have
like
a
Twitter
of
like
if
I
go
to
Twitter
and
Twitter
with
generates
on
the
fly?
My
news,
feed
and
yeah.
B
We
can't
know
the
hash
beforehand,
so
I'll
describe
it
in
terms
of
get
first
and
maybe
that'll
make
sense.
It
has
a
whole
bunch
of
static
content
inside
and
a
branch
pointer
just
moves
to
do
mutability
that
didn't
make
sense.
So
you
have
this
massive
amount
of
static
content,
but
what
you
need
to
do
to
get
mutability
is
that
you
need
a
pointer
that
can
move
somewhere,
that
it's
not
an
immutable
content,
pointer,
it's
a
mutable
pointer
and
we
get
that
with
naming.
B
We
call
them
names,
yes,
I
get
branches,
but
it's
we
call
them
names
and
so
in
the.
If
we
go
back
to
this
like
slide
way
back
about
the
links
here,
we
go.
You
see
there
that
I
have
ipfs
slash
a
hash.
That
is
a
hash
based
on
the
content
of
the
file.
Then
above
it
is
the
thing
called
/
IP
NS
for
namespace
there
is
a
DNS
name
right
example.com.
B
What
that
means
is
that
there
is
a
DNS
record
out
there
that
has
the
latest
version
that
has
aid
a
tax
record
pointing
to
that
content.
Of
course,
that's
not
the
best
solution.
You
don't
want
to
wait
for
DNS
resolution
to
get
updates
right.
You
want
it
to
be
less
than
a
second,
ideally
less
than
milliseconds,
sometimes,
and
you
get
that
through
another
layer
which
I
don't
show
here,
but
that's
where
s
FS
naming
comes
in
a
name
or
a
branch
in
in
ipfs
just
means
the
hash
of
a
public
key.
B
So
if
you
have
a
private
key,
that
means
you
can
create
a
pointer
to
some
content,
PFS
and
publisher
record
through
some
system,
and
people
can
trust
that
name.
So
the
name
is
ugly
right.
It's
a
huge
hash.
It's
not
a
nice
pre
name,
but
people
can
trust
that
name
because
it's
authenticated
right.
It's
like
the
name
itself
is
the
hash
of
a
public
key
from
that
they
fetch
the
public
key
and
they
can
verify
that
you
sign
the
record
correctly.
B
What's
it
like
an
onion,
you
are
oh
yeah,
exactly
actually
yeah,
so
as
if
his
naming
is
all
over
the
place,
it's
and
it's
an
onion.
It's
in
Farina
new
net
taco
FS,
there's
just
tons
of
things
that
use
it,
and
these
are
by
the
way
other
projects
that
are
related
to
ipfs
and
have
been
thinking
about
the
same
sort
of
social
about
yes,
so
definitely
check
those
out
and
yeah.
If
there's
anything
that
we
should
do
better,
please
come
tell
us
because
we
want
to
make
the
best
thing
possible.
So.