►
From YouTube: Filecoin’s Secondary Retrieval Market Architecture
Description
Learn about Filecoin’s plans for a Retrieval Market, and hear from four of the projects building it today.
Join us for Filecoin Liftoff Week, an action-packed series of talks, workshops, and panels curated by the web3 community to celebrate the Filecoin mainnet launch and chart the network’s future.
https://liftoff.filecoin.io/
Events take place all week, October 19-23, 2020. #FilecoinLiftoff
For more information on Filecoin
- visit the project website: https://filecoin.io/
- or follow Filecoin on Twitter: https://twitter.com/Filecoin
Get Filecoin community news and announcements in your inbox, monthly: http://eepurl.com/gbfn1n
Chapters
00:00 Retrieval Markets Meetup Intro
02:30 Peter Robinson of Consensys
04:00 Why use a Secondary Retrieval market?
11:30 Mike Goelzer Intro to Browser Retrieval Network
13:50 Demoing the Browser Extension
21:10 Thomas Chardin demoing Decentralized Filefinder
23:00 What is Filefinder?
28:35 Introducing David Ansermino + Elizabeth from Chainsafe
32:00 Key Requirements for a secondary market
34:54 So how do you optimize for profitability?
38:15 Demos
A
My
name
is
mike
gelser,
I'm
from
the
file
coin
project.
You
can
reach
me
at
m
gelser
on
falcon
slack,
and
I
just
want
to
sort
of
introduce
things
by
talking
about
first,
just
a
second.
What
what
is
the
retrieval
market?
So
if
you
go
back
to
the
original
file
coin
white
paper,
the
first
file
coin
paper
came
out
in,
I
think
2013,
it's
been
updated
several
times,
but
what
it
describes
is
actually
two
markets.
So
the
main
one
is
the
storage
market.
That's
the
one
that
is
launched
this
week.
A
A
You
know
a
large
number
of
people,
and
so
in
that
case,
what
you
want
actually
want
is
for
the
providers
of
retrieval
to
be
in
a
competitive
market
with
each
other,
where
they're
competing
to
provide
you
with.
You
know
certain
content
at
the
lowest
cost
at
the
highest
reliability,
the
same
idea
as
the
storage
market,
and
so
what
we're
gonna
do
tonight
is
hear
from
four
different
teams
who
have
taken
different
approaches
to
to
this
problem,
and
they
all
have
very
interesting
perspectives
on
it.
A
So
I
think
I
think
it's
gonna
be
a
good
event.
I
want
to
ask
the
presenters:
please
keep
yourself
to
10
minutes
and
for
the
audience,
please
feel
free
to
use
the
q
a
at
any
time.
You
can
ask
questions
and
presenters
can
address
them
either
in
their
presentation
or
in
writing
after
the
after
they
finish.
So
with
no
further
ado,
I
want
to
move
to
peter
david
and
robert
from
consensus
to
hear
about
their
retrieval
markets,
architecture.
B
So
when
you're
thinking
about
markets,
there
are
right,
there's
the
primary
retrieval
market
and
the
secondary
retrieval
market.
So
primary
retrieval
is
really
about
going
directly
to
a
storage
miner
to
get
your
content,
whereas
a
secondary
retrieval,
it's
all
about
discovery
and
having
the
information
in
highly
available
off-chain
sources,
and
so
it's
all
about
being
able
to
discover
where
the
content
is
and
being
able
to
get
it
and
by
market.
It's
all
about
having
choice
and
being
able
to
have
a
market
between
buyers
and
sellers
to
find
the
best
deal.
B
So
you
can
think
of
a
primary
retrieval
as
you've
got
a
lotus
node,
which
talks
directly
to
a
storage
miner
to
get
the
content.
So
when
we
were
thinking
about
this,
we
about
what
the
objectives
for
this
system
should
be.
Discovery
is
obviously
a
big
one.
You
know
we
need
to
discover
where
the
content
is.
We
need
to
promote
caching
in
the
network
so
that
you
might
have
content
cached
near
you.
B
B
B
So
you
have
much
better
latency,
and
so
these
retrieval
gateways
they
hold
the
offers
that
the
retrieval
providers
have
published.
So
the
retrieval
providers
are
connected
with
storage
miners
and
they
publish
these
offers,
and
then
the
gateways
hold
a
set
of
offers
and
the
clients
talk
to
the
gateway
and
then
they
get
the
offers
so
which
provider
will
provide
the
content
at
what
price.
B
B
And
not
all
gateways
might
have
all
of
the
information,
because
that
you've
got
essentially
a
publication
subscription
model
of
the
retrieval
providers.
Publishing
the
data
and
some
of
the
retrieval
gateways
might
choose
to
store
it,
but
they
might
not
choose
to
store
other
pieces
so
because
of
that
we've
created
a
dht
for
discovering
content,
and
so
if
a
gateway
doesn't
have
a
pcid
and
offers
for
a
pcid,
it
can
use
the
gateway
dht
to
locate
that
information
and
this
gateway.
B
Dht
is
not
your
typical
dht,
so
your
gateway
nodes
register
ron
chain
in
a
gateway
registration,
smart
contract
contract
and
when,
when
they're
registered
they're
staking
some
money,
some
file
coins,
so
they
can
be
slashed
if
they
do,
provably
do
bad
behavior
and
the
number
of
gateways
is
likely
to
scale
with
the
network.
But
you
know
it
would
have
seen
you'd
assume.
B
Tens
of
thousands
of
gateways
is
probably
going
to
be
your
upper
limit,
so
you're
going
to
be
able
to
have
a
table
of
all
the
possible
gateways
in
memory
in
one
of
those
nodes
so
because
of
the
registration
process
and
the
staking
money.
It
means
that
this
the
nodes
are
going
to
be
stable,
they're
going
to
be
long
lived
and
the
addresses
are
known,
and
what
this
means
is
that
we
should
be
able
to
do
a
single
hop
dht.
B
B
B
B
Also,
content
needs
to
be
paid
for,
and
so
typically
it
could.
You
could
do
it
through
a
gateway
if
they
have
payment
channels
set
up,
but,
additionally,
you
could
imagine
nodes
emerging,
called
payment
brokers
that
would
have
payment
channels
with
all
of
the
providers
as
well.
Not
everyone
might
have
file
coin
and
so
a
subscription
service,
where
maybe
you
pay
a
hundred
dollars
a
month
or
twenty
dollars
a
month,
some
amount
per
month
to
a
subscription
service,
and
then
it
retrieves
as
much
content
as
you
need
on
your
behalf.
A
I
am
going
to
present
on
the
browser
based
retrieval
network
that
we
built
the
filecam
project
built
over
the
summer,
and
so
I'm
just
going
to
set
my
timer,
and
this
is
I'm
also
going
to
do
a
demo
of
the
software.
It's
it's
very
close
to
being
usable
at
this
point.
It's
taken
us
a
while
to
get
there.
So,
basically,
let
me
just
explain
what
this
whole
thing
is.
The
the
idea
is
that
browser
clients.
A
A
You
can
download
cid
from
another
browser
client
who
is
storing
that
particular
piece
of
content,
or
you
can
make
a
request
to
the
file
coin
storage
network
and
a
retrieve
from
a
storage
miner,
and
the
way
that
you
talk
to
this
storage
network
is
through
something
called
a
cloud
lotus,
which
is
a
concept
that
we
invented.
But
it's
just
a
it's
a
public
lotus
instance.
A
It
doesn't
have
any
wallets
on
it
or
anything
you
you
only
use
it
to
you,
sign
a
message,
and
then
you
submit
it
to
the
cloud
lotus
and
that
message
will
go
into
the
message,
pool
and
get
mined
and
get
incorporated
onto
the
chain.
So
the
cloud
lotus
is
really
just
a
vehicle
for
you
to
publish
messages
to
the
chain.
So
the
advantages
of
this
model
are
there's
no
lotus
node
required.
You
don't
need
to
run
your
own
lotus.
A
You
also
don't
need
to
trust
the
cloud
lotus
by
the
way
your
signing
keys
stay
in
your
browser.
They
never
they're,
never
shared
with
the
public
lotus.
It's
just
you
just
send
it
pre-signed
messages
and
it
publishes
them
so
because
there's
no
lotus
and
require
you
don't
have
to
wait
to
sync
to
the
chain
which
can
take
like
a
week
or
something-
and
you
can
just
do
retrieval
mine
by
leaving
your
browser
running
so
that
was
kind
of
the
concept
behind
this.
A
Okay,
we'll
hear
it
I'm
going
to
go
to
the
demo.
Now
live
demos
almost
always
fail
for
some
reason.
So
here's
a
bitly
link
to
a
pre-recorded
version
of
the
demo.
I'm
about
to
show
you
you
can
check
that
out.
If,
if
I
fail
now,
but
let
me
let
me
try
to
give
this
a
try,
so
one
sec,
I
need
to
hide
a
bunch
of
controls,
okay,
so
this
is
what
the
browser
extension
looks
like.
A
Obviously
we
haven't
done
any
web
design
or-
or
you
know,
any
kind
of
put
any
thought
into
graphic
design
or
anything
most
of
the
complexity
of
this
is
kind
of
behind
the
scenes,
but
the
basic
idea
is
you're
on
a
peer-to-peer
network
right
now.
I'm
running
this
browser
extension
in
both
chrome
and
firefox.
So
each
node
sees
that
there's
one
other
node
on
the
network
and
that
other
node
is
just
the
other
browser,
and
so
the
you
can
do
two
things
you
can
either
request
data.
A
You
can
query
for
a
cid
and
try
and
request
it
from
the
other
nodes
on
the
peer-to-peer
network
or
you
can
be
a
provider
of
data.
So
you
see
that
this
firefox
is
storing
a
cid
already
it's
about
4
kilobytes
and
what
we're
going
to
do
is
request.
It
from
chrome
and
then
I'm
going
to
explain
a
little
bit
about
filecoin
payment
channels
and
what
what's
happening
behind
the
scenes.
So,
let's
take
this
copy
it
over
to
the
clipboard.
A
Go
go
to
this
chrome
one
this
this
guy
doesn't
have
any
cids
stored.
So
you
don't
you
don't
see
anything
there,
but
we
can
still
query
for
a
cid,
that's
going
to
send
a
message
out
to
all
of
the
other
nodes
on
the
on
the
peer-to-peer
network
and
that
message
is
requesting
the
cid.
Now,
in
this
case,
there's
only
one
other
node
and
it
happens
to
have
the
cid
since
I
copy
and
pasted
it
from
firefox.
So
this
is
the
firefox
node
responding
and
what
he's
saying
is?
A
Yes,
I
have
the
cid
here's
my
price
for
it,
3.7
million
autofill,
which
is
an
autofill,
is
one
to
the
10
to
the
negative
18
of
coin.
So
this
is
actually
very
small
price.
Just
to
show
you
quickly,
you,
you
have
configuration
options,
so
you
can
set
a
price
star
means
is
a
wild
card.
It
means
all
cids
are
going
to
be
priced
at
10
out
of
sorry
1
000
autofill
per
byte.
A
You
can
also
do
it
on
a
per
cid
basis,
and
so
anyway,
this
this
bank,
this
price,
makes
sense
right.
The
file
is
about
three
thousand
three
point:
seven
thousand
bytes
roughly
so
times
one
thousand
the
price
would
be
3.7
million
in
autofill.
So
now
we're
going
to
hit
the
buy
button
and
a
transfer
is
going
to
start,
and
you
can
see
this
guy
is.
A
Is
the
downloading
and
we
see
on
the
firefox,
he's
uploading
now,
there's
a
lot
happening
behind
the
scenes
and
you
can
see
it
in
the
logs,
actually,
both
both
have
logs
and
but
rather
than
walk
you
through
an
execution
trace
of
a
program
which
would
be
very
boring
in
a
demo
like
this.
I've
actually
made
a
slide
to
explain
what
what
is
happening
right
now.
So
here's
the
basic
idea,
the
chrome
extension
is
buying
a
cid
and
the
file
and
the
and
the
firefox
is
selling
it.
A
So
the
chrome,
the
the
chrome
extension-
needs
to
create
a
payment
channel
and
then
it
needs
to
exchange
signed
vouchers
with
the
firefox
in
exchange
for
chunks
of
data.
So
the
firefox
sends
a
chunk
of
data
and
then
the
then
the
chrome
extension
sends
back
a
signed
payment
voucher
and
then
this
you
can
go
around
in
this
loop,
two
three
four
loop
for
as
many
cycles
as
it
takes.
I
think
the
default
chunk
size
right
now
is
one
megabyte
and
it
increases
each
time
it's
a
model
of
escalating
trust.
A
So
anyway,
you
go
around
that
loop
for
a
while
and
for
a
four
kilobyte
file.
You're
only
going
to
go
around
that
loop
once,
but
then,
eventually,
all
the
data
is
transferred.
The
at
that
point
the
chrome
has
sent
all
of
the
payment
vouchers
he's
out
of
the
picture
now,
but
the
firefox
who's,
the
blue
tax,
has
a
bunch
of
things
that
he
needs
to
do
in
order
to
get
paid.
A
So
first
he
calls
a
method
on
the
payment,
channel
actor
called
settle,
and
that
means
that
the
transfer
is
over
and
it
opens
up
a
12
hour
window
during
which
the
firefox
extension
can
submit
signed
vouchers
to
the
channel
and
that's
an
on-chain
operation
gets
published
on
chain
and
that
during
that
12-hour
period,
the
firefox
browser
would
submit
the
highest
value
voucher.
It
has
received
each
each
time
you
go
around
that
loop,
the
voucher
values
get
higher
and
higher.
A
So
there's
always
one
highest
value,
voucher
that
the
provider
of
data
will
have,
and
so
he
submits
that
it's
published
on
chain,
then
12
hours
or
more
later,
the
firefox
can
call
collect
on
the
payment
channel
actor,
and
that
is
what
actually
distributes
the
funds
tears
down
the
payment
channel,
and
so
so,
if
the
transfer
was
successful,
100
of
the
funds
would
go
to
the
firefox.
If
the
transfer
broke
down
halfway,
then
50
of
the
funds
would
go
to
the
firefox.
A
A
I
think
up
here
he
checks
the
validity
of
the
voucher,
which
is
a
signature
check
to
make
sure
the
voucher
is
really
coming
from
the
from
the
chrome
and
that
it's
yeah
that
it's
real
and
also
a
verification
that
the
amount
is
correct,
and
so
I
just
I
just
want
to
wrap
up
by
saying
this
is
an
open
source
project.
Anyone
can
participate.
We
have
probably
about
10
contributors
right
now
and
we
we
are
very
close
to
having
everything
working.
A
I
think
the
one
thing
we
don't
fully
have
working
yet
is
the
retrieval
from
storage
miners,
but
I
think
a
lot
of
that
stuff
will
come
together
this
week
and
I'm
I'm
hoping
to
get
more
people
excited.
So
this
is
the
repo
and
there
are
instructions
for
how
to
install
the
how
to
you
know
you
get
clone
the
repo
and
then
how
to
build
it,
how
to
install
it
in
your
browser
and
there's
a
little
more
information
down
here
about,
what's
complete
and
what's
incomplete,
so
that's
that's
all!
I
have
right
now.
C
C
We
have
the
client
which
are
exactly
the
same
app
we're
just
running
two
different
walls,
so
I'll
go
ahead
and
and
load
some
data
here
in
my
provider,
so
both
are
running
ipfs
nodes,
and
so,
when
I'm
loading
this,
I
get
automatically
a
content,
cid
and
so
for
retrieval
here
we're
only.
We
are
only
supporting
by
content
id,
but
obviously
we
will
be
supporting
different
types
of
naming
systems
in
the
future.
C
So,
as
I
make
a
query
for
this
content,
I
can
immediately
see
this
local
peer,
which
is
found
through
local
discovery.
Right
now,
and
so
I
can
just
kind
of
load
and
get
a
quick
preview
of
the
content
directly
through
ipfs
then
get
the
ask
and
so
the
same
kind
of
parameters
as
mike
outlined,
and
so
we're
ready
to
just
kind
of
hit
retrieve
so
right
away.
We
are
loading,
the
content
and
creating
the
the
initial
payment
channel.
C
First,
we
want
to
kind
of
increase
the
speed
of
retrieval
on
fat
coin,
and
then
we
want
to
enable
anyone
to
just
monetize
their
extra
storage
and
create
noise,
new
sources
of
income,
and
obviously
we
want
to
get
get
those
solutions
to
market
as
soon
as
possible.
And
so
you
know,
we've
been
focusing
on
on
two
challenges
is
as
a
provider.
How
do
you
find
content
to
serve
which,
as
we've
seen
in
the
previous
presentation,
is
definitely
a
challenge
and
then,
as
a
provider?
C
How
do
I
serve
the
best
deals,
which
means:
how
do
I
most
utilize-
the
storage,
the
available
storage,
that
I
have
to
essentially
maximize
the
income?
And
so
our
strategy
is
really
for
our
network
of
providers
to
collaborate
instead
of
competing.
So
how
do
they
collaborate?
They
collaborate
by
sharing
information
and
they
share
information
about
deals
about
content
so
as
soon
as
you've
shared
a
content
id
to
one
peer
in
the
network.
This
information
about
content
is
kind
of
spread,
and
essentially
we
have
this
idea
of
this
concept
of
a
deal
faucet.
C
So
as
soon
as
as
you
join
the
network,
you
automatically
get
propagated
through
gossip
sub,
a
feed
of
content,
and
so
each
any
myopier
can
become
kind
of
a
faucet.
So
what
that
means
is
that
they
run
additional
caching
of
information
through
events,
and
then
they
run
some
type
of
local
optimization
that
we
are,
we
kind
of
we
built
the
prototype
of
it
during
hack,
fs,
actually
over
the
summer,
and
so
now
we're
kind
of
experimenting
with
different
models
to
optimize
this
process,
and
so
right
now.
This
is
very
simple.
C
It's
it's
kind
of
manual
a
little
bit,
but
we
try
to
make
it
really
a
very
sim,
simple,
app
that
anybody
can
download
and
tweak
their
parameters
and
then
gradually
we're
gonna
automize
things.
So
essentially
you
just
kind
of
run
the
app
and,
and
it
automatically
optimizes
the
best
deals
for
you.
So
behind
the
hood,
it's
really
using.
You
know
your
your
normal
ipfs
nodes.
It's
using
each
of
those
two
apps
are
running
p2p
nodes
and
so
we
kind
of
reuse.
C
The
same
kind
of
data
transfer
protocol
as
as
lotus
is
using,
for
example,
when
you
retrieve
from
lotus.
You
have
the
same
kind
of
graph
sync
data
transfer
where
you
can
kind
of
track
the
number
of
bytes
and
get
events,
and
so
what
really
is
it
is
is
interesting
for
us
is
to
kind
of
record
as
much
about
those
events
and
understand
how
the
the
retrieval
deal
happens
and
and
what
what
was
the
speed?
What
was
the
feature,
and
how
can
we
optimize
this?
C
C
You
can
essentially
put
more
funds
in
a
payment
channel,
for
example,
and
reuse,
the
same
payment
channel
to
avoid
having
to
wait,
for
example,
for
this
message,
so
we're
really
kind
of
exploring
different
relationships
between
peers
and
and
ways
to
kind
of
build
trust
in
the
long
term.
To
kind
of
you
know,
go
faster
and
not
have
to
go
through
all
those
different
processes
to
verify
that
content
is
stored,
etc.
C
D
So,
as
we've
kind
of
talked
about
already
looking
at
the
existing
protocol,
this
is
kind
of
the
picture
that
we
see.
We
have
clients
interacting
directly
with
storage
miners
to
both
store
and
retrieve
files,
but
when
we
start
to
look
at
the
secondary
market,
the
picture
starts
to
look
a
bit
more
like
this,
where
we're,
depending
on
these
providers,
that
sit
just
outside
the
immediate
network
and
then
clients
are
interacting
with
them
and
they
are
in
turn
interacting
with
the
network
as
necessary.
D
D
D
One
key
benefit
of
doing
things.
This
way
is
that
we're
really
avoiding
some
of
the
complexity
in
the
core
protocol.
It
provides
all
the
kind
of
fundamentals
that
we
need,
and
then
we
can
really
build
a
more
powerful
and
dynamic
system.
As
an
extension
of
that,
and
one
of
the
really
cool
points
here
is
that
this
data
does
not
necessarily
need
to
be
stored
on
filecoin,
we
can
certainly
construct
a
mechanism
in
which
any
data
can
be
referred
by
cid
and,
assuming
that
it's
profitable
to
provide
you
can
kind
of
expect
it's.
D
Looking
at
some
of
the
key
requirements
for
building
this
sort
of
secondary
retrieval
system,
some
of
these
things
we
already
have
so
content.
Addressing
is
definitely
a
big
one.
It
really
allows
us
to
guarantee
that
the
data
that
we're
retrieving
is
the
data
that
we're
asking
for.
It
also
allows
us
to
address
it
in
a
global
context,
so
that
any
retrieval
provider
who's
coming
onto
the
network
will
immediately
be
able
to
identify
any
kind
of
any
cid
effectively.
D
Looking
at
the
basic
flow
here,
we
have
a
client
she's
looking
to
retrieve
some
data
and
several
providers
that
clients
will
simply
submit
a
query
and
the
providers
will
gossip
that
to
their
peers,
any
provider
that
is
willing
to
offer
the
retrieval
can
do
so
by
directly
engaging
with
the
clients
and
then
the
client
can
select
which
of
those
to
follow
through
with
and
perform.
The
data
exchange
with
them.
D
D
And
then
it's
also
important
to
consider
that
the
locality
of
a
provider
could
definitely
have
a
big
impact
on
this.
Any
one
provider
is
going
to
have
a
very
different
perspective
on
the
network,
and
so
it's
important
to
kind
of
take
that
into
consideration
when
you're
looking
for
optimizations
here,
just
some
very
simple
things
that
we've
done
to
just
kind
of
make
this
a
little
more
usable
and
explorative.
D
Network
our
work
can
be
found
here
on
github
under
chainsafe
bill
secondary
retrieval
markets
in
its
present
state.
It
is
effectively
the
query
and
response
mechanism.
Kind
of
our
next
aim
here
is
to
integrate
some
of
the
existing
go
tooling
for
markets,
so
that
you
can
actually
do
the
retrieval
within
the
same
process,
and
now
that
we've
actually
completed
another
interface.
D
E
E
E
I'm
gonna
connect
it
to
the
lotus
node
and
then
query
for
this
cid
that
I
just
imported
and
then
we
should
be
getting
a
response
from
the
node
okay
cool.
So
we
have
a
response
from
the
secondary
provider
inside
the
node.
So
yay,
that's
that's
good!
Okay!
So
then,
I'm
also
going
to
very
quickly
show
just
the
provider
on
its
own
as
well
so
without
lotus.
E
Okay,
so
so
I'm
gonna
launch,
I
think
I'll
just
do
two.
For
now,
I'm
gonna
launch
one
of
the
providers
with
some
data
or
I'll.
Let
it
know
that
it
has
access
to
some
data.
It
doesn't
really
need
to
necessarily
know
about
the
data.
Yet
at
this
point,
since
it's
just
letting
the
client
know
that
it
has
it
so
sample,
data.json,
okay.
So
I'm
starting
this
with
some
data.
E
E
Okay,
okay,
so
okay
cool,
so
that
one
has
no
data
and
it
is
also
connected.
So
now
I'm
just
going
to
connect
this
and
then
query
for
the
data
that
the
first
one
has.
Then
we
should
only
get
one
response
from
this
guy,
so
yeah
cool.
So
we
got
it
the
parameters
that
it
would
like
for
the
retrieval
yeah,
so
that
is
the
demo
for
now
so
cool.
Thank
you.
A
While
we're
waiting
for
to
see
if
there
are
any
audience
questions
I'll,
just
ask
a
few
questions.
The
first
one
I
had
was
for
the
consensus
team.
I
think
one
of
the
things
you've
put
a
lot
of
thought
into
is
different
parts
of
your
system
where
people
entrepreneurs
could
come
in
and
it
would
be
profitable
for
them
to
you
know.
They'd
have
some
way
to
make
profit.
So
can
you
just
explain
a
little
bit
about
what
the
incentives
are
in
that
system?
Peter.
B
Sure
so
you've
got
for
a
retrieval
gateway.
You've
got
incentivization
for
you
to
do
essentially
the
lookup
between
a
pcid
and
cid
offer.
So
who
can
give
me
that
content
you've
also
got
incentivisation
for
participating
in
the
dht,
though
you,
if
you're
going
to
run
a
gateway?
You
have
to
do
that,
but
you
get
recompensed
for
that.
So
you
get
recompensed
for
seeing
if
you've
got
something
and
if
you
don't
have
it
you'll
still
get
something
and
you
get
recompensed
for
giving
the
content
as
well.
B
A
Okay,
great,
thank
you.
I
also
had
an
actually
another
quick.
Oh
here
I
have
a
question
for
tomah
from
mile.
Can
you
tell
us
a
little
bit
about?
I
know
your
project
started
hack.
Fs,
so
is
it?
Is
everything
open
source
or
is
part
of
it
open
source
like
open
core
or?
What's
what's
your
model
like
there.
C
Yeah,
so
we
have,
we
haven't,
invented
any
new
protocols,
so
it's
basically
using
all
the
available.
You
know
lotus
and
and
fat
coin.
Fpfs
storage
legos.
So
it's
been
really
about.
You
know
stripping
everything
down
to
exactly
using
what
we
needed
so
there's
actually
open
replays
where
most
of
the
the
back
end
that's
happening
is
is
all
available
and
but
the
ui
is
is
closed
source.
So
we
we
have
been
developing
our
own
native
app
on
in
the
private
repo.
A
C
Yeah
so
I've
been,
I
I
used
to
run
on
my
mac,
a
lotus
note,
and
then
I
I
just
ran
out
of
space,
and
so
it
really
made
me
realize
that
I
needed
that
to
be
stored
somewhere
else
so
yeah,
it's
kind
of
the
same
as
as
your
node
very
like
I
mean
right
now.
It's
a
full
node,
but
eventually
I
mean
it's
just
like
a
light.
C
A
A
Apis
yeah
but
but
yeah,
so
that
could
that
could
be
interesting.
We
can
talk
about
that
offline
yeah.
I
just
I
just
wanted
to
ask
chainsafe.
So
what
can
you
talk
a
little
bit
more
david?
Maybe
about
the
the
next
steps.
I
know
you
want
to
decouple
the
storage
miner
from
the
retrieval,
the
retrieval
markets
code,
so
that
you,
you
don't
have
to
also
be
running
a
full
storage
miner
and
just
talk
a
little
bit
about
the
direction
you'd
like
to
go
with
this.