►
Description
Filecoin presents many promising business opportunities for IPFS pinning services. In this talk, Pooja Shah and Juan Benet describe how Filecoin works, how it leads to unique properties in the underlying protocol, network, and token, and how these properties can lead to valuable opportunities for IPFS pinning services and other Web3 infrastructure providers.
Read more about the inaugural IPFS Pinning Summit on the website: https://ipfspinningsummit.com/
For more information on IPFS
- visit the project website: https://ipfs.io
- or follow IPFS on Twitter: https://twitter.com/IPFS
Sign up to get IPFS news, including releases, ecosystem updates, and community announcements in your inbox, each Tuesday: http://eepurl.com/gL2Pi5
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
A
So,
starting
with
a
little
bit
of
the
how
that
file.
Queen
protocol
works
if
I
appoint
as
a
market
that
matches
clients
who
want
storage
services
with
miners
who
provide
storage
services.
The
network
manages
the
interactions
between
the
two
sides
of
the
market,
so
first
clients
can
decide
which
storage
miners
they
want
to
make
deals
with
choosing
across
dimensions
such
as
storage,
price,
available
storage
capacity
and
location,
and
once
the
storage
deal
is
made,
clients
will
transfer
their
data
to
the
miners.
A
Clients
will
also
set
up
lock
up
file
coin
tokens
in
payment
channels
that
they
set
up
to
pay
the
miners
for
successful
storage
over
of
their
files
over
time.
The
file
coin
protocol
repeatedly
asks
miners
to
generate
proofs,
which
we
call
proofs
of
space-time
to
verify
that
the
miners
are
continuing
to
store
the
data
correctly
over
time
and
then,
after
miners
submit
these
proofs
and
they're
verified
to
be
correct.
A
The
network
contracts
will
pay
the
miners
the
correct
amount
of
file
coin
tokens,
and
this
is
a
repeated
process
throughout
the
lifetime
of
a
storage
contract
that
exists
on
the
final
point
network.
In
order
for
the
system
to
work,
clients
need
to
be
able
to
trust
that
the
storage
proofs
work
as
intended
and
that
the
network
protocols
themselves
and
incentivize
honest
behavior
on
behalf
of
all
the
participants
in
the
system,
including
clients
and
miners
infile
coin.
A
The
way
we
enable
this
is
by
using
our
secure
cryptographic
proofs,
which
are
the
proofs
of
replication,
improves
us
face
time
and
also
crypto
economic
mechanisms.
These
systems
work
together
to
ensure
that
all
participants
in
the
network
are
behaving
honestly
and,
ultimately
that,
when
a
client
sends
real
money
to
pay
for
storage
on
a
decentralized
storage
network,
they
can
be
guaranteed
that
their
data
is
being
stored
correctly.
A
So
there
are
certain
mechanisms
in
the
file
Queen
protocol
that
results
in
really
interesting
and
important
features,
network
properties
and
then,
ultimately,
participant
behaviors.
So
I
wanted
to
dive
into
these,
because
I
think
it's
provides
an
important
foundation
as
we
talk
about
some
pinning
service
business
opportunities.
A
Our
expectations
are
that,
around
the
time
of
main
net
launch,
we
should
see
at
least
one
exabyte
of
storage
that
will
join
the
network
around
Maine
at
launch
to
like
one
year
after
main
net
launch
and
then
many
exit
bytes
of
storage
capacity,
to
follow
the
vast
amounts
of
storage
capacity,
just
like
in
any
sort
of
two-sided
marketplace.
This
is
the
supply
side
of
the
market
place
just
having
lots
of
storage
capacity
is
going
to
drive
down
the
price
of
storage
on
the
network,
resulting
in
ultimately
relatively
cheap
storage.
A
The
other
thing
that's
kind
of
unique
about
file
coin
is
that,
in
contrast
to
other
protocols
like
Bitcoin
and
aetherium,
which
require
miners
should
generate
proofs
of
work
which
are
not
necessarily
tied
to
the
underlying
asset
of
the
of
the
system.
File
coin
cryptographic:
proofs
are
what
we
call
useful
proofs
of
work.
This
means
that
miners
are
generating
proofs
over
storage
space,
which
is
the
same
underlying
asset
that
they're
also
selling
on
the
network.
A
So
the
next
one
is,
you
know,
because
we
want
file
coin,
to
be
a
decentralized
protocol
that
has
these
really,
you
know
really
clear:
data
storage
and
availability
guarantees.
This
is
the
reason
why
we've
used
cryptographic
proofs
like
gruesel,
replication
and
space-time,
but
it's
just
you
know
a
property
of
these
cryptographic
proofs
that,
while
they
allow
participants
in
the
system
to
store
network
really
cheaply,
retrieval
on
the
network
is
comparable
slow.
So
you
know
we
have
a
process
in
our
proofs
called.
A
That's
called
unsealing,
so
when
you
first
commit
storage
to
the
network,
you
seal
the
data
in
order
to
retrieve
the
data
you
need
to
unseal
it
and
around
the
time
of
main
that
we're
going
to
have
unsealing
take
on
the
order
of
tens
of
seconds,
and
so
it's
a
lot
slower
than
what
pinning
services
and
ipfs
users
might
be
used
to.
This
is
one
of
the
properties
that
comes
out
of
their
cryptographic
proofs
of
the
system,
but
this
is
also
you
know.
A
I
think
it
creates
a
tremendous
opportunity
for
pinning
services
to
provide
a
lot
of
value
in
the
early
file
core
network
as
well
and
we'll
dive
into
that
a
little
bit
more
later
on
the
file.
Queen
token
is
the
medium
of
exchange
in
the
file
coin
storage
network,
and
so
this
means
that
the
tokens
underlying
value
is
also
fundamentally
tied
to
the
asset
that
it
is
used
to
pay,
for.
A
You
can
use
a
cerium
to
pay
for
file
Quinn
storage
one
day
but
or
you
know,
even
in
the
short
term,
but
you
can
also
use
file
coin
to
pay.
For
a
final
point,
storage-
and
this
has
a
lot
of
interesting
properties
as
well,
and
then.
Lastly,
the
file
Queen
protocol
has
implicit
incentives
for
a
geographic
distribution
of
network
notes.
So
the
reason
for
this
is,
you
know
many
clients.
Storage
clients
are
much
more
likely
to
hire
storage
miners
that
are
geographically
closer
to
them.
A
A
A
But
you
know,
as
we
all
know,
ipfs
appending
services
provide
this
really
important
service
for
applications
in
the
ecosystem
that
want
to
use
if
you
centralized
peer-to-peer
content
address
storage
network
network,
but
who
also
want
data
persistence,
and
so
this
list
view
of
you
know.
A
generalized
pinning
service
looks
like
an
IP
FS
painting,
API,
maybe
similar
to
the
one
that
one
discussed
earlier,
that
developers
can
call
into
from
their
platforms
and
and
from
their
applications,
and
the
pinning
service
itself
holds
all
of
the
business
logic.
A
That's
necessary
to
manage
user
accounts
and
user
pin
sets
as
well
as
managing
the
underlying
IP
FS
notes.
So
the
first
integration
opportunity
that
is
pretty
straightforward,
is
it's
just
taking
all
of
the
data
that
is
currently
stored
on
ipfs
nodes
and
backing
it
up
to
the
file
point
Network.
The
second
option
is
which
you
know
deserves
a
little
bit
more
explanation
and
we
think
of
these
as
truly
file
point
backed
ipfs
pinning
services.
A
In
this
model
you
can
think
of
ipfs
kind
of
as
the
hot
storage
layer
or
caching
layer
and
the
file
coin
network
as
the
colder
storage
layer.
The
way
that
we
envision
some
service,
like
this
working
with
using
file
coin
storage,
is
the
when
an
application
makes
a
storage
request
to
an
IP
FS
pinning
service.
The
service
automatically
will
store
data
on
both
the
IP
FS,
hot
storage
layer
and
the
file
coin.
A
Colder
storage
layer
and
the
data
may
remain
on
the
ipfs
side,
IFS
storage
layer
for
some
time
until
it's
cleared
out
by
some
sort
of
intelligent
garbage
collection
strategies,
and
then,
after
that
point
that
particular
CID
would
only
be
stored
on
the
file
coin
network.
Then,
when
the
data
needs
to
be
received,
sorry
retrieved,
pinning
services
can
pull
the
data
from
the
hottest
layer
that
has
the
storage.
A
A
You
know
intelligent
caching
strategies
to
determine
when
does
it
make
sense
to
pull
data
back
from
file
coin
into
this
hot
storage,
layer
and
so
on,
and
these
caching
layers
are
also
an
interesting
point
of
differentiation
for
pinning
services
that
might
choose
to
adopt
this
kind
of
architecture.
But
there
there's
also
relevant
to
note
that
both
IP
FS
and
file
coin
transfer,
I
peeled
ed
I,
peeled
D
data
objects
back
and
forth
with
each
other,
and
they
also
both
speak
lat
p2p
protocols.
A
So
the
sort
of
manipulation
of
data
and
networking
between
IP
FS
and
file
coin
networks
should
be
pretty
straightforward
and
then
I'm
in
depending
on
whatever
architecture
you
decide
to
use.
So,
let's
talk
about
why
this
is
compelling
first
file,
Queen
back
skinning
services
will
still
offer
sub-second
retrieval
to
existing
pending
services.
You
service
users
that
depend
on
this
sort
of
fast
sub-second,
retrieval
speed,
but
the
other
benefit
is
that
these
sorts
of
pinning
services
will
allow
greater
alignment
with
your
web
3
application
users
all
of
the
underlying
data.
A
It's
a
compelling
business
opportunity
to
in
a
number
of
ways
you
know.
First,
there's
just
as
we
mentioned,
five-point
storage
is
expected
to
be
pretty
cheap
and
especially
for
data
that's
stored
and
frequently
we
anticipate
that
the
cost
will
be
quite
low,
and
these
sorts
of
cost
savings
will
both
make
the
operations
cost
of
running
a
pinning
service.
A
lot
cheaper
and
also
these
cost
savings
can
be
passed
on
to
pinning
service
users,
and
that's
that
will
also
create
this
interesting
flywheel.
A
You
know
if
you're
able
to
pass
on
cost
savings
to
users
might
be
able
to
grow
the
market
for
pinning
service
users
as
well
and
allowing
you
to
then
have
additional.
You
know:
bulk
storage
at
cheaper
cost
and
so
on,
and
so
we
think
that
that's
like
a
really
compelling
opportunity
if
you're
able
to
use
file
coin
as
the
underlying
storage
layer.
The
other
reason
is
that
pinning
ipfs
data
on
a
decentralized
network
well
like
file
coin,
will
strengthen
the
pinning
service
value
prop
within
their
growing
web
3
market
of
applications.
A
We
expect
this
to
exist
organically
kind
of
just
because
of
this
greater
product
market
fit
and
user
alignment
that
we
discussed
earlier.
But
it's
also
because
you
know
the
file
crank
project
is
starting
to
engage
in
business
development
efforts
to
expect
expand
the
file
point
user
base,
and
in
fact
this
is
already
happening.
We
plan
to
direct
application
developers
to
pinning
services
that
are
backing
up
or
storing
their
data
on
file
point.
So
this
would
be.
A
You
know
just
like
cumulus,
collectively
growing
all
of
these
ecosystems
with
effort
from
number
a
number
of
groups,
including
the
Battle
Point
project,
and
then
another
benefit
is
that
a
lot
of
the
tooling
that's
needed
in
or
in
order
to
enable
these
file
Queen
backs.
Pinning
services
is
also
needed
to
enable
other
business
opportunities.
A
You
can
reach
out
to
me
directly
pooja
in
order
to
get
more
information
on
what
these
grants
look
like,
or
you
can
also
email,
Deb
ramps
at
50.0
org,
one
easy
way
to
get
started
with
building
this
type
of
file.
Crane
back
to
connect
service
is
to
use
textiles
power,
gate
which
andrew
is
and
Aaron
are
going
to
talk
a
lot
more
about
in
the
next
presentation.
But
the
high
level
is
that
power,
power,
gate
embeds
and
ipfs
node
and
a
file
coin
node,
and
it
exposes
this
really
exposes
this
really
intuitive
API.
A
To
interact
with
both
target
is
really
valuable
because
it
abstracts
away
a
lot
of
the
complexity
of
storing
data
on
the
file
point
network,
such
as
how
do
you
select
miners?
How
do
you
maintain
storage
deals
over
time
and
once
the
storage
contract
has
expired?
How
do
you
renew
that
storage
contract?
A
The
next
opportunity
I'd
like
to
talk
about
is
storage
mining,
so
we
believe
that
there's
a
massive
opportunity
for
file
coins,
storage,
mining.
Some
estimates
expect
that
by
2025
we're
gonna
be
generating
something
like
have
generated:
175
Zeta
bytes
of
data
in
the
world,
while
we'll
only
have
around
four
and
a
half
Zeta
bytes
of
data
storage
supply,
that's
I
think
the
diff
between
those
two.
It's
it's
around
40
X,
more
data
that
we
are
expected
to
generate.
A
Then
we
will
have
space
for
and
so
clearly
there's
a
huge
opportunity,
even
if
you
assume
the
appropriate
haircuts
and
maybe
we're
not
actually
going
to
generate
175
data
bytes
of
data.
But
it's
like
half
or
quarter
of
that,
and
even
if
a
lot
of
that
data
is
discarded,
there
will
clearly
still
in
the
future,
be
large
demand
for
storage
services,
and
this
is
not
in
the
far
future,
but
over
the
next
few
years.
A
File
coin
as
a
market
enables
a
host
of
services
to
put
storage
up
on
the
network.
You
know
this
is
this
could
be
the
top
three
cloud
storage
providers,
but
it
also
creates
an
opportunity
for
a
host
of
other
groups.
You
know
other
cloud
storage
providers
that
maybe
aren't
you
know
the
large,
the
largest
players
in
the
ecosystem,
but
also
service.
You
know,
businesses
that
do
storage,
backups,
telcos
and
ISPs
data,
warehouses,
storage,
hardware,
manufacturers,
consumers
who
may
have
their
own
sort
of
setups
at
home,
crypto
miners
and
also,
we
believe,
pinning
services.
So
file.
A
Queens
aim
is
to
provide
a
marketplace
where
players
in
any
of
these
verticals
can
provide
their
storage
capacity
for
rent
on
a
marketplace
and
abstracts
it.
You
know
it
straps
away
all
of
the
marketing
meeting
to
maintain
nodes
in
different
regions
and
so
on
and
provides
a
really
seamless
user
experience
for
storage
providers
that
want
to
sell
storage
to
users
who
need
it
touched
on
this
a
little
bit
earlier.
But
the
network
compensates
storage
miners
in
a
number
of
ways
and
just
to
touch
on
a
few
of
these.
A
You
know,
though,
probably
at
least
early
on,
we
anticipate
a
large,
a
large
part
of
the
rewards
or
the
compensation
for
storage.
Miners
will
come
through
block
rewards
and
these
are
earned
proportional
to
the
amount
of
data
that
you
as
a
miner
would
be
providing
to
the
network.
So
locker
words
we
anticipate
are
going
to
be
extremely
profitable
or
early
on,
and
so
that's
like
just,
but
that's
just
one
portion
of
the
file
coin
storage,
miner,
business
opportunity,
the
other
is
transaction
fees
for
processing
messages.
A
A
All
of
these
messages
need
to
be
processed
into
final
point
blocks
and
storage
miners
or
any
miners
on
the
network
that
produce
blocks
can
earn
transaction
fees
for
doing
this
work,
a
block
reduction,
and
then
the
third
is
it's
probably
the
most
important
like
definitely
in
the
long
term,
but
file
coins.
Forge
miners
also
get
paid.
A
Of
course,
you
know
for
the
service
of
storing
and
retrieving
data
on
the
network,
and
so
the
payments
that
come
over
time
from
the
network
contracts
as
miner
submit
their
proofs,
the
space/time
they
get
paid
in
file
coin
tokens
as
well.
Those
deal
payments
are
a
recurring
source
of
revenue
for
storage
miners
that
provides
storage
services
on
the
network
as
well.
So
you
know,
there's
a
lot
more.
A
We
could
say
on
the
file
quick
storage
mining
site,
it's
not
directly
at
I
into
pinning
services
and
what
pinning
services
do
today,
but
I
think
there's
a
really
compelling
opportunity
here
and
especially
if
your
service-
or
you
run
a
service
that
has
you
know,
storage
capacity
that
you
maintain
or
some
of
it
as
idle.
While
coin
storage
mining
could
be
a
great
opportunity
for
you
as
well
I'm.
B
Going
to
talk
about
the
how
to
take
payments,
how
to
operate
a
serial
market
operator
and
and
how
to
consider
potentially
selling
other
kinds
of
services
so
for
the
competence
part,
if
you're
familiar
with
that
with
a
lot
of
crota
payments,
then
this
is
going
to
be
very
similar
to
to
everything
else.
We
were
used
to.
We
followed
all
the
normal
standards
around
around
assigning
and
so
on.
B
B
And
you
end
up
with
with
Parkman
tokens
the
once
you
have.
These
tokens
are
kind
of
these
units
of
account
on
on
the
blockchain.
They
end
up
in
a
kind
of
spread
out
in
in
in
potential
I'm
a
number
of
accounts.
If
people
are
having
many
different
machines
running
this
and
so
on,
and
and
so
there
are,
these
kind
of
tools
called
being
across.
There's
that
then
I
kind
of
gather
that
together
and
then
cannot
do
accounting
for
figuring
out
what
what
deals
actually
happen
and
and
and
whatnot.
B
And
then
you
can
clear
the
and
then
you
can
effectively
trade,
the
trade,
the
popcorn
for
yup,
when
people
have
a
choice
of
whether
or
not
who
those
my
name
is
hang
on
to
that
pal
coin
and
then
use
it
in
the
network
or
something
or
then
to
kind
of
sell
it
to
two
clients
and
users
that
didn't
want
to
use
it
for
for
storage
right.
So
this
is
a
the
medium
of
exchange
can
be,
can
be
sold
through
through
a
will
will
be
kind
of
sellable
through,
through
whatever
systems
kind
of
support
that
I.
B
Imagine
that
a
it's
likely
that
in
aetherium
there
will
be
kind
of
some
like
wrapped
wrapped
flop,
going
kind
of
thing,
and
so
people
will
be
able
to
use
unit
swap
and
all
that
kind
of
stuff
to
to
move,
to
move
the
tokens
around,
and
so
this.
This
means
that
it.
It
ends
up
fairly
easy
for
parties
to
accept
falcone
payments,
with
use
of
payment
processor,
to
figure
out
what
what
happened
and
then
move
to
whatever
medium
of
whatever
sort
of
value.
B
They
prefer
to
keep
whether
it's
keeping
giving
five
point
or
being
some
other
cryptocurrencies
or
moving
to
a
stable
coin
or
moving
to
normal
fiat
currency,
either
by
going
through
a
currency
on
on
like
something
like
a
USC
T
or
something
like
that
or
or
directly
clearing
through
through
some
kind
of
fiat
exchange,
or
something
like
that.
So
those
kinds
of
systems
are
on
gonna,
be
kind
of
online.
B
At
some
point,
and
and
so
it
means
that
it's
very
fairly
easy
for
a
pinning
service
or
provider
to
accept
Popham
payments
without
having
to
kind
of
bare
cryptocurrency
market
risk.
If
you're,
not
if
you
don't
wanna,
do
that.
So
you
can
clear
it
to
whatever
the
currency
of
choice
is
and
get
the
popcorn
into
the
hands
of
clients
so
that
they
can
use
it
for
for
paying
for
for
the
storage
and
there's
a
number
of
like
libraries
and
tools
and
so
on.
Good
that'll
make
all
of
us
all
this
easy.
So
the
next
part.
B
So
the
remark-
and
we
keep
kind
of
talking
about
this-
the
way
to
think
about
it
is
that
if
you
have
the
ability
to
have
a
fully
open
market,
where
any
party
that
hears
about
a
file
being
requested,
a
lot
can
grab
it
and
then
serve
it
to
users
downstream.
Then
you
would
get
a
sort
of
automatically
scaling
CDN
that
responds
to
to
demand
on
the
network
right.
B
You
know
kind
of
close
to
them,
and
so
this
is
where
geographical
location
will
matter
a
lot,
because
it's
it's
not
about
kind
of
having
a
lot
of
storage
is
really
about
exactly
where
you
are
in
the
network.
And
how
quickly
can
you
deliver
something
to
do
the
parties
requesting
it?
It's
not
only
the
long-term,
though,
in
order
to
get
there
a
whole
network
has
to
be
built
out
of
these
virtual
parties
who
are
spread
out
across
the
globe.
So
think
of
this
as
kind
of
day.
B
That's
way
too
slow.
You
want
to
be
able
to
do
a
payment,
real-time
in
one
round-trip,
so
you
can
think
of,
like
you
know,
a
large
gossip
network
where,
where
requests
are
being
are
being
forwarded
and
know
if
somebody
responds
to
the
request
and
they
don't
pass
it
on
to
other
parties.
So
you
kind
of
see
kind
of
like
the
hotspots
of
content
being
addressed
by
by
the
content
moving
closer
to
those
those
areas.
B
And
so
then,
there's
a
whole
set
of
other
potential
services
that
that
can
be
kind
of
sold
through
the
network
and
into
this
a
whole
bunch
of
different
kinds
of
developer
processes.
I
like
to
call
them
and
the
be
necessary.
So
things
like
aggregation,
as
which
I
mentioned
so
the
the
normal
Falcone
stored
long-term
storage
I,
think
it
takes
a
while
to
seal
steel
sectors
and
in
takes
a
while
to
kind
of
store
that
sort
of
mean
kind
of
this
long-term,
long-term
storage.
And
so
that
means
that
it
means
that
doing
it
in
doing.
B
Storage
in
bulk
is
a
lot
cheaper,
trying
to
transaction
wise.
So
it
means
that
if
parties
want
to
sort
very
little
amounts
in
data,
it's
useful
to
first
kind
of
put
them
all
in
like
a
long
log
and
IRA
gave
them
along
with
other
pieces
of
data
and
then
batch.
All
of
that
and
put
all
of
that
together
into
a
deal.
B
So
you
can
think
of
normal
computers,
like
the
CPU
needing
kind
of
like
the
really
really
hot
memory
and
being
able
to
kind
of
pin
in
like
the
local
cache
and
then
going
out
to
ram.
If,
if
you
need
to
when
when,
when
a
you
need
kind
of
like
a
larger
address
space
or
going
out
to
disk,
if
and
so
on,
and
each
of
these
steps
is
dramatically
slower.
B
But
it
gives
you
access
to
much
more
memory
so
think
of
a
file
in
the
same
way
to
where
there
are
different
kinds
of
service
providers
at
different
tiers,
providing
different
parts
of
the
equation.
So
there's
like
long
term
really
cheap
storage,
that's
potentially
in
some
specific
parts
of
the
world,
then
there
are
much
faster
caches
that
are
closer
to
users
that
are
responding
to
the
man
only
so
the
things
are
not
doing
that
they
mattador
there's
a
kind
of
a
prior
agreement
to
keep
it
hot
things.
B
So
a
bunch
of
the
techniques
have
direct
analogs
you
can
even
think
of
like
so
the
term
pinning
and
ipfs
comes
from
the
idea
of
pinning
a
memory
page
in
the
CPU
and
keeping
it
really
hot.
So
the
idea
of
opening
service
keeping
things
really
hot
is
like
super
apt.
So
it's
like
exactly
like
a
pizza.
That's
a
bill.
B
There
are
also
other
kinds
of
things
that
the
price
can
provide,
and
so
this
is
think
of.
Think
of
there
being,
you
know,
classes
services
around
mining,
so,
for
example,
miners
have
to
do
this
set
of
expensive
computations
in
GPUs
that,
in
order
to
produce
some
of
the
proofs,
smaller
miners
are
not
going
to
want
to
buy
a
lot
of
GPUs
to
do
these
kinds
of
computations.
So
they're
gonna
hire
some
other
parties
to
to
run
these
these
these
computations.
B
A
B
Suddenly
so
so
that's
exactly
exactly
what
what
an
excellent
through
minor
would
would
do
so
so
really
think
of
kind
of
how
the
current
CDN
CDN
market
works
and,
and
just
think
of
like
mirroring
that
and
in
in
an
open
marketplace
where
we're
on
you
don't
have
to
write.
You
know
you
don't
have
to
have
a
huge
organization
and
write
contracts.
You
can
just
take
a
ticket
machine
and
set
it
up
point
it
to
the
blockchain
run
some
code,
and
then
it
spits
out
cryptocurrency
and
that's
it
like
that's.
That's
kind
of
like
really.
B
My
question
on:
do
you
see
storage
minor
separation
between
the
GPU
seeking
to
the
CPU
storage
only,
and
so
on
some
amount,
so
it
is
very
likely
that
that
large
scale
miners
are
gonna
want
to
have
their
whole
setups
in-house
and-
and
the
reason
is
that,
once
you
get
to
a
certain
scale,
you
really
need
those
operations
to
run
very
efficiently
and
trusting
another
group
to
to
do.
It
might
be
maybe
potentially
like
well
yeah
party,
just
missing
delivering
a
second
or
two.
B
You
might
like
cause
you
to
lose
a
block
which
then
loses
a
block
reward,
which
is
adds
up
really
quick.
So
we
expect
the
kind
of
large
might
as
well
we'll
just
kind
of
run
the
most
of
the
operations
in-house.
That's
that
medium
and
smaller
miners
are
unlikely
to
to
have
to
produce
that
many
ZJ
Starks,
and
so
they
did.
B
They
just
need
to
produce
them
quickly
enough,
but
at
a
given
moment-
and
so
this
is
exactly
what
we're
a
service
could
be-
could
be
extremely
useful,
where
we
sort
of
expect
to
see
these
kind
of
like
SAS
services.
Snarks
ask
the
service
things
popping
up
where
you
know
you
just
submit.
You
know
it's
kind
of
like
an
RPC.
You
submit
like
the
data
you
need
and
I
pop
them
in
and
then
you
get
back
the
the
snark
computed.
B
Pay
me
channels
just
small,
sequential
payments,
or
do
they
have
some
form
of
status
in
the
theorem
state
annals
today,
it's
mostly
sequential
payments
I
believe
it
had
a
memo
field,
I'll
go
back
and
look.
We
debated
like
a
fair
amount
of
time.
They
don't
know
might
have
gone
removed,
but
for
now
it's
mostly
just
payments.
We
want
to
do
stay
channels
at
some
point,
but
we
also
need
to
needed
to
keep
the
complexity
of
pot
going
down.
B
So
that's
one
of
the
the
Buckman
chain
ends
up
being
way
more
complex
than
then
something
like
a
Karen
because
of
a
cruise
and
the
consensus
and
so
on.
That's
why
we
kind
of
didn't
have
don't
have
a
p.m.
either
because
launching
all
of
that
all
those
layers
all
at
once
is
a
much
riskier.
So
we
want
to
kind
of
defer
that
out
to
the
to
later
on.
B
So
a
penny
service
is
like
an
advanced
betrayal,
minor,
longer-term
small
rill,
but
also
become
a
lighter-weight
homegrown,
pinning
service
and
CD
on
eve,
easy
enough
to
self
Costanera
yeah,
that's
right,
so
so
think
of
like
finding
services
fitting
like
the
right
spot
to
be
serving
kind
of,
like
really
sort
of
like
the
s3
use
case
now,
and
maybe
later
the
quad
front
use
case.
But
then
smaller
smaller
virtual
miners
might
be
even
closer
to
end-user,
so
they're
just
gonna
have
less
of
the
market.
It's
really
really
useful
to
think
of
it
as
like.
B
A
big
grapevine
right.
So
the
internet
is
shaped,
like
a
like
a
very,
very
large
grapevine
with
many
many
many
tiers,
and
so
you
can
think
of
anywhere
in
the
vertices
in
one
direction.
You
have
users
downstream
from
you
that
are
requesting
data
or
potentially
uploading
data
and
the
other
direction
you
have.
You
might
have
like
that.
B
So
both
very
high
levels,
where
you
need
very
high
performance
machines
and
you're
dealing
with
massive
amounts
of
traffic
because
tons
of
people
downstream
from
you
or
a
closer
and
closer
to
the
user.
Where
suddenly
latency
is
the
advantage
late,
you're
able
to
serve
things
extremely
quickly
to
another
party
close
to
you,
but
you
have
a
much
smaller
slice
of
the
market
downstream
from
you
and
so
that,
therefore
you
don't
you're
yeah
you're.
It's
probably
like
a
limited,
limited
level
of
revenue.
B
Now
that
what
we
would
like
to
get
is
just
an
automatic
scaling
where
content
flows
to
the
right
vertices
through
the
grapevine
where
it
really
should
be
to
kind
of
optimize
the
whole
distribution
based
on
what
what
is
actually
being
requested.
Real-Time-
but
you
know
getting
there-
is
gonna-
require
some
so
amount
of
time
with,
like
really
perfecting
those
those
kind
of
announcement
channels
and
all
that
kind
of
stuff.
B
And
so
you
can
use,
you
can
select
miners
based
on
a
lot
of
features,
so
price
is
likely
to
be
one
of
the
main
features
by
which
people
select
miners.
Well,
there's
a
whole
host
of
other
reasons
like
you
might
want
it
to
be
specifically
in
this
particular
region,
or
your
miners
might
offer
other
kinds
of
things
like
sir
kinds
of
security
thresholds.
So
some
of
the
stuff
that
we've
been
thinking
about
is
that
a
lot
of
data
really
requires
certain
degree
of
confidentiality
or
or
or
or
regulatory
levels
of
security.
B
It's
likely
that
phones
may
be
okay,
retrieval
miners
and
but
let
people's
phones.
If
you
keeping
connected
a
peer-to-peer
network,
often
I've
got
a
lose
brain
battery
and
use
a
lot
of
data.
So
people
probably
won't
do
that.
What
a?
But
there
are
a
lot
of
phones
like
many
more
phones
than
there
are
people
trying
to
use
them
so,
and
they
usually
have
really
good
GPUs
and
really
could
get
a
good
machinery.
B
So
I
wouldn't
be
surprised
if
people
set
up
like
farms
with
phones,
eventually
with
to
do
different
kinds
of
processing,
but
I
wouldn't
expect
it
to
be
people's
normal
phones.
I
think
what
phones
will
be
good
for
is
moving
around
content
in
almost
offline
settings,
where
the
fact
that
you're
close
to
another
person's
phone
means
that
you
have
access
to
the
network,
even
if
you're
not
close
enough
to
an
attendant
yourself.