►
Description
Juan Benet talks about composable market primitives at the Retrieval Market Builders Mini-Summit held in April 2021.
A
All
right,
so
this
is
going
to
be
a
a
session
on
order.
Books
and
bids
asks
deals,
and
so
on
and
kind
of
trying
to
arrive
at
a
very
composable
framework
for
representing
request,
reply,
requests
and
offers
for
storage
and
all
that
kind
of
stuff,
and
this
is
not
very
prescriptive,
but
it's
rather
more
vision,
oriented
and
emphasizing
kind
of
the
direction
that
defy
has
taken
in
the
last
year
and
a
half
and
just
the
explosion
that
was
facilitated
by
arriving
at
good
primitives.
A
A
So
what
to
start
off
with
kind
of
pointing
to
commodity
markets,
that
one
of
the
big
things
that
that
falcon
is
doing
is
just
making
space
time
a
commodity
and
and
just
digital
storage
is
to
today
not
not
really
commodity.
You
have
to
kind
of
create
all
these
service
contracts
and
so
on.
It's
not.
Actually.
A
So
there's
a
a
ton
of
stuff
going
on
many
different
kind
of
problems
being
solved
through
you
know.
Well-Composed
contracts
and
this
kind
of
explosion
in
in
the
design
space
has
been
made
possible
by
just
being
able
to
make
a
thing
ship
it
and
then
let
other
teams
go
and
solve
some
other
problem
downstream,
as
opposed
to
trying
to
kind
of
build
one
one
large
large
thing
so
the
long-term
goal.
So
it's
been
kind
of
like
a
long-term
goal
for
for
falcon.
A
You
get
to
you
really
make
these
proper
markets
right,
make
a
proper
storage
market
make
a
proper
retrieval
market
and
so
on.
So
you
can
think
of
you
know
the
the
requests
for
for
storage
or
the
request
for
for
piece
of
data
as
bids
and
masks
on
on
a
market
for
either
storage
space
or
for
the
data,
the
relevant
data
that
parties
are
looking
for
and
like
today,
the
the
it's
pretty
primitive
like
you
only
have
deals
represented
on
the
chain
we
ended
up.
A
We
had
implemented
a
lot
of
the
order,
book
dynamics
with
sort
bids
asks
and
so
on,
and
we
ended
up
ripping
those
out
and
just
in
order
to
kind
of
get
to
get
the
launch
and
so
on,
but
it's
time
to
kind
of
go
back
and
revisit
a
lot
of
this
stuff
and
and
and
start
using
it
yeah.
So
so
this
is,
you
know,
kind
of
the
old
falcon
paper
has
kind
of
a
lot
of
it's.
A
What
we
ideally
could
arrive
at
is
some
very
straightforward
definition
of
what
an
ask
is
what
a
bit
is
and
what
a
deal
is
that
are
all
extensible
to
represent
the
different
components
that
that
parties
might
might
want.
So,
for
example,
some
requests
for
storage
might
include
my
prescribed
price,
or
that
might
not
be
there.
It
might
be
a
request
on
a
specific
region.
There
might
be
other
features
that
people
people
care
about
right,
so
things
like
certain
certifications
around
the
storage.
A
So
like
api
compliance
and
so
on
or
a
miner
might
be
offering
some
like
added
services,
along
with
the
with
the
the
proposal
rating,
so
the
the
one
of
the
things
that's
true
about
all
these.
These,
these
ask
bits
and
and
potential
deals
is
that
they
need
some
way
of
encoding,
a
feature
space
that
clients
and
miners
might
care
about,
but
that
not
all
clients
and
not
all
miners
care
about.
So
it's
it's
some.
A
A
A
A
The
idea
the
pc
id
is
a
is
a
it's
an
obstruction
that
is
required
by
the
proofs
and
then
there's
notions
like
the
price,
the
start
time
and
end
time,
collaterals
and
and
so
on.
But
there's
there's
also
like
a
set
of
other
properties
that
that
people
might
care
about
that.
Already
people
care
about
quite
a
bit
and
they're
feeding
into
their
reputation
market.
Things
like
the
world
region
in
which
the
data
is
going
to
be
started.
A
Content
policy
so
mutually
agreeing
upon
a
policy
that
governs
the
content
to
kind
of
serve
to
deal
with
dispute
resolution
on
bad
bets,
potentially
certifications
like
hipaa
compliance
and
others
that
that
people
might
might
need
to
require.
Then
some
deals
might
be
self-contractible
and
some
deals
may
not
be
so.
A
A
So
you
know
the
same
kind
of
basics
and
in
this
case
it's
not
a
request
for
data
storage,
but
you
know
for
the
space
time
it's
a
request
for
the
data
itself,
and
so
here
the
cds
that
the
people
care
about
might
be
the
deal,
maybe
the
pcid
or
might
just
be
the
payload
cids
or
some
underlying
cds
that
are
within
within
a
deal
or
hey.
By
the
way,
this
the
cids
might
not
be
at
all
stored
in.
A
In
the
pipeline
storage
market,
the
these
cids
might
be,
you
know,
stored
in
like
an
iphone
spinning
service
or
some
other
component.
So
there's
a
possibility
here
that,
like
decoupling,
these
markets
and
not
forcing
that
the
retrieval
cids
are
specifically
deal
pcids
and
so
on
is
is
a
very
useful,
useful
way
to
make
the
retrieval
market
much
more
yeah
be
able
to
leverage
in
other
applications
and,
again,
you
know
different
kind
of
possible
properties.
A
I
didn't
listen
to
here,
but
a
lot
of
them
have
been
talked
about
in
the
last
few
days.
You
know
like
one
as
an
example
is
being
able
to
kind
of
have
the
ability
to
kind
of
use,
prepaid
vouchers
and
so
on,
where,
like
the
payment
comes
from
a
third
party
like
the
publisher
as
opposed
to
the
the
party
retrieving
and
so
on
right.
So
then
these
these
requests
would
potentially
like
look
different.
A
So
one
other
kind
of
interesting
way
to
look
at
this
whole
space
is
to
think
about
the
the
kinds
of
markets
that
might
develop
so,
for
example,
the
the
most
important
thing
like
the
most
the
most
people
have
been
talking
about.
So
far
is
going
from
one
chain
to
off
chain.
So,
like
that's,
you
know
one
straightforward
view,
but
there's
other
kinds
of
features
that
that
might
be
pretty
relevant.
A
So
just
looking
at
it
in
terms
of
the
number
of
market
participants
against
the
like
deal,
size
or
complexity
of
the
deal
already
yields
very
different,
different
categories
of
markets.
So
when
there's
not
that
many
market
participants
and
the
deals
are
small
or
the
complexity
of
the
deal
is
very
small,
you
can
get
away
you
can
you
can
get
really
really
far
with
just
basic
basic
deal
defaults.
A
As
soon
as
you
start
increasing
the
size,
though
like
if
you
get
into
terabytes
or
petabytes,
then
at
that
point
clients
care
a
lot
more
about
where
the
data
is
going
to
be,
who
the
miners
really
are.
What
kind
of
other
services
they're
going
to
allow
if
they
need
to
do
any
kind
of
offline
shipping?
Of
hard
drives,
that's
a
much
bigger
deal,
so
you
go
from
like
the
basic
defaults
in
the
tools
to
now
needing
marketplaces
and
marketplaces
that
are
not
automated
a
lot
of
times.
These
are
kind
of
more
oriented
towards
humans.
A
Where
and
kind
of
like
we
saw
phil
swann
as
an
example
a
couple
days
ago,
as
kind
of
a
marketplace
where
you
can
see
all
the
different
miners,
you
can
see
even
their
picture
and
you
can
see
their.
You
know
the
properties
that
they're
they're
kind
of
proposing
and
so
on,
and-
and
you
know
when,
when
there
is,
you
know
high
complexity
to
these
deals,
and
you
have
a
lot
of
or
or
they're
really
big.
A
A
There's
so
much
at
stake
here
and
and
so
much
cost
associated
with
each
deal
that
at
that
point
the
parties
will
want
to
know
much
more
about
each
other,
so
kind
of
marketplaces
start
making
sense
now
in
the
in
another
part
of
the
access.
If
the
deals
are
not
very
complex
like
they
don't
require,
a
bunch
of
complex
features
is
mostly
basic,
but
there's
a
lot
of
market
participants
and
the
data
is
small.
Then
you
can
start
doing
automated
clearinghouses
and
exchanges
right.
A
A
That's
really,
where
kind
of
like
the
realm
of
d5
comes
in,
and
so
on
and
sort
of
expect
a
lot
of
that
to
be
a
big
part
of
the
falcon
world
in
in
the
in
the
future.
And
don't
know
exactly
when,
but
but
it's
sort
of
dependent
on
a
lot
of
the
things
that
people
are
talking
about
getting
implemented
but
but
yeah,
that's
sort
of
like
where
that's
that's
headed,
so
you
can
do
you
know?
A
Kind
of
a
picture
of
the
world
might
be
something
like
this,
where
there
are
a
bunch
of
different
markets
with
asks
bids
and
deals
coming
out
of
those
markets.
Some
of
these
might
end
up
on
chain.
A
lot
of
them
may
not,
especially
in
the
retrieval
side,
and
most
of
that
stuff
is
not
not
going
to
hit
the
chain
only
kind
of
the
final
payment
vouchers
might,
and
you
can.
A
You
know
then
start
looking
at
some
of
the
other
things
going
on
in
the
ecosystem,
with
this
kind
of
market-oriented
view
and
say
hey,
maybe
all
of
these
things
have
an
associated
five
coin
market
which
is
sort
of
like
this
bridge
component
or
maybe
when
we
talk
about
retrieval
and
topologies
for
retrieval
and
hubs
and
so
on.
Maybe
all
of
these
things
are
kind
of
like
a
small
little
market
and
whether
or
not
you
connect.
A
This
is
again
a
separate,
separate
question
and,
and
yes,
then
interesting
gears
for
a
moment.
Here's
like
some
of
the
stuff
that
we've
been
talking
about
in
the
last
couple
of
days
like
different.
This
is
ways
of
collecting
information
on
miners
that
ends
up
being
fed
into
markets
and
marketplaces.
Yeah.
Here's
like
the
phil
swann
example
with
with
the
various
features
of
the
miner,
and
so
it's
for
these
marketplaces,
especially
it's
cool
too.
A
It's
good
to
look
at
other,
very
successful,
worldwide
global
marketplaces
and
and
see
kind
of
what
what
parts
of
the
product
might
be
relevant.
So,
for
example,
everybody
has
like
a
very
domain
specific
search
where
you
can
filter
for
the
features
that
that
that
the
consumers
of
the
market
marketplace
really
care
about,
and
you
can
visually
compare
a
bunch
of
the
of
the
offers
right
like
the
kind
of
like
service
providers.
A
Then
you
know
they
have
like
quantifiable
ratings
and,
like
that's,
really
useful
we're
already
there
like,
we
have
a
much
more
sophisticated,
we're
building
towards
a
much
more
sophisticated
set
of
ratings
airbnb,
because
it's
mostly
humans,
transacting
with
each
other,
has
a
whole
bunch
of
qualitative
ratings
and
there's
a
lot
there,
because
policies
and
and
requirements
sometimes
don't
fit
like
the
are
very
nuanced,
and
so
there's
there's
more
of
a
need
there.
This
might
be
relevant
for
market
falcon
marketplaces
where
there's
huge
amounts
of
data
involved
like
so.
A
If
a
party
wants
to
hire
miners
to
store
tens
of
petabytes,
then
there
might
be
a
lot
of
qualitative
feedback
about
about
that
relationship
that
might
be
relevant
to
other
other
parties
like
that.
So
again,
maybe
marketplaces
are
only
really
relevant
when,
when
data's
above
certain
size-
and
then
you
there's
also
ways
of
compressing
reputation
into
a
very
easy
representation,
that's
kind
of
like
part
of
the
part
of
the
network.
A
So
these
are
like
simple
badges
right
and
you
create
a
way
of
earning
like
like
doing
a
lot
of
work
over
time
to
earn
a
very
compressed
representation
of
quality,
so
that
the
marketplace
people
browsing
the
marketplace
can
very
quickly
and
easily
navigate
to
to
pay
for
pay.
For
that
kind
of
added
at
the
service,
then
yeah.
This
is
kind
of
a
way
of
looking
at
some
of
the
features
that
I
was
describing
earlier,
like
people
may
care
about.
A
So
you
have
like
a
a
very
large
market,
but
once
you
start
kind
of
narrowing
down
and
and
and
you
segmented
by
these
kinds
of
these
kinds
of
features,
as
you
can
think
of
like
uis,
that
have
you
know,
depth,
charts
and
so
on
that
are
that
are
segmented
by
by
some
of
these
some
of
these
teachers.
I
see
the
view
this
come
here.
I
think
we
have
five
more
minutes
right.
A
A
A
So
yeah
we've
talked
about
a
lot
about
these
components.
Once
you
have
these
primitives
like
these
asks,
and
and
so
on,
you
can
start
making
other
kinds
of
clearing
structures
like
auctions
and
and
so
on
and
and
start
experimenting
with
a
lot
of
that
and
and
it
should
be
fairly
easy
for
people
to
you,
know,
take
these
like
feeds
of
of
bids
and
and
and
ass
and
so
on
and
and
auction
them
and
whatnot.
A
A
So
we
have
this
right
now
lopsided
structure,
where
there's
a
lot
of
people
in
some
regions
of
the
world
and
not
a
lot
of
people
in
other
regions
of
the
world
and
and
maybe
that
demand
might
be
different,
like
the
distribution
of
supply
and
the
distribution
demand
might
be
different,
and
so
the
areas
where
there
is
a
lot
of
demand,
but
not
enough
supply,
creating
space
for
capturing
those
bids
and
and
being
having
a
structure.
That's
clear
with
the
clients
to
say
hey.
A
This
is
a
bid
that
will
be
fulfilled
in
the
future.
It's
not
automatic,
but
it's
allowing
those
clients
to
express
hey
falcon
network.
We
would
totally
use
a
lot
of
storage
here
if
it
was
in
this
specific
region
and
fit
these
specific
set
of
conditions
and-
and
we
allowed
that
network
to
kind
of
gather
these
bids
and,
if
they're,
if
they're,
if
they're
legitimate
bids
that
will
kind
of
cause
miners
to
go
and
build
out
facilities
and
provide
certain
kinds
of
services
to
meet
the
meet
the
demand
right.