►
From YouTube: State of the IPFS DevTools ecosystem-Brendan O'Brien, Brooklyn Zelenka, Carson Farmer & Sara Feenan
Description
This panel talk was given at IPFS Camp 2022 in Lisbon, Portugal.
A
I
think
we
should
probably
start
off
with
a
round
of
introductions.
I
think
we
have
quite
a
like
a
diverse
crew
here,
different
backgrounds,
different
areas
of
expertise
and
I
think
there's
a
Common
Thread
that
you'll
share,
and
that
is
you're
all
involved
in
developer,
tooling
and
sort
of
building
these
tools
that
have
this
multiplying
effect.
So
I
don't
know
Brooklyn.
Do
you.
B
A
To
sure.
C
Test
test,
yeah,
okay,
I,
just
have
to
hold
it
much
closer
to
my
face
than
I
thought
so
I'm
Brooke
I'm,
if
you
were
around
earlier
I'm,
boris's,
co-founder
and
I'm
working
on
Vision,
and
we
provide
sort
of
a
whole
Suite
of
developer
tooling
from
a
publishing
platform
initially
and
then
realize
that
we
need
to
actually
move
down
a
few
layers
and
start
fixing
and
changing
and
adapting
data
transfer
protocols.
C
D
Good
morning
everybody
hi
I'm
B5
I
work
at
a
company
called
number
zero
and
we
make
a
tool
called
iro
and
we're
on
a
mission
to
make
the
most
efficient
implementation
of
ipfs
on
any
planet.
D
This
one's
the
most
important
one
for
obvious
reasons,
but
yeah
I
came
to
this
because
I
originally
started
a
company
built
on
top
of
ipfs
and
I'm
now
trying
to
build
the
version
of
ipfs
with
our
team
that
I
wish
that
I
had
when
we
started
and
make
something
that
is
really
focused
on
application
developers
and
solving
some
of
their
pain
points
when
it
comes
to
working
with
highly
distributed.
Technologies.
B
Cool
hi,
I'm,
Carson
farmer,
I'm
I'm,
with
the
with
textile
and
I,
already
knew
this,
but
listening
to
my
fellow
panelists
describing
their
sort
of
like
how
they
got
and
why
they're
on
this
panel
today,
literally
the
exact
same
reason,
textile
has
been
around
for
quite
a
while.
We
started
trying
to
build
a
consumer
app
on
top
of
ipfs
on
mobile,
and
then
we
said,
oh
geez,
that
was
hard,
so
we
went
down
a
stack
and
we
started
building
developer
tooly
and
then
we're
like
well.
That
was
hard.
B
Let's
go
down
a
bit
further
and
now
we've
started
developing
sort
of
more
lower
level
protocol
tooling,
and
so
the
journey
is
obviously
one
that
you
just
have
to
do
and
it
and
it's
a
bit
of
a
battle
and
then
you
sort
of
find
a
good
spot
somewhere
lower
down
the
stack
and
start
building
there,
hoping
that
someone
who
comes
along
after
you
has
a
little
bit
of
an
easier
time.
So
yeah
same
journey,
and
here
I
am.
E
Hi
I'm
Sarah
I
work
at
infuria
I'm
going
to
break
ranks.
We
haven't
built
on
top
of
ipfs.
We
just
provide
RPC
access
to
ipfs,
so
yeah
we
want
to
you
know.
Obviously
we
want
to
make
the
developer
experience
as
as
good
as
it
can
possibly
be
as
well
so
I'm
really
on
this
panel,
just
to
learn.
A
C
Encryption
and
access
control
do
you
make
anything
that
helps
for
that?
We
we
do
so.
Several
things.
Web
native
file
system
gives
you
encryption
at
rest
and
like
a
granular,
Access
Control
within
that
so
single
file,
or
you
know
entire
directory.
We
are
rewriting
it
in
Rust,
currently
in
webassembly,
so
it's
going
to
run
literally
everywhere
a
distributed
database
with
the
same
properties.
C
So
if
you
don't
want
file
system,
you
want
database
type
things
we'll
be
able
to
do
that
quite
soon
as
well,
and
you
can
which
I'm
really
happy
to
see
showing
up
in
a
fair
number
of
talks
is
totally
decentralized,
auth
and
trustless
interoperation.
That
is
hopefully
going
to
find
its
way
into
things
like
ipfs
nodes
directly,
and
then
we
will
be
able
to
plug
all
of
our
systems
together
without
having
to
do
pre-negotiation
and
figure
out.
You
know
all
these
apis
and
systems.
B
B
So
it's
databases.
It's
like
applications,
it's
all
of
those.
It's
it's
like
Ides
and
developer,
tooling.
You
know
how
many
Ides
do
have
like
Integrations
for
ipfs,
yet
probably
zero.
So
it's
like
all
of
all
of
those
things
that
I
expect
on
my
desktop
I
think
we're
slowly
going
to
start
to
see
those
things
bubble
up
as
useful
tools.
You
know
as
part
of
the
ipfs
stack,
so
yeah
lots
of
things
is
my
answer.
Thank.
A
You
so
if
we
back
up
a
little
bit,
a
lot
of
the
decentralization
space
is
not
purely
focused
on
ipfs
there's
other
Solutions
I
mean
it's
hard
to
not
mention
blockchains
and
I.
Think
one
of
the
first
questions
that
developers
are
confronted
by
is
do
I
need
a
blockchain
for
this
I
mean
blockchains
are
extremely
expensive.
They
don't
really
offer
a
lot
of
space
so
to
sort
of
simplify
the
question.
How
do
you
know
as
a
developer
if
you
need
a
blockchain?
Oh.
E
Well,
that's
a
big
question:
isn't
it
as
a
developer?
Do
you
need
blockchain,
so
I
guess
yeah
I
mean
as
an
ipfs
developer?
No,
of
course
it's
a
completely
separate
Network
and
it's
chain
agnostic.
If
you
do
need
a
chain
but
I,
guess
that
I
just
spring
to
mind
all
those
flow
charts
that
you
see
do
you
need
a
blockchain
and
invariably
they
resolve
to
know
but
I.
E
Think
if
you
want
to
have
some
kind
of
you
know
single
source
of
Truth
or
something
that
is,
does
have
this
a
kind
of
cryptographically
secured
state
that
needs
to
transition.
Then
yes,
if
you
just
need
a
snapshot
in
time
or
something
that's
you
know
entirely
mutable
or
doesn't
matter
with
with
all
of
those
features.
Then
you
probably
have
been
here
to
blockchain
safe,
safe
space
right.
B
B
The
rest
is
just
build
your
app
with
the
tools
that
make
a
lot
of
sense,
and
if
those
tools
don't
exist,
you
can
come
and
join
us
and
build
them.
C
Ipfs
is
a
really
great
Discovery
and
transport
protocol.
It
is
not
in
its
Self
Storage.
B
No,
you
made
it
available
over
ipfs
or,
like
you
announced
some
to
the
network,
that
you
have
something,
but
you
didn't
put
it
on
ipfs
and
unfortunately,
that's
just
like
nomenclature
that
we
as
an
ecosystem
use
and
it's
very
confusing
for
end
users
and
as
a
as
a
team
of
developers
that
build
ipfse
type
things
a
lot
of
time.
I'll
get
you
know
a
new
person
to
the
ecosystem
come
and
they're
like
yeah.
No,
no
I'm,
building,
ipfs
I've
got
this
app.
B
Where
you
know
it's
going
to
work
great
because
they're
using
localhost
3000
and
it
does
work
great
and
then
they're
like
no
no
it's
on
ipfs.
So
it's
just
and
then
you
have
to
explain
to
them
like
well.
B
You
would
have
told
me
this
before
and
you
know,
there's
that's
a
part
of
our
job
as
tool
developers
is
to
try
and
like
smooth
over
some
of
that
stuff.
Make
some
of
that
stuff
more
obvious
and
just
like
a
sort
of
clear
up
those
common
misconceptions.
But
Brooke
is
100
right.
That's
the
biggest
thing
is
not
a
storage.
It's
a
it's
a
it's
a
discovery
and
transport
protocol,
and
if
we
get
that
across
really
clearly,
then
it
makes
it
so
much
easier
to
develop.
B
D
Hello,
I
would
plus
one
all
of
that
and
add
to
it
that
I
think
I
like
to
think
of
this,
in
terms
of
like
the
evolution
of
our
space,
a
little
more
broadly
temporally,
you're
hearing
it
from
a
panel,
a
group
of
panelists
who
are
saying
hey,
we
all
moved
down
to
try
and
help
people
coming
up
like.
D
How
do
we
align
on
clarifying
to
people
hey
this
like
the
way
this
should
end
is
really
should
look
similar
to
like
the
emergence
of
react
right
like
it's
like
cool.
You
follow
this
tutorial.
This
is
how
this
works
and
the
number
of
people
who
have
gone
into
put
time
into
removing
all
the
foot
guns
from
that
tutorial
and
making
it
such
that
it's
like.
Okay,
cool
and
last
step
is
choosing
your
pending
service
because
you
need.
D
One
of
those
type
thing
is
ideally
where
we
end
that,
if
we
thinking
sort
of
in
the
macro
time
scale
that's
starting
to
emerge
now,
because
we
have
we,
as
a
network
of
teams,
are
starting
to
learn
how
to
work
together
to
build
that
which
is
exciting
and
I.
Think
really
one
of
the
Watershed
things
that
comes
out
of
Converses
like
this.
But
my
expectation
is
that
next
year,
we'll
ideally
start
to
see
some
of
those
tutorials
emerge
and
those
will
be
the
product
of
cross-organization
collaboration.
Saying
I,
don't
know,
I've
talked
to
the
people.
D
We
need
to
really
explain
ipfs
differently
and
we
really
need
to
have
a
conversation
about
how
people
deal
with
persistence,
and
you
know
someone's
just
figured
out
how
to
make
the
whole
Nat
traversal
problem
kind
of
go
away
in
a
way
that
you
know
we're
finally
getting
hole
punching
happening
at
a
rate.
That's
effective
enough
that
this
is
just
less
of
an
issue,
so
we
can
kind
of
skirt
around
that
one.
B
A
C
Yeah
I
mean
so
HTTP
is
a
single.
You
know
point-to-point
transfer
protocol
right,
maybe
uncontroversially
focused
around
hypertext,
it's
in
the
name,
the
there's
not
there's
nothing
saying
that
we
couldn't
add
HTTP
as
a
little
P2P
transport.
It's
a
thing.
We
could
do
right.
It's
a
thing
that
my
team
is
literally
working
on
because
in
if
you
want
to
support
browsers,
you
need
to
use
some
mature
transports.
Sometimes
right,
browser's,
not
dialable.
It
does
really
great
job
of
client
server
today.
C
So
what
if
we
could
just
go
and
say,
like
hey
other
node,
just
give
me
data
and
not
over
a
HTTP
Gateway.
Yes,
we'll
use
HTTP
but
like
let's
still
maintain
the
ipld
graph
as
we're
pulling
in
car
files.
Things
like
this
right,
so
I
think
that
in
the
next
year-
and
actually
we
were
just
talking
directly
for
before
the
panel
I
think
this
keeps
coming
up
as
a
theme
of
A.
C
D
A
But
also
the
other
part
of
it
was
you
one
of
the
most
common
ways
to
use
ipfs
by
newcomers
is
ipfs
gateways
and,
and
so
we're
you
know,
we've
improved
the
specification.
There's
been
a
lot
of
work
there
we're
discussing
writable
gateways,
so
it's
like
HTTP
just
seems
to
come
up
everywhere.
You
look.
C
So,
on
on
the
topic
of
gateways,
in
particular,
I
think
that
if
we
upload
over
hcp
to
a
server
and
we
download
from
a
server
and
it's
the
same
server,
we've
created
an
Apache
server
with
extra
steps.
C
D
I
could
disagree.
Do
you
want
me
to
disagree?
I'll
disagree.
I,
think
that
I,
the
only
thing
I
don't
disagree
with
the
the
point.
I
disagree
with
where
to
put
the
like
what
counts
as
ipfs
I,
think
it
as
long
as
it
verifies
the
hash
locally.
D
The
thing
The
crucial
thing:
you
need
to
do
for
it
to
be
decentralizable
or
distributable
to
pick
your
buzzword
is
you
need
to
fetch
content
and
check
that
the
thing
you
got
is
the
thing
you
asked
for
and,
and
that's
that's
the
to
me
is
the
thing
that
like
if
I
could
only
say
one
thing
to
application
developers.
Your
system
needs
to
do
that.
I'm
less
concerned
about
the
transfer
protocol
I'm
more
concerned
about
the
characteristic
of
hash
integrity.
B
Okay,
so
on
on
that
front,
then
so
how
many
people
here
use
Brave,
good
chunk
of
people,
so
brave
has
an
embedded
ipfs
node,
which
is
really
just
a
local
Gateway,
and
so
by
that
logic,
does
that
does
that?
Count
then,
is
that
is
that
good,
we're
good
there?
Yeah
cool,
okay,
I,
think
that's
a
good
answer.
There,
too,
is
like,
as
we
see,
ipfs
nodes
becoming
more
ubiquitous
on
the
edges.
B
A
B
So
with
brave,
if
I
have
my
little
like
toggle
for
ipfs
turned
on
and
I
have
like
an
ipfs
node
running,
you
can
use
the
embedded
one
or
I
have
ipfs
desktop,
so
I
think
I
default
to
that
one
I,
just
it
just
works
yeah,
but
it's
DNS
link,
that's
doing
it
yeah,
as
is
the
technology
that's
being
used
yeah,
but
also
you
know
just
do
it.
It
works.
A
So
what
do
you
think
are
some
of
the
inherent
trade-offs
that
are
here
to
stay
with
us
as
developers
application
developers,
building
on
top
of
ivfs
I
mean
there's
certain
costs
as
per
your
definition,
it
was
verifying
locally,
that's
quite
expensive
to
do,
especially
if
we
consider
you
know
low
power
devices
I'm
sure
you
know
we
can
come
up
with
better.
You
know
hashing,
algorithms,
perhaps
and
and
more
optimizations,
but
like
what
are
some
of
these
inherent
trade-offs
that
you
know
an
application
developer
is
looking
to
build
these
decentralized
apps.
C
So
in
the
long
term,
I
think
that
we
can
get
to
a
place
where
it
is
not
just
at
feature
parity,
but
significantly
better
where
you've
downloaded
data.
Recently,
you
don't
have
to
go
back
out
to
the
network
to
fetch
it
again
because
it
has
the
hash
or
you
know
the
the
dream
that
we
all
have
I
think
is
you're
sitting
next
to
somebody
at
a
cafe.
C
They
have
a
you
know
news
article
up,
you
load
the
same
news
article
and
it
just
goes
direct
from
device
to
device
right
like
that's
like
you
know,
Future
Vision,
right
today,
everything
is
so
early,
so
the
trade-off
is.
B
Yeah
I
mean
there's
a
good
point
and
you
know
I
whenever
I
try
to
explain
how
the
internet
works
to
my
mom
and
I've
got
like
my
phone
here
and
she's
got
her
phone
and
she's
like
well
I
want
that
picture
on
there
and
I'm
like
well.
I
can't
just
send
it
directly
to
you
she's
like
We're
physically
next
to
each
other.
Surely
that's
how
the
internet
should
work.
It's
not
like.
B
That's
that's
the
sort
of
intuitive
thing,
and
it
is
true
that
it's
going
to
be
a
while
before
we
get
there
because
I
have
to
explain
to
her.
No,
no,
it
goes
to
somewhere
in
maybe
like
Chicago
or
Cupertino
somewhere,
and
then
someone
does
some
work
over
there
and
then
they
send
it
to
you
and
it's
like.
Oh
it's,
this
insane
network
of
of
tubes,
yeah
internet
tubes,
which
can
get
clogged
right.
So
you
know
it's
avoiding!
That
clogging
is
what
we're
after
and
that's
how
the
internet
works.
C
A
So
this
leads
me
to
the
next
question.
Many
of
you
shared
the
belief
that
encryption
is
one
of
the
missing
pieces
for
ipfs
and
you
know
for
any
kind
of
doing
any
kind
of
thing
that
involves
pii
personally
identifiable
information.
So
in
five
years
from
now,
how
do
you
think
users
will
manage
their
keys?
I'm
sure
that
you
have
some
opinions
about
this
but
yeah?
Let's,
let's
take
that
question.
C
So
pki
is
getting
better
really
fast,
so
even
in
say
the
you
know,
web
3
space
people
have
their
metamask
wallet
and
that's
about
50
million
users
today
and
they're
already
coming
with
their
keys,
they're
already
doing.
Key
Management
amazing,
with
things
like
pass
Keys,
which
is
Apple,
Google
Microsoft
and
a
few
others
it's
built
into
browsers.
Now
you
have
Key
Management
without
passwords
in
all
of
your
systems,
and
that
is
for
doing
signing
both
of
those
systems
right
I'm,
going
to
assign
some
some
data.
C
There's
work
happening
at
the
chain:
agnostic
standards,
Alliance
Casa
for
standardizing
encryption
decryption
with
polits,
and
we
have
the
web
crypto
API
in
browsers
for
doing
Key,
Management
I
think
the
the
Real
Techniques
are
around.
How
do
we
manage
as
few
Keys
as
possible
while
giving
people
access
to
as
much
data
as
they
can
have?
There's
a
great
quote
from
Google,
which
is
the
great
thing
about
cryptography,
is
it
turns
all
problems
into
Key
Management
problems,
which
is
terrifying
because
key
Management's
super
hard.
C
Then
it
just
makes
it
like
much
easier
and
we
have
this
abstraction
layer
now,
where
we
can
say
like
yeah.
This
is
stored
in
I,
owe
a
in
a
hardware
wallet
or
directly
in
the
browser
or
on
a
piece
of
paper
where
you
type
it
in
every
time
or
we
drive
it
from.
You
know
a
passphrase.
It
doesn't
really
matter
anymore
and
we
can
make
this
much
much
much
easier.
C
If
it's
self-sovereign
and
self-owned,
you
need
a
key
recovery,
Story
and
there's
a
bunch
of
ways
of
doing
that.
But
there
isn't
a.
You
know
there's
no
one
right
way
of
doing
that
today.
A
D
A
B
True,
that's
true
yeah
because
you
can
fall
in
you've
got
your
phone.
You
probably
have
your
Hardware
wallet,
because
we
all
keep
that
with
us
all
the
time
and
yeah
you're
you're
host,
basically
don't
go
boating,
no,
but
it.
But
there
there's
I
mean
literally
everything
Brook
said
is
100
right
and
I,
probably
don't
even
need
to
add
anything.
B
All
the
helpful
hardware
and
browser
like
sort
of
abstraction
is
just
going
to
get
better
and
better
and
better
losing
your
keys
is
a
bad
thing.
I
think
there's
a
ton
of
really
great
HC
or
human
computer
interaction.
Research.
That's
happening
right
now
around
key
recovery,
like
you
know,
if
you
forget
something,
you
ask
your
friends
for
it.
B
There's
a
pretty
great
model
where
you
recover
portions,
of
your
keys
from
your
network
of
friends
that
you
have
so
there's
lots
of
like
ways
to
leverage
being
a
human
to
actually
interact
with
your
keys
and
recover
them,
and
things
like
that.
You
also
have
more
than
one
device.
B
E
I
mean
I,
don't
really
have
anything
to
add.
Obviously,
apart
from
just
saying
yeah
the
same,
it's
the
human
problems
that
are
difficult
that
that
element
of
the
you
know
we
can
solve
the
technical
and
one
of
the
things
just
to
sort
of
shell.
This
conference
for
a
second
one
of
the
things
I
think
is
really
good,
is
seeing
the
level
of
use
cases
that
are
presented
here
as
well,
not
just
the
protocol
streams
and
that's
suddenly
finding
ourselves
on
a
Tech
Tool.
Looking
panel.
D
Just
I
know
this
is
dragging
this
question
on
a
little
longer,
but
I
think
just
to
like
take
the
I
think
we've
covered
the
technical
aspects.
Why
are
we
doing
the
pki
stuff
right,
we're
doing
pki
and
voicing
these
conversations
about
Keys
onto
users,
because
this
really
hits
at
the
heart
of
the
sort
of
political
aspects
of?
Why
we're
doing
this
right?
We're
introducing
this
complexity?
D
This
is
complexity
that
users
are
going
to
have
to
deal
with
on
their
own,
but
this
is
also
an
evolution
of
our
online
presence
as
people
and
we're
sort
of
saying
hey.
In
order
for
you
to
do
this,
you
do
have
to
actually
reconcile
with
a
new
set
of
concerns
and
we're
going
to
have
to
service
a
little
more
complexity
to
you,
because
our
inner,
more
and
more
of
our
lives
are
on
the
internet
and
more
and
more
of
our
lives
are
being
tangled
up
in
these
interactions
and
I.
D
Think
a
lot
of
the
like
social
reasons.
Why
we're
forcing
this
complexity
on
you
really
does
poke
into
HCI,
because
we're
saying
hey,
you
need
to
contest
with
this,
and-
and
this
is
part
of
the
whole,
like
reason,
we're
doing
all
this
to
begin
with,
and
that's
why
you
know
we
say
like:
oh,
this
is
just
a
ux
challenge
like
no.
This
isn't
just
a
ux
challenge.
This
is
like
a
really
crucial
thing
that
we
need
to
not
only
get
right.
C
And
just
to
extend
that
a
little
bit
if
you've
shown
a
user
a
did
or
a
public
key
you've
already
lost
it's
already
over
the
people
in
this
room
care
enough
to
get
educated
about
these
things.
Nobody
outside
of
this
building
does
zero
percent
people
say
that
they
care
about
privacy
and
security,
and
then
they
go
and
they
use
Facebook
right
like
it's.
C
We
have
to
make
this
stuff
so
easy
to
use
that
it
doesn't
matter
that
it's
using
the
stuff
underneath
and
it
happens
to
be
secure
right
and
by
having
that
security,
and
you
know
encryption
at
rest
and
all
these
things
we
get
all
these
other
really
great
conveniences
out
of
the
system,
but
we
still
have
to
solve
this
hard
problem
of
how
do
I
as
a
regular
human
being
without
a
technical
background.
How
do
I
well
I
mean
not
not
literally
me
but
like?
How
would
one
do
that.
E
Yeah,
it
kind
of
reminds
me
I
used
to
teach
journalists,
investigative
investigative
journalists,
how
to
use
things
like
tour
and
veracrypt,
and
things
like
that,
and
they
couldn't
get
their
head
around
things
like
don't
you
know,
resize
the
browser
use
different
sessions,
don't
log
on
with
Facebook,
whilst
you're
in
tour.
Like
all
of
these
things-
and
there
are,
there
are,
of
course
people
tracking
what's
going
on
with
investigative
journalists,
and
they
don't
kind
of
they
can't
understand
it.
It's
I
don't
mean
to
sound
rude
and
patronizing.
E
It
just
kind
of
shocked
me
I'm,
like
a
semi-technical
person,
which
is
what
I'm
really
weird
to
be
on
this
panel,
but
people
safe
space,
yeah
normal
people,
people
that
don't
work
in
technology.
They
get
terrified.
It's
like
me
trying
to
do
my
tax
return.
E
It
just
you
know,
and
my
mum's
been
in
crypto
for
years,
I've
been
in
the
space
for
like
five
or
six
years,
and
she
still
can't
do
anything
with
metamask
without
me
sitting
next
to
her
and
that's
a
pretty
easy
ux
like
it's
hard
but
yeah,
like
you
say,
once
you've
once
you've
introduced
these
people
to
something
like
a
you
know:
private
queue
or
a
public
key
yeah.
A
So
we've
spoken
a
lot
about
now,
I
think
with
key
management
about
sort
of
human,
the
HDI,
the
human
device
interaction
aspect
of
it.
I
want
to
take
this
now
in
a
slightly
more
technical
Direction,
especially
with
the
technology
that
I
think.
A
lot
of
us
are
very
hopeful
about.
I
know
that
some
of
you
are
have
either
already
interacted
and
used
this
technology
and
some
of
your
building
already
tooling,
on
top
of
it,
so
crdts
underrated
overrated.
Why
would
love
to
hear
your
thoughts
who
wants
to
take
this?
A
D
We
first
like
just
Prime
for
anybody
who
doesn't
know
what
that
acronym
is,
can
I
do
the
yes
totally
do
100
000
foot
conflict-free,
replicated
data
types,
the
idea
of
being
a
system
by
which
you
can
deal
with
multiple
people
writing
to
the
same
thing,
Fair
like
how
do
you
deal
with
it
like
so?
The
classic
example
is
like
the
Google
Doc
problem.
You've
got
four
people:
okay,
let's
not
use
Google
doc,
a
collaborative
document
and
you've
got
five
people.
They
all
want
to
type
into
that
same
thing.
D
D
Yeah,
so,
where
we're
at
with
them,
it's
the
only
horse
in
the
race
in
a
lot
of
ways
right
like
we,
we
don't
want
to
deal
with
it.
I,
don't
I,
don't
like
complexity
if
I
can
avoid
it,
but
this
feels
like
the
kind
of
thing
that
it's
the
best
candidate,
that,
in
my
view,
that
we
have
to
get
something
like
that
that
live
collaborative
process
is
one
really
great
use
case.
Just
any
time
where
you
really
have
kind
of
non-blockchain
synchronized
stuff
crdts
are
really
like.
D
It
seem
to
be
the
thing
that
people
keep
turning
to
the
challenge
with
these
consist
like
for
me.
The
biggest
challenge
here
is
the
the
problems
that
we're
trying
to
address
into
finding
a
really
nice
abstraction
that
application
developers
can
build.
On
top
of
the
problem
you
run
into
with
crdts,
is
it's
really
easy
to
make
something?
That's
consistent!
D
How
how
do
we
deal
with
that,
and
how
do
we
come
out
with
something
that
feels
consistent
and
like
a
good
user
experience,
I'm,
confident
that,
like
the
research,
is
moving
forward
really
well
on
the
like
underlying
text,
side
of
things,
I
think
that
the
last
piece
of
the
puzzle
is
going
to
be
really
hard
to
land.
Is
that
how
do
I
express
conflict
resolution
as
an
application
developer
without
exposing
them
to
all
of
the
guts
of
what
it
means
to
like?
Do
conflictally
replicated
data
types
across
multiple
parties.
B
Yeah
I,
so
I
actually
have
some
some
opinions
here,
and
this
might
make
the
panel
exciting,
because
I
definitely
think
that
crdts
are
really
interesting.
Research
topic
and
I
read
about
that's,
like
my,
you
know
fun
thing
to
read
about,
but
in
practice
they
always
end
up
falling,
pretty
short
of
what
I
really
need.
B
So
my
team,
we
developed
a
database
thing
called
thread
DB
and
it
used
crdts
as
a
way
to
sort
of
like
lazily
evaluate
State
like
on
the
Node
that
needed
to
look
at
it,
and
we
implemented
this
like
really
complex
framework
that
allowed
you
to
express.
You
know:
State
materialization,
as
the
as
a
crdt
and
the
one
that
got
used.
B
All
the
time
was
an
operational
transform
like
implementation,
which
is
basically
like
where
someone
decides
you
know,
sort
of
takes
an
update
and
sort
of
re-transforms
it
to
match
the
current
state
and
anyway
it's
the
simplest
possible
sort
of
like
scenario
and
then
the
other
simple
possible
scenario
that
we
saw
like
99
of
the
time
was
just
I
will
take.
The
latest
update.
Please,
like
that's
fine
I,
don't
need
anything
more
complicated
than
that,
because
it
turns
out,
like
semantics,
are
very
hard,
like
my
intention
to
like
move.
B
B
Is
that
they're
very
use
case
specific,
like
there's
a
there's
like
so
many
different
crdts
out
there
and
if
I
want
to
work
with
text,
I
need
one
and
if
I
want
to
work
with
numbers,
I
need
another
one
and
and
I
don't
want
to
think
about
that.
Like
I,
just
have
some
data
that
I
want
to.
Like
you
know,
sync,
that's
what
I
want
so,
unfortunately,
you
know
I
think
a
lot
of
developer
tooling
has
to
be
developed
in
order
to
make
that
whole
process
a
lot
easier.
C
Strong
agree,
basically,
so
the
things
I'll
sort
of
extend
on
on
that
in
in
addition
right
so
crgts,
are
solving
the
problem.
How
do
we
agree
on
a
convergent
State
once
we've
made
our
own
updates
right,
without
ideally
without
user
intervention
right?
So
it's
like
git,
but
without
the
merge
conflict,
part
right,
they're,
relatively
new
in
the
grand
scheme
of
things,
people
are
working
on
them
longer
than
this,
but
the
term
was
really
coined
in
paper
in
2011's
we're
at
11
years
of
crdts.
C
C
If
you're
in
a
distributed
system
with
an
unknown
number
of
participants,
you
can
never
garbage
collect
any
of
your
States
right
and
there
are
ways
of
compressing
parts
of
that
right,
but
somebody
could
have
a
right
that
depends
on
something
from
three
years
ago
and
now,
if
you're
doing
this
on
a
low
powered,
Android
device,
you've
got
to
pull
down
all
of
this
data
to
do
the
reconciliation
right.
There's
active
research.
C
This
literally
this
year
in
the
past
10
months
on
how
do
we
do
Byzantine
fault
tolerance,
your
DTS,
as
far
as
I'm
aware
fission
is
the
only
team
that
has
invested
in
encrypted
at
rest,
CRTs,
whether
it
has
semantics
at
the
encrypted
and
the
decrypted
layers.
These
are
really
foundational
fundamental
things
right
that
are
like
very,
very
new,
and
so
just
like
you
know,
should
you
use
a
blockchain
or
not?
Should
you
use
crdts
like
it's?
Ideally
not
right.
C
If
you
have
a
single
source
of
truth,
if
you
have
blockchain
data,
you
don't
have
to
worry
about
these
things,
because
you
have
consensus
and
you
don't
get.
Blockchains
are
one
style
of
consensus.
You
can
do
lots
of
them
in
their
trade-offs
because
consensus
to
slow,
because
the
speed
of
light
is
finite.
C
Things
like
that
I'm
just
going
to
plug
really
quickly
not
available
for
General
release
yet,
but
we're
working
on
a
essentially
a
crdt
framework
based
on
a
database
that
comes
with
all
of
these
things.
That
has
incremental
view
maintenance
that
does
all
of
this
stuff
and
encryption
out
of
the
box
and
has
a
standard
library
for
these
things.
But
it's
coming
out
of
a
research
lab
right,
so
there's
going
to
be
a
lot
of
learning
along
the
way
once
that
is
released
and
comes
out,
and
even
the
guarantee
is
that
cooties
give.
C
A
Thank
you
for
all
the
thoughtful
answers.
One
final
question:
if
you
had
a
magic
wand
that
you
could
sort
of
wave
and
solve
one
key
problem
in
this
space,
what
a
problem
would
you
solve?
Identity.
E
E
C
D
Wanna
marinate
for
a
minute
cross
organizational
collaboration.
B
Yeah
we're
just
brainly
I
like
that
one
I
like
that
one,
a
lot.
The
truth
is
all
of
the
like
things
that
I
find
really
interesting.
I
would
hate
to
just
like
wave
a
magic
wand
and
have
that
done
because
then
I
don't
get
to
work
on
it
anymore,
So.
My
answer
would
have
to
be
something
like
really
boring
that
I
don't
want
to
do
so
that
I
can
just
sort
of
like
not
deal
with
that
right.
B
So,
like
I,
like
working
on
databases,
yeah
it'll
be
something
like
that.
It
would
be
like
where
my
network
stack
just
works,
and
so
I
know
that,
like
if
I
publish
some
data,
I
will
be
able
to
get
it
on
the
other
side
or
some
peer
like
I,
think
it's
it's
like
basically
live
P2P,
just
like
ubiquitously
Works
pole
punching
everything
done
that
would
be
yeah,
yeah,
yeah
yeah,
exactly
yeah,
just
some
Quantum
just.
C
My
original
answer
was
actually
also
going
to
be
crossword
collaboration,
but
I
wanted
to
Echo
the
questions
about
accounts
and
identity
and
just
sort
of
broadly
key
management
and
Recovery,
which
are
really
super
hard
problems
that
there
are
not
good
solved
things
for
today
that
we
probably
won't
have
good
solutions
for
for
a
while
and
that
once
it's
solved
it
unlocks.
So
many
things
and
the
space
can
really
just
like
explode,
which
should
be
great.