►
From YouTube: 🖧 IPLD Every-two-weeks Sync 🙌🏽 2023-03-28
Description
An every two weeks meeting to sync up on all IPLD (https://ipld.io) related topics. It's open for everyone and recorded. https://github.com/ipld/team-mgmt
A
Welcome
everyone
to
this
week's
IP,
iPad,
sync
and
community
meeting
it's
March
27th
and
as
every
two
weeks
we
go
over
the
stuff
that
people
have
worked
on
or
discuss
any
open,
gen
items
or
if
you
have
questions
and
today
it's
special
because
we
even
have
a
talk
today
and
I.
Guess
we
just
start
with
the
talk
so
yeah
I
hand
over
to
Jay
Chris,
because
he
will
tell
us
what
he
will
be
talking
about
and
so
on.
B
Cool
thanks,
Walker,
thanks
for
having
me
thanks
for
being
here,
Rod
thanks
for
being
here.
Everyone
on
the
live
stream
I
am
excited
to
share
this
new
project.
I'm
working
on
that.
B
In
my
head,
distance,
I,
don't
know
like
current
form
and
since,
like
2000
10
or
something
before
that,
so
I've
been
thinking
about
how
to
build
this
thing
for
a
while
and
some
of
the
pieces
kind
of
fell
into
place
recently
that
made
it
something
that
one
could
hope
to
build.
B
So
you
know
there's
a
lot
of
stuff,
that's
hard
that
you
can't
just
do
yourself
and
there's
a
couple
of
ingredients
that
I
mixed
together
to
make
this
possible.
But
the
thing
that
I
think
this
brings
to
the
conversation
about
ipld
data
structures
is
that
they
can
be
packaged
in
a
way
where
you're
using
it.
B
You
don't
even
know
so
here
you
are
as
like:
a
quick
start
user
and
you're
just
npm
installing
a
thing
and
doing
stuff
with
it,
and
you
can
even
like
put
it
on
your
window
if
you're
wild
like
that,
and
it
basically
worked
right
away
when
I
just
did
this
remix
between
a
couple
of
these
components
that
I'll
talk
about
in
a
little
bit
and
but
I
worked
right
away,
I
mean
it
took
me
like
not
too
long
a
month
or
so
to
get
essentially
like
a
couch
like
implemented,
and
then
some
of
the
stuff.
B
That's
neat
about
doing
it
with
Merkel
instead
of
couch,
is
a
whole
bunch
of
efficiencies
in
the
replication
and
all
the
provability
that
you
get
from.
You
know
kind
of
the
stuff
you
have
to
do
if
you're
gonna
get.
B
B
Right,
it's
like
you're
building
an
app
you
need
cool
stuff
about
that.
Like
I,
didn't
choose,
it
chose
me
or
maybe
like
in
the
macro
picture
right
the
fact
that
it's
a
local
storage
based
Pro,
you
know
Merkel,
probably
tree
implementation,
means
that,
like
let's
say
you
were
doing
a
user
profile
page
that
you
already
had
written
like
you've
got
users,
and
now
you
want
to
add
features
to
it.
B
You
could
build
an
editor
UI
for
the
user,
where
they
do
arbitrary
stuff,
save
arbitrary
State
and
then
that's
just
like
a
little
Merkel
cookie
size
thing
that
you
have
to
keep
on
your
back
end
to
rehydrate
that
state
for
your
build
when
you're
building
a
profile
page
or
something.
So
it
allows
you
to
add
arbitrary
features
to
your
app
and
get
to
them
from
on
the
front
end
or
the
back
end
and
like
extend
whatever
kind
of
data
topology.
B
Your
app
already
has
that's
a
super
easy
thing
you
can
do
because
you
can
drop
new
UI
into
pages
and
power
it
like
via
having
this
data,
be
globally
accessible
and
I.
B
Wasn't
expecting
it
to
be
that
easy
and
the
other
thing
that's
super
y'all
know
a
lot
about
right
is
like,
for
instance,
smart
contracts
and
file
coin,
both
use
a
lot
of
Merkel,
and
so,
if
somebody's
like
doing
their
own
storage
aggregator,
they
might
use
fireproof
for
the
way
that
they
just
like,
keep
metadata
around
inside
of
the
car
files
that
they
ship
and
like
same
thing
for
anyone
doing
fvm
contracts.
B
That
makes
a
lot
of
sense
because
you
can
side
chain
it
safely
and
then
I
think
that
as
a
community,
we
could
talk
more
about
why
proofs
matter
for
AI
so
like
it.
Just
is
the
case
that
if
you
do
the
stuff
the
way
we
do
it,
then
you
can
get
like
a
whole
bunch
of
I,
guess
reproducibility.
B
So
people
can
rerun
models
with
the
same
inputs
and
random
seed
and
then
kind
of
like
have
tools
that
allow
them
to
through
that
AI
stuff.
So.
B
Yeah
I
think
it's
going
to
be
real
big,
but
I've
done
this
before
and
I
want
to
do
like
I
think
it's
gonna
be
bigger
than
the
last
time,
which
was
couch
based,
so
I
wanted
to
be
like
jQuery
like
every
app
out
there.
Just
does
this,
because
why?
Wouldn't
you
it's
super
easy
and
then
I'm
not
really
trying
to
make
money
on
your
infra
right,
like
let
three
storage
make
that
money
and
then
what
I
want
is
to
just
like
unlock
the
stuff
that
I
haven't
written
yet.
C
B
So
I
want
to
build
the
community
of
users
at
like
a
basically
like
they're
verified,
essentially
and
that's
my
short
term
plan,
but
like
the
idea
that
that
is
long-term
aligned
with
like
the
future
of
data
dials
and
stuff
that
we
might
see
in
that
World,
so
yeah
how
it
works.
B
The
it
uses
here.
I'll
do
this:
it
uses
Alan,
Shaw's,
pail,
clock
and
Michael's,
probably
true
implementation
and
I
kind
of
tore
into
both
of
those
to,
and
this
one
to
add,
like
a
bunch
of
edge
case
handling
and
this
one
to
tear
out
like
the
minimal
core
of
it
that
I
actually
use.
B
But
what
happens
is
events
come
in
like
a
writing,
a
document
change,
key
value
change
comes
in
and
then
the
causation
is
tracked
in
Alan's
clock
which
can,
like
you
know,
resolve
forks
and
it's
Auto
emerges
right.
If
changes
on
different
Keys
happen,
it
doesn't
care
about
those,
but
basically
how
it
does
all
that
is
outside
of
its
purview.
It's
pretty
it's
pretty
cool.
How
the
just
the
causal
tracking
gives
you
the
ability,
then
to
like
keep
one
of
these
probably
trees
around.
B
That
corresponds
to
here
and
then
like
Advance
it
to
here
whenever
you
feel
like
it
and
then
query
that
one
in
memory
and
like
most
your
blocks
are
already
cached
and
then
like
maybe
put
this
one
in
your
local
storage
cookie
eventually,
so
you
can
have
these
probably
trees.
You
know
they're
just
kind
of
like
cheap
Forks
of
all
these
different
update
operations.
So
as
you're
resolving
conflicts,
you
can
get
a
view
into
what
that
full
database
state
looks
like
real
cheaply
because,
most
of
it,
you
already
have.
B
And
then
you
know,
you
all
know
what
it
probably
looks
like,
but
this
is
my
bad
art
of
trying
to
say
that.
Well,
it's
just
like
essentially
a
b
tree
with
leaves
and
inner
nodes.
But
what
makes
it
different
is
when
it
constructs
itself.
It
doesn't
matter
the
order
of
insertion,
you're
going
to
get
the
same
Merkel
root
and
that
hash
is
like
enforced.
B
Essentially,
the
reason
it's
deterministic
book
called
prolly
is
because
the
way
these
nodes
decide
on
the
cutoff
boundary
for
the
next
block
is
based
on
their
content,
and
so
you
could
set
like
the
Threshold
at
which
how
much
content
is
going
to
be
next
block
making.
B
So
everyone's
gonna
end
up
with
the
same
thing,
which
makes
all
the
other
stuff
in
couch
way
easy
like
the
replication
is
just
telling
people
about
these
right
and
if
the
blocks
are
accessible
on
like
a
Gateway
or
something
I'll
show
a
demo
in
a
minute,
but
one
of
the
first
things
I
did
when
I
was
playing
with
it.
B
Myself
was,
after
I,
got
the
replication
to
Cloud
working,
just
delete
local
storage
and
refresh
the
page
and
my
app
loaded,
just
like
one
block
at
a
time
out
of
web3
storage,
and
it
worked
so
there's
optimizations
to
be
done
there,
but
that's
logically,
what's
happening
both
of
these
share
a
block
store.
B
You
know,
I,
guess
like
farther
more
before
we
care
about
the
Block
store.
The
indexes
are
just
a
JavaScript
function
that
can
pull
this
chain
since
the
last
time
you
had
an
index
head
and
you
know,
run
and
be
like.
Oh
this
document
you
know,
gets
indexed
under
this
key
this
document
under
that
key,
so
you
can
do
stuff
like
find
people
by
ZIP
code
or
whatever,
and
so
we've
got
to
open
issue
on
this
implementation
to
make
the
inner
node
aggregations
programmable.
B
But
once
you
do
that,
then
you
can
do
that
really
sweet
like
Dynamic
windowing
queries
where
the
is
logarithmic
cost,
because
most
of
the
work
is
already
memorized,
and
so
you
can
do
stuff
like
custom
aggregations.
What's
the
standard
deviation
of
people's
height
off
these
profile
records
or
whatever?
Okay?
So
now
we're
down
in
The
Block
store.
All
these
things
share
a
common
block
store.
You
can
segregate
out
the
views
to
a
different
Block
store,
if
you
don't
want
to
replicate
them
and
the
blocks
get
read
individually.
B
I
wrote
the
whole
thing
on
top
of
just
key
like
an
in-memory,
key
value
and
then
I
and
then
I
wrapped
my
transactions
in
car
files.
So
now,
every
time
you
do
an
update,
it's
probably
going
to
touch
like
XYZ
number
of
blocks,
and
so
those
blocks
are
going
to
go
into
a
car
file
which
is
Tiny
for
a
car
file.
It's
like
every
time
you
touch
something
you're,
sending
a
car
file,
web3,
storage
and
so
the
next
that's
my
transaction
boundary.
B
But
then
there's
like
an
acceleration
that
I
can
do
where
I
take
all
the
car
files
per
whole
database
and
make
it
into
one
car
file
and,
like
usually
that's
going
to
be
small
in
the
context
of
an
HTML
payload.
So
you
can
just
hydrate
your
HTML
from
the
car
file
that
you
attach
to
it
for
like
an
accelerated
loader.
B
So
that's
like
where
I
am
now
encryption
is
on
the
car
file
level
and
then
it's
I
guess
optional,
but
it's
gonna
be
on
by
default
and
it
uses
a
like,
essentially
a
one-time
pad
a
shared
key
for
symmetric
encryption
for
the
car
file.
B
And
then
it's
up
to
you
how
you
manage
that
I
think
people
will
want
to
have
like
a
Medusa
net
style,
re-encrypt
service
on
a
smart
contract
that
makes
copies
of
the
database
available
for
people,
but
because
of
the
way
that
I
split
the
transactions
like
this
and
the
way
that
everything's
plugable.
What
you
can
do
is
you
can
send
you,
can
encrypt
car
files
and
send
it
to
web3
storage
and
that's
under
your
own
personal
key
that
you're
not
really
going
to
share.
B
And
so,
as
long
as
that
you
know
channel
is
secure,
then
you
don't
have
to
worry
about
key
sharing
across
the
collaborators.
So
there's
a
bunch.
It
gives
you
options
and
I.
Don't
know
how
much
that
I'm
going
to
make
up
to
the
application,
but
yeah
I
mean
I,
guess
like
if
I
and
the
event
feed
can
do
your
chat,
room
or
whatever.
B
So
if
I
want
to
build
a
business
like
or
snowflake
yeah
I
guess
like
I'm
hoping
I
can
get
everybody
who's
developing
on
it
to
sign
up
for
this
almost
free
or
like
make
a
docs
pull
request
or
something,
and
then
the
and
then
I
have
a
demo
to
give
so
yeah.
Let
me
just
switch
over
to
that.
B
I
just
had
a
kid,
come
in
and
say:
hi,
okay,
so
I
built
this
as
kind
of
my
test
ground
and
I
can
go
into
I've.
Had
you
know
it's
kind
of
based
on
an
old
to
do,
MVC
that
I
had
around
that
had
multiple
lists
and
so
I'll
go
into
my
building.
Apps
list
and
I
can
check
off
things
like
you
know.
Basically,
I
could
check
everything
off
and
then
you'll
see
over
here.
B
B
But
now,
let's
like
it,
I
just
need
to
fix
the
UI
here.
There's
a
tree
happening
right.
So
if
I
go
like
and
do
some
other
pattern
and
now
here's
the
beginning
of
time
and
here's
the
the
pattern
I
did
and
here's
everything
right
and
here's
like
coming
up
from
the
tree
on
the
pattern.
B
So
I
can
you
know
kind
of
and
then
yeah.
If
you
just
do
like,
oh
I,
guess
yeah
we're
up
we're
already
logged
in.
So
if
I
look
under
the
network,
tab,
there's
gonna
be
maybe
some
car
files
going
Upstream,
but
I'll
do
that
in
another
demo,
so
I
think
that's
everything
I
have
to
share.
B
A
Things
yeah
so
when
I
first
heard
about
it,
I
basically
tweeted
that.
Finally,
someone
is
building
that
I
also
had
on
my
mind
for
years,
because
I'm
also
come
from
like
coach
TB
world
and
then
when
I
joined
protocol
apps,
it
was
like
Hey
this,
like
culturally
and
steroids.
So
I
need
to
build
something
like
this.
So
it's
great
to
hear
and
now
that
I
basically
got
an
in-depth
dive.
A
It's
like
it's
even
more
so
what
I
want
to
do
to
build
because,
like
like,
did
I
understand
correctly
that
you're,
basically
with
the
prolly
trees,
you
want
to
do
the
same
thing
that
couchdb
is
doing
with
the
having
the
aggregations
inside
the
tree.
Basically,
so,
basically,
if
you
aggregate
everything
you
know
already
on
the
route,
what
the
result
is.
B
Yeah
that
plus
the
replication
is
so
much
easier
because
everyone's
using
a
global
address
space
with
the
prolly
tree
style,
deterministic,
IDs.
A
And
one
thing
which
I
remember
being
a
problem
in
the
coach
DB
days
was
which
I
the
truth
is
the
problem
here
as
well
is
the
we
thought
about
doing
Ripley
like
if
you
do
duplications
of
a
bigger
data
set
that
you
can
basically
prioritize
things.
So
you
can
say
I,
don't
know
like
first
send
the
text
messages
and
then
send
the
image
or
something
like
this.
B
You
know
the
intelligent
loading
of
stuff
is
going
to
be,
there's
all
kinds
of
Parts.
You
can
accelerate
that
I
haven't
got
into
yet
so
I
talked
about
the
car
file
acceleration.
Basically,
loading
a
whole
data
set
from
only
the
identifier
of
its
root
node
with
no
like
kind
of
graph.
Sync
acceleration
is
good.
It's
just
slow
right.
It's
like
that.
B
Many
round
trips,
however
deep
the
graph
is,
and
so
it
works,
but
it's
not
I
mean
even
if
it
does
work,
there's
so
many
hops
that
you
start
like
running
into
timeouts
or,
like
you're
inevitable.
B
You
know
kind
of
Gateway
errors
or
something
to
get
a
whole
app
up
so
doing
the
car
file
acceleration,
which
I'm
pretty
sure
like
web
I,
just
haven't
written
it
yet,
but
I
can
load
a
whole
I
can
load
like
like
instead
of
loading
700
blocks,
I
can
load
65
car
files
and
that'll
help,
but
that's
why,
as
soon
as
I
can
do
that,
there's
gonna
be
motivation
to
just
load
one
car
file,
which
is
super
easy
to
repack,
like
it
kind
of
like
have
the
code
already
just
have
to
give
it
all.
A
B
Yes,
the
semantics
that
are
different
compared
to
normal
ipfs
is
it's
kind
of
a
closed
World
semantic.
You
assume
that
if
you're
collaborating
with
somebody,
they
have
the
car
files
that
you
need
them
to
have
and
so
yeah
you
can
go
to
the
network
and
get
them
if
you
don't
have
them.
B
But
if
you're
using
this
like
kind
of
webrtc
for
who's
in
the
live
pool
type
metaphor,
then
people
are
going
to
already
have
the
car
file
that
you
want
them
to
have,
and
so
that's
why
the
small
car
transaction
boundaries
matter,
even
if
there
is
that
accelerator
thing,
but
then,
with
the
accelerator
like
I,
guess,
yeah,
that's
like
at
some
point.
One
of
the
cloud
services
will
just
be
doing
the
repack
for
you,
so
it
doesn't
have
to
happen
on
the
browser
yeah.
C
D
Cool,
do
you
think
that
I
really
like
the
way,
you're
positioning
it
by
the
way
that
that
the
value
proposition
is
in
the
particular
audience
that
that
can
get
value
from
content?
Addressing
do
you
think
that's
going
to
limit
your
audience
ultimately,
or
do
you
think
you
have
a
strategy
to
like?
Do
you
think
the
value
proposition
is
large
enough
to
get
Beyond
those
that
can
understand,
see
and
and
and
feel
the
the
value
immediately
of
content
dressing
to.
D
A
decent,
a
decent,
you
know
web
to
back
database
back
end
in
their
browser.
I.
B
Think
that
it
positions
the
like
web
3
benefits
as
only
benefits
without
the
like
fussy
costs
right
you're
not
having
to
it,
doesn't
even
feel
like
it's
an
immutable
data
structure,
you're
just
using
it
like
you
would.
But
it's
easier
to
install
right.
The
default
experience
like
is
in
the
user's
local
storage
and
so
the
way
that
people
the
way
that,
like
it's
gonna
the
shit's
gonna
hit.
B
The
fan,
is
a
good
kind
of
when
there's
enough
developers
out
there
who
don't
know
and
don't
care
who
are
using
it
to
where
the
users
start
being
like
what
happened.
Because
Apple
takes
away
local
storage
without
asking
you
know,
and
so
it'll
be
that's
when
you
as
an
app
developer,
will
learn
to
turn
on
the
web,
3
replication
or
whatever.
B
I
had
a
hard
time
over
the
last
week,
doing
this
debug
exercise
and
I.
Think
it's
because
when
your
project
gets
too
big
to
fit
in
your
head,
it
should
be
in
typescript,
and
it's
not
right
now
and
so
I
went
through
and
made
it
throw
when
it
was
getting
stuff.
It
didn't
want
and
all
of
a
sudden
I
could
write
code
using
it,
and
so
mostly
it's
just
and
I'm
kind
of
new
to
IPL
these,
so
I
may
be
somebody
who's
been
in.
B
It
wouldn't
have
to
do
all
that,
but
it
the
you
know
like
because
the
protocols
are
so
flexible,
like
your
stuff,
you're,
expecting
to
be
your
stuff
being
a
CID
all
of.
D
A
really
very
ambitious
and
exciting
to
see
you've
made
so
much
progress
yeah.
So
thanks.
C
D
You
say
you're
going
to
be
presenting
this
again
at
ipfs
thing.
What
sort
of
progress
do
you
think
you'll
make
by
the
next
time
you
get
to
present.
B
I
have
a
bunch
of
like
a
kind
of
a
rich
and
varied
issue
tracker,
but
I'm,
really
hoping
now
that
I'm
done
with
this
core
kind
of
basically
I
was
having
a
problem
with
the
inner
node
merge
on
the
prollytree,
where
you
could
throw
data
at
it.
That
would
cause
it
to
like
not
be
able
to
read
itself
and
so
I
fuzz
tested
it
and
found
as
many
of
those
cases,
I
could
and
I
think
I'm
done.
You
know,
chat.
B
Gpt
4
is
really
good
at
fuzz
testing,
and
so
we
got
you
know
and
the
library
is
100
coverage
on
the
tooling
like
before
you
can
merge
anything.
So
we
keep
all
the
tests
there,
but
I
just
kind
of
like
restructured
the
recursion
to
be
something
that
I
could
understand
and
added
the
part
when
you
were.
B
You
know
basically
splitting
internodes,
and
that
was
a
I
mean
it's
not
much
to
look
at
when
you're
done,
but
I
think
I
was
also
learning
ipld
doing
it,
and
now
that
that's
done
all
I
want
to
do
is
write,
like
example,
apps
and
feature
unfeature
flag.
These
things
that
have
turned
off
like
the
query
engine
that
I
was
showing
you
that
was
powering
the
to-do
list
app
for
every
query.
B
It
rebuilds
the
index
from
scratch
and
I
joined
Apache
couchdb
early
enough
to
remember
when
it
did
that,
but
you
all
you
got
to
do
is
unfeature
flag.
It
it'll
hang
on
to
its
latest
index
route.
That
you
know
goes
with
that
query
and
that
introduces
semantics.
That
I
got
to
deal
with
right
like
now.
The
clock
root
of
the
whole
database
tracks
not
only
the
database
but
also
optionally,
the
indexes
that
you
have
defined.
B
Do
we
want
to
keep
the
index
definitions
around
or
right,
and
so
that's
like
more
of
a
plumbing
question
than
a
database
question.
But
I
got
to
do
that
kind
of
stuff,
better
support
for
snapshots.
You
can
grab
anywhere
on
that
tree
on
Alan's
clock
and
have
a
consistent
snapshot,
and
so
I
have
optional
rev
style,
like
updates
on
couch,
where
you
can
do
mvcc.
B
If
you
want
it,
but
I
flipped
the
bit
on
it
where,
if
you
want
to
write-
and
you
don't
care-
that's
cool
with
me-
it'll
come
up
in
Conflict
resolve
time,
so
the
there's
optional
mvcc
that
you
know
it's
easy
to
opt
into
and
then
but
there's
yeah
little
things
about
how
like
utilities
for
managing
database
snapshots.
B
That
would
make
a
lot
of
sense,
so
I,
basically
there's
some
old
apis
I
got
to
tear
out
that
probably
like
go
where
you
shouldn't
go
and
yeah
mostly
example:
apps
I
want
to
do.
Okay,
here's
where
I
could
use
help
from
this
community
it'd
be
super
fun.
B
There's
like
two
or
three
candidates
for
oh
dang,
I'm
frustrated
enough.
The
chat,
gbt
won't
get
me
access
to
my
chat
log
history
in
the
sidebar
that
I
built
my
own,
that
works
with
local
storage
right.
B
Those
are
out
there.
I
haven't
had
a
chance
to
review
them,
I'm
sure,
they're,
all
okay,
so
picking
one
of
those
and
then
making
it
use
fireproof
would
be
a
great
example
of
how
oh
yeah
I
plan
to
do
that
eventually,
but
it'd
be
a
great
example
of
how
AI
matters.
You
know
how
proofs
matter
for
AI.
A
B
I
already
added
full
text,
there's
full
it
took
me
like
you
know,
like
50
lines
of
code
or
something
so
yeah.
If
there's
a
Geo
indexer
that
supports
incremental
that
you
know
that's
out
there,
like
all
npm
install
it,
I'll
put
it
on
the
Dev
dependencies
and
right
plugin,
for
it.
A
Okay,
yeah
because,
like
that's,
obviously
what
I'm
interested
in
but
yeah,
but
it's
like,
but
it
would
need
to
be
but
like
like.
How
is
your
full
text
like
it's
a
full
text
thing
then
also
working
with
a
Merkel
tree,
or
is
it
just
like
an
usual.
B
It's
just
an
image
if
I
wanted
it
to
be
big.
You'd
have
to
write
it
on
top
I'm
excited
to
see.
I
haven't
like
looked
closely
enough
at
what
rhizome's
doing
and
so
I'm
excited
to
see
if,
like
rhizome
indexes,
can
just
plug
and
play
I
could
track
them.
With
my
clock,
you
know,
and
so
there's
gonna
be
a
lot
of
things
you
can
do
if
you're
design
goal
isn't
making
like
a
something
that
everyone
who
does
react,
apps
with
npm
install.
No
questions
asked
like
I.
B
A
A
Yeah,
you
can
also
do
that.
Yeah
yeah,
that's
the
Hot
Stuff,
currently
yeah,
that's
true,
but
yeah
so
yeah
yeah
I
mean
it's
also
something
that
that
Mo
wants
to
wants
to
look
into
effort.
C
B
Yeah
I
mean
I,
think
I
wish
we
had
a
vector
index
already,
because
people
can
use
that,
like
especially
in
this
context,
you
don't
even
need
to
have
the
horsepower
to
run
a
vectorization
engine.
You
know
like
you.
C
Could
if
you
had
a.
B
A
D
B
I
just
need
the
AI
vendors
to
start
signing
their
output
right.
Like
the
same
you
just
you
can't
use
the
https
session
for
non-repudiation,
so
they
would
need
to
actually
do
like
an
API
signature.
Key
right.
D
B
C
D
Yeah
on
and
you
gotta,
you
got
you
gotta
hook
in
with
AI
to
get
the
all
the
buzzwords
but
there's
definitely
a
fun
buzzword.
B
Yeah
well,
I
mean
I
I'm
Legit
excited
about
it,
I
think
the
the
GPT
like
kind
of
local
or
like
a
any
one
of
the
image
generation
things
especially
any
of
the
ones
where
there's
a
community,
that's
making
alternative
uis.
B
B
D
Yeah
this
will
be
on
YouTube.
It
does
get
some
views
afterwards,
so
there
might
be
comments,
questions
on
there,
but
I
I
assume
there
were
links
in
the
presentation.
Where
should
people?
Where
do
you?
Where
would
you
direct
people
to
go
if
they
had
more
questions.
B
Oh
yeah,
if
you
wanted
to
go
to
fireproof
dot
storage,
that's
going
to
get
you
to
everything
right.
B
B
So
what
we're
doing
is
we're
just
getting
the
changes
off
of
the
changes
API
that
are
like,
since
this
clock
that
we
just
keep
in
memory
for
this
right,
the
lifetime
of
the
page
and
then
every
time
you
query,
you
bring
it
up
to
date
by
pulling
those
changes
in
and
then
run
the
search
on
the
latest
version
of
the
full
text,
but
yeah,
it's
not
much
code
with
it
and
you
know
it
works.
A
C
B
Yeah
it
would
make
it
would
probably
end
up
being
like
super
fast
on
maps,
because
in
the
past
I've
done
map
work,
where
the
answer
turned
out
to
be
to
just
load
the
CSV
file
and
window
off
of
that.
But
I
bet
this
could
be
faster.
A
C
A
D
Past
so
I
I
think
I.
Last
time,
I
talked
about
the
parallel
traversal
stuff
that
I
was
working
on
and
that's
getting
getting
ready
to
land.
So
it's
you
know,
go
rpld,
Prime,
pull
request.
Four
five
two
I
think
the
interface
is
is
is
right
that
we
want
to
export
there.
It's
just
it's
a
preload
of
thing
for
traversals
and
then
in
Lassie,
as
I
think
I
mentioned.
D
There's
a
implementation
of
that
that
parallelizes
the
the
pre-fetching
of
blocks,
so
that
you
can
do
you
can
do
things
like
traversals
over
bit,
Swap
and
control
the
parallelism
of
that
and
have
stuff
running
in
the
background,
and
you
can
do
yeah,
so
you
can
end
up
having
really
good
performance
of
yeah.
D
So
so
the
the
aim
for
this
one
is
it
lands
in
in
Saturn
where
these
L1
nodes
on
Saturn
are
doing
the
thousands
of
thousands
of
fetches
from
ipfs
and
falcoin
at
a
time
and
each
one
of
those
fetches
could
be
parallelized.
So
this
gives
sort
of
the
control
to
to
do
to
to
just
to
to
spread
out
the
fetching
for
each
of
those.
D
Anyway,
it's
it's
working
quite
nicely.
It's
a
good
model
there
I
think
and
something
to
build
up.
There's
a
there's,
a
whole
framework
for
controlling
the
parallelism
in
Lassie.
That
I
will
eventually
pull
out
into
guy
building
Prime
for
others
to
use.
But
if
anyone
who's
listening
to
this
has
feedback
on
that
Now's,
the
Time
to
jump
in
because
that
it
might
get
merged
very
soon.
D
C
A
I
got
that's
the
main
thing:
I've
been
working
on,
that's
I
feel
like
do.
I
spend
a
bit
of
time
in
on
Rust
multi
hash,
because
there's
a
PR
open,
which
is
a
huge
refactoring
into
different
crates
and
so
on,
and
but
I
I
will
talk
about
it
once
it's
done
because
like
it
would
still
take
a
while,
and
but
this
is
more
like
really
a
refactoring,
but
it's
the
idea
is
that
people
need
less
braking.
A
So
if
we
change
something,
we
need
less
breaking
change
releases
because
it's
split
into
pieces
and
the
core
piece
should
usually
stay
the
same.
So
that's
the
whole
idea
because,
like
as
more
and
more
people
depend
on
multi-ash,
it's
really
a
pain.
If
we
just
yeah
make
a
small
change
and
then
it's
a
breaking
release
technically
and
then
everyone
has
to
update-
and
this
should
make
things
better
and.
C
A
Happy
to
get
those
third-party
contributors
to
actually
yeah
do
those
work,
so
I
only
I'm
only
reviewing
it
but
yeah
cool,
then
I,
guess.
The
important
thing
for
this
meeting
is
that
it's
usually
is
it
used
to
be
every
two
weeks
but
moving
on,
we
will
do
it
every
month,
because
it's
more
convenient
for
Rod
and
me
M,
which
means
that
the
next
meeting
will
then
be
on
April,
the
24th.
A
Fourth,
yes
yeah,
that
will
be
the
next
meeting
and
if
it
turns
out
that
we
will
basically
have
like
more
people
survival,
we
have
too
much
to
discuss
within
one
hour.
We
can
always
go
back
to
every
two
weeks,
but
so
far
we
haven't
had
that
much.
So
it
just
makes
sense
to
make
it
monthly.
D
B
A
D
So
maybe
we
should
have
a
just
a
very
quick
brainstorm
of
the
kinds
of
things
that
we'd
like
we'd
like
to
have
covered,
so
people
would
like
to
invite
to
talk
about
their
thing
if
we've
got
any
ideas,
because
I've
got
a
couple
ideas,
if
you
want
to
hear
them.
A
Yeah,
yes,
true,
yeah,
so
yeah
because,
like
yeah,
there's
also
like
four
basically
people
watching
this
recording.
We
would
like
to
have
like
something
like
take.
Christmas
talk
for
every
meeting,
because
it's
just
yeah
very
interesting
and
yes,
if
anyone
has
ideas,
feel
free
to
reach
out
our
channels
or
issues
or
whatever
but
yeah.
What?
If
you
have
ideas,
then.
D
Yeah,
it's
a
top
of
mind
for
me.
I
I,
I'd
like
to
get
Iraqi
in
to
talk
about
the
work
he's
been
doing
with
ucans
and
that
ecosystem
on
rpld
and
the
way
he's
been
using
the
tooling
to
solve
problems.
There
he's
doing
some
interesting
stuff
in
schemas,
which
is
he's
he's
he's
he's
currently
got
the
most
advanced
usage
of
schemas
in
JavaScript
and
he's
making
heavy
use
of
them
for
ucans,
which
is
really
exciting.
D
D
I
I
wouldn't
mind
at
some
stage
when
the
dust
is
settled
to
if
we
could
get
Stephen
on
to
talk
about
fem
and
ipld
and
how
and
how
that
works,
because
there's
some
there's
some
really
interesting
juggling
of
ipld
as
a
as
a
state
storage
for
fem,
smart
contracts
and
the
way
that
interacts
with
the
chain
and
all
that
whole
dynamic
of
how
to
how
to
do
wasm
rust,
interacting
with
the
capture
chain
and
all
that,
like
there's,
some
crazy
mechanics
going
on
with
that
people.
D
Do
you
would
love
to
hear
about,
and
the
other
and
related
to
that
I
wouldn't
mind
having
some
kind
of
I
don't
know
if
it's
a
panel
or
just
a
group
discussion
between
maybe
Volker
yourself
and
Stephen,
maybe
others
about
the
the
future
rapidly
in
Rust,
and
what
that?
What
the
potential
is
there
for
to
do?
Some
rethinking,
because
I
know
Steven's
done
a
lot
of
rethinking
about
where
that
could
go
and
I'd
love
to
get
that
on
the
table.
A
Yeah,
that's
that's
yeah
sounds
sounds
pretty
good
yeah.
Also
what
likes?
What
might
be
like?
It's
not
directly
ipd
related,
but
kind
of.
It's
also
like
the
the
work
Michael
is
currently
doing
with
his
I
forgot.
The
name
of
the
project
but
kind
of
like
car
files,
but
like
Next,
Generation,
clarifies
I
would
say
which
is
also
kind
of
interesting,
which
I
I
didn't
fully
grasp
it.
Yet,
but
yeah.
D
B
Yeah,
okay,
so
there's
this
thing,
I
noticed
in
his
new
style
of
stuff
is
in
the
current
or,
like
you
know,
a
lot
of
the
ipld
code
I've
been
working
with,
has
like
a
block,
a
a
put
style
semantic
relationship
with
the
blocks.
Like
you
know
your
own
CID
before
we
put
it
and
then
the
stuff
he's
been
doing
lately
is
more
of
a
post
semantic
right.
You
just
post
to
the
content,
and
it
gives
you
back
the
CID
and
they're
equivalent,
but
I
feel
like
the
post.
B
Semantic
is
a
lot
more
like
forgiving
in
the
application
and
so
I.
Don't
know
that
might
be
a
thing
to
be
interesting.
To
talk
about.
A
D
A
D
B
Yeah
so
I
might
he's
got
a
car
transaction
library
that
I
can't
use
because
I'm
like
hard
coded
with
put
semantics,
but
that
might
be
if
I
was
going
to
do
a
heavy
lift.
You
know
that
doesn't
change
my
test
Suite.
It
might
be
moving
to
that
car
transaction
Library.
A
Yeah
and
and
of
course,
the
checkers,
if
you
discover
like
any
like,
like
yeah
usability
issues
or
edge
cases
or
whatever
like
which
just
doesn't
work,
you
have
feel
free
to
open
issues,
because
yeah
I
would
also
say
the
like.
The
yeah,
like
the
like
ipld,
like
especially
I,
think
in
JavaScript.
It's
I'm
really
sure
it's
true
but
like
it's
not
like
it's
used,
but
it's
also
used
by
people
that
are
just
too
deep
into
it.
So
it's
basically
yeah.
C
B
B
Like
the
thing
I
just
said,
I
don't
know
where
I
would
find
like
to
put
my
finger
on
it.
I
just
know
like
because
it's
what
it
comes
up
when
you're
like
deep
in
the
middle
of
debugging
like.
Why
is
this
thing
serializing
like
that?
Not
like
that
and
then
you're
just
Json
inspecting
your
stuff,
I
mean
I.
Guess
yes,
Json
inspect
comes
out
weird,
unless
you
you
know,
do
it
right
and
nobody
does
things
right
and.
B
It's
like
it's
an
unnecessary
await,
which
now
it's
like,
infecting
all
your
Upstream
stuff
with
the
you
know
like.
If
you
do,
if
you
move
the
strict
post
semantics,
then
you
can
assume
that
you
have
the
CID
and
you
don't
need
to
like
verify
it
or
whatever.
D
D
D
Right,
yeah
I.
There
has
been
a
discussion
about
possibly
even
moving
away
from
that
and
using
some
of
these
user
land
implemented.
Crypto
algorithms.
B
Async
crypto
makes
sense,
it's
just
I,
don't
think
it
actually
has
to
implement
it
has
to
affect
the
CID
implementation.
You
know.
B
D
I
mean
so
that'll
be
in
JS
multi-formance.
If
you,
if
you
can
put
your
finger
on
the
thing
that
maybe
open
an
issue
and
we
can
probably
connect
it
with
some
existing
discussions
that
we're
having
because
Iraqi
also
has
ideas
about
refactoring
the
CID
interface
to
be
I,
don't
know
more
interesting,
there's,
there's
alternative
ways
of
representing
a
CID
that
we've
mainly
he
has
explored
yeah.
B
I
mean
I
think,
there's
how
it
looks
and
then
there's
whether
or
not
you
have
to
await
it.
You
know,
and
getting
away
from
the
await
would
be
just
as
big
of
an
impact
yeah,
but.
A
C
C
A
B
I
think
so
I
mean
I.
Think
that's
why
I
want
to
move
to
these.
So
if
I
move
to
car
transactions,
I
have
to
go.
Do
the
line
level
changes
for
that
and
then
I
think
I.
Can
that
makes
me
forward
compatible
for
the
blocks
of
stuff.
A
All
right
is
there
anything
else,
not
then,
as
always,
we
have
a
after
party.
So
basically,
if
the
live
stream
closes,
we
still
hang
out
in
case.
Someone
has
something
that
they
want
to
share,
but
one
don't
want
to
share
publicly
so
feel
free
to
hang
out
and
yeah.
So
then
a
goodbye
everyone
on
the
stream
and
see
you
in
four
weeks
again
and
I'll,
see
y'all
in
Brussels
yeah,
bye,.