►
From YouTube: 🖧 IPLD Every-two-weeks Sync 🙌🏽 2022-08-02
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
Yeah,
I
don't
know
if
there
was
a
meeting
two
weeks
ago
or
not
so
perhaps
the
meeting
notes
are
carried
over
again
rod.
B
Yeah,
so
there
was
a
meeting
and-
and
it
was
just
me
meeting
with
myself-
okay,
it
was.
It
was
a
fascinating
meeting,
absolutely
fascinating.
So
the
notes
where's
my
tab.
B
The
notes
in
the
current
doc
are
the
leftovers
from
that
meeting
that
I
wrote
up
for
myself
that
I'm
gonna
repeat
today
so
and
I'm
just
going
to
add
another
thing
to
this
because
I've
just
made
something
else
happen:
hello.
B
So
my
notes
are-
and
this
is
this-
is
reaching
a
little
bit
back
now
so
goca
had
a
security
release
a
number
of
week
back
weeks
back
just
wanted
to
put
that
on
people's
radars.
If
they
haven't
seen
that
because
they
probably
should
upgrade
upgrade
there's
an
advisory
in
the
dock
or
you
can
just
go
to
the
go
car.
Repo
and
you'll
find
a
link
to
it.
It's
it's.
It
was
to
do
with
allocation
related
problems.
B
Jaropo
did
some
fuzzing
and
found
some
some
ways
that
our
bite
arrays
could
be
allocated
that
are
way
too
large
buy
a
malformed
car.
So
there's
a
few
places
there
was
a
there
was
a
couple
of
panics
really
and
a
few
of
them
are
different,
but
we
wrapped
them
all
up.
There's
still
issues
with
go
car
v2
indexes
it
is
possible
to
receive
a
malformed
go-kart
index
and
and
have
that
cause
problems
for
you.
B
But
as
part
of
this
release,
we've
sort
of
increased
the
wording
in
the
api
about
not
consuming
indexes
that
you
don't
trust.
It's
just
not
a
good
idea.
The
kv2
index
stuff
is
not
intended
to
be
something
that
you
pass
around
between
untrusted
parties.
It
was
meant
it's
sort
of
like
a
it's
something
that
you
use
for
your
own
concerns,
and
then
you
pass
around
the
v1
and
regenerate
the
index
as
you
need
it.
So
we've
increased
the
language
there.
B
We
do
have
a
to
do
to
to
deal
with
the
issues
in
the
indexes,
but
they're
a
little
bit
more
there's
a
little
bit
more
change.
That
needs
to
go
on
to
make
that
happen
nicely.
So
we'll
get
to
that.
I
have
another.
It's
an
another
thing
to
do
with
cars.
Is
I've
got
a
pull
request
in
each
of
gokar
and
js
car
this?
This
came.
It
was
it's
an
experiment.
B
They
draft
pr's,
I
don't
know
if
they'll
get
merged,
I
don't
know,
maybe
they'll
sit
open
for
a
while
collecting
a
bit
of
dust,
but
we'll
see
the
request
was
it
was
initially.
We
want
to
be
able
to
put
more
metadata
into
cars,
so
we
want
to
extend
the
car
header
and
just
pack
it
full
of
stuff
and
and
that's
a
bit
of
a
no-no
for
the
current
format,
because
it's
just
going
to
break
stuff
like
we
could.
We
could
do
that,
but
it
would
be
a
car
b3
kind
of
thing.
B
So
so
I
instead,
I
took
the
car
v2
flexibility
around
where
the
the
the
cavity
one
off
the
copy
one
is
placed
within
the
v2
and
sort
of
pushed
it
back
and
put
a
little
flag
in
to
say
that
there's
a
a
message:
object
that
is
embedded
after
the
kv-1
header.
B
So
this
then
allows
you
to
encode
an
arbitrary
dagsy
or
encoded
blob
as
a
as
a
message
in
a
car,
and
you
can
do
things
like
you-
can
store
metadata
about
the
kv
one.
So,
let's
say
you're
using
a
kv-1
for
a
far
coin
deal.
So
kv-1
needs
to
be
sort
of
pristine.
This
thing
gets
stored
and
it
has
a
compe
which
is
the
prove
the
proof
of
the
the
bytes.
B
You
could
put
that
compe
in
this
message
and
you
could
put
other
things
about
it
that
don't
necessarily
go
in
the
thing,
so
it's
a
good
place
for
metadata.
Whether
or
not
this
would
be
needed.
It's
a
different
question.
I
don't
really
want
to
push
ahead
with
it
unless
there's
a
real
need,
it's
sort
of
like
it's
slightly
unnatural,
but
it
does
work
so
anyway,
feedback
welcome.
But
again
I
don't
know
if
this
will
be
flipped
to
a
proper
pr
until
there's
a
real
need.
B
Lastly,
I
just
I
click
merge
on
go
up
with
d
prime
0
18
0,
and
I
will
put
some
extra
notes
in
in
here,
but
the
the
changes
are
all
in
by
node.
There's,
there's
three
major
things
that
come
in
by
node
one
one
is
custom
type
converters
where
so
by
node
is
really
good
with
inferring
how
to
translate
from
plain
go
types
to
data
model
nodes,
but
it's
often
the
case
that
you
want
to
work
with
more
than
plain
types
and
or
you
have
your
own
specific
ways
to
encode
things.
B
So,
there's
some
really
good
examples
in
filecoin
that
this
was
actually
one
of
the
work
came
out
of
things
like
there's,
there's
other
places
where
we
use
big,
big
dot,
int
that
and
we
can't
use
int
for
that,
but
but
we
can
use,
but
they
do
get
encoded
as
bytes.
So
by
with
this
change
you,
when
you
wire
up
a
type
with
bind
node,
you
say
that
the
big.int
or
otherwise
known
as
what
is
it
called
token
amount.
B
There's
a
big
wrapper,
there's
actually
two
layers
of
big
end
wrappers
in
there,
but
it's
all
the
same.
So
you'd
say
I.
Whenever
you
encounter
a
token
amount,
use
this
encoder
and
decode
a
pair
to
get
it
to
and
from
bytes.
So
as
a
go
type,
you
get
the
token
amount,
but
as
a
node,
when
you,
when
you
get
the
ipl
data
model,
form
of
it,
you
get
it
bytes,
and
so
this
allows
you
to
extend
the
kinds
of
types
that
you
can
have
in
your
go,
structs
and
and
etc
for
by
node.
B
Another
thing
it
does:
is
it
wires
up
that
ui
64
support
that
I've
talked
about
before,
so
that
in
his
inbind
node?
So
now
you
can
your
go
type.
You
can
use
un-64
and
get
the
full
range.
B
And,
lastly,
there
is
a
just
a
utility
registry
in
there
for
by
node,
allowing
you
to
a
central
place
to
register
and
perform
operations
on
your
go
types
with
by
node,
so
you
initially
set
up
a
type
with
a
schema.
So
this
is
my
go
type.
Here's
my
schema,
you
register
them
and
then
later
on,
you
perform
operations
based
on
that
type.
Either
you
give
it
that
type,
and
you
say
I
want.
B
I
want
the
data
model
form
of
it
or
I
want
it
to
be
encoded
with
dax
ebook
or
whatever,
and
you
say
this
is
my.
This
is
my
type,
and
this
is
what
I
want
to
do
with
it
and
yeah.
So
it's
just
it's
just
a
utility
thing:
it's
not
necessary,
but
it
does
make
your
life
a
little
bit
easier
when
you're
dealing
with
multiple
types
through
by
node,
so
they're
my
main
things
who
would
like
to
go
next,
I
didn't
you're
still
writing
said
you
want
to.
C
I
can
I
can
go
next,
but
I
had
a
question
about
or
a
question
slash
comment
about:
the
car
v2
indices,
which
is
so
even
though,
like
loading
them
all
loading
them
all
into
memory,
I
guess
is:
is
a
trusted
operation.
C
I
suspect
that
you
could
use
some
of
the
like
reader
app
for
indices
stuff
to
let
you
do
things
like
sort
of
like
the
equivalent
of
running
bitswap
against
a
car
file
yeah,
which
is
to
say
like
I
could
be
like.
Do
you
have
this
block?
B
So
so
giropo
actually
did
that
in
the
first
round
of
this
he
actually
came
up
with
a
a
thing
and
we
managed
to
make
it
work
nicely
where
it
was
like
a
lazy
load
of
indexes,
and
it
would
do
some
of
that
for
the
index
types
that
we
cared
about.
It's
just
that
they
required
a
big
api
change
and
that
api
change
was
a
bit
too
traumatic
for.
B
Yeah
it
was,
it
was,
and
it
was
also
like
like
to
be
clear
to
be
honest
about
this.
It
was
it
was
boost
that
was,
is
the
biggest
consumer
of
the
go
cars
v2,
api
and
and
boost
was
really
like.
It
was
already
a
big
change
to
get
boost
to
adopt
some
of
these.
B
These
security
fixes,
but
having
a
lazy
loading
of
the
indexing,
was
just
a
bit
too
much,
and
so
we
punted
on
that
for
now,
and
I
don't
know
whether,
because
because
there
are
a
lot
of
cases
where
you
just
want
to
say,
I
want
to
load
this
all
in
memory
because
it's
fast.
So
I
think
this
is
the
thing
where
we'll
probably
have
to
have
different
apis,
where,
if
you
want
to
read
an
untrusted
one,
then
here
you
can
do
it
this
way,
but
it
still
brings
it
back
to
the
thing
of.
B
C
Well,
unless
you
want
to
do
what
what
I'm
saying
right,
which
is
like
you,
have
a
car
file
and
I
want
to
download
some
subset
of
the
blocks
inside
of
the
car
file
right
like
basically,
I
want
to
have
like
you
just
host.
Like
the
you
know,
all
of
the
current
snapshot
of
wikipedia
as
a
car
file
with
a
humongous
index,
and
I
want
to
be
able
to
download
that
index.
C
I
won't
be
able
to
download
the
subset
of
wikipedia
that
I
want
over
http,
and
I
can
do
that
because,
like
I
can
grab
the
blocks
from
the
car
file
that
I
need
to
yeah
as
as
dictated
by
by
the
index
and
like
if
your
index
is
lying
to
me,
then
I'll
download
some
data.
I
don't
like
and
then
I'll
be
like.
I
don't
like
you
anymore,
no
more
using
you
yeah
that'll,
be
the
end
of
it.
C
Is
that
code?
Is
that
code
that
that
joppa
wrote
anywhere
at
the
moment
or
is
that,
like
a
branch,
a
closed
branch
and
like
a
security.
B
With
the
intention
was
to
pull
it
across
okay,
so
if
it's
not
there,
let
me
just
have
a
look.
Maybe
it's
not
so
there
is
a
there's
an
index.
That's
sorry,
there's
an
issue
number
three
and
four
that
draw
up
open
but
okay
and
he
links
to
the
go-kart
repo
with
the
you
should
have
access
to,
but
we
should
just.
B
C
Yeah
yeah,
that
makes
that
makes
sense
all
right.
I
guess
I'll
go
on
with
a
couple
of
things
that
were
on
my
list.
The
first
one
is
a
proposal
to
move
the
management
or
the
governance
of
the
cid
spec
to
the
ipfs
specs
process
or
the
ipip,
and
there's
an
issue
for
that
that
I
linked
in
multi-format
cid.
C
The
reason
I
am
asking
this
in
general
is
because,
even
though
nobody
has
made
a
multi-format
spec
pr
in
quite
a
long
time,
michael
has
put
one
out
for
a
cid
v2
thing
and
even
aside
from
any
technical
considerations
around
that
proposal,
the
question
is
like:
how
are
we
going
to
make?
How
are
we
going
to
figure
out
how
this
goes
and
while
those
folks
interested
in
ipld
and
who've
been
carrying
a
lot
of
the
burden
there
generally?
Are
the
ones
who've
been
carrying
a
lot
of
the
multi
multi
formats?
C
Aside
from
multi-outer
burden,
ipld
has
less
of
has
like
less
specs
process
around
it
so
figured.
We
would
just
lend
ipfs's
for
now
and
if
an
ipl
d1
emerges
then
we'll
hand
it
right
back
yeah.
If
you
have
thoughts
thumbs
up
thumbs
down
other
things,
that'd
be
a
good
place
to
to
drop
them.
D
C
I
think
we
should
leave
it
in
the
place
that
people
are
used
to
it
being
in
it's
more
just
like.
What's
the
process
we're
gonna
use
for
when
we
want
people
to
write
things
up
in
a
particular
way
that
gives
sufficient
information
which
of
the
various
ways
in
which
you
could
write
something
up.
Would
we
like
you
to
do
it
in
so
people
have
a
consistent
way
and
then,
when
we
need
to
decide
how
we're
going
to
do
our
our
version
of
you
know
rough
consensus,
you
know
how
we
gonna
do
this
and
figure.
C
We
need
to
figure
this
out
in
the
ip
in
the
ikfs
context
and
there's
a
larger
crowd
of
people
there
who
are
all
who
are
all
affected
by
basically
a
lot
of
the
ipld
specs
anyway.
So
might
as
well.
Do
it
there
yeah
again
if,
if
a,
if
a
larger,
if
there
was
more
of
this
in
the
ipld
context,
they'd
say
like
yeah,
just
the
cid
spec
is
used
by
a
lot
of
people,
but
ipld
seems
the
best
steward
of
that
spec
just
seems
like
in
the
meanwhile.
A
So
I
think
it
should
be
changed,
but
I
would
also
be
happy
to
basically
going
through
a
process
because,
like
I
wouldn't
feel
comfortable,
basically
just
like
doing
a
pr
on
the
spec
and
then
rod
says
it's
okay
and
then
it's
merged
and
like
because,
like
it's
such
a
central
spec,
it's
like
yeah.
If
it
go
through
the
ipfs
process,
apip
process,
I
think
it's
basically
yeah.
I
would
feel
better
about
making
changes
because
like
yeah,
then
we
have
agreement
somehow
so
yeah
cool.
B
I
just
I
have
an
item
on
my
task
board
to
do
governance
stuff
in
multi-formats.
It's
a
long-standing
one,
so
this
is
probably
a
good
step
since
I
haven't
got
to
it
yet
yeah.
I
I
I
is
it.
Is
it
simply
just
a
a
is
it
is
it
did?
Does
it
come
out
of
this
emergency
that
there's
because
of
this
proposal
that
we
don't
want
to
like
yeah.
C
My
first
comment
when
I
read
it
was
like
this
is
missing
sufficient
context
for
most
people
in
the
community
to
evaluate
this,
can
you
add
more
context
and
the
response
was
well.
I
don't
know
how
the
governance
process
for
this
works.
I
don't
want
to
invest
more
time
into
writing
up.
C
If
we
don't
have
like
a
mechanism
for
deciding
what
happens
at
the
end
of
this,
and
so
that's
kind
of
where
this
comes
from
again,
nobody
has
touched
the
multi-formats
specs
in
a
long
time.
So
no
one's
had
needed
to
think
too
hard
about
this.
C
E
C
Aside
from
just
like
making
a
new
thing
right,
like
anyone
can
always
specify
whatever
it
is
they're
doing,
no
one
can
stop
anyone
from
writing
any
code
right
but
yeah.
I
suspect
the
more
the
more
ingrained
a
particular
spec
is
in
the
concept
and
what
that
name
means
the
more
consensus
you
need
to
change
it
to
whatever.
The
next
thing
is.
I
can't
just
make
up
whatever
I
want
and
make
it
http
for
right.
I
mean
there's
like
a
lot
of
stakeholders
and
what
http
means
then
getting
from
one
to
one
point.
E
B
Yeah,
I
know
I
agree,
and
I
I
just
I
I
have
no
problem
as
well.
I
just
have
it
it's
instilled
in
my
backlog
to
look
at
the
pr
and
and
just
have
a
look
over
the
deals,
because
I'm
not
I'm
not
familiar
with
the
spec
process.
Ipip
it's
day
process,
so
I'll
have
to
look
at
that
and
yeah,
but
it
did
to
me
on
principles.
It's
fine
yeah
because
I
do
see
difficulty
in
moving
forward
with
the
cidv2
thing.
B
It's
just
and
it's
a
difficulty
just
in
terms
of
process,
given
that
it's
not
a
it's,
not
a
it
out
of
the
park
type
of
thing
where
everyone
looks
and
says,
yeah,
that's
awesome.
I
mean
a
lot
of
people
are,
but
a
lot
of
people
are
saying
so
that
kind
of
so
that
kind
of
situation
does
make
it
tricky.
Yeah.
C
Yeah,
and
even
I
guess
to
to
be
clear,
like
you
can
look
at
the
linked
stuff,
the
the
ipip
like
the
process
is
not
explicit
about.
You
know
it's
not
explicit
about
some
of
the
things
that
tend
to
end
up
being
fuzzier
anyway,
like
what
does
rough
consensus
mean
in
a
group
that
doesn't
have
like
right.
It's
not
like
a
formal
like
no
like
formalized
here.
Are
the
entities
and
their
respective
voting
powers
it's
just
like.
C
Do
we
tend
to
agree
on
this
thing
or
or
not,
but
at
least
the
rest
of
it,
that's
like
why?
Why
are
you
doing?
What
is
the
change?
Why
are
you
doing
it?
What
were
the
alternatives?
What
are
the
ramifications
that
kind
of
stuff
is
there.
C
Yeah,
okay,
so
that
was
that
was
that
one
another
yes
reporting
on
things
happening
happening
elsewhere
that
I
happen
to
have
seen.
There
are
some
people
in
the
iqs
community
who've
been
poking
at
some
webassembly
stuff
and
some
people
have
reached
out
who
are
like
hey.
You
should
try.
You
should
look
at
this
thing
polywrath
and
consider
that
for
how
you,
you
know,
write
your
code
and
deal
with
your
abis
and
and
whatever
so
they
will
be
having
a
chat
with
some
folks.
C
C
You
got
to
define
you
got
to
define
your
like
interfaces
to
the
outside
world
if
you've
written
ffi
code.
It's
the
same
deal
like
what
are
my
c
exports,
except
that
in
this
world
everything
is
an
integer
everything,
integers
integers,
as
far
as
the
eye
can
see
integers
and
there
have
been
various
specs
or
attempts
at
like
abstract
of
creating.
C
You
know
standardized
ways
to
do
things
that
aren't
integers
like
you
know,
white,
arrays
or
strings
or
floats,
or
whatever
all
these
things
that
you
might
may
sound
familiar
from
your
ipld
data
model
spelunking
and
none
of
those,
as
far
as
I
can
understand,
have
crystallized
into
a
real
thing
yet,
and
so
everyone
is
still
kind
of
yellowing
it,
including
the
stuff
that
I
I
yolowed
for
what
my
field
exists
today.
So,
ideally,
maybe
more
standards
less
yellow
and
then
you
know
we
have
to
define
our
own
abis.
C
On
top
of
that
for
the
functions
we
care
about,
but
like
at
least
some
of
the
basics,
you
know
strings
errors
whatever
could
be
covered
by
that.
Yes
and
speaking
of
the
wasn't
my
pld
stuff,
there's
like
a
pr
that
just
sort
of
diffs
like
the
origin
with
now.
If
you
have
questions
about
how,
if
you've
like
been
poking
at
the
code,
you
want
to
wonder
how
any
of
that
works.
You
can
ask
questions
there.
C
I'm
also
happy
to
give
people
a
more
synchronous,
walkthrough,
and
I
guess
calling
out
that
I
will
be
less
present,
starting
in
like
a
month
from
now
for
for
a
while,
so
yeah.
If
there
are
things
that
you're
like
hey.
What
were
you
doing
about
this,
then
you
want
to
ask:
let's
do
that
sooner
rather
than
later,
and
that
is
all
for
me.
Oh
also,
I
guess
there's
a
bunch
of
videos
and
stuff
and
and
interesting
talks
that
came
out
of
the
ipfs
thing
event
around
ipld
stuff.
C
Hopefully
they
will
be
relatively
soon
and
I
recommend
taking
a
look
because
I
feel
like
we
made
a
bunch
of
progress
in
terms
of
helping
people
understand
some
of
the
leverage
points
that
we
get
out
of
using
ipld
as
an
abstraction
layer
for
these
different
hash
link
data
structures,
rather
than
everyone
trying
to
make
their
own
next
best,
unix,
fs
and
then
trying
to,
I
don't
know,
advocate
for
them
to
be
what's
called
unix
fsv2
or
something
like
that
like
surely
we
have
better
ways
like
just
allowing
you
to
signal
here
is
here
is
my
better
version
of
a
file
system
structure
and
then
letting
people
use
it.
C
Yeah
it,
it
could
be
a
few
things.
The
one
thing
that
I
am
determined
to
get
to
the
bottom
of
is
like
at
the
beginning
of
every
one
of
these
events.
It's
like
okay.
Can
we
get
videos
like
the
day
of
so
that
we
can
see
what's
been
going
on
if
they're
just
raw,
so
we
can
like
at
least
get
the
info
out,
and
it's
like
oh
yeah,
yeah,
we're
totally
gonna.
C
Do
it
this
time,
and
it
never
happens
and
I'd
like
to
understand
if
this
is
like
an
actually
fixable
thing
or
if
this
is
important,
because
every
so
often
you
give
a
presentation-
and
you
make
a
mistake-
you
leak
your
like
machine's
credentials
or
you
show
an
email
you
weren't
supposed
to,
or
you
say
something
that
you
weren't
allowed
to
or
whatever
and
people
are
concerned
about
that,
and
they
want
to
get
those
all
cleaned
up
before
the
videos
go
out.
C
D
I
have
a
question
about
the
wasm
stuff,
so
you
mentioned
you're
gonna
be
shifting
focus
in
a
month.
From
now
is,
is
there
someone
that's
like
leading
the
way
on
all
the
wasm
stuff
other
than
you,
because
I
think
there's
the
ipvm
stuff
as
well,
where
we've
got
a
group
now,
but
what's
the
relationship
between
that
and
the
work
you're
doing
and
like
who
else
cares
about
this?
As
like
an
actual
project.
C
Hopefully
some
of
the
details
around
that
working
group
will
show
up
soon
give
a
little
idea
of
where,
where
people
where
people
are
actively
trying
to
do
work,
I
kind
of
wanted
to
have
some
functioning
code
that
I
could
show
people
to
work
from,
because
again
you
can
sit
there
and
like
bike
about
these
abis
like
all
day,
because
everything
is
integers
as
far
as
the
eye
can
see,
and
there's
no
good
way
to
do
anything,
but
you
can
just
like
choose
something
and
then
it
works.
C
But
what
you
do
need
is
to
figure
out
things
like
how
do
I
do
adl
signaling
anywhere
right
since
that
brings
us
to
like
the
thornier
stuff,
around
yeah
ipld,
whether
I
kill
the
uri
schemes
or
changing
the
icfs
uri
scheme
or
whatever
you
know,
like
those
sorts
of
things,
become
more
prevalent
or
things
like.
C
If
I
wanted
to
describe
a
file
system
more
abstractly
right,
how
would
I
do
it
right
so
that
I
could
I
can
do
things
like
export
as
a
tar
file,
both
unix
fs
and
bittorrent
directories
or
git
directories
right
like
how
do
I
do
both
of
those?
So
I
think
what's
interesting.
Is
that
those
sort
of
surfaces
like
the
real
questions
and
then
like?
C
C
There
are
folks
who
have
there's
some
like
interesting
trade-off
stuff
around
the
people
who
want
to
like
autoload
in
who
auto
want
to
like
basically
automatically
load
in
codecs
and
adls,
and
that
sort
of
thing
by
the
idea
of
the
webassembly
and
then
how
you
want
to
deal
with
different
versioning
around
all
of
that
and
apis,
and
that
sounds
like
another
level
of
complexity.
But
I
figured
a
step
at
a
time.
If
we
can
do
like
hard-coded,
we
can
do
hard-coded
stuff
and
we
can
evolve
the
abi.
C
Then
we
can
figure
out
a
better
abi
and
standardize,
and
and
all
of
that
that
was
that
was
sort
of
the
approach
I
was
taking.
There's
nobody
who's
like
designated
as
the
taker
over
of
the
wasm
ipld
repo.
I
will
happily
transfer
it
to
the
iplb
org
and
let
anybody
who
wants
to
play
around
with
it
play
around
with
it.
C
B
Sorry,
this
problem
of
of
the
interface,
you
know
transferring
anything
other
than
bytes
between
muslim
and
the
runtime.
This
has
been
a
long-standing
thing
for
wasn't
like
we
were
looking.
This
is
this
was
why
we
didn't
go
further
three
years
ago
with
ipld,
when
we
were
looking
at
it
like
this,
vodka
was
playing
with
this,
and
it
was
like.
C
I
guess
my
thought
is
to
some
extent
like
a
little
bit
like
who
who
cares
like
unl
as
long
as
you
don't
have
to,
and
everyone
wants
to
like
skip
to
the
end
step
where
they
like
have
dynamically.
They
have
dynamically
loaded,
cid
walls
and
things
with
like
the
what
you
know
and
everything
just
sort
of
like
auto
loads
and
works
for
now.
I
would
just
settle
with
like
not
having
to
write
every
adl
and
codec
in
like
go
and
rust
and
python
and
javascript
and
net
and
wherever
in
order
to
get
self-certifiable
data.
C
C
C
Ways
in
which
webassembly
helps
you
one
is
like.
I
don't
want
to
write
my
code
five
times.
I
want
to
write
sort
of
the
interface
wrapper
once
and
then
the
other
is
I
want
to
have.
You
know,
dynamic,
auto
code,
loading
things
flying
all
over
the
place
and
I
think
they're
both
pretty
cool,
but
I
think
we
can
do
one
at
a
time
and
that's
not
terrible.
B
I
think
I
think
we
do
have
a
we
have.
A
benefit
in
our
data
model
is
relatively
straightforward,
so
it
could
be
that
we
are
waiting
on
an
ecosystem
that
is
trying
to
solve
all
of
all
of
the
things,
and
we
just
want
to
split
the
subset
sold
because
we
sold
all
those
the
rest
of
those
things
on
top
of
our
data
model
kind
of
thing.
So
we
just
want
to
be
able
to
transfer
basic
data
model
stuff.
A
Yeah,
I
would
so
I'm
kind
of
following
what's
happening
a
bit
in
my
free
time
in
the
wasn't
world
since
I
yeah,
we
tried
to
do
it
three
years
ago.
It
didn't
happen,
and
so
I
would
say,
yeah
it's
like
it's
not
anything
inside
soon,
so
I
would
say
yeah.
A
I
would
also
say
the
benefit
is
that
we
have
the
data
model,
so
I
would
probably
just
say:
yeah
you
just
encode
it
as
probably
even
cbor,
because
well
that's
where
we
have
codecs
in
every
language
and
basically
just
use
zebra
as
a
transport
layer
for
like
whatever
you
pass
around
between
warzone
instances,
as
you
likely
in
the
ipod
world,
you
would
likely
depend
on
the
data
model
anyway.
So
yeah,
something
like
this
because
like
this
is
also
what
like.
A
So
I'm,
not
sure
if
it's
still
the
case,
but
basically
back,
then
the
exchange
format
for
rust
through
javascript
was
just
json.
Basically,
so
they
just
basically
encoded.
This
json
made
a
string
out
of
it
passed
over.
The
string
passed
the
json
again,
and
this
is
what
they
do
in
their
code
generation
and
busy.
We
could
do
the
same
thing
with
cbor
until
basically
people
agreed
on,
like
whatever
you
will
do,
and
I
think
the
benefit
also
would
be
that
I
think
like.
A
C
C
I
basically
I
just
I
defined
a
new
thing,
which
was
like
a
simplified
version
of
of
dag
cyborg
or
matched
a
little
bit
like
michael's,
simple,
dag
thing,
except
that
it
it
sort
of
matched
the
data
model
with
all
of
the
with
all
of
the
things
instead
of,
instead
of
trying
to
take
a
more
opinionated
version,
which
is
like
you
know,
all
of
all
of
sort
of
the
thorny
things
like
do.
I
have
sorting
in
my
maps
and
it's
like
well,
I
just
needed
to
like
cleanly
translate
across
the
boundary
right.
C
I
don't
need
to
do
any
of
the
other,
correct
things.
So
that
was
what
that
codec
is
for
everything
there
seems.
Okay,
I
think
with
people
except
for
I
think
I
was
like
I
don't
know
what
floats
are
rod
suggested
just
go
with
ieee
stuff
and
with
all
of
the
gross
nands,
and
it's
fine
call
it.
Somebody.
B
C
So
that
worked
for
codex,
what
became
more
difficult
or
more
interesting
was
like
adls,
because
the
the
objects
get
bigger
right,
like
I
can't
return
to
you
bytes,
I
have
to
return
to
you
like
operators
like
handles
to
operate
on
a
stream
of
bytes
because
they're
just
you
know
it's
a
gigabyte
of
stuff,
I'm
not
going
to
serialize
a
gigabyte
of
stuff
and
send
it
to
you
over
the
webassembly
boundary,
so
that
was
like.
Okay,
we
have
to
like
make
up
our
own.
C
C
Do
they
want
to
really
scope
down
the
api
so
that
the
probability
of
things
that
can
mess
up
is
much
lower?
I
don't
know
there's
like
a
few
different
ways
to
go
at
this.
I
did
notice
that
the
adls,
even
in
go,
which
I
I
mimicked
for
for
a
web
assembly,
were
missing
like
I
couldn't
download
a
map
all
at
once
or
do
any
sort
of
parallelism.
C
This
is
sort
of
like
an
outstanding
thing,
which
is
most
of
why
pld
prime
does
not
allow
parallel
parallel
downloading
in
a
lot
of
places,
and
generally
you
can
sort
of
figure
it
out,
but
in
the
the
adl
functions
really
like
hide
that
from
you.
If
you
don't,
you
can't
get
at
it,
you
can't
get
it,
and
so
there
are
probably
learnings
from
there
around.
Like
oh
yeah,
we
should
add
more,
like
rod
recently
added
a
few
optional
optional
interfaces
to
the
ipld
prime
stuff.
C
There's
probably
a
few
more
optional
interfaces
to
be
added
around
getting
better
map
access
and
list
access
than
we
have
right
now,
because
iterators
are
linear,
and
that
can
be
sad
for
parallelism.
D
Yeah,
I'm
also
interested
in
the
wasm
stuff,
so
I've
been
following
along
since
the
since
the
ipl
ipfs
thing
super
exciting.
I
really
want
to
talk
more
about
folks.
I
think
there's
the
ipvm
meeting
sometime
this
week,
where
we
might
also
talk
about
more
of
this
stuff,
which
could
be
useful
but
outside
of
following,
along
with
all
of
the
new
stuff.
I've
been
working
a
bit
more
on
the
ipld
gateway,
slash
url
spec,
and
the
latest
thing
there
is
we're
reconsidering
how
to
do
parameters
in
the
in
the
path
segments.
D
So
just
to
like
recap,
one
of
the
motivations
with
the
ipld
gateway,
spec,
slash,
url
spec,
was
to
have
signaling
just
be
like
an
obvious
part
of
each
segment
of
the
path
traversal.
D
So
we
could
say
like
at
this
segment
use
this
adl
or
at
this
segment
use
this
schema
or
whatever
else,
and
so
I
had
initially
pulled
some
syntax
out
of
the
ether
based
on
the
query,
params
and
using
square
brackets
inside
a
path
segment
to
kind
of,
like
have
some
query
params
looking
parameters,
however,
on
github
it
was
proposed
by
I
forget,
the
name
of
who
proposed
it
by
software
plumber.
D
I
don't
remember
their
human
name,
but
they
proposed
to
use
this
syntax
called
matrix
parameters
which
has
been
around
for
a
while
in
other
areas
of
dealing
with
urls
and
uris,
particularly
in
the
xml
space,
and
so
it's
pretty
much
the
same
as
query
parameters
in
that
you
have
key
value
pairs
where
you
have
a
key
name
equals
and
then
some
value,
but
rather
than
using
ampersands
they're
separated
by
semicolons,
and
we
can
also
have
that
in
a
path
parameter.
D
So
right
now
I'm
kind
of
reworking
some
of
my
thinking,
because
this
might
be
easier
to
parse
it
disambiguates,
where
the
parameters
go
inside
a
path
segment,
because
before
it
wasn't
obvious
where
those
square
brackets
would
go
relative
to
the
name
of
the
path
segment,
we're
trying
to
trigger
so
like
does
it
go
before
the
name
or
after
the
name
in
my
initial
exploration
report,
I'd
written
about
how
it
might
be
a
problem
and
this
matrix
uris
thing:
it's
nice
in
that
it's
an
existing
standard.
D
We
could
point
to
and
it's
obvious
where
those
parameters
go
and
how
they're
encoded
and
stuff,
so
I'm
actually
working
on
adding
that
to
the
ip
ipip
that
I'm
working
on
for
it
and
I'm
gonna
be
talking
about
how
the
escaping
and
stuff
works
with
url
encoding
and
all
of
that
so
not
a
huge
difference.
But
I
might
talk
to
a
dean
later
about
trying
to
run
that
demo.
D
He
did
and
seeing
if
I
could
do
a
pr
towards
it
to
update
the
parsing
one
thing
I'd
like
to
do
soonish
is
to
add,
I
think,
probably
a
go
library
for
actually
parsing
these
urls.
D
So
that
folks
can
see
how
it
works
and
maybe
use
it,
and
probably
also
javascript
library,
maybe
yeah,
so
that's
kind
of
progressing.
I've
also
been
reaching
out
to
some
other
gateway
authors
outside
of
kubo,
particularly
the
new
iro
thing
in
rust.
That's
been
working
on,
so
I'm
gonna
try
to
poke
and
pester
them
and
see
if
they're
interested
in
adopting
ipl
date,
the
ipld
in
their
core
gateway,
stuff
yeah.
So
that's
the
main
thing.
But
honestly,
a
lot
of
my
time
has
been
spent.
D
I've
been
just
following
up
on
stuff
from
the
ipld
thing
or
ipfs
thing,
so
I've
been
kind
of
busy
with
that
yeah.
So
if
folks
have
comments
about
ipld,
your
urls
or
gateway
ideas
feel
free
to
post
in
this
specs.
Pr
that
I
linked
to
in
the
hackmd
that's
actively
a
work
in
progress
yeah.
E
Yeah,
thank
you.
I've
made
a
few
final
updates
to
amend.
There
were
a
couple
of
questions
that
the
note
raised
that
we
should
probably
discuss
how
you
know
we
can
maybe
this
call
about
packaging
and
where
to
really
fit
that
code
and
other
than
that.
I
also
put
up
a
draft
pr
for
a
potential
way
to
do
map
and
list
access
that
is
less
complicated
than
assemblers
and
builders,
and
that
also
uses
a
men
underneath.
So
you
kind
of
get
modification
transparently.
E
And
mentioned
in
that
pr
about
iterators
and
maybe
concurrent
access-
and
I
mean
you've,
brought
that
up
in
the
past
as
well.
So
I
think
it
is
possible
with
the
model
and
I'm
trying
to
to
get
so.
Maybe
we
can
chat
about
that
as
well.
D
Should
we
talk
about
it
now
regarding
the
mn
thing,
I
think
the
last
little
bit
we
were
talking
about
is
the
actual
file
names
and
like.
Where
should
this
amend
code
live?
Yes,
I
think
he
made
a
good
point
about
how
amend
is
like
its
own
thing
and
it
might
be
good
to
take
it
out
of
traversal
and
again
either
have
it
like
top
level
or
be
inside
node
or
something
I
was
wondering
if
rod
or
a
dean
or
someone
else
had
thoughts
on
where
to
put
put
it.
D
I
just
know
for
sure
it
feels
weird
to
have
all
the
amend
related
things
inside
traversal,
with
just
like,
amend
prefix
to
each
file
name,
just
like
smells
a
lot
like
a
module.
B
I
think
amend
is
born
out
of
the
operations
that
traversal
does,
but
I
think
martin
you've
you've
raised
some
good
points.
Just
reading
your
threat,
the
latest
threat
you
had
on
your
map
and
list
idea
that
thinking
of
amend
as
an
adl.
B
That
is
an
interesting
idea,
because-
and
if
you
do
that,
then
it
certainly
comes
out
as
something
very
different
to
something
that
belongs
in
traversal,
so
yeah
it
it
possibly
is
something
I
don't.
I
don't
know
about
if
it
should
go
under
node,
but
possibly
it's
just
the
kinds
of
stuff
we
have.
The
subdirectories
under
node
are
all
they're
like
implementations
of
nodes.
This
is
sort
of,
but
it's
a
bit
lower
level
than
that.
So
maybe
but
yeah.
E
E
E
So
when
I
integrated
into
progress
or
when
you
get
into
focus,
progress
has
is
part
of
so
of
traversal
allows
you
to
kind
of
track
the
path
and
do
like
budgets
for
the
traversal,
yeah
and
and
that's
kind
of
what
so,
that's
like
that's
required
by
vote
traversal
and
by
if
and
then
is
to
follow
the
pattern
that
focus
does.
E
E
Depart
to
like
too
far
from
where
things
were.
D
E
Like
discussing
with
y'all,
so
I'm
totally
open
to
whatever
options,
I
think,
might
work.
C
I
had
a
I
had
one
more
hornet's
nest.
I
wanted
to
kick
while
we're
here.
So
again,
I
apologize
for
the
lack
of
videos.
I've
I've
been
bugging
my
hardest,
but
yeah.
So,
in
addition
to
sort
of
yeah
kicking
kicking
some
some
awesome
hornets
and
some
data
model
hornets
before
the
event
I
I
also
was
pitching
a
proposal
I
had
last
year
around
how
to
deal
with
some
of
the
security
ramifications
of
some
that
come
with
having
block
limits
that
you
know
or
not
having
block
limits.
C
The
people
seem
to
be
reasonably
on
board
with
like
the
mechanism
for
being
able
to
deal
with,
like
your
standard
merkel.
Damn
guard
based
functions
like
shot
two
and
and
shot
three
shot,
one,
maybe
not,
because
it's
already
a
sketchy
hash
function,
as
is
for
allowing
you
to
incrementally
verify
that
data.
I,
like
the
presentation,
in
my
notes,
the
one
of
the
sort
of
interesting
things
that
falls
out
of
this
is
that
I
think
for
a
while.
C
We've
been
hiding
behind
like
the
security
issues
as
the
reason
why
we
have
these
block
limits
that
prevent
interoperability
with
other
formats.
Like
you
know,
git
or
bittorrent
or
whatever,
that
shows
a
higher
block
number
than
we
did,
but
if
you
remove
the
security
concerns
you
now
have
to
start
talking
about
like
well.
Was
there
anything
more
important?
Was
there
anything
else
that
was
important
here
like
when
I
build
my
ipld
libraries
or
tooling?
C
Are
there
things
where
I'm
concerned
where
I
want
some
number
in
there
or
maybe
I
don't
care
right
like
as
an
example,
you
know
if
I
just
if
I
you
know
when
writing
like
the
wasm
ipld
stuff,
if
I
just
serialize
all
the
data
and
in
a
block
into
you,
know
into
like
the
web
assembly
codec
and
then
ship
it
across
the
boundary.
C
That
seems
fine
if
it's
only
like
a
couple
megs,
but
probably
not
fine,
if
it's
like
100
megs
right
or
similarly,
if
I
have
a
block
and
the
block
is
100
megs
and
even-
and
I
can
refer
to
it
in
some
way.
But
oh
there
are
people
who
want
to
ship
car
files
around
and
the
car
files
must
be
less
than
or
equal
to.
100
mags,
you
know,
and
their
tooling
will
start
to
break.
So
there's
some
questions
around
like
sort
of
now
that
we
can
remove
some
of
the
security
concerns.
C
But
there
may
be
other
considerations
about
how
we
build
our
stack
that
make
us
care
about
these
sorts
of
limits,
but
we,
which
we
haven't
defined
in
other
places,
either
like
what's
the
limit
size
of
a
cid
or
a
multi-hash
or
there's
like
a
lot
of
these.
A
lot
of
these
like
limits
that
people
sort
of
are
casually
assuming,
are
there,
but
might
not
need
to
be.
C
Although
they
do
sort
of
turn
into
each
other
a
little
bit,
because
if
I
start
them,
if
I
start
amending
into
a
block
and
the
block
exceeds
the
size
right
now,
I
could
just
yell
at
you,
but
maybe
I
wouldn't
just
like.
Maybe
the
the
way
in
which
I
yell
at
you
changes
right.
If
we
don't
have
like
strict
like
to
mib
or
go
home
kinds
of
limits,.
C
B
C
I
mean
it's
per
block:
it's
just
a
way
of
transferring
the
blocks
around
right.
So
it's
just
like
saying:
if
your
block
is
over
a
certain
size,
you
have
to
use
a
special
block
transfer
protocol,
and
maybe
you
have
to
store
them
a
little
differently.
If
you
want
the
special
block
transfer
protocol
to
work
nicely
you're
just
sort
of
dealing
with
it
at
the
transport
layer.
But
then
the
question
is,
I
don't
know
if
I
have
a
100
megabyte
object
and
then
I
wanted
to
interpret
it
as
dag
json.
C
Would
that
be
like
fair
game
or
not
fair
game?
And
the
thing
is
this
isn't
like
news
either,
because
if
I
allow
for
like
we'll
call
them
recursive
adls,
I
could
do
something
like
interpret
this
file
as
unix
fs
and
then
interpret
this
file
as
dag
json,
because
I
can
have
an
adl
called
interpret
me
as
dag
json
and
like
should
that
work
or
fail,
and
so
we
sort
of
end
up.
In
the
same.
C
I
don't
think
this
is
so
special.
Maybe
it
is
because
the
codec
layers
may
be
more
special
than
the
adl
layer,
but
it
feels
like
we
end
up
in
the
same
sort
of
place
regardless
or
maybe
we're
leaning
on
things
like,
like
gas
limits
to
figure.
This
figure
out
these
kind
of
like
you're,
using
too
many
resources.
This
is
out
of
control,
kind
of
stuff.
B
Yeah
the
resource
budgeting
resource
budgeting
with
the
the
smaller
block
limits
is
straightforward,
and
but,
as
we've
expanded
with
adls,
primarily
with
ideals,
but
still
these
layering
that
we
can
do
these
resource
problems
are
coming
in
the
way
and
we're
actually
dealing
with
this
now
with
even
with
some
of
the
data
transfer
stuff
with
bind
node
where
we
are
we're
layering
decodes
of
the
nodes-
and
that
means
we've
got
multiple
copies
of
the
nodes
hanging
around
but
they're
stacked,
and
that's
really
awkward,
and
that's
that's
only
that's
only
for
small
messages
too.
B
It's
just
that.
We've
got
a
lot
of
them,
but
yeah
this
resource
problem
does
seem
to
be
something
that
we
have
to
pay
a
lot
more
attention
to,
and
maybe
the
complexity
of
our
layering
makes
that
really
hard.
C
I
mean
to
some
extent
this
is
where,
like
maybe
some
of
the
webassembly
stuff
ends
up
helping,
instead
of
just
being
like
it's,
the
subtraction
layer,
we're
paying
a
cost
for.
So
I
don't
have
to
rewrite
my
code
and
you
know
fortran,
but
it
because
because
they,
you
know
it's
sort
of
easier.
They
have
they're
all
these
people
who
are
trying
to
do
instrumentation
and
other
things
around
the
web
assembly.
Binaries.
A
B
Everybody
I'll
just
add
from
your
the
pr
on
maps
and
lists
and
containers.
The
container
thing
really
interesting.
I'd
be
happy
for
you
to
open
that
up
against
the
iple
prime
repo.
You
can
do
it
as
a
draft
just
for
discussion
get
more
eyes
on
it.
It's
not
there's
not
that
many
of
us
that
are
even
looking
at
the
repo,
so
don't
don't
feel
like
you're
gonna,
it's
gonna
be
noisy,
so
feel
free
to
open
it.
B
There
you
go
and
you
can
frame
it
as
this
is
just
an
idea.
That's
fine,
but
I
I
yeah.
I
find
the
idea
really
interesting,
so
it
might
be
good
to
take
that
further.
Well,.
D
B
A
All
right,
then,
thanks
everybody
for
attending,
and
we
see
us
again
in
two
weeks
so
goodbye
everyone.