►
From YouTube: 🖧 IPLD Bi-Weekly Sync 🙌🏽 2018-12-10
Description
A bi-weekly meeting to sync up on all IPLD related topics. It's open for everyone and recorded. https://github.com/ipfs/team-mgmt/issues/720
A
Okay,
welcome
everyone
to
the
be
weekly
actually
meeting
its
assembled
at
hanging
because
making
well.
We
already
have
a
note-taker,
it's
Eric,
you
moment
you
have
to
do
it
and
then
please
put
your
names
on
the
attendees
list.
If
your
conditioning
and
your
updates,
if
you
have
any
bottom
and
I
guess
as
I,
don't
see
any
particular
agenda
items,
we
just
start
with
like
with
the
updates,
and
then
we
go
from
there
if
we
have
to
definitely
discuss
all
right.
It's
not,
let's
start
with
Eric.
B
B
That's
a
big
topic,
so
I'm
trying
to
do
something.
That's
very
basic
and
very
easy
to
reach
for
that
covers
well-known
topics
like
some
types
like
a
unions,
product
types,
aka
strokes
and
like
the
most
limited
recursive
types
maps
and
arrays
and
come
up
with
a
very
basic,
validate
method
that
can
take
a
schema,
apply
it
to
some
data
and
be
like
yes
or
no.
So
I've
got
some
early
drafts
on
that
and
I've
got
a
hypothetical
DSL
sort
of
syntax
which
I'm
not
going
to
take
too
seriously
anytime
soon.
B
C
A
I
have
a
question
and
I
hope:
I.
Don't
do
you
weigh
the
whole
thing
is
yours
like?
Is
it
also
about
how
we
then
know
which
data
has
which
schema?
Because
even
if
you
have
a
schema,
you
need
to
somehow
connect
the
schema
to
the
actual
data?
So
thus
your
proposal
also
include
this
card,
or
it's
just
about
this
unit.
All
right,
cool,
yeah,.
A
All
right
so
who's
next
on
the
list-
oh
it's
me
so
I
was
just
playing
around
with
the
new
JavaScript
icon,
the
API
and
two
years
old.
They
like
protocol
buttons
to
see
how
it
looks
like
in
the
new
format
and
then
also
now
I'm
working
on
that
choice.
I
call
this
stuff
and
then
I
will
combine
the
both
to
see
if
works,
but
it
looks
good
so
far,
I
think,
and
yet
this
probably
also
what
are
we
doing
next
week?
A
A
D
Yes,
all
right
good
most,
my
fealty
related
work
has
in
20
minutes
more
people
to
use
the
IDS
yeah
in
there
in
ENS
Syrian
name
system,
cuz,
never
liked
going
to
any
Ivy
League
like
data
or
any
merkel
dog
really.
But
we
have
something
like
that.
They
really
want
to
have
some
form
vacation
addressing
or
at
least
something
that
indicates
that
hey
like
this
day
and
it
lives
in
swarm.
I'm
convinced,
I'm,
hatin
like
it
doesn't
have
to
you,
can
address
it
my
content
and
then
include
some
hit
somewhere
but
yeah.
E
Aiden
Vick
to
you,
yeah,
so
yeah
last
week,
was
just
doing
some
playing
around
with
using
IPL
D
to
pass
messages
around
instead
of
using
photo
buffs.
Thank
you
for
everybody.
Helping
me
figure
out
what
I
was
doing
wrong.
That
was
appreciated,
I'm
still
following
the
Limpy
to
clear
up
roadmap
stuff
to
see
where
that
law
is
with
using,
you
know,
are
having
an
RPC
library
or
some
sort
of
message
passing
anything
that
makes
it
easy
to
move
messages
around.
F
E
With
Java
JavaScript
and
go
and
I'm
winning,
if
anybody
else
has
more
comment
on
my
the
I
guess:
interface
for
replicating
a
virgin
graph,
that'd
be
useful,
but
I
will
probably
loop
back
to
doing
implementation
stuff
on
that
at
some
point.
But
now
that
I
have
a
basic
one
working
and
focusing
on
other
things
to
know.
A
A
C
C
One
kind
of
big
thing
to
talk
about,
though,
is
that
we
we
are
engaging
with
someone
to
take
on
the
work
of
I,
doing
IETF
spectra
compliance
and
all
the
process
work
to
get
some
stuff
through
the
ITF
for
multi
formats
and
we're
actually
going
to
do
multi,
hash
and
CID
together.
So
there
will
be
like
a
visually
they're
like
CID
will
be
the
spec
that
we're
pushing,
and
then
there
will
also
be
a
spectrum
multi
hash
that
we
are
depending
on
and
sort
of
the
process
will
be
like.
Oh.
B
C
Dependency
here,
that's
why
these
things
are
going
to
push
through.
We've
had
a
lot
of
like
good
conversations
already
with
who's
gonna
drive
this.
It's
done
like
a
lot,
so
one
who's
done
like
a
lot
of
ITF
spec
work
in
the
past,
including
like
HTTP
two.
So
another
really
interesting
thing
here
is
that
so
the
codec
registry
stuff
that
we
have
currently
I
thought
that
it
was
going
to
be
like
a
huge
amount
of
work
to
figure
out
like
what
does
governance
look
like
and
blah
blah
blah.
C
It
turns
out
that
if
you
have
a
registry
requirement
in
an
ICF
spec,
you
just
say:
hey
I
need
an
INR
registry
for
this,
and
you
just
put
that
in
the
spec
and
then
win
the
spec.
It's
finalized.
You
automatically
get
an
I
in
a
registry
and
with
as
part
of
that
process,
you
basically
assign
experts
to
it
and
the
there
is
an
expert
review
for
adding
anything
to
the
registry.
So
there
actually
isn't
a
ton
of
work
there
and
we
will
eventually
get
an
I
on
a
registry.
C
But
that's
something
to
think
about
like
it
will
not
just
be
pull
requests
to
a
CSV
file
at
some
point
in
time
where
these
Kotick
registries
will
actually
have
some
kind
of
governance
around
them
and
then
we'll
just
be
sending
pull
requests
to
update
the
CSV
after
they
land
in
the
INR
is
cool,
so
yeah.
The
the
big
change
that
I
want
to
talk
about.
C
Now
that
we're,
through
all
the
updates,
I
just
caught
I
just
put
this
in
a
comment,
but
if
something
I
was
thinking
about
all
weekend,
based
on
some
of
the
comments
that
Stephen
had
a
little
while
back
about
how
we
might
compress
the
representation
of
files.
So
if
you
have
a
really
small
file,
just
in
letting
the
binary
objects,
rather
than
always
linking
to
them.
That
got
me
thinking
about
like
in
some
of
the
other
interfaces
that
we're
doing
around
IP
LD.
C
We
are
trying
to
move
people
towards
just
looking
at
paths
rather
than
trying
to
think
about
things
at
a
block
level.
One
thing
that
we
could
do
is
actually
change
the
university
to
spec
right
now.
It
basically
thinks
about
things
as
a
block
is
an
object,
and
here
what
the
patterns
are
in
each
individual
block
an
object.
C
For
instance,
you
can
actually,
if
you
have
a
really
small
directory
of
small
files,
you
could
compact
the
entire
thing
into
one
block
rather
than
doing
multiple
blocks
there,
like
a
huge
amount
of
trade-offs
to
think
about
here,
and
we
would
also
have
to
carve
out
an
exception
for
HAMP
and
just
say
like
also
any
of
these
names
base
camps
or
whatever
but
yeah.
It
would
be
a
sizable
change
but
may
end
up
opening
up
a
lot
of
performance.
Optimizations
and
people
could
do
in
the
future,
so
Stevens
hand
go
up.
D
By
point
with
it,
we
can
be
link
agnostic,
where
we
just
say
what
comes
in
transparent,
you
just
reverse
them.
You
have
a
big.
Like
effect,
we
have
one
big
object
and
it
may
be
broken
syntax,
but
the
other
point
I
had
was
like
I
said
the
other
question.
I
wasn't
something
about
strings
and
not
being
able
to
attach
them.
What
was
that.
C
D
C
D
A
I,
just
I
think
it
can
be
a
sting
but
I
think
the
canonical
seaboard,
the
lowest
ring,
and
currently
we
choir
iPod
EB
can
not
equal
Seaborn,
but
that's
right.
I've
never
seen
that
required,
but
I
can
check
I
think
but
but
I
you
get.
But
anyway,
that's
like
like
at
least
see
Bo
definitely
supports
having
experience
as
a
top,
but.
D
I
want
to
point
out.
This
does
require
respect.
We
really.
If
we
do
this,
we
have
to
move
from
supporting
just
see
IDs
to
full
links.
You
know,
I
feel
the
objects
we're
not
always
for
the
objects
will
support
this
team
or
at
least
need
to
because,
like
otherwise
I
don't
have
some
object
and
like
need
to
like
include
one
file
that
object
you
in
a
different
directory,
and
it
has
like
pull
that
file
with
a
subdirectory
out
of
the
object
services.
Well,.
C
C
D
D
It's,
although
it's
backwards
compatible,
we
can
use
like
for
kind
of
force,
compatible,
I,
guess
or
like.
If
objects,
don't
have
these
extra
length
you
won't
be
able
interpret,
they
have
e-excellent
you'll
be
able
to
put
them,
but
we
can
easily.
We
can
literally
just
be
catenate
a
linked
with
the
CID
and
it
just
work
there
probably
better
way
to
do
this,
but
we're
pork
is
annoyed.
B
D
Ms
and
I
had
my
file
and
it's
in
a
directory,
and
then
I
say
like
copy
this
file
over
here,
like
the
m
FS
would
like
you
would
I
would
be
simplify
the
link
to
like
the
shortest
link
you
could
get,
which
would
be
like
you
whenever
flashed
some
very
short
paths,
and
then
it
would
like
put
that
in
as
the
link
to
the
file
in
like
the
other
directory.
So.
C
I
think
that
we
should
probably
create
a
thread
in
the
specs
repo
about
like
a
link
spec
or
like
a
link
B
to
spec
or
something
and
just
say
that
CID
is
alone
or
link
v1,
where
we
start
to
specify
out
what
the
padding
structure
is
and
then
in
the
university
to
spec.
We
can
just
say
like
this
is
a
Lincoln
at
some
points
they
also
link
me
to,
and
then
you
can
point
in
to
stuff
yeah
I
mean
the
question
for
right
now
is
like
and
I.
Think.
C
C
So
to
me,
the
more
important
thing
is
that
the
implementations
for
reading
support
this,
because
in
the
future
this
would
allow
parsers
to
make
any
kind
of
crazy
trade-off
that
they
want,
like
a
really
good
example
of
this
right
is
that
if
you
have
a
website,
you
know
that
they
have
to
pull
the
index.html
right
away
and
then
the
rest
of
the
resources
you
could
inline
all
of
it.
Next,
at
HTML
into
the
first
object
in
the
route
and
all
the
rest
can
still
be
in
a
big
tree
going
off
and
getting
like
yeah.
C
D
D
So
ideally
we
put
that
in
the
directory
itself,
but
if
we
put
it
in
the
director
itself,
then
like
it's
harder,
just
link
to
the
file
but
like
if
we
stopped
caring
about
like
these
links,
boundaries,
and
we
can
just
like
link
to
like
the
it's,
the
correct
location
that
directory,
so
it's
the
correct
directory
entry
and
have
that
beaten,
umbilicus
the
beginning
of
the
file
at
this
end
points
to
the
final
data.
This
allows
us
to
like
make
you
dock
musicians,
but.
B
G
B
G
D
D
C
Yeah
but
I
mean
frankly,
I
really
like
a
moment
the
moment
that
we
say
that
it
you
can
represent
this
in
two
ways:
we
lose
a
hundred
percent
reproducibility
of
the
hashes
right,
like
even
even
just
allowing
HAMP's
or
objects
in
there.
Right
now
means
that
you,
you
are
never
guaranteed
that
the
exact
same
hat
like
the
two
hashes,
don't
actually
represent
the
same
file
structure.
Reg.
D
B
C
All
of
these
require
us
to
just
like
not
always
have
a
hundred
percent
deduplication
like
you
only
get
deduplication
when
people
are
creating
the
tree
roughly
the
same
way
or
you
know
you're
getting
it
only
at
the
at
the
binaries
like
block
level
and
the
like,
if
you're
using
the
same
chunker,
but
you
don't
get
it
for
all
the
intermediary
nodes
like
there.
There's
always
going
to
be
this
trade-off
and
I
think
that
we're
gonna
have
to
make
it
once
we
like
look
start
looking
into
like
sorted
collections
and
stuff
anyway.
C
So
it's
not
as
big
a
deal
to
me
that
were
they
were
making
it
now
and
and
I
mean,
and
the
reality
is
like.
If
you
run
ipfs
like
IVFs,
it's
probably
going
to
make
the
trade-off
stuff
like
it
wants
to
be
as
deep
deduplicate
as
much
as
possible,
and
so,
if
you
always
run
it
with
no
options,
you're
always
going
to
get
the
same
thing
and
they're
always
going
to
be
hash
compared.
It's
just
opening
up
the
opportunity
to
do
it
differently.
Also,
remember.
D
The
data
size
is
mostly
like
large
blocks
and
that
will
always
be
in
separate
objects.
Actually
that's
one
of
the
points
of
this
idea.
So
the
point
is
like
one
of
the
pointers
like
we
can
embed,
like
all
the
mitad
in
all
like
the
stuff
that
differs
and
like
wouldn't
really
do
the
same
between
different
files
in
like
a
directory.
That's
maybe
different
anyways
and
then,
like
the
big
pieces
of
data,
are
just
blocks
at
the
end
of
it.
It
doesn't.
D
You
can
have
separate
hashes,
which
is
not
good,
but
I
I,
don't
think,
there's
anything
we
can
do
about
that.
Anyways.
If
you
have
a
filesystem,
we're
like
people
are
gonna,
be
adding
metadata
are
changing
things
really.
You
want
to
make
it
as
reproduced
as
possible,
and
you
also
won't
have
options
so
obvious.
Like
say
like
make
like
do
a
reproducible
minimal.
C
Like
if
you
take
like
the
the
index,
dot
HTML,
one
right
where
we're
optimizing
by
in
lining
all
of
that,
like
nobody's
gonna,
have
the
exact
same
website
to
me,
or
at
least
I
assume
they
don't
have
the
exact
same.
Each
team,
l
with
me
and
so
like
I'm,
not
as
worried
about
the
duplication
as
that
as
I
am
of
all
of
the
sub
resources
like
all
of
the
dependencies
and
all
the
the
images
and
fonts
and
all
the
other
that
I
linked
to
in
the
site
like
that.
B
Mean
at
this
point,
I
would
also
be
willing
to
rest
my
case
by
saying
yes,
I,
agree
with
all
of
the
above,
but
I,
don't
think
all
efficiently
enumerated
all
over
the
parameter
spaces.
We
have
ready
creep
usability
and
it
makes
me
super
nervous
to
talk
about
doing
more
of
them
before
even
having
made
that
about
how
much.
D
Well,
well,
we
do
care
about
having
greatest
reproducibility.
This
is
best
results
where
the
whole
idea
of
like
keeping
track
of
all
the
parameters
came
in
for
like
if
you
get
somewhere,
then
it
becomes
like
Norfolk
reproducible,
but
in
my
ideal
world
you
don't
have
to
care
about
any
of
those,
because
everything
just
goes
back
to
anyways.
When
you
get
a
file,
you
already
get
it
from
my
PMS.
So
there's
no
like,
like
most
people
reporting
data,
that's
not
so
yeah.
C
I
also
think
that,
like
the
reproducibility
side
like
that
is
an
IPSS
concern
for
the
trees
that
it
creates,
and
we
have
a
lot
more
control
there
to
change
what
we're
using
in
the
spec
like.
We
can
always
create
these.
The
exact
same
way
always
use
hams
for
directories
like
just
make
that
trade-off
and
then
they'll
always
be
reproducible.
We
can
limit
the
vectors
by
which
we're
like
introducing
on
reproducibility
in
ipfs,
without
containing
what
the
specs
can
theoretically
allow,
or
what
I
CFS
could
theoretically
parts
right.
E
E
About
the
like
dirt,
the
directories
and
metadata
like
what
what
were
you
envisioning,
it
seemed
like
you
were
envisioning
like
sometimes
people
want
to
cover.
Sometimes
I
want
to
send
you
a
folder
and
I.
Want
you
to
take
the
metadata
of
the
files
with
you
and,
like
sometimes
I.
Don't
is
that
is
that
what
you're
trying
to
allow
for.
C
G
D
D
Okay,
so
the
the
idea
is
basically,
we
need
to
be
able
to
support
duplicating
the
data
blocks,
even
if
people
have
different
metadata.
So
this
is
this
is
where
the
large
sort
of
motivation
of
the
questions
we've
netted
in
director
comes
from.
The
second
half
of
motivation
is
performance,
where
we
want
two
people,
except
someone
metadata
to
be
able
to
list
the
directory
and
like
with
the
metadata.
Let's
do
one
be
able
to
that
without
fetching
all
the
actual
data
files.
These
are
really
bit
a
to
motivation
to
come
from
there.
D
The
other
way
of
doing
this
is
you
can
actually
you
can
get
the
duplication
which
is
separating
off
the
manage
and
different
block.
We
have
like
directory
listing
metadata
metadata
minute
of
metadata
and
then
data
box,
but
that's
that's
all
I
have
informatician.
Have
these
tiny
little
metadata
blocks,
we
have
accession
them
and
then
fetch
them
in
the
block.
The
nice
thing
about,
like
start
saying,
okay,
forget
the
blocks.
We
just
have
a
crash
or
really
we
have
like.
We
have
a
dag.
D
D
It's
not
like
we've
actually
talked
about
doing
like
a
step
further
in
the
past,
we're
like
we
actually
build
in
a
sharding
concept
to
I
still
get,
though
for
like
it,
especially
charted
charted
back
charted,
array
and
stuff
like
that.
We
actually
sided
to
move
or
move
away
from
that,
because
it
was
too
and
like
we're
building
too
many
things
in
tactically
and
now
we're
planning
to
have
like
a
small
layer.
On
top
of
that,
with
this
I
can
be
sure
said
that
UNICEF
is
going
to
use
to
do
that
kind
of
stuff
budge
pretty.
E
I
cut
out
for
for
some
of
that,
because
my
internet
was
misbehaving
but
I'm
wondering
if,
like
if
you're,
you
can
always
send
the
metadata
separately
in
a
sense
right
like
you,
can
have
like
a
trace
in
the
same
way
that
if,
if
you
you
know
you
are
constrained
to
not
be
allowed,
you
can
only
put
primitives
into
a
map
right.
Well,
you
can
still
create.
You
can
still
store
a
struct
in
a
map
by
just
having
like
a
sequence
of
maps
that
each
contain
one
of
the
fields
in
the
stop.
Here.
D
You
could
do
that
that
gets
like
tricky
to
link
tubes.
Now,
if
you
want
to
like
someone
wants,
includes
like
a
sub
part
of
the
graph.
Now
you
need
to
create
this
new
join
node
that
make
sense.
Really
you
want
a
conclusive
file
or
some
some
trees.
Now
you
need
to
like
pull
out
the
sub
tree
from
pole
graphs
and
then
create
a
join
node
to
do
that.
Yeah.
E
I
mean
it's
definitely
like
it's.
It's
very
use,
kiss
dependent
in
terms
of
I
want
this
thing
to
be
content-addressable
with,
and
without
this
particular
chunk
of
data
added
to
it
right,
you
can't
sort
of
have-your-cake-
and-eat-it-too
and
have
it,
but
you
can
except
I,
feel
like
if,
if
I
understood
better
what
the
use
cases
were
of
like
when
we
want
those
and
when
we
didn't
right,
then
we'd
know
how
much
we
cared
well.
D
This
is
also
thinking
about
like
pulling
or
separating
the
data
blocks
from
the
rest
of
the
tree.
There's
it
at
least
we
can
do
to
placate
the
data
blocks.
This
does
cause
a
problem
for
content
routing
where
it's
much
like.
If
I
had
the
entire
tree,
it's
much
easier
to
have
like
everyone
like
tell
or
not
the
work
they
have
in
territory,
then
it
that
caught
it
round
just
works
beautifully.