►
From YouTube: 🖧 IPLD weekly Sync 🙌🏽 2019-08-19
Description
A weekly 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
And
that's
every
week
we
just
report
what
we've
been
up
to
and
then
discuss
any
open
issues.
If
you
want
to
talk
about
yeah
and
I
start
right
away
with
stuff
I've
been
working
on
and
I
start
with
great
news,
because
there
was
this
issue
that
in
the
Seaboard
world,
you
can
register
for
certain
tags
and
someone
register
for
take
42,
and
it
turns
out
that
they
are
really
nice
people
and
they
decided
that
we
can
have
this
tag
and
they
just
move
to
some
other
other
tag.
A
So
the
CDs
in
seaboard
will
be
take
40
to
it,
which
we
currently
use.
We
will
get
the
official
stamp
on
it
and
it's
especially
when
I
spent
time
on
and
now
I've
also
first
draft,
because
they
need
to
have
a
spec
and
man.
This
is
what
I'm
working
on
and
hopefully
I.
So
my
plans
reacted
to
get
it
finished
this
week,
because
I
will
also
be
in
a
conference
for
the
next
week.
So
for
10
days,
I
won't
be
and
we'll
get
conference,
so
I
hope
hopefully
posted
it
this
week.
A
So
then
we
are
good
to
go
and
yeah
have
this
tag
officially
registered
at
E
I.
You
know
other
than
that.
I
yeah
did
a
bit
of
I,
give
a
stuff
to
get
their
problem
with
the
single
with
endeavor
I
thought.
So
what
I
basically
did
is
that
I
did
this
work
and
I
felt
I
needed
it
for
the
iPhone
the
world,
but
it
turns
out
like
I,
can't
recall
why
I
thought
I
would
need
it,
but
anyway
I
did
the
work
and
works.
So
they
are
happy
yeah.
B
So
I
had
a
long
dark,
teatime
of
the
soul,
a
very
thoughtful,
more
contemplative
than
I
wanted,
even
time
trying
to
figure
out
what
we
should
do
with
map
ordering
I.
Think
that's
an
issue.
We've
we
know
is
going
to
be
an
issue
and
we've
kind
of
tried
to
dodge
it
for
the
most
part
so
far
fairly
successfully,
so
like
good
and
I
was
trying
to
work
on
decoding
logic
that
will
work
well
and
be
really
efficient
in
the
code
generation
stuff
that
I'm
working
on
around
structs
that
schemas
and
I.
B
B
B
We
have
to
do
something
because
we
have
iterators
in
the
data
model.
They
have
to
have
a
behavior
wherever
it
is.
Some
codecs
have
ordering
opinions
if
the
data
model,
like
the
storage
that
we
have
in
memory,
has
different
ordering
opinions
than
like
just
naturally,
some
cost
will
emerge
somewhere.
You
have
to
do
a
snort
or
rearrangement
to
some
other
predefined
order.
B
There's
just
no
obvious
slam-dunk,
which
reconciles
all
of
these
things
so
I'm,
still
kind
of
trying
to
churn
through
that
and
figure
out
which
things
are
even
interface,
design,
issues
in
the
code
versus
spec
issues
and
where
the
spec
issues
should
land.
So
I
think
this
just
come
out
in
one
big
exploration
document
first
and
then
we'll
like
we
have
to
go
through
a
couple
of
rounds
of
figuring
out
which
parts
of
it
even
go
where
what
idea
that
I've
had
so
far.
That
seems
to
simplify
a
lot
of
things.
B
It's
just
making
sure
that
ordering
opinions
stay
mostly
bound
to
the
codec
layer
like
when
we're
using
deck
C.
What
we
care
about,
having
sorted
order
for
caching
that
is
consistent
with
deck
c1,
but
if
you're
in
another
codec
here
I
are
gonna,
have
a
different
a
so
if
you
have
a
different
canonical
order,
that's
fine
and
that
answers
some
questions,
but
not
even
remotely
all
of
them.
So
so.
C
D
C
Like
like,
for
instance,
like
I,
you
know
the
Jason
spec
now
says
that
these
need
to
be
order
in
utf-8,
like
when
you
encode
them,
because
that
isn't
defined
an
adjacent
spec.
So,
of
course,
does
that
ordering,
if
you
put
them
into
a
Python
object
like
every
deserialize,
it
will
do.
They
are
now
in
a
new
random
order,
because
internally
inside
of
that
structure
in
Python,
it's
using
some
kind
of
set
and
there's
some
kind
of
under
than
behind,
like
doering
yep
yeah.
So
you.
B
Made
those
notes
in
my
document
in
progress
yeah,
but
that's
yeah,
that's
a
very
real
problem
for
several
languages.
Some
languages
make
sort
of
maps
and
reasonably
internal
implementation,
efficient
ways,
and
some
of
us
are
just
like
kind
of
out
of
luck
on
that
right,
like
I,
think
Python
might
be
one
of
the
kind
of
out
of
luck.
One
is
to
to
my
great
chagrin.
C
Well,
in
a
in
JavaScript,
there's
a
preserved
ordering,
but
it's
based
on
the
insertion
order.
So
this
has
the
benefit
that
like,
if
we
have
an
ordering
that
we
encode
it
will
end
up
in
the
deserialized
structure.
So
that's
good,
but
we
also
need
to
when
we're
encoding
aria
sort
them
if
they
were
inserted.
Out
of
you
know,
utf-8
character
ordering
so
they're
still
like
work.
Do
we
have
again
yeah.
A
But
but
isn't
ordered
Maps
something
like
it's
just
Ruby
their
maps
and
shouldn't
be
there
ordered
maps,
which
is
something
different.
Like
I
mean,
like
you
guys
like
but
baby,
do
you
still
need
to
have
an
iterator
over
maps?
Certainly,
but
so
yeah
I,
like
the
order
does
matter
so
it's
more
like
like
so
for
somebody
will
do
filter
something
maybe
ordering
doesn't
matter
so
yeah.
C
So
I
think,
like
honestly
with
the
multi
block
collections.
This
is
actually
easier
because
we
have
to
go
and
write
the
iterator
for
each
of
these.
So
we
can
specify
the
behavior
pretty
little
bit
I
named.
The
problematic
part
is:
is
it
the
codec
layer
where
we're
leveraging
a
lot
of
existing
language
level
assumptions
to
like
make
their
way
into
these
encoders,
and
then
it
could
potentially
give
us
back
the
same
semantic
data
with
it,
a
perhaps
even
with
the
same
code?
Id.
That's
what
we
need
to
worry
about
at
the
most
yeah.
B
Yeah,
there's
lots
of
things
that
are
scary
in
that
area
and
just
really
an
interface
design.
What
we
present
to
the
other
humans
using
these
libraries
so
I'm.
Looking
at
this
as
I've
got
this
node
interface
right,
where
I'm.
If
I,
have
a
single
tree
of
data,
that's
in
one
block
a
net,
so
no
advanced
layout
stuff
in
it.
Then
then,
that's
just
going
to
serialize
this
map
and
so
like.
B
Maybe
if
we
say
okay,
everything
has
this
canonical
order,
then
that
one
will
be
some
sorted
order,
a
terrain,
your
but
then,
if
I
have
another
thing,
that's
implemented
by
hampt
and
I'm,
exposing
that
as
a
node
and
I
ask
for
the
same
iterator
method,
then
that
one's
going
to
give
me
ham
decoration
order,
so
how
on
earth
do
we
write
our
documentation
of
this
to
not
surprise
user,
that
when
they're
using
these
different
kinds
of
things
underneath
they
get
consistent
but
consistently
different
orders?
It's
just
yeah
it's
tricky,
but.
A
But
doesn't
like
wouldn't
be
the
be
order
different
like
to
these
three
or
depends
on
a
codec
basically,
but
is
it
a
problem
like
like
like
in
like,
especially
for
JSON,
for
example?
There
is
no
no
order
specified
for
a
reason,
probably
because,
like
that's
the
little
bit
problem
says
so,
what's
the
format,
if
we
not
specify
it,
and
if
you
want
to
have
ordered
maps,
you
use
an
order
map
and
not
a
normal
bit.
So
so,
why
would
you
need
order
in
a
map
because.
C
A
C
C
B
Agree
with
that
would
be
good.
That
would
be
super
correct,
but
it
also
has
serious
costs,
because,
if
you're
parsing
in
the
canonical
order
of
one
codec
and
like
let's
say
it,
uses
some
I,
don't
know
it
has
some
sorted
order.
But
it's
one.
That's
just
like
this
codec
we
made
a
weird-ass
choice.
No.
C
B
So
then,
it
stores
everything.
Let's
say
the
data
model
has
and
no
opinions
the
data
model.
We
implement
everything
to
keep
exactly
whatever
order
was
fed,
so
it
keeps
that
one
and
then
we
serialize
it
out
in
another
codec
that
has
a
different
sorting
order
shoot.
Now
you
have
to
sort
you
have
to
allocate
all
the
additional
memory
needed
to
do
the
sort
on
the
side,
yeah.
A
B
B
B
B
That
may
well
be
part
of
the
answer,
but
it's
also
pretty
rough
I,
don't
know,
there's
more
than
one
concern
active
here
to
it.
I
think
we
were
like
a
document
which
just
enumerate,
which
the
concerns
are
is
important
because
we've
hammered
the
heck
out
of
one
concern
right
now,
but
one
of
the
things
that
originally
got
me
here
is
a
very
different
concern,
which
was
just
playing
parser
efficiency.
B
So
the
only
thing
it
needs
is
an
integer
word
to
say
how
many
fields
have
I
launched
already
and
if
I
have
to
accept
things
in
random
order,
it
gets
more
costly,
no
there's
just
a
whole
bunch
of
things
like
this,
which
show
up
in
kind
of
a
fractal
of
places
where
it's
not
immediately
expected.
So
yeah.
A
So
basically
it
busy
it
happens
and
we
don't
have
a
choice
so
basically
yeah,
let's
yeah,
so
it's
we
can
work
on
it
and
the
only
solution,
I
guess
is,
if
you
really
have
ordered
sets,
which
is
sew
em
in
from
the
pivots,
because,
like
we
take
the
D
issue,
also
with
a
couchdb
touch,
DB
back
in
the
days
because,
like
Erlang
doesn't
really
have
maps.
So
what
you
do
is
you
use
an
array
of
objects?
A
Then
that
question
is
well
then
suddenly
their
map
has
an
order
like
so
in
Erlang.
It
has
a
fixed
order,
but
in
chess
it
doesn't
in
any
and
all
those
weird
things
it
yeah
and
therefore
yeah
anyway
yeah
but
I
mean
I
I'm,
looking
forward
to
see
us
expression
of
human
and
then
you
can
go
from
there.
Yeah.
B
A
A
By
the
way,
but
the
but
the
order
changes
in
the
new
version
of
seaboard,
just
listen
for
information
so
is,
and
can
Tony
calls
sea
bordering
is
the
old
version,
so
I
just
basically
the
on
this.
Yet
so
it's
the
the
person
that
made
sure
we
get
the
tag
pointed
out
to
me
that
there's
a
new
version
or
github
I
can
put
it
into
them.
Notes
which
is
clear
about
the
ordering,
and
they
do
it
a
bit
differently,
just
yeah.
So
keep
this
in
mind
am
its
the
upcoming
work.
B
A
I
can
put
it
to
it.
Yeah
and
I
want
to
say
one
last
thing
yeah,
so
basically,
but
what
we
need
to
we
need
for
our
tourists.
We
need
some
ordering
for
all
colleagues
to
make
sure
that
the
hashes
are
always
the
same.
So
we
sort
of
need
to
have
some
alright
yeah,
okay,
alright
yeah
I
will
put
it
in
the
notes.
D
D
It
was
I,
I
put
it
down
as
a
learning
experience,
I
learned
a
lot
about
blockchain
blocks
and
I
count
with
a
new
binary
decoding
technique
that
I
think
could
be
useful
elsewhere,
but
yeah
I'll
tighten
it
up
and
figure
out
how
to
propose
that
get
it
integrated
with
the
about
I
peeled
easy
cash
thing.
I
did
a
little
bit
more
habit
work
coming
up
with
this,
because
the
spec
of
the
HAMP
distant
is
not
in
it's
not
in
in
perfect
sync,
with
the
implementation
I'm
having
some
trouble
with
the
testing
done.
D
C
C
I
really
need
to
rewrite
the
roadmap
in
the
okrs,
based
on
the
last
meeting
that
I
had
and
do
some
other
follow-ups
from
the
PAC
meeting
and
send
out
an
override
update
plenty
of
doing
that
last
week,
but
I
got
caught
up
with
some
metric
stuff
and
some
and
then
I
was
out
sick
for
a
few
days,
so
hopefully
that
I'll
get
done
within
the
next
day
or
two.
We
should
be
pleased.
C
One
cool
thing,
though,
about
the
metrics
was
that
I
was
doing
that.
You
know
you
want
to
check
out
so
github
actions
are
pretty
amazing
and
you
can
do
roughly
anything
with
them.
So
you
can
kick
off
actions
win.
New
issues
come
in
new
pull
requests
and
then
you
can
just
run
arbitrary
code
in
the
docker
file-like
in
any
language,
and
it
also
gets
minted
a
new
github
token.
C
C
A
Cool
things
and
I
forget
to
add
that,
on
my
so
also
a
little
dings
on
the
craft
for
the
spec
for
the
Seaboard
take
everyone
to
get
your
feedback
as
soon
as
possible
to
really
get
get
a
push
through
this
week.
So
yeah
yeah
five
past
iterations,
please
yeah,
okay
other
than
that
yeah,
so
I
yeah
I
will
be
out
next
week.
B
One
thing
I
just
saw:
we
should
may
be
able
look
at
the
IRC
and
IP
LD.
Someone
was
thinking
out
loud
about
some
protobuf
stuff
and
I.
Don't
know
if
I
have
instantaneous
good
answers
for
that.
I
think
we
may
want
to
I
have
remembered
that.
We
also
had
some
discussions
that
there
aren't
many
protobuf
users
and
we
were
seriously
considering
doing
some
retconning
on
what
some
of
the
codec
numberings
meaning
around
that
basically
be
like.
B
This
is
Unix
FS,
v1
style
parsing,
and
if
we're
gonna
do
those
things
we
should
I,
don't
want
to
go
full
depth
on
protobuf
codecs
right
now,
because
I
just
don't
have
the
time
and
attention
to
do
well.
But
maybe
we
should
outline
a
note
about
how,
like
the
old,
AG
protobuf
has
like
these
particular
weird
opinions
about
links
and
stuff
which
are
no
longer
in
sync,
with
the
IPA
OT
day
tomorrow.
By
the
way,
that's
a
thing,
and
it's
over
here
like
at
least
make
a
placeholder
for
a
new
protobufs.
C
Because
it
would
be
maybe
so
this
person-
that's
in
this
thing,
so
they're
saying
that
they're
they're
using
their
own
portable
thing,
not
Doug
TV.
So
maybe.