►
A
Okay,
welcome
everyone,
hello
again:
it's
the
ipfs
core
implementations,
weekly
sync,
it's
November,
the
25th,
we're
running
out
of
2019
time
and
now
it's
time
to
talk
about
wait.
We've
got
note
taker
its
Alex
today,
because
Jake
is
not
here
and
yeah.
So
if
you're
here
for
your
name
on
a
copepod
which
I
think
you've
done
already
under
the
attendees
list,
and
we
will
go
through
our
initiatives
in
a
moment
author
of
explain
the
bid
at
the
bottom,
where
you
should
put
your
weekly
update
for
async
review
and
we
won't
go
through
that.
A
But
today,
but
people
can
look
at
it
and
asynchronously
when
at
their
leisure
cool.
Let's
do
it,
then
initiatives.
So
what
what
I
took
the
liberty
of
doing
this
week
was
removing
the
updated
release
process
initiative,
because
we've
largely
kind
of
done
that
and
been
and
we're
now,
following
the
new
release
process
that
we
PR
to
both
go
up
to
passenger
side.
A
Bfs
and
I
replaced
it
with
a
section
called
upcoming,
slashed
shipped
releases,
and
so
this
is
a
place
where
you
can
put
things
that
are
going
to
be
shipped
soon
or
have
just
shipped
and
people
need
to
know
about
it.
So
I've
got
a
couple
of
things
down
here
already,
so
news
is
that
jsrf
s,
HTTP
client
version
40
has
been
released
under
the
hood.
It
is
all
fetch
and
async/await
now
so
hooray
for
that
we
still
have
cool
streams
and
nodejs
streams
in
there
and
we
still
retain
the
callback
api.
A
The
next
step
for
for
that
for
the
HTTP
client
is
to
then
is
to
now
make
some
changes
to
remove
some
of
the
streaming
API
s
that
we
have
in
favour
of
just
plain
old,
async
iterables,
which
would
be
super
cool
and
remove
callback
interfaces
and
remove
well.
This
is
a
thing
right:
I've
opened
a
PR
and
it
it
actually
reduces
the
size
of
the
ipfs
HTTP
client
bundle
by
over
half,
so
it
was
like
240
K
and
now
it's
less
than
196
I
think
so
that
is
really
cool.
A
But
honestly,
it's
mostly
because
I
removed
peer,
ID
and
period
flow
from
the
bundle,
and
what
is
happening
is
that
the
p2p
crypto
is
being
pulled
in
because
of
including
those
those
dependencies
and
Limpy.
2P
crypto
is
huge
and
I
believe
not
very
much
used
when
you
get
a
peer
ID
from
an
HTTP,
HTTP
client.
Sorry,
people
are
messaging
me,
that's
always
happens
right,
so
I've
got
a
proposal
for
removing
those
and
instead
replacing
them
with
just
an
object
which
describes
the
peer
info
and
a
CID
instead
of
a
peer
I.
A
B
A
I
kind
of
paused
on
that
a
little
bit
while
I
was
fixing
up
the
chair.
So
if
s
httpclient,
because
I
was
just
a
lot
of
work
and
I
wanted
to
get
it
all
out
the
way,
and
so
that
is
a
little
bit
delayed.
But
we
are
currently
on
stage
2,
which
is
community
dev
testing,
and
in
a
few
days
we
will
transition
to
stage
free
unless
people
have
big
issues.
A
I've
actually
already
found
out
a
number
of
issues
which
I
have
sent
PRS
to
fix
already,
so
that's
good,
so
that
should
hopefully
ship
by
at
the
end
of
the
month,
beginning
of
December
yeah.
So
you
can
look
at
that
and
that's
all
that
I
have
to
say
about
upcoming
and
shipped
things
releases
in
coin
from
patience.
Does
anyone
else
have
anything
to
share
that
spin?
Ship
tour
is
coming
up
through
it
or
have
any
questions
all
right,
I'm,
moving
on
moving.
C
Yeah
we
announced
tit,
but
one
owns
it
just
one
more
time
where
we
have
new
bootstrappers
and
we
are
starting
the
process
of
retiring,
the
old
bootstrappers,
so
I
through
kind
of
the
initial
PR
to
JSF
ipfs.
There's
some
notes
and
I
can
tweak
that
and
make
it
a
little
more
useful
feel
free
to
adopt
that
if
you'd
like
but
we'll
be
doing
the
same
for
IPSS
and
then
making
the
world
better.
A
C
A
B
A
Nice,
that
means
that
people
who
are
storing
large
amounts
of
data
and
need
to
retain
the
modified
times,
because
maybe
they're
sinking
some
things
and
want
to
know
when
things
have
been
updated,
can
can
actually
know
that
information
because
is
currently
doesn't
have
that
that
sort
of
thing
baked
in
and
file
modes.
So
that's
fun
times.
We
had
a
spec
where
we
created
a
spec
for
UNIX
of
s
1.5.
This
is
a
stopgap
between
unit
before
Unix,
FS,
v2
lands,
which
I'm
sure
is
coming
soon,
but
yeah
yeah.
A
D
A
A
Okay,
only
new
nodes,
okay,
but
I,
guess
the
good.
The
good
part
about
that
is
that
we
had
the
spec
builds
in
this
option
for
there
being
nodes
that
are
linked
to
from
the
metadata
node.
So
even
if
we
couldn't
read
the
destination
node,
you
could
probably
still
read
that
that
P
being
look
at
the
links
and
get
the
whole
of
the
file
anyway
or
most
of
the
file,
depending
on
how
its
constructed
anyway
enough
enough
of
that
that's
to
do
with,
like
you
can
put
some
of
the
file
data
in
the
metadata
note.
A
If
you
want-
or
none
of
it
anyway,
let's
move
on
because
I'm
not
helping
things
I
will
try
and
drop
a
link
to
the
UNIX
FS
1.5
spec
in
the
notes
for
the
show
cool,
okay.
So
the
upgrading,
let's
move
on
because
we
go
like
15
minutes
in
idomeni,
you
done
a
little
bit.
Upgrading
testing
an
infra
and
process.
Is
anyone
around
to
talk
to
that?
A
E
A
D
Yeah
sure
so
I
did
some
hacking
week,
a
little
bit
on
the
airport
on
the
airplane
on
the
way
back.
Having
file
size
versus
can
have
become
dramatically
faster.
So
what
I
did
was
a
few
things,
so
the
first
thing
I
did
was
when
you're
importing
a
file
at
the
moment
it
just
like
imports.
One
file
was
the
chunky
crates
today,
cousin
D
then
moves
on
to
the
next
flop.
So
now
it
it'll
do
like
ten
at
once.
Well
do
an
arbitral
number
of
one
and
then
for
each
file.
D
It
will
be
all
the
it
will
do
a
whole
bunch
of
nodes
at
once
when
it
creates
the
dag.
So
it
takes
the
buffers
and
it
hashes
them
and
turns
the
first
noise
case
to
go
be
like
all
that
kind
of
stuff
and
when
we're
doing
all
that
sequentially,
so
now,
that's
all
done
in
parallel
and
then
I
I
made
it
so
that
the
pinning
the
flushing
the
pinning
becomes
debounce.
So
at
the
moment,
every
time
you
add
a
file
it
pins
up
file
there
it
creates.
D
So,
yes,
another
time
is
spent
just
flush,
which
is
super
tedious,
but
I'm
sorry
do
bounced
it
like
standard
kind
of
thing
that
you
do
as
a
developer.
So
now,
every
time
you
say
flash
you
know
wait
a
little
bit
and
then,
if
something
else
says
please
flush,
they'll
flush,
the
second
time
round,
because
the
thing
is
when
the
second
thing
says:
please
lunch
it's
because
it's
changed
the
pins,
so
the
the
initial
call
to
to
flush
pins
now,
I
would
say
because
the
princess
been
affected.
D
So
you
don't
even
like
you
just
redundantly
wrong
data
again
and
again
and
again
so
now
it
would
just
like
she
got
a
whole
list
of
fuzzy
we're
doing.
It
will
just
wait,
wait,
wait
and
then
it
will
just
flush
at
the
end,
which
is
much
faster,
so
that
is,
has
basically
doubled
the
performance
of
adding
so
the
the
add
performance
at
the
moment.
Without
these
improvements
is
comparable
to
go
equivalents,
that's
half
the
speed,
which
is
quite
much
the
other
thing
that
I've
done
is
either
this
idea
of
non-atomic
far
right
to
the
datastore.
D
This
means
that,
like
in
practice,
this
means
the
the
far
stored
data
store.
We
wait
for
all
bytes
of
a
file
to
be
flushed
to
disk.
Before
we
consider
the
operation
complete.
We
do
that
with
a
in
which
record
fast
right
tonic,
which
is
not
fast,
but
it
does
write.
Atomically
just
cool,
except
for
its
ring,
slow
and
say:
node
has
its
own
right.
D
So
you
could
just
use
an
or
atomic
far
right
in
that
case,
and
that
has
delivered
an
enormous
speed
up.
So
passing
that,
as
an
option
has
basically
increased
the
speed
of
adding
by
an
order
of
magnitude
which
is
super
cool,
so
I'm
just
trying
to
rationalize
those
into
a
bunch
of
PS,
all
the
different
libraries
that
we
can
use
to
get
that.
You
know
super
good.
Any
question.
A
So
I
guess
the
caveat
is
that
if
you
are
writing
non
atomically
to
the
repo
and
if
you,
if
someone
tries
to
get
something
from
from
my
PFS,
get
that
CIT.
Ok,
a
CID
like
if
it's
quick
enough!
So
if
it
gets
the
result
back,
it's
quick
enough
to
try
and
get
the
CID.
It
could
ask
the
repo
and
say:
have
you
got
this
and
the
repo
you
can
say
No,
and
so
it
will
go
to
the
network
and
ask
start
searching
the
DHT
yeah.
A
Cool,
ok,
so
that's
that
Aiden
is
not
here
today
for
any
other
I
didn't
performance
stuff.
The
migration
tomorrow
keys
in
the
Block
store
is
currently
I.
Just
haven't
had
a
time
to
look
at
that.
Yet
so
it's
it's
it's
on
hold
for
now,
just
because
other
things
have
become
more
important,
but
I'll
get
better
soon
duck.
Do
you
have
any
bits,
WAP
updates
for
us
I
know.
E
A
I
think
Libby
to
me
is
the
one
that's
lacking
at
the
moment,
so
you
might
want
to
check
in
I,
know
Jacobs
not
not
in
today,
but
he
will
be
around
tomorrow,
I
think
so
you
could
check
in
with
him
yeah
in
terms
of
like
Jase
IVFs,
it's
literally
just
the
remaining
we're
all
the
way
up
to
ipfs
level.
Now
so
we're
doing
pretty
well
but
yeah.
That
would
be
super
useful.
A
There
we
go
already.
Next
up,
we
have
a
sink,
a
weight,
retractor
good,
we're
already
talking
about
it.
That,
like
I,
said
it's
merely
completing
the
PTP,
are
currently
bubbling
their
stuff
up.
I
did
a
little
bit
of
work
last
week,
but
a
week
before
on
a
module
called
link
for
it
length
prefixed,
which
is
just
a
fork
of
pull
length
free-fixed.
So
it's
just
an
async
iterables
version
of
of
a
an
async
iterable.
A
Where
you
can
read
you
can
encode
and
decode
messages
that
are
prefixed
with
a
number
to
say
how
long
they
are
and
that
prefix
is
usually
a
variant.
So
you
can,
it
can
be
of
arbitrary
size,
but
one
thing
that
the
Lapita
peak
guys
realized
they
needed
for
sec.
Io
was
actually
I,
think
it
uses
it
actually
doesn't
use
variants
for
its
length
free,
fixed
messages.
A
A
E
A
A
About
the
message
lengths
like
it
I
I,
didn't
know,
the
key
would
be
nice
to
just
be
able
to
send
the
messages
at
the
moment.
I
know
that
if
you
exceed
that,
if
you
give
it
a
chunk
of
buffer,
that's
more
than
four
Meg,
then
it
will
throw
it
will
not
let
you
send
that,
but
then,
but
then
now
I'm
thinking
like,
why
would
you
have
a
chunk
in
memory
that
is
more
than
four
Meg
of
data
I'll.
D
Because
the
question
was,
how
do
we
inform
the
client
because
they
just
send
the
message
that
the
sort
of
explain
you,
the
remote
mode,
they've
sent
a
message?
That's
too
big.
Do
we
not
just
want
to
like
a
chop
the
drop,
the
connection
or
something,
because
only
if
they're
sending
you
bad
things
and
they're,
probably
a
bad
actor,
you
probably
don't
to
be
connected
to
them
anymore,
I.
A
A
E
A
A
A
If
you
say
I
mean
anyway.
This
is
a.
This
is
a
conversation
for
sometime
else
anyway.
So
yeah
async
wait
reflected
that
we
we've
bothered
stuff
all
the
way
up
the
J
side,
because
the
HTTP
client
has
now
been
almost
all
converted,
but
from
like
it
still
has
full
streams
and
edge
streams
and
callbacks.
So
we're
really
close,
which
is
super
good
news
on
that
front.
A
A
A
So
that's
that
super
good
news
I've
tried
to
tag
people
in
an
issue,
but
github
will
only
mention
the
first
50
people,
so
so,
if
I
I
need
to
send
a
link
here,
but
basically
the
the
next
step
is
to
have
contribute
to
sign
off.
So
everyone
needs
to
agree
that
that
everything
they've
contributed
in
the
past
and
in
the
future
will
now
be
licensed
as
MIT
and
Apache.
So
that
would
be
be
cool
if
you
have
not
received
a
notification
for
that
particular
problem,
then
no
problem
on
that
particular
issue.