►
Description
Details here: https://github.com/ipfs/dynamic-data-and-capabilities/issues/44
A
A
A
So
this
last
last
two
weeks
were
mostly
filled
with
preparation
for
protocol
protocol
a
protocol
week
and
ipfs
days
which
happen
in
Glasgow
last
week,
so
some
people
are
still
recovering
from
jet
lag
for
that
and
a
lot
of
interesting
discussions,
mostly
on
the
FS
days,
segue
into
the
discussion
of
persistent
strategies
for
pierced
are
related
applications.
A
lot
of
things
were
discussed.
There's
a
bunch
of
issues
on
aren't
pierced,
Iraq.
C
A
A
Authorities,
so
that
there
is
this
issue,
we're
all
primitive
values
are
unique
in
JavaScript,
so
to
set
value.
Neatness
is
based
on
that
JavaScript
uniqueness.
Some
people
complained
about
that
and
rightfully
and
so
right
now
the
started:
there's
a
pull
request,
waiting
to
happen
strategy
to
use
the
hash
value
of
the
object
to
determine
the
key
for
that
objects
for
being
unique
in
in
asset,
and
this
has
implications,
are
multiple
different
reality
types
at
the
ones
that
implement
sets
and
the
ones
that
use
a
dot
set
for
implementing
some
some
see
oddity.
A
So
there's
a
like
the
there's.
A
discussion
also
happening
here
for
object
uniqueness.
If
you
want
to
chime
in.
If
you
have
an
opinion,
your
opinion
is
appreciated
right
now
the
strategy
is
like
I,
said,
use
a
hash
value
for
that
and
so
I'm
hoping
that
someone
will
review
it,
I,
tagged
and
but
I
wish
to
review
it.
If
you
have
time
also
anyone
that
is
interested
in
the
image
I
mean
I,
also
some
other
work
on
PSR
at
which
is
work.
That
I
was
doing
started.
A
Doing
on
last,
call,
which
is
the
gossip
Pierre
addresses
in
bed
in
a
strong
eventual,
consistent
manner,
I'm,
mostly
a
work
in
progress
and
at
the
face
of
Eddie
unit
tests
for
membership
and
discoverability,
so
that
things
are
instead
of
testing
this
using
integration
tests
I'm
going
to
isolate
this
into
tests
that
are
specific
for
for
membership
and
simulates.
All
the
dependencies
outside
of
of
the
thing
that
we're
testing
and
my
plan
for
next
week
is
to
continue
on
on
these
two
and
most
probably
the
most
probably
on
finishing
the
gossip
Pierre
Gosling.
B
Yeah
I
got
a
question
regarding
the
uniqueness
and
you
talked
about
the
sets,
so
you
said
you're
using
some
kind
of
function
that
slashes
an
object
and
computes
the
the
the
value
for
it
and
use
it
as
an
identifier
of
the
the
entry.
Are
you
planning
on
adding
a
feature
to
customize
that
behavior,
so
that
users
can
specify
the
ID
themselves?
B
A
A
B
This
one
minor
thing
did
you
think
about
the
user,
making
the
hash
code
property
known
known
in
normal
enumerable,
so
that
it
doesn't
appear
in
console
logs
and
and
stuff
like
that,
and,
if
so,
have
you
thought
about
when
you
dis
realizing
the
state
that
is
being
replicated
to
also
put
it,
as
you
know,
in
umaril,
or
is
just
that
I'm
complicating
stuff.
Just
by
saying
that
no,
I.
A
B
Mean
yeah
I
just
raise
the
dead
up,
because
usually
people
don't
use
that
property
for
uniqueness.
They
use
something
like
ID
or
in
my
case
CID
and
so
on.
So
people
have
to
put
that
property
into
the
objects
themselves
and
whose
willy
people
put
put
them
non
enumerable
so
that
don't
they
don't
really
belong
to
the
objects
and
he's
being
used
just
by
by
pier
star
to
to
identify
the
items.
But
it's
something
that's
you
can
discuss
afterwards.
Yes,.
B
So
for
for
everyone
that
doesn't
know
I'm
working
on
discussing
Phi,
which
is
decentralized
up
where
you
are
able
to
comment
on
any
site,
I'm
developing
an
extension
for
Chrome
and
Firefox.
So
in
my
previous,
during
these
these
two
weeks,
I've
concluded
with
the
five
months
of
the
application
itself,
including,
but
not
limited,
to
20
factor
that
I've
made
so
that
we
ducks
that
I
was
using
in
the
background
script
as
well
as
this
license
themselves,
and
the
slices
by
slices
I
mean
the
side
bar
dot
Education,
the
mains.
B
The
main
script
set
has
the
the
D
on
the
bottom,
so
this
license
don't
use
with
ax
anymore.
They
use
react,
react
component
that
connects
to
the
background
script
and
pulls
the
states
continuously
and
in
listen,
cottage
state
changes
as
well,
so
the
redux
is
only
is
only
on
the
background
script.
Now
this
is
Jesus
required.
Simple
final
order
of
things,
I've
also
completed
the
polite
and
remove
comments
feature
in
the
wide.
The
user
interface
I've
set
up
the
peer
stirrup
library
in
there
in
the
background
script.
B
So
if
you
guys
remember
in
the
last,
you
know
last
call
I
was
in
doubt
I
had
my
doubts
regarding
the
type
FS
being
incinerated
in
in
each
table,
and
we
settle
on
having
just
one
one:
peer,
startup
business
and
one
ffs
instance
running
on
the
back
home
script
and
I've
set
up,
and
also
some
minor
pull
requests
that
I've
made
on
pure
startup.
There's
an
example
up
that
shows
the
counter
and
a
discussion.
B
B
You
will
also
be
able
to
do
that
and
that's
it
and
for
my
next
steps
after
this
is
to
implement
the
replies
in
the
user
interface.
I've
skip
this
one
and
and
went
through
and
through
the
in
end,
to
the
CEO
duty
on
a
model.
First
I
need
to
implement
the
show
history
in
the
user
interface
that
will
use
their
history,
derived
missions
and
also
I
will
I
will
hopefully
had
persistence
of
the
data
based
on
the
work
that
Pedro
I,
know
patron
team
and
I.
B
B
These
problems
regarding
the
ipfs
instances,
it's
not
only
related
to
extensions,
it
made
just
more
evident,
but
if
you
think
about,
if
using
a
profess
on
the
on
the
web
right
now,
you
have
at
least
20
per
fess
instance
per
tab,
and
that's
that's
not
really
the
way
it
should
work.
It
was
just
by
having
one
instance.
That's
why
there
is
the
extension
where
you
can
use
the
your
demon.
They're
installed
is
installed
on
your
computer
and
that's
the
only
ace
to
instance
running,
but
of
course
that
will
require
users
to
have
annexation
and
I.
B
Think,
as
I
told
in
some
previous
meeting
there's
a
working
group
that
wants
to
take
on
that
by
using
a
bridge.
So
especially
when
you're
using
the
preference
on
each
tab,
it
isn't
IPF
assistance
itself,
it's
just
some
proxy
to
a
service
worker
running
on
ipfs
domain
and
it
behaves
like
one,
but
it
isn't
the
profess
instance.
It's
running
just
a
one
instance
per
browser.
That's
the
real,
the
real
deal
there,
and
hopefully
the
working
group
responsible
for
that
I
think
it's.
B
A
Yes,
the
Service,
Worker
gateway
and
pattern
is,
is
I,
think
a
really
interesting.
One
thing
is
that
we
don't
pierce.
Our
app
is
not
supported
on
that
model
because
we're
using
a
custom
protocol,
but
if
they
can
provide
like,
if
you
can
like
push
for
approval
or
supporting
custom,
we
b2b
protocols
on
top
of
that
of
that
proxy
model,
I
think
we
could
use
that
even.
B
B
D
D
I'm
gonna
be
working
on
for
now,
I
guess
we
can
call
it
multi
writer,
IP
NS,
but
basically
this
idea
that
somehow
I
need
to
propagate
changes
to
multiple
I
need
multiple
people
to
simultaneously
be
able
to
make
changes
and
have
one
of
them
not
Club
or
the
other
one.
As
what
happen
right
now
with
IP
NS,
the
current
model
that
I'm,
looking
at
after
talking
with
a
few
people
last
week,
including
Pedro,
Stephen
and
and
Cole,
are.
D
You
know
probably
get
the
updates
and
and
sort
of
synchronize
the
different
graphs
so
that
you
don't
have
to
keep
keep
quite
everything.
So
you
may
you
may
expect
some
some
showing
up
on
various
various
threads
I
I
just
got
referenced
on
on
one
of
the
threads
that
are
Dirk
is
working
on
and
there's
there's
another
thread.
There's
like
an
there's
another
another
thread
from
the
NGO
library.
D
A
D
Yeah,
like
one
of
the
one
of
the
one
of
the
reports
I'm
looking
at,
is
under
the
like.
The
go
flood
sub
library
that
started
in
like
October
2017
is
when
the
requests
started.
These
phones
have
been
going
on
for,
while
so
knowing
that
there's
like
a
few
people
that
are
involved,
that
I
can
reach
out
to
for
resources
to
help,
try
and
actually
hold
this
thing
somewhere
is
is
really
good,
so
it
expect
to
be
bothered.
A
Super
and,
and
just
just
as
a
side
thing,
there's
a
bunch
of
things
that
that
are
potentially
going
to
allow
us
to
converge
more
to
that
BFS
infrastructure
so
already
mentioned
in
the
I.
Think
last
few
calls
rendezvous
protocol,
which
was
also
discussed
in
last
week.
The
evolution
of
the
rendezvous
protocol,
which
is
something
that
will
replace
the
HD
for
this,
come
from
discoverability
so
for
discovery,
service
discovery,
discovery,
discovery
and
it
will
be
as
the
last
discussion
it
will
be
federated
protocol.
A
A
It's
is
very
good
to
know
that
it's
actually
being
worked
on
and
the
other
one
is
also
circuitry
like
with
with
that.
So
we
can
have
like
the
NOS
required
web
sockets,
star
or
web
sockets,
anything
or
sorry
or
whatever,
to
see
star,
for
instance,
and
just
being
able
to
relay
and
find
out
through
the
circuit
relay.
That
is
going
to
be
provided
by
well-known
Wow,
discovered,
discoverable
peers
on
the
network,
potentially
using
also
the
rendezvous
protocol
for
finding
those
those
peers
that
are
able
to
rely.
That
will
help.
A
The
Jas
side
of
things
were
where
we
need
to
we're.
Very
limited
in
discoverability
were
very
limited
in
well
in
the
browser,
at
least,
very
limited
in
discoverability
are
limited
in
in
how
to
connect
to
each
other.
So
this
is
exciting.
Also
that
bridges
the
gap
and
allows
us
to
cover
tomorr
with
the
global
ipfs
infrastructure.
F
F
Probably
me,
okay,
so
before
I
went
to
lab
week,
a
fix
I
had
a
bunch
of
fixes
in
the
tree,
I'm
working
on
for
the
pure
pad
Nano
and
I,
put
them
into
the
peer,
strap
and
Pedram
fixed
merged
them
in
they're,
mostly
sort
of
really
sort
of
blatant
sort
of
bugs.
So
they
weren't
they
weren't,
challenging,
like
the
design
of
anything,
so
I
think
they're,
pretty
easy
to
put
in
and
then
spent
the
whole
week
in
Glasgow
at
the
All
Hands
protocol
lab
week
thing
and
I
give
a
presentation.
F
F
A
If
I've,
not
not
the
questions,
it's
just
that
yeah
come
on
over
conference
again
for
the
Pinner,
maybe
surfacing
membership
cycle
membership
override
cycle
on
the
Sunda
memberships,
charity
I.
Think
I'm.
Detective!
That's
why
I
resorted
to
to
go
into
the
unit
tests
for
the
membership
and
trying
to
isolate
this
I.
Think
if
I
conclude
my
work
in
time,
we
also
have
your
bug
fixed.
If
the
causes
is
the
membership,
but
I
encourage
you
to
dig
into
more
into
exactly
why,
probably
those
will
be
membership.
Gossip
message,
yeah
that
you're
getting
I'm
almost
sure
yeah.
F
F
A
You
should
see
the
the
reflection
should
see
that
inside
the
memberships
charity
there's
a
bunch
of
vector
clocks
and
you
should
see
the
vector
clock
counts
like
going
up
really
fast,
it's
at
least
what
I
saw
on
certain
occasions.
So
if
that's
that's
the
case,
just
ping
me
when
and
we
can
work
once
you
detect
the
cause
we
can
work
through.
If
that's
the
me
any
more
questions
for
Jim.
So
it's
my
comment.
C
Cool,
so
now
is
the
bride
Susan
okay,
so
for
this
week
I
was
focused
more
on
pure
path
in
terms
of
UX
and
UI.
This
is
the
uncritical
crib
that
it
will
be
presented
all
my
okay
ours
regarding
the
topic
number
five,
which
is
p
r--
bad.
So
I
created
a
live
demo
where
at
which
I
will
display
afterwards
the
regarding
the
some
some
details
for
the
app.
C
C
Is
it
let
me
know
once
you
can
see
my
screen,
can
you
can
see
yep
yeah,
okay,
so
I'll
play
the
video
and
then
I
can
pause
it
one
any
any
single
moment
that
is
quite
different
from
what
is
being
implemented
right
now.
Okay,
this
is
the
current
structure.
As
I
said
previously,
the
skinny
mark
banner
might
be
this
mis'able,
so
it's
not
so
permanent.
On
the
page,
this
this
relation
between
the
toggles
and
the
section
underneath
I
think
it's
much
better.
C
C
There
to
here
there's
a
possibility
to
rename
the
documents
much
more
friendly
and
related
with
the
they
added
itself.
This
was
the
notification.
I
was
referring
to
so
there's
a
one
person
online
and
then
someone
just
you
have
a
notification
number
increases
your
both
are
online.
You
can
hover
and
see
which,
which
person
has
has
entered.
You
can
yeah
edit,
your
name
set
it
and
once
it
will
be
set,
you
can
just
have
the
option
to
change
it,
because
this
is
the
first
time
you
do
this
action.
C
You
don't
need
to
have
that
as
a
repeat
step,
so
the
unpure
path
right
now,
there's
always
the
option
to
set
your
name
set
their
name.
That's
it's
really
weird
and
then
yeah,
just
a
small
change
between
what
is
implanted
implemented
right
now.
The
offline
has
this
dot
associated
with
the
gray
area
and
yeah
I
was
just
taking
advantage
for
of
the
the
screen
and
try
to
accommodate
everything,
a
bit
better
and
well
organised
right
now,
another
detail
led
I'm
working
on
it's.
The
history
I
just
started
like
off
an
hour
ago
and
yeah.
C
You
have
a
new
new
hiking
here,
which
is
the
version
history
itself,
and
then
a
drawer
will
appear.
You
can
show
changes
on
the
documents
itself,
just
like
Google
Docs.
Does
it
like,
with
a
highlight
color
I,
like
between
any
user?
Imagine
that
okay,
this
version
was
changed
by
shadows
or
you
will
have
like
a
rectangle
with
this
color
and
then
just
look
like
github.
So
does
when
you
add,
or
remove
something
yeah,
just
try
to
accommodate
those
relations
of
colors,
with
the
changes
itself
and
yeah.
For
now.
That's
that's!
C
B
Don't
mean
that
I
pick
the
type
indicator
I
mean
them
the
propagation
of
the
changes.
So
we
have
five
like
everyone
on
this
on.
This
call
are
editing
these
documents
and
perhaps
Pedro
is
having
a
slow
connection
or
is
momentarily
offline.
I
want
to
know
if
my
changes
were
propagated
to
everyone
and
if,
if
it
doesn't,
who
didn't
have
the
changes?
Okay,
okay,.
C
C
So
yeah
I'm
locked
on
the
it's
clear
that
where
data
is
persisted
and
it's
clear
when
that
sorry,
where
data
is
persisted
and
it's
clear
when
that
these
procedures
I
have
a
few
questions
regarding,
is
it's
the
first
one?
It's
the.
How
does
this
will
influence
design
and
look
okay
and
I
approach
regarding
these
two
points
to
understand
what
should
I
do
afterwards?
I.
A
Can
I
can
answer
that
well,
but
was
partially
so
I
think
that
there's
a
bunch
of
interesting
UX
things
we
can
do
but
I
think
the
really
useful
thing
is
like
I
said
you
know
the
sphere
X
have
my
my
latest
changes.
That's
one
indicator
that
is
useful
when
we're
collaborating
to
know
to
know
that,
and
the
other
is,
is
this
persistent
on?
A
A
Currently
we
can
expose
that
in
in
the
Pierre
part,
sorry,
a
pierced,
our
app
API
and
the
last
one
is
all
fully
a
consequence
of
all
the
persistence
of
the
persistence
solution
that
that
we
do
and
I
think
I
think
those
are
the
interesting
ones
there.
Another
another
pattern
that
that
I
thought
of
could
be
interesting,
but
this
is
app
for
discussion
and
I
like
er.
A
My
changes
like
propagated
to
more
than
half,
for
instance,
of
my
of
my
membership,
and
this
is
may
be
useful
or
behalf
like
if
the
percent
is
like
as
good
as
any
I'd
say,
I
mean
arbitrary
number
here
or
everyone,
for
instance,
everyone
in
the
collaboration
before
just
to
collaborating
50
percent.
It's
just
it's
just
me.
So
something
do
to
keep
me
in
the
back
of
your
heads.
I
think
the
first,
the
first
one
is.
First,
you
are
really
important.
The
last
one,
maybe
in
the
absence
of
of
real
persistence.
A
A
B
I
will
I
will
give
my
opinion
if
possible
wishes.
We
should
think
about
the
the
point
that
Pedro
raised
as
possibly
as
a
pact
solution
instead
of
resolving
each
one
separately.
So
this
is
something
that,
as
Butler
says,
doesn't
happen
in
centralize
world,
because
in
centralized
rules
we
just
have
a
server
and
things
are
persistent
there,
and
once
you
get
a
confirmation
that
is
persisted,
it's
all
fine.
So
it's
just
one
one
one
thing
to
worry
about,
but
when
we
are
dealing
with
decentralize
there's
two
options
or
three:
we
we
either
have
a
persistence.
B
Disabled
and
my
persistence.
Disabled
I
mean
there's
no
like
no
the
persisting
the
data
and
then
the
today
they
persisted
by
the
peer
themselves
and
there's
also
another
scenario
which
is
having
both
combine,
which
is
both
peers
and
both
dedicated
no
to
persistent,
persistence
and
there's
another
scenario,
which
is
just
the
persistence
notes,
and
you
are
alone
editing
the
W,
so
I
think
all
these
three
things
should
be
combined
in
a
new.
Why
wax
pattern?
B
That's
not
only
will
will
work
for
pure
star
pure
pure
path,
sorry,
but
also
for
the
specify
and
and
other
other
other
interfaces
as
well
all
our
other
apps
as
well.
So
don't
we
don't
really
need
to
repeat
them
or
make
different
Y
approaches
for
these
kind
of
problems.
We
should
look
into
develop
a
unified
approach
to
all
of
these
and
triplicate
and
reuse
that
great.
A
C
A
A
Think
people
then
can
implement
different
different
patterns.
I
mean
you
can,
if
you're
dependent
on,
depending
on
the
type
of
collaboration
that
that
you're
having
you
can
have
different
requirements,
if
it's
just
photo
sharing
virtually
consistent
thing
that
takes
a
long
time.
If
it's
text
it's
more
real
time
more
of
a
real
time
thing,
if
it's
geared
so
as
they're
just
working
on
on
a
guitar
like
saying,
probably
different,
a
bit
of
different
requirements,
so
I
think.
D
Which
is
you
you're,
pointing
out
the
sort
of
doing
the
text
highlighting
for
who
made
what
changes
and
I
was
wondering?
If
you
have
any
thoughts
or
just
fun
that
you,
you
may
run
into
issues
with
the
Delta
C
Rd
tease,
because
when
you
squash
them
you're,
not
gonna,
know
who
you
may
not
know
who
made
the
changes
or
if
you
did,
then
you
missing
signatures
missing
signatures
might
be
okay,
but
you're
gonna.
D
D
A
Now
is
it
just
just
say
that
there's
options
in
so
I
was
just
yeah.
That
was
one
of
my
points
against
the
Delta
charities.
Initially,
it
was
well
what
about
history,
what
about
authorship,
and
but
the
thing
is
that
that
you
can
control
well,
you
have
to
control
it
at
the
edges,
but
you
can't
control
whether
to
to
to
merge
the
deltas
or
not
so
it's
kind
of
a
choice
like
I
tried
off.
If
you
want
to
keep,
you
can
keep
history
up
to
a
certain
point
and
then
just
course
the
rest
going
going
back.
A
We
have.
We
have
a
bunch
of
options.
Also,
you
can
embed
like
we
can
do
wealth.
We
can
embed
some
history
like
a
side
channel
if
you
will,
but
in
in
terms
of
like
a
side
sanity
with
with
history.
That
is
causally
consistent
with
with
the
other
one
which
resemble
more
of
an
operation
by
sanity
and
there's
a
bunch
of.
C
There
was
a
I
was
trying
to
reply,
but
I
seen
a
detail
today
that
which
is
the
snapshot
concept
for
any
change
that
is
created
on
a
document
where
you
can
like
create
a
stack
of
this
was
changed
today,
but
inside
of
today
you
ever
add
another
level
of
history
inside
over
they
imagine
it
creates
more
than
just
one
level
than
history
and
then
you
can
almost
save
it
and
then
just
continue
with
an
other
change.
Yeah,
it's
a
bit
messy,
but
it
works.
A
E
Oh
so,
yes,
as
you
guys
have
mentioned,
I've,
been
working
on
trying
to
implement
a
kind
of
proof
of
concepts
for
persistence
and,
firstly,
I
just
want
to
say
thank
you.
Everyone
do
like
inputed
and
gave
comments
and
stuff
you
made
some
really
nice,
diagrams
and
stuff
and
a
bunch
of
people
put
in
comments.
So
thank
you
for
that
work.
E
So
there's
sort
of
like
a
lot
of
different
variables
and
autoclave
ways
of
doing
this,
so
I
decided
to
choose
something
very
specific
and
you
come
at
it
as
a
proof
of
concept
and
basically
the
scenario
is
that
a
bunch
of
people
are
making
changes
to
southern
CRT.
To
let's
say
it's
a
bad
document,
then
everyone
goes
offline
and
then
someone
comes
online
again
or
someone
new
and
they
need
to
be
able
to
see
all
everything.
That's
happened,
or
at
least
the
current
state
of
the
document
so
messed
up.
E
As
always
mentioning
there's
kind
of
like
a
couple
of
different
options.
We
could
have
a
persistent
service,
that's
a
single
entity,
that's
running
all
the
time
and
which
is
kind
of
what
the
pinning
services
and
then
another
option
is
that
we
can
have
a
service
that
runs
within
each
node
and
and
so
each
node
has
a
service
that
can
persist
its
local
changes
plus
any
state
that
it
sees
that
comes
in
across
the
network.
E
So
I'm
excited
that
a
team
is
working
on
multi
writer,
IP
NS,
because
that's
gonna
take
care
of
a
bunch
of
concurrency
problems
that
we
have,
and
so
for
the
time
being,
there's
a
couple
of
strategies
that
we're
using
to
use
single
writer,
IP
NS.
One
of
those
is
basically
that
we
have
two
levels
of
IP
and
X
record.
E
One
keeps
track
of
all
the
participants
or
the
members,
and
then
at
the
second
level,
each
member
has
their
own
IP
n
s
key,
which
they
write,
their
local
state
and
the
local
updates
to,
and
so
when
a
new
member
comes
in,
they
get
the
first
level.
Ip
NS
record,
find
all
the
members
and
then
query
each
of
the
members
IP
s
keys
to
get
each
of
the
individual
states.
Imagine
the
second
option
we're
working
on
is
to
elect
a
leader,
and
so
there's
only
one
node.
E
That's
persisting
everything
basically
and
then,
when
that
node
disappears
or
goes
offline,
the
other
nodes
detect
that
and
they
elect
a
new
leader.
So
the
nice
thing
about
that
approach
is
that
because
the
state
is
already
being
sent
around
and
everyone
is
sharing
the
same
state
at
the
collaboration
level
through
gossip,
we
can
kind
of
sidestep
having
to
every
single
node
having
to
record
the
same
information.
So
it's
just
a
single
leader,
so
both
of
these
strategies
work
fine.
As
long
as
there
is
no
partition
in
the
network,
once
we
do
have
a
partition.
E
There
is
a
couple
of
different
scenarios
depending
if
we're
doing
a
leader
election
or
if
every
node
is
writing
and
I've
kind
of
outlined
those
scenarios
in
the
PR.
So
you
can
have
a
look
in
that
and
your
comments
to
them.
But
the
nice
thing
about
the
gusset
protocol
is:
it
ends
up
just
kind
of
peeling
everything
in
the
end,
as
long
as
there's
and
node
on
each
side
of
the
partition
comes
alive
at
some
stage
you
can
guarantee
the
state
will
converge.
A
F
E
For
sure
I
mean
I'm
kind
of
like
trying
out
a
few
different
things
in
that
PR.
So
I
see
it
very
much
a
Wiccan,
Fergus
and
love
to
hear
you
guys
comments
and
kind
of
suggestions
on
different
strategies.
My
guess
is:
I'll
probably
end
up
implementing
it
a
few
different
ways,
and
then
we
can
try
them
out
and
see
what
works
best
with.
F
E
You're
right,
thank
you
got
to
mention
that
at
the
moment,
I
pns
will
only
work
locally.
Essentially
it's
for
any
anyone.
Who's
sharing
this
store
the
data
store
within
my
penis,
which
is
anyone
who's
on
the
same
boat,
but
as
I
understand
it.
If
they're
working
on
connecting
up
IPS
with
the
DHT
as
well
as
I
penis
with
first
up,
can
you
speak
one
of
that
bitter?
E
A
The
the
Jessica
fest-
and
we
are
working
on
that
so
that
between
this
weekend
and
next
week
they
am
to
have
the
DHT
sorted
out,
and
then
they
can
link
the
mechanist
with
the
HT
quite
easily
I.
Think
after
after
that,
and
so
so
that
there's
there's
this
whole
DHT.
Never
that
I
can
link
to
I
have
to
find
it
first,
there's
a
like
an
awesome
endeavor,
which
a
name
that
JSF
FS
at
least
seems
describe
like
things
that
span
multiple
issues
and
pull
requests.
A
The
DHT
part
to
awesome
endeavor
is:
is
that
that
that
lasts
step
required
for
initially
to
be
to
be
usable
on
GFS
and
from
then
on,
as
I
said,
a
penis
that
uses
the
DT
or
before
that
they
also
are
implementing
a
penis
gossip
through
a
penis
closets
through
pops
up
I.
Believe
that's
also,
one
I
think
that
they're
actively
working
in
so
yeah
it's
looking
at
it's
a
hard.
It's
a
hard
Walker
for
you,
I
believe
they're
right.
E
F
D
D
D
Let's
just
have
this
multiple
naming
system
thing,
and
but
if
there's
this
idea
that
I'm
I'm
still
processing
right
now
about
whether
it
makes
sense
to
use
something
else,
because
we're
not
really
trying
to
broadcast
for
information
to
everyone,
we're
more
trying
to
sustain
it
for
this
group.
That
already
knows
about
each
other.
E
D
You
I
guess
to
clarify
what
I
meant
is,
there's
some
sort
of
external.
What
call
it
state
there's
some
sort
of
external
state
that
has
to
get
to
you
anyway,
as
a
function
of
me,
sharing
with
you
all
right,
especially
for
using
encryption
and
I
need
to
get
you
an
encryption
key
of
some
sort
or
or
just
let.
You
know
that
by
the
way
this
file
has
been
shared
with
you
in
the
same
way
that
Dropbox
or
whoever
like
they
send
you
an
email,
letting
you
know,
there's
a
new
file.
D
And
I
think
that
that
you
know
he
can't
see
the
air
quotes,
but
that
external
method
that
I'm
talking
about
should
eventually
be
part
of
ipn
I
like
be
part
of
the
protocol,
was
data
that's
stored
on
IPSS,
whether
that
looks
like
you
know:
sort
of
internal
like
email
over
IP
FS
or
some
sort
of
basic
chat
over
IP
FS.
That
lets
you
know:
hey,
there's
this
new
file
or
or
similarly,
if
I
want
to
like
kick
someone
off
a
document.
D
D
Let's,
let's
do
is:
let's
do
this
whole
thing
again,
but
like
without
pedra
this
time
that
sort
of
thing,
so
it's
tricky
but
I-
think
there's
there's
like
enough
information
flow
around
that
we
may
not
want
to
use
IPS,
except
when
we
have
to,
for
instance,
if
Li
you
may
actually
want
for
things
like
access
control
groups.
The
last
rider
wins
semantics
of
IP
ns.
D
C
E
A
B
Just
one
one
thing:
I
have
a
quick,
quick
Lokhande
on
the
pull
request
and
just
to
to
see
if
I'm
correct.
The
current
implementation
Allah
allows
people
or
developers
to
add
new
persistence,
implementations
of
strategies,
because
I've
seen
there's
a
main
memory,
implementation
and
there's
another
one
that
there
is
IPS.
E
Ipfs
is
kind
of
like
the
pod
implementation
of
the
content,
addressable
persistence
and
I
pns
is
the
hot
integrated
implementation
of
the
naming,
yeah
I'd
like
to
try
with
a
few
different
kinds
of
strategy
there.
For
example,
we
might
want
to
use
a
different
naming
strategy
that
supports
multi
write
in
a
more
natural
way.
Oh
yeah.
B
Yeah
even
like,
if,
if
a
particular
application
wants
to
go
for
the
centralized
approach
for
a
specific
use
case,
it
will
be
good
to
just
implement
the
interface
and
specify
the
strategy,
and
it
will
just
work
out
and
call
the
functions
of
the
implementation.
And,
if
it's
well
done,
it
will
just
work
out
for
everyone.
B
This
is
more
like
a
future-proof,
in
the
sense
that
it
allows
has
to
add
new
strategies
that
we
might
not
be
sure
about
them
right
now,
but
we
could
experiment
them
with
them,
but
just
providing
an
option
hey.
This
is
a
persistent
instance
that
has
these
methods
and
the
underlying
pure
star
just
calls
them,
and
it
will
work
out.
Yeah.
E
B
B
If
you
think
about
documents
that
you
share
between
your
friends
or
our
colleagues,
you
already
know
them,
and
if
they
provide
a
storage,
there
is
persistence
and
is
always
online.
You
can
always
retrieve
the
latest
state,
four
of
them
and
merge
all
together
using
the
Delta
Z
oddities
join
mechanism.
Just
to
you
know,
to
to
merge
all
the
state
of
your
of
the
people
that
are
working
on
the
document
and
it
will
work
out.
B
We
don't
need
twelve,
like
a
single
source
of
truth
of
the
state,
but,
as
you
might
assume,
as
image,
there's
a
strategy,
these
decent
another
strategy,
and
we
might
want
to
experiment
with
other
strategies
that
we
with
Arendt's
think
about
right
now
and
I.
Think
that
would
be
good
to
add
into
the
they
pierced
our
interface.
B
To
emission
something
there's
a
project
that
belongs
to
the
thief:
thief
foundation:
decentralize
Ida
Foundation,
which
is
essentially
a
storage
storage
service.
For
that
each
user
can
run
on
its
own
machine
and
of
course,
there
will
be
service
providers
where
people
can
host
their
or
you
can
run
on
your
computer
whatever.
And
if
you
think
about
what
what
adding
sets
it
could
be
a
strategy.
B
So
you
have
this
storage
that
each
user,
that's
edited
the
document
or
belongs
to
the
documents,
is
reliable
and
always
online.
You
can
just
query
the
lightest,
Titan
pooling
and
merge
to
say
together.
That
will
be
a
perfectly
valid
solution.
Of
course,
it
won't
be
using
IP,
NS
or
FS
capabilities
and
so
on.
It
will
probably
using
I,
don't
know
something
like
HTTP
and
and
some
key
value
storage
underneath
to
sort
the
the
information,
but
it
is
essentially
centralized
because
every
every
peer
provides
the
storage
to
to
which
we
even
start
states.
B
E
E
A
A
B
So
the
idea
is
that,
if
you
think
about
how
how
you
connect
people,
you
have
your
friends
and
the
friends,
your
friends
have
also
their
friends
and
the
friends
also
their
friends
and
so
on.
So
imagine
if
essentially,
each
each
person
will
be
able
to
provide
storage,
ok,
a
small
culture,
of
course,
for
each
of
your
friends
or
partial
list
of
friends
and
even
a
second-level
deep
of
your
friends
of
your
friends.
B
So
this
is
just
an
idea.
Of
course
it's
not
something
for
now,
so
something
that's
I
wanted
to.
You
know
point
out,
so
that's
the
current
persistent
layer
that
is
being
developed,
developed.
Allow
allow
us
to
experiment
with
these
future
concepts
of
of
the
upcoming
years
that
will
probably
tackle
on.
A
Yeah,
there's
the
victor
yushchenko,
just
just
passing
the
link
to
a
paper
from
97
from
work
cunningham,
which
is
an
architecture
for
this
thing
that
that
your
your
yard
is
kind
of
a
folk
memory
or
something
it's
also
like
prior
art.
Well,
more
recent
priorities,
which
is
the
well
not
exactly
this,
but
this
was
cuddle
but
secure
scuttlebutt.
A
Remember:
there's
a
p2p
file-sharing,
social
network
thing
where
you,
where
you
guarantee
that
they're,
like
you
offer
to
enter
the
network,
you
offer
storage,
but
the
thing
is
like
for
ipfs
falcone
solves
the
proof
that
you
provide
the
storage
right
so
that
that's
that's
the
whole
goal
of
the
project
is
to
for
you
to
prove
that
that
you
provided
the
storage.
So
the
proof
here
is
I.
B
Just
mentioned
mention
it's
just
to
open
possibilities
in
the
future
and
not
and
create
some
kind
of
interface
for
the
persistent
player
that
could
be
easily
iterated
on
and
and
it
each
of
course
will
be
solid.
So
that's
a
different
permutation
could
be
easily
implemented,
but
it's
something
that
we
probably
want
to
have
from
the
phone
to
start
going.
I
think
I
will
I
will
comment
on
the
pull
requests.
Okay,.
D
There's
there's
only
which
I
think
is
just
using
the
opportunity
to
like
collect
information
on
what
some
of
the
use
cases
people
are
looking
for
are
right.
I
mean
like
one
that
happens
to
appeal
to
me,
but
may
not
appeal
to
other
people
is.
I
would
have
been
reasonably
happy
with
a
cloud
with
like
having
amazon
store
my
data
instead
of
doing
it
peer-to-peer
because
the
encryption
would
protect
me
and
all
of
that.
D
C
F
This
is
sort
of
a
broad
thing,
but
I
think
an
architecture
that
I
am
seeing
the
might
evolve
out
of
some
of
this
work
is
like.
We
have
make
sure
that,
like
fat,
clients
that
have
like
IP
fasting
I'll
search
the
features
built
into
the
client
and
they're
completely
independent,
and
you
can
basically
default
peer-to-peer
networking
with
it,
but
I
time
everyone
sees
their
phones
and
they
don't.
They
actually
sort
of
want
thin
software
on
their
phones,
but
there's
concepts
such
as
service
workers,
where
you
can
sort
of
have
your
thin
client.
F
B
So,
as
I
said,
this
is
a
complex
topic,
we
need
to
I
think
we
need
to
come
up
with
a
document
somewhere
to
point
out
our
our
requirements,
our
concerns
and
construct
some
solutions
that
will
solve
different
points
and
some
others.
We
will
probably
don't
solve
those
some
of
those
problems
and
perhaps
we
could
converse
into
something
I
think
would
be.
It
will
be
good
just.
A
B
A
I
can
do
that.
I
can
do
like,
like
I'll,
have
to
review
the
video
and
then
and
then
just
just
put
a
snapchat
of
this
conversation
on
that
on
github
on
a
video
of
issue,
and
then
we
can,
we
can
take
the
call
from
that.
I
think
it's
like
a
long-running
discussion
probably
belongs
to
ipfs
notes
or
or
research,
or
something
like
that,
but
I
think
it's
really
with
also
to
take
away
points
right.