►
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
What
are
we
talking
about
today?
Private
content,
routing,
so
there's
a
couple
things
we
can
start
with
content
routing
and
the
specific
subset
of
this
overall
distributed
network
thing
that
we
are
are
dealing
with
right
now.
The
the
main
structure
that
we're
seeing
in
data
transfer
in
something
like
ipfs
is
you've
got
a
client.
You've
already
got
some
connected
peers
and
in
the
current
world.
A
What
routing
and
looks
like
when
you're
trying
to
look
up
a
sid?
Is
you
first
talk
to
your
connected
peers?
You
see
if
any
of
them
have
it.
If
you
don't
have
it
from
there,
you
fall
back
to
a
dht
lookup,
which
gives
you
more
connected
peers
right.
So
so
the
the
initial
lookup
is
through
is
over
bitswap
on
your
on
your
current
connected
peers.
A
When,
when
your
current
sort
of
set
of
connections
doesn't
work,
you
fall
back
to
a
dht,
because
it's
you
know
more
expensive
right.
It
takes
longer
than
those
immediate
single
hop
to
to
your
current
swarm
over
time
as
we
make
content
routing
a
fast
thing.
The
hope
is
that
that
is
is
no
longer
right
like
this.
This
I
start
by
talking
to
my
current
peers
and
then
I
fall
back
to
a
content,
routing
subsystem.
A
That's
not
something
that
you
want
to
do
every
sid,
that's
just
like
not
going
to
cut
it,
but
if,
if
this
was
fast,
if
this
meets
the
stuff,
we
were
talking
about
yesterday,
where
it's
10
milliseconds
that
stops
being
nearly
as
scary
and
in
fact
it
might
be
faster
to
first
go
to
content,
routing
and
say
who's
near
me.
Who
should
I
be
talking
to
if
that's
only
10
milliseconds,
then?
Knowing
that
I
go
to
the
right
peer?
A
Can
save
me
a
lot
of
bandwidth
and
time
versus
trying
all
of
my
connected
peers
before
I
make
that
10
millisecond
call.
So
at
least
you
would
want
to
do
them
in
parallel,
and
maybe
you
even
just
start
with
your
content.
Routing
look
up
on
a
sip
that
said,
if
it's
like
a
graph
and
you're,
pretty
sure
that
the
person
who
gave
you
the
block
before
also
has
the
blocks
under
it.
A
Maybe
you
can
continue
still
directly
doing
it
so,
but
there's
some
interplay,
that's
going
to
keep
changing
between
the
the
data
transfer
and
content
routing
and-
and
so
then,
what
is
this
interface
that
we're
building
around
content,
routing
you're
asking
for
a
sid
and
then
you're
getting
back
and
this-
and
this
is
the
part
that
isn't
in
the
interface
right.
Do
you
get
back
all
peers
that
have
that
set
all
potential
nodes
in
the
interplanetary?
Do
you
get
the
ones
that
are
closest
to
the
client?
A
We
have
not
really
set
up
semantics
yet
on
the
the
client
specificity
of
that
answer
in
the
dht
today,
it's
a
global
client,
agnostic
response,
but
but
in
order
to
get
our
performance
targets,
there's
a
lot
of
expectation
that
that
response
actually
cares
somewhat
about
the
client
like
who's
asking.
A
If,
if
I
ask
who
has
this
sid
in
north
america,
I
probably
expect
a
different
answer
than
if
I
ask
it
in
asia
pacific,
because
giving
the
full
list
of
everyone
who
has
that
sid
for
a
popular
sit
is
really
big
enough,
that
you
probably
are
going
to
need
to
truncate
it,
and
then
the
truncation,
where
you're
prioritizing,
is
a
different
priority
in
different
places,
and
so
that's
going
to,
I
think,
be
one
of
the
conflict
interplays
that
that
becomes
important
as
we
get
into
privacy.
A
Okay.
So
so
that
was
like
the
content
routing
but
then
what's
the
privacy
part?
Well,
there's
there's
a
few
different
definitions
of
privacy
and
we'll
see
those
as
we
get
into
different
mechanisms.
Today,
one
of
the
ways
that
we
say
what
is
privacy
is
a
concept
of
linkability,
so
linkability
is:
can
I
make
a
link?
Can
I
do
I
have
some
knowledge
between
the
the
client
and
then
what
they
asked
for
right.
A
So,
who
asked
for
what
or
like
the
did,
this
client
and
the
server
talk
like
if
I
can
create
links
between
two
things:
either
either
the
the
content,
layer
and
the
people
that
are
asking
for
it?
That's
a
you,
can
talk
about
linkability
there
or
the
two
sides
of
a
connection
if
you
can
link
them
together,
so
the
communications
coming
from
a
client
and
the
communications
on
client
or
a
server.
A
You
can
also
talk
about
that
as
a
linkability
problem,
and
so
that's
that's
sort
of
like
a
one
to
zero
like
like
it's
a
boolean
of
like.
Do
we
in
in
what
conditions
do
we
have
linkability
or
not?
Who
can
make
that
link,
there's
more
gradated
forms
of
privacy
that
we
can
take
as
well,
though,
so
there's
a
concept
of
k,
anonymity,
which
is
how
many
other
people
can.
A
That's
probably
the
biggest
number
of
confusion
that
you
can
get,
and
so
you
can
think
about
this
in
terms
of
that
k,
differential
privacy
is
another
sort
of
structuring
of
this
privacy
problem
and
that's
looking
at
the
distribution
of
a
set
of
logs
and
talking
about
how
do
you
inject
noise
or
think
about
the
the
perturbation
such
that
if
an
entry
is
removed
or
added
like
any
individual
entry,
doesn't
change
this
distribution?
So
so
what
do
we
need
to
perturb
and
how
much
distinguishability
or
how
much
differential
change?
A
A
The
other
thing
that
privacy
is
really
framed
by
is
your
threat
model,
and
so
we'll
probably
hear
this
either
implicitly
or
sometimes
called
out
explicitly,
which
is
what
is
it
that
we're
actually
protecting
about,
and
so
you
have
to
have
some
adversary
that
you
are
saying
you
know
this
system
is
designed
to
have
this
property
such
that
an
adversary
with
some
capabilities
can't
do
something,
and
so
that's
really.
You
know
what
what
what
do
we
let
the
adversary
do
and
what?
What
is
that
threat?
A
So
is
it
an
isp
who
can
monitor
traffic
going
over
wires
versus?
Can
they
actually
compromise
some
number
of
nodes
in
the
network?
Are
they
a
node
in
the
network?
So
is
it?
Is
it
someone
who's
running
an
ipfs
entity
or
a
you
know,
some
node
that's
participating
in
the
system
actively
or
you
know:
do
they
get
a
global,
passive
adversary?
They
can
see
actually
all
of
the
network
traffic.
So
so
this
is.
This
is,
I
think,
really
yeah.
A
A
Okay,
why
do
we
care-
and
I
think
we'll
we'll
get
more
of
this,
so
so
there's
one
thing
which
is
okay?
Can
we
just
encrypt
the
content
and
call
that
good
and
there's
a
couple
comment
like
memes
that
that
we
go
on
for
why
we
say
encryption
is
not
enough.
One
is
right:
metadata
has
become
a
big
meme.
This
is
this
is
the
place
where
it's
like.
Okay,
like
so
signal,
is
end
encrypted.
Is
that
good
enough?
What
else
do
we
need
to
do?
A
I
don't
know,
comment
on
this,
that
I've
heard
right,
which
is
we're
in
this
interesting
time,
where
we've
got
a
huge
amount
of
data
right
data
data
goes
up
and
has
gone
up
where
we
don't
yet
have
the
automated
data
analytics
that
have
pushed
us
into
a
surveillance
state
to
the
level
that
we
can't
get
out
yet.
A
But
that's
happening
right
like
as
ai,
is
moving
faster,
the
the
amount,
the
ability
to
just
like
capture
all
of
this
data-
that's
out
there
is
going
to
also
go
exponentially
while
the
data
might
not,
and
so,
if
we
don't
get
privacy
soon,
we're
going
to
have
a
lot
of
trouble
getting
privacy
like
that.
This
is
not
going
to
get
easier.
A
It's
also
true
that
there's
just
a
lot
of
applications
when
we
think
about
interactive
or
types
of
data
that
are
going
on
networks
that
that
have
privacy
implications
and
privacy
concerns,
and
so
we
should
be
aware
about
like
if,
if
we
want
these
to
be
applicable
to
the
interplanetary
stack,
we
need
to
think
about
what
the
privacy
story
is
right.
So,
if
you're
doing
e-commerce
or
things
with
payments,
people
don't
want
their
credit
card
just
on
an
ipfs
network
where
every
participant
can
see
it
right
and
so
great
okay.
A
We
can
like
encrypt
the
block
so
that
you
can't
just
like
see
my
credit
card.
That's
a
good
start,
but
but
the
there's
a
whole
bunch
of
other
things
around
transaction
volume.
Like
can
I
see
my
competitors
like
how
many
transactions
they're
making
over
time
where
you
start
caring
about
metadata
in
a
lot
of
these
cases
as
the
merchant
anything
that's
got,
chat
or
messaging,
where
it's
people
talking
they
care
about
their
privacy.
A
That
could
mine
people's
social
graphs
on
web2
they're,
not
going
to
like
that
that's
possible
in
a
web
3,
because
you
can
see
the
interactions
of
who's
requesting
what
sids
over
over
an
interplanetary
stack,
and
so
we
start
to
have
to
care
about
how
do
we
delink
stuff
and
then,
if
you've
got
file,
storage
or
collaboration,
either
in
an
office
environment
or
like
if
I'm
putting
my
backups
like,
I
probably
want
some
protection
there
now.
What
is
that
protection?
A
Is
that
protection
just
that
I've
encrypted
it
or
is
there
also
access
pattern,
stuff
or
or
more
informed?
That's
that's
a
a
deeper
question.
That's
going
to
be
nuanced!
I
think
a
lot
of
users
aren't
going
to
be
able
to
like
express
their
threat
model
and
what
expectations
around
k
anonymity.
They
have
there
like
that.
That's
not
like
the
way
that
people
talk
about,
but
they
want
privacy
in
some
level,
and
so
we
need
to
think
about
what
is
enough
that
we
feel
comfortable,
saying
great
you've
got
a
secure
and
private
connection.
A
Okay.
So
what
are
we
talking
about
today?
We've
got.
We've
got
a
bunch
of
different
talks
around
different
parts
of
this
privacy
as
it
relates
to
content,
routing
and
and
getting
into
data
transfer,
so
we'll
get
some
on
bitswap
and
the
at
the
actual
data
transfer
level
and
then
we'll
get
a
bunch
around
content,
routing
things.
So
thinking
about
mixnets
thinking
about
how
do
we
do
content
routing
where
the
content
is
not
leaked
to
the
network
which
is
double
hashing
gets
into,
and
then
some
of
the
other
exciting?
A
What
what
the
the
the
few?
The
more
researchy
forms
that
we
could
that
this
could
take.
There's
a
bunch
of
expectation
of
other
mechanisms
that
we're
not
going
to
talk
about
as
well.
So
so
this
is
not
an
exhaustive
list
either
of
what
mechanisms
to
be
thinking
about
or
that
we're
even
actively
considering.
So
so
I
want
to
like
throw
up
a
few,
but
there's
there's
actually
a
lot
more
as
well,
and
with
that
I'm
going
to
turn
over
to
phone
on
time.
So
there
we
go.