►
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
So
I'm
gonna,
I'm
gonna,
talk
about
an
effort
that
we're
that
we're
working
on
in
pl
for
funding
private
retrieval
this,
this
sort
of
is,
is
a
a
next
step
or
a
more
action-oriented
thing.
That
comes
out
of
a
lot
of
the
motivation
that
we
got
from
han
earlier
today.
A
So
in
terms
of
context,
there
have
been
a
couple
events
over
the
last
year
around
funding
the
commons,
and
this
came
out
of
plv8.
As
we
talked
about
you
know.
How
do
we
help
and
and
drive
this
innovation
chasm
of
getting
more
good
ideas
into
production,
faster
and
and
in
particular
the
one
one
of
the
blue
funds?
That's
sort
of
becoming
a
reality
is:
is
in
this
space
of
private
content,
routing
and
retrieval,
which
is
how?
A
How
do
we
encourage
more
mechanisms
and
more
possibilities
that
we
can
use
for
for
getting
to
where
we
want
to
be
with
private
access
to
content?
Address
data
I'm
going
to
keep
this
context
fairly
short,
because
we
just
got
sort
of
the
the
actual,
like
definition
out
of
out
of
juan
in
his
talk
about
why
this
is
such
an
important
proper.
Like
thing
that
we
need
to
be
focused
on,
but
but
logistically
this
is
within
sort
of
the
pl
context,
going
to
be
one
of
one
of
the
blue
funds.
A
Thinking
about
broadly
the
this
question
of
private
private
data,
retrieval
with
a
specific
interest
in
content,
address
data,
and
so
what
does?
What?
Does
that
mean?
A
We
put
together
a
hypothesis
or
sort
of
an
investment
thesis
of
why
this
makes
a
good
blue
fund
and
essentially
boiling
this
down
to
to
the
one
sentence
version.
A
We
think
that
an
increased
diversity
of
mechanisms
and
a
focused
effort
on
private
communication
is
going
to
give
us
better
options,
in
particular
on
reader
privacy,
and
so
we
got
that
definition
from
one
and
I'll
I'll
dive
into
specifically
what
that,
as
actually
meaning
right
now,
when
we
look
at
what
our
options
are
for
sort
of
private
access
to
data,
the
focus
has
not
been
on
content
address
data
right,
the
focus
for
for
low
latency
communication
and
access
to
privacy,
there's
some
on
messaging,
but
mostly
it's
web
right
and
it's
it's
the
web
2
web
of
like
a
single
origin
where
to
get
stuff,
I
need
to
go
to
that
origin,
who's,
the
authority,
who's
got
the
certificate,
and
so
in
that
focus
you
end
up
with
things
like
onion,
routing
and
mixnets,
which,
as
the
only
things
that
have
really
scaled
or
have
any
production
use,
because
they're
they're
tunneling
a
real-time
traffic
back
to
this
origin.
A
They
need
to
get
the
tls
to
keep
working
in
order
to
get
sort
of
that
end-to-end
property.
We
see
sort
of
like
this.
You
know
there's
some
success
here.
Right,
like
tor,
has
has
three
million
users
like
that.
That's
not
nothing,
but
it's
not
the
whole
web
by
a
long
shot.
Centralized
things
like
signal
have
an
order
of
magnitude
more
for
for
just
end-end
encryption.
A
When
you
look
at
getting
into
some
of
this
metadata
level,
which
is
what
onion
routing
is
trying
to
do,
the
the
signal
like
world
is
pushing
stuff
into
an
sgx
container
right.
That's
how
it's
doing
address
book
matching!
That's
how
it's
doing
attempts
at
some
of
the
protection
of
the
social
network
graph,
which
is
like
great
as
an
abstraction
right
like.
A
Let's,
let's
have
this
enclave
like
maybe
we
can
get
to
fully
homomorphic
encryption
instead
of
sgx
we're
instead
of
trusting
intel
we're
trusting
some
crypto
that
would
be
great
and,
and
the
nice
thing
about
sgx
is.
It
is
a
very
similar
abstraction
that
there's
the
the
potential
for
crypto
to
come
in
and
say.
Actually
we
can
do
this
in
a
in
a
in
a
better
way.
I
think
one.
One
of
the
you
know
interesting
things
is:
is
there
have
been
some
attempts
things
like
freenet
that
that
proactively
push
content
in
caching
they've?
A
Had
a
lot
of
liability
type
issues
of
that,
that
is
scary,
but
but
that's
really
like
the
only
thing.
That's
like
a
release
piece
of
software.
You
could
download
and
use
that's
doing
something
other
than
this
onion
routing
and
that's
an
interesting
place
that
we've
ended
on
okay,
so
so
so
the
so
that's
the
we
need
more
diversity
right
like
and,
and
so
we
can
say
that
mixed
nets
are
probably
not
going
to
solve
our
problem
like
long
term,
which
is
the
the
balance
that
a
mixed
net
is
giving.
A
You
is
like
we
can
trade
off
like
latency
through
additional
hops,
like
we
we're
laundering
our
connection
through
a
bunch
of
nodes
as
a
way
to
de-link
and
and
so
that
structure
is
just
inherently
a
more
latency
versus
privacy
and
and
that's
not
going
to
let
us
compete
with
cdns
like
if
we're
trying
to
do
content
routing
in
10
milliseconds
a
mixnet
is
gonna,
be
pretty
tough,
as
as
the
answer
to
that
and
we're
also
not
taking
advantage
of
content
addressability
in
them.
A
A
We
should
be
able
to
do
quite
a
bit
better
in
how
we
would
de-link,
given
that
there
are
a
bunch
of
copies
of
this
content
and
that
we
don't
need
to
go
to
a
specific
origin
and
we're
not
thinking
about
like
we
haven't
thought
about
how
you
would
take
a
mixnet
and
either
do
source
routing
that
understands
that
there's
a
bunch
of
places
or
that
the
the
routers
or
the
exit
could
potentially
find
the
copy
closest
to
it.
It's
still
going
back
to
the
single
origin.
So
we
haven't,
we
haven't.
A
You
know
there
is
stuff
to
do
for
how
do
you
do
better,
with
with
content
address
data
and
mixnets?
But
even
then
it's
it's
unclear.
That's
like
the
right
answer
for
especially
content
routing,
if
not
data
transfer,
and
then
also.
We
know
that
there
are
other
academic
approaches
and
we
just
don't
have
prototypes
and
we
haven't
developed
them
to
a
level
where
they
seem
applicable
today
or
something
that
we
would
consider
implementing.
A
We
know
that
that
there
are
other
agencies
that
see
this
problem
in
various
aspects
and
and
also
see
this
as
like
worth
investing
in
there's
there's
a
big
grand
challenge
from
darpa
around
fully
homomorphic
encryption
that
is,
spending
tens
of
millions
on
trying
to
get
that
within
a
factor
of
ten
or
two
like
it.
Like
the
they're
they've
identified
a
bunch
of
orders
of
magnitude
of
speed
improvement
that
they
think
they
can
get
through
a
pretty
stack,
integrated
approach
of
like
asic
design,
plus
crypto
improvements,
plus
etc.
A
That
they're
just
investing
across
that
whole
stack
and
think
they
can
really
speed
up
what's
possible
with
fhe.
A
You've
got
things
around
pir
that
have
gotten
to
tens
of
millions
or
like
roughly
what
hardware
is
capable
of,
and
that
has
sort
of
continued
to
expand.
But
there
hasn't
really
been
a
focus
there
on
on.
Can
you
make
use
of
sort
of
more
expensive
hardware
to
continue
scaling
that?
Where
did
the
the
limits
actually
end
up,
as
you
try
and
productionize
that
and
and
then
we're
seeing
across?
A
You
know
zero
knowledge
and
across
a
bunch
of
these
things,
that
the
primitives
that
we
have
are
actually
advancing
pretty
quickly
like
there's
been
several
new
bloom
filter
constructions
in
the
last
couple
years,
like
that,
like
all
of
these
different,
both
statistical
and
cryptographic,
data
structures
are
like
moving
pretty
fast.
A
So
we
have
other
options
in
sacrificing
latency
for
privacy
and
we
should
be
investigating
those.
So
what
is
the
mechanics?
The
the
fund
is
focused
on
five
years
out,
as
sort
of
like
the
goal,
which
is
like
the
the
goal
is
not
like.
Let's
make
content
routing
in
ipfs
be
more
private.
Next
year,
like
we
don't
like,
like
there's
a
bunch
of
stuff
that
pl
is
doing
directly,
that's
going
to
be
focusing
on
a
bunch
of
those
properties
like
let's
add
these
incremental
improvements,
because
those
are
interesting
and
solve
the
problems
today.
A
So
that's
going
to
be
earlier
stage
stuff.
It's
going
to
be
then
outreach
to
those
communities
to
let
them
know
that
that
we
think
this
is
a
problem
and
in
particular
that
shifting
to
also
think
about
content
address
data
is
worth
their
time.
That
gets
there
to
be
more
focused
and
more
work
there
and
then
coordinating
with
with
the
existing
funders
as
well.
To
sort
of
you
know
express
why
this
is
an
important
focus.
A
Where
are
we
now?
There
is
an
open
problem
statement
that
we're
beginning
to
put
on
pll
research.
We
are
expecting
to
fund
five
plus
academic
things
and
probably
three
plus,
like
prototype
implementation,
things
each
year
and
yeah.
So
what's
in
scope,
we've
got
sort
of
a
set
of
examples
on
the
academic
side
of
previous
work.
That
we
think
would
be
the
right
sort
of
thing
that
that
is,
you
know,
so
this
is
like
the
retroactive
like
if
we
were
doing
this
five
years
ago
or
ten
years
ago.
A
What
would
we
be?
You
know
what
would
have
been
the
great
like
home
runs
of
like
yeah,
that's
cool.
That
was
a
new
mechanism,
because
then
we
can
sort
of
say,
hey
if
you're
doing
something
that
you
think
is
like
meant
to
to
be
in
the
same
line.
These
are
the
sorts
of
things,
and
so
some
examples
here
you
know
things
like
express
is
injecting
noise
there's
the
whole
line
of
like.
A
Can
we
have
the
real
requests
be
indistinguishable
from
a
set
of
sort
of
noisy
requests,
or
at
least
statistically
express
with
zela
et
cetera?
A
We've
got
a
set
of
pir
type
things
that
that
have
been
used.
A
So
this
is,
I
want
to
get
an
item
out
of
a
database
without
the
server
knowing
which
item
there's
both
computational
things,
some
of
which
use
fully
homomorphic
style
constructions
and
then
there's
also
information
theoretic
where
you're
using
multiple
servers
and
you're
saying
they
don't
collude
that
one
still
has
some
latency
slash
some
like
overhead,
because
if
both
of
those
servers
are
in
the
database
next
to
you,
it's
harder
to
argue
that
they
might
not
collude
or
that
someone
can't
go
to
that
data
center
and
get
both
of
them.
A
There's
a
bunch
of
o-ram
and
like
oblivious
access
stuff
that
potentially
again
helps
sort
of
spread
out
and
make
it
harder
to
to
understand
what
was
accessed,
although
again
has
potentially
some
like
communication,
throughput
trade-offs
and
then
and
then
there
there's
sort
of
like
a
bunch
of
these,
like
direct
in
crypto
like
how
do
I
make
queries
that
are
hard
for
the
evaluator
to
understand?
A
A
I
guess
I
have
less
of
like
the
this
list
of
like
what
what
are
implementations
that
we
would
have
funded
if
we
were
doing
this
ten
years
ago,
like
you,
you
might
point
to
like
things
like
signal,
but
that
seems
probably
too
far
along
so
instead,
it's
we,
we
sort
of
had
what
what
are
things
that
like
we
think
we
could
be
doing.
That
would
be
interesting.
A
One
of
the
examples
is:
could
we
partner
with
nim
and
have
nim,
be
content,
aware,
like
let's
get
nim
exits,
to
understand,
ipfs,
that
that
seems
like
an
interesting
place
to
start
looking
at
researching
not
going
to
a
single
origin
but
saying
hey,
we
already
have
browsers
understanding
this
ipfs
url
format.
Why
not
have
the
exits
also
understand
it?
There's
probably
something
about
getting
the
tls
to
work
with
that
which
is
right
now,
because
the
the
requester
is
probably
going
to
tls
before
they
push
it
into
a
mixnet.
A
The
exit
just
sees
the
tls
wrapper
and
doesn't
realize
it's
an
ipfs
url.
So
there's
going
to
be
some
trickiness
to
figure
out.
How
do
you
just
not
protect
it,
but
then
you're
leaking
the
cid.
So
there's
there's
some
interactions
there,
but
it
gives
us
a
playground.
That's
that's
going
to
be
at
scale
where
we
can
start
to
understand
the
right,
caching
and
and
the
right
way
to
speed
this
up
as
much
as
you
can.
A
Another
example
is,
is
taking
some
of
the
messaging
systems
that
are
being
made
right
now
over.
You
know
onion
services
or
mixnets
and
and
asking
them
what
is
the
right
network
layer
that
you
would
actually
want
to
use?
A
lot
of
them
have
expressed
that
that
they
don't
actually
see
the
onion
routing,
as
necessarily
the
right
cryptographic
answer
for
them,
and
so,
if
you
look
at
lizard
or
switch
there's
these
sets
of
or
ricochet
it's
the
set
of
like
anonymous
messaging,
that
that
each
user
is
a
tor
onion
service.
A
And
then
they
end
up
doing
these
interesting
cryptographic
constructions
for
how
they
can
allow
offline
or
store
and
forward.
On
top
of
that
which
is
interesting.
But
then
it's
not
the
right
one,
and
they
also
realize
that
it's
not
the
right
one,
and
so
those
groups
that
are
building
that
are
interested
in
exploring
other
options
for
what
the?
A
What
that
network
level
construction
is,
and
so
that's
that's
something
that
would
be
in
the
implementation
level
to
give
them
as
practitioners,
the
ability
to
to
innovate,
cool
we've
got
and
are
continuing
to
assemble
an
advisory
committee,
and
so
the
basic
structure
is
proposals
go
in
to
an
rfp
area.
So
this
external
one
is
sort
of
a
set
trying
to
keep
this
fairly
broad
and
not
just
our
current
pl
advisors.
But
but
we
will
continue
to
expand
this.
A
There
is
a
link
to
the
pl
research
blog
that
has
the
open
problem
statement
that
will
evolve
with
links
to
the
rfp
and
places
to
submit,
as
that
comes
on,
but
already
has
sort
of
this
basic
problem.
A
If
you
have
things
you
think
are
in
scope
that
you
think
we
should
be
following
up
on
that.
You
think
are
exciting
mechanisms
that
we
shouldn't
be
forgetting
if
you
want
to
modify
this
pitch,
feel
free
to
reach
out
to
me
we're
we're
in
network
goods,
and
we
are
on
slack
and
so
forth,
so
happy
to
take
questions
and
thank
you.
A
So
we're
going
to
try
and
reach
out
to
so
pets
is
happening
this
week,
it's
theoretically
in
australia,
but
mostly
virtual
and
then
in
august
is
foci.
The
free
and
open
communication
workshop-
or
I
think
I
think,
they're
doing
hot
foci
in
august
and
then
the
real
focai
will
be
later
in
the
fall
so
we'll
reach
out.
A
We
think
those
are
probably
relevant
academic
groups
and
then
I'll
also
be
giving
this
talk,
but
with
expanded
context,
since
fawn
won't
be
there
at
ecc
next
week
and
then
we'll
we'll
be
trying
to
hit
some
of
these
practitioner
groups
as
well.
So
that's
that's
the
current
outreach
plan,
but
there's
always
more
to
do.
A
There's
a
notion,
but
I
think
I
think
we
could
probably
do
more
on
in
terms
of
public,
probably
linking
on
that
blog
post
is,
is
what
is
the
public
facing
thing
right
now
that
that's
like
the
link?
That
is
a
qr
code
and
probably
then
the
right
place,
yeah.
A
Yeah,
so
so
we've
got
this
open
problem
linked
already
as
as
a
as
an
issue
and
and
are
happy
to.
Let's
see
where
did
it
end
up
so
so
here's
here's
sort
of
this
draft
of
the
of
the
private
retrieval
open
problem,
and
so
then
we're
happy
to
take
prs
and
issues
and
so
forth,
as
as
continued
refinement
of
that
it's
it's
meant
to
be
a
living
document.
So
that's
the
current
thought
there
cool.
Thank
you
for
doing
this.