►
From YouTube: 2023-09-14 IPFS Implementers Working Group
Description
Sync of those involved with creating IPFS Implementations: https://docs.ipfs.tech/concepts/ipfs-implementations/
Notes:
- https://pl-strflt.notion.site/IPFS-Implementers-Sync-2023-09-14-eb21002dd43d40379f2aaf3eaa747095?pvs=4
- For more on the IPFS Implementers Working Group including calendar information see: https://pl-strflt.notion.site/IPFS-Implementers-Working-Group-f102a74802b34529a759ffbc3ea20303
A
Cool
well
good,
welcome
everybody
to
September
14th
implementer
ipfs,
implementers
working
group
yeah,
where
we
have
different
rtfs
folks
that
are
involved
in
ipfs
implementations
and
those
deeply
affected
by
it
to
come
together
and
you
connect
a
line,
share,
updates
Etc,
and
this
is
a
lot
of
where
specs
updates.
It's
that
also
get
discussed
as
well,
and
so
with
that
we
will,
we
will
jump
in
yeah
per
usual
just
start
to
start
start
with.
Anyone
who
wants
to
give
any
updates
on
iqfs
implementations.
A
So
on
Kubo
there's
a
023
release
coming
this
next
week.
We
expect
to
do
the
RC
on
Monday.
We
just
had
a
conversation
about
that
and
then
with
assuming
no
major
issues
found
that
then
do
the
release
Itself
by
the
end
of
the
for
the
end
of
the
week.
I,
don't
know
anything
in
particular.
You
want
to
call
out
for
this
audience.
That's
coming
not
that
they're
asked
do
I
know
we
have
a
link
to
the
release,
notes,
yeah.
B
I
think
for
people
who
want
to
set
up
host
there
will
be
experimental
opt-in
configuration
option
to
expose
existing
routing
system
for
content,
ipns
and
peers
under
the
same
port
that
we
have
for
HTTP
Gateway.
So,
just
like
you
have
slash
ipfs
slash
ipns
there
there
will
be
like
a
slash
routing
V1,
and
that
would
be
a
way
for
you
to
shut
down
the
HD
clients
with
one
kubernet
and
kind
of
like
delegate
it
to
a
dedicated
routing
cluster.
C
Yeah
and
I
guess
a
couple:
other
ones
are
deprecating
some
old
things,
so
so
mplex
and
quick
draft
order,
29,
which
are
sort
of
mplex,
is
no
longer
being
supported
by
default.
It'll
get
dropped
entirely
later,
and
the
old
quick
draft
is
not
being
supported
anymore.
C
It's
not
being
supported
in
Google,
P2P
or
quick
go,
and
so
we're
dropping
it
as
well
as
part
of
updating,
but
enough
of
the
network
has
upgraded
updated
to
quickly
one
that
should
be
fine
and
if
they
don't,
they
still
have
trustee
TCP.
So
shouldn't
be
any
problems
there.
A
C
A
Does
make
a
change
log
update?
So
that's
something
that
won't
show
up
here,
too
yep
cool,
yeah,
very,
very
good,
and
on
the
Helia
front,
there's
this
issue
just
focused
on
how
do
we
ensuring
we've
got
reliable,
retrieval
happening
in
a
browser
with
from
Helia
to
be
able
to
talk
to,
for
example,
any
modern
Cooper
know.
A
That's
providing
content
doesn't
have
to
be
cool,
but
just
using
that
as
a
example
to
be
able
to
show
end
to
end,
and
so
there's
quite
a
few
work
streams
related
to
this,
some
of
a
lot
of
which
happen
in
the
JSI.
But
there
are
back-end
things
needed
to,
but
that's
where
a
lot
of
the
Helia
development
is
going
right
now
and
we're
linking
the
specific
subtasks
from
this
kind
of
master
tracker.
A
But
you
know
that's
kind
of
that's
the
real
quick
summary
of
what's
happening
in
Helia
world
and
happy
to
talk
more
on
that
in
the
it-js,
fill
slack
Channel
or
in
the
corresponding
issues
do
do
any
other
yeah
ipfs
implementations
want
to
share
anything,
that's
been
happening
in
their
side
of
the
world,
yeah
not
required,
but
Alan
or
Brendan.
If
there's
things
you
want
to
share
love
to
hear
it,
but
your
call-
and
it
looks
like
lidl's-
got
one
he's
typing.
B
Into
yeah
I
thought
that
it's
worth
mentioning
that
we've
started
phase
two
of
Ip
of
a
scrummy
project
and
still
figuring
out
the
prioritization
and
ordering
but
tldr
it's
a
a
set
of
patches
on
top
of
chromium,
open
source
project
to
add
native
ipfs
and
ipns
protocol
handlers,
but
with
a
Twist
in
that
there's
no
lipid2p
at
this
right
now
and
it
purely
uses,
trustless,
Gateway
specification
and
just
http,
and
what
we
want
to
explore
in
this
phase
is
somehow
related
to
the
thing
that
you
just
mentioned
for
helium
that
the
phase
one
it
was.
B
You
provided
this
client
a
set
of
gateways
and
it
will
fetch
blocks
from
them.
It
was
a
proof
of
concept
to
demonstrate
it's
a
good
enough
for
a
single
person
to
browse
websites.
The
phase
two
is
to
make
it
more
robust
and
close
the
future
parity
with
existing
gateways
or
something
like
Kubo
in
ipfs
desktop.
B
So
if
the
none
of
gateways
has
the
CID,
you
want
now
that,
with
the
things,
that's
the
routing
V1,
that
will
be
shipping
the
next
couple
or
you
will
be
able
to
learn
about
providers
that
speak,
HTTP
and
and
then
ipfs
chromium.
What
we
want
to
do
in
this
phase
at
least
explore
the
possibility
is
to
leverage
HTTP
providers
and
figure
out
flesh
out
specification,
details
and
write
them
down.
So
people
can
leverage
that
as
well.
So.
D
We're
gonna
jump
in
cool
I
did
not,
and
I
did
not
get
on
the
agenda.
I'm,
sorry,
I'm
late,
nobody,
you're
good
yeah,
we're
cruising
Along
on
iro.
We
have
a
0.60
release
that
we
have
cut
two
alpha
releases
for
and
the
bigger
one
will
be
coming
out
in
at
the
end
of
September.
You
know
the
proper
r06
will
release.
This
includes
the
our
new
document,
synchronization
stuff
and
our
mobile
stuff,
I
think
notable
stuff
for
this
community.
That
would
be
fun
to
talk
about.
D
D
All
of
that
which
has
been
a
massive
lift
to
be
able
to
actually
have
your
physical
internet
vertical
address
change
and
still
maintain
a
peer-to-peer
connection
to
your
existing
peers
and
so
that
PR,
if
you
want
to
check
it,
I'll
link
it
in
the
notes,
but
all
in
all
things
are
really
cooking.
Along
on
that
side
and
we're
very
excited
about
some
of
those
bits
to
come
out,
we've
also
landed
up
a
bunch
of
new
stuff
around
downloading
prioritization
and
bandwidth,
prioritization
and
constraints.
D
So
if
you
want
to
see
any
of
that,
there's
also
you
have
a
nice
mechanism
for
queuing
and
setting
up
such
that
only
five
downloads
are
happening
simultaneously,
but
that
we
are
constantly
supplying
candidate
peers
meaningfully
to
the
background
of
that
queue,
including
retries
and
back
offs,
and
a
couple
of
techniques
for
improving
the
odds
of
choosing
the
right
tier
to
try
and
fetch
from.
If
anyone
want
to
see
more
details
about
those
happy
to
get
into
it.
But
I
I
can
link
that
in
the
notes
and
maybe
take
the
discussion.
A
second.
A
A
Okie
doke
I
guess
we'll
switch
over
to
ipip
Corner.
You
want
to
take
this
one.
B
This
will
be
like
a
short
one.
There
are
two
Ipswich
got
ratified
and
merged.
One
is
about
interrupt
between
the
path
and
subdomain
Gateway
tldr.
Is
that
all
the
extra
logic
should
happen
after
the
redirect
to
subdomain
Gateway,
because
then
you
have
a
origin
pairs
root,
CID.
The
second
one
is
the
tank
of
technical
debts
and
Legacy
of
ips
record
formats
and
ip4
to
8
got
merged,
and
it
has
updated
sections
about
how
to
create
a
pns
record
and
how
to
validate
it.
It's.
B
I
guess
like
the
gist
is
that
it
enables
people
who
don't
need
backwards.
Compatibility
with
super
super
old
nodes
that
speak
ipns
to
create
a
very
lean
records
that
no
longer
like
duplicate
the
same
values
in
C
board
and
plot
above
and
it
shipped
in
Kubo,
22
cubot
still
produces
V1,
plus
V2
the
same.
We
have
a
pull
request
in
JS
ipns
for
Helia,
and
it
the
library
will
still
produce
V1
plus
V2
by
default.
B
But
the
V2
only
is
an
opt-in
that
now
is
possible,
and
if
people
need
those
Lin
records,
they
can
create
and
use
them.
There
is
one
IP
ready
for
final
reviews.
It's
been
in
the
ratification
queue
for
a
while.
It's
a
work
by
Enrique
on
adding
peer
loading
and
point
to
the
routing
V1
for
delegated
routing
system,
which
closes
the
Gap.
Not
with
this
one.
We
will
have
future
parity.
We
can
look
up
content,
IPS
and
peer
records
and
effectively
with
Kubo
23,
be
implementing
this
ipip.
B
You
can
delegate
all
the
routing
to
a
different
node
and
shut
down
local,
like
DHD
clients,
if
you
want
probably
will
get
merged
on
Friday,
because
it
was
not
very
controversial
and
we've
been
working
on
that
for
a
while,
but
I
mentioned
it
because
it's
still
open
and
there
are
ongoing
discussions
more
on
the
board.
But
the
things
I
listed
here
are
probably
more
the
more
most
like,
topical
or
recently
updated.
B
It's
a
need
from
Project,
Spark,
I,
believe
I
think
we
found
a
way
that
will
not
create
too
many
waves
in
the
ecosystem.
While
we
are
waiting
for
car
V3,
the
second
one
is
codifying:
how
do
we
expose
trustless
Gateway
of
early
P2P
sockets
instead
of
like
TCP
socket
and
the
last
one
is?
B
How
do
we
signal
which
features
Gateway?
Has
there
may
be
a
Gateway
which
supports
specific
hash
functions
or
only
supports
blocks
and
not
cars,
and
it
would
be
nice
for
clients
to
be
able
to
quickly
probe
it
without
making
like
tests
with
various
requests
for
success
and
failure?
I
guess
that's
the
speedrun
update
on
the
ipps
I
think
there's
nothing
requiring
immediate
attention,
but
if
you
are
interested
in
any
ongoing
discussions,
the
time
to
provide
feedback
is
now
I.
Think.
E
When
he
said,
do
you
can
we
discuss
them?
Do
you
meet
on
the
on
the
issue
or,
like
is
like
now
free
time
for
doing
that
or
like
what
is
the.
B
C
I
think,
realistically,
there
are
a
number
of
asks
that
result
in
needing
like
a
car
V3,
and
this
is
like
there
is
an
extensibility
point
that
we
have
now,
which
is
that
you
can
use
the
you,
can
use
the
accept
headers
and
you
can
Define
that
I
would
like
the
data.
As
you
know,
format
equals
Car
Plus,
a
Dean's
wacky
and
you
know
wacky
inflatable
format
and
I
can
put
whatever
I
want
in
there
and
then
I
guess
people
will
understand
it.
C
As
long
as
it's
you
know,
looks
like
a
valid
car
data,
and
this
is
like,
if
you'd
like,
to
do
that
for
now.
Here
is
a
way
that
you
can
do
that,
but
I
think
realistically
we're
gonna
need
a
car
V3
thing
to
support
things
like
you
have
reached
the
end
of
your
file
or
there
is
an
error
in
sending
you.
The
data
here
is
the
error.
Type
I
don't
want
to
Traverse
anymore,
because
I
don't
want
to
give
you
this
block,
because
I
legally
have
been
told
to
not
give
you
the
block.
C
C
Those
like
the
the
current
format
just
has
like
a
number
of
failures
there
that
seem
like
they're,
not
that
hard
to
fix
together,
but
I
think
the
script
didn't
want
to
wait
on
that,
and
so
this
was
then
saying:
here's
here's
how
we're
gonna
do
it
until
that
lands
in
a
way
that
nobody
else
should
have
to
support
in
the
short
or
long
term,
and
then
hopefully,
it
fades
away.
B
Yeah,
it's
also
like
in
I
think
the
version
that
we
are
approaching
on
that
IP
is
it's
a
sensible
trade-off
in
that
it's
effectively,
the
Manifest
becomes
the
Json
at
the
end
of
the
after.
A
car
stream
make
making
it
explicit
opt-in
by
a
client
creates
like
a
control
environment.
So
the
most
important
part
is
to
not
impact
existing
clients
and
by
the
fact
that
this
is
like
an
explicit
opt-in
via
HTTP
content
negotiation.
B
We
don't
risk
creating
incompatibility
in
the
ecosystem,
at
least
not
for
the
default
side
yeah,
so
waiting
for
car
V3,
so.
E
Far
it's
the
best
there
was
like
I
think
there
was
some
talk
on
like
I
mean
you
know,
it
kind
of
sounds
fine
when
you're
sort
of
dynamically
generating
the
car
as
in
a
trustless
Gateway
but
I,
think
I,
don't
know
if
it
was
here
or
I
saw
somewhere
else
that,
like
if
you're
you're,
if
you
receive
a
car
with
this
data
in
it,
then
it
needs
to
be
stripped
out.
Somehow.
B
E
B
Why
yeah
that's
like
if.
C
You
have
a
problem
with
this
whole
scheme.
Right,
like
the
reason
we
need
a
car
V3
is
because
the
data
plane
and
like
the
control
plane
are
getting
mixed
up
right
what's
happening.
Is
you
want
to
give
me
some
metadata
about
the
car
and
you're
shoving
it
in
as
a
regular
looking
block
at
the
end
right.
C
Yeah,
so,
and
and
and
I
and
I
stole
that
particular
framing
of
the
complaint
from
from
Juan
when
he
saw
a
similar
proposal
like
seven
months
ago,
so
this
is
I.
This
is
why
I
think
this
is
like
this
is
very
unfortunate
and
I
think
also
why
we,
it
is
good
motivation
for
us
to
do
something
that
is
better
and
allows
you
know
separating
those
concerns
out
and
if
you've
used.
If
you,
if
you've
used
some
amount
of
the
specs
thing
is
like
I,
am
doing.
B
Is
it
comes
from
like
how
they
use
cars
when
someone
like
sends
you
a
car
with
a
stream
with
this
thing
at
the
end,
how
they
should
behave,
and
they
also
have
a
concept
of
car
cids,
which
is
a
hash
of
the
car
stream,
so
I
think
the
the
version
of
the
this
IP
after
a
suggestion
from
Rod
to
use
this
like
zero
length,
variance
at
the
end
before
the
Json
makes
the
makes
it
easier
for
like
for
you
Alan,
because
you
can
I,
then
you
can
simply
like
just
discard
anything
after
that
and
you
only
care.
E
B
C
I
do
know
I
curl,
I
curl,
some
and
point
I
curl
up
some
endpoints
I
get
a
car.
Then
I
go
over
to
my
I
I,
like
I
round
trip
some
data
right,
I
took
a
car
file.
I
put
it
into
that
storage
that
it
got
shipped
out
as
the
car
filed
and
I
fetched
it
and
then
like
somebody
else,
went
and
put
it
back
in
and
it's
not
even
the
same
thing
anymore,
because
it's
got
this
random
signature
blob
at.
B
F
B
Yeah,
but
you
know
if
it's
a
car,
if
we
say
it's
still
a
car,
but
it
just
has
a
manifest
at
the
end.
We
are
like
you
know
back
to
the
what's:
the
Lesser
evil.
Do
we
create
a
totally
new
content
type
and
we
risk
like
splintering
ecosystem
on
something
that
we
hope
it's
temporarily
until
car
V3
arrives
of
doing
the
the
other
thing,
I
think
it's
less
like
I
I,
elaborated
on
the
argue
that
I
think
this
is
like
a
lesser
evil
and
I.
B
Think
like
for
for
Allen,
gives
a
very
clear
signal
where
the
car
ends.
There's
no
ambiguity
is
the
last
block
manifest
or
not,
and
I
think
that's
probably
better
in
the
long
run,
yeah.
A
On
the
so
I
guess
find
this
bit
more
time.
Just
really
want
to
check.
Did
anyone
have
anything
else
they
wanted
to
add
to
the
agenda
or
bring
up,
so
we
use
the
appropriate
time
correctly,
not
saying
anything
else,
please,
please
feel
free
to
add
I
guess
I
I.
This
has
been
the
sorry.
Ipip431
has
been
framed
as
a
as
a
stop
Gap
that
you
know
keeps
the
damage
kind
of
contained
until
car
V3
comes
around
I
guess
what
are
the
natural
systemic
driving
forces
that
are
going
to
make
car
V3
actually
happen?
C
I'm
hoping
that
I'm,
hoping
that
one
of
the
one
of
the
three
driving
forces
I
mentioned
earlier,
will
will
start
propelling
people.
So
if
you
need
the,
if
you
want
the
the
large
Blake
three
blocks
in
the
car
file,
we
need
a
new
format
for
the
car
file,
so
that
would
be
one
group
that
might
push
if
you
need.
You
know.
Oh.
C
It's
it's
groups
together
right,
so
that's
one!
That's
one
push
if
you're
like
I've
noticed
that
I
keep
trying
to
serve
car
files
in
my
car
responses,
but
I
also
have
a
block
list
and
sometimes
nodes
in
the
middle
of
this
block
list
show
up
and
I
have
no
way
to
Signal
the
error
and
that
causes
the
systems
that
this
moves
through
to
freak
out.
C
Maybe
maybe
maybe
the
people
who
need
errors
want
this
right
if
you
have
HTTP
caching
stuff
in
the
middle
and
you'd
like
to
know
if
you're
stream
terminated
for
like
a
good
reason
or
a
bad
reason,
it
seems
like
a
third
group
that
would
that
would
maybe
push
on
this.
If,
if
none
of
those
groups
care
enough
to
do
this,
then
I
guess
we
are
okay.
Dealing
with
the
status
bow
right,
but
I'm,
hoping
at
least
one,
if
not
multiple
are
involved
in
care
to
fix.
A
E
D
D
F
F
So
they
would
like
to
attach
a
signature
in
some
form
at
the
end,
so
that
clients
to
boost
or
to
file
coin
storage
providers
get
that
and
can
potentially
submit
that
as
a
measurement
sort
of
token,
to
show
that
they
did
in
fact
get
it
from
the
expected
SP
not
from
someone
else.
E
A
E
You
know
what
it
was
largely
answered
for
Brendan's
or
a
Christian
yeah.
It
sounds
like
I'm
I'm
unlikely
to
have
to
implement
it.
So,
okay.
E
B
I
I
was
delegate
to
a
wall
of
text
on
the
IP
I.
Think
tldr
is
that
we
had
a
prior
art
when
we
said.
Oh,
this
is
just
a
temporary
measure
until
the
next
week
and
five
years
later,
we
still
have
things
like
preload
notes
and
the
list
of
examples
like
that
is
around,
so
it's
just
like
be
playing
it
safe
and
smarter.
This
way
it
just
in
case
we
are
stuck
with
this.
We
don't
end
up,
saying
hey.
If
you
want
to
support
ipfs,
you
need
to
implement
this.
E
D
Yeah
but
like
to
me
the
root
of
this
problem,
like
just,
if
we're
talking
about
we're
waiting
for
car
V3,
the
design
constraints
that
this
conversation
puts
on
a
an
archive
format
that
has
to
deal
with
error
responses.
All
of
a
sudden
feels
like
a
pretty
large
design
space.
That,
like
is
that
what
we're
doing
like
you
know,
that's
not.
That
seems
like
an
inherent
conflation
of
the
so
so
it
might
be,
and
the
address
format
isn't
it
like.
C
C
Be
that
it
might
be
that
what
comes
as
a
result
of
that
is
that
we
say
okay,
this
is
actually
two
different
pieces.
We
need
a
storage
format
and
we
need
a
streaming
format
and
the
streaming
format
has
these
things
in
the
storage
format.
As
those
and
I
understand
that
everybody
has
been
hoping
that
the
streaming
for
like
there
have
been
a
bunch
of
tools
that
are
trying
to
treat
the
streaming
format
and
the
storage
format
is
the
same
but
I'm.
C
Sorry,
that's
not
always
going
to
work
out
that
way
and
I
think
that's
reasonable,
I.
Think
one
of
the
proposals
around
the
streaming
errors
thing
was
like
I
think
one
was
like
you
may
consider
wanting
a
thing:
that's
just
a
car,
a
separate
car
stream
format
and
I
think
that's
like
I
think
that
could
naturally
come
out
of
that
I
suspect
it
won't
look
super
different
from
what
the
archive
format
looks
like,
except
that
maybe
you
say
like
this
type
of
metadata,
no
good
here
in
this
type
of
metadata
good
here.
A
D
It
right
like
that
is
this
is
the
thing
that's
that's
networking
Centric
right
so,
like
I,
think
I'd
I'd
behoove
us
to
think
about
the
V3
design
space
as
two
separate
things
and
not
call
it
a
streaming
or
wire-oriented
archive,
that's
a
confusing
thing,
and
so
like.
Let's
do
a
wire
protocol
cool.
We
can
totally
do
a
wire
protocol
and
then,
let's
do
an
archive
format
and
as
long
as
we
keep
those
kids
separate,
I
think
we'll.
Yes,
some
people
will
maybe
not
have
a
great
time
because
cars
over
the
wire
got
thus
far.
E
C
Although
I
think
they're,
they
I
think
I
think
the
relationship
between
them
can't
probably
can't
bifurcate
that
much
because
you're
you're,
you
may
end
up
with
questions
like
okay,
I'm
doing
I'm
I'm
doing
like
you
know,
I
want
some
like
you
know
some
index
or
I.
Have
my
my
large
block,
which
has
this
like
tree
structure.
D
Yeah,
it's
a
convenience
for
sure,
but
it
seems
like
if
I
understand
correctly
a
lot
of
these
new
requirements
really
aside
from
us
and
our
weirdo
Big
Blocks
are
really
or
is
it
on
the
wire
problems?
Is
that
fair,
like
Alan,
you
guys
seem
like
you're
having
you're
doing
a
lot
of
cool
stuff
on.
D
Yeah
I'm
just
trying
to
understand
like
what
we
should
be
focusing
on
right
like
do.
We
need
a
new
at
rest,
storage
format
or
do
we
need
to
sell
some
of
these
issues
on
how
we
add
metadata
to
responses
and
deal
with
the
burdens
of
check,
something
and
signing
which
are
hard
requirements
of
a
lot
of
our
systems?.
A
Yes,
I,
don't
know
if
yeah
other
comments
here,
certainly
welcome
I,
think
just
for
maybe
to
help
my
understanding
like
the
important
point
about
hey.
Should
we
just
maybe
accept
some
of
the
reality
that
there's
an
at
rest,
any
a
streaming
format
here
and
like
if
you
take
that
mindset
like
if
we
think
about
that,
what
what
causes
people
that
are
hitting
some
of
the
wire
issues
today
not
to
just
go
start
creating.
A
You
know
the
car
streaming
respect
like.
Is
it
just
such
a?
Is
it
too
big
of
a
lift
and
too
many
changes
in
the
tooling
to
get
something
like
that
supported.
B
B
It's
a.
We
got
the
hammer
it's
a
nail,
but
at
the
same
time,
by
by
reusing
cars,
you
reduce
complexity,
around
user
support
testing,
for
example,
we
have
Gateway
conformance
tests
that
can
make
sure
that
the
the
basic
production
use
of
car
format
for
reading
directories
files
works.
If
you
invent
a
new
format,
you
need
to
now
have
exactly
the
same
test
cases
against
these
new
format.
You
need
to
have
test
vectors
for
that,
and
now
people
need
to
implement
those
two
formats.
B
Instead
of
one
ipfs
chromium
right
now
uses
blocks,
it
may
be
fetching
cars.
It
does
not
matter
for
this
project
but,
for
example,
for
other
projects
in
ipfs
desktop.
You
have
an
option
to
fetch
an
archive
in
some
web
panel
of
a
service
that
provides
ipf
support.
You
may
also
have
a
button
says
fetch
this
dog
as
a
car
or
and
that's
the
car
that
you
can
import
to
your
node.
So
that's
the
use
case.
I
exported
the
file
and
I
got
it
back
during
that
export
over
the
network.
B
B
That's
like
a
placeholder
for
answering
the
question:
do
we
need
two
things
one,
and
if
so,
in
which
use
cases
we
would
be
using
them
at
the
end
of
the
day,
I
feel
we'll
be
still
hitting
the
situation
when
you
have
the
streaming
look
the
format
tailored
to
streaming,
but
the
user
wants
an
archive
and
error
detection
of
that
archive
figuring
out.
Oh
something
went
wrong
because
that
archive
got
fetched
through
multiple
hdp
hubs.
B
B
When
we
have
time
it
may
be
one
thing
with
maybe
two
things,
but
just
just
being
cognizant
that
there's
there
are
no
edge
cases
when,
when
you'll
be
always
mixing
the
two.
Thank
you.
F
F
The
in
the
specific
case
of
why
is
this
going
the
way
it
is
the
part
of
the
requirement
from
spark
in
order
to
make
their
thing
useful?
F
Is
that
the
boost
servers,
the
SPs
so
spark
is
the
is
the
measurement
subsystem
that
wants
this
authentication
from
Storage
providers
in
order
to
know
are
they
actually
serving
the
content
they're
supposed
to
as
part
of
a
reputation
feedback
loop
and
in
order
to
do
these
tests
and
figure
that
out
in
a
way
where
the
clients
can
subsequently,
you
know,
tell
some
third
party:
yes,
I
got
it
from
this
provider
or
no
I
didn't
and
get
and
have
a
signature
that
they
can
pass
on.
F
So
there's
this,
you
know
attestation
that
becomes
possible
over
these
requests
from
SPS.
That
are,
you
know,
under
a
protocol
that
that
provides
that
they'd
really
like
that
to
be
a
thing
that
all
clients
end
up
asking
for
and
getting
because
you
also
want
those
measurement
things
to
be
undifferentiated.
Requests
from
your
actual
real
user
traffic,
because.
E
F
It
needs
to
be
a
thing
where
that's
also
close
enough
to
a
standard
that
at
least
things
like
Lassie
and
other
clients
that
are
going
to
these
SPS
are
also
able
to
ask
for
and
understand
and
get
rid
of
that
signature,
because
that
signature
needs
to
become
sort
of
a
default
thing
that
happens
on
all
requests
to
these
SPS.
If
that's
going
to
be
a
meaningful
attestation
that
can
be
used
so
that
that's
why
it
went
into
that
standard
process
and
and
trying
to
figure
out.
A
Okay
well
and
to
rest
yeah,
there's,
definitely
interesting
conversation.
I
know.
A
lot
of
thought
has
gone
into
this.
A
lot
of
things
have
been
written
in
this
apip
I.
Guess
Brandon.
Did
you
get
your
answer
about?
What
spark
is
I
did
okay,
cool
yeah
I
just
make
sure
it
was
kind
of
yeah.
What
will
I
was
saying
very
good.
D
F
C
I
think
that
means
we
did
the
extensibility
okay
right,
if
we've
managed
to
put
it
in
a
scenario
where
someone's
been
able
to
add
in
a
thing
that
they
can
use
that
people
don't
need,
don't
have
to
support
the
call
out
to
ipft
425,
which
I
think
would
be
very
helpful.
Also
in
furthering
this
kind
of
thing.
A
So
for
yeah
and
so
for
moving
431
forward,
I,
I,
guess
I,
don't
know
if
there's
already
implementations
in
place
that
are
following
this
what's
been
outlined
here
will
but
I
guess
that's
that's
having
that
in
in
Lassie
and
in
spark
code
would
be
I
guess
what
we
would
want
to
see.
F
I
believe
I
believe
we
have
frisbee
as
a
server
implementation
that
supports
it
and
I
think
spark
client
supports
it.
I
don't
know
about
Lassie
fully
yet,
but
I
think
we're
working
on
that.
A
Great
both
thanks
everyone
for,
for
your
time,
yeah
I,
know,
I,
didn't
take
great
great
notes.
Here
I
will
we'll
get
the
transcripts
and
stuff
pasted
in
I
was
just
trying
to
focus
on
being
able
to
listen.
So
thanks
for
that,
we'll
post
the
recording
here
shortly
as
well
I
hope
everyone
has
a
good
rest
of
the
day
and
we'll
talk
more
next
time.
Ciao
for
now,
thanks
everybody
good
to
see
you
likewise,
bye-bye
foreign.