►
From YouTube: 2023-07-11 Content Routing Workgroup #15
Description
The team discusses content routing improvements to IPFS nodes including bootstrap node persistence released with Kubo v0.21, the state of deployment of the new Kubo version across the Bifrost cluster, and ongoing architectural efforts on IPNI. Additionally we address handling HTTP records for supporting multiple routers.
A
A
A
If
you
look
at
set
reps
or
the
Bedrock
metric
stocks
that
we
publish
routinely
that
documents
here,
it's
a
Google
doc
feel
free
to
leave
comments.
There
I'll
take
ownership
of
responding
with
questions
you
might
have
about
them,
but
those
are
freely
available
for
people
to
take
a
look
at
we'll
start.
A
The
meeting
off
with
our
team
updates
we've
got
a
lot
of
stuff
over
on
ipni,
so
I'll
rush
through
those
and
then,
if
the
folks
from
by
froster
ipfest
would
like
to
chime
in
and
let
us
know
what's
going
on
with
their
their
teams.
We'd
appreciate
that
as
well,
so
over
on
the
IP
and
I
side
of
the
house,
we
are
going
through
a
foundation
DB
ingestion
process
so
that
we
can
fully
replace
the
production
currently
Pebble
DV
with
Foundation
DB,
it's
at
about
65
ingestion
of
the
entire
index.
A
So
it's
working
its
way
through
kind
of
some
of
the
the
bigger
provider
data
to
be
fully
ingested
right
now
and
that'll
represent
a
snapshot
of
all
the
queryable
data
on
ipni.
That's
been
advertised
to
ipni
I
expect,
based
on
previous
iterations
of
us,
bringing
up
a
new
ipni
instance
that
this
will
probably
be
done
by
the
end
of
the
week.
A
The
count
function
that
we
represent
in
our
sit
rep,
because
the
ipni
key
value
store
is
so
large.
It's
not
something
you
can
just
simply
query
the
entire
thing
and
pull
Counts
from
also
the
ingestion
happens
very
very
fast,
so
we're
talking
hundreds
of
Records
I
think
a
second,
but
that
obviously
multiplies
very
rapidly
when
you
get
into
the
minutes
and
hours
range
and
so
counting
while
you're
ingesting
those
records
is
also
very
difficult
to
do
without
damaging
the
performance
or
creating
like
costly
operations.
A
Just
because
of
the
scale
of
ingestion
that
we
do
the
scale
of
records
that
we
onboard
and
so
our
account
function.
It
pulls
data
from
a
lot
of
different
places,
because
we've
got
multiple
nodes.
We
have
a
very
scalable
architecture,
and
so,
in
order
to
do
that,
we
kind
of
have
to
count
from
these
multiple
places
and
bring
that
data
together,
and
that
creates
a
bit
of
confusion
around
the
numbers
because
they
end
up
being
accurate
but
imprecise.
A
That
I
shared
it's
kind
of
an
ongoing
battle
that
we're
fighting
to
try
to
make
our
our
data
a
lot
easier
to
understand
for
folks
outside
the
team
that
aren't
routinely
working
with
it.
We're
also
doing
a
bit
of
cost
optimization
due
to
our
ingestion
activities.
I.
Think
those
of
you
that
are
on
the
cost,
optimization
work
group
probably
would
have
noticed.
A
We've
had
a
pretty
big
uptick
in
our
AWS
activity
as
we're
going
through
that
so
we've
kind
of
looked
at
some
of
the
redundancies
that
we
maintain
and
whether
or
not
those
redundancies
are
entirely
necessary
based
on
you
know
our
kind
of
current
backup
that
we
have
and
the
way
that
we're
you
know
launching
indexes
I
think
we've
probably
pruned
some
overly
cautious
stuff.
That
was
there
in
our
infra.
A
A
We
want
to
maintain,
obviously
the
the
right
amount
of
defensiveness
with
our
infrastructure,
but
simultaneously
only
using
what
we
absolutely
need,
and
so
it's
kind
of
a
constant
balancing
act
that,
thankfully
the
folks
on
our
team
are
pretty
good
at
doing.
But
it's
always
an
effort,
so
we've
been
working
on
that
and
then
one
thing
I
wanted
to
point
out
to
y'all
is
Mossy
was
kind
enough
to
work
with
the
Boost
team
to
get
a
dashboard
together.
A
That
gives
us
a
little
bit
more
visibility
into
the
uptake
of
boosts
across
the
storage
provider
Network
for
filecoin,
which
is
really
neat.
Actually
it
lets
us
kind
of
understand
when
boost
is
updated
like
how
broadly
adopted
that
new
version
is,
and
that's
really
beneficial
to
the
ipni
team,
because
for
those
of
you
that
aren't
familiar
with
Boost
as
an
operation,
it's
kind
of
a
package
deal
with
lotus
that
lets
storage
providers
manage
their
filecoin
deals.
A
That
I
wanted
to
call
out,
which
is
these
hay
fill,
which
is
our
API,
that
we
well
it's
a
service
that
we
have
that
can
be
exposed
via
API,
which
looks
at
snapshots
of
filecoin
and
lets.
You
do
some
kind
of
queryable
wearable
actions
that
we've
associated
with
a
an
API
endpoint.
You
can
look
at
so
there's
some
new
stuff.
You
can
look
at
Via
this,
which
is
you
can
look
at
the
known
quantity
of
known
storage
providers.
A
You
can
get
the
quantity
of
information
stored
and
list
them
by
their
peer
ID
and
there's
some
examples
of
this
in
this
link
here
go
take
a
look,
there's
actually
quite
a
bit
of
other
stuff.
That's
been
added
that
exposes
details
about
boost,
and
it's
just
a
lot
of
really
neat
and
helpful
stuff
to
folks
that
are
trying
to
look
at.
A
What's
going
on
at
any
given
point
on
filed
one,
that's
the
ipni
I
would
say
a
week
and
a
half
to
two
weeks
in
a
nutshell:
there's
been
a
lot
of
other
stuff,
but
it's
all
very
granular
and
I
would
say
mostly
internal
operational
stuff
that
we've
been
wrestling
with,
but
I'd
like
to
pass
the
mic
over
to
it
looks
like
ipfs
has
an
update.
Little
did
you
want
to
jump
in
and
add
some
color
to
that
one.
B
B
If
you
turn
on
your
node
every
so
often,
then
you
you
will
still
be
able
to
get
to
the
network,
even
if
say
like
the
bootstrap
nodes
are,
are
down
or
inaccessible
for
your
region
or
you
know,
DNS
got
black
hole
or
any
of
the
variety
of
things
that
could
possibly
go
wrong
when
you
have
like
five
nodes
or
so
serving
as
bootstrap
nodes
might
be,
that
other
groups
interested
in
in
running
routing
networks
like
this
have
have
similar
needs.
So
that's,
that's.
A
B
B
C
D
B
D
B
A
I
think
regarding
the
topic
of
ambient
Discovery
and
I
won't
dig
us
too
deep
into
this
hole,
but
it's
one
thing
that
I
want
to
keep
just
kind
of
at
the
front
of
people's
minds
just
because
we're
a
little
bit
delayed
in
the
sense
that
we're
not
working
towards
a
lot
of
new
features.
This
quarter
on
ipni
but
I
suspect
that
we
will
achieve
a
lot
of
those
features
but
demand
ambient
Discovery
pretty
quickly.
A
Once
we
start
to
focus
on
it
again
and
so
I
think
we'll
will
potentially
be
in
a
scenario
where
we
approach
the
need
for
ambient
Discovery
much
more
quickly
than
we
may
expect.
It'll
it'll,
sneak
up
on
us
and
I
want
to
make
sure
we're
all
kind
of
aware
that
it's
coming
down
the
pipe
we
want
to
keep
pushing
for
it
bifrost
team.
Would
you
like
to
share
an
update
with
us.
E
So
I
saw
the
the
thread
that
you
were.
You
were
talking
with
George
about
what
metrics
not
not
appearing.
This
is
what
I've
added
to
the
versions
that
we
are
currently
running,
at
least
as
far
as
our
resource
control
is
concerned.
So
I
am
assuming
that
that
is
current,
basically
yeah.
We
should
be
updating
to
0
21
over
the
next
few
days,
probably
at
least
from
what
I
saw
about
George.
E
A
A
And
sharing
that
information,
the
the
thread
that
that
we
just
mentioned
just
so
everybody's
kind
of
on
the
same
page
was
a
request
for
feedback
on
the
versions
of
Kubo
which
are
running
on
the
bifrost
Clusters.
And
so
there
is
a
work
effort
underway
that
we
don't
really
have
a
timeline
established
right
now.
I
think
unless
that's
changed
over
the
last
two
weeks.
A
Let
me
know
if
I'm
wrong
on
that
by
Frost
team,
but
in
order
to
add
this
as
a
metric
that
could
be
viewed
in
a
dashboard
and
then
potentially
to
recognize
I.
Think,
loss
of
traffic
is
another
thing
that
we
were
looking
for
is
like
some
alerting
around
that
from
gateways
to
ipni
from
outside
the
ipni.
A
F
F
A
A
All
right
so
I
threw
a
few
topics
up
here,
the
first
of
which
is
kind
of
an
ongoing
discussion
that
has
been
going
on.
That
I
think
we
should
kind
of
wrap
up
like
what
our
path
of
action
is
going
to
be,
but
there's
been
an
ongoing
discussion
which
was
HTTP
records,
ludel
brought
up
to
the
team,
ipip0388
routing,
HTTP
API
support
for
querying
multiple
routers.
A
So
there's
a
spec
linked
to
that
description.
There
that
you
can
take
a
look
at
when
we
last
left
off.
We
had
a
meeting
to
discuss
whether
or
not
we
were
going
to
do
this
and
the
magnitude
of
the
impact,
I,
think
and
I.
Think
monster
had
proposed
during
that
discussion
that
we
come
check
out
the
the
magnitude
of
these
lookups.
A
This
is
going
to
be
a
little
bit
difficult
now,
because
everything
we've
got
is
basically
critical.
This
quarter,
and
if
this
can
wait
till
Q4,
that's
probably
going
to
be
a
little
bit
easier
to
approach.
But
if
it's
something
that
needs
to
really
be
done
in
Q3,
even
I
think
we
should
understand
that
now
and
then
try
to
wrestle
with
that
urgency.
A
B
So
I
I
guess
the
short
version
is
like
I
would
expect
that
the
amount
of
stuff
to
be
done
to
support
this
is
pretty
low.
Given
that
either
you
know
we
we
could
send
PRS
for
the
things
that
Implement.
You
know
most
of
the
like
any
any
of
the
changes
here,
but
also
because
things
about
like
oh,
how
much
traffic
are
we
going
to
serve
and
whatever
are
are
somewhat
Mo,
given
that
this
is
the
status
quo.
B
So
this
is
there
it's
in
use
if
people
are
plugging
it
into
Helia,
because
they
don't
want
to
do
DHT
crawls
in
a
browser.
This
is
what's
already
happening
so,
like
the
the
status
quo
is,
gonna
is
Gonna.
Keep
On
keep
on
doing
its
thing
unless
we
want
to
make
it
change,
and
so
we
have
to
like
change
away
from
the
status
quo.
If
that's
what
we
want
to
do
and
then
otherwise,
this
is,
if
it's
just
like
changing,
you
know
around
a
little
bit
of
like
okay.
B
A
D
The
only
thing
that
comes
to
mind
is
you
know
the
the
current
Cascade
in
ipni
always
looks
up
ipni
in
addition
to
something
else
right
and
then
the
main
rationale
for
that
was
to
stop
people
from
using
sit.contact
to
only
look
up
DHT
so
that
you
you.
It
always
includes
life,
because
you're
calling
an
iPhone
is.
Is
there
something
in
the
proposal
or
do
you
think
we
need
this
type
of
protection
in
the
API
designs
in
the
dedicated
one.
B
If
I
ask
someone
if
I,
if
I
ask
you
know,
if
I
ask
George
to
do
the
work,
is
he
doing
it
as
well
as
if
I
did
it
or
like
good
enough,
and
if
I
can't
ask
that
question,
then
I
sort
of
have
I'm
sort
of
just
guessing
or
I'm
or
I'm
over
asking
I'm
like
you
know
what
provider
records
I
just
need
to
get
enough
of
them.
Ideally
they'd
be
the
the
best
records
in
a
good,
sorted
order,
but
I'm
willing
to
settle
for
getting
any
of
them.
So
I
get
my
data.
D
B
D
B
Think
I
don't
know
if
they're
a
separate,
so
I
don't
know
if
this
is
quite
this,
the
same
thing-
maybe
it's
not
the
HTTP
over
this,
the
line
about
HTTP
over
lip
P2P
is
fine,
but
pure
HTTP
implementations
is
up
for
addressing.
Is
that
specific
to
this
ipip
or
is
this
this?
Is
this
like
a
separate
ipni
thing?
This.
A
Is
the
discussion
outside
of
the
I
I
pulled
kind
of
the
relevant
points
from
the
previous
discussions
we've
had
so
we've
like
kind
of
talked
about
it
twice
and
I
tried
to
kind
of
gather
what
stood
out
to
me
as
like
important
details
from
those
discussions.
I
think
the
TLs
or
thing
is
kind
of
that's
Aaron
from
what
we're
discussing
now.
Oh,
what
we're
here.
B
A
Don't
I
don't
know
that
this
evolved
from
the
state
of
really
like
cascading
vhd
queries
being
like
a
desirable
Behavior.
So
much
as
a
like
I
mean
we've
been
talking
for
a
while
about
how
do
we
deprecate
cascading
and
stop
performing
that
that
function,
which
this
is
kind
of
a
another
direction,
to
take
right.
D
A
separate,
I
agree:
I
think
this
is
a
separate
thing.
There
are
two
separate
things
here:
one
is
just
API
over
cascading,
the
other
one
is
just
semantics
of
Transport
level.
How
do
you
handle
TLS
in
the
P2P
world
right
going
back
to
your
question?
I
didn't
pretty
much.
If
it
has
https,
then
just
ignore
the
prid
and
go
go
fetch.
The
thing
that's
how
it
works.
B
B
I
I
guess
I
have
I,
have
sort
of
like
two
concerns
with
the
current
state
of
like
the
the
fact
that
prids
are
sort
of
like
mandated,
and
then
we
ignore
them.
One
is
that
the
P2P
folks
have
a
proposal
for
enabling,
like
authentication
of
the
server
with
a
peer
ID,
and
so
now
you
sort
of
won't
have
a
wet.
You
won't
have
a
an
explicit
signal
to
know.
B
D
I
agree,
I
think
I
think
largely.
This
is
just
lack
of
specification
just
because
it
hasn't
been
needed
not
because
we
missed
something
and
lip
P2P
right.
So
I
think
there's
a
there's,
a
bunch
of
work
here
for
lip
TP
Forks
to
figure
this
out
and
I.
Think
they're
in
the
best
position
to
figure
it
out
is
in
excellent
hands.
D
As
for
the
ipni
specific
stuff,
it
goes
and
reaches
out,
but
in
ipni
messages
that
so,
but
then
in
the
message
that
I
read
back,
there's
expectations
that
there
is
peer
ID
and
you
know
their
signature,
that
gets
verified,
verified
and
there's
a
whole
I
think
three
different
variations
of
it
across
different
apis.
So
it's
like
a
you
know:
peer
ID
is
respected,
but
indirectly
so
going
back
to
the
main
problem.
D
I
think
the
main
issue
to
point
out
is
that
there
is
a
need
for
Liberty
to
be
specifications
on
handling
of
HTTP
over
lipitupi
I.
Think
Marco
is
already
on
it.
He's
doing
a
lot
of
work,
I've
seen
a
little
PR,
so
I
I
would
keep
an
eye
on
that
to
see
how
it
evolves
and
how
we
can.
Then
retrofit
it
to
everything
else,.
B
A
All
right
so
I
think
the
the
summary
there
is
is
that
this
is
important.
We've
got
like
pretty
significant
adoption.
It's
also
dependent
on
work
that
the
P2P
team
is
doing.
For
now.
It
sounds
like
on
the
IP
and
I
said.
We
probably
agree
that
this
isn't
like
a
major
undertaking,
so
maybe
our
commitment
from
the
IP
and
I
side
will
be
that
we
we
try
to
sneak
this
in
along,
along
with
some
of
our
other
work,
where
we
see
fits
once.
C
B
D
Already
have
the
adder,
so
there's
nothing
stopping
you
from
having
PID
inside
of
that,
but
we
also
have
an
ID
right.
I.
Think
we
I
think
there
is
some
negotiations
that
happen
here
which
are
non-technical.
In
my
opinion,
they're
just
mostly
specification,
because
the
main
thing
we
care
about
is
not
to
break
people
right
and
that
I
think
should
come
from
the
P2P
folks
happy
to
get
involved.
I
think
Marco
attacked
me
in
a
in
a
HTTP
spec
thing.
So
probably
we
should
talk
to
hdb
people.
D
B
D
D
D
A
My
response
to
that
is
that
it's
not
something
we
would
get
to
in
Q3
I
think
we
actually
don't
have
a
backlog
item
for
this
one
specifically
right
now,
but
I'm
going
to
add
one
but
I.
Don't
have
any
any
expectation
that
we'll
be
able
to
get
to
this
in
in
Q3
I
think
we
can
think
about
the
problem
a
bit,
but
I
don't
know
that
we're
gonna
have
the
resources
to
approach
wrestling
with
that
I.
C
Did
want
to
I
mean
I,
guess
I
guess
torpin.
The
other
question
is:
is
that
something
that
is
an
ipni
team
action
item?
So
maybe
that's
a
discussion
of
who
who
owns
that,
because
that
client
library
was
something
that
the
stewards
and
the
the
Kubo
developers
wanted
to
do
for
the
non-double
hashed
variant.
So
I
guess
it
is
unclear
to
me
that
that
is
a
thing
that
immediately
lands
on
whether
the
ipni
team
has
bandwidth.
A
A
B
I
mean
I.
Guess
you
still
get
something
out
of
the
double
hashing
work
that
that
you've
done
because
I
guess
you're,
you
know
you've
you've
you've
made
a
commitment
and
said
you
got
contact
to
say
we're
not
going
to
store
all
the
plain
text
records
but
you're
still
getting
all
the
plain
text
queries
so
I.
D
B
C
Think
that
I
guess
that
the
thing
I'd
say
is
I,
don't
have
a
clear
sense
of
when
you
expect
the
double
hash
DHT
work
to
land,
but
that's
that
refactor
in
boxo
of
understanding
how
to
do
double
hashing
in
general.
In
boxo,
as
you
have,
these
two
different
implementations
likely
then
drives
what
the
concrete
implementation
is
for,
how
those
get
why
interleaved
and
when.
B
C
But
are
you
going
to
do
it
in
parallel?
Are
you
going
to
do
double
hash
and
then,
if
you
think
you
get
it
you're
good
and
you
have
some
delay
of
I.
Don't
have
the
results
so
I'm
going
to
now
go
back
to
the
non-double
hash
non-private
one.
Is
that
bad?
So
it's
a
config
option,
and
so
we
just
have
it
as
a
config
thing
of.
Do
you
fall
back
or
not.
B
B
A
function
like
right
now
right,
if
we
did
this
I'd,
be
like
okay.
Cid.Contact
is
a
thing
that
supports
the
double
hashing
right.
Therefore,
if
I'm
just
sending
requests
to
city.contact
just
use
double
hashing
and
then
fail,
if
otherwise-
and
that's
like
that's,
the
status
quo
does
this
anyway,
so
it's
basically
like
is
there?
Is
there
an
options?
Is
there
some
way
for
me
to
discover
so
that
you
know
this
set?
This
ipni
instance
supports
double
hashing.
If
so,
use
it
if
not
go
away,
and
things
only
become
more
complicated.
B
D
So
inside
Kubo
what
you're,
imagining
I
didn't
just
just
so
I
understand
it's
basically
another
routing
system
which
implements
delegated
routing,
but
internally
does
double
hashing,
and
then
you
just
have
multiple
routers,
just
like
we
do
now,
which
is
DHD
and
c.contact
would
have
say
that
contact
double
hash
to
something
like
that.
Is
that
what
you're,
imagining.
B
Yeah
to
some
extent,
and
then
you
know
the
order
in
which
you
query
things
and
how
long
and
whatever
will
have
the
same
set
of
games
as
now,
which
is
like
do
I
want
to
wait.
Do
I
want
to
do
the
DHT
and
and
city.contact
in
parallel,
or
do
I
wait
if
I
had
multiple
IP
like
delegated
routers
first
for
ipni
would
I
query
them
all
in
parallel
when
I
do
them
one
at
a
time
you
have
the
same
set
of
games.
D
Okay,
so
I
guess
there
are
three
different
things
that
I
see.
One
is
the
specification
for
delegated
routing
for
double
hatched
records.
Third,
one
is
the
implementation
of
the
client
for
that
HTTP
specification
and
the
other
one
that
is
in
there
is
the
DHT
implementation
of
it
and
I
think
we
want
all
of
those
to
work
in
Harmony
in
that
for
the.
B
B
We
start
adding
support
for,
like
you
know,
BitTorrent
records
or
or
the
things
that
like
number
zero
is
working
on
inside
of
Kubo
I
have
no
guarantee
that
they're
going
to
do
double
hashing
or
that
they'll
do
it
in
the
same
way,
and
that's
okay,
because,
like
we're
just
trying
to
get
some
records
and
move
on
and
the
client
has
to
make
some
educated
choices
around,
you
know
where
they're
willing
to
make
their
privacy
versus
latency
trade-offs.
Yep.
D
D
B
A
I
mean
I
I
think
even
in
this
scenario,
probably
this
feels
to
me
like
something.
We
would
try
to
start
hacking
at
potentially
in
Q4.
If
we
were
to
assuming
that
the
implementation
itself
would
actually
live
with
ipni,
it
sounds
like
I
mean
we
really
do
need
to
nail
down
like
whether
or
not
this
would
be
I.
A
I
think
this
this
feels
like
a
Q4
thing
to
me
just
based
on
how
busy
we
are
right
now,
but
I'll
I'll
talk
to
Steve
about
that
too,
since
he
brought
up
this
question
so
I'll
follow
up
with
him,
but
I
agree
on
this
back.
Will
that's
a
good
call
I,
don't
think
anything
should
stop
us
from
hacking
at
specs.