►
From YouTube: Kubernetes WG IoT Edge 20230712
Description
July 12, 2023 meeting. Open discussion of upcoming edge related conferences, efforts to define edge, AI and ML for edge. WebAssembly related to edge
A
Hi
welcome
to
the
July
12th
meeting
of
the
cncf
iot
edge
working
group.
Last
I
looked.
We
have
a
kind
of
light
agenda
today,
but
after
we
covered
the
items
in
there
now,
which
I
think
are
primarily
covering
upcoming,
open
source
and
Edge
related
conferences,
we'll
move
on
to
just
freeform
birds
of
a
feather
discussion
of
anything
anybody
wants
to
talk
about,
including
interesting
things,
you've
seen
lately
or
questions
you
have
were
things
you'd
like
to
open
for
group
sharing
and
discussion.
A
A
A
B
B
Those
working
on
cloud
native,
apps
and
development
for
the
Telco
space
might
be
worth
just
looking
at
the
agenda
as
well
as
it
develops
to
see,
what's
happening
in
that
space
and
it
is
I
think
it's
to
some
degree,
concurrent
with
kubernetes
on
edge.
So
I
just
wanted
to
point
that
out
to
folks
we're
considering
that
and
we've
been
a
part
of
it
in
the
past,
and
it's
been
fairly
successful
for
us.
So
one
of
the
groups
is
to
be
aware
of
that
upcoming
deadline
as
well.
A
Okay,
great
thanks
Brandon,
just
FYI
the
I'm,
assuming
they're,
going
to
enable
web
assembly
day
as
well,
and
there
has
been
a
fair
amount
of
overlap
with
webassembly
and
Edge
as
well,
and
all
of
those
events
are
on
the
same
day
it's
one
registration
and
you're
free
to
look
at
the
schedule
and
shift
back
and
forth
between
rooms.
The
other
thing
that
I've
noticed,
particularly
in
Europe,
is
it
sold
out
very
early,
so
they
have
a
limited
number,
a
limited
amount
of
space
for
those
day.
A
One
events
that
shared
across
all
the
events
I
think
that
it's
in
the
range
of
holding
2000
to
3
500
people
and
in
Europe
it
sold
out
before
they
even
had
the
schedule
up.
So
people
went
and
paid
for
those
extra
cost
registrations
before
they
could
even
see
fully
what
they
were
getting
into,
but
it
still
sold
out
so
I
kind
of
expect
that
that
is
likely
to
continue.
Historically,
the
North
America
conference
has
had
higher
attendance
than
the
Europe
one.
A
So
if
you're
interested
in
going
as
an
attendee,
you
might
want
to
get
your
registration
in
early
or
clear.
The
process,
if
you're
with
a
bureaucratic
org
that
requires
pre-approval
for
conference
attendance,
go
get
that
in
the
pipeline
for
your
management
to
approve
early.
So
you
don't
get
left
out.
C
C
D
Yeah
I
just
did
want
to
chime
in
I'm
chair
of
webassembly
day
and
also
a
chair
of
waslam
Khan.
So
if
there's
interest
I
know,
there
is
a
lot
of
overlap
between
those
conferences
and
even
Telco
con
web
to
somebody's
just
bleeding
into
all
of
the
related
deployment
areas
from
wasam
Khan.
Cfp
is
closed.
We
got
almost
100
submissions,
which
was
incredible
for
a
first
time
conference
it'll
be
in
Bellevue,
September,
6th
and
7,
and
then
on
the
8th,
we're
actually
doing
a
bike
code.
D
Alliance
hackathon,
where
we'll
work
on
componentizing
the
world.
The
big
thing
that's
launching
is
component
support.
So
Think,
like
plugable
Lego
blocks,
including
language
support
for
JavaScript,
go
python
as
well
as
rust,
and
then
all
these
statically
compiled
languages
of
course
are
supported
as
well.
C,
C,
plus
plus
all
that
stuff.
So
it
should
be.
A
lot
of
fun.
Tickets
are
available.
D
Now
I
we
may
sell
out
I,
don't
know
we're
planning
on
around
400
for
the
main
was
of
con
and
then
200
for
the
hackathon
on
the
following
Friday,
but
all
are
welcome,
just
like
with
Telco
day
and
The
Edge
kubernetes
on
edge
day.
Cfp
is
still
open
for
wasm
day.
So
I
there's
another
yet
another
form
where
you
can
submit
talks
to
there
as
well.
D
A
D
I'll
put
it
there
for
wasm
Khan
we're
working
through
all
the
ratings
now
and
there's
still
a
little
room
for
invited
talks.
There's
a
few
folks
that
we're
trying
to
get
on
the
schedule
like
the
Chrome
v18,
we
think,
would
be
great
to
have
on
there
and
a
few
others.
So
if
you
have
something
that
you
feel
is
especially
relevant,
a
ping
me
and
I
can
socialize
it
across
the
cfp.
Reviewers.
E
A
A
So
historically,
we've
given
talks
on
just
broad
coverage
I
would
think
the
things
that
would
be
valid,
for
that
would
be
to
cover
the
white
paper
draft.
We
have
underway
also
that
spreadsheet
that
I
started
on
op.
You
know
a
summary
of
all
the
open
source
Edge
related
projects.
We
could
perhaps
give
a
talk
on
that
or
combine
the
two
and
any
perspective
thing
that
this
group
wanted
to
take
on
as
a
group.
It's
it's
really
for
kind
of
group,
Centric
I
activity,
or
we
have
done
just
broad
coverage.
A
A
I
think
the
attitude
is
in
terms
of
speakers
like
I,
say
the
actual
form
has
to
be
filled
out
by
leadership,
but
they
don't
have
to
be
even
one
of
the
speakers,
although
I
think
generally,
it
has
been
that
one
of
the
leaders
would
be
on
there
and
anybody
else
who
joins
I.
Think
generally,
as
a
reviewer
myself,
you
look
for
people
who
are
speaking
to
have
been
actually
active
in
the
group.
You
don't
want
somebody
who
never
joins
a
meeting
and
just
came
in
out
of
nowhere.
F
A
A
It's
not
an
issue
if
you
have
one
speaker,
but
as
soon
as
you
get
a
group
of
speakers,
they're
going
to
look
at
the
the
mix
of
who
is
there
and
a
number
of
people
have
worked
on
that
white
paper
graph,
but
I
don't
think
it
will
be
viable
to
have
the
whole
group
show
up
it's
yeah.
You
know
you'd
be
looking
to
compose
something
that
would
extend
up
to
say
three
people.
Frankly,
I
think
those
sessions
are
only
30
minutes.
30.
A
E
A
E
A
Could
and
then,
then,
maybe
you
could
have
four
there's
additional
rules
on
panel
discussions
by
the
way,
or
at
least
there
have
been
historically
so
a
panel
discussion
might
work
the
other
things
that
favor
acceptance.
You
know
they.
Certainly
they
look
at
diversity
as
a
criteria
for
Acceptance
in
my
experience,
but
there's
also
a
real
desire
to
get
user
talks.
Those
are
typically
viewed
as
underrepresented
at
the
kubecon
events,
so
that,
if
you
were
to
do
a
panel,
I
think
that
yeah
and
expect
to
get
it
accepted.
A
A
A
G
But
there
was
already
a
discussion
earlier
on,
like
there
is
a
set
of
different
constraints
at
the
edge
some
of
our
Network
related.
Some
are
data
related,
some
are
resource
related
and
some
are
security
or
policy
related
and
would
be
good
to
go
and
list
those
and
well.
There
is
now
a
small
section
in
the
document
that
does
exactly
that.
G
I
separated
that
out
into
a
section
at
the
very
end
of
the
document,
so
that
we
have
a
clean
part
at
the
very
beginning
and
then
a
section
that
is
currently
called
draft
where
text
that
is
not
fully
complete,
where
people
just
provided
a
skeleton
or
the
likes
or
topics
were
just
under
discussion
that
this
part,
like
that
needs
clarification,
is
clearly
separated
from
the
the
main
body
of
the
text.
So
those
were
the
two
quick
updates
and
from
the
the
chat
with
with
Brandon
earlier
on,
he.
G
He
proposed
that
we
kind
of
chew
on
this
a
little
further
and
then
maybe
do
a
proper
review
on
the
in
the
next
call
and
I'm
not
sure
what
do
you
think
from
a
maturity
perspective?
Unfortunately,
we
don't
really
have
everybody
here,
like
troll,
couldn't
make
it
Brandon
had
a
draw,
but
do
you
think
that's
feasible.
A
A
lot
depends
on
people
showing
up
to
do
that
final
review,
and
just
as
an
aside
since
Andy
brought
up
the
possibility
that
we
would
talk
about
this
in
a
maintainer
correct
session.
I
think
that
the
way
to
do
that,
if
it
goes
for
maintainer
track,
is
it
should
be.
That
would
be
a
good
place
to
announce
the
publication
of
it
in
final
form.
A
In
my
opinion,
which
means
it
has
to
be
in
final
form
and
technically
for
these
talks,
you
are
supposed
to
have
your
slide
deck
and
things
submitted
several
weeks
before
the
event.
My
observation
has
been
that,
if
you
ignore
all
of
those
demands
from
the
Linux
Foundation
emails,
they
don't
kick
you
out
of
the
conference
and
I've
seen
plenty
of
examples
where
people
get
away
with
it,
but
you're
supposed
to
actually
have
it
done
in
advance
of
the
conference
not
day
of
so
I.
A
B
A
G
G
A
A
A
Let's
try
to
get
a
review
pass
done
either
in
the
meeting
or
sidebar
in
a
supplemental
meeting,
then
take
it
to
something
where
you
would
allow
offline
voting,
just
put
a
link
to
a
survey
or
something
and
send
yeah
make
sure
that
everybody
whose
name
is
in
there
as
a
reviewer,
gets
invited
to
just
give
a
thumbs
up
approval
or
a
thumbs
down.
It's.
A
And
that
way
you
can
record
it
and
when
you
get
to
the
pr
process,
where
we're
going
to
check
it
in
under
the
Masthead
of
the
cncf
structure
of
working
groups,
I
think
it's
best
to
be
able
to
say
in
that
PR
that
you
had
this
survey
of
the
reviewers
and
here's
the
results
of
the
poll
that
you
know
not
just
one
person
who
puts
in
the
pr
just
does
it
on
their
own
name
that
you.
You
indicate
that
there's
actually
a
widespread
consensus
for
this
thing
to
be
published
and
I.
A
A
A
G
A
Okay
and
then
Andy,
if
you
want
to
get
this
into
a
maintainer
track,
we
we.
We
should
talk
about
that,
because
I
think
we
got
to
get
that
submitted
and
maybe
we
could
do
a
maintainer
track
where
you
know.
We
share
this
with
coverage
of
that
list
of
Open
Source
projects
or
something
so
that
this
doesn't
have
to
fill
the
whole
thing.
If.
A
Direction
of
a
panel
I
think
that
you
have
to
identify
the
panel
participants
in
that
submit
and
at
this
stage,
having
gathered
panel
participants
in
the
past,
man
I
think
that's
a
big
ask
to
get
to
assemble
a
panel
in
less
than
a
week
and
get
the
people
you
have
to
have
their
names,
their
emails,
their
permission
that
they're
actually
going
to
be
there
and
I.
Think
that
would
be
a
challenge.
But
if
somebody
else
other
than
me
wants
to
take
on
assembling
that
I'm
not
against
it.
E
E
E
A
Yeah,
it
might
be
that
it's
in
a
separate
category
but
I,
don't
know
you,
but
you
most
definitely
should
look
before
you
make
an
assumption
that
you'd
be
the
one
doing
it
because
it
there
is
definitely
a
quota
and
there's
a
hard
limit
that
is
published
and
then
I.
Think
practically
speaking,
the
program
committee
tries
to
favor
a
broader
array
of
speakers
coming
in
the
conference
so
that
even
below
the
limit,
it
probably
reduces
your
acceptance
of
getting
an
accept
if
you've
been,
if
you're
already
speaking
in
some
other
session,
yeah.
E
C
G
C
G
A
D
Yeah
Andy
were
all
your
submissions
to
kubecon
General,
or
did
they
include
some
of
the
the
days
I?
My
experience
in
the
past
has
been
that
that
not
a
way
around
the
limit,
but
that
the
days
are
all
counted
separately.
So
we
had
a
speaker
previously
from
our
team
that
spoke
in
two
different
kubecon
sessions,
as
well
as
three
different
days
on
the
same
day.
D
However,
I'm
unclear
for
wasmcon
we
just
switched
over
to
sessionize,
which,
as
a
cfp
Reviewer
is
amazing.
It
was
just
a
real
easy
tool
to
use
it
gives
you
Consolidated,
metrics
and
Reporting
all
baked
in
you
know,
if
you've,
if
you've
done
a
bunch
of
cfp
work,
I
I
was
told
that
that
kubecon
will
be
switching
to
sessionized
as
well.
A
A
A
A
F
So
I
found
some
definition
of
posting
that
in
the
channel
in
a
slack
channel,
just
anyone
aware
of
what's
going
on
in
the
like
the
overall,
the
whole
ecosystem,
not
not
only
cncf,
but
about
like
what's
being
worked
on
when
it
comes
to
Edge
but
I'm,
aware
of
like,
for
example,
the
RF
Edge
have
some
discussions,
but
they
actually,
it
looks
like
Earth.
Edge
is
not
that
active.
Actually,
maybe
just
the
slack
channel
is
not
that
active,
so
I
don't
know.
F
A
Okay,
I'll
cover
some
of
those
and
I'll
invite
others
to
speak
too,
but
in
terms
of
activity
of
Elliott,
said
Shin
slack.
If
you're
looking
at
the
cncf
slack
yeah
it's
quiet
there
because
they
have
their
own
slack,
so
they
have
an
independent
slack,
that's
different
from
the
slack
where
cncf
and
kubernetes
lives,
and
that
said,
I
think
it's
relatively
quiet,
except
for
individual
projects.
So
I
think
that
slack
is
it's
is
actually
called
LFA
LF
space
edge
and
then
they
have
a
channel
for
individual
projects
in
general.
A
The
activity
within
those
individual
projects.
It
varies
quite
a
bit
where
some
of
them
are
ghost
towns
and
some
of
them
have
multiple
posts
per
day.
The
projects
I
don't
know.
I
just
opened
the
slack
myself
to
look
because
I'm
in
there,
but
the
channels
I
joined
were
a
cradle,
a
crano
badal
Eve,
the
Fido,
which
is
a
secure
device
on
board
fledge
general
home
Edge,
open
Horizon
Etc.
So
you
could
go
there
for
specific
LF
Edge.
An
elephant
does
publish
a
definition
of
edge,
so
does
eclipse
foundation.
A
So
do
the
white
papers
that
this
group
has
put
out
there's
a
lot
of
intersection,
but
the
bottom
line
I'm
getting
kind
of
cynical
on
these
Edge
definitions,
because
to
me
I've
come
to
the
conclusion
that
definitions
are
not
they.
They
really
aren't
something,
that's
right
or
wrong.
There's
something
that's
useful
or
not
useful
and
useful
depends
on
the
context
you're
in
so,
if
you're
in
a
specific
industry
like
Telco
or
something
like
that,
using
a
definition
of
edge,
that's
useful
to
you.
A
If
you're
in
industrial
iot
use
a
definition,
that's
useful
to
you,
they
might
vary
so
just
to
give
you
a
sampling
of
what
some
some
of
them
are
based
on
resource.
So
Edge
is
co-aligned
with
having
low
resource,
meaning
low,
compute,
low
power,
low
or
no
internet
connectivity.
Some
others
are
dependent
on
a
definition
that
they're
far
away
from
cloud
but
close
to
users
and
devices.
The
definitions
vary
quite
a
bit
and
I
think
I
have
come
to
the
conclusion.
A
E
E
It
varies
greatly
as
we've
spoken
to
lots
of
people
who
think
that
edges
are
are
something
that
are
low
lightweight
very
little
in
terms
of
compute
power.
Maybe
also
have
an
extra
modality
attached
to
them
like
a
camera
or
some
other
type
of
sensor,
visual
audio
Etc
and
then
there's
other
people
that
could
take
the
exact
opposite
and
make
it.
A
A
Is
that
if
you're
going
to
try
to
manage
these
over
time,
you're
going
to
have
a
lot
of
them
and
the
real
Crux
of
this
is
coming
up
with
the
management
infrastructure
that
lets
you
deploy
them
and
manage
them
at
that
kind
of
scale,
which
is
quite
a
bit
different
than
what
you
might
have
in
a
public
Cloud.
You
have
scale
there
too,
but
it's
a
different
type
of
scale
and
dealing
with
the
issues
of
low,
no
or
intermittent
connectivity.
F
So
the
the
way
I
I
see
I,
read
just
I,
guess
listening
to
a
lot
of
talk
about
the
big
guy
like
CEOs,
so
there's
three
areas,
that's
really
popular
topic
and
I.
Think
most
people
agree
is
number
one
AIML
number
two
security
and
in
a
way
number
three
is
Edge
AIML
and
security.
There's
a
lot
of
different
organizations
working
on
them.
There
are
a
lot
of
overlapping
areas.
However,
there
is
a
tendency
to
gradually
you
know.
Have
it
well
defined,
you
know.
F
F
E
No
I
was
just
going
to
say
is
that
for
me
it
depends
on
the
properties
right,
I
think
that
people
are
somewhat
agreeable
on
properties,
but
not
necessarily
attributing
it
to
any
one
specific
Edge
definition
right
now
you
brought
up
like
disconnected
operations.
I
know
you
were
talking
about
security
and
AI
in
another
space,
but
there's
also
it's
all
the
considerations
that
I
think
that
we
drew
down
on
the
white
paper
that
we've
been
working
on
right.
E
What
makes
an
edge
native
application
and
one
of
the
properties
of
of
edge
devices
that
you
have
to
consider
when
building
an
edge
native
application.
So
I
think
that's
what
we
were
kind
of
trying
to
get
our
hands
around
on
that
in
that
white
paper.
So
if
I
was
I,
couldn't
give
you
a
60
character
or
less
definition,
but
I
think
that
white
paper
intends
to
encapsulate
that
that
just
just
my
foray
into
the
discussion,
sorry
Steve
good
yeah.
A
No
I
agree
with
you,
and
part
of
that
is
there
are
all
these
characteristics
and
I
think
the
M
drops
some
of
these
in
the
chat
of
disrupted
in
a
practical
at
a
single
edge
location.
It
could
be
that
of
that
list
of
low
connectivity,
low
resource.
What
desire
for
latency
desire
for
data
sovereignty,
maybe
only
half.
A
That
list
applies
to
you
in
your
use
case,
but
you
can
still
get
great
value
out
of
these
Management
systems
and
the
articles
that
are
put
in
place
on
how
to
go
about
approaching
that
in
an
efficient
way
and
Victor.
The
other
things
I
think
you
brought
up.
Aiml
I
think
that's
a
brand
new
activity
for
ranch.
So
this
is
one
thing
that
I
think
is
dramatically
undercovered,
because
people
are
just
getting
there.
You
know
sort
of
like,
even
though
AIML
has
been
out
there
for
I
think
two
decades.
A
A
lot
of
the
resources
required
to
allow
this
to
take
off
technology
wise
have
just
take
they.
They
have
gone
around
a
curve
and
it's
a
combination
of
Technology
enablement
along
with
the
world
just
being
aware
of
it.
It's
sort
of
like
the
whole
chat,
GPT
thing
woke
the
world's
eyes
up,
or
maybe
the
world
I
mean,
is
corporate
management
and
yeah.
A
Frankly,
it's
changing
so
rapidly
that
if
somebody
wrote
a
white
paper
attempting
to
cover
the
AIML
for
ads
a
year
ago,
it
would
be
way
past
its
sell
by
date.
It
would
smell
pretty
bad
at
this
point,
because
it's
moving
so
rapidly.
It's
probably
one
of
these
things
where
my
opinion
is
in
AIM
alpharedge.
It's
a
time
to
get
in
there
and
play
with
it
and
do
proofs
of
Concepts
rather
than
start
writing
things
down.
That
will
be
obsolete
in
a
year
anyway,
and
it's
it's
a
real,
interesting
area,
but
I,
don't
think
I.
A
Think
if
you
use
Google
search
you'll
find
a
bunch
of
stale
articles.
If
any,
maybe
asking
chat.
Gpt
gives
you
something
better,
but
I'd
be
skeptical
and
it's
kind
of
like
just
go,
get
your
hands
on
some
equipment
and
start
playing
with
it,
because
you
probably
know
about
as
much
as
anybody
else.
At
this
point.
F
Yeah,
maybe
just
like
Andy
said
right:
the
paper
you're
working
on
is
probably
going
to
be
the
the
how
to
say
the
the
first
Splash
into
the
whole
ecosystem,
like
you
know,
really
getting
attraction
attention
into
this
area
because
talking
about
all
the
stuff
that's
being
worked
on,
you
know
cncf
and
everywhere
open
ssf
security
AIML.
When
you
move
all
that
into
the
edge
slash,
iot
and
mobile
world,
you
can
get
10
times,
not
100,
more
complicated
right.
So
it's
just
an
interesting
area.
F
D
Well,
you
know
Steve
I,
think
I
would
agree
with
you
from
a
technology
perspective
that
it's
moving
so
fastly
that
anything
would
be
past
Itself
by
date,
but
I
think
if
you
take
a
step
back
from
the
you
know,
the
procedures
to
you
know
the
underlying
reasons.
I
think
that
those
are
pretty
Timeless
and
those
have
been
a
consistent
Trend
that
have
been
pushing
us
towards
the
edges
or
towards
distributed,
and
it's
not
just
because
the
data's
on
the
edge
it's
latency
and
determinism.
You
know
it's:
privacy
and
security.
D
Look
at
Facebook
getting
whacked
for
pulling
all
of
the
French
data
back
to
the
US,
limited
or
delimited
autonomy,
regulatory
reasons.
I
think
those
are
some
of
the
compelling
motivators
that
we
have
that
are
driving
intelligence
out
towards
the
edges.
You
know
the
last
10
years
the
last
decade
was
dominated
by
this
great
lifting
shift
into
the
clouds
and
I
really
believe
that
the
next
10
years
are
about.
You
know,
bringing
the
compute
to
the
data
would
be
the
high
level.
D
You
know
shingle
that
I
would
hang
that
under.
So
that's
just
my
two
cents
on
on
that
and
then
underneath
that
I
do
agree
that
the
world
is
is
moving
so
fast
and
so
quickly
that,
if
you
even
just
consider
the
like
the
value
of
edge
deployed
ml
llamas
changed
the
game.
You
know
I
mean
when
you
can
have
a
an
ml
model
on
on
your
desktop.
That
is
nearly
as
powerful
as
something
that
you
could
previously
host
in
the
cloud
I
mean
it
does
really
change
the
game.
D
You
know
before
I
dropped
a
link
to
a
video
we
did
at
arm
Summit
last
year,
where
we
talked
about
running
lightweight
machine
learning,
models
on
the
edge
that
could
then
using
Blossom
Cloud.
That
could
then
you
know,
leverage
cloud-based
models
depending
on
privacy,
security,
performance
availability
or
other
reasons,
and
we
sort
of
walked
through
the
technical.
You
know,
how
did
we
get
this
to
work
and
all
those
other
all
the
underlying
reasons?
D
But
when
we
take
a
step
back,
I
think
that
the
technology
is
moving
so
fast.
That
now
you
might,
you
know
it's
strategically
move
compute
back
to
the
edges,
to
save
money
or
for
cost
or
all
kinds
of
other
reasons.
So
I
I
think
it's
super
interesting
and
Steve.
We
actually
work
with
that
Sony
team.
They
are
a
big
web
assembly
shop
and,
if
you've
seen
their
platform
that
they're
building
it's
a
really
neat
that
kind
of
stuff
on
on
on
their
approach
for
the
world.
A
Yeah
the
reason
I
brought
that
up
for
those
who
weren't
don't
have
the
chat
window.
Written
I
talked
in
this
group
a
few
weeks
ago,
shortly
after
kubecon,
Europe
I
went
and
saw
their
booth,
and
it
was
kind
of
interesting.
What
they
were
doing
was
a
an
attempt
to
use
webassembly
to
write
a
what
they
called
a
virtualization
wrapper
around
ml
accelerators
for
Edge
devices
and
I.
A
Think
Sony's
angle
in
this
is
that
they
make
camera
chips
so
that
this
would
be
used
in
their
context
or
machine
learning
related
to
image
processing,
but
it
could
be
open-ended.
The
concept
is
just
virtualize
an
accelerator,
because
these
Edge
accelerators
are
all
going
to
be
just
a
little
bit
different
and
it
makes
it
tough
to
write
a
higher
level
app.
A
If
you
have
to
deal
with
all
of
these
differences,
they
said
there
that
they
intended
to
open
source
this
in
the
June
time
frame
and
the
way
I
came
across
this
article
is
I,
just
went
and
looked
and
to
try
to
figure
out.
If
that
had
happened,
it
doesn't
appear
that
it
has
because
I
didn't
see
anything
about
it
in
their
GitHub
repo,
but
I
assume
that
it's
still
in
the
pipeline,
because
they
posted
this.
D
So
the
the
virtualization
of
capability
providers
is
something
that
wasmcloud
has
done
since
it's
been
in
the
cncf
so
a
few
years.
You
know
where
you
define
a
contract
like
SQL
and
you
define
you
use.
You
know:
contract
driven
development
to
Define,
an
interface
for
what
that
looks
like,
and
then
you
can
have
pluggable
you
know.
Providers
Dapper
is
very
similar.
They
do
that
as
well
from
a
webassembly
perspective.
D
This
is
actually
a
part
of
a
bigger
standard,
that's
referred
to
as
the
component
model
and
it
uses
a
language
called
wit,
web
assembly
interface
types
to
let
you
define
those
interfaces,
so
I
was
working
on
the
blog
post
for
the
webassembly
roadmap
and
we're
announcing
it
should
hopefully
go
out
this
week.
But
those
are
the
formal
specifications
that
I
believe
even
Sony's,
relying
on
underneath
the
hood
in
order
to
define
those
and
groups
of
collections.
D
Of
these
you
know,
interfaces
are
defined
into
what
you
call
worlds
so,
for
example,
the
Two
Worlds
that
we
are
shipping
our
yzcli,
which
includes
things
like
wall,
clock
and
file
system
types
and
Network
and
TCP
things
like
that
and
then
HTTP
as
well,
which
has
obvious
huge
implications
for
you
know
Faz
and
functions
as
a
service,
and
things
like
that
there
is
another
set
of
Standards.
That's
called
Wazi
cloud
awazi
webassembly
system
interfaces
that
defines
common
Cloud
interfaces
like
blob
store
plugins
could
be
file
system
S3.
D
You
know
anything
that
you
can
Implement
a
blob
store
on,
but
those
have
been
deferred
and
are
moving
just
at
a
parallel
track
outside
of
the
standard.
One
of
the
reasons
why
we
pulled
those
out
of
the
standard
was
because
we
didn't
want
all
of
the
webassembly
engines
to
have
to
support
100.
You
know
new
interfaces
on
day
one.
We
wanted
to
keep
some
optional
outside
of
the
standard
and
to
accelerate
that
work.
There's
a
new
definition,
a
virtualization
layer
that
lets
you
virtualize.
D
Those
interfaces
for
testing
cicd
or
you
know
like
essentially
have
placeholders
in
there
so
watch
a
little
blog
post,
that'll
be
on
the
bytecode
alliance.
I'll,
try
to
drop
it
into
slack
when
I
think
about
it
or,
if
I
remember,
to
and
Link
it
in.
The
notes
here,
but
it'll
have
links
to
the
blog
post
does
already
have
links
to
you
know
where
we
are
in
garbage
collection,
where
we
are
on
threads
and
all
of
the
other
related
underlying
standards.
A
G
E
D
A
D
Both
yeah,
we
we
do
both
I
mean.
If
you
look
at
any
of
the
previous
Watson
Cloud
talks,
we
at
the
time
we
created
it.
We
you
know
it
didn't
exist.
The
web
assembly
interface
types,
the
IDL,
that
you
use
to
define
the
interfaces,
so
we
use
the
tool
from
Amazon
called
Smithy
and
we're
in
the
process
now
of
converting
all
of
those
to
wit,
so
that
so
that
you
have
you
know
fully
standards.
D
Compliance
within
Blossom
Cloud
now
wasn't
cloud
is
an
application
runtime
that
sits
on
top
of
a
webassembly
runtime
called
wasn't
time
and
the
web
assembly
you
know
wasn't
time
is
the
engine
that
runs,
but
what
wasn't
Cloud
gives
you
is,
you
know
tooling,
for
you
know
starting
scaling
running
operating.
It
runs
on
top
of
kubernetes,
but
it
also
runs
independently
of
itself
I'm,
more
than
happy
to
I,
think
there's,
probably
two
opportunities
here.
A
That
sounds
great
and
I
think
maybe
even
a
general
context
of
webassembly,
because
I
think
there's
great
potential
at
Edge
for
this
to
be
of
value,
given
that
one
of
the
characteristics
of
edge
is
that
device
Hardware
life
cycles
tend
to
be
really
long.
Historically,
you
know
in
a
public
cloud
or
a
big
commercial
data
center,
that
equipment
gets
replaced
every
three
to
five
years.
It
you
rarely
see
something.
Seven
years
old,
it's
kind
of
warranty
runs
out.
D
A
D
Right,
but
it's
there's
even
a
greater
benefit
here
when
we
think
about
you,
know
the
layers
of
abstract
action
that
we've
climbed
through
the
next
big
layer,
that's
available
to
abstract
is
the
application.
Libraries.
Those
are
the
things
you
know.
We
pull
in
all
this
open
source
code
and
we
statically
compile
it
at
at
compile
time
and
then
embed
it
and
bake
all
those
vulnerabilities
in
the
app
in
a
container
and
ship
it.
D
What
webassembly
is
doing
is
trying
to
drive
that
next
level
of
abstraction,
where
capabilities
are
actually
loaded
at
runtime
and
the
big
Advantage.
There,
then,
is
you
ship
business
logic
in
any
language?
You
say
this
needs
a
web
server.
It
needs
a
SQL
database,
it
needs
a
file
system,
it
needs
a
blob
store
and
it's
going
to
open
these
ports
and
then
at
runtime.
The
platform
gives
you
the
latest
version
of
those
components
or
the
version
of
those
components
that
work
in
that
particular
domain.
So
how
does
is
this
all
science
fiction?
D
No,
this
is
exactly
how
fastly
and
cloudflare
work
if
you've
ever
programmed
on
workers
or
compute,
underneath
the
hood.
That's
just
webassembled
cloudflare
runs
V8,
which
is
the
in
genetic
Chrome
and
fastly
runs,
wasn't
taught,
so
this
is
I
think
a
very
real
Tech,
that's
being
used
to
deliver
trillions
of
transactions
in
a
given
quarter,
vastly
posted
a
Blog
on
their
first
trillion
transactions
last
year.
The
year
before
and
I
really
think
that
across
Cloud
native
this
is
the
next
epic
of
tech
that
we're
going
to
slide
into
yeah.
A
A
Know
what
it
is
and
kind
of
the
shiny
object
of
ml-
maybe
is
distracting
them
now,
but
there's
an
opportunity
for
those
things
to
be
related
and
I.
Think
that
a
lot
of
this
talk
about
webassembly,
being
the
translation
Force
to
hide
differences
in
the
underlying
Hardware,
maybe
has
to
overcome
people's
attitudes
of
having
heard
that
story
before
with
Java
yeah
getting
burnt
out
on
the
experience.
A
Maybe
not
because
the
concept
was
bad,
but
maybe
a
lot
of
that
is
because
the
garbage
collection
of
memory
and
Java
kind
of
burned
it
for
a
lot
of
edge
things
that
needed
determinism
I,
don't
know
that
webassembly
is
guaranteed
to
be
different
in
that
context,
but
I
think
you
can
implement
it
in
a
way
that
you're
not
exposed
to
that.
So.
D
Yeah
I
think
you
make
a
number
of
compelling
points,
but,
to
be
honest,
I
think
that
I
mean
like
last
year.
One
of
the
things
that
we
analyzed
was
that
the
web
assembly
standards
were
moving
too
slowly.
So
I
don't
know
if
anybody
on
this
call
does
Bailey
Hayes,
but
we
recruited
her
and
she
was
writing
the
web
assembly
plugins
for
single
store.
You
know
bring
your
your
compute
to
the
data.
D
You
know
ship
your
logic
into
your
database
and
we
I
cosmotic
hired
her
and
has
really
surged
her
into
the
standards
to
help
organize
and
collaborate
and
coordinate.
Earlier
this
year
we
did
a
private
road
mapping
Summit,
and
what
it's
resulted
in
is
an
incredible
amount
of
focus
on
a
smaller
number
of
standards
and
it's.
D
What
it's
meant
is
is
that
you're
are
seeing
Now
webassembly
language
tooling,
coming
across
the
line
for
python
for
go
121,
which
is
the
next
release
for
a
kotlin.net,
a
JavaScript
there's
two
different
ways
to
bring
it
across
and
the
standards
have
really
accelerated
pretty
dramatically
and
it's
Bailey's
not
solely
responsible
for
it
at
all.
But
she
spends
a
lot
of
her
time
in
the
ecosystem
and,
like
you
know,
creating
I
don't
want
to
share
the
roadmap
because
we're
about
to
launch
the
the
blog
but
I.
D
They
are
super
geeked
up
about
this
Tech
as
well
as
I,
think
is
most
of
the
people
that
are
in
webassembly,
because
it's
still
that
Better
Together
moment
in
web
assembly,
where
we
see
all
this
great
opportunity
and
collaborating
to
get
to
agree
on
the
standards,
because,
what's
really
neat
is
and
I
think
we
were
all
unprepared
for
this
last
year.
If
components
are
the
new
container,
it's
going
to
be
an
amazing
world.
When
you
can
ship.
D
You
know
units
of
logic
across
any
of
those
platforms
cosmotic,
which
is
my
company,
a
fastly,
cloudflare
wasm
cloud
or
any
of
the
other
runtimes
that
are
out
there.
Kate
Golden
Ring
works
for
a
great
company
called
fermion.
They've
got
an
awesome,
a
fast
runtime
called
spin.
All
of
us
now
will
be
able
to
ship
those
components.
I,
don't
think
the
world's
really
ready
for
something
that
small
and
that
portable
and
as
to
what
it
means.
A
F
F
Assembly
is
a
very
interesting
area,
so
I've
been
helping
customers
doing
database
and
migration
to
the
cloud
for
the
past
10
years.
So
most
of
the
time
customer
just
left
with
this
Edge,
that's
the
only
place
that
customer
has
and
and
that's
where
they're
innovating
and
the
rfat
Really
there's
nothing
going
on
discussion
over
there.
So
I
was
hoping.
This
could
be
the
really
the
place
where
I
know.
The
edge
related
topic
is
including
the
landscape.
Discussion
can
start
from
here.