►
From YouTube: ONNX Foundation WG - LF JDF intro meeting 20190920
Description
Meeting recorded by Jim Spohrer (IBM) on September 20, 2019. This is intro meeting with LF JDF.
A
You
hello,
this
is
gyms
for
co-chair
of
the
onyx
working
group
of
onyx
and
a
foundation
Ryan
Loney
from
Intel
is
also
on
from
the
onyx
community
and
Scott
and
Seth.
Thank
you
very
much.
I
know
I
sent
you
an
email
early
early
this
morning
and
said
hey.
Would
you
help
the
onyx
community
learn
a
little
bit
more
about
the
joint
linux
foundation?
Jdf,
and
I
really
want
to
thank
you
for
being
so
responsive
I.
Think
Scott
I
got
your
email
response.
A
30
minutes
after
I
sent
a
note
saying
happy
to
help
and
let
me
provide
a
little
context.
Then
then
I'll
ask
you
to
introduce
yourself
and,
and
we
can
proceed
with
a
deck
that
you've
prepared,
but
I
know
you
both
know
the
Onyx
community
well
and
onyx,
originally
created
by
Microsoft,
Facebook
and
AWS
as
an
interchange
format
for
deep
learning
models.
A
The
community
has
grown
very
well
in
the
last
community
meeting
I
think
there
were
over
a
hundred
and
fifty
people
there.
Lots
of
good
presentations
and
I
was
asked
by
the
Onyx
steering
committee
to
work
with
Ryan
to
start
this
onyx
and
a
Foundation
working
group,
because
they're
eager
to
get
the
intellectual
property
in
the
assets
into
a
foundation.
We've
had
calls
previously
with
Apache
Eclipse
and,
of
course,
Abraham
from
Linux
Foundation,
AI
and
Mike
Dolan
also
presented.
A
So
thank
you
for
making
the
time
to
present
to
us
and
at
this
point,
I'll
turn
it
over
you
to
take
us
through
the
presentation
that
you
prepared
and
educate
us
a
little
bit
about
the
Onyx
I
mean
JDF,
and
if
you
have
any
questions
so
for
Ryan
and
I,
please
go
ahead
and
ask
those
as
well
and
I'll
come
over
to
you.
Scott
that'd.
B
Be
great,
thank
you.
So
much
Jim
and
Ryan
I'm
Scott,
Nicholas
I'm
senior
director
of
strategic
programs
for
the
LF
I,
spend
a
lot
of
time.
I've
focused
on
setting
up
collaborations
on
the
standard
side
and
on
the
open
source
side,
so
very
pleased
to
be
talking
to
Seth.
If
you've
got
a
new
button,
handy
I'll,
have
you
introduce
yourself
as
well?
Seth
is
our
general
manager
of
standards.
B
So
the
traditional
challenge
and
the
reason
the
JTF
was
formed
a
number
of
years
ago-
is
that
existing
approaches
for
creation
of
standards
bodies
didn't
fit
the
needs
of
all
projects
they
were
and
and
can
be
highly
customizable,
but
very
complicated.
B
Let's
zoom
through
some
of
these
and
the
the
goal
and
the
vision
of
JDF
is
to
create
a
standards
organization
in
a
box.
So
what
JDF
is?
Is
it's
a
framework
for
coming
together
to
create
a
standard,
either
one
specification
or
a
series
of
specifications
and
to
do
so
in
a
very
fast
manner,
with
clearly
defined
and
understood
rules
of
operation
and
intellectual
property
frameworks
that
are
custom,
tailored
with
a
specification
in
mind?
B
B
The
unique
aspect
of
JDF
is
that
the
governance
documents
are
a
template.
So
while
we
have
every
capability
of
doing
fully
customized
community
specific
documents
on
the
LF
side,
the
vision
for
JDF
is
to
enable
rapid,
spin
up
of
a
standards
organization
and
therefore
the
terms
aren't
something
that
are
negotiated
from
agreement
to
agreement.
There
is
flexibility
in
choosing
IP
regimes
within
a
set
of
check
boxes,
but
the
other
than
that.
B
B
The
structure
that
the
JD
of
template's
follows
is
one
where
there's
a
project
structure,
not
sorry,
a
project
charter
that
talks
about
how
the
steering
committee
will
make
decisions.
But
then
the
specific
work
is
done
at
the
working
group
level
and
each
working
group
can
have
its
own
IP
structure
and
its
own
set
of
specific
deliverables.
B
A
time
on
the
working
group
charter,
so
when
a
project
is
set
up,
if
there's
a
specific
specification
to
be
done,
that
would
be
done
in
the
JTF
model
as
part
of
a
working
group.
The
working
group
charter
is
a
short
governance
document
that
again
it's
not
negotiated,
but
there's
some
check-the-box
items
you
would
on
the
working
group
charter.
Put
the
name
of
the
working
group,
talk
about
the
scope
and
then
choose
a
copyright
policy
for
the
project
and
indicate
whether
there
was
a
particular
patent
approach
for
the
project
as
well.
B
Workflow
JDF
is
intended
to
be
to
enable
the
rapid
setup
of
a
standards
organization,
but
also
one
that's
capable
of
developing
a
comprehensive
range
of
specifications.
The
workflow
within
the
JTF
structure
is
contemplated
as
follows.
So
initial
drafting
of
the
specification
is
done
at
the
working
group
level.
This
is
in
blue
here.
B
Draft
deliverables
would
be
approved
by
that
working
group
and
then
they
would
go
up
to
the
steering
committee.
The
steering
committee
is
established
as
part
of
the
project
charter,
and
the
steering
committee
would
then
take
action
to
approve
the
draft
deliverables
as
approved
deliverables.
From
that
point,
the
steering
committee
can
decide
to
take
some
additional
actions,
such
as
publisher,
publishing
the
specification
or
submitting
it
to
an
international
standards
organization.
B
If
that's
something
that
the
group
wants
to
do,
the
intellectual
property
structure
selected
by
the
working
group
will
have
in
many
cases,
requirements
that
will
kick
in
when
a
specification
is
finalized
and
in
this
depiction
here,
finalization
event
is
when
the
steering
committee
approves
it.
So
when
projects
are
looking
at
the
JTF,
there
are
several
advantages
to
the
JDF
structure:
it's
not
the
right
fit
for
every
project.
B
Where
projects
need
to
have
a
rapid
way
of
spinning
up
with
the
ability
to
tap
into
the
full
resources
of
the
Linux
Foundation,
both
from
a
personnel
and
responsiveness,
but
then
also
from
a
standpoint
of
being
within
the
ecosystem
and
cross
connecting
with
other
projects
that's
available.
If
there
are
specific
specification,
oriented,
IPR
protections
that
the
project
is
looking
for,
that
are
open-source
friendly
JDF
has
a
comprehensive
set
of
options
for
for
that
situation.
C
No,
this
is
Seth,
actually
just
one
small
clarification
on
the
steering
committee.
This
is
doing
he
is
still
inside
the
project.
It's
so
it's
not
an
external
steering
committee.
The
the
founders
of
the
group
are
fully
in
control
of
the
specification
and
its
approval.
Thank
you
for
that
clarification
set
yes
great.
A
B
B
B
B
To
the
IP
commitments
that
are
made
so
movie
step
involved
of
getting
those
signed
up,
this
is
akin
to
we
don't
JDF
doesn't
call
them
clas,
but
it's
it's
similar.
You
know
it's
similar
from
a
process
standpoint
and
it
will
want
a
feedback
agreement
or
a
membership
agreement
for
contributions
to
the
specification.
This
is
that's
a
that
takes
some
time,
but
that's
as
fast
as,
as
you
know,
the
participants
can
turn
documents
right,
good.
A
Self-Governance
of
the
community,
how
does
how
does
that
work?
I
think
I
mentioned
neonics
community
is
a
very
well-run
community.
You
know
150
or
so
people
at
the
last
meeting
as
well
established
steering
committee
with
you
know:
bylaws
operating
principles,
all
of
that
good
stuff
with
Microsoft
Facebook,
AWS,
Intel
and
NVIDIA
on
the
steering
committee.
So
could
you
say
a
little
bit
about
the
ability
in
the
prosperous
process
for
onyx
community
to
continue
to
self-govern.
B
Certainly
so
in
the
in
the
checkbox
options
on
the
on
the
project
charter,
there's
there's
two
approaches
that
are
provided.
One
is
to
have
each
steering
member
appoint
someone
to
the
steering
committee.
The
other
approach
is
to
use
a
model
developed
by
consensus
of
the
steering
members
and
that's
to
enable
flexibility
for
a
situation
when,
when
a
group
wants
to
govern
itself
through
a
different
approach,
so
the
basic
template
is
structured,
a
particular
way.
If
the
community
wanted
more
flexibility
there,
there
is
some
flexibility
built
into
the
JD
of
templates.
B
A
Ryan,
did
you
have
any
questions,
no
I
guess
just
to
clarify
in
the
last
plane
it
sounds
like
we
would
be
more
restricted
in
our
ability
to
self
govern
and
some
of
our
other
options,
but
it
was
I
hearing
that
everything
had
to
be
approved.
You
know
sort
of
a
level
up
from
the
steering
committee,
though
the.
B
Steering
committee
is
the
level
up
I
see,
okay,
the
way
JDS
is
structured,
and
the
trade-off
here
is
a
speed
of
implementing
a
new,
creating
a
new
community
versus
of
absolute
flexibility
that
that's
the
trade-off
JDF
is,
is
focused
on
enabling
a
rapid
spin
up
with
consistent
terms
and
therefore
that
uses
that
structure
of
having
civic
technical
work
done
at
a
working
group
level,
and
something
is
ready
to
be
approved
as
a
fine
specification.
It
would
go
up
to
the
steering
committee
for
approval.
C
A
A
So
I
think
this
is
super
helpful
and
again
I
want
to
thank
you,
Scott
and
stuff,
because
when
Ryan
and
I
were
on
the
Onyx
steering
committee
meeting
yesterday,
they
said
what
about
JDF
can.
Can
we
learn
a
little
bit
more
about
that?
So
I
think
it's
a
good
thing
that
the
Onyx
committee
wanted
to
learn
more
community
wanted
to
learn
more
steering
committee
wanted
to
learn
more
about
onyx,
but
I
also
know
I.
Think
I
have
this
right.
A
B
Is
that
is
correct,
I'll
just
put
an
asterisk
there
and
that
we
actually
use
the
Linux
Foundation
for
seed
processing
and
fundraising,
okay.
We
would,
for
example,
here's
a
publicly
operating
project.
The
graph
QL
specification
uses
the
patent
constructs
for
specification
development
of
the
joint
development
foundation.
The
money
raising
is
done
at
the
L
app
and
that's
that's
the
structure
we
use
going
forward.
It
has
a
variety
of
advantages
for
everyone.
Okay,.
A
And
I
think
the
you
know
when
I,
when
we-
and
this
is
kind
of
Ryan
and
I
as
we
take
this
back
to
the
community
I
think
you
know
both
Ryan
and
I-
see
the
Linux
Foundation
as
a
kind
of
a
whole
and
Apache
foundation
is
kind
of
a
Kohl
and
Eclipse
Foundation
is
kind
of
a
hole.
So
we've
got
these
three
kind
of
holes,
but
I
do
think.
There's
probably
one
question
the
steering
committee
is
going
to
ask,
is
you
know
so
Linux
Foundation?
A
What's
the
advantages
of
doing
LFA,
I,
/,
Linux,
Foundation,
JDF
and
I?
Think
you
know,
Ryan
and
I
are
trying
to
net
this
out
for
the
community
to
make
it.
You
know
simple
and
clear
and
I
think
that
the
the
the
benefits
that
we
see
for
the
Linux
Foundation
ai
is
like
voids
already
in
there.
A
key
Mouse
is
already
using
onyx.
A
If
we
went
with
the
JDF,
it
feels
like
onyx
community
would
kind
of
be
in
a
in
a
in
a
somewhat
more
isolated
position
in
linux
foundation
than
if
it
was
part
of
LFA
I.
It
would
be.
You
know
part
of
an
AI
group
with
you
know,
vote
adopted
by
a
chemos.
Is
there
any
thoughts
that
you
have
Scott
or
Stefan?
You
know
the
pros
and
cons
of
going
I
mean
going
with
Linux
Foundation,
but
AI
versus
JDF
with
anything.
B
Preserve
your
options
and
thinking
about
things,
but
just
share
a
best
of
my
own
about
some
of
this
I
think
ecosystem
benefits.
You
know
we're
focused
on
providing
that
to
all
of
our
projects
in
all
areas.
So
whether
something
is
in
JDF
or
something
is
is
elsewhere
we
we.
We
want
there
to
be
up
streams
and
down
streams
and
collaboration
and
sharing
of
ideas.
Okay,.
C
B
B
So,
if
you're
looking
at
a
specification-
and
you
need
a
patent
commitment
and
patent
standstills
of
that
apply
to
all
the
members
and
all
the
contributors,
whether
or
not
the
contribution
comes
from
them-
and
you
want
the
the
robust
set
of
patent
protections
in
an
open
source,
complimentary
approach
and
JAF
is-
is
great
at
delivering
that
type
of
outcome,
and
let
me
say
that
type
of
outcome
in
a
ready
to
go
template
on
the
LF
side,
we
can
structure
any
project
any
way
that
a
community
needs.
B
C
B
Any
arrangement
in
terms
of
IP
structures
of
JDF
we
can
replicate
in
the
LF
and
any
community
governance
in
the
LF
we
can.
We
can
replicate
that
the
challenge
with
JDF
is
in
order
to
enable
the
speed
of
spinning
up
the
sto,
and
that
has
many
advantages.
It
does
leverage
a
membership-based
working
group
and
steering
committee
structure
and
there
is
a
trade-off
there.
C
If
I
could
just
add
a
comment
here,
I
think
that
if
you,
if
you
start
from
it
LF
perspective
and
you
think
about
the
JTF
specification
setting
activities
as
a
service
that
can
be
kind
of
vertically
integrated
into
the
LF-
that's
not
a
bad
way
of
thinking
about
it.
So
the
JTF
is
available
to
you.
If
you
start
it,
if
you
start
from
inside
LF
and
workout,
if
you
start
from
JDF
and
work
in
that
works
too,
but
you
get
a
you
get
kind
of
a
more.
A
Yeah
yeah
I
think
that's,
that's
really
helpful
and
I
understand
and
I
I
think
Ryan
thing
that
we
have
to
figure
out
when
we
go
back
to
the
onyx
steering
committee
on
this
is,
it
seems
to
me
like
we
should
probably
have
the
options,
be.
You
know
LF,
Apache
and
Eclipse.
Those
three
I
think
from
my
perspective,.
A
You
know
you
really
take
a
group,
that's
not
used
to
working
together
and
you
you
get
them.
You
know
into
the
whole
culture
and
tradition
of
working
together
and
a
non-profit
foundation,
whereas
in
the
case
of
onyx
community
they've
already
passed
all
of
many
of
those
early
stage,
various
barriers
and,
while
I
think
the
you
know,
JDF
still
be
appropriate
for
things
that
are
more
mature.
Really.
You
know
when
I
look
at
what
the
Onyx
community
needs
at
the
next
stage.
It's
more,
you
know
how
do
we
get
more
adoption
into
related
projects?
A
A
Whereas
if
you
have
something
that's
more
fully
formed,
there
can
still
be
some
benefits,
but
there
could
also
be
benefits
with
an
organization
that
already
has
more
more
aligned
projects.
So
I
think
that
is
the.
That
is
the
kind
of
discussion
that
might
happen
in
the
Onyx
community,
around
Linux,
Foundation
and
JDF
and
AI
Ryan.
Did
you
have
I
know.
A
Intel
has
some
concerns.
The
current
onyx
community
has
clas.
Could
you
say
a
little
bit
more
about
intell's
preference
going
forward,
not
cap
clas,
and
how
that
might
meet
the
perspective
from
Intel
legal
is
that
we
prefer
not
to
have
to
have
CL
age
for
contributions
and
the
preference
is,
you
know,
Apache
2.0
license
without
a
CLA.
That's
you
know
that
they're
they're,
very
flexible,
on
working
with
Linux,
Foundation
and
I.
A
Think
the
key
for
us
right
now
is
that
there's
we
don't
have
a
licensee
in
place
for
what
was
previously
the
CLA
for
onyx
that
was
set
up
by
my
Microsoft
and
Facebook.
So
you
know
it's
the
any
of
the
number
one
preference
from
Intel
Eagles
perspective
is
to
not
have
a
CLA,
but
if
a
CLA
is
required,
the
licensee
should
be
a
non-profit
foundation
and
that
I
think
that
our
legal
team
wants
to
be
involved
and
maybe
Judea
might
be
a
little
bit
too
locked
in
where
or
they
won't
be
able
to
give.
A
You
know,
here's
here's,
a
change
that
we
want,
or
here
this
doesn't
work
for
us
and
it
sounds
like
maybe
there'd
be
more
flexibility
with
with
El
FAI
about
some
of
those
those
decisions
about
IP
and
how
the
license
agreements
are
structured,
but
I,
yeah
I
think
that
the
the
most
important
takeaway
from
Intel
legal
is
preference
for
Linux,
Foundation
and
everything
else.
After
that,
you
know
their
experience
has
been
very
good
with
Linux
Foundation,
so
contemplated
projects
and
I
think
that's
that's.
The
most
important
thing
is
we'll
figure
out
how
to
make
it
work.
A
Assuming
we
go
down
either
at
the
pass
right
now
for
Linux
Foundation,
okay,
we're
good
yeah
right,
I,
see
we're
a
little
over
time,
but
I
just
wanted
to
thank
everybody
again,
Scott's
at
and
I'm
for
joining
the
call-
and
this
will
make
this
available
in
the
Onyx
community
and
will
definitely
be
circling
back
Scott
and
sets
if
we
have
any
further
questions
or
next
steps
from
the
community.
So
thanks
again
for
making
time
anything
before
I
stop
the
recording
Scott
stop
just.