►
From YouTube: IETF102-CDNI-20180720-0930
Description
CDNI meeting session at IETF102
2018/07/20 0930
https://datatracker.ietf.org/meeting/102/proceedings/
A
A
So
passing
around
the
blue
sheets
and
we
have
jabber
and
taking
so
we're
good
to
go,
Kevin
isn't
gonna,
be
helping
me
out
remotely
today
and
Francois
could
not
make
it.
So
this
is
the
no
well.
This
is
Friday,
so
I
assume
you've
seen
this
many
times
already.
If
not
I
definitely
suggest
you
read
it
governs.
How
governs
your
participation
here
at
the
idea.
A
B
A
A
A
E
E
So
here's
the
list
of
the
remaining
milestones,
the
room.
The
only
remaining
milestone
currently
is
URI
signing,
which
Phil's
going
to
give
an
update
about
in
a
minute.
Next
page,
please,
and
then
we
have
our
beyond
working
group
milestones
the
rate
pacing
HTTP
delegation,
which
Sanjay
is
going
to
talk
about
and
the
SVA
open
caching
extensions
that
Ori
is
going
to
discuss
as
well.
You
can
page
through
all.
F
A
A
To
come
and
talk
about
this
and
I
was
fully
expecting
to
come
here
this
week
and
say
that
this
was
not
happening
for
a
while,
but
Tuesday
they
came
and
said,
hey.
You
can
continue
work
on
this,
so
scurry
did
a
bunch
of
work
and
came
up
with
these
slides
to
talk
to
you
guys
about
today,
so
this
might
be
more
like
10
minutes
and
set
5s
so
really
feel
like
we're
close
to
the
final
draft.
Maybe
why
is
this
not.
A
It's
not
better.
Okay,
so
have
three
open
issues.
One
of
them
was
a
little
bit
contentious
and
I
want
to
talk
about
it
today
and
that's
a
new
base
claim.
This
was
discovered
by
Chris
lemons
he's
the
one
who
implemented
the
URI
signing
with
the
change
with
the
token
renewal
for
Comcast
and
the
issue
was
and
I'll.
Let
him
come
up
to
the
mic
and
explain
this
a
little
bit
more
depth,
but
essentially,
when
generating
an
HTTP
cookie
to
send
back
to
the
client.
A
You
have
to
give
it
a
path,
and
when
you
have
a
regex
for
your
URI,
it's
not
really
obvious
what
that
path
should
be,
and
you
don't
want
to
reverse
and
reverse
engineer
or
reg
X
to
come
up
with
the
path
and
using
path.
Slash
is
not
really
a
good
solution,
because
then
you
pollute
the
namespace
with
cookies,
so
chris
has
a
PR.
It's
number
44
on
the
github
repo
and
I
wanted
him
to
come
up
to
the
mic.
If
you
would
and
explain.
G
All
right
so
I'll
start
with
apologies,
I'm
new
guy,
I'm,
gonna
claim
new
guy
status
for
at
least
a
little
bit
longer
so
I
was
I
was
probably
violated
a
whole
bunch
of
social
norms
when
I
opened
this
PR.
But
the
basic
problem
is
that
when
we
send
the
cookie
back
for
signed
token
renewal,
we
have
to
have
the
path
and
it's
you
can
optionally
put
the
host
in
there
for
our
particular
use
case.
G
We
didn't
need
to
have
the
host
and
so
I
needed
some
way
to
pass
the
the
host
in
the
path
and
the
scheme.
Well,
the
most
sane
way
of
passing
a
scheme,
a
host
in
a
path
is
via
URI,
had
a
little
bit
of
trouble
with
that,
because
you
don't
always
need
all
of
them.
You
do
always
need
a
path,
but
you
it's
likely
that
you
don't
need
a
host
or
a
scheme
for,
like
the
CDN
use
case
that
we
had
again.
Other
people
may
have
have
different
requirements.
G
So
I
put
in
some
some
language
around
that
so
that
you
know
you
could
take
the
URI,
but
you
could
drop
pieces
and
maybe
that's
not
necessarily
the
best
way,
but
there
should
be
some
thinking
around
that
and
so
I
added
this
base
claim.
That's
that
gave
all
of
that
information
that
I
needed
for
the
cookie
renewal
and
then
I
realized
that
every
time
I
use
a
base
claim
that
says
this
particular
cookie
should
be
renewed
at
this
particular
path.
G
Every
time
I
said
that
I
always
meant,
and
the
cookie
is
valid
for
that
path
too,
and
I
thought
about
some
some
complex
ways
of
handling
it,
but
basically
anytime
you
want
the
cookie
for
that.
You,
you
want
a
cookie
to
be
renewed
at
a
path.
It
doesn't
really
make
sense
to
have
the
cookie
outside
the
path,
and
so
either
you
can
use
the
regular
expression.
A
G
G
G
You
would
put
this
container
in
alright,
so
the
base
claim
is
not
a
container
right.
The
regex
is
a
container,
so
you
put
the
base
claim
in
that,
tells
you
what
path
your
cookies
should
be
renewed
at
if
you
have
a
regex
for
your
UC
claim,
because
you
only
get
one
UC
claim
right.
So
if
you
have
a
regex
for
your
UC
claim,
then
the
regex
controls
what
it's
valid
for,
if
you
so
regex
that
says
this
cookie
is
not
valid.
It
is.
G
This
is
only
valid
outside
the
path
of
the
that
you
passed
for
the
cookie
renewal,
then
cookie
renewal
is
not
going
to
work
very
well
for
you.
So
it's
up
to
you
as
the
as
the
cookie
maker
at
that
point
or
the
token
maker
to
keep
those
in
sync.
So
so,
where
does
the
other
container
commitment,
because
that
there.
G
There
is
so
the
purpose
of
the
base
container.
It's
an
optimization
that
says
instead
of
having,
instead
of
repeating
the
text,
it's
strictly
less
powerful
than
the
regex
container,
instead
of
repeating
the
text
as
a
regex
just
go
see
that
data.
Otherwise,
every
time
you
use
sign
token
chaining
you'd
be
required
to
repeat
the
path
twice
in
the
cookie
or
in
the
token
okay.
G
If
we
allow
reg
X's,
we
can't
really
do
that.
This
provides
a
fairly
simple
way
to
not
bring
back
the
globbing
format
that
that
I
hate
so
much
but
allow
a
a
path
prefix
container
effectively,
so
that
we
can
potentially
disable
regex
acceptance
for
certain
keys
that
we
don't
trust
quite
so
much.
Okay,
let.
A
A
F
G
So
technically,
at
the
Maya,
so
at
the
moment
the
base
claim
is
primarily
useful
for
cookies.
We
do
allow
sign
to
containing
is
the
only
token
renewal
that
we
have
defined
in
the
spec,
but
the
it's
a
little
open-ended,
so
there
might
be
other
ways
of
renewing
that
token.
If
those
other
ways
needed
a
path
at
which
that
token
should
be
valid.
This
would
serve
as
information
for
that
as
well.
G
A
The
the
alternative
here,
I
think,
is
to
just
have
a
you
know,
cookie
base,
or
something
like
that
claim
that
just
has
a
path
in
it
for
basically
handing
the
path
off
to
to
the
degeneration
of
the
cookie.
Is
that
something
like
do
you
feel
like
that's
too
much
extra
baggage
in
the
in
the
token,
and
that
we
need
to
not
do
that,
because
I
think
that's
the
primary
driver
for
what
Chris
is
trying
to
do
I'm.
F
A
You
don't
have
to
use
the
the
changing
at
all,
but
if
you
do
use
the
chaining
and
you
do
need
a
path
basically
you're
like
the
the
path
is
likely
in
the
regex
and
then
it
would
be
repeated
in
a
more
specified
form
in
another
claim
that
says
hey,
and
this
is
the
path
that
you've
set
the
cookie
to
so
that
you
only
get
it
handed
back
to
you
when
you're,
when
the
user
is
accessing
something
in
that
scope.
So
it's
like.
A
Think
that's
where
Chris
is
coming
from
to
say:
hey,
we
don't
need
that
second
copy
of
the
path,
but
it's
to
meet
to
me
personally.
It
seems
like
a
lot
of
work
to
get
rid
of
those
characters,
but
I
know
that
people
in
the
working
group
were
concerned.
People
outside
the
working
group
are
concerned
about
the
size
of
the
token
and
also
the
efficiency
of
it.
So
no.
F
A
So
not
in
all
cases
is
it
duplicated,
but
if
you
are
making
a
very
specific
right,
jex,
it's
likely
duplicated
and
trying
to
reverse
engineer
the
regex
and
say
well,
you
know
by
processing
this.
This
would
be
the
path
that
matches
so
we'll
set
that
took
the
cookie.
That
is
way
way
too
much
complexity,
I
think
and.
G
A
Right
well,
unless
anybody
else
has
more
comments,
you
have
more
comments,
okay,
but
unless
anybody
else
has
anything
else
to
say,
I
think
at
a
minimum.
We
need
to
work
on
the
wording,
because
there's
some
concern
that
it's
undone
implementable
as
written
due
to
the
complexity
of
it,
so
we'll
have
to
take
this
offline
and
and
try
and
come
up
with
the
least
and
better
wording
yeah.
If.
G
A
G
A
G
So
this
person,
so
the
person
who
has
that
issuer
and
that
key
does
have
authorization
to
sign
tokens
and
I,
will
ask
I
will
check
those
tokens
when
someone
is
accessing
a
site
or
a
resource
that
is
in
scope
for
the
issuer
at
hand
right,
but
I.
Don't
necessarily
trust
that
issuer
not
to
hand
me
malicious
tokens.
I.
A
G
Even
ask
the
question
right
because
it
comes
back
with
an
I
I.
You
know:
I
I
would
keep
a
list
of
issuers
and
domains
and
if
I
get
a
I
get
a
request
for
that
domain
and
I
give
it
and
it
comes
along
with
the
token
I'm
gonna
check
to
make
sure
the
issuer
is
allowed
to
authorize
for
that
domain.
If
not
I
reject
it.
I
think.
A
All
right,
I
will
move
on
to
the
next
slide.
Here
then
this
is
a
another
issues
was
brought
up
by
life,
saying
that
we
want
to
remove
the
simple
URI
container
current
recently
we
added
a
hash
method
that
was
to
make
smaller
tokens.
So
if
the
URI
was
very
long,
you
could
hash
the
URI
in
as
long
as
the
hashes
match
awesomeness.
A
So,
if
we're
going
to
keep
things
simple
and
it's
something
that
nobody's
ever
going
to
use,
we
should
probably
remove
it.
It
is
kind
of
like
the
simple
obvious
case.
You
know,
if
you're
going
to
show
somebody
an
example:
hey
here's
the
literal
URI
you
want
to
put
in
it.
But
what
are
some
other
reasons
to
keep
it?
It
kind
of
feels
like
training
wheels
like
oh,
but
they're,
always
there.
While
we
want
to
take
them
off.
Well,
just
simplicity
and
not
meeting
them,
but
I'm,
not
sure
I.
A
I
D
I
Matthew
Miller
so
Mike.
My
comment
with
this
was
it
is
possible
when
you're
dealing
with
the
hash,
then
you
had
to
have
a
canonicalized
representation
of
thing.
You're,
hashing
and
your
eyes.
Don't
have
that,
so
you
can
run
into
some
issues
like
there's.
There's
the
possibility
of
case
mismatch.
I
mean
the
biggest
ones
going
to
be
case
mismatch.
I
A
I
feel
like
there's
going
to
be,
you
know
between
a
token
producer
and
a
token
consumer,
there's
going
to
be
some
kind
of
integration
step
and
it's
not
going
to
be
in
a
vacuum,
but
I
think
a
good
point
and
I
think
having
that
extra
step
there.
Even
if
you
want
to
say
hey
we're
gonna,
try
it
with
the
URI
form.
First
make
sure
that's
working
and
then
go
to
the
hash
form
would
possibly
make
it
easier
for
implementers.
Like
lay.
F
I
I
G
G
J
Yeah,
well
maybe
it
should
say
your
I
compared
the
same
and
then
point
to
the
document,
because
this
is
sort
of
a
part
of
your
right
registration
template,
but
they're
also
put
possibly
multiple
levels
of
your
I
compared
the
same
day
now
I
think
we're
gtp
is
a
bit
more
is
like
scheme.
Part
is
case.
Insensitive
course
part
miss
case
insensitive.
Okay,.
J
A
I
A
A
Yeah,
so
the
point
of
this
I
got
a
lot
of
people
saying
hey,
but
I
want
to
add
this
that
or
the
other
thing
into
this
token.
You
know
I
want
to
add
a
kink.
Can
we
put
some
kind
of
like
opaque
handle
in
there
or
something
and
I
was
like?
Oh,
that
sounds
like
tracking.
Let's
not
do
that,
but
people
were
coming
up
with
some
some
valid
reasons
to
do
this
too,
and
they,
you
know,
want
to
add
extensions,
but
things
that
not
are
not
necessarily
worthy
of
being
in
the
draft.
A
So
this
this
is
pulled
out
of
the
JW
s.
Spec
and
the
JW
suspect
references
it
directly
and
it's
essentially
a
header
that
allows
you
to
say
hey.
This
is
the
extension
and-
and
you
were
required
to
understand
this
or
invalidate
or
reject
a
token
and
it's
different
than
version.
So
it's
not
really
about
upgrading
it's
about
extending
the
current
version.
So
if
somebody
wants
to
add
something
they
could
say:
okay,
I'm,
adding
this
field
and
I'm
putting
it
in
the
creek.
A
I
That
matthew,
miller
again
so
I
think
I
mean
yeah,
I,
I
think
serrated
reiterating
some
off
awfulest
comments.
It
was
I,
didn't
see
in
a
so
a
problem
of
just
calling
this
crit.
You
can
use
it
similar
to
how
crit
is
used
for
data.
Yes,
it
it
has
the
same
meaning
here
as
it
has
their
like.
This
claim
is
not
valid
unless
you
understand
these
other
things
and
can
validate
them.
I
A
So
so
those
are
the
only
three
open
issues.
It
sounds
like
two
of
them
are
mostly
okay.
One
of
them
needs
a
little
bit
more
work.
I
want
to
put
them
all
on
the
list
and
give
people
time
to
to
look
at
them
and
provide
feedback,
especially
about
number
44,
but
I
really
really
really
want
to
be
done
with
this
and
I.
Think
a
lot
of
people
in
this
room
really
want
that
too.
A
So
so
you
know,
I
would
love
to
see
those
three
issues
get
the
gate
closed
and
we
have
a
new
version
of
the
draft
within
two
weeks.
I,
don't
know
if
that's
that's,
maybe
crazy,
based
on
the
conversations
we
had
with
the
number
forty-four,
but
that's
what
I'd
like
to
see
there
anything
that
people
think
other
areas
of
review.
We
had
GWT
experts.
Look
at
this
already.
E
I'm
gonna
move
to
the
mic.
All
right.
Can
you
see
me
yeah
yeah,
you
have
full
control
I,
don't
need
to
press
the
button
for
you,
oh
sweet,
so
I
think
it
sounds
like
a
good
plan.
I
think.
If
you
can
get
out
the
next
version,
then
I
will
I'm
the
Shepherd
I'll
go
ahead
and
do
a
quick
Shepherd
review
and
make
sure
everything's
kosher,
I
haven't
looked
at
these
latest
PRS,
so
I
will
go
over
those
and
then
once
that
looks
good
I
think
we
can
probably
go
to
a
working
group.
B
Okay,
so
the
agenda,
both
with
the
motivation
and
then
walked
through
the
updates
in
the
last
draft
and
some
examples
from
the
draft
that
we
have
here
so
really
wanted
to
lay
something
out
is
far.
What
is
the
motivation
for
this
draft
and
if
we
look
at
it,
the
first
question
really
is
that
you
know
in
the
current
RFC's
that
we
have
DCD
ends
currently
are
not
aware
of
supported
delegation
methods
by
you
CDN.
So
that
was
one
question
and
so
adding
delegation
metadata.
B
If
we
do
want
to
add
that
what
does
it
give
us,
you
know.
Basically,
it
gives
the
downstream
CDN
to
become
aware
of
any
delegation
methods
that
you
CDN
prefers
and,
and
so
basically
having
a
place
in
metadata
would
be.
One
I
want
one
area
where
the
CD
ends
can
at
least
communicate
about
this.
So
if
we
do
that,
and
obviously
you
have
a
need
for
developing
a
common
framework
to
handle
the
HTTP
delegation,
so
that's
kind
of
what
we
were
thinking
about
that.
B
B
Starters
one
which
currently
has
is
in
the
working
group
last
call
in
the
Acme
and
then
sub
certificates
is
another
option,
which
is
a
working
group
document
in
the
TLS
working
group
and
Lord
there.
There
are
drafts
out
there,
but
this
is.
It
does
not
belong
to
any
working
group.
Yet
so
we'll
have
to
see
how
that
goes.
B
So
the
the
changes
from
the
last
draft
there
were
some
updates
to
both
star
and
delegated
credentials,
so
some
minor
tweaks
there
and
then
from
the
previous
version.
We
did
not
have
lurk,
but
given
that
there's
a
log
drafts
1.2
and
1.3,
which
extends
the
TLS
options
to
to
use
log,
so
we
added
text
in
this
draft
corresponding
to
the
changes
that
have
been
come
through
the
Lord
draft,
just
at
a
at
a
very
high
level.
B
If
the
you
CDN
wants
to
delegate
content
to
the
downstream
CDN
in
this
example,
the
the
addition
that
we
have
done
for
the
lurk
is
essentially
the
support
the
meta
data
object.
So
we
would
specify
the
delegation
method
and
additional
information
would
be
provided,
such
as
the
key
server,
and
we
have
to
do
a
little
bit
more
running
of
this
concept
to
see
if
there's
anything
additional
needed
and
I
know,
Emile
is
looking
into
that
to
see
if
there's
any
other
additional
information
needed,
I
don't
know
Emil.
K
B
K
B
B
So
this
slide
six
and
seven
I'm
going
to
skip
over,
because
this
is
how
we
had
originally
proposed,
adding
a
secure
delegation
and
then
the
the
specific
methods,
but
on
the
mailing
list.
Kevin
had
some
suggestions.
So
if
we
ignore
six
and
seven,
then
basically
not
having
the
the
the
mi
doc
secure
delegation,
if
you
don't
have
that
path
and
then
the
specific
methods
an
alternate
way
to
do.
B
This
is
basically
listing
out
the
each
of
the
individual
methods-
the
for
example,
the
acne
star
delegation,
the
log
allegation
method
and,
of
course
the
sub
search
would
be
the
very
similar
to
what
we
have
here.
But
we
didn't
list
it
down.
In
this
example,
there
was
an
error,
so
I
think
this
was
Kevin
based
on
suggestion
that
you
had
in
a
mailing
list.
So
that's
kind
of
what
we're
thinking
about
as
well,
that
you
know
this
might
be
a
more
trimmed
way
of
handling.
B
E
E
E
Have
a
question:
if
no
one's
at
the
mic,
so
I
guess
I
asked
this
a
few
idf's
ago,
but
I'll
ask
it
again
now:
what
is
the
plan
for
maintaining
all
of
the
different
methods
in
a
single
document?
I
think
it
becomes
more
difficult,
as
you
add,
more
new
methods,
because
then
you
will
never
be
able
to
finish
this
document.
If
it
always
has
you
know
new
stuff
that
have
normative
references
to
something
else.
E
K
Emiliana
yeah
as
far
I'm
concerned.
Actually
we
don't
have
any
favorite
solution.
What
we
want
to
do
is
to
have
a
registry
of
the
existing
solution
on
the
minimum
metadata
to
enable
enough
spins
at
the
end,
to
inform
them
swing
above
the
meter.
That
is,
he
is
support,
and
so,
okay,
with
a
public
release
tree,
we
can
add
new
one
in
the
past
without
redoing
all
the
process,
yeah
I.
E
Think,
that's
that's
fair.
We
have
a
registry
already
for
payload
type
right
and
if
you're
going
with
metadata,
then
you
can
always
add
more
metadata.
I
think
my
question
was
more
towards.
If
acne
star
is
closed,
but
lurk
is
not
at
all
close.
Are
you
going
to
look
to
try
and
push
forward
with
an
RFC
just
with
the
metadata
for
acne,
or
are
we
going
to
wait
for
its
to
solidify?
More
now.
K
This
is
an
open
point,
and
actually
we
have
subsets
on
star
which
are
locking
group
document
in
TLS
on
acne
on
lower
key
still
pending
for
the
ITF
landing
roofing.
So
we
don't
know
how
we
have
to
at
least
so.
It's
enough
to
justify
the
racially,
but
the
work
is
part
of
those,
for
us
is
part
of
the
I
would
say
the
legacy
a
so
is
one
of
the
squeeze
at
our
envision
further
for
the
future
I'm.
E
All
for
that
I,
my
only
concern
is
how
we
progress
the
documents
and
if
we
have
to
wait
for
lurk,
that's
fine
I
just
want
to
know.
If
that's
your
plan
to
to
continue
working
on
this
until
waiting
for
lurk
to
get
a
working
group
or
an
RFC
versus
if
Acme
is
closed
and
people
want
to
use
it.
Do
we
want
to
move
forward
with
that.
K
K
E
If
you're
just
doing
metadata,
there's
a
registry
for
metadata
already,
we
use
payload
type
right
yep,
so
you
can
register
new
metadata
and
if
people
support
it,
then
they
support
it.
If
they
don't
they
don't
and
then
they
will
be
able
to
advertise
the
new
metadata
objects
being
defined
in
these
Doc's.
Now.
K
E
B
K
E
K
E
A
A
K
E
A
I
think
what
it
comes
down
to
right
now
is
if
it's
one
document
with
lurk,
yep
we're
not
going
to
do
anything
about
working
group
adoption
until
that's
ironed
out.
But
if
you
want
to
separate
that
out
into
a
separate
document
or
couple
separate
documents
that
that's
something
that
we
can
discuss,
some
less,
how
what
it
should
look
like.
But
if
you
want
to
separate
that
out
to
something
that
we
take
action
on
now,
we
could
do
that
much
sooner.
So.
K
K
B
K
E
Individual
I
agree
with
Phil
I,
don't
know
about
the
timelines
and
I
haven't
personally
looked
yet
at
the
what
stages
sub
certs
and
Acme
are
in.
So
the
the
question
of
one
versus
two
versus
three
documents.
I,
don't
feel
like
it
can
make
that
comment
on
that
currently,
but
I
think
it's
a
probably
a
useful
thing
to
have
good.
B
Letter
yeah
I
mean
those
those
two
documents
are
progressing
Rich's
here
and
he
can
confirm
that
it's
already
on
the
last.
You
know,
working
group
last
call
and
sub
certs
is
also
making
progress
in,
but
I
have
not
checked
the
email
in
the
last
day
or
two
to
see
where
it
is
now,
but
it's
also
progressing
so
I.
Think
I
expect
you
pretty
much
safe
to
move
forward
with
these
two
methods,
at
least
in
our
CDN
I
scope,
because
they
are
progressing
forward.
A
L
L
L
What
if
protein
is
a
new
extensions
to
regret,
routing
and
four?
There
are
four
figures
and
up
coming
from
requirement
that
we
are
coming
from
the
obligations
and
specification
that
will
be
more
normal
writing
an
infinitive
in
the
commune
video
rights.
So
we're
going
to
go
to
graphs.
One
relates
to
talking.
L
L
L
L
What
we
are
requirement
was
to
I'd
be
able
to
advertise
the
little
egg
target
to
the
use
in
the
end,
and
the
dumb
songs
in
the
end
is
telling
the
optimist.
In
the
end
where
to
send
traffic
to
the
redirect
house,
it
can
vary
by
footprint
for
different
regions
and
different
places.
We
might
have
different
directorates.
The
direct
a
legit
can
change
over
time
for
different
reasons,
and
it
should
be
possible
for
the
dancers
CDN
request
out
to
resolve
the
absolute
we're
electing
others
from
directly
from
the
county
address.
L
L
The
direct
target
objects,
the
DNS
targets
target
properties
is
just
the
host
them.
Basically,
this
is
the
hospital
that
you
use
for
the
cname
to
delegate
the
traffic
to
the
damsels.
In
here
we
have
the
HTTP
the
target
properties.
It
has
a
cost
which
is
the
hostname
or
IP
address
with
the
optional
port.
Basically,
it's
it's
an
endpoint
and
we
have
a
past
perfect
that
allows
the
decision
to
request
for
a
specific
path,
prefix
to
be
added
or
the
original.
H
E
E
L
So
that's
was
just
a
comment
for
the
actual
text
in
the
document
and
rather
than
the
in
the
draft,
rather
than
how
the
object
looks
like
okay,
okay,
so
maybe
that's
the
question
for
everybody
else.
Does
anyone
see
they
need
to
have
a
more
flexible
configuration
that
we
are
then?
What
we're
proposing
here.
L
L
L
L
We
just
think
a
way
to
select
the
content
for
triggers
by
regex
and
by
video
playlist,
and
we
are
going
to
add
generic
extension
objects,
which
is
a
list
of
general
extension
objects
that
allows
you
to
do
all
kind
of
specific
manipulation
over
the
target
trigger
for
currently,
we
propose
first,
two
initial
ones,
which
is
the
location
policy
and
policy
and
well
adding
the
support
or
FCI,
without
meant
for
the
extensions
next
life
is,
and
so,
let's
start
with
versioning
we're
introducing
via
v2,
which
has
three
new
healthy
new
properties.
It's
the
fondant
projections.
H
C
H
L
The
list
of
extensions
in
order
to
be
able
to
use
this
trigger
v2
we're
introducing
version
2
of
the
interface
we
have
the
ID
radio
commands
me
to
that
uses
bigger
B
to
instead
of
trigger.
We
have
the
I
finger
status,
v2
that
using
finger
to
do
instead
of
tuna
and
the
CI
trigger
collection,
has
no
change
so
and
with
adding
an
FC
I
and
to
advertise
each
version.
Unsupported,
we'll
talk
about
that
later.
L
What's
important
here
is
to
understand
the
versioning
scheme
and
what
were
in
town
suggesting
is
those
three
interfaces,
the
CI
trigger
command
V
through
the
statue
in
the
status
v2,
and
this
bigger
connection
together
defined
this
version,
two
of
the
interface
of
the
interface
and
when
somebody
says,
hey,
I'm,
supporting
a
vision
tube.
It
means
they
supposed
o
three.
L
L
Basically,
we
add
a
property
called
contradictions,
which
is
a
list
of
bring
it's
match
objects,
the
logics
matched
object,
has
expression,
K
sensitive,
a
boolean
that
sensitive
or
not,
and
a
boolean
that
says
you're
gonna
do
match
matching
on
the
first
being
or
not
so
the
object
looks
like
the
one
here.
We
have
an
example
for
reject
the
case
sensitive
in
the
match,
burning
unions,
taking
the
feedback
that,
with
neurons
point
about
metrics
lets
it
be
attack.
We
probably
need
to
think
about
that
here.
Also.
L
We
know
that
the
parties
that
passes
this
value
between
them
is
the
usually
ended
this
idiom,
which
are
supposed
to
have
trust
relationship
with
Windham,
but
this
can
also
be
delegated
along
the
paths
of
two
other
decisions,
and
we
need
to
think
about.
The
implication
admitted
I
haven't
thought
about
that
yet
I'm
taking
it
to
the
pigment.
Do
you
think
about
I?
Think
wow
use
things
next
week,
so
the
next
way
to
select
content.
L
A
requirement
use
playlist.
The
main
reason
we
want
to
do
that
is
better
positioning.
It
is
very
easy
for
video
applications,
video
CDN
and
video
caches
to
talk
in
the
language
of
at
least
so
we
added
contemplative
operators
and
from
and
that
playlist
has
the
following
properties.
It
has
playlist
and
then
the
type
of
ABR
protocol,
as
you
can
see,
on
the
on
the
right
side
sample
the
ABR
protocol
comes
or
a
new
registry.
L
Hr
at
an
expense
and
of
course
you
can
add
more
in
the
future.
If
there's
any
reason
to
do
so,
one
thing
back
that
I
got
and
I
think
starts
with.
We
want
to
use
just
the
names
Internet
and
the
or
handed
version
into
them,
which
means
that
in
the
future,
wonder
the
nature
like
v8,
we're
going
to
add
another
one
to
the
registry.
E
Think
well,
with
all
these
things,
the
first
version
you
can
really
do
whatever
you
want
and
it
only
matters
when
it
gets
revved.
So
it's
probably
ok
but
I
wouldn't
mind
seeing
your
version
there
and,
as
I
think
about
it
now,
I
think
a
B.
Our
protocol
is
probably
not
a
good
is
not
as
generic
a
name
as
we
can
go
with.
It
could
just
be
something
else:
protocol
or
protocol.
L
L
L
Generate
extension,
so
I
follow
the
line
of
the
generate
metadata.
It's
pretty
much
the
same,
but
after
I
even
thought
about
giving
giving
them
instead
of
defining
new
ones.
Now,
but
I've
been
talking
with
Kevin
about
that,
we
agree
that
it
makes
more
sense
to
and
separates
and
all
this
definition,
but
they
look
very
much
like
they've
benefited
ain't
I,
so
we
have
an
extension
type.
We
have,
it
will
be
taking
value
the
moniker.
It
wouldn't
fall
straight
when
we
distribute
and
incomprehensible
and
one
thing
that
that
I
got
from
Kevin
and
I
thought.
L
The
dispatcher
point
is:
do
we
actually
need
to
say?
Can
we
distribute
so?
Do
we
think
confused
cases
in
which
this
BN
one
passes
between
you
command
the
next
one?
Also,
and
do
we
think
of
these
cases
in
which,
if
the
N
one
can
comprehend
behavior
for
a
missing
person,
throw
together
a
specific
extension
object?
Is
it
safe
for
him
for
that
there
are
you
bein,
want
to
pass
it
on
to
the
community
I
think
it
is
because
I
have
an
example.
E
So
re
I
think
your
your
example
was
valid.
I
think
that
one
made
sense
I
think.
The
only
thing
that
concerned
me
was
that
the
immediate
D
CDN
is
still
going
to
return
that
it's
failed
right,
even
if
it
passes
it
downstream,
and
it's
not
clear
to
me
that
the
ultimate
you
CDN
can
tell
where
it
failed
and
where
it
didn't.
E
L
E
E
L
L
But
that
means
that
we
mean
separate
meters
for
each
and
every
time
zone,
because
there
is
no
inclusion
of
extending
one
side,
the
other
we
haven't
done.
It's
done
so
I've
never
done
them
yet
so,
assuming
we
are
not
going
to
do
inclusion
of
extensions
or
billion
expansion,
which
means
that,
if
we're
going
to
do
for
it
4:00
a.m.
in
the
morning
for
each
time
zone,
we
need
run
because
we
give
permission
every
arms
of
the
other
eternity.
L
A
Don't
over
speaking
as
an
individual,
so
it
seems
like
time
policy
and
location
pause.
They
are
really
like.
You
have
to
go
into
so
much
detail
to
cover
every
everything
and
then
local
time
doesn't
give
you
a
lot
of
options.
It's
just
saying:
4
a.m.
local
time.
It
might
be
nice
to
do
a
default
policy
of
4
a.m.
local
time,
and
then
you
know
for
a
given
location.
Here's
a
here's,
an
alternate
time
policy
to
do
overrides.
L
A
Don't
really
have
see
it
in
any
particular
way.
I
can
maybe
come
up
with
something
an
example
for
you,
but
I,
don't
I'm,
just
thinking
that
time,
policy
and
location
policy
is
is
overly
specific
and
location.
Local
time
is
very
generic
and
it
would
be
nice
to
be
able
to
use
both
I
have
one
and
say
this
is
the
default
is
local
time
or
maybe
just
to
be
implicit?
The
default
is
local
time.
E
Yeah
I
think,
if
you
take
the
time
policy-
and
you
add
a
boolean
attribute
that
says
this
is
local
time
and
it
can
default
the
true
or
false,
and
if
it
is
local
time,
then
you
know
at
4:00
a.m.
your
time.
Do
this.
Otherwise
excuse
me
otherwise
it's
UTC
I
think
we
can
support
both,
but
it
makes
sense
because
of
the
use
case
for
local
time
to
support
local
time.
In
my
opinion,.
A
L
So
the
location
both
the
object
has
one
property.
It's
called
locations,
it's
a
list
of
occasions
for
each
location.
We
have
an
action
which
is
allow
or
deny
which
basically
means
include
or
exclude,
and
maybe
this
was
like
kind
of
copic
in
her
face
from
the
metadata.
Maybe
it
makes
no.
Actually
we
leave
it
like
that.
I'm
using
the
location,
rule,
objects,
omitted
anything
I
think
there
is
no
good
reason
to
define
the
new
object
object.
L
L
L
If
anybody
thinks
there
is
a
good
reason
to
do
that,
for
example,
if
you
want
to
do
something,
but
you
want
to
do
it
always
between
every
day,
between
6
a.m.
to
7
a.m.
in
the
morning,
then
maybe
we
bring
a
list
of
Windows
I,
didn't
think
that,
for
our
use
case
make
sense,
but
if
anybody
helping
throw
them,
let's
discuss
it.
L
E
L
L
The
versions
are
just
strings.
You
can
put
anything
in
the
next.
Is
the
playlist
protocols
again
I'll
change
that
to
media
protocols
instead
of
Sylve
BR
and
here
I
give
a
list
of
the
support
protocols
and
if
we
go
for
versions
which
I
mean
that
one
now
it's
gonna
be
it's
version
7,
but
in
the
future
of
urban,
a
you're
going
to
have
to
state
each
and
every
version
that
to
support
and
the
last
one
is
the
generic
conventions.
A
A
A
So
so
we
have
a
plan
for
URI
signing.
We
have
a
path
forward
for
HTTP
delegation,
specifically
we're
not
sure
we
need
to
figure
out
how
we're
gonna
go
with
the
number
of
drafts
and
the
content,
but
we
have
a
path
forward
and
then
extension,
so
I
believe
it
is.
The
the
conversations
had
about
the
triggers
draft
was
that
we
would
need
to
reach
ardor
if
we
were
going
to
adopt
that.
E
A
A
A
A
E
Then
the
the
chairs
have
a
little
bit
of
work
to
do
and
and
I
think
the
draft
has
some
work
that
it
needs
as
well.
I,
don't
think
we
have
to
address
adoption
for
that.
One
I,
don't
think
we're
ready,
but
the
chairs
will
go
ahead
and
start
looking
at
what
it
will
take
to
recharter
to
get
a
new
version
of
triggers.