►
From YouTube: IETF115-JMAP-20221110-1300
Description
JMAP meeting session at IETF115
2022/11/10 1300
https://datatracker.ietf.org/meeting/115/proceedings/
B
A
C
A
B
D
F
G
B
A
B
Quickly,
we're
doing
this
double
X
very
well,
I
mean.
Should
we
get
straight
into
the
first
item
on
the
agenda,
which
is
blob,
I,
don't
have
any
slides
for
blob.
I
had
one
review
from
sector
which
I
have
replied
to
and
I've
submitted
a
new
version
with
the
fixes
from
that
I.
Don't
think,
there's
anything
else
to
say.
Unless
our
area
director
has
any
comments
on.
I
H
I
J
J
Yeah
so
I
don't
know
if
people
notice
there
is
a
an
email
to
the
list.
Someone
had
requested
two
other
items
in
there,
which
I
think
were
probably
better
off.
As
extensions
I
forgot
to
tell
my
head
what
they
were.
Oh
one
of
them,
they
wanted
ability
to
recall,
failed
action.
Why
why
an
action
had
failed
through
this
protocol?
J
I
think
it's
probably
too
far
along
to
add
that
now
it
seems
right
for
an
extension.
If
somebody
wants
that
the
only
other
issue
that
was
brought
up
that
might
want
to
be
discussed
here
currently
to
deactivate
a
script,
you
basically
use
the
activate
script
with
a
empty
string.
The
request
was
to
have
a
second
another
property
which
specifically
disables
a
script
as
opposed
to
using
enable
with
an
empty
string
which
follows
what
manage
Civ
did.
I
have
no
strong
opinion,
either
way.
I'll
leave
it
up
to
the
group
to
decide.
J
B
I
J
Yeah,
the
the
spec
as
I
wrote,
it
was
following
the
model
outlined
by
manage
Civ
for
better
for
worse,
and
this
model
is
exactly
what
that
does
again.
If
people
feel
strongly
about
having
another
property
I'm
willing
to
do
it,
but
at
this
late
in
the
game,
if
there
aren't
strong
opinions,
I'm
inclined
is
to
leave
it,
as
is.
A
K
Hello,
first
of
all,
it
shouldn't
accept
empty
string,
the
spec
actually
says,
or
null
it's
ID
or
null,
and
an
empty
string
is
not
an
ID
and
it
is
not
null
so
that
that
should
be
an
invalid
argument.
So
if
you
sent
that
the
advantage
to
having
a
separate
property
you
I
could
see
would
be,
if,
if
you
don't
want
it
to
change,
you
don't
have
to
admit
to
the
whole
thing.
You
can
always
submit
not
every
argument
which
I
could
see
being
potentially
a
thing
for
strictly
typed
languages.
K
L
K
K
B
M
B
M
Hi
assistant,
York,
I,
remember
Ken.
That
is
quite
some
time
ago.
I
think
we
had
a
very
small
exchange
on
a
minor
topic
on
this
civ4j
map
thing
and
I
just
relooked
it
because
it
somehow
went
dormant
a
little
bit
and
it's
related
to
what
we
discussed
in
Vienna
about
Sif
in
in
non-j
map
context
with
you
know
how
people
build
safe
rules
together
as
a
system
from
what's
coming,
basically
and
I.
M
M
Yeah,
so
the
background
of
the
question
is
so
there
is
so.
The
observation
is
that
empirically
in
practice,
Sif
is
very
much
structured
by
the
use
of
the
underlying
software.
So
there
is,
for
instance,
certain
web
mailers
that
have
a
very
clear
structure
of
the
Sif
rules
which
they
create,
and
it's
helpful
for
a
consuming
client
to
know
that,
and
typically,
when
you
use
manage
sift
directly,
you
have
a
greeting.
So
you
know.
M
Actually
this
is
coming
from
certain
server
implementation
like
a
Zimbra
or
whatever
kind
of
thing,
and
you
can
do
actually
apply
your
changes.
Sometimes
you
have
it
within
the
sift
script,
but
in
case
of
jmap
I
think
it
might
be
the
case
that
search
for
jmap
might
be
wrapping
existing
underlying
skim
servers
so
that
information
somehow
is
currently
not
able
to
be
retrieved
in
my
opinion,
and
it
would
be
good
to
have
maybe
a
mechanism
for
that.
J
B
K
B
K
K
Cool
clever,
so
I
think
this
was
hopefully
going
to
be
fairly
brief.
So
we
can
move
us
along
I,
published
version
nine
draft
about
a
month
ago,
which
had
a
few
updates
in
it.
So
the
probably
the
biggest
one
is
we've.
Had
this
big
discussion,
I've
been
calyx
about
adding
schedule
ID
to
JS
calendar
to
participants
which
identifies
that
participant
kind
of
globally.
K
This
would
often
just
be
essentially
a
mail
to
URI,
with
someone's
email
address
or
something
like
that,
but
basically
separating
out
the
ID
of
a
participant
from
the
methods
of
contacting
them,
which
is
in
the
sanitary
property,
and
why
this
is
important
for
jmap
calendars
is.
We
need
to
be
able
to
look
at
an
event
and
go
which
of
these
participants?
Is
you
the
logged
in
user
and
before
we
were
looking
through
all
of
the
send
to
options
and
going
would
I
received
any
of
these?
K
But
this
kind
of
simplifies
that,
because
we
can
now
just
look
at
the
schedule
ID
and
go,
is
this
I'd
ID,
one
of
the
user
schedule
IDs?
You
may
have
multiple
because
you
might
have
like
different
email
addresses
that
aliases
they're.
All
you,
but
it's
still,
comparing
you're
not
doing
a
nested
Loop
kind
of
comparison.
So
it's
it's
a
simplifies
things,
but
I
think
so
that
was
the
biggest
change
we
also
added.
This
is
origin
Property.
K
We
discussed
this
at
the
last
ITF
meeting,
which
is
just
the
server
notifying
putting
a
property
on
the
event
to
say
it
is
the
authoritative
source
for
this
event.
So,
if
you
make
changes,
it's
going
to
send
out
it
messages
to
all
the
other
participants
and
expect
to
receive
RSVPs
as
opposed
to.
If
that's,
of
course,
then
you've
received
this
from
someone
else
and
you'll
be
sending
out
RSVPs
back
to
them.
K
Finally,
the
other
kind
of
bigger
change
was
moving.
This
Sherry's
actors,
property
which
simplified
the
spec.
We
added
this
originally
to
give
flexibility
for
kind
of
different
modes
of
sharing,
but
we
looked
around
it
didn't
seem
that
the
the
two
modes
were
not
that
different
and
didn't
seem.
Anyone
actually
needed
both
and
from
existing
implementations
and
removing
complexity
is
I,
think
a
worthwhile
thing.
So
we
took
that
out
and
then
I
added
quite
a
few
API
examples.
So
with
that,
I
believe
that
I'll
mute
the
next
slide.
This
is
basically
finished.
K
I
need
to
fix
up
some
editorial
stuff,
including
section
references
that
are
missing
and
just
say
section
XXX,
rather
than
property
linking,
but
do
all
that
and
we're
gonna
definitely
need
to
wait
to
ensure
the
JS
Canada
schedule.
Id
stuff
is
finalized
because
we
can't
have
that
critical
to
how
a
lot
of
this
works.
So
we
can't
have
that
changing
underneath
us,
if
that
doesn't
end
up
doing
what
we
think
it's
going
to
do
so.
K
H
E
Are
a
lot
of
commonalities
yeah,
so
one
thing
I
stumbled
upon
is
the
calendar
preferences
object
for
now.
I
rev
refrain
from
tagging
along
here.
So
my
take
on
this
in
in
jmap
mail,
there's
a
role
property
right.
What
the
folder
is
supposed
to
be.
Is
it
an
inbox?
Is
it
a
draft
and
for
calendars?
This
also
was
there?
Do
you
have
the
role
property
of
a
of
a
calendar
folder?
E
Was
it
I
think
it
was
inbox
only
for
now
right
basically
to
to
say
this:
is
the
default
folder
of
a
user
and
in
the
newest
version?
You
move
this
to
a
separate
object,
so
I
just
wanted
to
to
understand
the.
So
what
I?
Don't
really
like
about
this?
That
we
now
have
two
approaches.
One
is
in
in
mail.
This
is
inside
the
folder
object.
It's
it's
a
it's
a
one
property
and
in
calendars.
This
is
a
separate
object.
Yeah.
K
G
Yeah,
so
regarding
the
in
short,
Chase
kind
of
work
is
finished
part
and
the
dependency
now
of
gmap
calendars
requiring
scheduler
D
to
be
defined
for
JS
calendar.
So
on
Tuesday
at
calex
we
were
setting
the
deadline
for
GS
cannabis
to
Summer
next
year
and
so
I
and-
and
this
was
done
when
we
were
under
the
assumption
that
there
is
no
hard
dependency
of
jmf
calendars
on
JS
calendar
base.
B
G
G
Well,
we
because
we
said
we
only
want
to
have
one
default.
One
of
the
reasons
was
that
it
is
that
it
was
more
straightforward
to
us
to
have
one
property
where
this
is
defined
rather
than
implementations,
requiring
to
look
at
all
the
roles
of
their
calendars
and
make
sure
that
their
requirements
are
satisfied.
G
K
E
So
one
insurance,
so
on
crazy
thing
one
could
do
is
basically
have
both.
You
know
like
it's.
The
calendar
preferences
object
is
already
defined
as
an
extension
to
the
core
calendar
spec,
so
the
system
could
say
Okay
I
have
this
I
only
implement
the
core
capabilities
and
I
only
have
the
role
inside
the
as
a
property
inside
the
folder
or
you
could
say.
Oh
I
also
have
this
fancy
Kanda
preferences
object
and
you
don't
need
to
go
through
all
the
calendar
folders
to
find
out.
What's
the
default,
you
just
fetch.
E
D
K
M
Hi
this
is
Sans
York
I
didn't
have
the
time
to
look
it
up.
Direct
Kings
back
and
I.
Don't
know
it
by
heart,
maybe
it's
inside,
but
it
sounded
a
little
bit
like
what
Robert
said
like
the
default
folder,
you
see
only
role,
you
have
in
mind
here
for
calendars.
Is
it
correct
that
there's
just
one
such
role,
which
you
currently
consider.
M
So
point
I
would
like
to
make
is
that
Sam
might
also
encounters
be
other
roads
like
you
know,
you
have
this
kind
of
auto-generated
birthday
calendars
and
there
is
I
think
two
or
three
other
examples.
I
would
need
to
look
that
up.
So
probably
that
would
speak
more
for
World
Properties
and
if
there's
multiple
ones,
then
you
know
having
a
fixed
set
of
properties
in
the
calendar
preferences
which
you
had
to
extend
I,
don't
know,
I
need
to
look
it
up,
but
just
want
to
make
the
point.
M
D
K
B
K
I
mean
this
has
even
less
to
say
about
it,
because
it's
basically
I
did
publish
an
update,
but
it
was
mainly
kind
of
going
to
and
doing
copy
edits.
There
was
only
one
actual
change,
I
think
which
was
adding
this
accounts
property,
which
is
to
the
principal
object,
and
basically
this
was
about
I
realized.
If
you
were
the
sharing
information
from
calendars
and
mail
and,
like
you
know,
different
things,
they
were
all
duplicating.
K
This
whole
account
object
inside
them,
which
was
getting
messy,
and
you
know
needless
duplication,
so
kind
of
move
that
up
to
top
level
property
that
the
other
ones
could
just
reference.
The
idea
of
that's,
that's
it
and
I
think
you
know
this
is
a
fairly
much
simpler
spec.
In
some
ways
it
was
extracted
out
of
calendars
because
we
were
so
it
was
pretty
generic
across
the
different
things
you
could
do.
K
B
J
K
So
it
kind
of
it's
a
slightly
abstract
spec.
It
defines
the
things
that
are
common,
that
you
also
have
to
say
I'm
going
to
share
this
stuff,
because
you
have
to
Define
some
things,
but
you
certainly
could
apply
to
mail,
but
that
would
actually
be
a
separate
spec
in
terms
of
letting
you
manage
your
mailbox
sharing
using
this
model,
but
yeah
I
think
it
definitely
could
be
published
actually
because
it's
it's
not
dependent
on
it's
the
other
way
around
calendars.
One
is
dependent
on
this
one.
B
B
C
Hello
sure
hello
turn
off
on
my
screen:
yeah,
hello,
guys.
This
is
Renee,
so
I've
been
about
Gemma
coupon.
Actually
it's
in
last
call
as
well,
so
it's
gonna
be
a
pretty
quick,
so
yeah
there
is
some
minor
tapos
to
to
fix.
Murray
thanks
for
reviewing
it,
I
think
I
will
I
will
push
something
before
the
end
of
the
of
the
week,
then
yeah
we're
just
saying
as
well.
C
C
C
All
right
so
I
just
recorded
the
video
I
want
to
be
sure
to
not
have
problems
and
make
it
fast,
because
we
know
how
it
goes
right
during
demos
in
real
time,
so
yeah,
just
a
user
with
his
mailbox.
Just
so
getting
his
quotas.
He
has
nokutas
for
no
defined.
The
list
is
empty,
doing
a
good
so
checking
with
the
just
a
web
admin
from
the
server
side
that
there
is
no
quota
setup.
M
C
C
C
C
Let's
do
it
now:
okay,
yeah
sure.
So,
regarding
gemap
identities,
when
our
front-end
team
started
implementing
the
identities
with
their
with
the
mobile
application,
they
ask
us
yeah.
We
don't
see
anything
to
get
the
foot
identity
to
show
it
like
the
users
have
multiple
identities
and
after
checking
against
the
spec,
realized
that
Yes.
C
Actually,
there
is
a
nothing
defined
in
jmap
for
that,
so
we
pulled
it
a
little
bit
and
what
we
think
is
just
adding
salt
or
the
field
like
for
the
mailbox.
For
example,
where
I
mean
the
lowest
number
would
be
your
first
identity,
so
it
would
be
the
default
one
to
show
and
then
the
higher
it
goes
and
you
get
the
list
of
identity
depending
on
the
of
the
ore.
E
E
E
C
C
K
You
have
to
order
property
that
would
imply
to
me
that
you
expect
me
to
be
able
to
manually
order
these
things.
I
mean
we
can
do
that
in
your
product,
but
that
would
be
surprising
to
me
like
I've,
not
like
I've,
seen
lots
of
things.
Let
you
drag
and
drop
mailboxes
around
to
to
find
a
custom
sort
order,
but
when
you
have
a
list
of
like
things
or
identities,
generally
they
just
sorted
alphabetical
or
whatever
by
a
client.
So
if
you're
not
going
to
support
that,
that
seems
adding.
C
K
So,
but
so
specifying
default
is
fine,
but
that
would
just
be
like
we
were
discussing
with
the
calendar
stuff
like
a
single
property,
as
opposed
to
a
sort
of
defect
means
that
every
single
identity
you
have
has
some.
You
know
ordering
that's
strictly
defined,
and
that
starts
to
quite
context,
I
think,
probably
more
than
it's
that's
very
important.
That's.
F
F
When
I'm
replying
to
a
particular
person,
I
mean
they
start
to
get
a
little
bit
fiddly
and
I'm,
not
saying
that
you
want
to
specify
all
of
these
things,
but
but
but
really
what
you
want
is
you
want
default
and
non-default
in
a
given
context,
and
so
I
think
you
could
say,
here's
the
generic
context.
If
you
don't
know
what
context
you're
in
here's
the
default
specify
that
here,
it's
just
a
flag
and
then
that
leaves
room
for
somebody
to
say.
F
C
C
K
C
A
B
E
I
just
wanted
to
mention
one
thing
for
for
user,
as
it
is
right
now
in
jmap
you,
you
could
just
come
up
with
your
own
extension
right,
your
own
capability
and
then
yeah.
Do
that
but
yeah,
of
course,
if
there's
interest
that
makes
sort
of
I
think
it
makes
sense
to
discuss
it
in
the
IDF
context
to
come
up
with
a
well-defined
thing.
Yeah
just
want
to
mention
that
possibility
in
case
some
people.
B
E
E
The
idea
here
for
tasks
is
to
make
it
a
common
standard,
not
only
for
grouper
systems
focused
on
mail,
but
also
for
kanban
boards
or
issue
trackers.
This
is
one
of
the
core
things
we're
working
on,
but
of
course,
we
also
want
to
include
grouper
next
slide,
please
yeah.
So
what's
new
in
the
newest
version,
as
I
said
before,
I
tagged
along
with
most
changes
in
calendars,
then
I
worked
in
some
feedback
from
the
last
ITF.
E
One
of
them
was
that
estimated
work
should
only
mention
the
direction
and
not
I,
don't
know
what
did
I
have
Fibonacci
scale
or
something
so
now
it's
only
defined
larger
means
more
work
for
the
estimated
work
property,
which
is
a
property
of
a
task.
Another
thing
that
was
mentioned
at
last
ITF
was
that
there
are
now
we,
the
the
tasks
back
now
has
a
bunch
of
new
extensions,
defined
or
or
capabilities
defined.
E
F
E
So
we
have
a
bunch
of
groups
for
use
cases
that
are
using
a
common
feature
set
and
basically
the
ideas
that
implementers
can
then
choose
which
group
they
want
to
implement
and
last
ITF
there.
It
came
up
that
there
was
some
some
clarification
needed
so
now.
The
text
also
includes
it
or
mentions
that
in
case
a
capability.
E
What
happens
when
a
client
is
using
an
unsupported
capability?
Then
the
server
should
throw
an
unknown
capability
errors
or
it's
just
as
it's
defined
in
the
in
the
jmap
calls
back
already.
It's
just
mentioned
again
in
this
spec
yeah
and
what
happens
if
a
client
uses
extensions
that
the
server
doesn't
understand
in
this
case
the
server
should
store
those
properties
and
just
ignore
them.
That's
yeah
how
I
extended
the
spec
or
that's
I,
clarify
that
in
respect.
E
Okay,
next
slide,
please
so
one
thing
I
want
to
discuss
about
today
is
what
could
be
new
in
Jama
for
tasks
version
6.
here
you
can
see
a
task
from
Trello
and
it
contains
actually
two
checklists
that
are
inside
this
task.
One
checklist
is
called
keep
in
mind,
and
then
you
see
two
checklist
items
that
are
not
done
yet
and
the
other
one
is
another
checklist
with
two
other
check
those
items
here.
So
you
there
are
actually
systems
where
you
can
have
this
as
a
feature
and
actually
widely
used
systems.
E
Next
slide,
please
yeah,
so
it
would
be
nice
to
have
that.
Also
in
general
tasks,
we
see
that
in
Google
workspace
Microsoft
to
do
planner,
Factor
click
up
a
lot
of
systems
use
checklists,
not
a
lot
of
views.
Multiple
named
checklists
assigned
to
a
single
task.
So
as
I
here
I
have
the
observed,
features
in
those
systems
for
now,
I
only
found
those
that
feature
within
Trello,
but
Trello
is
an
important
server.
E
Most
most
systems
have
a
title
for
checklists
or
that's
actually
a
checklist
item
that
has
a
title
or
all
I'll
check.
This
items
have
a
title.
Actually,
so
all
Checkers
items
have
a
simple
done
or
not
done.
State
there's
a
position
last
modified
stamp,
that's
only
visible
in
the
UI
usually
and
an
sine
is
also
sometimes
used.
So
we
have
a
bunch
of
features
for
this
functionality
that
are
very
common
common
or
not
so
common,
but
still
important
next
slide,
please.
E
So
the
question
here
is
how
to
actually
support
that
here
is
an
example
how
to
do
that
in
I
calendar.
Somehow
this
is
from
feedback
that
we
got
from
one
of
the
task
list,
the
task
system,
vendors.
So,
on
the
left
side
we
see
the
first
vehicle
in
the
item,
which
is
basically
a
checklist
that
is
in
relation
to
another
another
task.
E
Another,
oh
sorry,
it's
not
a
v
event.
That's
my
mistake.
It's
a
v
to
do
that's
in
relation
to
another
to-do
item,
which
is
a
checklist
item.
So
what
you
can
use
in
iconetter
right
now
is
to
have
related
to
basically,
and
then
you
have
the
checklist
as
a
special
kind
of
task.
That
is
actually
a
folder.
E
Okay,
the
question
is:
can
we
do
better
than
or
as
and
also
there's
this
X
type
thing
that
also
defines?
Is
it
a
checklist
or
not
so
the
questions
can
we
do
better
than
that
and
basically
the
idea
here
is
to
find
out.
How
do
we
do
that
in
jmap
tasks?
If
you
have
some
opinions
on
that
so
next
slide.
Please.
E
So
my
first
proposal
is
to
since
checklists
are
so
simple,
basically
and
check
items
are
also
very
simple.
We
could
just
have
a
tasks
on
the
right
hand,
side,
and
we
could
just
say
a
task
can
also
contain
most
one
or
more
checklists,
basically,
which
is
an
own
object,
and
this
checklist
can
then
contain
all
objects
on
itself
on
its
own,
which
are
check
items,
and
so
both
of
them
are
very
simple
objects
and
the
task
can
have
those
inside
the
checklist
property.
E
E
So
that's
like
basically
the
second
proposal,
yeah
I,
also
there's
also
an
alternative
that
you
could
add
another
add
type,
maybe
as
well.
So
it's
not
just
a
simple
task,
then
it's
it's
the
whole.
The
whole
task
object
then
gets
checklist
object
as
an
as
a
type
or
a
checklist
item
object
as
a
type
just
yeah,
but
I'm
not
sure
about
this
idea.
Just
put
it
on
there.
So
next
slide.
Please
now
the
question:
do
you
have
any
preference
between
those
two
things?
I
hope
you
were
able
to
follow.
G
A
G
G
D
G
A
E
A
A
So
if
you
have
a
client
that
only
that
does
not
support
multiple
checklists
yeah,
do
we
have
a
way
of
determining
the
client
capabilities
and
somehow
handling
that
or
does
do
like
the
the
second
and
subsequent
checklists
just
disappear
or
or
what
I
mean
that
that
would
seem
to
be
bad?
It
would
be
better
if
you
could
even
take
them
multiple,
checklists
and
and
merge
them
together
somehow
for
representation.
If
the
client
doesn't
have
that
capability,
you
don't
want
to
just
lose
the
lose.
The
other
checklists
yeah.
E
A
E
I
think
the
question
here
is
how
fine-grained
we
want
this
capability
mechan,
but
okay,
what
we
could
do,
maybe
is
have
like,
like
a
flag
of
the
capability
to
say:
oh
I'm,
a
server
and
I'm
using
multiple
checklists
by
the
way,
so
please
make
sure
make
sure
you
handle
this
somehow.
This
is
what
I
can
imagine
here
in
this
scenario,
and
then
the
client
sees
that
that
flag
and
then
says:
okay.
O
E
Just
yeah,
that's
the
thing
like
a
checklist
item
would
then
be
much
simpler
right.
So
I
try
to
keep
it
as
simple
as
possible
in
this
checklist
item
to
not
pull
in
the
whole
complexity
of
big
blown
task,
basically
so
yeah.
Well,
it's
just
much
simpler.
The
assignee
is
just
a
user.
I
user
ID
actually
needs
to
be
a
participant.
Then
right,
yeah,
you're,
right,
that's
or
it's
a
yeah
anyway.
The
goal
is
to
make
it
much
simpler.
It's
not
perfectly
written
down,
yet
I
didn't
put
it
in
the
spec.
E
Yet
it's
it's
yeah.
Try
to
gather
some
feedback
on
how
I
approached
it.
So,
on
the
on
the
left,
hand,
side
I'm,
not
sure
if
I
actually
explained
that
well
each
the
task,
the
checklist
to
check
this
item.
It's
it's
each
a
full-blown
task.
Basically,
you
know
the
big
difference
is
the
relation
between
them
and,
on
the
right
hand,
side.
The
checklist
item
is
much
simpler.
It's
not
a
full-blown
task.
You
cannot
do
all
the
things
with
it
that
you
can
do
with
a
task.
E
E
G
But
probably
there
is
a
middle
ground
where
one
can
Define
what
parts
of
the
scheduling
of
a
task
must
be
implemented
for
to
achieve
checklist,
for
example.
So
maybe
one
could
still
use
the
the
already
defined
scheduling
properties
and
just
not
require
applications
to
implement
all
of
scheduling.
If
I
understand
this
that
this
might
be
too
much
for
some
applications.
F
A
E
M
This
is
santiarc
two
Jim's
question,
I
think.
Actually
the
Trello
case
was
a
multiple
checklist
seems
to
be
like
yours
at
a
little
bit
of
a
particular
special
case
within
those
supporting
checklists
at
all,
I-
probably
personally,
not
over
engineered
here.
In
that
extent,
so
I
don't
think
we
necessarily
need
to
have
a
capability
or
something
like
that
in
the
standard.
I
think
what
you
need
to
consider
a
little
bit.
We
are
typically
coming
as
Rodrigo
from
the
portability
side.
M
So
we
see
this
also,
as
in
from
the
data
format,
as
an
exchange
format
that
you
can
use
to
import
export
data
from
different
systems,
so
interoperability
between
different
systems
is
obviously
something
you
can
also
later
on
the
news
with
that.
What
is
actually
where
is
initially
coming
from,
so
we
sometimes
might
have
a
slightly
different
perspective
where
we
think
about
I
would
then
see
this
as
an
issue
for
the
client.
M
M
Is
you
know
we
are
even
a
third
party,
usually
in
the
middle,
so
we
have
said
very
often
that
we
migrate
from
systems
that
support
whatever
infinite
number
of
phone
numbers,
whereas
many
systems
only
support
three
of
them
or
something
like
that,
and
then
we
do
something
like
you
suggested.
We
merge
it
together
or
something
like
that
so,
but
this
is
nothing
that
strictly
I
think
needs
to
be
within
the
spec.
It's
it's
up
to
the
import
export
mechanism
at
stage
stage.
Okay,
that
was
to
Jim
to
that
slide.
M
So
that's
a
similar
thing,
sometimes
in
a
calendaring
when
you
have
modified
events
and
so
on
in
the
Stream
of
all
instances.
So
that's
sometimes
a
little
bit
tedious
from
a
client
side.
I'm,
not
sure
if
the
jmap
batch
API
would
support,
for
instance,
currently
to
retrieve
something
like
a
task
and
its
related
items.
That
would
be
an
interesting
question
to
people
that
know
more
about
the
nested
gmf
stuff
here.
B
Okay,
thank
you
next
on
this
list,
so
yeah.
What
we've
done
for
similar
types
of
things,
I
think
this
came
up
in
JS
contact
for
the
names
as
well.
If
you
don't
know
what
all
the
different
parts
are,
there's
still
an
array
of
items
kind
of
in
order,
and
it
seems
to
find
us
as
what
they
are
the
the
idea
of
again
having.
H
E
H
E
B
E
Been
a
while
is
there
us,
oh
yeah,
future
work,
yeah
right,
so
I
didn't
yet
touch.
The
task
list
still
need
to
refine
that
that's
the
future
work.
Also
the
icon.
The
task
spec
brings
a
grouping
mechanism
that
might
be
beneficial
for
the
jmap
tasks
back
as
well.
Yeah
still
wanted
to
to
keep
track.
How
you
know
I
didn't
want
to
redefine
it
and
look
how
it
so
and
kind
of
waiting
for
it
to
to
settle
here.
E
N
D
E
B
B
E
E
E
Yeah,
so
again
from
last
time.
Basically,
what
is
the
motivation?
Why
are
we
doing
this?
We
basically
when
our
business
or
our
use
case
is
moving
existing
user
data
between
systems
or
virtual
generic
API.
This
is,
or
this
is
actually
what
we're
looking
for.
So
yeah.
We
found
J
map,
which
is
really
nice
for
that.
E
E
So
what
should
the
spec
provide
that
we
still
don't
have
yet
guidance
on
how
to
develop
a
jmap
API
in
a
very
simple
way,
so
we
want
to
have
low
requirements
here.
We
want
to
have
propose
a
reduced
jmap
feature
set
like
the
minimum
that
implementers
need
to
do,
and
also
some
further
extensions
for
migration
rate
problems.
Next
slide.
Please.
E
Yeah
so,
first
of
all
the
the
we
call
it
the
central
profile
for
Rapid
development.
Basically,
the
idea
is:
how
do
you
quick
start
gen
web
API
development,
so
we
want
to
define
a
bare
minimum
for
like
the
one
time
use.
Actually,
so
it's
it's
a
bit
weird
word
or
a
bit
weird
wording.
So
our
use
case.
Sometimes
people
have
a
legacy
system
and
they
won't
just
want
to
migrate
the
data
away
once
and
then
they
want
to
shut
down
the
system.
E
So
for
this
they
need
something
really
simple,
really
easy
to
implement
and
we
want
to
give
some
guidance
on
Basics
to
start
with,
so
people
will
more
easily
be
able
to
dive
into
the
jmap
world
and
then
extend
upon
this
bare
minimum
yeah.
So
rough
outline
is
here
as
I
already
had
it
last
time.
Basically,
only
focus
on
key
objects
skip
a
session
object
in
case
the
username
equals
the
account
introduce
a
simplified
request
as
already
teased
last
time,
and
also
no
batching
yeah
for
now,
as
a
first
step.
E
E
E
E
So
the
problem
we
ran
into,
we
have
an
API
software,
is
running
also
the
jmap
API
software
is
running
that
that
you
write
yourself
is
running
on
a
third-party
infrastructure,
but
the
log
messages
from
this
third-party
infrastructure
cannot
be
sent
to
a
logging
service
that
you
can
use
yourself
so,
but
what
you
can
do
you
have
a
clearly
client
that
has
access
to
the
Gmail
API.
So
how
do
you
get
access
to
the
logs
or
to
more
information
to
see
what's
happening
on
the
server
side?
E
E
Somewhere,
so
it's
basically,
the
idea
is
basically
to
send
logs
to
send
logs
alongside
the
Gmail
response.
No
that's
so
yeah!
So
yes,
so
there
might
be
some
other
use
cases.
So
here
you
can
see
how
we
currently
implement
it.
It's
it's
like
an
initial
start,
so
we
have
actually
log
lines
here
in
the
response
and
you
see
files
like
a
file,
a
log
level
which
log
line
message
and
so
on.
E
As
you
can
imagine,
it's
like
need
to
take
care
that
this
is
a
secure
Channel
with
restricted
access,
because
it's
logs
it's
sensitive
information
and
yeah.
It's
it's
for
now.
It's
just
our
own
extension,
but
we
might
want
to
include
this
like
a
more
generic,
more
a
better
version
in
in
this
migration
draft
and
there
might
be
other
uses,
use
cases,
for
example
the
in
the
thief
case.
You
could
also
say
I
run
the
runtime
I'm
I
mean
the
runtime
errors
that
were
an
issue
with
a
thief
draft.
E
E
Yes,
so
backend
info
is
basically
to
what
Hunter
was
saying
before,
so
no
no,
so
we
have
the
issue
as
a
client
supporting
a
lot
of
servers
that
no
server
software
is
perfect,
even
though
they
say
they
implement
the
same
thing
they
they
have
their
own
quirks
and
we
want
to
be
able
to
handle
that
in
an
automated
fashion.
So
we
need
to
identify
which
server
is.
What
I
don't
know
is
a
version
two.
E
Is
it
version
2.1
things
like
that
and
so
basically
to
be
able
to
handle
also
like
small
bugs
in
server-side
software,
that
people
will
not
just
not
fix
themselves
yeah.
So
what
what
do
we
need
to
do
before?
We
can
do
that?
We
need
to
get
information.
What
software
is
actually
running
on
those
servers
without
that
info?
We
cannot
do
that.
Next
Step,
please.
E
So
here's
a
proposal,
basically
how
to
do
that
within
the
session
object
one
could
just
after
login
server
could
just
expose
which
back-end,
which
software
basically
is
running
on
it
or
which
so,
which
software
versions
it's
using.
So,
for
example,
we
have
PHP,
we
have
harder
versions
here
or
then
we
also
have
the
sieve
script
type.
This
is
just
a
very
rough
first
idea.
We
want
to
include
in
this
draft
and
I
think
the
main
question
I
have
is:
does
it
make
sense
to
compile
everything
in
this
initial
draft
I?
E
M
M
So
we
see
very
often
customers
coming
to
us
with
some
dump
of
emails
somewhere,
some
mail
dear
folder,
something
like
that
and
we
are
increasingly
considering
a
list
of
Sif
files
stored
there
and
we
are
increasingly
seeing
jmap
as
a
very
good
option
to
lift
those
on
an
HTTP
level
for
migration
purposes,
and
so
that
is
basically
the
background
of
all
these
proposals.
Here.
You
know
that
for
people
that
have
a
similar
use
case
in
a
similar
situation
that
might
find
that
also
very
useful,
obviously
like
Ken
already
discuss.
O
B
M
B
D
A
B
B
K
B
O
K
B
B
B
B
Tasks
is
a
thing:
that's
got
stuff
already
identities,
we're
going
to
start
discussion
on
mailing
lists.
I
guess
we
can.
We
can
add
a
milestone
for
that
when
we
decide
to
do
something
quota
sharing
their
all
things.
We
already
have
milestones
for
calendars.
We
have
math
stands
for
I.
Think
good,
then
I
think
let
me
move.
That's
everything
any
other
Milestone
things.
People
want
to
add
no
review
changes.
B
B
Thank
you,
everybody!
That's
the
end
of
the
jam
up
section.
Let's
start
the
extra
section:
do
we
want
to
switch
chairs
or
just
hang
out
where
you
are
to
both
look
happy
all
right
rolling
into
the
extra
now
I
think
we
only
have
a
couple
of
sets
of
slides
for
this.
So
let's
just
go
back
and
have
a
look
at
go
away.
Keep
using
at
the
chair,
slides
and
see
what
was
on
our
agenda
for
extra.
I
I
I
So
yeah
I
think
the
history
of
this
is
basically
bronze
started.
Last
call
in
July,
August
Barry
had
a
look
and
said.
Oh,
there
is
a
bunch
of
comments
that
you
haven't
addressed.
I
said:
okay,
partial
seems
to
be
done.
Let
me
split
it
off,
and
this
is
the
other
part
of
it
yeah,
which
still
have
open
issues
we'll
need
to
discuss
so
yeah.
Basically,
the
idea
here
is.
I
I
I
I
For
the
client
to
support
this
client
needs
to
opt-in
and
I.
Think
the
mechanism
for
the
client
to
obtain
is
to
use
enable
yeah
I
think
there
is
quite
a
bit
a
big
desire
to
be
able
to
do
this
without
client
opting
in.
But
you
know
if
we
can
find
another
way,
and
probably
next
next
slide
would
be
better
next
slide.
B
I
I
I
N
This
is
this
is
barely,
but
that
was
why
I
got
into
the
queue
a
client
that
wants
to
copy
10
000
messages
when
the
message
limit
is
a
thousand
is
just
going
to
do
it
with
10
copy
commands.
So
where
does
the
server
win
on
this?
You
wound
up
making
the
client
more
complicated
and
the
server
does
not
ultimately
have
to
do
any
less
work
in
the
process.
I
N
D
L
D
D
G
I
So
can
I
actually,
if
we
just
want
to
solve
the
copy
timing
out,
there
are
other
mechanisms
right.
We
can
send
untagged
response
in
the
middle
of
copy
and
actually
I
know
that
some
servers
do
it
every
within
a
minute
to
make
sure
that
the
client
doesn't
time
out.
Cyrus
server
does
that.
So,
if
we
just
want
to
address
this,
that
is
fixable
in
a
different
way.
Okay,.
D
B
N
And
and
what
I'm
thinking
that
we
might
do
with
copy
is,
have
it
sort
of
a
checkpointable?
It
might
be
part
of
this.
It
could
be
a
separate
document,
but
where
the
the
server
the
server
just
sends
a
checkpoint,
every
thousand
messages,
or
something
that
says
these
were
copied.
Okay,
these
were
copied
okay,
so
the
client
doesn't
have
to
retry
it.
But
if
there's
a
problem
and
it
crashes
in
the
middle,
the
client
knows
some
of
the
messages
were
successfully
copied.
N
I
D
N
J
Yeah,
this
is
kind
of
I
I
want
to
clarify.
We
are
still
going
to
allow
a
server
to
return.
No
message
limit
on
a
copy
if
it's
too
large
for
the
server
to
handle
correct
I,
know
we're
not
going
to
do
the
client
side
weirdness,
where
we
can
do
part
of
the
copy,
but
the
server
can
still
say
I'm
not
doing
this,
because
you've
asked
for
too
many
copies
and
return.
This
message:
okay,.
I
N
N
D
N
D
D
I
I
I
I
So
the
bottom
of
the
mailbox
or
to
the
top
you
can
you
have
to
do
this
by
doing
a
search,
getting
the
uid
back
from
the
server
as
a
number
then
constructing
another
search.
Obviously
you
cannot
pipeline
it
with
this.
You
can
actually
pipeline
the
two
searches
command.
You
say.
First,
you
get
the
you
know
whatever
you're
searching
for
you
find
your
message
using
the
save
option.
You
say
that
dollar
becomes
this
message
that
you
found
and
then
you
can
say
well.
I
want.
I
I
I
I
I
Oh
yes,
the
wonders
of
IMAP.
There
are
so
many
extensions.
There
are
always
so
many
interactions
between
them
so
as
it
happened,
actually
I
just
realized.
After
writing.
This
example
I
realized
message,
limit
and
search
res
or
IMAP
4f
to
have
an
interaction
because
What's
the
way
message
limit
is
written
at
the
moment.
I
B
M
J
H
I
B
J
J
So
I'm
I'm
ready
to
go
to
ask
a
call
on
that
unless
there's
different
opinions
in
the
room
as
far
as
sifts
news
goes,
I
don't
see
Pete
in
the
room,
but
he
he
had
an
issue
with
a
presentation
of
my
colleague.
Rick
did
about
implementing
snooze
via
IMAP
in
regards
to
having
a
special
purpose
mailbox
and
that
is
kind
of
what
Sif
snooze
is
doing,
and
I
asked
Pete
to
review
that
document.
B
B
And
I'm
just
going
to
do
it
from
here.
So
basically,
the
problem
here
is
that
there's
millions
of
existing
IMAP
clients
out
there,
which
connect
to
IMAP
servers
and
people
are
accessing
their
email
through
them.
We
would
love
them
to
upgrade
to
Jam
app
and
we
would
not
love
them
to
have
to
read
credential
re-authenticate
refiddle
with
their
settings,
ideally
also
for
quite
a
while.
B
I've
been
looking
at
what
you
would
do
here
if
your
IMAP
server
had
object,
ID
support
and
you
already
knew
the
IDS
for
every
single
email,
which
there
might
be
some
things
that
you
would
want
to
do
via
jmap,
to
manipulate
those
messages
while
still
keeping
the
synchronization
code,
and
that
means
you
need
some
way
of
of
saying
how
to
access
this
same
mailbox.
You
can
take
this
thing
off
via
jmap
as
well.
B
If
you
have
a
server
that
supports
both
IMAP
and
jmap,
and
you
already
have
an
existing
IMAP
connection
to
it,
you
might
want
to
change
that
to
a
jam
up
connection.
I
know
there
are
clients
that
already
do
that
kind
of
thing,
with
some
IMAP
servers
where
they
switch
from
having
a
password
to
having
an
oauth
connection,
and
they
ask
the
server
using
proprietary
methods
to
say.
Please
convert
this
password
into
an
oauth
token
and
then
I'll
save
that
and
go
ahead
with
that.
B
B
Both
the
URI
and
a
token
and
say,
connect
this
URI
using
this
token
and
do
jmap
with
that,
and
then
this
question
of
whether
you
need
to
pop
up
a
consent
screen
and
have
some
will
you
allow
this
client
to
connect
using
the
same
way
that
you
do
with
any
other
OS
connection,
because
it
may
actually
appear
as
a
connected
device
in
in
the
UI
of
your
server
if
it
depending
how
it
stores
things.
So
is
there
any
confusion
there?
That's
kind
of
an
unanswered
question
at
this
point
for
that
Barry.
N
B
Instead
in
number
three,
you
return
a
token
and
say
use
this
one,
which
may
be
the
same
token
or
it
may
be
something
generated
specifically,
if
you
have
a
password
connection,
number
three
is
certainly
the
the
most
flexible
of
the
token
based
one
and
number
one
is
the
easiest,
because
you're
creating
a
URL
with
a
token
in
it,
then
the
client
needs
no
smarts
at
all,
just
go
here
and
and
make
a
call
all
right.
So.
B
One
possibility
is
to
separate
things
out
and
to
say
in
the
IMAP
connection
or
via
some
other
endpoint
I
have
this
username
and
password.
Please
convert
it
into
an
oauth
token
for
me,
so
you
could
just
say
get
oauth
or
whatever
in
your
own
connection,
and
it
returns
saying
this
username
this
token
you
have
an
oauth
token
and
then
I've
done
some
examples
for
gmap
loss.
If
that's
readable
up
there,
you
could
say
no,
you
could
say.
B
Yes,
here's
that
example
number
one
his
his
a
path
with
the
token
at
number,
two
I
see
a
rather
than
just
giving
the
the
gem
up.
Uri
I
gave
both
which
yeah
okay,
that's
wrong.
That
should
just
be
slash,
Jam
app
in
that
case,
and
then
the
third
one
was
just
use
your
same
token,
please
so
that
those
examples
are
reversed.
This
is
what
I
get
for
writing
the
slides
five
minutes
beforehand.
B
J
B
That
it's
a
good
question,
certainly
if
it's,
if
you
need
to
generate
something
for
it,
then
it's
an
expensive
thing
to
create.
If
you
don't
know,
if
it's
wanted
or
not,
if
someone's
logging
in
with
a
username
password,
you
may
be
generating
an
oauth
session
for
them
at
the
point
they
request
one
of
these
things
you
don't
want
to
do
one
of
them
for
every
single
IMAP
connection.
M
Hi
assistant,
York
yeah-
this
is
kind
of
interesting
I
mean
I
and
it
rang
a
bell
with
me.
I
mean
I
have
actually
a
different
use
case,
for
that
I
mean
of
course,
I
see
your
use
case
here,
but
you
know
you
have
generally
the
use
case
when
you
have
a
platform,
migration
and
people
switching,
for
instance
to
another
migrated
server,
or
something
like
that.
M
B
So
the
the
motivation
behind
this
is
that
there's
a
bunch
of
clients
now
who
support
IMAP
and
would
like
to
enable
some
of
the
features
that
are
J
map
only,
but
don't
want
to
replace
their
entire
synchronization
stack
straight
away.
So
this
is
a
step
in
that
direction.
So
I
can
add
jmap
support
and
create
the
new
features
using
jmap
calls
without
replacing
their
entire
IMAP
stack
immediately.
Yeah.
M
I
D
I
I
If
we
are
going
there,
you
automatically
generate
all
token
when
requested
I'm
kind
of
a
bit
worried
about
this,
because
I
think
it
goes
against
their
old
philosophy,
because
if
there
is
a
user
interaction
that
is
needed
How,
what
you're
saying
you
always
will
get
a
token
or
it
will
fail.
And
then
the
user
might
never
see
the
nice
page
saying
that
you're
connecting
to
this
branded
service
and
all
of
the
rest,
so
I'm
kind
of
I'm
not
entirely
sure
whether
it's
in
spirit
with
oauth
to
be.
I
Which
case
you
are
actually
what
you're
doing
is
you're
implementing
just
such
a
loss-
and
you
know-
and
if
you
do,
then
you
will
get
a
token
anyway.
So
why
do
you
need
to
stick
it
there
in
in
the
first
place,
unless
you
maybe
need
to
want
to
put
some
or
specific
configuration
information
in
Amma
protocol,
because
it's
not
always
available
and
you
want
to
make
it
workable
with
lots
of
clients
and
yeah.
B
B
M
M
So
what
Alex
said
just
said
yeah,
maybe
there
is
a
slight
relation
to
that
draft
presented
in
the
I,
think
HTTP
API
working
group
by
Benny,
Derry,
Schwartz,
I,
think
about
or
off
by
command
line
applications,
yeah
yeah,
just
I,
don't
know,
maybe
keep
in
mind
yeah
and
to
your
further
questions.
You
know
like
what
could
this
imply
on
a
larger
scale?
So,
for
instance,
what
I
mean
here
I
mean
you
look
at
the
very
specific
thing
now
like
getting
from
a
password
based
to
an
auto
space
author.
M
So
the
question
would
be
what
if,
for
some
reason,
another
provider
would
change
from
X
to
Y
authentication
scheme.
So
would
it
be
worth
thinking
about
supporting
that
in
a
in
a
general
fashion
and
for
instance,
also
in
terms
of
like,
for
instance,
a
provider
deciding
to
shut
down
the
iPad
service
or
whatever
else
service,
and
this
is
also
something
that
might
be
communicated
so
that,
for
instance,
a
client,
then
can
say
the
user?
Okay,
you
need
to
know.
Do
that
or
something
like
that
and
question
related
to
that
also
I.
M
B
That's
an
open
question
still
whether
the
the
user
needs
to
explicitly
opt
into
this.
Ideally,
for
me,
a
client
that
adds
say
snooze
as
a
feature
rather
than
trying
to
to
back
Port
snooze
from
Jam
app
into
IMAP.
They
could
just
say
if
I'm
going
to
snooze
this
message,
then
I'll
make
a
snooze
call
with
the
object.
Id
email
ID
that
I
fetched
via
IMAP
and
say
please
snooze
this
message
to
the
server
and
the
server
will
then
make
the
changes
to
mailbox,
and
then
the
regular
synchronization
code
will
notice.
B
M
J
B
B
N
Hell
with
it
too
fussy
to
do
the
thing
this
is
Barry
I,
don't
agree
with
that,
because
I
I,
don't
think
things
need
to
be
fully
baked
to
be
adopted.
I
think
that
the
idea
is
that
there's
enough
here
to
get
us
thinking
and
then
we
can
use
that
to
develop
the
document
there.
Everything
is
still
up
for
grabs.
I
N
B
O
I
just
have
a
question:
I
have
a
draft
to
revivify
the
expires
header
for
mail
messages
and
someone
this
morning
asked.
Does
that
mean
we
can
have
an
expires
keyword
in
IMAP
so
that
people
can
tell
it
can
ask
for
the
messages
that
are
expired
or
not,
and
I
said
I,
don't
know,
but
we
have
a
meeting
and
I
can
find
out.
So
assuming
that
people
use
this
header,
which
I
think
is
fairly
likely
because
there
was
a
lot
of
interest
in
it,
MOG.