►
From YouTube: IETF103-JMAP-20181107-1350
Description
JMAP meeting session at IETF103
2018/11/07 1350
https://datatracker.ietf.org/meeting/103/proceedings/
A
B
B
C
Is
this
the
one
yeah
okay,
well,
I'm
hoping
this
is
actually
probably
quicker
than
we
had
on
the
agenda.
I'm
just
gonna
give
a
quick
status,
update,
I,
guess
so.
J
map
core
is
in
last
call
has
been
for
a
while,
and
we
plan
to
finish
that
on
16th.
So
about
a
week
after
this
ayah
TF
conference
finishes
have
been
small
revisions
since
since
ITF
102,
but
mainly
its
editorial,
fixes,
I've
gone
through
and
made
sure
all
the
JSON
now
passes
and
isn't
full
of
syntax
errors
and
they
fix
some
grammatical.
C
Knits
and
people
have
sent
small
amounts
of
feedback
which
have
been
incorporated,
such
as
making
account
ID
explicit,
rather
than
letting
it
be
optional
because
they
make
it
changed.
What
was
the
default
on
the
server?
You
could
get
very
confused
as
a
client
and
added
some
state
management,
so
copy
and
import
are
consistent
with
set.
We've
got
a
couple
of
implementations
out
there
in
Syrus
in
C
and
X,
which
is
a
pearl
implementation
and
Patrick
James
of
implantation
as
well.
A
slightly
older
version
of
the
specter.
C
Now,
there's
no
open
issues
on
github
and
so
basically
we're
just
waiting
for
any
more
final
reviews
from
people
coming
in
I
know
a
few
more
people
are
reading,
as
back
at
the
moment.
Well,
hopefully
see
that
on
the
list
over
the
next
few
days,
again,
I'm
not
expecting
any
major
issues.
Just
again
small
speck
nips
grammatical
changes,
things
that
need
to
be
clarified
and
then
hopefully,
yeah
sixteenth
will
close
and
submit
to
the
next
step.
C
C
D
C
Well,
what
turns
what
you
mean
by
validation?
I
mean
we
we've
certainly
got
experience
using
a
number
of
different
contexts
that
aren't
mail.
Quite
a
few
I
think
what
has
happened
is
it's
it's
in
terms
of
the
two
specs.
They
very
much
become
data
model
in
mail,
which
is
just
like
the
objects,
I
think
or
is
all
the
kind
of
methods
and
synchronization
primitive,
assert
and
stuff,
and
that
that
seems
to
be
very
clean
separation,
I,
don't
so.
D
C
E
You
can
have
some
requests
that
span
several
account
ID,
for
instance,
if
I'm
doing
some
text
search
in
the
mailbox
and
I
wish
to
aggregate
results
across
different
accounts,
for
example
in
case
of
main
share,
and
the
current
model
with
a
quantity
does
not
hold
this.
So
I
think
that
is
a
point
that
might
need
to
be
discussed
as
part
of
Jim
at
core
I
assume,
which
is
not
specific
to
email
object.
That's
why
I
raised
the
point
I.
C
C
So
j-mac
mail,
the
other
spec,
is
basically
same
status,
had
small
revisions,
but
I
think
it's
in
a
very
good
shape.
Last
call
finishes.
At
the
same
time
again,
there
have
been
at
least
three
implementations.
I
know
of
various
versions
of
the
spec
as
it
has
developed.
Cyrus
is
the
one
that's
most
up-to-date
with
the
current
and
it's
fully
up
to
date
with
the
current
spec.
C
The
other
two
are
based
on
older
versions,
but
I
think
we
have
got
significant
implementation
experience,
especially
for
a
spec
at
this
stage,
again
no
open
issues
on
github
so
awaiting
any
final
Nicks
and
reviews
on
the
mailing
list
or
elsewhere,
and
then
we'll
move
them
on
on
16th
November.
So
again,
are
there
any
specific
comments
or
queries
or
concerns
about
mail,
no
cool,
all
right.
So
if
we
go
into
the
next
slide,
this
is
the
point
that
my
raised.
C
But,
like
you
don't
have
the
same.
Consistency
guarantees
like
IDs
must
be
unique
within
an
account.
You
can
kind
of
lock
the
account
as
you're
making
a
change.
You
have
an
old
state
in
new
state.
No
only
your
changes
happen
within
that
and
to
make
this
scalable
across
not
lots
of
users
and
stuff.
This
is
a
really
nice
property.
C
At
the
same
time
as
a
user,
you
can,
you
might
be
given
access
to
multiple
accounts,
for
example,
if
I'm
sharing
my
mail
with
Ron,
he
might
be
able
to
see
mail
for
my
account
as
well.
So
the
issue
was
raised
is,
if
you
want
to
present
some
single
view
of
stuff
between
different
accounts
like
a
unified
inbox.
What's
the
best
way
to
achieve
this
now?
One
way
you
could
do
this
is
just
to
aggregate
in
the
client.
You
make
a
query
to
both
the
accounts.
C
Take
the
data
you
merge
virtually
to
show
it,
but
for
some
clients
they
may
want
to
put
this
on
the
server
instead
of
me,
it's
so
that
they
can
be
simpler.
So,
what's
the
best
way
to
achieve
this,
so
I
came
up
with
a
couple
of
ideas
which
I
put
on
the
on
the
main.
This
one
is
just
to
simply
set
up
your
server
so
that
the
users
the
do
share
stuff
of
all
within
a
single
account,
but
that
loses
you
various
properties
that
make
that
a
pain,
I,
don't
think.
That's
actually
a
good
solution.
C
As
the
start
you'd
have
named
flashes.
You
know
if
you
both
want
to
mingle
in
box,
but
they're
in
the
same
account.
Well,
that's
a
clash,
so
you
couldn't
have
that
so
doesn't
really
work
having
two
different
users
stuff
share
that
so
I
guess.
The
other
solution,
then,
is:
do
we
want
some
way
for
the
server
to
aggregate
this
data
from
the
different
accounts?
Is
this
best
as
an
extension?
Is
this
something
that's
very
generic
and
any
ideas
or
comments
on
this
I
guess
Kurt.
F
Anderson
I
think
the
problem
here
is
that
it
is
conflating
the
user
experience
with
the
implementation
details
that
I
might
have
accounts
on
multiple
servers,
not
to
mention
multiple
services,
each
of
which
have
multiple
servers
and
want
a
unified
view
in
some
sort
of
a
client.
There
is
absolutely
no
way
that
the
SPAC
can
say:
oh
yeah,
let's
let's
just
aggregate
the
world
and
give
you
one
view
and
so
I
don't
think
we
should
try
absolutely.
C
I
mean
the
that
was
my
initial
view
and
that's
why
we
I
always
presumed
a
client
would
do
this.
If
you
need
to
do
so,
it
could
do
the
same
across
accounts.
Why
it's
not
in
that,
but
there
is
still
a
use
case
potentially
on
a
single
server
where
you
know,
which
does
have
access
to
both
yeah
I
guess.
Is
this
worth
addressing
I
I
think
to
be
clear:
if
we
did,
it
would
definitely
be
an
extension.
This
doesn't
seem
something
that's
integral
to
the
spec
as
a
whole.
C
C
C
G
B
G
C
C
E
C
Sounds
to
me
a
lot
like
search
snippets
in
that
it's
a
extra
data
item.
That
only
applies
to
the
thing
within
the
context
of
the
search,
and
so
you
could
definitely
yeah
just
add
an
extension
or
something
to
fetch
that
relevant
score,
but
you're
talking
about
for
your
search,
get
search
relevant,
sir,
so
they
get
social
snippets
and
that
would
then
yes
give
you
the
information.
You
need
to
do
this
on
the
client.
That
might
be
a
lot
simpler
than
trying
to
do
this
application
thing.
So
cool
well,
I.
Think
then
terms
of.
B
We
had
a
thing
called
ex-con
multi-source,
which
is
an
IMAP
extension
and
it
returns
a
list
of
folder
names
that
the
messages
are
in
at
the
start,
and
then
it
returns
a
list
of
tuples
which
are
UID
and
a
number
into
that.
Folder
named
lists
for
each
item,
which
is,
is
basically
that
tuples
of
account
ID.
C
G
Just
a
general
comment
that,
given
this
house
cut,
a
complex
aspect
is
already
I
really
would
be
happy
to
have.
Some
of
you
say
here
is
running
here's
here's
some
running
code
that
we
can
work
on.
Particularly
for
this.
You
know
you
can
happen
your
server,
you
could,
you
could
mock
it
up
as
a
ship
yeah,
it's
it
seems,
it
seems
to
me
like,
like
write,
write,
write
some
code
and
bring
it
back
to
us,
and
if
we
like
it
we'll
say
grateful
a
bit
yeah.
C
I
mean
III,
think
I
seen
three
general
consensus
here
that
this
is
definitely
not
part
of
the
core
spec
or
a
general
thing.
It's
a
extra
edition
that
could
be
added
on,
and
so
it
would
be
an
extension
and
yeah.
At
this
point,
the
best
thing
is
yeah
perhaps
been
why,
if
you
have
want
to
write
up
how
you
think
it
should
work,
whether
that's
just
fetching
the
relevance
with
the
search
or
whether
it's
something
that
aggregates
results,
you
know
there's
completely
different
options,
but
if
then,
then
we
can
also
you
look
at
that.
C
That's
a
group
from
the
list
and
see
if
that's
something
we
think
is
useful
enough
generally
to
standardize
as
an
extension,
but
I
don't
think
this
is
seems
to
be
part
of
the
current
specs
of
core
mail
and
so
I.
Don't
think
this
blocking
what
we're
doing
at
the
moment,
moving
forward
with
those
okay.
C
So
Jay
macking
sanctions
so
with
mail
and
call
hopefully
essentially
done
they've,
already
been
few
proposals
for
extensions.
That
people
would
like
to
see
that
we
might
want
to
work
on
within
the
group
and
I
think,
hopefully
still
within
the
chart.
Although
we'll
talk
about
the
charters
later
so
just
got,
I
didn't
got
to
want
to
briefly
bring
up,
so
the
first
one
is
about
generating
mDM's.
C
E
Use
K
of
how
to
use
MTN
reference
to
the
MDM
sent
keyword,
maybe
to
reference
to
header
for
having
an
ID
from
an
implementer
point
of
view
of
how
to
be
using
J
map
to
be
sending
requesting
Indian
and
managing
state
of
India.
So
just
adding
some
little
sentences
just
put
mdn
context
in
that
spec.
Okay,.
C
C
C
I
think
yeah
again,
we'll
look,
hopefully
to
I'll,
send
through
a
bit
more
the
review
about
just
the
the
J
map,
style,
knits
and
if
we
can
have
those
fixed
and
maybe
add
a
ad
in
the
discussion
that
you
you
want
to
add
in
that
you're
talking
about
and
then
we
can
review
that
and
go
from
there.
But
this
looks
like
yes
and
that'll
be
good.
C
C
So
there's
a
couple
of
open
questions
on
this
the
so
the
proposal
at
the
moment
is
to
be
able
to
use
Webb's
office
as
an
alternative
force
and
the
API
requests
back
and
forth.
That's
all
you
structured
data
which
I
think
would
be
potentially
useful.
It
gives
you
certainly
some
performance
advantages
at
the
moment
in
terms
of
not
having
to
be
authenticate.
C
Every
request,
which
could
be
expensive
depending
on
your
authentication
mechanism
and
also
WebSockets,
are
having
pretty
good
widespread
support
for
compressing
uploads,
as
well
as
the
Downloads,
whereas
HTTP
generally
not
much
support
for
gzipping
uploads
at
the
moment
in
compliance.
So
that's
so
I
think
it
would
be
a
nice
thing
to
sport.
C
I
think
the
extensions
worthwhile,
the
open
questions
the
moment
are
there
two
main
ones,
firstly,
about
out-of-order
responses,
so
it
with
HTTP,
you
might
make
concurrent
connections
to
the
api
server
api
was
resource
to
send
method
calls
and
currently
there's
or
say
they
could
interleave
an
execution
they're
not
guaranteed
to
anything
about
how
the
concurrency
works,
in
terms
of
which
one
executes
first.
But
that's
fine
if
you
know
that
the
two
requests
are
independent
and
that
give
you
books
with
WebSockets.
Do
you
presumably
don't
want
to
create
multiple
WebSocket
connections
to
the
server?
C
That
seems
a
waste,
but
does
that
mean
if
you
send
one
request
down
and
then
send
another
request
before
the
first
response
comes
back?
Is
the
server
allowed,
or
indeed
meant
to
execute
those
in
parallel
and
just
return
the
responses
in
the
order
that
they
happen
or
each
of
those
sequential
and
only
execute
the
next
requests?
Once
it's
finished
that
previous
one
you
sent?
That's
does
that
make
sense.
C
Iii
personally,
I
think
it
makes
you
you
want
to
be
able
to
replace,
can
come
an
ATP
requests
with
a
single
WebSocket,
so
it
should
execute
that
water,
just
as
soon
as
it
receives
them
and
then
send
the
results
back.
The
I
guess
the
issue,
potentially
you
have
here,
is
rate
limiting.
So
at
the
moment
you
have,
you
can
define
a
limit
on
the
number
of
concurrent
HTTP
requests
and
the
number
of
method
calls
you
can
make
within
a
single
request.
How
does
that
interact
with
this?
C
C
A
C
C
C
C
Although
it
is
oh
yeah,
so
this
is
just
for
API
requests
sending
and
receiving,
but
we
also
have
the
event
source
connection
you
can
make
in
order
to
get
push
notifications,
the
state
has
changed
and
therefore
you
should
make
a
new
API
request
to
update
your
state
to
the
latest
naazy
web
ii.
Sockets
are
bi-directional,
so
should
you
be
able
to
send
the
pushes
straight
down
this
as
well,
so
you
replace
your
event
source
connection
as
well
as
your
API
regress.
B
C
C
I
I
think
that
sounds
reasonable
as
well.
So
in
that
case,
we'll
need
some
kind
of
again
we'll
need
some
kind
of
tag
on
the
response
object
to
say
what
is
this.
Is
this
a
response
to
an
API
request,
or
is
this
a
notification
yeah
unlisted
thing?
So
it's
only
some
kind
of
common
property
on
the
top-level
JSON
object,
probably
app
type
might
be
a
it's
been
used
in
other
specs.
C
To
say
what
is
this
type
of
message
you
just
receive
it
yeah,
okay,
I
think
I'm,
happy
that
yeah
I
think
that
both
of
those
questions
that
we
have
good
solutions
and
yeah
ask
Ken
to
write
that
up,
but
then
I
think
it's
it's
not
actually
a
complicated,
it's
not
along
spec.
Mostly
it's
just.
You
use
the
stack
so
at
this
Ken
merchants
in
this
draft,
so
well
yeah
it
just
uses
the
standard
upgrade
a
connection
to
WebSocket
and
then
everything
else
just
happens
normally
just
enjoyment
would.
C
A
C
C
C
C
D
D
D
B
D
B
C
D
D
C
D
D
C
C
D
C
C
B
I
This
is
dkg,
so
the
sign
verified
in
addition
to
the
question
about
trust
management
on
the
server-side.
There
could
easily
be
two
certificates
that
match
all
the
things
identified
in
this
paragraph
right,
so
sanger.
If
I
just
says
we
believe
something
some
certificate
matched,
but
there's
no,
there's
nothing
that
you
can
use
to
on
the
client-side
to
log,
for
example,
key
continuity,
discovery
or
something
like
that.
H
C
I
think
dkg
is
going
to
just
point
to
the
general.
Probably
things
is
you
have
a
board.
You
can
go
from
a
simple
to
very
complex,
depending
on
what
the
client
wants
to
do
and
kind
of
picking
what
this
extensions
for.
As
long
as
this
has
a
good
use
case,
that's
fine.
Do
that
and
then
don't
over,
complicate
it
and
then
set
the
stuff
signed.
D
C
D
C
D
C
A
B
A
I
I
C
I
mean
it's
not
just
for
UI,
but
I
did
have
a
similar
question
of
what
is
the
client
meant
to
do
if
you
get
signed
slash
failed
by
what
is
the
expectation?
There
is
an
instance.
Tell
the
user,
don't
read
as
message
at
all
or
not
show
it
to
the
user
or
just
not
show
anything
as
it
said,
and
just
display
as
normal
and.
D
A
C
D
A
D
A
C
A
C
C
E
Exception
remained
the
exact
name,
but
we
do
it
on
a
roll
when
a
quota
is
exited
and
currently
a
client
have
no
idea
of
what
that
quota
is.
Would
we
implemented
that
in
ago
is
an
additional
field
as
part
of
the
mailbox,
a
twitter
inverse
information?
C
A
E
C
Even
mailbox
so,
for
example,
if
you're
using
a
labels,
type
approach
well,
one
message
come
in
multiple
mailboxes.
You
only
may
only
count
it
once
for
your
quota,
even
though
it's
in
multiple
mailboxes.
So
if
you
just
add
all
the
quota
of
mailboxes
together,
you
end
up
with
two
higher
quota
usage
and
also
similarly
I
think
to
what
Chris
was
saying.
So
you
might
have
a
single
quote
rude
that
also
encompasses
other
non
mail
data
types.
C
D
C
B
A
B
We
actually
undertaking
the
work,
and
the
other
thing
is
back
to
the
idea
that
a
JMO
client
is
often
a
display
shimmer
over
the
data
model.
Most
of
what
the
user
cares
about
is
am
I
using
5%
of
my
quota
or
95
percent
of
my
quota.
They
don't
care
about
the
exact
details
of
how
that's
calculated
I
just
want
a
bar
that
shows
approximate
usage
and
being
exact
to
a
bike
doesn't
matter.
The
server
will
still
reject
it
if
you're
right
at
the
edge.
C
C
That's
that's
a
good
point,
I
think
and
actually
that
meets
interesting
potential.
Other
way
of
doing
this,
which
is
just
you,
have
one
or
more
quota
object
returned
by
the
server
up
to
the
server,
and
it
just
says
how
big
it
is,
how
much
you
can't
using
and
a
label
for
it,
because
that
label
is,
you
know
you
tell
the
user.
Is
this
full
male
quota?
Is
this
your
whole
account
quota
whatever
and
then
client
sure
I,
don't
know?
That's
it.
It
certainly
might
actually
be
more
useful
to
the
user
than.
C
B
So
yeah,
once
these
ones
have
been
completed,
take
into
account
that
there's
value
in
having
a
single
connection
and
single
endpoint,
the
working
group,
the
person
that
we
now
work
on
other
extensions
for
later
data
types,
including
calendars
contacts
files
and
that
other
extensions,
maybe
credit
to
access
the
other
features
of
IMAP
and
sieve
quota
being
entirely
within
that
scope,
as
well
saying
that
we
should
still
be
bound
by
the
constraints
that
we're
developing
a
protocol.
First
synchronizing
clients
server
that
we
use
existing
IETF
work,
who
are
possible.
B
So
jeaious
calendar
already
exists
us
as
they
calendaring
format
over
JSON
and
that
we
work
with
whatever
IETF
groups
are
defining
data
types
in
the
areas
to
interesting,
transferring
and
the
core
principles
that
the
server
doesn't
do,
work
that
hasn't
been
requested
of
it,
and
things
are
explicit
rather
than
implicit.
Where
possible
service.
The
server
is
going
to
take
further
actions
based
on
the
data
being
uploaded.
B
C
B
Undertake
scheduling,
actions
and
based
on
who
the
authenticated
user
is
and
where
they're,
putting
the
data
automatically
and
there's.
In
some
cases
you
can't
even
turn
it
off
and
that
the
Jai
map
approach
to
things
is
more
than
if
you
want
something
done
by
the
server
you
ask
for
it,
and
if
you
don't
ask
for
the
default
is
not
to
do
anything
other
than
just
store.
C
C
C
C
If
you
have
one
protocol
and
again
that
also
means
you
avoid
the
issues
where
my
contacts
were
calendars,
don't
because
my
conflicts
slightly
out
in
that
respect,
all
the
authentication
needed
is
different
or
anything
like
that.
So
we
think
it
makes
a
lot
of
sense
for
this
to
be
a
single
protocol.
Also,
then
it
competes
properly
with
active
sync
and
other
proprietary
things
like
that
and
which
part
of
their
advantages.
Yes,
they
have
a
single
protocol
for
accessing
all
these
different
types.
B
F
Yeah
I'm
I'm
also
wondering,
though,
even
given
the
guidelines
of
where
does
this
sort
of
leverage
other
specifications
and
what
are
their
kind
of
core
principles
that
governs
a
gmap
interaction
with
a
server
I'm
wondering
if
we
need
to
bound
this
base
a
little
bit
more
and
do
we
want
to
talk
about
J
map,
RT
or
something
I
mean
or
J
map
track.
I
mean
it's,
maybe
not
I,
don't
know.
D
With
my
ad
hat
on
now,
this
is
actually
similar
to
what
I
was
about
to
say
is.
This
seems
like
a
lot
of
work
potentially,
and
one
thing
which
is
not
clear
is
priority
of
different
items.
For
example,
we're
probably
still
want
to
do
gmf
extensions
before
other
documents,
unless
we
really
collectively
can
I
think
that
there
is
an
out
bandwidth
in
the
working
group
to
work
on
all
of
them
in
parallel.
So
probably
some
text
should
be
added
about
sequencing.
D
A
D
A
D
A
C
Don't
you
Robin
yet
have
a
focus.
My
a
lot
of
person
in
terms
of
ordering
I
would
say.
J-Mac
calendars
makes
a
fair
bit
of
sense.
Do
because
I
don't
think
it's
going
to
be
too
complex
with
Jays
calendar
already
in
last
call
that
collects
as
well.
It
is
pretty
much
combination
of
that
data
format,
plus
J
map
call
for
the
desynchronization
and
a
small
mountaintops
to
talk
about
scheduling.
That's
it's
fairly,
straightforward,
yeah.
F
C
You
know
again,
I
think
the
extensions
proposed
have
been
like
so
WebSocket
extension
just
extends
call.
It
would
make
no
difference
which
in
use
on
top
of
it
I,
don't
think
the
mini
any
conflict,
whereas
yeah,
whereas
it's
not
any
extension,
is
an
alternative.
Where
is
like
email
send.
Mdn
is
clearly
only
relevant.
You
know,
I
I'm,
sorry
points
to
me
so
far
the
extensions
of
a
clear
whether
they're
doing
how
they
would
interact
well.
F
B
C
It's
which
is
a
slightly
odd
when
some
aspects
and
yes,
it's
kind
of
I,
see
email
but
the
same
time.
The
chord
doesn't
say
anything
about
email
like
it
doesn't
have
to
be
email
and
that's
why
we're
using
for
other
types
as
well,
and
then
that
was
the
point
raised
by
sodium
on
the
lift
list.
There
was
email
focused
in
this.
C
Well,
that
will
be
recharging
again
later.
Calendar
in
context
clearly
makes
the
most
sense
and
would
be
a
really
good
thing
to
have
together
with
email
on
on
this
protocol,
so
yeah.
If
we
focus
on
that
and
then
minor
extensions
to
the
email,
it
all
then
I
think
that
encompasses
all
the
things
we
actually
want
to
do
and
keeps
it
more
constrained.
B
Having
just
stood
up
in
the
chairs
meeting
and
and
said
that
we
should
keep
her
milestones
up
to
date,
us
working
group
chairs,
I've
gone
through
and
teak
done
for
submitting
the
documents
describing
the
protocol.
Since
that's
done,
no
one
demanded
a
problem.
Statements
turns
off
the
document
to
for
guidance
to
implementation
of
service.