►
From YouTube: IETF114 NFSV4 20220726 1900
Description
No description was provided for this meeting.
If this is YOUR meeting, an easy way to fix this is to add a description to your video, wherever mtngs.io found it (probably YouTube).
A
A
Okay,
I
got
the
meeting
I
started
the
meeting
in
this
thing
with
the
thing.
B
B
C
C
I'm
brian
palaski-
this
is
the
nfs
version
for
working
group
meeting.
Is
the
sound
coming
out
okay
to
remote
participants?
Please
let
me
know
I'm
worried
that
we
have
too
many
microphones.
C
Okay,
so
this
is
brian
blasey
again
thomas
haynes
is
with
me,
he'll,
be
taking
a
minute.
I
have
no
clue
how
to
get
there
from
the
reading
materials
thing.
Click
on
the
notepad
for
fs
version
4..
C
We're
having
technical
difficulties,
no
we're
fine
now,
so
I
just
want
to
say
that
tom
haynes
is
with
me
here.
Taking
minutes
I'll,
be
driving
the
slides.
All
the
presenters
are
on
site
right,
yes,
and.
D
C
Take
minutes
while
you're
talking
and
dave
novak
is
not
here
today
and
I
he's
not
in
remotely
either.
Thank
you
all
the
remote
participants
for
joining
us
and
again,
if
you
have
a
problem
with
audio
and
you
can't
hear
something
or
we're
not
being
clear,
just
send
a
message
over
the
thing
and
you're
doing.
C
Richard
and
I
can't
pronounce
your
last
name,
richard
n
is
going
to
be
looking
and
monitoring
for
people
that
want
to
ask
questions
remotely
okay,
so
it's
tuesday,
everybody
in
the
room
is
messed
up,
except
for
the
guy
in
the
back
row.
C
Man,
I'm
a
beanie,
okay
and
the
I
have
a
beautiful
butterfly
mask.
C
In
the
room
here
and
95,
mass
or
km95
master
required
blue
sheets
are
signed
when
you
log
in
to
meet
echo
for
all
your
people
already
logged
in,
and
the
local
people
have
also
logged
into
I'm
gonna
touch
on
a
note
well
and
then
we'll
do
agenda
bash
at
the
end
of
the
agenda.
C
Is
a
discussion
and
finish
one
big
thing:
we're
not
gonna
cover
the
this
document,
which
is
under
dave,
novak
he's
not
available
today,
I'm
going
to
take
that
to
the
mail
list,
and
I
need
to
after
this
meeting
and
talking
as
we
go
through
the
discussion
today,
I'll
send
out
to
the
mail
list
the
an
update
to
the
overall
document
status
and
where
we
stand
from
the
comments
coming
out
today
and
any
follow-ups
after
this
meeting
I'll.
Take
it
to
the
list.
C
There's
the
note.
Well,
I
can
sorry
I
had
another
one.
C
The
session
is
being
recorded
everyone,
so
everything
I've
said
has
been
recorded
for
posterity.
You
signed
it,
I'm
just
looking
at
the
general
rules
here
from
the
slides.
I
can't
get
to
see
you
to
see
yeah.
C
Okay,
that's
it
that's
enough
back
to
the
regular
schedule
program,
so
everything
you
say
is
kind
of
recorded
property
of
iotf
is
a
very
simple
summary
of
the
note
well,
but
there
are
a
lot
of
documents
below
by
the
way,
which
is
very
useful
in
the
links.
If
you
haven't
looked
through
the
documents
for
the
meeting
and
ietf
sessions
and
ietf
work,
they
actually
do
change
once
in
a
while.
So
you
might
want
to
refresh
your
memory
if
you
haven't,
but
you
saw
this
when
you
registered.
C
All
right,
this
is
the
agenda
a
little
bit
over
on
mice.
Oh.
C
C
No
well
agenda
bashing.
Is
there
any
additions
from
anyone?
We
have
kristoff
covering
layout
issues,
chucks
on
the
news,
security
and
rdma
tom.
F
C
F
So
under
agenda
I
thought,
like
I
mean
the
last
last
one
you
have
like
was
the
road
map
of
the
working
group
would
would
need
more
than
20
minutes
to.
I
shall
come
to
an
prospective
conclusion,
so
I
was
thinking
like
if
we
can
run
some
of
the
discussions
really
short
and
have
more
time
allocation
for
the
last
discussion.
C
Thanks
and
then
I
I'm
sorry
I
really
do
have
to,
and
then
I
just
I
clear
you
after
you.
C
E
C
All
right
up
is
kristoff.
C
G
So
background
story
is
five
years
ago
we
got
rfc
8154
published,
which
is
basically
a
revision
of
the
block
layout.
That
ties
deeply
into
scuzzy
concept,
most
importantly
for
his
persistent
reservations
and
device
identification,
it
turns
out
just
around
the
same
time.
The
whole
storage
industry
is
all
on
the
nvme
hype.
G
That
does
the
same
thing
just
different,
and
we
somehow
need
to
support
that
as
well,
and
because
of
that,
I
pretty
soon
after
started
a
document
that
explains
how
to
use
that
layout
with
nvme
in
a
way
that
doesn't
change
the
underwire
protocol,
but
requires
a
lot
of
normative
language.
How
to
map
nvme
to
scuzzy
concepts
that
fundamentally,
are
the
same,
but
expressed
very
differently
and
the
way
the
initial
version
did.
That
was
to
by
referencing
a
nvm
express
nvme
document.
G
That's
called
discussion,
translation
reference
that
does
this
mapping,
but
it
does
it
rather
sloppily
and
isn't
really
up
to
the
standards
that
the
ietf
has
for
normative
documents
and
because
of
that,
that
document
lingered
for
a
while.
The
next
slide-
and
I
know
dave
and
sarin-
did
another
take
on
it
and
but
not
a
whole
lot
of
things
happened
for
years,
so
this
year
after
chuck
started
prodding.
We
put
a
lot
of
work
into
it
like
chuck
and
dave,
and
me,
and
especially
dave,
did
a
lot
of
work
to
me.
G
So
we
know
yeah,
but
you
also
did
a
whole
lot
of
work
and
chuck
also
did
some
nice
work
and
dave.
So
we
we
finally
got
a
team
instead
of
just
people,
doing
it
and
put
out
a
new
version
of
it,
which
gets
rid
of
gets
rid
of
that
reference
to
the
nvme
stlr
document
and
instead
puts
in
its
own
normative
language,
and
I
think,
we're
more
or
less
ready
to
have
working
group
adoption
for
a
status
track
document.
We
have
like
two
open
issues,
and
that
goes
to
the
next
slide.
G
So
one
is
the
on
the
wire
protocol
has
the
concept
of
and
designator
for
the
device
ids.
That
is
basically
that
binary
blob
that
identifies
the
device.
What
type
is
a
discussion
has
a
bunch
of
different
types,
so
the
two
two
of
the
three
types
relevant
in
nvme,
actually
three
of
the
three
types
defined
in
nvme
map,
one
two,
one
to
the
concepts
in
skz.
So
what
the
current
document
does?
Is
it
reuses
them
and
doesn't
do
any
on
the
wire
changes?
G
So
dave
brought
up
an
idea
that
we
would
use
different
identifiers
to
show
that
we're
talking
about
an
nvme
device
which
reduces
the
amount
that
the
scope,
how
the
client
has
to
search
for
the
device
and
might
be
a
little
easier
to
understand
for
administrators
or
devops?
The
downside
is,
instead
of
just
being
a
mapping
document.
This
actually
adds
new
xdr
on
the
wire.
G
I
mean
basically
just
two
new
code
points,
but
it's
still
a
change
of
the
on
the
wire
protocol
and
the
other
question
that's
kind
of
related
to
it
is:
do
we
want
to
add
a
new
uuid,
designator
type,
so
nvme
and
scuzzy
both
added
a
way
to
define
device
to
do
device
identification
with
an
I
with
an
ietf
was
at
4122
uuid,
and
we
don't
have
that
in
the
previous
rfc
yet
because
scuzzy
hadn't
edited
at
the
times,
because
he
was
adding
it
right
around
the
time
and
nvme
also
edited
it
around
the
time.
G
So
if
we
want
to
change
the
wire
protocol
anyway,
we
should
support
that
as
well,
and
preferably
also
for
scuzzy
and
not
just
friendly
me
and
that's
basically
the
two
interesting
items
we
have
left
open
and
otherwise
I
think
we're
ready
for
nitpicking
review
and
bashing.
The
document
into
submission.
C
Are
there
sorry,
if
you
can
hear
me,
are
there
any
questions?
I
I
guess
I
just
want
to
say
I
disagree
with
exactly
one
thing.
You
said
he
said
you
started
off
by
saying.
There's
this
fad
or
something
like
that
of
nvme.
So
the
thing
is
is
that
it's
probably
become
a
lot
more
important
moving
forward
for
this
type
of
application
than
a
lot
of
other
things
because
of
mb
media
fabric
stuff.
G
G
G
C
I'm
sorry,
I
got
chuck
lever
at
the
microphone.
H
Hi
chuck
lever
a
couple
of
comments.
I
would
like
to
propose
moving
this
to
a
working
group
document.
I
think
somebody
has
to
do
that.
So
let
me
let
me
do
that.
The
second
comment
is
right.
Now
it's
an
informational,
the
status
of
document
is
informational.
I
would
like
to
see
it
become
a
standards
track
document,
because
it
it
does
have
some
normative
statements
in
it
that
I
think
really
don't
belong
in
an
informative
document.
So
those
are
my
two
comments.
C
Okay,
thank
you.
The
chair
recognizes
even
with
his
mask
on
david
black.
A
I'm
in
deep
trouble,
so
I
completely
agree
with
everything
chuck
just
said.
In
addition,
having
been
the
source
of
some
of
the
open
issues,
I'm
going
to
agree
with
where
the
draft
currently
is,
let's
get
this
document
done
done
quickly
with
no
change
to
the
wire
protocol
and
then
and
then,
if
somebody
sees
value
in
the
uuid
bring
in
a
separate
draft
to
add
the
uuid.
I
Richard
schaffenegger,
just
I
haven't,
heard
anything
that
would
block
this
from
not
being
adopted
and
quite
the
opposite.
I
would
also
think
that
having
this
document
on
the
standards
track
is
would
be
very
valuable.
C
F
A
C
All
right
next
up
is
thank
you,
chuck
weber,
discuss
security
rpc
with
tls.
Let
me
pull
you
slide
that.
H
It's
it's
always
been
a
problem
of
these
three
normative
documents
listed
on
this
slide,
not
being
ready.
Yet
these
normatively
cited
in
this
particular
document,
the
two
of
them
at
the
bottom
of
the
slide,
9146
and
9147,
are
now
published,
so
those
are
no
longer
blocking
this
document.
H
The
kitten
document
is
now
on
off
48,
at
least
when
I
checked
a
week
and
a
half
ago,
which
means
it's
well
on
its
way
should
be
published
any
very,
very
soon
in
within
weeks
or
less
so.
This
document
is
now
in
ref,
meaning
we're
we're
this
close
to
to
to
being
off
48
and
mile
48.
It
just
means
for
those
who
don't
know.
That's
just
the
authors,
making
final
comments
before
publication
next
slide.
H
There
are
still
these
two
matters
that
were
discussed
probably
18
months
ago,
and
I
did
I
did
mock
these
up
in
the
github
repo
for
this
document.
So
if
anybody
wants
to
go
and
look
at
them,
they're
available
in
in
the
github
version
of
this,
we
haven't
made
any
changes
to
the
actual
cued
document
that
will
happen
during
off
48..
H
I
should
say
I
did
check
to
see
how
long
this
document
has
been
waiting
in
the
rfc
editor's
queue.
It's
been
88
weeks.
This
is
not
the
longest
awaiting
document
in
the
queue
right
now.
There's
one
that's
been
waiting
since
2015.,
so
I
count
I
count
us
as
being
lucky.
H
I
just
wanted
to
mention
the
implementations
of
this
standard
that
we
have
so
far.
I
think
three
of
these
are
already
listed
in
the
actual
draft.
The
freebsd
implementation
is
both
client
server.
H
I'm
told,
has
both
client
and
server,
although
it's
not
for
nfs
right
now,
I'm
working
on
the
linux
client
prototype,
which
is
for
for
nfs
and
ben
connington,
has
produced
an
nginx
module
that
can
act
as
a
front
end
for
a
box
standard,
nfs
server
of
any
flavor
and
do
the
rpc
with
tls
translation
in
front
of
it,
and
he
recently
told
me
that
he's
he's
put
up
on
github,
I
believe
so
that
it's
open
source
and
people
can
look
at
it
and
he
intends
to
ask
for
the
nginx
folks
who
to
merge
it
and
he's
actually
online.
H
We've
had
decent
community
testing
exposure
at
the
most
recent
one.
We
had
all
the
implementations
present
from
the
last
slide
that
I
mentioned
and
we
were
interoperating.
So
that's
going
well
next
slide.
H
H
H
Okay,
we
discussed
among
the
various
implementers
what
this
administrative
interface
will
look
like,
and
this
is
sort
of
what
we've
come
up
with
so
we'll
just
add
a
mount
option
that
allows
you
to
specify
a
tls
or
or
not,
and
the
prototype
is
available
on
in
my
repo
github
and
on
kernel.org
next
slide.
F
H
Last
we
talked
about
this.
There
were
some
other
standard
actions
that
we
were
considering,
for
example,
how
exchange
id
and
friends
could
use
tls
peer
authentication
in
the
same
way
that
they
use
kerberos.
Today,
kerberos
principles
there
hasn't
been
any
progress
there
and
the
other
thing
we
had
sort
of
extended
conversations
about
was
how
servers
can
tell
clients
that
if
the
client
wants
to
use
offices
it
must
use
tls
or
some
of
some
form
to
protect
the
transport.
H
We've
had
a
couple
of
proposals
in
that
arena,
but
they're
not
well
liked.
I
had
one
called
rpc
tail
suit
of
flavors.
I've
gotten
two
thumbs
down
on
that
one.
So
I'm
going
to
let
that
document
expire
and
we
can
continue
to
explore
basic,
basically
not
wire
changes,
but
rather
just
conventions
for
servers
to
communicate
their
security
policy
requirements
via
existing
protocol
elements
rather
than
adding
new
ones,
and
we
can
take
that
conversation
to
the
list
next
slide.
C
Yeah,
I
was
wondering
I
do
get
previews
over
here,
all
right
and
and
so
what's
the
ask
here:
what's
the.
H
This
is
a
progress
report
implement
implementation
status.
I
guess
I'm
asking
whether
people
agree
that
letting
rpc
tlsu
flavors
expire
gracefully
is
what
we
would
like
to
do
and
then
continue
the
conversation
on
soliciting
other
proposals
for
that.
For
that
requirement
about
servers,
nfs
servers
indicating
that
they
want
stronger
transport
layer,
security
for
offices.
C
No,
no,
I
got
I
got
so
first,
let's
suit
the
flavors
expire,
any
any
objection
to
that
online.
Can
you
check
it.
H
Should
we
hear
from
ben
cadec,
first.
C
K
C
So
I
don't
think
it
was
an
objection
there.
I
think
nobody
objected.
The
next
thing
was:
let's
take
this
to
the
list
and
discuss
the
do
you
want
to?
How
do
we
handle
this
and
just
description.
H
Yeah,
mr
novak
and
rick
macklin
both
had
some
ideas
about
how
they
wanted
to
go.
Rick's
proposal
was
he
had
a
couple
of
he
used
some
pre-existing
protocol
elements
for
both
nfs
v3
and
nfs
v4
servers
to
indicate
requirements
for
transport
layer
security.
C
Okay,
okay
cool
anything
else
online
in
room.
E
H
So
I've
been
working
on
this
version
two
of
rpc.
Over
already
may
I
did
a
little
prototyping
about
a
year
ago,
it's
very
limited
in
scope
and
hasn't
really
changed
since
last
time.
I
presented
on
this
topic
next
slide.
H
The
concern
a
year
ago
was
that
if
we
wanted
to
support
message
chaining
properly,
we
really
need
to
have
a
a
full,
a
full-on
flow
control
protocol
built
into
our
pc
of
already
made.
We
don't
have
that
in
version
one.
So
over
the
last
year
I've
worked
with
john
a
gar
to
fix
that.
To
rectify
that
we
have
it.
The
description
is
written
in
section
4.2.1.
H
H
H
H
So
his
proposal
was
to
add
a
mechanism
in
the
protocol
to
exchange
certificate,
material
or
authentication
with
their
material.
I
would
rather
see
us
use
build
on
existing
security
protocols
that
are
more
sophisticated,
like
tlsv,
1.3.
H
So
that
we
have
underneath
rpc
over
rdma,
the
transport
of
the
rdma,
verbs,
etc,
etc
are
protected
in
the
same
way
that
tls
protects
communications
on
tcp
next
slide.
H
I
suggested
this
tom
talpy
recently
and
he
directed
me
to
section
5.4.2
of
rfc
5042,
which
I've
read
several
times
and
I
don't
find
it
convincing.
H
That's
not
something
we
need
to
go
into
here,
but
I
just
thought.
Maybe
it
would
be
interesting
to
consider
specification
of
ddp
and
rd
map
on
quick,
but
maybe
that's
more
than
we
ever
want
to
take
on
so
at
this
point
yeah.
So
I
don't
know
what
to
it's
an
open
question,
how
we
will
deal
with
transport
layer,
security
for
rpc
already
rdma.
If
we
want
to
continue
work
on
this
protocol,
which
I
will
get
to
in
a
moment,
but
first
they
have
black.
A
The
layering
on
quick
might
not
be
anywhere
near
as
bad
as
you
think.
It
would
be
because
the
time
that
we
did
our,
I
did
I
warp
which
is
in
on
mpa
and
ddp.
We
also
did
a
fairly
quick
mapping
of
that
onto
sctp.
A
H
H
A
C
Yeah
tom.
J
Good
chuck
yeah,
I
mean
the
there's
a
couple
of
concerns
about
putting
rpc
over
our
dma
in
charge
of
security.
In
that
rdma
will
be
basically
transferring
you
know
encrypted
bits,
so
the
software
layer
of
rpc
rdma
has
to
decrypt
it.
So
you
loot
some
performance.
That's
not
a
very
compelling
architecture.
J
I
agree.
Okay,
second,
you
know
there's
rocky,
which
has
no
security
whatsoever
I
wore
which
referred
to
ipsec.
I
work
over
quick.
I
agree
with
david.
It's
a
tractable
problem.
I
think
it's
a
fairly
significant
problem.
We
deferred
that
work
because
at
the
time
we
wanted
to
do
it
a
few
years
ago.
Quick
wasn't
fully
specified
now
that
it
is
maybe
not
the
datagram
service
part
of
it.
J
But
you
know
there's
there's
a
possibility
now,
but
the
discussion
at
the
time
was
to
have
a
quick
extension
specifically
for
rdma
and
that's
a
bit
more
of
a
problem
than
just
simply
mapping
rdma
to
a
quick
stream.
J
H
J
H
Yeah,
I'm
happy
to
entertain
more
conversation
about
this
on
the
mailing
list
or
privately.
C
Go
ahead!
Okay,
so.
I
Richard
just
just
in
high
level
observation,
I
would
strongly
refrain
from
doing
any
private
security
whatsoever
in
a
protocol
at
this
level.
So
I'm
sorry
that
he's
not
here,
but
I
would
really
not
support
that
how
the
layering
at
the
end
of
the
day
then
is
done.
Yes,
there
will
be
secure
performance
implications
if
you
do
it
with
rdma
over
tls,
you
can
do
the
layering
the
other
way
around,
so
that
you
can
do
some
hardware
offloading
there.
I
D
H
Color
me
a
pollyanna,
but
you
know
I
like
to
think
that
in
some
day
in
the
near
future
that
we
can
have
both
security
and
performance,
especially
with
offload
solutions.
So
I'm
optimistic.
I
Security
will
always
be
a
bump
in
the
road,
so
yeah,
but
anyway
my
main
point
was
not
to
do
an
any
any
custom
solution
to
this
problem
here
in
this
group,
but
use
a
standard
industry
standard
security
protocol
and
again
how
the
layering
is
then
done
ultimately,
is
that
that's
the
sound?
That's
something
that
the
group
needs
to
decide.
C
Okay,
I
have
christoph
hellwig
on
mike
and
then
tom
kelby.
G
Yeah,
so
I
I
don't
think
security
and
performance
really
are
that
far
apart
right,
both
both
with
existing
tls
offloads
with
ssds.
We
absolutely
know
there's
hardware
encryption
engines
that
do
line
rate
without
any
problems,
but
for
that
we
need
a
good
enough
standard
that
they
can
implement.
Which
brings
me
to
the
next
question
I'll.
G
I'll
just
emphasize
and
someone's
kind
of
specified
this,
and
I
don't
think
this
is
the
nfs
people
and
given
that
storm
has
disappeared,
and
I
know
there's
a
lot
of
discussions
about
this
working
group
disappearing.
I
think
we
need
to
have
a
good
conversation
about
maybe
grouping
that
kind
of
work
in
a
new
working
group
that
includes
a
broader
group
of
people
with
applicable
skills.
J
Okay,
so
christoph,
I
totally
agree.
We
want
the
broadest
skills
when,
when
nfs
v4
was
going
to
take
over
rdma,
the
the
discussion
was
that
all
the
rdma
experts
were
already
hanging
around
nfsv4,
so
to
the
extent
that
we
branch
out
of
nfs
and
rdma
uses,
we
need
the
additional
expertise
that's
present
in
itf
like
if
we
go
for
quick.
Obviously
we
engage
the
quick
group
but
richard.
If,
if
somebody
deploys
rdma,
they
absolutely
want
performance.
That's
the
reason
to
deploy.
Rdma
and
security
is
mandatory
to
implement
these
days.
E
H
Originally,
we
looked
at
read
plus
and
said:
that's
going
to
be
difficult
to
fit
into
the
rdma
paradigm,
because
the
client
doesn't
know
in
advance
what
the
reply
is
going
to
look
like.
It
could
look
like
a
single
content
data
segment
or
it
could
be
some
set
of
content,
data
and
content
whole
and
there's
no
predicting,
and
so
it
was
impossible
for
the
client
to
be
able
to
set
up
a
reasonable
set
of
rdma
registered
memory
to
receive
replies
in
general.
H
So
our
upper
layer
bindings
have
therefore
not
permitted
any
resulting
items
in
replus
to
be
eligible
for
a
direct
data
placement.
H
However,
it's
been
pointed
out
to
me
that
certain
file
systems
will
start
doing
a
whole.
Instantiation
when
reads
happen
over
unwritten
extents,
so
rate
plus
is
a
is
a
really
good
idea
for
large,
sparse
files
where
you
want
to
keep
the
sparseness.
H
So
I
think
we
and
of
course
we
do
want
to
serve
large,
sparse
files
over
rdma.
Those
are
important,
of
course,
in
hvc,
as
well
as
in
the
cloud
for
virtual
disks.
So
I'd
like
to
just
sort
of
throw
this
out
here
and
say:
we
need
to
think
a
little
bit
more
carefully
about
how
replus
needs
to
work
over
rdma
either.
We
need
a
convention
that
explains
how
to
make
it
work,
at
least
for
the
simple
cases,
or
we
actually
need
to
extend
either
rdma
or
nfsv
4.2
replus.
H
We
had
some
performance
goals
for
v2
when
we
started
this
project
a
few
years
ago
and
I
think
that
most
of
them
have
been
met.
They'll
either
be
met
by
the
piano
bus
layout
type
that
kristoff
described
earlier,
or
they
will
be
met
in
practice
with
the
private
data
extension
that
we
did
in
rfc
8797.
H
So
I
think
the
urgency
for
our
version
2
is
less
than
it
used
to
be,
and
my
concern
is,
as
it
was
last
time
represented,
that
we
don't
have
more
than
one
version
two
prototype
right
now
I
don't
see
more
than
that
happening,
so
there
really
isn't
a
an
interoperability
issue
there.
There
are
loads
of
other
projects
in
front
of
the
working
group
right
now
that
I
think
are
higher
priority,
especially
tls
related
issues
and
there's
really
only
one
person
who's
available
to
prototype
author
review
and
perhaps
steward
this
document.
H
So
here's
where
I
do
put
this
in
front
of
the
group-
maybe
we
should
just
let
this
work
expire
for
now
and
leave
it
to
come
back
at
some
later
point
when
we
might
have
more
resources
to
bring
to
bear.
J
I
think
letting
it
expire
is
the
wrong
term.
It
just
kind
of
let
it
sit
right,
it's
well,
the
you
know.
Physically,
the
draft
will
expire,
but
it's
still
out
there
and
it
can
still
be
discussed
and
we
can
reactivate
it
when
it
makes
sense
so
yeah.
What
you
described
makes
good
sense.
I
just
the
terminology
of
expired,
doesn't
sound
right
to
me.
H
J
C
C
G
So
I
I
don't
think
great
plus
really
is
a
protocol
problem.
I
know
we
have
some
implementation
problems
on
the
linux
site,
but
read
plus,
can
always
do
a
short
read,
so
you
can
always
just
return
one
chunk
of
data
in
one
hole
and
let
the
client
retry,
which
I
guess
for
most
cases,
is
simply
the
right
thing
to
do.
G
H
H
There
are
some
more
subtle
issues
about
direct
data
placement
that
make
me
a
little
concerned,
but
if
we
can
find
a
way
of
making
it
work,
I
think
the
the
benefits
of
not
triggering
whole
instantiation
are
are
real
and
strong
and
important.
So.
C
Any
objection
to
letting
this
draft
expire
in
the
ietf
sense
and
and
if
needed,
we
will
pick
up
the
work
in
the
future.
But
at
this
point,
chuck
is
saying
that
the
gains
are
not
there
to
warrant
the
effort
nor
the
resources
to
take
forward.
The
v2
rpc
already
made
route
any
objection
to
dropping
work
and
freeing
up
bandwidth.
So
we
can
focus
on
more
important
things.
H
That's
for
I
think
we
should
deliver,
remove
the
milestone
for
delivering
these
documents,
because
that's
not
going
to
happen.
B
C
B
B
So
I
thought
I'd
give
some
some
dates.
We
first
published
it
in
april
2018
we've
had
a
hammer.
Space
has
had
a
private
server
application
since
that
time,
we've
also
had
an
open
source.
Client
implementation.
Since
that
time
and
the
linux
flight,
maintainer
kind
of
is
a.
I
want
to
say,
hard-ass
well
I'll,
say
hard-ass
about
this,
and
he
he
doesn't
like
to
take
changes
that
haven't
been
ietf
standard
track
documents
because
it
clears
up
ipr
concerns
and
open
standards
equates
to
open
source
right.
B
So
I'm
trying
to
get
this
the
him
to
push
the
changes
upstream
as
they
call
it,
and
he
won't
do
this
until
I
get
this
more
solid.
B
E
B
C
Forward
to
a
working
group
last
call
and
attaching
a
document
shepard
to
it
any
any.
F
J
F
B
B
J
I
know
it's
not
it's
not
your
responsibility
to
convince
others,
but
I
just
curious
how
broad
the
interest
in
this
might
be.
If
it's
been
idle
for
a
year
or
two
one
might
ask,
is
it
gonna
reawaken
outside
of
hammer
space?
I
don't
know.
C
Thank
you.
I'm.
E
B
So
my
next
presentation
is
about
layout
competing
file
recovery.
If
we
go
to
the
next
slide,
please.
B
So
the
problem
statement
is
when
a
server
reboots,
we
have
a
grace
period
for
the
that
allows
the
client
to
reclaim
open
files.
The
server
doesn't
necessarily
know
that
which
files
are
opened
by
the
client,
so
the
client
just
is
informed
that
the
the
server
is
rebooted
and
it
knows
it
has
a
set
amount
of
time
to
to
reclaim
those
files
if
it's
interested
in
reclaiming
them,
and
it
can
use
the
claim
previous
with
the
open
to
which
is
typically
only
valid
during
that
reclaim
period
to
say
I
had
this
file
open.
B
B
B
It's
it's
private
protocol
in
which
they
can
communicate
state
changes
and
errors
that
have
occurred
and
everything
a
loosely
one
is
one
in
which
we
use
a
storage
protocol
like
nfs
in
order
to
communicate
those
changes
and
that's
very
limiting
right.
So
the
loosely
a
loosely
coupled
system
is
one
pre,
pretty
much,
which
you
have
several
different
proprietors
proprietary
code
bases
for
the
servers.
B
B
B
So
can't
we
just
get
a
new
layout
and
then
report
the
error.
Well
again,
this
we
have
a
complication
if
we
have
client-side
mirroring
actually
even
without
client-side
mirroring
the
server
may
have
moved
the
file
the
data
instances
of
the
file
to
different
data
servers.
So
it's
no
longer
possible
to
report
errors
that
had
occurred.
B
So,
what's
the
importance
of
reporting
the
errors?
Well,
if
we
report
the
errors
it
allows
with
client-side
mirroring
it
allows
the
the
mds
to
pick
which
one
of
the
the
instances
are
valid
and
should
be
re-silvered.
If
it
doesn't
have
that,
then
it
has
to
scan
all
the
files
assume
all
the
ones
that
weren't
reported
as
errors
may
be
an
error
and
go
across
all
the
files
and
the
the
cost
of
it
depends
on
the
size
of
the
files
and
the
number
of
open
files.
B
I
feel
like
I'm
going
fast,
but
I
always
do
next
slide.
Please
so
I'll
present
two
potential
solutions
and
of
course
I've
picked
one
already,
but
I'm
just
trying
to
show
that
we
did
consider
this.
We
could
introduce
new
operations
which
would
allow
the
server
and
the
client
to
agree
that
the
server
has
this
capability,
and
then
the
client
knows
that,
since
the
server
is
advertising
this
ability,
it
can
then
use
that
ability
or
we
can
use
the
special
state
id
on
layout
return
and
define
some
new
semantics.
During
the
grace
period.
B
I
picked
to
find
the
new
semantics
with
the
special
state
id.
B
So
the
special
state
id
is
defined
not
to
be
valid
with
layout
return
and
if
it
is
presented
with
layout
return,
the
server
returns
nfs4
error
that
state
id,
but
if
it's
inside
the
grace
period
we
could
have
it.
We
could
understand
that
this,
the
client
is
saying.
I
encountered
an
error
when
I
was
trying
to
when
I
was
writing
to
this
layout.
I
want
to
report
it
and,
if
you're
capable
of
using
that
information
use
it
if
not
go
ahead,
so
that
if
the
server
was
inside
the
grace
period
was
returned.
B
Bad
state
id.
The
the
client
would
know
that
the
feature
was
not
supported.
There's
no
need
for
layout
error,
because
layout
error
implies
that
you're
going
to
reuse
the
layout
state
id
afterwards
and
we're
not
going
to
and
then
there's
no
need
to
reclaim
the
file,
because
the
client
has
already
said.
Here's
what's
happened
with
the
file.
B
B
The
big
question
is
where
to
put
this:
I
could
tack
it
on
to
dell
state
id,
but
I
don't
like
that
because
it
del
state
id
is
pretty
close
to
being
done
and
I
don't
think
it
fits
in
with
it
all
right.
It
is
about
open
files,
but
that's
kind
of
stretching
it.
For
me,
I
could
do
a
new
document,
but
it
starts
a
new
cycle.
It's
only
two
to
three
pages
and
the
next
topic
and
the
working
in
is
about
what
are
we
going
to
do
with
the
working
group?
J
B
It's
about
what,
if
we
were
to
go
back
a
couple
of
slides
about
the
the
presence
of
the
special
state
id
during
the
grace
period
versus
not
in
the
grace
period
and
how
the
client
can
determine
whether
or
not
the
server
supports
this
feature.
J
E
J
B
C
Take
a
step
back
and
put
make
a
proposal
unless
to
to
the
working
to
the
working
rebellious
in
the
next
two
weeks:
okay,
but
chuck
leaver,
please
chuck
lever.
H
I
agree
with
that,
following
up
on
what
mr
talby
said,
making
this
a
convention
seems
appropriate
rather
than
something
that's
normative.
It
seems
pretty
straightforward
for
a
client
to
try
this
and
the
server
says
what
the
the
client
goes.
Okay,
this
is
isn't
working.
I
fall
back.
E
H
Whatever
the
old,
and
otherwise,
it
will
do
something
more
reasonable
right,
so,
in
other
words,
it
wouldn't
require
any
action
by
this
by
the
working
group
at
all.
B
B
So,
but
all
right
I'll
talk
later,
I'm
I'm
so
used
to
normative
documents.
I
don't
go
with
the
other
workflow.
C
J
B
J
Well,
so
you
know
it's
always
better
to
be
clear
and
write
a
document,
I'm
not
trying
to
say
get
around
it
or
anything
like
that
it
just
if
it
doesn't
rise
to
a
requirement
that
everybody
must
do
even
if
it's
optional
it
it's
like.
Maybe
this
isn't
really
a
protocol
question?
Okay,
just
it's
just
food
for
thought!
That's
all!
Thank
you!
Oh.
C
E
C
C
C
C
Yeah,
that's
the
end
of
the
prepared
presentations.
Now
you
have
a
little
bit
under
an
hour
left
in
the
meeting.
This
is
an
time
for
an
open
discussion
on
the
next
steps
and
work.
What's
open
items
in
the
working
group
and
how
do
we
deal
with
them
and
generally
frame
what
we're
going
to
do
here.
I
would
hope
that
tom's
going
to
help
me
out
a
lot
here
by
the
way.
C
C
Yeah,
so
that's
actually
all
right.
Let
me
try
so
so.
We've
been
negligent
in
meeting
our
deliverables
to
our.
C
To
our
stated
milestones
on
the
the
description
of
our
working
group
for
those
of
you
who
have
been
involved
with
this
for
a
while.
C
This
is
not
a
surprise
given
the
sometimes
our
change
in
direction
and
just
distractions,
but
I
think
at
this
point
we're
coming
up
to
a
question
of
along
I'm,
not
gonna,
I'm
just
gonna
use
as
an
example,
but
not
it's
not
the
most
important
thing,
but
this
discussion
we
had
about
the
sorry,
the
rpcr
over
rdma,
the
v,
the
v2
document,
whether
it
was
absolutely
required
that
we
complete
that
work
at
this
time
and
is.
C
Is
it
holding
us
up
from
actually
delivering
on
implementations
and
making
things
just
work
with
what
we
have
and
what
we
know.
So
we
decided
basically
to
put
that
on
the
shelf
until
something
would
make
us
go
back,
reopen
that
discussion
and
because
there's
something
critical
that
we
need
to
do
in
that
in
our
pc.
Over
already,
I
think
that's
kind
of
the
discussion
we
had.
C
So
I
think
what
I'm
looking
here
at
the
working
group
is
to
just
come
up
with
the
list
of,
must
complete
items
must
do
and
define
those
attach
dates
to
them.
Find
out,
identify
the
ones
where
we're
struggling
to
we're
struggling
to
apply
the
people
necessary
to
complete
the
document
work.
Pretty
much.
Everything
in
the
I
tip
is
document
work
for
the
most
part.
That's
all
I
care
about,
but
backing
the
document.
Work
is
sometimes
from
our
side.
Implementation
work
to
prove
out
the
ideas.
C
I
don't
know
that
the
implementation
work
is
what's
holding
us
back
at
this
time.
I
think
what
we
have
is
a
pile
of
documentation
work,
probably
the
biggest
overhang
we
have
is
likely
the
completion
of
the
bis
document.
That's
just
a
bunch
of
clarifications.
It's
just
a
lot
of
work,
so
I
need
some
help
here.
B
F
F
Please
please
so
yeah
you
had
here
transported
with
my
id
at
on
I
mean
I
mean
I
I
can
tell
tell
you
like
what
my
my
what
I
see
happening
here
and
then
then,
which
is
basically
what
you
said.
Brian
I
mean
like
like
that's.
This
working
move
is
like
really
really
I
mean
I
have
seen
like
a
couple
of
times
that
people
have
finding
reviewers
like
that,
for
the
beast.
F
One
which
seems
to
be
when
I
took
took
over
this
working
group
seems
to
be
very
important,
but
then
I
failed
to
see
like
anybody
and
you
have
any
interest
in
it,
and
there
are
a
couple
of
documents
there
they're
out
there.
I
have
been
asking
the
like,
whoever
I
I
know
cares
about
actually
like
how
much
it
is
critical
for
implementation
to
get
it
done.
I
didn't
get
a
really
right
answer,
so
that's
that's
like
this.
This.
This
is
an
I
understand
this
is.
This
is
like
historically
like
really
long.
F
It
has
been
there.
This
working
has
been
there
for
for
quite
a
long
time,
and
you
you
guys,
are
really
I
see
it
today,
like
you
know
each
other,
you
you
work
together.
You
have
done
tremendous
amount
of
work,
but
now
I,
but
I
see
this
one
actually
really
declining
and
failed
to
deliver.
That's
that's
the
one
part.
Another
part
is
like
when
it
comes
to
errata
and
all
this
thing
I
have
seen
like
we
have
like
huge.
E
F
Of
ira
down
fs
before
the
I
have
talked
to
three
past
eddies,
like
what
happened
to
those
like.
Why
didn't
anybody
have
to
clear
those
aorta,
at
least
for
document
hold
up
for
document
update
that
could
have
been
like
purified
you're
waiting
for
something
to
happen?
That
didn't
happen.
F
So
I
I
also
wanted
to
see
like
how
the
working
group
actually
think
and
how
much
that's
critical,
because
because
this
is
a
huge
amount
of
irata
existing,
the
other
thing
is
like
how
we're
running
the
meetings
and
all
these
things,
but
that
I
can
definitely
work
with
on
my
chairs
to
make
it
more
workable.
So,
let's,
for
example,
like
the
business
document
that
has
been
like
this,
like
documented
and
dev,
had
worked
on
it.
F
It's
not
even
in
the
working
group
documentation
list,
so
I
had
to
push
it
for,
like
okay
call
for
a
working
goal,
adaptation
it
got
adopted,
but
the
record
doesn't
see.
Tell
me
anything
like
that.
So
I
think
this
all
these
things
is
happening
like
whether
there
is
the
energy
in
this
working
group
to
actually
carry
on
the
rest
of
the
make
the
thing
and
or
what
should
we
do.
That's
basically,
what
I
want
to
hear
from
you
guys
and
I'll
I'll
shut
the
shut
up,
and
let
you
guys
to
talk.
C
Why
why
why
don't,
I
believe
you
anyway,
so
so?
Is
there
a
way,
david
you're,
the
zen
master,
on
this
stuff?
Is
there
a
way
to
show
a
a
screen
on
beat
echo?
C
A
C
Oh
okay,
so
yes
allow,
okay,
that
isn't
very
good!
I'm
gonna
screen
snap
that
got
it
that
looks
like
my
entire
life
right
now.
Did
that
do
it?
That
did
it
okay,
so
I
think
one
of
our
one
of
our
issues
is
just.
C
Frankly,
let's
just
be
clear
here:
our
a
minor
issue
we
have
is
updating
this
page
yeah,
that's
not
actually
the
problem,
but
it
doesn't
help
that
this
page
is
not
up
to
date,
because
that
just
makes
people
confused
so
there's
the
documents
we
discussed
today
up
on
top.
C
C
The
v2
rdma
open
files
and
yeah,
so
that's
both
the
ulba
and
the
rdma
documents
which
we
are
going
to
shelve
until
something
changes,
but
it's
these
bottom
documents.
I
think
that
we're
mostly
talking
about
there's
a
list
of
internet
drafts.
We
touched
on
quick
a
bit
today,
I'm
not
sure
if
that
is.
C
Let's
ignore
that
the
big
document
here
is
the
the
nfs
v4
version,
one
minor
protocol.
This
document,
I
believe,
is
the
one
that
is
being
reworked
and
addresses.
C
I
believe
is,
I
think,
is
where
those
are
appearing,
and
I
I
would
just
put
out
that
dave
novak
needs
help,
and
I
would
have
to
figure
out
how
to
make
that
work.
Mcdavis
dave
owns
that
document
and
he's
been
soldiering
on
a
lot
on
this.
So
what
what.
I
This
is
richard.
What's
curious
to
me
is
a
if
this
is
a
beast
document.
Why
is
it
still
as
a
individual
document?
It
should
basically
go
straight
straight
into
the
adopted
into
working
group,
and
the
working
group
can
then
decide
a
number
of
editors,
so
even
people
that
have
no
technical
inside
knowledge-
and
I
just
remember
I-
should
keep
my
mouth
shut.
C
A
A
Okay,
sorry
yeah
and
you
did
you
didn't
want
to
do
that.
Bp
yeah,
I
know,
but
I
don't
know
but
yeah
yeah
television
is
such
fun.
So
what
you're
seeing
here
is
a
fairly
small
community
that
in
the
past,
has
managed
to
generate
the
largest
docks
the
ietf
has
ever
seen.
I
got
ribbed
for
iscs
he's
being
300.
C
A
Like
I
said
two
country
miles-
and
you
might
be
looking
at
one
of
the
things
that
may
be
going
on
here
is
there's
definitely
enthusiasm
for
smaller
docs
I
mean
give
kristoff
and
I
and
everybody
else
interested
in
the
green
light
and
and
that
scuzzy
layout
update
draft
will
be
what
will
be
in
your
hands
before
the
end
of
the
year.
A
I
I
don't
know
what
to
say
about
this.
Aside
from
the
fact
that
yeah
the
specs
need
maintenance
yeah,
there
ought
to
be
some
interesting
stuff
done
there,
but
the
extent
I
can
I
can
see
enthusiasm
or
speaking
myself,
enthusiasm
it's
for
this
small
stuff,
like
the
scuzzy
layout
update.
That
makes
a
big
difference.
F
Yeah,
so
just
just
to
respond
your.
C
F
F
Sorry,
yeah,
okay,
okay,
okay,
you're
transported
so
yeah
just
to
respond
to
that
one,
because
even
if
you
were
like
laughing
about
it,
but
you
made
a
good
point
right.
I
mean
this
is
this:
is
a
small
working
group
trying
to
take
it
like
600
press,
the
documentation,
update
and
all
this
thing
and
my
question
to
this
like
if
this
is
so
important,
why
there
is
no
my
much
progress
in
it?
F
So
if
this
is
like
really
important
thing
to
do
the
beast
thing,
do
we
have
the
industry
ready
to
I
mean
when
this
is
published,
they
will
be
like
updating
their
servers
and
clowns
to
to
the
new
version
that
will
be
updating.
I
don't
see
that
enthusiasm
I
mean
into
the
work.
I
know
this
is
important.
I
know
this
is
a
maintenance
is
important,
but
well
I
don't
see
any
visible
indication,
that's
the
case
and
suddenly,
you
said
like
we
have
now
a
smaller
document
that
becomes
more
important.
F
Maybe
those
will,
if
we
get
through
those
those
will
get
have
more
probability
getting
deployed
in
the
in
the
internet
than
the
the
updated
base.
So
I
would
rather
be
happy
that
if
this
working
group
focuses
on
the
smaller
document,
get
them
done
and
say
like
well,
we've
done
our
so
far.
We
can
and
then
rest
in
peace.
B
F
B
D
There
are
a
lot
of
work
lingering
because
we
were
accident
and
we
didn't
finish
what
we
started
and
I
think
that
needs
to
be
addressed.
I
know
that
we're
exigent.
I
know
that
we
aren't
the
perfect
thing
yet
there
are
several
documents
that
were
not
accepted,
including
mine,
for
example,
for
data
reduction
as
an
example,
but
the
fact
is
that
there
is
work
that
can
be
done
and
we
kind
of
didn't
attach
it
attack
it.
So
what
I
suggest
is
to
try
to
get
to
the
point.
D
Then
we
are
finishing
the
4-2,
so
I
know
that
it
will
take
time.
I
don't
know,
there's
a
four-three
in
the
future,
I
don't
think
so.
Maybe
there
is
maybe
there
isn't.
There
are
new
technologies
in
the
network
in
the
memories
and
all
this
that
could
justify.
But
at
this
point
in
time
we
have
to
finish
what
we
started
and
I
would
be
very
sad
to
see
this
group
closed,
because
I
think
we
have
something
important
that
I
can
tell
you
that
my
current
company
were
using
some
of
the
the
nvme
stuff.
D
C
Okay,
so,
okay,
thank
you
siren.
This
is
brian
plowsky
co-chair
of
the
nfs
version
for
working
group.
So
I
I
agree
with
you
when
you
said.
I
think
that
this
very
much
summarizes.
C
I
think,
where
we
are
finish,
what
we
started,
that
that
absolutely
should
be
our
friggin
tagline
and
we
should
have
t-shirts
made
because
that's,
I
think,
what
we're
discussing
I
no
one
is
proposing
shutting
down
the
working
group
and
dropping
stuff
on
the
floor
and
watching
it
bounce.
That's
not
it.
The
discussion
is
what
are
the
things
we
need
to.
E
C
D
C
R.I.P
yeah
before
I
am
r.I.p
and
very
very
quiet
and
six
feet
underground.
So
I
think
that's
that's
just
where
we
are
and-
and
I
think
part
of
that
goes
to
I'm
extraordinarily
reluctant
to
for
proposals
of
let's
do
new
work
of
any
substantial
size
like
a
let's
redo,
the
entire
nfs
version
4
protocol.
That
would
I
I
could
imagine
somebody
proposing.
E
C
And
I
would
be
saying
that
no
that's
just
not
where
we
are
okay,
so
the
small
documents
are.
I
would
call
not
the
diminished
amount
of
work
or
thinking,
but
the
nvme
document
highly
leverages
the
previous
work
done
for
pnfs
and
the
layouts,
where
a
lot
of
thinking
went
into
it
and
we're
not
revisiting
that
thinking.
We're
just
reflecting
where
the
industry
is
in
the
this
is
the
third
decade
of
20
sec
21st
century
I
think,
is
that
over
here
2022..
C
So
so
I
think
that
that
level
of
work
that
I
think
the
best
document
just
needs.
I
think
the
best
document
needs
just
to
finish
what
we
started.
Completion,
I'm
afraid
to
drop
david
help
me
sorry.
Can
I
interrupt
you
kristoff
for
a
second
david.
Can
someone
give
me
the
strict
definition
of
a
missed
document.
C
A
Strict
definition,
yeah
folklore,
it
evolved
as
a
as
a
convenient
way
to
tell
someone
who's
reading
the
document.
What
it
replaces.
E
C
A
Right
and
and
and
let's
see,
okay-
well,
I'm
trying
to
stare
at
the
screen
where
the
font's
too
small
and,
as
I
said
this
head
the.
What
does
this
mean?
I
said
it,
it
really
is
folklore
it.
It
became
a
convenient
way
to
tell
some
to
tell
the
reader
that
it's
a
example.
This
draft
is
intended
to
replace
rfc
5661
and
that
that's
it.
I
I
Having
had
some
experience
with
being
an
editor
for
this
document-
and
I
can
fully
agree
with
with
david
here-
a
biz
document
is
really
just
an
convenient
way
to
say
something
that
this
document
will
become
the
new
standard
instead
of
an
old
standard.
It
has
no
meaning
appearing
whatsoever
of
the
technical
contents
that
are
being
changed
or
if
it
is
just
an
update
because
of
errata.
I
That's
that
has
no
bearing
to
one
other
than
this
it
a
this
document
that
is
really
just
an
update
for
errata
for
editorial
things
can
become
a
very,
very
lengthy
work.
Speaking
of
experience
to
very
minor
gain,
so
I've
been
updating
the
1323
document
to
73
23..
Nobody
actually
refers
to
20
73
23
by
the
document
name
itself,
because
everybody
knows
the
old
document.
I
However,
if
you
would
nowadays
implement
the
old
document
specs,
you
would
basically
not
be
compatible
with
tcp
stacks
all
over
the
world
or
you
would
have
actual
bugs
in
your
implementation.
So
a
biz
document
that
addresses
technical
and
and
editorial
errata
is
an
important
aspect,
but
it's
a
kind
of
an
unloved
child
kind
of
thing.
A
lot
of
work
for
very
little
gain
so
from
a
point
by
making
from
a
priority
perspective.
I
C
Okay,
thank
you
wait.
Is
there
anybody,
I'm
sorry,
I
gotta
bother
everybody
in
a
second
nobody.
Okay,
sorry
kristoff.
G
G
Then
the
group
spent
a
lot
of
effort
on
doing
the
best
document
for
nfs4
that
no
one
really
cared
about
and
now
we're
trying
to
do
the
same
for
4.1
and
if
it
was
just
catching
up
with
the
errata,
it
would
be
very
simple
document,
but
I
think
the
problem
is
it's
trying
to
bring
over
the
internationalization
rewrite
from
the
four
biz
document,
which
in
theory
is
very
important
in
practice,
doesn't
matter
what
the
implementations
do
anyway,
so
yeah,
I'm
not
sure,
that's
a
good
priority
for
the
group.
F
So
yeah
I
mean
I
I'd
like
to
understand
the
priority.
Here
I
mean
the
definition
of
priority.
I
mean
prediction
of
priorities
like
what
documents
we
should
take
on
first
and
complete
or
priority
based
on
like
what
is
needed
to
get
things
secure
and
get
things
interoperable
and
get
things
going.
So
I
I'd
like
to
understand
that
part
as
well,
because
I
mean
the
what
I'm
hearing
like
this.
F
This
document
is
important
and
it
is
addressing
lots
of
error
time,
but
nobody
cares
that
doesn't
tell
me
that,
like
this,
this
should
get
like
a
lot
of
attention
right
now.
It
could
just
be
there
and
we
can
have
like
very
low
updates
and
all
this
thing,
but
at
the
same
time,
if
this
is
this,
is
this
somebody
tells
me
like.
I
cannot
really
run
this
nfsv
nfsv4,
one
version,
because
I
don't.
F
I
have
all
this
error
I
have
like
hundred
and
some
some
hundred
plus
and
if
I
implemented
that,
then
it's
like
it's
not
going
to
work
and
that's
this
is
really
important
that
so
I
please
please
mention
like
what
is
your
priority
and
try
to
try
to
understand
like?
What's
the
implication
we
have
if
nobody
cares
about
this
document?
F
C
Like
sorry,
there's
one
person
in
front
of
you
and
then
I'll
get
to
you
hold
on,
but
I
want
to
make
a
comment.
This
is
why
I
was
asking
about
the
definition
of
this.
I
thought
that
the
best
document
started
from
addressing
errata
and
unclear
portions
of
the
specification.
L
A
Protocol,
because,
because
you
bumped
the
minor
number,
I've
also
been
involved
in
verb.
Well,
I
wouldn't
say
involved
at
the
same
credit
in
a
working
group
that
I
run.
Unfortunately,
other
people
did
most
of
the
work.
We
just
turned
out
a
best
document
for
sctp,
and
that
was
a
rather
large
doc.
I'm
sure
it's
not
as
a
head
start
saw
it
go
by
and
it
was.
A
It
was
not
not
all
errata,
but
it
was
all
minor
stuff
that
needed
to
be
fixed
that
the
implementers
agreed
this
is
this
is
this
is
how
this
is
how
it
ought
to
work.
So
I
I
guess
what
I
would
recommend
to
us
ahead
is
to
structure
his
thinking
as
an
ad
around
two
types
of
docs.
We
got
this.
A
We
got
these
relatively
small
stuff
in
which
there
is
strong
interest,
and
then
we've
got
this
abyss
and
the
maintenance
of
the
old
stuff
and
the
errata,
and
which
is
what
seems
to
be
causing
him
loss
of
sleep
and
those
two
things
ought
to
be
thought
about
separately
and
figuring
out.
What
happens
here.
C
Okay
and-
and
I
I
need
thank
you-
david
black
dot,
soren-
are
you
still
there?
Are
you
with
us
yep
yep?
Please
go.
D
Okay,
so
look,
I
am
going
to
go
back
to
some
changes
that
the
industry
went
through.
That
will
require
attention
that
I'm
focusing
on
my
draft
for
data
reduction
because
that's
the
place
I
live
in
today,
and
this
is
the
clouds,
so
all
the
the
the
clouds,
the
one
that
are
relevant
amazon,
aws,
google,
gcp
and
microsoft
azure.
They
all
are
talking
about
data
reduction
because
that's
an
important
part
of
moving
to
the
cloud.
D
So
I
think
from
that
perspective,
this
is
a
new
work
that
should
be
done
in
this
group,
because
you
know
the
clouds
are
moving
faster
and
we
are
staying
behind
if
we
don't
keep
up
with
that
again,
I
advocate
for
my
draft
as
an
example,
but
that
should
be
one
of
the
directions
that
should
happen
today
because
it
is
what's
happening
in
the
industry.
It's
not
me
it's!
You
know.
I
work
in
a
company
that
does
data
migration
to
the
cloud
and
I
can
observe
all
these
facts.
So
that's
my
point.
F
Your
transparency
again
again
so
david,
I
I
think
you
made
a
good
point
right.
I
mean
we
have
a
couple
of
a
couple
of
issues
here
and
I
say,
like
I
started
with
three
like
I
mean
the
question
is
like
I
can
I
can.
F
I
can
ask
the
question
about
the
process
and
all
this
thing
and
previously
made
decisions
like
why
this
is
that-
and
this
is
that,
like
you,
have
like
100,
plus
unverified
errata,
that
you
started
to
fix
it
without
any
kind
of
like
like
resolution
on
this
like,
even
if
it's
not
not
like
verified
hold
it
for
documents
like
nobody
did
anything,
it
has
been
there
for
quite
a
long
and
then,
basically
now,
if
I
look
and
into
this
whole
thing
the
piece
I
kind
of
understand
like
these
are-
these
are
two
this
this.
F
This
is
supposed
to
take
care
of
irritates
which
are
not
really.
We
didn't
agree
on
how
to
solve
them.
Now
this
when
this
piece
is
happening,
we
I
expect
that
discussion
to
happen,
that's
not
even
happening.
It's
like
dave
is
trying
pushing
his
work,
he's
trying
to
get
reviews.
There
were
a
couple
of
the
ietf
meetings
and
incrementing.
There
was
no
review,
so
I
I
kind
of
felt
like
this
is
not
like
the
thing
that
the
working
group
should
focus
on,
and
then
it
comes
to
the
smaller
documents.
F
I
would
appreciate,
and
the
authors
of
this
smaller
document,
to
get
these
things
to
some
sort
of
a
state
where,
like
I
mean,
I
think
today,
we
we
said
like
we
will
we're
getting
getting
this
nvme
document
as
a
working
group
and
then
dave.
You
said
like
yeah
I'll.
D
F
Something
really
soon-
and
there
are
a
couple
of
documents
that
we
discussed
today
is
very
it's
really
like
tom
said
it's
almost
done
now.
It's
just
working
with
last
call.
So
do
those
things
go
for
those
things
and
see
like
where
it
ends
up,
because,
because
I'm
I
don't
I
see,
I
would
like
to
see
this
working
group
progress.
I'm
not
here
to
just
click
on
this
button
says
close
this
working
group-
I'm
not
here-
for
that.
I
want
to
show
this
working
group
this
progressing.
F
If
it
is
a
management
issue,
how
we
are
running
things,
we
can
talk
about
that
one.
If
it
is
like
I'm
encouraging
people
to
do
that,
we
can
we
can.
We
need
to
do
something
about
it.
That's
what
I'm
asking
and
today
I'm
exactly
I'm
feeling
it
better
like
we're
talking
about
it.
We
see
like
this
work
need
to
be
done.
We
want
to
continue
this
one
and
that's
exactly.
I
want
out
of
this
discussion
out
of
this
working
group.
C
Thank
you,
david
black
is
going
to
say
something.
A
Yeah
david
blackman
is
going
to
say
something
so
there's
a
lot
of
nfs
out
there
there's
a
lot
of
running
code.
A
lot
of
it
is
not
exactly
close
to
the
latest
docs.
This
working
group
has
produced,
to
put
it
politely,
and
unfortunately,
I
got
a
plus
one,
christoph
snide
remark
about
internationalization
that
that
that
that
has
that
has
a
particularly
that
has
a
particularly
sorry,
sorry,
his
sorry
history,
nfs
that
I
don't
care
to
go
over.
C
Further
without
adult
beverages,
it
had
a
sorry
history
before
the
itf
right
and
that.
A
A
A
E
A
So
I
think
the
structure
here
is
figure
out
how
to
take
some
of
this,
how
to
take
some
of
the
smaller
drafts
forward
and
then
look
at
the
sort
of
maintenance
of
the
big
docks
and
the
and
of
the
the
bigger
drafts
as
sort
of
sort
of
a
separate
matter.
And
if
you
tease
those
two
apart
at
least,
has
a
structure
to
start
grappling
with
with
what
what?
A
C
And
and
yes,
and
and
so
let's
just
for
everybody
on
the
online,
just
a
reminder-
dave
novak
who
is
doing
most
of
the
work
on
this
document
is
not
here
and
I
had
a
slide
early
on
which
I
said
I
didn't
want
to
talk
about
this
document
without
dave
not
being
here,
because
we're
going
to
be
talking
about
stuff
without
the
background
to
complete
the
discussion
so
and
dave
did
tell
me-
he's
not
going
to
be
able
to
give
us
an
assessment
of
that
until
a
month
out.
C
Okay,
so
I
believe
that
was
you
know.
We
had
a
brief
talk
before
we
talked
so
I
got
it.
I
got
it.
I
think
we
decided
today
to
shelve
two
documents
that
are
on
this
list
related
to
the
already
already
maywork
the
v2
stuff
and
then
move
forward
with
the
nvme
dock.
These
are,
and
people
have
put
their
hands
up
to
work
on
it,
and
tom
haynes
on
the
state.
Id
document
is
moving
forward.
So
I
think
we
made
some
decisions
on
the
small
documents.
Then
we
just
gotta.
C
Put
them
in
the
milestone
list,
I
believe,
is
what
one
thing
we
have
to
do
clean.
What's
been
the
milestone
that
doesn't
need
to
be
there
anymore
and
then
get
dave
novak
to
come
in
and
give
us
a
do
not
do
on
the
this
document
and
assess
the
what's
critical
and
what's
not,
is
that
and
we're
separating
out
the
decisions
made
on
the
smaller
documents
today
from
the
larger
pile
of
abyss
document?
C
I
don't
think
there's
anything
else
and
I
brian
plowsky,
I'm
not
asking
for
new
work
right
now,
except
when
kristoff
proposes
new
work.
Kristoff.
G
Not
proposing
new
work,
but
I
think
one
of
the
big
items
our
ad
gave
us
was
errata
handling,
and
I
mean
I
think
one
of
the
problems
is.
We
have
a
lot
of
old,
big
documents
that
aren't
particularly
relevant
that
have
a
lot
of
errata
and
a
lot
of
people
that
worked
on
them
are
not
actively
part
of
the
working
group.
C
B
G
C
Okay,
can
I
so
christoph
I'm
gonna,
I'm
gonna
summarize
that
and
tom's
gonna
put
a
note.
Take
a
note
tom
that
we
just
gotta
scrub
the
errata
list
and
and
just
really
scrub
it
so
that
people
aren't
getting
nervous
and
then
determine
to
me
determine
two
things:
one
which
errada
we
don't
care
about
anymore.
C
You
know
what
nobody's
using
it
we're
not
going
to
update
documents
that
no
one
even
cares
about,
and
let's
mark
and
just
I
assume,
there's
a
way
to
close
those
errata
out
and
then
my
thinking
was
that
the
this
document
that
dave
owns
has
been
primarily
working
on
the
existing
errata
for
what
people
are
interested
in,
using
which
a
lot
of
what
I
thought
we
discussed
in
the
past
by
the
way
was
just
clarifications
of
the
specification
describing
what
we
actually
meant
when
we
said
something
I
thought
there
was
a
big
pile
of
those
and
maybe
dave's
gotten
through
muslim,
but
I
I
let's
let
me
again
follow
up
with
dave
he's
not
here.
C
I
This
richard
again
just
one
one
minor
point
that
you
that
I've
not
hear
you
say,
and
that
is
adopting
of
this
document
as
a
work
as
a
full
featured
working
group
document,
so
that
the
worker
group
has
a
formal
change
control
over
the
document
and
not
the
author
alone.
I
Take
the
bis
as
an
working
group
document.
E
F
C
F
F
What
do
you
just
say
and-
and
I
see
one
volunteer
to
actually
look-
go
through
that
one
that
would
be
great,
then
we
know
exactly
what
you're
doing
what
what
have
already
been
taken
care
of
and
what
are
the
different
things
and-
and
I
have
a
clear
idea
like
this-
is
like
something
that
that
I
can
reflect
to
when
I'm
looking
through
to
the
documents
and
all
this
thing.
That
would
be
a
great
thing
to
do.
F
F
C
I
got
I
got
it,
I
think
chuck's
going
to
say
something.
That's
going
to
remind
us
of
something
that
dave's
working
on.
H
One
of
the
reasons
why
this
is
still
a
personal
draft
and
not
a
working
group
document
is
because
part
of
this
part
of
the
mission
of
doing
this
business
to
recognize
that
this
is
an
enormous,
ass
piece
of
work
that
needs
to
be
split
so
dave
had
proposed.
I
think
splitting
it
four
ways
into
four
different
documents,
so
I'm
not
sure
about
process
here,
but
I
don't
know
if
we
want
to
adopt
that.
C
Right
and
I
yeah
and
I'm
sorry,
that's
and
that
was
discussed
a
very
long
time
ago.
Yes,
it
was
and
agreed
to
by
the
working
group,
primarily
because,
frankly,
anything
that's
done
outside
of
this
working
group
in
terms
of
editor
reviews
and
everything
else
that
the
size
of
the
document
is
daunting.
C
It
was
felt
that
if
it
was
broken
apart
with
no
changes
to
the
protocol
itself
and
that
that
just
the
act
of
breaking
the
document
into
four
pieces
focusing
on
specific
areas,
which
I
correct
me
from
wrong
truck
but
like
one
of
them,
one
of
the
four
documents
would
be
specifically
a
security
document
which
allows
outside
review
of
the
security
portions
of
the
protocol,
which
aren't
necessarily
in
one
location
on
the
document
today.
So
it
was
going
to
be
like
pull,
pull,
pull
focus.
C
H
H
And
like
we
don't
see
on
this
window
in
front
of
us
the
documents
that
have
expired
since
january,
I
think
at
least
three
of
them.
Security
is
not
up
here
because
it's
expired.
H
So
it's
hard
for
us
to
track
the
versions
of
the
documents
that
were
the
progeny
of
5661
bis.
I
don't
think
dave
ever
intended
5661
bis
to
be
adopted
as
a
working
group
document.
He
just
meant
it
as
the
base
for
doing
the
split
okay,
so
just
to
be
clear,
I'm
not
objecting
to
to
this
work
becoming
working
group
working
group
documents,
but
I
think
that
there's
some
history
here
that
we
need.
C
To
identify
you're
being
very
specific
about
what
we
choose
to
adopt
exactly:
okay,
okay,
so
again
without
dave.
Here
I
don't
wanna.
I
don't
wanna,
say
things
that
I
can't
speak
to,
but
I
have
the
list
of
questions
and
the
concern
and
the
asks
around.
C
And
priorities
for
the
group,
I
think
general
discussion
here
in
terms
of
documents.
Okay,
but
I
I
need
a
lot
more
information
from
dave
and
I
needed
yeah.
A
And
the
other
day,
yeah
david
black,
yes
and
yeah
dave's
going
to
have
a
cow
when
he
understands
what
we
did
in
his
absence
I
mean
and
there's
there's
even
bigger
question
on
the
table,
which
is.
C
You
know
talking
about
cows.
This
is
brian
pulaski
chair
the
version
four.
C
C
A
C
The
outcome,
so
I
will
say
sorry
david
s,
very
good
question-
is
that
whole
idea
worth
the
outcome.
I
would
say
when
we
discuss
the
split-
and
I
cannot
remember
how
long
ago,
but
this
was
not
in
the
past
three
months.
This
was
in
the
past
three
years.
We
discussed
this
split.
I
don't
think
anybody
thought
we'd
be
having
this
discussion
today.
C
I
thought
I
think
we
thought
that
the
split
would
be
completed
by
now,
and
things
would
be
going
along
smoothly
and
also
would
have
allowed
other
people
within
the
working
group
to
take
the
areas
they're
interested
in
to
review
right,
because
we
wouldn't
be
having
a
discussion
of
you
know
a
bunch
of
people
trying
to
find
the
area
they
need
to
worry
about.
So
I
do
are
you?
Did
you
scan
in.
C
In
the
name
they
might
not
disband
us,
though.
Thank
you,
okay.
Is
there
any
okay?
I
I
think
I
have
tom's
been
taking
notes.
I
hope
I
think
david
black
rewrote
all
of
my
notes.
I
think
that
was
him
typing
and
which
is
kind
of
funny,
but
I
think
I
have
a
lot
of
follow-up
week
taking
some
actions
when
the
minutes
get
published,
we'll
communicate
the
actions
we
took
on
the
documents.
C
Today
I
will
talk
to
priorities
on
the
working
group
list
and
and
I'm
buying
into
david's
model
of
let's
consider
two
piles,
one
of
the
smaller
docks
that
people
have
gravitated
towards
to
drive
to
completion
that
we
discussed
today
and
then
the
best
document
to
step
back
for
a
second
and
find
out
and
have
that
discussion
on.
Where
are
we
what's
its
purpose
and
since
nobody's
yeah?
Let's
just
leave
it
at
that?
Okay,
but
I
again
I
do
need
dave
novak
in
that
discussion.
Okay
and
then
at
the
there.
F
Yeah
so
so
yeah.
I
think
I
think
this
has
been
very
fruitful
discussion
so
far.
We
have
at
least
an
idea
like
what
we
are
supposed
to
do.
We
have
some
action
points
I
would
love
to
have
you
or,
or
the
secretary
actually
write,
write
an
email
to
the
mailing
list
for
the
record
like
this
is
what.
F
F
My
personal
opinion
is
like
we
we
talk
about
whenever
we
have
a
business
document
is
obsoleting,
that's
a
different
case,
but
if
it's
updating
and
stuff
like
that,
the
isg
usually
asks
for,
like
whatever
be
clear,
what
you
are
updating
and
all
these
things.
So
the
structure
of
document
really
matters
for
that
case.
But
that's
a
that's
the
later
case.
F
But
let's
discuss
on
this
when
dave
is
back
okay
about
the
split
and
then
do
the
right
thing
like
split
or
not
split,
I
don't
care
just
add
up
the
documents
and
carry
on
the
work
and
the
background,
and
please
keep
the
data
updated.
Yeah.
C
F
C
Okay,
thank
you.
Is
there
anything
any
tom
talpy.
J
C
C
J
E
C
I'm
going
to
say
going
well,
I'm
going
to
say
two
things.
This
is
the
first
time
we've
gotten
together
since
before
covet.
Is
that
face-to-face
we've
gotten
ahead
of
interims
and
everything
virtual
from
thank
you
tom,
so
much
for
making
this
a
lot
smoother
than
I
would
have
been
with
only
me
here
and
no
it's
it's
been
a
lot
more.
C
It's
been
a
much
higher
bandwidth
discussion
than
I
think
we
we've
had,
and
I
mean-
and
I
appreciate
the
guys
virtually,
but
managing
the
managing
the
crowd
with
some
of
the
crowd
here
and
some
of
the
people
online
is
to
me
a
lot
easier
from
as
a
from
a
chair
position.
C
So
this
has
been
a
great
thank
you
for
the
people
that
joined
virtually.
Thank
you
to
people
who
showed
up
physically
and
for
getting
all
the
drafts
in
before
the
meeting
started.
That
was
super
helpful.
So
this
has
been
great
and
I
just
want
to
say
going
once
going
twice.
C
E
C
C
C
E
C
What
are
you
leaving
tomorrow,
we're
gonna
drink
tonight?
Right
apparently,
can
you.