NFS version 4 - Vancouver, BC (IETF 84)
Monday, July 25, 2011
0900 - 1130 : Room 2103
--------------------------------

=== Intro / Blue Sheets

	Noted Well etc.

=== Agenda bashing

No additions.

=== Charter Updates, Documents in Shepherding, IESG actions

New charter posted. 4.2 work moving.

4.0 BIS document is rather large with many areas that haven't changed.
We are to deliver a document that navigates the chang. Want a diff
presentation, similar to STORM chair...

Other in process work is the FedFS document under review by AD Martin,
presenting today.

=== NFSv4.2 (Haynes) (20 minutes)

We've gone through entire document in weekly review session.

	Design group meeting this afternoon. Meeting room 2-5 for
	line-by-line review.

Major changes are editorial. Protocol issues... server developers proposed
most features, open items are typically from things client developers
reviewed later.

Open issues:

	Server side copy - editorial change. Not defined as push or pull.
	More about authentication.

Really see slides. Space reservations. Issue with Linux fallocate().

Holes. INITIALIZE, or WRITE_PLUS or SETCONTENT. See slides for both
protocol chnages and editorial. David Black owes some information update
on server side copy URLs - clean up language.

Shooting for Last Call in two months.

=== FedFS    (Haynes and Lever) (30 minutes)
   General (Haynes)
    NSDB Open Issues (Lever)

641 is out, do not had to worry about it. Document is in AD reveew, some
issues have come back. External review of the document exposed security
concerns. Martin has requested both docs move together simultaneously.
(main items on both docs are security concerns).

	Why is this taking so long - Shepler? Allocating time - backlog
	of issues.

FS_LOCATIONS cannot pass major and minor numbers, so we plan to remove
fields. October concerns on UUID variants: time variant... Is there a 
security problem here - David Black yes, we need to look at sooner. We
have to carefully document the problem in security considerations.

	Black is to review that security section.

See slides on all this.

Howard Chu sent issues on use of LDAP. See slide.

	LDAP server administration differences mentioned.

Storing URLs for file servers and services? Can specify inward facing
approaches. Things not specified in the NSDB service name.

Spencer asked Lever how best to move forward. We should've caught the
issues earlier in our review process, since we're dealing with LDAP here
and not NFS - we need some LDAP expertise to clean up. LDAP experts
harder to find - Martin to try and help us.

=== NFSv4 Migration Issues (Noveck) (20 minutes)
      draft-ietf-nfsv4-migration-issues-01

See slides. Basic - multiserver features are harder than they look.

List of multiserver interactions (see slides). Devil in details as
in state reclaim in migration.

	Use different client strings based on server defeats
	client state merge on migration - example by Dave.

See Dave's latest document. But there is the question of what to put into
the normative documents. Backed by implementation experience.

=== Thoughts on construction of nfs_client_id4 id (Noveck/Lever) (10 minutes)

Chuck on the implementation experience around the draft migration language.

	See slides on details for distinguishing reboot from
	update.

Long description of the details of the client_id definition and uniqueness
problem.

Chuck suggests recommendations for virtualized environment for generating
client_id information.

What are next steps - Shepler? Most of this captured in the migration draft,
though not as explicit.

=== End-to-end Data Integrity (Lever) (20 minutes)

Personal draft on end-to-end data integrity. Gave talk in Paris on this.
Application generates the integrity information making it available in
every step of the I/O path.

Block storage has T10 PI. Been around for a bit. Eight bytes. Distinguishes
data, metadata content. Open standard. David Black - magic bit to not store
third field... to make it plausibly implementable...

	ATO mode in SCSI... makes it implementable...

We shoud be able to use the T10 PI ideas for inspiration and background
for our work here according to David Black response to Brian Pawlowski..

Some white paper from a vendor on the T10 PI:

    http://www.seagate.com/files/staticfiles/docs/pdf/whitepaper/tp621-safeguarding-data-from-corruption-us.pdf

Even if not at application level generation through POSIX (extensions),
Database and Hypervisor not POSIX restricted, can use.

There is much more information in the slides!

=== FedFS Schema Updates for SMB (Thurlow) (20 minutes)

See slides. 

=== Ideas on Layout usage (Matt Benjamin / Peter Honeyman) (20 minutes)

Peter was explaining the use of different experimental layouts under
pNFS. (Ceph).

=== RPCSEC_GSSv3 (Haynes / All)

Not covered due to time.

=== Wrapup (Shepler)