Network File System Version 4 (nfsv4)
-------------------------------------

Chair(s):

    Spencer Shepler <Spencer.Shepler@sun.com>
    Brian Pawlowski <beepy@netapp.com>

AGENDA:  Tuesday, August 2, 2005: 14:00pm - 16:00 (2 hour)
Subject to change

    Welcome and Introduction            (beepy)            2 min
    Agenda bash
        - Blue Sheets
        - NOTE WELL

Agenda bashed, blue sheeted and noted well.  All your IP is belong
to us.

    June Bakeathon results              (Shepler)         10 min
    NFS V4 interoperability testing

No issues found, two new participants in interoperability testing.

Sun is exploring ACL model in context of new local file system.

Next IETF meeting is early in November.  Some discussion on best
practices for setting up Active Directory for security thingies.

Action Item: Next Bakeathon Active Directory and KDC testing
and issue identification.

Honeyman suggested we're getting to point where maybe we need
only 2 Bakeathons a year. Shepler to take to mail list.

Looking for a sponsor for next Bakeathon at CITI/UMich after
Austin.


    pNFS                                                  20 min
        http://www.ietf.org/internet-drafts/draft-black-pnfs-block-01.txt
        http://www.ietf.org/internet-drafts/draft-zelenka-pnfs-obj-01.txt
        http://www.ietf.org/internet-drafts/draft-welch-pnfs-ops-03.txt
        - June discussion summary
        - drafts review
        - adding to the NFSv4 charter.

Garth led discussion.  Better support of cluster file systems?
Noveck responded on mail list - maybe the issues shoul be factored
out of the pNFS draft.

Other issue is possible race conditions between LAYOUT_GET and other
calls.  Because of two connections.  Likely a STATE_ID or SEQUENCE_ID?
David Black said Highroad experience suggested that sequencing needed?

Byte range delegations - should be done in one way.  Whole
idea came out of pNFS but useful outside of it. Delegation
needed to safely cache layouts - coupled to byte range delegation.
Pull out of pNFS and factor into NFS V4.1.

Dragon in the work (David Black) caching only valid over the byte
range delegation.

Linux EXT3 may have issues with change time... as a way to implement
consistency. The V4 protocol requires a real change attribute. Bruce
Fields said there is a client checking change time in a single compoound
after mod (?) and it is failing to detect change? Spencer observes
we shouldn't try to fix broken file systems in protocol. (Aside
from Brent Welch - is there a client flush operation? - yes, session
bring down signals to server to reap all state held for that client).

David Black said blocks draft is hakey jakey.

Brent Welch thinks byte range delegations are good - may need stronger
cache consistency metric rather than change time. Some back and forth
regarding whether NFS cache consistency (strong) is sufficient for
apps - thought is keep it simple and let apps do what apps do anyway.
Can you give apps serialization with advisory locks - yes. 
        
Brent Welch summarized.  Main ops draft looks good. Metadata issue
with clustering may need address.  There are a handful of issues
in main ops draft.  State ID on layouts...
        
Three docs need to be smoothed out amongst themselves. Garth Goodson
observed we may have to add striping and mirroring concessions to
main doc.

Need to take this discussion to alias.

Mark Eshel raised striping and parallel READDIRPLUS. Caused people
distress.

Need to convert pNFS drafts to working group drafts. 


    Work to combine SECINFO, Sessions,  (Shepler)         20 min
    and directory delegations into
    the first draft of NFSv4 minor version 1. 
        http://www.ietf.org/internet-drafts/draft-ietf-nfsv4-sess-02.txt
        - Next steps
        - Plan for standards track entry
        - Owners 

Shepler made weak and feebles.  His new commitment is September 6.
Collapsing individual drafts.  Cleaning up. 

Sessions draft etc. have lots of background that needs pruning.
Shepler will generate an XDR file for .1 extensions. Bruce Fields -
why do all things have to go into one minor version?  Maybe overhead
of discussion should be amortized...

Interoperability testing to be done in February likely. Shepler 
wants to have two independent implementations before WG last call.

Is there a chance for pNFS to get into 4.1... Or should it be 4.2?
We could do in parallel...

We need directory delegations implementation before February.

        Concern in room about lack of interest?  I'm
        confused.

September 15... for WG ready and vetted pNFS - we will discuss at that
time whether 4.1 or not.


    NFS DNS Considerations for identities and ACLS        10 min
        http://www.ietf.org/internet-drafts/draft-mesta-nfs4id-dns-rr-01.txt
        http://www.ietf.org/internet-drafts/draft-mesta-nfsv4-domain-01.txt
        - Discussion

DNS config proposal. Brent Welch - isn't LDAP better? User and Group
ID mapping - no, just domain I am playing in. Which domain.

UID/GID name space at sun.com - regardless of sub-domains...
Want a single record to present the "realm" binding for identity.


    NFS Referrals                                         10 min
        http://www.ietf.org/internet-drafts/draft-ietf-nfsv4-referrals-00.txt

Carl Burnett and Dave Noveck to move forward as informational RFC
explaining how to use "referrals" for migration/replication.
Shepler will tell them to move forward.

(Nico mentioned minor version negotiation? Shepler to see if
he can get consensus.)


    RFC3530 I18N issues                 (Williams)         5 min

There are none?  Nico was going to try and convince people of a
string profile...  But he is giving up?  There is no desire
in related communities.


    SPKM alternative for NFSv4          (Williams)         5 min
        
It will be a lot easier to specify a DTLS (?) mechanism for
SPKM.

Need a meeting to iron out with relevant parties on moving forward
SPKM.  beepy observed we might need a respin if we don't simply
drop.  Andy Adamson and Peter Honeyman are sympathetic to DTLS.
        
Lots of discussion to be had...
        

    Status of NFS/RDMA                  (Talpey)           5 min
        - Next steps
        - Plan for standards track entry

Problem Statement is done deal.  Move forward.  NFS RDMA is a
RPC tweak document.  Direct Data Placement drafts.

Mallikarjun's comments from March need merging... Talpey to
be spanked.

Incosistencies in text and protocol description - of sessions.
Talpey to fix. By September 6. 

Several implementations.  Problem Statement to go to last call.
WG call for final comments on the September 6 drafts. 


    WG Goals/Milestones                (Shepler)          15 min

Spencer Shepler...

API advancement draft removed from work queue.  Migration/replication
is in charter - IBM is whacking on something - need to reconcile with
CITI stuff.  beepy to try and wrest something out of these guys.

MIB work. What to do... Adamson wants to review the MIB draft...
he will give to student (expired draft) to revise. Shepler to
send expired draft.

The testing document to advance draft.  For next summer consideration
because we want some more maturity in the interoperability.

Move forward DNS record stuff, etc.  See slides.  WG last call for
minor version in late spring proposed. Hufferd asked about testing
of RDDP...

pNFS to be moved forward with approval of area directors - beepy
to drive...  Garth Goodson to own driving con calls... September 15.

Security considerations for DNS record... Nico raised - will have
to be resolved... Need consideration - Spencer Shepler owns.

Kitten chair... asked about channel bindings document... asking
about channel bindings work... Shepler responded we're at point
of doing implementation work - need to follow up with Mike Eisler
on state of doc...

    Wrapup                              (beepy)            1 m

Well done everybody...

(Small asides on color blindedness and blue sheets)...

_______________________________________________
nfsv4 mailing list
nfsv4@ietf.org
https://www1.ietf.org/mailman/listinfo/nfsv4