Re: Issue 05 was (lease expiration, stable storage (was "questions/comments on draft-ietf-nfsv4-rfc3010bis-02-04.txt")

New Message Reply About this list Date view Thread view Subject view Author view Attachment view

From: mike.kupfer@sun.com
Date: 08/20/02-04:09:08 PM Z


Message-Id: <200208202109.g7KL99XE197848@athyra.eng.sun.com>
From: mike.kupfer@sun.com
Subject: Re: Issue 05 was (lease expiration, stable storage (was "questions/comments on draft-ietf-nfsv4-rfc3010bis-02-04.txt") 
Date: Tue, 20 Aug 2002 14:09:08 -0700

After digesting the flurry of mail on this topic, here are my
thoughts:

- I think that using the NLM/NSM protocols as a standard is setting
  the bar a bit low.  On the other hand, I guess they provide us with
  enough operational experience to conclude that incorrect lock
  reclaims are not a big problem in practice.

- I don't understand why the problem should be considered more severe
  when the server implements mandatory locking.  It's not like the
  application gets to choose which model (mandatory or advisory) the
  server is using, or even detect reliably which model the server is
  using.

- Spencer wrote

  > I don't want to mislead the client implementor into thinking that
  > receiving a STALE_STATEID within the lease period is death to all
  > state.  It may be that the client's state for that single file has
  > been revoked and nothing else.

  Is this a statement that the client needs to handle STALE_STATEID
  differently, depending on whether it thinks it still has an
  unexpired lease?  That would be a pretty major departure from how
  STALE_STATEID is currently dealt with in the spec.

- I didn't see consensus on either of the following two issues:

  - administrative commands and their impact on client recovery

  - what the spec should say about stable storage and lock recovery

  I doubt that the Last Call period is long enough to make the
  necessary changes to provide single-file recovery (to support
  administrative commands).  I think this is an issue that should be
  deferred.

  I guess the stable storage issue is small enough (given the history
  we have with NLM/NSM) that it can be deferred.  On the other hand,
  given that we know about the issue and that the consequences are
  pretty severe (file corruption), I don't see how v4 can get past
  Proposed Standard until this is addressed.  I'm willing to let
  others with a better understanding of IETF policies and politics
  make the call on what to do here.

mike


New Message Reply About this list Date view Thread view Subject view Author view Attachment view

This archive was generated by hypermail 2.1.2 : 03/04/05-01:50:15 AM Z CST