From: Peter Staubach (staubach@nyday.eng.sun.com)
Date: 07/09/98-08:09:02 PM Z
Date: Thu, 9 Jul 1998 18:09:02 -0700 From: staubach@nyday.eng.sun.com (Peter Staubach) Message-Id: <199807100109.SAA11794@nyday.eng.sun.com> Subject: Re: NFSv4 draft (new version) > > > > > > > > (2) It also needs a way of supporting ATOMIC APPENDS - atomically append > > > to the end of a file > > > > Atomic appends would be great. Please suggest a solution that would > > correctly handle idempotency and transmission. > > Would something similar to the Setattr case work? > Not unless you can come up with some new spin on it. We tried very hard when we constructed the NFS Version 3 protocol to include append mode writes, but we couldn't find a way of handling retransmissions. You really have to know whether the data got written once or not at all or you end up with multiple copies of the data at different offsets in the file. For some applications, this is okay, but for others, it is intolerable. ps
This archive was generated by hypermail 2.1.2 : 03/04/05-01:45:53 AM Z CST