From: Noveck, Dave (Dave.Noveck@netapp.com)
Date: 02/14/01-10:42:28 AM Z
Message-ID: <8C610D86AF6CD4119C9800B0D0499E331A6EB7@red.nane.netapp.com> From: "Noveck, Dave" <Dave.Noveck@netapp.com> Subject: V4 at Connectathon -- What protocol *exactly* Date: Wed, 14 Feb 2001 08:42:28 -0800 Obvious answer is the one in RFC3010, but that leaves some open issues. One issue has already been discussed. Everybody is going with per-(lockowner+file) stateid's. Another issue is the updating of seqid on errors. At the last bakeoff the rule was "errors don't count". Neil Brown's position that that won't work has general acceptance, but what are we going to do for Connectathon? Before you answer, remember that one (lovely) simplification from the previous approach was that it limits the amount of information you have to save for retransmission. In particular, saving the other-locker denial information involves a lockowner which is an unbounded-length string, so you have to either copy that and allocate space or reference count the lockowner structure to prevent it from going away (doing all the reference-counting correctly much preferred). We need to agree on this or we will have interoperability problems. Are people willing to sign up to have this working in the new (and correct) way? One other note: Using RFC3010 means we will not add the type to the denial information for Connectathon, despite the fact that everybody agrees this is a good thing to do in the spec eventually. Comments?
This archive was generated by hypermail 2.1.2 : 03/04/05-01:48:36 AM Z CST